xennet0: rx no buf under NetBSD 6.1.5 domU

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

Good afternoon folks,

I've setup a domU to process NetFlow data using wip/pmacct, and have 
begun running into periods where the domU kernel starts complaining with 
a dmesg stream of:
xennet0: rx no buf
xennet0: rx no buf
xennet0: rx no buf
xennet0: rx no buf
xennet0: rx no buf

I've had a Google around, and found thread [1] which describes the 
behavior, but no indication of resolution. Based on the troubleshooting 
suggestions there, I'll include the following:

pmacct01$ sudo netstat -m
57 mbufs in use:
        27 mbufs allocated to data
        3 mbufs allocated to packet headers
        27 mbufs allocated to socket names and addresses
438 calls to protocol drain routines

pmacct01$ sudo vmstat -m | grep -e mbpl -e mclpl -e Name
vmstat: Kmem statistics are not being gathered by the kernel.
Name        Size Requests Fail Releases Pgreq Pgrel Npage Hiwat Minpg 
Maxpg Idle
mbpl         512    53230    0    52826   646   593    53    62     2 
inf    2
mclpl       2048      437    0      436   219   214     5    11     4 
131072   4

pmacct01$ netstat -a | grep -e Proto -e tcp -e udp
Proto Recv-Q Send-Q  Local Address          Foreign Address        State
tcp       18      0  pmacct01.postgresql    kibana01.65285 
ESTABLISHED
tcp        0      0  pmacct01.postgresql    kibana01.65383 
ESTABLISHED
tcp        0      0  *.postgresql           *.*                    LISTEN
tcp        0      0  pmacct01.ssh           10.11.83.66.51046 
ESTABLISHED
tcp        0      0  *.ssh                  *.*                    LISTEN
udp        0      0  localhost.ntp          *.*
udp        0      0  pmacct01.ntp           *.*
udp        0      0  *.ntp                  *.*
udp     2037      0  *.iop                  *.*
Proto Recv-Q Send-Q  Local Address          Foreign Address        (state)
tcp6       0      0  *.postgresql           *.*                    LISTEN
tcp6       0      0  *.ssh                  *.*                    LISTEN
udp6       0      0  fe80::1%lo0.ntp        *.*
udp6       0      0  localhost.ntp          *.*
udp6       0      0  fe80::216:3eff:f.ntp   *.*
udp6       0      0  localhost.65534        localhost.65534
udp6       0      0  *.ntp                  *.*

Given Manuel's suggestion in the referenced thread, I'm wondering if the 
*.iop queue is what leading to the condition; which is the NetFlow 
listener. Running netstat a handful of times rapidly shows the queue 
value jumping anywhere from 0 up to 10k entries.

Is the suggestion that the process we have listening there isn't 
collecting the waiting connections rapidly enough? I'd like the 
ascertain if it's something which is more "correct" or feasible to 
address by allocating additional buffer somewhere OS side, or if I need 
to look at the software side and see if it's not allocating something 
adequately for NetBSD specifically. (Which is entirely possible... the 
import of wip/pmacct is my doing.)

Suggestions welcomed.

Best,

Mike.

[1] http://mail-index.netbsd.org/netbsd-users/2012/11/19/msg012003.html