[vortex] 3c59x Driver lockup

Bogdan Costescu bogdan.costescu@iwr.uni-heidelberg.de
Tue Feb 4 10:04:00 2003


On Fri, 31 Jan 2003, Ryan Moore wrote:

> Too much work in interrupt, status e003.

e003 = host error - either PCI master abort or PCI target abort

> PCI bus error, bus status 80000020

Master abort (+upRxEarlyEnable which is a normal setting)

> Host error, FIFO diagnostic register 0000.

This means that nothing is wrong.

> PCI bus error, bus status 800000a0

Master abort (+dnInProg which is a normal indication during Tx + 
upRxEarlyEnable)

> Host error, FIFO diagnostic register 8000.

This means that the NIC is receiving a frame into the FIFO, which is 
normal during Rx.

> (note: didn't get the errors in that exact sequence, it is just a sample 
> of different errors I'm getting)

Well, the driver can only spit them in a certain sequence, so it doesn't 
matter...

> Briefly looking at the archives and searching via google, it looks like 
> this problem has been run into a decent amount by more than just myself.

Hmmm, not really. The first message, "too much work in interrupt" is 
indeed common for systems where there is some device/driver that keep the 
PCI bus busy for too long, so that the NIC cannot use it. But your system 
reports a different status which is an error condition.

> At first I thought it was a hardware problem, but I've tried several 
> different cards (all 905x) and two other motherboards and the problem is 
> appearing on them as well (all three mobos are Supermicro P6DGU/P6DBU 
> series.

Well, to me it looks like a hardware problem or maybe something to do 
with the PCI settings (f.e. latency), but maybe somebody like 
Donald who has much more experience with PCI than me can diagnoze it 
better. You should probably also provide more information about your 
system, like any other PCI devices that you are using.

> I've also tried the `bomerang_rx(dev);` suggestion

I don't know what suggestion this is, but I don't expect it to work. 
3Com's documentation specifies what should be done after these errors and 
the driver does it, so it recovers.

-- 
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