msm8916-openwrt/target/linux/generic/backport-6.1/729-16-v6.3-net-ethernet-mtk_wed-get-rid-of-queue-lock-for-rx-qu.patch
Christian Marangi fa79baf4a6
generic: copy backport, hack, pending patch and config from 5.15 to 6.1
Copy backport, hack, pending patch and config from 5.15 to 6.1.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
2023-05-22 22:28:13 +02:00

53 lines
1.7 KiB
Diff

From: Lorenzo Bianconi <lorenzo@kernel.org>
Date: Tue, 10 Jan 2023 10:31:26 +0100
Subject: [PATCH] net: ethernet: mtk_wed: get rid of queue lock for rx queue
Queue spinlock is currently held in mtk_wed_wo_queue_rx_clean and
mtk_wed_wo_queue_refill routines for MTK Wireless Ethernet Dispatcher
MCU rx queue. mtk_wed_wo_queue_refill() is running during initialization
and in rx tasklet while mtk_wed_wo_queue_rx_clean() is running in
mtk_wed_wo_hw_deinit() during hw de-init phase after rx tasklet has been
disabled. Since mtk_wed_wo_queue_rx_clean and mtk_wed_wo_queue_refill
routines can't run concurrently get rid of spinlock for mcu rx queue.
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Link: https://lore.kernel.org/r/36ec3b729542ea60898471d890796f745479ba32.1673342990.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
--- a/drivers/net/ethernet/mediatek/mtk_wed_wo.c
+++ b/drivers/net/ethernet/mediatek/mtk_wed_wo.c
@@ -138,7 +138,6 @@ mtk_wed_wo_queue_refill(struct mtk_wed_w
enum dma_data_direction dir = rx ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
int n_buf = 0;
- spin_lock_bh(&q->lock);
while (q->queued < q->n_desc) {
struct mtk_wed_wo_queue_entry *entry;
dma_addr_t addr;
@@ -172,7 +171,6 @@ mtk_wed_wo_queue_refill(struct mtk_wed_w
q->queued++;
n_buf++;
}
- spin_unlock_bh(&q->lock);
return n_buf;
}
@@ -316,7 +314,6 @@ mtk_wed_wo_queue_rx_clean(struct mtk_wed
{
struct page *page;
- spin_lock_bh(&q->lock);
for (;;) {
void *buf = mtk_wed_wo_dequeue(wo, q, NULL, true);
@@ -325,7 +322,6 @@ mtk_wed_wo_queue_rx_clean(struct mtk_wed
skb_free_frag(buf);
}
- spin_unlock_bh(&q->lock);
if (!q->cache.va)
return;