Some notes on RootOnZFS article in wiki
serenity at exscape.org
Tue Dec 22 12:22:44 UTC 2009
On Dec 22, 2009, at 12:48 PM, Dag-Erling Smørgrav wrote:
> "James R. Van Artsdalen" <james-freebsd-current at jrv.org> writes:
>> Ollivier Robert <roberto at keltia.freenix.fr> writes:
>>> On modern machines, system will boot from the GPT "freebsd-boot"
>>> partition w/o having it active
>> A correctly-written PC BIOS does not even look at (or for) a partition
>> table of any sort when booting. That's been the case for a
>> quarter-century. A system that does not boot without the active bit
>> set is buggy, not new vs. old.
> Wrong, wrong, wrong.
> Some MBRs look for the active bit, some don't. It doesn't mean they're
> buggy; it's a design decision. FWIW, ours does.
> Some BIOSes *do* read the partition table; there was an issue some years
> ago with ThinkPads that froze at boot if you installed FreeBSD on them
> because they misidentified the FreeBSD partition as a suspend-to-disk
Surely this doesn't contradict his statements in any way?
He said that a correctly written *BIOS* doesn't look for at a partition table; the fact that the FreeBSD MBR does is clearly irrelevant. I took "system" to mean hardware/BIOS, by the way.
The fact that some BIOSes do read it and freeze is obviously a bug, further strengthening his argument that the BIOS shouldn't care about the partition table.
(I have this problem myself; I have two computers with nForce4 chipsets, which cannot boot from GPT; if the disk channels aren't set to "disabled" in the BIOS, it freezes on POST. You can have GPT storage disks, if you disable the channels before plugging the disks in, but it won't boot from a disk it doesn't find.)
More information about the freebsd-current