diff options
238 files changed, 6402 insertions, 4103 deletions
@@ -2365,8 +2365,6 @@ E: acme@redhat.com | |||
2365 | W: http://oops.ghostprotocols.net:81/blog/ | 2365 | W: http://oops.ghostprotocols.net:81/blog/ |
2366 | P: 1024D/9224DF01 D5DF E3BB E3C8 BCBB F8AD 841A B6AB 4681 9224 DF01 | 2366 | P: 1024D/9224DF01 D5DF E3BB E3C8 BCBB F8AD 841A B6AB 4681 9224 DF01 |
2367 | D: IPX, LLC, DCCP, cyc2x, wl3501_cs, net/ hacks | 2367 | D: IPX, LLC, DCCP, cyc2x, wl3501_cs, net/ hacks |
2368 | S: R. Brasílio Itiberê, 4270/1010 - Água Verde | ||
2369 | S: 80240-060 - Curitiba - Paraná | ||
2370 | S: Brazil | 2368 | S: Brazil |
2371 | 2369 | ||
2372 | N: Karsten Merker | 2370 | N: Karsten Merker |
diff --git a/Documentation/RCU/trace.txt b/Documentation/RCU/trace.txt index a851118775d..6a8c73f55b8 100644 --- a/Documentation/RCU/trace.txt +++ b/Documentation/RCU/trace.txt | |||
@@ -1,18 +1,22 @@ | |||
1 | CONFIG_RCU_TRACE debugfs Files and Formats | 1 | CONFIG_RCU_TRACE debugfs Files and Formats |
2 | 2 | ||
3 | 3 | ||
4 | The rcutree implementation of RCU provides debugfs trace output that | 4 | The rcutree and rcutiny implementations of RCU provide debugfs trace |
5 | summarizes counters and state. This information is useful for debugging | 5 | output that summarizes counters and state. This information is useful for |
6 | RCU itself, and can sometimes also help to debug abuses of RCU. | 6 | debugging RCU itself, and can sometimes also help to debug abuses of RCU. |
7 | The following sections describe the debugfs files and formats. | 7 | The following sections describe the debugfs files and formats, first |
8 | for rcutree and next for rcutiny. | ||
8 | 9 | ||
9 | 10 | ||
10 | Hierarchical RCU debugfs Files and Formats | 11 | CONFIG_TREE_RCU and CONFIG_TREE_PREEMPT_RCU debugfs Files and Formats |
11 | 12 | ||
12 | This implementation of RCU provides three debugfs files under the | 13 | These implementations of RCU provides five debugfs files under the |
13 | top-level directory RCU: rcu/rcudata (which displays fields in struct | 14 | top-level directory RCU: rcu/rcudata (which displays fields in struct |
14 | rcu_data), rcu/rcugp (which displays grace-period counters), and | 15 | rcu_data), rcu/rcudata.csv (which is a .csv spreadsheet version of |
15 | rcu/rcuhier (which displays the struct rcu_node hierarchy). | 16 | rcu/rcudata), rcu/rcugp (which displays grace-period counters), |
17 | rcu/rcuhier (which displays the struct rcu_node hierarchy), and | ||
18 | rcu/rcu_pending (which displays counts of the reasons that the | ||
19 | rcu_pending() function decided that there was core RCU work to do). | ||
16 | 20 | ||
17 | The output of "cat rcu/rcudata" looks as follows: | 21 | The output of "cat rcu/rcudata" looks as follows: |
18 | 22 | ||
@@ -130,7 +134,8 @@ o "ci" is the number of RCU callbacks that have been invoked for | |||
130 | been registered in absence of CPU-hotplug activity. | 134 | been registered in absence of CPU-hotplug activity. |
131 | 135 | ||
132 | o "co" is the number of RCU callbacks that have been orphaned due to | 136 | o "co" is the number of RCU callbacks that have been orphaned due to |
133 | this CPU going offline. | 137 | this CPU going offline. These orphaned callbacks have been moved |
138 | to an arbitrarily chosen online CPU. | ||
134 | 139 | ||
135 | o "ca" is the number of RCU callbacks that have been adopted due to | 140 | o "ca" is the number of RCU callbacks that have been adopted due to |
136 | other CPUs going offline. Note that ci+co-ca+ql is the number of | 141 | other CPUs going offline. Note that ci+co-ca+ql is the number of |
@@ -168,12 +173,12 @@ o "gpnum" is the number of grace periods that have started. It is | |||
168 | 173 | ||
169 | The output of "cat rcu/rcuhier" looks as follows, with very long lines: | 174 | The output of "cat rcu/rcuhier" looks as follows, with very long lines: |
170 | 175 | ||
171 | c=6902 g=6903 s=2 jfq=3 j=72c7 nfqs=13142/nfqsng=0(13142) fqlh=6 oqlen=0 | 176 | c=6902 g=6903 s=2 jfq=3 j=72c7 nfqs=13142/nfqsng=0(13142) fqlh=6 |
172 | 1/1 .>. 0:127 ^0 | 177 | 1/1 .>. 0:127 ^0 |
173 | 3/3 .>. 0:35 ^0 0/0 .>. 36:71 ^1 0/0 .>. 72:107 ^2 0/0 .>. 108:127 ^3 | 178 | 3/3 .>. 0:35 ^0 0/0 .>. 36:71 ^1 0/0 .>. 72:107 ^2 0/0 .>. 108:127 ^3 |
174 | 3/3f .>. 0:5 ^0 2/3 .>. 6:11 ^1 0/0 .>. 12:17 ^2 0/0 .>. 18:23 ^3 0/0 .>. 24:29 ^4 0/0 .>. 30:35 ^5 0/0 .>. 36:41 ^0 0/0 .>. 42:47 ^1 0/0 .>. 48:53 ^2 0/0 .>. 54:59 ^3 0/0 .>. 60:65 ^4 0/0 .>. 66:71 ^5 0/0 .>. 72:77 ^0 0/0 .>. 78:83 ^1 0/0 .>. 84:89 ^2 0/0 .>. 90:95 ^3 0/0 .>. 96:101 ^4 0/0 .>. 102:107 ^5 0/0 .>. 108:113 ^0 0/0 .>. 114:119 ^1 0/0 .>. 120:125 ^2 0/0 .>. 126:127 ^3 | 179 | 3/3f .>. 0:5 ^0 2/3 .>. 6:11 ^1 0/0 .>. 12:17 ^2 0/0 .>. 18:23 ^3 0/0 .>. 24:29 ^4 0/0 .>. 30:35 ^5 0/0 .>. 36:41 ^0 0/0 .>. 42:47 ^1 0/0 .>. 48:53 ^2 0/0 .>. 54:59 ^3 0/0 .>. 60:65 ^4 0/0 .>. 66:71 ^5 0/0 .>. 72:77 ^0 0/0 .>. 78:83 ^1 0/0 .>. 84:89 ^2 0/0 .>. 90:95 ^3 0/0 .>. 96:101 ^4 0/0 .>. 102:107 ^5 0/0 .>. 108:113 ^0 0/0 .>. 114:119 ^1 0/0 .>. 120:125 ^2 0/0 .>. 126:127 ^3 |
175 | rcu_bh: | 180 | rcu_bh: |
176 | c=-226 g=-226 s=1 jfq=-5701 j=72c7 nfqs=88/nfqsng=0(88) fqlh=0 oqlen=0 | 181 | c=-226 g=-226 s=1 jfq=-5701 j=72c7 nfqs=88/nfqsng=0(88) fqlh=0 |
177 | 0/1 .>. 0:127 ^0 | 182 | 0/1 .>. 0:127 ^0 |
178 | 0/3 .>. 0:35 ^0 0/0 .>. 36:71 ^1 0/0 .>. 72:107 ^2 0/0 .>. 108:127 ^3 | 183 | 0/3 .>. 0:35 ^0 0/0 .>. 36:71 ^1 0/0 .>. 72:107 ^2 0/0 .>. 108:127 ^3 |
179 | 0/3f .>. 0:5 ^0 0/3 .>. 6:11 ^1 0/0 .>. 12:17 ^2 0/0 .>. 18:23 ^3 0/0 .>. 24:29 ^4 0/0 .>. 30:35 ^5 0/0 .>. 36:41 ^0 0/0 .>. 42:47 ^1 0/0 .>. 48:53 ^2 0/0 .>. 54:59 ^3 0/0 .>. 60:65 ^4 0/0 .>. 66:71 ^5 0/0 .>. 72:77 ^0 0/0 .>. 78:83 ^1 0/0 .>. 84:89 ^2 0/0 .>. 90:95 ^3 0/0 .>. 96:101 ^4 0/0 .>. 102:107 ^5 0/0 .>. 108:113 ^0 0/0 .>. 114:119 ^1 0/0 .>. 120:125 ^2 0/0 .>. 126:127 ^3 | 184 | 0/3f .>. 0:5 ^0 0/3 .>. 6:11 ^1 0/0 .>. 12:17 ^2 0/0 .>. 18:23 ^3 0/0 .>. 24:29 ^4 0/0 .>. 30:35 ^5 0/0 .>. 36:41 ^0 0/0 .>. 42:47 ^1 0/0 .>. 48:53 ^2 0/0 .>. 54:59 ^3 0/0 .>. 60:65 ^4 0/0 .>. 66:71 ^5 0/0 .>. 72:77 ^0 0/0 .>. 78:83 ^1 0/0 .>. 84:89 ^2 0/0 .>. 90:95 ^3 0/0 .>. 96:101 ^4 0/0 .>. 102:107 ^5 0/0 .>. 108:113 ^0 0/0 .>. 114:119 ^1 0/0 .>. 120:125 ^2 0/0 .>. 126:127 ^3 |
@@ -212,11 +217,6 @@ o "fqlh" is the number of calls to force_quiescent_state() that | |||
212 | exited immediately (without even being counted in nfqs above) | 217 | exited immediately (without even being counted in nfqs above) |
213 | due to contention on ->fqslock. | 218 | due to contention on ->fqslock. |
214 | 219 | ||
215 | o "oqlen" is the number of callbacks on the "orphan" callback | ||
216 | list. RCU callbacks are placed on this list by CPUs going | ||
217 | offline, and are "adopted" either by the CPU helping the outgoing | ||
218 | CPU or by the next rcu_barrier*() call, whichever comes first. | ||
219 | |||
220 | o Each element of the form "1/1 0:127 ^0" represents one struct | 220 | o Each element of the form "1/1 0:127 ^0" represents one struct |
221 | rcu_node. Each line represents one level of the hierarchy, from | 221 | rcu_node. Each line represents one level of the hierarchy, from |
222 | root to leaves. It is best to think of the rcu_data structures | 222 | root to leaves. It is best to think of the rcu_data structures |
@@ -326,3 +326,115 @@ o "nn" is the number of times that this CPU needed nothing. Alert | |||
326 | readers will note that the rcu "nn" number for a given CPU very | 326 | readers will note that the rcu "nn" number for a given CPU very |
327 | closely matches the rcu_bh "np" number for that same CPU. This | 327 | closely matches the rcu_bh "np" number for that same CPU. This |
328 | is due to short-circuit evaluation in rcu_pending(). | 328 | is due to short-circuit evaluation in rcu_pending(). |
329 | |||
330 | |||
331 | CONFIG_TINY_RCU and CONFIG_TINY_PREEMPT_RCU debugfs Files and Formats | ||
332 | |||
333 | These implementations of RCU provides a single debugfs file under the | ||
334 | top-level directory RCU, namely rcu/rcudata, which displays fields in | ||
335 | rcu_bh_ctrlblk, rcu_sched_ctrlblk and, for CONFIG_TINY_PREEMPT_RCU, | ||
336 | rcu_preempt_ctrlblk. | ||
337 | |||
338 | The output of "cat rcu/rcudata" is as follows: | ||
339 | |||
340 | rcu_preempt: qlen=24 gp=1097669 g197/p197/c197 tasks=... | ||
341 | ttb=. btg=no ntb=184 neb=0 nnb=183 j=01f7 bt=0274 | ||
342 | normal balk: nt=1097669 gt=0 bt=371 b=0 ny=25073378 nos=0 | ||
343 | exp balk: bt=0 nos=0 | ||
344 | rcu_sched: qlen: 0 | ||
345 | rcu_bh: qlen: 0 | ||
346 | |||
347 | This is split into rcu_preempt, rcu_sched, and rcu_bh sections, with the | ||
348 | rcu_preempt section appearing only in CONFIG_TINY_PREEMPT_RCU builds. | ||
349 | The last three lines of the rcu_preempt section appear only in | ||
350 | CONFIG_RCU_BOOST kernel builds. The fields are as follows: | ||
351 | |||
352 | o "qlen" is the number of RCU callbacks currently waiting either | ||
353 | for an RCU grace period or waiting to be invoked. This is the | ||
354 | only field present for rcu_sched and rcu_bh, due to the | ||
355 | short-circuiting of grace period in those two cases. | ||
356 | |||
357 | o "gp" is the number of grace periods that have completed. | ||
358 | |||
359 | o "g197/p197/c197" displays the grace-period state, with the | ||
360 | "g" number being the number of grace periods that have started | ||
361 | (mod 256), the "p" number being the number of grace periods | ||
362 | that the CPU has responded to (also mod 256), and the "c" | ||
363 | number being the number of grace periods that have completed | ||
364 | (once again mode 256). | ||
365 | |||
366 | Why have both "gp" and "g"? Because the data flowing into | ||
367 | "gp" is only present in a CONFIG_RCU_TRACE kernel. | ||
368 | |||
369 | o "tasks" is a set of bits. The first bit is "T" if there are | ||
370 | currently tasks that have recently blocked within an RCU | ||
371 | read-side critical section, the second bit is "N" if any of the | ||
372 | aforementioned tasks are blocking the current RCU grace period, | ||
373 | and the third bit is "E" if any of the aforementioned tasks are | ||
374 | blocking the current expedited grace period. Each bit is "." | ||
375 | if the corresponding condition does not hold. | ||
376 | |||
377 | o "ttb" is a single bit. It is "B" if any of the blocked tasks | ||
378 | need to be priority boosted and "." otherwise. | ||
379 | |||
380 | o "btg" indicates whether boosting has been carried out during | ||
381 | the current grace period, with "exp" indicating that boosting | ||
382 | is in progress for an expedited grace period, "no" indicating | ||
383 | that boosting has not yet started for a normal grace period, | ||
384 | "begun" indicating that boosting has bebug for a normal grace | ||
385 | period, and "done" indicating that boosting has completed for | ||
386 | a normal grace period. | ||
387 | |||
388 | o "ntb" is the total number of tasks subjected to RCU priority boosting | ||
389 | periods since boot. | ||
390 | |||
391 | o "neb" is the number of expedited grace periods that have had | ||
392 | to resort to RCU priority boosting since boot. | ||
393 | |||
394 | o "nnb" is the number of normal grace periods that have had | ||
395 | to resort to RCU priority boosting since boot. | ||
396 | |||
397 | o "j" is the low-order 12 bits of the jiffies counter in hexadecimal. | ||
398 | |||
399 | o "bt" is the low-order 12 bits of the value that the jiffies counter | ||
400 | will have at the next time that boosting is scheduled to begin. | ||
401 | |||
402 | o In the line beginning with "normal balk", the fields are as follows: | ||
403 | |||
404 | o "nt" is the number of times that the system balked from | ||
405 | boosting because there were no blocked tasks to boost. | ||
406 | Note that the system will balk from boosting even if the | ||
407 | grace period is overdue when the currently running task | ||
408 | is looping within an RCU read-side critical section. | ||
409 | There is no point in boosting in this case, because | ||
410 | boosting a running task won't make it run any faster. | ||
411 | |||
412 | o "gt" is the number of times that the system balked | ||
413 | from boosting because, although there were blocked tasks, | ||
414 | none of them were preventing the current grace period | ||
415 | from completing. | ||
416 | |||
417 | o "bt" is the number of times that the system balked | ||
418 | from boosting because boosting was already in progress. | ||
419 | |||
420 | o "b" is the number of times that the system balked from | ||
421 | boosting because boosting had already completed for | ||
422 | the grace period in question. | ||
423 | |||
424 | o "ny" is the number of times that the system balked from | ||
425 | boosting because it was not yet time to start boosting | ||
426 | the grace period in question. | ||
427 | |||
428 | o "nos" is the number of times that the system balked from | ||
429 | boosting for inexplicable ("not otherwise specified") | ||
430 | reasons. This can actually happen due to races involving | ||
431 | increments of the jiffies counter. | ||
432 | |||
433 | o In the line beginning with "exp balk", the fields are as follows: | ||
434 | |||
435 | o "bt" is the number of times that the system balked from | ||
436 | boosting because there were no blocked tasks to boost. | ||
437 | |||
438 | o "nos" is the number of times that the system balked from | ||
439 | boosting for inexplicable ("not otherwise specified") | ||
440 | reasons. | ||
diff --git a/Documentation/dontdiff b/Documentation/dontdiff index d9bcffd5943..470d3dba1a6 100644 --- a/Documentation/dontdiff +++ b/Documentation/dontdiff | |||
@@ -62,6 +62,10 @@ aic7*reg_print.c* | |||
62 | aic7*seq.h* | 62 | aic7*seq.h* |
63 | aicasm | 63 | aicasm |
64 | aicdb.h* | 64 | aicdb.h* |
65 | altivec1.c | ||
66 | altivec2.c | ||
67 | altivec4.c | ||
68 | altivec8.c | ||
65 | asm-offsets.h | 69 | asm-offsets.h |
66 | asm_offsets.h | 70 | asm_offsets.h |
67 | autoconf.h* | 71 | autoconf.h* |
@@ -76,6 +80,7 @@ btfixupprep | |||
76 | build | 80 | build |
77 | bvmlinux | 81 | bvmlinux |
78 | bzImage* | 82 | bzImage* |
83 | capflags.c | ||
79 | classlist.h* | 84 | classlist.h* |
80 | comp*.log | 85 | comp*.log |
81 | compile.h* | 86 | compile.h* |
@@ -94,6 +99,7 @@ devlist.h* | |||
94 | docproc | 99 | docproc |
95 | elf2ecoff | 100 | elf2ecoff |
96 | elfconfig.h* | 101 | elfconfig.h* |
102 | evergreen_reg_safe.h | ||
97 | fixdep | 103 | fixdep |
98 | flask.h | 104 | flask.h |
99 | fore200e_mkfirm | 105 | fore200e_mkfirm |
@@ -108,9 +114,16 @@ genksyms | |||
108 | *_gray256.c | 114 | *_gray256.c |
109 | ihex2fw | 115 | ihex2fw |
110 | ikconfig.h* | 116 | ikconfig.h* |
117 | inat-tables.c | ||
111 | initramfs_data.cpio | 118 | initramfs_data.cpio |
112 | initramfs_data.cpio.gz | 119 | initramfs_data.cpio.gz |
113 | initramfs_list | 120 | initramfs_list |
121 | int16.c | ||
122 | int1.c | ||
123 | int2.c | ||
124 | int32.c | ||
125 | int4.c | ||
126 | int8.c | ||
114 | kallsyms | 127 | kallsyms |
115 | kconfig | 128 | kconfig |
116 | keywords.c | 129 | keywords.c |
@@ -140,6 +153,7 @@ mkprep | |||
140 | mktables | 153 | mktables |
141 | mktree | 154 | mktree |
142 | modpost | 155 | modpost |
156 | modules.builtin | ||
143 | modules.order | 157 | modules.order |
144 | modversions.h* | 158 | modversions.h* |
145 | ncscope.* | 159 | ncscope.* |
@@ -153,14 +167,23 @@ pca200e.bin | |||
153 | pca200e_ecd.bin2 | 167 | pca200e_ecd.bin2 |
154 | piggy.gz | 168 | piggy.gz |
155 | piggyback | 169 | piggyback |
170 | piggy.S | ||
156 | pnmtologo | 171 | pnmtologo |
157 | ppc_defs.h* | 172 | ppc_defs.h* |
158 | pss_boot.h | 173 | pss_boot.h |
159 | qconf | 174 | qconf |
175 | r100_reg_safe.h | ||
176 | r200_reg_safe.h | ||
177 | r300_reg_safe.h | ||
178 | r420_reg_safe.h | ||
179 | r600_reg_safe.h | ||
160 | raid6altivec*.c | 180 | raid6altivec*.c |
161 | raid6int*.c | 181 | raid6int*.c |
162 | raid6tables.c | 182 | raid6tables.c |
163 | relocs | 183 | relocs |
184 | rn50_reg_safe.h | ||
185 | rs600_reg_safe.h | ||
186 | rv515_reg_safe.h | ||
164 | series | 187 | series |
165 | setup | 188 | setup |
166 | setup.bin | 189 | setup.bin |
@@ -169,6 +192,7 @@ sImage | |||
169 | sm_tbl* | 192 | sm_tbl* |
170 | split-include | 193 | split-include |
171 | syscalltab.h | 194 | syscalltab.h |
195 | tables.c | ||
172 | tags | 196 | tags |
173 | tftpboot.img | 197 | tftpboot.img |
174 | timeconst.h | 198 | timeconst.h |
@@ -190,6 +214,7 @@ vmlinux | |||
190 | vmlinux-* | 214 | vmlinux-* |
191 | vmlinux.aout | 215 | vmlinux.aout |
192 | vmlinux.lds | 216 | vmlinux.lds |
217 | voffset.h | ||
193 | vsyscall.lds | 218 | vsyscall.lds |
194 | vsyscall_32.lds | 219 | vsyscall_32.lds |
195 | wanxlfw.inc | 220 | wanxlfw.inc |
@@ -200,3 +225,4 @@ wakeup.elf | |||
200 | wakeup.lds | 225 | wakeup.lds |
201 | zImage* | 226 | zImage* |
202 | zconf.hash.c | 227 | zconf.hash.c |
228 | zoffset.h | ||
diff --git a/Documentation/kernel-docs.txt b/Documentation/kernel-docs.txt index 715eaaf1519..9a8674629a0 100644 --- a/Documentation/kernel-docs.txt +++ b/Documentation/kernel-docs.txt | |||
@@ -537,7 +537,7 @@ | |||
537 | Notes: Further information in | 537 | Notes: Further information in |
538 | http://www.oreilly.com/catalog/linuxdrive2/ | 538 | http://www.oreilly.com/catalog/linuxdrive2/ |
539 | 539 | ||
540 | * Title: "Linux Device Drivers, 3nd Edition" | 540 | * Title: "Linux Device Drivers, 3rd Edition" |
541 | Authors: Jonathan Corbet, Alessandro Rubini, and Greg Kroah-Hartman | 541 | Authors: Jonathan Corbet, Alessandro Rubini, and Greg Kroah-Hartman |
542 | Publisher: O'Reilly & Associates. | 542 | Publisher: O'Reilly & Associates. |
543 | Date: 2005. | 543 | Date: 2005. |
@@ -592,14 +592,6 @@ | |||
592 | Pages: 600. | 592 | Pages: 600. |
593 | ISBN: 0-13-101908-2 | 593 | ISBN: 0-13-101908-2 |
594 | 594 | ||
595 | * Title: "The Design and Implementation of the 4.4 BSD UNIX | ||
596 | Operating System" | ||
597 | Author: Marshall Kirk McKusick, Keith Bostic, Michael J. Karels, | ||
598 | John S. Quarterman. | ||
599 | Publisher: Addison-Wesley. | ||
600 | Date: 1996. | ||
601 | ISBN: 0-201-54979-4 | ||
602 | |||
603 | * Title: "Programming for the real world - POSIX.4" | 595 | * Title: "Programming for the real world - POSIX.4" |
604 | Author: Bill O. Gallmeister. | 596 | Author: Bill O. Gallmeister. |
605 | Publisher: O'Reilly & Associates, Inc.. | 597 | Publisher: O'Reilly & Associates, Inc.. |
@@ -610,28 +602,13 @@ | |||
610 | POSIX. Good reference. | 602 | POSIX. Good reference. |
611 | 603 | ||
612 | * Title: "UNIX Systems for Modern Architectures: Symmetric | 604 | * Title: "UNIX Systems for Modern Architectures: Symmetric |
613 | Multiprocesssing and Caching for Kernel Programmers" | 605 | Multiprocessing and Caching for Kernel Programmers" |
614 | Author: Curt Schimmel. | 606 | Author: Curt Schimmel. |
615 | Publisher: Addison Wesley. | 607 | Publisher: Addison Wesley. |
616 | Date: June, 1994. | 608 | Date: June, 1994. |
617 | Pages: 432. | 609 | Pages: 432. |
618 | ISBN: 0-201-63338-8 | 610 | ISBN: 0-201-63338-8 |
619 | 611 | ||
620 | * Title: "The Design and Implementation of the 4.3 BSD UNIX | ||
621 | Operating System" | ||
622 | Author: Samuel J. Leffler, Marshall Kirk McKusick, Michael J. | ||
623 | Karels, John S. Quarterman. | ||
624 | Publisher: Addison-Wesley. | ||
625 | Date: 1989 (reprinted with corrections on October, 1990). | ||
626 | ISBN: 0-201-06196-1 | ||
627 | |||
628 | * Title: "The Design of the UNIX Operating System" | ||
629 | Author: Maurice J. Bach. | ||
630 | Publisher: Prentice Hall. | ||
631 | Date: 1986. | ||
632 | Pages: 471. | ||
633 | ISBN: 0-13-201757-1 | ||
634 | |||
635 | MISCELLANEOUS: | 612 | MISCELLANEOUS: |
636 | 613 | ||
637 | * Name: linux/Documentation | 614 | * Name: linux/Documentation |
diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt index 1031923f525..d6496fde618 100644 --- a/Documentation/kernel-parameters.txt +++ b/Documentation/kernel-parameters.txt | |||
@@ -1579,20 +1579,12 @@ and is between 256 and 4096 characters. It is defined in the file | |||
1579 | 1579 | ||
1580 | nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels | 1580 | nmi_watchdog= [KNL,BUGS=X86] Debugging features for SMP kernels |
1581 | Format: [panic,][num] | 1581 | Format: [panic,][num] |
1582 | Valid num: 0,1,2 | 1582 | Valid num: 0 |
1583 | 0 - turn nmi_watchdog off | 1583 | 0 - turn nmi_watchdog off |
1584 | 1 - use the IO-APIC timer for the NMI watchdog | ||
1585 | 2 - use the local APIC for the NMI watchdog using | ||
1586 | a performance counter. Note: This will use one | ||
1587 | performance counter and the local APIC's performance | ||
1588 | vector. | ||
1589 | When panic is specified, panic when an NMI watchdog | 1584 | When panic is specified, panic when an NMI watchdog |
1590 | timeout occurs. | 1585 | timeout occurs. |
1591 | This is useful when you use a panic=... timeout and | 1586 | This is useful when you use a panic=... timeout and |
1592 | need the box quickly up again. | 1587 | need the box quickly up again. |
1593 | Instead of 1 and 2 it is possible to use the following | ||
1594 | symbolic names: lapic and ioapic | ||
1595 | Example: nmi_watchdog=2 or nmi_watchdog=panic,lapic | ||
1596 | 1588 | ||
1597 | netpoll.carrier_timeout= | 1589 | netpoll.carrier_timeout= |
1598 | [NET] Specifies amount of time (in seconds) that | 1590 | [NET] Specifies amount of time (in seconds) that |
diff --git a/Documentation/trace/events-power.txt b/Documentation/trace/events-power.txt new file mode 100644 index 00000000000..96d87b67fe3 --- /dev/null +++ b/Documentation/trace/events-power.txt | |||
@@ -0,0 +1,90 @@ | |||
1 | |||
2 | Subsystem Trace Points: power | ||
3 | |||
4 | The power tracing system captures events related to power transitions | ||
5 | within the kernel. Broadly speaking there are three major subheadings: | ||
6 | |||
7 | o Power state switch which reports events related to suspend (S-states), | ||
8 | cpuidle (C-states) and cpufreq (P-states) | ||
9 | o System clock related changes | ||
10 | o Power domains related changes and transitions | ||
11 | |||
12 | This document describes what each of the tracepoints is and why they | ||
13 | might be useful. | ||
14 | |||
15 | Cf. include/trace/events/power.h for the events definitions. | ||
16 | |||
17 | 1. Power state switch events | ||
18 | ============================ | ||
19 | |||
20 | 1.1 New trace API | ||
21 | ----------------- | ||
22 | |||
23 | A 'cpu' event class gathers the CPU-related events: cpuidle and | ||
24 | cpufreq. | ||
25 | |||
26 | cpu_idle "state=%lu cpu_id=%lu" | ||
27 | cpu_frequency "state=%lu cpu_id=%lu" | ||
28 | |||
29 | A suspend event is used to indicate the system going in and out of the | ||
30 | suspend mode: | ||
31 | |||
32 | machine_suspend "state=%lu" | ||
33 | |||
34 | |||
35 | Note: the value of '-1' or '4294967295' for state means an exit from the current state, | ||
36 | i.e. trace_cpu_idle(4, smp_processor_id()) means that the system | ||
37 | enters the idle state 4, while trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id()) | ||
38 | means that the system exits the previous idle state. | ||
39 | |||
40 | The event which has 'state=4294967295' in the trace is very important to the user | ||
41 | space tools which are using it to detect the end of the current state, and so to | ||
42 | correctly draw the states diagrams and to calculate accurate statistics etc. | ||
43 | |||
44 | 1.2 DEPRECATED trace API | ||
45 | ------------------------ | ||
46 | |||
47 | A new Kconfig option CONFIG_EVENT_POWER_TRACING_DEPRECATED with the default value of | ||
48 | 'y' has been created. This allows the legacy trace power API to be used conjointly | ||
49 | with the new trace API. | ||
50 | The Kconfig option, the old trace API (in include/trace/events/power.h) and the | ||
51 | old trace points will disappear in a future release (namely 2.6.41). | ||
52 | |||
53 | power_start "type=%lu state=%lu cpu_id=%lu" | ||
54 | power_frequency "type=%lu state=%lu cpu_id=%lu" | ||
55 | power_end "cpu_id=%lu" | ||
56 | |||
57 | The 'type' parameter takes one of those macros: | ||
58 | . POWER_NONE = 0, | ||
59 | . POWER_CSTATE = 1, /* C-State */ | ||
60 | . POWER_PSTATE = 2, /* Fequency change or DVFS */ | ||
61 | |||
62 | The 'state' parameter is set depending on the type: | ||
63 | . Target C-state for type=POWER_CSTATE, | ||
64 | . Target frequency for type=POWER_PSTATE, | ||
65 | |||
66 | power_end is used to indicate the exit of a state, corresponding to the latest | ||
67 | power_start event. | ||
68 | |||
69 | 2. Clocks events | ||
70 | ================ | ||
71 | The clock events are used for clock enable/disable and for | ||
72 | clock rate change. | ||
73 | |||
74 | clock_enable "%s state=%lu cpu_id=%lu" | ||
75 | clock_disable "%s state=%lu cpu_id=%lu" | ||
76 | clock_set_rate "%s state=%lu cpu_id=%lu" | ||
77 | |||
78 | The first parameter gives the clock name (e.g. "gpio1_iclk"). | ||
79 | The second parameter is '1' for enable, '0' for disable, the target | ||
80 | clock rate for set_rate. | ||
81 | |||
82 | 3. Power domains events | ||
83 | ======================= | ||
84 | The power domain events are used for power domains transitions | ||
85 | |||
86 | power_domain_target "%s state=%lu cpu_id=%lu" | ||
87 | |||
88 | The first parameter gives the power domain name (e.g. "mpu_pwrdm"). | ||
89 | The second parameter is the power domain target state. | ||
90 | |||
diff --git a/MAINTAINERS b/MAINTAINERS index 7585e9dd835..b1dda78a1e7 100644 --- a/MAINTAINERS +++ b/MAINTAINERS | |||
@@ -4627,7 +4627,7 @@ PERFORMANCE EVENTS SUBSYSTEM | |||
4627 | M: Peter Zijlstra <a.p.zijlstra@chello.nl> | 4627 | M: Peter Zijlstra <a.p.zijlstra@chello.nl> |
4628 | M: Paul Mackerras <paulus@samba.org> | 4628 | M: Paul Mackerras <paulus@samba.org> |
4629 | M: Ingo Molnar <mingo@elte.hu> | 4629 | M: Ingo Molnar <mingo@elte.hu> |
4630 | M: Arnaldo Carvalho de Melo <acme@redhat.com> | 4630 | M: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> |
4631 | S: Supported | 4631 | S: Supported |
4632 | F: kernel/perf_event*.c | 4632 | F: kernel/perf_event*.c |
4633 | F: include/linux/perf_event.h | 4633 | F: include/linux/perf_event.h |
diff --git a/arch/alpha/include/asm/perf_event.h b/arch/alpha/include/asm/perf_event.h index fe792ca818f..5996e7a6757 100644 --- a/arch/alpha/include/asm/perf_event.h +++ b/arch/alpha/include/asm/perf_event.h | |||
@@ -1,10 +1,4 @@ | |||
1 | #ifndef __ASM_ALPHA_PERF_EVENT_H | 1 | #ifndef __ASM_ALPHA_PERF_EVENT_H |
2 | #define __ASM_ALPHA_PERF_EVENT_H | 2 | #define __ASM_ALPHA_PERF_EVENT_H |
3 | 3 | ||
4 | #ifdef CONFIG_PERF_EVENTS | ||
5 | extern void init_hw_perf_events(void); | ||
6 | #else | ||
7 | static inline void init_hw_perf_events(void) { } | ||
8 | #endif | ||
9 | |||
10 | #endif /* __ASM_ALPHA_PERF_EVENT_H */ | 4 | #endif /* __ASM_ALPHA_PERF_EVENT_H */ |
diff --git a/arch/alpha/kernel/irq_alpha.c b/arch/alpha/kernel/irq_alpha.c index 5f77afb88e8..4c8bb374eb0 100644 --- a/arch/alpha/kernel/irq_alpha.c +++ b/arch/alpha/kernel/irq_alpha.c | |||
@@ -112,8 +112,6 @@ init_IRQ(void) | |||
112 | wrent(entInt, 0); | 112 | wrent(entInt, 0); |
113 | 113 | ||
114 | alpha_mv.init_irq(); | 114 | alpha_mv.init_irq(); |
115 | |||
116 | init_hw_perf_events(); | ||
117 | } | 115 | } |
118 | 116 | ||
119 | /* | 117 | /* |
diff --git a/arch/alpha/kernel/perf_event.c b/arch/alpha/kernel/perf_event.c index 1cc49683fb6..90561c45e7d 100644 --- a/arch/alpha/kernel/perf_event.c +++ b/arch/alpha/kernel/perf_event.c | |||
@@ -14,6 +14,7 @@ | |||
14 | #include <linux/kernel.h> | 14 | #include <linux/kernel.h> |
15 | #include <linux/kdebug.h> | 15 | #include <linux/kdebug.h> |
16 | #include <linux/mutex.h> | 16 | #include <linux/mutex.h> |
17 | #include <linux/init.h> | ||
17 | 18 | ||
18 | #include <asm/hwrpb.h> | 19 | #include <asm/hwrpb.h> |
19 | #include <asm/atomic.h> | 20 | #include <asm/atomic.h> |
@@ -863,13 +864,13 @@ static void alpha_perf_event_irq_handler(unsigned long la_ptr, | |||
863 | /* | 864 | /* |
864 | * Init call to initialise performance events at kernel startup. | 865 | * Init call to initialise performance events at kernel startup. |
865 | */ | 866 | */ |
866 | void __init init_hw_perf_events(void) | 867 | int __init init_hw_perf_events(void) |
867 | { | 868 | { |
868 | pr_info("Performance events: "); | 869 | pr_info("Performance events: "); |
869 | 870 | ||
870 | if (!supported_cpu()) { | 871 | if (!supported_cpu()) { |
871 | pr_cont("No support for your CPU.\n"); | 872 | pr_cont("No support for your CPU.\n"); |
872 | return; | 873 | return 0; |
873 | } | 874 | } |
874 | 875 | ||
875 | pr_cont("Supported CPU type!\n"); | 876 | pr_cont("Supported CPU type!\n"); |
@@ -881,6 +882,8 @@ void __init init_hw_perf_events(void) | |||
881 | /* And set up PMU specification */ | 882 | /* And set up PMU specification */ |
882 | alpha_pmu = &ev67_pmu; | 883 | alpha_pmu = &ev67_pmu; |
883 | 884 | ||
884 | perf_pmu_register(&pmu); | 885 | perf_pmu_register(&pmu, "cpu", PERF_TYPE_RAW); |
885 | } | ||
886 | 886 | ||
887 | return 0; | ||
888 | } | ||
889 | early_initcall(init_hw_perf_events); | ||
diff --git a/arch/arm/kernel/perf_event.c b/arch/arm/kernel/perf_event.c index 07a50357492..fdfa4976b0b 100644 --- a/arch/arm/kernel/perf_event.c +++ b/arch/arm/kernel/perf_event.c | |||
@@ -3034,11 +3034,11 @@ init_hw_perf_events(void) | |||
3034 | pr_info("no hardware support available\n"); | 3034 | pr_info("no hardware support available\n"); |
3035 | } | 3035 | } |
3036 | 3036 | ||
3037 | perf_pmu_register(&pmu); | 3037 | perf_pmu_register(&pmu, "cpu", PERF_TYPE_RAW); |
3038 | 3038 | ||
3039 | return 0; | 3039 | return 0; |
3040 | } | 3040 | } |
3041 | arch_initcall(init_hw_perf_events); | 3041 | early_initcall(init_hw_perf_events); |
3042 | 3042 | ||
3043 | /* | 3043 | /* |
3044 | * Callchain handling code. | 3044 | * Callchain handling code. |
diff --git a/arch/mips/kernel/perf_event_mipsxx.c b/arch/mips/kernel/perf_event_mipsxx.c index 5c7c6fc0756..183e0d22666 100644 --- a/arch/mips/kernel/perf_event_mipsxx.c +++ b/arch/mips/kernel/perf_event_mipsxx.c | |||
@@ -1047,6 +1047,6 @@ init_hw_perf_events(void) | |||
1047 | 1047 | ||
1048 | return 0; | 1048 | return 0; |
1049 | } | 1049 | } |
1050 | arch_initcall(init_hw_perf_events); | 1050 | early_initcall(init_hw_perf_events); |
1051 | 1051 | ||
1052 | #endif /* defined(CONFIG_CPU_MIPS32)... */ | 1052 | #endif /* defined(CONFIG_CPU_MIPS32)... */ |
diff --git a/arch/powerpc/kernel/e500-pmu.c b/arch/powerpc/kernel/e500-pmu.c index 7c07de0d894..b150b510510 100644 --- a/arch/powerpc/kernel/e500-pmu.c +++ b/arch/powerpc/kernel/e500-pmu.c | |||
@@ -126,4 +126,4 @@ static int init_e500_pmu(void) | |||
126 | return register_fsl_emb_pmu(&e500_pmu); | 126 | return register_fsl_emb_pmu(&e500_pmu); |
127 | } | 127 | } |
128 | 128 | ||
129 | arch_initcall(init_e500_pmu); | 129 | early_initcall(init_e500_pmu); |
diff --git a/arch/powerpc/kernel/mpc7450-pmu.c b/arch/powerpc/kernel/mpc7450-pmu.c index 09d72028f31..2cc5e0301d0 100644 --- a/arch/powerpc/kernel/mpc7450-pmu.c +++ b/arch/powerpc/kernel/mpc7450-pmu.c | |||
@@ -414,4 +414,4 @@ static int init_mpc7450_pmu(void) | |||
414 | return register_power_pmu(&mpc7450_pmu); | 414 | return register_power_pmu(&mpc7450_pmu); |
415 | } | 415 | } |
416 | 416 | ||
417 | arch_initcall(init_mpc7450_pmu); | 417 | early_initcall(init_mpc7450_pmu); |
diff --git a/arch/powerpc/kernel/perf_event.c b/arch/powerpc/kernel/perf_event.c index 3129c855933..56748070578 100644 --- a/arch/powerpc/kernel/perf_event.c +++ b/arch/powerpc/kernel/perf_event.c | |||
@@ -1379,7 +1379,7 @@ int register_power_pmu(struct power_pmu *pmu) | |||
1379 | freeze_events_kernel = MMCR0_FCHV; | 1379 | freeze_events_kernel = MMCR0_FCHV; |
1380 | #endif /* CONFIG_PPC64 */ | 1380 | #endif /* CONFIG_PPC64 */ |
1381 | 1381 | ||
1382 | perf_pmu_register(&power_pmu); | 1382 | perf_pmu_register(&power_pmu, "cpu", PERF_TYPE_RAW); |
1383 | perf_cpu_notifier(power_pmu_notifier); | 1383 | perf_cpu_notifier(power_pmu_notifier); |
1384 | 1384 | ||
1385 | return 0; | 1385 | return 0; |
diff --git a/arch/powerpc/kernel/perf_event_fsl_emb.c b/arch/powerpc/kernel/perf_event_fsl_emb.c index 7ecca59ddf7..4dcf5f831e9 100644 --- a/arch/powerpc/kernel/perf_event_fsl_emb.c +++ b/arch/powerpc/kernel/perf_event_fsl_emb.c | |||
@@ -681,7 +681,7 @@ int register_fsl_emb_pmu(struct fsl_emb_pmu *pmu) | |||
681 | pr_info("%s performance monitor hardware support registered\n", | 681 | pr_info("%s performance monitor hardware support registered\n", |
682 | pmu->name); | 682 | pmu->name); |
683 | 683 | ||
684 | perf_pmu_register(&fsl_emb_pmu); | 684 | perf_pmu_register(&fsl_emb_pmu, "cpu", PERF_TYPE_RAW); |
685 | 685 | ||
686 | return 0; | 686 | return 0; |
687 | } | 687 | } |
diff --git a/arch/powerpc/kernel/power4-pmu.c b/arch/powerpc/kernel/power4-pmu.c index 2a361cdda63..ead8b3c2649 100644 --- a/arch/powerpc/kernel/power4-pmu.c +++ b/arch/powerpc/kernel/power4-pmu.c | |||
@@ -613,4 +613,4 @@ static int init_power4_pmu(void) | |||
613 | return register_power_pmu(&power4_pmu); | 613 | return register_power_pmu(&power4_pmu); |
614 | } | 614 | } |
615 | 615 | ||
616 | arch_initcall(init_power4_pmu); | 616 | early_initcall(init_power4_pmu); |
diff --git a/arch/powerpc/kernel/power5+-pmu.c b/arch/powerpc/kernel/power5+-pmu.c index 199de527d41..eca0ac595cb 100644 --- a/arch/powerpc/kernel/power5+-pmu.c +++ b/arch/powerpc/kernel/power5+-pmu.c | |||
@@ -682,4 +682,4 @@ static int init_power5p_pmu(void) | |||
682 | return register_power_pmu(&power5p_pmu); | 682 | return register_power_pmu(&power5p_pmu); |
683 | } | 683 | } |
684 | 684 | ||
685 | arch_initcall(init_power5p_pmu); | 685 | early_initcall(init_power5p_pmu); |
diff --git a/arch/powerpc/kernel/power5-pmu.c b/arch/powerpc/kernel/power5-pmu.c index 98b6a729a9d..d5ff0f64a5e 100644 --- a/arch/powerpc/kernel/power5-pmu.c +++ b/arch/powerpc/kernel/power5-pmu.c | |||
@@ -621,4 +621,4 @@ static int init_power5_pmu(void) | |||
621 | return register_power_pmu(&power5_pmu); | 621 | return register_power_pmu(&power5_pmu); |
622 | } | 622 | } |
623 | 623 | ||
624 | arch_initcall(init_power5_pmu); | 624 | early_initcall(init_power5_pmu); |
diff --git a/arch/powerpc/kernel/power6-pmu.c b/arch/powerpc/kernel/power6-pmu.c index 84a607bda8f..31603927e37 100644 --- a/arch/powerpc/kernel/power6-pmu.c +++ b/arch/powerpc/kernel/power6-pmu.c | |||
@@ -544,4 +544,4 @@ static int init_power6_pmu(void) | |||
544 | return register_power_pmu(&power6_pmu); | 544 | return register_power_pmu(&power6_pmu); |
545 | } | 545 | } |
546 | 546 | ||
547 | arch_initcall(init_power6_pmu); | 547 | early_initcall(init_power6_pmu); |
diff --git a/arch/powerpc/kernel/power7-pmu.c b/arch/powerpc/kernel/power7-pmu.c index 852f7b7f6b4..593740fcb79 100644 --- a/arch/powerpc/kernel/power7-pmu.c +++ b/arch/powerpc/kernel/power7-pmu.c | |||
@@ -369,4 +369,4 @@ static int init_power7_pmu(void) | |||
369 | return register_power_pmu(&power7_pmu); | 369 | return register_power_pmu(&power7_pmu); |
370 | } | 370 | } |
371 | 371 | ||
372 | arch_initcall(init_power7_pmu); | 372 | early_initcall(init_power7_pmu); |
diff --git a/arch/powerpc/kernel/ppc970-pmu.c b/arch/powerpc/kernel/ppc970-pmu.c index 3fee685de4d..9a6e093858f 100644 --- a/arch/powerpc/kernel/ppc970-pmu.c +++ b/arch/powerpc/kernel/ppc970-pmu.c | |||
@@ -494,4 +494,4 @@ static int init_ppc970_pmu(void) | |||
494 | return register_power_pmu(&ppc970_pmu); | 494 | return register_power_pmu(&ppc970_pmu); |
495 | } | 495 | } |
496 | 496 | ||
497 | arch_initcall(init_ppc970_pmu); | 497 | early_initcall(init_ppc970_pmu); |
diff --git a/arch/sh/kernel/cpu/sh4/perf_event.c b/arch/sh/kernel/cpu/sh4/perf_event.c index dbf3b4bb71f..748955df018 100644 --- a/arch/sh/kernel/cpu/sh4/perf_event.c +++ b/arch/sh/kernel/cpu/sh4/perf_event.c | |||
@@ -250,4 +250,4 @@ static int __init sh7750_pmu_init(void) | |||
250 | 250 | ||
251 | return register_sh_pmu(&sh7750_pmu); | 251 | return register_sh_pmu(&sh7750_pmu); |
252 | } | 252 | } |
253 | arch_initcall(sh7750_pmu_init); | 253 | early_initcall(sh7750_pmu_init); |
diff --git a/arch/sh/kernel/cpu/sh4a/perf_event.c b/arch/sh/kernel/cpu/sh4a/perf_event.c index 58027652573..17e6bebfede 100644 --- a/arch/sh/kernel/cpu/sh4a/perf_event.c +++ b/arch/sh/kernel/cpu/sh4a/perf_event.c | |||
@@ -284,4 +284,4 @@ static int __init sh4a_pmu_init(void) | |||
284 | 284 | ||
285 | return register_sh_pmu(&sh4a_pmu); | 285 | return register_sh_pmu(&sh4a_pmu); |
286 | } | 286 | } |
287 | arch_initcall(sh4a_pmu_init); | 287 | early_initcall(sh4a_pmu_init); |
diff --git a/arch/sh/kernel/perf_event.c b/arch/sh/kernel/perf_event.c index 5a4b3343565..2ee21a47b5a 100644 --- a/arch/sh/kernel/perf_event.c +++ b/arch/sh/kernel/perf_event.c | |||
@@ -389,7 +389,7 @@ int __cpuinit register_sh_pmu(struct sh_pmu *_pmu) | |||
389 | 389 | ||
390 | WARN_ON(_pmu->num_events > MAX_HWEVENTS); | 390 | WARN_ON(_pmu->num_events > MAX_HWEVENTS); |
391 | 391 | ||
392 | perf_pmu_register(&pmu); | 392 | perf_pmu_register(&pmu, "cpu", PERF_TYPE_RAW); |
393 | perf_cpu_notifier(sh_pmu_notifier); | 393 | perf_cpu_notifier(sh_pmu_notifier); |
394 | return 0; | 394 | return 0; |
395 | } | 395 | } |
diff --git a/arch/sparc/include/asm/perf_event.h b/arch/sparc/include/asm/perf_event.h index 6e8bfa1786d..4d3dbe3703e 100644 --- a/arch/sparc/include/asm/perf_event.h +++ b/arch/sparc/include/asm/perf_event.h | |||
@@ -4,8 +4,6 @@ | |||
4 | #ifdef CONFIG_PERF_EVENTS | 4 | #ifdef CONFIG_PERF_EVENTS |
5 | #include <asm/ptrace.h> | 5 | #include <asm/ptrace.h> |
6 | 6 | ||
7 | extern void init_hw_perf_events(void); | ||
8 | |||
9 | #define perf_arch_fetch_caller_regs(regs, ip) \ | 7 | #define perf_arch_fetch_caller_regs(regs, ip) \ |
10 | do { \ | 8 | do { \ |
11 | unsigned long _pstate, _asi, _pil, _i7, _fp; \ | 9 | unsigned long _pstate, _asi, _pil, _i7, _fp; \ |
@@ -26,8 +24,6 @@ do { \ | |||
26 | (regs)->u_regs[UREG_I6] = _fp; \ | 24 | (regs)->u_regs[UREG_I6] = _fp; \ |
27 | (regs)->u_regs[UREG_I7] = _i7; \ | 25 | (regs)->u_regs[UREG_I7] = _i7; \ |
28 | } while (0) | 26 | } while (0) |
29 | #else | ||
30 | static inline void init_hw_perf_events(void) { } | ||
31 | #endif | 27 | #endif |
32 | 28 | ||
33 | #endif | 29 | #endif |
diff --git a/arch/sparc/kernel/nmi.c b/arch/sparc/kernel/nmi.c index a4bd7ba74c8..300f810142f 100644 --- a/arch/sparc/kernel/nmi.c +++ b/arch/sparc/kernel/nmi.c | |||
@@ -270,8 +270,6 @@ int __init nmi_init(void) | |||
270 | atomic_set(&nmi_active, -1); | 270 | atomic_set(&nmi_active, -1); |
271 | } | 271 | } |
272 | } | 272 | } |
273 | if (!err) | ||
274 | init_hw_perf_events(); | ||
275 | 273 | ||
276 | return err; | 274 | return err; |
277 | } | 275 | } |
diff --git a/arch/sparc/kernel/perf_event.c b/arch/sparc/kernel/perf_event.c index 0d6deb55a2a..760578687e7 100644 --- a/arch/sparc/kernel/perf_event.c +++ b/arch/sparc/kernel/perf_event.c | |||
@@ -1307,20 +1307,23 @@ static bool __init supported_pmu(void) | |||
1307 | return false; | 1307 | return false; |
1308 | } | 1308 | } |
1309 | 1309 | ||
1310 | void __init init_hw_perf_events(void) | 1310 | int __init init_hw_perf_events(void) |
1311 | { | 1311 | { |
1312 | pr_info("Performance events: "); | 1312 | pr_info("Performance events: "); |
1313 | 1313 | ||
1314 | if (!supported_pmu()) { | 1314 | if (!supported_pmu()) { |
1315 | pr_cont("No support for PMU type '%s'\n", sparc_pmu_type); | 1315 | pr_cont("No support for PMU type '%s'\n", sparc_pmu_type); |
1316 | return; | 1316 | return 0; |
1317 | } | 1317 | } |
1318 | 1318 | ||
1319 | pr_cont("Supported PMU type is '%s'\n", sparc_pmu_type); | 1319 | pr_cont("Supported PMU type is '%s'\n", sparc_pmu_type); |
1320 | 1320 | ||
1321 | perf_pmu_register(&pmu); | 1321 | perf_pmu_register(&pmu, "cpu", PERF_TYPE_RAW); |
1322 | register_die_notifier(&perf_event_nmi_notifier); | 1322 | register_die_notifier(&perf_event_nmi_notifier); |
1323 | |||
1324 | return 0; | ||
1323 | } | 1325 | } |
1326 | early_initcall(init_hw_perf_events); | ||
1324 | 1327 | ||
1325 | void perf_callchain_kernel(struct perf_callchain_entry *entry, | 1328 | void perf_callchain_kernel(struct perf_callchain_entry *entry, |
1326 | struct pt_regs *regs) | 1329 | struct pt_regs *regs) |
diff --git a/arch/x86/include/asm/alternative.h b/arch/x86/include/asm/alternative.h index 76561d20ea2..4a2adaa9aef 100644 --- a/arch/x86/include/asm/alternative.h +++ b/arch/x86/include/asm/alternative.h | |||
@@ -180,8 +180,15 @@ extern void *text_poke_early(void *addr, const void *opcode, size_t len); | |||
180 | * On the local CPU you need to be protected again NMI or MCE handlers seeing an | 180 | * On the local CPU you need to be protected again NMI or MCE handlers seeing an |
181 | * inconsistent instruction while you patch. | 181 | * inconsistent instruction while you patch. |
182 | */ | 182 | */ |
183 | struct text_poke_param { | ||
184 | void *addr; | ||
185 | const void *opcode; | ||
186 | size_t len; | ||
187 | }; | ||
188 | |||
183 | extern void *text_poke(void *addr, const void *opcode, size_t len); | 189 | extern void *text_poke(void *addr, const void *opcode, size_t len); |
184 | extern void *text_poke_smp(void *addr, const void *opcode, size_t len); | 190 | extern void *text_poke_smp(void *addr, const void *opcode, size_t len); |
191 | extern void text_poke_smp_batch(struct text_poke_param *params, int n); | ||
185 | 192 | ||
186 | #if defined(CONFIG_DYNAMIC_FTRACE) || defined(HAVE_JUMP_LABEL) | 193 | #if defined(CONFIG_DYNAMIC_FTRACE) || defined(HAVE_JUMP_LABEL) |
187 | #define IDEAL_NOP_SIZE_5 5 | 194 | #define IDEAL_NOP_SIZE_5 5 |
diff --git a/arch/x86/include/asm/irq.h b/arch/x86/include/asm/irq.h index 13b0ebaa512..ba870bb6dd8 100644 --- a/arch/x86/include/asm/irq.h +++ b/arch/x86/include/asm/irq.h | |||
@@ -15,10 +15,6 @@ static inline int irq_canonicalize(int irq) | |||
15 | return ((irq == 2) ? 9 : irq); | 15 | return ((irq == 2) ? 9 : irq); |
16 | } | 16 | } |
17 | 17 | ||
18 | #ifdef CONFIG_X86_LOCAL_APIC | ||
19 | # define ARCH_HAS_NMI_WATCHDOG | ||
20 | #endif | ||
21 | |||
22 | #ifdef CONFIG_X86_32 | 18 | #ifdef CONFIG_X86_32 |
23 | extern void irq_ctx_init(int cpu); | 19 | extern void irq_ctx_init(int cpu); |
24 | #else | 20 | #else |
diff --git a/arch/x86/include/asm/kdebug.h b/arch/x86/include/asm/kdebug.h index 5bdfca86581..f23eb252846 100644 --- a/arch/x86/include/asm/kdebug.h +++ b/arch/x86/include/asm/kdebug.h | |||
@@ -28,7 +28,7 @@ extern void die(const char *, struct pt_regs *,long); | |||
28 | extern int __must_check __die(const char *, struct pt_regs *, long); | 28 | extern int __must_check __die(const char *, struct pt_regs *, long); |
29 | extern void show_registers(struct pt_regs *regs); | 29 | extern void show_registers(struct pt_regs *regs); |
30 | extern void show_trace(struct task_struct *t, struct pt_regs *regs, | 30 | extern void show_trace(struct task_struct *t, struct pt_regs *regs, |
31 | unsigned long *sp, unsigned long bp); | 31 | unsigned long *sp); |
32 | extern void __show_regs(struct pt_regs *regs, int all); | 32 | extern void __show_regs(struct pt_regs *regs, int all); |
33 | extern void show_regs(struct pt_regs *regs); | 33 | extern void show_regs(struct pt_regs *regs); |
34 | extern unsigned long oops_begin(void); | 34 | extern unsigned long oops_begin(void); |
diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h index 6b89f5e8602..86030f63ba0 100644 --- a/arch/x86/include/asm/msr-index.h +++ b/arch/x86/include/asm/msr-index.h | |||
@@ -123,6 +123,10 @@ | |||
123 | #define MSR_AMD64_IBSCTL 0xc001103a | 123 | #define MSR_AMD64_IBSCTL 0xc001103a |
124 | #define MSR_AMD64_IBSBRTARGET 0xc001103b | 124 | #define MSR_AMD64_IBSBRTARGET 0xc001103b |
125 | 125 | ||
126 | /* Fam 15h MSRs */ | ||
127 | #define MSR_F15H_PERF_CTL 0xc0010200 | ||
128 | #define MSR_F15H_PERF_CTR 0xc0010201 | ||
129 | |||
126 | /* Fam 10h MSRs */ | 130 | /* Fam 10h MSRs */ |
127 | #define MSR_FAM10H_MMIO_CONF_BASE 0xc0010058 | 131 | #define MSR_FAM10H_MMIO_CONF_BASE 0xc0010058 |
128 | #define FAM10H_MMIO_CONF_ENABLE (1<<0) | 132 | #define FAM10H_MMIO_CONF_ENABLE (1<<0) |
diff --git a/arch/x86/include/asm/nmi.h b/arch/x86/include/asm/nmi.h index 932f0f86b4b..c4021b95351 100644 --- a/arch/x86/include/asm/nmi.h +++ b/arch/x86/include/asm/nmi.h | |||
@@ -5,41 +5,15 @@ | |||
5 | #include <asm/irq.h> | 5 | #include <asm/irq.h> |
6 | #include <asm/io.h> | 6 | #include <asm/io.h> |
7 | 7 | ||
8 | #ifdef ARCH_HAS_NMI_WATCHDOG | 8 | #ifdef CONFIG_X86_LOCAL_APIC |
9 | |||
10 | /** | ||
11 | * do_nmi_callback | ||
12 | * | ||
13 | * Check to see if a callback exists and execute it. Return 1 | ||
14 | * if the handler exists and was handled successfully. | ||
15 | */ | ||
16 | int do_nmi_callback(struct pt_regs *regs, int cpu); | ||
17 | 9 | ||
18 | extern void die_nmi(char *str, struct pt_regs *regs, int do_panic); | 10 | extern void die_nmi(char *str, struct pt_regs *regs, int do_panic); |
19 | extern int check_nmi_watchdog(void); | ||
20 | #if !defined(CONFIG_LOCKUP_DETECTOR) | ||
21 | extern int nmi_watchdog_enabled; | ||
22 | #endif | ||
23 | extern int avail_to_resrv_perfctr_nmi_bit(unsigned int); | 11 | extern int avail_to_resrv_perfctr_nmi_bit(unsigned int); |
24 | extern int reserve_perfctr_nmi(unsigned int); | 12 | extern int reserve_perfctr_nmi(unsigned int); |
25 | extern void release_perfctr_nmi(unsigned int); | 13 | extern void release_perfctr_nmi(unsigned int); |
26 | extern int reserve_evntsel_nmi(unsigned int); | 14 | extern int reserve_evntsel_nmi(unsigned int); |
27 | extern void release_evntsel_nmi(unsigned int); | 15 | extern void release_evntsel_nmi(unsigned int); |
28 | 16 | ||
29 | extern void setup_apic_nmi_watchdog(void *); | ||
30 | extern void stop_apic_nmi_watchdog(void *); | ||
31 | extern void disable_timer_nmi_watchdog(void); | ||
32 | extern void enable_timer_nmi_watchdog(void); | ||
33 | extern int nmi_watchdog_tick(struct pt_regs *regs, unsigned reason); | ||
34 | extern void cpu_nmi_set_wd_enabled(void); | ||
35 | |||
36 | extern atomic_t nmi_active; | ||
37 | extern unsigned int nmi_watchdog; | ||
38 | #define NMI_NONE 0 | ||
39 | #define NMI_IO_APIC 1 | ||
40 | #define NMI_LOCAL_APIC 2 | ||
41 | #define NMI_INVALID 3 | ||
42 | |||
43 | struct ctl_table; | 17 | struct ctl_table; |
44 | extern int proc_nmi_enabled(struct ctl_table *, int , | 18 | extern int proc_nmi_enabled(struct ctl_table *, int , |
45 | void __user *, size_t *, loff_t *); | 19 | void __user *, size_t *, loff_t *); |
@@ -47,33 +21,8 @@ extern int unknown_nmi_panic; | |||
47 | 21 | ||
48 | void arch_trigger_all_cpu_backtrace(void); | 22 | void arch_trigger_all_cpu_backtrace(void); |
49 | #define arch_trigger_all_cpu_backtrace arch_trigger_all_cpu_backtrace | 23 | #define arch_trigger_all_cpu_backtrace arch_trigger_all_cpu_backtrace |
50 | |||
51 | static inline void localise_nmi_watchdog(void) | ||
52 | { | ||
53 | if (nmi_watchdog == NMI_IO_APIC) | ||
54 | nmi_watchdog = NMI_LOCAL_APIC; | ||
55 | } | ||
56 | |||
57 | /* check if nmi_watchdog is active (ie was specified at boot) */ | ||
58 | static inline int nmi_watchdog_active(void) | ||
59 | { | ||
60 | /* | ||
61 | * actually it should be: | ||
62 | * return (nmi_watchdog == NMI_LOCAL_APIC || | ||
63 | * nmi_watchdog == NMI_IO_APIC) | ||
64 | * but since they are power of two we could use a | ||
65 | * cheaper way --cvg | ||
66 | */ | ||
67 | return nmi_watchdog & (NMI_LOCAL_APIC | NMI_IO_APIC); | ||
68 | } | ||
69 | #endif | 24 | #endif |
70 | 25 | ||
71 | void lapic_watchdog_stop(void); | ||
72 | int lapic_watchdog_init(unsigned nmi_hz); | ||
73 | int lapic_wd_event(unsigned nmi_hz); | ||
74 | unsigned lapic_adjust_nmi_hz(unsigned hz); | ||
75 | void disable_lapic_nmi_watchdog(void); | ||
76 | void enable_lapic_nmi_watchdog(void); | ||
77 | void stop_nmi(void); | 26 | void stop_nmi(void); |
78 | void restart_nmi(void); | 27 | void restart_nmi(void); |
79 | 28 | ||
diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h index 550e26b1dbb..d9d4dae305f 100644 --- a/arch/x86/include/asm/perf_event.h +++ b/arch/x86/include/asm/perf_event.h | |||
@@ -125,7 +125,6 @@ union cpuid10_edx { | |||
125 | #define IBS_OP_MAX_CNT_EXT 0x007FFFFFULL /* not a register bit mask */ | 125 | #define IBS_OP_MAX_CNT_EXT 0x007FFFFFULL /* not a register bit mask */ |
126 | 126 | ||
127 | #ifdef CONFIG_PERF_EVENTS | 127 | #ifdef CONFIG_PERF_EVENTS |
128 | extern void init_hw_perf_events(void); | ||
129 | extern void perf_events_lapic_init(void); | 128 | extern void perf_events_lapic_init(void); |
130 | 129 | ||
131 | #define PERF_EVENT_INDEX_OFFSET 0 | 130 | #define PERF_EVENT_INDEX_OFFSET 0 |
@@ -156,7 +155,6 @@ extern unsigned long perf_misc_flags(struct pt_regs *regs); | |||
156 | } | 155 | } |
157 | 156 | ||
158 | #else | 157 | #else |
159 | static inline void init_hw_perf_events(void) { } | ||
160 | static inline void perf_events_lapic_init(void) { } | 158 | static inline void perf_events_lapic_init(void) { } |
161 | #endif | 159 | #endif |
162 | 160 | ||
diff --git a/arch/x86/include/asm/perf_event_p4.h b/arch/x86/include/asm/perf_event_p4.h index a70cd216be5..295e2ff18a6 100644 --- a/arch/x86/include/asm/perf_event_p4.h +++ b/arch/x86/include/asm/perf_event_p4.h | |||
@@ -744,14 +744,6 @@ enum P4_ESCR_EMASKS { | |||
744 | }; | 744 | }; |
745 | 745 | ||
746 | /* | 746 | /* |
747 | * P4 PEBS specifics (Replay Event only) | ||
748 | * | ||
749 | * Format (bits): | ||
750 | * 0-6: metric from P4_PEBS_METRIC enum | ||
751 | * 7 : reserved | ||
752 | * 8 : reserved | ||
753 | * 9-11 : reserved | ||
754 | * | ||
755 | * Note we have UOP and PEBS bits reserved for now | 747 | * Note we have UOP and PEBS bits reserved for now |
756 | * just in case if we will need them once | 748 | * just in case if we will need them once |
757 | */ | 749 | */ |
@@ -788,5 +780,60 @@ enum P4_PEBS_METRIC { | |||
788 | P4_PEBS_METRIC__max | 780 | P4_PEBS_METRIC__max |
789 | }; | 781 | }; |
790 | 782 | ||
783 | /* | ||
784 | * Notes on internal configuration of ESCR+CCCR tuples | ||
785 | * | ||
786 | * Since P4 has quite the different architecture of | ||
787 | * performance registers in compare with "architectural" | ||
788 | * once and we have on 64 bits to keep configuration | ||
789 | * of performance event, the following trick is used. | ||
790 | * | ||
791 | * 1) Since both ESCR and CCCR registers have only low | ||
792 | * 32 bits valuable, we pack them into a single 64 bit | ||
793 | * configuration. Low 32 bits of such config correspond | ||
794 | * to low 32 bits of CCCR register and high 32 bits | ||
795 | * correspond to low 32 bits of ESCR register. | ||
796 | * | ||
797 | * 2) The meaning of every bit of such config field can | ||
798 | * be found in Intel SDM but it should be noted that | ||
799 | * we "borrow" some reserved bits for own usage and | ||
800 | * clean them or set to a proper value when we do | ||
801 | * a real write to hardware registers. | ||
802 | * | ||
803 | * 3) The format of bits of config is the following | ||
804 | * and should be either 0 or set to some predefined | ||
805 | * values: | ||
806 | * | ||
807 | * Low 32 bits | ||
808 | * ----------- | ||
809 | * 0-6: P4_PEBS_METRIC enum | ||
810 | * 7-11: reserved | ||
811 | * 12: reserved (Enable) | ||
812 | * 13-15: reserved (ESCR select) | ||
813 | * 16-17: Active Thread | ||
814 | * 18: Compare | ||
815 | * 19: Complement | ||
816 | * 20-23: Threshold | ||
817 | * 24: Edge | ||
818 | * 25: reserved (FORCE_OVF) | ||
819 | * 26: reserved (OVF_PMI_T0) | ||
820 | * 27: reserved (OVF_PMI_T1) | ||
821 | * 28-29: reserved | ||
822 | * 30: reserved (Cascade) | ||
823 | * 31: reserved (OVF) | ||
824 | * | ||
825 | * High 32 bits | ||
826 | * ------------ | ||
827 | * 0: reserved (T1_USR) | ||
828 | * 1: reserved (T1_OS) | ||
829 | * 2: reserved (T0_USR) | ||
830 | * 3: reserved (T0_OS) | ||
831 | * 4: Tag Enable | ||
832 | * 5-8: Tag Value | ||
833 | * 9-24: Event Mask (may use P4_ESCR_EMASK_BIT helper) | ||
834 | * 25-30: enum P4_EVENTS | ||
835 | * 31: reserved (HT thread) | ||
836 | */ | ||
837 | |||
791 | #endif /* PERF_EVENT_P4_H */ | 838 | #endif /* PERF_EVENT_P4_H */ |
792 | 839 | ||
diff --git a/arch/x86/include/asm/smpboot_hooks.h b/arch/x86/include/asm/smpboot_hooks.h index 1def6011490..6c22bf353f2 100644 --- a/arch/x86/include/asm/smpboot_hooks.h +++ b/arch/x86/include/asm/smpboot_hooks.h | |||
@@ -48,7 +48,6 @@ static inline void __init smpboot_setup_io_apic(void) | |||
48 | setup_IO_APIC(); | 48 | setup_IO_APIC(); |
49 | else { | 49 | else { |
50 | nr_ioapics = 0; | 50 | nr_ioapics = 0; |
51 | localise_nmi_watchdog(); | ||
52 | } | 51 | } |
53 | #endif | 52 | #endif |
54 | } | 53 | } |
diff --git a/arch/x86/include/asm/stacktrace.h b/arch/x86/include/asm/stacktrace.h index 2b16a2ad23d..52b5c7ed360 100644 --- a/arch/x86/include/asm/stacktrace.h +++ b/arch/x86/include/asm/stacktrace.h | |||
@@ -7,6 +7,7 @@ | |||
7 | #define _ASM_X86_STACKTRACE_H | 7 | #define _ASM_X86_STACKTRACE_H |
8 | 8 | ||
9 | #include <linux/uaccess.h> | 9 | #include <linux/uaccess.h> |
10 | #include <linux/ptrace.h> | ||
10 | 11 | ||
11 | extern int kstack_depth_to_print; | 12 | extern int kstack_depth_to_print; |
12 | 13 | ||
@@ -46,7 +47,7 @@ struct stacktrace_ops { | |||
46 | }; | 47 | }; |
47 | 48 | ||
48 | void dump_trace(struct task_struct *tsk, struct pt_regs *regs, | 49 | void dump_trace(struct task_struct *tsk, struct pt_regs *regs, |
49 | unsigned long *stack, unsigned long bp, | 50 | unsigned long *stack, |
50 | const struct stacktrace_ops *ops, void *data); | 51 | const struct stacktrace_ops *ops, void *data); |
51 | 52 | ||
52 | #ifdef CONFIG_X86_32 | 53 | #ifdef CONFIG_X86_32 |
@@ -57,13 +58,39 @@ void dump_trace(struct task_struct *tsk, struct pt_regs *regs, | |||
57 | #define get_bp(bp) asm("movq %%rbp, %0" : "=r" (bp) :) | 58 | #define get_bp(bp) asm("movq %%rbp, %0" : "=r" (bp) :) |
58 | #endif | 59 | #endif |
59 | 60 | ||
61 | #ifdef CONFIG_FRAME_POINTER | ||
62 | static inline unsigned long | ||
63 | stack_frame(struct task_struct *task, struct pt_regs *regs) | ||
64 | { | ||
65 | unsigned long bp; | ||
66 | |||
67 | if (regs) | ||
68 | return regs->bp; | ||
69 | |||
70 | if (task == current) { | ||
71 | /* Grab bp right from our regs */ | ||
72 | get_bp(bp); | ||
73 | return bp; | ||
74 | } | ||
75 | |||
76 | /* bp is the last reg pushed by switch_to */ | ||
77 | return *(unsigned long *)task->thread.sp; | ||
78 | } | ||
79 | #else | ||
80 | static inline unsigned long | ||
81 | stack_frame(struct task_struct *task, struct pt_regs *regs) | ||
82 | { | ||
83 | return 0; | ||
84 | } | ||
85 | #endif | ||
86 | |||
60 | extern void | 87 | extern void |
61 | show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, | 88 | show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, |
62 | unsigned long *stack, unsigned long bp, char *log_lvl); | 89 | unsigned long *stack, char *log_lvl); |
63 | 90 | ||
64 | extern void | 91 | extern void |
65 | show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, | 92 | show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, |
66 | unsigned long *sp, unsigned long bp, char *log_lvl); | 93 | unsigned long *sp, char *log_lvl); |
67 | 94 | ||
68 | extern unsigned int code_bytes; | 95 | extern unsigned int code_bytes; |
69 | 96 | ||
diff --git a/arch/x86/include/asm/timer.h b/arch/x86/include/asm/timer.h index 5469630b27f..fa7b9176b76 100644 --- a/arch/x86/include/asm/timer.h +++ b/arch/x86/include/asm/timer.h | |||
@@ -10,12 +10,6 @@ | |||
10 | unsigned long long native_sched_clock(void); | 10 | unsigned long long native_sched_clock(void); |
11 | extern int recalibrate_cpu_khz(void); | 11 | extern int recalibrate_cpu_khz(void); |
12 | 12 | ||
13 | #if defined(CONFIG_X86_32) && defined(CONFIG_X86_IO_APIC) | ||
14 | extern int timer_ack; | ||
15 | #else | ||
16 | # define timer_ack (0) | ||
17 | #endif | ||
18 | |||
19 | extern int no_timer_check; | 13 | extern int no_timer_check; |
20 | 14 | ||
21 | /* Accelerators for sched_clock() | 15 | /* Accelerators for sched_clock() |
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c index 5079f24c955..553d0b0d639 100644 --- a/arch/x86/kernel/alternative.c +++ b/arch/x86/kernel/alternative.c | |||
@@ -591,17 +591,21 @@ static atomic_t stop_machine_first; | |||
591 | static int wrote_text; | 591 | static int wrote_text; |
592 | 592 | ||
593 | struct text_poke_params { | 593 | struct text_poke_params { |
594 | void *addr; | 594 | struct text_poke_param *params; |
595 | const void *opcode; | 595 | int nparams; |
596 | size_t len; | ||
597 | }; | 596 | }; |
598 | 597 | ||
599 | static int __kprobes stop_machine_text_poke(void *data) | 598 | static int __kprobes stop_machine_text_poke(void *data) |
600 | { | 599 | { |
601 | struct text_poke_params *tpp = data; | 600 | struct text_poke_params *tpp = data; |
601 | struct text_poke_param *p; | ||
602 | int i; | ||
602 | 603 | ||
603 | if (atomic_dec_and_test(&stop_machine_first)) { | 604 | if (atomic_dec_and_test(&stop_machine_first)) { |
604 | text_poke(tpp->addr, tpp->opcode, tpp->len); | 605 | for (i = 0; i < tpp->nparams; i++) { |
606 | p = &tpp->params[i]; | ||
607 | text_poke(p->addr, p->opcode, p->len); | ||
608 | } | ||
605 | smp_wmb(); /* Make sure other cpus see that this has run */ | 609 | smp_wmb(); /* Make sure other cpus see that this has run */ |
606 | wrote_text = 1; | 610 | wrote_text = 1; |
607 | } else { | 611 | } else { |
@@ -610,8 +614,12 @@ static int __kprobes stop_machine_text_poke(void *data) | |||
610 | smp_mb(); /* Load wrote_text before following execution */ | 614 | smp_mb(); /* Load wrote_text before following execution */ |
611 | } | 615 | } |
612 | 616 | ||
613 | flush_icache_range((unsigned long)tpp->addr, | 617 | for (i = 0; i < tpp->nparams; i++) { |
614 | (unsigned long)tpp->addr + tpp->len); | 618 | p = &tpp->params[i]; |
619 | flush_icache_range((unsigned long)p->addr, | ||
620 | (unsigned long)p->addr + p->len); | ||
621 | } | ||
622 | |||
615 | return 0; | 623 | return 0; |
616 | } | 624 | } |
617 | 625 | ||
@@ -631,10 +639,13 @@ static int __kprobes stop_machine_text_poke(void *data) | |||
631 | void *__kprobes text_poke_smp(void *addr, const void *opcode, size_t len) | 639 | void *__kprobes text_poke_smp(void *addr, const void *opcode, size_t len) |
632 | { | 640 | { |
633 | struct text_poke_params tpp; | 641 | struct text_poke_params tpp; |
642 | struct text_poke_param p; | ||
634 | 643 | ||
635 | tpp.addr = addr; | 644 | p.addr = addr; |
636 | tpp.opcode = opcode; | 645 | p.opcode = opcode; |
637 | tpp.len = len; | 646 | p.len = len; |
647 | tpp.params = &p; | ||
648 | tpp.nparams = 1; | ||
638 | atomic_set(&stop_machine_first, 1); | 649 | atomic_set(&stop_machine_first, 1); |
639 | wrote_text = 0; | 650 | wrote_text = 0; |
640 | /* Use __stop_machine() because the caller already got online_cpus. */ | 651 | /* Use __stop_machine() because the caller already got online_cpus. */ |
@@ -642,6 +653,26 @@ void *__kprobes text_poke_smp(void *addr, const void *opcode, size_t len) | |||
642 | return addr; | 653 | return addr; |
643 | } | 654 | } |
644 | 655 | ||
656 | /** | ||
657 | * text_poke_smp_batch - Update instructions on a live kernel on SMP | ||
658 | * @params: an array of text_poke parameters | ||
659 | * @n: the number of elements in params. | ||
660 | * | ||
661 | * Modify multi-byte instruction by using stop_machine() on SMP. Since the | ||
662 | * stop_machine() is heavy task, it is better to aggregate text_poke requests | ||
663 | * and do it once if possible. | ||
664 | * | ||
665 | * Note: Must be called under get_online_cpus() and text_mutex. | ||
666 | */ | ||
667 | void __kprobes text_poke_smp_batch(struct text_poke_param *params, int n) | ||
668 | { | ||
669 | struct text_poke_params tpp = {.params = params, .nparams = n}; | ||
670 | |||
671 | atomic_set(&stop_machine_first, 1); | ||
672 | wrote_text = 0; | ||
673 | stop_machine(stop_machine_text_poke, (void *)&tpp, NULL); | ||
674 | } | ||
675 | |||
645 | #if defined(CONFIG_DYNAMIC_FTRACE) || defined(HAVE_JUMP_LABEL) | 676 | #if defined(CONFIG_DYNAMIC_FTRACE) || defined(HAVE_JUMP_LABEL) |
646 | 677 | ||
647 | #ifdef CONFIG_X86_64 | 678 | #ifdef CONFIG_X86_64 |
diff --git a/arch/x86/kernel/apic/Makefile b/arch/x86/kernel/apic/Makefile index 910f20b457c..3966b564ea4 100644 --- a/arch/x86/kernel/apic/Makefile +++ b/arch/x86/kernel/apic/Makefile | |||
@@ -3,10 +3,7 @@ | |||
3 | # | 3 | # |
4 | 4 | ||
5 | obj-$(CONFIG_X86_LOCAL_APIC) += apic.o apic_noop.o probe_$(BITS).o ipi.o | 5 | obj-$(CONFIG_X86_LOCAL_APIC) += apic.o apic_noop.o probe_$(BITS).o ipi.o |
6 | ifneq ($(CONFIG_HARDLOCKUP_DETECTOR),y) | 6 | obj-y += hw_nmi.o |
7 | obj-$(CONFIG_X86_LOCAL_APIC) += nmi.o | ||
8 | endif | ||
9 | obj-$(CONFIG_HARDLOCKUP_DETECTOR) += hw_nmi.o | ||
10 | 7 | ||
11 | obj-$(CONFIG_X86_IO_APIC) += io_apic.o | 8 | obj-$(CONFIG_X86_IO_APIC) += io_apic.o |
12 | obj-$(CONFIG_SMP) += ipi.o | 9 | obj-$(CONFIG_SMP) += ipi.o |
diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c index 78218135b48..fb7657822aa 100644 --- a/arch/x86/kernel/apic/apic.c +++ b/arch/x86/kernel/apic/apic.c | |||
@@ -31,7 +31,6 @@ | |||
31 | #include <linux/init.h> | 31 | #include <linux/init.h> |
32 | #include <linux/cpu.h> | 32 | #include <linux/cpu.h> |
33 | #include <linux/dmi.h> | 33 | #include <linux/dmi.h> |
34 | #include <linux/nmi.h> | ||
35 | #include <linux/smp.h> | 34 | #include <linux/smp.h> |
36 | #include <linux/mm.h> | 35 | #include <linux/mm.h> |
37 | 36 | ||
@@ -799,11 +798,7 @@ void __init setup_boot_APIC_clock(void) | |||
799 | * PIT/HPET going. Otherwise register lapic as a dummy | 798 | * PIT/HPET going. Otherwise register lapic as a dummy |
800 | * device. | 799 | * device. |
801 | */ | 800 | */ |
802 | if (nmi_watchdog != NMI_IO_APIC) | 801 | lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY; |
803 | lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY; | ||
804 | else | ||
805 | pr_warning("APIC timer registered as dummy," | ||
806 | " due to nmi_watchdog=%d!\n", nmi_watchdog); | ||
807 | 802 | ||
808 | /* Setup the lapic or request the broadcast */ | 803 | /* Setup the lapic or request the broadcast */ |
809 | setup_APIC_timer(); | 804 | setup_APIC_timer(); |
@@ -1387,7 +1382,6 @@ void __cpuinit end_local_APIC_setup(void) | |||
1387 | } | 1382 | } |
1388 | #endif | 1383 | #endif |
1389 | 1384 | ||
1390 | setup_apic_nmi_watchdog(NULL); | ||
1391 | apic_pm_activate(); | 1385 | apic_pm_activate(); |
1392 | 1386 | ||
1393 | /* | 1387 | /* |
@@ -1758,17 +1752,10 @@ int __init APIC_init_uniprocessor(void) | |||
1758 | setup_IO_APIC(); | 1752 | setup_IO_APIC(); |
1759 | else { | 1753 | else { |
1760 | nr_ioapics = 0; | 1754 | nr_ioapics = 0; |
1761 | localise_nmi_watchdog(); | ||
1762 | } | 1755 | } |
1763 | #else | ||
1764 | localise_nmi_watchdog(); | ||
1765 | #endif | 1756 | #endif |
1766 | 1757 | ||
1767 | x86_init.timers.setup_percpu_clockev(); | 1758 | x86_init.timers.setup_percpu_clockev(); |
1768 | #ifdef CONFIG_X86_64 | ||
1769 | check_nmi_watchdog(); | ||
1770 | #endif | ||
1771 | |||
1772 | return 0; | 1759 | return 0; |
1773 | } | 1760 | } |
1774 | 1761 | ||
diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c index 62f6e1e55b9..72ec29e1ae0 100644 --- a/arch/x86/kernel/apic/hw_nmi.c +++ b/arch/x86/kernel/apic/hw_nmi.c | |||
@@ -17,20 +17,31 @@ | |||
17 | #include <linux/nmi.h> | 17 | #include <linux/nmi.h> |
18 | #include <linux/module.h> | 18 | #include <linux/module.h> |
19 | 19 | ||
20 | #ifdef CONFIG_HARDLOCKUP_DETECTOR | ||
20 | u64 hw_nmi_get_sample_period(void) | 21 | u64 hw_nmi_get_sample_period(void) |
21 | { | 22 | { |
22 | return (u64)(cpu_khz) * 1000 * 60; | 23 | return (u64)(cpu_khz) * 1000 * 60; |
23 | } | 24 | } |
25 | #endif | ||
24 | 26 | ||
25 | #ifdef ARCH_HAS_NMI_WATCHDOG | 27 | #ifdef arch_trigger_all_cpu_backtrace |
26 | |||
27 | /* For reliability, we're prepared to waste bits here. */ | 28 | /* For reliability, we're prepared to waste bits here. */ |
28 | static DECLARE_BITMAP(backtrace_mask, NR_CPUS) __read_mostly; | 29 | static DECLARE_BITMAP(backtrace_mask, NR_CPUS) __read_mostly; |
29 | 30 | ||
31 | /* "in progress" flag of arch_trigger_all_cpu_backtrace */ | ||
32 | static unsigned long backtrace_flag; | ||
33 | |||
30 | void arch_trigger_all_cpu_backtrace(void) | 34 | void arch_trigger_all_cpu_backtrace(void) |
31 | { | 35 | { |
32 | int i; | 36 | int i; |
33 | 37 | ||
38 | if (test_and_set_bit(0, &backtrace_flag)) | ||
39 | /* | ||
40 | * If there is already a trigger_all_cpu_backtrace() in progress | ||
41 | * (backtrace_flag == 1), don't output double cpu dump infos. | ||
42 | */ | ||
43 | return; | ||
44 | |||
34 | cpumask_copy(to_cpumask(backtrace_mask), cpu_online_mask); | 45 | cpumask_copy(to_cpumask(backtrace_mask), cpu_online_mask); |
35 | 46 | ||
36 | printk(KERN_INFO "sending NMI to all CPUs:\n"); | 47 | printk(KERN_INFO "sending NMI to all CPUs:\n"); |
@@ -42,6 +53,9 @@ void arch_trigger_all_cpu_backtrace(void) | |||
42 | break; | 53 | break; |
43 | mdelay(1); | 54 | mdelay(1); |
44 | } | 55 | } |
56 | |||
57 | clear_bit(0, &backtrace_flag); | ||
58 | smp_mb__after_clear_bit(); | ||
45 | } | 59 | } |
46 | 60 | ||
47 | static int __kprobes | 61 | static int __kprobes |
@@ -50,7 +64,7 @@ arch_trigger_all_cpu_backtrace_handler(struct notifier_block *self, | |||
50 | { | 64 | { |
51 | struct die_args *args = __args; | 65 | struct die_args *args = __args; |
52 | struct pt_regs *regs; | 66 | struct pt_regs *regs; |
53 | int cpu = smp_processor_id(); | 67 | int cpu; |
54 | 68 | ||
55 | switch (cmd) { | 69 | switch (cmd) { |
56 | case DIE_NMI: | 70 | case DIE_NMI: |
@@ -62,6 +76,7 @@ arch_trigger_all_cpu_backtrace_handler(struct notifier_block *self, | |||
62 | } | 76 | } |
63 | 77 | ||
64 | regs = args->regs; | 78 | regs = args->regs; |
79 | cpu = smp_processor_id(); | ||
65 | 80 | ||
66 | if (cpumask_test_cpu(cpu, to_cpumask(backtrace_mask))) { | 81 | if (cpumask_test_cpu(cpu, to_cpumask(backtrace_mask))) { |
67 | static arch_spinlock_t lock = __ARCH_SPIN_LOCK_UNLOCKED; | 82 | static arch_spinlock_t lock = __ARCH_SPIN_LOCK_UNLOCKED; |
@@ -91,18 +106,3 @@ static int __init register_trigger_all_cpu_backtrace(void) | |||
91 | } | 106 | } |
92 | early_initcall(register_trigger_all_cpu_backtrace); | 107 | early_initcall(register_trigger_all_cpu_backtrace); |
93 | #endif | 108 | #endif |
94 | |||
95 | /* STUB calls to mimic old nmi_watchdog behaviour */ | ||
96 | #if defined(CONFIG_X86_LOCAL_APIC) | ||
97 | unsigned int nmi_watchdog = NMI_NONE; | ||
98 | EXPORT_SYMBOL(nmi_watchdog); | ||
99 | void acpi_nmi_enable(void) { return; } | ||
100 | void acpi_nmi_disable(void) { return; } | ||
101 | #endif | ||
102 | atomic_t nmi_active = ATOMIC_INIT(0); /* oprofile uses this */ | ||
103 | EXPORT_SYMBOL(nmi_active); | ||
104 | int unknown_nmi_panic; | ||
105 | void cpu_nmi_set_wd_enabled(void) { return; } | ||
106 | void stop_apic_nmi_watchdog(void *unused) { return; } | ||
107 | void setup_apic_nmi_watchdog(void *unused) { return; } | ||
108 | int __init check_nmi_watchdog(void) { return 0; } | ||
diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c index fadcd743a74..16c2db8750a 100644 --- a/arch/x86/kernel/apic/io_apic.c +++ b/arch/x86/kernel/apic/io_apic.c | |||
@@ -54,7 +54,6 @@ | |||
54 | #include <asm/dma.h> | 54 | #include <asm/dma.h> |
55 | #include <asm/timer.h> | 55 | #include <asm/timer.h> |
56 | #include <asm/i8259.h> | 56 | #include <asm/i8259.h> |
57 | #include <asm/nmi.h> | ||
58 | #include <asm/msidef.h> | 57 | #include <asm/msidef.h> |
59 | #include <asm/hypertransport.h> | 58 | #include <asm/hypertransport.h> |
60 | #include <asm/setup.h> | 59 | #include <asm/setup.h> |
@@ -2642,24 +2641,6 @@ static void lapic_register_intr(int irq) | |||
2642 | "edge"); | 2641 | "edge"); |
2643 | } | 2642 | } |
2644 | 2643 | ||
2645 | static void __init setup_nmi(void) | ||
2646 | { | ||
2647 | /* | ||
2648 | * Dirty trick to enable the NMI watchdog ... | ||
2649 | * We put the 8259A master into AEOI mode and | ||
2650 | * unmask on all local APICs LVT0 as NMI. | ||
2651 | * | ||
2652 | * The idea to use the 8259A in AEOI mode ('8259A Virtual Wire') | ||
2653 | * is from Maciej W. Rozycki - so we do not have to EOI from | ||
2654 | * the NMI handler or the timer interrupt. | ||
2655 | */ | ||
2656 | apic_printk(APIC_VERBOSE, KERN_INFO "activating NMI Watchdog ..."); | ||
2657 | |||
2658 | enable_NMI_through_LVT0(); | ||
2659 | |||
2660 | apic_printk(APIC_VERBOSE, " done.\n"); | ||
2661 | } | ||
2662 | |||
2663 | /* | 2644 | /* |
2664 | * This looks a bit hackish but it's about the only one way of sending | 2645 | * This looks a bit hackish but it's about the only one way of sending |
2665 | * a few INTA cycles to 8259As and any associated glue logic. ICR does | 2646 | * a few INTA cycles to 8259As and any associated glue logic. ICR does |
@@ -2765,15 +2746,6 @@ static inline void __init check_timer(void) | |||
2765 | */ | 2746 | */ |
2766 | apic_write(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_EXTINT); | 2747 | apic_write(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_EXTINT); |
2767 | legacy_pic->init(1); | 2748 | legacy_pic->init(1); |
2768 | #ifdef CONFIG_X86_32 | ||
2769 | { | ||
2770 | unsigned int ver; | ||
2771 | |||
2772 | ver = apic_read(APIC_LVR); | ||
2773 | ver = GET_APIC_VERSION(ver); | ||
2774 | timer_ack = (nmi_watchdog == NMI_IO_APIC && !APIC_INTEGRATED(ver)); | ||
2775 | } | ||
2776 | #endif | ||
2777 | 2749 | ||
2778 | pin1 = find_isa_irq_pin(0, mp_INT); | 2750 | pin1 = find_isa_irq_pin(0, mp_INT); |
2779 | apic1 = find_isa_irq_apic(0, mp_INT); | 2751 | apic1 = find_isa_irq_apic(0, mp_INT); |
@@ -2821,10 +2793,6 @@ static inline void __init check_timer(void) | |||
2821 | unmask_ioapic(cfg); | 2793 | unmask_ioapic(cfg); |
2822 | } | 2794 | } |
2823 | if (timer_irq_works()) { | 2795 | if (timer_irq_works()) { |
2824 | if (nmi_watchdog == NMI_IO_APIC) { | ||
2825 | setup_nmi(); | ||
2826 | legacy_pic->unmask(0); | ||
2827 | } | ||
2828 | if (disable_timer_pin_1 > 0) | 2796 | if (disable_timer_pin_1 > 0) |
2829 | clear_IO_APIC_pin(0, pin1); | 2797 | clear_IO_APIC_pin(0, pin1); |
2830 | goto out; | 2798 | goto out; |
@@ -2850,11 +2818,6 @@ static inline void __init check_timer(void) | |||
2850 | if (timer_irq_works()) { | 2818 | if (timer_irq_works()) { |
2851 | apic_printk(APIC_QUIET, KERN_INFO "....... works.\n"); | 2819 | apic_printk(APIC_QUIET, KERN_INFO "....... works.\n"); |
2852 | timer_through_8259 = 1; | 2820 | timer_through_8259 = 1; |
2853 | if (nmi_watchdog == NMI_IO_APIC) { | ||
2854 | legacy_pic->mask(0); | ||
2855 | setup_nmi(); | ||
2856 | legacy_pic->unmask(0); | ||
2857 | } | ||
2858 | goto out; | 2821 | goto out; |
2859 | } | 2822 | } |
2860 | /* | 2823 | /* |
@@ -2866,15 +2829,6 @@ static inline void __init check_timer(void) | |||
2866 | apic_printk(APIC_QUIET, KERN_INFO "....... failed.\n"); | 2829 | apic_printk(APIC_QUIET, KERN_INFO "....... failed.\n"); |
2867 | } | 2830 | } |
2868 | 2831 | ||
2869 | if (nmi_watchdog == NMI_IO_APIC) { | ||
2870 | apic_printk(APIC_QUIET, KERN_WARNING "timer doesn't work " | ||
2871 | "through the IO-APIC - disabling NMI Watchdog!\n"); | ||
2872 | nmi_watchdog = NMI_NONE; | ||
2873 | } | ||
2874 | #ifdef CONFIG_X86_32 | ||
2875 | timer_ack = 0; | ||
2876 | #endif | ||
2877 | |||
2878 | apic_printk(APIC_QUIET, KERN_INFO | 2832 | apic_printk(APIC_QUIET, KERN_INFO |
2879 | "...trying to set up timer as Virtual Wire IRQ...\n"); | 2833 | "...trying to set up timer as Virtual Wire IRQ...\n"); |
2880 | 2834 | ||
diff --git a/arch/x86/kernel/apic/nmi.c b/arch/x86/kernel/apic/nmi.c deleted file mode 100644 index c90041ccb74..00000000000 --- a/arch/x86/kernel/apic/nmi.c +++ /dev/null | |||
@@ -1,567 +0,0 @@ | |||
1 | /* | ||
2 | * NMI watchdog support on APIC systems | ||
3 | * | ||
4 | * Started by Ingo Molnar <mingo@redhat.com> | ||
5 | * | ||
6 | * Fixes: | ||
7 | * Mikael Pettersson : AMD K7 support for local APIC NMI watchdog. | ||
8 | * Mikael Pettersson : Power Management for local APIC NMI watchdog. | ||
9 | * Mikael Pettersson : Pentium 4 support for local APIC NMI watchdog. | ||
10 | * Pavel Machek and | ||
11 | * Mikael Pettersson : PM converted to driver model. Disable/enable API. | ||
12 | */ | ||
13 | |||
14 | #include <asm/apic.h> | ||
15 | |||
16 | #include <linux/nmi.h> | ||
17 | #include <linux/mm.h> | ||
18 | #include <linux/delay.h> | ||
19 | #include <linux/interrupt.h> | ||
20 | #include <linux/module.h> | ||
21 | #include <linux/slab.h> | ||
22 | #include <linux/sysdev.h> | ||
23 | #include <linux/sysctl.h> | ||
24 | #include <linux/percpu.h> | ||
25 | #include <linux/kprobes.h> | ||
26 | #include <linux/cpumask.h> | ||
27 | #include <linux/kernel_stat.h> | ||
28 | #include <linux/kdebug.h> | ||
29 | #include <linux/smp.h> | ||
30 | |||
31 | #include <asm/i8259.h> | ||
32 | #include <asm/io_apic.h> | ||
33 | #include <asm/proto.h> | ||
34 | #include <asm/timer.h> | ||
35 | |||
36 | #include <asm/mce.h> | ||
37 | |||
38 | #include <asm/mach_traps.h> | ||
39 | |||
40 | int unknown_nmi_panic; | ||
41 | int nmi_watchdog_enabled; | ||
42 | |||
43 | /* For reliability, we're prepared to waste bits here. */ | ||
44 | static DECLARE_BITMAP(backtrace_mask, NR_CPUS) __read_mostly; | ||
45 | |||
46 | /* nmi_active: | ||
47 | * >0: the lapic NMI watchdog is active, but can be disabled | ||
48 | * <0: the lapic NMI watchdog has not been set up, and cannot | ||
49 | * be enabled | ||
50 | * 0: the lapic NMI watchdog is disabled, but can be enabled | ||
51 | */ | ||
52 | atomic_t nmi_active = ATOMIC_INIT(0); /* oprofile uses this */ | ||
53 | EXPORT_SYMBOL(nmi_active); | ||
54 | |||
55 | unsigned int nmi_watchdog = NMI_NONE; | ||
56 | EXPORT_SYMBOL(nmi_watchdog); | ||
57 | |||
58 | static int panic_on_timeout; | ||
59 | |||
60 | static unsigned int nmi_hz = HZ; | ||
61 | static DEFINE_PER_CPU(short, wd_enabled); | ||
62 | static int endflag __initdata; | ||
63 | |||
64 | static inline unsigned int get_nmi_count(int cpu) | ||
65 | { | ||
66 | return per_cpu(irq_stat, cpu).__nmi_count; | ||
67 | } | ||
68 | |||
69 | static inline int mce_in_progress(void) | ||
70 | { | ||
71 | #if defined(CONFIG_X86_MCE) | ||
72 | return atomic_read(&mce_entry) > 0; | ||
73 | #endif | ||
74 | return 0; | ||
75 | } | ||
76 | |||
77 | /* | ||
78 | * Take the local apic timer and PIT/HPET into account. We don't | ||
79 | * know which one is active, when we have highres/dyntick on | ||
80 | */ | ||
81 | static inline unsigned int get_timer_irqs(int cpu) | ||
82 | { | ||
83 | return per_cpu(irq_stat, cpu).apic_timer_irqs + | ||
84 | per_cpu(irq_stat, cpu).irq0_irqs; | ||
85 | } | ||
86 | |||
87 | #ifdef CONFIG_SMP | ||
88 | /* | ||
89 | * The performance counters used by NMI_LOCAL_APIC don't trigger when | ||
90 | * the CPU is idle. To make sure the NMI watchdog really ticks on all | ||
91 | * CPUs during the test make them busy. | ||
92 | */ | ||
93 | static __init void nmi_cpu_busy(void *data) | ||
94 | { | ||
95 | local_irq_enable_in_hardirq(); | ||
96 | /* | ||
97 | * Intentionally don't use cpu_relax here. This is | ||
98 | * to make sure that the performance counter really ticks, | ||
99 | * even if there is a simulator or similar that catches the | ||
100 | * pause instruction. On a real HT machine this is fine because | ||
101 | * all other CPUs are busy with "useless" delay loops and don't | ||
102 | * care if they get somewhat less cycles. | ||
103 | */ | ||
104 | while (endflag == 0) | ||
105 | mb(); | ||
106 | } | ||
107 | #endif | ||
108 | |||
109 | static void report_broken_nmi(int cpu, unsigned int *prev_nmi_count) | ||
110 | { | ||
111 | printk(KERN_CONT "\n"); | ||
112 | |||
113 | printk(KERN_WARNING | ||
114 | "WARNING: CPU#%d: NMI appears to be stuck (%d->%d)!\n", | ||
115 | cpu, prev_nmi_count[cpu], get_nmi_count(cpu)); | ||
116 | |||
117 | printk(KERN_WARNING | ||
118 | "Please report this to bugzilla.kernel.org,\n"); | ||
119 | printk(KERN_WARNING | ||
120 | "and attach the output of the 'dmesg' command.\n"); | ||
121 | |||
122 | per_cpu(wd_enabled, cpu) = 0; | ||
123 | atomic_dec(&nmi_active); | ||
124 | } | ||
125 | |||
126 | static void __acpi_nmi_disable(void *__unused) | ||
127 | { | ||
128 | apic_write(APIC_LVT0, APIC_DM_NMI | APIC_LVT_MASKED); | ||
129 | } | ||
130 | |||
131 | int __init check_nmi_watchdog(void) | ||
132 | { | ||
133 | unsigned int *prev_nmi_count; | ||
134 | int cpu; | ||
135 | |||
136 | if (!nmi_watchdog_active() || !atomic_read(&nmi_active)) | ||
137 | return 0; | ||
138 | |||
139 | prev_nmi_count = kmalloc(nr_cpu_ids * sizeof(int), GFP_KERNEL); | ||
140 | if (!prev_nmi_count) | ||
141 | goto error; | ||
142 | |||
143 | printk(KERN_INFO "Testing NMI watchdog ... "); | ||
144 | |||
145 | #ifdef CONFIG_SMP | ||
146 | if (nmi_watchdog == NMI_LOCAL_APIC) | ||
147 | smp_call_function(nmi_cpu_busy, (void *)&endflag, 0); | ||
148 | #endif | ||
149 | |||
150 | for_each_possible_cpu(cpu) | ||
151 | prev_nmi_count[cpu] = get_nmi_count(cpu); | ||
152 | local_irq_enable(); | ||
153 | mdelay((20 * 1000) / nmi_hz); /* wait 20 ticks */ | ||
154 | |||
155 | for_each_online_cpu(cpu) { | ||
156 | if (!per_cpu(wd_enabled, cpu)) | ||
157 | continue; | ||
158 | if (get_nmi_count(cpu) - prev_nmi_count[cpu] <= 5) | ||
159 | report_broken_nmi(cpu, prev_nmi_count); | ||
160 | } | ||
161 | endflag = 1; | ||
162 | if (!atomic_read(&nmi_active)) { | ||
163 | kfree(prev_nmi_count); | ||
164 | atomic_set(&nmi_active, -1); | ||
165 | goto error; | ||
166 | } | ||
167 | printk("OK.\n"); | ||
168 | |||
169 | /* | ||
170 | * now that we know it works we can reduce NMI frequency to | ||
171 | * something more reasonable; makes a difference in some configs | ||
172 | */ | ||
173 | if (nmi_watchdog == NMI_LOCAL_APIC) | ||
174 | nmi_hz = lapic_adjust_nmi_hz(1); | ||
175 | |||
176 | kfree(prev_nmi_count); | ||
177 | return 0; | ||
178 | error: | ||
179 | if (nmi_watchdog == NMI_IO_APIC) { | ||
180 | if (!timer_through_8259) | ||
181 | legacy_pic->mask(0); | ||
182 | on_each_cpu(__acpi_nmi_disable, NULL, 1); | ||
183 | } | ||
184 | |||
185 | #ifdef CONFIG_X86_32 | ||
186 | timer_ack = 0; | ||
187 | #endif | ||
188 | return -1; | ||
189 | } | ||
190 | |||
191 | static int __init setup_nmi_watchdog(char *str) | ||
192 | { | ||
193 | unsigned int nmi; | ||
194 | |||
195 | if (!strncmp(str, "panic", 5)) { | ||
196 | panic_on_timeout = 1; | ||
197 | str = strchr(str, ','); | ||
198 | if (!str) | ||
199 | return 1; | ||
200 | ++str; | ||
201 | } | ||
202 | |||
203 | if (!strncmp(str, "lapic", 5)) | ||
204 | nmi_watchdog = NMI_LOCAL_APIC; | ||
205 | else if (!strncmp(str, "ioapic", 6)) | ||
206 | nmi_watchdog = NMI_IO_APIC; | ||
207 | else { | ||
208 | get_option(&str, &nmi); | ||
209 | if (nmi >= NMI_INVALID) | ||
210 | return 0; | ||
211 | nmi_watchdog = nmi; | ||
212 | } | ||
213 | |||
214 | return 1; | ||
215 | } | ||
216 | __setup("nmi_watchdog=", setup_nmi_watchdog); | ||
217 | |||
218 | /* | ||
219 | * Suspend/resume support | ||
220 | */ | ||
221 | #ifdef CONFIG_PM | ||
222 | |||
223 | static int nmi_pm_active; /* nmi_active before suspend */ | ||
224 | |||
225 | static int lapic_nmi_suspend(struct sys_device *dev, pm_message_t state) | ||
226 | { | ||
227 | /* only CPU0 goes here, other CPUs should be offline */ | ||
228 | nmi_pm_active = atomic_read(&nmi_active); | ||
229 | stop_apic_nmi_watchdog(NULL); | ||
230 | BUG_ON(atomic_read(&nmi_active) != 0); | ||
231 | return 0; | ||
232 | } | ||
233 | |||
234 | static int lapic_nmi_resume(struct sys_device *dev) | ||
235 | { | ||
236 | /* only CPU0 goes here, other CPUs should be offline */ | ||
237 | if (nmi_pm_active > 0) { | ||
238 | setup_apic_nmi_watchdog(NULL); | ||
239 | touch_nmi_watchdog(); | ||
240 | } | ||
241 | return 0; | ||
242 | } | ||
243 | |||
244 | static struct sysdev_class nmi_sysclass = { | ||
245 | .name = "lapic_nmi", | ||
246 | .resume = lapic_nmi_resume, | ||
247 | .suspend = lapic_nmi_suspend, | ||
248 | }; | ||
249 | |||
250 | static struct sys_device device_lapic_nmi = { | ||
251 | .id = 0, | ||
252 | .cls = &nmi_sysclass, | ||
253 | }; | ||
254 | |||
255 | static int __init init_lapic_nmi_sysfs(void) | ||
256 | { | ||
257 | int error; | ||
258 | |||
259 | /* | ||
260 | * should really be a BUG_ON but b/c this is an | ||
261 | * init call, it just doesn't work. -dcz | ||
262 | */ | ||
263 | if (nmi_watchdog != NMI_LOCAL_APIC) | ||
264 | return 0; | ||
265 | |||
266 | if (atomic_read(&nmi_active) < 0) | ||
267 | return 0; | ||
268 | |||
269 | error = sysdev_class_register(&nmi_sysclass); | ||
270 | if (!error) | ||
271 | error = sysdev_register(&device_lapic_nmi); | ||
272 | return error; | ||
273 | } | ||
274 | |||
275 | /* must come after the local APIC's device_initcall() */ | ||
276 | late_initcall(init_lapic_nmi_sysfs); | ||
277 | |||
278 | #endif /* CONFIG_PM */ | ||
279 | |||
280 | static void __acpi_nmi_enable(void *__unused) | ||
281 | { | ||
282 | apic_write(APIC_LVT0, APIC_DM_NMI); | ||
283 | } | ||
284 | |||
285 | /* | ||
286 | * Enable timer based NMIs on all CPUs: | ||
287 | */ | ||
288 | void acpi_nmi_enable(void) | ||
289 | { | ||
290 | if (atomic_read(&nmi_active) && nmi_watchdog == NMI_IO_APIC) | ||
291 | on_each_cpu(__acpi_nmi_enable, NULL, 1); | ||
292 | } | ||
293 | |||
294 | /* | ||
295 | * Disable timer based NMIs on all CPUs: | ||
296 | */ | ||
297 | void acpi_nmi_disable(void) | ||
298 | { | ||
299 | if (atomic_read(&nmi_active) && nmi_watchdog == NMI_IO_APIC) | ||
300 | on_each_cpu(__acpi_nmi_disable, NULL, 1); | ||
301 | } | ||
302 | |||
303 | /* | ||
304 | * This function is called as soon the LAPIC NMI watchdog driver has everything | ||
305 | * in place and it's ready to check if the NMIs belong to the NMI watchdog | ||
306 | */ | ||
307 | void cpu_nmi_set_wd_enabled(void) | ||
308 | { | ||
309 | __get_cpu_var(wd_enabled) = 1; | ||
310 | } | ||
311 | |||
312 | void setup_apic_nmi_watchdog(void *unused) | ||
313 | { | ||
314 | if (__get_cpu_var(wd_enabled)) | ||
315 | return; | ||
316 | |||
317 | /* cheap hack to support suspend/resume */ | ||
318 | /* if cpu0 is not active neither should the other cpus */ | ||
319 | if (smp_processor_id() != 0 && atomic_read(&nmi_active) <= 0) | ||
320 | return; | ||
321 | |||
322 | switch (nmi_watchdog) { | ||
323 | case NMI_LOCAL_APIC: | ||
324 | if (lapic_watchdog_init(nmi_hz) < 0) { | ||
325 | __get_cpu_var(wd_enabled) = 0; | ||
326 | return; | ||
327 | } | ||
328 | /* FALL THROUGH */ | ||
329 | case NMI_IO_APIC: | ||
330 | __get_cpu_var(wd_enabled) = 1; | ||
331 | atomic_inc(&nmi_active); | ||
332 | } | ||
333 | } | ||
334 | |||
335 | void stop_apic_nmi_watchdog(void *unused) | ||
336 | { | ||
337 | /* only support LOCAL and IO APICs for now */ | ||
338 | if (!nmi_watchdog_active()) | ||
339 | return; | ||
340 | if (__get_cpu_var(wd_enabled) == 0) | ||
341 | return; | ||
342 | if (nmi_watchdog == NMI_LOCAL_APIC) | ||
343 | lapic_watchdog_stop(); | ||
344 | else | ||
345 | __acpi_nmi_disable(NULL); | ||
346 | __get_cpu_var(wd_enabled) = 0; | ||
347 | atomic_dec(&nmi_active); | ||
348 | } | ||
349 | |||
350 | /* | ||
351 | * the best way to detect whether a CPU has a 'hard lockup' problem | ||
352 | * is to check it's local APIC timer IRQ counts. If they are not | ||
353 | * changing then that CPU has some problem. | ||
354 | * | ||
355 | * as these watchdog NMI IRQs are generated on every CPU, we only | ||
356 | * have to check the current processor. | ||
357 | * | ||
358 | * since NMIs don't listen to _any_ locks, we have to be extremely | ||
359 | * careful not to rely on unsafe variables. The printk might lock | ||
360 | * up though, so we have to break up any console locks first ... | ||
361 | * [when there will be more tty-related locks, break them up here too!] | ||
362 | */ | ||
363 | |||
364 | static DEFINE_PER_CPU(unsigned, last_irq_sum); | ||
365 | static DEFINE_PER_CPU(long, alert_counter); | ||
366 | static DEFINE_PER_CPU(int, nmi_touch); | ||
367 | |||
368 | void touch_nmi_watchdog(void) | ||
369 | { | ||
370 | if (nmi_watchdog_active()) { | ||
371 | unsigned cpu; | ||
372 | |||
373 | /* | ||
374 | * Tell other CPUs to reset their alert counters. We cannot | ||
375 | * do it ourselves because the alert count increase is not | ||
376 | * atomic. | ||
377 | */ | ||
378 | for_each_present_cpu(cpu) { | ||
379 | if (per_cpu(nmi_touch, cpu) != 1) | ||
380 | per_cpu(nmi_touch, cpu) = 1; | ||
381 | } | ||
382 | } | ||
383 | |||
384 | /* | ||
385 | * Tickle the softlockup detector too: | ||
386 | */ | ||
387 | touch_softlockup_watchdog(); | ||
388 | } | ||
389 | EXPORT_SYMBOL(touch_nmi_watchdog); | ||
390 | |||
391 | notrace __kprobes int | ||
392 | nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) | ||
393 | { | ||
394 | /* | ||
395 | * Since current_thread_info()-> is always on the stack, and we | ||
396 | * always switch the stack NMI-atomically, it's safe to use | ||
397 | * smp_processor_id(). | ||
398 | */ | ||
399 | unsigned int sum; | ||
400 | int touched = 0; | ||
401 | int cpu = smp_processor_id(); | ||
402 | int rc = 0; | ||
403 | |||
404 | sum = get_timer_irqs(cpu); | ||
405 | |||
406 | if (__get_cpu_var(nmi_touch)) { | ||
407 | __get_cpu_var(nmi_touch) = 0; | ||
408 | touched = 1; | ||
409 | } | ||
410 | |||
411 | /* We can be called before check_nmi_watchdog, hence NULL check. */ | ||
412 | if (cpumask_test_cpu(cpu, to_cpumask(backtrace_mask))) { | ||
413 | static DEFINE_RAW_SPINLOCK(lock); /* Serialise the printks */ | ||
414 | |||
415 | raw_spin_lock(&lock); | ||
416 | printk(KERN_WARNING "NMI backtrace for cpu %d\n", cpu); | ||
417 | show_regs(regs); | ||
418 | dump_stack(); | ||
419 | raw_spin_unlock(&lock); | ||
420 | cpumask_clear_cpu(cpu, to_cpumask(backtrace_mask)); | ||
421 | |||
422 | rc = 1; | ||
423 | } | ||
424 | |||
425 | /* Could check oops_in_progress here too, but it's safer not to */ | ||
426 | if (mce_in_progress()) | ||
427 | touched = 1; | ||
428 | |||
429 | /* if the none of the timers isn't firing, this cpu isn't doing much */ | ||
430 | if (!touched && __get_cpu_var(last_irq_sum) == sum) { | ||
431 | /* | ||
432 | * Ayiee, looks like this CPU is stuck ... | ||
433 | * wait a few IRQs (5 seconds) before doing the oops ... | ||
434 | */ | ||
435 | __this_cpu_inc(alert_counter); | ||
436 | if (__this_cpu_read(alert_counter) == 5 * nmi_hz) | ||
437 | /* | ||
438 | * die_nmi will return ONLY if NOTIFY_STOP happens.. | ||
439 | */ | ||
440 | die_nmi("BUG: NMI Watchdog detected LOCKUP", | ||
441 | regs, panic_on_timeout); | ||
442 | } else { | ||
443 | __get_cpu_var(last_irq_sum) = sum; | ||
444 | __this_cpu_write(alert_counter, 0); | ||
445 | } | ||
446 | |||
447 | /* see if the nmi watchdog went off */ | ||
448 | if (!__get_cpu_var(wd_enabled)) | ||
449 | return rc; | ||
450 | switch (nmi_watchdog) { | ||
451 | case NMI_LOCAL_APIC: | ||
452 | rc |= lapic_wd_event(nmi_hz); | ||
453 | break; | ||
454 | case NMI_IO_APIC: | ||
455 | /* | ||
456 | * don't know how to accurately check for this. | ||
457 | * just assume it was a watchdog timer interrupt | ||
458 | * This matches the old behaviour. | ||
459 | */ | ||
460 | rc = 1; | ||
461 | break; | ||
462 | } | ||
463 | return rc; | ||
464 | } | ||
465 | |||
466 | #ifdef CONFIG_SYSCTL | ||
467 | |||
468 | static void enable_ioapic_nmi_watchdog_single(void *unused) | ||
469 | { | ||
470 | __get_cpu_var(wd_enabled) = 1; | ||
471 | atomic_inc(&nmi_active); | ||
472 | __acpi_nmi_enable(NULL); | ||
473 | } | ||
474 | |||
475 | static void enable_ioapic_nmi_watchdog(void) | ||
476 | { | ||
477 | on_each_cpu(enable_ioapic_nmi_watchdog_single, NULL, 1); | ||
478 | touch_nmi_watchdog(); | ||
479 | } | ||
480 | |||
481 | static void disable_ioapic_nmi_watchdog(void) | ||
482 | { | ||
483 | on_each_cpu(stop_apic_nmi_watchdog, NULL, 1); | ||
484 | } | ||
485 | |||
486 | static int __init setup_unknown_nmi_panic(char *str) | ||
487 | { | ||
488 | unknown_nmi_panic = 1; | ||
489 | return 1; | ||
490 | } | ||
491 | __setup("unknown_nmi_panic", setup_unknown_nmi_panic); | ||
492 | |||
493 | static int unknown_nmi_panic_callback(struct pt_regs *regs, int cpu) | ||
494 | { | ||
495 | unsigned char reason = get_nmi_reason(); | ||
496 | char buf[64]; | ||
497 | |||
498 | sprintf(buf, "NMI received for unknown reason %02x\n", reason); | ||
499 | die_nmi(buf, regs, 1); /* Always panic here */ | ||
500 | return 0; | ||
501 | } | ||
502 | |||
503 | /* | ||
504 | * proc handler for /proc/sys/kernel/nmi | ||
505 | */ | ||
506 | int proc_nmi_enabled(struct ctl_table *table, int write, | ||
507 | void __user *buffer, size_t *length, loff_t *ppos) | ||
508 | { | ||
509 | int old_state; | ||
510 | |||
511 | nmi_watchdog_enabled = (atomic_read(&nmi_active) > 0) ? 1 : 0; | ||
512 | old_state = nmi_watchdog_enabled; | ||
513 | proc_dointvec(table, write, buffer, length, ppos); | ||
514 | if (!!old_state == !!nmi_watchdog_enabled) | ||
515 | return 0; | ||
516 | |||
517 | if (atomic_read(&nmi_active) < 0 || !nmi_watchdog_active()) { | ||
518 | printk(KERN_WARNING | ||
519 | "NMI watchdog is permanently disabled\n"); | ||
520 | return -EIO; | ||
521 | } | ||
522 | |||
523 | if (nmi_watchdog == NMI_LOCAL_APIC) { | ||
524 | if (nmi_watchdog_enabled) | ||
525 | enable_lapic_nmi_watchdog(); | ||
526 | else | ||
527 | disable_lapic_nmi_watchdog(); | ||
528 | } else if (nmi_watchdog == NMI_IO_APIC) { | ||
529 | if (nmi_watchdog_enabled) | ||
530 | enable_ioapic_nmi_watchdog(); | ||
531 | else | ||
532 | disable_ioapic_nmi_watchdog(); | ||
533 | } else { | ||
534 | printk(KERN_WARNING | ||
535 | "NMI watchdog doesn't know what hardware to touch\n"); | ||
536 | return -EIO; | ||
537 | } | ||
538 | return 0; | ||
539 | } | ||
540 | |||
541 | #endif /* CONFIG_SYSCTL */ | ||
542 | |||
543 | int do_nmi_callback(struct pt_regs *regs, int cpu) | ||
544 | { | ||
545 | #ifdef CONFIG_SYSCTL | ||
546 | if (unknown_nmi_panic) | ||
547 | return unknown_nmi_panic_callback(regs, cpu); | ||
548 | #endif | ||
549 | return 0; | ||
550 | } | ||
551 | |||
552 | void arch_trigger_all_cpu_backtrace(void) | ||
553 | { | ||
554 | int i; | ||
555 | |||
556 | cpumask_copy(to_cpumask(backtrace_mask), cpu_online_mask); | ||
557 | |||
558 | printk(KERN_INFO "sending NMI to all CPUs:\n"); | ||
559 | apic->send_IPI_all(NMI_VECTOR); | ||
560 | |||
561 | /* Wait for up to 10 seconds for all CPUs to do the backtrace */ | ||
562 | for (i = 0; i < 10 * 1000; i++) { | ||
563 | if (cpumask_empty(to_cpumask(backtrace_mask))) | ||
564 | break; | ||
565 | mdelay(1); | ||
566 | } | ||
567 | } | ||
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 4b68bda3093..1d59834396b 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c | |||
@@ -894,7 +894,6 @@ void __init identify_boot_cpu(void) | |||
894 | #else | 894 | #else |
895 | vgetcpu_set_mode(); | 895 | vgetcpu_set_mode(); |
896 | #endif | 896 | #endif |
897 | init_hw_perf_events(); | ||
898 | } | 897 | } |
899 | 898 | ||
900 | void __cpuinit identify_secondary_cpu(struct cpuinfo_x86 *c) | 899 | void __cpuinit identify_secondary_cpu(struct cpuinfo_x86 *c) |
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 6d75b9145b1..0a360d14659 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c | |||
@@ -330,9 +330,6 @@ static bool reserve_pmc_hardware(void) | |||
330 | { | 330 | { |
331 | int i; | 331 | int i; |
332 | 332 | ||
333 | if (nmi_watchdog == NMI_LOCAL_APIC) | ||
334 | disable_lapic_nmi_watchdog(); | ||
335 | |||
336 | for (i = 0; i < x86_pmu.num_counters; i++) { | 333 | for (i = 0; i < x86_pmu.num_counters; i++) { |
337 | if (!reserve_perfctr_nmi(x86_pmu.perfctr + i)) | 334 | if (!reserve_perfctr_nmi(x86_pmu.perfctr + i)) |
338 | goto perfctr_fail; | 335 | goto perfctr_fail; |
@@ -355,9 +352,6 @@ perfctr_fail: | |||
355 | for (i--; i >= 0; i--) | 352 | for (i--; i >= 0; i--) |
356 | release_perfctr_nmi(x86_pmu.perfctr + i); | 353 | release_perfctr_nmi(x86_pmu.perfctr + i); |
357 | 354 | ||
358 | if (nmi_watchdog == NMI_LOCAL_APIC) | ||
359 | enable_lapic_nmi_watchdog(); | ||
360 | |||
361 | return false; | 355 | return false; |
362 | } | 356 | } |
363 | 357 | ||
@@ -369,9 +363,6 @@ static void release_pmc_hardware(void) | |||
369 | release_perfctr_nmi(x86_pmu.perfctr + i); | 363 | release_perfctr_nmi(x86_pmu.perfctr + i); |
370 | release_evntsel_nmi(x86_pmu.eventsel + i); | 364 | release_evntsel_nmi(x86_pmu.eventsel + i); |
371 | } | 365 | } |
372 | |||
373 | if (nmi_watchdog == NMI_LOCAL_APIC) | ||
374 | enable_lapic_nmi_watchdog(); | ||
375 | } | 366 | } |
376 | 367 | ||
377 | #else | 368 | #else |
@@ -384,15 +375,53 @@ static void release_pmc_hardware(void) {} | |||
384 | static bool check_hw_exists(void) | 375 | static bool check_hw_exists(void) |
385 | { | 376 | { |
386 | u64 val, val_new = 0; | 377 | u64 val, val_new = 0; |
387 | int ret = 0; | 378 | int i, reg, ret = 0; |
379 | |||
380 | /* | ||
381 | * Check to see if the BIOS enabled any of the counters, if so | ||
382 | * complain and bail. | ||
383 | */ | ||
384 | for (i = 0; i < x86_pmu.num_counters; i++) { | ||
385 | reg = x86_pmu.eventsel + i; | ||
386 | ret = rdmsrl_safe(reg, &val); | ||
387 | if (ret) | ||
388 | goto msr_fail; | ||
389 | if (val & ARCH_PERFMON_EVENTSEL_ENABLE) | ||
390 | goto bios_fail; | ||
391 | } | ||
388 | 392 | ||
393 | if (x86_pmu.num_counters_fixed) { | ||
394 | reg = MSR_ARCH_PERFMON_FIXED_CTR_CTRL; | ||
395 | ret = rdmsrl_safe(reg, &val); | ||
396 | if (ret) | ||
397 | goto msr_fail; | ||
398 | for (i = 0; i < x86_pmu.num_counters_fixed; i++) { | ||
399 | if (val & (0x03 << i*4)) | ||
400 | goto bios_fail; | ||
401 | } | ||
402 | } | ||
403 | |||
404 | /* | ||
405 | * Now write a value and read it back to see if it matches, | ||
406 | * this is needed to detect certain hardware emulators (qemu/kvm) | ||
407 | * that don't trap on the MSR access and always return 0s. | ||
408 | */ | ||
389 | val = 0xabcdUL; | 409 | val = 0xabcdUL; |
390 | ret |= checking_wrmsrl(x86_pmu.perfctr, val); | 410 | ret = checking_wrmsrl(x86_pmu.perfctr, val); |
391 | ret |= rdmsrl_safe(x86_pmu.perfctr, &val_new); | 411 | ret |= rdmsrl_safe(x86_pmu.perfctr, &val_new); |
392 | if (ret || val != val_new) | 412 | if (ret || val != val_new) |
393 | return false; | 413 | goto msr_fail; |
394 | 414 | ||
395 | return true; | 415 | return true; |
416 | |||
417 | bios_fail: | ||
418 | printk(KERN_CONT "Broken BIOS detected, using software events only.\n"); | ||
419 | printk(KERN_ERR FW_BUG "the BIOS has corrupted hw-PMU resources (MSR %x is %Lx)\n", reg, val); | ||
420 | return false; | ||
421 | |||
422 | msr_fail: | ||
423 | printk(KERN_CONT "Broken PMU hardware detected, using software events only.\n"); | ||
424 | return false; | ||
396 | } | 425 | } |
397 | 426 | ||
398 | static void reserve_ds_buffers(void); | 427 | static void reserve_ds_buffers(void); |
@@ -451,7 +480,7 @@ static int x86_setup_perfctr(struct perf_event *event) | |||
451 | struct hw_perf_event *hwc = &event->hw; | 480 | struct hw_perf_event *hwc = &event->hw; |
452 | u64 config; | 481 | u64 config; |
453 | 482 | ||
454 | if (!hwc->sample_period) { | 483 | if (!is_sampling_event(event)) { |
455 | hwc->sample_period = x86_pmu.max_period; | 484 | hwc->sample_period = x86_pmu.max_period; |
456 | hwc->last_period = hwc->sample_period; | 485 | hwc->last_period = hwc->sample_period; |
457 | local64_set(&hwc->period_left, hwc->sample_period); | 486 | local64_set(&hwc->period_left, hwc->sample_period); |
@@ -1362,7 +1391,7 @@ static void __init pmu_check_apic(void) | |||
1362 | pr_info("no hardware sampling interrupt available.\n"); | 1391 | pr_info("no hardware sampling interrupt available.\n"); |
1363 | } | 1392 | } |
1364 | 1393 | ||
1365 | void __init init_hw_perf_events(void) | 1394 | int __init init_hw_perf_events(void) |
1366 | { | 1395 | { |
1367 | struct event_constraint *c; | 1396 | struct event_constraint *c; |
1368 | int err; | 1397 | int err; |
@@ -1377,20 +1406,18 @@ void __init init_hw_perf_events(void) | |||
1377 | err = amd_pmu_init(); | 1406 | err = amd_pmu_init(); |
1378 | break; | 1407 | break; |
1379 | default: | 1408 | default: |
1380 | return; | 1409 | return 0; |
1381 | } | 1410 | } |
1382 | if (err != 0) { | 1411 | if (err != 0) { |
1383 | pr_cont("no PMU driver, software events only.\n"); | 1412 | pr_cont("no PMU driver, software events only.\n"); |
1384 | return; | 1413 | return 0; |
1385 | } | 1414 | } |
1386 | 1415 | ||
1387 | pmu_check_apic(); | 1416 | pmu_check_apic(); |
1388 | 1417 | ||
1389 | /* sanity check that the hardware exists or is emulated */ | 1418 | /* sanity check that the hardware exists or is emulated */ |
1390 | if (!check_hw_exists()) { | 1419 | if (!check_hw_exists()) |
1391 | pr_cont("Broken PMU hardware detected, software events only.\n"); | 1420 | return 0; |
1392 | return; | ||
1393 | } | ||
1394 | 1421 | ||
1395 | pr_cont("%s PMU driver.\n", x86_pmu.name); | 1422 | pr_cont("%s PMU driver.\n", x86_pmu.name); |
1396 | 1423 | ||
@@ -1438,9 +1465,12 @@ void __init init_hw_perf_events(void) | |||
1438 | pr_info("... fixed-purpose events: %d\n", x86_pmu.num_counters_fixed); | 1465 | pr_info("... fixed-purpose events: %d\n", x86_pmu.num_counters_fixed); |
1439 | pr_info("... event mask: %016Lx\n", x86_pmu.intel_ctrl); | 1466 | pr_info("... event mask: %016Lx\n", x86_pmu.intel_ctrl); |
1440 | 1467 | ||
1441 | perf_pmu_register(&pmu); | 1468 | perf_pmu_register(&pmu, "cpu", PERF_TYPE_RAW); |
1442 | perf_cpu_notifier(x86_pmu_notifier); | 1469 | perf_cpu_notifier(x86_pmu_notifier); |
1470 | |||
1471 | return 0; | ||
1443 | } | 1472 | } |
1473 | early_initcall(init_hw_perf_events); | ||
1444 | 1474 | ||
1445 | static inline void x86_pmu_read(struct perf_event *event) | 1475 | static inline void x86_pmu_read(struct perf_event *event) |
1446 | { | 1476 | { |
@@ -1686,7 +1716,7 @@ perf_callchain_kernel(struct perf_callchain_entry *entry, struct pt_regs *regs) | |||
1686 | 1716 | ||
1687 | perf_callchain_store(entry, regs->ip); | 1717 | perf_callchain_store(entry, regs->ip); |
1688 | 1718 | ||
1689 | dump_trace(NULL, regs, NULL, regs->bp, &backtrace_ops, entry); | 1719 | dump_trace(NULL, regs, NULL, &backtrace_ops, entry); |
1690 | } | 1720 | } |
1691 | 1721 | ||
1692 | #ifdef CONFIG_COMPAT | 1722 | #ifdef CONFIG_COMPAT |
diff --git a/arch/x86/kernel/cpu/perf_event_amd.c b/arch/x86/kernel/cpu/perf_event_amd.c index e421b8cd694..67e2202a603 100644 --- a/arch/x86/kernel/cpu/perf_event_amd.c +++ b/arch/x86/kernel/cpu/perf_event_amd.c | |||
@@ -1,7 +1,5 @@ | |||
1 | #ifdef CONFIG_CPU_SUP_AMD | 1 | #ifdef CONFIG_CPU_SUP_AMD |
2 | 2 | ||
3 | static DEFINE_RAW_SPINLOCK(amd_nb_lock); | ||
4 | |||
5 | static __initconst const u64 amd_hw_cache_event_ids | 3 | static __initconst const u64 amd_hw_cache_event_ids |
6 | [PERF_COUNT_HW_CACHE_MAX] | 4 | [PERF_COUNT_HW_CACHE_MAX] |
7 | [PERF_COUNT_HW_CACHE_OP_MAX] | 5 | [PERF_COUNT_HW_CACHE_OP_MAX] |
@@ -275,7 +273,7 @@ done: | |||
275 | return &emptyconstraint; | 273 | return &emptyconstraint; |
276 | } | 274 | } |
277 | 275 | ||
278 | static struct amd_nb *amd_alloc_nb(int cpu, int nb_id) | 276 | static struct amd_nb *amd_alloc_nb(int cpu) |
279 | { | 277 | { |
280 | struct amd_nb *nb; | 278 | struct amd_nb *nb; |
281 | int i; | 279 | int i; |
@@ -285,7 +283,7 @@ static struct amd_nb *amd_alloc_nb(int cpu, int nb_id) | |||
285 | if (!nb) | 283 | if (!nb) |
286 | return NULL; | 284 | return NULL; |
287 | 285 | ||
288 | nb->nb_id = nb_id; | 286 | nb->nb_id = -1; |
289 | 287 | ||
290 | /* | 288 | /* |
291 | * initialize all possible NB constraints | 289 | * initialize all possible NB constraints |
@@ -306,7 +304,7 @@ static int amd_pmu_cpu_prepare(int cpu) | |||
306 | if (boot_cpu_data.x86_max_cores < 2) | 304 | if (boot_cpu_data.x86_max_cores < 2) |
307 | return NOTIFY_OK; | 305 | return NOTIFY_OK; |
308 | 306 | ||
309 | cpuc->amd_nb = amd_alloc_nb(cpu, -1); | 307 | cpuc->amd_nb = amd_alloc_nb(cpu); |
310 | if (!cpuc->amd_nb) | 308 | if (!cpuc->amd_nb) |
311 | return NOTIFY_BAD; | 309 | return NOTIFY_BAD; |
312 | 310 | ||
@@ -325,8 +323,6 @@ static void amd_pmu_cpu_starting(int cpu) | |||
325 | nb_id = amd_get_nb_id(cpu); | 323 | nb_id = amd_get_nb_id(cpu); |
326 | WARN_ON_ONCE(nb_id == BAD_APICID); | 324 | WARN_ON_ONCE(nb_id == BAD_APICID); |
327 | 325 | ||
328 | raw_spin_lock(&amd_nb_lock); | ||
329 | |||
330 | for_each_online_cpu(i) { | 326 | for_each_online_cpu(i) { |
331 | nb = per_cpu(cpu_hw_events, i).amd_nb; | 327 | nb = per_cpu(cpu_hw_events, i).amd_nb; |
332 | if (WARN_ON_ONCE(!nb)) | 328 | if (WARN_ON_ONCE(!nb)) |
@@ -341,8 +337,6 @@ static void amd_pmu_cpu_starting(int cpu) | |||
341 | 337 | ||
342 | cpuc->amd_nb->nb_id = nb_id; | 338 | cpuc->amd_nb->nb_id = nb_id; |
343 | cpuc->amd_nb->refcnt++; | 339 | cpuc->amd_nb->refcnt++; |
344 | |||
345 | raw_spin_unlock(&amd_nb_lock); | ||
346 | } | 340 | } |
347 | 341 | ||
348 | static void amd_pmu_cpu_dead(int cpu) | 342 | static void amd_pmu_cpu_dead(int cpu) |
@@ -354,8 +348,6 @@ static void amd_pmu_cpu_dead(int cpu) | |||
354 | 348 | ||
355 | cpuhw = &per_cpu(cpu_hw_events, cpu); | 349 | cpuhw = &per_cpu(cpu_hw_events, cpu); |
356 | 350 | ||
357 | raw_spin_lock(&amd_nb_lock); | ||
358 | |||
359 | if (cpuhw->amd_nb) { | 351 | if (cpuhw->amd_nb) { |
360 | struct amd_nb *nb = cpuhw->amd_nb; | 352 | struct amd_nb *nb = cpuhw->amd_nb; |
361 | 353 | ||
@@ -364,8 +356,6 @@ static void amd_pmu_cpu_dead(int cpu) | |||
364 | 356 | ||
365 | cpuhw->amd_nb = NULL; | 357 | cpuhw->amd_nb = NULL; |
366 | } | 358 | } |
367 | |||
368 | raw_spin_unlock(&amd_nb_lock); | ||
369 | } | 359 | } |
370 | 360 | ||
371 | static __initconst const struct x86_pmu amd_pmu = { | 361 | static __initconst const struct x86_pmu amd_pmu = { |
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index c8f5c088cad..24e390e40f2 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c | |||
@@ -816,6 +816,32 @@ static int intel_pmu_hw_config(struct perf_event *event) | |||
816 | if (ret) | 816 | if (ret) |
817 | return ret; | 817 | return ret; |
818 | 818 | ||
819 | if (event->attr.precise_ip && | ||
820 | (event->hw.config & X86_RAW_EVENT_MASK) == 0x003c) { | ||
821 | /* | ||
822 | * Use an alternative encoding for CPU_CLK_UNHALTED.THREAD_P | ||
823 | * (0x003c) so that we can use it with PEBS. | ||
824 | * | ||
825 | * The regular CPU_CLK_UNHALTED.THREAD_P event (0x003c) isn't | ||
826 | * PEBS capable. However we can use INST_RETIRED.ANY_P | ||
827 | * (0x00c0), which is a PEBS capable event, to get the same | ||
828 | * count. | ||
829 | * | ||
830 | * INST_RETIRED.ANY_P counts the number of cycles that retires | ||
831 | * CNTMASK instructions. By setting CNTMASK to a value (16) | ||
832 | * larger than the maximum number of instructions that can be | ||
833 | * retired per cycle (4) and then inverting the condition, we | ||
834 | * count all cycles that retire 16 or less instructions, which | ||
835 | * is every cycle. | ||
836 | * | ||
837 | * Thereby we gain a PEBS capable cycle counter. | ||
838 | */ | ||
839 | u64 alt_config = 0x108000c0; /* INST_RETIRED.TOTAL_CYCLES */ | ||
840 | |||
841 | alt_config |= (event->hw.config & ~X86_RAW_EVENT_MASK); | ||
842 | event->hw.config = alt_config; | ||
843 | } | ||
844 | |||
819 | if (event->attr.type != PERF_TYPE_RAW) | 845 | if (event->attr.type != PERF_TYPE_RAW) |
820 | return 0; | 846 | return 0; |
821 | 847 | ||
diff --git a/arch/x86/kernel/cpu/perfctr-watchdog.c b/arch/x86/kernel/cpu/perfctr-watchdog.c index d9f4ff8fcd6..d5a23661550 100644 --- a/arch/x86/kernel/cpu/perfctr-watchdog.c +++ b/arch/x86/kernel/cpu/perfctr-watchdog.c | |||
@@ -16,32 +16,12 @@ | |||
16 | #include <linux/kernel.h> | 16 | #include <linux/kernel.h> |
17 | #include <linux/bitops.h> | 17 | #include <linux/bitops.h> |
18 | #include <linux/smp.h> | 18 | #include <linux/smp.h> |
19 | #include <linux/nmi.h> | 19 | #include <asm/nmi.h> |
20 | #include <linux/kprobes.h> | 20 | #include <linux/kprobes.h> |
21 | 21 | ||
22 | #include <asm/apic.h> | 22 | #include <asm/apic.h> |
23 | #include <asm/perf_event.h> | 23 | #include <asm/perf_event.h> |
24 | 24 | ||
25 | struct nmi_watchdog_ctlblk { | ||
26 | unsigned int cccr_msr; | ||
27 | unsigned int perfctr_msr; /* the MSR to reset in NMI handler */ | ||
28 | unsigned int evntsel_msr; /* the MSR to select the events to handle */ | ||
29 | }; | ||
30 | |||
31 | /* Interface defining a CPU specific perfctr watchdog */ | ||
32 | struct wd_ops { | ||
33 | int (*reserve)(void); | ||
34 | void (*unreserve)(void); | ||
35 | int (*setup)(unsigned nmi_hz); | ||
36 | void (*rearm)(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz); | ||
37 | void (*stop)(void); | ||
38 | unsigned perfctr; | ||
39 | unsigned evntsel; | ||
40 | u64 checkbit; | ||
41 | }; | ||
42 | |||
43 | static const struct wd_ops *wd_ops; | ||
44 | |||
45 | /* | 25 | /* |
46 | * this number is calculated from Intel's MSR_P4_CRU_ESCR5 register and it's | 26 | * this number is calculated from Intel's MSR_P4_CRU_ESCR5 register and it's |
47 | * offset from MSR_P4_BSU_ESCR0. | 27 | * offset from MSR_P4_BSU_ESCR0. |
@@ -60,8 +40,6 @@ static const struct wd_ops *wd_ops; | |||
60 | static DECLARE_BITMAP(perfctr_nmi_owner, NMI_MAX_COUNTER_BITS); | 40 | static DECLARE_BITMAP(perfctr_nmi_owner, NMI_MAX_COUNTER_BITS); |
61 | static DECLARE_BITMAP(evntsel_nmi_owner, NMI_MAX_COUNTER_BITS); | 41 | static DECLARE_BITMAP(evntsel_nmi_owner, NMI_MAX_COUNTER_BITS); |
62 | 42 | ||
63 | static DEFINE_PER_CPU(struct nmi_watchdog_ctlblk, nmi_watchdog_ctlblk); | ||
64 | |||
65 | /* converts an msr to an appropriate reservation bit */ | 43 | /* converts an msr to an appropriate reservation bit */ |
66 | static inline unsigned int nmi_perfctr_msr_to_bit(unsigned int msr) | 44 | static inline unsigned int nmi_perfctr_msr_to_bit(unsigned int msr) |
67 | { | 45 | { |
@@ -172,623 +150,3 @@ void release_evntsel_nmi(unsigned int msr) | |||
172 | clear_bit(counter, evntsel_nmi_owner); | 150 | clear_bit(counter, evntsel_nmi_owner); |
173 | } | 151 | } |
174 | EXPORT_SYMBOL(release_evntsel_nmi); | 152 | EXPORT_SYMBOL(release_evntsel_nmi); |
175 | |||
176 | void disable_lapic_nmi_watchdog(void) | ||
177 | { | ||
178 | BUG_ON(nmi_watchdog != NMI_LOCAL_APIC); | ||
179 | |||
180 | if (atomic_read(&nmi_active) <= 0) | ||
181 | return; | ||
182 | |||
183 | on_each_cpu(stop_apic_nmi_watchdog, NULL, 1); | ||
184 | |||
185 | if (wd_ops) | ||
186 | wd_ops->unreserve(); | ||
187 | |||
188 | BUG_ON(atomic_read(&nmi_active) != 0); | ||
189 | } | ||
190 | |||
191 | void enable_lapic_nmi_watchdog(void) | ||
192 | { | ||
193 | BUG_ON(nmi_watchdog != NMI_LOCAL_APIC); | ||
194 | |||
195 | /* are we already enabled */ | ||
196 | if (atomic_read(&nmi_active) != 0) | ||
197 | return; | ||
198 | |||
199 | /* are we lapic aware */ | ||
200 | if (!wd_ops) | ||
201 | return; | ||
202 | if (!wd_ops->reserve()) { | ||
203 | printk(KERN_ERR "NMI watchdog: cannot reserve perfctrs\n"); | ||
204 | return; | ||
205 | } | ||
206 | |||
207 | on_each_cpu(setup_apic_nmi_watchdog, NULL, 1); | ||
208 | touch_nmi_watchdog(); | ||
209 | } | ||
210 | |||
211 | /* | ||
212 | * Activate the NMI watchdog via the local APIC. | ||
213 | */ | ||
214 | |||
215 | static unsigned int adjust_for_32bit_ctr(unsigned int hz) | ||
216 | { | ||
217 | u64 counter_val; | ||
218 | unsigned int retval = hz; | ||
219 | |||
220 | /* | ||
221 | * On Intel CPUs with P6/ARCH_PERFMON only 32 bits in the counter | ||
222 | * are writable, with higher bits sign extending from bit 31. | ||
223 | * So, we can only program the counter with 31 bit values and | ||
224 | * 32nd bit should be 1, for 33.. to be 1. | ||
225 | * Find the appropriate nmi_hz | ||
226 | */ | ||
227 | counter_val = (u64)cpu_khz * 1000; | ||
228 | do_div(counter_val, retval); | ||
229 | if (counter_val > 0x7fffffffULL) { | ||
230 | u64 count = (u64)cpu_khz * 1000; | ||
231 | do_div(count, 0x7fffffffUL); | ||
232 | retval = count + 1; | ||
233 | } | ||
234 | return retval; | ||
235 | } | ||
236 | |||
237 | static void write_watchdog_counter(unsigned int perfctr_msr, | ||
238 | const char *descr, unsigned nmi_hz) | ||
239 | { | ||
240 | u64 count = (u64)cpu_khz * 1000; | ||
241 | |||
242 | do_div(count, nmi_hz); | ||
243 | if (descr) | ||
244 | pr_debug("setting %s to -0x%08Lx\n", descr, count); | ||
245 | wrmsrl(perfctr_msr, 0 - count); | ||
246 | } | ||
247 | |||
248 | static void write_watchdog_counter32(unsigned int perfctr_msr, | ||
249 | const char *descr, unsigned nmi_hz) | ||
250 | { | ||
251 | u64 count = (u64)cpu_khz * 1000; | ||
252 | |||
253 | do_div(count, nmi_hz); | ||
254 | if (descr) | ||
255 | pr_debug("setting %s to -0x%08Lx\n", descr, count); | ||
256 | wrmsr(perfctr_msr, (u32)(-count), 0); | ||
257 | } | ||
258 | |||
259 | /* | ||
260 | * AMD K7/K8/Family10h/Family11h support. | ||
261 | * AMD keeps this interface nicely stable so there is not much variety | ||
262 | */ | ||
263 | #define K7_EVNTSEL_ENABLE (1 << 22) | ||
264 | #define K7_EVNTSEL_INT (1 << 20) | ||
265 | #define K7_EVNTSEL_OS (1 << 17) | ||
266 | #define K7_EVNTSEL_USR (1 << 16) | ||
267 | #define K7_EVENT_CYCLES_PROCESSOR_IS_RUNNING 0x76 | ||
268 | #define K7_NMI_EVENT K7_EVENT_CYCLES_PROCESSOR_IS_RUNNING | ||
269 | |||
270 | static int setup_k7_watchdog(unsigned nmi_hz) | ||
271 | { | ||
272 | unsigned int perfctr_msr, evntsel_msr; | ||
273 | unsigned int evntsel; | ||
274 | struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk); | ||
275 | |||
276 | perfctr_msr = wd_ops->perfctr; | ||
277 | evntsel_msr = wd_ops->evntsel; | ||
278 | |||
279 | wrmsrl(perfctr_msr, 0UL); | ||
280 | |||
281 | evntsel = K7_EVNTSEL_INT | ||
282 | | K7_EVNTSEL_OS | ||
283 | | K7_EVNTSEL_USR | ||
284 | | K7_NMI_EVENT; | ||
285 | |||
286 | /* setup the timer */ | ||
287 | wrmsr(evntsel_msr, evntsel, 0); | ||
288 | write_watchdog_counter(perfctr_msr, "K7_PERFCTR0", nmi_hz); | ||
289 | |||
290 | /* initialize the wd struct before enabling */ | ||
291 | wd->perfctr_msr = perfctr_msr; | ||
292 | wd->evntsel_msr = evntsel_msr; | ||
293 | wd->cccr_msr = 0; /* unused */ | ||
294 | |||
295 | /* ok, everything is initialized, announce that we're set */ | ||
296 | cpu_nmi_set_wd_enabled(); | ||
297 | |||
298 | apic_write(APIC_LVTPC, APIC_DM_NMI); | ||
299 | evntsel |= K7_EVNTSEL_ENABLE; | ||
300 | wrmsr(evntsel_msr, evntsel, 0); | ||
301 | |||
302 | return 1; | ||
303 | } | ||
304 | |||
305 | static void single_msr_stop_watchdog(void) | ||
306 | { | ||
307 | struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk); | ||
308 | |||
309 | wrmsr(wd->evntsel_msr, 0, 0); | ||
310 | } | ||
311 | |||
312 | static int single_msr_reserve(void) | ||
313 | { | ||
314 | if (!reserve_perfctr_nmi(wd_ops->perfctr)) | ||
315 | return 0; | ||
316 | |||
317 | if (!reserve_evntsel_nmi(wd_ops->evntsel)) { | ||
318 | release_perfctr_nmi(wd_ops->perfctr); | ||
319 | return 0; | ||
320 | } | ||
321 | return 1; | ||
322 | } | ||
323 | |||
324 | static void single_msr_unreserve(void) | ||
325 | { | ||
326 | release_evntsel_nmi(wd_ops->evntsel); | ||
327 | release_perfctr_nmi(wd_ops->perfctr); | ||
328 | } | ||
329 | |||
330 | static void __kprobes | ||
331 | single_msr_rearm(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz) | ||
332 | { | ||
333 | /* start the cycle over again */ | ||
334 | write_watchdog_counter(wd->perfctr_msr, NULL, nmi_hz); | ||
335 | } | ||
336 | |||
337 | static const struct wd_ops k7_wd_ops = { | ||
338 | .reserve = single_msr_reserve, | ||
339 | .unreserve = single_msr_unreserve, | ||
340 | .setup = setup_k7_watchdog, | ||
341 | .rearm = single_msr_rearm, | ||
342 | .stop = single_msr_stop_watchdog, | ||
343 | .perfctr = MSR_K7_PERFCTR0, | ||
344 | .evntsel = MSR_K7_EVNTSEL0, | ||
345 | .checkbit = 1ULL << 47, | ||
346 | }; | ||
347 | |||
348 | /* | ||
349 | * Intel Model 6 (PPro+,P2,P3,P-M,Core1) | ||
350 | */ | ||
351 | #define P6_EVNTSEL0_ENABLE (1 << 22) | ||
352 | #define P6_EVNTSEL_INT (1 << 20) | ||
353 | #define P6_EVNTSEL_OS (1 << 17) | ||
354 | #define P6_EVNTSEL_USR (1 << 16) | ||
355 | #define P6_EVENT_CPU_CLOCKS_NOT_HALTED 0x79 | ||
356 | #define P6_NMI_EVENT P6_EVENT_CPU_CLOCKS_NOT_HALTED | ||
357 | |||
358 | static int setup_p6_watchdog(unsigned nmi_hz) | ||
359 | { | ||
360 | unsigned int perfctr_msr, evntsel_msr; | ||
361 | unsigned int evntsel; | ||
362 | struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk); | ||
363 | |||
364 | perfctr_msr = wd_ops->perfctr; | ||
365 | evntsel_msr = wd_ops->evntsel; | ||
366 | |||
367 | /* KVM doesn't implement this MSR */ | ||
368 | if (wrmsr_safe(perfctr_msr, 0, 0) < 0) | ||
369 | return 0; | ||
370 | |||
371 | evntsel = P6_EVNTSEL_INT | ||
372 | | P6_EVNTSEL_OS | ||
373 | | P6_EVNTSEL_USR | ||
374 | | P6_NMI_EVENT; | ||
375 | |||
376 | /* setup the timer */ | ||
377 | wrmsr(evntsel_msr, evntsel, 0); | ||
378 | nmi_hz = adjust_for_32bit_ctr(nmi_hz); | ||
379 | write_watchdog_counter32(perfctr_msr, "P6_PERFCTR0", nmi_hz); | ||
380 | |||
381 | /* initialize the wd struct before enabling */ | ||
382 | wd->perfctr_msr = perfctr_msr; | ||
383 | wd->evntsel_msr = evntsel_msr; | ||
384 | wd->cccr_msr = 0; /* unused */ | ||
385 | |||
386 | /* ok, everything is initialized, announce that we're set */ | ||
387 | cpu_nmi_set_wd_enabled(); | ||
388 | |||
389 | apic_write(APIC_LVTPC, APIC_DM_NMI); | ||
390 | evntsel |= P6_EVNTSEL0_ENABLE; | ||
391 | wrmsr(evntsel_msr, evntsel, 0); | ||
392 | |||
393 | return 1; | ||
394 | } | ||
395 | |||
396 | static void __kprobes p6_rearm(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz) | ||
397 | { | ||
398 | /* | ||
399 | * P6 based Pentium M need to re-unmask | ||
400 | * the apic vector but it doesn't hurt | ||
401 | * other P6 variant. | ||
402 | * ArchPerfom/Core Duo also needs this | ||
403 | */ | ||
404 | apic_write(APIC_LVTPC, APIC_DM_NMI); | ||
405 | |||
406 | /* P6/ARCH_PERFMON has 32 bit counter write */ | ||
407 | write_watchdog_counter32(wd->perfctr_msr, NULL, nmi_hz); | ||
408 | } | ||
409 | |||
410 | static const struct wd_ops p6_wd_ops = { | ||
411 | .reserve = single_msr_reserve, | ||
412 | .unreserve = single_msr_unreserve, | ||
413 | .setup = setup_p6_watchdog, | ||
414 | .rearm = p6_rearm, | ||
415 | .stop = single_msr_stop_watchdog, | ||
416 | .perfctr = MSR_P6_PERFCTR0, | ||
417 | .evntsel = MSR_P6_EVNTSEL0, | ||
418 | .checkbit = 1ULL << 39, | ||
419 | }; | ||
420 | |||
421 | /* | ||
422 | * Intel P4 performance counters. | ||
423 | * By far the most complicated of all. | ||
424 | */ | ||
425 | #define MSR_P4_MISC_ENABLE_PERF_AVAIL (1 << 7) | ||
426 | #define P4_ESCR_EVENT_SELECT(N) ((N) << 25) | ||
427 | #define P4_ESCR_OS (1 << 3) | ||
428 | #define P4_ESCR_USR (1 << 2) | ||
429 | #define P4_CCCR_OVF_PMI0 (1 << 26) | ||
430 | #define P4_CCCR_OVF_PMI1 (1 << 27) | ||
431 | #define P4_CCCR_THRESHOLD(N) ((N) << 20) | ||
432 | #define P4_CCCR_COMPLEMENT (1 << 19) | ||
433 | #define P4_CCCR_COMPARE (1 << 18) | ||
434 | #define P4_CCCR_REQUIRED (3 << 16) | ||
435 | #define P4_CCCR_ESCR_SELECT(N) ((N) << 13) | ||
436 | #define P4_CCCR_ENABLE (1 << 12) | ||
437 | #define P4_CCCR_OVF (1 << 31) | ||
438 | |||
439 | #define P4_CONTROLS 18 | ||
440 | static unsigned int p4_controls[18] = { | ||
441 | MSR_P4_BPU_CCCR0, | ||
442 | MSR_P4_BPU_CCCR1, | ||
443 | MSR_P4_BPU_CCCR2, | ||
444 | MSR_P4_BPU_CCCR3, | ||
445 | MSR_P4_MS_CCCR0, | ||
446 | MSR_P4_MS_CCCR1, | ||
447 | MSR_P4_MS_CCCR2, | ||
448 | MSR_P4_MS_CCCR3, | ||
449 | MSR_P4_FLAME_CCCR0, | ||
450 | MSR_P4_FLAME_CCCR1, | ||
451 | MSR_P4_FLAME_CCCR2, | ||
452 | MSR_P4_FLAME_CCCR3, | ||
453 | MSR_P4_IQ_CCCR0, | ||
454 | MSR_P4_IQ_CCCR1, | ||
455 | MSR_P4_IQ_CCCR2, | ||
456 | MSR_P4_IQ_CCCR3, | ||
457 | MSR_P4_IQ_CCCR4, | ||
458 | MSR_P4_IQ_CCCR5, | ||
459 | }; | ||
460 | /* | ||
461 | * Set up IQ_COUNTER0 to behave like a clock, by having IQ_CCCR0 filter | ||
462 | * CRU_ESCR0 (with any non-null event selector) through a complemented | ||
463 | * max threshold. [IA32-Vol3, Section 14.9.9] | ||
464 | */ | ||
465 | static int setup_p4_watchdog(unsigned nmi_hz) | ||
466 | { | ||
467 | unsigned int perfctr_msr, evntsel_msr, cccr_msr; | ||
468 | unsigned int evntsel, cccr_val; | ||
469 | unsigned int misc_enable, dummy; | ||
470 | unsigned int ht_num; | ||
471 | struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk); | ||
472 | |||
473 | rdmsr(MSR_IA32_MISC_ENABLE, misc_enable, dummy); | ||
474 | if (!(misc_enable & MSR_P4_MISC_ENABLE_PERF_AVAIL)) | ||
475 | return 0; | ||
476 | |||
477 | #ifdef CONFIG_SMP | ||
478 | /* detect which hyperthread we are on */ | ||
479 | if (smp_num_siblings == 2) { | ||
480 | unsigned int ebx, apicid; | ||
481 | |||
482 | ebx = cpuid_ebx(1); | ||
483 | apicid = (ebx >> 24) & 0xff; | ||
484 | ht_num = apicid & 1; | ||
485 | } else | ||
486 | #endif | ||
487 | ht_num = 0; | ||
488 | |||
489 | /* | ||
490 | * performance counters are shared resources | ||
491 | * assign each hyperthread its own set | ||
492 | * (re-use the ESCR0 register, seems safe | ||
493 | * and keeps the cccr_val the same) | ||
494 | */ | ||
495 | if (!ht_num) { | ||
496 | /* logical cpu 0 */ | ||
497 | perfctr_msr = MSR_P4_IQ_PERFCTR0; | ||
498 | evntsel_msr = MSR_P4_CRU_ESCR0; | ||
499 | cccr_msr = MSR_P4_IQ_CCCR0; | ||
500 | cccr_val = P4_CCCR_OVF_PMI0 | P4_CCCR_ESCR_SELECT(4); | ||
501 | |||
502 | /* | ||
503 | * If we're on the kdump kernel or other situation, we may | ||
504 | * still have other performance counter registers set to | ||
505 | * interrupt and they'll keep interrupting forever because | ||
506 | * of the P4_CCCR_OVF quirk. So we need to ACK all the | ||
507 | * pending interrupts and disable all the registers here, | ||
508 | * before reenabling the NMI delivery. Refer to p4_rearm() | ||
509 | * about the P4_CCCR_OVF quirk. | ||
510 | */ | ||
511 | if (reset_devices) { | ||
512 | unsigned int low, high; | ||
513 | int i; | ||
514 | |||
515 | for (i = 0; i < P4_CONTROLS; i++) { | ||
516 | rdmsr(p4_controls[i], low, high); | ||
517 | low &= ~(P4_CCCR_ENABLE | P4_CCCR_OVF); | ||
518 | wrmsr(p4_controls[i], low, high); | ||
519 | } | ||
520 | } | ||
521 | } else { | ||
522 | /* logical cpu 1 */ | ||
523 | perfctr_msr = MSR_P4_IQ_PERFCTR1; | ||
524 | evntsel_msr = MSR_P4_CRU_ESCR0; | ||
525 | cccr_msr = MSR_P4_IQ_CCCR1; | ||
526 | |||
527 | /* Pentium 4 D processors don't support P4_CCCR_OVF_PMI1 */ | ||
528 | if (boot_cpu_data.x86_model == 4 && boot_cpu_data.x86_mask == 4) | ||
529 | cccr_val = P4_CCCR_OVF_PMI0; | ||
530 | else | ||
531 | cccr_val = P4_CCCR_OVF_PMI1; | ||
532 | cccr_val |= P4_CCCR_ESCR_SELECT(4); | ||
533 | } | ||
534 | |||
535 | evntsel = P4_ESCR_EVENT_SELECT(0x3F) | ||
536 | | P4_ESCR_OS | ||
537 | | P4_ESCR_USR; | ||
538 | |||
539 | cccr_val |= P4_CCCR_THRESHOLD(15) | ||
540 | | P4_CCCR_COMPLEMENT | ||
541 | | P4_CCCR_COMPARE | ||
542 | | P4_CCCR_REQUIRED; | ||
543 | |||
544 | wrmsr(evntsel_msr, evntsel, 0); | ||
545 | wrmsr(cccr_msr, cccr_val, 0); | ||
546 | write_watchdog_counter(perfctr_msr, "P4_IQ_COUNTER0", nmi_hz); | ||
547 | |||
548 | wd->perfctr_msr = perfctr_msr; | ||
549 | wd->evntsel_msr = evntsel_msr; | ||
550 | wd->cccr_msr = cccr_msr; | ||
551 | |||
552 | /* ok, everything is initialized, announce that we're set */ | ||
553 | cpu_nmi_set_wd_enabled(); | ||
554 | |||
555 | apic_write(APIC_LVTPC, APIC_DM_NMI); | ||
556 | cccr_val |= P4_CCCR_ENABLE; | ||
557 | wrmsr(cccr_msr, cccr_val, 0); | ||
558 | return 1; | ||
559 | } | ||
560 | |||
561 | static void stop_p4_watchdog(void) | ||
562 | { | ||
563 | struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk); | ||
564 | wrmsr(wd->cccr_msr, 0, 0); | ||
565 | wrmsr(wd->evntsel_msr, 0, 0); | ||
566 | } | ||
567 | |||
568 | static int p4_reserve(void) | ||
569 | { | ||
570 | if (!reserve_perfctr_nmi(MSR_P4_IQ_PERFCTR0)) | ||
571 | return 0; | ||
572 | #ifdef CONFIG_SMP | ||
573 | if (smp_num_siblings > 1 && !reserve_perfctr_nmi(MSR_P4_IQ_PERFCTR1)) | ||
574 | goto fail1; | ||
575 | #endif | ||
576 | if (!reserve_evntsel_nmi(MSR_P4_CRU_ESCR0)) | ||
577 | goto fail2; | ||
578 | /* RED-PEN why is ESCR1 not reserved here? */ | ||
579 | return 1; | ||
580 | fail2: | ||
581 | #ifdef CONFIG_SMP | ||
582 | if (smp_num_siblings > 1) | ||
583 | release_perfctr_nmi(MSR_P4_IQ_PERFCTR1); | ||
584 | fail1: | ||
585 | #endif | ||
586 | release_perfctr_nmi(MSR_P4_IQ_PERFCTR0); | ||
587 | return 0; | ||
588 | } | ||
589 | |||
590 | static void p4_unreserve(void) | ||
591 | { | ||
592 | #ifdef CONFIG_SMP | ||
593 | if (smp_num_siblings > 1) | ||
594 | release_perfctr_nmi(MSR_P4_IQ_PERFCTR1); | ||
595 | #endif | ||
596 | release_evntsel_nmi(MSR_P4_CRU_ESCR0); | ||
597 | release_perfctr_nmi(MSR_P4_IQ_PERFCTR0); | ||
598 | } | ||
599 | |||
600 | static void __kprobes p4_rearm(struct nmi_watchdog_ctlblk *wd, unsigned nmi_hz) | ||
601 | { | ||
602 | unsigned dummy; | ||
603 | /* | ||
604 | * P4 quirks: | ||
605 | * - An overflown perfctr will assert its interrupt | ||
606 | * until the OVF flag in its CCCR is cleared. | ||
607 | * - LVTPC is masked on interrupt and must be | ||
608 | * unmasked by the LVTPC handler. | ||
609 | */ | ||
610 | rdmsrl(wd->cccr_msr, dummy); | ||
611 | dummy &= ~P4_CCCR_OVF; | ||
612 | wrmsrl(wd->cccr_msr, dummy); | ||
613 | apic_write(APIC_LVTPC, APIC_DM_NMI); | ||
614 | /* start the cycle over again */ | ||
615 | write_watchdog_counter(wd->perfctr_msr, NULL, nmi_hz); | ||
616 | } | ||
617 | |||
618 | static const struct wd_ops p4_wd_ops = { | ||
619 | .reserve = p4_reserve, | ||
620 | .unreserve = p4_unreserve, | ||
621 | .setup = setup_p4_watchdog, | ||
622 | .rearm = p4_rearm, | ||
623 | .stop = stop_p4_watchdog, | ||
624 | /* RED-PEN this is wrong for the other sibling */ | ||
625 | .perfctr = MSR_P4_BPU_PERFCTR0, | ||
626 | .evntsel = MSR_P4_BSU_ESCR0, | ||
627 | .checkbit = 1ULL << 39, | ||
628 | }; | ||
629 | |||
630 | /* | ||
631 | * Watchdog using the Intel architected PerfMon. | ||
632 | * Used for Core2 and hopefully all future Intel CPUs. | ||
633 | */ | ||
634 | #define ARCH_PERFMON_NMI_EVENT_SEL ARCH_PERFMON_UNHALTED_CORE_CYCLES_SEL | ||
635 | #define ARCH_PERFMON_NMI_EVENT_UMASK ARCH_PERFMON_UNHALTED_CORE_CYCLES_UMASK | ||
636 | |||
637 | static struct wd_ops intel_arch_wd_ops; | ||
638 | |||
639 | static int setup_intel_arch_watchdog(unsigned nmi_hz) | ||
640 | { | ||
641 | unsigned int ebx; | ||
642 | union cpuid10_eax eax; | ||
643 | unsigned int unused; | ||
644 | unsigned int perfctr_msr, evntsel_msr; | ||
645 | unsigned int evntsel; | ||
646 | struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk); | ||
647 | |||
648 | /* | ||
649 | * Check whether the Architectural PerfMon supports | ||
650 | * Unhalted Core Cycles Event or not. | ||
651 | * NOTE: Corresponding bit = 0 in ebx indicates event present. | ||
652 | */ | ||
653 | cpuid(10, &(eax.full), &ebx, &unused, &unused); | ||
654 | if ((eax.split.mask_length < | ||
655 | (ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX+1)) || | ||
656 | (ebx & ARCH_PERFMON_UNHALTED_CORE_CYCLES_PRESENT)) | ||
657 | return 0; | ||
658 | |||
659 | perfctr_msr = wd_ops->perfctr; | ||
660 | evntsel_msr = wd_ops->evntsel; | ||
661 | |||
662 | wrmsrl(perfctr_msr, 0UL); | ||
663 | |||
664 | evntsel = ARCH_PERFMON_EVENTSEL_INT | ||
665 | | ARCH_PERFMON_EVENTSEL_OS | ||
666 | | ARCH_PERFMON_EVENTSEL_USR | ||
667 | | ARCH_PERFMON_NMI_EVENT_SEL | ||
668 | | ARCH_PERFMON_NMI_EVENT_UMASK; | ||
669 | |||
670 | /* setup the timer */ | ||
671 | wrmsr(evntsel_msr, evntsel, 0); | ||
672 | nmi_hz = adjust_for_32bit_ctr(nmi_hz); | ||
673 | write_watchdog_counter32(perfctr_msr, "INTEL_ARCH_PERFCTR0", nmi_hz); | ||
674 | |||
675 | wd->perfctr_msr = perfctr_msr; | ||
676 | wd->evntsel_msr = evntsel_msr; | ||
677 | wd->cccr_msr = 0; /* unused */ | ||
678 | |||
679 | /* ok, everything is initialized, announce that we're set */ | ||
680 | cpu_nmi_set_wd_enabled(); | ||
681 | |||
682 | apic_write(APIC_LVTPC, APIC_DM_NMI); | ||
683 | evntsel |= ARCH_PERFMON_EVENTSEL_ENABLE; | ||
684 | wrmsr(evntsel_msr, evntsel, 0); | ||
685 | intel_arch_wd_ops.checkbit = 1ULL << (eax.split.bit_width - 1); | ||
686 | return 1; | ||
687 | } | ||
688 | |||
689 | static struct wd_ops intel_arch_wd_ops __read_mostly = { | ||
690 | .reserve = single_msr_reserve, | ||
691 | .unreserve = single_msr_unreserve, | ||
692 | .setup = setup_intel_arch_watchdog, | ||
693 | .rearm = p6_rearm, | ||
694 | .stop = single_msr_stop_watchdog, | ||
695 | .perfctr = MSR_ARCH_PERFMON_PERFCTR1, | ||
696 | .evntsel = MSR_ARCH_PERFMON_EVENTSEL1, | ||
697 | }; | ||
698 | |||
699 | static void probe_nmi_watchdog(void) | ||
700 | { | ||
701 | switch (boot_cpu_data.x86_vendor) { | ||
702 | case X86_VENDOR_AMD: | ||
703 | if (boot_cpu_data.x86 == 6 || | ||
704 | (boot_cpu_data.x86 >= 0xf && boot_cpu_data.x86 <= 0x15)) | ||
705 | wd_ops = &k7_wd_ops; | ||
706 | return; | ||
707 | case X86_VENDOR_INTEL: | ||
708 | /* Work around where perfctr1 doesn't have a working enable | ||
709 | * bit as described in the following errata: | ||
710 | * AE49 Core Duo and Intel Core Solo 65 nm | ||
711 | * AN49 Intel Pentium Dual-Core | ||
712 | * AF49 Dual-Core Intel Xeon Processor LV | ||
713 | */ | ||
714 | if ((boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model == 14) || | ||
715 | ((boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model == 15 && | ||
716 | boot_cpu_data.x86_mask == 4))) { | ||
717 | intel_arch_wd_ops.perfctr = MSR_ARCH_PERFMON_PERFCTR0; | ||
718 | intel_arch_wd_ops.evntsel = MSR_ARCH_PERFMON_EVENTSEL0; | ||
719 | } | ||
720 | if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) { | ||
721 | wd_ops = &intel_arch_wd_ops; | ||
722 | break; | ||
723 | } | ||
724 | switch (boot_cpu_data.x86) { | ||
725 | case 6: | ||
726 | if (boot_cpu_data.x86_model > 13) | ||
727 | return; | ||
728 | |||
729 | wd_ops = &p6_wd_ops; | ||
730 | break; | ||
731 | case 15: | ||
732 | wd_ops = &p4_wd_ops; | ||
733 | break; | ||
734 | default: | ||
735 | return; | ||
736 | } | ||
737 | break; | ||
738 | } | ||
739 | } | ||
740 | |||
741 | /* Interface to nmi.c */ | ||
742 | |||
743 | int lapic_watchdog_init(unsigned nmi_hz) | ||
744 | { | ||
745 | if (!wd_ops) { | ||
746 | probe_nmi_watchdog(); | ||
747 | if (!wd_ops) { | ||
748 | printk(KERN_INFO "NMI watchdog: CPU not supported\n"); | ||
749 | return -1; | ||
750 | } | ||
751 | |||
752 | if (!wd_ops->reserve()) { | ||
753 | printk(KERN_ERR | ||
754 | "NMI watchdog: cannot reserve perfctrs\n"); | ||
755 | return -1; | ||
756 | } | ||
757 | } | ||
758 | |||
759 | if (!(wd_ops->setup(nmi_hz))) { | ||
760 | printk(KERN_ERR "Cannot setup NMI watchdog on CPU %d\n", | ||
761 | raw_smp_processor_id()); | ||
762 | return -1; | ||
763 | } | ||
764 | |||
765 | return 0; | ||
766 | } | ||
767 | |||
768 | void lapic_watchdog_stop(void) | ||
769 | { | ||
770 | if (wd_ops) | ||
771 | wd_ops->stop(); | ||
772 | } | ||
773 | |||
774 | unsigned lapic_adjust_nmi_hz(unsigned hz) | ||
775 | { | ||
776 | struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk); | ||
777 | if (wd->perfctr_msr == MSR_P6_PERFCTR0 || | ||
778 | wd->perfctr_msr == MSR_ARCH_PERFMON_PERFCTR1) | ||
779 | hz = adjust_for_32bit_ctr(hz); | ||
780 | return hz; | ||
781 | } | ||
782 | |||
783 | int __kprobes lapic_wd_event(unsigned nmi_hz) | ||
784 | { | ||
785 | struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk); | ||
786 | u64 ctr; | ||
787 | |||
788 | rdmsrl(wd->perfctr_msr, ctr); | ||
789 | if (ctr & wd_ops->checkbit) /* perfctr still running? */ | ||
790 | return 0; | ||
791 | |||
792 | wd_ops->rearm(wd, nmi_hz); | ||
793 | return 1; | ||
794 | } | ||
diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c index 6e8752c1bd5..8474c998cbd 100644 --- a/arch/x86/kernel/dumpstack.c +++ b/arch/x86/kernel/dumpstack.c | |||
@@ -175,21 +175,21 @@ static const struct stacktrace_ops print_trace_ops = { | |||
175 | 175 | ||
176 | void | 176 | void |
177 | show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, | 177 | show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, |
178 | unsigned long *stack, unsigned long bp, char *log_lvl) | 178 | unsigned long *stack, char *log_lvl) |
179 | { | 179 | { |
180 | printk("%sCall Trace:\n", log_lvl); | 180 | printk("%sCall Trace:\n", log_lvl); |
181 | dump_trace(task, regs, stack, bp, &print_trace_ops, log_lvl); | 181 | dump_trace(task, regs, stack, &print_trace_ops, log_lvl); |
182 | } | 182 | } |
183 | 183 | ||
184 | void show_trace(struct task_struct *task, struct pt_regs *regs, | 184 | void show_trace(struct task_struct *task, struct pt_regs *regs, |
185 | unsigned long *stack, unsigned long bp) | 185 | unsigned long *stack) |
186 | { | 186 | { |
187 | show_trace_log_lvl(task, regs, stack, bp, ""); | 187 | show_trace_log_lvl(task, regs, stack, ""); |
188 | } | 188 | } |
189 | 189 | ||
190 | void show_stack(struct task_struct *task, unsigned long *sp) | 190 | void show_stack(struct task_struct *task, unsigned long *sp) |
191 | { | 191 | { |
192 | show_stack_log_lvl(task, NULL, sp, 0, ""); | 192 | show_stack_log_lvl(task, NULL, sp, ""); |
193 | } | 193 | } |
194 | 194 | ||
195 | /* | 195 | /* |
@@ -210,7 +210,7 @@ void dump_stack(void) | |||
210 | init_utsname()->release, | 210 | init_utsname()->release, |
211 | (int)strcspn(init_utsname()->version, " "), | 211 | (int)strcspn(init_utsname()->version, " "), |
212 | init_utsname()->version); | 212 | init_utsname()->version); |
213 | show_trace(NULL, NULL, &stack, bp); | 213 | show_trace(NULL, NULL, &stack); |
214 | } | 214 | } |
215 | EXPORT_SYMBOL(dump_stack); | 215 | EXPORT_SYMBOL(dump_stack); |
216 | 216 | ||
diff --git a/arch/x86/kernel/dumpstack_32.c b/arch/x86/kernel/dumpstack_32.c index 1bc7f75a5bd..74cc1eda384 100644 --- a/arch/x86/kernel/dumpstack_32.c +++ b/arch/x86/kernel/dumpstack_32.c | |||
@@ -17,11 +17,12 @@ | |||
17 | #include <asm/stacktrace.h> | 17 | #include <asm/stacktrace.h> |
18 | 18 | ||
19 | 19 | ||
20 | void dump_trace(struct task_struct *task, struct pt_regs *regs, | 20 | void dump_trace(struct task_struct *task, |
21 | unsigned long *stack, unsigned long bp, | 21 | struct pt_regs *regs, unsigned long *stack, |
22 | const struct stacktrace_ops *ops, void *data) | 22 | const struct stacktrace_ops *ops, void *data) |
23 | { | 23 | { |
24 | int graph = 0; | 24 | int graph = 0; |
25 | unsigned long bp; | ||
25 | 26 | ||
26 | if (!task) | 27 | if (!task) |
27 | task = current; | 28 | task = current; |
@@ -34,18 +35,7 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs, | |||
34 | stack = (unsigned long *)task->thread.sp; | 35 | stack = (unsigned long *)task->thread.sp; |
35 | } | 36 | } |
36 | 37 | ||
37 | #ifdef CONFIG_FRAME_POINTER | 38 | bp = stack_frame(task, regs); |
38 | if (!bp) { | ||
39 | if (task == current) { | ||
40 | /* Grab bp right from our regs */ | ||
41 | get_bp(bp); | ||
42 | } else { | ||
43 | /* bp is the last reg pushed by switch_to */ | ||
44 | bp = *(unsigned long *) task->thread.sp; | ||
45 | } | ||
46 | } | ||
47 | #endif | ||
48 | |||
49 | for (;;) { | 39 | for (;;) { |
50 | struct thread_info *context; | 40 | struct thread_info *context; |
51 | 41 | ||
@@ -65,7 +55,7 @@ EXPORT_SYMBOL(dump_trace); | |||
65 | 55 | ||
66 | void | 56 | void |
67 | show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, | 57 | show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, |
68 | unsigned long *sp, unsigned long bp, char *log_lvl) | 58 | unsigned long *sp, char *log_lvl) |
69 | { | 59 | { |
70 | unsigned long *stack; | 60 | unsigned long *stack; |
71 | int i; | 61 | int i; |
@@ -87,7 +77,7 @@ show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, | |||
87 | touch_nmi_watchdog(); | 77 | touch_nmi_watchdog(); |
88 | } | 78 | } |
89 | printk(KERN_CONT "\n"); | 79 | printk(KERN_CONT "\n"); |
90 | show_trace_log_lvl(task, regs, sp, bp, log_lvl); | 80 | show_trace_log_lvl(task, regs, sp, log_lvl); |
91 | } | 81 | } |
92 | 82 | ||
93 | 83 | ||
@@ -112,8 +102,7 @@ void show_registers(struct pt_regs *regs) | |||
112 | u8 *ip; | 102 | u8 *ip; |
113 | 103 | ||
114 | printk(KERN_EMERG "Stack:\n"); | 104 | printk(KERN_EMERG "Stack:\n"); |
115 | show_stack_log_lvl(NULL, regs, ®s->sp, | 105 | show_stack_log_lvl(NULL, regs, ®s->sp, KERN_EMERG); |
116 | 0, KERN_EMERG); | ||
117 | 106 | ||
118 | printk(KERN_EMERG "Code: "); | 107 | printk(KERN_EMERG "Code: "); |
119 | 108 | ||
diff --git a/arch/x86/kernel/dumpstack_64.c b/arch/x86/kernel/dumpstack_64.c index 6a340485249..64101335de1 100644 --- a/arch/x86/kernel/dumpstack_64.c +++ b/arch/x86/kernel/dumpstack_64.c | |||
@@ -139,8 +139,8 @@ fixup_bp_irq_link(unsigned long bp, unsigned long *stack, | |||
139 | * severe exception (double fault, nmi, stack fault, debug, mce) hardware stack | 139 | * severe exception (double fault, nmi, stack fault, debug, mce) hardware stack |
140 | */ | 140 | */ |
141 | 141 | ||
142 | void dump_trace(struct task_struct *task, struct pt_regs *regs, | 142 | void dump_trace(struct task_struct *task, |
143 | unsigned long *stack, unsigned long bp, | 143 | struct pt_regs *regs, unsigned long *stack, |
144 | const struct stacktrace_ops *ops, void *data) | 144 | const struct stacktrace_ops *ops, void *data) |
145 | { | 145 | { |
146 | const unsigned cpu = get_cpu(); | 146 | const unsigned cpu = get_cpu(); |
@@ -149,6 +149,7 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs, | |||
149 | unsigned used = 0; | 149 | unsigned used = 0; |
150 | struct thread_info *tinfo; | 150 | struct thread_info *tinfo; |
151 | int graph = 0; | 151 | int graph = 0; |
152 | unsigned long bp; | ||
152 | 153 | ||
153 | if (!task) | 154 | if (!task) |
154 | task = current; | 155 | task = current; |
@@ -160,18 +161,7 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs, | |||
160 | stack = (unsigned long *)task->thread.sp; | 161 | stack = (unsigned long *)task->thread.sp; |
161 | } | 162 | } |
162 | 163 | ||
163 | #ifdef CONFIG_FRAME_POINTER | 164 | bp = stack_frame(task, regs); |
164 | if (!bp) { | ||
165 | if (task == current) { | ||
166 | /* Grab bp right from our regs */ | ||
167 | get_bp(bp); | ||
168 | } else { | ||
169 | /* bp is the last reg pushed by switch_to */ | ||
170 | bp = *(unsigned long *) task->thread.sp; | ||
171 | } | ||
172 | } | ||
173 | #endif | ||
174 | |||
175 | /* | 165 | /* |
176 | * Print function call entries in all stacks, starting at the | 166 | * Print function call entries in all stacks, starting at the |
177 | * current stack address. If the stacks consist of nested | 167 | * current stack address. If the stacks consist of nested |
@@ -235,7 +225,7 @@ EXPORT_SYMBOL(dump_trace); | |||
235 | 225 | ||
236 | void | 226 | void |
237 | show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, | 227 | show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, |
238 | unsigned long *sp, unsigned long bp, char *log_lvl) | 228 | unsigned long *sp, char *log_lvl) |
239 | { | 229 | { |
240 | unsigned long *irq_stack_end; | 230 | unsigned long *irq_stack_end; |
241 | unsigned long *irq_stack; | 231 | unsigned long *irq_stack; |
@@ -279,7 +269,7 @@ show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, | |||
279 | preempt_enable(); | 269 | preempt_enable(); |
280 | 270 | ||
281 | printk(KERN_CONT "\n"); | 271 | printk(KERN_CONT "\n"); |
282 | show_trace_log_lvl(task, regs, sp, bp, log_lvl); | 272 | show_trace_log_lvl(task, regs, sp, log_lvl); |
283 | } | 273 | } |
284 | 274 | ||
285 | void show_registers(struct pt_regs *regs) | 275 | void show_registers(struct pt_regs *regs) |
@@ -308,7 +298,7 @@ void show_registers(struct pt_regs *regs) | |||
308 | 298 | ||
309 | printk(KERN_EMERG "Stack:\n"); | 299 | printk(KERN_EMERG "Stack:\n"); |
310 | show_stack_log_lvl(NULL, regs, (unsigned long *)sp, | 300 | show_stack_log_lvl(NULL, regs, (unsigned long *)sp, |
311 | regs->bp, KERN_EMERG); | 301 | KERN_EMERG); |
312 | 302 | ||
313 | printk(KERN_EMERG "Code: "); | 303 | printk(KERN_EMERG "Code: "); |
314 | 304 | ||
diff --git a/arch/x86/kernel/kprobes.c b/arch/x86/kernel/kprobes.c index 1cbd54c0df9..5940282bd2f 100644 --- a/arch/x86/kernel/kprobes.c +++ b/arch/x86/kernel/kprobes.c | |||
@@ -1184,6 +1184,10 @@ static void __kprobes optimized_callback(struct optimized_kprobe *op, | |||
1184 | { | 1184 | { |
1185 | struct kprobe_ctlblk *kcb = get_kprobe_ctlblk(); | 1185 | struct kprobe_ctlblk *kcb = get_kprobe_ctlblk(); |
1186 | 1186 | ||
1187 | /* This is possible if op is under delayed unoptimizing */ | ||
1188 | if (kprobe_disabled(&op->kp)) | ||
1189 | return; | ||
1190 | |||
1187 | preempt_disable(); | 1191 | preempt_disable(); |
1188 | if (kprobe_running()) { | 1192 | if (kprobe_running()) { |
1189 | kprobes_inc_nmissed_count(&op->kp); | 1193 | kprobes_inc_nmissed_count(&op->kp); |
@@ -1401,10 +1405,16 @@ int __kprobes arch_prepare_optimized_kprobe(struct optimized_kprobe *op) | |||
1401 | return 0; | 1405 | return 0; |
1402 | } | 1406 | } |
1403 | 1407 | ||
1404 | /* Replace a breakpoint (int3) with a relative jump. */ | 1408 | #define MAX_OPTIMIZE_PROBES 256 |
1405 | int __kprobes arch_optimize_kprobe(struct optimized_kprobe *op) | 1409 | static struct text_poke_param *jump_poke_params; |
1410 | static struct jump_poke_buffer { | ||
1411 | u8 buf[RELATIVEJUMP_SIZE]; | ||
1412 | } *jump_poke_bufs; | ||
1413 | |||
1414 | static void __kprobes setup_optimize_kprobe(struct text_poke_param *tprm, | ||
1415 | u8 *insn_buf, | ||
1416 | struct optimized_kprobe *op) | ||
1406 | { | 1417 | { |
1407 | unsigned char jmp_code[RELATIVEJUMP_SIZE]; | ||
1408 | s32 rel = (s32)((long)op->optinsn.insn - | 1418 | s32 rel = (s32)((long)op->optinsn.insn - |
1409 | ((long)op->kp.addr + RELATIVEJUMP_SIZE)); | 1419 | ((long)op->kp.addr + RELATIVEJUMP_SIZE)); |
1410 | 1420 | ||
@@ -1412,16 +1422,79 @@ int __kprobes arch_optimize_kprobe(struct optimized_kprobe *op) | |||
1412 | memcpy(op->optinsn.copied_insn, op->kp.addr + INT3_SIZE, | 1422 | memcpy(op->optinsn.copied_insn, op->kp.addr + INT3_SIZE, |
1413 | RELATIVE_ADDR_SIZE); | 1423 | RELATIVE_ADDR_SIZE); |
1414 | 1424 | ||
1415 | jmp_code[0] = RELATIVEJUMP_OPCODE; | 1425 | insn_buf[0] = RELATIVEJUMP_OPCODE; |
1416 | *(s32 *)(&jmp_code[1]) = rel; | 1426 | *(s32 *)(&insn_buf[1]) = rel; |
1427 | |||
1428 | tprm->addr = op->kp.addr; | ||
1429 | tprm->opcode = insn_buf; | ||
1430 | tprm->len = RELATIVEJUMP_SIZE; | ||
1431 | } | ||
1432 | |||
1433 | /* | ||
1434 | * Replace breakpoints (int3) with relative jumps. | ||
1435 | * Caller must call with locking kprobe_mutex and text_mutex. | ||
1436 | */ | ||
1437 | void __kprobes arch_optimize_kprobes(struct list_head *oplist) | ||
1438 | { | ||
1439 | struct optimized_kprobe *op, *tmp; | ||
1440 | int c = 0; | ||
1441 | |||
1442 | list_for_each_entry_safe(op, tmp, oplist, list) { | ||
1443 | WARN_ON(kprobe_disabled(&op->kp)); | ||
1444 | /* Setup param */ | ||
1445 | setup_optimize_kprobe(&jump_poke_params[c], | ||
1446 | jump_poke_bufs[c].buf, op); | ||
1447 | list_del_init(&op->list); | ||
1448 | if (++c >= MAX_OPTIMIZE_PROBES) | ||
1449 | break; | ||
1450 | } | ||
1417 | 1451 | ||
1418 | /* | 1452 | /* |
1419 | * text_poke_smp doesn't support NMI/MCE code modifying. | 1453 | * text_poke_smp doesn't support NMI/MCE code modifying. |
1420 | * However, since kprobes itself also doesn't support NMI/MCE | 1454 | * However, since kprobes itself also doesn't support NMI/MCE |
1421 | * code probing, it's not a problem. | 1455 | * code probing, it's not a problem. |
1422 | */ | 1456 | */ |
1423 | text_poke_smp(op->kp.addr, jmp_code, RELATIVEJUMP_SIZE); | 1457 | text_poke_smp_batch(jump_poke_params, c); |
1424 | return 0; | 1458 | } |
1459 | |||
1460 | static void __kprobes setup_unoptimize_kprobe(struct text_poke_param *tprm, | ||
1461 | u8 *insn_buf, | ||
1462 | struct optimized_kprobe *op) | ||
1463 | { | ||
1464 | /* Set int3 to first byte for kprobes */ | ||
1465 | insn_buf[0] = BREAKPOINT_INSTRUCTION; | ||
1466 | memcpy(insn_buf + 1, op->optinsn.copied_insn, RELATIVE_ADDR_SIZE); | ||
1467 | |||
1468 | tprm->addr = op->kp.addr; | ||
1469 | tprm->opcode = insn_buf; | ||
1470 | tprm->len = RELATIVEJUMP_SIZE; | ||
1471 | } | ||
1472 | |||
1473 | /* | ||
1474 | * Recover original instructions and breakpoints from relative jumps. | ||
1475 | * Caller must call with locking kprobe_mutex. | ||
1476 | */ | ||
1477 | extern void arch_unoptimize_kprobes(struct list_head *oplist, | ||
1478 | struct list_head *done_list) | ||
1479 | { | ||
1480 | struct optimized_kprobe *op, *tmp; | ||
1481 | int c = 0; | ||
1482 | |||
1483 | list_for_each_entry_safe(op, tmp, oplist, list) { | ||
1484 | /* Setup param */ | ||
1485 | setup_unoptimize_kprobe(&jump_poke_params[c], | ||
1486 | jump_poke_bufs[c].buf, op); | ||
1487 | list_move(&op->list, done_list); | ||
1488 | if (++c >= MAX_OPTIMIZE_PROBES) | ||
1489 | break; | ||
1490 | } | ||
1491 | |||
1492 | /* | ||
1493 | * text_poke_smp doesn't support NMI/MCE code modifying. | ||
1494 | * However, since kprobes itself also doesn't support NMI/MCE | ||
1495 | * code probing, it's not a problem. | ||
1496 | */ | ||
1497 | text_poke_smp_batch(jump_poke_params, c); | ||
1425 | } | 1498 | } |
1426 | 1499 | ||
1427 | /* Replace a relative jump with a breakpoint (int3). */ | 1500 | /* Replace a relative jump with a breakpoint (int3). */ |
@@ -1453,11 +1526,35 @@ static int __kprobes setup_detour_execution(struct kprobe *p, | |||
1453 | } | 1526 | } |
1454 | return 0; | 1527 | return 0; |
1455 | } | 1528 | } |
1529 | |||
1530 | static int __kprobes init_poke_params(void) | ||
1531 | { | ||
1532 | /* Allocate code buffer and parameter array */ | ||
1533 | jump_poke_bufs = kmalloc(sizeof(struct jump_poke_buffer) * | ||
1534 | MAX_OPTIMIZE_PROBES, GFP_KERNEL); | ||
1535 | if (!jump_poke_bufs) | ||
1536 | return -ENOMEM; | ||
1537 | |||
1538 | jump_poke_params = kmalloc(sizeof(struct text_poke_param) * | ||
1539 | MAX_OPTIMIZE_PROBES, GFP_KERNEL); | ||
1540 | if (!jump_poke_params) { | ||
1541 | kfree(jump_poke_bufs); | ||
1542 | jump_poke_bufs = NULL; | ||
1543 | return -ENOMEM; | ||
1544 | } | ||
1545 | |||
1546 | return 0; | ||
1547 | } | ||
1548 | #else /* !CONFIG_OPTPROBES */ | ||
1549 | static int __kprobes init_poke_params(void) | ||
1550 | { | ||
1551 | return 0; | ||
1552 | } | ||
1456 | #endif | 1553 | #endif |
1457 | 1554 | ||
1458 | int __init arch_init_kprobes(void) | 1555 | int __init arch_init_kprobes(void) |
1459 | { | 1556 | { |
1460 | return 0; | 1557 | return init_poke_params(); |
1461 | } | 1558 | } |
1462 | 1559 | ||
1463 | int __kprobes arch_trampoline_kprobe(struct kprobe *p) | 1560 | int __kprobes arch_trampoline_kprobe(struct kprobe *p) |
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c index 57d1868a86a..c852041bfc3 100644 --- a/arch/x86/kernel/process.c +++ b/arch/x86/kernel/process.c | |||
@@ -91,8 +91,7 @@ void exit_thread(void) | |||
91 | void show_regs(struct pt_regs *regs) | 91 | void show_regs(struct pt_regs *regs) |
92 | { | 92 | { |
93 | show_registers(regs); | 93 | show_registers(regs); |
94 | show_trace(NULL, regs, (unsigned long *)kernel_stack_pointer(regs), | 94 | show_trace(NULL, regs, (unsigned long *)kernel_stack_pointer(regs)); |
95 | regs->bp); | ||
96 | } | 95 | } |
97 | 96 | ||
98 | void show_regs_common(void) | 97 | void show_regs_common(void) |
@@ -374,6 +373,7 @@ void default_idle(void) | |||
374 | { | 373 | { |
375 | if (hlt_use_halt()) { | 374 | if (hlt_use_halt()) { |
376 | trace_power_start(POWER_CSTATE, 1, smp_processor_id()); | 375 | trace_power_start(POWER_CSTATE, 1, smp_processor_id()); |
376 | trace_cpu_idle(1, smp_processor_id()); | ||
377 | current_thread_info()->status &= ~TS_POLLING; | 377 | current_thread_info()->status &= ~TS_POLLING; |
378 | /* | 378 | /* |
379 | * TS_POLLING-cleared state must be visible before we | 379 | * TS_POLLING-cleared state must be visible before we |
@@ -444,6 +444,7 @@ EXPORT_SYMBOL_GPL(cpu_idle_wait); | |||
444 | void mwait_idle_with_hints(unsigned long ax, unsigned long cx) | 444 | void mwait_idle_with_hints(unsigned long ax, unsigned long cx) |
445 | { | 445 | { |
446 | trace_power_start(POWER_CSTATE, (ax>>4)+1, smp_processor_id()); | 446 | trace_power_start(POWER_CSTATE, (ax>>4)+1, smp_processor_id()); |
447 | trace_cpu_idle((ax>>4)+1, smp_processor_id()); | ||
447 | if (!need_resched()) { | 448 | if (!need_resched()) { |
448 | if (cpu_has(¤t_cpu_data, X86_FEATURE_CLFLUSH_MONITOR)) | 449 | if (cpu_has(¤t_cpu_data, X86_FEATURE_CLFLUSH_MONITOR)) |
449 | clflush((void *)¤t_thread_info()->flags); | 450 | clflush((void *)¤t_thread_info()->flags); |
@@ -460,6 +461,7 @@ static void mwait_idle(void) | |||
460 | { | 461 | { |
461 | if (!need_resched()) { | 462 | if (!need_resched()) { |
462 | trace_power_start(POWER_CSTATE, 1, smp_processor_id()); | 463 | trace_power_start(POWER_CSTATE, 1, smp_processor_id()); |
464 | trace_cpu_idle(1, smp_processor_id()); | ||
463 | if (cpu_has(¤t_cpu_data, X86_FEATURE_CLFLUSH_MONITOR)) | 465 | if (cpu_has(¤t_cpu_data, X86_FEATURE_CLFLUSH_MONITOR)) |
464 | clflush((void *)¤t_thread_info()->flags); | 466 | clflush((void *)¤t_thread_info()->flags); |
465 | 467 | ||
@@ -481,10 +483,12 @@ static void mwait_idle(void) | |||
481 | static void poll_idle(void) | 483 | static void poll_idle(void) |
482 | { | 484 | { |
483 | trace_power_start(POWER_CSTATE, 0, smp_processor_id()); | 485 | trace_power_start(POWER_CSTATE, 0, smp_processor_id()); |
486 | trace_cpu_idle(0, smp_processor_id()); | ||
484 | local_irq_enable(); | 487 | local_irq_enable(); |
485 | while (!need_resched()) | 488 | while (!need_resched()) |
486 | cpu_relax(); | 489 | cpu_relax(); |
487 | trace_power_end(0); | 490 | trace_power_end(smp_processor_id()); |
491 | trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id()); | ||
488 | } | 492 | } |
489 | 493 | ||
490 | /* | 494 | /* |
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c index 96586c3cbbb..4b9befa0e34 100644 --- a/arch/x86/kernel/process_32.c +++ b/arch/x86/kernel/process_32.c | |||
@@ -113,8 +113,8 @@ void cpu_idle(void) | |||
113 | stop_critical_timings(); | 113 | stop_critical_timings(); |
114 | pm_idle(); | 114 | pm_idle(); |
115 | start_critical_timings(); | 115 | start_critical_timings(); |
116 | |||
117 | trace_power_end(smp_processor_id()); | 116 | trace_power_end(smp_processor_id()); |
117 | trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id()); | ||
118 | } | 118 | } |
119 | tick_nohz_restart_sched_tick(); | 119 | tick_nohz_restart_sched_tick(); |
120 | preempt_enable_no_resched(); | 120 | preempt_enable_no_resched(); |
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index b3d7a3a04f3..4c818a73839 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c | |||
@@ -142,6 +142,8 @@ void cpu_idle(void) | |||
142 | start_critical_timings(); | 142 | start_critical_timings(); |
143 | 143 | ||
144 | trace_power_end(smp_processor_id()); | 144 | trace_power_end(smp_processor_id()); |
145 | trace_cpu_idle(PWR_EVENT_EXIT, | ||
146 | smp_processor_id()); | ||
145 | 147 | ||
146 | /* In many cases the interrupt that ended idle | 148 | /* In many cases the interrupt that ended idle |
147 | has already called exit_idle. But some idle | 149 | has already called exit_idle. But some idle |
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 083e99d1b7d..68f61ac632e 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c | |||
@@ -281,6 +281,13 @@ static void __cpuinit smp_callin(void) | |||
281 | */ | 281 | */ |
282 | smp_store_cpu_info(cpuid); | 282 | smp_store_cpu_info(cpuid); |
283 | 283 | ||
284 | /* | ||
285 | * This must be done before setting cpu_online_mask | ||
286 | * or calling notify_cpu_starting. | ||
287 | */ | ||
288 | set_cpu_sibling_map(raw_smp_processor_id()); | ||
289 | wmb(); | ||
290 | |||
284 | notify_cpu_starting(cpuid); | 291 | notify_cpu_starting(cpuid); |
285 | 292 | ||
286 | /* | 293 | /* |
@@ -316,16 +323,6 @@ notrace static void __cpuinit start_secondary(void *unused) | |||
316 | */ | 323 | */ |
317 | check_tsc_sync_target(); | 324 | check_tsc_sync_target(); |
318 | 325 | ||
319 | if (nmi_watchdog == NMI_IO_APIC) { | ||
320 | legacy_pic->mask(0); | ||
321 | enable_NMI_through_LVT0(); | ||
322 | legacy_pic->unmask(0); | ||
323 | } | ||
324 | |||
325 | /* This must be done before setting cpu_online_mask */ | ||
326 | set_cpu_sibling_map(raw_smp_processor_id()); | ||
327 | wmb(); | ||
328 | |||
329 | /* | 326 | /* |
330 | * We need to hold call_lock, so there is no inconsistency | 327 | * We need to hold call_lock, so there is no inconsistency |
331 | * between the time smp_call_function() determines number of | 328 | * between the time smp_call_function() determines number of |
@@ -1061,8 +1058,6 @@ static int __init smp_sanity_check(unsigned max_cpus) | |||
1061 | printk(KERN_INFO "SMP mode deactivated.\n"); | 1058 | printk(KERN_INFO "SMP mode deactivated.\n"); |
1062 | smpboot_clear_io_apic(); | 1059 | smpboot_clear_io_apic(); |
1063 | 1060 | ||
1064 | localise_nmi_watchdog(); | ||
1065 | |||
1066 | connect_bsp_APIC(); | 1061 | connect_bsp_APIC(); |
1067 | setup_local_APIC(); | 1062 | setup_local_APIC(); |
1068 | end_local_APIC_setup(); | 1063 | end_local_APIC_setup(); |
@@ -1196,7 +1191,6 @@ void __init native_smp_cpus_done(unsigned int max_cpus) | |||
1196 | #ifdef CONFIG_X86_IO_APIC | 1191 | #ifdef CONFIG_X86_IO_APIC |
1197 | setup_ioapic_dest(); | 1192 | setup_ioapic_dest(); |
1198 | #endif | 1193 | #endif |
1199 | check_nmi_watchdog(); | ||
1200 | mtrr_aps_init(); | 1194 | mtrr_aps_init(); |
1201 | } | 1195 | } |
1202 | 1196 | ||
@@ -1341,8 +1335,6 @@ int native_cpu_disable(void) | |||
1341 | if (cpu == 0) | 1335 | if (cpu == 0) |
1342 | return -EBUSY; | 1336 | return -EBUSY; |
1343 | 1337 | ||
1344 | if (nmi_watchdog == NMI_LOCAL_APIC) | ||
1345 | stop_apic_nmi_watchdog(NULL); | ||
1346 | clear_local_APIC(); | 1338 | clear_local_APIC(); |
1347 | 1339 | ||
1348 | cpu_disable_common(); | 1340 | cpu_disable_common(); |
diff --git a/arch/x86/kernel/stacktrace.c b/arch/x86/kernel/stacktrace.c index b53c525368a..938c8e10a19 100644 --- a/arch/x86/kernel/stacktrace.c +++ b/arch/x86/kernel/stacktrace.c | |||
@@ -73,22 +73,22 @@ static const struct stacktrace_ops save_stack_ops_nosched = { | |||
73 | */ | 73 | */ |
74 | void save_stack_trace(struct stack_trace *trace) | 74 | void save_stack_trace(struct stack_trace *trace) |
75 | { | 75 | { |
76 | dump_trace(current, NULL, NULL, 0, &save_stack_ops, trace); | 76 | dump_trace(current, NULL, NULL, &save_stack_ops, trace); |
77 | if (trace->nr_entries < trace->max_entries) | 77 | if (trace->nr_entries < trace->max_entries) |
78 | trace->entries[trace->nr_entries++] = ULONG_MAX; | 78 | trace->entries[trace->nr_entries++] = ULONG_MAX; |
79 | } | 79 | } |
80 | EXPORT_SYMBOL_GPL(save_stack_trace); | 80 | EXPORT_SYMBOL_GPL(save_stack_trace); |
81 | 81 | ||
82 | void save_stack_trace_bp(struct stack_trace *trace, unsigned long bp) | 82 | void save_stack_trace_regs(struct stack_trace *trace, struct pt_regs *regs) |
83 | { | 83 | { |
84 | dump_trace(current, NULL, NULL, bp, &save_stack_ops, trace); | 84 | dump_trace(current, regs, NULL, &save_stack_ops, trace); |
85 | if (trace->nr_entries < trace->max_entries) | 85 | if (trace->nr_entries < trace->max_entries) |
86 | trace->entries[trace->nr_entries++] = ULONG_MAX; | 86 | trace->entries[trace->nr_entries++] = ULONG_MAX; |
87 | } | 87 | } |
88 | 88 | ||
89 | void save_stack_trace_tsk(struct task_struct *tsk, struct stack_trace *trace) | 89 | void save_stack_trace_tsk(struct task_struct *tsk, struct stack_trace *trace) |
90 | { | 90 | { |
91 | dump_trace(tsk, NULL, NULL, 0, &save_stack_ops_nosched, trace); | 91 | dump_trace(tsk, NULL, NULL, &save_stack_ops_nosched, trace); |
92 | if (trace->nr_entries < trace->max_entries) | 92 | if (trace->nr_entries < trace->max_entries) |
93 | trace->entries[trace->nr_entries++] = ULONG_MAX; | 93 | trace->entries[trace->nr_entries++] = ULONG_MAX; |
94 | } | 94 | } |
diff --git a/arch/x86/kernel/time.c b/arch/x86/kernel/time.c index fb5cc5e14cf..25a28a24593 100644 --- a/arch/x86/kernel/time.c +++ b/arch/x86/kernel/time.c | |||
@@ -22,10 +22,6 @@ | |||
22 | #include <asm/hpet.h> | 22 | #include <asm/hpet.h> |
23 | #include <asm/time.h> | 23 | #include <asm/time.h> |
24 | 24 | ||
25 | #if defined(CONFIG_X86_32) && defined(CONFIG_X86_IO_APIC) | ||
26 | int timer_ack; | ||
27 | #endif | ||
28 | |||
29 | #ifdef CONFIG_X86_64 | 25 | #ifdef CONFIG_X86_64 |
30 | volatile unsigned long __jiffies __section_jiffies = INITIAL_JIFFIES; | 26 | volatile unsigned long __jiffies __section_jiffies = INITIAL_JIFFIES; |
31 | #endif | 27 | #endif |
@@ -63,20 +59,6 @@ static irqreturn_t timer_interrupt(int irq, void *dev_id) | |||
63 | /* Keep nmi watchdog up to date */ | 59 | /* Keep nmi watchdog up to date */ |
64 | inc_irq_stat(irq0_irqs); | 60 | inc_irq_stat(irq0_irqs); |
65 | 61 | ||
66 | /* Optimized out for !IO_APIC and x86_64 */ | ||
67 | if (timer_ack) { | ||
68 | /* | ||
69 | * Subtle, when I/O APICs are used we have to ack timer IRQ | ||
70 | * manually to deassert NMI lines for the watchdog if run | ||
71 | * on an 82489DX-based system. | ||
72 | */ | ||
73 | raw_spin_lock(&i8259A_lock); | ||
74 | outb(0x0c, PIC_MASTER_OCW3); | ||
75 | /* Ack the IRQ; AEOI will end it automatically. */ | ||
76 | inb(PIC_MASTER_POLL); | ||
77 | raw_spin_unlock(&i8259A_lock); | ||
78 | } | ||
79 | |||
80 | global_clock_event->event_handler(global_clock_event); | 62 | global_clock_event->event_handler(global_clock_event); |
81 | 63 | ||
82 | /* MCA bus quirk: Acknowledge irq0 by setting bit 7 in port 0x61 */ | 64 | /* MCA bus quirk: Acknowledge irq0 by setting bit 7 in port 0x61 */ |
diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c index cb838ca42c9..c76aaca5694 100644 --- a/arch/x86/kernel/traps.c +++ b/arch/x86/kernel/traps.c | |||
@@ -83,6 +83,8 @@ EXPORT_SYMBOL_GPL(used_vectors); | |||
83 | 83 | ||
84 | static int ignore_nmis; | 84 | static int ignore_nmis; |
85 | 85 | ||
86 | int unknown_nmi_panic; | ||
87 | |||
86 | static inline void conditional_sti(struct pt_regs *regs) | 88 | static inline void conditional_sti(struct pt_regs *regs) |
87 | { | 89 | { |
88 | if (regs->flags & X86_EFLAGS_IF) | 90 | if (regs->flags & X86_EFLAGS_IF) |
@@ -300,6 +302,13 @@ gp_in_kernel: | |||
300 | die("general protection fault", regs, error_code); | 302 | die("general protection fault", regs, error_code); |
301 | } | 303 | } |
302 | 304 | ||
305 | static int __init setup_unknown_nmi_panic(char *str) | ||
306 | { | ||
307 | unknown_nmi_panic = 1; | ||
308 | return 1; | ||
309 | } | ||
310 | __setup("unknown_nmi_panic", setup_unknown_nmi_panic); | ||
311 | |||
303 | static notrace __kprobes void | 312 | static notrace __kprobes void |
304 | mem_parity_error(unsigned char reason, struct pt_regs *regs) | 313 | mem_parity_error(unsigned char reason, struct pt_regs *regs) |
305 | { | 314 | { |
@@ -342,9 +351,11 @@ io_check_error(unsigned char reason, struct pt_regs *regs) | |||
342 | reason = (reason & 0xf) | 8; | 351 | reason = (reason & 0xf) | 8; |
343 | outb(reason, 0x61); | 352 | outb(reason, 0x61); |
344 | 353 | ||
345 | i = 2000; | 354 | i = 20000; |
346 | while (--i) | 355 | while (--i) { |
347 | udelay(1000); | 356 | touch_nmi_watchdog(); |
357 | udelay(100); | ||
358 | } | ||
348 | 359 | ||
349 | reason &= ~8; | 360 | reason &= ~8; |
350 | outb(reason, 0x61); | 361 | outb(reason, 0x61); |
@@ -371,7 +382,7 @@ unknown_nmi_error(unsigned char reason, struct pt_regs *regs) | |||
371 | reason, smp_processor_id()); | 382 | reason, smp_processor_id()); |
372 | 383 | ||
373 | printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n"); | 384 | printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n"); |
374 | if (panic_on_unrecovered_nmi) | 385 | if (unknown_nmi_panic || panic_on_unrecovered_nmi) |
375 | panic("NMI: Not continuing"); | 386 | panic("NMI: Not continuing"); |
376 | 387 | ||
377 | printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); | 388 | printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); |
@@ -397,20 +408,8 @@ static notrace __kprobes void default_do_nmi(struct pt_regs *regs) | |||
397 | if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) | 408 | if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) |
398 | == NOTIFY_STOP) | 409 | == NOTIFY_STOP) |
399 | return; | 410 | return; |
400 | |||
401 | #ifndef CONFIG_LOCKUP_DETECTOR | ||
402 | /* | ||
403 | * Ok, so this is none of the documented NMI sources, | ||
404 | * so it must be the NMI watchdog. | ||
405 | */ | ||
406 | if (nmi_watchdog_tick(regs, reason)) | ||
407 | return; | ||
408 | if (!do_nmi_callback(regs, cpu)) | ||
409 | #endif /* !CONFIG_LOCKUP_DETECTOR */ | ||
410 | unknown_nmi_error(reason, regs); | ||
411 | #else | ||
412 | unknown_nmi_error(reason, regs); | ||
413 | #endif | 411 | #endif |
412 | unknown_nmi_error(reason, regs); | ||
414 | 413 | ||
415 | return; | 414 | return; |
416 | } | 415 | } |
@@ -446,14 +445,12 @@ do_nmi(struct pt_regs *regs, long error_code) | |||
446 | 445 | ||
447 | void stop_nmi(void) | 446 | void stop_nmi(void) |
448 | { | 447 | { |
449 | acpi_nmi_disable(); | ||
450 | ignore_nmis++; | 448 | ignore_nmis++; |
451 | } | 449 | } |
452 | 450 | ||
453 | void restart_nmi(void) | 451 | void restart_nmi(void) |
454 | { | 452 | { |
455 | ignore_nmis--; | 453 | ignore_nmis--; |
456 | acpi_nmi_enable(); | ||
457 | } | 454 | } |
458 | 455 | ||
459 | /* May run on IST stack. */ | 456 | /* May run on IST stack. */ |
diff --git a/arch/x86/mm/kmemcheck/error.c b/arch/x86/mm/kmemcheck/error.c index af3b6c8a436..704a37ceddd 100644 --- a/arch/x86/mm/kmemcheck/error.c +++ b/arch/x86/mm/kmemcheck/error.c | |||
@@ -185,7 +185,7 @@ void kmemcheck_error_save(enum kmemcheck_shadow state, | |||
185 | e->trace.entries = e->trace_entries; | 185 | e->trace.entries = e->trace_entries; |
186 | e->trace.max_entries = ARRAY_SIZE(e->trace_entries); | 186 | e->trace.max_entries = ARRAY_SIZE(e->trace_entries); |
187 | e->trace.skip = 0; | 187 | e->trace.skip = 0; |
188 | save_stack_trace_bp(&e->trace, regs->bp); | 188 | save_stack_trace_regs(&e->trace, regs); |
189 | 189 | ||
190 | /* Round address down to nearest 16 bytes */ | 190 | /* Round address down to nearest 16 bytes */ |
191 | shadow_copy = kmemcheck_shadow_lookup(address | 191 | shadow_copy = kmemcheck_shadow_lookup(address |
diff --git a/arch/x86/oprofile/backtrace.c b/arch/x86/oprofile/backtrace.c index 2d49d4e19a3..72cbec14d78 100644 --- a/arch/x86/oprofile/backtrace.c +++ b/arch/x86/oprofile/backtrace.c | |||
@@ -126,7 +126,7 @@ x86_backtrace(struct pt_regs * const regs, unsigned int depth) | |||
126 | if (!user_mode_vm(regs)) { | 126 | if (!user_mode_vm(regs)) { |
127 | unsigned long stack = kernel_stack_pointer(regs); | 127 | unsigned long stack = kernel_stack_pointer(regs); |
128 | if (depth) | 128 | if (depth) |
129 | dump_trace(NULL, regs, (unsigned long *)stack, 0, | 129 | dump_trace(NULL, regs, (unsigned long *)stack, |
130 | &backtrace_ops, &depth); | 130 | &backtrace_ops, &depth); |
131 | return; | 131 | return; |
132 | } | 132 | } |
diff --git a/arch/x86/oprofile/nmi_int.c b/arch/x86/oprofile/nmi_int.c index 4e8baad36d3..358c8b9c96a 100644 --- a/arch/x86/oprofile/nmi_int.c +++ b/arch/x86/oprofile/nmi_int.c | |||
@@ -732,6 +732,9 @@ int __init op_nmi_init(struct oprofile_operations *ops) | |||
732 | case 0x14: | 732 | case 0x14: |
733 | cpu_type = "x86-64/family14h"; | 733 | cpu_type = "x86-64/family14h"; |
734 | break; | 734 | break; |
735 | case 0x15: | ||
736 | cpu_type = "x86-64/family15h"; | ||
737 | break; | ||
735 | default: | 738 | default: |
736 | return -ENODEV; | 739 | return -ENODEV; |
737 | } | 740 | } |
diff --git a/arch/x86/oprofile/nmi_timer_int.c b/arch/x86/oprofile/nmi_timer_int.c index e3ecb71b579..0636dd93cef 100644 --- a/arch/x86/oprofile/nmi_timer_int.c +++ b/arch/x86/oprofile/nmi_timer_int.c | |||
@@ -58,9 +58,6 @@ static void timer_stop(void) | |||
58 | 58 | ||
59 | int __init op_nmi_timer_init(struct oprofile_operations *ops) | 59 | int __init op_nmi_timer_init(struct oprofile_operations *ops) |
60 | { | 60 | { |
61 | if ((nmi_watchdog != NMI_IO_APIC) || (atomic_read(&nmi_active) <= 0)) | ||
62 | return -ENODEV; | ||
63 | |||
64 | ops->start = timer_start; | 61 | ops->start = timer_start; |
65 | ops->stop = timer_stop; | 62 | ops->stop = timer_stop; |
66 | ops->cpu_type = "timer"; | 63 | ops->cpu_type = "timer"; |
diff --git a/arch/x86/oprofile/op_model_amd.c b/arch/x86/oprofile/op_model_amd.c index 7d90d47655b..51104b33fd5 100644 --- a/arch/x86/oprofile/op_model_amd.c +++ b/arch/x86/oprofile/op_model_amd.c | |||
@@ -29,11 +29,12 @@ | |||
29 | #include "op_x86_model.h" | 29 | #include "op_x86_model.h" |
30 | #include "op_counter.h" | 30 | #include "op_counter.h" |
31 | 31 | ||
32 | #define NUM_COUNTERS 4 | 32 | #define NUM_COUNTERS 4 |
33 | #define NUM_COUNTERS_F15H 6 | ||
33 | #ifdef CONFIG_OPROFILE_EVENT_MULTIPLEX | 34 | #ifdef CONFIG_OPROFILE_EVENT_MULTIPLEX |
34 | #define NUM_VIRT_COUNTERS 32 | 35 | #define NUM_VIRT_COUNTERS 32 |
35 | #else | 36 | #else |
36 | #define NUM_VIRT_COUNTERS NUM_COUNTERS | 37 | #define NUM_VIRT_COUNTERS 0 |
37 | #endif | 38 | #endif |
38 | 39 | ||
39 | #define OP_EVENT_MASK 0x0FFF | 40 | #define OP_EVENT_MASK 0x0FFF |
@@ -41,7 +42,8 @@ | |||
41 | 42 | ||
42 | #define MSR_AMD_EVENTSEL_RESERVED ((0xFFFFFCF0ULL<<32)|(1ULL<<21)) | 43 | #define MSR_AMD_EVENTSEL_RESERVED ((0xFFFFFCF0ULL<<32)|(1ULL<<21)) |
43 | 44 | ||
44 | static unsigned long reset_value[NUM_VIRT_COUNTERS]; | 45 | static int num_counters; |
46 | static unsigned long reset_value[OP_MAX_COUNTER]; | ||
45 | 47 | ||
46 | #define IBS_FETCH_SIZE 6 | 48 | #define IBS_FETCH_SIZE 6 |
47 | #define IBS_OP_SIZE 12 | 49 | #define IBS_OP_SIZE 12 |
@@ -387,7 +389,7 @@ static void op_mux_switch_ctrl(struct op_x86_model_spec const *model, | |||
387 | int i; | 389 | int i; |
388 | 390 | ||
389 | /* enable active counters */ | 391 | /* enable active counters */ |
390 | for (i = 0; i < NUM_COUNTERS; ++i) { | 392 | for (i = 0; i < num_counters; ++i) { |
391 | int virt = op_x86_phys_to_virt(i); | 393 | int virt = op_x86_phys_to_virt(i); |
392 | if (!reset_value[virt]) | 394 | if (!reset_value[virt]) |
393 | continue; | 395 | continue; |
@@ -406,7 +408,7 @@ static void op_amd_shutdown(struct op_msrs const * const msrs) | |||
406 | { | 408 | { |
407 | int i; | 409 | int i; |
408 | 410 | ||
409 | for (i = 0; i < NUM_COUNTERS; ++i) { | 411 | for (i = 0; i < num_counters; ++i) { |
410 | if (!msrs->counters[i].addr) | 412 | if (!msrs->counters[i].addr) |
411 | continue; | 413 | continue; |
412 | release_perfctr_nmi(MSR_K7_PERFCTR0 + i); | 414 | release_perfctr_nmi(MSR_K7_PERFCTR0 + i); |
@@ -418,7 +420,7 @@ static int op_amd_fill_in_addresses(struct op_msrs * const msrs) | |||
418 | { | 420 | { |
419 | int i; | 421 | int i; |
420 | 422 | ||
421 | for (i = 0; i < NUM_COUNTERS; i++) { | 423 | for (i = 0; i < num_counters; i++) { |
422 | if (!reserve_perfctr_nmi(MSR_K7_PERFCTR0 + i)) | 424 | if (!reserve_perfctr_nmi(MSR_K7_PERFCTR0 + i)) |
423 | goto fail; | 425 | goto fail; |
424 | if (!reserve_evntsel_nmi(MSR_K7_EVNTSEL0 + i)) { | 426 | if (!reserve_evntsel_nmi(MSR_K7_EVNTSEL0 + i)) { |
@@ -426,8 +428,13 @@ static int op_amd_fill_in_addresses(struct op_msrs * const msrs) | |||
426 | goto fail; | 428 | goto fail; |
427 | } | 429 | } |
428 | /* both registers must be reserved */ | 430 | /* both registers must be reserved */ |
429 | msrs->counters[i].addr = MSR_K7_PERFCTR0 + i; | 431 | if (num_counters == NUM_COUNTERS_F15H) { |
430 | msrs->controls[i].addr = MSR_K7_EVNTSEL0 + i; | 432 | msrs->counters[i].addr = MSR_F15H_PERF_CTR + (i << 1); |
433 | msrs->controls[i].addr = MSR_F15H_PERF_CTL + (i << 1); | ||
434 | } else { | ||
435 | msrs->controls[i].addr = MSR_K7_EVNTSEL0 + i; | ||
436 | msrs->counters[i].addr = MSR_K7_PERFCTR0 + i; | ||
437 | } | ||
431 | continue; | 438 | continue; |
432 | fail: | 439 | fail: |
433 | if (!counter_config[i].enabled) | 440 | if (!counter_config[i].enabled) |
@@ -447,7 +454,7 @@ static void op_amd_setup_ctrs(struct op_x86_model_spec const *model, | |||
447 | int i; | 454 | int i; |
448 | 455 | ||
449 | /* setup reset_value */ | 456 | /* setup reset_value */ |
450 | for (i = 0; i < NUM_VIRT_COUNTERS; ++i) { | 457 | for (i = 0; i < OP_MAX_COUNTER; ++i) { |
451 | if (counter_config[i].enabled | 458 | if (counter_config[i].enabled |
452 | && msrs->counters[op_x86_virt_to_phys(i)].addr) | 459 | && msrs->counters[op_x86_virt_to_phys(i)].addr) |
453 | reset_value[i] = counter_config[i].count; | 460 | reset_value[i] = counter_config[i].count; |
@@ -456,7 +463,7 @@ static void op_amd_setup_ctrs(struct op_x86_model_spec const *model, | |||
456 | } | 463 | } |
457 | 464 | ||
458 | /* clear all counters */ | 465 | /* clear all counters */ |
459 | for (i = 0; i < NUM_COUNTERS; ++i) { | 466 | for (i = 0; i < num_counters; ++i) { |
460 | if (!msrs->controls[i].addr) | 467 | if (!msrs->controls[i].addr) |
461 | continue; | 468 | continue; |
462 | rdmsrl(msrs->controls[i].addr, val); | 469 | rdmsrl(msrs->controls[i].addr, val); |
@@ -472,7 +479,7 @@ static void op_amd_setup_ctrs(struct op_x86_model_spec const *model, | |||
472 | } | 479 | } |
473 | 480 | ||
474 | /* enable active counters */ | 481 | /* enable active counters */ |
475 | for (i = 0; i < NUM_COUNTERS; ++i) { | 482 | for (i = 0; i < num_counters; ++i) { |
476 | int virt = op_x86_phys_to_virt(i); | 483 | int virt = op_x86_phys_to_virt(i); |
477 | if (!reset_value[virt]) | 484 | if (!reset_value[virt]) |
478 | continue; | 485 | continue; |
@@ -503,7 +510,7 @@ static int op_amd_check_ctrs(struct pt_regs * const regs, | |||
503 | u64 val; | 510 | u64 val; |
504 | int i; | 511 | int i; |
505 | 512 | ||
506 | for (i = 0; i < NUM_COUNTERS; ++i) { | 513 | for (i = 0; i < num_counters; ++i) { |
507 | int virt = op_x86_phys_to_virt(i); | 514 | int virt = op_x86_phys_to_virt(i); |
508 | if (!reset_value[virt]) | 515 | if (!reset_value[virt]) |
509 | continue; | 516 | continue; |
@@ -526,7 +533,7 @@ static void op_amd_start(struct op_msrs const * const msrs) | |||
526 | u64 val; | 533 | u64 val; |
527 | int i; | 534 | int i; |
528 | 535 | ||
529 | for (i = 0; i < NUM_COUNTERS; ++i) { | 536 | for (i = 0; i < num_counters; ++i) { |
530 | if (!reset_value[op_x86_phys_to_virt(i)]) | 537 | if (!reset_value[op_x86_phys_to_virt(i)]) |
531 | continue; | 538 | continue; |
532 | rdmsrl(msrs->controls[i].addr, val); | 539 | rdmsrl(msrs->controls[i].addr, val); |
@@ -546,7 +553,7 @@ static void op_amd_stop(struct op_msrs const * const msrs) | |||
546 | * Subtle: stop on all counters to avoid race with setting our | 553 | * Subtle: stop on all counters to avoid race with setting our |
547 | * pm callback | 554 | * pm callback |
548 | */ | 555 | */ |
549 | for (i = 0; i < NUM_COUNTERS; ++i) { | 556 | for (i = 0; i < num_counters; ++i) { |
550 | if (!reset_value[op_x86_phys_to_virt(i)]) | 557 | if (!reset_value[op_x86_phys_to_virt(i)]) |
551 | continue; | 558 | continue; |
552 | rdmsrl(msrs->controls[i].addr, val); | 559 | rdmsrl(msrs->controls[i].addr, val); |
@@ -706,18 +713,29 @@ static int setup_ibs_files(struct super_block *sb, struct dentry *root) | |||
706 | return 0; | 713 | return 0; |
707 | } | 714 | } |
708 | 715 | ||
716 | struct op_x86_model_spec op_amd_spec; | ||
717 | |||
709 | static int op_amd_init(struct oprofile_operations *ops) | 718 | static int op_amd_init(struct oprofile_operations *ops) |
710 | { | 719 | { |
711 | init_ibs(); | 720 | init_ibs(); |
712 | create_arch_files = ops->create_files; | 721 | create_arch_files = ops->create_files; |
713 | ops->create_files = setup_ibs_files; | 722 | ops->create_files = setup_ibs_files; |
723 | |||
724 | if (boot_cpu_data.x86 == 0x15) { | ||
725 | num_counters = NUM_COUNTERS_F15H; | ||
726 | } else { | ||
727 | num_counters = NUM_COUNTERS; | ||
728 | } | ||
729 | |||
730 | op_amd_spec.num_counters = num_counters; | ||
731 | op_amd_spec.num_controls = num_counters; | ||
732 | op_amd_spec.num_virt_counters = max(num_counters, NUM_VIRT_COUNTERS); | ||
733 | |||
714 | return 0; | 734 | return 0; |
715 | } | 735 | } |
716 | 736 | ||
717 | struct op_x86_model_spec op_amd_spec = { | 737 | struct op_x86_model_spec op_amd_spec = { |
718 | .num_counters = NUM_COUNTERS, | 738 | /* num_counters/num_controls filled in at runtime */ |
719 | .num_controls = NUM_COUNTERS, | ||
720 | .num_virt_counters = NUM_VIRT_COUNTERS, | ||
721 | .reserved = MSR_AMD_EVENTSEL_RESERVED, | 739 | .reserved = MSR_AMD_EVENTSEL_RESERVED, |
722 | .event_mask = OP_EVENT_MASK, | 740 | .event_mask = OP_EVENT_MASK, |
723 | .init = op_amd_init, | 741 | .init = op_amd_init, |
diff --git a/arch/x86/oprofile/op_model_p4.c b/arch/x86/oprofile/op_model_p4.c index 182558dd551..9fadec07414 100644 --- a/arch/x86/oprofile/op_model_p4.c +++ b/arch/x86/oprofile/op_model_p4.c | |||
@@ -11,7 +11,7 @@ | |||
11 | #include <linux/oprofile.h> | 11 | #include <linux/oprofile.h> |
12 | #include <linux/smp.h> | 12 | #include <linux/smp.h> |
13 | #include <linux/ptrace.h> | 13 | #include <linux/ptrace.h> |
14 | #include <linux/nmi.h> | 14 | #include <asm/nmi.h> |
15 | #include <asm/msr.h> | 15 | #include <asm/msr.h> |
16 | #include <asm/fixmap.h> | 16 | #include <asm/fixmap.h> |
17 | #include <asm/apic.h> | 17 | #include <asm/apic.h> |
diff --git a/drivers/acpi/acpica/nsinit.c b/drivers/acpi/acpica/nsinit.c index 660a2728908..0cac7ec0d2e 100644 --- a/drivers/acpi/acpica/nsinit.c +++ b/drivers/acpi/acpica/nsinit.c | |||
@@ -577,9 +577,7 @@ acpi_ns_init_one_device(acpi_handle obj_handle, | |||
577 | * as possible (without an NMI being received in the middle of | 577 | * as possible (without an NMI being received in the middle of |
578 | * this) - so disable NMIs and initialize the device: | 578 | * this) - so disable NMIs and initialize the device: |
579 | */ | 579 | */ |
580 | acpi_nmi_disable(); | ||
581 | status = acpi_ns_evaluate(info); | 580 | status = acpi_ns_evaluate(info); |
582 | acpi_nmi_enable(); | ||
583 | 581 | ||
584 | if (ACPI_SUCCESS(status)) { | 582 | if (ACPI_SUCCESS(status)) { |
585 | walk_info->num_INI++; | 583 | walk_info->num_INI++; |
diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index c63a4382374..1109f6848a4 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c | |||
@@ -355,6 +355,7 @@ void cpufreq_notify_transition(struct cpufreq_freqs *freqs, unsigned int state) | |||
355 | dprintk("FREQ: %lu - CPU: %lu", (unsigned long)freqs->new, | 355 | dprintk("FREQ: %lu - CPU: %lu", (unsigned long)freqs->new, |
356 | (unsigned long)freqs->cpu); | 356 | (unsigned long)freqs->cpu); |
357 | trace_power_frequency(POWER_PSTATE, freqs->new, freqs->cpu); | 357 | trace_power_frequency(POWER_PSTATE, freqs->new, freqs->cpu); |
358 | trace_cpu_frequency(freqs->new, freqs->cpu); | ||
358 | srcu_notifier_call_chain(&cpufreq_transition_notifier_list, | 359 | srcu_notifier_call_chain(&cpufreq_transition_notifier_list, |
359 | CPUFREQ_POSTCHANGE, freqs); | 360 | CPUFREQ_POSTCHANGE, freqs); |
360 | if (likely(policy) && likely(policy->cpu == freqs->cpu)) | 361 | if (likely(policy) && likely(policy->cpu == freqs->cpu)) |
diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c index a5071084337..08d5f05378d 100644 --- a/drivers/cpuidle/cpuidle.c +++ b/drivers/cpuidle/cpuidle.c | |||
@@ -107,6 +107,7 @@ static void cpuidle_idle_call(void) | |||
107 | if (cpuidle_curr_governor->reflect) | 107 | if (cpuidle_curr_governor->reflect) |
108 | cpuidle_curr_governor->reflect(dev); | 108 | cpuidle_curr_governor->reflect(dev); |
109 | trace_power_end(smp_processor_id()); | 109 | trace_power_end(smp_processor_id()); |
110 | trace_cpu_idle(PWR_EVENT_EXIT, smp_processor_id()); | ||
110 | } | 111 | } |
111 | 112 | ||
112 | /** | 113 | /** |
diff --git a/drivers/idle/intel_idle.c b/drivers/idle/intel_idle.c index c131d58bcb5..56ac09d6c93 100644 --- a/drivers/idle/intel_idle.c +++ b/drivers/idle/intel_idle.c | |||
@@ -220,9 +220,8 @@ static int intel_idle(struct cpuidle_device *dev, struct cpuidle_state *state) | |||
220 | kt_before = ktime_get_real(); | 220 | kt_before = ktime_get_real(); |
221 | 221 | ||
222 | stop_critical_timings(); | 222 | stop_critical_timings(); |
223 | #ifndef MODULE | ||
224 | trace_power_start(POWER_CSTATE, (eax >> 4) + 1, cpu); | 223 | trace_power_start(POWER_CSTATE, (eax >> 4) + 1, cpu); |
225 | #endif | 224 | trace_cpu_idle((eax >> 4) + 1, cpu); |
226 | if (!need_resched()) { | 225 | if (!need_resched()) { |
227 | 226 | ||
228 | __monitor((void *)¤t_thread_info()->flags, 0, 0); | 227 | __monitor((void *)¤t_thread_info()->flags, 0, 0); |
diff --git a/drivers/watchdog/hpwdt.c b/drivers/watchdog/hpwdt.c index 3d77116e463..dea7b5bf6e2 100644 --- a/drivers/watchdog/hpwdt.c +++ b/drivers/watchdog/hpwdt.c | |||
@@ -642,19 +642,14 @@ static struct notifier_block die_notifier = { | |||
642 | */ | 642 | */ |
643 | 643 | ||
644 | #ifdef CONFIG_HPWDT_NMI_DECODING | 644 | #ifdef CONFIG_HPWDT_NMI_DECODING |
645 | #ifdef ARCH_HAS_NMI_WATCHDOG | 645 | #ifdef CONFIG_X86_LOCAL_APIC |
646 | static void __devinit hpwdt_check_nmi_decoding(struct pci_dev *dev) | 646 | static void __devinit hpwdt_check_nmi_decoding(struct pci_dev *dev) |
647 | { | 647 | { |
648 | /* | 648 | /* |
649 | * If nmi_watchdog is turned off then we can turn on | 649 | * If nmi_watchdog is turned off then we can turn on |
650 | * our nmi decoding capability. | 650 | * our nmi decoding capability. |
651 | */ | 651 | */ |
652 | if (!nmi_watchdog_active()) | 652 | hpwdt_nmi_decoding = 1; |
653 | hpwdt_nmi_decoding = 1; | ||
654 | else | ||
655 | dev_warn(&dev->dev, "NMI decoding is disabled. To enable this " | ||
656 | "functionality you must reboot with nmi_watchdog=0 " | ||
657 | "and load the hpwdt driver with priority=1.\n"); | ||
658 | } | 653 | } |
659 | #else | 654 | #else |
660 | static void __devinit hpwdt_check_nmi_decoding(struct pci_dev *dev) | 655 | static void __devinit hpwdt_check_nmi_decoding(struct pci_dev *dev) |
@@ -662,7 +657,7 @@ static void __devinit hpwdt_check_nmi_decoding(struct pci_dev *dev) | |||
662 | dev_warn(&dev->dev, "NMI decoding is disabled. " | 657 | dev_warn(&dev->dev, "NMI decoding is disabled. " |
663 | "Your kernel does not support a NMI Watchdog.\n"); | 658 | "Your kernel does not support a NMI Watchdog.\n"); |
664 | } | 659 | } |
665 | #endif /* ARCH_HAS_NMI_WATCHDOG */ | 660 | #endif /* CONFIG_X86_LOCAL_APIC */ |
666 | 661 | ||
667 | static int __devinit hpwdt_init_nmi_decoding(struct pci_dev *dev) | 662 | static int __devinit hpwdt_init_nmi_decoding(struct pci_dev *dev) |
668 | { | 663 | { |
diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c index 5476c066d4e..3c4039d5eef 100644 --- a/fs/gfs2/bmap.c +++ b/fs/gfs2/bmap.c | |||
@@ -763,7 +763,7 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh, | |||
763 | int metadata; | 763 | int metadata; |
764 | unsigned int revokes = 0; | 764 | unsigned int revokes = 0; |
765 | int x; | 765 | int x; |
766 | int error; | 766 | int error = 0; |
767 | 767 | ||
768 | if (!*top) | 768 | if (!*top) |
769 | sm->sm_first = 0; | 769 | sm->sm_first = 0; |
@@ -780,7 +780,11 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh, | |||
780 | if (metadata) | 780 | if (metadata) |
781 | revokes = (height) ? sdp->sd_inptrs : sdp->sd_diptrs; | 781 | revokes = (height) ? sdp->sd_inptrs : sdp->sd_diptrs; |
782 | 782 | ||
783 | error = gfs2_rindex_hold(sdp, &ip->i_alloc->al_ri_gh); | 783 | if (ip != GFS2_I(sdp->sd_rindex)) |
784 | error = gfs2_rindex_hold(sdp, &ip->i_alloc->al_ri_gh); | ||
785 | else if (!sdp->sd_rgrps) | ||
786 | error = gfs2_ri_update(ip); | ||
787 | |||
784 | if (error) | 788 | if (error) |
785 | return error; | 789 | return error; |
786 | 790 | ||
@@ -879,7 +883,8 @@ out_rg_gunlock: | |||
879 | out_rlist: | 883 | out_rlist: |
880 | gfs2_rlist_free(&rlist); | 884 | gfs2_rlist_free(&rlist); |
881 | out: | 885 | out: |
882 | gfs2_glock_dq_uninit(&ip->i_alloc->al_ri_gh); | 886 | if (ip != GFS2_I(sdp->sd_rindex)) |
887 | gfs2_glock_dq_uninit(&ip->i_alloc->al_ri_gh); | ||
883 | return error; | 888 | return error; |
884 | } | 889 | } |
885 | 890 | ||
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index f92c1770416..08a8beb152e 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c | |||
@@ -541,21 +541,6 @@ out_locked: | |||
541 | spin_unlock(&gl->gl_spin); | 541 | spin_unlock(&gl->gl_spin); |
542 | } | 542 | } |
543 | 543 | ||
544 | static unsigned int gfs2_lm_lock(struct gfs2_sbd *sdp, void *lock, | ||
545 | unsigned int req_state, | ||
546 | unsigned int flags) | ||
547 | { | ||
548 | int ret = LM_OUT_ERROR; | ||
549 | |||
550 | if (!sdp->sd_lockstruct.ls_ops->lm_lock) | ||
551 | return req_state == LM_ST_UNLOCKED ? 0 : req_state; | ||
552 | |||
553 | if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) | ||
554 | ret = sdp->sd_lockstruct.ls_ops->lm_lock(lock, | ||
555 | req_state, flags); | ||
556 | return ret; | ||
557 | } | ||
558 | |||
559 | /** | 544 | /** |
560 | * do_xmote - Calls the DLM to change the state of a lock | 545 | * do_xmote - Calls the DLM to change the state of a lock |
561 | * @gl: The lock state | 546 | * @gl: The lock state |
@@ -575,13 +560,14 @@ __acquires(&gl->gl_spin) | |||
575 | 560 | ||
576 | lck_flags &= (LM_FLAG_TRY | LM_FLAG_TRY_1CB | LM_FLAG_NOEXP | | 561 | lck_flags &= (LM_FLAG_TRY | LM_FLAG_TRY_1CB | LM_FLAG_NOEXP | |
577 | LM_FLAG_PRIORITY); | 562 | LM_FLAG_PRIORITY); |
578 | BUG_ON(gl->gl_state == target); | 563 | GLOCK_BUG_ON(gl, gl->gl_state == target); |
579 | BUG_ON(gl->gl_state == gl->gl_target); | 564 | GLOCK_BUG_ON(gl, gl->gl_state == gl->gl_target); |
580 | if ((target == LM_ST_UNLOCKED || target == LM_ST_DEFERRED) && | 565 | if ((target == LM_ST_UNLOCKED || target == LM_ST_DEFERRED) && |
581 | glops->go_inval) { | 566 | glops->go_inval) { |
582 | set_bit(GLF_INVALIDATE_IN_PROGRESS, &gl->gl_flags); | 567 | set_bit(GLF_INVALIDATE_IN_PROGRESS, &gl->gl_flags); |
583 | do_error(gl, 0); /* Fail queued try locks */ | 568 | do_error(gl, 0); /* Fail queued try locks */ |
584 | } | 569 | } |
570 | gl->gl_req = target; | ||
585 | spin_unlock(&gl->gl_spin); | 571 | spin_unlock(&gl->gl_spin); |
586 | if (glops->go_xmote_th) | 572 | if (glops->go_xmote_th) |
587 | glops->go_xmote_th(gl); | 573 | glops->go_xmote_th(gl); |
@@ -594,15 +580,17 @@ __acquires(&gl->gl_spin) | |||
594 | gl->gl_state == LM_ST_DEFERRED) && | 580 | gl->gl_state == LM_ST_DEFERRED) && |
595 | !(lck_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB))) | 581 | !(lck_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB))) |
596 | lck_flags |= LM_FLAG_TRY_1CB; | 582 | lck_flags |= LM_FLAG_TRY_1CB; |
597 | ret = gfs2_lm_lock(sdp, gl, target, lck_flags); | ||
598 | 583 | ||
599 | if (!(ret & LM_OUT_ASYNC)) { | 584 | if (sdp->sd_lockstruct.ls_ops->lm_lock) { |
600 | finish_xmote(gl, ret); | 585 | /* lock_dlm */ |
586 | ret = sdp->sd_lockstruct.ls_ops->lm_lock(gl, target, lck_flags); | ||
587 | GLOCK_BUG_ON(gl, ret); | ||
588 | } else { /* lock_nolock */ | ||
589 | finish_xmote(gl, target); | ||
601 | if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) | 590 | if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) |
602 | gfs2_glock_put(gl); | 591 | gfs2_glock_put(gl); |
603 | } else { | ||
604 | GLOCK_BUG_ON(gl, ret != LM_OUT_ASYNC); | ||
605 | } | 592 | } |
593 | |||
606 | spin_lock(&gl->gl_spin); | 594 | spin_lock(&gl->gl_spin); |
607 | } | 595 | } |
608 | 596 | ||
@@ -951,17 +939,22 @@ int gfs2_glock_wait(struct gfs2_holder *gh) | |||
951 | 939 | ||
952 | void gfs2_print_dbg(struct seq_file *seq, const char *fmt, ...) | 940 | void gfs2_print_dbg(struct seq_file *seq, const char *fmt, ...) |
953 | { | 941 | { |
942 | struct va_format vaf; | ||
954 | va_list args; | 943 | va_list args; |
955 | 944 | ||
956 | va_start(args, fmt); | 945 | va_start(args, fmt); |
946 | |||
957 | if (seq) { | 947 | if (seq) { |
958 | struct gfs2_glock_iter *gi = seq->private; | 948 | struct gfs2_glock_iter *gi = seq->private; |
959 | vsprintf(gi->string, fmt, args); | 949 | vsprintf(gi->string, fmt, args); |
960 | seq_printf(seq, gi->string); | 950 | seq_printf(seq, gi->string); |
961 | } else { | 951 | } else { |
962 | printk(KERN_ERR " "); | 952 | vaf.fmt = fmt; |
963 | vprintk(fmt, args); | 953 | vaf.va = &args; |
954 | |||
955 | printk(KERN_ERR " %pV", &vaf); | ||
964 | } | 956 | } |
957 | |||
965 | va_end(args); | 958 | va_end(args); |
966 | } | 959 | } |
967 | 960 | ||
@@ -1361,24 +1354,28 @@ static int gfs2_should_freeze(const struct gfs2_glock *gl) | |||
1361 | * @gl: Pointer to the glock | 1354 | * @gl: Pointer to the glock |
1362 | * @ret: The return value from the dlm | 1355 | * @ret: The return value from the dlm |
1363 | * | 1356 | * |
1357 | * The gl_reply field is under the gl_spin lock so that it is ok | ||
1358 | * to use a bitfield shared with other glock state fields. | ||
1364 | */ | 1359 | */ |
1365 | 1360 | ||
1366 | void gfs2_glock_complete(struct gfs2_glock *gl, int ret) | 1361 | void gfs2_glock_complete(struct gfs2_glock *gl, int ret) |
1367 | { | 1362 | { |
1368 | struct lm_lockstruct *ls = &gl->gl_sbd->sd_lockstruct; | 1363 | struct lm_lockstruct *ls = &gl->gl_sbd->sd_lockstruct; |
1369 | 1364 | ||
1365 | spin_lock(&gl->gl_spin); | ||
1370 | gl->gl_reply = ret; | 1366 | gl->gl_reply = ret; |
1371 | 1367 | ||
1372 | if (unlikely(test_bit(DFL_BLOCK_LOCKS, &ls->ls_flags))) { | 1368 | if (unlikely(test_bit(DFL_BLOCK_LOCKS, &ls->ls_flags))) { |
1373 | spin_lock(&gl->gl_spin); | ||
1374 | if (gfs2_should_freeze(gl)) { | 1369 | if (gfs2_should_freeze(gl)) { |
1375 | set_bit(GLF_FROZEN, &gl->gl_flags); | 1370 | set_bit(GLF_FROZEN, &gl->gl_flags); |
1376 | spin_unlock(&gl->gl_spin); | 1371 | spin_unlock(&gl->gl_spin); |
1377 | return; | 1372 | return; |
1378 | } | 1373 | } |
1379 | spin_unlock(&gl->gl_spin); | ||
1380 | } | 1374 | } |
1375 | |||
1376 | spin_unlock(&gl->gl_spin); | ||
1381 | set_bit(GLF_REPLY_PENDING, &gl->gl_flags); | 1377 | set_bit(GLF_REPLY_PENDING, &gl->gl_flags); |
1378 | smp_wmb(); | ||
1382 | gfs2_glock_hold(gl); | 1379 | gfs2_glock_hold(gl); |
1383 | if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) | 1380 | if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) |
1384 | gfs2_glock_put(gl); | 1381 | gfs2_glock_put(gl); |
@@ -1626,18 +1623,17 @@ static const char *hflags2str(char *buf, unsigned flags, unsigned long iflags) | |||
1626 | static int dump_holder(struct seq_file *seq, const struct gfs2_holder *gh) | 1623 | static int dump_holder(struct seq_file *seq, const struct gfs2_holder *gh) |
1627 | { | 1624 | { |
1628 | struct task_struct *gh_owner = NULL; | 1625 | struct task_struct *gh_owner = NULL; |
1629 | char buffer[KSYM_SYMBOL_LEN]; | ||
1630 | char flags_buf[32]; | 1626 | char flags_buf[32]; |
1631 | 1627 | ||
1632 | sprint_symbol(buffer, gh->gh_ip); | ||
1633 | if (gh->gh_owner_pid) | 1628 | if (gh->gh_owner_pid) |
1634 | gh_owner = pid_task(gh->gh_owner_pid, PIDTYPE_PID); | 1629 | gh_owner = pid_task(gh->gh_owner_pid, PIDTYPE_PID); |
1635 | gfs2_print_dbg(seq, " H: s:%s f:%s e:%d p:%ld [%s] %s\n", | 1630 | gfs2_print_dbg(seq, " H: s:%s f:%s e:%d p:%ld [%s] %pS\n", |
1636 | state2str(gh->gh_state), | 1631 | state2str(gh->gh_state), |
1637 | hflags2str(flags_buf, gh->gh_flags, gh->gh_iflags), | 1632 | hflags2str(flags_buf, gh->gh_flags, gh->gh_iflags), |
1638 | gh->gh_error, | 1633 | gh->gh_error, |
1639 | gh->gh_owner_pid ? (long)pid_nr(gh->gh_owner_pid) : -1, | 1634 | gh->gh_owner_pid ? (long)pid_nr(gh->gh_owner_pid) : -1, |
1640 | gh_owner ? gh_owner->comm : "(ended)", buffer); | 1635 | gh_owner ? gh_owner->comm : "(ended)", |
1636 | (void *)gh->gh_ip); | ||
1641 | return 0; | 1637 | return 0; |
1642 | } | 1638 | } |
1643 | 1639 | ||
@@ -1782,12 +1778,13 @@ int __init gfs2_glock_init(void) | |||
1782 | } | 1778 | } |
1783 | #endif | 1779 | #endif |
1784 | 1780 | ||
1785 | glock_workqueue = alloc_workqueue("glock_workqueue", WQ_RESCUER | | 1781 | glock_workqueue = alloc_workqueue("glock_workqueue", WQ_MEM_RECLAIM | |
1786 | WQ_HIGHPRI | WQ_FREEZEABLE, 0); | 1782 | WQ_HIGHPRI | WQ_FREEZEABLE, 0); |
1787 | if (IS_ERR(glock_workqueue)) | 1783 | if (IS_ERR(glock_workqueue)) |
1788 | return PTR_ERR(glock_workqueue); | 1784 | return PTR_ERR(glock_workqueue); |
1789 | gfs2_delete_workqueue = alloc_workqueue("delete_workqueue", WQ_RESCUER | | 1785 | gfs2_delete_workqueue = alloc_workqueue("delete_workqueue", |
1790 | WQ_FREEZEABLE, 0); | 1786 | WQ_MEM_RECLAIM | WQ_FREEZEABLE, |
1787 | 0); | ||
1791 | if (IS_ERR(gfs2_delete_workqueue)) { | 1788 | if (IS_ERR(gfs2_delete_workqueue)) { |
1792 | destroy_workqueue(glock_workqueue); | 1789 | destroy_workqueue(glock_workqueue); |
1793 | return PTR_ERR(gfs2_delete_workqueue); | 1790 | return PTR_ERR(gfs2_delete_workqueue); |
diff --git a/fs/gfs2/glock.h b/fs/gfs2/glock.h index db1c26d6d22..691851ceb61 100644 --- a/fs/gfs2/glock.h +++ b/fs/gfs2/glock.h | |||
@@ -87,11 +87,10 @@ enum { | |||
87 | #define GL_ASYNC 0x00000040 | 87 | #define GL_ASYNC 0x00000040 |
88 | #define GL_EXACT 0x00000080 | 88 | #define GL_EXACT 0x00000080 |
89 | #define GL_SKIP 0x00000100 | 89 | #define GL_SKIP 0x00000100 |
90 | #define GL_ATIME 0x00000200 | ||
91 | #define GL_NOCACHE 0x00000400 | 90 | #define GL_NOCACHE 0x00000400 |
92 | 91 | ||
93 | /* | 92 | /* |
94 | * lm_lock() and lm_async_cb return flags | 93 | * lm_async_cb return flags |
95 | * | 94 | * |
96 | * LM_OUT_ST_MASK | 95 | * LM_OUT_ST_MASK |
97 | * Masks the lower two bits of lock state in the returned value. | 96 | * Masks the lower two bits of lock state in the returned value. |
@@ -99,15 +98,11 @@ enum { | |||
99 | * LM_OUT_CANCELED | 98 | * LM_OUT_CANCELED |
100 | * The lock request was canceled. | 99 | * The lock request was canceled. |
101 | * | 100 | * |
102 | * LM_OUT_ASYNC | ||
103 | * The result of the request will be returned in an LM_CB_ASYNC callback. | ||
104 | * | ||
105 | */ | 101 | */ |
106 | 102 | ||
107 | #define LM_OUT_ST_MASK 0x00000003 | 103 | #define LM_OUT_ST_MASK 0x00000003 |
108 | #define LM_OUT_CANCELED 0x00000008 | 104 | #define LM_OUT_CANCELED 0x00000008 |
109 | #define LM_OUT_ASYNC 0x00000080 | 105 | #define LM_OUT_ERROR 0x00000004 |
110 | #define LM_OUT_ERROR 0x00000100 | ||
111 | 106 | ||
112 | /* | 107 | /* |
113 | * lm_recovery_done() messages | 108 | * lm_recovery_done() messages |
@@ -124,25 +119,12 @@ struct lm_lockops { | |||
124 | void (*lm_unmount) (struct gfs2_sbd *sdp); | 119 | void (*lm_unmount) (struct gfs2_sbd *sdp); |
125 | void (*lm_withdraw) (struct gfs2_sbd *sdp); | 120 | void (*lm_withdraw) (struct gfs2_sbd *sdp); |
126 | void (*lm_put_lock) (struct kmem_cache *cachep, struct gfs2_glock *gl); | 121 | void (*lm_put_lock) (struct kmem_cache *cachep, struct gfs2_glock *gl); |
127 | unsigned int (*lm_lock) (struct gfs2_glock *gl, | 122 | int (*lm_lock) (struct gfs2_glock *gl, unsigned int req_state, |
128 | unsigned int req_state, unsigned int flags); | 123 | unsigned int flags); |
129 | void (*lm_cancel) (struct gfs2_glock *gl); | 124 | void (*lm_cancel) (struct gfs2_glock *gl); |
130 | const match_table_t *lm_tokens; | 125 | const match_table_t *lm_tokens; |
131 | }; | 126 | }; |
132 | 127 | ||
133 | #define LM_FLAG_TRY 0x00000001 | ||
134 | #define LM_FLAG_TRY_1CB 0x00000002 | ||
135 | #define LM_FLAG_NOEXP 0x00000004 | ||
136 | #define LM_FLAG_ANY 0x00000008 | ||
137 | #define LM_FLAG_PRIORITY 0x00000010 | ||
138 | |||
139 | #define GL_ASYNC 0x00000040 | ||
140 | #define GL_EXACT 0x00000080 | ||
141 | #define GL_SKIP 0x00000100 | ||
142 | #define GL_NOCACHE 0x00000400 | ||
143 | |||
144 | #define GLR_TRYFAILED 13 | ||
145 | |||
146 | extern struct workqueue_struct *gfs2_delete_workqueue; | 128 | extern struct workqueue_struct *gfs2_delete_workqueue; |
147 | static inline struct gfs2_holder *gfs2_glock_is_locked_by_me(struct gfs2_glock *gl) | 129 | static inline struct gfs2_holder *gfs2_glock_is_locked_by_me(struct gfs2_glock *gl) |
148 | { | 130 | { |
@@ -212,6 +194,8 @@ int gfs2_glock_nq_num(struct gfs2_sbd *sdp, | |||
212 | int gfs2_glock_nq_m(unsigned int num_gh, struct gfs2_holder *ghs); | 194 | int gfs2_glock_nq_m(unsigned int num_gh, struct gfs2_holder *ghs); |
213 | void gfs2_glock_dq_m(unsigned int num_gh, struct gfs2_holder *ghs); | 195 | void gfs2_glock_dq_m(unsigned int num_gh, struct gfs2_holder *ghs); |
214 | void gfs2_glock_dq_uninit_m(unsigned int num_gh, struct gfs2_holder *ghs); | 196 | void gfs2_glock_dq_uninit_m(unsigned int num_gh, struct gfs2_holder *ghs); |
197 | |||
198 | __attribute__ ((format(printf, 2, 3))) | ||
215 | void gfs2_print_dbg(struct seq_file *seq, const char *fmt, ...); | 199 | void gfs2_print_dbg(struct seq_file *seq, const char *fmt, ...); |
216 | 200 | ||
217 | /** | 201 | /** |
diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c index 0d149dcc04e..263561bf1a5 100644 --- a/fs/gfs2/glops.c +++ b/fs/gfs2/glops.c | |||
@@ -325,7 +325,6 @@ static void trans_go_sync(struct gfs2_glock *gl) | |||
325 | 325 | ||
326 | if (gl->gl_state != LM_ST_UNLOCKED && | 326 | if (gl->gl_state != LM_ST_UNLOCKED && |
327 | test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) { | 327 | test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) { |
328 | flush_workqueue(gfs2_delete_workqueue); | ||
329 | gfs2_meta_syncfs(sdp); | 328 | gfs2_meta_syncfs(sdp); |
330 | gfs2_log_shutdown(sdp); | 329 | gfs2_log_shutdown(sdp); |
331 | } | 330 | } |
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h index 764fbb49efc..8d3d2b4a0a7 100644 --- a/fs/gfs2/incore.h +++ b/fs/gfs2/incore.h | |||
@@ -207,12 +207,14 @@ struct gfs2_glock { | |||
207 | 207 | ||
208 | spinlock_t gl_spin; | 208 | spinlock_t gl_spin; |
209 | 209 | ||
210 | unsigned int gl_state; | 210 | /* State fields protected by gl_spin */ |
211 | unsigned int gl_target; | 211 | unsigned int gl_state:2, /* Current state */ |
212 | unsigned int gl_reply; | 212 | gl_target:2, /* Target state */ |
213 | gl_demote_state:2, /* State requested by remote node */ | ||
214 | gl_req:2, /* State in last dlm request */ | ||
215 | gl_reply:8; /* Last reply from the dlm */ | ||
216 | |||
213 | unsigned int gl_hash; | 217 | unsigned int gl_hash; |
214 | unsigned int gl_req; | ||
215 | unsigned int gl_demote_state; /* state requested by remote node */ | ||
216 | unsigned long gl_demote_time; /* time of first demote request */ | 218 | unsigned long gl_demote_time; /* time of first demote request */ |
217 | struct list_head gl_holders; | 219 | struct list_head gl_holders; |
218 | 220 | ||
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c index e1213f7f921..14e682dbe8b 100644 --- a/fs/gfs2/inode.c +++ b/fs/gfs2/inode.c | |||
@@ -916,17 +916,8 @@ static int __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr) | |||
916 | if (error) | 916 | if (error) |
917 | return error; | 917 | return error; |
918 | 918 | ||
919 | if ((attr->ia_valid & ATTR_SIZE) && | ||
920 | attr->ia_size != i_size_read(inode)) { | ||
921 | error = vmtruncate(inode, attr->ia_size); | ||
922 | if (error) | ||
923 | return error; | ||
924 | } | ||
925 | |||
926 | setattr_copy(inode, attr); | 919 | setattr_copy(inode, attr); |
927 | mark_inode_dirty(inode); | 920 | mark_inode_dirty(inode); |
928 | |||
929 | gfs2_assert_warn(GFS2_SB(inode), !error); | ||
930 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | 921 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); |
931 | gfs2_dinode_out(ip, dibh->b_data); | 922 | gfs2_dinode_out(ip, dibh->b_data); |
932 | brelse(dibh); | 923 | brelse(dibh); |
diff --git a/fs/gfs2/lock_dlm.c b/fs/gfs2/lock_dlm.c index 1c09425b45f..6e493aee28f 100644 --- a/fs/gfs2/lock_dlm.c +++ b/fs/gfs2/lock_dlm.c | |||
@@ -146,15 +146,13 @@ static u32 make_flags(const u32 lkid, const unsigned int gfs_flags, | |||
146 | return lkf; | 146 | return lkf; |
147 | } | 147 | } |
148 | 148 | ||
149 | static unsigned int gdlm_lock(struct gfs2_glock *gl, | 149 | static int gdlm_lock(struct gfs2_glock *gl, unsigned int req_state, |
150 | unsigned int req_state, unsigned int flags) | 150 | unsigned int flags) |
151 | { | 151 | { |
152 | struct lm_lockstruct *ls = &gl->gl_sbd->sd_lockstruct; | 152 | struct lm_lockstruct *ls = &gl->gl_sbd->sd_lockstruct; |
153 | int error; | ||
154 | int req; | 153 | int req; |
155 | u32 lkf; | 154 | u32 lkf; |
156 | 155 | ||
157 | gl->gl_req = req_state; | ||
158 | req = make_mode(req_state); | 156 | req = make_mode(req_state); |
159 | lkf = make_flags(gl->gl_lksb.sb_lkid, flags, req); | 157 | lkf = make_flags(gl->gl_lksb.sb_lkid, flags, req); |
160 | 158 | ||
@@ -162,13 +160,8 @@ static unsigned int gdlm_lock(struct gfs2_glock *gl, | |||
162 | * Submit the actual lock request. | 160 | * Submit the actual lock request. |
163 | */ | 161 | */ |
164 | 162 | ||
165 | error = dlm_lock(ls->ls_dlm, req, &gl->gl_lksb, lkf, gl->gl_strname, | 163 | return dlm_lock(ls->ls_dlm, req, &gl->gl_lksb, lkf, gl->gl_strname, |
166 | GDLM_STRNAME_BYTES - 1, 0, gdlm_ast, gl, gdlm_bast); | 164 | GDLM_STRNAME_BYTES - 1, 0, gdlm_ast, gl, gdlm_bast); |
167 | if (error == -EAGAIN) | ||
168 | return 0; | ||
169 | if (error) | ||
170 | return LM_OUT_ERROR; | ||
171 | return LM_OUT_ASYNC; | ||
172 | } | 165 | } |
173 | 166 | ||
174 | static void gdlm_put_lock(struct kmem_cache *cachep, struct gfs2_glock *gl) | 167 | static void gdlm_put_lock(struct kmem_cache *cachep, struct gfs2_glock *gl) |
diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c index 12cbea7502c..1db6b734322 100644 --- a/fs/gfs2/ops_inode.c +++ b/fs/gfs2/ops_inode.c | |||
@@ -1069,7 +1069,6 @@ static int setattr_chown(struct inode *inode, struct iattr *attr) | |||
1069 | { | 1069 | { |
1070 | struct gfs2_inode *ip = GFS2_I(inode); | 1070 | struct gfs2_inode *ip = GFS2_I(inode); |
1071 | struct gfs2_sbd *sdp = GFS2_SB(inode); | 1071 | struct gfs2_sbd *sdp = GFS2_SB(inode); |
1072 | struct buffer_head *dibh; | ||
1073 | u32 ouid, ogid, nuid, ngid; | 1072 | u32 ouid, ogid, nuid, ngid; |
1074 | int error; | 1073 | int error; |
1075 | 1074 | ||
@@ -1100,25 +1099,10 @@ static int setattr_chown(struct inode *inode, struct iattr *attr) | |||
1100 | if (error) | 1099 | if (error) |
1101 | goto out_gunlock_q; | 1100 | goto out_gunlock_q; |
1102 | 1101 | ||
1103 | error = gfs2_meta_inode_buffer(ip, &dibh); | 1102 | error = gfs2_setattr_simple(ip, attr); |
1104 | if (error) | 1103 | if (error) |
1105 | goto out_end_trans; | 1104 | goto out_end_trans; |
1106 | 1105 | ||
1107 | if ((attr->ia_valid & ATTR_SIZE) && | ||
1108 | attr->ia_size != i_size_read(inode)) { | ||
1109 | int error; | ||
1110 | |||
1111 | error = vmtruncate(inode, attr->ia_size); | ||
1112 | gfs2_assert_warn(sdp, !error); | ||
1113 | } | ||
1114 | |||
1115 | setattr_copy(inode, attr); | ||
1116 | mark_inode_dirty(inode); | ||
1117 | |||
1118 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | ||
1119 | gfs2_dinode_out(ip, dibh->b_data); | ||
1120 | brelse(dibh); | ||
1121 | |||
1122 | if (ouid != NO_QUOTA_CHANGE || ogid != NO_QUOTA_CHANGE) { | 1106 | if (ouid != NO_QUOTA_CHANGE || ogid != NO_QUOTA_CHANGE) { |
1123 | u64 blocks = gfs2_get_inode_blocks(&ip->i_inode); | 1107 | u64 blocks = gfs2_get_inode_blocks(&ip->i_inode); |
1124 | gfs2_quota_change(ip, -blocks, ouid, ogid); | 1108 | gfs2_quota_change(ip, -blocks, ouid, ogid); |
diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c index f606baf9ba7..a689901963d 100644 --- a/fs/gfs2/quota.c +++ b/fs/gfs2/quota.c | |||
@@ -666,6 +666,10 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc, | |||
666 | qp->qu_limit = cpu_to_be64(fdq->d_blk_hardlimit >> sdp->sd_fsb2bb_shift); | 666 | qp->qu_limit = cpu_to_be64(fdq->d_blk_hardlimit >> sdp->sd_fsb2bb_shift); |
667 | qd->qd_qb.qb_limit = qp->qu_limit; | 667 | qd->qd_qb.qb_limit = qp->qu_limit; |
668 | } | 668 | } |
669 | if (fdq->d_fieldmask & FS_DQ_BCOUNT) { | ||
670 | qp->qu_value = cpu_to_be64(fdq->d_bcount >> sdp->sd_fsb2bb_shift); | ||
671 | qd->qd_qb.qb_value = qp->qu_value; | ||
672 | } | ||
669 | } | 673 | } |
670 | 674 | ||
671 | /* Write the quota into the quota file on disk */ | 675 | /* Write the quota into the quota file on disk */ |
@@ -1509,7 +1513,7 @@ out: | |||
1509 | } | 1513 | } |
1510 | 1514 | ||
1511 | /* GFS2 only supports a subset of the XFS fields */ | 1515 | /* GFS2 only supports a subset of the XFS fields */ |
1512 | #define GFS2_FIELDMASK (FS_DQ_BSOFT|FS_DQ_BHARD) | 1516 | #define GFS2_FIELDMASK (FS_DQ_BSOFT|FS_DQ_BHARD|FS_DQ_BCOUNT) |
1513 | 1517 | ||
1514 | static int gfs2_set_dqblk(struct super_block *sb, int type, qid_t id, | 1518 | static int gfs2_set_dqblk(struct super_block *sb, int type, qid_t id, |
1515 | struct fs_disk_quota *fdq) | 1519 | struct fs_disk_quota *fdq) |
@@ -1569,9 +1573,15 @@ static int gfs2_set_dqblk(struct super_block *sb, int type, qid_t id, | |||
1569 | if ((fdq->d_fieldmask & FS_DQ_BSOFT) && | 1573 | if ((fdq->d_fieldmask & FS_DQ_BSOFT) && |
1570 | ((fdq->d_blk_softlimit >> sdp->sd_fsb2bb_shift) == be64_to_cpu(qd->qd_qb.qb_warn))) | 1574 | ((fdq->d_blk_softlimit >> sdp->sd_fsb2bb_shift) == be64_to_cpu(qd->qd_qb.qb_warn))) |
1571 | fdq->d_fieldmask ^= FS_DQ_BSOFT; | 1575 | fdq->d_fieldmask ^= FS_DQ_BSOFT; |
1576 | |||
1572 | if ((fdq->d_fieldmask & FS_DQ_BHARD) && | 1577 | if ((fdq->d_fieldmask & FS_DQ_BHARD) && |
1573 | ((fdq->d_blk_hardlimit >> sdp->sd_fsb2bb_shift) == be64_to_cpu(qd->qd_qb.qb_limit))) | 1578 | ((fdq->d_blk_hardlimit >> sdp->sd_fsb2bb_shift) == be64_to_cpu(qd->qd_qb.qb_limit))) |
1574 | fdq->d_fieldmask ^= FS_DQ_BHARD; | 1579 | fdq->d_fieldmask ^= FS_DQ_BHARD; |
1580 | |||
1581 | if ((fdq->d_fieldmask & FS_DQ_BCOUNT) && | ||
1582 | ((fdq->d_bcount >> sdp->sd_fsb2bb_shift) == be64_to_cpu(qd->qd_qb.qb_value))) | ||
1583 | fdq->d_fieldmask ^= FS_DQ_BCOUNT; | ||
1584 | |||
1575 | if (fdq->d_fieldmask == 0) | 1585 | if (fdq->d_fieldmask == 0) |
1576 | goto out_i; | 1586 | goto out_i; |
1577 | 1587 | ||
@@ -1620,4 +1630,3 @@ const struct quotactl_ops gfs2_quotactl_ops = { | |||
1620 | .get_dqblk = gfs2_get_dqblk, | 1630 | .get_dqblk = gfs2_get_dqblk, |
1621 | .set_dqblk = gfs2_set_dqblk, | 1631 | .set_dqblk = gfs2_set_dqblk, |
1622 | }; | 1632 | }; |
1623 | |||
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c index 33c8407b876..7293ea27020 100644 --- a/fs/gfs2/rgrp.c +++ b/fs/gfs2/rgrp.c | |||
@@ -500,7 +500,7 @@ u64 gfs2_ri_total(struct gfs2_sbd *sdp) | |||
500 | for (rgrps = 0;; rgrps++) { | 500 | for (rgrps = 0;; rgrps++) { |
501 | loff_t pos = rgrps * sizeof(struct gfs2_rindex); | 501 | loff_t pos = rgrps * sizeof(struct gfs2_rindex); |
502 | 502 | ||
503 | if (pos + sizeof(struct gfs2_rindex) >= i_size_read(inode)) | 503 | if (pos + sizeof(struct gfs2_rindex) > i_size_read(inode)) |
504 | break; | 504 | break; |
505 | error = gfs2_internal_read(ip, &ra_state, buf, &pos, | 505 | error = gfs2_internal_read(ip, &ra_state, buf, &pos, |
506 | sizeof(struct gfs2_rindex)); | 506 | sizeof(struct gfs2_rindex)); |
@@ -583,7 +583,7 @@ static int read_rindex_entry(struct gfs2_inode *ip, | |||
583 | * Returns: 0 on successful update, error code otherwise | 583 | * Returns: 0 on successful update, error code otherwise |
584 | */ | 584 | */ |
585 | 585 | ||
586 | static int gfs2_ri_update(struct gfs2_inode *ip) | 586 | int gfs2_ri_update(struct gfs2_inode *ip) |
587 | { | 587 | { |
588 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | 588 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
589 | struct inode *inode = &ip->i_inode; | 589 | struct inode *inode = &ip->i_inode; |
@@ -614,46 +614,6 @@ static int gfs2_ri_update(struct gfs2_inode *ip) | |||
614 | } | 614 | } |
615 | 615 | ||
616 | /** | 616 | /** |
617 | * gfs2_ri_update_special - Pull in a new resource index from the disk | ||
618 | * | ||
619 | * This is a special version that's safe to call from gfs2_inplace_reserve_i. | ||
620 | * In this case we know that we don't have any resource groups in memory yet. | ||
621 | * | ||
622 | * @ip: pointer to the rindex inode | ||
623 | * | ||
624 | * Returns: 0 on successful update, error code otherwise | ||
625 | */ | ||
626 | static int gfs2_ri_update_special(struct gfs2_inode *ip) | ||
627 | { | ||
628 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | ||
629 | struct inode *inode = &ip->i_inode; | ||
630 | struct file_ra_state ra_state; | ||
631 | struct gfs2_rgrpd *rgd; | ||
632 | unsigned int max_data = 0; | ||
633 | int error; | ||
634 | |||
635 | file_ra_state_init(&ra_state, inode->i_mapping); | ||
636 | for (sdp->sd_rgrps = 0;; sdp->sd_rgrps++) { | ||
637 | /* Ignore partials */ | ||
638 | if ((sdp->sd_rgrps + 1) * sizeof(struct gfs2_rindex) > | ||
639 | i_size_read(inode)) | ||
640 | break; | ||
641 | error = read_rindex_entry(ip, &ra_state); | ||
642 | if (error) { | ||
643 | clear_rgrpdi(sdp); | ||
644 | return error; | ||
645 | } | ||
646 | } | ||
647 | list_for_each_entry(rgd, &sdp->sd_rindex_list, rd_list) | ||
648 | if (rgd->rd_data > max_data) | ||
649 | max_data = rgd->rd_data; | ||
650 | sdp->sd_max_rg_data = max_data; | ||
651 | |||
652 | sdp->sd_rindex_uptodate = 1; | ||
653 | return 0; | ||
654 | } | ||
655 | |||
656 | /** | ||
657 | * gfs2_rindex_hold - Grab a lock on the rindex | 617 | * gfs2_rindex_hold - Grab a lock on the rindex |
658 | * @sdp: The GFS2 superblock | 618 | * @sdp: The GFS2 superblock |
659 | * @ri_gh: the glock holder | 619 | * @ri_gh: the glock holder |
@@ -1226,16 +1186,25 @@ int gfs2_inplace_reserve_i(struct gfs2_inode *ip, int hold_rindex, | |||
1226 | error = gfs2_rindex_hold(sdp, &al->al_ri_gh); | 1186 | error = gfs2_rindex_hold(sdp, &al->al_ri_gh); |
1227 | else if (!sdp->sd_rgrps) /* We may not have the rindex read | 1187 | else if (!sdp->sd_rgrps) /* We may not have the rindex read |
1228 | in, so: */ | 1188 | in, so: */ |
1229 | error = gfs2_ri_update_special(ip); | 1189 | error = gfs2_ri_update(ip); |
1230 | if (error) | 1190 | if (error) |
1231 | return error; | 1191 | return error; |
1232 | } | 1192 | } |
1233 | 1193 | ||
1194 | try_again: | ||
1234 | do { | 1195 | do { |
1235 | error = get_local_rgrp(ip, &last_unlinked); | 1196 | error = get_local_rgrp(ip, &last_unlinked); |
1236 | /* If there is no space, flushing the log may release some */ | 1197 | /* If there is no space, flushing the log may release some */ |
1237 | if (error) | 1198 | if (error) { |
1199 | if (ip == GFS2_I(sdp->sd_rindex) && | ||
1200 | !sdp->sd_rindex_uptodate) { | ||
1201 | error = gfs2_ri_update(ip); | ||
1202 | if (error) | ||
1203 | return error; | ||
1204 | goto try_again; | ||
1205 | } | ||
1238 | gfs2_log_flush(sdp, NULL); | 1206 | gfs2_log_flush(sdp, NULL); |
1207 | } | ||
1239 | } while (error && tries++ < 3); | 1208 | } while (error && tries++ < 3); |
1240 | 1209 | ||
1241 | if (error) { | 1210 | if (error) { |
diff --git a/fs/gfs2/rgrp.h b/fs/gfs2/rgrp.h index 0e35c0466f9..50c2bb04369 100644 --- a/fs/gfs2/rgrp.h +++ b/fs/gfs2/rgrp.h | |||
@@ -48,6 +48,7 @@ extern int gfs2_inplace_reserve_i(struct gfs2_inode *ip, int hold_rindex, | |||
48 | 48 | ||
49 | extern void gfs2_inplace_release(struct gfs2_inode *ip); | 49 | extern void gfs2_inplace_release(struct gfs2_inode *ip); |
50 | 50 | ||
51 | extern int gfs2_ri_update(struct gfs2_inode *ip); | ||
51 | extern int gfs2_alloc_block(struct gfs2_inode *ip, u64 *bn, unsigned int *n); | 52 | extern int gfs2_alloc_block(struct gfs2_inode *ip, u64 *bn, unsigned int *n); |
52 | extern int gfs2_alloc_di(struct gfs2_inode *ip, u64 *bn, u64 *generation); | 53 | extern int gfs2_alloc_di(struct gfs2_inode *ip, u64 *bn, u64 *generation); |
53 | 54 | ||
diff --git a/fs/gfs2/xattr.c b/fs/gfs2/xattr.c index 30b58f07c8a..439b61c0326 100644 --- a/fs/gfs2/xattr.c +++ b/fs/gfs2/xattr.c | |||
@@ -1296,10 +1296,8 @@ fail: | |||
1296 | 1296 | ||
1297 | int gfs2_xattr_acl_chmod(struct gfs2_inode *ip, struct iattr *attr, char *data) | 1297 | int gfs2_xattr_acl_chmod(struct gfs2_inode *ip, struct iattr *attr, char *data) |
1298 | { | 1298 | { |
1299 | struct inode *inode = &ip->i_inode; | ||
1300 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | 1299 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
1301 | struct gfs2_ea_location el; | 1300 | struct gfs2_ea_location el; |
1302 | struct buffer_head *dibh; | ||
1303 | int error; | 1301 | int error; |
1304 | 1302 | ||
1305 | error = gfs2_ea_find(ip, GFS2_EATYPE_SYS, GFS2_POSIX_ACL_ACCESS, &el); | 1303 | error = gfs2_ea_find(ip, GFS2_EATYPE_SYS, GFS2_POSIX_ACL_ACCESS, &el); |
@@ -1321,26 +1319,7 @@ int gfs2_xattr_acl_chmod(struct gfs2_inode *ip, struct iattr *attr, char *data) | |||
1321 | if (error) | 1319 | if (error) |
1322 | return error; | 1320 | return error; |
1323 | 1321 | ||
1324 | error = gfs2_meta_inode_buffer(ip, &dibh); | 1322 | error = gfs2_setattr_simple(ip, attr); |
1325 | if (error) | ||
1326 | goto out_trans_end; | ||
1327 | |||
1328 | if ((attr->ia_valid & ATTR_SIZE) && | ||
1329 | attr->ia_size != i_size_read(inode)) { | ||
1330 | int error; | ||
1331 | |||
1332 | error = vmtruncate(inode, attr->ia_size); | ||
1333 | gfs2_assert_warn(GFS2_SB(inode), !error); | ||
1334 | } | ||
1335 | |||
1336 | setattr_copy(inode, attr); | ||
1337 | mark_inode_dirty(inode); | ||
1338 | |||
1339 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | ||
1340 | gfs2_dinode_out(ip, dibh->b_data); | ||
1341 | brelse(dibh); | ||
1342 | |||
1343 | out_trans_end: | ||
1344 | gfs2_trans_end(sdp); | 1323 | gfs2_trans_end(sdp); |
1345 | return error; | 1324 | return error; |
1346 | } | 1325 | } |
diff --git a/include/linux/ftrace_event.h b/include/linux/ftrace_event.h index 8beabb958f6..47e3997f7b5 100644 --- a/include/linux/ftrace_event.h +++ b/include/linux/ftrace_event.h | |||
@@ -154,12 +154,14 @@ enum { | |||
154 | TRACE_EVENT_FL_ENABLED_BIT, | 154 | TRACE_EVENT_FL_ENABLED_BIT, |
155 | TRACE_EVENT_FL_FILTERED_BIT, | 155 | TRACE_EVENT_FL_FILTERED_BIT, |
156 | TRACE_EVENT_FL_RECORDED_CMD_BIT, | 156 | TRACE_EVENT_FL_RECORDED_CMD_BIT, |
157 | TRACE_EVENT_FL_CAP_ANY_BIT, | ||
157 | }; | 158 | }; |
158 | 159 | ||
159 | enum { | 160 | enum { |
160 | TRACE_EVENT_FL_ENABLED = (1 << TRACE_EVENT_FL_ENABLED_BIT), | 161 | TRACE_EVENT_FL_ENABLED = (1 << TRACE_EVENT_FL_ENABLED_BIT), |
161 | TRACE_EVENT_FL_FILTERED = (1 << TRACE_EVENT_FL_FILTERED_BIT), | 162 | TRACE_EVENT_FL_FILTERED = (1 << TRACE_EVENT_FL_FILTERED_BIT), |
162 | TRACE_EVENT_FL_RECORDED_CMD = (1 << TRACE_EVENT_FL_RECORDED_CMD_BIT), | 163 | TRACE_EVENT_FL_RECORDED_CMD = (1 << TRACE_EVENT_FL_RECORDED_CMD_BIT), |
164 | TRACE_EVENT_FL_CAP_ANY = (1 << TRACE_EVENT_FL_CAP_ANY_BIT), | ||
163 | }; | 165 | }; |
164 | 166 | ||
165 | struct ftrace_event_call { | 167 | struct ftrace_event_call { |
@@ -196,6 +198,14 @@ struct ftrace_event_call { | |||
196 | #endif | 198 | #endif |
197 | }; | 199 | }; |
198 | 200 | ||
201 | #define __TRACE_EVENT_FLAGS(name, value) \ | ||
202 | static int __init trace_init_flags_##name(void) \ | ||
203 | { \ | ||
204 | event_##name.flags = value; \ | ||
205 | return 0; \ | ||
206 | } \ | ||
207 | early_initcall(trace_init_flags_##name); | ||
208 | |||
199 | #define PERF_MAX_TRACE_SIZE 2048 | 209 | #define PERF_MAX_TRACE_SIZE 2048 |
200 | 210 | ||
201 | #define MAX_FILTER_PRED 32 | 211 | #define MAX_FILTER_PRED 32 |
@@ -215,6 +225,10 @@ enum { | |||
215 | FILTER_PTR_STRING, | 225 | FILTER_PTR_STRING, |
216 | }; | 226 | }; |
217 | 227 | ||
228 | #define EVENT_STORAGE_SIZE 128 | ||
229 | extern struct mutex event_storage_mutex; | ||
230 | extern char event_storage[EVENT_STORAGE_SIZE]; | ||
231 | |||
218 | extern int trace_event_raw_init(struct ftrace_event_call *call); | 232 | extern int trace_event_raw_init(struct ftrace_event_call *call); |
219 | extern int trace_define_field(struct ftrace_event_call *call, const char *type, | 233 | extern int trace_define_field(struct ftrace_event_call *call, const char *type, |
220 | const char *name, int offset, int size, | 234 | const char *name, int offset, int size, |
diff --git a/include/linux/init_task.h b/include/linux/init_task.h index 6ed8812bfe2..caa151fbebb 100644 --- a/include/linux/init_task.h +++ b/include/linux/init_task.h | |||
@@ -90,6 +90,12 @@ extern struct group_info init_groups; | |||
90 | */ | 90 | */ |
91 | # define CAP_INIT_BSET CAP_FULL_SET | 91 | # define CAP_INIT_BSET CAP_FULL_SET |
92 | 92 | ||
93 | #ifdef CONFIG_RCU_BOOST | ||
94 | #define INIT_TASK_RCU_BOOST() \ | ||
95 | .rcu_boost_mutex = NULL, | ||
96 | #else | ||
97 | #define INIT_TASK_RCU_BOOST() | ||
98 | #endif | ||
93 | #ifdef CONFIG_TREE_PREEMPT_RCU | 99 | #ifdef CONFIG_TREE_PREEMPT_RCU |
94 | #define INIT_TASK_RCU_TREE_PREEMPT() \ | 100 | #define INIT_TASK_RCU_TREE_PREEMPT() \ |
95 | .rcu_blocked_node = NULL, | 101 | .rcu_blocked_node = NULL, |
@@ -101,7 +107,8 @@ extern struct group_info init_groups; | |||
101 | .rcu_read_lock_nesting = 0, \ | 107 | .rcu_read_lock_nesting = 0, \ |
102 | .rcu_read_unlock_special = 0, \ | 108 | .rcu_read_unlock_special = 0, \ |
103 | .rcu_node_entry = LIST_HEAD_INIT(tsk.rcu_node_entry), \ | 109 | .rcu_node_entry = LIST_HEAD_INIT(tsk.rcu_node_entry), \ |
104 | INIT_TASK_RCU_TREE_PREEMPT() | 110 | INIT_TASK_RCU_TREE_PREEMPT() \ |
111 | INIT_TASK_RCU_BOOST() | ||
105 | #else | 112 | #else |
106 | #define INIT_TASK_RCU_PREEMPT(tsk) | 113 | #define INIT_TASK_RCU_PREEMPT(tsk) |
107 | #endif | 114 | #endif |
diff --git a/include/linux/kprobes.h b/include/linux/kprobes.h index e7d1b2e0070..b78edb58ee6 100644 --- a/include/linux/kprobes.h +++ b/include/linux/kprobes.h | |||
@@ -275,7 +275,9 @@ extern int arch_prepared_optinsn(struct arch_optimized_insn *optinsn); | |||
275 | extern int arch_check_optimized_kprobe(struct optimized_kprobe *op); | 275 | extern int arch_check_optimized_kprobe(struct optimized_kprobe *op); |
276 | extern int arch_prepare_optimized_kprobe(struct optimized_kprobe *op); | 276 | extern int arch_prepare_optimized_kprobe(struct optimized_kprobe *op); |
277 | extern void arch_remove_optimized_kprobe(struct optimized_kprobe *op); | 277 | extern void arch_remove_optimized_kprobe(struct optimized_kprobe *op); |
278 | extern int arch_optimize_kprobe(struct optimized_kprobe *op); | 278 | extern void arch_optimize_kprobes(struct list_head *oplist); |
279 | extern void arch_unoptimize_kprobes(struct list_head *oplist, | ||
280 | struct list_head *done_list); | ||
279 | extern void arch_unoptimize_kprobe(struct optimized_kprobe *op); | 281 | extern void arch_unoptimize_kprobe(struct optimized_kprobe *op); |
280 | extern kprobe_opcode_t *get_optinsn_slot(void); | 282 | extern kprobe_opcode_t *get_optinsn_slot(void); |
281 | extern void free_optinsn_slot(kprobe_opcode_t *slot, int dirty); | 283 | extern void free_optinsn_slot(kprobe_opcode_t *slot, int dirty); |
diff --git a/include/linux/nmi.h b/include/linux/nmi.h index 06aab5eee13..c536f8545f7 100644 --- a/include/linux/nmi.h +++ b/include/linux/nmi.h | |||
@@ -14,22 +14,14 @@ | |||
14 | * may be used to reset the timeout - for code which intentionally | 14 | * may be used to reset the timeout - for code which intentionally |
15 | * disables interrupts for a long time. This call is stateless. | 15 | * disables interrupts for a long time. This call is stateless. |
16 | */ | 16 | */ |
17 | #ifdef ARCH_HAS_NMI_WATCHDOG | 17 | #if defined(ARCH_HAS_NMI_WATCHDOG) || defined(CONFIG_HARDLOCKUP_DETECTOR) |
18 | #include <asm/nmi.h> | 18 | #include <asm/nmi.h> |
19 | extern void touch_nmi_watchdog(void); | 19 | extern void touch_nmi_watchdog(void); |
20 | extern void acpi_nmi_disable(void); | ||
21 | extern void acpi_nmi_enable(void); | ||
22 | #else | 20 | #else |
23 | #ifndef CONFIG_HARDLOCKUP_DETECTOR | ||
24 | static inline void touch_nmi_watchdog(void) | 21 | static inline void touch_nmi_watchdog(void) |
25 | { | 22 | { |
26 | touch_softlockup_watchdog(); | 23 | touch_softlockup_watchdog(); |
27 | } | 24 | } |
28 | #else | ||
29 | extern void touch_nmi_watchdog(void); | ||
30 | #endif | ||
31 | static inline void acpi_nmi_disable(void) { } | ||
32 | static inline void acpi_nmi_enable(void) { } | ||
33 | #endif | 25 | #endif |
34 | 26 | ||
35 | /* | 27 | /* |
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 4f1279e105e..dda5b0a3ff6 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h | |||
@@ -215,8 +215,9 @@ struct perf_event_attr { | |||
215 | */ | 215 | */ |
216 | precise_ip : 2, /* skid constraint */ | 216 | precise_ip : 2, /* skid constraint */ |
217 | mmap_data : 1, /* non-exec mmap data */ | 217 | mmap_data : 1, /* non-exec mmap data */ |
218 | sample_id_all : 1, /* sample_type all events */ | ||
218 | 219 | ||
219 | __reserved_1 : 46; | 220 | __reserved_1 : 45; |
220 | 221 | ||
221 | union { | 222 | union { |
222 | __u32 wakeup_events; /* wakeup every n events */ | 223 | __u32 wakeup_events; /* wakeup every n events */ |
@@ -327,6 +328,15 @@ struct perf_event_header { | |||
327 | enum perf_event_type { | 328 | enum perf_event_type { |
328 | 329 | ||
329 | /* | 330 | /* |
331 | * If perf_event_attr.sample_id_all is set then all event types will | ||
332 | * have the sample_type selected fields related to where/when | ||
333 | * (identity) an event took place (TID, TIME, ID, CPU, STREAM_ID) | ||
334 | * described in PERF_RECORD_SAMPLE below, it will be stashed just after | ||
335 | * the perf_event_header and the fields already present for the existing | ||
336 | * fields, i.e. at the end of the payload. That way a newer perf.data | ||
337 | * file will be supported by older perf tools, with these new optional | ||
338 | * fields being ignored. | ||
339 | * | ||
330 | * The MMAP events record the PROT_EXEC mappings so that we can | 340 | * The MMAP events record the PROT_EXEC mappings so that we can |
331 | * correlate userspace IPs to code. They have the following structure: | 341 | * correlate userspace IPs to code. They have the following structure: |
332 | * | 342 | * |
@@ -578,6 +588,10 @@ struct perf_event; | |||
578 | struct pmu { | 588 | struct pmu { |
579 | struct list_head entry; | 589 | struct list_head entry; |
580 | 590 | ||
591 | struct device *dev; | ||
592 | char *name; | ||
593 | int type; | ||
594 | |||
581 | int * __percpu pmu_disable_count; | 595 | int * __percpu pmu_disable_count; |
582 | struct perf_cpu_context * __percpu pmu_cpu_context; | 596 | struct perf_cpu_context * __percpu pmu_cpu_context; |
583 | int task_ctx_nr; | 597 | int task_ctx_nr; |
@@ -758,6 +772,9 @@ struct perf_event { | |||
758 | u64 shadow_ctx_time; | 772 | u64 shadow_ctx_time; |
759 | 773 | ||
760 | struct perf_event_attr attr; | 774 | struct perf_event_attr attr; |
775 | u16 header_size; | ||
776 | u16 id_header_size; | ||
777 | u16 read_size; | ||
761 | struct hw_perf_event hw; | 778 | struct hw_perf_event hw; |
762 | 779 | ||
763 | struct perf_event_context *ctx; | 780 | struct perf_event_context *ctx; |
@@ -903,7 +920,7 @@ struct perf_output_handle { | |||
903 | 920 | ||
904 | #ifdef CONFIG_PERF_EVENTS | 921 | #ifdef CONFIG_PERF_EVENTS |
905 | 922 | ||
906 | extern int perf_pmu_register(struct pmu *pmu); | 923 | extern int perf_pmu_register(struct pmu *pmu, char *name, int type); |
907 | extern void perf_pmu_unregister(struct pmu *pmu); | 924 | extern void perf_pmu_unregister(struct pmu *pmu); |
908 | 925 | ||
909 | extern int perf_num_counters(void); | 926 | extern int perf_num_counters(void); |
@@ -970,6 +987,11 @@ extern int perf_event_overflow(struct perf_event *event, int nmi, | |||
970 | struct perf_sample_data *data, | 987 | struct perf_sample_data *data, |
971 | struct pt_regs *regs); | 988 | struct pt_regs *regs); |
972 | 989 | ||
990 | static inline bool is_sampling_event(struct perf_event *event) | ||
991 | { | ||
992 | return event->attr.sample_period != 0; | ||
993 | } | ||
994 | |||
973 | /* | 995 | /* |
974 | * Return 1 for a software event, 0 for a hardware event | 996 | * Return 1 for a software event, 0 for a hardware event |
975 | */ | 997 | */ |
diff --git a/include/linux/rculist.h b/include/linux/rculist.h index f31ef61f1c6..2dea94fc440 100644 --- a/include/linux/rculist.h +++ b/include/linux/rculist.h | |||
@@ -241,11 +241,6 @@ static inline void list_splice_init_rcu(struct list_head *list, | |||
241 | #define list_first_entry_rcu(ptr, type, member) \ | 241 | #define list_first_entry_rcu(ptr, type, member) \ |
242 | list_entry_rcu((ptr)->next, type, member) | 242 | list_entry_rcu((ptr)->next, type, member) |
243 | 243 | ||
244 | #define __list_for_each_rcu(pos, head) \ | ||
245 | for (pos = rcu_dereference_raw(list_next_rcu(head)); \ | ||
246 | pos != (head); \ | ||
247 | pos = rcu_dereference_raw(list_next_rcu((pos))) | ||
248 | |||
249 | /** | 244 | /** |
250 | * list_for_each_entry_rcu - iterate over rcu list of given type | 245 | * list_for_each_entry_rcu - iterate over rcu list of given type |
251 | * @pos: the type * to use as a loop cursor. | 246 | * @pos: the type * to use as a loop cursor. |
diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h index 03cda7bed98..af561485628 100644 --- a/include/linux/rcupdate.h +++ b/include/linux/rcupdate.h | |||
@@ -47,6 +47,8 @@ | |||
47 | extern int rcutorture_runnable; /* for sysctl */ | 47 | extern int rcutorture_runnable; /* for sysctl */ |
48 | #endif /* #ifdef CONFIG_RCU_TORTURE_TEST */ | 48 | #endif /* #ifdef CONFIG_RCU_TORTURE_TEST */ |
49 | 49 | ||
50 | #define UINT_CMP_GE(a, b) (UINT_MAX / 2 >= (a) - (b)) | ||
51 | #define UINT_CMP_LT(a, b) (UINT_MAX / 2 < (a) - (b)) | ||
50 | #define ULONG_CMP_GE(a, b) (ULONG_MAX / 2 >= (a) - (b)) | 52 | #define ULONG_CMP_GE(a, b) (ULONG_MAX / 2 >= (a) - (b)) |
51 | #define ULONG_CMP_LT(a, b) (ULONG_MAX / 2 < (a) - (b)) | 53 | #define ULONG_CMP_LT(a, b) (ULONG_MAX / 2 < (a) - (b)) |
52 | 54 | ||
@@ -66,7 +68,6 @@ extern void call_rcu_sched(struct rcu_head *head, | |||
66 | extern void synchronize_sched(void); | 68 | extern void synchronize_sched(void); |
67 | extern void rcu_barrier_bh(void); | 69 | extern void rcu_barrier_bh(void); |
68 | extern void rcu_barrier_sched(void); | 70 | extern void rcu_barrier_sched(void); |
69 | extern void synchronize_sched_expedited(void); | ||
70 | extern int sched_expedited_torture_stats(char *page); | 71 | extern int sched_expedited_torture_stats(char *page); |
71 | 72 | ||
72 | static inline void __rcu_read_lock_bh(void) | 73 | static inline void __rcu_read_lock_bh(void) |
@@ -118,7 +119,6 @@ static inline int rcu_preempt_depth(void) | |||
118 | #endif /* #else #ifdef CONFIG_PREEMPT_RCU */ | 119 | #endif /* #else #ifdef CONFIG_PREEMPT_RCU */ |
119 | 120 | ||
120 | /* Internal to kernel */ | 121 | /* Internal to kernel */ |
121 | extern void rcu_init(void); | ||
122 | extern void rcu_sched_qs(int cpu); | 122 | extern void rcu_sched_qs(int cpu); |
123 | extern void rcu_bh_qs(int cpu); | 123 | extern void rcu_bh_qs(int cpu); |
124 | extern void rcu_check_callbacks(int cpu, int user); | 124 | extern void rcu_check_callbacks(int cpu, int user); |
diff --git a/include/linux/rcutiny.h b/include/linux/rcutiny.h index 13877cb93a6..30ebd7c8d87 100644 --- a/include/linux/rcutiny.h +++ b/include/linux/rcutiny.h | |||
@@ -27,7 +27,9 @@ | |||
27 | 27 | ||
28 | #include <linux/cache.h> | 28 | #include <linux/cache.h> |
29 | 29 | ||
30 | #define rcu_init_sched() do { } while (0) | 30 | static inline void rcu_init(void) |
31 | { | ||
32 | } | ||
31 | 33 | ||
32 | #ifdef CONFIG_TINY_RCU | 34 | #ifdef CONFIG_TINY_RCU |
33 | 35 | ||
@@ -58,6 +60,11 @@ static inline void synchronize_rcu_bh_expedited(void) | |||
58 | synchronize_sched(); | 60 | synchronize_sched(); |
59 | } | 61 | } |
60 | 62 | ||
63 | static inline void synchronize_sched_expedited(void) | ||
64 | { | ||
65 | synchronize_sched(); | ||
66 | } | ||
67 | |||
61 | #ifdef CONFIG_TINY_RCU | 68 | #ifdef CONFIG_TINY_RCU |
62 | 69 | ||
63 | static inline void rcu_preempt_note_context_switch(void) | 70 | static inline void rcu_preempt_note_context_switch(void) |
@@ -125,16 +132,12 @@ static inline void rcu_cpu_stall_reset(void) | |||
125 | } | 132 | } |
126 | 133 | ||
127 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | 134 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
128 | |||
129 | extern int rcu_scheduler_active __read_mostly; | 135 | extern int rcu_scheduler_active __read_mostly; |
130 | extern void rcu_scheduler_starting(void); | 136 | extern void rcu_scheduler_starting(void); |
131 | |||
132 | #else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */ | 137 | #else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */ |
133 | |||
134 | static inline void rcu_scheduler_starting(void) | 138 | static inline void rcu_scheduler_starting(void) |
135 | { | 139 | { |
136 | } | 140 | } |
137 | |||
138 | #endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */ | 141 | #endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */ |
139 | 142 | ||
140 | #endif /* __LINUX_RCUTINY_H */ | 143 | #endif /* __LINUX_RCUTINY_H */ |
diff --git a/include/linux/rcutree.h b/include/linux/rcutree.h index 95518e62879..3a933482734 100644 --- a/include/linux/rcutree.h +++ b/include/linux/rcutree.h | |||
@@ -30,6 +30,7 @@ | |||
30 | #ifndef __LINUX_RCUTREE_H | 30 | #ifndef __LINUX_RCUTREE_H |
31 | #define __LINUX_RCUTREE_H | 31 | #define __LINUX_RCUTREE_H |
32 | 32 | ||
33 | extern void rcu_init(void); | ||
33 | extern void rcu_note_context_switch(int cpu); | 34 | extern void rcu_note_context_switch(int cpu); |
34 | extern int rcu_needs_cpu(int cpu); | 35 | extern int rcu_needs_cpu(int cpu); |
35 | extern void rcu_cpu_stall_reset(void); | 36 | extern void rcu_cpu_stall_reset(void); |
@@ -47,6 +48,7 @@ static inline void exit_rcu(void) | |||
47 | #endif /* #else #ifdef CONFIG_TREE_PREEMPT_RCU */ | 48 | #endif /* #else #ifdef CONFIG_TREE_PREEMPT_RCU */ |
48 | 49 | ||
49 | extern void synchronize_rcu_bh(void); | 50 | extern void synchronize_rcu_bh(void); |
51 | extern void synchronize_sched_expedited(void); | ||
50 | extern void synchronize_rcu_expedited(void); | 52 | extern void synchronize_rcu_expedited(void); |
51 | 53 | ||
52 | static inline void synchronize_rcu_bh_expedited(void) | 54 | static inline void synchronize_rcu_bh_expedited(void) |
diff --git a/include/linux/sched.h b/include/linux/sched.h index 883ad10eadb..777cd01e240 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h | |||
@@ -316,6 +316,7 @@ extern int proc_dowatchdog_thresh(struct ctl_table *table, int write, | |||
316 | size_t *lenp, loff_t *ppos); | 316 | size_t *lenp, loff_t *ppos); |
317 | extern unsigned int softlockup_panic; | 317 | extern unsigned int softlockup_panic; |
318 | extern int softlockup_thresh; | 318 | extern int softlockup_thresh; |
319 | void lockup_detector_init(void); | ||
319 | #else | 320 | #else |
320 | static inline void touch_softlockup_watchdog(void) | 321 | static inline void touch_softlockup_watchdog(void) |
321 | { | 322 | { |
@@ -326,6 +327,9 @@ static inline void touch_softlockup_watchdog_sync(void) | |||
326 | static inline void touch_all_softlockup_watchdogs(void) | 327 | static inline void touch_all_softlockup_watchdogs(void) |
327 | { | 328 | { |
328 | } | 329 | } |
330 | static inline void lockup_detector_init(void) | ||
331 | { | ||
332 | } | ||
329 | #endif | 333 | #endif |
330 | 334 | ||
331 | #ifdef CONFIG_DETECT_HUNG_TASK | 335 | #ifdef CONFIG_DETECT_HUNG_TASK |
@@ -1234,6 +1238,9 @@ struct task_struct { | |||
1234 | #ifdef CONFIG_TREE_PREEMPT_RCU | 1238 | #ifdef CONFIG_TREE_PREEMPT_RCU |
1235 | struct rcu_node *rcu_blocked_node; | 1239 | struct rcu_node *rcu_blocked_node; |
1236 | #endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */ | 1240 | #endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */ |
1241 | #ifdef CONFIG_RCU_BOOST | ||
1242 | struct rt_mutex *rcu_boost_mutex; | ||
1243 | #endif /* #ifdef CONFIG_RCU_BOOST */ | ||
1237 | 1244 | ||
1238 | #if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) | 1245 | #if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) |
1239 | struct sched_info sched_info; | 1246 | struct sched_info sched_info; |
@@ -1766,7 +1773,8 @@ extern void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t * | |||
1766 | #ifdef CONFIG_PREEMPT_RCU | 1773 | #ifdef CONFIG_PREEMPT_RCU |
1767 | 1774 | ||
1768 | #define RCU_READ_UNLOCK_BLOCKED (1 << 0) /* blocked while in RCU read-side. */ | 1775 | #define RCU_READ_UNLOCK_BLOCKED (1 << 0) /* blocked while in RCU read-side. */ |
1769 | #define RCU_READ_UNLOCK_NEED_QS (1 << 1) /* RCU core needs CPU response. */ | 1776 | #define RCU_READ_UNLOCK_BOOSTED (1 << 1) /* boosted while in RCU read-side. */ |
1777 | #define RCU_READ_UNLOCK_NEED_QS (1 << 2) /* RCU core needs CPU response. */ | ||
1770 | 1778 | ||
1771 | static inline void rcu_copy_process(struct task_struct *p) | 1779 | static inline void rcu_copy_process(struct task_struct *p) |
1772 | { | 1780 | { |
@@ -1774,7 +1782,10 @@ static inline void rcu_copy_process(struct task_struct *p) | |||
1774 | p->rcu_read_unlock_special = 0; | 1782 | p->rcu_read_unlock_special = 0; |
1775 | #ifdef CONFIG_TREE_PREEMPT_RCU | 1783 | #ifdef CONFIG_TREE_PREEMPT_RCU |
1776 | p->rcu_blocked_node = NULL; | 1784 | p->rcu_blocked_node = NULL; |
1777 | #endif | 1785 | #endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */ |
1786 | #ifdef CONFIG_RCU_BOOST | ||
1787 | p->rcu_boost_mutex = NULL; | ||
1788 | #endif /* #ifdef CONFIG_RCU_BOOST */ | ||
1778 | INIT_LIST_HEAD(&p->rcu_node_entry); | 1789 | INIT_LIST_HEAD(&p->rcu_node_entry); |
1779 | } | 1790 | } |
1780 | 1791 | ||
diff --git a/include/linux/stacktrace.h b/include/linux/stacktrace.h index 51efbef38fb..25310f1d7f3 100644 --- a/include/linux/stacktrace.h +++ b/include/linux/stacktrace.h | |||
@@ -2,6 +2,7 @@ | |||
2 | #define __LINUX_STACKTRACE_H | 2 | #define __LINUX_STACKTRACE_H |
3 | 3 | ||
4 | struct task_struct; | 4 | struct task_struct; |
5 | struct pt_regs; | ||
5 | 6 | ||
6 | #ifdef CONFIG_STACKTRACE | 7 | #ifdef CONFIG_STACKTRACE |
7 | struct task_struct; | 8 | struct task_struct; |
@@ -13,7 +14,8 @@ struct stack_trace { | |||
13 | }; | 14 | }; |
14 | 15 | ||
15 | extern void save_stack_trace(struct stack_trace *trace); | 16 | extern void save_stack_trace(struct stack_trace *trace); |
16 | extern void save_stack_trace_bp(struct stack_trace *trace, unsigned long bp); | 17 | extern void save_stack_trace_regs(struct stack_trace *trace, |
18 | struct pt_regs *regs); | ||
17 | extern void save_stack_trace_tsk(struct task_struct *tsk, | 19 | extern void save_stack_trace_tsk(struct task_struct *tsk, |
18 | struct stack_trace *trace); | 20 | struct stack_trace *trace); |
19 | 21 | ||
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h index cacc27a0e28..18cd0684fc4 100644 --- a/include/linux/syscalls.h +++ b/include/linux/syscalls.h | |||
@@ -127,8 +127,6 @@ extern struct trace_event_functions exit_syscall_print_funcs; | |||
127 | #define SYSCALL_TRACE_ENTER_EVENT(sname) \ | 127 | #define SYSCALL_TRACE_ENTER_EVENT(sname) \ |
128 | static struct syscall_metadata \ | 128 | static struct syscall_metadata \ |
129 | __attribute__((__aligned__(4))) __syscall_meta_##sname; \ | 129 | __attribute__((__aligned__(4))) __syscall_meta_##sname; \ |
130 | static struct ftrace_event_call \ | ||
131 | __attribute__((__aligned__(4))) event_enter_##sname; \ | ||
132 | static struct ftrace_event_call __used \ | 130 | static struct ftrace_event_call __used \ |
133 | __attribute__((__aligned__(4))) \ | 131 | __attribute__((__aligned__(4))) \ |
134 | __attribute__((section("_ftrace_events"))) \ | 132 | __attribute__((section("_ftrace_events"))) \ |
@@ -137,13 +135,12 @@ extern struct trace_event_functions exit_syscall_print_funcs; | |||
137 | .class = &event_class_syscall_enter, \ | 135 | .class = &event_class_syscall_enter, \ |
138 | .event.funcs = &enter_syscall_print_funcs, \ | 136 | .event.funcs = &enter_syscall_print_funcs, \ |
139 | .data = (void *)&__syscall_meta_##sname,\ | 137 | .data = (void *)&__syscall_meta_##sname,\ |
140 | } | 138 | }; \ |
139 | __TRACE_EVENT_FLAGS(enter_##sname, TRACE_EVENT_FL_CAP_ANY) | ||
141 | 140 | ||
142 | #define SYSCALL_TRACE_EXIT_EVENT(sname) \ | 141 | #define SYSCALL_TRACE_EXIT_EVENT(sname) \ |
143 | static struct syscall_metadata \ | 142 | static struct syscall_metadata \ |
144 | __attribute__((__aligned__(4))) __syscall_meta_##sname; \ | 143 | __attribute__((__aligned__(4))) __syscall_meta_##sname; \ |
145 | static struct ftrace_event_call \ | ||
146 | __attribute__((__aligned__(4))) event_exit_##sname; \ | ||
147 | static struct ftrace_event_call __used \ | 144 | static struct ftrace_event_call __used \ |
148 | __attribute__((__aligned__(4))) \ | 145 | __attribute__((__aligned__(4))) \ |
149 | __attribute__((section("_ftrace_events"))) \ | 146 | __attribute__((section("_ftrace_events"))) \ |
@@ -152,7 +149,8 @@ extern struct trace_event_functions exit_syscall_print_funcs; | |||
152 | .class = &event_class_syscall_exit, \ | 149 | .class = &event_class_syscall_exit, \ |
153 | .event.funcs = &exit_syscall_print_funcs, \ | 150 | .event.funcs = &exit_syscall_print_funcs, \ |
154 | .data = (void *)&__syscall_meta_##sname,\ | 151 | .data = (void *)&__syscall_meta_##sname,\ |
155 | } | 152 | }; \ |
153 | __TRACE_EVENT_FLAGS(exit_##sname, TRACE_EVENT_FL_CAP_ANY) | ||
156 | 154 | ||
157 | #define SYSCALL_METADATA(sname, nb) \ | 155 | #define SYSCALL_METADATA(sname, nb) \ |
158 | SYSCALL_TRACE_ENTER_EVENT(sname); \ | 156 | SYSCALL_TRACE_ENTER_EVENT(sname); \ |
diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h index a4a90b6726c..d3e4f87e95c 100644 --- a/include/linux/tracepoint.h +++ b/include/linux/tracepoint.h | |||
@@ -106,6 +106,7 @@ static inline void tracepoint_update_probe_range(struct tracepoint *begin, | |||
106 | 106 | ||
107 | #define TP_PROTO(args...) args | 107 | #define TP_PROTO(args...) args |
108 | #define TP_ARGS(args...) args | 108 | #define TP_ARGS(args...) args |
109 | #define TP_CONDITION(args...) args | ||
109 | 110 | ||
110 | #ifdef CONFIG_TRACEPOINTS | 111 | #ifdef CONFIG_TRACEPOINTS |
111 | 112 | ||
@@ -119,12 +120,14 @@ static inline void tracepoint_update_probe_range(struct tracepoint *begin, | |||
119 | * as "(void *, void)". The DECLARE_TRACE_NOARGS() will pass in just | 120 | * as "(void *, void)". The DECLARE_TRACE_NOARGS() will pass in just |
120 | * "void *data", where as the DECLARE_TRACE() will pass in "void *data, proto". | 121 | * "void *data", where as the DECLARE_TRACE() will pass in "void *data, proto". |
121 | */ | 122 | */ |
122 | #define __DO_TRACE(tp, proto, args) \ | 123 | #define __DO_TRACE(tp, proto, args, cond) \ |
123 | do { \ | 124 | do { \ |
124 | struct tracepoint_func *it_func_ptr; \ | 125 | struct tracepoint_func *it_func_ptr; \ |
125 | void *it_func; \ | 126 | void *it_func; \ |
126 | void *__data; \ | 127 | void *__data; \ |
127 | \ | 128 | \ |
129 | if (!(cond)) \ | ||
130 | return; \ | ||
128 | rcu_read_lock_sched_notrace(); \ | 131 | rcu_read_lock_sched_notrace(); \ |
129 | it_func_ptr = rcu_dereference_sched((tp)->funcs); \ | 132 | it_func_ptr = rcu_dereference_sched((tp)->funcs); \ |
130 | if (it_func_ptr) { \ | 133 | if (it_func_ptr) { \ |
@@ -142,7 +145,7 @@ static inline void tracepoint_update_probe_range(struct tracepoint *begin, | |||
142 | * not add unwanted padding between the beginning of the section and the | 145 | * not add unwanted padding between the beginning of the section and the |
143 | * structure. Force alignment to the same alignment as the section start. | 146 | * structure. Force alignment to the same alignment as the section start. |
144 | */ | 147 | */ |
145 | #define __DECLARE_TRACE(name, proto, args, data_proto, data_args) \ | 148 | #define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \ |
146 | extern struct tracepoint __tracepoint_##name; \ | 149 | extern struct tracepoint __tracepoint_##name; \ |
147 | static inline void trace_##name(proto) \ | 150 | static inline void trace_##name(proto) \ |
148 | { \ | 151 | { \ |
@@ -151,7 +154,8 @@ static inline void tracepoint_update_probe_range(struct tracepoint *begin, | |||
151 | do_trace: \ | 154 | do_trace: \ |
152 | __DO_TRACE(&__tracepoint_##name, \ | 155 | __DO_TRACE(&__tracepoint_##name, \ |
153 | TP_PROTO(data_proto), \ | 156 | TP_PROTO(data_proto), \ |
154 | TP_ARGS(data_args)); \ | 157 | TP_ARGS(data_args), \ |
158 | TP_CONDITION(cond)); \ | ||
155 | } \ | 159 | } \ |
156 | static inline int \ | 160 | static inline int \ |
157 | register_trace_##name(void (*probe)(data_proto), void *data) \ | 161 | register_trace_##name(void (*probe)(data_proto), void *data) \ |
@@ -186,7 +190,7 @@ do_trace: \ | |||
186 | EXPORT_SYMBOL(__tracepoint_##name) | 190 | EXPORT_SYMBOL(__tracepoint_##name) |
187 | 191 | ||
188 | #else /* !CONFIG_TRACEPOINTS */ | 192 | #else /* !CONFIG_TRACEPOINTS */ |
189 | #define __DECLARE_TRACE(name, proto, args, data_proto, data_args) \ | 193 | #define __DECLARE_TRACE(name, proto, args, cond, data_proto, data_args) \ |
190 | static inline void trace_##name(proto) \ | 194 | static inline void trace_##name(proto) \ |
191 | { } \ | 195 | { } \ |
192 | static inline int \ | 196 | static inline int \ |
@@ -227,13 +231,20 @@ do_trace: \ | |||
227 | * "void *__data, proto" as the callback prototype. | 231 | * "void *__data, proto" as the callback prototype. |
228 | */ | 232 | */ |
229 | #define DECLARE_TRACE_NOARGS(name) \ | 233 | #define DECLARE_TRACE_NOARGS(name) \ |
230 | __DECLARE_TRACE(name, void, , void *__data, __data) | 234 | __DECLARE_TRACE(name, void, , 1, void *__data, __data) |
231 | 235 | ||
232 | #define DECLARE_TRACE(name, proto, args) \ | 236 | #define DECLARE_TRACE(name, proto, args) \ |
233 | __DECLARE_TRACE(name, PARAMS(proto), PARAMS(args), \ | 237 | __DECLARE_TRACE(name, PARAMS(proto), PARAMS(args), 1, \ |
234 | PARAMS(void *__data, proto), \ | 238 | PARAMS(void *__data, proto), \ |
235 | PARAMS(__data, args)) | 239 | PARAMS(__data, args)) |
236 | 240 | ||
241 | #define DECLARE_TRACE_CONDITION(name, proto, args, cond) \ | ||
242 | __DECLARE_TRACE(name, PARAMS(proto), PARAMS(args), PARAMS(cond), \ | ||
243 | PARAMS(void *__data, proto), \ | ||
244 | PARAMS(__data, args)) | ||
245 | |||
246 | #define TRACE_EVENT_FLAGS(event, flag) | ||
247 | |||
237 | #endif /* DECLARE_TRACE */ | 248 | #endif /* DECLARE_TRACE */ |
238 | 249 | ||
239 | #ifndef TRACE_EVENT | 250 | #ifndef TRACE_EVENT |
@@ -347,11 +358,21 @@ do_trace: \ | |||
347 | DECLARE_TRACE(name, PARAMS(proto), PARAMS(args)) | 358 | DECLARE_TRACE(name, PARAMS(proto), PARAMS(args)) |
348 | #define DEFINE_EVENT_PRINT(template, name, proto, args, print) \ | 359 | #define DEFINE_EVENT_PRINT(template, name, proto, args, print) \ |
349 | DECLARE_TRACE(name, PARAMS(proto), PARAMS(args)) | 360 | DECLARE_TRACE(name, PARAMS(proto), PARAMS(args)) |
361 | #define DEFINE_EVENT_CONDITION(template, name, proto, \ | ||
362 | args, cond) \ | ||
363 | DECLARE_TRACE_CONDITION(name, PARAMS(proto), \ | ||
364 | PARAMS(args), PARAMS(cond)) | ||
350 | 365 | ||
351 | #define TRACE_EVENT(name, proto, args, struct, assign, print) \ | 366 | #define TRACE_EVENT(name, proto, args, struct, assign, print) \ |
352 | DECLARE_TRACE(name, PARAMS(proto), PARAMS(args)) | 367 | DECLARE_TRACE(name, PARAMS(proto), PARAMS(args)) |
353 | #define TRACE_EVENT_FN(name, proto, args, struct, \ | 368 | #define TRACE_EVENT_FN(name, proto, args, struct, \ |
354 | assign, print, reg, unreg) \ | 369 | assign, print, reg, unreg) \ |
355 | DECLARE_TRACE(name, PARAMS(proto), PARAMS(args)) | 370 | DECLARE_TRACE(name, PARAMS(proto), PARAMS(args)) |
371 | #define TRACE_EVENT_CONDITION(name, proto, args, cond, \ | ||
372 | struct, assign, print) \ | ||
373 | DECLARE_TRACE_CONDITION(name, PARAMS(proto), \ | ||
374 | PARAMS(args), PARAMS(cond)) | ||
375 | |||
376 | #define TRACE_EVENT_FLAGS(event, flag) | ||
356 | 377 | ||
357 | #endif /* ifdef TRACE_EVENT (see note above) */ | 378 | #endif /* ifdef TRACE_EVENT (see note above) */ |
diff --git a/include/trace/define_trace.h b/include/trace/define_trace.h index 1dfab540151..b0b4eb24d59 100644 --- a/include/trace/define_trace.h +++ b/include/trace/define_trace.h | |||
@@ -26,6 +26,15 @@ | |||
26 | #define TRACE_EVENT(name, proto, args, tstruct, assign, print) \ | 26 | #define TRACE_EVENT(name, proto, args, tstruct, assign, print) \ |
27 | DEFINE_TRACE(name) | 27 | DEFINE_TRACE(name) |
28 | 28 | ||
29 | #undef TRACE_EVENT_CONDITION | ||
30 | #define TRACE_EVENT_CONDITION(name, proto, args, cond, tstruct, assign, print) \ | ||
31 | TRACE_EVENT(name, \ | ||
32 | PARAMS(proto), \ | ||
33 | PARAMS(args), \ | ||
34 | PARAMS(tstruct), \ | ||
35 | PARAMS(assign), \ | ||
36 | PARAMS(print)) | ||
37 | |||
29 | #undef TRACE_EVENT_FN | 38 | #undef TRACE_EVENT_FN |
30 | #define TRACE_EVENT_FN(name, proto, args, tstruct, \ | 39 | #define TRACE_EVENT_FN(name, proto, args, tstruct, \ |
31 | assign, print, reg, unreg) \ | 40 | assign, print, reg, unreg) \ |
@@ -39,6 +48,10 @@ | |||
39 | #define DEFINE_EVENT_PRINT(template, name, proto, args, print) \ | 48 | #define DEFINE_EVENT_PRINT(template, name, proto, args, print) \ |
40 | DEFINE_TRACE(name) | 49 | DEFINE_TRACE(name) |
41 | 50 | ||
51 | #undef DEFINE_EVENT_CONDITION | ||
52 | #define DEFINE_EVENT_CONDITION(template, name, proto, args, cond) \ | ||
53 | DEFINE_EVENT(template, name, PARAMS(proto), PARAMS(args)) | ||
54 | |||
42 | #undef DECLARE_TRACE | 55 | #undef DECLARE_TRACE |
43 | #define DECLARE_TRACE(name, proto, args) \ | 56 | #define DECLARE_TRACE(name, proto, args) \ |
44 | DEFINE_TRACE(name) | 57 | DEFINE_TRACE(name) |
@@ -75,9 +88,11 @@ | |||
75 | 88 | ||
76 | #undef TRACE_EVENT | 89 | #undef TRACE_EVENT |
77 | #undef TRACE_EVENT_FN | 90 | #undef TRACE_EVENT_FN |
91 | #undef TRACE_EVENT_CONDITION | ||
78 | #undef DECLARE_EVENT_CLASS | 92 | #undef DECLARE_EVENT_CLASS |
79 | #undef DEFINE_EVENT | 93 | #undef DEFINE_EVENT |
80 | #undef DEFINE_EVENT_PRINT | 94 | #undef DEFINE_EVENT_PRINT |
95 | #undef DEFINE_EVENT_CONDITION | ||
81 | #undef TRACE_HEADER_MULTI_READ | 96 | #undef TRACE_HEADER_MULTI_READ |
82 | #undef DECLARE_TRACE | 97 | #undef DECLARE_TRACE |
83 | 98 | ||
diff --git a/include/trace/events/power.h b/include/trace/events/power.h index 286784d69b8..1bcc2a8c00e 100644 --- a/include/trace/events/power.h +++ b/include/trace/events/power.h | |||
@@ -7,16 +7,67 @@ | |||
7 | #include <linux/ktime.h> | 7 | #include <linux/ktime.h> |
8 | #include <linux/tracepoint.h> | 8 | #include <linux/tracepoint.h> |
9 | 9 | ||
10 | #ifndef _TRACE_POWER_ENUM_ | 10 | DECLARE_EVENT_CLASS(cpu, |
11 | #define _TRACE_POWER_ENUM_ | 11 | |
12 | enum { | 12 | TP_PROTO(unsigned int state, unsigned int cpu_id), |
13 | POWER_NONE = 0, | 13 | |
14 | POWER_CSTATE = 1, /* C-State */ | 14 | TP_ARGS(state, cpu_id), |
15 | POWER_PSTATE = 2, /* Fequency change or DVFS */ | 15 | |
16 | POWER_SSTATE = 3, /* Suspend */ | 16 | TP_STRUCT__entry( |
17 | }; | 17 | __field( u32, state ) |
18 | __field( u32, cpu_id ) | ||
19 | ), | ||
20 | |||
21 | TP_fast_assign( | ||
22 | __entry->state = state; | ||
23 | __entry->cpu_id = cpu_id; | ||
24 | ), | ||
25 | |||
26 | TP_printk("state=%lu cpu_id=%lu", (unsigned long)__entry->state, | ||
27 | (unsigned long)__entry->cpu_id) | ||
28 | ); | ||
29 | |||
30 | DEFINE_EVENT(cpu, cpu_idle, | ||
31 | |||
32 | TP_PROTO(unsigned int state, unsigned int cpu_id), | ||
33 | |||
34 | TP_ARGS(state, cpu_id) | ||
35 | ); | ||
36 | |||
37 | /* This file can get included multiple times, TRACE_HEADER_MULTI_READ at top */ | ||
38 | #ifndef _PWR_EVENT_AVOID_DOUBLE_DEFINING | ||
39 | #define _PWR_EVENT_AVOID_DOUBLE_DEFINING | ||
40 | |||
41 | #define PWR_EVENT_EXIT -1 | ||
18 | #endif | 42 | #endif |
19 | 43 | ||
44 | DEFINE_EVENT(cpu, cpu_frequency, | ||
45 | |||
46 | TP_PROTO(unsigned int frequency, unsigned int cpu_id), | ||
47 | |||
48 | TP_ARGS(frequency, cpu_id) | ||
49 | ); | ||
50 | |||
51 | TRACE_EVENT(machine_suspend, | ||
52 | |||
53 | TP_PROTO(unsigned int state), | ||
54 | |||
55 | TP_ARGS(state), | ||
56 | |||
57 | TP_STRUCT__entry( | ||
58 | __field( u32, state ) | ||
59 | ), | ||
60 | |||
61 | TP_fast_assign( | ||
62 | __entry->state = state; | ||
63 | ), | ||
64 | |||
65 | TP_printk("state=%lu", (unsigned long)__entry->state) | ||
66 | ); | ||
67 | |||
68 | /* This code will be removed after deprecation time exceeded (2.6.41) */ | ||
69 | #ifdef CONFIG_EVENT_POWER_TRACING_DEPRECATED | ||
70 | |||
20 | /* | 71 | /* |
21 | * The power events are used for cpuidle & suspend (power_start, power_end) | 72 | * The power events are used for cpuidle & suspend (power_start, power_end) |
22 | * and for cpufreq (power_frequency) | 73 | * and for cpufreq (power_frequency) |
@@ -75,6 +126,36 @@ TRACE_EVENT(power_end, | |||
75 | 126 | ||
76 | ); | 127 | ); |
77 | 128 | ||
129 | /* Deprecated dummy functions must be protected against multi-declartion */ | ||
130 | #ifndef _PWR_EVENT_AVOID_DOUBLE_DEFINING_DEPRECATED | ||
131 | #define _PWR_EVENT_AVOID_DOUBLE_DEFINING_DEPRECATED | ||
132 | |||
133 | enum { | ||
134 | POWER_NONE = 0, | ||
135 | POWER_CSTATE = 1, | ||
136 | POWER_PSTATE = 2, | ||
137 | }; | ||
138 | #endif /* _PWR_EVENT_AVOID_DOUBLE_DEFINING_DEPRECATED */ | ||
139 | |||
140 | #else /* CONFIG_EVENT_POWER_TRACING_DEPRECATED */ | ||
141 | |||
142 | #ifndef _PWR_EVENT_AVOID_DOUBLE_DEFINING_DEPRECATED | ||
143 | #define _PWR_EVENT_AVOID_DOUBLE_DEFINING_DEPRECATED | ||
144 | enum { | ||
145 | POWER_NONE = 0, | ||
146 | POWER_CSTATE = 1, | ||
147 | POWER_PSTATE = 2, | ||
148 | }; | ||
149 | |||
150 | /* These dummy declaration have to be ripped out when the deprecated | ||
151 | events get removed */ | ||
152 | static inline void trace_power_start(u64 type, u64 state, u64 cpuid) {}; | ||
153 | static inline void trace_power_end(u64 cpuid) {}; | ||
154 | static inline void trace_power_frequency(u64 type, u64 state, u64 cpuid) {}; | ||
155 | #endif /* _PWR_EVENT_AVOID_DOUBLE_DEFINING_DEPRECATED */ | ||
156 | |||
157 | #endif /* CONFIG_EVENT_POWER_TRACING_DEPRECATED */ | ||
158 | |||
78 | /* | 159 | /* |
79 | * The clock events are used for clock enable/disable and for | 160 | * The clock events are used for clock enable/disable and for |
80 | * clock rate change | 161 | * clock rate change |
@@ -153,7 +234,6 @@ DEFINE_EVENT(power_domain, power_domain_target, | |||
153 | 234 | ||
154 | TP_ARGS(name, state, cpu_id) | 235 | TP_ARGS(name, state, cpu_id) |
155 | ); | 236 | ); |
156 | |||
157 | #endif /* _TRACE_POWER_H */ | 237 | #endif /* _TRACE_POWER_H */ |
158 | 238 | ||
159 | /* This part must be outside protection */ | 239 | /* This part must be outside protection */ |
diff --git a/include/trace/events/syscalls.h b/include/trace/events/syscalls.h index fb726ac7cae..5a4c04a75b3 100644 --- a/include/trace/events/syscalls.h +++ b/include/trace/events/syscalls.h | |||
@@ -40,6 +40,8 @@ TRACE_EVENT_FN(sys_enter, | |||
40 | syscall_regfunc, syscall_unregfunc | 40 | syscall_regfunc, syscall_unregfunc |
41 | ); | 41 | ); |
42 | 42 | ||
43 | TRACE_EVENT_FLAGS(sys_enter, TRACE_EVENT_FL_CAP_ANY) | ||
44 | |||
43 | TRACE_EVENT_FN(sys_exit, | 45 | TRACE_EVENT_FN(sys_exit, |
44 | 46 | ||
45 | TP_PROTO(struct pt_regs *regs, long ret), | 47 | TP_PROTO(struct pt_regs *regs, long ret), |
@@ -62,6 +64,8 @@ TRACE_EVENT_FN(sys_exit, | |||
62 | syscall_regfunc, syscall_unregfunc | 64 | syscall_regfunc, syscall_unregfunc |
63 | ); | 65 | ); |
64 | 66 | ||
67 | TRACE_EVENT_FLAGS(sys_exit, TRACE_EVENT_FL_CAP_ANY) | ||
68 | |||
65 | #endif /* CONFIG_HAVE_SYSCALL_TRACEPOINTS */ | 69 | #endif /* CONFIG_HAVE_SYSCALL_TRACEPOINTS */ |
66 | 70 | ||
67 | #endif /* _TRACE_EVENTS_SYSCALLS_H */ | 71 | #endif /* _TRACE_EVENTS_SYSCALLS_H */ |
diff --git a/include/trace/ftrace.h b/include/trace/ftrace.h index a9377c0083a..e16610c208c 100644 --- a/include/trace/ftrace.h +++ b/include/trace/ftrace.h | |||
@@ -82,6 +82,10 @@ | |||
82 | TRACE_EVENT(name, PARAMS(proto), PARAMS(args), \ | 82 | TRACE_EVENT(name, PARAMS(proto), PARAMS(args), \ |
83 | PARAMS(tstruct), PARAMS(assign), PARAMS(print)) \ | 83 | PARAMS(tstruct), PARAMS(assign), PARAMS(print)) \ |
84 | 84 | ||
85 | #undef TRACE_EVENT_FLAGS | ||
86 | #define TRACE_EVENT_FLAGS(name, value) \ | ||
87 | __TRACE_EVENT_FLAGS(name, value) | ||
88 | |||
85 | #include TRACE_INCLUDE(TRACE_INCLUDE_FILE) | 89 | #include TRACE_INCLUDE(TRACE_INCLUDE_FILE) |
86 | 90 | ||
87 | 91 | ||
@@ -129,6 +133,9 @@ | |||
129 | #define DEFINE_EVENT_PRINT(template, name, proto, args, print) \ | 133 | #define DEFINE_EVENT_PRINT(template, name, proto, args, print) \ |
130 | DEFINE_EVENT(template, name, PARAMS(proto), PARAMS(args)) | 134 | DEFINE_EVENT(template, name, PARAMS(proto), PARAMS(args)) |
131 | 135 | ||
136 | #undef TRACE_EVENT_FLAGS | ||
137 | #define TRACE_EVENT_FLAGS(event, flag) | ||
138 | |||
132 | #include TRACE_INCLUDE(TRACE_INCLUDE_FILE) | 139 | #include TRACE_INCLUDE(TRACE_INCLUDE_FILE) |
133 | 140 | ||
134 | /* | 141 | /* |
@@ -289,13 +296,19 @@ static struct trace_event_functions ftrace_event_type_funcs_##call = { \ | |||
289 | 296 | ||
290 | #undef __array | 297 | #undef __array |
291 | #define __array(type, item, len) \ | 298 | #define __array(type, item, len) \ |
292 | BUILD_BUG_ON(len > MAX_FILTER_STR_VAL); \ | 299 | do { \ |
293 | ret = trace_define_field(event_call, #type "[" #len "]", #item, \ | 300 | mutex_lock(&event_storage_mutex); \ |
301 | BUILD_BUG_ON(len > MAX_FILTER_STR_VAL); \ | ||
302 | snprintf(event_storage, sizeof(event_storage), \ | ||
303 | "%s[%d]", #type, len); \ | ||
304 | ret = trace_define_field(event_call, event_storage, #item, \ | ||
294 | offsetof(typeof(field), item), \ | 305 | offsetof(typeof(field), item), \ |
295 | sizeof(field.item), \ | 306 | sizeof(field.item), \ |
296 | is_signed_type(type), FILTER_OTHER); \ | 307 | is_signed_type(type), FILTER_OTHER); \ |
297 | if (ret) \ | 308 | mutex_unlock(&event_storage_mutex); \ |
298 | return ret; | 309 | if (ret) \ |
310 | return ret; \ | ||
311 | } while (0); | ||
299 | 312 | ||
300 | #undef __dynamic_array | 313 | #undef __dynamic_array |
301 | #define __dynamic_array(type, item, len) \ | 314 | #define __dynamic_array(type, item, len) \ |
diff --git a/init/Kconfig b/init/Kconfig index 62b71491dde..8dfd094e687 100644 --- a/init/Kconfig +++ b/init/Kconfig | |||
@@ -393,7 +393,6 @@ config PREEMPT_RCU | |||
393 | 393 | ||
394 | config RCU_TRACE | 394 | config RCU_TRACE |
395 | bool "Enable tracing for RCU" | 395 | bool "Enable tracing for RCU" |
396 | depends on TREE_RCU || TREE_PREEMPT_RCU | ||
397 | help | 396 | help |
398 | This option provides tracing in RCU which presents stats | 397 | This option provides tracing in RCU which presents stats |
399 | in debugfs for debugging RCU implementation. | 398 | in debugfs for debugging RCU implementation. |
@@ -459,6 +458,60 @@ config TREE_RCU_TRACE | |||
459 | TREE_PREEMPT_RCU implementations, permitting Makefile to | 458 | TREE_PREEMPT_RCU implementations, permitting Makefile to |
460 | trivially select kernel/rcutree_trace.c. | 459 | trivially select kernel/rcutree_trace.c. |
461 | 460 | ||
461 | config RCU_BOOST | ||
462 | bool "Enable RCU priority boosting" | ||
463 | depends on RT_MUTEXES && TINY_PREEMPT_RCU | ||
464 | default n | ||
465 | help | ||
466 | This option boosts the priority of preempted RCU readers that | ||
467 | block the current preemptible RCU grace period for too long. | ||
468 | This option also prevents heavy loads from blocking RCU | ||
469 | callback invocation for all flavors of RCU. | ||
470 | |||
471 | Say Y here if you are working with real-time apps or heavy loads | ||
472 | Say N here if you are unsure. | ||
473 | |||
474 | config RCU_BOOST_PRIO | ||
475 | int "Real-time priority to boost RCU readers to" | ||
476 | range 1 99 | ||
477 | depends on RCU_BOOST | ||
478 | default 1 | ||
479 | help | ||
480 | This option specifies the real-time priority to which preempted | ||
481 | RCU readers are to be boosted. If you are working with CPU-bound | ||
482 | real-time applications, you should specify a priority higher then | ||
483 | the highest-priority CPU-bound application. | ||
484 | |||
485 | Specify the real-time priority, or take the default if unsure. | ||
486 | |||
487 | config RCU_BOOST_DELAY | ||
488 | int "Milliseconds to delay boosting after RCU grace-period start" | ||
489 | range 0 3000 | ||
490 | depends on RCU_BOOST | ||
491 | default 500 | ||
492 | help | ||
493 | This option specifies the time to wait after the beginning of | ||
494 | a given grace period before priority-boosting preempted RCU | ||
495 | readers blocking that grace period. Note that any RCU reader | ||
496 | blocking an expedited RCU grace period is boosted immediately. | ||
497 | |||
498 | Accept the default if unsure. | ||
499 | |||
500 | config SRCU_SYNCHRONIZE_DELAY | ||
501 | int "Microseconds to delay before waiting for readers" | ||
502 | range 0 20 | ||
503 | default 10 | ||
504 | help | ||
505 | This option controls how long SRCU delays before entering its | ||
506 | loop waiting on SRCU readers. The purpose of this loop is | ||
507 | to avoid the unconditional context-switch penalty that would | ||
508 | otherwise be incurred if there was an active SRCU reader, | ||
509 | in a manner similar to adaptive locking schemes. This should | ||
510 | be set to be a bit longer than the common-case SRCU read-side | ||
511 | critical-section overhead. | ||
512 | |||
513 | Accept the default if unsure. | ||
514 | |||
462 | endmenu # "RCU Subsystem" | 515 | endmenu # "RCU Subsystem" |
463 | 516 | ||
464 | config IKCONFIG | 517 | config IKCONFIG |
diff --git a/init/main.c b/init/main.c index 8646401f7a0..ea51770c017 100644 --- a/init/main.c +++ b/init/main.c | |||
@@ -67,6 +67,7 @@ | |||
67 | #include <linux/sfi.h> | 67 | #include <linux/sfi.h> |
68 | #include <linux/shmem_fs.h> | 68 | #include <linux/shmem_fs.h> |
69 | #include <linux/slab.h> | 69 | #include <linux/slab.h> |
70 | #include <linux/perf_event.h> | ||
70 | 71 | ||
71 | #include <asm/io.h> | 72 | #include <asm/io.h> |
72 | #include <asm/bugs.h> | 73 | #include <asm/bugs.h> |
@@ -603,6 +604,8 @@ asmlinkage void __init start_kernel(void) | |||
603 | "enabled *very* early, fixing it\n"); | 604 | "enabled *very* early, fixing it\n"); |
604 | local_irq_disable(); | 605 | local_irq_disable(); |
605 | } | 606 | } |
607 | idr_init_cache(); | ||
608 | perf_event_init(); | ||
606 | rcu_init(); | 609 | rcu_init(); |
607 | radix_tree_init(); | 610 | radix_tree_init(); |
608 | /* init some links before init_ISA_irqs() */ | 611 | /* init some links before init_ISA_irqs() */ |
@@ -658,7 +661,6 @@ asmlinkage void __init start_kernel(void) | |||
658 | enable_debug_pagealloc(); | 661 | enable_debug_pagealloc(); |
659 | kmemleak_init(); | 662 | kmemleak_init(); |
660 | debug_objects_mem_init(); | 663 | debug_objects_mem_init(); |
661 | idr_init_cache(); | ||
662 | setup_per_cpu_pageset(); | 664 | setup_per_cpu_pageset(); |
663 | numa_policy_init(); | 665 | numa_policy_init(); |
664 | if (late_time_init) | 666 | if (late_time_init) |
@@ -882,6 +884,7 @@ static int __init kernel_init(void * unused) | |||
882 | smp_prepare_cpus(setup_max_cpus); | 884 | smp_prepare_cpus(setup_max_cpus); |
883 | 885 | ||
884 | do_pre_smp_initcalls(); | 886 | do_pre_smp_initcalls(); |
887 | lockup_detector_init(); | ||
885 | 888 | ||
886 | smp_init(); | 889 | smp_init(); |
887 | sched_init_smp(); | 890 | sched_init_smp(); |
diff --git a/kernel/futex.c b/kernel/futex.c index 40a8777a27d..3019b92e691 100644 --- a/kernel/futex.c +++ b/kernel/futex.c | |||
@@ -69,6 +69,14 @@ int __read_mostly futex_cmpxchg_enabled; | |||
69 | #define FUTEX_HASHBITS (CONFIG_BASE_SMALL ? 4 : 8) | 69 | #define FUTEX_HASHBITS (CONFIG_BASE_SMALL ? 4 : 8) |
70 | 70 | ||
71 | /* | 71 | /* |
72 | * Futex flags used to encode options to functions and preserve them across | ||
73 | * restarts. | ||
74 | */ | ||
75 | #define FLAGS_SHARED 0x01 | ||
76 | #define FLAGS_CLOCKRT 0x02 | ||
77 | #define FLAGS_HAS_TIMEOUT 0x04 | ||
78 | |||
79 | /* | ||
72 | * Priority Inheritance state: | 80 | * Priority Inheritance state: |
73 | */ | 81 | */ |
74 | struct futex_pi_state { | 82 | struct futex_pi_state { |
@@ -123,6 +131,12 @@ struct futex_q { | |||
123 | u32 bitset; | 131 | u32 bitset; |
124 | }; | 132 | }; |
125 | 133 | ||
134 | static const struct futex_q futex_q_init = { | ||
135 | /* list gets initialized in queue_me()*/ | ||
136 | .key = FUTEX_KEY_INIT, | ||
137 | .bitset = FUTEX_BITSET_MATCH_ANY | ||
138 | }; | ||
139 | |||
126 | /* | 140 | /* |
127 | * Hash buckets are shared by all the futex_keys that hash to the same | 141 | * Hash buckets are shared by all the futex_keys that hash to the same |
128 | * location. Each key may have multiple futex_q structures, one for each task | 142 | * location. Each key may have multiple futex_q structures, one for each task |
@@ -283,8 +297,7 @@ again: | |||
283 | return 0; | 297 | return 0; |
284 | } | 298 | } |
285 | 299 | ||
286 | static inline | 300 | static inline void put_futex_key(union futex_key *key) |
287 | void put_futex_key(int fshared, union futex_key *key) | ||
288 | { | 301 | { |
289 | drop_futex_key_refs(key); | 302 | drop_futex_key_refs(key); |
290 | } | 303 | } |
@@ -870,7 +883,8 @@ double_unlock_hb(struct futex_hash_bucket *hb1, struct futex_hash_bucket *hb2) | |||
870 | /* | 883 | /* |
871 | * Wake up waiters matching bitset queued on this futex (uaddr). | 884 | * Wake up waiters matching bitset queued on this futex (uaddr). |
872 | */ | 885 | */ |
873 | static int futex_wake(u32 __user *uaddr, int fshared, int nr_wake, u32 bitset) | 886 | static int |
887 | futex_wake(u32 __user *uaddr, unsigned int flags, int nr_wake, u32 bitset) | ||
874 | { | 888 | { |
875 | struct futex_hash_bucket *hb; | 889 | struct futex_hash_bucket *hb; |
876 | struct futex_q *this, *next; | 890 | struct futex_q *this, *next; |
@@ -881,7 +895,7 @@ static int futex_wake(u32 __user *uaddr, int fshared, int nr_wake, u32 bitset) | |||
881 | if (!bitset) | 895 | if (!bitset) |
882 | return -EINVAL; | 896 | return -EINVAL; |
883 | 897 | ||
884 | ret = get_futex_key(uaddr, fshared, &key); | 898 | ret = get_futex_key(uaddr, flags & FLAGS_SHARED, &key); |
885 | if (unlikely(ret != 0)) | 899 | if (unlikely(ret != 0)) |
886 | goto out; | 900 | goto out; |
887 | 901 | ||
@@ -907,7 +921,7 @@ static int futex_wake(u32 __user *uaddr, int fshared, int nr_wake, u32 bitset) | |||
907 | } | 921 | } |
908 | 922 | ||
909 | spin_unlock(&hb->lock); | 923 | spin_unlock(&hb->lock); |
910 | put_futex_key(fshared, &key); | 924 | put_futex_key(&key); |
911 | out: | 925 | out: |
912 | return ret; | 926 | return ret; |
913 | } | 927 | } |
@@ -917,7 +931,7 @@ out: | |||
917 | * to this virtual address: | 931 | * to this virtual address: |
918 | */ | 932 | */ |
919 | static int | 933 | static int |
920 | futex_wake_op(u32 __user *uaddr1, int fshared, u32 __user *uaddr2, | 934 | futex_wake_op(u32 __user *uaddr1, unsigned int flags, u32 __user *uaddr2, |
921 | int nr_wake, int nr_wake2, int op) | 935 | int nr_wake, int nr_wake2, int op) |
922 | { | 936 | { |
923 | union futex_key key1 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT; | 937 | union futex_key key1 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT; |
@@ -927,10 +941,10 @@ futex_wake_op(u32 __user *uaddr1, int fshared, u32 __user *uaddr2, | |||
927 | int ret, op_ret; | 941 | int ret, op_ret; |
928 | 942 | ||
929 | retry: | 943 | retry: |
930 | ret = get_futex_key(uaddr1, fshared, &key1); | 944 | ret = get_futex_key(uaddr1, flags & FLAGS_SHARED, &key1); |
931 | if (unlikely(ret != 0)) | 945 | if (unlikely(ret != 0)) |
932 | goto out; | 946 | goto out; |
933 | ret = get_futex_key(uaddr2, fshared, &key2); | 947 | ret = get_futex_key(uaddr2, flags & FLAGS_SHARED, &key2); |
934 | if (unlikely(ret != 0)) | 948 | if (unlikely(ret != 0)) |
935 | goto out_put_key1; | 949 | goto out_put_key1; |
936 | 950 | ||
@@ -962,11 +976,11 @@ retry_private: | |||
962 | if (ret) | 976 | if (ret) |
963 | goto out_put_keys; | 977 | goto out_put_keys; |
964 | 978 | ||
965 | if (!fshared) | 979 | if (!(flags & FLAGS_SHARED)) |
966 | goto retry_private; | 980 | goto retry_private; |
967 | 981 | ||
968 | put_futex_key(fshared, &key2); | 982 | put_futex_key(&key2); |
969 | put_futex_key(fshared, &key1); | 983 | put_futex_key(&key1); |
970 | goto retry; | 984 | goto retry; |
971 | } | 985 | } |
972 | 986 | ||
@@ -996,9 +1010,9 @@ retry_private: | |||
996 | 1010 | ||
997 | double_unlock_hb(hb1, hb2); | 1011 | double_unlock_hb(hb1, hb2); |
998 | out_put_keys: | 1012 | out_put_keys: |
999 | put_futex_key(fshared, &key2); | 1013 | put_futex_key(&key2); |
1000 | out_put_key1: | 1014 | out_put_key1: |
1001 | put_futex_key(fshared, &key1); | 1015 | put_futex_key(&key1); |
1002 | out: | 1016 | out: |
1003 | return ret; | 1017 | return ret; |
1004 | } | 1018 | } |
@@ -1133,13 +1147,13 @@ static int futex_proxy_trylock_atomic(u32 __user *pifutex, | |||
1133 | /** | 1147 | /** |
1134 | * futex_requeue() - Requeue waiters from uaddr1 to uaddr2 | 1148 | * futex_requeue() - Requeue waiters from uaddr1 to uaddr2 |
1135 | * @uaddr1: source futex user address | 1149 | * @uaddr1: source futex user address |
1136 | * @fshared: 0 for a PROCESS_PRIVATE futex, 1 for PROCESS_SHARED | 1150 | * @flags: futex flags (FLAGS_SHARED, etc.) |
1137 | * @uaddr2: target futex user address | 1151 | * @uaddr2: target futex user address |
1138 | * @nr_wake: number of waiters to wake (must be 1 for requeue_pi) | 1152 | * @nr_wake: number of waiters to wake (must be 1 for requeue_pi) |
1139 | * @nr_requeue: number of waiters to requeue (0-INT_MAX) | 1153 | * @nr_requeue: number of waiters to requeue (0-INT_MAX) |
1140 | * @cmpval: @uaddr1 expected value (or %NULL) | 1154 | * @cmpval: @uaddr1 expected value (or %NULL) |
1141 | * @requeue_pi: if we are attempting to requeue from a non-pi futex to a | 1155 | * @requeue_pi: if we are attempting to requeue from a non-pi futex to a |
1142 | * pi futex (pi to pi requeue is not supported) | 1156 | * pi futex (pi to pi requeue is not supported) |
1143 | * | 1157 | * |
1144 | * Requeue waiters on uaddr1 to uaddr2. In the requeue_pi case, try to acquire | 1158 | * Requeue waiters on uaddr1 to uaddr2. In the requeue_pi case, try to acquire |
1145 | * uaddr2 atomically on behalf of the top waiter. | 1159 | * uaddr2 atomically on behalf of the top waiter. |
@@ -1148,9 +1162,9 @@ static int futex_proxy_trylock_atomic(u32 __user *pifutex, | |||
1148 | * >=0 - on success, the number of tasks requeued or woken | 1162 | * >=0 - on success, the number of tasks requeued or woken |
1149 | * <0 - on error | 1163 | * <0 - on error |
1150 | */ | 1164 | */ |
1151 | static int futex_requeue(u32 __user *uaddr1, int fshared, u32 __user *uaddr2, | 1165 | static int futex_requeue(u32 __user *uaddr1, unsigned int flags, |
1152 | int nr_wake, int nr_requeue, u32 *cmpval, | 1166 | u32 __user *uaddr2, int nr_wake, int nr_requeue, |
1153 | int requeue_pi) | 1167 | u32 *cmpval, int requeue_pi) |
1154 | { | 1168 | { |
1155 | union futex_key key1 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT; | 1169 | union futex_key key1 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT; |
1156 | int drop_count = 0, task_count = 0, ret; | 1170 | int drop_count = 0, task_count = 0, ret; |
@@ -1191,10 +1205,10 @@ retry: | |||
1191 | pi_state = NULL; | 1205 | pi_state = NULL; |
1192 | } | 1206 | } |
1193 | 1207 | ||
1194 | ret = get_futex_key(uaddr1, fshared, &key1); | 1208 | ret = get_futex_key(uaddr1, flags & FLAGS_SHARED, &key1); |
1195 | if (unlikely(ret != 0)) | 1209 | if (unlikely(ret != 0)) |
1196 | goto out; | 1210 | goto out; |
1197 | ret = get_futex_key(uaddr2, fshared, &key2); | 1211 | ret = get_futex_key(uaddr2, flags & FLAGS_SHARED, &key2); |
1198 | if (unlikely(ret != 0)) | 1212 | if (unlikely(ret != 0)) |
1199 | goto out_put_key1; | 1213 | goto out_put_key1; |
1200 | 1214 | ||
@@ -1216,11 +1230,11 @@ retry_private: | |||
1216 | if (ret) | 1230 | if (ret) |
1217 | goto out_put_keys; | 1231 | goto out_put_keys; |
1218 | 1232 | ||
1219 | if (!fshared) | 1233 | if (!(flags & FLAGS_SHARED)) |
1220 | goto retry_private; | 1234 | goto retry_private; |
1221 | 1235 | ||
1222 | put_futex_key(fshared, &key2); | 1236 | put_futex_key(&key2); |
1223 | put_futex_key(fshared, &key1); | 1237 | put_futex_key(&key1); |
1224 | goto retry; | 1238 | goto retry; |
1225 | } | 1239 | } |
1226 | if (curval != *cmpval) { | 1240 | if (curval != *cmpval) { |
@@ -1260,8 +1274,8 @@ retry_private: | |||
1260 | break; | 1274 | break; |
1261 | case -EFAULT: | 1275 | case -EFAULT: |
1262 | double_unlock_hb(hb1, hb2); | 1276 | double_unlock_hb(hb1, hb2); |
1263 | put_futex_key(fshared, &key2); | 1277 | put_futex_key(&key2); |
1264 | put_futex_key(fshared, &key1); | 1278 | put_futex_key(&key1); |
1265 | ret = fault_in_user_writeable(uaddr2); | 1279 | ret = fault_in_user_writeable(uaddr2); |
1266 | if (!ret) | 1280 | if (!ret) |
1267 | goto retry; | 1281 | goto retry; |
@@ -1269,8 +1283,8 @@ retry_private: | |||
1269 | case -EAGAIN: | 1283 | case -EAGAIN: |
1270 | /* The owner was exiting, try again. */ | 1284 | /* The owner was exiting, try again. */ |
1271 | double_unlock_hb(hb1, hb2); | 1285 | double_unlock_hb(hb1, hb2); |
1272 | put_futex_key(fshared, &key2); | 1286 | put_futex_key(&key2); |
1273 | put_futex_key(fshared, &key1); | 1287 | put_futex_key(&key1); |
1274 | cond_resched(); | 1288 | cond_resched(); |
1275 | goto retry; | 1289 | goto retry; |
1276 | default: | 1290 | default: |
@@ -1352,9 +1366,9 @@ out_unlock: | |||
1352 | drop_futex_key_refs(&key1); | 1366 | drop_futex_key_refs(&key1); |
1353 | 1367 | ||
1354 | out_put_keys: | 1368 | out_put_keys: |
1355 | put_futex_key(fshared, &key2); | 1369 | put_futex_key(&key2); |
1356 | out_put_key1: | 1370 | out_put_key1: |
1357 | put_futex_key(fshared, &key1); | 1371 | put_futex_key(&key1); |
1358 | out: | 1372 | out: |
1359 | if (pi_state != NULL) | 1373 | if (pi_state != NULL) |
1360 | free_pi_state(pi_state); | 1374 | free_pi_state(pi_state); |
@@ -1494,7 +1508,7 @@ static void unqueue_me_pi(struct futex_q *q) | |||
1494 | * private futexes. | 1508 | * private futexes. |
1495 | */ | 1509 | */ |
1496 | static int fixup_pi_state_owner(u32 __user *uaddr, struct futex_q *q, | 1510 | static int fixup_pi_state_owner(u32 __user *uaddr, struct futex_q *q, |
1497 | struct task_struct *newowner, int fshared) | 1511 | struct task_struct *newowner) |
1498 | { | 1512 | { |
1499 | u32 newtid = task_pid_vnr(newowner) | FUTEX_WAITERS; | 1513 | u32 newtid = task_pid_vnr(newowner) | FUTEX_WAITERS; |
1500 | struct futex_pi_state *pi_state = q->pi_state; | 1514 | struct futex_pi_state *pi_state = q->pi_state; |
@@ -1587,20 +1601,11 @@ handle_fault: | |||
1587 | goto retry; | 1601 | goto retry; |
1588 | } | 1602 | } |
1589 | 1603 | ||
1590 | /* | ||
1591 | * In case we must use restart_block to restart a futex_wait, | ||
1592 | * we encode in the 'flags' shared capability | ||
1593 | */ | ||
1594 | #define FLAGS_SHARED 0x01 | ||
1595 | #define FLAGS_CLOCKRT 0x02 | ||
1596 | #define FLAGS_HAS_TIMEOUT 0x04 | ||
1597 | |||
1598 | static long futex_wait_restart(struct restart_block *restart); | 1604 | static long futex_wait_restart(struct restart_block *restart); |
1599 | 1605 | ||
1600 | /** | 1606 | /** |
1601 | * fixup_owner() - Post lock pi_state and corner case management | 1607 | * fixup_owner() - Post lock pi_state and corner case management |
1602 | * @uaddr: user address of the futex | 1608 | * @uaddr: user address of the futex |
1603 | * @fshared: whether the futex is shared (1) or not (0) | ||
1604 | * @q: futex_q (contains pi_state and access to the rt_mutex) | 1609 | * @q: futex_q (contains pi_state and access to the rt_mutex) |
1605 | * @locked: if the attempt to take the rt_mutex succeeded (1) or not (0) | 1610 | * @locked: if the attempt to take the rt_mutex succeeded (1) or not (0) |
1606 | * | 1611 | * |
@@ -1613,8 +1618,7 @@ static long futex_wait_restart(struct restart_block *restart); | |||
1613 | * 0 - success, lock not taken | 1618 | * 0 - success, lock not taken |
1614 | * <0 - on error (-EFAULT) | 1619 | * <0 - on error (-EFAULT) |
1615 | */ | 1620 | */ |
1616 | static int fixup_owner(u32 __user *uaddr, int fshared, struct futex_q *q, | 1621 | static int fixup_owner(u32 __user *uaddr, struct futex_q *q, int locked) |
1617 | int locked) | ||
1618 | { | 1622 | { |
1619 | struct task_struct *owner; | 1623 | struct task_struct *owner; |
1620 | int ret = 0; | 1624 | int ret = 0; |
@@ -1625,7 +1629,7 @@ static int fixup_owner(u32 __user *uaddr, int fshared, struct futex_q *q, | |||
1625 | * did a lock-steal - fix up the PI-state in that case: | 1629 | * did a lock-steal - fix up the PI-state in that case: |
1626 | */ | 1630 | */ |
1627 | if (q->pi_state->owner != current) | 1631 | if (q->pi_state->owner != current) |
1628 | ret = fixup_pi_state_owner(uaddr, q, current, fshared); | 1632 | ret = fixup_pi_state_owner(uaddr, q, current); |
1629 | goto out; | 1633 | goto out; |
1630 | } | 1634 | } |
1631 | 1635 | ||
@@ -1652,7 +1656,7 @@ static int fixup_owner(u32 __user *uaddr, int fshared, struct futex_q *q, | |||
1652 | * lock. Fix the state up. | 1656 | * lock. Fix the state up. |
1653 | */ | 1657 | */ |
1654 | owner = rt_mutex_owner(&q->pi_state->pi_mutex); | 1658 | owner = rt_mutex_owner(&q->pi_state->pi_mutex); |
1655 | ret = fixup_pi_state_owner(uaddr, q, owner, fshared); | 1659 | ret = fixup_pi_state_owner(uaddr, q, owner); |
1656 | goto out; | 1660 | goto out; |
1657 | } | 1661 | } |
1658 | 1662 | ||
@@ -1715,7 +1719,7 @@ static void futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q *q, | |||
1715 | * futex_wait_setup() - Prepare to wait on a futex | 1719 | * futex_wait_setup() - Prepare to wait on a futex |
1716 | * @uaddr: the futex userspace address | 1720 | * @uaddr: the futex userspace address |
1717 | * @val: the expected value | 1721 | * @val: the expected value |
1718 | * @fshared: whether the futex is shared (1) or not (0) | 1722 | * @flags: futex flags (FLAGS_SHARED, etc.) |
1719 | * @q: the associated futex_q | 1723 | * @q: the associated futex_q |
1720 | * @hb: storage for hash_bucket pointer to be returned to caller | 1724 | * @hb: storage for hash_bucket pointer to be returned to caller |
1721 | * | 1725 | * |
@@ -1728,7 +1732,7 @@ static void futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q *q, | |||
1728 | * 0 - uaddr contains val and hb has been locked | 1732 | * 0 - uaddr contains val and hb has been locked |
1729 | * <1 - -EFAULT or -EWOULDBLOCK (uaddr does not contain val) and hb is unlcoked | 1733 | * <1 - -EFAULT or -EWOULDBLOCK (uaddr does not contain val) and hb is unlcoked |
1730 | */ | 1734 | */ |
1731 | static int futex_wait_setup(u32 __user *uaddr, u32 val, int fshared, | 1735 | static int futex_wait_setup(u32 __user *uaddr, u32 val, unsigned int flags, |
1732 | struct futex_q *q, struct futex_hash_bucket **hb) | 1736 | struct futex_q *q, struct futex_hash_bucket **hb) |
1733 | { | 1737 | { |
1734 | u32 uval; | 1738 | u32 uval; |
@@ -1752,8 +1756,7 @@ static int futex_wait_setup(u32 __user *uaddr, u32 val, int fshared, | |||
1752 | * rare, but normal. | 1756 | * rare, but normal. |
1753 | */ | 1757 | */ |
1754 | retry: | 1758 | retry: |
1755 | q->key = FUTEX_KEY_INIT; | 1759 | ret = get_futex_key(uaddr, flags & FLAGS_SHARED, &q->key); |
1756 | ret = get_futex_key(uaddr, fshared, &q->key); | ||
1757 | if (unlikely(ret != 0)) | 1760 | if (unlikely(ret != 0)) |
1758 | return ret; | 1761 | return ret; |
1759 | 1762 | ||
@@ -1769,10 +1772,10 @@ retry_private: | |||
1769 | if (ret) | 1772 | if (ret) |
1770 | goto out; | 1773 | goto out; |
1771 | 1774 | ||
1772 | if (!fshared) | 1775 | if (!(flags & FLAGS_SHARED)) |
1773 | goto retry_private; | 1776 | goto retry_private; |
1774 | 1777 | ||
1775 | put_futex_key(fshared, &q->key); | 1778 | put_futex_key(&q->key); |
1776 | goto retry; | 1779 | goto retry; |
1777 | } | 1780 | } |
1778 | 1781 | ||
@@ -1783,32 +1786,29 @@ retry_private: | |||
1783 | 1786 | ||
1784 | out: | 1787 | out: |
1785 | if (ret) | 1788 | if (ret) |
1786 | put_futex_key(fshared, &q->key); | 1789 | put_futex_key(&q->key); |
1787 | return ret; | 1790 | return ret; |
1788 | } | 1791 | } |
1789 | 1792 | ||
1790 | static int futex_wait(u32 __user *uaddr, int fshared, | 1793 | static int futex_wait(u32 __user *uaddr, unsigned int flags, u32 val, |
1791 | u32 val, ktime_t *abs_time, u32 bitset, int clockrt) | 1794 | ktime_t *abs_time, u32 bitset) |
1792 | { | 1795 | { |
1793 | struct hrtimer_sleeper timeout, *to = NULL; | 1796 | struct hrtimer_sleeper timeout, *to = NULL; |
1794 | struct restart_block *restart; | 1797 | struct restart_block *restart; |
1795 | struct futex_hash_bucket *hb; | 1798 | struct futex_hash_bucket *hb; |
1796 | struct futex_q q; | 1799 | struct futex_q q = futex_q_init; |
1797 | int ret; | 1800 | int ret; |
1798 | 1801 | ||
1799 | if (!bitset) | 1802 | if (!bitset) |
1800 | return -EINVAL; | 1803 | return -EINVAL; |
1801 | |||
1802 | q.pi_state = NULL; | ||
1803 | q.bitset = bitset; | 1804 | q.bitset = bitset; |
1804 | q.rt_waiter = NULL; | ||
1805 | q.requeue_pi_key = NULL; | ||
1806 | 1805 | ||
1807 | if (abs_time) { | 1806 | if (abs_time) { |
1808 | to = &timeout; | 1807 | to = &timeout; |
1809 | 1808 | ||
1810 | hrtimer_init_on_stack(&to->timer, clockrt ? CLOCK_REALTIME : | 1809 | hrtimer_init_on_stack(&to->timer, (flags & FLAGS_CLOCKRT) ? |
1811 | CLOCK_MONOTONIC, HRTIMER_MODE_ABS); | 1810 | CLOCK_REALTIME : CLOCK_MONOTONIC, |
1811 | HRTIMER_MODE_ABS); | ||
1812 | hrtimer_init_sleeper(to, current); | 1812 | hrtimer_init_sleeper(to, current); |
1813 | hrtimer_set_expires_range_ns(&to->timer, *abs_time, | 1813 | hrtimer_set_expires_range_ns(&to->timer, *abs_time, |
1814 | current->timer_slack_ns); | 1814 | current->timer_slack_ns); |
@@ -1819,7 +1819,7 @@ retry: | |||
1819 | * Prepare to wait on uaddr. On success, holds hb lock and increments | 1819 | * Prepare to wait on uaddr. On success, holds hb lock and increments |
1820 | * q.key refs. | 1820 | * q.key refs. |
1821 | */ | 1821 | */ |
1822 | ret = futex_wait_setup(uaddr, val, fshared, &q, &hb); | 1822 | ret = futex_wait_setup(uaddr, val, flags, &q, &hb); |
1823 | if (ret) | 1823 | if (ret) |
1824 | goto out; | 1824 | goto out; |
1825 | 1825 | ||
@@ -1852,12 +1852,7 @@ retry: | |||
1852 | restart->futex.val = val; | 1852 | restart->futex.val = val; |
1853 | restart->futex.time = abs_time->tv64; | 1853 | restart->futex.time = abs_time->tv64; |
1854 | restart->futex.bitset = bitset; | 1854 | restart->futex.bitset = bitset; |
1855 | restart->futex.flags = FLAGS_HAS_TIMEOUT; | 1855 | restart->futex.flags = flags; |
1856 | |||
1857 | if (fshared) | ||
1858 | restart->futex.flags |= FLAGS_SHARED; | ||
1859 | if (clockrt) | ||
1860 | restart->futex.flags |= FLAGS_CLOCKRT; | ||
1861 | 1856 | ||
1862 | ret = -ERESTART_RESTARTBLOCK; | 1857 | ret = -ERESTART_RESTARTBLOCK; |
1863 | 1858 | ||
@@ -1873,7 +1868,6 @@ out: | |||
1873 | static long futex_wait_restart(struct restart_block *restart) | 1868 | static long futex_wait_restart(struct restart_block *restart) |
1874 | { | 1869 | { |
1875 | u32 __user *uaddr = restart->futex.uaddr; | 1870 | u32 __user *uaddr = restart->futex.uaddr; |
1876 | int fshared = 0; | ||
1877 | ktime_t t, *tp = NULL; | 1871 | ktime_t t, *tp = NULL; |
1878 | 1872 | ||
1879 | if (restart->futex.flags & FLAGS_HAS_TIMEOUT) { | 1873 | if (restart->futex.flags & FLAGS_HAS_TIMEOUT) { |
@@ -1881,11 +1875,9 @@ static long futex_wait_restart(struct restart_block *restart) | |||
1881 | tp = &t; | 1875 | tp = &t; |
1882 | } | 1876 | } |
1883 | restart->fn = do_no_restart_syscall; | 1877 | restart->fn = do_no_restart_syscall; |
1884 | if (restart->futex.flags & FLAGS_SHARED) | 1878 | |
1885 | fshared = 1; | 1879 | return (long)futex_wait(uaddr, restart->futex.flags, |
1886 | return (long)futex_wait(uaddr, fshared, restart->futex.val, tp, | 1880 | restart->futex.val, tp, restart->futex.bitset); |
1887 | restart->futex.bitset, | ||
1888 | restart->futex.flags & FLAGS_CLOCKRT); | ||
1889 | } | 1881 | } |
1890 | 1882 | ||
1891 | 1883 | ||
@@ -1895,12 +1887,12 @@ static long futex_wait_restart(struct restart_block *restart) | |||
1895 | * if there are waiters then it will block, it does PI, etc. (Due to | 1887 | * if there are waiters then it will block, it does PI, etc. (Due to |
1896 | * races the kernel might see a 0 value of the futex too.) | 1888 | * races the kernel might see a 0 value of the futex too.) |
1897 | */ | 1889 | */ |
1898 | static int futex_lock_pi(u32 __user *uaddr, int fshared, | 1890 | static int futex_lock_pi(u32 __user *uaddr, unsigned int flags, int detect, |
1899 | int detect, ktime_t *time, int trylock) | 1891 | ktime_t *time, int trylock) |
1900 | { | 1892 | { |
1901 | struct hrtimer_sleeper timeout, *to = NULL; | 1893 | struct hrtimer_sleeper timeout, *to = NULL; |
1902 | struct futex_hash_bucket *hb; | 1894 | struct futex_hash_bucket *hb; |
1903 | struct futex_q q; | 1895 | struct futex_q q = futex_q_init; |
1904 | int res, ret; | 1896 | int res, ret; |
1905 | 1897 | ||
1906 | if (refill_pi_state_cache()) | 1898 | if (refill_pi_state_cache()) |
@@ -1914,12 +1906,8 @@ static int futex_lock_pi(u32 __user *uaddr, int fshared, | |||
1914 | hrtimer_set_expires(&to->timer, *time); | 1906 | hrtimer_set_expires(&to->timer, *time); |
1915 | } | 1907 | } |
1916 | 1908 | ||
1917 | q.pi_state = NULL; | ||
1918 | q.rt_waiter = NULL; | ||
1919 | q.requeue_pi_key = NULL; | ||
1920 | retry: | 1909 | retry: |
1921 | q.key = FUTEX_KEY_INIT; | 1910 | ret = get_futex_key(uaddr, flags & FLAGS_SHARED, &q.key); |
1922 | ret = get_futex_key(uaddr, fshared, &q.key); | ||
1923 | if (unlikely(ret != 0)) | 1911 | if (unlikely(ret != 0)) |
1924 | goto out; | 1912 | goto out; |
1925 | 1913 | ||
@@ -1941,7 +1929,7 @@ retry_private: | |||
1941 | * exit to complete. | 1929 | * exit to complete. |
1942 | */ | 1930 | */ |
1943 | queue_unlock(&q, hb); | 1931 | queue_unlock(&q, hb); |
1944 | put_futex_key(fshared, &q.key); | 1932 | put_futex_key(&q.key); |
1945 | cond_resched(); | 1933 | cond_resched(); |
1946 | goto retry; | 1934 | goto retry; |
1947 | default: | 1935 | default: |
@@ -1971,7 +1959,7 @@ retry_private: | |||
1971 | * Fixup the pi_state owner and possibly acquire the lock if we | 1959 | * Fixup the pi_state owner and possibly acquire the lock if we |
1972 | * haven't already. | 1960 | * haven't already. |
1973 | */ | 1961 | */ |
1974 | res = fixup_owner(uaddr, fshared, &q, !ret); | 1962 | res = fixup_owner(uaddr, &q, !ret); |
1975 | /* | 1963 | /* |
1976 | * If fixup_owner() returned an error, proprogate that. If it acquired | 1964 | * If fixup_owner() returned an error, proprogate that. If it acquired |
1977 | * the lock, clear our -ETIMEDOUT or -EINTR. | 1965 | * the lock, clear our -ETIMEDOUT or -EINTR. |
@@ -1995,7 +1983,7 @@ out_unlock_put_key: | |||
1995 | queue_unlock(&q, hb); | 1983 | queue_unlock(&q, hb); |
1996 | 1984 | ||
1997 | out_put_key: | 1985 | out_put_key: |
1998 | put_futex_key(fshared, &q.key); | 1986 | put_futex_key(&q.key); |
1999 | out: | 1987 | out: |
2000 | if (to) | 1988 | if (to) |
2001 | destroy_hrtimer_on_stack(&to->timer); | 1989 | destroy_hrtimer_on_stack(&to->timer); |
@@ -2008,10 +1996,10 @@ uaddr_faulted: | |||
2008 | if (ret) | 1996 | if (ret) |
2009 | goto out_put_key; | 1997 | goto out_put_key; |
2010 | 1998 | ||
2011 | if (!fshared) | 1999 | if (!(flags & FLAGS_SHARED)) |
2012 | goto retry_private; | 2000 | goto retry_private; |
2013 | 2001 | ||
2014 | put_futex_key(fshared, &q.key); | 2002 | put_futex_key(&q.key); |
2015 | goto retry; | 2003 | goto retry; |
2016 | } | 2004 | } |
2017 | 2005 | ||
@@ -2020,7 +2008,7 @@ uaddr_faulted: | |||
2020 | * This is the in-kernel slowpath: we look up the PI state (if any), | 2008 | * This is the in-kernel slowpath: we look up the PI state (if any), |
2021 | * and do the rt-mutex unlock. | 2009 | * and do the rt-mutex unlock. |
2022 | */ | 2010 | */ |
2023 | static int futex_unlock_pi(u32 __user *uaddr, int fshared) | 2011 | static int futex_unlock_pi(u32 __user *uaddr, unsigned int flags) |
2024 | { | 2012 | { |
2025 | struct futex_hash_bucket *hb; | 2013 | struct futex_hash_bucket *hb; |
2026 | struct futex_q *this, *next; | 2014 | struct futex_q *this, *next; |
@@ -2038,7 +2026,7 @@ retry: | |||
2038 | if ((uval & FUTEX_TID_MASK) != task_pid_vnr(current)) | 2026 | if ((uval & FUTEX_TID_MASK) != task_pid_vnr(current)) |
2039 | return -EPERM; | 2027 | return -EPERM; |
2040 | 2028 | ||
2041 | ret = get_futex_key(uaddr, fshared, &key); | 2029 | ret = get_futex_key(uaddr, flags & FLAGS_SHARED, &key); |
2042 | if (unlikely(ret != 0)) | 2030 | if (unlikely(ret != 0)) |
2043 | goto out; | 2031 | goto out; |
2044 | 2032 | ||
@@ -2093,14 +2081,14 @@ retry: | |||
2093 | 2081 | ||
2094 | out_unlock: | 2082 | out_unlock: |
2095 | spin_unlock(&hb->lock); | 2083 | spin_unlock(&hb->lock); |
2096 | put_futex_key(fshared, &key); | 2084 | put_futex_key(&key); |
2097 | 2085 | ||
2098 | out: | 2086 | out: |
2099 | return ret; | 2087 | return ret; |
2100 | 2088 | ||
2101 | pi_faulted: | 2089 | pi_faulted: |
2102 | spin_unlock(&hb->lock); | 2090 | spin_unlock(&hb->lock); |
2103 | put_futex_key(fshared, &key); | 2091 | put_futex_key(&key); |
2104 | 2092 | ||
2105 | ret = fault_in_user_writeable(uaddr); | 2093 | ret = fault_in_user_writeable(uaddr); |
2106 | if (!ret) | 2094 | if (!ret) |
@@ -2160,7 +2148,7 @@ int handle_early_requeue_pi_wakeup(struct futex_hash_bucket *hb, | |||
2160 | /** | 2148 | /** |
2161 | * futex_wait_requeue_pi() - Wait on uaddr and take uaddr2 | 2149 | * futex_wait_requeue_pi() - Wait on uaddr and take uaddr2 |
2162 | * @uaddr: the futex we initially wait on (non-pi) | 2150 | * @uaddr: the futex we initially wait on (non-pi) |
2163 | * @fshared: whether the futexes are shared (1) or not (0). They must be | 2151 | * @flags: futex flags (FLAGS_SHARED, FLAGS_CLOCKRT, etc.), they must be |
2164 | * the same type, no requeueing from private to shared, etc. | 2152 | * the same type, no requeueing from private to shared, etc. |
2165 | * @val: the expected value of uaddr | 2153 | * @val: the expected value of uaddr |
2166 | * @abs_time: absolute timeout | 2154 | * @abs_time: absolute timeout |
@@ -2198,16 +2186,16 @@ int handle_early_requeue_pi_wakeup(struct futex_hash_bucket *hb, | |||
2198 | * 0 - On success | 2186 | * 0 - On success |
2199 | * <0 - On error | 2187 | * <0 - On error |
2200 | */ | 2188 | */ |
2201 | static int futex_wait_requeue_pi(u32 __user *uaddr, int fshared, | 2189 | static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags, |
2202 | u32 val, ktime_t *abs_time, u32 bitset, | 2190 | u32 val, ktime_t *abs_time, u32 bitset, |
2203 | int clockrt, u32 __user *uaddr2) | 2191 | u32 __user *uaddr2) |
2204 | { | 2192 | { |
2205 | struct hrtimer_sleeper timeout, *to = NULL; | 2193 | struct hrtimer_sleeper timeout, *to = NULL; |
2206 | struct rt_mutex_waiter rt_waiter; | 2194 | struct rt_mutex_waiter rt_waiter; |
2207 | struct rt_mutex *pi_mutex = NULL; | 2195 | struct rt_mutex *pi_mutex = NULL; |
2208 | struct futex_hash_bucket *hb; | 2196 | struct futex_hash_bucket *hb; |
2209 | union futex_key key2; | 2197 | union futex_key key2 = FUTEX_KEY_INIT; |
2210 | struct futex_q q; | 2198 | struct futex_q q = futex_q_init; |
2211 | int res, ret; | 2199 | int res, ret; |
2212 | 2200 | ||
2213 | if (!bitset) | 2201 | if (!bitset) |
@@ -2215,8 +2203,9 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, int fshared, | |||
2215 | 2203 | ||
2216 | if (abs_time) { | 2204 | if (abs_time) { |
2217 | to = &timeout; | 2205 | to = &timeout; |
2218 | hrtimer_init_on_stack(&to->timer, clockrt ? CLOCK_REALTIME : | 2206 | hrtimer_init_on_stack(&to->timer, (flags & FLAGS_CLOCKRT) ? |
2219 | CLOCK_MONOTONIC, HRTIMER_MODE_ABS); | 2207 | CLOCK_REALTIME : CLOCK_MONOTONIC, |
2208 | HRTIMER_MODE_ABS); | ||
2220 | hrtimer_init_sleeper(to, current); | 2209 | hrtimer_init_sleeper(to, current); |
2221 | hrtimer_set_expires_range_ns(&to->timer, *abs_time, | 2210 | hrtimer_set_expires_range_ns(&to->timer, *abs_time, |
2222 | current->timer_slack_ns); | 2211 | current->timer_slack_ns); |
@@ -2229,12 +2218,10 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, int fshared, | |||
2229 | debug_rt_mutex_init_waiter(&rt_waiter); | 2218 | debug_rt_mutex_init_waiter(&rt_waiter); |
2230 | rt_waiter.task = NULL; | 2219 | rt_waiter.task = NULL; |
2231 | 2220 | ||
2232 | key2 = FUTEX_KEY_INIT; | 2221 | ret = get_futex_key(uaddr2, flags & FLAGS_SHARED, &key2); |
2233 | ret = get_futex_key(uaddr2, fshared, &key2); | ||
2234 | if (unlikely(ret != 0)) | 2222 | if (unlikely(ret != 0)) |
2235 | goto out; | 2223 | goto out; |
2236 | 2224 | ||
2237 | q.pi_state = NULL; | ||
2238 | q.bitset = bitset; | 2225 | q.bitset = bitset; |
2239 | q.rt_waiter = &rt_waiter; | 2226 | q.rt_waiter = &rt_waiter; |
2240 | q.requeue_pi_key = &key2; | 2227 | q.requeue_pi_key = &key2; |
@@ -2243,7 +2230,7 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, int fshared, | |||
2243 | * Prepare to wait on uaddr. On success, increments q.key (key1) ref | 2230 | * Prepare to wait on uaddr. On success, increments q.key (key1) ref |
2244 | * count. | 2231 | * count. |
2245 | */ | 2232 | */ |
2246 | ret = futex_wait_setup(uaddr, val, fshared, &q, &hb); | 2233 | ret = futex_wait_setup(uaddr, val, flags, &q, &hb); |
2247 | if (ret) | 2234 | if (ret) |
2248 | goto out_key2; | 2235 | goto out_key2; |
2249 | 2236 | ||
@@ -2273,8 +2260,7 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, int fshared, | |||
2273 | */ | 2260 | */ |
2274 | if (q.pi_state && (q.pi_state->owner != current)) { | 2261 | if (q.pi_state && (q.pi_state->owner != current)) { |
2275 | spin_lock(q.lock_ptr); | 2262 | spin_lock(q.lock_ptr); |
2276 | ret = fixup_pi_state_owner(uaddr2, &q, current, | 2263 | ret = fixup_pi_state_owner(uaddr2, &q, current); |
2277 | fshared); | ||
2278 | spin_unlock(q.lock_ptr); | 2264 | spin_unlock(q.lock_ptr); |
2279 | } | 2265 | } |
2280 | } else { | 2266 | } else { |
@@ -2293,7 +2279,7 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, int fshared, | |||
2293 | * Fixup the pi_state owner and possibly acquire the lock if we | 2279 | * Fixup the pi_state owner and possibly acquire the lock if we |
2294 | * haven't already. | 2280 | * haven't already. |
2295 | */ | 2281 | */ |
2296 | res = fixup_owner(uaddr2, fshared, &q, !ret); | 2282 | res = fixup_owner(uaddr2, &q, !ret); |
2297 | /* | 2283 | /* |
2298 | * If fixup_owner() returned an error, proprogate that. If it | 2284 | * If fixup_owner() returned an error, proprogate that. If it |
2299 | * acquired the lock, clear -ETIMEDOUT or -EINTR. | 2285 | * acquired the lock, clear -ETIMEDOUT or -EINTR. |
@@ -2324,9 +2310,9 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, int fshared, | |||
2324 | } | 2310 | } |
2325 | 2311 | ||
2326 | out_put_keys: | 2312 | out_put_keys: |
2327 | put_futex_key(fshared, &q.key); | 2313 | put_futex_key(&q.key); |
2328 | out_key2: | 2314 | out_key2: |
2329 | put_futex_key(fshared, &key2); | 2315 | put_futex_key(&key2); |
2330 | 2316 | ||
2331 | out: | 2317 | out: |
2332 | if (to) { | 2318 | if (to) { |
@@ -2551,58 +2537,57 @@ void exit_robust_list(struct task_struct *curr) | |||
2551 | long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout, | 2537 | long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout, |
2552 | u32 __user *uaddr2, u32 val2, u32 val3) | 2538 | u32 __user *uaddr2, u32 val2, u32 val3) |
2553 | { | 2539 | { |
2554 | int clockrt, ret = -ENOSYS; | 2540 | int ret = -ENOSYS, cmd = op & FUTEX_CMD_MASK; |
2555 | int cmd = op & FUTEX_CMD_MASK; | 2541 | unsigned int flags = 0; |
2556 | int fshared = 0; | ||
2557 | 2542 | ||
2558 | if (!(op & FUTEX_PRIVATE_FLAG)) | 2543 | if (!(op & FUTEX_PRIVATE_FLAG)) |
2559 | fshared = 1; | 2544 | flags |= FLAGS_SHARED; |
2560 | 2545 | ||
2561 | clockrt = op & FUTEX_CLOCK_REALTIME; | 2546 | if (op & FUTEX_CLOCK_REALTIME) { |
2562 | if (clockrt && cmd != FUTEX_WAIT_BITSET && cmd != FUTEX_WAIT_REQUEUE_PI) | 2547 | flags |= FLAGS_CLOCKRT; |
2563 | return -ENOSYS; | 2548 | if (cmd != FUTEX_WAIT_BITSET && cmd != FUTEX_WAIT_REQUEUE_PI) |
2549 | return -ENOSYS; | ||
2550 | } | ||
2564 | 2551 | ||
2565 | switch (cmd) { | 2552 | switch (cmd) { |
2566 | case FUTEX_WAIT: | 2553 | case FUTEX_WAIT: |
2567 | val3 = FUTEX_BITSET_MATCH_ANY; | 2554 | val3 = FUTEX_BITSET_MATCH_ANY; |
2568 | case FUTEX_WAIT_BITSET: | 2555 | case FUTEX_WAIT_BITSET: |
2569 | ret = futex_wait(uaddr, fshared, val, timeout, val3, clockrt); | 2556 | ret = futex_wait(uaddr, flags, val, timeout, val3); |
2570 | break; | 2557 | break; |
2571 | case FUTEX_WAKE: | 2558 | case FUTEX_WAKE: |
2572 | val3 = FUTEX_BITSET_MATCH_ANY; | 2559 | val3 = FUTEX_BITSET_MATCH_ANY; |
2573 | case FUTEX_WAKE_BITSET: | 2560 | case FUTEX_WAKE_BITSET: |
2574 | ret = futex_wake(uaddr, fshared, val, val3); | 2561 | ret = futex_wake(uaddr, flags, val, val3); |
2575 | break; | 2562 | break; |
2576 | case FUTEX_REQUEUE: | 2563 | case FUTEX_REQUEUE: |
2577 | ret = futex_requeue(uaddr, fshared, uaddr2, val, val2, NULL, 0); | 2564 | ret = futex_requeue(uaddr, flags, uaddr2, val, val2, NULL, 0); |
2578 | break; | 2565 | break; |
2579 | case FUTEX_CMP_REQUEUE: | 2566 | case FUTEX_CMP_REQUEUE: |
2580 | ret = futex_requeue(uaddr, fshared, uaddr2, val, val2, &val3, | 2567 | ret = futex_requeue(uaddr, flags, uaddr2, val, val2, &val3, 0); |
2581 | 0); | ||
2582 | break; | 2568 | break; |
2583 | case FUTEX_WAKE_OP: | 2569 | case FUTEX_WAKE_OP: |
2584 | ret = futex_wake_op(uaddr, fshared, uaddr2, val, val2, val3); | 2570 | ret = futex_wake_op(uaddr, flags, uaddr2, val, val2, val3); |
2585 | break; | 2571 | break; |
2586 | case FUTEX_LOCK_PI: | 2572 | case FUTEX_LOCK_PI: |
2587 | if (futex_cmpxchg_enabled) | 2573 | if (futex_cmpxchg_enabled) |
2588 | ret = futex_lock_pi(uaddr, fshared, val, timeout, 0); | 2574 | ret = futex_lock_pi(uaddr, flags, val, timeout, 0); |
2589 | break; | 2575 | break; |
2590 | case FUTEX_UNLOCK_PI: | 2576 | case FUTEX_UNLOCK_PI: |
2591 | if (futex_cmpxchg_enabled) | 2577 | if (futex_cmpxchg_enabled) |
2592 | ret = futex_unlock_pi(uaddr, fshared); | 2578 | ret = futex_unlock_pi(uaddr, flags); |
2593 | break; | 2579 | break; |
2594 | case FUTEX_TRYLOCK_PI: | 2580 | case FUTEX_TRYLOCK_PI: |
2595 | if (futex_cmpxchg_enabled) | 2581 | if (futex_cmpxchg_enabled) |
2596 | ret = futex_lock_pi(uaddr, fshared, 0, timeout, 1); | 2582 | ret = futex_lock_pi(uaddr, flags, 0, timeout, 1); |
2597 | break; | 2583 | break; |
2598 | case FUTEX_WAIT_REQUEUE_PI: | 2584 | case FUTEX_WAIT_REQUEUE_PI: |
2599 | val3 = FUTEX_BITSET_MATCH_ANY; | 2585 | val3 = FUTEX_BITSET_MATCH_ANY; |
2600 | ret = futex_wait_requeue_pi(uaddr, fshared, val, timeout, val3, | 2586 | ret = futex_wait_requeue_pi(uaddr, flags, val, timeout, val3, |
2601 | clockrt, uaddr2); | 2587 | uaddr2); |
2602 | break; | 2588 | break; |
2603 | case FUTEX_CMP_REQUEUE_PI: | 2589 | case FUTEX_CMP_REQUEUE_PI: |
2604 | ret = futex_requeue(uaddr, fshared, uaddr2, val, val2, &val3, | 2590 | ret = futex_requeue(uaddr, flags, uaddr2, val, val2, &val3, 1); |
2605 | 1); | ||
2606 | break; | 2591 | break; |
2607 | default: | 2592 | default: |
2608 | ret = -ENOSYS; | 2593 | ret = -ENOSYS; |
diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index e5325825aeb..086adf25a55 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c | |||
@@ -641,7 +641,7 @@ int __init init_hw_breakpoint(void) | |||
641 | 641 | ||
642 | constraints_initialized = 1; | 642 | constraints_initialized = 1; |
643 | 643 | ||
644 | perf_pmu_register(&perf_breakpoint); | 644 | perf_pmu_register(&perf_breakpoint, "breakpoint", PERF_TYPE_BREAKPOINT); |
645 | 645 | ||
646 | return register_die_notifier(&hw_breakpoint_exceptions_nb); | 646 | return register_die_notifier(&hw_breakpoint_exceptions_nb); |
647 | 647 | ||
diff --git a/kernel/kprobes.c b/kernel/kprobes.c index 9737a76e106..7663e5df0e6 100644 --- a/kernel/kprobes.c +++ b/kernel/kprobes.c | |||
@@ -354,13 +354,20 @@ static inline int kprobe_aggrprobe(struct kprobe *p) | |||
354 | return p->pre_handler == aggr_pre_handler; | 354 | return p->pre_handler == aggr_pre_handler; |
355 | } | 355 | } |
356 | 356 | ||
357 | /* Return true(!0) if the kprobe is unused */ | ||
358 | static inline int kprobe_unused(struct kprobe *p) | ||
359 | { | ||
360 | return kprobe_aggrprobe(p) && kprobe_disabled(p) && | ||
361 | list_empty(&p->list); | ||
362 | } | ||
363 | |||
357 | /* | 364 | /* |
358 | * Keep all fields in the kprobe consistent | 365 | * Keep all fields in the kprobe consistent |
359 | */ | 366 | */ |
360 | static inline void copy_kprobe(struct kprobe *old_p, struct kprobe *p) | 367 | static inline void copy_kprobe(struct kprobe *ap, struct kprobe *p) |
361 | { | 368 | { |
362 | memcpy(&p->opcode, &old_p->opcode, sizeof(kprobe_opcode_t)); | 369 | memcpy(&p->opcode, &ap->opcode, sizeof(kprobe_opcode_t)); |
363 | memcpy(&p->ainsn, &old_p->ainsn, sizeof(struct arch_specific_insn)); | 370 | memcpy(&p->ainsn, &ap->ainsn, sizeof(struct arch_specific_insn)); |
364 | } | 371 | } |
365 | 372 | ||
366 | #ifdef CONFIG_OPTPROBES | 373 | #ifdef CONFIG_OPTPROBES |
@@ -384,6 +391,17 @@ void __kprobes opt_pre_handler(struct kprobe *p, struct pt_regs *regs) | |||
384 | } | 391 | } |
385 | } | 392 | } |
386 | 393 | ||
394 | /* Free optimized instructions and optimized_kprobe */ | ||
395 | static __kprobes void free_aggr_kprobe(struct kprobe *p) | ||
396 | { | ||
397 | struct optimized_kprobe *op; | ||
398 | |||
399 | op = container_of(p, struct optimized_kprobe, kp); | ||
400 | arch_remove_optimized_kprobe(op); | ||
401 | arch_remove_kprobe(p); | ||
402 | kfree(op); | ||
403 | } | ||
404 | |||
387 | /* Return true(!0) if the kprobe is ready for optimization. */ | 405 | /* Return true(!0) if the kprobe is ready for optimization. */ |
388 | static inline int kprobe_optready(struct kprobe *p) | 406 | static inline int kprobe_optready(struct kprobe *p) |
389 | { | 407 | { |
@@ -397,6 +415,33 @@ static inline int kprobe_optready(struct kprobe *p) | |||
397 | return 0; | 415 | return 0; |
398 | } | 416 | } |
399 | 417 | ||
418 | /* Return true(!0) if the kprobe is disarmed. Note: p must be on hash list */ | ||
419 | static inline int kprobe_disarmed(struct kprobe *p) | ||
420 | { | ||
421 | struct optimized_kprobe *op; | ||
422 | |||
423 | /* If kprobe is not aggr/opt probe, just return kprobe is disabled */ | ||
424 | if (!kprobe_aggrprobe(p)) | ||
425 | return kprobe_disabled(p); | ||
426 | |||
427 | op = container_of(p, struct optimized_kprobe, kp); | ||
428 | |||
429 | return kprobe_disabled(p) && list_empty(&op->list); | ||
430 | } | ||
431 | |||
432 | /* Return true(!0) if the probe is queued on (un)optimizing lists */ | ||
433 | static int __kprobes kprobe_queued(struct kprobe *p) | ||
434 | { | ||
435 | struct optimized_kprobe *op; | ||
436 | |||
437 | if (kprobe_aggrprobe(p)) { | ||
438 | op = container_of(p, struct optimized_kprobe, kp); | ||
439 | if (!list_empty(&op->list)) | ||
440 | return 1; | ||
441 | } | ||
442 | return 0; | ||
443 | } | ||
444 | |||
400 | /* | 445 | /* |
401 | * Return an optimized kprobe whose optimizing code replaces | 446 | * Return an optimized kprobe whose optimizing code replaces |
402 | * instructions including addr (exclude breakpoint). | 447 | * instructions including addr (exclude breakpoint). |
@@ -422,30 +467,23 @@ static struct kprobe *__kprobes get_optimized_kprobe(unsigned long addr) | |||
422 | 467 | ||
423 | /* Optimization staging list, protected by kprobe_mutex */ | 468 | /* Optimization staging list, protected by kprobe_mutex */ |
424 | static LIST_HEAD(optimizing_list); | 469 | static LIST_HEAD(optimizing_list); |
470 | static LIST_HEAD(unoptimizing_list); | ||
425 | 471 | ||
426 | static void kprobe_optimizer(struct work_struct *work); | 472 | static void kprobe_optimizer(struct work_struct *work); |
427 | static DECLARE_DELAYED_WORK(optimizing_work, kprobe_optimizer); | 473 | static DECLARE_DELAYED_WORK(optimizing_work, kprobe_optimizer); |
474 | static DECLARE_COMPLETION(optimizer_comp); | ||
428 | #define OPTIMIZE_DELAY 5 | 475 | #define OPTIMIZE_DELAY 5 |
429 | 476 | ||
430 | /* Kprobe jump optimizer */ | 477 | /* |
431 | static __kprobes void kprobe_optimizer(struct work_struct *work) | 478 | * Optimize (replace a breakpoint with a jump) kprobes listed on |
479 | * optimizing_list. | ||
480 | */ | ||
481 | static __kprobes void do_optimize_kprobes(void) | ||
432 | { | 482 | { |
433 | struct optimized_kprobe *op, *tmp; | 483 | /* Optimization never be done when disarmed */ |
434 | 484 | if (kprobes_all_disarmed || !kprobes_allow_optimization || | |
435 | /* Lock modules while optimizing kprobes */ | 485 | list_empty(&optimizing_list)) |
436 | mutex_lock(&module_mutex); | 486 | return; |
437 | mutex_lock(&kprobe_mutex); | ||
438 | if (kprobes_all_disarmed || !kprobes_allow_optimization) | ||
439 | goto end; | ||
440 | |||
441 | /* | ||
442 | * Wait for quiesence period to ensure all running interrupts | ||
443 | * are done. Because optprobe may modify multiple instructions | ||
444 | * there is a chance that Nth instruction is interrupted. In that | ||
445 | * case, running interrupt can return to 2nd-Nth byte of jump | ||
446 | * instruction. This wait is for avoiding it. | ||
447 | */ | ||
448 | synchronize_sched(); | ||
449 | 487 | ||
450 | /* | 488 | /* |
451 | * The optimization/unoptimization refers online_cpus via | 489 | * The optimization/unoptimization refers online_cpus via |
@@ -459,17 +497,111 @@ static __kprobes void kprobe_optimizer(struct work_struct *work) | |||
459 | */ | 497 | */ |
460 | get_online_cpus(); | 498 | get_online_cpus(); |
461 | mutex_lock(&text_mutex); | 499 | mutex_lock(&text_mutex); |
462 | list_for_each_entry_safe(op, tmp, &optimizing_list, list) { | 500 | arch_optimize_kprobes(&optimizing_list); |
463 | WARN_ON(kprobe_disabled(&op->kp)); | 501 | mutex_unlock(&text_mutex); |
464 | if (arch_optimize_kprobe(op) < 0) | 502 | put_online_cpus(); |
465 | op->kp.flags &= ~KPROBE_FLAG_OPTIMIZED; | 503 | } |
466 | list_del_init(&op->list); | 504 | |
505 | /* | ||
506 | * Unoptimize (replace a jump with a breakpoint and remove the breakpoint | ||
507 | * if need) kprobes listed on unoptimizing_list. | ||
508 | */ | ||
509 | static __kprobes void do_unoptimize_kprobes(struct list_head *free_list) | ||
510 | { | ||
511 | struct optimized_kprobe *op, *tmp; | ||
512 | |||
513 | /* Unoptimization must be done anytime */ | ||
514 | if (list_empty(&unoptimizing_list)) | ||
515 | return; | ||
516 | |||
517 | /* Ditto to do_optimize_kprobes */ | ||
518 | get_online_cpus(); | ||
519 | mutex_lock(&text_mutex); | ||
520 | arch_unoptimize_kprobes(&unoptimizing_list, free_list); | ||
521 | /* Loop free_list for disarming */ | ||
522 | list_for_each_entry_safe(op, tmp, free_list, list) { | ||
523 | /* Disarm probes if marked disabled */ | ||
524 | if (kprobe_disabled(&op->kp)) | ||
525 | arch_disarm_kprobe(&op->kp); | ||
526 | if (kprobe_unused(&op->kp)) { | ||
527 | /* | ||
528 | * Remove unused probes from hash list. After waiting | ||
529 | * for synchronization, these probes are reclaimed. | ||
530 | * (reclaiming is done by do_free_cleaned_kprobes.) | ||
531 | */ | ||
532 | hlist_del_rcu(&op->kp.hlist); | ||
533 | } else | ||
534 | list_del_init(&op->list); | ||
467 | } | 535 | } |
468 | mutex_unlock(&text_mutex); | 536 | mutex_unlock(&text_mutex); |
469 | put_online_cpus(); | 537 | put_online_cpus(); |
470 | end: | 538 | } |
539 | |||
540 | /* Reclaim all kprobes on the free_list */ | ||
541 | static __kprobes void do_free_cleaned_kprobes(struct list_head *free_list) | ||
542 | { | ||
543 | struct optimized_kprobe *op, *tmp; | ||
544 | |||
545 | list_for_each_entry_safe(op, tmp, free_list, list) { | ||
546 | BUG_ON(!kprobe_unused(&op->kp)); | ||
547 | list_del_init(&op->list); | ||
548 | free_aggr_kprobe(&op->kp); | ||
549 | } | ||
550 | } | ||
551 | |||
552 | /* Start optimizer after OPTIMIZE_DELAY passed */ | ||
553 | static __kprobes void kick_kprobe_optimizer(void) | ||
554 | { | ||
555 | if (!delayed_work_pending(&optimizing_work)) | ||
556 | schedule_delayed_work(&optimizing_work, OPTIMIZE_DELAY); | ||
557 | } | ||
558 | |||
559 | /* Kprobe jump optimizer */ | ||
560 | static __kprobes void kprobe_optimizer(struct work_struct *work) | ||
561 | { | ||
562 | LIST_HEAD(free_list); | ||
563 | |||
564 | /* Lock modules while optimizing kprobes */ | ||
565 | mutex_lock(&module_mutex); | ||
566 | mutex_lock(&kprobe_mutex); | ||
567 | |||
568 | /* | ||
569 | * Step 1: Unoptimize kprobes and collect cleaned (unused and disarmed) | ||
570 | * kprobes before waiting for quiesence period. | ||
571 | */ | ||
572 | do_unoptimize_kprobes(&free_list); | ||
573 | |||
574 | /* | ||
575 | * Step 2: Wait for quiesence period to ensure all running interrupts | ||
576 | * are done. Because optprobe may modify multiple instructions | ||
577 | * there is a chance that Nth instruction is interrupted. In that | ||
578 | * case, running interrupt can return to 2nd-Nth byte of jump | ||
579 | * instruction. This wait is for avoiding it. | ||
580 | */ | ||
581 | synchronize_sched(); | ||
582 | |||
583 | /* Step 3: Optimize kprobes after quiesence period */ | ||
584 | do_optimize_kprobes(); | ||
585 | |||
586 | /* Step 4: Free cleaned kprobes after quiesence period */ | ||
587 | do_free_cleaned_kprobes(&free_list); | ||
588 | |||
471 | mutex_unlock(&kprobe_mutex); | 589 | mutex_unlock(&kprobe_mutex); |
472 | mutex_unlock(&module_mutex); | 590 | mutex_unlock(&module_mutex); |
591 | |||
592 | /* Step 5: Kick optimizer again if needed */ | ||
593 | if (!list_empty(&optimizing_list) || !list_empty(&unoptimizing_list)) | ||
594 | kick_kprobe_optimizer(); | ||
595 | else | ||
596 | /* Wake up all waiters */ | ||
597 | complete_all(&optimizer_comp); | ||
598 | } | ||
599 | |||
600 | /* Wait for completing optimization and unoptimization */ | ||
601 | static __kprobes void wait_for_kprobe_optimizer(void) | ||
602 | { | ||
603 | if (delayed_work_pending(&optimizing_work)) | ||
604 | wait_for_completion(&optimizer_comp); | ||
473 | } | 605 | } |
474 | 606 | ||
475 | /* Optimize kprobe if p is ready to be optimized */ | 607 | /* Optimize kprobe if p is ready to be optimized */ |
@@ -495,42 +627,99 @@ static __kprobes void optimize_kprobe(struct kprobe *p) | |||
495 | /* Check if it is already optimized. */ | 627 | /* Check if it is already optimized. */ |
496 | if (op->kp.flags & KPROBE_FLAG_OPTIMIZED) | 628 | if (op->kp.flags & KPROBE_FLAG_OPTIMIZED) |
497 | return; | 629 | return; |
498 | |||
499 | op->kp.flags |= KPROBE_FLAG_OPTIMIZED; | 630 | op->kp.flags |= KPROBE_FLAG_OPTIMIZED; |
500 | list_add(&op->list, &optimizing_list); | 631 | |
501 | if (!delayed_work_pending(&optimizing_work)) | 632 | if (!list_empty(&op->list)) |
502 | schedule_delayed_work(&optimizing_work, OPTIMIZE_DELAY); | 633 | /* This is under unoptimizing. Just dequeue the probe */ |
634 | list_del_init(&op->list); | ||
635 | else { | ||
636 | list_add(&op->list, &optimizing_list); | ||
637 | kick_kprobe_optimizer(); | ||
638 | } | ||
639 | } | ||
640 | |||
641 | /* Short cut to direct unoptimizing */ | ||
642 | static __kprobes void force_unoptimize_kprobe(struct optimized_kprobe *op) | ||
643 | { | ||
644 | get_online_cpus(); | ||
645 | arch_unoptimize_kprobe(op); | ||
646 | put_online_cpus(); | ||
647 | if (kprobe_disabled(&op->kp)) | ||
648 | arch_disarm_kprobe(&op->kp); | ||
503 | } | 649 | } |
504 | 650 | ||
505 | /* Unoptimize a kprobe if p is optimized */ | 651 | /* Unoptimize a kprobe if p is optimized */ |
506 | static __kprobes void unoptimize_kprobe(struct kprobe *p) | 652 | static __kprobes void unoptimize_kprobe(struct kprobe *p, bool force) |
507 | { | 653 | { |
508 | struct optimized_kprobe *op; | 654 | struct optimized_kprobe *op; |
509 | 655 | ||
510 | if ((p->flags & KPROBE_FLAG_OPTIMIZED) && kprobe_aggrprobe(p)) { | 656 | if (!kprobe_aggrprobe(p) || kprobe_disarmed(p)) |
511 | op = container_of(p, struct optimized_kprobe, kp); | 657 | return; /* This is not an optprobe nor optimized */ |
512 | if (!list_empty(&op->list)) | 658 | |
513 | /* Dequeue from the optimization queue */ | 659 | op = container_of(p, struct optimized_kprobe, kp); |
660 | if (!kprobe_optimized(p)) { | ||
661 | /* Unoptimized or unoptimizing case */ | ||
662 | if (force && !list_empty(&op->list)) { | ||
663 | /* | ||
664 | * Only if this is unoptimizing kprobe and forced, | ||
665 | * forcibly unoptimize it. (No need to unoptimize | ||
666 | * unoptimized kprobe again :) | ||
667 | */ | ||
514 | list_del_init(&op->list); | 668 | list_del_init(&op->list); |
515 | else | 669 | force_unoptimize_kprobe(op); |
516 | /* Replace jump with break */ | 670 | } |
517 | arch_unoptimize_kprobe(op); | 671 | return; |
518 | op->kp.flags &= ~KPROBE_FLAG_OPTIMIZED; | 672 | } |
673 | |||
674 | op->kp.flags &= ~KPROBE_FLAG_OPTIMIZED; | ||
675 | if (!list_empty(&op->list)) { | ||
676 | /* Dequeue from the optimization queue */ | ||
677 | list_del_init(&op->list); | ||
678 | return; | ||
679 | } | ||
680 | /* Optimized kprobe case */ | ||
681 | if (force) | ||
682 | /* Forcibly update the code: this is a special case */ | ||
683 | force_unoptimize_kprobe(op); | ||
684 | else { | ||
685 | list_add(&op->list, &unoptimizing_list); | ||
686 | kick_kprobe_optimizer(); | ||
519 | } | 687 | } |
520 | } | 688 | } |
521 | 689 | ||
690 | /* Cancel unoptimizing for reusing */ | ||
691 | static void reuse_unused_kprobe(struct kprobe *ap) | ||
692 | { | ||
693 | struct optimized_kprobe *op; | ||
694 | |||
695 | BUG_ON(!kprobe_unused(ap)); | ||
696 | /* | ||
697 | * Unused kprobe MUST be on the way of delayed unoptimizing (means | ||
698 | * there is still a relative jump) and disabled. | ||
699 | */ | ||
700 | op = container_of(ap, struct optimized_kprobe, kp); | ||
701 | if (unlikely(list_empty(&op->list))) | ||
702 | printk(KERN_WARNING "Warning: found a stray unused " | ||
703 | "aggrprobe@%p\n", ap->addr); | ||
704 | /* Enable the probe again */ | ||
705 | ap->flags &= ~KPROBE_FLAG_DISABLED; | ||
706 | /* Optimize it again (remove from op->list) */ | ||
707 | BUG_ON(!kprobe_optready(ap)); | ||
708 | optimize_kprobe(ap); | ||
709 | } | ||
710 | |||
522 | /* Remove optimized instructions */ | 711 | /* Remove optimized instructions */ |
523 | static void __kprobes kill_optimized_kprobe(struct kprobe *p) | 712 | static void __kprobes kill_optimized_kprobe(struct kprobe *p) |
524 | { | 713 | { |
525 | struct optimized_kprobe *op; | 714 | struct optimized_kprobe *op; |
526 | 715 | ||
527 | op = container_of(p, struct optimized_kprobe, kp); | 716 | op = container_of(p, struct optimized_kprobe, kp); |
528 | if (!list_empty(&op->list)) { | 717 | if (!list_empty(&op->list)) |
529 | /* Dequeue from the optimization queue */ | 718 | /* Dequeue from the (un)optimization queue */ |
530 | list_del_init(&op->list); | 719 | list_del_init(&op->list); |
531 | op->kp.flags &= ~KPROBE_FLAG_OPTIMIZED; | 720 | |
532 | } | 721 | op->kp.flags &= ~KPROBE_FLAG_OPTIMIZED; |
533 | /* Don't unoptimize, because the target code will be freed. */ | 722 | /* Don't touch the code, because it is already freed. */ |
534 | arch_remove_optimized_kprobe(op); | 723 | arch_remove_optimized_kprobe(op); |
535 | } | 724 | } |
536 | 725 | ||
@@ -543,16 +732,6 @@ static __kprobes void prepare_optimized_kprobe(struct kprobe *p) | |||
543 | arch_prepare_optimized_kprobe(op); | 732 | arch_prepare_optimized_kprobe(op); |
544 | } | 733 | } |
545 | 734 | ||
546 | /* Free optimized instructions and optimized_kprobe */ | ||
547 | static __kprobes void free_aggr_kprobe(struct kprobe *p) | ||
548 | { | ||
549 | struct optimized_kprobe *op; | ||
550 | |||
551 | op = container_of(p, struct optimized_kprobe, kp); | ||
552 | arch_remove_optimized_kprobe(op); | ||
553 | kfree(op); | ||
554 | } | ||
555 | |||
556 | /* Allocate new optimized_kprobe and try to prepare optimized instructions */ | 735 | /* Allocate new optimized_kprobe and try to prepare optimized instructions */ |
557 | static __kprobes struct kprobe *alloc_aggr_kprobe(struct kprobe *p) | 736 | static __kprobes struct kprobe *alloc_aggr_kprobe(struct kprobe *p) |
558 | { | 737 | { |
@@ -587,7 +766,8 @@ static __kprobes void try_to_optimize_kprobe(struct kprobe *p) | |||
587 | op = container_of(ap, struct optimized_kprobe, kp); | 766 | op = container_of(ap, struct optimized_kprobe, kp); |
588 | if (!arch_prepared_optinsn(&op->optinsn)) { | 767 | if (!arch_prepared_optinsn(&op->optinsn)) { |
589 | /* If failed to setup optimizing, fallback to kprobe */ | 768 | /* If failed to setup optimizing, fallback to kprobe */ |
590 | free_aggr_kprobe(ap); | 769 | arch_remove_optimized_kprobe(op); |
770 | kfree(op); | ||
591 | return; | 771 | return; |
592 | } | 772 | } |
593 | 773 | ||
@@ -631,21 +811,16 @@ static void __kprobes unoptimize_all_kprobes(void) | |||
631 | return; | 811 | return; |
632 | 812 | ||
633 | kprobes_allow_optimization = false; | 813 | kprobes_allow_optimization = false; |
634 | printk(KERN_INFO "Kprobes globally unoptimized\n"); | ||
635 | get_online_cpus(); /* For avoiding text_mutex deadlock */ | ||
636 | mutex_lock(&text_mutex); | ||
637 | for (i = 0; i < KPROBE_TABLE_SIZE; i++) { | 814 | for (i = 0; i < KPROBE_TABLE_SIZE; i++) { |
638 | head = &kprobe_table[i]; | 815 | head = &kprobe_table[i]; |
639 | hlist_for_each_entry_rcu(p, node, head, hlist) { | 816 | hlist_for_each_entry_rcu(p, node, head, hlist) { |
640 | if (!kprobe_disabled(p)) | 817 | if (!kprobe_disabled(p)) |
641 | unoptimize_kprobe(p); | 818 | unoptimize_kprobe(p, false); |
642 | } | 819 | } |
643 | } | 820 | } |
644 | 821 | /* Wait for unoptimizing completion */ | |
645 | mutex_unlock(&text_mutex); | 822 | wait_for_kprobe_optimizer(); |
646 | put_online_cpus(); | 823 | printk(KERN_INFO "Kprobes globally unoptimized\n"); |
647 | /* Allow all currently running kprobes to complete */ | ||
648 | synchronize_sched(); | ||
649 | } | 824 | } |
650 | 825 | ||
651 | int sysctl_kprobes_optimization; | 826 | int sysctl_kprobes_optimization; |
@@ -669,44 +844,60 @@ int proc_kprobes_optimization_handler(struct ctl_table *table, int write, | |||
669 | } | 844 | } |
670 | #endif /* CONFIG_SYSCTL */ | 845 | #endif /* CONFIG_SYSCTL */ |
671 | 846 | ||
847 | /* Put a breakpoint for a probe. Must be called with text_mutex locked */ | ||
672 | static void __kprobes __arm_kprobe(struct kprobe *p) | 848 | static void __kprobes __arm_kprobe(struct kprobe *p) |
673 | { | 849 | { |
674 | struct kprobe *old_p; | 850 | struct kprobe *_p; |
675 | 851 | ||
676 | /* Check collision with other optimized kprobes */ | 852 | /* Check collision with other optimized kprobes */ |
677 | old_p = get_optimized_kprobe((unsigned long)p->addr); | 853 | _p = get_optimized_kprobe((unsigned long)p->addr); |
678 | if (unlikely(old_p)) | 854 | if (unlikely(_p)) |
679 | unoptimize_kprobe(old_p); /* Fallback to unoptimized kprobe */ | 855 | /* Fallback to unoptimized kprobe */ |
856 | unoptimize_kprobe(_p, true); | ||
680 | 857 | ||
681 | arch_arm_kprobe(p); | 858 | arch_arm_kprobe(p); |
682 | optimize_kprobe(p); /* Try to optimize (add kprobe to a list) */ | 859 | optimize_kprobe(p); /* Try to optimize (add kprobe to a list) */ |
683 | } | 860 | } |
684 | 861 | ||
685 | static void __kprobes __disarm_kprobe(struct kprobe *p) | 862 | /* Remove the breakpoint of a probe. Must be called with text_mutex locked */ |
863 | static void __kprobes __disarm_kprobe(struct kprobe *p, bool reopt) | ||
686 | { | 864 | { |
687 | struct kprobe *old_p; | 865 | struct kprobe *_p; |
688 | 866 | ||
689 | unoptimize_kprobe(p); /* Try to unoptimize */ | 867 | unoptimize_kprobe(p, false); /* Try to unoptimize */ |
690 | arch_disarm_kprobe(p); | ||
691 | 868 | ||
692 | /* If another kprobe was blocked, optimize it. */ | 869 | if (!kprobe_queued(p)) { |
693 | old_p = get_optimized_kprobe((unsigned long)p->addr); | 870 | arch_disarm_kprobe(p); |
694 | if (unlikely(old_p)) | 871 | /* If another kprobe was blocked, optimize it. */ |
695 | optimize_kprobe(old_p); | 872 | _p = get_optimized_kprobe((unsigned long)p->addr); |
873 | if (unlikely(_p) && reopt) | ||
874 | optimize_kprobe(_p); | ||
875 | } | ||
876 | /* TODO: reoptimize others after unoptimized this probe */ | ||
696 | } | 877 | } |
697 | 878 | ||
698 | #else /* !CONFIG_OPTPROBES */ | 879 | #else /* !CONFIG_OPTPROBES */ |
699 | 880 | ||
700 | #define optimize_kprobe(p) do {} while (0) | 881 | #define optimize_kprobe(p) do {} while (0) |
701 | #define unoptimize_kprobe(p) do {} while (0) | 882 | #define unoptimize_kprobe(p, f) do {} while (0) |
702 | #define kill_optimized_kprobe(p) do {} while (0) | 883 | #define kill_optimized_kprobe(p) do {} while (0) |
703 | #define prepare_optimized_kprobe(p) do {} while (0) | 884 | #define prepare_optimized_kprobe(p) do {} while (0) |
704 | #define try_to_optimize_kprobe(p) do {} while (0) | 885 | #define try_to_optimize_kprobe(p) do {} while (0) |
705 | #define __arm_kprobe(p) arch_arm_kprobe(p) | 886 | #define __arm_kprobe(p) arch_arm_kprobe(p) |
706 | #define __disarm_kprobe(p) arch_disarm_kprobe(p) | 887 | #define __disarm_kprobe(p, o) arch_disarm_kprobe(p) |
888 | #define kprobe_disarmed(p) kprobe_disabled(p) | ||
889 | #define wait_for_kprobe_optimizer() do {} while (0) | ||
890 | |||
891 | /* There should be no unused kprobes can be reused without optimization */ | ||
892 | static void reuse_unused_kprobe(struct kprobe *ap) | ||
893 | { | ||
894 | printk(KERN_ERR "Error: There should be no unused kprobe here.\n"); | ||
895 | BUG_ON(kprobe_unused(ap)); | ||
896 | } | ||
707 | 897 | ||
708 | static __kprobes void free_aggr_kprobe(struct kprobe *p) | 898 | static __kprobes void free_aggr_kprobe(struct kprobe *p) |
709 | { | 899 | { |
900 | arch_remove_kprobe(p); | ||
710 | kfree(p); | 901 | kfree(p); |
711 | } | 902 | } |
712 | 903 | ||
@@ -732,11 +923,10 @@ static void __kprobes arm_kprobe(struct kprobe *kp) | |||
732 | /* Disarm a kprobe with text_mutex */ | 923 | /* Disarm a kprobe with text_mutex */ |
733 | static void __kprobes disarm_kprobe(struct kprobe *kp) | 924 | static void __kprobes disarm_kprobe(struct kprobe *kp) |
734 | { | 925 | { |
735 | get_online_cpus(); /* For avoiding text_mutex deadlock */ | 926 | /* Ditto */ |
736 | mutex_lock(&text_mutex); | 927 | mutex_lock(&text_mutex); |
737 | __disarm_kprobe(kp); | 928 | __disarm_kprobe(kp, true); |
738 | mutex_unlock(&text_mutex); | 929 | mutex_unlock(&text_mutex); |
739 | put_online_cpus(); | ||
740 | } | 930 | } |
741 | 931 | ||
742 | /* | 932 | /* |
@@ -942,7 +1132,7 @@ static int __kprobes add_new_kprobe(struct kprobe *ap, struct kprobe *p) | |||
942 | BUG_ON(kprobe_gone(ap) || kprobe_gone(p)); | 1132 | BUG_ON(kprobe_gone(ap) || kprobe_gone(p)); |
943 | 1133 | ||
944 | if (p->break_handler || p->post_handler) | 1134 | if (p->break_handler || p->post_handler) |
945 | unoptimize_kprobe(ap); /* Fall back to normal kprobe */ | 1135 | unoptimize_kprobe(ap, true); /* Fall back to normal kprobe */ |
946 | 1136 | ||
947 | if (p->break_handler) { | 1137 | if (p->break_handler) { |
948 | if (ap->break_handler) | 1138 | if (ap->break_handler) |
@@ -993,19 +1183,21 @@ static void __kprobes init_aggr_kprobe(struct kprobe *ap, struct kprobe *p) | |||
993 | * This is the second or subsequent kprobe at the address - handle | 1183 | * This is the second or subsequent kprobe at the address - handle |
994 | * the intricacies | 1184 | * the intricacies |
995 | */ | 1185 | */ |
996 | static int __kprobes register_aggr_kprobe(struct kprobe *old_p, | 1186 | static int __kprobes register_aggr_kprobe(struct kprobe *orig_p, |
997 | struct kprobe *p) | 1187 | struct kprobe *p) |
998 | { | 1188 | { |
999 | int ret = 0; | 1189 | int ret = 0; |
1000 | struct kprobe *ap = old_p; | 1190 | struct kprobe *ap = orig_p; |
1001 | 1191 | ||
1002 | if (!kprobe_aggrprobe(old_p)) { | 1192 | if (!kprobe_aggrprobe(orig_p)) { |
1003 | /* If old_p is not an aggr_kprobe, create new aggr_kprobe. */ | 1193 | /* If orig_p is not an aggr_kprobe, create new aggr_kprobe. */ |
1004 | ap = alloc_aggr_kprobe(old_p); | 1194 | ap = alloc_aggr_kprobe(orig_p); |
1005 | if (!ap) | 1195 | if (!ap) |
1006 | return -ENOMEM; | 1196 | return -ENOMEM; |
1007 | init_aggr_kprobe(ap, old_p); | 1197 | init_aggr_kprobe(ap, orig_p); |
1008 | } | 1198 | } else if (kprobe_unused(ap)) |
1199 | /* This probe is going to die. Rescue it */ | ||
1200 | reuse_unused_kprobe(ap); | ||
1009 | 1201 | ||
1010 | if (kprobe_gone(ap)) { | 1202 | if (kprobe_gone(ap)) { |
1011 | /* | 1203 | /* |
@@ -1039,23 +1231,6 @@ static int __kprobes register_aggr_kprobe(struct kprobe *old_p, | |||
1039 | return add_new_kprobe(ap, p); | 1231 | return add_new_kprobe(ap, p); |
1040 | } | 1232 | } |
1041 | 1233 | ||
1042 | /* Try to disable aggr_kprobe, and return 1 if succeeded.*/ | ||
1043 | static int __kprobes try_to_disable_aggr_kprobe(struct kprobe *p) | ||
1044 | { | ||
1045 | struct kprobe *kp; | ||
1046 | |||
1047 | list_for_each_entry_rcu(kp, &p->list, list) { | ||
1048 | if (!kprobe_disabled(kp)) | ||
1049 | /* | ||
1050 | * There is an active probe on the list. | ||
1051 | * We can't disable aggr_kprobe. | ||
1052 | */ | ||
1053 | return 0; | ||
1054 | } | ||
1055 | p->flags |= KPROBE_FLAG_DISABLED; | ||
1056 | return 1; | ||
1057 | } | ||
1058 | |||
1059 | static int __kprobes in_kprobes_functions(unsigned long addr) | 1234 | static int __kprobes in_kprobes_functions(unsigned long addr) |
1060 | { | 1235 | { |
1061 | struct kprobe_blackpoint *kb; | 1236 | struct kprobe_blackpoint *kb; |
@@ -1098,34 +1273,33 @@ static kprobe_opcode_t __kprobes *kprobe_addr(struct kprobe *p) | |||
1098 | /* Check passed kprobe is valid and return kprobe in kprobe_table. */ | 1273 | /* Check passed kprobe is valid and return kprobe in kprobe_table. */ |
1099 | static struct kprobe * __kprobes __get_valid_kprobe(struct kprobe *p) | 1274 | static struct kprobe * __kprobes __get_valid_kprobe(struct kprobe *p) |
1100 | { | 1275 | { |
1101 | struct kprobe *old_p, *list_p; | 1276 | struct kprobe *ap, *list_p; |
1102 | 1277 | ||
1103 | old_p = get_kprobe(p->addr); | 1278 | ap = get_kprobe(p->addr); |
1104 | if (unlikely(!old_p)) | 1279 | if (unlikely(!ap)) |
1105 | return NULL; | 1280 | return NULL; |
1106 | 1281 | ||
1107 | if (p != old_p) { | 1282 | if (p != ap) { |
1108 | list_for_each_entry_rcu(list_p, &old_p->list, list) | 1283 | list_for_each_entry_rcu(list_p, &ap->list, list) |
1109 | if (list_p == p) | 1284 | if (list_p == p) |
1110 | /* kprobe p is a valid probe */ | 1285 | /* kprobe p is a valid probe */ |
1111 | goto valid; | 1286 | goto valid; |
1112 | return NULL; | 1287 | return NULL; |
1113 | } | 1288 | } |
1114 | valid: | 1289 | valid: |
1115 | return old_p; | 1290 | return ap; |
1116 | } | 1291 | } |
1117 | 1292 | ||
1118 | /* Return error if the kprobe is being re-registered */ | 1293 | /* Return error if the kprobe is being re-registered */ |
1119 | static inline int check_kprobe_rereg(struct kprobe *p) | 1294 | static inline int check_kprobe_rereg(struct kprobe *p) |
1120 | { | 1295 | { |
1121 | int ret = 0; | 1296 | int ret = 0; |
1122 | struct kprobe *old_p; | ||
1123 | 1297 | ||
1124 | mutex_lock(&kprobe_mutex); | 1298 | mutex_lock(&kprobe_mutex); |
1125 | old_p = __get_valid_kprobe(p); | 1299 | if (__get_valid_kprobe(p)) |
1126 | if (old_p) | ||
1127 | ret = -EINVAL; | 1300 | ret = -EINVAL; |
1128 | mutex_unlock(&kprobe_mutex); | 1301 | mutex_unlock(&kprobe_mutex); |
1302 | |||
1129 | return ret; | 1303 | return ret; |
1130 | } | 1304 | } |
1131 | 1305 | ||
@@ -1229,67 +1403,121 @@ fail_with_jump_label: | |||
1229 | } | 1403 | } |
1230 | EXPORT_SYMBOL_GPL(register_kprobe); | 1404 | EXPORT_SYMBOL_GPL(register_kprobe); |
1231 | 1405 | ||
1406 | /* Check if all probes on the aggrprobe are disabled */ | ||
1407 | static int __kprobes aggr_kprobe_disabled(struct kprobe *ap) | ||
1408 | { | ||
1409 | struct kprobe *kp; | ||
1410 | |||
1411 | list_for_each_entry_rcu(kp, &ap->list, list) | ||
1412 | if (!kprobe_disabled(kp)) | ||
1413 | /* | ||
1414 | * There is an active probe on the list. | ||
1415 | * We can't disable this ap. | ||
1416 | */ | ||
1417 | return 0; | ||
1418 | |||
1419 | return 1; | ||
1420 | } | ||
1421 | |||
1422 | /* Disable one kprobe: Make sure called under kprobe_mutex is locked */ | ||
1423 | static struct kprobe *__kprobes __disable_kprobe(struct kprobe *p) | ||
1424 | { | ||
1425 | struct kprobe *orig_p; | ||
1426 | |||
1427 | /* Get an original kprobe for return */ | ||
1428 | orig_p = __get_valid_kprobe(p); | ||
1429 | if (unlikely(orig_p == NULL)) | ||
1430 | return NULL; | ||
1431 | |||
1432 | if (!kprobe_disabled(p)) { | ||
1433 | /* Disable probe if it is a child probe */ | ||
1434 | if (p != orig_p) | ||
1435 | p->flags |= KPROBE_FLAG_DISABLED; | ||
1436 | |||
1437 | /* Try to disarm and disable this/parent probe */ | ||
1438 | if (p == orig_p || aggr_kprobe_disabled(orig_p)) { | ||
1439 | disarm_kprobe(orig_p); | ||
1440 | orig_p->flags |= KPROBE_FLAG_DISABLED; | ||
1441 | } | ||
1442 | } | ||
1443 | |||
1444 | return orig_p; | ||
1445 | } | ||
1446 | |||
1232 | /* | 1447 | /* |
1233 | * Unregister a kprobe without a scheduler synchronization. | 1448 | * Unregister a kprobe without a scheduler synchronization. |
1234 | */ | 1449 | */ |
1235 | static int __kprobes __unregister_kprobe_top(struct kprobe *p) | 1450 | static int __kprobes __unregister_kprobe_top(struct kprobe *p) |
1236 | { | 1451 | { |
1237 | struct kprobe *old_p, *list_p; | 1452 | struct kprobe *ap, *list_p; |
1238 | 1453 | ||
1239 | old_p = __get_valid_kprobe(p); | 1454 | /* Disable kprobe. This will disarm it if needed. */ |
1240 | if (old_p == NULL) | 1455 | ap = __disable_kprobe(p); |
1456 | if (ap == NULL) | ||
1241 | return -EINVAL; | 1457 | return -EINVAL; |
1242 | 1458 | ||
1243 | if (old_p == p || | 1459 | if (ap == p) |
1244 | (kprobe_aggrprobe(old_p) && | ||
1245 | list_is_singular(&old_p->list))) { | ||
1246 | /* | 1460 | /* |
1247 | * Only probe on the hash list. Disarm only if kprobes are | 1461 | * This probe is an independent(and non-optimized) kprobe |
1248 | * enabled and not gone - otherwise, the breakpoint would | 1462 | * (not an aggrprobe). Remove from the hash list. |
1249 | * already have been removed. We save on flushing icache. | ||
1250 | */ | 1463 | */ |
1251 | if (!kprobes_all_disarmed && !kprobe_disabled(old_p)) | 1464 | goto disarmed; |
1252 | disarm_kprobe(old_p); | 1465 | |
1253 | hlist_del_rcu(&old_p->hlist); | 1466 | /* Following process expects this probe is an aggrprobe */ |
1254 | } else { | 1467 | WARN_ON(!kprobe_aggrprobe(ap)); |
1468 | |||
1469 | if (list_is_singular(&ap->list) && kprobe_disarmed(ap)) | ||
1470 | /* | ||
1471 | * !disarmed could be happen if the probe is under delayed | ||
1472 | * unoptimizing. | ||
1473 | */ | ||
1474 | goto disarmed; | ||
1475 | else { | ||
1476 | /* If disabling probe has special handlers, update aggrprobe */ | ||
1255 | if (p->break_handler && !kprobe_gone(p)) | 1477 | if (p->break_handler && !kprobe_gone(p)) |
1256 | old_p->break_handler = NULL; | 1478 | ap->break_handler = NULL; |
1257 | if (p->post_handler && !kprobe_gone(p)) { | 1479 | if (p->post_handler && !kprobe_gone(p)) { |
1258 | list_for_each_entry_rcu(list_p, &old_p->list, list) { | 1480 | list_for_each_entry_rcu(list_p, &ap->list, list) { |
1259 | if ((list_p != p) && (list_p->post_handler)) | 1481 | if ((list_p != p) && (list_p->post_handler)) |
1260 | goto noclean; | 1482 | goto noclean; |
1261 | } | 1483 | } |
1262 | old_p->post_handler = NULL; | 1484 | ap->post_handler = NULL; |
1263 | } | 1485 | } |
1264 | noclean: | 1486 | noclean: |
1487 | /* | ||
1488 | * Remove from the aggrprobe: this path will do nothing in | ||
1489 | * __unregister_kprobe_bottom(). | ||
1490 | */ | ||
1265 | list_del_rcu(&p->list); | 1491 | list_del_rcu(&p->list); |
1266 | if (!kprobe_disabled(old_p)) { | 1492 | if (!kprobe_disabled(ap) && !kprobes_all_disarmed) |
1267 | try_to_disable_aggr_kprobe(old_p); | 1493 | /* |
1268 | if (!kprobes_all_disarmed) { | 1494 | * Try to optimize this probe again, because post |
1269 | if (kprobe_disabled(old_p)) | 1495 | * handler may have been changed. |
1270 | disarm_kprobe(old_p); | 1496 | */ |
1271 | else | 1497 | optimize_kprobe(ap); |
1272 | /* Try to optimize this probe again */ | ||
1273 | optimize_kprobe(old_p); | ||
1274 | } | ||
1275 | } | ||
1276 | } | 1498 | } |
1277 | return 0; | 1499 | return 0; |
1500 | |||
1501 | disarmed: | ||
1502 | BUG_ON(!kprobe_disarmed(ap)); | ||
1503 | hlist_del_rcu(&ap->hlist); | ||
1504 | return 0; | ||
1278 | } | 1505 | } |
1279 | 1506 | ||
1280 | static void __kprobes __unregister_kprobe_bottom(struct kprobe *p) | 1507 | static void __kprobes __unregister_kprobe_bottom(struct kprobe *p) |
1281 | { | 1508 | { |
1282 | struct kprobe *old_p; | 1509 | struct kprobe *ap; |
1283 | 1510 | ||
1284 | if (list_empty(&p->list)) | 1511 | if (list_empty(&p->list)) |
1512 | /* This is an independent kprobe */ | ||
1285 | arch_remove_kprobe(p); | 1513 | arch_remove_kprobe(p); |
1286 | else if (list_is_singular(&p->list)) { | 1514 | else if (list_is_singular(&p->list)) { |
1287 | /* "p" is the last child of an aggr_kprobe */ | 1515 | /* This is the last child of an aggrprobe */ |
1288 | old_p = list_entry(p->list.next, struct kprobe, list); | 1516 | ap = list_entry(p->list.next, struct kprobe, list); |
1289 | list_del(&p->list); | 1517 | list_del(&p->list); |
1290 | arch_remove_kprobe(old_p); | 1518 | free_aggr_kprobe(ap); |
1291 | free_aggr_kprobe(old_p); | ||
1292 | } | 1519 | } |
1520 | /* Otherwise, do nothing. */ | ||
1293 | } | 1521 | } |
1294 | 1522 | ||
1295 | int __kprobes register_kprobes(struct kprobe **kps, int num) | 1523 | int __kprobes register_kprobes(struct kprobe **kps, int num) |
@@ -1607,29 +1835,13 @@ static void __kprobes kill_kprobe(struct kprobe *p) | |||
1607 | int __kprobes disable_kprobe(struct kprobe *kp) | 1835 | int __kprobes disable_kprobe(struct kprobe *kp) |
1608 | { | 1836 | { |
1609 | int ret = 0; | 1837 | int ret = 0; |
1610 | struct kprobe *p; | ||
1611 | 1838 | ||
1612 | mutex_lock(&kprobe_mutex); | 1839 | mutex_lock(&kprobe_mutex); |
1613 | 1840 | ||
1614 | /* Check whether specified probe is valid. */ | 1841 | /* Disable this kprobe */ |
1615 | p = __get_valid_kprobe(kp); | 1842 | if (__disable_kprobe(kp) == NULL) |
1616 | if (unlikely(p == NULL)) { | ||
1617 | ret = -EINVAL; | 1843 | ret = -EINVAL; |
1618 | goto out; | ||
1619 | } | ||
1620 | 1844 | ||
1621 | /* If the probe is already disabled (or gone), just return */ | ||
1622 | if (kprobe_disabled(kp)) | ||
1623 | goto out; | ||
1624 | |||
1625 | kp->flags |= KPROBE_FLAG_DISABLED; | ||
1626 | if (p != kp) | ||
1627 | /* When kp != p, p is always enabled. */ | ||
1628 | try_to_disable_aggr_kprobe(p); | ||
1629 | |||
1630 | if (!kprobes_all_disarmed && kprobe_disabled(p)) | ||
1631 | disarm_kprobe(p); | ||
1632 | out: | ||
1633 | mutex_unlock(&kprobe_mutex); | 1845 | mutex_unlock(&kprobe_mutex); |
1634 | return ret; | 1846 | return ret; |
1635 | } | 1847 | } |
@@ -1927,36 +2139,27 @@ static void __kprobes disarm_all_kprobes(void) | |||
1927 | mutex_lock(&kprobe_mutex); | 2139 | mutex_lock(&kprobe_mutex); |
1928 | 2140 | ||
1929 | /* If kprobes are already disarmed, just return */ | 2141 | /* If kprobes are already disarmed, just return */ |
1930 | if (kprobes_all_disarmed) | 2142 | if (kprobes_all_disarmed) { |
1931 | goto already_disabled; | 2143 | mutex_unlock(&kprobe_mutex); |
2144 | return; | ||
2145 | } | ||
1932 | 2146 | ||
1933 | kprobes_all_disarmed = true; | 2147 | kprobes_all_disarmed = true; |
1934 | printk(KERN_INFO "Kprobes globally disabled\n"); | 2148 | printk(KERN_INFO "Kprobes globally disabled\n"); |
1935 | 2149 | ||
1936 | /* | ||
1937 | * Here we call get_online_cpus() for avoiding text_mutex deadlock, | ||
1938 | * because disarming may also unoptimize kprobes. | ||
1939 | */ | ||
1940 | get_online_cpus(); | ||
1941 | mutex_lock(&text_mutex); | 2150 | mutex_lock(&text_mutex); |
1942 | for (i = 0; i < KPROBE_TABLE_SIZE; i++) { | 2151 | for (i = 0; i < KPROBE_TABLE_SIZE; i++) { |
1943 | head = &kprobe_table[i]; | 2152 | head = &kprobe_table[i]; |
1944 | hlist_for_each_entry_rcu(p, node, head, hlist) { | 2153 | hlist_for_each_entry_rcu(p, node, head, hlist) { |
1945 | if (!arch_trampoline_kprobe(p) && !kprobe_disabled(p)) | 2154 | if (!arch_trampoline_kprobe(p) && !kprobe_disabled(p)) |
1946 | __disarm_kprobe(p); | 2155 | __disarm_kprobe(p, false); |
1947 | } | 2156 | } |
1948 | } | 2157 | } |
1949 | |||
1950 | mutex_unlock(&text_mutex); | 2158 | mutex_unlock(&text_mutex); |
1951 | put_online_cpus(); | ||
1952 | mutex_unlock(&kprobe_mutex); | 2159 | mutex_unlock(&kprobe_mutex); |
1953 | /* Allow all currently running kprobes to complete */ | ||
1954 | synchronize_sched(); | ||
1955 | return; | ||
1956 | 2160 | ||
1957 | already_disabled: | 2161 | /* Wait for disarming all kprobes by optimizer */ |
1958 | mutex_unlock(&kprobe_mutex); | 2162 | wait_for_kprobe_optimizer(); |
1959 | return; | ||
1960 | } | 2163 | } |
1961 | 2164 | ||
1962 | /* | 2165 | /* |
diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 2870feee81d..11847bf1e8c 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c | |||
@@ -13,6 +13,7 @@ | |||
13 | #include <linux/mm.h> | 13 | #include <linux/mm.h> |
14 | #include <linux/cpu.h> | 14 | #include <linux/cpu.h> |
15 | #include <linux/smp.h> | 15 | #include <linux/smp.h> |
16 | #include <linux/idr.h> | ||
16 | #include <linux/file.h> | 17 | #include <linux/file.h> |
17 | #include <linux/poll.h> | 18 | #include <linux/poll.h> |
18 | #include <linux/slab.h> | 19 | #include <linux/slab.h> |
@@ -21,7 +22,9 @@ | |||
21 | #include <linux/dcache.h> | 22 | #include <linux/dcache.h> |
22 | #include <linux/percpu.h> | 23 | #include <linux/percpu.h> |
23 | #include <linux/ptrace.h> | 24 | #include <linux/ptrace.h> |
25 | #include <linux/reboot.h> | ||
24 | #include <linux/vmstat.h> | 26 | #include <linux/vmstat.h> |
27 | #include <linux/device.h> | ||
25 | #include <linux/vmalloc.h> | 28 | #include <linux/vmalloc.h> |
26 | #include <linux/hardirq.h> | 29 | #include <linux/hardirq.h> |
27 | #include <linux/rculist.h> | 30 | #include <linux/rculist.h> |
@@ -133,6 +136,28 @@ static void unclone_ctx(struct perf_event_context *ctx) | |||
133 | } | 136 | } |
134 | } | 137 | } |
135 | 138 | ||
139 | static u32 perf_event_pid(struct perf_event *event, struct task_struct *p) | ||
140 | { | ||
141 | /* | ||
142 | * only top level events have the pid namespace they were created in | ||
143 | */ | ||
144 | if (event->parent) | ||
145 | event = event->parent; | ||
146 | |||
147 | return task_tgid_nr_ns(p, event->ns); | ||
148 | } | ||
149 | |||
150 | static u32 perf_event_tid(struct perf_event *event, struct task_struct *p) | ||
151 | { | ||
152 | /* | ||
153 | * only top level events have the pid namespace they were created in | ||
154 | */ | ||
155 | if (event->parent) | ||
156 | event = event->parent; | ||
157 | |||
158 | return task_pid_nr_ns(p, event->ns); | ||
159 | } | ||
160 | |||
136 | /* | 161 | /* |
137 | * If we inherit events we want to return the parent event id | 162 | * If we inherit events we want to return the parent event id |
138 | * to userspace. | 163 | * to userspace. |
@@ -312,9 +337,84 @@ list_add_event(struct perf_event *event, struct perf_event_context *ctx) | |||
312 | ctx->nr_stat++; | 337 | ctx->nr_stat++; |
313 | } | 338 | } |
314 | 339 | ||
340 | /* | ||
341 | * Called at perf_event creation and when events are attached/detached from a | ||
342 | * group. | ||
343 | */ | ||
344 | static void perf_event__read_size(struct perf_event *event) | ||
345 | { | ||
346 | int entry = sizeof(u64); /* value */ | ||
347 | int size = 0; | ||
348 | int nr = 1; | ||
349 | |||
350 | if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) | ||
351 | size += sizeof(u64); | ||
352 | |||
353 | if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) | ||
354 | size += sizeof(u64); | ||
355 | |||
356 | if (event->attr.read_format & PERF_FORMAT_ID) | ||
357 | entry += sizeof(u64); | ||
358 | |||
359 | if (event->attr.read_format & PERF_FORMAT_GROUP) { | ||
360 | nr += event->group_leader->nr_siblings; | ||
361 | size += sizeof(u64); | ||
362 | } | ||
363 | |||
364 | size += entry * nr; | ||
365 | event->read_size = size; | ||
366 | } | ||
367 | |||
368 | static void perf_event__header_size(struct perf_event *event) | ||
369 | { | ||
370 | struct perf_sample_data *data; | ||
371 | u64 sample_type = event->attr.sample_type; | ||
372 | u16 size = 0; | ||
373 | |||
374 | perf_event__read_size(event); | ||
375 | |||
376 | if (sample_type & PERF_SAMPLE_IP) | ||
377 | size += sizeof(data->ip); | ||
378 | |||
379 | if (sample_type & PERF_SAMPLE_ADDR) | ||
380 | size += sizeof(data->addr); | ||
381 | |||
382 | if (sample_type & PERF_SAMPLE_PERIOD) | ||
383 | size += sizeof(data->period); | ||
384 | |||
385 | if (sample_type & PERF_SAMPLE_READ) | ||
386 | size += event->read_size; | ||
387 | |||
388 | event->header_size = size; | ||
389 | } | ||
390 | |||
391 | static void perf_event__id_header_size(struct perf_event *event) | ||
392 | { | ||
393 | struct perf_sample_data *data; | ||
394 | u64 sample_type = event->attr.sample_type; | ||
395 | u16 size = 0; | ||
396 | |||
397 | if (sample_type & PERF_SAMPLE_TID) | ||
398 | size += sizeof(data->tid_entry); | ||
399 | |||
400 | if (sample_type & PERF_SAMPLE_TIME) | ||
401 | size += sizeof(data->time); | ||
402 | |||
403 | if (sample_type & PERF_SAMPLE_ID) | ||
404 | size += sizeof(data->id); | ||
405 | |||
406 | if (sample_type & PERF_SAMPLE_STREAM_ID) | ||
407 | size += sizeof(data->stream_id); | ||
408 | |||
409 | if (sample_type & PERF_SAMPLE_CPU) | ||
410 | size += sizeof(data->cpu_entry); | ||
411 | |||
412 | event->id_header_size = size; | ||
413 | } | ||
414 | |||
315 | static void perf_group_attach(struct perf_event *event) | 415 | static void perf_group_attach(struct perf_event *event) |
316 | { | 416 | { |
317 | struct perf_event *group_leader = event->group_leader; | 417 | struct perf_event *group_leader = event->group_leader, *pos; |
318 | 418 | ||
319 | /* | 419 | /* |
320 | * We can have double attach due to group movement in perf_event_open. | 420 | * We can have double attach due to group movement in perf_event_open. |
@@ -333,6 +433,11 @@ static void perf_group_attach(struct perf_event *event) | |||
333 | 433 | ||
334 | list_add_tail(&event->group_entry, &group_leader->sibling_list); | 434 | list_add_tail(&event->group_entry, &group_leader->sibling_list); |
335 | group_leader->nr_siblings++; | 435 | group_leader->nr_siblings++; |
436 | |||
437 | perf_event__header_size(group_leader); | ||
438 | |||
439 | list_for_each_entry(pos, &group_leader->sibling_list, group_entry) | ||
440 | perf_event__header_size(pos); | ||
336 | } | 441 | } |
337 | 442 | ||
338 | /* | 443 | /* |
@@ -391,7 +496,7 @@ static void perf_group_detach(struct perf_event *event) | |||
391 | if (event->group_leader != event) { | 496 | if (event->group_leader != event) { |
392 | list_del_init(&event->group_entry); | 497 | list_del_init(&event->group_entry); |
393 | event->group_leader->nr_siblings--; | 498 | event->group_leader->nr_siblings--; |
394 | return; | 499 | goto out; |
395 | } | 500 | } |
396 | 501 | ||
397 | if (!list_empty(&event->group_entry)) | 502 | if (!list_empty(&event->group_entry)) |
@@ -410,6 +515,12 @@ static void perf_group_detach(struct perf_event *event) | |||
410 | /* Inherit group flags from the previous leader */ | 515 | /* Inherit group flags from the previous leader */ |
411 | sibling->group_flags = event->group_flags; | 516 | sibling->group_flags = event->group_flags; |
412 | } | 517 | } |
518 | |||
519 | out: | ||
520 | perf_event__header_size(event->group_leader); | ||
521 | |||
522 | list_for_each_entry(tmp, &event->group_leader->sibling_list, group_entry) | ||
523 | perf_event__header_size(tmp); | ||
413 | } | 524 | } |
414 | 525 | ||
415 | static inline int | 526 | static inline int |
@@ -1073,7 +1184,7 @@ static int perf_event_refresh(struct perf_event *event, int refresh) | |||
1073 | /* | 1184 | /* |
1074 | * not supported on inherited events | 1185 | * not supported on inherited events |
1075 | */ | 1186 | */ |
1076 | if (event->attr.inherit) | 1187 | if (event->attr.inherit || !is_sampling_event(event)) |
1077 | return -EINVAL; | 1188 | return -EINVAL; |
1078 | 1189 | ||
1079 | atomic_add(refresh, &event->event_limit); | 1190 | atomic_add(refresh, &event->event_limit); |
@@ -2289,31 +2400,6 @@ static int perf_release(struct inode *inode, struct file *file) | |||
2289 | return perf_event_release_kernel(event); | 2400 | return perf_event_release_kernel(event); |
2290 | } | 2401 | } |
2291 | 2402 | ||
2292 | static int perf_event_read_size(struct perf_event *event) | ||
2293 | { | ||
2294 | int entry = sizeof(u64); /* value */ | ||
2295 | int size = 0; | ||
2296 | int nr = 1; | ||
2297 | |||
2298 | if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) | ||
2299 | size += sizeof(u64); | ||
2300 | |||
2301 | if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) | ||
2302 | size += sizeof(u64); | ||
2303 | |||
2304 | if (event->attr.read_format & PERF_FORMAT_ID) | ||
2305 | entry += sizeof(u64); | ||
2306 | |||
2307 | if (event->attr.read_format & PERF_FORMAT_GROUP) { | ||
2308 | nr += event->group_leader->nr_siblings; | ||
2309 | size += sizeof(u64); | ||
2310 | } | ||
2311 | |||
2312 | size += entry * nr; | ||
2313 | |||
2314 | return size; | ||
2315 | } | ||
2316 | |||
2317 | u64 perf_event_read_value(struct perf_event *event, u64 *enabled, u64 *running) | 2403 | u64 perf_event_read_value(struct perf_event *event, u64 *enabled, u64 *running) |
2318 | { | 2404 | { |
2319 | struct perf_event *child; | 2405 | struct perf_event *child; |
@@ -2428,7 +2514,7 @@ perf_read_hw(struct perf_event *event, char __user *buf, size_t count) | |||
2428 | if (event->state == PERF_EVENT_STATE_ERROR) | 2514 | if (event->state == PERF_EVENT_STATE_ERROR) |
2429 | return 0; | 2515 | return 0; |
2430 | 2516 | ||
2431 | if (count < perf_event_read_size(event)) | 2517 | if (count < event->read_size) |
2432 | return -ENOSPC; | 2518 | return -ENOSPC; |
2433 | 2519 | ||
2434 | WARN_ON_ONCE(event->ctx->parent_ctx); | 2520 | WARN_ON_ONCE(event->ctx->parent_ctx); |
@@ -2514,7 +2600,7 @@ static int perf_event_period(struct perf_event *event, u64 __user *arg) | |||
2514 | int ret = 0; | 2600 | int ret = 0; |
2515 | u64 value; | 2601 | u64 value; |
2516 | 2602 | ||
2517 | if (!event->attr.sample_period) | 2603 | if (!is_sampling_event(event)) |
2518 | return -EINVAL; | 2604 | return -EINVAL; |
2519 | 2605 | ||
2520 | if (copy_from_user(&value, arg, sizeof(value))) | 2606 | if (copy_from_user(&value, arg, sizeof(value))) |
@@ -3305,6 +3391,73 @@ __always_inline void perf_output_copy(struct perf_output_handle *handle, | |||
3305 | } while (len); | 3391 | } while (len); |
3306 | } | 3392 | } |
3307 | 3393 | ||
3394 | static void __perf_event_header__init_id(struct perf_event_header *header, | ||
3395 | struct perf_sample_data *data, | ||
3396 | struct perf_event *event) | ||
3397 | { | ||
3398 | u64 sample_type = event->attr.sample_type; | ||
3399 | |||
3400 | data->type = sample_type; | ||
3401 | header->size += event->id_header_size; | ||
3402 | |||
3403 | if (sample_type & PERF_SAMPLE_TID) { | ||
3404 | /* namespace issues */ | ||
3405 | data->tid_entry.pid = perf_event_pid(event, current); | ||
3406 | data->tid_entry.tid = perf_event_tid(event, current); | ||
3407 | } | ||
3408 | |||
3409 | if (sample_type & PERF_SAMPLE_TIME) | ||
3410 | data->time = perf_clock(); | ||
3411 | |||
3412 | if (sample_type & PERF_SAMPLE_ID) | ||
3413 | data->id = primary_event_id(event); | ||
3414 | |||
3415 | if (sample_type & PERF_SAMPLE_STREAM_ID) | ||
3416 | data->stream_id = event->id; | ||
3417 | |||
3418 | if (sample_type & PERF_SAMPLE_CPU) { | ||
3419 | data->cpu_entry.cpu = raw_smp_processor_id(); | ||
3420 | data->cpu_entry.reserved = 0; | ||
3421 | } | ||
3422 | } | ||
3423 | |||
3424 | static void perf_event_header__init_id(struct perf_event_header *header, | ||
3425 | struct perf_sample_data *data, | ||
3426 | struct perf_event *event) | ||
3427 | { | ||
3428 | if (event->attr.sample_id_all) | ||
3429 | __perf_event_header__init_id(header, data, event); | ||
3430 | } | ||
3431 | |||
3432 | static void __perf_event__output_id_sample(struct perf_output_handle *handle, | ||
3433 | struct perf_sample_data *data) | ||
3434 | { | ||
3435 | u64 sample_type = data->type; | ||
3436 | |||
3437 | if (sample_type & PERF_SAMPLE_TID) | ||
3438 | perf_output_put(handle, data->tid_entry); | ||
3439 | |||
3440 | if (sample_type & PERF_SAMPLE_TIME) | ||
3441 | perf_output_put(handle, data->time); | ||
3442 | |||
3443 | if (sample_type & PERF_SAMPLE_ID) | ||
3444 | perf_output_put(handle, data->id); | ||
3445 | |||
3446 | if (sample_type & PERF_SAMPLE_STREAM_ID) | ||
3447 | perf_output_put(handle, data->stream_id); | ||
3448 | |||
3449 | if (sample_type & PERF_SAMPLE_CPU) | ||
3450 | perf_output_put(handle, data->cpu_entry); | ||
3451 | } | ||
3452 | |||
3453 | static void perf_event__output_id_sample(struct perf_event *event, | ||
3454 | struct perf_output_handle *handle, | ||
3455 | struct perf_sample_data *sample) | ||
3456 | { | ||
3457 | if (event->attr.sample_id_all) | ||
3458 | __perf_event__output_id_sample(handle, sample); | ||
3459 | } | ||
3460 | |||
3308 | int perf_output_begin(struct perf_output_handle *handle, | 3461 | int perf_output_begin(struct perf_output_handle *handle, |
3309 | struct perf_event *event, unsigned int size, | 3462 | struct perf_event *event, unsigned int size, |
3310 | int nmi, int sample) | 3463 | int nmi, int sample) |
@@ -3312,6 +3465,7 @@ int perf_output_begin(struct perf_output_handle *handle, | |||
3312 | struct perf_buffer *buffer; | 3465 | struct perf_buffer *buffer; |
3313 | unsigned long tail, offset, head; | 3466 | unsigned long tail, offset, head; |
3314 | int have_lost; | 3467 | int have_lost; |
3468 | struct perf_sample_data sample_data; | ||
3315 | struct { | 3469 | struct { |
3316 | struct perf_event_header header; | 3470 | struct perf_event_header header; |
3317 | u64 id; | 3471 | u64 id; |
@@ -3338,8 +3492,12 @@ int perf_output_begin(struct perf_output_handle *handle, | |||
3338 | goto out; | 3492 | goto out; |
3339 | 3493 | ||
3340 | have_lost = local_read(&buffer->lost); | 3494 | have_lost = local_read(&buffer->lost); |
3341 | if (have_lost) | 3495 | if (have_lost) { |
3342 | size += sizeof(lost_event); | 3496 | lost_event.header.size = sizeof(lost_event); |
3497 | perf_event_header__init_id(&lost_event.header, &sample_data, | ||
3498 | event); | ||
3499 | size += lost_event.header.size; | ||
3500 | } | ||
3343 | 3501 | ||
3344 | perf_output_get_handle(handle); | 3502 | perf_output_get_handle(handle); |
3345 | 3503 | ||
@@ -3370,11 +3528,11 @@ int perf_output_begin(struct perf_output_handle *handle, | |||
3370 | if (have_lost) { | 3528 | if (have_lost) { |
3371 | lost_event.header.type = PERF_RECORD_LOST; | 3529 | lost_event.header.type = PERF_RECORD_LOST; |
3372 | lost_event.header.misc = 0; | 3530 | lost_event.header.misc = 0; |
3373 | lost_event.header.size = sizeof(lost_event); | ||
3374 | lost_event.id = event->id; | 3531 | lost_event.id = event->id; |
3375 | lost_event.lost = local_xchg(&buffer->lost, 0); | 3532 | lost_event.lost = local_xchg(&buffer->lost, 0); |
3376 | 3533 | ||
3377 | perf_output_put(handle, lost_event); | 3534 | perf_output_put(handle, lost_event); |
3535 | perf_event__output_id_sample(event, handle, &sample_data); | ||
3378 | } | 3536 | } |
3379 | 3537 | ||
3380 | return 0; | 3538 | return 0; |
@@ -3407,28 +3565,6 @@ void perf_output_end(struct perf_output_handle *handle) | |||
3407 | rcu_read_unlock(); | 3565 | rcu_read_unlock(); |
3408 | } | 3566 | } |
3409 | 3567 | ||
3410 | static u32 perf_event_pid(struct perf_event *event, struct task_struct *p) | ||
3411 | { | ||
3412 | /* | ||
3413 | * only top level events have the pid namespace they were created in | ||
3414 | */ | ||
3415 | if (event->parent) | ||
3416 | event = event->parent; | ||
3417 | |||
3418 | return task_tgid_nr_ns(p, event->ns); | ||
3419 | } | ||
3420 | |||
3421 | static u32 perf_event_tid(struct perf_event *event, struct task_struct *p) | ||
3422 | { | ||
3423 | /* | ||
3424 | * only top level events have the pid namespace they were created in | ||
3425 | */ | ||
3426 | if (event->parent) | ||
3427 | event = event->parent; | ||
3428 | |||
3429 | return task_pid_nr_ns(p, event->ns); | ||
3430 | } | ||
3431 | |||
3432 | static void perf_output_read_one(struct perf_output_handle *handle, | 3568 | static void perf_output_read_one(struct perf_output_handle *handle, |
3433 | struct perf_event *event, | 3569 | struct perf_event *event, |
3434 | u64 enabled, u64 running) | 3570 | u64 enabled, u64 running) |
@@ -3603,61 +3739,16 @@ void perf_prepare_sample(struct perf_event_header *header, | |||
3603 | { | 3739 | { |
3604 | u64 sample_type = event->attr.sample_type; | 3740 | u64 sample_type = event->attr.sample_type; |
3605 | 3741 | ||
3606 | data->type = sample_type; | ||
3607 | |||
3608 | header->type = PERF_RECORD_SAMPLE; | 3742 | header->type = PERF_RECORD_SAMPLE; |
3609 | header->size = sizeof(*header); | 3743 | header->size = sizeof(*header) + event->header_size; |
3610 | 3744 | ||
3611 | header->misc = 0; | 3745 | header->misc = 0; |
3612 | header->misc |= perf_misc_flags(regs); | 3746 | header->misc |= perf_misc_flags(regs); |
3613 | 3747 | ||
3614 | if (sample_type & PERF_SAMPLE_IP) { | 3748 | __perf_event_header__init_id(header, data, event); |
3615 | data->ip = perf_instruction_pointer(regs); | ||
3616 | |||
3617 | header->size += sizeof(data->ip); | ||
3618 | } | ||
3619 | |||
3620 | if (sample_type & PERF_SAMPLE_TID) { | ||
3621 | /* namespace issues */ | ||
3622 | data->tid_entry.pid = perf_event_pid(event, current); | ||
3623 | data->tid_entry.tid = perf_event_tid(event, current); | ||
3624 | |||
3625 | header->size += sizeof(data->tid_entry); | ||
3626 | } | ||
3627 | |||
3628 | if (sample_type & PERF_SAMPLE_TIME) { | ||
3629 | data->time = perf_clock(); | ||
3630 | |||
3631 | header->size += sizeof(data->time); | ||
3632 | } | ||
3633 | |||
3634 | if (sample_type & PERF_SAMPLE_ADDR) | ||
3635 | header->size += sizeof(data->addr); | ||
3636 | |||
3637 | if (sample_type & PERF_SAMPLE_ID) { | ||
3638 | data->id = primary_event_id(event); | ||
3639 | |||
3640 | header->size += sizeof(data->id); | ||
3641 | } | ||
3642 | |||
3643 | if (sample_type & PERF_SAMPLE_STREAM_ID) { | ||
3644 | data->stream_id = event->id; | ||
3645 | |||
3646 | header->size += sizeof(data->stream_id); | ||
3647 | } | ||
3648 | |||
3649 | if (sample_type & PERF_SAMPLE_CPU) { | ||
3650 | data->cpu_entry.cpu = raw_smp_processor_id(); | ||
3651 | data->cpu_entry.reserved = 0; | ||
3652 | |||
3653 | header->size += sizeof(data->cpu_entry); | ||
3654 | } | ||
3655 | |||
3656 | if (sample_type & PERF_SAMPLE_PERIOD) | ||
3657 | header->size += sizeof(data->period); | ||
3658 | 3749 | ||
3659 | if (sample_type & PERF_SAMPLE_READ) | 3750 | if (sample_type & PERF_SAMPLE_IP) |
3660 | header->size += perf_event_read_size(event); | 3751 | data->ip = perf_instruction_pointer(regs); |
3661 | 3752 | ||
3662 | if (sample_type & PERF_SAMPLE_CALLCHAIN) { | 3753 | if (sample_type & PERF_SAMPLE_CALLCHAIN) { |
3663 | int size = 1; | 3754 | int size = 1; |
@@ -3722,23 +3813,26 @@ perf_event_read_event(struct perf_event *event, | |||
3722 | struct task_struct *task) | 3813 | struct task_struct *task) |
3723 | { | 3814 | { |
3724 | struct perf_output_handle handle; | 3815 | struct perf_output_handle handle; |
3816 | struct perf_sample_data sample; | ||
3725 | struct perf_read_event read_event = { | 3817 | struct perf_read_event read_event = { |
3726 | .header = { | 3818 | .header = { |
3727 | .type = PERF_RECORD_READ, | 3819 | .type = PERF_RECORD_READ, |
3728 | .misc = 0, | 3820 | .misc = 0, |
3729 | .size = sizeof(read_event) + perf_event_read_size(event), | 3821 | .size = sizeof(read_event) + event->read_size, |
3730 | }, | 3822 | }, |
3731 | .pid = perf_event_pid(event, task), | 3823 | .pid = perf_event_pid(event, task), |
3732 | .tid = perf_event_tid(event, task), | 3824 | .tid = perf_event_tid(event, task), |
3733 | }; | 3825 | }; |
3734 | int ret; | 3826 | int ret; |
3735 | 3827 | ||
3828 | perf_event_header__init_id(&read_event.header, &sample, event); | ||
3736 | ret = perf_output_begin(&handle, event, read_event.header.size, 0, 0); | 3829 | ret = perf_output_begin(&handle, event, read_event.header.size, 0, 0); |
3737 | if (ret) | 3830 | if (ret) |
3738 | return; | 3831 | return; |
3739 | 3832 | ||
3740 | perf_output_put(&handle, read_event); | 3833 | perf_output_put(&handle, read_event); |
3741 | perf_output_read(&handle, event); | 3834 | perf_output_read(&handle, event); |
3835 | perf_event__output_id_sample(event, &handle, &sample); | ||
3742 | 3836 | ||
3743 | perf_output_end(&handle); | 3837 | perf_output_end(&handle); |
3744 | } | 3838 | } |
@@ -3768,14 +3862,16 @@ static void perf_event_task_output(struct perf_event *event, | |||
3768 | struct perf_task_event *task_event) | 3862 | struct perf_task_event *task_event) |
3769 | { | 3863 | { |
3770 | struct perf_output_handle handle; | 3864 | struct perf_output_handle handle; |
3865 | struct perf_sample_data sample; | ||
3771 | struct task_struct *task = task_event->task; | 3866 | struct task_struct *task = task_event->task; |
3772 | int size, ret; | 3867 | int ret, size = task_event->event_id.header.size; |
3773 | 3868 | ||
3774 | size = task_event->event_id.header.size; | 3869 | perf_event_header__init_id(&task_event->event_id.header, &sample, event); |
3775 | ret = perf_output_begin(&handle, event, size, 0, 0); | ||
3776 | 3870 | ||
3871 | ret = perf_output_begin(&handle, event, | ||
3872 | task_event->event_id.header.size, 0, 0); | ||
3777 | if (ret) | 3873 | if (ret) |
3778 | return; | 3874 | goto out; |
3779 | 3875 | ||
3780 | task_event->event_id.pid = perf_event_pid(event, task); | 3876 | task_event->event_id.pid = perf_event_pid(event, task); |
3781 | task_event->event_id.ppid = perf_event_pid(event, current); | 3877 | task_event->event_id.ppid = perf_event_pid(event, current); |
@@ -3785,7 +3881,11 @@ static void perf_event_task_output(struct perf_event *event, | |||
3785 | 3881 | ||
3786 | perf_output_put(&handle, task_event->event_id); | 3882 | perf_output_put(&handle, task_event->event_id); |
3787 | 3883 | ||
3884 | perf_event__output_id_sample(event, &handle, &sample); | ||
3885 | |||
3788 | perf_output_end(&handle); | 3886 | perf_output_end(&handle); |
3887 | out: | ||
3888 | task_event->event_id.header.size = size; | ||
3789 | } | 3889 | } |
3790 | 3890 | ||
3791 | static int perf_event_task_match(struct perf_event *event) | 3891 | static int perf_event_task_match(struct perf_event *event) |
@@ -3900,11 +4000,16 @@ static void perf_event_comm_output(struct perf_event *event, | |||
3900 | struct perf_comm_event *comm_event) | 4000 | struct perf_comm_event *comm_event) |
3901 | { | 4001 | { |
3902 | struct perf_output_handle handle; | 4002 | struct perf_output_handle handle; |
4003 | struct perf_sample_data sample; | ||
3903 | int size = comm_event->event_id.header.size; | 4004 | int size = comm_event->event_id.header.size; |
3904 | int ret = perf_output_begin(&handle, event, size, 0, 0); | 4005 | int ret; |
4006 | |||
4007 | perf_event_header__init_id(&comm_event->event_id.header, &sample, event); | ||
4008 | ret = perf_output_begin(&handle, event, | ||
4009 | comm_event->event_id.header.size, 0, 0); | ||
3905 | 4010 | ||
3906 | if (ret) | 4011 | if (ret) |
3907 | return; | 4012 | goto out; |
3908 | 4013 | ||
3909 | comm_event->event_id.pid = perf_event_pid(event, comm_event->task); | 4014 | comm_event->event_id.pid = perf_event_pid(event, comm_event->task); |
3910 | comm_event->event_id.tid = perf_event_tid(event, comm_event->task); | 4015 | comm_event->event_id.tid = perf_event_tid(event, comm_event->task); |
@@ -3912,7 +4017,12 @@ static void perf_event_comm_output(struct perf_event *event, | |||
3912 | perf_output_put(&handle, comm_event->event_id); | 4017 | perf_output_put(&handle, comm_event->event_id); |
3913 | perf_output_copy(&handle, comm_event->comm, | 4018 | perf_output_copy(&handle, comm_event->comm, |
3914 | comm_event->comm_size); | 4019 | comm_event->comm_size); |
4020 | |||
4021 | perf_event__output_id_sample(event, &handle, &sample); | ||
4022 | |||
3915 | perf_output_end(&handle); | 4023 | perf_output_end(&handle); |
4024 | out: | ||
4025 | comm_event->event_id.header.size = size; | ||
3916 | } | 4026 | } |
3917 | 4027 | ||
3918 | static int perf_event_comm_match(struct perf_event *event) | 4028 | static int perf_event_comm_match(struct perf_event *event) |
@@ -3957,7 +4067,6 @@ static void perf_event_comm_event(struct perf_comm_event *comm_event) | |||
3957 | comm_event->comm_size = size; | 4067 | comm_event->comm_size = size; |
3958 | 4068 | ||
3959 | comm_event->event_id.header.size = sizeof(comm_event->event_id) + size; | 4069 | comm_event->event_id.header.size = sizeof(comm_event->event_id) + size; |
3960 | |||
3961 | rcu_read_lock(); | 4070 | rcu_read_lock(); |
3962 | list_for_each_entry_rcu(pmu, &pmus, entry) { | 4071 | list_for_each_entry_rcu(pmu, &pmus, entry) { |
3963 | cpuctx = get_cpu_ptr(pmu->pmu_cpu_context); | 4072 | cpuctx = get_cpu_ptr(pmu->pmu_cpu_context); |
@@ -4038,11 +4147,15 @@ static void perf_event_mmap_output(struct perf_event *event, | |||
4038 | struct perf_mmap_event *mmap_event) | 4147 | struct perf_mmap_event *mmap_event) |
4039 | { | 4148 | { |
4040 | struct perf_output_handle handle; | 4149 | struct perf_output_handle handle; |
4150 | struct perf_sample_data sample; | ||
4041 | int size = mmap_event->event_id.header.size; | 4151 | int size = mmap_event->event_id.header.size; |
4042 | int ret = perf_output_begin(&handle, event, size, 0, 0); | 4152 | int ret; |
4043 | 4153 | ||
4154 | perf_event_header__init_id(&mmap_event->event_id.header, &sample, event); | ||
4155 | ret = perf_output_begin(&handle, event, | ||
4156 | mmap_event->event_id.header.size, 0, 0); | ||
4044 | if (ret) | 4157 | if (ret) |
4045 | return; | 4158 | goto out; |
4046 | 4159 | ||
4047 | mmap_event->event_id.pid = perf_event_pid(event, current); | 4160 | mmap_event->event_id.pid = perf_event_pid(event, current); |
4048 | mmap_event->event_id.tid = perf_event_tid(event, current); | 4161 | mmap_event->event_id.tid = perf_event_tid(event, current); |
@@ -4050,7 +4163,12 @@ static void perf_event_mmap_output(struct perf_event *event, | |||
4050 | perf_output_put(&handle, mmap_event->event_id); | 4163 | perf_output_put(&handle, mmap_event->event_id); |
4051 | perf_output_copy(&handle, mmap_event->file_name, | 4164 | perf_output_copy(&handle, mmap_event->file_name, |
4052 | mmap_event->file_size); | 4165 | mmap_event->file_size); |
4166 | |||
4167 | perf_event__output_id_sample(event, &handle, &sample); | ||
4168 | |||
4053 | perf_output_end(&handle); | 4169 | perf_output_end(&handle); |
4170 | out: | ||
4171 | mmap_event->event_id.header.size = size; | ||
4054 | } | 4172 | } |
4055 | 4173 | ||
4056 | static int perf_event_mmap_match(struct perf_event *event, | 4174 | static int perf_event_mmap_match(struct perf_event *event, |
@@ -4205,6 +4323,7 @@ void perf_event_mmap(struct vm_area_struct *vma) | |||
4205 | static void perf_log_throttle(struct perf_event *event, int enable) | 4323 | static void perf_log_throttle(struct perf_event *event, int enable) |
4206 | { | 4324 | { |
4207 | struct perf_output_handle handle; | 4325 | struct perf_output_handle handle; |
4326 | struct perf_sample_data sample; | ||
4208 | int ret; | 4327 | int ret; |
4209 | 4328 | ||
4210 | struct { | 4329 | struct { |
@@ -4226,11 +4345,15 @@ static void perf_log_throttle(struct perf_event *event, int enable) | |||
4226 | if (enable) | 4345 | if (enable) |
4227 | throttle_event.header.type = PERF_RECORD_UNTHROTTLE; | 4346 | throttle_event.header.type = PERF_RECORD_UNTHROTTLE; |
4228 | 4347 | ||
4229 | ret = perf_output_begin(&handle, event, sizeof(throttle_event), 1, 0); | 4348 | perf_event_header__init_id(&throttle_event.header, &sample, event); |
4349 | |||
4350 | ret = perf_output_begin(&handle, event, | ||
4351 | throttle_event.header.size, 1, 0); | ||
4230 | if (ret) | 4352 | if (ret) |
4231 | return; | 4353 | return; |
4232 | 4354 | ||
4233 | perf_output_put(&handle, throttle_event); | 4355 | perf_output_put(&handle, throttle_event); |
4356 | perf_event__output_id_sample(event, &handle, &sample); | ||
4234 | perf_output_end(&handle); | 4357 | perf_output_end(&handle); |
4235 | } | 4358 | } |
4236 | 4359 | ||
@@ -4246,6 +4369,13 @@ static int __perf_event_overflow(struct perf_event *event, int nmi, | |||
4246 | struct hw_perf_event *hwc = &event->hw; | 4369 | struct hw_perf_event *hwc = &event->hw; |
4247 | int ret = 0; | 4370 | int ret = 0; |
4248 | 4371 | ||
4372 | /* | ||
4373 | * Non-sampling counters might still use the PMI to fold short | ||
4374 | * hardware counters, ignore those. | ||
4375 | */ | ||
4376 | if (unlikely(!is_sampling_event(event))) | ||
4377 | return 0; | ||
4378 | |||
4249 | if (!throttle) { | 4379 | if (!throttle) { |
4250 | hwc->interrupts++; | 4380 | hwc->interrupts++; |
4251 | } else { | 4381 | } else { |
@@ -4391,7 +4521,7 @@ static void perf_swevent_event(struct perf_event *event, u64 nr, | |||
4391 | if (!regs) | 4521 | if (!regs) |
4392 | return; | 4522 | return; |
4393 | 4523 | ||
4394 | if (!hwc->sample_period) | 4524 | if (!is_sampling_event(event)) |
4395 | return; | 4525 | return; |
4396 | 4526 | ||
4397 | if (nr == 1 && hwc->sample_period == 1 && !event->attr.freq) | 4527 | if (nr == 1 && hwc->sample_period == 1 && !event->attr.freq) |
@@ -4554,7 +4684,7 @@ static int perf_swevent_add(struct perf_event *event, int flags) | |||
4554 | struct hw_perf_event *hwc = &event->hw; | 4684 | struct hw_perf_event *hwc = &event->hw; |
4555 | struct hlist_head *head; | 4685 | struct hlist_head *head; |
4556 | 4686 | ||
4557 | if (hwc->sample_period) { | 4687 | if (is_sampling_event(event)) { |
4558 | hwc->last_period = hwc->sample_period; | 4688 | hwc->last_period = hwc->sample_period; |
4559 | perf_swevent_set_period(event); | 4689 | perf_swevent_set_period(event); |
4560 | } | 4690 | } |
@@ -4811,15 +4941,6 @@ static int perf_tp_event_init(struct perf_event *event) | |||
4811 | if (event->attr.type != PERF_TYPE_TRACEPOINT) | 4941 | if (event->attr.type != PERF_TYPE_TRACEPOINT) |
4812 | return -ENOENT; | 4942 | return -ENOENT; |
4813 | 4943 | ||
4814 | /* | ||
4815 | * Raw tracepoint data is a severe data leak, only allow root to | ||
4816 | * have these. | ||
4817 | */ | ||
4818 | if ((event->attr.sample_type & PERF_SAMPLE_RAW) && | ||
4819 | perf_paranoid_tracepoint_raw() && | ||
4820 | !capable(CAP_SYS_ADMIN)) | ||
4821 | return -EPERM; | ||
4822 | |||
4823 | err = perf_trace_init(event); | 4944 | err = perf_trace_init(event); |
4824 | if (err) | 4945 | if (err) |
4825 | return err; | 4946 | return err; |
@@ -4842,7 +4963,7 @@ static struct pmu perf_tracepoint = { | |||
4842 | 4963 | ||
4843 | static inline void perf_tp_register(void) | 4964 | static inline void perf_tp_register(void) |
4844 | { | 4965 | { |
4845 | perf_pmu_register(&perf_tracepoint); | 4966 | perf_pmu_register(&perf_tracepoint, "tracepoint", PERF_TYPE_TRACEPOINT); |
4846 | } | 4967 | } |
4847 | 4968 | ||
4848 | static int perf_event_set_filter(struct perf_event *event, void __user *arg) | 4969 | static int perf_event_set_filter(struct perf_event *event, void __user *arg) |
@@ -4932,31 +5053,33 @@ static enum hrtimer_restart perf_swevent_hrtimer(struct hrtimer *hrtimer) | |||
4932 | static void perf_swevent_start_hrtimer(struct perf_event *event) | 5053 | static void perf_swevent_start_hrtimer(struct perf_event *event) |
4933 | { | 5054 | { |
4934 | struct hw_perf_event *hwc = &event->hw; | 5055 | struct hw_perf_event *hwc = &event->hw; |
5056 | s64 period; | ||
5057 | |||
5058 | if (!is_sampling_event(event)) | ||
5059 | return; | ||
4935 | 5060 | ||
4936 | hrtimer_init(&hwc->hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); | 5061 | hrtimer_init(&hwc->hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
4937 | hwc->hrtimer.function = perf_swevent_hrtimer; | 5062 | hwc->hrtimer.function = perf_swevent_hrtimer; |
4938 | if (hwc->sample_period) { | ||
4939 | s64 period = local64_read(&hwc->period_left); | ||
4940 | 5063 | ||
4941 | if (period) { | 5064 | period = local64_read(&hwc->period_left); |
4942 | if (period < 0) | 5065 | if (period) { |
4943 | period = 10000; | 5066 | if (period < 0) |
5067 | period = 10000; | ||
4944 | 5068 | ||
4945 | local64_set(&hwc->period_left, 0); | 5069 | local64_set(&hwc->period_left, 0); |
4946 | } else { | 5070 | } else { |
4947 | period = max_t(u64, 10000, hwc->sample_period); | 5071 | period = max_t(u64, 10000, hwc->sample_period); |
4948 | } | 5072 | } |
4949 | __hrtimer_start_range_ns(&hwc->hrtimer, | 5073 | __hrtimer_start_range_ns(&hwc->hrtimer, |
4950 | ns_to_ktime(period), 0, | 5074 | ns_to_ktime(period), 0, |
4951 | HRTIMER_MODE_REL_PINNED, 0); | 5075 | HRTIMER_MODE_REL_PINNED, 0); |
4952 | } | ||
4953 | } | 5076 | } |
4954 | 5077 | ||
4955 | static void perf_swevent_cancel_hrtimer(struct perf_event *event) | 5078 | static void perf_swevent_cancel_hrtimer(struct perf_event *event) |
4956 | { | 5079 | { |
4957 | struct hw_perf_event *hwc = &event->hw; | 5080 | struct hw_perf_event *hwc = &event->hw; |
4958 | 5081 | ||
4959 | if (hwc->sample_period) { | 5082 | if (is_sampling_event(event)) { |
4960 | ktime_t remaining = hrtimer_get_remaining(&hwc->hrtimer); | 5083 | ktime_t remaining = hrtimer_get_remaining(&hwc->hrtimer); |
4961 | local64_set(&hwc->period_left, ktime_to_ns(remaining)); | 5084 | local64_set(&hwc->period_left, ktime_to_ns(remaining)); |
4962 | 5085 | ||
@@ -5184,8 +5307,61 @@ static void free_pmu_context(struct pmu *pmu) | |||
5184 | out: | 5307 | out: |
5185 | mutex_unlock(&pmus_lock); | 5308 | mutex_unlock(&pmus_lock); |
5186 | } | 5309 | } |
5310 | static struct idr pmu_idr; | ||
5311 | |||
5312 | static ssize_t | ||
5313 | type_show(struct device *dev, struct device_attribute *attr, char *page) | ||
5314 | { | ||
5315 | struct pmu *pmu = dev_get_drvdata(dev); | ||
5316 | |||
5317 | return snprintf(page, PAGE_SIZE-1, "%d\n", pmu->type); | ||
5318 | } | ||
5319 | |||
5320 | static struct device_attribute pmu_dev_attrs[] = { | ||
5321 | __ATTR_RO(type), | ||
5322 | __ATTR_NULL, | ||
5323 | }; | ||
5324 | |||
5325 | static int pmu_bus_running; | ||
5326 | static struct bus_type pmu_bus = { | ||
5327 | .name = "event_source", | ||
5328 | .dev_attrs = pmu_dev_attrs, | ||
5329 | }; | ||
5330 | |||
5331 | static void pmu_dev_release(struct device *dev) | ||
5332 | { | ||
5333 | kfree(dev); | ||
5334 | } | ||
5335 | |||
5336 | static int pmu_dev_alloc(struct pmu *pmu) | ||
5337 | { | ||
5338 | int ret = -ENOMEM; | ||
5339 | |||
5340 | pmu->dev = kzalloc(sizeof(struct device), GFP_KERNEL); | ||
5341 | if (!pmu->dev) | ||
5342 | goto out; | ||
5343 | |||
5344 | device_initialize(pmu->dev); | ||
5345 | ret = dev_set_name(pmu->dev, "%s", pmu->name); | ||
5346 | if (ret) | ||
5347 | goto free_dev; | ||
5348 | |||
5349 | dev_set_drvdata(pmu->dev, pmu); | ||
5350 | pmu->dev->bus = &pmu_bus; | ||
5351 | pmu->dev->release = pmu_dev_release; | ||
5352 | ret = device_add(pmu->dev); | ||
5353 | if (ret) | ||
5354 | goto free_dev; | ||
5355 | |||
5356 | out: | ||
5357 | return ret; | ||
5358 | |||
5359 | free_dev: | ||
5360 | put_device(pmu->dev); | ||
5361 | goto out; | ||
5362 | } | ||
5187 | 5363 | ||
5188 | int perf_pmu_register(struct pmu *pmu) | 5364 | int perf_pmu_register(struct pmu *pmu, char *name, int type) |
5189 | { | 5365 | { |
5190 | int cpu, ret; | 5366 | int cpu, ret; |
5191 | 5367 | ||
@@ -5195,13 +5371,38 @@ int perf_pmu_register(struct pmu *pmu) | |||
5195 | if (!pmu->pmu_disable_count) | 5371 | if (!pmu->pmu_disable_count) |
5196 | goto unlock; | 5372 | goto unlock; |
5197 | 5373 | ||
5374 | pmu->type = -1; | ||
5375 | if (!name) | ||
5376 | goto skip_type; | ||
5377 | pmu->name = name; | ||
5378 | |||
5379 | if (type < 0) { | ||
5380 | int err = idr_pre_get(&pmu_idr, GFP_KERNEL); | ||
5381 | if (!err) | ||
5382 | goto free_pdc; | ||
5383 | |||
5384 | err = idr_get_new_above(&pmu_idr, pmu, PERF_TYPE_MAX, &type); | ||
5385 | if (err) { | ||
5386 | ret = err; | ||
5387 | goto free_pdc; | ||
5388 | } | ||
5389 | } | ||
5390 | pmu->type = type; | ||
5391 | |||
5392 | if (pmu_bus_running) { | ||
5393 | ret = pmu_dev_alloc(pmu); | ||
5394 | if (ret) | ||
5395 | goto free_idr; | ||
5396 | } | ||
5397 | |||
5398 | skip_type: | ||
5198 | pmu->pmu_cpu_context = find_pmu_context(pmu->task_ctx_nr); | 5399 | pmu->pmu_cpu_context = find_pmu_context(pmu->task_ctx_nr); |
5199 | if (pmu->pmu_cpu_context) | 5400 | if (pmu->pmu_cpu_context) |
5200 | goto got_cpu_context; | 5401 | goto got_cpu_context; |
5201 | 5402 | ||
5202 | pmu->pmu_cpu_context = alloc_percpu(struct perf_cpu_context); | 5403 | pmu->pmu_cpu_context = alloc_percpu(struct perf_cpu_context); |
5203 | if (!pmu->pmu_cpu_context) | 5404 | if (!pmu->pmu_cpu_context) |
5204 | goto free_pdc; | 5405 | goto free_dev; |
5205 | 5406 | ||
5206 | for_each_possible_cpu(cpu) { | 5407 | for_each_possible_cpu(cpu) { |
5207 | struct perf_cpu_context *cpuctx; | 5408 | struct perf_cpu_context *cpuctx; |
@@ -5245,6 +5446,14 @@ unlock: | |||
5245 | 5446 | ||
5246 | return ret; | 5447 | return ret; |
5247 | 5448 | ||
5449 | free_dev: | ||
5450 | device_del(pmu->dev); | ||
5451 | put_device(pmu->dev); | ||
5452 | |||
5453 | free_idr: | ||
5454 | if (pmu->type >= PERF_TYPE_MAX) | ||
5455 | idr_remove(&pmu_idr, pmu->type); | ||
5456 | |||
5248 | free_pdc: | 5457 | free_pdc: |
5249 | free_percpu(pmu->pmu_disable_count); | 5458 | free_percpu(pmu->pmu_disable_count); |
5250 | goto unlock; | 5459 | goto unlock; |
@@ -5264,6 +5473,10 @@ void perf_pmu_unregister(struct pmu *pmu) | |||
5264 | synchronize_rcu(); | 5473 | synchronize_rcu(); |
5265 | 5474 | ||
5266 | free_percpu(pmu->pmu_disable_count); | 5475 | free_percpu(pmu->pmu_disable_count); |
5476 | if (pmu->type >= PERF_TYPE_MAX) | ||
5477 | idr_remove(&pmu_idr, pmu->type); | ||
5478 | device_del(pmu->dev); | ||
5479 | put_device(pmu->dev); | ||
5267 | free_pmu_context(pmu); | 5480 | free_pmu_context(pmu); |
5268 | } | 5481 | } |
5269 | 5482 | ||
@@ -5273,6 +5486,13 @@ struct pmu *perf_init_event(struct perf_event *event) | |||
5273 | int idx; | 5486 | int idx; |
5274 | 5487 | ||
5275 | idx = srcu_read_lock(&pmus_srcu); | 5488 | idx = srcu_read_lock(&pmus_srcu); |
5489 | |||
5490 | rcu_read_lock(); | ||
5491 | pmu = idr_find(&pmu_idr, event->attr.type); | ||
5492 | rcu_read_unlock(); | ||
5493 | if (pmu) | ||
5494 | goto unlock; | ||
5495 | |||
5276 | list_for_each_entry_rcu(pmu, &pmus, entry) { | 5496 | list_for_each_entry_rcu(pmu, &pmus, entry) { |
5277 | int ret = pmu->event_init(event); | 5497 | int ret = pmu->event_init(event); |
5278 | if (!ret) | 5498 | if (!ret) |
@@ -5738,6 +5958,12 @@ SYSCALL_DEFINE5(perf_event_open, | |||
5738 | mutex_unlock(¤t->perf_event_mutex); | 5958 | mutex_unlock(¤t->perf_event_mutex); |
5739 | 5959 | ||
5740 | /* | 5960 | /* |
5961 | * Precalculate sample_data sizes | ||
5962 | */ | ||
5963 | perf_event__header_size(event); | ||
5964 | perf_event__id_header_size(event); | ||
5965 | |||
5966 | /* | ||
5741 | * Drop the reference on the group_event after placing the | 5967 | * Drop the reference on the group_event after placing the |
5742 | * new event on the sibling_list. This ensures destruction | 5968 | * new event on the sibling_list. This ensures destruction |
5743 | * of the group leader will find the pointer to itself in | 5969 | * of the group leader will find the pointer to itself in |
@@ -6090,6 +6316,12 @@ inherit_event(struct perf_event *parent_event, | |||
6090 | child_event->overflow_handler = parent_event->overflow_handler; | 6316 | child_event->overflow_handler = parent_event->overflow_handler; |
6091 | 6317 | ||
6092 | /* | 6318 | /* |
6319 | * Precalculate sample_data sizes | ||
6320 | */ | ||
6321 | perf_event__header_size(child_event); | ||
6322 | perf_event__id_header_size(child_event); | ||
6323 | |||
6324 | /* | ||
6093 | * Link it up in the child's context: | 6325 | * Link it up in the child's context: |
6094 | */ | 6326 | */ |
6095 | raw_spin_lock_irqsave(&child_ctx->lock, flags); | 6327 | raw_spin_lock_irqsave(&child_ctx->lock, flags); |
@@ -6320,7 +6552,7 @@ static void __cpuinit perf_event_init_cpu(int cpu) | |||
6320 | mutex_unlock(&swhash->hlist_mutex); | 6552 | mutex_unlock(&swhash->hlist_mutex); |
6321 | } | 6553 | } |
6322 | 6554 | ||
6323 | #ifdef CONFIG_HOTPLUG_CPU | 6555 | #if defined CONFIG_HOTPLUG_CPU || defined CONFIG_KEXEC |
6324 | static void perf_pmu_rotate_stop(struct pmu *pmu) | 6556 | static void perf_pmu_rotate_stop(struct pmu *pmu) |
6325 | { | 6557 | { |
6326 | struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context); | 6558 | struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context); |
@@ -6374,6 +6606,26 @@ static void perf_event_exit_cpu(int cpu) | |||
6374 | static inline void perf_event_exit_cpu(int cpu) { } | 6606 | static inline void perf_event_exit_cpu(int cpu) { } |
6375 | #endif | 6607 | #endif |
6376 | 6608 | ||
6609 | static int | ||
6610 | perf_reboot(struct notifier_block *notifier, unsigned long val, void *v) | ||
6611 | { | ||
6612 | int cpu; | ||
6613 | |||
6614 | for_each_online_cpu(cpu) | ||
6615 | perf_event_exit_cpu(cpu); | ||
6616 | |||
6617 | return NOTIFY_OK; | ||
6618 | } | ||
6619 | |||
6620 | /* | ||
6621 | * Run the perf reboot notifier at the very last possible moment so that | ||
6622 | * the generic watchdog code runs as long as possible. | ||
6623 | */ | ||
6624 | static struct notifier_block perf_reboot_notifier = { | ||
6625 | .notifier_call = perf_reboot, | ||
6626 | .priority = INT_MIN, | ||
6627 | }; | ||
6628 | |||
6377 | static int __cpuinit | 6629 | static int __cpuinit |
6378 | perf_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu) | 6630 | perf_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu) |
6379 | { | 6631 | { |
@@ -6402,14 +6654,45 @@ void __init perf_event_init(void) | |||
6402 | { | 6654 | { |
6403 | int ret; | 6655 | int ret; |
6404 | 6656 | ||
6657 | idr_init(&pmu_idr); | ||
6658 | |||
6405 | perf_event_init_all_cpus(); | 6659 | perf_event_init_all_cpus(); |
6406 | init_srcu_struct(&pmus_srcu); | 6660 | init_srcu_struct(&pmus_srcu); |
6407 | perf_pmu_register(&perf_swevent); | 6661 | perf_pmu_register(&perf_swevent, "software", PERF_TYPE_SOFTWARE); |
6408 | perf_pmu_register(&perf_cpu_clock); | 6662 | perf_pmu_register(&perf_cpu_clock, NULL, -1); |
6409 | perf_pmu_register(&perf_task_clock); | 6663 | perf_pmu_register(&perf_task_clock, NULL, -1); |
6410 | perf_tp_register(); | 6664 | perf_tp_register(); |
6411 | perf_cpu_notifier(perf_cpu_notify); | 6665 | perf_cpu_notifier(perf_cpu_notify); |
6666 | register_reboot_notifier(&perf_reboot_notifier); | ||
6412 | 6667 | ||
6413 | ret = init_hw_breakpoint(); | 6668 | ret = init_hw_breakpoint(); |
6414 | WARN(ret, "hw_breakpoint initialization failed with: %d", ret); | 6669 | WARN(ret, "hw_breakpoint initialization failed with: %d", ret); |
6415 | } | 6670 | } |
6671 | |||
6672 | static int __init perf_event_sysfs_init(void) | ||
6673 | { | ||
6674 | struct pmu *pmu; | ||
6675 | int ret; | ||
6676 | |||
6677 | mutex_lock(&pmus_lock); | ||
6678 | |||
6679 | ret = bus_register(&pmu_bus); | ||
6680 | if (ret) | ||
6681 | goto unlock; | ||
6682 | |||
6683 | list_for_each_entry(pmu, &pmus, entry) { | ||
6684 | if (!pmu->name || pmu->type < 0) | ||
6685 | continue; | ||
6686 | |||
6687 | ret = pmu_dev_alloc(pmu); | ||
6688 | WARN(ret, "Failed to register pmu: %s, reason %d\n", pmu->name, ret); | ||
6689 | } | ||
6690 | pmu_bus_running = 1; | ||
6691 | ret = 0; | ||
6692 | |||
6693 | unlock: | ||
6694 | mutex_unlock(&pmus_lock); | ||
6695 | |||
6696 | return ret; | ||
6697 | } | ||
6698 | device_initcall(perf_event_sysfs_init); | ||
diff --git a/kernel/power/suspend.c b/kernel/power/suspend.c index ecf770509d0..031d5e3a619 100644 --- a/kernel/power/suspend.c +++ b/kernel/power/suspend.c | |||
@@ -22,6 +22,7 @@ | |||
22 | #include <linux/mm.h> | 22 | #include <linux/mm.h> |
23 | #include <linux/slab.h> | 23 | #include <linux/slab.h> |
24 | #include <linux/suspend.h> | 24 | #include <linux/suspend.h> |
25 | #include <trace/events/power.h> | ||
25 | 26 | ||
26 | #include "power.h" | 27 | #include "power.h" |
27 | 28 | ||
@@ -201,6 +202,7 @@ int suspend_devices_and_enter(suspend_state_t state) | |||
201 | if (!suspend_ops) | 202 | if (!suspend_ops) |
202 | return -ENOSYS; | 203 | return -ENOSYS; |
203 | 204 | ||
205 | trace_machine_suspend(state); | ||
204 | if (suspend_ops->begin) { | 206 | if (suspend_ops->begin) { |
205 | error = suspend_ops->begin(state); | 207 | error = suspend_ops->begin(state); |
206 | if (error) | 208 | if (error) |
@@ -229,6 +231,7 @@ int suspend_devices_and_enter(suspend_state_t state) | |||
229 | Close: | 231 | Close: |
230 | if (suspend_ops->end) | 232 | if (suspend_ops->end) |
231 | suspend_ops->end(); | 233 | suspend_ops->end(); |
234 | trace_machine_suspend(PWR_EVENT_EXIT); | ||
232 | return error; | 235 | return error; |
233 | 236 | ||
234 | Recover_platform: | 237 | Recover_platform: |
diff --git a/kernel/rcutiny.c b/kernel/rcutiny.c index d806735342a..03449372474 100644 --- a/kernel/rcutiny.c +++ b/kernel/rcutiny.c | |||
@@ -36,31 +36,16 @@ | |||
36 | #include <linux/time.h> | 36 | #include <linux/time.h> |
37 | #include <linux/cpu.h> | 37 | #include <linux/cpu.h> |
38 | 38 | ||
39 | /* Global control variables for rcupdate callback mechanism. */ | 39 | /* Controls for rcu_kthread() kthread, replacing RCU_SOFTIRQ used previously. */ |
40 | struct rcu_ctrlblk { | 40 | static struct task_struct *rcu_kthread_task; |
41 | struct rcu_head *rcucblist; /* List of pending callbacks (CBs). */ | 41 | static DECLARE_WAIT_QUEUE_HEAD(rcu_kthread_wq); |
42 | struct rcu_head **donetail; /* ->next pointer of last "done" CB. */ | 42 | static unsigned long have_rcu_kthread_work; |
43 | struct rcu_head **curtail; /* ->next pointer of last CB. */ | 43 | static void invoke_rcu_kthread(void); |
44 | }; | ||
45 | |||
46 | /* Definition for rcupdate control block. */ | ||
47 | static struct rcu_ctrlblk rcu_sched_ctrlblk = { | ||
48 | .donetail = &rcu_sched_ctrlblk.rcucblist, | ||
49 | .curtail = &rcu_sched_ctrlblk.rcucblist, | ||
50 | }; | ||
51 | |||
52 | static struct rcu_ctrlblk rcu_bh_ctrlblk = { | ||
53 | .donetail = &rcu_bh_ctrlblk.rcucblist, | ||
54 | .curtail = &rcu_bh_ctrlblk.rcucblist, | ||
55 | }; | ||
56 | |||
57 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | ||
58 | int rcu_scheduler_active __read_mostly; | ||
59 | EXPORT_SYMBOL_GPL(rcu_scheduler_active); | ||
60 | #endif /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */ | ||
61 | 44 | ||
62 | /* Forward declarations for rcutiny_plugin.h. */ | 45 | /* Forward declarations for rcutiny_plugin.h. */ |
63 | static void __rcu_process_callbacks(struct rcu_ctrlblk *rcp); | 46 | struct rcu_ctrlblk; |
47 | static void rcu_process_callbacks(struct rcu_ctrlblk *rcp); | ||
48 | static int rcu_kthread(void *arg); | ||
64 | static void __call_rcu(struct rcu_head *head, | 49 | static void __call_rcu(struct rcu_head *head, |
65 | void (*func)(struct rcu_head *rcu), | 50 | void (*func)(struct rcu_head *rcu), |
66 | struct rcu_ctrlblk *rcp); | 51 | struct rcu_ctrlblk *rcp); |
@@ -123,7 +108,7 @@ void rcu_sched_qs(int cpu) | |||
123 | { | 108 | { |
124 | if (rcu_qsctr_help(&rcu_sched_ctrlblk) + | 109 | if (rcu_qsctr_help(&rcu_sched_ctrlblk) + |
125 | rcu_qsctr_help(&rcu_bh_ctrlblk)) | 110 | rcu_qsctr_help(&rcu_bh_ctrlblk)) |
126 | raise_softirq(RCU_SOFTIRQ); | 111 | invoke_rcu_kthread(); |
127 | } | 112 | } |
128 | 113 | ||
129 | /* | 114 | /* |
@@ -132,7 +117,7 @@ void rcu_sched_qs(int cpu) | |||
132 | void rcu_bh_qs(int cpu) | 117 | void rcu_bh_qs(int cpu) |
133 | { | 118 | { |
134 | if (rcu_qsctr_help(&rcu_bh_ctrlblk)) | 119 | if (rcu_qsctr_help(&rcu_bh_ctrlblk)) |
135 | raise_softirq(RCU_SOFTIRQ); | 120 | invoke_rcu_kthread(); |
136 | } | 121 | } |
137 | 122 | ||
138 | /* | 123 | /* |
@@ -152,13 +137,14 @@ void rcu_check_callbacks(int cpu, int user) | |||
152 | } | 137 | } |
153 | 138 | ||
154 | /* | 139 | /* |
155 | * Helper function for rcu_process_callbacks() that operates on the | 140 | * Invoke the RCU callbacks on the specified rcu_ctrlkblk structure |
156 | * specified rcu_ctrlkblk structure. | 141 | * whose grace period has elapsed. |
157 | */ | 142 | */ |
158 | static void __rcu_process_callbacks(struct rcu_ctrlblk *rcp) | 143 | static void rcu_process_callbacks(struct rcu_ctrlblk *rcp) |
159 | { | 144 | { |
160 | struct rcu_head *next, *list; | 145 | struct rcu_head *next, *list; |
161 | unsigned long flags; | 146 | unsigned long flags; |
147 | RCU_TRACE(int cb_count = 0); | ||
162 | 148 | ||
163 | /* If no RCU callbacks ready to invoke, just return. */ | 149 | /* If no RCU callbacks ready to invoke, just return. */ |
164 | if (&rcp->rcucblist == rcp->donetail) | 150 | if (&rcp->rcucblist == rcp->donetail) |
@@ -180,19 +166,58 @@ static void __rcu_process_callbacks(struct rcu_ctrlblk *rcp) | |||
180 | next = list->next; | 166 | next = list->next; |
181 | prefetch(next); | 167 | prefetch(next); |
182 | debug_rcu_head_unqueue(list); | 168 | debug_rcu_head_unqueue(list); |
169 | local_bh_disable(); | ||
183 | list->func(list); | 170 | list->func(list); |
171 | local_bh_enable(); | ||
184 | list = next; | 172 | list = next; |
173 | RCU_TRACE(cb_count++); | ||
185 | } | 174 | } |
175 | RCU_TRACE(rcu_trace_sub_qlen(rcp, cb_count)); | ||
186 | } | 176 | } |
187 | 177 | ||
188 | /* | 178 | /* |
189 | * Invoke any callbacks whose grace period has completed. | 179 | * This kthread invokes RCU callbacks whose grace periods have |
180 | * elapsed. It is awakened as needed, and takes the place of the | ||
181 | * RCU_SOFTIRQ that was used previously for this purpose. | ||
182 | * This is a kthread, but it is never stopped, at least not until | ||
183 | * the system goes down. | ||
190 | */ | 184 | */ |
191 | static void rcu_process_callbacks(struct softirq_action *unused) | 185 | static int rcu_kthread(void *arg) |
192 | { | 186 | { |
193 | __rcu_process_callbacks(&rcu_sched_ctrlblk); | 187 | unsigned long work; |
194 | __rcu_process_callbacks(&rcu_bh_ctrlblk); | 188 | unsigned long morework; |
195 | rcu_preempt_process_callbacks(); | 189 | unsigned long flags; |
190 | |||
191 | for (;;) { | ||
192 | wait_event(rcu_kthread_wq, have_rcu_kthread_work != 0); | ||
193 | morework = rcu_boost(); | ||
194 | local_irq_save(flags); | ||
195 | work = have_rcu_kthread_work; | ||
196 | have_rcu_kthread_work = morework; | ||
197 | local_irq_restore(flags); | ||
198 | if (work) { | ||
199 | rcu_process_callbacks(&rcu_sched_ctrlblk); | ||
200 | rcu_process_callbacks(&rcu_bh_ctrlblk); | ||
201 | rcu_preempt_process_callbacks(); | ||
202 | } | ||
203 | schedule_timeout_interruptible(1); /* Leave CPU for others. */ | ||
204 | } | ||
205 | |||
206 | return 0; /* Not reached, but needed to shut gcc up. */ | ||
207 | } | ||
208 | |||
209 | /* | ||
210 | * Wake up rcu_kthread() to process callbacks now eligible for invocation | ||
211 | * or to boost readers. | ||
212 | */ | ||
213 | static void invoke_rcu_kthread(void) | ||
214 | { | ||
215 | unsigned long flags; | ||
216 | |||
217 | local_irq_save(flags); | ||
218 | have_rcu_kthread_work = 1; | ||
219 | wake_up(&rcu_kthread_wq); | ||
220 | local_irq_restore(flags); | ||
196 | } | 221 | } |
197 | 222 | ||
198 | /* | 223 | /* |
@@ -230,6 +255,7 @@ static void __call_rcu(struct rcu_head *head, | |||
230 | local_irq_save(flags); | 255 | local_irq_save(flags); |
231 | *rcp->curtail = head; | 256 | *rcp->curtail = head; |
232 | rcp->curtail = &head->next; | 257 | rcp->curtail = &head->next; |
258 | RCU_TRACE(rcp->qlen++); | ||
233 | local_irq_restore(flags); | 259 | local_irq_restore(flags); |
234 | } | 260 | } |
235 | 261 | ||
@@ -282,7 +308,16 @@ void rcu_barrier_sched(void) | |||
282 | } | 308 | } |
283 | EXPORT_SYMBOL_GPL(rcu_barrier_sched); | 309 | EXPORT_SYMBOL_GPL(rcu_barrier_sched); |
284 | 310 | ||
285 | void __init rcu_init(void) | 311 | /* |
312 | * Spawn the kthread that invokes RCU callbacks. | ||
313 | */ | ||
314 | static int __init rcu_spawn_kthreads(void) | ||
286 | { | 315 | { |
287 | open_softirq(RCU_SOFTIRQ, rcu_process_callbacks); | 316 | struct sched_param sp; |
317 | |||
318 | rcu_kthread_task = kthread_run(rcu_kthread, NULL, "rcu_kthread"); | ||
319 | sp.sched_priority = RCU_BOOST_PRIO; | ||
320 | sched_setscheduler_nocheck(rcu_kthread_task, SCHED_FIFO, &sp); | ||
321 | return 0; | ||
288 | } | 322 | } |
323 | early_initcall(rcu_spawn_kthreads); | ||
diff --git a/kernel/rcutiny_plugin.h b/kernel/rcutiny_plugin.h index 6ceca4f745f..015abaea962 100644 --- a/kernel/rcutiny_plugin.h +++ b/kernel/rcutiny_plugin.h | |||
@@ -22,6 +22,40 @@ | |||
22 | * Author: Paul E. McKenney <paulmck@linux.vnet.ibm.com> | 22 | * Author: Paul E. McKenney <paulmck@linux.vnet.ibm.com> |
23 | */ | 23 | */ |
24 | 24 | ||
25 | #include <linux/kthread.h> | ||
26 | #include <linux/debugfs.h> | ||
27 | #include <linux/seq_file.h> | ||
28 | |||
29 | #ifdef CONFIG_RCU_TRACE | ||
30 | #define RCU_TRACE(stmt) stmt | ||
31 | #else /* #ifdef CONFIG_RCU_TRACE */ | ||
32 | #define RCU_TRACE(stmt) | ||
33 | #endif /* #else #ifdef CONFIG_RCU_TRACE */ | ||
34 | |||
35 | /* Global control variables for rcupdate callback mechanism. */ | ||
36 | struct rcu_ctrlblk { | ||
37 | struct rcu_head *rcucblist; /* List of pending callbacks (CBs). */ | ||
38 | struct rcu_head **donetail; /* ->next pointer of last "done" CB. */ | ||
39 | struct rcu_head **curtail; /* ->next pointer of last CB. */ | ||
40 | RCU_TRACE(long qlen); /* Number of pending CBs. */ | ||
41 | }; | ||
42 | |||
43 | /* Definition for rcupdate control block. */ | ||
44 | static struct rcu_ctrlblk rcu_sched_ctrlblk = { | ||
45 | .donetail = &rcu_sched_ctrlblk.rcucblist, | ||
46 | .curtail = &rcu_sched_ctrlblk.rcucblist, | ||
47 | }; | ||
48 | |||
49 | static struct rcu_ctrlblk rcu_bh_ctrlblk = { | ||
50 | .donetail = &rcu_bh_ctrlblk.rcucblist, | ||
51 | .curtail = &rcu_bh_ctrlblk.rcucblist, | ||
52 | }; | ||
53 | |||
54 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | ||
55 | int rcu_scheduler_active __read_mostly; | ||
56 | EXPORT_SYMBOL_GPL(rcu_scheduler_active); | ||
57 | #endif /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */ | ||
58 | |||
25 | #ifdef CONFIG_TINY_PREEMPT_RCU | 59 | #ifdef CONFIG_TINY_PREEMPT_RCU |
26 | 60 | ||
27 | #include <linux/delay.h> | 61 | #include <linux/delay.h> |
@@ -46,17 +80,45 @@ struct rcu_preempt_ctrlblk { | |||
46 | struct list_head *gp_tasks; | 80 | struct list_head *gp_tasks; |
47 | /* Pointer to the first task blocking the */ | 81 | /* Pointer to the first task blocking the */ |
48 | /* current grace period, or NULL if there */ | 82 | /* current grace period, or NULL if there */ |
49 | /* is not such task. */ | 83 | /* is no such task. */ |
50 | struct list_head *exp_tasks; | 84 | struct list_head *exp_tasks; |
51 | /* Pointer to first task blocking the */ | 85 | /* Pointer to first task blocking the */ |
52 | /* current expedited grace period, or NULL */ | 86 | /* current expedited grace period, or NULL */ |
53 | /* if there is no such task. If there */ | 87 | /* if there is no such task. If there */ |
54 | /* is no current expedited grace period, */ | 88 | /* is no current expedited grace period, */ |
55 | /* then there cannot be any such task. */ | 89 | /* then there cannot be any such task. */ |
90 | #ifdef CONFIG_RCU_BOOST | ||
91 | struct list_head *boost_tasks; | ||
92 | /* Pointer to first task that needs to be */ | ||
93 | /* priority-boosted, or NULL if no priority */ | ||
94 | /* boosting is needed. If there is no */ | ||
95 | /* current or expedited grace period, there */ | ||
96 | /* can be no such task. */ | ||
97 | #endif /* #ifdef CONFIG_RCU_BOOST */ | ||
56 | u8 gpnum; /* Current grace period. */ | 98 | u8 gpnum; /* Current grace period. */ |
57 | u8 gpcpu; /* Last grace period blocked by the CPU. */ | 99 | u8 gpcpu; /* Last grace period blocked by the CPU. */ |
58 | u8 completed; /* Last grace period completed. */ | 100 | u8 completed; /* Last grace period completed. */ |
59 | /* If all three are equal, RCU is idle. */ | 101 | /* If all three are equal, RCU is idle. */ |
102 | #ifdef CONFIG_RCU_BOOST | ||
103 | s8 boosted_this_gp; /* Has boosting already happened? */ | ||
104 | unsigned long boost_time; /* When to start boosting (jiffies) */ | ||
105 | #endif /* #ifdef CONFIG_RCU_BOOST */ | ||
106 | #ifdef CONFIG_RCU_TRACE | ||
107 | unsigned long n_grace_periods; | ||
108 | #ifdef CONFIG_RCU_BOOST | ||
109 | unsigned long n_tasks_boosted; | ||
110 | unsigned long n_exp_boosts; | ||
111 | unsigned long n_normal_boosts; | ||
112 | unsigned long n_normal_balk_blkd_tasks; | ||
113 | unsigned long n_normal_balk_gp_tasks; | ||
114 | unsigned long n_normal_balk_boost_tasks; | ||
115 | unsigned long n_normal_balk_boosted; | ||
116 | unsigned long n_normal_balk_notyet; | ||
117 | unsigned long n_normal_balk_nos; | ||
118 | unsigned long n_exp_balk_blkd_tasks; | ||
119 | unsigned long n_exp_balk_nos; | ||
120 | #endif /* #ifdef CONFIG_RCU_BOOST */ | ||
121 | #endif /* #ifdef CONFIG_RCU_TRACE */ | ||
60 | }; | 122 | }; |
61 | 123 | ||
62 | static struct rcu_preempt_ctrlblk rcu_preempt_ctrlblk = { | 124 | static struct rcu_preempt_ctrlblk rcu_preempt_ctrlblk = { |
@@ -122,6 +184,210 @@ static int rcu_preempt_gp_in_progress(void) | |||
122 | } | 184 | } |
123 | 185 | ||
124 | /* | 186 | /* |
187 | * Advance a ->blkd_tasks-list pointer to the next entry, instead | ||
188 | * returning NULL if at the end of the list. | ||
189 | */ | ||
190 | static struct list_head *rcu_next_node_entry(struct task_struct *t) | ||
191 | { | ||
192 | struct list_head *np; | ||
193 | |||
194 | np = t->rcu_node_entry.next; | ||
195 | if (np == &rcu_preempt_ctrlblk.blkd_tasks) | ||
196 | np = NULL; | ||
197 | return np; | ||
198 | } | ||
199 | |||
200 | #ifdef CONFIG_RCU_TRACE | ||
201 | |||
202 | #ifdef CONFIG_RCU_BOOST | ||
203 | static void rcu_initiate_boost_trace(void); | ||
204 | static void rcu_initiate_exp_boost_trace(void); | ||
205 | #endif /* #ifdef CONFIG_RCU_BOOST */ | ||
206 | |||
207 | /* | ||
208 | * Dump additional statistice for TINY_PREEMPT_RCU. | ||
209 | */ | ||
210 | static void show_tiny_preempt_stats(struct seq_file *m) | ||
211 | { | ||
212 | seq_printf(m, "rcu_preempt: qlen=%ld gp=%lu g%u/p%u/c%u tasks=%c%c%c\n", | ||
213 | rcu_preempt_ctrlblk.rcb.qlen, | ||
214 | rcu_preempt_ctrlblk.n_grace_periods, | ||
215 | rcu_preempt_ctrlblk.gpnum, | ||
216 | rcu_preempt_ctrlblk.gpcpu, | ||
217 | rcu_preempt_ctrlblk.completed, | ||
218 | "T."[list_empty(&rcu_preempt_ctrlblk.blkd_tasks)], | ||
219 | "N."[!rcu_preempt_ctrlblk.gp_tasks], | ||
220 | "E."[!rcu_preempt_ctrlblk.exp_tasks]); | ||
221 | #ifdef CONFIG_RCU_BOOST | ||
222 | seq_printf(m, " ttb=%c btg=", | ||
223 | "B."[!rcu_preempt_ctrlblk.boost_tasks]); | ||
224 | switch (rcu_preempt_ctrlblk.boosted_this_gp) { | ||
225 | case -1: | ||
226 | seq_puts(m, "exp"); | ||
227 | break; | ||
228 | case 0: | ||
229 | seq_puts(m, "no"); | ||
230 | break; | ||
231 | case 1: | ||
232 | seq_puts(m, "begun"); | ||
233 | break; | ||
234 | case 2: | ||
235 | seq_puts(m, "done"); | ||
236 | break; | ||
237 | default: | ||
238 | seq_printf(m, "?%d?", rcu_preempt_ctrlblk.boosted_this_gp); | ||
239 | } | ||
240 | seq_printf(m, " ntb=%lu neb=%lu nnb=%lu j=%04x bt=%04x\n", | ||
241 | rcu_preempt_ctrlblk.n_tasks_boosted, | ||
242 | rcu_preempt_ctrlblk.n_exp_boosts, | ||
243 | rcu_preempt_ctrlblk.n_normal_boosts, | ||
244 | (int)(jiffies & 0xffff), | ||
245 | (int)(rcu_preempt_ctrlblk.boost_time & 0xffff)); | ||
246 | seq_printf(m, " %s: nt=%lu gt=%lu bt=%lu b=%lu ny=%lu nos=%lu\n", | ||
247 | "normal balk", | ||
248 | rcu_preempt_ctrlblk.n_normal_balk_blkd_tasks, | ||
249 | rcu_preempt_ctrlblk.n_normal_balk_gp_tasks, | ||
250 | rcu_preempt_ctrlblk.n_normal_balk_boost_tasks, | ||
251 | rcu_preempt_ctrlblk.n_normal_balk_boosted, | ||
252 | rcu_preempt_ctrlblk.n_normal_balk_notyet, | ||
253 | rcu_preempt_ctrlblk.n_normal_balk_nos); | ||
254 | seq_printf(m, " exp balk: bt=%lu nos=%lu\n", | ||
255 | rcu_preempt_ctrlblk.n_exp_balk_blkd_tasks, | ||
256 | rcu_preempt_ctrlblk.n_exp_balk_nos); | ||
257 | #endif /* #ifdef CONFIG_RCU_BOOST */ | ||
258 | } | ||
259 | |||
260 | #endif /* #ifdef CONFIG_RCU_TRACE */ | ||
261 | |||
262 | #ifdef CONFIG_RCU_BOOST | ||
263 | |||
264 | #include "rtmutex_common.h" | ||
265 | |||
266 | /* | ||
267 | * Carry out RCU priority boosting on the task indicated by ->boost_tasks, | ||
268 | * and advance ->boost_tasks to the next task in the ->blkd_tasks list. | ||
269 | */ | ||
270 | static int rcu_boost(void) | ||
271 | { | ||
272 | unsigned long flags; | ||
273 | struct rt_mutex mtx; | ||
274 | struct list_head *np; | ||
275 | struct task_struct *t; | ||
276 | |||
277 | if (rcu_preempt_ctrlblk.boost_tasks == NULL) | ||
278 | return 0; /* Nothing to boost. */ | ||
279 | raw_local_irq_save(flags); | ||
280 | rcu_preempt_ctrlblk.boosted_this_gp++; | ||
281 | t = container_of(rcu_preempt_ctrlblk.boost_tasks, struct task_struct, | ||
282 | rcu_node_entry); | ||
283 | np = rcu_next_node_entry(t); | ||
284 | rt_mutex_init_proxy_locked(&mtx, t); | ||
285 | t->rcu_boost_mutex = &mtx; | ||
286 | t->rcu_read_unlock_special |= RCU_READ_UNLOCK_BOOSTED; | ||
287 | raw_local_irq_restore(flags); | ||
288 | rt_mutex_lock(&mtx); | ||
289 | RCU_TRACE(rcu_preempt_ctrlblk.n_tasks_boosted++); | ||
290 | rcu_preempt_ctrlblk.boosted_this_gp++; | ||
291 | rt_mutex_unlock(&mtx); | ||
292 | return rcu_preempt_ctrlblk.boost_tasks != NULL; | ||
293 | } | ||
294 | |||
295 | /* | ||
296 | * Check to see if it is now time to start boosting RCU readers blocking | ||
297 | * the current grace period, and, if so, tell the rcu_kthread_task to | ||
298 | * start boosting them. If there is an expedited boost in progress, | ||
299 | * we wait for it to complete. | ||
300 | * | ||
301 | * If there are no blocked readers blocking the current grace period, | ||
302 | * return 0 to let the caller know, otherwise return 1. Note that this | ||
303 | * return value is independent of whether or not boosting was done. | ||
304 | */ | ||
305 | static int rcu_initiate_boost(void) | ||
306 | { | ||
307 | if (!rcu_preempt_blocked_readers_cgp()) { | ||
308 | RCU_TRACE(rcu_preempt_ctrlblk.n_normal_balk_blkd_tasks++); | ||
309 | return 0; | ||
310 | } | ||
311 | if (rcu_preempt_ctrlblk.gp_tasks != NULL && | ||
312 | rcu_preempt_ctrlblk.boost_tasks == NULL && | ||
313 | rcu_preempt_ctrlblk.boosted_this_gp == 0 && | ||
314 | ULONG_CMP_GE(jiffies, rcu_preempt_ctrlblk.boost_time)) { | ||
315 | rcu_preempt_ctrlblk.boost_tasks = rcu_preempt_ctrlblk.gp_tasks; | ||
316 | invoke_rcu_kthread(); | ||
317 | RCU_TRACE(rcu_preempt_ctrlblk.n_normal_boosts++); | ||
318 | } else | ||
319 | RCU_TRACE(rcu_initiate_boost_trace()); | ||
320 | return 1; | ||
321 | } | ||
322 | |||
323 | /* | ||
324 | * Initiate boosting for an expedited grace period. | ||
325 | */ | ||
326 | static void rcu_initiate_expedited_boost(void) | ||
327 | { | ||
328 | unsigned long flags; | ||
329 | |||
330 | raw_local_irq_save(flags); | ||
331 | if (!list_empty(&rcu_preempt_ctrlblk.blkd_tasks)) { | ||
332 | rcu_preempt_ctrlblk.boost_tasks = | ||
333 | rcu_preempt_ctrlblk.blkd_tasks.next; | ||
334 | rcu_preempt_ctrlblk.boosted_this_gp = -1; | ||
335 | invoke_rcu_kthread(); | ||
336 | RCU_TRACE(rcu_preempt_ctrlblk.n_exp_boosts++); | ||
337 | } else | ||
338 | RCU_TRACE(rcu_initiate_exp_boost_trace()); | ||
339 | raw_local_irq_restore(flags); | ||
340 | } | ||
341 | |||
342 | #define RCU_BOOST_DELAY_JIFFIES DIV_ROUND_UP(CONFIG_RCU_BOOST_DELAY * HZ, 1000); | ||
343 | |||
344 | /* | ||
345 | * Do priority-boost accounting for the start of a new grace period. | ||
346 | */ | ||
347 | static void rcu_preempt_boost_start_gp(void) | ||
348 | { | ||
349 | rcu_preempt_ctrlblk.boost_time = jiffies + RCU_BOOST_DELAY_JIFFIES; | ||
350 | if (rcu_preempt_ctrlblk.boosted_this_gp > 0) | ||
351 | rcu_preempt_ctrlblk.boosted_this_gp = 0; | ||
352 | } | ||
353 | |||
354 | #else /* #ifdef CONFIG_RCU_BOOST */ | ||
355 | |||
356 | /* | ||
357 | * If there is no RCU priority boosting, we don't boost. | ||
358 | */ | ||
359 | static int rcu_boost(void) | ||
360 | { | ||
361 | return 0; | ||
362 | } | ||
363 | |||
364 | /* | ||
365 | * If there is no RCU priority boosting, we don't initiate boosting, | ||
366 | * but we do indicate whether there are blocked readers blocking the | ||
367 | * current grace period. | ||
368 | */ | ||
369 | static int rcu_initiate_boost(void) | ||
370 | { | ||
371 | return rcu_preempt_blocked_readers_cgp(); | ||
372 | } | ||
373 | |||
374 | /* | ||
375 | * If there is no RCU priority boosting, we don't initiate expedited boosting. | ||
376 | */ | ||
377 | static void rcu_initiate_expedited_boost(void) | ||
378 | { | ||
379 | } | ||
380 | |||
381 | /* | ||
382 | * If there is no RCU priority boosting, nothing to do at grace-period start. | ||
383 | */ | ||
384 | static void rcu_preempt_boost_start_gp(void) | ||
385 | { | ||
386 | } | ||
387 | |||
388 | #endif /* else #ifdef CONFIG_RCU_BOOST */ | ||
389 | |||
390 | /* | ||
125 | * Record a preemptible-RCU quiescent state for the specified CPU. Note | 391 | * Record a preemptible-RCU quiescent state for the specified CPU. Note |
126 | * that this just means that the task currently running on the CPU is | 392 | * that this just means that the task currently running on the CPU is |
127 | * in a quiescent state. There might be any number of tasks blocked | 393 | * in a quiescent state. There might be any number of tasks blocked |
@@ -148,11 +414,14 @@ static void rcu_preempt_cpu_qs(void) | |||
148 | rcu_preempt_ctrlblk.gpcpu = rcu_preempt_ctrlblk.gpnum; | 414 | rcu_preempt_ctrlblk.gpcpu = rcu_preempt_ctrlblk.gpnum; |
149 | current->rcu_read_unlock_special &= ~RCU_READ_UNLOCK_NEED_QS; | 415 | current->rcu_read_unlock_special &= ~RCU_READ_UNLOCK_NEED_QS; |
150 | 416 | ||
417 | /* If there is no GP then there is nothing more to do. */ | ||
418 | if (!rcu_preempt_gp_in_progress()) | ||
419 | return; | ||
151 | /* | 420 | /* |
152 | * If there is no GP, or if blocked readers are still blocking GP, | 421 | * Check up on boosting. If there are no readers blocking the |
153 | * then there is nothing more to do. | 422 | * current grace period, leave. |
154 | */ | 423 | */ |
155 | if (!rcu_preempt_gp_in_progress() || rcu_preempt_blocked_readers_cgp()) | 424 | if (rcu_initiate_boost()) |
156 | return; | 425 | return; |
157 | 426 | ||
158 | /* Advance callbacks. */ | 427 | /* Advance callbacks. */ |
@@ -164,9 +433,9 @@ static void rcu_preempt_cpu_qs(void) | |||
164 | if (!rcu_preempt_blocked_readers_any()) | 433 | if (!rcu_preempt_blocked_readers_any()) |
165 | rcu_preempt_ctrlblk.rcb.donetail = rcu_preempt_ctrlblk.nexttail; | 434 | rcu_preempt_ctrlblk.rcb.donetail = rcu_preempt_ctrlblk.nexttail; |
166 | 435 | ||
167 | /* If there are done callbacks, make RCU_SOFTIRQ process them. */ | 436 | /* If there are done callbacks, cause them to be invoked. */ |
168 | if (*rcu_preempt_ctrlblk.rcb.donetail != NULL) | 437 | if (*rcu_preempt_ctrlblk.rcb.donetail != NULL) |
169 | raise_softirq(RCU_SOFTIRQ); | 438 | invoke_rcu_kthread(); |
170 | } | 439 | } |
171 | 440 | ||
172 | /* | 441 | /* |
@@ -178,12 +447,16 @@ static void rcu_preempt_start_gp(void) | |||
178 | 447 | ||
179 | /* Official start of GP. */ | 448 | /* Official start of GP. */ |
180 | rcu_preempt_ctrlblk.gpnum++; | 449 | rcu_preempt_ctrlblk.gpnum++; |
450 | RCU_TRACE(rcu_preempt_ctrlblk.n_grace_periods++); | ||
181 | 451 | ||
182 | /* Any blocked RCU readers block new GP. */ | 452 | /* Any blocked RCU readers block new GP. */ |
183 | if (rcu_preempt_blocked_readers_any()) | 453 | if (rcu_preempt_blocked_readers_any()) |
184 | rcu_preempt_ctrlblk.gp_tasks = | 454 | rcu_preempt_ctrlblk.gp_tasks = |
185 | rcu_preempt_ctrlblk.blkd_tasks.next; | 455 | rcu_preempt_ctrlblk.blkd_tasks.next; |
186 | 456 | ||
457 | /* Set up for RCU priority boosting. */ | ||
458 | rcu_preempt_boost_start_gp(); | ||
459 | |||
187 | /* If there is no running reader, CPU is done with GP. */ | 460 | /* If there is no running reader, CPU is done with GP. */ |
188 | if (!rcu_preempt_running_reader()) | 461 | if (!rcu_preempt_running_reader()) |
189 | rcu_preempt_cpu_qs(); | 462 | rcu_preempt_cpu_qs(); |
@@ -304,14 +577,16 @@ static void rcu_read_unlock_special(struct task_struct *t) | |||
304 | */ | 577 | */ |
305 | empty = !rcu_preempt_blocked_readers_cgp(); | 578 | empty = !rcu_preempt_blocked_readers_cgp(); |
306 | empty_exp = rcu_preempt_ctrlblk.exp_tasks == NULL; | 579 | empty_exp = rcu_preempt_ctrlblk.exp_tasks == NULL; |
307 | np = t->rcu_node_entry.next; | 580 | np = rcu_next_node_entry(t); |
308 | if (np == &rcu_preempt_ctrlblk.blkd_tasks) | ||
309 | np = NULL; | ||
310 | list_del(&t->rcu_node_entry); | 581 | list_del(&t->rcu_node_entry); |
311 | if (&t->rcu_node_entry == rcu_preempt_ctrlblk.gp_tasks) | 582 | if (&t->rcu_node_entry == rcu_preempt_ctrlblk.gp_tasks) |
312 | rcu_preempt_ctrlblk.gp_tasks = np; | 583 | rcu_preempt_ctrlblk.gp_tasks = np; |
313 | if (&t->rcu_node_entry == rcu_preempt_ctrlblk.exp_tasks) | 584 | if (&t->rcu_node_entry == rcu_preempt_ctrlblk.exp_tasks) |
314 | rcu_preempt_ctrlblk.exp_tasks = np; | 585 | rcu_preempt_ctrlblk.exp_tasks = np; |
586 | #ifdef CONFIG_RCU_BOOST | ||
587 | if (&t->rcu_node_entry == rcu_preempt_ctrlblk.boost_tasks) | ||
588 | rcu_preempt_ctrlblk.boost_tasks = np; | ||
589 | #endif /* #ifdef CONFIG_RCU_BOOST */ | ||
315 | INIT_LIST_HEAD(&t->rcu_node_entry); | 590 | INIT_LIST_HEAD(&t->rcu_node_entry); |
316 | 591 | ||
317 | /* | 592 | /* |
@@ -331,6 +606,14 @@ static void rcu_read_unlock_special(struct task_struct *t) | |||
331 | if (!empty_exp && rcu_preempt_ctrlblk.exp_tasks == NULL) | 606 | if (!empty_exp && rcu_preempt_ctrlblk.exp_tasks == NULL) |
332 | rcu_report_exp_done(); | 607 | rcu_report_exp_done(); |
333 | } | 608 | } |
609 | #ifdef CONFIG_RCU_BOOST | ||
610 | /* Unboost self if was boosted. */ | ||
611 | if (special & RCU_READ_UNLOCK_BOOSTED) { | ||
612 | t->rcu_read_unlock_special &= ~RCU_READ_UNLOCK_BOOSTED; | ||
613 | rt_mutex_unlock(t->rcu_boost_mutex); | ||
614 | t->rcu_boost_mutex = NULL; | ||
615 | } | ||
616 | #endif /* #ifdef CONFIG_RCU_BOOST */ | ||
334 | local_irq_restore(flags); | 617 | local_irq_restore(flags); |
335 | } | 618 | } |
336 | 619 | ||
@@ -374,7 +657,7 @@ static void rcu_preempt_check_callbacks(void) | |||
374 | rcu_preempt_cpu_qs(); | 657 | rcu_preempt_cpu_qs(); |
375 | if (&rcu_preempt_ctrlblk.rcb.rcucblist != | 658 | if (&rcu_preempt_ctrlblk.rcb.rcucblist != |
376 | rcu_preempt_ctrlblk.rcb.donetail) | 659 | rcu_preempt_ctrlblk.rcb.donetail) |
377 | raise_softirq(RCU_SOFTIRQ); | 660 | invoke_rcu_kthread(); |
378 | if (rcu_preempt_gp_in_progress() && | 661 | if (rcu_preempt_gp_in_progress() && |
379 | rcu_cpu_blocking_cur_gp() && | 662 | rcu_cpu_blocking_cur_gp() && |
380 | rcu_preempt_running_reader()) | 663 | rcu_preempt_running_reader()) |
@@ -383,7 +666,7 @@ static void rcu_preempt_check_callbacks(void) | |||
383 | 666 | ||
384 | /* | 667 | /* |
385 | * TINY_PREEMPT_RCU has an extra callback-list tail pointer to | 668 | * TINY_PREEMPT_RCU has an extra callback-list tail pointer to |
386 | * update, so this is invoked from __rcu_process_callbacks() to | 669 | * update, so this is invoked from rcu_process_callbacks() to |
387 | * handle that case. Of course, it is invoked for all flavors of | 670 | * handle that case. Of course, it is invoked for all flavors of |
388 | * RCU, but RCU callbacks can appear only on one of the lists, and | 671 | * RCU, but RCU callbacks can appear only on one of the lists, and |
389 | * neither ->nexttail nor ->donetail can possibly be NULL, so there | 672 | * neither ->nexttail nor ->donetail can possibly be NULL, so there |
@@ -400,7 +683,7 @@ static void rcu_preempt_remove_callbacks(struct rcu_ctrlblk *rcp) | |||
400 | */ | 683 | */ |
401 | static void rcu_preempt_process_callbacks(void) | 684 | static void rcu_preempt_process_callbacks(void) |
402 | { | 685 | { |
403 | __rcu_process_callbacks(&rcu_preempt_ctrlblk.rcb); | 686 | rcu_process_callbacks(&rcu_preempt_ctrlblk.rcb); |
404 | } | 687 | } |
405 | 688 | ||
406 | /* | 689 | /* |
@@ -417,6 +700,7 @@ void call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu)) | |||
417 | local_irq_save(flags); | 700 | local_irq_save(flags); |
418 | *rcu_preempt_ctrlblk.nexttail = head; | 701 | *rcu_preempt_ctrlblk.nexttail = head; |
419 | rcu_preempt_ctrlblk.nexttail = &head->next; | 702 | rcu_preempt_ctrlblk.nexttail = &head->next; |
703 | RCU_TRACE(rcu_preempt_ctrlblk.rcb.qlen++); | ||
420 | rcu_preempt_start_gp(); /* checks to see if GP needed. */ | 704 | rcu_preempt_start_gp(); /* checks to see if GP needed. */ |
421 | local_irq_restore(flags); | 705 | local_irq_restore(flags); |
422 | } | 706 | } |
@@ -532,6 +816,7 @@ void synchronize_rcu_expedited(void) | |||
532 | 816 | ||
533 | /* Wait for tail of ->blkd_tasks list to drain. */ | 817 | /* Wait for tail of ->blkd_tasks list to drain. */ |
534 | if (rcu_preempted_readers_exp()) | 818 | if (rcu_preempted_readers_exp()) |
819 | rcu_initiate_expedited_boost(); | ||
535 | wait_event(sync_rcu_preempt_exp_wq, | 820 | wait_event(sync_rcu_preempt_exp_wq, |
536 | !rcu_preempted_readers_exp()); | 821 | !rcu_preempted_readers_exp()); |
537 | 822 | ||
@@ -572,6 +857,27 @@ void exit_rcu(void) | |||
572 | 857 | ||
573 | #else /* #ifdef CONFIG_TINY_PREEMPT_RCU */ | 858 | #else /* #ifdef CONFIG_TINY_PREEMPT_RCU */ |
574 | 859 | ||
860 | #ifdef CONFIG_RCU_TRACE | ||
861 | |||
862 | /* | ||
863 | * Because preemptible RCU does not exist, it is not necessary to | ||
864 | * dump out its statistics. | ||
865 | */ | ||
866 | static void show_tiny_preempt_stats(struct seq_file *m) | ||
867 | { | ||
868 | } | ||
869 | |||
870 | #endif /* #ifdef CONFIG_RCU_TRACE */ | ||
871 | |||
872 | /* | ||
873 | * Because preemptible RCU does not exist, it is never necessary to | ||
874 | * boost preempted RCU readers. | ||
875 | */ | ||
876 | static int rcu_boost(void) | ||
877 | { | ||
878 | return 0; | ||
879 | } | ||
880 | |||
575 | /* | 881 | /* |
576 | * Because preemptible RCU does not exist, it never has any callbacks | 882 | * Because preemptible RCU does not exist, it never has any callbacks |
577 | * to check. | 883 | * to check. |
@@ -599,17 +905,116 @@ static void rcu_preempt_process_callbacks(void) | |||
599 | #endif /* #else #ifdef CONFIG_TINY_PREEMPT_RCU */ | 905 | #endif /* #else #ifdef CONFIG_TINY_PREEMPT_RCU */ |
600 | 906 | ||
601 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | 907 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
602 | |||
603 | #include <linux/kernel_stat.h> | 908 | #include <linux/kernel_stat.h> |
604 | 909 | ||
605 | /* | 910 | /* |
606 | * During boot, we forgive RCU lockdep issues. After this function is | 911 | * During boot, we forgive RCU lockdep issues. After this function is |
607 | * invoked, we start taking RCU lockdep issues seriously. | 912 | * invoked, we start taking RCU lockdep issues seriously. |
608 | */ | 913 | */ |
609 | void rcu_scheduler_starting(void) | 914 | void __init rcu_scheduler_starting(void) |
610 | { | 915 | { |
611 | WARN_ON(nr_context_switches() > 0); | 916 | WARN_ON(nr_context_switches() > 0); |
612 | rcu_scheduler_active = 1; | 917 | rcu_scheduler_active = 1; |
613 | } | 918 | } |
614 | 919 | ||
615 | #endif /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */ | 920 | #endif /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */ |
921 | |||
922 | #ifdef CONFIG_RCU_BOOST | ||
923 | #define RCU_BOOST_PRIO CONFIG_RCU_BOOST_PRIO | ||
924 | #else /* #ifdef CONFIG_RCU_BOOST */ | ||
925 | #define RCU_BOOST_PRIO 1 | ||
926 | #endif /* #else #ifdef CONFIG_RCU_BOOST */ | ||
927 | |||
928 | #ifdef CONFIG_RCU_TRACE | ||
929 | |||
930 | #ifdef CONFIG_RCU_BOOST | ||
931 | |||
932 | static void rcu_initiate_boost_trace(void) | ||
933 | { | ||
934 | if (rcu_preempt_ctrlblk.gp_tasks == NULL) | ||
935 | rcu_preempt_ctrlblk.n_normal_balk_gp_tasks++; | ||
936 | else if (rcu_preempt_ctrlblk.boost_tasks != NULL) | ||
937 | rcu_preempt_ctrlblk.n_normal_balk_boost_tasks++; | ||
938 | else if (rcu_preempt_ctrlblk.boosted_this_gp != 0) | ||
939 | rcu_preempt_ctrlblk.n_normal_balk_boosted++; | ||
940 | else if (!ULONG_CMP_GE(jiffies, rcu_preempt_ctrlblk.boost_time)) | ||
941 | rcu_preempt_ctrlblk.n_normal_balk_notyet++; | ||
942 | else | ||
943 | rcu_preempt_ctrlblk.n_normal_balk_nos++; | ||
944 | } | ||
945 | |||
946 | static void rcu_initiate_exp_boost_trace(void) | ||
947 | { | ||
948 | if (list_empty(&rcu_preempt_ctrlblk.blkd_tasks)) | ||
949 | rcu_preempt_ctrlblk.n_exp_balk_blkd_tasks++; | ||
950 | else | ||
951 | rcu_preempt_ctrlblk.n_exp_balk_nos++; | ||
952 | } | ||
953 | |||
954 | #endif /* #ifdef CONFIG_RCU_BOOST */ | ||
955 | |||
956 | static void rcu_trace_sub_qlen(struct rcu_ctrlblk *rcp, int n) | ||
957 | { | ||
958 | unsigned long flags; | ||
959 | |||
960 | raw_local_irq_save(flags); | ||
961 | rcp->qlen -= n; | ||
962 | raw_local_irq_restore(flags); | ||
963 | } | ||
964 | |||
965 | /* | ||
966 | * Dump statistics for TINY_RCU, such as they are. | ||
967 | */ | ||
968 | static int show_tiny_stats(struct seq_file *m, void *unused) | ||
969 | { | ||
970 | show_tiny_preempt_stats(m); | ||
971 | seq_printf(m, "rcu_sched: qlen: %ld\n", rcu_sched_ctrlblk.qlen); | ||
972 | seq_printf(m, "rcu_bh: qlen: %ld\n", rcu_bh_ctrlblk.qlen); | ||
973 | return 0; | ||
974 | } | ||
975 | |||
976 | static int show_tiny_stats_open(struct inode *inode, struct file *file) | ||
977 | { | ||
978 | return single_open(file, show_tiny_stats, NULL); | ||
979 | } | ||
980 | |||
981 | static const struct file_operations show_tiny_stats_fops = { | ||
982 | .owner = THIS_MODULE, | ||
983 | .open = show_tiny_stats_open, | ||
984 | .read = seq_read, | ||
985 | .llseek = seq_lseek, | ||
986 | .release = single_release, | ||
987 | }; | ||
988 | |||
989 | static struct dentry *rcudir; | ||
990 | |||
991 | static int __init rcutiny_trace_init(void) | ||
992 | { | ||
993 | struct dentry *retval; | ||
994 | |||
995 | rcudir = debugfs_create_dir("rcu", NULL); | ||
996 | if (!rcudir) | ||
997 | goto free_out; | ||
998 | retval = debugfs_create_file("rcudata", 0444, rcudir, | ||
999 | NULL, &show_tiny_stats_fops); | ||
1000 | if (!retval) | ||
1001 | goto free_out; | ||
1002 | return 0; | ||
1003 | free_out: | ||
1004 | debugfs_remove_recursive(rcudir); | ||
1005 | return 1; | ||
1006 | } | ||
1007 | |||
1008 | static void __exit rcutiny_trace_cleanup(void) | ||
1009 | { | ||
1010 | debugfs_remove_recursive(rcudir); | ||
1011 | } | ||
1012 | |||
1013 | module_init(rcutiny_trace_init); | ||
1014 | module_exit(rcutiny_trace_cleanup); | ||
1015 | |||
1016 | MODULE_AUTHOR("Paul E. McKenney"); | ||
1017 | MODULE_DESCRIPTION("Read-Copy Update tracing for tiny implementation"); | ||
1018 | MODULE_LICENSE("GPL"); | ||
1019 | |||
1020 | #endif /* #ifdef CONFIG_RCU_TRACE */ | ||
diff --git a/kernel/rcutorture.c b/kernel/rcutorture.c index 9d8e8fb2515..89613f97ff2 100644 --- a/kernel/rcutorture.c +++ b/kernel/rcutorture.c | |||
@@ -47,6 +47,7 @@ | |||
47 | #include <linux/srcu.h> | 47 | #include <linux/srcu.h> |
48 | #include <linux/slab.h> | 48 | #include <linux/slab.h> |
49 | #include <asm/byteorder.h> | 49 | #include <asm/byteorder.h> |
50 | #include <linux/sched.h> | ||
50 | 51 | ||
51 | MODULE_LICENSE("GPL"); | 52 | MODULE_LICENSE("GPL"); |
52 | MODULE_AUTHOR("Paul E. McKenney <paulmck@us.ibm.com> and " | 53 | MODULE_AUTHOR("Paul E. McKenney <paulmck@us.ibm.com> and " |
@@ -64,6 +65,9 @@ static int irqreader = 1; /* RCU readers from irq (timers). */ | |||
64 | static int fqs_duration = 0; /* Duration of bursts (us), 0 to disable. */ | 65 | static int fqs_duration = 0; /* Duration of bursts (us), 0 to disable. */ |
65 | static int fqs_holdoff = 0; /* Hold time within burst (us). */ | 66 | static int fqs_holdoff = 0; /* Hold time within burst (us). */ |
66 | static int fqs_stutter = 3; /* Wait time between bursts (s). */ | 67 | static int fqs_stutter = 3; /* Wait time between bursts (s). */ |
68 | static int test_boost = 1; /* Test RCU prio boost: 0=no, 1=maybe, 2=yes. */ | ||
69 | static int test_boost_interval = 7; /* Interval between boost tests, seconds. */ | ||
70 | static int test_boost_duration = 4; /* Duration of each boost test, seconds. */ | ||
67 | static char *torture_type = "rcu"; /* What RCU implementation to torture. */ | 71 | static char *torture_type = "rcu"; /* What RCU implementation to torture. */ |
68 | 72 | ||
69 | module_param(nreaders, int, 0444); | 73 | module_param(nreaders, int, 0444); |
@@ -88,6 +92,12 @@ module_param(fqs_holdoff, int, 0444); | |||
88 | MODULE_PARM_DESC(fqs_holdoff, "Holdoff time within fqs bursts (us)"); | 92 | MODULE_PARM_DESC(fqs_holdoff, "Holdoff time within fqs bursts (us)"); |
89 | module_param(fqs_stutter, int, 0444); | 93 | module_param(fqs_stutter, int, 0444); |
90 | MODULE_PARM_DESC(fqs_stutter, "Wait time between fqs bursts (s)"); | 94 | MODULE_PARM_DESC(fqs_stutter, "Wait time between fqs bursts (s)"); |
95 | module_param(test_boost, int, 0444); | ||
96 | MODULE_PARM_DESC(test_boost, "Test RCU prio boost: 0=no, 1=maybe, 2=yes."); | ||
97 | module_param(test_boost_interval, int, 0444); | ||
98 | MODULE_PARM_DESC(test_boost_interval, "Interval between boost tests, seconds."); | ||
99 | module_param(test_boost_duration, int, 0444); | ||
100 | MODULE_PARM_DESC(test_boost_duration, "Duration of each boost test, seconds."); | ||
91 | module_param(torture_type, charp, 0444); | 101 | module_param(torture_type, charp, 0444); |
92 | MODULE_PARM_DESC(torture_type, "Type of RCU to torture (rcu, rcu_bh, srcu)"); | 102 | MODULE_PARM_DESC(torture_type, "Type of RCU to torture (rcu, rcu_bh, srcu)"); |
93 | 103 | ||
@@ -109,6 +119,7 @@ static struct task_struct *stats_task; | |||
109 | static struct task_struct *shuffler_task; | 119 | static struct task_struct *shuffler_task; |
110 | static struct task_struct *stutter_task; | 120 | static struct task_struct *stutter_task; |
111 | static struct task_struct *fqs_task; | 121 | static struct task_struct *fqs_task; |
122 | static struct task_struct *boost_tasks[NR_CPUS]; | ||
112 | 123 | ||
113 | #define RCU_TORTURE_PIPE_LEN 10 | 124 | #define RCU_TORTURE_PIPE_LEN 10 |
114 | 125 | ||
@@ -134,6 +145,12 @@ static atomic_t n_rcu_torture_alloc_fail; | |||
134 | static atomic_t n_rcu_torture_free; | 145 | static atomic_t n_rcu_torture_free; |
135 | static atomic_t n_rcu_torture_mberror; | 146 | static atomic_t n_rcu_torture_mberror; |
136 | static atomic_t n_rcu_torture_error; | 147 | static atomic_t n_rcu_torture_error; |
148 | static long n_rcu_torture_boost_ktrerror; | ||
149 | static long n_rcu_torture_boost_rterror; | ||
150 | static long n_rcu_torture_boost_allocerror; | ||
151 | static long n_rcu_torture_boost_afferror; | ||
152 | static long n_rcu_torture_boost_failure; | ||
153 | static long n_rcu_torture_boosts; | ||
137 | static long n_rcu_torture_timers; | 154 | static long n_rcu_torture_timers; |
138 | static struct list_head rcu_torture_removed; | 155 | static struct list_head rcu_torture_removed; |
139 | static cpumask_var_t shuffle_tmp_mask; | 156 | static cpumask_var_t shuffle_tmp_mask; |
@@ -147,6 +164,16 @@ static int stutter_pause_test; | |||
147 | #endif | 164 | #endif |
148 | int rcutorture_runnable = RCUTORTURE_RUNNABLE_INIT; | 165 | int rcutorture_runnable = RCUTORTURE_RUNNABLE_INIT; |
149 | 166 | ||
167 | #ifdef CONFIG_RCU_BOOST | ||
168 | #define rcu_can_boost() 1 | ||
169 | #else /* #ifdef CONFIG_RCU_BOOST */ | ||
170 | #define rcu_can_boost() 0 | ||
171 | #endif /* #else #ifdef CONFIG_RCU_BOOST */ | ||
172 | |||
173 | static unsigned long boost_starttime; /* jiffies of next boost test start. */ | ||
174 | DEFINE_MUTEX(boost_mutex); /* protect setting boost_starttime */ | ||
175 | /* and boost task create/destroy. */ | ||
176 | |||
150 | /* Mediate rmmod and system shutdown. Concurrent rmmod & shutdown illegal! */ | 177 | /* Mediate rmmod and system shutdown. Concurrent rmmod & shutdown illegal! */ |
151 | 178 | ||
152 | #define FULLSTOP_DONTSTOP 0 /* Normal operation. */ | 179 | #define FULLSTOP_DONTSTOP 0 /* Normal operation. */ |
@@ -277,6 +304,7 @@ struct rcu_torture_ops { | |||
277 | void (*fqs)(void); | 304 | void (*fqs)(void); |
278 | int (*stats)(char *page); | 305 | int (*stats)(char *page); |
279 | int irq_capable; | 306 | int irq_capable; |
307 | int can_boost; | ||
280 | char *name; | 308 | char *name; |
281 | }; | 309 | }; |
282 | 310 | ||
@@ -366,6 +394,7 @@ static struct rcu_torture_ops rcu_ops = { | |||
366 | .fqs = rcu_force_quiescent_state, | 394 | .fqs = rcu_force_quiescent_state, |
367 | .stats = NULL, | 395 | .stats = NULL, |
368 | .irq_capable = 1, | 396 | .irq_capable = 1, |
397 | .can_boost = rcu_can_boost(), | ||
369 | .name = "rcu" | 398 | .name = "rcu" |
370 | }; | 399 | }; |
371 | 400 | ||
@@ -408,6 +437,7 @@ static struct rcu_torture_ops rcu_sync_ops = { | |||
408 | .fqs = rcu_force_quiescent_state, | 437 | .fqs = rcu_force_quiescent_state, |
409 | .stats = NULL, | 438 | .stats = NULL, |
410 | .irq_capable = 1, | 439 | .irq_capable = 1, |
440 | .can_boost = rcu_can_boost(), | ||
411 | .name = "rcu_sync" | 441 | .name = "rcu_sync" |
412 | }; | 442 | }; |
413 | 443 | ||
@@ -424,6 +454,7 @@ static struct rcu_torture_ops rcu_expedited_ops = { | |||
424 | .fqs = rcu_force_quiescent_state, | 454 | .fqs = rcu_force_quiescent_state, |
425 | .stats = NULL, | 455 | .stats = NULL, |
426 | .irq_capable = 1, | 456 | .irq_capable = 1, |
457 | .can_boost = rcu_can_boost(), | ||
427 | .name = "rcu_expedited" | 458 | .name = "rcu_expedited" |
428 | }; | 459 | }; |
429 | 460 | ||
@@ -684,6 +715,110 @@ static struct rcu_torture_ops sched_expedited_ops = { | |||
684 | }; | 715 | }; |
685 | 716 | ||
686 | /* | 717 | /* |
718 | * RCU torture priority-boost testing. Runs one real-time thread per | ||
719 | * CPU for moderate bursts, repeatedly registering RCU callbacks and | ||
720 | * spinning waiting for them to be invoked. If a given callback takes | ||
721 | * too long to be invoked, we assume that priority inversion has occurred. | ||
722 | */ | ||
723 | |||
724 | struct rcu_boost_inflight { | ||
725 | struct rcu_head rcu; | ||
726 | int inflight; | ||
727 | }; | ||
728 | |||
729 | static void rcu_torture_boost_cb(struct rcu_head *head) | ||
730 | { | ||
731 | struct rcu_boost_inflight *rbip = | ||
732 | container_of(head, struct rcu_boost_inflight, rcu); | ||
733 | |||
734 | smp_mb(); /* Ensure RCU-core accesses precede clearing ->inflight */ | ||
735 | rbip->inflight = 0; | ||
736 | } | ||
737 | |||
738 | static int rcu_torture_boost(void *arg) | ||
739 | { | ||
740 | unsigned long call_rcu_time; | ||
741 | unsigned long endtime; | ||
742 | unsigned long oldstarttime; | ||
743 | struct rcu_boost_inflight rbi = { .inflight = 0 }; | ||
744 | struct sched_param sp; | ||
745 | |||
746 | VERBOSE_PRINTK_STRING("rcu_torture_boost started"); | ||
747 | |||
748 | /* Set real-time priority. */ | ||
749 | sp.sched_priority = 1; | ||
750 | if (sched_setscheduler(current, SCHED_FIFO, &sp) < 0) { | ||
751 | VERBOSE_PRINTK_STRING("rcu_torture_boost RT prio failed!"); | ||
752 | n_rcu_torture_boost_rterror++; | ||
753 | } | ||
754 | |||
755 | /* Each pass through the following loop does one boost-test cycle. */ | ||
756 | do { | ||
757 | /* Wait for the next test interval. */ | ||
758 | oldstarttime = boost_starttime; | ||
759 | while (jiffies - oldstarttime > ULONG_MAX / 2) { | ||
760 | schedule_timeout_uninterruptible(1); | ||
761 | rcu_stutter_wait("rcu_torture_boost"); | ||
762 | if (kthread_should_stop() || | ||
763 | fullstop != FULLSTOP_DONTSTOP) | ||
764 | goto checkwait; | ||
765 | } | ||
766 | |||
767 | /* Do one boost-test interval. */ | ||
768 | endtime = oldstarttime + test_boost_duration * HZ; | ||
769 | call_rcu_time = jiffies; | ||
770 | while (jiffies - endtime > ULONG_MAX / 2) { | ||
771 | /* If we don't have a callback in flight, post one. */ | ||
772 | if (!rbi.inflight) { | ||
773 | smp_mb(); /* RCU core before ->inflight = 1. */ | ||
774 | rbi.inflight = 1; | ||
775 | call_rcu(&rbi.rcu, rcu_torture_boost_cb); | ||
776 | if (jiffies - call_rcu_time > | ||
777 | test_boost_duration * HZ - HZ / 2) { | ||
778 | VERBOSE_PRINTK_STRING("rcu_torture_boost boosting failed"); | ||
779 | n_rcu_torture_boost_failure++; | ||
780 | } | ||
781 | call_rcu_time = jiffies; | ||
782 | } | ||
783 | cond_resched(); | ||
784 | rcu_stutter_wait("rcu_torture_boost"); | ||
785 | if (kthread_should_stop() || | ||
786 | fullstop != FULLSTOP_DONTSTOP) | ||
787 | goto checkwait; | ||
788 | } | ||
789 | |||
790 | /* | ||
791 | * Set the start time of the next test interval. | ||
792 | * Yes, this is vulnerable to long delays, but such | ||
793 | * delays simply cause a false negative for the next | ||
794 | * interval. Besides, we are running at RT priority, | ||
795 | * so delays should be relatively rare. | ||
796 | */ | ||
797 | while (oldstarttime == boost_starttime) { | ||
798 | if (mutex_trylock(&boost_mutex)) { | ||
799 | boost_starttime = jiffies + | ||
800 | test_boost_interval * HZ; | ||
801 | n_rcu_torture_boosts++; | ||
802 | mutex_unlock(&boost_mutex); | ||
803 | break; | ||
804 | } | ||
805 | schedule_timeout_uninterruptible(1); | ||
806 | } | ||
807 | |||
808 | /* Go do the stutter. */ | ||
809 | checkwait: rcu_stutter_wait("rcu_torture_boost"); | ||
810 | } while (!kthread_should_stop() && fullstop == FULLSTOP_DONTSTOP); | ||
811 | |||
812 | /* Clean up and exit. */ | ||
813 | VERBOSE_PRINTK_STRING("rcu_torture_boost task stopping"); | ||
814 | rcutorture_shutdown_absorb("rcu_torture_boost"); | ||
815 | while (!kthread_should_stop() || rbi.inflight) | ||
816 | schedule_timeout_uninterruptible(1); | ||
817 | smp_mb(); /* order accesses to ->inflight before stack-frame death. */ | ||
818 | return 0; | ||
819 | } | ||
820 | |||
821 | /* | ||
687 | * RCU torture force-quiescent-state kthread. Repeatedly induces | 822 | * RCU torture force-quiescent-state kthread. Repeatedly induces |
688 | * bursts of calls to force_quiescent_state(), increasing the probability | 823 | * bursts of calls to force_quiescent_state(), increasing the probability |
689 | * of occurrence of some important types of race conditions. | 824 | * of occurrence of some important types of race conditions. |
@@ -933,7 +1068,8 @@ rcu_torture_printk(char *page) | |||
933 | cnt += sprintf(&page[cnt], "%s%s ", torture_type, TORTURE_FLAG); | 1068 | cnt += sprintf(&page[cnt], "%s%s ", torture_type, TORTURE_FLAG); |
934 | cnt += sprintf(&page[cnt], | 1069 | cnt += sprintf(&page[cnt], |
935 | "rtc: %p ver: %ld tfle: %d rta: %d rtaf: %d rtf: %d " | 1070 | "rtc: %p ver: %ld tfle: %d rta: %d rtaf: %d rtf: %d " |
936 | "rtmbe: %d nt: %ld", | 1071 | "rtmbe: %d rtbke: %ld rtbre: %ld rtbae: %ld rtbafe: %ld " |
1072 | "rtbf: %ld rtb: %ld nt: %ld", | ||
937 | rcu_torture_current, | 1073 | rcu_torture_current, |
938 | rcu_torture_current_version, | 1074 | rcu_torture_current_version, |
939 | list_empty(&rcu_torture_freelist), | 1075 | list_empty(&rcu_torture_freelist), |
@@ -941,8 +1077,19 @@ rcu_torture_printk(char *page) | |||
941 | atomic_read(&n_rcu_torture_alloc_fail), | 1077 | atomic_read(&n_rcu_torture_alloc_fail), |
942 | atomic_read(&n_rcu_torture_free), | 1078 | atomic_read(&n_rcu_torture_free), |
943 | atomic_read(&n_rcu_torture_mberror), | 1079 | atomic_read(&n_rcu_torture_mberror), |
1080 | n_rcu_torture_boost_ktrerror, | ||
1081 | n_rcu_torture_boost_rterror, | ||
1082 | n_rcu_torture_boost_allocerror, | ||
1083 | n_rcu_torture_boost_afferror, | ||
1084 | n_rcu_torture_boost_failure, | ||
1085 | n_rcu_torture_boosts, | ||
944 | n_rcu_torture_timers); | 1086 | n_rcu_torture_timers); |
945 | if (atomic_read(&n_rcu_torture_mberror) != 0) | 1087 | if (atomic_read(&n_rcu_torture_mberror) != 0 || |
1088 | n_rcu_torture_boost_ktrerror != 0 || | ||
1089 | n_rcu_torture_boost_rterror != 0 || | ||
1090 | n_rcu_torture_boost_allocerror != 0 || | ||
1091 | n_rcu_torture_boost_afferror != 0 || | ||
1092 | n_rcu_torture_boost_failure != 0) | ||
946 | cnt += sprintf(&page[cnt], " !!!"); | 1093 | cnt += sprintf(&page[cnt], " !!!"); |
947 | cnt += sprintf(&page[cnt], "\n%s%s ", torture_type, TORTURE_FLAG); | 1094 | cnt += sprintf(&page[cnt], "\n%s%s ", torture_type, TORTURE_FLAG); |
948 | if (i > 1) { | 1095 | if (i > 1) { |
@@ -1094,22 +1241,91 @@ rcu_torture_stutter(void *arg) | |||
1094 | } | 1241 | } |
1095 | 1242 | ||
1096 | static inline void | 1243 | static inline void |
1097 | rcu_torture_print_module_parms(char *tag) | 1244 | rcu_torture_print_module_parms(struct rcu_torture_ops *cur_ops, char *tag) |
1098 | { | 1245 | { |
1099 | printk(KERN_ALERT "%s" TORTURE_FLAG | 1246 | printk(KERN_ALERT "%s" TORTURE_FLAG |
1100 | "--- %s: nreaders=%d nfakewriters=%d " | 1247 | "--- %s: nreaders=%d nfakewriters=%d " |
1101 | "stat_interval=%d verbose=%d test_no_idle_hz=%d " | 1248 | "stat_interval=%d verbose=%d test_no_idle_hz=%d " |
1102 | "shuffle_interval=%d stutter=%d irqreader=%d " | 1249 | "shuffle_interval=%d stutter=%d irqreader=%d " |
1103 | "fqs_duration=%d fqs_holdoff=%d fqs_stutter=%d\n", | 1250 | "fqs_duration=%d fqs_holdoff=%d fqs_stutter=%d " |
1251 | "test_boost=%d/%d test_boost_interval=%d " | ||
1252 | "test_boost_duration=%d\n", | ||
1104 | torture_type, tag, nrealreaders, nfakewriters, | 1253 | torture_type, tag, nrealreaders, nfakewriters, |
1105 | stat_interval, verbose, test_no_idle_hz, shuffle_interval, | 1254 | stat_interval, verbose, test_no_idle_hz, shuffle_interval, |
1106 | stutter, irqreader, fqs_duration, fqs_holdoff, fqs_stutter); | 1255 | stutter, irqreader, fqs_duration, fqs_holdoff, fqs_stutter, |
1256 | test_boost, cur_ops->can_boost, | ||
1257 | test_boost_interval, test_boost_duration); | ||
1107 | } | 1258 | } |
1108 | 1259 | ||
1109 | static struct notifier_block rcutorture_nb = { | 1260 | static struct notifier_block rcutorture_shutdown_nb = { |
1110 | .notifier_call = rcutorture_shutdown_notify, | 1261 | .notifier_call = rcutorture_shutdown_notify, |
1111 | }; | 1262 | }; |
1112 | 1263 | ||
1264 | static void rcutorture_booster_cleanup(int cpu) | ||
1265 | { | ||
1266 | struct task_struct *t; | ||
1267 | |||
1268 | if (boost_tasks[cpu] == NULL) | ||
1269 | return; | ||
1270 | mutex_lock(&boost_mutex); | ||
1271 | VERBOSE_PRINTK_STRING("Stopping rcu_torture_boost task"); | ||
1272 | t = boost_tasks[cpu]; | ||
1273 | boost_tasks[cpu] = NULL; | ||
1274 | mutex_unlock(&boost_mutex); | ||
1275 | |||
1276 | /* This must be outside of the mutex, otherwise deadlock! */ | ||
1277 | kthread_stop(t); | ||
1278 | } | ||
1279 | |||
1280 | static int rcutorture_booster_init(int cpu) | ||
1281 | { | ||
1282 | int retval; | ||
1283 | |||
1284 | if (boost_tasks[cpu] != NULL) | ||
1285 | return 0; /* Already created, nothing more to do. */ | ||
1286 | |||
1287 | /* Don't allow time recalculation while creating a new task. */ | ||
1288 | mutex_lock(&boost_mutex); | ||
1289 | VERBOSE_PRINTK_STRING("Creating rcu_torture_boost task"); | ||
1290 | boost_tasks[cpu] = kthread_create(rcu_torture_boost, NULL, | ||
1291 | "rcu_torture_boost"); | ||
1292 | if (IS_ERR(boost_tasks[cpu])) { | ||
1293 | retval = PTR_ERR(boost_tasks[cpu]); | ||
1294 | VERBOSE_PRINTK_STRING("rcu_torture_boost task create failed"); | ||
1295 | n_rcu_torture_boost_ktrerror++; | ||
1296 | boost_tasks[cpu] = NULL; | ||
1297 | mutex_unlock(&boost_mutex); | ||
1298 | return retval; | ||
1299 | } | ||
1300 | kthread_bind(boost_tasks[cpu], cpu); | ||
1301 | wake_up_process(boost_tasks[cpu]); | ||
1302 | mutex_unlock(&boost_mutex); | ||
1303 | return 0; | ||
1304 | } | ||
1305 | |||
1306 | static int rcutorture_cpu_notify(struct notifier_block *self, | ||
1307 | unsigned long action, void *hcpu) | ||
1308 | { | ||
1309 | long cpu = (long)hcpu; | ||
1310 | |||
1311 | switch (action) { | ||
1312 | case CPU_ONLINE: | ||
1313 | case CPU_DOWN_FAILED: | ||
1314 | (void)rcutorture_booster_init(cpu); | ||
1315 | break; | ||
1316 | case CPU_DOWN_PREPARE: | ||
1317 | rcutorture_booster_cleanup(cpu); | ||
1318 | break; | ||
1319 | default: | ||
1320 | break; | ||
1321 | } | ||
1322 | return NOTIFY_OK; | ||
1323 | } | ||
1324 | |||
1325 | static struct notifier_block rcutorture_cpu_nb = { | ||
1326 | .notifier_call = rcutorture_cpu_notify, | ||
1327 | }; | ||
1328 | |||
1113 | static void | 1329 | static void |
1114 | rcu_torture_cleanup(void) | 1330 | rcu_torture_cleanup(void) |
1115 | { | 1331 | { |
@@ -1127,7 +1343,7 @@ rcu_torture_cleanup(void) | |||
1127 | } | 1343 | } |
1128 | fullstop = FULLSTOP_RMMOD; | 1344 | fullstop = FULLSTOP_RMMOD; |
1129 | mutex_unlock(&fullstop_mutex); | 1345 | mutex_unlock(&fullstop_mutex); |
1130 | unregister_reboot_notifier(&rcutorture_nb); | 1346 | unregister_reboot_notifier(&rcutorture_shutdown_nb); |
1131 | if (stutter_task) { | 1347 | if (stutter_task) { |
1132 | VERBOSE_PRINTK_STRING("Stopping rcu_torture_stutter task"); | 1348 | VERBOSE_PRINTK_STRING("Stopping rcu_torture_stutter task"); |
1133 | kthread_stop(stutter_task); | 1349 | kthread_stop(stutter_task); |
@@ -1184,6 +1400,12 @@ rcu_torture_cleanup(void) | |||
1184 | kthread_stop(fqs_task); | 1400 | kthread_stop(fqs_task); |
1185 | } | 1401 | } |
1186 | fqs_task = NULL; | 1402 | fqs_task = NULL; |
1403 | if ((test_boost == 1 && cur_ops->can_boost) || | ||
1404 | test_boost == 2) { | ||
1405 | unregister_cpu_notifier(&rcutorture_cpu_nb); | ||
1406 | for_each_possible_cpu(i) | ||
1407 | rcutorture_booster_cleanup(i); | ||
1408 | } | ||
1187 | 1409 | ||
1188 | /* Wait for all RCU callbacks to fire. */ | 1410 | /* Wait for all RCU callbacks to fire. */ |
1189 | 1411 | ||
@@ -1195,9 +1417,9 @@ rcu_torture_cleanup(void) | |||
1195 | if (cur_ops->cleanup) | 1417 | if (cur_ops->cleanup) |
1196 | cur_ops->cleanup(); | 1418 | cur_ops->cleanup(); |
1197 | if (atomic_read(&n_rcu_torture_error)) | 1419 | if (atomic_read(&n_rcu_torture_error)) |
1198 | rcu_torture_print_module_parms("End of test: FAILURE"); | 1420 | rcu_torture_print_module_parms(cur_ops, "End of test: FAILURE"); |
1199 | else | 1421 | else |
1200 | rcu_torture_print_module_parms("End of test: SUCCESS"); | 1422 | rcu_torture_print_module_parms(cur_ops, "End of test: SUCCESS"); |
1201 | } | 1423 | } |
1202 | 1424 | ||
1203 | static int __init | 1425 | static int __init |
@@ -1242,7 +1464,7 @@ rcu_torture_init(void) | |||
1242 | nrealreaders = nreaders; | 1464 | nrealreaders = nreaders; |
1243 | else | 1465 | else |
1244 | nrealreaders = 2 * num_online_cpus(); | 1466 | nrealreaders = 2 * num_online_cpus(); |
1245 | rcu_torture_print_module_parms("Start of test"); | 1467 | rcu_torture_print_module_parms(cur_ops, "Start of test"); |
1246 | fullstop = FULLSTOP_DONTSTOP; | 1468 | fullstop = FULLSTOP_DONTSTOP; |
1247 | 1469 | ||
1248 | /* Set up the freelist. */ | 1470 | /* Set up the freelist. */ |
@@ -1263,6 +1485,12 @@ rcu_torture_init(void) | |||
1263 | atomic_set(&n_rcu_torture_free, 0); | 1485 | atomic_set(&n_rcu_torture_free, 0); |
1264 | atomic_set(&n_rcu_torture_mberror, 0); | 1486 | atomic_set(&n_rcu_torture_mberror, 0); |
1265 | atomic_set(&n_rcu_torture_error, 0); | 1487 | atomic_set(&n_rcu_torture_error, 0); |
1488 | n_rcu_torture_boost_ktrerror = 0; | ||
1489 | n_rcu_torture_boost_rterror = 0; | ||
1490 | n_rcu_torture_boost_allocerror = 0; | ||
1491 | n_rcu_torture_boost_afferror = 0; | ||
1492 | n_rcu_torture_boost_failure = 0; | ||
1493 | n_rcu_torture_boosts = 0; | ||
1266 | for (i = 0; i < RCU_TORTURE_PIPE_LEN + 1; i++) | 1494 | for (i = 0; i < RCU_TORTURE_PIPE_LEN + 1; i++) |
1267 | atomic_set(&rcu_torture_wcount[i], 0); | 1495 | atomic_set(&rcu_torture_wcount[i], 0); |
1268 | for_each_possible_cpu(cpu) { | 1496 | for_each_possible_cpu(cpu) { |
@@ -1376,7 +1604,27 @@ rcu_torture_init(void) | |||
1376 | goto unwind; | 1604 | goto unwind; |
1377 | } | 1605 | } |
1378 | } | 1606 | } |
1379 | register_reboot_notifier(&rcutorture_nb); | 1607 | if (test_boost_interval < 1) |
1608 | test_boost_interval = 1; | ||
1609 | if (test_boost_duration < 2) | ||
1610 | test_boost_duration = 2; | ||
1611 | if ((test_boost == 1 && cur_ops->can_boost) || | ||
1612 | test_boost == 2) { | ||
1613 | int retval; | ||
1614 | |||
1615 | boost_starttime = jiffies + test_boost_interval * HZ; | ||
1616 | register_cpu_notifier(&rcutorture_cpu_nb); | ||
1617 | for_each_possible_cpu(i) { | ||
1618 | if (cpu_is_offline(i)) | ||
1619 | continue; /* Heuristic: CPU can go offline. */ | ||
1620 | retval = rcutorture_booster_init(i); | ||
1621 | if (retval < 0) { | ||
1622 | firsterr = retval; | ||
1623 | goto unwind; | ||
1624 | } | ||
1625 | } | ||
1626 | } | ||
1627 | register_reboot_notifier(&rcutorture_shutdown_nb); | ||
1380 | mutex_unlock(&fullstop_mutex); | 1628 | mutex_unlock(&fullstop_mutex); |
1381 | return 0; | 1629 | return 0; |
1382 | 1630 | ||
diff --git a/kernel/rcutree.c b/kernel/rcutree.c index ccdc04c4798..d0ddfea6579 100644 --- a/kernel/rcutree.c +++ b/kernel/rcutree.c | |||
@@ -67,9 +67,6 @@ static struct lock_class_key rcu_node_class[NUM_RCU_LVLS]; | |||
67 | .gpnum = -300, \ | 67 | .gpnum = -300, \ |
68 | .completed = -300, \ | 68 | .completed = -300, \ |
69 | .onofflock = __RAW_SPIN_LOCK_UNLOCKED(&structname.onofflock), \ | 69 | .onofflock = __RAW_SPIN_LOCK_UNLOCKED(&structname.onofflock), \ |
70 | .orphan_cbs_list = NULL, \ | ||
71 | .orphan_cbs_tail = &structname.orphan_cbs_list, \ | ||
72 | .orphan_qlen = 0, \ | ||
73 | .fqslock = __RAW_SPIN_LOCK_UNLOCKED(&structname.fqslock), \ | 70 | .fqslock = __RAW_SPIN_LOCK_UNLOCKED(&structname.fqslock), \ |
74 | .n_force_qs = 0, \ | 71 | .n_force_qs = 0, \ |
75 | .n_force_qs_ngp = 0, \ | 72 | .n_force_qs_ngp = 0, \ |
@@ -620,9 +617,17 @@ static void __init check_cpu_stall_init(void) | |||
620 | static void __note_new_gpnum(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp) | 617 | static void __note_new_gpnum(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp) |
621 | { | 618 | { |
622 | if (rdp->gpnum != rnp->gpnum) { | 619 | if (rdp->gpnum != rnp->gpnum) { |
623 | rdp->qs_pending = 1; | 620 | /* |
624 | rdp->passed_quiesc = 0; | 621 | * If the current grace period is waiting for this CPU, |
622 | * set up to detect a quiescent state, otherwise don't | ||
623 | * go looking for one. | ||
624 | */ | ||
625 | rdp->gpnum = rnp->gpnum; | 625 | rdp->gpnum = rnp->gpnum; |
626 | if (rnp->qsmask & rdp->grpmask) { | ||
627 | rdp->qs_pending = 1; | ||
628 | rdp->passed_quiesc = 0; | ||
629 | } else | ||
630 | rdp->qs_pending = 0; | ||
626 | } | 631 | } |
627 | } | 632 | } |
628 | 633 | ||
@@ -681,6 +686,24 @@ __rcu_process_gp_end(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_dat | |||
681 | 686 | ||
682 | /* Remember that we saw this grace-period completion. */ | 687 | /* Remember that we saw this grace-period completion. */ |
683 | rdp->completed = rnp->completed; | 688 | rdp->completed = rnp->completed; |
689 | |||
690 | /* | ||
691 | * If we were in an extended quiescent state, we may have | ||
692 | * missed some grace periods that others CPUs handled on | ||
693 | * our behalf. Catch up with this state to avoid noting | ||
694 | * spurious new grace periods. If another grace period | ||
695 | * has started, then rnp->gpnum will have advanced, so | ||
696 | * we will detect this later on. | ||
697 | */ | ||
698 | if (ULONG_CMP_LT(rdp->gpnum, rdp->completed)) | ||
699 | rdp->gpnum = rdp->completed; | ||
700 | |||
701 | /* | ||
702 | * If RCU does not need a quiescent state from this CPU, | ||
703 | * then make sure that this CPU doesn't go looking for one. | ||
704 | */ | ||
705 | if ((rnp->qsmask & rdp->grpmask) == 0) | ||
706 | rdp->qs_pending = 0; | ||
684 | } | 707 | } |
685 | } | 708 | } |
686 | 709 | ||
@@ -984,53 +1007,31 @@ rcu_check_quiescent_state(struct rcu_state *rsp, struct rcu_data *rdp) | |||
984 | #ifdef CONFIG_HOTPLUG_CPU | 1007 | #ifdef CONFIG_HOTPLUG_CPU |
985 | 1008 | ||
986 | /* | 1009 | /* |
987 | * Move a dying CPU's RCU callbacks to the ->orphan_cbs_list for the | 1010 | * Move a dying CPU's RCU callbacks to online CPU's callback list. |
988 | * specified flavor of RCU. The callbacks will be adopted by the next | 1011 | * Synchronization is not required because this function executes |
989 | * _rcu_barrier() invocation or by the CPU_DEAD notifier, whichever | 1012 | * in stop_machine() context. |
990 | * comes first. Because this is invoked from the CPU_DYING notifier, | ||
991 | * irqs are already disabled. | ||
992 | */ | 1013 | */ |
993 | static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp) | 1014 | static void rcu_send_cbs_to_online(struct rcu_state *rsp) |
994 | { | 1015 | { |
995 | int i; | 1016 | int i; |
1017 | /* current DYING CPU is cleared in the cpu_online_mask */ | ||
1018 | int receive_cpu = cpumask_any(cpu_online_mask); | ||
996 | struct rcu_data *rdp = this_cpu_ptr(rsp->rda); | 1019 | struct rcu_data *rdp = this_cpu_ptr(rsp->rda); |
1020 | struct rcu_data *receive_rdp = per_cpu_ptr(rsp->rda, receive_cpu); | ||
997 | 1021 | ||
998 | if (rdp->nxtlist == NULL) | 1022 | if (rdp->nxtlist == NULL) |
999 | return; /* irqs disabled, so comparison is stable. */ | 1023 | return; /* irqs disabled, so comparison is stable. */ |
1000 | raw_spin_lock(&rsp->onofflock); /* irqs already disabled. */ | 1024 | |
1001 | *rsp->orphan_cbs_tail = rdp->nxtlist; | 1025 | *receive_rdp->nxttail[RCU_NEXT_TAIL] = rdp->nxtlist; |
1002 | rsp->orphan_cbs_tail = rdp->nxttail[RCU_NEXT_TAIL]; | 1026 | receive_rdp->nxttail[RCU_NEXT_TAIL] = rdp->nxttail[RCU_NEXT_TAIL]; |
1027 | receive_rdp->qlen += rdp->qlen; | ||
1028 | receive_rdp->n_cbs_adopted += rdp->qlen; | ||
1029 | rdp->n_cbs_orphaned += rdp->qlen; | ||
1030 | |||
1003 | rdp->nxtlist = NULL; | 1031 | rdp->nxtlist = NULL; |
1004 | for (i = 0; i < RCU_NEXT_SIZE; i++) | 1032 | for (i = 0; i < RCU_NEXT_SIZE; i++) |
1005 | rdp->nxttail[i] = &rdp->nxtlist; | 1033 | rdp->nxttail[i] = &rdp->nxtlist; |
1006 | rsp->orphan_qlen += rdp->qlen; | ||
1007 | rdp->n_cbs_orphaned += rdp->qlen; | ||
1008 | rdp->qlen = 0; | 1034 | rdp->qlen = 0; |
1009 | raw_spin_unlock(&rsp->onofflock); /* irqs remain disabled. */ | ||
1010 | } | ||
1011 | |||
1012 | /* | ||
1013 | * Adopt previously orphaned RCU callbacks. | ||
1014 | */ | ||
1015 | static void rcu_adopt_orphan_cbs(struct rcu_state *rsp) | ||
1016 | { | ||
1017 | unsigned long flags; | ||
1018 | struct rcu_data *rdp; | ||
1019 | |||
1020 | raw_spin_lock_irqsave(&rsp->onofflock, flags); | ||
1021 | rdp = this_cpu_ptr(rsp->rda); | ||
1022 | if (rsp->orphan_cbs_list == NULL) { | ||
1023 | raw_spin_unlock_irqrestore(&rsp->onofflock, flags); | ||
1024 | return; | ||
1025 | } | ||
1026 | *rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_list; | ||
1027 | rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_tail; | ||
1028 | rdp->qlen += rsp->orphan_qlen; | ||
1029 | rdp->n_cbs_adopted += rsp->orphan_qlen; | ||
1030 | rsp->orphan_cbs_list = NULL; | ||
1031 | rsp->orphan_cbs_tail = &rsp->orphan_cbs_list; | ||
1032 | rsp->orphan_qlen = 0; | ||
1033 | raw_spin_unlock_irqrestore(&rsp->onofflock, flags); | ||
1034 | } | 1035 | } |
1035 | 1036 | ||
1036 | /* | 1037 | /* |
@@ -1081,8 +1082,6 @@ static void __rcu_offline_cpu(int cpu, struct rcu_state *rsp) | |||
1081 | raw_spin_unlock_irqrestore(&rnp->lock, flags); | 1082 | raw_spin_unlock_irqrestore(&rnp->lock, flags); |
1082 | if (need_report & RCU_OFL_TASKS_EXP_GP) | 1083 | if (need_report & RCU_OFL_TASKS_EXP_GP) |
1083 | rcu_report_exp_rnp(rsp, rnp); | 1084 | rcu_report_exp_rnp(rsp, rnp); |
1084 | |||
1085 | rcu_adopt_orphan_cbs(rsp); | ||
1086 | } | 1085 | } |
1087 | 1086 | ||
1088 | /* | 1087 | /* |
@@ -1100,11 +1099,7 @@ static void rcu_offline_cpu(int cpu) | |||
1100 | 1099 | ||
1101 | #else /* #ifdef CONFIG_HOTPLUG_CPU */ | 1100 | #else /* #ifdef CONFIG_HOTPLUG_CPU */ |
1102 | 1101 | ||
1103 | static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp) | 1102 | static void rcu_send_cbs_to_online(struct rcu_state *rsp) |
1104 | { | ||
1105 | } | ||
1106 | |||
1107 | static void rcu_adopt_orphan_cbs(struct rcu_state *rsp) | ||
1108 | { | 1103 | { |
1109 | } | 1104 | } |
1110 | 1105 | ||
@@ -1440,22 +1435,11 @@ __call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu), | |||
1440 | */ | 1435 | */ |
1441 | local_irq_save(flags); | 1436 | local_irq_save(flags); |
1442 | rdp = this_cpu_ptr(rsp->rda); | 1437 | rdp = this_cpu_ptr(rsp->rda); |
1443 | rcu_process_gp_end(rsp, rdp); | ||
1444 | check_for_new_grace_period(rsp, rdp); | ||
1445 | 1438 | ||
1446 | /* Add the callback to our list. */ | 1439 | /* Add the callback to our list. */ |
1447 | *rdp->nxttail[RCU_NEXT_TAIL] = head; | 1440 | *rdp->nxttail[RCU_NEXT_TAIL] = head; |
1448 | rdp->nxttail[RCU_NEXT_TAIL] = &head->next; | 1441 | rdp->nxttail[RCU_NEXT_TAIL] = &head->next; |
1449 | 1442 | ||
1450 | /* Start a new grace period if one not already started. */ | ||
1451 | if (!rcu_gp_in_progress(rsp)) { | ||
1452 | unsigned long nestflag; | ||
1453 | struct rcu_node *rnp_root = rcu_get_root(rsp); | ||
1454 | |||
1455 | raw_spin_lock_irqsave(&rnp_root->lock, nestflag); | ||
1456 | rcu_start_gp(rsp, nestflag); /* releases rnp_root->lock. */ | ||
1457 | } | ||
1458 | |||
1459 | /* | 1443 | /* |
1460 | * Force the grace period if too many callbacks or too long waiting. | 1444 | * Force the grace period if too many callbacks or too long waiting. |
1461 | * Enforce hysteresis, and don't invoke force_quiescent_state() | 1445 | * Enforce hysteresis, and don't invoke force_quiescent_state() |
@@ -1464,12 +1448,27 @@ __call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu), | |||
1464 | * is the only one waiting for a grace period to complete. | 1448 | * is the only one waiting for a grace period to complete. |
1465 | */ | 1449 | */ |
1466 | if (unlikely(++rdp->qlen > rdp->qlen_last_fqs_check + qhimark)) { | 1450 | if (unlikely(++rdp->qlen > rdp->qlen_last_fqs_check + qhimark)) { |
1467 | rdp->blimit = LONG_MAX; | 1451 | |
1468 | if (rsp->n_force_qs == rdp->n_force_qs_snap && | 1452 | /* Are we ignoring a completed grace period? */ |
1469 | *rdp->nxttail[RCU_DONE_TAIL] != head) | 1453 | rcu_process_gp_end(rsp, rdp); |
1470 | force_quiescent_state(rsp, 0); | 1454 | check_for_new_grace_period(rsp, rdp); |
1471 | rdp->n_force_qs_snap = rsp->n_force_qs; | 1455 | |
1472 | rdp->qlen_last_fqs_check = rdp->qlen; | 1456 | /* Start a new grace period if one not already started. */ |
1457 | if (!rcu_gp_in_progress(rsp)) { | ||
1458 | unsigned long nestflag; | ||
1459 | struct rcu_node *rnp_root = rcu_get_root(rsp); | ||
1460 | |||
1461 | raw_spin_lock_irqsave(&rnp_root->lock, nestflag); | ||
1462 | rcu_start_gp(rsp, nestflag); /* rlses rnp_root->lock */ | ||
1463 | } else { | ||
1464 | /* Give the grace period a kick. */ | ||
1465 | rdp->blimit = LONG_MAX; | ||
1466 | if (rsp->n_force_qs == rdp->n_force_qs_snap && | ||
1467 | *rdp->nxttail[RCU_DONE_TAIL] != head) | ||
1468 | force_quiescent_state(rsp, 0); | ||
1469 | rdp->n_force_qs_snap = rsp->n_force_qs; | ||
1470 | rdp->qlen_last_fqs_check = rdp->qlen; | ||
1471 | } | ||
1473 | } else if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies)) | 1472 | } else if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies)) |
1474 | force_quiescent_state(rsp, 1); | 1473 | force_quiescent_state(rsp, 1); |
1475 | local_irq_restore(flags); | 1474 | local_irq_restore(flags); |
@@ -1699,13 +1698,12 @@ static void _rcu_barrier(struct rcu_state *rsp, | |||
1699 | * decrement rcu_barrier_cpu_count -- otherwise the first CPU | 1698 | * decrement rcu_barrier_cpu_count -- otherwise the first CPU |
1700 | * might complete its grace period before all of the other CPUs | 1699 | * might complete its grace period before all of the other CPUs |
1701 | * did their increment, causing this function to return too | 1700 | * did their increment, causing this function to return too |
1702 | * early. | 1701 | * early. Note that on_each_cpu() disables irqs, which prevents |
1702 | * any CPUs from coming online or going offline until each online | ||
1703 | * CPU has queued its RCU-barrier callback. | ||
1703 | */ | 1704 | */ |
1704 | atomic_set(&rcu_barrier_cpu_count, 1); | 1705 | atomic_set(&rcu_barrier_cpu_count, 1); |
1705 | preempt_disable(); /* stop CPU_DYING from filling orphan_cbs_list */ | ||
1706 | rcu_adopt_orphan_cbs(rsp); | ||
1707 | on_each_cpu(rcu_barrier_func, (void *)call_rcu_func, 1); | 1706 | on_each_cpu(rcu_barrier_func, (void *)call_rcu_func, 1); |
1708 | preempt_enable(); /* CPU_DYING can again fill orphan_cbs_list */ | ||
1709 | if (atomic_dec_and_test(&rcu_barrier_cpu_count)) | 1707 | if (atomic_dec_and_test(&rcu_barrier_cpu_count)) |
1710 | complete(&rcu_barrier_completion); | 1708 | complete(&rcu_barrier_completion); |
1711 | wait_for_completion(&rcu_barrier_completion); | 1709 | wait_for_completion(&rcu_barrier_completion); |
@@ -1831,18 +1829,13 @@ static int __cpuinit rcu_cpu_notify(struct notifier_block *self, | |||
1831 | case CPU_DYING: | 1829 | case CPU_DYING: |
1832 | case CPU_DYING_FROZEN: | 1830 | case CPU_DYING_FROZEN: |
1833 | /* | 1831 | /* |
1834 | * preempt_disable() in _rcu_barrier() prevents stop_machine(), | 1832 | * The whole machine is "stopped" except this CPU, so we can |
1835 | * so when "on_each_cpu(rcu_barrier_func, (void *)type, 1);" | 1833 | * touch any data without introducing corruption. We send the |
1836 | * returns, all online cpus have queued rcu_barrier_func(). | 1834 | * dying CPU's callbacks to an arbitrarily chosen online CPU. |
1837 | * The dying CPU clears its cpu_online_mask bit and | ||
1838 | * moves all of its RCU callbacks to ->orphan_cbs_list | ||
1839 | * in the context of stop_machine(), so subsequent calls | ||
1840 | * to _rcu_barrier() will adopt these callbacks and only | ||
1841 | * then queue rcu_barrier_func() on all remaining CPUs. | ||
1842 | */ | 1835 | */ |
1843 | rcu_send_cbs_to_orphanage(&rcu_bh_state); | 1836 | rcu_send_cbs_to_online(&rcu_bh_state); |
1844 | rcu_send_cbs_to_orphanage(&rcu_sched_state); | 1837 | rcu_send_cbs_to_online(&rcu_sched_state); |
1845 | rcu_preempt_send_cbs_to_orphanage(); | 1838 | rcu_preempt_send_cbs_to_online(); |
1846 | break; | 1839 | break; |
1847 | case CPU_DEAD: | 1840 | case CPU_DEAD: |
1848 | case CPU_DEAD_FROZEN: | 1841 | case CPU_DEAD_FROZEN: |
@@ -1880,8 +1873,9 @@ static void __init rcu_init_levelspread(struct rcu_state *rsp) | |||
1880 | { | 1873 | { |
1881 | int i; | 1874 | int i; |
1882 | 1875 | ||
1883 | for (i = NUM_RCU_LVLS - 1; i >= 0; i--) | 1876 | for (i = NUM_RCU_LVLS - 1; i > 0; i--) |
1884 | rsp->levelspread[i] = CONFIG_RCU_FANOUT; | 1877 | rsp->levelspread[i] = CONFIG_RCU_FANOUT; |
1878 | rsp->levelspread[0] = RCU_FANOUT_LEAF; | ||
1885 | } | 1879 | } |
1886 | #else /* #ifdef CONFIG_RCU_FANOUT_EXACT */ | 1880 | #else /* #ifdef CONFIG_RCU_FANOUT_EXACT */ |
1887 | static void __init rcu_init_levelspread(struct rcu_state *rsp) | 1881 | static void __init rcu_init_levelspread(struct rcu_state *rsp) |
diff --git a/kernel/rcutree.h b/kernel/rcutree.h index 91d4170c5c1..e8f057e44e3 100644 --- a/kernel/rcutree.h +++ b/kernel/rcutree.h | |||
@@ -31,46 +31,51 @@ | |||
31 | /* | 31 | /* |
32 | * Define shape of hierarchy based on NR_CPUS and CONFIG_RCU_FANOUT. | 32 | * Define shape of hierarchy based on NR_CPUS and CONFIG_RCU_FANOUT. |
33 | * In theory, it should be possible to add more levels straightforwardly. | 33 | * In theory, it should be possible to add more levels straightforwardly. |
34 | * In practice, this has not been tested, so there is probably some | 34 | * In practice, this did work well going from three levels to four. |
35 | * bug somewhere. | 35 | * Of course, your mileage may vary. |
36 | */ | 36 | */ |
37 | #define MAX_RCU_LVLS 4 | 37 | #define MAX_RCU_LVLS 4 |
38 | #define RCU_FANOUT (CONFIG_RCU_FANOUT) | 38 | #if CONFIG_RCU_FANOUT > 16 |
39 | #define RCU_FANOUT_SQ (RCU_FANOUT * RCU_FANOUT) | 39 | #define RCU_FANOUT_LEAF 16 |
40 | #define RCU_FANOUT_CUBE (RCU_FANOUT_SQ * RCU_FANOUT) | 40 | #else /* #if CONFIG_RCU_FANOUT > 16 */ |
41 | #define RCU_FANOUT_FOURTH (RCU_FANOUT_CUBE * RCU_FANOUT) | 41 | #define RCU_FANOUT_LEAF (CONFIG_RCU_FANOUT) |
42 | 42 | #endif /* #else #if CONFIG_RCU_FANOUT > 16 */ | |
43 | #if NR_CPUS <= RCU_FANOUT | 43 | #define RCU_FANOUT_1 (RCU_FANOUT_LEAF) |
44 | #define RCU_FANOUT_2 (RCU_FANOUT_1 * CONFIG_RCU_FANOUT) | ||
45 | #define RCU_FANOUT_3 (RCU_FANOUT_2 * CONFIG_RCU_FANOUT) | ||
46 | #define RCU_FANOUT_4 (RCU_FANOUT_3 * CONFIG_RCU_FANOUT) | ||
47 | |||
48 | #if NR_CPUS <= RCU_FANOUT_1 | ||
44 | # define NUM_RCU_LVLS 1 | 49 | # define NUM_RCU_LVLS 1 |
45 | # define NUM_RCU_LVL_0 1 | 50 | # define NUM_RCU_LVL_0 1 |
46 | # define NUM_RCU_LVL_1 (NR_CPUS) | 51 | # define NUM_RCU_LVL_1 (NR_CPUS) |
47 | # define NUM_RCU_LVL_2 0 | 52 | # define NUM_RCU_LVL_2 0 |
48 | # define NUM_RCU_LVL_3 0 | 53 | # define NUM_RCU_LVL_3 0 |
49 | # define NUM_RCU_LVL_4 0 | 54 | # define NUM_RCU_LVL_4 0 |
50 | #elif NR_CPUS <= RCU_FANOUT_SQ | 55 | #elif NR_CPUS <= RCU_FANOUT_2 |
51 | # define NUM_RCU_LVLS 2 | 56 | # define NUM_RCU_LVLS 2 |
52 | # define NUM_RCU_LVL_0 1 | 57 | # define NUM_RCU_LVL_0 1 |
53 | # define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT) | 58 | # define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1) |
54 | # define NUM_RCU_LVL_2 (NR_CPUS) | 59 | # define NUM_RCU_LVL_2 (NR_CPUS) |
55 | # define NUM_RCU_LVL_3 0 | 60 | # define NUM_RCU_LVL_3 0 |
56 | # define NUM_RCU_LVL_4 0 | 61 | # define NUM_RCU_LVL_4 0 |
57 | #elif NR_CPUS <= RCU_FANOUT_CUBE | 62 | #elif NR_CPUS <= RCU_FANOUT_3 |
58 | # define NUM_RCU_LVLS 3 | 63 | # define NUM_RCU_LVLS 3 |
59 | # define NUM_RCU_LVL_0 1 | 64 | # define NUM_RCU_LVL_0 1 |
60 | # define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_SQ) | 65 | # define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_2) |
61 | # define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT) | 66 | # define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1) |
62 | # define NUM_RCU_LVL_3 NR_CPUS | 67 | # define NUM_RCU_LVL_3 (NR_CPUS) |
63 | # define NUM_RCU_LVL_4 0 | 68 | # define NUM_RCU_LVL_4 0 |
64 | #elif NR_CPUS <= RCU_FANOUT_FOURTH | 69 | #elif NR_CPUS <= RCU_FANOUT_4 |
65 | # define NUM_RCU_LVLS 4 | 70 | # define NUM_RCU_LVLS 4 |
66 | # define NUM_RCU_LVL_0 1 | 71 | # define NUM_RCU_LVL_0 1 |
67 | # define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_CUBE) | 72 | # define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_3) |
68 | # define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_SQ) | 73 | # define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_2) |
69 | # define NUM_RCU_LVL_3 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT) | 74 | # define NUM_RCU_LVL_3 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1) |
70 | # define NUM_RCU_LVL_4 NR_CPUS | 75 | # define NUM_RCU_LVL_4 (NR_CPUS) |
71 | #else | 76 | #else |
72 | # error "CONFIG_RCU_FANOUT insufficient for NR_CPUS" | 77 | # error "CONFIG_RCU_FANOUT insufficient for NR_CPUS" |
73 | #endif /* #if (NR_CPUS) <= RCU_FANOUT */ | 78 | #endif /* #if (NR_CPUS) <= RCU_FANOUT_1 */ |
74 | 79 | ||
75 | #define RCU_SUM (NUM_RCU_LVL_0 + NUM_RCU_LVL_1 + NUM_RCU_LVL_2 + NUM_RCU_LVL_3 + NUM_RCU_LVL_4) | 80 | #define RCU_SUM (NUM_RCU_LVL_0 + NUM_RCU_LVL_1 + NUM_RCU_LVL_2 + NUM_RCU_LVL_3 + NUM_RCU_LVL_4) |
76 | #define NUM_RCU_NODES (RCU_SUM - NR_CPUS) | 81 | #define NUM_RCU_NODES (RCU_SUM - NR_CPUS) |
@@ -203,8 +208,8 @@ struct rcu_data { | |||
203 | long qlen_last_fqs_check; | 208 | long qlen_last_fqs_check; |
204 | /* qlen at last check for QS forcing */ | 209 | /* qlen at last check for QS forcing */ |
205 | unsigned long n_cbs_invoked; /* count of RCU cbs invoked. */ | 210 | unsigned long n_cbs_invoked; /* count of RCU cbs invoked. */ |
206 | unsigned long n_cbs_orphaned; /* RCU cbs sent to orphanage. */ | 211 | unsigned long n_cbs_orphaned; /* RCU cbs orphaned by dying CPU */ |
207 | unsigned long n_cbs_adopted; /* RCU cbs adopted from orphanage. */ | 212 | unsigned long n_cbs_adopted; /* RCU cbs adopted from dying CPU */ |
208 | unsigned long n_force_qs_snap; | 213 | unsigned long n_force_qs_snap; |
209 | /* did other CPU force QS recently? */ | 214 | /* did other CPU force QS recently? */ |
210 | long blimit; /* Upper limit on a processed batch */ | 215 | long blimit; /* Upper limit on a processed batch */ |
@@ -309,15 +314,7 @@ struct rcu_state { | |||
309 | /* End of fields guarded by root rcu_node's lock. */ | 314 | /* End of fields guarded by root rcu_node's lock. */ |
310 | 315 | ||
311 | raw_spinlock_t onofflock; /* exclude on/offline and */ | 316 | raw_spinlock_t onofflock; /* exclude on/offline and */ |
312 | /* starting new GP. Also */ | 317 | /* starting new GP. */ |
313 | /* protects the following */ | ||
314 | /* orphan_cbs fields. */ | ||
315 | struct rcu_head *orphan_cbs_list; /* list of rcu_head structs */ | ||
316 | /* orphaned by all CPUs in */ | ||
317 | /* a given leaf rcu_node */ | ||
318 | /* going offline. */ | ||
319 | struct rcu_head **orphan_cbs_tail; /* And tail pointer. */ | ||
320 | long orphan_qlen; /* Number of orphaned cbs. */ | ||
321 | raw_spinlock_t fqslock; /* Only one task forcing */ | 318 | raw_spinlock_t fqslock; /* Only one task forcing */ |
322 | /* quiescent states. */ | 319 | /* quiescent states. */ |
323 | unsigned long jiffies_force_qs; /* Time at which to invoke */ | 320 | unsigned long jiffies_force_qs; /* Time at which to invoke */ |
@@ -390,7 +387,7 @@ static void rcu_report_exp_rnp(struct rcu_state *rsp, struct rcu_node *rnp); | |||
390 | static int rcu_preempt_pending(int cpu); | 387 | static int rcu_preempt_pending(int cpu); |
391 | static int rcu_preempt_needs_cpu(int cpu); | 388 | static int rcu_preempt_needs_cpu(int cpu); |
392 | static void __cpuinit rcu_preempt_init_percpu_data(int cpu); | 389 | static void __cpuinit rcu_preempt_init_percpu_data(int cpu); |
393 | static void rcu_preempt_send_cbs_to_orphanage(void); | 390 | static void rcu_preempt_send_cbs_to_online(void); |
394 | static void __init __rcu_init_preempt(void); | 391 | static void __init __rcu_init_preempt(void); |
395 | static void rcu_needs_cpu_flush(void); | 392 | static void rcu_needs_cpu_flush(void); |
396 | 393 | ||
diff --git a/kernel/rcutree_plugin.h b/kernel/rcutree_plugin.h index 71a4147473f..a3638710dc6 100644 --- a/kernel/rcutree_plugin.h +++ b/kernel/rcutree_plugin.h | |||
@@ -25,6 +25,7 @@ | |||
25 | */ | 25 | */ |
26 | 26 | ||
27 | #include <linux/delay.h> | 27 | #include <linux/delay.h> |
28 | #include <linux/stop_machine.h> | ||
28 | 29 | ||
29 | /* | 30 | /* |
30 | * Check the RCU kernel configuration parameters and print informative | 31 | * Check the RCU kernel configuration parameters and print informative |
@@ -773,11 +774,11 @@ static void __cpuinit rcu_preempt_init_percpu_data(int cpu) | |||
773 | } | 774 | } |
774 | 775 | ||
775 | /* | 776 | /* |
776 | * Move preemptable RCU's callbacks to ->orphan_cbs_list. | 777 | * Move preemptable RCU's callbacks from dying CPU to other online CPU. |
777 | */ | 778 | */ |
778 | static void rcu_preempt_send_cbs_to_orphanage(void) | 779 | static void rcu_preempt_send_cbs_to_online(void) |
779 | { | 780 | { |
780 | rcu_send_cbs_to_orphanage(&rcu_preempt_state); | 781 | rcu_send_cbs_to_online(&rcu_preempt_state); |
781 | } | 782 | } |
782 | 783 | ||
783 | /* | 784 | /* |
@@ -1001,7 +1002,7 @@ static void __cpuinit rcu_preempt_init_percpu_data(int cpu) | |||
1001 | /* | 1002 | /* |
1002 | * Because there is no preemptable RCU, there are no callbacks to move. | 1003 | * Because there is no preemptable RCU, there are no callbacks to move. |
1003 | */ | 1004 | */ |
1004 | static void rcu_preempt_send_cbs_to_orphanage(void) | 1005 | static void rcu_preempt_send_cbs_to_online(void) |
1005 | { | 1006 | { |
1006 | } | 1007 | } |
1007 | 1008 | ||
@@ -1014,6 +1015,132 @@ static void __init __rcu_init_preempt(void) | |||
1014 | 1015 | ||
1015 | #endif /* #else #ifdef CONFIG_TREE_PREEMPT_RCU */ | 1016 | #endif /* #else #ifdef CONFIG_TREE_PREEMPT_RCU */ |
1016 | 1017 | ||
1018 | #ifndef CONFIG_SMP | ||
1019 | |||
1020 | void synchronize_sched_expedited(void) | ||
1021 | { | ||
1022 | cond_resched(); | ||
1023 | } | ||
1024 | EXPORT_SYMBOL_GPL(synchronize_sched_expedited); | ||
1025 | |||
1026 | #else /* #ifndef CONFIG_SMP */ | ||
1027 | |||
1028 | static atomic_t sync_sched_expedited_started = ATOMIC_INIT(0); | ||
1029 | static atomic_t sync_sched_expedited_done = ATOMIC_INIT(0); | ||
1030 | |||
1031 | static int synchronize_sched_expedited_cpu_stop(void *data) | ||
1032 | { | ||
1033 | /* | ||
1034 | * There must be a full memory barrier on each affected CPU | ||
1035 | * between the time that try_stop_cpus() is called and the | ||
1036 | * time that it returns. | ||
1037 | * | ||
1038 | * In the current initial implementation of cpu_stop, the | ||
1039 | * above condition is already met when the control reaches | ||
1040 | * this point and the following smp_mb() is not strictly | ||
1041 | * necessary. Do smp_mb() anyway for documentation and | ||
1042 | * robustness against future implementation changes. | ||
1043 | */ | ||
1044 | smp_mb(); /* See above comment block. */ | ||
1045 | return 0; | ||
1046 | } | ||
1047 | |||
1048 | /* | ||
1049 | * Wait for an rcu-sched grace period to elapse, but use "big hammer" | ||
1050 | * approach to force grace period to end quickly. This consumes | ||
1051 | * significant time on all CPUs, and is thus not recommended for | ||
1052 | * any sort of common-case code. | ||
1053 | * | ||
1054 | * Note that it is illegal to call this function while holding any | ||
1055 | * lock that is acquired by a CPU-hotplug notifier. Failing to | ||
1056 | * observe this restriction will result in deadlock. | ||
1057 | * | ||
1058 | * This implementation can be thought of as an application of ticket | ||
1059 | * locking to RCU, with sync_sched_expedited_started and | ||
1060 | * sync_sched_expedited_done taking on the roles of the halves | ||
1061 | * of the ticket-lock word. Each task atomically increments | ||
1062 | * sync_sched_expedited_started upon entry, snapshotting the old value, | ||
1063 | * then attempts to stop all the CPUs. If this succeeds, then each | ||
1064 | * CPU will have executed a context switch, resulting in an RCU-sched | ||
1065 | * grace period. We are then done, so we use atomic_cmpxchg() to | ||
1066 | * update sync_sched_expedited_done to match our snapshot -- but | ||
1067 | * only if someone else has not already advanced past our snapshot. | ||
1068 | * | ||
1069 | * On the other hand, if try_stop_cpus() fails, we check the value | ||
1070 | * of sync_sched_expedited_done. If it has advanced past our | ||
1071 | * initial snapshot, then someone else must have forced a grace period | ||
1072 | * some time after we took our snapshot. In this case, our work is | ||
1073 | * done for us, and we can simply return. Otherwise, we try again, | ||
1074 | * but keep our initial snapshot for purposes of checking for someone | ||
1075 | * doing our work for us. | ||
1076 | * | ||
1077 | * If we fail too many times in a row, we fall back to synchronize_sched(). | ||
1078 | */ | ||
1079 | void synchronize_sched_expedited(void) | ||
1080 | { | ||
1081 | int firstsnap, s, snap, trycount = 0; | ||
1082 | |||
1083 | /* Note that atomic_inc_return() implies full memory barrier. */ | ||
1084 | firstsnap = snap = atomic_inc_return(&sync_sched_expedited_started); | ||
1085 | get_online_cpus(); | ||
1086 | |||
1087 | /* | ||
1088 | * Each pass through the following loop attempts to force a | ||
1089 | * context switch on each CPU. | ||
1090 | */ | ||
1091 | while (try_stop_cpus(cpu_online_mask, | ||
1092 | synchronize_sched_expedited_cpu_stop, | ||
1093 | NULL) == -EAGAIN) { | ||
1094 | put_online_cpus(); | ||
1095 | |||
1096 | /* No joy, try again later. Or just synchronize_sched(). */ | ||
1097 | if (trycount++ < 10) | ||
1098 | udelay(trycount * num_online_cpus()); | ||
1099 | else { | ||
1100 | synchronize_sched(); | ||
1101 | return; | ||
1102 | } | ||
1103 | |||
1104 | /* Check to see if someone else did our work for us. */ | ||
1105 | s = atomic_read(&sync_sched_expedited_done); | ||
1106 | if (UINT_CMP_GE((unsigned)s, (unsigned)firstsnap)) { | ||
1107 | smp_mb(); /* ensure test happens before caller kfree */ | ||
1108 | return; | ||
1109 | } | ||
1110 | |||
1111 | /* | ||
1112 | * Refetching sync_sched_expedited_started allows later | ||
1113 | * callers to piggyback on our grace period. We subtract | ||
1114 | * 1 to get the same token that the last incrementer got. | ||
1115 | * We retry after they started, so our grace period works | ||
1116 | * for them, and they started after our first try, so their | ||
1117 | * grace period works for us. | ||
1118 | */ | ||
1119 | get_online_cpus(); | ||
1120 | snap = atomic_read(&sync_sched_expedited_started) - 1; | ||
1121 | smp_mb(); /* ensure read is before try_stop_cpus(). */ | ||
1122 | } | ||
1123 | |||
1124 | /* | ||
1125 | * Everyone up to our most recent fetch is covered by our grace | ||
1126 | * period. Update the counter, but only if our work is still | ||
1127 | * relevant -- which it won't be if someone who started later | ||
1128 | * than we did beat us to the punch. | ||
1129 | */ | ||
1130 | do { | ||
1131 | s = atomic_read(&sync_sched_expedited_done); | ||
1132 | if (UINT_CMP_GE((unsigned)s, (unsigned)snap)) { | ||
1133 | smp_mb(); /* ensure test happens before caller kfree */ | ||
1134 | break; | ||
1135 | } | ||
1136 | } while (atomic_cmpxchg(&sync_sched_expedited_done, s, snap) != s); | ||
1137 | |||
1138 | put_online_cpus(); | ||
1139 | } | ||
1140 | EXPORT_SYMBOL_GPL(synchronize_sched_expedited); | ||
1141 | |||
1142 | #endif /* #else #ifndef CONFIG_SMP */ | ||
1143 | |||
1017 | #if !defined(CONFIG_RCU_FAST_NO_HZ) | 1144 | #if !defined(CONFIG_RCU_FAST_NO_HZ) |
1018 | 1145 | ||
1019 | /* | 1146 | /* |
diff --git a/kernel/rcutree_trace.c b/kernel/rcutree_trace.c index d15430b9d12..c8e97853b97 100644 --- a/kernel/rcutree_trace.c +++ b/kernel/rcutree_trace.c | |||
@@ -166,13 +166,13 @@ static void print_one_rcu_state(struct seq_file *m, struct rcu_state *rsp) | |||
166 | 166 | ||
167 | gpnum = rsp->gpnum; | 167 | gpnum = rsp->gpnum; |
168 | seq_printf(m, "c=%lu g=%lu s=%d jfq=%ld j=%x " | 168 | seq_printf(m, "c=%lu g=%lu s=%d jfq=%ld j=%x " |
169 | "nfqs=%lu/nfqsng=%lu(%lu) fqlh=%lu oqlen=%ld\n", | 169 | "nfqs=%lu/nfqsng=%lu(%lu) fqlh=%lu\n", |
170 | rsp->completed, gpnum, rsp->signaled, | 170 | rsp->completed, gpnum, rsp->signaled, |
171 | (long)(rsp->jiffies_force_qs - jiffies), | 171 | (long)(rsp->jiffies_force_qs - jiffies), |
172 | (int)(jiffies & 0xffff), | 172 | (int)(jiffies & 0xffff), |
173 | rsp->n_force_qs, rsp->n_force_qs_ngp, | 173 | rsp->n_force_qs, rsp->n_force_qs_ngp, |
174 | rsp->n_force_qs - rsp->n_force_qs_ngp, | 174 | rsp->n_force_qs - rsp->n_force_qs_ngp, |
175 | rsp->n_force_qs_lh, rsp->orphan_qlen); | 175 | rsp->n_force_qs_lh); |
176 | for (rnp = &rsp->node[0]; rnp - &rsp->node[0] < NUM_RCU_NODES; rnp++) { | 176 | for (rnp = &rsp->node[0]; rnp - &rsp->node[0] < NUM_RCU_NODES; rnp++) { |
177 | if (rnp->level != level) { | 177 | if (rnp->level != level) { |
178 | seq_puts(m, "\n"); | 178 | seq_puts(m, "\n"); |
@@ -300,7 +300,7 @@ static const struct file_operations rcu_pending_fops = { | |||
300 | 300 | ||
301 | static struct dentry *rcudir; | 301 | static struct dentry *rcudir; |
302 | 302 | ||
303 | static int __init rcuclassic_trace_init(void) | 303 | static int __init rcutree_trace_init(void) |
304 | { | 304 | { |
305 | struct dentry *retval; | 305 | struct dentry *retval; |
306 | 306 | ||
@@ -337,14 +337,14 @@ free_out: | |||
337 | return 1; | 337 | return 1; |
338 | } | 338 | } |
339 | 339 | ||
340 | static void __exit rcuclassic_trace_cleanup(void) | 340 | static void __exit rcutree_trace_cleanup(void) |
341 | { | 341 | { |
342 | debugfs_remove_recursive(rcudir); | 342 | debugfs_remove_recursive(rcudir); |
343 | } | 343 | } |
344 | 344 | ||
345 | 345 | ||
346 | module_init(rcuclassic_trace_init); | 346 | module_init(rcutree_trace_init); |
347 | module_exit(rcuclassic_trace_cleanup); | 347 | module_exit(rcutree_trace_cleanup); |
348 | 348 | ||
349 | MODULE_AUTHOR("Paul E. McKenney"); | 349 | MODULE_AUTHOR("Paul E. McKenney"); |
350 | MODULE_DESCRIPTION("Read-Copy Update tracing for hierarchical implementation"); | 350 | MODULE_DESCRIPTION("Read-Copy Update tracing for hierarchical implementation"); |
diff --git a/kernel/sched.c b/kernel/sched.c index 114a0deb2b0..04949089e76 100644 --- a/kernel/sched.c +++ b/kernel/sched.c | |||
@@ -8067,8 +8067,6 @@ void __init sched_init(void) | |||
8067 | zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT); | 8067 | zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT); |
8068 | #endif /* SMP */ | 8068 | #endif /* SMP */ |
8069 | 8069 | ||
8070 | perf_event_init(); | ||
8071 | |||
8072 | scheduler_running = 1; | 8070 | scheduler_running = 1; |
8073 | } | 8071 | } |
8074 | 8072 | ||
@@ -9241,72 +9239,3 @@ struct cgroup_subsys cpuacct_subsys = { | |||
9241 | }; | 9239 | }; |
9242 | #endif /* CONFIG_CGROUP_CPUACCT */ | 9240 | #endif /* CONFIG_CGROUP_CPUACCT */ |
9243 | 9241 | ||
9244 | #ifndef CONFIG_SMP | ||
9245 | |||
9246 | void synchronize_sched_expedited(void) | ||
9247 | { | ||
9248 | barrier(); | ||
9249 | } | ||
9250 | EXPORT_SYMBOL_GPL(synchronize_sched_expedited); | ||
9251 | |||
9252 | #else /* #ifndef CONFIG_SMP */ | ||
9253 | |||
9254 | static atomic_t synchronize_sched_expedited_count = ATOMIC_INIT(0); | ||
9255 | |||
9256 | static int synchronize_sched_expedited_cpu_stop(void *data) | ||
9257 | { | ||
9258 | /* | ||
9259 | * There must be a full memory barrier on each affected CPU | ||
9260 | * between the time that try_stop_cpus() is called and the | ||
9261 | * time that it returns. | ||
9262 | * | ||
9263 | * In the current initial implementation of cpu_stop, the | ||
9264 | * above condition is already met when the control reaches | ||
9265 | * this point and the following smp_mb() is not strictly | ||
9266 | * necessary. Do smp_mb() anyway for documentation and | ||
9267 | * robustness against future implementation changes. | ||
9268 | */ | ||
9269 | smp_mb(); /* See above comment block. */ | ||
9270 | return 0; | ||
9271 | } | ||
9272 | |||
9273 | /* | ||
9274 | * Wait for an rcu-sched grace period to elapse, but use "big hammer" | ||
9275 | * approach to force grace period to end quickly. This consumes | ||
9276 | * significant time on all CPUs, and is thus not recommended for | ||
9277 | * any sort of common-case code. | ||
9278 | * | ||
9279 | * Note that it is illegal to call this function while holding any | ||
9280 | * lock that is acquired by a CPU-hotplug notifier. Failing to | ||
9281 | * observe this restriction will result in deadlock. | ||
9282 | */ | ||
9283 | void synchronize_sched_expedited(void) | ||
9284 | { | ||
9285 | int snap, trycount = 0; | ||
9286 | |||
9287 | smp_mb(); /* ensure prior mod happens before capturing snap. */ | ||
9288 | snap = atomic_read(&synchronize_sched_expedited_count) + 1; | ||
9289 | get_online_cpus(); | ||
9290 | while (try_stop_cpus(cpu_online_mask, | ||
9291 | synchronize_sched_expedited_cpu_stop, | ||
9292 | NULL) == -EAGAIN) { | ||
9293 | put_online_cpus(); | ||
9294 | if (trycount++ < 10) | ||
9295 | udelay(trycount * num_online_cpus()); | ||
9296 | else { | ||
9297 | synchronize_sched(); | ||
9298 | return; | ||
9299 | } | ||
9300 | if (atomic_read(&synchronize_sched_expedited_count) - snap > 0) { | ||
9301 | smp_mb(); /* ensure test happens before caller kfree */ | ||
9302 | return; | ||
9303 | } | ||
9304 | get_online_cpus(); | ||
9305 | } | ||
9306 | atomic_inc(&synchronize_sched_expedited_count); | ||
9307 | smp_mb__after_atomic_inc(); /* ensure post-GP actions seen after GP. */ | ||
9308 | put_online_cpus(); | ||
9309 | } | ||
9310 | EXPORT_SYMBOL_GPL(synchronize_sched_expedited); | ||
9311 | |||
9312 | #endif /* #else #ifndef CONFIG_SMP */ | ||
diff --git a/kernel/srcu.c b/kernel/srcu.c index c71e0750053..98d8c1e80ed 100644 --- a/kernel/srcu.c +++ b/kernel/srcu.c | |||
@@ -31,6 +31,7 @@ | |||
31 | #include <linux/rcupdate.h> | 31 | #include <linux/rcupdate.h> |
32 | #include <linux/sched.h> | 32 | #include <linux/sched.h> |
33 | #include <linux/smp.h> | 33 | #include <linux/smp.h> |
34 | #include <linux/delay.h> | ||
34 | #include <linux/srcu.h> | 35 | #include <linux/srcu.h> |
35 | 36 | ||
36 | static int init_srcu_struct_fields(struct srcu_struct *sp) | 37 | static int init_srcu_struct_fields(struct srcu_struct *sp) |
@@ -203,9 +204,14 @@ static void __synchronize_srcu(struct srcu_struct *sp, void (*sync_func)(void)) | |||
203 | * all srcu_read_lock() calls using the old counters have completed. | 204 | * all srcu_read_lock() calls using the old counters have completed. |
204 | * Their corresponding critical sections might well be still | 205 | * Their corresponding critical sections might well be still |
205 | * executing, but the srcu_read_lock() primitives themselves | 206 | * executing, but the srcu_read_lock() primitives themselves |
206 | * will have finished executing. | 207 | * will have finished executing. We initially give readers |
208 | * an arbitrarily chosen 10 microseconds to get out of their | ||
209 | * SRCU read-side critical sections, then loop waiting 1/HZ | ||
210 | * seconds per iteration. | ||
207 | */ | 211 | */ |
208 | 212 | ||
213 | if (srcu_readers_active_idx(sp, idx)) | ||
214 | udelay(CONFIG_SRCU_SYNCHRONIZE_DELAY); | ||
209 | while (srcu_readers_active_idx(sp, idx)) | 215 | while (srcu_readers_active_idx(sp, idx)) |
210 | schedule_timeout_interruptible(1); | 216 | schedule_timeout_interruptible(1); |
211 | 217 | ||
diff --git a/kernel/sysctl.c b/kernel/sysctl.c index 121e4fff03d..ae5cbb1e3ce 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c | |||
@@ -744,21 +744,21 @@ static struct ctl_table kern_table[] = { | |||
744 | .extra1 = &zero, | 744 | .extra1 = &zero, |
745 | .extra2 = &one, | 745 | .extra2 = &one, |
746 | }, | 746 | }, |
747 | #endif | ||
748 | #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86) && !defined(CONFIG_LOCKUP_DETECTOR) | ||
749 | { | 747 | { |
750 | .procname = "unknown_nmi_panic", | 748 | .procname = "nmi_watchdog", |
751 | .data = &unknown_nmi_panic, | 749 | .data = &watchdog_enabled, |
752 | .maxlen = sizeof (int), | 750 | .maxlen = sizeof (int), |
753 | .mode = 0644, | 751 | .mode = 0644, |
754 | .proc_handler = proc_dointvec, | 752 | .proc_handler = proc_dowatchdog_enabled, |
755 | }, | 753 | }, |
754 | #endif | ||
755 | #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86) | ||
756 | { | 756 | { |
757 | .procname = "nmi_watchdog", | 757 | .procname = "unknown_nmi_panic", |
758 | .data = &nmi_watchdog_enabled, | 758 | .data = &unknown_nmi_panic, |
759 | .maxlen = sizeof (int), | 759 | .maxlen = sizeof (int), |
760 | .mode = 0644, | 760 | .mode = 0644, |
761 | .proc_handler = proc_nmi_enabled, | 761 | .proc_handler = proc_dointvec, |
762 | }, | 762 | }, |
763 | #endif | 763 | #endif |
764 | #if defined(CONFIG_X86) | 764 | #if defined(CONFIG_X86) |
diff --git a/kernel/sysctl_binary.c b/kernel/sysctl_binary.c index 1357c578606..4b2545a136f 100644 --- a/kernel/sysctl_binary.c +++ b/kernel/sysctl_binary.c | |||
@@ -136,7 +136,6 @@ static const struct bin_table bin_kern_table[] = { | |||
136 | { CTL_INT, KERN_IA64_UNALIGNED, "ignore-unaligned-usertrap" }, | 136 | { CTL_INT, KERN_IA64_UNALIGNED, "ignore-unaligned-usertrap" }, |
137 | { CTL_INT, KERN_COMPAT_LOG, "compat-log" }, | 137 | { CTL_INT, KERN_COMPAT_LOG, "compat-log" }, |
138 | { CTL_INT, KERN_MAX_LOCK_DEPTH, "max_lock_depth" }, | 138 | { CTL_INT, KERN_MAX_LOCK_DEPTH, "max_lock_depth" }, |
139 | { CTL_INT, KERN_NMI_WATCHDOG, "nmi_watchdog" }, | ||
140 | { CTL_INT, KERN_PANIC_ON_NMI, "panic_on_unrecovered_nmi" }, | 139 | { CTL_INT, KERN_PANIC_ON_NMI, "panic_on_unrecovered_nmi" }, |
141 | {} | 140 | {} |
142 | }; | 141 | }; |
diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig index ea37e2ff416..14674dce77a 100644 --- a/kernel/trace/Kconfig +++ b/kernel/trace/Kconfig | |||
@@ -69,6 +69,21 @@ config EVENT_TRACING | |||
69 | select CONTEXT_SWITCH_TRACER | 69 | select CONTEXT_SWITCH_TRACER |
70 | bool | 70 | bool |
71 | 71 | ||
72 | config EVENT_POWER_TRACING_DEPRECATED | ||
73 | depends on EVENT_TRACING | ||
74 | bool "Deprecated power event trace API, to be removed" | ||
75 | default y | ||
76 | help | ||
77 | Provides old power event types: | ||
78 | C-state/idle accounting events: | ||
79 | power:power_start | ||
80 | power:power_end | ||
81 | and old cpufreq accounting event: | ||
82 | power:power_frequency | ||
83 | This is for userspace compatibility | ||
84 | and will vanish after 5 kernel iterations, | ||
85 | namely 2.6.41. | ||
86 | |||
72 | config CONTEXT_SWITCH_TRACER | 87 | config CONTEXT_SWITCH_TRACER |
73 | bool | 88 | bool |
74 | 89 | ||
diff --git a/kernel/trace/power-traces.c b/kernel/trace/power-traces.c index a22582a0616..f55fcf61b22 100644 --- a/kernel/trace/power-traces.c +++ b/kernel/trace/power-traces.c | |||
@@ -13,5 +13,8 @@ | |||
13 | #define CREATE_TRACE_POINTS | 13 | #define CREATE_TRACE_POINTS |
14 | #include <trace/events/power.h> | 14 | #include <trace/events/power.h> |
15 | 15 | ||
16 | EXPORT_TRACEPOINT_SYMBOL_GPL(power_frequency); | 16 | #ifdef EVENT_POWER_TRACING_DEPRECATED |
17 | EXPORT_TRACEPOINT_SYMBOL_GPL(power_start); | ||
18 | #endif | ||
19 | EXPORT_TRACEPOINT_SYMBOL_GPL(cpu_idle); | ||
17 | 20 | ||
diff --git a/kernel/trace/trace_event_perf.c b/kernel/trace/trace_event_perf.c index 39c059ca670..19a359d5e6d 100644 --- a/kernel/trace/trace_event_perf.c +++ b/kernel/trace/trace_event_perf.c | |||
@@ -21,17 +21,46 @@ typedef typeof(unsigned long [PERF_MAX_TRACE_SIZE / sizeof(unsigned long)]) | |||
21 | /* Count the events in use (per event id, not per instance) */ | 21 | /* Count the events in use (per event id, not per instance) */ |
22 | static int total_ref_count; | 22 | static int total_ref_count; |
23 | 23 | ||
24 | static int perf_trace_event_perm(struct ftrace_event_call *tp_event, | ||
25 | struct perf_event *p_event) | ||
26 | { | ||
27 | /* No tracing, just counting, so no obvious leak */ | ||
28 | if (!(p_event->attr.sample_type & PERF_SAMPLE_RAW)) | ||
29 | return 0; | ||
30 | |||
31 | /* Some events are ok to be traced by non-root users... */ | ||
32 | if (p_event->attach_state == PERF_ATTACH_TASK) { | ||
33 | if (tp_event->flags & TRACE_EVENT_FL_CAP_ANY) | ||
34 | return 0; | ||
35 | } | ||
36 | |||
37 | /* | ||
38 | * ...otherwise raw tracepoint data can be a severe data leak, | ||
39 | * only allow root to have these. | ||
40 | */ | ||
41 | if (perf_paranoid_tracepoint_raw() && !capable(CAP_SYS_ADMIN)) | ||
42 | return -EPERM; | ||
43 | |||
44 | return 0; | ||
45 | } | ||
46 | |||
24 | static int perf_trace_event_init(struct ftrace_event_call *tp_event, | 47 | static int perf_trace_event_init(struct ftrace_event_call *tp_event, |
25 | struct perf_event *p_event) | 48 | struct perf_event *p_event) |
26 | { | 49 | { |
27 | struct hlist_head __percpu *list; | 50 | struct hlist_head __percpu *list; |
28 | int ret = -ENOMEM; | 51 | int ret; |
29 | int cpu; | 52 | int cpu; |
30 | 53 | ||
54 | ret = perf_trace_event_perm(tp_event, p_event); | ||
55 | if (ret) | ||
56 | return ret; | ||
57 | |||
31 | p_event->tp_event = tp_event; | 58 | p_event->tp_event = tp_event; |
32 | if (tp_event->perf_refcount++ > 0) | 59 | if (tp_event->perf_refcount++ > 0) |
33 | return 0; | 60 | return 0; |
34 | 61 | ||
62 | ret = -ENOMEM; | ||
63 | |||
35 | list = alloc_percpu(struct hlist_head); | 64 | list = alloc_percpu(struct hlist_head); |
36 | if (!list) | 65 | if (!list) |
37 | goto fail; | 66 | goto fail; |
diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c index 0725eeab193..35fde09b81d 100644 --- a/kernel/trace/trace_events.c +++ b/kernel/trace/trace_events.c | |||
@@ -27,6 +27,12 @@ | |||
27 | 27 | ||
28 | DEFINE_MUTEX(event_mutex); | 28 | DEFINE_MUTEX(event_mutex); |
29 | 29 | ||
30 | DEFINE_MUTEX(event_storage_mutex); | ||
31 | EXPORT_SYMBOL_GPL(event_storage_mutex); | ||
32 | |||
33 | char event_storage[EVENT_STORAGE_SIZE]; | ||
34 | EXPORT_SYMBOL_GPL(event_storage); | ||
35 | |||
30 | LIST_HEAD(ftrace_events); | 36 | LIST_HEAD(ftrace_events); |
31 | LIST_HEAD(ftrace_common_fields); | 37 | LIST_HEAD(ftrace_common_fields); |
32 | 38 | ||
diff --git a/kernel/trace/trace_export.c b/kernel/trace/trace_export.c index 4ba44deaac2..4b74d71705c 100644 --- a/kernel/trace/trace_export.c +++ b/kernel/trace/trace_export.c | |||
@@ -83,13 +83,19 @@ static void __always_unused ____ftrace_check_##name(void) \ | |||
83 | 83 | ||
84 | #undef __array | 84 | #undef __array |
85 | #define __array(type, item, len) \ | 85 | #define __array(type, item, len) \ |
86 | BUILD_BUG_ON(len > MAX_FILTER_STR_VAL); \ | 86 | do { \ |
87 | ret = trace_define_field(event_call, #type "[" #len "]", #item, \ | 87 | BUILD_BUG_ON(len > MAX_FILTER_STR_VAL); \ |
88 | mutex_lock(&event_storage_mutex); \ | ||
89 | snprintf(event_storage, sizeof(event_storage), \ | ||
90 | "%s[%d]", #type, len); \ | ||
91 | ret = trace_define_field(event_call, event_storage, #item, \ | ||
88 | offsetof(typeof(field), item), \ | 92 | offsetof(typeof(field), item), \ |
89 | sizeof(field.item), \ | 93 | sizeof(field.item), \ |
90 | is_signed_type(type), FILTER_OTHER); \ | 94 | is_signed_type(type), FILTER_OTHER); \ |
91 | if (ret) \ | 95 | mutex_unlock(&event_storage_mutex); \ |
92 | return ret; | 96 | if (ret) \ |
97 | return ret; \ | ||
98 | } while (0); | ||
93 | 99 | ||
94 | #undef __array_desc | 100 | #undef __array_desc |
95 | #define __array_desc(type, container, item, len) \ | 101 | #define __array_desc(type, container, item, len) \ |
diff --git a/kernel/watchdog.c b/kernel/watchdog.c index c812c4927ca..6e7b575ac33 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c | |||
@@ -57,6 +57,8 @@ static int __init hardlockup_panic_setup(char *str) | |||
57 | { | 57 | { |
58 | if (!strncmp(str, "panic", 5)) | 58 | if (!strncmp(str, "panic", 5)) |
59 | hardlockup_panic = 1; | 59 | hardlockup_panic = 1; |
60 | else if (!strncmp(str, "0", 1)) | ||
61 | no_watchdog = 1; | ||
60 | return 1; | 62 | return 1; |
61 | } | 63 | } |
62 | __setup("nmi_watchdog=", hardlockup_panic_setup); | 64 | __setup("nmi_watchdog=", hardlockup_panic_setup); |
@@ -548,13 +550,13 @@ static struct notifier_block __cpuinitdata cpu_nfb = { | |||
548 | .notifier_call = cpu_callback | 550 | .notifier_call = cpu_callback |
549 | }; | 551 | }; |
550 | 552 | ||
551 | static int __init spawn_watchdog_task(void) | 553 | void __init lockup_detector_init(void) |
552 | { | 554 | { |
553 | void *cpu = (void *)(long)smp_processor_id(); | 555 | void *cpu = (void *)(long)smp_processor_id(); |
554 | int err; | 556 | int err; |
555 | 557 | ||
556 | if (no_watchdog) | 558 | if (no_watchdog) |
557 | return 0; | 559 | return; |
558 | 560 | ||
559 | err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu); | 561 | err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu); |
560 | WARN_ON(notifier_to_errno(err)); | 562 | WARN_ON(notifier_to_errno(err)); |
@@ -562,6 +564,5 @@ static int __init spawn_watchdog_task(void) | |||
562 | cpu_callback(&cpu_nfb, CPU_ONLINE, cpu); | 564 | cpu_callback(&cpu_nfb, CPU_ONLINE, cpu); |
563 | register_cpu_notifier(&cpu_nfb); | 565 | register_cpu_notifier(&cpu_nfb); |
564 | 566 | ||
565 | return 0; | 567 | return; |
566 | } | 568 | } |
567 | early_initcall(spawn_watchdog_task); | ||
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index 28b42b9274d..2d05adb9840 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug | |||
@@ -173,7 +173,8 @@ config LOCKUP_DETECTOR | |||
173 | An NMI is generated every 60 seconds or so to check for hardlockups. | 173 | An NMI is generated every 60 seconds or so to check for hardlockups. |
174 | 174 | ||
175 | config HARDLOCKUP_DETECTOR | 175 | config HARDLOCKUP_DETECTOR |
176 | def_bool LOCKUP_DETECTOR && PERF_EVENTS && HAVE_PERF_EVENTS_NMI | 176 | def_bool LOCKUP_DETECTOR && PERF_EVENTS && HAVE_PERF_EVENTS_NMI && \ |
177 | !ARCH_HAS_NMI_WATCHDOG | ||
177 | 178 | ||
178 | config BOOTPARAM_SOFTLOCKUP_PANIC | 179 | config BOOTPARAM_SOFTLOCKUP_PANIC |
179 | bool "Panic (Reboot) On Soft Lockups" | 180 | bool "Panic (Reboot) On Soft Lockups" |
diff --git a/scripts/Makefile.build b/scripts/Makefile.build index 5ad25e17b6c..4eb99ab3405 100644 --- a/scripts/Makefile.build +++ b/scripts/Makefile.build | |||
@@ -214,17 +214,22 @@ ifdef BUILD_C_RECORDMCOUNT | |||
214 | # The empty.o file is created in the make process in order to determine | 214 | # The empty.o file is created in the make process in order to determine |
215 | # the target endianness and word size. It is made before all other C | 215 | # the target endianness and word size. It is made before all other C |
216 | # files, including recordmcount. | 216 | # files, including recordmcount. |
217 | cmd_record_mcount = if [ $(@) != "scripts/mod/empty.o" ]; then \ | 217 | sub_cmd_record_mcount = \ |
218 | $(objtree)/scripts/recordmcount "$(@)"; \ | 218 | if [ $(@) != "scripts/mod/empty.o" ]; then \ |
219 | fi; | 219 | $(objtree)/scripts/recordmcount "$(@)"; \ |
220 | fi; | ||
220 | else | 221 | else |
221 | cmd_record_mcount = set -e ; perl $(srctree)/scripts/recordmcount.pl "$(ARCH)" \ | 222 | sub_cmd_record_mcount = set -e ; perl $(srctree)/scripts/recordmcount.pl "$(ARCH)" \ |
222 | "$(if $(CONFIG_CPU_BIG_ENDIAN),big,little)" \ | 223 | "$(if $(CONFIG_CPU_BIG_ENDIAN),big,little)" \ |
223 | "$(if $(CONFIG_64BIT),64,32)" \ | 224 | "$(if $(CONFIG_64BIT),64,32)" \ |
224 | "$(OBJDUMP)" "$(OBJCOPY)" "$(CC) $(KBUILD_CFLAGS)" \ | 225 | "$(OBJDUMP)" "$(OBJCOPY)" "$(CC) $(KBUILD_CFLAGS)" \ |
225 | "$(LD)" "$(NM)" "$(RM)" "$(MV)" \ | 226 | "$(LD)" "$(NM)" "$(RM)" "$(MV)" \ |
226 | "$(if $(part-of-module),1,0)" "$(@)"; | 227 | "$(if $(part-of-module),1,0)" "$(@)"; |
227 | endif | 228 | endif |
229 | cmd_record_mcount = \ | ||
230 | if [ "$(findstring -pg,$(_c_flags))" = "-pg" ]; then \ | ||
231 | $(sub_cmd_record_mcount) \ | ||
232 | fi; | ||
228 | endif | 233 | endif |
229 | 234 | ||
230 | define rule_cc_o_c | 235 | define rule_cc_o_c |
diff --git a/scripts/kernel-doc b/scripts/kernel-doc index 39580a5dc5d..9f85012acf0 100755 --- a/scripts/kernel-doc +++ b/scripts/kernel-doc | |||
@@ -155,6 +155,8 @@ use strict; | |||
155 | # '@parameter' - name of a parameter | 155 | # '@parameter' - name of a parameter |
156 | # '%CONST' - name of a constant. | 156 | # '%CONST' - name of a constant. |
157 | 157 | ||
158 | ## init lots of data | ||
159 | |||
158 | my $errors = 0; | 160 | my $errors = 0; |
159 | my $warnings = 0; | 161 | my $warnings = 0; |
160 | my $anon_struct_union = 0; | 162 | my $anon_struct_union = 0; |
@@ -218,21 +220,14 @@ my %highlights_list = ( $type_constant, "\$1", | |||
218 | $type_param, "\$1" ); | 220 | $type_param, "\$1" ); |
219 | my $blankline_list = ""; | 221 | my $blankline_list = ""; |
220 | 222 | ||
221 | sub usage { | ||
222 | print "Usage: $0 [ -v ] [ -docbook | -html | -text | -man | -list ]\n"; | ||
223 | print " [ -no-doc-sections ]\n"; | ||
224 | print " [ -function funcname [ -function funcname ...] ]\n"; | ||
225 | print " [ -nofunction funcname [ -nofunction funcname ...] ]\n"; | ||
226 | print " c source file(s) > outputfile\n"; | ||
227 | print " -v : verbose output, more warnings & other info listed\n"; | ||
228 | exit 1; | ||
229 | } | ||
230 | |||
231 | # read arguments | 223 | # read arguments |
232 | if ($#ARGV == -1) { | 224 | if ($#ARGV == -1) { |
233 | usage(); | 225 | usage(); |
234 | } | 226 | } |
235 | 227 | ||
228 | my $kernelversion; | ||
229 | my $dohighlight = ""; | ||
230 | |||
236 | my $verbose = 0; | 231 | my $verbose = 0; |
237 | my $output_mode = "man"; | 232 | my $output_mode = "man"; |
238 | my $no_doc_sections = 0; | 233 | my $no_doc_sections = 0; |
@@ -245,7 +240,7 @@ my $man_date = ('January', 'February', 'March', 'April', 'May', 'June', | |||
245 | 'November', 'December')[(localtime)[4]] . | 240 | 'November', 'December')[(localtime)[4]] . |
246 | " " . ((localtime)[5]+1900); | 241 | " " . ((localtime)[5]+1900); |
247 | 242 | ||
248 | # Essentially these are globals | 243 | # Essentially these are globals. |
249 | # They probably want to be tidied up, made more localised or something. | 244 | # They probably want to be tidied up, made more localised or something. |
250 | # CAVEAT EMPTOR! Some of the others I localised may not want to be, which | 245 | # CAVEAT EMPTOR! Some of the others I localised may not want to be, which |
251 | # could cause "use of undefined value" or other bugs. | 246 | # could cause "use of undefined value" or other bugs. |
@@ -353,6 +348,18 @@ while ($ARGV[0] =~ m/^-(.*)/) { | |||
353 | } | 348 | } |
354 | } | 349 | } |
355 | 350 | ||
351 | # continue execution near EOF; | ||
352 | |||
353 | sub usage { | ||
354 | print "Usage: $0 [ -v ] [ -docbook | -html | -text | -man | -list ]\n"; | ||
355 | print " [ -no-doc-sections ]\n"; | ||
356 | print " [ -function funcname [ -function funcname ...] ]\n"; | ||
357 | print " [ -nofunction funcname [ -nofunction funcname ...] ]\n"; | ||
358 | print " c source file(s) > outputfile\n"; | ||
359 | print " -v : verbose output, more warnings & other info listed\n"; | ||
360 | exit 1; | ||
361 | } | ||
362 | |||
356 | # get kernel version from env | 363 | # get kernel version from env |
357 | sub get_kernel_version() { | 364 | sub get_kernel_version() { |
358 | my $version = 'unknown kernel version'; | 365 | my $version = 'unknown kernel version'; |
@@ -362,15 +369,6 @@ sub get_kernel_version() { | |||
362 | } | 369 | } |
363 | return $version; | 370 | return $version; |
364 | } | 371 | } |
365 | my $kernelversion = get_kernel_version(); | ||
366 | |||
367 | # generate a sequence of code that will splice in highlighting information | ||
368 | # using the s// operator. | ||
369 | my $dohighlight = ""; | ||
370 | foreach my $pattern (keys %highlights) { | ||
371 | # print STDERR "scanning pattern:$pattern, highlight:($highlights{$pattern})\n"; | ||
372 | $dohighlight .= "\$contents =~ s:$pattern:$highlights{$pattern}:gs;\n"; | ||
373 | } | ||
374 | 372 | ||
375 | ## | 373 | ## |
376 | # dumps section contents to arrays/hashes intended for that purpose. | 374 | # dumps section contents to arrays/hashes intended for that purpose. |
@@ -1851,34 +1849,6 @@ sub dump_function($$) { | |||
1851 | }); | 1849 | }); |
1852 | } | 1850 | } |
1853 | 1851 | ||
1854 | sub process_file($); | ||
1855 | |||
1856 | # Read the file that maps relative names to absolute names for | ||
1857 | # separate source and object directories and for shadow trees. | ||
1858 | if (open(SOURCE_MAP, "<.tmp_filelist.txt")) { | ||
1859 | my ($relname, $absname); | ||
1860 | while(<SOURCE_MAP>) { | ||
1861 | chop(); | ||
1862 | ($relname, $absname) = (split())[0..1]; | ||
1863 | $relname =~ s:^/+::; | ||
1864 | $source_map{$relname} = $absname; | ||
1865 | } | ||
1866 | close(SOURCE_MAP); | ||
1867 | } | ||
1868 | |||
1869 | foreach (@ARGV) { | ||
1870 | chomp; | ||
1871 | process_file($_); | ||
1872 | } | ||
1873 | if ($verbose && $errors) { | ||
1874 | print STDERR "$errors errors\n"; | ||
1875 | } | ||
1876 | if ($verbose && $warnings) { | ||
1877 | print STDERR "$warnings warnings\n"; | ||
1878 | } | ||
1879 | |||
1880 | exit($errors); | ||
1881 | |||
1882 | sub reset_state { | 1852 | sub reset_state { |
1883 | $function = ""; | 1853 | $function = ""; |
1884 | %constants = (); | 1854 | %constants = (); |
@@ -2285,3 +2255,39 @@ sub process_file($) { | |||
2285 | } | 2255 | } |
2286 | } | 2256 | } |
2287 | } | 2257 | } |
2258 | |||
2259 | |||
2260 | $kernelversion = get_kernel_version(); | ||
2261 | |||
2262 | # generate a sequence of code that will splice in highlighting information | ||
2263 | # using the s// operator. | ||
2264 | foreach my $pattern (keys %highlights) { | ||
2265 | # print STDERR "scanning pattern:$pattern, highlight:($highlights{$pattern})\n"; | ||
2266 | $dohighlight .= "\$contents =~ s:$pattern:$highlights{$pattern}:gs;\n"; | ||
2267 | } | ||
2268 | |||
2269 | # Read the file that maps relative names to absolute names for | ||
2270 | # separate source and object directories and for shadow trees. | ||
2271 | if (open(SOURCE_MAP, "<.tmp_filelist.txt")) { | ||
2272 | my ($relname, $absname); | ||
2273 | while(<SOURCE_MAP>) { | ||
2274 | chop(); | ||
2275 | ($relname, $absname) = (split())[0..1]; | ||
2276 | $relname =~ s:^/+::; | ||
2277 | $source_map{$relname} = $absname; | ||
2278 | } | ||
2279 | close(SOURCE_MAP); | ||
2280 | } | ||
2281 | |||
2282 | foreach (@ARGV) { | ||
2283 | chomp; | ||
2284 | process_file($_); | ||
2285 | } | ||
2286 | if ($verbose && $errors) { | ||
2287 | print STDERR "$errors errors\n"; | ||
2288 | } | ||
2289 | if ($verbose && $warnings) { | ||
2290 | print STDERR "$warnings warnings\n"; | ||
2291 | } | ||
2292 | |||
2293 | exit($errors); | ||
diff --git a/tools/perf/Documentation/perf-annotate.txt b/tools/perf/Documentation/perf-annotate.txt index b2c63309a65..6f5a498608b 100644 --- a/tools/perf/Documentation/perf-annotate.txt +++ b/tools/perf/Documentation/perf-annotate.txt | |||
@@ -24,12 +24,47 @@ OPTIONS | |||
24 | --input=:: | 24 | --input=:: |
25 | Input file name. (default: perf.data) | 25 | Input file name. (default: perf.data) |
26 | 26 | ||
27 | -d:: | ||
28 | --dsos=<dso[,dso...]>:: | ||
29 | Only consider symbols in these dsos. | ||
30 | -s:: | ||
31 | --symbol=<symbol>:: | ||
32 | Symbol to annotate. | ||
33 | |||
34 | -f:: | ||
35 | --force:: | ||
36 | Don't complain, do it. | ||
37 | |||
38 | -v:: | ||
39 | --verbose:: | ||
40 | Be more verbose. (Show symbol address, etc) | ||
41 | |||
42 | -D:: | ||
43 | --dump-raw-trace:: | ||
44 | Dump raw trace in ASCII. | ||
45 | |||
46 | -k:: | ||
47 | --vmlinux=<file>:: | ||
48 | vmlinux pathname. | ||
49 | |||
50 | -m:: | ||
51 | --modules:: | ||
52 | Load module symbols. WARNING: use only with -k and LIVE kernel. | ||
53 | |||
54 | -l:: | ||
55 | --print-line:: | ||
56 | Print matching source lines (may be slow). | ||
57 | |||
58 | -P:: | ||
59 | --full-paths:: | ||
60 | Don't shorten the displayed pathnames. | ||
61 | |||
27 | --stdio:: Use the stdio interface. | 62 | --stdio:: Use the stdio interface. |
28 | 63 | ||
29 | --tui:: Use the TUI interface Use of --tui requires a tty, if one is not | 64 | --tui:: Use the TUI interface Use of --tui requires a tty, if one is not |
30 | present, as when piping to other commands, the stdio interface is | 65 | present, as when piping to other commands, the stdio interface is |
31 | used. This interfaces starts by centering on the line with more | 66 | used. This interfaces starts by centering on the line with more |
32 | samples, TAB/UNTAB cycles thru the lines with more samples. | 67 | samples, TAB/UNTAB cycles through the lines with more samples. |
33 | 68 | ||
34 | SEE ALSO | 69 | SEE ALSO |
35 | -------- | 70 | -------- |
diff --git a/tools/perf/Documentation/perf-buildid-list.txt b/tools/perf/Documentation/perf-buildid-list.txt index 01b642c0bf8..5eaac6f26d5 100644 --- a/tools/perf/Documentation/perf-buildid-list.txt +++ b/tools/perf/Documentation/perf-buildid-list.txt | |||
@@ -18,6 +18,9 @@ perf report. | |||
18 | 18 | ||
19 | OPTIONS | 19 | OPTIONS |
20 | ------- | 20 | ------- |
21 | -H:: | ||
22 | --with-hits:: | ||
23 | Show only DSOs with hits. | ||
21 | -i:: | 24 | -i:: |
22 | --input=:: | 25 | --input=:: |
23 | Input file name. (default: perf.data) | 26 | Input file name. (default: perf.data) |
diff --git a/tools/perf/Documentation/perf-diff.txt b/tools/perf/Documentation/perf-diff.txt index 20d97d84ea1..74d7481ed7a 100644 --- a/tools/perf/Documentation/perf-diff.txt +++ b/tools/perf/Documentation/perf-diff.txt | |||
@@ -19,6 +19,18 @@ If no parameters are passed it will assume perf.data.old and perf.data. | |||
19 | 19 | ||
20 | OPTIONS | 20 | OPTIONS |
21 | ------- | 21 | ------- |
22 | -M:: | ||
23 | --displacement:: | ||
24 | Show position displacement relative to baseline. | ||
25 | |||
26 | -D:: | ||
27 | --dump-raw-trace:: | ||
28 | Dump raw trace in ASCII. | ||
29 | |||
30 | -m:: | ||
31 | --modules:: | ||
32 | Load module symbols. WARNING: use only with -k and LIVE kernel | ||
33 | |||
22 | -d:: | 34 | -d:: |
23 | --dsos=:: | 35 | --dsos=:: |
24 | Only consider symbols in these dsos. CSV that understands | 36 | Only consider symbols in these dsos. CSV that understands |
@@ -42,7 +54,7 @@ OPTIONS | |||
42 | --field-separator=:: | 54 | --field-separator=:: |
43 | 55 | ||
44 | Use a special separator character and don't pad with spaces, replacing | 56 | Use a special separator character and don't pad with spaces, replacing |
45 | all occurances of this separator in symbol names (and other output) | 57 | all occurrences of this separator in symbol names (and other output) |
46 | with a '.' character, that thus it's the only non valid separator. | 58 | with a '.' character, that thus it's the only non valid separator. |
47 | 59 | ||
48 | -v:: | 60 | -v:: |
@@ -50,6 +62,13 @@ OPTIONS | |||
50 | Be verbose, for instance, show the raw counts in addition to the | 62 | Be verbose, for instance, show the raw counts in addition to the |
51 | diff. | 63 | diff. |
52 | 64 | ||
65 | -f:: | ||
66 | --force:: | ||
67 | Don't complain, do it. | ||
68 | |||
69 | --symfs=<directory>:: | ||
70 | Look for files with symbols relative to this directory. | ||
71 | |||
53 | SEE ALSO | 72 | SEE ALSO |
54 | -------- | 73 | -------- |
55 | linkperf:perf-record[1] | 74 | linkperf:perf-record[1] |
diff --git a/tools/perf/Documentation/perf-kvm.txt b/tools/perf/Documentation/perf-kvm.txt index d004e19fe6d..dd84cb2f0a8 100644 --- a/tools/perf/Documentation/perf-kvm.txt +++ b/tools/perf/Documentation/perf-kvm.txt | |||
@@ -22,7 +22,7 @@ There are a couple of variants of perf kvm: | |||
22 | a performance counter profile of guest os in realtime | 22 | a performance counter profile of guest os in realtime |
23 | of an arbitrary workload. | 23 | of an arbitrary workload. |
24 | 24 | ||
25 | 'perf kvm record <command>' to record the performance couinter profile | 25 | 'perf kvm record <command>' to record the performance counter profile |
26 | of an arbitrary workload and save it into a perf data file. If both | 26 | of an arbitrary workload and save it into a perf data file. If both |
27 | --host and --guest are input, the perf data file name is perf.data.kvm. | 27 | --host and --guest are input, the perf data file name is perf.data.kvm. |
28 | If there is no --host but --guest, the file name is perf.data.guest. | 28 | If there is no --host but --guest, the file name is perf.data.guest. |
@@ -40,6 +40,12 @@ There are a couple of variants of perf kvm: | |||
40 | 40 | ||
41 | OPTIONS | 41 | OPTIONS |
42 | ------- | 42 | ------- |
43 | -i:: | ||
44 | --input=:: | ||
45 | Input file name. | ||
46 | -o:: | ||
47 | --output:: | ||
48 | Output file name. | ||
43 | --host=:: | 49 | --host=:: |
44 | Collect host side performance profile. | 50 | Collect host side performance profile. |
45 | --guest=:: | 51 | --guest=:: |
diff --git a/tools/perf/Documentation/perf-lock.txt b/tools/perf/Documentation/perf-lock.txt index b317102138c..921de259ea1 100644 --- a/tools/perf/Documentation/perf-lock.txt +++ b/tools/perf/Documentation/perf-lock.txt | |||
@@ -24,6 +24,21 @@ and statistics with this 'perf lock' command. | |||
24 | 24 | ||
25 | 'perf lock report' reports statistical data. | 25 | 'perf lock report' reports statistical data. |
26 | 26 | ||
27 | OPTIONS | ||
28 | ------- | ||
29 | |||
30 | -i:: | ||
31 | --input=<file>:: | ||
32 | Input file name. | ||
33 | |||
34 | -v:: | ||
35 | --verbose:: | ||
36 | Be more verbose (show symbol address, etc). | ||
37 | |||
38 | -D:: | ||
39 | --dump-raw-trace:: | ||
40 | Dump raw trace in ASCII. | ||
41 | |||
27 | SEE ALSO | 42 | SEE ALSO |
28 | -------- | 43 | -------- |
29 | linkperf:perf[1] | 44 | linkperf:perf[1] |
diff --git a/tools/perf/Documentation/perf-probe.txt b/tools/perf/Documentation/perf-probe.txt index 62de1b7f4e7..86b797a35aa 100644 --- a/tools/perf/Documentation/perf-probe.txt +++ b/tools/perf/Documentation/perf-probe.txt | |||
@@ -115,9 +115,9 @@ Each probe argument follows below syntax. | |||
115 | 115 | ||
116 | LINE SYNTAX | 116 | LINE SYNTAX |
117 | ----------- | 117 | ----------- |
118 | Line range is descripted by following syntax. | 118 | Line range is described by following syntax. |
119 | 119 | ||
120 | "FUNC[:RLN[+NUM|-RLN2]]|SRC:ALN[+NUM|-ALN2]" | 120 | "FUNC[:RLN[+NUM|-RLN2]]|SRC[:ALN[+NUM|-ALN2]]" |
121 | 121 | ||
122 | FUNC specifies the function name of showing lines. 'RLN' is the start line | 122 | FUNC specifies the function name of showing lines. 'RLN' is the start line |
123 | number from function entry line, and 'RLN2' is the end line number. As same as | 123 | number from function entry line, and 'RLN2' is the end line number. As same as |
diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt index a91f9f9e6e5..52462ae2645 100644 --- a/tools/perf/Documentation/perf-record.txt +++ b/tools/perf/Documentation/perf-record.txt | |||
@@ -39,15 +39,24 @@ OPTIONS | |||
39 | be passed as follows: '\mem:addr[:[r][w][x]]'. | 39 | be passed as follows: '\mem:addr[:[r][w][x]]'. |
40 | If you want to profile read-write accesses in 0x1000, just set | 40 | If you want to profile read-write accesses in 0x1000, just set |
41 | 'mem:0x1000:rw'. | 41 | 'mem:0x1000:rw'. |
42 | |||
43 | --filter=<filter>:: | ||
44 | Event filter. | ||
45 | |||
42 | -a:: | 46 | -a:: |
43 | System-wide collection. | 47 | --all-cpus:: |
48 | System-wide collection from all CPUs. | ||
44 | 49 | ||
45 | -l:: | 50 | -l:: |
46 | Scale counter values. | 51 | Scale counter values. |
47 | 52 | ||
48 | -p:: | 53 | -p:: |
49 | --pid=:: | 54 | --pid=:: |
50 | Record events on existing pid. | 55 | Record events on existing process ID. |
56 | |||
57 | -t:: | ||
58 | --tid=:: | ||
59 | Record events on existing thread ID. | ||
51 | 60 | ||
52 | -r:: | 61 | -r:: |
53 | --realtime=:: | 62 | --realtime=:: |
@@ -99,6 +108,11 @@ OPTIONS | |||
99 | --data:: | 108 | --data:: |
100 | Sample addresses. | 109 | Sample addresses. |
101 | 110 | ||
111 | -T:: | ||
112 | --timestamp:: | ||
113 | Sample timestamps. Use it with 'perf report -D' to see the timestamps, | ||
114 | for instance. | ||
115 | |||
102 | -n:: | 116 | -n:: |
103 | --no-samples:: | 117 | --no-samples:: |
104 | Don't sample. | 118 | Don't sample. |
@@ -109,8 +123,8 @@ Collect raw sample records from all opened counters (default for tracepoint coun | |||
109 | 123 | ||
110 | -C:: | 124 | -C:: |
111 | --cpu:: | 125 | --cpu:: |
112 | Collect samples only on the list of cpus provided. Multiple CPUs can be provided as a | 126 | Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a |
113 | comma-sperated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. | 127 | comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. |
114 | In per-thread mode with inheritance mode on (default), samples are captured only when | 128 | In per-thread mode with inheritance mode on (default), samples are captured only when |
115 | the thread executes on the designated CPUs. Default is to monitor all CPUs. | 129 | the thread executes on the designated CPUs. Default is to monitor all CPUs. |
116 | 130 | ||
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt index 12052c9ed0b..8ba03d6e539 100644 --- a/tools/perf/Documentation/perf-report.txt +++ b/tools/perf/Documentation/perf-report.txt | |||
@@ -20,6 +20,11 @@ OPTIONS | |||
20 | -i:: | 20 | -i:: |
21 | --input=:: | 21 | --input=:: |
22 | Input file name. (default: perf.data) | 22 | Input file name. (default: perf.data) |
23 | |||
24 | -v:: | ||
25 | --verbose:: | ||
26 | Be more verbose. (show symbol address, etc) | ||
27 | |||
23 | -d:: | 28 | -d:: |
24 | --dsos=:: | 29 | --dsos=:: |
25 | Only consider symbols in these dsos. CSV that understands | 30 | Only consider symbols in these dsos. CSV that understands |
@@ -27,6 +32,10 @@ OPTIONS | |||
27 | -n:: | 32 | -n:: |
28 | --show-nr-samples:: | 33 | --show-nr-samples:: |
29 | Show the number of samples for each symbol | 34 | Show the number of samples for each symbol |
35 | |||
36 | --showcpuutilization:: | ||
37 | Show sample percentage for different cpu modes. | ||
38 | |||
30 | -T:: | 39 | -T:: |
31 | --threads:: | 40 | --threads:: |
32 | Show per-thread event counters | 41 | Show per-thread event counters |
@@ -39,12 +48,24 @@ OPTIONS | |||
39 | Only consider these symbols. CSV that understands | 48 | Only consider these symbols. CSV that understands |
40 | file://filename entries. | 49 | file://filename entries. |
41 | 50 | ||
51 | -U:: | ||
52 | --hide-unresolved:: | ||
53 | Only display entries resolved to a symbol. | ||
54 | |||
42 | -s:: | 55 | -s:: |
43 | --sort=:: | 56 | --sort=:: |
44 | Sort by key(s): pid, comm, dso, symbol, parent. | 57 | Sort by key(s): pid, comm, dso, symbol, parent. |
45 | 58 | ||
59 | -p:: | ||
60 | --parent=<regex>:: | ||
61 | regex filter to identify parent, see: '--sort parent' | ||
62 | |||
63 | -x:: | ||
64 | --exclude-other:: | ||
65 | Only display entries with parent-match. | ||
66 | |||
46 | -w:: | 67 | -w:: |
47 | --field-width=:: | 68 | --column-widths=<width[,width...]>:: |
48 | Force each column width to the provided list, for large terminal | 69 | Force each column width to the provided list, for large terminal |
49 | readability. | 70 | readability. |
50 | 71 | ||
@@ -52,19 +73,26 @@ OPTIONS | |||
52 | --field-separator=:: | 73 | --field-separator=:: |
53 | 74 | ||
54 | Use a special separator character and don't pad with spaces, replacing | 75 | Use a special separator character and don't pad with spaces, replacing |
55 | all occurances of this separator in symbol names (and other output) | 76 | all occurrences of this separator in symbol names (and other output) |
56 | with a '.' character, that thus it's the only non valid separator. | 77 | with a '.' character, that thus it's the only non valid separator. |
57 | 78 | ||
79 | -D:: | ||
80 | --dump-raw-trace:: | ||
81 | Dump raw trace in ASCII. | ||
82 | |||
58 | -g [type,min]:: | 83 | -g [type,min]:: |
59 | --call-graph:: | 84 | --call-graph:: |
60 | Display callchains using type and min percent threshold. | 85 | Display call chains using type and min percent threshold. |
61 | type can be either: | 86 | type can be either: |
62 | - flat: single column, linear exposure of callchains. | 87 | - flat: single column, linear exposure of call chains. |
63 | - graph: use a graph tree, displaying absolute overhead rates. | 88 | - graph: use a graph tree, displaying absolute overhead rates. |
64 | - fractal: like graph, but displays relative rates. Each branch of | 89 | - fractal: like graph, but displays relative rates. Each branch of |
65 | the tree is considered as a new profiled object. + | 90 | the tree is considered as a new profiled object. + |
66 | Default: fractal,0.5. | 91 | Default: fractal,0.5. |
67 | 92 | ||
93 | --pretty=<key>:: | ||
94 | Pretty printing style. key: normal, raw | ||
95 | |||
68 | --stdio:: Use the stdio interface. | 96 | --stdio:: Use the stdio interface. |
69 | 97 | ||
70 | --tui:: Use the TUI interface, that is integrated with annotate and allows | 98 | --tui:: Use the TUI interface, that is integrated with annotate and allows |
@@ -72,6 +100,25 @@ OPTIONS | |||
72 | requires a tty, if one is not present, as when piping to other | 100 | requires a tty, if one is not present, as when piping to other |
73 | commands, the stdio interface is used. | 101 | commands, the stdio interface is used. |
74 | 102 | ||
103 | -k:: | ||
104 | --vmlinux=<file>:: | ||
105 | vmlinux pathname | ||
106 | |||
107 | --kallsyms=<file>:: | ||
108 | kallsyms pathname | ||
109 | |||
110 | -m:: | ||
111 | --modules:: | ||
112 | Load module symbols. WARNING: This should only be used with -k and | ||
113 | a LIVE kernel. | ||
114 | |||
115 | -f:: | ||
116 | --force:: | ||
117 | Don't complain, do it. | ||
118 | |||
119 | --symfs=<directory>:: | ||
120 | Look for files with symbols relative to this directory. | ||
121 | |||
75 | SEE ALSO | 122 | SEE ALSO |
76 | -------- | 123 | -------- |
77 | linkperf:perf-stat[1] | 124 | linkperf:perf-stat[1] |
diff --git a/tools/perf/Documentation/perf-sched.txt b/tools/perf/Documentation/perf-sched.txt index 8417644a616..46822d5fde1 100644 --- a/tools/perf/Documentation/perf-sched.txt +++ b/tools/perf/Documentation/perf-sched.txt | |||
@@ -8,11 +8,11 @@ perf-sched - Tool to trace/measure scheduler properties (latencies) | |||
8 | SYNOPSIS | 8 | SYNOPSIS |
9 | -------- | 9 | -------- |
10 | [verse] | 10 | [verse] |
11 | 'perf sched' {record|latency|replay|trace} | 11 | 'perf sched' {record|latency|map|replay|trace} |
12 | 12 | ||
13 | DESCRIPTION | 13 | DESCRIPTION |
14 | ----------- | 14 | ----------- |
15 | There are four variants of perf sched: | 15 | There are five variants of perf sched: |
16 | 16 | ||
17 | 'perf sched record <command>' to record the scheduling events | 17 | 'perf sched record <command>' to record the scheduling events |
18 | of an arbitrary workload. | 18 | of an arbitrary workload. |
@@ -30,8 +30,22 @@ There are four variants of perf sched: | |||
30 | of the workload as it occurred when it was recorded - and can repeat | 30 | of the workload as it occurred when it was recorded - and can repeat |
31 | it a number of times, measuring its performance.) | 31 | it a number of times, measuring its performance.) |
32 | 32 | ||
33 | 'perf sched map' to print a textual context-switching outline of | ||
34 | workload captured via perf sched record. Columns stand for | ||
35 | individual CPUs, and the two-letter shortcuts stand for tasks that | ||
36 | are running on a CPU. A '*' denotes the CPU that had the event, and | ||
37 | a dot signals an idle CPU. | ||
38 | |||
33 | OPTIONS | 39 | OPTIONS |
34 | ------- | 40 | ------- |
41 | -i:: | ||
42 | --input=<file>:: | ||
43 | Input file name. (default: perf.data) | ||
44 | |||
45 | -v:: | ||
46 | --verbose:: | ||
47 | Be more verbose. (show symbol address, etc) | ||
48 | |||
35 | -D:: | 49 | -D:: |
36 | --dump-raw-trace=:: | 50 | --dump-raw-trace=:: |
37 | Display verbose dump of the sched data. | 51 | Display verbose dump of the sched data. |
diff --git a/tools/perf/Documentation/perf-trace-perl.txt b/tools/perf/Documentation/perf-script-perl.txt index ee6525ee6d6..5bb41e55a3a 100644 --- a/tools/perf/Documentation/perf-trace-perl.txt +++ b/tools/perf/Documentation/perf-script-perl.txt | |||
@@ -1,19 +1,19 @@ | |||
1 | perf-trace-perl(1) | 1 | perf-script-perl(1) |
2 | ================== | 2 | ================== |
3 | 3 | ||
4 | NAME | 4 | NAME |
5 | ---- | 5 | ---- |
6 | perf-trace-perl - Process trace data with a Perl script | 6 | perf-script-perl - Process trace data with a Perl script |
7 | 7 | ||
8 | SYNOPSIS | 8 | SYNOPSIS |
9 | -------- | 9 | -------- |
10 | [verse] | 10 | [verse] |
11 | 'perf trace' [-s [Perl]:script[.pl] ] | 11 | 'perf script' [-s [Perl]:script[.pl] ] |
12 | 12 | ||
13 | DESCRIPTION | 13 | DESCRIPTION |
14 | ----------- | 14 | ----------- |
15 | 15 | ||
16 | This perf trace option is used to process perf trace data using perf's | 16 | This perf script option is used to process perf script data using perf's |
17 | built-in Perl interpreter. It reads and processes the input file and | 17 | built-in Perl interpreter. It reads and processes the input file and |
18 | displays the results of the trace analysis implemented in the given | 18 | displays the results of the trace analysis implemented in the given |
19 | Perl script, if any. | 19 | Perl script, if any. |
@@ -21,7 +21,7 @@ Perl script, if any. | |||
21 | STARTER SCRIPTS | 21 | STARTER SCRIPTS |
22 | --------------- | 22 | --------------- |
23 | 23 | ||
24 | You can avoid reading the rest of this document by running 'perf trace | 24 | You can avoid reading the rest of this document by running 'perf script |
25 | -g perl' in the same directory as an existing perf.data trace file. | 25 | -g perl' in the same directory as an existing perf.data trace file. |
26 | That will generate a starter script containing a handler for each of | 26 | That will generate a starter script containing a handler for each of |
27 | the event types in the trace file; it simply prints every available | 27 | the event types in the trace file; it simply prints every available |
@@ -30,13 +30,13 @@ field for each event in the trace file. | |||
30 | You can also look at the existing scripts in | 30 | You can also look at the existing scripts in |
31 | ~/libexec/perf-core/scripts/perl for typical examples showing how to | 31 | ~/libexec/perf-core/scripts/perl for typical examples showing how to |
32 | do basic things like aggregate event data, print results, etc. Also, | 32 | do basic things like aggregate event data, print results, etc. Also, |
33 | the check-perf-trace.pl script, while not interesting for its results, | 33 | the check-perf-script.pl script, while not interesting for its results, |
34 | attempts to exercise all of the main scripting features. | 34 | attempts to exercise all of the main scripting features. |
35 | 35 | ||
36 | EVENT HANDLERS | 36 | EVENT HANDLERS |
37 | -------------- | 37 | -------------- |
38 | 38 | ||
39 | When perf trace is invoked using a trace script, a user-defined | 39 | When perf script is invoked using a trace script, a user-defined |
40 | 'handler function' is called for each event in the trace. If there's | 40 | 'handler function' is called for each event in the trace. If there's |
41 | no handler function defined for a given event type, the event is | 41 | no handler function defined for a given event type, the event is |
42 | ignored (or passed to a 'trace_handled' function, see below) and the | 42 | ignored (or passed to a 'trace_handled' function, see below) and the |
@@ -112,13 +112,13 @@ write a useful trace script. The sections below cover the rest. | |||
112 | SCRIPT LAYOUT | 112 | SCRIPT LAYOUT |
113 | ------------- | 113 | ------------- |
114 | 114 | ||
115 | Every perf trace Perl script should start by setting up a Perl module | 115 | Every perf script Perl script should start by setting up a Perl module |
116 | search path and 'use'ing a few support modules (see module | 116 | search path and 'use'ing a few support modules (see module |
117 | descriptions below): | 117 | descriptions below): |
118 | 118 | ||
119 | ---- | 119 | ---- |
120 | use lib "$ENV{'PERF_EXEC_PATH'}/scripts/perl/Perf-Trace-Util/lib"; | 120 | use lib "$ENV{'PERF_EXEC_PATH'}/scripts/perl/perf-script-Util/lib"; |
121 | use lib "./Perf-Trace-Util/lib"; | 121 | use lib "./perf-script-Util/lib"; |
122 | use Perf::Trace::Core; | 122 | use Perf::Trace::Core; |
123 | use Perf::Trace::Context; | 123 | use Perf::Trace::Context; |
124 | use Perf::Trace::Util; | 124 | use Perf::Trace::Util; |
@@ -162,7 +162,7 @@ sub trace_unhandled | |||
162 | ---- | 162 | ---- |
163 | 163 | ||
164 | The remaining sections provide descriptions of each of the available | 164 | The remaining sections provide descriptions of each of the available |
165 | built-in perf trace Perl modules and their associated functions. | 165 | built-in perf script Perl modules and their associated functions. |
166 | 166 | ||
167 | AVAILABLE MODULES AND FUNCTIONS | 167 | AVAILABLE MODULES AND FUNCTIONS |
168 | ------------------------------- | 168 | ------------------------------- |
@@ -170,7 +170,7 @@ AVAILABLE MODULES AND FUNCTIONS | |||
170 | The following sections describe the functions and variables available | 170 | The following sections describe the functions and variables available |
171 | via the various Perf::Trace::* Perl modules. To use the functions and | 171 | via the various Perf::Trace::* Perl modules. To use the functions and |
172 | variables from the given module, add the corresponding 'use | 172 | variables from the given module, add the corresponding 'use |
173 | Perf::Trace::XXX' line to your perf trace script. | 173 | Perf::Trace::XXX' line to your perf script script. |
174 | 174 | ||
175 | Perf::Trace::Core Module | 175 | Perf::Trace::Core Module |
176 | ~~~~~~~~~~~~~~~~~~~~~~~~ | 176 | ~~~~~~~~~~~~~~~~~~~~~~~~ |
@@ -204,7 +204,7 @@ argument. | |||
204 | Perf::Trace::Util Module | 204 | Perf::Trace::Util Module |
205 | ~~~~~~~~~~~~~~~~~~~~~~~~ | 205 | ~~~~~~~~~~~~~~~~~~~~~~~~ |
206 | 206 | ||
207 | Various utility functions for use with perf trace: | 207 | Various utility functions for use with perf script: |
208 | 208 | ||
209 | nsecs($secs, $nsecs) - returns total nsecs given secs/nsecs pair | 209 | nsecs($secs, $nsecs) - returns total nsecs given secs/nsecs pair |
210 | nsecs_secs($nsecs) - returns whole secs portion given nsecs | 210 | nsecs_secs($nsecs) - returns whole secs portion given nsecs |
@@ -214,4 +214,4 @@ Various utility functions for use with perf trace: | |||
214 | 214 | ||
215 | SEE ALSO | 215 | SEE ALSO |
216 | -------- | 216 | -------- |
217 | linkperf:perf-trace[1] | 217 | linkperf:perf-script[1] |
diff --git a/tools/perf/Documentation/perf-trace-python.txt b/tools/perf/Documentation/perf-script-python.txt index 693be804dd3..36b38277422 100644 --- a/tools/perf/Documentation/perf-trace-python.txt +++ b/tools/perf/Documentation/perf-script-python.txt | |||
@@ -1,19 +1,19 @@ | |||
1 | perf-trace-python(1) | 1 | perf-script-python(1) |
2 | ==================== | 2 | ==================== |
3 | 3 | ||
4 | NAME | 4 | NAME |
5 | ---- | 5 | ---- |
6 | perf-trace-python - Process trace data with a Python script | 6 | perf-script-python - Process trace data with a Python script |
7 | 7 | ||
8 | SYNOPSIS | 8 | SYNOPSIS |
9 | -------- | 9 | -------- |
10 | [verse] | 10 | [verse] |
11 | 'perf trace' [-s [Python]:script[.py] ] | 11 | 'perf script' [-s [Python]:script[.py] ] |
12 | 12 | ||
13 | DESCRIPTION | 13 | DESCRIPTION |
14 | ----------- | 14 | ----------- |
15 | 15 | ||
16 | This perf trace option is used to process perf trace data using perf's | 16 | This perf script option is used to process perf script data using perf's |
17 | built-in Python interpreter. It reads and processes the input file and | 17 | built-in Python interpreter. It reads and processes the input file and |
18 | displays the results of the trace analysis implemented in the given | 18 | displays the results of the trace analysis implemented in the given |
19 | Python script, if any. | 19 | Python script, if any. |
@@ -23,15 +23,15 @@ A QUICK EXAMPLE | |||
23 | 23 | ||
24 | This section shows the process, start to finish, of creating a working | 24 | This section shows the process, start to finish, of creating a working |
25 | Python script that aggregates and extracts useful information from a | 25 | Python script that aggregates and extracts useful information from a |
26 | raw perf trace stream. You can avoid reading the rest of this | 26 | raw perf script stream. You can avoid reading the rest of this |
27 | document if an example is enough for you; the rest of the document | 27 | document if an example is enough for you; the rest of the document |
28 | provides more details on each step and lists the library functions | 28 | provides more details on each step and lists the library functions |
29 | available to script writers. | 29 | available to script writers. |
30 | 30 | ||
31 | This example actually details the steps that were used to create the | 31 | This example actually details the steps that were used to create the |
32 | 'syscall-counts' script you see when you list the available perf trace | 32 | 'syscall-counts' script you see when you list the available perf script |
33 | scripts via 'perf trace -l'. As such, this script also shows how to | 33 | scripts via 'perf script -l'. As such, this script also shows how to |
34 | integrate your script into the list of general-purpose 'perf trace' | 34 | integrate your script into the list of general-purpose 'perf script' |
35 | scripts listed by that command. | 35 | scripts listed by that command. |
36 | 36 | ||
37 | The syscall-counts script is a simple script, but demonstrates all the | 37 | The syscall-counts script is a simple script, but demonstrates all the |
@@ -105,31 +105,31 @@ That single stream will be recorded in a file in the current directory | |||
105 | called perf.data. | 105 | called perf.data. |
106 | 106 | ||
107 | Once we have a perf.data file containing our data, we can use the -g | 107 | Once we have a perf.data file containing our data, we can use the -g |
108 | 'perf trace' option to generate a Python script that will contain a | 108 | 'perf script' option to generate a Python script that will contain a |
109 | callback handler for each event type found in the perf.data trace | 109 | callback handler for each event type found in the perf.data trace |
110 | stream (for more details, see the STARTER SCRIPTS section). | 110 | stream (for more details, see the STARTER SCRIPTS section). |
111 | 111 | ||
112 | ---- | 112 | ---- |
113 | # perf trace -g python | 113 | # perf script -g python |
114 | generated Python script: perf-trace.py | 114 | generated Python script: perf-script.py |
115 | 115 | ||
116 | The output file created also in the current directory is named | 116 | The output file created also in the current directory is named |
117 | perf-trace.py. Here's the file in its entirety: | 117 | perf-script.py. Here's the file in its entirety: |
118 | 118 | ||
119 | # perf trace event handlers, generated by perf trace -g python | 119 | # perf script event handlers, generated by perf script -g python |
120 | # Licensed under the terms of the GNU GPL License version 2 | 120 | # Licensed under the terms of the GNU GPL License version 2 |
121 | 121 | ||
122 | # The common_* event handler fields are the most useful fields common to | 122 | # The common_* event handler fields are the most useful fields common to |
123 | # all events. They don't necessarily correspond to the 'common_*' fields | 123 | # all events. They don't necessarily correspond to the 'common_*' fields |
124 | # in the format files. Those fields not available as handler params can | 124 | # in the format files. Those fields not available as handler params can |
125 | # be retrieved using Python functions of the form common_*(context). | 125 | # be retrieved using Python functions of the form common_*(context). |
126 | # See the perf-trace-python Documentation for the list of available functions. | 126 | # See the perf-script-python Documentation for the list of available functions. |
127 | 127 | ||
128 | import os | 128 | import os |
129 | import sys | 129 | import sys |
130 | 130 | ||
131 | sys.path.append(os.environ['PERF_EXEC_PATH'] + \ | 131 | sys.path.append(os.environ['PERF_EXEC_PATH'] + \ |
132 | '/scripts/python/Perf-Trace-Util/lib/Perf/Trace') | 132 | '/scripts/python/perf-script-Util/lib/Perf/Trace') |
133 | 133 | ||
134 | from perf_trace_context import * | 134 | from perf_trace_context import * |
135 | from Core import * | 135 | from Core import * |
@@ -160,7 +160,7 @@ def print_header(event_name, cpu, secs, nsecs, pid, comm): | |||
160 | ---- | 160 | ---- |
161 | 161 | ||
162 | At the top is a comment block followed by some import statements and a | 162 | At the top is a comment block followed by some import statements and a |
163 | path append which every perf trace script should include. | 163 | path append which every perf script script should include. |
164 | 164 | ||
165 | Following that are a couple generated functions, trace_begin() and | 165 | Following that are a couple generated functions, trace_begin() and |
166 | trace_end(), which are called at the beginning and the end of the | 166 | trace_end(), which are called at the beginning and the end of the |
@@ -189,8 +189,8 @@ simply a utility function used for that purpose. Let's rename the | |||
189 | script and run it to see the default output: | 189 | script and run it to see the default output: |
190 | 190 | ||
191 | ---- | 191 | ---- |
192 | # mv perf-trace.py syscall-counts.py | 192 | # mv perf-script.py syscall-counts.py |
193 | # perf trace -s syscall-counts.py | 193 | # perf script -s syscall-counts.py |
194 | 194 | ||
195 | raw_syscalls__sys_enter 1 00840.847582083 7506 perf id=1, args= | 195 | raw_syscalls__sys_enter 1 00840.847582083 7506 perf id=1, args= |
196 | raw_syscalls__sys_enter 1 00840.847595764 7506 perf id=1, args= | 196 | raw_syscalls__sys_enter 1 00840.847595764 7506 perf id=1, args= |
@@ -216,7 +216,7 @@ import os | |||
216 | import sys | 216 | import sys |
217 | 217 | ||
218 | sys.path.append(os.environ['PERF_EXEC_PATH'] + \ | 218 | sys.path.append(os.environ['PERF_EXEC_PATH'] + \ |
219 | '/scripts/python/Perf-Trace-Util/lib/Perf/Trace') | 219 | '/scripts/python/perf-script-Util/lib/Perf/Trace') |
220 | 220 | ||
221 | from perf_trace_context import * | 221 | from perf_trace_context import * |
222 | from Core import * | 222 | from Core import * |
@@ -279,7 +279,7 @@ import os | |||
279 | import sys | 279 | import sys |
280 | 280 | ||
281 | sys.path.append(os.environ['PERF_EXEC_PATH'] + \ | 281 | sys.path.append(os.environ['PERF_EXEC_PATH'] + \ |
282 | '/scripts/python/Perf-Trace-Util/lib/Perf/Trace') | 282 | '/scripts/python/perf-script-Util/lib/Perf/Trace') |
283 | 283 | ||
284 | from perf_trace_context import * | 284 | from perf_trace_context import * |
285 | from Core import * | 285 | from Core import * |
@@ -315,7 +315,7 @@ def print_syscall_totals(): | |||
315 | 315 | ||
316 | The script can be run just as before: | 316 | The script can be run just as before: |
317 | 317 | ||
318 | # perf trace -s syscall-counts.py | 318 | # perf script -s syscall-counts.py |
319 | 319 | ||
320 | So those are the essential steps in writing and running a script. The | 320 | So those are the essential steps in writing and running a script. The |
321 | process can be generalized to any tracepoint or set of tracepoints | 321 | process can be generalized to any tracepoint or set of tracepoints |
@@ -324,17 +324,17 @@ interested in by looking at the list of available events shown by | |||
324 | 'perf list' and/or look in /sys/kernel/debug/tracing events for | 324 | 'perf list' and/or look in /sys/kernel/debug/tracing events for |
325 | detailed event and field info, record the corresponding trace data | 325 | detailed event and field info, record the corresponding trace data |
326 | using 'perf record', passing it the list of interesting events, | 326 | using 'perf record', passing it the list of interesting events, |
327 | generate a skeleton script using 'perf trace -g python' and modify the | 327 | generate a skeleton script using 'perf script -g python' and modify the |
328 | code to aggregate and display it for your particular needs. | 328 | code to aggregate and display it for your particular needs. |
329 | 329 | ||
330 | After you've done that you may end up with a general-purpose script | 330 | After you've done that you may end up with a general-purpose script |
331 | that you want to keep around and have available for future use. By | 331 | that you want to keep around and have available for future use. By |
332 | writing a couple of very simple shell scripts and putting them in the | 332 | writing a couple of very simple shell scripts and putting them in the |
333 | right place, you can have your script listed alongside the other | 333 | right place, you can have your script listed alongside the other |
334 | scripts listed by the 'perf trace -l' command e.g.: | 334 | scripts listed by the 'perf script -l' command e.g.: |
335 | 335 | ||
336 | ---- | 336 | ---- |
337 | root@tropicana:~# perf trace -l | 337 | root@tropicana:~# perf script -l |
338 | List of available trace scripts: | 338 | List of available trace scripts: |
339 | workqueue-stats workqueue stats (ins/exe/create/destroy) | 339 | workqueue-stats workqueue stats (ins/exe/create/destroy) |
340 | wakeup-latency system-wide min/max/avg wakeup latency | 340 | wakeup-latency system-wide min/max/avg wakeup latency |
@@ -365,14 +365,14 @@ perf record -a -e raw_syscalls:sys_enter | |||
365 | The 'report' script is also a shell script with the same base name as | 365 | The 'report' script is also a shell script with the same base name as |
366 | your script, but with -report appended. It should also be located in | 366 | your script, but with -report appended. It should also be located in |
367 | the perf/scripts/python/bin directory. In that script, you write the | 367 | the perf/scripts/python/bin directory. In that script, you write the |
368 | 'perf trace -s' command-line needed for running your script: | 368 | 'perf script -s' command-line needed for running your script: |
369 | 369 | ||
370 | ---- | 370 | ---- |
371 | # cat kernel-source/tools/perf/scripts/python/bin/syscall-counts-report | 371 | # cat kernel-source/tools/perf/scripts/python/bin/syscall-counts-report |
372 | 372 | ||
373 | #!/bin/bash | 373 | #!/bin/bash |
374 | # description: system-wide syscall counts | 374 | # description: system-wide syscall counts |
375 | perf trace -s ~/libexec/perf-core/scripts/python/syscall-counts.py | 375 | perf script -s ~/libexec/perf-core/scripts/python/syscall-counts.py |
376 | ---- | 376 | ---- |
377 | 377 | ||
378 | Note that the location of the Python script given in the shell script | 378 | Note that the location of the Python script given in the shell script |
@@ -390,17 +390,17 @@ total 32 | |||
390 | drwxr-xr-x 4 trz trz 4096 2010-01-26 22:30 . | 390 | drwxr-xr-x 4 trz trz 4096 2010-01-26 22:30 . |
391 | drwxr-xr-x 4 trz trz 4096 2010-01-26 22:29 .. | 391 | drwxr-xr-x 4 trz trz 4096 2010-01-26 22:29 .. |
392 | drwxr-xr-x 2 trz trz 4096 2010-01-26 22:29 bin | 392 | drwxr-xr-x 2 trz trz 4096 2010-01-26 22:29 bin |
393 | -rw-r--r-- 1 trz trz 2548 2010-01-26 22:29 check-perf-trace.py | 393 | -rw-r--r-- 1 trz trz 2548 2010-01-26 22:29 check-perf-script.py |
394 | drwxr-xr-x 3 trz trz 4096 2010-01-26 22:49 Perf-Trace-Util | 394 | drwxr-xr-x 3 trz trz 4096 2010-01-26 22:49 perf-script-Util |
395 | -rw-r--r-- 1 trz trz 1462 2010-01-26 22:30 syscall-counts.py | 395 | -rw-r--r-- 1 trz trz 1462 2010-01-26 22:30 syscall-counts.py |
396 | ---- | 396 | ---- |
397 | 397 | ||
398 | Once you've done that (don't forget to do a new 'make install', | 398 | Once you've done that (don't forget to do a new 'make install', |
399 | otherwise your script won't show up at run-time), 'perf trace -l' | 399 | otherwise your script won't show up at run-time), 'perf script -l' |
400 | should show a new entry for your script: | 400 | should show a new entry for your script: |
401 | 401 | ||
402 | ---- | 402 | ---- |
403 | root@tropicana:~# perf trace -l | 403 | root@tropicana:~# perf script -l |
404 | List of available trace scripts: | 404 | List of available trace scripts: |
405 | workqueue-stats workqueue stats (ins/exe/create/destroy) | 405 | workqueue-stats workqueue stats (ins/exe/create/destroy) |
406 | wakeup-latency system-wide min/max/avg wakeup latency | 406 | wakeup-latency system-wide min/max/avg wakeup latency |
@@ -409,19 +409,19 @@ List of available trace scripts: | |||
409 | syscall-counts system-wide syscall counts | 409 | syscall-counts system-wide syscall counts |
410 | ---- | 410 | ---- |
411 | 411 | ||
412 | You can now perform the record step via 'perf trace record': | 412 | You can now perform the record step via 'perf script record': |
413 | 413 | ||
414 | # perf trace record syscall-counts | 414 | # perf script record syscall-counts |
415 | 415 | ||
416 | and display the output using 'perf trace report': | 416 | and display the output using 'perf script report': |
417 | 417 | ||
418 | # perf trace report syscall-counts | 418 | # perf script report syscall-counts |
419 | 419 | ||
420 | STARTER SCRIPTS | 420 | STARTER SCRIPTS |
421 | --------------- | 421 | --------------- |
422 | 422 | ||
423 | You can quickly get started writing a script for a particular set of | 423 | You can quickly get started writing a script for a particular set of |
424 | trace data by generating a skeleton script using 'perf trace -g | 424 | trace data by generating a skeleton script using 'perf script -g |
425 | python' in the same directory as an existing perf.data trace file. | 425 | python' in the same directory as an existing perf.data trace file. |
426 | That will generate a starter script containing a handler for each of | 426 | That will generate a starter script containing a handler for each of |
427 | the event types in the trace file; it simply prints every available | 427 | the event types in the trace file; it simply prints every available |
@@ -430,13 +430,13 @@ field for each event in the trace file. | |||
430 | You can also look at the existing scripts in | 430 | You can also look at the existing scripts in |
431 | ~/libexec/perf-core/scripts/python for typical examples showing how to | 431 | ~/libexec/perf-core/scripts/python for typical examples showing how to |
432 | do basic things like aggregate event data, print results, etc. Also, | 432 | do basic things like aggregate event data, print results, etc. Also, |
433 | the check-perf-trace.py script, while not interesting for its results, | 433 | the check-perf-script.py script, while not interesting for its results, |
434 | attempts to exercise all of the main scripting features. | 434 | attempts to exercise all of the main scripting features. |
435 | 435 | ||
436 | EVENT HANDLERS | 436 | EVENT HANDLERS |
437 | -------------- | 437 | -------------- |
438 | 438 | ||
439 | When perf trace is invoked using a trace script, a user-defined | 439 | When perf script is invoked using a trace script, a user-defined |
440 | 'handler function' is called for each event in the trace. If there's | 440 | 'handler function' is called for each event in the trace. If there's |
441 | no handler function defined for a given event type, the event is | 441 | no handler function defined for a given event type, the event is |
442 | ignored (or passed to a 'trace_handled' function, see below) and the | 442 | ignored (or passed to a 'trace_handled' function, see below) and the |
@@ -510,7 +510,7 @@ write a useful trace script. The sections below cover the rest. | |||
510 | SCRIPT LAYOUT | 510 | SCRIPT LAYOUT |
511 | ------------- | 511 | ------------- |
512 | 512 | ||
513 | Every perf trace Python script should start by setting up a Python | 513 | Every perf script Python script should start by setting up a Python |
514 | module search path and 'import'ing a few support modules (see module | 514 | module search path and 'import'ing a few support modules (see module |
515 | descriptions below): | 515 | descriptions below): |
516 | 516 | ||
@@ -519,7 +519,7 @@ descriptions below): | |||
519 | import sys | 519 | import sys |
520 | 520 | ||
521 | sys.path.append(os.environ['PERF_EXEC_PATH'] + \ | 521 | sys.path.append(os.environ['PERF_EXEC_PATH'] + \ |
522 | '/scripts/python/Perf-Trace-Util/lib/Perf/Trace') | 522 | '/scripts/python/perf-script-Util/lib/Perf/Trace') |
523 | 523 | ||
524 | from perf_trace_context import * | 524 | from perf_trace_context import * |
525 | from Core import * | 525 | from Core import * |
@@ -559,15 +559,15 @@ def trace_unhandled(event_name, context, common_cpu, common_secs, | |||
559 | ---- | 559 | ---- |
560 | 560 | ||
561 | The remaining sections provide descriptions of each of the available | 561 | The remaining sections provide descriptions of each of the available |
562 | built-in perf trace Python modules and their associated functions. | 562 | built-in perf script Python modules and their associated functions. |
563 | 563 | ||
564 | AVAILABLE MODULES AND FUNCTIONS | 564 | AVAILABLE MODULES AND FUNCTIONS |
565 | ------------------------------- | 565 | ------------------------------- |
566 | 566 | ||
567 | The following sections describe the functions and variables available | 567 | The following sections describe the functions and variables available |
568 | via the various perf trace Python modules. To use the functions and | 568 | via the various perf script Python modules. To use the functions and |
569 | variables from the given module, add the corresponding 'from XXXX | 569 | variables from the given module, add the corresponding 'from XXXX |
570 | import' line to your perf trace script. | 570 | import' line to your perf script script. |
571 | 571 | ||
572 | Core.py Module | 572 | Core.py Module |
573 | ~~~~~~~~~~~~~~ | 573 | ~~~~~~~~~~~~~~ |
@@ -610,7 +610,7 @@ argument. | |||
610 | Util.py Module | 610 | Util.py Module |
611 | ~~~~~~~~~~~~~~ | 611 | ~~~~~~~~~~~~~~ |
612 | 612 | ||
613 | Various utility functions for use with perf trace: | 613 | Various utility functions for use with perf script: |
614 | 614 | ||
615 | nsecs(secs, nsecs) - returns total nsecs given secs/nsecs pair | 615 | nsecs(secs, nsecs) - returns total nsecs given secs/nsecs pair |
616 | nsecs_secs(nsecs) - returns whole secs portion given nsecs | 616 | nsecs_secs(nsecs) - returns whole secs portion given nsecs |
@@ -620,4 +620,4 @@ Various utility functions for use with perf trace: | |||
620 | 620 | ||
621 | SEE ALSO | 621 | SEE ALSO |
622 | -------- | 622 | -------- |
623 | linkperf:perf-trace[1] | 623 | linkperf:perf-script[1] |
diff --git a/tools/perf/Documentation/perf-trace.txt b/tools/perf/Documentation/perf-script.txt index 26aff6bf9e5..29ad94293cd 100644 --- a/tools/perf/Documentation/perf-trace.txt +++ b/tools/perf/Documentation/perf-script.txt | |||
@@ -1,71 +1,71 @@ | |||
1 | perf-trace(1) | 1 | perf-script(1) |
2 | ============= | 2 | ============= |
3 | 3 | ||
4 | NAME | 4 | NAME |
5 | ---- | 5 | ---- |
6 | perf-trace - Read perf.data (created by perf record) and display trace output | 6 | perf-script - Read perf.data (created by perf record) and display trace output |
7 | 7 | ||
8 | SYNOPSIS | 8 | SYNOPSIS |
9 | -------- | 9 | -------- |
10 | [verse] | 10 | [verse] |
11 | 'perf trace' [<options>] | 11 | 'perf script' [<options>] |
12 | 'perf trace' [<options>] record <script> [<record-options>] <command> | 12 | 'perf script' [<options>] record <script> [<record-options>] <command> |
13 | 'perf trace' [<options>] report <script> [script-args] | 13 | 'perf script' [<options>] report <script> [script-args] |
14 | 'perf trace' [<options>] <script> <required-script-args> [<record-options>] <command> | 14 | 'perf script' [<options>] <script> <required-script-args> [<record-options>] <command> |
15 | 'perf trace' [<options>] <top-script> [script-args] | 15 | 'perf script' [<options>] <top-script> [script-args] |
16 | 16 | ||
17 | DESCRIPTION | 17 | DESCRIPTION |
18 | ----------- | 18 | ----------- |
19 | This command reads the input file and displays the trace recorded. | 19 | This command reads the input file and displays the trace recorded. |
20 | 20 | ||
21 | There are several variants of perf trace: | 21 | There are several variants of perf script: |
22 | 22 | ||
23 | 'perf trace' to see a detailed trace of the workload that was | 23 | 'perf script' to see a detailed trace of the workload that was |
24 | recorded. | 24 | recorded. |
25 | 25 | ||
26 | You can also run a set of pre-canned scripts that aggregate and | 26 | You can also run a set of pre-canned scripts that aggregate and |
27 | summarize the raw trace data in various ways (the list of scripts is | 27 | summarize the raw trace data in various ways (the list of scripts is |
28 | available via 'perf trace -l'). The following variants allow you to | 28 | available via 'perf script -l'). The following variants allow you to |
29 | record and run those scripts: | 29 | record and run those scripts: |
30 | 30 | ||
31 | 'perf trace record <script> <command>' to record the events required | 31 | 'perf script record <script> <command>' to record the events required |
32 | for 'perf trace report'. <script> is the name displayed in the | 32 | for 'perf script report'. <script> is the name displayed in the |
33 | output of 'perf trace --list' i.e. the actual script name minus any | 33 | output of 'perf script --list' i.e. the actual script name minus any |
34 | language extension. If <command> is not specified, the events are | 34 | language extension. If <command> is not specified, the events are |
35 | recorded using the -a (system-wide) 'perf record' option. | 35 | recorded using the -a (system-wide) 'perf record' option. |
36 | 36 | ||
37 | 'perf trace report <script> [args]' to run and display the results | 37 | 'perf script report <script> [args]' to run and display the results |
38 | of <script>. <script> is the name displayed in the output of 'perf | 38 | of <script>. <script> is the name displayed in the output of 'perf |
39 | trace --list' i.e. the actual script name minus any language | 39 | trace --list' i.e. the actual script name minus any language |
40 | extension. The perf.data output from a previous run of 'perf trace | 40 | extension. The perf.data output from a previous run of 'perf script |
41 | record <script>' is used and should be present for this command to | 41 | record <script>' is used and should be present for this command to |
42 | succeed. [args] refers to the (mainly optional) args expected by | 42 | succeed. [args] refers to the (mainly optional) args expected by |
43 | the script. | 43 | the script. |
44 | 44 | ||
45 | 'perf trace <script> <required-script-args> <command>' to both | 45 | 'perf script <script> <required-script-args> <command>' to both |
46 | record the events required for <script> and to run the <script> | 46 | record the events required for <script> and to run the <script> |
47 | using 'live-mode' i.e. without writing anything to disk. <script> | 47 | using 'live-mode' i.e. without writing anything to disk. <script> |
48 | is the name displayed in the output of 'perf trace --list' i.e. the | 48 | is the name displayed in the output of 'perf script --list' i.e. the |
49 | actual script name minus any language extension. If <command> is | 49 | actual script name minus any language extension. If <command> is |
50 | not specified, the events are recorded using the -a (system-wide) | 50 | not specified, the events are recorded using the -a (system-wide) |
51 | 'perf record' option. If <script> has any required args, they | 51 | 'perf record' option. If <script> has any required args, they |
52 | should be specified before <command>. This mode doesn't allow for | 52 | should be specified before <command>. This mode doesn't allow for |
53 | optional script args to be specified; if optional script args are | 53 | optional script args to be specified; if optional script args are |
54 | desired, they can be specified using separate 'perf trace record' | 54 | desired, they can be specified using separate 'perf script record' |
55 | and 'perf trace report' commands, with the stdout of the record step | 55 | and 'perf script report' commands, with the stdout of the record step |
56 | piped to the stdin of the report script, using the '-o -' and '-i -' | 56 | piped to the stdin of the report script, using the '-o -' and '-i -' |
57 | options of the corresponding commands. | 57 | options of the corresponding commands. |
58 | 58 | ||
59 | 'perf trace <top-script>' to both record the events required for | 59 | 'perf script <top-script>' to both record the events required for |
60 | <top-script> and to run the <top-script> using 'live-mode' | 60 | <top-script> and to run the <top-script> using 'live-mode' |
61 | i.e. without writing anything to disk. <top-script> is the name | 61 | i.e. without writing anything to disk. <top-script> is the name |
62 | displayed in the output of 'perf trace --list' i.e. the actual | 62 | displayed in the output of 'perf script --list' i.e. the actual |
63 | script name minus any language extension; a <top-script> is defined | 63 | script name minus any language extension; a <top-script> is defined |
64 | as any script name ending with the string 'top'. | 64 | as any script name ending with the string 'top'. |
65 | 65 | ||
66 | [<record-options>] can be passed to the record steps of 'perf trace | 66 | [<record-options>] can be passed to the record steps of 'perf script |
67 | record' and 'live-mode' variants; this isn't possible however for | 67 | record' and 'live-mode' variants; this isn't possible however for |
68 | <top-script> 'live-mode' or 'perf trace report' variants. | 68 | <top-script> 'live-mode' or 'perf script report' variants. |
69 | 69 | ||
70 | See the 'SEE ALSO' section for links to language-specific | 70 | See the 'SEE ALSO' section for links to language-specific |
71 | information on how to write and run your own trace scripts. | 71 | information on how to write and run your own trace scripts. |
@@ -76,7 +76,7 @@ OPTIONS | |||
76 | Any command you can specify in a shell. | 76 | Any command you can specify in a shell. |
77 | 77 | ||
78 | -D:: | 78 | -D:: |
79 | --dump-raw-trace=:: | 79 | --dump-raw-script=:: |
80 | Display verbose dump of the trace data. | 80 | Display verbose dump of the trace data. |
81 | 81 | ||
82 | -L:: | 82 | -L:: |
@@ -95,7 +95,7 @@ OPTIONS | |||
95 | 95 | ||
96 | -g:: | 96 | -g:: |
97 | --gen-script=:: | 97 | --gen-script=:: |
98 | Generate perf-trace.[ext] starter script for given language, | 98 | Generate perf-script.[ext] starter script for given language, |
99 | using current perf.data. | 99 | using current perf.data. |
100 | 100 | ||
101 | -a:: | 101 | -a:: |
@@ -104,8 +104,15 @@ OPTIONS | |||
104 | normally don't - this option allows the latter to be run in | 104 | normally don't - this option allows the latter to be run in |
105 | system-wide mode. | 105 | system-wide mode. |
106 | 106 | ||
107 | -i:: | ||
108 | --input=:: | ||
109 | Input file name. | ||
110 | |||
111 | -d:: | ||
112 | --debug-mode:: | ||
113 | Do various checks like samples ordering and lost events. | ||
107 | 114 | ||
108 | SEE ALSO | 115 | SEE ALSO |
109 | -------- | 116 | -------- |
110 | linkperf:perf-record[1], linkperf:perf-trace-perl[1], | 117 | linkperf:perf-record[1], linkperf:perf-script-perl[1], |
111 | linkperf:perf-trace-python[1] | 118 | linkperf:perf-script-python[1] |
diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt index 4b3a2d46b43..b6da7affbbe 100644 --- a/tools/perf/Documentation/perf-stat.txt +++ b/tools/perf/Documentation/perf-stat.txt | |||
@@ -8,8 +8,8 @@ perf-stat - Run a command and gather performance counter statistics | |||
8 | SYNOPSIS | 8 | SYNOPSIS |
9 | -------- | 9 | -------- |
10 | [verse] | 10 | [verse] |
11 | 'perf stat' [-e <EVENT> | --event=EVENT] [-S] [-a] <command> | 11 | 'perf stat' [-e <EVENT> | --event=EVENT] [-a] <command> |
12 | 'perf stat' [-e <EVENT> | --event=EVENT] [-S] [-a] -- <command> [<options>] | 12 | 'perf stat' [-e <EVENT> | --event=EVENT] [-a] -- <command> [<options>] |
13 | 13 | ||
14 | DESCRIPTION | 14 | DESCRIPTION |
15 | ----------- | 15 | ----------- |
@@ -35,24 +35,54 @@ OPTIONS | |||
35 | child tasks do not inherit counters | 35 | child tasks do not inherit counters |
36 | -p:: | 36 | -p:: |
37 | --pid=<pid>:: | 37 | --pid=<pid>:: |
38 | stat events on existing pid | 38 | stat events on existing process id |
39 | |||
40 | -t:: | ||
41 | --tid=<tid>:: | ||
42 | stat events on existing thread id | ||
43 | |||
39 | 44 | ||
40 | -a:: | 45 | -a:: |
41 | system-wide collection | 46 | --all-cpus:: |
47 | system-wide collection from all CPUs | ||
42 | 48 | ||
43 | -c:: | 49 | -c:: |
44 | scale counter values | 50 | --scale:: |
51 | scale/normalize counter values | ||
52 | |||
53 | -r:: | ||
54 | --repeat=<n>:: | ||
55 | repeat command and print average + stddev (max: 100) | ||
45 | 56 | ||
46 | -B:: | 57 | -B:: |
58 | --big-num:: | ||
47 | print large numbers with thousands' separators according to locale | 59 | print large numbers with thousands' separators according to locale |
48 | 60 | ||
49 | -C:: | 61 | -C:: |
50 | --cpu=:: | 62 | --cpu=:: |
51 | Count only on the list of cpus provided. Multiple CPUs can be provided as a | 63 | Count only on the list of CPUs provided. Multiple CPUs can be provided as a |
52 | comma-sperated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. | 64 | comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. |
53 | In per-thread mode, this option is ignored. The -a option is still necessary | 65 | In per-thread mode, this option is ignored. The -a option is still necessary |
54 | to activate system-wide monitoring. Default is to count on all CPUs. | 66 | to activate system-wide monitoring. Default is to count on all CPUs. |
55 | 67 | ||
68 | -A:: | ||
69 | --no-aggr:: | ||
70 | Do not aggregate counts across all monitored CPUs in system-wide mode (-a). | ||
71 | This option is only valid in system-wide mode. | ||
72 | |||
73 | -n:: | ||
74 | --null:: | ||
75 | null run - don't start any counters | ||
76 | |||
77 | -v:: | ||
78 | --verbose:: | ||
79 | be more verbose (show counter open errors, etc) | ||
80 | |||
81 | -x SEP:: | ||
82 | --field-separator SEP:: | ||
83 | print counts using a CSV-style output to make it easy to import directly into | ||
84 | spreadsheets. Columns are separated by the string specified in SEP. | ||
85 | |||
56 | EXAMPLES | 86 | EXAMPLES |
57 | -------- | 87 | -------- |
58 | 88 | ||
diff --git a/tools/perf/Documentation/perf-test.txt b/tools/perf/Documentation/perf-test.txt index 1c4b5f5b7f7..2c3b462f64b 100644 --- a/tools/perf/Documentation/perf-test.txt +++ b/tools/perf/Documentation/perf-test.txt | |||
@@ -12,7 +12,7 @@ SYNOPSIS | |||
12 | 12 | ||
13 | DESCRIPTION | 13 | DESCRIPTION |
14 | ----------- | 14 | ----------- |
15 | This command does assorted sanity tests, initially thru linked routines but | 15 | This command does assorted sanity tests, initially through linked routines but |
16 | also will look for a directory with more tests in the form of scripts. | 16 | also will look for a directory with more tests in the form of scripts. |
17 | 17 | ||
18 | OPTIONS | 18 | OPTIONS |
diff --git a/tools/perf/Documentation/perf-timechart.txt b/tools/perf/Documentation/perf-timechart.txt index 4b1788355ec..d7b79e2ba2a 100644 --- a/tools/perf/Documentation/perf-timechart.txt +++ b/tools/perf/Documentation/perf-timechart.txt | |||
@@ -38,6 +38,8 @@ OPTIONS | |||
38 | --process:: | 38 | --process:: |
39 | Select the processes to display, by name or PID | 39 | Select the processes to display, by name or PID |
40 | 40 | ||
41 | --symfs=<directory>:: | ||
42 | Look for files with symbols relative to this directory. | ||
41 | 43 | ||
42 | SEE ALSO | 44 | SEE ALSO |
43 | -------- | 45 | -------- |
diff --git a/tools/perf/Documentation/perf-top.txt b/tools/perf/Documentation/perf-top.txt index 1f9687663f2..f6eb1cdafb7 100644 --- a/tools/perf/Documentation/perf-top.txt +++ b/tools/perf/Documentation/perf-top.txt | |||
@@ -12,7 +12,7 @@ SYNOPSIS | |||
12 | 12 | ||
13 | DESCRIPTION | 13 | DESCRIPTION |
14 | ----------- | 14 | ----------- |
15 | This command generates and displays a performance counter profile in realtime. | 15 | This command generates and displays a performance counter profile in real time. |
16 | 16 | ||
17 | 17 | ||
18 | OPTIONS | 18 | OPTIONS |
@@ -27,8 +27,8 @@ OPTIONS | |||
27 | 27 | ||
28 | -C <cpu-list>:: | 28 | -C <cpu-list>:: |
29 | --cpu=<cpu>:: | 29 | --cpu=<cpu>:: |
30 | Monitor only on the list of cpus provided. Multiple CPUs can be provided as a | 30 | Monitor only on the list of CPUs provided. Multiple CPUs can be provided as a |
31 | comma-sperated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. | 31 | comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. |
32 | Default is to monitor all CPUS. | 32 | Default is to monitor all CPUS. |
33 | 33 | ||
34 | -d <seconds>:: | 34 | -d <seconds>:: |
@@ -50,6 +50,10 @@ Default is to monitor all CPUS. | |||
50 | --count-filter=<count>:: | 50 | --count-filter=<count>:: |
51 | Only display functions with more events than this. | 51 | Only display functions with more events than this. |
52 | 52 | ||
53 | -g:: | ||
54 | --group:: | ||
55 | Put the counters into a counter group. | ||
56 | |||
53 | -F <freq>:: | 57 | -F <freq>:: |
54 | --freq=<freq>:: | 58 | --freq=<freq>:: |
55 | Profile at this frequency. | 59 | Profile at this frequency. |
@@ -68,7 +72,11 @@ Default is to monitor all CPUS. | |||
68 | 72 | ||
69 | -p <pid>:: | 73 | -p <pid>:: |
70 | --pid=<pid>:: | 74 | --pid=<pid>:: |
71 | Profile events on existing pid. | 75 | Profile events on existing Process ID. |
76 | |||
77 | -t <tid>:: | ||
78 | --tid=<tid>:: | ||
79 | Profile events on existing thread ID. | ||
72 | 80 | ||
73 | -r <priority>:: | 81 | -r <priority>:: |
74 | --realtime=<priority>:: | 82 | --realtime=<priority>:: |
@@ -78,6 +86,18 @@ Default is to monitor all CPUS. | |||
78 | --sym-annotate=<symbol>:: | 86 | --sym-annotate=<symbol>:: |
79 | Annotate this symbol. | 87 | Annotate this symbol. |
80 | 88 | ||
89 | -K:: | ||
90 | --hide_kernel_symbols:: | ||
91 | Hide kernel symbols. | ||
92 | |||
93 | -U:: | ||
94 | --hide_user_symbols:: | ||
95 | Hide user symbols. | ||
96 | |||
97 | -D:: | ||
98 | --dump-symtab:: | ||
99 | Dump the symbol table used for profiling. | ||
100 | |||
81 | -v:: | 101 | -v:: |
82 | --verbose:: | 102 | --verbose:: |
83 | Be more verbose (show counter open errors, etc). | 103 | Be more verbose (show counter open errors, etc). |
diff --git a/tools/perf/MANIFEST b/tools/perf/MANIFEST index 8c7fc0c8f0b..c12659d8cb2 100644 --- a/tools/perf/MANIFEST +++ b/tools/perf/MANIFEST | |||
@@ -7,6 +7,7 @@ include/linux/stringify.h | |||
7 | lib/rbtree.c | 7 | lib/rbtree.c |
8 | include/linux/swab.h | 8 | include/linux/swab.h |
9 | arch/*/include/asm/unistd*.h | 9 | arch/*/include/asm/unistd*.h |
10 | arch/*/lib/memcpy*.S | ||
10 | include/linux/poison.h | 11 | include/linux/poison.h |
11 | include/linux/magic.h | 12 | include/linux/magic.h |
12 | include/linux/hw_breakpoint.h | 13 | include/linux/hw_breakpoint.h |
diff --git a/tools/perf/Makefile b/tools/perf/Makefile index d1db0f676a4..1b9b13ee2a7 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile | |||
@@ -185,7 +185,10 @@ ifeq ($(ARCH),i386) | |||
185 | ARCH := x86 | 185 | ARCH := x86 |
186 | endif | 186 | endif |
187 | ifeq ($(ARCH),x86_64) | 187 | ifeq ($(ARCH),x86_64) |
188 | RAW_ARCH := x86_64 | ||
188 | ARCH := x86 | 189 | ARCH := x86 |
190 | ARCH_CFLAGS := -DARCH_X86_64 | ||
191 | ARCH_INCLUDE = ../../arch/x86/lib/memcpy_64.S | ||
189 | endif | 192 | endif |
190 | 193 | ||
191 | # CFLAGS and LDFLAGS are for the users to override from the command line. | 194 | # CFLAGS and LDFLAGS are for the users to override from the command line. |
@@ -375,6 +378,7 @@ LIB_H += util/include/linux/prefetch.h | |||
375 | LIB_H += util/include/linux/rbtree.h | 378 | LIB_H += util/include/linux/rbtree.h |
376 | LIB_H += util/include/linux/string.h | 379 | LIB_H += util/include/linux/string.h |
377 | LIB_H += util/include/linux/types.h | 380 | LIB_H += util/include/linux/types.h |
381 | LIB_H += util/include/linux/linkage.h | ||
378 | LIB_H += util/include/asm/asm-offsets.h | 382 | LIB_H += util/include/asm/asm-offsets.h |
379 | LIB_H += util/include/asm/bug.h | 383 | LIB_H += util/include/asm/bug.h |
380 | LIB_H += util/include/asm/byteorder.h | 384 | LIB_H += util/include/asm/byteorder.h |
@@ -383,6 +387,8 @@ LIB_H += util/include/asm/swab.h | |||
383 | LIB_H += util/include/asm/system.h | 387 | LIB_H += util/include/asm/system.h |
384 | LIB_H += util/include/asm/uaccess.h | 388 | LIB_H += util/include/asm/uaccess.h |
385 | LIB_H += util/include/dwarf-regs.h | 389 | LIB_H += util/include/dwarf-regs.h |
390 | LIB_H += util/include/asm/dwarf2.h | ||
391 | LIB_H += util/include/asm/cpufeature.h | ||
386 | LIB_H += perf.h | 392 | LIB_H += perf.h |
387 | LIB_H += util/cache.h | 393 | LIB_H += util/cache.h |
388 | LIB_H += util/callchain.h | 394 | LIB_H += util/callchain.h |
@@ -390,6 +396,7 @@ LIB_H += util/build-id.h | |||
390 | LIB_H += util/debug.h | 396 | LIB_H += util/debug.h |
391 | LIB_H += util/debugfs.h | 397 | LIB_H += util/debugfs.h |
392 | LIB_H += util/event.h | 398 | LIB_H += util/event.h |
399 | LIB_H += util/evsel.h | ||
393 | LIB_H += util/exec_cmd.h | 400 | LIB_H += util/exec_cmd.h |
394 | LIB_H += util/types.h | 401 | LIB_H += util/types.h |
395 | LIB_H += util/levenshtein.h | 402 | LIB_H += util/levenshtein.h |
@@ -398,6 +405,7 @@ LIB_H += util/parse-options.h | |||
398 | LIB_H += util/parse-events.h | 405 | LIB_H += util/parse-events.h |
399 | LIB_H += util/quote.h | 406 | LIB_H += util/quote.h |
400 | LIB_H += util/util.h | 407 | LIB_H += util/util.h |
408 | LIB_H += util/xyarray.h | ||
401 | LIB_H += util/header.h | 409 | LIB_H += util/header.h |
402 | LIB_H += util/help.h | 410 | LIB_H += util/help.h |
403 | LIB_H += util/session.h | 411 | LIB_H += util/session.h |
@@ -417,6 +425,7 @@ LIB_H += util/probe-finder.h | |||
417 | LIB_H += util/probe-event.h | 425 | LIB_H += util/probe-event.h |
418 | LIB_H += util/pstack.h | 426 | LIB_H += util/pstack.h |
419 | LIB_H += util/cpumap.h | 427 | LIB_H += util/cpumap.h |
428 | LIB_H += $(ARCH_INCLUDE) | ||
420 | 429 | ||
421 | LIB_OBJS += $(OUTPUT)util/abspath.o | 430 | LIB_OBJS += $(OUTPUT)util/abspath.o |
422 | LIB_OBJS += $(OUTPUT)util/alias.o | 431 | LIB_OBJS += $(OUTPUT)util/alias.o |
@@ -426,6 +435,7 @@ LIB_OBJS += $(OUTPUT)util/ctype.o | |||
426 | LIB_OBJS += $(OUTPUT)util/debugfs.o | 435 | LIB_OBJS += $(OUTPUT)util/debugfs.o |
427 | LIB_OBJS += $(OUTPUT)util/environment.o | 436 | LIB_OBJS += $(OUTPUT)util/environment.o |
428 | LIB_OBJS += $(OUTPUT)util/event.o | 437 | LIB_OBJS += $(OUTPUT)util/event.o |
438 | LIB_OBJS += $(OUTPUT)util/evsel.o | ||
429 | LIB_OBJS += $(OUTPUT)util/exec_cmd.o | 439 | LIB_OBJS += $(OUTPUT)util/exec_cmd.o |
430 | LIB_OBJS += $(OUTPUT)util/help.o | 440 | LIB_OBJS += $(OUTPUT)util/help.o |
431 | LIB_OBJS += $(OUTPUT)util/levenshtein.o | 441 | LIB_OBJS += $(OUTPUT)util/levenshtein.o |
@@ -463,6 +473,7 @@ LIB_OBJS += $(OUTPUT)util/sort.o | |||
463 | LIB_OBJS += $(OUTPUT)util/hist.o | 473 | LIB_OBJS += $(OUTPUT)util/hist.o |
464 | LIB_OBJS += $(OUTPUT)util/probe-event.o | 474 | LIB_OBJS += $(OUTPUT)util/probe-event.o |
465 | LIB_OBJS += $(OUTPUT)util/util.o | 475 | LIB_OBJS += $(OUTPUT)util/util.o |
476 | LIB_OBJS += $(OUTPUT)util/xyarray.o | ||
466 | LIB_OBJS += $(OUTPUT)util/cpumap.o | 477 | LIB_OBJS += $(OUTPUT)util/cpumap.o |
467 | 478 | ||
468 | BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o | 479 | BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o |
@@ -472,6 +483,9 @@ BUILTIN_OBJS += $(OUTPUT)builtin-bench.o | |||
472 | # Benchmark modules | 483 | # Benchmark modules |
473 | BUILTIN_OBJS += $(OUTPUT)bench/sched-messaging.o | 484 | BUILTIN_OBJS += $(OUTPUT)bench/sched-messaging.o |
474 | BUILTIN_OBJS += $(OUTPUT)bench/sched-pipe.o | 485 | BUILTIN_OBJS += $(OUTPUT)bench/sched-pipe.o |
486 | ifeq ($(RAW_ARCH),x86_64) | ||
487 | BUILTIN_OBJS += $(OUTPUT)bench/mem-memcpy-x86-64-asm.o | ||
488 | endif | ||
475 | BUILTIN_OBJS += $(OUTPUT)bench/mem-memcpy.o | 489 | BUILTIN_OBJS += $(OUTPUT)bench/mem-memcpy.o |
476 | 490 | ||
477 | BUILTIN_OBJS += $(OUTPUT)builtin-diff.o | 491 | BUILTIN_OBJS += $(OUTPUT)builtin-diff.o |
@@ -485,7 +499,7 @@ BUILTIN_OBJS += $(OUTPUT)builtin-report.o | |||
485 | BUILTIN_OBJS += $(OUTPUT)builtin-stat.o | 499 | BUILTIN_OBJS += $(OUTPUT)builtin-stat.o |
486 | BUILTIN_OBJS += $(OUTPUT)builtin-timechart.o | 500 | BUILTIN_OBJS += $(OUTPUT)builtin-timechart.o |
487 | BUILTIN_OBJS += $(OUTPUT)builtin-top.o | 501 | BUILTIN_OBJS += $(OUTPUT)builtin-top.o |
488 | BUILTIN_OBJS += $(OUTPUT)builtin-trace.o | 502 | BUILTIN_OBJS += $(OUTPUT)builtin-script.o |
489 | BUILTIN_OBJS += $(OUTPUT)builtin-probe.o | 503 | BUILTIN_OBJS += $(OUTPUT)builtin-probe.o |
490 | BUILTIN_OBJS += $(OUTPUT)builtin-kmem.o | 504 | BUILTIN_OBJS += $(OUTPUT)builtin-kmem.o |
491 | BUILTIN_OBJS += $(OUTPUT)builtin-lock.o | 505 | BUILTIN_OBJS += $(OUTPUT)builtin-lock.o |
@@ -507,7 +521,7 @@ PERFLIBS = $(LIB_FILE) | |||
507 | -include config.mak | 521 | -include config.mak |
508 | 522 | ||
509 | ifndef NO_DWARF | 523 | ifndef NO_DWARF |
510 | FLAGS_DWARF=$(ALL_CFLAGS) -I/usr/include/elfutils -ldw -lelf $(ALL_LDFLAGS) $(EXTLIBS) | 524 | FLAGS_DWARF=$(ALL_CFLAGS) -ldw -lelf $(ALL_LDFLAGS) $(EXTLIBS) |
511 | ifneq ($(call try-cc,$(SOURCE_DWARF),$(FLAGS_DWARF)),y) | 525 | ifneq ($(call try-cc,$(SOURCE_DWARF),$(FLAGS_DWARF)),y) |
512 | msg := $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev); | 526 | msg := $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev); |
513 | NO_DWARF := 1 | 527 | NO_DWARF := 1 |
@@ -554,7 +568,7 @@ ifndef NO_DWARF | |||
554 | ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined) | 568 | ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined) |
555 | msg := $(warning DWARF register mappings have not been defined for architecture $(ARCH), DWARF support disabled); | 569 | msg := $(warning DWARF register mappings have not been defined for architecture $(ARCH), DWARF support disabled); |
556 | else | 570 | else |
557 | BASIC_CFLAGS += -I/usr/include/elfutils -DDWARF_SUPPORT | 571 | BASIC_CFLAGS += -DDWARF_SUPPORT |
558 | EXTLIBS += -lelf -ldw | 572 | EXTLIBS += -lelf -ldw |
559 | LIB_OBJS += $(OUTPUT)util/probe-finder.o | 573 | LIB_OBJS += $(OUTPUT)util/probe-finder.o |
560 | endif # PERF_HAVE_DWARF_REGS | 574 | endif # PERF_HAVE_DWARF_REGS |
@@ -891,13 +905,14 @@ prefix_SQ = $(subst ','\'',$(prefix)) | |||
891 | SHELL_PATH_SQ = $(subst ','\'',$(SHELL_PATH)) | 905 | SHELL_PATH_SQ = $(subst ','\'',$(SHELL_PATH)) |
892 | PERL_PATH_SQ = $(subst ','\'',$(PERL_PATH)) | 906 | PERL_PATH_SQ = $(subst ','\'',$(PERL_PATH)) |
893 | 907 | ||
894 | LIBS = $(PERFLIBS) $(EXTLIBS) | 908 | LIBS = -Wl,--whole-archive $(PERFLIBS) -Wl,--no-whole-archive $(EXTLIBS) |
895 | 909 | ||
896 | BASIC_CFLAGS += -DSHA1_HEADER='$(SHA1_HEADER_SQ)' \ | 910 | BASIC_CFLAGS += -DSHA1_HEADER='$(SHA1_HEADER_SQ)' \ |
897 | $(COMPAT_CFLAGS) | 911 | $(COMPAT_CFLAGS) |
898 | LIB_OBJS += $(COMPAT_OBJS) | 912 | LIB_OBJS += $(COMPAT_OBJS) |
899 | 913 | ||
900 | ALL_CFLAGS += $(BASIC_CFLAGS) | 914 | ALL_CFLAGS += $(BASIC_CFLAGS) |
915 | ALL_CFLAGS += $(ARCH_CFLAGS) | ||
901 | ALL_LDFLAGS += $(BASIC_LDFLAGS) | 916 | ALL_LDFLAGS += $(BASIC_LDFLAGS) |
902 | 917 | ||
903 | export TAR INSTALL DESTDIR SHELL_PATH | 918 | export TAR INSTALL DESTDIR SHELL_PATH |
diff --git a/tools/perf/bench/mem-memcpy-arch.h b/tools/perf/bench/mem-memcpy-arch.h new file mode 100644 index 00000000000..a72e36cb539 --- /dev/null +++ b/tools/perf/bench/mem-memcpy-arch.h | |||
@@ -0,0 +1,12 @@ | |||
1 | |||
2 | #ifdef ARCH_X86_64 | ||
3 | |||
4 | #define MEMCPY_FN(fn, name, desc) \ | ||
5 | extern void *fn(void *, const void *, size_t); | ||
6 | |||
7 | #include "mem-memcpy-x86-64-asm-def.h" | ||
8 | |||
9 | #undef MEMCPY_FN | ||
10 | |||
11 | #endif | ||
12 | |||
diff --git a/tools/perf/bench/mem-memcpy-x86-64-asm-def.h b/tools/perf/bench/mem-memcpy-x86-64-asm-def.h new file mode 100644 index 00000000000..d588b87696f --- /dev/null +++ b/tools/perf/bench/mem-memcpy-x86-64-asm-def.h | |||
@@ -0,0 +1,4 @@ | |||
1 | |||
2 | MEMCPY_FN(__memcpy, | ||
3 | "x86-64-unrolled", | ||
4 | "unrolled memcpy() in arch/x86/lib/memcpy_64.S") | ||
diff --git a/tools/perf/bench/mem-memcpy-x86-64-asm.S b/tools/perf/bench/mem-memcpy-x86-64-asm.S new file mode 100644 index 00000000000..a57b66e853c --- /dev/null +++ b/tools/perf/bench/mem-memcpy-x86-64-asm.S | |||
@@ -0,0 +1,2 @@ | |||
1 | |||
2 | #include "../../../arch/x86/lib/memcpy_64.S" | ||
diff --git a/tools/perf/bench/mem-memcpy.c b/tools/perf/bench/mem-memcpy.c index 38dae746514..db82021f4b9 100644 --- a/tools/perf/bench/mem-memcpy.c +++ b/tools/perf/bench/mem-memcpy.c | |||
@@ -12,6 +12,7 @@ | |||
12 | #include "../util/parse-options.h" | 12 | #include "../util/parse-options.h" |
13 | #include "../util/header.h" | 13 | #include "../util/header.h" |
14 | #include "bench.h" | 14 | #include "bench.h" |
15 | #include "mem-memcpy-arch.h" | ||
15 | 16 | ||
16 | #include <stdio.h> | 17 | #include <stdio.h> |
17 | #include <stdlib.h> | 18 | #include <stdlib.h> |
@@ -23,8 +24,10 @@ | |||
23 | 24 | ||
24 | static const char *length_str = "1MB"; | 25 | static const char *length_str = "1MB"; |
25 | static const char *routine = "default"; | 26 | static const char *routine = "default"; |
26 | static bool use_clock = false; | 27 | static bool use_clock; |
27 | static int clock_fd; | 28 | static int clock_fd; |
29 | static bool only_prefault; | ||
30 | static bool no_prefault; | ||
28 | 31 | ||
29 | static const struct option options[] = { | 32 | static const struct option options[] = { |
30 | OPT_STRING('l', "length", &length_str, "1MB", | 33 | OPT_STRING('l', "length", &length_str, "1MB", |
@@ -34,19 +37,33 @@ static const struct option options[] = { | |||
34 | "Specify routine to copy"), | 37 | "Specify routine to copy"), |
35 | OPT_BOOLEAN('c', "clock", &use_clock, | 38 | OPT_BOOLEAN('c', "clock", &use_clock, |
36 | "Use CPU clock for measuring"), | 39 | "Use CPU clock for measuring"), |
40 | OPT_BOOLEAN('o', "only-prefault", &only_prefault, | ||
41 | "Show only the result with page faults before memcpy()"), | ||
42 | OPT_BOOLEAN('n', "no-prefault", &no_prefault, | ||
43 | "Show only the result without page faults before memcpy()"), | ||
37 | OPT_END() | 44 | OPT_END() |
38 | }; | 45 | }; |
39 | 46 | ||
47 | typedef void *(*memcpy_t)(void *, const void *, size_t); | ||
48 | |||
40 | struct routine { | 49 | struct routine { |
41 | const char *name; | 50 | const char *name; |
42 | const char *desc; | 51 | const char *desc; |
43 | void * (*fn)(void *dst, const void *src, size_t len); | 52 | memcpy_t fn; |
44 | }; | 53 | }; |
45 | 54 | ||
46 | struct routine routines[] = { | 55 | struct routine routines[] = { |
47 | { "default", | 56 | { "default", |
48 | "Default memcpy() provided by glibc", | 57 | "Default memcpy() provided by glibc", |
49 | memcpy }, | 58 | memcpy }, |
59 | #ifdef ARCH_X86_64 | ||
60 | |||
61 | #define MEMCPY_FN(fn, name, desc) { name, desc, fn }, | ||
62 | #include "mem-memcpy-x86-64-asm-def.h" | ||
63 | #undef MEMCPY_FN | ||
64 | |||
65 | #endif | ||
66 | |||
50 | { NULL, | 67 | { NULL, |
51 | NULL, | 68 | NULL, |
52 | NULL } | 69 | NULL } |
@@ -89,29 +106,98 @@ static double timeval2double(struct timeval *ts) | |||
89 | (double)ts->tv_usec / (double)1000000; | 106 | (double)ts->tv_usec / (double)1000000; |
90 | } | 107 | } |
91 | 108 | ||
109 | static void alloc_mem(void **dst, void **src, size_t length) | ||
110 | { | ||
111 | *dst = zalloc(length); | ||
112 | if (!dst) | ||
113 | die("memory allocation failed - maybe length is too large?\n"); | ||
114 | |||
115 | *src = zalloc(length); | ||
116 | if (!src) | ||
117 | die("memory allocation failed - maybe length is too large?\n"); | ||
118 | } | ||
119 | |||
120 | static u64 do_memcpy_clock(memcpy_t fn, size_t len, bool prefault) | ||
121 | { | ||
122 | u64 clock_start = 0ULL, clock_end = 0ULL; | ||
123 | void *src = NULL, *dst = NULL; | ||
124 | |||
125 | alloc_mem(&src, &dst, len); | ||
126 | |||
127 | if (prefault) | ||
128 | fn(dst, src, len); | ||
129 | |||
130 | clock_start = get_clock(); | ||
131 | fn(dst, src, len); | ||
132 | clock_end = get_clock(); | ||
133 | |||
134 | free(src); | ||
135 | free(dst); | ||
136 | return clock_end - clock_start; | ||
137 | } | ||
138 | |||
139 | static double do_memcpy_gettimeofday(memcpy_t fn, size_t len, bool prefault) | ||
140 | { | ||
141 | struct timeval tv_start, tv_end, tv_diff; | ||
142 | void *src = NULL, *dst = NULL; | ||
143 | |||
144 | alloc_mem(&src, &dst, len); | ||
145 | |||
146 | if (prefault) | ||
147 | fn(dst, src, len); | ||
148 | |||
149 | BUG_ON(gettimeofday(&tv_start, NULL)); | ||
150 | fn(dst, src, len); | ||
151 | BUG_ON(gettimeofday(&tv_end, NULL)); | ||
152 | |||
153 | timersub(&tv_end, &tv_start, &tv_diff); | ||
154 | |||
155 | free(src); | ||
156 | free(dst); | ||
157 | return (double)((double)len / timeval2double(&tv_diff)); | ||
158 | } | ||
159 | |||
160 | #define pf (no_prefault ? 0 : 1) | ||
161 | |||
162 | #define print_bps(x) do { \ | ||
163 | if (x < K) \ | ||
164 | printf(" %14lf B/Sec", x); \ | ||
165 | else if (x < K * K) \ | ||
166 | printf(" %14lfd KB/Sec", x / K); \ | ||
167 | else if (x < K * K * K) \ | ||
168 | printf(" %14lf MB/Sec", x / K / K); \ | ||
169 | else \ | ||
170 | printf(" %14lf GB/Sec", x / K / K / K); \ | ||
171 | } while (0) | ||
172 | |||
92 | int bench_mem_memcpy(int argc, const char **argv, | 173 | int bench_mem_memcpy(int argc, const char **argv, |
93 | const char *prefix __used) | 174 | const char *prefix __used) |
94 | { | 175 | { |
95 | int i; | 176 | int i; |
96 | void *dst, *src; | 177 | size_t len; |
97 | size_t length; | 178 | double result_bps[2]; |
98 | double bps = 0.0; | 179 | u64 result_clock[2]; |
99 | struct timeval tv_start, tv_end, tv_diff; | ||
100 | u64 clock_start, clock_end, clock_diff; | ||
101 | 180 | ||
102 | clock_start = clock_end = clock_diff = 0ULL; | ||
103 | argc = parse_options(argc, argv, options, | 181 | argc = parse_options(argc, argv, options, |
104 | bench_mem_memcpy_usage, 0); | 182 | bench_mem_memcpy_usage, 0); |
105 | 183 | ||
106 | tv_diff.tv_sec = 0; | 184 | if (use_clock) |
107 | tv_diff.tv_usec = 0; | 185 | init_clock(); |
108 | length = (size_t)perf_atoll((char *)length_str); | 186 | |
187 | len = (size_t)perf_atoll((char *)length_str); | ||
109 | 188 | ||
110 | if ((s64)length <= 0) { | 189 | result_clock[0] = result_clock[1] = 0ULL; |
190 | result_bps[0] = result_bps[1] = 0.0; | ||
191 | |||
192 | if ((s64)len <= 0) { | ||
111 | fprintf(stderr, "Invalid length:%s\n", length_str); | 193 | fprintf(stderr, "Invalid length:%s\n", length_str); |
112 | return 1; | 194 | return 1; |
113 | } | 195 | } |
114 | 196 | ||
197 | /* same to without specifying either of prefault and no-prefault */ | ||
198 | if (only_prefault && no_prefault) | ||
199 | only_prefault = no_prefault = false; | ||
200 | |||
115 | for (i = 0; routines[i].name; i++) { | 201 | for (i = 0; routines[i].name; i++) { |
116 | if (!strcmp(routines[i].name, routine)) | 202 | if (!strcmp(routines[i].name, routine)) |
117 | break; | 203 | break; |
@@ -126,61 +212,80 @@ int bench_mem_memcpy(int argc, const char **argv, | |||
126 | return 1; | 212 | return 1; |
127 | } | 213 | } |
128 | 214 | ||
129 | dst = zalloc(length); | 215 | if (bench_format == BENCH_FORMAT_DEFAULT) |
130 | if (!dst) | 216 | printf("# Copying %s Bytes ...\n\n", length_str); |
131 | die("memory allocation failed - maybe length is too large?\n"); | ||
132 | |||
133 | src = zalloc(length); | ||
134 | if (!src) | ||
135 | die("memory allocation failed - maybe length is too large?\n"); | ||
136 | |||
137 | if (bench_format == BENCH_FORMAT_DEFAULT) { | ||
138 | printf("# Copying %s Bytes from %p to %p ...\n\n", | ||
139 | length_str, src, dst); | ||
140 | } | ||
141 | |||
142 | if (use_clock) { | ||
143 | init_clock(); | ||
144 | clock_start = get_clock(); | ||
145 | } else { | ||
146 | BUG_ON(gettimeofday(&tv_start, NULL)); | ||
147 | } | ||
148 | |||
149 | routines[i].fn(dst, src, length); | ||
150 | 217 | ||
151 | if (use_clock) { | 218 | if (!only_prefault && !no_prefault) { |
152 | clock_end = get_clock(); | 219 | /* show both of results */ |
153 | clock_diff = clock_end - clock_start; | 220 | if (use_clock) { |
221 | result_clock[0] = | ||
222 | do_memcpy_clock(routines[i].fn, len, false); | ||
223 | result_clock[1] = | ||
224 | do_memcpy_clock(routines[i].fn, len, true); | ||
225 | } else { | ||
226 | result_bps[0] = | ||
227 | do_memcpy_gettimeofday(routines[i].fn, | ||
228 | len, false); | ||
229 | result_bps[1] = | ||
230 | do_memcpy_gettimeofday(routines[i].fn, | ||
231 | len, true); | ||
232 | } | ||
154 | } else { | 233 | } else { |
155 | BUG_ON(gettimeofday(&tv_end, NULL)); | 234 | if (use_clock) { |
156 | timersub(&tv_end, &tv_start, &tv_diff); | 235 | result_clock[pf] = |
157 | bps = (double)((double)length / timeval2double(&tv_diff)); | 236 | do_memcpy_clock(routines[i].fn, |
237 | len, only_prefault); | ||
238 | } else { | ||
239 | result_bps[pf] = | ||
240 | do_memcpy_gettimeofday(routines[i].fn, | ||
241 | len, only_prefault); | ||
242 | } | ||
158 | } | 243 | } |
159 | 244 | ||
160 | switch (bench_format) { | 245 | switch (bench_format) { |
161 | case BENCH_FORMAT_DEFAULT: | 246 | case BENCH_FORMAT_DEFAULT: |
162 | if (use_clock) { | 247 | if (!only_prefault && !no_prefault) { |
163 | printf(" %14lf Clock/Byte\n", | 248 | if (use_clock) { |
164 | (double)clock_diff / (double)length); | 249 | printf(" %14lf Clock/Byte\n", |
165 | } else { | 250 | (double)result_clock[0] |
166 | if (bps < K) | 251 | / (double)len); |
167 | printf(" %14lf B/Sec\n", bps); | 252 | printf(" %14lf Clock/Byte (with prefault)\n", |
168 | else if (bps < K * K) | 253 | (double)result_clock[1] |
169 | printf(" %14lfd KB/Sec\n", bps / 1024); | 254 | / (double)len); |
170 | else if (bps < K * K * K) | 255 | } else { |
171 | printf(" %14lf MB/Sec\n", bps / 1024 / 1024); | 256 | print_bps(result_bps[0]); |
172 | else { | 257 | printf("\n"); |
173 | printf(" %14lf GB/Sec\n", | 258 | print_bps(result_bps[1]); |
174 | bps / 1024 / 1024 / 1024); | 259 | printf(" (with prefault)\n"); |
175 | } | 260 | } |
261 | } else { | ||
262 | if (use_clock) { | ||
263 | printf(" %14lf Clock/Byte", | ||
264 | (double)result_clock[pf] | ||
265 | / (double)len); | ||
266 | } else | ||
267 | print_bps(result_bps[pf]); | ||
268 | |||
269 | printf("%s\n", only_prefault ? " (with prefault)" : ""); | ||
176 | } | 270 | } |
177 | break; | 271 | break; |
178 | case BENCH_FORMAT_SIMPLE: | 272 | case BENCH_FORMAT_SIMPLE: |
179 | if (use_clock) { | 273 | if (!only_prefault && !no_prefault) { |
180 | printf("%14lf\n", | 274 | if (use_clock) { |
181 | (double)clock_diff / (double)length); | 275 | printf("%lf %lf\n", |
182 | } else | 276 | (double)result_clock[0] / (double)len, |
183 | printf("%lf\n", bps); | 277 | (double)result_clock[1] / (double)len); |
278 | } else { | ||
279 | printf("%lf %lf\n", | ||
280 | result_bps[0], result_bps[1]); | ||
281 | } | ||
282 | } else { | ||
283 | if (use_clock) { | ||
284 | printf("%lf\n", (double)result_clock[pf] | ||
285 | / (double)len); | ||
286 | } else | ||
287 | printf("%lf\n", result_bps[pf]); | ||
288 | } | ||
184 | break; | 289 | break; |
185 | default: | 290 | default: |
186 | /* reaching this means there's some disaster: */ | 291 | /* reaching this means there's some disaster: */ |
diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index 6d5604d8df9..c056cdc0691 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c | |||
@@ -58,12 +58,12 @@ static int hists__add_entry(struct hists *self, struct addr_location *al) | |||
58 | return hist_entry__inc_addr_samples(he, al->addr); | 58 | return hist_entry__inc_addr_samples(he, al->addr); |
59 | } | 59 | } |
60 | 60 | ||
61 | static int process_sample_event(event_t *event, struct perf_session *session) | 61 | static int process_sample_event(event_t *event, struct sample_data *sample, |
62 | struct perf_session *session) | ||
62 | { | 63 | { |
63 | struct addr_location al; | 64 | struct addr_location al; |
64 | struct sample_data data; | ||
65 | 65 | ||
66 | if (event__preprocess_sample(event, session, &al, &data, NULL) < 0) { | 66 | if (event__preprocess_sample(event, session, &al, sample, NULL) < 0) { |
67 | pr_warning("problem processing %d event, skipping it.\n", | 67 | pr_warning("problem processing %d event, skipping it.\n", |
68 | event->header.type); | 68 | event->header.type); |
69 | return -1; | 69 | return -1; |
@@ -375,6 +375,8 @@ static struct perf_event_ops event_ops = { | |||
375 | .mmap = event__process_mmap, | 375 | .mmap = event__process_mmap, |
376 | .comm = event__process_comm, | 376 | .comm = event__process_comm, |
377 | .fork = event__process_task, | 377 | .fork = event__process_task, |
378 | .ordered_samples = true, | ||
379 | .ordering_requires_timestamps = true, | ||
378 | }; | 380 | }; |
379 | 381 | ||
380 | static int __cmd_annotate(void) | 382 | static int __cmd_annotate(void) |
@@ -382,7 +384,7 @@ static int __cmd_annotate(void) | |||
382 | int ret; | 384 | int ret; |
383 | struct perf_session *session; | 385 | struct perf_session *session; |
384 | 386 | ||
385 | session = perf_session__new(input_name, O_RDONLY, force, false); | 387 | session = perf_session__new(input_name, O_RDONLY, force, false, &event_ops); |
386 | if (session == NULL) | 388 | if (session == NULL) |
387 | return -ENOMEM; | 389 | return -ENOMEM; |
388 | 390 | ||
diff --git a/tools/perf/builtin-buildid-list.c b/tools/perf/builtin-buildid-list.c index c49837de7d3..5af32ae9031 100644 --- a/tools/perf/builtin-buildid-list.c +++ b/tools/perf/builtin-buildid-list.c | |||
@@ -38,7 +38,8 @@ static int __cmd_buildid_list(void) | |||
38 | { | 38 | { |
39 | struct perf_session *session; | 39 | struct perf_session *session; |
40 | 40 | ||
41 | session = perf_session__new(input_name, O_RDONLY, force, false); | 41 | session = perf_session__new(input_name, O_RDONLY, force, false, |
42 | &build_id__mark_dso_hit_ops); | ||
42 | if (session == NULL) | 43 | if (session == NULL) |
43 | return -1; | 44 | return -1; |
44 | 45 | ||
diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c index fca1d440291..3153e492dbc 100644 --- a/tools/perf/builtin-diff.c +++ b/tools/perf/builtin-diff.c | |||
@@ -30,12 +30,13 @@ static int hists__add_entry(struct hists *self, | |||
30 | return -ENOMEM; | 30 | return -ENOMEM; |
31 | } | 31 | } |
32 | 32 | ||
33 | static int diff__process_sample_event(event_t *event, struct perf_session *session) | 33 | static int diff__process_sample_event(event_t *event, |
34 | struct sample_data *sample, | ||
35 | struct perf_session *session) | ||
34 | { | 36 | { |
35 | struct addr_location al; | 37 | struct addr_location al; |
36 | struct sample_data data = { .period = 1, }; | ||
37 | 38 | ||
38 | if (event__preprocess_sample(event, session, &al, &data, NULL) < 0) { | 39 | if (event__preprocess_sample(event, session, &al, sample, NULL) < 0) { |
39 | pr_warning("problem processing %d event, skipping it.\n", | 40 | pr_warning("problem processing %d event, skipping it.\n", |
40 | event->header.type); | 41 | event->header.type); |
41 | return -1; | 42 | return -1; |
@@ -44,12 +45,12 @@ static int diff__process_sample_event(event_t *event, struct perf_session *sessi | |||
44 | if (al.filtered || al.sym == NULL) | 45 | if (al.filtered || al.sym == NULL) |
45 | return 0; | 46 | return 0; |
46 | 47 | ||
47 | if (hists__add_entry(&session->hists, &al, data.period)) { | 48 | if (hists__add_entry(&session->hists, &al, sample->period)) { |
48 | pr_warning("problem incrementing symbol period, skipping event\n"); | 49 | pr_warning("problem incrementing symbol period, skipping event\n"); |
49 | return -1; | 50 | return -1; |
50 | } | 51 | } |
51 | 52 | ||
52 | session->hists.stats.total_period += data.period; | 53 | session->hists.stats.total_period += sample->period; |
53 | return 0; | 54 | return 0; |
54 | } | 55 | } |
55 | 56 | ||
@@ -60,6 +61,8 @@ static struct perf_event_ops event_ops = { | |||
60 | .exit = event__process_task, | 61 | .exit = event__process_task, |
61 | .fork = event__process_task, | 62 | .fork = event__process_task, |
62 | .lost = event__process_lost, | 63 | .lost = event__process_lost, |
64 | .ordered_samples = true, | ||
65 | .ordering_requires_timestamps = true, | ||
63 | }; | 66 | }; |
64 | 67 | ||
65 | static void perf_session__insert_hist_entry_by_name(struct rb_root *root, | 68 | static void perf_session__insert_hist_entry_by_name(struct rb_root *root, |
@@ -141,8 +144,8 @@ static int __cmd_diff(void) | |||
141 | int ret, i; | 144 | int ret, i; |
142 | struct perf_session *session[2]; | 145 | struct perf_session *session[2]; |
143 | 146 | ||
144 | session[0] = perf_session__new(input_old, O_RDONLY, force, false); | 147 | session[0] = perf_session__new(input_old, O_RDONLY, force, false, &event_ops); |
145 | session[1] = perf_session__new(input_new, O_RDONLY, force, false); | 148 | session[1] = perf_session__new(input_new, O_RDONLY, force, false, &event_ops); |
146 | if (session[0] == NULL || session[1] == NULL) | 149 | if (session[0] == NULL || session[1] == NULL) |
147 | return -ENOMEM; | 150 | return -ENOMEM; |
148 | 151 | ||
@@ -173,7 +176,7 @@ static const char * const diff_usage[] = { | |||
173 | static const struct option options[] = { | 176 | static const struct option options[] = { |
174 | OPT_INCR('v', "verbose", &verbose, | 177 | OPT_INCR('v', "verbose", &verbose, |
175 | "be more verbose (show symbol address, etc)"), | 178 | "be more verbose (show symbol address, etc)"), |
176 | OPT_BOOLEAN('m', "displacement", &show_displacement, | 179 | OPT_BOOLEAN('M', "displacement", &show_displacement, |
177 | "Show position displacement relative to baseline"), | 180 | "Show position displacement relative to baseline"), |
178 | OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace, | 181 | OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace, |
179 | "dump raw trace in ASCII"), | 182 | "dump raw trace in ASCII"), |
@@ -191,6 +194,8 @@ static const struct option options[] = { | |||
191 | OPT_STRING('t', "field-separator", &symbol_conf.field_sep, "separator", | 194 | OPT_STRING('t', "field-separator", &symbol_conf.field_sep, "separator", |
192 | "separator for columns, no spaces will be added between " | 195 | "separator for columns, no spaces will be added between " |
193 | "columns '.' is reserved."), | 196 | "columns '.' is reserved."), |
197 | OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory", | ||
198 | "Look for files with symbols relative to this directory"), | ||
194 | OPT_END() | 199 | OPT_END() |
195 | }; | 200 | }; |
196 | 201 | ||
diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c index 8e3e47b064c..0c78ffa7bf6 100644 --- a/tools/perf/builtin-inject.c +++ b/tools/perf/builtin-inject.c | |||
@@ -16,8 +16,8 @@ | |||
16 | static char const *input_name = "-"; | 16 | static char const *input_name = "-"; |
17 | static bool inject_build_ids; | 17 | static bool inject_build_ids; |
18 | 18 | ||
19 | static int event__repipe(event_t *event __used, | 19 | static int event__repipe_synth(event_t *event, |
20 | struct perf_session *session __used) | 20 | struct perf_session *session __used) |
21 | { | 21 | { |
22 | uint32_t size; | 22 | uint32_t size; |
23 | void *buf = event; | 23 | void *buf = event; |
@@ -36,22 +36,30 @@ static int event__repipe(event_t *event __used, | |||
36 | return 0; | 36 | return 0; |
37 | } | 37 | } |
38 | 38 | ||
39 | static int event__repipe_mmap(event_t *self, struct perf_session *session) | 39 | static int event__repipe(event_t *event, struct sample_data *sample __used, |
40 | struct perf_session *session) | ||
41 | { | ||
42 | return event__repipe_synth(event, session); | ||
43 | } | ||
44 | |||
45 | static int event__repipe_mmap(event_t *self, struct sample_data *sample, | ||
46 | struct perf_session *session) | ||
40 | { | 47 | { |
41 | int err; | 48 | int err; |
42 | 49 | ||
43 | err = event__process_mmap(self, session); | 50 | err = event__process_mmap(self, sample, session); |
44 | event__repipe(self, session); | 51 | event__repipe(self, sample, session); |
45 | 52 | ||
46 | return err; | 53 | return err; |
47 | } | 54 | } |
48 | 55 | ||
49 | static int event__repipe_task(event_t *self, struct perf_session *session) | 56 | static int event__repipe_task(event_t *self, struct sample_data *sample, |
57 | struct perf_session *session) | ||
50 | { | 58 | { |
51 | int err; | 59 | int err; |
52 | 60 | ||
53 | err = event__process_task(self, session); | 61 | err = event__process_task(self, sample, session); |
54 | event__repipe(self, session); | 62 | event__repipe(self, sample, session); |
55 | 63 | ||
56 | return err; | 64 | return err; |
57 | } | 65 | } |
@@ -61,7 +69,7 @@ static int event__repipe_tracing_data(event_t *self, | |||
61 | { | 69 | { |
62 | int err; | 70 | int err; |
63 | 71 | ||
64 | event__repipe(self, session); | 72 | event__repipe_synth(self, session); |
65 | err = event__process_tracing_data(self, session); | 73 | err = event__process_tracing_data(self, session); |
66 | 74 | ||
67 | return err; | 75 | return err; |
@@ -111,7 +119,8 @@ static int dso__inject_build_id(struct dso *self, struct perf_session *session) | |||
111 | return 0; | 119 | return 0; |
112 | } | 120 | } |
113 | 121 | ||
114 | static int event__inject_buildid(event_t *event, struct perf_session *session) | 122 | static int event__inject_buildid(event_t *event, struct sample_data *sample, |
123 | struct perf_session *session) | ||
115 | { | 124 | { |
116 | struct addr_location al; | 125 | struct addr_location al; |
117 | struct thread *thread; | 126 | struct thread *thread; |
@@ -146,7 +155,7 @@ static int event__inject_buildid(event_t *event, struct perf_session *session) | |||
146 | } | 155 | } |
147 | 156 | ||
148 | repipe: | 157 | repipe: |
149 | event__repipe(event, session); | 158 | event__repipe(event, sample, session); |
150 | return 0; | 159 | return 0; |
151 | } | 160 | } |
152 | 161 | ||
@@ -160,10 +169,10 @@ struct perf_event_ops inject_ops = { | |||
160 | .read = event__repipe, | 169 | .read = event__repipe, |
161 | .throttle = event__repipe, | 170 | .throttle = event__repipe, |
162 | .unthrottle = event__repipe, | 171 | .unthrottle = event__repipe, |
163 | .attr = event__repipe, | 172 | .attr = event__repipe_synth, |
164 | .event_type = event__repipe, | 173 | .event_type = event__repipe_synth, |
165 | .tracing_data = event__repipe, | 174 | .tracing_data = event__repipe_synth, |
166 | .build_id = event__repipe, | 175 | .build_id = event__repipe_synth, |
167 | }; | 176 | }; |
168 | 177 | ||
169 | extern volatile int session_done; | 178 | extern volatile int session_done; |
@@ -187,7 +196,7 @@ static int __cmd_inject(void) | |||
187 | inject_ops.tracing_data = event__repipe_tracing_data; | 196 | inject_ops.tracing_data = event__repipe_tracing_data; |
188 | } | 197 | } |
189 | 198 | ||
190 | session = perf_session__new(input_name, O_RDONLY, false, true); | 199 | session = perf_session__new(input_name, O_RDONLY, false, true, &inject_ops); |
191 | if (session == NULL) | 200 | if (session == NULL) |
192 | return -ENOMEM; | 201 | return -ENOMEM; |
193 | 202 | ||
diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c index 31f60a2535e..def7ddc2fd4 100644 --- a/tools/perf/builtin-kmem.c +++ b/tools/perf/builtin-kmem.c | |||
@@ -304,22 +304,11 @@ process_raw_event(event_t *raw_event __used, void *data, | |||
304 | } | 304 | } |
305 | } | 305 | } |
306 | 306 | ||
307 | static int process_sample_event(event_t *event, struct perf_session *session) | 307 | static int process_sample_event(event_t *event, struct sample_data *sample, |
308 | struct perf_session *session) | ||
308 | { | 309 | { |
309 | struct sample_data data; | 310 | struct thread *thread = perf_session__findnew(session, event->ip.pid); |
310 | struct thread *thread; | ||
311 | 311 | ||
312 | memset(&data, 0, sizeof(data)); | ||
313 | data.time = -1; | ||
314 | data.cpu = -1; | ||
315 | data.period = 1; | ||
316 | |||
317 | event__parse_sample(event, session->sample_type, &data); | ||
318 | |||
319 | dump_printf("(IP, %d): %d/%d: %#Lx period: %Ld\n", event->header.misc, | ||
320 | data.pid, data.tid, data.ip, data.period); | ||
321 | |||
322 | thread = perf_session__findnew(session, event->ip.pid); | ||
323 | if (thread == NULL) { | 312 | if (thread == NULL) { |
324 | pr_debug("problem processing %d event, skipping it.\n", | 313 | pr_debug("problem processing %d event, skipping it.\n", |
325 | event->header.type); | 314 | event->header.type); |
@@ -328,8 +317,8 @@ static int process_sample_event(event_t *event, struct perf_session *session) | |||
328 | 317 | ||
329 | dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); | 318 | dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); |
330 | 319 | ||
331 | process_raw_event(event, data.raw_data, data.cpu, | 320 | process_raw_event(event, sample->raw_data, sample->cpu, |
332 | data.time, thread); | 321 | sample->time, thread); |
333 | 322 | ||
334 | return 0; | 323 | return 0; |
335 | } | 324 | } |
@@ -492,7 +481,8 @@ static void sort_result(void) | |||
492 | static int __cmd_kmem(void) | 481 | static int __cmd_kmem(void) |
493 | { | 482 | { |
494 | int err = -EINVAL; | 483 | int err = -EINVAL; |
495 | struct perf_session *session = perf_session__new(input_name, O_RDONLY, 0, false); | 484 | struct perf_session *session = perf_session__new(input_name, O_RDONLY, |
485 | 0, false, &event_ops); | ||
496 | if (session == NULL) | 486 | if (session == NULL) |
497 | return -ENOMEM; | 487 | return -ENOMEM; |
498 | 488 | ||
@@ -747,6 +737,9 @@ static int __cmd_record(int argc, const char **argv) | |||
747 | rec_argc = ARRAY_SIZE(record_args) + argc - 1; | 737 | rec_argc = ARRAY_SIZE(record_args) + argc - 1; |
748 | rec_argv = calloc(rec_argc + 1, sizeof(char *)); | 738 | rec_argv = calloc(rec_argc + 1, sizeof(char *)); |
749 | 739 | ||
740 | if (rec_argv == NULL) | ||
741 | return -ENOMEM; | ||
742 | |||
750 | for (i = 0; i < ARRAY_SIZE(record_args); i++) | 743 | for (i = 0; i < ARRAY_SIZE(record_args); i++) |
751 | rec_argv[i] = strdup(record_args[i]); | 744 | rec_argv[i] = strdup(record_args[i]); |
752 | 745 | ||
diff --git a/tools/perf/builtin-lock.c b/tools/perf/builtin-lock.c index 821c1586a22..b9c6e543297 100644 --- a/tools/perf/builtin-lock.c +++ b/tools/perf/builtin-lock.c | |||
@@ -834,22 +834,18 @@ static void dump_info(void) | |||
834 | die("Unknown type of information\n"); | 834 | die("Unknown type of information\n"); |
835 | } | 835 | } |
836 | 836 | ||
837 | static int process_sample_event(event_t *self, struct perf_session *s) | 837 | static int process_sample_event(event_t *self, struct sample_data *sample, |
838 | struct perf_session *s) | ||
838 | { | 839 | { |
839 | struct sample_data data; | 840 | struct thread *thread = perf_session__findnew(s, sample->tid); |
840 | struct thread *thread; | ||
841 | 841 | ||
842 | bzero(&data, sizeof(data)); | ||
843 | event__parse_sample(self, s->sample_type, &data); | ||
844 | |||
845 | thread = perf_session__findnew(s, data.tid); | ||
846 | if (thread == NULL) { | 842 | if (thread == NULL) { |
847 | pr_debug("problem processing %d event, skipping it.\n", | 843 | pr_debug("problem processing %d event, skipping it.\n", |
848 | self->header.type); | 844 | self->header.type); |
849 | return -1; | 845 | return -1; |
850 | } | 846 | } |
851 | 847 | ||
852 | process_raw_event(data.raw_data, data.cpu, data.time, thread); | 848 | process_raw_event(sample->raw_data, sample->cpu, sample->time, thread); |
853 | 849 | ||
854 | return 0; | 850 | return 0; |
855 | } | 851 | } |
@@ -862,7 +858,7 @@ static struct perf_event_ops eops = { | |||
862 | 858 | ||
863 | static int read_events(void) | 859 | static int read_events(void) |
864 | { | 860 | { |
865 | session = perf_session__new(input_name, O_RDONLY, 0, false); | 861 | session = perf_session__new(input_name, O_RDONLY, 0, false, &eops); |
866 | if (!session) | 862 | if (!session) |
867 | die("Initializing perf session failed\n"); | 863 | die("Initializing perf session failed\n"); |
868 | 864 | ||
@@ -947,6 +943,9 @@ static int __cmd_record(int argc, const char **argv) | |||
947 | rec_argc = ARRAY_SIZE(record_args) + argc - 1; | 943 | rec_argc = ARRAY_SIZE(record_args) + argc - 1; |
948 | rec_argv = calloc(rec_argc + 1, sizeof(char *)); | 944 | rec_argv = calloc(rec_argc + 1, sizeof(char *)); |
949 | 945 | ||
946 | if (rec_argv == NULL) | ||
947 | return -ENOMEM; | ||
948 | |||
950 | for (i = 0; i < ARRAY_SIZE(record_args); i++) | 949 | for (i = 0; i < ARRAY_SIZE(record_args); i++) |
951 | rec_argv[i] = strdup(record_args[i]); | 950 | rec_argv[i] = strdup(record_args[i]); |
952 | 951 | ||
@@ -982,9 +981,9 @@ int cmd_lock(int argc, const char **argv, const char *prefix __used) | |||
982 | usage_with_options(report_usage, report_options); | 981 | usage_with_options(report_usage, report_options); |
983 | } | 982 | } |
984 | __cmd_report(); | 983 | __cmd_report(); |
985 | } else if (!strcmp(argv[0], "trace")) { | 984 | } else if (!strcmp(argv[0], "script")) { |
986 | /* Aliased to 'perf trace' */ | 985 | /* Aliased to 'perf script' */ |
987 | return cmd_trace(argc, argv, prefix); | 986 | return cmd_script(argc, argv, prefix); |
988 | } else if (!strcmp(argv[0], "info")) { | 987 | } else if (!strcmp(argv[0], "info")) { |
989 | if (argc) { | 988 | if (argc) { |
990 | argc = parse_options(argc, argv, | 989 | argc = parse_options(argc, argv, |
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 564491fa18b..7bc04903548 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c | |||
@@ -18,6 +18,7 @@ | |||
18 | 18 | ||
19 | #include "util/header.h" | 19 | #include "util/header.h" |
20 | #include "util/event.h" | 20 | #include "util/event.h" |
21 | #include "util/evsel.h" | ||
21 | #include "util/debug.h" | 22 | #include "util/debug.h" |
22 | #include "util/session.h" | 23 | #include "util/session.h" |
23 | #include "util/symbol.h" | 24 | #include "util/symbol.h" |
@@ -27,17 +28,18 @@ | |||
27 | #include <sched.h> | 28 | #include <sched.h> |
28 | #include <sys/mman.h> | 29 | #include <sys/mman.h> |
29 | 30 | ||
31 | #define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y)) | ||
32 | |||
30 | enum write_mode_t { | 33 | enum write_mode_t { |
31 | WRITE_FORCE, | 34 | WRITE_FORCE, |
32 | WRITE_APPEND | 35 | WRITE_APPEND |
33 | }; | 36 | }; |
34 | 37 | ||
35 | static int *fd[MAX_NR_CPUS][MAX_COUNTERS]; | ||
36 | |||
37 | static u64 user_interval = ULLONG_MAX; | 38 | static u64 user_interval = ULLONG_MAX; |
38 | static u64 default_interval = 0; | 39 | static u64 default_interval = 0; |
40 | static u64 sample_type; | ||
39 | 41 | ||
40 | static int nr_cpus = 0; | 42 | static struct cpu_map *cpus; |
41 | static unsigned int page_size; | 43 | static unsigned int page_size; |
42 | static unsigned int mmap_pages = 128; | 44 | static unsigned int mmap_pages = 128; |
43 | static unsigned int user_freq = UINT_MAX; | 45 | static unsigned int user_freq = UINT_MAX; |
@@ -48,11 +50,11 @@ static const char *output_name = "perf.data"; | |||
48 | static int group = 0; | 50 | static int group = 0; |
49 | static int realtime_prio = 0; | 51 | static int realtime_prio = 0; |
50 | static bool raw_samples = false; | 52 | static bool raw_samples = false; |
53 | static bool sample_id_all_avail = true; | ||
51 | static bool system_wide = false; | 54 | static bool system_wide = false; |
52 | static pid_t target_pid = -1; | 55 | static pid_t target_pid = -1; |
53 | static pid_t target_tid = -1; | 56 | static pid_t target_tid = -1; |
54 | static pid_t *all_tids = NULL; | 57 | static struct thread_map *threads; |
55 | static int thread_num = 0; | ||
56 | static pid_t child_pid = -1; | 58 | static pid_t child_pid = -1; |
57 | static bool no_inherit = false; | 59 | static bool no_inherit = false; |
58 | static enum write_mode_t write_mode = WRITE_FORCE; | 60 | static enum write_mode_t write_mode = WRITE_FORCE; |
@@ -60,7 +62,9 @@ static bool call_graph = false; | |||
60 | static bool inherit_stat = false; | 62 | static bool inherit_stat = false; |
61 | static bool no_samples = false; | 63 | static bool no_samples = false; |
62 | static bool sample_address = false; | 64 | static bool sample_address = false; |
65 | static bool sample_time = false; | ||
63 | static bool no_buildid = false; | 66 | static bool no_buildid = false; |
67 | static bool no_buildid_cache = false; | ||
64 | 68 | ||
65 | static long samples = 0; | 69 | static long samples = 0; |
66 | static u64 bytes_written = 0; | 70 | static u64 bytes_written = 0; |
@@ -77,7 +81,6 @@ static struct perf_session *session; | |||
77 | static const char *cpu_list; | 81 | static const char *cpu_list; |
78 | 82 | ||
79 | struct mmap_data { | 83 | struct mmap_data { |
80 | int counter; | ||
81 | void *base; | 84 | void *base; |
82 | unsigned int mask; | 85 | unsigned int mask; |
83 | unsigned int prev; | 86 | unsigned int prev; |
@@ -128,6 +131,7 @@ static void write_output(void *buf, size_t size) | |||
128 | } | 131 | } |
129 | 132 | ||
130 | static int process_synthesized_event(event_t *event, | 133 | static int process_synthesized_event(event_t *event, |
134 | struct sample_data *sample __used, | ||
131 | struct perf_session *self __used) | 135 | struct perf_session *self __used) |
132 | { | 136 | { |
133 | write_output(event, event->header.size); | 137 | write_output(event, event->header.size); |
@@ -224,12 +228,12 @@ static struct perf_header_attr *get_header_attr(struct perf_event_attr *a, int n | |||
224 | return h_attr; | 228 | return h_attr; |
225 | } | 229 | } |
226 | 230 | ||
227 | static void create_counter(int counter, int cpu) | 231 | static void create_counter(struct perf_evsel *evsel, int cpu) |
228 | { | 232 | { |
229 | char *filter = filters[counter]; | 233 | char *filter = evsel->filter; |
230 | struct perf_event_attr *attr = attrs + counter; | 234 | struct perf_event_attr *attr = &evsel->attr; |
231 | struct perf_header_attr *h_attr; | 235 | struct perf_header_attr *h_attr; |
232 | int track = !counter; /* only the first counter needs these */ | 236 | int track = !evsel->idx; /* only the first counter needs these */ |
233 | int thread_index; | 237 | int thread_index; |
234 | int ret; | 238 | int ret; |
235 | struct { | 239 | struct { |
@@ -238,6 +242,19 @@ static void create_counter(int counter, int cpu) | |||
238 | u64 time_running; | 242 | u64 time_running; |
239 | u64 id; | 243 | u64 id; |
240 | } read_data; | 244 | } read_data; |
245 | /* | ||
246 | * Check if parse_single_tracepoint_event has already asked for | ||
247 | * PERF_SAMPLE_TIME. | ||
248 | * | ||
249 | * XXX this is kludgy but short term fix for problems introduced by | ||
250 | * eac23d1c that broke 'perf script' by having different sample_types | ||
251 | * when using multiple tracepoint events when we use a perf binary | ||
252 | * that tries to use sample_id_all on an older kernel. | ||
253 | * | ||
254 | * We need to move counter creation to perf_session, support | ||
255 | * different sample_types, etc. | ||
256 | */ | ||
257 | bool time_needed = attr->sample_type & PERF_SAMPLE_TIME; | ||
241 | 258 | ||
242 | attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED | | 259 | attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED | |
243 | PERF_FORMAT_TOTAL_TIME_RUNNING | | 260 | PERF_FORMAT_TOTAL_TIME_RUNNING | |
@@ -280,6 +297,10 @@ static void create_counter(int counter, int cpu) | |||
280 | if (system_wide) | 297 | if (system_wide) |
281 | attr->sample_type |= PERF_SAMPLE_CPU; | 298 | attr->sample_type |= PERF_SAMPLE_CPU; |
282 | 299 | ||
300 | if (sample_id_all_avail && | ||
301 | (sample_time || system_wide || !no_inherit || cpu_list)) | ||
302 | attr->sample_type |= PERF_SAMPLE_TIME; | ||
303 | |||
283 | if (raw_samples) { | 304 | if (raw_samples) { |
284 | attr->sample_type |= PERF_SAMPLE_TIME; | 305 | attr->sample_type |= PERF_SAMPLE_TIME; |
285 | attr->sample_type |= PERF_SAMPLE_RAW; | 306 | attr->sample_type |= PERF_SAMPLE_RAW; |
@@ -293,13 +314,14 @@ static void create_counter(int counter, int cpu) | |||
293 | attr->disabled = 1; | 314 | attr->disabled = 1; |
294 | attr->enable_on_exec = 1; | 315 | attr->enable_on_exec = 1; |
295 | } | 316 | } |
317 | retry_sample_id: | ||
318 | attr->sample_id_all = sample_id_all_avail ? 1 : 0; | ||
296 | 319 | ||
297 | for (thread_index = 0; thread_index < thread_num; thread_index++) { | 320 | for (thread_index = 0; thread_index < threads->nr; thread_index++) { |
298 | try_again: | 321 | try_again: |
299 | fd[nr_cpu][counter][thread_index] = sys_perf_event_open(attr, | 322 | FD(evsel, nr_cpu, thread_index) = sys_perf_event_open(attr, threads->map[thread_index], cpu, group_fd, 0); |
300 | all_tids[thread_index], cpu, group_fd, 0); | ||
301 | 323 | ||
302 | if (fd[nr_cpu][counter][thread_index] < 0) { | 324 | if (FD(evsel, nr_cpu, thread_index) < 0) { |
303 | int err = errno; | 325 | int err = errno; |
304 | 326 | ||
305 | if (err == EPERM || err == EACCES) | 327 | if (err == EPERM || err == EACCES) |
@@ -309,6 +331,15 @@ try_again: | |||
309 | else if (err == ENODEV && cpu_list) { | 331 | else if (err == ENODEV && cpu_list) { |
310 | die("No such device - did you specify" | 332 | die("No such device - did you specify" |
311 | " an out-of-range profile CPU?\n"); | 333 | " an out-of-range profile CPU?\n"); |
334 | } else if (err == EINVAL && sample_id_all_avail) { | ||
335 | /* | ||
336 | * Old kernel, no attr->sample_id_type_all field | ||
337 | */ | ||
338 | sample_id_all_avail = false; | ||
339 | if (!sample_time && !raw_samples && !time_needed) | ||
340 | attr->sample_type &= ~PERF_SAMPLE_TIME; | ||
341 | |||
342 | goto retry_sample_id; | ||
312 | } | 343 | } |
313 | 344 | ||
314 | /* | 345 | /* |
@@ -326,8 +357,8 @@ try_again: | |||
326 | goto try_again; | 357 | goto try_again; |
327 | } | 358 | } |
328 | printf("\n"); | 359 | printf("\n"); |
329 | error("perfcounter syscall returned with %d (%s)\n", | 360 | error("sys_perf_event_open() syscall returned with %d (%s). /bin/dmesg may provide additional information.\n", |
330 | fd[nr_cpu][counter][thread_index], strerror(err)); | 361 | FD(evsel, nr_cpu, thread_index), strerror(err)); |
331 | 362 | ||
332 | #if defined(__i386__) || defined(__x86_64__) | 363 | #if defined(__i386__) || defined(__x86_64__) |
333 | if (attr->type == PERF_TYPE_HARDWARE && err == EOPNOTSUPP) | 364 | if (attr->type == PERF_TYPE_HARDWARE && err == EOPNOTSUPP) |
@@ -341,7 +372,7 @@ try_again: | |||
341 | exit(-1); | 372 | exit(-1); |
342 | } | 373 | } |
343 | 374 | ||
344 | h_attr = get_header_attr(attr, counter); | 375 | h_attr = get_header_attr(attr, evsel->idx); |
345 | if (h_attr == NULL) | 376 | if (h_attr == NULL) |
346 | die("nomem\n"); | 377 | die("nomem\n"); |
347 | 378 | ||
@@ -352,7 +383,7 @@ try_again: | |||
352 | } | 383 | } |
353 | } | 384 | } |
354 | 385 | ||
355 | if (read(fd[nr_cpu][counter][thread_index], &read_data, sizeof(read_data)) == -1) { | 386 | if (read(FD(evsel, nr_cpu, thread_index), &read_data, sizeof(read_data)) == -1) { |
356 | perror("Unable to read perf file descriptor"); | 387 | perror("Unable to read perf file descriptor"); |
357 | exit(-1); | 388 | exit(-1); |
358 | } | 389 | } |
@@ -362,43 +393,44 @@ try_again: | |||
362 | exit(-1); | 393 | exit(-1); |
363 | } | 394 | } |
364 | 395 | ||
365 | assert(fd[nr_cpu][counter][thread_index] >= 0); | 396 | assert(FD(evsel, nr_cpu, thread_index) >= 0); |
366 | fcntl(fd[nr_cpu][counter][thread_index], F_SETFL, O_NONBLOCK); | 397 | fcntl(FD(evsel, nr_cpu, thread_index), F_SETFL, O_NONBLOCK); |
367 | 398 | ||
368 | /* | 399 | /* |
369 | * First counter acts as the group leader: | 400 | * First counter acts as the group leader: |
370 | */ | 401 | */ |
371 | if (group && group_fd == -1) | 402 | if (group && group_fd == -1) |
372 | group_fd = fd[nr_cpu][counter][thread_index]; | 403 | group_fd = FD(evsel, nr_cpu, thread_index); |
373 | 404 | ||
374 | if (counter || thread_index) { | 405 | if (evsel->idx || thread_index) { |
375 | ret = ioctl(fd[nr_cpu][counter][thread_index], | 406 | struct perf_evsel *first; |
376 | PERF_EVENT_IOC_SET_OUTPUT, | 407 | first = list_entry(evsel_list.next, struct perf_evsel, node); |
377 | fd[nr_cpu][0][0]); | 408 | ret = ioctl(FD(evsel, nr_cpu, thread_index), |
409 | PERF_EVENT_IOC_SET_OUTPUT, | ||
410 | FD(first, nr_cpu, 0)); | ||
378 | if (ret) { | 411 | if (ret) { |
379 | error("failed to set output: %d (%s)\n", errno, | 412 | error("failed to set output: %d (%s)\n", errno, |
380 | strerror(errno)); | 413 | strerror(errno)); |
381 | exit(-1); | 414 | exit(-1); |
382 | } | 415 | } |
383 | } else { | 416 | } else { |
384 | mmap_array[nr_cpu].counter = counter; | ||
385 | mmap_array[nr_cpu].prev = 0; | 417 | mmap_array[nr_cpu].prev = 0; |
386 | mmap_array[nr_cpu].mask = mmap_pages*page_size - 1; | 418 | mmap_array[nr_cpu].mask = mmap_pages*page_size - 1; |
387 | mmap_array[nr_cpu].base = mmap(NULL, (mmap_pages+1)*page_size, | 419 | mmap_array[nr_cpu].base = mmap(NULL, (mmap_pages+1)*page_size, |
388 | PROT_READ|PROT_WRITE, MAP_SHARED, fd[nr_cpu][counter][thread_index], 0); | 420 | PROT_READ | PROT_WRITE, MAP_SHARED, FD(evsel, nr_cpu, thread_index), 0); |
389 | if (mmap_array[nr_cpu].base == MAP_FAILED) { | 421 | if (mmap_array[nr_cpu].base == MAP_FAILED) { |
390 | error("failed to mmap with %d (%s)\n", errno, strerror(errno)); | 422 | error("failed to mmap with %d (%s)\n", errno, strerror(errno)); |
391 | exit(-1); | 423 | exit(-1); |
392 | } | 424 | } |
393 | 425 | ||
394 | event_array[nr_poll].fd = fd[nr_cpu][counter][thread_index]; | 426 | event_array[nr_poll].fd = FD(evsel, nr_cpu, thread_index); |
395 | event_array[nr_poll].events = POLLIN; | 427 | event_array[nr_poll].events = POLLIN; |
396 | nr_poll++; | 428 | nr_poll++; |
397 | } | 429 | } |
398 | 430 | ||
399 | if (filter != NULL) { | 431 | if (filter != NULL) { |
400 | ret = ioctl(fd[nr_cpu][counter][thread_index], | 432 | ret = ioctl(FD(evsel, nr_cpu, thread_index), |
401 | PERF_EVENT_IOC_SET_FILTER, filter); | 433 | PERF_EVENT_IOC_SET_FILTER, filter); |
402 | if (ret) { | 434 | if (ret) { |
403 | error("failed to set filter with %d (%s)\n", errno, | 435 | error("failed to set filter with %d (%s)\n", errno, |
404 | strerror(errno)); | 436 | strerror(errno)); |
@@ -406,15 +438,19 @@ try_again: | |||
406 | } | 438 | } |
407 | } | 439 | } |
408 | } | 440 | } |
441 | |||
442 | if (!sample_type) | ||
443 | sample_type = attr->sample_type; | ||
409 | } | 444 | } |
410 | 445 | ||
411 | static void open_counters(int cpu) | 446 | static void open_counters(int cpu) |
412 | { | 447 | { |
413 | int counter; | 448 | struct perf_evsel *pos; |
414 | 449 | ||
415 | group_fd = -1; | 450 | group_fd = -1; |
416 | for (counter = 0; counter < nr_counters; counter++) | 451 | |
417 | create_counter(counter, cpu); | 452 | list_for_each_entry(pos, &evsel_list, node) |
453 | create_counter(pos, cpu); | ||
418 | 454 | ||
419 | nr_cpu++; | 455 | nr_cpu++; |
420 | } | 456 | } |
@@ -437,7 +473,8 @@ static void atexit_header(void) | |||
437 | if (!pipe_output) { | 473 | if (!pipe_output) { |
438 | session->header.data_size += bytes_written; | 474 | session->header.data_size += bytes_written; |
439 | 475 | ||
440 | process_buildids(); | 476 | if (!no_buildid) |
477 | process_buildids(); | ||
441 | perf_header__write(&session->header, output, true); | 478 | perf_header__write(&session->header, output, true); |
442 | perf_session__delete(session); | 479 | perf_session__delete(session); |
443 | symbol__exit(); | 480 | symbol__exit(); |
@@ -500,7 +537,7 @@ static void mmap_read_all(void) | |||
500 | 537 | ||
501 | static int __cmd_record(int argc, const char **argv) | 538 | static int __cmd_record(int argc, const char **argv) |
502 | { | 539 | { |
503 | int i, counter; | 540 | int i; |
504 | struct stat st; | 541 | struct stat st; |
505 | int flags; | 542 | int flags; |
506 | int err; | 543 | int err; |
@@ -552,19 +589,22 @@ static int __cmd_record(int argc, const char **argv) | |||
552 | } | 589 | } |
553 | 590 | ||
554 | session = perf_session__new(output_name, O_WRONLY, | 591 | session = perf_session__new(output_name, O_WRONLY, |
555 | write_mode == WRITE_FORCE, false); | 592 | write_mode == WRITE_FORCE, false, NULL); |
556 | if (session == NULL) { | 593 | if (session == NULL) { |
557 | pr_err("Not enough memory for reading perf file header\n"); | 594 | pr_err("Not enough memory for reading perf file header\n"); |
558 | return -1; | 595 | return -1; |
559 | } | 596 | } |
560 | 597 | ||
598 | if (!no_buildid) | ||
599 | perf_header__set_feat(&session->header, HEADER_BUILD_ID); | ||
600 | |||
561 | if (!file_new) { | 601 | if (!file_new) { |
562 | err = perf_header__read(session, output); | 602 | err = perf_header__read(session, output); |
563 | if (err < 0) | 603 | if (err < 0) |
564 | goto out_delete_session; | 604 | goto out_delete_session; |
565 | } | 605 | } |
566 | 606 | ||
567 | if (have_tracepoints(attrs, nr_counters)) | 607 | if (have_tracepoints(&evsel_list)) |
568 | perf_header__set_feat(&session->header, HEADER_TRACE_INFO); | 608 | perf_header__set_feat(&session->header, HEADER_TRACE_INFO); |
569 | 609 | ||
570 | /* | 610 | /* |
@@ -612,7 +652,7 @@ static int __cmd_record(int argc, const char **argv) | |||
612 | } | 652 | } |
613 | 653 | ||
614 | if (!system_wide && target_tid == -1 && target_pid == -1) | 654 | if (!system_wide && target_tid == -1 && target_pid == -1) |
615 | all_tids[0] = child_pid; | 655 | threads->map[0] = child_pid; |
616 | 656 | ||
617 | close(child_ready_pipe[1]); | 657 | close(child_ready_pipe[1]); |
618 | close(go_pipe[0]); | 658 | close(go_pipe[0]); |
@@ -626,19 +666,15 @@ static int __cmd_record(int argc, const char **argv) | |||
626 | close(child_ready_pipe[0]); | 666 | close(child_ready_pipe[0]); |
627 | } | 667 | } |
628 | 668 | ||
629 | nr_cpus = read_cpu_map(cpu_list); | ||
630 | if (nr_cpus < 1) { | ||
631 | perror("failed to collect number of CPUs"); | ||
632 | return -1; | ||
633 | } | ||
634 | |||
635 | if (!system_wide && no_inherit && !cpu_list) { | 669 | if (!system_wide && no_inherit && !cpu_list) { |
636 | open_counters(-1); | 670 | open_counters(-1); |
637 | } else { | 671 | } else { |
638 | for (i = 0; i < nr_cpus; i++) | 672 | for (i = 0; i < cpus->nr; i++) |
639 | open_counters(cpumap[i]); | 673 | open_counters(cpus->map[i]); |
640 | } | 674 | } |
641 | 675 | ||
676 | perf_session__set_sample_type(session, sample_type); | ||
677 | |||
642 | if (pipe_output) { | 678 | if (pipe_output) { |
643 | err = perf_header__write_pipe(output); | 679 | err = perf_header__write_pipe(output); |
644 | if (err < 0) | 680 | if (err < 0) |
@@ -651,6 +687,8 @@ static int __cmd_record(int argc, const char **argv) | |||
651 | 687 | ||
652 | post_processing_offset = lseek(output, 0, SEEK_CUR); | 688 | post_processing_offset = lseek(output, 0, SEEK_CUR); |
653 | 689 | ||
690 | perf_session__set_sample_id_all(session, sample_id_all_avail); | ||
691 | |||
654 | if (pipe_output) { | 692 | if (pipe_output) { |
655 | err = event__synthesize_attrs(&session->header, | 693 | err = event__synthesize_attrs(&session->header, |
656 | process_synthesized_event, | 694 | process_synthesized_event, |
@@ -667,7 +705,7 @@ static int __cmd_record(int argc, const char **argv) | |||
667 | return err; | 705 | return err; |
668 | } | 706 | } |
669 | 707 | ||
670 | if (have_tracepoints(attrs, nr_counters)) { | 708 | if (have_tracepoints(&evsel_list)) { |
671 | /* | 709 | /* |
672 | * FIXME err <= 0 here actually means that | 710 | * FIXME err <= 0 here actually means that |
673 | * there were no tracepoints so its not really | 711 | * there were no tracepoints so its not really |
@@ -676,8 +714,7 @@ static int __cmd_record(int argc, const char **argv) | |||
676 | * return this more properly and also | 714 | * return this more properly and also |
677 | * propagate errors that now are calling die() | 715 | * propagate errors that now are calling die() |
678 | */ | 716 | */ |
679 | err = event__synthesize_tracing_data(output, attrs, | 717 | err = event__synthesize_tracing_data(output, &evsel_list, |
680 | nr_counters, | ||
681 | process_synthesized_event, | 718 | process_synthesized_event, |
682 | session); | 719 | session); |
683 | if (err <= 0) { | 720 | if (err <= 0) { |
@@ -751,13 +788,13 @@ static int __cmd_record(int argc, const char **argv) | |||
751 | 788 | ||
752 | if (done) { | 789 | if (done) { |
753 | for (i = 0; i < nr_cpu; i++) { | 790 | for (i = 0; i < nr_cpu; i++) { |
754 | for (counter = 0; | 791 | struct perf_evsel *pos; |
755 | counter < nr_counters; | 792 | |
756 | counter++) { | 793 | list_for_each_entry(pos, &evsel_list, node) { |
757 | for (thread = 0; | 794 | for (thread = 0; |
758 | thread < thread_num; | 795 | thread < threads->nr; |
759 | thread++) | 796 | thread++) |
760 | ioctl(fd[i][counter][thread], | 797 | ioctl(FD(pos, i, thread), |
761 | PERF_EVENT_IOC_DISABLE); | 798 | PERF_EVENT_IOC_DISABLE); |
762 | } | 799 | } |
763 | } | 800 | } |
@@ -831,16 +868,20 @@ const struct option record_options[] = { | |||
831 | "per thread counts"), | 868 | "per thread counts"), |
832 | OPT_BOOLEAN('d', "data", &sample_address, | 869 | OPT_BOOLEAN('d', "data", &sample_address, |
833 | "Sample addresses"), | 870 | "Sample addresses"), |
871 | OPT_BOOLEAN('T', "timestamp", &sample_time, "Sample timestamps"), | ||
834 | OPT_BOOLEAN('n', "no-samples", &no_samples, | 872 | OPT_BOOLEAN('n', "no-samples", &no_samples, |
835 | "don't sample"), | 873 | "don't sample"), |
836 | OPT_BOOLEAN('N', "no-buildid-cache", &no_buildid, | 874 | OPT_BOOLEAN('N', "no-buildid-cache", &no_buildid_cache, |
837 | "do not update the buildid cache"), | 875 | "do not update the buildid cache"), |
876 | OPT_BOOLEAN('B', "no-buildid", &no_buildid, | ||
877 | "do not collect buildids in perf.data"), | ||
838 | OPT_END() | 878 | OPT_END() |
839 | }; | 879 | }; |
840 | 880 | ||
841 | int cmd_record(int argc, const char **argv, const char *prefix __used) | 881 | int cmd_record(int argc, const char **argv, const char *prefix __used) |
842 | { | 882 | { |
843 | int i, j, err = -ENOMEM; | 883 | int err = -ENOMEM; |
884 | struct perf_evsel *pos; | ||
844 | 885 | ||
845 | argc = parse_options(argc, argv, record_options, record_usage, | 886 | argc = parse_options(argc, argv, record_options, record_usage, |
846 | PARSE_OPT_STOP_AT_NON_OPTION); | 887 | PARSE_OPT_STOP_AT_NON_OPTION); |
@@ -859,41 +900,36 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) | |||
859 | } | 900 | } |
860 | 901 | ||
861 | symbol__init(); | 902 | symbol__init(); |
862 | if (no_buildid) | 903 | |
904 | if (no_buildid_cache || no_buildid) | ||
863 | disable_buildid_cache(); | 905 | disable_buildid_cache(); |
864 | 906 | ||
865 | if (!nr_counters) { | 907 | if (list_empty(&evsel_list) && perf_evsel_list__create_default() < 0) { |
866 | nr_counters = 1; | 908 | pr_err("Not enough memory for event selector list\n"); |
867 | attrs[0].type = PERF_TYPE_HARDWARE; | 909 | goto out_symbol_exit; |
868 | attrs[0].config = PERF_COUNT_HW_CPU_CYCLES; | ||
869 | } | 910 | } |
870 | 911 | ||
871 | if (target_pid != -1) { | 912 | if (target_pid != -1) |
872 | target_tid = target_pid; | 913 | target_tid = target_pid; |
873 | thread_num = find_all_tid(target_pid, &all_tids); | ||
874 | if (thread_num <= 0) { | ||
875 | fprintf(stderr, "Can't find all threads of pid %d\n", | ||
876 | target_pid); | ||
877 | usage_with_options(record_usage, record_options); | ||
878 | } | ||
879 | } else { | ||
880 | all_tids=malloc(sizeof(pid_t)); | ||
881 | if (!all_tids) | ||
882 | goto out_symbol_exit; | ||
883 | 914 | ||
884 | all_tids[0] = target_tid; | 915 | threads = thread_map__new(target_pid, target_tid); |
885 | thread_num = 1; | 916 | if (threads == NULL) { |
917 | pr_err("Problems finding threads of monitor\n"); | ||
918 | usage_with_options(record_usage, record_options); | ||
886 | } | 919 | } |
887 | 920 | ||
888 | for (i = 0; i < MAX_NR_CPUS; i++) { | 921 | cpus = cpu_map__new(cpu_list); |
889 | for (j = 0; j < MAX_COUNTERS; j++) { | 922 | if (cpus == NULL) { |
890 | fd[i][j] = malloc(sizeof(int)*thread_num); | 923 | perror("failed to parse CPUs map"); |
891 | if (!fd[i][j]) | 924 | return -1; |
892 | goto out_free_fd; | ||
893 | } | ||
894 | } | 925 | } |
895 | event_array = malloc( | 926 | |
896 | sizeof(struct pollfd)*MAX_NR_CPUS*MAX_COUNTERS*thread_num); | 927 | list_for_each_entry(pos, &evsel_list, node) { |
928 | if (perf_evsel__alloc_fd(pos, cpus->nr, threads->nr) < 0) | ||
929 | goto out_free_fd; | ||
930 | } | ||
931 | event_array = malloc((sizeof(struct pollfd) * MAX_NR_CPUS * | ||
932 | MAX_COUNTERS * threads->nr)); | ||
897 | if (!event_array) | 933 | if (!event_array) |
898 | goto out_free_fd; | 934 | goto out_free_fd; |
899 | 935 | ||
@@ -920,12 +956,8 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) | |||
920 | out_free_event_array: | 956 | out_free_event_array: |
921 | free(event_array); | 957 | free(event_array); |
922 | out_free_fd: | 958 | out_free_fd: |
923 | for (i = 0; i < MAX_NR_CPUS; i++) { | 959 | thread_map__delete(threads); |
924 | for (j = 0; j < MAX_COUNTERS; j++) | 960 | threads = NULL; |
925 | free(fd[i][j]); | ||
926 | } | ||
927 | free(all_tids); | ||
928 | all_tids = NULL; | ||
929 | out_symbol_exit: | 961 | out_symbol_exit: |
930 | symbol__exit(); | 962 | symbol__exit(); |
931 | return err; | 963 | return err; |
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 5de405d4523..75183a4518e 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c | |||
@@ -150,13 +150,13 @@ static int add_event_total(struct perf_session *session, | |||
150 | return 0; | 150 | return 0; |
151 | } | 151 | } |
152 | 152 | ||
153 | static int process_sample_event(event_t *event, struct perf_session *session) | 153 | static int process_sample_event(event_t *event, struct sample_data *sample, |
154 | struct perf_session *session) | ||
154 | { | 155 | { |
155 | struct sample_data data = { .period = 1, }; | ||
156 | struct addr_location al; | 156 | struct addr_location al; |
157 | struct perf_event_attr *attr; | 157 | struct perf_event_attr *attr; |
158 | 158 | ||
159 | if (event__preprocess_sample(event, session, &al, &data, NULL) < 0) { | 159 | if (event__preprocess_sample(event, session, &al, sample, NULL) < 0) { |
160 | fprintf(stderr, "problem processing %d event, skipping it.\n", | 160 | fprintf(stderr, "problem processing %d event, skipping it.\n", |
161 | event->header.type); | 161 | event->header.type); |
162 | return -1; | 162 | return -1; |
@@ -165,14 +165,14 @@ static int process_sample_event(event_t *event, struct perf_session *session) | |||
165 | if (al.filtered || (hide_unresolved && al.sym == NULL)) | 165 | if (al.filtered || (hide_unresolved && al.sym == NULL)) |
166 | return 0; | 166 | return 0; |
167 | 167 | ||
168 | if (perf_session__add_hist_entry(session, &al, &data)) { | 168 | if (perf_session__add_hist_entry(session, &al, sample)) { |
169 | pr_debug("problem incrementing symbol period, skipping event\n"); | 169 | pr_debug("problem incrementing symbol period, skipping event\n"); |
170 | return -1; | 170 | return -1; |
171 | } | 171 | } |
172 | 172 | ||
173 | attr = perf_header__find_attr(data.id, &session->header); | 173 | attr = perf_header__find_attr(sample->id, &session->header); |
174 | 174 | ||
175 | if (add_event_total(session, &data, attr)) { | 175 | if (add_event_total(session, sample, attr)) { |
176 | pr_debug("problem adding event period\n"); | 176 | pr_debug("problem adding event period\n"); |
177 | return -1; | 177 | return -1; |
178 | } | 178 | } |
@@ -180,7 +180,8 @@ static int process_sample_event(event_t *event, struct perf_session *session) | |||
180 | return 0; | 180 | return 0; |
181 | } | 181 | } |
182 | 182 | ||
183 | static int process_read_event(event_t *event, struct perf_session *session __used) | 183 | static int process_read_event(event_t *event, struct sample_data *sample __used, |
184 | struct perf_session *session __used) | ||
184 | { | 185 | { |
185 | struct perf_event_attr *attr; | 186 | struct perf_event_attr *attr; |
186 | 187 | ||
@@ -243,6 +244,8 @@ static struct perf_event_ops event_ops = { | |||
243 | .event_type = event__process_event_type, | 244 | .event_type = event__process_event_type, |
244 | .tracing_data = event__process_tracing_data, | 245 | .tracing_data = event__process_tracing_data, |
245 | .build_id = event__process_build_id, | 246 | .build_id = event__process_build_id, |
247 | .ordered_samples = true, | ||
248 | .ordering_requires_timestamps = true, | ||
246 | }; | 249 | }; |
247 | 250 | ||
248 | extern volatile int session_done; | 251 | extern volatile int session_done; |
@@ -307,7 +310,7 @@ static int __cmd_report(void) | |||
307 | 310 | ||
308 | signal(SIGINT, sig_handler); | 311 | signal(SIGINT, sig_handler); |
309 | 312 | ||
310 | session = perf_session__new(input_name, O_RDONLY, force, false); | 313 | session = perf_session__new(input_name, O_RDONLY, force, false, &event_ops); |
311 | if (session == NULL) | 314 | if (session == NULL) |
312 | return -ENOMEM; | 315 | return -ENOMEM; |
313 | 316 | ||
@@ -442,6 +445,8 @@ static const struct option options[] = { | |||
442 | "dump raw trace in ASCII"), | 445 | "dump raw trace in ASCII"), |
443 | OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name, | 446 | OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name, |
444 | "file", "vmlinux pathname"), | 447 | "file", "vmlinux pathname"), |
448 | OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name, | ||
449 | "file", "kallsyms pathname"), | ||
445 | OPT_BOOLEAN('f', "force", &force, "don't complain, do it"), | 450 | OPT_BOOLEAN('f', "force", &force, "don't complain, do it"), |
446 | OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules, | 451 | OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules, |
447 | "load module symbols - WARNING: use only with -k and LIVE kernel"), | 452 | "load module symbols - WARNING: use only with -k and LIVE kernel"), |
@@ -478,6 +483,8 @@ static const struct option options[] = { | |||
478 | "columns '.' is reserved."), | 483 | "columns '.' is reserved."), |
479 | OPT_BOOLEAN('U', "hide-unresolved", &hide_unresolved, | 484 | OPT_BOOLEAN('U', "hide-unresolved", &hide_unresolved, |
480 | "Only display entries resolved to a symbol"), | 485 | "Only display entries resolved to a symbol"), |
486 | OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory", | ||
487 | "Look for files with symbols relative to this directory"), | ||
481 | OPT_END() | 488 | OPT_END() |
482 | }; | 489 | }; |
483 | 490 | ||
diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 55f3b5dcc73..7a4ebeb8b01 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c | |||
@@ -1606,25 +1606,15 @@ process_raw_event(event_t *raw_event __used, struct perf_session *session, | |||
1606 | process_sched_migrate_task_event(data, session, event, cpu, timestamp, thread); | 1606 | process_sched_migrate_task_event(data, session, event, cpu, timestamp, thread); |
1607 | } | 1607 | } |
1608 | 1608 | ||
1609 | static int process_sample_event(event_t *event, struct perf_session *session) | 1609 | static int process_sample_event(event_t *event, struct sample_data *sample, |
1610 | struct perf_session *session) | ||
1610 | { | 1611 | { |
1611 | struct sample_data data; | ||
1612 | struct thread *thread; | 1612 | struct thread *thread; |
1613 | 1613 | ||
1614 | if (!(session->sample_type & PERF_SAMPLE_RAW)) | 1614 | if (!(session->sample_type & PERF_SAMPLE_RAW)) |
1615 | return 0; | 1615 | return 0; |
1616 | 1616 | ||
1617 | memset(&data, 0, sizeof(data)); | 1617 | thread = perf_session__findnew(session, sample->pid); |
1618 | data.time = -1; | ||
1619 | data.cpu = -1; | ||
1620 | data.period = -1; | ||
1621 | |||
1622 | event__parse_sample(event, session->sample_type, &data); | ||
1623 | |||
1624 | dump_printf("(IP, %d): %d/%d: %#Lx period: %Ld\n", event->header.misc, | ||
1625 | data.pid, data.tid, data.ip, data.period); | ||
1626 | |||
1627 | thread = perf_session__findnew(session, data.pid); | ||
1628 | if (thread == NULL) { | 1618 | if (thread == NULL) { |
1629 | pr_debug("problem processing %d event, skipping it.\n", | 1619 | pr_debug("problem processing %d event, skipping it.\n", |
1630 | event->header.type); | 1620 | event->header.type); |
@@ -1633,10 +1623,11 @@ static int process_sample_event(event_t *event, struct perf_session *session) | |||
1633 | 1623 | ||
1634 | dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); | 1624 | dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid); |
1635 | 1625 | ||
1636 | if (profile_cpu != -1 && profile_cpu != (int)data.cpu) | 1626 | if (profile_cpu != -1 && profile_cpu != (int)sample->cpu) |
1637 | return 0; | 1627 | return 0; |
1638 | 1628 | ||
1639 | process_raw_event(event, session, data.raw_data, data.cpu, data.time, thread); | 1629 | process_raw_event(event, session, sample->raw_data, sample->cpu, |
1630 | sample->time, thread); | ||
1640 | 1631 | ||
1641 | return 0; | 1632 | return 0; |
1642 | } | 1633 | } |
@@ -1652,7 +1643,8 @@ static struct perf_event_ops event_ops = { | |||
1652 | static int read_events(void) | 1643 | static int read_events(void) |
1653 | { | 1644 | { |
1654 | int err = -EINVAL; | 1645 | int err = -EINVAL; |
1655 | struct perf_session *session = perf_session__new(input_name, O_RDONLY, 0, false); | 1646 | struct perf_session *session = perf_session__new(input_name, O_RDONLY, |
1647 | 0, false, &event_ops); | ||
1656 | if (session == NULL) | 1648 | if (session == NULL) |
1657 | return -ENOMEM; | 1649 | return -ENOMEM; |
1658 | 1650 | ||
@@ -1869,6 +1861,9 @@ static int __cmd_record(int argc, const char **argv) | |||
1869 | rec_argc = ARRAY_SIZE(record_args) + argc - 1; | 1861 | rec_argc = ARRAY_SIZE(record_args) + argc - 1; |
1870 | rec_argv = calloc(rec_argc + 1, sizeof(char *)); | 1862 | rec_argv = calloc(rec_argc + 1, sizeof(char *)); |
1871 | 1863 | ||
1864 | if (rec_argv) | ||
1865 | return -ENOMEM; | ||
1866 | |||
1872 | for (i = 0; i < ARRAY_SIZE(record_args); i++) | 1867 | for (i = 0; i < ARRAY_SIZE(record_args); i++) |
1873 | rec_argv[i] = strdup(record_args[i]); | 1868 | rec_argv[i] = strdup(record_args[i]); |
1874 | 1869 | ||
@@ -1888,10 +1883,10 @@ int cmd_sched(int argc, const char **argv, const char *prefix __used) | |||
1888 | usage_with_options(sched_usage, sched_options); | 1883 | usage_with_options(sched_usage, sched_options); |
1889 | 1884 | ||
1890 | /* | 1885 | /* |
1891 | * Aliased to 'perf trace' for now: | 1886 | * Aliased to 'perf script' for now: |
1892 | */ | 1887 | */ |
1893 | if (!strcmp(argv[0], "trace")) | 1888 | if (!strcmp(argv[0], "script")) |
1894 | return cmd_trace(argc, argv, prefix); | 1889 | return cmd_script(argc, argv, prefix); |
1895 | 1890 | ||
1896 | symbol__init(); | 1891 | symbol__init(); |
1897 | if (!strncmp(argv[0], "rec", 3)) { | 1892 | if (!strncmp(argv[0], "rec", 3)) { |
diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-script.c index 86cfe3800e6..150a606002e 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-script.c | |||
@@ -56,29 +56,18 @@ static void setup_scripting(void) | |||
56 | 56 | ||
57 | static int cleanup_scripting(void) | 57 | static int cleanup_scripting(void) |
58 | { | 58 | { |
59 | pr_debug("\nperf trace script stopped\n"); | 59 | pr_debug("\nperf script stopped\n"); |
60 | 60 | ||
61 | return scripting_ops->stop_script(); | 61 | return scripting_ops->stop_script(); |
62 | } | 62 | } |
63 | 63 | ||
64 | static char const *input_name = "perf.data"; | 64 | static char const *input_name = "perf.data"; |
65 | 65 | ||
66 | static int process_sample_event(event_t *event, struct perf_session *session) | 66 | static int process_sample_event(event_t *event, struct sample_data *sample, |
67 | struct perf_session *session) | ||
67 | { | 68 | { |
68 | struct sample_data data; | 69 | struct thread *thread = perf_session__findnew(session, event->ip.pid); |
69 | struct thread *thread; | ||
70 | 70 | ||
71 | memset(&data, 0, sizeof(data)); | ||
72 | data.time = -1; | ||
73 | data.cpu = -1; | ||
74 | data.period = 1; | ||
75 | |||
76 | event__parse_sample(event, session->sample_type, &data); | ||
77 | |||
78 | dump_printf("(IP, %d): %d/%d: %#Lx period: %Ld\n", event->header.misc, | ||
79 | data.pid, data.tid, data.ip, data.period); | ||
80 | |||
81 | thread = perf_session__findnew(session, event->ip.pid); | ||
82 | if (thread == NULL) { | 71 | if (thread == NULL) { |
83 | pr_debug("problem processing %d event, skipping it.\n", | 72 | pr_debug("problem processing %d event, skipping it.\n", |
84 | event->header.type); | 73 | event->header.type); |
@@ -87,13 +76,13 @@ static int process_sample_event(event_t *event, struct perf_session *session) | |||
87 | 76 | ||
88 | if (session->sample_type & PERF_SAMPLE_RAW) { | 77 | if (session->sample_type & PERF_SAMPLE_RAW) { |
89 | if (debug_mode) { | 78 | if (debug_mode) { |
90 | if (data.time < last_timestamp) { | 79 | if (sample->time < last_timestamp) { |
91 | pr_err("Samples misordered, previous: %llu " | 80 | pr_err("Samples misordered, previous: %llu " |
92 | "this: %llu\n", last_timestamp, | 81 | "this: %llu\n", last_timestamp, |
93 | data.time); | 82 | sample->time); |
94 | nr_unordered++; | 83 | nr_unordered++; |
95 | } | 84 | } |
96 | last_timestamp = data.time; | 85 | last_timestamp = sample->time; |
97 | return 0; | 86 | return 0; |
98 | } | 87 | } |
99 | /* | 88 | /* |
@@ -101,21 +90,12 @@ static int process_sample_event(event_t *event, struct perf_session *session) | |||
101 | * field, although it should be the same than this perf | 90 | * field, although it should be the same than this perf |
102 | * event pid | 91 | * event pid |
103 | */ | 92 | */ |
104 | scripting_ops->process_event(data.cpu, data.raw_data, | 93 | scripting_ops->process_event(sample->cpu, sample->raw_data, |
105 | data.raw_size, | 94 | sample->raw_size, |
106 | data.time, thread->comm); | 95 | sample->time, thread->comm); |
107 | } | 96 | } |
108 | 97 | ||
109 | session->hists.stats.total_period += data.period; | 98 | session->hists.stats.total_period += sample->period; |
110 | return 0; | ||
111 | } | ||
112 | |||
113 | static u64 nr_lost; | ||
114 | |||
115 | static int process_lost_event(event_t *event, struct perf_session *session __used) | ||
116 | { | ||
117 | nr_lost += event->lost.lost; | ||
118 | |||
119 | return 0; | 99 | return 0; |
120 | } | 100 | } |
121 | 101 | ||
@@ -126,7 +106,7 @@ static struct perf_event_ops event_ops = { | |||
126 | .event_type = event__process_event_type, | 106 | .event_type = event__process_event_type, |
127 | .tracing_data = event__process_tracing_data, | 107 | .tracing_data = event__process_tracing_data, |
128 | .build_id = event__process_build_id, | 108 | .build_id = event__process_build_id, |
129 | .lost = process_lost_event, | 109 | .ordering_requires_timestamps = true, |
130 | .ordered_samples = true, | 110 | .ordered_samples = true, |
131 | }; | 111 | }; |
132 | 112 | ||
@@ -137,7 +117,7 @@ static void sig_handler(int sig __unused) | |||
137 | session_done = 1; | 117 | session_done = 1; |
138 | } | 118 | } |
139 | 119 | ||
140 | static int __cmd_trace(struct perf_session *session) | 120 | static int __cmd_script(struct perf_session *session) |
141 | { | 121 | { |
142 | int ret; | 122 | int ret; |
143 | 123 | ||
@@ -145,10 +125,8 @@ static int __cmd_trace(struct perf_session *session) | |||
145 | 125 | ||
146 | ret = perf_session__process_events(session, &event_ops); | 126 | ret = perf_session__process_events(session, &event_ops); |
147 | 127 | ||
148 | if (debug_mode) { | 128 | if (debug_mode) |
149 | pr_err("Misordered timestamps: %llu\n", nr_unordered); | 129 | pr_err("Misordered timestamps: %llu\n", nr_unordered); |
150 | pr_err("Lost events: %llu\n", nr_lost); | ||
151 | } | ||
152 | 130 | ||
153 | return ret; | 131 | return ret; |
154 | } | 132 | } |
@@ -159,7 +137,7 @@ struct script_spec { | |||
159 | char spec[0]; | 137 | char spec[0]; |
160 | }; | 138 | }; |
161 | 139 | ||
162 | LIST_HEAD(script_specs); | 140 | static LIST_HEAD(script_specs); |
163 | 141 | ||
164 | static struct script_spec *script_spec__new(const char *spec, | 142 | static struct script_spec *script_spec__new(const char *spec, |
165 | struct scripting_ops *ops) | 143 | struct scripting_ops *ops) |
@@ -247,7 +225,7 @@ static void list_available_languages(void) | |||
247 | 225 | ||
248 | fprintf(stderr, "\n"); | 226 | fprintf(stderr, "\n"); |
249 | fprintf(stderr, "Scripting language extensions (used in " | 227 | fprintf(stderr, "Scripting language extensions (used in " |
250 | "perf trace -s [spec:]script.[spec]):\n\n"); | 228 | "perf script -s [spec:]script.[spec]):\n\n"); |
251 | 229 | ||
252 | list_for_each_entry(s, &script_specs, node) | 230 | list_for_each_entry(s, &script_specs, node) |
253 | fprintf(stderr, " %-42s [%s]\n", s->spec, s->ops->name); | 231 | fprintf(stderr, " %-42s [%s]\n", s->spec, s->ops->name); |
@@ -301,17 +279,34 @@ static int parse_scriptname(const struct option *opt __used, | |||
301 | return 0; | 279 | return 0; |
302 | } | 280 | } |
303 | 281 | ||
304 | #define for_each_lang(scripts_dir, lang_dirent, lang_next) \ | 282 | /* Helper function for filesystems that return a dent->d_type DT_UNKNOWN */ |
283 | static int is_directory(const char *base_path, const struct dirent *dent) | ||
284 | { | ||
285 | char path[PATH_MAX]; | ||
286 | struct stat st; | ||
287 | |||
288 | sprintf(path, "%s/%s", base_path, dent->d_name); | ||
289 | if (stat(path, &st)) | ||
290 | return 0; | ||
291 | |||
292 | return S_ISDIR(st.st_mode); | ||
293 | } | ||
294 | |||
295 | #define for_each_lang(scripts_path, scripts_dir, lang_dirent, lang_next)\ | ||
305 | while (!readdir_r(scripts_dir, &lang_dirent, &lang_next) && \ | 296 | while (!readdir_r(scripts_dir, &lang_dirent, &lang_next) && \ |
306 | lang_next) \ | 297 | lang_next) \ |
307 | if (lang_dirent.d_type == DT_DIR && \ | 298 | if ((lang_dirent.d_type == DT_DIR || \ |
299 | (lang_dirent.d_type == DT_UNKNOWN && \ | ||
300 | is_directory(scripts_path, &lang_dirent))) && \ | ||
308 | (strcmp(lang_dirent.d_name, ".")) && \ | 301 | (strcmp(lang_dirent.d_name, ".")) && \ |
309 | (strcmp(lang_dirent.d_name, ".."))) | 302 | (strcmp(lang_dirent.d_name, ".."))) |
310 | 303 | ||
311 | #define for_each_script(lang_dir, script_dirent, script_next) \ | 304 | #define for_each_script(lang_path, lang_dir, script_dirent, script_next)\ |
312 | while (!readdir_r(lang_dir, &script_dirent, &script_next) && \ | 305 | while (!readdir_r(lang_dir, &script_dirent, &script_next) && \ |
313 | script_next) \ | 306 | script_next) \ |
314 | if (script_dirent.d_type != DT_DIR) | 307 | if (script_dirent.d_type != DT_DIR && \ |
308 | (script_dirent.d_type != DT_UNKNOWN || \ | ||
309 | !is_directory(lang_path, &script_dirent))) | ||
315 | 310 | ||
316 | 311 | ||
317 | #define RECORD_SUFFIX "-record" | 312 | #define RECORD_SUFFIX "-record" |
@@ -324,7 +319,7 @@ struct script_desc { | |||
324 | char *args; | 319 | char *args; |
325 | }; | 320 | }; |
326 | 321 | ||
327 | LIST_HEAD(script_descs); | 322 | static LIST_HEAD(script_descs); |
328 | 323 | ||
329 | static struct script_desc *script_desc__new(const char *name) | 324 | static struct script_desc *script_desc__new(const char *name) |
330 | { | 325 | { |
@@ -380,10 +375,10 @@ out_delete_desc: | |||
380 | return NULL; | 375 | return NULL; |
381 | } | 376 | } |
382 | 377 | ||
383 | static char *ends_with(char *str, const char *suffix) | 378 | static const char *ends_with(const char *str, const char *suffix) |
384 | { | 379 | { |
385 | size_t suffix_len = strlen(suffix); | 380 | size_t suffix_len = strlen(suffix); |
386 | char *p = str; | 381 | const char *p = str; |
387 | 382 | ||
388 | if (strlen(str) > suffix_len) { | 383 | if (strlen(str) > suffix_len) { |
389 | p = str + strlen(str) - suffix_len; | 384 | p = str + strlen(str) - suffix_len; |
@@ -466,16 +461,16 @@ static int list_available_scripts(const struct option *opt __used, | |||
466 | if (!scripts_dir) | 461 | if (!scripts_dir) |
467 | return -1; | 462 | return -1; |
468 | 463 | ||
469 | for_each_lang(scripts_dir, lang_dirent, lang_next) { | 464 | for_each_lang(scripts_path, scripts_dir, lang_dirent, lang_next) { |
470 | snprintf(lang_path, MAXPATHLEN, "%s/%s/bin", scripts_path, | 465 | snprintf(lang_path, MAXPATHLEN, "%s/%s/bin", scripts_path, |
471 | lang_dirent.d_name); | 466 | lang_dirent.d_name); |
472 | lang_dir = opendir(lang_path); | 467 | lang_dir = opendir(lang_path); |
473 | if (!lang_dir) | 468 | if (!lang_dir) |
474 | continue; | 469 | continue; |
475 | 470 | ||
476 | for_each_script(lang_dir, script_dirent, script_next) { | 471 | for_each_script(lang_path, lang_dir, script_dirent, script_next) { |
477 | script_root = strdup(script_dirent.d_name); | 472 | script_root = strdup(script_dirent.d_name); |
478 | str = ends_with(script_root, REPORT_SUFFIX); | 473 | str = (char *)ends_with(script_root, REPORT_SUFFIX); |
479 | if (str) { | 474 | if (str) { |
480 | *str = '\0'; | 475 | *str = '\0'; |
481 | desc = script_desc__findnew(script_root); | 476 | desc = script_desc__findnew(script_root); |
@@ -514,16 +509,16 @@ static char *get_script_path(const char *script_root, const char *suffix) | |||
514 | if (!scripts_dir) | 509 | if (!scripts_dir) |
515 | return NULL; | 510 | return NULL; |
516 | 511 | ||
517 | for_each_lang(scripts_dir, lang_dirent, lang_next) { | 512 | for_each_lang(scripts_path, scripts_dir, lang_dirent, lang_next) { |
518 | snprintf(lang_path, MAXPATHLEN, "%s/%s/bin", scripts_path, | 513 | snprintf(lang_path, MAXPATHLEN, "%s/%s/bin", scripts_path, |
519 | lang_dirent.d_name); | 514 | lang_dirent.d_name); |
520 | lang_dir = opendir(lang_path); | 515 | lang_dir = opendir(lang_path); |
521 | if (!lang_dir) | 516 | if (!lang_dir) |
522 | continue; | 517 | continue; |
523 | 518 | ||
524 | for_each_script(lang_dir, script_dirent, script_next) { | 519 | for_each_script(lang_path, lang_dir, script_dirent, script_next) { |
525 | __script_root = strdup(script_dirent.d_name); | 520 | __script_root = strdup(script_dirent.d_name); |
526 | str = ends_with(__script_root, suffix); | 521 | str = (char *)ends_with(__script_root, suffix); |
527 | if (str) { | 522 | if (str) { |
528 | *str = '\0'; | 523 | *str = '\0'; |
529 | if (strcmp(__script_root, script_root)) | 524 | if (strcmp(__script_root, script_root)) |
@@ -543,7 +538,7 @@ static char *get_script_path(const char *script_root, const char *suffix) | |||
543 | 538 | ||
544 | static bool is_top_script(const char *script_path) | 539 | static bool is_top_script(const char *script_path) |
545 | { | 540 | { |
546 | return ends_with((char *)script_path, "top") == NULL ? false : true; | 541 | return ends_with(script_path, "top") == NULL ? false : true; |
547 | } | 542 | } |
548 | 543 | ||
549 | static int has_required_arg(char *script_path) | 544 | static int has_required_arg(char *script_path) |
@@ -569,12 +564,12 @@ out: | |||
569 | return n_args; | 564 | return n_args; |
570 | } | 565 | } |
571 | 566 | ||
572 | static const char * const trace_usage[] = { | 567 | static const char * const script_usage[] = { |
573 | "perf trace [<options>]", | 568 | "perf script [<options>]", |
574 | "perf trace [<options>] record <script> [<record-options>] <command>", | 569 | "perf script [<options>] record <script> [<record-options>] <command>", |
575 | "perf trace [<options>] report <script> [script-args]", | 570 | "perf script [<options>] report <script> [script-args]", |
576 | "perf trace [<options>] <script> [<record-options>] <command>", | 571 | "perf script [<options>] <script> [<record-options>] <command>", |
577 | "perf trace [<options>] <top-script> [script-args]", | 572 | "perf script [<options>] <top-script> [script-args]", |
578 | NULL | 573 | NULL |
579 | }; | 574 | }; |
580 | 575 | ||
@@ -591,7 +586,7 @@ static const struct option options[] = { | |||
591 | "script file name (lang:script name, script name, or *)", | 586 | "script file name (lang:script name, script name, or *)", |
592 | parse_scriptname), | 587 | parse_scriptname), |
593 | OPT_STRING('g', "gen-script", &generate_script_lang, "lang", | 588 | OPT_STRING('g', "gen-script", &generate_script_lang, "lang", |
594 | "generate perf-trace.xx script in specified language"), | 589 | "generate perf-script.xx script in specified language"), |
595 | OPT_STRING('i', "input", &input_name, "file", | 590 | OPT_STRING('i', "input", &input_name, "file", |
596 | "input file name"), | 591 | "input file name"), |
597 | OPT_BOOLEAN('d', "debug-mode", &debug_mode, | 592 | OPT_BOOLEAN('d', "debug-mode", &debug_mode, |
@@ -614,7 +609,7 @@ static bool have_cmd(int argc, const char **argv) | |||
614 | return argc != 0; | 609 | return argc != 0; |
615 | } | 610 | } |
616 | 611 | ||
617 | int cmd_trace(int argc, const char **argv, const char *prefix __used) | 612 | int cmd_script(int argc, const char **argv, const char *prefix __used) |
618 | { | 613 | { |
619 | char *rec_script_path = NULL; | 614 | char *rec_script_path = NULL; |
620 | char *rep_script_path = NULL; | 615 | char *rep_script_path = NULL; |
@@ -626,7 +621,7 @@ int cmd_trace(int argc, const char **argv, const char *prefix __used) | |||
626 | 621 | ||
627 | setup_scripting(); | 622 | setup_scripting(); |
628 | 623 | ||
629 | argc = parse_options(argc, argv, options, trace_usage, | 624 | argc = parse_options(argc, argv, options, script_usage, |
630 | PARSE_OPT_STOP_AT_NON_OPTION); | 625 | PARSE_OPT_STOP_AT_NON_OPTION); |
631 | 626 | ||
632 | if (argc > 1 && !strncmp(argv[0], "rec", strlen("rec"))) { | 627 | if (argc > 1 && !strncmp(argv[0], "rec", strlen("rec"))) { |
@@ -640,7 +635,7 @@ int cmd_trace(int argc, const char **argv, const char *prefix __used) | |||
640 | if (!rep_script_path) { | 635 | if (!rep_script_path) { |
641 | fprintf(stderr, | 636 | fprintf(stderr, |
642 | "Please specify a valid report script" | 637 | "Please specify a valid report script" |
643 | "(see 'perf trace -l' for listing)\n"); | 638 | "(see 'perf script -l' for listing)\n"); |
644 | return -1; | 639 | return -1; |
645 | } | 640 | } |
646 | } | 641 | } |
@@ -658,8 +653,8 @@ int cmd_trace(int argc, const char **argv, const char *prefix __used) | |||
658 | 653 | ||
659 | if (!rec_script_path && !rep_script_path) { | 654 | if (!rec_script_path && !rep_script_path) { |
660 | fprintf(stderr, " Couldn't find script %s\n\n See perf" | 655 | fprintf(stderr, " Couldn't find script %s\n\n See perf" |
661 | " trace -l for available scripts.\n", argv[0]); | 656 | " script -l for available scripts.\n", argv[0]); |
662 | usage_with_options(trace_usage, options); | 657 | usage_with_options(script_usage, options); |
663 | } | 658 | } |
664 | 659 | ||
665 | if (is_top_script(argv[0])) { | 660 | if (is_top_script(argv[0])) { |
@@ -671,9 +666,9 @@ int cmd_trace(int argc, const char **argv, const char *prefix __used) | |||
671 | rec_args = (argc - 1) - rep_args; | 666 | rec_args = (argc - 1) - rep_args; |
672 | if (rec_args < 0) { | 667 | if (rec_args < 0) { |
673 | fprintf(stderr, " %s script requires options." | 668 | fprintf(stderr, " %s script requires options." |
674 | "\n\n See perf trace -l for available " | 669 | "\n\n See perf script -l for available " |
675 | "scripts and options.\n", argv[0]); | 670 | "scripts and options.\n", argv[0]); |
676 | usage_with_options(trace_usage, options); | 671 | usage_with_options(script_usage, options); |
677 | } | 672 | } |
678 | } | 673 | } |
679 | 674 | ||
@@ -772,7 +767,7 @@ int cmd_trace(int argc, const char **argv, const char *prefix __used) | |||
772 | if (!script_name) | 767 | if (!script_name) |
773 | setup_pager(); | 768 | setup_pager(); |
774 | 769 | ||
775 | session = perf_session__new(input_name, O_RDONLY, 0, false); | 770 | session = perf_session__new(input_name, O_RDONLY, 0, false, &event_ops); |
776 | if (session == NULL) | 771 | if (session == NULL) |
777 | return -ENOMEM; | 772 | return -ENOMEM; |
778 | 773 | ||
@@ -806,7 +801,7 @@ int cmd_trace(int argc, const char **argv, const char *prefix __used) | |||
806 | return -1; | 801 | return -1; |
807 | } | 802 | } |
808 | 803 | ||
809 | err = scripting_ops->generate_script("perf-trace"); | 804 | err = scripting_ops->generate_script("perf-script"); |
810 | goto out; | 805 | goto out; |
811 | } | 806 | } |
812 | 807 | ||
@@ -814,10 +809,10 @@ int cmd_trace(int argc, const char **argv, const char *prefix __used) | |||
814 | err = scripting_ops->start_script(script_name, argc, argv); | 809 | err = scripting_ops->start_script(script_name, argc, argv); |
815 | if (err) | 810 | if (err) |
816 | goto out; | 811 | goto out; |
817 | pr_debug("perf trace started with script %s\n\n", script_name); | 812 | pr_debug("perf script started with script %s\n\n", script_name); |
818 | } | 813 | } |
819 | 814 | ||
820 | err = __cmd_trace(session); | 815 | err = __cmd_script(session); |
821 | 816 | ||
822 | perf_session__delete(session); | 817 | perf_session__delete(session); |
823 | cleanup_scripting(); | 818 | cleanup_scripting(); |
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index a6b4d44f950..02b2d8013a6 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c | |||
@@ -43,6 +43,7 @@ | |||
43 | #include "util/parse-options.h" | 43 | #include "util/parse-options.h" |
44 | #include "util/parse-events.h" | 44 | #include "util/parse-events.h" |
45 | #include "util/event.h" | 45 | #include "util/event.h" |
46 | #include "util/evsel.h" | ||
46 | #include "util/debug.h" | 47 | #include "util/debug.h" |
47 | #include "util/header.h" | 48 | #include "util/header.h" |
48 | #include "util/cpumap.h" | 49 | #include "util/cpumap.h" |
@@ -52,6 +53,8 @@ | |||
52 | #include <math.h> | 53 | #include <math.h> |
53 | #include <locale.h> | 54 | #include <locale.h> |
54 | 55 | ||
56 | #define DEFAULT_SEPARATOR " " | ||
57 | |||
55 | static struct perf_event_attr default_attrs[] = { | 58 | static struct perf_event_attr default_attrs[] = { |
56 | 59 | ||
57 | { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_TASK_CLOCK }, | 60 | { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_TASK_CLOCK }, |
@@ -69,25 +72,23 @@ static struct perf_event_attr default_attrs[] = { | |||
69 | }; | 72 | }; |
70 | 73 | ||
71 | static bool system_wide = false; | 74 | static bool system_wide = false; |
72 | static int nr_cpus = 0; | 75 | static struct cpu_map *cpus; |
73 | static int run_idx = 0; | 76 | static int run_idx = 0; |
74 | 77 | ||
75 | static int run_count = 1; | 78 | static int run_count = 1; |
76 | static bool no_inherit = false; | 79 | static bool no_inherit = false; |
77 | static bool scale = true; | 80 | static bool scale = true; |
81 | static bool no_aggr = false; | ||
78 | static pid_t target_pid = -1; | 82 | static pid_t target_pid = -1; |
79 | static pid_t target_tid = -1; | 83 | static pid_t target_tid = -1; |
80 | static pid_t *all_tids = NULL; | 84 | static struct thread_map *threads; |
81 | static int thread_num = 0; | ||
82 | static pid_t child_pid = -1; | 85 | static pid_t child_pid = -1; |
83 | static bool null_run = false; | 86 | static bool null_run = false; |
84 | static bool big_num = false; | 87 | static bool big_num = true; |
88 | static int big_num_opt = -1; | ||
85 | static const char *cpu_list; | 89 | static const char *cpu_list; |
86 | 90 | static const char *csv_sep = NULL; | |
87 | 91 | static bool csv_output = false; | |
88 | static int *fd[MAX_NR_CPUS][MAX_COUNTERS]; | ||
89 | |||
90 | static int event_scaled[MAX_COUNTERS]; | ||
91 | 92 | ||
92 | static volatile int done = 0; | 93 | static volatile int done = 0; |
93 | 94 | ||
@@ -96,6 +97,22 @@ struct stats | |||
96 | double n, mean, M2; | 97 | double n, mean, M2; |
97 | }; | 98 | }; |
98 | 99 | ||
100 | struct perf_stat { | ||
101 | struct stats res_stats[3]; | ||
102 | }; | ||
103 | |||
104 | static int perf_evsel__alloc_stat_priv(struct perf_evsel *evsel) | ||
105 | { | ||
106 | evsel->priv = zalloc(sizeof(struct perf_stat)); | ||
107 | return evsel->priv == NULL ? -ENOMEM : 0; | ||
108 | } | ||
109 | |||
110 | static void perf_evsel__free_stat_priv(struct perf_evsel *evsel) | ||
111 | { | ||
112 | free(evsel->priv); | ||
113 | evsel->priv = NULL; | ||
114 | } | ||
115 | |||
99 | static void update_stats(struct stats *stats, u64 val) | 116 | static void update_stats(struct stats *stats, u64 val) |
100 | { | 117 | { |
101 | double delta; | 118 | double delta; |
@@ -135,69 +152,38 @@ static double stddev_stats(struct stats *stats) | |||
135 | return sqrt(variance_mean); | 152 | return sqrt(variance_mean); |
136 | } | 153 | } |
137 | 154 | ||
138 | struct stats event_res_stats[MAX_COUNTERS][3]; | 155 | struct stats runtime_nsecs_stats[MAX_NR_CPUS]; |
139 | struct stats runtime_nsecs_stats; | 156 | struct stats runtime_cycles_stats[MAX_NR_CPUS]; |
157 | struct stats runtime_branches_stats[MAX_NR_CPUS]; | ||
140 | struct stats walltime_nsecs_stats; | 158 | struct stats walltime_nsecs_stats; |
141 | struct stats runtime_cycles_stats; | ||
142 | struct stats runtime_branches_stats; | ||
143 | 159 | ||
144 | #define MATCH_EVENT(t, c, counter) \ | 160 | static int create_perf_stat_counter(struct perf_evsel *evsel) |
145 | (attrs[counter].type == PERF_TYPE_##t && \ | ||
146 | attrs[counter].config == PERF_COUNT_##c) | ||
147 | |||
148 | #define ERR_PERF_OPEN \ | ||
149 | "Error: counter %d, sys_perf_event_open() syscall returned with %d (%s)\n" | ||
150 | |||
151 | static int create_perf_stat_counter(int counter) | ||
152 | { | 161 | { |
153 | struct perf_event_attr *attr = attrs + counter; | 162 | struct perf_event_attr *attr = &evsel->attr; |
154 | int thread; | ||
155 | int ncreated = 0; | ||
156 | 163 | ||
157 | if (scale) | 164 | if (scale) |
158 | attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED | | 165 | attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED | |
159 | PERF_FORMAT_TOTAL_TIME_RUNNING; | 166 | PERF_FORMAT_TOTAL_TIME_RUNNING; |
160 | 167 | ||
161 | if (system_wide) { | 168 | if (system_wide) |
162 | int cpu; | 169 | return perf_evsel__open_per_cpu(evsel, cpus); |
163 | 170 | ||
164 | for (cpu = 0; cpu < nr_cpus; cpu++) { | 171 | attr->inherit = !no_inherit; |
165 | fd[cpu][counter][0] = sys_perf_event_open(attr, | 172 | if (target_pid == -1 && target_tid == -1) { |
166 | -1, cpumap[cpu], -1, 0); | 173 | attr->disabled = 1; |
167 | if (fd[cpu][counter][0] < 0) | 174 | attr->enable_on_exec = 1; |
168 | pr_debug(ERR_PERF_OPEN, counter, | ||
169 | fd[cpu][counter][0], strerror(errno)); | ||
170 | else | ||
171 | ++ncreated; | ||
172 | } | ||
173 | } else { | ||
174 | attr->inherit = !no_inherit; | ||
175 | if (target_pid == -1 && target_tid == -1) { | ||
176 | attr->disabled = 1; | ||
177 | attr->enable_on_exec = 1; | ||
178 | } | ||
179 | for (thread = 0; thread < thread_num; thread++) { | ||
180 | fd[0][counter][thread] = sys_perf_event_open(attr, | ||
181 | all_tids[thread], -1, -1, 0); | ||
182 | if (fd[0][counter][thread] < 0) | ||
183 | pr_debug(ERR_PERF_OPEN, counter, | ||
184 | fd[0][counter][thread], | ||
185 | strerror(errno)); | ||
186 | else | ||
187 | ++ncreated; | ||
188 | } | ||
189 | } | 175 | } |
190 | 176 | ||
191 | return ncreated; | 177 | return perf_evsel__open_per_thread(evsel, threads); |
192 | } | 178 | } |
193 | 179 | ||
194 | /* | 180 | /* |
195 | * Does the counter have nsecs as a unit? | 181 | * Does the counter have nsecs as a unit? |
196 | */ | 182 | */ |
197 | static inline int nsec_counter(int counter) | 183 | static inline int nsec_counter(struct perf_evsel *evsel) |
198 | { | 184 | { |
199 | if (MATCH_EVENT(SOFTWARE, SW_CPU_CLOCK, counter) || | 185 | if (perf_evsel__match(evsel, SOFTWARE, SW_CPU_CLOCK) || |
200 | MATCH_EVENT(SOFTWARE, SW_TASK_CLOCK, counter)) | 186 | perf_evsel__match(evsel, SOFTWARE, SW_TASK_CLOCK)) |
201 | return 1; | 187 | return 1; |
202 | 188 | ||
203 | return 0; | 189 | return 0; |
@@ -205,55 +191,19 @@ static inline int nsec_counter(int counter) | |||
205 | 191 | ||
206 | /* | 192 | /* |
207 | * Read out the results of a single counter: | 193 | * Read out the results of a single counter: |
194 | * aggregate counts across CPUs in system-wide mode | ||
208 | */ | 195 | */ |
209 | static void read_counter(int counter) | 196 | static int read_counter_aggr(struct perf_evsel *counter) |
210 | { | 197 | { |
211 | u64 count[3], single_count[3]; | 198 | struct perf_stat *ps = counter->priv; |
212 | int cpu; | 199 | u64 *count = counter->counts->aggr.values; |
213 | size_t res, nv; | 200 | int i; |
214 | int scaled; | ||
215 | int i, thread; | ||
216 | |||
217 | count[0] = count[1] = count[2] = 0; | ||
218 | |||
219 | nv = scale ? 3 : 1; | ||
220 | for (cpu = 0; cpu < nr_cpus; cpu++) { | ||
221 | for (thread = 0; thread < thread_num; thread++) { | ||
222 | if (fd[cpu][counter][thread] < 0) | ||
223 | continue; | ||
224 | |||
225 | res = read(fd[cpu][counter][thread], | ||
226 | single_count, nv * sizeof(u64)); | ||
227 | assert(res == nv * sizeof(u64)); | ||
228 | |||
229 | close(fd[cpu][counter][thread]); | ||
230 | fd[cpu][counter][thread] = -1; | ||
231 | |||
232 | count[0] += single_count[0]; | ||
233 | if (scale) { | ||
234 | count[1] += single_count[1]; | ||
235 | count[2] += single_count[2]; | ||
236 | } | ||
237 | } | ||
238 | } | ||
239 | |||
240 | scaled = 0; | ||
241 | if (scale) { | ||
242 | if (count[2] == 0) { | ||
243 | event_scaled[counter] = -1; | ||
244 | count[0] = 0; | ||
245 | return; | ||
246 | } | ||
247 | 201 | ||
248 | if (count[2] < count[1]) { | 202 | if (__perf_evsel__read(counter, cpus->nr, threads->nr, scale) < 0) |
249 | event_scaled[counter] = 1; | 203 | return -1; |
250 | count[0] = (unsigned long long) | ||
251 | ((double)count[0] * count[1] / count[2] + 0.5); | ||
252 | } | ||
253 | } | ||
254 | 204 | ||
255 | for (i = 0; i < 3; i++) | 205 | for (i = 0; i < 3; i++) |
256 | update_stats(&event_res_stats[counter][i], count[i]); | 206 | update_stats(&ps->res_stats[i], count[i]); |
257 | 207 | ||
258 | if (verbose) { | 208 | if (verbose) { |
259 | fprintf(stderr, "%s: %Ld %Ld %Ld\n", event_name(counter), | 209 | fprintf(stderr, "%s: %Ld %Ld %Ld\n", event_name(counter), |
@@ -263,26 +213,51 @@ static void read_counter(int counter) | |||
263 | /* | 213 | /* |
264 | * Save the full runtime - to allow normalization during printout: | 214 | * Save the full runtime - to allow normalization during printout: |
265 | */ | 215 | */ |
266 | if (MATCH_EVENT(SOFTWARE, SW_TASK_CLOCK, counter)) | 216 | if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK)) |
267 | update_stats(&runtime_nsecs_stats, count[0]); | 217 | update_stats(&runtime_nsecs_stats[0], count[0]); |
268 | if (MATCH_EVENT(HARDWARE, HW_CPU_CYCLES, counter)) | 218 | if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES)) |
269 | update_stats(&runtime_cycles_stats, count[0]); | 219 | update_stats(&runtime_cycles_stats[0], count[0]); |
270 | if (MATCH_EVENT(HARDWARE, HW_BRANCH_INSTRUCTIONS, counter)) | 220 | if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS)) |
271 | update_stats(&runtime_branches_stats, count[0]); | 221 | update_stats(&runtime_branches_stats[0], count[0]); |
222 | |||
223 | return 0; | ||
224 | } | ||
225 | |||
226 | /* | ||
227 | * Read out the results of a single counter: | ||
228 | * do not aggregate counts across CPUs in system-wide mode | ||
229 | */ | ||
230 | static int read_counter(struct perf_evsel *counter) | ||
231 | { | ||
232 | u64 *count; | ||
233 | int cpu; | ||
234 | |||
235 | for (cpu = 0; cpu < cpus->nr; cpu++) { | ||
236 | if (__perf_evsel__read_on_cpu(counter, cpu, 0, scale) < 0) | ||
237 | return -1; | ||
238 | |||
239 | count = counter->counts->cpu[cpu].values; | ||
240 | |||
241 | if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK)) | ||
242 | update_stats(&runtime_nsecs_stats[cpu], count[0]); | ||
243 | if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES)) | ||
244 | update_stats(&runtime_cycles_stats[cpu], count[0]); | ||
245 | if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS)) | ||
246 | update_stats(&runtime_branches_stats[cpu], count[0]); | ||
247 | } | ||
248 | |||
249 | return 0; | ||
272 | } | 250 | } |
273 | 251 | ||
274 | static int run_perf_stat(int argc __used, const char **argv) | 252 | static int run_perf_stat(int argc __used, const char **argv) |
275 | { | 253 | { |
276 | unsigned long long t0, t1; | 254 | unsigned long long t0, t1; |
255 | struct perf_evsel *counter; | ||
277 | int status = 0; | 256 | int status = 0; |
278 | int counter, ncreated = 0; | ||
279 | int child_ready_pipe[2], go_pipe[2]; | 257 | int child_ready_pipe[2], go_pipe[2]; |
280 | const bool forks = (argc > 0); | 258 | const bool forks = (argc > 0); |
281 | char buf; | 259 | char buf; |
282 | 260 | ||
283 | if (!system_wide) | ||
284 | nr_cpus = 1; | ||
285 | |||
286 | if (forks && (pipe(child_ready_pipe) < 0 || pipe(go_pipe) < 0)) { | 261 | if (forks && (pipe(child_ready_pipe) < 0 || pipe(go_pipe) < 0)) { |
287 | perror("failed to create pipes"); | 262 | perror("failed to create pipes"); |
288 | exit(1); | 263 | exit(1); |
@@ -322,7 +297,7 @@ static int run_perf_stat(int argc __used, const char **argv) | |||
322 | } | 297 | } |
323 | 298 | ||
324 | if (target_tid == -1 && target_pid == -1 && !system_wide) | 299 | if (target_tid == -1 && target_pid == -1 && !system_wide) |
325 | all_tids[0] = child_pid; | 300 | threads->map[0] = child_pid; |
326 | 301 | ||
327 | /* | 302 | /* |
328 | * Wait for the child to be ready to exec. | 303 | * Wait for the child to be ready to exec. |
@@ -334,16 +309,23 @@ static int run_perf_stat(int argc __used, const char **argv) | |||
334 | close(child_ready_pipe[0]); | 309 | close(child_ready_pipe[0]); |
335 | } | 310 | } |
336 | 311 | ||
337 | for (counter = 0; counter < nr_counters; counter++) | 312 | list_for_each_entry(counter, &evsel_list, node) { |
338 | ncreated += create_perf_stat_counter(counter); | 313 | if (create_perf_stat_counter(counter) < 0) { |
339 | 314 | if (errno == -EPERM || errno == -EACCES) { | |
340 | if (ncreated == 0) { | 315 | error("You may not have permission to collect %sstats.\n" |
341 | pr_err("No permission to collect %sstats.\n" | 316 | "\t Consider tweaking" |
342 | "Consider tweaking /proc/sys/kernel/perf_event_paranoid.\n", | 317 | " /proc/sys/kernel/perf_event_paranoid or running as root.", |
343 | system_wide ? "system-wide " : ""); | 318 | system_wide ? "system-wide " : ""); |
344 | if (child_pid != -1) | 319 | } else { |
345 | kill(child_pid, SIGTERM); | 320 | error("open_counter returned with %d (%s). " |
346 | return -1; | 321 | "/bin/dmesg may provide additional information.\n", |
322 | errno, strerror(errno)); | ||
323 | } | ||
324 | if (child_pid != -1) | ||
325 | kill(child_pid, SIGTERM); | ||
326 | die("Not all events could be opened.\n"); | ||
327 | return -1; | ||
328 | } | ||
347 | } | 329 | } |
348 | 330 | ||
349 | /* | 331 | /* |
@@ -362,60 +344,97 @@ static int run_perf_stat(int argc __used, const char **argv) | |||
362 | 344 | ||
363 | update_stats(&walltime_nsecs_stats, t1 - t0); | 345 | update_stats(&walltime_nsecs_stats, t1 - t0); |
364 | 346 | ||
365 | for (counter = 0; counter < nr_counters; counter++) | 347 | if (no_aggr) { |
366 | read_counter(counter); | 348 | list_for_each_entry(counter, &evsel_list, node) { |
349 | read_counter(counter); | ||
350 | perf_evsel__close_fd(counter, cpus->nr, 1); | ||
351 | } | ||
352 | } else { | ||
353 | list_for_each_entry(counter, &evsel_list, node) { | ||
354 | read_counter_aggr(counter); | ||
355 | perf_evsel__close_fd(counter, cpus->nr, threads->nr); | ||
356 | } | ||
357 | } | ||
367 | 358 | ||
368 | return WEXITSTATUS(status); | 359 | return WEXITSTATUS(status); |
369 | } | 360 | } |
370 | 361 | ||
371 | static void print_noise(int counter, double avg) | 362 | static void print_noise(struct perf_evsel *evsel, double avg) |
372 | { | 363 | { |
364 | struct perf_stat *ps; | ||
365 | |||
373 | if (run_count == 1) | 366 | if (run_count == 1) |
374 | return; | 367 | return; |
375 | 368 | ||
369 | ps = evsel->priv; | ||
376 | fprintf(stderr, " ( +- %7.3f%% )", | 370 | fprintf(stderr, " ( +- %7.3f%% )", |
377 | 100 * stddev_stats(&event_res_stats[counter][0]) / avg); | 371 | 100 * stddev_stats(&ps->res_stats[0]) / avg); |
378 | } | 372 | } |
379 | 373 | ||
380 | static void nsec_printout(int counter, double avg) | 374 | static void nsec_printout(int cpu, struct perf_evsel *evsel, double avg) |
381 | { | 375 | { |
382 | double msecs = avg / 1e6; | 376 | double msecs = avg / 1e6; |
377 | char cpustr[16] = { '\0', }; | ||
378 | const char *fmt = csv_output ? "%s%.6f%s%s" : "%s%18.6f%s%-24s"; | ||
383 | 379 | ||
384 | fprintf(stderr, " %18.6f %-24s", msecs, event_name(counter)); | 380 | if (no_aggr) |
381 | sprintf(cpustr, "CPU%*d%s", | ||
382 | csv_output ? 0 : -4, | ||
383 | cpus->map[cpu], csv_sep); | ||
384 | |||
385 | fprintf(stderr, fmt, cpustr, msecs, csv_sep, event_name(evsel)); | ||
386 | |||
387 | if (csv_output) | ||
388 | return; | ||
385 | 389 | ||
386 | if (MATCH_EVENT(SOFTWARE, SW_TASK_CLOCK, counter)) { | 390 | if (perf_evsel__match(evsel, SOFTWARE, SW_TASK_CLOCK)) |
387 | fprintf(stderr, " # %10.3f CPUs ", | 391 | fprintf(stderr, " # %10.3f CPUs ", |
388 | avg / avg_stats(&walltime_nsecs_stats)); | 392 | avg / avg_stats(&walltime_nsecs_stats)); |
389 | } | ||
390 | } | 393 | } |
391 | 394 | ||
392 | static void abs_printout(int counter, double avg) | 395 | static void abs_printout(int cpu, struct perf_evsel *evsel, double avg) |
393 | { | 396 | { |
394 | double total, ratio = 0.0; | 397 | double total, ratio = 0.0; |
398 | char cpustr[16] = { '\0', }; | ||
399 | const char *fmt; | ||
400 | |||
401 | if (csv_output) | ||
402 | fmt = "%s%.0f%s%s"; | ||
403 | else if (big_num) | ||
404 | fmt = "%s%'18.0f%s%-24s"; | ||
405 | else | ||
406 | fmt = "%s%18.0f%s%-24s"; | ||
395 | 407 | ||
396 | if (big_num) | 408 | if (no_aggr) |
397 | fprintf(stderr, " %'18.0f %-24s", avg, event_name(counter)); | 409 | sprintf(cpustr, "CPU%*d%s", |
410 | csv_output ? 0 : -4, | ||
411 | cpus->map[cpu], csv_sep); | ||
398 | else | 412 | else |
399 | fprintf(stderr, " %18.0f %-24s", avg, event_name(counter)); | 413 | cpu = 0; |
414 | |||
415 | fprintf(stderr, fmt, cpustr, avg, csv_sep, event_name(evsel)); | ||
400 | 416 | ||
401 | if (MATCH_EVENT(HARDWARE, HW_INSTRUCTIONS, counter)) { | 417 | if (csv_output) |
402 | total = avg_stats(&runtime_cycles_stats); | 418 | return; |
419 | |||
420 | if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) { | ||
421 | total = avg_stats(&runtime_cycles_stats[cpu]); | ||
403 | 422 | ||
404 | if (total) | 423 | if (total) |
405 | ratio = avg / total; | 424 | ratio = avg / total; |
406 | 425 | ||
407 | fprintf(stderr, " # %10.3f IPC ", ratio); | 426 | fprintf(stderr, " # %10.3f IPC ", ratio); |
408 | } else if (MATCH_EVENT(HARDWARE, HW_BRANCH_MISSES, counter) && | 427 | } else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES) && |
409 | runtime_branches_stats.n != 0) { | 428 | runtime_branches_stats[cpu].n != 0) { |
410 | total = avg_stats(&runtime_branches_stats); | 429 | total = avg_stats(&runtime_branches_stats[cpu]); |
411 | 430 | ||
412 | if (total) | 431 | if (total) |
413 | ratio = avg * 100 / total; | 432 | ratio = avg * 100 / total; |
414 | 433 | ||
415 | fprintf(stderr, " # %10.3f %% ", ratio); | 434 | fprintf(stderr, " # %10.3f %% ", ratio); |
416 | 435 | ||
417 | } else if (runtime_nsecs_stats.n != 0) { | 436 | } else if (runtime_nsecs_stats[cpu].n != 0) { |
418 | total = avg_stats(&runtime_nsecs_stats); | 437 | total = avg_stats(&runtime_nsecs_stats[cpu]); |
419 | 438 | ||
420 | if (total) | 439 | if (total) |
421 | ratio = 1000.0 * avg / total; | 440 | ratio = 1000.0 * avg / total; |
@@ -426,30 +445,38 @@ static void abs_printout(int counter, double avg) | |||
426 | 445 | ||
427 | /* | 446 | /* |
428 | * Print out the results of a single counter: | 447 | * Print out the results of a single counter: |
448 | * aggregated counts in system-wide mode | ||
429 | */ | 449 | */ |
430 | static void print_counter(int counter) | 450 | static void print_counter_aggr(struct perf_evsel *counter) |
431 | { | 451 | { |
432 | double avg = avg_stats(&event_res_stats[counter][0]); | 452 | struct perf_stat *ps = counter->priv; |
433 | int scaled = event_scaled[counter]; | 453 | double avg = avg_stats(&ps->res_stats[0]); |
454 | int scaled = counter->counts->scaled; | ||
434 | 455 | ||
435 | if (scaled == -1) { | 456 | if (scaled == -1) { |
436 | fprintf(stderr, " %18s %-24s\n", | 457 | fprintf(stderr, "%*s%s%-24s\n", |
437 | "<not counted>", event_name(counter)); | 458 | csv_output ? 0 : 18, |
459 | "<not counted>", csv_sep, event_name(counter)); | ||
438 | return; | 460 | return; |
439 | } | 461 | } |
440 | 462 | ||
441 | if (nsec_counter(counter)) | 463 | if (nsec_counter(counter)) |
442 | nsec_printout(counter, avg); | 464 | nsec_printout(-1, counter, avg); |
443 | else | 465 | else |
444 | abs_printout(counter, avg); | 466 | abs_printout(-1, counter, avg); |
467 | |||
468 | if (csv_output) { | ||
469 | fputc('\n', stderr); | ||
470 | return; | ||
471 | } | ||
445 | 472 | ||
446 | print_noise(counter, avg); | 473 | print_noise(counter, avg); |
447 | 474 | ||
448 | if (scaled) { | 475 | if (scaled) { |
449 | double avg_enabled, avg_running; | 476 | double avg_enabled, avg_running; |
450 | 477 | ||
451 | avg_enabled = avg_stats(&event_res_stats[counter][1]); | 478 | avg_enabled = avg_stats(&ps->res_stats[1]); |
452 | avg_running = avg_stats(&event_res_stats[counter][2]); | 479 | avg_running = avg_stats(&ps->res_stats[2]); |
453 | 480 | ||
454 | fprintf(stderr, " (scaled from %.2f%%)", | 481 | fprintf(stderr, " (scaled from %.2f%%)", |
455 | 100 * avg_running / avg_enabled); | 482 | 100 * avg_running / avg_enabled); |
@@ -458,40 +485,92 @@ static void print_counter(int counter) | |||
458 | fprintf(stderr, "\n"); | 485 | fprintf(stderr, "\n"); |
459 | } | 486 | } |
460 | 487 | ||
488 | /* | ||
489 | * Print out the results of a single counter: | ||
490 | * does not use aggregated count in system-wide | ||
491 | */ | ||
492 | static void print_counter(struct perf_evsel *counter) | ||
493 | { | ||
494 | u64 ena, run, val; | ||
495 | int cpu; | ||
496 | |||
497 | for (cpu = 0; cpu < cpus->nr; cpu++) { | ||
498 | val = counter->counts->cpu[cpu].val; | ||
499 | ena = counter->counts->cpu[cpu].ena; | ||
500 | run = counter->counts->cpu[cpu].run; | ||
501 | if (run == 0 || ena == 0) { | ||
502 | fprintf(stderr, "CPU%*d%s%*s%s%-24s", | ||
503 | csv_output ? 0 : -4, | ||
504 | cpus->map[cpu], csv_sep, | ||
505 | csv_output ? 0 : 18, | ||
506 | "<not counted>", csv_sep, | ||
507 | event_name(counter)); | ||
508 | |||
509 | fprintf(stderr, "\n"); | ||
510 | continue; | ||
511 | } | ||
512 | |||
513 | if (nsec_counter(counter)) | ||
514 | nsec_printout(cpu, counter, val); | ||
515 | else | ||
516 | abs_printout(cpu, counter, val); | ||
517 | |||
518 | if (!csv_output) { | ||
519 | print_noise(counter, 1.0); | ||
520 | |||
521 | if (run != ena) { | ||
522 | fprintf(stderr, " (scaled from %.2f%%)", | ||
523 | 100.0 * run / ena); | ||
524 | } | ||
525 | } | ||
526 | fprintf(stderr, "\n"); | ||
527 | } | ||
528 | } | ||
529 | |||
461 | static void print_stat(int argc, const char **argv) | 530 | static void print_stat(int argc, const char **argv) |
462 | { | 531 | { |
463 | int i, counter; | 532 | struct perf_evsel *counter; |
533 | int i; | ||
464 | 534 | ||
465 | fflush(stdout); | 535 | fflush(stdout); |
466 | 536 | ||
467 | fprintf(stderr, "\n"); | 537 | if (!csv_output) { |
468 | fprintf(stderr, " Performance counter stats for "); | 538 | fprintf(stderr, "\n"); |
469 | if(target_pid == -1 && target_tid == -1) { | 539 | fprintf(stderr, " Performance counter stats for "); |
470 | fprintf(stderr, "\'%s", argv[0]); | 540 | if(target_pid == -1 && target_tid == -1) { |
471 | for (i = 1; i < argc; i++) | 541 | fprintf(stderr, "\'%s", argv[0]); |
472 | fprintf(stderr, " %s", argv[i]); | 542 | for (i = 1; i < argc; i++) |
473 | } else if (target_pid != -1) | 543 | fprintf(stderr, " %s", argv[i]); |
474 | fprintf(stderr, "process id \'%d", target_pid); | 544 | } else if (target_pid != -1) |
475 | else | 545 | fprintf(stderr, "process id \'%d", target_pid); |
476 | fprintf(stderr, "thread id \'%d", target_tid); | 546 | else |
477 | 547 | fprintf(stderr, "thread id \'%d", target_tid); | |
478 | fprintf(stderr, "\'"); | 548 | |
479 | if (run_count > 1) | 549 | fprintf(stderr, "\'"); |
480 | fprintf(stderr, " (%d runs)", run_count); | 550 | if (run_count > 1) |
481 | fprintf(stderr, ":\n\n"); | 551 | fprintf(stderr, " (%d runs)", run_count); |
552 | fprintf(stderr, ":\n\n"); | ||
553 | } | ||
482 | 554 | ||
483 | for (counter = 0; counter < nr_counters; counter++) | 555 | if (no_aggr) { |
484 | print_counter(counter); | 556 | list_for_each_entry(counter, &evsel_list, node) |
557 | print_counter(counter); | ||
558 | } else { | ||
559 | list_for_each_entry(counter, &evsel_list, node) | ||
560 | print_counter_aggr(counter); | ||
561 | } | ||
485 | 562 | ||
486 | fprintf(stderr, "\n"); | 563 | if (!csv_output) { |
487 | fprintf(stderr, " %18.9f seconds time elapsed", | 564 | fprintf(stderr, "\n"); |
488 | avg_stats(&walltime_nsecs_stats)/1e9); | 565 | fprintf(stderr, " %18.9f seconds time elapsed", |
489 | if (run_count > 1) { | 566 | avg_stats(&walltime_nsecs_stats)/1e9); |
490 | fprintf(stderr, " ( +- %7.3f%% )", | 567 | if (run_count > 1) { |
568 | fprintf(stderr, " ( +- %7.3f%% )", | ||
491 | 100*stddev_stats(&walltime_nsecs_stats) / | 569 | 100*stddev_stats(&walltime_nsecs_stats) / |
492 | avg_stats(&walltime_nsecs_stats)); | 570 | avg_stats(&walltime_nsecs_stats)); |
571 | } | ||
572 | fprintf(stderr, "\n\n"); | ||
493 | } | 573 | } |
494 | fprintf(stderr, "\n\n"); | ||
495 | } | 574 | } |
496 | 575 | ||
497 | static volatile int signr = -1; | 576 | static volatile int signr = -1; |
@@ -521,6 +600,13 @@ static const char * const stat_usage[] = { | |||
521 | NULL | 600 | NULL |
522 | }; | 601 | }; |
523 | 602 | ||
603 | static int stat__set_big_num(const struct option *opt __used, | ||
604 | const char *s __used, int unset) | ||
605 | { | ||
606 | big_num_opt = unset ? 0 : 1; | ||
607 | return 0; | ||
608 | } | ||
609 | |||
524 | static const struct option options[] = { | 610 | static const struct option options[] = { |
525 | OPT_CALLBACK('e', "event", NULL, "event", | 611 | OPT_CALLBACK('e', "event", NULL, "event", |
526 | "event selector. use 'perf list' to list available events", | 612 | "event selector. use 'perf list' to list available events", |
@@ -541,64 +627,96 @@ static const struct option options[] = { | |||
541 | "repeat command and print average + stddev (max: 100)"), | 627 | "repeat command and print average + stddev (max: 100)"), |
542 | OPT_BOOLEAN('n', "null", &null_run, | 628 | OPT_BOOLEAN('n', "null", &null_run, |
543 | "null run - dont start any counters"), | 629 | "null run - dont start any counters"), |
544 | OPT_BOOLEAN('B', "big-num", &big_num, | 630 | OPT_CALLBACK_NOOPT('B', "big-num", NULL, NULL, |
545 | "print large numbers with thousands\' separators"), | 631 | "print large numbers with thousands\' separators", |
632 | stat__set_big_num), | ||
546 | OPT_STRING('C', "cpu", &cpu_list, "cpu", | 633 | OPT_STRING('C', "cpu", &cpu_list, "cpu", |
547 | "list of cpus to monitor in system-wide"), | 634 | "list of cpus to monitor in system-wide"), |
635 | OPT_BOOLEAN('A', "no-aggr", &no_aggr, | ||
636 | "disable CPU count aggregation"), | ||
637 | OPT_STRING('x', "field-separator", &csv_sep, "separator", | ||
638 | "print counts with custom separator"), | ||
548 | OPT_END() | 639 | OPT_END() |
549 | }; | 640 | }; |
550 | 641 | ||
551 | int cmd_stat(int argc, const char **argv, const char *prefix __used) | 642 | int cmd_stat(int argc, const char **argv, const char *prefix __used) |
552 | { | 643 | { |
553 | int status; | 644 | struct perf_evsel *pos; |
554 | int i,j; | 645 | int status = -ENOMEM; |
555 | 646 | ||
556 | setlocale(LC_ALL, ""); | 647 | setlocale(LC_ALL, ""); |
557 | 648 | ||
558 | argc = parse_options(argc, argv, options, stat_usage, | 649 | argc = parse_options(argc, argv, options, stat_usage, |
559 | PARSE_OPT_STOP_AT_NON_OPTION); | 650 | PARSE_OPT_STOP_AT_NON_OPTION); |
651 | |||
652 | if (csv_sep) | ||
653 | csv_output = true; | ||
654 | else | ||
655 | csv_sep = DEFAULT_SEPARATOR; | ||
656 | |||
657 | /* | ||
658 | * let the spreadsheet do the pretty-printing | ||
659 | */ | ||
660 | if (csv_output) { | ||
661 | /* User explicitely passed -B? */ | ||
662 | if (big_num_opt == 1) { | ||
663 | fprintf(stderr, "-B option not supported with -x\n"); | ||
664 | usage_with_options(stat_usage, options); | ||
665 | } else /* Nope, so disable big number formatting */ | ||
666 | big_num = false; | ||
667 | } else if (big_num_opt == 0) /* User passed --no-big-num */ | ||
668 | big_num = false; | ||
669 | |||
560 | if (!argc && target_pid == -1 && target_tid == -1) | 670 | if (!argc && target_pid == -1 && target_tid == -1) |
561 | usage_with_options(stat_usage, options); | 671 | usage_with_options(stat_usage, options); |
562 | if (run_count <= 0) | 672 | if (run_count <= 0) |
563 | usage_with_options(stat_usage, options); | 673 | usage_with_options(stat_usage, options); |
564 | 674 | ||
675 | /* no_aggr is for system-wide only */ | ||
676 | if (no_aggr && !system_wide) | ||
677 | usage_with_options(stat_usage, options); | ||
678 | |||
565 | /* Set attrs and nr_counters if no event is selected and !null_run */ | 679 | /* Set attrs and nr_counters if no event is selected and !null_run */ |
566 | if (!null_run && !nr_counters) { | 680 | if (!null_run && !nr_counters) { |
567 | memcpy(attrs, default_attrs, sizeof(default_attrs)); | 681 | size_t c; |
682 | |||
568 | nr_counters = ARRAY_SIZE(default_attrs); | 683 | nr_counters = ARRAY_SIZE(default_attrs); |
684 | |||
685 | for (c = 0; c < ARRAY_SIZE(default_attrs); ++c) { | ||
686 | pos = perf_evsel__new(default_attrs[c].type, | ||
687 | default_attrs[c].config, | ||
688 | nr_counters); | ||
689 | if (pos == NULL) | ||
690 | goto out; | ||
691 | list_add(&pos->node, &evsel_list); | ||
692 | } | ||
569 | } | 693 | } |
570 | 694 | ||
571 | if (system_wide) | 695 | if (target_pid != -1) |
572 | nr_cpus = read_cpu_map(cpu_list); | 696 | target_tid = target_pid; |
573 | else | ||
574 | nr_cpus = 1; | ||
575 | 697 | ||
576 | if (nr_cpus < 1) | 698 | threads = thread_map__new(target_pid, target_tid); |
699 | if (threads == NULL) { | ||
700 | pr_err("Problems finding threads of monitor\n"); | ||
577 | usage_with_options(stat_usage, options); | 701 | usage_with_options(stat_usage, options); |
702 | } | ||
578 | 703 | ||
579 | if (target_pid != -1) { | 704 | if (system_wide) |
580 | target_tid = target_pid; | 705 | cpus = cpu_map__new(cpu_list); |
581 | thread_num = find_all_tid(target_pid, &all_tids); | 706 | else |
582 | if (thread_num <= 0) { | 707 | cpus = cpu_map__dummy_new(); |
583 | fprintf(stderr, "Can't find all threads of pid %d\n", | ||
584 | target_pid); | ||
585 | usage_with_options(stat_usage, options); | ||
586 | } | ||
587 | } else { | ||
588 | all_tids=malloc(sizeof(pid_t)); | ||
589 | if (!all_tids) | ||
590 | return -ENOMEM; | ||
591 | 708 | ||
592 | all_tids[0] = target_tid; | 709 | if (cpus == NULL) { |
593 | thread_num = 1; | 710 | perror("failed to parse CPUs map"); |
711 | usage_with_options(stat_usage, options); | ||
712 | return -1; | ||
594 | } | 713 | } |
595 | 714 | ||
596 | for (i = 0; i < MAX_NR_CPUS; i++) { | 715 | list_for_each_entry(pos, &evsel_list, node) { |
597 | for (j = 0; j < MAX_COUNTERS; j++) { | 716 | if (perf_evsel__alloc_stat_priv(pos) < 0 || |
598 | fd[i][j] = malloc(sizeof(int)*thread_num); | 717 | perf_evsel__alloc_counts(pos, cpus->nr) < 0 || |
599 | if (!fd[i][j]) | 718 | perf_evsel__alloc_fd(pos, cpus->nr, threads->nr) < 0) |
600 | return -ENOMEM; | 719 | goto out_free_fd; |
601 | } | ||
602 | } | 720 | } |
603 | 721 | ||
604 | /* | 722 | /* |
@@ -621,6 +739,11 @@ int cmd_stat(int argc, const char **argv, const char *prefix __used) | |||
621 | 739 | ||
622 | if (status != -1) | 740 | if (status != -1) |
623 | print_stat(argc, argv); | 741 | print_stat(argc, argv); |
624 | 742 | out_free_fd: | |
743 | list_for_each_entry(pos, &evsel_list, node) | ||
744 | perf_evsel__free_stat_priv(pos); | ||
745 | out: | ||
746 | thread_map__delete(threads); | ||
747 | threads = NULL; | ||
625 | return status; | 748 | return status; |
626 | } | 749 | } |
diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c index 035b9fa063a..1c984342a57 100644 --- a/tools/perf/builtin-test.c +++ b/tools/perf/builtin-test.c | |||
@@ -119,10 +119,16 @@ static int test__vmlinux_matches_kallsyms(void) | |||
119 | * end addresses too. | 119 | * end addresses too. |
120 | */ | 120 | */ |
121 | for (nd = rb_first(&vmlinux_map->dso->symbols[type]); nd; nd = rb_next(nd)) { | 121 | for (nd = rb_first(&vmlinux_map->dso->symbols[type]); nd; nd = rb_next(nd)) { |
122 | struct symbol *pair; | 122 | struct symbol *pair, *first_pair; |
123 | bool backwards = true; | ||
123 | 124 | ||
124 | sym = rb_entry(nd, struct symbol, rb_node); | 125 | sym = rb_entry(nd, struct symbol, rb_node); |
125 | pair = machine__find_kernel_symbol(&kallsyms, type, sym->start, NULL, NULL); | 126 | |
127 | if (sym->start == sym->end) | ||
128 | continue; | ||
129 | |||
130 | first_pair = machine__find_kernel_symbol(&kallsyms, type, sym->start, NULL, NULL); | ||
131 | pair = first_pair; | ||
126 | 132 | ||
127 | if (pair && pair->start == sym->start) { | 133 | if (pair && pair->start == sym->start) { |
128 | next_pair: | 134 | next_pair: |
@@ -143,8 +149,10 @@ next_pair: | |||
143 | pr_debug("%#Lx: diff end addr for %s v: %#Lx k: %#Lx\n", | 149 | pr_debug("%#Lx: diff end addr for %s v: %#Lx k: %#Lx\n", |
144 | sym->start, sym->name, sym->end, pair->end); | 150 | sym->start, sym->name, sym->end, pair->end); |
145 | } else { | 151 | } else { |
146 | struct rb_node *nnd = rb_prev(&pair->rb_node); | 152 | struct rb_node *nnd; |
147 | 153 | detour: | |
154 | nnd = backwards ? rb_prev(&pair->rb_node) : | ||
155 | rb_next(&pair->rb_node); | ||
148 | if (nnd) { | 156 | if (nnd) { |
149 | struct symbol *next = rb_entry(nnd, struct symbol, rb_node); | 157 | struct symbol *next = rb_entry(nnd, struct symbol, rb_node); |
150 | 158 | ||
@@ -153,6 +161,13 @@ next_pair: | |||
153 | goto next_pair; | 161 | goto next_pair; |
154 | } | 162 | } |
155 | } | 163 | } |
164 | |||
165 | if (backwards) { | ||
166 | backwards = false; | ||
167 | pair = first_pair; | ||
168 | goto detour; | ||
169 | } | ||
170 | |||
156 | pr_debug("%#Lx: diff name v: %s k: %s\n", | 171 | pr_debug("%#Lx: diff name v: %s k: %s\n", |
157 | sym->start, sym->name, pair->name); | 172 | sym->start, sym->name, pair->name); |
158 | } | 173 | } |
@@ -219,6 +234,89 @@ out: | |||
219 | return err; | 234 | return err; |
220 | } | 235 | } |
221 | 236 | ||
237 | #include "util/evsel.h" | ||
238 | #include <sys/types.h> | ||
239 | |||
240 | static int trace_event__id(const char *event_name) | ||
241 | { | ||
242 | char *filename; | ||
243 | int err = -1, fd; | ||
244 | |||
245 | if (asprintf(&filename, | ||
246 | "/sys/kernel/debug/tracing/events/syscalls/%s/id", | ||
247 | event_name) < 0) | ||
248 | return -1; | ||
249 | |||
250 | fd = open(filename, O_RDONLY); | ||
251 | if (fd >= 0) { | ||
252 | char id[16]; | ||
253 | if (read(fd, id, sizeof(id)) > 0) | ||
254 | err = atoi(id); | ||
255 | close(fd); | ||
256 | } | ||
257 | |||
258 | free(filename); | ||
259 | return err; | ||
260 | } | ||
261 | |||
262 | static int test__open_syscall_event(void) | ||
263 | { | ||
264 | int err = -1, fd; | ||
265 | struct thread_map *threads; | ||
266 | struct perf_evsel *evsel; | ||
267 | unsigned int nr_open_calls = 111, i; | ||
268 | int id = trace_event__id("sys_enter_open"); | ||
269 | |||
270 | if (id < 0) { | ||
271 | pr_debug("is debugfs mounted on /sys/kernel/debug?\n"); | ||
272 | return -1; | ||
273 | } | ||
274 | |||
275 | threads = thread_map__new(-1, getpid()); | ||
276 | if (threads == NULL) { | ||
277 | pr_debug("thread_map__new\n"); | ||
278 | return -1; | ||
279 | } | ||
280 | |||
281 | evsel = perf_evsel__new(PERF_TYPE_TRACEPOINT, id, 0); | ||
282 | if (evsel == NULL) { | ||
283 | pr_debug("perf_evsel__new\n"); | ||
284 | goto out_thread_map_delete; | ||
285 | } | ||
286 | |||
287 | if (perf_evsel__open_per_thread(evsel, threads) < 0) { | ||
288 | pr_debug("failed to open counter: %s, " | ||
289 | "tweak /proc/sys/kernel/perf_event_paranoid?\n", | ||
290 | strerror(errno)); | ||
291 | goto out_evsel_delete; | ||
292 | } | ||
293 | |||
294 | for (i = 0; i < nr_open_calls; ++i) { | ||
295 | fd = open("/etc/passwd", O_RDONLY); | ||
296 | close(fd); | ||
297 | } | ||
298 | |||
299 | if (perf_evsel__read_on_cpu(evsel, 0, 0) < 0) { | ||
300 | pr_debug("perf_evsel__open_read_on_cpu\n"); | ||
301 | goto out_close_fd; | ||
302 | } | ||
303 | |||
304 | if (evsel->counts->cpu[0].val != nr_open_calls) { | ||
305 | pr_debug("perf_evsel__read_on_cpu: expected to intercept %d calls, got %Ld\n", | ||
306 | nr_open_calls, evsel->counts->cpu[0].val); | ||
307 | goto out_close_fd; | ||
308 | } | ||
309 | |||
310 | err = 0; | ||
311 | out_close_fd: | ||
312 | perf_evsel__close_fd(evsel, 1, threads->nr); | ||
313 | out_evsel_delete: | ||
314 | perf_evsel__delete(evsel); | ||
315 | out_thread_map_delete: | ||
316 | thread_map__delete(threads); | ||
317 | return err; | ||
318 | } | ||
319 | |||
222 | static struct test { | 320 | static struct test { |
223 | const char *desc; | 321 | const char *desc; |
224 | int (*func)(void); | 322 | int (*func)(void); |
@@ -228,6 +326,10 @@ static struct test { | |||
228 | .func = test__vmlinux_matches_kallsyms, | 326 | .func = test__vmlinux_matches_kallsyms, |
229 | }, | 327 | }, |
230 | { | 328 | { |
329 | .desc = "detect open syscall event", | ||
330 | .func = test__open_syscall_event, | ||
331 | }, | ||
332 | { | ||
231 | .func = NULL, | 333 | .func = NULL, |
232 | }, | 334 | }, |
233 | }; | 335 | }; |
diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index 9bcc38f0b70..746cf03cb05 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c | |||
@@ -32,6 +32,10 @@ | |||
32 | #include "util/session.h" | 32 | #include "util/session.h" |
33 | #include "util/svghelper.h" | 33 | #include "util/svghelper.h" |
34 | 34 | ||
35 | #define SUPPORT_OLD_POWER_EVENTS 1 | ||
36 | #define PWR_EVENT_EXIT -1 | ||
37 | |||
38 | |||
35 | static char const *input_name = "perf.data"; | 39 | static char const *input_name = "perf.data"; |
36 | static char const *output_name = "output.svg"; | 40 | static char const *output_name = "output.svg"; |
37 | 41 | ||
@@ -272,19 +276,22 @@ static int cpus_cstate_state[MAX_CPUS]; | |||
272 | static u64 cpus_pstate_start_times[MAX_CPUS]; | 276 | static u64 cpus_pstate_start_times[MAX_CPUS]; |
273 | static u64 cpus_pstate_state[MAX_CPUS]; | 277 | static u64 cpus_pstate_state[MAX_CPUS]; |
274 | 278 | ||
275 | static int process_comm_event(event_t *event, struct perf_session *session __used) | 279 | static int process_comm_event(event_t *event, struct sample_data *sample __used, |
280 | struct perf_session *session __used) | ||
276 | { | 281 | { |
277 | pid_set_comm(event->comm.tid, event->comm.comm); | 282 | pid_set_comm(event->comm.tid, event->comm.comm); |
278 | return 0; | 283 | return 0; |
279 | } | 284 | } |
280 | 285 | ||
281 | static int process_fork_event(event_t *event, struct perf_session *session __used) | 286 | static int process_fork_event(event_t *event, struct sample_data *sample __used, |
287 | struct perf_session *session __used) | ||
282 | { | 288 | { |
283 | pid_fork(event->fork.pid, event->fork.ppid, event->fork.time); | 289 | pid_fork(event->fork.pid, event->fork.ppid, event->fork.time); |
284 | return 0; | 290 | return 0; |
285 | } | 291 | } |
286 | 292 | ||
287 | static int process_exit_event(event_t *event, struct perf_session *session __used) | 293 | static int process_exit_event(event_t *event, struct sample_data *sample __used, |
294 | struct perf_session *session __used) | ||
288 | { | 295 | { |
289 | pid_exit(event->fork.pid, event->fork.time); | 296 | pid_exit(event->fork.pid, event->fork.time); |
290 | return 0; | 297 | return 0; |
@@ -298,12 +305,21 @@ struct trace_entry { | |||
298 | int lock_depth; | 305 | int lock_depth; |
299 | }; | 306 | }; |
300 | 307 | ||
301 | struct power_entry { | 308 | #ifdef SUPPORT_OLD_POWER_EVENTS |
309 | static int use_old_power_events; | ||
310 | struct power_entry_old { | ||
302 | struct trace_entry te; | 311 | struct trace_entry te; |
303 | u64 type; | 312 | u64 type; |
304 | u64 value; | 313 | u64 value; |
305 | u64 cpu_id; | 314 | u64 cpu_id; |
306 | }; | 315 | }; |
316 | #endif | ||
317 | |||
318 | struct power_processor_entry { | ||
319 | struct trace_entry te; | ||
320 | u32 state; | ||
321 | u32 cpu_id; | ||
322 | }; | ||
307 | 323 | ||
308 | #define TASK_COMM_LEN 16 | 324 | #define TASK_COMM_LEN 16 |
309 | struct wakeup_entry { | 325 | struct wakeup_entry { |
@@ -470,48 +486,65 @@ static void sched_switch(int cpu, u64 timestamp, struct trace_entry *te) | |||
470 | } | 486 | } |
471 | 487 | ||
472 | 488 | ||
473 | static int process_sample_event(event_t *event, struct perf_session *session) | 489 | static int process_sample_event(event_t *event __used, |
490 | struct sample_data *sample, | ||
491 | struct perf_session *session) | ||
474 | { | 492 | { |
475 | struct sample_data data; | ||
476 | struct trace_entry *te; | 493 | struct trace_entry *te; |
477 | 494 | ||
478 | memset(&data, 0, sizeof(data)); | ||
479 | |||
480 | event__parse_sample(event, session->sample_type, &data); | ||
481 | |||
482 | if (session->sample_type & PERF_SAMPLE_TIME) { | 495 | if (session->sample_type & PERF_SAMPLE_TIME) { |
483 | if (!first_time || first_time > data.time) | 496 | if (!first_time || first_time > sample->time) |
484 | first_time = data.time; | 497 | first_time = sample->time; |
485 | if (last_time < data.time) | 498 | if (last_time < sample->time) |
486 | last_time = data.time; | 499 | last_time = sample->time; |
487 | } | 500 | } |
488 | 501 | ||
489 | te = (void *)data.raw_data; | 502 | te = (void *)sample->raw_data; |
490 | if (session->sample_type & PERF_SAMPLE_RAW && data.raw_size > 0) { | 503 | if (session->sample_type & PERF_SAMPLE_RAW && sample->raw_size > 0) { |
491 | char *event_str; | 504 | char *event_str; |
492 | struct power_entry *pe; | 505 | #ifdef SUPPORT_OLD_POWER_EVENTS |
493 | 506 | struct power_entry_old *peo; | |
494 | pe = (void *)te; | 507 | peo = (void *)te; |
495 | 508 | #endif | |
496 | event_str = perf_header__find_event(te->type); | 509 | event_str = perf_header__find_event(te->type); |
497 | 510 | ||
498 | if (!event_str) | 511 | if (!event_str) |
499 | return 0; | 512 | return 0; |
500 | 513 | ||
501 | if (strcmp(event_str, "power:power_start") == 0) | 514 | if (strcmp(event_str, "power:cpu_idle") == 0) { |
502 | c_state_start(pe->cpu_id, data.time, pe->value); | 515 | struct power_processor_entry *ppe = (void *)te; |
516 | if (ppe->state == (u32)PWR_EVENT_EXIT) | ||
517 | c_state_end(ppe->cpu_id, sample->time); | ||
518 | else | ||
519 | c_state_start(ppe->cpu_id, sample->time, | ||
520 | ppe->state); | ||
521 | } | ||
522 | else if (strcmp(event_str, "power:cpu_frequency") == 0) { | ||
523 | struct power_processor_entry *ppe = (void *)te; | ||
524 | p_state_change(ppe->cpu_id, sample->time, ppe->state); | ||
525 | } | ||
526 | |||
527 | else if (strcmp(event_str, "sched:sched_wakeup") == 0) | ||
528 | sched_wakeup(sample->cpu, sample->time, sample->pid, te); | ||
503 | 529 | ||
504 | if (strcmp(event_str, "power:power_end") == 0) | 530 | else if (strcmp(event_str, "sched:sched_switch") == 0) |
505 | c_state_end(pe->cpu_id, data.time); | 531 | sched_switch(sample->cpu, sample->time, te); |
506 | 532 | ||
507 | if (strcmp(event_str, "power:power_frequency") == 0) | 533 | #ifdef SUPPORT_OLD_POWER_EVENTS |
508 | p_state_change(pe->cpu_id, data.time, pe->value); | 534 | if (use_old_power_events) { |
535 | if (strcmp(event_str, "power:power_start") == 0) | ||
536 | c_state_start(peo->cpu_id, sample->time, | ||
537 | peo->value); | ||
509 | 538 | ||
510 | if (strcmp(event_str, "sched:sched_wakeup") == 0) | 539 | else if (strcmp(event_str, "power:power_end") == 0) |
511 | sched_wakeup(data.cpu, data.time, data.pid, te); | 540 | c_state_end(sample->cpu, sample->time); |
512 | 541 | ||
513 | if (strcmp(event_str, "sched:sched_switch") == 0) | 542 | else if (strcmp(event_str, |
514 | sched_switch(data.cpu, data.time, te); | 543 | "power:power_frequency") == 0) |
544 | p_state_change(peo->cpu_id, sample->time, | ||
545 | peo->value); | ||
546 | } | ||
547 | #endif | ||
515 | } | 548 | } |
516 | return 0; | 549 | return 0; |
517 | } | 550 | } |
@@ -937,7 +970,8 @@ static struct perf_event_ops event_ops = { | |||
937 | 970 | ||
938 | static int __cmd_timechart(void) | 971 | static int __cmd_timechart(void) |
939 | { | 972 | { |
940 | struct perf_session *session = perf_session__new(input_name, O_RDONLY, 0, false); | 973 | struct perf_session *session = perf_session__new(input_name, O_RDONLY, |
974 | 0, false, &event_ops); | ||
941 | int ret = -EINVAL; | 975 | int ret = -EINVAL; |
942 | 976 | ||
943 | if (session == NULL) | 977 | if (session == NULL) |
@@ -968,7 +1002,8 @@ static const char * const timechart_usage[] = { | |||
968 | NULL | 1002 | NULL |
969 | }; | 1003 | }; |
970 | 1004 | ||
971 | static const char *record_args[] = { | 1005 | #ifdef SUPPORT_OLD_POWER_EVENTS |
1006 | static const char * const record_old_args[] = { | ||
972 | "record", | 1007 | "record", |
973 | "-a", | 1008 | "-a", |
974 | "-R", | 1009 | "-R", |
@@ -980,16 +1015,43 @@ static const char *record_args[] = { | |||
980 | "-e", "sched:sched_wakeup", | 1015 | "-e", "sched:sched_wakeup", |
981 | "-e", "sched:sched_switch", | 1016 | "-e", "sched:sched_switch", |
982 | }; | 1017 | }; |
1018 | #endif | ||
1019 | |||
1020 | static const char * const record_new_args[] = { | ||
1021 | "record", | ||
1022 | "-a", | ||
1023 | "-R", | ||
1024 | "-f", | ||
1025 | "-c", "1", | ||
1026 | "-e", "power:cpu_frequency", | ||
1027 | "-e", "power:cpu_idle", | ||
1028 | "-e", "sched:sched_wakeup", | ||
1029 | "-e", "sched:sched_switch", | ||
1030 | }; | ||
983 | 1031 | ||
984 | static int __cmd_record(int argc, const char **argv) | 1032 | static int __cmd_record(int argc, const char **argv) |
985 | { | 1033 | { |
986 | unsigned int rec_argc, i, j; | 1034 | unsigned int rec_argc, i, j; |
987 | const char **rec_argv; | 1035 | const char **rec_argv; |
1036 | const char * const *record_args = record_new_args; | ||
1037 | unsigned int record_elems = ARRAY_SIZE(record_new_args); | ||
1038 | |||
1039 | #ifdef SUPPORT_OLD_POWER_EVENTS | ||
1040 | if (!is_valid_tracepoint("power:cpu_idle") && | ||
1041 | is_valid_tracepoint("power:power_start")) { | ||
1042 | use_old_power_events = 1; | ||
1043 | record_args = record_old_args; | ||
1044 | record_elems = ARRAY_SIZE(record_old_args); | ||
1045 | } | ||
1046 | #endif | ||
988 | 1047 | ||
989 | rec_argc = ARRAY_SIZE(record_args) + argc - 1; | 1048 | rec_argc = record_elems + argc - 1; |
990 | rec_argv = calloc(rec_argc + 1, sizeof(char *)); | 1049 | rec_argv = calloc(rec_argc + 1, sizeof(char *)); |
991 | 1050 | ||
992 | for (i = 0; i < ARRAY_SIZE(record_args); i++) | 1051 | if (rec_argv == NULL) |
1052 | return -ENOMEM; | ||
1053 | |||
1054 | for (i = 0; i < record_elems; i++) | ||
993 | rec_argv[i] = strdup(record_args[i]); | 1055 | rec_argv[i] = strdup(record_args[i]); |
994 | 1056 | ||
995 | for (j = 1; j < (unsigned int)argc; j++, i++) | 1057 | for (j = 1; j < (unsigned int)argc; j++, i++) |
@@ -1018,6 +1080,8 @@ static const struct option options[] = { | |||
1018 | OPT_CALLBACK('p', "process", NULL, "process", | 1080 | OPT_CALLBACK('p', "process", NULL, "process", |
1019 | "process selector. Pass a pid or process name.", | 1081 | "process selector. Pass a pid or process name.", |
1020 | parse_process), | 1082 | parse_process), |
1083 | OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory", | ||
1084 | "Look for files with symbols relative to this directory"), | ||
1021 | OPT_END() | 1085 | OPT_END() |
1022 | }; | 1086 | }; |
1023 | 1087 | ||
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index dd625808c2a..1e67ab9c7eb 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c | |||
@@ -21,6 +21,7 @@ | |||
21 | #include "perf.h" | 21 | #include "perf.h" |
22 | 22 | ||
23 | #include "util/color.h" | 23 | #include "util/color.h" |
24 | #include "util/evsel.h" | ||
24 | #include "util/session.h" | 25 | #include "util/session.h" |
25 | #include "util/symbol.h" | 26 | #include "util/symbol.h" |
26 | #include "util/thread.h" | 27 | #include "util/thread.h" |
@@ -29,6 +30,7 @@ | |||
29 | #include "util/parse-options.h" | 30 | #include "util/parse-options.h" |
30 | #include "util/parse-events.h" | 31 | #include "util/parse-events.h" |
31 | #include "util/cpumap.h" | 32 | #include "util/cpumap.h" |
33 | #include "util/xyarray.h" | ||
32 | 34 | ||
33 | #include "util/debug.h" | 35 | #include "util/debug.h" |
34 | 36 | ||
@@ -55,7 +57,7 @@ | |||
55 | #include <linux/unistd.h> | 57 | #include <linux/unistd.h> |
56 | #include <linux/types.h> | 58 | #include <linux/types.h> |
57 | 59 | ||
58 | static int *fd[MAX_NR_CPUS][MAX_COUNTERS]; | 60 | #define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y)) |
59 | 61 | ||
60 | static bool system_wide = false; | 62 | static bool system_wide = false; |
61 | 63 | ||
@@ -66,10 +68,9 @@ static int print_entries; | |||
66 | 68 | ||
67 | static int target_pid = -1; | 69 | static int target_pid = -1; |
68 | static int target_tid = -1; | 70 | static int target_tid = -1; |
69 | static pid_t *all_tids = NULL; | 71 | static struct thread_map *threads; |
70 | static int thread_num = 0; | ||
71 | static bool inherit = false; | 72 | static bool inherit = false; |
72 | static int nr_cpus = 0; | 73 | static struct cpu_map *cpus; |
73 | static int realtime_prio = 0; | 74 | static int realtime_prio = 0; |
74 | static bool group = false; | 75 | static bool group = false; |
75 | static unsigned int page_size; | 76 | static unsigned int page_size; |
@@ -100,6 +101,7 @@ struct sym_entry *sym_filter_entry = NULL; | |||
100 | struct sym_entry *sym_filter_entry_sched = NULL; | 101 | struct sym_entry *sym_filter_entry_sched = NULL; |
101 | static int sym_pcnt_filter = 5; | 102 | static int sym_pcnt_filter = 5; |
102 | static int sym_counter = 0; | 103 | static int sym_counter = 0; |
104 | static struct perf_evsel *sym_evsel = NULL; | ||
103 | static int display_weighted = -1; | 105 | static int display_weighted = -1; |
104 | static const char *cpu_list; | 106 | static const char *cpu_list; |
105 | 107 | ||
@@ -353,7 +355,7 @@ static void show_details(struct sym_entry *syme) | |||
353 | return; | 355 | return; |
354 | 356 | ||
355 | symbol = sym_entry__symbol(syme); | 357 | symbol = sym_entry__symbol(syme); |
356 | printf("Showing %s for %s\n", event_name(sym_counter), symbol->name); | 358 | printf("Showing %s for %s\n", event_name(sym_evsel), symbol->name); |
357 | printf(" Events Pcnt (>=%d%%)\n", sym_pcnt_filter); | 359 | printf(" Events Pcnt (>=%d%%)\n", sym_pcnt_filter); |
358 | 360 | ||
359 | pthread_mutex_lock(&syme->src->lock); | 361 | pthread_mutex_lock(&syme->src->lock); |
@@ -460,7 +462,8 @@ static void rb_insert_active_sym(struct rb_root *tree, struct sym_entry *se) | |||
460 | static void print_sym_table(void) | 462 | static void print_sym_table(void) |
461 | { | 463 | { |
462 | int printed = 0, j; | 464 | int printed = 0, j; |
463 | int counter, snap = !display_weighted ? sym_counter : 0; | 465 | struct perf_evsel *counter; |
466 | int snap = !display_weighted ? sym_counter : 0; | ||
464 | float samples_per_sec = samples/delay_secs; | 467 | float samples_per_sec = samples/delay_secs; |
465 | float ksamples_per_sec = kernel_samples/delay_secs; | 468 | float ksamples_per_sec = kernel_samples/delay_secs; |
466 | float us_samples_per_sec = (us_samples)/delay_secs; | 469 | float us_samples_per_sec = (us_samples)/delay_secs; |
@@ -532,7 +535,9 @@ static void print_sym_table(void) | |||
532 | } | 535 | } |
533 | 536 | ||
534 | if (nr_counters == 1 || !display_weighted) { | 537 | if (nr_counters == 1 || !display_weighted) { |
535 | printf("%Ld", (u64)attrs[0].sample_period); | 538 | struct perf_evsel *first; |
539 | first = list_entry(evsel_list.next, struct perf_evsel, node); | ||
540 | printf("%Ld", first->attr.sample_period); | ||
536 | if (freq) | 541 | if (freq) |
537 | printf("Hz "); | 542 | printf("Hz "); |
538 | else | 543 | else |
@@ -540,9 +545,9 @@ static void print_sym_table(void) | |||
540 | } | 545 | } |
541 | 546 | ||
542 | if (!display_weighted) | 547 | if (!display_weighted) |
543 | printf("%s", event_name(sym_counter)); | 548 | printf("%s", event_name(sym_evsel)); |
544 | else for (counter = 0; counter < nr_counters; counter++) { | 549 | else list_for_each_entry(counter, &evsel_list, node) { |
545 | if (counter) | 550 | if (counter->idx) |
546 | printf("/"); | 551 | printf("/"); |
547 | 552 | ||
548 | printf("%s", event_name(counter)); | 553 | printf("%s", event_name(counter)); |
@@ -558,12 +563,12 @@ static void print_sym_table(void) | |||
558 | printf(" (all"); | 563 | printf(" (all"); |
559 | 564 | ||
560 | if (cpu_list) | 565 | if (cpu_list) |
561 | printf(", CPU%s: %s)\n", nr_cpus > 1 ? "s" : "", cpu_list); | 566 | printf(", CPU%s: %s)\n", cpus->nr > 1 ? "s" : "", cpu_list); |
562 | else { | 567 | else { |
563 | if (target_tid != -1) | 568 | if (target_tid != -1) |
564 | printf(")\n"); | 569 | printf(")\n"); |
565 | else | 570 | else |
566 | printf(", %d CPU%s)\n", nr_cpus, nr_cpus > 1 ? "s" : ""); | 571 | printf(", %d CPU%s)\n", cpus->nr, cpus->nr > 1 ? "s" : ""); |
567 | } | 572 | } |
568 | 573 | ||
569 | printf("%-*.*s\n", win_width, win_width, graph_dotted_line); | 574 | printf("%-*.*s\n", win_width, win_width, graph_dotted_line); |
@@ -739,7 +744,7 @@ static void print_mapped_keys(void) | |||
739 | fprintf(stdout, "\t[e] display entries (lines). \t(%d)\n", print_entries); | 744 | fprintf(stdout, "\t[e] display entries (lines). \t(%d)\n", print_entries); |
740 | 745 | ||
741 | if (nr_counters > 1) | 746 | if (nr_counters > 1) |
742 | fprintf(stdout, "\t[E] active event counter. \t(%s)\n", event_name(sym_counter)); | 747 | fprintf(stdout, "\t[E] active event counter. \t(%s)\n", event_name(sym_evsel)); |
743 | 748 | ||
744 | fprintf(stdout, "\t[f] profile display filter (count). \t(%d)\n", count_filter); | 749 | fprintf(stdout, "\t[f] profile display filter (count). \t(%d)\n", count_filter); |
745 | 750 | ||
@@ -826,19 +831,23 @@ static void handle_keypress(struct perf_session *session, int c) | |||
826 | break; | 831 | break; |
827 | case 'E': | 832 | case 'E': |
828 | if (nr_counters > 1) { | 833 | if (nr_counters > 1) { |
829 | int i; | ||
830 | |||
831 | fprintf(stderr, "\nAvailable events:"); | 834 | fprintf(stderr, "\nAvailable events:"); |
832 | for (i = 0; i < nr_counters; i++) | 835 | |
833 | fprintf(stderr, "\n\t%d %s", i, event_name(i)); | 836 | list_for_each_entry(sym_evsel, &evsel_list, node) |
837 | fprintf(stderr, "\n\t%d %s", sym_evsel->idx, event_name(sym_evsel)); | ||
834 | 838 | ||
835 | prompt_integer(&sym_counter, "Enter details event counter"); | 839 | prompt_integer(&sym_counter, "Enter details event counter"); |
836 | 840 | ||
837 | if (sym_counter >= nr_counters) { | 841 | if (sym_counter >= nr_counters) { |
838 | fprintf(stderr, "Sorry, no such event, using %s.\n", event_name(0)); | 842 | sym_evsel = list_entry(evsel_list.next, struct perf_evsel, node); |
839 | sym_counter = 0; | 843 | sym_counter = 0; |
844 | fprintf(stderr, "Sorry, no such event, using %s.\n", event_name(sym_evsel)); | ||
840 | sleep(1); | 845 | sleep(1); |
846 | break; | ||
841 | } | 847 | } |
848 | list_for_each_entry(sym_evsel, &evsel_list, node) | ||
849 | if (sym_evsel->idx == sym_counter) | ||
850 | break; | ||
842 | } else sym_counter = 0; | 851 | } else sym_counter = 0; |
843 | break; | 852 | break; |
844 | case 'f': | 853 | case 'f': |
@@ -977,12 +986,13 @@ static int symbol_filter(struct map *map, struct symbol *sym) | |||
977 | } | 986 | } |
978 | 987 | ||
979 | static void event__process_sample(const event_t *self, | 988 | static void event__process_sample(const event_t *self, |
980 | struct perf_session *session, int counter) | 989 | struct sample_data *sample, |
990 | struct perf_session *session, | ||
991 | struct perf_evsel *evsel) | ||
981 | { | 992 | { |
982 | u64 ip = self->ip.ip; | 993 | u64 ip = self->ip.ip; |
983 | struct sym_entry *syme; | 994 | struct sym_entry *syme; |
984 | struct addr_location al; | 995 | struct addr_location al; |
985 | struct sample_data data; | ||
986 | struct machine *machine; | 996 | struct machine *machine; |
987 | u8 origin = self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; | 997 | u8 origin = self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; |
988 | 998 | ||
@@ -1025,7 +1035,7 @@ static void event__process_sample(const event_t *self, | |||
1025 | if (self->header.misc & PERF_RECORD_MISC_EXACT_IP) | 1035 | if (self->header.misc & PERF_RECORD_MISC_EXACT_IP) |
1026 | exact_samples++; | 1036 | exact_samples++; |
1027 | 1037 | ||
1028 | if (event__preprocess_sample(self, session, &al, &data, | 1038 | if (event__preprocess_sample(self, session, &al, sample, |
1029 | symbol_filter) < 0 || | 1039 | symbol_filter) < 0 || |
1030 | al.filtered) | 1040 | al.filtered) |
1031 | return; | 1041 | return; |
@@ -1071,9 +1081,9 @@ static void event__process_sample(const event_t *self, | |||
1071 | 1081 | ||
1072 | syme = symbol__priv(al.sym); | 1082 | syme = symbol__priv(al.sym); |
1073 | if (!syme->skip) { | 1083 | if (!syme->skip) { |
1074 | syme->count[counter]++; | 1084 | syme->count[evsel->idx]++; |
1075 | syme->origin = origin; | 1085 | syme->origin = origin; |
1076 | record_precise_ip(syme, counter, ip); | 1086 | record_precise_ip(syme, evsel->idx, ip); |
1077 | pthread_mutex_lock(&active_symbols_lock); | 1087 | pthread_mutex_lock(&active_symbols_lock); |
1078 | if (list_empty(&syme->node) || !syme->node.next) | 1088 | if (list_empty(&syme->node) || !syme->node.next) |
1079 | __list_insert_active_sym(syme); | 1089 | __list_insert_active_sym(syme); |
@@ -1082,12 +1092,24 @@ static void event__process_sample(const event_t *self, | |||
1082 | } | 1092 | } |
1083 | 1093 | ||
1084 | struct mmap_data { | 1094 | struct mmap_data { |
1085 | int counter; | ||
1086 | void *base; | 1095 | void *base; |
1087 | int mask; | 1096 | int mask; |
1088 | unsigned int prev; | 1097 | unsigned int prev; |
1089 | }; | 1098 | }; |
1090 | 1099 | ||
1100 | static int perf_evsel__alloc_mmap_per_thread(struct perf_evsel *evsel, | ||
1101 | int ncpus, int nthreads) | ||
1102 | { | ||
1103 | evsel->priv = xyarray__new(ncpus, nthreads, sizeof(struct mmap_data)); | ||
1104 | return evsel->priv != NULL ? 0 : -ENOMEM; | ||
1105 | } | ||
1106 | |||
1107 | static void perf_evsel__free_mmap(struct perf_evsel *evsel) | ||
1108 | { | ||
1109 | xyarray__delete(evsel->priv); | ||
1110 | evsel->priv = NULL; | ||
1111 | } | ||
1112 | |||
1091 | static unsigned int mmap_read_head(struct mmap_data *md) | 1113 | static unsigned int mmap_read_head(struct mmap_data *md) |
1092 | { | 1114 | { |
1093 | struct perf_event_mmap_page *pc = md->base; | 1115 | struct perf_event_mmap_page *pc = md->base; |
@@ -1100,11 +1122,15 @@ static unsigned int mmap_read_head(struct mmap_data *md) | |||
1100 | } | 1122 | } |
1101 | 1123 | ||
1102 | static void perf_session__mmap_read_counter(struct perf_session *self, | 1124 | static void perf_session__mmap_read_counter(struct perf_session *self, |
1103 | struct mmap_data *md) | 1125 | struct perf_evsel *evsel, |
1126 | int cpu, int thread_idx) | ||
1104 | { | 1127 | { |
1128 | struct xyarray *mmap_array = evsel->priv; | ||
1129 | struct mmap_data *md = xyarray__entry(mmap_array, cpu, thread_idx); | ||
1105 | unsigned int head = mmap_read_head(md); | 1130 | unsigned int head = mmap_read_head(md); |
1106 | unsigned int old = md->prev; | 1131 | unsigned int old = md->prev; |
1107 | unsigned char *data = md->base + page_size; | 1132 | unsigned char *data = md->base + page_size; |
1133 | struct sample_data sample; | ||
1108 | int diff; | 1134 | int diff; |
1109 | 1135 | ||
1110 | /* | 1136 | /* |
@@ -1152,10 +1178,11 @@ static void perf_session__mmap_read_counter(struct perf_session *self, | |||
1152 | event = &event_copy; | 1178 | event = &event_copy; |
1153 | } | 1179 | } |
1154 | 1180 | ||
1181 | event__parse_sample(event, self, &sample); | ||
1155 | if (event->header.type == PERF_RECORD_SAMPLE) | 1182 | if (event->header.type == PERF_RECORD_SAMPLE) |
1156 | event__process_sample(event, self, md->counter); | 1183 | event__process_sample(event, &sample, self, evsel); |
1157 | else | 1184 | else |
1158 | event__process(event, self); | 1185 | event__process(event, &sample, self); |
1159 | old += size; | 1186 | old += size; |
1160 | } | 1187 | } |
1161 | 1188 | ||
@@ -1163,36 +1190,39 @@ static void perf_session__mmap_read_counter(struct perf_session *self, | |||
1163 | } | 1190 | } |
1164 | 1191 | ||
1165 | static struct pollfd *event_array; | 1192 | static struct pollfd *event_array; |
1166 | static struct mmap_data *mmap_array[MAX_NR_CPUS][MAX_COUNTERS]; | ||
1167 | 1193 | ||
1168 | static void perf_session__mmap_read(struct perf_session *self) | 1194 | static void perf_session__mmap_read(struct perf_session *self) |
1169 | { | 1195 | { |
1170 | int i, counter, thread_index; | 1196 | struct perf_evsel *counter; |
1197 | int i, thread_index; | ||
1171 | 1198 | ||
1172 | for (i = 0; i < nr_cpus; i++) { | 1199 | for (i = 0; i < cpus->nr; i++) { |
1173 | for (counter = 0; counter < nr_counters; counter++) | 1200 | list_for_each_entry(counter, &evsel_list, node) { |
1174 | for (thread_index = 0; | 1201 | for (thread_index = 0; |
1175 | thread_index < thread_num; | 1202 | thread_index < threads->nr; |
1176 | thread_index++) { | 1203 | thread_index++) { |
1177 | perf_session__mmap_read_counter(self, | 1204 | perf_session__mmap_read_counter(self, |
1178 | &mmap_array[i][counter][thread_index]); | 1205 | counter, i, thread_index); |
1179 | } | 1206 | } |
1207 | } | ||
1180 | } | 1208 | } |
1181 | } | 1209 | } |
1182 | 1210 | ||
1183 | int nr_poll; | 1211 | int nr_poll; |
1184 | int group_fd; | 1212 | int group_fd; |
1185 | 1213 | ||
1186 | static void start_counter(int i, int counter) | 1214 | static void start_counter(int i, struct perf_evsel *evsel) |
1187 | { | 1215 | { |
1216 | struct xyarray *mmap_array = evsel->priv; | ||
1217 | struct mmap_data *mm; | ||
1188 | struct perf_event_attr *attr; | 1218 | struct perf_event_attr *attr; |
1189 | int cpu = -1; | 1219 | int cpu = -1; |
1190 | int thread_index; | 1220 | int thread_index; |
1191 | 1221 | ||
1192 | if (target_tid == -1) | 1222 | if (target_tid == -1) |
1193 | cpu = cpumap[i]; | 1223 | cpu = cpus->map[i]; |
1194 | 1224 | ||
1195 | attr = attrs + counter; | 1225 | attr = &evsel->attr; |
1196 | 1226 | ||
1197 | attr->sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID; | 1227 | attr->sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID; |
1198 | 1228 | ||
@@ -1205,16 +1235,18 @@ static void start_counter(int i, int counter) | |||
1205 | attr->inherit = (cpu < 0) && inherit; | 1235 | attr->inherit = (cpu < 0) && inherit; |
1206 | attr->mmap = 1; | 1236 | attr->mmap = 1; |
1207 | 1237 | ||
1208 | for (thread_index = 0; thread_index < thread_num; thread_index++) { | 1238 | for (thread_index = 0; thread_index < threads->nr; thread_index++) { |
1209 | try_again: | 1239 | try_again: |
1210 | fd[i][counter][thread_index] = sys_perf_event_open(attr, | 1240 | FD(evsel, i, thread_index) = sys_perf_event_open(attr, |
1211 | all_tids[thread_index], cpu, group_fd, 0); | 1241 | threads->map[thread_index], cpu, group_fd, 0); |
1212 | 1242 | ||
1213 | if (fd[i][counter][thread_index] < 0) { | 1243 | if (FD(evsel, i, thread_index) < 0) { |
1214 | int err = errno; | 1244 | int err = errno; |
1215 | 1245 | ||
1216 | if (err == EPERM || err == EACCES) | 1246 | if (err == EPERM || err == EACCES) |
1217 | die("No permission - are you root?\n"); | 1247 | die("Permission error - are you root?\n" |
1248 | "\t Consider tweaking" | ||
1249 | " /proc/sys/kernel/perf_event_paranoid.\n"); | ||
1218 | /* | 1250 | /* |
1219 | * If it's cycles then fall back to hrtimer | 1251 | * If it's cycles then fall back to hrtimer |
1220 | * based cpu-clock-tick sw counter, which | 1252 | * based cpu-clock-tick sw counter, which |
@@ -1231,30 +1263,30 @@ try_again: | |||
1231 | goto try_again; | 1263 | goto try_again; |
1232 | } | 1264 | } |
1233 | printf("\n"); | 1265 | printf("\n"); |
1234 | error("perfcounter syscall returned with %d (%s)\n", | 1266 | error("sys_perf_event_open() syscall returned with %d (%s). /bin/dmesg may provide additional information.\n", |
1235 | fd[i][counter][thread_index], strerror(err)); | 1267 | FD(evsel, i, thread_index), strerror(err)); |
1236 | die("No CONFIG_PERF_EVENTS=y kernel support configured?\n"); | 1268 | die("No CONFIG_PERF_EVENTS=y kernel support configured?\n"); |
1237 | exit(-1); | 1269 | exit(-1); |
1238 | } | 1270 | } |
1239 | assert(fd[i][counter][thread_index] >= 0); | 1271 | assert(FD(evsel, i, thread_index) >= 0); |
1240 | fcntl(fd[i][counter][thread_index], F_SETFL, O_NONBLOCK); | 1272 | fcntl(FD(evsel, i, thread_index), F_SETFL, O_NONBLOCK); |
1241 | 1273 | ||
1242 | /* | 1274 | /* |
1243 | * First counter acts as the group leader: | 1275 | * First counter acts as the group leader: |
1244 | */ | 1276 | */ |
1245 | if (group && group_fd == -1) | 1277 | if (group && group_fd == -1) |
1246 | group_fd = fd[i][counter][thread_index]; | 1278 | group_fd = FD(evsel, i, thread_index); |
1247 | 1279 | ||
1248 | event_array[nr_poll].fd = fd[i][counter][thread_index]; | 1280 | event_array[nr_poll].fd = FD(evsel, i, thread_index); |
1249 | event_array[nr_poll].events = POLLIN; | 1281 | event_array[nr_poll].events = POLLIN; |
1250 | nr_poll++; | 1282 | nr_poll++; |
1251 | 1283 | ||
1252 | mmap_array[i][counter][thread_index].counter = counter; | 1284 | mm = xyarray__entry(mmap_array, i, thread_index); |
1253 | mmap_array[i][counter][thread_index].prev = 0; | 1285 | mm->prev = 0; |
1254 | mmap_array[i][counter][thread_index].mask = mmap_pages*page_size - 1; | 1286 | mm->mask = mmap_pages*page_size - 1; |
1255 | mmap_array[i][counter][thread_index].base = mmap(NULL, (mmap_pages+1)*page_size, | 1287 | mm->base = mmap(NULL, (mmap_pages+1)*page_size, |
1256 | PROT_READ, MAP_SHARED, fd[i][counter][thread_index], 0); | 1288 | PROT_READ, MAP_SHARED, FD(evsel, i, thread_index), 0); |
1257 | if (mmap_array[i][counter][thread_index].base == MAP_FAILED) | 1289 | if (mm->base == MAP_FAILED) |
1258 | die("failed to mmap with %d (%s)\n", errno, strerror(errno)); | 1290 | die("failed to mmap with %d (%s)\n", errno, strerror(errno)); |
1259 | } | 1291 | } |
1260 | } | 1292 | } |
@@ -1262,13 +1294,13 @@ try_again: | |||
1262 | static int __cmd_top(void) | 1294 | static int __cmd_top(void) |
1263 | { | 1295 | { |
1264 | pthread_t thread; | 1296 | pthread_t thread; |
1265 | int i, counter; | 1297 | struct perf_evsel *counter; |
1266 | int ret; | 1298 | int i, ret; |
1267 | /* | 1299 | /* |
1268 | * FIXME: perf_session__new should allow passing a O_MMAP, so that all this | 1300 | * FIXME: perf_session__new should allow passing a O_MMAP, so that all this |
1269 | * mmap reading, etc is encapsulated in it. Use O_WRONLY for now. | 1301 | * mmap reading, etc is encapsulated in it. Use O_WRONLY for now. |
1270 | */ | 1302 | */ |
1271 | struct perf_session *session = perf_session__new(NULL, O_WRONLY, false, false); | 1303 | struct perf_session *session = perf_session__new(NULL, O_WRONLY, false, false, NULL); |
1272 | if (session == NULL) | 1304 | if (session == NULL) |
1273 | return -ENOMEM; | 1305 | return -ENOMEM; |
1274 | 1306 | ||
@@ -1277,9 +1309,9 @@ static int __cmd_top(void) | |||
1277 | else | 1309 | else |
1278 | event__synthesize_threads(event__process, session); | 1310 | event__synthesize_threads(event__process, session); |
1279 | 1311 | ||
1280 | for (i = 0; i < nr_cpus; i++) { | 1312 | for (i = 0; i < cpus->nr; i++) { |
1281 | group_fd = -1; | 1313 | group_fd = -1; |
1282 | for (counter = 0; counter < nr_counters; counter++) | 1314 | list_for_each_entry(counter, &evsel_list, node) |
1283 | start_counter(i, counter); | 1315 | start_counter(i, counter); |
1284 | } | 1316 | } |
1285 | 1317 | ||
@@ -1368,8 +1400,8 @@ static const struct option options[] = { | |||
1368 | 1400 | ||
1369 | int cmd_top(int argc, const char **argv, const char *prefix __used) | 1401 | int cmd_top(int argc, const char **argv, const char *prefix __used) |
1370 | { | 1402 | { |
1371 | int counter; | 1403 | struct perf_evsel *pos; |
1372 | int i,j; | 1404 | int status = -ENOMEM; |
1373 | 1405 | ||
1374 | page_size = sysconf(_SC_PAGE_SIZE); | 1406 | page_size = sysconf(_SC_PAGE_SIZE); |
1375 | 1407 | ||
@@ -1377,34 +1409,17 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) | |||
1377 | if (argc) | 1409 | if (argc) |
1378 | usage_with_options(top_usage, options); | 1410 | usage_with_options(top_usage, options); |
1379 | 1411 | ||
1380 | if (target_pid != -1) { | 1412 | if (target_pid != -1) |
1381 | target_tid = target_pid; | 1413 | target_tid = target_pid; |
1382 | thread_num = find_all_tid(target_pid, &all_tids); | ||
1383 | if (thread_num <= 0) { | ||
1384 | fprintf(stderr, "Can't find all threads of pid %d\n", | ||
1385 | target_pid); | ||
1386 | usage_with_options(top_usage, options); | ||
1387 | } | ||
1388 | } else { | ||
1389 | all_tids=malloc(sizeof(pid_t)); | ||
1390 | if (!all_tids) | ||
1391 | return -ENOMEM; | ||
1392 | 1414 | ||
1393 | all_tids[0] = target_tid; | 1415 | threads = thread_map__new(target_pid, target_tid); |
1394 | thread_num = 1; | 1416 | if (threads == NULL) { |
1417 | pr_err("Problems finding threads of monitor\n"); | ||
1418 | usage_with_options(top_usage, options); | ||
1395 | } | 1419 | } |
1396 | 1420 | ||
1397 | for (i = 0; i < MAX_NR_CPUS; i++) { | 1421 | event_array = malloc((sizeof(struct pollfd) * |
1398 | for (j = 0; j < MAX_COUNTERS; j++) { | 1422 | MAX_NR_CPUS * MAX_COUNTERS * threads->nr)); |
1399 | fd[i][j] = malloc(sizeof(int)*thread_num); | ||
1400 | mmap_array[i][j] = zalloc( | ||
1401 | sizeof(struct mmap_data)*thread_num); | ||
1402 | if (!fd[i][j] || !mmap_array[i][j]) | ||
1403 | return -ENOMEM; | ||
1404 | } | ||
1405 | } | ||
1406 | event_array = malloc( | ||
1407 | sizeof(struct pollfd)*MAX_NR_CPUS*MAX_COUNTERS*thread_num); | ||
1408 | if (!event_array) | 1423 | if (!event_array) |
1409 | return -ENOMEM; | 1424 | return -ENOMEM; |
1410 | 1425 | ||
@@ -1415,15 +1430,10 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) | |||
1415 | cpu_list = NULL; | 1430 | cpu_list = NULL; |
1416 | } | 1431 | } |
1417 | 1432 | ||
1418 | if (!nr_counters) | 1433 | if (!nr_counters && perf_evsel_list__create_default() < 0) { |
1419 | nr_counters = 1; | 1434 | pr_err("Not enough memory for event selector list\n"); |
1420 | 1435 | return -ENOMEM; | |
1421 | symbol_conf.priv_size = (sizeof(struct sym_entry) + | 1436 | } |
1422 | (nr_counters + 1) * sizeof(unsigned long)); | ||
1423 | |||
1424 | symbol_conf.try_vmlinux_path = (symbol_conf.vmlinux_name == NULL); | ||
1425 | if (symbol__init() < 0) | ||
1426 | return -1; | ||
1427 | 1437 | ||
1428 | if (delay_secs < 1) | 1438 | if (delay_secs < 1) |
1429 | delay_secs = 1; | 1439 | delay_secs = 1; |
@@ -1440,23 +1450,33 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) | |||
1440 | exit(EXIT_FAILURE); | 1450 | exit(EXIT_FAILURE); |
1441 | } | 1451 | } |
1442 | 1452 | ||
1443 | /* | 1453 | if (target_tid != -1) |
1444 | * Fill in the ones not specifically initialized via -c: | 1454 | cpus = cpu_map__dummy_new(); |
1445 | */ | 1455 | else |
1446 | for (counter = 0; counter < nr_counters; counter++) { | 1456 | cpus = cpu_map__new(cpu_list); |
1447 | if (attrs[counter].sample_period) | 1457 | |
1458 | if (cpus == NULL) | ||
1459 | usage_with_options(top_usage, options); | ||
1460 | |||
1461 | list_for_each_entry(pos, &evsel_list, node) { | ||
1462 | if (perf_evsel__alloc_mmap_per_thread(pos, cpus->nr, threads->nr) < 0 || | ||
1463 | perf_evsel__alloc_fd(pos, cpus->nr, threads->nr) < 0) | ||
1464 | goto out_free_fd; | ||
1465 | /* | ||
1466 | * Fill in the ones not specifically initialized via -c: | ||
1467 | */ | ||
1468 | if (pos->attr.sample_period) | ||
1448 | continue; | 1469 | continue; |
1449 | 1470 | ||
1450 | attrs[counter].sample_period = default_interval; | 1471 | pos->attr.sample_period = default_interval; |
1451 | } | 1472 | } |
1452 | 1473 | ||
1453 | if (target_tid != -1) | 1474 | symbol_conf.priv_size = (sizeof(struct sym_entry) + |
1454 | nr_cpus = 1; | 1475 | (nr_counters + 1) * sizeof(unsigned long)); |
1455 | else | ||
1456 | nr_cpus = read_cpu_map(cpu_list); | ||
1457 | 1476 | ||
1458 | if (nr_cpus < 1) | 1477 | symbol_conf.try_vmlinux_path = (symbol_conf.vmlinux_name == NULL); |
1459 | usage_with_options(top_usage, options); | 1478 | if (symbol__init() < 0) |
1479 | return -1; | ||
1460 | 1480 | ||
1461 | get_term_dimensions(&winsize); | 1481 | get_term_dimensions(&winsize); |
1462 | if (print_entries == 0) { | 1482 | if (print_entries == 0) { |
@@ -1464,5 +1484,10 @@ int cmd_top(int argc, const char **argv, const char *prefix __used) | |||
1464 | signal(SIGWINCH, sig_winch_handler); | 1484 | signal(SIGWINCH, sig_winch_handler); |
1465 | } | 1485 | } |
1466 | 1486 | ||
1467 | return __cmd_top(); | 1487 | status = __cmd_top(); |
1488 | out_free_fd: | ||
1489 | list_for_each_entry(pos, &evsel_list, node) | ||
1490 | perf_evsel__free_mmap(pos); | ||
1491 | |||
1492 | return status; | ||
1468 | } | 1493 | } |
diff --git a/tools/perf/builtin.h b/tools/perf/builtin.h index 921245b2858..c7798c7f24e 100644 --- a/tools/perf/builtin.h +++ b/tools/perf/builtin.h | |||
@@ -27,7 +27,7 @@ extern int cmd_report(int argc, const char **argv, const char *prefix); | |||
27 | extern int cmd_stat(int argc, const char **argv, const char *prefix); | 27 | extern int cmd_stat(int argc, const char **argv, const char *prefix); |
28 | extern int cmd_timechart(int argc, const char **argv, const char *prefix); | 28 | extern int cmd_timechart(int argc, const char **argv, const char *prefix); |
29 | extern int cmd_top(int argc, const char **argv, const char *prefix); | 29 | extern int cmd_top(int argc, const char **argv, const char *prefix); |
30 | extern int cmd_trace(int argc, const char **argv, const char *prefix); | 30 | extern int cmd_script(int argc, const char **argv, const char *prefix); |
31 | extern int cmd_version(int argc, const char **argv, const char *prefix); | 31 | extern int cmd_version(int argc, const char **argv, const char *prefix); |
32 | extern int cmd_probe(int argc, const char **argv, const char *prefix); | 32 | extern int cmd_probe(int argc, const char **argv, const char *prefix); |
33 | extern int cmd_kmem(int argc, const char **argv, const char *prefix); | 33 | extern int cmd_kmem(int argc, const char **argv, const char *prefix); |
diff --git a/tools/perf/command-list.txt b/tools/perf/command-list.txt index 949d77fc0b9..16b5088cf8f 100644 --- a/tools/perf/command-list.txt +++ b/tools/perf/command-list.txt | |||
@@ -16,7 +16,7 @@ perf-report mainporcelain common | |||
16 | perf-stat mainporcelain common | 16 | perf-stat mainporcelain common |
17 | perf-timechart mainporcelain common | 17 | perf-timechart mainporcelain common |
18 | perf-top mainporcelain common | 18 | perf-top mainporcelain common |
19 | perf-trace mainporcelain common | 19 | perf-script mainporcelain common |
20 | perf-probe mainporcelain common | 20 | perf-probe mainporcelain common |
21 | perf-kmem mainporcelain common | 21 | perf-kmem mainporcelain common |
22 | perf-lock mainporcelain common | 22 | perf-lock mainporcelain common |
diff --git a/tools/perf/feature-tests.mak b/tools/perf/feature-tests.mak index b253db634f0..b041ca67a2c 100644 --- a/tools/perf/feature-tests.mak +++ b/tools/perf/feature-tests.mak | |||
@@ -9,8 +9,8 @@ endef | |||
9 | ifndef NO_DWARF | 9 | ifndef NO_DWARF |
10 | define SOURCE_DWARF | 10 | define SOURCE_DWARF |
11 | #include <dwarf.h> | 11 | #include <dwarf.h> |
12 | #include <libdw.h> | 12 | #include <elfutils/libdw.h> |
13 | #include <version.h> | 13 | #include <elfutils/version.h> |
14 | #ifndef _ELFUTILS_PREREQ | 14 | #ifndef _ELFUTILS_PREREQ |
15 | #error | 15 | #error |
16 | #endif | 16 | #endif |
diff --git a/tools/perf/perf.c b/tools/perf/perf.c index cdd6c03f1e1..5b1ecd66bb3 100644 --- a/tools/perf/perf.c +++ b/tools/perf/perf.c | |||
@@ -286,6 +286,8 @@ static int run_builtin(struct cmd_struct *p, int argc, const char **argv) | |||
286 | status = p->fn(argc, argv, prefix); | 286 | status = p->fn(argc, argv, prefix); |
287 | exit_browser(status); | 287 | exit_browser(status); |
288 | 288 | ||
289 | perf_evsel_list__delete(); | ||
290 | |||
289 | if (status) | 291 | if (status) |
290 | return status & 0xff; | 292 | return status & 0xff; |
291 | 293 | ||
@@ -323,7 +325,7 @@ static void handle_internal_command(int argc, const char **argv) | |||
323 | { "top", cmd_top, 0 }, | 325 | { "top", cmd_top, 0 }, |
324 | { "annotate", cmd_annotate, 0 }, | 326 | { "annotate", cmd_annotate, 0 }, |
325 | { "version", cmd_version, 0 }, | 327 | { "version", cmd_version, 0 }, |
326 | { "trace", cmd_trace, 0 }, | 328 | { "script", cmd_script, 0 }, |
327 | { "sched", cmd_sched, 0 }, | 329 | { "sched", cmd_sched, 0 }, |
328 | { "probe", cmd_probe, 0 }, | 330 | { "probe", cmd_probe, 0 }, |
329 | { "kmem", cmd_kmem, 0 }, | 331 | { "kmem", cmd_kmem, 0 }, |
diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/Context.c b/tools/perf/scripts/perl/Perf-Trace-Util/Context.c index 01a64ad693f..790ceba6ad3 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/Context.c +++ b/tools/perf/scripts/perl/Perf-Trace-Util/Context.c | |||
@@ -8,7 +8,7 @@ | |||
8 | 8 | ||
9 | #line 1 "Context.xs" | 9 | #line 1 "Context.xs" |
10 | /* | 10 | /* |
11 | * Context.xs. XS interfaces for perf trace. | 11 | * Context.xs. XS interfaces for perf script. |
12 | * | 12 | * |
13 | * Copyright (C) 2009 Tom Zanussi <tzanussi@gmail.com> | 13 | * Copyright (C) 2009 Tom Zanussi <tzanussi@gmail.com> |
14 | * | 14 | * |
diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/Context.xs b/tools/perf/scripts/perl/Perf-Trace-Util/Context.xs index 549cf0467d3..c1e2ed1ed34 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/Context.xs +++ b/tools/perf/scripts/perl/Perf-Trace-Util/Context.xs | |||
@@ -1,5 +1,5 @@ | |||
1 | /* | 1 | /* |
2 | * Context.xs. XS interfaces for perf trace. | 2 | * Context.xs. XS interfaces for perf script. |
3 | * | 3 | * |
4 | * Copyright (C) 2009 Tom Zanussi <tzanussi@gmail.com> | 4 | * Copyright (C) 2009 Tom Zanussi <tzanussi@gmail.com> |
5 | * | 5 | * |
@@ -23,7 +23,7 @@ | |||
23 | #include "perl.h" | 23 | #include "perl.h" |
24 | #include "XSUB.h" | 24 | #include "XSUB.h" |
25 | #include "../../../perf.h" | 25 | #include "../../../perf.h" |
26 | #include "../../../util/trace-event.h" | 26 | #include "../../../util/script-event.h" |
27 | 27 | ||
28 | MODULE = Perf::Trace::Context PACKAGE = Perf::Trace::Context | 28 | MODULE = Perf::Trace::Context PACKAGE = Perf::Trace::Context |
29 | PROTOTYPES: ENABLE | 29 | PROTOTYPES: ENABLE |
diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/README b/tools/perf/scripts/perl/Perf-Trace-Util/README index 9a970763079..2f0c7f3043e 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/README +++ b/tools/perf/scripts/perl/Perf-Trace-Util/README | |||
@@ -1,7 +1,7 @@ | |||
1 | Perf-Trace-Util version 0.01 | 1 | Perf-Trace-Util version 0.01 |
2 | ============================ | 2 | ============================ |
3 | 3 | ||
4 | This module contains utility functions for use with perf trace. | 4 | This module contains utility functions for use with perf script. |
5 | 5 | ||
6 | Core.pm and Util.pm are pure Perl modules; Core.pm contains routines | 6 | Core.pm and Util.pm are pure Perl modules; Core.pm contains routines |
7 | that the core perf support for Perl calls on and should always be | 7 | that the core perf support for Perl calls on and should always be |
@@ -33,7 +33,7 @@ After you do that: | |||
33 | 33 | ||
34 | INSTALLATION | 34 | INSTALLATION |
35 | 35 | ||
36 | Building perf with perf trace Perl scripting should install this | 36 | Building perf with perf script Perl scripting should install this |
37 | module in the right place. | 37 | module in the right place. |
38 | 38 | ||
39 | You should make sure libperl and ExtUtils/Embed.pm are installed first | 39 | You should make sure libperl and ExtUtils/Embed.pm are installed first |
diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Context.pm b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Context.pm index 6c7f3659cb1..4e2f6039ac9 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Context.pm +++ b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Context.pm | |||
@@ -34,7 +34,7 @@ Perf::Trace::Context - Perl extension for accessing functions in perf. | |||
34 | 34 | ||
35 | =head1 SEE ALSO | 35 | =head1 SEE ALSO |
36 | 36 | ||
37 | Perf (trace) documentation | 37 | Perf (script) documentation |
38 | 38 | ||
39 | =head1 AUTHOR | 39 | =head1 AUTHOR |
40 | 40 | ||
diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm index 9df376a9f62..9158458d3ee 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm +++ b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Core.pm | |||
@@ -163,7 +163,7 @@ sub dump_symbolic_fields | |||
163 | __END__ | 163 | __END__ |
164 | =head1 NAME | 164 | =head1 NAME |
165 | 165 | ||
166 | Perf::Trace::Core - Perl extension for perf trace | 166 | Perf::Trace::Core - Perl extension for perf script |
167 | 167 | ||
168 | =head1 SYNOPSIS | 168 | =head1 SYNOPSIS |
169 | 169 | ||
@@ -171,7 +171,7 @@ Perf::Trace::Core - Perl extension for perf trace | |||
171 | 171 | ||
172 | =head1 SEE ALSO | 172 | =head1 SEE ALSO |
173 | 173 | ||
174 | Perf (trace) documentation | 174 | Perf (script) documentation |
175 | 175 | ||
176 | =head1 AUTHOR | 176 | =head1 AUTHOR |
177 | 177 | ||
diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Util.pm b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Util.pm index d94b40c8ac8..05350011462 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Util.pm +++ b/tools/perf/scripts/perl/Perf-Trace-Util/lib/Perf/Trace/Util.pm | |||
@@ -65,7 +65,7 @@ sub clear_term | |||
65 | __END__ | 65 | __END__ |
66 | =head1 NAME | 66 | =head1 NAME |
67 | 67 | ||
68 | Perf::Trace::Util - Perl extension for perf trace | 68 | Perf::Trace::Util - Perl extension for perf script |
69 | 69 | ||
70 | =head1 SYNOPSIS | 70 | =head1 SYNOPSIS |
71 | 71 | ||
@@ -73,7 +73,7 @@ Perf::Trace::Util - Perl extension for perf trace | |||
73 | 73 | ||
74 | =head1 SEE ALSO | 74 | =head1 SEE ALSO |
75 | 75 | ||
76 | Perf (trace) documentation | 76 | Perf (script) documentation |
77 | 77 | ||
78 | =head1 AUTHOR | 78 | =head1 AUTHOR |
79 | 79 | ||
diff --git a/tools/perf/scripts/perl/bin/failed-syscalls-report b/tools/perf/scripts/perl/bin/failed-syscalls-report index 4028d92dc4a..9f83cc1ad8b 100644 --- a/tools/perf/scripts/perl/bin/failed-syscalls-report +++ b/tools/perf/scripts/perl/bin/failed-syscalls-report | |||
@@ -7,4 +7,4 @@ if [ $# -gt 0 ] ; then | |||
7 | shift | 7 | shift |
8 | fi | 8 | fi |
9 | fi | 9 | fi |
10 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/failed-syscalls.pl $comm | 10 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/perl/failed-syscalls.pl $comm |
diff --git a/tools/perf/scripts/perl/bin/rw-by-file-report b/tools/perf/scripts/perl/bin/rw-by-file-report index ba25f4d41fb..77200b3f310 100644 --- a/tools/perf/scripts/perl/bin/rw-by-file-report +++ b/tools/perf/scripts/perl/bin/rw-by-file-report | |||
@@ -7,7 +7,4 @@ if [ $# -lt 1 ] ; then | |||
7 | fi | 7 | fi |
8 | comm=$1 | 8 | comm=$1 |
9 | shift | 9 | shift |
10 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/rw-by-file.pl $comm | 10 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/perl/rw-by-file.pl $comm |
11 | |||
12 | |||
13 | |||
diff --git a/tools/perf/scripts/perl/bin/rw-by-pid-report b/tools/perf/scripts/perl/bin/rw-by-pid-report index 641a3f5d085..a27b9f311f9 100644 --- a/tools/perf/scripts/perl/bin/rw-by-pid-report +++ b/tools/perf/scripts/perl/bin/rw-by-pid-report | |||
@@ -1,6 +1,3 @@ | |||
1 | #!/bin/bash | 1 | #!/bin/bash |
2 | # description: system-wide r/w activity | 2 | # description: system-wide r/w activity |
3 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/rw-by-pid.pl | 3 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/perl/rw-by-pid.pl |
4 | |||
5 | |||
6 | |||
diff --git a/tools/perf/scripts/perl/bin/rwtop-report b/tools/perf/scripts/perl/bin/rwtop-report index 4918dba7702..83e11ec2e19 100644 --- a/tools/perf/scripts/perl/bin/rwtop-report +++ b/tools/perf/scripts/perl/bin/rwtop-report | |||
@@ -17,7 +17,4 @@ if [ "$n_args" -gt 0 ] ; then | |||
17 | interval=$1 | 17 | interval=$1 |
18 | shift | 18 | shift |
19 | fi | 19 | fi |
20 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/rwtop.pl $interval | 20 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/perl/rwtop.pl $interval |
21 | |||
22 | |||
23 | |||
diff --git a/tools/perf/scripts/perl/bin/wakeup-latency-report b/tools/perf/scripts/perl/bin/wakeup-latency-report index 49052ebcb63..889e8130cca 100644 --- a/tools/perf/scripts/perl/bin/wakeup-latency-report +++ b/tools/perf/scripts/perl/bin/wakeup-latency-report | |||
@@ -1,6 +1,3 @@ | |||
1 | #!/bin/bash | 1 | #!/bin/bash |
2 | # description: system-wide min/max/avg wakeup latency | 2 | # description: system-wide min/max/avg wakeup latency |
3 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/wakeup-latency.pl | 3 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/perl/wakeup-latency.pl |
4 | |||
5 | |||
6 | |||
diff --git a/tools/perf/scripts/perl/bin/workqueue-stats-report b/tools/perf/scripts/perl/bin/workqueue-stats-report index df0c65f4ca9..6d91411d248 100644 --- a/tools/perf/scripts/perl/bin/workqueue-stats-report +++ b/tools/perf/scripts/perl/bin/workqueue-stats-report | |||
@@ -1,7 +1,3 @@ | |||
1 | #!/bin/bash | 1 | #!/bin/bash |
2 | # description: workqueue stats (ins/exe/create/destroy) | 2 | # description: workqueue stats (ins/exe/create/destroy) |
3 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/perl/workqueue-stats.pl | 3 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/perl/workqueue-stats.pl |
4 | |||
5 | |||
6 | |||
7 | |||
diff --git a/tools/perf/scripts/perl/check-perf-trace.pl b/tools/perf/scripts/perl/check-perf-trace.pl index 4e7dc0a407a..4e7076c2061 100644 --- a/tools/perf/scripts/perl/check-perf-trace.pl +++ b/tools/perf/scripts/perl/check-perf-trace.pl | |||
@@ -1,4 +1,4 @@ | |||
1 | # perf trace event handlers, generated by perf trace -g perl | 1 | # perf script event handlers, generated by perf script -g perl |
2 | # (c) 2009, Tom Zanussi <tzanussi@gmail.com> | 2 | # (c) 2009, Tom Zanussi <tzanussi@gmail.com> |
3 | # Licensed under the terms of the GNU GPL License version 2 | 3 | # Licensed under the terms of the GNU GPL License version 2 |
4 | 4 | ||
diff --git a/tools/perf/scripts/perl/rw-by-file.pl b/tools/perf/scripts/perl/rw-by-file.pl index 2a39097687b..74844ee2be3 100644 --- a/tools/perf/scripts/perl/rw-by-file.pl +++ b/tools/perf/scripts/perl/rw-by-file.pl | |||
@@ -18,7 +18,7 @@ use lib "./Perf-Trace-Util/lib"; | |||
18 | use Perf::Trace::Core; | 18 | use Perf::Trace::Core; |
19 | use Perf::Trace::Util; | 19 | use Perf::Trace::Util; |
20 | 20 | ||
21 | my $usage = "perf trace -s rw-by-file.pl <comm>\n"; | 21 | my $usage = "perf script -s rw-by-file.pl <comm>\n"; |
22 | 22 | ||
23 | my $for_comm = shift or die $usage; | 23 | my $for_comm = shift or die $usage; |
24 | 24 | ||
diff --git a/tools/perf/scripts/perl/workqueue-stats.pl b/tools/perf/scripts/perl/workqueue-stats.pl index b84b12699b7..a8eaff5119e 100644 --- a/tools/perf/scripts/perl/workqueue-stats.pl +++ b/tools/perf/scripts/perl/workqueue-stats.pl | |||
@@ -10,7 +10,7 @@ | |||
10 | # workqueue:workqueue_destruction -e workqueue:workqueue_execution | 10 | # workqueue:workqueue_destruction -e workqueue:workqueue_execution |
11 | # -e workqueue:workqueue_insertion | 11 | # -e workqueue:workqueue_insertion |
12 | # | 12 | # |
13 | # perf trace -p -s tools/perf/scripts/perl/workqueue-stats.pl | 13 | # perf script -p -s tools/perf/scripts/perl/workqueue-stats.pl |
14 | 14 | ||
15 | use 5.010000; | 15 | use 5.010000; |
16 | use strict; | 16 | use strict; |
diff --git a/tools/perf/scripts/python/Perf-Trace-Util/Context.c b/tools/perf/scripts/python/Perf-Trace-Util/Context.c index 957085dd5d8..315067b8f55 100644 --- a/tools/perf/scripts/python/Perf-Trace-Util/Context.c +++ b/tools/perf/scripts/python/Perf-Trace-Util/Context.c | |||
@@ -1,5 +1,5 @@ | |||
1 | /* | 1 | /* |
2 | * Context.c. Python interfaces for perf trace. | 2 | * Context.c. Python interfaces for perf script. |
3 | * | 3 | * |
4 | * Copyright (C) 2010 Tom Zanussi <tzanussi@gmail.com> | 4 | * Copyright (C) 2010 Tom Zanussi <tzanussi@gmail.com> |
5 | * | 5 | * |
diff --git a/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Core.py b/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Core.py index aad7525bca1..de7211e4fa4 100644 --- a/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Core.py +++ b/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Core.py | |||
@@ -1,4 +1,4 @@ | |||
1 | # Core.py - Python extension for perf trace, core functions | 1 | # Core.py - Python extension for perf script, core functions |
2 | # | 2 | # |
3 | # Copyright (C) 2010 by Tom Zanussi <tzanussi@gmail.com> | 3 | # Copyright (C) 2010 by Tom Zanussi <tzanussi@gmail.com> |
4 | # | 4 | # |
diff --git a/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/SchedGui.py b/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/SchedGui.py index ae9a56e43e0..fdd92f69905 100644 --- a/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/SchedGui.py +++ b/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/SchedGui.py | |||
@@ -1,4 +1,4 @@ | |||
1 | # SchedGui.py - Python extension for perf trace, basic GUI code for | 1 | # SchedGui.py - Python extension for perf script, basic GUI code for |
2 | # traces drawing and overview. | 2 | # traces drawing and overview. |
3 | # | 3 | # |
4 | # Copyright (C) 2010 by Frederic Weisbecker <fweisbec@gmail.com> | 4 | # Copyright (C) 2010 by Frederic Weisbecker <fweisbec@gmail.com> |
diff --git a/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Util.py b/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Util.py index 13cc02b5893..15c8400240f 100644 --- a/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Util.py +++ b/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Util.py | |||
@@ -1,4 +1,4 @@ | |||
1 | # Util.py - Python extension for perf trace, miscellaneous utility code | 1 | # Util.py - Python extension for perf script, miscellaneous utility code |
2 | # | 2 | # |
3 | # Copyright (C) 2010 by Tom Zanussi <tzanussi@gmail.com> | 3 | # Copyright (C) 2010 by Tom Zanussi <tzanussi@gmail.com> |
4 | # | 4 | # |
diff --git a/tools/perf/scripts/python/bin/failed-syscalls-by-pid-report b/tools/perf/scripts/python/bin/failed-syscalls-by-pid-report index 03587021463..fda5096d0cb 100644 --- a/tools/perf/scripts/python/bin/failed-syscalls-by-pid-report +++ b/tools/perf/scripts/python/bin/failed-syscalls-by-pid-report | |||
@@ -7,4 +7,4 @@ if [ $# -gt 0 ] ; then | |||
7 | shift | 7 | shift |
8 | fi | 8 | fi |
9 | fi | 9 | fi |
10 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/failed-syscalls-by-pid.py $comm | 10 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/python/failed-syscalls-by-pid.py $comm |
diff --git a/tools/perf/scripts/python/bin/futex-contention-report b/tools/perf/scripts/python/bin/futex-contention-report index c8268138fb7..6c44271091a 100644 --- a/tools/perf/scripts/python/bin/futex-contention-report +++ b/tools/perf/scripts/python/bin/futex-contention-report | |||
@@ -1,4 +1,4 @@ | |||
1 | #!/bin/bash | 1 | #!/bin/bash |
2 | # description: futext contention measurement | 2 | # description: futext contention measurement |
3 | 3 | ||
4 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/futex-contention.py | 4 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/python/futex-contention.py |
diff --git a/tools/perf/scripts/python/bin/netdev-times-report b/tools/perf/scripts/python/bin/netdev-times-report index 4ad361b3124..8f759291da8 100644 --- a/tools/perf/scripts/python/bin/netdev-times-report +++ b/tools/perf/scripts/python/bin/netdev-times-report | |||
@@ -2,4 +2,4 @@ | |||
2 | # description: display a process of packet and processing time | 2 | # description: display a process of packet and processing time |
3 | # args: [tx] [rx] [dev=] [debug] | 3 | # args: [tx] [rx] [dev=] [debug] |
4 | 4 | ||
5 | perf trace -s "$PERF_EXEC_PATH"/scripts/python/netdev-times.py $@ | 5 | perf script -s "$PERF_EXEC_PATH"/scripts/python/netdev-times.py $@ |
diff --git a/tools/perf/scripts/python/bin/sched-migration-report b/tools/perf/scripts/python/bin/sched-migration-report index df1791f07c2..68b037a1849 100644 --- a/tools/perf/scripts/python/bin/sched-migration-report +++ b/tools/perf/scripts/python/bin/sched-migration-report | |||
@@ -1,3 +1,3 @@ | |||
1 | #!/bin/bash | 1 | #!/bin/bash |
2 | # description: sched migration overview | 2 | # description: sched migration overview |
3 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/sched-migration.py | 3 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/python/sched-migration.py |
diff --git a/tools/perf/scripts/python/bin/sctop-report b/tools/perf/scripts/python/bin/sctop-report index 36b409c05e5..c32db294124 100644 --- a/tools/perf/scripts/python/bin/sctop-report +++ b/tools/perf/scripts/python/bin/sctop-report | |||
@@ -21,4 +21,4 @@ elif [ "$n_args" -gt 0 ] ; then | |||
21 | interval=$1 | 21 | interval=$1 |
22 | shift | 22 | shift |
23 | fi | 23 | fi |
24 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/sctop.py $comm $interval | 24 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/python/sctop.py $comm $interval |
diff --git a/tools/perf/scripts/python/bin/syscall-counts-by-pid-report b/tools/perf/scripts/python/bin/syscall-counts-by-pid-report index 4eb88c9fc83..16eb8d65c54 100644 --- a/tools/perf/scripts/python/bin/syscall-counts-by-pid-report +++ b/tools/perf/scripts/python/bin/syscall-counts-by-pid-report | |||
@@ -7,4 +7,4 @@ if [ $# -gt 0 ] ; then | |||
7 | shift | 7 | shift |
8 | fi | 8 | fi |
9 | fi | 9 | fi |
10 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/syscall-counts-by-pid.py $comm | 10 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/python/syscall-counts-by-pid.py $comm |
diff --git a/tools/perf/scripts/python/bin/syscall-counts-report b/tools/perf/scripts/python/bin/syscall-counts-report index cb2f9c5cf17..0f0e9d453bb 100644 --- a/tools/perf/scripts/python/bin/syscall-counts-report +++ b/tools/perf/scripts/python/bin/syscall-counts-report | |||
@@ -7,4 +7,4 @@ if [ $# -gt 0 ] ; then | |||
7 | shift | 7 | shift |
8 | fi | 8 | fi |
9 | fi | 9 | fi |
10 | perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/syscall-counts.py $comm | 10 | perf script $@ -s "$PERF_EXEC_PATH"/scripts/python/syscall-counts.py $comm |
diff --git a/tools/perf/scripts/python/check-perf-trace.py b/tools/perf/scripts/python/check-perf-trace.py index d9f7893e315..4647a7694cf 100644 --- a/tools/perf/scripts/python/check-perf-trace.py +++ b/tools/perf/scripts/python/check-perf-trace.py | |||
@@ -1,4 +1,4 @@ | |||
1 | # perf trace event handlers, generated by perf trace -g python | 1 | # perf script event handlers, generated by perf script -g python |
2 | # (c) 2010, Tom Zanussi <tzanussi@gmail.com> | 2 | # (c) 2010, Tom Zanussi <tzanussi@gmail.com> |
3 | # Licensed under the terms of the GNU GPL License version 2 | 3 | # Licensed under the terms of the GNU GPL License version 2 |
4 | # | 4 | # |
diff --git a/tools/perf/scripts/python/failed-syscalls-by-pid.py b/tools/perf/scripts/python/failed-syscalls-by-pid.py index acd7848717b..85805fac411 100644 --- a/tools/perf/scripts/python/failed-syscalls-by-pid.py +++ b/tools/perf/scripts/python/failed-syscalls-by-pid.py | |||
@@ -15,7 +15,7 @@ from perf_trace_context import * | |||
15 | from Core import * | 15 | from Core import * |
16 | from Util import * | 16 | from Util import * |
17 | 17 | ||
18 | usage = "perf trace -s syscall-counts-by-pid.py [comm|pid]\n"; | 18 | usage = "perf script -s syscall-counts-by-pid.py [comm|pid]\n"; |
19 | 19 | ||
20 | for_comm = None | 20 | for_comm = None |
21 | for_pid = None | 21 | for_pid = None |
diff --git a/tools/perf/scripts/python/sched-migration.py b/tools/perf/scripts/python/sched-migration.py index b934383c336..74d55ec08ae 100644 --- a/tools/perf/scripts/python/sched-migration.py +++ b/tools/perf/scripts/python/sched-migration.py | |||
@@ -4,7 +4,7 @@ | |||
4 | # | 4 | # |
5 | # Copyright (C) 2010 Frederic Weisbecker <fweisbec@gmail.com> | 5 | # Copyright (C) 2010 Frederic Weisbecker <fweisbec@gmail.com> |
6 | # | 6 | # |
7 | # perf trace event handlers have been generated by perf trace -g python | 7 | # perf script event handlers have been generated by perf script -g python |
8 | # | 8 | # |
9 | # This software is distributed under the terms of the GNU General | 9 | # This software is distributed under the terms of the GNU General |
10 | # Public License ("GPL") version 2 as published by the Free Software | 10 | # Public License ("GPL") version 2 as published by the Free Software |
diff --git a/tools/perf/scripts/python/sctop.py b/tools/perf/scripts/python/sctop.py index 7a6ec2c7d8a..42c267e292f 100644 --- a/tools/perf/scripts/python/sctop.py +++ b/tools/perf/scripts/python/sctop.py | |||
@@ -17,7 +17,7 @@ from perf_trace_context import * | |||
17 | from Core import * | 17 | from Core import * |
18 | from Util import * | 18 | from Util import * |
19 | 19 | ||
20 | usage = "perf trace -s sctop.py [comm] [interval]\n"; | 20 | usage = "perf script -s sctop.py [comm] [interval]\n"; |
21 | 21 | ||
22 | for_comm = None | 22 | for_comm = None |
23 | default_interval = 3 | 23 | default_interval = 3 |
diff --git a/tools/perf/scripts/python/syscall-counts-by-pid.py b/tools/perf/scripts/python/syscall-counts-by-pid.py index d1ee3ec10cf..c64d1c55d74 100644 --- a/tools/perf/scripts/python/syscall-counts-by-pid.py +++ b/tools/perf/scripts/python/syscall-counts-by-pid.py | |||
@@ -14,7 +14,7 @@ from perf_trace_context import * | |||
14 | from Core import * | 14 | from Core import * |
15 | from Util import syscall_name | 15 | from Util import syscall_name |
16 | 16 | ||
17 | usage = "perf trace -s syscall-counts-by-pid.py [comm]\n"; | 17 | usage = "perf script -s syscall-counts-by-pid.py [comm]\n"; |
18 | 18 | ||
19 | for_comm = None | 19 | for_comm = None |
20 | for_pid = None | 20 | for_pid = None |
diff --git a/tools/perf/scripts/python/syscall-counts.py b/tools/perf/scripts/python/syscall-counts.py index ea183dc82d2..b435d3f188e 100644 --- a/tools/perf/scripts/python/syscall-counts.py +++ b/tools/perf/scripts/python/syscall-counts.py | |||
@@ -15,7 +15,7 @@ from perf_trace_context import * | |||
15 | from Core import * | 15 | from Core import * |
16 | from Util import syscall_name | 16 | from Util import syscall_name |
17 | 17 | ||
18 | usage = "perf trace -s syscall-counts.py [comm]\n"; | 18 | usage = "perf script -s syscall-counts.py [comm]\n"; |
19 | 19 | ||
20 | for_comm = None | 20 | for_comm = None |
21 | 21 | ||
diff --git a/tools/perf/util/build-id.c b/tools/perf/util/build-id.c index e437edb7241..deffb8c9607 100644 --- a/tools/perf/util/build-id.c +++ b/tools/perf/util/build-id.c | |||
@@ -14,7 +14,9 @@ | |||
14 | #include <linux/kernel.h> | 14 | #include <linux/kernel.h> |
15 | #include "debug.h" | 15 | #include "debug.h" |
16 | 16 | ||
17 | static int build_id__mark_dso_hit(event_t *event, struct perf_session *session) | 17 | static int build_id__mark_dso_hit(event_t *event, |
18 | struct sample_data *sample __used, | ||
19 | struct perf_session *session) | ||
18 | { | 20 | { |
19 | struct addr_location al; | 21 | struct addr_location al; |
20 | u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; | 22 | u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; |
@@ -35,7 +37,8 @@ static int build_id__mark_dso_hit(event_t *event, struct perf_session *session) | |||
35 | return 0; | 37 | return 0; |
36 | } | 38 | } |
37 | 39 | ||
38 | static int event__exit_del_thread(event_t *self, struct perf_session *session) | 40 | static int event__exit_del_thread(event_t *self, struct sample_data *sample __used, |
41 | struct perf_session *session) | ||
39 | { | 42 | { |
40 | struct thread *thread = perf_session__findnew(session, self->fork.tid); | 43 | struct thread *thread = perf_session__findnew(session, self->fork.tid); |
41 | 44 | ||
diff --git a/tools/perf/util/cpumap.c b/tools/perf/util/cpumap.c index 0f9b8d7a7d7..3ccaa104338 100644 --- a/tools/perf/util/cpumap.c +++ b/tools/perf/util/cpumap.c | |||
@@ -4,32 +4,53 @@ | |||
4 | #include <assert.h> | 4 | #include <assert.h> |
5 | #include <stdio.h> | 5 | #include <stdio.h> |
6 | 6 | ||
7 | int cpumap[MAX_NR_CPUS]; | 7 | static struct cpu_map *cpu_map__default_new(void) |
8 | |||
9 | static int default_cpu_map(void) | ||
10 | { | 8 | { |
11 | int nr_cpus, i; | 9 | struct cpu_map *cpus; |
10 | int nr_cpus; | ||
12 | 11 | ||
13 | nr_cpus = sysconf(_SC_NPROCESSORS_ONLN); | 12 | nr_cpus = sysconf(_SC_NPROCESSORS_ONLN); |
14 | assert(nr_cpus <= MAX_NR_CPUS); | 13 | if (nr_cpus < 0) |
15 | assert((int)nr_cpus >= 0); | 14 | return NULL; |
15 | |||
16 | cpus = malloc(sizeof(*cpus) + nr_cpus * sizeof(int)); | ||
17 | if (cpus != NULL) { | ||
18 | int i; | ||
19 | for (i = 0; i < nr_cpus; ++i) | ||
20 | cpus->map[i] = i; | ||
16 | 21 | ||
17 | for (i = 0; i < nr_cpus; ++i) | 22 | cpus->nr = nr_cpus; |
18 | cpumap[i] = i; | 23 | } |
19 | 24 | ||
20 | return nr_cpus; | 25 | return cpus; |
21 | } | 26 | } |
22 | 27 | ||
23 | static int read_all_cpu_map(void) | 28 | static struct cpu_map *cpu_map__trim_new(int nr_cpus, int *tmp_cpus) |
24 | { | 29 | { |
30 | size_t payload_size = nr_cpus * sizeof(int); | ||
31 | struct cpu_map *cpus = malloc(sizeof(*cpus) + payload_size); | ||
32 | |||
33 | if (cpus != NULL) { | ||
34 | cpus->nr = nr_cpus; | ||
35 | memcpy(cpus->map, tmp_cpus, payload_size); | ||
36 | } | ||
37 | |||
38 | return cpus; | ||
39 | } | ||
40 | |||
41 | static struct cpu_map *cpu_map__read_all_cpu_map(void) | ||
42 | { | ||
43 | struct cpu_map *cpus = NULL; | ||
25 | FILE *onlnf; | 44 | FILE *onlnf; |
26 | int nr_cpus = 0; | 45 | int nr_cpus = 0; |
46 | int *tmp_cpus = NULL, *tmp; | ||
47 | int max_entries = 0; | ||
27 | int n, cpu, prev; | 48 | int n, cpu, prev; |
28 | char sep; | 49 | char sep; |
29 | 50 | ||
30 | onlnf = fopen("/sys/devices/system/cpu/online", "r"); | 51 | onlnf = fopen("/sys/devices/system/cpu/online", "r"); |
31 | if (!onlnf) | 52 | if (!onlnf) |
32 | return default_cpu_map(); | 53 | return cpu_map__default_new(); |
33 | 54 | ||
34 | sep = 0; | 55 | sep = 0; |
35 | prev = -1; | 56 | prev = -1; |
@@ -38,12 +59,28 @@ static int read_all_cpu_map(void) | |||
38 | if (n <= 0) | 59 | if (n <= 0) |
39 | break; | 60 | break; |
40 | if (prev >= 0) { | 61 | if (prev >= 0) { |
41 | assert(nr_cpus + cpu - prev - 1 < MAX_NR_CPUS); | 62 | int new_max = nr_cpus + cpu - prev - 1; |
63 | |||
64 | if (new_max >= max_entries) { | ||
65 | max_entries = new_max + MAX_NR_CPUS / 2; | ||
66 | tmp = realloc(tmp_cpus, max_entries * sizeof(int)); | ||
67 | if (tmp == NULL) | ||
68 | goto out_free_tmp; | ||
69 | tmp_cpus = tmp; | ||
70 | } | ||
71 | |||
42 | while (++prev < cpu) | 72 | while (++prev < cpu) |
43 | cpumap[nr_cpus++] = prev; | 73 | tmp_cpus[nr_cpus++] = prev; |
74 | } | ||
75 | if (nr_cpus == max_entries) { | ||
76 | max_entries += MAX_NR_CPUS; | ||
77 | tmp = realloc(tmp_cpus, max_entries * sizeof(int)); | ||
78 | if (tmp == NULL) | ||
79 | goto out_free_tmp; | ||
80 | tmp_cpus = tmp; | ||
44 | } | 81 | } |
45 | assert (nr_cpus < MAX_NR_CPUS); | 82 | |
46 | cpumap[nr_cpus++] = cpu; | 83 | tmp_cpus[nr_cpus++] = cpu; |
47 | if (n == 2 && sep == '-') | 84 | if (n == 2 && sep == '-') |
48 | prev = cpu; | 85 | prev = cpu; |
49 | else | 86 | else |
@@ -51,24 +88,31 @@ static int read_all_cpu_map(void) | |||
51 | if (n == 1 || sep == '\n') | 88 | if (n == 1 || sep == '\n') |
52 | break; | 89 | break; |
53 | } | 90 | } |
54 | fclose(onlnf); | ||
55 | if (nr_cpus > 0) | ||
56 | return nr_cpus; | ||
57 | 91 | ||
58 | return default_cpu_map(); | 92 | if (nr_cpus > 0) |
93 | cpus = cpu_map__trim_new(nr_cpus, tmp_cpus); | ||
94 | else | ||
95 | cpus = cpu_map__default_new(); | ||
96 | out_free_tmp: | ||
97 | free(tmp_cpus); | ||
98 | fclose(onlnf); | ||
99 | return cpus; | ||
59 | } | 100 | } |
60 | 101 | ||
61 | int read_cpu_map(const char *cpu_list) | 102 | struct cpu_map *cpu_map__new(const char *cpu_list) |
62 | { | 103 | { |
104 | struct cpu_map *cpus = NULL; | ||
63 | unsigned long start_cpu, end_cpu = 0; | 105 | unsigned long start_cpu, end_cpu = 0; |
64 | char *p = NULL; | 106 | char *p = NULL; |
65 | int i, nr_cpus = 0; | 107 | int i, nr_cpus = 0; |
108 | int *tmp_cpus = NULL, *tmp; | ||
109 | int max_entries = 0; | ||
66 | 110 | ||
67 | if (!cpu_list) | 111 | if (!cpu_list) |
68 | return read_all_cpu_map(); | 112 | return cpu_map__read_all_cpu_map(); |
69 | 113 | ||
70 | if (!isdigit(*cpu_list)) | 114 | if (!isdigit(*cpu_list)) |
71 | goto invalid; | 115 | goto out; |
72 | 116 | ||
73 | while (isdigit(*cpu_list)) { | 117 | while (isdigit(*cpu_list)) { |
74 | p = NULL; | 118 | p = NULL; |
@@ -94,21 +138,42 @@ int read_cpu_map(const char *cpu_list) | |||
94 | for (; start_cpu <= end_cpu; start_cpu++) { | 138 | for (; start_cpu <= end_cpu; start_cpu++) { |
95 | /* check for duplicates */ | 139 | /* check for duplicates */ |
96 | for (i = 0; i < nr_cpus; i++) | 140 | for (i = 0; i < nr_cpus; i++) |
97 | if (cpumap[i] == (int)start_cpu) | 141 | if (tmp_cpus[i] == (int)start_cpu) |
98 | goto invalid; | 142 | goto invalid; |
99 | 143 | ||
100 | assert(nr_cpus < MAX_NR_CPUS); | 144 | if (nr_cpus == max_entries) { |
101 | cpumap[nr_cpus++] = (int)start_cpu; | 145 | max_entries += MAX_NR_CPUS; |
146 | tmp = realloc(tmp_cpus, max_entries * sizeof(int)); | ||
147 | if (tmp == NULL) | ||
148 | goto invalid; | ||
149 | tmp_cpus = tmp; | ||
150 | } | ||
151 | tmp_cpus[nr_cpus++] = (int)start_cpu; | ||
102 | } | 152 | } |
103 | if (*p) | 153 | if (*p) |
104 | ++p; | 154 | ++p; |
105 | 155 | ||
106 | cpu_list = p; | 156 | cpu_list = p; |
107 | } | 157 | } |
108 | if (nr_cpus > 0) | ||
109 | return nr_cpus; | ||
110 | 158 | ||
111 | return default_cpu_map(); | 159 | if (nr_cpus > 0) |
160 | cpus = cpu_map__trim_new(nr_cpus, tmp_cpus); | ||
161 | else | ||
162 | cpus = cpu_map__default_new(); | ||
112 | invalid: | 163 | invalid: |
113 | return -1; | 164 | free(tmp_cpus); |
165 | out: | ||
166 | return cpus; | ||
167 | } | ||
168 | |||
169 | struct cpu_map *cpu_map__dummy_new(void) | ||
170 | { | ||
171 | struct cpu_map *cpus = malloc(sizeof(*cpus) + sizeof(int)); | ||
172 | |||
173 | if (cpus != NULL) { | ||
174 | cpus->nr = 1; | ||
175 | cpus->map[0] = -1; | ||
176 | } | ||
177 | |||
178 | return cpus; | ||
114 | } | 179 | } |
diff --git a/tools/perf/util/cpumap.h b/tools/perf/util/cpumap.h index 3e60f56e490..f7a4f42f630 100644 --- a/tools/perf/util/cpumap.h +++ b/tools/perf/util/cpumap.h | |||
@@ -1,7 +1,13 @@ | |||
1 | #ifndef __PERF_CPUMAP_H | 1 | #ifndef __PERF_CPUMAP_H |
2 | #define __PERF_CPUMAP_H | 2 | #define __PERF_CPUMAP_H |
3 | 3 | ||
4 | extern int read_cpu_map(const char *cpu_list); | 4 | struct cpu_map { |
5 | extern int cpumap[]; | 5 | int nr; |
6 | int map[]; | ||
7 | }; | ||
8 | |||
9 | struct cpu_map *cpu_map__new(const char *cpu_list); | ||
10 | struct cpu_map *cpu_map__dummy_new(void); | ||
11 | void *cpu_map__delete(struct cpu_map *map); | ||
6 | 12 | ||
7 | #endif /* __PERF_CPUMAP_H */ | 13 | #endif /* __PERF_CPUMAP_H */ |
diff --git a/tools/perf/util/debug.c b/tools/perf/util/debug.c index c8d81b00089..01bbe8ecec3 100644 --- a/tools/perf/util/debug.c +++ b/tools/perf/util/debug.c | |||
@@ -46,20 +46,16 @@ int dump_printf(const char *fmt, ...) | |||
46 | return ret; | 46 | return ret; |
47 | } | 47 | } |
48 | 48 | ||
49 | static int dump_printf_color(const char *fmt, const char *color, ...) | 49 | #ifdef NO_NEWT_SUPPORT |
50 | void ui__warning(const char *format, ...) | ||
50 | { | 51 | { |
51 | va_list args; | 52 | va_list args; |
52 | int ret = 0; | ||
53 | 53 | ||
54 | if (dump_trace) { | 54 | va_start(args, format); |
55 | va_start(args, color); | 55 | vfprintf(stderr, format, args); |
56 | ret = color_vfprintf(stdout, color, fmt, args); | 56 | va_end(args); |
57 | va_end(args); | ||
58 | } | ||
59 | |||
60 | return ret; | ||
61 | } | 57 | } |
62 | 58 | #endif | |
63 | 59 | ||
64 | void trace_event(event_t *event) | 60 | void trace_event(event_t *event) |
65 | { | 61 | { |
@@ -70,29 +66,29 @@ void trace_event(event_t *event) | |||
70 | if (!dump_trace) | 66 | if (!dump_trace) |
71 | return; | 67 | return; |
72 | 68 | ||
73 | dump_printf("."); | 69 | printf("."); |
74 | dump_printf_color("\n. ... raw event: size %d bytes\n", color, | 70 | color_fprintf(stdout, color, "\n. ... raw event: size %d bytes\n", |
75 | event->header.size); | 71 | event->header.size); |
76 | 72 | ||
77 | for (i = 0; i < event->header.size; i++) { | 73 | for (i = 0; i < event->header.size; i++) { |
78 | if ((i & 15) == 0) { | 74 | if ((i & 15) == 0) { |
79 | dump_printf("."); | 75 | printf("."); |
80 | dump_printf_color(" %04x: ", color, i); | 76 | color_fprintf(stdout, color, " %04x: ", i); |
81 | } | 77 | } |
82 | 78 | ||
83 | dump_printf_color(" %02x", color, raw_event[i]); | 79 | color_fprintf(stdout, color, " %02x", raw_event[i]); |
84 | 80 | ||
85 | if (((i & 15) == 15) || i == event->header.size-1) { | 81 | if (((i & 15) == 15) || i == event->header.size-1) { |
86 | dump_printf_color(" ", color); | 82 | color_fprintf(stdout, color, " "); |
87 | for (j = 0; j < 15-(i & 15); j++) | 83 | for (j = 0; j < 15-(i & 15); j++) |
88 | dump_printf_color(" ", color); | 84 | color_fprintf(stdout, color, " "); |
89 | for (j = i & ~15; j <= i; j++) { | 85 | for (j = i & ~15; j <= i; j++) { |
90 | dump_printf_color("%c", color, | 86 | color_fprintf(stdout, color, "%c", |
91 | isprint(raw_event[j]) ? | 87 | isprint(raw_event[j]) ? |
92 | raw_event[j] : '.'); | 88 | raw_event[j] : '.'); |
93 | } | 89 | } |
94 | dump_printf_color("\n", color); | 90 | color_fprintf(stdout, color, "\n"); |
95 | } | 91 | } |
96 | } | 92 | } |
97 | dump_printf(".\n"); | 93 | printf(".\n"); |
98 | } | 94 | } |
diff --git a/tools/perf/util/debug.h b/tools/perf/util/debug.h index 7b514082bba..ca35fd66b5d 100644 --- a/tools/perf/util/debug.h +++ b/tools/perf/util/debug.h | |||
@@ -35,4 +35,6 @@ int ui_helpline__show_help(const char *format, va_list ap); | |||
35 | #include "ui/progress.h" | 35 | #include "ui/progress.h" |
36 | #endif | 36 | #endif |
37 | 37 | ||
38 | void ui__warning(const char *format, ...) __attribute__((format(printf, 1, 2))); | ||
39 | |||
38 | #endif /* __PERF_DEBUG_H */ | 40 | #endif /* __PERF_DEBUG_H */ |
diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index dab9e754a28..2302ec051bb 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c | |||
@@ -7,7 +7,7 @@ | |||
7 | #include "strlist.h" | 7 | #include "strlist.h" |
8 | #include "thread.h" | 8 | #include "thread.h" |
9 | 9 | ||
10 | const char *event__name[] = { | 10 | static const char *event__name[] = { |
11 | [0] = "TOTAL", | 11 | [0] = "TOTAL", |
12 | [PERF_RECORD_MMAP] = "MMAP", | 12 | [PERF_RECORD_MMAP] = "MMAP", |
13 | [PERF_RECORD_LOST] = "LOST", | 13 | [PERF_RECORD_LOST] = "LOST", |
@@ -22,13 +22,31 @@ const char *event__name[] = { | |||
22 | [PERF_RECORD_HEADER_EVENT_TYPE] = "EVENT_TYPE", | 22 | [PERF_RECORD_HEADER_EVENT_TYPE] = "EVENT_TYPE", |
23 | [PERF_RECORD_HEADER_TRACING_DATA] = "TRACING_DATA", | 23 | [PERF_RECORD_HEADER_TRACING_DATA] = "TRACING_DATA", |
24 | [PERF_RECORD_HEADER_BUILD_ID] = "BUILD_ID", | 24 | [PERF_RECORD_HEADER_BUILD_ID] = "BUILD_ID", |
25 | [PERF_RECORD_FINISHED_ROUND] = "FINISHED_ROUND", | ||
25 | }; | 26 | }; |
26 | 27 | ||
27 | static pid_t event__synthesize_comm(pid_t pid, int full, | 28 | const char *event__get_event_name(unsigned int id) |
29 | { | ||
30 | if (id >= ARRAY_SIZE(event__name)) | ||
31 | return "INVALID"; | ||
32 | if (!event__name[id]) | ||
33 | return "UNKNOWN"; | ||
34 | return event__name[id]; | ||
35 | } | ||
36 | |||
37 | static struct sample_data synth_sample = { | ||
38 | .pid = -1, | ||
39 | .tid = -1, | ||
40 | .time = -1, | ||
41 | .stream_id = -1, | ||
42 | .cpu = -1, | ||
43 | .period = 1, | ||
44 | }; | ||
45 | |||
46 | static pid_t event__synthesize_comm(event_t *event, pid_t pid, int full, | ||
28 | event__handler_t process, | 47 | event__handler_t process, |
29 | struct perf_session *session) | 48 | struct perf_session *session) |
30 | { | 49 | { |
31 | event_t ev; | ||
32 | char filename[PATH_MAX]; | 50 | char filename[PATH_MAX]; |
33 | char bf[BUFSIZ]; | 51 | char bf[BUFSIZ]; |
34 | FILE *fp; | 52 | FILE *fp; |
@@ -49,34 +67,39 @@ out_race: | |||
49 | return 0; | 67 | return 0; |
50 | } | 68 | } |
51 | 69 | ||
52 | memset(&ev.comm, 0, sizeof(ev.comm)); | 70 | memset(&event->comm, 0, sizeof(event->comm)); |
53 | while (!ev.comm.comm[0] || !ev.comm.pid) { | 71 | |
54 | if (fgets(bf, sizeof(bf), fp) == NULL) | 72 | while (!event->comm.comm[0] || !event->comm.pid) { |
55 | goto out_failure; | 73 | if (fgets(bf, sizeof(bf), fp) == NULL) { |
74 | pr_warning("couldn't get COMM and pgid, malformed %s\n", filename); | ||
75 | goto out; | ||
76 | } | ||
56 | 77 | ||
57 | if (memcmp(bf, "Name:", 5) == 0) { | 78 | if (memcmp(bf, "Name:", 5) == 0) { |
58 | char *name = bf + 5; | 79 | char *name = bf + 5; |
59 | while (*name && isspace(*name)) | 80 | while (*name && isspace(*name)) |
60 | ++name; | 81 | ++name; |
61 | size = strlen(name) - 1; | 82 | size = strlen(name) - 1; |
62 | memcpy(ev.comm.comm, name, size++); | 83 | memcpy(event->comm.comm, name, size++); |
63 | } else if (memcmp(bf, "Tgid:", 5) == 0) { | 84 | } else if (memcmp(bf, "Tgid:", 5) == 0) { |
64 | char *tgids = bf + 5; | 85 | char *tgids = bf + 5; |
65 | while (*tgids && isspace(*tgids)) | 86 | while (*tgids && isspace(*tgids)) |
66 | ++tgids; | 87 | ++tgids; |
67 | tgid = ev.comm.pid = atoi(tgids); | 88 | tgid = event->comm.pid = atoi(tgids); |
68 | } | 89 | } |
69 | } | 90 | } |
70 | 91 | ||
71 | ev.comm.header.type = PERF_RECORD_COMM; | 92 | event->comm.header.type = PERF_RECORD_COMM; |
72 | size = ALIGN(size, sizeof(u64)); | 93 | size = ALIGN(size, sizeof(u64)); |
73 | ev.comm.header.size = sizeof(ev.comm) - (sizeof(ev.comm.comm) - size); | 94 | memset(event->comm.comm + size, 0, session->id_hdr_size); |
74 | 95 | event->comm.header.size = (sizeof(event->comm) - | |
96 | (sizeof(event->comm.comm) - size) + | ||
97 | session->id_hdr_size); | ||
75 | if (!full) { | 98 | if (!full) { |
76 | ev.comm.tid = pid; | 99 | event->comm.tid = pid; |
77 | 100 | ||
78 | process(&ev, session); | 101 | process(event, &synth_sample, session); |
79 | goto out_fclose; | 102 | goto out; |
80 | } | 103 | } |
81 | 104 | ||
82 | snprintf(filename, sizeof(filename), "/proc/%d/task", pid); | 105 | snprintf(filename, sizeof(filename), "/proc/%d/task", pid); |
@@ -91,22 +114,19 @@ out_race: | |||
91 | if (*end) | 114 | if (*end) |
92 | continue; | 115 | continue; |
93 | 116 | ||
94 | ev.comm.tid = pid; | 117 | event->comm.tid = pid; |
95 | 118 | ||
96 | process(&ev, session); | 119 | process(event, &synth_sample, session); |
97 | } | 120 | } |
98 | closedir(tasks); | ||
99 | 121 | ||
100 | out_fclose: | 122 | closedir(tasks); |
123 | out: | ||
101 | fclose(fp); | 124 | fclose(fp); |
102 | return tgid; | ||
103 | 125 | ||
104 | out_failure: | 126 | return tgid; |
105 | pr_warning("couldn't get COMM and pgid, malformed %s\n", filename); | ||
106 | return -1; | ||
107 | } | 127 | } |
108 | 128 | ||
109 | static int event__synthesize_mmap_events(pid_t pid, pid_t tgid, | 129 | static int event__synthesize_mmap_events(event_t *event, pid_t pid, pid_t tgid, |
110 | event__handler_t process, | 130 | event__handler_t process, |
111 | struct perf_session *session) | 131 | struct perf_session *session) |
112 | { | 132 | { |
@@ -124,29 +144,25 @@ static int event__synthesize_mmap_events(pid_t pid, pid_t tgid, | |||
124 | return -1; | 144 | return -1; |
125 | } | 145 | } |
126 | 146 | ||
147 | event->header.type = PERF_RECORD_MMAP; | ||
148 | /* | ||
149 | * Just like the kernel, see __perf_event_mmap in kernel/perf_event.c | ||
150 | */ | ||
151 | event->header.misc = PERF_RECORD_MISC_USER; | ||
152 | |||
127 | while (1) { | 153 | while (1) { |
128 | char bf[BUFSIZ], *pbf = bf; | 154 | char bf[BUFSIZ], *pbf = bf; |
129 | event_t ev = { | ||
130 | .header = { | ||
131 | .type = PERF_RECORD_MMAP, | ||
132 | /* | ||
133 | * Just like the kernel, see __perf_event_mmap | ||
134 | * in kernel/perf_event.c | ||
135 | */ | ||
136 | .misc = PERF_RECORD_MISC_USER, | ||
137 | }, | ||
138 | }; | ||
139 | int n; | 155 | int n; |
140 | size_t size; | 156 | size_t size; |
141 | if (fgets(bf, sizeof(bf), fp) == NULL) | 157 | if (fgets(bf, sizeof(bf), fp) == NULL) |
142 | break; | 158 | break; |
143 | 159 | ||
144 | /* 00400000-0040c000 r-xp 00000000 fd:01 41038 /bin/cat */ | 160 | /* 00400000-0040c000 r-xp 00000000 fd:01 41038 /bin/cat */ |
145 | n = hex2u64(pbf, &ev.mmap.start); | 161 | n = hex2u64(pbf, &event->mmap.start); |
146 | if (n < 0) | 162 | if (n < 0) |
147 | continue; | 163 | continue; |
148 | pbf += n + 1; | 164 | pbf += n + 1; |
149 | n = hex2u64(pbf, &ev.mmap.len); | 165 | n = hex2u64(pbf, &event->mmap.len); |
150 | if (n < 0) | 166 | if (n < 0) |
151 | continue; | 167 | continue; |
152 | pbf += n + 3; | 168 | pbf += n + 3; |
@@ -161,19 +177,21 @@ static int event__synthesize_mmap_events(pid_t pid, pid_t tgid, | |||
161 | continue; | 177 | continue; |
162 | 178 | ||
163 | pbf += 3; | 179 | pbf += 3; |
164 | n = hex2u64(pbf, &ev.mmap.pgoff); | 180 | n = hex2u64(pbf, &event->mmap.pgoff); |
165 | 181 | ||
166 | size = strlen(execname); | 182 | size = strlen(execname); |
167 | execname[size - 1] = '\0'; /* Remove \n */ | 183 | execname[size - 1] = '\0'; /* Remove \n */ |
168 | memcpy(ev.mmap.filename, execname, size); | 184 | memcpy(event->mmap.filename, execname, size); |
169 | size = ALIGN(size, sizeof(u64)); | 185 | size = ALIGN(size, sizeof(u64)); |
170 | ev.mmap.len -= ev.mmap.start; | 186 | event->mmap.len -= event->mmap.start; |
171 | ev.mmap.header.size = (sizeof(ev.mmap) - | 187 | event->mmap.header.size = (sizeof(event->mmap) - |
172 | (sizeof(ev.mmap.filename) - size)); | 188 | (sizeof(event->mmap.filename) - size)); |
173 | ev.mmap.pid = tgid; | 189 | memset(event->mmap.filename + size, 0, session->id_hdr_size); |
174 | ev.mmap.tid = pid; | 190 | event->mmap.header.size += session->id_hdr_size; |
175 | 191 | event->mmap.pid = tgid; | |
176 | process(&ev, session); | 192 | event->mmap.tid = pid; |
193 | |||
194 | process(event, &synth_sample, session); | ||
177 | } | 195 | } |
178 | } | 196 | } |
179 | 197 | ||
@@ -187,20 +205,27 @@ int event__synthesize_modules(event__handler_t process, | |||
187 | { | 205 | { |
188 | struct rb_node *nd; | 206 | struct rb_node *nd; |
189 | struct map_groups *kmaps = &machine->kmaps; | 207 | struct map_groups *kmaps = &machine->kmaps; |
190 | u16 misc; | 208 | event_t *event = zalloc(sizeof(event->mmap) + session->id_hdr_size); |
209 | |||
210 | if (event == NULL) { | ||
211 | pr_debug("Not enough memory synthesizing mmap event " | ||
212 | "for kernel modules\n"); | ||
213 | return -1; | ||
214 | } | ||
215 | |||
216 | event->header.type = PERF_RECORD_MMAP; | ||
191 | 217 | ||
192 | /* | 218 | /* |
193 | * kernel uses 0 for user space maps, see kernel/perf_event.c | 219 | * kernel uses 0 for user space maps, see kernel/perf_event.c |
194 | * __perf_event_mmap | 220 | * __perf_event_mmap |
195 | */ | 221 | */ |
196 | if (machine__is_host(machine)) | 222 | if (machine__is_host(machine)) |
197 | misc = PERF_RECORD_MISC_KERNEL; | 223 | event->header.misc = PERF_RECORD_MISC_KERNEL; |
198 | else | 224 | else |
199 | misc = PERF_RECORD_MISC_GUEST_KERNEL; | 225 | event->header.misc = PERF_RECORD_MISC_GUEST_KERNEL; |
200 | 226 | ||
201 | for (nd = rb_first(&kmaps->maps[MAP__FUNCTION]); | 227 | for (nd = rb_first(&kmaps->maps[MAP__FUNCTION]); |
202 | nd; nd = rb_next(nd)) { | 228 | nd; nd = rb_next(nd)) { |
203 | event_t ev; | ||
204 | size_t size; | 229 | size_t size; |
205 | struct map *pos = rb_entry(nd, struct map, rb_node); | 230 | struct map *pos = rb_entry(nd, struct map, rb_node); |
206 | 231 | ||
@@ -208,39 +233,78 @@ int event__synthesize_modules(event__handler_t process, | |||
208 | continue; | 233 | continue; |
209 | 234 | ||
210 | size = ALIGN(pos->dso->long_name_len + 1, sizeof(u64)); | 235 | size = ALIGN(pos->dso->long_name_len + 1, sizeof(u64)); |
211 | memset(&ev, 0, sizeof(ev)); | 236 | event->mmap.header.type = PERF_RECORD_MMAP; |
212 | ev.mmap.header.misc = misc; | 237 | event->mmap.header.size = (sizeof(event->mmap) - |
213 | ev.mmap.header.type = PERF_RECORD_MMAP; | 238 | (sizeof(event->mmap.filename) - size)); |
214 | ev.mmap.header.size = (sizeof(ev.mmap) - | 239 | memset(event->mmap.filename + size, 0, session->id_hdr_size); |
215 | (sizeof(ev.mmap.filename) - size)); | 240 | event->mmap.header.size += session->id_hdr_size; |
216 | ev.mmap.start = pos->start; | 241 | event->mmap.start = pos->start; |
217 | ev.mmap.len = pos->end - pos->start; | 242 | event->mmap.len = pos->end - pos->start; |
218 | ev.mmap.pid = machine->pid; | 243 | event->mmap.pid = machine->pid; |
219 | 244 | ||
220 | memcpy(ev.mmap.filename, pos->dso->long_name, | 245 | memcpy(event->mmap.filename, pos->dso->long_name, |
221 | pos->dso->long_name_len + 1); | 246 | pos->dso->long_name_len + 1); |
222 | process(&ev, session); | 247 | process(event, &synth_sample, session); |
223 | } | 248 | } |
224 | 249 | ||
250 | free(event); | ||
225 | return 0; | 251 | return 0; |
226 | } | 252 | } |
227 | 253 | ||
228 | int event__synthesize_thread(pid_t pid, event__handler_t process, | 254 | static int __event__synthesize_thread(event_t *comm_event, event_t *mmap_event, |
229 | struct perf_session *session) | 255 | pid_t pid, event__handler_t process, |
256 | struct perf_session *session) | ||
230 | { | 257 | { |
231 | pid_t tgid = event__synthesize_comm(pid, 1, process, session); | 258 | pid_t tgid = event__synthesize_comm(comm_event, pid, 1, process, |
259 | session); | ||
232 | if (tgid == -1) | 260 | if (tgid == -1) |
233 | return -1; | 261 | return -1; |
234 | return event__synthesize_mmap_events(pid, tgid, process, session); | 262 | return event__synthesize_mmap_events(mmap_event, pid, tgid, |
263 | process, session); | ||
264 | } | ||
265 | |||
266 | int event__synthesize_thread(pid_t pid, event__handler_t process, | ||
267 | struct perf_session *session) | ||
268 | { | ||
269 | event_t *comm_event, *mmap_event; | ||
270 | int err = -1; | ||
271 | |||
272 | comm_event = malloc(sizeof(comm_event->comm) + session->id_hdr_size); | ||
273 | if (comm_event == NULL) | ||
274 | goto out; | ||
275 | |||
276 | mmap_event = malloc(sizeof(mmap_event->mmap) + session->id_hdr_size); | ||
277 | if (mmap_event == NULL) | ||
278 | goto out_free_comm; | ||
279 | |||
280 | err = __event__synthesize_thread(comm_event, mmap_event, pid, | ||
281 | process, session); | ||
282 | free(mmap_event); | ||
283 | out_free_comm: | ||
284 | free(comm_event); | ||
285 | out: | ||
286 | return err; | ||
235 | } | 287 | } |
236 | 288 | ||
237 | void event__synthesize_threads(event__handler_t process, | 289 | int event__synthesize_threads(event__handler_t process, |
238 | struct perf_session *session) | 290 | struct perf_session *session) |
239 | { | 291 | { |
240 | DIR *proc; | 292 | DIR *proc; |
241 | struct dirent dirent, *next; | 293 | struct dirent dirent, *next; |
294 | event_t *comm_event, *mmap_event; | ||
295 | int err = -1; | ||
296 | |||
297 | comm_event = malloc(sizeof(comm_event->comm) + session->id_hdr_size); | ||
298 | if (comm_event == NULL) | ||
299 | goto out; | ||
300 | |||
301 | mmap_event = malloc(sizeof(mmap_event->mmap) + session->id_hdr_size); | ||
302 | if (mmap_event == NULL) | ||
303 | goto out_free_comm; | ||
242 | 304 | ||
243 | proc = opendir("/proc"); | 305 | proc = opendir("/proc"); |
306 | if (proc == NULL) | ||
307 | goto out_free_mmap; | ||
244 | 308 | ||
245 | while (!readdir_r(proc, &dirent, &next) && next) { | 309 | while (!readdir_r(proc, &dirent, &next) && next) { |
246 | char *end; | 310 | char *end; |
@@ -249,10 +313,18 @@ void event__synthesize_threads(event__handler_t process, | |||
249 | if (*end) /* only interested in proper numerical dirents */ | 313 | if (*end) /* only interested in proper numerical dirents */ |
250 | continue; | 314 | continue; |
251 | 315 | ||
252 | event__synthesize_thread(pid, process, session); | 316 | __event__synthesize_thread(comm_event, mmap_event, pid, |
317 | process, session); | ||
253 | } | 318 | } |
254 | 319 | ||
255 | closedir(proc); | 320 | closedir(proc); |
321 | err = 0; | ||
322 | out_free_mmap: | ||
323 | free(mmap_event); | ||
324 | out_free_comm: | ||
325 | free(comm_event); | ||
326 | out: | ||
327 | return err; | ||
256 | } | 328 | } |
257 | 329 | ||
258 | struct process_symbol_args { | 330 | struct process_symbol_args { |
@@ -260,7 +332,8 @@ struct process_symbol_args { | |||
260 | u64 start; | 332 | u64 start; |
261 | }; | 333 | }; |
262 | 334 | ||
263 | static int find_symbol_cb(void *arg, const char *name, char type, u64 start) | 335 | static int find_symbol_cb(void *arg, const char *name, char type, |
336 | u64 start, u64 end __used) | ||
264 | { | 337 | { |
265 | struct process_symbol_args *args = arg; | 338 | struct process_symbol_args *args = arg; |
266 | 339 | ||
@@ -286,18 +359,20 @@ int event__synthesize_kernel_mmap(event__handler_t process, | |||
286 | char path[PATH_MAX]; | 359 | char path[PATH_MAX]; |
287 | char name_buff[PATH_MAX]; | 360 | char name_buff[PATH_MAX]; |
288 | struct map *map; | 361 | struct map *map; |
289 | 362 | int err; | |
290 | event_t ev = { | ||
291 | .header = { | ||
292 | .type = PERF_RECORD_MMAP, | ||
293 | }, | ||
294 | }; | ||
295 | /* | 363 | /* |
296 | * We should get this from /sys/kernel/sections/.text, but till that is | 364 | * We should get this from /sys/kernel/sections/.text, but till that is |
297 | * available use this, and after it is use this as a fallback for older | 365 | * available use this, and after it is use this as a fallback for older |
298 | * kernels. | 366 | * kernels. |
299 | */ | 367 | */ |
300 | struct process_symbol_args args = { .name = symbol_name, }; | 368 | struct process_symbol_args args = { .name = symbol_name, }; |
369 | event_t *event = zalloc(sizeof(event->mmap) + session->id_hdr_size); | ||
370 | |||
371 | if (event == NULL) { | ||
372 | pr_debug("Not enough memory synthesizing mmap event " | ||
373 | "for kernel modules\n"); | ||
374 | return -1; | ||
375 | } | ||
301 | 376 | ||
302 | mmap_name = machine__mmap_name(machine, name_buff, sizeof(name_buff)); | 377 | mmap_name = machine__mmap_name(machine, name_buff, sizeof(name_buff)); |
303 | if (machine__is_host(machine)) { | 378 | if (machine__is_host(machine)) { |
@@ -305,10 +380,10 @@ int event__synthesize_kernel_mmap(event__handler_t process, | |||
305 | * kernel uses PERF_RECORD_MISC_USER for user space maps, | 380 | * kernel uses PERF_RECORD_MISC_USER for user space maps, |
306 | * see kernel/perf_event.c __perf_event_mmap | 381 | * see kernel/perf_event.c __perf_event_mmap |
307 | */ | 382 | */ |
308 | ev.header.misc = PERF_RECORD_MISC_KERNEL; | 383 | event->header.misc = PERF_RECORD_MISC_KERNEL; |
309 | filename = "/proc/kallsyms"; | 384 | filename = "/proc/kallsyms"; |
310 | } else { | 385 | } else { |
311 | ev.header.misc = PERF_RECORD_MISC_GUEST_KERNEL; | 386 | event->header.misc = PERF_RECORD_MISC_GUEST_KERNEL; |
312 | if (machine__is_default_guest(machine)) | 387 | if (machine__is_default_guest(machine)) |
313 | filename = (char *) symbol_conf.default_guest_kallsyms; | 388 | filename = (char *) symbol_conf.default_guest_kallsyms; |
314 | else { | 389 | else { |
@@ -321,17 +396,21 @@ int event__synthesize_kernel_mmap(event__handler_t process, | |||
321 | return -ENOENT; | 396 | return -ENOENT; |
322 | 397 | ||
323 | map = machine->vmlinux_maps[MAP__FUNCTION]; | 398 | map = machine->vmlinux_maps[MAP__FUNCTION]; |
324 | size = snprintf(ev.mmap.filename, sizeof(ev.mmap.filename), | 399 | size = snprintf(event->mmap.filename, sizeof(event->mmap.filename), |
325 | "%s%s", mmap_name, symbol_name) + 1; | 400 | "%s%s", mmap_name, symbol_name) + 1; |
326 | size = ALIGN(size, sizeof(u64)); | 401 | size = ALIGN(size, sizeof(u64)); |
327 | ev.mmap.header.size = (sizeof(ev.mmap) - | 402 | event->mmap.header.type = PERF_RECORD_MMAP; |
328 | (sizeof(ev.mmap.filename) - size)); | 403 | event->mmap.header.size = (sizeof(event->mmap) - |
329 | ev.mmap.pgoff = args.start; | 404 | (sizeof(event->mmap.filename) - size) + session->id_hdr_size); |
330 | ev.mmap.start = map->start; | 405 | event->mmap.pgoff = args.start; |
331 | ev.mmap.len = map->end - ev.mmap.start; | 406 | event->mmap.start = map->start; |
332 | ev.mmap.pid = machine->pid; | 407 | event->mmap.len = map->end - event->mmap.start; |
333 | 408 | event->mmap.pid = machine->pid; | |
334 | return process(&ev, session); | 409 | |
410 | err = process(event, &synth_sample, session); | ||
411 | free(event); | ||
412 | |||
413 | return err; | ||
335 | } | 414 | } |
336 | 415 | ||
337 | static void thread__comm_adjust(struct thread *self, struct hists *hists) | 416 | static void thread__comm_adjust(struct thread *self, struct hists *hists) |
@@ -361,7 +440,8 @@ static int thread__set_comm_adjust(struct thread *self, const char *comm, | |||
361 | return 0; | 440 | return 0; |
362 | } | 441 | } |
363 | 442 | ||
364 | int event__process_comm(event_t *self, struct perf_session *session) | 443 | int event__process_comm(event_t *self, struct sample_data *sample __used, |
444 | struct perf_session *session) | ||
365 | { | 445 | { |
366 | struct thread *thread = perf_session__findnew(session, self->comm.tid); | 446 | struct thread *thread = perf_session__findnew(session, self->comm.tid); |
367 | 447 | ||
@@ -376,7 +456,8 @@ int event__process_comm(event_t *self, struct perf_session *session) | |||
376 | return 0; | 456 | return 0; |
377 | } | 457 | } |
378 | 458 | ||
379 | int event__process_lost(event_t *self, struct perf_session *session) | 459 | int event__process_lost(event_t *self, struct sample_data *sample __used, |
460 | struct perf_session *session) | ||
380 | { | 461 | { |
381 | dump_printf(": id:%Ld: lost:%Ld\n", self->lost.id, self->lost.lost); | 462 | dump_printf(": id:%Ld: lost:%Ld\n", self->lost.id, self->lost.lost); |
382 | session->hists.stats.total_lost += self->lost.lost; | 463 | session->hists.stats.total_lost += self->lost.lost; |
@@ -392,7 +473,7 @@ static void event_set_kernel_mmap_len(struct map **maps, event_t *self) | |||
392 | * a zero sized synthesized MMAP event for the kernel. | 473 | * a zero sized synthesized MMAP event for the kernel. |
393 | */ | 474 | */ |
394 | if (maps[MAP__FUNCTION]->end == 0) | 475 | if (maps[MAP__FUNCTION]->end == 0) |
395 | maps[MAP__FUNCTION]->end = ~0UL; | 476 | maps[MAP__FUNCTION]->end = ~0ULL; |
396 | } | 477 | } |
397 | 478 | ||
398 | static int event__process_kernel_mmap(event_t *self, | 479 | static int event__process_kernel_mmap(event_t *self, |
@@ -485,7 +566,8 @@ out_problem: | |||
485 | return -1; | 566 | return -1; |
486 | } | 567 | } |
487 | 568 | ||
488 | int event__process_mmap(event_t *self, struct perf_session *session) | 569 | int event__process_mmap(event_t *self, struct sample_data *sample __used, |
570 | struct perf_session *session) | ||
489 | { | 571 | { |
490 | struct machine *machine; | 572 | struct machine *machine; |
491 | struct thread *thread; | 573 | struct thread *thread; |
@@ -526,7 +608,8 @@ out_problem: | |||
526 | return 0; | 608 | return 0; |
527 | } | 609 | } |
528 | 610 | ||
529 | int event__process_task(event_t *self, struct perf_session *session) | 611 | int event__process_task(event_t *self, struct sample_data *sample __used, |
612 | struct perf_session *session) | ||
530 | { | 613 | { |
531 | struct thread *thread = perf_session__findnew(session, self->fork.tid); | 614 | struct thread *thread = perf_session__findnew(session, self->fork.tid); |
532 | struct thread *parent = perf_session__findnew(session, self->fork.ptid); | 615 | struct thread *parent = perf_session__findnew(session, self->fork.ptid); |
@@ -548,18 +631,19 @@ int event__process_task(event_t *self, struct perf_session *session) | |||
548 | return 0; | 631 | return 0; |
549 | } | 632 | } |
550 | 633 | ||
551 | int event__process(event_t *event, struct perf_session *session) | 634 | int event__process(event_t *event, struct sample_data *sample, |
635 | struct perf_session *session) | ||
552 | { | 636 | { |
553 | switch (event->header.type) { | 637 | switch (event->header.type) { |
554 | case PERF_RECORD_COMM: | 638 | case PERF_RECORD_COMM: |
555 | event__process_comm(event, session); | 639 | event__process_comm(event, sample, session); |
556 | break; | 640 | break; |
557 | case PERF_RECORD_MMAP: | 641 | case PERF_RECORD_MMAP: |
558 | event__process_mmap(event, session); | 642 | event__process_mmap(event, sample, session); |
559 | break; | 643 | break; |
560 | case PERF_RECORD_FORK: | 644 | case PERF_RECORD_FORK: |
561 | case PERF_RECORD_EXIT: | 645 | case PERF_RECORD_EXIT: |
562 | event__process_task(event, session); | 646 | event__process_task(event, sample, session); |
563 | break; | 647 | break; |
564 | default: | 648 | default: |
565 | break; | 649 | break; |
@@ -674,32 +758,8 @@ int event__preprocess_sample(const event_t *self, struct perf_session *session, | |||
674 | symbol_filter_t filter) | 758 | symbol_filter_t filter) |
675 | { | 759 | { |
676 | u8 cpumode = self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; | 760 | u8 cpumode = self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; |
677 | struct thread *thread; | 761 | struct thread *thread = perf_session__findnew(session, self->ip.pid); |
678 | |||
679 | event__parse_sample(self, session->sample_type, data); | ||
680 | |||
681 | dump_printf("(IP, %d): %d/%d: %#Lx period: %Ld cpu:%d\n", | ||
682 | self->header.misc, data->pid, data->tid, data->ip, | ||
683 | data->period, data->cpu); | ||
684 | |||
685 | if (session->sample_type & PERF_SAMPLE_CALLCHAIN) { | ||
686 | unsigned int i; | ||
687 | |||
688 | dump_printf("... chain: nr:%Lu\n", data->callchain->nr); | ||
689 | 762 | ||
690 | if (!ip_callchain__valid(data->callchain, self)) { | ||
691 | pr_debug("call-chain problem with event, " | ||
692 | "skipping it.\n"); | ||
693 | goto out_filtered; | ||
694 | } | ||
695 | |||
696 | if (dump_trace) { | ||
697 | for (i = 0; i < data->callchain->nr; i++) | ||
698 | dump_printf("..... %2d: %016Lx\n", | ||
699 | i, data->callchain->ips[i]); | ||
700 | } | ||
701 | } | ||
702 | thread = perf_session__findnew(session, self->ip.pid); | ||
703 | if (thread == NULL) | 763 | if (thread == NULL) |
704 | return -1; | 764 | return -1; |
705 | 765 | ||
@@ -766,9 +826,65 @@ out_filtered: | |||
766 | return 0; | 826 | return 0; |
767 | } | 827 | } |
768 | 828 | ||
769 | int event__parse_sample(const event_t *event, u64 type, struct sample_data *data) | 829 | static int event__parse_id_sample(const event_t *event, |
830 | struct perf_session *session, | ||
831 | struct sample_data *sample) | ||
770 | { | 832 | { |
771 | const u64 *array = event->sample.array; | 833 | const u64 *array; |
834 | u64 type; | ||
835 | |||
836 | sample->cpu = sample->pid = sample->tid = -1; | ||
837 | sample->stream_id = sample->id = sample->time = -1ULL; | ||
838 | |||
839 | if (!session->sample_id_all) | ||
840 | return 0; | ||
841 | |||
842 | array = event->sample.array; | ||
843 | array += ((event->header.size - | ||
844 | sizeof(event->header)) / sizeof(u64)) - 1; | ||
845 | type = session->sample_type; | ||
846 | |||
847 | if (type & PERF_SAMPLE_CPU) { | ||
848 | u32 *p = (u32 *)array; | ||
849 | sample->cpu = *p; | ||
850 | array--; | ||
851 | } | ||
852 | |||
853 | if (type & PERF_SAMPLE_STREAM_ID) { | ||
854 | sample->stream_id = *array; | ||
855 | array--; | ||
856 | } | ||
857 | |||
858 | if (type & PERF_SAMPLE_ID) { | ||
859 | sample->id = *array; | ||
860 | array--; | ||
861 | } | ||
862 | |||
863 | if (type & PERF_SAMPLE_TIME) { | ||
864 | sample->time = *array; | ||
865 | array--; | ||
866 | } | ||
867 | |||
868 | if (type & PERF_SAMPLE_TID) { | ||
869 | u32 *p = (u32 *)array; | ||
870 | sample->pid = p[0]; | ||
871 | sample->tid = p[1]; | ||
872 | } | ||
873 | |||
874 | return 0; | ||
875 | } | ||
876 | |||
877 | int event__parse_sample(const event_t *event, struct perf_session *session, | ||
878 | struct sample_data *data) | ||
879 | { | ||
880 | const u64 *array; | ||
881 | u64 type; | ||
882 | |||
883 | if (event->header.type != PERF_RECORD_SAMPLE) | ||
884 | return event__parse_id_sample(event, session, data); | ||
885 | |||
886 | array = event->sample.array; | ||
887 | type = session->sample_type; | ||
772 | 888 | ||
773 | if (type & PERF_SAMPLE_IP) { | 889 | if (type & PERF_SAMPLE_IP) { |
774 | data->ip = event->ip.ip; | 890 | data->ip = event->ip.ip; |
diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index 8e790dae702..2b7e91902f1 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h | |||
@@ -85,6 +85,7 @@ struct build_id_event { | |||
85 | }; | 85 | }; |
86 | 86 | ||
87 | enum perf_user_event_type { /* above any possible kernel type */ | 87 | enum perf_user_event_type { /* above any possible kernel type */ |
88 | PERF_RECORD_USER_TYPE_START = 64, | ||
88 | PERF_RECORD_HEADER_ATTR = 64, | 89 | PERF_RECORD_HEADER_ATTR = 64, |
89 | PERF_RECORD_HEADER_EVENT_TYPE = 65, | 90 | PERF_RECORD_HEADER_EVENT_TYPE = 65, |
90 | PERF_RECORD_HEADER_TRACING_DATA = 66, | 91 | PERF_RECORD_HEADER_TRACING_DATA = 66, |
@@ -135,12 +136,15 @@ void event__print_totals(void); | |||
135 | 136 | ||
136 | struct perf_session; | 137 | struct perf_session; |
137 | 138 | ||
138 | typedef int (*event__handler_t)(event_t *event, struct perf_session *session); | 139 | typedef int (*event__handler_synth_t)(event_t *event, |
140 | struct perf_session *session); | ||
141 | typedef int (*event__handler_t)(event_t *event, struct sample_data *sample, | ||
142 | struct perf_session *session); | ||
139 | 143 | ||
140 | int event__synthesize_thread(pid_t pid, event__handler_t process, | 144 | int event__synthesize_thread(pid_t pid, event__handler_t process, |
141 | struct perf_session *session); | 145 | struct perf_session *session); |
142 | void event__synthesize_threads(event__handler_t process, | 146 | int event__synthesize_threads(event__handler_t process, |
143 | struct perf_session *session); | 147 | struct perf_session *session); |
144 | int event__synthesize_kernel_mmap(event__handler_t process, | 148 | int event__synthesize_kernel_mmap(event__handler_t process, |
145 | struct perf_session *session, | 149 | struct perf_session *session, |
146 | struct machine *machine, | 150 | struct machine *machine, |
@@ -150,18 +154,24 @@ int event__synthesize_modules(event__handler_t process, | |||
150 | struct perf_session *session, | 154 | struct perf_session *session, |
151 | struct machine *machine); | 155 | struct machine *machine); |
152 | 156 | ||
153 | int event__process_comm(event_t *self, struct perf_session *session); | 157 | int event__process_comm(event_t *self, struct sample_data *sample, |
154 | int event__process_lost(event_t *self, struct perf_session *session); | 158 | struct perf_session *session); |
155 | int event__process_mmap(event_t *self, struct perf_session *session); | 159 | int event__process_lost(event_t *self, struct sample_data *sample, |
156 | int event__process_task(event_t *self, struct perf_session *session); | 160 | struct perf_session *session); |
157 | int event__process(event_t *event, struct perf_session *session); | 161 | int event__process_mmap(event_t *self, struct sample_data *sample, |
162 | struct perf_session *session); | ||
163 | int event__process_task(event_t *self, struct sample_data *sample, | ||
164 | struct perf_session *session); | ||
165 | int event__process(event_t *event, struct sample_data *sample, | ||
166 | struct perf_session *session); | ||
158 | 167 | ||
159 | struct addr_location; | 168 | struct addr_location; |
160 | int event__preprocess_sample(const event_t *self, struct perf_session *session, | 169 | int event__preprocess_sample(const event_t *self, struct perf_session *session, |
161 | struct addr_location *al, struct sample_data *data, | 170 | struct addr_location *al, struct sample_data *data, |
162 | symbol_filter_t filter); | 171 | symbol_filter_t filter); |
163 | int event__parse_sample(const event_t *event, u64 type, struct sample_data *data); | 172 | int event__parse_sample(const event_t *event, struct perf_session *session, |
173 | struct sample_data *sample); | ||
164 | 174 | ||
165 | extern const char *event__name[]; | 175 | const char *event__get_event_name(unsigned int id); |
166 | 176 | ||
167 | #endif /* __PERF_RECORD_H */ | 177 | #endif /* __PERF_RECORD_H */ |
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c new file mode 100644 index 00000000000..c95267e63c5 --- /dev/null +++ b/tools/perf/util/evsel.c | |||
@@ -0,0 +1,186 @@ | |||
1 | #include "evsel.h" | ||
2 | #include "../perf.h" | ||
3 | #include "util.h" | ||
4 | #include "cpumap.h" | ||
5 | #include "thread.h" | ||
6 | |||
7 | #define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y)) | ||
8 | |||
9 | struct perf_evsel *perf_evsel__new(u32 type, u64 config, int idx) | ||
10 | { | ||
11 | struct perf_evsel *evsel = zalloc(sizeof(*evsel)); | ||
12 | |||
13 | if (evsel != NULL) { | ||
14 | evsel->idx = idx; | ||
15 | evsel->attr.type = type; | ||
16 | evsel->attr.config = config; | ||
17 | INIT_LIST_HEAD(&evsel->node); | ||
18 | } | ||
19 | |||
20 | return evsel; | ||
21 | } | ||
22 | |||
23 | int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads) | ||
24 | { | ||
25 | evsel->fd = xyarray__new(ncpus, nthreads, sizeof(int)); | ||
26 | return evsel->fd != NULL ? 0 : -ENOMEM; | ||
27 | } | ||
28 | |||
29 | int perf_evsel__alloc_counts(struct perf_evsel *evsel, int ncpus) | ||
30 | { | ||
31 | evsel->counts = zalloc((sizeof(*evsel->counts) + | ||
32 | (ncpus * sizeof(struct perf_counts_values)))); | ||
33 | return evsel->counts != NULL ? 0 : -ENOMEM; | ||
34 | } | ||
35 | |||
36 | void perf_evsel__free_fd(struct perf_evsel *evsel) | ||
37 | { | ||
38 | xyarray__delete(evsel->fd); | ||
39 | evsel->fd = NULL; | ||
40 | } | ||
41 | |||
42 | void perf_evsel__close_fd(struct perf_evsel *evsel, int ncpus, int nthreads) | ||
43 | { | ||
44 | int cpu, thread; | ||
45 | |||
46 | for (cpu = 0; cpu < ncpus; cpu++) | ||
47 | for (thread = 0; thread < nthreads; ++thread) { | ||
48 | close(FD(evsel, cpu, thread)); | ||
49 | FD(evsel, cpu, thread) = -1; | ||
50 | } | ||
51 | } | ||
52 | |||
53 | void perf_evsel__delete(struct perf_evsel *evsel) | ||
54 | { | ||
55 | assert(list_empty(&evsel->node)); | ||
56 | xyarray__delete(evsel->fd); | ||
57 | free(evsel); | ||
58 | } | ||
59 | |||
60 | int __perf_evsel__read_on_cpu(struct perf_evsel *evsel, | ||
61 | int cpu, int thread, bool scale) | ||
62 | { | ||
63 | struct perf_counts_values count; | ||
64 | size_t nv = scale ? 3 : 1; | ||
65 | |||
66 | if (FD(evsel, cpu, thread) < 0) | ||
67 | return -EINVAL; | ||
68 | |||
69 | if (evsel->counts == NULL && perf_evsel__alloc_counts(evsel, cpu + 1) < 0) | ||
70 | return -ENOMEM; | ||
71 | |||
72 | if (readn(FD(evsel, cpu, thread), &count, nv * sizeof(u64)) < 0) | ||
73 | return -errno; | ||
74 | |||
75 | if (scale) { | ||
76 | if (count.run == 0) | ||
77 | count.val = 0; | ||
78 | else if (count.run < count.ena) | ||
79 | count.val = (u64)((double)count.val * count.ena / count.run + 0.5); | ||
80 | } else | ||
81 | count.ena = count.run = 0; | ||
82 | |||
83 | evsel->counts->cpu[cpu] = count; | ||
84 | return 0; | ||
85 | } | ||
86 | |||
87 | int __perf_evsel__read(struct perf_evsel *evsel, | ||
88 | int ncpus, int nthreads, bool scale) | ||
89 | { | ||
90 | size_t nv = scale ? 3 : 1; | ||
91 | int cpu, thread; | ||
92 | struct perf_counts_values *aggr = &evsel->counts->aggr, count; | ||
93 | |||
94 | aggr->val = 0; | ||
95 | |||
96 | for (cpu = 0; cpu < ncpus; cpu++) { | ||
97 | for (thread = 0; thread < nthreads; thread++) { | ||
98 | if (FD(evsel, cpu, thread) < 0) | ||
99 | continue; | ||
100 | |||
101 | if (readn(FD(evsel, cpu, thread), | ||
102 | &count, nv * sizeof(u64)) < 0) | ||
103 | return -errno; | ||
104 | |||
105 | aggr->val += count.val; | ||
106 | if (scale) { | ||
107 | aggr->ena += count.ena; | ||
108 | aggr->run += count.run; | ||
109 | } | ||
110 | } | ||
111 | } | ||
112 | |||
113 | evsel->counts->scaled = 0; | ||
114 | if (scale) { | ||
115 | if (aggr->run == 0) { | ||
116 | evsel->counts->scaled = -1; | ||
117 | aggr->val = 0; | ||
118 | return 0; | ||
119 | } | ||
120 | |||
121 | if (aggr->run < aggr->ena) { | ||
122 | evsel->counts->scaled = 1; | ||
123 | aggr->val = (u64)((double)aggr->val * aggr->ena / aggr->run + 0.5); | ||
124 | } | ||
125 | } else | ||
126 | aggr->ena = aggr->run = 0; | ||
127 | |||
128 | return 0; | ||
129 | } | ||
130 | |||
131 | int perf_evsel__open_per_cpu(struct perf_evsel *evsel, struct cpu_map *cpus) | ||
132 | { | ||
133 | int cpu; | ||
134 | |||
135 | if (evsel->fd == NULL && perf_evsel__alloc_fd(evsel, cpus->nr, 1) < 0) | ||
136 | return -1; | ||
137 | |||
138 | for (cpu = 0; cpu < cpus->nr; cpu++) { | ||
139 | FD(evsel, cpu, 0) = sys_perf_event_open(&evsel->attr, -1, | ||
140 | cpus->map[cpu], -1, 0); | ||
141 | if (FD(evsel, cpu, 0) < 0) | ||
142 | goto out_close; | ||
143 | } | ||
144 | |||
145 | return 0; | ||
146 | |||
147 | out_close: | ||
148 | while (--cpu >= 0) { | ||
149 | close(FD(evsel, cpu, 0)); | ||
150 | FD(evsel, cpu, 0) = -1; | ||
151 | } | ||
152 | return -1; | ||
153 | } | ||
154 | |||
155 | int perf_evsel__open_per_thread(struct perf_evsel *evsel, struct thread_map *threads) | ||
156 | { | ||
157 | int thread; | ||
158 | |||
159 | if (evsel->fd == NULL && perf_evsel__alloc_fd(evsel, 1, threads->nr)) | ||
160 | return -1; | ||
161 | |||
162 | for (thread = 0; thread < threads->nr; thread++) { | ||
163 | FD(evsel, 0, thread) = sys_perf_event_open(&evsel->attr, | ||
164 | threads->map[thread], -1, -1, 0); | ||
165 | if (FD(evsel, 0, thread) < 0) | ||
166 | goto out_close; | ||
167 | } | ||
168 | |||
169 | return 0; | ||
170 | |||
171 | out_close: | ||
172 | while (--thread >= 0) { | ||
173 | close(FD(evsel, 0, thread)); | ||
174 | FD(evsel, 0, thread) = -1; | ||
175 | } | ||
176 | return -1; | ||
177 | } | ||
178 | |||
179 | int perf_evsel__open(struct perf_evsel *evsel, | ||
180 | struct cpu_map *cpus, struct thread_map *threads) | ||
181 | { | ||
182 | if (threads == NULL) | ||
183 | return perf_evsel__open_per_cpu(evsel, cpus); | ||
184 | |||
185 | return perf_evsel__open_per_thread(evsel, threads); | ||
186 | } | ||
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h new file mode 100644 index 00000000000..a0ccd69c3fc --- /dev/null +++ b/tools/perf/util/evsel.h | |||
@@ -0,0 +1,115 @@ | |||
1 | #ifndef __PERF_EVSEL_H | ||
2 | #define __PERF_EVSEL_H 1 | ||
3 | |||
4 | #include <linux/list.h> | ||
5 | #include <stdbool.h> | ||
6 | #include "../../../include/linux/perf_event.h" | ||
7 | #include "types.h" | ||
8 | #include "xyarray.h" | ||
9 | |||
10 | struct perf_counts_values { | ||
11 | union { | ||
12 | struct { | ||
13 | u64 val; | ||
14 | u64 ena; | ||
15 | u64 run; | ||
16 | }; | ||
17 | u64 values[3]; | ||
18 | }; | ||
19 | }; | ||
20 | |||
21 | struct perf_counts { | ||
22 | s8 scaled; | ||
23 | struct perf_counts_values aggr; | ||
24 | struct perf_counts_values cpu[]; | ||
25 | }; | ||
26 | |||
27 | struct perf_evsel { | ||
28 | struct list_head node; | ||
29 | struct perf_event_attr attr; | ||
30 | char *filter; | ||
31 | struct xyarray *fd; | ||
32 | struct perf_counts *counts; | ||
33 | int idx; | ||
34 | void *priv; | ||
35 | }; | ||
36 | |||
37 | struct cpu_map; | ||
38 | struct thread_map; | ||
39 | |||
40 | struct perf_evsel *perf_evsel__new(u32 type, u64 config, int idx); | ||
41 | void perf_evsel__delete(struct perf_evsel *evsel); | ||
42 | |||
43 | int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads); | ||
44 | int perf_evsel__alloc_counts(struct perf_evsel *evsel, int ncpus); | ||
45 | void perf_evsel__free_fd(struct perf_evsel *evsel); | ||
46 | void perf_evsel__close_fd(struct perf_evsel *evsel, int ncpus, int nthreads); | ||
47 | |||
48 | int perf_evsel__open_per_cpu(struct perf_evsel *evsel, struct cpu_map *cpus); | ||
49 | int perf_evsel__open_per_thread(struct perf_evsel *evsel, struct thread_map *threads); | ||
50 | int perf_evsel__open(struct perf_evsel *evsel, | ||
51 | struct cpu_map *cpus, struct thread_map *threads); | ||
52 | |||
53 | #define perf_evsel__match(evsel, t, c) \ | ||
54 | (evsel->attr.type == PERF_TYPE_##t && \ | ||
55 | evsel->attr.config == PERF_COUNT_##c) | ||
56 | |||
57 | int __perf_evsel__read_on_cpu(struct perf_evsel *evsel, | ||
58 | int cpu, int thread, bool scale); | ||
59 | |||
60 | /** | ||
61 | * perf_evsel__read_on_cpu - Read out the results on a CPU and thread | ||
62 | * | ||
63 | * @evsel - event selector to read value | ||
64 | * @cpu - CPU of interest | ||
65 | * @thread - thread of interest | ||
66 | */ | ||
67 | static inline int perf_evsel__read_on_cpu(struct perf_evsel *evsel, | ||
68 | int cpu, int thread) | ||
69 | { | ||
70 | return __perf_evsel__read_on_cpu(evsel, cpu, thread, false); | ||
71 | } | ||
72 | |||
73 | /** | ||
74 | * perf_evsel__read_on_cpu_scaled - Read out the results on a CPU and thread, scaled | ||
75 | * | ||
76 | * @evsel - event selector to read value | ||
77 | * @cpu - CPU of interest | ||
78 | * @thread - thread of interest | ||
79 | */ | ||
80 | static inline int perf_evsel__read_on_cpu_scaled(struct perf_evsel *evsel, | ||
81 | int cpu, int thread) | ||
82 | { | ||
83 | return __perf_evsel__read_on_cpu(evsel, cpu, thread, true); | ||
84 | } | ||
85 | |||
86 | int __perf_evsel__read(struct perf_evsel *evsel, int ncpus, int nthreads, | ||
87 | bool scale); | ||
88 | |||
89 | /** | ||
90 | * perf_evsel__read - Read the aggregate results on all CPUs | ||
91 | * | ||
92 | * @evsel - event selector to read value | ||
93 | * @ncpus - Number of cpus affected, from zero | ||
94 | * @nthreads - Number of threads affected, from zero | ||
95 | */ | ||
96 | static inline int perf_evsel__read(struct perf_evsel *evsel, | ||
97 | int ncpus, int nthreads) | ||
98 | { | ||
99 | return __perf_evsel__read(evsel, ncpus, nthreads, false); | ||
100 | } | ||
101 | |||
102 | /** | ||
103 | * perf_evsel__read_scaled - Read the aggregate results on all CPUs, scaled | ||
104 | * | ||
105 | * @evsel - event selector to read value | ||
106 | * @ncpus - Number of cpus affected, from zero | ||
107 | * @nthreads - Number of threads affected, from zero | ||
108 | */ | ||
109 | static inline int perf_evsel__read_scaled(struct perf_evsel *evsel, | ||
110 | int ncpus, int nthreads) | ||
111 | { | ||
112 | return __perf_evsel__read(evsel, ncpus, nthreads, true); | ||
113 | } | ||
114 | |||
115 | #endif /* __PERF_EVSEL_H */ | ||
diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 7cba0551a56..989fa2dee2f 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c | |||
@@ -152,6 +152,11 @@ void perf_header__set_feat(struct perf_header *self, int feat) | |||
152 | set_bit(feat, self->adds_features); | 152 | set_bit(feat, self->adds_features); |
153 | } | 153 | } |
154 | 154 | ||
155 | void perf_header__clear_feat(struct perf_header *self, int feat) | ||
156 | { | ||
157 | clear_bit(feat, self->adds_features); | ||
158 | } | ||
159 | |||
155 | bool perf_header__has_feat(const struct perf_header *self, int feat) | 160 | bool perf_header__has_feat(const struct perf_header *self, int feat) |
156 | { | 161 | { |
157 | return test_bit(feat, self->adds_features); | 162 | return test_bit(feat, self->adds_features); |
@@ -433,8 +438,10 @@ static int perf_header__adds_write(struct perf_header *self, int fd) | |||
433 | int idx = 0, err; | 438 | int idx = 0, err; |
434 | 439 | ||
435 | session = container_of(self, struct perf_session, header); | 440 | session = container_of(self, struct perf_session, header); |
436 | if (perf_session__read_build_ids(session, true)) | 441 | |
437 | perf_header__set_feat(self, HEADER_BUILD_ID); | 442 | if (perf_header__has_feat(self, HEADER_BUILD_ID && |
443 | !perf_session__read_build_ids(session, true))) | ||
444 | perf_header__clear_feat(self, HEADER_BUILD_ID); | ||
438 | 445 | ||
439 | nr_sections = bitmap_weight(self->adds_features, HEADER_FEAT_BITS); | 446 | nr_sections = bitmap_weight(self->adds_features, HEADER_FEAT_BITS); |
440 | if (!nr_sections) | 447 | if (!nr_sections) |
@@ -456,7 +463,7 @@ static int perf_header__adds_write(struct perf_header *self, int fd) | |||
456 | 463 | ||
457 | /* Write trace info */ | 464 | /* Write trace info */ |
458 | trace_sec->offset = lseek(fd, 0, SEEK_CUR); | 465 | trace_sec->offset = lseek(fd, 0, SEEK_CUR); |
459 | read_tracing_data(fd, attrs, nr_counters); | 466 | read_tracing_data(fd, &evsel_list); |
460 | trace_sec->size = lseek(fd, 0, SEEK_CUR) - trace_sec->offset; | 467 | trace_sec->size = lseek(fd, 0, SEEK_CUR) - trace_sec->offset; |
461 | } | 468 | } |
462 | 469 | ||
@@ -599,7 +606,7 @@ int perf_header__write(struct perf_header *self, int fd, bool at_exit) | |||
599 | static int perf_header__getbuffer64(struct perf_header *self, | 606 | static int perf_header__getbuffer64(struct perf_header *self, |
600 | int fd, void *buf, size_t size) | 607 | int fd, void *buf, size_t size) |
601 | { | 608 | { |
602 | if (do_read(fd, buf, size) <= 0) | 609 | if (readn(fd, buf, size) <= 0) |
603 | return -1; | 610 | return -1; |
604 | 611 | ||
605 | if (self->needs_swap) | 612 | if (self->needs_swap) |
@@ -655,7 +662,7 @@ int perf_file_header__read(struct perf_file_header *self, | |||
655 | { | 662 | { |
656 | lseek(fd, 0, SEEK_SET); | 663 | lseek(fd, 0, SEEK_SET); |
657 | 664 | ||
658 | if (do_read(fd, self, sizeof(*self)) <= 0 || | 665 | if (readn(fd, self, sizeof(*self)) <= 0 || |
659 | memcmp(&self->magic, __perf_magic, sizeof(self->magic))) | 666 | memcmp(&self->magic, __perf_magic, sizeof(self->magic))) |
660 | return -1; | 667 | return -1; |
661 | 668 | ||
@@ -816,7 +823,7 @@ static int perf_file_header__read_pipe(struct perf_pipe_file_header *self, | |||
816 | struct perf_header *ph, int fd, | 823 | struct perf_header *ph, int fd, |
817 | bool repipe) | 824 | bool repipe) |
818 | { | 825 | { |
819 | if (do_read(fd, self, sizeof(*self)) <= 0 || | 826 | if (readn(fd, self, sizeof(*self)) <= 0 || |
820 | memcmp(&self->magic, __perf_magic, sizeof(self->magic))) | 827 | memcmp(&self->magic, __perf_magic, sizeof(self->magic))) |
821 | return -1; | 828 | return -1; |
822 | 829 | ||
@@ -941,6 +948,24 @@ u64 perf_header__sample_type(struct perf_header *header) | |||
941 | return type; | 948 | return type; |
942 | } | 949 | } |
943 | 950 | ||
951 | bool perf_header__sample_id_all(const struct perf_header *header) | ||
952 | { | ||
953 | bool value = false, first = true; | ||
954 | int i; | ||
955 | |||
956 | for (i = 0; i < header->attrs; i++) { | ||
957 | struct perf_header_attr *attr = header->attr[i]; | ||
958 | |||
959 | if (first) { | ||
960 | value = attr->attr.sample_id_all; | ||
961 | first = false; | ||
962 | } else if (value != attr->attr.sample_id_all) | ||
963 | die("non matching sample_id_all"); | ||
964 | } | ||
965 | |||
966 | return value; | ||
967 | } | ||
968 | |||
944 | struct perf_event_attr * | 969 | struct perf_event_attr * |
945 | perf_header__find_attr(u64 id, struct perf_header *header) | 970 | perf_header__find_attr(u64 id, struct perf_header *header) |
946 | { | 971 | { |
@@ -987,21 +1012,23 @@ int event__synthesize_attr(struct perf_event_attr *attr, u16 ids, u64 *id, | |||
987 | 1012 | ||
988 | ev = malloc(size); | 1013 | ev = malloc(size); |
989 | 1014 | ||
1015 | if (ev == NULL) | ||
1016 | return -ENOMEM; | ||
1017 | |||
990 | ev->attr.attr = *attr; | 1018 | ev->attr.attr = *attr; |
991 | memcpy(ev->attr.id, id, ids * sizeof(u64)); | 1019 | memcpy(ev->attr.id, id, ids * sizeof(u64)); |
992 | 1020 | ||
993 | ev->attr.header.type = PERF_RECORD_HEADER_ATTR; | 1021 | ev->attr.header.type = PERF_RECORD_HEADER_ATTR; |
994 | ev->attr.header.size = size; | 1022 | ev->attr.header.size = size; |
995 | 1023 | ||
996 | err = process(ev, session); | 1024 | err = process(ev, NULL, session); |
997 | 1025 | ||
998 | free(ev); | 1026 | free(ev); |
999 | 1027 | ||
1000 | return err; | 1028 | return err; |
1001 | } | 1029 | } |
1002 | 1030 | ||
1003 | int event__synthesize_attrs(struct perf_header *self, | 1031 | int event__synthesize_attrs(struct perf_header *self, event__handler_t process, |
1004 | event__handler_t process, | ||
1005 | struct perf_session *session) | 1032 | struct perf_session *session) |
1006 | { | 1033 | { |
1007 | struct perf_header_attr *attr; | 1034 | struct perf_header_attr *attr; |
@@ -1071,7 +1098,7 @@ int event__synthesize_event_type(u64 event_id, char *name, | |||
1071 | ev.event_type.header.size = sizeof(ev.event_type) - | 1098 | ev.event_type.header.size = sizeof(ev.event_type) - |
1072 | (sizeof(ev.event_type.event_type.name) - size); | 1099 | (sizeof(ev.event_type.event_type.name) - size); |
1073 | 1100 | ||
1074 | err = process(&ev, session); | 1101 | err = process(&ev, NULL, session); |
1075 | 1102 | ||
1076 | return err; | 1103 | return err; |
1077 | } | 1104 | } |
@@ -1106,8 +1133,7 @@ int event__process_event_type(event_t *self, | |||
1106 | return 0; | 1133 | return 0; |
1107 | } | 1134 | } |
1108 | 1135 | ||
1109 | int event__synthesize_tracing_data(int fd, struct perf_event_attr *pattrs, | 1136 | int event__synthesize_tracing_data(int fd, struct list_head *pattrs, |
1110 | int nb_events, | ||
1111 | event__handler_t process, | 1137 | event__handler_t process, |
1112 | struct perf_session *session __unused) | 1138 | struct perf_session *session __unused) |
1113 | { | 1139 | { |
@@ -1118,7 +1144,7 @@ int event__synthesize_tracing_data(int fd, struct perf_event_attr *pattrs, | |||
1118 | memset(&ev, 0, sizeof(ev)); | 1144 | memset(&ev, 0, sizeof(ev)); |
1119 | 1145 | ||
1120 | ev.tracing_data.header.type = PERF_RECORD_HEADER_TRACING_DATA; | 1146 | ev.tracing_data.header.type = PERF_RECORD_HEADER_TRACING_DATA; |
1121 | size = read_tracing_data_size(fd, pattrs, nb_events); | 1147 | size = read_tracing_data_size(fd, pattrs); |
1122 | if (size <= 0) | 1148 | if (size <= 0) |
1123 | return size; | 1149 | return size; |
1124 | aligned_size = ALIGN(size, sizeof(u64)); | 1150 | aligned_size = ALIGN(size, sizeof(u64)); |
@@ -1126,9 +1152,9 @@ int event__synthesize_tracing_data(int fd, struct perf_event_attr *pattrs, | |||
1126 | ev.tracing_data.header.size = sizeof(ev.tracing_data); | 1152 | ev.tracing_data.header.size = sizeof(ev.tracing_data); |
1127 | ev.tracing_data.size = aligned_size; | 1153 | ev.tracing_data.size = aligned_size; |
1128 | 1154 | ||
1129 | process(&ev, session); | 1155 | process(&ev, NULL, session); |
1130 | 1156 | ||
1131 | err = read_tracing_data(fd, pattrs, nb_events); | 1157 | err = read_tracing_data(fd, pattrs); |
1132 | write_padded(fd, NULL, 0, padding); | 1158 | write_padded(fd, NULL, 0, padding); |
1133 | 1159 | ||
1134 | return aligned_size; | 1160 | return aligned_size; |
@@ -1186,7 +1212,7 @@ int event__synthesize_build_id(struct dso *pos, u16 misc, | |||
1186 | ev.build_id.header.size = sizeof(ev.build_id) + len; | 1212 | ev.build_id.header.size = sizeof(ev.build_id) + len; |
1187 | memcpy(&ev.build_id.filename, pos->long_name, pos->long_name_len); | 1213 | memcpy(&ev.build_id.filename, pos->long_name, pos->long_name_len); |
1188 | 1214 | ||
1189 | err = process(&ev, session); | 1215 | err = process(&ev, NULL, session); |
1190 | 1216 | ||
1191 | return err; | 1217 | return err; |
1192 | } | 1218 | } |
diff --git a/tools/perf/util/header.h b/tools/perf/util/header.h index 402ac2454cf..33f16be7b72 100644 --- a/tools/perf/util/header.h +++ b/tools/perf/util/header.h | |||
@@ -81,9 +81,11 @@ void perf_header_attr__delete(struct perf_header_attr *self); | |||
81 | int perf_header_attr__add_id(struct perf_header_attr *self, u64 id); | 81 | int perf_header_attr__add_id(struct perf_header_attr *self, u64 id); |
82 | 82 | ||
83 | u64 perf_header__sample_type(struct perf_header *header); | 83 | u64 perf_header__sample_type(struct perf_header *header); |
84 | bool perf_header__sample_id_all(const struct perf_header *header); | ||
84 | struct perf_event_attr * | 85 | struct perf_event_attr * |
85 | perf_header__find_attr(u64 id, struct perf_header *header); | 86 | perf_header__find_attr(u64 id, struct perf_header *header); |
86 | void perf_header__set_feat(struct perf_header *self, int feat); | 87 | void perf_header__set_feat(struct perf_header *self, int feat); |
88 | void perf_header__clear_feat(struct perf_header *self, int feat); | ||
87 | bool perf_header__has_feat(const struct perf_header *self, int feat); | 89 | bool perf_header__has_feat(const struct perf_header *self, int feat); |
88 | 90 | ||
89 | int perf_header__process_sections(struct perf_header *self, int fd, | 91 | int perf_header__process_sections(struct perf_header *self, int fd, |
@@ -111,8 +113,7 @@ int event__synthesize_event_types(event__handler_t process, | |||
111 | int event__process_event_type(event_t *self, | 113 | int event__process_event_type(event_t *self, |
112 | struct perf_session *session); | 114 | struct perf_session *session); |
113 | 115 | ||
114 | int event__synthesize_tracing_data(int fd, struct perf_event_attr *pattrs, | 116 | int event__synthesize_tracing_data(int fd, struct list_head *pattrs, |
115 | int nb_events, | ||
116 | event__handler_t process, | 117 | event__handler_t process, |
117 | struct perf_session *session); | 118 | struct perf_session *session); |
118 | int event__process_tracing_data(event_t *self, | 119 | int event__process_tracing_data(event_t *self, |
diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c index 76bcc35cf9b..c749ba6136a 100644 --- a/tools/perf/util/hist.c +++ b/tools/perf/util/hist.c | |||
@@ -1092,6 +1092,12 @@ int hist_entry__annotate(struct hist_entry *self, struct list_head *head, | |||
1092 | FILE *file; | 1092 | FILE *file; |
1093 | int err = 0; | 1093 | int err = 0; |
1094 | u64 len; | 1094 | u64 len; |
1095 | char symfs_filename[PATH_MAX]; | ||
1096 | |||
1097 | if (filename) { | ||
1098 | snprintf(symfs_filename, sizeof(symfs_filename), "%s%s", | ||
1099 | symbol_conf.symfs, filename); | ||
1100 | } | ||
1095 | 1101 | ||
1096 | if (filename == NULL) { | 1102 | if (filename == NULL) { |
1097 | if (dso->has_build_id) { | 1103 | if (dso->has_build_id) { |
@@ -1100,9 +1106,9 @@ int hist_entry__annotate(struct hist_entry *self, struct list_head *head, | |||
1100 | return -ENOMEM; | 1106 | return -ENOMEM; |
1101 | } | 1107 | } |
1102 | goto fallback; | 1108 | goto fallback; |
1103 | } else if (readlink(filename, command, sizeof(command)) < 0 || | 1109 | } else if (readlink(symfs_filename, command, sizeof(command)) < 0 || |
1104 | strstr(command, "[kernel.kallsyms]") || | 1110 | strstr(command, "[kernel.kallsyms]") || |
1105 | access(filename, R_OK)) { | 1111 | access(symfs_filename, R_OK)) { |
1106 | free(filename); | 1112 | free(filename); |
1107 | fallback: | 1113 | fallback: |
1108 | /* | 1114 | /* |
@@ -1111,6 +1117,8 @@ fallback: | |||
1111 | * DSO is the same as when 'perf record' ran. | 1117 | * DSO is the same as when 'perf record' ran. |
1112 | */ | 1118 | */ |
1113 | filename = dso->long_name; | 1119 | filename = dso->long_name; |
1120 | snprintf(symfs_filename, sizeof(symfs_filename), "%s%s", | ||
1121 | symbol_conf.symfs, filename); | ||
1114 | free_filename = false; | 1122 | free_filename = false; |
1115 | } | 1123 | } |
1116 | 1124 | ||
@@ -1137,7 +1145,7 @@ fallback: | |||
1137 | "objdump --start-address=0x%016Lx --stop-address=0x%016Lx -dS -C %s|grep -v %s|expand", | 1145 | "objdump --start-address=0x%016Lx --stop-address=0x%016Lx -dS -C %s|grep -v %s|expand", |
1138 | map__rip_2objdump(map, sym->start), | 1146 | map__rip_2objdump(map, sym->start), |
1139 | map__rip_2objdump(map, sym->end), | 1147 | map__rip_2objdump(map, sym->end), |
1140 | filename, filename); | 1148 | symfs_filename, filename); |
1141 | 1149 | ||
1142 | pr_debug("Executing: %s\n", command); | 1150 | pr_debug("Executing: %s\n", command); |
1143 | 1151 | ||
@@ -1168,10 +1176,13 @@ size_t hists__fprintf_nr_events(struct hists *self, FILE *fp) | |||
1168 | size_t ret = 0; | 1176 | size_t ret = 0; |
1169 | 1177 | ||
1170 | for (i = 0; i < PERF_RECORD_HEADER_MAX; ++i) { | 1178 | for (i = 0; i < PERF_RECORD_HEADER_MAX; ++i) { |
1171 | if (!event__name[i]) | 1179 | const char *name = event__get_event_name(i); |
1180 | |||
1181 | if (!strcmp(name, "UNKNOWN")) | ||
1172 | continue; | 1182 | continue; |
1173 | ret += fprintf(fp, "%10s events: %10d\n", | 1183 | |
1174 | event__name[i], self->stats.nr_events[i]); | 1184 | ret += fprintf(fp, "%16s events: %10d\n", name, |
1185 | self->stats.nr_events[i]); | ||
1175 | } | 1186 | } |
1176 | 1187 | ||
1177 | return ret; | 1188 | return ret; |
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h index 587d375d343..ee789856a8c 100644 --- a/tools/perf/util/hist.h +++ b/tools/perf/util/hist.h | |||
@@ -52,8 +52,10 @@ struct sym_priv { | |||
52 | struct events_stats { | 52 | struct events_stats { |
53 | u64 total_period; | 53 | u64 total_period; |
54 | u64 total_lost; | 54 | u64 total_lost; |
55 | u64 total_invalid_chains; | ||
55 | u32 nr_events[PERF_RECORD_HEADER_MAX]; | 56 | u32 nr_events[PERF_RECORD_HEADER_MAX]; |
56 | u32 nr_unknown_events; | 57 | u32 nr_unknown_events; |
58 | u32 nr_invalid_chains; | ||
57 | }; | 59 | }; |
58 | 60 | ||
59 | enum hist_column { | 61 | enum hist_column { |
diff --git a/tools/perf/util/include/asm/cpufeature.h b/tools/perf/util/include/asm/cpufeature.h new file mode 100644 index 00000000000..acffd5e4d1d --- /dev/null +++ b/tools/perf/util/include/asm/cpufeature.h | |||
@@ -0,0 +1,9 @@ | |||
1 | |||
2 | #ifndef PERF_CPUFEATURE_H | ||
3 | #define PERF_CPUFEATURE_H | ||
4 | |||
5 | /* cpufeature.h ... dummy header file for including arch/x86/lib/memcpy_64.S */ | ||
6 | |||
7 | #define X86_FEATURE_REP_GOOD 0 | ||
8 | |||
9 | #endif /* PERF_CPUFEATURE_H */ | ||
diff --git a/tools/perf/util/include/asm/dwarf2.h b/tools/perf/util/include/asm/dwarf2.h new file mode 100644 index 00000000000..bb4198e7837 --- /dev/null +++ b/tools/perf/util/include/asm/dwarf2.h | |||
@@ -0,0 +1,11 @@ | |||
1 | |||
2 | #ifndef PERF_DWARF2_H | ||
3 | #define PERF_DWARF2_H | ||
4 | |||
5 | /* dwarf2.h ... dummy header file for including arch/x86/lib/memcpy_64.S */ | ||
6 | |||
7 | #define CFI_STARTPROC | ||
8 | #define CFI_ENDPROC | ||
9 | |||
10 | #endif /* PERF_DWARF2_H */ | ||
11 | |||
diff --git a/tools/perf/util/include/linux/bitops.h b/tools/perf/util/include/linux/bitops.h index bb4ac2e0538..8be0b968ca0 100644 --- a/tools/perf/util/include/linux/bitops.h +++ b/tools/perf/util/include/linux/bitops.h | |||
@@ -13,6 +13,11 @@ static inline void set_bit(int nr, unsigned long *addr) | |||
13 | addr[nr / BITS_PER_LONG] |= 1UL << (nr % BITS_PER_LONG); | 13 | addr[nr / BITS_PER_LONG] |= 1UL << (nr % BITS_PER_LONG); |
14 | } | 14 | } |
15 | 15 | ||
16 | static inline void clear_bit(int nr, unsigned long *addr) | ||
17 | { | ||
18 | addr[nr / BITS_PER_LONG] &= ~(1UL << (nr % BITS_PER_LONG)); | ||
19 | } | ||
20 | |||
16 | static __always_inline int test_bit(unsigned int nr, const unsigned long *addr) | 21 | static __always_inline int test_bit(unsigned int nr, const unsigned long *addr) |
17 | { | 22 | { |
18 | return ((1UL << (nr % BITS_PER_LONG)) & | 23 | return ((1UL << (nr % BITS_PER_LONG)) & |
diff --git a/tools/perf/util/include/linux/linkage.h b/tools/perf/util/include/linux/linkage.h new file mode 100644 index 00000000000..06387cffe12 --- /dev/null +++ b/tools/perf/util/include/linux/linkage.h | |||
@@ -0,0 +1,13 @@ | |||
1 | |||
2 | #ifndef PERF_LINUX_LINKAGE_H_ | ||
3 | #define PERF_LINUX_LINKAGE_H_ | ||
4 | |||
5 | /* linkage.h ... for including arch/x86/lib/memcpy_64.S */ | ||
6 | |||
7 | #define ENTRY(name) \ | ||
8 | .globl name; \ | ||
9 | name: | ||
10 | |||
11 | #define ENDPROC(name) | ||
12 | |||
13 | #endif /* PERF_LINUX_LINKAGE_H_ */ | ||
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index 4af5bd59cfd..649083f27e0 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c | |||
@@ -1,6 +1,7 @@ | |||
1 | #include "../../../include/linux/hw_breakpoint.h" | 1 | #include "../../../include/linux/hw_breakpoint.h" |
2 | #include "util.h" | 2 | #include "util.h" |
3 | #include "../perf.h" | 3 | #include "../perf.h" |
4 | #include "evsel.h" | ||
4 | #include "parse-options.h" | 5 | #include "parse-options.h" |
5 | #include "parse-events.h" | 6 | #include "parse-events.h" |
6 | #include "exec_cmd.h" | 7 | #include "exec_cmd.h" |
@@ -12,8 +13,7 @@ | |||
12 | 13 | ||
13 | int nr_counters; | 14 | int nr_counters; |
14 | 15 | ||
15 | struct perf_event_attr attrs[MAX_COUNTERS]; | 16 | LIST_HEAD(evsel_list); |
16 | char *filters[MAX_COUNTERS]; | ||
17 | 17 | ||
18 | struct event_symbol { | 18 | struct event_symbol { |
19 | u8 type; | 19 | u8 type; |
@@ -266,10 +266,10 @@ static char *event_cache_name(u8 cache_type, u8 cache_op, u8 cache_result) | |||
266 | return name; | 266 | return name; |
267 | } | 267 | } |
268 | 268 | ||
269 | const char *event_name(int counter) | 269 | const char *event_name(struct perf_evsel *evsel) |
270 | { | 270 | { |
271 | u64 config = attrs[counter].config; | 271 | u64 config = evsel->attr.config; |
272 | int type = attrs[counter].type; | 272 | int type = evsel->attr.type; |
273 | 273 | ||
274 | return __event_name(type, config); | 274 | return __event_name(type, config); |
275 | } | 275 | } |
@@ -434,7 +434,7 @@ parse_single_tracepoint_event(char *sys_name, | |||
434 | id = atoll(id_buf); | 434 | id = atoll(id_buf); |
435 | attr->config = id; | 435 | attr->config = id; |
436 | attr->type = PERF_TYPE_TRACEPOINT; | 436 | attr->type = PERF_TYPE_TRACEPOINT; |
437 | *strp = evt_name + evt_length; | 437 | *strp += strlen(sys_name) + evt_length + 1; /* + 1 for the ':' */ |
438 | 438 | ||
439 | attr->sample_type |= PERF_SAMPLE_RAW; | 439 | attr->sample_type |= PERF_SAMPLE_RAW; |
440 | attr->sample_type |= PERF_SAMPLE_TIME; | 440 | attr->sample_type |= PERF_SAMPLE_TIME; |
@@ -495,7 +495,7 @@ static enum event_result parse_tracepoint_event(const char **strp, | |||
495 | struct perf_event_attr *attr) | 495 | struct perf_event_attr *attr) |
496 | { | 496 | { |
497 | const char *evt_name; | 497 | const char *evt_name; |
498 | char *flags; | 498 | char *flags = NULL, *comma_loc; |
499 | char sys_name[MAX_EVENT_LENGTH]; | 499 | char sys_name[MAX_EVENT_LENGTH]; |
500 | unsigned int sys_length, evt_length; | 500 | unsigned int sys_length, evt_length; |
501 | 501 | ||
@@ -514,6 +514,11 @@ static enum event_result parse_tracepoint_event(const char **strp, | |||
514 | sys_name[sys_length] = '\0'; | 514 | sys_name[sys_length] = '\0'; |
515 | evt_name = evt_name + 1; | 515 | evt_name = evt_name + 1; |
516 | 516 | ||
517 | comma_loc = strchr(evt_name, ','); | ||
518 | if (comma_loc) { | ||
519 | /* take the event name up to the comma */ | ||
520 | evt_name = strndup(evt_name, comma_loc - evt_name); | ||
521 | } | ||
517 | flags = strchr(evt_name, ':'); | 522 | flags = strchr(evt_name, ':'); |
518 | if (flags) { | 523 | if (flags) { |
519 | /* split it out: */ | 524 | /* split it out: */ |
@@ -524,9 +529,8 @@ static enum event_result parse_tracepoint_event(const char **strp, | |||
524 | evt_length = strlen(evt_name); | 529 | evt_length = strlen(evt_name); |
525 | if (evt_length >= MAX_EVENT_LENGTH) | 530 | if (evt_length >= MAX_EVENT_LENGTH) |
526 | return EVT_FAILED; | 531 | return EVT_FAILED; |
527 | |||
528 | if (strpbrk(evt_name, "*?")) { | 532 | if (strpbrk(evt_name, "*?")) { |
529 | *strp = evt_name + evt_length; | 533 | *strp += strlen(sys_name) + evt_length; |
530 | return parse_multiple_tracepoint_event(sys_name, evt_name, | 534 | return parse_multiple_tracepoint_event(sys_name, evt_name, |
531 | flags); | 535 | flags); |
532 | } else | 536 | } else |
@@ -810,9 +814,6 @@ int parse_events(const struct option *opt __used, const char *str, int unset __u | |||
810 | return -1; | 814 | return -1; |
811 | 815 | ||
812 | for (;;) { | 816 | for (;;) { |
813 | if (nr_counters == MAX_COUNTERS) | ||
814 | return -1; | ||
815 | |||
816 | memset(&attr, 0, sizeof(attr)); | 817 | memset(&attr, 0, sizeof(attr)); |
817 | ret = parse_event_symbols(&str, &attr); | 818 | ret = parse_event_symbols(&str, &attr); |
818 | if (ret == EVT_FAILED) | 819 | if (ret == EVT_FAILED) |
@@ -822,8 +823,13 @@ int parse_events(const struct option *opt __used, const char *str, int unset __u | |||
822 | return -1; | 823 | return -1; |
823 | 824 | ||
824 | if (ret != EVT_HANDLED_ALL) { | 825 | if (ret != EVT_HANDLED_ALL) { |
825 | attrs[nr_counters] = attr; | 826 | struct perf_evsel *evsel; |
826 | nr_counters++; | 827 | evsel = perf_evsel__new(attr.type, attr.config, |
828 | nr_counters); | ||
829 | if (evsel == NULL) | ||
830 | return -1; | ||
831 | list_add_tail(&evsel->node, &evsel_list); | ||
832 | ++nr_counters; | ||
827 | } | 833 | } |
828 | 834 | ||
829 | if (*str == 0) | 835 | if (*str == 0) |
@@ -840,21 +846,22 @@ int parse_events(const struct option *opt __used, const char *str, int unset __u | |||
840 | int parse_filter(const struct option *opt __used, const char *str, | 846 | int parse_filter(const struct option *opt __used, const char *str, |
841 | int unset __used) | 847 | int unset __used) |
842 | { | 848 | { |
843 | int i = nr_counters - 1; | 849 | struct perf_evsel *last = NULL; |
844 | int len = strlen(str); | ||
845 | 850 | ||
846 | if (i < 0 || attrs[i].type != PERF_TYPE_TRACEPOINT) { | 851 | if (!list_empty(&evsel_list)) |
852 | last = list_entry(evsel_list.prev, struct perf_evsel, node); | ||
853 | |||
854 | if (last == NULL || last->attr.type != PERF_TYPE_TRACEPOINT) { | ||
847 | fprintf(stderr, | 855 | fprintf(stderr, |
848 | "-F option should follow a -e tracepoint option\n"); | 856 | "-F option should follow a -e tracepoint option\n"); |
849 | return -1; | 857 | return -1; |
850 | } | 858 | } |
851 | 859 | ||
852 | filters[i] = malloc(len + 1); | 860 | last->filter = strdup(str); |
853 | if (!filters[i]) { | 861 | if (last->filter == NULL) { |
854 | fprintf(stderr, "not enough memory to hold filter string\n"); | 862 | fprintf(stderr, "not enough memory to hold filter string\n"); |
855 | return -1; | 863 | return -1; |
856 | } | 864 | } |
857 | strcpy(filters[i], str); | ||
858 | 865 | ||
859 | return 0; | 866 | return 0; |
860 | } | 867 | } |
@@ -906,6 +913,47 @@ static void print_tracepoint_events(void) | |||
906 | } | 913 | } |
907 | 914 | ||
908 | /* | 915 | /* |
916 | * Check whether event is in <debugfs_mount_point>/tracing/events | ||
917 | */ | ||
918 | |||
919 | int is_valid_tracepoint(const char *event_string) | ||
920 | { | ||
921 | DIR *sys_dir, *evt_dir; | ||
922 | struct dirent *sys_next, *evt_next, sys_dirent, evt_dirent; | ||
923 | char evt_path[MAXPATHLEN]; | ||
924 | char dir_path[MAXPATHLEN]; | ||
925 | |||
926 | if (debugfs_valid_mountpoint(debugfs_path)) | ||
927 | return 0; | ||
928 | |||
929 | sys_dir = opendir(debugfs_path); | ||
930 | if (!sys_dir) | ||
931 | return 0; | ||
932 | |||
933 | for_each_subsystem(sys_dir, sys_dirent, sys_next) { | ||
934 | |||
935 | snprintf(dir_path, MAXPATHLEN, "%s/%s", debugfs_path, | ||
936 | sys_dirent.d_name); | ||
937 | evt_dir = opendir(dir_path); | ||
938 | if (!evt_dir) | ||
939 | continue; | ||
940 | |||
941 | for_each_event(sys_dirent, evt_dir, evt_dirent, evt_next) { | ||
942 | snprintf(evt_path, MAXPATHLEN, "%s:%s", | ||
943 | sys_dirent.d_name, evt_dirent.d_name); | ||
944 | if (!strcmp(evt_path, event_string)) { | ||
945 | closedir(evt_dir); | ||
946 | closedir(sys_dir); | ||
947 | return 1; | ||
948 | } | ||
949 | } | ||
950 | closedir(evt_dir); | ||
951 | } | ||
952 | closedir(sys_dir); | ||
953 | return 0; | ||
954 | } | ||
955 | |||
956 | /* | ||
909 | * Print the help text for the event symbols: | 957 | * Print the help text for the event symbols: |
910 | */ | 958 | */ |
911 | void print_events(void) | 959 | void print_events(void) |
@@ -963,3 +1011,26 @@ void print_events(void) | |||
963 | 1011 | ||
964 | exit(129); | 1012 | exit(129); |
965 | } | 1013 | } |
1014 | |||
1015 | int perf_evsel_list__create_default(void) | ||
1016 | { | ||
1017 | struct perf_evsel *evsel = perf_evsel__new(PERF_TYPE_HARDWARE, | ||
1018 | PERF_COUNT_HW_CPU_CYCLES, 0); | ||
1019 | if (evsel == NULL) | ||
1020 | return -ENOMEM; | ||
1021 | |||
1022 | list_add(&evsel->node, &evsel_list); | ||
1023 | ++nr_counters; | ||
1024 | return 0; | ||
1025 | } | ||
1026 | |||
1027 | void perf_evsel_list__delete(void) | ||
1028 | { | ||
1029 | struct perf_evsel *pos, *n; | ||
1030 | |||
1031 | list_for_each_entry_safe(pos, n, &evsel_list, node) { | ||
1032 | list_del_init(&pos->node); | ||
1033 | perf_evsel__delete(pos); | ||
1034 | } | ||
1035 | nr_counters = 0; | ||
1036 | } | ||
diff --git a/tools/perf/util/parse-events.h b/tools/perf/util/parse-events.h index fc4ab3fe877..b82cafb8377 100644 --- a/tools/perf/util/parse-events.h +++ b/tools/perf/util/parse-events.h | |||
@@ -4,6 +4,16 @@ | |||
4 | * Parse symbolic events/counts passed in as options: | 4 | * Parse symbolic events/counts passed in as options: |
5 | */ | 5 | */ |
6 | 6 | ||
7 | #include "../../../include/linux/perf_event.h" | ||
8 | |||
9 | struct list_head; | ||
10 | struct perf_evsel; | ||
11 | |||
12 | extern struct list_head evsel_list; | ||
13 | |||
14 | int perf_evsel_list__create_default(void); | ||
15 | void perf_evsel_list__delete(void); | ||
16 | |||
7 | struct option; | 17 | struct option; |
8 | 18 | ||
9 | struct tracepoint_path { | 19 | struct tracepoint_path { |
@@ -13,14 +23,11 @@ struct tracepoint_path { | |||
13 | }; | 23 | }; |
14 | 24 | ||
15 | extern struct tracepoint_path *tracepoint_id_to_path(u64 config); | 25 | extern struct tracepoint_path *tracepoint_id_to_path(u64 config); |
16 | extern bool have_tracepoints(struct perf_event_attr *pattrs, int nb_events); | 26 | extern bool have_tracepoints(struct list_head *evsel_list); |
17 | 27 | ||
18 | extern int nr_counters; | 28 | extern int nr_counters; |
19 | 29 | ||
20 | extern struct perf_event_attr attrs[MAX_COUNTERS]; | 30 | const char *event_name(struct perf_evsel *event); |
21 | extern char *filters[MAX_COUNTERS]; | ||
22 | |||
23 | extern const char *event_name(int ctr); | ||
24 | extern const char *__event_name(int type, u64 config); | 31 | extern const char *__event_name(int type, u64 config); |
25 | 32 | ||
26 | extern int parse_events(const struct option *opt, const char *str, int unset); | 33 | extern int parse_events(const struct option *opt, const char *str, int unset); |
@@ -29,9 +36,9 @@ extern int parse_filter(const struct option *opt, const char *str, int unset); | |||
29 | #define EVENTS_HELP_MAX (128*1024) | 36 | #define EVENTS_HELP_MAX (128*1024) |
30 | 37 | ||
31 | extern void print_events(void); | 38 | extern void print_events(void); |
39 | extern int is_valid_tracepoint(const char *event_string); | ||
32 | 40 | ||
33 | extern char debugfs_path[]; | 41 | extern char debugfs_path[]; |
34 | extern int valid_debugfs_mount(const char *debugfs); | 42 | extern int valid_debugfs_mount(const char *debugfs); |
35 | 43 | ||
36 | |||
37 | #endif /* __PERF_PARSE_EVENTS_H */ | 44 | #endif /* __PERF_PARSE_EVENTS_H */ |
diff --git a/tools/perf/util/parse-options.h b/tools/perf/util/parse-options.h index c7d72dce54b..abc31a1dac1 100644 --- a/tools/perf/util/parse-options.h +++ b/tools/perf/util/parse-options.h | |||
@@ -119,6 +119,10 @@ struct option { | |||
119 | { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .flags = PARSE_OPT_NOARG } | 119 | { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .flags = PARSE_OPT_NOARG } |
120 | #define OPT_CALLBACK_DEFAULT(s, l, v, a, h, f, d) \ | 120 | #define OPT_CALLBACK_DEFAULT(s, l, v, a, h, f, d) \ |
121 | { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .defval = (intptr_t)d, .flags = PARSE_OPT_LASTARG_DEFAULT } | 121 | { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l), .value = (v), (a), .help = (h), .callback = (f), .defval = (intptr_t)d, .flags = PARSE_OPT_LASTARG_DEFAULT } |
122 | #define OPT_CALLBACK_DEFAULT_NOOPT(s, l, v, a, h, f, d) \ | ||
123 | { .type = OPTION_CALLBACK, .short_name = (s), .long_name = (l),\ | ||
124 | .value = (v), (a), .help = (h), .callback = (f), .defval = (intptr_t)d,\ | ||
125 | .flags = PARSE_OPT_LASTARG_DEFAULT | PARSE_OPT_NOARG} | ||
122 | 126 | ||
123 | /* parse_options() will filter out the processed options and leave the | 127 | /* parse_options() will filter out the processed options and leave the |
124 | * non-option argments in argv[]. | 128 | * non-option argments in argv[]. |
diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c index 61191c6cbe7..128aaab0aed 100644 --- a/tools/perf/util/probe-event.c +++ b/tools/perf/util/probe-event.c | |||
@@ -95,7 +95,7 @@ static int init_vmlinux(void) | |||
95 | goto out; | 95 | goto out; |
96 | 96 | ||
97 | if (machine__create_kernel_maps(&machine) < 0) { | 97 | if (machine__create_kernel_maps(&machine) < 0) { |
98 | pr_debug("machine__create_kernel_maps "); | 98 | pr_debug("machine__create_kernel_maps() failed.\n"); |
99 | goto out; | 99 | goto out; |
100 | } | 100 | } |
101 | out: | 101 | out: |
@@ -149,7 +149,8 @@ static int open_vmlinux(const char *module) | |||
149 | { | 149 | { |
150 | const char *path = kernel_get_module_path(module); | 150 | const char *path = kernel_get_module_path(module); |
151 | if (!path) { | 151 | if (!path) { |
152 | pr_err("Failed to find path of %s module", module ?: "kernel"); | 152 | pr_err("Failed to find path of %s module.\n", |
153 | module ?: "kernel"); | ||
153 | return -ENOENT; | 154 | return -ENOENT; |
154 | } | 155 | } |
155 | pr_debug("Try to open %s\n", path); | 156 | pr_debug("Try to open %s\n", path); |
@@ -226,7 +227,7 @@ static int try_to_find_probe_trace_events(struct perf_probe_event *pev, | |||
226 | pr_warning("Warning: No dwarf info found in the vmlinux - " | 227 | pr_warning("Warning: No dwarf info found in the vmlinux - " |
227 | "please rebuild kernel with CONFIG_DEBUG_INFO=y.\n"); | 228 | "please rebuild kernel with CONFIG_DEBUG_INFO=y.\n"); |
228 | if (!need_dwarf) { | 229 | if (!need_dwarf) { |
229 | pr_debug("Trying to use symbols.\nn"); | 230 | pr_debug("Trying to use symbols.\n"); |
230 | return 0; | 231 | return 0; |
231 | } | 232 | } |
232 | } | 233 | } |
@@ -295,42 +296,49 @@ static int get_real_path(const char *raw_path, const char *comp_dir, | |||
295 | #define LINEBUF_SIZE 256 | 296 | #define LINEBUF_SIZE 256 |
296 | #define NR_ADDITIONAL_LINES 2 | 297 | #define NR_ADDITIONAL_LINES 2 |
297 | 298 | ||
298 | static int show_one_line(FILE *fp, int l, bool skip, bool show_num) | 299 | static int __show_one_line(FILE *fp, int l, bool skip, bool show_num) |
299 | { | 300 | { |
300 | char buf[LINEBUF_SIZE]; | 301 | char buf[LINEBUF_SIZE]; |
301 | const char *color = PERF_COLOR_BLUE; | 302 | const char *color = show_num ? "" : PERF_COLOR_BLUE; |
303 | const char *prefix = NULL; | ||
302 | 304 | ||
303 | if (fgets(buf, LINEBUF_SIZE, fp) == NULL) | 305 | do { |
304 | goto error; | ||
305 | if (!skip) { | ||
306 | if (show_num) | ||
307 | fprintf(stdout, "%7d %s", l, buf); | ||
308 | else | ||
309 | color_fprintf(stdout, color, " %s", buf); | ||
310 | } | ||
311 | |||
312 | while (strlen(buf) == LINEBUF_SIZE - 1 && | ||
313 | buf[LINEBUF_SIZE - 2] != '\n') { | ||
314 | if (fgets(buf, LINEBUF_SIZE, fp) == NULL) | 306 | if (fgets(buf, LINEBUF_SIZE, fp) == NULL) |
315 | goto error; | 307 | goto error; |
316 | if (!skip) { | 308 | if (skip) |
317 | if (show_num) | 309 | continue; |
318 | fprintf(stdout, "%s", buf); | 310 | if (!prefix) { |
319 | else | 311 | prefix = show_num ? "%7d " : " "; |
320 | color_fprintf(stdout, color, "%s", buf); | 312 | color_fprintf(stdout, color, prefix, l); |
321 | } | 313 | } |
322 | } | 314 | color_fprintf(stdout, color, "%s", buf); |
323 | 315 | ||
324 | return 0; | 316 | } while (strchr(buf, '\n') == NULL); |
317 | |||
318 | return 1; | ||
325 | error: | 319 | error: |
326 | if (feof(fp)) | 320 | if (ferror(fp)) { |
327 | pr_warning("Source file is shorter than expected.\n"); | ||
328 | else | ||
329 | pr_warning("File read error: %s\n", strerror(errno)); | 321 | pr_warning("File read error: %s\n", strerror(errno)); |
322 | return -1; | ||
323 | } | ||
324 | return 0; | ||
325 | } | ||
330 | 326 | ||
331 | return -1; | 327 | static int _show_one_line(FILE *fp, int l, bool skip, bool show_num) |
328 | { | ||
329 | int rv = __show_one_line(fp, l, skip, show_num); | ||
330 | if (rv == 0) { | ||
331 | pr_warning("Source file is shorter than expected.\n"); | ||
332 | rv = -1; | ||
333 | } | ||
334 | return rv; | ||
332 | } | 335 | } |
333 | 336 | ||
337 | #define show_one_line_with_num(f,l) _show_one_line(f,l,false,true) | ||
338 | #define show_one_line(f,l) _show_one_line(f,l,false,false) | ||
339 | #define skip_one_line(f,l) _show_one_line(f,l,true,false) | ||
340 | #define show_one_line_or_eof(f,l) __show_one_line(f,l,false,false) | ||
341 | |||
334 | /* | 342 | /* |
335 | * Show line-range always requires debuginfo to find source file and | 343 | * Show line-range always requires debuginfo to find source file and |
336 | * line number. | 344 | * line number. |
@@ -379,7 +387,7 @@ int show_line_range(struct line_range *lr, const char *module) | |||
379 | fprintf(stdout, "<%s:%d>\n", lr->function, | 387 | fprintf(stdout, "<%s:%d>\n", lr->function, |
380 | lr->start - lr->offset); | 388 | lr->start - lr->offset); |
381 | else | 389 | else |
382 | fprintf(stdout, "<%s:%d>\n", lr->file, lr->start); | 390 | fprintf(stdout, "<%s:%d>\n", lr->path, lr->start); |
383 | 391 | ||
384 | fp = fopen(lr->path, "r"); | 392 | fp = fopen(lr->path, "r"); |
385 | if (fp == NULL) { | 393 | if (fp == NULL) { |
@@ -388,26 +396,30 @@ int show_line_range(struct line_range *lr, const char *module) | |||
388 | return -errno; | 396 | return -errno; |
389 | } | 397 | } |
390 | /* Skip to starting line number */ | 398 | /* Skip to starting line number */ |
391 | while (l < lr->start && ret >= 0) | 399 | while (l < lr->start) { |
392 | ret = show_one_line(fp, l++, true, false); | 400 | ret = skip_one_line(fp, l++); |
393 | if (ret < 0) | 401 | if (ret < 0) |
394 | goto end; | 402 | goto end; |
403 | } | ||
395 | 404 | ||
396 | list_for_each_entry(ln, &lr->line_list, list) { | 405 | list_for_each_entry(ln, &lr->line_list, list) { |
397 | while (ln->line > l && ret >= 0) | 406 | for (; ln->line > l; l++) { |
398 | ret = show_one_line(fp, (l++) - lr->offset, | 407 | ret = show_one_line(fp, l - lr->offset); |
399 | false, false); | 408 | if (ret < 0) |
400 | if (ret >= 0) | 409 | goto end; |
401 | ret = show_one_line(fp, (l++) - lr->offset, | 410 | } |
402 | false, true); | 411 | ret = show_one_line_with_num(fp, l++ - lr->offset); |
403 | if (ret < 0) | 412 | if (ret < 0) |
404 | goto end; | 413 | goto end; |
405 | } | 414 | } |
406 | 415 | ||
407 | if (lr->end == INT_MAX) | 416 | if (lr->end == INT_MAX) |
408 | lr->end = l + NR_ADDITIONAL_LINES; | 417 | lr->end = l + NR_ADDITIONAL_LINES; |
409 | while (l <= lr->end && !feof(fp) && ret >= 0) | 418 | while (l <= lr->end) { |
410 | ret = show_one_line(fp, (l++) - lr->offset, false, false); | 419 | ret = show_one_line_or_eof(fp, l++ - lr->offset); |
420 | if (ret <= 0) | ||
421 | break; | ||
422 | } | ||
411 | end: | 423 | end: |
412 | fclose(fp); | 424 | fclose(fp); |
413 | return ret; | 425 | return ret; |
@@ -466,7 +478,7 @@ int show_available_vars(struct perf_probe_event *pevs, int npevs, | |||
466 | 478 | ||
467 | fd = open_vmlinux(module); | 479 | fd = open_vmlinux(module); |
468 | if (fd < 0) { | 480 | if (fd < 0) { |
469 | pr_warning("Failed to open debuginfo file.\n"); | 481 | pr_warning("Failed to open debug information file.\n"); |
470 | return fd; | 482 | return fd; |
471 | } | 483 | } |
472 | 484 | ||
@@ -526,56 +538,87 @@ int show_available_vars(struct perf_probe_event *pevs __unused, | |||
526 | } | 538 | } |
527 | #endif | 539 | #endif |
528 | 540 | ||
541 | static int parse_line_num(char **ptr, int *val, const char *what) | ||
542 | { | ||
543 | const char *start = *ptr; | ||
544 | |||
545 | errno = 0; | ||
546 | *val = strtol(*ptr, ptr, 0); | ||
547 | if (errno || *ptr == start) { | ||
548 | semantic_error("'%s' is not a valid number.\n", what); | ||
549 | return -EINVAL; | ||
550 | } | ||
551 | return 0; | ||
552 | } | ||
553 | |||
554 | /* | ||
555 | * Stuff 'lr' according to the line range described by 'arg'. | ||
556 | * The line range syntax is described by: | ||
557 | * | ||
558 | * SRC[:SLN[+NUM|-ELN]] | ||
559 | * FNC[:SLN[+NUM|-ELN]] | ||
560 | */ | ||
529 | int parse_line_range_desc(const char *arg, struct line_range *lr) | 561 | int parse_line_range_desc(const char *arg, struct line_range *lr) |
530 | { | 562 | { |
531 | const char *ptr; | 563 | char *range, *name = strdup(arg); |
532 | char *tmp; | 564 | int err; |
533 | /* | 565 | |
534 | * <Syntax> | 566 | if (!name) |
535 | * SRC:SLN[+NUM|-ELN] | 567 | return -ENOMEM; |
536 | * FUNC[:SLN[+NUM|-ELN]] | 568 | |
537 | */ | 569 | lr->start = 0; |
538 | ptr = strchr(arg, ':'); | 570 | lr->end = INT_MAX; |
539 | if (ptr) { | 571 | |
540 | lr->start = (int)strtoul(ptr + 1, &tmp, 0); | 572 | range = strchr(name, ':'); |
541 | if (*tmp == '+') { | 573 | if (range) { |
542 | lr->end = lr->start + (int)strtoul(tmp + 1, &tmp, 0); | 574 | *range++ = '\0'; |
543 | lr->end--; /* | 575 | |
544 | * Adjust the number of lines here. | 576 | err = parse_line_num(&range, &lr->start, "start line"); |
545 | * If the number of lines == 1, the | 577 | if (err) |
546 | * the end of line should be equal to | 578 | goto err; |
547 | * the start of line. | 579 | |
548 | */ | 580 | if (*range == '+' || *range == '-') { |
549 | } else if (*tmp == '-') | 581 | const char c = *range++; |
550 | lr->end = (int)strtoul(tmp + 1, &tmp, 0); | 582 | |
551 | else | 583 | err = parse_line_num(&range, &lr->end, "end line"); |
552 | lr->end = INT_MAX; | 584 | if (err) |
585 | goto err; | ||
586 | |||
587 | if (c == '+') { | ||
588 | lr->end += lr->start; | ||
589 | /* | ||
590 | * Adjust the number of lines here. | ||
591 | * If the number of lines == 1, the | ||
592 | * the end of line should be equal to | ||
593 | * the start of line. | ||
594 | */ | ||
595 | lr->end--; | ||
596 | } | ||
597 | } | ||
598 | |||
553 | pr_debug("Line range is %d to %d\n", lr->start, lr->end); | 599 | pr_debug("Line range is %d to %d\n", lr->start, lr->end); |
600 | |||
601 | err = -EINVAL; | ||
554 | if (lr->start > lr->end) { | 602 | if (lr->start > lr->end) { |
555 | semantic_error("Start line must be smaller" | 603 | semantic_error("Start line must be smaller" |
556 | " than end line.\n"); | 604 | " than end line.\n"); |
557 | return -EINVAL; | 605 | goto err; |
558 | } | 606 | } |
559 | if (*tmp != '\0') { | 607 | if (*range != '\0') { |
560 | semantic_error("Tailing with invalid character '%d'.\n", | 608 | semantic_error("Tailing with invalid str '%s'.\n", range); |
561 | *tmp); | 609 | goto err; |
562 | return -EINVAL; | ||
563 | } | 610 | } |
564 | tmp = strndup(arg, (ptr - arg)); | ||
565 | } else { | ||
566 | tmp = strdup(arg); | ||
567 | lr->end = INT_MAX; | ||
568 | } | 611 | } |
569 | 612 | ||
570 | if (tmp == NULL) | 613 | if (strchr(name, '.')) |
571 | return -ENOMEM; | 614 | lr->file = name; |
572 | |||
573 | if (strchr(tmp, '.')) | ||
574 | lr->file = tmp; | ||
575 | else | 615 | else |
576 | lr->function = tmp; | 616 | lr->function = name; |
577 | 617 | ||
578 | return 0; | 618 | return 0; |
619 | err: | ||
620 | free(name); | ||
621 | return err; | ||
579 | } | 622 | } |
580 | 623 | ||
581 | /* Check the name is good for event/group */ | 624 | /* Check the name is good for event/group */ |
@@ -699,39 +742,40 @@ static int parse_perf_probe_point(char *arg, struct perf_probe_event *pev) | |||
699 | 742 | ||
700 | /* Exclusion check */ | 743 | /* Exclusion check */ |
701 | if (pp->lazy_line && pp->line) { | 744 | if (pp->lazy_line && pp->line) { |
702 | semantic_error("Lazy pattern can't be used with line number."); | 745 | semantic_error("Lazy pattern can't be used with" |
746 | " line number.\n"); | ||
703 | return -EINVAL; | 747 | return -EINVAL; |
704 | } | 748 | } |
705 | 749 | ||
706 | if (pp->lazy_line && pp->offset) { | 750 | if (pp->lazy_line && pp->offset) { |
707 | semantic_error("Lazy pattern can't be used with offset."); | 751 | semantic_error("Lazy pattern can't be used with offset.\n"); |
708 | return -EINVAL; | 752 | return -EINVAL; |
709 | } | 753 | } |
710 | 754 | ||
711 | if (pp->line && pp->offset) { | 755 | if (pp->line && pp->offset) { |
712 | semantic_error("Offset can't be used with line number."); | 756 | semantic_error("Offset can't be used with line number.\n"); |
713 | return -EINVAL; | 757 | return -EINVAL; |
714 | } | 758 | } |
715 | 759 | ||
716 | if (!pp->line && !pp->lazy_line && pp->file && !pp->function) { | 760 | if (!pp->line && !pp->lazy_line && pp->file && !pp->function) { |
717 | semantic_error("File always requires line number or " | 761 | semantic_error("File always requires line number or " |
718 | "lazy pattern."); | 762 | "lazy pattern.\n"); |
719 | return -EINVAL; | 763 | return -EINVAL; |
720 | } | 764 | } |
721 | 765 | ||
722 | if (pp->offset && !pp->function) { | 766 | if (pp->offset && !pp->function) { |
723 | semantic_error("Offset requires an entry function."); | 767 | semantic_error("Offset requires an entry function.\n"); |
724 | return -EINVAL; | 768 | return -EINVAL; |
725 | } | 769 | } |
726 | 770 | ||
727 | if (pp->retprobe && !pp->function) { | 771 | if (pp->retprobe && !pp->function) { |
728 | semantic_error("Return probe requires an entry function."); | 772 | semantic_error("Return probe requires an entry function.\n"); |
729 | return -EINVAL; | 773 | return -EINVAL; |
730 | } | 774 | } |
731 | 775 | ||
732 | if ((pp->offset || pp->line || pp->lazy_line) && pp->retprobe) { | 776 | if ((pp->offset || pp->line || pp->lazy_line) && pp->retprobe) { |
733 | semantic_error("Offset/Line/Lazy pattern can't be used with " | 777 | semantic_error("Offset/Line/Lazy pattern can't be used with " |
734 | "return probe."); | 778 | "return probe.\n"); |
735 | return -EINVAL; | 779 | return -EINVAL; |
736 | } | 780 | } |
737 | 781 | ||
@@ -1005,7 +1049,7 @@ int synthesize_perf_probe_arg(struct perf_probe_arg *pa, char *buf, size_t len) | |||
1005 | 1049 | ||
1006 | return tmp - buf; | 1050 | return tmp - buf; |
1007 | error: | 1051 | error: |
1008 | pr_debug("Failed to synthesize perf probe argument: %s", | 1052 | pr_debug("Failed to synthesize perf probe argument: %s\n", |
1009 | strerror(-ret)); | 1053 | strerror(-ret)); |
1010 | return ret; | 1054 | return ret; |
1011 | } | 1055 | } |
@@ -1033,13 +1077,13 @@ static char *synthesize_perf_probe_point(struct perf_probe_point *pp) | |||
1033 | goto error; | 1077 | goto error; |
1034 | } | 1078 | } |
1035 | if (pp->file) { | 1079 | if (pp->file) { |
1036 | len = strlen(pp->file) - 31; | 1080 | tmp = pp->file; |
1037 | if (len < 0) | 1081 | len = strlen(tmp); |
1038 | len = 0; | 1082 | if (len > 30) { |
1039 | tmp = strchr(pp->file + len, '/'); | 1083 | tmp = strchr(pp->file + len - 30, '/'); |
1040 | if (!tmp) | 1084 | tmp = tmp ? tmp + 1 : pp->file + len - 30; |
1041 | tmp = pp->file + len; | 1085 | } |
1042 | ret = e_snprintf(file, 32, "@%s", tmp + 1); | 1086 | ret = e_snprintf(file, 32, "@%s", tmp); |
1043 | if (ret <= 0) | 1087 | if (ret <= 0) |
1044 | goto error; | 1088 | goto error; |
1045 | } | 1089 | } |
@@ -1055,7 +1099,7 @@ static char *synthesize_perf_probe_point(struct perf_probe_point *pp) | |||
1055 | 1099 | ||
1056 | return buf; | 1100 | return buf; |
1057 | error: | 1101 | error: |
1058 | pr_debug("Failed to synthesize perf probe point: %s", | 1102 | pr_debug("Failed to synthesize perf probe point: %s\n", |
1059 | strerror(-ret)); | 1103 | strerror(-ret)); |
1060 | if (buf) | 1104 | if (buf) |
1061 | free(buf); | 1105 | free(buf); |
@@ -1796,7 +1840,7 @@ static int del_trace_probe_event(int fd, const char *group, | |||
1796 | 1840 | ||
1797 | ret = e_snprintf(buf, 128, "%s:%s", group, event); | 1841 | ret = e_snprintf(buf, 128, "%s:%s", group, event); |
1798 | if (ret < 0) { | 1842 | if (ret < 0) { |
1799 | pr_err("Failed to copy event."); | 1843 | pr_err("Failed to copy event.\n"); |
1800 | return ret; | 1844 | return ret; |
1801 | } | 1845 | } |
1802 | 1846 | ||
diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index ddf4d455632..ab83b6ac5d6 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c | |||
@@ -652,8 +652,8 @@ static_var: | |||
652 | regs = get_arch_regstr(regn); | 652 | regs = get_arch_regstr(regn); |
653 | if (!regs) { | 653 | if (!regs) { |
654 | /* This should be a bug in DWARF or this tool */ | 654 | /* This should be a bug in DWARF or this tool */ |
655 | pr_warning("Mapping for DWARF register number %u " | 655 | pr_warning("Mapping for the register number %u " |
656 | "missing on this architecture.", regn); | 656 | "missing on this architecture.\n", regn); |
657 | return -ERANGE; | 657 | return -ERANGE; |
658 | } | 658 | } |
659 | 659 | ||
@@ -699,13 +699,14 @@ static int convert_variable_type(Dwarf_Die *vr_die, | |||
699 | if (ret != DW_TAG_pointer_type && | 699 | if (ret != DW_TAG_pointer_type && |
700 | ret != DW_TAG_array_type) { | 700 | ret != DW_TAG_array_type) { |
701 | pr_warning("Failed to cast into string: " | 701 | pr_warning("Failed to cast into string: " |
702 | "%s(%s) is not a pointer nor array.", | 702 | "%s(%s) is not a pointer nor array.\n", |
703 | dwarf_diename(vr_die), dwarf_diename(&type)); | 703 | dwarf_diename(vr_die), dwarf_diename(&type)); |
704 | return -EINVAL; | 704 | return -EINVAL; |
705 | } | 705 | } |
706 | if (ret == DW_TAG_pointer_type) { | 706 | if (ret == DW_TAG_pointer_type) { |
707 | if (die_get_real_type(&type, &type) == NULL) { | 707 | if (die_get_real_type(&type, &type) == NULL) { |
708 | pr_warning("Failed to get a type information."); | 708 | pr_warning("Failed to get a type" |
709 | " information.\n"); | ||
709 | return -ENOENT; | 710 | return -ENOENT; |
710 | } | 711 | } |
711 | while (*ref_ptr) | 712 | while (*ref_ptr) |
@@ -720,7 +721,7 @@ static int convert_variable_type(Dwarf_Die *vr_die, | |||
720 | if (!die_compare_name(&type, "char") && | 721 | if (!die_compare_name(&type, "char") && |
721 | !die_compare_name(&type, "unsigned char")) { | 722 | !die_compare_name(&type, "unsigned char")) { |
722 | pr_warning("Failed to cast into string: " | 723 | pr_warning("Failed to cast into string: " |
723 | "%s is not (unsigned) char *.", | 724 | "%s is not (unsigned) char *.\n", |
724 | dwarf_diename(vr_die)); | 725 | dwarf_diename(vr_die)); |
725 | return -EINVAL; | 726 | return -EINVAL; |
726 | } | 727 | } |
@@ -830,8 +831,8 @@ static int convert_variable_fields(Dwarf_Die *vr_die, const char *varname, | |||
830 | return -EINVAL; | 831 | return -EINVAL; |
831 | } | 832 | } |
832 | if (field->name[0] == '[') { | 833 | if (field->name[0] == '[') { |
833 | pr_err("Semantic error: %s is not a pointor nor array.", | 834 | pr_err("Semantic error: %s is not a pointor" |
834 | varname); | 835 | " nor array.\n", varname); |
835 | return -EINVAL; | 836 | return -EINVAL; |
836 | } | 837 | } |
837 | if (field->ref) { | 838 | if (field->ref) { |
@@ -978,7 +979,7 @@ static int convert_to_trace_point(Dwarf_Die *sp_die, Dwarf_Addr paddr, | |||
978 | name = dwarf_diename(sp_die); | 979 | name = dwarf_diename(sp_die); |
979 | if (name) { | 980 | if (name) { |
980 | if (dwarf_entrypc(sp_die, &eaddr) != 0) { | 981 | if (dwarf_entrypc(sp_die, &eaddr) != 0) { |
981 | pr_warning("Failed to get entry pc of %s\n", | 982 | pr_warning("Failed to get entry address of %s\n", |
982 | dwarf_diename(sp_die)); | 983 | dwarf_diename(sp_die)); |
983 | return -ENOENT; | 984 | return -ENOENT; |
984 | } | 985 | } |
@@ -994,7 +995,7 @@ static int convert_to_trace_point(Dwarf_Die *sp_die, Dwarf_Addr paddr, | |||
994 | if (retprobe) { | 995 | if (retprobe) { |
995 | if (eaddr != paddr) { | 996 | if (eaddr != paddr) { |
996 | pr_warning("Return probe must be on the head of" | 997 | pr_warning("Return probe must be on the head of" |
997 | " a real function\n"); | 998 | " a real function.\n"); |
998 | return -EINVAL; | 999 | return -EINVAL; |
999 | } | 1000 | } |
1000 | tp->retprobe = true; | 1001 | tp->retprobe = true; |
@@ -1033,7 +1034,7 @@ static int call_probe_finder(Dwarf_Die *sp_die, struct probe_finder *pf) | |||
1033 | Dwarf_Frame *frame; | 1034 | Dwarf_Frame *frame; |
1034 | if (dwarf_cfi_addrframe(pf->cfi, pf->addr, &frame) != 0 || | 1035 | if (dwarf_cfi_addrframe(pf->cfi, pf->addr, &frame) != 0 || |
1035 | dwarf_frame_cfa(frame, &pf->fb_ops, &nops) != 0) { | 1036 | dwarf_frame_cfa(frame, &pf->fb_ops, &nops) != 0) { |
1036 | pr_warning("Failed to get CFA on 0x%jx\n", | 1037 | pr_warning("Failed to get call frame on 0x%jx\n", |
1037 | (uintmax_t)pf->addr); | 1038 | (uintmax_t)pf->addr); |
1038 | return -ENOENT; | 1039 | return -ENOENT; |
1039 | } | 1040 | } |
@@ -1060,7 +1061,7 @@ static int find_probe_point_by_line(struct probe_finder *pf) | |||
1060 | int ret = 0; | 1061 | int ret = 0; |
1061 | 1062 | ||
1062 | if (dwarf_getsrclines(&pf->cu_die, &lines, &nlines) != 0) { | 1063 | if (dwarf_getsrclines(&pf->cu_die, &lines, &nlines) != 0) { |
1063 | pr_warning("No source lines found in this CU.\n"); | 1064 | pr_warning("No source lines found.\n"); |
1064 | return -ENOENT; | 1065 | return -ENOENT; |
1065 | } | 1066 | } |
1066 | 1067 | ||
@@ -1162,7 +1163,7 @@ static int find_probe_point_lazy(Dwarf_Die *sp_die, struct probe_finder *pf) | |||
1162 | } | 1163 | } |
1163 | 1164 | ||
1164 | if (dwarf_getsrclines(&pf->cu_die, &lines, &nlines) != 0) { | 1165 | if (dwarf_getsrclines(&pf->cu_die, &lines, &nlines) != 0) { |
1165 | pr_warning("No source lines found in this CU.\n"); | 1166 | pr_warning("No source lines found.\n"); |
1166 | return -ENOENT; | 1167 | return -ENOENT; |
1167 | } | 1168 | } |
1168 | 1169 | ||
@@ -1220,7 +1221,7 @@ static int probe_point_inline_cb(Dwarf_Die *in_die, void *data) | |||
1220 | else { | 1221 | else { |
1221 | /* Get probe address */ | 1222 | /* Get probe address */ |
1222 | if (dwarf_entrypc(in_die, &addr) != 0) { | 1223 | if (dwarf_entrypc(in_die, &addr) != 0) { |
1223 | pr_warning("Failed to get entry pc of %s.\n", | 1224 | pr_warning("Failed to get entry address of %s.\n", |
1224 | dwarf_diename(in_die)); | 1225 | dwarf_diename(in_die)); |
1225 | param->retval = -ENOENT; | 1226 | param->retval = -ENOENT; |
1226 | return DWARF_CB_ABORT; | 1227 | return DWARF_CB_ABORT; |
@@ -1261,8 +1262,8 @@ static int probe_point_search_cb(Dwarf_Die *sp_die, void *data) | |||
1261 | param->retval = find_probe_point_lazy(sp_die, pf); | 1262 | param->retval = find_probe_point_lazy(sp_die, pf); |
1262 | else { | 1263 | else { |
1263 | if (dwarf_entrypc(sp_die, &pf->addr) != 0) { | 1264 | if (dwarf_entrypc(sp_die, &pf->addr) != 0) { |
1264 | pr_warning("Failed to get entry pc of %s.\n", | 1265 | pr_warning("Failed to get entry address of " |
1265 | dwarf_diename(sp_die)); | 1266 | "%s.\n", dwarf_diename(sp_die)); |
1266 | param->retval = -ENOENT; | 1267 | param->retval = -ENOENT; |
1267 | return DWARF_CB_ABORT; | 1268 | return DWARF_CB_ABORT; |
1268 | } | 1269 | } |
@@ -1304,7 +1305,7 @@ static int find_probes(int fd, struct probe_finder *pf) | |||
1304 | 1305 | ||
1305 | dbg = dwfl_init_offline_dwarf(fd, &dwfl, &bias); | 1306 | dbg = dwfl_init_offline_dwarf(fd, &dwfl, &bias); |
1306 | if (!dbg) { | 1307 | if (!dbg) { |
1307 | pr_warning("No dwarf info found in the vmlinux - " | 1308 | pr_warning("No debug information found in the vmlinux - " |
1308 | "please rebuild with CONFIG_DEBUG_INFO=y.\n"); | 1309 | "please rebuild with CONFIG_DEBUG_INFO=y.\n"); |
1309 | return -EBADF; | 1310 | return -EBADF; |
1310 | } | 1311 | } |
@@ -1549,7 +1550,7 @@ int find_perf_probe_point(unsigned long addr, struct perf_probe_point *ppt) | |||
1549 | /* Open the live linux kernel */ | 1550 | /* Open the live linux kernel */ |
1550 | dbg = dwfl_init_live_kernel_dwarf(addr, &dwfl, &bias); | 1551 | dbg = dwfl_init_live_kernel_dwarf(addr, &dwfl, &bias); |
1551 | if (!dbg) { | 1552 | if (!dbg) { |
1552 | pr_warning("No dwarf info found in the vmlinux - " | 1553 | pr_warning("No debug information found in the vmlinux - " |
1553 | "please rebuild with CONFIG_DEBUG_INFO=y.\n"); | 1554 | "please rebuild with CONFIG_DEBUG_INFO=y.\n"); |
1554 | ret = -EINVAL; | 1555 | ret = -EINVAL; |
1555 | goto end; | 1556 | goto end; |
@@ -1559,7 +1560,8 @@ int find_perf_probe_point(unsigned long addr, struct perf_probe_point *ppt) | |||
1559 | addr += bias; | 1560 | addr += bias; |
1560 | /* Find cu die */ | 1561 | /* Find cu die */ |
1561 | if (!dwarf_addrdie(dbg, (Dwarf_Addr)addr - bias, &cudie)) { | 1562 | if (!dwarf_addrdie(dbg, (Dwarf_Addr)addr - bias, &cudie)) { |
1562 | pr_warning("No CU DIE is found at %lx\n", addr); | 1563 | pr_warning("Failed to find debug information for address %lx\n", |
1564 | addr); | ||
1563 | ret = -EINVAL; | 1565 | ret = -EINVAL; |
1564 | goto end; | 1566 | goto end; |
1565 | } | 1567 | } |
@@ -1684,7 +1686,7 @@ static int find_line_range_by_line(Dwarf_Die *sp_die, struct line_finder *lf) | |||
1684 | 1686 | ||
1685 | line_list__init(&lf->lr->line_list); | 1687 | line_list__init(&lf->lr->line_list); |
1686 | if (dwarf_getsrclines(&lf->cu_die, &lines, &nlines) != 0) { | 1688 | if (dwarf_getsrclines(&lf->cu_die, &lines, &nlines) != 0) { |
1687 | pr_warning("No source lines found in this CU.\n"); | 1689 | pr_warning("No source lines found.\n"); |
1688 | return -ENOENT; | 1690 | return -ENOENT; |
1689 | } | 1691 | } |
1690 | 1692 | ||
@@ -1809,7 +1811,7 @@ int find_line_range(int fd, struct line_range *lr) | |||
1809 | 1811 | ||
1810 | dbg = dwfl_init_offline_dwarf(fd, &dwfl, &bias); | 1812 | dbg = dwfl_init_offline_dwarf(fd, &dwfl, &bias); |
1811 | if (!dbg) { | 1813 | if (!dbg) { |
1812 | pr_warning("No dwarf info found in the vmlinux - " | 1814 | pr_warning("No debug information found in the vmlinux - " |
1813 | "please rebuild with CONFIG_DEBUG_INFO=y.\n"); | 1815 | "please rebuild with CONFIG_DEBUG_INFO=y.\n"); |
1814 | return -EBADF; | 1816 | return -EBADF; |
1815 | } | 1817 | } |
diff --git a/tools/perf/util/probe-finder.h b/tools/perf/util/probe-finder.h index bba69d45569..beaefc3c122 100644 --- a/tools/perf/util/probe-finder.h +++ b/tools/perf/util/probe-finder.h | |||
@@ -34,9 +34,9 @@ extern int find_available_vars_at(int fd, struct perf_probe_event *pev, | |||
34 | bool externs); | 34 | bool externs); |
35 | 35 | ||
36 | #include <dwarf.h> | 36 | #include <dwarf.h> |
37 | #include <libdw.h> | 37 | #include <elfutils/libdw.h> |
38 | #include <libdwfl.h> | 38 | #include <elfutils/libdwfl.h> |
39 | #include <version.h> | 39 | #include <elfutils/version.h> |
40 | 40 | ||
41 | struct probe_finder { | 41 | struct probe_finder { |
42 | struct perf_probe_event *pev; /* Target probe event */ | 42 | struct perf_probe_event *pev; /* Target probe event */ |
diff --git a/tools/perf/util/scripting-engines/trace-event-perl.c b/tools/perf/util/scripting-engines/trace-event-perl.c index b059dc50cc2..93680818e24 100644 --- a/tools/perf/util/scripting-engines/trace-event-perl.c +++ b/tools/perf/util/scripting-engines/trace-event-perl.c | |||
@@ -1,5 +1,5 @@ | |||
1 | /* | 1 | /* |
2 | * trace-event-perl. Feed perf trace events to an embedded Perl interpreter. | 2 | * trace-event-perl. Feed perf script events to an embedded Perl interpreter. |
3 | * | 3 | * |
4 | * Copyright (C) 2009 Tom Zanussi <tzanussi@gmail.com> | 4 | * Copyright (C) 2009 Tom Zanussi <tzanussi@gmail.com> |
5 | * | 5 | * |
@@ -411,8 +411,8 @@ static int perl_generate_script(const char *outfile) | |||
411 | return -1; | 411 | return -1; |
412 | } | 412 | } |
413 | 413 | ||
414 | fprintf(ofp, "# perf trace event handlers, " | 414 | fprintf(ofp, "# perf script event handlers, " |
415 | "generated by perf trace -g perl\n"); | 415 | "generated by perf script -g perl\n"); |
416 | 416 | ||
417 | fprintf(ofp, "# Licensed under the terms of the GNU GPL" | 417 | fprintf(ofp, "# Licensed under the terms of the GNU GPL" |
418 | " License version 2\n\n"); | 418 | " License version 2\n\n"); |
diff --git a/tools/perf/util/scripting-engines/trace-event-python.c b/tools/perf/util/scripting-engines/trace-event-python.c index 33a63252374..c6d99334bdf 100644 --- a/tools/perf/util/scripting-engines/trace-event-python.c +++ b/tools/perf/util/scripting-engines/trace-event-python.c | |||
@@ -442,8 +442,8 @@ static int python_generate_script(const char *outfile) | |||
442 | fprintf(stderr, "couldn't open %s\n", fname); | 442 | fprintf(stderr, "couldn't open %s\n", fname); |
443 | return -1; | 443 | return -1; |
444 | } | 444 | } |
445 | fprintf(ofp, "# perf trace event handlers, " | 445 | fprintf(ofp, "# perf script event handlers, " |
446 | "generated by perf trace -g python\n"); | 446 | "generated by perf script -g python\n"); |
447 | 447 | ||
448 | fprintf(ofp, "# Licensed under the terms of the GNU GPL" | 448 | fprintf(ofp, "# Licensed under the terms of the GNU GPL" |
449 | " License version 2\n\n"); | 449 | " License version 2\n\n"); |
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index fa9d652c2dc..6fb4694d05f 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c | |||
@@ -65,9 +65,49 @@ out_close: | |||
65 | return -1; | 65 | return -1; |
66 | } | 66 | } |
67 | 67 | ||
68 | static void perf_session__id_header_size(struct perf_session *session) | ||
69 | { | ||
70 | struct sample_data *data; | ||
71 | u64 sample_type = session->sample_type; | ||
72 | u16 size = 0; | ||
73 | |||
74 | if (!session->sample_id_all) | ||
75 | goto out; | ||
76 | |||
77 | if (sample_type & PERF_SAMPLE_TID) | ||
78 | size += sizeof(data->tid) * 2; | ||
79 | |||
80 | if (sample_type & PERF_SAMPLE_TIME) | ||
81 | size += sizeof(data->time); | ||
82 | |||
83 | if (sample_type & PERF_SAMPLE_ID) | ||
84 | size += sizeof(data->id); | ||
85 | |||
86 | if (sample_type & PERF_SAMPLE_STREAM_ID) | ||
87 | size += sizeof(data->stream_id); | ||
88 | |||
89 | if (sample_type & PERF_SAMPLE_CPU) | ||
90 | size += sizeof(data->cpu) * 2; | ||
91 | out: | ||
92 | session->id_hdr_size = size; | ||
93 | } | ||
94 | |||
95 | void perf_session__set_sample_id_all(struct perf_session *session, bool value) | ||
96 | { | ||
97 | session->sample_id_all = value; | ||
98 | perf_session__id_header_size(session); | ||
99 | } | ||
100 | |||
101 | void perf_session__set_sample_type(struct perf_session *session, u64 type) | ||
102 | { | ||
103 | session->sample_type = type; | ||
104 | } | ||
105 | |||
68 | void perf_session__update_sample_type(struct perf_session *self) | 106 | void perf_session__update_sample_type(struct perf_session *self) |
69 | { | 107 | { |
70 | self->sample_type = perf_header__sample_type(&self->header); | 108 | self->sample_type = perf_header__sample_type(&self->header); |
109 | self->sample_id_all = perf_header__sample_id_all(&self->header); | ||
110 | perf_session__id_header_size(self); | ||
71 | } | 111 | } |
72 | 112 | ||
73 | int perf_session__create_kernel_maps(struct perf_session *self) | 113 | int perf_session__create_kernel_maps(struct perf_session *self) |
@@ -85,7 +125,9 @@ static void perf_session__destroy_kernel_maps(struct perf_session *self) | |||
85 | machines__destroy_guest_kernel_maps(&self->machines); | 125 | machines__destroy_guest_kernel_maps(&self->machines); |
86 | } | 126 | } |
87 | 127 | ||
88 | struct perf_session *perf_session__new(const char *filename, int mode, bool force, bool repipe) | 128 | struct perf_session *perf_session__new(const char *filename, int mode, |
129 | bool force, bool repipe, | ||
130 | struct perf_event_ops *ops) | ||
89 | { | 131 | { |
90 | size_t len = filename ? strlen(filename) + 1 : 0; | 132 | size_t len = filename ? strlen(filename) + 1 : 0; |
91 | struct perf_session *self = zalloc(sizeof(*self) + len); | 133 | struct perf_session *self = zalloc(sizeof(*self) + len); |
@@ -101,10 +143,20 @@ struct perf_session *perf_session__new(const char *filename, int mode, bool forc | |||
101 | INIT_LIST_HEAD(&self->dead_threads); | 143 | INIT_LIST_HEAD(&self->dead_threads); |
102 | self->hists_tree = RB_ROOT; | 144 | self->hists_tree = RB_ROOT; |
103 | self->last_match = NULL; | 145 | self->last_match = NULL; |
104 | self->mmap_window = 32; | 146 | /* |
147 | * On 64bit we can mmap the data file in one go. No need for tiny mmap | ||
148 | * slices. On 32bit we use 32MB. | ||
149 | */ | ||
150 | #if BITS_PER_LONG == 64 | ||
151 | self->mmap_window = ULLONG_MAX; | ||
152 | #else | ||
153 | self->mmap_window = 32 * 1024 * 1024ULL; | ||
154 | #endif | ||
105 | self->machines = RB_ROOT; | 155 | self->machines = RB_ROOT; |
106 | self->repipe = repipe; | 156 | self->repipe = repipe; |
107 | INIT_LIST_HEAD(&self->ordered_samples.samples_head); | 157 | INIT_LIST_HEAD(&self->ordered_samples.samples); |
158 | INIT_LIST_HEAD(&self->ordered_samples.sample_cache); | ||
159 | INIT_LIST_HEAD(&self->ordered_samples.to_free); | ||
108 | machine__init(&self->host_machine, "", HOST_KERNEL_ID); | 160 | machine__init(&self->host_machine, "", HOST_KERNEL_ID); |
109 | 161 | ||
110 | if (mode == O_RDONLY) { | 162 | if (mode == O_RDONLY) { |
@@ -120,6 +172,13 @@ struct perf_session *perf_session__new(const char *filename, int mode, bool forc | |||
120 | } | 172 | } |
121 | 173 | ||
122 | perf_session__update_sample_type(self); | 174 | perf_session__update_sample_type(self); |
175 | |||
176 | if (ops && ops->ordering_requires_timestamps && | ||
177 | ops->ordered_samples && !self->sample_id_all) { | ||
178 | dump_printf("WARNING: No sample_id_all support, falling back to unordered processing\n"); | ||
179 | ops->ordered_samples = false; | ||
180 | } | ||
181 | |||
123 | out: | 182 | out: |
124 | return self; | 183 | return self; |
125 | out_free: | 184 | out_free: |
@@ -230,7 +289,15 @@ struct map_symbol *perf_session__resolve_callchain(struct perf_session *self, | |||
230 | return syms; | 289 | return syms; |
231 | } | 290 | } |
232 | 291 | ||
292 | static int process_event_synth_stub(event_t *event __used, | ||
293 | struct perf_session *session __used) | ||
294 | { | ||
295 | dump_printf(": unhandled!\n"); | ||
296 | return 0; | ||
297 | } | ||
298 | |||
233 | static int process_event_stub(event_t *event __used, | 299 | static int process_event_stub(event_t *event __used, |
300 | struct sample_data *sample __used, | ||
234 | struct perf_session *session __used) | 301 | struct perf_session *session __used) |
235 | { | 302 | { |
236 | dump_printf(": unhandled!\n"); | 303 | dump_printf(": unhandled!\n"); |
@@ -262,7 +329,7 @@ static void perf_event_ops__fill_defaults(struct perf_event_ops *handler) | |||
262 | if (handler->exit == NULL) | 329 | if (handler->exit == NULL) |
263 | handler->exit = process_event_stub; | 330 | handler->exit = process_event_stub; |
264 | if (handler->lost == NULL) | 331 | if (handler->lost == NULL) |
265 | handler->lost = process_event_stub; | 332 | handler->lost = event__process_lost; |
266 | if (handler->read == NULL) | 333 | if (handler->read == NULL) |
267 | handler->read = process_event_stub; | 334 | handler->read = process_event_stub; |
268 | if (handler->throttle == NULL) | 335 | if (handler->throttle == NULL) |
@@ -270,13 +337,13 @@ static void perf_event_ops__fill_defaults(struct perf_event_ops *handler) | |||
270 | if (handler->unthrottle == NULL) | 337 | if (handler->unthrottle == NULL) |
271 | handler->unthrottle = process_event_stub; | 338 | handler->unthrottle = process_event_stub; |
272 | if (handler->attr == NULL) | 339 | if (handler->attr == NULL) |
273 | handler->attr = process_event_stub; | 340 | handler->attr = process_event_synth_stub; |
274 | if (handler->event_type == NULL) | 341 | if (handler->event_type == NULL) |
275 | handler->event_type = process_event_stub; | 342 | handler->event_type = process_event_synth_stub; |
276 | if (handler->tracing_data == NULL) | 343 | if (handler->tracing_data == NULL) |
277 | handler->tracing_data = process_event_stub; | 344 | handler->tracing_data = process_event_synth_stub; |
278 | if (handler->build_id == NULL) | 345 | if (handler->build_id == NULL) |
279 | handler->build_id = process_event_stub; | 346 | handler->build_id = process_event_synth_stub; |
280 | if (handler->finished_round == NULL) { | 347 | if (handler->finished_round == NULL) { |
281 | if (handler->ordered_samples) | 348 | if (handler->ordered_samples) |
282 | handler->finished_round = process_finished_round; | 349 | handler->finished_round = process_finished_round; |
@@ -386,33 +453,61 @@ static event__swap_op event__swap_ops[] = { | |||
386 | 453 | ||
387 | struct sample_queue { | 454 | struct sample_queue { |
388 | u64 timestamp; | 455 | u64 timestamp; |
389 | struct sample_event *event; | 456 | u64 file_offset; |
457 | event_t *event; | ||
390 | struct list_head list; | 458 | struct list_head list; |
391 | }; | 459 | }; |
392 | 460 | ||
461 | static void perf_session_free_sample_buffers(struct perf_session *session) | ||
462 | { | ||
463 | struct ordered_samples *os = &session->ordered_samples; | ||
464 | |||
465 | while (!list_empty(&os->to_free)) { | ||
466 | struct sample_queue *sq; | ||
467 | |||
468 | sq = list_entry(os->to_free.next, struct sample_queue, list); | ||
469 | list_del(&sq->list); | ||
470 | free(sq); | ||
471 | } | ||
472 | } | ||
473 | |||
474 | static int perf_session_deliver_event(struct perf_session *session, | ||
475 | event_t *event, | ||
476 | struct sample_data *sample, | ||
477 | struct perf_event_ops *ops, | ||
478 | u64 file_offset); | ||
479 | |||
393 | static void flush_sample_queue(struct perf_session *s, | 480 | static void flush_sample_queue(struct perf_session *s, |
394 | struct perf_event_ops *ops) | 481 | struct perf_event_ops *ops) |
395 | { | 482 | { |
396 | struct list_head *head = &s->ordered_samples.samples_head; | 483 | struct ordered_samples *os = &s->ordered_samples; |
397 | u64 limit = s->ordered_samples.next_flush; | 484 | struct list_head *head = &os->samples; |
398 | struct sample_queue *tmp, *iter; | 485 | struct sample_queue *tmp, *iter; |
486 | struct sample_data sample; | ||
487 | u64 limit = os->next_flush; | ||
488 | u64 last_ts = os->last_sample ? os->last_sample->timestamp : 0ULL; | ||
399 | 489 | ||
400 | if (!ops->ordered_samples || !limit) | 490 | if (!ops->ordered_samples || !limit) |
401 | return; | 491 | return; |
402 | 492 | ||
403 | list_for_each_entry_safe(iter, tmp, head, list) { | 493 | list_for_each_entry_safe(iter, tmp, head, list) { |
404 | if (iter->timestamp > limit) | 494 | if (iter->timestamp > limit) |
405 | return; | 495 | break; |
406 | 496 | ||
407 | if (iter == s->ordered_samples.last_inserted) | 497 | event__parse_sample(iter->event, s, &sample); |
408 | s->ordered_samples.last_inserted = NULL; | 498 | perf_session_deliver_event(s, iter->event, &sample, ops, |
499 | iter->file_offset); | ||
409 | 500 | ||
410 | ops->sample((event_t *)iter->event, s); | 501 | os->last_flush = iter->timestamp; |
411 | |||
412 | s->ordered_samples.last_flush = iter->timestamp; | ||
413 | list_del(&iter->list); | 502 | list_del(&iter->list); |
414 | free(iter->event); | 503 | list_add(&iter->list, &os->sample_cache); |
415 | free(iter); | 504 | } |
505 | |||
506 | if (list_empty(head)) { | ||
507 | os->last_sample = NULL; | ||
508 | } else if (last_ts <= limit) { | ||
509 | os->last_sample = | ||
510 | list_entry(head->prev, struct sample_queue, list); | ||
416 | } | 511 | } |
417 | } | 512 | } |
418 | 513 | ||
@@ -465,178 +560,265 @@ static int process_finished_round(event_t *event __used, | |||
465 | return 0; | 560 | return 0; |
466 | } | 561 | } |
467 | 562 | ||
468 | static void __queue_sample_end(struct sample_queue *new, struct list_head *head) | ||
469 | { | ||
470 | struct sample_queue *iter; | ||
471 | |||
472 | list_for_each_entry_reverse(iter, head, list) { | ||
473 | if (iter->timestamp < new->timestamp) { | ||
474 | list_add(&new->list, &iter->list); | ||
475 | return; | ||
476 | } | ||
477 | } | ||
478 | |||
479 | list_add(&new->list, head); | ||
480 | } | ||
481 | |||
482 | static void __queue_sample_before(struct sample_queue *new, | ||
483 | struct sample_queue *iter, | ||
484 | struct list_head *head) | ||
485 | { | ||
486 | list_for_each_entry_continue_reverse(iter, head, list) { | ||
487 | if (iter->timestamp < new->timestamp) { | ||
488 | list_add(&new->list, &iter->list); | ||
489 | return; | ||
490 | } | ||
491 | } | ||
492 | |||
493 | list_add(&new->list, head); | ||
494 | } | ||
495 | |||
496 | static void __queue_sample_after(struct sample_queue *new, | ||
497 | struct sample_queue *iter, | ||
498 | struct list_head *head) | ||
499 | { | ||
500 | list_for_each_entry_continue(iter, head, list) { | ||
501 | if (iter->timestamp > new->timestamp) { | ||
502 | list_add_tail(&new->list, &iter->list); | ||
503 | return; | ||
504 | } | ||
505 | } | ||
506 | list_add_tail(&new->list, head); | ||
507 | } | ||
508 | |||
509 | /* The queue is ordered by time */ | 563 | /* The queue is ordered by time */ |
510 | static void __queue_sample_event(struct sample_queue *new, | 564 | static void __queue_event(struct sample_queue *new, struct perf_session *s) |
511 | struct perf_session *s) | ||
512 | { | 565 | { |
513 | struct sample_queue *last_inserted = s->ordered_samples.last_inserted; | 566 | struct ordered_samples *os = &s->ordered_samples; |
514 | struct list_head *head = &s->ordered_samples.samples_head; | 567 | struct sample_queue *sample = os->last_sample; |
568 | u64 timestamp = new->timestamp; | ||
569 | struct list_head *p; | ||
515 | 570 | ||
571 | os->last_sample = new; | ||
516 | 572 | ||
517 | if (!last_inserted) { | 573 | if (!sample) { |
518 | __queue_sample_end(new, head); | 574 | list_add(&new->list, &os->samples); |
575 | os->max_timestamp = timestamp; | ||
519 | return; | 576 | return; |
520 | } | 577 | } |
521 | 578 | ||
522 | /* | 579 | /* |
523 | * Most of the time the current event has a timestamp | 580 | * last_sample might point to some random place in the list as it's |
524 | * very close to the last event inserted, unless we just switched | 581 | * the last queued event. We expect that the new event is close to |
525 | * to another event buffer. Having a sorting based on a list and | 582 | * this. |
526 | * on the last inserted event that is close to the current one is | ||
527 | * probably more efficient than an rbtree based sorting. | ||
528 | */ | 583 | */ |
529 | if (last_inserted->timestamp >= new->timestamp) | 584 | if (sample->timestamp <= timestamp) { |
530 | __queue_sample_before(new, last_inserted, head); | 585 | while (sample->timestamp <= timestamp) { |
531 | else | 586 | p = sample->list.next; |
532 | __queue_sample_after(new, last_inserted, head); | 587 | if (p == &os->samples) { |
588 | list_add_tail(&new->list, &os->samples); | ||
589 | os->max_timestamp = timestamp; | ||
590 | return; | ||
591 | } | ||
592 | sample = list_entry(p, struct sample_queue, list); | ||
593 | } | ||
594 | list_add_tail(&new->list, &sample->list); | ||
595 | } else { | ||
596 | while (sample->timestamp > timestamp) { | ||
597 | p = sample->list.prev; | ||
598 | if (p == &os->samples) { | ||
599 | list_add(&new->list, &os->samples); | ||
600 | return; | ||
601 | } | ||
602 | sample = list_entry(p, struct sample_queue, list); | ||
603 | } | ||
604 | list_add(&new->list, &sample->list); | ||
605 | } | ||
533 | } | 606 | } |
534 | 607 | ||
535 | static int queue_sample_event(event_t *event, struct sample_data *data, | 608 | #define MAX_SAMPLE_BUFFER (64 * 1024 / sizeof(struct sample_queue)) |
536 | struct perf_session *s) | 609 | |
610 | static int perf_session_queue_event(struct perf_session *s, event_t *event, | ||
611 | struct sample_data *data, u64 file_offset) | ||
537 | { | 612 | { |
613 | struct ordered_samples *os = &s->ordered_samples; | ||
614 | struct list_head *sc = &os->sample_cache; | ||
538 | u64 timestamp = data->time; | 615 | u64 timestamp = data->time; |
539 | struct sample_queue *new; | 616 | struct sample_queue *new; |
540 | 617 | ||
618 | if (!timestamp || timestamp == ~0ULL) | ||
619 | return -ETIME; | ||
541 | 620 | ||
542 | if (timestamp < s->ordered_samples.last_flush) { | 621 | if (timestamp < s->ordered_samples.last_flush) { |
543 | printf("Warning: Timestamp below last timeslice flush\n"); | 622 | printf("Warning: Timestamp below last timeslice flush\n"); |
544 | return -EINVAL; | 623 | return -EINVAL; |
545 | } | 624 | } |
546 | 625 | ||
547 | new = malloc(sizeof(*new)); | 626 | if (!list_empty(sc)) { |
548 | if (!new) | 627 | new = list_entry(sc->next, struct sample_queue, list); |
549 | return -ENOMEM; | 628 | list_del(&new->list); |
629 | } else if (os->sample_buffer) { | ||
630 | new = os->sample_buffer + os->sample_buffer_idx; | ||
631 | if (++os->sample_buffer_idx == MAX_SAMPLE_BUFFER) | ||
632 | os->sample_buffer = NULL; | ||
633 | } else { | ||
634 | os->sample_buffer = malloc(MAX_SAMPLE_BUFFER * sizeof(*new)); | ||
635 | if (!os->sample_buffer) | ||
636 | return -ENOMEM; | ||
637 | list_add(&os->sample_buffer->list, &os->to_free); | ||
638 | os->sample_buffer_idx = 2; | ||
639 | new = os->sample_buffer + 1; | ||
640 | } | ||
550 | 641 | ||
551 | new->timestamp = timestamp; | 642 | new->timestamp = timestamp; |
643 | new->file_offset = file_offset; | ||
644 | new->event = event; | ||
552 | 645 | ||
553 | new->event = malloc(event->header.size); | 646 | __queue_event(new, s); |
554 | if (!new->event) { | ||
555 | free(new); | ||
556 | return -ENOMEM; | ||
557 | } | ||
558 | 647 | ||
559 | memcpy(new->event, event, event->header.size); | 648 | return 0; |
649 | } | ||
560 | 650 | ||
561 | __queue_sample_event(new, s); | 651 | static void callchain__printf(struct sample_data *sample) |
562 | s->ordered_samples.last_inserted = new; | 652 | { |
653 | unsigned int i; | ||
563 | 654 | ||
564 | if (new->timestamp > s->ordered_samples.max_timestamp) | 655 | printf("... chain: nr:%Lu\n", sample->callchain->nr); |
565 | s->ordered_samples.max_timestamp = new->timestamp; | ||
566 | 656 | ||
567 | return 0; | 657 | for (i = 0; i < sample->callchain->nr; i++) |
658 | printf("..... %2d: %016Lx\n", i, sample->callchain->ips[i]); | ||
568 | } | 659 | } |
569 | 660 | ||
570 | static int perf_session__process_sample(event_t *event, struct perf_session *s, | 661 | static void perf_session__print_tstamp(struct perf_session *session, |
571 | struct perf_event_ops *ops) | 662 | event_t *event, |
663 | struct sample_data *sample) | ||
572 | { | 664 | { |
573 | struct sample_data data; | 665 | if (event->header.type != PERF_RECORD_SAMPLE && |
666 | !session->sample_id_all) { | ||
667 | fputs("-1 -1 ", stdout); | ||
668 | return; | ||
669 | } | ||
574 | 670 | ||
575 | if (!ops->ordered_samples) | 671 | if ((session->sample_type & PERF_SAMPLE_CPU)) |
576 | return ops->sample(event, s); | 672 | printf("%u ", sample->cpu); |
577 | 673 | ||
578 | bzero(&data, sizeof(struct sample_data)); | 674 | if (session->sample_type & PERF_SAMPLE_TIME) |
579 | event__parse_sample(event, s->sample_type, &data); | 675 | printf("%Lu ", sample->time); |
676 | } | ||
580 | 677 | ||
581 | queue_sample_event(event, &data, s); | 678 | static void dump_event(struct perf_session *session, event_t *event, |
679 | u64 file_offset, struct sample_data *sample) | ||
680 | { | ||
681 | if (!dump_trace) | ||
682 | return; | ||
582 | 683 | ||
583 | return 0; | 684 | printf("\n%#Lx [%#x]: event: %d\n", file_offset, event->header.size, |
685 | event->header.type); | ||
686 | |||
687 | trace_event(event); | ||
688 | |||
689 | if (sample) | ||
690 | perf_session__print_tstamp(session, event, sample); | ||
691 | |||
692 | printf("%#Lx [%#x]: PERF_RECORD_%s", file_offset, event->header.size, | ||
693 | event__get_event_name(event->header.type)); | ||
584 | } | 694 | } |
585 | 695 | ||
586 | static int perf_session__process_event(struct perf_session *self, | 696 | static void dump_sample(struct perf_session *session, event_t *event, |
587 | event_t *event, | 697 | struct sample_data *sample) |
588 | struct perf_event_ops *ops, | ||
589 | u64 offset, u64 head) | ||
590 | { | 698 | { |
591 | trace_event(event); | 699 | if (!dump_trace) |
700 | return; | ||
592 | 701 | ||
593 | if (event->header.type < PERF_RECORD_HEADER_MAX) { | 702 | printf("(IP, %d): %d/%d: %#Lx period: %Ld\n", event->header.misc, |
594 | dump_printf("%#Lx [%#x]: PERF_RECORD_%s", | 703 | sample->pid, sample->tid, sample->ip, sample->period); |
595 | offset + head, event->header.size, | ||
596 | event__name[event->header.type]); | ||
597 | hists__inc_nr_events(&self->hists, event->header.type); | ||
598 | } | ||
599 | 704 | ||
600 | if (self->header.needs_swap && event__swap_ops[event->header.type]) | 705 | if (session->sample_type & PERF_SAMPLE_CALLCHAIN) |
601 | event__swap_ops[event->header.type](event); | 706 | callchain__printf(sample); |
707 | } | ||
708 | |||
709 | static int perf_session_deliver_event(struct perf_session *session, | ||
710 | event_t *event, | ||
711 | struct sample_data *sample, | ||
712 | struct perf_event_ops *ops, | ||
713 | u64 file_offset) | ||
714 | { | ||
715 | dump_event(session, event, file_offset, sample); | ||
602 | 716 | ||
603 | switch (event->header.type) { | 717 | switch (event->header.type) { |
604 | case PERF_RECORD_SAMPLE: | 718 | case PERF_RECORD_SAMPLE: |
605 | return perf_session__process_sample(event, self, ops); | 719 | dump_sample(session, event, sample); |
720 | return ops->sample(event, sample, session); | ||
606 | case PERF_RECORD_MMAP: | 721 | case PERF_RECORD_MMAP: |
607 | return ops->mmap(event, self); | 722 | return ops->mmap(event, sample, session); |
608 | case PERF_RECORD_COMM: | 723 | case PERF_RECORD_COMM: |
609 | return ops->comm(event, self); | 724 | return ops->comm(event, sample, session); |
610 | case PERF_RECORD_FORK: | 725 | case PERF_RECORD_FORK: |
611 | return ops->fork(event, self); | 726 | return ops->fork(event, sample, session); |
612 | case PERF_RECORD_EXIT: | 727 | case PERF_RECORD_EXIT: |
613 | return ops->exit(event, self); | 728 | return ops->exit(event, sample, session); |
614 | case PERF_RECORD_LOST: | 729 | case PERF_RECORD_LOST: |
615 | return ops->lost(event, self); | 730 | return ops->lost(event, sample, session); |
616 | case PERF_RECORD_READ: | 731 | case PERF_RECORD_READ: |
617 | return ops->read(event, self); | 732 | return ops->read(event, sample, session); |
618 | case PERF_RECORD_THROTTLE: | 733 | case PERF_RECORD_THROTTLE: |
619 | return ops->throttle(event, self); | 734 | return ops->throttle(event, sample, session); |
620 | case PERF_RECORD_UNTHROTTLE: | 735 | case PERF_RECORD_UNTHROTTLE: |
621 | return ops->unthrottle(event, self); | 736 | return ops->unthrottle(event, sample, session); |
737 | default: | ||
738 | ++session->hists.stats.nr_unknown_events; | ||
739 | return -1; | ||
740 | } | ||
741 | } | ||
742 | |||
743 | static int perf_session__preprocess_sample(struct perf_session *session, | ||
744 | event_t *event, struct sample_data *sample) | ||
745 | { | ||
746 | if (event->header.type != PERF_RECORD_SAMPLE || | ||
747 | !(session->sample_type & PERF_SAMPLE_CALLCHAIN)) | ||
748 | return 0; | ||
749 | |||
750 | if (!ip_callchain__valid(sample->callchain, event)) { | ||
751 | pr_debug("call-chain problem with event, skipping it.\n"); | ||
752 | ++session->hists.stats.nr_invalid_chains; | ||
753 | session->hists.stats.total_invalid_chains += sample->period; | ||
754 | return -EINVAL; | ||
755 | } | ||
756 | return 0; | ||
757 | } | ||
758 | |||
759 | static int perf_session__process_user_event(struct perf_session *session, event_t *event, | ||
760 | struct perf_event_ops *ops, u64 file_offset) | ||
761 | { | ||
762 | dump_event(session, event, file_offset, NULL); | ||
763 | |||
764 | /* These events are processed right away */ | ||
765 | switch (event->header.type) { | ||
622 | case PERF_RECORD_HEADER_ATTR: | 766 | case PERF_RECORD_HEADER_ATTR: |
623 | return ops->attr(event, self); | 767 | return ops->attr(event, session); |
624 | case PERF_RECORD_HEADER_EVENT_TYPE: | 768 | case PERF_RECORD_HEADER_EVENT_TYPE: |
625 | return ops->event_type(event, self); | 769 | return ops->event_type(event, session); |
626 | case PERF_RECORD_HEADER_TRACING_DATA: | 770 | case PERF_RECORD_HEADER_TRACING_DATA: |
627 | /* setup for reading amidst mmap */ | 771 | /* setup for reading amidst mmap */ |
628 | lseek(self->fd, offset + head, SEEK_SET); | 772 | lseek(session->fd, file_offset, SEEK_SET); |
629 | return ops->tracing_data(event, self); | 773 | return ops->tracing_data(event, session); |
630 | case PERF_RECORD_HEADER_BUILD_ID: | 774 | case PERF_RECORD_HEADER_BUILD_ID: |
631 | return ops->build_id(event, self); | 775 | return ops->build_id(event, session); |
632 | case PERF_RECORD_FINISHED_ROUND: | 776 | case PERF_RECORD_FINISHED_ROUND: |
633 | return ops->finished_round(event, self, ops); | 777 | return ops->finished_round(event, session, ops); |
634 | default: | 778 | default: |
635 | ++self->hists.stats.nr_unknown_events; | 779 | return -EINVAL; |
636 | return -1; | ||
637 | } | 780 | } |
638 | } | 781 | } |
639 | 782 | ||
783 | static int perf_session__process_event(struct perf_session *session, | ||
784 | event_t *event, | ||
785 | struct perf_event_ops *ops, | ||
786 | u64 file_offset) | ||
787 | { | ||
788 | struct sample_data sample; | ||
789 | int ret; | ||
790 | |||
791 | if (session->header.needs_swap && event__swap_ops[event->header.type]) | ||
792 | event__swap_ops[event->header.type](event); | ||
793 | |||
794 | if (event->header.type >= PERF_RECORD_HEADER_MAX) | ||
795 | return -EINVAL; | ||
796 | |||
797 | hists__inc_nr_events(&session->hists, event->header.type); | ||
798 | |||
799 | if (event->header.type >= PERF_RECORD_USER_TYPE_START) | ||
800 | return perf_session__process_user_event(session, event, ops, file_offset); | ||
801 | |||
802 | /* | ||
803 | * For all kernel events we get the sample data | ||
804 | */ | ||
805 | event__parse_sample(event, session, &sample); | ||
806 | |||
807 | /* Preprocess sample records - precheck callchains */ | ||
808 | if (perf_session__preprocess_sample(session, event, &sample)) | ||
809 | return 0; | ||
810 | |||
811 | if (ops->ordered_samples) { | ||
812 | ret = perf_session_queue_event(session, event, &sample, | ||
813 | file_offset); | ||
814 | if (ret != -ETIME) | ||
815 | return ret; | ||
816 | } | ||
817 | |||
818 | return perf_session_deliver_event(session, event, &sample, ops, | ||
819 | file_offset); | ||
820 | } | ||
821 | |||
640 | void perf_event_header__bswap(struct perf_event_header *self) | 822 | void perf_event_header__bswap(struct perf_event_header *self) |
641 | { | 823 | { |
642 | self->type = bswap_32(self->type); | 824 | self->type = bswap_32(self->type); |
@@ -656,21 +838,33 @@ static struct thread *perf_session__register_idle_thread(struct perf_session *se | |||
656 | return thread; | 838 | return thread; |
657 | } | 839 | } |
658 | 840 | ||
659 | int do_read(int fd, void *buf, size_t size) | 841 | static void perf_session__warn_about_errors(const struct perf_session *session, |
842 | const struct perf_event_ops *ops) | ||
660 | { | 843 | { |
661 | void *buf_start = buf; | 844 | if (ops->lost == event__process_lost && |
662 | 845 | session->hists.stats.total_lost != 0) { | |
663 | while (size) { | 846 | ui__warning("Processed %Lu events and LOST %Lu!\n\n" |
664 | int ret = read(fd, buf, size); | 847 | "Check IO/CPU overload!\n\n", |
665 | 848 | session->hists.stats.total_period, | |
666 | if (ret <= 0) | 849 | session->hists.stats.total_lost); |
667 | return ret; | 850 | } |
668 | 851 | ||
669 | size -= ret; | 852 | if (session->hists.stats.nr_unknown_events != 0) { |
670 | buf += ret; | 853 | ui__warning("Found %u unknown events!\n\n" |
854 | "Is this an older tool processing a perf.data " | ||
855 | "file generated by a more recent tool?\n\n" | ||
856 | "If that is not the case, consider " | ||
857 | "reporting to linux-kernel@vger.kernel.org.\n\n", | ||
858 | session->hists.stats.nr_unknown_events); | ||
671 | } | 859 | } |
672 | 860 | ||
673 | return buf - buf_start; | 861 | if (session->hists.stats.nr_invalid_chains != 0) { |
862 | ui__warning("Found invalid callchains!\n\n" | ||
863 | "%u out of %u events were discarded for this reason.\n\n" | ||
864 | "Consider reporting to linux-kernel@vger.kernel.org.\n\n", | ||
865 | session->hists.stats.nr_invalid_chains, | ||
866 | session->hists.stats.nr_events[PERF_RECORD_SAMPLE]); | ||
867 | } | ||
674 | } | 868 | } |
675 | 869 | ||
676 | #define session_done() (*(volatile int *)(&session_done)) | 870 | #define session_done() (*(volatile int *)(&session_done)) |
@@ -690,7 +884,7 @@ static int __perf_session__process_pipe_events(struct perf_session *self, | |||
690 | 884 | ||
691 | head = 0; | 885 | head = 0; |
692 | more: | 886 | more: |
693 | err = do_read(self->fd, &event, sizeof(struct perf_event_header)); | 887 | err = readn(self->fd, &event, sizeof(struct perf_event_header)); |
694 | if (err <= 0) { | 888 | if (err <= 0) { |
695 | if (err == 0) | 889 | if (err == 0) |
696 | goto done; | 890 | goto done; |
@@ -710,8 +904,7 @@ more: | |||
710 | p += sizeof(struct perf_event_header); | 904 | p += sizeof(struct perf_event_header); |
711 | 905 | ||
712 | if (size - sizeof(struct perf_event_header)) { | 906 | if (size - sizeof(struct perf_event_header)) { |
713 | err = do_read(self->fd, p, | 907 | err = readn(self->fd, p, size - sizeof(struct perf_event_header)); |
714 | size - sizeof(struct perf_event_header)); | ||
715 | if (err <= 0) { | 908 | if (err <= 0) { |
716 | if (err == 0) { | 909 | if (err == 0) { |
717 | pr_err("unexpected end of event stream\n"); | 910 | pr_err("unexpected end of event stream\n"); |
@@ -724,8 +917,7 @@ more: | |||
724 | } | 917 | } |
725 | 918 | ||
726 | if (size == 0 || | 919 | if (size == 0 || |
727 | (skip = perf_session__process_event(self, &event, ops, | 920 | (skip = perf_session__process_event(self, &event, ops, head)) < 0) { |
728 | 0, head)) < 0) { | ||
729 | dump_printf("%#Lx [%#x]: skipping unknown header type: %d\n", | 921 | dump_printf("%#Lx [%#x]: skipping unknown header type: %d\n", |
730 | head, event.header.size, event.header.type); | 922 | head, event.header.size, event.header.type); |
731 | /* | 923 | /* |
@@ -740,9 +932,6 @@ more: | |||
740 | 932 | ||
741 | head += size; | 933 | head += size; |
742 | 934 | ||
743 | dump_printf("\n%#Lx [%#x]: event: %d\n", | ||
744 | head, event.header.size, event.header.type); | ||
745 | |||
746 | if (skip > 0) | 935 | if (skip > 0) |
747 | head += skip; | 936 | head += skip; |
748 | 937 | ||
@@ -751,82 +940,91 @@ more: | |||
751 | done: | 940 | done: |
752 | err = 0; | 941 | err = 0; |
753 | out_err: | 942 | out_err: |
943 | perf_session__warn_about_errors(self, ops); | ||
944 | perf_session_free_sample_buffers(self); | ||
754 | return err; | 945 | return err; |
755 | } | 946 | } |
756 | 947 | ||
757 | int __perf_session__process_events(struct perf_session *self, | 948 | int __perf_session__process_events(struct perf_session *session, |
758 | u64 data_offset, u64 data_size, | 949 | u64 data_offset, u64 data_size, |
759 | u64 file_size, struct perf_event_ops *ops) | 950 | u64 file_size, struct perf_event_ops *ops) |
760 | { | 951 | { |
761 | int err, mmap_prot, mmap_flags; | 952 | u64 head, page_offset, file_offset, file_pos, progress_next; |
762 | u64 head, shift; | 953 | int err, mmap_prot, mmap_flags, map_idx = 0; |
763 | u64 offset = 0; | 954 | struct ui_progress *progress; |
764 | size_t page_size; | 955 | size_t page_size, mmap_size; |
956 | char *buf, *mmaps[8]; | ||
765 | event_t *event; | 957 | event_t *event; |
766 | uint32_t size; | 958 | uint32_t size; |
767 | char *buf; | ||
768 | struct ui_progress *progress = ui_progress__new("Processing events...", | ||
769 | self->size); | ||
770 | if (progress == NULL) | ||
771 | return -1; | ||
772 | 959 | ||
773 | perf_event_ops__fill_defaults(ops); | 960 | perf_event_ops__fill_defaults(ops); |
774 | 961 | ||
775 | page_size = sysconf(_SC_PAGESIZE); | 962 | page_size = sysconf(_SC_PAGESIZE); |
776 | 963 | ||
777 | head = data_offset; | 964 | page_offset = page_size * (data_offset / page_size); |
778 | shift = page_size * (head / page_size); | 965 | file_offset = page_offset; |
779 | offset += shift; | 966 | head = data_offset - page_offset; |
780 | head -= shift; | 967 | |
968 | if (data_offset + data_size < file_size) | ||
969 | file_size = data_offset + data_size; | ||
970 | |||
971 | progress_next = file_size / 16; | ||
972 | progress = ui_progress__new("Processing events...", file_size); | ||
973 | if (progress == NULL) | ||
974 | return -1; | ||
975 | |||
976 | mmap_size = session->mmap_window; | ||
977 | if (mmap_size > file_size) | ||
978 | mmap_size = file_size; | ||
979 | |||
980 | memset(mmaps, 0, sizeof(mmaps)); | ||
781 | 981 | ||
782 | mmap_prot = PROT_READ; | 982 | mmap_prot = PROT_READ; |
783 | mmap_flags = MAP_SHARED; | 983 | mmap_flags = MAP_SHARED; |
784 | 984 | ||
785 | if (self->header.needs_swap) { | 985 | if (session->header.needs_swap) { |
786 | mmap_prot |= PROT_WRITE; | 986 | mmap_prot |= PROT_WRITE; |
787 | mmap_flags = MAP_PRIVATE; | 987 | mmap_flags = MAP_PRIVATE; |
788 | } | 988 | } |
789 | remap: | 989 | remap: |
790 | buf = mmap(NULL, page_size * self->mmap_window, mmap_prot, | 990 | buf = mmap(NULL, mmap_size, mmap_prot, mmap_flags, session->fd, |
791 | mmap_flags, self->fd, offset); | 991 | file_offset); |
792 | if (buf == MAP_FAILED) { | 992 | if (buf == MAP_FAILED) { |
793 | pr_err("failed to mmap file\n"); | 993 | pr_err("failed to mmap file\n"); |
794 | err = -errno; | 994 | err = -errno; |
795 | goto out_err; | 995 | goto out_err; |
796 | } | 996 | } |
997 | mmaps[map_idx] = buf; | ||
998 | map_idx = (map_idx + 1) & (ARRAY_SIZE(mmaps) - 1); | ||
999 | file_pos = file_offset + head; | ||
797 | 1000 | ||
798 | more: | 1001 | more: |
799 | event = (event_t *)(buf + head); | 1002 | event = (event_t *)(buf + head); |
800 | ui_progress__update(progress, offset); | ||
801 | 1003 | ||
802 | if (self->header.needs_swap) | 1004 | if (session->header.needs_swap) |
803 | perf_event_header__bswap(&event->header); | 1005 | perf_event_header__bswap(&event->header); |
804 | size = event->header.size; | 1006 | size = event->header.size; |
805 | if (size == 0) | 1007 | if (size == 0) |
806 | size = 8; | 1008 | size = 8; |
807 | 1009 | ||
808 | if (head + event->header.size >= page_size * self->mmap_window) { | 1010 | if (head + event->header.size >= mmap_size) { |
809 | int munmap_ret; | 1011 | if (mmaps[map_idx]) { |
810 | 1012 | munmap(mmaps[map_idx], mmap_size); | |
811 | shift = page_size * (head / page_size); | 1013 | mmaps[map_idx] = NULL; |
812 | 1014 | } | |
813 | munmap_ret = munmap(buf, page_size * self->mmap_window); | ||
814 | assert(munmap_ret == 0); | ||
815 | 1015 | ||
816 | offset += shift; | 1016 | page_offset = page_size * (head / page_size); |
817 | head -= shift; | 1017 | file_offset += page_offset; |
1018 | head -= page_offset; | ||
818 | goto remap; | 1019 | goto remap; |
819 | } | 1020 | } |
820 | 1021 | ||
821 | size = event->header.size; | 1022 | size = event->header.size; |
822 | 1023 | ||
823 | dump_printf("\n%#Lx [%#x]: event: %d\n", | ||
824 | offset + head, event->header.size, event->header.type); | ||
825 | |||
826 | if (size == 0 || | 1024 | if (size == 0 || |
827 | perf_session__process_event(self, event, ops, offset, head) < 0) { | 1025 | perf_session__process_event(session, event, ops, file_pos) < 0) { |
828 | dump_printf("%#Lx [%#x]: skipping unknown header type: %d\n", | 1026 | dump_printf("%#Lx [%#x]: skipping unknown header type: %d\n", |
829 | offset + head, event->header.size, | 1027 | file_offset + head, event->header.size, |
830 | event->header.type); | 1028 | event->header.type); |
831 | /* | 1029 | /* |
832 | * assume we lost track of the stream, check alignment, and | 1030 | * assume we lost track of the stream, check alignment, and |
@@ -839,19 +1037,24 @@ more: | |||
839 | } | 1037 | } |
840 | 1038 | ||
841 | head += size; | 1039 | head += size; |
1040 | file_pos += size; | ||
842 | 1041 | ||
843 | if (offset + head >= data_offset + data_size) | 1042 | if (file_pos >= progress_next) { |
844 | goto done; | 1043 | progress_next += file_size / 16; |
1044 | ui_progress__update(progress, file_pos); | ||
1045 | } | ||
845 | 1046 | ||
846 | if (offset + head < file_size) | 1047 | if (file_pos < file_size) |
847 | goto more; | 1048 | goto more; |
848 | done: | 1049 | |
849 | err = 0; | 1050 | err = 0; |
850 | /* do the final flush for ordered samples */ | 1051 | /* do the final flush for ordered samples */ |
851 | self->ordered_samples.next_flush = ULLONG_MAX; | 1052 | session->ordered_samples.next_flush = ULLONG_MAX; |
852 | flush_sample_queue(self, ops); | 1053 | flush_sample_queue(session, ops); |
853 | out_err: | 1054 | out_err: |
854 | ui_progress__delete(progress); | 1055 | ui_progress__delete(progress); |
1056 | perf_session__warn_about_errors(session, ops); | ||
1057 | perf_session_free_sample_buffers(session); | ||
855 | return err; | 1058 | return err; |
856 | } | 1059 | } |
857 | 1060 | ||
diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h index 9fa0fc2a863..decd83f274f 100644 --- a/tools/perf/util/session.h +++ b/tools/perf/util/session.h | |||
@@ -17,8 +17,12 @@ struct ordered_samples { | |||
17 | u64 last_flush; | 17 | u64 last_flush; |
18 | u64 next_flush; | 18 | u64 next_flush; |
19 | u64 max_timestamp; | 19 | u64 max_timestamp; |
20 | struct list_head samples_head; | 20 | struct list_head samples; |
21 | struct sample_queue *last_inserted; | 21 | struct list_head sample_cache; |
22 | struct list_head to_free; | ||
23 | struct sample_queue *sample_buffer; | ||
24 | struct sample_queue *last_sample; | ||
25 | int sample_buffer_idx; | ||
22 | }; | 26 | }; |
23 | 27 | ||
24 | struct perf_session { | 28 | struct perf_session { |
@@ -42,6 +46,8 @@ struct perf_session { | |||
42 | int fd; | 46 | int fd; |
43 | bool fd_pipe; | 47 | bool fd_pipe; |
44 | bool repipe; | 48 | bool repipe; |
49 | bool sample_id_all; | ||
50 | u16 id_hdr_size; | ||
45 | int cwdlen; | 51 | int cwdlen; |
46 | char *cwd; | 52 | char *cwd; |
47 | struct ordered_samples ordered_samples; | 53 | struct ordered_samples ordered_samples; |
@@ -50,7 +56,9 @@ struct perf_session { | |||
50 | 56 | ||
51 | struct perf_event_ops; | 57 | struct perf_event_ops; |
52 | 58 | ||
53 | typedef int (*event_op)(event_t *self, struct perf_session *session); | 59 | typedef int (*event_op)(event_t *self, struct sample_data *sample, |
60 | struct perf_session *session); | ||
61 | typedef int (*event_synth_op)(event_t *self, struct perf_session *session); | ||
54 | typedef int (*event_op2)(event_t *self, struct perf_session *session, | 62 | typedef int (*event_op2)(event_t *self, struct perf_session *session, |
55 | struct perf_event_ops *ops); | 63 | struct perf_event_ops *ops); |
56 | 64 | ||
@@ -63,16 +71,19 @@ struct perf_event_ops { | |||
63 | lost, | 71 | lost, |
64 | read, | 72 | read, |
65 | throttle, | 73 | throttle, |
66 | unthrottle, | 74 | unthrottle; |
67 | attr, | 75 | event_synth_op attr, |
68 | event_type, | 76 | event_type, |
69 | tracing_data, | 77 | tracing_data, |
70 | build_id; | 78 | build_id; |
71 | event_op2 finished_round; | 79 | event_op2 finished_round; |
72 | bool ordered_samples; | 80 | bool ordered_samples; |
81 | bool ordering_requires_timestamps; | ||
73 | }; | 82 | }; |
74 | 83 | ||
75 | struct perf_session *perf_session__new(const char *filename, int mode, bool force, bool repipe); | 84 | struct perf_session *perf_session__new(const char *filename, int mode, |
85 | bool force, bool repipe, | ||
86 | struct perf_event_ops *ops); | ||
76 | void perf_session__delete(struct perf_session *self); | 87 | void perf_session__delete(struct perf_session *self); |
77 | 88 | ||
78 | void perf_event_header__bswap(struct perf_event_header *self); | 89 | void perf_event_header__bswap(struct perf_event_header *self); |
@@ -98,8 +109,9 @@ void mem_bswap_64(void *src, int byte_size); | |||
98 | 109 | ||
99 | int perf_session__create_kernel_maps(struct perf_session *self); | 110 | int perf_session__create_kernel_maps(struct perf_session *self); |
100 | 111 | ||
101 | int do_read(int fd, void *buf, size_t size); | ||
102 | void perf_session__update_sample_type(struct perf_session *self); | 112 | void perf_session__update_sample_type(struct perf_session *self); |
113 | void perf_session__set_sample_id_all(struct perf_session *session, bool value); | ||
114 | void perf_session__set_sample_type(struct perf_session *session, u64 type); | ||
103 | void perf_session__remove_thread(struct perf_session *self, struct thread *th); | 115 | void perf_session__remove_thread(struct perf_session *self, struct thread *th); |
104 | 116 | ||
105 | static inline | 117 | static inline |
diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c index b62a553cc67..f44fa541d56 100644 --- a/tools/perf/util/sort.c +++ b/tools/perf/util/sort.c | |||
@@ -170,7 +170,7 @@ static int hist_entry__dso_snprintf(struct hist_entry *self, char *bf, | |||
170 | return repsep_snprintf(bf, size, "%-*s", width, dso_name); | 170 | return repsep_snprintf(bf, size, "%-*s", width, dso_name); |
171 | } | 171 | } |
172 | 172 | ||
173 | return repsep_snprintf(bf, size, "%*Lx", width, self->ip); | 173 | return repsep_snprintf(bf, size, "%-*s", width, "[unknown]"); |
174 | } | 174 | } |
175 | 175 | ||
176 | /* --sort symbol */ | 176 | /* --sort symbol */ |
@@ -196,7 +196,7 @@ static int hist_entry__sym_snprintf(struct hist_entry *self, char *bf, | |||
196 | 196 | ||
197 | if (verbose) { | 197 | if (verbose) { |
198 | char o = self->ms.map ? dso__symtab_origin(self->ms.map->dso) : '!'; | 198 | char o = self->ms.map ? dso__symtab_origin(self->ms.map->dso) : '!'; |
199 | ret += repsep_snprintf(bf, size, "%*Lx %c ", | 199 | ret += repsep_snprintf(bf, size, "%-#*llx %c ", |
200 | BITS_PER_LONG / 4, self->ip, o); | 200 | BITS_PER_LONG / 4, self->ip, o); |
201 | } | 201 | } |
202 | 202 | ||
@@ -205,7 +205,7 @@ static int hist_entry__sym_snprintf(struct hist_entry *self, char *bf, | |||
205 | ret += repsep_snprintf(bf + ret, size - ret, "%s", | 205 | ret += repsep_snprintf(bf + ret, size - ret, "%s", |
206 | self->ms.sym->name); | 206 | self->ms.sym->name); |
207 | else | 207 | else |
208 | ret += repsep_snprintf(bf + ret, size - ret, "%*Lx", | 208 | ret += repsep_snprintf(bf + ret, size - ret, "%-#*llx", |
209 | BITS_PER_LONG / 4, self->ip); | 209 | BITS_PER_LONG / 4, self->ip); |
210 | 210 | ||
211 | return ret; | 211 | return ret; |
diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 439ab947daf..15ccfba8cdf 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c | |||
@@ -22,6 +22,10 @@ | |||
22 | #include <limits.h> | 22 | #include <limits.h> |
23 | #include <sys/utsname.h> | 23 | #include <sys/utsname.h> |
24 | 24 | ||
25 | #ifndef KSYM_NAME_LEN | ||
26 | #define KSYM_NAME_LEN 128 | ||
27 | #endif | ||
28 | |||
25 | #ifndef NT_GNU_BUILD_ID | 29 | #ifndef NT_GNU_BUILD_ID |
26 | #define NT_GNU_BUILD_ID 3 | 30 | #define NT_GNU_BUILD_ID 3 |
27 | #endif | 31 | #endif |
@@ -41,6 +45,7 @@ struct symbol_conf symbol_conf = { | |||
41 | .exclude_other = true, | 45 | .exclude_other = true, |
42 | .use_modules = true, | 46 | .use_modules = true, |
43 | .try_vmlinux_path = true, | 47 | .try_vmlinux_path = true, |
48 | .symfs = "", | ||
44 | }; | 49 | }; |
45 | 50 | ||
46 | int dso__name_len(const struct dso *self) | 51 | int dso__name_len(const struct dso *self) |
@@ -92,7 +97,7 @@ static void symbols__fixup_end(struct rb_root *self) | |||
92 | prev = curr; | 97 | prev = curr; |
93 | curr = rb_entry(nd, struct symbol, rb_node); | 98 | curr = rb_entry(nd, struct symbol, rb_node); |
94 | 99 | ||
95 | if (prev->end == prev->start) | 100 | if (prev->end == prev->start && prev->end != curr->start) |
96 | prev->end = curr->start - 1; | 101 | prev->end = curr->start - 1; |
97 | } | 102 | } |
98 | 103 | ||
@@ -121,7 +126,7 @@ static void __map_groups__fixup_end(struct map_groups *self, enum map_type type) | |||
121 | * We still haven't the actual symbols, so guess the | 126 | * We still haven't the actual symbols, so guess the |
122 | * last map final address. | 127 | * last map final address. |
123 | */ | 128 | */ |
124 | curr->end = ~0UL; | 129 | curr->end = ~0ULL; |
125 | } | 130 | } |
126 | 131 | ||
127 | static void map_groups__fixup_end(struct map_groups *self) | 132 | static void map_groups__fixup_end(struct map_groups *self) |
@@ -425,16 +430,25 @@ size_t dso__fprintf(struct dso *self, enum map_type type, FILE *fp) | |||
425 | 430 | ||
426 | int kallsyms__parse(const char *filename, void *arg, | 431 | int kallsyms__parse(const char *filename, void *arg, |
427 | int (*process_symbol)(void *arg, const char *name, | 432 | int (*process_symbol)(void *arg, const char *name, |
428 | char type, u64 start)) | 433 | char type, u64 start, u64 end)) |
429 | { | 434 | { |
430 | char *line = NULL; | 435 | char *line = NULL; |
431 | size_t n; | 436 | size_t n; |
432 | int err = 0; | 437 | int err = -1; |
438 | u64 prev_start = 0; | ||
439 | char prev_symbol_type = 0; | ||
440 | char *prev_symbol_name; | ||
433 | FILE *file = fopen(filename, "r"); | 441 | FILE *file = fopen(filename, "r"); |
434 | 442 | ||
435 | if (file == NULL) | 443 | if (file == NULL) |
436 | goto out_failure; | 444 | goto out_failure; |
437 | 445 | ||
446 | prev_symbol_name = malloc(KSYM_NAME_LEN); | ||
447 | if (prev_symbol_name == NULL) | ||
448 | goto out_close; | ||
449 | |||
450 | err = 0; | ||
451 | |||
438 | while (!feof(file)) { | 452 | while (!feof(file)) { |
439 | u64 start; | 453 | u64 start; |
440 | int line_len, len; | 454 | int line_len, len; |
@@ -454,14 +468,33 @@ int kallsyms__parse(const char *filename, void *arg, | |||
454 | continue; | 468 | continue; |
455 | 469 | ||
456 | symbol_type = toupper(line[len]); | 470 | symbol_type = toupper(line[len]); |
457 | symbol_name = line + len + 2; | 471 | len += 2; |
472 | symbol_name = line + len; | ||
473 | len = line_len - len; | ||
458 | 474 | ||
459 | err = process_symbol(arg, symbol_name, symbol_type, start); | 475 | if (len >= KSYM_NAME_LEN) { |
460 | if (err) | 476 | err = -1; |
461 | break; | 477 | break; |
478 | } | ||
479 | |||
480 | if (prev_symbol_type) { | ||
481 | u64 end = start; | ||
482 | if (end != prev_start) | ||
483 | --end; | ||
484 | err = process_symbol(arg, prev_symbol_name, | ||
485 | prev_symbol_type, prev_start, end); | ||
486 | if (err) | ||
487 | break; | ||
488 | } | ||
489 | |||
490 | memcpy(prev_symbol_name, symbol_name, len + 1); | ||
491 | prev_symbol_type = symbol_type; | ||
492 | prev_start = start; | ||
462 | } | 493 | } |
463 | 494 | ||
495 | free(prev_symbol_name); | ||
464 | free(line); | 496 | free(line); |
497 | out_close: | ||
465 | fclose(file); | 498 | fclose(file); |
466 | return err; | 499 | return err; |
467 | 500 | ||
@@ -483,7 +516,7 @@ static u8 kallsyms2elf_type(char type) | |||
483 | } | 516 | } |
484 | 517 | ||
485 | static int map__process_kallsym_symbol(void *arg, const char *name, | 518 | static int map__process_kallsym_symbol(void *arg, const char *name, |
486 | char type, u64 start) | 519 | char type, u64 start, u64 end) |
487 | { | 520 | { |
488 | struct symbol *sym; | 521 | struct symbol *sym; |
489 | struct process_kallsyms_args *a = arg; | 522 | struct process_kallsyms_args *a = arg; |
@@ -492,11 +525,8 @@ static int map__process_kallsym_symbol(void *arg, const char *name, | |||
492 | if (!symbol_type__is_a(type, a->map->type)) | 525 | if (!symbol_type__is_a(type, a->map->type)) |
493 | return 0; | 526 | return 0; |
494 | 527 | ||
495 | /* | 528 | sym = symbol__new(start, end - start + 1, |
496 | * Will fix up the end later, when we have all symbols sorted. | 529 | kallsyms2elf_type(type), name); |
497 | */ | ||
498 | sym = symbol__new(start, 0, kallsyms2elf_type(type), name); | ||
499 | |||
500 | if (sym == NULL) | 530 | if (sym == NULL) |
501 | return -ENOMEM; | 531 | return -ENOMEM; |
502 | /* | 532 | /* |
@@ -649,7 +679,6 @@ int dso__load_kallsyms(struct dso *self, const char *filename, | |||
649 | if (dso__load_all_kallsyms(self, filename, map) < 0) | 679 | if (dso__load_all_kallsyms(self, filename, map) < 0) |
650 | return -1; | 680 | return -1; |
651 | 681 | ||
652 | symbols__fixup_end(&self->symbols[map->type]); | ||
653 | if (self->kernel == DSO_TYPE_GUEST_KERNEL) | 682 | if (self->kernel == DSO_TYPE_GUEST_KERNEL) |
654 | self->origin = DSO__ORIG_GUEST_KERNEL; | 683 | self->origin = DSO__ORIG_GUEST_KERNEL; |
655 | else | 684 | else |
@@ -839,8 +868,11 @@ static int dso__synthesize_plt_symbols(struct dso *self, struct map *map, | |||
839 | char sympltname[1024]; | 868 | char sympltname[1024]; |
840 | Elf *elf; | 869 | Elf *elf; |
841 | int nr = 0, symidx, fd, err = 0; | 870 | int nr = 0, symidx, fd, err = 0; |
871 | char name[PATH_MAX]; | ||
842 | 872 | ||
843 | fd = open(self->long_name, O_RDONLY); | 873 | snprintf(name, sizeof(name), "%s%s", |
874 | symbol_conf.symfs, self->long_name); | ||
875 | fd = open(name, O_RDONLY); | ||
844 | if (fd < 0) | 876 | if (fd < 0) |
845 | goto out; | 877 | goto out; |
846 | 878 | ||
@@ -1452,16 +1484,19 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) | |||
1452 | self->origin++) { | 1484 | self->origin++) { |
1453 | switch (self->origin) { | 1485 | switch (self->origin) { |
1454 | case DSO__ORIG_BUILD_ID_CACHE: | 1486 | case DSO__ORIG_BUILD_ID_CACHE: |
1455 | if (dso__build_id_filename(self, name, size) == NULL) | 1487 | /* skip the locally configured cache if a symfs is given */ |
1488 | if (symbol_conf.symfs[0] || | ||
1489 | (dso__build_id_filename(self, name, size) == NULL)) { | ||
1456 | continue; | 1490 | continue; |
1491 | } | ||
1457 | break; | 1492 | break; |
1458 | case DSO__ORIG_FEDORA: | 1493 | case DSO__ORIG_FEDORA: |
1459 | snprintf(name, size, "/usr/lib/debug%s.debug", | 1494 | snprintf(name, size, "%s/usr/lib/debug%s.debug", |
1460 | self->long_name); | 1495 | symbol_conf.symfs, self->long_name); |
1461 | break; | 1496 | break; |
1462 | case DSO__ORIG_UBUNTU: | 1497 | case DSO__ORIG_UBUNTU: |
1463 | snprintf(name, size, "/usr/lib/debug%s", | 1498 | snprintf(name, size, "%s/usr/lib/debug%s", |
1464 | self->long_name); | 1499 | symbol_conf.symfs, self->long_name); |
1465 | break; | 1500 | break; |
1466 | case DSO__ORIG_BUILDID: { | 1501 | case DSO__ORIG_BUILDID: { |
1467 | char build_id_hex[BUILD_ID_SIZE * 2 + 1]; | 1502 | char build_id_hex[BUILD_ID_SIZE * 2 + 1]; |
@@ -1473,19 +1508,26 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) | |||
1473 | sizeof(self->build_id), | 1508 | sizeof(self->build_id), |
1474 | build_id_hex); | 1509 | build_id_hex); |
1475 | snprintf(name, size, | 1510 | snprintf(name, size, |
1476 | "/usr/lib/debug/.build-id/%.2s/%s.debug", | 1511 | "%s/usr/lib/debug/.build-id/%.2s/%s.debug", |
1477 | build_id_hex, build_id_hex + 2); | 1512 | symbol_conf.symfs, build_id_hex, build_id_hex + 2); |
1478 | } | 1513 | } |
1479 | break; | 1514 | break; |
1480 | case DSO__ORIG_DSO: | 1515 | case DSO__ORIG_DSO: |
1481 | snprintf(name, size, "%s", self->long_name); | 1516 | snprintf(name, size, "%s%s", |
1517 | symbol_conf.symfs, self->long_name); | ||
1482 | break; | 1518 | break; |
1483 | case DSO__ORIG_GUEST_KMODULE: | 1519 | case DSO__ORIG_GUEST_KMODULE: |
1484 | if (map->groups && map->groups->machine) | 1520 | if (map->groups && map->groups->machine) |
1485 | root_dir = map->groups->machine->root_dir; | 1521 | root_dir = map->groups->machine->root_dir; |
1486 | else | 1522 | else |
1487 | root_dir = ""; | 1523 | root_dir = ""; |
1488 | snprintf(name, size, "%s%s", root_dir, self->long_name); | 1524 | snprintf(name, size, "%s%s%s", symbol_conf.symfs, |
1525 | root_dir, self->long_name); | ||
1526 | break; | ||
1527 | |||
1528 | case DSO__ORIG_KMODULE: | ||
1529 | snprintf(name, size, "%s%s", symbol_conf.symfs, | ||
1530 | self->long_name); | ||
1489 | break; | 1531 | break; |
1490 | 1532 | ||
1491 | default: | 1533 | default: |
@@ -1784,17 +1826,20 @@ int dso__load_vmlinux(struct dso *self, struct map *map, | |||
1784 | const char *vmlinux, symbol_filter_t filter) | 1826 | const char *vmlinux, symbol_filter_t filter) |
1785 | { | 1827 | { |
1786 | int err = -1, fd; | 1828 | int err = -1, fd; |
1829 | char symfs_vmlinux[PATH_MAX]; | ||
1787 | 1830 | ||
1788 | fd = open(vmlinux, O_RDONLY); | 1831 | snprintf(symfs_vmlinux, sizeof(symfs_vmlinux), "%s/%s", |
1832 | symbol_conf.symfs, vmlinux); | ||
1833 | fd = open(symfs_vmlinux, O_RDONLY); | ||
1789 | if (fd < 0) | 1834 | if (fd < 0) |
1790 | return -1; | 1835 | return -1; |
1791 | 1836 | ||
1792 | dso__set_loaded(self, map->type); | 1837 | dso__set_loaded(self, map->type); |
1793 | err = dso__load_sym(self, map, vmlinux, fd, filter, 0, 0); | 1838 | err = dso__load_sym(self, map, symfs_vmlinux, fd, filter, 0, 0); |
1794 | close(fd); | 1839 | close(fd); |
1795 | 1840 | ||
1796 | if (err > 0) | 1841 | if (err > 0) |
1797 | pr_debug("Using %s for symbols\n", vmlinux); | 1842 | pr_debug("Using %s for symbols\n", symfs_vmlinux); |
1798 | 1843 | ||
1799 | return err; | 1844 | return err; |
1800 | } | 1845 | } |
@@ -1836,8 +1881,8 @@ static int dso__load_kernel_sym(struct dso *self, struct map *map, | |||
1836 | const char *kallsyms_filename = NULL; | 1881 | const char *kallsyms_filename = NULL; |
1837 | char *kallsyms_allocated_filename = NULL; | 1882 | char *kallsyms_allocated_filename = NULL; |
1838 | /* | 1883 | /* |
1839 | * Step 1: if the user specified a vmlinux filename, use it and only | 1884 | * Step 1: if the user specified a kallsyms or vmlinux filename, use |
1840 | * it, reporting errors to the user if it cannot be used. | 1885 | * it and only it, reporting errors to the user if it cannot be used. |
1841 | * | 1886 | * |
1842 | * For instance, try to analyse an ARM perf.data file _without_ a | 1887 | * For instance, try to analyse an ARM perf.data file _without_ a |
1843 | * build-id, or if the user specifies the wrong path to the right | 1888 | * build-id, or if the user specifies the wrong path to the right |
@@ -1850,6 +1895,11 @@ static int dso__load_kernel_sym(struct dso *self, struct map *map, | |||
1850 | * validation in dso__load_vmlinux and will bail out if they don't | 1895 | * validation in dso__load_vmlinux and will bail out if they don't |
1851 | * match. | 1896 | * match. |
1852 | */ | 1897 | */ |
1898 | if (symbol_conf.kallsyms_name != NULL) { | ||
1899 | kallsyms_filename = symbol_conf.kallsyms_name; | ||
1900 | goto do_kallsyms; | ||
1901 | } | ||
1902 | |||
1853 | if (symbol_conf.vmlinux_name != NULL) { | 1903 | if (symbol_conf.vmlinux_name != NULL) { |
1854 | err = dso__load_vmlinux(self, map, | 1904 | err = dso__load_vmlinux(self, map, |
1855 | symbol_conf.vmlinux_name, filter); | 1905 | symbol_conf.vmlinux_name, filter); |
@@ -1867,6 +1917,10 @@ static int dso__load_kernel_sym(struct dso *self, struct map *map, | |||
1867 | goto out_fixup; | 1917 | goto out_fixup; |
1868 | } | 1918 | } |
1869 | 1919 | ||
1920 | /* do not try local files if a symfs was given */ | ||
1921 | if (symbol_conf.symfs[0] != 0) | ||
1922 | return -1; | ||
1923 | |||
1870 | /* | 1924 | /* |
1871 | * Say the kernel DSO was created when processing the build-id header table, | 1925 | * Say the kernel DSO was created when processing the build-id header table, |
1872 | * we have a build-id, so check if it is the same as the running kernel, | 1926 | * we have a build-id, so check if it is the same as the running kernel, |
@@ -2136,7 +2190,7 @@ struct process_args { | |||
2136 | }; | 2190 | }; |
2137 | 2191 | ||
2138 | static int symbol__in_kernel(void *arg, const char *name, | 2192 | static int symbol__in_kernel(void *arg, const char *name, |
2139 | char type __used, u64 start) | 2193 | char type __used, u64 start, u64 end __used) |
2140 | { | 2194 | { |
2141 | struct process_args *args = arg; | 2195 | struct process_args *args = arg; |
2142 | 2196 | ||
@@ -2257,9 +2311,6 @@ static int vmlinux_path__init(void) | |||
2257 | struct utsname uts; | 2311 | struct utsname uts; |
2258 | char bf[PATH_MAX]; | 2312 | char bf[PATH_MAX]; |
2259 | 2313 | ||
2260 | if (uname(&uts) < 0) | ||
2261 | return -1; | ||
2262 | |||
2263 | vmlinux_path = malloc(sizeof(char *) * 5); | 2314 | vmlinux_path = malloc(sizeof(char *) * 5); |
2264 | if (vmlinux_path == NULL) | 2315 | if (vmlinux_path == NULL) |
2265 | return -1; | 2316 | return -1; |
@@ -2272,6 +2323,14 @@ static int vmlinux_path__init(void) | |||
2272 | if (vmlinux_path[vmlinux_path__nr_entries] == NULL) | 2323 | if (vmlinux_path[vmlinux_path__nr_entries] == NULL) |
2273 | goto out_fail; | 2324 | goto out_fail; |
2274 | ++vmlinux_path__nr_entries; | 2325 | ++vmlinux_path__nr_entries; |
2326 | |||
2327 | /* only try running kernel version if no symfs was given */ | ||
2328 | if (symbol_conf.symfs[0] != 0) | ||
2329 | return 0; | ||
2330 | |||
2331 | if (uname(&uts) < 0) | ||
2332 | return -1; | ||
2333 | |||
2275 | snprintf(bf, sizeof(bf), "/boot/vmlinux-%s", uts.release); | 2334 | snprintf(bf, sizeof(bf), "/boot/vmlinux-%s", uts.release); |
2276 | vmlinux_path[vmlinux_path__nr_entries] = strdup(bf); | 2335 | vmlinux_path[vmlinux_path__nr_entries] = strdup(bf); |
2277 | if (vmlinux_path[vmlinux_path__nr_entries] == NULL) | 2336 | if (vmlinux_path[vmlinux_path__nr_entries] == NULL) |
@@ -2331,6 +2390,8 @@ static int setup_list(struct strlist **list, const char *list_str, | |||
2331 | 2390 | ||
2332 | int symbol__init(void) | 2391 | int symbol__init(void) |
2333 | { | 2392 | { |
2393 | const char *symfs; | ||
2394 | |||
2334 | if (symbol_conf.initialized) | 2395 | if (symbol_conf.initialized) |
2335 | return 0; | 2396 | return 0; |
2336 | 2397 | ||
@@ -2359,6 +2420,18 @@ int symbol__init(void) | |||
2359 | symbol_conf.sym_list_str, "symbol") < 0) | 2420 | symbol_conf.sym_list_str, "symbol") < 0) |
2360 | goto out_free_comm_list; | 2421 | goto out_free_comm_list; |
2361 | 2422 | ||
2423 | /* | ||
2424 | * A path to symbols of "/" is identical to "" | ||
2425 | * reset here for simplicity. | ||
2426 | */ | ||
2427 | symfs = realpath(symbol_conf.symfs, NULL); | ||
2428 | if (symfs == NULL) | ||
2429 | symfs = symbol_conf.symfs; | ||
2430 | if (strcmp(symfs, "/") == 0) | ||
2431 | symbol_conf.symfs = ""; | ||
2432 | if (symfs != symbol_conf.symfs) | ||
2433 | free((void *)symfs); | ||
2434 | |||
2362 | symbol_conf.initialized = true; | 2435 | symbol_conf.initialized = true; |
2363 | return 0; | 2436 | return 0; |
2364 | 2437 | ||
diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 6c6eafdb932..670cd1c88f5 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h | |||
@@ -72,6 +72,7 @@ struct symbol_conf { | |||
72 | show_cpu_utilization, | 72 | show_cpu_utilization, |
73 | initialized; | 73 | initialized; |
74 | const char *vmlinux_name, | 74 | const char *vmlinux_name, |
75 | *kallsyms_name, | ||
75 | *source_prefix, | 76 | *source_prefix, |
76 | *field_sep; | 77 | *field_sep; |
77 | const char *default_guest_vmlinux_name, | 78 | const char *default_guest_vmlinux_name, |
@@ -85,6 +86,7 @@ struct symbol_conf { | |||
85 | struct strlist *dso_list, | 86 | struct strlist *dso_list, |
86 | *comm_list, | 87 | *comm_list, |
87 | *sym_list; | 88 | *sym_list; |
89 | const char *symfs; | ||
88 | }; | 90 | }; |
89 | 91 | ||
90 | extern struct symbol_conf symbol_conf; | 92 | extern struct symbol_conf symbol_conf; |
@@ -215,7 +217,7 @@ bool __dsos__read_build_ids(struct list_head *head, bool with_hits); | |||
215 | int build_id__sprintf(const u8 *self, int len, char *bf); | 217 | int build_id__sprintf(const u8 *self, int len, char *bf); |
216 | int kallsyms__parse(const char *filename, void *arg, | 218 | int kallsyms__parse(const char *filename, void *arg, |
217 | int (*process_symbol)(void *arg, const char *name, | 219 | int (*process_symbol)(void *arg, const char *name, |
218 | char type, u64 start)); | 220 | char type, u64 start, u64 end)); |
219 | 221 | ||
220 | void machine__destroy_kernel_maps(struct machine *self); | 222 | void machine__destroy_kernel_maps(struct machine *self); |
221 | int __machine__create_kernel_maps(struct machine *self, struct dso *kernel); | 223 | int __machine__create_kernel_maps(struct machine *self, struct dso *kernel); |
diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c index 8c72d888e44..00f4eade2e3 100644 --- a/tools/perf/util/thread.c +++ b/tools/perf/util/thread.c | |||
@@ -16,35 +16,50 @@ static int filter(const struct dirent *dir) | |||
16 | return 1; | 16 | return 1; |
17 | } | 17 | } |
18 | 18 | ||
19 | int find_all_tid(int pid, pid_t ** all_tid) | 19 | struct thread_map *thread_map__new_by_pid(pid_t pid) |
20 | { | 20 | { |
21 | struct thread_map *threads; | ||
21 | char name[256]; | 22 | char name[256]; |
22 | int items; | 23 | int items; |
23 | struct dirent **namelist = NULL; | 24 | struct dirent **namelist = NULL; |
24 | int ret = 0; | ||
25 | int i; | 25 | int i; |
26 | 26 | ||
27 | sprintf(name, "/proc/%d/task", pid); | 27 | sprintf(name, "/proc/%d/task", pid); |
28 | items = scandir(name, &namelist, filter, NULL); | 28 | items = scandir(name, &namelist, filter, NULL); |
29 | if (items <= 0) | 29 | if (items <= 0) |
30 | return -ENOENT; | 30 | return NULL; |
31 | *all_tid = malloc(sizeof(pid_t) * items); | ||
32 | if (!*all_tid) { | ||
33 | ret = -ENOMEM; | ||
34 | goto failure; | ||
35 | } | ||
36 | |||
37 | for (i = 0; i < items; i++) | ||
38 | (*all_tid)[i] = atoi(namelist[i]->d_name); | ||
39 | 31 | ||
40 | ret = items; | 32 | threads = malloc(sizeof(*threads) + sizeof(pid_t) * items); |
33 | if (threads != NULL) { | ||
34 | for (i = 0; i < items; i++) | ||
35 | threads->map[i] = atoi(namelist[i]->d_name); | ||
36 | threads->nr = items; | ||
37 | } | ||
41 | 38 | ||
42 | failure: | ||
43 | for (i=0; i<items; i++) | 39 | for (i=0; i<items; i++) |
44 | free(namelist[i]); | 40 | free(namelist[i]); |
45 | free(namelist); | 41 | free(namelist); |
46 | 42 | ||
47 | return ret; | 43 | return threads; |
44 | } | ||
45 | |||
46 | struct thread_map *thread_map__new_by_tid(pid_t tid) | ||
47 | { | ||
48 | struct thread_map *threads = malloc(sizeof(*threads) + sizeof(pid_t)); | ||
49 | |||
50 | if (threads != NULL) { | ||
51 | threads->map[0] = tid; | ||
52 | threads->nr = 1; | ||
53 | } | ||
54 | |||
55 | return threads; | ||
56 | } | ||
57 | |||
58 | struct thread_map *thread_map__new(pid_t pid, pid_t tid) | ||
59 | { | ||
60 | if (pid != -1) | ||
61 | return thread_map__new_by_pid(pid); | ||
62 | return thread_map__new_by_tid(tid); | ||
48 | } | 63 | } |
49 | 64 | ||
50 | static struct thread *thread__new(pid_t pid) | 65 | static struct thread *thread__new(pid_t pid) |
diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h index 688500ff826..d7574101054 100644 --- a/tools/perf/util/thread.h +++ b/tools/perf/util/thread.h | |||
@@ -18,11 +18,24 @@ struct thread { | |||
18 | int comm_len; | 18 | int comm_len; |
19 | }; | 19 | }; |
20 | 20 | ||
21 | struct thread_map { | ||
22 | int nr; | ||
23 | int map[]; | ||
24 | }; | ||
25 | |||
21 | struct perf_session; | 26 | struct perf_session; |
22 | 27 | ||
23 | void thread__delete(struct thread *self); | 28 | void thread__delete(struct thread *self); |
24 | 29 | ||
25 | int find_all_tid(int pid, pid_t ** all_tid); | 30 | struct thread_map *thread_map__new_by_pid(pid_t pid); |
31 | struct thread_map *thread_map__new_by_tid(pid_t tid); | ||
32 | struct thread_map *thread_map__new(pid_t pid, pid_t tid); | ||
33 | |||
34 | static inline void thread_map__delete(struct thread_map *threads) | ||
35 | { | ||
36 | free(threads); | ||
37 | } | ||
38 | |||
26 | int thread__set_comm(struct thread *self, const char *comm); | 39 | int thread__set_comm(struct thread *self, const char *comm); |
27 | int thread__comm_len(struct thread *self); | 40 | int thread__comm_len(struct thread *self); |
28 | struct thread *perf_session__findnew(struct perf_session *self, pid_t pid); | 41 | struct thread *perf_session__findnew(struct perf_session *self, pid_t pid); |
diff --git a/tools/perf/util/trace-event-info.c b/tools/perf/util/trace-event-info.c index b1572601286..35729f4c40c 100644 --- a/tools/perf/util/trace-event-info.c +++ b/tools/perf/util/trace-event-info.c | |||
@@ -34,11 +34,13 @@ | |||
34 | #include <ctype.h> | 34 | #include <ctype.h> |
35 | #include <errno.h> | 35 | #include <errno.h> |
36 | #include <stdbool.h> | 36 | #include <stdbool.h> |
37 | #include <linux/list.h> | ||
37 | #include <linux/kernel.h> | 38 | #include <linux/kernel.h> |
38 | 39 | ||
39 | #include "../perf.h" | 40 | #include "../perf.h" |
40 | #include "trace-event.h" | 41 | #include "trace-event.h" |
41 | #include "debugfs.h" | 42 | #include "debugfs.h" |
43 | #include "evsel.h" | ||
42 | 44 | ||
43 | #define VERSION "0.5" | 45 | #define VERSION "0.5" |
44 | 46 | ||
@@ -469,16 +471,17 @@ out: | |||
469 | } | 471 | } |
470 | 472 | ||
471 | static struct tracepoint_path * | 473 | static struct tracepoint_path * |
472 | get_tracepoints_path(struct perf_event_attr *pattrs, int nb_events) | 474 | get_tracepoints_path(struct list_head *pattrs) |
473 | { | 475 | { |
474 | struct tracepoint_path path, *ppath = &path; | 476 | struct tracepoint_path path, *ppath = &path; |
475 | int i, nr_tracepoints = 0; | 477 | struct perf_evsel *pos; |
478 | int nr_tracepoints = 0; | ||
476 | 479 | ||
477 | for (i = 0; i < nb_events; i++) { | 480 | list_for_each_entry(pos, pattrs, node) { |
478 | if (pattrs[i].type != PERF_TYPE_TRACEPOINT) | 481 | if (pos->attr.type != PERF_TYPE_TRACEPOINT) |
479 | continue; | 482 | continue; |
480 | ++nr_tracepoints; | 483 | ++nr_tracepoints; |
481 | ppath->next = tracepoint_id_to_path(pattrs[i].config); | 484 | ppath->next = tracepoint_id_to_path(pos->attr.config); |
482 | if (!ppath->next) | 485 | if (!ppath->next) |
483 | die("%s\n", "No memory to alloc tracepoints list"); | 486 | die("%s\n", "No memory to alloc tracepoints list"); |
484 | ppath = ppath->next; | 487 | ppath = ppath->next; |
@@ -487,21 +490,21 @@ get_tracepoints_path(struct perf_event_attr *pattrs, int nb_events) | |||
487 | return nr_tracepoints > 0 ? path.next : NULL; | 490 | return nr_tracepoints > 0 ? path.next : NULL; |
488 | } | 491 | } |
489 | 492 | ||
490 | bool have_tracepoints(struct perf_event_attr *pattrs, int nb_events) | 493 | bool have_tracepoints(struct list_head *pattrs) |
491 | { | 494 | { |
492 | int i; | 495 | struct perf_evsel *pos; |
493 | 496 | ||
494 | for (i = 0; i < nb_events; i++) | 497 | list_for_each_entry(pos, pattrs, node) |
495 | if (pattrs[i].type == PERF_TYPE_TRACEPOINT) | 498 | if (pos->attr.type == PERF_TYPE_TRACEPOINT) |
496 | return true; | 499 | return true; |
497 | 500 | ||
498 | return false; | 501 | return false; |
499 | } | 502 | } |
500 | 503 | ||
501 | int read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events) | 504 | int read_tracing_data(int fd, struct list_head *pattrs) |
502 | { | 505 | { |
503 | char buf[BUFSIZ]; | 506 | char buf[BUFSIZ]; |
504 | struct tracepoint_path *tps = get_tracepoints_path(pattrs, nb_events); | 507 | struct tracepoint_path *tps = get_tracepoints_path(pattrs); |
505 | 508 | ||
506 | /* | 509 | /* |
507 | * What? No tracepoints? No sense writing anything here, bail out. | 510 | * What? No tracepoints? No sense writing anything here, bail out. |
@@ -545,14 +548,13 @@ int read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events) | |||
545 | return 0; | 548 | return 0; |
546 | } | 549 | } |
547 | 550 | ||
548 | ssize_t read_tracing_data_size(int fd, struct perf_event_attr *pattrs, | 551 | ssize_t read_tracing_data_size(int fd, struct list_head *pattrs) |
549 | int nb_events) | ||
550 | { | 552 | { |
551 | ssize_t size; | 553 | ssize_t size; |
552 | int err = 0; | 554 | int err = 0; |
553 | 555 | ||
554 | calc_data_size = 1; | 556 | calc_data_size = 1; |
555 | err = read_tracing_data(fd, pattrs, nb_events); | 557 | err = read_tracing_data(fd, pattrs); |
556 | size = calc_data_size - 1; | 558 | size = calc_data_size - 1; |
557 | calc_data_size = 0; | 559 | calc_data_size = 0; |
558 | 560 | ||
diff --git a/tools/perf/util/trace-event.h b/tools/perf/util/trace-event.h index b3e86b1e444..b5f12ca24d9 100644 --- a/tools/perf/util/trace-event.h +++ b/tools/perf/util/trace-event.h | |||
@@ -262,9 +262,8 @@ raw_field_value(struct event *event, const char *name, void *data); | |||
262 | void *raw_field_ptr(struct event *event, const char *name, void *data); | 262 | void *raw_field_ptr(struct event *event, const char *name, void *data); |
263 | unsigned long long eval_flag(const char *flag); | 263 | unsigned long long eval_flag(const char *flag); |
264 | 264 | ||
265 | int read_tracing_data(int fd, struct perf_event_attr *pattrs, int nb_events); | 265 | int read_tracing_data(int fd, struct list_head *pattrs); |
266 | ssize_t read_tracing_data_size(int fd, struct perf_event_attr *pattrs, | 266 | ssize_t read_tracing_data_size(int fd, struct list_head *pattrs); |
267 | int nb_events); | ||
268 | 267 | ||
269 | /* taken from kernel/trace/trace.h */ | 268 | /* taken from kernel/trace/trace.h */ |
270 | enum trace_flag_type { | 269 | enum trace_flag_type { |
diff --git a/tools/perf/util/ui/util.c b/tools/perf/util/ui/util.c index 056c69521a3..7b5a8926624 100644 --- a/tools/perf/util/ui/util.c +++ b/tools/perf/util/ui/util.c | |||
@@ -104,10 +104,24 @@ out_destroy_form: | |||
104 | return rc; | 104 | return rc; |
105 | } | 105 | } |
106 | 106 | ||
107 | static const char yes[] = "Yes", no[] = "No"; | 107 | static const char yes[] = "Yes", no[] = "No", |
108 | warning_str[] = "Warning!", ok[] = "Ok"; | ||
108 | 109 | ||
109 | bool ui__dialog_yesno(const char *msg) | 110 | bool ui__dialog_yesno(const char *msg) |
110 | { | 111 | { |
111 | /* newtWinChoice should really be accepting const char pointers... */ | 112 | /* newtWinChoice should really be accepting const char pointers... */ |
112 | return newtWinChoice(NULL, (char *)yes, (char *)no, (char *)msg) == 1; | 113 | return newtWinChoice(NULL, (char *)yes, (char *)no, (char *)msg) == 1; |
113 | } | 114 | } |
115 | |||
116 | void ui__warning(const char *format, ...) | ||
117 | { | ||
118 | va_list args; | ||
119 | |||
120 | va_start(args, format); | ||
121 | if (use_browser > 0) | ||
122 | newtWinMessagev((char *)warning_str, (char *)ok, | ||
123 | (char *)format, args); | ||
124 | else | ||
125 | vfprintf(stderr, format, args); | ||
126 | va_end(args); | ||
127 | } | ||
diff --git a/tools/perf/util/util.c b/tools/perf/util/util.c index 214265674dd..5b3ea49aa63 100644 --- a/tools/perf/util/util.c +++ b/tools/perf/util/util.c | |||
@@ -114,3 +114,20 @@ unsigned long convert_unit(unsigned long value, char *unit) | |||
114 | 114 | ||
115 | return value; | 115 | return value; |
116 | } | 116 | } |
117 | |||
118 | int readn(int fd, void *buf, size_t n) | ||
119 | { | ||
120 | void *buf_start = buf; | ||
121 | |||
122 | while (n) { | ||
123 | int ret = read(fd, buf, n); | ||
124 | |||
125 | if (ret <= 0) | ||
126 | return ret; | ||
127 | |||
128 | n -= ret; | ||
129 | buf += ret; | ||
130 | } | ||
131 | |||
132 | return buf - buf_start; | ||
133 | } | ||
diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h index 7562707ddd1..e833f26f3bf 100644 --- a/tools/perf/util/util.h +++ b/tools/perf/util/util.h | |||
@@ -265,6 +265,7 @@ void argv_free(char **argv); | |||
265 | bool strglobmatch(const char *str, const char *pat); | 265 | bool strglobmatch(const char *str, const char *pat); |
266 | bool strlazymatch(const char *str, const char *pat); | 266 | bool strlazymatch(const char *str, const char *pat); |
267 | unsigned long convert_unit(unsigned long value, char *unit); | 267 | unsigned long convert_unit(unsigned long value, char *unit); |
268 | int readn(int fd, void *buf, size_t size); | ||
268 | 269 | ||
269 | #define _STR(x) #x | 270 | #define _STR(x) #x |
270 | #define STR(x) _STR(x) | 271 | #define STR(x) _STR(x) |
diff --git a/tools/perf/util/xyarray.c b/tools/perf/util/xyarray.c new file mode 100644 index 00000000000..22afbf6c536 --- /dev/null +++ b/tools/perf/util/xyarray.c | |||
@@ -0,0 +1,20 @@ | |||
1 | #include "xyarray.h" | ||
2 | #include "util.h" | ||
3 | |||
4 | struct xyarray *xyarray__new(int xlen, int ylen, size_t entry_size) | ||
5 | { | ||
6 | size_t row_size = ylen * entry_size; | ||
7 | struct xyarray *xy = zalloc(sizeof(*xy) + xlen * row_size); | ||
8 | |||
9 | if (xy != NULL) { | ||
10 | xy->entry_size = entry_size; | ||
11 | xy->row_size = row_size; | ||
12 | } | ||
13 | |||
14 | return xy; | ||
15 | } | ||
16 | |||
17 | void xyarray__delete(struct xyarray *xy) | ||
18 | { | ||
19 | free(xy); | ||
20 | } | ||
diff --git a/tools/perf/util/xyarray.h b/tools/perf/util/xyarray.h new file mode 100644 index 00000000000..c488a07275d --- /dev/null +++ b/tools/perf/util/xyarray.h | |||
@@ -0,0 +1,20 @@ | |||
1 | #ifndef _PERF_XYARRAY_H_ | ||
2 | #define _PERF_XYARRAY_H_ 1 | ||
3 | |||
4 | #include <sys/types.h> | ||
5 | |||
6 | struct xyarray { | ||
7 | size_t row_size; | ||
8 | size_t entry_size; | ||
9 | char contents[]; | ||
10 | }; | ||
11 | |||
12 | struct xyarray *xyarray__new(int xlen, int ylen, size_t entry_size); | ||
13 | void xyarray__delete(struct xyarray *xy); | ||
14 | |||
15 | static inline void *xyarray__entry(struct xyarray *xy, int x, int y) | ||
16 | { | ||
17 | return &xy->contents[x * xy->row_size + y * xy->entry_size]; | ||
18 | } | ||
19 | |||
20 | #endif /* _PERF_XYARRAY_H_ */ | ||