Re: boot-floppies on amiga/apus. friday cvs version.
On Fri, Apr 21, 2000 at 12:34:35PM -0400, Adam Di Carlo wrote:
> Sven LUTHER <email@example.com> writes:
> > no the modules are not set right, since we need loop.o to install the modules
> > which are in rescue.bin. so this is a chicken and egg problem.
> Then put loop.o on the root disk, just as we do on i386 for the module
> enabling syslog.
Where do i do that, ...
> > > > * why not install the modules from drviers.tgz directly, if present, this
> > > > way there should be no need to do stuff with the loop device and
> > > > rescue.bin file. On apus and other arch/subarch who don't support lilo or
> > > > similar, rescue.bin is overkill anyway.
> > >
> > > Do not comprehend.
> > Well, since there is no lilo, there is no need to install the kernel into
> > /boot, since it will reside into an affs partition.
> Still, you need the kernel from somewhere, don't you?
Yes, from the amiga side of things. and the current install places a kernel in
> > Also the driver modules
> > are pretty small, compared to i386, only 325928. So there is really no need to
> > make people download a 1.4MB disk image, when the only reason to use it will
> > be to umpack drivers.tgz.
> Well, if it's so small, then it will be quick to download (the rest of
> the image being zero padded, it should download quickly).
> I really do not believe in special-casing our install procedure for
> such minor gains. The maintenance implications of what you are
> proposing is monstrous -- and things are already monstrous.
just have a "install only modules and no kernel" option in the menu, that does
a tar xzvf drivers.tgz or something such, in total, perhaps 10/20 lines of
code. nothing monstrous really, ...
> > Also the kernel is available alone also (in
> > subarch/linux).
> Still, you need the kernel from somewhere, don't you? Where do you
> get it from if not rescue.bin ?
I just said it, the kernel got copiedto apus/linux. (in the
disks-powerpc/2.2.x.xxx dir. TO use it, we just do :
bootstrap --apus -k apus/linux -r apus/images-1.44/root.bin root=/dev/ram ...
clearly no need for rescue.bin, and all the problem coming with in what format
to store it, etc ...
There are no true floppies to boot from for amigaos anyway (since the standard
floppy drive is 880 KB only anyway, though i think you could put a kernel +
boot strap there. Care about supporting 880KB image set ? :)))
> > > > * once i hand insmodded loop.o from the drivers.tgz, i try to install os &
> > > > modules, but get a Can't mount the rescue floppy, i try to loop mount it
> > > > by hand, but it didn't work also, maybe the rescue.bin is broken
> > > > ?
> > >
> > > More likely m68k loop device is broken. The CVS area has been
> > > building pretty much fine all week.
> > Well, no loop device works pretty well, i can mount the root partition, and
> > build the whole stuff without problem. I more think it is brain dead i386
> > centric rescue.bin behavior.
> I think you just don't understand it.
> Is it brain dead to depend on loop.o to be available in the kernel you
> are booting with ? I don't think so. Even so, we allow extra kernel
> modules on root.bin for arches which need it. I can't understand why
> any of this is i386 centric.
Well, yes i had loops included in the kernel at first, to solve this selfsame
problem, but boot floppies wanted to insert a loop.o module, which was
naturally not available. this was at the begining, and i didn't knew my way
around boot floppies, so i just started using a kernel with loop modules.
The i386 centricness is the dependance on floppies, i guess no other arch is
truly using floppies, or are they ?
(at least not powerpc nor m68k, i don't think sparc uses floppies a lot, maybe
> > BTW, i am running a ppc, not an m68k.
> Then have you been talking to Daniel Jacobowitz? You are telling me
> very stubbornly that everything is broken and the sky is falling,
> while he is telling me that it all works fine on ppc.
Sure i have, ...
he solved the segfault bug in 2.2.11.
And i don't tell you nothing about everything being broken, i just tell you
that using rescue.bin for arches that are not using using a lilo like setup
but a loadlin like setup, there is never a need to install the kernel on the
linux side of things, so you can just untar the drivers.tgz tarball which is
anyway included in the release as of today. This should represent 10 to 20
lines of code, and when i spoke of it some time ago, others than me agreed
that this could be useful. look at the debian-boot archives. I would have
implemented, had not my system decided to not boot anymore with scsi disks ...
> > > > * With lots of other menus, when checking the state of the install, i get :
> > > >
> > > > libfdisk : error reading sector 538988403 : no such file or directory.
> > >
> > > Doens't seem to happen on other arches AFAIK. Wierd.
> > Well, it don't hinder the process or anything.
> > but then maybe i have problem with my disk ...
> > > > * after all of this, when viewing the partition table, only part of the
> > > > table is shown, i have 7 partitions (amiga partitioning), and /dev/sda4 is
> > > > target for install, prior to mounting it, all of them get shown, but after
> > > > i only get partitions 1 to 3 shown, as well as a reference to sdb, which
> > > > has no partition table (being my super floppy syjet cartridge).
> > >
> > > Yikes.
> > You said it.
> > That said, unless someone else reports this and the preceding problem also, It
> > could well be done to hardware problem on my setup. Since then it seems for
> > some strange reason my scsi disk is no more wanting to be recognized (or more,
> > my system won't boot when scsi is enabled and/or the disk present).
> > Will try to get a ide disk to continue checking this.
> Well, I'm not a powerpc user so I don't think I can shed any light
No, thanks anyway, ...
i hope to get a cheap ide disk, at least until the potato release, or else fix
the stuff in cvs, wait for dan to build it, and then ask one of my fellow apus
user to test it, not very quick process, but well, ...