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>
diff --git a/drivers/md/linear.c b/drivers/md/linear.c
index df6f2c9..5f1eb91 100644
--- a/drivers/md/linear.c
+++ b/drivers/md/linear.c
@@ -245,7 +245,7 @@ static void linear_free(struct mddev *mddev, void *priv)
kfree(conf);
}
-static void linear_make_request(struct mddev *mddev, struct bio *bio)
+static bool linear_make_request(struct mddev *mddev, struct bio *bio)
{
char b[BDEVNAME_SIZE];
struct dev_info *tmp_dev;
@@ -254,7 +254,7 @@ static void linear_make_request(struct mddev *mddev, struct bio *bio)
if (unlikely(bio->bi_opf & REQ_PREFLUSH)) {
md_flush_request(mddev, bio);
- return;
+ return true;
}
tmp_dev = which_dev(mddev, bio_sector);
@@ -292,7 +292,7 @@ static void linear_make_request(struct mddev *mddev, struct bio *bio)
mddev_check_write_zeroes(mddev, bio);
generic_make_request(bio);
}
- return;
+ return true;
out_of_bounds:
pr_err("md/linear:%s: make_request: Sector %llu out of bounds on dev %s: %llu sectors, offset %llu\n",
@@ -302,6 +302,7 @@ static void linear_make_request(struct mddev *mddev, struct bio *bio)
(unsigned long long)tmp_dev->rdev->sectors,
(unsigned long long)start_sector);
bio_io_error(bio);
+ return true;
}
static void linear_status (struct seq_file *seq, struct mddev *mddev)