(no subject)
I think I've straightened out the disk corruption issue. After considerable testing, I've determined that it only happens when using both SATA channels simultaneously, which means the controller (or its driver) is at fault. The disk controller is manufactured by Silicon Image, which is the new name of CMD Technologies-manufacturer of the old CMD640 chip, which was notorious for a similar problem. (And it sounds like the same design team may still be at work there.)
It turns out that what I'm observing is a known interaction between the SiI3112 SATA controller and the nForce2 north bridge. (That page talks about Abit, but other mainboard vendors also had this problem.) In fact, my BIOS even had a setting to address it, but the documentation was in Engrish, so I didn't realize that. Selecting the most conservative value for the setting appears to have fixed things.
This does not give me a warm and fuzzy feeling about the server at work which is doing software RAID1 on the same combination of chipset and SATA controller.
It turns out that what I'm observing is a known interaction between the SiI3112 SATA controller and the nForce2 north bridge. (That page talks about Abit, but other mainboard vendors also had this problem.) In fact, my BIOS even had a setting to address it, but the documentation was in Engrish, so I didn't realize that. Selecting the most conservative value for the setting appears to have fixed things.
This does not give me a warm and fuzzy feeling about the server at work which is doing software RAID1 on the same combination of chipset and SATA controller.