Comment 75 for bug 285892

Ok, after biting the bullet, I called it a day again... everything I wrote before about fglrx, google-chrome, flash player is crap... meaning that it was a huge coincidence... Forget also BIOS settings, RAID, AHCI, and stuff like that, coz this is only in the kernel...

I have dumped Ubuntu 10.04 and moved on to Archlinux. You know, with this one, you need to dig things deep... very deep... I'm not saying that Ubuntu is not good... no, in fact, is the best distro around... but I was kinda missing the do-yourself approach for a long time now.... but I still have a Ubuntu live pendrive for other tasks.

Turns out that this bug is also present in Archlinux current kernel. And it manifests this way:

ata3: softreset failed (device not ready)
ata3: applying SB600 PMP SRST workaround and retrying
ata3: SATA link up 3.0 Gbps (SStatus 123 SControl 300)

This bug is surely related to the AMD SB600/700 controller mixed with certain hard disks (Samsung, mine).

Like I said some previous posts... something to do with the kernel configuration CONFIG_SATA_PMP, some say that if you turn it off, this bug stops. I have yet to test this in Arch. You know, it's hard to compile a kernel in Ubuntu, following the traditional way and at the same time keeping all things together without breaking the package manager or something. In a similar fashion you need to be careful in Arch too, even though you can easily compile the traditional way, you need to properly create kernel headers package, if you, like me, use the fglrx beast, because it wants to compile a new module every new installed kernel.

But if there is a way to avoid all this kernel mambo-jambo, I am gonna try it first:
I have turned off ncq in my system by adding this line to GRUB kernel paramenters:

libata.force=noncq

I am currently testing this with no errors so far, next time I come here, it will be with another parameter.
In the meanwhile, CONFIG_SATA_PMP may be your next adventure...