On Wed, Aug 12, 2020 at 11:17:29AM +0900, YASUOKA Masahiko wrote:
> On Tue, 11 Aug 2020 23:06:45 +0300
> Vitaliy Makkoveev <m...@openbsd.org> wrote:
> > We removed `pipex{in,out}q'. So now we can destroy pppac(4) session just
> > like we do in pppx(4) case. Also there is no reason to allow
> > pipex_timer() to destroy sessions - userland will do this by
> > PIPEXDSESSION. This permit us to use existing pipex_get_closed() for
> > both pppac(4) and pppx(4) without any modifications.
> >
> > So, I propose pipex_close_session() and pipex_timer() be like below.
>
> It doesn't seem to fix "idle-timeout".
>
Yes it's not. It's "pre-" step which makes following fix easier.
We don't need to mark pppx(4) sessions because there is no special cases
for them. We just need to kill pppx(4) related "pr_timeout_sec != 0"
checks and call pipex_get_closed() by pppx_get_closed().
> > We simplify pppac(4) session destruction. We unify behavior with pppx(4)
> > - we killing session just now. There is no reason to modify
> > pipex_get_closed() and pipex_link_session(). pppx(4) related sessions
> > can be processed by pipex_timer(). There is no performance impact.
>
> We need to modify pppx_get_closed() to implement idle-timeout.
>
> > Do you like this? We can do two diffs. The first to unify destruction
> > and the second to re-enable in-kernel timeout for pppx(4) and revert man
> > pages modifications.
>
> I have no objection to your "unify destruction".
>
> I'll rebase my diff after that work.
Thanks. I posted "unify destruction" here [1].
1. https://marc.info/?l=openbsd-tech&m=159722447900893&w=2
