There's an I/O error on fsync() in a detached loop device if it has been previously attached. The issue is write cache is enabled in the attach path in loop_configure() but it isn't disabled in the detach path; thus it remains enabled in the block device regardless of whether it is attached or not. Now fsync() can get an I/O request that will just be failed later in loop_queue_rq() as device's state is not 'Lo_bound'. So, disable write cache in the detach path. Do so based on the queue flag, not the loop device flag for read-only (used to enable) as the queue flag can be changed via sysfs even on read-only loop devices (e.g., losetup -r.) Test-case: # DEV=/dev/loop7 # IMG=/tmp/image # truncate --size 1M $IMG # losetup $DEV $IMG # losetup -d $DEV Before: # strace -e fsync parted -s $DEV print 2>&1 | grep fsync fsync(3) = -1 EIO (Input/output error) Warning: Error fsyncing/closing /dev/loop7: Input/output error [ 982.529929] blk_update_request: I/O error, dev loop7, sector 0 op 0x1:(WRITE) flags 0x800 phys_seg 0 prio class 0 After: # strace -e fsync parted -s $DEV print 2>&1 | grep fsync fsync(3) = 0 Co-developed-by: Eric Desrochers <eric.desrochers@xxxxxxxxxxxxx> Signed-off-by: Eric Desrochers <eric.desrochers@xxxxxxxxxxxxx> Signed-off-by: Mauricio Faria de Oliveira <mfo@xxxxxxxxxxxxx> Tested-by: Gabriel Krisman Bertazi <krisman@xxxxxxxxxxxxx> Reviewed-by: Ming Lei <ming.lei@xxxxxxxxxx> --- v4: - Add Reviewed-By: tag (thanks, Ming.) v3: - Disable write cache based on QUEUE_FLAG_WC, not just !LO_FLAGS_READ_ONLY, as the former can be enabled in sysfs even for LO_FLAGS_READ_ONLY, thus not disabled later. Mention that in commit message. (thanks, Ming) v2: - Fix ordering of Co-developed-by:/Signed-off-by: tags. (thanks, Krisman) - Add Tested-by: tag. (likewise.) drivers/block/loop.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/block/loop.c b/drivers/block/loop.c index e5ff328f0917..e94a11dbb5bd 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -1212,6 +1212,9 @@ static int __loop_clr_fd(struct loop_device *lo, bool release) goto out_unlock; } + if (test_bit(QUEUE_FLAG_WC, &lo->lo_queue->queue_flags)) + blk_queue_write_cache(lo->lo_queue, false, false); + /* freeze request queue during the transition */ blk_mq_freeze_queue(lo->lo_queue); -- 2.27.0