Re: recent current panic

On Thursday 27 November 2003 06:38 am, Bjoern A. Zeeb wrote:
> On Thu, 27 Nov 2003, Yuriy Tsibizov wrote:
>
> Hi,
>
> > >i got a panic on recent -CURRENT:
> > >
> > ># tcpdump -i lo0 port 23 &
> > >[1] 507
> > >listening on lo0
> > >
> > ># telnet localhost
> > >Trying ::1...
> > >
> > >Wed Nov 26 14:51:23 MSK 2003
> > >Debugger+0x55:  xchgl   %ebx,in_Debugger.0
> >
> > [....]
> >
> > It looks very similar to
> > http://www.freebsd.org/cgi/query-pr.cgi?pr=59576. You can revert to
> > version 1.28 of net/bpf.h if you need tcpdump on lo, tun, ic, plip, disc
> > or gif interface right now.
>
> just to note: I ran tcpdump on tunN (which was a pppoe interface) w/o
> problems two days ago (with a recent world) while debugging and
> testing other net related things.

As was noted earlier, the issue is the addition of M_ASSERTVALID to the 
BPF_MTAP macro.  This fails in numerous cases because existing practice was 
to allocate a "fake mbuf" on the stack for passing local data in to bpf.  If 
this mbuf is not zero'd (or at least m_flags) then the assertion may trip.  
It's arguable what the correct solution is for these problems but you should 
be able to safely remove the assert until we can make a sweep over the 
drivers as the existing code has been around for a while (not to imply this 
makes it right :)).

	Sam