[PATCH 3.8 076/116] xfs: ioctl check for capabilities in the current user namespace

[Eric W. Biederman]

Kamal Mostafa <kamal at canonical.com> writes:

> On Wed, 2014-07-23 at 09:12 +1000, Dave Chinner wrote:
>> On Tue, Jul 22, 2014 at 03:21:27PM -0700, Kamal Mostafa wrote:
>> > 3.8.13.27 -stable review patch.  If anyone has any objections, please let me know.
>> > 
>> > ------------------
>> > 
>> > From: Dwight Engen <dwight.engen at oracle.com>
>> > 
>> > commit fd5e2aa8653665ae1cc60f7aca1069abdbcad3f6 upstream.
>> > 
>> > Use inode_capable() to check if SUID|SGID bits should be cleared to match
>> > [...]
>> 
>> Why are you backporting this to 3.8? namespace support didn't come
>> along until much later, so grabbing one patch out of themiddle of a
>> patch series to allow userns support in XFS is likely to cause
>> problems because there's no supporting code in XFS it.
>> 
>> Please don't randomly cherry pick userns support patches that change
>> permission checks back into kernels that don't have userns support.
>
> Yup, that's why we ask for reviews all right!  I've dropped these from
> the 3.8-stable queue:
>
>     fs,userns: Change inode_capable to capable_wrt_inode_uidgid

The fs,userns: Chage inode_capable to capable_wrt_inode_uidgid is
appropriate for 3.8.  I think that one is applicable all of the way
back to 3.4

I don't know if xfs in 3.8 called inode_capable in xfs and if it didn't
you can remove that hunk.  To keep things very simple you could just
skip the rename of inode_capable to capable_wrt_uidgid and just
include the one line change to add kgid_has_mapping.

But that bug fix is very much applicable to older kernels.

Eric