Re: [RFC] linux: Use fchmodat2 on fchmod for flags different than 0 (BZ 26401)

[Florian Weimer <fweimer@redhat.com>]

* Adhemerval Zanella:

> Linux 6.6 (09da082b07bbae1c) added support for fchmodat2, which is
> has similar semantic of fchmodat with an extra flag argument.  This
> allow fchmodat to implement AT_SYMLINK_NOFOLLOW and AT_EMPTY_PATH
> without the need to procfs.
>
> The syscall is registered on all architectures (with value of 452
> except on alpha which is 562, commit 78252deb023cf087).
>
> Checked on x86_64-linux-gnu on a 6.6 kernel.
>
> PS: setting it as RFC because there is no Linux 6.6 release yet.

Thanks for getting the work here started.

The syscall list update should definitely be in a separate commit.

> diff --git a/sysdeps/unix/sysv/linux/fchmodat.c b/sysdeps/unix/sysv/linux/fchmodat.c
> index 99527a3727..99d3df6440 100644
> --- a/sysdeps/unix/sysv/linux/fchmodat.c
> +++ b/sysdeps/unix/sysv/linux/fchmodat.c
> @@ -26,66 +26,80 @@
>  #include <sysdep.h>
>  #include <unistd.h>

diff -w makes the diff more manageable.

> +  /* The kernel system call does not have a mode argument.
> +     However, we can create an O_PATH descriptor and change that
> +     via /proc (which does not resolve symbolic links).  */

I would update the comment to mention the old fchmodat system call
explicitly.

The mechanics look okay to me.

I wonder how much breakage we are going to see due to EPERM instead of
ENOSYS on broken container runtimes.  We can certainly try to get the
world cleaned up for the glibc 2.39 release, we still have some time.

Thanks,
Florian