Re: [PATCH v6 6/6] fs: introduce file_getattr and file_setattr syscalls

[Christian Brauner]

On Thu, Jul 03, 2025 at 10:42:27AM +0200, Amir Goldstein wrote:
> On Thu, Jul 3, 2025 at 10:28 AM Christian Brauner <brauner@xxxxxxxxxx> wrote:
> >
> > On Wed, Jul 02, 2025 at 11:37:50AM -0700, Darrick J. Wong wrote:
> > > On Wed, Jul 02, 2025 at 03:43:28PM +0200, Amir Goldstein wrote:
> > > > On Wed, Jul 2, 2025 at 2:40 PM Christian Brauner <brauner@xxxxxxxxxx> wrote:
> > > > >
> > > > > > Er... "fsx_fileattr" is the struct that the system call uses?
> > > > > >
> > > > > > That's a little confusing considering that xfs already has a
> > > > > > xfs_fill_fsxattr function that actually fills a struct fileattr.
> > > > > > That could be renamed xfs_fill_fileattr.
> > > > > >
> > > > > > I dunno.  There's a part of me that would really rather that the
> > > > > > file_getattr and file_setattr syscalls operate on a struct file_attr.
> > > > >
> > > > > Agreed, I'm pretty sure I suggested this during an earlier review. Fits
> > > > > in line with struct mount_attr and others. Fwiw, struct fileattr (the
> > > > > kernel internal thing) should've really been struct file_kattr or struct
> > > > > kernel_file_attr. This is a common pattern now:
> > > > >
> > > > > struct mount_attr vs struct mount_kattr
> > > > >
> > > > > struct clone_args vs struct kernel_clone_kargs
> > > > >
> > > > > etc.
> > > > >file_attr
> > > >
> > > > I can see the allure, but we have a long history here with fsxattr,
> > > > so I think it serves the users better to reference this history with
> > > > fsxattr64.
> > >
> > > <shrug> XFS has a long history with 'struct fsxattr' (the structure you
> > > passed to XFS_IOC_FSGETXATTR) but the rest of the kernel needn't be so
> > > fixated upon the historical name.  ext4/f2fs/overlay afaict are just
> > > going along for the ride.
> > >
> > > IOWs I like brauner's struct file_attr and struct file_kattr
> > > suggestions.
> > >
> > > > That, and also, avoid the churn of s/fileattr/file_kattr/
> > > > If you want to do this renaming, please do it in the same PR
> > > > because I don't like the idea of having both file_attr and fileattr
> > > > in the tree for an unknown period.
> > >
> > > But yeah, that ought to be a treewide change done at the same time.
> >
> > Why do you all hate me? ;)
> > See the appended patch.
> 
> This looks obviously fine, but I wonder how much conflicts that would
> cause in linux-next?
> It may just be small enough to get by.

With such changes that's always a possibility but really I'll just
provide a branch with the resolutions for Linus to pull.