[j-nsp] m20 RE crashing on interface link changes

[Kevin Day]

I've got an M20 with dual RE's running JUNOS 6.2-20050122.3. I've got  
a pretty consistent crash of the master RE whenever any gigabit  
interface goes down. All it takes is me unplugging one of the cables  
and I get a kernel panic on the RE.

This is sent to the console:

Fatal double fault:
eip = 0xc03e0590
esp = 0xe4e84000
ebp = 0xe4e8403c
panic: double fault
dblfault_handler(c058e3c3,e4e84000,0,0,0) at dblfault_handler+0x5f
dblfault_handler(0,0,0,0,0) at dblfault_handler+0x5f

syncing disks... kernel trap 12 with interrupts disabled
kernel trap 12 with interrupts disabled
kernel trap 12 with interrupts disabled
(repeated for about 60 seconds at full console speed)



After a few minutes the backup RE kicked in, assumed mastership then  
crashed "Cause unknown...." as well. I didn't have a serial console  
attached to that one, but I did get core dumps in /var/crash on both.


This is a relatively new development, so I'm not sure what I might  
have done to start this, but I've rolled my config back as far as I  
can and it's still happening. Has anyone seen this before?