Re: [PATCH mptcp-next v5 0/5] Squash to "Add mptcp_subflow bpf_iter support"

[Matthieu Baerts <matttbe@kernel.org>]

Hi Geliang, Mat,

On 03/03/2025 11:33, Geliang Tang wrote:
> From: Geliang Tang <tanggeliang@kylinos.cn>
> 
> v5:
>  - check bpf_iter_task in mptcp_subflow_new() as Mat suggested.

Thank you for the modification, but would it not be easier not to do
that in this series, not to increase the complexity?

This can be done in the "use bpf_iter in bpf schedulers" instead, no?

For me, the v4 (with my squash-to patch) is enough to be used with CG
[gs]etsocktopt.

Also, I think the idea is set this marker before calling the future
struct_ops, in places where we know the msk is owned, but not in
bpf_mptcp_sock_from_sock() which doesn't check if the msk is owned, no?

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