repeated panics/lock-ups while tcpdump-ing over ssh

[Edgar Fuß <ef%math.uni-bonn.de@localhost>]

I posted this on tech-kern, but was asked to post to tech-net:

I repeatedly experience panics and lock-ups while tcpdump-ing on an 
8.1/amd64 machine over ssh. The net I am tracing is different from 
the one the ssh connection was on.

1. panic: tp->tlastm == NULL (in sshd)
backtrace:
breakpoint
vpanic
snprintf
tcp_output
tcp_send_wrapper
sosend
soo_write
dofilewrite
sys_write
syscall

2. lock-up (in pid 0.3): wscons switching not working, numlock working, 
cursor flashes on keypresses. Could enter ddb:
backtrace (before interrupt)
sbappendaddr
udp4_sendup
udp_input
ipintr
softint_dispatch
DDB lost frame
Xsoftintr

I could cont from there and re-enter ddb, giving the same backtrace.

3. panic: m_copym: m == 0, len 152 [!COPYALL] (in squid)
backtrace:
breakpoint
vpanic
snprintf
m_copym0
tcp_output
tcp_send_wrapper
sosend
soo_write
dofilewrite
sys_write
syscall

4. panic: sbappendaddr
backtrace:
breakpoint
vpanic
snprintf
sbappendaddrchain
udp4_sendup
udp_input
ipintr
softint_dispatch
DDB lost frame
Xsoftintr

I do have show reg output of all incidents.

This is getting scary. About every other time I tcpdump over a ssh connection 
to a 8.1 server, it locks up or panics. I never experienced anything close 
with 6.1.