Re: [gentoo-user] Kernel2.6.33: ATA failed command: READ FPDMA QUEUED, hard resetting link

[Paul Hartman <paul.hartman+gentoo@gmail.com>]

On Fri, Mar 26, 2010 at 10:39 PM, Stroller
<stroller@stellar.eclipse.co.uk> wrote:
>
> On 26 Mar 2010, at 22:21, Paul Hartman wrote:
>>
>> ...
>> Well,  I was thinking more about something like alteriong IOH/ICH
>> voltage, or whichever voltage powers the SATA controllers. It has 12
>> SATA headers on this motherboard but I don't know how much it can
>> realistically handle at once. Hopefully all of my disks. :)
>
> Honestly, I'm not sure that I'd do that.
>
> If I thought it was a power issue - and that seems quite a reasonable
> possibility - I would replace the PSU first.
>
> What PSU are you using at the moment? Brand / wattage?
>
> I think these are probably overkill:
>
> http://www.bluepoint.net/AKAPSU066
> http://www.bluepoint.net/AKAPSU071
>
> However, if you're using a 450W PSU at the moment, you can get an OCZ
> branded 600W for less than £50. I tend to be suspicious of cheap unbranded
> and Wong Fu 350W - 450W PSUs. Often a 350W PSU from a manufacturer with a
> half-decent brand (eg Trust) will be better than a no-name 450W PSU.
>
> It seems important not only the actual wattage that the PSU gives out
> (measured with an analogue-needle multimeter) but also how stable that
> voltage is, how smooth, consistent and reliable it is. Unless your PSU is
> absolute top-notch quality, try to under-utilise it.
>
> With 6 x hard-drives, I would be generous with supplying power - I think a
> 600W PSU would easily be justified.
>
> Stroller.
>

I have a 750W Corsair 750TX which should be plenty (if not overkill).
Seems to have all the bells and whistles (though I wish I had bought
the modular version). I definitely agree about the cheap no-name PSUs,
I've seen some USD$10 ones go up in smoke after just a few minutes of
use.

Based on the little info I could find Googling it seems the most
likely explanations are:

IRQ handling weirdness (but that was more probable in older
motherboards and kernels)
ahci driver issues (I will reboot with the controllers in "legacy"
mode to use the non-ahci drivers and see what happens)
a bad drive, bad cable, bad controller, bad sector, something physically bad.

Overnight I copied several hundred gigs of data (and still going) and
the error has not been reproduced. Before installing, I did a full
SMART test (a few hours) and a badblocks read-only test (even more
hours) and neither showed any errors. I did not do a full read/write
test, though.

I haven't heard any unexpected noises coming from the drives (no
clicks of doom), though I have so many fans going it may be hard to
hear.

Thanks.