Re: [PATCH net v4 1/3] octeon_ep: fix skb frags overflow in the RX path

[Simon Horman <horms@kernel.org>]

On Mon, Jul 06, 2026 at 11:02:06PM +0800, Maoyi Xie wrote:
> __octep_oq_process_rx() builds an skb for a multi-buffer packet by adding
> one fragment per buffer_size chunk:
> 
> 	data_len = buff_info->len - oq->max_single_buffer_size;
> 	while (data_len) {
> 		...
> 		skb_add_rx_frag(skb, shinfo->nr_frags, buff_info->page, 0,
> 				buff_info->len, buff_info->len);
> 		...
> 	}
> 
> buff_info->len comes from the device response header
> (be64_to_cpu(resp_hw->length)). Nothing bounds the fragment count against
> MAX_SKB_FRAGS. data_len can be close to 65535. buffer_size defaults to
> about 3776 on 4K pages, so a full packet yields about 18 fragments. That
> is one more than the default MAX_SKB_FRAGS of 17, so skb_add_rx_frag()
> writes past shinfo->frags[].
> 
> The fragment count is now checked before build_skb(). A packet that needs
> more fragments than the skb can hold is dropped. octep_oq_drop_rx()
> consumes its descriptors like the build_skb failure path. The same class
> was fixed in other RX paths, including commit 5ffcb7b890f6 ("net: atlantic:
> fix fragment overflow handling in RX path") and commit f0813bcd2d9d ("net:
> wwan: t7xx: fix potential skb->frags overflow in RX path").
> 
> Fixes: 37d79d059606 ("octeon_ep: add Tx/Rx processing and interrupt support")
> Co-developed-by: Kaixuan Li <kaixuan.li@ntu.edu.sg>
> Signed-off-by: Kaixuan Li <kaixuan.li@ntu.edu.sg>
> Signed-off-by: Maoyi Xie <maoyixie.tju@gmail.com>
> Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> ---
>  drivers/net/ethernet/marvell/octeon_ep/octep_rx.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/drivers/net/ethernet/marvell/octeon_ep/octep_rx.c b/drivers/net/ethernet/marvell/octeon_ep/octep_rx.c
> index e6ebc7e44a..bdbed58c7b 100644
> --- a/drivers/net/ethernet/marvell/octeon_ep/octep_rx.c
> +++ b/drivers/net/ethernet/marvell/octeon_ep/octep_rx.c
> @@ -453,6 +453,15 @@ static int __octep_oq_process_rx(struct octep_device *oct,
>  
>  		octep_oq_next_pkt(oq, buff_info, &read_idx, &desc_used);
>  
> +		if (buff_info->len > oq->max_single_buffer_size) {
> +			u16 data_len = buff_info->len - oq->max_single_buffer_size;
> +
> +			if (DIV_ROUND_UP(data_len, oq->buffer_size) > MAX_SKB_FRAGS) {
> +				octep_oq_drop_rx(oq, buff_info, &read_idx, &desc_used);
> +				continue;
> +			}
> +		}

The AI-generated review of this patch on sashiko.dev flags that
this cannot occur for 4K pages. I believe this can be
addressed by changing the type of data_len to u32.
I also believe that a similar problem exists in patch 3/3.

   "Does truncating the data_len calculation to u16 introduce a regression
    by bypassing the MAX_SKB_FRAGS check?

    If buff_info->len is unexpectedly large, the u16 truncation of data_len
    could wrap around to a small value. This would bypass the check,
    leaving unconsumed fragments in the ring that might be incorrectly
    parsed as new packet headers."

> +
>  		skb = build_skb((void *)resp_hw, PAGE_SIZE);
>  		if (!skb) {
>  			octep_oq_drop_rx(oq, buff_info,
> -- 
> 2.34.1
>