[vortex] 3c905 oddities

[Bogdan Costescu]

On Tue, 12 Dec 2000, Kevin Kress wrote:

> 3c905C X -TXM

Gotcha! Are you using this card in a fast (>= 700 MHz) computer ?

> I have the serial on irq 3.  and only the card on 4, but I would prefer
> to have it on 11 (where NIC's belong or is it 10?) is there an easy way
> to set that?

If this is the only one on IRQ 4, there should be nothing wrong with it.
You can check by looking at /proc/interrupts.
You can usually assign interrupts to PCI slots in BIOS, or at least can
reserve IRQ 4 for ISA, so that it's not used for PCI.

>  Manufacture date (MM/DD/YYYY) 10/14/2000, division H, product HN.

These are very close to the other reports of CX cards...

>  MII PHY found at address 1, status 0020.
>  MII PHY found at address 2, status 0020.
>  MII PHY found at address 3, status 0020.
>  MII PHY found at address 4, status 0020.

... and these too. If you are connecting to a device that supports
autonegotiation, everything should be handled automatically, so the MII
address is not really needed to be correct.

However, your report of not being able to receive packets is similar to
what Andrew said about Richard Gooch's card. Please read the CX related
threads from
http://www.scyld.com/pipermail/vortex/2000-November/thread.html

One of the messages describes a workaround: increasing a loop count to
4000000; although highly imperfect, this is the only way to use the CX
cards.

To Andrew: Did you receive the data from Richard w.r.t. timing of
different commands ?

Sincerely,

Bogdan Costescu

IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: Bogdan.Costescu@IWR.Uni-Heidelberg.De