Place Ghantoos

A week ago, I encountered some space problems on my FreeBSD system during a system wide portupgrade. Fortunately, I had backed up my system just before the adventure, so my system was saved.

The procedure I followed to restore my data, was:
1- recreate my partitioning, giving 10 more Gigs to the system, using the FreeBSD installation CD
2- copy all the data on it

Since I backed up my FreeBSD system on an external ext3 hard drive, I needed to copy it back from “outside” my freebsd system. My first option was to use Fixit from the FreeBSD installation CD set, but I had problems mounting my partition because of the inode block size of my ext3 partition. Bref. I decided to mount everything from my Linux (ubuntu) system, and copy everything back from there!

I only way to do this was to recompile my kernel with the right “UFS write” flag. And it JUST WORKED!! I copied back __ALL__ my FreeBSD system on the new slices, rebooted my machine, and here I am writing this post from my old-new FreeBSD system! :)

What I did

By default, up to the Linux kernel 2.6.29, UFS file system write support is considered as “dangereous/experimental” and so, not set. Which means that, by default, a Linux machine will only mount your UFS partitions in readonly:

$ sudo mount -t ufs -o ufstype=ufs2 /dev/sdb5 /mnt/
mount: wrong fs type, bad option, bad superblock on /dev/sdb5,
       missing codepage or helper program, or other error
       In some cases useful info is found in syslog - try
       dmesg | tail  or so

$ dmesg | tail -n 1
[  871.777769] ufs was compiled with read-only support, can't be mounted as read-write

In order to actually mount it, you need to specify the readonly flag:

$ sudo mount -t ufs -o ro,ufstype=ufs2 /dev/sdb2 /mnt/

First of all, you need to download you kernel’s sources, and copy the .config of your current kernel:

$ sudo -s
# apt-get install linux-source-2.6.27
# cd /usr/src
# bzip2 -dc linux-source-2.6.27.tar.bz2 | tar xf -
# cp /boot/config-2.6.27-2-686 .config

To add the UFS write support to your kernel, uncomment/add the CONFIG_UFS_FS_WRITE flag in the config of your kernel, then re-compile it:

$ grep  CONFIG_UFS_FS_WRITE /usr/src/linux-source-2.6.27/.config
CONFIG_UFS_FS_WRITE=y

To compile my kernel, I followed the instruction here. I just added CONCURRENCY_LEVEL=2 to use the 2 cores of my CPU.

$ sudo -s
# CONCURRENCY_LEVEL=2 make-kpkg --initrd --append-to-version=-2-686-ufs kernel-image kernel-headers

Once the compilation is through, install your freshly built kernel (you should find the .deb files in /usr/src/) , and reboot on it.

$ sudo dpkg -i /usr/src/linux-image-2.6.27-1-686-ufs_2.6.27-1-686-ufs-10.00.Custom_i386.deb

NOTE for debian users: If you get kernel panics when booting. It could be because of the missing initrd associated to the new kernel. In this case, if you are using kernel-package-12.009 (or higher?) make sure you copy initramfs hooks so that your package generates the initrd uppon installation (and removes it when package is uninstalled):

cp /usr/share/kernel-package/examples/etc/kernel/postinst.d/initramfs /etc/kernel/postinst.d/
cp /usr/share/kernel-package/examples/etc/kernel/postrm.d/initramfs /etc/kernel/postrm.d/

Once you have booted on your new “UFS-writable” kernel:

$ sudo fdisk -l /dev/sdb

Disk /dev/sdb: 500.1 GB, 500107862016 bytes
255 heads, 63 sectors/track, 60801 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Disk identifier: 0x8f8000b1

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1               1        1958    15727603+  83  Linux
/dev/sdb2   *        1959        4569    20972857+  a5  FreeBSD
/dev/sdb3            4570        4830     2096482+  82  Linux swap / Solaris
/dev/sdb4            4831       60801   449587057+  83  Linux

In this case, /dev/sdb2 contains my UFS slices.
If we take a closer look at /dev/ we will be able to see all the slices (this actually works by default)

$ ls -l /dev/sdb*
brw-rw---- 1 root disk 8, 16 2009-04-04 16:20 /dev/sdb
brw-rw---- 1 root disk 8, 17 2009-04-04 14:20 /dev/sdb1
brw-rw---- 1 root disk 8, 18 2009-04-04 16:20 /dev/sdb2  <--- FreeBSD paartition
brw-rw---- 1 root disk 8, 19 2009-04-04 16:20 /dev/sdb3
brw-rw---- 1 root disk 8, 20 2009-04-04 14:20 /dev/sdb4
brw-rw---- 1 root disk 8, 21 2009-04-04 16:20 /dev/sdb5  <--- FreeBSD / slice
brw-rw---- 1 root disk 8, 22 2009-04-04 16:20 /dev/sdb6  <--- FreeBSD swap
brw-rw---- 1 root disk 8, 23 2009-04-04 16:20 /dev/sdb7  <--- FreeBSD /var slice
brw-rw---- 1 root disk 8, 24 2009-04-04 16:20 /dev/sdb8  <--- FreeBSD /tmp slice
brw-rw---- 1 root disk 8, 25 2009-04-04 16:20 /dev/sdb9  <--- FreeBSD /usr slice

Then, I mounted the slices in the right order and to the right mount points (in /mnt/) to have my FreeBSD exact filesystem environment. Then I copied back all the data using the old and nice tar technique, as root:

# ( cd /path/to/backup && tar cpf - . ) | ( cd /mnt/ && tar xvfp - )

Once the "un"-tar was done, I rebooted on my FreeBSD system (from which I am writing this post), forced the fsck of all my slices: everything seems to be OK!

Conclusion

The CONFIG_UFS_FS_WRITE kernel flag is still set as experimental, but it did work like charm for me (using kernel 2.6.27).
If you do use it, and you find a bug, please report it refering to this doc :)

Hope this was helpful.

Cheers,
Ignace M -ghantoos-

Sources

http://www.quietearth.us/articles/2006/09/15/Ubuntu-Compiling-a-custom-kernel
http://wiki.netbsd.se/How_to_mount_FFS_partition_under_Linux
http://kernel.xc.net/html/kconfig-2.6.29_x86.html