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

[Lukasz Majewski <lukma at denx dot de>]

Hi Arnd,

> On Thu, Jun 27, 2019 at 12:36 PM Lukasz Majewski <lukma@denx.de>
> wrote:
> > > On Thu, Jun 27, 2019 at 12:55 AM Lukasz Majewski <lukma@denx.de>  
> 
> > After the discussion I do believe that it would be correct to change
> > the proposed patch [1] '__clock_settime' to:
> >
> > #if __TIMESIZE != 64  
> 
> This would hide the function on 64-bit architectures, right?

Yes, this function would be hidden on architectures with __WORDSIZE==64
(e.g. x86_64) (+ x32 ).

> 
> Maybe
> #if (__TIMESIZE == __WORDSIZE) || (defined __x86_64__ && defined
> __ILP32__)
> 
> or shortening that by adding a new macro
> 
> #if __ASSUME_TIME32_SYSCALLS

Please see below comment.

> 
> > int
> > __clock_settime (clockid_t clock_id, const struct timespec *tp)
> > {
> > /* For archs with WORDSIZE==32, which do not support clock_settime64
> > the clock_settime supporting 32 bit time ABI will be used */  
> 
> Same for the comment: __clock_settime would be used both
> on 64-bit architectures as the normal path, and for compatiblity
> with applications built without _TIME_BITS=64.

With the patch [1] - it would be used in such a way:

1. For archs with WORDSIZE==32 and __TIMESIZE==32 (legacy ones - e.g.
arm 32 bits)

#if __TIMESIZE != 64
int
__clock_settime (clockid_t clock_id, const struct timespec *tp)
{
	return clock_settime(....)
}
#endif


2. For arch with WORDSIZE==64 (and WORDSIZE_SYSCALL=64 -> x32) and
__TIMESIZE==64

__clock_settime64 (clockid_t clock_id, const struct __timespec64 *tp)
{
#ifdef __ASSUME_TIME64_SYSCALLS
# ifndef __NR_clock_settime64
#  define __NR_clock_settime64 __NR_clock_settime
# endif
  return INLINE_SYSCALL_CALL (clock_settime64, clock_id, tp);
#else
....


}

The __ASSUME_TIME64_SYSCALLS is always defined in this case [2]

Here the assumption is that for such architectures (like aarch64 or
x86_64) the __NR_clock_settime64 (if ever introduced) has the same ABI
as __NR_clock_settime, so there will be no need to adjust the glibc code.


3. New archs added or ones being Y2038 safe (with __NR_clock_settime64
available in the kernel -> starting from 5.1+).

For them the __ASSUME_TIME64_SYSCALLS is defined [2] and execution
patch from point 2. is used with potential fallback to 32 bit version
of clock_settime:

__clock_settime64 (clockid_t clock_id, const struct __timespec64 *tp)
{
#ifdef __ASSUME_TIME64_SYSCALLS
# ifndef __NR_clock_settime64
#  define __NR_clock_settime64 __NR_clock_settime
# endif
  return INLINE_SYSCALL_CALL (clock_settime64, clock_id, tp);
#else
# ifdef __NR_clock_settime64
  int ret = INLINE_SYSCALL_CALL (clock_settime64, clock_id, tp);
  if (ret == 0 || errno != ENOSYS)
    return ret;
# endif
  if (! in_time_t_range (tp->tv_sec))
    {
      __set_errno (EOVERFLOW);
      return -1;
    }

  struct timespec ts32;
  valid_timespec64_to_timespec (tp, &ts32);
  return INLINE_SYSCALL_CALL (clock_settime, clock_id, &ts32);
#endif
}

Note for the above code:

The __NR_clock_settime64 is used whenever possible and if not available
the fallback to 32 bit time supporting clock_settime is executed (with
overflow check).


> 
> > return INLINE_SYSCALL_CALL (clock_settime, clock_id, tp);
> > }
> > #endif
> >
> > The "__clock_settime64()" part could be left untouched.  
> 
> Ok.
> 
>       Arnd

Note:

[1] - https://patchwork.ozlabs.org/patch/1107235/
[2] - https://patchwork.ozlabs.org/patch/1117100/


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

Attachment: pgpHjpE2Zswd6.pgp
Description: OpenPGP digital signature