svn commit: r228071 - head/sys/net

[Gleb Smirnoff]

On Mon, Nov 28, 2011 at 10:23:07PM +0200, Jaakko Heinonen wrote:
J> On 2011-11-28, Bjoern A. Zeeb wrote:
J> > > static void
J> > > if_clone_free(struct if_clone *ifc)
J> > > {
J> > > -	for (int bytoff = 0; bytoff < ifc->ifc_bmlen; bytoff++) {
J> > > -		KASSERT(ifc->ifc_units[bytoff] == 0x00,
J> > > -		    ("ifc_units[%d] is not empty", bytoff));
J> > > -	}
J> > > 
J> > > 	KASSERT(LIST_EMPTY(&ifc->ifc_iflist),
J> > > 	    ("%s: ifc_iflist not empty", __func__));
J> > > 
J> > > 	IF_CLONE_LOCK_DESTROY(ifc);
J> > > -	free(ifc->ifc_units, M_CLONE);
J> > > +	delete_unrhdr(ifc->ifc_unrhdr);
J> > > }
J> 
J> delete_unrhdr() KASSERTs that all allocations has been freed. Thus if
J> the leak below has occurred, the result will be a panic at this point
J> with INVARIANTS enabled kernel. If INVARIANTS is disabled, a memory leak
J> is possible.

Thanks, Jaakko! Good catch. I will try to fix that.

-- 
Totus tuus, Glebius.