cat /proc/mounts was correct: after checking the root fs it was marked
as ext3.
But now it works!
I missed one thing: util-linux package.
I had util-linux-2.10f-7 that doesn't contain pivot_root (called inside
linuxrc in the initrd image to switch the root).
I upgraded to util-linux-2.11f-9.
rpm --rebuild gave problems about info and man files (changed path from
6.2 to 7.2 I think), so:
- get src.rpm from 7.2 distro
- rpm -ivh .src.rpm
- unpacked and changed the .tar.bz2 file substituting in MCONFIG
from
MAN_DIR= /usr/share/man
INFO_DIR= /usr/share/info
into
MAN_DIR= /usr/man
INFO_DIR= /usr/info
- cd /usr/src/redhat/SPECS
- change the spec file: duplicate entry of
%{_mandir}/man1/arch.1*
(probably a bug of the src.rpm itself undependently by the distro???
rpm -ba complained about it and didn't complete successfully...)
- rpm -ba util-linux.spec
- installed the corresponding rpm binary generated in
/usr/src/redhat/RPMS/i386.
-voila'
-great sw rpm!!!
-great sw rpm!!!
Look for example at: http://www.rpm.org/max-rpm/index.html
And so getting also the pivot_root, after recompiling kernel, modules
(for safety) and installing, my glibc21 rh 6.2 based system can quickly
recover also my root ext3 file system without problem even after an
unclean power off. Great!
ext2 is compiled statically into the kernel,
ext3 is compiled as a module (no entry in modules.conf: ext3.o and
jbd.o are automatically included in the initrd image when mkinitrd
is launched)
Bye,
Gianluca
Stephen C. Tweedie wrote:
> Hi,
>
> On Thu, Jan 03, 2002 at 03:38:06PM +0100, Gianluca Cecchi wrote:
>
>
>>I upgraded my filesystems to ext3 using tune2fs.
>>
>>Now I'm using kernel 2.4.17 with ext2 in core and ext3 as a module, so I
>>have the two jbd.o and ext3.o files under /lib/modules tree.
>>
>>If I power off cleanly my system I have no problem rebooting and the
>>filesystem type of them is ext3 (I put ext3 for root and auto for the
>>others in fstab).
>>
>>If I power off uncleanly the machine, root has to make fsck (in the long
>>ext2 way and complaining also), while for the other file systems the
>>journal is applied and in few seconds they are ok and after booting they
>>are all ext3.
>>
>>What is the problem with root filesystem?
>>
>
> It could be several things. Does /proc/mounts show the root fs as
> ext2 or ext3? Getting the wrong root fs type is the most likely
> problem.
>
>
>>I have read in some places something like this:
>>
>>WARNING : If u are converting your existing ext2 root file system to
>> ext3 than compile ext3 support in the kernel since u
>> will need it on startup [:-)]
>>
>
>>Is it right?
>>
>
> Not strictly correct, but it's much easier to get things working that
> way as you don't have to fiddle with getting the initrd correct.
>
> Cheers,
> Stephen
>
