Re: [PATCH v2] tcp: fastopen: check rsk_drop_req() in tcp_fastopen_create_child()

["Yilin Zhang" <yilinzhang@moonshot.ai>]

Hi Jiayuan, Paolo, Matt,

On 11/09/2026, Jiayuan Chen wrote:
> Should we accept this subflow ?
>
> It's just a SYN with MPJ + valid fastopen cookie, replying SYNACK
> and fallback to 3-way handshake may be easier.
> (I'm not sure whether RFC define it or not.)

This looks good to me. Falling back like this is also standard TFO
behavior (RFC 7413, sec. 4.2.2), so IIUC the RFCs don't object.

On 10/09/2026, Paolo Abeni wrote:
> If so mptcp should reject entirely MPJ + fastopen and no addtional
> code required on the TCP side.

With Jiayuan's SYN-flag check in subflow_syn_recv_sock(), we can do
exactly that: return NULL there, and tcp_conn_request() falls back to
the regular MP_JOIN handshake. No TCP-side changes needed.

> A minor process note: the changelog should come after the tag area
> and a '---' separator.

Noted, thanks for the reminder. I'll draft a v3 shortly along these
lines.

Thanks,
Yilin