[tulip] device or resource busy

[Dan Hollis]

On Fri, 17 Nov 2000, Donald Becker wrote:
> On Fri, 17 Nov 2000, Dan Hollis wrote:
> > On Fri, 17 Nov 2000, Donald Becker wrote:
> > > I'm spending the rest of today working on and testing the Tulip driver,
> > > incorporating several of the recent bug fixes:
> > >    21143 autonegotiation when there is no EEPROM table
> > >    PNIC and PNIC-II lossage when link beat is temporarily lost (two distinct
> > >       problems)
> > ADMtek centaur tx lockups?
> When the Tx threshold is increased due to a FIFO underrun?
> I haven't been able to reproduce this problem, but I'll try again.

Some motherboards simply wont exhibit the problem. I suspect related to
BIOS initializations / northbridge differences.

5 out of 10 of our servers experienced tx lockup problems with the ADMtek
Centaur.

> My current theory is that the FIFO underrun is a symptom of a prior
> problem, and bumping up the threshold register is unrelated.

Enabling CSR18 bit 1 *eliminates* the tx lockup problem on *all* of our
servers. Just a data point.

Did you read the message from Bill Paul about ADMtek hardware flaws he ran
into while writing the FreeBSD driver?

-Dan