> On 17 Jul 2021, at 19:32, Hrvoje Popovski <[email protected]> wrote: > > On 17.7.2021. 1:02, Alexander Bluhm wrote: >> On Fri, Jul 16, 2021 at 10:57:24PM +0200, Alexander Bluhm wrote: >>> All I said is more or less theory, I did not test it yet. >> I should not send untested diffs. New version one does not crash >> immediately. I removed a netlock that is already taken due to not >> queuing. This also fixes the tdb->tdb_odrops++ spotted by mvs@. >> >> Note that avoiding queues is the fastest way do IPsec. >> http://bluhm.genua.de/perform/results/2021-07-15T23:54:11Z/perform.html >> >> This diff is the middle column. >> http://bluhm.genua.de/perform/results/2021-07-15T23:54:11Z/gnuplot/ipsec.html >> >> bluhm > > Hi, > > with this diff i'm getting very stable traffic over tunnel and it's > little faster. > Even with your last diff on tech@ > https://marc.info/?l=openbsd-tech&m=162645141414262&w=2 > i'm seeing traffic drops, less frequent, but i'm seeing it... > > Do you want me to test this diff combined with your ipsec diff > on tech@ ? > And this diff with parallel forwarding?
Hi, Parallel forwarding breaks at least 'ipo->ipo_tdb’ which is not yet ready for parallelisation. I posted the diff to tech@ which documents current locking scheme for some ipsec(4) structures. It still pending for review.
