Re: [PATCH] powerpc/vdso: fix clock_gettime_fallback for vdso32

[Christophe Leroy <christophe.leroy@csgroup.eu>]

Le 11/01/2021 à 16:18, Andreas Schwab a écrit :
> The second argument of __kernel_clock_gettime64 points to a struct
> __kernel_timespec, with 64-bit time_t, so use the clock_gettime64 syscall
> in the fallback function for the 32-bit vdso.  Similarily,
> clock_getres_fallback should use the clock_getres_time64 syscall, though
> it isn't yet called from the 32-bit vdso.
> 
> Signed-off-by: Andreas Schwab <schwab@linux-m68k.org>

Fixes: d0e3fc69d00d ("powerpc/vdso: Provide __kernel_clock_gettime64() on vdso32")

> ---
>   arch/powerpc/include/asm/vdso/gettimeofday.h | 8 ++++++++
>   1 file changed, 8 insertions(+)
> 
> diff --git a/arch/powerpc/include/asm/vdso/gettimeofday.h b/arch/powerpc/include/asm/vdso/gettimeofday.h
> index 81671aa365b3..a26c12df2c9a 100644
> --- a/arch/powerpc/include/asm/vdso/gettimeofday.h
> +++ b/arch/powerpc/include/asm/vdso/gettimeofday.h
> @@ -106,13 +106,21 @@ int gettimeofday_fallback(struct __kernel_old_timeval *_tv, struct timezone *_tz
>   static __always_inline
>   int clock_gettime_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
>   {
> +#ifndef __powerpc64__

I think we should do like x86, create a second version of clock_gettime_fallback() inside the #ifdef 
CONFIG_VDSO32, and move that one in a matching (non existing yet) #else

> +	return do_syscall_2(__NR_clock_gettime64, _clkid, (unsigned long)_ts);
> +#else
>   	return do_syscall_2(__NR_clock_gettime, _clkid, (unsigned long)_ts);
> +#endif
>   }
>   
>   static __always_inline
>   int clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
>   {
> +#ifndef __powerpc64__

Same.

> +	return do_syscall_2(__NR_clock_getres_time64, _clkid, (unsigned long)_ts);
> +#else
>   	return do_syscall_2(__NR_clock_getres, _clkid, (unsigned long)_ts);
> +#endif
>   }
>   
>   #ifdef CONFIG_VDSO32
> 

Christophe