From dcf6e2e38a1c7ccbc535de5e1d9b14998847499d Mon Sep 17 00:00:00 2001 From: Zachary Hays Date: Thu, 7 Feb 2019 10:03:08 -0500 Subject: mmc: block: handle complete_work on separate workqueue The kblockd workqueue is created with the WQ_MEM_RECLAIM flag set. This generates a rescuer thread for that queue that will trigger when the CPU is under heavy load and collect the uncompleted work. In the case of mmc, this creates the possibility of a deadlock when there are multiple partitions on the device as other blk-mq work is also run on the same queue. For example: - worker 0 claims the mmc host to work on partition 1 - worker 1 attempts to claim the host for partition 2 but has to wait for worker 0 to finish - worker 0 schedules complete_work to release the host - rescuer thread is triggered after time-out and collects the dangling work - rescuer thread attempts to complete the work in order starting with claim host - the task to release host is now blocked by a task to claim it and will never be called The above results in multiple hung tasks that lead to failures to mount partitions. Handling complete_work on a separate workqueue avoids this by keeping the work completion tasks separate from the other blk-mq work. This allows the host to be released without getting blocked by other tasks attempting to claim the host. Signed-off-by: Zachary Hays Fixes: 81196976ed94 ("mmc: block: Add blk-mq support") Cc: Signed-off-by: Ulf Hansson --- include/linux/mmc/card.h | 1 + 1 file changed, 1 insertion(+) (limited to 'include/linux/mmc') diff --git a/include/linux/mmc/card.h b/include/linux/mmc/card.h index de7377815b6b..8ef330027b13 100644 --- a/include/linux/mmc/card.h +++ b/include/linux/mmc/card.h @@ -308,6 +308,7 @@ struct mmc_card { unsigned int nr_parts; unsigned int bouncesz; /* Bounce buffer size */ + struct workqueue_struct *complete_wq; /* Private workqueue */ }; static inline bool mmc_large_sector(struct mmc_card *card) -- cgit v1.2.3-55-g7522