Re: [PATCH 2/7] md: fix a race condition when stopping the sync thread

[Yu Kuai]

Hi,

在 2024/01/18 2:18, Mikulas Patocka 写道:

Note that md_wakeup_thread_directly is racy - it will do nothing if the
thread is already running or it may cause spurious wake-up if the thread
is blocked in another subsystem.

No, as the comment said, md_wakeup_thread_directly() is just to prevent
that md_wakeup_thread() can't wake up md_do_sync() if it's waiting for
metadata update.

In order to eliminate the race condition, we will retry
md_wakeup_thread_directly after 0.1 seconds.

And what you changed is not md_wakeup_thread_directly() at all.

Thanks,
Kuai

Signed-off-by: Mikulas Patocka <mpato...@redhat.com>
Cc: sta...@vger.kernel.org      # v6.7
Fixes: 130443d60b1b ("md: refactor idle/frozen_sync_thread() to fix deadlock")

---
  drivers/md/md.c |    7 +++++--
  1 file changed, 5 insertions(+), 2 deletions(-)

Index: linux-2.6/drivers/md/md.c
===================================================================
--- linux-2.6.orig/drivers/md/md.c
+++ linux-2.6/drivers/md/md.c
@@ -4889,6 +4889,7 @@ static void stop_sync_thread(struct mdde

mddev_unlock(mddev); +retry:

        set_bit(MD_RECOVERY_INTR, &mddev->recovery);
        /*
         * Thread might be blocked waiting for metadata update which will now
@@ -4898,9 +4899,11 @@ static void stop_sync_thread(struct mdde
        if (work_pending(&mddev->sync_work))
                flush_work(&mddev->sync_work);

- wait_event(resync_wait,

+       if (!wait_event_timeout(resync_wait,
                   !test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
-                  (check_seq && sync_seq != atomic_read(&mddev->sync_seq)));
+                  (check_seq && sync_seq != atomic_read(&mddev->sync_seq)),
+                  HZ / 10))
+               goto retry;

if (locked)

                mddev_lock_nointr(mddev);

.