aboutsummaryrefslogtreecommitdiffstats
path: root/kernel
diff options
context:
space:
mode:
authorIngo Molnar <mingo@elte.hu>2009-10-13 03:31:28 -0400
committerIngo Molnar <mingo@elte.hu>2009-10-13 03:31:34 -0400
commit9dbdd6c41c12fb42ee7188eafa7e1917b192af3a (patch)
tree06a9eb894bc976c5c20e84ccd74fd82b9b71aed4 /kernel
parent7a693d3f0d10f978ebdf3082c41404ab97106567 (diff)
parent161291396e76e0832c08f617eb9bd364d1648148 (diff)
Merge commit 'v2.6.32-rc4' into perf/core
Merge reason: we were on an -rc1 base, merge up to -rc4. Signed-off-by: Ingo Molnar <mingo@elte.hu>
Diffstat (limited to 'kernel')
-rw-r--r--kernel/cgroup.c15
-rw-r--r--kernel/exit.c2
-rw-r--r--kernel/fork.c10
-rw-r--r--kernel/futex.c3
-rw-r--r--kernel/hrtimer.c2
-rw-r--r--kernel/irq/handle.c1
-rw-r--r--kernel/kprobes.c4
-rw-r--r--kernel/module.c7
-rw-r--r--kernel/mutex-debug.c1
-rw-r--r--kernel/panic.c3
-rw-r--r--kernel/rcupdate.c140
-rw-r--r--kernel/rcutorture.c4
-rw-r--r--kernel/rcutree.c330
-rw-r--r--kernel/rcutree.h86
-rw-r--r--kernel/rcutree_plugin.h103
-rw-r--r--kernel/rcutree_trace.c14
-rw-r--r--kernel/res_counter.c18
-rw-r--r--kernel/sched.c22
-rw-r--r--kernel/sched_clock.c4
-rw-r--r--kernel/time/tick-sched.c9
-rw-r--r--kernel/time/timekeeping.c1
-rw-r--r--kernel/time/timer_list.c2
-rw-r--r--kernel/time/timer_stats.c2
-rw-r--r--kernel/trace/blktrace.c39
-rw-r--r--kernel/trace/ftrace.c27
-rw-r--r--kernel/trace/kmemtrace.c2
-rw-r--r--kernel/trace/trace_branch.c8
-rw-r--r--kernel/trace/trace_event_profile.c15
-rw-r--r--kernel/trace/trace_hw_branches.c8
-rw-r--r--kernel/trace/trace_output.c18
30 files changed, 513 insertions, 387 deletions
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 7ccba4bc5e3b..ca83b73fba19 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -703,7 +703,7 @@ static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, int mode);
703static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry); 703static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry);
704static int cgroup_populate_dir(struct cgroup *cgrp); 704static int cgroup_populate_dir(struct cgroup *cgrp);
705static const struct inode_operations cgroup_dir_inode_operations; 705static const struct inode_operations cgroup_dir_inode_operations;
706static struct file_operations proc_cgroupstats_operations; 706static const struct file_operations proc_cgroupstats_operations;
707 707
708static struct backing_dev_info cgroup_backing_dev_info = { 708static struct backing_dev_info cgroup_backing_dev_info = {
709 .name = "cgroup", 709 .name = "cgroup",
@@ -1863,7 +1863,7 @@ static int cgroup_seqfile_release(struct inode *inode, struct file *file)
1863 return single_release(inode, file); 1863 return single_release(inode, file);
1864} 1864}
1865 1865
1866static struct file_operations cgroup_seqfile_operations = { 1866static const struct file_operations cgroup_seqfile_operations = {
1867 .read = seq_read, 1867 .read = seq_read,
1868 .write = cgroup_file_write, 1868 .write = cgroup_file_write,
1869 .llseek = seq_lseek, 1869 .llseek = seq_lseek,
@@ -1922,7 +1922,7 @@ static int cgroup_rename(struct inode *old_dir, struct dentry *old_dentry,
1922 return simple_rename(old_dir, old_dentry, new_dir, new_dentry); 1922 return simple_rename(old_dir, old_dentry, new_dir, new_dentry);
1923} 1923}
1924 1924
1925static struct file_operations cgroup_file_operations = { 1925static const struct file_operations cgroup_file_operations = {
1926 .read = cgroup_file_read, 1926 .read = cgroup_file_read,
1927 .write = cgroup_file_write, 1927 .write = cgroup_file_write,
1928 .llseek = generic_file_llseek, 1928 .llseek = generic_file_llseek,
@@ -3369,7 +3369,7 @@ static int cgroup_open(struct inode *inode, struct file *file)
3369 return single_open(file, proc_cgroup_show, pid); 3369 return single_open(file, proc_cgroup_show, pid);
3370} 3370}
3371 3371
3372struct file_operations proc_cgroup_operations = { 3372const struct file_operations proc_cgroup_operations = {
3373 .open = cgroup_open, 3373 .open = cgroup_open,
3374 .read = seq_read, 3374 .read = seq_read,
3375 .llseek = seq_lseek, 3375 .llseek = seq_lseek,
@@ -3398,7 +3398,7 @@ static int cgroupstats_open(struct inode *inode, struct file *file)
3398 return single_open(file, proc_cgroupstats_show, NULL); 3398 return single_open(file, proc_cgroupstats_show, NULL);
3399} 3399}
3400 3400
3401static struct file_operations proc_cgroupstats_operations = { 3401static const struct file_operations proc_cgroupstats_operations = {
3402 .open = cgroupstats_open, 3402 .open = cgroupstats_open,
3403 .read = seq_read, 3403 .read = seq_read,
3404 .llseek = seq_lseek, 3404 .llseek = seq_lseek,
@@ -3708,8 +3708,10 @@ static void check_for_release(struct cgroup *cgrp)
3708void __css_put(struct cgroup_subsys_state *css) 3708void __css_put(struct cgroup_subsys_state *css)
3709{ 3709{
3710 struct cgroup *cgrp = css->cgroup; 3710 struct cgroup *cgrp = css->cgroup;
3711 int val;
3711 rcu_read_lock(); 3712 rcu_read_lock();
3712 if (atomic_dec_return(&css->refcnt) == 1) { 3713 val = atomic_dec_return(&css->refcnt);
3714 if (val == 1) {
3713 if (notify_on_release(cgrp)) { 3715 if (notify_on_release(cgrp)) {
3714 set_bit(CGRP_RELEASABLE, &cgrp->flags); 3716 set_bit(CGRP_RELEASABLE, &cgrp->flags);
3715 check_for_release(cgrp); 3717 check_for_release(cgrp);
@@ -3717,6 +3719,7 @@ void __css_put(struct cgroup_subsys_state *css)
3717 cgroup_wakeup_rmdir_waiter(cgrp); 3719 cgroup_wakeup_rmdir_waiter(cgrp);
3718 } 3720 }
3719 rcu_read_unlock(); 3721 rcu_read_unlock();
3722 WARN_ON_ONCE(val < 1);
3720} 3723}
3721 3724
3722/* 3725/*
diff --git a/kernel/exit.c b/kernel/exit.c
index 5859f598c951..e61891f80123 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -991,8 +991,6 @@ NORET_TYPE void do_exit(long code)
991 tsk->mempolicy = NULL; 991 tsk->mempolicy = NULL;
992#endif 992#endif
993#ifdef CONFIG_FUTEX 993#ifdef CONFIG_FUTEX
994 if (unlikely(!list_empty(&tsk->pi_state_list)))
995 exit_pi_state_list(tsk);
996 if (unlikely(current->pi_state_cache)) 994 if (unlikely(current->pi_state_cache))
997 kfree(current->pi_state_cache); 995 kfree(current->pi_state_cache);
998#endif 996#endif
diff --git a/kernel/fork.c b/kernel/fork.c
index 266c6af6ef1b..4c20fff8c13a 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -570,12 +570,18 @@ void mm_release(struct task_struct *tsk, struct mm_struct *mm)
570 570
571 /* Get rid of any futexes when releasing the mm */ 571 /* Get rid of any futexes when releasing the mm */
572#ifdef CONFIG_FUTEX 572#ifdef CONFIG_FUTEX
573 if (unlikely(tsk->robust_list)) 573 if (unlikely(tsk->robust_list)) {
574 exit_robust_list(tsk); 574 exit_robust_list(tsk);
575 tsk->robust_list = NULL;
576 }
575#ifdef CONFIG_COMPAT 577#ifdef CONFIG_COMPAT
576 if (unlikely(tsk->compat_robust_list)) 578 if (unlikely(tsk->compat_robust_list)) {
577 compat_exit_robust_list(tsk); 579 compat_exit_robust_list(tsk);
580 tsk->compat_robust_list = NULL;
581 }
578#endif 582#endif
583 if (unlikely(!list_empty(&tsk->pi_state_list)))
584 exit_pi_state_list(tsk);
579#endif 585#endif
580 586
581 /* Get rid of any cached register state */ 587 /* Get rid of any cached register state */
diff --git a/kernel/futex.c b/kernel/futex.c
index b911adceb2c4..4949d336d88d 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -916,8 +916,8 @@ retry:
916 hb1 = hash_futex(&key1); 916 hb1 = hash_futex(&key1);
917 hb2 = hash_futex(&key2); 917 hb2 = hash_futex(&key2);
918 918
919 double_lock_hb(hb1, hb2);
920retry_private: 919retry_private:
920 double_lock_hb(hb1, hb2);
921 op_ret = futex_atomic_op_inuser(op, uaddr2); 921 op_ret = futex_atomic_op_inuser(op, uaddr2);
922 if (unlikely(op_ret < 0)) { 922 if (unlikely(op_ret < 0)) {
923 923
@@ -2117,7 +2117,6 @@ int handle_early_requeue_pi_wakeup(struct futex_hash_bucket *hb,
2117 * Unqueue the futex_q and determine which it was. 2117 * Unqueue the futex_q and determine which it was.
2118 */ 2118 */
2119 plist_del(&q->list, &q->list.plist); 2119 plist_del(&q->list, &q->list.plist);
2120 drop_futex_key_refs(&q->key);
2121 2120
2122 if (timeout && !timeout->task) 2121 if (timeout && !timeout->task)
2123 ret = -ETIMEDOUT; 2122 ret = -ETIMEDOUT;
diff --git a/kernel/hrtimer.c b/kernel/hrtimer.c
index 6d7020490f94..3e1c36e7998f 100644
--- a/kernel/hrtimer.c
+++ b/kernel/hrtimer.c
@@ -726,8 +726,6 @@ static int hrtimer_switch_to_hres(void)
726 /* "Retrigger" the interrupt to get things going */ 726 /* "Retrigger" the interrupt to get things going */
727 retrigger_next_event(NULL); 727 retrigger_next_event(NULL);
728 local_irq_restore(flags); 728 local_irq_restore(flags);
729 printk(KERN_DEBUG "Switched to high resolution mode on CPU %d\n",
730 smp_processor_id());
731 return 1; 729 return 1;
732} 730}
733 731
diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c
index a81cf80554db..17c71bb565c6 100644
--- a/kernel/irq/handle.c
+++ b/kernel/irq/handle.c
@@ -11,6 +11,7 @@
11 */ 11 */
12 12
13#include <linux/irq.h> 13#include <linux/irq.h>
14#include <linux/sched.h>
14#include <linux/slab.h> 15#include <linux/slab.h>
15#include <linux/module.h> 16#include <linux/module.h>
16#include <linux/random.h> 17#include <linux/random.h>
diff --git a/kernel/kprobes.c b/kernel/kprobes.c
index cfadc1291d0b..5240d75f4c60 100644
--- a/kernel/kprobes.c
+++ b/kernel/kprobes.c
@@ -1333,7 +1333,7 @@ static int __kprobes kprobes_open(struct inode *inode, struct file *filp)
1333 return seq_open(filp, &kprobes_seq_ops); 1333 return seq_open(filp, &kprobes_seq_ops);
1334} 1334}
1335 1335
1336static struct file_operations debugfs_kprobes_operations = { 1336static const struct file_operations debugfs_kprobes_operations = {
1337 .open = kprobes_open, 1337 .open = kprobes_open,
1338 .read = seq_read, 1338 .read = seq_read,
1339 .llseek = seq_lseek, 1339 .llseek = seq_lseek,
@@ -1515,7 +1515,7 @@ static ssize_t write_enabled_file_bool(struct file *file,
1515 return count; 1515 return count;
1516} 1516}
1517 1517
1518static struct file_operations fops_kp = { 1518static const struct file_operations fops_kp = {
1519 .read = read_enabled_file_bool, 1519 .read = read_enabled_file_bool,
1520 .write = write_enabled_file_bool, 1520 .write = write_enabled_file_bool,
1521}; 1521};
diff --git a/kernel/module.c b/kernel/module.c
index fe748a86d452..8b7d8805819d 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -1992,12 +1992,14 @@ static inline unsigned long layout_symtab(struct module *mod,
1992 Elf_Shdr *sechdrs, 1992 Elf_Shdr *sechdrs,
1993 unsigned int symindex, 1993 unsigned int symindex,
1994 unsigned int strindex, 1994 unsigned int strindex,
1995 const Elf_Hdr *hdr, 1995 const Elf_Ehdr *hdr,
1996 const char *secstrings, 1996 const char *secstrings,
1997 unsigned long *pstroffs, 1997 unsigned long *pstroffs,
1998 unsigned long *strmap) 1998 unsigned long *strmap)
1999{ 1999{
2000 return 0;
2000} 2001}
2002
2001static inline void add_kallsyms(struct module *mod, 2003static inline void add_kallsyms(struct module *mod,
2002 Elf_Shdr *sechdrs, 2004 Elf_Shdr *sechdrs,
2003 unsigned int shnum, 2005 unsigned int shnum,
@@ -2081,9 +2083,8 @@ static noinline struct module *load_module(void __user *umod,
2081 struct module *mod; 2083 struct module *mod;
2082 long err = 0; 2084 long err = 0;
2083 void *percpu = NULL, *ptr = NULL; /* Stops spurious gcc warning */ 2085 void *percpu = NULL, *ptr = NULL; /* Stops spurious gcc warning */
2084#ifdef CONFIG_KALLSYMS
2085 unsigned long symoffs, stroffs, *strmap; 2086 unsigned long symoffs, stroffs, *strmap;
2086#endif 2087
2087 mm_segment_t old_fs; 2088 mm_segment_t old_fs;
2088 2089
2089 DEBUGP("load_module: umod=%p, len=%lu, uargs=%p\n", 2090 DEBUGP("load_module: umod=%p, len=%lu, uargs=%p\n",
diff --git a/kernel/mutex-debug.c b/kernel/mutex-debug.c
index 50d022e5a560..ec815a960b5d 100644
--- a/kernel/mutex-debug.c
+++ b/kernel/mutex-debug.c
@@ -16,6 +16,7 @@
16#include <linux/delay.h> 16#include <linux/delay.h>
17#include <linux/module.h> 17#include <linux/module.h>
18#include <linux/poison.h> 18#include <linux/poison.h>
19#include <linux/sched.h>
19#include <linux/spinlock.h> 20#include <linux/spinlock.h>
20#include <linux/kallsyms.h> 21#include <linux/kallsyms.h>
21#include <linux/interrupt.h> 22#include <linux/interrupt.h>
diff --git a/kernel/panic.c b/kernel/panic.c
index bcdef26e3332..96b45d0b4ba5 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -90,6 +90,8 @@ NORET_TYPE void panic(const char * fmt, ...)
90 90
91 atomic_notifier_call_chain(&panic_notifier_list, 0, buf); 91 atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
92 92
93 bust_spinlocks(0);
94
93 if (!panic_blink) 95 if (!panic_blink)
94 panic_blink = no_blink; 96 panic_blink = no_blink;
95 97
@@ -136,7 +138,6 @@ NORET_TYPE void panic(const char * fmt, ...)
136 mdelay(1); 138 mdelay(1);
137 i++; 139 i++;
138 } 140 }
139 bust_spinlocks(0);
140} 141}
141 142
142EXPORT_SYMBOL(panic); 143EXPORT_SYMBOL(panic);
diff --git a/kernel/rcupdate.c b/kernel/rcupdate.c
index 37ac45483082..400183346ad2 100644
--- a/kernel/rcupdate.c
+++ b/kernel/rcupdate.c
@@ -46,22 +46,15 @@
46#include <linux/module.h> 46#include <linux/module.h>
47#include <linux/kernel_stat.h> 47#include <linux/kernel_stat.h>
48 48
49enum rcu_barrier { 49#ifdef CONFIG_DEBUG_LOCK_ALLOC
50 RCU_BARRIER_STD, 50static struct lock_class_key rcu_lock_key;
51 RCU_BARRIER_BH, 51struct lockdep_map rcu_lock_map =
52 RCU_BARRIER_SCHED, 52 STATIC_LOCKDEP_MAP_INIT("rcu_read_lock", &rcu_lock_key);
53}; 53EXPORT_SYMBOL_GPL(rcu_lock_map);
54#endif
54 55
55static DEFINE_PER_CPU(struct rcu_head, rcu_barrier_head) = {NULL};
56static atomic_t rcu_barrier_cpu_count;
57static DEFINE_MUTEX(rcu_barrier_mutex);
58static struct completion rcu_barrier_completion;
59int rcu_scheduler_active __read_mostly; 56int rcu_scheduler_active __read_mostly;
60 57
61static atomic_t rcu_migrate_type_count = ATOMIC_INIT(0);
62static struct rcu_head rcu_migrate_head[3];
63static DECLARE_WAIT_QUEUE_HEAD(rcu_migrate_wq);
64
65/* 58/*
66 * Awaken the corresponding synchronize_rcu() instance now that a 59 * Awaken the corresponding synchronize_rcu() instance now that a
67 * grace period has elapsed. 60 * grace period has elapsed.
@@ -164,129 +157,10 @@ void synchronize_rcu_bh(void)
164} 157}
165EXPORT_SYMBOL_GPL(synchronize_rcu_bh); 158EXPORT_SYMBOL_GPL(synchronize_rcu_bh);
166 159
167static void rcu_barrier_callback(struct rcu_head *notused)
168{
169 if (atomic_dec_and_test(&rcu_barrier_cpu_count))
170 complete(&rcu_barrier_completion);
171}
172
173/*
174 * Called with preemption disabled, and from cross-cpu IRQ context.
175 */
176static void rcu_barrier_func(void *type)
177{
178 int cpu = smp_processor_id();
179 struct rcu_head *head = &per_cpu(rcu_barrier_head, cpu);
180
181 atomic_inc(&rcu_barrier_cpu_count);
182 switch ((enum rcu_barrier)type) {
183 case RCU_BARRIER_STD:
184 call_rcu(head, rcu_barrier_callback);
185 break;
186 case RCU_BARRIER_BH:
187 call_rcu_bh(head, rcu_barrier_callback);
188 break;
189 case RCU_BARRIER_SCHED:
190 call_rcu_sched(head, rcu_barrier_callback);
191 break;
192 }
193}
194
195static inline void wait_migrated_callbacks(void)
196{
197 wait_event(rcu_migrate_wq, !atomic_read(&rcu_migrate_type_count));
198 smp_mb(); /* In case we didn't sleep. */
199}
200
201/*
202 * Orchestrate the specified type of RCU barrier, waiting for all
203 * RCU callbacks of the specified type to complete.
204 */
205static void _rcu_barrier(enum rcu_barrier type)
206{
207 BUG_ON(in_interrupt());
208 /* Take cpucontrol mutex to protect against CPU hotplug */
209 mutex_lock(&rcu_barrier_mutex);
210 init_completion(&rcu_barrier_completion);
211 /*
212 * Initialize rcu_barrier_cpu_count to 1, then invoke
213 * rcu_barrier_func() on each CPU, so that each CPU also has
214 * incremented rcu_barrier_cpu_count. Only then is it safe to
215 * decrement rcu_barrier_cpu_count -- otherwise the first CPU
216 * might complete its grace period before all of the other CPUs
217 * did their increment, causing this function to return too
218 * early.
219 */
220 atomic_set(&rcu_barrier_cpu_count, 1);
221 on_each_cpu(rcu_barrier_func, (void *)type, 1);
222 if (atomic_dec_and_test(&rcu_barrier_cpu_count))
223 complete(&rcu_barrier_completion);
224 wait_for_completion(&rcu_barrier_completion);
225 mutex_unlock(&rcu_barrier_mutex);
226 wait_migrated_callbacks();
227}
228
229/**
230 * rcu_barrier - Wait until all in-flight call_rcu() callbacks complete.
231 */
232void rcu_barrier(void)
233{
234 _rcu_barrier(RCU_BARRIER_STD);
235}
236EXPORT_SYMBOL_GPL(rcu_barrier);
237
238/**
239 * rcu_barrier_bh - Wait until all in-flight call_rcu_bh() callbacks complete.
240 */
241void rcu_barrier_bh(void)
242{
243 _rcu_barrier(RCU_BARRIER_BH);
244}
245EXPORT_SYMBOL_GPL(rcu_barrier_bh);
246
247/**
248 * rcu_barrier_sched - Wait for in-flight call_rcu_sched() callbacks.
249 */
250void rcu_barrier_sched(void)
251{
252 _rcu_barrier(RCU_BARRIER_SCHED);
253}
254EXPORT_SYMBOL_GPL(rcu_barrier_sched);
255
256static void rcu_migrate_callback(struct rcu_head *notused)
257{
258 if (atomic_dec_and_test(&rcu_migrate_type_count))
259 wake_up(&rcu_migrate_wq);
260}
261
262extern int rcu_cpu_notify(struct notifier_block *self,
263 unsigned long action, void *hcpu);
264
265static int __cpuinit rcu_barrier_cpu_hotplug(struct notifier_block *self, 160static int __cpuinit rcu_barrier_cpu_hotplug(struct notifier_block *self,
266 unsigned long action, void *hcpu) 161 unsigned long action, void *hcpu)
267{ 162{
268 rcu_cpu_notify(self, action, hcpu); 163 return rcu_cpu_notify(self, action, hcpu);
269 if (action == CPU_DYING) {
270 /*
271 * preempt_disable() in on_each_cpu() prevents stop_machine(),
272 * so when "on_each_cpu(rcu_barrier_func, (void *)type, 1);"
273 * returns, all online cpus have queued rcu_barrier_func(),
274 * and the dead cpu(if it exist) queues rcu_migrate_callback()s.
275 *
276 * These callbacks ensure _rcu_barrier() waits for all
277 * RCU callbacks of the specified type to complete.
278 */
279 atomic_set(&rcu_migrate_type_count, 3);
280 call_rcu_bh(rcu_migrate_head, rcu_migrate_callback);
281 call_rcu_sched(rcu_migrate_head + 1, rcu_migrate_callback);
282 call_rcu(rcu_migrate_head + 2, rcu_migrate_callback);
283 } else if (action == CPU_DOWN_PREPARE) {
284 /* Don't need to wait until next removal operation. */
285 /* rcu_migrate_head is protected by cpu_add_remove_lock */
286 wait_migrated_callbacks();
287 }
288
289 return NOTIFY_OK;
290} 164}
291 165
292void __init rcu_init(void) 166void __init rcu_init(void)
diff --git a/kernel/rcutorture.c b/kernel/rcutorture.c
index 233768f21f97..697c0a0229d4 100644
--- a/kernel/rcutorture.c
+++ b/kernel/rcutorture.c
@@ -606,8 +606,6 @@ static struct rcu_torture_ops sched_ops_sync = {
606 .name = "sched_sync" 606 .name = "sched_sync"
607}; 607};
608 608
609extern int rcu_expedited_torture_stats(char *page);
610
611static struct rcu_torture_ops sched_expedited_ops = { 609static struct rcu_torture_ops sched_expedited_ops = {
612 .init = rcu_sync_torture_init, 610 .init = rcu_sync_torture_init,
613 .cleanup = NULL, 611 .cleanup = NULL,
@@ -650,7 +648,7 @@ rcu_torture_writer(void *arg)
650 old_rp = rcu_torture_current; 648 old_rp = rcu_torture_current;
651 rp->rtort_mbtest = 1; 649 rp->rtort_mbtest = 1;
652 rcu_assign_pointer(rcu_torture_current, rp); 650 rcu_assign_pointer(rcu_torture_current, rp);
653 smp_wmb(); 651 smp_wmb(); /* Mods to old_rp must follow rcu_assign_pointer() */
654 if (old_rp) { 652 if (old_rp) {
655 i = old_rp->rtort_pipe_count; 653 i = old_rp->rtort_pipe_count;
656 if (i > RCU_TORTURE_PIPE_LEN) 654 if (i > RCU_TORTURE_PIPE_LEN)
diff --git a/kernel/rcutree.c b/kernel/rcutree.c
index 52b06f6e158c..705f02ac7433 100644
--- a/kernel/rcutree.c
+++ b/kernel/rcutree.c
@@ -49,13 +49,6 @@
49 49
50#include "rcutree.h" 50#include "rcutree.h"
51 51
52#ifdef CONFIG_DEBUG_LOCK_ALLOC
53static struct lock_class_key rcu_lock_key;
54struct lockdep_map rcu_lock_map =
55 STATIC_LOCKDEP_MAP_INIT("rcu_read_lock", &rcu_lock_key);
56EXPORT_SYMBOL_GPL(rcu_lock_map);
57#endif
58
59/* Data structures. */ 52/* Data structures. */
60 53
61#define RCU_STATE_INITIALIZER(name) { \ 54#define RCU_STATE_INITIALIZER(name) { \
@@ -70,6 +63,9 @@ EXPORT_SYMBOL_GPL(rcu_lock_map);
70 .gpnum = -300, \ 63 .gpnum = -300, \
71 .completed = -300, \ 64 .completed = -300, \
72 .onofflock = __SPIN_LOCK_UNLOCKED(&name.onofflock), \ 65 .onofflock = __SPIN_LOCK_UNLOCKED(&name.onofflock), \
66 .orphan_cbs_list = NULL, \
67 .orphan_cbs_tail = &name.orphan_cbs_list, \
68 .orphan_qlen = 0, \
73 .fqslock = __SPIN_LOCK_UNLOCKED(&name.fqslock), \ 69 .fqslock = __SPIN_LOCK_UNLOCKED(&name.fqslock), \
74 .n_force_qs = 0, \ 70 .n_force_qs = 0, \
75 .n_force_qs_ngp = 0, \ 71 .n_force_qs_ngp = 0, \
@@ -81,24 +77,16 @@ DEFINE_PER_CPU(struct rcu_data, rcu_sched_data);
81struct rcu_state rcu_bh_state = RCU_STATE_INITIALIZER(rcu_bh_state); 77struct rcu_state rcu_bh_state = RCU_STATE_INITIALIZER(rcu_bh_state);
82DEFINE_PER_CPU(struct rcu_data, rcu_bh_data); 78DEFINE_PER_CPU(struct rcu_data, rcu_bh_data);
83 79
84extern long rcu_batches_completed_sched(void);
85static struct rcu_node *rcu_get_root(struct rcu_state *rsp);
86static void cpu_quiet_msk(unsigned long mask, struct rcu_state *rsp,
87 struct rcu_node *rnp, unsigned long flags);
88static void cpu_quiet_msk_finish(struct rcu_state *rsp, unsigned long flags);
89#ifdef CONFIG_HOTPLUG_CPU
90static void __rcu_offline_cpu(int cpu, struct rcu_state *rsp);
91#endif /* #ifdef CONFIG_HOTPLUG_CPU */
92static void __rcu_process_callbacks(struct rcu_state *rsp,
93 struct rcu_data *rdp);
94static void __call_rcu(struct rcu_head *head,
95 void (*func)(struct rcu_head *rcu),
96 struct rcu_state *rsp);
97static int __rcu_pending(struct rcu_state *rsp, struct rcu_data *rdp);
98static void __cpuinit rcu_init_percpu_data(int cpu, struct rcu_state *rsp,
99 int preemptable);
100 80
101#include "rcutree_plugin.h" 81/*
82 * Return true if an RCU grace period is in progress. The ACCESS_ONCE()s
83 * permit this function to be invoked without holding the root rcu_node
84 * structure's ->lock, but of course results can be subject to change.
85 */
86static int rcu_gp_in_progress(struct rcu_state *rsp)
87{
88 return ACCESS_ONCE(rsp->completed) != ACCESS_ONCE(rsp->gpnum);
89}
102 90
103/* 91/*
104 * Note a quiescent state. Because we do not need to know 92 * Note a quiescent state. Because we do not need to know
@@ -137,6 +125,10 @@ static int blimit = 10; /* Maximum callbacks per softirq. */
137static int qhimark = 10000; /* If this many pending, ignore blimit. */ 125static int qhimark = 10000; /* If this many pending, ignore blimit. */
138static int qlowmark = 100; /* Once only this many pending, use blimit. */ 126static int qlowmark = 100; /* Once only this many pending, use blimit. */
139 127
128module_param(blimit, int, 0);
129module_param(qhimark, int, 0);
130module_param(qlowmark, int, 0);
131
140static void force_quiescent_state(struct rcu_state *rsp, int relaxed); 132static void force_quiescent_state(struct rcu_state *rsp, int relaxed);
141static int rcu_pending(int cpu); 133static int rcu_pending(int cpu);
142 134
@@ -173,9 +165,7 @@ cpu_has_callbacks_ready_to_invoke(struct rcu_data *rdp)
173static int 165static int
174cpu_needs_another_gp(struct rcu_state *rsp, struct rcu_data *rdp) 166cpu_needs_another_gp(struct rcu_state *rsp, struct rcu_data *rdp)
175{ 167{
176 /* ACCESS_ONCE() because we are accessing outside of lock. */ 168 return *rdp->nxttail[RCU_DONE_TAIL] && !rcu_gp_in_progress(rsp);
177 return *rdp->nxttail[RCU_DONE_TAIL] &&
178 ACCESS_ONCE(rsp->completed) == ACCESS_ONCE(rsp->gpnum);
179} 169}
180 170
181/* 171/*
@@ -369,7 +359,7 @@ static long dyntick_recall_completed(struct rcu_state *rsp)
369/* 359/*
370 * Snapshot the specified CPU's dynticks counter so that we can later 360 * Snapshot the specified CPU's dynticks counter so that we can later
371 * credit them with an implicit quiescent state. Return 1 if this CPU 361 * credit them with an implicit quiescent state. Return 1 if this CPU
372 * is already in a quiescent state courtesy of dynticks idle mode. 362 * is in dynticks idle mode, which is an extended quiescent state.
373 */ 363 */
374static int dyntick_save_progress_counter(struct rcu_data *rdp) 364static int dyntick_save_progress_counter(struct rcu_data *rdp)
375{ 365{
@@ -475,30 +465,34 @@ static void print_other_cpu_stall(struct rcu_state *rsp)
475 long delta; 465 long delta;
476 unsigned long flags; 466 unsigned long flags;
477 struct rcu_node *rnp = rcu_get_root(rsp); 467 struct rcu_node *rnp = rcu_get_root(rsp);
478 struct rcu_node *rnp_cur = rsp->level[NUM_RCU_LVLS - 1];
479 struct rcu_node *rnp_end = &rsp->node[NUM_RCU_NODES];
480 468
481 /* Only let one CPU complain about others per time interval. */ 469 /* Only let one CPU complain about others per time interval. */
482 470
483 spin_lock_irqsave(&rnp->lock, flags); 471 spin_lock_irqsave(&rnp->lock, flags);
484 delta = jiffies - rsp->jiffies_stall; 472 delta = jiffies - rsp->jiffies_stall;
485 if (delta < RCU_STALL_RAT_DELAY || rsp->gpnum == rsp->completed) { 473 if (delta < RCU_STALL_RAT_DELAY || !rcu_gp_in_progress(rsp)) {
486 spin_unlock_irqrestore(&rnp->lock, flags); 474 spin_unlock_irqrestore(&rnp->lock, flags);
487 return; 475 return;
488 } 476 }
489 rsp->jiffies_stall = jiffies + RCU_SECONDS_TILL_STALL_RECHECK; 477 rsp->jiffies_stall = jiffies + RCU_SECONDS_TILL_STALL_RECHECK;
478
479 /*
480 * Now rat on any tasks that got kicked up to the root rcu_node
481 * due to CPU offlining.
482 */
483 rcu_print_task_stall(rnp);
490 spin_unlock_irqrestore(&rnp->lock, flags); 484 spin_unlock_irqrestore(&rnp->lock, flags);
491 485
492 /* OK, time to rat on our buddy... */ 486 /* OK, time to rat on our buddy... */
493 487
494 printk(KERN_ERR "INFO: RCU detected CPU stalls:"); 488 printk(KERN_ERR "INFO: RCU detected CPU stalls:");
495 for (; rnp_cur < rnp_end; rnp_cur++) { 489 rcu_for_each_leaf_node(rsp, rnp) {
496 rcu_print_task_stall(rnp); 490 rcu_print_task_stall(rnp);
497 if (rnp_cur->qsmask == 0) 491 if (rnp->qsmask == 0)
498 continue; 492 continue;
499 for (cpu = 0; cpu <= rnp_cur->grphi - rnp_cur->grplo; cpu++) 493 for (cpu = 0; cpu <= rnp->grphi - rnp->grplo; cpu++)
500 if (rnp_cur->qsmask & (1UL << cpu)) 494 if (rnp->qsmask & (1UL << cpu))
501 printk(" %d", rnp_cur->grplo + cpu); 495 printk(" %d", rnp->grplo + cpu);
502 } 496 }
503 printk(" (detected by %d, t=%ld jiffies)\n", 497 printk(" (detected by %d, t=%ld jiffies)\n",
504 smp_processor_id(), (long)(jiffies - rsp->gp_start)); 498 smp_processor_id(), (long)(jiffies - rsp->gp_start));
@@ -537,8 +531,7 @@ static void check_cpu_stall(struct rcu_state *rsp, struct rcu_data *rdp)
537 /* We haven't checked in, so go dump stack. */ 531 /* We haven't checked in, so go dump stack. */
538 print_cpu_stall(rsp); 532 print_cpu_stall(rsp);
539 533
540 } else if (rsp->gpnum != rsp->completed && 534 } else if (rcu_gp_in_progress(rsp) && delta >= RCU_STALL_RAT_DELAY) {
541 delta >= RCU_STALL_RAT_DELAY) {
542 535
543 /* They had two time units to dump stack, so complain. */ 536 /* They had two time units to dump stack, so complain. */
544 print_other_cpu_stall(rsp); 537 print_other_cpu_stall(rsp);
@@ -617,9 +610,15 @@ rcu_start_gp(struct rcu_state *rsp, unsigned long flags)
617 note_new_gpnum(rsp, rdp); 610 note_new_gpnum(rsp, rdp);
618 611
619 /* 612 /*
620 * Because we are first, we know that all our callbacks will 613 * Because this CPU just now started the new grace period, we know
621 * be covered by this upcoming grace period, even the ones 614 * that all of its callbacks will be covered by this upcoming grace
622 * that were registered arbitrarily recently. 615 * period, even the ones that were registered arbitrarily recently.
616 * Therefore, advance all outstanding callbacks to RCU_WAIT_TAIL.
617 *
618 * Other CPUs cannot be sure exactly when the grace period started.
619 * Therefore, their recently registered callbacks must pass through
620 * an additional RCU_NEXT_READY stage, so that they will be handled
621 * by the next RCU grace period.
623 */ 622 */
624 rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL]; 623 rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];
625 rdp->nxttail[RCU_WAIT_TAIL] = rdp->nxttail[RCU_NEXT_TAIL]; 624 rdp->nxttail[RCU_WAIT_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];
@@ -657,7 +656,7 @@ rcu_start_gp(struct rcu_state *rsp, unsigned long flags)
657 * one corresponding to this CPU, due to the fact that we have 656 * one corresponding to this CPU, due to the fact that we have
658 * irqs disabled. 657 * irqs disabled.
659 */ 658 */
660 for (rnp = &rsp->node[0]; rnp < &rsp->node[NUM_RCU_NODES]; rnp++) { 659 rcu_for_each_node_breadth_first(rsp, rnp) {
661 spin_lock(&rnp->lock); /* irqs already disabled. */ 660 spin_lock(&rnp->lock); /* irqs already disabled. */
662 rcu_preempt_check_blocked_tasks(rnp); 661 rcu_preempt_check_blocked_tasks(rnp);
663 rnp->qsmask = rnp->qsmaskinit; 662 rnp->qsmask = rnp->qsmaskinit;
@@ -703,9 +702,9 @@ rcu_process_gp_end(struct rcu_state *rsp, struct rcu_data *rdp)
703 * hold rnp->lock, as required by rcu_start_gp(), which will release it. 702 * hold rnp->lock, as required by rcu_start_gp(), which will release it.
704 */ 703 */
705static void cpu_quiet_msk_finish(struct rcu_state *rsp, unsigned long flags) 704static void cpu_quiet_msk_finish(struct rcu_state *rsp, unsigned long flags)
706 __releases(rnp->lock) 705 __releases(rcu_get_root(rsp)->lock)
707{ 706{
708 WARN_ON_ONCE(rsp->completed == rsp->gpnum); 707 WARN_ON_ONCE(!rcu_gp_in_progress(rsp));
709 rsp->completed = rsp->gpnum; 708 rsp->completed = rsp->gpnum;
710 rcu_process_gp_end(rsp, rsp->rda[smp_processor_id()]); 709 rcu_process_gp_end(rsp, rsp->rda[smp_processor_id()]);
711 rcu_start_gp(rsp, flags); /* releases root node's rnp->lock. */ 710 rcu_start_gp(rsp, flags); /* releases root node's rnp->lock. */
@@ -842,17 +841,63 @@ rcu_check_quiescent_state(struct rcu_state *rsp, struct rcu_data *rdp)
842#ifdef CONFIG_HOTPLUG_CPU 841#ifdef CONFIG_HOTPLUG_CPU
843 842
844/* 843/*
844 * Move a dying CPU's RCU callbacks to the ->orphan_cbs_list for the
845 * specified flavor of RCU. The callbacks will be adopted by the next
846 * _rcu_barrier() invocation or by the CPU_DEAD notifier, whichever
847 * comes first. Because this is invoked from the CPU_DYING notifier,
848 * irqs are already disabled.
849 */
850static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp)
851{
852 int i;
853 struct rcu_data *rdp = rsp->rda[smp_processor_id()];
854
855 if (rdp->nxtlist == NULL)
856 return; /* irqs disabled, so comparison is stable. */
857 spin_lock(&rsp->onofflock); /* irqs already disabled. */
858 *rsp->orphan_cbs_tail = rdp->nxtlist;
859 rsp->orphan_cbs_tail = rdp->nxttail[RCU_NEXT_TAIL];
860 rdp->nxtlist = NULL;
861 for (i = 0; i < RCU_NEXT_SIZE; i++)
862 rdp->nxttail[i] = &rdp->nxtlist;
863 rsp->orphan_qlen += rdp->qlen;
864 rdp->qlen = 0;
865 spin_unlock(&rsp->onofflock); /* irqs remain disabled. */
866}
867
868/*
869 * Adopt previously orphaned RCU callbacks.
870 */
871static void rcu_adopt_orphan_cbs(struct rcu_state *rsp)
872{
873 unsigned long flags;
874 struct rcu_data *rdp;
875
876 spin_lock_irqsave(&rsp->onofflock, flags);
877 rdp = rsp->rda[smp_processor_id()];
878 if (rsp->orphan_cbs_list == NULL) {
879 spin_unlock_irqrestore(&rsp->onofflock, flags);
880 return;
881 }
882 *rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_list;
883 rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_tail;
884 rdp->qlen += rsp->orphan_qlen;
885 rsp->orphan_cbs_list = NULL;
886 rsp->orphan_cbs_tail = &rsp->orphan_cbs_list;
887 rsp->orphan_qlen = 0;
888 spin_unlock_irqrestore(&rsp->onofflock, flags);
889}
890
891/*
845 * Remove the outgoing CPU from the bitmasks in the rcu_node hierarchy 892 * Remove the outgoing CPU from the bitmasks in the rcu_node hierarchy
846 * and move all callbacks from the outgoing CPU to the current one. 893 * and move all callbacks from the outgoing CPU to the current one.
847 */ 894 */
848static void __rcu_offline_cpu(int cpu, struct rcu_state *rsp) 895static void __rcu_offline_cpu(int cpu, struct rcu_state *rsp)
849{ 896{
850 int i;
851 unsigned long flags; 897 unsigned long flags;
852 long lastcomp; 898 long lastcomp;
853 unsigned long mask; 899 unsigned long mask;
854 struct rcu_data *rdp = rsp->rda[cpu]; 900 struct rcu_data *rdp = rsp->rda[cpu];
855 struct rcu_data *rdp_me;
856 struct rcu_node *rnp; 901 struct rcu_node *rnp;
857 902
858 /* Exclude any attempts to start a new grace period. */ 903 /* Exclude any attempts to start a new grace period. */
@@ -875,32 +920,9 @@ static void __rcu_offline_cpu(int cpu, struct rcu_state *rsp)
875 } while (rnp != NULL); 920 } while (rnp != NULL);
876 lastcomp = rsp->completed; 921 lastcomp = rsp->completed;
877 922
878 spin_unlock(&rsp->onofflock); /* irqs remain disabled. */ 923 spin_unlock_irqrestore(&rsp->onofflock, flags);
879 924
880 /* 925 rcu_adopt_orphan_cbs(rsp);
881 * Move callbacks from the outgoing CPU to the running CPU.
882 * Note that the outgoing CPU is now quiscent, so it is now
883 * (uncharacteristically) safe to access its rcu_data structure.
884 * Note also that we must carefully retain the order of the
885 * outgoing CPU's callbacks in order for rcu_barrier() to work
886 * correctly. Finally, note that we start all the callbacks
887 * afresh, even those that have passed through a grace period
888 * and are therefore ready to invoke. The theory is that hotplug
889 * events are rare, and that if they are frequent enough to
890 * indefinitely delay callbacks, you have far worse things to
891 * be worrying about.
892 */
893 rdp_me = rsp->rda[smp_processor_id()];
894 if (rdp->nxtlist != NULL) {
895 *rdp_me->nxttail[RCU_NEXT_TAIL] = rdp->nxtlist;
896 rdp_me->nxttail[RCU_NEXT_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];
897 rdp->nxtlist = NULL;
898 for (i = 0; i < RCU_NEXT_SIZE; i++)
899 rdp->nxttail[i] = &rdp->nxtlist;
900 rdp_me->qlen += rdp->qlen;
901 rdp->qlen = 0;
902 }
903 local_irq_restore(flags);
904} 926}
905 927
906/* 928/*
@@ -918,6 +940,14 @@ static void rcu_offline_cpu(int cpu)
918 940
919#else /* #ifdef CONFIG_HOTPLUG_CPU */ 941#else /* #ifdef CONFIG_HOTPLUG_CPU */
920 942
943static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp)
944{
945}
946
947static void rcu_adopt_orphan_cbs(struct rcu_state *rsp)
948{
949}
950
921static void rcu_offline_cpu(int cpu) 951static void rcu_offline_cpu(int cpu)
922{ 952{
923} 953}
@@ -1050,33 +1080,32 @@ static int rcu_process_dyntick(struct rcu_state *rsp, long lastcomp,
1050 int cpu; 1080 int cpu;
1051 unsigned long flags; 1081 unsigned long flags;
1052 unsigned long mask; 1082 unsigned long mask;
1053 struct rcu_node *rnp_cur = rsp->level[NUM_RCU_LVLS - 1]; 1083 struct rcu_node *rnp;
1054 struct rcu_node *rnp_end = &rsp->node[NUM_RCU_NODES];
1055 1084
1056 for (; rnp_cur < rnp_end; rnp_cur++) { 1085 rcu_for_each_leaf_node(rsp, rnp) {
1057 mask = 0; 1086 mask = 0;
1058 spin_lock_irqsave(&rnp_cur->lock, flags); 1087 spin_lock_irqsave(&rnp->lock, flags);
1059 if (rsp->completed != lastcomp) { 1088 if (rsp->completed != lastcomp) {
1060 spin_unlock_irqrestore(&rnp_cur->lock, flags); 1089 spin_unlock_irqrestore(&rnp->lock, flags);
1061 return 1; 1090 return 1;
1062 } 1091 }
1063 if (rnp_cur->qsmask == 0) { 1092 if (rnp->qsmask == 0) {
1064 spin_unlock_irqrestore(&rnp_cur->lock, flags); 1093 spin_unlock_irqrestore(&rnp->lock, flags);
1065 continue; 1094 continue;
1066 } 1095 }
1067 cpu = rnp_cur->grplo; 1096 cpu = rnp->grplo;
1068 bit = 1; 1097 bit = 1;
1069 for (; cpu <= rnp_cur->grphi; cpu++, bit <<= 1) { 1098 for (; cpu <= rnp->grphi; cpu++, bit <<= 1) {
1070 if ((rnp_cur->qsmask & bit) != 0 && f(rsp->rda[cpu])) 1099 if ((rnp->qsmask & bit) != 0 && f(rsp->rda[cpu]))
1071 mask |= bit; 1100 mask |= bit;
1072 } 1101 }
1073 if (mask != 0 && rsp->completed == lastcomp) { 1102 if (mask != 0 && rsp->completed == lastcomp) {
1074 1103
1075 /* cpu_quiet_msk() releases rnp_cur->lock. */ 1104 /* cpu_quiet_msk() releases rnp->lock. */
1076 cpu_quiet_msk(mask, rsp, rnp_cur, flags); 1105 cpu_quiet_msk(mask, rsp, rnp, flags);
1077 continue; 1106 continue;
1078 } 1107 }
1079 spin_unlock_irqrestore(&rnp_cur->lock, flags); 1108 spin_unlock_irqrestore(&rnp->lock, flags);
1080 } 1109 }
1081 return 0; 1110 return 0;
1082} 1111}
@@ -1092,7 +1121,7 @@ static void force_quiescent_state(struct rcu_state *rsp, int relaxed)
1092 struct rcu_node *rnp = rcu_get_root(rsp); 1121 struct rcu_node *rnp = rcu_get_root(rsp);
1093 u8 signaled; 1122 u8 signaled;
1094 1123
1095 if (ACCESS_ONCE(rsp->completed) == ACCESS_ONCE(rsp->gpnum)) 1124 if (!rcu_gp_in_progress(rsp))
1096 return; /* No grace period in progress, nothing to force. */ 1125 return; /* No grace period in progress, nothing to force. */
1097 if (!spin_trylock_irqsave(&rsp->fqslock, flags)) { 1126 if (!spin_trylock_irqsave(&rsp->fqslock, flags)) {
1098 rsp->n_force_qs_lh++; /* Inexact, can lose counts. Tough! */ 1127 rsp->n_force_qs_lh++; /* Inexact, can lose counts. Tough! */
@@ -1251,7 +1280,7 @@ __call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu),
1251 rdp->nxttail[RCU_NEXT_TAIL] = &head->next; 1280 rdp->nxttail[RCU_NEXT_TAIL] = &head->next;
1252 1281
1253 /* Start a new grace period if one not already started. */ 1282 /* Start a new grace period if one not already started. */
1254 if (ACCESS_ONCE(rsp->completed) == ACCESS_ONCE(rsp->gpnum)) { 1283 if (!rcu_gp_in_progress(rsp)) {
1255 unsigned long nestflag; 1284 unsigned long nestflag;
1256 struct rcu_node *rnp_root = rcu_get_root(rsp); 1285 struct rcu_node *rnp_root = rcu_get_root(rsp);
1257 1286
@@ -1331,7 +1360,7 @@ static int __rcu_pending(struct rcu_state *rsp, struct rcu_data *rdp)
1331 } 1360 }
1332 1361
1333 /* Has an RCU GP gone long enough to send resched IPIs &c? */ 1362 /* Has an RCU GP gone long enough to send resched IPIs &c? */
1334 if (ACCESS_ONCE(rsp->completed) != ACCESS_ONCE(rsp->gpnum) && 1363 if (rcu_gp_in_progress(rsp) &&
1335 ((long)(ACCESS_ONCE(rsp->jiffies_force_qs) - jiffies) < 0)) { 1364 ((long)(ACCESS_ONCE(rsp->jiffies_force_qs) - jiffies) < 0)) {
1336 rdp->n_rp_need_fqs++; 1365 rdp->n_rp_need_fqs++;
1337 return 1; 1366 return 1;
@@ -1368,6 +1397,82 @@ int rcu_needs_cpu(int cpu)
1368 rcu_preempt_needs_cpu(cpu); 1397 rcu_preempt_needs_cpu(cpu);
1369} 1398}
1370 1399
1400static DEFINE_PER_CPU(struct rcu_head, rcu_barrier_head) = {NULL};
1401static atomic_t rcu_barrier_cpu_count;
1402static DEFINE_MUTEX(rcu_barrier_mutex);
1403static struct completion rcu_barrier_completion;
1404
1405static void rcu_barrier_callback(struct rcu_head *notused)
1406{
1407 if (atomic_dec_and_test(&rcu_barrier_cpu_count))
1408 complete(&rcu_barrier_completion);
1409}
1410
1411/*
1412 * Called with preemption disabled, and from cross-cpu IRQ context.
1413 */
1414static void rcu_barrier_func(void *type)
1415{
1416 int cpu = smp_processor_id();
1417 struct rcu_head *head = &per_cpu(rcu_barrier_head, cpu);
1418 void (*call_rcu_func)(struct rcu_head *head,
1419 void (*func)(struct rcu_head *head));
1420
1421 atomic_inc(&rcu_barrier_cpu_count);
1422 call_rcu_func = type;
1423 call_rcu_func(head, rcu_barrier_callback);
1424}
1425
1426/*
1427 * Orchestrate the specified type of RCU barrier, waiting for all
1428 * RCU callbacks of the specified type to complete.
1429 */
1430static void _rcu_barrier(struct rcu_state *rsp,
1431 void (*call_rcu_func)(struct rcu_head *head,
1432 void (*func)(struct rcu_head *head)))
1433{
1434 BUG_ON(in_interrupt());
1435 /* Take mutex to serialize concurrent rcu_barrier() requests. */
1436 mutex_lock(&rcu_barrier_mutex);
1437 init_completion(&rcu_barrier_completion);
1438 /*
1439 * Initialize rcu_barrier_cpu_count to 1, then invoke
1440 * rcu_barrier_func() on each CPU, so that each CPU also has
1441 * incremented rcu_barrier_cpu_count. Only then is it safe to
1442 * decrement rcu_barrier_cpu_count -- otherwise the first CPU
1443 * might complete its grace period before all of the other CPUs
1444 * did their increment, causing this function to return too
1445 * early.
1446 */
1447 atomic_set(&rcu_barrier_cpu_count, 1);
1448 preempt_disable(); /* stop CPU_DYING from filling orphan_cbs_list */
1449 rcu_adopt_orphan_cbs(rsp);
1450 on_each_cpu(rcu_barrier_func, (void *)call_rcu_func, 1);
1451 preempt_enable(); /* CPU_DYING can again fill orphan_cbs_list */
1452 if (atomic_dec_and_test(&rcu_barrier_cpu_count))
1453 complete(&rcu_barrier_completion);
1454 wait_for_completion(&rcu_barrier_completion);
1455 mutex_unlock(&rcu_barrier_mutex);
1456}
1457
1458/**
1459 * rcu_barrier_bh - Wait until all in-flight call_rcu_bh() callbacks complete.
1460 */
1461void rcu_barrier_bh(void)
1462{
1463 _rcu_barrier(&rcu_bh_state, call_rcu_bh);
1464}
1465EXPORT_SYMBOL_GPL(rcu_barrier_bh);
1466
1467/**
1468 * rcu_barrier_sched - Wait for in-flight call_rcu_sched() callbacks.
1469 */
1470void rcu_barrier_sched(void)
1471{
1472 _rcu_barrier(&rcu_sched_state, call_rcu_sched);
1473}
1474EXPORT_SYMBOL_GPL(rcu_barrier_sched);
1475
1371/* 1476/*
1372 * Do boot-time initialization of a CPU's per-CPU RCU data. 1477 * Do boot-time initialization of a CPU's per-CPU RCU data.
1373 */ 1478 */
@@ -1464,6 +1569,22 @@ int __cpuinit rcu_cpu_notify(struct notifier_block *self,
1464 case CPU_UP_PREPARE_FROZEN: 1569 case CPU_UP_PREPARE_FROZEN:
1465 rcu_online_cpu(cpu); 1570 rcu_online_cpu(cpu);
1466 break; 1571 break;
1572 case CPU_DYING:
1573 case CPU_DYING_FROZEN:
1574 /*
1575 * preempt_disable() in _rcu_barrier() prevents stop_machine(),
1576 * so when "on_each_cpu(rcu_barrier_func, (void *)type, 1);"
1577 * returns, all online cpus have queued rcu_barrier_func().
1578 * The dying CPU clears its cpu_online_mask bit and
1579 * moves all of its RCU callbacks to ->orphan_cbs_list
1580 * in the context of stop_machine(), so subsequent calls
1581 * to _rcu_barrier() will adopt these callbacks and only
1582 * then queue rcu_barrier_func() on all remaining CPUs.
1583 */
1584 rcu_send_cbs_to_orphanage(&rcu_bh_state);
1585 rcu_send_cbs_to_orphanage(&rcu_sched_state);
1586 rcu_preempt_send_cbs_to_orphanage();
1587 break;
1467 case CPU_DEAD: 1588 case CPU_DEAD:
1468 case CPU_DEAD_FROZEN: 1589 case CPU_DEAD_FROZEN:
1469 case CPU_UP_CANCELED: 1590 case CPU_UP_CANCELED:
@@ -1526,7 +1647,8 @@ static void __init rcu_init_one(struct rcu_state *rsp)
1526 cpustride *= rsp->levelspread[i]; 1647 cpustride *= rsp->levelspread[i];
1527 rnp = rsp->level[i]; 1648 rnp = rsp->level[i];
1528 for (j = 0; j < rsp->levelcnt[i]; j++, rnp++) { 1649 for (j = 0; j < rsp->levelcnt[i]; j++, rnp++) {
1529 spin_lock_init(&rnp->lock); 1650 if (rnp != rcu_get_root(rsp))
1651 spin_lock_init(&rnp->lock);
1530 rnp->gpnum = 0; 1652 rnp->gpnum = 0;
1531 rnp->qsmask = 0; 1653 rnp->qsmask = 0;
1532 rnp->qsmaskinit = 0; 1654 rnp->qsmaskinit = 0;
@@ -1549,6 +1671,7 @@ static void __init rcu_init_one(struct rcu_state *rsp)
1549 INIT_LIST_HEAD(&rnp->blocked_tasks[1]); 1671 INIT_LIST_HEAD(&rnp->blocked_tasks[1]);
1550 } 1672 }
1551 } 1673 }
1674 spin_lock_init(&rcu_get_root(rsp)->lock);
1552} 1675}
1553 1676
1554/* 1677/*
@@ -1558,6 +1681,10 @@ static void __init rcu_init_one(struct rcu_state *rsp)
1558 */ 1681 */
1559#define RCU_INIT_FLAVOR(rsp, rcu_data) \ 1682#define RCU_INIT_FLAVOR(rsp, rcu_data) \
1560do { \ 1683do { \
1684 int i; \
1685 int j; \
1686 struct rcu_node *rnp; \
1687 \
1561 rcu_init_one(rsp); \ 1688 rcu_init_one(rsp); \
1562 rnp = (rsp)->level[NUM_RCU_LVLS - 1]; \ 1689 rnp = (rsp)->level[NUM_RCU_LVLS - 1]; \
1563 j = 0; \ 1690 j = 0; \
@@ -1570,31 +1697,8 @@ do { \
1570 } \ 1697 } \
1571} while (0) 1698} while (0)
1572 1699
1573#ifdef CONFIG_TREE_PREEMPT_RCU
1574
1575void __init __rcu_init_preempt(void)
1576{
1577 int i; /* All used by RCU_INIT_FLAVOR(). */
1578 int j;
1579 struct rcu_node *rnp;
1580
1581 RCU_INIT_FLAVOR(&rcu_preempt_state, rcu_preempt_data);
1582}
1583
1584#else /* #ifdef CONFIG_TREE_PREEMPT_RCU */
1585
1586void __init __rcu_init_preempt(void)
1587{
1588}
1589
1590#endif /* #else #ifdef CONFIG_TREE_PREEMPT_RCU */
1591
1592void __init __rcu_init(void) 1700void __init __rcu_init(void)
1593{ 1701{
1594 int i; /* All used by RCU_INIT_FLAVOR(). */
1595 int j;
1596 struct rcu_node *rnp;
1597
1598 rcu_bootup_announce(); 1702 rcu_bootup_announce();
1599#ifdef CONFIG_RCU_CPU_STALL_DETECTOR 1703#ifdef CONFIG_RCU_CPU_STALL_DETECTOR
1600 printk(KERN_INFO "RCU-based detection of stalled CPUs is enabled.\n"); 1704 printk(KERN_INFO "RCU-based detection of stalled CPUs is enabled.\n");
@@ -1605,6 +1709,4 @@ void __init __rcu_init(void)
1605 open_softirq(RCU_SOFTIRQ, rcu_process_callbacks); 1709 open_softirq(RCU_SOFTIRQ, rcu_process_callbacks);
1606} 1710}
1607 1711
1608module_param(blimit, int, 0); 1712#include "rcutree_plugin.h"
1609module_param(qhimark, int, 0);
1610module_param(qlowmark, int, 0);
diff --git a/kernel/rcutree.h b/kernel/rcutree.h
index 8e8287a983c2..b40ac5706040 100644
--- a/kernel/rcutree.h
+++ b/kernel/rcutree.h
@@ -48,14 +48,14 @@
48#elif NR_CPUS <= RCU_FANOUT_SQ 48#elif NR_CPUS <= RCU_FANOUT_SQ
49# define NUM_RCU_LVLS 2 49# define NUM_RCU_LVLS 2
50# define NUM_RCU_LVL_0 1 50# define NUM_RCU_LVL_0 1
51# define NUM_RCU_LVL_1 (((NR_CPUS) + RCU_FANOUT - 1) / RCU_FANOUT) 51# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT)
52# define NUM_RCU_LVL_2 (NR_CPUS) 52# define NUM_RCU_LVL_2 (NR_CPUS)
53# define NUM_RCU_LVL_3 0 53# define NUM_RCU_LVL_3 0
54#elif NR_CPUS <= RCU_FANOUT_CUBE 54#elif NR_CPUS <= RCU_FANOUT_CUBE
55# define NUM_RCU_LVLS 3 55# define NUM_RCU_LVLS 3
56# define NUM_RCU_LVL_0 1 56# define NUM_RCU_LVL_0 1
57# define NUM_RCU_LVL_1 (((NR_CPUS) + RCU_FANOUT_SQ - 1) / RCU_FANOUT_SQ) 57# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_SQ)
58# define NUM_RCU_LVL_2 (((NR_CPUS) + (RCU_FANOUT) - 1) / (RCU_FANOUT)) 58# define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT)
59# define NUM_RCU_LVL_3 NR_CPUS 59# define NUM_RCU_LVL_3 NR_CPUS
60#else 60#else
61# error "CONFIG_RCU_FANOUT insufficient for NR_CPUS" 61# error "CONFIG_RCU_FANOUT insufficient for NR_CPUS"
@@ -79,15 +79,21 @@ struct rcu_dynticks {
79 * Definition for node within the RCU grace-period-detection hierarchy. 79 * Definition for node within the RCU grace-period-detection hierarchy.
80 */ 80 */
81struct rcu_node { 81struct rcu_node {
82 spinlock_t lock; 82 spinlock_t lock; /* Root rcu_node's lock protects some */
83 /* rcu_state fields as well as following. */
83 long gpnum; /* Current grace period for this node. */ 84 long gpnum; /* Current grace period for this node. */
84 /* This will either be equal to or one */ 85 /* This will either be equal to or one */
85 /* behind the root rcu_node's gpnum. */ 86 /* behind the root rcu_node's gpnum. */
86 unsigned long qsmask; /* CPUs or groups that need to switch in */ 87 unsigned long qsmask; /* CPUs or groups that need to switch in */
87 /* order for current grace period to proceed.*/ 88 /* order for current grace period to proceed.*/
89 /* In leaf rcu_node, each bit corresponds to */
90 /* an rcu_data structure, otherwise, each */
91 /* bit corresponds to a child rcu_node */
92 /* structure. */
88 unsigned long qsmaskinit; 93 unsigned long qsmaskinit;
89 /* Per-GP initialization for qsmask. */ 94 /* Per-GP initialization for qsmask. */
90 unsigned long grpmask; /* Mask to apply to parent qsmask. */ 95 unsigned long grpmask; /* Mask to apply to parent qsmask. */
96 /* Only one bit will be set in this mask. */
91 int grplo; /* lowest-numbered CPU or group here. */ 97 int grplo; /* lowest-numbered CPU or group here. */
92 int grphi; /* highest-numbered CPU or group here. */ 98 int grphi; /* highest-numbered CPU or group here. */
93 u8 grpnum; /* CPU/group number for next level up. */ 99 u8 grpnum; /* CPU/group number for next level up. */
@@ -95,8 +101,23 @@ struct rcu_node {
95 struct rcu_node *parent; 101 struct rcu_node *parent;
96 struct list_head blocked_tasks[2]; 102 struct list_head blocked_tasks[2];
97 /* Tasks blocked in RCU read-side critsect. */ 103 /* Tasks blocked in RCU read-side critsect. */
104 /* Grace period number (->gpnum) x blocked */
105 /* by tasks on the (x & 0x1) element of the */
106 /* blocked_tasks[] array. */
98} ____cacheline_internodealigned_in_smp; 107} ____cacheline_internodealigned_in_smp;
99 108
109/*
110 * Do a full breadth-first scan of the rcu_node structures for the
111 * specified rcu_state structure.
112 */
113#define rcu_for_each_node_breadth_first(rsp, rnp) \
114 for ((rnp) = &(rsp)->node[0]; \
115 (rnp) < &(rsp)->node[NUM_RCU_NODES]; (rnp)++)
116
117#define rcu_for_each_leaf_node(rsp, rnp) \
118 for ((rnp) = (rsp)->level[NUM_RCU_LVLS - 1]; \
119 (rnp) < &(rsp)->node[NUM_RCU_NODES]; (rnp)++)
120
100/* Index values for nxttail array in struct rcu_data. */ 121/* Index values for nxttail array in struct rcu_data. */
101#define RCU_DONE_TAIL 0 /* Also RCU_WAIT head. */ 122#define RCU_DONE_TAIL 0 /* Also RCU_WAIT head. */
102#define RCU_WAIT_TAIL 1 /* Also RCU_NEXT_READY head. */ 123#define RCU_WAIT_TAIL 1 /* Also RCU_NEXT_READY head. */
@@ -126,19 +147,22 @@ struct rcu_data {
126 * Any of the partitions might be empty, in which case the 147 * Any of the partitions might be empty, in which case the
127 * pointer to that partition will be equal to the pointer for 148 * pointer to that partition will be equal to the pointer for
128 * the following partition. When the list is empty, all of 149 * the following partition. When the list is empty, all of
129 * the nxttail elements point to nxtlist, which is NULL. 150 * the nxttail elements point to the ->nxtlist pointer itself,
151 * which in that case is NULL.
130 * 152 *
131 * [*nxttail[RCU_NEXT_READY_TAIL], NULL = *nxttail[RCU_NEXT_TAIL]):
132 * Entries that might have arrived after current GP ended
133 * [*nxttail[RCU_WAIT_TAIL], *nxttail[RCU_NEXT_READY_TAIL]):
134 * Entries known to have arrived before current GP ended
135 * [*nxttail[RCU_DONE_TAIL], *nxttail[RCU_WAIT_TAIL]):
136 * Entries that batch # <= ->completed - 1: waiting for current GP
137 * [nxtlist, *nxttail[RCU_DONE_TAIL]): 153 * [nxtlist, *nxttail[RCU_DONE_TAIL]):
138 * Entries that batch # <= ->completed 154 * Entries that batch # <= ->completed
139 * The grace period for these entries has completed, and 155 * The grace period for these entries has completed, and
140 * the other grace-period-completed entries may be moved 156 * the other grace-period-completed entries may be moved
141 * here temporarily in rcu_process_callbacks(). 157 * here temporarily in rcu_process_callbacks().
158 * [*nxttail[RCU_DONE_TAIL], *nxttail[RCU_WAIT_TAIL]):
159 * Entries that batch # <= ->completed - 1: waiting for current GP
160 * [*nxttail[RCU_WAIT_TAIL], *nxttail[RCU_NEXT_READY_TAIL]):
161 * Entries known to have arrived before current GP ended
162 * [*nxttail[RCU_NEXT_READY_TAIL], *nxttail[RCU_NEXT_TAIL]):
163 * Entries that might have arrived after current GP ended
164 * Note that the value of *nxttail[RCU_NEXT_TAIL] will
165 * always be NULL, as this is the end of the list.
142 */ 166 */
143 struct rcu_head *nxtlist; 167 struct rcu_head *nxtlist;
144 struct rcu_head **nxttail[RCU_NEXT_SIZE]; 168 struct rcu_head **nxttail[RCU_NEXT_SIZE];
@@ -216,8 +240,19 @@ struct rcu_state {
216 /* Force QS state. */ 240 /* Force QS state. */
217 long gpnum; /* Current gp number. */ 241 long gpnum; /* Current gp number. */
218 long completed; /* # of last completed gp. */ 242 long completed; /* # of last completed gp. */
243
244 /* End of fields guarded by root rcu_node's lock. */
245
219 spinlock_t onofflock; /* exclude on/offline and */ 246 spinlock_t onofflock; /* exclude on/offline and */
220 /* starting new GP. */ 247 /* starting new GP. Also */
248 /* protects the following */
249 /* orphan_cbs fields. */
250 struct rcu_head *orphan_cbs_list; /* list of rcu_head structs */
251 /* orphaned by all CPUs in */
252 /* a given leaf rcu_node */
253 /* going offline. */
254 struct rcu_head **orphan_cbs_tail; /* And tail pointer. */
255 long orphan_qlen; /* Number of orphaned cbs. */
221 spinlock_t fqslock; /* Only one task forcing */ 256 spinlock_t fqslock; /* Only one task forcing */
222 /* quiescent states. */ 257 /* quiescent states. */
223 unsigned long jiffies_force_qs; /* Time at which to invoke */ 258 unsigned long jiffies_force_qs; /* Time at which to invoke */
@@ -255,5 +290,30 @@ extern struct rcu_state rcu_preempt_state;
255DECLARE_PER_CPU(struct rcu_data, rcu_preempt_data); 290DECLARE_PER_CPU(struct rcu_data, rcu_preempt_data);
256#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */ 291#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */
257 292
258#endif /* #ifdef RCU_TREE_NONCORE */ 293#else /* #ifdef RCU_TREE_NONCORE */
294
295/* Forward declarations for rcutree_plugin.h */
296static inline void rcu_bootup_announce(void);
297long rcu_batches_completed(void);
298static void rcu_preempt_note_context_switch(int cpu);
299static int rcu_preempted_readers(struct rcu_node *rnp);
300#ifdef CONFIG_RCU_CPU_STALL_DETECTOR
301static void rcu_print_task_stall(struct rcu_node *rnp);
302#endif /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */
303static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp);
304#ifdef CONFIG_HOTPLUG_CPU
305static void rcu_preempt_offline_tasks(struct rcu_state *rsp,
306 struct rcu_node *rnp,
307 struct rcu_data *rdp);
308static void rcu_preempt_offline_cpu(int cpu);
309#endif /* #ifdef CONFIG_HOTPLUG_CPU */
310static void rcu_preempt_check_callbacks(int cpu);
311static void rcu_preempt_process_callbacks(void);
312void call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu));
313static int rcu_preempt_pending(int cpu);
314static int rcu_preempt_needs_cpu(int cpu);
315static void __cpuinit rcu_preempt_init_percpu_data(int cpu);
316static void rcu_preempt_send_cbs_to_orphanage(void);
317static void __init __rcu_init_preempt(void);
259 318
319#endif /* #else #ifdef RCU_TREE_NONCORE */
diff --git a/kernel/rcutree_plugin.h b/kernel/rcutree_plugin.h
index 1cee04f627eb..c0cb783aa16a 100644
--- a/kernel/rcutree_plugin.h
+++ b/kernel/rcutree_plugin.h
@@ -150,6 +150,16 @@ void __rcu_read_lock(void)
150} 150}
151EXPORT_SYMBOL_GPL(__rcu_read_lock); 151EXPORT_SYMBOL_GPL(__rcu_read_lock);
152 152
153/*
154 * Check for preempted RCU readers blocking the current grace period
155 * for the specified rcu_node structure. If the caller needs a reliable
156 * answer, it must hold the rcu_node's ->lock.
157 */
158static int rcu_preempted_readers(struct rcu_node *rnp)
159{
160 return !list_empty(&rnp->blocked_tasks[rnp->gpnum & 0x1]);
161}
162
153static void rcu_read_unlock_special(struct task_struct *t) 163static void rcu_read_unlock_special(struct task_struct *t)
154{ 164{
155 int empty; 165 int empty;
@@ -196,7 +206,7 @@ static void rcu_read_unlock_special(struct task_struct *t)
196 break; 206 break;
197 spin_unlock(&rnp->lock); /* irqs remain disabled. */ 207 spin_unlock(&rnp->lock); /* irqs remain disabled. */
198 } 208 }
199 empty = list_empty(&rnp->blocked_tasks[rnp->gpnum & 0x1]); 209 empty = !rcu_preempted_readers(rnp);
200 list_del_init(&t->rcu_node_entry); 210 list_del_init(&t->rcu_node_entry);
201 t->rcu_blocked_node = NULL; 211 t->rcu_blocked_node = NULL;
202 212
@@ -207,7 +217,7 @@ static void rcu_read_unlock_special(struct task_struct *t)
207 * drop rnp->lock and restore irq. 217 * drop rnp->lock and restore irq.
208 */ 218 */
209 if (!empty && rnp->qsmask == 0 && 219 if (!empty && rnp->qsmask == 0 &&
210 list_empty(&rnp->blocked_tasks[rnp->gpnum & 0x1])) { 220 !rcu_preempted_readers(rnp)) {
211 struct rcu_node *rnp_p; 221 struct rcu_node *rnp_p;
212 222
213 if (rnp->parent == NULL) { 223 if (rnp->parent == NULL) {
@@ -257,12 +267,12 @@ static void rcu_print_task_stall(struct rcu_node *rnp)
257{ 267{
258 unsigned long flags; 268 unsigned long flags;
259 struct list_head *lp; 269 struct list_head *lp;
260 int phase = rnp->gpnum & 0x1; 270 int phase;
261 struct task_struct *t; 271 struct task_struct *t;
262 272
263 if (!list_empty(&rnp->blocked_tasks[phase])) { 273 if (rcu_preempted_readers(rnp)) {
264 spin_lock_irqsave(&rnp->lock, flags); 274 spin_lock_irqsave(&rnp->lock, flags);
265 phase = rnp->gpnum & 0x1; /* re-read under lock. */ 275 phase = rnp->gpnum & 0x1;
266 lp = &rnp->blocked_tasks[phase]; 276 lp = &rnp->blocked_tasks[phase];
267 list_for_each_entry(t, lp, rcu_node_entry) 277 list_for_each_entry(t, lp, rcu_node_entry)
268 printk(" P%d", t->pid); 278 printk(" P%d", t->pid);
@@ -281,20 +291,10 @@ static void rcu_print_task_stall(struct rcu_node *rnp)
281 */ 291 */
282static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp) 292static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp)
283{ 293{
284 WARN_ON_ONCE(!list_empty(&rnp->blocked_tasks[rnp->gpnum & 0x1])); 294 WARN_ON_ONCE(rcu_preempted_readers(rnp));
285 WARN_ON_ONCE(rnp->qsmask); 295 WARN_ON_ONCE(rnp->qsmask);
286} 296}
287 297
288/*
289 * Check for preempted RCU readers for the specified rcu_node structure.
290 * If the caller needs a reliable answer, it must hold the rcu_node's
291 * >lock.
292 */
293static int rcu_preempted_readers(struct rcu_node *rnp)
294{
295 return !list_empty(&rnp->blocked_tasks[rnp->gpnum & 0x1]);
296}
297
298#ifdef CONFIG_HOTPLUG_CPU 298#ifdef CONFIG_HOTPLUG_CPU
299 299
300/* 300/*
@@ -410,6 +410,15 @@ static int rcu_preempt_needs_cpu(int cpu)
410 return !!per_cpu(rcu_preempt_data, cpu).nxtlist; 410 return !!per_cpu(rcu_preempt_data, cpu).nxtlist;
411} 411}
412 412
413/**
414 * rcu_barrier - Wait until all in-flight call_rcu() callbacks complete.
415 */
416void rcu_barrier(void)
417{
418 _rcu_barrier(&rcu_preempt_state, call_rcu);
419}
420EXPORT_SYMBOL_GPL(rcu_barrier);
421
413/* 422/*
414 * Initialize preemptable RCU's per-CPU data. 423 * Initialize preemptable RCU's per-CPU data.
415 */ 424 */
@@ -419,6 +428,22 @@ static void __cpuinit rcu_preempt_init_percpu_data(int cpu)
419} 428}
420 429
421/* 430/*
431 * Move preemptable RCU's callbacks to ->orphan_cbs_list.
432 */
433static void rcu_preempt_send_cbs_to_orphanage(void)
434{
435 rcu_send_cbs_to_orphanage(&rcu_preempt_state);
436}
437
438/*
439 * Initialize preemptable RCU's state structures.
440 */
441static void __init __rcu_init_preempt(void)
442{
443 RCU_INIT_FLAVOR(&rcu_preempt_state, rcu_preempt_data);
444}
445
446/*
422 * Check for a task exiting while in a preemptable-RCU read-side 447 * Check for a task exiting while in a preemptable-RCU read-side
423 * critical section, clean up if so. No need to issue warnings, 448 * critical section, clean up if so. No need to issue warnings,
424 * as debug_check_no_locks_held() already does this if lockdep 449 * as debug_check_no_locks_held() already does this if lockdep
@@ -461,6 +486,15 @@ static void rcu_preempt_note_context_switch(int cpu)
461{ 486{
462} 487}
463 488
489/*
490 * Because preemptable RCU does not exist, there are never any preempted
491 * RCU readers.
492 */
493static int rcu_preempted_readers(struct rcu_node *rnp)
494{
495 return 0;
496}
497
464#ifdef CONFIG_RCU_CPU_STALL_DETECTOR 498#ifdef CONFIG_RCU_CPU_STALL_DETECTOR
465 499
466/* 500/*
@@ -483,15 +517,6 @@ static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp)
483 WARN_ON_ONCE(rnp->qsmask); 517 WARN_ON_ONCE(rnp->qsmask);
484} 518}
485 519
486/*
487 * Because preemptable RCU does not exist, there are never any preempted
488 * RCU readers.
489 */
490static int rcu_preempted_readers(struct rcu_node *rnp)
491{
492 return 0;
493}
494
495#ifdef CONFIG_HOTPLUG_CPU 520#ifdef CONFIG_HOTPLUG_CPU
496 521
497/* 522/*
@@ -518,7 +543,7 @@ static void rcu_preempt_offline_cpu(int cpu)
518 * Because preemptable RCU does not exist, it never has any callbacks 543 * Because preemptable RCU does not exist, it never has any callbacks
519 * to check. 544 * to check.
520 */ 545 */
521void rcu_preempt_check_callbacks(int cpu) 546static void rcu_preempt_check_callbacks(int cpu)
522{ 547{
523} 548}
524 549
@@ -526,7 +551,7 @@ void rcu_preempt_check_callbacks(int cpu)
526 * Because preemptable RCU does not exist, it never has any callbacks 551 * Because preemptable RCU does not exist, it never has any callbacks
527 * to process. 552 * to process.
528 */ 553 */
529void rcu_preempt_process_callbacks(void) 554static void rcu_preempt_process_callbacks(void)
530{ 555{
531} 556}
532 557
@@ -556,6 +581,16 @@ static int rcu_preempt_needs_cpu(int cpu)
556} 581}
557 582
558/* 583/*
584 * Because preemptable RCU does not exist, rcu_barrier() is just
585 * another name for rcu_barrier_sched().
586 */
587void rcu_barrier(void)
588{
589 rcu_barrier_sched();
590}
591EXPORT_SYMBOL_GPL(rcu_barrier);
592
593/*
559 * Because preemptable RCU does not exist, there is no per-CPU 594 * Because preemptable RCU does not exist, there is no per-CPU
560 * data to initialize. 595 * data to initialize.
561 */ 596 */
@@ -563,4 +598,18 @@ static void __cpuinit rcu_preempt_init_percpu_data(int cpu)
563{ 598{
564} 599}
565 600
601/*
602 * Because there is no preemptable RCU, there are no callbacks to move.
603 */
604static void rcu_preempt_send_cbs_to_orphanage(void)
605{
606}
607
608/*
609 * Because preemptable RCU does not exist, it need not be initialized.
610 */
611static void __init __rcu_init_preempt(void)
612{
613}
614
566#endif /* #else #ifdef CONFIG_TREE_PREEMPT_RCU */ 615#endif /* #else #ifdef CONFIG_TREE_PREEMPT_RCU */
diff --git a/kernel/rcutree_trace.c b/kernel/rcutree_trace.c
index c89f5e9fd173..4b31c779e62e 100644
--- a/kernel/rcutree_trace.c
+++ b/kernel/rcutree_trace.c
@@ -93,7 +93,7 @@ static int rcudata_open(struct inode *inode, struct file *file)
93 return single_open(file, show_rcudata, NULL); 93 return single_open(file, show_rcudata, NULL);
94} 94}
95 95
96static struct file_operations rcudata_fops = { 96static const struct file_operations rcudata_fops = {
97 .owner = THIS_MODULE, 97 .owner = THIS_MODULE,
98 .open = rcudata_open, 98 .open = rcudata_open,
99 .read = seq_read, 99 .read = seq_read,
@@ -145,7 +145,7 @@ static int rcudata_csv_open(struct inode *inode, struct file *file)
145 return single_open(file, show_rcudata_csv, NULL); 145 return single_open(file, show_rcudata_csv, NULL);
146} 146}
147 147
148static struct file_operations rcudata_csv_fops = { 148static const struct file_operations rcudata_csv_fops = {
149 .owner = THIS_MODULE, 149 .owner = THIS_MODULE,
150 .open = rcudata_csv_open, 150 .open = rcudata_csv_open,
151 .read = seq_read, 151 .read = seq_read,
@@ -159,13 +159,13 @@ static void print_one_rcu_state(struct seq_file *m, struct rcu_state *rsp)
159 struct rcu_node *rnp; 159 struct rcu_node *rnp;
160 160
161 seq_printf(m, "c=%ld g=%ld s=%d jfq=%ld j=%x " 161 seq_printf(m, "c=%ld g=%ld s=%d jfq=%ld j=%x "
162 "nfqs=%lu/nfqsng=%lu(%lu) fqlh=%lu\n", 162 "nfqs=%lu/nfqsng=%lu(%lu) fqlh=%lu oqlen=%ld\n",
163 rsp->completed, rsp->gpnum, rsp->signaled, 163 rsp->completed, rsp->gpnum, rsp->signaled,
164 (long)(rsp->jiffies_force_qs - jiffies), 164 (long)(rsp->jiffies_force_qs - jiffies),
165 (int)(jiffies & 0xffff), 165 (int)(jiffies & 0xffff),
166 rsp->n_force_qs, rsp->n_force_qs_ngp, 166 rsp->n_force_qs, rsp->n_force_qs_ngp,
167 rsp->n_force_qs - rsp->n_force_qs_ngp, 167 rsp->n_force_qs - rsp->n_force_qs_ngp,
168 rsp->n_force_qs_lh); 168 rsp->n_force_qs_lh, rsp->orphan_qlen);
169 for (rnp = &rsp->node[0]; rnp - &rsp->node[0] < NUM_RCU_NODES; rnp++) { 169 for (rnp = &rsp->node[0]; rnp - &rsp->node[0] < NUM_RCU_NODES; rnp++) {
170 if (rnp->level != level) { 170 if (rnp->level != level) {
171 seq_puts(m, "\n"); 171 seq_puts(m, "\n");
@@ -196,7 +196,7 @@ static int rcuhier_open(struct inode *inode, struct file *file)
196 return single_open(file, show_rcuhier, NULL); 196 return single_open(file, show_rcuhier, NULL);
197} 197}
198 198
199static struct file_operations rcuhier_fops = { 199static const struct file_operations rcuhier_fops = {
200 .owner = THIS_MODULE, 200 .owner = THIS_MODULE,
201 .open = rcuhier_open, 201 .open = rcuhier_open,
202 .read = seq_read, 202 .read = seq_read,
@@ -222,7 +222,7 @@ static int rcugp_open(struct inode *inode, struct file *file)
222 return single_open(file, show_rcugp, NULL); 222 return single_open(file, show_rcugp, NULL);
223} 223}
224 224
225static struct file_operations rcugp_fops = { 225static const struct file_operations rcugp_fops = {
226 .owner = THIS_MODULE, 226 .owner = THIS_MODULE,
227 .open = rcugp_open, 227 .open = rcugp_open,
228 .read = seq_read, 228 .read = seq_read,
@@ -276,7 +276,7 @@ static int rcu_pending_open(struct inode *inode, struct file *file)
276 return single_open(file, show_rcu_pending, NULL); 276 return single_open(file, show_rcu_pending, NULL);
277} 277}
278 278
279static struct file_operations rcu_pending_fops = { 279static const struct file_operations rcu_pending_fops = {
280 .owner = THIS_MODULE, 280 .owner = THIS_MODULE,
281 .open = rcu_pending_open, 281 .open = rcu_pending_open,
282 .read = seq_read, 282 .read = seq_read,
diff --git a/kernel/res_counter.c b/kernel/res_counter.c
index 88faec23e833..bcdabf37c40b 100644
--- a/kernel/res_counter.c
+++ b/kernel/res_counter.c
@@ -37,27 +37,17 @@ int res_counter_charge_locked(struct res_counter *counter, unsigned long val)
37} 37}
38 38
39int res_counter_charge(struct res_counter *counter, unsigned long val, 39int res_counter_charge(struct res_counter *counter, unsigned long val,
40 struct res_counter **limit_fail_at, 40 struct res_counter **limit_fail_at)
41 struct res_counter **soft_limit_fail_at)
42{ 41{
43 int ret; 42 int ret;
44 unsigned long flags; 43 unsigned long flags;
45 struct res_counter *c, *u; 44 struct res_counter *c, *u;
46 45
47 *limit_fail_at = NULL; 46 *limit_fail_at = NULL;
48 if (soft_limit_fail_at)
49 *soft_limit_fail_at = NULL;
50 local_irq_save(flags); 47 local_irq_save(flags);
51 for (c = counter; c != NULL; c = c->parent) { 48 for (c = counter; c != NULL; c = c->parent) {
52 spin_lock(&c->lock); 49 spin_lock(&c->lock);
53 ret = res_counter_charge_locked(c, val); 50 ret = res_counter_charge_locked(c, val);
54 /*
55 * With soft limits, we return the highest ancestor
56 * that exceeds its soft limit
57 */
58 if (soft_limit_fail_at &&
59 !res_counter_soft_limit_check_locked(c))
60 *soft_limit_fail_at = c;
61 spin_unlock(&c->lock); 51 spin_unlock(&c->lock);
62 if (ret < 0) { 52 if (ret < 0) {
63 *limit_fail_at = c; 53 *limit_fail_at = c;
@@ -85,8 +75,7 @@ void res_counter_uncharge_locked(struct res_counter *counter, unsigned long val)
85 counter->usage -= val; 75 counter->usage -= val;
86} 76}
87 77
88void res_counter_uncharge(struct res_counter *counter, unsigned long val, 78void res_counter_uncharge(struct res_counter *counter, unsigned long val)
89 bool *was_soft_limit_excess)
90{ 79{
91 unsigned long flags; 80 unsigned long flags;
92 struct res_counter *c; 81 struct res_counter *c;
@@ -94,9 +83,6 @@ void res_counter_uncharge(struct res_counter *counter, unsigned long val,
94 local_irq_save(flags); 83 local_irq_save(flags);
95 for (c = counter; c != NULL; c = c->parent) { 84 for (c = counter; c != NULL; c = c->parent) {
96 spin_lock(&c->lock); 85 spin_lock(&c->lock);
97 if (was_soft_limit_excess)
98 *was_soft_limit_excess =
99 !res_counter_soft_limit_check_locked(c);
100 res_counter_uncharge_locked(c, val); 86 res_counter_uncharge_locked(c, val);
101 spin_unlock(&c->lock); 87 spin_unlock(&c->lock);
102 } 88 }
diff --git a/kernel/sched.c b/kernel/sched.c
index ee61f454a98b..76c0e9691fc0 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -780,7 +780,7 @@ static int sched_feat_open(struct inode *inode, struct file *filp)
780 return single_open(filp, sched_feat_show, NULL); 780 return single_open(filp, sched_feat_show, NULL);
781} 781}
782 782
783static struct file_operations sched_feat_fops = { 783static const struct file_operations sched_feat_fops = {
784 .open = sched_feat_open, 784 .open = sched_feat_open,
785 .write = sched_feat_write, 785 .write = sched_feat_write,
786 .read = seq_read, 786 .read = seq_read,
@@ -2515,22 +2515,17 @@ void sched_fork(struct task_struct *p, int clone_flags)
2515 __sched_fork(p); 2515 __sched_fork(p);
2516 2516
2517 /* 2517 /*
2518 * Make sure we do not leak PI boosting priority to the child.
2519 */
2520 p->prio = current->normal_prio;
2521
2522 /*
2523 * Revert to default priority/policy on fork if requested. 2518 * Revert to default priority/policy on fork if requested.
2524 */ 2519 */
2525 if (unlikely(p->sched_reset_on_fork)) { 2520 if (unlikely(p->sched_reset_on_fork)) {
2526 if (p->policy == SCHED_FIFO || p->policy == SCHED_RR) 2521 if (p->policy == SCHED_FIFO || p->policy == SCHED_RR) {
2527 p->policy = SCHED_NORMAL; 2522 p->policy = SCHED_NORMAL;
2528 2523 p->normal_prio = p->static_prio;
2529 if (p->normal_prio < DEFAULT_PRIO) 2524 }
2530 p->prio = DEFAULT_PRIO;
2531 2525
2532 if (PRIO_TO_NICE(p->static_prio) < 0) { 2526 if (PRIO_TO_NICE(p->static_prio) < 0) {
2533 p->static_prio = NICE_TO_PRIO(0); 2527 p->static_prio = NICE_TO_PRIO(0);
2528 p->normal_prio = p->static_prio;
2534 set_load_weight(p); 2529 set_load_weight(p);
2535 } 2530 }
2536 2531
@@ -2541,6 +2536,11 @@ void sched_fork(struct task_struct *p, int clone_flags)
2541 p->sched_reset_on_fork = 0; 2536 p->sched_reset_on_fork = 0;
2542 } 2537 }
2543 2538
2539 /*
2540 * Make sure we do not leak PI boosting priority to the child.
2541 */
2542 p->prio = current->normal_prio;
2543
2544 if (!rt_prio(p->prio)) 2544 if (!rt_prio(p->prio))
2545 p->sched_class = &fair_sched_class; 2545 p->sched_class = &fair_sched_class;
2546 2546
@@ -2581,8 +2581,6 @@ void wake_up_new_task(struct task_struct *p, unsigned long clone_flags)
2581 BUG_ON(p->state != TASK_RUNNING); 2581 BUG_ON(p->state != TASK_RUNNING);
2582 update_rq_clock(rq); 2582 update_rq_clock(rq);
2583 2583
2584 p->prio = effective_prio(p);
2585
2586 if (!p->sched_class->task_new || !current->se.on_rq) { 2584 if (!p->sched_class->task_new || !current->se.on_rq) {
2587 activate_task(rq, p, 0); 2585 activate_task(rq, p, 0);
2588 } else { 2586 } else {
diff --git a/kernel/sched_clock.c b/kernel/sched_clock.c
index ac2e1dc708bd..479ce5682d7c 100644
--- a/kernel/sched_clock.c
+++ b/kernel/sched_clock.c
@@ -127,7 +127,7 @@ again:
127 clock = wrap_max(clock, min_clock); 127 clock = wrap_max(clock, min_clock);
128 clock = wrap_min(clock, max_clock); 128 clock = wrap_min(clock, max_clock);
129 129
130 if (cmpxchg(&scd->clock, old_clock, clock) != old_clock) 130 if (cmpxchg64(&scd->clock, old_clock, clock) != old_clock)
131 goto again; 131 goto again;
132 132
133 return clock; 133 return clock;
@@ -163,7 +163,7 @@ again:
163 val = remote_clock; 163 val = remote_clock;
164 } 164 }
165 165
166 if (cmpxchg(ptr, old_val, val) != old_val) 166 if (cmpxchg64(ptr, old_val, val) != old_val)
167 goto again; 167 goto again;
168 168
169 return val; 169 return val;
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index e0f59a21c061..89aed5933ed4 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -231,6 +231,13 @@ void tick_nohz_stop_sched_tick(int inidle)
231 if (!inidle && !ts->inidle) 231 if (!inidle && !ts->inidle)
232 goto end; 232 goto end;
233 233
234 /*
235 * Set ts->inidle unconditionally. Even if the system did not
236 * switch to NOHZ mode the cpu frequency governers rely on the
237 * update of the idle time accounting in tick_nohz_start_idle().
238 */
239 ts->inidle = 1;
240
234 now = tick_nohz_start_idle(ts); 241 now = tick_nohz_start_idle(ts);
235 242
236 /* 243 /*
@@ -248,8 +255,6 @@ void tick_nohz_stop_sched_tick(int inidle)
248 if (unlikely(ts->nohz_mode == NOHZ_MODE_INACTIVE)) 255 if (unlikely(ts->nohz_mode == NOHZ_MODE_INACTIVE))
249 goto end; 256 goto end;
250 257
251 ts->inidle = 1;
252
253 if (need_resched()) 258 if (need_resched())
254 goto end; 259 goto end;
255 260
diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
index fb0f46fa1ecd..c3a4e2907eaa 100644
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -13,6 +13,7 @@
13#include <linux/percpu.h> 13#include <linux/percpu.h>
14#include <linux/init.h> 14#include <linux/init.h>
15#include <linux/mm.h> 15#include <linux/mm.h>
16#include <linux/sched.h>
16#include <linux/sysdev.h> 17#include <linux/sysdev.h>
17#include <linux/clocksource.h> 18#include <linux/clocksource.h>
18#include <linux/jiffies.h> 19#include <linux/jiffies.h>
diff --git a/kernel/time/timer_list.c b/kernel/time/timer_list.c
index fddd69d16e03..1b5b7aa2fdfd 100644
--- a/kernel/time/timer_list.c
+++ b/kernel/time/timer_list.c
@@ -275,7 +275,7 @@ static int timer_list_open(struct inode *inode, struct file *filp)
275 return single_open(filp, timer_list_show, NULL); 275 return single_open(filp, timer_list_show, NULL);
276} 276}
277 277
278static struct file_operations timer_list_fops = { 278static const struct file_operations timer_list_fops = {
279 .open = timer_list_open, 279 .open = timer_list_open,
280 .read = seq_read, 280 .read = seq_read,
281 .llseek = seq_lseek, 281 .llseek = seq_lseek,
diff --git a/kernel/time/timer_stats.c b/kernel/time/timer_stats.c
index 4cde8b9c716f..ee5681f8d7ec 100644
--- a/kernel/time/timer_stats.c
+++ b/kernel/time/timer_stats.c
@@ -395,7 +395,7 @@ static int tstats_open(struct inode *inode, struct file *filp)
395 return single_open(filp, tstats_show, NULL); 395 return single_open(filp, tstats_show, NULL);
396} 396}
397 397
398static struct file_operations tstats_fops = { 398static const struct file_operations tstats_fops = {
399 .open = tstats_open, 399 .open = tstats_open,
400 .read = seq_read, 400 .read = seq_read,
401 .write = tstats_write, 401 .write = tstats_write,
diff --git a/kernel/trace/blktrace.c b/kernel/trace/blktrace.c
index 3eb159c277c8..d9d6206e0b14 100644
--- a/kernel/trace/blktrace.c
+++ b/kernel/trace/blktrace.c
@@ -856,6 +856,37 @@ static void blk_add_trace_remap(struct request_queue *q, struct bio *bio,
856} 856}
857 857
858/** 858/**
859 * blk_add_trace_rq_remap - Add a trace for a request-remap operation
860 * @q: queue the io is for
861 * @rq: the source request
862 * @dev: target device
863 * @from: source sector
864 *
865 * Description:
866 * Device mapper remaps request to other devices.
867 * Add a trace for that action.
868 *
869 **/
870static void blk_add_trace_rq_remap(struct request_queue *q,
871 struct request *rq, dev_t dev,
872 sector_t from)
873{
874 struct blk_trace *bt = q->blk_trace;
875 struct blk_io_trace_remap r;
876
877 if (likely(!bt))
878 return;
879
880 r.device_from = cpu_to_be32(dev);
881 r.device_to = cpu_to_be32(disk_devt(rq->rq_disk));
882 r.sector_from = cpu_to_be64(from);
883
884 __blk_add_trace(bt, blk_rq_pos(rq), blk_rq_bytes(rq),
885 rq_data_dir(rq), BLK_TA_REMAP, !!rq->errors,
886 sizeof(r), &r);
887}
888
889/**
859 * blk_add_driver_data - Add binary message with driver-specific data 890 * blk_add_driver_data - Add binary message with driver-specific data
860 * @q: queue the io is for 891 * @q: queue the io is for
861 * @rq: io request 892 * @rq: io request
@@ -922,10 +953,13 @@ static void blk_register_tracepoints(void)
922 WARN_ON(ret); 953 WARN_ON(ret);
923 ret = register_trace_block_remap(blk_add_trace_remap); 954 ret = register_trace_block_remap(blk_add_trace_remap);
924 WARN_ON(ret); 955 WARN_ON(ret);
956 ret = register_trace_block_rq_remap(blk_add_trace_rq_remap);
957 WARN_ON(ret);
925} 958}
926 959
927static void blk_unregister_tracepoints(void) 960static void blk_unregister_tracepoints(void)
928{ 961{
962 unregister_trace_block_rq_remap(blk_add_trace_rq_remap);
929 unregister_trace_block_remap(blk_add_trace_remap); 963 unregister_trace_block_remap(blk_add_trace_remap);
930 unregister_trace_block_split(blk_add_trace_split); 964 unregister_trace_block_split(blk_add_trace_split);
931 unregister_trace_block_unplug_io(blk_add_trace_unplug_io); 965 unregister_trace_block_unplug_io(blk_add_trace_unplug_io);
@@ -1657,6 +1691,11 @@ int blk_trace_init_sysfs(struct device *dev)
1657 return sysfs_create_group(&dev->kobj, &blk_trace_attr_group); 1691 return sysfs_create_group(&dev->kobj, &blk_trace_attr_group);
1658} 1692}
1659 1693
1694void blk_trace_remove_sysfs(struct device *dev)
1695{
1696 sysfs_remove_group(&dev->kobj, &blk_trace_attr_group);
1697}
1698
1660#endif /* CONFIG_BLK_DEV_IO_TRACE */ 1699#endif /* CONFIG_BLK_DEV_IO_TRACE */
1661 1700
1662#ifdef CONFIG_EVENT_TRACING 1701#ifdef CONFIG_EVENT_TRACING
diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
index 46592feab5a6..37ba67e33265 100644
--- a/kernel/trace/ftrace.c
+++ b/kernel/trace/ftrace.c
@@ -225,7 +225,11 @@ static void ftrace_update_pid_func(void)
225 if (ftrace_trace_function == ftrace_stub) 225 if (ftrace_trace_function == ftrace_stub)
226 return; 226 return;
227 227
228#ifdef CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST
228 func = ftrace_trace_function; 229 func = ftrace_trace_function;
230#else
231 func = __ftrace_trace_function;
232#endif
229 233
230 if (ftrace_pid_trace) { 234 if (ftrace_pid_trace) {
231 set_ftrace_pid_function(func); 235 set_ftrace_pid_function(func);
@@ -1074,14 +1078,9 @@ static void ftrace_replace_code(int enable)
1074 failed = __ftrace_replace_code(rec, enable); 1078 failed = __ftrace_replace_code(rec, enable);
1075 if (failed) { 1079 if (failed) {
1076 rec->flags |= FTRACE_FL_FAILED; 1080 rec->flags |= FTRACE_FL_FAILED;
1077 if ((system_state == SYSTEM_BOOTING) || 1081 ftrace_bug(failed, rec->ip);
1078 !core_kernel_text(rec->ip)) { 1082 /* Stop processing */
1079 ftrace_free_rec(rec); 1083 return;
1080 } else {
1081 ftrace_bug(failed, rec->ip);
1082 /* Stop processing */
1083 return;
1084 }
1085 } 1084 }
1086 } while_for_each_ftrace_rec(); 1085 } while_for_each_ftrace_rec();
1087} 1086}
@@ -2658,19 +2657,17 @@ static int ftrace_convert_nops(struct module *mod,
2658} 2657}
2659 2658
2660#ifdef CONFIG_MODULES 2659#ifdef CONFIG_MODULES
2661void ftrace_release(void *start, void *end) 2660void ftrace_release_mod(struct module *mod)
2662{ 2661{
2663 struct dyn_ftrace *rec; 2662 struct dyn_ftrace *rec;
2664 struct ftrace_page *pg; 2663 struct ftrace_page *pg;
2665 unsigned long s = (unsigned long)start;
2666 unsigned long e = (unsigned long)end;
2667 2664
2668 if (ftrace_disabled || !start || start == end) 2665 if (ftrace_disabled)
2669 return; 2666 return;
2670 2667
2671 mutex_lock(&ftrace_lock); 2668 mutex_lock(&ftrace_lock);
2672 do_for_each_ftrace_rec(pg, rec) { 2669 do_for_each_ftrace_rec(pg, rec) {
2673 if ((rec->ip >= s) && (rec->ip < e)) { 2670 if (within_module_core(rec->ip, mod)) {
2674 /* 2671 /*
2675 * rec->ip is changed in ftrace_free_rec() 2672 * rec->ip is changed in ftrace_free_rec()
2676 * It should not between s and e if record was freed. 2673 * It should not between s and e if record was freed.
@@ -2702,9 +2699,7 @@ static int ftrace_module_notify(struct notifier_block *self,
2702 mod->num_ftrace_callsites); 2699 mod->num_ftrace_callsites);
2703 break; 2700 break;
2704 case MODULE_STATE_GOING: 2701 case MODULE_STATE_GOING:
2705 ftrace_release(mod->ftrace_callsites, 2702 ftrace_release_mod(mod);
2706 mod->ftrace_callsites +
2707 mod->num_ftrace_callsites);
2708 break; 2703 break;
2709 } 2704 }
2710 2705
diff --git a/kernel/trace/kmemtrace.c b/kernel/trace/kmemtrace.c
index 81b1645c8549..a91da69f153a 100644
--- a/kernel/trace/kmemtrace.c
+++ b/kernel/trace/kmemtrace.c
@@ -501,7 +501,7 @@ static int __init init_kmem_tracer(void)
501 return 1; 501 return 1;
502 } 502 }
503 503
504 if (!register_tracer(&kmem_tracer)) { 504 if (register_tracer(&kmem_tracer) != 0) {
505 pr_warning("Warning: could not register the kmem tracer\n"); 505 pr_warning("Warning: could not register the kmem tracer\n");
506 return 1; 506 return 1;
507 } 507 }
diff --git a/kernel/trace/trace_branch.c b/kernel/trace/trace_branch.c
index 7a7a9fd249a9..4a194f08f88c 100644
--- a/kernel/trace/trace_branch.c
+++ b/kernel/trace/trace_branch.c
@@ -34,6 +34,7 @@ probe_likely_condition(struct ftrace_branch_data *f, int val, int expect)
34 struct trace_array *tr = branch_tracer; 34 struct trace_array *tr = branch_tracer;
35 struct ring_buffer_event *event; 35 struct ring_buffer_event *event;
36 struct trace_branch *entry; 36 struct trace_branch *entry;
37 struct ring_buffer *buffer;
37 unsigned long flags; 38 unsigned long flags;
38 int cpu, pc; 39 int cpu, pc;
39 const char *p; 40 const char *p;
@@ -54,7 +55,8 @@ probe_likely_condition(struct ftrace_branch_data *f, int val, int expect)
54 goto out; 55 goto out;
55 56
56 pc = preempt_count(); 57 pc = preempt_count();
57 event = trace_buffer_lock_reserve(tr, TRACE_BRANCH, 58 buffer = tr->buffer;
59 event = trace_buffer_lock_reserve(buffer, TRACE_BRANCH,
58 sizeof(*entry), flags, pc); 60 sizeof(*entry), flags, pc);
59 if (!event) 61 if (!event)
60 goto out; 62 goto out;
@@ -74,8 +76,8 @@ probe_likely_condition(struct ftrace_branch_data *f, int val, int expect)
74 entry->line = f->line; 76 entry->line = f->line;
75 entry->correct = val == expect; 77 entry->correct = val == expect;
76 78
77 if (!filter_check_discard(call, entry, tr->buffer, event)) 79 if (!filter_check_discard(call, entry, buffer, event))
78 ring_buffer_unlock_commit(tr->buffer, event); 80 ring_buffer_unlock_commit(buffer, event);
79 81
80 out: 82 out:
81 atomic_dec(&tr->data[cpu]->disabled); 83 atomic_dec(&tr->data[cpu]->disabled);
diff --git a/kernel/trace/trace_event_profile.c b/kernel/trace/trace_event_profile.c
index dd44b8768867..8d5c171cc998 100644
--- a/kernel/trace/trace_event_profile.c
+++ b/kernel/trace/trace_event_profile.c
@@ -31,7 +31,7 @@ static int ftrace_profile_enable_event(struct ftrace_event_call *event)
31 if (atomic_inc_return(&event->profile_count)) 31 if (atomic_inc_return(&event->profile_count))
32 return 0; 32 return 0;
33 33
34 if (!total_profile_count++) { 34 if (!total_profile_count) {
35 buf = (char *)alloc_percpu(profile_buf_t); 35 buf = (char *)alloc_percpu(profile_buf_t);
36 if (!buf) 36 if (!buf)
37 goto fail_buf; 37 goto fail_buf;
@@ -46,14 +46,19 @@ static int ftrace_profile_enable_event(struct ftrace_event_call *event)
46 } 46 }
47 47
48 ret = event->profile_enable(); 48 ret = event->profile_enable();
49 if (!ret) 49 if (!ret) {
50 total_profile_count++;
50 return 0; 51 return 0;
52 }
51 53
52 kfree(trace_profile_buf_nmi);
53fail_buf_nmi: 54fail_buf_nmi:
54 kfree(trace_profile_buf); 55 if (!total_profile_count) {
56 free_percpu(trace_profile_buf_nmi);
57 free_percpu(trace_profile_buf);
58 trace_profile_buf_nmi = NULL;
59 trace_profile_buf = NULL;
60 }
55fail_buf: 61fail_buf:
56 total_profile_count--;
57 atomic_dec(&event->profile_count); 62 atomic_dec(&event->profile_count);
58 63
59 return ret; 64 return ret;
diff --git a/kernel/trace/trace_hw_branches.c b/kernel/trace/trace_hw_branches.c
index 23b63859130e..69543a905cd5 100644
--- a/kernel/trace/trace_hw_branches.c
+++ b/kernel/trace/trace_hw_branches.c
@@ -165,6 +165,7 @@ void trace_hw_branch(u64 from, u64 to)
165 struct ftrace_event_call *call = &event_hw_branch; 165 struct ftrace_event_call *call = &event_hw_branch;
166 struct trace_array *tr = hw_branch_trace; 166 struct trace_array *tr = hw_branch_trace;
167 struct ring_buffer_event *event; 167 struct ring_buffer_event *event;
168 struct ring_buffer *buf;
168 struct hw_branch_entry *entry; 169 struct hw_branch_entry *entry;
169 unsigned long irq1; 170 unsigned long irq1;
170 int cpu; 171 int cpu;
@@ -180,7 +181,8 @@ void trace_hw_branch(u64 from, u64 to)
180 if (atomic_inc_return(&tr->data[cpu]->disabled) != 1) 181 if (atomic_inc_return(&tr->data[cpu]->disabled) != 1)
181 goto out; 182 goto out;
182 183
183 event = trace_buffer_lock_reserve(tr, TRACE_HW_BRANCHES, 184 buf = tr->buffer;
185 event = trace_buffer_lock_reserve(buf, TRACE_HW_BRANCHES,
184 sizeof(*entry), 0, 0); 186 sizeof(*entry), 0, 0);
185 if (!event) 187 if (!event)
186 goto out; 188 goto out;
@@ -189,8 +191,8 @@ void trace_hw_branch(u64 from, u64 to)
189 entry->ent.type = TRACE_HW_BRANCHES; 191 entry->ent.type = TRACE_HW_BRANCHES;
190 entry->from = from; 192 entry->from = from;
191 entry->to = to; 193 entry->to = to;
192 if (!filter_check_discard(call, entry, tr->buffer, event)) 194 if (!filter_check_discard(call, entry, buf, event))
193 trace_buffer_unlock_commit(tr, event, 0, 0); 195 trace_buffer_unlock_commit(buf, event, 0, 0);
194 196
195 out: 197 out:
196 atomic_dec(&tr->data[cpu]->disabled); 198 atomic_dec(&tr->data[cpu]->disabled);
diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c
index f572f44c6e1e..ed17565826b0 100644
--- a/kernel/trace/trace_output.c
+++ b/kernel/trace/trace_output.c
@@ -486,16 +486,18 @@ int trace_print_lat_fmt(struct trace_seq *s, struct trace_entry *entry)
486 hardirq ? 'h' : softirq ? 's' : '.')) 486 hardirq ? 'h' : softirq ? 's' : '.'))
487 return 0; 487 return 0;
488 488
489 if (entry->lock_depth < 0) 489 if (entry->preempt_count)
490 ret = trace_seq_putc(s, '.'); 490 ret = trace_seq_printf(s, "%x", entry->preempt_count);
491 else 491 else
492 ret = trace_seq_printf(s, "%d", entry->lock_depth); 492 ret = trace_seq_putc(s, '.');
493
493 if (!ret) 494 if (!ret)
494 return 0; 495 return 0;
495 496
496 if (entry->preempt_count) 497 if (entry->lock_depth < 0)
497 return trace_seq_printf(s, "%x", entry->preempt_count); 498 return trace_seq_putc(s, '.');
498 return trace_seq_putc(s, '.'); 499
500 return trace_seq_printf(s, "%d", entry->lock_depth);
499} 501}
500 502
501static int 503static int
@@ -883,7 +885,7 @@ static int trace_ctxwake_raw(struct trace_iterator *iter, char S)
883 trace_assign_type(field, iter->ent); 885 trace_assign_type(field, iter->ent);
884 886
885 if (!S) 887 if (!S)
886 task_state_char(field->prev_state); 888 S = task_state_char(field->prev_state);
887 T = task_state_char(field->next_state); 889 T = task_state_char(field->next_state);
888 if (!trace_seq_printf(&iter->seq, "%d %d %c %d %d %d %c\n", 890 if (!trace_seq_printf(&iter->seq, "%d %d %c %d %d %d %c\n",
889 field->prev_pid, 891 field->prev_pid,
@@ -918,7 +920,7 @@ static int trace_ctxwake_hex(struct trace_iterator *iter, char S)
918 trace_assign_type(field, iter->ent); 920 trace_assign_type(field, iter->ent);
919 921
920 if (!S) 922 if (!S)
921 task_state_char(field->prev_state); 923 S = task_state_char(field->prev_state);
922 T = task_state_char(field->next_state); 924 T = task_state_char(field->next_state);
923 925
924 SEQ_PUT_HEX_FIELD_RET(s, field->prev_pid); 926 SEQ_PUT_HEX_FIELD_RET(s, field->prev_pid);
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-27 07:25:44 -0500