[PATCH] xfs: in _attrlist_by_handle, copy the cursor back to userspace

[Mark Tinguely]

On 08/02/16 16:41, Dave Chinner wrote:
> On Tue, Aug 02, 2016 at 08:40:35AM -0700, Darrick J. Wong wrote:
>> On Tue, Aug 02, 2016 at 05:25:17AM -0700, Christoph Hellwig wrote:
>>> On Mon, Aug 01, 2016 at 08:56:51PM -0700, Darrick J. Wong wrote:
>>>> When we're iterating inode xattrs by handle, we have to copy the
>>>> cursor back to userspace so that a subsequent invocation actually
>>>> retrieves subsequent contents.
>>> Testcase?
>> Found it while continuing development of xfs_scrub.  I'll send along the
>> xfstest patch when I've finished polishing it.  (It was harder than usual since
>> xfs_io doesn't /use/ the attr-by-handle interface... nothing does.)
> It was (and probably still is) used by SGI's HSM. I thought there
> was some coverage of the interface in the dmapi part of the xfstests
> suite, but perhaps it's only tested by SGI's internal hsm test
> suite...
>
> Cheers,
>
> Dave.

You may be thinking of the DMAPI specific extended attribute interface 
and that is tested in xfstests. That function cannot cursor.

This change would be appropriate to xfsdump (dumping extended attributes 
via jdm_attr_list()) than our HSM.

I don't see a SGI test for attr_list_by_handle() nor jdm_attr_list(). 
There is mention in the xfstests source file src/open_unlink.c but that 
does not cursor.

--Mark Tinguely.