Re: [patch] voluntary-preempt-2.6.9-rc1-bk4-R1

[Stephen Smalley]

On Tue, 2004-09-14 at 14:32, Ingo Molnar wrote:
> it seems a bit odd. avc_has_perm_noaudit() calls avc_insert(), and from
> the entry to avc_insert(), to the memcpy() call done by
> avc_has_perm_noaudit() there were 204 usecs spent. That's alot of time -
> 175 thousand cycles! Now if you check out the code between line 1009 and
> 1019 in security/selinux/avc.c, there's little that could cause this
> amount of overhead. avc_insert() itself is rather simple - it does an
> avc_claim_node() call which is an inlined function so it doesnt show up
> in the trace. That function _might_ have called acv_reclaim_node() which
> seems to do a linear scan over a list - i dont know how long that list
> is typically but it could be long. Could you add "#define inline" near
> the top of avc.c (but below the #include lines) so that we can see how
> the overhead is distributed in the future?

May be rendered moot by the RCU patches for SELinux, see
http://marc.theaimsgroup.com/?l=linux-kernel&m=109386501021596&w=2

-- 
Stephen Smalley <sds@xxxxxxxxxxxxxx>
National Security Agency

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/