Re: [RFC v2 03/20] y2038: linux: Provide __clock_settime64 implementation

[Alistair Francis <alistair23 at gmail dot com>]

On Thu, Jun 27, 2019 at 3:08 PM Lukasz Majewski <lukma@denx.de> wrote:
>
> Hi Arnd,
>
> > On Thu, Jun 27, 2019 at 6:17 PM Lukasz Majewski <lukma@denx.de> wrote:
> >
> > > >
> > > > Ok, so you redirect both __clock_settime64 to __clock_settime
> > > > and __NR_settime64 to __NR_settime, in order to get back to the
> > > > same definition that you have today on 64-bit.
> > > >
> > > > Sorry for taking so long to understand that.
> > >
> > > Not only you had a hard time to understand it.
> >
> > Here is how I had imagined it would be done in a way that I find
> > easier to understand:
> >
> > #if __WORDSIZE == 32 /* yes, all of them including rv32 and x32 */
> > __clock_settime64 (clockid_t clock_id, const struct __timespec64 *tp)
> > {
> >    int ret;
> >
> >    /* Make sure the time cvalue is OK.  */
> >    if (tp->tv_nsec < 0 || tp->tv_nsec >= 1000000000)
> >     {
> >        __set_errno (EINVAL);
> >        return -1;
> >      }
> > #ifdef __NR_clock_settime64
> >  ret = INLINE_SYSCALL_CALL (clock_settime64, clock_id, tp);
> > #endif
> >
> > #ifndef __ASSUME_TIME64_SYSCALLS
> >   if (ret == 0 || errno != ENOSYS)
> >     return ret;
> >
> >   if (! in_time_t_range (tp->tv_sec))
> >     {
> >       __set_errno (EOVERFLOW);
> >       return -1;
> >     }
> >
> >   struct timespec ts32;
> >   valid_timespec64_to_timespec (tp, &ts32);
> >   ret = INLINE_SYSCALL_CALL (clock_settime, clock_id, &ts32);
> > #endif
> >
> >   return ret;
> > }
> > #endif
> >
> > This should be functionally identical to what you have, but avoid
> > some of the complexity, especially once linux-5.1 becomes the
> > minimum version.

This implementation looks good to me. It shouldn't be too hard to
extend this to all the problematic y2038 syscalls.

>
> The semantics for __ASSUME_TIME64_SYSCALLS is under the discussion for
> some time [1].

I don't see any comments on this, do you know the status?

>
> Also the __clock_settime64() implementation from [2] takes into
> account some corner cases (like support for all archs since 3.2
> kernel, x32, legacy systems, etc). The implementation from [2] IMHO
> seems to be more concise and fits into the 64 bit conversion paradigm
> for glibc [3].

This one also works for RV32 :)

Alistair

>
> However, this is only my opinion and we shall wait for other community
> members to express their opinions.
>
> >
> >       Arnd
>
>
> Note:
>
> [1] - https://patchwork.ozlabs.org/patch/1117100/
> [2] - https://patchwork.ozlabs.org/patch/1107235/
> [3] -
> https://www.gnu.org/software/libc/manual/html_mono/libc.html
> "D.2.1 64-bit time symbol handling in the GNU C Library"
>
>
>
> Best regards,
>
> Lukasz Majewski
>
> --
>
> DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
> HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
> Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma@denx.de