Re: [PATCH v3] erofs: add sysfs feature entry for xattr prefixes

[Gao Xiang]

On Wed, Sep 02, 2026 at 01:38:16PM +0200, Greg KH wrote:
> On Wed, Sep 02, 2026 at 06:53:16PM +0800, Gao Xiang wrote:
> > Hi Greg,
> > 
> > On Wed, Sep 02, 2026 at 07:36:13AM +0200, Greg KH wrote:
> > > On Wed, Sep 02, 2026 at 09:50:39AM +0800, Jingbo Xu wrote:
> > > > Let /sys/fs/erofs/features/xattr_prefixes advertise that this kernel
> > > > supports the EROFS_FEATURE_INCOMPAT_XATTR_PREFIXES on-disk format.
> > > > 
> > > > Fixes: 6a318ccd7e08 ("erofs: enable long extended attribute name prefixes")
> > > > Cc: <stable@xxxxxxxxxxxxxxx> # 6.4+
> > > > Reviewed-by: Gao Xiang <xiang@xxxxxxxxxx>
> > > > Signed-off-by: Jingbo Xu <jefflexu@xxxxxxxxxxxxxxxxx>
> > > > ---
> > > > changes since v2:
> > > > - add Reviewed-by tag (Gao Xiang)
> > > > - update list in Documentation/ABI/testing/sysfs-fs-erofs
> > > > (https://sashiko.dev/#/patchset/20260901062243.109172-1-jefflexu%40linux.alibaba.com)
> > > > ---
> > > >  Documentation/ABI/testing/sysfs-fs-erofs | 2 +-
> > > >  fs/erofs/sysfs.c                         | 2 ++
> > > >  2 files changed, 3 insertions(+), 1 deletion(-)
> > > 
> > > Why is adding a new feature a stable thing?
> > 
> > It's not a new feature (Also, this patch has 3 lines diff):
> > This is a bugfix to address the missing part of
> > commit 6a318ccd7e08 ("erofs: enable long extended
> > attribute name prefixes"") since every on-disk features
> > should be shown in `/sys/fs/erofs/features/*` according to
> > the definition of:
> > 
> > What:           /sys/fs/erofs/features/
> > Date:           November 2021
> > Contact:        "Huang Jianan" <huangjianan@xxxxxxxx>
> > Description:    Shows all enabled kernel features.
> > ...
> > 
> > IOWs, this sysfs entry should be part of commit 6a318ccd7e08
> > to indicate this feature is supported since Linux 6.4+
> 
> Ok, but as this has not been present for a very long time, did anyone
> even notice and need this now in stable kernels?  :)

I think Jingbo needs it (btw, I'm no longer in Alibaba Cloud
anymore).

Also it would be better to port this into stable kernels so
that users can get the supported on-disk features in a more
unique obvious way.

Thanks,
Gao Xiang

> 
> thanks,
> 
> greg k-h
>