[LWN Logo]
[LWN.net]
From:	 "Stephen C. Tweedie" <sct@redhat.com>
To:	 ext3-users@redhat.com, linux-fsdevel@vger.kernel.org,
	 ext2-devel@lists.sourceforge.net
Subject: ext3-0.0.7a for 2.2.19 is released
Date:	 Wed, 6 Jun 2001 19:06:36 +0100
Cc:	 Stephen Tweedie <sct@redhat.com>

Hi all,

ext3-0.0.7a is now out, at the usual places:

	ftp.uk.linux.org:/pub/linux/sct/fs/jfs/

and propagating now from

	ftp.*.kernel.org:/pub/linux/kernel/people/sct/ext3/

This fixes one major bug in ext3 recovery for metadata-only
journaling.  Because recovery can also happen in e2fsck, users should
also upgrade to at least e2fsprogs-1.21-WIP-0601 (also provided at the
above ftp sites).  Thanks to Andrew Morton for finding that bug, but
unfortunately it was found just after e2fsprogs-1.20 was released.

Full changes for this release are included below.

Cheers,
 Stephen



Changes in this release
-----------------------

0.0.7a:

**************************************************************** 
The major change in this release is a missing byte-endian conversion
in the recovery code.  This problem affects both the ext3 kernel
recovery and the recovery code in e2fsck.  Please upgrade to
e2fsprogs-1.21-WIP-0601.tar.gz or later.  You can find this on
ftp://download.sourceforge.net/pub/sourceforge/e2fsprogs
**************************************************************** 

The other significant change is a fix when running shared mmaped files
on ext3 with more than one ext3 filesystem active: it was possible for
ext3 to get confused about which filesystem was active if the VM tried
to swap out a page on one filesystem while processing a filesystem
update on the other filesystem.

The other changes include:

 Pin the buffer heads backing an inode to a transaction when the
 transaction starts, to prevent the inode being flushed to disk before
 transaction commit.

 Add a debugging check to trap any attempts to run a nested
transaction
 inside a transaction on a different filesystem.

 Fix an assertion failure that could occur after allocating an
 indirect block but failing to allocate a block beneath it due to disk
 full.

 Fix a bug which meant that writes larger than 4k were never being
 coalesced into a single transaction.

 Rename the confusingly-named "tid_ge" transaction-greater-than
 function to "tid_gt"

 Update the README to describe the new ext3 features of current
 e2fsprogs.


-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org