Solaris 10 10/08 x86 (Ultra 20 M2) install issues

Question

Morning everyone,
After backing my workstation up, I performed a full update of a fully patched (OS and BIOS) 11/06 Solaris 10 Ultra 20 M2 machine (Opteron 1210). However, shortly after installing I get repeated start failures, and even rebooting off of the 10/08 DVD gives me the exact same error. The error, which occurs right after detection of
ng0 and ng1, states (my comments are in <>):
WARNING: /PCI@0,0/PCI108e,534d@5/disk@0,0 (sd0):
Error for command: read(10) Error Level: Retryable
Requested Block: <number always varies. i.e. 138462966) Error Block:<same as requested block>
Vendor: ATA Serial Number: <nothing>
Sense Key: Aborted Command
ASC: 0x0 (no additional sense info), ASCQ: 0x0, FRU 0x0
Now, after first installed I got this error, but several reboot attempts got me into the working system, and I rebooted several times again and didn't see the error. I performed my regular restore (/export/home, /opt, ect..) and updated the system from Update Manager. However from that point forward, the machine will not get past this error. Now even performing a full re-install of 10/08 lands me with the same exact error.
The md5 checksum of the iso image checks out, and I've written/verified the data on the DVD with two different burning programs. I've even written the information to two different brands of DVD-R will no luck. The slices are standard UFS, and it's a plain Sun Workstation with 4G of memory.
Suggestions as to what the error is referring to? I've Googled and searched this forum in vain to see any thing that might relate, but no luck.
Dave

Answer 1

An update for anyone who might run into a similar issue:
I narrowed down to an issue with the ata controller and 10/08 ( 05/08 works like a charm). In the release notes, the following entry appears:
"x86: ata Timeouts During Boot (6586621)
ata driver timeouts might occur during system boot on Intel multiprocessor systems. These timeouts occur when the root device is on a drive with the HBA controller bound to the legacy ata driver. These timeouts lead to a momentary hang, hard hang, or a panic during system boot with console messages similar to the following:"
So, even though this machine is running an AMD 1210 Opteron (perhaps the entry should read: ".. on x86 multiprocessor machines..." instead), I am affected by this bug.
I was able to successfully load u6 on the machine once, and once that happened 10/08 would fail with that ugly error message. However, by disabling the MP support in Solaris (in kbd mode so that Solaris only sees a single processor), it booted just fine.
Once I got into the working system (I assume I could have done it from a live CD like milax) I disabled the microcode support by following the instructions listed in Workaround 5 (mv/renaming the microcode entry). After multiple warm and cold restarts to test the functionality of the fix, the machine passed with flying colors.
Lastly, the other option, listed as Workaround 1, was to specify in the BIOS the use of AHCI mode. Unfortunately, the 1.52 BIOS does not offer this feature.
Dave
Edited by: WeatherDave on Nov 13, 2008 3:00 PM