Messages in this thread | | | Date | Mon, 6 Dec 1999 15:13:37 -0800 (PST) | From | Brion Vibber <> | Subject | Re: Ext3 Filesystem |
| |
On Mon, 6 Dec 1999, Byron R. Stanoszek wrote:
> My thoughts on ext3: > > When looking into ext3, I thought it would be a completely different file > system apart from ext2. However, I was surprised to see that most of > the filesystem is the same except for using that journal.dat file. So I was > wondering, couldn't this journaling filesystem really just be a flag added > to one of the reserve bits in the ext2 header that labels it as such?
Actually that's exactly what it is... giving it a separate name just makes it safer since the stable ext2 code can be run simultaneously on other filesystems that don't need the journaling while the experimental ext3 eats your test partitions if something goes wrong.
> Also, I've read recently about the problems with root seeing journal.dat, > being able to modify it, etc. I perfectly agree with making this file a hidden
Yeah, that was me. :) Turns out making the journal file immutable (chattr +i journal.dat) works fine, it protects the file from casual overwriting without affecting its actual usage. Of course you can still overwrite it deliberately with a chattr -i... but then if you really want you can write straight to the block device anyway and cause all kinds of havoc!
> inode, but I agree to the point where it does not show up in any directory, > and is instead a fixed inode number that is 1) a statically set # either on > creation of the filesystem,
Mmm, this would make it difficult/impossible to migrate an existing ext2 filesystem to journaling. Being able to eliminate fsck with ten minutes of kernel recompiling and 30 seconds of command line work was my favorite part of ext3...
> or 2) an inode number physically stored into the > superblock when a ext2 filesystem is converted to journaling.
That's how it works actually, you only need to specify the journal inode on the first mount.
[snip] > I feel the file shouldn't actually be part of the directory hierarchy, but > stored within an normally-unaccessible portion of the ext2 partition.
I'm not sure how much I like having hidden mysterious things in my filesystems that I can't see... of course if I can't see them, I can't step on them. :) If there's a nice tool to create and configure the hidden journal, then I don't think I'd mind.
Is there any way to change the journal size later on though? It seems like it ought to be possible to do if for some reason one must. Harder with things hidden...
e2fsck would probably need to be taught not to treat the journal as a lost file too, otherwise you just have your journal as /lost+found/#12 if you ever do run fsck. "Hmm, that file doesn't look important, I'll chuck it. Whoops..."
-- brion vibber (brion@pobox.com)
- 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/
| |