[tulip-bug] Very slow performance with FA310TX
John William
jw2357@hotmail.com
Sun, 24 Sep 2000 21:48:14 GMT
>From: Donald Becker <becker@scyld.com>
>On Sun, 24 Sep 2000, John William wrote:
>
> > I am having problems with a Netgear FA310TX card. Transfer rates to/from
>the
...
>
>This sounds like either a
> Duplex mismatch
> Bad cable pairing, which works at 10BaseT but not 100baseTx
The card is connected to a FS105 10/100 switch on port #1. Both the switch
and card indicate 100MBS and full duplex. They are connected with a 2 meter
CAT5 pre-made cable. I'm 99% sure the hardware setup is good because I can
pull the card and pop it into a Win98 box and get >8MB/sec with the same
card/cable/switch.
> > I'm not even sure where to start looking for a problem since everything
> > works. The only hint that something might be wrong is under heavy load,
>the
> > 310TX starts throwing a lot of frame errors:
> >
> > eth1 Link encap:Ethernet HWaddr 00:A0:CC:39:B1:F9
> > RX packets:530996 errors:0 dropped:0 overruns:0 frame:5735
> > TX packets:495731 errors:1 dropped:0 overruns:1 carrier:0
>...
> > Under light load, no frame errors.
>
>Verify that that you are not forcing the duplex on this end. That will
>turn
>off autonegotiation, so your link partner think the link is
>100baseTx-half-duplex.
As far as I know I'm not forcing duplex. I am not passing any parameters to
the tulip.o module. "tulip-diag" output shows:
tulip-diag.c:v2.03 7/31/2000 Donald Becker (becker@scyld.com)
http://www.scyld.com/diag/index.html
Index #1: Found a Lite-On 82c168 PNIC adapter at 0xfe00.
Port selection is MII, full-duplex.
Transmit started, Receive started, full-duplex.
The Rx process state is 'Waiting for packets'.
The Tx process state is 'Idle'.
The transmit threshold is 1024.
Use '-a' or '-aa' to show device registers,
'-e' to show EEPROM contents, -ee for parsed contents,
or '-m' or '-mm' to show MII management registers.
So the card thinks it's full-duplex. The switch indicates full-duplex, so I
have to assume it's correct.
After a little tweaking, outbound transfers from the machine are much
better, but inbound is still very poor. Light network activity seems to be
OK, but past a certain threshold performance really nose-dives and the
"frame" errors start cropping up. The harder I try and "push" data into the
machine the slower it gets (and the more frame error packets I get).
Is there anything else I can look at or try to do to narrow the problem
down? When I can get some time on the machine, I will try and reboot it
uniprocessor to see if there is a SMP problem but other than that I'm at a
loss of how to proceed to debug this problem.
_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
Share information about yourself, create your own public profile at
http://profiles.msn.com.