| Message ID | 20240120103734.4155446-6-yukuai1@huaweicloud.com (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Delegated to: | Mike Snitzer |
| Headers | show |
| Series | md: fix/prevent dm-raid regressions | expand |
Hi, 在 2024/01/20 18:37, Yu Kuai 写道: > The root cause is still not clear yet, however, let's convert dm-raid > back to use md_reap_sync_thread() directly. This is not safe but at > least there won't be new regressions. We can decide what to do after > figuring out the root cause. I think I finally figure out the root cause here. This patch is no longer needed after following patch. I already verified in my VM for 3 times that lvconvert-raid-reshape.sh won't fail(with raid6 patch 2c265ac5ffde reverted). I'll run more tests in case there are new regression. Meanwhile I'll try to locate root cause of the problem decribed in patch 4. Thanks, Kuai diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c index eb009d6bb03a..108e7e313631 100644 --- a/drivers/md/dm-raid.c +++ b/drivers/md/dm-raid.c @@ -3241,7 +3241,7 @@ static int raid_ctr(struct dm_target *ti, unsigned int argc, char **argv) rs->md.in_sync = 1; /* Keep array frozen until resume. */ - set_bit(MD_RECOVERY_FROZEN, &rs->md.recovery); + md_frozen_sync_thread(&rs->md); /* Has to be held on running the array */ mddev_suspend_and_lock_nointr(&rs->md); @@ -3722,6 +3722,9 @@ static int raid_message(struct dm_target *ti, unsigned int argc, char **argv, if (!mddev->pers || !mddev->pers->sync_request) return -EINVAL; + if (test_bit(RT_FLAG_RS_SUSPENDED, &rs->runtime_flags)) + return -EBUSY; + if (!strcasecmp(argv[0], "frozen")) set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); else @@ -3796,10 +3799,8 @@ 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)) { - /* Writes have to be stopped before suspending to avoid deadlocks. */ - if (!test_bit(MD_RECOVERY_FROZEN, &rs->md.recovery)) - md_stop_writes(&rs->md); - + md_frozen_sync_thread(&rs->md); + md_stop_writes(&rs->md); mddev_suspend(&rs->md, false); } } @@ -4011,9 +4012,6 @@ static int raid_preresume(struct dm_target *ti) DMERR("Failed to resize bitmap"); } - /* Check for any resize/reshape on @rs and adjust/initiate */ - /* Be prepared for mddev_resume() in raid_resume() */ - set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); if (mddev->recovery_cp && mddev->recovery_cp < MaxSector) { set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); mddev->resync_min = mddev->recovery_cp; @@ -4056,10 +4054,11 @@ static void raid_resume(struct dm_target *ti) rs_set_capacity(rs); mddev_lock_nointr(mddev); - clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); mddev->ro = 0; mddev->in_sync = 0; mddev_unlock_and_resume(mddev); + + md_unfrozen_sync_thread(mddev); } } diff --git a/drivers/md/md.c b/drivers/md/md.c index 9ef17a769cc2..0638d104fe26 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -4939,7 +4939,7 @@ static void idle_sync_thread(struct mddev *mddev) mutex_unlock(&mddev->sync_mutex); } -static void frozen_sync_thread(struct mddev *mddev) +void md_frozen_sync_thread(struct mddev *mddev) { mutex_lock(&mddev->sync_mutex); set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); @@ -4952,6 +4952,18 @@ static void frozen_sync_thread(struct mddev *mddev) stop_sync_thread(mddev, false, false); mutex_unlock(&mddev->sync_mutex); } +EXPORT_SYMBOL_GPL(md_frozen_sync_thread); + +void md_unfrozen_sync_thread(struct mddev *mddev) +{ + mutex_lock(&mddev->sync_mutex); + clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); + set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); + md_wakeup_thread(mddev->thread); + sysfs_notify_dirent_safe(mddev->sysfs_action); + mutex_unlock(&mddev->sync_mutex); +} +EXPORT_SYMBOL_GPL(md_unfrozen_sync_thread); static ssize_t action_store(struct mddev *mddev, const char *page, size_t len) @@ -4963,7 +4975,7 @@ action_store(struct mddev *mddev, const char *page, size_t len) if (cmd_match(page, "idle")) idle_sync_thread(mddev); else if (cmd_match(page, "frozen")) - frozen_sync_thread(mddev); + md_frozen_sync_thread(mddev); else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) return -EBUSY; else if (cmd_match(page, "resync")) diff --git a/drivers/md/md.h b/drivers/md/md.h index 8d881cc59799..332520595ed8 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -781,6 +781,8 @@ extern void md_rdev_clear(struct md_rdev *rdev); extern void md_handle_request(struct mddev *mddev, struct bio *bio); extern int mddev_suspend(struct mddev *mddev, bool interruptible); extern void mddev_resume(struct mddev *mddev); +extern void md_frozen_sync_thread(struct mddev *mddev); +extern void md_unfrozen_sync_thread(struct mddev *mddev); extern void md_reload_sb(struct mddev *mddev, int raid_disk); extern void md_update_sb(struct mddev *mddev, int force);
diff --git a/drivers/md/md.c b/drivers/md/md.c index 7db749ba7e60..3e8dd020bf9f 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -4909,6 +4909,14 @@ static void stop_sync_thread(struct mddev *mddev, bool locked, bool check_seq) if (work_pending(&mddev->sync_work)) flush_work(&mddev->sync_work); + if (!mddev->gendisk) { + mddev_lock_nointr(mddev); + md_reap_sync_thread(mddev); + if (!locked) + mddev_unlock(mddev); + return; + } + wait_event(resync_wait, !test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) || (check_seq && sync_seq != atomic_read(&mddev->sync_seq)));