[vortex] Re: DnStall period

[Bogdan Costescu]

On Sat, 13 May 2000, Donald Becker wrote:

> > - The NIC has started to transmit a packet.
> > - The next packet in memory is, say, 64 bytes.
> > - The NIC sees >32 bytes spare in the FIFO (but <64).
> 
> This doesn't happen on the Boomerang.  The TxFreeThreshold is set to 1536,
> which assures us that a packet transfer will not start until there is room
> in the FIFO.

We are testing on Cyclone/Tornado cards (B and C). These cards DO NOT HAVE
the TxFreeThreshold register (page 110 in 90x/B docs). For the same
effect, you can use DnBurstThresh (page 97 in C docs, default 256 bytes)
or (slightly different) DnPriorityThresh (page 100 in C docs, default 128
bytes). So, you're right saying "This doesn't happen on the Boomerang.",
but you're wrong assuming that this is also true for Cyclone/Tornado.

> > Bogdan,  I am at a loss to explain why increasing the loop count from
> > 2,000 to 4,000 changed anything for you.   You're on switched 100bT,
> > right?  You shouldn't be getting _any_ collisions (and when in full
> > duplex mode the NIC doesn't even look for collisions).  So what's going
> > on?
> > 
> > I suspect that you're mistaken and that upping the loop counter was not
> > the source of your success.

Adding wait_for_completion with loop counter = 4000 was the _only_ thing
that I've done to the driver. I tested it with normal ring sizes and then
with them reduced to 2; the first case was OK for more than 4 days (then I
rebooted) and the second one was OK for more than 7 days (then again I
rebooted); in both cases I was continuously running parallel jobs on them.

> I still suspect that something else is going on, and slowing down the
> machine, driver and network now hides the real problem.

Can you elaborate on the "slowing down..." part ?

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