Red Hat Bugzilla – Bug 56026
mkbootdisk fails on recent "smp" and "enterprise" kernels
Last modified: 2007-04-18 12:38:12 EDT
Description of Problem:
Recent "smp" and "enterprise" kernels (plus respective initrd's) are too
big to fit on floppy. As a result mkbootdisk fails. (This is also
applies to "install" for rh7.2 which uses mkbootdisk to make a bootfloppy.)
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1.find an i686 with a SCSI or two (or whatever makes initrd bigger)
2.become root, install "enterprise" kernel if it is not
3.mkinitrd --device /dev/fd0 2.4.9-13enterprise
It fails during initrd creation with "No space left on the device"
It works with no error messages.
[root@cholla boot]# ls -l /boot/initrd-2.4.9-13enterprise.img
-rw-r--r-- 1 root root 455871 Nov 10 17:47
[root@cholla boot]# ls -l /boot/vmlinuz-2.4.9-13enterprise
-rw-r--r-- 1 root root 1044432 Oct 30 17:53
This also happens with the standard 2.4.9-13 kernel (at least the Athlon version).
I find that 2.4.9-13 for UP athlon is the last kernel for which I can make a
boot disk. Both -21 and -31 fail. Maybe it is time for a mkbootcd command.
I'm using the SGI XFS installer based on either RedHat 7.1 or RedHat 7.2. With
the XFS enabled kernel it was never possible to create a bootdisk because of the
XFS code which enlarged the kernel too much. It seems that 'mainstream' RedHat
kernels suffer the same problem now. I have created patched RPMs of mkbootdisk
which allows to create overformatted bootdisks and I have not found any pc until
now which could not boot such a disk.
Updated mkbootdisk RPMs are here:
Create bootdisk as usual, but use a different device:
mkbootdisk --device /dev/fd0u1722 2.4.9-31enterprise
Would be nice to see the patch integrated into the next RedHat release.
It doesnt work for 2.4.9-31 on a intel system either. However a grub boot disk
can hold the current size of the vmlinuz image (952 kb). For automated boot you
would have make the appropriate entries into menu file. There of course is the
rescue mode on the iso image cdroms. Of course this boots the image of the iso,
but it can help you fix a boot problem.
Maybe someone could write a utility to create boot cd rom from your current
"out of space" isn't a bug in mkbootdisk, though it is a limitation that's
becoming more and more important.
"Out of space" isn't a bug in mkbootdisk, well is then a limitation
in glibc? This is incredible annoying, I am running Red Hat Linux
Enterprise 2.1 AS 2.4.9-e.27 SMP and I was able to create the
bootdisk when I ran at the 2.4.9-e.3 SMP levl, no I am left outside
Red Hat Support tells me that this bug has been seen in Red Hat 9 as
well, but I am running Red Hat Linux Enterprise 2.1 AS 2.4.9-e.27 so
I really don't care, I have purchased a product and I expect it to
If my server craches how do I then get access the a kernel having my
detected SCSI devices discovered?
Now I have been told to play with the mkbootdisk --iso --
device /root/boot.img command and I still keep asking myself, why on
earth shall I start to use time on this.
Have anyone of you played with the iso-stuff?
I have a Red Hat Linux Enterprise 3.0 AS running next to me and there
it works fine, but I really need to get something going on the Red
Hat Linux Enterprise 2.1 AS level