On Thu, 2006-01-19 at 10:53 -0500, Amy Griffis wrote:
On Wed, Jan 18, 2006 at 12:02:10PM -0600, Serge E. Hallyn wrote:
> Quoting Timothy R. Chavez (tinytim(a)us.ibm.com):
> > On Tue, 2006-01-17 at 17:19 -0500, Amy Griffis wrote:
> > > Here is an update that incorporates changes based on Tim's feedback:
> > > - sanity check for path with trailing /
> > > - call path_release
> > > - use audit_compare_watch
> > >
> > > and fixes panics due to the assumption of the existence of
> > > rule->watch.
> > >
> > > diff --git a/kernel/audit.h b/kernel/audit.h
> > > index f3b2a00..cc979e9 100644
> > > --- a/kernel/audit.h
> > > +++ b/kernel/audit.h
> > > @@ -52,6 +52,12 @@ enum audit_state {
> > > };
> > >
> > > /* Rule lists */
> > > +struct audit_watch {
> > > + char *path; /* watch insertion path */
> > > + struct list_head mlist; /* entry in master_watchlist */
> > > + struct list_head rules; /* associated rules */
> > > +};
> > > +
> > > struct audit_field {
> > > u32 type;
> > > u32 val;
> > > @@ -67,6 +73,8 @@ struct audit_krule {
> > > u32 buflen; /* for data alloc on list rules */
> > > u32 field_count;
> > > struct audit_field fields[AUDIT_MAX_FIELDS];
> > > + struct audit_watch *watch; /* associated watch */
> > > + struct list_head rlist; /* entry in audit_watch.rules list */
> > > };
> >
> > This may not really be that important, but if you switch to hlist_head
> > you have a 4-byte savings, which is something...
> >
> > AUDIT_MAX_FIELDS defaults to 64, sizeof(audit_field) is 12-bytes...
> > 768-bytes... 788-bytes per audit_krule?
>
> Agree.
Sorry, I'm not following how the list_head applies to the fields[]
array.
I was attempting to highlight the issue of unnecessary space consumption
by first offering a way to reduce the size of audit_krule in an easy way
(switching from list_head to hlist_head) and then bringing to everyone's
attention the larger problem of having such a large static array which
is probably mostly unused on average. Sorry I didn't make that
clearer :-)
-tim