On Friday 09 May 2008 17:29:04 Valdis.Kletnieks(a)vt.edu wrote:
On Fri, 09 May 2008 16:20:44 EDT, Jeremy Leonard said:
> -a exit,always -S sched_setparam -S sched_setscheduler -k RULE7
>
> type=SYSCALL msg=audit(04/25/08 16:37:48.568:194518) : arch=i386
> syscall=_newselect
OK, I'll bite - why is a select() syscall tripping sched_setparam or
sched_setschdeduler?
This record has a personality bit set unlike most events I ever see:
"arch=i386 syscall=_newselect per=400000"
I don't know if that is affecting the syscalls or not. Assuming it doesn't,
_newselect only occurs on ppc as far as I know. Its syscall 142. On x86_64,
sched_setparam is syscall 142. Not sure if that is the connection or not. But
something is wrong in this audit event. :) Also notice that the subject is
unconstrained, so they must be running some special SE Linux policy or have a
very unique kernel.
Or more importantly - are those two cutting audit events for the
wrong
reasons?
Yeah, could be. I think I'd also want to audit the setting of the personality
as that could be an attempt at masking the real actions of the user. Most
systems never do this. If you have a system doing it, you might want to keep
an eye on that.
-Steve