On 9/30/20 3:01 PM, Matheus Castanho via Libc-alpha wrote:
> Hi Adhemerval,
> 
> sysvipc/test-sysvsem started failing on ppc64le on kernels older than
> 5.1 after:
> 
> commit aaa12e9ff02b32d5fbb2f367d7d6b6985a2176d6
> Author: Adhemerval Zanella <adhemerval.zanella@linaro.org>
> Date:   Fri Sep 25 15:04:34 2020 -0300
> 
>     sysvipc: Fix semtimeop for !__ASSUME_DIRECT_SYSVIPC_SYSCALLS
> 
>     The __NR_ipc syscall does not support 64-bit time operations.  It
>     fixes 7c437d3778.
> 
>     Checked on i686-linux-gnu on a Linux 5.4.
> 
> It fails with:
> FAIL: sysvipc/test-sysvsem
> original exit status 1
> error: test-sysvsem.c:101: semop failed (errno=38)
> error: 1 test failures
> 
> Looks like semtimedop was added on Linux 5.1, so it makes sense that
> older kernels will fail with ENOSYS when calling that. So in such cases
> should we also apply time convertion and fallback to semtimedop/ipc in
> __semtimedop64 as done when !__ASSUME_DIRECT_SYSVIPC_SYSCALLS?
                        -----------------^

Sorry, I meant !_ASSUME_TIME64_SYSCALLS

> 
> The patch below seems to solve the issue, at least on ppc64le.
> 
> Suggestions?
> 
> Thanks,
> Matheus Castanho
> 
> --- 8< ---
> 
> diff --git a/sysdeps/unix/sysv/linux/semtimedop.c
> b/sysdeps/unix/sysv/linux/semtimedop.c
> index a9ad922ee2..510fea1852 100644
> --- a/sysdeps/unix/sysv/linux/semtimedop.c
> +++ b/sysdeps/unix/sysv/linux/semtimedop.c
> @@ -32,7 +32,7 @@ __semtimedop64 (int semid, struct sembuf *sops, size_t
> nsops,
>    int r = INLINE_SYSCALL_CALL (semtimedop_time64, semid, sops, nsops,
>                                timeout);
> 
> -#ifndef __ASSUME_TIME64_SYSCALLS
> +#if !(defined __ASSUME_TIME64_SYSCALLS) || __LINUX_KERNEL_VERSION <
> 0x050100
>    if (r == 0 || errno != ENOSYS)
>      return r;
> 

Also, looks like my email client messed up the diff *sigh*. I'm sending
a proper patch attached this time.

--
Matheus Castanho