Commit 7fc4da6a authored by Ming Lei's avatar Ming Lei Committed by Jens Axboe
Browse files

ublk: scan partition in async way 



Implement async partition scan to avoid IO hang when reading partition
tables. Similar to nvme_partition_scan_work(), partition scanning is
deferred to a work queue to prevent deadlocks.

When partition scan happens synchronously during add_disk(), IO errors
can cause the partition scan to wait while holding ub->mutex, which
can deadlock with other operations that need the mutex.

Changes:
- Add partition_scan_work to ublk_device structure
- Implement ublk_partition_scan_work() to perform async scan
- Always suppress sync partition scan during add_disk()
- Schedule async work after add_disk() for trusted daemons
- Add flush_work() in ublk_stop_dev() before grabbing ub->mutex

Reviewed-by: default avatarCaleb Sander Mateos <csander@purestorage.com>
Reported-by: default avatarYoav Cohen <yoav@nvidia.com>
Closes: https://lore.kernel.org/linux-block/DM4PR12MB63280C5637917C071C2F0D65A9A8A@DM4PR12MB6328.namprd12.prod.outlook.com/


Fixes: 71f28f31 ("ublk_drv: add io_uring based userspace block driver")
Signed-off-by: default avatarMing Lei <ming.lei@redhat.com>
Signed-off-by: default avatarJens Axboe <axboe@kernel.dk>
parent 04bdb1a0

drivers/block/ublk_drv.c

+32 −3
Original line number Diff line number Diff line
@@ -237,6 +237,7 @@ struct ublk_device { 
	bool canceling;

	pid_t 	ublksrv_tgid;

	struct delayed_work	exit_work;

	struct work_struct	partition_scan_work;



	struct ublk_queue       *queues[];

};
@@ -254,6 +255,20 @@ static inline struct request *__ublk_check_and_get_req(struct ublk_device *ub, 
		u16 q_id, u16 tag, struct ublk_io *io, size_t offset);

static inline unsigned int ublk_req_build_flags(struct request *req);



static void ublk_partition_scan_work(struct work_struct *work)

{

	struct ublk_device *ub =

		container_of(work, struct ublk_device, partition_scan_work);



	if (WARN_ON_ONCE(!test_and_clear_bit(GD_SUPPRESS_PART_SCAN,

					     &ub->ub_disk->state)))

		return;



	mutex_lock(&ub->ub_disk->open_mutex);

	bdev_disk_changed(ub->ub_disk, false);

	mutex_unlock(&ub->ub_disk->open_mutex);

}



static inline struct ublksrv_io_desc *

ublk_get_iod(const struct ublk_queue *ubq, unsigned tag)

{
@@ -2026,6 +2041,7 @@ static void ublk_stop_dev(struct ublk_device *ub) 
	mutex_lock(&ub->mutex);

	ublk_stop_dev_unlocked(ub);

	mutex_unlock(&ub->mutex);

	flush_work(&ub->partition_scan_work);

	ublk_cancel_dev(ub);

}
@@ -2954,8 +2970,16 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub, 


	ublk_apply_params(ub);



	/* don't probe partitions if any daemon task is un-trusted */

	if (ub->unprivileged_daemons)

	/*

	 * Suppress partition scan to avoid potential IO hang.

	 *

	 * If ublk server error occurs during partition scan, the IO may

	 * wait while holding ub->mutex, which can deadlock with other

	 * operations that need the mutex. Defer partition scan to async

	 * work.

	 * For unprivileged daemons, keep GD_SUPPRESS_PART_SCAN set

	 * permanently.

	 */

	set_bit(GD_SUPPRESS_PART_SCAN, &disk->state);



	ublk_get_device(ub);
@@ -2973,6 +2997,10 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub, 


	set_bit(UB_STATE_USED, &ub->state);



	/* Schedule async partition scan for trusted daemons */

	if (!ub->unprivileged_daemons)

		schedule_work(&ub->partition_scan_work);



out_put_cdev:

	if (ret) {

		ublk_detach_disk(ub);
@@ -3138,6 +3166,7 @@ static int ublk_ctrl_add_dev(const struct ublksrv_ctrl_cmd *header) 
	mutex_init(&ub->mutex);

	spin_lock_init(&ub->lock);

	mutex_init(&ub->cancel_mutex);

	INIT_WORK(&ub->partition_scan_work, ublk_partition_scan_work);



	ret = ublk_alloc_dev_number(ub, header->dev_id);

	if (ret < 0)