Re: Bug#396741: gnumach: General protection trap on network load

[<olafBuddenhagen@gmx.net>]

Hi,

On Mon, Mar 05, 2007 at 10:20:54AM +0100, Samuel Thibault wrote:

> Now that I fixed kdb's trace command (2007-02-20), could you try
> reproducing it and get a trace?

I already traced it manually a while back. (Sorry for not reporting my
findings...)

It turns out the crash happens in the line

   ipc_kmsg_enqueue(&net_queue_high, kmsg);

in net_packet() (net_io.c), which in turn is called from netif_rx() in
linux/dev/glue/net.c.

The most useful information I was able to gather from that, is that the
corruption occurs to net_queue_high -- but why? I tried fencing in the
problem, with some asserts checking the consistency of net_queue_high at
"random" places; but not much progress so far.

-antrik-