Date: Wed, 27 Sep 2006 12:14:32 +0400 From: Gleb Smirnoff <glebius@FreeBSD.org> To: Maxim Sobolev <sobomax@FreeBSD.org> Cc: cvs-src@FreeBSD.org, src-committers@FreeBSD.org, cvs-all@FreeBSD.org, Paolo Pisati <piso@FreeBSD.org> Subject: Re: cvs commit: src/etc Makefile libalias.conf src/lib/libalias Makefile src/lib/libalias/libalias Makefile src/lib/libalias/modules Makefile Makefile.inc src/lib/libalias/modules/cuseeme Makefile src/lib/libalias/modules/dummy ... Message-ID: <20060927081432.GH59833@FreeBSD.org> In-Reply-To: <4519BE2F.4000503@FreeBSD.org> References: <200609262326.k8QNQrtW098027@repoman.freebsd.org> <4519BE2F.4000503@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Sep 26, 2006 at 04:56:31PM -0700, Maxim Sobolev wrote: M> Looks cool. Do you have any numbers comparing performance of the old M> libalias and a new one running in kernel/userland mode? AFAIK, this change doesn't affect performance of the userland NAT. The main aim of this step was not performance, but separation of specific protocol NATing into separate pluggable modules. The kernel nat (I mean ng_nat) is probably pessimized be this change, because Paolo added locking to the list of plugins, this adds two atomic ops per packet. I hope, Paolo will rethink this. -- Totus tuus, Glebius. GLEBIUS-RIPN GLEB-RIPE
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060927081432.GH59833>