Motherboard RAID problem

[Bill Blue]

I'm not sure if I'm expecting too much, or this is a real bug.

Using FreeBSD 6.1 release, CVSup'd to current. The motherboard is a Supermicro P4SCT0 with a 3.2Ghz P4 and 2 DDR400 1G sticks of RAM.  On the MB is a built-in RAID controller (Adaptec chip) for the SATA drives.  You set it for discrete SATA or RAID.  If RAID is set, on the next boot you have essentially a BIOS configuration for that 'device' consisting of the two SATA devices in either RAID 0 (striped) or RAID 1 (mirrored).

I already had a functional SATA boot drive (first SATA device) which shows up on the OS as device ad4s1a-f.  I added a second identical drive as the second SATA device, configured the BIOS setting for SATA RAID, rebooted, set up an array of RAID 1 with the first drive as the master.  The second device was then built up as a mirror of the first and everything looked normal.

Boot the OS now and all goes well with the device still showing up on /dev/ad4* but I couldn't tell if the mirroring was really working since the drives have no individual led indications.  I then noticed that there was a new ad6* device, and guess what -- it was the second SATA drive and a mirror image of the *original* first drive.  Watching it with DF for size changes when copying a large file to my home directory, it didn't change at all.

ad6* were the only new devices seen in the OS.

Went back to the BIOS which displayed that instead of two SATA drives there was one RAID unit as far as it's concerned.

So it appears as though FreeBSD doesn't have the right drivers to actually see the RAID volume built by the BIOS and considers its members to be individual devices.

Does this ring any bells with anyone?  Suggestions?


A separate and semi-related question, this system also has 16 other drives with two 3Ware 9500SML controllers and 8x500 and 8x750GB arrays.  It's managed remotely with X.  In this type of application, is Hyperthreading best on or off?

Of course for just reading files it's not a big issue, but receiving large files seems to be largely limited by processing interrupts from the 1G ether.  They shoot up to 50% and stay pretty consistent regardless of variations in the incoming stream.  If Hyperthreading was off, wouldn't there be better performance inbound?  Or would that be a tradeoff when the entire picture is taken into account?

Thanks for any insights.

--Bill