Re: [PATCH ipsec 1/1] xfrm: espintcp: wait for pending partial send before reuse

[Sabrina Dubroca <sd@queasysnail.net>]

2026-05-30, 13:24:55 +0800, Ren Wei wrote:
> From: Wyatt Feng <bronzed_45_vested@icloud.com>
> 
> espintcp keeps a single in-flight transmit in ctx->partial.  Before
> building a new sk_msg, espintcp_sendmsg() first tries to flush that
> state through espintcp_push_msgs().
> 
> For blocking callers, espintcp_push_msgs() may return success while the
> previous partial send is still pending.  espintcp_sendmsg() then
> reinitializes emsg->skmsg and reuses the same ctx->partial slot even
> though the old send has not completed yet.
> 
> Wait for the pending partial send to drain before reusing ctx->partial.
> Nonblocking callers keep the existing error handling, while blocking
> callers sleep until the partial send can make progress.
> 
> This preserves the one-message-at-a-time design already used by the
> espintcp transmit path and avoids overwriting an in-flight partial send.
> 
> Fixes: e27cca96cd68 ("xfrm: add espintcp (RFC 8229)")
> Cc: stable@kernel.org
> Reported-by: Yuan Tan <yuantan098@gmail.com>
> Reported-by: Yifan Wu <yifanwucs@gmail.com>
> Reported-by: Juefei Pu <tomapufckgml@gmail.com>
> Reported-by: Zhengchuan Liang <zcliangcn@gmail.com>
> Reported-by: Xin Liu <bird@lzu.edu.cn>
> Assisted-by: Codex:GPT-5.4
> Signed-off-by: Wyatt Feng <bronzed_45_vested@icloud.com>
> Signed-off-by: Ren Wei <n05ec@lzu.edu.cn>
> ---
>  net/xfrm/espintcp.c | 19 ++++++++++++++-----
>  1 file changed, 14 insertions(+), 5 deletions(-)
> 
> diff --git a/net/xfrm/espintcp.c b/net/xfrm/espintcp.c
> index a2756186e13a..f8513a19a3c8 100644
> --- a/net/xfrm/espintcp.c
> +++ b/net/xfrm/espintcp.c
> @@ -340,11 +340,20 @@ static int espintcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
>  
>  	lock_sock(sk);
>  
> -	err = espintcp_push_msgs(sk, msg->msg_flags & MSG_DONTWAIT);
> -	if (err < 0) {
> -		if (err != -EAGAIN || !(msg->msg_flags & MSG_DONTWAIT))
> -			err = -ENOBUFS;
> -		goto unlock;
> +	while (emsg->len) {
> +		err = espintcp_push_msgs(sk, msg->msg_flags & MSG_DONTWAIT);
> +		if (err < 0) {
> +			if (err != -EAGAIN || !(msg->msg_flags & MSG_DONTWAIT))
> +				err = -ENOBUFS;
> +			goto unlock;
> +		}
> +
> +		if (!emsg->len)
> +			break;
> +
> +		err = sk_stream_wait_memory(sk, &timeo);

Is this really useful? espintcp_push_msgs ends up in
tcp_sendmsg_locked, which already has a retry loop that calls
sk_stream_wait_memory. So we just need better handling of in-progress
transfer on return of espintcp_push_msgs (something like
espintcp_push_skb does)?

-- 
Sabrina