md/raid5: avoid deadlock when raid5 array has unack badblocks during md_stop_writes.

When raid5 recovery hits a fresh badblock, this badblock will flagged as unack
badblock until md_update_sb() is called.
But md_stop will take reconfig lock which means raid5d can't call
md_update_sb() in md_check_recovery(), the badblock will always
be unack, so raid5d thread enters an infinite loop and md_stop_write()
can never stop sync_thread. This causes deadlock.

To solve this, when STOP_ARRAY ioctl is issued and sync_thread is
running, we need set md->recovery FROZEN and INTR flags and wait for
sync_thread to stop before we (re)take reconfig lock.

This requires that raid5 reshape_request notices MD_RECOVERY_INTR
(which it probably should have noticed anyway) and stops waiting for a
metadata update in that case.

Reported-by: Jianpeng Ma <majianpeng@gmail.com>
Reported-by: Bian Yu <bianyu@kedacom.com>
Signed-off-by: NeilBrown <neilb@suse.de>

diff --git a/drivers/md/md.c b/drivers/md/md.c
index a74045d..47e7bc7 100644 (file)
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -5340,20 +5340,35 @@ EXPORT_SYMBOL_GPL(md_stop);
 static int md_set_readonly(struct mddev *mddev, struct block_device *bdev)
 {
        int err = 0;
+       int did_freeze = 0;
+
+       if (!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) {
+               did_freeze = 1;
+               set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+               md_wakeup_thread(mddev->thread);
+       }
+       if (mddev->sync_thread) {
+               set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+               /* Thread might be blocked waiting for metadata update
+                * which will now never happen */
+               wake_up_process(mddev->sync_thread->tsk);
+       }
+       mddev_unlock(mddev);
+       wait_event(resync_wait, mddev->sync_thread == NULL);
+       mddev_lock_nointr(mddev);
+
        mutex_lock(&mddev->open_mutex);
-       if (atomic_read(&mddev->openers) > !!bdev) {
+       if (atomic_read(&mddev->openers) > !!bdev ||
+           mddev->sync_thread ||
+           (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags))) {
                printk("md: %s still in use.\n",mdname(mddev));
+               if (did_freeze) {
+                       clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+                       md_wakeup_thread(mddev->thread);
+               }
                err = -EBUSY;
                goto out;
        }
-       if (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags)) {
-               /* Someone opened the device since we flushed it
-                * so page cache could be dirty and it is too late
-                * to flush.  So abort
-                */
-               mutex_unlock(&mddev->open_mutex);
-               return -EBUSY;
-       }
        if (mddev->pers) {
                __md_stop_writes(mddev);
 
@@ -5364,7 +5379,7 @@ static int md_set_readonly(struct mddev *mddev, struct block_device *bdev)
                set_disk_ro(mddev->gendisk, 1);
                clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
                sysfs_notify_dirent_safe(mddev->sysfs_state);
-               err = 0;        
+               err = 0;
        }
 out:
        mutex_unlock(&mddev->open_mutex);
@@ -5380,20 +5395,34 @@ static int do_md_stop(struct mddev * mddev, int mode,
 {
        struct gendisk *disk = mddev->gendisk;
        struct md_rdev *rdev;
+       int did_freeze = 0;
+
+       if (!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) {
+               did_freeze = 1;
+               set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+               md_wakeup_thread(mddev->thread);
+       }
+       if (mddev->sync_thread) {
+               set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+               /* Thread might be blocked waiting for metadata update
+                * which will now never happen */
+               wake_up_process(mddev->sync_thread->tsk);
+       }
+       mddev_unlock(mddev);
+       wait_event(resync_wait, mddev->sync_thread == NULL);
+       mddev_lock_nointr(mddev);
 
        mutex_lock(&mddev->open_mutex);
        if (atomic_read(&mddev->openers) > !!bdev ||
-           mddev->sysfs_active) {
+           mddev->sysfs_active ||
+           mddev->sync_thread ||
+           (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags))) {
                printk("md: %s still in use.\n",mdname(mddev));
                mutex_unlock(&mddev->open_mutex);
-               return -EBUSY;
-       }
-       if (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags)) {
-               /* Someone opened the device since we flushed it
-                * so page cache could be dirty and it is too late
-                * to flush.  So abort
-                */
-               mutex_unlock(&mddev->open_mutex);
+               if (did_freeze) {
+                       clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+                       md_wakeup_thread(mddev->thread);
+               }
                return -EBUSY;
        }
        if (mddev->pers) {
@@ -7931,6 +7960,7 @@ void md_reap_sync_thread(struct mddev *mddev)
 
        /* resync has finished, collect result */
        md_unregister_thread(&mddev->sync_thread);
+       wake_up(&resync_wait);
        if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
            !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
                /* success...*/