Ongoing SATA woes

[Giles Orr]

My apologies once again if this is double-posted.  gmail and/or tlug
doesn't seem to like it when I post twice on the same subject(?).
Please let me know if this post has seen the light of day before.

---------- Forwarded message ----------
From: Giles Orr <gilesorr-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org>
Date: Oct 3, 2007 7:16 PM
Subject: Re: [TLUG]: First SATA drive - not working
To: tlug-lxSQFCZeNF4 at public.gmane.org

On 10/1/07, Lennart Sorensen <lsorense-1wCw9BSqJbv44Nm34jS7GywD8/FfD2ys at public.gmane.org> wrote:
> On Mon, Oct 01, 2007 at 10:06:51AM -0400, Giles Orr wrote:
> > Debian testing, kernel 2.6.21-2-k7.  It's auto-loading the sata_via
> > kernel module.
[SNIP]
> > This is really making me think that I should buy a SATA PCI card ...
> > That would allow me to use SATA drives in any of my other computers,
> > so it might solve not only the current problem but potentially some
> > future ones as well.  Does that make sense?
>
> It does make some sense if the onboard controller is really that much of
> a pain.
>
> What model is the drive again?

The hard drive is a Western Digital 500 Gb.  WD5000AAKS, WD Caviar SE16.

I bought a Bytecc Serial ATA PCI card yesterday on College, with one
internal and one external connector for $20.  I chose it because it
said "Silicon Image Sil3112a" right on the box, and that's a supported
chip.  And it worked, first try - the "sata_sil" kernel module was
autoloaded.

But the problems continue.  When copying my backup from an external
USB hard drive to this new SATA drive, I occasionally get the
following error:

  EXT3-fs error (device sda6): ext3_new_block: Allocating
  block in system zone - blocks from 42500102, length 1.

Not surprisingly, this completely munges the file system (which is
indeed ext3).  fsck spits millions of errors.  So I reformatted the
HD, I did two complete rounds of read/write bad blocks checking.  The
disk seems to be fine.  But big copies still cause it to spit this
same error.  Google suggests that this isn't a particularly common
error, but it seems to be associated with Silicon Image SATA
controllers.  Worse, I found this post:
http://lkml.org/lkml/2005/3/12/65 - same error, two years ago.
Apparently it hasn't been dealt with, or at least not completely.

I am at this point seriously considering buying a 500 Gb IDE drive,
maybe use the SATA in an external case.  Except with my current
success rate, there will be problems with the IDE too.

P.S. One theory to consider: is it possible that disabling the  (not
working) motherboard SATA and/or not loading the kernel "sata_via"
module would stop the issues?  I'm tired enough of grasping at straws
that I haven't tried this yet.

--
Giles
http://www.gilesorr.com/
gilesorr-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists