aboutsummaryrefslogtreecommitdiffstats
path: root/tools
diff options
context:
space:
mode:
authorDavid S. Miller <davem@davemloft.net>2012-10-09 13:14:32 -0400
committerDavid S. Miller <davem@davemloft.net>2012-10-09 13:14:32 -0400
commit8dd9117cc7a021ced1c5cf177e2d44dd92b88617 (patch)
treecad990f58f9ec6d400226dda86718fc10781416e /tools
parent16e310ae6ed352c4963b1f2413fcd88fa693eeda (diff)
parent547b1e81afe3119f7daf702cc03b158495535a25 (diff)
Merge git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux
Pulled mainline in order to get the UAPI infrastructure already merged before I pull in David Howells's UAPI trees for networking. Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'tools')
-rw-r--r--tools/lguest/lguest.c1
-rw-r--r--tools/perf/Makefile2
-rw-r--r--tools/perf/util/include/linux/rbtree.h1
-rw-r--r--tools/power/acpi/Makefile18
-rw-r--r--tools/power/acpi/acpidump.859
-rw-r--r--tools/power/acpi/acpidump.c560
-rw-r--r--tools/power/cpupower/Makefile2
-rw-r--r--tools/power/x86/turbostat/turbostat.855
-rw-r--r--tools/power/x86/turbostat/turbostat.c214
-rw-r--r--tools/virtio/virtio-trace/Makefile13
-rw-r--r--tools/virtio/virtio-trace/README118
-rw-r--r--tools/virtio/virtio-trace/trace-agent-ctl.c137
-rw-r--r--tools/virtio/virtio-trace/trace-agent-rw.c192
-rw-r--r--tools/virtio/virtio-trace/trace-agent.c270
-rw-r--r--tools/virtio/virtio-trace/trace-agent.h75
15 files changed, 1679 insertions, 38 deletions
diff --git a/tools/lguest/lguest.c b/tools/lguest/lguest.c
index f759f4f097c7..fd2f9221b241 100644
--- a/tools/lguest/lguest.c
+++ b/tools/lguest/lguest.c
@@ -1299,6 +1299,7 @@ static struct device *new_device(const char *name, u16 type)
1299 dev->feature_len = 0; 1299 dev->feature_len = 0;
1300 dev->num_vq = 0; 1300 dev->num_vq = 0;
1301 dev->running = false; 1301 dev->running = false;
1302 dev->next = NULL;
1302 1303
1303 /* 1304 /*
1304 * Append to device list. Prepending to a single-linked list is 1305 * Append to device list. Prepending to a single-linked list is
diff --git a/tools/perf/Makefile b/tools/perf/Makefile
index e5e71e7d95a0..86258c2a2c23 100644
--- a/tools/perf/Makefile
+++ b/tools/perf/Makefile
@@ -72,7 +72,7 @@ ifeq ($(ARCH),x86_64)
72 override ARCH := x86 72 override ARCH := x86
73 IS_X86_64 := 0 73 IS_X86_64 := 0
74 ifeq (, $(findstring m32,$(EXTRA_CFLAGS))) 74 ifeq (, $(findstring m32,$(EXTRA_CFLAGS)))
75 IS_X86_64 := $(shell echo __x86_64__ | ${CC} -E -xc - | tail -n 1) 75 IS_X86_64 := $(shell echo __x86_64__ | ${CC} -E -x c - | tail -n 1)
76 endif 76 endif
77 ifeq (${IS_X86_64}, 1) 77 ifeq (${IS_X86_64}, 1)
78 RAW_ARCH := x86_64 78 RAW_ARCH := x86_64
diff --git a/tools/perf/util/include/linux/rbtree.h b/tools/perf/util/include/linux/rbtree.h
index 2a030c5af3aa..9bcdc844b330 100644
--- a/tools/perf/util/include/linux/rbtree.h
+++ b/tools/perf/util/include/linux/rbtree.h
@@ -1,2 +1,3 @@
1#include <stdbool.h> 1#include <stdbool.h>
2#include <stdbool.h>
2#include "../../../../include/linux/rbtree.h" 3#include "../../../../include/linux/rbtree.h"
diff --git a/tools/power/acpi/Makefile b/tools/power/acpi/Makefile
new file mode 100644
index 000000000000..6b9cf7a987c7
--- /dev/null
+++ b/tools/power/acpi/Makefile
@@ -0,0 +1,18 @@
1PROG= acpidump
2SRCS= acpidump.c
3KERNEL_INCLUDE := ../../../include
4CFLAGS += -Wall -Wstrict-prototypes -Wdeclaration-after-statement -Os -s -D_LINUX -DDEFINE_ALTERNATE_TYPES -I$(KERNEL_INCLUDE)
5
6all: acpidump
7$(PROG) : $(SRCS)
8 $(CC) $(CFLAGS) $(SRCS) -o $(PROG)
9
10CLEANFILES= $(PROG)
11
12clean :
13 rm -f $(CLEANFILES) $(patsubst %.c,%.o, $(SRCS)) *~
14
15install :
16 install acpidump /usr/bin/acpidump
17 install acpidump.8 /usr/share/man/man8
18
diff --git a/tools/power/acpi/acpidump.8 b/tools/power/acpi/acpidump.8
new file mode 100644
index 000000000000..adfa99166e5e
--- /dev/null
+++ b/tools/power/acpi/acpidump.8
@@ -0,0 +1,59 @@
1.TH ACPIDUMP 8
2.SH NAME
3acpidump \- Dump system's ACPI tables to an ASCII file.
4.SH SYNOPSIS
5.ft B
6.B acpidump > acpidump.out
7.SH DESCRIPTION
8\fBacpidump \fP dumps the systems ACPI tables to an ASCII file
9appropriate for attaching to a bug report.
10
11Subsequently, they can be processed by utilities in the ACPICA package.
12.SS Options
13no options worth worrying about.
14.PP
15.SH EXAMPLE
16
17.nf
18# acpidump > acpidump.out
19
20$ acpixtract -a acpidump.out
21 Acpi table [DSDT] - 15974 bytes written to DSDT.dat
22 Acpi table [FACS] - 64 bytes written to FACS.dat
23 Acpi table [FACP] - 116 bytes written to FACP.dat
24 Acpi table [APIC] - 120 bytes written to APIC.dat
25 Acpi table [MCFG] - 60 bytes written to MCFG.dat
26 Acpi table [SSDT] - 444 bytes written to SSDT1.dat
27 Acpi table [SSDT] - 439 bytes written to SSDT2.dat
28 Acpi table [SSDT] - 439 bytes written to SSDT3.dat
29 Acpi table [SSDT] - 439 bytes written to SSDT4.dat
30 Acpi table [SSDT] - 439 bytes written to SSDT5.dat
31 Acpi table [RSDT] - 76 bytes written to RSDT.dat
32 Acpi table [RSDP] - 20 bytes written to RSDP.dat
33
34$ iasl -d *.dat
35...
36.fi
37creates *.dsl, a human readable form which can be edited
38and compiled using iasl.
39
40
41.SH NOTES
42
43.B "acpidump "
44must be run as root.
45
46.SH REFERENCES
47ACPICA: https://acpica.org/
48
49.SH FILES
50.ta
51.nf
52/dev/mem
53/sys/firmware/acpi/tables/dynamic/*
54.fi
55
56.PP
57.SH AUTHOR
58.nf
59Written by Len Brown <len.brown@intel.com>
diff --git a/tools/power/acpi/acpidump.c b/tools/power/acpi/acpidump.c
new file mode 100644
index 000000000000..07779871421c
--- /dev/null
+++ b/tools/power/acpi/acpidump.c
@@ -0,0 +1,560 @@
1/*
2 * (c) Alexey Starikovskiy, Intel, 2005-2006.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions, and the following disclaimer,
10 * without modification.
11 * 2. Redistributions in binary form must reproduce at minimum a disclaimer
12 * substantially similar to the "NO WARRANTY" disclaimer below
13 * ("Disclaimer") and any redistribution must be conditioned upon
14 * including a substantially similar Disclaimer requirement for further
15 * binary redistribution.
16 * 3. Neither the names of the above-listed copyright holders nor the names
17 * of any contributors may be used to endorse or promote products derived
18 * from this software without specific prior written permission.
19 *
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
23 *
24 * NO WARRANTY
25 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
26 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
27 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR
28 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
29 * HOLDERS OR CONTRIBUTORS BE LIABLE FOR SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
33 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
34 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGES.
36 */
37
38#ifdef DEFINE_ALTERNATE_TYPES
39/* hack to enable building old application with new headers -lenb */
40#define acpi_fadt_descriptor acpi_table_fadt
41#define acpi_rsdp_descriptor acpi_table_rsdp
42#define DSDT_SIG ACPI_SIG_DSDT
43#define FACS_SIG ACPI_SIG_FACS
44#define FADT_SIG ACPI_SIG_FADT
45#define xfirmware_ctrl Xfacs
46#define firmware_ctrl facs
47
48typedef int s32;
49typedef unsigned char u8;
50typedef unsigned short u16;
51typedef unsigned int u32;
52typedef unsigned long long u64;
53typedef long long s64;
54#endif
55
56#include <sys/mman.h>
57#include <sys/types.h>
58#include <sys/stat.h>
59#include <fcntl.h>
60#include <stdio.h>
61#include <string.h>
62#include <unistd.h>
63#include <getopt.h>
64
65#include <sys/types.h>
66#include <dirent.h>
67
68#include <acpi/acconfig.h>
69#include <acpi/platform/acenv.h>
70#include <acpi/actypes.h>
71#include <acpi/actbl.h>
72
73static inline u8 checksum(u8 * buffer, u32 length)
74{
75 u8 sum = 0, *i = buffer;
76 buffer += length;
77 for (; i < buffer; sum += *(i++));
78 return sum;
79}
80
81static unsigned long psz, addr, length;
82static int print, connect, skip;
83static u8 select_sig[4];
84
85static unsigned long read_efi_systab( void )
86{
87 char buffer[80];
88 unsigned long addr;
89 FILE *f = fopen("/sys/firmware/efi/systab", "r");
90 if (f) {
91 while (fgets(buffer, 80, f)) {
92 if (sscanf(buffer, "ACPI20=0x%lx", &addr) == 1)
93 return addr;
94 }
95 fclose(f);
96 }
97 return 0;
98}
99
100static u8 *acpi_map_memory(unsigned long where, unsigned length)
101{
102 unsigned long offset;
103 u8 *there;
104 int fd = open("/dev/mem", O_RDONLY);
105 if (fd < 0) {
106 fprintf(stderr, "acpi_os_map_memory: cannot open /dev/mem\n");
107 exit(1);
108 }
109 offset = where % psz;
110 there = mmap(NULL, length + offset, PROT_READ, MAP_PRIVATE,
111 fd, where - offset);
112 close(fd);
113 if (there == MAP_FAILED) return 0;
114 return (there + offset);
115}
116
117static void acpi_unmap_memory(u8 * there, unsigned length)
118{
119 unsigned long offset = (unsigned long)there % psz;
120 munmap(there - offset, length + offset);
121}
122
123static struct acpi_table_header *acpi_map_table(unsigned long where, char *sig)
124{
125 unsigned size;
126 struct acpi_table_header *tbl = (struct acpi_table_header *)
127 acpi_map_memory(where, sizeof(struct acpi_table_header));
128 if (!tbl || (sig && memcmp(sig, tbl->signature, 4))) return 0;
129 size = tbl->length;
130 acpi_unmap_memory((u8 *) tbl, sizeof(struct acpi_table_header));
131 return (struct acpi_table_header *)acpi_map_memory(where, size);
132}
133
134static void acpi_unmap_table(struct acpi_table_header *tbl)
135{
136 acpi_unmap_memory((u8 *)tbl, tbl->length);
137}
138
139static struct acpi_rsdp_descriptor *acpi_scan_for_rsdp(u8 *begin, u32 length)
140{
141 struct acpi_rsdp_descriptor *rsdp;
142 u8 *i, *end = begin + length;
143 /* Search from given start address for the requested length */
144 for (i = begin; i < end; i += ACPI_RSDP_SCAN_STEP) {
145 /* The signature and checksum must both be correct */
146 if (memcmp((char *)i, "RSD PTR ", 8)) continue;
147 rsdp = (struct acpi_rsdp_descriptor *)i;
148 /* Signature matches, check the appropriate checksum */
149 if (!checksum((u8 *) rsdp, (rsdp->revision < 2) ?
150 ACPI_RSDP_CHECKSUM_LENGTH :
151 ACPI_RSDP_XCHECKSUM_LENGTH))
152 /* Checksum valid, we have found a valid RSDP */
153 return rsdp;
154 }
155 /* Searched entire block, no RSDP was found */
156 return 0;
157}
158
159/*
160 * Output data
161 */
162static void acpi_show_data(int fd, u8 * data, int size)
163{
164 char buffer[256];
165 int len;
166 int i, remain = size;
167 while (remain > 0) {
168 len = snprintf(buffer, 256, " %04x:", size - remain);
169 for (i = 0; i < 16 && i < remain; i++) {
170 len +=
171 snprintf(&buffer[len], 256 - len, " %02x", data[i]);
172 }
173 for (; i < 16; i++) {
174 len += snprintf(&buffer[len], 256 - len, " ");
175 }
176 len += snprintf(&buffer[len], 256 - len, " ");
177 for (i = 0; i < 16 && i < remain; i++) {
178 buffer[len++] = (isprint(data[i])) ? data[i] : '.';
179 }
180 buffer[len++] = '\n';
181 write(fd, buffer, len);
182 data += 16;
183 remain -= 16;
184 }
185}
186
187/*
188 * Output ACPI table
189 */
190static void acpi_show_table(int fd, struct acpi_table_header *table, unsigned long addr)
191{
192 char buff[80];
193 int len = snprintf(buff, 80, "%.4s @ %p\n", table->signature, (void *)addr);
194 write(fd, buff, len);
195 acpi_show_data(fd, (u8 *) table, table->length);
196 buff[0] = '\n';
197 write(fd, buff, 1);
198}
199
200static void write_table(int fd, struct acpi_table_header *tbl, unsigned long addr)
201{
202 static int select_done = 0;
203 if (!select_sig[0]) {
204 if (print) {
205 acpi_show_table(fd, tbl, addr);
206 } else {
207 write(fd, tbl, tbl->length);
208 }
209 } else if (!select_done && !memcmp(select_sig, tbl->signature, 4)) {
210 if (skip > 0) {
211 --skip;
212 return;
213 }
214 if (print) {
215 acpi_show_table(fd, tbl, addr);
216 } else {
217 write(fd, tbl, tbl->length);
218 }
219 select_done = 1;
220 }
221}
222
223static void acpi_dump_FADT(int fd, struct acpi_table_header *tbl, unsigned long xaddr) {
224 struct acpi_fadt_descriptor x;
225 unsigned long addr;
226 size_t len = sizeof(struct acpi_fadt_descriptor);
227 if (len > tbl->length) len = tbl->length;
228 memcpy(&x, tbl, len);
229 x.header.length = len;
230 if (checksum((u8 *)tbl, len)) {
231 fprintf(stderr, "Wrong checksum for FADT!\n");
232 }
233 if (x.header.length >= 148 && x.Xdsdt) {
234 addr = (unsigned long)x.Xdsdt;
235 if (connect) {
236 x.Xdsdt = lseek(fd, 0, SEEK_CUR);
237 }
238 } else if (x.header.length >= 44 && x.dsdt) {
239 addr = (unsigned long)x.dsdt;
240 if (connect) {
241 x.dsdt = lseek(fd, 0, SEEK_CUR);
242 }
243 } else {
244 fprintf(stderr, "No DSDT in FADT!\n");
245 goto no_dsdt;
246 }
247 tbl = acpi_map_table(addr, DSDT_SIG);
248 if (!tbl) goto no_dsdt;
249 if (checksum((u8 *)tbl, tbl->length))
250 fprintf(stderr, "Wrong checksum for DSDT!\n");
251 write_table(fd, tbl, addr);
252 acpi_unmap_table(tbl);
253no_dsdt:
254 if (x.header.length >= 140 && x.xfirmware_ctrl) {
255 addr = (unsigned long)x.xfirmware_ctrl;
256 if (connect) {
257 x.xfirmware_ctrl = lseek(fd, 0, SEEK_CUR);
258 }
259 } else if (x.header.length >= 40 && x.firmware_ctrl) {
260 addr = (unsigned long)x.firmware_ctrl;
261 if (connect) {
262 x.firmware_ctrl = lseek(fd, 0, SEEK_CUR);
263 }
264 } else {
265 fprintf(stderr, "No FACS in FADT!\n");
266 goto no_facs;
267 }
268 tbl = acpi_map_table(addr, FACS_SIG);
269 if (!tbl) goto no_facs;
270 /* do not checksum FACS */
271 write_table(fd, tbl, addr);
272 acpi_unmap_table(tbl);
273no_facs:
274 write_table(fd, (struct acpi_table_header *)&x, xaddr);
275}
276
277static int acpi_dump_SDT(int fd, struct acpi_rsdp_descriptor *rsdp)
278{
279 struct acpi_table_header *sdt, *tbl = 0;
280 int xsdt = 1, i, num;
281 char *offset;
282 unsigned long addr;
283 if (rsdp->revision > 1 && rsdp->xsdt_physical_address) {
284 tbl = acpi_map_table(rsdp->xsdt_physical_address, "XSDT");
285 }
286 if (!tbl && rsdp->rsdt_physical_address) {
287 xsdt = 0;
288 tbl = acpi_map_table(rsdp->rsdt_physical_address, "RSDT");
289 }
290 if (!tbl) return 0;
291 sdt = malloc(tbl->length);
292 memcpy(sdt, tbl, tbl->length);
293 acpi_unmap_table(tbl);
294 if (checksum((u8 *)sdt, sdt->length))
295 fprintf(stderr, "Wrong checksum for %s!\n", (xsdt)?"XSDT":"RSDT");
296 num = (sdt->length - sizeof(struct acpi_table_header))/((xsdt)?sizeof(u64):sizeof(u32));
297 offset = (char *)sdt + sizeof(struct acpi_table_header);
298 for (i = 0; i < num; ++i, offset += ((xsdt) ? sizeof(u64) : sizeof(u32))) {
299 addr = (xsdt) ? (unsigned long)(*(u64 *)offset):
300 (unsigned long)(*(u32 *)offset);
301 if (!addr) continue;
302 tbl = acpi_map_table(addr, 0);
303 if (!tbl) continue;
304 if (!memcmp(tbl->signature, FADT_SIG, 4)) {
305 acpi_dump_FADT(fd, tbl, addr);
306 } else {
307 if (checksum((u8 *)tbl, tbl->length))
308 fprintf(stderr, "Wrong checksum for generic table!\n");
309 write_table(fd, tbl, addr);
310 }
311 acpi_unmap_table(tbl);
312 if (connect) {
313 if (xsdt)
314 (*(u64*)offset) = lseek(fd, 0, SEEK_CUR);
315 else
316 (*(u32*)offset) = lseek(fd, 0, SEEK_CUR);
317 }
318 }
319 if (xsdt) {
320 addr = (unsigned long)rsdp->xsdt_physical_address;
321 if (connect) {
322 rsdp->xsdt_physical_address = lseek(fd, 0, SEEK_CUR);
323 }
324 } else {
325 addr = (unsigned long)rsdp->rsdt_physical_address;
326 if (connect) {
327 rsdp->rsdt_physical_address = lseek(fd, 0, SEEK_CUR);
328 }
329 }
330 write_table(fd, sdt, addr);
331 free (sdt);
332 return 1;
333}
334
335#define DYNAMIC_SSDT "/sys/firmware/acpi/tables/dynamic"
336
337static void acpi_dump_dynamic_SSDT(int fd)
338{
339 struct stat file_stat;
340 char filename[256], *ptr;
341 DIR *tabledir;
342 struct dirent *entry;
343 FILE *fp;
344 int count, readcount, length;
345 struct acpi_table_header table_header, *ptable;
346
347 if (stat(DYNAMIC_SSDT, &file_stat) == -1) {
348 /* The directory doesn't exist */
349 return;
350 }
351 tabledir = opendir(DYNAMIC_SSDT);
352 if(!tabledir){
353 /*can't open the directory */
354 return;
355 }
356
357 while ((entry = readdir(tabledir)) != 0){
358 /* skip the file of . /.. */
359 if (entry->d_name[0] == '.')
360 continue;
361
362 sprintf(filename, "%s/%s", DYNAMIC_SSDT, entry->d_name);
363 fp = fopen(filename, "r");
364 if (fp == NULL) {
365 fprintf(stderr, "Can't open the file of %s\n",
366 filename);
367 continue;
368 }
369 /* Read the Table header to parse the table length */
370 count = fread(&table_header, 1, sizeof(struct acpi_table_header), fp);
371 if (count < sizeof(table_header)) {
372 /* the length is lessn than ACPI table header. skip it */
373 fclose(fp);
374 continue;
375 }
376 length = table_header.length;
377 ptr = malloc(table_header.length);
378 fseek(fp, 0, SEEK_SET);
379 readcount = 0;
380 while(!feof(fp) && readcount < length) {
381 count = fread(ptr + readcount, 1, 256, fp);
382 readcount += count;
383 }
384 fclose(fp);
385 ptable = (struct acpi_table_header *) ptr;
386 if (checksum((u8 *) ptable, ptable->length))
387 fprintf(stderr, "Wrong checksum "
388 "for dynamic SSDT table!\n");
389 write_table(fd, ptable, 0);
390 free(ptr);
391 }
392 closedir(tabledir);
393 return;
394}
395
396static void usage(const char *progname)
397{
398 puts("Usage:");
399 printf("%s [--addr 0x1234][--table DSDT][--output filename]"
400 "[--binary][--length 0x456][--help]\n", progname);
401 puts("\t--addr 0x1234 or -a 0x1234 -- look for tables at this physical address");
402 puts("\t--table DSDT or -t DSDT -- only dump table with DSDT signature");
403 puts("\t--output filename or -o filename -- redirect output from stdin to filename");
404 puts("\t--binary or -b -- dump data in binary form rather than in hex-dump format");
405 puts("\t--length 0x456 or -l 0x456 -- works only with --addr, dump physical memory"
406 "\n\t\tregion without trying to understand it's contents");
407 puts("\t--skip 2 or -s 2 -- skip 2 tables of the given name and output only 3rd one");
408 puts("\t--help or -h -- this help message");
409 exit(0);
410}
411
412static struct option long_options[] = {
413 {"addr", 1, 0, 0},
414 {"table", 1, 0, 0},
415 {"output", 1, 0, 0},
416 {"binary", 0, 0, 0},
417 {"length", 1, 0, 0},
418 {"skip", 1, 0, 0},
419 {"help", 0, 0, 0},
420 {0, 0, 0, 0}
421};
422int main(int argc, char **argv)
423{
424 int option_index, c, fd;
425 u8 *raw;
426 struct acpi_rsdp_descriptor rsdpx, *x = 0;
427 char *filename = 0;
428 char buff[80];
429 memset(select_sig, 0, 4);
430 print = 1;
431 connect = 0;
432 addr = length = 0;
433 skip = 0;
434 while (1) {
435 option_index = 0;
436 c = getopt_long(argc, argv, "a:t:o:bl:s:h",
437 long_options, &option_index);
438 if (c == -1)
439 break;
440
441 switch (c) {
442 case 0:
443 switch (option_index) {
444 case 0:
445 addr = strtoul(optarg, (char **)NULL, 16);
446 break;
447 case 1:
448 memcpy(select_sig, optarg, 4);
449 break;
450 case 2:
451 filename = optarg;
452 break;
453 case 3:
454 print = 0;
455 break;
456 case 4:
457 length = strtoul(optarg, (char **)NULL, 16);
458 break;
459 case 5:
460 skip = strtoul(optarg, (char **)NULL, 10);
461 break;
462 case 6:
463 usage(argv[0]);
464 exit(0);
465 }
466 break;
467 case 'a':
468 addr = strtoul(optarg, (char **)NULL, 16);
469 break;
470 case 't':
471 memcpy(select_sig, optarg, 4);
472 break;
473 case 'o':
474 filename = optarg;
475 break;
476 case 'b':
477 print = 0;
478 break;
479 case 'l':
480 length = strtoul(optarg, (char **)NULL, 16);
481 break;
482 case 's':
483 skip = strtoul(optarg, (char **)NULL, 10);
484 break;
485 case 'h':
486 usage(argv[0]);
487 exit(0);
488 default:
489 printf("Unknown option!\n");
490 usage(argv[0]);
491 exit(0);
492 }
493 }
494
495 fd = STDOUT_FILENO;
496 if (filename) {
497 fd = creat(filename, S_IRUSR|S_IWUSR|S_IRGRP|S_IROTH);
498 if (fd < 0)
499 return fd;
500 }
501
502 if (!select_sig[0] && !print) {
503 connect = 1;
504 }
505
506 psz = sysconf(_SC_PAGESIZE);
507 if (length && addr) {
508 /* We know length and address, it means we just want a memory dump */
509 if (!(raw = acpi_map_memory(addr, length)))
510 goto not_found;
511 write(fd, raw, length);
512 acpi_unmap_memory(raw, length);
513 close(fd);
514 return 0;
515 }
516
517 length = sizeof(struct acpi_rsdp_descriptor);
518 if (!addr) {
519 addr = read_efi_systab();
520 if (!addr) {
521 addr = ACPI_HI_RSDP_WINDOW_BASE;
522 length = ACPI_HI_RSDP_WINDOW_SIZE;
523 }
524 }
525
526 if (!(raw = acpi_map_memory(addr, length)) ||
527 !(x = acpi_scan_for_rsdp(raw, length)))
528 goto not_found;
529
530 /* Find RSDP and print all found tables */
531 memcpy(&rsdpx, x, sizeof(struct acpi_rsdp_descriptor));
532 acpi_unmap_memory(raw, length);
533 if (connect) {
534 lseek(fd, sizeof(struct acpi_rsdp_descriptor), SEEK_SET);
535 }
536 if (!acpi_dump_SDT(fd, &rsdpx))
537 goto not_found;
538 if (connect) {
539 lseek(fd, 0, SEEK_SET);
540 write(fd, x, (rsdpx.revision < 2) ?
541 ACPI_RSDP_CHECKSUM_LENGTH : ACPI_RSDP_XCHECKSUM_LENGTH);
542 } else if (!select_sig[0] || !memcmp("RSD PTR ", select_sig, 4)) {
543 addr += (long)x - (long)raw;
544 length = snprintf(buff, 80, "RSD PTR @ %p\n", (void *)addr);
545 write(fd, buff, length);
546 acpi_show_data(fd, (u8 *) & rsdpx, (rsdpx.revision < 2) ?
547 ACPI_RSDP_CHECKSUM_LENGTH : ACPI_RSDP_XCHECKSUM_LENGTH);
548 buff[0] = '\n';
549 write(fd, buff, 1);
550 }
551 acpi_dump_dynamic_SSDT(fd);
552 close(fd);
553 return 0;
554not_found:
555 close(fd);
556 fprintf(stderr, "ACPI tables were not found. If you know location "
557 "of RSD PTR table (from dmesg, etc), "
558 "supply it with either --addr or -a option\n");
559 return 1;
560}
diff --git a/tools/power/cpupower/Makefile b/tools/power/cpupower/Makefile
index a93e06cfcc2a..cf397bd26d0c 100644
--- a/tools/power/cpupower/Makefile
+++ b/tools/power/cpupower/Makefile
@@ -111,7 +111,7 @@ GMO_FILES = ${shell for HLANG in ${LANGUAGES}; do echo $(OUTPUT)po/$$HLANG.gmo;
111export CROSS CC AR STRIP RANLIB CFLAGS LDFLAGS LIB_OBJS 111export CROSS CC AR STRIP RANLIB CFLAGS LDFLAGS LIB_OBJS
112 112
113# check if compiler option is supported 113# check if compiler option is supported
114cc-supports = ${shell if $(CC) ${1} -S -o /dev/null -xc /dev/null > /dev/null 2>&1; then echo "$(1)"; fi;} 114cc-supports = ${shell if $(CC) ${1} -S -o /dev/null -x c /dev/null > /dev/null 2>&1; then echo "$(1)"; fi;}
115 115
116# use '-Os' optimization if available, else use -O2 116# use '-Os' optimization if available, else use -O2
117OPTIMIZATION := $(call cc-supports,-Os,-O2) 117OPTIMIZATION := $(call cc-supports,-Os,-O2)
diff --git a/tools/power/x86/turbostat/turbostat.8 b/tools/power/x86/turbostat/turbostat.8
index 74e44507dfe9..e4d0690cccf9 100644
--- a/tools/power/x86/turbostat/turbostat.8
+++ b/tools/power/x86/turbostat/turbostat.8
@@ -4,15 +4,11 @@ turbostat \- Report processor frequency and idle statistics
4.SH SYNOPSIS 4.SH SYNOPSIS
5.ft B 5.ft B
6.B turbostat 6.B turbostat
7.RB [ "\-s" ] 7.RB [ Options ]
8.RB [ "\-v" ]
9.RB [ "\-M MSR#" ]
10.RB command 8.RB command
11.br 9.br
12.B turbostat 10.B turbostat
13.RB [ "\-s" ] 11.RB [ Options ]
14.RB [ "\-v" ]
15.RB [ "\-M MSR#" ]
16.RB [ "\-i interval_sec" ] 12.RB [ "\-i interval_sec" ]
17.SH DESCRIPTION 13.SH DESCRIPTION
18\fBturbostat \fP reports processor topology, frequency 14\fBturbostat \fP reports processor topology, frequency
@@ -27,16 +23,23 @@ supports an "invariant" TSC, plus the APERF and MPERF MSRs.
27on processors that additionally support C-state residency counters. 23on processors that additionally support C-state residency counters.
28 24
29.SS Options 25.SS Options
30The \fB-s\fP option limits output to a 1-line system summary for each interval. 26The \fB-p\fP option limits output to the 1st thread in 1st core of each package.
31.PP 27.PP
32The \fB-c\fP option limits output to the 1st thread in each core. 28The \fB-P\fP option limits output to the 1st thread in each Package.
33.PP 29.PP
34The \fB-p\fP option limits output to the 1st thread in each package. 30The \fB-S\fP option limits output to a 1-line System Summary for each interval.
35.PP 31.PP
36The \fB-v\fP option increases verbosity. 32The \fB-v\fP option increases verbosity.
37.PP 33.PP
38The \fB-M MSR#\fP option dumps the specified MSR, 34The \fB-s\fP option prints the SMI counter, equivalent to "-c 0x34"
39in addition to the usual frequency and idle statistics. 35.PP
36The \fB-c MSR#\fP option includes the delta of the specified 32-bit MSR counter.
37.PP
38The \fB-C MSR#\fP option includes the delta of the specified 64-bit MSR counter.
39.PP
40The \fB-m MSR#\fP option includes the the specified 32-bit MSR value.
41.PP
42The \fB-M MSR#\fP option includes the the specified 64-bit MSR value.
40.PP 43.PP
41The \fB-i interval_sec\fP option prints statistics every \fiinterval_sec\fP seconds. 44The \fB-i interval_sec\fP option prints statistics every \fiinterval_sec\fP seconds.
42The default is 5 seconds. 45The default is 5 seconds.
@@ -150,6 +153,29 @@ Note that turbostat reports average GHz of 3.63, while
150the arithmetic average of the GHz column above is lower. 153the arithmetic average of the GHz column above is lower.
151This is a weighted average, where the weight is %c0. ie. it is the total number of 154This is a weighted average, where the weight is %c0. ie. it is the total number of
152un-halted cycles elapsed per time divided by the number of CPUs. 155un-halted cycles elapsed per time divided by the number of CPUs.
156.SH SMI COUNTING EXAMPLE
157On Intel Nehalem and newer processors, MSR 0x34 is a System Management Mode Interrupt (SMI) counter.
158Using the -m option, you can display how many SMIs have fired since reset, or if there
159are SMIs during the measurement interval, you can display the delta using the -d option.
160.nf
161[root@x980 ~]# turbostat -m 0x34
162cor CPU %c0 GHz TSC MSR 0x034 %c1 %c3 %c6 %pc3 %pc6
163 1.41 1.82 3.38 0x00000000 8.92 37.82 51.85 17.37 0.55
164 0 0 3.73 2.03 3.38 0x00000055 1.72 48.25 46.31 17.38 0.55
165 0 6 0.14 1.63 3.38 0x00000056 5.30
166 1 2 2.51 1.80 3.38 0x00000056 15.65 29.33 52.52
167 1 8 0.10 1.65 3.38 0x00000056 18.05
168 2 4 1.16 1.68 3.38 0x00000056 5.87 24.47 68.50
169 2 10 0.10 1.63 3.38 0x00000056 6.93
170 8 1 3.84 1.91 3.38 0x00000056 1.36 50.65 44.16
171 8 7 0.08 1.64 3.38 0x00000056 5.12
172 9 3 1.82 1.73 3.38 0x00000056 7.59 24.21 66.38
173 9 9 0.09 1.68 3.38 0x00000056 9.32
174 10 5 1.66 1.65 3.38 0x00000056 15.10 50.00 33.23
175 10 11 1.72 1.65 3.38 0x00000056 15.05
176^C
177[root@x980 ~]#
178.fi
153.SH NOTES 179.SH NOTES
154 180
155.B "turbostat " 181.B "turbostat "
@@ -165,6 +191,13 @@ may work poorly on Linux-2.6.20 through 2.6.29,
165as \fBacpi-cpufreq \fPperiodically cleared the APERF and MPERF 191as \fBacpi-cpufreq \fPperiodically cleared the APERF and MPERF
166in those kernels. 192in those kernels.
167 193
194If the TSC column does not make sense, then
195the other numbers will also make no sense.
196Turbostat is lightweight, and its data collection is not atomic.
197These issues are usually caused by an extremely short measurement
198interval (much less than 1 second), or system activity that prevents
199turbostat from being able to run on all CPUS to quickly collect data.
200
168The APERF, MPERF MSRs are defined to count non-halted cycles. 201The APERF, MPERF MSRs are defined to count non-halted cycles.
169Although it is not guaranteed by the architecture, turbostat assumes 202Although it is not guaranteed by the architecture, turbostat assumes
170that they count at TSC rate, which is true on all processors tested to date. 203that they count at TSC rate, which is true on all processors tested to date.
diff --git a/tools/power/x86/turbostat/turbostat.c b/tools/power/x86/turbostat/turbostat.c
index 861d77190206..2655ae9a3ad8 100644
--- a/tools/power/x86/turbostat/turbostat.c
+++ b/tools/power/x86/turbostat/turbostat.c
@@ -35,9 +35,9 @@
35#include <ctype.h> 35#include <ctype.h>
36#include <sched.h> 36#include <sched.h>
37 37
38#define MSR_TSC 0x10
39#define MSR_NEHALEM_PLATFORM_INFO 0xCE 38#define MSR_NEHALEM_PLATFORM_INFO 0xCE
40#define MSR_NEHALEM_TURBO_RATIO_LIMIT 0x1AD 39#define MSR_NEHALEM_TURBO_RATIO_LIMIT 0x1AD
40#define MSR_IVT_TURBO_RATIO_LIMIT 0x1AE
41#define MSR_APERF 0xE8 41#define MSR_APERF 0xE8
42#define MSR_MPERF 0xE7 42#define MSR_MPERF 0xE7
43#define MSR_PKG_C2_RESIDENCY 0x60D /* SNB only */ 43#define MSR_PKG_C2_RESIDENCY 0x60D /* SNB only */
@@ -62,7 +62,11 @@ unsigned int genuine_intel;
62unsigned int has_invariant_tsc; 62unsigned int has_invariant_tsc;
63unsigned int do_nehalem_platform_info; 63unsigned int do_nehalem_platform_info;
64unsigned int do_nehalem_turbo_ratio_limit; 64unsigned int do_nehalem_turbo_ratio_limit;
65unsigned int extra_msr_offset; 65unsigned int do_ivt_turbo_ratio_limit;
66unsigned int extra_msr_offset32;
67unsigned int extra_msr_offset64;
68unsigned int extra_delta_offset32;
69unsigned int extra_delta_offset64;
66double bclk; 70double bclk;
67unsigned int show_pkg; 71unsigned int show_pkg;
68unsigned int show_core; 72unsigned int show_core;
@@ -83,7 +87,10 @@ struct thread_data {
83 unsigned long long aperf; 87 unsigned long long aperf;
84 unsigned long long mperf; 88 unsigned long long mperf;
85 unsigned long long c1; /* derived */ 89 unsigned long long c1; /* derived */
86 unsigned long long extra_msr; 90 unsigned long long extra_msr64;
91 unsigned long long extra_delta64;
92 unsigned long long extra_msr32;
93 unsigned long long extra_delta32;
87 unsigned int cpu_id; 94 unsigned int cpu_id;
88 unsigned int flags; 95 unsigned int flags;
89#define CPU_IS_FIRST_THREAD_IN_CORE 0x2 96#define CPU_IS_FIRST_THREAD_IN_CORE 0x2
@@ -222,6 +229,14 @@ void print_header(void)
222 if (has_aperf) 229 if (has_aperf)
223 outp += sprintf(outp, " GHz"); 230 outp += sprintf(outp, " GHz");
224 outp += sprintf(outp, " TSC"); 231 outp += sprintf(outp, " TSC");
232 if (extra_delta_offset32)
233 outp += sprintf(outp, " count 0x%03X", extra_delta_offset32);
234 if (extra_delta_offset64)
235 outp += sprintf(outp, " COUNT 0x%03X", extra_delta_offset64);
236 if (extra_msr_offset32)
237 outp += sprintf(outp, " MSR 0x%03X", extra_msr_offset32);
238 if (extra_msr_offset64)
239 outp += sprintf(outp, " MSR 0x%03X", extra_msr_offset64);
225 if (do_nhm_cstates) 240 if (do_nhm_cstates)
226 outp += sprintf(outp, " %%c1"); 241 outp += sprintf(outp, " %%c1");
227 if (do_nhm_cstates) 242 if (do_nhm_cstates)
@@ -238,8 +253,6 @@ void print_header(void)
238 outp += sprintf(outp, " %%pc6"); 253 outp += sprintf(outp, " %%pc6");
239 if (do_snb_cstates) 254 if (do_snb_cstates)
240 outp += sprintf(outp, " %%pc7"); 255 outp += sprintf(outp, " %%pc7");
241 if (extra_msr_offset)
242 outp += sprintf(outp, " MSR 0x%x ", extra_msr_offset);
243 256
244 outp += sprintf(outp, "\n"); 257 outp += sprintf(outp, "\n");
245} 258}
@@ -255,8 +268,14 @@ int dump_counters(struct thread_data *t, struct core_data *c,
255 fprintf(stderr, "aperf: %016llX\n", t->aperf); 268 fprintf(stderr, "aperf: %016llX\n", t->aperf);
256 fprintf(stderr, "mperf: %016llX\n", t->mperf); 269 fprintf(stderr, "mperf: %016llX\n", t->mperf);
257 fprintf(stderr, "c1: %016llX\n", t->c1); 270 fprintf(stderr, "c1: %016llX\n", t->c1);
271 fprintf(stderr, "msr0x%x: %08llX\n",
272 extra_delta_offset32, t->extra_delta32);
258 fprintf(stderr, "msr0x%x: %016llX\n", 273 fprintf(stderr, "msr0x%x: %016llX\n",
259 extra_msr_offset, t->extra_msr); 274 extra_delta_offset64, t->extra_delta64);
275 fprintf(stderr, "msr0x%x: %08llX\n",
276 extra_msr_offset32, t->extra_msr32);
277 fprintf(stderr, "msr0x%x: %016llX\n",
278 extra_msr_offset64, t->extra_msr64);
260 } 279 }
261 280
262 if (c) { 281 if (c) {
@@ -360,6 +379,21 @@ int format_counters(struct thread_data *t, struct core_data *c,
360 /* TSC */ 379 /* TSC */
361 outp += sprintf(outp, "%5.2f", 1.0 * t->tsc/units/interval_float); 380 outp += sprintf(outp, "%5.2f", 1.0 * t->tsc/units/interval_float);
362 381
382 /* delta */
383 if (extra_delta_offset32)
384 outp += sprintf(outp, " %11llu", t->extra_delta32);
385
386 /* DELTA */
387 if (extra_delta_offset64)
388 outp += sprintf(outp, " %11llu", t->extra_delta64);
389 /* msr */
390 if (extra_msr_offset32)
391 outp += sprintf(outp, " 0x%08llx", t->extra_msr32);
392
393 /* MSR */
394 if (extra_msr_offset64)
395 outp += sprintf(outp, " 0x%016llx", t->extra_msr64);
396
363 if (do_nhm_cstates) { 397 if (do_nhm_cstates) {
364 if (!skip_c1) 398 if (!skip_c1)
365 outp += sprintf(outp, " %6.2f", 100.0 * t->c1/t->tsc); 399 outp += sprintf(outp, " %6.2f", 100.0 * t->c1/t->tsc);
@@ -391,8 +425,6 @@ int format_counters(struct thread_data *t, struct core_data *c,
391 if (do_snb_cstates) 425 if (do_snb_cstates)
392 outp += sprintf(outp, " %6.2f", 100.0 * p->pc7/t->tsc); 426 outp += sprintf(outp, " %6.2f", 100.0 * p->pc7/t->tsc);
393done: 427done:
394 if (extra_msr_offset)
395 outp += sprintf(outp, " 0x%016llx", t->extra_msr);
396 outp += sprintf(outp, "\n"); 428 outp += sprintf(outp, "\n");
397 429
398 return 0; 430 return 0;
@@ -502,10 +534,16 @@ delta_thread(struct thread_data *new, struct thread_data *old,
502 old->mperf = 1; /* divide by 0 protection */ 534 old->mperf = 1; /* divide by 0 protection */
503 } 535 }
504 536
537 old->extra_delta32 = new->extra_delta32 - old->extra_delta32;
538 old->extra_delta32 &= 0xFFFFFFFF;
539
540 old->extra_delta64 = new->extra_delta64 - old->extra_delta64;
541
505 /* 542 /*
506 * for "extra msr", just copy the latest w/o subtracting 543 * Extra MSR is just a snapshot, simply copy latest w/o subtracting
507 */ 544 */
508 old->extra_msr = new->extra_msr; 545 old->extra_msr32 = new->extra_msr32;
546 old->extra_msr64 = new->extra_msr64;
509} 547}
510 548
511int delta_cpu(struct thread_data *t, struct core_data *c, 549int delta_cpu(struct thread_data *t, struct core_data *c,
@@ -533,6 +571,9 @@ void clear_counters(struct thread_data *t, struct core_data *c, struct pkg_data
533 t->mperf = 0; 571 t->mperf = 0;
534 t->c1 = 0; 572 t->c1 = 0;
535 573
574 t->extra_delta32 = 0;
575 t->extra_delta64 = 0;
576
536 /* tells format_counters to dump all fields from this set */ 577 /* tells format_counters to dump all fields from this set */
537 t->flags = CPU_IS_FIRST_THREAD_IN_CORE | CPU_IS_FIRST_CORE_IN_PACKAGE; 578 t->flags = CPU_IS_FIRST_THREAD_IN_CORE | CPU_IS_FIRST_CORE_IN_PACKAGE;
538 579
@@ -553,6 +594,9 @@ int sum_counters(struct thread_data *t, struct core_data *c,
553 average.threads.mperf += t->mperf; 594 average.threads.mperf += t->mperf;
554 average.threads.c1 += t->c1; 595 average.threads.c1 += t->c1;
555 596
597 average.threads.extra_delta32 += t->extra_delta32;
598 average.threads.extra_delta64 += t->extra_delta64;
599
556 /* sum per-core values only for 1st thread in core */ 600 /* sum per-core values only for 1st thread in core */
557 if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE)) 601 if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE))
558 return 0; 602 return 0;
@@ -588,6 +632,11 @@ void compute_average(struct thread_data *t, struct core_data *c,
588 average.threads.mperf /= topo.num_cpus; 632 average.threads.mperf /= topo.num_cpus;
589 average.threads.c1 /= topo.num_cpus; 633 average.threads.c1 /= topo.num_cpus;
590 634
635 average.threads.extra_delta32 /= topo.num_cpus;
636 average.threads.extra_delta32 &= 0xFFFFFFFF;
637
638 average.threads.extra_delta64 /= topo.num_cpus;
639
591 average.cores.c3 /= topo.num_cores; 640 average.cores.c3 /= topo.num_cores;
592 average.cores.c6 /= topo.num_cores; 641 average.cores.c6 /= topo.num_cores;
593 average.cores.c7 /= topo.num_cores; 642 average.cores.c7 /= topo.num_cores;
@@ -629,8 +678,24 @@ int get_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
629 return -4; 678 return -4;
630 } 679 }
631 680
632 if (extra_msr_offset) 681 if (extra_delta_offset32) {
633 if (get_msr(cpu, extra_msr_offset, &t->extra_msr)) 682 if (get_msr(cpu, extra_delta_offset32, &t->extra_delta32))
683 return -5;
684 t->extra_delta32 &= 0xFFFFFFFF;
685 }
686
687 if (extra_delta_offset64)
688 if (get_msr(cpu, extra_delta_offset64, &t->extra_delta64))
689 return -5;
690
691 if (extra_msr_offset32) {
692 if (get_msr(cpu, extra_msr_offset32, &t->extra_msr32))
693 return -5;
694 t->extra_msr32 &= 0xFFFFFFFF;
695 }
696
697 if (extra_msr_offset64)
698 if (get_msr(cpu, extra_msr_offset64, &t->extra_msr64))
634 return -5; 699 return -5;
635 700
636 /* collect core counters only for 1st thread in core */ 701 /* collect core counters only for 1st thread in core */
@@ -677,6 +742,9 @@ void print_verbose_header(void)
677 742
678 get_msr(0, MSR_NEHALEM_PLATFORM_INFO, &msr); 743 get_msr(0, MSR_NEHALEM_PLATFORM_INFO, &msr);
679 744
745 if (verbose > 1)
746 fprintf(stderr, "MSR_NEHALEM_PLATFORM_INFO: 0x%llx\n", msr);
747
680 ratio = (msr >> 40) & 0xFF; 748 ratio = (msr >> 40) & 0xFF;
681 fprintf(stderr, "%d * %.0f = %.0f MHz max efficiency\n", 749 fprintf(stderr, "%d * %.0f = %.0f MHz max efficiency\n",
682 ratio, bclk, ratio * bclk); 750 ratio, bclk, ratio * bclk);
@@ -685,14 +753,84 @@ void print_verbose_header(void)
685 fprintf(stderr, "%d * %.0f = %.0f MHz TSC frequency\n", 753 fprintf(stderr, "%d * %.0f = %.0f MHz TSC frequency\n",
686 ratio, bclk, ratio * bclk); 754 ratio, bclk, ratio * bclk);
687 755
756 if (!do_ivt_turbo_ratio_limit)
757 goto print_nhm_turbo_ratio_limits;
758
759 get_msr(0, MSR_IVT_TURBO_RATIO_LIMIT, &msr);
760
688 if (verbose > 1) 761 if (verbose > 1)
689 fprintf(stderr, "MSR_NEHALEM_PLATFORM_INFO: 0x%llx\n", msr); 762 fprintf(stderr, "MSR_IVT_TURBO_RATIO_LIMIT: 0x%llx\n", msr);
763
764 ratio = (msr >> 56) & 0xFF;
765 if (ratio)
766 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 16 active cores\n",
767 ratio, bclk, ratio * bclk);
768
769 ratio = (msr >> 48) & 0xFF;
770 if (ratio)
771 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 15 active cores\n",
772 ratio, bclk, ratio * bclk);
773
774 ratio = (msr >> 40) & 0xFF;
775 if (ratio)
776 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 14 active cores\n",
777 ratio, bclk, ratio * bclk);
778
779 ratio = (msr >> 32) & 0xFF;
780 if (ratio)
781 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 13 active cores\n",
782 ratio, bclk, ratio * bclk);
783
784 ratio = (msr >> 24) & 0xFF;
785 if (ratio)
786 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 12 active cores\n",
787 ratio, bclk, ratio * bclk);
788
789 ratio = (msr >> 16) & 0xFF;
790 if (ratio)
791 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 11 active cores\n",
792 ratio, bclk, ratio * bclk);
793
794 ratio = (msr >> 8) & 0xFF;
795 if (ratio)
796 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 10 active cores\n",
797 ratio, bclk, ratio * bclk);
798
799 ratio = (msr >> 0) & 0xFF;
800 if (ratio)
801 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 9 active cores\n",
802 ratio, bclk, ratio * bclk);
803
804print_nhm_turbo_ratio_limits:
690 805
691 if (!do_nehalem_turbo_ratio_limit) 806 if (!do_nehalem_turbo_ratio_limit)
692 return; 807 return;
693 808
694 get_msr(0, MSR_NEHALEM_TURBO_RATIO_LIMIT, &msr); 809 get_msr(0, MSR_NEHALEM_TURBO_RATIO_LIMIT, &msr);
695 810
811 if (verbose > 1)
812 fprintf(stderr, "MSR_NEHALEM_TURBO_RATIO_LIMIT: 0x%llx\n", msr);
813
814 ratio = (msr >> 56) & 0xFF;
815 if (ratio)
816 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 8 active cores\n",
817 ratio, bclk, ratio * bclk);
818
819 ratio = (msr >> 48) & 0xFF;
820 if (ratio)
821 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 7 active cores\n",
822 ratio, bclk, ratio * bclk);
823
824 ratio = (msr >> 40) & 0xFF;
825 if (ratio)
826 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 6 active cores\n",
827 ratio, bclk, ratio * bclk);
828
829 ratio = (msr >> 32) & 0xFF;
830 if (ratio)
831 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 5 active cores\n",
832 ratio, bclk, ratio * bclk);
833
696 ratio = (msr >> 24) & 0xFF; 834 ratio = (msr >> 24) & 0xFF;
697 if (ratio) 835 if (ratio)
698 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 4 active cores\n", 836 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 4 active cores\n",
@@ -712,7 +850,6 @@ void print_verbose_header(void)
712 if (ratio) 850 if (ratio)
713 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 1 active cores\n", 851 fprintf(stderr, "%d * %.0f = %.0f MHz max turbo 1 active cores\n",
714 ratio, bclk, ratio * bclk); 852 ratio, bclk, ratio * bclk);
715
716} 853}
717 854
718void free_all_buffers(void) 855void free_all_buffers(void)
@@ -1038,7 +1175,7 @@ int has_nehalem_turbo_ratio_limit(unsigned int family, unsigned int model)
1038 case 0x2A: /* SNB */ 1175 case 0x2A: /* SNB */
1039 case 0x2D: /* SNB Xeon */ 1176 case 0x2D: /* SNB Xeon */
1040 case 0x3A: /* IVB */ 1177 case 0x3A: /* IVB */
1041 case 0x3D: /* IVB Xeon */ 1178 case 0x3E: /* IVB Xeon */
1042 return 1; 1179 return 1;
1043 case 0x2E: /* Nehalem-EX Xeon - Beckton */ 1180 case 0x2E: /* Nehalem-EX Xeon - Beckton */
1044 case 0x2F: /* Westmere-EX Xeon - Eagleton */ 1181 case 0x2F: /* Westmere-EX Xeon - Eagleton */
@@ -1046,6 +1183,22 @@ int has_nehalem_turbo_ratio_limit(unsigned int family, unsigned int model)
1046 return 0; 1183 return 0;
1047 } 1184 }
1048} 1185}
1186int has_ivt_turbo_ratio_limit(unsigned int family, unsigned int model)
1187{
1188 if (!genuine_intel)
1189 return 0;
1190
1191 if (family != 6)
1192 return 0;
1193
1194 switch (model) {
1195 case 0x3E: /* IVB Xeon */
1196 return 1;
1197 default:
1198 return 0;
1199 }
1200}
1201
1049 1202
1050int is_snb(unsigned int family, unsigned int model) 1203int is_snb(unsigned int family, unsigned int model)
1051{ 1204{
@@ -1056,7 +1209,7 @@ int is_snb(unsigned int family, unsigned int model)
1056 case 0x2A: 1209 case 0x2A:
1057 case 0x2D: 1210 case 0x2D:
1058 case 0x3A: /* IVB */ 1211 case 0x3A: /* IVB */
1059 case 0x3D: /* IVB Xeon */ 1212 case 0x3E: /* IVB Xeon */
1060 return 1; 1213 return 1;
1061 } 1214 }
1062 return 0; 1215 return 0;
@@ -1145,12 +1298,13 @@ void check_cpuid()
1145 bclk = discover_bclk(family, model); 1298 bclk = discover_bclk(family, model);
1146 1299
1147 do_nehalem_turbo_ratio_limit = has_nehalem_turbo_ratio_limit(family, model); 1300 do_nehalem_turbo_ratio_limit = has_nehalem_turbo_ratio_limit(family, model);
1301 do_ivt_turbo_ratio_limit = has_ivt_turbo_ratio_limit(family, model);
1148} 1302}
1149 1303
1150 1304
1151void usage() 1305void usage()
1152{ 1306{
1153 fprintf(stderr, "%s: [-v] [-M MSR#] [-i interval_sec | command ...]\n", 1307 fprintf(stderr, "%s: [-v][-p|-P|-S][-c MSR# | -s]][-C MSR#][-m MSR#][-M MSR#][-i interval_sec | command ...]\n",
1154 progname); 1308 progname);
1155 exit(1); 1309 exit(1);
1156} 1310}
@@ -1440,15 +1594,15 @@ void cmdline(int argc, char **argv)
1440 1594
1441 progname = argv[0]; 1595 progname = argv[0];
1442 1596
1443 while ((opt = getopt(argc, argv, "+cpsvi:M:")) != -1) { 1597 while ((opt = getopt(argc, argv, "+pPSvisc:sC:m:M:")) != -1) {
1444 switch (opt) { 1598 switch (opt) {
1445 case 'c': 1599 case 'p':
1446 show_core_only++; 1600 show_core_only++;
1447 break; 1601 break;
1448 case 'p': 1602 case 'P':
1449 show_pkg_only++; 1603 show_pkg_only++;
1450 break; 1604 break;
1451 case 's': 1605 case 'S':
1452 summary_only++; 1606 summary_only++;
1453 break; 1607 break;
1454 case 'v': 1608 case 'v':
@@ -1457,10 +1611,20 @@ void cmdline(int argc, char **argv)
1457 case 'i': 1611 case 'i':
1458 interval_sec = atoi(optarg); 1612 interval_sec = atoi(optarg);
1459 break; 1613 break;
1614 case 'c':
1615 sscanf(optarg, "%x", &extra_delta_offset32);
1616 break;
1617 case 's':
1618 extra_delta_offset32 = 0x34; /* SMI counter */
1619 break;
1620 case 'C':
1621 sscanf(optarg, "%x", &extra_delta_offset64);
1622 break;
1623 case 'm':
1624 sscanf(optarg, "%x", &extra_msr_offset32);
1625 break;
1460 case 'M': 1626 case 'M':
1461 sscanf(optarg, "%x", &extra_msr_offset); 1627 sscanf(optarg, "%x", &extra_msr_offset64);
1462 if (verbose > 1)
1463 fprintf(stderr, "MSR 0x%X\n", extra_msr_offset);
1464 break; 1628 break;
1465 default: 1629 default:
1466 usage(); 1630 usage();
@@ -1473,7 +1637,7 @@ int main(int argc, char **argv)
1473 cmdline(argc, argv); 1637 cmdline(argc, argv);
1474 1638
1475 if (verbose > 1) 1639 if (verbose > 1)
1476 fprintf(stderr, "turbostat v2.0 May 16, 2012" 1640 fprintf(stderr, "turbostat v2.1 October 6, 2012"
1477 " - Len Brown <lenb@kernel.org>\n"); 1641 " - Len Brown <lenb@kernel.org>\n");
1478 1642
1479 turbostat_init(); 1643 turbostat_init();
diff --git a/tools/virtio/virtio-trace/Makefile b/tools/virtio/virtio-trace/Makefile
new file mode 100644
index 000000000000..0d2381633475
--- /dev/null
+++ b/tools/virtio/virtio-trace/Makefile
@@ -0,0 +1,13 @@
1CC = gcc
2CFLAGS = -O2 -Wall -pthread
3
4all: trace-agent
5
6.c.o:
7 $(CC) $(CFLAGS) -c $^ -o $@
8
9trace-agent: trace-agent.o trace-agent-ctl.o trace-agent-rw.o
10 $(CC) $(CFLAGS) -o $@ $^
11
12clean:
13 rm -f *.o trace-agent
diff --git a/tools/virtio/virtio-trace/README b/tools/virtio/virtio-trace/README
new file mode 100644
index 000000000000..b64845b823ab
--- /dev/null
+++ b/tools/virtio/virtio-trace/README
@@ -0,0 +1,118 @@
1Trace Agent for virtio-trace
2============================
3
4Trace agent is a user tool for sending trace data of a guest to a Host in low
5overhead. Trace agent has the following functions:
6 - splice a page of ring-buffer to read_pipe without memory copying
7 - splice the page from write_pipe to virtio-console without memory copying
8 - write trace data to stdout by using -o option
9 - controlled by start/stop orders from a Host
10
11The trace agent operates as follows:
12 1) Initialize all structures.
13 2) Create a read/write thread per CPU. Each thread is bound to a CPU.
14 The read/write threads hold it.
15 3) A controller thread does poll() for a start order of a host.
16 4) After the controller of the trace agent receives a start order from a host,
17 the controller wake read/write threads.
18 5) The read/write threads start to read trace data from ring-buffers and
19 write the data to virtio-serial.
20 6) If the controller receives a stop order from a host, the read/write threads
21 stop to read trace data.
22
23
24Files
25=====
26
27README: this file
28Makefile: Makefile of trace agent for virtio-trace
29trace-agent.c: includes main function, sets up for operating trace agent
30trace-agent.h: includes all structures and some macros
31trace-agent-ctl.c: includes controller function for read/write threads
32trace-agent-rw.c: includes read/write threads function
33
34
35Setup
36=====
37
38To use this trace agent for virtio-trace, we need to prepare some virtio-serial
39I/Fs.
40
411) Make FIFO in a host
42 virtio-trace uses virtio-serial pipe as trace data paths as to the number
43of CPUs and a control path, so FIFO (named pipe) should be created as follows:
44 # mkdir /tmp/virtio-trace/
45 # mkfifo /tmp/virtio-trace/trace-path-cpu{0,1,2,...,X}.{in,out}
46 # mkfifo /tmp/virtio-trace/agent-ctl-path.{in,out}
47
48For example, if a guest use three CPUs, the names are
49 trace-path-cpu{0,1,2}.{in.out}
50and
51 agent-ctl-path.{in,out}.
52
532) Set up of virtio-serial pipe in a host
54 Add qemu option to use virtio-serial pipe.
55
56 ##virtio-serial device##
57 -device virtio-serial-pci,id=virtio-serial0\
58 ##control path##
59 -chardev pipe,id=charchannel0,path=/tmp/virtio-trace/agent-ctl-path\
60 -device virtserialport,bus=virtio-serial0.0,nr=1,chardev=charchannel0,\
61 id=channel0,name=agent-ctl-path\
62 ##data path##
63 -chardev pipe,id=charchannel1,path=/tmp/virtio-trace/trace-path-cpu0\
64 -device virtserialport,bus=virtio-serial0.0,nr=2,chardev=charchannel0,\
65 id=channel1,name=trace-path-cpu0\
66 ...
67
68If you manage guests with libvirt, add the following tags to domain XML files.
69Then, libvirt passes the same command option to qemu.
70
71 <channel type='pipe'>
72 <source path='/tmp/virtio-trace/agent-ctl-path'/>
73 <target type='virtio' name='agent-ctl-path'/>
74 <address type='virtio-serial' controller='0' bus='0' port='0'/>
75 </channel>
76 <channel type='pipe'>
77 <source path='/tmp/virtio-trace/trace-path-cpu0'/>
78 <target type='virtio' name='trace-path-cpu0'/>
79 <address type='virtio-serial' controller='0' bus='0' port='1'/>
80 </channel>
81 ...
82Here, chardev names are restricted to trace-path-cpuX and agent-ctl-path. For
83example, if a guest use three CPUs, chardev names should be trace-path-cpu0,
84trace-path-cpu1, trace-path-cpu2, and agent-ctl-path.
85
863) Boot the guest
87 You can find some chardev in /dev/virtio-ports/ in the guest.
88
89
90Run
91===
92
930) Build trace agent in a guest
94 $ make
95
961) Enable ftrace in the guest
97 <Example>
98 # echo 1 > /sys/kernel/debug/tracing/events/sched/enable
99
1002) Run trace agent in the guest
101 This agent must be operated as root.
102 # ./trace-agent
103read/write threads in the agent wait for start order from host. If you add -o
104option, trace data are output via stdout in the guest.
105
1063) Open FIFO in a host
107 # cat /tmp/virtio-trace/trace-path-cpu0.out
108If a host does not open these, trace data get stuck in buffers of virtio. Then,
109the guest will stop by specification of chardev in QEMU. This blocking mode may
110be solved in the future.
111
1124) Start to read trace data by ordering from a host
113 A host injects read start order to the guest via virtio-serial.
114 # echo 1 > /tmp/virtio-trace/agent-ctl-path.in
115
1165) Stop to read trace data by ordering from a host
117 A host injects read stop order to the guest via virtio-serial.
118 # echo 0 > /tmp/virtio-trace/agent-ctl-path.in
diff --git a/tools/virtio/virtio-trace/trace-agent-ctl.c b/tools/virtio/virtio-trace/trace-agent-ctl.c
new file mode 100644
index 000000000000..a2d0403c4f94
--- /dev/null
+++ b/tools/virtio/virtio-trace/trace-agent-ctl.c
@@ -0,0 +1,137 @@
1/*
2 * Controller of read/write threads for virtio-trace
3 *
4 * Copyright (C) 2012 Hitachi, Ltd.
5 * Created by Yoshihiro Yunomae <yoshihiro.yunomae.ez@hitachi.com>
6 * Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
7 *
8 * Licensed under GPL version 2 only.
9 *
10 */
11
12#define _GNU_SOURCE
13#include <fcntl.h>
14#include <poll.h>
15#include <signal.h>
16#include <stdio.h>
17#include <stdlib.h>
18#include <unistd.h>
19#include "trace-agent.h"
20
21#define HOST_MSG_SIZE 256
22#define EVENT_WAIT_MSEC 100
23
24static volatile sig_atomic_t global_signal_val;
25bool global_sig_receive; /* default false */
26bool global_run_operation; /* default false*/
27
28/* Handle SIGTERM/SIGINT/SIGQUIT to exit */
29static void signal_handler(int sig)
30{
31 global_signal_val = sig;
32}
33
34int rw_ctl_init(const char *ctl_path)
35{
36 int ctl_fd;
37
38 ctl_fd = open(ctl_path, O_RDONLY);
39 if (ctl_fd == -1) {
40 pr_err("Cannot open ctl_fd\n");
41 goto error;
42 }
43
44 return ctl_fd;
45
46error:
47 exit(EXIT_FAILURE);
48}
49
50static int wait_order(int ctl_fd)
51{
52 struct pollfd poll_fd;
53 int ret = 0;
54
55 while (!global_sig_receive) {
56 poll_fd.fd = ctl_fd;
57 poll_fd.events = POLLIN;
58
59 ret = poll(&poll_fd, 1, EVENT_WAIT_MSEC);
60
61 if (global_signal_val) {
62 global_sig_receive = true;
63 pr_info("Receive interrupt %d\n", global_signal_val);
64
65 /* Wakes rw-threads when they are sleeping */
66 if (!global_run_operation)
67 pthread_cond_broadcast(&cond_wakeup);
68
69 ret = -1;
70 break;
71 }
72
73 if (ret < 0) {
74 pr_err("Polling error\n");
75 goto error;
76 }
77
78 if (ret)
79 break;
80 };
81
82 return ret;
83
84error:
85 exit(EXIT_FAILURE);
86}
87
88/*
89 * contol read/write threads by handling global_run_operation
90 */
91void *rw_ctl_loop(int ctl_fd)
92{
93 ssize_t rlen;
94 char buf[HOST_MSG_SIZE];
95 int ret;
96
97 /* Setup signal handlers */
98 signal(SIGTERM, signal_handler);
99 signal(SIGINT, signal_handler);
100 signal(SIGQUIT, signal_handler);
101
102 while (!global_sig_receive) {
103
104 ret = wait_order(ctl_fd);
105 if (ret < 0)
106 break;
107
108 rlen = read(ctl_fd, buf, sizeof(buf));
109 if (rlen < 0) {
110 pr_err("read data error in ctl thread\n");
111 goto error;
112 }
113
114 if (rlen == 2 && buf[0] == '1') {
115 /*
116 * If host writes '1' to a control path,
117 * this controller wakes all read/write threads.
118 */
119 global_run_operation = true;
120 pthread_cond_broadcast(&cond_wakeup);
121 pr_debug("Wake up all read/write threads\n");
122 } else if (rlen == 2 && buf[0] == '0') {
123 /*
124 * If host writes '0' to a control path, read/write
125 * threads will wait for notification from Host.
126 */
127 global_run_operation = false;
128 pr_debug("Stop all read/write threads\n");
129 } else
130 pr_info("Invalid host notification: %s\n", buf);
131 }
132
133 return NULL;
134
135error:
136 exit(EXIT_FAILURE);
137}
diff --git a/tools/virtio/virtio-trace/trace-agent-rw.c b/tools/virtio/virtio-trace/trace-agent-rw.c
new file mode 100644
index 000000000000..3aace5ea4842
--- /dev/null
+++ b/tools/virtio/virtio-trace/trace-agent-rw.c
@@ -0,0 +1,192 @@
1/*
2 * Read/write thread of a guest agent for virtio-trace
3 *
4 * Copyright (C) 2012 Hitachi, Ltd.
5 * Created by Yoshihiro Yunomae <yoshihiro.yunomae.ez@hitachi.com>
6 * Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
7 *
8 * Licensed under GPL version 2 only.
9 *
10 */
11
12#define _GNU_SOURCE
13#include <fcntl.h>
14#include <stdio.h>
15#include <stdlib.h>
16#include <unistd.h>
17#include <sys/syscall.h>
18#include "trace-agent.h"
19
20#define READ_WAIT_USEC 100000
21
22void *rw_thread_info_new(void)
23{
24 struct rw_thread_info *rw_ti;
25
26 rw_ti = zalloc(sizeof(struct rw_thread_info));
27 if (rw_ti == NULL) {
28 pr_err("rw_thread_info zalloc error\n");
29 exit(EXIT_FAILURE);
30 }
31
32 rw_ti->cpu_num = -1;
33 rw_ti->in_fd = -1;
34 rw_ti->out_fd = -1;
35 rw_ti->read_pipe = -1;
36 rw_ti->write_pipe = -1;
37 rw_ti->pipe_size = PIPE_INIT;
38
39 return rw_ti;
40}
41
42void *rw_thread_init(int cpu, const char *in_path, const char *out_path,
43 bool stdout_flag, unsigned long pipe_size,
44 struct rw_thread_info *rw_ti)
45{
46 int data_pipe[2];
47
48 rw_ti->cpu_num = cpu;
49
50 /* set read(input) fd */
51 rw_ti->in_fd = open(in_path, O_RDONLY);
52 if (rw_ti->in_fd == -1) {
53 pr_err("Could not open in_fd (CPU:%d)\n", cpu);
54 goto error;
55 }
56
57 /* set write(output) fd */
58 if (!stdout_flag) {
59 /* virtio-serial output mode */
60 rw_ti->out_fd = open(out_path, O_WRONLY);
61 if (rw_ti->out_fd == -1) {
62 pr_err("Could not open out_fd (CPU:%d)\n", cpu);
63 goto error;
64 }
65 } else
66 /* stdout mode */
67 rw_ti->out_fd = STDOUT_FILENO;
68
69 if (pipe2(data_pipe, O_NONBLOCK) < 0) {
70 pr_err("Could not create pipe in rw-thread(%d)\n", cpu);
71 goto error;
72 }
73
74 /*
75 * Size of pipe is 64kB in default based on fs/pipe.c.
76 * To read/write trace data speedy, pipe size is changed.
77 */
78 if (fcntl(*data_pipe, F_SETPIPE_SZ, pipe_size) < 0) {
79 pr_err("Could not change pipe size in rw-thread(%d)\n", cpu);
80 goto error;
81 }
82
83 rw_ti->read_pipe = data_pipe[1];
84 rw_ti->write_pipe = data_pipe[0];
85 rw_ti->pipe_size = pipe_size;
86
87 return NULL;
88
89error:
90 exit(EXIT_FAILURE);
91}
92
93/* Bind a thread to a cpu */
94static void bind_cpu(int cpu_num)
95{
96 cpu_set_t mask;
97
98 CPU_ZERO(&mask);
99 CPU_SET(cpu_num, &mask);
100
101 /* bind my thread to cpu_num by assigning zero to the first argument */
102 if (sched_setaffinity(0, sizeof(mask), &mask) == -1)
103 pr_err("Could not set CPU#%d affinity\n", (int)cpu_num);
104}
105
106static void *rw_thread_main(void *thread_info)
107{
108 ssize_t rlen, wlen;
109 ssize_t ret;
110 struct rw_thread_info *ts = (struct rw_thread_info *)thread_info;
111
112 bind_cpu(ts->cpu_num);
113
114 while (1) {
115 /* Wait for a read order of trace data by Host OS */
116 if (!global_run_operation) {
117 pthread_mutex_lock(&mutex_notify);
118 pthread_cond_wait(&cond_wakeup, &mutex_notify);
119 pthread_mutex_unlock(&mutex_notify);
120 }
121
122 if (global_sig_receive)
123 break;
124
125 /*
126 * Each thread read trace_pipe_raw of each cpu bounding the
127 * thread, so contention of multi-threads does not occur.
128 */
129 rlen = splice(ts->in_fd, NULL, ts->read_pipe, NULL,
130 ts->pipe_size, SPLICE_F_MOVE | SPLICE_F_MORE);
131
132 if (rlen < 0) {
133 pr_err("Splice_read in rw-thread(%d)\n", ts->cpu_num);
134 goto error;
135 } else if (rlen == 0) {
136 /*
137 * If trace data do not exist or are unreadable not
138 * for exceeding the page size, splice_read returns
139 * NULL. Then, this waits for being filled the data in a
140 * ring-buffer.
141 */
142 usleep(READ_WAIT_USEC);
143 pr_debug("Read retry(cpu:%d)\n", ts->cpu_num);
144 continue;
145 }
146
147 wlen = 0;
148
149 do {
150 ret = splice(ts->write_pipe, NULL, ts->out_fd, NULL,
151 rlen - wlen,
152 SPLICE_F_MOVE | SPLICE_F_MORE);
153
154 if (ret < 0) {
155 pr_err("Splice_write in rw-thread(%d)\n",
156 ts->cpu_num);
157 goto error;
158 } else if (ret == 0)
159 /*
160 * When host reader is not in time for reading
161 * trace data, guest will be stopped. This is
162 * because char dev in QEMU is not supported
163 * non-blocking mode. Then, writer might be
164 * sleep in that case.
165 * This sleep will be removed by supporting
166 * non-blocking mode.
167 */
168 sleep(1);
169 wlen += ret;
170 } while (wlen < rlen);
171 }
172
173 return NULL;
174
175error:
176 exit(EXIT_FAILURE);
177}
178
179
180pthread_t rw_thread_run(struct rw_thread_info *rw_ti)
181{
182 int ret;
183 pthread_t rw_thread_per_cpu;
184
185 ret = pthread_create(&rw_thread_per_cpu, NULL, rw_thread_main, rw_ti);
186 if (ret != 0) {
187 pr_err("Could not create a rw thread(%d)\n", rw_ti->cpu_num);
188 exit(EXIT_FAILURE);
189 }
190
191 return rw_thread_per_cpu;
192}
diff --git a/tools/virtio/virtio-trace/trace-agent.c b/tools/virtio/virtio-trace/trace-agent.c
new file mode 100644
index 000000000000..0a0a7dd4eff7
--- /dev/null
+++ b/tools/virtio/virtio-trace/trace-agent.c
@@ -0,0 +1,270 @@
1/*
2 * Guest agent for virtio-trace
3 *
4 * Copyright (C) 2012 Hitachi, Ltd.
5 * Created by Yoshihiro Yunomae <yoshihiro.yunomae.ez@hitachi.com>
6 * Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
7 *
8 * Licensed under GPL version 2 only.
9 *
10 */
11
12#define _GNU_SOURCE
13#include <limits.h>
14#include <stdio.h>
15#include <stdlib.h>
16#include <unistd.h>
17#include "trace-agent.h"
18
19#define PAGE_SIZE (sysconf(_SC_PAGE_SIZE))
20#define PIPE_DEF_BUFS 16
21#define PIPE_MIN_SIZE (PAGE_SIZE*PIPE_DEF_BUFS)
22#define PIPE_MAX_SIZE (1024*1024)
23#define READ_PATH_FMT \
24 "/sys/kernel/debug/tracing/per_cpu/cpu%d/trace_pipe_raw"
25#define WRITE_PATH_FMT "/dev/virtio-ports/trace-path-cpu%d"
26#define CTL_PATH "/dev/virtio-ports/agent-ctl-path"
27
28pthread_mutex_t mutex_notify = PTHREAD_MUTEX_INITIALIZER;
29pthread_cond_t cond_wakeup = PTHREAD_COND_INITIALIZER;
30
31static int get_total_cpus(void)
32{
33 int nr_cpus = (int)sysconf(_SC_NPROCESSORS_CONF);
34
35 if (nr_cpus <= 0) {
36 pr_err("Could not read cpus\n");
37 goto error;
38 } else if (nr_cpus > MAX_CPUS) {
39 pr_err("Exceed max cpus(%d)\n", (int)MAX_CPUS);
40 goto error;
41 }
42
43 return nr_cpus;
44
45error:
46 exit(EXIT_FAILURE);
47}
48
49static void *agent_info_new(void)
50{
51 struct agent_info *s;
52 int i;
53
54 s = zalloc(sizeof(struct agent_info));
55 if (s == NULL) {
56 pr_err("agent_info zalloc error\n");
57 exit(EXIT_FAILURE);
58 }
59
60 s->pipe_size = PIPE_INIT;
61 s->use_stdout = false;
62 s->cpus = get_total_cpus();
63 s->ctl_fd = -1;
64
65 /* read/write threads init */
66 for (i = 0; i < s->cpus; i++)
67 s->rw_ti[i] = rw_thread_info_new();
68
69 return s;
70}
71
72static unsigned long parse_size(const char *arg)
73{
74 unsigned long value, round;
75 char *ptr;
76
77 value = strtoul(arg, &ptr, 10);
78 switch (*ptr) {
79 case 'K': case 'k':
80 value <<= 10;
81 break;
82 case 'M': case 'm':
83 value <<= 20;
84 break;
85 default:
86 break;
87 }
88
89 if (value > PIPE_MAX_SIZE) {
90 pr_err("Pipe size must be less than 1MB\n");
91 goto error;
92 } else if (value < PIPE_MIN_SIZE) {
93 pr_err("Pipe size must be over 64KB\n");
94 goto error;
95 }
96
97 /* Align buffer size with page unit */
98 round = value & (PAGE_SIZE - 1);
99 value = value - round;
100
101 return value;
102error:
103 return 0;
104}
105
106static void usage(char const *prg)
107{
108 pr_err("usage: %s [-h] [-o] [-s <size of pipe>]\n", prg);
109}
110
111static const char *make_path(int cpu_num, bool this_is_write_path)
112{
113 int ret;
114 char *buf;
115
116 buf = zalloc(PATH_MAX);
117 if (buf == NULL) {
118 pr_err("Could not allocate buffer\n");
119 goto error;
120 }
121
122 if (this_is_write_path)
123 /* write(output) path */
124 ret = snprintf(buf, PATH_MAX, WRITE_PATH_FMT, cpu_num);
125 else
126 /* read(input) path */
127 ret = snprintf(buf, PATH_MAX, READ_PATH_FMT, cpu_num);
128
129 if (ret <= 0) {
130 pr_err("Failed to generate %s path(CPU#%d):%d\n",
131 this_is_write_path ? "read" : "write", cpu_num, ret);
132 goto error;
133 }
134
135 return buf;
136
137error:
138 free(buf);
139 return NULL;
140}
141
142static const char *make_input_path(int cpu_num)
143{
144 return make_path(cpu_num, false);
145}
146
147static const char *make_output_path(int cpu_num)
148{
149 return make_path(cpu_num, true);
150}
151
152static void *agent_info_init(struct agent_info *s)
153{
154 int cpu;
155 const char *in_path = NULL;
156 const char *out_path = NULL;
157
158 /* init read/write threads */
159 for (cpu = 0; cpu < s->cpus; cpu++) {
160 /* set read(input) path per read/write thread */
161 in_path = make_input_path(cpu);
162 if (in_path == NULL)
163 goto error;
164
165 /* set write(output) path per read/write thread*/
166 if (!s->use_stdout) {
167 out_path = make_output_path(cpu);
168 if (out_path == NULL)
169 goto error;
170 } else
171 /* stdout mode */
172 pr_debug("stdout mode\n");
173
174 rw_thread_init(cpu, in_path, out_path, s->use_stdout,
175 s->pipe_size, s->rw_ti[cpu]);
176 }
177
178 /* init controller of read/write threads */
179 s->ctl_fd = rw_ctl_init((const char *)CTL_PATH);
180
181 return NULL;
182
183error:
184 exit(EXIT_FAILURE);
185}
186
187static void *parse_args(int argc, char *argv[], struct agent_info *s)
188{
189 int cmd;
190 unsigned long size;
191
192 while ((cmd = getopt(argc, argv, "hos:")) != -1) {
193 switch (cmd) {
194 /* stdout mode */
195 case 'o':
196 s->use_stdout = true;
197 break;
198 /* size of pipe */
199 case 's':
200 size = parse_size(optarg);
201 if (size == 0)
202 goto error;
203 s->pipe_size = size;
204 break;
205 case 'h':
206 default:
207 usage(argv[0]);
208 goto error;
209 }
210 }
211
212 agent_info_init(s);
213
214 return NULL;
215
216error:
217 exit(EXIT_FAILURE);
218}
219
220static void agent_main_loop(struct agent_info *s)
221{
222 int cpu;
223 pthread_t rw_thread_per_cpu[MAX_CPUS];
224
225 /* Start all read/write threads */
226 for (cpu = 0; cpu < s->cpus; cpu++)
227 rw_thread_per_cpu[cpu] = rw_thread_run(s->rw_ti[cpu]);
228
229 rw_ctl_loop(s->ctl_fd);
230
231 /* Finish all read/write threads */
232 for (cpu = 0; cpu < s->cpus; cpu++) {
233 int ret;
234
235 ret = pthread_join(rw_thread_per_cpu[cpu], NULL);
236 if (ret != 0) {
237 pr_err("pthread_join() error:%d (cpu %d)\n", ret, cpu);
238 exit(EXIT_FAILURE);
239 }
240 }
241}
242
243static void agent_info_free(struct agent_info *s)
244{
245 int i;
246
247 close(s->ctl_fd);
248 for (i = 0; i < s->cpus; i++) {
249 close(s->rw_ti[i]->in_fd);
250 close(s->rw_ti[i]->out_fd);
251 close(s->rw_ti[i]->read_pipe);
252 close(s->rw_ti[i]->write_pipe);
253 free(s->rw_ti[i]);
254 }
255 free(s);
256}
257
258int main(int argc, char *argv[])
259{
260 struct agent_info *s = NULL;
261
262 s = agent_info_new();
263 parse_args(argc, argv, s);
264
265 agent_main_loop(s);
266
267 agent_info_free(s);
268
269 return 0;
270}
diff --git a/tools/virtio/virtio-trace/trace-agent.h b/tools/virtio/virtio-trace/trace-agent.h
new file mode 100644
index 000000000000..8de79bfeaa73
--- /dev/null
+++ b/tools/virtio/virtio-trace/trace-agent.h
@@ -0,0 +1,75 @@
1#ifndef __TRACE_AGENT_H__
2#define __TRACE_AGENT_H__
3#include <pthread.h>
4#include <stdbool.h>
5
6#define MAX_CPUS 256
7#define PIPE_INIT (1024*1024)
8
9/*
10 * agent_info - structure managing total information of guest agent
11 * @pipe_size: size of pipe (default 1MB)
12 * @use_stdout: set to true when o option is added (default false)
13 * @cpus: total number of CPUs
14 * @ctl_fd: fd of control path, /dev/virtio-ports/agent-ctl-path
15 * @rw_ti: structure managing information of read/write threads
16 */
17struct agent_info {
18 unsigned long pipe_size;
19 bool use_stdout;
20 int cpus;
21 int ctl_fd;
22 struct rw_thread_info *rw_ti[MAX_CPUS];
23};
24
25/*
26 * rw_thread_info - structure managing a read/write thread a cpu
27 * @cpu_num: cpu number operating this read/write thread
28 * @in_fd: fd of reading trace data path in cpu_num
29 * @out_fd: fd of writing trace data path in cpu_num
30 * @read_pipe: fd of read pipe
31 * @write_pipe: fd of write pipe
32 * @pipe_size: size of pipe (default 1MB)
33 */
34struct rw_thread_info {
35 int cpu_num;
36 int in_fd;
37 int out_fd;
38 int read_pipe;
39 int write_pipe;
40 unsigned long pipe_size;
41};
42
43/* use for stopping rw threads */
44extern bool global_sig_receive;
45
46/* use for notification */
47extern bool global_run_operation;
48extern pthread_mutex_t mutex_notify;
49extern pthread_cond_t cond_wakeup;
50
51/* for controller of read/write threads */
52extern int rw_ctl_init(const char *ctl_path);
53extern void *rw_ctl_loop(int ctl_fd);
54
55/* for trace read/write thread */
56extern void *rw_thread_info_new(void);
57extern void *rw_thread_init(int cpu, const char *in_path, const char *out_path,
58 bool stdout_flag, unsigned long pipe_size,
59 struct rw_thread_info *rw_ti);
60extern pthread_t rw_thread_run(struct rw_thread_info *rw_ti);
61
62static inline void *zalloc(size_t size)
63{
64 return calloc(1, size);
65}
66
67#define pr_err(format, ...) fprintf(stderr, format, ## __VA_ARGS__)
68#define pr_info(format, ...) fprintf(stdout, format, ## __VA_ARGS__)
69#ifdef DEBUG
70#define pr_debug(format, ...) fprintf(stderr, format, ## __VA_ARGS__)
71#else
72#define pr_debug(format, ...) do {} while (0)
73#endif
74
75#endif /*__TRACE_AGENT_H__*/
generated by cgit v1.2.2 (git 2.25.0) at 2025-08-23 10:11:02 -0400