Hi all, I've been working with the ext3-fs for several months (since 0.0.2c), and it works pretty fine for me, but I haven't found out, how I works to migrate my root-partition from ext2 to ext3. I tried it with the commands at the lilo-prompt, but it didn't work. I only received a kernel panic. Any ideas, how I can migrate to ext3 on my root-partition? tnx, Joachim -- Joachim Kunze Alte Marktstrasse 16 Tel.: +49-7042-830006 D-71665 Horrheim Fax: +49-7042-830006 Germany eMail: joachim@resistance.de
Hi, On Wed, Nov 01, 2000 at 04:49:35PM +0100, Joachim Kunze wrote:> Hi all, > > I've been working with the ext3-fs for several months (since 0.0.2c), > and it works pretty fine for me, but I haven't found out, how I works to > migrate my root-partition from ext2 to ext3. > > I tried it with the commands at the lilo-prompt, but it didn't work. > I only received a kernel panic.Well, it works fine for me, so unless you tell me what you did exactly and what happened, I can't really debug this! Cheers, Stephen
On Wed, Nov 01, 2000 at 04:49:35PM +0100, Joachim Kunze wrote:> Hi all, > > I've been working with the ext3-fs for several months (since 0.0.2c), > and it works pretty fine for me, but I haven't found out, how I works to > migrate my root-partition from ext2 to ext3. > > I tried it with the commands at the lilo-prompt, but it didn't work. > I only received a kernel panic. > > Any ideas, how I can migrate to ext3 on my root-partition? > > tnx, >I almost don't want to say how I did it, but it does in fact work. Machine has a 30G disk, with 128MB swap partition, 30MB boot partition, and the rest is root. ("But you should have 9 very tiny partitions. That root is way too big, let me tell you why blah blah blah." Whatever.) What I did was disabled swap, reformatted the swap partition as ext2, set the type from swap to "linux" and copied just enough of the real root partition (bin, sbin, etc, lib, ... usr wasn't needed) to boot. Just add a stanza to your lilo.conf for this "alternate" root partition. Boot. (I didn't even boot single user - it errors all over the place but who cares). Then mkdir /mnt/realroot mount -t ext2 /dev/hdaX /mnt/realroot dd if=/dev/zero ..... exactly as from the doc ls -i /mnt/realroot/.journal.dat umount /mnt/realroot mount -t ext3 -o noload,journal=XXXX /dev/hdaX /mnt/realroot It should mount fine now. Be wary of a too small journal file. I recommend following the ext3 docs on beta.redhat.com exactly. [Linux machines (and probably *BSD) can be maintained by a good sysadmin without any of the delusions in /usr. This situation is why. (And network mounted /usr partitions, but thats another story) Never attempt running a "real" unix without /usr.] Remember to change the real root partitions /etc/fstab, rerun lilo, and then reboot. It should come up with / being an ext3 partition type. I would cat /proc/mounts just to make SURE that its an ext3. If all goes well, take the fake root out of lilo.conf, mkswap on your swap partition, and change the swap partitions type back to "swap". Rerun lilo. Reboot to regain your swap partition. (I have to reboot to regain swap due to being conservative about the warning that fdisk prints out when I write out the new partition table. YMMV] Viola. No warranty express or implied. Always wear your seatbelt. Professional driver on closed course. Do not attempt these maneuvers. Not all investments go up, some go down. Not FDIC insured. If it breaks in two, keep both pieces.:) -- Hunter Matthews Unix / Network Administrator Office: BioScience 222/244 Duke Univ. Biology Department Key: F0F88438 / FFB5 34C0 B350 99A4 BB02 9779 A5DB 8B09 F0F8 8438 Never take candy from strangers. Especially on the internet.
Hi Stephen, "Stephen C. Tweedie" schrieb:> > Hi, > > On Thu, Nov 02, 2000 at 11:25:54PM +0100, Joachim Kunze wrote: > > > > Afterwards I used the following command: > > > > /sbin/lilo -R linux_ext3 rw rootflags=journal=2828 > > > > I receive the following message: > > > > EXT3-fs: error (device ramdisk(1,0)): ext3_get_inode_loc: bad inode > > number: 2828 > > EXT3-fs: invalid journal file > > That simply indicates that "2828" was not the number of a valid inode > when ext3 came to build a journal on your root partition.That's what I also thought, but with #ls -i I get the result 2828 /journal.dat> > > EXT3-fs: get root inode failed > > EXT3-fs: Unrecognized mount option journal > > kernel panic ... > > Known problem, fixed since 0.0.4b I think (it's impossible to be > certain it is the same problem if you don't supply the kernel oops > trace).I don't get a kernel oops. The only message I get is: kernel panic:VFS: Unable to mount rootfs on 01:00 and no more tracing is possible - or what do I have to do to get more results? SysRq also didn't function.> The current development tree still has a few problems of its > own to be ironed out, but once I release 0.0.5 this should be fixed. > The fix simply removes the oops when you perform a journal-create on > an illegal inode, though: it doesn't automatically stop you from > specifying the wrong inode in the first place!I already set up ext3 on the root-partition of two further machines and it works pretty well there, but on this machine I couldn't set it up, because of this error. Any suggestions? cheers, Joachim -- Joachim Kunze Alte Marktstrasse 16 Tel.: +49-7042-830006 D-71665 Horrheim Fax: +49-7042-830006 Germany eMail: joachim@resistance.de
Hi, On Sat, Nov 04, 2000 at 09:20:35PM +0100, Joachim Kunze wrote:> > > On Thu, Nov 02, 2000 at 11:25:54PM +0100, Joachim Kunze wrote: > > > > > > Afterwards I used the following command: > > > /sbin/lilo -R linux_ext3 rw rootflags=journal=2828 > > > I receive the following message: > > > EXT3-fs: error (device ramdisk(1,0)): ext3_get_inode_loc: bad inode > > > number: 2828Are you using an initrd? An init ramdisk setup means that the root filesystem is actually inside the kernel image, and the on-disk root only gets mounted later. In that case, the "rootflags" may end up being applied to the wrong filesystem. Cheers, Stephen