[PATCH net v2] net: airoha: Fix page recycling in airoha_qdma_rx_process()

[PATCH net v2] net: airoha: Fix page recycling in airoha_qdma_rx_process()

[Lorenzo Bianconi]

Do not recycle the page twice in airoha_qdma_rx_process routine in case
of error. Just run dev_kfree_skb() if the skb has been allocated and marked
for recycling. Run page_pool_put_full_page() directly if the skb has not
been allocated yet.
Moreover, rely on DMA address from queue entry element instead of reading
it from the DMA descriptor for DMA syncing in airoha_qdma_rx_process().

Fixes: e12182ddb6e71 ("net: airoha: Enable Rx Scatter-Gather")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
---
Changes in v2:
- Add missing signed-off-by
- Link to v1: https://lore.kernel.org/r/20250513-airoha-fix-rx-process-error-condition-v1-1-e5d70cd7c066@kernel.org
---
 drivers/net/ethernet/airoha/airoha_eth.c | 22 +++++++++-------------
 1 file changed, 9 insertions(+), 13 deletions(-)

diff --git a/drivers/net/ethernet/airoha/airoha_eth.c b/drivers/net/ethernet/airoha/airoha_eth.c
index d748dc6de92367365db9f9548f9af52a7fdac187..1e9ab65218ff144d99b47f5d4ad5ff4f9c227418 100644
--- a/drivers/net/ethernet/airoha/airoha_eth.c
+++ b/drivers/net/ethernet/airoha/airoha_eth.c
@@ -614,7 +614,6 @@ static int airoha_qdma_rx_process(struct airoha_queue *q, int budget)
 		struct airoha_queue_entry *e = &q->entry[q->tail];
 		struct airoha_qdma_desc *desc = &q->desc[q->tail];
 		u32 hash, reason, msg1 = le32_to_cpu(desc->msg1);
-		dma_addr_t dma_addr = le32_to_cpu(desc->addr);
 		struct page *page = virt_to_head_page(e->buf);
 		u32 desc_ctrl = le32_to_cpu(desc->ctrl);
 		struct airoha_gdm_port *port;
@@ -623,22 +622,16 @@ static int airoha_qdma_rx_process(struct airoha_queue *q, int budget)
 		if (!(desc_ctrl & QDMA_DESC_DONE_MASK))
 			break;
 
-		if (!dma_addr)
-			break;
-
-		len = FIELD_GET(QDMA_DESC_LEN_MASK, desc_ctrl);
-		if (!len)
-			break;
-
 		q->tail = (q->tail + 1) % q->ndesc;
 		q->queued--;
 
-		dma_sync_single_for_cpu(eth->dev, dma_addr,
+		dma_sync_single_for_cpu(eth->dev, e->dma_addr,
 					SKB_WITH_OVERHEAD(q->buf_size), dir);
 
+		len = FIELD_GET(QDMA_DESC_LEN_MASK, desc_ctrl);
 		data_len = q->skb ? q->buf_size
 				  : SKB_WITH_OVERHEAD(q->buf_size);
-		if (data_len < len)
+		if (!len || data_len < len)
 			goto free_frag;
 
 		p = airoha_qdma_get_gdm_port(eth, desc);
@@ -701,9 +694,12 @@ static int airoha_qdma_rx_process(struct airoha_queue *q, int budget)
 		q->skb = NULL;
 		continue;
 free_frag:
-		page_pool_put_full_page(q->page_pool, page, true);
-		dev_kfree_skb(q->skb);
-		q->skb = NULL;
+		if (q->skb) {
+			dev_kfree_skb(q->skb);
+			q->skb = NULL;
+		} else {
+			page_pool_put_full_page(q->page_pool, page, true);
+		}
 	}
 	airoha_qdma_fill_rx_queue(q);
 

---
base-commit: 09db7a4d287d1a2bcfc04df023c103d1213a0518
change-id: 20250513-airoha-fix-rx-process-error-condition-b6e5db52fd4f

Best regards,
-- 
Lorenzo Bianconi <lorenzo@kernel.org>

Re: [PATCH net v2] net: airoha: Fix page recycling in airoha_qdma_rx_process()

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Thu, 15 May 2025 08:33:06 +0200 you wrote:
> Do not recycle the page twice in airoha_qdma_rx_process routine in case
> of error. Just run dev_kfree_skb() if the skb has been allocated and marked
> for recycling. Run page_pool_put_full_page() directly if the skb has not
> been allocated yet.
> Moreover, rely on DMA address from queue entry element instead of reading
> it from the DMA descriptor for DMA syncing in airoha_qdma_rx_process().
> 
> [...]

Here is the summary with links:
  - [net,v2] net: airoha: Fix page recycling in airoha_qdma_rx_process()
    https://git.kernel.org/netdev/net/c/d6d2b0e1538d

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html