Linux 2.2.14 and FA310TA cards

[Donald Becker]

On 26 Apr 2000, Chris Martin wrote:

> I have two FA310TX cards, one in a 150 MHz Pentium machine and the
> other in a 600 MHz dual Pentium III machine linked by an FE104 hub (a
> Netgear FB104 starter kit).
...
> No problems were reported when I tested the connection under DOS using
> the DIAG.EXE on the diagnostic disk.
...
>   53 packets transmitted, 21 packets received, 60% packet loss
>   round-trip min/avg/max = 0.3/0.3/0.3 ms

Good RT time, but horrible reliability.  Ping doesn't stress the duplex,
something else is going on.

> While pinging the slow machine from the fast machine shows
> 
>   51 packets transmitted, 23 packets received, +10 corrupted, 54% packet loss
>   round-trip min/avg/max = 0.2/0.2/0.2 ms

Oohhhhh, this is bad.  Very bad.

> The corrupted packets show an output like
> 
> 64 bytes from 192.168.1.1: icmp_seq=40 ttl=255 time=0.2 ms (BAD CHECKSUM!)

The Ethernet chip guards against packet corruption on the wire with the CRC.
If you get bad data from noise, the card discards it before the driver even
sees the contents.

So you are getting corruption either in a switch (but you have a repeater,
so rule that out) or on the PCI bus or memory system.

> I originally used the driver (tulip.c) from the kernel tarball, this
> was 0.91g-ppc, I then replaced it in the kernel source with the driver
> supplied on the Netgear driver/diagnostic disk, 0.89K, and then
> downloaded the development driver, 0.91g.

I don't think that the driver version will change the behavior.  This looks
more like a memory problem.

> Netgear support say they only know Windows and ask if the cards are in
> a bus mastering slot.

There are not-fully-bus-mastering slots that will seem to work.  But that's
a pretty rare problem.

Do you have a third system or card to test with, so that you can identify
which machine is the problem?

Donald Becker				becker@scyld.com
Scyld Computing Corporation
410 Severn Ave. Suite 210
Annapolis MD 21403


-------------------------------------------------------------------
To unsubscribe send a message body containing "unsubscribe"
to linux-tulip-request@beowulf.org