Re: [Y2038] [PATCH v3 14/22] time:Introduce the do_sys_settimeofday64() function with timespec64 type

[Arnd Bergmann <arnd@arndb.de>]

On Wednesday 13 May 2015 10:43:25 Baolin Wang wrote:
> > > -int do_sys_settimeofday(const struct timespec *tv, const struct
> > timezone *tz)
> > > +int do_sys_settimeofday64(const struct timespec64 *tv, const struct
> > timezone *tz)
> > >  {
> > >       static int firsttime = 1;
> > >       int error = 0;
> > > +     struct timespec ts;
> > >
> > > -     if (tv && !timespec_valid(tv))
> > > +     if (tv && !timespec64_valid(tv))
> > >               return -EINVAL;
> > >
> > > -     error = security_settime(tv, tz);
> > > +     ts = timespec64_to_timespec(*tv);
> > > +     error = security_settime(&ts, tz);
> >
> > How is that 2038 safe? Not at all.
> >
> > I told you before that we fix the underlying functions before we add
> > such bogosities.
> >
> >
> Thanks for your comments, and i'll continue to repair my patch.
> But could you explain what the defects in do_sys_settimeofday64
> function? Thanks.

What Thomas meant here is that you should first fix security_settime
to take a timespec64 argument. I guess you could introduce a
security_settime64() function first, using another inline wrapper
to provide the existing security_settime() API. security_settime()
does not actually use the time argument, so this is a fairly easy
thing to do.

	Arnd