Why can't ext4 file systems be used as a boot partition?

On Sep 18, 6:43 am, Aragorn <arag... DeleteThis @chatfactory.invalid> wrote:

> There are of course reasons as to why you might prefer GRUB over LILO.
> For one, GRUB offers more possibilities - such as an interactive boot
> shell and the ability to load many other types of UNIX operating
> systems, including GNU/Hurd, Free-/Net-/OpenBSD and Solaris - and if
> you're going to use the Xen hypervisor, then using LILO requires some
> special trick to get Xen to load the privileged domain's kernel,
> because LILO does not support module loading, while GRUB does.  
>
> Still, somehow I don't think that would be an issue in your case, so if
> you want to have /ext4/ on the */boot* partition and you can't install
> any of the later GRUB versions, try going with LILO instead.  It's not
> as featureful as GRUB but it'll do the job of booting up your
> system.

I'm glad Aragorn is around. He gives good answers, that I can
occasionally add a bit of extra thought to. For example, grub is far
easier to manage. You load the master boot record (MBR) once, then
that gives you enough to read all those cool configuration files and
you don't have to reload it everytiime you have a new kernel. It makes
alternative kernel management and dual-booting much, much safer. And
it doesn't have the roughly 15 character limits in the descriptions,
so you can say "RHEL 4 Xen test kernel 2.5.0.3.14152695, mirror disk"
and have it mean something useful.

grub lacks one feature I really miss from LILO, the command line "use
the old kernel still there as the default", then "use the new kernel
one time and one time only to reboot with". (You had to do them on
different commands, on the same command didn't work, dang it!!!) This
let you reboot once with the new kernel, and if it failed, the boot
procedure would revert to the old kernel merely by someone hitting the
power button. That was incredibly helpful when testing out new kernels
or sending new kernels to weird, remote hardware: having a cage monkey
hit the power button is much safer than making them connect up a
keyboard and console.

On Sep 18, 8:07 am, Aragorn <arag....TakeThisOut@chatfactory.invalid> wrote:
> On Friday 18 September 2009 13:52, someone identifying as *Nico
> Kadel-Garcia* wrote in /comp.os.linux.setup:/

> > grub lacks one feature I really miss from LILO, the command line "use
> > the old kernel still there as the default", then "use the new kernel
> > one time and one time only to reboot with". (You had to do them on
> > different commands, on the same command didn't work, dang it!!!) This
> > let you reboot once with the new kernel, and if it failed, the boot
> > procedure would revert to the old kernel merely by someone hitting the
> > power button.
>
> GRUB does have a "boot once" option, doesn't it?

If it's there, it's relatively new: I've had to work with RHEL 4 and
RHEL 5 the last few years, and it certainly wasn't there.

I'm looking right now at http://www.gnu.org/software/grub/manual/grub.html#Booting-once_002donly.
And I'm saying 'eeeewwwww!!!!' Hey, let's go in and manually edit our
grub.conf and add an indirectioi for the 'saved' kernel Then when we
do a kernel installation on RHEL and it renumbers the boot loaders to
make the new kernel first, it can be completely confused! That'll fix
things!

Just, just.... ewwww. The new 'grub 2' does seem to have a better
mechanism of handling this, but RHEL 5 is still stuck at version 0.7of
grub, so I'm reluctant to play with this for a while.

> > That was incredibly helpful when testing out new kernels or sending
> > new kernels to weird, remote hardware: having a cage monkey hit the
> > power button is much safer than making them connect up a keyboard and
> > console.
>
> It was once postulated by chaoticians that a million monkeys sitting at
> a computer would eventually come up with the collected works of William
> Shakespeare.  Now, with the internet, we know that this is not true. :p

Now, now, be nice to the cage monkeys. I'm restraining a lot of odd
humor about them at the moment, but they've usually done their flat-
out best for me when I've called and I appreciate their on-site hands
and configuration insights tremendously.

On Friday 18 September 2009 07:21, someone identifying as *piscesboy*
wrote in /comp.os.linux.setup:/

> Newbie question.

As Allen noted, the most recent versions of GRUB can handle /ext4/
filesystems, but the older versions can't, as /ext4/ is still fairly
new.

That said, there *is* an alternative, and that is to use LILO as the
bootloader instead of GRUB. GRUB loads the kernel via a filesystem
driver, but LILO does it by hardcoding the logical block offset
instead. It is totally oblivious of any filesystems and could
therefore be used with any kind of filesystem on the */boot* partition.

There are of course reasons as to why you might prefer GRUB over LILO.
For one, GRUB offers more possibilities - such as an interactive boot
shell and the ability to load many other types of UNIX operating
systems, including GNU/Hurd, Free-/Net-/OpenBSD and Solaris - and if
you're going to use the Xen hypervisor, then using LILO requires some
special trick to get Xen to load the privileged domain's kernel,
because LILO does not support module loading, while GRUB does.

Still, somehow I don't think that would be an issue in your case, so if
you want to have /ext4/ on the */boot* partition and you can't install
any of the later GRUB versions, try going with LILO instead. It's not
as featureful as GRUB but it'll do the job of booting up your
system.

--
*Aragorn*
(registered GNU/Linux user #223157)

On Friday 18 September 2009 13:52, someone identifying as *Nico
Kadel-Garcia* wrote in /comp.os.linux.setup:/

> On Sep 18, 6:43 am, Aragorn <arag....TakeThisOut@chatfactory.invalid> wrote:
>
>> There are of course reasons as to why you might prefer GRUB over
>> LILO. For one, GRUB offers more possibilities - such as an
>> interactive boot shell and the ability to load many other types of
>> UNIX operating systems, including GNU/Hurd, Free-/Net-/OpenBSD and
>> Solaris - and if you're going to use the Xen hypervisor, then using
>> LILO requires some special trick to get Xen to load the privileged
>> domain's kernel, because LILO does not support module loading, while
>> GRUB does.
>>
>> Still, somehow I don't think that would be an issue in your case, so
>> if you want to have /ext4/ on the */boot* partition and you can't
>> install any of the later GRUB versions, try going with LILO instead.
>> It's not as featureful as GRUB but it'll do the job of booting up
>> your system.
>
> I'm glad Aragorn is around. He gives good answers, that I can
> occasionally add a bit of extra thought to.

Why thank you, I really appreciate that.

> For example, grub is far easier to manage. You load the master boot
> record (MBR) once, then that gives you enough to read all those cool
> configuration files and you don't have to reload it everytiime you
> have a new kernel. It makes alternative kernel management and
> dual-booting much, much safer. And it doesn't have the roughly 15
> character limits in the descriptions, so you can say "RHEL 4 Xen test
> kernel 2.5.0.3.14152695, mirror disk" and have it mean something
> useful.

It also offers encryption for the password protected entries in the
configuration file. LILO offers password protection, but the password
is stored in */etc/lilo.conf* unencrypted, which necessitates that
*/etc/lilo.conf* be set up with 600 permissions on multi-user systems.
Not that this is such a troublesome thing to do of course, but still,
it's another area where GRUB is better.

> grub lacks one feature I really miss from LILO, the command line "use
> the old kernel still there as the default", then "use the new kernel
> one time and one time only to reboot with". (You had to do them on
> different commands, on the same command didn't work, dang it!!!) This
> let you reboot once with the new kernel, and if it failed, the boot
> procedure would revert to the old kernel merely by someone hitting the
> power button.

GRUB does have a "boot once" option, doesn't it?

> That was incredibly helpful when testing out new kernels or sending
> new kernels to weird, remote hardware: having a cage monkey hit the
> power button is much safer than making them connect up a keyboard and
> console.

It was once postulated by chaoticians that a million monkeys sitting at
a computer would eventually come up with the collected works of William
Shakespeare. Now, with the internet, we know that this is not true. :p

--
*Aragorn*
(registered GNU/Linux user #223157)

On Fri, 18 Sep 2009 02:58:38 -0500, Allen Kistler wrote:

> piscesboy wrote:
>> Newbie question.
>
> Very brand new versions of grub (like in Fedora 12, scheduled to be
> released in November) can read ext4, but older ones can't.
>
> Can't read /boot = can't boot.

One could have a small ext2 /boot partition

--
GNU/Linux runs on IBM mainframes and on the world's fastest supercomputers
Windows supercomputers on the other hand are called botnets. <grin>

Aragorn schrieb:

> As Allen noted, the most recent versions of GRUB can handle /ext4/
> filesystems, but the older versions can't, as /ext4/ is still fairly
> new.
>
> That said, there *is* an alternative, and that is to use LILO as the
> bootloader instead of GRUB. GRUB loads the kernel via a filesystem
> driver, but LILO does it by hardcoding the logical block offset
> instead. It is totally oblivious of any filesystems and could
> therefore be used with any kind of filesystem on the */boot* partition.

Since GRUB can boot e.g. Windows systems as well - isn't there a GRUB
option, that a logical drive shall be booted in the "traditional" (LILO)
way?

DoDi

On Friday 18 September 2009 15:00, someone identifying as *Hans-Peter
Diettrich* wrote in /comp.os.linux.setup:/

> Aragorn schrieb:
>
>> As Allen noted, the most recent versions of GRUB can handle /ext4/
>> filesystems, but the older versions can't, as /ext4/ is still fairly
>> new.
>>
>> That said, there *is* an alternative, and that is to use LILO as the
>> bootloader instead of GRUB. GRUB loads the kernel via a filesystem
>> driver, but LILO does it by hardcoding the logical block offset
>> instead. It is totally oblivious of any filesystems and could
>> therefore be used with any kind of filesystem on the */boot*
>> partition.
>
> Since GRUB can boot e.g. Windows systems as well - isn't there a GRUB
> option, that a logical drive shall be booted in the "traditional"
> (LILO) way?

I'm not sure I understand your question... :-/ Both GRUB and LILO boot
Windows in the same manner, i.e. by loading the Windows partition's
bootsector into memory and passing control of the machine onto whatever
code resides there, and this code in question would then be the Windows
bootloader.

Now, Windows traditionally requires that its bootloader be installed in
the active primary partition on the hard disk seen as the BIOS as being
the bootable one, but there are indeed tricks to fool Windows - and DOS
for that matter - into having it believe that it resides on an active
primary partition on the bootable hard disk while it in fact resides in
a logical partition in an extended partition container, yes.

While the above is perfectly possible - and has been done - I do however
wonder how one could trick Windows into being *installed* - i.e.
including its bootloader - in a logical partition of the extended
partition container. I know it has been done with Windows 98, which
was still DOS-based, but I don't see how Win98 could have been
installed in that partition, unless it was a primary partition during
Windows install time and was then converted into a logical partition in
an extended container via some very elaborate operation.

As your question was pretty vague, I'm not sure whether my reply is
adequate, though.

--
*Aragorn*
(registered GNU/Linux user #223157)

Nico Kadel-Garcia wrote:

> grub lacks one feature I really miss from LILO, the command line "use
> the old kernel still there as the default", then "use the new kernel
> one time and one time only to reboot with". (You had to do them on
> different commands, on the same command didn't work, dang it!!!) This
> let you reboot once with the new kernel, and if it failed, the boot
> procedure would revert to the old kernel merely by someone hitting the
> power button. That was incredibly helpful when testing out new kernels
> or sending new kernels to weird, remote hardware: having a cage monkey
> hit the power button is much safer than making them connect up a
> keyboard and console.

Oops, you got that backwards, grub allows you to reboot once, not lilo.

Nico Kadel-Garcia wrote:

>
> If it's there, it's relatively new: I've had to work with RHEL 4 and
> RHEL 5 the last few years, and it certainly wasn't there.

I've not used lilo for so long, I forget that it had a reboot once
option. Anyway, they both do and it's nothing new at all for grub.

Aragorn wrote:
> On Friday 18 September 2009 13:52, someone identifying as *Nico
> Kadel-Garcia* wrote in /comp.os.linux.setup:/
>
>> On Sep 18, 6:43 am, Aragorn <arag... DeleteThis @chatfactory.invalid> wrote:
>>
>>> There are of course reasons as to why you might prefer GRUB over
>>> LILO. For one, GRUB offers more possibilities - such as an
>>> interactive boot shell and the ability to load many other types of
>>> UNIX operating systems, including GNU/Hurd, Free-/Net-/OpenBSD and
>>> Solaris - and if you're going to use the Xen hypervisor, then using
>>> LILO requires some special trick to get Xen to load the privileged
>>> domain's kernel, because LILO does not support module loading, while
>>> GRUB does.
>>>
>>> Still, somehow I don't think that would be an issue in your case, so
>>> if you want to have /ext4/ on the */boot* partition and you can't
>>> install any of the later GRUB versions, try going with LILO instead.
>>> It's not as featureful as GRUB but it'll do the job of booting up
>>> your system.
>> I'm glad Aragorn is around. He gives good answers, that I can
>> occasionally add a bit of extra thought to.
>
> Why thank you, I really appreciate that.
> #

well add my welcome too, as I appreciate your style and knowledge.

Nico Kadel-Garcia wrote:
> On Sep 18, 8:07 am, Aragorn <arag....RemoveThis@chatfactory.invalid> wrote:

>> It was once postulated by chaoticians that a million monkeys sitting at
>> a computer would eventually come up with the collected works of William
>> Shakespeare. Now, with the internet, we know that this is not true. :p
>
> Now, now, be nice to the cage monkeys. I'm restraining a lot of odd
> humor about them at the moment, but they've usually done their flat-
> out best for me when I've called and I appreciate their on-site hands
> and configuration insights tremendously.

Ah, but those were RANDOM monkeys.

The internet species is anything but..its more the BANDAR-LOG species,
if you read your Kipling... Cf the Hunting of Kaa.

>On Sep 18, 8:07=A0am, Aragorn <arag....RemoveThis@chatfactory.invalid> wrote:
>>
>> It was once postulated by chaoticians that a million monkeys sitting at
>> a computer would eventually come up with the collected works of William
>> Shakespeare. =A0Now, with the internet, we know that this is not true. :p

Nope. they gave us Wikipedia instead.

unruh wrote:
>> On Sep 18, 8:07=A0am, Aragorn <arag... RemoveThis @chatfactory.invalid> wrote:
>>> It was once postulated by chaoticians that a million monkeys sitting at
>>> a computer would eventually come up with the collected works of William
>>> Shakespeare. =A0Now, with the internet, we know that this is not true. :p
>
> Nope. they gave us Wikipedia instead.
>
ITYM facebook.