blk-mq: dynamic h/w context count

The hardware's provided queue count may change at runtime with resource
provisioning. This patch allows a block driver to alter the number of
h/w queues available when its resource count changes.

The main part is a new blk-mq API to request a new number of h/w queues
for a given live tag set. The new API freezes all queues using that set,
then adjusts the allocated count prior to remapping these to CPUs.

The bulk of the rest just shifts where h/w contexts and all their
artifacts are allocated and freed.

The number of max h/w contexts is capped to the number of possible cpus
since there is no use for more than that. As such, all pre-allocated
memory for pointers need to account for the max possible rather than
the initial number of queues.

A side effect of this is that the blk-mq will proceed successfully as
long as it can allocate at least one h/w context. Previously it would
fail request queue initialization if less than the requested number
was allocated.

Signed-off-by: Keith Busch <keith.busch@intel.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: Jon Derrick <jonathan.derrick@intel.com>
Signed-off-by: Jens Axboe <axboe@fb.com>

1 files changed, 104 insertions, 69 deletions

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 4c0622f..645eb9e 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -1742,31 +1742,6 @@ static int blk_mq_init_hctx(struct request_queue *q,
 	return -1;
 }
 
-static int blk_mq_init_hw_queues(struct request_queue *q,
-		struct blk_mq_tag_set *set)
-{
-	struct blk_mq_hw_ctx *hctx;
-	unsigned int i;
-
-	/*
-	 * Initialize hardware queues
-	 */
-	queue_for_each_hw_ctx(q, hctx, i) {
-		if (blk_mq_init_hctx(q, set, hctx, i))
-			break;
-	}
-
-	if (i == q->nr_hw_queues)
-		return 0;
-
-	/*
-	 * Init failed
-	 */
-	blk_mq_exit_hw_queues(q, set, i);
-
-	return 1;
-}
-
 static void blk_mq_init_cpu_queues(struct request_queue *q,
 				   unsigned int nr_hw_queues)
 {
@@ -1824,6 +1799,7 @@ static void blk_mq_map_swqueue(struct request_queue *q,
 			continue;
 
 		hctx = q->mq_ops->map_queue(q, i);
+
 		cpumask_set_cpu(i, hctx->cpumask);
 		ctx->index_hw = hctx->nr_ctx;
 		hctx->ctxs[hctx->nr_ctx++] = ctx;
@@ -1972,54 +1948,89 @@ struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *set)
 }
 EXPORT_SYMBOL(blk_mq_init_queue);
 
-struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
-						  struct request_queue *q)
+static void blk_mq_realloc_hw_ctxs(struct blk_mq_tag_set *set,
+						struct request_queue *q)
 {
-	struct blk_mq_hw_ctx **hctxs;
-	struct blk_mq_ctx __percpu *ctx;
-	unsigned int *map;
-	int i;
-
-	ctx = alloc_percpu(struct blk_mq_ctx);
-	if (!ctx)
-		return ERR_PTR(-ENOMEM);
-
-	hctxs = kmalloc_node(set->nr_hw_queues * sizeof(*hctxs), GFP_KERNEL,
-			set->numa_node);
-
-	if (!hctxs)
-		goto err_percpu;
-
-	map = blk_mq_make_queue_map(set);
-	if (!map)
-		goto err_map;
+	int i, j;
+	struct blk_mq_hw_ctx **hctxs = q->queue_hw_ctx;
 
+	blk_mq_sysfs_unregister(q);
 	for (i = 0; i < set->nr_hw_queues; i++) {
-		int node = blk_mq_hw_queue_to_node(map, i);
+		int node;
 
+		if (hctxs[i])
+			continue;
+
+		node = blk_mq_hw_queue_to_node(q->mq_map, i);
 		hctxs[i] = kzalloc_node(sizeof(struct blk_mq_hw_ctx),
 					GFP_KERNEL, node);
 		if (!hctxs[i])
-			goto err_hctxs;
+			break;
 
 		if (!zalloc_cpumask_var_node(&hctxs[i]->cpumask, GFP_KERNEL,
-						node))
-			goto err_hctxs;
+						node)) {
+			kfree(hctxs[i]);
+			hctxs[i] = NULL;
+			break;
+		}
 
 		atomic_set(&hctxs[i]->nr_active, 0);
 		hctxs[i]->numa_node = node;
 		hctxs[i]->queue_num = i;
+
+		if (blk_mq_init_hctx(q, set, hctxs[i], i)) {
+			free_cpumask_var(hctxs[i]->cpumask);
+			kfree(hctxs[i]);
+			hctxs[i] = NULL;
+			break;
+		}
+		blk_mq_hctx_kobj_init(hctxs[i]);
 	}
+	for (j = i; j < q->nr_hw_queues; j++) {
+		struct blk_mq_hw_ctx *hctx = hctxs[j];
+
+		if (hctx) {
+			if (hctx->tags) {
+				blk_mq_free_rq_map(set, hctx->tags, j);
+				set->tags[j] = NULL;
+			}
+			blk_mq_exit_hctx(q, set, hctx, j);
+			free_cpumask_var(hctx->cpumask);
+			kobject_put(&hctx->kobj);
+			kfree(hctx->ctxs);
+			kfree(hctx);
+			hctxs[j] = NULL;
+
+		}
+	}
+	q->nr_hw_queues = i;
+	blk_mq_sysfs_register(q);
+}
+
+struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
+						  struct request_queue *q)
+{
+	q->queue_ctx = alloc_percpu(struct blk_mq_ctx);
+	if (!q->queue_ctx)
+		return ERR_PTR(-ENOMEM);
+
+	q->queue_hw_ctx = kzalloc_node(nr_cpu_ids * sizeof(*(q->queue_hw_ctx)),
+						GFP_KERNEL, set->numa_node);
+	if (!q->queue_hw_ctx)
+		goto err_percpu;
+
+	q->mq_map = blk_mq_make_queue_map(set);
+	if (!q->mq_map)
+		goto err_map;
+
+	blk_mq_realloc_hw_ctxs(set, q);
+	if (!q->nr_hw_queues)
+		goto err_hctxs;
 
 	INIT_WORK(&q->timeout_work, blk_mq_timeout_work);
 	blk_queue_rq_timeout(q, set->timeout ? set->timeout : 30 * HZ);
 
 	q->nr_queues = nr_cpu_ids;
-	q->nr_hw_queues = set->nr_hw_queues;
-	q->mq_map = map;
-
-	q->queue_ctx = ctx;
-	q->queue_hw_ctx = hctxs;
 
 	q->mq_ops = set->ops;
 	q->queue_flags |= QUEUE_FLAG_MQ_DEFAULT;
@@ -2048,9 +2059,6 @@ struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
 
 	blk_mq_init_cpu_queues(q, set->nr_hw_queues);
 
-	if (blk_mq_init_hw_queues(q, set))
-		goto err_hctxs;
-
 	get_online_cpus();
 	mutex_lock(&all_q_mutex);
 
@@ -2064,17 +2072,11 @@ struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
 	return q;
 
 err_hctxs:
-	kfree(map);
-	for (i = 0; i < set->nr_hw_queues; i++) {
-		if (!hctxs[i])
-			break;
-		free_cpumask_var(hctxs[i]->cpumask);
-		kfree(hctxs[i]);
-	}
+	kfree(q->mq_map);
 err_map:
-	kfree(hctxs);
+	kfree(q->queue_hw_ctx);
 err_percpu:
-	free_percpu(ctx);
+	free_percpu(q->queue_ctx);
 	return ERR_PTR(-ENOMEM);
 }
 EXPORT_SYMBOL(blk_mq_init_allocated_queue);
@@ -2282,9 +2284,13 @@ int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set)
 		set->nr_hw_queues = 1;
 		set->queue_depth = min(64U, set->queue_depth);
 	}
+	/*
+	 * There is no use for more h/w queues than cpus.
+	 */
+	if (set->nr_hw_queues > nr_cpu_ids)
+		set->nr_hw_queues = nr_cpu_ids;
 
-	set->tags = kmalloc_node(set->nr_hw_queues *
-				 sizeof(struct blk_mq_tags *),
+	set->tags = kzalloc_node(nr_cpu_ids * sizeof(struct blk_mq_tags *),
 				 GFP_KERNEL, set->numa_node);
 	if (!set->tags)
 		return -ENOMEM;
@@ -2307,7 +2313,7 @@ void blk_mq_free_tag_set(struct blk_mq_tag_set *set)
 {
 	int i;
 
-	for (i = 0; i < set->nr_hw_queues; i++) {
+	for (i = 0; i < nr_cpu_ids; i++) {
 		if (set->tags[i])
 			blk_mq_free_rq_map(set, set->tags[i], i);
 	}
@@ -2339,6 +2345,35 @@ int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
 	return ret;
 }
 
+void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues)
+{
+	struct request_queue *q;
+
+	if (nr_hw_queues > nr_cpu_ids)
+		nr_hw_queues = nr_cpu_ids;
+	if (nr_hw_queues < 1 || nr_hw_queues == set->nr_hw_queues)
+		return;
+
+	list_for_each_entry(q, &set->tag_list, tag_set_list)
+		blk_mq_freeze_queue(q);
+
+	set->nr_hw_queues = nr_hw_queues;
+	list_for_each_entry(q, &set->tag_list, tag_set_list) {
+		blk_mq_realloc_hw_ctxs(set, q);
+
+		if (q->nr_hw_queues > 1)
+			blk_queue_make_request(q, blk_mq_make_request);
+		else
+			blk_queue_make_request(q, blk_sq_make_request);
+
+		blk_mq_queue_reinit(q, cpu_online_mask);
+	}
+
+	list_for_each_entry(q, &set->tag_list, tag_set_list)
+		blk_mq_unfreeze_queue(q);
+}
+EXPORT_SYMBOL_GPL(blk_mq_update_nr_hw_queues);
+
 void blk_mq_disable_hotplug(void)
 {
 	mutex_lock(&all_q_mutex);