[PATCH 1/5] linux: Implement fstatat with __fstatat64_time64

[Adhemerval Zanella]

On 26/03/2021 06:24, Stefan Liebler via Libc-alpha wrote:
> On 19/03/2021 19:31, Adhemerval Zanella via Libc-alpha wrote:
>> It makes fstatat use __NR_statx, which fix the s390 issue with
>> missing nanoxsecond support on compat stat syscalls (at least
>> on recent kernels) and limits the statx call to only one function
>> (which simplifies the __ASSUME_STATX support).
>>
>> Checked on i686-linux-gnu and on powerpc-linux-gnu.
>> ---
>>  sysdeps/unix/sysv/linux/fstatat.c | 52 +++++++------------------------
>>  1 file changed, 11 insertions(+), 41 deletions(-)
>>
>> diff --git a/sysdeps/unix/sysv/linux/fstatat.c b/sysdeps/unix/sysv/linux/fstatat.c
>> index 59efff615f..618c254d6f 100644
>> --- a/sysdeps/unix/sysv/linux/fstatat.c
>> +++ b/sysdeps/unix/sysv/linux/fstatat.c
>> @@ -26,33 +26,19 @@
>>  int
>>  __fstatat (int fd, const char *file, struct stat *buf, int flag)
>>  {
>> -  int r;
>> -
>> -# if STAT_IS_KERNEL_STAT
>> -  /* New kABIs which uses generic pre 64-bit time Linux ABI, e.g.
>> -     csky, nios2  */
>> -  r = INTERNAL_SYSCALL_CALL (fstatat64, fd, file, buf, flag);
>> -  if (r == 0 && (buf->__st_ino_pad != 0
>> -		 || buf->__st_size_pad != 0
>> -		 || buf->__st_blocks_pad != 0))
>> -    return INLINE_SYSCALL_ERROR_RETURN_VALUE (EOVERFLOW);
>> -# else
>> -#  ifdef __NR_fstatat64
>> -  /* Old KABIs with old non-LFS support, e.g. arm, i386, hppa, m68k, mips32,
>> -     microblaze, s390, sh, powerpc, and sparc.  */
>> -  struct stat64 st64;
>> -  r = INTERNAL_SYSCALL_CALL (fstatat64, fd, file, &st64, flag);
>> +  struct __stat64_t64 st64;
>> +  int r = __fstatat64_time64 (fd, file, &st64, flag);
>>    if (r == 0)
>>      {
>>        if (! in_ino_t_range (st64.st_ino)
>>  	  || ! in_off_t_range (st64.st_size)
>> -	  || ! in_blkcnt_t_range (st64.st_blocks))
>> +	  || ! in_blkcnt_t_range (st64.st_blocks)
>> +	  || ! in_time_t_range (st64.st_atim.tv_sec)
>> +	  || ! in_time_t_range (st64.st_mtim.tv_sec)
>> +	  || ! in_time_t_range (st64.st_ctim.tv_sec))
>>  	return INLINE_SYSCALL_ERROR_RETURN_VALUE (EOVERFLOW);
> OK
>>
>> -      /* Clear internal pad and reserved fields.  */
>> -      memset (buf, 0, sizeof (*buf));
> Do we have to clear the user provided struct as before?

I think it is not strictly required, but some architecture does
have padding and reserved fields.  I will reinstate it, the compiler
should have enough information to only zero the fields not actively
set.