public inbox for linux-wireless@vger.kernel.org
 help / color / mirror / Atom feed
From: Felix Fietkau <nbd@nbd.name>
To: linux-wireless@vger.kernel.org
Subject: [PATCH 5/8] mt76: dma: reduce lock contention in mt76_dma_tx_cleanup
Date: Fri, 25 Mar 2022 22:10:28 +0100	[thread overview]
Message-ID: <20220325211031.24171-5-nbd@nbd.name> (raw)
In-Reply-To: <20220325211031.24171-1-nbd@nbd.name>

Instead of taking q->lock for every single cleaned up frame, batch up to
16 cleaned up frames before taking it.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
 drivers/net/wireless/mediatek/mt76/dma.c | 26 +++++++++++++++++++++---
 1 file changed, 23 insertions(+), 3 deletions(-)

diff --git a/drivers/net/wireless/mediatek/mt76/dma.c b/drivers/net/wireless/mediatek/mt76/dma.c
index 09dc37bbf112..07065816262f 100644
--- a/drivers/net/wireless/mediatek/mt76/dma.c
+++ b/drivers/net/wireless/mediatek/mt76/dma.c
@@ -233,7 +233,9 @@ static void
 mt76_dma_tx_cleanup(struct mt76_dev *dev, struct mt76_queue *q, bool flush)
 {
 	struct mt76_queue_entry entry;
+	int done = 0;
 	int last;
+	int tail;
 
 	if (!q || !q->ndesc)
 		return;
@@ -244,21 +246,39 @@ mt76_dma_tx_cleanup(struct mt76_dev *dev, struct mt76_queue *q, bool flush)
 	else
 		last = readl(&q->regs->dma_idx);
 
-	while (q->queued > 0 && q->tail != last) {
+	tail = q->tail;
+	while (q->queued - done > 0 && tail != last) {
 		mt76_dma_tx_cleanup_idx(dev, q, q->tail, &entry);
-		mt76_queue_tx_complete(dev, q, &entry);
+		if (entry.skb)
+			dev->drv->tx_complete_skb(dev, &entry);
+		tail = (tail + 1) % q->ndesc;
+		done++;
 
 		if (entry.txwi) {
 			if (!(dev->drv->drv_flags & MT_DRV_TXWI_NO_FREE))
 				mt76_put_txwi(dev, entry.txwi);
 		}
 
-		if (!flush && q->tail == last)
+		if (!flush && tail == last)
 			last = readl(&q->regs->dma_idx);
 
+		if (done > 16) {
+			spin_lock_bh(&q->lock);
+			q->queued -= done;
+			q->tail = tail;
+			spin_unlock_bh(&q->lock);
+			done = 0;
+		}
 	}
 	spin_unlock_bh(&q->cleanup_lock);
 
+	if (done) {
+		spin_lock_bh(&q->lock);
+		q->queued -= done;
+		q->tail = tail;
+		spin_unlock_bh(&q->lock);
+	}
+
 	if (flush) {
 		spin_lock_bh(&q->lock);
 		mt76_dma_sync_idx(dev, q);
-- 
2.35.1


  parent reply	other threads:[~2022-03-25 21:10 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-25 21:10 [PATCH 1/8] mt76: mt7915: fix DBDC default band selection on MT7915D Felix Fietkau
2022-03-25 21:10 ` [PATCH 2/8] mt76: mt7915: rework hardware/phy initialization Felix Fietkau
2022-03-25 21:10 ` [PATCH 3/8] mt76: reduce tx queue lock hold time Felix Fietkau
2022-03-25 21:10 ` [PATCH 4/8] mt76: dma: use kzalloc instead of devm_kzalloc for txwi Felix Fietkau
2022-03-25 21:10 ` Felix Fietkau [this message]
2022-03-25 21:10 ` [PATCH 6/8] mt76: mt7915: accept rx frames with non-standard VHT MCS10-11 Felix Fietkau
2022-03-25 21:10 ` [PATCH 7/8] mt76: mt7921: " Felix Fietkau
2022-03-25 21:10 ` [PATCH 8/8] mt76: fix use-after-free by removing a non-RCU wcid pointer Felix Fietkau

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220325211031.24171-5-nbd@nbd.name \
    --to=nbd@nbd.name \
    --cc=linux-wireless@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox