nforce 430, AMD64, and SATA woes...

[cheuschober]

Hi. First time gentoo user trying to set it up on my media machine but I can't seem to get as far as getting a kernel built or even unpacked.

So my woes begin with my machine.
AMD Athlon64 X2
Generic nforce 430 + 6100 based board with four nvidia sata ports.
4 HDD > 3 Samsung Spinpoint-P SATA & 1 36GB WD Raptor SATA (10000RPM - /dev/hda)

I use the Spinpoint drives in mdadm raid for storage. The Raptor is the 'system drive' that everything gets installed on. I should note I used parted to create the partitions so I know the tables are in good shape according to disk geometry and made several passes of a destructive read-write badblocks test just to confirm that the disk is solid. No bad blocks were found.

I started with the 2006.1 ~amd64 minimal install cd. Following the handbook I successfully mounted my relevant drive and its partitions (/dev/hda1 = ext2 boot, /dev/hda2 = swap, /dev/hda3 = reiserfs /)

I downloaded the stage3 tarball, did an md5 checksum, and untar'ed. No problem. Followed suit with the portage tree and had some kind of crazy disk i/o problem in the midst of the untaring that (unfortunately) spit out more error messages than the terminal buffer held so I can't be certain where the problem started. Restarted. Reformatted (just to be safe). Try again. Untar'ed both. No problems. Get to compiling the kernel another disk problem and I have a kernel panic.

Rinse and repeat.

The errors are never the same nor do they occur a the time time (sometimes its the stage untar, others the portage, others during an rsync, others during compilation) but always involve high disk use with the drive working at full. Some leave me with a blinking (crippled) terminal that can't even shut itself down. Some are full-on panics.

To troubleshoot I've removed all non-sata_nv sata modules (ahci doesn't work with the nforce 430) in case of conflicts. I've tried noapic and nolapic but then I get an irq error every operation. I've even gone so far as to use SuSe and Ubuntu x86_64 live cd's to give me a 'working' boot environment that I later chroot into. In all cases high disk read/write causes some kind of instability. Out of curiosity I tested both the SuSe and Ubuntu x86_64 environments more fully and they both proved to be just as unstable. I have, on many occasions, used a knoppix cd for pre-setup before an OS install including setting up the raid and lvm on my spinpoint drives -- I never had a lick of a problem with sata_nv and it seems that individuals running x86 kernels have (generally) good luck with the nforce 430 and sata_nv.

This leads me to believe there's something fishy going on with the ~x86_64 implementation of sata_nv -- though I fully recognize I'm not an expert on the matter. I'll stuck at this point as to what to do. If this was anything other than a media machine that could obviously benefit from the wider memory bus during decoding/encoding I would just throw an x86 on it.

Can anyone offer any suggestions? More ways to troubleshoot? Relevant information threads? Beads and glimmers of hope?

Is there anything I can do? (Code dumps or otherwise).

Any and all is appreciated.

Best,
~Chad

[weaksauce]

Hey, I have installed amd64 gentoo flavor on three different machines with no problems. I think that sata_nv works fine, though sometimes you don't turn it on in the kernel the right way and need to go back, but here, since the livecd is failing, that screams out hardware problem to me. Especially if the problem is disk i/o errors, that makes it look like the controller on your motherboard is bad, the connection to the hard drive is bad, or something. I don't think the harddrive is failing because then it would probably just not work. Like I said, I've seen this before, it scared the crap out of me and I thought my harddrive was dying, but it was the IDE controller. In your case it may be the SATA controller. I just don't know what else could cause unrepeatable disk i/o errors like that.

[cheuschober]

I'd agree if I wasn't getting perfect performance in x86 mode.

Clearly something is screwy here but if it was the hd controller I'd think I'd have problems in x86... wouldn't I?

I'm also fairly certain it's not the drive since the badblocks destructive read/write (run in x86 knoppix) with consecutive runs didn't turn up anything.

Could it possibly have something to do with specific drive / model? As a 10k rpm drive it can write/read considerably quicker and since these errors always end up happening during full-spin disc use...

I'm just plum stumped on this one.

Best,
~Chad

[weaksauce]

sorry i can't be of any more help man. I have a raptor working fine, and I did have problems with my friends' raptor, though not like yours. On his, the live cd and then actual kernel would boot up incredibly slowly, and then when we turned of apic in his mobo options, it worked fine. On my new motherboard, i had to give the linux kernel option apic=off just to get the damn thing to boot up. Try combos of kernel option and motherboard option, but if it doesn't work i'm stumped. Sorry.