[PATCH] md: don't unregister sync_thread with reconfig_mutex held

[PATCH] md: don't unregister sync_thread with reconfig_mutex held

[Guoqing Jiang]

Unregister sync_thread doesn't need to hold reconfig_mutex since it
doesn't reconfigure array.

And it could cause deadlock problem for raid5 as follows:

1. process A tried to reap sync thread with reconfig_mutex held after echo
   idle to sync_action.
2. raid5 sync thread was blocked if there were too many active stripes.
3. SB_CHANGE_PENDING was set (because of write IO comes from upper layer)
   which causes the number of active stripes can't be decreased.
4. SB_CHANGE_PENDING can't be cleared since md_check_recovery was not able
   to hold reconfig_mutex.

More details in the link:
issu://lore.kernel.org/linux-raid/5ed54ffc-ce82-bf66-4eff-390cb23bc1ac@molgen.mpg.de/T/#t

Reported-and-tested-by: Donald Buczek <buczek@molgen.mpg.de>
Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
---
 drivers/md/md.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/md/md.c b/drivers/md/md.c
index ca40942..eec8c27 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -9365,13 +9365,18 @@ void md_check_recovery(struct mddev *mddev)
 EXPORT_SYMBOL(md_check_recovery);
 
 void md_reap_sync_thread(struct mddev *mddev)
+	__releases(&mddev->reconfig_mutex)
+	__acquires(&mddev->reconfig_mutex)
+
 {
 	struct md_rdev *rdev;
 	sector_t old_dev_sectors = mddev->dev_sectors;
 	bool is_reshaped = false;
 
 	/* resync has finished, collect result */
+	mddev_unlock(mddev);
 	md_unregister_thread(&mddev->sync_thread);
+	mddev_lock_nointr(mddev);
 	if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
 	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
 	    mddev->degraded != mddev->raid_disks) {
-- 
2.7.4

Re: [PATCH] md: don't unregister sync_thread with reconfig_mutex held

[Song Liu]

On Tue, Feb 9, 2021 at 6:22 PM Guoqing Jiang
<guoqing.jiang@cloud.ionos.com> wrote:
>
> Unregister sync_thread doesn't need to hold reconfig_mutex since it
> doesn't reconfigure array.
>
> And it could cause deadlock problem for raid5 as follows:
>
> 1. process A tried to reap sync thread with reconfig_mutex held after echo
>    idle to sync_action.
> 2. raid5 sync thread was blocked if there were too many active stripes.
> 3. SB_CHANGE_PENDING was set (because of write IO comes from upper layer)
>    which causes the number of active stripes can't be decreased.
> 4. SB_CHANGE_PENDING can't be cleared since md_check_recovery was not able
>    to hold reconfig_mutex.
>
> More details in the link:
> issu://lore.kernel.org/linux-raid/5ed54ffc-ce82-bf66-4eff-390cb23bc1ac@molgen.mpg.de/T/#t
>
> Reported-and-tested-by: Donald Buczek <buczek@molgen.mpg.de>
> Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>

Thanks for debugging the issue. However, I am not sure whether this is
the proper
fix. For example, would this break dm-raid.c:raid_message()? IIUC,
raid_message()
calls md_reap_sync_thread() without holding reconfigure_mutex, no?

Thanks,
Song

> ---
>  drivers/md/md.c | 5 +++++
>  1 file changed, 5 insertions(+)
>
> diff --git a/drivers/md/md.c b/drivers/md/md.c
> index ca40942..eec8c27 100644
> --- a/drivers/md/md.c
> +++ b/drivers/md/md.c
> @@ -9365,13 +9365,18 @@ void md_check_recovery(struct mddev *mddev)
>  EXPORT_SYMBOL(md_check_recovery);
>
>  void md_reap_sync_thread(struct mddev *mddev)
> +       __releases(&mddev->reconfig_mutex)
> +       __acquires(&mddev->reconfig_mutex)
> +
>  {
>         struct md_rdev *rdev;
>         sector_t old_dev_sectors = mddev->dev_sectors;
>         bool is_reshaped = false;
>
>         /* resync has finished, collect result */
> +       mddev_unlock(mddev);
>         md_unregister_thread(&mddev->sync_thread);
> +       mddev_lock_nointr(mddev);
>         if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
>             !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
>             mddev->degraded != mddev->raid_disks) {
> --
> 2.7.4
>

Re: [PATCH] md: don't unregister sync_thread with reconfig_mutex held

[Jack Wang]

Song Liu <song@kernel.org> 于2021年2月11日周四 上午8:31写道：
>
> On Tue, Feb 9, 2021 at 6:22 PM Guoqing Jiang
> <guoqing.jiang@cloud.ionos.com> wrote:
> >
> > Unregister sync_thread doesn't need to hold reconfig_mutex since it
> > doesn't reconfigure array.
> >
> > And it could cause deadlock problem for raid5 as follows:
> >
> > 1. process A tried to reap sync thread with reconfig_mutex held after echo
> >    idle to sync_action.
> > 2. raid5 sync thread was blocked if there were too many active stripes.
> > 3. SB_CHANGE_PENDING was set (because of write IO comes from upper layer)
> >    which causes the number of active stripes can't be decreased.
> > 4. SB_CHANGE_PENDING can't be cleared since md_check_recovery was not able
> >    to hold reconfig_mutex.
> >
> > More details in the link:
> > issu://lore.kernel.org/linux-raid/5ed54ffc-ce82-bf66-4eff-390cb23bc1ac@molgen.mpg.de/T/#t
> >
> > Reported-and-tested-by: Donald Buczek <buczek@molgen.mpg.de>
> > Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
>
> Thanks for debugging the issue. However, I am not sure whether this is
> the proper
> fix. For example, would this break dm-raid.c:raid_message()? IIUC,
> raid_message()
> calls md_reap_sync_thread() without holding reconfigure_mutex, no?
>
> Thanks,
> Song.
right.
A simple solution would be add a parameter to md_reap_sync_thread to
indicate if a reconfigure_mutex lock is held.

Regards!
Jack
>
> > ---
> >  drivers/md/md.c | 5 +++++
> >  1 file changed, 5 insertions(+)
> >
> > diff --git a/drivers/md/md.c b/drivers/md/md.c
> > index ca40942..eec8c27 100644
> > --- a/drivers/md/md.c
> > +++ b/drivers/md/md.c
> > @@ -9365,13 +9365,18 @@ void md_check_recovery(struct mddev *mddev)
> >  EXPORT_SYMBOL(md_check_recovery);
> >
> >  void md_reap_sync_thread(struct mddev *mddev)
> > +       __releases(&mddev->reconfig_mutex)
> > +       __acquires(&mddev->reconfig_mutex)
> > +
> >  {
> >         struct md_rdev *rdev;
> >         sector_t old_dev_sectors = mddev->dev_sectors;
> >         bool is_reshaped = false;
> >
> >         /* resync has finished, collect result */
> > +       mddev_unlock(mddev);
> >         md_unregister_thread(&mddev->sync_thread);
> > +       mddev_lock_nointr(mddev);
> >         if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
> >             !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
> >             mddev->degraded != mddev->raid_disks) {
> > --
> > 2.7.4
> >

Re: [PATCH] md: don't unregister sync_thread with reconfig_mutex held

[Guoqing Jiang]

On 2/11/21 08:28, Song Liu wrote:
> On Tue, Feb 9, 2021 at 6:22 PM Guoqing Jiang
> <guoqing.jiang@cloud.ionos.com> wrote:
>>
>> Unregister sync_thread doesn't need to hold reconfig_mutex since it
>> doesn't reconfigure array.
>>
>> And it could cause deadlock problem for raid5 as follows:
>>
>> 1. process A tried to reap sync thread with reconfig_mutex held after echo
>>     idle to sync_action.
>> 2. raid5 sync thread was blocked if there were too many active stripes.
>> 3. SB_CHANGE_PENDING was set (because of write IO comes from upper layer)
>>     which causes the number of active stripes can't be decreased.
>> 4. SB_CHANGE_PENDING can't be cleared since md_check_recovery was not able
>>     to hold reconfig_mutex.
>>
>> More details in the link:
>> issu://lore.kernel.org/linux-raid/5ed54ffc-ce82-bf66-4eff-390cb23bc1ac@molgen.mpg.de/T/#t
>>
>> Reported-and-tested-by: Donald Buczek <buczek@molgen.mpg.de>
>> Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
> 
> Thanks for debugging the issue. However, I am not sure whether this is
> the proper
> fix. For example, would this break dm-raid.c:raid_message()? IIUC,
> raid_message()
> calls md_reap_sync_thread() without holding reconfigure_mutex, no?

Oops, I didn't notice dm-raid calls it though md did call it with 
reconfig_mutex held. But on the other side, it proves we don't need to 
call md_reap_sync_thread with the mutex held.

Thanks,
Guoqing