Skip to contentSkip to conversation

nvme_fc: minor fixes on sqsize

PATCH New James Smart 3 participants 3 messages Mar 6, 2018 – Mar 7, 2018

Subsystems (2)
NVMe NVMe FC transport
lore
Share & tools
lore

James Smart

author
#0 lore raw

PATCH 1/1 nvme_fc: minor fixes on sqsize

Corrected two outstanding issues in the transport around sqsize.

The first is the Create Connection LS, which is to use a sqvalue which mirror what will be in the fabric connect, is sending count rather than count - 1.

The second is the hw queue for the admin queue. It's using the define that corresponds to the count - 1 value. It should be count.

Signed-off-by James Smart <james.smart at broadcom.com>
---
 drivers/nvme/host/fc.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/nvme/host/fc.c +2 -2
index 8cacaccac90d..fc00c609c695 100644
@@ -1323,7 +1323,7 @@ nvme_fc_connect_queue(struct nvme_fc_ctrl *ctrl, struct nvme_fc_queue *queue,
 				sizeof(struct fcnvme_lsdesc_cr_conn_cmd));
 	conn_rqst->connect_cmd.ersp_ratio = cpu_to_be16(ersp_ratio);
 	conn_rqst->connect_cmd.qid  = cpu_to_be16(queue->qnum);
-	conn_rqst->connect_cmd.sqsize = cpu_to_be16(qsize);
+	conn_rqst->connect_cmd.sqsize = cpu_to_be16(qsize - 1);
 
 	lsop->queue = queue;
 	lsreq->rqstaddr = conn_rqst;

@@ -2643,7 +2643,7 @@ nvme_fc_create_association(struct nvme_fc_ctrl *ctrl)
 	nvme_fc_init_queue(ctrl, 0);
 
 	ret = __nvme_fc_create_hw_queue(ctrl, &ctrl->queues[0], 0,
-				NVME_AQ_BLK_MQ_DEPTH);
+				NVME_AQ_BLK_MQ_DEPTH + 1);
 	if (ret)
 		goto out_free_queue;
Signature · 1 line
2.13.1
#1 lore raw

PATCH 1/1 nvme_fc: minor fixes on sqsize

drivers/nvme/host/fc.c nvme_fc_create_association(struct nvme_fc_ctrl *ctrl) · line 2646

Quoted 3 lines · James Smart
drivers/nvme/host/fc.c +1 -1
 	ret = __nvme_fc_create_hw_queue(ctrl, &ctrl->queues[0], 0,
-				NVME_AQ_BLK_MQ_DEPTH);
+				NVME_AQ_BLK_MQ_DEPTH + 1);

A bit of a nit, but I think you'd prefer to use NVME_AQ_DEPTH here instead of assuming the +1.

#2 lore raw

PATCH 1/1 nvme_fc: minor fixes on sqsize

Quoted 6 lines · Keith Busch

drivers/nvme/host/fc.c nvme_fc_create_association(struct nvme_fc_ctrl *ctrl) · line 2646

Quoted 3 lines · James Smart

ret = __nvme_fc_create_hw_queue(ctrl, &ctrl->queues[0], 0,

drivers/nvme/host/fc.c +1 -1
-				NVME_AQ_BLK_MQ_DEPTH);
+				NVME_AQ_BLK_MQ_DEPTH + 1);

A bit of a nit, but I think you'd prefer to use NVME_AQ_DEPTH here instead of assuming the +1.

NP.?? I'll address it. I need to rework the patch as I found another area when we tune for controller limits that isn't right either.

-- james