Enabling linux ext3 full data journal mode

[cleanbash]

I was wondering if anyone could help me with the following problem.

For a project I'm building some kind of failsafe linux system which users can
switch off with the power button (no monitor attached). Because the system
must be failsafe I'm using a journaling file system (ext3) (in case of a power failure
or someone who accidentally disconnects the power plug).

As far as I know ext3 is using ordered mode by default so it only journals
metadata. Because I want a failsafe system it's better to use full data journaling
(data=journal).

I've searched almost a whole day on google and tried several options but it
still doesn't work.

On several sites I see that I have to add the following flag to my bootloader

rootflags=data=journal

and add this to my fstab:

data=journal

- If I don't add the rootflags then the system is getting read/write errors because the filesystem is read only.
- If I don't add the line to my fstab but only the line in my bootloader then the system gets a kernel panic.
- If I add both lines the system also gets a kernel panic.

System info: linux box (mandrake 10.1) kernel 2.6.8.1 with lilo as boot loader, kernel support
for ext3 filesystems (compiled in kernel itself and not as a module).

Who can help me to enable full data journaling?

[SirDice]

If I don't add the rootflags then the system is getting read/write errors because the filesystem is read only.

If the filesystem is read-only, why enable journaling?
If the filesystem is read-only it cannot get corrupted? (except during a harddisk failure)

Not sure if linux uses the same scheme but on BSD I would..

mount / and /usr read-only, mount /tmp on a memory disk and /usr/home and /var on a journaling system. /home (/usr/home) and /var should be the only filesystems that change (during normal operations) and are important enough to save. It shouldn't matter if /tmp gets nuked..

[cleanbash]

@SirDice

Sorry I forgot to mention that readonly is the error message I get. I want the
root filesystem to be read/write but when I set one of these options then the
system complains that it can't write to the disk because it's mounted ro.

[SirDice]

odd... What does a mount tell you when you get that message? Is it really mounted read-only?

[cleanbash]

Tomorrow I will look what the error messages are and I shall post the boot.log
and the exact settings I made.

By the way I'm still looking on the internet but there's not very much I can find
about this problem.

[SirDice]

I'm not a linux user, so I'm not sure about certain options, but I found these in a couple of minutes:
http://www-128.ibm.com/developerwork...one=opensource
http://www.zip.com.au/~akpm/linux/ext3/ext3-usage.html

[Trevoke]

reiser4 performs 'atomic' operations; in other words, data cannot get corrupted through a power-off during transfer.
On the other hand, I don't know how much sits in swap before the transfer is done. It is not in the kernel yet though it's easy enough to patch it; and I've never had a problem with reiser4 getting corrupted (and I have had some unclean shutdowns).

[cleanbash]

Small update:

@SirDice thanks but I've already read those two links ;-)

A few days ago I posted that I should upload the logfiles. Because the filesystem is in read
only mode (god knows why) the logfiles cannot be updated.

I know that the system first mounts the filesystem in ro mode but the it should switch to
rw mode which is not happening at my system when I turn full journaling on. Anyway
I'm still searching for a solution, when I find it I'll post it here.

[cleanbash]

I solved the problem,

First use tune2fs to turn full journaling on on the desired partition:

tune2fs -o +journal_data /dev/hdaX

then change the /etc/fstab file, add the following line in the option field of the desired partition:

data=journal

Reboot and check the journaling mode with:

tune2fs -l /dev/hdaX

it should display the journaling mode with a lot of other info.