[ale] Possible errors on Hard Drive

[Greg Freemyer]

On Dec 24, 2007 2:04 PM, Scott Castaline <hscast at charter.net> wrote:
>
> Greg Freemyer wrote:
> > Scott,
> >
> > I don't know enough to interpret all those code values, but I see that
> > you just moved from Sata-150 to Sata-300 and amongst the output you
> > have:  HSM Violation.
> >
> > Sata-150 (Sata 1) does not support NCQ.
> >
> > Sata-300 (Sata 2) does and is enabled by default for Linux.
> >
> > Unfortunately all of the stable Linux kernels have bugs in the NCQ
> > implementation and report HSM violations under normal conditions if
> > NCQ is in use.
> >
> > You can disable NCQ (how?) or you can ignore the errors.
> >
> > The latest 2.6.24-rc5 (iirc) has eliminated the HSM violation noise,
> > but I don't know how stable it is yet.  They put in a lot of ACPI
> > power saving stuff in 2.6.24.  I'm not so sure its going to be a good
> > kernel.  Not sure if only laptops are affected by ACPI, or everything.
> >
> > 2.6.22 and 2.6.23 are supposed to get updates to the stable versions
> > to eliminate the HSM violation noise.  You may be better off waiting
> > for one of those to get the cleanup patch.
> >
> > Hope that helps.
> >
> > Greg
> >
> > On Dec 21, 2007 9:30 PM, Scott Castaline <hscast at charter.net> wrote:
> >
> >> Checking my logs and noticed that I'm suddenly getting what looks like
> >> drive errors but not sure. I've noticed the following repeated about 3
> >> times every once in a while:
> >> ata4.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2
> >> ata4.00: port_status 0x20080000
> >> ata4.00: cmd c8/00:30:ef:e8:cb/00:00:00:00:00/e6 tag 0 cdb 0x0 data 24576 in
> >>          res 50/00:00:1e:e9:cb/00:00:00:00:00/e6 Emask 0x2 (HSM violation)
> >> ata4: soft resetting port
> >> ata4: SATA link up 3.0 Gbps (SStatus 123 SControl 300)
> >> ata4.00: configured for UDMA/133
> >> ata4: EH complete
> >>
> >> followed by:
> >> sd 3:0:0:0: [sdd] 976773168 512-byte hardware sectors (500108 MB)
> >> sd 3:0:0:0: [sdd] Write Protect is off
> >> sd 3:0:0:0: [sdd] Write cache: enabled, read cache: enabled, doesn't
> >> support DPO or FUA
> >>
> >> I just recently changed HA from an Adaptec 2 port SATA 150 to a Promise
> >> SATA 300TX4, 4 port SATA 300. I have all WD drives.
> >> port 1 = WD1600  (150)
> >> port 2 = WD5000  (300)
> >> port 3 = WD360    (150)  Raptor 10K rpm
> >> port 4 = WD360    (150)  Raptor 10K rpm
> >>
> >> Any ideas? Everything I've found is close, but no cigar. Also any
> >> documentation I've found is like from 2004. I'm not too sure how current
> >> that is.
> >> _______________________________________________
> >> Ale mailing list
> >> Ale at ale.org
> >> http://www.ale.org/mailman/listinfo/ale
> >>
> >>
> >
> >
> >
> >
> Yeh I do have a mix of 150 and 300 drives. The problem is with the 300
> drive though. I do remember reading somewhere of a bug in the NCQ part
> of the kernel and also an issue with Promise boards in particular. That
> was some time ago though and I thought that it was already fixed. Based
> upon what you are saying it sounds like it hasn't. I did notice that the
> problem seemed to have started at the time I added the 4 port card
> replacing a 2 port 150 card.
>

NCQ support itself was just added in the 2.6.20 timeframe (or newer?).

NCQ is a Sata-2 feature.  I'm not sure but I think Sata-2 features are
often only negotiated if a Sata-300 speed can be negotiated.

Anyway, that spurious HSM violation has been around since NCQ went
into the kernel.  It is only in the last few weeks that it was
determined to be spurious and a patch was sent to Linus to remove it.
ie. 2.6.24-rc5 I think.

And I think that patch is being backported to the 2.6.23 and 2.6.22 kernels.

Greg
-- 
Greg Freemyer
Litigation Triage Solutions Specialist
http://www.linkedin.com/in/gregfreemyer
First 99 Days Litigation White Paper -
http://www.norcrossgroup.com/forms/whitepapers/99%20Days%20whitepaper.pdf

The Norcross Group
The Intersection of Evidence & Technology
http://www.norcrossgroup.com