md: don't check MD_SB_CHANGE_CLEAN in md_allow_write

Only MD_SB_CHANGE_PENDING should be used to wait for transition from
clean to dirty. Checking also MD_SB_CHANGE_CLEAN is unnecessary and can
race with e.g. md_do_sync(). This sporadically causes a hang when
changing consistency policy during resync:

INFO: task mdadm:6183 blocked for more than 30 seconds.
      Not tainted 4.14.0-rc3+ #391
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
mdadm           D12752  6183   6022 0x00000000
Call Trace:
 __schedule+0x93f/0x990
 schedule+0x6b/0x90
 md_allow_write+0x100/0x130 [md_mod]
 ? do_wait_intr_irq+0x90/0x90
 resize_stripes+0x3a/0x5b0 [raid456]
 ? kernfs_fop_write+0xbe/0x180
 raid5_change_consistency_policy+0xa6/0x200 [raid456]
 consistency_policy_store+0x2e/0x70 [md_mod]
 md_attr_store+0x90/0xc0 [md_mod]
 sysfs_kf_write+0x42/0x50
 kernfs_fop_write+0x119/0x180
 __vfs_write+0x28/0x110
 ? rcu_sync_lockdep_assert+0x12/0x60
 ? __sb_start_write+0x15a/0x1c0
 ? vfs_write+0xa3/0x1a0
 vfs_write+0xb4/0x1a0
 SyS_write+0x49/0xa0
 entry_SYSCALL_64_fastpath+0x18/0xad

Fixes: 2214c260c72b ("md: don't return -EAGAIN in md_allow_write for external metadata arrays")
Cc: <stable@vger.kernel.org>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Shaohua Li <shli@fb.com>

diff --git a/drivers/md/md.c b/drivers/md/md.c
index 5a0ec1d1a6e8c569a82eb44ebfce408a615af99a..2bf4cc41b4f81756ac6243fc4e2e5b4c58f44944 100644 (file)
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -8125,7 +8125,6 @@ void md_allow_write(struct mddev *mddev)
                sysfs_notify_dirent_safe(mddev->sysfs_state);
                /* wait for the dirty state to be recorded in the metadata */
                wait_event(mddev->sb_wait,
-                          !test_bit(MD_SB_CHANGE_CLEAN, &mddev->sb_flags) &&
                           !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
        } else
                spin_unlock(&mddev->lock);