Re: squeeze more performance out of natd?

2002-02-12 Thread Richard A Steenbergen
On Mon, Feb 11, 2002 at 11:26:45AM -0800, Alfred Perlstein wrote: > failing that, there's always moving it into the kernel where the perf > would most likely get better by several orders of magnitude by avoiding > copies and userspace/kernel context switching. Of course copying the entire packet

socket options (struct sockopt)

2002-02-12 Thread Aleksander Rozman - Andy
Hi ! I am working on implementation of AX.25 on fbsd (as you probably already know)... I need to know how to create socket options (for use with xxx_ctlinput, xxx_ctloutput, getsockopt, setsockopt)? In which part of code could I see how socket options are created... Andy

gif0 tunnel and rip (using zebra)

2002-02-12 Thread Jay Austad
I'm trying to propagate rip routes using zebra (ripd) across a gif0 tunnel interface to another freebsd box running zebra. I have zebra and ripd running, but they only seem to broadcast routes out the fxp0 interface, and not across the gif0 tunnel. It's quite annoying. A "show ip protocols" i

~40 DupAcks And No Fast Retransmit !!

2002-02-12 Thread murthy kn
Hello all, I am using 4.3 BSD and from the below capture, I feel that there is some problem with the Fast Retransmit (unless I am missing something). I also recall that there was a short thread about some problems with Fast Retransmit earlier in Nov 2001 - that sender was resorting to the cos

Re: HEADS UP: device polling code merged into -stable

2002-02-12 Thread Luigi Rizzo
On Tue, Feb 12, 2002 at 07:43:35AM -0600, Greg Panula wrote: > I still get watchdog time-outs with the dc driver when I use a SMP kernel. > Hardware platform: Compaq DL360 with dual-cpus > NIC: D-Link's DFE570TX (4-port nic, chipset Intel 21143) > > cvsup'd the src tree this morning(~2am CST), b

Re: HEADS UP: device polling code merged into -stable

2002-02-12 Thread Greg Panula
Luigi Rizzo wrote: > > [Bcc to -stable as relevant there] > > I have merged into -stable the device polling code that has > been in -current for the past couple of months, plus a minor > change that addresses the 'device going deaf' problems that > some have reported with early patches of this c

RE: squeeze more performance out of natd?

2002-02-12 Thread Tariq Rashid
thanks for the suggestions, ideas and info everyone... i'll start by looking more closely at the source code and looking at what kqueue and kevents actually are (!) before i dive in and change natd. (for all those who suggested ipf/ipnat - yes i know, but unfortunately natd if part of a bigg