1. Patchew
  2. linux
  3. [PATCH 0/2] md: flush deadlock bugfix
  4. View patch

[PATCH 1/2] md: change the return value type of md_write_start to void

linan666@huaweicloud.com posted 2 patches 1 year, 8 months ago
[PATCH 1/2] md: change the return value type of md_write_start to void
Posted by linan666@huaweicloud.com 1 year, 8 months ago 
From: Li Nan <linan122@huawei.com>

Commit cc27b0c78c79 ("md: fix deadlock between mddev_suspend() and
md_write_start()") aborted md_write_start() with false when mddev is
suspended, which fixed a deadlock if calling mddev_suspend() with
holding reconfig_mutex(). Since mddev_suspend() now includes
lockdep_assert_not_held(), it no longer holds the reconfig_mutex. This
makes previous abort unnecessary. Now, remove unnecessary abort and
change function return value to void.

Signed-off-by: Li Nan <linan122@huawei.com>
---
 drivers/md/md.h     |  2 +-
 drivers/md/md.c     | 14 ++++----------
 drivers/md/raid1.c  |  3 +--
 drivers/md/raid10.c |  3 +--
 drivers/md/raid5.c  |  3 +--
 5 files changed, 8 insertions(+), 17 deletions(-)

diff --git a/drivers/md/md.h b/drivers/md/md.h
index ca085ecad504..487582058f74 100644
--- a/drivers/md/md.h
+++ b/drivers/md/md.h
@@ -785,7 +785,7 @@ extern void md_unregister_thread(struct mddev *mddev, struct md_thread __rcu **t
 extern void md_wakeup_thread(struct md_thread __rcu *thread);
 extern void md_check_recovery(struct mddev *mddev);
 extern void md_reap_sync_thread(struct mddev *mddev);
-extern bool md_write_start(struct mddev *mddev, struct bio *bi);
+extern void md_write_start(struct mddev *mddev, struct bio *bi);
 extern void md_write_inc(struct mddev *mddev, struct bio *bi);
 extern void md_write_end(struct mddev *mddev);
 extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 509e5638cea1..14d6e615bcbb 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -8638,12 +8638,12 @@ EXPORT_SYMBOL(md_done_sync);
  * A return value of 'false' means that the write wasn't recorded
  * and cannot proceed as the array is being suspend.
  */
-bool md_write_start(struct mddev *mddev, struct bio *bi)
+void md_write_start(struct mddev *mddev, struct bio *bi)
 {
 	int did_change = 0;
 
 	if (bio_data_dir(bi) != WRITE)
-		return true;
+		return;
 
 	BUG_ON(mddev->ro == MD_RDONLY);
 	if (mddev->ro == MD_AUTO_READ) {
@@ -8676,15 +8676,9 @@ bool md_write_start(struct mddev *mddev, struct bio *bi)
 	if (did_change)
 		sysfs_notify_dirent_safe(mddev->sysfs_state);
 	if (!mddev->has_superblocks)
-		return true;
+		return;
 	wait_event(mddev->sb_wait,
-		   !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags) ||
-		   is_md_suspended(mddev));
-	if (test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) {
-		percpu_ref_put(&mddev->writes_pending);
-		return false;
-	}
-	return true;
+		   !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
 }
 EXPORT_SYMBOL(md_write_start);
 
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index 7b8a71ca66dd..0d80ff471c73 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -1687,8 +1687,7 @@ static bool raid1_make_request(struct mddev *mddev, struct bio *bio)
 	if (bio_data_dir(bio) == READ)
 		raid1_read_request(mddev, bio, sectors, NULL);
 	else {
-		if (!md_write_start(mddev,bio))
-			return false;
+		md_write_start(mddev,bio);
 		raid1_write_request(mddev, bio, sectors);
 	}
 	return true;
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index a4556d2e46bf..f8d7c02c6ed5 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -1836,8 +1836,7 @@ static bool raid10_make_request(struct mddev *mddev, struct bio *bio)
 	    && md_flush_request(mddev, bio))
 		return true;
 
-	if (!md_write_start(mddev, bio))
-		return false;
+	md_write_start(mddev, bio);
 
 	if (unlikely(bio_op(bio) == REQ_OP_DISCARD))
 		if (!raid10_handle_discard(mddev, bio))
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 2bd1ce9b3922..a84389311dd1 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -6078,8 +6078,7 @@ static bool raid5_make_request(struct mddev *mddev, struct bio * bi)
 		ctx.do_flush = bi->bi_opf & REQ_PREFLUSH;
 	}
 
-	if (!md_write_start(mddev, bi))
-		return false;
+	md_write_start(mddev, bi);
 	/*
 	 * If array is degraded, better not do chunk aligned read because
 	 * later we might have to read it again in order to reconstruct
-- 
2.39.2
Re: [PATCH 1/2] md: change the return value type of md_write_start to void
Posted by Yu Kuai 1 year, 8 months ago 
在 2024/05/26 2:52, linan666@huaweicloud.com 写道:
> From: Li Nan <linan122@huawei.com>
> 
> Commit cc27b0c78c79 ("md: fix deadlock between mddev_suspend() and
> md_write_start()") aborted md_write_start() with false when mddev is
> suspended, which fixed a deadlock if calling mddev_suspend() with
> holding reconfig_mutex(). Since mddev_suspend() now includes
> lockdep_assert_not_held(), it no longer holds the reconfig_mutex. This
> makes previous abort unnecessary. Now, remove unnecessary abort and
> change function return value to void.

Nice cleanup, feel free to add:

Reviewed-by: Yu Kuai <yukuai3@huawei.com>
> 
> Signed-off-by: Li Nan <linan122@huawei.com>
> ---
>   drivers/md/md.h     |  2 +-
>   drivers/md/md.c     | 14 ++++----------
>   drivers/md/raid1.c  |  3 +--
>   drivers/md/raid10.c |  3 +--
>   drivers/md/raid5.c  |  3 +--
>   5 files changed, 8 insertions(+), 17 deletions(-)
> 
> diff --git a/drivers/md/md.h b/drivers/md/md.h
> index ca085ecad504..487582058f74 100644
> --- a/drivers/md/md.h
> +++ b/drivers/md/md.h
> @@ -785,7 +785,7 @@ extern void md_unregister_thread(struct mddev *mddev, struct md_thread __rcu **t
>   extern void md_wakeup_thread(struct md_thread __rcu *thread);
>   extern void md_check_recovery(struct mddev *mddev);
>   extern void md_reap_sync_thread(struct mddev *mddev);
> -extern bool md_write_start(struct mddev *mddev, struct bio *bi);
> +extern void md_write_start(struct mddev *mddev, struct bio *bi);
>   extern void md_write_inc(struct mddev *mddev, struct bio *bi);
>   extern void md_write_end(struct mddev *mddev);
>   extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
> diff --git a/drivers/md/md.c b/drivers/md/md.c
> index 509e5638cea1..14d6e615bcbb 100644
> --- a/drivers/md/md.c
> +++ b/drivers/md/md.c
> @@ -8638,12 +8638,12 @@ EXPORT_SYMBOL(md_done_sync);
>    * A return value of 'false' means that the write wasn't recorded
>    * and cannot proceed as the array is being suspend.
>    */
> -bool md_write_start(struct mddev *mddev, struct bio *bi)
> +void md_write_start(struct mddev *mddev, struct bio *bi)
>   {
>   	int did_change = 0;
>   
>   	if (bio_data_dir(bi) != WRITE)
> -		return true;
> +		return;
>   
>   	BUG_ON(mddev->ro == MD_RDONLY);
>   	if (mddev->ro == MD_AUTO_READ) {
> @@ -8676,15 +8676,9 @@ bool md_write_start(struct mddev *mddev, struct bio *bi)
>   	if (did_change)
>   		sysfs_notify_dirent_safe(mddev->sysfs_state);
>   	if (!mddev->has_superblocks)
> -		return true;
> +		return;
>   	wait_event(mddev->sb_wait,
> -		   !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags) ||
> -		   is_md_suspended(mddev));
> -	if (test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) {
> -		percpu_ref_put(&mddev->writes_pending);
> -		return false;
> -	}
> -	return true;
> +		   !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags));
>   }
>   EXPORT_SYMBOL(md_write_start);
>   
> diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
> index 7b8a71ca66dd..0d80ff471c73 100644
> --- a/drivers/md/raid1.c
> +++ b/drivers/md/raid1.c
> @@ -1687,8 +1687,7 @@ static bool raid1_make_request(struct mddev *mddev, struct bio *bio)
>   	if (bio_data_dir(bio) == READ)
>   		raid1_read_request(mddev, bio, sectors, NULL);
>   	else {
> -		if (!md_write_start(mddev,bio))
> -			return false;
> +		md_write_start(mddev,bio);
>   		raid1_write_request(mddev, bio, sectors);
>   	}
>   	return true;
> diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
> index a4556d2e46bf..f8d7c02c6ed5 100644
> --- a/drivers/md/raid10.c
> +++ b/drivers/md/raid10.c
> @@ -1836,8 +1836,7 @@ static bool raid10_make_request(struct mddev *mddev, struct bio *bio)
>   	    && md_flush_request(mddev, bio))
>   		return true;
>   
> -	if (!md_write_start(mddev, bio))
> -		return false;
> +	md_write_start(mddev, bio);
>   
>   	if (unlikely(bio_op(bio) == REQ_OP_DISCARD))
>   		if (!raid10_handle_discard(mddev, bio))
> diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
> index 2bd1ce9b3922..a84389311dd1 100644
> --- a/drivers/md/raid5.c
> +++ b/drivers/md/raid5.c
> @@ -6078,8 +6078,7 @@ static bool raid5_make_request(struct mddev *mddev, struct bio * bi)
>   		ctx.do_flush = bi->bi_opf & REQ_PREFLUSH;
>   	}
>   
> -	if (!md_write_start(mddev, bi))
> -		return false;
> +	md_write_start(mddev, bi);
>   	/*
>   	 * If array is degraded, better not do chunk aligned read because
>   	 * later we might have to read it again in order to reconstruct
>

Patchew 2.1-devel-b67e4c2

© 2016 - 2026 Red Hat, Inc.