Offlining memory may block forever, waiting for kswapd() to wake up because
kswapd() does not check the event kthread->should_stop before sleeping.
The proper pattern, from Documentation/memory-barriers.txt, is:
--- waker ---
event_indicated = 1;
wake_up_process(event_daemon);
--- sleeper ---
for (;;) {
set_current_state(TASK_UNINTERRUPTIBLE);
if (event_indicated)
break;
schedule();
}
set_current_state() may be wrapped by:
prepare_to_wait();
In the kswapd() case, event_indicated is kthread->should_stop.
--- offlining memory (waker) ---
kswapd_stop()
kthread_stop()
kthread->should_stop = 1
wake_up_process()
wait_for_completion()
--- kswapd_try_to_sleep (sleeper) ---
kswapd_try_to_sleep()
prepare_to_wait()
.
.
schedule()
.
.
finish_wait()
The schedule() needs to be protected by a test of kthread->should_stop,
which is wrapped by kthread_should_stop().
Reproducer:
Do heavy file I/O in background.
Do a memory offline/online in a tight loop
Signed-off-by: Aaditya Kumar <[email protected]>
---
diff --git a/mm/vmscan.c b/mm/vmscan.c
index eeb3bc9..b60691e 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -2688,7 +2688,10 @@ static void kswapd_try_to_sleep(pg_data_t
*pgdat, int order, int classzone_idx)
* them before going back to sleep.
*/
set_pgdat_percpu_threshold(pgdat, calculate_normal_threshold);
- schedule();
+
+ if (!kthread_should_stop())
+ schedule();
+
set_pgdat_percpu_threshold(pgdat, calculate_pressure_threshold);
} else {
if (remaining)
On Wed, Jun 20, 2012 at 09:53:31PM +0530, Aaditya Kumar wrote:
> Offlining memory may block forever, waiting for kswapd() to wake up because
> kswapd() does not check the event kthread->should_stop before sleeping.
>
> The proper pattern, from Documentation/memory-barriers.txt, is:
> --- waker ---
> event_indicated = 1;
> wake_up_process(event_daemon);
>
> --- sleeper ---
> for (;;) {
> set_current_state(TASK_UNINTERRUPTIBLE);
> if (event_indicated)
> break;
> schedule();
> }
>
> set_current_state() may be wrapped by:
> prepare_to_wait();
>
> In the kswapd() case, event_indicated is kthread->should_stop.
> --- offlining memory (waker) ---
> kswapd_stop()
> kthread_stop()
> kthread->should_stop = 1
> wake_up_process()
> wait_for_completion()
>
>
> --- kswapd_try_to_sleep (sleeper) ---
> kswapd_try_to_sleep()
> prepare_to_wait()
> .
> .
> schedule()
> .
> .
> finish_wait()
>
> The schedule() needs to be protected by a test of kthread->should_stop,
> which is wrapped by kthread_should_stop().
>
> Reproducer:
> Do heavy file I/O in background.
> Do a memory offline/online in a tight loop
>
>
> Signed-off-by: Aaditya Kumar <[email protected]>
<formletter>
This is not the correct way to submit patches for inclusion in the
stable kernel tree. Please read Documentation/stable_kernel_rules.txt
for how to do this properly.
</formletter>
On 6/20/2012 12:23 PM, Aaditya Kumar wrote:
> Offlining memory may block forever, waiting for kswapd() to wake up because
> kswapd() does not check the event kthread->should_stop before sleeping.
>
> The proper pattern, from Documentation/memory-barriers.txt, is:
> --- waker ---
> event_indicated = 1;
> wake_up_process(event_daemon);
>
> --- sleeper ---
> for (;;) {
> set_current_state(TASK_UNINTERRUPTIBLE);
> if (event_indicated)
> break;
> schedule();
> }
>
> set_current_state() may be wrapped by:
> prepare_to_wait();
>
> In the kswapd() case, event_indicated is kthread->should_stop.
> --- offlining memory (waker) ---
Please avoid "---". This is used for a separator between a patch
description and code.
Other than that,
Acked-by: KOSAKI Motohiro <[email protected]>
> kswapd_stop()
> kthread_stop()
> kthread->should_stop = 1
> wake_up_process()
> wait_for_completion()
>
>
> --- kswapd_try_to_sleep (sleeper) ---
> kswapd_try_to_sleep()
> prepare_to_wait()
> .
> .
> schedule()
> .
> .
> finish_wait()
>
> The schedule() needs to be protected by a test of kthread->should_stop,
> which is wrapped by kthread_should_stop().
>
> Reproducer:
> Do heavy file I/O in background.
> Do a memory offline/online in a tight loop
>
>
> Signed-off-by: Aaditya Kumar <[email protected]>
>
> ---
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index eeb3bc9..b60691e 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -2688,7 +2688,10 @@ static void kswapd_try_to_sleep(pg_data_t
> *pgdat, int order, int classzone_idx)
> * them before going back to sleep.
> */
> set_pgdat_percpu_threshold(pgdat, calculate_normal_threshold);
> - schedule();
> +
> + if (!kthread_should_stop())
> + schedule();
> +
> set_pgdat_percpu_threshold(pgdat, calculate_pressure_threshold);
> } else {
> if (remaining)
>
On 06/21/2012 01:23 AM, Aaditya Kumar wrote:
> Offlining memory may block forever, waiting for kswapd() to wake up because
> kswapd() does not check the event kthread->should_stop before sleeping.
>
> The proper pattern, from Documentation/memory-barriers.txt, is:
> --- waker ---
> event_indicated = 1;
> wake_up_process(event_daemon);
>
> --- sleeper ---
> for (;;) {
> set_current_state(TASK_UNINTERRUPTIBLE);
> if (event_indicated)
> break;
> schedule();
> }
>
> set_current_state() may be wrapped by:
> prepare_to_wait();
>
> In the kswapd() case, event_indicated is kthread->should_stop.
> --- offlining memory (waker) ---
> kswapd_stop()
> kthread_stop()
> kthread->should_stop = 1
> wake_up_process()
> wait_for_completion()
>
>
> --- kswapd_try_to_sleep (sleeper) ---
> kswapd_try_to_sleep()
> prepare_to_wait()
> .
> .
> schedule()
> .
> .
> finish_wait()
>
> The schedule() needs to be protected by a test of kthread->should_stop,
> which is wrapped by kthread_should_stop().
>
> Reproducer:
> Do heavy file I/O in background.
> Do a memory offline/online in a tight loop
>
>
> Signed-off-by: Aaditya Kumar <[email protected]>
Reviewed-by: Minchan Kim <[email protected]>
Nitpick: We can remove kthread_should_stop check earlier in kswapd_try_to_sleep.
But it's no biggie. And I hope you change patch title
Title : Fix loss of kswapd wakeup in kswapd_stop
Description: Offlining memory may block forever because blah, blah, blah.
--
Kind regards,
Minchan Kim
On Wed, Jun 20, 2012 at 09:53:31PM +0530, Aaditya Kumar wrote:
> Offlining memory may block forever, waiting for kswapd() to wake up because
> kswapd() does not check the event kthread->should_stop before sleeping.
>
Acked-by: Mel Gorman <[email protected]>
--
Mel Gorman
SUSE Labs