I have included the whole text of this message because there is so much useful
info concentrated here, I figured it wouldn't hurt
to have it exposed to the list again.
I only wanted to comment on a couple of things. First, of course when I said
"boot partition" I wasn't referring to a /boot
partition, I meant the Apple_boot partition. This is a very confusing topic
for newbies to powerpc arch. In the man page for
bootstrap(8), there is a whole big section on OLD WORLD MACs. That would imply
that it has some use on these machines. But
according to this, and I believe it's true, yaboot isn't any use on these OWM
machines, only quik and ~shiver~ miboot. [Don't
forget BootX.] Logic classes in college ~:^) teach me that the conclusion
therefore is that bootstrap partitions aren't useful on
OWM, and therefore should there be any mention in the man page about OWMs
execpt to say "don't bother, use quik on your linux
partition with the kernel?"
You mention that I should stop whining and volunteer. I hear that. Where do I
sign up? I don't get a rash writing
documentation. The install documention is _really_ confusing on powermacs, and
it is now clear to me why, but I definitely
wouldn't mind taking a whack at fixing the install/boot parts of the docs. Be
prepared to proof read!
I do have one whine about yaboot, quik, and mkofboot, which is that they all
should add a testing option, like -t in lilo (I
apologise for having to use lilo as an example). This is very helpful when
trying to figure out if the options/configuration you
have specified is messed up or not, and is VERY helpful when you have changed
things and neglected to edit your config file! A
mistake in there parameters can sometimes be catastrofic and so therefore I am
suggesting that this is a very useful addition to
these programs. Again, I volunteer if needed, but it probably would go a lot
faster by someone who is more familiar with these
programs.
For anyone who cares, I am currently using the BootX approach where it starts
to boot macos, then bootx takes over and it boots
linux. Sometimes called the "BootX extentions" method. Kinda like if you put
linload in your windows startup directory. OK, I
know there isn't necessarily a lot of Windohs users on this list ~:^) It was
easy to set up and fairly flexible when it comes to
adding new kernels. I wrote a pathetically simple shell script to copy new
kernels over to the right place in the macos
partition. Of course, bootx is pretty useless if you want to wipe macos off
your system.
Cheers,
a
Ethan Benson wrote:
> On Sun, Jan 07, 2001 at 06:06:41PM -0800, Andrew Sharp wrote:
> > >
> > > where /dev/sda2 is your root partition. Perform this step after
> > > running Make-Linux-Bootable-Directly-From-Hard-Disk.
> >
> > um, root partition or boot partition?
>
> root partition. /boot partitions are a bad idea in general on powerpc
> (and unecessary) bootstrap partitions as in newworld setups are not
> usable on oldworld.
>
> you can also use partition 0
>
> nvsetev boot-device "$(ofpath /dev/sda)0"
>
> like that.
>
> > Perhaps it can be said too much, however. You may have missed the spot
> > where I mentioned that I had read the install docs
> > backwards and forwards, and many other sources of info floating around the
> > web just to figure out what the heck a .sit file
> > is. But I did also read the part where it says that the 'ofpath' command
> > doesn't work if you booted with BootX, and that
> > appears to be true on my machine when I tried it. So I am not able to make
> > use of the above instructions. It's sort of a
> > chicken and egg thing.
>
> ofpath only breaks with bootx on newworld systems.
>
> as for .sit Apple has been shipping a desitter with the OS for years
> now, but i do agree that archiving on macos is a horrid mess.
>
> > As for the idea that the installation procedure did something to install
> > quik, but that something isn't effective until you do
> > something else that's documented in the installation docs, it still isn't a
> > good reason to leave a somewhat misleading
> > exchange in the installation process. It could just as easily mention
> > after it's finished installing quik that the user now
> > needs to consult the docs to finish the job. For a person like me, I might
> > have taken a week to fully read the installation
>
> talk to the boot floppies maintainers about this, im not one of them.
> dbootstrap could use ofpath and nvsetenv to set the boot-device
> variable as part of the quik install process. thats half the reason i
> bothered to put oldworld support in ofpath in the first place.
>
> however the debian boot-floppies code is well a mess, there is no
> other way to say it, so i understand the logic behind not screwing
> with that pile of spaghetti code any more then necessary for a stable
> update. i do seriously hope that make-bootable-from-disk will be
> fixed to take care of the nvsetenv task on quik setup and setup
> ybin/yaboot on newworlds for woody. (i am not good enough at C to
> attempt to fix it myself)
>
> woody will probably be the last to use this boot-floppies code,
> debian-installer (the complete from scratch rewrite of boot-floppies)
> should make building proper bootloader setups much simpler.
> unfortuantly it doesn't look like debian-installer will be ready in
> time.
>
> > docs, as a background task, and may not remember every word by the time I
> > finally get the system to boot into Linux so I can
> > even do the install, so the more helpful the messages that pop up, the
> > better, if anyone is asking me. Some of the more
> > complicated parts might be hard to remember at all if you have no idea what
> > they are talking about, for instance, if you've
> > never heard of quik or yaboot or BootX or "OpenFirmware device paths"
> > before. OK, so I was already familiar with OF, but you
> > get my drift.
>
> documentation does need to be improved, but this is a volunteer
> effort. when you see holes volunteering to help is the best approach
> to getting them fixed.
>
> > I'm skeptical that it actually did install quik on my system, but I won't
> > know until I get the following question answered:
> >
> > I don't know which partition to specify as the boot partition when running
> > ybin/quik. On the original MacOS disk, there are 5
> > partitions:
> >
> > part1: the partition map
>
> apple partition tables have the peculiar property that the partition
> table itself *is* a partition.
>
> > part2: Apple_Driver43 (or something close to that -- 32KB in size, if
> > memory serves)
> > part3: Apple_Driver43 (part duex? this one is 128KB in size, if I remember
> > right)
> > part4: Apple_patches (don't remember the size, but less than 1MB, I think)
>
> macos cruft. required for macos to boot. (but not quik or any other
> OF bootloader)
>
> > part5: Apple_HFS whatever, the main partition with MacOS installed on it.
>
> you macos partition of course.
>
> > So, what does it all mean? ~:^) There is no Apple_Boot partition, I'm
> > guessing that this is another benefit of trying to
>
> Apple_Bootstrap partitions are only needed on Newworlds because
> Newworld OpenFirmware is broken in that it won't load a bootblock.
> Oldworld OF will load the first 2 blocks from any partition, so the
> quik first stage loader is installed in the bootblock of your ext2fs
> root partition (ext2fs like most filesystems leaves 2 512 byte blocks
> empty at the start of the partition to leave room for a bootloader)
>
> > reuse wicked obsolete hardware ~:^) I'm used to jumping through hoops to
> > get old hardware working, now if I can just figure
> > out which hoop to jump through. I installed Potato on a second disk, and I
> > left myself a 65MB partition on that disk with
> > type Apple_Boot, but haven't quite figured out how to switch over
> > completely to that disk. [I am thinking of changing the
> > SCSI id's on the two disks so that the linux disk is ends up sda instead of
> > sdb, and I'm hoping that the Apple ROM will be
> > nice enough to notice that the first disk isn't bootable but the second is,
> > and then I can run ybin and remove the MacOS drive
> > w/o any problems.] Right now I'm afraid to run ybin/quik and specify any
> > of the Apple_{Driver,patches} partitions as the boot
> > partition for fear that [I don't know what I'm doing and] I'll hose the
> > MacOS disk and then I will be completely screwed.
>
> ybin won't do you any good on oldworld as its purpose is to install
> yaboot, which is newworld only (ybin should warn when run on oldworlds
> that the bootstrap installed won't boot)
>
> quik should automatically install the bootblock on your root partition
> unless you have something wierd like a /boot partition separate from /
> (not recommended)
>
> you need a quik.conf first:
>
> partition=6 ## your root partition
> root=/dev/sda6 ## your root partition
>
> image=/boot/vmlinux-2.2.18pre21 ## your kernel, don't use a symlink it might
> not work
> label=Linux
> read-only
>
> then run /sbin/quik no arguments are needed.
>
> > Cheers,
> >
> > a
> >
> > PS. Three hoorahs for Debian Potato supporting PPC. I now have the
> > ability to make excellent use of this machine, which was
> > serving as a door stop before this, and the only thing which isn't perfect
> > is the current booting procedure, but even if I can
> > never remove the MacOS disk and always have to boot MacOS first, that would
> > hardly be the end of the world, since Linux has
> > yet to crash on this machine. It makes a fine web applications server for
> > very lite web apps. As long as there are no
> > graphics to render or quadratic equations to solve .... ~:^)
>
> quik can usually be made to work. one other option that you could
> look into is miboot, but its fickle, hard to configure/maintain and is
> undocumented. its not made to be a general purpose bootloader but
> made for bootfloppies.
>
> miboot however requires the proprietary Apple disk drivers (all those
> Apple_Driver partitions)
>
> --
> Ethan Benson
> http://www.alaska.net/~erbenson/
>
> ------------------------------------------------------------------------
> Part 1.2Type: application/pgp-signature
