What we really nee to do is to always set the NOWAIT flag (we have a
reserved tag for connect anyway), and thus never trigger the code
deep down in bt_get that might switch to a different hctx.
Below is the version that I've tested together with the NVMe change
to use the NOWAIT flag:
---
>From 2346a4fdcc57c70a671e1329f7550d91d4e9d8a8 Mon Sep 17 00:00:00 2001
From: Ming Lin <ming.l@xxxxxxxxxxxxxxx>
Date: Mon, 30 Nov 2015 19:45:48 +0100
Subject: blk-mq: add blk_mq_alloc_request_hctx
For some protocols like NVMe over Fabrics we need to be able to send
initialization commands to a specific queue.
Based on an earlier patch from Christoph Hellwig <hch@xxxxxx>.
Signed-off-by: Ming Lin <ming.l@xxxxxxxxxxxxxxx>
[hch: disallow sleeping allocation, req_op fixes]
Signed-off-by: Christoph Hellwig <hch@xxxxxx>
---
block/blk-mq.c | 39 +++++++++++++++++++++++++++++++++++++++
include/linux/blk-mq.h | 2 ++
2 files changed, 41 insertions(+)
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 13f4603..7aa60c4 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -267,6 +267,45 @@ struct request *blk_mq_alloc_request(struct request_queue *q, int rw,
}
EXPORT_SYMBOL(blk_mq_alloc_request);
+struct request *blk_mq_alloc_request_hctx(struct request_queue *q, int rw,
+ unsigned int flags, unsigned int hctx_idx)
+{
+ struct blk_mq_hw_ctx *hctx;
+ struct blk_mq_ctx *ctx;
+ struct request *rq;
+ struct blk_mq_alloc_data alloc_data;
+ int ret;
+
+ /*
+ * If the tag allocator sleeps we could get an allocation for a
+ * different hardware context. No need to complicate the low level
+ * allocator for this for the rare use case of a command tied to
+ * a specific queue.
+ */
+ if (WARN_ON_ONCE(!(flags & BLK_MQ_REQ_NOWAIT)))
+ return ERR_PTR(-EINVAL);
+
+ if (hctx_idx >= q->nr_hw_queues)
+ return ERR_PTR(-EIO);
+
+ ret = blk_queue_enter(q, true);
+ if (ret)
+ return ERR_PTR(ret);
+
+ hctx = q->queue_hw_ctx[hctx_idx];
+ ctx = __blk_mq_get_ctx(q, cpumask_first(hctx->cpumask));
+
+ blk_mq_set_alloc_data(&alloc_data, q, flags, ctx, hctx);
+ rq = __blk_mq_alloc_request(&alloc_data, rw, 0);
+ if (!rq) {
+ blk_queue_exit(q);
+ return ERR_PTR(-EWOULDBLOCK);
+ }
+
+ return rq;
+}
+EXPORT_SYMBOL_GPL(blk_mq_alloc_request_hctx);
+
static void __blk_mq_free_request(struct blk_mq_hw_ctx *hctx,
struct blk_mq_ctx *ctx, struct request *rq)
{
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index faa7d5c2..e43bbff 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -198,6 +198,8 @@ enum {
struct request *blk_mq_alloc_request(struct request_queue *q, int rw,
unsigned int flags);
+struct request *blk_mq_alloc_request_hctx(struct request_queue *q, int op,
+ unsigned int flags, unsigned int hctx_idx);
struct request *blk_mq_tag_to_rq(struct blk_mq_tags *tags, unsigned int tag);
struct cpumask *blk_mq_tags_cpumask(struct blk_mq_tags *tags);
--
2.1.4
--
To unsubscribe from this list: send the line "unsubscribe linux-block" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
