dm: dec_pending() needs locking to save error value
PATCH 1/1 dm: dec_pending() needs locking to save error value
Hi Alasdair,
While implementing the barrier support patch for request-based dm, I found a possible bug in bio-based dm.
In the patch posted here:
http://patchwork.kernel.org/patch/48189/ I added 'barrier_error_lock' to struct mapped_device to save error values from multiple barrier I/Os safely like below: if (error) { spin_lock_irqsave(&md->barrier_error_lock, flags); if (!md->barrier_error || error == -EOPNOTSUPP || (md->barrier_error != -EOPNOTSUPP && error == DM_ENDIO_REQUEUE)) md->barrier_error = error; spin_unlock_irqrestore(&md->barrier_error_lock, flags); }
On the other hand, bio-based dm has been doing the same thing in dec_pending() without any lock:
if (error && !(io->error > 0 && __noflush_suspending(md)))
io->error = error;Though I have never experienced actual problem without locking and just found this during code inspection, I believe it needs locking here, too.
This patch adds the locking.
I've done compile, boot and basic I/O testings.
Signed-off-by Kiyoshi Ueda <[email protected]>
Signed-off-by Jun'ichi Nomura <[email protected]>
Cc Alasdair G Kergon <[email protected]>
--- drivers/md/dm.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)
drivers/md/dm.c +8 -2
@@ -47,6 +47,7 @@ struct dm_io {atomic_t io_count;struct bio *bio;unsigned long start_time; +spinlock_t endio_lock; };/*
@@ -590,8 +591,12 @@ static void dec_pending(struct dm_io *iostruct mapped_device *md = io->md;/* Push-back supersedes any I/O errors */ -if (error && !(io->error >0 &&__noflush_suspending (md))) - io->error = error; +if (unlikely (error)) { +spin_lock_irqsave (&io->endio_lock, flags); +if (!(io->error >0 &&__noflush_suspending (md))) + io->error = error; +spin_unlock_irqrestore (&io->endio_lock, flags); + }if (atomic_dec_and_test (&io->io_count)) {if (io->error ==DM_ENDIO_REQUEUE ) {
@@ -1300,6 +1305,7 @@ static void __split_and_process_bio(struatomic_set (&ci.io->io_count,1 ); ci.io->bio = bio; ci.io->md = md; +spin_lock_init (&ci.io->endio_lock); ci.sector = bio->bi_sector; ci.sector_count =bio_sectors (bio);if (unlikely (bio_empty_barrier (bio)))