Re: [PATCH 2/4] UBIFS: Add xattr support for symlinks

[Artem Bityutskiy <dedekind1@gmail.com>]

On Mon, 2014-11-10 at 11:12 -0600, Ben Shelton wrote:
> On 11/10, Artem Bityutskiy wrote:
> > Could you please re-test this with any kernel and carefully verify
> > symlinks. I think this should not work, because in case of symlinks we
> > already store the link target path in the inode, and with this patch the
> > target patch will be over-written with the SELinux label. I expect this
> > to be seen easily on testing - symlink targets should be corrupted.
> > 
> > Artem.
> > 
> 
> I retested this with a 3.18-rc3 kernel on one of our ARM-based targets.
> The kernel has patch 1/4 with your changes, plus patches 2/4, 3/4, and
> 4/4 as posted.

Ben, thanks for re-testing. And yes, I was wrong. Now I checked again
and finally remembered how it works.

So yes, symlink inode has data, and the target is in the data.

If we add an extattr to a symlink, we create a separate inode for that
xattr under the symlink inode. Any new xattr gets a new inode.

Artem.