* [RFC V2 PATCH 0/3] timer: patchset focus on del_timer_sync()
@ 2010-08-25 13:45 Yong Zhang
2010-08-25 13:45 ` [RFC V2 PATCH 1/3] timer: make try_to_del_timer_sync() safe on both hardirq context and UP Yong Zhang
` (2 more replies)
0 siblings, 3 replies; 6+ messages in thread
From: Yong Zhang @ 2010-08-25 13:45 UTC (permalink / raw)
To: linux-kernel; +Cc: tglx, akpm, mingo, oleg, peterz
From: Yong Zhang <yong.zhang@windriver.com>
This is inspired by http://lkml.org/lkml/2010/8/16/291
which catch a lockdep false positive on fake_timer_lock.
When I go into del_timer_sync(), but don't find anything
which prevent del_timer_sync() from using in softirq context,
and indeed it's been used in softirq for some timer(such as
__dst_free()).
Thus, tell others it can't be used in softirq context, and
teach lockdep about that. It's realized by patch-0002.
Your comments are very appreciated.
Thanks,
Yong
V2 <- V1
Update patch 0001, the new version make try_to_del_timer_sync()
safe on both hardirq context and UP. Thanks Oleg.
---
Yong Zhang (3):
timer: make try_to_del_timer_sync() safe on both hardirq context and UP
timer: del_timer_sync() can be used in softirq context
timer: warn when del_timer_sync() used in hardirq context
include/linux/timer.h | 4 ++--
kernel/timer.c | 20 +++++++++-----------
2 files changed, 11 insertions(+), 13 deletions(-)
^ permalink raw reply [flat|nested] 6+ messages in thread* [RFC V2 PATCH 1/3] timer: make try_to_del_timer_sync() safe on both hardirq context and UP
2010-08-25 13:45 [RFC V2 PATCH 0/3] timer: patchset focus on del_timer_sync() Yong Zhang
@ 2010-08-25 13:45 ` Yong Zhang
2010-08-25 18:10 ` Oleg Nesterov
2010-08-25 13:45 ` [RFC V2 PATCH 2/3] timer: del_timer_sync() can be used in softirq context Yong Zhang
2010-08-25 13:45 ` [RFC V2 PATCH 3/3] timer: warn when del_timer_sync() used in hardirq context Yong Zhang
2 siblings, 1 reply; 6+ messages in thread
From: Yong Zhang @ 2010-08-25 13:45 UTC (permalink / raw)
To: linux-kernel; +Cc: tglx, akpm, mingo, oleg, peterz
From: Yong Zhang <yong.zhang@windriver.com>
In commit fd450b7318b75343fd76b3d95416853e34e72c95, it was saying
try_to_del_timer_sync() can be used in interrupt context.
But because base->running_timer is SMP special, this lead to
unsymmetry try_to_del_timer_sync() on UP and SMP.
We can make running_timer live on everywhere, then try_to_del_timer_sync()
is also UP safe and can be used in hardirq context.
Signed-off-by: Yong Zhang <yong.zhang0@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Oleg Nesterov <oleg@redhat.com>
---
Hi Oleg,
How do you think about this one?
Yong
include/linux/timer.h | 4 ++--
kernel/timer.c | 6 +-----
2 files changed, 3 insertions(+), 7 deletions(-)
diff --git a/include/linux/timer.h b/include/linux/timer.h
index 38cf093..225c4ce 100644
--- a/include/linux/timer.h
+++ b/include/linux/timer.h
@@ -248,11 +248,11 @@ static inline void timer_stats_timer_clear_start_info(struct timer_list *timer)
extern void add_timer(struct timer_list *timer);
+extern int try_to_del_timer_sync(struct timer_list *timer);
+
#ifdef CONFIG_SMP
- extern int try_to_del_timer_sync(struct timer_list *timer);
extern int del_timer_sync(struct timer_list *timer);
#else
-# define try_to_del_timer_sync(t) del_timer(t)
# define del_timer_sync(t) del_timer(t)
#endif
diff --git a/kernel/timer.c b/kernel/timer.c
index 97bf05b..e031eb4 100644
--- a/kernel/timer.c
+++ b/kernel/timer.c
@@ -347,9 +347,7 @@ EXPORT_SYMBOL_GPL(set_timer_slack);
static inline void set_running_timer(struct tvec_base *base,
struct timer_list *timer)
{
-#ifdef CONFIG_SMP
base->running_timer = timer;
-#endif
}
static void internal_add_timer(struct tvec_base *base, struct timer_list *timer)
@@ -936,15 +934,12 @@ int del_timer(struct timer_list *timer)
}
EXPORT_SYMBOL(del_timer);
-#ifdef CONFIG_SMP
/**
* try_to_del_timer_sync - Try to deactivate a timer
* @timer: timer do del
*
* This function tries to deactivate a timer. Upon successful (ret >= 0)
* exit the timer is not queued and the handler is not running on any CPU.
- *
- * It must not be called from interrupt contexts.
*/
int try_to_del_timer_sync(struct timer_list *timer)
{
@@ -973,6 +968,7 @@ out:
}
EXPORT_SYMBOL(try_to_del_timer_sync);
+#ifdef CONFIG_SMP
/**
* del_timer_sync - deactivate a timer and wait for the handler to finish.
* @timer: the timer to be deactivated
--
1.7.0.4
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [RFC V2 PATCH 1/3] timer: make try_to_del_timer_sync() safe on both hardirq context and UP
2010-08-25 13:45 ` [RFC V2 PATCH 1/3] timer: make try_to_del_timer_sync() safe on both hardirq context and UP Yong Zhang
@ 2010-08-25 18:10 ` Oleg Nesterov
0 siblings, 0 replies; 6+ messages in thread
From: Oleg Nesterov @ 2010-08-25 18:10 UTC (permalink / raw)
To: Yong Zhang; +Cc: linux-kernel, tglx, akpm, mingo, peterz
On 08/25, Yong Zhang wrote:
>
> From: Yong Zhang <yong.zhang@windriver.com>
>
> In commit fd450b7318b75343fd76b3d95416853e34e72c95, it was saying
> try_to_del_timer_sync() can be used in interrupt context.
>
> But because base->running_timer is SMP special, this lead to
> unsymmetry try_to_del_timer_sync() on UP and SMP.
>
> We can make running_timer live on everywhere, then try_to_del_timer_sync()
> is also UP safe and can be used in hardirq context.
Acked-by: Oleg Nesterov <oleg@redhat.com>
> Signed-off-by: Yong Zhang <yong.zhang0@gmail.com>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Ingo Molnar <mingo@elte.hu>
> Cc: Oleg Nesterov <oleg@redhat.com>
> ---
> Hi Oleg,
>
> How do you think about this one?
>
> Yong
>
> include/linux/timer.h | 4 ++--
> kernel/timer.c | 6 +-----
> 2 files changed, 3 insertions(+), 7 deletions(-)
>
> diff --git a/include/linux/timer.h b/include/linux/timer.h
> index 38cf093..225c4ce 100644
> --- a/include/linux/timer.h
> +++ b/include/linux/timer.h
> @@ -248,11 +248,11 @@ static inline void timer_stats_timer_clear_start_info(struct timer_list *timer)
>
> extern void add_timer(struct timer_list *timer);
>
> +extern int try_to_del_timer_sync(struct timer_list *timer);
> +
> #ifdef CONFIG_SMP
> - extern int try_to_del_timer_sync(struct timer_list *timer);
> extern int del_timer_sync(struct timer_list *timer);
> #else
> -# define try_to_del_timer_sync(t) del_timer(t)
> # define del_timer_sync(t) del_timer(t)
> #endif
>
> diff --git a/kernel/timer.c b/kernel/timer.c
> index 97bf05b..e031eb4 100644
> --- a/kernel/timer.c
> +++ b/kernel/timer.c
> @@ -347,9 +347,7 @@ EXPORT_SYMBOL_GPL(set_timer_slack);
> static inline void set_running_timer(struct tvec_base *base,
> struct timer_list *timer)
> {
> -#ifdef CONFIG_SMP
> base->running_timer = timer;
> -#endif
> }
>
> static void internal_add_timer(struct tvec_base *base, struct timer_list *timer)
> @@ -936,15 +934,12 @@ int del_timer(struct timer_list *timer)
> }
> EXPORT_SYMBOL(del_timer);
>
> -#ifdef CONFIG_SMP
> /**
> * try_to_del_timer_sync - Try to deactivate a timer
> * @timer: timer do del
> *
> * This function tries to deactivate a timer. Upon successful (ret >= 0)
> * exit the timer is not queued and the handler is not running on any CPU.
> - *
> - * It must not be called from interrupt contexts.
> */
> int try_to_del_timer_sync(struct timer_list *timer)
> {
> @@ -973,6 +968,7 @@ out:
> }
> EXPORT_SYMBOL(try_to_del_timer_sync);
>
> +#ifdef CONFIG_SMP
> /**
> * del_timer_sync - deactivate a timer and wait for the handler to finish.
> * @timer: the timer to be deactivated
> --
> 1.7.0.4
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* [RFC V2 PATCH 2/3] timer: del_timer_sync() can be used in softirq context
2010-08-25 13:45 [RFC V2 PATCH 0/3] timer: patchset focus on del_timer_sync() Yong Zhang
2010-08-25 13:45 ` [RFC V2 PATCH 1/3] timer: make try_to_del_timer_sync() safe on both hardirq context and UP Yong Zhang
@ 2010-08-25 13:45 ` Yong Zhang
2010-08-25 13:45 ` [RFC V2 PATCH 3/3] timer: warn when del_timer_sync() used in hardirq context Yong Zhang
2 siblings, 0 replies; 6+ messages in thread
From: Yong Zhang @ 2010-08-25 13:45 UTC (permalink / raw)
To: linux-kernel; +Cc: tglx, akpm, mingo, oleg, peterz
From: Yong Zhang <yong.zhang@windriver.com>
Actually we have used del_timer_sync() in softirq context for a long
time, such as: __dst_free()::cancel_delayed_work() and maybe the
DEBUG_OBJECTS_TIMERS fundamental code.
So change the comments of it to warn on hardirq context only,
and make lockdep know about this change.
Signed-off-by: Yong Zhang <yong.zhang0@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
---
kernel/timer.c | 8 +++-----
1 files changed, 3 insertions(+), 5 deletions(-)
diff --git a/kernel/timer.c b/kernel/timer.c
index 55b6a2d..08c9559 100644
--- a/kernel/timer.c
+++ b/kernel/timer.c
@@ -981,7 +981,7 @@ EXPORT_SYMBOL(try_to_del_timer_sync);
*
* Synchronization rules: Callers must prevent restarting of the timer,
* otherwise this function is meaningless. It must not be called from
- * interrupt contexts. The caller must not hold locks which would prevent
+ * hardirq contexts. The caller must not hold locks which would prevent
* completion of the timer's handler. The timer's handler must not call
* add_timer_on(). Upon exit the timer is not queued and the handler is
* not running on any CPU.
@@ -991,12 +991,10 @@ EXPORT_SYMBOL(try_to_del_timer_sync);
int del_timer_sync(struct timer_list *timer)
{
#ifdef CONFIG_LOCKDEP
- unsigned long flags;
-
- local_irq_save(flags);
+ local_bh_disable();
lock_map_acquire(&timer->lockdep_map);
lock_map_release(&timer->lockdep_map);
- local_irq_restore(flags);
+ local_bh_enable();
#endif
for (;;) {
--
1.7.0.4
^ permalink raw reply related [flat|nested] 6+ messages in thread* [RFC V2 PATCH 3/3] timer: warn when del_timer_sync() used in hardirq context
2010-08-25 13:45 [RFC V2 PATCH 0/3] timer: patchset focus on del_timer_sync() Yong Zhang
2010-08-25 13:45 ` [RFC V2 PATCH 1/3] timer: make try_to_del_timer_sync() safe on both hardirq context and UP Yong Zhang
2010-08-25 13:45 ` [RFC V2 PATCH 2/3] timer: del_timer_sync() can be used in softirq context Yong Zhang
@ 2010-08-25 13:45 ` Yong Zhang
2010-08-25 18:22 ` Oleg Nesterov
2 siblings, 1 reply; 6+ messages in thread
From: Yong Zhang @ 2010-08-25 13:45 UTC (permalink / raw)
To: linux-kernel; +Cc: tglx, akpm, mingo, oleg, peterz
From: Yong Zhang <yong.zhang@windriver.com>
Add explict warning to prevent del_timer_sync() from using
in hardirq context.
Signed-off-by: Yong Zhang <yong.zhang0@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Ingo Molnar <mingo@elte.hu>
---
kernel/timer.c | 6 +++++-
1 files changed, 5 insertions(+), 1 deletions(-)
diff --git a/kernel/timer.c b/kernel/timer.c
index 08c9559..1bc4b4a 100644
--- a/kernel/timer.c
+++ b/kernel/timer.c
@@ -996,7 +996,11 @@ int del_timer_sync(struct timer_list *timer)
lock_map_release(&timer->lockdep_map);
local_bh_enable();
#endif
-
+ /*
+ * don't use it in hardirq context, because it
+ * could lead to deadlock.
+ */
+ WARN_ON(in_irq());
for (;;) {
int ret = try_to_del_timer_sync(timer);
if (ret >= 0)
--
1.7.0.4
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [RFC V2 PATCH 3/3] timer: warn when del_timer_sync() used in hardirq context
2010-08-25 13:45 ` [RFC V2 PATCH 3/3] timer: warn when del_timer_sync() used in hardirq context Yong Zhang
@ 2010-08-25 18:22 ` Oleg Nesterov
0 siblings, 0 replies; 6+ messages in thread
From: Oleg Nesterov @ 2010-08-25 18:22 UTC (permalink / raw)
To: Yong Zhang; +Cc: linux-kernel, tglx, akpm, mingo, peterz
On 08/25, Yong Zhang wrote:
>
> From: Yong Zhang <yong.zhang@windriver.com>
>
> Add explict warning to prevent del_timer_sync() from using
> in hardirq context.
Agreed, del_timer_sync() from softirq should be OK.
> Signed-off-by: Yong Zhang <yong.zhang0@gmail.com>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Ingo Molnar <mingo@elte.hu>
> ---
> kernel/timer.c | 6 +++++-
> 1 files changed, 5 insertions(+), 1 deletions(-)
>
> diff --git a/kernel/timer.c b/kernel/timer.c
> index 08c9559..1bc4b4a 100644
> --- a/kernel/timer.c
> +++ b/kernel/timer.c
> @@ -996,7 +996,11 @@ int del_timer_sync(struct timer_list *timer)
> lock_map_release(&timer->lockdep_map);
> local_bh_enable();
> #endif
> -
> + /*
> + * don't use it in hardirq context, because it
> + * could lead to deadlock.
> + */
> + WARN_ON(in_irq());
> for (;;) {
> int ret = try_to_del_timer_sync(timer);
> if (ret >= 0)
> --
> 1.7.0.4
>
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2010-08-25 18:25 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-08-25 13:45 [RFC V2 PATCH 0/3] timer: patchset focus on del_timer_sync() Yong Zhang
2010-08-25 13:45 ` [RFC V2 PATCH 1/3] timer: make try_to_del_timer_sync() safe on both hardirq context and UP Yong Zhang
2010-08-25 18:10 ` Oleg Nesterov
2010-08-25 13:45 ` [RFC V2 PATCH 2/3] timer: del_timer_sync() can be used in softirq context Yong Zhang
2010-08-25 13:45 ` [RFC V2 PATCH 3/3] timer: warn when del_timer_sync() used in hardirq context Yong Zhang
2010-08-25 18:22 ` Oleg Nesterov
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox