[PATCH] dm/dax: Fix table reference counts

From: Dan Williams
Date: Fri Sep 18 2020 - 16:09:46 EST

A recent fix to the dm_dax_supported() flow uncovered a latent bug. When
dm_get_live_table() fails it is still required to drop the
srcu_read_lock(). Without this change the lvm2 test-suite triggers this

# lvm2-testsuite --only pvmove-abort-all.sh

WARNING: lock held when returning to user space!
5.9.0-rc5+ #251 Tainted: G OE
lvm/1318 is leaving the kernel with locks still held!
1 lock held by lvm/1318:
#0: ffff9372abb5a340 (&md->io_barrier){....}-{0:0}, at: dm_get_live_table+0x5/0xb0 [dm_mod]

...and later on this hang signature:

INFO: task lvm:1344 blocked for more than 122 seconds.
Tainted: G OE 5.9.0-rc5+ #251
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
task:lvm state:D stack: 0 pid: 1344 ppid: 1 flags:0x00004000
Call Trace:
? finish_task_switch+0x249/0x2c0
? wait_for_completion+0x86/0x110
? __schedule+0x467/0xa80
? wait_for_completion+0x86/0x110
? __bpf_trace_rcu_utilization+0x10/0x10
__dm_suspend+0x6d/0x210 [dm_mod]
dm_suspend+0xf6/0x140 [dm_mod]

Fixes: 7bf7eac8d648 ("dax: Arrange for dax_supported check to span multiple devices")
Cc: <stable@xxxxxxxxxxxxxxx>
Cc: Jan Kara <jack@xxxxxxx>
Cc: Alasdair Kergon <agk@xxxxxxxxxx>
Cc: Mike Snitzer <snitzer@xxxxxxxxxx>
Reported-by: Adrian Huang <ahuang12@xxxxxxxxxx>
Signed-off-by: Dan Williams <dan.j.williams@xxxxxxxxx>
drivers/md/dm.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index fb0255d25e4b..4a40df8af7d3 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -1136,15 +1136,16 @@ static bool dm_dax_supported(struct dax_device *dax_dev, struct block_device *bd
struct mapped_device *md = dax_get_private(dax_dev);
struct dm_table *map;
+ bool ret = false;
int srcu_idx;
- bool ret;

map = dm_get_live_table(md, &srcu_idx);
if (!map)
- return false;
+ goto out;

ret = dm_table_supports_dax(map, device_supports_dax, &blocksize);

dm_put_live_table(md, srcu_idx);

return ret;