Re: [PATCH 2/2] vfs: support statx(..., NULL, AT_EMPTY_PATH, ...)

[Huacai Chen <chenhuacai@kernel.org>]

On Thu, Jul 4, 2024 at 12:54 AM Xi Ruoyao <xry111@xry111.site> wrote:
>
> On Wed, 2024-07-03 at 09:31 -0700, Linus Torvalds wrote:
> > On Wed, 3 Jul 2024 at 01:46, Christian Brauner <brauner@kernel.org>
> > wrote:
> > >
> > > We've now added AT_EMPTY_PATH support with NULL names because we
> > > want to
> > > allow that generically. But I clearly remember that this was
> > > requested
> > > to make statx() work with these sandboxes. So the kernel has done
> > > its
> > > part. Now it's for the sandbox to allow statx() with NULL paths and
> > > AT_EMPTY_PATH but certainly not for the kernel to start reenabling
> > > old
> > > system calls.
> >
> > Those old system calls are still used.
> >
> > Just enable them.
> >
> > statx isn't the promised land. Existing applications matter. And there
> > is absolutely nothing wrong with plain old 'stat' (well, we call it
> > "newstat" in the kernel for historical reasons) on 64-bit
> > architectures.
> >
> > Honestly, 'statx' is disgusting. I don't understand why anybody pushes
> > that thing that nobody actually uses or cares about.
>
> Hmm why it was added in the first place then?  Why not just NAK it?  If
> someone tries to add a "seccomp sandbox" into my project I'll
> immediately NAK it anyway :).
>
> And should we add stat_time64, fstat_time64, and fstatat_time64 to stop
> using statx on 32-bit platforms too as it's disgusting?
>
> Also some bad news: Glibc has this:
>
> #if (__WORDSIZE == 32 \
>      && (!defined __SYSCALL_WORDSIZE || __SYSCALL_WORDSIZE == 32)) \
>     || defined STAT_HAS_TIME32 \
>     || (!defined __NR_newfstatat && !defined __NR_fstatat64)
> # define FSTATAT_USE_STATX 1
> #else
> # define FSTATAT_USE_STATX 0
> #endif
>
> So if a LoongArch Glibc is built with Linux kernel headers >= 6.11,
> it'll use fstatat **even configured --with-kernel=5.19** and fail to run
> on Linux kernel <= 6.10.  This will immediately blow up building Linux
> From Scratch on a host distro with an "old" kernel.
The patch which adds newstat back will CC the stable list and be
backported to old kernels.

Huacai

>
> <sarcasm>Alright, some Google project matters but Glibc does not matter
> because it uses a disgusting syscall in the first place.</sarcasm>
>
> We have to add some __ASSUME_blah_blah here now.
>
> To make things worse Glibc 2.40 is being frozen today :(.  Copying to
> libc-alpha and the RM.
>
> --
> Xi Ruoyao <xry111@xry111.site>
> School of Aerospace Science and Technology, Xidian University