block: move the poll flag to queue_limits

Move the poll flag into the queue_limits feature field so that it can
be set atomically with the queue frozen.

Stacking drivers are simplified in that they now can simply set the
flag, and blk_stack_limits will clear it when the features is not
supported by any of the underlying devices.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-22-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block/blk-core.c

@@ -791,7 +791,7 @@ void submit_bio_noacct(struct bio *bio)
 		}
 	}
 
-	if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
+	if (!(q->limits.features & BLK_FEAT_POLL))
 		bio_clear_polled(bio);
 
 	switch (bio_op(bio)) {
@@ -915,8 +915,7 @@ int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags)
 		return 0;
 
 	q = bdev_get_queue(bdev);
-	if (cookie == BLK_QC_T_NONE ||
-	    !test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
+	if (cookie == BLK_QC_T_NONE || !(q->limits.features & BLK_FEAT_POLL))
 		return 0;
 
 	blk_flush_plug(current->plug, false);

block/blk-mq-debugfs.c

@@ -87,7 +87,6 @@ static const char *const blk_queue_flag_name[] = {
 	QUEUE_FLAG_NAME(NOXMERGES),
 	QUEUE_FLAG_NAME(SAME_FORCE),
 	QUEUE_FLAG_NAME(INIT_DONE),
-	QUEUE_FLAG_NAME(POLL),
 	QUEUE_FLAG_NAME(STATS),
 	QUEUE_FLAG_NAME(REGISTERED),
 	QUEUE_FLAG_NAME(QUIESCED),

block/blk-mq.c

@@ -4109,6 +4109,12 @@ void blk_mq_release(struct request_queue *q)
 	blk_mq_sysfs_deinit(q);
 }
 
+static bool blk_mq_can_poll(struct blk_mq_tag_set *set)
+{
+	return set->nr_maps > HCTX_TYPE_POLL &&
+		set->map[HCTX_TYPE_POLL].nr_queues;
+}
+
 struct request_queue *blk_mq_alloc_queue(struct blk_mq_tag_set *set,
 		struct queue_limits *lim, void *queuedata)
 {
@@ -4119,6 +4125,8 @@ struct request_queue *blk_mq_alloc_queue(struct blk_mq_tag_set *set,
 	if (!lim)
 		lim = &default_lim;
 	lim->features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT;
+	if (blk_mq_can_poll(set))
+		lim->features |= BLK_FEAT_POLL;
 
 	q = blk_alloc_queue(lim, set->numa_node);
 	if (IS_ERR(q))
@@ -4273,17 +4281,6 @@ static void blk_mq_realloc_hw_ctxs(struct blk_mq_tag_set *set,
 	mutex_unlock(&q->sysfs_lock);
 }
 
-static void blk_mq_update_poll_flag(struct request_queue *q)
-{
-	struct blk_mq_tag_set *set = q->tag_set;
-
-	if (set->nr_maps > HCTX_TYPE_POLL &&
-	    set->map[HCTX_TYPE_POLL].nr_queues)
-		blk_queue_flag_set(QUEUE_FLAG_POLL, q);
-	else
-		blk_queue_flag_clear(QUEUE_FLAG_POLL, q);
-}
-
 int blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
 		struct request_queue *q)
 {
@@ -4311,7 +4308,6 @@ int blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
 	q->tag_set = set;
 
 	q->queue_flags |= QUEUE_FLAG_MQ_DEFAULT;
-	blk_mq_update_poll_flag(q);
 
 	INIT_DELAYED_WORK(&q->requeue_work, blk_mq_requeue_work);
 	INIT_LIST_HEAD(&q->flush_list);
@@ -4798,8 +4794,10 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
 fallback:
 	blk_mq_update_queue_map(set);
 	list_for_each_entry(q, &set->tag_list, tag_set_list) {
+		struct queue_limits lim;
+
 		blk_mq_realloc_hw_ctxs(set, q);
-		blk_mq_update_poll_flag(q);
+
 		if (q->nr_hw_queues != set->nr_hw_queues) {
 			int i = prev_nr_hw_queues;
 
@@ -4811,6 +4809,13 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
 			set->nr_hw_queues = prev_nr_hw_queues;
 			goto fallback;
 		}
+		lim = queue_limits_start_update(q);
+		if (blk_mq_can_poll(set))
+			lim.features |= BLK_FEAT_POLL;
+		else
+			lim.features &= ~BLK_FEAT_POLL;
+		if (queue_limits_commit_update(q, &lim) < 0)
+			pr_warn("updating the poll flag failed\n");
 		blk_mq_map_swqueue(q);
 	}
 

block/blk-settings.c

@@ -460,13 +460,15 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
 	t->features |= (b->features & BLK_FEAT_INHERIT_MASK);
 
 	/*
-	 * BLK_FEAT_NOWAIT needs to be supported both by the stacking driver
-	 * and all underlying devices.  The stacking driver sets the flag
-	 * before stacking the limits, and this will clear the flag if any
-	 * of the underlying devices does not support it.
+	 * BLK_FEAT_NOWAIT and BLK_FEAT_POLL need to be supported both by the
+	 * stacking driver and all underlying devices.  The stacking driver sets
+	 * the flags before stacking the limits, and this will clear the flags
+	 * if any of the underlying devices does not support it.
 	 */
 	if (!(b->features & BLK_FEAT_NOWAIT))
 		t->features &= ~BLK_FEAT_NOWAIT;
+	if (!(b->features & BLK_FEAT_POLL))
+		t->features &= ~BLK_FEAT_POLL;
 
 	t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
 	t->max_user_sectors = min_not_zero(t->max_user_sectors,

block/blk-sysfs.c

@@ -394,13 +394,13 @@ static ssize_t queue_poll_delay_store(struct request_queue *q, const char *page,
 
 static ssize_t queue_poll_show(struct request_queue *q, char *page)
 {
-	return queue_var_show(test_bit(QUEUE_FLAG_POLL, &q->queue_flags), page);
+	return queue_var_show(q->limits.features & BLK_FEAT_POLL, page);
 }
 
 static ssize_t queue_poll_store(struct request_queue *q, const char *page,
 				size_t count)
 {
-	if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
+	if (!(q->limits.features & BLK_FEAT_POLL))
 		return -EINVAL;
 	pr_info_ratelimited("writes to the poll attribute are ignored.\n");
 	pr_info_ratelimited("please use driver specific parameters instead.\n");

drivers/md/dm-table.c

@@ -582,7 +582,7 @@ int dm_split_args(int *argc, char ***argvp, char *input)
 static void dm_set_stacking_limits(struct queue_limits *limits)
 {
 	blk_set_stacking_limits(limits);
-	limits->features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT;
+	limits->features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT | BLK_FEAT_POLL;
 }
 
 /*
@@ -1024,14 +1024,13 @@ bool dm_table_request_based(struct dm_table *t)
 	return __table_type_request_based(dm_table_get_type(t));
 }
 
-static bool dm_table_supports_poll(struct dm_table *t);
-
 static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *md)
 {
 	enum dm_queue_mode type = dm_table_get_type(t);
 	unsigned int per_io_data_size = 0, front_pad, io_front_pad;
 	unsigned int min_pool_size = 0, pool_size;
 	struct dm_md_mempools *pools;
+	unsigned int bioset_flags = 0;
 
 	if (unlikely(type == DM_TYPE_NONE)) {
 		DMERR("no table type is set, can't allocate mempools");
@@ -1048,6 +1047,9 @@ static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *
 		goto init_bs;
 	}
 
+	if (md->queue->limits.features & BLK_FEAT_POLL)
+		bioset_flags |= BIOSET_PERCPU_CACHE;
+
 	for (unsigned int i = 0; i < t->num_targets; i++) {
 		struct dm_target *ti = dm_table_get_target(t, i);
 
@@ -1060,8 +1062,7 @@ static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *
 
 	io_front_pad = roundup(per_io_data_size,
 		__alignof__(struct dm_io)) + DM_IO_BIO_OFFSET;
-	if (bioset_init(&pools->io_bs, pool_size, io_front_pad,
-			dm_table_supports_poll(t) ? BIOSET_PERCPU_CACHE : 0))
+	if (bioset_init(&pools->io_bs, pool_size, io_front_pad, bioset_flags))
 		goto out_free_pools;
 	if (t->integrity_supported &&
 	    bioset_integrity_create(&pools->io_bs, pool_size))
@@ -1404,14 +1405,6 @@ struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
 	return &t->targets[(KEYS_PER_NODE * n) + k];
 }
 
-static int device_not_poll_capable(struct dm_target *ti, struct dm_dev *dev,
-				   sector_t start, sector_t len, void *data)
-{
-	struct request_queue *q = bdev_get_queue(dev->bdev);
-
-	return !test_bit(QUEUE_FLAG_POLL, &q->queue_flags);
-}
-
 /*
  * type->iterate_devices() should be called when the sanity check needs to
  * iterate and check all underlying data devices. iterate_devices() will
@@ -1459,19 +1452,6 @@ static int count_device(struct dm_target *ti, struct dm_dev *dev,
 	return 0;
 }
 
-static bool dm_table_supports_poll(struct dm_table *t)
-{
-	for (unsigned int i = 0; i < t->num_targets; i++) {
-		struct dm_target *ti = dm_table_get_target(t, i);
-
-		if (!ti->type->iterate_devices ||
-		    ti->type->iterate_devices(ti, device_not_poll_capable, NULL))
-			return false;
-	}
-
-	return true;
-}
-
 /*
  * Check whether a table has no data devices attached using each
  * target's iterate_devices method.
@@ -1817,6 +1797,13 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
 	if (!dm_table_supports_nowait(t))
 		limits->features &= ~BLK_FEAT_NOWAIT;
 
+	/*
+	 * The current polling impementation does not support request based
+	 * stacking.
+	 */
+	if (!__table_type_bio_based(t->type))
+		limits->features &= ~BLK_FEAT_POLL;
+
 	if (!dm_table_supports_discards(t)) {
 		limits->max_hw_discard_sectors = 0;
 		limits->discard_granularity = 0;
@@ -1858,21 +1845,6 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
 		return r;
 
 	dm_update_crypto_profile(q, t);
-
-	/*
-	 * Check for request-based device is left to
-	 * dm_mq_init_request_queue()->blk_mq_init_allocated_queue().
-	 *
-	 * For bio-based device, only set QUEUE_FLAG_POLL when all
-	 * underlying devices supporting polling.
-	 */
-	if (__table_type_bio_based(t->type)) {
-		if (dm_table_supports_poll(t))
-			blk_queue_flag_set(QUEUE_FLAG_POLL, q);
-		else
-			blk_queue_flag_clear(QUEUE_FLAG_POLL, q);
-	}
-
 	return 0;
 }
 

drivers/nvme/host/multipath.c

@@ -538,7 +538,7 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head)
 
 	blk_set_stacking_limits(&lim);
 	lim.dma_alignment = 3;
-	lim.features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT;
+	lim.features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT | BLK_FEAT_POLL;
 	if (head->ids.csi != NVME_CSI_ZNS)
 		lim.max_zone_append_sectors = 0;
 
@@ -549,16 +549,6 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head)
 	head->disk->private_data = head;
 	sprintf(head->disk->disk_name, "nvme%dn%d",
 			ctrl->subsys->instance, head->instance);
-
-	/*
-	 * This assumes all controllers that refer to a namespace either
-	 * support poll queues or not.  That is not a strict guarantee,
-	 * but if the assumption is wrong the effect is only suboptimal
-	 * performance but not correctness problem.
-	 */
-	if (ctrl->tagset->nr_maps > HCTX_TYPE_POLL &&
-	    ctrl->tagset->map[HCTX_TYPE_POLL].nr_queues)
-		blk_queue_flag_set(QUEUE_FLAG_POLL, head->disk->queue);
 	return 0;
 }
 

include/linux/blkdev.h

@@ -310,6 +310,9 @@ enum {
 
 	/* supports DAX */
 	BLK_FEAT_DAX				= (1u << 8),
+
+	/* supports I/O polling */
+	BLK_FEAT_POLL				= (1u << 9),
 };
 
 /*
@@ -577,7 +580,6 @@ struct request_queue {
 #define QUEUE_FLAG_NOXMERGES	9	/* No extended merges */
 #define QUEUE_FLAG_SAME_FORCE	12	/* force complete on same CPU */
 #define QUEUE_FLAG_INIT_DONE	14	/* queue is initialized */
-#define QUEUE_FLAG_POLL		16	/* IO polling enabled if set */
 #define QUEUE_FLAG_STATS	20	/* track IO start and completion times */
 #define QUEUE_FLAG_REGISTERED	22	/* queue has been registered to a disk */
 #define QUEUE_FLAG_QUIESCED	24	/* queue has been quiesced */