Re: [PATCH net-next v6 02/12] net: airoha: Reserve RX headroom to avoid skb reallocation

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] DMA synchronization size mismatch in `airoha_qdma_rx_process()`.
--

commit bc83d88c9a8c6fa433da204ea6d6867b3be7fa46
Author: Lorenzo Bianconi <lorenzo@kernel.org>
net: airoha: Reserve RX headroom to avoid skb reallocation

This commit reserves headroom for received packets to avoid reallocation
when pushing protocol headers. It adjusts the buffer offset and decreases
the DMA length accordingly.

> diff --git a/drivers/net/ethernet/airoha/airoha_eth.c b/drivers/net/ethernet/airoha/airoha_eth.c
> index f71fb18197ecd..3fe2561c85f17 100644
> --- a/drivers/net/ethernet/airoha/airoha_eth.c
> +++ b/drivers/net/ethernet/airoha/airoha_eth.c
> @@ -543,9 +543,10 @@ static int airoha_qdma_fill_rx_queue(struct airoha_queue *q)
>  		q->queued++;
>  		nframes++;
>  
> +		offset += AIROHA_RX_HEADROOM;
>  		e->buf = page_address(page) + offset;
>  		e->dma_addr = page_pool_get_dma_addr(page) + offset;
> -		e->dma_len = SKB_WITH_OVERHEAD(q->buf_size);
> +		e->dma_len = SKB_WITH_OVERHEAD(AIROHA_RX_LEN(q->buf_size));

[ ... ]

> @@ -616,8 +617,9 @@ static int airoha_qdma_rx_process(struct airoha_queue *q, int budget)
>  
>  		page = virt_to_head_page(e->buf);
>  		len = FIELD_GET(QDMA_DESC_LEN_MASK, desc_ctrl);
> -		data_len = q->skb ? q->buf_size
> -				  : SKB_WITH_OVERHEAD(q->buf_size);
> +		data_len = q->skb
> +			   ? AIROHA_RX_LEN(q->buf_size)
> +			   : SKB_WITH_OVERHEAD(AIROHA_RX_LEN(q->buf_size));

Does the length passed to dma_sync_single_for_cpu() also need to be updated
to match the new DMA length?

Looking at the code just above this context in airoha_qdma_rx_process():

		dma_sync_single_for_cpu(eth->dev, e->dma_addr,
					SKB_WITH_OVERHEAD(q->buf_size), dir);

Since e->dma_addr is now shifted forward by AIROHA_RX_HEADROOM bytes during
allocation, syncing the original unreduced length causes the synchronization
region to extend past the intended DMA data buffer by AIROHA_RX_HEADROOM
bytes.

Could syncing this extra memory lead to unintended cache invalidation
boundaries if fragment sizes change?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260511-airoha-eth-multi-serdes-v6-0-c899462c4f75@kernel.org?part=2