The card in question is a pcmcia card being used thru the pcmcia
support within the cardbus code.  ie. ne0 -> pcmcia0 -> cbb0 This
card worked fine using the cardbus patches to a current kernel
before the cardbus changes were integrated.  However, using a
-current kernel as of today, the cardbus bridge configures, finds
it's slots and finds the ne2k card, but as soon as the pcmcia code
disables the socket (pccbb_pcmcia_socket_disable) it goes nuts,
triggering the (correct) intr over and over until the machine
reboots (a split second later).  There is no card status event
associated with the intr, as far as the code can tell.

Has anyone seen this behavior before?  If not, I'll try again to
track it down, but I wanted to check because the last time I debugged
the cardbus code it took a week of commenting out random bits of
code to find what hack needed to be performed.

Ezra