Re: polling(4) and rl(4)

2004-04-12 Thread Ruslan Ermilov
On Mon, Apr 12, 2004 at 10:44:37AM +0400, Roman Kurakin wrote: > Ruslan Ermilov wrote: > > >As an aside, I've started working on the ``[-]polling'' option for > >ifconfig(8) that, when done, will allow changing the polling status > >of individual interfaces in run-time, e.g., the following command

Re: Per-interface polling(4) controls

2004-04-12 Thread Ruslan Ermilov
Dear -STABLE users, I've backported my patches that implement per-interface polling(4) controls. The RELENG_4 patchset for testing is available here: http://people.FreeBSD.org/~ru/patches/polling.patch The patchset also includes an updated vr(4) driver with polling(4) support. Cheers,

Re: polling(4) and rl(4)

2004-04-12 Thread Roman Kurakin
Ruslan Ermilov wrote: On Mon, Apr 12, 2004 at 10:44:37AM +0400, Roman Kurakin wrote: Ruslan Ermilov wrote: As an aside, I've started working on the ``[-]polling'' option for ifconfig(8) that, when done, will allow changing the polling status of individual interfaces in run-time, e.g., the

re(4): puzzling netperf result

2004-04-12 Thread Christian Weisgerber
I just did some quick and dirty checks with netperf and noticed a somewhat surprising result. Machine A: Alpha PC164, FreeBSD 5.2-CURRENT, re(4), 1000Base-T Machine B: Alpha PC164SX, OpenBSD 3.5, de(4), 100Base-TX Switch:StarChip SGS-1008 (10/100/1000Base-T) Running netperf -t UDP_STREAM on m

Re: suggested patches for netinet6/

2004-04-12 Thread JINMEI Tatuya / 神明達哉
(cc'ing to [EMAIL PROTECTED]) > On Fri, 9 Apr 2004 04:27:20 -0700, > Luigi Rizzo <[EMAIL PROTECTED]> said: > While adapting to ipv6 the new arp table code I am developing > following Andre's ideas, i hit a few places that would deserve a > fix independently of that: > + nd6_nud_hint() i

Re: re(4): puzzling netperf result

2004-04-12 Thread Ruslan Ermilov
On Mon, Apr 12, 2004 at 02:33:39PM +, Christian Weisgerber wrote: > I just did some quick and dirty checks with netperf and noticed a > somewhat surprising result. > > Machine A: Alpha PC164, FreeBSD 5.2-CURRENT, re(4), 1000Base-T > Machine B: Alpha PC164SX, OpenBSD 3.5, de(4), 100Base-TX > Sw

Re: suggested patches for netinet6/

2004-04-12 Thread Luigi Rizzo
On Mon, Apr 12, 2004 at 11:34:04PM +0900, JINMEI Tatuya / [EMAIL PROTECTED]@C#:H?(B wrote: > (cc'ing to [EMAIL PROTECTED]) [thanks for the cc] > > + nd6_nud_hint() is only called as nd6_nud_hint(NULL, NULL, 0); > > in some cases from netinet/tcp_input.c > > With these args, the routine is a

Re: re(4): puzzling netperf result

2004-04-12 Thread Christian Weisgerber
Ruslan Ermilov: > > How does the machine get the idea it is pushing 200 Mbit/s down a > > 100 Mbit/s link? > > Does ``netstat -I re0 -w 1'' show the same numbers while you're > running the UDP_STREAM test? Yes, it does (~2600 bytes). I have two further GigE-equipped OpenBSD boxes on the LAN

Re: [nwebe] How to track TCP socket variables? (cwnd, ssthresh)

2004-04-12 Thread Bruce M Simpson
On Sun, Apr 11, 2004 at 11:27:11AM +0200, Knocke wrote: > trpt on my system keeps on saying : > > % trpt: /boot/kernel/kernel: no namelist > > so probably no sockets are currently SO_DEBUG ready. You probably need to recompile your kernel with makeoptions TCPDEBUG. Diffing up something simple l

Current problem reports assigned to you

2004-04-12 Thread FreeBSD bugmaster
Current FreeBSD problem reports Critical problems Serious problems Non-critical problems S Submitted Tracker Resp. Description --- o [2003/07/11] kern/54383 net NFS root configurations without dynamic

Re: re(4): puzzling netperf result

2004-04-12 Thread Christian Weisgerber
Christian Weisgerber <[EMAIL PROTECTED]> wrote: > Running netperf -t UDP_STREAM on machine A with target B reports a > throughput of ~200(!) Mbit/s. By popular request: [EMAIL PROTECTED] /usr/local/netperf/netperf -H bardioc -t UDP_STREAM UDP UNIDIRECTIONAL SEND TEST to bardioc : histogram Socke

Re: suggested patches for netinet6/

2004-04-12 Thread JINMEI Tatuya / 神明達哉
> On Mon, 12 Apr 2004 07:56:38 -0700, > Luigi Rizzo <[EMAIL PROTECTED]> said: >> > + nd6_nud_hint() is only called as nd6_nud_hint(NULL, NULL, 0); >> > in some cases from netinet/tcp_input.c >> > With these args, the routine is a NOP. I propose to remove it >> > (and the associated

Re: suggested patches for netinet6/

2004-04-12 Thread Jun-ichiro itojun Hagino
> >> > + nd6_nud_hint() is only called as nd6_nud_hint(NULL, NULL, 0); > >> > in some cases from netinet/tcp_input.c > >> > With these args, the routine is a NOP. I propose to remove it > >> > (and the associated field, ln_byhint, in struct llinfo_nd6) on other OSes we call nd6_nud_h

Re: suggested patches for netinet6/

2004-04-12 Thread JINMEI Tatuya / 神明達哉
> On Mon, 12 Apr 2004 07:56:38 -0700, > Luigi Rizzo <[EMAIL PROTECTED]> said: >> > + is it ok to remove the __P() from the header files, ANSIfy >> > the function declarations and make them static as appropriate ? >> > Of course this ought to be done as a separate step. >> >> I myself

Re: suggested patches for netinet6/

2004-04-12 Thread Luigi Rizzo
On Tue, Apr 13, 2004 at 03:16:55PM +0900, JINMEI Tatuya / [EMAIL PROTECTED]@C#:H?(B wrote: ... > >> > + is it ok to remove the __P() from the header files, ANSIfy > >> > the function declarations and make them static as appropriate ? > >> > Of course this ought to be done as a separate step. .