I tried booting with *both* hard drives disconnected, to test whether the (software) RAID filesystem present on them was causing my problems. The first time I tried it (no break=top, no "modprobe piix"), it booted up! Always skeptical, I tried it again, and it failed, just as before. I did try the piix thing, but to no avail.
So, from this I deduce:
1. It's not md0 (RAID) related, and in fact it's not related to disk contents at all.
2. It's intermittent. Either it's a race, or possibly its affected by previous boots of the machine (i.e., hardware state retained across warm boots). Pretty sure it's a race.
3. At least on my machine, the piix fix is not absolutely required for a successful boot, as I've seen it work without.
I *did* see some ATA error messages in the log, during the failed attempts. Something about bad status (0x7F). I suppose that that could just be a symptom of the disconnected hds, or perhaps a separate issue that the piix thing would fix.
Notably, there seems to be a pause of maybe 20 seconds or so while it seems to be looking for devices. I'm guessing that this is a period during which the SCSI controller is looking for devices, but I'm not sure that this is so.
I tried booting with *both* hard drives disconnected, to test whether the (software) RAID filesystem present on them was causing my problems. The first time I tried it (no break=top, no "modprobe piix"), it booted up! Always skeptical, I tried it again, and it failed, just as before. I did try the piix thing, but to no avail.
So, from this I deduce:
1. It's not md0 (RAID) related, and in fact it's not related to disk contents at all.
2. It's intermittent. Either it's a race, or possibly its affected by previous boots of the machine (i.e., hardware state retained across warm boots). Pretty sure it's a race.
3. At least on my machine, the piix fix is not absolutely required for a successful boot, as I've seen it work without.
I *did* see some ATA error messages in the log, during the failed attempts. Something about bad status (0x7F). I suppose that that could just be a symptom of the disconnected hds, or perhaps a separate issue that the piix thing would fix.
Notably, there seems to be a pause of maybe 20 seconds or so while it seems to be looking for devices. I'm guessing that this is a period during which the SCSI controller is looking for devices, but I'm not sure that this is so.