Upgraded 2 of the 4 1TB Samsungs to 2TB WD EARS drives under embedded-0.7.1.5127.img (32 or 64 bit versions) and writing to the 1.8TB ext2 partitions beyond a certain (but predictable) point on a volume just under half full will trigger a critical system reboot.
There appears to be a serious bug involving the two WD EARS drives since all 4 disks (not setup as a RAID) are flagged dirty and will only mount after they've been repaired with FSCK. Strangely, FSCK itself seems to cope with the 2TB WD EARS drives ok.
Observing the console, I saw error messages relating to allocation units appear just before it rebooted. These looked similar to errors thrown up by simply selecting all files and folders in the root of one of these two drives from a win2k client and clicking on the properties option. Although this doesn't crash FreeNAS, there is a definite shortfall in the number of files, folders and bytes reported versus reality.
I am now using Knoppix 5.3 as a temporary replacement OS on the server box whilst this serious bug is sorted out. I mention this fact to highlight that Knoppix has no issue with these drives and the Asrock "AliveNF7G-FullHD" motherboard and the Sempron LE-1250 (2.2GHz rated but clocked at 1.2GHz) CPU that I've been using for the past 12 months or so.
I'm sorry I can't supply any more detail on the errors since the box is now running a Knoppix Live CD session (DVD version on an 8GB flash drive!) and there's no way I'm going to push my luck with those EARS drives which now have over 1TB's worth of data stored on each one.
In case you're wondering, I am well aware of the 4K sector alignment issue with these drives and, under a knoppix live session, used fdisk to start the partition on sector 64 before copying some 900GB's worth across from each of the 1TB Samsungs they were replacing. I use Ext2 since this is a FS that can be used by OpenBSD, Linux and win2k (using the Ext2 4 Win driver) and I don't want to be held to ransom by the use of FreeNAS's "native FS". FreeNAS has coped very nicely with Ext2 on the 1TB Samsung drives over the last 12 months or so and I'm hoping this will (bugs aside) eventually prove to hold true when the next revision of FreeNAS becomes available.
I've since changed from using Ext2 to using UFS and the problem has totally disappeared. The bug would appear to be a wraparound issue at the 1Tb mark in the Ext2 driver.