nfe driver bad performance on FreeBSD 7

[kordex -]

I found the reason for bad performance. It also affected the non-free nvidia
glx driver. It was options _KPOSIX_PRIORITY_SCHEDULING # POSIX P1003_1B
real-time extensions not set in the kernel configuration.

--Mikko Kortelainen

2008/10/25 Pyun YongHyeon <pyunyh at gmail.com>

> On Thu, Oct 23, 2008 at 04:02:48PM +0300, kordex - wrote:
>  > Same issue as
>  >
> http://unix.derkeiler.com/Mailing-Lists/FreeBSD/net/2008-09/msg00278.html
>  >
>  > I got HP Pavilion dv6646eo laptop with same network chip. Max throughput
> is
>  > 800kB/s with scp. same with generic kernel.
>  >
>  > nfe0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu
> 1500
>  >         options=8<VLAN_MTU>
>  >         ether 00:00:00:00:00:00
>  >         inet 192.168.1.200 netmask 0xffffff00 broadcast 192.168.1.255
>  >         media: Ethernet autoselect (100baseTX <full-duplex>)
>  >         status: active
>  > lilian#
>  >
>  > MAC being all zeros is not done by me. It's BIOS doing that. I wonder if
>  > that can cause things like this. No PXE Boot for me :| Should I send
> this to
>  > warranty for that?
>  >
>
> Don't use all zeroed ethernet address. You can assign fake ethernet
> address to nfe(4) with "ifconfig nfe0 ether 00:01:02:03:04:05".
> The problem is why nfe(4) got all zeord station address from
> controller. Does nve(4) also show the same ethernet address?
>
>  > lilian# uname -a
>  > FreeBSD lilian.xnet.kx 7.0-RELEASE FreeBSD 7.0-RELEASE #3: Sat Sep 27
>  > 20:36:01 UTC 2008     root at lilian.xnet.kx
> :/usr/obj/usr/src/sys/LILIAN_KERN
>  > i386
>  >
>  >
>  > lilian# dmesg
>  > Copyright (c) 1992-2008 The FreeBSD Project.
>  > Copyright (c) 1979, 1980, 1983, 1986, 1988, 1989, 1991, 1992, 1993, 1994
>  >         The Regents of the University of California. All rights
> reserved.
>  > FreeBSD is a registered trademark of The FreeBSD Foundation.
>  > FreeBSD 7.0-RELEASE #3: Sat Sep 27 20:36:01 UTC 2008
>  >     root at lilian.xnet.kx:/usr/obj/usr/src/sys/LILIAN_KERN
>  > Timecounter "i8254" frequency 1193182 Hz quality 0
>  > CPU: AMD Turion(tm) 64 X2 Mobile Technology TL-58 (1908.67-MHz 686-class
>  > CPU)
>  >   Origin = "AuthenticAMD"  Id = 0x60f81  Stepping = 1
>  >
>
> [...]
>
>  > nfe0: <NVIDIA nForce MCP65 Networking Adapter> port 0x30e0-0x30e7 mem
>  > 0xf2487000-0xf2487fff irq 20 at device 6.0 on pci0
>  > miibus0: <MII bus> on nfe0
>  > rgephy0: <RTL8169S/8110S/8211B media interface> PHY 1 on miibus0
>  > rgephy0:  10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT,
>  > 1000baseT-FDX, auto
>  > nfe0: [FILTER]
>
> There was one report of MCP65's poor performance but he said it
> happens on 1000Mbps link only. Normally poor network performance
> comes from speed/duplex mismatch. Does link partner also agree on
> resolved speed/duplex of nfe(4)? Would you show me the output of
> "netstat -ndI nfe0"?
>
> nfe(4) in 8-CURRENT supports hardware MAC counters and the counters
> give us very valuable information to diagnose driver issues. Would
> you try latest CURRENT and show me the output of
> "sysctl dev.nfe.0.stats" after some network activities?
>
> --
> Regards,
> Pyun YongHyeon
>