> On Wed, Dec 08, 1999 at 12:38:19PM +0100, Josef Höök wrote:
> > I must disagree on that. If your disk where you've put the journal
> > breaks
> > then you have a long night ahead restoring backups.
> > I can only refer to Aix which has a journal log for every disk,
> > one journal / disk and all filesystems within.
Well, of course, I'm (used to be) very used to AIX, so this is where some
of my thoughts have their origin.
I think, what I would like to have for ext3 (or ext2) is a mount option
like 'journal=<filename>', where <filename> is an existing file on ANY
filesystem or a <devicenode> (full disk, disk partition, software raid
disk....)
If the same <filename> is used as a journal by several filesystems it
should nicely interact and know which journal entry belongs to which
filesystems. It is definitely difficult to replay/finalize a journal if
not all filesystems are mounted, maybe even unmounted ones could be
replayed as long as the disk is accessible. It's still a nice/good to
have, IMHO.
This way, in the end state, there won't be a real ext3 fs, but ext2 with
journal option would be ext3. An unclean ext2 with journal should just not
mount w/o journal option, a clean ext2 should not replay a journal maybe
given to it in a new mount (but start to use it in the future)
There should be a force option to e2fsck that does a normal ext2 fsck even
on an unclean unmounted ext3 where the journal is not given/accessible
(and converts it to a plain ext2). This is to recover from a lost journal.
A used journal file (or the intermediate journal.dat file of a fixed
visible name) should probably give ETXTBUSY on any attempt to remove or
modify it. Maybe that's a good feature even for the intermediate
journal.dat file.
*I* think this is what you want. You can then setup a journal per disk, or
one for the system (but maybe on a raid system) or whatever. Leave it to
the user to decide the level of security versus consumpution of space
versus speed. However, I just *think* noone wants a fixed journal on EACH
fs on the same disk. Now, I may be wrong, nor do I want to force anyone
not to have a jornal per fs even on the same disk.
All I mean is: rather than spent much time to invent odd ways to hide a
journal.dat file from a user to keep him from shooting himself in the
food, better see you can put a journal on each disk (or even one per
system). I understood the ext3 docs in a way that this is the real goal
and journal.dat file is nothing else than a debugging aid right now. You
can not design a fool proof system (that's what windows tries, see how
well they do). I simply do not want/accept a loss of feature/flexibility
just to prevent someone from shooting himself in the foot. Those people
better install Win98, I do not care about them. A separate journal
partition per disk is IMHO what anyone and any distrib is going to setup.
Now, IMHO...
Finally, I'm not familiar with the ext2/3 code, nor do I have time
currently to get familiar and add stuff, although I'd love too (really!).
I hate to get stupid comments on my own linux programs from people that
never wrote a line of code for it and are clueless. So, I will now
definitely shut up.
(It's still a pity that noone answered my embedded question: does a
read-only mount of an ext3 work now or not?)
Michael.
--Michael Weller: eowmob@exp-math.uni-essen.de, eowmob@ms.exp-math.uni-essen.de, or even mat42b@spi.power.uni-essen.de. If you encounter an eowmob account on any machine in the net, it's very likely it's me.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/