Messages in this thread | | | Date | Wed, 02 Jul 1997 08:27:33 +0000 | From | Jason Flynn 605619 <> | Subject | Re: Bug#10954: Kernel panic: skpush:under (fwd) |
| |
Eric.Schenk@dna.lth.se wrote: > > Phil Karn <karn@ka9q.ampr.org> writes: > >This morning my machine panicked with the following on the console: > > > >Kernel panic: skpush:under: 0014140c:84772916 > >In swapper task - not syncing > > > >According to my symbol table and my reading of the source code, this > >came from a call to skb_push() from within packet_rcv(). packet_rcv > >is at 001413c8 and packet_sendmsg is at 00141484. > > This indicates that either some device driver blew it when > setting up the ethernet headers on an incoming packet. > > >The machine has two Ethernet interfaces, a 3C590 and a NE2000. > >It's used as an IP router with a fairly complex configuration; both > >IP masquerading and IP-in-IP tunneling are used. At the time of > >the crash I had tcpdump running on the NE2000 interface. > > Can you give me a bit more information about this setup? > > My first guess is that the tunneling code is the culprit, > but it's a bit late for me to go digging yet tonight.
I've seen this happen quite a bit after a fragmented packet is re-assembled and put through the de-tunneler and then passed back down to be 're-received'. This is especially the case when bridging was in the mix too. I sent a patch to Alan to fix this with the bridging for 2.0 a couple of months ago. - It looks like a similar problem here.
J
| |