> > Is it generally feasible to pass either argument or environment
> > strings to the kernel?  (I know some systems support this.)
> 
> this rubs me in an interesting way:
> 
> i think i'd like to see the boot program/locore changed to support
> this.  it can be useful; for instance, unless i'm mistaken, you
> can boot BSDI systems with '-autodebug', and it'll print autoconfig
> debugging information...
> 
> now, i don't know if that flag is handled in the bootstrap, or in the
> kernel, but something like that would be a good idea, and i think the
> more general it is, the better...
> 
> 
> cgd

well this also fits with the machine-dependent hints problem.
Basically, the machine-independent console stuff, and the
nfs_diskless_chose_which_interface code both want hints as to what
the hardware or bootstrapping software wants to do.

re: console problem.  some hardware has its own sense of what the
console is setup to be.  the machine-independent cnsole stuff makes
its own decision w/o knowledge of this.  If hints were available, then
maybe it could do the right thing.  This problem is half of why the
sparc code doesn't use sys/dev/cons.c but its own hack.

re: nfsdiskless.  my new code does reverse arps to find out the addr
of an interface.  however it may chose a different interface than was
used by the bootprom/bootstrapping program because it just doesn't
know.

later,
Adam





------------------------------------------------------------------------------