md: fix deadlock between mddev_suspend() and md_write_start()

If mddev_suspend() races with md_write_start() we can deadlock
with mddev_suspend() waiting for the request that is currently
in md_write_start() to complete the ->make_request() call,
and md_write_start() waiting for the metadata to be updated
to mark the array as 'dirty'.
As metadata updates done by md_check_recovery() only happen then
the mddev_lock() can be claimed, and as mddev_suspend() is often
called with the lock held, these threads wait indefinitely for each
other.

We fix this by having md_write_start() abort if mddev_suspend()
is happening, and ->make_request() aborts if md_write_start()
aborted.
md_make_request() can detect this abort, decrease the ->active_io
count, and wait for mddev_suspend().

Reported-by: Nix <nix@esperi.org.uk>
Fix: 68866e425be2(MD: no sync IO while suspended)
Cc: stable@vger.kernel.org
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Shaohua Li <shli@fb.com>

1 files changed, 9 insertions, 8 deletions

diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index ec0f951ae19f..b218a42fd702 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -5479,7 +5479,6 @@ static void make_discard_request(struct mddev *mddev, struct bio *bi)
 	last_sector = bi->bi_iter.bi_sector + (bi->bi_iter.bi_size>>9);
 
 	bi->bi_next = NULL;
-	md_write_start(mddev, bi);
 
 	stripe_sectors = conf->chunk_sectors *
 		(conf->raid_disks - conf->max_degraded);
@@ -5549,11 +5548,10 @@ static void make_discard_request(struct mddev *mddev, struct bio *bi)
 		release_stripe_plug(mddev, sh);
 	}
 
-	md_write_end(mddev);
 	bio_endio(bi);
 }
 
-static void raid5_make_request(struct mddev *mddev, struct bio * bi)
+static bool raid5_make_request(struct mddev *mddev, struct bio * bi)
 {
 	struct r5conf *conf = mddev->private;
 	int dd_idx;
@@ -5569,10 +5567,10 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
 		int ret = r5l_handle_flush_request(conf->log, bi);
 
 		if (ret == 0)
-			return;
+			return true;
 		if (ret == -ENODEV) {
 			md_flush_request(mddev, bi);
-			return;
+			return true;
 		}
 		/* ret == -EAGAIN, fallback */
 		/*
@@ -5582,6 +5580,8 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
 		do_flush = bi->bi_opf & REQ_PREFLUSH;
 	}
 
+	if (!md_write_start(mddev, bi))
+		return false;
 	/*
 	 * If array is degraded, better not do chunk aligned read because
 	 * later we might have to read it again in order to reconstruct
@@ -5591,18 +5591,18 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
 	    mddev->reshape_position == MaxSector) {
 		bi = chunk_aligned_read(mddev, bi);
 		if (!bi)
-			return;
+			return true;
 	}
 
 	if (unlikely(bio_op(bi) == REQ_OP_DISCARD)) {
 		make_discard_request(mddev, bi);
-		return;
+		md_write_end(mddev);
+		return true;
 	}
 
 	logical_sector = bi->bi_iter.bi_sector & ~((sector_t)STRIPE_SECTORS-1);
 	last_sector = bio_end_sector(bi);
 	bi->bi_next = NULL;
-	md_write_start(mddev, bi);
 
 	prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
 	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
@@ -5740,6 +5740,7 @@ static void raid5_make_request(struct mddev *mddev, struct bio * bi)
 	if (rw == WRITE)
 		md_write_end(mddev);
 	bio_endio(bi);
+	return true;
 }
 
 static sector_t raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks);