Andrew writes:
> It's probably worth thinking about adding a fourth journalling
> mode: `journal=none'.

Yes, I had added this (at least in skeleton form) in my ext3 tree.
If only I could keep up with you and Daniel for both the ext3 and
indexed directory stuff, I might be able to submit it...

Basically, I wrapped all of the ext3 journal operations in an inline
ext3_<journal_op> and did nothing if handle was NULL (except
for the journal_dirty_metadata, which went to mark_buffer_dirty().
You could also just make them (mostly) no-ops if CONFIG_EXT3_FS
was not defined and we were living in the ext2 tree.

We may want to keep the orphan list handling even for ext2, because
e2fsck does orphan cleanup regardless of whether the filesystem has
a journal, and it would the amount of output from e2fsck. Doesn't
matter much either way.

> Roll it all up into a single codebase and call it ext4.

Rather just stick with ext3 for now.  Don't want to confuse
the issue even more.  If we can get the ext3 code mounting ext2
filesystems (i.e. without a journal), then we can slowly merge
the changes back to stock ext2 surrounded by CONFIG_EXT3_FS
(or not, as Linus dictates).

> It rather depends on where the buffercache ends up.  ext3 is a
> client of JBD (nee JFS).  JBD does *block* level journalling.  Any
> major change at that level will take rather some adjusting to.

Well, Daniel's part of the code still uses buffer_heads, but they
are backed by the page cache and not the buffer cache.  This is
the direction Linus wants to go (AFAICS), that we use the page
cache for cacheing, and buffer_heads for I/O handles only.  Al's
page-cache directory stuff does not use Daniel's buffer_head
abstraction, so it may need to get changed a bit to work with JBD.

At this point, it is probably worth doing a global search-and-replace
for all of the jfs_* functions, and rename them jbd_*, to avoid
conflicts with IBM JFS.

Cheers, Andreas
-- 
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert
-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to [EMAIL PROTECTED]

Reply via email to