[PATCH net v1] net: stmmac: set PP_FLAG_DMA_SYNC_DEV only if XDP is enabled

Furong Xu posted 1 patch 2 months, 2 weeks ago
There is a newer version of this series
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
[PATCH net v1] net: stmmac: set PP_FLAG_DMA_SYNC_DEV only if XDP is enabled
Posted by Furong Xu 2 months, 2 weeks ago
When XDP is not enabled, the page which holds the received buffer
will be recycled once the buffer is copied into SKB by
skb_copy_to_linear_data(), then the MAC core will never reuse this
page any longer. Set PP_FLAG_DMA_SYNC_DEV wastes CPU cycles.

This patch brings up to 9% noticeable performance improvement on
certain platforms.

Fixes: 5fabb01207a2 ("net: stmmac: Add initial XDP support")
Signed-off-by: Furong Xu <0x1207@gmail.com>
---
 drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index f3a1b179aaea..95d3d1081727 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -2022,7 +2022,7 @@ static int __alloc_dma_rx_desc_resources(struct stmmac_priv *priv,
 	rx_q->queue_index = queue;
 	rx_q->priv_data = priv;
 
-	pp_params.flags = PP_FLAG_DMA_MAP | PP_FLAG_DMA_SYNC_DEV;
+	pp_params.flags = PP_FLAG_DMA_MAP | (xdp_prog ? PP_FLAG_DMA_SYNC_DEV : 0);
 	pp_params.pool_size = dma_conf->dma_rx_size;
 	num_pages = DIV_ROUND_UP(dma_conf->dma_buf_sz, PAGE_SIZE);
 	pp_params.order = ilog2(num_pages);
-- 
2.34.1
Re: [PATCH net v1] net: stmmac: set PP_FLAG_DMA_SYNC_DEV only if XDP is enabled
Posted by Paolo Abeni 2 months, 1 week ago
On 9/13/24 13:02, Furong Xu wrote:
> When XDP is not enabled, the page which holds the received buffer
> will be recycled once the buffer is copied into SKB by
> skb_copy_to_linear_data(), then the MAC core will never reuse this
> page any longer. Set PP_FLAG_DMA_SYNC_DEV wastes CPU cycles.
> 
> This patch brings up to 9% noticeable performance improvement on
> certain platforms.
> 
> Fixes: 5fabb01207a2 ("net: stmmac: Add initial XDP support")
> Signed-off-by: Furong Xu <0x1207@gmail.com>

I'm quite unconvinced that every performance improvement would be 
eligible to be considered a fix.

Reading the code it looks like this change actually addresses a 
regression introduced by the blamed commit, is that correct? If so 
please re-phrase the commit message accordingly.

Thanks,

Paolo