Hello,
my statement here is just for the record. we should have a follow up
discussion in a different thread, which is yet to be started (when time
will come).
</snip>
>
> > - Make ARP MP safe. Currently we need the kernel lock there or
> > it crashes. This creates latency for all kind of packets.
> > - Convert the rwlock in pf to mutex. I think your argument counts
> > much more there. But I cannot prove it.
>
> My argument about what? Didn't we all agree about converting the rwlock
> to a mutex for now?
>
I disagree here. If we will trade current rw-lock, which protects
state table in pf(4), we will see significant drop in throughput.
great deal of packets is forwarded as readers on pf_state_lock.
only those, which don't match state are taking slow path. to turn
rw-lock to mutex, we must break single giant tree into smaller
buckets protected by mutexes. Or something like that must happen
first.
thanks and
regards
sashan
