On 2007/07/19 12:15, Jeff Santos wrote: > > net.inet.ip.ifq.len=0 > net.inet.ip.ifq.maxlen=300 > net.inet.ip.ifq.drops=0 > > However, the problem did not go away altogether. I would like to > know if it has anything to do with the "memory" or "state > mismatch" statistics below:
"state-mismatch" is TCP packets which arrive out of a reasonable window around the expected sequence number. To examine them in detail, use pfctl -xm and look in syslog. "memory" is PFRES_MEMORY, this could well be it. the description is "Dropped due to lacking mem", it's triggered in quite a few places (grep for PFRES_MEMORY in /sys/dev/net/pf*). I wonder if it's worth looking at 'fragment crop'.
