bnx panic with > 1 domU

[Mike Bowie <mbowie%buzmo.com@localhost>]

Good day all,

My PowerEdge 1950 seems to have a hard time (aka it panics) running 
4.99.72 and Xen 3.3.  It's been doing this for a few weeks, actually 
since I updated from 4.99.54... which had other issues preventing me 
from getting this far.

If I have more than one domU running, the box will panic after about two 
hours.  The only thing running under dom0 is snmpd and the same goes for 
the domUs.  Here's what I'm seeing:

From the console:
db> bt
breakpoint() at netbsd:breakpoint+0x5
panic() at netbsd:panic+0x255
bnx_rx_intr() at netbsd:bnx_rx_intr+0x37f
bnx_intr() at netbsd:bnx_intr+0xd7
evtchn_do_event() at netbsd:evtchn_do_event+0xfa
call_evtchn_do_event() at netbsd:call_evtchn_do_event+0xd
hypervisor_callback() at netbsd:hypervisor_callback+0xa3
idle_loop() at netbsd:idle_loop+0x148
db>

From the end of dmesg:
WARNING: mclpool limit reached; increase NMBCLUSTERS
panic: bnx0: Double mbuf allocation failure!
fatal breakpoint trap in supervisor mode
trap type 1 code 0 rip ffffffff804b675d cs e030 rflags 246 cr2 
7f7ffdfdc000 cpl
 6 rsp ffffa000459ce898
?

sysctl returns "kern.mbuf.nmbclusters = 2048", which I could clearly 
double or quadruple, but I did read a post which discussed a similar 
behavior:
http://mail-index.netbsd.org/current-users/2008/06/27/msg003233.html

Now that got me thinking that perhaps the issue is SMP related, because 
so far, if I only run one domU, it will stay up for a couple of days. 
(When I generally start messing with it again.)

Clearly there's more poking I need to do to isolate the problem... so if 
anyone has any requests for specific details or suggestions of avenues 
to explore, I'd be most appreciative.

Cheers,

Mike.