Re: [PATCH net 3/4] xsk: drain continuation descs after overflow in xsk_build_skb()

[Stanislav Fomichev <sdf.kernel@gmail.com>]

On 05/10, Jason Xing wrote:
> From: Jason Xing <kernelxing@tencent.com>
> 
> When a multi-buffer packet exceeds MAX_SKB_FRAGS and triggers -EOVERFLOW,
> only the current descriptor is released from the TX ring. The remaining
> continuation descriptors of the same packet stay in the ring. Since
> xs->skb is set to NULL after the drop, the TX loop picks up these
> leftover frags and misinterprets each one as the beginning of a new
> packet, corrupting the packet stream.
> 
> Fix this by adding a drain_cont flag to xdp_sock. When overflow occurs
> and the dropped descriptor has XDP_PKT_CONTD set, the flag is raised.
> The main TX loop in __xsk_generic_xmit() then handles continuation
> descriptors one at a time: each gets a normal CQ reservation (with
> backpressure), its address is submitted to the completion queue, and
> the descriptor is released from the TX ring. When the last fragment
> (without XDP_PKT_CONTD) is processed, the flag is cleared and the
> function returns -EOVERFLOW so the next call starts with a fresh
> budget for normal packets.
> 
> This reuses the existing CQ backpressure and budget mechanisms, so if
> the CQ is full the function returns -EAGAIN and userspace drains the
> CQ before retrying. Zero buffer leakage, zero packet stream corruption.
> 
> Closes: https://lore.kernel.org/all/20260425041726.85FB3C2BCB2@smtp.kernel.org/
> Fixes: cf24f5a5feea ("xsk: add support for AF_XDP multi-buffer on Tx path")
> Signed-off-by: Jason Xing <kernelxing@tencent.com>
> ---
>  include/net/xdp_sock.h |  1 +
>  net/xdp/xsk.c          | 22 ++++++++++++++++++++++
>  2 files changed, 23 insertions(+)
> 
> diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
> index 23e8861e8b25..1958d19d9925 100644
> --- a/include/net/xdp_sock.h
> +++ b/include/net/xdp_sock.h
> @@ -80,6 +80,7 @@ struct xdp_sock {
>  	 * call of __xsk_generic_xmit().
>  	 */
>  	struct sk_buff *skb;
> +	bool drain_cont;
>  
>  	struct list_head map_list;
>  	/* Protects map_list */
> diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
> index 3f1e590c855d..232dd7126905 100644
> --- a/net/xdp/xsk.c
> +++ b/net/xdp/xsk.c
> @@ -936,6 +936,8 @@ static struct sk_buff *xsk_build_skb(struct xdp_sock *xs,
>  			xs->tx->invalid_descs++;
>  		}
>  		xskq_cons_release(xs->tx);
> +		if (xp_mb_desc(desc))
> +			xs->drain_cont = true;
>  	} else {
>  		/* Let application retry */
>  		xsk_cq_cancel_locked(xs->pool, 1);
> @@ -982,6 +984,26 @@ static int __xsk_generic_xmit(struct sock *sk)
>  			goto out;
>  		}
>  
> +		if (unlikely(xs->drain_cont)) {
> +			unsigned long flags;
> +			u32 idx;
> +

[..]

> +			spin_lock_irqsave(&xs->pool->cq_prod_lock, flags);
> +			idx = xskq_get_prod(xs->pool->cq);
> +			xskq_prod_write_addr(xs->pool->cq, idx, desc.addr);
> +			xskq_prod_submit_n(xs->pool->cq, 1);
> +			spin_unlock_irqrestore(&xs->pool->cq_prod_lock, flags);

Not sure I understand why you want this if you're still marking the desc
invalid.

> +
> +			xs->tx->invalid_descs++;
> +			xskq_cons_release(xs->tx);
> +			if (!xp_mb_desc(&desc)) {
> +				xs->drain_cont = false;
> +				err = -EOVERFLOW;
> +				goto out;
> +			}

I also don't understand why you want to return -EOVERFLOW again? Why not
(quietly) swallow these invalid xp_mb_desc from the previous packet and move
on?

Also, have you considered doing this drain in xsk_build_skb error path?
Is it more messy? Carrying this drain_cont across the calls seems a bit
too complicated (but idk, maybe the alternative drain in xsk_build_skb
is even messier).