On 27.08.2013 01:07, Craig Rodrigues wrote: > Hi, > > On box 2, since I this is a debug kernel with WITNESS and INVARIANTS > enabled, I get a kernel panic. (see attached core.txt.gz)
It seems the log was stripped by maillist.
> The panic occurs here on line 1779:
>
> 1768 static struct ifnet *
> 1769 in6p_lookup_mcast_ifp(const struct inpcb *in6p,
> 1770 const struct sockaddr_in6 *gsin6)
> 1771 {
> 1772 struct route_in6 ro6;
> 1773 struct ifnet *ifp;
> 1774
> 1775 KASSERT(in6p->inp_vflag & INP_IPV6,
> 1776 ("%s: not INP_IPV6 inpcb", __func__));
> 1777 KASSERT(gsin6->sin6_family == AF_INET6,
> 1778 ("%s: not AF_INET6 group", __func__));
> 1779 KASSERT(IN6_IS_ADDR_MULTICAST(&gsin6->sin6_addr),
> 1780 ("%s: not multicast", __func__));
>
> If I look at gsin6->sin6_addr inside kgdb,
> I see:
>
> (kgdb) p gsin6->sin6_addr
> $1 = {__u6_addr = {__u6_addr8 =
> "\000\000\000\000\000\000\000\000\000\000���M|�", __u6_addr16 = {0, 0, 0,
> 0, 0, 65535, 19951, 54652}, __u6_addr32 = {0, 0,
> 4294901760, 3581693423}}}
>
>
> I am not so familiar with this part of the networking code.
> Can someone recommend where is the best place to fix
> this would be?
AFAIR, I already saw similar report here.
This is V4 mapped IPv6 address ::ffff:239.77.124.213.
I guess application is trying to use setsockopt with IPV6_JOIN_GROUP
option. And since outgoing interface isn't specified, the kernel is
trying to determine it from routing table. But this mapped address
triggers assert in in6p_lookup_mcast_ifp() function. It seems to me,
that v4mapped addresses isn't supported in the multicast code. If you
remove KASSERT from in6p_lookup_mcast_ifp(), this address will be
treated as invalid later.
--
WBR, Andrey V. Elsukov
signature.asc
Description: OpenPGP digital signature
