Bootable CD idea

Bootable CD idea

[John Bradford]

Just a random idea that occurred to me...

Since the El-Torito bootable CD standard supports multiple floppy
images on a single CD, it would be possible to write a script that
takes a .config file, and the source to, say all the -pre and -rc
versions of a particular kernel, compiles multiple kernels with the
same .config, and writes a CD with them all on, set to boot from an
arbitrary disk partition.

It would make:

> > > Foo doesn't work in -rc2
> > Did it work in -rc1
> Not sure

E-Mail exchanges a thing of the past.

Note, that as each floppy image is separate, it's not the same as
trying to cram multiple kernels on a 2.88 MB floppy image, and is
therefore actually do-able :-).

John.

Re: Bootable CD idea

[Richard B. Johnson]

On Wed, 30 Apr 2003, John Bradford wrote:

> Just a random idea that occurred to me...
>
> Since the El-Torito bootable CD standard supports multiple floppy
> images on a single CD, it would be possible to write a script that
> takes a .config file, and the source to, say all the -pre and -rc
> versions of a particular kernel, compiles multiple kernels with the
> same .config, and writes a CD with them all on, set to boot from an
> arbitrary disk partition.
>
> It would make:
>
> > > > Foo doesn't work in -rc2
> > > Did it work in -rc1
> > Not sure
>
> E-Mail exchanges a thing of the past.
>
> Note, that as each floppy image is separate, it's not the same as
> trying to cram multiple kernels on a 2.88 MB floppy image, and is
> therefore actually do-able :-).
>
> John.

Well, it's a boot-loader problem that has to be solved in one
of the boot-loader programs like grub or lilo. When booting a
CD, the BIOS is made to 'think' that a CD is a floppy. It
loads the boot-record and jumps to its code. From that point
on, whatever happens is based upon how, what, and where the
boot record and subsequent code executes. With the large
data space available, you could certainly have multiple
operating systems. It's simply a coding problem. If you
want to modify one of the existing boot-loader programs
let me know. I may be able to help, and certainly can test.

FYI I have a bootable CR/ROM (who doesn't), that contains
a limited root file-system with ramdisks that mount on
/tmp and /var. I use this to boot any linux machine and
repair it. It would be nice to be able to select different
operating system versions as well.


Cheers,
Dick Johnson
Penguin : Linux version 2.4.20 on an i686 machine (797.90 BogoMips).
Why is the government concerned about the lunatic fringe? Think about it.

Re: Bootable CD idea

[John Bradford]

> > Just a random idea that occurred to me...
> >
> > Since the El-Torito bootable CD standard supports multiple floppy
> > images on a single CD, it would be possible to write a script that
> > takes a .config file, and the source to, say all the -pre and -rc
> > versions of a particular kernel, compiles multiple kernels with the
> > same .config, and writes a CD with them all on, set to boot from an
> > arbitrary disk partition.
> >
> > It would make:
> >
> > > > > Foo doesn't work in -rc2
> > > > Did it work in -rc1
> > > Not sure
> >
> > E-Mail exchanges a thing of the past.
> >
> > Note, that as each floppy image is separate, it's not the same as
> > trying to cram multiple kernels on a 2.88 MB floppy image, and is
> > therefore actually do-able :-).
> >
> > John.
> 
> Well, it's a boot-loader problem that has to be solved in one
> of the boot-loader programs like grub or lilo. When booting a
> CD, the BIOS is made to 'think' that a CD is a floppy. It
> loads the boot-record and jumps to its code. From that point
> on, whatever happens is based upon how, what, and where the
> boot record and subsequent code executes. With the large
> data space available, you could certainly have multiple
> operating systems. It's simply a coding problem. If you
> want to modify one of the existing boot-loader programs
> let me know. I may be able to help, and certainly can test.

I didn't think we'd need to modify anything - the BIOS lets us choose
the floppy image we want, so each image only needs to contain a
bootloader[1], and a kernel.  The root filesystem doesn't need to be
on the CD - if the disk is to be used for quickly visiting another
kernel to see if something breaks, you'd want your normal root
filesystem.  If you set the root filesystem to /dev/fd0, you can
actually use a real floppy, as the kernel bypasses the BIOS floppy
emulation :-).  So, you can boot from a CD, and use a floppy that
you've prepared separately as your root filesystem.

Were you suggesting that we could have the root filesystem as a set of
blocks on the CD, which are accessed directly, and not via El-Torito
floppy emulation?  That would be pretty useful.  Heh, initial ram disk
on a CD would be even better...

[1] I originally thought that the 2.4 kernel's in-built floppy
bootloader used BIOS calls to access the disk, and that a 2.4 kernel
image as the El-Torito boot image would work, as the kernel would be
accessing the emulated disk, but it didn't seem to when I tried it
just now - it failed with an error saying something along the lines of
it had run out of data to decompress.

> FYI I have a bootable CR/ROM (who doesn't), that contains
> a limited root file-system with ramdisks that mount on
> /tmp and /var. I use this to boot any linux machine and
> repair it. It would be nice to be able to select different
> operating system versions as well.

Yeah, maybe I was just being lazy, but I imagined something like:

make allprepatches 

running the same config through 2.4.21-pre1 to -pre7, and -rc1, and
writing a CD that boots and gives you a menu to select from.

Kind of like make bzdisk, but for 2003 instead of 1991 :-).

I mean, if we're killing of the in-kernel bootloader in 2.5, why not?

John.

Re: Bootable CD idea

[Tomas Szepe]

> [john@grabjohn.com]
> 
> I didn't think we'd need to modify anything - the BIOS lets us choose
> the floppy image we want, so each image only needs to contain a
> bootloader[1], and a kernel.

Why not use isolinux[1] to boot however many images you feel like booting?

-- 
Tomas Szepe <szepe@pinerecords.com>

[1] http://syslinux.zytor.com/iso.php

Re: Bootable CD idea

[John Bradford]

> > I didn't think we'd need to modify anything - the BIOS lets us choose
> > the floppy image we want, so each image only needs to contain a
> > bootloader[1], and a kernel.
> 
> Why not use isolinux[1] to boot however many images you feel like booting?

Good point :-)

John.

Re: Bootable CD idea

[Tigran Aivazian]

On Wed, 30 Apr 2003, John Bradford wrote:
> [1] I originally thought that the 2.4 kernel's in-built floppy
> bootloader used BIOS calls to access the disk, and that a 2.4 kernel
> image as the El-Torito boot image would work, as the kernel would be
> accessing the emulated disk, but it didn't seem to when I tried it
> just now - it failed with an error saying something along the lines of
> it had run out of data to decompress.

when you did "make bzImage", are you sure you didn't get the message about 
the kernel being too big for floppy booting?

Regards
Tigran

Re: Bootable CD idea

[John Bradford]

> > [1] I originally thought that the 2.4 kernel's in-built floppy
> > bootloader used BIOS calls to access the disk, and that a 2.4 kernel
> > image as the El-Torito boot image would work, as the kernel would be
> > accessing the emulated disk, but it didn't seem to when I tried it
> > just now - it failed with an error saying something along the lines of
> > it had run out of data to decompress.
> 
> when you did "make bzImage", are you sure you didn't get the message about 
> the kernel being too big for floppy booting?

No, I've just checked - the same kernel image boots fine from a real floppy.

The CD image just gives:

Loading ...........................
Uncompressing Linux ...

Ran out of input data

--- System halted

Looking at bootsect.S, I am suprised it doesn't work, as it's done via
Int 0x13, so I can't see why it doesn't see the BIOS-emulated floppy
correctly.

John.

Re: Bootable CD idea

[H. Peter Anvin]

Followup to:  <200304301325.h3UDPtla000141@81-2-122-30.bradfords.org.uk>
By author:    John Bradford <john@grabjohn.com>
In newsgroup: linux.dev.kernel
>
> > > [1] I originally thought that the 2.4 kernel's in-built floppy
> > > bootloader used BIOS calls to access the disk, and that a 2.4 kernel
> > > image as the El-Torito boot image would work, as the kernel would be
> > > accessing the emulated disk, but it didn't seem to when I tried it
> > > just now - it failed with an error saying something along the lines of
> > > it had run out of data to decompress.
> > 
> > when you did "make bzImage", are you sure you didn't get the message about 
> > the kernel being too big for floppy booting?
> 
> No, I've just checked - the same kernel image boots fine from a real floppy.
> 

The boot sector bootloader is broken for anything but genuine legacy
floppies, because it relies on getting the proper sector not found in
order to determine the geometry.  Most LBA<->CHS conversions -- and
that includes El Torito, IDE floppies, USB floppies, and just about
anything else that isn't a classical legacy floppy -- simply spill
into the next track, confusing bootsect.S.  This is part of why
bootsect.S is gone in 2.5.

	-hpa

-- 
<hpa@transmeta.com> at work, <hpa@zytor.com> in private!
"Unix gives you enough rope to shoot yourself in the foot."
Architectures needed: ia64 m68k mips64 ppc ppc64 s390 s390x sh v850 x86-64

Re: Bootable CD idea

[John Bradford]

> > > > [1] I originally thought that the 2.4 kernel's in-built floppy
> > > > bootloader used BIOS calls to access the disk, and that a 2.4 kernel
> > > > image as the El-Torito boot image would work, as the kernel would be
> > > > accessing the emulated disk, but it didn't seem to when I tried it
> > > > just now - it failed with an error saying something along the lines of
> > > > it had run out of data to decompress.
> > > 
> > > when you did "make bzImage", are you sure you didn't get the message about 
> > > the kernel being too big for floppy booting?
> > 
> > No, I've just checked - the same kernel image boots fine from a real floppy.
> > 
> 
> The boot sector bootloader is broken for anything but genuine legacy
> floppies, because it relies on getting the proper sector not found in
> order to determine the geometry.  Most LBA<->CHS conversions -- and
> that includes El Torito, IDE floppies, USB floppies, and just about
> anything else that isn't a classical legacy floppy -- simply spill
> into the next track, confusing bootsect.S.  This is part of why
> bootsect.S is gone in 2.5.

Ah, it makes more sense now :-).  So, could I bodge 2.4 in to working
by modifying bootsect.S with something like this?


        movw    $disksizes+1, %si         # Force 18 sectors/track
probe_loop:
        lodsb
        cbtw                            # extend to word
        xchgw   %cx, %ax                # %cx = track and sector
        xorw    %dx, %dx                # drive 0, head 0
        movw    $0x0200, %bx            # address = 512, in INITSEG (%es = %cs)
        movw    $0x0201, %ax            # service 2, 1 sector
        int     $0x13
        movb    $0x03, %ah              # read cursor pos
        xorb    %bh, %bh

John.

Re: Bootable CD idea

[H. Peter Anvin]

John Bradford wrote:
> 
> Ah, it makes more sense now :-).  So, could I bodge 2.4 in to working
> by modifying bootsect.S with something like this?
> 
> 
>         movw    $disksizes+1, %si         # Force 18 sectors/track
> probe_loop:
>         lodsb
>         cbtw                            # extend to word
>         xchgw   %cx, %ax                # %cx = track and sector
>         xorw    %dx, %dx                # drive 0, head 0
>         movw    $0x0200, %bx            # address = 512, in INITSEG (%es = %cs)
>         movw    $0x0201, %ax            # service 2, 1 sector
>         int     $0x13
>         movb    $0x03, %ah              # read cursor pos
>         xorb    %bh, %bh
> 
> John.

Perhaps you could, but why the fsck do you want to beat a dead horse
over and over again?  Use a real bootloader instead.

	-hpa