Re: Sata controller headache

Earlier, I wrote in -current:

> I've been seeing the same kinds of errors [as Paul Bliss was having]
> with a Promise SATA300 TX4 controller and a pair of Seagate 300GB
> SATA drives.  Apparently, people have been having similar problems
> with SATA drives on Promise controllers for quite some time now, in
> both FreeBSD and Linux systems.  Lots of reports and requests for
> help, but no one so far has admitted to having a clue as to what is
> causing it.

I wanted to let people know that I managed to fix (or, at least, work
around) my problem by adjusting the BIOS settings for my (old "Slot A"
Athlon system) motherboard.  Specifically, I disabled PCI master burst
mode, and although this slowed down disk I/O significantly, it made
the instabilities w/r/t the Promise card go away completely.

I'm not sure whether the fundamental problem is flaky PCI bus design
in some motherboards, or overly picky bus expectations by Promise, but
this experience suggests to me that people who are having timeouts and
hanging errors with Promise SATA controllers might want to try playing
with the PCI-related BIOS settings on their motherboards and see if
that gives them relief.  If anyone is having trouble with a Promise
card in a recent-design motherboard with normal BIOS settings, of
course, that would strongly point to Promise as the guilty party.

Whether anything can be done to relieve this problem in the device
driver is a question I'm not in a position to answer.

Rich Wales
Palo Alto, CA, USA
richw_at_richw.org
http://www.richw.org