abs@netbsd.org (David Brownlee) writes:
> 	This is definitely good stuff - any objection to it being added?

(1) the relevant bits in MD code need to be removed from all ports;
the patch is not good as-is.

(2) it should be defopted.

(3) it should probably be a patchable variable, initted by an #ifdef,
rather than just an #ifdef.  In general, this is true of most kernel
configuration bits.

(4) it's kind-of losing.  it's quite reasonable to boot an install
kernel with the boot flag that sets RB_ASKNAME, then select an
alternate root from which to boot.  You may wish to boot single-user,
or you may in fact wish to boot multi-user, in that case, and it
shouldn't be forced by a value hard-wired into the kernel (rather, it
should be controlled by the command line).


If this is going to be a flag, then it should probably:

	(1) be 'md'-specific, i.e. hack all of the MD 'md' code
	which sets RB_SINGLE, and

	(2) have its sense reversed, so that existing configs
	continue to work properly.


Thinking about it more, it's not obvious to me that we really need to
worry about setting RB_SINGLE for most miniroot boots anyway -- the
'small' init used doesn't even support multi-user, IIRC.  somebody
should check that, verify that it's being used everyplace, and GC that
RB_SINGLE-setting code entirely (or make it an option if there's
really some desire to support it).


cgd