Sil3112 and Seagate ST3160023AS

Sil3112 and Seagate ST3160023AS

[Julien Langer]

Hi,

I know, Sil SATA controllers and Seagate disks are a bad combination.
However my Seagate drive worked flawlessly for about half a year and
hdparm -t tells my, it's working with approx. 50 mb/s.

I just switched from 2.6.7 to 2.6.10RC3 yesterday and noticed that my
drive now only works with 15 mb/s. Also I get a new message "applying
pessimistic Seagate errata fix" when booting, which wasn't there with
2.6.7.

Is there a way to disable this fix, which slows down my drive, since it
worked fine for a long time without this fix on older kernel versions?
I'm using the deprecated ide driver for the sil controller, not libata.

Thanks in advance,
Julien Langer

PS: Please CC me, I'm not on the list

Re: Sil3112 and Seagate ST3160023AS

[Jeff Garzik]

Julien Langer wrote:
> Is there a way to disable this fix, which slows down my drive, since it
> worked fine for a long time without this fix on older kernel versions?
> I'm using the deprecated ide driver for the sil controller, not libata.


Unfortunately it's just a matter of time until you hit a problem, 
without the errata fix that causes the performance loss.

	Jeff

Re: Sil3112 and Seagate ST3160023AS

[Måns Rullgård]

Jeff Garzik <jgarzik@pobox.com> writes:

> Julien Langer wrote:
>> Is there a way to disable this fix, which slows down my drive, since it
>> worked fine for a long time without this fix on older kernel versions?
>> I'm using the deprecated ide driver for the sil controller, not libata.
>
> Unfortunately it's just a matter of time until you hit a problem,

How long a time would that be a matter of?  I've been using two
ST3160827AS drives (firmware 3.00 and 3.03) connected to an SiI3114
controller using the libata driver (kernel 2.6.9) for a few weeks, at
full speed, without noticing anything out of the ordinary.  I can't
say I have any heavy load, but I have done some moving around of huge
files, and the usual compilation jobs.

The drives are both new.  Is there a chance Seagate have fixed the
problems?

-- 
Måns Rullgård
mru@inprovide.com

Re: Sil3112 and Seagate ST3160023AS

[Eric Wong]

[-- Attachment #1: Type: text/plain, Size: 1562 bytes --]

Jeff Garzik <jgarzik@pobox.com> wrote:
> Julien Langer wrote:
> >Is there a way to disable this fix, which slows down my drive, since it
> >worked fine for a long time without this fix on older kernel versions?
> >I'm using the deprecated ide driver for the sil controller, not libata.
> 
> Unfortunately it's just a matter of time until you hit a problem, 
> without the errata fix that causes the performance loss.
 
It's been 11 months with my ST3160023AS 3.05 and SiI 3112 (rev 02) under
fairly heavy use and I haven't noticed anything wrong.

I think I've read somewhere that rev 2 of the SiI 3112 is safe, and that
might be why I'm alright.  Unfortunately, the sata_sil blacklist
implementation I wrote at the time doesn't seem to account for the
revision of either the drive nor the controller.

My experiences: (purely anecdotal evidence, ymmv)

Stability has been nothing but solid, the box they're on is on 24/7 as
an NFS server housing mainly FLAC audio and multiple Arch archives and
a build daemon (which means revision libraries and working trees exist
too)

Neither Arch archives/working trees/revision libraries nor my FLAC audio
collection has shown any inconsistency (but then again MD5 used by both
Arch and FLAC has been proven broken lately).  I'll fix up an Arch
script to check the SHA1 and the MD5 checksums sometime this weekend
(the newer commits are double checksummed).  Arch revision library
consistency is checked by tla via inode signatures, and those haven't
burped on me, either.

-- 
Eric Wong

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]