Whamcloud - gitweb
tools/e2fsprogs.git
16 years agoTeach ext2fs_extent_insert() to split the current node if necessary
Eric Sandeen [Tue, 20 May 2008 15:15:27 +0000 (10:15 -0500)]
Teach ext2fs_extent_insert() to split the current node if necessary

If ext2fs_extent_insert finds that the requested node
for insertion is full, it will currently fail.

With this patch it will split as necessary to make room, unless an
EXT2_EXTENT_INSERT_NOSPLIT flag is passed to it.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agolibext2fs: Teach extent.c how to split nodes
Eric Sandeen [Tue, 20 May 2008 15:14:20 +0000 (10:14 -0500)]
libext2fs: Teach extent.c how to split nodes

When called for a given handle, the new function extent_node_split()
will split the current node such that half of the node's entries will
be moved to a new tree block.  The parent will then be updated to
point to the (now smaller) original node as well as the new node.

If the root node is requested to be split, it will move all
entries out to a new node, and leave a single entry in the
root pointing to that new node.

If the reqested split node's parent is full it will recursively
split up to the root to make room for the new node's insertion.

If you ask to split a non-root node with only one entry,
it will refuse (we'd have an empty node otherwise).

It also updates the i_blocks count when a new block has
successfully been connected to the tree.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agoext2fs_extent_open: If the inode is empty, initialize the extent tree
Theodore Ts'o [Mon, 2 Jun 2008 04:08:19 +0000 (00:08 -0400)]
ext2fs_extent_open: If the inode is empty, initialize the extent tree

If the inode's i_block[] array is completely empty, create an empty
extent tree in the in-core inode and set the EXT4_EXTENT_FL inode
flag.  This makes it easy to create a new inode using extents.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext2fs_extent_replace: Support uninit extents and check validity of e_len
Theodore Ts'o [Mon, 2 Jun 2008 02:56:37 +0000 (22:56 -0400)]
ext2fs_extent_replace: Support uninit extents and check validity of e_len

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoe2fsck: Don't double count an extent after deleting the last extent
Theodore Ts'o [Fri, 30 May 2008 19:23:24 +0000 (15:23 -0400)]
e2fsck: Don't double count an extent after deleting the last extent

ext2fs_extent_delete() will leave the extent handle pointing at the
next extent --- except if the last extent in the node.  To deal with
this last case, call ext2fs_get_extent_info() and stop scanning after
processing info->num_entries extents.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoe2fsck: Don't skip an extent after deleting an invalid extent
Theodore Ts'o [Wed, 28 May 2008 08:54:44 +0000 (04:54 -0400)]
e2fsck: Don't skip an extent after deleting an invalid extent

ext2fs_delete_extent() deletes the current extent and moves to the
next extent (if present).  So we need to skip moving to the next
extent and get the (new) current extent and check it before moving on.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoRemove bashism in test script for libblkid
Theodore Ts'o [Sun, 25 May 2008 23:45:29 +0000 (19:45 -0400)]
Remove bashism in test script for libblkid

Thanks to Christian Kujau for pointing this out.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoAdd 64-bit support to the test_io manager
Theodore Ts'o [Sat, 24 May 2008 23:17:01 +0000 (19:17 -0400)]
Add 64-bit support to the test_io manager

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoFix m_large_file and r_inline_xattr tests to match with recent changes
Theodore Ts'o [Sat, 24 May 2008 23:10:49 +0000 (19:10 -0400)]
Fix m_large_file and r_inline_xattr tests to match with recent changes

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoMerge branch 'maint'
Theodore Ts'o [Wed, 21 May 2008 20:56:42 +0000 (16:56 -0400)]
Merge branch 'maint'

Conflicts:

README
version.h

16 years agoUpdate release notes and debian changelog for 1.40.10 release v1.40.10
Theodore Ts'o [Wed, 21 May 2008 17:23:04 +0000 (13:23 -0400)]
Update release notes and debian changelog for 1.40.10 release

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoRemove default sizeof sizes in configure script when cross-compiling
Maciej W. Rozycki [Tue, 20 May 2008 23:38:12 +0000 (00:38 +0100)]
Remove default sizeof sizes in configure script when cross-compiling

Since version 2.50 autoconf fully supports checking sizes of types
(with AC_CHECK_SIZEOF) when cross-compiling.  Therefore there is no
need to preset the respective cache variables anymore.  The following
patch removes the special case.  There is no need to adjust AC_PREREQ
as it's set to 2.50 already.

Tested successfully cross-building for the mips64el-linux-gnu host on
an i386-linux-gnu build system, removing the following warning
(because of a mismatch for the "long" type):

Sizeof(__U64__TYPEDEF) is 4 should be 8
Problem detected with asm_types.h

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoblkid: If the device mtime is newer that the cache time, force a revalidation
Theodore Ts'o [Tue, 20 May 2008 21:28:54 +0000 (17:28 -0400)]
blkid: If the device mtime is newer that the cache time, force a revalidation

This fixes problems turned up by a test case written by Erez Zadok's
group which constantly reformats filesystems.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoe2fsck: Fix potential data corruptor bug in journal recovery
Theodore Ts'o [Tue, 20 May 2008 18:51:14 +0000 (14:51 -0400)]
e2fsck: Fix potential data corruptor bug in journal recovery

While synchronizing e2fsck's recovery.c with the latest 2.6 kernel
sources, I discovered a serious bug that apparently had been fixed in
the kernel sometime between Deceber 2003 and April 2005, but which had
not been carried over to e2fsprogs.  Specifically, when blocks whose
first 4 bytes are JFS_MAGIC_NUMBER (0xc03b3998) are written into the
journal, the first 4 bytes zero'ed out.  A one character typo meant
that when the blocks were replayed by e2fsck, the JFS_MAGIC_NUMBER
would not be restored.

Oops.

Fortunately, it is *highly* unlikely that ext4 metadata blocks will
contain that magic number in the first four bytes, and data=journalled
is a relatively rarely used.

This commit fixes this bug, as well as updating e2fsck's recovery.c to
be in sync with 2.6.25.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agodebugfs: Print the nanosecond field and i_version field of an inode
Theodore Ts'o [Fri, 16 May 2008 02:17:06 +0000 (22:17 -0400)]
debugfs: Print the nanosecond field and i_version field of an inode

Add support for dumping out the new inode fields added in ext4

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext2fs.h: Add l_i_file_acl_high and l_version_hi to on-disk inode structure
Theodore Ts'o [Fri, 16 May 2008 02:14:52 +0000 (22:14 -0400)]
ext2fs.h: Add l_i_file_acl_high and l_version_hi to on-disk inode structure

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agopo: update vi.po (from translationproject.org)
Clytie Siddall [Wed, 14 May 2008 22:42:45 +0000 (18:42 -0400)]
po: update vi.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agopo: update sv.po (from translationproject.org)
Göran Uddeborg [Wed, 14 May 2008 22:42:45 +0000 (18:42 -0400)]
po: update sv.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agopo: update nl.po (from translationproject.org)
Benno Schulenberg [Wed, 14 May 2008 22:42:45 +0000 (18:42 -0400)]
po: update nl.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agopo: update de.po (from translationproject.org)
Philipp Thomas [Wed, 14 May 2008 22:42:45 +0000 (18:42 -0400)]
po: update de.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agoInstall {mkfs,fsck}.{ext4,ext4dev} binary and man pages
Theodore Ts'o [Wed, 14 May 2008 22:23:47 +0000 (18:23 -0400)]
Install {mkfs,fsck}.{ext4,ext4dev} binary and man pages

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoAdd ext4dev stanza to mke2fs.conf
Theodore Ts'o [Wed, 14 May 2008 22:23:16 +0000 (18:23 -0400)]
Add ext4dev stanza to mke2fs.conf

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agomke2fs: Add the ability to configure default extended options in mke2fs.conf
Theodore Ts'o [Wed, 14 May 2008 22:09:37 +0000 (18:09 -0400)]
mke2fs: Add the ability to configure default extended options in mke2fs.conf

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agofix extent_goto for non-0 leaf levels
Eric Sandeen [Mon, 12 May 2008 23:13:49 +0000 (18:13 -0500)]
fix extent_goto for non-0 leaf levels

The logic for stopping at the right level in extent_goto was wrong,
so if you asked it to go to any level other than 0 (the leaf
level) it would fail.

Also add this argument to the tst_extents goto command to test it.

(I thought this was a failure in my split code but it was this
helper that was causing problems...)

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years ago- fix swap sanity tests in blkid, and blkid tests
Eric Sandeen [Mon, 12 May 2008 18:26:51 +0000 (13:26 -0500)]
- fix swap sanity tests in blkid, and blkid tests

Swap is actually native-endian on disk, and with the latest
swapspace sanity checks I added we need to have native swapspace
examples in the blkid tests, so re-mkswap them during testing.

One one other required change, though; mkswap requires at least
10 pages of swap, so the image needs to be increased to 10x64k
if mkswap is to succeed...

Maybe it'd be better to just dd it out on the fly?

Addresses-redhat-bugzilla: 445786

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agoFix German translation bug (y,n) -> (j,n)
Theodore Ts'o [Mon, 12 May 2008 13:51:17 +0000 (09:51 -0400)]
Fix German translation bug (y,n) -> (j,n)

Addresses-Sourceforge-Bug: #1961831

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agolibuuid: don't use unitialized variable
Matthias Koenig [Tue, 6 May 2008 11:38:31 +0000 (13:38 +0200)]
libuuid: don't use unitialized variable

Signed-off-by: Matthias Koenig <mkoenig@suse.de>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agoFix parallel build problem
Theodore Ts'o [Mon, 12 May 2008 00:34:12 +0000 (20:34 -0400)]
Fix parallel build problem

Problem was introduced by commit a4b69b7f1809898c87bdc11502ca258a8f16a220

Thanks to Eric Sandeen from Red Hat for pointing out this problem.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoUpdate version.h and README for 1.41-WIP-0427 release v1.41-WIP-0427
Theodore Ts'o [Mon, 28 Apr 2008 02:40:17 +0000 (22:40 -0400)]
Update version.h and README for 1.41-WIP-0427 release

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoMerge branch 'next' into tt/flex-bg
Theodore Ts'o [Mon, 28 Apr 2008 02:24:36 +0000 (22:24 -0400)]
Merge branch 'next' into tt/flex-bg

16 years agoAdd test cases for undoe2fs: u_undoe2fs_mke2fs and u_undoe2fs_tune2fs
Aneesh Kumar K.V [Sun, 27 Apr 2008 22:36:38 +0000 (18:36 -0400)]
Add test cases for undoe2fs: u_undoe2fs_mke2fs and u_undoe2fs_tune2fs

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoFix the resize inode test case
Aneesh Kumar K.V [Mon, 13 Aug 2007 10:26:25 +0000 (15:56 +0530)]
Fix the resize inode test case

With the new mke2fs changes the output of the
command differs if we run mke2fs on a device that
already have the file system. So erase the file system
before running mke2fs so that output remain as expected.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agotune2fs: Support for large inode migration.
Aneesh Kumar K.V [Mon, 13 Aug 2007 10:26:24 +0000 (15:56 +0530)]
tune2fs: Support for large inode migration.

Add new option -I <inode_size> to tune2fs.  This is used to change the
inode size. The size need to be multiple of 2 and we don't allow to
decrease the inode size.

As a part of increasing the inode size we increase the inode table
size. We also move the used data blocks around and update the
respective inodes to point to the new block

tune2fs uses the undo I/O manager when migrating to large inode. This
helps in reverting the changes if end results are not correct.  The
environment variable TUNE2FS_UNDO_DIR is used to indicate the
directory within which the tdb file need to be created. The file will
be named tune2fs-<device-name> If TUNE2FS_UNDO_DIR is not set
/var/lib/e2fsprogs is used

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agomke2fs: Add support for the undo I/O manager.
Aneesh Kumar K.V [Mon, 13 Aug 2007 10:26:26 +0000 (15:56 +0530)]
mke2fs: Add support for the undo I/O manager.

When running mke2fs, if a file system is detected
on the device, we use Undo I/O manager as the io manager.
This helps in reverting the changes made to the filesystem
in case we wrongly selected the device.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext2fs_add_journal_inode: Optimize writing the journal data blocks
Theodore Ts'o [Sun, 27 Apr 2008 20:10:32 +0000 (16:10 -0400)]
ext2fs_add_journal_inode: Optimize writing the journal data blocks

Add a new function ext2fs_zero_blocks(), and use it so that journal
data blocks is written in larger chunks to speed up the creation of
the journal.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoAdd e2undo command
Aneesh Kumar K.V [Mon, 13 Aug 2007 10:26:22 +0000 (15:56 +0530)]
Add e2undo command

The e2undo command can be used to replay the transaction saved in the
transaction file using undo I/O Manager.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agolibext2fs: Add undo I/O manager
Aneesh Kumar K.V [Mon, 13 Aug 2007 10:26:21 +0000 (15:56 +0530)]
libext2fs: Add undo I/O manager

This I/O manager saves the contents of the location being overwritten
to a tdb database. This helps in undoing the changes done to the
file system.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agomke2fs: Make argument to the extended option lazy_itable_init optional
Theodore Ts'o [Sun, 27 Apr 2008 23:38:02 +0000 (19:38 -0400)]
mke2fs: Make argument to the extended option lazy_itable_init optional

Also remove some debugging printf's left over from a previous patch.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoMerge branch 'maint'
Theodore Ts'o [Sun, 27 Apr 2008 11:59:28 +0000 (07:59 -0400)]
Merge branch 'maint'

Conflicts:

tests/m_raid_opt/expect.1

16 years agodebian: The doc-base section should be Programming, not Apps/Programming
Theodore Ts'o [Sun, 27 Apr 2008 11:53:02 +0000 (07:53 -0400)]
debian: The doc-base section should be Programming, not Apps/Programming

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoUpdate release notes and debian changelog for 1.40.9 release v1.40.9
Theodore Ts'o [Sun, 27 Apr 2008 04:53:27 +0000 (00:53 -0400)]
Update release notes and debian changelog for 1.40.9 release

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agofsck does not accept 'j' with German locale
Makoto Dei [Mon, 21 Apr 2008 07:49:48 +0000 (16:49 +0900)]
fsck does not accept 'j' with German locale

"yY" should be translated to "jJ" instead of "yJ"

Addresses-Sourceforge-Bug: #1947683

Signed-off-by: Makoto Dei <makoto@turbolinux.co.jp>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agomke2fs: Don't allow the combination of the meta_bg and resize_inode features
Theodore Ts'o [Sun, 27 Apr 2008 04:08:14 +0000 (00:08 -0400)]
mke2fs: Don't allow the combination of the meta_bg and resize_inode features

The combination of meta_bg and resize_inode leads to a corrupt
filesystem, and it's not really clear it makes any logical sense.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agomke2fs: New bitmap and inode table allocation for FLEX_BG
Theodore Ts'o [Tue, 22 Apr 2008 12:27:01 +0000 (08:27 -0400)]
mke2fs: New bitmap and inode table allocation for FLEX_BG

Change the way we allocate bitmaps and inode tables if the FLEX_BG
feature is used at mke2fs time.  It places calculates a new offset for
bitmaps and inode table base on the number of groups that the user
wishes to pack together using the new "-G" option.  Creating a
filesystem with 64 block groups in a flex group can be done by:

mke2fs -j -I 256 -O flex_bg -G 32 /dev/sdX

Signed-off-by: Jose R. Santos <jrs@us.ibm.com>
Signed-off-by: Valerie Clement <valerie.clement@bull.net>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agoBasic flexible block group support
Theodore Ts'o [Tue, 22 Apr 2008 11:47:25 +0000 (07:47 -0400)]
Basic flexible block group support

Add superblock definition, and dumpe2fs and debugfs support.

Signed-off-by: Jose R. Santos <jrs@us.ibm.com>
Signed-off-by: Valerie Clement <valerie.clement@bull.net>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agoext2fs_open_inode_scan: Handle an non-zero bg_itable_used in block group 0
Theodore Ts'o [Wed, 23 Apr 2008 03:22:17 +0000 (23:22 -0400)]
ext2fs_open_inode_scan: Handle an non-zero bg_itable_used in block group 0

Previously, the portion of the inode table for block group 0 was
always completely zero'ed out, so the ext2fs_open_inode_scan() didn't
handle a non-zero bg_itable_used value for the first block group.  Fix
this.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agomke2fs/libext2fs: Fix lazy inode table initialization
Theodore Ts'o [Wed, 23 Apr 2008 03:18:37 +0000 (23:18 -0400)]
mke2fs/libext2fs: Fix lazy inode table initialization

This fixes some bugs which I introduced recently while revamping the
uninit_bg code.  Since mke2fs is no longer calling
ext2fs_set_gdt_csum(), it's important that ext2fs_initialize()
correctly initialize bg_itable_unused for all block group descriptors.
In addition, mke2fs needs to zero out the the reserved inodes based on
the values of bg_itable_unused set by ext2fs_initialize().

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoMerge branch 'maint'
Theodore Ts'o [Tue, 22 Apr 2008 01:35:44 +0000 (21:35 -0400)]
Merge branch 'maint'

Conflicts:

configure

16 years agoUpdate texinfo.tex file newer version from the FSF
Theodore Ts'o [Tue, 22 Apr 2008 01:24:43 +0000 (21:24 -0400)]
Update texinfo.tex file newer version from the FSF

The one we had in the tree before was from circa 1996.  Lots of
updates and bug fixes since then....

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agolibext2fs.texinfo: Fix @deftypefun definitions so they are all on one line
Theodore Ts'o [Tue, 22 Apr 2008 01:21:57 +0000 (21:21 -0400)]
libext2fs.texinfo: Fix @deftypefun definitions so they are all on one line

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoFix "make docs" rule in Makefile for doc folder.
Manish Katiyar [Mon, 14 Apr 2008 11:50:03 +0000 (17:20 +0530)]
Fix "make docs" rule in Makefile for doc folder.

Signed-off-by: Manish Katiyar <mkatiyar@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoblkid: Keep cached filesystem information on EACCES and ENOENT errors
Theodore Ts'o [Mon, 21 Apr 2008 23:22:49 +0000 (19:22 -0400)]
blkid: Keep cached filesystem information on EACCES and ENOENT errors

When a nonprivileged user uses the blkid command, we want to keep the
cached filesystem information, and opening a device file could result
in an EACCESS or ENOENT (if an intervening directory is mode 700).  We
were previously testing for EPERM, which was really the wrong error
code to be testing against.

Addresses-Launchpad-Bug: #220275

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoTransfer responsibility of setting the *_UNINIT flags to libext2fs
Theodore Ts'o [Mon, 21 Apr 2008 05:29:01 +0000 (01:29 -0400)]
Transfer responsibility of setting the *_UNINIT flags to libext2fs

Mke2fs used to have special case, ugly code in
setup_lazy_bg/setup_uninit_bg flag which set the flags based on all
sorts of special cases.  Change it so that it is done in libext2fs,
and fix mke2fs to use alloc_stats functions which will take care of
clearing the *_UNINIT flags automatically as needed.

This is preparatory work to make the flex_bg allocation patch much
cleaner.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoRemove LAZY_BG feature
Theodore Ts'o [Mon, 21 Apr 2008 03:33:34 +0000 (23:33 -0400)]
Remove LAZY_BG feature

This simplifies the code, and using the uninit_bg with the inode table
lazily initialized is just as good.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoAdd comments documenting ext2fs_[reserve_]super_and_bgd_loc()
Theodore Ts'o [Sun, 20 Apr 2008 23:22:49 +0000 (19:22 -0400)]
Add comments documenting ext2fs_[reserve_]super_and_bgd_loc()

Add an explanation of exactly what ext2fs_super_and_bgd_loc() and
ext2fs_reserve_super_and_bgd_loc() do, and more importantly, exactly
what they return.  Note that most callers should *not* rely on the
return value since it's rarely useful, especially once the flex_bg
feature is enabled and inode table and allocation bitmap blocks may
not be in the block group.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoe2fsck: Fix pass5 handling of meta_bg and uninit_bg combination
Theodore Ts'o [Sat, 19 Apr 2008 14:30:59 +0000 (10:30 -0400)]
e2fsck: Fix pass5 handling of meta_bg and uninit_bg combination

The pass5 lazy block group support didn't understand meta_bg formats
correctly.  Fix this.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext2fs_set_gdt_csum: Remove setting of BLOCK_UNINIT
Theodore Ts'o [Sat, 19 Apr 2008 13:47:37 +0000 (09:47 -0400)]
ext2fs_set_gdt_csum: Remove setting of BLOCK_UNINIT

This function tried to set BLOCK_UNINIT based on the return value of
ext2fs_super_and_bgd_loc.  That's not something that works once we
start allowing flex_bg since the block group metadata might not be
located in the blockgroup itself.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext2fs_set_gdt_csum: Remove bogus setting of ITABLE_ZEROED
Theodore Ts'o [Sat, 19 Apr 2008 13:28:47 +0000 (09:28 -0400)]
ext2fs_set_gdt_csum: Remove bogus setting of ITABLE_ZEROED

It used to be the case that ext2fs_set_gdt_csum set the ITABLE_ZEROED
flag if the INODE_UNINIT is not set.  This assumed that the only
caller of ext2fs_set_gdt_csum was e2fsck (which was not true), and
that e2fsck would take care of zeroing the inode table (whic was also
not true).

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agolibe2p: Print the s_min_extra_isize and s_wanted_extra_isize fields
Theodore Ts'o [Sun, 20 Apr 2008 20:10:07 +0000 (16:10 -0400)]
libe2p: Print the s_min_extra_isize and s_wanted_extra_isize fields

Make dumpe2fs and debugfs print out the s_min_extra_isize and
s_wanted_extra_isize fields from the superblock.

Update tests expect files as appropriate.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agolibext2fs: Initialize s_min_extra_isize and s_wanted_extra_isize
Theodore Ts'o [Sun, 20 Apr 2008 20:02:13 +0000 (16:02 -0400)]
libext2fs: Initialize s_min_extra_isize and s_wanted_extra_isize

Set the s_min_extra_isize and s_wanted_extra_isize superblock fields
to reasonable defaults in ext2fs_initialize().

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoblkid: more sanity checks for swap v1
Eric Sandeen [Fri, 18 Apr 2008 15:28:44 +0000 (10:28 -0500)]
blkid: more sanity checks for swap v1

mkswap followed by pvcreate on a block device
will still turn up as "swap" in blkid, because
pvcreate isn't particularly careful about zeroing
old signatures.  (neither is mkswap, for that matter).

Testing for appropriate version and page counts
gives us a bit more confidence that we have a
real swap (v1) partition.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agomke2fs: Disentangle lazy_bg from uninit_bg's lazy inode table initialization
Theodore Ts'o [Fri, 18 Apr 2008 14:19:27 +0000 (10:19 -0400)]
mke2fs: Disentangle lazy_bg from uninit_bg's lazy inode table initialization

Instead of using -O lazy_bg,uninit_bg as the way of requesting that
the inode table be lazy unitialized, use the parameter
lazy_itable_init, which can either be set via mke2fs's -E option, or
via /etc/mke2fs.conf.

Also fix some random problems in mke2fs's man page, including
documenting the extent feature, which had been missing.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoChange the primary name of the extents feature to be 'extent'
Theodore Ts'o [Fri, 18 Apr 2008 03:33:05 +0000 (23:33 -0400)]
Change the primary name of the extents feature to be 'extent'

This was the original name used by Lustre's patches; keep the plural
when converting feature names to a feature mask for compatibility's
sake.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoRename the feature uninit_groups to uninit_bg
Theodore Ts'o [Fri, 18 Apr 2008 03:31:44 +0000 (23:31 -0400)]
Rename the feature uninit_groups to uninit_bg

Allow the old name of uninit_groups when converting feature names for
backwards compatibility for scripts running mke2fs and tune2fs.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoe2fsck: Only check PR_1_EXTENT_ENDS_BEYOND for leaf nodes
Eric Sandeen [Wed, 2 Apr 2008 01:38:58 +0000 (20:38 -0500)]
e2fsck: Only check PR_1_EXTENT_ENDS_BEYOND for leaf nodes

pass1 was checking that an "extent's" start+len did not extend
past the last filesystem block, but unless we are at a leaf
block, the physical block is that of a node in the tree, and
the length may include sparseness.  The test is only valid
for leaf blocks.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agoMake extent_goto() deterministic when logical block not found
Eric Sandeen [Tue, 8 Apr 2008 03:00:59 +0000 (22:00 -0500)]
Make extent_goto() deterministic when logical block not found

Make sure that extent_goto() leaves us at the last extent
prior to the requested logical block, if the logical block
requested lands in a hole.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agoFix ext2fs_extent_insert when at last extent in node
Eric Sandeen [Tue, 8 Apr 2008 03:00:58 +0000 (22:00 -0500)]
Fix ext2fs_extent_insert when at last extent in node

ext2fs_extent_insert() only did a memmove if path->left
was > 0, but if we are at the last extent in the node,
path->left == 0, and this node must be moved before the
current extent is replaced with the newly inserted node.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agoprint a bit more info for the tst_extents info command
Eric Sandeen [Tue, 8 Apr 2008 03:00:57 +0000 (22:00 -0500)]
print a bit more info for the tst_extents info command

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoFix tst_extents build when building w/o dynamic libraries
Theodore Ts'o [Thu, 17 Apr 2008 20:54:24 +0000 (16:54 -0400)]
Fix tst_extents build when building w/o dynamic libraries

$(LIBSS) should automatically include @DLOPEN_LIB@ so the right thing
happens for programs that need to use the ss library.

Reorder the library link order for tst_extents since the blkid library
uses libuuid functions.

Thanks to Eric Sandeen for pointing this problem out!

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoAdd support for the HUGE_FILE feature
Theodore Ts'o [Wed, 9 Apr 2008 15:39:11 +0000 (11:39 -0400)]
Add support for the HUGE_FILE feature

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoMerge branch 'maint'
Theodore Ts'o [Thu, 17 Apr 2008 20:32:28 +0000 (16:32 -0400)]
Merge branch 'maint'

16 years agopo: update sv.po (from translationproject.org)
Göran Uddeborg [Sun, 13 Apr 2008 17:31:34 +0000 (13:31 -0400)]
po: update sv.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agopo: update pl.po (from translationproject.org)
Jakub Bogusz [Sun, 13 Apr 2008 17:31:34 +0000 (13:31 -0400)]
po: update pl.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agopo: update fr.po (from translationproject.org)
Samuel Thibault [Sun, 13 Apr 2008 17:31:34 +0000 (13:31 -0400)]
po: update fr.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agopo: update de.po (from translationproject.org)
Philipp Thomas [Sun, 13 Apr 2008 17:31:34 +0000 (13:31 -0400)]
po: update de.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agopo: update cs.po (from translationproject.org)
Petr Pisar [Sun, 13 Apr 2008 17:31:34 +0000 (13:31 -0400)]
po: update cs.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agoFix non-POSIX ">&" in the Makefile for lib/ss's "make check" rule
Theodore Ts'o [Wed, 9 Apr 2008 19:40:43 +0000 (15:40 -0400)]
Fix non-POSIX ">&" in the Makefile for lib/ss's "make check" rule

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoMerge branch 'maint'
Theodore Ts'o [Mon, 7 Apr 2008 16:38:16 +0000 (12:38 -0400)]
Merge branch 'maint'

16 years agoe2fsck: avoid double free of ctx->filesystem_name
Matthias Koenig [Mon, 7 Apr 2008 15:49:34 +0000 (11:49 -0400)]
e2fsck: avoid double free of ctx->filesystem_name

In current git there is a double free on ctx->filesystem_name in the
end of main() and in e2fsck_free_context, causing e2fsck to abort at
the end of pass5.

Signed-off-by: Matthias Koenig <mkoenig@suse.de>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoMerge branch 'es/warnings-fix' into next
Theodore Ts'o [Sun, 6 Apr 2008 21:22:08 +0000 (17:22 -0400)]
Merge branch 'es/warnings-fix' into next

16 years agoFix a couple of implicit function declarations
Eric Sandeen [Sun, 30 Mar 2008 04:01:52 +0000 (23:01 -0500)]
Fix a couple of implicit function declarations

Fedora seems to be gearing up to add
-Werror-implicit-function-declaration
to the standard build flags, so I thought I'd get out ahead
of this one...

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoNew mke2fs filesystem and usage types support
Theodore Ts'o [Tue, 19 Feb 2008 13:32:58 +0000 (08:32 -0500)]
New mke2fs filesystem and usage types support

Provide mke2fs with a much more sophisticated system for controlling
configuration parameters of a newly created filesystem based on a
split filesystem and usage type system.  The -t option to mke2fs was a
deprecated alias to -c; it now specifies a filesystem type (ext2,
ext3, ext4, etc.), while the -T option can now be a comma separated
usage list.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoMerge branch 'maint' into next
Theodore Ts'o [Sun, 6 Apr 2008 21:10:49 +0000 (17:10 -0400)]
Merge branch 'maint' into next

16 years agoFix m_no_opt regression test
Theodore Ts'o [Sun, 6 Apr 2008 21:07:04 +0000 (17:07 -0400)]
Fix m_no_opt regression test

The m_no_opt test has been silently failing since the addition of
mke2fs.conf, since it was trying to use resize_inode on a non-sparse
filesystem.  Fix it so it works correctly.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agolibext2fs: Fix resize inode creation with non-standard s_first_data_block
Theodore Ts'o [Sun, 6 Apr 2008 21:00:23 +0000 (17:00 -0400)]
libext2fs: Fix resize inode creation with non-standard s_first_data_block

Thanks to Max Lindner (lokimuh) for pointing this out.

   I'm playing around a bit with ext2 and multi-user encryption and I
   need space for my key management. So I set s_first_data_block to 4000
   or something like that.

   This way mke2fs segfaults when executing
   ext2fs_create_resize_inode() because

   blk_t goal = 3 + sb->s_reserved_gdt_blocks + fs->desc_blocks +
   fs->inode_blocks_per_group;

   will produce a integer underrun later and segfault then in the
   ext2fs_test_bit assembler inline function.

   when exchanging 3 with 2 + sb->s_first_data_block, mke2fs does not
   segfault.

   I'm not 100% sure if thats the correct way dealing with this issue
   but I think its a proper solution.

Addresses-Sourceforge-Bug: #1935847

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agopo: update vi.po (from translationproject.org)
Clytie Siddall [Sun, 6 Apr 2008 20:11:56 +0000 (16:11 -0400)]
po: update vi.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agopo: update nl.po (from translationproject.org)
Benno Schulenberg [Sun, 6 Apr 2008 20:11:56 +0000 (16:11 -0400)]
po: update nl.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agopo: update fr.po (from translationproject.org)
Samuel Thibault [Wed, 2 Apr 2008 00:34:24 +0000 (20:34 -0400)]
po: update fr.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 years agoRemove bashism from configure script
Theodore Ts'o [Wed, 2 Apr 2008 00:32:55 +0000 (20:32 -0400)]
Remove bashism from configure script

Thanks to Mike Frysinger for pointing this out.

Addresses-Sourceforge-Bug: 1921969

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agolibext2fs: Fix big-endian bug in tst_csum.c
Theodore Ts'o [Tue, 1 Apr 2008 21:03:00 +0000 (17:03 -0400)]
libext2fs: Fix big-endian bug in tst_csum.c

It's not necessary to byte-swap the checksum.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoAdd dependency rule so that static library when necessary for "make check"
Theodore Ts'o [Tue, 1 Apr 2008 20:45:37 +0000 (16:45 -0400)]
Add dependency rule so that static library when necessary for "make check"

Some library makefiles use $(LIB)/$(LIBRARY).a as a dependency for the
static library built by the makefile fragment.  Add it as a target
created when building $(LIBRARY).a so that it is rebuilt when necessary.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agolibext2fs: Make all test programs link only against the static library
Theodore Ts'o [Tue, 1 Apr 2008 20:43:24 +0000 (16:43 -0400)]
libext2fs: Make all test programs link only against the static library

This is needed so that all of the test programs compile correctly on
big-endian systems.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoe2fsck: Fix extent flag validity tests in pass1 on big endian boxes.
Eric Sandeen [Tue, 1 Apr 2008 19:18:44 +0000 (15:18 -0400)]
e2fsck: Fix extent flag validity tests in pass1 on big endian boxes.

Extent data is shared with the i_block[] space in the inode,
but it is always swapped on access, not when the inode is read.

In e2fsck/pass1.c we must be careful when checking validity
of the extents flag on the inode.  If the flag was set when
the inode was read & swapped, then the extents data itself
(in ->i_block[]) was NOT swapped, so testing for a valid
extent header requires some swapping first.  Then, if we
ultimately set the extents flag, all of i_block[] must be
re/un-swapped.

This passes the f_extent regression test on both ppc & x86.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agodumpe2fs: Print the group checksum and the block options in a nicer way
Theodore Ts'o [Tue, 1 Apr 2008 18:59:47 +0000 (14:59 -0400)]
dumpe2fs: Print the group checksum and the block options in a nicer way

Also add support for printing the ITABLE_ZERO flag

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoAdd new regression test: f_uninit_last_uninit
Andreas Dilger [Mon, 31 Mar 2008 18:32:39 +0000 (14:32 -0400)]
Add new regression test: f_uninit_last_uninit

Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoFix trailing whitespace in e2fsck/problem.[ch]
Theodore Ts'o [Mon, 31 Mar 2008 16:16:51 +0000 (12:16 -0400)]
Fix trailing whitespace in e2fsck/problem.[ch]

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoe2fsck: Add check to enforce a valid block bitmap in last block group
Andreas Dilger [Mon, 31 Mar 2008 16:14:22 +0000 (12:14 -0400)]
e2fsck: Add check to enforce a valid block bitmap in last block group

Add a check for the UNINIT_BLOCKS flag set in the last group.  The kernel
patch doesn't handle this gracefully, because it assumes there are a full
set of blocks in each group marked UNINIT_BLOCKS.  The kernel should be
fixed up, but in the meantime this avoids hitting the problem, and is
more consistent with lazy_bg not marking the last group UNINIT.

Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoSplit the m_lazy test case into two cases: m_lazy and m_lazy_resize
Andreas Dilger [Mon, 31 Mar 2008 14:59:27 +0000 (10:59 -0400)]
Split the m_lazy test case into two cases: m_lazy and m_lazy_resize

Split the m_lazy testcase into two tests - one that has resize enabled,
and one that does not.  There were previously problems with lazy+resize
because the resize feature consumed blocks in some of the groups.

Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agolibext2fs: Micro-optimization in inode scan code
Andreas Dilger [Mon, 31 Mar 2008 14:57:38 +0000 (10:57 -0400)]
libext2fs: Micro-optimization in inode scan code

Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext2fs_set_gdt_csum(): Return an error code on errors instead of void
Andreas Dilger [Mon, 31 Mar 2008 14:50:19 +0000 (10:50 -0400)]
ext2fs_set_gdt_csum(): Return an error code on errors instead of void

Change the function signature so that ext2fs_set_gdt_csum() returns an
error code.

If the inode bitmap hasn't been loaded return EXT2_ET_NO_INODE_BITMAP.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>