Re: [PATCH v3 1/1] pipe: only enable the extra wake_up(rd_wait) for edge-triggered consumers

Mateusz, Breno,

thanks, I'll try to think about it later, but...

On 07/30, Mateusz Guzik wrote:
>
> On Thu, Jul 30, 2026 at 4:38 PM Oleg Nesterov <[email protected]> wrote:
> > What exactly do you think the comment should say? I agree with anything in
> > advance. I thought that "edge-triggered" provides enough info, but I would
> > be happy to improve the docs.
> >
>
> how about: There is userspace depending on the extra wake up, see
> commit 3a34b13a88caeb28 ("pipe: make pipe writes always wake up
> readers") for details.

To me this looks confusing.

IMO, the comment like this (with the reference to the commit) would make
sense to document the unconditional/undocumented kill_fasync(fasync_readers)
in anon_pipe_write(), this SIGIO is even worse in some sense and I would like
to discuss it another time ;)

But as for poll_usage/poll_et... We have the established API, and (afaics) it
works as documented. It doesn't matter if EPOLLET behaviour is good or bad.
We only need to document what ->poll_et means for pipes.

> > And the helper's name? pipe_enable_poll_et() ?
>
> perhaps pipe_enable_epoll_semantics()?

Again, contrary to the current comments this is not Epoll-only...

Oleg.