dmaengine.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] dmaengine: idxd: assign MSIX vectors to each WQ rather than roundrobin
@ 2021-06-24 20:39 Dave Jiang
  2021-07-14  6:58 ` Vinod Koul
  0 siblings, 1 reply; 2+ messages in thread
From: Dave Jiang @ 2021-06-24 20:39 UTC (permalink / raw)
  To: vkoul; +Cc: Konstantin Ananyev, Konstantin Ananyev, dmaengine

IOPS increased when changing MSIX vector to per WQ from roundrobin.
Allows descriptor to be completed by the submitter improves caching
locality.

Suggested-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
---
 drivers/dma/idxd/idxd.h   |    2 --
 drivers/dma/idxd/submit.c |   36 ++++++++----------------------------
 2 files changed, 8 insertions(+), 30 deletions(-)

diff --git a/drivers/dma/idxd/idxd.h b/drivers/dma/idxd/idxd.h
index 5516ea31ff07..cdefd304d396 100644
--- a/drivers/dma/idxd/idxd.h
+++ b/drivers/dma/idxd/idxd.h
@@ -178,7 +178,6 @@ struct idxd_wq {
 	enum idxd_wq_state state;
 	unsigned long flags;
 	union wqcfg *wqcfg;
-	u32 vec_ptr;		/* interrupt steering */
 	struct dsa_hw_desc **hw_descs;
 	int num_descs;
 	union {
@@ -314,7 +313,6 @@ struct idxd_desc {
 	struct list_head list;
 	int id;
 	int cpu;
-	unsigned int vector;
 	struct idxd_wq *wq;
 };
 
diff --git a/drivers/dma/idxd/submit.c b/drivers/dma/idxd/submit.c
index 19afb62abaff..406f15a84f11 100644
--- a/drivers/dma/idxd/submit.c
+++ b/drivers/dma/idxd/submit.c
@@ -22,22 +22,13 @@ static struct idxd_desc *__get_desc(struct idxd_wq *wq, int idx, int cpu)
 		desc->hw->pasid = idxd->pasid;
 
 	/*
-	 * Descriptor completion vectors are 1...N for MSIX. We will round
-	 * robin through the N vectors.
+	 * On host, MSIX vecotr 0 is used for misc interrupt. Therefore when we match
+	 * vector 1:1 to the WQ id, we need to add 1
 	 */
-	wq->vec_ptr = (wq->vec_ptr % idxd->num_wq_irqs) + 1;
-	if (!idxd->int_handles) {
-		desc->hw->int_handle = wq->vec_ptr;
-	} else {
-		desc->vector = wq->vec_ptr;
-		/*
-		 * int_handles are only for descriptor completion. However for device
-		 * MSIX enumeration, vec 0 is used for misc interrupts. Therefore even
-		 * though we are rotating through 1...N for descriptor interrupts, we
-		 * need to acqurie the int_handles from 0..N-1.
-		 */
-		desc->hw->int_handle = idxd->int_handles[desc->vector - 1];
-	}
+	if (!idxd->int_handles)
+		desc->hw->int_handle = wq->id + 1;
+	else
+		desc->hw->int_handle = idxd->int_handles[wq->id];
 
 	return desc;
 }
@@ -128,19 +119,8 @@ int idxd_submit_desc(struct idxd_wq *wq, struct idxd_desc *desc)
 	 * Pending the descriptor to the lockless list for the irq_entry
 	 * that we designated the descriptor to.
 	 */
-	if (desc->hw->flags & IDXD_OP_FLAG_RCI) {
-		int vec;
-
-		/*
-		 * If the driver is on host kernel, it would be the value
-		 * assigned to interrupt handle, which is index for MSIX
-		 * vector. If it's guest then can't use the int_handle since
-		 * that is the index to IMS for the entire device. The guest
-		 * device local index will be used.
-		 */
-		vec = !idxd->int_handles ? desc->hw->int_handle : desc->vector;
-		llist_add(&desc->llnode, &idxd->irq_entries[vec].pending_llist);
-	}
+	if (desc->hw->flags & IDXD_OP_FLAG_RCI)
+		llist_add(&desc->llnode, &idxd->irq_entries[wq->id + 1].pending_llist);
 
 	return 0;
 }



^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH] dmaengine: idxd: assign MSIX vectors to each WQ rather than roundrobin
  2021-06-24 20:39 [PATCH] dmaengine: idxd: assign MSIX vectors to each WQ rather than roundrobin Dave Jiang
@ 2021-07-14  6:58 ` Vinod Koul
  0 siblings, 0 replies; 2+ messages in thread
From: Vinod Koul @ 2021-07-14  6:58 UTC (permalink / raw)
  To: Dave Jiang; +Cc: Konstantin Ananyev, dmaengine

On 24-06-21, 13:39, Dave Jiang wrote:
> IOPS increased when changing MSIX vector to per WQ from roundrobin.
> Allows descriptor to be completed by the submitter improves caching
> locality.

Applied, thanks

-- 
~Vinod

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2021-07-14  6:58 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-06-24 20:39 [PATCH] dmaengine: idxd: assign MSIX vectors to each WQ rather than roundrobin Dave Jiang
2021-07-14  6:58 ` Vinod Koul

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).