To: bsd-dk@bsd-dk.dk Subject: Re: FreeBSD som router Date: Sun, 04 Nov 2001 19:22:24 +0100 From: Poul-Henning Kamp <none@phk--critter.freebsd.dk.lh.bsd-dk.dk>
In message <none@20011104181037.H57188--vinyl.catpipe.net.lh.bsd-dk.dk>, Phil Regnauld writes:
>> pcib2: <DEC 21152 PCI-PCI bridge> at device 16.0 on pci0
>> pci2: <PCI bus> on pcib2
>> dc0: <Intel 21143 10/100BaseTX> port 0xec00-0xec7f mem 0xfdfff800-0xfdfffbff irq 11 at device 4.0 on pci2
>
> Jeg har selv so-so holdning til D-Link -- måske er det host systemet,
> (underdimensioneret CPU ?) eller hvad, men anyway:
>
> Nov 4 18:00:29 gw /kernel: dc1: TX underrun -- increasing TX threshold
> Nov 4 18:01:22 gw /kernel: dc0: TX underrun -- increasing TX threshold
> Nov 4 18:05:01 gw /kernel: dc3: TX underrun -- increasing TX threshold
> Nov 4 18:05:01 gw /kernel: dc1: TX underrun -- increasing TX threshold
> Nov 4 18:05:02 gw /kernel: dc3: TX underrun -- increasing TX threshold
> Nov 4 18:05:02 gw /kernel: dc1: TX underrun -- increasing TX threshold
> Nov 4 18:05:03 gw /kernel: dc0: TX underrun -- increasing TX threshold
> Nov 4 18:05:03 gw /kernel: dc1: TX underrun -- increasing TX threshold
>
> hver 1-2 minutter...
>
> (dc1, dc0 og dc3 plejer at køre fuld load det meste af tiden).
Det betyder at chippen ikke har kunnet hente data via PCI bussen
hurtigt nok og derfor bruger nogle mere konservative settings fra
nu af.
Der er lige lavet en ændring som sandsynligvis forbedre den situation
en anelse:
| src/sys/pci/if_dc.c:
| Revision 1.57 Sat Oct 27 00:59:17 2001 UTC (8 days, 17 hours ago) by luigi
| Branch: MAIN
| CVS Tags: HEAD
| Changes since 1.56: +7 -2 lines
| Diff to previous 1.56 (colored)
|
| Enable round-robin arbitration between transmit and receive unit
| in the 21143, instead of giving priority to the receive unit.
| This gives a 10-15% performance improvement in the forwarding rate
| under heavy load.
|
| Reviewed-by: Bill Paul
Den er MFC'ed til -stable et par dage senere.
-- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 phk@FreeBSD.ORG | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence.
This archive was generated by hypermail 2b30 : Wed 15 Nov 2006 - 18:24:16 CET