Re: [PATCH 1/6] blk-mq: figure out correct numa node for hw queue
From: John Garry <hidden>
Date: 2022-03-01 19:19:52
On 28/02/2022 09:04, Ming Lei wrote:
The current code always uses default queue map and hw queue index for figuring out the numa node for hw queue, this way isn't correct because blk-mq supports three queue maps, and the correct queue map should be used for the specified hw queue. Signed-off-by: Ming Lei <redacted> ---
Hi Ming, Just some small comments to consider if you need to respin. Thanks, John
quoted hunk ↗ jump to hunk
block/blk-mq.c | 36 ++++++++++++++++++++++++++++++------ 1 file changed, 30 insertions(+), 6 deletions(-)diff --git a/block/blk-mq.c b/block/blk-mq.c index a05ce7725031..931add81813b 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c@@ -3107,15 +3107,41 @@ void blk_mq_free_rq_map(struct blk_mq_tags *tags) blk_mq_free_tags(tags); } +static int
enum hctx_type?
hctx_idx_to_type(struct blk_mq_tag_set *set,
+ unsigned int hctx_idx)
+{
+ int j;super nit: normally use i
+
+ for (j = 0; j < set->nr_maps; j++) {
+ unsigned int start = set->map[j].queue_offset;nit: double whitespace intentional?
+ unsigned int end = start + set->map[j].nr_queues;
+
+ if (hctx_idx >= start && hctx_idx < end)
+ break;
+ }
+
+ if (j >= set->nr_maps)
+ j = HCTX_TYPE_DEFAULT;
+
+ return j;
+}
+
+static int blk_mq_get_hctx_node(struct blk_mq_tag_set *set,
+ unsigned int hctx_idx)
+{
+ int type = hctx_idx_to_type(set, hctx_idx);
+
+ return blk_mq_hw_queue_to_node(&set->map[type], hctx_idx);
+}
+
static struct blk_mq_tags *blk_mq_alloc_rq_map(struct blk_mq_tag_set *set,
unsigned int hctx_idx,
unsigned int nr_tags,
unsigned int reserved_tags)
{
struct blk_mq_tags *tags;
- int node;
+ int node = blk_mq_get_hctx_node(set, hctx_idx);nit: the code originally had reverse firtree ordering, which I suppose is not by mistake
quoted hunk ↗ jump to hunk
- node = blk_mq_hw_queue_to_node(&set->map[HCTX_TYPE_DEFAULT], hctx_idx); if (node == NUMA_NO_NODE) node = set->numa_node;@@ -3165,9 +3191,8 @@ static int blk_mq_alloc_rqs(struct blk_mq_tag_set *set, { unsigned int i, j, entries_per_page, max_order = 4; size_t rq_size, left; - int node; + int node = blk_mq_get_hctx_node(set, hctx_idx);
and here
quoted hunk ↗ jump to hunk
- node = blk_mq_hw_queue_to_node(&set->map[HCTX_TYPE_DEFAULT], hctx_idx); if (node == NUMA_NO_NODE) node = set->numa_node;@@ -3941,10 +3966,9 @@ static void blk_mq_realloc_hw_ctxs(struct blk_mq_tag_set *set, /* protect against switching io scheduler */ mutex_lock(&q->sysfs_lock); for (i = 0; i < set->nr_hw_queues; i++) { - int node; + int node = blk_mq_get_hctx_node(set, i); struct blk_mq_hw_ctx *hctx; - node = blk_mq_hw_queue_to_node(&set->map[HCTX_TYPE_DEFAULT], i); /* * If the hw queue has been mapped to another numa node, * we need to realloc the hctx. If allocation fails, fallback