boot problem on snap 520 platform.

[etherfish]

Trying to help another member, I started replying in another thread. Unfortunately, my reply quickly became a request for assistance all of it's own. I'll repost it's contents here:


I've got a Snap 520 with four 1TB drives. Specifically, they're ST31000340AS. Unfortunately, we've been frustated that it will hang on boot while running the Broadcom HT-1000 option rom. Since the machine uses a usb flash drive for it's root filesystem, the crude workaround we've devised is to pull all four drives out about half a centimeter - enough to disconnect them form the backplane until we reach the linux grub screen. At the grub screen, we reinsert the drives and then boot the kernel. The kernel has no problem detecting the drives. We're also not concerned about inserting the drives while the system is live - previous inspection revelead that the the snap 520 actually has a hotplug SAS backplane.

With regards to the hang on boot, I would be keenly interested to know if your snap hangs at the same place in the boot process. To reduce the noise of the fans, I upgraded the bios on our snap from the one it had to the most recent, generic supermicro bios, specifically dspr6306.rom dated 6/30/2006. I was concerned that i'd broken my snap with the new bios. Does your snap hang if the ht-1000 sata is set to "RAID" mode in the bios instead of IDE or MMIO? Our will in a way that requires clearing the cmos and I can't recall if it worked before the bios upgrade either.

One last clue regarding the boot failure. The machine had no issues booting when the drives were first installed new and before a partition table had been written to them. I'm beginning to suspect a bug in the ht-1000's option rom master boot record parser. One of these days I mean to migrate the disks to GPT hoping it will avoid the bug and allow the machine to boot normally.

[etherfish]

I'm replying to myself, in essence, to reply to blue68f100's response to my post in the other thread:

>I do not know of any one that is not using the factory release BIOS. The
>hardware has a special chip that identifies it as Snap Hardware. the GOS checks
>for this. So your ROM is probably the problem. I do not know if your bios allow you
>to monitor the boot process with a null modem cable off com1 with a terminal. >This is how I check things out when I have a problem and can not access the gui.

In our 520, the onboard VGA does not have a port to attach a VGA cable so we moved a jumper to disable the onboard VGA card - a rage128 - and installed a regular PCI video card. With a usb keyboard, we were able to use the machine as a regular PC. This is how we watch the boot process, modified the BIOS settings. We also upgraded the RAM to 4 GB.

For what it's worth, we originally used an IDE CD-ROM connected to the motherboard's IDE header to install Solaris 11. After encountering problems with the HT1000 SATA controller, we PXE booted a debian install using the onboard NICs.

Has anyone else done anything like this?

To be frank, the hardware so far has done a fantastic job. We're currently using the onboard Ultra320 SCSI to run a set of 14 10K 73G drives in a Dell PowerVault 220S rack.

The curious part is that the Broadcom BIOS for running the HT1000 did not hang on the drives before we partitioned them. If it really is a BIOS bug, I'm surprised that SuperMicro hasn't released an update.

I suspect that GOS must be checking the DMI tables or similar. Unfortunately, I cannot check as I didn't consider the old bios image worth keeping around after I was happy the new BIOS worked well.