[LTP] [PATCH 3/5] syscalls: Don't use tst_syscall() unnecessarily

[Arnd Bergmann]

On Thu, May 21, 2020 at 4:20 PM Cyril Hrubis <chrubis@suse.cz> wrote:
> > > I guess that at the current time we do support distros that are at max
> > > 10 years old, mostly because enterprise support cycles are about 10
> > > years in lenght.
> > >
> > > The issue here is that glibc needed -lrt passed to linker couple of
> > > years ago and we wanted to avoid the need of linking everything with
> > > -lrt, as calling the raw syscall was just easier fix.
> >
> > To conclude the discussion, is this patch okay or not ? The reason why I am
> > asking this is because this file still uses the old test framework and so can't
> > include tst_timer.h, and so doesn't have access to struct __kernel_old_timespec.
> >
> > What do you suggest I do here ?
>
> Well if it was only about __kernel_old_timespec we could easily just
> declare it locally in the file and be done with it.
>
> But I guess that newer 32bit architectures will have only the 64 bit
> syscall present, I think that somebody pointed out that this is the case
> for 32bit RiscV. If that's true we will have to have a fallback for that
> case as well.

Yes, that is correct. Also, the 32-bit time_t syscalls can be disabled
on newer kernels when you have a libc implementation that
uses the 64-bit calls exclusively.

I guess the fallback to to libc gettimeofday() isn't really needed
unless it needs to run on ancient kernels.

      Arnd