On 28/08/16(Sun) 13:39, Martijn van Duren wrote:
> Hello bugs,
>
> I just had the following panic switching from wireless to wired when
> doing a dhclient.
What do you mean by "switching"? What did you do?
Could you please tell me if the diff below fix your issue? The
'flushclone' call when adding a route is unnecessary and exposes a bug
introduced when we added support for multiple RTF_CONNECTED.
Index: net/route.c
===================================================================
RCS file: /cvs/src/sys/net/route.c,v
retrieving revision 1.317
diff -u -p -r1.317 route.c
--- net/route.c 22 Aug 2016 16:53:59 -0000 1.317
+++ net/route.c 29 Aug 2016 11:07:13 -0000
@@ -927,12 +927,13 @@ rtrequest_delete(struct rt_addrinfo *inf
return (ESRCH);
}
- /* clean up any cloned children */
- if ((rt->rt_flags & RTF_CLONING) != 0)
- rtflushclone(tableid, rt);
-
+ /* Release next hop cache before flushing cloned entries. */
rt_putgwroute(rt);
+ /* Clean up any cloned children. */
+ if (ISSET(rt->rt_flags, RTF_CLONING))
+ rtflushclone(tableid, rt);
+
rtfree(rt->rt_parent);
rt->rt_parent = NULL;
@@ -1177,11 +1178,6 @@ rtrequest(int req, struct rt_addrinfo *i
return (EEXIST);
}
ifp->if_rtrequest(ifp, req, rt);
-
- if ((rt->rt_flags & RTF_CLONING) != 0) {
- /* clean up any cloned children */
- rtflushclone(tableid, rt);
- }
if_group_routechange(info->rti_info[RTAX_DST],
info->rti_info[RTAX_NETMASK]);
