Re: [RESEND PATCH mptcp-next v2 2/3] mptcp: implement .splice_eof

[Matthieu Baerts <matttbe@kernel.org>]

Hi Geliang,

On 17/03/2026 10:41, Geliang Tang wrote:
> From: Geliang Tang <tanggeliang@kylinos.cn>
> 
> This patch implements the .splice_eof interface for MPTCP, namely
> mptcp_splice_eof(), which calls do_tcp_splice_eof() for each active
> subflow when a sendfile() operation reaches end-of-file.
> 
> do_tcp_splice_eof() flushes any remaining data in the TCP send queue.
> MPTCP operates over multiple TCP subflows, and each subflow may have
> pending data in its send buffer that needs to be properly finalized
> when splicing data through an MPTCP socket.
> 
> sock_splice_eof() calls the .splice_eof interface from struct proto_ops.
> To maintain consistency with regular TCP behavior, the .splice_eof
> interface of mptcp_stream_ops is set to inet_splice_eof, which will
> switch to the protocol-specific implementation (sk->sk_prot->splice_eof)
> - for MPTCP, that is mptcp_splice_eof().
> 
> This is an improvement; nothing was broken before. MPTCP previously did not
> handle the splice EOF notification, while TCP did. Without .splice_eof()
> support, the queue is not flushed immediately when sendfile() reaches EOF,
> but it will eventually be flushed after a small delay.

Thank you for this patch!

Do you mind checking the comments from sashiko please?

https://sashiko.dev/#/patchset/cover.1773740361.git.tanggeliang%40kylinos.cn

They look plausible, but better to double-check.

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.