Re: [PATCH v5 03/14] md: make sure md_do_sync() will set MD_RECOVERY_DONE

[Xiao Ni <xni@xxxxxxxxxx>]

On Thu, Feb 1, 2024 at 5:30 PM Yu Kuai <yukuai1@xxxxxxxxxxxxxxx> wrote:
>
> From: Yu Kuai <yukuai3@xxxxxxxxxx>
>
> stop_sync_thread() will interrupt md_do_sync(), and md_do_sync() must
> set MD_RECOVERY_DONE, so that follow up md_check_recovery() will
> unregister sync_thread, clear MD_RECOVERY_RUNNING and wake up
> stop_sync_thread().
>
> If MD_RECOVERY_WAIT is set or the array is read-only, md_do_sync() will
> return without setting MD_RECOVERY_DONE, and after commit f52f5c71f3d4
> ("md: fix stopping sync thread"), dm-raid switch from
> md_reap_sync_thread() to stop_sync_thread() to unregister sync_thread
> from md_stop() and md_stop_writes(), causing the test
> shell/lvconvert-raid-reshape.sh hang.
>
> We shouldn't switch back to md_reap_sync_thread() because it's
> problematic in the first place. Fix the problem by making sure
> md_do_sync() will set MD_RECOVERY_DONE.
>
> Reported-by: Mikulas Patocka <mpatocka@xxxxxxxxxx>
> Closes: https://lore.kernel.org/all/ece2b06f-d647-6613-a534-ff4c9bec1142@xxxxxxxxxx/
> Fixes: d5d885fd514f ("md: introduce new personality funciton start()")
> Fixes: 5fd6c1dce06e ("[PATCH] md: allow checkpoint of recovery with version-1 superblock")
> Fixes: f52f5c71f3d4 ("md: fix stopping sync thread")
> Signed-off-by: Yu Kuai <yukuai3@xxxxxxxxxx>
> ---
>  drivers/md/md.c | 12 ++++++++----
>  1 file changed, 8 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/md/md.c b/drivers/md/md.c
> index 6906d023f1d6..c65dfd156090 100644
> --- a/drivers/md/md.c
> +++ b/drivers/md/md.c
> @@ -8788,12 +8788,16 @@ void md_do_sync(struct md_thread *thread)
>         int ret;
>
>         /* just incase thread restarts... */
> -       if (test_bit(MD_RECOVERY_DONE, &mddev->recovery) ||
> -           test_bit(MD_RECOVERY_WAIT, &mddev->recovery))
> +       if (test_bit(MD_RECOVERY_DONE, &mddev->recovery))
>                 return;
> -       if (!md_is_rdwr(mddev)) {/* never try to sync a read-only array */
> +
> +       if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
> +               goto skip;
> +
> +       if (test_bit(MD_RECOVERY_WAIT, &mddev->recovery) ||
> +           !md_is_rdwr(mddev)) {/* never try to sync a read-only array */
>                 set_bit(MD_RECOVERY_INTR, &mddev->recovery);
> -               return;
> +               goto skip;
>         }

Hi all

I have a question here. The codes above means if MD_RECOVERY_WAIT is
set, it sets MD_RECOVERY_INTR. If so, the sync thread can't happen.
But from the codes in md_start function:

                set_bit(MD_RECOVERY_WAIT, &mddev->recovery);
                md_wakeup_thread(mddev->thread);
                ret = mddev->pers->start(mddev);
                clear_bit(MD_RECOVERY_WAIT, &mddev->recovery);
                md_wakeup_thread(mddev->sync_thread);

MD_RECOVERY_WAIT means "it'll run sync thread later not interrupt it".
I guess this patch can introduce a new bug for raid5 journal?

And to resolve this deadlock, we can use this patch:

--- a/drivers/md/dm-raid.c
+++ b/drivers/md/dm-raid.c
@@ -3796,8 +3796,10 @@ static void raid_postsuspend(struct dm_target *ti)
        struct raid_set *rs = ti->private;

        if (!test_and_set_bit(RT_FLAG_RS_SUSPENDED, &rs->runtime_flags)) {
+               if (test_bit(MD_RECOVERY_WAIT, &rs->md.recovery))
+                       clear_bit(MD_RECOVERY_WAIT, &rs->md.recovery);

Regards
Xiao
>
>         if (mddev_is_clustered(mddev)) {
> --
> 2.39.2
>