System:
Fedora Core 2, using kernel 2.6.8.1 which I compiled myself
Opteron242x2
Iwill DK8X
Silicon Image 3114 SATA controller (integrated)
40GB Western Digital (/dev/hda, /)
80GB Western Digital (/dev/hdb)
200GB Seagate SATA (/dev/sda1)
200GB Seagate SATA (/dev/sdb1)

Error:
Sep 13 21:05:27 localhost kernel: ata2: command 0x35 timeout, stat 0xd8 host_stat 0x61
Sep 13 21:05:27 localhost kernel: scsi1: ERROR on channel 0, id 0, lun 0, CDB: 0x2a 00 0b da 19 9f 00 04 00 00
Sep 13 21:05:27 localhost kernel: Current sdb: sense = 70 3
Sep 13 21:05:27 localhost kernel: ASC= c ASCQ= 2
Sep 13 21:05:27 localhost kernel: Raw sense data:0x70 0x00 0x03 0x00 0x00 0x00 0x00 0x06 0x00 0x00 0x00 0x00 0x0c 0x02
Sep 13 21:05:27 localhost kernel: end_request: I/O error, dev sdb, sector 198842783
Sep 13 21:05:27 localhost kernel: ATA: abnormal status 0xD8 on port 0xF882CCC7
Sep 13 21:05:27 localhost last message repeated 2 times

Error occurs when writing large files (~2GB+) to any of the two SATA Seagate drives. Reading off the drives is fine. First noticed the errors when transferring data from a Windows computer via Samba, but it turns out that writing large files in general causes this. Following that, the mount becomes unusable until the computer is rebooted.
The controller is using the sata_sil module. Googling turned up the following:

lists.suse.com/archive/suse-linux-e/2004-Jun/2331.html
ussg.iu.edu/hypermail/linux/kernel/0405.1/0533.html
slashdot.org/comments.pl?sid=100115&threshold=1&commentsort=0&tid=106&mode=thread&pid=8532018#8533725

(forum won't let me post urls until 5 posts...)
So far I've found no solutions from anyone. It seems that the siimage approach may work though.
How would I go about telling FC2 to use the older siimage module instead of the default sata_sil?
Or if someone knows a way to fix this, that would be much appreciated.

I'm having the same exact problem, except with the nv version of the sata driver (even though i have the sil 3114 controller too). Except that it doesn't happen to me when writing large files. It just, well, happens. Sometimes it happens while booting up at random times, sometimes I can log in, and it'll die when I start running apt-get.

I've heard hints that the sata support is bugged right now, so I think I'm going to try a different kernel version (2.6.5).

However, I didn't have this problem when mounting the drive as /dev/hde instead of /dev/sda. The only problem is that my current kernel doesn't recognized /dev/hde as a valid block device, but the installer's kernel did.

So, I'm thinking that for now, you need to mount the sata drive as a normal ata drive using the device /dev/hde (or what have you - hdf, hdg, hdh, etc), instead of /dev/sda (or sdb, sdc, etc).


I just don't know how to make /dev/hde a valid block device.

I believe I have found a solution to this problem, so here goes.
marc.theaimsgroup.com/?l=linux-kernel&m=108751565331668&w=2 (another url)

Basically, the pair of seagate 200GB SATA drives don't play nice with the SiImage 3114 controller without a small fix. It turns out there is a listing of all the drives in the sata_sil.c file which need the fix, and being a semi-recent drive, my 200 wasn't added. Adding the drive seems to have fixed the problem.

If your drives are seagate you might want to check out the blacklist. However, if its using the nv driver I don't think I may be of much help. My best advice would be to do some digging on the kernel.org mailing list as well.

Add noapic to the append line for lilo or in the kernel line for grub. If that does not work, buy either IBM/Hitachi or Western Digital hard drives. Seagate is never good for IDE or SATA because Seagate use those interfaces as their test bed for their SCSI series. Also Silicon Image controllers are not well programmed for Linux. I suggest controllers form Highpoint, Promise, 3ware, and many others. Though I prefer to use Highpoint because they have open source drivers that should work in kernel version 2.6.x.