RedHat 9 on Michael’s Domain

Changing the Runlevel (inittab, telinit)

Sat, 26 May 2007 00:00:00 +0000

Usually you change the default runlevel by editing /etc/inittab. However, sometimes you boot into runlevel 3 (e.g. by selecting safe mode from the grub menu) and then want to change to runlevel 5. This you accomplish on the fly with the command: telinit 5In RedHat 9, just change in the /etc/inittab file the runlevel from 5 to 3 and you will have only the command line left by default upon booting. No X11 graphical login etc. In newer Ubuntu releases, this apparently doesn’t work anymore, although telinit + number should still change the runlevel, but at least in the Precise Pangolin (12.04), telinit 2-5 appears not to do anything. Telinit 1 tries to shutdown the machine, but the shutdown got stuck halfway when I tested it. To achive something similar to changing to runlevel 3, you will have to shutdown the GUI (meaning to kill the gdm or lightdm process). The runlevel concept seems to be dead.

Mounting disk image files and encrypted filesystems/partitions

Sat, 26 May 2007 00:00:00 +0000

To dump a disk (or a partition) into a file, you use the following command:dd if=/dev/hda3 of=file.binThis command writes the complete data on the hda3 partition into the file file.bin.To mount the filesystem that is in the file, you need to create a loopback device:losetup /dev/loop0 file.binNow you can mount the filesystem as usual:mount -r -t filesystemtype /dev/loop0 /mnt/mountpointApparently, one can combine the last two commands into one:mount -t filesystemtype -o loop ./file.bin /mnt/mountpointIf you want to encrypt the data on a file, that contains a whole filesystem, it is getting a bit more complicated:sudo mkdir /mnt/secure (create mount point for the filesystem)dd if=/dev/zero of=path/to/file bs=1k count=409600 (create an empty file with 400 MB size) sudo /sbin/losetup -e xor /dev/loop0 path/to/file/sbin/mkfs -t ext2 /dev/loop0 409600 (format the device as ext2)sudo mount -t ext2 /dev/loop0 /mnt/secure (mount the device file)cd /mnt/securechown username . (change the owner of the top level directory of the filesystem)If you want to unmount the filesystem:sudo umount /dev/loop0If you want to get rid of the filesystem, you have to un-associate it from the loop device 0: /sbin/losetup -d /dev/loop0For some reason RedHat 9 doesn’t come with DES support, so for the time being (until I patch the kernel or move to Suse Linux) I am using the faster, but much weaker xor encryption.Suse 9 comes with inbuilt strong encryption and offers already during the installation the possibility to create an encrypted partition. Suse 9 asks during booting for the passphrase to mount the encrypted filesystem. The boot process stops and waits for 2 minutes before continuing if you don’t type in the password. In order to reduce this time, you can edit the file /etc/init.d/boot.crypto. Change in the following line 120 to e.g. 10::${TIMEOUT:=120}If you have missed your chance to type in the passphrase during boot time, you can mount the encrypted partition as follows:/sbin/losetup -e twofish /dev/loop0 /dev/hda7 mount /dev/loop0 /media/confBTW: The information about encrypted filesystems resides in /etc/cryptotab.

rio500-0.7-1.i386.rpm works under RedHat 9

Sat, 26 May 2007 00:00:00 +0000

I again tried to compile rio500-0.8.1, but without success. Compiling fails under kernel 2.4. This is an unresolved issue and mentioned in the discussion forums. So I installed the Rio500 package rio500-0.7-1.i386.rpm. Interestingly I managed via the command line to a) format the internal memory: rio_formatb) create a folder (which is necessary before one can upload a song): rio_add_folder c) upload mp3s: rio_add_song ). That’s a big advantage to the crappy Windows Rioport interface. Next time I try to install the Gnome GUI.When listening to the first songs on the Rio500 that I transferred under Linux, they appeared very silent. When I encode next time some songs I should try the lame option –scale . However, I don’t know what argument to use to get some decent amplification of the sound.

Suse 9

Fri, 25 May 2007 00:00:00 +0000

I switched to Suse 9. I can’t figure out from RedHat’s announcements what they are really up to. So I better get used to an alternative right now. I also switched from Gnome to KDE and at least on Suse 9, I like KDE more than Gnome on Red Hat 9.

Adding a Suse 9 to a Red Hat 9 (two different distros on the same computer)

Wed, 23 May 2007 00:00:00 +0000

Red Hat 9 is installed and the following partitioning is present:hdd1 /boot (ext3) 100MBhdd2 /home (reiserfs) 7.8GBhdd3 / (ext3) 7.8GBhdd4 EXThdd5 swap 1GBhdd6 /mnt/documents (reiserfs) 12GBhdd7 /mnt/music (reiserfs) 90GBThe task is to add a second Linux OS (Suse 9) onto this hard disk without destroying anything. The root partition of this system should be 10 GB. Additionally I want to put all user-created files on one partition (now they are on three different partitions: hdd2, hdd6 and hdd7). First I got a 200 MB external USB2/firewire hard drive (Maxtor, mounted as /mnt/sdb1) and copied the data from hdd1, hdd2, hdd6 and hdd7 onto it with “cp -ax /home /mnt/sdb1”, etc. Then I rebooted using the Knoppix 3.3 CD and from a root terminal executed partimage. I tried to save the image of the root partition directly to my external usb drive, but without success: everytime the program halted somewhere during the process. I then saved the image on /mnt/music and that succeeded. I rebooted into Red Hat 9 and copied the image manually from /mnt/music to the usb drive (/mnt/sdb1). Because I didn’t trust the partimage program, I additionally copied the filesystem manually to the usb drive using the following commands: mkdir /mnt/sdb1/redhat_root cd / find . -xdev -print | cpio -padm /mnt/sdb1/redhat_root Apparently this copies everything including special files, mount points, etc. preserving all the file meta data. After this I disconnected the external usb drive (to be on the safe side), rebooted into Knoppix and wanted to try out QTParted. It should be able to resize partitions. From a root terminal: “qtparted”. However, I was not able to resize anything (the options were greyed out). I could delete and recreate though (which didn’t help much). So I decided to go for a clean sweep, popped the Suse 9 CD into the drive and rebooted into the Suse 9 installer. I choose expert partitioning and deleted all partitions that were present and recreated the following partitions. I made sure to use exactly the same amount of blocks (1019) for hdb3 as had been used for hdd3 because I knew that partimage can only restore images onto partitions that have exactly the same size or are bigger:/dev/hdb1 1 8 64228+ 83 Linux (ext3) /boot/dev/hdb2 9 1314 10490445 83 Linux (reiserfs) //dev/hdb3 1315 2333 8185117+ 83 Linux (reiserfs) /redhat/dev/hdb4 2334 14945 101305890 f Win95 Ext'd (LBA)/dev/hdb5 2334 2464 1052226 82 Linux swap/dev/hdb6 2465 14945 100253601 83 Linux (reiserfs) /homeAfter the install had finished, I rebooted into Knoppix and tried to restore the Red Hat root partition onto hdb3. Partimage told me that the partition was smaller than the image, thus a restore was not possible. So rebooted into Suse 9 and manually copied the Red Hat 9 root system back to hdb3 (with the same command sequnce that was used to copy it manually to the usb drive). Then I changed /redhat/etc/fstab to reflect the new partitioning and copied all the files from /boot (but not the subdirectories from /boot) onto the new boot partition (except those files that had already equivalents in /boot). Then I modified /boot/grub/menu.lst by adding a line for the Red Hat install:title Red Hat 9 Linuxroot (hd1,0)kernel /vmlinuz-2.4.20-19.9 root=/dev/hdb3initrd /initrd-2.4.20-19.9.imgI also create the two user folders in the /home directory (which used to be the mount point for the /home partition). But I copy all the hidden files and directories (those starting with a .) into the userfolders.Then I tried to reboot into Red Hat 9. Without success. Kernel panic. No init found. Try passing init=option to the kernel. Actually the file system check of hdb3 failed (no superblock found). There are several error messages and the important ones come first and fast (and you don’t see them) and the bad, misleading error messages come late. Something like that you should manually repair the partition. And that root is mounted read-only. And that you can make it read-write by “mount -n -o remount,rw /”. And that only with Ctrl-D one can reboot into this state. Obviously that’s what I am going to do. However there is no hint about what command to use. Several attempts to do anything while having mounted from hdb failed (if reiserfs is read-only it cannot be repaired, if it is read-write it cannot be checked). So I reboot into Knoppix and do the repair from there: fsck (which executes reiserfsck) doesn’t do the job: it only checks, but doesn’t do any repairs. I figured that flagging reiserfsck –rebuild-sb /dev/hdb3 does repair the superblock. Still no go, so I do the slow complete rebuild of the filesystem from scratch with reiserfsck –rebuild-tree /dev/hdb3. It still doesn’t boot into RH9. Then I get the idea: the RH9 kernel might not support reiserfs (or the module is not loaded). Or (alternatively) copying some special files from a ext3 partition to a reiserfs partition might simply not work. So I delete (from Knoppix) the whole hdb3 and recreate it as ext3. Copy back all the data (QTParted still complains about too small size). And it finally works. I create two directories for the two users on the /home partitons and manually copy into them the non-hidden files of the old userfolders. I don’t copy the hidden files. There are many that would screw up Suse and make it impossible to log in (I had tried that out before). I log into Suse 9 as root and create the two users with exactly the same name as on RH9. There is a warning that the directory exists and that all its content will be owned by the new user. That is OK of course.