aboutsummaryrefslogtreecommitdiffstats
path: root/kernel
diff options
context:
space:
mode:
authorIngo Molnar <mingo@elte.hu>2008-07-26 09:30:40 -0400
committerIngo Molnar <mingo@elte.hu>2008-07-26 09:30:40 -0400
commit1503af661947b7a4a09355cc2ae6aa0d43f16776 (patch)
tree5bfcfadf2dd2d98c2ad251c96d7ee43a6903819a /kernel
parenta31863168660c6b6f6c7ffe05bb6a38e97803326 (diff)
parent024e8ac04453b3525448c31ef39848cf675ba6db (diff)
Merge branch 'linus' into x86/header-guards
Conflicts: include/asm-x86/gpio.h include/asm-x86/ide.h Signed-off-by: Ingo Molnar <mingo@elte.hu>
Diffstat (limited to 'kernel')
-rw-r--r--kernel/Kconfig.hz2
-rw-r--r--kernel/Makefile5
-rw-r--r--kernel/acct.c222
-rw-r--r--kernel/auditsc.c3
-rw-r--r--kernel/capability.c338
-rw-r--r--kernel/cgroup.c309
-rw-r--r--kernel/cpu.c47
-rw-r--r--kernel/cpuset.c363
-rw-r--r--kernel/delayacct.c16
-rw-r--r--kernel/exec_domain.c2
-rw-r--r--kernel/exit.c61
-rw-r--r--kernel/fork.c44
-rw-r--r--kernel/irq/manage.c107
-rw-r--r--kernel/kallsyms.c2
-rw-r--r--kernel/kmod.c15
-rw-r--r--kernel/kprobes.c132
-rw-r--r--kernel/kthread.c4
-rw-r--r--kernel/marker.c25
-rw-r--r--kernel/module.c336
-rw-r--r--kernel/ns_cgroup.c8
-rw-r--r--kernel/nsproxy.c8
-rw-r--r--kernel/panic.c22
-rw-r--r--kernel/pid.c10
-rw-r--r--kernel/pid_namespace.c10
-rw-r--r--kernel/posix-timers.c21
-rw-r--r--kernel/power/Kconfig11
-rw-r--r--kernel/power/main.c194
-rw-r--r--kernel/power/poweroff.c4
-rw-r--r--kernel/power/process.c2
-rw-r--r--kernel/power/snapshot.c88
-rw-r--r--kernel/printk.c19
-rw-r--r--kernel/profile.c4
-rw-r--r--kernel/rcuclassic.c2
-rw-r--r--kernel/rcupreempt.c10
-rw-r--r--kernel/res_counter.c48
-rw-r--r--kernel/rtmutex-tester.c7
-rw-r--r--kernel/sched.c359
-rw-r--r--kernel/sched_fair.c10
-rw-r--r--kernel/sched_rt.c83
-rw-r--r--kernel/signal.c80
-rw-r--r--kernel/softirq.c2
-rw-r--r--kernel/softlockup.c45
-rw-r--r--kernel/stop_machine.c3
-rw-r--r--kernel/sys.c4
-rw-r--r--kernel/sys_ni.c7
-rw-r--r--kernel/sysctl.c38
-rw-r--r--kernel/sysctl_check.c2
-rw-r--r--kernel/taskstats.c6
-rw-r--r--kernel/time/clocksource.c12
-rw-r--r--kernel/time/tick-broadcast.c3
-rw-r--r--kernel/time/tick-common.c14
-rw-r--r--kernel/time/tick-sched.c16
-rw-r--r--kernel/trace/trace_sysprof.c6
-rw-r--r--kernel/tsacct.c25
-rw-r--r--kernel/workqueue.c149
55 files changed, 1992 insertions, 1373 deletions
diff --git a/kernel/Kconfig.hz b/kernel/Kconfig.hz
index 526128a2e622..382dd5a8b2d7 100644
--- a/kernel/Kconfig.hz
+++ b/kernel/Kconfig.hz
@@ -55,4 +55,4 @@ config HZ
55 default 1000 if HZ_1000 55 default 1000 if HZ_1000
56 56
57config SCHED_HRTICK 57config SCHED_HRTICK
58 def_bool HIGH_RES_TIMERS && X86 58 def_bool HIGH_RES_TIMERS && USE_GENERIC_SMP_HELPERS
diff --git a/kernel/Makefile b/kernel/Makefile
index 985ddb7da4d0..54f69837d35a 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -2,7 +2,7 @@
2# Makefile for the linux kernel. 2# Makefile for the linux kernel.
3# 3#
4 4
5obj-y = sched.o fork.o exec_domain.o panic.o printk.o profile.o \ 5obj-y = sched.o fork.o exec_domain.o panic.o printk.o \
6 cpu.o exit.o itimer.o time.o softirq.o resource.o \ 6 cpu.o exit.o itimer.o time.o softirq.o resource.o \
7 sysctl.o capability.o ptrace.o timer.o user.o \ 7 sysctl.o capability.o ptrace.o timer.o user.o \
8 signal.o sys.o kmod.o workqueue.o pid.o \ 8 signal.o sys.o kmod.o workqueue.o pid.o \
@@ -11,6 +11,8 @@ obj-y = sched.o fork.o exec_domain.o panic.o printk.o profile.o \
11 hrtimer.o rwsem.o nsproxy.o srcu.o semaphore.o \ 11 hrtimer.o rwsem.o nsproxy.o srcu.o semaphore.o \
12 notifier.o ksysfs.o pm_qos_params.o sched_clock.o 12 notifier.o ksysfs.o pm_qos_params.o sched_clock.o
13 13
14CFLAGS_REMOVE_sched.o = -mno-spe
15
14ifdef CONFIG_FTRACE 16ifdef CONFIG_FTRACE
15# Do not trace debug files and internal ftrace files 17# Do not trace debug files and internal ftrace files
16CFLAGS_REMOVE_lockdep.o = -pg 18CFLAGS_REMOVE_lockdep.o = -pg
@@ -22,6 +24,7 @@ CFLAGS_REMOVE_sched_clock.o = -pg
22CFLAGS_REMOVE_sched.o = -mno-spe -pg 24CFLAGS_REMOVE_sched.o = -mno-spe -pg
23endif 25endif
24 26
27obj-$(CONFIG_PROFILING) += profile.o
25obj-$(CONFIG_SYSCTL_SYSCALL_CHECK) += sysctl_check.o 28obj-$(CONFIG_SYSCTL_SYSCALL_CHECK) += sysctl_check.o
26obj-$(CONFIG_STACKTRACE) += stacktrace.o 29obj-$(CONFIG_STACKTRACE) += stacktrace.o
27obj-y += time/ 30obj-y += time/
diff --git a/kernel/acct.c b/kernel/acct.c
index 91e1cfd734d2..dd68b9059418 100644
--- a/kernel/acct.c
+++ b/kernel/acct.c
@@ -75,37 +75,39 @@ int acct_parm[3] = {4, 2, 30};
75/* 75/*
76 * External references and all of the globals. 76 * External references and all of the globals.
77 */ 77 */
78static void do_acct_process(struct pid_namespace *ns, struct file *); 78static void do_acct_process(struct bsd_acct_struct *acct,
79 struct pid_namespace *ns, struct file *);
79 80
80/* 81/*
81 * This structure is used so that all the data protected by lock 82 * This structure is used so that all the data protected by lock
82 * can be placed in the same cache line as the lock. This primes 83 * can be placed in the same cache line as the lock. This primes
83 * the cache line to have the data after getting the lock. 84 * the cache line to have the data after getting the lock.
84 */ 85 */
85struct acct_glbs { 86struct bsd_acct_struct {
86 spinlock_t lock;
87 volatile int active; 87 volatile int active;
88 volatile int needcheck; 88 volatile int needcheck;
89 struct file *file; 89 struct file *file;
90 struct pid_namespace *ns; 90 struct pid_namespace *ns;
91 struct timer_list timer; 91 struct timer_list timer;
92 struct list_head list;
92}; 93};
93 94
94static struct acct_glbs acct_globals __cacheline_aligned = 95static DEFINE_SPINLOCK(acct_lock);
95 {__SPIN_LOCK_UNLOCKED(acct_globals.lock)}; 96static LIST_HEAD(acct_list);
96 97
97/* 98/*
98 * Called whenever the timer says to check the free space. 99 * Called whenever the timer says to check the free space.
99 */ 100 */
100static void acct_timeout(unsigned long unused) 101static void acct_timeout(unsigned long x)
101{ 102{
102 acct_globals.needcheck = 1; 103 struct bsd_acct_struct *acct = (struct bsd_acct_struct *)x;
104 acct->needcheck = 1;
103} 105}
104 106
105/* 107/*
106 * Check the amount of free space and suspend/resume accordingly. 108 * Check the amount of free space and suspend/resume accordingly.
107 */ 109 */
108static int check_free_space(struct file *file) 110static int check_free_space(struct bsd_acct_struct *acct, struct file *file)
109{ 111{
110 struct kstatfs sbuf; 112 struct kstatfs sbuf;
111 int res; 113 int res;
@@ -113,11 +115,11 @@ static int check_free_space(struct file *file)
113 sector_t resume; 115 sector_t resume;
114 sector_t suspend; 116 sector_t suspend;
115 117
116 spin_lock(&acct_globals.lock); 118 spin_lock(&acct_lock);
117 res = acct_globals.active; 119 res = acct->active;
118 if (!file || !acct_globals.needcheck) 120 if (!file || !acct->needcheck)
119 goto out; 121 goto out;
120 spin_unlock(&acct_globals.lock); 122 spin_unlock(&acct_lock);
121 123
122 /* May block */ 124 /* May block */
123 if (vfs_statfs(file->f_path.dentry, &sbuf)) 125 if (vfs_statfs(file->f_path.dentry, &sbuf))
@@ -136,35 +138,35 @@ static int check_free_space(struct file *file)
136 act = 0; 138 act = 0;
137 139
138 /* 140 /*
139 * If some joker switched acct_globals.file under us we'ld better be 141 * If some joker switched acct->file under us we'ld better be
140 * silent and _not_ touch anything. 142 * silent and _not_ touch anything.
141 */ 143 */
142 spin_lock(&acct_globals.lock); 144 spin_lock(&acct_lock);
143 if (file != acct_globals.file) { 145 if (file != acct->file) {
144 if (act) 146 if (act)
145 res = act>0; 147 res = act>0;
146 goto out; 148 goto out;
147 } 149 }
148 150
149 if (acct_globals.active) { 151 if (acct->active) {
150 if (act < 0) { 152 if (act < 0) {
151 acct_globals.active = 0; 153 acct->active = 0;
152 printk(KERN_INFO "Process accounting paused\n"); 154 printk(KERN_INFO "Process accounting paused\n");
153 } 155 }
154 } else { 156 } else {
155 if (act > 0) { 157 if (act > 0) {
156 acct_globals.active = 1; 158 acct->active = 1;
157 printk(KERN_INFO "Process accounting resumed\n"); 159 printk(KERN_INFO "Process accounting resumed\n");
158 } 160 }
159 } 161 }
160 162
161 del_timer(&acct_globals.timer); 163 del_timer(&acct->timer);
162 acct_globals.needcheck = 0; 164 acct->needcheck = 0;
163 acct_globals.timer.expires = jiffies + ACCT_TIMEOUT*HZ; 165 acct->timer.expires = jiffies + ACCT_TIMEOUT*HZ;
164 add_timer(&acct_globals.timer); 166 add_timer(&acct->timer);
165 res = acct_globals.active; 167 res = acct->active;
166out: 168out:
167 spin_unlock(&acct_globals.lock); 169 spin_unlock(&acct_lock);
168 return res; 170 return res;
169} 171}
170 172
@@ -172,39 +174,41 @@ out:
172 * Close the old accounting file (if currently open) and then replace 174 * Close the old accounting file (if currently open) and then replace
173 * it with file (if non-NULL). 175 * it with file (if non-NULL).
174 * 176 *
175 * NOTE: acct_globals.lock MUST be held on entry and exit. 177 * NOTE: acct_lock MUST be held on entry and exit.
176 */ 178 */
177static void acct_file_reopen(struct file *file) 179static void acct_file_reopen(struct bsd_acct_struct *acct, struct file *file,
180 struct pid_namespace *ns)
178{ 181{
179 struct file *old_acct = NULL; 182 struct file *old_acct = NULL;
180 struct pid_namespace *old_ns = NULL; 183 struct pid_namespace *old_ns = NULL;
181 184
182 if (acct_globals.file) { 185 if (acct->file) {
183 old_acct = acct_globals.file; 186 old_acct = acct->file;
184 old_ns = acct_globals.ns; 187 old_ns = acct->ns;
185 del_timer(&acct_globals.timer); 188 del_timer(&acct->timer);
186 acct_globals.active = 0; 189 acct->active = 0;
187 acct_globals.needcheck = 0; 190 acct->needcheck = 0;
188 acct_globals.file = NULL; 191 acct->file = NULL;
192 acct->ns = NULL;
193 list_del(&acct->list);
189 } 194 }
190 if (file) { 195 if (file) {
191 acct_globals.file = file; 196 acct->file = file;
192 acct_globals.ns = get_pid_ns(task_active_pid_ns(current)); 197 acct->ns = ns;
193 acct_globals.needcheck = 0; 198 acct->needcheck = 0;
194 acct_globals.active = 1; 199 acct->active = 1;
200 list_add(&acct->list, &acct_list);
195 /* It's been deleted if it was used before so this is safe */ 201 /* It's been deleted if it was used before so this is safe */
196 init_timer(&acct_globals.timer); 202 setup_timer(&acct->timer, acct_timeout, (unsigned long)acct);
197 acct_globals.timer.function = acct_timeout; 203 acct->timer.expires = jiffies + ACCT_TIMEOUT*HZ;
198 acct_globals.timer.expires = jiffies + ACCT_TIMEOUT*HZ; 204 add_timer(&acct->timer);
199 add_timer(&acct_globals.timer);
200 } 205 }
201 if (old_acct) { 206 if (old_acct) {
202 mnt_unpin(old_acct->f_path.mnt); 207 mnt_unpin(old_acct->f_path.mnt);
203 spin_unlock(&acct_globals.lock); 208 spin_unlock(&acct_lock);
204 do_acct_process(old_ns, old_acct); 209 do_acct_process(acct, old_ns, old_acct);
205 filp_close(old_acct, NULL); 210 filp_close(old_acct, NULL);
206 put_pid_ns(old_ns); 211 spin_lock(&acct_lock);
207 spin_lock(&acct_globals.lock);
208 } 212 }
209} 213}
210 214
@@ -212,6 +216,8 @@ static int acct_on(char *name)
212{ 216{
213 struct file *file; 217 struct file *file;
214 int error; 218 int error;
219 struct pid_namespace *ns;
220 struct bsd_acct_struct *acct = NULL;
215 221
216 /* Difference from BSD - they don't do O_APPEND */ 222 /* Difference from BSD - they don't do O_APPEND */
217 file = filp_open(name, O_WRONLY|O_APPEND|O_LARGEFILE, 0); 223 file = filp_open(name, O_WRONLY|O_APPEND|O_LARGEFILE, 0);
@@ -228,18 +234,34 @@ static int acct_on(char *name)
228 return -EIO; 234 return -EIO;
229 } 235 }
230 236
237 ns = task_active_pid_ns(current);
238 if (ns->bacct == NULL) {
239 acct = kzalloc(sizeof(struct bsd_acct_struct), GFP_KERNEL);
240 if (acct == NULL) {
241 filp_close(file, NULL);
242 return -ENOMEM;
243 }
244 }
245
231 error = security_acct(file); 246 error = security_acct(file);
232 if (error) { 247 if (error) {
248 kfree(acct);
233 filp_close(file, NULL); 249 filp_close(file, NULL);
234 return error; 250 return error;
235 } 251 }
236 252
237 spin_lock(&acct_globals.lock); 253 spin_lock(&acct_lock);
254 if (ns->bacct == NULL) {
255 ns->bacct = acct;
256 acct = NULL;
257 }
258
238 mnt_pin(file->f_path.mnt); 259 mnt_pin(file->f_path.mnt);
239 acct_file_reopen(file); 260 acct_file_reopen(ns->bacct, file, ns);
240 spin_unlock(&acct_globals.lock); 261 spin_unlock(&acct_lock);
241 262
242 mntput(file->f_path.mnt); /* it's pinned, now give up active reference */ 263 mntput(file->f_path.mnt); /* it's pinned, now give up active reference */
264 kfree(acct);
243 265
244 return 0; 266 return 0;
245} 267}
@@ -269,11 +291,17 @@ asmlinkage long sys_acct(const char __user *name)
269 error = acct_on(tmp); 291 error = acct_on(tmp);
270 putname(tmp); 292 putname(tmp);
271 } else { 293 } else {
294 struct bsd_acct_struct *acct;
295
296 acct = task_active_pid_ns(current)->bacct;
297 if (acct == NULL)
298 return 0;
299
272 error = security_acct(NULL); 300 error = security_acct(NULL);
273 if (!error) { 301 if (!error) {
274 spin_lock(&acct_globals.lock); 302 spin_lock(&acct_lock);
275 acct_file_reopen(NULL); 303 acct_file_reopen(acct, NULL, NULL);
276 spin_unlock(&acct_globals.lock); 304 spin_unlock(&acct_lock);
277 } 305 }
278 } 306 }
279 return error; 307 return error;
@@ -288,10 +316,16 @@ asmlinkage long sys_acct(const char __user *name)
288 */ 316 */
289void acct_auto_close_mnt(struct vfsmount *m) 317void acct_auto_close_mnt(struct vfsmount *m)
290{ 318{
291 spin_lock(&acct_globals.lock); 319 struct bsd_acct_struct *acct;
292 if (acct_globals.file && acct_globals.file->f_path.mnt == m) 320
293 acct_file_reopen(NULL); 321 spin_lock(&acct_lock);
294 spin_unlock(&acct_globals.lock); 322restart:
323 list_for_each_entry(acct, &acct_list, list)
324 if (acct->file && acct->file->f_path.mnt == m) {
325 acct_file_reopen(acct, NULL, NULL);
326 goto restart;
327 }
328 spin_unlock(&acct_lock);
295} 329}
296 330
297/** 331/**
@@ -303,12 +337,31 @@ void acct_auto_close_mnt(struct vfsmount *m)
303 */ 337 */
304void acct_auto_close(struct super_block *sb) 338void acct_auto_close(struct super_block *sb)
305{ 339{
306 spin_lock(&acct_globals.lock); 340 struct bsd_acct_struct *acct;
307 if (acct_globals.file && 341
308 acct_globals.file->f_path.mnt->mnt_sb == sb) { 342 spin_lock(&acct_lock);
309 acct_file_reopen(NULL); 343restart:
344 list_for_each_entry(acct, &acct_list, list)
345 if (acct->file && acct->file->f_path.mnt->mnt_sb == sb) {
346 acct_file_reopen(acct, NULL, NULL);
347 goto restart;
348 }
349 spin_unlock(&acct_lock);
350}
351
352void acct_exit_ns(struct pid_namespace *ns)
353{
354 struct bsd_acct_struct *acct;
355
356 spin_lock(&acct_lock);
357 acct = ns->bacct;
358 if (acct != NULL) {
359 if (acct->file != NULL)
360 acct_file_reopen(acct, NULL, NULL);
361
362 kfree(acct);
310 } 363 }
311 spin_unlock(&acct_globals.lock); 364 spin_unlock(&acct_lock);
312} 365}
313 366
314/* 367/*
@@ -425,7 +478,8 @@ static u32 encode_float(u64 value)
425/* 478/*
426 * do_acct_process does all actual work. Caller holds the reference to file. 479 * do_acct_process does all actual work. Caller holds the reference to file.
427 */ 480 */
428static void do_acct_process(struct pid_namespace *ns, struct file *file) 481static void do_acct_process(struct bsd_acct_struct *acct,
482 struct pid_namespace *ns, struct file *file)
429{ 483{
430 struct pacct_struct *pacct = &current->signal->pacct; 484 struct pacct_struct *pacct = &current->signal->pacct;
431 acct_t ac; 485 acct_t ac;
@@ -440,7 +494,7 @@ static void do_acct_process(struct pid_namespace *ns, struct file *file)
440 * First check to see if there is enough free_space to continue 494 * First check to see if there is enough free_space to continue
441 * the process accounting system. 495 * the process accounting system.
442 */ 496 */
443 if (!check_free_space(file)) 497 if (!check_free_space(acct, file))
444 return; 498 return;
445 499
446 /* 500 /*
@@ -577,34 +631,46 @@ void acct_collect(long exitcode, int group_dead)
577 spin_unlock_irq(&current->sighand->siglock); 631 spin_unlock_irq(&current->sighand->siglock);
578} 632}
579 633
580/** 634static void acct_process_in_ns(struct pid_namespace *ns)
581 * acct_process - now just a wrapper around do_acct_process
582 * @exitcode: task exit code
583 *
584 * handles process accounting for an exiting task
585 */
586void acct_process(void)
587{ 635{
588 struct file *file = NULL; 636 struct file *file = NULL;
589 struct pid_namespace *ns; 637 struct bsd_acct_struct *acct;
590 638
639 acct = ns->bacct;
591 /* 640 /*
592 * accelerate the common fastpath: 641 * accelerate the common fastpath:
593 */ 642 */
594 if (!acct_globals.file) 643 if (!acct || !acct->file)
595 return; 644 return;
596 645
597 spin_lock(&acct_globals.lock); 646 spin_lock(&acct_lock);
598 file = acct_globals.file; 647 file = acct->file;
599 if (unlikely(!file)) { 648 if (unlikely(!file)) {
600 spin_unlock(&acct_globals.lock); 649 spin_unlock(&acct_lock);
601 return; 650 return;
602 } 651 }
603 get_file(file); 652 get_file(file);
604 ns = get_pid_ns(acct_globals.ns); 653 spin_unlock(&acct_lock);
605 spin_unlock(&acct_globals.lock);
606 654
607 do_acct_process(ns, file); 655 do_acct_process(acct, ns, file);
608 fput(file); 656 fput(file);
609 put_pid_ns(ns); 657}
658
659/**
660 * acct_process - now just a wrapper around acct_process_in_ns,
661 * which in turn is a wrapper around do_acct_process.
662 *
663 * handles process accounting for an exiting task
664 */
665void acct_process(void)
666{
667 struct pid_namespace *ns;
668
669 /*
670 * This loop is safe lockless, since current is still
671 * alive and holds its namespace, which in turn holds
672 * its parent.
673 */
674 for (ns = task_active_pid_ns(current); ns != NULL; ns = ns->parent)
675 acct_process_in_ns(ns);
610} 676}
diff --git a/kernel/auditsc.c b/kernel/auditsc.c
index c10e7aae04d7..4699950e65bd 100644
--- a/kernel/auditsc.c
+++ b/kernel/auditsc.c
@@ -1476,7 +1476,8 @@ void audit_syscall_entry(int arch, int major,
1476 struct audit_context *context = tsk->audit_context; 1476 struct audit_context *context = tsk->audit_context;
1477 enum audit_state state; 1477 enum audit_state state;
1478 1478
1479 BUG_ON(!context); 1479 if (unlikely(!context))
1480 return;
1480 1481
1481 /* 1482 /*
1482 * This happens only on certain architectures that make system 1483 * This happens only on certain architectures that make system
diff --git a/kernel/capability.c b/kernel/capability.c
index 901e0fdc3fff..0101e847603e 100644
--- a/kernel/capability.c
+++ b/kernel/capability.c
@@ -115,11 +115,208 @@ static int cap_validate_magic(cap_user_header_t header, unsigned *tocopy)
115 return 0; 115 return 0;
116} 116}
117 117
118#ifndef CONFIG_SECURITY_FILE_CAPABILITIES
119
120/*
121 * Without filesystem capability support, we nominally support one process
122 * setting the capabilities of another
123 */
124static inline int cap_get_target_pid(pid_t pid, kernel_cap_t *pEp,
125 kernel_cap_t *pIp, kernel_cap_t *pPp)
126{
127 struct task_struct *target;
128 int ret;
129
130 spin_lock(&task_capability_lock);
131 read_lock(&tasklist_lock);
132
133 if (pid && pid != task_pid_vnr(current)) {
134 target = find_task_by_vpid(pid);
135 if (!target) {
136 ret = -ESRCH;
137 goto out;
138 }
139 } else
140 target = current;
141
142 ret = security_capget(target, pEp, pIp, pPp);
143
144out:
145 read_unlock(&tasklist_lock);
146 spin_unlock(&task_capability_lock);
147
148 return ret;
149}
150
151/*
152 * cap_set_pg - set capabilities for all processes in a given process
153 * group. We call this holding task_capability_lock and tasklist_lock.
154 */
155static inline int cap_set_pg(int pgrp_nr, kernel_cap_t *effective,
156 kernel_cap_t *inheritable,
157 kernel_cap_t *permitted)
158{
159 struct task_struct *g, *target;
160 int ret = -EPERM;
161 int found = 0;
162 struct pid *pgrp;
163
164 spin_lock(&task_capability_lock);
165 read_lock(&tasklist_lock);
166
167 pgrp = find_vpid(pgrp_nr);
168 do_each_pid_task(pgrp, PIDTYPE_PGID, g) {
169 target = g;
170 while_each_thread(g, target) {
171 if (!security_capset_check(target, effective,
172 inheritable, permitted)) {
173 security_capset_set(target, effective,
174 inheritable, permitted);
175 ret = 0;
176 }
177 found = 1;
178 }
179 } while_each_pid_task(pgrp, PIDTYPE_PGID, g);
180
181 read_unlock(&tasklist_lock);
182 spin_unlock(&task_capability_lock);
183
184 if (!found)
185 ret = 0;
186 return ret;
187}
188
118/* 189/*
119 * For sys_getproccap() and sys_setproccap(), any of the three 190 * cap_set_all - set capabilities for all processes other than init
120 * capability set pointers may be NULL -- indicating that that set is 191 * and self. We call this holding task_capability_lock and tasklist_lock.
121 * uninteresting and/or not to be changed.
122 */ 192 */
193static inline int cap_set_all(kernel_cap_t *effective,
194 kernel_cap_t *inheritable,
195 kernel_cap_t *permitted)
196{
197 struct task_struct *g, *target;
198 int ret = -EPERM;
199 int found = 0;
200
201 spin_lock(&task_capability_lock);
202 read_lock(&tasklist_lock);
203
204 do_each_thread(g, target) {
205 if (target == current
206 || is_container_init(target->group_leader))
207 continue;
208 found = 1;
209 if (security_capset_check(target, effective, inheritable,
210 permitted))
211 continue;
212 ret = 0;
213 security_capset_set(target, effective, inheritable, permitted);
214 } while_each_thread(g, target);
215
216 read_unlock(&tasklist_lock);
217 spin_unlock(&task_capability_lock);
218
219 if (!found)
220 ret = 0;
221
222 return ret;
223}
224
225/*
226 * Given the target pid does not refer to the current process we
227 * need more elaborate support... (This support is not present when
228 * filesystem capabilities are configured.)
229 */
230static inline int do_sys_capset_other_tasks(pid_t pid, kernel_cap_t *effective,
231 kernel_cap_t *inheritable,
232 kernel_cap_t *permitted)
233{
234 struct task_struct *target;
235 int ret;
236
237 if (!capable(CAP_SETPCAP))
238 return -EPERM;
239
240 if (pid == -1) /* all procs other than current and init */
241 return cap_set_all(effective, inheritable, permitted);
242
243 else if (pid < 0) /* all procs in process group */
244 return cap_set_pg(-pid, effective, inheritable, permitted);
245
246 /* target != current */
247 spin_lock(&task_capability_lock);
248 read_lock(&tasklist_lock);
249
250 target = find_task_by_vpid(pid);
251 if (!target)
252 ret = -ESRCH;
253 else {
254 ret = security_capset_check(target, effective, inheritable,
255 permitted);
256
257 /* having verified that the proposed changes are legal,
258 we now put them into effect. */
259 if (!ret)
260 security_capset_set(target, effective, inheritable,
261 permitted);
262 }
263
264 read_unlock(&tasklist_lock);
265 spin_unlock(&task_capability_lock);
266
267 return ret;
268}
269
270#else /* ie., def CONFIG_SECURITY_FILE_CAPABILITIES */
271
272/*
273 * If we have configured with filesystem capability support, then the
274 * only thing that can change the capabilities of the current process
275 * is the current process. As such, we can't be in this code at the
276 * same time as we are in the process of setting capabilities in this
277 * process. The net result is that we can limit our use of locks to
278 * when we are reading the caps of another process.
279 */
280static inline int cap_get_target_pid(pid_t pid, kernel_cap_t *pEp,
281 kernel_cap_t *pIp, kernel_cap_t *pPp)
282{
283 int ret;
284
285 if (pid && (pid != task_pid_vnr(current))) {
286 struct task_struct *target;
287
288 spin_lock(&task_capability_lock);
289 read_lock(&tasklist_lock);
290
291 target = find_task_by_vpid(pid);
292 if (!target)
293 ret = -ESRCH;
294 else
295 ret = security_capget(target, pEp, pIp, pPp);
296
297 read_unlock(&tasklist_lock);
298 spin_unlock(&task_capability_lock);
299 } else
300 ret = security_capget(current, pEp, pIp, pPp);
301
302 return ret;
303}
304
305/*
306 * With filesystem capability support configured, the kernel does not
307 * permit the changing of capabilities in one process by another
308 * process. (CAP_SETPCAP has much less broad semantics when configured
309 * this way.)
310 */
311static inline int do_sys_capset_other_tasks(pid_t pid,
312 kernel_cap_t *effective,
313 kernel_cap_t *inheritable,
314 kernel_cap_t *permitted)
315{
316 return -EPERM;
317}
318
319#endif /* ie., ndef CONFIG_SECURITY_FILE_CAPABILITIES */
123 320
124/* 321/*
125 * Atomically modify the effective capabilities returning the original 322 * Atomically modify the effective capabilities returning the original
@@ -155,7 +352,6 @@ asmlinkage long sys_capget(cap_user_header_t header, cap_user_data_t dataptr)
155{ 352{
156 int ret = 0; 353 int ret = 0;
157 pid_t pid; 354 pid_t pid;
158 struct task_struct *target;
159 unsigned tocopy; 355 unsigned tocopy;
160 kernel_cap_t pE, pI, pP; 356 kernel_cap_t pE, pI, pP;
161 357
@@ -169,23 +365,7 @@ asmlinkage long sys_capget(cap_user_header_t header, cap_user_data_t dataptr)
169 if (pid < 0) 365 if (pid < 0)
170 return -EINVAL; 366 return -EINVAL;
171 367
172 spin_lock(&task_capability_lock); 368 ret = cap_get_target_pid(pid, &pE, &pI, &pP);
173 read_lock(&tasklist_lock);
174
175 if (pid && pid != task_pid_vnr(current)) {
176 target = find_task_by_vpid(pid);
177 if (!target) {
178 ret = -ESRCH;
179 goto out;
180 }
181 } else
182 target = current;
183
184 ret = security_capget(target, &pE, &pI, &pP);
185
186out:
187 read_unlock(&tasklist_lock);
188 spin_unlock(&task_capability_lock);
189 369
190 if (!ret) { 370 if (!ret) {
191 struct __user_cap_data_struct kdata[_KERNEL_CAPABILITY_U32S]; 371 struct __user_cap_data_struct kdata[_KERNEL_CAPABILITY_U32S];
@@ -216,7 +396,6 @@ out:
216 * before modification is attempted and the application 396 * before modification is attempted and the application
217 * fails. 397 * fails.
218 */ 398 */
219
220 if (copy_to_user(dataptr, kdata, tocopy 399 if (copy_to_user(dataptr, kdata, tocopy
221 * sizeof(struct __user_cap_data_struct))) { 400 * sizeof(struct __user_cap_data_struct))) {
222 return -EFAULT; 401 return -EFAULT;
@@ -226,70 +405,8 @@ out:
226 return ret; 405 return ret;
227} 406}
228 407
229/*
230 * cap_set_pg - set capabilities for all processes in a given process
231 * group. We call this holding task_capability_lock and tasklist_lock.
232 */
233static inline int cap_set_pg(int pgrp_nr, kernel_cap_t *effective,
234 kernel_cap_t *inheritable,
235 kernel_cap_t *permitted)
236{
237 struct task_struct *g, *target;
238 int ret = -EPERM;
239 int found = 0;
240 struct pid *pgrp;
241
242 pgrp = find_vpid(pgrp_nr);
243 do_each_pid_task(pgrp, PIDTYPE_PGID, g) {
244 target = g;
245 while_each_thread(g, target) {
246 if (!security_capset_check(target, effective,
247 inheritable,
248 permitted)) {
249 security_capset_set(target, effective,
250 inheritable,
251 permitted);
252 ret = 0;
253 }
254 found = 1;
255 }
256 } while_each_pid_task(pgrp, PIDTYPE_PGID, g);
257
258 if (!found)
259 ret = 0;
260 return ret;
261}
262
263/*
264 * cap_set_all - set capabilities for all processes other than init
265 * and self. We call this holding task_capability_lock and tasklist_lock.
266 */
267static inline int cap_set_all(kernel_cap_t *effective,
268 kernel_cap_t *inheritable,
269 kernel_cap_t *permitted)
270{
271 struct task_struct *g, *target;
272 int ret = -EPERM;
273 int found = 0;
274
275 do_each_thread(g, target) {
276 if (target == current || is_container_init(target->group_leader))
277 continue;
278 found = 1;
279 if (security_capset_check(target, effective, inheritable,
280 permitted))
281 continue;
282 ret = 0;
283 security_capset_set(target, effective, inheritable, permitted);
284 } while_each_thread(g, target);
285
286 if (!found)
287 ret = 0;
288 return ret;
289}
290
291/** 408/**
292 * sys_capset - set capabilities for a process or a group of processes 409 * sys_capset - set capabilities for a process or (*) a group of processes
293 * @header: pointer to struct that contains capability version and 410 * @header: pointer to struct that contains capability version and
294 * target pid data 411 * target pid data
295 * @data: pointer to struct that contains the effective, permitted, 412 * @data: pointer to struct that contains the effective, permitted,
@@ -313,7 +430,6 @@ asmlinkage long sys_capset(cap_user_header_t header, const cap_user_data_t data)
313 struct __user_cap_data_struct kdata[_KERNEL_CAPABILITY_U32S]; 430 struct __user_cap_data_struct kdata[_KERNEL_CAPABILITY_U32S];
314 unsigned i, tocopy; 431 unsigned i, tocopy;
315 kernel_cap_t inheritable, permitted, effective; 432 kernel_cap_t inheritable, permitted, effective;
316 struct task_struct *target;
317 int ret; 433 int ret;
318 pid_t pid; 434 pid_t pid;
319 435
@@ -324,9 +440,6 @@ asmlinkage long sys_capset(cap_user_header_t header, const cap_user_data_t data)
324 if (get_user(pid, &header->pid)) 440 if (get_user(pid, &header->pid))
325 return -EFAULT; 441 return -EFAULT;
326 442
327 if (pid && pid != task_pid_vnr(current) && !capable(CAP_SETPCAP))
328 return -EPERM;
329
330 if (copy_from_user(&kdata, data, tocopy 443 if (copy_from_user(&kdata, data, tocopy
331 * sizeof(struct __user_cap_data_struct))) { 444 * sizeof(struct __user_cap_data_struct))) {
332 return -EFAULT; 445 return -EFAULT;
@@ -344,40 +457,31 @@ asmlinkage long sys_capset(cap_user_header_t header, const cap_user_data_t data)
344 i++; 457 i++;
345 } 458 }
346 459
347 spin_lock(&task_capability_lock); 460 if (pid && (pid != task_pid_vnr(current)))
348 read_lock(&tasklist_lock); 461 ret = do_sys_capset_other_tasks(pid, &effective, &inheritable,
349 462 &permitted);
350 if (pid > 0 && pid != task_pid_vnr(current)) { 463 else {
351 target = find_task_by_vpid(pid); 464 /*
352 if (!target) { 465 * This lock is required even when filesystem
353 ret = -ESRCH; 466 * capability support is configured - it protects the
354 goto out; 467 * sys_capget() call from returning incorrect data in
355 } 468 * the case that the targeted process is not the
356 } else 469 * current one.
357 target = current; 470 */
358 471 spin_lock(&task_capability_lock);
359 ret = 0;
360
361 /* having verified that the proposed changes are legal,
362 we now put them into effect. */
363 if (pid < 0) {
364 if (pid == -1) /* all procs other than current and init */
365 ret = cap_set_all(&effective, &inheritable, &permitted);
366 472
367 else /* all procs in process group */ 473 ret = security_capset_check(current, &effective, &inheritable,
368 ret = cap_set_pg(-pid, &effective, &inheritable,
369 &permitted);
370 } else {
371 ret = security_capset_check(target, &effective, &inheritable,
372 &permitted); 474 &permitted);
475 /*
476 * Having verified that the proposed changes are
477 * legal, we now put them into effect.
478 */
373 if (!ret) 479 if (!ret)
374 security_capset_set(target, &effective, &inheritable, 480 security_capset_set(current, &effective, &inheritable,
375 &permitted); 481 &permitted);
482 spin_unlock(&task_capability_lock);
376 } 483 }
377 484
378out:
379 read_unlock(&tasklist_lock);
380 spin_unlock(&task_capability_lock);
381 485
382 return ret; 486 return ret;
383} 487}
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 15ac0e1e4f4d..66ec9fd21e0c 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -89,11 +89,7 @@ struct cgroupfs_root {
89 /* Hierarchy-specific flags */ 89 /* Hierarchy-specific flags */
90 unsigned long flags; 90 unsigned long flags;
91 91
92 /* The path to use for release notifications. No locking 92 /* The path to use for release notifications. */
93 * between setting and use - so if userspace updates this
94 * while child cgroups exist, you could miss a
95 * notification. We ensure that it's always a valid
96 * NUL-terminated string */
97 char release_agent_path[PATH_MAX]; 93 char release_agent_path[PATH_MAX];
98}; 94};
99 95
@@ -118,7 +114,7 @@ static int root_count;
118 * extra work in the fork/exit path if none of the subsystems need to 114 * extra work in the fork/exit path if none of the subsystems need to
119 * be called. 115 * be called.
120 */ 116 */
121static int need_forkexit_callback; 117static int need_forkexit_callback __read_mostly;
122static int need_mm_owner_callback __read_mostly; 118static int need_mm_owner_callback __read_mostly;
123 119
124/* convenient tests for these bits */ 120/* convenient tests for these bits */
@@ -220,7 +216,7 @@ static struct hlist_head *css_set_hash(struct cgroup_subsys_state *css[])
220 * task until after the first call to cgroup_iter_start(). This 216 * task until after the first call to cgroup_iter_start(). This
221 * reduces the fork()/exit() overhead for people who have cgroups 217 * reduces the fork()/exit() overhead for people who have cgroups
222 * compiled into their kernel but not actually in use */ 218 * compiled into their kernel but not actually in use */
223static int use_task_css_set_links; 219static int use_task_css_set_links __read_mostly;
224 220
225/* When we create or destroy a css_set, the operation simply 221/* When we create or destroy a css_set, the operation simply
226 * takes/releases a reference count on all the cgroups referenced 222 * takes/releases a reference count on all the cgroups referenced
@@ -241,17 +237,20 @@ static int use_task_css_set_links;
241 */ 237 */
242static void unlink_css_set(struct css_set *cg) 238static void unlink_css_set(struct css_set *cg)
243{ 239{
240 struct cg_cgroup_link *link;
241 struct cg_cgroup_link *saved_link;
242
244 write_lock(&css_set_lock); 243 write_lock(&css_set_lock);
245 hlist_del(&cg->hlist); 244 hlist_del(&cg->hlist);
246 css_set_count--; 245 css_set_count--;
247 while (!list_empty(&cg->cg_links)) { 246
248 struct cg_cgroup_link *link; 247 list_for_each_entry_safe(link, saved_link, &cg->cg_links,
249 link = list_entry(cg->cg_links.next, 248 cg_link_list) {
250 struct cg_cgroup_link, cg_link_list);
251 list_del(&link->cg_link_list); 249 list_del(&link->cg_link_list);
252 list_del(&link->cgrp_link_list); 250 list_del(&link->cgrp_link_list);
253 kfree(link); 251 kfree(link);
254 } 252 }
253
255 write_unlock(&css_set_lock); 254 write_unlock(&css_set_lock);
256} 255}
257 256
@@ -363,15 +362,14 @@ static struct css_set *find_existing_css_set(
363static int allocate_cg_links(int count, struct list_head *tmp) 362static int allocate_cg_links(int count, struct list_head *tmp)
364{ 363{
365 struct cg_cgroup_link *link; 364 struct cg_cgroup_link *link;
365 struct cg_cgroup_link *saved_link;
366 int i; 366 int i;
367 INIT_LIST_HEAD(tmp); 367 INIT_LIST_HEAD(tmp);
368 for (i = 0; i < count; i++) { 368 for (i = 0; i < count; i++) {
369 link = kmalloc(sizeof(*link), GFP_KERNEL); 369 link = kmalloc(sizeof(*link), GFP_KERNEL);
370 if (!link) { 370 if (!link) {
371 while (!list_empty(tmp)) { 371 list_for_each_entry_safe(link, saved_link, tmp,
372 link = list_entry(tmp->next, 372 cgrp_link_list) {
373 struct cg_cgroup_link,
374 cgrp_link_list);
375 list_del(&link->cgrp_link_list); 373 list_del(&link->cgrp_link_list);
376 kfree(link); 374 kfree(link);
377 } 375 }
@@ -384,11 +382,10 @@ static int allocate_cg_links(int count, struct list_head *tmp)
384 382
385static void free_cg_links(struct list_head *tmp) 383static void free_cg_links(struct list_head *tmp)
386{ 384{
387 while (!list_empty(tmp)) { 385 struct cg_cgroup_link *link;
388 struct cg_cgroup_link *link; 386 struct cg_cgroup_link *saved_link;
389 link = list_entry(tmp->next, 387
390 struct cg_cgroup_link, 388 list_for_each_entry_safe(link, saved_link, tmp, cgrp_link_list) {
391 cgrp_link_list);
392 list_del(&link->cgrp_link_list); 389 list_del(&link->cgrp_link_list);
393 kfree(link); 390 kfree(link);
394 } 391 }
@@ -415,11 +412,11 @@ static struct css_set *find_css_set(
415 412
416 /* First see if we already have a cgroup group that matches 413 /* First see if we already have a cgroup group that matches
417 * the desired set */ 414 * the desired set */
418 write_lock(&css_set_lock); 415 read_lock(&css_set_lock);
419 res = find_existing_css_set(oldcg, cgrp, template); 416 res = find_existing_css_set(oldcg, cgrp, template);
420 if (res) 417 if (res)
421 get_css_set(res); 418 get_css_set(res);
422 write_unlock(&css_set_lock); 419 read_unlock(&css_set_lock);
423 420
424 if (res) 421 if (res)
425 return res; 422 return res;
@@ -507,10 +504,6 @@ static struct css_set *find_css_set(
507 * knows that the cgroup won't be removed, as cgroup_rmdir() 504 * knows that the cgroup won't be removed, as cgroup_rmdir()
508 * needs that mutex. 505 * needs that mutex.
509 * 506 *
510 * The cgroup_common_file_write handler for operations that modify
511 * the cgroup hierarchy holds cgroup_mutex across the entire operation,
512 * single threading all such cgroup modifications across the system.
513 *
514 * The fork and exit callbacks cgroup_fork() and cgroup_exit(), don't 507 * The fork and exit callbacks cgroup_fork() and cgroup_exit(), don't
515 * (usually) take cgroup_mutex. These are the two most performance 508 * (usually) take cgroup_mutex. These are the two most performance
516 * critical pieces of code here. The exception occurs on cgroup_exit(), 509 * critical pieces of code here. The exception occurs on cgroup_exit(),
@@ -1093,6 +1086,8 @@ static void cgroup_kill_sb(struct super_block *sb) {
1093 struct cgroupfs_root *root = sb->s_fs_info; 1086 struct cgroupfs_root *root = sb->s_fs_info;
1094 struct cgroup *cgrp = &root->top_cgroup; 1087 struct cgroup *cgrp = &root->top_cgroup;
1095 int ret; 1088 int ret;
1089 struct cg_cgroup_link *link;
1090 struct cg_cgroup_link *saved_link;
1096 1091
1097 BUG_ON(!root); 1092 BUG_ON(!root);
1098 1093
@@ -1112,10 +1107,9 @@ static void cgroup_kill_sb(struct super_block *sb) {
1112 * root cgroup 1107 * root cgroup
1113 */ 1108 */
1114 write_lock(&css_set_lock); 1109 write_lock(&css_set_lock);
1115 while (!list_empty(&cgrp->css_sets)) { 1110
1116 struct cg_cgroup_link *link; 1111 list_for_each_entry_safe(link, saved_link, &cgrp->css_sets,
1117 link = list_entry(cgrp->css_sets.next, 1112 cgrp_link_list) {
1118 struct cg_cgroup_link, cgrp_link_list);
1119 list_del(&link->cg_link_list); 1113 list_del(&link->cg_link_list);
1120 list_del(&link->cgrp_link_list); 1114 list_del(&link->cgrp_link_list);
1121 kfree(link); 1115 kfree(link);
@@ -1281,18 +1275,14 @@ int cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
1281} 1275}
1282 1276
1283/* 1277/*
1284 * Attach task with pid 'pid' to cgroup 'cgrp'. Call with 1278 * Attach task with pid 'pid' to cgroup 'cgrp'. Call with cgroup_mutex
1285 * cgroup_mutex, may take task_lock of task 1279 * held. May take task_lock of task
1286 */ 1280 */
1287static int attach_task_by_pid(struct cgroup *cgrp, char *pidbuf) 1281static int attach_task_by_pid(struct cgroup *cgrp, u64 pid)
1288{ 1282{
1289 pid_t pid;
1290 struct task_struct *tsk; 1283 struct task_struct *tsk;
1291 int ret; 1284 int ret;
1292 1285
1293 if (sscanf(pidbuf, "%d", &pid) != 1)
1294 return -EIO;
1295
1296 if (pid) { 1286 if (pid) {
1297 rcu_read_lock(); 1287 rcu_read_lock();
1298 tsk = find_task_by_vpid(pid); 1288 tsk = find_task_by_vpid(pid);
@@ -1318,6 +1308,16 @@ static int attach_task_by_pid(struct cgroup *cgrp, char *pidbuf)
1318 return ret; 1308 return ret;
1319} 1309}
1320 1310
1311static int cgroup_tasks_write(struct cgroup *cgrp, struct cftype *cft, u64 pid)
1312{
1313 int ret;
1314 if (!cgroup_lock_live_group(cgrp))
1315 return -ENODEV;
1316 ret = attach_task_by_pid(cgrp, pid);
1317 cgroup_unlock();
1318 return ret;
1319}
1320
1321/* The various types of files and directories in a cgroup file system */ 1321/* The various types of files and directories in a cgroup file system */
1322enum cgroup_filetype { 1322enum cgroup_filetype {
1323 FILE_ROOT, 1323 FILE_ROOT,
@@ -1327,12 +1327,54 @@ enum cgroup_filetype {
1327 FILE_RELEASE_AGENT, 1327 FILE_RELEASE_AGENT,
1328}; 1328};
1329 1329
1330/**
1331 * cgroup_lock_live_group - take cgroup_mutex and check that cgrp is alive.
1332 * @cgrp: the cgroup to be checked for liveness
1333 *
1334 * On success, returns true; the lock should be later released with
1335 * cgroup_unlock(). On failure returns false with no lock held.
1336 */
1337bool cgroup_lock_live_group(struct cgroup *cgrp)
1338{
1339 mutex_lock(&cgroup_mutex);
1340 if (cgroup_is_removed(cgrp)) {
1341 mutex_unlock(&cgroup_mutex);
1342 return false;
1343 }
1344 return true;
1345}
1346
1347static int cgroup_release_agent_write(struct cgroup *cgrp, struct cftype *cft,
1348 const char *buffer)
1349{
1350 BUILD_BUG_ON(sizeof(cgrp->root->release_agent_path) < PATH_MAX);
1351 if (!cgroup_lock_live_group(cgrp))
1352 return -ENODEV;
1353 strcpy(cgrp->root->release_agent_path, buffer);
1354 cgroup_unlock();
1355 return 0;
1356}
1357
1358static int cgroup_release_agent_show(struct cgroup *cgrp, struct cftype *cft,
1359 struct seq_file *seq)
1360{
1361 if (!cgroup_lock_live_group(cgrp))
1362 return -ENODEV;
1363 seq_puts(seq, cgrp->root->release_agent_path);
1364 seq_putc(seq, '\n');
1365 cgroup_unlock();
1366 return 0;
1367}
1368
1369/* A buffer size big enough for numbers or short strings */
1370#define CGROUP_LOCAL_BUFFER_SIZE 64
1371
1330static ssize_t cgroup_write_X64(struct cgroup *cgrp, struct cftype *cft, 1372static ssize_t cgroup_write_X64(struct cgroup *cgrp, struct cftype *cft,
1331 struct file *file, 1373 struct file *file,
1332 const char __user *userbuf, 1374 const char __user *userbuf,
1333 size_t nbytes, loff_t *unused_ppos) 1375 size_t nbytes, loff_t *unused_ppos)
1334{ 1376{
1335 char buffer[64]; 1377 char buffer[CGROUP_LOCAL_BUFFER_SIZE];
1336 int retval = 0; 1378 int retval = 0;
1337 char *end; 1379 char *end;
1338 1380
@@ -1361,68 +1403,36 @@ static ssize_t cgroup_write_X64(struct cgroup *cgrp, struct cftype *cft,
1361 return retval; 1403 return retval;
1362} 1404}
1363 1405
1364static ssize_t cgroup_common_file_write(struct cgroup *cgrp, 1406static ssize_t cgroup_write_string(struct cgroup *cgrp, struct cftype *cft,
1365 struct cftype *cft, 1407 struct file *file,
1366 struct file *file, 1408 const char __user *userbuf,
1367 const char __user *userbuf, 1409 size_t nbytes, loff_t *unused_ppos)
1368 size_t nbytes, loff_t *unused_ppos)
1369{ 1410{
1370 enum cgroup_filetype type = cft->private; 1411 char local_buffer[CGROUP_LOCAL_BUFFER_SIZE];
1371 char *buffer;
1372 int retval = 0; 1412 int retval = 0;
1413 size_t max_bytes = cft->max_write_len;
1414 char *buffer = local_buffer;
1373 1415
1374 if (nbytes >= PATH_MAX) 1416 if (!max_bytes)
1417 max_bytes = sizeof(local_buffer) - 1;
1418 if (nbytes >= max_bytes)
1375 return -E2BIG; 1419 return -E2BIG;
1376 1420 /* Allocate a dynamic buffer if we need one */
1377 /* +1 for nul-terminator */ 1421 if (nbytes >= sizeof(local_buffer)) {
1378 buffer = kmalloc(nbytes + 1, GFP_KERNEL); 1422 buffer = kmalloc(nbytes + 1, GFP_KERNEL);
1379 if (buffer == NULL) 1423 if (buffer == NULL)
1380 return -ENOMEM; 1424 return -ENOMEM;
1381
1382 if (copy_from_user(buffer, userbuf, nbytes)) {
1383 retval = -EFAULT;
1384 goto out1;
1385 } 1425 }
1386 buffer[nbytes] = 0; /* nul-terminate */ 1426 if (nbytes && copy_from_user(buffer, userbuf, nbytes))
1387 strstrip(buffer); /* strip -just- trailing whitespace */ 1427 return -EFAULT;
1388
1389 mutex_lock(&cgroup_mutex);
1390 1428
1391 /* 1429 buffer[nbytes] = 0; /* nul-terminate */
1392 * This was already checked for in cgroup_file_write(), but 1430 strstrip(buffer);
1393 * check again now we're holding cgroup_mutex. 1431 retval = cft->write_string(cgrp, cft, buffer);
1394 */ 1432 if (!retval)
1395 if (cgroup_is_removed(cgrp)) {
1396 retval = -ENODEV;
1397 goto out2;
1398 }
1399
1400 switch (type) {
1401 case FILE_TASKLIST:
1402 retval = attach_task_by_pid(cgrp, buffer);
1403 break;
1404 case FILE_NOTIFY_ON_RELEASE:
1405 clear_bit(CGRP_RELEASABLE, &cgrp->flags);
1406 if (simple_strtoul(buffer, NULL, 10) != 0)
1407 set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
1408 else
1409 clear_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
1410 break;
1411 case FILE_RELEASE_AGENT:
1412 BUILD_BUG_ON(sizeof(cgrp->root->release_agent_path) < PATH_MAX);
1413 strcpy(cgrp->root->release_agent_path, buffer);
1414 break;
1415 default:
1416 retval = -EINVAL;
1417 goto out2;
1418 }
1419
1420 if (retval == 0)
1421 retval = nbytes; 1433 retval = nbytes;
1422out2: 1434 if (buffer != local_buffer)
1423 mutex_unlock(&cgroup_mutex); 1435 kfree(buffer);
1424out1:
1425 kfree(buffer);
1426 return retval; 1436 return retval;
1427} 1437}
1428 1438
@@ -1438,6 +1448,8 @@ static ssize_t cgroup_file_write(struct file *file, const char __user *buf,
1438 return cft->write(cgrp, cft, file, buf, nbytes, ppos); 1448 return cft->write(cgrp, cft, file, buf, nbytes, ppos);
1439 if (cft->write_u64 || cft->write_s64) 1449 if (cft->write_u64 || cft->write_s64)
1440 return cgroup_write_X64(cgrp, cft, file, buf, nbytes, ppos); 1450 return cgroup_write_X64(cgrp, cft, file, buf, nbytes, ppos);
1451 if (cft->write_string)
1452 return cgroup_write_string(cgrp, cft, file, buf, nbytes, ppos);
1441 if (cft->trigger) { 1453 if (cft->trigger) {
1442 int ret = cft->trigger(cgrp, (unsigned int)cft->private); 1454 int ret = cft->trigger(cgrp, (unsigned int)cft->private);
1443 return ret ? ret : nbytes; 1455 return ret ? ret : nbytes;
@@ -1450,7 +1462,7 @@ static ssize_t cgroup_read_u64(struct cgroup *cgrp, struct cftype *cft,
1450 char __user *buf, size_t nbytes, 1462 char __user *buf, size_t nbytes,
1451 loff_t *ppos) 1463 loff_t *ppos)
1452{ 1464{
1453 char tmp[64]; 1465 char tmp[CGROUP_LOCAL_BUFFER_SIZE];
1454 u64 val = cft->read_u64(cgrp, cft); 1466 u64 val = cft->read_u64(cgrp, cft);
1455 int len = sprintf(tmp, "%llu\n", (unsigned long long) val); 1467 int len = sprintf(tmp, "%llu\n", (unsigned long long) val);
1456 1468
@@ -1462,56 +1474,13 @@ static ssize_t cgroup_read_s64(struct cgroup *cgrp, struct cftype *cft,
1462 char __user *buf, size_t nbytes, 1474 char __user *buf, size_t nbytes,
1463 loff_t *ppos) 1475 loff_t *ppos)
1464{ 1476{
1465 char tmp[64]; 1477 char tmp[CGROUP_LOCAL_BUFFER_SIZE];
1466 s64 val = cft->read_s64(cgrp, cft); 1478 s64 val = cft->read_s64(cgrp, cft);
1467 int len = sprintf(tmp, "%lld\n", (long long) val); 1479 int len = sprintf(tmp, "%lld\n", (long long) val);
1468 1480
1469 return simple_read_from_buffer(buf, nbytes, ppos, tmp, len); 1481 return simple_read_from_buffer(buf, nbytes, ppos, tmp, len);
1470} 1482}
1471 1483
1472static ssize_t cgroup_common_file_read(struct cgroup *cgrp,
1473 struct cftype *cft,
1474 struct file *file,
1475 char __user *buf,
1476 size_t nbytes, loff_t *ppos)
1477{
1478 enum cgroup_filetype type = cft->private;
1479 char *page;
1480 ssize_t retval = 0;
1481 char *s;
1482
1483 if (!(page = (char *)__get_free_page(GFP_KERNEL)))
1484 return -ENOMEM;
1485
1486 s = page;
1487
1488 switch (type) {
1489 case FILE_RELEASE_AGENT:
1490 {
1491 struct cgroupfs_root *root;
1492 size_t n;
1493 mutex_lock(&cgroup_mutex);
1494 root = cgrp->root;
1495 n = strnlen(root->release_agent_path,
1496 sizeof(root->release_agent_path));
1497 n = min(n, (size_t) PAGE_SIZE);
1498 strncpy(s, root->release_agent_path, n);
1499 mutex_unlock(&cgroup_mutex);
1500 s += n;
1501 break;
1502 }
1503 default:
1504 retval = -EINVAL;
1505 goto out;
1506 }
1507 *s++ = '\n';
1508
1509 retval = simple_read_from_buffer(buf, nbytes, ppos, page, s - page);
1510out:
1511 free_page((unsigned long)page);
1512 return retval;
1513}
1514
1515static ssize_t cgroup_file_read(struct file *file, char __user *buf, 1484static ssize_t cgroup_file_read(struct file *file, char __user *buf,
1516 size_t nbytes, loff_t *ppos) 1485 size_t nbytes, loff_t *ppos)
1517{ 1486{
@@ -1569,6 +1538,7 @@ int cgroup_seqfile_release(struct inode *inode, struct file *file)
1569 1538
1570static struct file_operations cgroup_seqfile_operations = { 1539static struct file_operations cgroup_seqfile_operations = {
1571 .read = seq_read, 1540 .read = seq_read,
1541 .write = cgroup_file_write,
1572 .llseek = seq_lseek, 1542 .llseek = seq_lseek,
1573 .release = cgroup_seqfile_release, 1543 .release = cgroup_seqfile_release,
1574}; 1544};
@@ -1756,15 +1726,11 @@ int cgroup_add_files(struct cgroup *cgrp,
1756int cgroup_task_count(const struct cgroup *cgrp) 1726int cgroup_task_count(const struct cgroup *cgrp)
1757{ 1727{
1758 int count = 0; 1728 int count = 0;
1759 struct list_head *l; 1729 struct cg_cgroup_link *link;
1760 1730
1761 read_lock(&css_set_lock); 1731 read_lock(&css_set_lock);
1762 l = cgrp->css_sets.next; 1732 list_for_each_entry(link, &cgrp->css_sets, cgrp_link_list) {
1763 while (l != &cgrp->css_sets) {
1764 struct cg_cgroup_link *link =
1765 list_entry(l, struct cg_cgroup_link, cgrp_link_list);
1766 count += atomic_read(&link->cg->ref.refcount); 1733 count += atomic_read(&link->cg->ref.refcount);
1767 l = l->next;
1768 } 1734 }
1769 read_unlock(&css_set_lock); 1735 read_unlock(&css_set_lock);
1770 return count; 1736 return count;
@@ -2227,6 +2193,18 @@ static u64 cgroup_read_notify_on_release(struct cgroup *cgrp,
2227 return notify_on_release(cgrp); 2193 return notify_on_release(cgrp);
2228} 2194}
2229 2195
2196static int cgroup_write_notify_on_release(struct cgroup *cgrp,
2197 struct cftype *cft,
2198 u64 val)
2199{
2200 clear_bit(CGRP_RELEASABLE, &cgrp->flags);
2201 if (val)
2202 set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
2203 else
2204 clear_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
2205 return 0;
2206}
2207
2230/* 2208/*
2231 * for the common functions, 'private' gives the type of file 2209 * for the common functions, 'private' gives the type of file
2232 */ 2210 */
@@ -2235,7 +2213,7 @@ static struct cftype files[] = {
2235 .name = "tasks", 2213 .name = "tasks",
2236 .open = cgroup_tasks_open, 2214 .open = cgroup_tasks_open,
2237 .read = cgroup_tasks_read, 2215 .read = cgroup_tasks_read,
2238 .write = cgroup_common_file_write, 2216 .write_u64 = cgroup_tasks_write,
2239 .release = cgroup_tasks_release, 2217 .release = cgroup_tasks_release,
2240 .private = FILE_TASKLIST, 2218 .private = FILE_TASKLIST,
2241 }, 2219 },
@@ -2243,15 +2221,16 @@ static struct cftype files[] = {
2243 { 2221 {
2244 .name = "notify_on_release", 2222 .name = "notify_on_release",
2245 .read_u64 = cgroup_read_notify_on_release, 2223 .read_u64 = cgroup_read_notify_on_release,
2246 .write = cgroup_common_file_write, 2224 .write_u64 = cgroup_write_notify_on_release,
2247 .private = FILE_NOTIFY_ON_RELEASE, 2225 .private = FILE_NOTIFY_ON_RELEASE,
2248 }, 2226 },
2249}; 2227};
2250 2228
2251static struct cftype cft_release_agent = { 2229static struct cftype cft_release_agent = {
2252 .name = "release_agent", 2230 .name = "release_agent",
2253 .read = cgroup_common_file_read, 2231 .read_seq_string = cgroup_release_agent_show,
2254 .write = cgroup_common_file_write, 2232 .write_string = cgroup_release_agent_write,
2233 .max_write_len = PATH_MAX,
2255 .private = FILE_RELEASE_AGENT, 2234 .private = FILE_RELEASE_AGENT,
2256}; 2235};
2257 2236
@@ -2869,16 +2848,17 @@ void cgroup_exit(struct task_struct *tsk, int run_callbacks)
2869 * cgroup_clone - clone the cgroup the given subsystem is attached to 2848 * cgroup_clone - clone the cgroup the given subsystem is attached to
2870 * @tsk: the task to be moved 2849 * @tsk: the task to be moved
2871 * @subsys: the given subsystem 2850 * @subsys: the given subsystem
2851 * @nodename: the name for the new cgroup
2872 * 2852 *
2873 * Duplicate the current cgroup in the hierarchy that the given 2853 * Duplicate the current cgroup in the hierarchy that the given
2874 * subsystem is attached to, and move this task into the new 2854 * subsystem is attached to, and move this task into the new
2875 * child. 2855 * child.
2876 */ 2856 */
2877int cgroup_clone(struct task_struct *tsk, struct cgroup_subsys *subsys) 2857int cgroup_clone(struct task_struct *tsk, struct cgroup_subsys *subsys,
2858 char *nodename)
2878{ 2859{
2879 struct dentry *dentry; 2860 struct dentry *dentry;
2880 int ret = 0; 2861 int ret = 0;
2881 char nodename[MAX_CGROUP_TYPE_NAMELEN];
2882 struct cgroup *parent, *child; 2862 struct cgroup *parent, *child;
2883 struct inode *inode; 2863 struct inode *inode;
2884 struct css_set *cg; 2864 struct css_set *cg;
@@ -2903,8 +2883,6 @@ int cgroup_clone(struct task_struct *tsk, struct cgroup_subsys *subsys)
2903 cg = tsk->cgroups; 2883 cg = tsk->cgroups;
2904 parent = task_cgroup(tsk, subsys->subsys_id); 2884 parent = task_cgroup(tsk, subsys->subsys_id);
2905 2885
2906 snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "%d", tsk->pid);
2907
2908 /* Pin the hierarchy */ 2886 /* Pin the hierarchy */
2909 atomic_inc(&parent->root->sb->s_active); 2887 atomic_inc(&parent->root->sb->s_active);
2910 2888
@@ -3078,27 +3056,24 @@ static void cgroup_release_agent(struct work_struct *work)
3078 while (!list_empty(&release_list)) { 3056 while (!list_empty(&release_list)) {
3079 char *argv[3], *envp[3]; 3057 char *argv[3], *envp[3];
3080 int i; 3058 int i;
3081 char *pathbuf; 3059 char *pathbuf = NULL, *agentbuf = NULL;
3082 struct cgroup *cgrp = list_entry(release_list.next, 3060 struct cgroup *cgrp = list_entry(release_list.next,
3083 struct cgroup, 3061 struct cgroup,
3084 release_list); 3062 release_list);
3085 list_del_init(&cgrp->release_list); 3063 list_del_init(&cgrp->release_list);
3086 spin_unlock(&release_list_lock); 3064 spin_unlock(&release_list_lock);
3087 pathbuf = kmalloc(PAGE_SIZE, GFP_KERNEL); 3065 pathbuf = kmalloc(PAGE_SIZE, GFP_KERNEL);
3088 if (!pathbuf) { 3066 if (!pathbuf)
3089 spin_lock(&release_list_lock); 3067 goto continue_free;
3090 continue; 3068 if (cgroup_path(cgrp, pathbuf, PAGE_SIZE) < 0)
3091 } 3069 goto continue_free;
3092 3070 agentbuf = kstrdup(cgrp->root->release_agent_path, GFP_KERNEL);
3093 if (cgroup_path(cgrp, pathbuf, PAGE_SIZE) < 0) { 3071 if (!agentbuf)
3094 kfree(pathbuf); 3072 goto continue_free;
3095 spin_lock(&release_list_lock);
3096 continue;
3097 }
3098 3073
3099 i = 0; 3074 i = 0;
3100 argv[i++] = cgrp->root->release_agent_path; 3075 argv[i++] = agentbuf;
3101 argv[i++] = (char *)pathbuf; 3076 argv[i++] = pathbuf;
3102 argv[i] = NULL; 3077 argv[i] = NULL;
3103 3078
3104 i = 0; 3079 i = 0;
@@ -3112,8 +3087,10 @@ static void cgroup_release_agent(struct work_struct *work)
3112 * be a slow process */ 3087 * be a slow process */
3113 mutex_unlock(&cgroup_mutex); 3088 mutex_unlock(&cgroup_mutex);
3114 call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC); 3089 call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
3115 kfree(pathbuf);
3116 mutex_lock(&cgroup_mutex); 3090 mutex_lock(&cgroup_mutex);
3091 continue_free:
3092 kfree(pathbuf);
3093 kfree(agentbuf);
3117 spin_lock(&release_list_lock); 3094 spin_lock(&release_list_lock);
3118 } 3095 }
3119 spin_unlock(&release_list_lock); 3096 spin_unlock(&release_list_lock);
diff --git a/kernel/cpu.c b/kernel/cpu.c
index cfb1d43ab801..10ba5f1004a5 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -64,6 +64,8 @@ void __init cpu_hotplug_init(void)
64 cpu_hotplug.refcount = 0; 64 cpu_hotplug.refcount = 0;
65} 65}
66 66
67cpumask_t cpu_active_map;
68
67#ifdef CONFIG_HOTPLUG_CPU 69#ifdef CONFIG_HOTPLUG_CPU
68 70
69void get_online_cpus(void) 71void get_online_cpus(void)
@@ -283,6 +285,11 @@ out_allowed:
283 set_cpus_allowed_ptr(current, &old_allowed); 285 set_cpus_allowed_ptr(current, &old_allowed);
284out_release: 286out_release:
285 cpu_hotplug_done(); 287 cpu_hotplug_done();
288 if (!err) {
289 if (raw_notifier_call_chain(&cpu_chain, CPU_POST_DEAD | mod,
290 hcpu) == NOTIFY_BAD)
291 BUG();
292 }
286 return err; 293 return err;
287} 294}
288 295
@@ -291,11 +298,30 @@ int __ref cpu_down(unsigned int cpu)
291 int err = 0; 298 int err = 0;
292 299
293 cpu_maps_update_begin(); 300 cpu_maps_update_begin();
294 if (cpu_hotplug_disabled) 301
302 if (cpu_hotplug_disabled) {
295 err = -EBUSY; 303 err = -EBUSY;
296 else 304 goto out;
297 err = _cpu_down(cpu, 0); 305 }
306
307 cpu_clear(cpu, cpu_active_map);
308
309 /*
310 * Make sure the all cpus did the reschedule and are not
311 * using stale version of the cpu_active_map.
312 * This is not strictly necessary becuase stop_machine()
313 * that we run down the line already provides the required
314 * synchronization. But it's really a side effect and we do not
315 * want to depend on the innards of the stop_machine here.
316 */
317 synchronize_sched();
318
319 err = _cpu_down(cpu, 0);
320
321 if (cpu_online(cpu))
322 cpu_set(cpu, cpu_active_map);
298 323
324out:
299 cpu_maps_update_done(); 325 cpu_maps_update_done();
300 return err; 326 return err;
301} 327}
@@ -355,11 +381,18 @@ int __cpuinit cpu_up(unsigned int cpu)
355 } 381 }
356 382
357 cpu_maps_update_begin(); 383 cpu_maps_update_begin();
358 if (cpu_hotplug_disabled) 384
385 if (cpu_hotplug_disabled) {
359 err = -EBUSY; 386 err = -EBUSY;
360 else 387 goto out;
361 err = _cpu_up(cpu, 0); 388 }
389
390 err = _cpu_up(cpu, 0);
362 391
392 if (cpu_online(cpu))
393 cpu_set(cpu, cpu_active_map);
394
395out:
363 cpu_maps_update_done(); 396 cpu_maps_update_done();
364 return err; 397 return err;
365} 398}
@@ -413,7 +446,7 @@ void __ref enable_nonboot_cpus(void)
413 goto out; 446 goto out;
414 447
415 printk("Enabling non-boot CPUs ...\n"); 448 printk("Enabling non-boot CPUs ...\n");
416 for_each_cpu_mask(cpu, frozen_cpus) { 449 for_each_cpu_mask_nr(cpu, frozen_cpus) {
417 error = _cpu_up(cpu, 1); 450 error = _cpu_up(cpu, 1);
418 if (!error) { 451 if (!error) {
419 printk("CPU%d is up\n", cpu); 452 printk("CPU%d is up\n", cpu);
diff --git a/kernel/cpuset.c b/kernel/cpuset.c
index 459d601947a8..91cf85b36dd5 100644
--- a/kernel/cpuset.c
+++ b/kernel/cpuset.c
@@ -227,10 +227,6 @@ static struct cpuset top_cpuset = {
227 * The task_struct fields mems_allowed and mems_generation may only 227 * The task_struct fields mems_allowed and mems_generation may only
228 * be accessed in the context of that task, so require no locks. 228 * be accessed in the context of that task, so require no locks.
229 * 229 *
230 * The cpuset_common_file_write handler for operations that modify
231 * the cpuset hierarchy holds cgroup_mutex across the entire operation,
232 * single threading all such cpuset modifications across the system.
233 *
234 * The cpuset_common_file_read() handlers only hold callback_mutex across 230 * The cpuset_common_file_read() handlers only hold callback_mutex across
235 * small pieces of code, such as when reading out possibly multi-word 231 * small pieces of code, such as when reading out possibly multi-word
236 * cpumasks and nodemasks. 232 * cpumasks and nodemasks.
@@ -369,7 +365,7 @@ void cpuset_update_task_memory_state(void)
369 my_cpusets_mem_gen = top_cpuset.mems_generation; 365 my_cpusets_mem_gen = top_cpuset.mems_generation;
370 } else { 366 } else {
371 rcu_read_lock(); 367 rcu_read_lock();
372 my_cpusets_mem_gen = task_cs(current)->mems_generation; 368 my_cpusets_mem_gen = task_cs(tsk)->mems_generation;
373 rcu_read_unlock(); 369 rcu_read_unlock();
374 } 370 }
375 371
@@ -500,11 +496,16 @@ update_domain_attr(struct sched_domain_attr *dattr, struct cpuset *c)
500/* 496/*
501 * rebuild_sched_domains() 497 * rebuild_sched_domains()
502 * 498 *
503 * If the flag 'sched_load_balance' of any cpuset with non-empty 499 * This routine will be called to rebuild the scheduler's dynamic
504 * 'cpus' changes, or if the 'cpus' allowed changes in any cpuset 500 * sched domains:
505 * which has that flag enabled, or if any cpuset with a non-empty 501 * - if the flag 'sched_load_balance' of any cpuset with non-empty
506 * 'cpus' is removed, then call this routine to rebuild the 502 * 'cpus' changes,
507 * scheduler's dynamic sched domains. 503 * - or if the 'cpus' allowed changes in any cpuset which has that
504 * flag enabled,
505 * - or if the 'sched_relax_domain_level' of any cpuset which has
506 * that flag enabled and with non-empty 'cpus' changes,
507 * - or if any cpuset with non-empty 'cpus' is removed,
508 * - or if a cpu gets offlined.
508 * 509 *
509 * This routine builds a partial partition of the systems CPUs 510 * This routine builds a partial partition of the systems CPUs
510 * (the set of non-overlappping cpumask_t's in the array 'part' 511 * (the set of non-overlappping cpumask_t's in the array 'part'
@@ -564,7 +565,7 @@ update_domain_attr(struct sched_domain_attr *dattr, struct cpuset *c)
564 * partition_sched_domains(). 565 * partition_sched_domains().
565 */ 566 */
566 567
567static void rebuild_sched_domains(void) 568void rebuild_sched_domains(void)
568{ 569{
569 struct kfifo *q; /* queue of cpusets to be scanned */ 570 struct kfifo *q; /* queue of cpusets to be scanned */
570 struct cpuset *cp; /* scans q */ 571 struct cpuset *cp; /* scans q */
@@ -609,8 +610,13 @@ static void rebuild_sched_domains(void)
609 while (__kfifo_get(q, (void *)&cp, sizeof(cp))) { 610 while (__kfifo_get(q, (void *)&cp, sizeof(cp))) {
610 struct cgroup *cont; 611 struct cgroup *cont;
611 struct cpuset *child; /* scans child cpusets of cp */ 612 struct cpuset *child; /* scans child cpusets of cp */
613
614 if (cpus_empty(cp->cpus_allowed))
615 continue;
616
612 if (is_sched_load_balance(cp)) 617 if (is_sched_load_balance(cp))
613 csa[csn++] = cp; 618 csa[csn++] = cp;
619
614 list_for_each_entry(cont, &cp->css.cgroup->children, sibling) { 620 list_for_each_entry(cont, &cp->css.cgroup->children, sibling) {
615 child = cgroup_cs(cont); 621 child = cgroup_cs(cont);
616 __kfifo_put(q, (void *)&child, sizeof(cp)); 622 __kfifo_put(q, (void *)&child, sizeof(cp));
@@ -679,7 +685,9 @@ restart:
679 if (apn == b->pn) { 685 if (apn == b->pn) {
680 cpus_or(*dp, *dp, b->cpus_allowed); 686 cpus_or(*dp, *dp, b->cpus_allowed);
681 b->pn = -1; 687 b->pn = -1;
682 update_domain_attr(dattr, b); 688 if (dattr)
689 update_domain_attr(dattr
690 + nslot, b);
683 } 691 }
684 } 692 }
685 nslot++; 693 nslot++;
@@ -701,36 +709,6 @@ done:
701 /* Don't kfree(dattr) -- partition_sched_domains() does that. */ 709 /* Don't kfree(dattr) -- partition_sched_domains() does that. */
702} 710}
703 711
704static inline int started_after_time(struct task_struct *t1,
705 struct timespec *time,
706 struct task_struct *t2)
707{
708 int start_diff = timespec_compare(&t1->start_time, time);
709 if (start_diff > 0) {
710 return 1;
711 } else if (start_diff < 0) {
712 return 0;
713 } else {
714 /*
715 * Arbitrarily, if two processes started at the same
716 * time, we'll say that the lower pointer value
717 * started first. Note that t2 may have exited by now
718 * so this may not be a valid pointer any longer, but
719 * that's fine - it still serves to distinguish
720 * between two tasks started (effectively)
721 * simultaneously.
722 */
723 return t1 > t2;
724 }
725}
726
727static inline int started_after(void *p1, void *p2)
728{
729 struct task_struct *t1 = p1;
730 struct task_struct *t2 = p2;
731 return started_after_time(t1, &t2->start_time, t2);
732}
733
734/** 712/**
735 * cpuset_test_cpumask - test a task's cpus_allowed versus its cpuset's 713 * cpuset_test_cpumask - test a task's cpus_allowed versus its cpuset's
736 * @tsk: task to test 714 * @tsk: task to test
@@ -766,15 +744,49 @@ static void cpuset_change_cpumask(struct task_struct *tsk,
766} 744}
767 745
768/** 746/**
747 * update_tasks_cpumask - Update the cpumasks of tasks in the cpuset.
748 * @cs: the cpuset in which each task's cpus_allowed mask needs to be changed
749 *
750 * Called with cgroup_mutex held
751 *
752 * The cgroup_scan_tasks() function will scan all the tasks in a cgroup,
753 * calling callback functions for each.
754 *
755 * Return 0 if successful, -errno if not.
756 */
757static int update_tasks_cpumask(struct cpuset *cs)
758{
759 struct cgroup_scanner scan;
760 struct ptr_heap heap;
761 int retval;
762
763 /*
764 * cgroup_scan_tasks() will initialize heap->gt for us.
765 * heap_init() is still needed here for we should not change
766 * cs->cpus_allowed when heap_init() fails.
767 */
768 retval = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, NULL);
769 if (retval)
770 return retval;
771
772 scan.cg = cs->css.cgroup;
773 scan.test_task = cpuset_test_cpumask;
774 scan.process_task = cpuset_change_cpumask;
775 scan.heap = &heap;
776 retval = cgroup_scan_tasks(&scan);
777
778 heap_free(&heap);
779 return retval;
780}
781
782/**
769 * update_cpumask - update the cpus_allowed mask of a cpuset and all tasks in it 783 * update_cpumask - update the cpus_allowed mask of a cpuset and all tasks in it
770 * @cs: the cpuset to consider 784 * @cs: the cpuset to consider
771 * @buf: buffer of cpu numbers written to this cpuset 785 * @buf: buffer of cpu numbers written to this cpuset
772 */ 786 */
773static int update_cpumask(struct cpuset *cs, char *buf) 787static int update_cpumask(struct cpuset *cs, const char *buf)
774{ 788{
775 struct cpuset trialcs; 789 struct cpuset trialcs;
776 struct cgroup_scanner scan;
777 struct ptr_heap heap;
778 int retval; 790 int retval;
779 int is_load_balanced; 791 int is_load_balanced;
780 792
@@ -790,7 +802,6 @@ static int update_cpumask(struct cpuset *cs, char *buf)
790 * that parsing. The validate_change() call ensures that cpusets 802 * that parsing. The validate_change() call ensures that cpusets
791 * with tasks have cpus. 803 * with tasks have cpus.
792 */ 804 */
793 buf = strstrip(buf);
794 if (!*buf) { 805 if (!*buf) {
795 cpus_clear(trialcs.cpus_allowed); 806 cpus_clear(trialcs.cpus_allowed);
796 } else { 807 } else {
@@ -809,10 +820,6 @@ static int update_cpumask(struct cpuset *cs, char *buf)
809 if (cpus_equal(cs->cpus_allowed, trialcs.cpus_allowed)) 820 if (cpus_equal(cs->cpus_allowed, trialcs.cpus_allowed))
810 return 0; 821 return 0;
811 822
812 retval = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, &started_after);
813 if (retval)
814 return retval;
815
816 is_load_balanced = is_sched_load_balance(&trialcs); 823 is_load_balanced = is_sched_load_balance(&trialcs);
817 824
818 mutex_lock(&callback_mutex); 825 mutex_lock(&callback_mutex);
@@ -823,12 +830,9 @@ static int update_cpumask(struct cpuset *cs, char *buf)
823 * Scan tasks in the cpuset, and update the cpumasks of any 830 * Scan tasks in the cpuset, and update the cpumasks of any
824 * that need an update. 831 * that need an update.
825 */ 832 */
826 scan.cg = cs->css.cgroup; 833 retval = update_tasks_cpumask(cs);
827 scan.test_task = cpuset_test_cpumask; 834 if (retval < 0)
828 scan.process_task = cpuset_change_cpumask; 835 return retval;
829 scan.heap = &heap;
830 cgroup_scan_tasks(&scan);
831 heap_free(&heap);
832 836
833 if (is_load_balanced) 837 if (is_load_balanced)
834 rebuild_sched_domains(); 838 rebuild_sched_domains();
@@ -884,74 +888,25 @@ static void cpuset_migrate_mm(struct mm_struct *mm, const nodemask_t *from,
884 mutex_unlock(&callback_mutex); 888 mutex_unlock(&callback_mutex);
885} 889}
886 890
887/*
888 * Handle user request to change the 'mems' memory placement
889 * of a cpuset. Needs to validate the request, update the
890 * cpusets mems_allowed and mems_generation, and for each
891 * task in the cpuset, rebind any vma mempolicies and if
892 * the cpuset is marked 'memory_migrate', migrate the tasks
893 * pages to the new memory.
894 *
895 * Call with cgroup_mutex held. May take callback_mutex during call.
896 * Will take tasklist_lock, scan tasklist for tasks in cpuset cs,
897 * lock each such tasks mm->mmap_sem, scan its vma's and rebind
898 * their mempolicies to the cpusets new mems_allowed.
899 */
900
901static void *cpuset_being_rebound; 891static void *cpuset_being_rebound;
902 892
903static int update_nodemask(struct cpuset *cs, char *buf) 893/**
894 * update_tasks_nodemask - Update the nodemasks of tasks in the cpuset.
895 * @cs: the cpuset in which each task's mems_allowed mask needs to be changed
896 * @oldmem: old mems_allowed of cpuset cs
897 *
898 * Called with cgroup_mutex held
899 * Return 0 if successful, -errno if not.
900 */
901static int update_tasks_nodemask(struct cpuset *cs, const nodemask_t *oldmem)
904{ 902{
905 struct cpuset trialcs;
906 nodemask_t oldmem;
907 struct task_struct *p; 903 struct task_struct *p;
908 struct mm_struct **mmarray; 904 struct mm_struct **mmarray;
909 int i, n, ntasks; 905 int i, n, ntasks;
910 int migrate; 906 int migrate;
911 int fudge; 907 int fudge;
912 int retval;
913 struct cgroup_iter it; 908 struct cgroup_iter it;
914 909 int retval;
915 /*
916 * top_cpuset.mems_allowed tracks node_stats[N_HIGH_MEMORY];
917 * it's read-only
918 */
919 if (cs == &top_cpuset)
920 return -EACCES;
921
922 trialcs = *cs;
923
924 /*
925 * An empty mems_allowed is ok iff there are no tasks in the cpuset.
926 * Since nodelist_parse() fails on an empty mask, we special case
927 * that parsing. The validate_change() call ensures that cpusets
928 * with tasks have memory.
929 */
930 buf = strstrip(buf);
931 if (!*buf) {
932 nodes_clear(trialcs.mems_allowed);
933 } else {
934 retval = nodelist_parse(buf, trialcs.mems_allowed);
935 if (retval < 0)
936 goto done;
937
938 if (!nodes_subset(trialcs.mems_allowed,
939 node_states[N_HIGH_MEMORY]))
940 return -EINVAL;
941 }
942 oldmem = cs->mems_allowed;
943 if (nodes_equal(oldmem, trialcs.mems_allowed)) {
944 retval = 0; /* Too easy - nothing to do */
945 goto done;
946 }
947 retval = validate_change(cs, &trialcs);
948 if (retval < 0)
949 goto done;
950
951 mutex_lock(&callback_mutex);
952 cs->mems_allowed = trialcs.mems_allowed;
953 cs->mems_generation = cpuset_mems_generation++;
954 mutex_unlock(&callback_mutex);
955 910
956 cpuset_being_rebound = cs; /* causes mpol_dup() rebind */ 911 cpuset_being_rebound = cs; /* causes mpol_dup() rebind */
957 912
@@ -1018,7 +973,7 @@ static int update_nodemask(struct cpuset *cs, char *buf)
1018 973
1019 mpol_rebind_mm(mm, &cs->mems_allowed); 974 mpol_rebind_mm(mm, &cs->mems_allowed);
1020 if (migrate) 975 if (migrate)
1021 cpuset_migrate_mm(mm, &oldmem, &cs->mems_allowed); 976 cpuset_migrate_mm(mm, oldmem, &cs->mems_allowed);
1022 mmput(mm); 977 mmput(mm);
1023 } 978 }
1024 979
@@ -1030,6 +985,70 @@ done:
1030 return retval; 985 return retval;
1031} 986}
1032 987
988/*
989 * Handle user request to change the 'mems' memory placement
990 * of a cpuset. Needs to validate the request, update the
991 * cpusets mems_allowed and mems_generation, and for each
992 * task in the cpuset, rebind any vma mempolicies and if
993 * the cpuset is marked 'memory_migrate', migrate the tasks
994 * pages to the new memory.
995 *
996 * Call with cgroup_mutex held. May take callback_mutex during call.
997 * Will take tasklist_lock, scan tasklist for tasks in cpuset cs,
998 * lock each such tasks mm->mmap_sem, scan its vma's and rebind
999 * their mempolicies to the cpusets new mems_allowed.
1000 */
1001static int update_nodemask(struct cpuset *cs, const char *buf)
1002{
1003 struct cpuset trialcs;
1004 nodemask_t oldmem;
1005 int retval;
1006
1007 /*
1008 * top_cpuset.mems_allowed tracks node_stats[N_HIGH_MEMORY];
1009 * it's read-only
1010 */
1011 if (cs == &top_cpuset)
1012 return -EACCES;
1013
1014 trialcs = *cs;
1015
1016 /*
1017 * An empty mems_allowed is ok iff there are no tasks in the cpuset.
1018 * Since nodelist_parse() fails on an empty mask, we special case
1019 * that parsing. The validate_change() call ensures that cpusets
1020 * with tasks have memory.
1021 */
1022 if (!*buf) {
1023 nodes_clear(trialcs.mems_allowed);
1024 } else {
1025 retval = nodelist_parse(buf, trialcs.mems_allowed);
1026 if (retval < 0)
1027 goto done;
1028
1029 if (!nodes_subset(trialcs.mems_allowed,
1030 node_states[N_HIGH_MEMORY]))
1031 return -EINVAL;
1032 }
1033 oldmem = cs->mems_allowed;
1034 if (nodes_equal(oldmem, trialcs.mems_allowed)) {
1035 retval = 0; /* Too easy - nothing to do */
1036 goto done;
1037 }
1038 retval = validate_change(cs, &trialcs);
1039 if (retval < 0)
1040 goto done;
1041
1042 mutex_lock(&callback_mutex);
1043 cs->mems_allowed = trialcs.mems_allowed;
1044 cs->mems_generation = cpuset_mems_generation++;
1045 mutex_unlock(&callback_mutex);
1046
1047 retval = update_tasks_nodemask(cs, &oldmem);
1048done:
1049 return retval;
1050}
1051
1033int current_cpuset_is_being_rebound(void) 1052int current_cpuset_is_being_rebound(void)
1034{ 1053{
1035 return task_cs(current) == cpuset_being_rebound; 1054 return task_cs(current) == cpuset_being_rebound;
@@ -1042,7 +1061,8 @@ static int update_relax_domain_level(struct cpuset *cs, s64 val)
1042 1061
1043 if (val != cs->relax_domain_level) { 1062 if (val != cs->relax_domain_level) {
1044 cs->relax_domain_level = val; 1063 cs->relax_domain_level = val;
1045 rebuild_sched_domains(); 1064 if (!cpus_empty(cs->cpus_allowed) && is_sched_load_balance(cs))
1065 rebuild_sched_domains();
1046 } 1066 }
1047 1067
1048 return 0; 1068 return 0;
@@ -1254,72 +1274,14 @@ typedef enum {
1254 FILE_SPREAD_SLAB, 1274 FILE_SPREAD_SLAB,
1255} cpuset_filetype_t; 1275} cpuset_filetype_t;
1256 1276
1257static ssize_t cpuset_common_file_write(struct cgroup *cont,
1258 struct cftype *cft,
1259 struct file *file,
1260 const char __user *userbuf,
1261 size_t nbytes, loff_t *unused_ppos)
1262{
1263 struct cpuset *cs = cgroup_cs(cont);
1264 cpuset_filetype_t type = cft->private;
1265 char *buffer;
1266 int retval = 0;
1267
1268 /* Crude upper limit on largest legitimate cpulist user might write. */
1269 if (nbytes > 100U + 6 * max(NR_CPUS, MAX_NUMNODES))
1270 return -E2BIG;
1271
1272 /* +1 for nul-terminator */
1273 buffer = kmalloc(nbytes + 1, GFP_KERNEL);
1274 if (!buffer)
1275 return -ENOMEM;
1276
1277 if (copy_from_user(buffer, userbuf, nbytes)) {
1278 retval = -EFAULT;
1279 goto out1;
1280 }
1281 buffer[nbytes] = 0; /* nul-terminate */
1282
1283 cgroup_lock();
1284
1285 if (cgroup_is_removed(cont)) {
1286 retval = -ENODEV;
1287 goto out2;
1288 }
1289
1290 switch (type) {
1291 case FILE_CPULIST:
1292 retval = update_cpumask(cs, buffer);
1293 break;
1294 case FILE_MEMLIST:
1295 retval = update_nodemask(cs, buffer);
1296 break;
1297 default:
1298 retval = -EINVAL;
1299 goto out2;
1300 }
1301
1302 if (retval == 0)
1303 retval = nbytes;
1304out2:
1305 cgroup_unlock();
1306out1:
1307 kfree(buffer);
1308 return retval;
1309}
1310
1311static int cpuset_write_u64(struct cgroup *cgrp, struct cftype *cft, u64 val) 1277static int cpuset_write_u64(struct cgroup *cgrp, struct cftype *cft, u64 val)
1312{ 1278{
1313 int retval = 0; 1279 int retval = 0;
1314 struct cpuset *cs = cgroup_cs(cgrp); 1280 struct cpuset *cs = cgroup_cs(cgrp);
1315 cpuset_filetype_t type = cft->private; 1281 cpuset_filetype_t type = cft->private;
1316 1282
1317 cgroup_lock(); 1283 if (!cgroup_lock_live_group(cgrp))
1318
1319 if (cgroup_is_removed(cgrp)) {
1320 cgroup_unlock();
1321 return -ENODEV; 1284 return -ENODEV;
1322 }
1323 1285
1324 switch (type) { 1286 switch (type) {
1325 case FILE_CPU_EXCLUSIVE: 1287 case FILE_CPU_EXCLUSIVE:
@@ -1365,12 +1327,9 @@ static int cpuset_write_s64(struct cgroup *cgrp, struct cftype *cft, s64 val)
1365 struct cpuset *cs = cgroup_cs(cgrp); 1327 struct cpuset *cs = cgroup_cs(cgrp);
1366 cpuset_filetype_t type = cft->private; 1328 cpuset_filetype_t type = cft->private;
1367 1329
1368 cgroup_lock(); 1330 if (!cgroup_lock_live_group(cgrp))
1369
1370 if (cgroup_is_removed(cgrp)) {
1371 cgroup_unlock();
1372 return -ENODEV; 1331 return -ENODEV;
1373 } 1332
1374 switch (type) { 1333 switch (type) {
1375 case FILE_SCHED_RELAX_DOMAIN_LEVEL: 1334 case FILE_SCHED_RELAX_DOMAIN_LEVEL:
1376 retval = update_relax_domain_level(cs, val); 1335 retval = update_relax_domain_level(cs, val);
@@ -1384,6 +1343,32 @@ static int cpuset_write_s64(struct cgroup *cgrp, struct cftype *cft, s64 val)
1384} 1343}
1385 1344
1386/* 1345/*
1346 * Common handling for a write to a "cpus" or "mems" file.
1347 */
1348static int cpuset_write_resmask(struct cgroup *cgrp, struct cftype *cft,
1349 const char *buf)
1350{
1351 int retval = 0;
1352
1353 if (!cgroup_lock_live_group(cgrp))
1354 return -ENODEV;
1355
1356 switch (cft->private) {
1357 case FILE_CPULIST:
1358 retval = update_cpumask(cgroup_cs(cgrp), buf);
1359 break;
1360 case FILE_MEMLIST:
1361 retval = update_nodemask(cgroup_cs(cgrp), buf);
1362 break;
1363 default:
1364 retval = -EINVAL;
1365 break;
1366 }
1367 cgroup_unlock();
1368 return retval;
1369}
1370
1371/*
1387 * These ascii lists should be read in a single call, by using a user 1372 * These ascii lists should be read in a single call, by using a user
1388 * buffer large enough to hold the entire map. If read in smaller 1373 * buffer large enough to hold the entire map. If read in smaller
1389 * chunks, there is no guarantee of atomicity. Since the display format 1374 * chunks, there is no guarantee of atomicity. Since the display format
@@ -1502,14 +1487,16 @@ static struct cftype files[] = {
1502 { 1487 {
1503 .name = "cpus", 1488 .name = "cpus",
1504 .read = cpuset_common_file_read, 1489 .read = cpuset_common_file_read,
1505 .write = cpuset_common_file_write, 1490 .write_string = cpuset_write_resmask,
1491 .max_write_len = (100U + 6 * NR_CPUS),
1506 .private = FILE_CPULIST, 1492 .private = FILE_CPULIST,
1507 }, 1493 },
1508 1494
1509 { 1495 {
1510 .name = "mems", 1496 .name = "mems",
1511 .read = cpuset_common_file_read, 1497 .read = cpuset_common_file_read,
1512 .write = cpuset_common_file_write, 1498 .write_string = cpuset_write_resmask,
1499 .max_write_len = (100U + 6 * MAX_NUMNODES),
1513 .private = FILE_MEMLIST, 1500 .private = FILE_MEMLIST,
1514 }, 1501 },
1515 1502
@@ -1790,7 +1777,7 @@ static void move_member_tasks_to_cpuset(struct cpuset *from, struct cpuset *to)
1790 scan.scan.heap = NULL; 1777 scan.scan.heap = NULL;
1791 scan.to = to->css.cgroup; 1778 scan.to = to->css.cgroup;
1792 1779
1793 if (cgroup_scan_tasks((struct cgroup_scanner *)&scan)) 1780 if (cgroup_scan_tasks(&scan.scan))
1794 printk(KERN_ERR "move_member_tasks_to_cpuset: " 1781 printk(KERN_ERR "move_member_tasks_to_cpuset: "
1795 "cgroup_scan_tasks failed\n"); 1782 "cgroup_scan_tasks failed\n");
1796} 1783}
@@ -1850,6 +1837,7 @@ static void scan_for_empty_cpusets(const struct cpuset *root)
1850 struct cpuset *child; /* scans child cpusets of cp */ 1837 struct cpuset *child; /* scans child cpusets of cp */
1851 struct list_head queue; 1838 struct list_head queue;
1852 struct cgroup *cont; 1839 struct cgroup *cont;
1840 nodemask_t oldmems;
1853 1841
1854 INIT_LIST_HEAD(&queue); 1842 INIT_LIST_HEAD(&queue);
1855 1843
@@ -1869,6 +1857,8 @@ static void scan_for_empty_cpusets(const struct cpuset *root)
1869 nodes_subset(cp->mems_allowed, node_states[N_HIGH_MEMORY])) 1857 nodes_subset(cp->mems_allowed, node_states[N_HIGH_MEMORY]))
1870 continue; 1858 continue;
1871 1859
1860 oldmems = cp->mems_allowed;
1861
1872 /* Remove offline cpus and mems from this cpuset. */ 1862 /* Remove offline cpus and mems from this cpuset. */
1873 mutex_lock(&callback_mutex); 1863 mutex_lock(&callback_mutex);
1874 cpus_and(cp->cpus_allowed, cp->cpus_allowed, cpu_online_map); 1864 cpus_and(cp->cpus_allowed, cp->cpus_allowed, cpu_online_map);
@@ -1880,6 +1870,10 @@ static void scan_for_empty_cpusets(const struct cpuset *root)
1880 if (cpus_empty(cp->cpus_allowed) || 1870 if (cpus_empty(cp->cpus_allowed) ||
1881 nodes_empty(cp->mems_allowed)) 1871 nodes_empty(cp->mems_allowed))
1882 remove_tasks_in_empty_cpuset(cp); 1872 remove_tasks_in_empty_cpuset(cp);
1873 else {
1874 update_tasks_cpumask(cp);
1875 update_tasks_nodemask(cp, &oldmems);
1876 }
1883 } 1877 }
1884} 1878}
1885 1879
@@ -1972,7 +1966,6 @@ void __init cpuset_init_smp(void)
1972} 1966}
1973 1967
1974/** 1968/**
1975
1976 * cpuset_cpus_allowed - return cpus_allowed mask from a tasks cpuset. 1969 * cpuset_cpus_allowed - return cpus_allowed mask from a tasks cpuset.
1977 * @tsk: pointer to task_struct from which to obtain cpuset->cpus_allowed. 1970 * @tsk: pointer to task_struct from which to obtain cpuset->cpus_allowed.
1978 * @pmask: pointer to cpumask_t variable to receive cpus_allowed set. 1971 * @pmask: pointer to cpumask_t variable to receive cpus_allowed set.
diff --git a/kernel/delayacct.c b/kernel/delayacct.c
index 10e43fd8b721..b3179dad71be 100644
--- a/kernel/delayacct.c
+++ b/kernel/delayacct.c
@@ -145,8 +145,11 @@ int __delayacct_add_tsk(struct taskstats *d, struct task_struct *tsk)
145 d->blkio_delay_total = (tmp < d->blkio_delay_total) ? 0 : tmp; 145 d->blkio_delay_total = (tmp < d->blkio_delay_total) ? 0 : tmp;
146 tmp = d->swapin_delay_total + tsk->delays->swapin_delay; 146 tmp = d->swapin_delay_total + tsk->delays->swapin_delay;
147 d->swapin_delay_total = (tmp < d->swapin_delay_total) ? 0 : tmp; 147 d->swapin_delay_total = (tmp < d->swapin_delay_total) ? 0 : tmp;
148 tmp = d->freepages_delay_total + tsk->delays->freepages_delay;
149 d->freepages_delay_total = (tmp < d->freepages_delay_total) ? 0 : tmp;
148 d->blkio_count += tsk->delays->blkio_count; 150 d->blkio_count += tsk->delays->blkio_count;
149 d->swapin_count += tsk->delays->swapin_count; 151 d->swapin_count += tsk->delays->swapin_count;
152 d->freepages_count += tsk->delays->freepages_count;
150 spin_unlock_irqrestore(&tsk->delays->lock, flags); 153 spin_unlock_irqrestore(&tsk->delays->lock, flags);
151 154
152done: 155done:
@@ -165,3 +168,16 @@ __u64 __delayacct_blkio_ticks(struct task_struct *tsk)
165 return ret; 168 return ret;
166} 169}
167 170
171void __delayacct_freepages_start(void)
172{
173 delayacct_start(&current->delays->freepages_start);
174}
175
176void __delayacct_freepages_end(void)
177{
178 delayacct_end(&current->delays->freepages_start,
179 &current->delays->freepages_end,
180 &current->delays->freepages_delay,
181 &current->delays->freepages_count);
182}
183
diff --git a/kernel/exec_domain.c b/kernel/exec_domain.c
index a9e6bad9f706..c1ef192aa655 100644
--- a/kernel/exec_domain.c
+++ b/kernel/exec_domain.c
@@ -65,7 +65,7 @@ lookup_exec_domain(u_long personality)
65 goto out; 65 goto out;
66 } 66 }
67 67
68#ifdef CONFIG_KMOD 68#ifdef CONFIG_MODULES
69 read_unlock(&exec_domains_lock); 69 read_unlock(&exec_domains_lock);
70 request_module("personality-%ld", pers); 70 request_module("personality-%ld", pers);
71 read_lock(&exec_domains_lock); 71 read_lock(&exec_domains_lock);
diff --git a/kernel/exit.c b/kernel/exit.c
index 93d2711b9381..ad933bb29ec7 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -85,7 +85,6 @@ static void __exit_signal(struct task_struct *tsk)
85 BUG_ON(!sig); 85 BUG_ON(!sig);
86 BUG_ON(!atomic_read(&sig->count)); 86 BUG_ON(!atomic_read(&sig->count));
87 87
88 rcu_read_lock();
89 sighand = rcu_dereference(tsk->sighand); 88 sighand = rcu_dereference(tsk->sighand);
90 spin_lock(&sighand->siglock); 89 spin_lock(&sighand->siglock);
91 90
@@ -121,6 +120,18 @@ static void __exit_signal(struct task_struct *tsk)
121 sig->nivcsw += tsk->nivcsw; 120 sig->nivcsw += tsk->nivcsw;
122 sig->inblock += task_io_get_inblock(tsk); 121 sig->inblock += task_io_get_inblock(tsk);
123 sig->oublock += task_io_get_oublock(tsk); 122 sig->oublock += task_io_get_oublock(tsk);
123#ifdef CONFIG_TASK_XACCT
124 sig->rchar += tsk->rchar;
125 sig->wchar += tsk->wchar;
126 sig->syscr += tsk->syscr;
127 sig->syscw += tsk->syscw;
128#endif /* CONFIG_TASK_XACCT */
129#ifdef CONFIG_TASK_IO_ACCOUNTING
130 sig->ioac.read_bytes += tsk->ioac.read_bytes;
131 sig->ioac.write_bytes += tsk->ioac.write_bytes;
132 sig->ioac.cancelled_write_bytes +=
133 tsk->ioac.cancelled_write_bytes;
134#endif /* CONFIG_TASK_IO_ACCOUNTING */
124 sig->sum_sched_runtime += tsk->se.sum_exec_runtime; 135 sig->sum_sched_runtime += tsk->se.sum_exec_runtime;
125 sig = NULL; /* Marker for below. */ 136 sig = NULL; /* Marker for below. */
126 } 137 }
@@ -136,7 +147,6 @@ static void __exit_signal(struct task_struct *tsk)
136 tsk->signal = NULL; 147 tsk->signal = NULL;
137 tsk->sighand = NULL; 148 tsk->sighand = NULL;
138 spin_unlock(&sighand->siglock); 149 spin_unlock(&sighand->siglock);
139 rcu_read_unlock();
140 150
141 __cleanup_sighand(sighand); 151 __cleanup_sighand(sighand);
142 clear_tsk_thread_flag(tsk,TIF_SIGPENDING); 152 clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
@@ -432,7 +442,7 @@ void daemonize(const char *name, ...)
432 * We don't want to have TIF_FREEZE set if the system-wide hibernation 442 * We don't want to have TIF_FREEZE set if the system-wide hibernation
433 * or suspend transition begins right now. 443 * or suspend transition begins right now.
434 */ 444 */
435 current->flags |= PF_NOFREEZE; 445 current->flags |= (PF_NOFREEZE | PF_KTHREAD);
436 446
437 if (current->nsproxy != &init_nsproxy) { 447 if (current->nsproxy != &init_nsproxy) {
438 get_nsproxy(&init_nsproxy); 448 get_nsproxy(&init_nsproxy);
@@ -666,26 +676,40 @@ assign_new_owner:
666static void exit_mm(struct task_struct * tsk) 676static void exit_mm(struct task_struct * tsk)
667{ 677{
668 struct mm_struct *mm = tsk->mm; 678 struct mm_struct *mm = tsk->mm;
679 struct core_state *core_state;
669 680
670 mm_release(tsk, mm); 681 mm_release(tsk, mm);
671 if (!mm) 682 if (!mm)
672 return; 683 return;
673 /* 684 /*
674 * Serialize with any possible pending coredump. 685 * Serialize with any possible pending coredump.
675 * We must hold mmap_sem around checking core_waiters 686 * We must hold mmap_sem around checking core_state
676 * and clearing tsk->mm. The core-inducing thread 687 * and clearing tsk->mm. The core-inducing thread
677 * will increment core_waiters for each thread in the 688 * will increment ->nr_threads for each thread in the
678 * group with ->mm != NULL. 689 * group with ->mm != NULL.
679 */ 690 */
680 down_read(&mm->mmap_sem); 691 down_read(&mm->mmap_sem);
681 if (mm->core_waiters) { 692 core_state = mm->core_state;
693 if (core_state) {
694 struct core_thread self;
682 up_read(&mm->mmap_sem); 695 up_read(&mm->mmap_sem);
683 down_write(&mm->mmap_sem);
684 if (!--mm->core_waiters)
685 complete(mm->core_startup_done);
686 up_write(&mm->mmap_sem);
687 696
688 wait_for_completion(&mm->core_done); 697 self.task = tsk;
698 self.next = xchg(&core_state->dumper.next, &self);
699 /*
700 * Implies mb(), the result of xchg() must be visible
701 * to core_state->dumper.
702 */
703 if (atomic_dec_and_test(&core_state->nr_threads))
704 complete(&core_state->startup);
705
706 for (;;) {
707 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
708 if (!self.task) /* see coredump_finish() */
709 break;
710 schedule();
711 }
712 __set_task_state(tsk, TASK_RUNNING);
689 down_read(&mm->mmap_sem); 713 down_read(&mm->mmap_sem);
690 } 714 }
691 atomic_inc(&mm->mm_count); 715 atomic_inc(&mm->mm_count);
@@ -1354,6 +1378,21 @@ static int wait_task_zombie(struct task_struct *p, int options,
1354 psig->coublock += 1378 psig->coublock +=
1355 task_io_get_oublock(p) + 1379 task_io_get_oublock(p) +
1356 sig->oublock + sig->coublock; 1380 sig->oublock + sig->coublock;
1381#ifdef CONFIG_TASK_XACCT
1382 psig->rchar += p->rchar + sig->rchar;
1383 psig->wchar += p->wchar + sig->wchar;
1384 psig->syscr += p->syscr + sig->syscr;
1385 psig->syscw += p->syscw + sig->syscw;
1386#endif /* CONFIG_TASK_XACCT */
1387#ifdef CONFIG_TASK_IO_ACCOUNTING
1388 psig->ioac.read_bytes +=
1389 p->ioac.read_bytes + sig->ioac.read_bytes;
1390 psig->ioac.write_bytes +=
1391 p->ioac.write_bytes + sig->ioac.write_bytes;
1392 psig->ioac.cancelled_write_bytes +=
1393 p->ioac.cancelled_write_bytes +
1394 sig->ioac.cancelled_write_bytes;
1395#endif /* CONFIG_TASK_IO_ACCOUNTING */
1357 spin_unlock_irq(&p->parent->sighand->siglock); 1396 spin_unlock_irq(&p->parent->sighand->siglock);
1358 } 1397 }
1359 1398
diff --git a/kernel/fork.c b/kernel/fork.c
index adefc1131f27..b99d73e971a4 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -33,6 +33,7 @@
33#include <linux/cpu.h> 33#include <linux/cpu.h>
34#include <linux/cgroup.h> 34#include <linux/cgroup.h>
35#include <linux/security.h> 35#include <linux/security.h>
36#include <linux/hugetlb.h>
36#include <linux/swap.h> 37#include <linux/swap.h>
37#include <linux/syscalls.h> 38#include <linux/syscalls.h>
38#include <linux/jiffies.h> 39#include <linux/jiffies.h>
@@ -92,6 +93,23 @@ int nr_processes(void)
92static struct kmem_cache *task_struct_cachep; 93static struct kmem_cache *task_struct_cachep;
93#endif 94#endif
94 95
96#ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
97static inline struct thread_info *alloc_thread_info(struct task_struct *tsk)
98{
99#ifdef CONFIG_DEBUG_STACK_USAGE
100 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
101#else
102 gfp_t mask = GFP_KERNEL;
103#endif
104 return (struct thread_info *)__get_free_pages(mask, THREAD_SIZE_ORDER);
105}
106
107static inline void free_thread_info(struct thread_info *ti)
108{
109 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
110}
111#endif
112
95/* SLAB cache for signal_struct structures (tsk->signal) */ 113/* SLAB cache for signal_struct structures (tsk->signal) */
96static struct kmem_cache *signal_cachep; 114static struct kmem_cache *signal_cachep;
97 115
@@ -307,6 +325,14 @@ static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
307 } 325 }
308 326
309 /* 327 /*
328 * Clear hugetlb-related page reserves for children. This only
329 * affects MAP_PRIVATE mappings. Faults generated by the child
330 * are not guaranteed to succeed, even if read-only
331 */
332 if (is_vm_hugetlb_page(tmp))
333 reset_vma_resv_huge_pages(tmp);
334
335 /*
310 * Link in the new vma and copy the page table entries. 336 * Link in the new vma and copy the page table entries.
311 */ 337 */
312 *pprev = tmp; 338 *pprev = tmp;
@@ -374,7 +400,7 @@ static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
374 INIT_LIST_HEAD(&mm->mmlist); 400 INIT_LIST_HEAD(&mm->mmlist);
375 mm->flags = (current->mm) ? current->mm->flags 401 mm->flags = (current->mm) ? current->mm->flags
376 : MMF_DUMP_FILTER_DEFAULT; 402 : MMF_DUMP_FILTER_DEFAULT;
377 mm->core_waiters = 0; 403 mm->core_state = NULL;
378 mm->nr_ptes = 0; 404 mm->nr_ptes = 0;
379 set_mm_counter(mm, file_rss, 0); 405 set_mm_counter(mm, file_rss, 0);
380 set_mm_counter(mm, anon_rss, 0); 406 set_mm_counter(mm, anon_rss, 0);
@@ -448,7 +474,7 @@ EXPORT_SYMBOL_GPL(mmput);
448/** 474/**
449 * get_task_mm - acquire a reference to the task's mm 475 * get_task_mm - acquire a reference to the task's mm
450 * 476 *
451 * Returns %NULL if the task has no mm. Checks PF_BORROWED_MM (meaning 477 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
452 * this kernel workthread has transiently adopted a user mm with use_mm, 478 * this kernel workthread has transiently adopted a user mm with use_mm,
453 * to do its AIO) is not set and if so returns a reference to it, after 479 * to do its AIO) is not set and if so returns a reference to it, after
454 * bumping up the use count. User must release the mm via mmput() 480 * bumping up the use count. User must release the mm via mmput()
@@ -461,7 +487,7 @@ struct mm_struct *get_task_mm(struct task_struct *task)
461 task_lock(task); 487 task_lock(task);
462 mm = task->mm; 488 mm = task->mm;
463 if (mm) { 489 if (mm) {
464 if (task->flags & PF_BORROWED_MM) 490 if (task->flags & PF_KTHREAD)
465 mm = NULL; 491 mm = NULL;
466 else 492 else
467 atomic_inc(&mm->mm_users); 493 atomic_inc(&mm->mm_users);
@@ -786,6 +812,12 @@ static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
786 sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0; 812 sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
787 sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0; 813 sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
788 sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0; 814 sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0;
815#ifdef CONFIG_TASK_XACCT
816 sig->rchar = sig->wchar = sig->syscr = sig->syscw = 0;
817#endif
818#ifdef CONFIG_TASK_IO_ACCOUNTING
819 memset(&sig->ioac, 0, sizeof(sig->ioac));
820#endif
789 sig->sum_sched_runtime = 0; 821 sig->sum_sched_runtime = 0;
790 INIT_LIST_HEAD(&sig->cpu_timers[0]); 822 INIT_LIST_HEAD(&sig->cpu_timers[0]);
791 INIT_LIST_HEAD(&sig->cpu_timers[1]); 823 INIT_LIST_HEAD(&sig->cpu_timers[1]);
@@ -1081,6 +1113,12 @@ static struct task_struct *copy_process(unsigned long clone_flags,
1081 if (clone_flags & CLONE_THREAD) 1113 if (clone_flags & CLONE_THREAD)
1082 p->tgid = current->tgid; 1114 p->tgid = current->tgid;
1083 1115
1116 if (current->nsproxy != p->nsproxy) {
1117 retval = ns_cgroup_clone(p, pid);
1118 if (retval)
1119 goto bad_fork_free_pid;
1120 }
1121
1084 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; 1122 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1085 /* 1123 /*
1086 * Clear TID on mm_release()? 1124 * Clear TID on mm_release()?
diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c
index 77a51be36010..f8914b92b664 100644
--- a/kernel/irq/manage.c
+++ b/kernel/irq/manage.c
@@ -217,6 +217,17 @@ void enable_irq(unsigned int irq)
217} 217}
218EXPORT_SYMBOL(enable_irq); 218EXPORT_SYMBOL(enable_irq);
219 219
220int set_irq_wake_real(unsigned int irq, unsigned int on)
221{
222 struct irq_desc *desc = irq_desc + irq;
223 int ret = -ENXIO;
224
225 if (desc->chip->set_wake)
226 ret = desc->chip->set_wake(irq, on);
227
228 return ret;
229}
230
220/** 231/**
221 * set_irq_wake - control irq power management wakeup 232 * set_irq_wake - control irq power management wakeup
222 * @irq: interrupt to control 233 * @irq: interrupt to control
@@ -233,30 +244,32 @@ int set_irq_wake(unsigned int irq, unsigned int on)
233{ 244{
234 struct irq_desc *desc = irq_desc + irq; 245 struct irq_desc *desc = irq_desc + irq;
235 unsigned long flags; 246 unsigned long flags;
236 int ret = -ENXIO; 247 int ret = 0;
237 int (*set_wake)(unsigned, unsigned) = desc->chip->set_wake;
238 248
239 /* wakeup-capable irqs can be shared between drivers that 249 /* wakeup-capable irqs can be shared between drivers that
240 * don't need to have the same sleep mode behaviors. 250 * don't need to have the same sleep mode behaviors.
241 */ 251 */
242 spin_lock_irqsave(&desc->lock, flags); 252 spin_lock_irqsave(&desc->lock, flags);
243 if (on) { 253 if (on) {
244 if (desc->wake_depth++ == 0) 254 if (desc->wake_depth++ == 0) {
245 desc->status |= IRQ_WAKEUP; 255 ret = set_irq_wake_real(irq, on);
246 else 256 if (ret)
247 set_wake = NULL; 257 desc->wake_depth = 0;
258 else
259 desc->status |= IRQ_WAKEUP;
260 }
248 } else { 261 } else {
249 if (desc->wake_depth == 0) { 262 if (desc->wake_depth == 0) {
250 printk(KERN_WARNING "Unbalanced IRQ %d " 263 WARN(1, "Unbalanced IRQ %d wake disable\n", irq);
251 "wake disable\n", irq); 264 } else if (--desc->wake_depth == 0) {
252 WARN_ON(1); 265 ret = set_irq_wake_real(irq, on);
253 } else if (--desc->wake_depth == 0) 266 if (ret)
254 desc->status &= ~IRQ_WAKEUP; 267 desc->wake_depth = 1;
255 else 268 else
256 set_wake = NULL; 269 desc->status &= ~IRQ_WAKEUP;
270 }
257 } 271 }
258 if (set_wake) 272
259 ret = desc->chip->set_wake(irq, on);
260 spin_unlock_irqrestore(&desc->lock, flags); 273 spin_unlock_irqrestore(&desc->lock, flags);
261 return ret; 274 return ret;
262} 275}
@@ -293,6 +306,30 @@ void compat_irq_chip_set_default_handler(struct irq_desc *desc)
293 desc->handle_irq = NULL; 306 desc->handle_irq = NULL;
294} 307}
295 308
309static int __irq_set_trigger(struct irq_chip *chip, unsigned int irq,
310 unsigned long flags)
311{
312 int ret;
313
314 if (!chip || !chip->set_type) {
315 /*
316 * IRQF_TRIGGER_* but the PIC does not support multiple
317 * flow-types?
318 */
319 pr_warning("No set_type function for IRQ %d (%s)\n", irq,
320 chip ? (chip->name ? : "unknown") : "unknown");
321 return 0;
322 }
323
324 ret = chip->set_type(irq, flags & IRQF_TRIGGER_MASK);
325
326 if (ret)
327 pr_err("setting flow type for irq %u failed (%pF)\n",
328 irq, chip->set_type);
329
330 return ret;
331}
332
296/* 333/*
297 * Internal function to register an irqaction - typically used to 334 * Internal function to register an irqaction - typically used to
298 * allocate special interrupts that are part of the architecture. 335 * allocate special interrupts that are part of the architecture.
@@ -304,6 +341,7 @@ int setup_irq(unsigned int irq, struct irqaction *new)
304 const char *old_name = NULL; 341 const char *old_name = NULL;
305 unsigned long flags; 342 unsigned long flags;
306 int shared = 0; 343 int shared = 0;
344 int ret;
307 345
308 if (irq >= NR_IRQS) 346 if (irq >= NR_IRQS)
309 return -EINVAL; 347 return -EINVAL;
@@ -361,35 +399,23 @@ int setup_irq(unsigned int irq, struct irqaction *new)
361 shared = 1; 399 shared = 1;
362 } 400 }
363 401
364 *p = new;
365
366 /* Exclude IRQ from balancing */
367 if (new->flags & IRQF_NOBALANCING)
368 desc->status |= IRQ_NO_BALANCING;
369
370 if (!shared) { 402 if (!shared) {
371 irq_chip_set_defaults(desc->chip); 403 irq_chip_set_defaults(desc->chip);
372 404
373#if defined(CONFIG_IRQ_PER_CPU)
374 if (new->flags & IRQF_PERCPU)
375 desc->status |= IRQ_PER_CPU;
376#endif
377
378 /* Setup the type (level, edge polarity) if configured: */ 405 /* Setup the type (level, edge polarity) if configured: */
379 if (new->flags & IRQF_TRIGGER_MASK) { 406 if (new->flags & IRQF_TRIGGER_MASK) {
380 if (desc->chip->set_type) 407 ret = __irq_set_trigger(desc->chip, irq, new->flags);
381 desc->chip->set_type(irq, 408
382 new->flags & IRQF_TRIGGER_MASK); 409 if (ret) {
383 else 410 spin_unlock_irqrestore(&desc->lock, flags);
384 /* 411 return ret;
385 * IRQF_TRIGGER_* but the PIC does not support 412 }
386 * multiple flow-types?
387 */
388 printk(KERN_WARNING "No IRQF_TRIGGER set_type "
389 "function for IRQ %d (%s)\n", irq,
390 desc->chip->name);
391 } else 413 } else
392 compat_irq_chip_set_default_handler(desc); 414 compat_irq_chip_set_default_handler(desc);
415#if defined(CONFIG_IRQ_PER_CPU)
416 if (new->flags & IRQF_PERCPU)
417 desc->status |= IRQ_PER_CPU;
418#endif
393 419
394 desc->status &= ~(IRQ_AUTODETECT | IRQ_WAITING | 420 desc->status &= ~(IRQ_AUTODETECT | IRQ_WAITING |
395 IRQ_INPROGRESS | IRQ_SPURIOUS_DISABLED); 421 IRQ_INPROGRESS | IRQ_SPURIOUS_DISABLED);
@@ -408,6 +434,13 @@ int setup_irq(unsigned int irq, struct irqaction *new)
408 /* Set default affinity mask once everything is setup */ 434 /* Set default affinity mask once everything is setup */
409 irq_select_affinity(irq); 435 irq_select_affinity(irq);
410 } 436 }
437
438 *p = new;
439
440 /* Exclude IRQ from balancing */
441 if (new->flags & IRQF_NOBALANCING)
442 desc->status |= IRQ_NO_BALANCING;
443
411 /* Reset broken irq detection when installing new handler */ 444 /* Reset broken irq detection when installing new handler */
412 desc->irq_count = 0; 445 desc->irq_count = 0;
413 desc->irqs_unhandled = 0; 446 desc->irqs_unhandled = 0;
diff --git a/kernel/kallsyms.c b/kernel/kallsyms.c
index 6fc0040f3e3a..38fc10ac7541 100644
--- a/kernel/kallsyms.c
+++ b/kernel/kallsyms.c
@@ -176,7 +176,7 @@ static unsigned long get_symbol_pos(unsigned long addr,
176 high = kallsyms_num_syms; 176 high = kallsyms_num_syms;
177 177
178 while (high - low > 1) { 178 while (high - low > 1) {
179 mid = (low + high) / 2; 179 mid = low + (high - low) / 2;
180 if (kallsyms_addresses[mid] <= addr) 180 if (kallsyms_addresses[mid] <= addr)
181 low = mid; 181 low = mid;
182 else 182 else
diff --git a/kernel/kmod.c b/kernel/kmod.c
index 8df97d3dfda8..2456d1a0befb 100644
--- a/kernel/kmod.c
+++ b/kernel/kmod.c
@@ -42,7 +42,7 @@ extern int max_threads;
42 42
43static struct workqueue_struct *khelper_wq; 43static struct workqueue_struct *khelper_wq;
44 44
45#ifdef CONFIG_KMOD 45#ifdef CONFIG_MODULES
46 46
47/* 47/*
48 modprobe_path is set via /proc/sys. 48 modprobe_path is set via /proc/sys.
@@ -352,16 +352,17 @@ static inline void register_pm_notifier_callback(void) {}
352 * @path: path to usermode executable 352 * @path: path to usermode executable
353 * @argv: arg vector for process 353 * @argv: arg vector for process
354 * @envp: environment for process 354 * @envp: environment for process
355 * @gfp_mask: gfp mask for memory allocation
355 * 356 *
356 * Returns either %NULL on allocation failure, or a subprocess_info 357 * Returns either %NULL on allocation failure, or a subprocess_info
357 * structure. This should be passed to call_usermodehelper_exec to 358 * structure. This should be passed to call_usermodehelper_exec to
358 * exec the process and free the structure. 359 * exec the process and free the structure.
359 */ 360 */
360struct subprocess_info *call_usermodehelper_setup(char *path, 361struct subprocess_info *call_usermodehelper_setup(char *path, char **argv,
361 char **argv, char **envp) 362 char **envp, gfp_t gfp_mask)
362{ 363{
363 struct subprocess_info *sub_info; 364 struct subprocess_info *sub_info;
364 sub_info = kzalloc(sizeof(struct subprocess_info), GFP_ATOMIC); 365 sub_info = kzalloc(sizeof(struct subprocess_info), gfp_mask);
365 if (!sub_info) 366 if (!sub_info)
366 goto out; 367 goto out;
367 368
@@ -417,12 +418,12 @@ int call_usermodehelper_stdinpipe(struct subprocess_info *sub_info,
417{ 418{
418 struct file *f; 419 struct file *f;
419 420
420 f = create_write_pipe(); 421 f = create_write_pipe(0);
421 if (IS_ERR(f)) 422 if (IS_ERR(f))
422 return PTR_ERR(f); 423 return PTR_ERR(f);
423 *filp = f; 424 *filp = f;
424 425
425 f = create_read_pipe(f); 426 f = create_read_pipe(f, 0);
426 if (IS_ERR(f)) { 427 if (IS_ERR(f)) {
427 free_write_pipe(*filp); 428 free_write_pipe(*filp);
428 return PTR_ERR(f); 429 return PTR_ERR(f);
@@ -494,7 +495,7 @@ int call_usermodehelper_pipe(char *path, char **argv, char **envp,
494 struct subprocess_info *sub_info; 495 struct subprocess_info *sub_info;
495 int ret; 496 int ret;
496 497
497 sub_info = call_usermodehelper_setup(path, argv, envp); 498 sub_info = call_usermodehelper_setup(path, argv, envp, GFP_KERNEL);
498 if (sub_info == NULL) 499 if (sub_info == NULL)
499 return -ENOMEM; 500 return -ENOMEM;
500 501
diff --git a/kernel/kprobes.c b/kernel/kprobes.c
index 1485ca8d0e00..75bc2cd9ebc6 100644
--- a/kernel/kprobes.c
+++ b/kernel/kprobes.c
@@ -62,6 +62,7 @@
62 addr = ((kprobe_opcode_t *)(kallsyms_lookup_name(name))) 62 addr = ((kprobe_opcode_t *)(kallsyms_lookup_name(name)))
63#endif 63#endif
64 64
65static int kprobes_initialized;
65static struct hlist_head kprobe_table[KPROBE_TABLE_SIZE]; 66static struct hlist_head kprobe_table[KPROBE_TABLE_SIZE];
66static struct hlist_head kretprobe_inst_table[KPROBE_TABLE_SIZE]; 67static struct hlist_head kretprobe_inst_table[KPROBE_TABLE_SIZE];
67 68
@@ -69,8 +70,15 @@ static struct hlist_head kretprobe_inst_table[KPROBE_TABLE_SIZE];
69static bool kprobe_enabled; 70static bool kprobe_enabled;
70 71
71DEFINE_MUTEX(kprobe_mutex); /* Protects kprobe_table */ 72DEFINE_MUTEX(kprobe_mutex); /* Protects kprobe_table */
72DEFINE_SPINLOCK(kretprobe_lock); /* Protects kretprobe_inst_table */
73static DEFINE_PER_CPU(struct kprobe *, kprobe_instance) = NULL; 73static DEFINE_PER_CPU(struct kprobe *, kprobe_instance) = NULL;
74static struct {
75 spinlock_t lock ____cacheline_aligned;
76} kretprobe_table_locks[KPROBE_TABLE_SIZE];
77
78static spinlock_t *kretprobe_table_lock_ptr(unsigned long hash)
79{
80 return &(kretprobe_table_locks[hash].lock);
81}
74 82
75/* 83/*
76 * Normally, functions that we'd want to prohibit kprobes in, are marked 84 * Normally, functions that we'd want to prohibit kprobes in, are marked
@@ -368,26 +376,53 @@ void __kprobes kprobes_inc_nmissed_count(struct kprobe *p)
368 return; 376 return;
369} 377}
370 378
371/* Called with kretprobe_lock held */
372void __kprobes recycle_rp_inst(struct kretprobe_instance *ri, 379void __kprobes recycle_rp_inst(struct kretprobe_instance *ri,
373 struct hlist_head *head) 380 struct hlist_head *head)
374{ 381{
382 struct kretprobe *rp = ri->rp;
383
375 /* remove rp inst off the rprobe_inst_table */ 384 /* remove rp inst off the rprobe_inst_table */
376 hlist_del(&ri->hlist); 385 hlist_del(&ri->hlist);
377 if (ri->rp) { 386 INIT_HLIST_NODE(&ri->hlist);
378 /* remove rp inst off the used list */ 387 if (likely(rp)) {
379 hlist_del(&ri->uflist); 388 spin_lock(&rp->lock);
380 /* put rp inst back onto the free list */ 389 hlist_add_head(&ri->hlist, &rp->free_instances);
381 INIT_HLIST_NODE(&ri->uflist); 390 spin_unlock(&rp->lock);
382 hlist_add_head(&ri->uflist, &ri->rp->free_instances);
383 } else 391 } else
384 /* Unregistering */ 392 /* Unregistering */
385 hlist_add_head(&ri->hlist, head); 393 hlist_add_head(&ri->hlist, head);
386} 394}
387 395
388struct hlist_head __kprobes *kretprobe_inst_table_head(struct task_struct *tsk) 396void kretprobe_hash_lock(struct task_struct *tsk,
397 struct hlist_head **head, unsigned long *flags)
389{ 398{
390 return &kretprobe_inst_table[hash_ptr(tsk, KPROBE_HASH_BITS)]; 399 unsigned long hash = hash_ptr(tsk, KPROBE_HASH_BITS);
400 spinlock_t *hlist_lock;
401
402 *head = &kretprobe_inst_table[hash];
403 hlist_lock = kretprobe_table_lock_ptr(hash);
404 spin_lock_irqsave(hlist_lock, *flags);
405}
406
407void kretprobe_table_lock(unsigned long hash, unsigned long *flags)
408{
409 spinlock_t *hlist_lock = kretprobe_table_lock_ptr(hash);
410 spin_lock_irqsave(hlist_lock, *flags);
411}
412
413void kretprobe_hash_unlock(struct task_struct *tsk, unsigned long *flags)
414{
415 unsigned long hash = hash_ptr(tsk, KPROBE_HASH_BITS);
416 spinlock_t *hlist_lock;
417
418 hlist_lock = kretprobe_table_lock_ptr(hash);
419 spin_unlock_irqrestore(hlist_lock, *flags);
420}
421
422void kretprobe_table_unlock(unsigned long hash, unsigned long *flags)
423{
424 spinlock_t *hlist_lock = kretprobe_table_lock_ptr(hash);
425 spin_unlock_irqrestore(hlist_lock, *flags);
391} 426}
392 427
393/* 428/*
@@ -401,17 +436,21 @@ void __kprobes kprobe_flush_task(struct task_struct *tk)
401 struct kretprobe_instance *ri; 436 struct kretprobe_instance *ri;
402 struct hlist_head *head, empty_rp; 437 struct hlist_head *head, empty_rp;
403 struct hlist_node *node, *tmp; 438 struct hlist_node *node, *tmp;
404 unsigned long flags = 0; 439 unsigned long hash, flags = 0;
405 440
406 INIT_HLIST_HEAD(&empty_rp); 441 if (unlikely(!kprobes_initialized))
407 spin_lock_irqsave(&kretprobe_lock, flags); 442 /* Early boot. kretprobe_table_locks not yet initialized. */
408 head = kretprobe_inst_table_head(tk); 443 return;
444
445 hash = hash_ptr(tk, KPROBE_HASH_BITS);
446 head = &kretprobe_inst_table[hash];
447 kretprobe_table_lock(hash, &flags);
409 hlist_for_each_entry_safe(ri, node, tmp, head, hlist) { 448 hlist_for_each_entry_safe(ri, node, tmp, head, hlist) {
410 if (ri->task == tk) 449 if (ri->task == tk)
411 recycle_rp_inst(ri, &empty_rp); 450 recycle_rp_inst(ri, &empty_rp);
412 } 451 }
413 spin_unlock_irqrestore(&kretprobe_lock, flags); 452 kretprobe_table_unlock(hash, &flags);
414 453 INIT_HLIST_HEAD(&empty_rp);
415 hlist_for_each_entry_safe(ri, node, tmp, &empty_rp, hlist) { 454 hlist_for_each_entry_safe(ri, node, tmp, &empty_rp, hlist) {
416 hlist_del(&ri->hlist); 455 hlist_del(&ri->hlist);
417 kfree(ri); 456 kfree(ri);
@@ -423,24 +462,29 @@ static inline void free_rp_inst(struct kretprobe *rp)
423 struct kretprobe_instance *ri; 462 struct kretprobe_instance *ri;
424 struct hlist_node *pos, *next; 463 struct hlist_node *pos, *next;
425 464
426 hlist_for_each_entry_safe(ri, pos, next, &rp->free_instances, uflist) { 465 hlist_for_each_entry_safe(ri, pos, next, &rp->free_instances, hlist) {
427 hlist_del(&ri->uflist); 466 hlist_del(&ri->hlist);
428 kfree(ri); 467 kfree(ri);
429 } 468 }
430} 469}
431 470
432static void __kprobes cleanup_rp_inst(struct kretprobe *rp) 471static void __kprobes cleanup_rp_inst(struct kretprobe *rp)
433{ 472{
434 unsigned long flags; 473 unsigned long flags, hash;
435 struct kretprobe_instance *ri; 474 struct kretprobe_instance *ri;
436 struct hlist_node *pos, *next; 475 struct hlist_node *pos, *next;
476 struct hlist_head *head;
477
437 /* No race here */ 478 /* No race here */
438 spin_lock_irqsave(&kretprobe_lock, flags); 479 for (hash = 0; hash < KPROBE_TABLE_SIZE; hash++) {
439 hlist_for_each_entry_safe(ri, pos, next, &rp->used_instances, uflist) { 480 kretprobe_table_lock(hash, &flags);
440 ri->rp = NULL; 481 head = &kretprobe_inst_table[hash];
441 hlist_del(&ri->uflist); 482 hlist_for_each_entry_safe(ri, pos, next, head, hlist) {
483 if (ri->rp == rp)
484 ri->rp = NULL;
485 }
486 kretprobe_table_unlock(hash, &flags);
442 } 487 }
443 spin_unlock_irqrestore(&kretprobe_lock, flags);
444 free_rp_inst(rp); 488 free_rp_inst(rp);
445} 489}
446 490
@@ -831,32 +875,37 @@ static int __kprobes pre_handler_kretprobe(struct kprobe *p,
831 struct pt_regs *regs) 875 struct pt_regs *regs)
832{ 876{
833 struct kretprobe *rp = container_of(p, struct kretprobe, kp); 877 struct kretprobe *rp = container_of(p, struct kretprobe, kp);
834 unsigned long flags = 0; 878 unsigned long hash, flags = 0;
879 struct kretprobe_instance *ri;
835 880
836 /*TODO: consider to only swap the RA after the last pre_handler fired */ 881 /*TODO: consider to only swap the RA after the last pre_handler fired */
837 spin_lock_irqsave(&kretprobe_lock, flags); 882 hash = hash_ptr(current, KPROBE_HASH_BITS);
883 spin_lock_irqsave(&rp->lock, flags);
838 if (!hlist_empty(&rp->free_instances)) { 884 if (!hlist_empty(&rp->free_instances)) {
839 struct kretprobe_instance *ri;
840
841 ri = hlist_entry(rp->free_instances.first, 885 ri = hlist_entry(rp->free_instances.first,
842 struct kretprobe_instance, uflist); 886 struct kretprobe_instance, hlist);
887 hlist_del(&ri->hlist);
888 spin_unlock_irqrestore(&rp->lock, flags);
889
843 ri->rp = rp; 890 ri->rp = rp;
844 ri->task = current; 891 ri->task = current;
845 892
846 if (rp->entry_handler && rp->entry_handler(ri, regs)) { 893 if (rp->entry_handler && rp->entry_handler(ri, regs)) {
847 spin_unlock_irqrestore(&kretprobe_lock, flags); 894 spin_unlock_irqrestore(&rp->lock, flags);
848 return 0; 895 return 0;
849 } 896 }
850 897
851 arch_prepare_kretprobe(ri, regs); 898 arch_prepare_kretprobe(ri, regs);
852 899
853 /* XXX(hch): why is there no hlist_move_head? */ 900 /* XXX(hch): why is there no hlist_move_head? */
854 hlist_del(&ri->uflist); 901 INIT_HLIST_NODE(&ri->hlist);
855 hlist_add_head(&ri->uflist, &ri->rp->used_instances); 902 kretprobe_table_lock(hash, &flags);
856 hlist_add_head(&ri->hlist, kretprobe_inst_table_head(ri->task)); 903 hlist_add_head(&ri->hlist, &kretprobe_inst_table[hash]);
857 } else 904 kretprobe_table_unlock(hash, &flags);
905 } else {
858 rp->nmissed++; 906 rp->nmissed++;
859 spin_unlock_irqrestore(&kretprobe_lock, flags); 907 spin_unlock_irqrestore(&rp->lock, flags);
908 }
860 return 0; 909 return 0;
861} 910}
862 911
@@ -892,7 +941,7 @@ static int __kprobes __register_kretprobe(struct kretprobe *rp,
892 rp->maxactive = NR_CPUS; 941 rp->maxactive = NR_CPUS;
893#endif 942#endif
894 } 943 }
895 INIT_HLIST_HEAD(&rp->used_instances); 944 spin_lock_init(&rp->lock);
896 INIT_HLIST_HEAD(&rp->free_instances); 945 INIT_HLIST_HEAD(&rp->free_instances);
897 for (i = 0; i < rp->maxactive; i++) { 946 for (i = 0; i < rp->maxactive; i++) {
898 inst = kmalloc(sizeof(struct kretprobe_instance) + 947 inst = kmalloc(sizeof(struct kretprobe_instance) +
@@ -901,8 +950,8 @@ static int __kprobes __register_kretprobe(struct kretprobe *rp,
901 free_rp_inst(rp); 950 free_rp_inst(rp);
902 return -ENOMEM; 951 return -ENOMEM;
903 } 952 }
904 INIT_HLIST_NODE(&inst->uflist); 953 INIT_HLIST_NODE(&inst->hlist);
905 hlist_add_head(&inst->uflist, &rp->free_instances); 954 hlist_add_head(&inst->hlist, &rp->free_instances);
906 } 955 }
907 956
908 rp->nmissed = 0; 957 rp->nmissed = 0;
@@ -1009,6 +1058,7 @@ static int __init init_kprobes(void)
1009 for (i = 0; i < KPROBE_TABLE_SIZE; i++) { 1058 for (i = 0; i < KPROBE_TABLE_SIZE; i++) {
1010 INIT_HLIST_HEAD(&kprobe_table[i]); 1059 INIT_HLIST_HEAD(&kprobe_table[i]);
1011 INIT_HLIST_HEAD(&kretprobe_inst_table[i]); 1060 INIT_HLIST_HEAD(&kretprobe_inst_table[i]);
1061 spin_lock_init(&(kretprobe_table_locks[i].lock));
1012 } 1062 }
1013 1063
1014 /* 1064 /*
@@ -1050,6 +1100,7 @@ static int __init init_kprobes(void)
1050 err = arch_init_kprobes(); 1100 err = arch_init_kprobes();
1051 if (!err) 1101 if (!err)
1052 err = register_die_notifier(&kprobe_exceptions_nb); 1102 err = register_die_notifier(&kprobe_exceptions_nb);
1103 kprobes_initialized = (err == 0);
1053 1104
1054 if (!err) 1105 if (!err)
1055 init_test_probes(); 1106 init_test_probes();
@@ -1286,13 +1337,8 @@ EXPORT_SYMBOL_GPL(register_jprobe);
1286EXPORT_SYMBOL_GPL(unregister_jprobe); 1337EXPORT_SYMBOL_GPL(unregister_jprobe);
1287EXPORT_SYMBOL_GPL(register_jprobes); 1338EXPORT_SYMBOL_GPL(register_jprobes);
1288EXPORT_SYMBOL_GPL(unregister_jprobes); 1339EXPORT_SYMBOL_GPL(unregister_jprobes);
1289#ifdef CONFIG_KPROBES
1290EXPORT_SYMBOL_GPL(jprobe_return); 1340EXPORT_SYMBOL_GPL(jprobe_return);
1291#endif
1292
1293#ifdef CONFIG_KPROBES
1294EXPORT_SYMBOL_GPL(register_kretprobe); 1341EXPORT_SYMBOL_GPL(register_kretprobe);
1295EXPORT_SYMBOL_GPL(unregister_kretprobe); 1342EXPORT_SYMBOL_GPL(unregister_kretprobe);
1296EXPORT_SYMBOL_GPL(register_kretprobes); 1343EXPORT_SYMBOL_GPL(register_kretprobes);
1297EXPORT_SYMBOL_GPL(unregister_kretprobes); 1344EXPORT_SYMBOL_GPL(unregister_kretprobes);
1298#endif
diff --git a/kernel/kthread.c b/kernel/kthread.c
index ac3fb7326641..6111c27491b1 100644
--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -106,7 +106,7 @@ static void create_kthread(struct kthread_create_info *create)
106 */ 106 */
107 sched_setscheduler(create->result, SCHED_NORMAL, &param); 107 sched_setscheduler(create->result, SCHED_NORMAL, &param);
108 set_user_nice(create->result, KTHREAD_NICE_LEVEL); 108 set_user_nice(create->result, KTHREAD_NICE_LEVEL);
109 set_cpus_allowed(create->result, CPU_MASK_ALL); 109 set_cpus_allowed_ptr(create->result, CPU_MASK_ALL_PTR);
110 } 110 }
111 complete(&create->done); 111 complete(&create->done);
112} 112}
@@ -233,7 +233,7 @@ int kthreadd(void *unused)
233 set_task_comm(tsk, "kthreadd"); 233 set_task_comm(tsk, "kthreadd");
234 ignore_signals(tsk); 234 ignore_signals(tsk);
235 set_user_nice(tsk, KTHREAD_NICE_LEVEL); 235 set_user_nice(tsk, KTHREAD_NICE_LEVEL);
236 set_cpus_allowed(tsk, CPU_MASK_ALL); 236 set_cpus_allowed_ptr(tsk, CPU_MASK_ALL_PTR);
237 237
238 current->flags |= PF_NOFREEZE | PF_FREEZER_NOSIG; 238 current->flags |= PF_NOFREEZE | PF_FREEZER_NOSIG;
239 239
diff --git a/kernel/marker.c b/kernel/marker.c
index 1abfb923b761..971da5317903 100644
--- a/kernel/marker.c
+++ b/kernel/marker.c
@@ -441,7 +441,7 @@ static int remove_marker(const char *name)
441 hlist_del(&e->hlist); 441 hlist_del(&e->hlist);
442 /* Make sure the call_rcu has been executed */ 442 /* Make sure the call_rcu has been executed */
443 if (e->rcu_pending) 443 if (e->rcu_pending)
444 rcu_barrier(); 444 rcu_barrier_sched();
445 kfree(e); 445 kfree(e);
446 return 0; 446 return 0;
447} 447}
@@ -476,7 +476,7 @@ static int marker_set_format(struct marker_entry **entry, const char *format)
476 hlist_del(&(*entry)->hlist); 476 hlist_del(&(*entry)->hlist);
477 /* Make sure the call_rcu has been executed */ 477 /* Make sure the call_rcu has been executed */
478 if ((*entry)->rcu_pending) 478 if ((*entry)->rcu_pending)
479 rcu_barrier(); 479 rcu_barrier_sched();
480 kfree(*entry); 480 kfree(*entry);
481 *entry = e; 481 *entry = e;
482 trace_mark(core_marker_format, "name %s format %s", 482 trace_mark(core_marker_format, "name %s format %s",
@@ -655,7 +655,7 @@ int marker_probe_register(const char *name, const char *format,
655 * make sure it's executed now. 655 * make sure it's executed now.
656 */ 656 */
657 if (entry->rcu_pending) 657 if (entry->rcu_pending)
658 rcu_barrier(); 658 rcu_barrier_sched();
659 old = marker_entry_add_probe(entry, probe, probe_private); 659 old = marker_entry_add_probe(entry, probe, probe_private);
660 if (IS_ERR(old)) { 660 if (IS_ERR(old)) {
661 ret = PTR_ERR(old); 661 ret = PTR_ERR(old);
@@ -670,10 +670,7 @@ int marker_probe_register(const char *name, const char *format,
670 entry->rcu_pending = 1; 670 entry->rcu_pending = 1;
671 /* write rcu_pending before calling the RCU callback */ 671 /* write rcu_pending before calling the RCU callback */
672 smp_wmb(); 672 smp_wmb();
673#ifdef CONFIG_PREEMPT_RCU 673 call_rcu_sched(&entry->rcu, free_old_closure);
674 synchronize_sched(); /* Until we have the call_rcu_sched() */
675#endif
676 call_rcu(&entry->rcu, free_old_closure);
677end: 674end:
678 mutex_unlock(&markers_mutex); 675 mutex_unlock(&markers_mutex);
679 return ret; 676 return ret;
@@ -704,7 +701,7 @@ int marker_probe_unregister(const char *name,
704 if (!entry) 701 if (!entry)
705 goto end; 702 goto end;
706 if (entry->rcu_pending) 703 if (entry->rcu_pending)
707 rcu_barrier(); 704 rcu_barrier_sched();
708 old = marker_entry_remove_probe(entry, probe, probe_private); 705 old = marker_entry_remove_probe(entry, probe, probe_private);
709 mutex_unlock(&markers_mutex); 706 mutex_unlock(&markers_mutex);
710 marker_update_probes(); /* may update entry */ 707 marker_update_probes(); /* may update entry */
@@ -716,10 +713,7 @@ int marker_probe_unregister(const char *name,
716 entry->rcu_pending = 1; 713 entry->rcu_pending = 1;
717 /* write rcu_pending before calling the RCU callback */ 714 /* write rcu_pending before calling the RCU callback */
718 smp_wmb(); 715 smp_wmb();
719#ifdef CONFIG_PREEMPT_RCU 716 call_rcu_sched(&entry->rcu, free_old_closure);
720 synchronize_sched(); /* Until we have the call_rcu_sched() */
721#endif
722 call_rcu(&entry->rcu, free_old_closure);
723 remove_marker(name); /* Ignore busy error message */ 717 remove_marker(name); /* Ignore busy error message */
724 ret = 0; 718 ret = 0;
725end: 719end:
@@ -786,7 +780,7 @@ int marker_probe_unregister_private_data(marker_probe_func *probe,
786 goto end; 780 goto end;
787 } 781 }
788 if (entry->rcu_pending) 782 if (entry->rcu_pending)
789 rcu_barrier(); 783 rcu_barrier_sched();
790 old = marker_entry_remove_probe(entry, NULL, probe_private); 784 old = marker_entry_remove_probe(entry, NULL, probe_private);
791 mutex_unlock(&markers_mutex); 785 mutex_unlock(&markers_mutex);
792 marker_update_probes(); /* may update entry */ 786 marker_update_probes(); /* may update entry */
@@ -797,10 +791,7 @@ int marker_probe_unregister_private_data(marker_probe_func *probe,
797 entry->rcu_pending = 1; 791 entry->rcu_pending = 1;
798 /* write rcu_pending before calling the RCU callback */ 792 /* write rcu_pending before calling the RCU callback */
799 smp_wmb(); 793 smp_wmb();
800#ifdef CONFIG_PREEMPT_RCU 794 call_rcu_sched(&entry->rcu, free_old_closure);
801 synchronize_sched(); /* Until we have the call_rcu_sched() */
802#endif
803 call_rcu(&entry->rcu, free_old_closure);
804 remove_marker(entry->name); /* Ignore busy error message */ 795 remove_marker(entry->name); /* Ignore busy error message */
805end: 796end:
806 mutex_unlock(&markers_mutex); 797 mutex_unlock(&markers_mutex);
diff --git a/kernel/module.c b/kernel/module.c
index 5f80478b746d..d8b5605132a0 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -70,6 +70,9 @@ static DECLARE_WAIT_QUEUE_HEAD(module_wq);
70 70
71static BLOCKING_NOTIFIER_HEAD(module_notify_list); 71static BLOCKING_NOTIFIER_HEAD(module_notify_list);
72 72
73/* Bounds of module allocation, for speeding __module_text_address */
74static unsigned long module_addr_min = -1UL, module_addr_max = 0;
75
73int register_module_notifier(struct notifier_block * nb) 76int register_module_notifier(struct notifier_block * nb)
74{ 77{
75 return blocking_notifier_chain_register(&module_notify_list, nb); 78 return blocking_notifier_chain_register(&module_notify_list, nb);
@@ -134,17 +137,19 @@ extern const struct kernel_symbol __start___ksymtab_gpl[];
134extern const struct kernel_symbol __stop___ksymtab_gpl[]; 137extern const struct kernel_symbol __stop___ksymtab_gpl[];
135extern const struct kernel_symbol __start___ksymtab_gpl_future[]; 138extern const struct kernel_symbol __start___ksymtab_gpl_future[];
136extern const struct kernel_symbol __stop___ksymtab_gpl_future[]; 139extern const struct kernel_symbol __stop___ksymtab_gpl_future[];
137extern const struct kernel_symbol __start___ksymtab_unused[];
138extern const struct kernel_symbol __stop___ksymtab_unused[];
139extern const struct kernel_symbol __start___ksymtab_unused_gpl[];
140extern const struct kernel_symbol __stop___ksymtab_unused_gpl[];
141extern const struct kernel_symbol __start___ksymtab_gpl_future[]; 140extern const struct kernel_symbol __start___ksymtab_gpl_future[];
142extern const struct kernel_symbol __stop___ksymtab_gpl_future[]; 141extern const struct kernel_symbol __stop___ksymtab_gpl_future[];
143extern const unsigned long __start___kcrctab[]; 142extern const unsigned long __start___kcrctab[];
144extern const unsigned long __start___kcrctab_gpl[]; 143extern const unsigned long __start___kcrctab_gpl[];
145extern const unsigned long __start___kcrctab_gpl_future[]; 144extern const unsigned long __start___kcrctab_gpl_future[];
145#ifdef CONFIG_UNUSED_SYMBOLS
146extern const struct kernel_symbol __start___ksymtab_unused[];
147extern const struct kernel_symbol __stop___ksymtab_unused[];
148extern const struct kernel_symbol __start___ksymtab_unused_gpl[];
149extern const struct kernel_symbol __stop___ksymtab_unused_gpl[];
146extern const unsigned long __start___kcrctab_unused[]; 150extern const unsigned long __start___kcrctab_unused[];
147extern const unsigned long __start___kcrctab_unused_gpl[]; 151extern const unsigned long __start___kcrctab_unused_gpl[];
152#endif
148 153
149#ifndef CONFIG_MODVERSIONS 154#ifndef CONFIG_MODVERSIONS
150#define symversion(base, idx) NULL 155#define symversion(base, idx) NULL
@@ -152,156 +157,186 @@ extern const unsigned long __start___kcrctab_unused_gpl[];
152#define symversion(base, idx) ((base != NULL) ? ((base) + (idx)) : NULL) 157#define symversion(base, idx) ((base != NULL) ? ((base) + (idx)) : NULL)
153#endif 158#endif
154 159
155/* lookup symbol in given range of kernel_symbols */
156static const struct kernel_symbol *lookup_symbol(const char *name,
157 const struct kernel_symbol *start,
158 const struct kernel_symbol *stop)
159{
160 const struct kernel_symbol *ks = start;
161 for (; ks < stop; ks++)
162 if (strcmp(ks->name, name) == 0)
163 return ks;
164 return NULL;
165}
166
167static bool always_ok(bool gplok, bool warn, const char *name)
168{
169 return true;
170}
171
172static bool printk_unused_warning(bool gplok, bool warn, const char *name)
173{
174 if (warn) {
175 printk(KERN_WARNING "Symbol %s is marked as UNUSED, "
176 "however this module is using it.\n", name);
177 printk(KERN_WARNING
178 "This symbol will go away in the future.\n");
179 printk(KERN_WARNING
180 "Please evalute if this is the right api to use and if "
181 "it really is, submit a report the linux kernel "
182 "mailinglist together with submitting your code for "
183 "inclusion.\n");
184 }
185 return true;
186}
187
188static bool gpl_only_unused_warning(bool gplok, bool warn, const char *name)
189{
190 if (!gplok)
191 return false;
192 return printk_unused_warning(gplok, warn, name);
193}
194
195static bool gpl_only(bool gplok, bool warn, const char *name)
196{
197 return gplok;
198}
199
200static bool warn_if_not_gpl(bool gplok, bool warn, const char *name)
201{
202 if (!gplok && warn) {
203 printk(KERN_WARNING "Symbol %s is being used "
204 "by a non-GPL module, which will not "
205 "be allowed in the future\n", name);
206 printk(KERN_WARNING "Please see the file "
207 "Documentation/feature-removal-schedule.txt "
208 "in the kernel source tree for more details.\n");
209 }
210 return true;
211}
212
213struct symsearch { 160struct symsearch {
214 const struct kernel_symbol *start, *stop; 161 const struct kernel_symbol *start, *stop;
215 const unsigned long *crcs; 162 const unsigned long *crcs;
216 bool (*check)(bool gplok, bool warn, const char *name); 163 enum {
164 NOT_GPL_ONLY,
165 GPL_ONLY,
166 WILL_BE_GPL_ONLY,
167 } licence;
168 bool unused;
217}; 169};
218 170
219/* Look through this array of symbol tables for a symbol match which 171static bool each_symbol_in_section(const struct symsearch *arr,
220 * passes the check function. */ 172 unsigned int arrsize,
221static const struct kernel_symbol *search_symarrays(const struct symsearch *arr, 173 struct module *owner,
222 unsigned int num, 174 bool (*fn)(const struct symsearch *syms,
223 const char *name, 175 struct module *owner,
224 bool gplok, 176 unsigned int symnum, void *data),
225 bool warn, 177 void *data)
226 const unsigned long **crc)
227{ 178{
228 unsigned int i; 179 unsigned int i, j;
229 const struct kernel_symbol *ks;
230 180
231 for (i = 0; i < num; i++) { 181 for (j = 0; j < arrsize; j++) {
232 ks = lookup_symbol(name, arr[i].start, arr[i].stop); 182 for (i = 0; i < arr[j].stop - arr[j].start; i++)
233 if (!ks || !arr[i].check(gplok, warn, name)) 183 if (fn(&arr[j], owner, i, data))
234 continue; 184 return true;
235
236 if (crc)
237 *crc = symversion(arr[i].crcs, ks - arr[i].start);
238 return ks;
239 } 185 }
240 return NULL; 186
187 return false;
241} 188}
242 189
243/* Find a symbol, return value, (optional) crc and (optional) module 190/* Returns true as soon as fn returns true, otherwise false. */
244 * which owns it */ 191static bool each_symbol(bool (*fn)(const struct symsearch *arr,
245static unsigned long find_symbol(const char *name, 192 struct module *owner,
246 struct module **owner, 193 unsigned int symnum, void *data),
247 const unsigned long **crc, 194 void *data)
248 bool gplok,
249 bool warn)
250{ 195{
251 struct module *mod; 196 struct module *mod;
252 const struct kernel_symbol *ks;
253 const struct symsearch arr[] = { 197 const struct symsearch arr[] = {
254 { __start___ksymtab, __stop___ksymtab, __start___kcrctab, 198 { __start___ksymtab, __stop___ksymtab, __start___kcrctab,
255 always_ok }, 199 NOT_GPL_ONLY, false },
256 { __start___ksymtab_gpl, __stop___ksymtab_gpl, 200 { __start___ksymtab_gpl, __stop___ksymtab_gpl,
257 __start___kcrctab_gpl, gpl_only }, 201 __start___kcrctab_gpl,
202 GPL_ONLY, false },
258 { __start___ksymtab_gpl_future, __stop___ksymtab_gpl_future, 203 { __start___ksymtab_gpl_future, __stop___ksymtab_gpl_future,
259 __start___kcrctab_gpl_future, warn_if_not_gpl }, 204 __start___kcrctab_gpl_future,
205 WILL_BE_GPL_ONLY, false },
206#ifdef CONFIG_UNUSED_SYMBOLS
260 { __start___ksymtab_unused, __stop___ksymtab_unused, 207 { __start___ksymtab_unused, __stop___ksymtab_unused,
261 __start___kcrctab_unused, printk_unused_warning }, 208 __start___kcrctab_unused,
209 NOT_GPL_ONLY, true },
262 { __start___ksymtab_unused_gpl, __stop___ksymtab_unused_gpl, 210 { __start___ksymtab_unused_gpl, __stop___ksymtab_unused_gpl,
263 __start___kcrctab_unused_gpl, gpl_only_unused_warning }, 211 __start___kcrctab_unused_gpl,
212 GPL_ONLY, true },
213#endif
264 }; 214 };
265 215
266 /* Core kernel first. */ 216 if (each_symbol_in_section(arr, ARRAY_SIZE(arr), NULL, fn, data))
267 ks = search_symarrays(arr, ARRAY_SIZE(arr), name, gplok, warn, crc); 217 return true;
268 if (ks) {
269 if (owner)
270 *owner = NULL;
271 return ks->value;
272 }
273 218
274 /* Now try modules. */
275 list_for_each_entry(mod, &modules, list) { 219 list_for_each_entry(mod, &modules, list) {
276 struct symsearch arr[] = { 220 struct symsearch arr[] = {
277 { mod->syms, mod->syms + mod->num_syms, mod->crcs, 221 { mod->syms, mod->syms + mod->num_syms, mod->crcs,
278 always_ok }, 222 NOT_GPL_ONLY, false },
279 { mod->gpl_syms, mod->gpl_syms + mod->num_gpl_syms, 223 { mod->gpl_syms, mod->gpl_syms + mod->num_gpl_syms,
280 mod->gpl_crcs, gpl_only }, 224 mod->gpl_crcs,
225 GPL_ONLY, false },
281 { mod->gpl_future_syms, 226 { mod->gpl_future_syms,
282 mod->gpl_future_syms + mod->num_gpl_future_syms, 227 mod->gpl_future_syms + mod->num_gpl_future_syms,
283 mod->gpl_future_crcs, warn_if_not_gpl }, 228 mod->gpl_future_crcs,
229 WILL_BE_GPL_ONLY, false },
230#ifdef CONFIG_UNUSED_SYMBOLS
284 { mod->unused_syms, 231 { mod->unused_syms,
285 mod->unused_syms + mod->num_unused_syms, 232 mod->unused_syms + mod->num_unused_syms,
286 mod->unused_crcs, printk_unused_warning }, 233 mod->unused_crcs,
234 NOT_GPL_ONLY, true },
287 { mod->unused_gpl_syms, 235 { mod->unused_gpl_syms,
288 mod->unused_gpl_syms + mod->num_unused_gpl_syms, 236 mod->unused_gpl_syms + mod->num_unused_gpl_syms,
289 mod->unused_gpl_crcs, gpl_only_unused_warning }, 237 mod->unused_gpl_crcs,
238 GPL_ONLY, true },
239#endif
290 }; 240 };
291 241
292 ks = search_symarrays(arr, ARRAY_SIZE(arr), 242 if (each_symbol_in_section(arr, ARRAY_SIZE(arr), mod, fn, data))
293 name, gplok, warn, crc); 243 return true;
294 if (ks) { 244 }
295 if (owner) 245 return false;
296 *owner = mod; 246}
297 return ks->value; 247
248struct find_symbol_arg {
249 /* Input */
250 const char *name;
251 bool gplok;
252 bool warn;
253
254 /* Output */
255 struct module *owner;
256 const unsigned long *crc;
257 unsigned long value;
258};
259
260static bool find_symbol_in_section(const struct symsearch *syms,
261 struct module *owner,
262 unsigned int symnum, void *data)
263{
264 struct find_symbol_arg *fsa = data;
265
266 if (strcmp(syms->start[symnum].name, fsa->name) != 0)
267 return false;
268
269 if (!fsa->gplok) {
270 if (syms->licence == GPL_ONLY)
271 return false;
272 if (syms->licence == WILL_BE_GPL_ONLY && fsa->warn) {
273 printk(KERN_WARNING "Symbol %s is being used "
274 "by a non-GPL module, which will not "
275 "be allowed in the future\n", fsa->name);
276 printk(KERN_WARNING "Please see the file "
277 "Documentation/feature-removal-schedule.txt "
278 "in the kernel source tree for more details.\n");
298 } 279 }
299 } 280 }
300 281
282#ifdef CONFIG_UNUSED_SYMBOLS
283 if (syms->unused && fsa->warn) {
284 printk(KERN_WARNING "Symbol %s is marked as UNUSED, "
285 "however this module is using it.\n", fsa->name);
286 printk(KERN_WARNING
287 "This symbol will go away in the future.\n");
288 printk(KERN_WARNING
289 "Please evalute if this is the right api to use and if "
290 "it really is, submit a report the linux kernel "
291 "mailinglist together with submitting your code for "
292 "inclusion.\n");
293 }
294#endif
295
296 fsa->owner = owner;
297 fsa->crc = symversion(syms->crcs, symnum);
298 fsa->value = syms->start[symnum].value;
299 return true;
300}
301
302/* Find a symbol, return value, (optional) crc and (optional) module
303 * which owns it */
304static unsigned long find_symbol(const char *name,
305 struct module **owner,
306 const unsigned long **crc,
307 bool gplok,
308 bool warn)
309{
310 struct find_symbol_arg fsa;
311
312 fsa.name = name;
313 fsa.gplok = gplok;
314 fsa.warn = warn;
315
316 if (each_symbol(find_symbol_in_section, &fsa)) {
317 if (owner)
318 *owner = fsa.owner;
319 if (crc)
320 *crc = fsa.crc;
321 return fsa.value;
322 }
323
301 DEBUGP("Failed to find symbol %s\n", name); 324 DEBUGP("Failed to find symbol %s\n", name);
302 return -ENOENT; 325 return -ENOENT;
303} 326}
304 327
328/* lookup symbol in given range of kernel_symbols */
329static const struct kernel_symbol *lookup_symbol(const char *name,
330 const struct kernel_symbol *start,
331 const struct kernel_symbol *stop)
332{
333 const struct kernel_symbol *ks = start;
334 for (; ks < stop; ks++)
335 if (strcmp(ks->name, name) == 0)
336 return ks;
337 return NULL;
338}
339
305/* Search for module by name: must hold module_mutex. */ 340/* Search for module by name: must hold module_mutex. */
306static struct module *find_module(const char *name) 341static struct module *find_module(const char *name)
307{ 342{
@@ -639,8 +674,8 @@ static int __try_stop_module(void *_sref)
639{ 674{
640 struct stopref *sref = _sref; 675 struct stopref *sref = _sref;
641 676
642 /* If it's not unused, quit unless we are told to block. */ 677 /* If it's not unused, quit unless we're forcing. */
643 if ((sref->flags & O_NONBLOCK) && module_refcount(sref->mod) != 0) { 678 if (module_refcount(sref->mod) != 0) {
644 if (!(*sref->forced = try_force_unload(sref->flags))) 679 if (!(*sref->forced = try_force_unload(sref->flags)))
645 return -EWOULDBLOCK; 680 return -EWOULDBLOCK;
646 } 681 }
@@ -652,9 +687,16 @@ static int __try_stop_module(void *_sref)
652 687
653static int try_stop_module(struct module *mod, int flags, int *forced) 688static int try_stop_module(struct module *mod, int flags, int *forced)
654{ 689{
655 struct stopref sref = { mod, flags, forced }; 690 if (flags & O_NONBLOCK) {
691 struct stopref sref = { mod, flags, forced };
656 692
657 return stop_machine_run(__try_stop_module, &sref, NR_CPUS); 693 return stop_machine_run(__try_stop_module, &sref, NR_CPUS);
694 } else {
695 /* We don't need to stop the machine for this. */
696 mod->state = MODULE_STATE_GOING;
697 synchronize_sched();
698 return 0;
699 }
658} 700}
659 701
660unsigned int module_refcount(struct module *mod) 702unsigned int module_refcount(struct module *mod)
@@ -1445,8 +1487,10 @@ static int verify_export_symbols(struct module *mod)
1445 { mod->syms, mod->num_syms }, 1487 { mod->syms, mod->num_syms },
1446 { mod->gpl_syms, mod->num_gpl_syms }, 1488 { mod->gpl_syms, mod->num_gpl_syms },
1447 { mod->gpl_future_syms, mod->num_gpl_future_syms }, 1489 { mod->gpl_future_syms, mod->num_gpl_future_syms },
1490#ifdef CONFIG_UNUSED_SYMBOLS
1448 { mod->unused_syms, mod->num_unused_syms }, 1491 { mod->unused_syms, mod->num_unused_syms },
1449 { mod->unused_gpl_syms, mod->num_unused_gpl_syms }, 1492 { mod->unused_gpl_syms, mod->num_unused_gpl_syms },
1493#endif
1450 }; 1494 };
1451 1495
1452 for (i = 0; i < ARRAY_SIZE(arr); i++) { 1496 for (i = 0; i < ARRAY_SIZE(arr); i++) {
@@ -1526,7 +1570,7 @@ static int simplify_symbols(Elf_Shdr *sechdrs,
1526} 1570}
1527 1571
1528/* Update size with this section: return offset. */ 1572/* Update size with this section: return offset. */
1529static long get_offset(unsigned long *size, Elf_Shdr *sechdr) 1573static long get_offset(unsigned int *size, Elf_Shdr *sechdr)
1530{ 1574{
1531 long ret; 1575 long ret;
1532 1576
@@ -1738,6 +1782,20 @@ static inline void add_kallsyms(struct module *mod,
1738} 1782}
1739#endif /* CONFIG_KALLSYMS */ 1783#endif /* CONFIG_KALLSYMS */
1740 1784
1785static void *module_alloc_update_bounds(unsigned long size)
1786{
1787 void *ret = module_alloc(size);
1788
1789 if (ret) {
1790 /* Update module bounds. */
1791 if ((unsigned long)ret < module_addr_min)
1792 module_addr_min = (unsigned long)ret;
1793 if ((unsigned long)ret + size > module_addr_max)
1794 module_addr_max = (unsigned long)ret + size;
1795 }
1796 return ret;
1797}
1798
1741/* Allocate and load the module: note that size of section 0 is always 1799/* Allocate and load the module: note that size of section 0 is always
1742 zero, and we rely on this for optional sections. */ 1800 zero, and we rely on this for optional sections. */
1743static struct module *load_module(void __user *umod, 1801static struct module *load_module(void __user *umod,
@@ -1764,10 +1822,12 @@ static struct module *load_module(void __user *umod,
1764 unsigned int gplfutureindex; 1822 unsigned int gplfutureindex;
1765 unsigned int gplfuturecrcindex; 1823 unsigned int gplfuturecrcindex;
1766 unsigned int unwindex = 0; 1824 unsigned int unwindex = 0;
1825#ifdef CONFIG_UNUSED_SYMBOLS
1767 unsigned int unusedindex; 1826 unsigned int unusedindex;
1768 unsigned int unusedcrcindex; 1827 unsigned int unusedcrcindex;
1769 unsigned int unusedgplindex; 1828 unsigned int unusedgplindex;
1770 unsigned int unusedgplcrcindex; 1829 unsigned int unusedgplcrcindex;
1830#endif
1771 unsigned int markersindex; 1831 unsigned int markersindex;
1772 unsigned int markersstringsindex; 1832 unsigned int markersstringsindex;
1773 struct module *mod; 1833 struct module *mod;
@@ -1850,13 +1910,15 @@ static struct module *load_module(void __user *umod,
1850 exportindex = find_sec(hdr, sechdrs, secstrings, "__ksymtab"); 1910 exportindex = find_sec(hdr, sechdrs, secstrings, "__ksymtab");
1851 gplindex = find_sec(hdr, sechdrs, secstrings, "__ksymtab_gpl"); 1911 gplindex = find_sec(hdr, sechdrs, secstrings, "__ksymtab_gpl");
1852 gplfutureindex = find_sec(hdr, sechdrs, secstrings, "__ksymtab_gpl_future"); 1912 gplfutureindex = find_sec(hdr, sechdrs, secstrings, "__ksymtab_gpl_future");
1853 unusedindex = find_sec(hdr, sechdrs, secstrings, "__ksymtab_unused");
1854 unusedgplindex = find_sec(hdr, sechdrs, secstrings, "__ksymtab_unused_gpl");
1855 crcindex = find_sec(hdr, sechdrs, secstrings, "__kcrctab"); 1913 crcindex = find_sec(hdr, sechdrs, secstrings, "__kcrctab");
1856 gplcrcindex = find_sec(hdr, sechdrs, secstrings, "__kcrctab_gpl"); 1914 gplcrcindex = find_sec(hdr, sechdrs, secstrings, "__kcrctab_gpl");
1857 gplfuturecrcindex = find_sec(hdr, sechdrs, secstrings, "__kcrctab_gpl_future"); 1915 gplfuturecrcindex = find_sec(hdr, sechdrs, secstrings, "__kcrctab_gpl_future");
1916#ifdef CONFIG_UNUSED_SYMBOLS
1917 unusedindex = find_sec(hdr, sechdrs, secstrings, "__ksymtab_unused");
1918 unusedgplindex = find_sec(hdr, sechdrs, secstrings, "__ksymtab_unused_gpl");
1858 unusedcrcindex = find_sec(hdr, sechdrs, secstrings, "__kcrctab_unused"); 1919 unusedcrcindex = find_sec(hdr, sechdrs, secstrings, "__kcrctab_unused");
1859 unusedgplcrcindex = find_sec(hdr, sechdrs, secstrings, "__kcrctab_unused_gpl"); 1920 unusedgplcrcindex = find_sec(hdr, sechdrs, secstrings, "__kcrctab_unused_gpl");
1921#endif
1860 setupindex = find_sec(hdr, sechdrs, secstrings, "__param"); 1922 setupindex = find_sec(hdr, sechdrs, secstrings, "__param");
1861 exindex = find_sec(hdr, sechdrs, secstrings, "__ex_table"); 1923 exindex = find_sec(hdr, sechdrs, secstrings, "__ex_table");
1862 obsparmindex = find_sec(hdr, sechdrs, secstrings, "__obsparm"); 1924 obsparmindex = find_sec(hdr, sechdrs, secstrings, "__obsparm");
@@ -1935,7 +1997,7 @@ static struct module *load_module(void __user *umod,
1935 layout_sections(mod, hdr, sechdrs, secstrings); 1997 layout_sections(mod, hdr, sechdrs, secstrings);
1936 1998
1937 /* Do the allocs. */ 1999 /* Do the allocs. */
1938 ptr = module_alloc(mod->core_size); 2000 ptr = module_alloc_update_bounds(mod->core_size);
1939 if (!ptr) { 2001 if (!ptr) {
1940 err = -ENOMEM; 2002 err = -ENOMEM;
1941 goto free_percpu; 2003 goto free_percpu;
@@ -1943,7 +2005,7 @@ static struct module *load_module(void __user *umod,
1943 memset(ptr, 0, mod->core_size); 2005 memset(ptr, 0, mod->core_size);
1944 mod->module_core = ptr; 2006 mod->module_core = ptr;
1945 2007
1946 ptr = module_alloc(mod->init_size); 2008 ptr = module_alloc_update_bounds(mod->init_size);
1947 if (!ptr && mod->init_size) { 2009 if (!ptr && mod->init_size) {
1948 err = -ENOMEM; 2010 err = -ENOMEM;
1949 goto free_core; 2011 goto free_core;
@@ -2018,14 +2080,15 @@ static struct module *load_module(void __user *umod,
2018 mod->gpl_crcs = (void *)sechdrs[gplcrcindex].sh_addr; 2080 mod->gpl_crcs = (void *)sechdrs[gplcrcindex].sh_addr;
2019 mod->num_gpl_future_syms = sechdrs[gplfutureindex].sh_size / 2081 mod->num_gpl_future_syms = sechdrs[gplfutureindex].sh_size /
2020 sizeof(*mod->gpl_future_syms); 2082 sizeof(*mod->gpl_future_syms);
2021 mod->num_unused_syms = sechdrs[unusedindex].sh_size /
2022 sizeof(*mod->unused_syms);
2023 mod->num_unused_gpl_syms = sechdrs[unusedgplindex].sh_size /
2024 sizeof(*mod->unused_gpl_syms);
2025 mod->gpl_future_syms = (void *)sechdrs[gplfutureindex].sh_addr; 2083 mod->gpl_future_syms = (void *)sechdrs[gplfutureindex].sh_addr;
2026 if (gplfuturecrcindex) 2084 if (gplfuturecrcindex)
2027 mod->gpl_future_crcs = (void *)sechdrs[gplfuturecrcindex].sh_addr; 2085 mod->gpl_future_crcs = (void *)sechdrs[gplfuturecrcindex].sh_addr;
2028 2086
2087#ifdef CONFIG_UNUSED_SYMBOLS
2088 mod->num_unused_syms = sechdrs[unusedindex].sh_size /
2089 sizeof(*mod->unused_syms);
2090 mod->num_unused_gpl_syms = sechdrs[unusedgplindex].sh_size /
2091 sizeof(*mod->unused_gpl_syms);
2029 mod->unused_syms = (void *)sechdrs[unusedindex].sh_addr; 2092 mod->unused_syms = (void *)sechdrs[unusedindex].sh_addr;
2030 if (unusedcrcindex) 2093 if (unusedcrcindex)
2031 mod->unused_crcs = (void *)sechdrs[unusedcrcindex].sh_addr; 2094 mod->unused_crcs = (void *)sechdrs[unusedcrcindex].sh_addr;
@@ -2033,13 +2096,17 @@ static struct module *load_module(void __user *umod,
2033 if (unusedgplcrcindex) 2096 if (unusedgplcrcindex)
2034 mod->unused_gpl_crcs 2097 mod->unused_gpl_crcs
2035 = (void *)sechdrs[unusedgplcrcindex].sh_addr; 2098 = (void *)sechdrs[unusedgplcrcindex].sh_addr;
2099#endif
2036 2100
2037#ifdef CONFIG_MODVERSIONS 2101#ifdef CONFIG_MODVERSIONS
2038 if ((mod->num_syms && !crcindex) || 2102 if ((mod->num_syms && !crcindex)
2039 (mod->num_gpl_syms && !gplcrcindex) || 2103 || (mod->num_gpl_syms && !gplcrcindex)
2040 (mod->num_gpl_future_syms && !gplfuturecrcindex) || 2104 || (mod->num_gpl_future_syms && !gplfuturecrcindex)
2041 (mod->num_unused_syms && !unusedcrcindex) || 2105#ifdef CONFIG_UNUSED_SYMBOLS
2042 (mod->num_unused_gpl_syms && !unusedgplcrcindex)) { 2106 || (mod->num_unused_syms && !unusedcrcindex)
2107 || (mod->num_unused_gpl_syms && !unusedgplcrcindex)
2108#endif
2109 ) {
2043 printk(KERN_WARNING "%s: No versions for exported symbols.\n", mod->name); 2110 printk(KERN_WARNING "%s: No versions for exported symbols.\n", mod->name);
2044 err = try_to_force_load(mod, "nocrc"); 2111 err = try_to_force_load(mod, "nocrc");
2045 if (err) 2112 if (err)
@@ -2512,7 +2579,7 @@ static int m_show(struct seq_file *m, void *p)
2512 struct module *mod = list_entry(p, struct module, list); 2579 struct module *mod = list_entry(p, struct module, list);
2513 char buf[8]; 2580 char buf[8];
2514 2581
2515 seq_printf(m, "%s %lu", 2582 seq_printf(m, "%s %u",
2516 mod->name, mod->init_size + mod->core_size); 2583 mod->name, mod->init_size + mod->core_size);
2517 print_unload_info(m, mod); 2584 print_unload_info(m, mod);
2518 2585
@@ -2595,6 +2662,9 @@ struct module *__module_text_address(unsigned long addr)
2595{ 2662{
2596 struct module *mod; 2663 struct module *mod;
2597 2664
2665 if (addr < module_addr_min || addr > module_addr_max)
2666 return NULL;
2667
2598 list_for_each_entry(mod, &modules, list) 2668 list_for_each_entry(mod, &modules, list)
2599 if (within(addr, mod->module_init, mod->init_text_size) 2669 if (within(addr, mod->module_init, mod->init_text_size)
2600 || within(addr, mod->module_core, mod->core_text_size)) 2670 || within(addr, mod->module_core, mod->core_text_size))
diff --git a/kernel/ns_cgroup.c b/kernel/ns_cgroup.c
index 48d7ed6fc3a4..43c2111cd54d 100644
--- a/kernel/ns_cgroup.c
+++ b/kernel/ns_cgroup.c
@@ -7,6 +7,7 @@
7#include <linux/module.h> 7#include <linux/module.h>
8#include <linux/cgroup.h> 8#include <linux/cgroup.h>
9#include <linux/fs.h> 9#include <linux/fs.h>
10#include <linux/proc_fs.h>
10#include <linux/slab.h> 11#include <linux/slab.h>
11#include <linux/nsproxy.h> 12#include <linux/nsproxy.h>
12 13
@@ -24,9 +25,12 @@ static inline struct ns_cgroup *cgroup_to_ns(
24 struct ns_cgroup, css); 25 struct ns_cgroup, css);
25} 26}
26 27
27int ns_cgroup_clone(struct task_struct *task) 28int ns_cgroup_clone(struct task_struct *task, struct pid *pid)
28{ 29{
29 return cgroup_clone(task, &ns_subsys); 30 char name[PROC_NUMBUF];
31
32 snprintf(name, PROC_NUMBUF, "%d", pid_vnr(pid));
33 return cgroup_clone(task, &ns_subsys, name);
30} 34}
31 35
32/* 36/*
diff --git a/kernel/nsproxy.c b/kernel/nsproxy.c
index adc785146a1c..21575fc46d05 100644
--- a/kernel/nsproxy.c
+++ b/kernel/nsproxy.c
@@ -157,12 +157,6 @@ int copy_namespaces(unsigned long flags, struct task_struct *tsk)
157 goto out; 157 goto out;
158 } 158 }
159 159
160 err = ns_cgroup_clone(tsk);
161 if (err) {
162 put_nsproxy(new_ns);
163 goto out;
164 }
165
166 tsk->nsproxy = new_ns; 160 tsk->nsproxy = new_ns;
167 161
168out: 162out:
@@ -209,7 +203,7 @@ int unshare_nsproxy_namespaces(unsigned long unshare_flags,
209 goto out; 203 goto out;
210 } 204 }
211 205
212 err = ns_cgroup_clone(current); 206 err = ns_cgroup_clone(current, task_pid(current));
213 if (err) 207 if (err)
214 put_nsproxy(*new_nsp); 208 put_nsproxy(*new_nsp);
215 209
diff --git a/kernel/panic.c b/kernel/panic.c
index 425567f45b9f..12c5a0a6c89b 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -318,6 +318,28 @@ void warn_on_slowpath(const char *file, int line)
318 add_taint(TAINT_WARN); 318 add_taint(TAINT_WARN);
319} 319}
320EXPORT_SYMBOL(warn_on_slowpath); 320EXPORT_SYMBOL(warn_on_slowpath);
321
322
323void warn_slowpath(const char *file, int line, const char *fmt, ...)
324{
325 va_list args;
326 char function[KSYM_SYMBOL_LEN];
327 unsigned long caller = (unsigned long)__builtin_return_address(0);
328 sprint_symbol(function, caller);
329
330 printk(KERN_WARNING "------------[ cut here ]------------\n");
331 printk(KERN_WARNING "WARNING: at %s:%d %s()\n", file,
332 line, function);
333 va_start(args, fmt);
334 vprintk(fmt, args);
335 va_end(args);
336
337 print_modules();
338 dump_stack();
339 print_oops_end_marker();
340 add_taint(TAINT_WARN);
341}
342EXPORT_SYMBOL(warn_slowpath);
321#endif 343#endif
322 344
323#ifdef CONFIG_CC_STACKPROTECTOR 345#ifdef CONFIG_CC_STACKPROTECTOR
diff --git a/kernel/pid.c b/kernel/pid.c
index 30bd5d4b2ac7..064e76afa507 100644
--- a/kernel/pid.c
+++ b/kernel/pid.c
@@ -309,12 +309,6 @@ struct pid *find_vpid(int nr)
309} 309}
310EXPORT_SYMBOL_GPL(find_vpid); 310EXPORT_SYMBOL_GPL(find_vpid);
311 311
312struct pid *find_pid(int nr)
313{
314 return find_pid_ns(nr, &init_pid_ns);
315}
316EXPORT_SYMBOL_GPL(find_pid);
317
318/* 312/*
319 * attach_pid() must be called with the tasklist_lock write-held. 313 * attach_pid() must be called with the tasklist_lock write-held.
320 */ 314 */
@@ -435,6 +429,7 @@ struct pid *find_get_pid(pid_t nr)
435 429
436 return pid; 430 return pid;
437} 431}
432EXPORT_SYMBOL_GPL(find_get_pid);
438 433
439pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns) 434pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns)
440{ 435{
@@ -482,7 +477,7 @@ EXPORT_SYMBOL(task_session_nr_ns);
482/* 477/*
483 * Used by proc to find the first pid that is greater then or equal to nr. 478 * Used by proc to find the first pid that is greater then or equal to nr.
484 * 479 *
485 * If there is a pid at nr this function is exactly the same as find_pid. 480 * If there is a pid at nr this function is exactly the same as find_pid_ns.
486 */ 481 */
487struct pid *find_ge_pid(int nr, struct pid_namespace *ns) 482struct pid *find_ge_pid(int nr, struct pid_namespace *ns)
488{ 483{
@@ -497,7 +492,6 @@ struct pid *find_ge_pid(int nr, struct pid_namespace *ns)
497 492
498 return pid; 493 return pid;
499} 494}
500EXPORT_SYMBOL_GPL(find_get_pid);
501 495
502/* 496/*
503 * The pid hash table is scaled according to the amount of memory in the 497 * The pid hash table is scaled according to the amount of memory in the
diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c
index 98702b4b8851..ea567b78d1aa 100644
--- a/kernel/pid_namespace.c
+++ b/kernel/pid_namespace.c
@@ -12,6 +12,7 @@
12#include <linux/pid_namespace.h> 12#include <linux/pid_namespace.h>
13#include <linux/syscalls.h> 13#include <linux/syscalls.h>
14#include <linux/err.h> 14#include <linux/err.h>
15#include <linux/acct.h>
15 16
16#define BITS_PER_PAGE (PAGE_SIZE*8) 17#define BITS_PER_PAGE (PAGE_SIZE*8)
17 18
@@ -71,7 +72,7 @@ static struct pid_namespace *create_pid_namespace(unsigned int level)
71 struct pid_namespace *ns; 72 struct pid_namespace *ns;
72 int i; 73 int i;
73 74
74 ns = kmem_cache_alloc(pid_ns_cachep, GFP_KERNEL); 75 ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL);
75 if (ns == NULL) 76 if (ns == NULL)
76 goto out; 77 goto out;
77 78
@@ -84,17 +85,13 @@ static struct pid_namespace *create_pid_namespace(unsigned int level)
84 goto out_free_map; 85 goto out_free_map;
85 86
86 kref_init(&ns->kref); 87 kref_init(&ns->kref);
87 ns->last_pid = 0;
88 ns->child_reaper = NULL;
89 ns->level = level; 88 ns->level = level;
90 89
91 set_bit(0, ns->pidmap[0].page); 90 set_bit(0, ns->pidmap[0].page);
92 atomic_set(&ns->pidmap[0].nr_free, BITS_PER_PAGE - 1); 91 atomic_set(&ns->pidmap[0].nr_free, BITS_PER_PAGE - 1);
93 92
94 for (i = 1; i < PIDMAP_ENTRIES; i++) { 93 for (i = 1; i < PIDMAP_ENTRIES; i++)
95 ns->pidmap[i].page = NULL;
96 atomic_set(&ns->pidmap[i].nr_free, BITS_PER_PAGE); 94 atomic_set(&ns->pidmap[i].nr_free, BITS_PER_PAGE);
97 }
98 95
99 return ns; 96 return ns;
100 97
@@ -185,6 +182,7 @@ void zap_pid_ns_processes(struct pid_namespace *pid_ns)
185 182
186 /* Child reaper for the pid namespace is going away */ 183 /* Child reaper for the pid namespace is going away */
187 pid_ns->child_reaper = NULL; 184 pid_ns->child_reaper = NULL;
185 acct_exit_ns(pid_ns);
188 return; 186 return;
189} 187}
190 188
diff --git a/kernel/posix-timers.c b/kernel/posix-timers.c
index dbd8398ddb0b..9a21681aa80f 100644
--- a/kernel/posix-timers.c
+++ b/kernel/posix-timers.c
@@ -449,9 +449,6 @@ static void release_posix_timer(struct k_itimer *tmr, int it_id_set)
449 spin_unlock_irqrestore(&idr_lock, flags); 449 spin_unlock_irqrestore(&idr_lock, flags);
450 } 450 }
451 sigqueue_free(tmr->sigq); 451 sigqueue_free(tmr->sigq);
452 if (unlikely(tmr->it_process) &&
453 tmr->it_sigev_notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID))
454 put_task_struct(tmr->it_process);
455 kmem_cache_free(posix_timers_cache, tmr); 452 kmem_cache_free(posix_timers_cache, tmr);
456} 453}
457 454
@@ -856,11 +853,10 @@ retry_delete:
856 * This keeps any tasks waiting on the spin lock from thinking 853 * This keeps any tasks waiting on the spin lock from thinking
857 * they got something (see the lock code above). 854 * they got something (see the lock code above).
858 */ 855 */
859 if (timer->it_process) { 856 if (timer->it_sigev_notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID))
860 if (timer->it_sigev_notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID)) 857 put_task_struct(timer->it_process);
861 put_task_struct(timer->it_process); 858 timer->it_process = NULL;
862 timer->it_process = NULL; 859
863 }
864 unlock_timer(timer, flags); 860 unlock_timer(timer, flags);
865 release_posix_timer(timer, IT_ID_SET); 861 release_posix_timer(timer, IT_ID_SET);
866 return 0; 862 return 0;
@@ -885,11 +881,10 @@ retry_delete:
885 * This keeps any tasks waiting on the spin lock from thinking 881 * This keeps any tasks waiting on the spin lock from thinking
886 * they got something (see the lock code above). 882 * they got something (see the lock code above).
887 */ 883 */
888 if (timer->it_process) { 884 if (timer->it_sigev_notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID))
889 if (timer->it_sigev_notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID)) 885 put_task_struct(timer->it_process);
890 put_task_struct(timer->it_process); 886 timer->it_process = NULL;
891 timer->it_process = NULL; 887
892 }
893 unlock_timer(timer, flags); 888 unlock_timer(timer, flags);
894 release_posix_timer(timer, IT_ID_SET); 889 release_posix_timer(timer, IT_ID_SET);
895} 890}
diff --git a/kernel/power/Kconfig b/kernel/power/Kconfig
index 59dfdf1e1d20..dcd165f92a88 100644
--- a/kernel/power/Kconfig
+++ b/kernel/power/Kconfig
@@ -94,6 +94,17 @@ config SUSPEND
94 powered and thus its contents are preserved, such as the 94 powered and thus its contents are preserved, such as the
95 suspend-to-RAM state (e.g. the ACPI S3 state). 95 suspend-to-RAM state (e.g. the ACPI S3 state).
96 96
97config PM_TEST_SUSPEND
98 bool "Test suspend/resume and wakealarm during bootup"
99 depends on SUSPEND && PM_DEBUG && RTC_LIB=y
100 ---help---
101 This option will let you suspend your machine during bootup, and
102 make it wake up a few seconds later using an RTC wakeup alarm.
103 Enable this with a kernel parameter like "test_suspend=mem".
104
105 You probably want to have your system's RTC driver statically
106 linked, ensuring that it's available when this test runs.
107
97config SUSPEND_FREEZER 108config SUSPEND_FREEZER
98 bool "Enable freezer for suspend to RAM/standby" \ 109 bool "Enable freezer for suspend to RAM/standby" \
99 if ARCH_WANTS_FREEZER_CONTROL || BROKEN 110 if ARCH_WANTS_FREEZER_CONTROL || BROKEN
diff --git a/kernel/power/main.c b/kernel/power/main.c
index 3398f4651aa1..95bff23ecdaa 100644
--- a/kernel/power/main.c
+++ b/kernel/power/main.c
@@ -132,6 +132,61 @@ static inline int suspend_test(int level) { return 0; }
132 132
133#ifdef CONFIG_SUSPEND 133#ifdef CONFIG_SUSPEND
134 134
135#ifdef CONFIG_PM_TEST_SUSPEND
136
137/*
138 * We test the system suspend code by setting an RTC wakealarm a short
139 * time in the future, then suspending. Suspending the devices won't
140 * normally take long ... some systems only need a few milliseconds.
141 *
142 * The time it takes is system-specific though, so when we test this
143 * during system bootup we allow a LOT of time.
144 */
145#define TEST_SUSPEND_SECONDS 5
146
147static unsigned long suspend_test_start_time;
148
149static void suspend_test_start(void)
150{
151 /* FIXME Use better timebase than "jiffies", ideally a clocksource.
152 * What we want is a hardware counter that will work correctly even
153 * during the irqs-are-off stages of the suspend/resume cycle...
154 */
155 suspend_test_start_time = jiffies;
156}
157
158static void suspend_test_finish(const char *label)
159{
160 long nj = jiffies - suspend_test_start_time;
161 unsigned msec;
162
163 msec = jiffies_to_msecs(abs(nj));
164 pr_info("PM: %s took %d.%03d seconds\n", label,
165 msec / 1000, msec % 1000);
166
167 /* Warning on suspend means the RTC alarm period needs to be
168 * larger -- the system was sooo slooowwww to suspend that the
169 * alarm (should have) fired before the system went to sleep!
170 *
171 * Warning on either suspend or resume also means the system
172 * has some performance issues. The stack dump of a WARN_ON
173 * is more likely to get the right attention than a printk...
174 */
175 WARN_ON(msec > (TEST_SUSPEND_SECONDS * 1000));
176}
177
178#else
179
180static void suspend_test_start(void)
181{
182}
183
184static void suspend_test_finish(const char *label)
185{
186}
187
188#endif
189
135/* This is just an arbitrary number */ 190/* This is just an arbitrary number */
136#define FREE_PAGE_NUMBER (100) 191#define FREE_PAGE_NUMBER (100)
137 192
@@ -266,12 +321,13 @@ int suspend_devices_and_enter(suspend_state_t state)
266 goto Close; 321 goto Close;
267 } 322 }
268 suspend_console(); 323 suspend_console();
324 suspend_test_start();
269 error = device_suspend(PMSG_SUSPEND); 325 error = device_suspend(PMSG_SUSPEND);
270 if (error) { 326 if (error) {
271 printk(KERN_ERR "PM: Some devices failed to suspend\n"); 327 printk(KERN_ERR "PM: Some devices failed to suspend\n");
272 goto Recover_platform; 328 goto Recover_platform;
273 } 329 }
274 330 suspend_test_finish("suspend devices");
275 if (suspend_test(TEST_DEVICES)) 331 if (suspend_test(TEST_DEVICES))
276 goto Recover_platform; 332 goto Recover_platform;
277 333
@@ -293,7 +349,9 @@ int suspend_devices_and_enter(suspend_state_t state)
293 if (suspend_ops->finish) 349 if (suspend_ops->finish)
294 suspend_ops->finish(); 350 suspend_ops->finish();
295 Resume_devices: 351 Resume_devices:
352 suspend_test_start();
296 device_resume(PMSG_RESUME); 353 device_resume(PMSG_RESUME);
354 suspend_test_finish("resume devices");
297 resume_console(); 355 resume_console();
298 Close: 356 Close:
299 if (suspend_ops->end) 357 if (suspend_ops->end)
@@ -521,3 +579,137 @@ static int __init pm_init(void)
521} 579}
522 580
523core_initcall(pm_init); 581core_initcall(pm_init);
582
583
584#ifdef CONFIG_PM_TEST_SUSPEND
585
586#include <linux/rtc.h>
587
588/*
589 * To test system suspend, we need a hands-off mechanism to resume the
590 * system. RTCs wake alarms are a common self-contained mechanism.
591 */
592
593static void __init test_wakealarm(struct rtc_device *rtc, suspend_state_t state)
594{
595 static char err_readtime[] __initdata =
596 KERN_ERR "PM: can't read %s time, err %d\n";
597 static char err_wakealarm [] __initdata =
598 KERN_ERR "PM: can't set %s wakealarm, err %d\n";
599 static char err_suspend[] __initdata =
600 KERN_ERR "PM: suspend test failed, error %d\n";
601 static char info_test[] __initdata =
602 KERN_INFO "PM: test RTC wakeup from '%s' suspend\n";
603
604 unsigned long now;
605 struct rtc_wkalrm alm;
606 int status;
607
608 /* this may fail if the RTC hasn't been initialized */
609 status = rtc_read_time(rtc, &alm.time);
610 if (status < 0) {
611 printk(err_readtime, rtc->dev.bus_id, status);
612 return;
613 }
614 rtc_tm_to_time(&alm.time, &now);
615
616 memset(&alm, 0, sizeof alm);
617 rtc_time_to_tm(now + TEST_SUSPEND_SECONDS, &alm.time);
618 alm.enabled = true;
619
620 status = rtc_set_alarm(rtc, &alm);
621 if (status < 0) {
622 printk(err_wakealarm, rtc->dev.bus_id, status);
623 return;
624 }
625
626 if (state == PM_SUSPEND_MEM) {
627 printk(info_test, pm_states[state]);
628 status = pm_suspend(state);
629 if (status == -ENODEV)
630 state = PM_SUSPEND_STANDBY;
631 }
632 if (state == PM_SUSPEND_STANDBY) {
633 printk(info_test, pm_states[state]);
634 status = pm_suspend(state);
635 }
636 if (status < 0)
637 printk(err_suspend, status);
638}
639
640static int __init has_wakealarm(struct device *dev, void *name_ptr)
641{
642 struct rtc_device *candidate = to_rtc_device(dev);
643
644 if (!candidate->ops->set_alarm)
645 return 0;
646 if (!device_may_wakeup(candidate->dev.parent))
647 return 0;
648
649 *(char **)name_ptr = dev->bus_id;
650 return 1;
651}
652
653/*
654 * Kernel options like "test_suspend=mem" force suspend/resume sanity tests
655 * at startup time. They're normally disabled, for faster boot and because
656 * we can't know which states really work on this particular system.
657 */
658static suspend_state_t test_state __initdata = PM_SUSPEND_ON;
659
660static char warn_bad_state[] __initdata =
661 KERN_WARNING "PM: can't test '%s' suspend state\n";
662
663static int __init setup_test_suspend(char *value)
664{
665 unsigned i;
666
667 /* "=mem" ==> "mem" */
668 value++;
669 for (i = 0; i < PM_SUSPEND_MAX; i++) {
670 if (!pm_states[i])
671 continue;
672 if (strcmp(pm_states[i], value) != 0)
673 continue;
674 test_state = (__force suspend_state_t) i;
675 return 0;
676 }
677 printk(warn_bad_state, value);
678 return 0;
679}
680__setup("test_suspend", setup_test_suspend);
681
682static int __init test_suspend(void)
683{
684 static char warn_no_rtc[] __initdata =
685 KERN_WARNING "PM: no wakealarm-capable RTC driver is ready\n";
686
687 char *pony = NULL;
688 struct rtc_device *rtc = NULL;
689
690 /* PM is initialized by now; is that state testable? */
691 if (test_state == PM_SUSPEND_ON)
692 goto done;
693 if (!valid_state(test_state)) {
694 printk(warn_bad_state, pm_states[test_state]);
695 goto done;
696 }
697
698 /* RTCs have initialized by now too ... can we use one? */
699 class_find_device(rtc_class, NULL, &pony, has_wakealarm);
700 if (pony)
701 rtc = rtc_class_open(pony);
702 if (!rtc) {
703 printk(warn_no_rtc);
704 goto done;
705 }
706
707 /* go for it */
708 test_wakealarm(rtc, test_state);
709 rtc_class_close(rtc);
710done:
711 return 0;
712}
713late_initcall(test_suspend);
714
715#endif /* CONFIG_PM_TEST_SUSPEND */
diff --git a/kernel/power/poweroff.c b/kernel/power/poweroff.c
index 678ec736076b..72016f051477 100644
--- a/kernel/power/poweroff.c
+++ b/kernel/power/poweroff.c
@@ -10,6 +10,7 @@
10#include <linux/pm.h> 10#include <linux/pm.h>
11#include <linux/workqueue.h> 11#include <linux/workqueue.h>
12#include <linux/reboot.h> 12#include <linux/reboot.h>
13#include <linux/cpumask.h>
13 14
14/* 15/*
15 * When the user hits Sys-Rq o to power down the machine this is the 16 * When the user hits Sys-Rq o to power down the machine this is the
@@ -25,7 +26,8 @@ static DECLARE_WORK(poweroff_work, do_poweroff);
25 26
26static void handle_poweroff(int key, struct tty_struct *tty) 27static void handle_poweroff(int key, struct tty_struct *tty)
27{ 28{
28 schedule_work(&poweroff_work); 29 /* run sysrq poweroff on boot cpu */
30 schedule_work_on(first_cpu(cpu_online_map), &poweroff_work);
29} 31}
30 32
31static struct sysrq_key_op sysrq_poweroff_op = { 33static struct sysrq_key_op sysrq_poweroff_op = {
diff --git a/kernel/power/process.c b/kernel/power/process.c
index 5fb87652f214..278946aecaf0 100644
--- a/kernel/power/process.c
+++ b/kernel/power/process.c
@@ -149,7 +149,7 @@ static int try_to_freeze_tasks(bool sig_only)
149 unsigned long end_time; 149 unsigned long end_time;
150 unsigned int todo; 150 unsigned int todo;
151 struct timeval start, end; 151 struct timeval start, end;
152 s64 elapsed_csecs64; 152 u64 elapsed_csecs64;
153 unsigned int elapsed_csecs; 153 unsigned int elapsed_csecs;
154 154
155 do_gettimeofday(&start); 155 do_gettimeofday(&start);
diff --git a/kernel/power/snapshot.c b/kernel/power/snapshot.c
index 5f91a07c4eac..5d2ab836e998 100644
--- a/kernel/power/snapshot.c
+++ b/kernel/power/snapshot.c
@@ -205,8 +205,7 @@ static void chain_free(struct chain_allocator *ca, int clear_page_nosave)
205 * objects. The main list's elements are of type struct zone_bitmap 205 * objects. The main list's elements are of type struct zone_bitmap
206 * and each of them corresonds to one zone. For each zone bitmap 206 * and each of them corresonds to one zone. For each zone bitmap
207 * object there is a list of objects of type struct bm_block that 207 * object there is a list of objects of type struct bm_block that
208 * represent each blocks of bit chunks in which information is 208 * represent each blocks of bitmap in which information is stored.
209 * stored.
210 * 209 *
211 * struct memory_bitmap contains a pointer to the main list of zone 210 * struct memory_bitmap contains a pointer to the main list of zone
212 * bitmap objects, a struct bm_position used for browsing the bitmap, 211 * bitmap objects, a struct bm_position used for browsing the bitmap,
@@ -224,26 +223,27 @@ static void chain_free(struct chain_allocator *ca, int clear_page_nosave)
224 * pfns that correspond to the start and end of the represented zone. 223 * pfns that correspond to the start and end of the represented zone.
225 * 224 *
226 * struct bm_block contains a pointer to the memory page in which 225 * struct bm_block contains a pointer to the memory page in which
227 * information is stored (in the form of a block of bit chunks 226 * information is stored (in the form of a block of bitmap)
228 * of type unsigned long each). It also contains the pfns that 227 * It also contains the pfns that correspond to the start and end of
229 * correspond to the start and end of the represented memory area and 228 * the represented memory area.
230 * the number of bit chunks in the block.
231 */ 229 */
232 230
233#define BM_END_OF_MAP (~0UL) 231#define BM_END_OF_MAP (~0UL)
234 232
235#define BM_CHUNKS_PER_BLOCK (PAGE_SIZE / sizeof(long))
236#define BM_BITS_PER_CHUNK (sizeof(long) << 3)
237#define BM_BITS_PER_BLOCK (PAGE_SIZE << 3) 233#define BM_BITS_PER_BLOCK (PAGE_SIZE << 3)
238 234
239struct bm_block { 235struct bm_block {
240 struct bm_block *next; /* next element of the list */ 236 struct bm_block *next; /* next element of the list */
241 unsigned long start_pfn; /* pfn represented by the first bit */ 237 unsigned long start_pfn; /* pfn represented by the first bit */
242 unsigned long end_pfn; /* pfn represented by the last bit plus 1 */ 238 unsigned long end_pfn; /* pfn represented by the last bit plus 1 */
243 unsigned int size; /* number of bit chunks */ 239 unsigned long *data; /* bitmap representing pages */
244 unsigned long *data; /* chunks of bits representing pages */
245}; 240};
246 241
242static inline unsigned long bm_block_bits(struct bm_block *bb)
243{
244 return bb->end_pfn - bb->start_pfn;
245}
246
247struct zone_bitmap { 247struct zone_bitmap {
248 struct zone_bitmap *next; /* next element of the list */ 248 struct zone_bitmap *next; /* next element of the list */
249 unsigned long start_pfn; /* minimal pfn in this zone */ 249 unsigned long start_pfn; /* minimal pfn in this zone */
@@ -257,7 +257,6 @@ struct zone_bitmap {
257struct bm_position { 257struct bm_position {
258 struct zone_bitmap *zone_bm; 258 struct zone_bitmap *zone_bm;
259 struct bm_block *block; 259 struct bm_block *block;
260 int chunk;
261 int bit; 260 int bit;
262}; 261};
263 262
@@ -272,12 +271,6 @@ struct memory_bitmap {
272 271
273/* Functions that operate on memory bitmaps */ 272/* Functions that operate on memory bitmaps */
274 273
275static inline void memory_bm_reset_chunk(struct memory_bitmap *bm)
276{
277 bm->cur.chunk = 0;
278 bm->cur.bit = -1;
279}
280
281static void memory_bm_position_reset(struct memory_bitmap *bm) 274static void memory_bm_position_reset(struct memory_bitmap *bm)
282{ 275{
283 struct zone_bitmap *zone_bm; 276 struct zone_bitmap *zone_bm;
@@ -285,7 +278,7 @@ static void memory_bm_position_reset(struct memory_bitmap *bm)
285 zone_bm = bm->zone_bm_list; 278 zone_bm = bm->zone_bm_list;
286 bm->cur.zone_bm = zone_bm; 279 bm->cur.zone_bm = zone_bm;
287 bm->cur.block = zone_bm->bm_blocks; 280 bm->cur.block = zone_bm->bm_blocks;
288 memory_bm_reset_chunk(bm); 281 bm->cur.bit = 0;
289} 282}
290 283
291static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free); 284static void memory_bm_free(struct memory_bitmap *bm, int clear_nosave_free);
@@ -394,12 +387,10 @@ memory_bm_create(struct memory_bitmap *bm, gfp_t gfp_mask, int safe_needed)
394 bb->start_pfn = pfn; 387 bb->start_pfn = pfn;
395 if (nr >= BM_BITS_PER_BLOCK) { 388 if (nr >= BM_BITS_PER_BLOCK) {
396 pfn += BM_BITS_PER_BLOCK; 389 pfn += BM_BITS_PER_BLOCK;
397 bb->size = BM_CHUNKS_PER_BLOCK;
398 nr -= BM_BITS_PER_BLOCK; 390 nr -= BM_BITS_PER_BLOCK;
399 } else { 391 } else {
400 /* This is executed only once in the loop */ 392 /* This is executed only once in the loop */
401 pfn += nr; 393 pfn += nr;
402 bb->size = DIV_ROUND_UP(nr, BM_BITS_PER_CHUNK);
403 } 394 }
404 bb->end_pfn = pfn; 395 bb->end_pfn = pfn;
405 bb = bb->next; 396 bb = bb->next;
@@ -478,8 +469,8 @@ static int memory_bm_find_bit(struct memory_bitmap *bm, unsigned long pfn,
478 } 469 }
479 zone_bm->cur_block = bb; 470 zone_bm->cur_block = bb;
480 pfn -= bb->start_pfn; 471 pfn -= bb->start_pfn;
481 *bit_nr = pfn % BM_BITS_PER_CHUNK; 472 *bit_nr = pfn;
482 *addr = bb->data + pfn / BM_BITS_PER_CHUNK; 473 *addr = bb->data;
483 return 0; 474 return 0;
484} 475}
485 476
@@ -528,36 +519,6 @@ static int memory_bm_test_bit(struct memory_bitmap *bm, unsigned long pfn)
528 return test_bit(bit, addr); 519 return test_bit(bit, addr);
529} 520}
530 521
531/* Two auxiliary functions for memory_bm_next_pfn */
532
533/* Find the first set bit in the given chunk, if there is one */
534
535static inline int next_bit_in_chunk(int bit, unsigned long *chunk_p)
536{
537 bit++;
538 while (bit < BM_BITS_PER_CHUNK) {
539 if (test_bit(bit, chunk_p))
540 return bit;
541
542 bit++;
543 }
544 return -1;
545}
546
547/* Find a chunk containing some bits set in given block of bits */
548
549static inline int next_chunk_in_block(int n, struct bm_block *bb)
550{
551 n++;
552 while (n < bb->size) {
553 if (bb->data[n])
554 return n;
555
556 n++;
557 }
558 return -1;
559}
560
561/** 522/**
562 * memory_bm_next_pfn - find the pfn that corresponds to the next set bit 523 * memory_bm_next_pfn - find the pfn that corresponds to the next set bit
563 * in the bitmap @bm. If the pfn cannot be found, BM_END_OF_MAP is 524 * in the bitmap @bm. If the pfn cannot be found, BM_END_OF_MAP is
@@ -571,40 +532,33 @@ static unsigned long memory_bm_next_pfn(struct memory_bitmap *bm)
571{ 532{
572 struct zone_bitmap *zone_bm; 533 struct zone_bitmap *zone_bm;
573 struct bm_block *bb; 534 struct bm_block *bb;
574 int chunk;
575 int bit; 535 int bit;
576 536
577 do { 537 do {
578 bb = bm->cur.block; 538 bb = bm->cur.block;
579 do { 539 do {
580 chunk = bm->cur.chunk;
581 bit = bm->cur.bit; 540 bit = bm->cur.bit;
582 do { 541 bit = find_next_bit(bb->data, bm_block_bits(bb), bit);
583 bit = next_bit_in_chunk(bit, bb->data + chunk); 542 if (bit < bm_block_bits(bb))
584 if (bit >= 0) 543 goto Return_pfn;
585 goto Return_pfn; 544
586
587 chunk = next_chunk_in_block(chunk, bb);
588 bit = -1;
589 } while (chunk >= 0);
590 bb = bb->next; 545 bb = bb->next;
591 bm->cur.block = bb; 546 bm->cur.block = bb;
592 memory_bm_reset_chunk(bm); 547 bm->cur.bit = 0;
593 } while (bb); 548 } while (bb);
594 zone_bm = bm->cur.zone_bm->next; 549 zone_bm = bm->cur.zone_bm->next;
595 if (zone_bm) { 550 if (zone_bm) {
596 bm->cur.zone_bm = zone_bm; 551 bm->cur.zone_bm = zone_bm;
597 bm->cur.block = zone_bm->bm_blocks; 552 bm->cur.block = zone_bm->bm_blocks;
598 memory_bm_reset_chunk(bm); 553 bm->cur.bit = 0;
599 } 554 }
600 } while (zone_bm); 555 } while (zone_bm);
601 memory_bm_position_reset(bm); 556 memory_bm_position_reset(bm);
602 return BM_END_OF_MAP; 557 return BM_END_OF_MAP;
603 558
604 Return_pfn: 559 Return_pfn:
605 bm->cur.chunk = chunk; 560 bm->cur.bit = bit + 1;
606 bm->cur.bit = bit; 561 return bb->start_pfn + bit;
607 return bb->start_pfn + chunk * BM_BITS_PER_CHUNK + bit;
608} 562}
609 563
610/** 564/**
diff --git a/kernel/printk.c b/kernel/printk.c
index 07ad9e7f7a66..a7f7559c5f6c 100644
--- a/kernel/printk.c
+++ b/kernel/printk.c
@@ -933,7 +933,7 @@ void suspend_console(void)
933{ 933{
934 if (!console_suspend_enabled) 934 if (!console_suspend_enabled)
935 return; 935 return;
936 printk("Suspending console(s)\n"); 936 printk("Suspending console(s) (use no_console_suspend to debug)\n");
937 acquire_console_sem(); 937 acquire_console_sem();
938 console_suspended = 1; 938 console_suspended = 1;
939} 939}
@@ -1308,6 +1308,8 @@ void tty_write_message(struct tty_struct *tty, char *msg)
1308} 1308}
1309 1309
1310#if defined CONFIG_PRINTK 1310#if defined CONFIG_PRINTK
1311
1312DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
1311/* 1313/*
1312 * printk rate limiting, lifted from the networking subsystem. 1314 * printk rate limiting, lifted from the networking subsystem.
1313 * 1315 *
@@ -1315,22 +1317,9 @@ void tty_write_message(struct tty_struct *tty, char *msg)
1315 * every printk_ratelimit_jiffies to make a denial-of-service 1317 * every printk_ratelimit_jiffies to make a denial-of-service
1316 * attack impossible. 1318 * attack impossible.
1317 */ 1319 */
1318int __printk_ratelimit(int ratelimit_jiffies, int ratelimit_burst)
1319{
1320 return __ratelimit(ratelimit_jiffies, ratelimit_burst);
1321}
1322EXPORT_SYMBOL(__printk_ratelimit);
1323
1324/* minimum time in jiffies between messages */
1325int printk_ratelimit_jiffies = 5 * HZ;
1326
1327/* number of messages we send before ratelimiting */
1328int printk_ratelimit_burst = 10;
1329
1330int printk_ratelimit(void) 1320int printk_ratelimit(void)
1331{ 1321{
1332 return __printk_ratelimit(printk_ratelimit_jiffies, 1322 return __ratelimit(&printk_ratelimit_state);
1333 printk_ratelimit_burst);
1334} 1323}
1335EXPORT_SYMBOL(printk_ratelimit); 1324EXPORT_SYMBOL(printk_ratelimit);
1336 1325
diff --git a/kernel/profile.c b/kernel/profile.c
index 58926411eb2a..cd26bed4cc26 100644
--- a/kernel/profile.c
+++ b/kernel/profile.c
@@ -112,8 +112,6 @@ void __init profile_init(void)
112 112
113/* Profile event notifications */ 113/* Profile event notifications */
114 114
115#ifdef CONFIG_PROFILING
116
117static BLOCKING_NOTIFIER_HEAD(task_exit_notifier); 115static BLOCKING_NOTIFIER_HEAD(task_exit_notifier);
118static ATOMIC_NOTIFIER_HEAD(task_free_notifier); 116static ATOMIC_NOTIFIER_HEAD(task_free_notifier);
119static BLOCKING_NOTIFIER_HEAD(munmap_notifier); 117static BLOCKING_NOTIFIER_HEAD(munmap_notifier);
@@ -203,8 +201,6 @@ void unregister_timer_hook(int (*hook)(struct pt_regs *))
203} 201}
204EXPORT_SYMBOL_GPL(unregister_timer_hook); 202EXPORT_SYMBOL_GPL(unregister_timer_hook);
205 203
206#endif /* CONFIG_PROFILING */
207
208 204
209#ifdef CONFIG_SMP 205#ifdef CONFIG_SMP
210/* 206/*
diff --git a/kernel/rcuclassic.c b/kernel/rcuclassic.c
index 16eeeaa9d618..6f8696c502f4 100644
--- a/kernel/rcuclassic.c
+++ b/kernel/rcuclassic.c
@@ -106,7 +106,7 @@ static void force_quiescent_state(struct rcu_data *rdp,
106 */ 106 */
107 cpus_and(cpumask, rcp->cpumask, cpu_online_map); 107 cpus_and(cpumask, rcp->cpumask, cpu_online_map);
108 cpu_clear(rdp->cpu, cpumask); 108 cpu_clear(rdp->cpu, cpumask);
109 for_each_cpu_mask(cpu, cpumask) 109 for_each_cpu_mask_nr(cpu, cpumask)
110 smp_send_reschedule(cpu); 110 smp_send_reschedule(cpu);
111 } 111 }
112} 112}
diff --git a/kernel/rcupreempt.c b/kernel/rcupreempt.c
index 6f62b77d93c4..27827931ca0d 100644
--- a/kernel/rcupreempt.c
+++ b/kernel/rcupreempt.c
@@ -756,7 +756,7 @@ rcu_try_flip_idle(void)
756 756
757 /* Now ask each CPU for acknowledgement of the flip. */ 757 /* Now ask each CPU for acknowledgement of the flip. */
758 758
759 for_each_cpu_mask(cpu, rcu_cpu_online_map) { 759 for_each_cpu_mask_nr(cpu, rcu_cpu_online_map) {
760 per_cpu(rcu_flip_flag, cpu) = rcu_flipped; 760 per_cpu(rcu_flip_flag, cpu) = rcu_flipped;
761 dyntick_save_progress_counter(cpu); 761 dyntick_save_progress_counter(cpu);
762 } 762 }
@@ -774,7 +774,7 @@ rcu_try_flip_waitack(void)
774 int cpu; 774 int cpu;
775 775
776 RCU_TRACE_ME(rcupreempt_trace_try_flip_a1); 776 RCU_TRACE_ME(rcupreempt_trace_try_flip_a1);
777 for_each_cpu_mask(cpu, rcu_cpu_online_map) 777 for_each_cpu_mask_nr(cpu, rcu_cpu_online_map)
778 if (rcu_try_flip_waitack_needed(cpu) && 778 if (rcu_try_flip_waitack_needed(cpu) &&
779 per_cpu(rcu_flip_flag, cpu) != rcu_flip_seen) { 779 per_cpu(rcu_flip_flag, cpu) != rcu_flip_seen) {
780 RCU_TRACE_ME(rcupreempt_trace_try_flip_ae1); 780 RCU_TRACE_ME(rcupreempt_trace_try_flip_ae1);
@@ -806,7 +806,7 @@ rcu_try_flip_waitzero(void)
806 /* Check to see if the sum of the "last" counters is zero. */ 806 /* Check to see if the sum of the "last" counters is zero. */
807 807
808 RCU_TRACE_ME(rcupreempt_trace_try_flip_z1); 808 RCU_TRACE_ME(rcupreempt_trace_try_flip_z1);
809 for_each_cpu_mask(cpu, rcu_cpu_online_map) 809 for_each_cpu_mask_nr(cpu, rcu_cpu_online_map)
810 sum += RCU_DATA_CPU(cpu)->rcu_flipctr[lastidx]; 810 sum += RCU_DATA_CPU(cpu)->rcu_flipctr[lastidx];
811 if (sum != 0) { 811 if (sum != 0) {
812 RCU_TRACE_ME(rcupreempt_trace_try_flip_ze1); 812 RCU_TRACE_ME(rcupreempt_trace_try_flip_ze1);
@@ -821,7 +821,7 @@ rcu_try_flip_waitzero(void)
821 smp_mb(); /* ^^^^^^^^^^^^ */ 821 smp_mb(); /* ^^^^^^^^^^^^ */
822 822
823 /* Call for a memory barrier from each CPU. */ 823 /* Call for a memory barrier from each CPU. */
824 for_each_cpu_mask(cpu, rcu_cpu_online_map) { 824 for_each_cpu_mask_nr(cpu, rcu_cpu_online_map) {
825 per_cpu(rcu_mb_flag, cpu) = rcu_mb_needed; 825 per_cpu(rcu_mb_flag, cpu) = rcu_mb_needed;
826 dyntick_save_progress_counter(cpu); 826 dyntick_save_progress_counter(cpu);
827 } 827 }
@@ -841,7 +841,7 @@ rcu_try_flip_waitmb(void)
841 int cpu; 841 int cpu;
842 842
843 RCU_TRACE_ME(rcupreempt_trace_try_flip_m1); 843 RCU_TRACE_ME(rcupreempt_trace_try_flip_m1);
844 for_each_cpu_mask(cpu, rcu_cpu_online_map) 844 for_each_cpu_mask_nr(cpu, rcu_cpu_online_map)
845 if (rcu_try_flip_waitmb_needed(cpu) && 845 if (rcu_try_flip_waitmb_needed(cpu) &&
846 per_cpu(rcu_mb_flag, cpu) != rcu_mb_done) { 846 per_cpu(rcu_mb_flag, cpu) != rcu_mb_done) {
847 RCU_TRACE_ME(rcupreempt_trace_try_flip_me1); 847 RCU_TRACE_ME(rcupreempt_trace_try_flip_me1);
diff --git a/kernel/res_counter.c b/kernel/res_counter.c
index d3c61b4ebef2..f275c8eca772 100644
--- a/kernel/res_counter.c
+++ b/kernel/res_counter.c
@@ -13,6 +13,7 @@
13#include <linux/slab.h> 13#include <linux/slab.h>
14#include <linux/res_counter.h> 14#include <linux/res_counter.h>
15#include <linux/uaccess.h> 15#include <linux/uaccess.h>
16#include <linux/mm.h>
16 17
17void res_counter_init(struct res_counter *counter) 18void res_counter_init(struct res_counter *counter)
18{ 19{
@@ -102,44 +103,37 @@ u64 res_counter_read_u64(struct res_counter *counter, int member)
102 return *res_counter_member(counter, member); 103 return *res_counter_member(counter, member);
103} 104}
104 105
105ssize_t res_counter_write(struct res_counter *counter, int member, 106int res_counter_memparse_write_strategy(const char *buf,
106 const char __user *userbuf, size_t nbytes, loff_t *pos, 107 unsigned long long *res)
107 int (*write_strategy)(char *st_buf, unsigned long long *val))
108{ 108{
109 int ret; 109 char *end;
110 char *buf, *end; 110 /* FIXME - make memparse() take const char* args */
111 unsigned long flags; 111 *res = memparse((char *)buf, &end);
112 unsigned long long tmp, *val; 112 if (*end != '\0')
113 113 return -EINVAL;
114 buf = kmalloc(nbytes + 1, GFP_KERNEL);
115 ret = -ENOMEM;
116 if (buf == NULL)
117 goto out;
118 114
119 buf[nbytes] = '\0'; 115 *res = PAGE_ALIGN(*res);
120 ret = -EFAULT; 116 return 0;
121 if (copy_from_user(buf, userbuf, nbytes)) 117}
122 goto out_free;
123 118
124 ret = -EINVAL; 119int res_counter_write(struct res_counter *counter, int member,
120 const char *buf, write_strategy_fn write_strategy)
121{
122 char *end;
123 unsigned long flags;
124 unsigned long long tmp, *val;
125 125
126 strstrip(buf);
127 if (write_strategy) { 126 if (write_strategy) {
128 if (write_strategy(buf, &tmp)) { 127 if (write_strategy(buf, &tmp))
129 goto out_free; 128 return -EINVAL;
130 }
131 } else { 129 } else {
132 tmp = simple_strtoull(buf, &end, 10); 130 tmp = simple_strtoull(buf, &end, 10);
133 if (*end != '\0') 131 if (*end != '\0')
134 goto out_free; 132 return -EINVAL;
135 } 133 }
136 spin_lock_irqsave(&counter->lock, flags); 134 spin_lock_irqsave(&counter->lock, flags);
137 val = res_counter_member(counter, member); 135 val = res_counter_member(counter, member);
138 *val = tmp; 136 *val = tmp;
139 spin_unlock_irqrestore(&counter->lock, flags); 137 spin_unlock_irqrestore(&counter->lock, flags);
140 ret = nbytes; 138 return 0;
141out_free:
142 kfree(buf);
143out:
144 return ret;
145} 139}
diff --git a/kernel/rtmutex-tester.c b/kernel/rtmutex-tester.c
index 092e4c620af9..a56f629b057a 100644
--- a/kernel/rtmutex-tester.c
+++ b/kernel/rtmutex-tester.c
@@ -297,8 +297,8 @@ static int test_func(void *data)
297 * 297 *
298 * opcode:data 298 * opcode:data
299 */ 299 */
300static ssize_t sysfs_test_command(struct sys_device *dev, const char *buf, 300static ssize_t sysfs_test_command(struct sys_device *dev, struct sysdev_attribute *attr,
301 size_t count) 301 const char *buf, size_t count)
302{ 302{
303 struct sched_param schedpar; 303 struct sched_param schedpar;
304 struct test_thread_data *td; 304 struct test_thread_data *td;
@@ -360,7 +360,8 @@ static ssize_t sysfs_test_command(struct sys_device *dev, const char *buf,
360 * @dev: thread to query 360 * @dev: thread to query
361 * @buf: char buffer to be filled with thread status info 361 * @buf: char buffer to be filled with thread status info
362 */ 362 */
363static ssize_t sysfs_test_status(struct sys_device *dev, char *buf) 363static ssize_t sysfs_test_status(struct sys_device *dev, struct sysdev_attribute *attr,
364 char *buf)
364{ 365{
365 struct test_thread_data *td; 366 struct test_thread_data *td;
366 struct task_struct *tsk; 367 struct task_struct *tsk;
diff --git a/kernel/sched.c b/kernel/sched.c
index 99e6d850ecab..0047bd9b96aa 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -571,8 +571,10 @@ struct rq {
571#endif 571#endif
572 572
573#ifdef CONFIG_SCHED_HRTICK 573#ifdef CONFIG_SCHED_HRTICK
574 unsigned long hrtick_flags; 574#ifdef CONFIG_SMP
575 ktime_t hrtick_expire; 575 int hrtick_csd_pending;
576 struct call_single_data hrtick_csd;
577#endif
576 struct hrtimer hrtick_timer; 578 struct hrtimer hrtick_timer;
577#endif 579#endif
578 580
@@ -983,13 +985,6 @@ static struct rq *this_rq_lock(void)
983 return rq; 985 return rq;
984} 986}
985 987
986static void __resched_task(struct task_struct *p, int tif_bit);
987
988static inline void resched_task(struct task_struct *p)
989{
990 __resched_task(p, TIF_NEED_RESCHED);
991}
992
993#ifdef CONFIG_SCHED_HRTICK 988#ifdef CONFIG_SCHED_HRTICK
994/* 989/*
995 * Use HR-timers to deliver accurate preemption points. 990 * Use HR-timers to deliver accurate preemption points.
@@ -1001,25 +996,6 @@ static inline void resched_task(struct task_struct *p)
1001 * When we get rescheduled we reprogram the hrtick_timer outside of the 996 * When we get rescheduled we reprogram the hrtick_timer outside of the
1002 * rq->lock. 997 * rq->lock.
1003 */ 998 */
1004static inline void resched_hrt(struct task_struct *p)
1005{
1006 __resched_task(p, TIF_HRTICK_RESCHED);
1007}
1008
1009static inline void resched_rq(struct rq *rq)
1010{
1011 unsigned long flags;
1012
1013 spin_lock_irqsave(&rq->lock, flags);
1014 resched_task(rq->curr);
1015 spin_unlock_irqrestore(&rq->lock, flags);
1016}
1017
1018enum {
1019 HRTICK_SET, /* re-programm hrtick_timer */
1020 HRTICK_RESET, /* not a new slice */
1021 HRTICK_BLOCK, /* stop hrtick operations */
1022};
1023 999
1024/* 1000/*
1025 * Use hrtick when: 1001 * Use hrtick when:
@@ -1030,40 +1006,11 @@ static inline int hrtick_enabled(struct rq *rq)
1030{ 1006{
1031 if (!sched_feat(HRTICK)) 1007 if (!sched_feat(HRTICK))
1032 return 0; 1008 return 0;
1033 if (unlikely(test_bit(HRTICK_BLOCK, &rq->hrtick_flags))) 1009 if (!cpu_active(cpu_of(rq)))
1034 return 0; 1010 return 0;
1035 return hrtimer_is_hres_active(&rq->hrtick_timer); 1011 return hrtimer_is_hres_active(&rq->hrtick_timer);
1036} 1012}
1037 1013
1038/*
1039 * Called to set the hrtick timer state.
1040 *
1041 * called with rq->lock held and irqs disabled
1042 */
1043static void hrtick_start(struct rq *rq, u64 delay, int reset)
1044{
1045 assert_spin_locked(&rq->lock);
1046
1047 /*
1048 * preempt at: now + delay
1049 */
1050 rq->hrtick_expire =
1051 ktime_add_ns(rq->hrtick_timer.base->get_time(), delay);
1052 /*
1053 * indicate we need to program the timer
1054 */
1055 __set_bit(HRTICK_SET, &rq->hrtick_flags);
1056 if (reset)
1057 __set_bit(HRTICK_RESET, &rq->hrtick_flags);
1058
1059 /*
1060 * New slices are called from the schedule path and don't need a
1061 * forced reschedule.
1062 */
1063 if (reset)
1064 resched_hrt(rq->curr);
1065}
1066
1067static void hrtick_clear(struct rq *rq) 1014static void hrtick_clear(struct rq *rq)
1068{ 1015{
1069 if (hrtimer_active(&rq->hrtick_timer)) 1016 if (hrtimer_active(&rq->hrtick_timer))
@@ -1071,32 +1018,6 @@ static void hrtick_clear(struct rq *rq)
1071} 1018}
1072 1019
1073/* 1020/*
1074 * Update the timer from the possible pending state.
1075 */
1076static void hrtick_set(struct rq *rq)
1077{
1078 ktime_t time;
1079 int set, reset;
1080 unsigned long flags;
1081
1082 WARN_ON_ONCE(cpu_of(rq) != smp_processor_id());
1083
1084 spin_lock_irqsave(&rq->lock, flags);
1085 set = __test_and_clear_bit(HRTICK_SET, &rq->hrtick_flags);
1086 reset = __test_and_clear_bit(HRTICK_RESET, &rq->hrtick_flags);
1087 time = rq->hrtick_expire;
1088 clear_thread_flag(TIF_HRTICK_RESCHED);
1089 spin_unlock_irqrestore(&rq->lock, flags);
1090
1091 if (set) {
1092 hrtimer_start(&rq->hrtick_timer, time, HRTIMER_MODE_ABS);
1093 if (reset && !hrtimer_active(&rq->hrtick_timer))
1094 resched_rq(rq);
1095 } else
1096 hrtick_clear(rq);
1097}
1098
1099/*
1100 * High-resolution timer tick. 1021 * High-resolution timer tick.
1101 * Runs from hardirq context with interrupts disabled. 1022 * Runs from hardirq context with interrupts disabled.
1102 */ 1023 */
@@ -1115,27 +1036,37 @@ static enum hrtimer_restart hrtick(struct hrtimer *timer)
1115} 1036}
1116 1037
1117#ifdef CONFIG_SMP 1038#ifdef CONFIG_SMP
1118static void hotplug_hrtick_disable(int cpu) 1039/*
1040 * called from hardirq (IPI) context
1041 */
1042static void __hrtick_start(void *arg)
1119{ 1043{
1120 struct rq *rq = cpu_rq(cpu); 1044 struct rq *rq = arg;
1121 unsigned long flags;
1122 1045
1123 spin_lock_irqsave(&rq->lock, flags); 1046 spin_lock(&rq->lock);
1124 rq->hrtick_flags = 0; 1047 hrtimer_restart(&rq->hrtick_timer);
1125 __set_bit(HRTICK_BLOCK, &rq->hrtick_flags); 1048 rq->hrtick_csd_pending = 0;
1126 spin_unlock_irqrestore(&rq->lock, flags); 1049 spin_unlock(&rq->lock);
1127
1128 hrtick_clear(rq);
1129} 1050}
1130 1051
1131static void hotplug_hrtick_enable(int cpu) 1052/*
1053 * Called to set the hrtick timer state.
1054 *
1055 * called with rq->lock held and irqs disabled
1056 */
1057static void hrtick_start(struct rq *rq, u64 delay)
1132{ 1058{
1133 struct rq *rq = cpu_rq(cpu); 1059 struct hrtimer *timer = &rq->hrtick_timer;
1134 unsigned long flags; 1060 ktime_t time = ktime_add_ns(timer->base->get_time(), delay);
1135 1061
1136 spin_lock_irqsave(&rq->lock, flags); 1062 timer->expires = time;
1137 __clear_bit(HRTICK_BLOCK, &rq->hrtick_flags); 1063
1138 spin_unlock_irqrestore(&rq->lock, flags); 1064 if (rq == this_rq()) {
1065 hrtimer_restart(timer);
1066 } else if (!rq->hrtick_csd_pending) {
1067 __smp_call_function_single(cpu_of(rq), &rq->hrtick_csd);
1068 rq->hrtick_csd_pending = 1;
1069 }
1139} 1070}
1140 1071
1141static int 1072static int
@@ -1150,16 +1081,7 @@ hotplug_hrtick(struct notifier_block *nfb, unsigned long action, void *hcpu)
1150 case CPU_DOWN_PREPARE_FROZEN: 1081 case CPU_DOWN_PREPARE_FROZEN:
1151 case CPU_DEAD: 1082 case CPU_DEAD:
1152 case CPU_DEAD_FROZEN: 1083 case CPU_DEAD_FROZEN:
1153 hotplug_hrtick_disable(cpu); 1084 hrtick_clear(cpu_rq(cpu));
1154 return NOTIFY_OK;
1155
1156 case CPU_UP_PREPARE:
1157 case CPU_UP_PREPARE_FROZEN:
1158 case CPU_DOWN_FAILED:
1159 case CPU_DOWN_FAILED_FROZEN:
1160 case CPU_ONLINE:
1161 case CPU_ONLINE_FROZEN:
1162 hotplug_hrtick_enable(cpu);
1163 return NOTIFY_OK; 1085 return NOTIFY_OK;
1164 } 1086 }
1165 1087
@@ -1170,46 +1092,45 @@ static void init_hrtick(void)
1170{ 1092{
1171 hotcpu_notifier(hotplug_hrtick, 0); 1093 hotcpu_notifier(hotplug_hrtick, 0);
1172} 1094}
1173#endif /* CONFIG_SMP */ 1095#else
1096/*
1097 * Called to set the hrtick timer state.
1098 *
1099 * called with rq->lock held and irqs disabled
1100 */
1101static void hrtick_start(struct rq *rq, u64 delay)
1102{
1103 hrtimer_start(&rq->hrtick_timer, ns_to_ktime(delay), HRTIMER_MODE_REL);
1104}
1174 1105
1175static void init_rq_hrtick(struct rq *rq) 1106static void init_hrtick(void)
1176{ 1107{
1177 rq->hrtick_flags = 0;
1178 hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
1179 rq->hrtick_timer.function = hrtick;
1180 rq->hrtick_timer.cb_mode = HRTIMER_CB_IRQSAFE_NO_SOFTIRQ;
1181} 1108}
1109#endif /* CONFIG_SMP */
1182 1110
1183void hrtick_resched(void) 1111static void init_rq_hrtick(struct rq *rq)
1184{ 1112{
1185 struct rq *rq; 1113#ifdef CONFIG_SMP
1186 unsigned long flags; 1114 rq->hrtick_csd_pending = 0;
1187 1115
1188 if (!test_thread_flag(TIF_HRTICK_RESCHED)) 1116 rq->hrtick_csd.flags = 0;
1189 return; 1117 rq->hrtick_csd.func = __hrtick_start;
1118 rq->hrtick_csd.info = rq;
1119#endif
1190 1120
1191 local_irq_save(flags); 1121 hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
1192 rq = cpu_rq(smp_processor_id()); 1122 rq->hrtick_timer.function = hrtick;
1193 hrtick_set(rq); 1123 rq->hrtick_timer.cb_mode = HRTIMER_CB_IRQSAFE_NO_SOFTIRQ;
1194 local_irq_restore(flags);
1195} 1124}
1196#else 1125#else
1197static inline void hrtick_clear(struct rq *rq) 1126static inline void hrtick_clear(struct rq *rq)
1198{ 1127{
1199} 1128}
1200 1129
1201static inline void hrtick_set(struct rq *rq)
1202{
1203}
1204
1205static inline void init_rq_hrtick(struct rq *rq) 1130static inline void init_rq_hrtick(struct rq *rq)
1206{ 1131{
1207} 1132}
1208 1133
1209void hrtick_resched(void)
1210{
1211}
1212
1213static inline void init_hrtick(void) 1134static inline void init_hrtick(void)
1214{ 1135{
1215} 1136}
@@ -1228,16 +1149,16 @@ static inline void init_hrtick(void)
1228#define tsk_is_polling(t) test_tsk_thread_flag(t, TIF_POLLING_NRFLAG) 1149#define tsk_is_polling(t) test_tsk_thread_flag(t, TIF_POLLING_NRFLAG)
1229#endif 1150#endif
1230 1151
1231static void __resched_task(struct task_struct *p, int tif_bit) 1152static void resched_task(struct task_struct *p)
1232{ 1153{
1233 int cpu; 1154 int cpu;
1234 1155
1235 assert_spin_locked(&task_rq(p)->lock); 1156 assert_spin_locked(&task_rq(p)->lock);
1236 1157
1237 if (unlikely(test_tsk_thread_flag(p, tif_bit))) 1158 if (unlikely(test_tsk_thread_flag(p, TIF_NEED_RESCHED)))
1238 return; 1159 return;
1239 1160
1240 set_tsk_thread_flag(p, tif_bit); 1161 set_tsk_thread_flag(p, TIF_NEED_RESCHED);
1241 1162
1242 cpu = task_cpu(p); 1163 cpu = task_cpu(p);
1243 if (cpu == smp_processor_id()) 1164 if (cpu == smp_processor_id())
@@ -1303,10 +1224,10 @@ void wake_up_idle_cpu(int cpu)
1303#endif /* CONFIG_NO_HZ */ 1224#endif /* CONFIG_NO_HZ */
1304 1225
1305#else /* !CONFIG_SMP */ 1226#else /* !CONFIG_SMP */
1306static void __resched_task(struct task_struct *p, int tif_bit) 1227static void resched_task(struct task_struct *p)
1307{ 1228{
1308 assert_spin_locked(&task_rq(p)->lock); 1229 assert_spin_locked(&task_rq(p)->lock);
1309 set_tsk_thread_flag(p, tif_bit); 1230 set_tsk_need_resched(p);
1310} 1231}
1311#endif /* CONFIG_SMP */ 1232#endif /* CONFIG_SMP */
1312 1233
@@ -2108,7 +2029,7 @@ find_idlest_group(struct sched_domain *sd, struct task_struct *p, int this_cpu)
2108 /* Tally up the load of all CPUs in the group */ 2029 /* Tally up the load of all CPUs in the group */
2109 avg_load = 0; 2030 avg_load = 0;
2110 2031
2111 for_each_cpu_mask(i, group->cpumask) { 2032 for_each_cpu_mask_nr(i, group->cpumask) {
2112 /* Bias balancing toward cpus of our domain */ 2033 /* Bias balancing toward cpus of our domain */
2113 if (local_group) 2034 if (local_group)
2114 load = source_load(i, load_idx); 2035 load = source_load(i, load_idx);
@@ -2150,7 +2071,7 @@ find_idlest_cpu(struct sched_group *group, struct task_struct *p, int this_cpu,
2150 /* Traverse only the allowed CPUs */ 2071 /* Traverse only the allowed CPUs */
2151 cpus_and(*tmp, group->cpumask, p->cpus_allowed); 2072 cpus_and(*tmp, group->cpumask, p->cpus_allowed);
2152 2073
2153 for_each_cpu_mask(i, *tmp) { 2074 for_each_cpu_mask_nr(i, *tmp) {
2154 load = weighted_cpuload(i); 2075 load = weighted_cpuload(i);
2155 2076
2156 if (load < min_load || (load == min_load && i == this_cpu)) { 2077 if (load < min_load || (load == min_load && i == this_cpu)) {
@@ -2881,7 +2802,7 @@ static void sched_migrate_task(struct task_struct *p, int dest_cpu)
2881 2802
2882 rq = task_rq_lock(p, &flags); 2803 rq = task_rq_lock(p, &flags);
2883 if (!cpu_isset(dest_cpu, p->cpus_allowed) 2804 if (!cpu_isset(dest_cpu, p->cpus_allowed)
2884 || unlikely(cpu_is_offline(dest_cpu))) 2805 || unlikely(!cpu_active(dest_cpu)))
2885 goto out; 2806 goto out;
2886 2807
2887 /* force the process onto the specified CPU */ 2808 /* force the process onto the specified CPU */
@@ -3168,7 +3089,7 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
3168 max_cpu_load = 0; 3089 max_cpu_load = 0;
3169 min_cpu_load = ~0UL; 3090 min_cpu_load = ~0UL;
3170 3091
3171 for_each_cpu_mask(i, group->cpumask) { 3092 for_each_cpu_mask_nr(i, group->cpumask) {
3172 struct rq *rq; 3093 struct rq *rq;
3173 3094
3174 if (!cpu_isset(i, *cpus)) 3095 if (!cpu_isset(i, *cpus))
@@ -3447,7 +3368,7 @@ find_busiest_queue(struct sched_group *group, enum cpu_idle_type idle,
3447 unsigned long max_load = 0; 3368 unsigned long max_load = 0;
3448 int i; 3369 int i;
3449 3370
3450 for_each_cpu_mask(i, group->cpumask) { 3371 for_each_cpu_mask_nr(i, group->cpumask) {
3451 unsigned long wl; 3372 unsigned long wl;
3452 3373
3453 if (!cpu_isset(i, *cpus)) 3374 if (!cpu_isset(i, *cpus))
@@ -3849,7 +3770,7 @@ int select_nohz_load_balancer(int stop_tick)
3849 /* 3770 /*
3850 * If we are going offline and still the leader, give up! 3771 * If we are going offline and still the leader, give up!
3851 */ 3772 */
3852 if (cpu_is_offline(cpu) && 3773 if (!cpu_active(cpu) &&
3853 atomic_read(&nohz.load_balancer) == cpu) { 3774 atomic_read(&nohz.load_balancer) == cpu) {
3854 if (atomic_cmpxchg(&nohz.load_balancer, cpu, -1) != cpu) 3775 if (atomic_cmpxchg(&nohz.load_balancer, cpu, -1) != cpu)
3855 BUG(); 3776 BUG();
@@ -3989,7 +3910,7 @@ static void run_rebalance_domains(struct softirq_action *h)
3989 int balance_cpu; 3910 int balance_cpu;
3990 3911
3991 cpu_clear(this_cpu, cpus); 3912 cpu_clear(this_cpu, cpus);
3992 for_each_cpu_mask(balance_cpu, cpus) { 3913 for_each_cpu_mask_nr(balance_cpu, cpus) {
3993 /* 3914 /*
3994 * If this cpu gets work to do, stop the load balancing 3915 * If this cpu gets work to do, stop the load balancing
3995 * work being done for other cpus. Next load 3916 * work being done for other cpus. Next load
@@ -4125,6 +4046,8 @@ void account_user_time(struct task_struct *p, cputime_t cputime)
4125 cpustat->nice = cputime64_add(cpustat->nice, tmp); 4046 cpustat->nice = cputime64_add(cpustat->nice, tmp);
4126 else 4047 else
4127 cpustat->user = cputime64_add(cpustat->user, tmp); 4048 cpustat->user = cputime64_add(cpustat->user, tmp);
4049 /* Account for user time used */
4050 acct_update_integrals(p);
4128} 4051}
4129 4052
4130/* 4053/*
@@ -4395,7 +4318,7 @@ asmlinkage void __sched schedule(void)
4395 struct task_struct *prev, *next; 4318 struct task_struct *prev, *next;
4396 unsigned long *switch_count; 4319 unsigned long *switch_count;
4397 struct rq *rq; 4320 struct rq *rq;
4398 int cpu, hrtick = sched_feat(HRTICK); 4321 int cpu;
4399 4322
4400need_resched: 4323need_resched:
4401 preempt_disable(); 4324 preempt_disable();
@@ -4410,7 +4333,7 @@ need_resched_nonpreemptible:
4410 4333
4411 schedule_debug(prev); 4334 schedule_debug(prev);
4412 4335
4413 if (hrtick) 4336 if (sched_feat(HRTICK))
4414 hrtick_clear(rq); 4337 hrtick_clear(rq);
4415 4338
4416 /* 4339 /*
@@ -4457,9 +4380,6 @@ need_resched_nonpreemptible:
4457 } else 4380 } else
4458 spin_unlock_irq(&rq->lock); 4381 spin_unlock_irq(&rq->lock);
4459 4382
4460 if (hrtick)
4461 hrtick_set(rq);
4462
4463 if (unlikely(reacquire_kernel_lock(current) < 0)) 4383 if (unlikely(reacquire_kernel_lock(current) < 0))
4464 goto need_resched_nonpreemptible; 4384 goto need_resched_nonpreemptible;
4465 4385
@@ -5876,7 +5796,7 @@ static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
5876 struct rq *rq_dest, *rq_src; 5796 struct rq *rq_dest, *rq_src;
5877 int ret = 0, on_rq; 5797 int ret = 0, on_rq;
5878 5798
5879 if (unlikely(cpu_is_offline(dest_cpu))) 5799 if (unlikely(!cpu_active(dest_cpu)))
5880 return ret; 5800 return ret;
5881 5801
5882 rq_src = cpu_rq(src_cpu); 5802 rq_src = cpu_rq(src_cpu);
@@ -6768,7 +6688,8 @@ static cpumask_t cpu_isolated_map = CPU_MASK_NONE;
6768/* Setup the mask of cpus configured for isolated domains */ 6688/* Setup the mask of cpus configured for isolated domains */
6769static int __init isolated_cpu_setup(char *str) 6689static int __init isolated_cpu_setup(char *str)
6770{ 6690{
6771 int ints[NR_CPUS], i; 6691 static int __initdata ints[NR_CPUS];
6692 int i;
6772 6693
6773 str = get_options(str, ARRAY_SIZE(ints), ints); 6694 str = get_options(str, ARRAY_SIZE(ints), ints);
6774 cpus_clear(cpu_isolated_map); 6695 cpus_clear(cpu_isolated_map);
@@ -6802,7 +6723,7 @@ init_sched_build_groups(const cpumask_t *span, const cpumask_t *cpu_map,
6802 6723
6803 cpus_clear(*covered); 6724 cpus_clear(*covered);
6804 6725
6805 for_each_cpu_mask(i, *span) { 6726 for_each_cpu_mask_nr(i, *span) {
6806 struct sched_group *sg; 6727 struct sched_group *sg;
6807 int group = group_fn(i, cpu_map, &sg, tmpmask); 6728 int group = group_fn(i, cpu_map, &sg, tmpmask);
6808 int j; 6729 int j;
@@ -6813,7 +6734,7 @@ init_sched_build_groups(const cpumask_t *span, const cpumask_t *cpu_map,
6813 cpus_clear(sg->cpumask); 6734 cpus_clear(sg->cpumask);
6814 sg->__cpu_power = 0; 6735 sg->__cpu_power = 0;
6815 6736
6816 for_each_cpu_mask(j, *span) { 6737 for_each_cpu_mask_nr(j, *span) {
6817 if (group_fn(j, cpu_map, NULL, tmpmask) != group) 6738 if (group_fn(j, cpu_map, NULL, tmpmask) != group)
6818 continue; 6739 continue;
6819 6740
@@ -7013,7 +6934,7 @@ static void init_numa_sched_groups_power(struct sched_group *group_head)
7013 if (!sg) 6934 if (!sg)
7014 return; 6935 return;
7015 do { 6936 do {
7016 for_each_cpu_mask(j, sg->cpumask) { 6937 for_each_cpu_mask_nr(j, sg->cpumask) {
7017 struct sched_domain *sd; 6938 struct sched_domain *sd;
7018 6939
7019 sd = &per_cpu(phys_domains, j); 6940 sd = &per_cpu(phys_domains, j);
@@ -7038,7 +6959,7 @@ static void free_sched_groups(const cpumask_t *cpu_map, cpumask_t *nodemask)
7038{ 6959{
7039 int cpu, i; 6960 int cpu, i;
7040 6961
7041 for_each_cpu_mask(cpu, *cpu_map) { 6962 for_each_cpu_mask_nr(cpu, *cpu_map) {
7042 struct sched_group **sched_group_nodes 6963 struct sched_group **sched_group_nodes
7043 = sched_group_nodes_bycpu[cpu]; 6964 = sched_group_nodes_bycpu[cpu];
7044 6965
@@ -7277,7 +7198,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map,
7277 /* 7198 /*
7278 * Set up domains for cpus specified by the cpu_map. 7199 * Set up domains for cpus specified by the cpu_map.
7279 */ 7200 */
7280 for_each_cpu_mask(i, *cpu_map) { 7201 for_each_cpu_mask_nr(i, *cpu_map) {
7281 struct sched_domain *sd = NULL, *p; 7202 struct sched_domain *sd = NULL, *p;
7282 SCHED_CPUMASK_VAR(nodemask, allmasks); 7203 SCHED_CPUMASK_VAR(nodemask, allmasks);
7283 7204
@@ -7344,7 +7265,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map,
7344 7265
7345#ifdef CONFIG_SCHED_SMT 7266#ifdef CONFIG_SCHED_SMT
7346 /* Set up CPU (sibling) groups */ 7267 /* Set up CPU (sibling) groups */
7347 for_each_cpu_mask(i, *cpu_map) { 7268 for_each_cpu_mask_nr(i, *cpu_map) {
7348 SCHED_CPUMASK_VAR(this_sibling_map, allmasks); 7269 SCHED_CPUMASK_VAR(this_sibling_map, allmasks);
7349 SCHED_CPUMASK_VAR(send_covered, allmasks); 7270 SCHED_CPUMASK_VAR(send_covered, allmasks);
7350 7271
@@ -7361,7 +7282,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map,
7361 7282
7362#ifdef CONFIG_SCHED_MC 7283#ifdef CONFIG_SCHED_MC
7363 /* Set up multi-core groups */ 7284 /* Set up multi-core groups */
7364 for_each_cpu_mask(i, *cpu_map) { 7285 for_each_cpu_mask_nr(i, *cpu_map) {
7365 SCHED_CPUMASK_VAR(this_core_map, allmasks); 7286 SCHED_CPUMASK_VAR(this_core_map, allmasks);
7366 SCHED_CPUMASK_VAR(send_covered, allmasks); 7287 SCHED_CPUMASK_VAR(send_covered, allmasks);
7367 7288
@@ -7428,7 +7349,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map,
7428 goto error; 7349 goto error;
7429 } 7350 }
7430 sched_group_nodes[i] = sg; 7351 sched_group_nodes[i] = sg;
7431 for_each_cpu_mask(j, *nodemask) { 7352 for_each_cpu_mask_nr(j, *nodemask) {
7432 struct sched_domain *sd; 7353 struct sched_domain *sd;
7433 7354
7434 sd = &per_cpu(node_domains, j); 7355 sd = &per_cpu(node_domains, j);
@@ -7474,21 +7395,21 @@ static int __build_sched_domains(const cpumask_t *cpu_map,
7474 7395
7475 /* Calculate CPU power for physical packages and nodes */ 7396 /* Calculate CPU power for physical packages and nodes */
7476#ifdef CONFIG_SCHED_SMT 7397#ifdef CONFIG_SCHED_SMT
7477 for_each_cpu_mask(i, *cpu_map) { 7398 for_each_cpu_mask_nr(i, *cpu_map) {
7478 struct sched_domain *sd = &per_cpu(cpu_domains, i); 7399 struct sched_domain *sd = &per_cpu(cpu_domains, i);
7479 7400
7480 init_sched_groups_power(i, sd); 7401 init_sched_groups_power(i, sd);
7481 } 7402 }
7482#endif 7403#endif
7483#ifdef CONFIG_SCHED_MC 7404#ifdef CONFIG_SCHED_MC
7484 for_each_cpu_mask(i, *cpu_map) { 7405 for_each_cpu_mask_nr(i, *cpu_map) {
7485 struct sched_domain *sd = &per_cpu(core_domains, i); 7406 struct sched_domain *sd = &per_cpu(core_domains, i);
7486 7407
7487 init_sched_groups_power(i, sd); 7408 init_sched_groups_power(i, sd);
7488 } 7409 }
7489#endif 7410#endif
7490 7411
7491 for_each_cpu_mask(i, *cpu_map) { 7412 for_each_cpu_mask_nr(i, *cpu_map) {
7492 struct sched_domain *sd = &per_cpu(phys_domains, i); 7413 struct sched_domain *sd = &per_cpu(phys_domains, i);
7493 7414
7494 init_sched_groups_power(i, sd); 7415 init_sched_groups_power(i, sd);
@@ -7508,7 +7429,7 @@ static int __build_sched_domains(const cpumask_t *cpu_map,
7508#endif 7429#endif
7509 7430
7510 /* Attach the domains */ 7431 /* Attach the domains */
7511 for_each_cpu_mask(i, *cpu_map) { 7432 for_each_cpu_mask_nr(i, *cpu_map) {
7512 struct sched_domain *sd; 7433 struct sched_domain *sd;
7513#ifdef CONFIG_SCHED_SMT 7434#ifdef CONFIG_SCHED_SMT
7514 sd = &per_cpu(cpu_domains, i); 7435 sd = &per_cpu(cpu_domains, i);
@@ -7553,18 +7474,6 @@ void __attribute__((weak)) arch_update_cpu_topology(void)
7553} 7474}
7554 7475
7555/* 7476/*
7556 * Free current domain masks.
7557 * Called after all cpus are attached to NULL domain.
7558 */
7559static void free_sched_domains(void)
7560{
7561 ndoms_cur = 0;
7562 if (doms_cur != &fallback_doms)
7563 kfree(doms_cur);
7564 doms_cur = &fallback_doms;
7565}
7566
7567/*
7568 * Set up scheduler domains and groups. Callers must hold the hotplug lock. 7477 * Set up scheduler domains and groups. Callers must hold the hotplug lock.
7569 * For now this just excludes isolated cpus, but could be used to 7478 * For now this just excludes isolated cpus, but could be used to
7570 * exclude other special cases in the future. 7479 * exclude other special cases in the future.
@@ -7603,7 +7512,7 @@ static void detach_destroy_domains(const cpumask_t *cpu_map)
7603 7512
7604 unregister_sched_domain_sysctl(); 7513 unregister_sched_domain_sysctl();
7605 7514
7606 for_each_cpu_mask(i, *cpu_map) 7515 for_each_cpu_mask_nr(i, *cpu_map)
7607 cpu_attach_domain(NULL, &def_root_domain, i); 7516 cpu_attach_domain(NULL, &def_root_domain, i);
7608 synchronize_sched(); 7517 synchronize_sched();
7609 arch_destroy_sched_domains(cpu_map, &tmpmask); 7518 arch_destroy_sched_domains(cpu_map, &tmpmask);
@@ -7642,7 +7551,7 @@ static int dattrs_equal(struct sched_domain_attr *cur, int idx_cur,
7642 * ownership of it and will kfree it when done with it. If the caller 7551 * ownership of it and will kfree it when done with it. If the caller
7643 * failed the kmalloc call, then it can pass in doms_new == NULL, 7552 * failed the kmalloc call, then it can pass in doms_new == NULL,
7644 * and partition_sched_domains() will fallback to the single partition 7553 * and partition_sched_domains() will fallback to the single partition
7645 * 'fallback_doms'. 7554 * 'fallback_doms', it also forces the domains to be rebuilt.
7646 * 7555 *
7647 * Call with hotplug lock held 7556 * Call with hotplug lock held
7648 */ 7557 */
@@ -7656,12 +7565,8 @@ void partition_sched_domains(int ndoms_new, cpumask_t *doms_new,
7656 /* always unregister in case we don't destroy any domains */ 7565 /* always unregister in case we don't destroy any domains */
7657 unregister_sched_domain_sysctl(); 7566 unregister_sched_domain_sysctl();
7658 7567
7659 if (doms_new == NULL) { 7568 if (doms_new == NULL)
7660 ndoms_new = 1; 7569 ndoms_new = 0;
7661 doms_new = &fallback_doms;
7662 cpus_andnot(doms_new[0], cpu_online_map, cpu_isolated_map);
7663 dattr_new = NULL;
7664 }
7665 7570
7666 /* Destroy deleted domains */ 7571 /* Destroy deleted domains */
7667 for (i = 0; i < ndoms_cur; i++) { 7572 for (i = 0; i < ndoms_cur; i++) {
@@ -7676,6 +7581,14 @@ match1:
7676 ; 7581 ;
7677 } 7582 }
7678 7583
7584 if (doms_new == NULL) {
7585 ndoms_cur = 0;
7586 ndoms_new = 1;
7587 doms_new = &fallback_doms;
7588 cpus_andnot(doms_new[0], cpu_online_map, cpu_isolated_map);
7589 dattr_new = NULL;
7590 }
7591
7679 /* Build new domains */ 7592 /* Build new domains */
7680 for (i = 0; i < ndoms_new; i++) { 7593 for (i = 0; i < ndoms_new; i++) {
7681 for (j = 0; j < ndoms_cur; j++) { 7594 for (j = 0; j < ndoms_cur; j++) {
@@ -7706,17 +7619,10 @@ match2:
7706#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) 7619#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
7707int arch_reinit_sched_domains(void) 7620int arch_reinit_sched_domains(void)
7708{ 7621{
7709 int err;
7710
7711 get_online_cpus(); 7622 get_online_cpus();
7712 mutex_lock(&sched_domains_mutex); 7623 rebuild_sched_domains();
7713 detach_destroy_domains(&cpu_online_map);
7714 free_sched_domains();
7715 err = arch_init_sched_domains(&cpu_online_map);
7716 mutex_unlock(&sched_domains_mutex);
7717 put_online_cpus(); 7624 put_online_cpus();
7718 7625 return 0;
7719 return err;
7720} 7626}
7721 7627
7722static ssize_t sched_power_savings_store(const char *buf, size_t count, int smt) 7628static ssize_t sched_power_savings_store(const char *buf, size_t count, int smt)
@@ -7737,11 +7643,13 @@ static ssize_t sched_power_savings_store(const char *buf, size_t count, int smt)
7737} 7643}
7738 7644
7739#ifdef CONFIG_SCHED_MC 7645#ifdef CONFIG_SCHED_MC
7740static ssize_t sched_mc_power_savings_show(struct sys_device *dev, char *page) 7646static ssize_t sched_mc_power_savings_show(struct sys_device *dev,
7647 struct sysdev_attribute *attr, char *page)
7741{ 7648{
7742 return sprintf(page, "%u\n", sched_mc_power_savings); 7649 return sprintf(page, "%u\n", sched_mc_power_savings);
7743} 7650}
7744static ssize_t sched_mc_power_savings_store(struct sys_device *dev, 7651static ssize_t sched_mc_power_savings_store(struct sys_device *dev,
7652 struct sysdev_attribute *attr,
7745 const char *buf, size_t count) 7653 const char *buf, size_t count)
7746{ 7654{
7747 return sched_power_savings_store(buf, count, 0); 7655 return sched_power_savings_store(buf, count, 0);
@@ -7751,11 +7659,13 @@ static SYSDEV_ATTR(sched_mc_power_savings, 0644, sched_mc_power_savings_show,
7751#endif 7659#endif
7752 7660
7753#ifdef CONFIG_SCHED_SMT 7661#ifdef CONFIG_SCHED_SMT
7754static ssize_t sched_smt_power_savings_show(struct sys_device *dev, char *page) 7662static ssize_t sched_smt_power_savings_show(struct sys_device *dev,
7663 struct sysdev_attribute *attr, char *page)
7755{ 7664{
7756 return sprintf(page, "%u\n", sched_smt_power_savings); 7665 return sprintf(page, "%u\n", sched_smt_power_savings);
7757} 7666}
7758static ssize_t sched_smt_power_savings_store(struct sys_device *dev, 7667static ssize_t sched_smt_power_savings_store(struct sys_device *dev,
7668 struct sysdev_attribute *attr,
7759 const char *buf, size_t count) 7669 const char *buf, size_t count)
7760{ 7670{
7761 return sched_power_savings_store(buf, count, 1); 7671 return sched_power_savings_store(buf, count, 1);
@@ -7782,59 +7692,49 @@ int sched_create_sysfs_power_savings_entries(struct sysdev_class *cls)
7782} 7692}
7783#endif /* CONFIG_SCHED_MC || CONFIG_SCHED_SMT */ 7693#endif /* CONFIG_SCHED_MC || CONFIG_SCHED_SMT */
7784 7694
7695#ifndef CONFIG_CPUSETS
7785/* 7696/*
7786 * Force a reinitialization of the sched domains hierarchy. The domains 7697 * Add online and remove offline CPUs from the scheduler domains.
7787 * and groups cannot be updated in place without racing with the balancing 7698 * When cpusets are enabled they take over this function.
7788 * code, so we temporarily attach all running cpus to the NULL domain
7789 * which will prevent rebalancing while the sched domains are recalculated.
7790 */ 7699 */
7791static int update_sched_domains(struct notifier_block *nfb, 7700static int update_sched_domains(struct notifier_block *nfb,
7792 unsigned long action, void *hcpu) 7701 unsigned long action, void *hcpu)
7793{ 7702{
7703 switch (action) {
7704 case CPU_ONLINE:
7705 case CPU_ONLINE_FROZEN:
7706 case CPU_DEAD:
7707 case CPU_DEAD_FROZEN:
7708 partition_sched_domains(0, NULL, NULL);
7709 return NOTIFY_OK;
7710
7711 default:
7712 return NOTIFY_DONE;
7713 }
7714}
7715#endif
7716
7717static int update_runtime(struct notifier_block *nfb,
7718 unsigned long action, void *hcpu)
7719{
7794 int cpu = (int)(long)hcpu; 7720 int cpu = (int)(long)hcpu;
7795 7721
7796 switch (action) { 7722 switch (action) {
7797 case CPU_DOWN_PREPARE: 7723 case CPU_DOWN_PREPARE:
7798 case CPU_DOWN_PREPARE_FROZEN: 7724 case CPU_DOWN_PREPARE_FROZEN:
7799 disable_runtime(cpu_rq(cpu)); 7725 disable_runtime(cpu_rq(cpu));
7800 /* fall-through */
7801 case CPU_UP_PREPARE:
7802 case CPU_UP_PREPARE_FROZEN:
7803 detach_destroy_domains(&cpu_online_map);
7804 free_sched_domains();
7805 return NOTIFY_OK; 7726 return NOTIFY_OK;
7806 7727
7807
7808 case CPU_DOWN_FAILED: 7728 case CPU_DOWN_FAILED:
7809 case CPU_DOWN_FAILED_FROZEN: 7729 case CPU_DOWN_FAILED_FROZEN:
7810 case CPU_ONLINE: 7730 case CPU_ONLINE:
7811 case CPU_ONLINE_FROZEN: 7731 case CPU_ONLINE_FROZEN:
7812 enable_runtime(cpu_rq(cpu)); 7732 enable_runtime(cpu_rq(cpu));
7813 /* fall-through */ 7733 return NOTIFY_OK;
7814 case CPU_UP_CANCELED: 7734
7815 case CPU_UP_CANCELED_FROZEN:
7816 case CPU_DEAD:
7817 case CPU_DEAD_FROZEN:
7818 /*
7819 * Fall through and re-initialise the domains.
7820 */
7821 break;
7822 default: 7735 default:
7823 return NOTIFY_DONE; 7736 return NOTIFY_DONE;
7824 } 7737 }
7825
7826#ifndef CONFIG_CPUSETS
7827 /*
7828 * Create default domain partitioning if cpusets are disabled.
7829 * Otherwise we let cpusets rebuild the domains based on the
7830 * current setup.
7831 */
7832
7833 /* The hotplug lock is already held by cpu_up/cpu_down */
7834 arch_init_sched_domains(&cpu_online_map);
7835#endif
7836
7837 return NOTIFY_OK;
7838} 7738}
7839 7739
7840void __init sched_init_smp(void) 7740void __init sched_init_smp(void)
@@ -7854,8 +7754,15 @@ void __init sched_init_smp(void)
7854 cpu_set(smp_processor_id(), non_isolated_cpus); 7754 cpu_set(smp_processor_id(), non_isolated_cpus);
7855 mutex_unlock(&sched_domains_mutex); 7755 mutex_unlock(&sched_domains_mutex);
7856 put_online_cpus(); 7756 put_online_cpus();
7757
7758#ifndef CONFIG_CPUSETS
7857 /* XXX: Theoretical race here - CPU may be hotplugged now */ 7759 /* XXX: Theoretical race here - CPU may be hotplugged now */
7858 hotcpu_notifier(update_sched_domains, 0); 7760 hotcpu_notifier(update_sched_domains, 0);
7761#endif
7762
7763 /* RT runtime code needs to handle some hotplug events */
7764 hotcpu_notifier(update_runtime, 0);
7765
7859 init_hrtick(); 7766 init_hrtick();
7860 7767
7861 /* Move init over to a non-isolated CPU */ 7768 /* Move init over to a non-isolated CPU */
diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index f2aa987027d6..cf2cd6ce4cb2 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -878,7 +878,6 @@ entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued)
878#ifdef CONFIG_SCHED_HRTICK 878#ifdef CONFIG_SCHED_HRTICK
879static void hrtick_start_fair(struct rq *rq, struct task_struct *p) 879static void hrtick_start_fair(struct rq *rq, struct task_struct *p)
880{ 880{
881 int requeue = rq->curr == p;
882 struct sched_entity *se = &p->se; 881 struct sched_entity *se = &p->se;
883 struct cfs_rq *cfs_rq = cfs_rq_of(se); 882 struct cfs_rq *cfs_rq = cfs_rq_of(se);
884 883
@@ -899,10 +898,10 @@ static void hrtick_start_fair(struct rq *rq, struct task_struct *p)
899 * Don't schedule slices shorter than 10000ns, that just 898 * Don't schedule slices shorter than 10000ns, that just
900 * doesn't make sense. Rely on vruntime for fairness. 899 * doesn't make sense. Rely on vruntime for fairness.
901 */ 900 */
902 if (!requeue) 901 if (rq->curr != p)
903 delta = max(10000LL, delta); 902 delta = max(10000LL, delta);
904 903
905 hrtick_start(rq, delta, requeue); 904 hrtick_start(rq, delta);
906 } 905 }
907} 906}
908#else /* !CONFIG_SCHED_HRTICK */ 907#else /* !CONFIG_SCHED_HRTICK */
@@ -1004,6 +1003,8 @@ static void yield_task_fair(struct rq *rq)
1004 * not idle and an idle cpu is available. The span of cpus to 1003 * not idle and an idle cpu is available. The span of cpus to
1005 * search starts with cpus closest then further out as needed, 1004 * search starts with cpus closest then further out as needed,
1006 * so we always favor a closer, idle cpu. 1005 * so we always favor a closer, idle cpu.
1006 * Domains may include CPUs that are not usable for migration,
1007 * hence we need to mask them out (cpu_active_map)
1007 * 1008 *
1008 * Returns the CPU we should wake onto. 1009 * Returns the CPU we should wake onto.
1009 */ 1010 */
@@ -1031,7 +1032,8 @@ static int wake_idle(int cpu, struct task_struct *p)
1031 || ((sd->flags & SD_WAKE_IDLE_FAR) 1032 || ((sd->flags & SD_WAKE_IDLE_FAR)
1032 && !task_hot(p, task_rq(p)->clock, sd))) { 1033 && !task_hot(p, task_rq(p)->clock, sd))) {
1033 cpus_and(tmp, sd->span, p->cpus_allowed); 1034 cpus_and(tmp, sd->span, p->cpus_allowed);
1034 for_each_cpu_mask(i, tmp) { 1035 cpus_and(tmp, tmp, cpu_active_map);
1036 for_each_cpu_mask_nr(i, tmp) {
1035 if (idle_cpu(i)) { 1037 if (idle_cpu(i)) {
1036 if (i != task_cpu(p)) { 1038 if (i != task_cpu(p)) {
1037 schedstat_inc(p, 1039 schedstat_inc(p,
diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c
index 47ceac9e8552..908c04f9dad0 100644
--- a/kernel/sched_rt.c
+++ b/kernel/sched_rt.c
@@ -240,7 +240,7 @@ static int do_balance_runtime(struct rt_rq *rt_rq)
240 240
241 spin_lock(&rt_b->rt_runtime_lock); 241 spin_lock(&rt_b->rt_runtime_lock);
242 rt_period = ktime_to_ns(rt_b->rt_period); 242 rt_period = ktime_to_ns(rt_b->rt_period);
243 for_each_cpu_mask(i, rd->span) { 243 for_each_cpu_mask_nr(i, rd->span) {
244 struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i); 244 struct rt_rq *iter = sched_rt_period_rt_rq(rt_b, i);
245 s64 diff; 245 s64 diff;
246 246
@@ -253,7 +253,7 @@ static int do_balance_runtime(struct rt_rq *rt_rq)
253 253
254 diff = iter->rt_runtime - iter->rt_time; 254 diff = iter->rt_runtime - iter->rt_time;
255 if (diff > 0) { 255 if (diff > 0) {
256 do_div(diff, weight); 256 diff = div_u64((u64)diff, weight);
257 if (rt_rq->rt_runtime + diff > rt_period) 257 if (rt_rq->rt_runtime + diff > rt_period)
258 diff = rt_period - rt_rq->rt_runtime; 258 diff = rt_period - rt_rq->rt_runtime;
259 iter->rt_runtime -= diff; 259 iter->rt_runtime -= diff;
@@ -505,7 +505,9 @@ void inc_rt_tasks(struct sched_rt_entity *rt_se, struct rt_rq *rt_rq)
505 rt_rq->rt_nr_running++; 505 rt_rq->rt_nr_running++;
506#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED 506#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
507 if (rt_se_prio(rt_se) < rt_rq->highest_prio) { 507 if (rt_se_prio(rt_se) < rt_rq->highest_prio) {
508#ifdef CONFIG_SMP
508 struct rq *rq = rq_of_rt_rq(rt_rq); 509 struct rq *rq = rq_of_rt_rq(rt_rq);
510#endif
509 511
510 rt_rq->highest_prio = rt_se_prio(rt_se); 512 rt_rq->highest_prio = rt_se_prio(rt_se);
511#ifdef CONFIG_SMP 513#ifdef CONFIG_SMP
@@ -599,11 +601,7 @@ static void __enqueue_rt_entity(struct sched_rt_entity *rt_se)
599 if (group_rq && (rt_rq_throttled(group_rq) || !group_rq->rt_nr_running)) 601 if (group_rq && (rt_rq_throttled(group_rq) || !group_rq->rt_nr_running))
600 return; 602 return;
601 603
602 if (rt_se->nr_cpus_allowed == 1) 604 list_add_tail(&rt_se->run_list, queue);
603 list_add(&rt_se->run_list, queue);
604 else
605 list_add_tail(&rt_se->run_list, queue);
606
607 __set_bit(rt_se_prio(rt_se), array->bitmap); 605 __set_bit(rt_se_prio(rt_se), array->bitmap);
608 606
609 inc_rt_tasks(rt_se, rt_rq); 607 inc_rt_tasks(rt_se, rt_rq);
@@ -688,32 +686,34 @@ static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int sleep)
688 * Put task to the end of the run list without the overhead of dequeue 686 * Put task to the end of the run list without the overhead of dequeue
689 * followed by enqueue. 687 * followed by enqueue.
690 */ 688 */
691static 689static void
692void requeue_rt_entity(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se) 690requeue_rt_entity(struct rt_rq *rt_rq, struct sched_rt_entity *rt_se, int head)
693{ 691{
694 struct rt_prio_array *array = &rt_rq->active;
695
696 if (on_rt_rq(rt_se)) { 692 if (on_rt_rq(rt_se)) {
697 list_del_init(&rt_se->run_list); 693 struct rt_prio_array *array = &rt_rq->active;
698 list_add_tail(&rt_se->run_list, 694 struct list_head *queue = array->queue + rt_se_prio(rt_se);
699 array->queue + rt_se_prio(rt_se)); 695
696 if (head)
697 list_move(&rt_se->run_list, queue);
698 else
699 list_move_tail(&rt_se->run_list, queue);
700 } 700 }
701} 701}
702 702
703static void requeue_task_rt(struct rq *rq, struct task_struct *p) 703static void requeue_task_rt(struct rq *rq, struct task_struct *p, int head)
704{ 704{
705 struct sched_rt_entity *rt_se = &p->rt; 705 struct sched_rt_entity *rt_se = &p->rt;
706 struct rt_rq *rt_rq; 706 struct rt_rq *rt_rq;
707 707
708 for_each_sched_rt_entity(rt_se) { 708 for_each_sched_rt_entity(rt_se) {
709 rt_rq = rt_rq_of_se(rt_se); 709 rt_rq = rt_rq_of_se(rt_se);
710 requeue_rt_entity(rt_rq, rt_se); 710 requeue_rt_entity(rt_rq, rt_se, head);
711 } 711 }
712} 712}
713 713
714static void yield_task_rt(struct rq *rq) 714static void yield_task_rt(struct rq *rq)
715{ 715{
716 requeue_task_rt(rq, rq->curr); 716 requeue_task_rt(rq, rq->curr, 0);
717} 717}
718 718
719#ifdef CONFIG_SMP 719#ifdef CONFIG_SMP
@@ -753,6 +753,30 @@ static int select_task_rq_rt(struct task_struct *p, int sync)
753 */ 753 */
754 return task_cpu(p); 754 return task_cpu(p);
755} 755}
756
757static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p)
758{
759 cpumask_t mask;
760
761 if (rq->curr->rt.nr_cpus_allowed == 1)
762 return;
763
764 if (p->rt.nr_cpus_allowed != 1
765 && cpupri_find(&rq->rd->cpupri, p, &mask))
766 return;
767
768 if (!cpupri_find(&rq->rd->cpupri, rq->curr, &mask))
769 return;
770
771 /*
772 * There appears to be other cpus that can accept
773 * current and none to run 'p', so lets reschedule
774 * to try and push current away:
775 */
776 requeue_task_rt(rq, p, 1);
777 resched_task(rq->curr);
778}
779
756#endif /* CONFIG_SMP */ 780#endif /* CONFIG_SMP */
757 781
758/* 782/*
@@ -778,18 +802,8 @@ static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p)
778 * to move current somewhere else, making room for our non-migratable 802 * to move current somewhere else, making room for our non-migratable
779 * task. 803 * task.
780 */ 804 */
781 if((p->prio == rq->curr->prio) 805 if (p->prio == rq->curr->prio && !need_resched())
782 && p->rt.nr_cpus_allowed == 1 806 check_preempt_equal_prio(rq, p);
783 && rq->curr->rt.nr_cpus_allowed != 1) {
784 cpumask_t mask;
785
786 if (cpupri_find(&rq->rd->cpupri, rq->curr, &mask))
787 /*
788 * There appears to be other cpus that can accept
789 * current, so lets reschedule to try and push it away
790 */
791 resched_task(rq->curr);
792 }
793#endif 807#endif
794} 808}
795 809
@@ -922,6 +936,13 @@ static int find_lowest_rq(struct task_struct *task)
922 return -1; /* No targets found */ 936 return -1; /* No targets found */
923 937
924 /* 938 /*
939 * Only consider CPUs that are usable for migration.
940 * I guess we might want to change cpupri_find() to ignore those
941 * in the first place.
942 */
943 cpus_and(*lowest_mask, *lowest_mask, cpu_active_map);
944
945 /*
925 * At this point we have built a mask of cpus representing the 946 * At this point we have built a mask of cpus representing the
926 * lowest priority tasks in the system. Now we want to elect 947 * lowest priority tasks in the system. Now we want to elect
927 * the best one based on our affinity and topology. 948 * the best one based on our affinity and topology.
@@ -1107,7 +1128,7 @@ static int pull_rt_task(struct rq *this_rq)
1107 1128
1108 next = pick_next_task_rt(this_rq); 1129 next = pick_next_task_rt(this_rq);
1109 1130
1110 for_each_cpu_mask(cpu, this_rq->rd->rto_mask) { 1131 for_each_cpu_mask_nr(cpu, this_rq->rd->rto_mask) {
1111 if (this_cpu == cpu) 1132 if (this_cpu == cpu)
1112 continue; 1133 continue;
1113 1134
@@ -1415,7 +1436,7 @@ static void task_tick_rt(struct rq *rq, struct task_struct *p, int queued)
1415 * on the queue: 1436 * on the queue:
1416 */ 1437 */
1417 if (p->rt.run_list.prev != p->rt.run_list.next) { 1438 if (p->rt.run_list.prev != p->rt.run_list.next) {
1418 requeue_task_rt(rq, p); 1439 requeue_task_rt(rq, p, 0);
1419 set_tsk_need_resched(p); 1440 set_tsk_need_resched(p);
1420 } 1441 }
1421} 1442}
diff --git a/kernel/signal.c b/kernel/signal.c
index 6c0958e52ea7..82c3545596c5 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -338,13 +338,9 @@ unblock_all_signals(void)
338 spin_unlock_irqrestore(&current->sighand->siglock, flags); 338 spin_unlock_irqrestore(&current->sighand->siglock, flags);
339} 339}
340 340
341static int collect_signal(int sig, struct sigpending *list, siginfo_t *info) 341static void collect_signal(int sig, struct sigpending *list, siginfo_t *info)
342{ 342{
343 struct sigqueue *q, *first = NULL; 343 struct sigqueue *q, *first = NULL;
344 int still_pending = 0;
345
346 if (unlikely(!sigismember(&list->signal, sig)))
347 return 0;
348 344
349 /* 345 /*
350 * Collect the siginfo appropriate to this signal. Check if 346 * Collect the siginfo appropriate to this signal. Check if
@@ -352,33 +348,30 @@ static int collect_signal(int sig, struct sigpending *list, siginfo_t *info)
352 */ 348 */
353 list_for_each_entry(q, &list->list, list) { 349 list_for_each_entry(q, &list->list, list) {
354 if (q->info.si_signo == sig) { 350 if (q->info.si_signo == sig) {
355 if (first) { 351 if (first)
356 still_pending = 1; 352 goto still_pending;
357 break;
358 }
359 first = q; 353 first = q;
360 } 354 }
361 } 355 }
356
357 sigdelset(&list->signal, sig);
358
362 if (first) { 359 if (first) {
360still_pending:
363 list_del_init(&first->list); 361 list_del_init(&first->list);
364 copy_siginfo(info, &first->info); 362 copy_siginfo(info, &first->info);
365 __sigqueue_free(first); 363 __sigqueue_free(first);
366 if (!still_pending)
367 sigdelset(&list->signal, sig);
368 } else { 364 } else {
369
370 /* Ok, it wasn't in the queue. This must be 365 /* Ok, it wasn't in the queue. This must be
371 a fast-pathed signal or we must have been 366 a fast-pathed signal or we must have been
372 out of queue space. So zero out the info. 367 out of queue space. So zero out the info.
373 */ 368 */
374 sigdelset(&list->signal, sig);
375 info->si_signo = sig; 369 info->si_signo = sig;
376 info->si_errno = 0; 370 info->si_errno = 0;
377 info->si_code = 0; 371 info->si_code = 0;
378 info->si_pid = 0; 372 info->si_pid = 0;
379 info->si_uid = 0; 373 info->si_uid = 0;
380 } 374 }
381 return 1;
382} 375}
383 376
384static int __dequeue_signal(struct sigpending *pending, sigset_t *mask, 377static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
@@ -396,8 +389,7 @@ static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
396 } 389 }
397 } 390 }
398 391
399 if (!collect_signal(sig, pending, info)) 392 collect_signal(sig, pending, info);
400 sig = 0;
401 } 393 }
402 394
403 return sig; 395 return sig;
@@ -462,8 +454,7 @@ int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info)
462 * is to alert stop-signal processing code when another 454 * is to alert stop-signal processing code when another
463 * processor has come along and cleared the flag. 455 * processor has come along and cleared the flag.
464 */ 456 */
465 if (!(tsk->signal->flags & SIGNAL_GROUP_EXIT)) 457 tsk->signal->flags |= SIGNAL_STOP_DEQUEUED;
466 tsk->signal->flags |= SIGNAL_STOP_DEQUEUED;
467 } 458 }
468 if ((info->si_code & __SI_MASK) == __SI_TIMER && info->si_sys_private) { 459 if ((info->si_code & __SI_MASK) == __SI_TIMER && info->si_sys_private) {
469 /* 460 /*
@@ -1125,7 +1116,7 @@ EXPORT_SYMBOL_GPL(kill_pid_info_as_uid);
1125 * is probably wrong. Should make it like BSD or SYSV. 1116 * is probably wrong. Should make it like BSD or SYSV.
1126 */ 1117 */
1127 1118
1128static int kill_something_info(int sig, struct siginfo *info, int pid) 1119static int kill_something_info(int sig, struct siginfo *info, pid_t pid)
1129{ 1120{
1130 int ret; 1121 int ret;
1131 1122
@@ -1237,17 +1228,6 @@ int kill_pid(struct pid *pid, int sig, int priv)
1237} 1228}
1238EXPORT_SYMBOL(kill_pid); 1229EXPORT_SYMBOL(kill_pid);
1239 1230
1240int
1241kill_proc(pid_t pid, int sig, int priv)
1242{
1243 int ret;
1244
1245 rcu_read_lock();
1246 ret = kill_pid_info(sig, __si_special(priv), find_pid(pid));
1247 rcu_read_unlock();
1248 return ret;
1249}
1250
1251/* 1231/*
1252 * These functions support sending signals using preallocated sigqueue 1232 * These functions support sending signals using preallocated sigqueue
1253 * structures. This is needed "because realtime applications cannot 1233 * structures. This is needed "because realtime applications cannot
@@ -1379,10 +1359,9 @@ void do_notify_parent(struct task_struct *tsk, int sig)
1379 1359
1380 info.si_uid = tsk->uid; 1360 info.si_uid = tsk->uid;
1381 1361
1382 /* FIXME: find out whether or not this is supposed to be c*time. */ 1362 info.si_utime = cputime_to_clock_t(cputime_add(tsk->utime,
1383 info.si_utime = cputime_to_jiffies(cputime_add(tsk->utime,
1384 tsk->signal->utime)); 1363 tsk->signal->utime));
1385 info.si_stime = cputime_to_jiffies(cputime_add(tsk->stime, 1364 info.si_stime = cputime_to_clock_t(cputime_add(tsk->stime,
1386 tsk->signal->stime)); 1365 tsk->signal->stime));
1387 1366
1388 info.si_status = tsk->exit_code & 0x7f; 1367 info.si_status = tsk->exit_code & 0x7f;
@@ -1450,9 +1429,8 @@ static void do_notify_parent_cldstop(struct task_struct *tsk, int why)
1450 1429
1451 info.si_uid = tsk->uid; 1430 info.si_uid = tsk->uid;
1452 1431
1453 /* FIXME: find out whether or not this is supposed to be c*time. */ 1432 info.si_utime = cputime_to_clock_t(tsk->utime);
1454 info.si_utime = cputime_to_jiffies(tsk->utime); 1433 info.si_stime = cputime_to_clock_t(tsk->stime);
1455 info.si_stime = cputime_to_jiffies(tsk->stime);
1456 1434
1457 info.si_code = why; 1435 info.si_code = why;
1458 switch (why) { 1436 switch (why) {
@@ -1491,10 +1469,10 @@ static inline int may_ptrace_stop(void)
1491 * is a deadlock situation, and pointless because our tracer 1469 * is a deadlock situation, and pointless because our tracer
1492 * is dead so don't allow us to stop. 1470 * is dead so don't allow us to stop.
1493 * If SIGKILL was already sent before the caller unlocked 1471 * If SIGKILL was already sent before the caller unlocked
1494 * ->siglock we must see ->core_waiters != 0. Otherwise it 1472 * ->siglock we must see ->core_state != NULL. Otherwise it
1495 * is safe to enter schedule(). 1473 * is safe to enter schedule().
1496 */ 1474 */
1497 if (unlikely(current->mm->core_waiters) && 1475 if (unlikely(current->mm->core_state) &&
1498 unlikely(current->mm == current->parent->mm)) 1476 unlikely(current->mm == current->parent->mm))
1499 return 0; 1477 return 0;
1500 1478
@@ -1507,9 +1485,8 @@ static inline int may_ptrace_stop(void)
1507 */ 1485 */
1508static int sigkill_pending(struct task_struct *tsk) 1486static int sigkill_pending(struct task_struct *tsk)
1509{ 1487{
1510 return ((sigismember(&tsk->pending.signal, SIGKILL) || 1488 return sigismember(&tsk->pending.signal, SIGKILL) ||
1511 sigismember(&tsk->signal->shared_pending.signal, SIGKILL)) && 1489 sigismember(&tsk->signal->shared_pending.signal, SIGKILL);
1512 !unlikely(sigismember(&tsk->blocked, SIGKILL)));
1513} 1490}
1514 1491
1515/* 1492/*
@@ -1525,8 +1502,6 @@ static int sigkill_pending(struct task_struct *tsk)
1525 */ 1502 */
1526static void ptrace_stop(int exit_code, int clear_code, siginfo_t *info) 1503static void ptrace_stop(int exit_code, int clear_code, siginfo_t *info)
1527{ 1504{
1528 int killed = 0;
1529
1530 if (arch_ptrace_stop_needed(exit_code, info)) { 1505 if (arch_ptrace_stop_needed(exit_code, info)) {
1531 /* 1506 /*
1532 * The arch code has something special to do before a 1507 * The arch code has something special to do before a
@@ -1542,7 +1517,8 @@ static void ptrace_stop(int exit_code, int clear_code, siginfo_t *info)
1542 spin_unlock_irq(&current->sighand->siglock); 1517 spin_unlock_irq(&current->sighand->siglock);
1543 arch_ptrace_stop(exit_code, info); 1518 arch_ptrace_stop(exit_code, info);
1544 spin_lock_irq(&current->sighand->siglock); 1519 spin_lock_irq(&current->sighand->siglock);
1545 killed = sigkill_pending(current); 1520 if (sigkill_pending(current))
1521 return;
1546 } 1522 }
1547 1523
1548 /* 1524 /*
@@ -1559,7 +1535,7 @@ static void ptrace_stop(int exit_code, int clear_code, siginfo_t *info)
1559 __set_current_state(TASK_TRACED); 1535 __set_current_state(TASK_TRACED);
1560 spin_unlock_irq(&current->sighand->siglock); 1536 spin_unlock_irq(&current->sighand->siglock);
1561 read_lock(&tasklist_lock); 1537 read_lock(&tasklist_lock);
1562 if (!unlikely(killed) && may_ptrace_stop()) { 1538 if (may_ptrace_stop()) {
1563 do_notify_parent_cldstop(current, CLD_TRAPPED); 1539 do_notify_parent_cldstop(current, CLD_TRAPPED);
1564 read_unlock(&tasklist_lock); 1540 read_unlock(&tasklist_lock);
1565 schedule(); 1541 schedule();
@@ -1658,8 +1634,7 @@ static int do_signal_stop(int signr)
1658 } else { 1634 } else {
1659 struct task_struct *t; 1635 struct task_struct *t;
1660 1636
1661 if (unlikely((sig->flags & (SIGNAL_STOP_DEQUEUED | SIGNAL_UNKILLABLE)) 1637 if (!likely(sig->flags & SIGNAL_STOP_DEQUEUED) ||
1662 != SIGNAL_STOP_DEQUEUED) ||
1663 unlikely(signal_group_exit(sig))) 1638 unlikely(signal_group_exit(sig)))
1664 return 0; 1639 return 0;
1665 /* 1640 /*
@@ -1920,7 +1895,6 @@ EXPORT_SYMBOL(recalc_sigpending);
1920EXPORT_SYMBOL_GPL(dequeue_signal); 1895EXPORT_SYMBOL_GPL(dequeue_signal);
1921EXPORT_SYMBOL(flush_signals); 1896EXPORT_SYMBOL(flush_signals);
1922EXPORT_SYMBOL(force_sig); 1897EXPORT_SYMBOL(force_sig);
1923EXPORT_SYMBOL(kill_proc);
1924EXPORT_SYMBOL(ptrace_notify); 1898EXPORT_SYMBOL(ptrace_notify);
1925EXPORT_SYMBOL(send_sig); 1899EXPORT_SYMBOL(send_sig);
1926EXPORT_SYMBOL(send_sig_info); 1900EXPORT_SYMBOL(send_sig_info);
@@ -2196,7 +2170,7 @@ sys_rt_sigtimedwait(const sigset_t __user *uthese,
2196} 2170}
2197 2171
2198asmlinkage long 2172asmlinkage long
2199sys_kill(int pid, int sig) 2173sys_kill(pid_t pid, int sig)
2200{ 2174{
2201 struct siginfo info; 2175 struct siginfo info;
2202 2176
@@ -2209,7 +2183,7 @@ sys_kill(int pid, int sig)
2209 return kill_something_info(sig, &info, pid); 2183 return kill_something_info(sig, &info, pid);
2210} 2184}
2211 2185
2212static int do_tkill(int tgid, int pid, int sig) 2186static int do_tkill(pid_t tgid, pid_t pid, int sig)
2213{ 2187{
2214 int error; 2188 int error;
2215 struct siginfo info; 2189 struct siginfo info;
@@ -2255,7 +2229,7 @@ static int do_tkill(int tgid, int pid, int sig)
2255 * exists but it's not belonging to the target process anymore. This 2229 * exists but it's not belonging to the target process anymore. This
2256 * method solves the problem of threads exiting and PIDs getting reused. 2230 * method solves the problem of threads exiting and PIDs getting reused.
2257 */ 2231 */
2258asmlinkage long sys_tgkill(int tgid, int pid, int sig) 2232asmlinkage long sys_tgkill(pid_t tgid, pid_t pid, int sig)
2259{ 2233{
2260 /* This is only valid for single tasks */ 2234 /* This is only valid for single tasks */
2261 if (pid <= 0 || tgid <= 0) 2235 if (pid <= 0 || tgid <= 0)
@@ -2268,7 +2242,7 @@ asmlinkage long sys_tgkill(int tgid, int pid, int sig)
2268 * Send a signal to only one task, even if it's a CLONE_THREAD task. 2242 * Send a signal to only one task, even if it's a CLONE_THREAD task.
2269 */ 2243 */
2270asmlinkage long 2244asmlinkage long
2271sys_tkill(int pid, int sig) 2245sys_tkill(pid_t pid, int sig)
2272{ 2246{
2273 /* This is only valid for single tasks */ 2247 /* This is only valid for single tasks */
2274 if (pid <= 0) 2248 if (pid <= 0)
@@ -2278,7 +2252,7 @@ sys_tkill(int pid, int sig)
2278} 2252}
2279 2253
2280asmlinkage long 2254asmlinkage long
2281sys_rt_sigqueueinfo(int pid, int sig, siginfo_t __user *uinfo) 2255sys_rt_sigqueueinfo(pid_t pid, int sig, siginfo_t __user *uinfo)
2282{ 2256{
2283 siginfo_t info; 2257 siginfo_t info;
2284 2258
diff --git a/kernel/softirq.c b/kernel/softirq.c
index 81e2fe0f983a..f6b03d56c2bf 100644
--- a/kernel/softirq.c
+++ b/kernel/softirq.c
@@ -286,7 +286,7 @@ void irq_exit(void)
286#ifdef CONFIG_NO_HZ 286#ifdef CONFIG_NO_HZ
287 /* Make sure that timer wheel updates are propagated */ 287 /* Make sure that timer wheel updates are propagated */
288 if (!in_interrupt() && idle_cpu(smp_processor_id()) && !need_resched()) 288 if (!in_interrupt() && idle_cpu(smp_processor_id()) && !need_resched())
289 tick_nohz_stop_sched_tick(); 289 tick_nohz_stop_sched_tick(0);
290 rcu_irq_exit(); 290 rcu_irq_exit();
291#endif 291#endif
292 preempt_enable_no_resched(); 292 preempt_enable_no_resched();
diff --git a/kernel/softlockup.c b/kernel/softlockup.c
index a272d78185eb..7bd8d1aadd5d 100644
--- a/kernel/softlockup.c
+++ b/kernel/softlockup.c
@@ -13,6 +13,7 @@
13#include <linux/delay.h> 13#include <linux/delay.h>
14#include <linux/freezer.h> 14#include <linux/freezer.h>
15#include <linux/kthread.h> 15#include <linux/kthread.h>
16#include <linux/lockdep.h>
16#include <linux/notifier.h> 17#include <linux/notifier.h>
17#include <linux/module.h> 18#include <linux/module.h>
18 19
@@ -25,7 +26,22 @@ static DEFINE_PER_CPU(unsigned long, print_timestamp);
25static DEFINE_PER_CPU(struct task_struct *, watchdog_task); 26static DEFINE_PER_CPU(struct task_struct *, watchdog_task);
26 27
27static int __read_mostly did_panic; 28static int __read_mostly did_panic;
28unsigned long __read_mostly softlockup_thresh = 60; 29int __read_mostly softlockup_thresh = 60;
30
31/*
32 * Should we panic (and reboot, if panic_timeout= is set) when a
33 * soft-lockup occurs:
34 */
35unsigned int __read_mostly softlockup_panic =
36 CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC_VALUE;
37
38static int __init softlockup_panic_setup(char *str)
39{
40 softlockup_panic = simple_strtoul(str, NULL, 0);
41
42 return 1;
43}
44__setup("softlockup_panic=", softlockup_panic_setup);
29 45
30static int 46static int
31softlock_panic(struct notifier_block *this, unsigned long event, void *ptr) 47softlock_panic(struct notifier_block *this, unsigned long event, void *ptr)
@@ -84,6 +100,14 @@ void softlockup_tick(void)
84 struct pt_regs *regs = get_irq_regs(); 100 struct pt_regs *regs = get_irq_regs();
85 unsigned long now; 101 unsigned long now;
86 102
103 /* Is detection switched off? */
104 if (!per_cpu(watchdog_task, this_cpu) || softlockup_thresh <= 0) {
105 /* Be sure we don't false trigger if switched back on */
106 if (touch_timestamp)
107 per_cpu(touch_timestamp, this_cpu) = 0;
108 return;
109 }
110
87 if (touch_timestamp == 0) { 111 if (touch_timestamp == 0) {
88 __touch_softlockup_watchdog(); 112 __touch_softlockup_watchdog();
89 return; 113 return;
@@ -92,11 +116,8 @@ void softlockup_tick(void)
92 print_timestamp = per_cpu(print_timestamp, this_cpu); 116 print_timestamp = per_cpu(print_timestamp, this_cpu);
93 117
94 /* report at most once a second */ 118 /* report at most once a second */
95 if ((print_timestamp >= touch_timestamp && 119 if (print_timestamp == touch_timestamp || did_panic)
96 print_timestamp < (touch_timestamp + 1)) ||
97 did_panic || !per_cpu(watchdog_task, this_cpu)) {
98 return; 120 return;
99 }
100 121
101 /* do not print during early bootup: */ 122 /* do not print during early bootup: */
102 if (unlikely(system_state != SYSTEM_RUNNING)) { 123 if (unlikely(system_state != SYSTEM_RUNNING)) {
@@ -106,8 +127,11 @@ void softlockup_tick(void)
106 127
107 now = get_timestamp(this_cpu); 128 now = get_timestamp(this_cpu);
108 129
109 /* Wake up the high-prio watchdog task every second: */ 130 /*
110 if (now > (touch_timestamp + 1)) 131 * Wake up the high-prio watchdog task twice per
132 * threshold timespan.
133 */
134 if (now > touch_timestamp + softlockup_thresh/2)
111 wake_up_process(per_cpu(watchdog_task, this_cpu)); 135 wake_up_process(per_cpu(watchdog_task, this_cpu));
112 136
113 /* Warn about unreasonable delays: */ 137 /* Warn about unreasonable delays: */
@@ -121,11 +145,15 @@ void softlockup_tick(void)
121 this_cpu, now - touch_timestamp, 145 this_cpu, now - touch_timestamp,
122 current->comm, task_pid_nr(current)); 146 current->comm, task_pid_nr(current));
123 print_modules(); 147 print_modules();
148 print_irqtrace_events(current);
124 if (regs) 149 if (regs)
125 show_regs(regs); 150 show_regs(regs);
126 else 151 else
127 dump_stack(); 152 dump_stack();
128 spin_unlock(&print_lock); 153 spin_unlock(&print_lock);
154
155 if (softlockup_panic)
156 panic("softlockup: hung tasks");
129} 157}
130 158
131/* 159/*
@@ -178,6 +206,9 @@ static void check_hung_task(struct task_struct *t, unsigned long now)
178 206
179 t->last_switch_timestamp = now; 207 t->last_switch_timestamp = now;
180 touch_nmi_watchdog(); 208 touch_nmi_watchdog();
209
210 if (softlockup_panic)
211 panic("softlockup: blocked tasks");
181} 212}
182 213
183/* 214/*
diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c
index ba9b2054ecbd..738b411ff2d3 100644
--- a/kernel/stop_machine.c
+++ b/kernel/stop_machine.c
@@ -33,8 +33,9 @@ static int stopmachine(void *cpu)
33{ 33{
34 int irqs_disabled = 0; 34 int irqs_disabled = 0;
35 int prepared = 0; 35 int prepared = 0;
36 cpumask_of_cpu_ptr(cpumask, (int)(long)cpu);
36 37
37 set_cpus_allowed_ptr(current, &cpumask_of_cpu((int)(long)cpu)); 38 set_cpus_allowed_ptr(current, cpumask);
38 39
39 /* Ack: we are alive */ 40 /* Ack: we are alive */
40 smp_mb(); /* Theoretically the ack = 0 might not be on this CPU yet. */ 41 smp_mb(); /* Theoretically the ack = 0 might not be on this CPU yet. */
diff --git a/kernel/sys.c b/kernel/sys.c
index 14e97282eb6c..0c9d3fa1f5ff 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -1343,8 +1343,6 @@ EXPORT_SYMBOL(in_egroup_p);
1343 1343
1344DECLARE_RWSEM(uts_sem); 1344DECLARE_RWSEM(uts_sem);
1345 1345
1346EXPORT_SYMBOL(uts_sem);
1347
1348asmlinkage long sys_newuname(struct new_utsname __user * name) 1346asmlinkage long sys_newuname(struct new_utsname __user * name)
1349{ 1347{
1350 int errno = 0; 1348 int errno = 0;
@@ -1795,7 +1793,7 @@ int orderly_poweroff(bool force)
1795 goto out; 1793 goto out;
1796 } 1794 }
1797 1795
1798 info = call_usermodehelper_setup(argv[0], argv, envp); 1796 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1799 if (info == NULL) { 1797 if (info == NULL) {
1800 argv_free(argv); 1798 argv_free(argv);
1801 goto out; 1799 goto out;
diff --git a/kernel/sys_ni.c b/kernel/sys_ni.c
index 5b9b467de070..08d6e1bb99ac 100644
--- a/kernel/sys_ni.c
+++ b/kernel/sys_ni.c
@@ -31,6 +31,7 @@ cond_syscall(sys_socketpair);
31cond_syscall(sys_bind); 31cond_syscall(sys_bind);
32cond_syscall(sys_listen); 32cond_syscall(sys_listen);
33cond_syscall(sys_accept); 33cond_syscall(sys_accept);
34cond_syscall(sys_paccept);
34cond_syscall(sys_connect); 35cond_syscall(sys_connect);
35cond_syscall(sys_getsockname); 36cond_syscall(sys_getsockname);
36cond_syscall(sys_getpeername); 37cond_syscall(sys_getpeername);
@@ -56,9 +57,11 @@ cond_syscall(compat_sys_set_robust_list);
56cond_syscall(sys_get_robust_list); 57cond_syscall(sys_get_robust_list);
57cond_syscall(compat_sys_get_robust_list); 58cond_syscall(compat_sys_get_robust_list);
58cond_syscall(sys_epoll_create); 59cond_syscall(sys_epoll_create);
60cond_syscall(sys_epoll_create1);
59cond_syscall(sys_epoll_ctl); 61cond_syscall(sys_epoll_ctl);
60cond_syscall(sys_epoll_wait); 62cond_syscall(sys_epoll_wait);
61cond_syscall(sys_epoll_pwait); 63cond_syscall(sys_epoll_pwait);
64cond_syscall(compat_sys_epoll_pwait);
62cond_syscall(sys_semget); 65cond_syscall(sys_semget);
63cond_syscall(sys_semop); 66cond_syscall(sys_semop);
64cond_syscall(sys_semtimedop); 67cond_syscall(sys_semtimedop);
@@ -94,6 +97,7 @@ cond_syscall(sys_keyctl);
94cond_syscall(compat_sys_keyctl); 97cond_syscall(compat_sys_keyctl);
95cond_syscall(compat_sys_socketcall); 98cond_syscall(compat_sys_socketcall);
96cond_syscall(sys_inotify_init); 99cond_syscall(sys_inotify_init);
100cond_syscall(sys_inotify_init1);
97cond_syscall(sys_inotify_add_watch); 101cond_syscall(sys_inotify_add_watch);
98cond_syscall(sys_inotify_rm_watch); 102cond_syscall(sys_inotify_rm_watch);
99cond_syscall(sys_migrate_pages); 103cond_syscall(sys_migrate_pages);
@@ -154,10 +158,13 @@ cond_syscall(sys_ioprio_get);
154 158
155/* New file descriptors */ 159/* New file descriptors */
156cond_syscall(sys_signalfd); 160cond_syscall(sys_signalfd);
161cond_syscall(sys_signalfd4);
157cond_syscall(compat_sys_signalfd); 162cond_syscall(compat_sys_signalfd);
163cond_syscall(compat_sys_signalfd4);
158cond_syscall(sys_timerfd_create); 164cond_syscall(sys_timerfd_create);
159cond_syscall(sys_timerfd_settime); 165cond_syscall(sys_timerfd_settime);
160cond_syscall(sys_timerfd_gettime); 166cond_syscall(sys_timerfd_gettime);
161cond_syscall(compat_sys_timerfd_settime); 167cond_syscall(compat_sys_timerfd_settime);
162cond_syscall(compat_sys_timerfd_gettime); 168cond_syscall(compat_sys_timerfd_gettime);
163cond_syscall(sys_eventfd); 169cond_syscall(sys_eventfd);
170cond_syscall(sys_eventfd2);
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 6b16e16428d8..35a50db9b6ce 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -43,6 +43,7 @@
43#include <linux/limits.h> 43#include <linux/limits.h>
44#include <linux/dcache.h> 44#include <linux/dcache.h>
45#include <linux/syscalls.h> 45#include <linux/syscalls.h>
46#include <linux/vmstat.h>
46#include <linux/nfs_fs.h> 47#include <linux/nfs_fs.h>
47#include <linux/acpi.h> 48#include <linux/acpi.h>
48#include <linux/reboot.h> 49#include <linux/reboot.h>
@@ -80,7 +81,6 @@ extern int sysctl_drop_caches;
80extern int percpu_pagelist_fraction; 81extern int percpu_pagelist_fraction;
81extern int compat_log; 82extern int compat_log;
82extern int maps_protect; 83extern int maps_protect;
83extern int sysctl_stat_interval;
84extern int latencytop_enabled; 84extern int latencytop_enabled;
85extern int sysctl_nr_open_min, sysctl_nr_open_max; 85extern int sysctl_nr_open_min, sysctl_nr_open_max;
86#ifdef CONFIG_RCU_TORTURE_TEST 86#ifdef CONFIG_RCU_TORTURE_TEST
@@ -88,12 +88,13 @@ extern int rcutorture_runnable;
88#endif /* #ifdef CONFIG_RCU_TORTURE_TEST */ 88#endif /* #ifdef CONFIG_RCU_TORTURE_TEST */
89 89
90/* Constants used for minimum and maximum */ 90/* Constants used for minimum and maximum */
91#if defined(CONFIG_DETECT_SOFTLOCKUP) || defined(CONFIG_HIGHMEM) 91#if defined(CONFIG_HIGHMEM) || defined(CONFIG_DETECT_SOFTLOCKUP)
92static int one = 1; 92static int one = 1;
93#endif 93#endif
94 94
95#ifdef CONFIG_DETECT_SOFTLOCKUP 95#ifdef CONFIG_DETECT_SOFTLOCKUP
96static int sixty = 60; 96static int sixty = 60;
97static int neg_one = -1;
97#endif 98#endif
98 99
99#ifdef CONFIG_MMU 100#ifdef CONFIG_MMU
@@ -110,7 +111,7 @@ static int min_percpu_pagelist_fract = 8;
110 111
111static int ngroups_max = NGROUPS_MAX; 112static int ngroups_max = NGROUPS_MAX;
112 113
113#ifdef CONFIG_KMOD 114#ifdef CONFIG_MODULES
114extern char modprobe_path[]; 115extern char modprobe_path[];
115#endif 116#endif
116#ifdef CONFIG_CHR_DEV_SG 117#ifdef CONFIG_CHR_DEV_SG
@@ -475,7 +476,7 @@ static struct ctl_table kern_table[] = {
475 .proc_handler = &ftrace_enable_sysctl, 476 .proc_handler = &ftrace_enable_sysctl,
476 }, 477 },
477#endif 478#endif
478#ifdef CONFIG_KMOD 479#ifdef CONFIG_MODULES
479 { 480 {
480 .ctl_name = KERN_MODPROBE, 481 .ctl_name = KERN_MODPROBE,
481 .procname = "modprobe", 482 .procname = "modprobe",
@@ -623,7 +624,7 @@ static struct ctl_table kern_table[] = {
623 { 624 {
624 .ctl_name = KERN_PRINTK_RATELIMIT, 625 .ctl_name = KERN_PRINTK_RATELIMIT,
625 .procname = "printk_ratelimit", 626 .procname = "printk_ratelimit",
626 .data = &printk_ratelimit_jiffies, 627 .data = &printk_ratelimit_state.interval,
627 .maxlen = sizeof(int), 628 .maxlen = sizeof(int),
628 .mode = 0644, 629 .mode = 0644,
629 .proc_handler = &proc_dointvec_jiffies, 630 .proc_handler = &proc_dointvec_jiffies,
@@ -632,7 +633,7 @@ static struct ctl_table kern_table[] = {
632 { 633 {
633 .ctl_name = KERN_PRINTK_RATELIMIT_BURST, 634 .ctl_name = KERN_PRINTK_RATELIMIT_BURST,
634 .procname = "printk_ratelimit_burst", 635 .procname = "printk_ratelimit_burst",
635 .data = &printk_ratelimit_burst, 636 .data = &printk_ratelimit_state.burst,
636 .maxlen = sizeof(int), 637 .maxlen = sizeof(int),
637 .mode = 0644, 638 .mode = 0644,
638 .proc_handler = &proc_dointvec, 639 .proc_handler = &proc_dointvec,
@@ -739,13 +740,24 @@ static struct ctl_table kern_table[] = {
739#ifdef CONFIG_DETECT_SOFTLOCKUP 740#ifdef CONFIG_DETECT_SOFTLOCKUP
740 { 741 {
741 .ctl_name = CTL_UNNUMBERED, 742 .ctl_name = CTL_UNNUMBERED,
743 .procname = "softlockup_panic",
744 .data = &softlockup_panic,
745 .maxlen = sizeof(int),
746 .mode = 0644,
747 .proc_handler = &proc_dointvec_minmax,
748 .strategy = &sysctl_intvec,
749 .extra1 = &zero,
750 .extra2 = &one,
751 },
752 {
753 .ctl_name = CTL_UNNUMBERED,
742 .procname = "softlockup_thresh", 754 .procname = "softlockup_thresh",
743 .data = &softlockup_thresh, 755 .data = &softlockup_thresh,
744 .maxlen = sizeof(unsigned long), 756 .maxlen = sizeof(int),
745 .mode = 0644, 757 .mode = 0644,
746 .proc_handler = &proc_doulongvec_minmax, 758 .proc_handler = &proc_dointvec_minmax,
747 .strategy = &sysctl_intvec, 759 .strategy = &sysctl_intvec,
748 .extra1 = &one, 760 .extra1 = &neg_one,
749 .extra2 = &sixty, 761 .extra2 = &sixty,
750 }, 762 },
751 { 763 {
@@ -947,7 +959,7 @@ static struct ctl_table vm_table[] = {
947#ifdef CONFIG_HUGETLB_PAGE 959#ifdef CONFIG_HUGETLB_PAGE
948 { 960 {
949 .procname = "nr_hugepages", 961 .procname = "nr_hugepages",
950 .data = &max_huge_pages, 962 .data = NULL,
951 .maxlen = sizeof(unsigned long), 963 .maxlen = sizeof(unsigned long),
952 .mode = 0644, 964 .mode = 0644,
953 .proc_handler = &hugetlb_sysctl_handler, 965 .proc_handler = &hugetlb_sysctl_handler,
@@ -973,10 +985,12 @@ static struct ctl_table vm_table[] = {
973 { 985 {
974 .ctl_name = CTL_UNNUMBERED, 986 .ctl_name = CTL_UNNUMBERED,
975 .procname = "nr_overcommit_hugepages", 987 .procname = "nr_overcommit_hugepages",
976 .data = &sysctl_overcommit_huge_pages, 988 .data = NULL,
977 .maxlen = sizeof(sysctl_overcommit_huge_pages), 989 .maxlen = sizeof(unsigned long),
978 .mode = 0644, 990 .mode = 0644,
979 .proc_handler = &hugetlb_overcommit_handler, 991 .proc_handler = &hugetlb_overcommit_handler,
992 .extra1 = (void *)&hugetlb_zero,
993 .extra2 = (void *)&hugetlb_infinity,
980 }, 994 },
981#endif 995#endif
982 { 996 {
diff --git a/kernel/sysctl_check.c b/kernel/sysctl_check.c
index c09350d564f2..c35da23ab8fb 100644
--- a/kernel/sysctl_check.c
+++ b/kernel/sysctl_check.c
@@ -1532,6 +1532,8 @@ int sysctl_check_table(struct nsproxy *namespaces, struct ctl_table *table)
1532 sysctl_check_leaf(namespaces, table, &fail); 1532 sysctl_check_leaf(namespaces, table, &fail);
1533 } 1533 }
1534 sysctl_check_bin_path(table, &fail); 1534 sysctl_check_bin_path(table, &fail);
1535 if (table->mode > 0777)
1536 set_fail(&fail, table, "bogus .mode");
1535 if (fail) { 1537 if (fail) {
1536 set_fail(&fail, table, NULL); 1538 set_fail(&fail, table, NULL);
1537 error = -EINVAL; 1539 error = -EINVAL;
diff --git a/kernel/taskstats.c b/kernel/taskstats.c
index 4a23517169a6..bd6be76303cf 100644
--- a/kernel/taskstats.c
+++ b/kernel/taskstats.c
@@ -35,7 +35,7 @@
35 */ 35 */
36#define TASKSTATS_CPUMASK_MAXLEN (100+6*NR_CPUS) 36#define TASKSTATS_CPUMASK_MAXLEN (100+6*NR_CPUS)
37 37
38static DEFINE_PER_CPU(__u32, taskstats_seqnum) = { 0 }; 38static DEFINE_PER_CPU(__u32, taskstats_seqnum);
39static int family_registered; 39static int family_registered;
40struct kmem_cache *taskstats_cache; 40struct kmem_cache *taskstats_cache;
41 41
@@ -301,7 +301,7 @@ static int add_del_listener(pid_t pid, cpumask_t *maskp, int isadd)
301 return -EINVAL; 301 return -EINVAL;
302 302
303 if (isadd == REGISTER) { 303 if (isadd == REGISTER) {
304 for_each_cpu_mask(cpu, mask) { 304 for_each_cpu_mask_nr(cpu, mask) {
305 s = kmalloc_node(sizeof(struct listener), GFP_KERNEL, 305 s = kmalloc_node(sizeof(struct listener), GFP_KERNEL,
306 cpu_to_node(cpu)); 306 cpu_to_node(cpu));
307 if (!s) 307 if (!s)
@@ -320,7 +320,7 @@ static int add_del_listener(pid_t pid, cpumask_t *maskp, int isadd)
320 320
321 /* Deregister or cleanup */ 321 /* Deregister or cleanup */
322cleanup: 322cleanup:
323 for_each_cpu_mask(cpu, mask) { 323 for_each_cpu_mask_nr(cpu, mask) {
324 listeners = &per_cpu(listener_array, cpu); 324 listeners = &per_cpu(listener_array, cpu);
325 down_write(&listeners->sem); 325 down_write(&listeners->sem);
326 list_for_each_entry_safe(s, tmp, &listeners->list, list) { 326 list_for_each_entry_safe(s, tmp, &listeners->list, list) {
diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c
index dadde5361f32..093d4acf993b 100644
--- a/kernel/time/clocksource.c
+++ b/kernel/time/clocksource.c
@@ -145,9 +145,9 @@ static void clocksource_watchdog(unsigned long data)
145 * Cycle through CPUs to check if the CPUs stay 145 * Cycle through CPUs to check if the CPUs stay
146 * synchronized to each other. 146 * synchronized to each other.
147 */ 147 */
148 int next_cpu = next_cpu(raw_smp_processor_id(), cpu_online_map); 148 int next_cpu = next_cpu_nr(raw_smp_processor_id(), cpu_online_map);
149 149
150 if (next_cpu >= NR_CPUS) 150 if (next_cpu >= nr_cpu_ids)
151 next_cpu = first_cpu(cpu_online_map); 151 next_cpu = first_cpu(cpu_online_map);
152 watchdog_timer.expires += WATCHDOG_INTERVAL; 152 watchdog_timer.expires += WATCHDOG_INTERVAL;
153 add_timer_on(&watchdog_timer, next_cpu); 153 add_timer_on(&watchdog_timer, next_cpu);
@@ -376,7 +376,8 @@ void clocksource_unregister(struct clocksource *cs)
376 * Provides sysfs interface for listing current clocksource. 376 * Provides sysfs interface for listing current clocksource.
377 */ 377 */
378static ssize_t 378static ssize_t
379sysfs_show_current_clocksources(struct sys_device *dev, char *buf) 379sysfs_show_current_clocksources(struct sys_device *dev,
380 struct sysdev_attribute *attr, char *buf)
380{ 381{
381 ssize_t count = 0; 382 ssize_t count = 0;
382 383
@@ -397,6 +398,7 @@ sysfs_show_current_clocksources(struct sys_device *dev, char *buf)
397 * clocksource selction. 398 * clocksource selction.
398 */ 399 */
399static ssize_t sysfs_override_clocksource(struct sys_device *dev, 400static ssize_t sysfs_override_clocksource(struct sys_device *dev,
401 struct sysdev_attribute *attr,
400 const char *buf, size_t count) 402 const char *buf, size_t count)
401{ 403{
402 struct clocksource *ovr = NULL; 404 struct clocksource *ovr = NULL;
@@ -449,7 +451,9 @@ static ssize_t sysfs_override_clocksource(struct sys_device *dev,
449 * Provides sysfs interface for listing registered clocksources 451 * Provides sysfs interface for listing registered clocksources
450 */ 452 */
451static ssize_t 453static ssize_t
452sysfs_show_available_clocksources(struct sys_device *dev, char *buf) 454sysfs_show_available_clocksources(struct sys_device *dev,
455 struct sysdev_attribute *attr,
456 char *buf)
453{ 457{
454 struct clocksource *src; 458 struct clocksource *src;
455 ssize_t count = 0; 459 ssize_t count = 0;
diff --git a/kernel/time/tick-broadcast.c b/kernel/time/tick-broadcast.c
index f48d0f09d32f..31463d370b94 100644
--- a/kernel/time/tick-broadcast.c
+++ b/kernel/time/tick-broadcast.c
@@ -399,8 +399,7 @@ again:
399 mask = CPU_MASK_NONE; 399 mask = CPU_MASK_NONE;
400 now = ktime_get(); 400 now = ktime_get();
401 /* Find all expired events */ 401 /* Find all expired events */
402 for (cpu = first_cpu(tick_broadcast_oneshot_mask); cpu != NR_CPUS; 402 for_each_cpu_mask_nr(cpu, tick_broadcast_oneshot_mask) {
403 cpu = next_cpu(cpu, tick_broadcast_oneshot_mask)) {
404 td = &per_cpu(tick_cpu_device, cpu); 403 td = &per_cpu(tick_cpu_device, cpu);
405 if (td->evtdev->next_event.tv64 <= now.tv64) 404 if (td->evtdev->next_event.tv64 <= now.tv64)
406 cpu_set(cpu, mask); 405 cpu_set(cpu, mask);
diff --git a/kernel/time/tick-common.c b/kernel/time/tick-common.c
index 4f3886562b8c..bf43284d6855 100644
--- a/kernel/time/tick-common.c
+++ b/kernel/time/tick-common.c
@@ -135,7 +135,7 @@ void tick_setup_periodic(struct clock_event_device *dev, int broadcast)
135 */ 135 */
136static void tick_setup_device(struct tick_device *td, 136static void tick_setup_device(struct tick_device *td,
137 struct clock_event_device *newdev, int cpu, 137 struct clock_event_device *newdev, int cpu,
138 cpumask_t cpumask) 138 const cpumask_t *cpumask)
139{ 139{
140 ktime_t next_event; 140 ktime_t next_event;
141 void (*handler)(struct clock_event_device *) = NULL; 141 void (*handler)(struct clock_event_device *) = NULL;
@@ -169,8 +169,8 @@ static void tick_setup_device(struct tick_device *td,
169 * When the device is not per cpu, pin the interrupt to the 169 * When the device is not per cpu, pin the interrupt to the
170 * current cpu: 170 * current cpu:
171 */ 171 */
172 if (!cpus_equal(newdev->cpumask, cpumask)) 172 if (!cpus_equal(newdev->cpumask, *cpumask))
173 irq_set_affinity(newdev->irq, cpumask); 173 irq_set_affinity(newdev->irq, *cpumask);
174 174
175 /* 175 /*
176 * When global broadcasting is active, check if the current 176 * When global broadcasting is active, check if the current
@@ -196,20 +196,20 @@ static int tick_check_new_device(struct clock_event_device *newdev)
196 struct tick_device *td; 196 struct tick_device *td;
197 int cpu, ret = NOTIFY_OK; 197 int cpu, ret = NOTIFY_OK;
198 unsigned long flags; 198 unsigned long flags;
199 cpumask_t cpumask; 199 cpumask_of_cpu_ptr_declare(cpumask);
200 200
201 spin_lock_irqsave(&tick_device_lock, flags); 201 spin_lock_irqsave(&tick_device_lock, flags);
202 202
203 cpu = smp_processor_id(); 203 cpu = smp_processor_id();
204 cpumask_of_cpu_ptr_next(cpumask, cpu);
204 if (!cpu_isset(cpu, newdev->cpumask)) 205 if (!cpu_isset(cpu, newdev->cpumask))
205 goto out_bc; 206 goto out_bc;
206 207
207 td = &per_cpu(tick_cpu_device, cpu); 208 td = &per_cpu(tick_cpu_device, cpu);
208 curdev = td->evtdev; 209 curdev = td->evtdev;
209 cpumask = cpumask_of_cpu(cpu);
210 210
211 /* cpu local device ? */ 211 /* cpu local device ? */
212 if (!cpus_equal(newdev->cpumask, cpumask)) { 212 if (!cpus_equal(newdev->cpumask, *cpumask)) {
213 213
214 /* 214 /*
215 * If the cpu affinity of the device interrupt can not 215 * If the cpu affinity of the device interrupt can not
@@ -222,7 +222,7 @@ static int tick_check_new_device(struct clock_event_device *newdev)
222 * If we have a cpu local device already, do not replace it 222 * If we have a cpu local device already, do not replace it
223 * by a non cpu local device 223 * by a non cpu local device
224 */ 224 */
225 if (curdev && cpus_equal(curdev->cpumask, cpumask)) 225 if (curdev && cpus_equal(curdev->cpumask, *cpumask))
226 goto out_bc; 226 goto out_bc;
227 } 227 }
228 228
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index beef7ccdf842..825b4c00fe44 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -140,8 +140,6 @@ void tick_nohz_update_jiffies(void)
140 if (!ts->tick_stopped) 140 if (!ts->tick_stopped)
141 return; 141 return;
142 142
143 touch_softlockup_watchdog();
144
145 cpu_clear(cpu, nohz_cpu_mask); 143 cpu_clear(cpu, nohz_cpu_mask);
146 now = ktime_get(); 144 now = ktime_get();
147 ts->idle_waketime = now; 145 ts->idle_waketime = now;
@@ -149,6 +147,8 @@ void tick_nohz_update_jiffies(void)
149 local_irq_save(flags); 147 local_irq_save(flags);
150 tick_do_update_jiffies64(now); 148 tick_do_update_jiffies64(now);
151 local_irq_restore(flags); 149 local_irq_restore(flags);
150
151 touch_softlockup_watchdog();
152} 152}
153 153
154void tick_nohz_stop_idle(int cpu) 154void tick_nohz_stop_idle(int cpu)
@@ -195,7 +195,7 @@ u64 get_cpu_idle_time_us(int cpu, u64 *last_update_time)
195 * Called either from the idle loop or from irq_exit() when an idle period was 195 * Called either from the idle loop or from irq_exit() when an idle period was
196 * just interrupted by an interrupt which did not cause a reschedule. 196 * just interrupted by an interrupt which did not cause a reschedule.
197 */ 197 */
198void tick_nohz_stop_sched_tick(void) 198void tick_nohz_stop_sched_tick(int inidle)
199{ 199{
200 unsigned long seq, last_jiffies, next_jiffies, delta_jiffies, flags; 200 unsigned long seq, last_jiffies, next_jiffies, delta_jiffies, flags;
201 struct tick_sched *ts; 201 struct tick_sched *ts;
@@ -224,6 +224,11 @@ void tick_nohz_stop_sched_tick(void)
224 if (unlikely(ts->nohz_mode == NOHZ_MODE_INACTIVE)) 224 if (unlikely(ts->nohz_mode == NOHZ_MODE_INACTIVE))
225 goto end; 225 goto end;
226 226
227 if (!inidle && !ts->inidle)
228 goto end;
229
230 ts->inidle = 1;
231
227 if (need_resched()) 232 if (need_resched())
228 goto end; 233 goto end;
229 234
@@ -373,11 +378,14 @@ void tick_nohz_restart_sched_tick(void)
373 local_irq_disable(); 378 local_irq_disable();
374 tick_nohz_stop_idle(cpu); 379 tick_nohz_stop_idle(cpu);
375 380
376 if (!ts->tick_stopped) { 381 if (!ts->inidle || !ts->tick_stopped) {
382 ts->inidle = 0;
377 local_irq_enable(); 383 local_irq_enable();
378 return; 384 return;
379 } 385 }
380 386
387 ts->inidle = 0;
388
381 rcu_exit_nohz(); 389 rcu_exit_nohz();
382 390
383 /* Update jiffies first */ 391 /* Update jiffies first */
diff --git a/kernel/trace/trace_sysprof.c b/kernel/trace/trace_sysprof.c
index 2301e1e7c606..ce2d723c10e1 100644
--- a/kernel/trace/trace_sysprof.c
+++ b/kernel/trace/trace_sysprof.c
@@ -161,7 +161,7 @@ static void timer_notify(struct pt_regs *regs, int cpu)
161 __trace_special(tr, data, 2, regs->ip, 0); 161 __trace_special(tr, data, 2, regs->ip, 0);
162 162
163 while (i < sample_max_depth) { 163 while (i < sample_max_depth) {
164 frame.next_fp = 0; 164 frame.next_fp = NULL;
165 frame.return_address = 0; 165 frame.return_address = 0;
166 if (!copy_stack_frame(fp, &frame)) 166 if (!copy_stack_frame(fp, &frame))
167 break; 167 break;
@@ -213,7 +213,9 @@ static void start_stack_timers(void)
213 int cpu; 213 int cpu;
214 214
215 for_each_online_cpu(cpu) { 215 for_each_online_cpu(cpu) {
216 set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu)); 216 cpumask_of_cpu_ptr(new_mask, cpu);
217
218 set_cpus_allowed_ptr(current, new_mask);
217 start_stack_timer(cpu); 219 start_stack_timer(cpu);
218 } 220 }
219 set_cpus_allowed_ptr(current, &saved_mask); 221 set_cpus_allowed_ptr(current, &saved_mask);
diff --git a/kernel/tsacct.c b/kernel/tsacct.c
index 4ab1b584961b..3da47ccdc5e5 100644
--- a/kernel/tsacct.c
+++ b/kernel/tsacct.c
@@ -28,14 +28,14 @@
28void bacct_add_tsk(struct taskstats *stats, struct task_struct *tsk) 28void bacct_add_tsk(struct taskstats *stats, struct task_struct *tsk)
29{ 29{
30 struct timespec uptime, ts; 30 struct timespec uptime, ts;
31 s64 ac_etime; 31 u64 ac_etime;
32 32
33 BUILD_BUG_ON(TS_COMM_LEN < TASK_COMM_LEN); 33 BUILD_BUG_ON(TS_COMM_LEN < TASK_COMM_LEN);
34 34
35 /* calculate task elapsed time in timespec */ 35 /* calculate task elapsed time in timespec */
36 do_posix_clock_monotonic_gettime(&uptime); 36 do_posix_clock_monotonic_gettime(&uptime);
37 ts = timespec_sub(uptime, tsk->start_time); 37 ts = timespec_sub(uptime, tsk->start_time);
38 /* rebase elapsed time to usec */ 38 /* rebase elapsed time to usec (should never be negative) */
39 ac_etime = timespec_to_ns(&ts); 39 ac_etime = timespec_to_ns(&ts);
40 do_div(ac_etime, NSEC_PER_USEC); 40 do_div(ac_etime, NSEC_PER_USEC);
41 stats->ac_etime = ac_etime; 41 stats->ac_etime = ac_etime;
@@ -84,9 +84,9 @@ void xacct_add_tsk(struct taskstats *stats, struct task_struct *p)
84{ 84{
85 struct mm_struct *mm; 85 struct mm_struct *mm;
86 86
87 /* convert pages-jiffies to Mbyte-usec */ 87 /* convert pages-usec to Mbyte-usec */
88 stats->coremem = jiffies_to_usecs(p->acct_rss_mem1) * PAGE_SIZE / MB; 88 stats->coremem = p->acct_rss_mem1 * PAGE_SIZE / MB;
89 stats->virtmem = jiffies_to_usecs(p->acct_vm_mem1) * PAGE_SIZE / MB; 89 stats->virtmem = p->acct_vm_mem1 * PAGE_SIZE / MB;
90 mm = get_task_mm(p); 90 mm = get_task_mm(p);
91 if (mm) { 91 if (mm) {
92 /* adjust to KB unit */ 92 /* adjust to KB unit */
@@ -118,12 +118,19 @@ void xacct_add_tsk(struct taskstats *stats, struct task_struct *p)
118void acct_update_integrals(struct task_struct *tsk) 118void acct_update_integrals(struct task_struct *tsk)
119{ 119{
120 if (likely(tsk->mm)) { 120 if (likely(tsk->mm)) {
121 long delta = cputime_to_jiffies( 121 cputime_t time, dtime;
122 cputime_sub(tsk->stime, tsk->acct_stimexpd)); 122 struct timeval value;
123 u64 delta;
124
125 time = tsk->stime + tsk->utime;
126 dtime = cputime_sub(time, tsk->acct_timexpd);
127 jiffies_to_timeval(cputime_to_jiffies(dtime), &value);
128 delta = value.tv_sec;
129 delta = delta * USEC_PER_SEC + value.tv_usec;
123 130
124 if (delta == 0) 131 if (delta == 0)
125 return; 132 return;
126 tsk->acct_stimexpd = tsk->stime; 133 tsk->acct_timexpd = time;
127 tsk->acct_rss_mem1 += delta * get_mm_rss(tsk->mm); 134 tsk->acct_rss_mem1 += delta * get_mm_rss(tsk->mm);
128 tsk->acct_vm_mem1 += delta * tsk->mm->total_vm; 135 tsk->acct_vm_mem1 += delta * tsk->mm->total_vm;
129 } 136 }
@@ -135,7 +142,7 @@ void acct_update_integrals(struct task_struct *tsk)
135 */ 142 */
136void acct_clear_integrals(struct task_struct *tsk) 143void acct_clear_integrals(struct task_struct *tsk)
137{ 144{
138 tsk->acct_stimexpd = 0; 145 tsk->acct_timexpd = 0;
139 tsk->acct_rss_mem1 = 0; 146 tsk->acct_rss_mem1 = 0;
140 tsk->acct_vm_mem1 = 0; 147 tsk->acct_vm_mem1 = 0;
141} 148}
diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index ce7799540c91..ec7e4f62aaff 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -125,7 +125,7 @@ struct cpu_workqueue_struct *get_wq_data(struct work_struct *work)
125} 125}
126 126
127static void insert_work(struct cpu_workqueue_struct *cwq, 127static void insert_work(struct cpu_workqueue_struct *cwq,
128 struct work_struct *work, int tail) 128 struct work_struct *work, struct list_head *head)
129{ 129{
130 set_wq_data(work, cwq); 130 set_wq_data(work, cwq);
131 /* 131 /*
@@ -133,21 +133,17 @@ static void insert_work(struct cpu_workqueue_struct *cwq,
133 * result of list_add() below, see try_to_grab_pending(). 133 * result of list_add() below, see try_to_grab_pending().
134 */ 134 */
135 smp_wmb(); 135 smp_wmb();
136 if (tail) 136 list_add_tail(&work->entry, head);
137 list_add_tail(&work->entry, &cwq->worklist);
138 else
139 list_add(&work->entry, &cwq->worklist);
140 wake_up(&cwq->more_work); 137 wake_up(&cwq->more_work);
141} 138}
142 139
143/* Preempt must be disabled. */
144static void __queue_work(struct cpu_workqueue_struct *cwq, 140static void __queue_work(struct cpu_workqueue_struct *cwq,
145 struct work_struct *work) 141 struct work_struct *work)
146{ 142{
147 unsigned long flags; 143 unsigned long flags;
148 144
149 spin_lock_irqsave(&cwq->lock, flags); 145 spin_lock_irqsave(&cwq->lock, flags);
150 insert_work(cwq, work, 1); 146 insert_work(cwq, work, &cwq->worklist);
151 spin_unlock_irqrestore(&cwq->lock, flags); 147 spin_unlock_irqrestore(&cwq->lock, flags);
152} 148}
153 149
@@ -163,17 +159,39 @@ static void __queue_work(struct cpu_workqueue_struct *cwq,
163 */ 159 */
164int queue_work(struct workqueue_struct *wq, struct work_struct *work) 160int queue_work(struct workqueue_struct *wq, struct work_struct *work)
165{ 161{
162 int ret;
163
164 ret = queue_work_on(get_cpu(), wq, work);
165 put_cpu();
166
167 return ret;
168}
169EXPORT_SYMBOL_GPL(queue_work);
170
171/**
172 * queue_work_on - queue work on specific cpu
173 * @cpu: CPU number to execute work on
174 * @wq: workqueue to use
175 * @work: work to queue
176 *
177 * Returns 0 if @work was already on a queue, non-zero otherwise.
178 *
179 * We queue the work to a specific CPU, the caller must ensure it
180 * can't go away.
181 */
182int
183queue_work_on(int cpu, struct workqueue_struct *wq, struct work_struct *work)
184{
166 int ret = 0; 185 int ret = 0;
167 186
168 if (!test_and_set_bit(WORK_STRUCT_PENDING, work_data_bits(work))) { 187 if (!test_and_set_bit(WORK_STRUCT_PENDING, work_data_bits(work))) {
169 BUG_ON(!list_empty(&work->entry)); 188 BUG_ON(!list_empty(&work->entry));
170 __queue_work(wq_per_cpu(wq, get_cpu()), work); 189 __queue_work(wq_per_cpu(wq, cpu), work);
171 put_cpu();
172 ret = 1; 190 ret = 1;
173 } 191 }
174 return ret; 192 return ret;
175} 193}
176EXPORT_SYMBOL_GPL(queue_work); 194EXPORT_SYMBOL_GPL(queue_work_on);
177 195
178static void delayed_work_timer_fn(unsigned long __data) 196static void delayed_work_timer_fn(unsigned long __data)
179{ 197{
@@ -337,14 +355,14 @@ static void wq_barrier_func(struct work_struct *work)
337} 355}
338 356
339static void insert_wq_barrier(struct cpu_workqueue_struct *cwq, 357static void insert_wq_barrier(struct cpu_workqueue_struct *cwq,
340 struct wq_barrier *barr, int tail) 358 struct wq_barrier *barr, struct list_head *head)
341{ 359{
342 INIT_WORK(&barr->work, wq_barrier_func); 360 INIT_WORK(&barr->work, wq_barrier_func);
343 __set_bit(WORK_STRUCT_PENDING, work_data_bits(&barr->work)); 361 __set_bit(WORK_STRUCT_PENDING, work_data_bits(&barr->work));
344 362
345 init_completion(&barr->done); 363 init_completion(&barr->done);
346 364
347 insert_work(cwq, &barr->work, tail); 365 insert_work(cwq, &barr->work, head);
348} 366}
349 367
350static int flush_cpu_workqueue(struct cpu_workqueue_struct *cwq) 368static int flush_cpu_workqueue(struct cpu_workqueue_struct *cwq)
@@ -364,7 +382,7 @@ static int flush_cpu_workqueue(struct cpu_workqueue_struct *cwq)
364 active = 0; 382 active = 0;
365 spin_lock_irq(&cwq->lock); 383 spin_lock_irq(&cwq->lock);
366 if (!list_empty(&cwq->worklist) || cwq->current_work != NULL) { 384 if (!list_empty(&cwq->worklist) || cwq->current_work != NULL) {
367 insert_wq_barrier(cwq, &barr, 1); 385 insert_wq_barrier(cwq, &barr, &cwq->worklist);
368 active = 1; 386 active = 1;
369 } 387 }
370 spin_unlock_irq(&cwq->lock); 388 spin_unlock_irq(&cwq->lock);
@@ -397,11 +415,62 @@ void flush_workqueue(struct workqueue_struct *wq)
397 might_sleep(); 415 might_sleep();
398 lock_acquire(&wq->lockdep_map, 0, 0, 0, 2, _THIS_IP_); 416 lock_acquire(&wq->lockdep_map, 0, 0, 0, 2, _THIS_IP_);
399 lock_release(&wq->lockdep_map, 1, _THIS_IP_); 417 lock_release(&wq->lockdep_map, 1, _THIS_IP_);
400 for_each_cpu_mask(cpu, *cpu_map) 418 for_each_cpu_mask_nr(cpu, *cpu_map)
401 flush_cpu_workqueue(per_cpu_ptr(wq->cpu_wq, cpu)); 419 flush_cpu_workqueue(per_cpu_ptr(wq->cpu_wq, cpu));
402} 420}
403EXPORT_SYMBOL_GPL(flush_workqueue); 421EXPORT_SYMBOL_GPL(flush_workqueue);
404 422
423/**
424 * flush_work - block until a work_struct's callback has terminated
425 * @work: the work which is to be flushed
426 *
427 * Returns false if @work has already terminated.
428 *
429 * It is expected that, prior to calling flush_work(), the caller has
430 * arranged for the work to not be requeued, otherwise it doesn't make
431 * sense to use this function.
432 */
433int flush_work(struct work_struct *work)
434{
435 struct cpu_workqueue_struct *cwq;
436 struct list_head *prev;
437 struct wq_barrier barr;
438
439 might_sleep();
440 cwq = get_wq_data(work);
441 if (!cwq)
442 return 0;
443
444 lock_acquire(&cwq->wq->lockdep_map, 0, 0, 0, 2, _THIS_IP_);
445 lock_release(&cwq->wq->lockdep_map, 1, _THIS_IP_);
446
447 prev = NULL;
448 spin_lock_irq(&cwq->lock);
449 if (!list_empty(&work->entry)) {
450 /*
451 * See the comment near try_to_grab_pending()->smp_rmb().
452 * If it was re-queued under us we are not going to wait.
453 */
454 smp_rmb();
455 if (unlikely(cwq != get_wq_data(work)))
456 goto out;
457 prev = &work->entry;
458 } else {
459 if (cwq->current_work != work)
460 goto out;
461 prev = &cwq->worklist;
462 }
463 insert_wq_barrier(cwq, &barr, prev->next);
464out:
465 spin_unlock_irq(&cwq->lock);
466 if (!prev)
467 return 0;
468
469 wait_for_completion(&barr.done);
470 return 1;
471}
472EXPORT_SYMBOL_GPL(flush_work);
473
405/* 474/*
406 * Upon a successful return (>= 0), the caller "owns" WORK_STRUCT_PENDING bit, 475 * Upon a successful return (>= 0), the caller "owns" WORK_STRUCT_PENDING bit,
407 * so this work can't be re-armed in any way. 476 * so this work can't be re-armed in any way.
@@ -449,7 +518,7 @@ static void wait_on_cpu_work(struct cpu_workqueue_struct *cwq,
449 518
450 spin_lock_irq(&cwq->lock); 519 spin_lock_irq(&cwq->lock);
451 if (unlikely(cwq->current_work == work)) { 520 if (unlikely(cwq->current_work == work)) {
452 insert_wq_barrier(cwq, &barr, 0); 521 insert_wq_barrier(cwq, &barr, cwq->worklist.next);
453 running = 1; 522 running = 1;
454 } 523 }
455 spin_unlock_irq(&cwq->lock); 524 spin_unlock_irq(&cwq->lock);
@@ -477,7 +546,7 @@ static void wait_on_work(struct work_struct *work)
477 wq = cwq->wq; 546 wq = cwq->wq;
478 cpu_map = wq_cpu_map(wq); 547 cpu_map = wq_cpu_map(wq);
479 548
480 for_each_cpu_mask(cpu, *cpu_map) 549 for_each_cpu_mask_nr(cpu, *cpu_map)
481 wait_on_cpu_work(per_cpu_ptr(wq->cpu_wq, cpu), work); 550 wait_on_cpu_work(per_cpu_ptr(wq->cpu_wq, cpu), work);
482} 551}
483 552
@@ -553,6 +622,19 @@ int schedule_work(struct work_struct *work)
553} 622}
554EXPORT_SYMBOL(schedule_work); 623EXPORT_SYMBOL(schedule_work);
555 624
625/*
626 * schedule_work_on - put work task on a specific cpu
627 * @cpu: cpu to put the work task on
628 * @work: job to be done
629 *
630 * This puts a job on a specific cpu
631 */
632int schedule_work_on(int cpu, struct work_struct *work)
633{
634 return queue_work_on(cpu, keventd_wq, work);
635}
636EXPORT_SYMBOL(schedule_work_on);
637
556/** 638/**
557 * schedule_delayed_work - put work task in global workqueue after delay 639 * schedule_delayed_work - put work task in global workqueue after delay
558 * @dwork: job to be done 640 * @dwork: job to be done
@@ -607,10 +689,10 @@ int schedule_on_each_cpu(work_func_t func)
607 struct work_struct *work = per_cpu_ptr(works, cpu); 689 struct work_struct *work = per_cpu_ptr(works, cpu);
608 690
609 INIT_WORK(work, func); 691 INIT_WORK(work, func);
610 set_bit(WORK_STRUCT_PENDING, work_data_bits(work)); 692 schedule_work_on(cpu, work);
611 __queue_work(per_cpu_ptr(keventd_wq->cpu_wq, cpu), work);
612 } 693 }
613 flush_workqueue(keventd_wq); 694 for_each_online_cpu(cpu)
695 flush_work(per_cpu_ptr(works, cpu));
614 put_online_cpus(); 696 put_online_cpus();
615 free_percpu(works); 697 free_percpu(works);
616 return 0; 698 return 0;
@@ -747,7 +829,7 @@ struct workqueue_struct *__create_workqueue_key(const char *name,
747 err = create_workqueue_thread(cwq, singlethread_cpu); 829 err = create_workqueue_thread(cwq, singlethread_cpu);
748 start_workqueue_thread(cwq, -1); 830 start_workqueue_thread(cwq, -1);
749 } else { 831 } else {
750 get_online_cpus(); 832 cpu_maps_update_begin();
751 spin_lock(&workqueue_lock); 833 spin_lock(&workqueue_lock);
752 list_add(&wq->list, &workqueues); 834 list_add(&wq->list, &workqueues);
753 spin_unlock(&workqueue_lock); 835 spin_unlock(&workqueue_lock);
@@ -759,7 +841,7 @@ struct workqueue_struct *__create_workqueue_key(const char *name,
759 err = create_workqueue_thread(cwq, cpu); 841 err = create_workqueue_thread(cwq, cpu);
760 start_workqueue_thread(cwq, cpu); 842 start_workqueue_thread(cwq, cpu);
761 } 843 }
762 put_online_cpus(); 844 cpu_maps_update_done();
763 } 845 }
764 846
765 if (err) { 847 if (err) {
@@ -773,8 +855,8 @@ EXPORT_SYMBOL_GPL(__create_workqueue_key);
773static void cleanup_workqueue_thread(struct cpu_workqueue_struct *cwq) 855static void cleanup_workqueue_thread(struct cpu_workqueue_struct *cwq)
774{ 856{
775 /* 857 /*
776 * Our caller is either destroy_workqueue() or CPU_DEAD, 858 * Our caller is either destroy_workqueue() or CPU_POST_DEAD,
777 * get_online_cpus() protects cwq->thread. 859 * cpu_add_remove_lock protects cwq->thread.
778 */ 860 */
779 if (cwq->thread == NULL) 861 if (cwq->thread == NULL)
780 return; 862 return;
@@ -784,7 +866,7 @@ static void cleanup_workqueue_thread(struct cpu_workqueue_struct *cwq)
784 866
785 flush_cpu_workqueue(cwq); 867 flush_cpu_workqueue(cwq);
786 /* 868 /*
787 * If the caller is CPU_DEAD and cwq->worklist was not empty, 869 * If the caller is CPU_POST_DEAD and cwq->worklist was not empty,
788 * a concurrent flush_workqueue() can insert a barrier after us. 870 * a concurrent flush_workqueue() can insert a barrier after us.
789 * However, in that case run_workqueue() won't return and check 871 * However, in that case run_workqueue() won't return and check
790 * kthread_should_stop() until it flushes all work_struct's. 872 * kthread_should_stop() until it flushes all work_struct's.
@@ -808,14 +890,14 @@ void destroy_workqueue(struct workqueue_struct *wq)
808 const cpumask_t *cpu_map = wq_cpu_map(wq); 890 const cpumask_t *cpu_map = wq_cpu_map(wq);
809 int cpu; 891 int cpu;
810 892
811 get_online_cpus(); 893 cpu_maps_update_begin();
812 spin_lock(&workqueue_lock); 894 spin_lock(&workqueue_lock);
813 list_del(&wq->list); 895 list_del(&wq->list);
814 spin_unlock(&workqueue_lock); 896 spin_unlock(&workqueue_lock);
815 897
816 for_each_cpu_mask(cpu, *cpu_map) 898 for_each_cpu_mask_nr(cpu, *cpu_map)
817 cleanup_workqueue_thread(per_cpu_ptr(wq->cpu_wq, cpu)); 899 cleanup_workqueue_thread(per_cpu_ptr(wq->cpu_wq, cpu));
818 put_online_cpus(); 900 cpu_maps_update_done();
819 901
820 free_percpu(wq->cpu_wq); 902 free_percpu(wq->cpu_wq);
821 kfree(wq); 903 kfree(wq);
@@ -829,6 +911,7 @@ static int __devinit workqueue_cpu_callback(struct notifier_block *nfb,
829 unsigned int cpu = (unsigned long)hcpu; 911 unsigned int cpu = (unsigned long)hcpu;
830 struct cpu_workqueue_struct *cwq; 912 struct cpu_workqueue_struct *cwq;
831 struct workqueue_struct *wq; 913 struct workqueue_struct *wq;
914 int ret = NOTIFY_OK;
832 915
833 action &= ~CPU_TASKS_FROZEN; 916 action &= ~CPU_TASKS_FROZEN;
834 917
@@ -836,7 +919,7 @@ static int __devinit workqueue_cpu_callback(struct notifier_block *nfb,
836 case CPU_UP_PREPARE: 919 case CPU_UP_PREPARE:
837 cpu_set(cpu, cpu_populated_map); 920 cpu_set(cpu, cpu_populated_map);
838 } 921 }
839 922undo:
840 list_for_each_entry(wq, &workqueues, list) { 923 list_for_each_entry(wq, &workqueues, list) {
841 cwq = per_cpu_ptr(wq->cpu_wq, cpu); 924 cwq = per_cpu_ptr(wq->cpu_wq, cpu);
842 925
@@ -846,7 +929,9 @@ static int __devinit workqueue_cpu_callback(struct notifier_block *nfb,
846 break; 929 break;
847 printk(KERN_ERR "workqueue [%s] for %i failed\n", 930 printk(KERN_ERR "workqueue [%s] for %i failed\n",
848 wq->name, cpu); 931 wq->name, cpu);
849 return NOTIFY_BAD; 932 action = CPU_UP_CANCELED;
933 ret = NOTIFY_BAD;
934 goto undo;
850 935
851 case CPU_ONLINE: 936 case CPU_ONLINE:
852 start_workqueue_thread(cwq, cpu); 937 start_workqueue_thread(cwq, cpu);
@@ -854,7 +939,7 @@ static int __devinit workqueue_cpu_callback(struct notifier_block *nfb,
854 939
855 case CPU_UP_CANCELED: 940 case CPU_UP_CANCELED:
856 start_workqueue_thread(cwq, -1); 941 start_workqueue_thread(cwq, -1);
857 case CPU_DEAD: 942 case CPU_POST_DEAD:
858 cleanup_workqueue_thread(cwq); 943 cleanup_workqueue_thread(cwq);
859 break; 944 break;
860 } 945 }
@@ -862,11 +947,11 @@ static int __devinit workqueue_cpu_callback(struct notifier_block *nfb,
862 947
863 switch (action) { 948 switch (action) {
864 case CPU_UP_CANCELED: 949 case CPU_UP_CANCELED:
865 case CPU_DEAD: 950 case CPU_POST_DEAD:
866 cpu_clear(cpu, cpu_populated_map); 951 cpu_clear(cpu, cpu_populated_map);
867 } 952 }
868 953
869 return NOTIFY_OK; 954 return ret;
870} 955}
871 956
872void __init init_workqueues(void) 957void __init init_workqueues(void)
generated by cgit v1.2.2 (git 2.25.0) at 2025-02-26 05:41:47 -0500