Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1571856 > unrolled thread

[PATCH V2 1/3] blk-mq: allocate blk_mq_tags and requests in correct node

Back to article view | Back to linux.kernel

Contents

  [PATCH V2 1/3] blk-mq: allocate blk_mq_tags and requests in correct node Shaohua Li <shli@fb.com> - 2017-02-01 19:00 +0100
    [PATCH V2 3/3] nvme: allocate nvme_queue in correct node Shaohua Li <shli@fb.com> - 2017-02-01 19:00 +0100
      Re: [PATCH V2 3/3] nvme: allocate nvme_queue in correct node Christoph Hellwig <hch@lst.de> - 2017-02-01 19:20 +0100
    Re: [PATCH V2 1/3] blk-mq: allocate blk_mq_tags and requests in         correct node Christoph Hellwig <hch@lst.de> - 2017-02-01 19:20 +0100

#1571856 — [PATCH V2 1/3] blk-mq: allocate blk_mq_tags and requests in correct node

blk_mq_tags/requests of specific hardware queue are mostly used in
specific cpus, which might not be in the same numa node as disk. For
example, a nvme card is in node 0. half hardware queue will be used by
node 0, the other node 1.

Signed-off-by: Shaohua Li <shli@fb.com>
---
 block/blk-mq.c | 21 +++++++++++++++------
 1 file changed, 15 insertions(+), 6 deletions(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 48df5fd..888077c 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -1666,16 +1666,20 @@ struct blk_mq_tags *blk_mq_alloc_rq_map(struct blk_mq_tag_set *set,
 					unsigned int reserved_tags)
 {
 	struct blk_mq_tags *tags;
+	int node;
+
+	node = blk_mq_hw_queue_to_node(set->mq_map, hctx_idx);
+	if (node == NUMA_NO_NODE)
+		node = set->numa_node;
 
-	tags = blk_mq_init_tags(nr_tags, reserved_tags,
-				set->numa_node,
+	tags = blk_mq_init_tags(nr_tags, reserved_tags, node,
 				BLK_MQ_FLAG_TO_ALLOC_POLICY(set->flags));
 	if (!tags)
 		return NULL;
 
 	tags->rqs = kzalloc_node(nr_tags * sizeof(struct request *),
 				 GFP_NOIO | __GFP_NOWARN | __GFP_NORETRY,
-				 set->numa_node);
+				 node);
 	if (!tags->rqs) {
 		blk_mq_free_tags(tags);
 		return NULL;
@@ -1683,7 +1687,7 @@ struct blk_mq_tags *blk_mq_alloc_rq_map(struct blk_mq_tag_set *set,
 
 	tags->static_rqs = kzalloc_node(nr_tags * sizeof(struct request *),
 				 GFP_NOIO | __GFP_NOWARN | __GFP_NORETRY,
-				 set->numa_node);
+				 node);
 	if (!tags->static_rqs) {
 		kfree(tags->rqs);
 		blk_mq_free_tags(tags);
@@ -1703,6 +1707,11 @@ int blk_mq_alloc_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
 {
 	unsigned int i, j, entries_per_page, max_order = 4;
 	size_t rq_size, left;
+	int node;
+
+	node = blk_mq_hw_queue_to_node(set->mq_map, hctx_idx);
+	if (node == NUMA_NO_NODE)
+		node = set->numa_node;
 
 	INIT_LIST_HEAD(&tags->page_list);
 
@@ -1724,7 +1733,7 @@ int blk_mq_alloc_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
 			this_order--;
 
 		do {
-			page = alloc_pages_node(set->numa_node,
+			page = alloc_pages_node(node,
 				GFP_NOIO | __GFP_NOWARN | __GFP_NORETRY | __GFP_ZERO,
 				this_order);
 			if (page)
@@ -1757,7 +1766,7 @@ int blk_mq_alloc_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
 			if (set->ops->init_request) {
 				if (set->ops->init_request(set->driver_data,
 						rq, hctx_idx, i,
-						set->numa_node)) {
+						node)) {
 					tags->static_rqs[i] = NULL;
 					goto fail;
 				}
-- 
2.9.3

[toc] | [next] | [standalone]

#1571858 — [PATCH V2 3/3] nvme: allocate nvme_queue in correct node

nvme_queue is per-cpu queue (mostly). Allocating it in node where blk-mq
will use it.

Signed-off-by: Shaohua Li <shli@fb.com>
---
 drivers/nvme/host/pci.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index 032237c..9733008 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -1040,9 +1040,10 @@ static int nvme_alloc_sq_cmds(struct nvme_dev *dev, struct nvme_queue *nvmeq,
 }
 
 static struct nvme_queue *nvme_alloc_queue(struct nvme_dev *dev, int qid,
-							int depth)
+							int depth, int node)
 {
-	struct nvme_queue *nvmeq = kzalloc(sizeof(*nvmeq), GFP_KERNEL);
+	struct nvme_queue *nvmeq = kzalloc_node(sizeof(*nvmeq), GFP_KERNEL,
+							node);
 	if (!nvmeq)
 		return NULL;
 
@@ -1219,7 +1220,8 @@ static int nvme_configure_admin_queue(struct nvme_dev *dev)
 
 	nvmeq = dev->queues[0];
 	if (!nvmeq) {
-		nvmeq = nvme_alloc_queue(dev, 0, NVME_AQ_DEPTH);
+		nvmeq = nvme_alloc_queue(dev, 0, NVME_AQ_DEPTH,
+					dev_to_node(dev->dev));
 		if (!nvmeq)
 			return -ENOMEM;
 	}
@@ -1311,7 +1313,9 @@ static int nvme_create_io_queues(struct nvme_dev *dev)
 	int ret = 0;
 
 	for (i = dev->queue_count; i <= dev->max_qid; i++) {
-		if (!nvme_alloc_queue(dev, i, dev->q_depth)) {
+		/* vector == qid - 1, match nvme_create_queue */
+		if (!nvme_alloc_queue(dev, i, dev->q_depth,
+		     pci_irq_get_node(to_pci_dev(dev->dev), i - 1))) {
 			ret = -ENOMEM;
 			break;
 		}
-- 
2.9.3

[toc] | [prev] | [next] | [standalone]

#1571873 — Re: [PATCH V2 3/3] nvme: allocate nvme_queue in correct node

Looks fine,

Reviewed-by: Christoph Hellwig <hch@lst.de>

[toc] | [prev] | [next] | [standalone]

#1571867 — Re: [PATCH V2 1/3] blk-mq: allocate blk_mq_tags and requests in correct node

Looks fine,

Reviewed-by: Christoph Hellwig <hch@lst.de>

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web