On 08/30/2011 11:03 AM, Daniel P. Berrange wrote:
> This is non-trivial code. While we've already determined
that
> SELinux doesn't need SetProcessFDLabel, is there any chance that
> app-armor still needs this approach? If so, that would argue for
> keeping the function, but making it a no-op stub for SELinux, and
> still calling it in all the right places for the benefit of
> app-armor.
>
> I'm not familiar enough with app-armor theory of operation to answer
> this question, and without an answer, I can't give ack or nack.
The app armour code here was just copied from the similarly name
SetImageFDLabel, which resolves the FD into a file path using
/proc/self/fd/$FDNUM. This actually never worked for TCP sockets
with apparmour, so I don't believe I'm making anything worse.
Fair enough. If we actually need something for tcp socket labeling in
apparmor, then we can add a working solution later; disabling the
questionable code now is okay. You've given me an answer good enough
that I feel comfortable for:
ACK.
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library
http://libvirt.org