0.99N and 3C905C

Bogdan Costescu Bogdan.Costescu@IWR.Uni-Heidelberg.De
Mon Apr 3 08:58:30 2000


Hi,

I have some problems when trying to use the new version of the driver with
kernel 2.2.15pre16, SMP. The setup is: dual PIII-450 MHz on a Gigabyte6BXD
board and a 3C905C card, diskless (NSF-root), BayStack 350-24T switch.
Using 0.99N (12/15/99) driver with pci-netif.c 0.05 (9/16/99) and
k_compat.h 0.02 (1999/09/14), I get 8 instances of the card at boot, then
of course a lot of error messages (some SMP related, like simultaneous
entry in interrupt handler). How can I get rid of this?

I tried to get out of the "while" loop after the first found card, but I
still have a lot of errors afterwards: TX timeouts then some strange
"neighbour table full" that I have never seen...

I have run for months kernel 2.2.13 with 0.99L driver with no such
problems. I had another problem with full-duplex setting that I described
in a message on this list (24 Nov '99) and some Tx timeouts (status e000
and resetting the Tx ring pointer) that happen only rarely (after weeks
of running), so I cannot track them down. I have to say this is a node in
a Beowulf cluster, so the network is really loaded.

I would like to continue to use these drivers (as opposed to 3Com's) as
they seem a bit faster which an important point in a cluster setup. Any
suggestions or patches are welcome!

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


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