On Thu, Sep 4, 2025 at 12:27 AM Jakub Kicinski [off-list ref] wrote:

On Thu, 4 Sep 2025 00:17:34 +0100 Dmitry Safonov wrote:

[..]

quoted

Looking at the code now, I guess what I could have done even more is
migrating tcp_sock::ao_info (and tcp_timewait_sock::ao_info) from
rcu_*() helpers to acquire/release ones. Somewhat feeling uneasy about
going that far just yet. Should I do it with another cleanup on the
top, what do you think?

No preference :)

Jakub,

I've sent v5, addressing minor v4 review comments.
I have a patch that migrates ao_info from rcu_*() to
smp_{store,release}*(), it seems to work. I'm going to send it as a
follow-up, once this gets into net-next as these two seem to be mostly
reviewed/ready.

While on it, I noticed that potentially I could trim (struct
tcp_timewait_sock) by 8 bytes if tw_md5_key and ao_info would union in
another helper structure. As on time-wait socket only one of MD5 or AO
could be set.

Planning to send these two patches in a separate thread/patches set.

Thanks,
            Dmitry