aboutsummaryrefslogtreecommitdiffstats
path: root/kernel/timer.c
diff options
context:
space:
mode:
authorTejun Heo <tj@kernel.org>2012-08-08 14:10:28 -0400
committerThomas Gleixner <tglx@linutronix.de>2012-08-21 10:28:31 -0400
commitc5f66e99b7cb091e3d51ae8e8156892e8feb7fa3 (patch)
treeb73b64ceabad685abd9a2bc24f0258d9dbfae367 /kernel/timer.c
parentfc683995a6c4e604d62ab9a488ac2c1ba94fa868 (diff)
timer: Implement TIMER_IRQSAFE
Timer internals are protected with irq-safe locks but timer execution isn't, so a timer being dequeued for execution and its execution aren't atomic against IRQs. This makes it impossible to wait for its completion from IRQ handlers and difficult to shoot down a timer from IRQ handlers. This issue caused some issues for delayed_work interface. Because there's no way to reliably shoot down delayed_work->timer from IRQ handlers, __cancel_delayed_work() can't share the logic to steal the target delayed_work with cancel_delayed_work_sync(), and can only steal delayed_works which are on queued on timer. Similarly, the pending mod_delayed_work() can't be used from IRQ handlers. This patch adds a new timer flag TIMER_IRQSAFE, which makes the timer to be executed without enabling IRQ after dequeueing such that its dequeueing and execution are atomic against IRQ handlers. This makes it safe to wait for the timer's completion from IRQ handlers, for example, using del_timer_sync(). It can never be executing on the local CPU and if executing on other CPUs it won't be interrupted until done. This will enable simplifying delayed_work cancel/mod interface. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: torvalds@linux-foundation.org Cc: peterz@infradead.org Link: http://lkml.kernel.org/r/1344449428-24962-5-git-send-email-tj@kernel.org Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Diffstat (limited to 'kernel/timer.c')
-rw-r--r--kernel/timer.c35
1 files changed, 24 insertions, 11 deletions
diff --git a/kernel/timer.c b/kernel/timer.c
index 8d185a1677cc..706fe4c53e82 100644
--- a/kernel/timer.c
+++ b/kernel/timer.c
@@ -95,6 +95,11 @@ static inline unsigned int tbase_get_deferrable(struct tvec_base *base)
95 return ((unsigned int)(unsigned long)base & TIMER_DEFERRABLE); 95 return ((unsigned int)(unsigned long)base & TIMER_DEFERRABLE);
96} 96}
97 97
98static inline unsigned int tbase_get_irqsafe(struct tvec_base *base)
99{
100 return ((unsigned int)(unsigned long)base & TIMER_IRQSAFE);
101}
102
98static inline struct tvec_base *tbase_get_base(struct tvec_base *base) 103static inline struct tvec_base *tbase_get_base(struct tvec_base *base)
99{ 104{
100 return ((struct tvec_base *)((unsigned long)base & ~TIMER_FLAG_MASK)); 105 return ((struct tvec_base *)((unsigned long)base & ~TIMER_FLAG_MASK));
@@ -1002,14 +1007,14 @@ EXPORT_SYMBOL(try_to_del_timer_sync);
1002 * 1007 *
1003 * Synchronization rules: Callers must prevent restarting of the timer, 1008 * Synchronization rules: Callers must prevent restarting of the timer,
1004 * otherwise this function is meaningless. It must not be called from 1009 * otherwise this function is meaningless. It must not be called from
1005 * interrupt contexts. The caller must not hold locks which would prevent 1010 * interrupt contexts unless the timer is an irqsafe one. The caller must
1006 * completion of the timer's handler. The timer's handler must not call 1011 * not hold locks which would prevent completion of the timer's
1007 * add_timer_on(). Upon exit the timer is not queued and the handler is 1012 * handler. The timer's handler must not call add_timer_on(). Upon exit the
1008 * not running on any CPU. 1013 * timer is not queued and the handler is not running on any CPU.
1009 * 1014 *
1010 * Note: You must not hold locks that are held in interrupt context 1015 * Note: For !irqsafe timers, you must not hold locks that are held in
1011 * while calling this function. Even if the lock has nothing to do 1016 * interrupt context while calling this function. Even if the lock has
1012 * with the timer in question. Here's why: 1017 * nothing to do with the timer in question. Here's why:
1013 * 1018 *
1014 * CPU0 CPU1 1019 * CPU0 CPU1
1015 * ---- ---- 1020 * ---- ----
@@ -1046,7 +1051,7 @@ int del_timer_sync(struct timer_list *timer)
1046 * don't use it in hardirq context, because it 1051 * don't use it in hardirq context, because it
1047 * could lead to deadlock. 1052 * could lead to deadlock.
1048 */ 1053 */
1049 WARN_ON(in_irq()); 1054 WARN_ON(in_irq() && !tbase_get_irqsafe(timer->base));
1050 for (;;) { 1055 for (;;) {
1051 int ret = try_to_del_timer_sync(timer); 1056 int ret = try_to_del_timer_sync(timer);
1052 if (ret >= 0) 1057 if (ret >= 0)
@@ -1153,19 +1158,27 @@ static inline void __run_timers(struct tvec_base *base)
1153 while (!list_empty(head)) { 1158 while (!list_empty(head)) {
1154 void (*fn)(unsigned long); 1159 void (*fn)(unsigned long);
1155 unsigned long data; 1160 unsigned long data;
1161 bool irqsafe;
1156 1162
1157 timer = list_first_entry(head, struct timer_list,entry); 1163 timer = list_first_entry(head, struct timer_list,entry);
1158 fn = timer->function; 1164 fn = timer->function;
1159 data = timer->data; 1165 data = timer->data;
1166 irqsafe = tbase_get_irqsafe(timer->base);
1160 1167
1161 timer_stats_account_timer(timer); 1168 timer_stats_account_timer(timer);
1162 1169
1163 base->running_timer = timer; 1170 base->running_timer = timer;
1164 detach_expired_timer(timer, base); 1171 detach_expired_timer(timer, base);
1165 1172
1166 spin_unlock_irq(&base->lock); 1173 if (irqsafe) {
1167 call_timer_fn(timer, fn, data); 1174 spin_unlock(&base->lock);
1168 spin_lock_irq(&base->lock); 1175 call_timer_fn(timer, fn, data);
1176 spin_lock(&base->lock);
1177 } else {
1178 spin_unlock_irq(&base->lock);
1179 call_timer_fn(timer, fn, data);
1180 spin_lock_irq(&base->lock);
1181 }
1169 } 1182 }
1170 } 1183 }
1171 base->running_timer = NULL; 1184 base->running_timer = NULL;