Re: [PATCH net-next v3 0/2] tcp: keep the route window metric when SO_RCVBUF is set before connect

[shashank Jain]

Yes, I read it, and I understand it: the missing information should go in
a reply, not a repost, and if the series needs another revision for other
reasons, it belongs in the commit messages. v3 was posted because the fix
itself changed (the AI review of v2 found that the socket-state check
missed the TCP_SYN_SENT paths), but I left the "how it was found / how it
was tested" part only in my reply to the bot and not in the changelog.
That was my mistake. v4 will have it in the commit message of patch 1.

> Looks like AI reviewer has some comments that need addressing.

Yes. The v3 review found that the SYN cookie paths (cookie_v4_check() and
cookie_v6_check()) still use the old rule, so the window scale of the
cookie SYN-ACK and of the child can disagree; that a single bit can leave
window_clamp below the requested value when SO_RCVBUF is raised later; and
that the bit is read on the listener without READ_ONCE(). I will fix those,
add the listener / SYN-ACK and cookie cases to the selftests, and post v4
as a new thread with a link to v3.

Thanks,
Shashank

On Tue, Oct 20, 2026 at 7:43 AM Jakub Kicinski <kuba@xxxxxxxxxx> wrote:
>
> On Mon, 12 Oct 2026 19:04:27 +0530 shashank Jain wrote:
> > > Please do not repost the series just to address the above. Instead,
> > > reply to this email with the missing information, so that reviewers
> > > can take it into account. If the series needs another revision for
> > > other reasons, please include the information in the commit messages
> > > then.
>
> Could you please confirm that you read and understood this paragraph.
>
> Looks like AI reviewer has some comments that need addressing.
> --
> pw-bot: cr