@@ -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);
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(-)