Re: Sata controller headache

[Rich Wales]

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
ri...@richw.org
http://www.richw.org
_______________________________________________
freebsd...@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-curre...@freebsd.org"

[Søren Schmidt]

Rich Wales wrote:
> 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.
>

As I told you (in private mail) some of the Promise chips does get close
to the edges of the PCI spec, but usually its only a problem on
motherboards that haunted as well with bugs in that area.

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

Thats a bridge I'd rather not cross in this life :)

-Søren

[Paul Bliss]

Thanks again for the lead, Rich. I've been unable to find the comparable
setting in my bios, though. I wonder if the burts mode specifically was
the issue, or if just the general slowing of the IO was what did the trick
for you.

FYI I've heard that the nvidia sata, that is onboard with so many new
boards, doesn't have this problem.

I think I may just go with IDE. For the amount that I've shelled for
controllers that haven't worked out...
pleh.

But, seriously, thanks for everyone who helped out.
-Paul

On Wed, 11 Oct 2006, Rich Wales wrote:

> 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
> ri...@richw.org
> http://www.richw.org