Re: utku's sata freeze and hard reset problem

[Tejun Heo <tj@kernel.org>]

Utku Somer wrote:
> Tejun Heo wrote:
>> Tiemouts without any link layer problem reported.  Hmmm.... Does it
>> help if you disable NCQ with the following command?
>>
>> # echo 1 > /sys/block/sdX/device/queue_depth
>>
>> Thanks
> I had the same idea after I read around the mailing list. (But I was
> ignorant
> on the way, I actually tried to echo 0 into queue_depth first. Good
> thing it doesn't
> let me.) Anyway, it has been running without the errors for more than 12
> hours
> now. I also tried by increasing the disk load, which increases the
> chances of the problem
> occurring. It has not happened yet.
> 
> So yes, disabling NCQ stops the problem from occurring.
> 
> Any ideas?

Timeouts w/ NCQ seem to be one of the most subtle failure modes.  It
can be caused by several different issues - firmware bug, power
problem or link layer problem.  Is there a firmware update available
for your drive?  If there's one, please save the output of "hdparm -I"
before going forward with firmware update.

Thanks.

-- 
tejun