block: simplify disk shutdown

Set the queue dying flag and call blk_mq_exit_queue from del_gendisk for
all disks that do not have separately allocated queues, and thus remove
the need to call blk_cleanup_queue for them.

Rename blk_cleanup_disk to blk_mq_destroy_queue to make it clear that
this function is intended only for separately allocated blk-mq queues.

This saves an extra queue freeze for devices without a separately
allocated queue.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20220619060552.1850436-6-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block/blk-core.c

@@ -284,43 +284,6 @@ void blk_queue_start_drain(struct request_queue *q)
 	wake_up_all(&q->mq_freeze_wq);
 }
 
-/**
- * blk_cleanup_queue - shutdown a request queue
- * @q: request queue to shutdown
- *
- * Mark @q DYING, drain all pending requests, mark @q DEAD, destroy and
- * put it.  All future requests will be failed immediately with -ENODEV.
- *
- * Context: can sleep
- */
-void blk_cleanup_queue(struct request_queue *q)
-{
-	/* cannot be called from atomic context */
-	might_sleep();
-
-	WARN_ON_ONCE(blk_queue_registered(q));
-
-	/* mark @q DYING, no new request or merges will be allowed afterwards */
-	blk_queue_flag_set(QUEUE_FLAG_DYING, q);
-	blk_queue_start_drain(q);
-
-	/*
-	 * Drain all requests queued before DYING marking. Set DEAD flag to
-	 * prevent that blk_mq_run_hw_queues() accesses the hardware queues
-	 * after draining finished.
-	 */
-	blk_freeze_queue(q);
-	blk_sync_queue(q);
-	if (queue_is_mq(q)) {
-		blk_mq_cancel_work_sync(q);
-		blk_mq_exit_queue(q);
-	}
-
-	/* @q is and will stay empty, shutdown and put */
-	blk_put_queue(q);
-}
-EXPORT_SYMBOL(blk_cleanup_queue);
-
 /**
  * blk_queue_enter() - try to increase q->q_usage_counter
  * @q: request queue pointer

block/blk-mq.c

@@ -3902,7 +3902,7 @@ static struct request_queue *blk_mq_init_queue_data(struct blk_mq_tag_set *set,
 	q->queuedata = queuedata;
 	ret = blk_mq_init_allocated_queue(set, q);
 	if (ret) {
-		blk_cleanup_queue(q);
+		blk_put_queue(q);
 		return ERR_PTR(ret);
 	}
 	return q;
@@ -3914,6 +3914,35 @@ struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *set)
 }
 EXPORT_SYMBOL(blk_mq_init_queue);
 
+/**
+ * blk_mq_destroy_queue - shutdown a request queue
+ * @q: request queue to shutdown
+ *
+ * This shuts down a request queue allocated by blk_mq_init_queue() and drops
+ * the initial reference.  All future requests will failed with -ENODEV.
+ *
+ * Context: can sleep
+ */
+void blk_mq_destroy_queue(struct request_queue *q)
+{
+	WARN_ON_ONCE(!queue_is_mq(q));
+	WARN_ON_ONCE(blk_queue_registered(q));
+
+	might_sleep();
+
+	blk_queue_flag_set(QUEUE_FLAG_DYING, q);
+	blk_queue_start_drain(q);
+	blk_freeze_queue(q);
+
+	blk_sync_queue(q);
+	blk_mq_cancel_work_sync(q);
+	blk_mq_exit_queue(q);
+
+	/* @q is and will stay empty, shutdown and put */
+	blk_put_queue(q);
+}
+EXPORT_SYMBOL(blk_mq_destroy_queue);
+
 struct gendisk *__blk_mq_alloc_disk(struct blk_mq_tag_set *set, void *queuedata,
 		struct lock_class_key *lkclass)
 {
@@ -3926,13 +3955,23 @@ struct gendisk *__blk_mq_alloc_disk(struct blk_mq_tag_set *set, void *queuedata,
 
 	disk = __alloc_disk_node(q, set->numa_node, lkclass);
 	if (!disk) {
-		blk_cleanup_queue(q);
+		blk_put_queue(q);
 		return ERR_PTR(-ENOMEM);
 	}
+	set_bit(GD_OWNS_QUEUE, &disk->state);
 	return disk;
 }
 EXPORT_SYMBOL(__blk_mq_alloc_disk);
 
+struct gendisk *blk_mq_alloc_disk_for_queue(struct request_queue *q,
+		struct lock_class_key *lkclass)
+{
+	if (!blk_get_queue(q))
+		return NULL;
+	return __alloc_disk_node(q, NUMA_NO_NODE, lkclass);
+}
+EXPORT_SYMBOL(blk_mq_alloc_disk_for_queue);
+
 static struct blk_mq_hw_ctx *blk_mq_alloc_and_init_hctx(
 		struct blk_mq_tag_set *set, struct request_queue *q,
 		int hctx_idx, int node)

block/blk-sysfs.c

@@ -755,11 +755,6 @@ static void blk_free_queue_rcu(struct rcu_head *rcu_head)
  * decremented with blk_put_queue(). Once the refcount reaches 0 this function
  * is called.
  *
- * For drivers that have a request_queue on a gendisk and added with
- * __device_add_disk() the refcount to request_queue will reach 0 with
- * the last put_disk() called by the driver. For drivers which don't use
- * __device_add_disk() this happens with blk_cleanup_queue().
- *
  * Drivers exist which depend on the release of the request_queue to be
  * synchronous, it should not be deferred.
  *

block/blk.h

@@ -424,6 +424,9 @@ int bdev_resize_partition(struct gendisk *disk, int partno, sector_t start,
 		sector_t length);
 void blk_drop_partitions(struct gendisk *disk);
 
+struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
+		struct lock_class_key *lkclass);
+
 int bio_add_hw_page(struct request_queue *q, struct bio *bio,
 		struct page *page, unsigned int len, unsigned int offset,
 		unsigned int max_sectors, bool *same_page);

block/bsg-lib.c

@@ -324,7 +324,7 @@ void bsg_remove_queue(struct request_queue *q)
 			container_of(q->tag_set, struct bsg_set, tag_set);
 
 		bsg_unregister_queue(bset->bd);
-		blk_cleanup_queue(q);
+		blk_mq_destroy_queue(q);
 		blk_mq_free_tag_set(&bset->tag_set);
 		kfree(bset);
 	}
@@ -399,7 +399,7 @@ struct request_queue *bsg_setup_queue(struct device *dev, const char *name,
 
 	return q;
 out_cleanup_queue:
-	blk_cleanup_queue(q);
+	blk_mq_destroy_queue(q);
 out_queue:
 	blk_mq_free_tag_set(set);
 out_tag_set:

block/genhd.c

@@ -617,6 +617,8 @@ void del_gendisk(struct gendisk *disk)
 	 * Fail any new I/O.
 	 */
 	set_bit(GD_DEAD, &disk->state);
+	if (test_bit(GD_OWNS_QUEUE, &disk->state))
+		blk_queue_flag_set(QUEUE_FLAG_DYING, q);
 	set_capacity(disk, 0);
 
 	/*
@@ -663,11 +665,16 @@ void del_gendisk(struct gendisk *disk)
 	blk_mq_unquiesce_queue(q);
 
 	/*
-	 * Allow using passthrough request again after the queue is torn down.
+	 * If the disk does not own the queue, allow using passthrough requests
+	 * again.  Else leave the queue frozen to fail all I/O.
 	 */
-	blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q);
-	__blk_mq_unfreeze_queue(q, true);
-
+	if (!test_bit(GD_OWNS_QUEUE, &disk->state)) {
+		blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q);
+		__blk_mq_unfreeze_queue(q, true);
+	} else {
+		if (queue_is_mq(q))
+			blk_mq_exit_queue(q);
+	}
 }
 EXPORT_SYMBOL(del_gendisk);
 
@@ -1338,9 +1345,6 @@ struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
 {
 	struct gendisk *disk;
 
-	if (!blk_get_queue(q))
-		return NULL;
-
 	disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
 	if (!disk)
 		goto out_put_queue;
@@ -1391,7 +1395,6 @@ out_put_queue:
 	blk_put_queue(q);
 	return NULL;
 }
-EXPORT_SYMBOL(__alloc_disk_node);
 
 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
 {
@@ -1404,9 +1407,10 @@ struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass)
 
 	disk = __alloc_disk_node(q, node, lkclass);
 	if (!disk) {
-		blk_cleanup_queue(q);
+		blk_put_queue(q);
 		return NULL;
 	}
+	set_bit(GD_OWNS_QUEUE, &disk->state);
 	return disk;
 }
 EXPORT_SYMBOL(__blk_alloc_disk);
@@ -1439,7 +1443,6 @@ EXPORT_SYMBOL(put_disk);
  */
 void blk_cleanup_disk(struct gendisk *disk)
 {
-	blk_cleanup_queue(disk->queue);
 	put_disk(disk);
 }
 EXPORT_SYMBOL(blk_cleanup_disk);

drivers/block/ataflop.c

@@ -2045,7 +2045,6 @@ static void atari_floppy_cleanup(void)
 			if (!unit[i].disk[type])
 				continue;
 			del_gendisk(unit[i].disk[type]);
-			blk_cleanup_queue(unit[i].disk[type]->queue);
 			put_disk(unit[i].disk[type]);
 		}
 		blk_mq_free_tag_set(&unit[i].tag_set);

drivers/block/loop.c

@@ -2057,7 +2057,6 @@ static void loop_remove(struct loop_device *lo)
 {
 	/* Make this loop device unreachable from pathname. */
 	del_gendisk(lo->lo_disk);
-	blk_cleanup_queue(lo->lo_disk->queue);
 	blk_mq_free_tag_set(&lo->tag_set);
 
 	mutex_lock(&loop_ctl_mutex);

drivers/block/mtip32xx/mtip32xx.c

@@ -3565,7 +3565,6 @@ static int mtip_block_shutdown(struct driver_data *dd)
 	if (test_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag))
 		del_gendisk(dd->disk);
 
-	blk_cleanup_queue(dd->queue);
 	blk_mq_free_tag_set(&dd->tags);
 	put_disk(dd->disk);
 	return 0;
@@ -3914,7 +3913,6 @@ static void mtip_pci_remove(struct pci_dev *pdev)
 		dev_info(&dd->pdev->dev, "device %s surprise removal\n",
 						dd->disk->disk_name);
 
-	blk_cleanup_queue(dd->queue);
 	blk_mq_free_tag_set(&dd->tags);
 
 	/* De-initialize the protocol layer. */

drivers/block/rnbd/rnbd-clt.c

@@ -1755,7 +1755,7 @@ static void rnbd_destroy_sessions(void)
 		list_for_each_entry_safe(dev, tn, &sess->devs_list, list) {
 			/*
 			 * Here unmap happens in parallel for only one reason:
-			 * blk_cleanup_queue() takes around half a second, so
+			 * del_gendisk() takes around half a second, so
 			 * on huge amount of devices the whole module unload
 			 * procedure takes minutes.
 			 */

drivers/block/sx8.c

@@ -1536,7 +1536,7 @@ err_out_free_majors:
 		clear_bit(0, &carm_major_alloc);
 	else if (host->major == 161)
 		clear_bit(1, &carm_major_alloc);
-	blk_cleanup_queue(host->oob_q);
+	blk_mq_destroy_queue(host->oob_q);
 	blk_mq_free_tag_set(&host->tag_set);
 err_out_dma_free:
 	dma_free_coherent(&pdev->dev, CARM_SHM_SIZE, host->shm, host->shm_dma);
@@ -1570,7 +1570,7 @@ static void carm_remove_one (struct pci_dev *pdev)
 		clear_bit(0, &carm_major_alloc);
 	else if (host->major == 161)
 		clear_bit(1, &carm_major_alloc);
-	blk_cleanup_queue(host->oob_q);
+	blk_mq_destroy_queue(host->oob_q);
 	blk_mq_free_tag_set(&host->tag_set);
 	dma_free_coherent(&pdev->dev, CARM_SHM_SIZE, host->shm, host->shm_dma);
 	iounmap(host->mmio);

drivers/block/virtio_blk.c

@@ -1111,7 +1111,6 @@ static void virtblk_remove(struct virtio_device *vdev)
 	flush_work(&vblk->config_work);
 
 	del_gendisk(vblk->disk);
-	blk_cleanup_queue(vblk->disk->queue);
 	blk_mq_free_tag_set(&vblk->tag_set);
 
 	mutex_lock(&vblk->vdev_mutex);

drivers/block/z2ram.c

@@ -384,7 +384,6 @@ static void __exit z2_exit(void)
 
 	for (i = 0; i < Z2MINOR_COUNT; i++) {
 		del_gendisk(z2ram_gendisk[i]);
-		blk_cleanup_queue(z2ram_gendisk[i]->queue);
 		put_disk(z2ram_gendisk[i]);
 	}
 	blk_mq_free_tag_set(&tag_set);

drivers/cdrom/gdrom.c

@@ -831,7 +831,6 @@ probe_fail_no_mem:
 
 static int remove_gdrom(struct platform_device *devptr)
 {
-	blk_cleanup_queue(gd.gdrom_rq);
 	blk_mq_free_tag_set(&gd.tag_set);
 	free_irq(HW_EVENT_GDROM_CMD, &gd);
 	free_irq(HW_EVENT_GDROM_DMA, &gd);

drivers/memstick/core/ms_block.c

@@ -2187,7 +2187,6 @@ static void msb_remove(struct memstick_dev *card)
 
 	/* Remove the disk */
 	del_gendisk(msb->disk);
-	blk_cleanup_queue(msb->queue);
 	blk_mq_free_tag_set(&msb->tag_set);
 	msb->queue = NULL;
 

drivers/memstick/core/mspro_block.c

@@ -1294,7 +1294,6 @@ static void mspro_block_remove(struct memstick_dev *card)
 	del_gendisk(msb->disk);
 	dev_dbg(&card->dev, "mspro block remove\n");
 
-	blk_cleanup_queue(msb->queue);
 	blk_mq_free_tag_set(&msb->tag_set);
 	msb->queue = NULL;
 

drivers/mmc/core/block.c

@@ -2509,7 +2509,6 @@ static struct mmc_blk_data *mmc_blk_alloc_req(struct mmc_card *card,
 	return md;
 
  err_cleanup_queue:
-	blk_cleanup_queue(md->disk->queue);
 	blk_mq_free_tag_set(&md->queue.tag_set);
  err_kfree:
 	kfree(md);

drivers/mmc/core/queue.c

@@ -494,7 +494,6 @@ void mmc_cleanup_queue(struct mmc_queue *mq)
 	if (blk_queue_quiesced(q))
 		blk_mq_unquiesce_queue(q);
 
-	blk_cleanup_queue(q);
 	blk_mq_free_tag_set(&mq->tag_set);
 
 	/*

drivers/nvme/host/apple.c

@@ -1502,7 +1502,7 @@ static int apple_nvme_probe(struct platform_device *pdev)
 
 	if (!blk_get_queue(anv->ctrl.admin_q)) {
 		nvme_start_admin_queue(&anv->ctrl);
-		blk_cleanup_queue(anv->ctrl.admin_q);
+		blk_mq_destroy_queue(anv->ctrl.admin_q);
 		anv->ctrl.admin_q = NULL;
 		ret = -ENODEV;
 		goto put_dev;

drivers/nvme/host/core.c

@@ -4103,7 +4103,6 @@ static void nvme_ns_remove(struct nvme_ns *ns)
 	if (!nvme_ns_head_multipath(ns->head))
 		nvme_cdev_del(&ns->cdev, &ns->cdev_device);
 	del_gendisk(ns->disk);
-	blk_cleanup_queue(ns->queue);
 
 	down_write(&ns->ctrl->namespaces_rwsem);
 	list_del_init(&ns->list);