Re: [PATCH] S390: Don't test nanoseconds in io/tst-stat.c

[Adhemerval Zanella <adhemerval.zanella@linaro.org>]

On 23/03/2021 13:13, Stefan Liebler wrote:
> On 18/03/2021 14:31, Adhemerval Zanella wrote:
>>
>>
>> On 17/03/2021 10:03, Stefan Liebler wrote:
>>> Both new tests io/tst-stat and io/tst-stat-lfs (_FILE_OFFSET_BITS=64)
>>> are comparing the nanosecond fields with the statx result.  Unfortunately
>>> on s390(31bit) those fields are always zero if old KABI with non-LFS
>>> support is used.  With _FILE_OFFSET_BITS=64 stat is using statx internally.
>>>
>>> As suggested by Adhemerval this patch disables the nanosecond check for
>>> s390(31bit).
>>
>> LGTM, the fstatat call does not call statx and even for LFS that call statx
>> it might ended calling old stat syscall in the fallback part that does not
>>
>> About the __ASSUME_STATX note Joseph has raised, I think we should add it
>> on Linux at least for fstatat64 implementation. However it does not really
>> help on the fstatat one.  I will try to spare some time to make fstatat.c
>> use statx as well, so we can tie the test to __ASSUME_STATX.
>>
>> Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>
>>
> 
> Hi Adhemerval,
> 
> Sorry for the delay, I was busy with another project.
> Thanks for your series
> "[PATCH 1/5] linux: Implement fstatat with __fstatat64_time64"
> https://sourceware.org/pipermail/libc-alpha/2021-March/124191.html
> 
> As also mentioned there, with your series, at least on my s390 systems,
> stat is then using statx and the nanosecond fields are not zero anymore.
> 
> Shall I commit my patch as is and as soon as you've commited your
> series, you can adjust support_stat_nanoseconds to return false if
> __ASSUME_STATX is not defined?

Yes, I can rebase on top your patch.  I think we still need to handle
the nanosecond missing support on older kernels.

Thanks for checking on s390, if you can review the patchset I would be
grateful