Re: [PATCH v3 1/2] epoll: add nsec timeout support with epoll_pwait2

[Willem de Bruijn <willemdebruijn.kernel@gmail.com>]

On Fri, Nov 20, 2020 at 2:23 PM Arnd Bergmann <arnd@kernel.org> wrote:
>
> On Fri, Nov 20, 2020 at 5:01 PM Willem de Bruijn
> <willemdebruijn.kernel@gmail.com> wrote:
> >
> > On Fri, Nov 20, 2020 at 3:13 AM Arnd Bergmann <arnd@kernel.org> wrote:
> > >
> > > On Thu, Nov 19, 2020 at 9:13 PM Willem de Bruijn
> > > <willemdebruijn.kernel@gmail.com> wrote:
> > > > On Thu, Nov 19, 2020 at 10:45 AM Arnd Bergmann <arnd@kernel.org> wrote:
>
> > Thanks for the suggestion.
> >
> > I do have an initial patchset. As expected, it does involve quite a
> > bit of code churn to pass slack through the callers. I'll take a look
> > at your suggestion to simplify it.
> >
> > As is, the patchset is not ready to send to the list for possible
> > merge. In the meantime, I did push the patchset to github at
> > https://github.com/wdebruij/linux/commits/epoll-nstimeo-1 . I can send
> > a version marked RFC to the list if that's easier.
>
> Looks all good to me, just two small things I noticed that you can
> address before sending the new series:
>
> * The div_u64_rem() in ep_timeout_to_timespec() looks wrong, as
>   you are actually dividing a 'long' that does not need it.
>
> * In "epoll: wire up syscall epoll_pwait2", the alpha syscall has the
> wrong number, it
>    should be 110 higher than the others, not 109.

Thanks! I'll fix these up.

> > Btw, the other change, to convert epoll implementation to timespec64
> > before adding the syscall, equally adds some code churn compared to
> > patch v3. But perhaps the end state is cleaner and more consistent.
>
> Right, that's what I meant. If it causes too much churn, don't worry
> about it it.

I think it'll be better to split the patchsets:

epoll: convert internal api to timespec64
epoll: add syscall epoll_pwait2
epoll: wire up syscall epoll_pwait2
selftests/filesystems: expand epoll with epoll_pwait2

and

select: compute slack based on relative time
epoll: compute slack based on relative time

and judge the slack conversion on its own merit.

I also would rather not tie this up with the compat deduplication.
Happy to take a stab at that though. On that note, when combining
functions like

  int core_sys_select(int n, fd_set __user *inp, fd_set __user *outp,
                           fd_set __user *exp, struct timespec64 *end_time,
                           u64 slack)

and

  static int compat_core_sys_select(int n, compat_ulong_t __user *inp,
        compat_ulong_t __user *outp, compat_ulong_t __user *exp,
        struct timespec64 *end_time, u64 slack)

by branching on in_compat_syscall() inside get_fd_set/set_fd_set and
deprecating their compat_.. counterparts, what would the argument
pointers look like? Or is that not the approach you have in mind?