I am able to work on Gentoo and SELinux about one hour per day. It's more in total time, but being a bit exhausted makes me act a bit more slowly which boils down to about one hour per day. And one hour per day isn't bad, you're able to do many things in that hour.

The last few days, I've been hunting SELinux denials. I've set my mind onto releasing the 2.20120215-r5 policy only when I've been able to boot a minimalistic Gentoo installation without any visible denials, so either dontaudit them or fix them. Of course, I only want to allow if I'm absolutely confident that they are needed on some systems, but I also only want to dontaudit when I understand what it is doing (and find that it isn't something needed).

Some of the denials are driving me up the walls, often having an entire evening hunting for a single one... and this is why you haven't seen many updates since a week or so. The one I'm hunting now, is shown in the logs as follows:

Mar 12 20:21:32 testsys kernel: [    6.550618] type=1400 audit(1331580090.874:4): avc:  denied  { getattr } 
  for  pid=1512 comm="fuser" path="socket:[3159]" dev=sockfs ino=3159 
  scontext=system_u:system_r:initrc_t tcontext=system_u:system_r:udev_t tclass=unix_stream_socket
Mar 12 20:21:32 testsys kernel: [    6.551232] type=1400 audit(1331580090.875:5): avc:  denied  { getattr }
  for  pid=1513 comm="fuser" path="socket:[3160]" dev=sockfs ino=3160
  scontext=system_u:system_r:initrc_t tcontext=system_u:system_r:udev_t tclass=netlink_kobject_uevent_socket
Mar 12 20:21:32 testsys kernel: [    6.562005] type=1400 audit(1331580090.885:6): avc:  denied  { getattr }
  for  pid=1530 comm="fuser" path="socket:[3705]" dev=sockfs ino=3705
  scontext=system_u:system_r:initrc_t tcontext=system_u:system_r:udev_t tclass=netlink_kobject_uevent_socket
... (these netlink_kobject_uevent_socket ones are repeated a few times)

I have no idea who (or what) is executing fuser to find some information. The shown PIDs are those of fuser, and of course that isn't running anymore when the system is booted. The timeframe shown also doesn't seem to provide much information, because it is the time that it is logged by the system logger apparently (I once hoped I was wrong here, but repeated tests and introducing delays and such seems to confirm it). And because the target is on dev=sockfs, it's hardly something I'm able to actively search for.

Or at least that I know of.

The source context is initrc_t, so it is started from an init script. And the target is always udev_t, so it might be triggered by the udev (or a udev-related) init script (as it seems to only look for these of udev, but that might be a coincidence). But alas, I still don't know what is calling it as I can't find a script or udev rule that calls fuser :-( It doesn't affect the runtime behavior of my system (everything seems to work just fine) so I might go on and dontaudit it. But I so want to know what this is about.

To be continued...


Comments

To comment as a guest, use "Or sign up with disqus" and then select the "I'd rather post as guest" option.

comments powered by Disqus