OSDN Git Service

android-x86/external-e2fsprogs.git
13 years agoMerge branch 'maint' into next
Theodore Ts'o [Mon, 28 Feb 2011 00:47:44 +0000 (19:47 -0500)]
Merge branch 'maint' into next

Conflicts:
lib/ext2fs/initialize.c

13 years agoAdd basic BIGALLOC support for cluster-based allocation
Theodore Ts'o [Sat, 26 Feb 2011 02:43:54 +0000 (21:43 -0500)]
Add basic BIGALLOC support for cluster-based allocation

This adds the superblock fields needed so that dumpe2fs works and the
code points and renames the superblock fields from describing
fragments to clusters.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoe2fsprogs: reserving code points for new ext4 quota feature
Aditya Kali [Tue, 15 Feb 2011 22:27:27 +0000 (14:27 -0800)]
e2fsprogs: reserving code points for new ext4 quota feature

This patch adds support for detecting the new 'quota' feature in ext4.
The patch reserves code points for usr and group quota inodes and also
for the feature flag EXT4_FEATURE_RO_COMPAT_QUOTA.

Signed-off-by: Aditya Kali <adityakali@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agomke2fs: Display progress report during the device discard
Lukas Czerner [Mon, 24 Jan 2011 19:52:00 +0000 (20:52 +0100)]
mke2fs: Display progress report during the device discard

For some time now we are doing initial discard of the device prior to
filesystem creation. However, there is no feedback for the user and
hence on some devices with slow TRIM implementation it may appear that
mke2fs is stuck.

This commit introduce new function mke2fs_discard_device(), which is a
wrapper for io_channel_discard(). The discard is done in chunks of
2GB, which seems reasonably well for both slow and fast devices, and
discard progress is reported back to the user.

I gave up on doing fancy things like align discard according to
discard_alignment, checking for discard granularity and computing
estimate time. First of all, because it would require either new ioctl
to retrieve those information or use of libudev library, none of it
seems to be worth it. Regarding discard_granularity, I doubt there is
any sane device with discard granularity that big it would affect this.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoMerge branch 'maint' into next
Theodore Ts'o [Mon, 21 Feb 2011 02:54:43 +0000 (21:54 -0500)]
Merge branch 'maint' into next

Conflicts:
misc/mke2fs.conf.5.in

13 years agomke2fs: Simple man page nodiscard option correction
Lukas Czerner [Tue, 1 Feb 2011 16:56:41 +0000 (17:56 +0100)]
mke2fs: Simple man page nodiscard option correction

It is not true that 'nodiscard' is set as default, so remove this
sentence. The default is 'discard' and it is properly documented in man
page.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe2fsprogs: enable user namespace xattrs by default
Eric Sandeen [Thu, 17 Feb 2011 21:56:17 +0000 (15:56 -0600)]
e2fsprogs: enable user namespace xattrs by default

User namespace xattrs are generally useful, and I think extN
is the only filesystem requiring a special mount option to
enable them, when xattrs are otherwise available.  So this
change sets that mount option into the defaults, via a
mke2fs.conf option.

Note that if xattrs are config'd off, this will lead to a
mostly-harmless:

   EXT4-fs (sdc1): (no)user_xattr options not supported

message at mount time...

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe2fsprogs: turn off enforced fsck intervals by default
Eric Sandeen [Thu, 17 Feb 2011 21:55:15 +0000 (15:55 -0600)]
e2fsprogs: turn off enforced fsck intervals by default

The forced fsck often comes at unexpected and inopportune moments,
and even enterprise customers are often caught by surprise when
this happens.  Because a filesystem with an error condition will
be marked as requiring fsck anyway, I submit that the time-based
and mount-based checks are not particularly useful, and that
administrators can schedule fscks on their own time, or tune2fs
the enforced intervals if they so choose.  This patch disables the
intervals by default, and I've added a new mkfs.conf option to
turn on the old behavior of random, unexpected, time-consuming
fscks at boot time.  ;)

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe2fsprogs: create com_err.h link in includedir
Eric Sandeen [Wed, 16 Feb 2011 18:01:39 +0000 (12:01 -0600)]
e2fsprogs: create com_err.h link in includedir

After debian bug #192277, debian/rules started making a symlink
to com_err.h in /usr/include.  Now I have Fedora bug #550889
for the same issue, and perhaps it's time to make this link
by default, rather than fixing it up in packaging steps?

[ Changed by tytso to remove the explicit -s option; this will default
  to creating a hard link by default, which slightly faster.  If
  people want to use symlinks for all links during the install
  process, they can use configure option --enable-symlink-install.
  The reason for this change is that some file systems, like AFS,
  don't support symlinks, and AFS users complain when they can't build
  or install into AFS.  So I don't want to use symlinks
  unconditionally without a way of switching things back and forth,
  and it's easier if we just make all links made during the install
  process to be hard links or sym links. ]

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agobadblocks: Add accounting for different types of errors
Theodore Ts'o [Sun, 20 Feb 2011 20:29:51 +0000 (15:29 -0500)]
badblocks: Add accounting for different types of errors

When using the -v option, report a breakdown of the number of read,
write, and comparison errors that were found by badblocks.

Thanks to Ragnar Kjørstad for providing this patch.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agobadblocks: Fix up recover_block handling in badblocks
Theodore Ts'o [Sun, 20 Feb 2011 20:19:47 +0000 (15:19 -0500)]
badblocks: Fix up recover_block handling in badblocks

If there was a bad block for block #0, badblocks would never switch
back testing blocks more efficiently.  In addition, we were
double-incrementing the blocks to be tested in the read/write test due
to failure to remove code.

Thanks to Ragnar Kjørstad for pointing these problems out.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoMerge branch 'maint' into next
Theodore Ts'o [Fri, 18 Feb 2011 06:19:05 +0000 (01:19 -0500)]
Merge branch 'maint' into next

13 years agobadblocks: Only report errors when reading/writing one block at a time
Theodore Ts'o [Fri, 18 Feb 2011 03:58:21 +0000 (22:58 -0500)]
badblocks: Only report errors when reading/writing one block at a time

With Direct I/O, the kernel can report 0 bytes read even though the
first block has no errors.  So there are any errors, we need try to
read/write blocks one at a time and to get an accurate report.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agobadblocks: Add the -B option which forces the use of buffered I/O
Theodore Ts'o [Thu, 17 Feb 2011 04:40:46 +0000 (23:40 -0500)]
badblocks: Add the -B option which forces the use of buffered I/O

If for some reason direct I/O does not work correctly, force the use
of buffered I/O.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agobadblocks: Fix bug so that O_DIRECT mode is correctly entered
Theodore Ts'o [Thu, 17 Feb 2011 04:35:49 +0000 (23:35 -0500)]
badblocks: Fix bug so that O_DIRECT mode is correctly entered

The check to see if the block number is properly aligned was not done
correctly.  Fix this.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoMerge branch 'maint' into next
Theodore Ts'o [Fri, 11 Feb 2011 00:04:46 +0000 (19:04 -0500)]
Merge branch 'maint' into next

13 years agopo: update fr.po (from translationproject.org)
Samuel Thibault [Fri, 11 Feb 2011 00:03:42 +0000 (19:03 -0500)]
po: update fr.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agofilefrag: remove useless assignment
Jim Meyering [Tue, 1 Feb 2011 18:25:35 +0000 (19:25 +0100)]
filefrag: remove useless assignment

The very next one memset's all bytes of fiemap to 0.

Signed-off-by: Jim Meyering <meyering@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agopo: update zh_CN.po (from translationproject.org)
Dark Raven [Wed, 26 Jan 2011 23:54:25 +0000 (18:54 -0500)]
po: update zh_CN.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agopo: update de.po (from translationproject.org)
Philipp Thomas [Wed, 26 Jan 2011 23:54:24 +0000 (18:54 -0500)]
po: update de.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agopo: update id.po (from translationproject.org)
Theodore Ts'o [Mon, 24 Jan 2011 20:23:56 +0000 (15:23 -0500)]
po: update id.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agopo: update vi.po (from translationproject.org)
Clytie Siddall [Mon, 24 Jan 2011 20:03:44 +0000 (15:03 -0500)]
po: update vi.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agopo: update sv.po (from translationproject.org)
Göran Uddeborg [Mon, 24 Jan 2011 20:03:43 +0000 (15:03 -0500)]
po: update sv.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agopo: update pl.po (from translationproject.org)
Jakub Bogusz [Mon, 24 Jan 2011 20:03:43 +0000 (15:03 -0500)]
po: update pl.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agopo: update nl.po (from translationproject.org)
Benno Schulenberg [Mon, 24 Jan 2011 20:03:42 +0000 (15:03 -0500)]
po: update nl.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agopo: update fr.po (from translationproject.org)
Samuel Thibault [Mon, 24 Jan 2011 20:03:42 +0000 (15:03 -0500)]
po: update fr.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agopo: update de.po (from translationproject.org)
Philipp Thomas [Mon, 24 Jan 2011 20:03:41 +0000 (15:03 -0500)]
po: update de.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agopo: update cs.po (from translationproject.org)
Petr Pisar [Mon, 24 Jan 2011 20:03:41 +0000 (15:03 -0500)]
po: update cs.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoresize2fs: fix computation of the real end of the bitmap to be 64-bit
Theodore Ts'o [Fri, 7 Jan 2011 21:38:13 +0000 (16:38 -0500)]
resize2fs: fix computation of the real end of the bitmap to be 64-bit

real_end had been previously declared with a bogus type, which is why
this was missed earlier.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoe4defrag: Use libext2fs to get the correct superblock information
Kazuya Mio [Sun, 19 Dec 2010 22:01:24 +0000 (22:01 +0000)]
e4defrag: Use libext2fs to get the correct superblock information

Currently, e4defrag always does byte-swapping when it gets superblock
information, so the calculation of the best extents count is not
correct on little endian machine. This doesn't cause data corruption,
but it may confuse users by showing the wrong extent count.  To solve
this problem, we use ext2fs_open() instead of get_superblock_info()
that is the original function.

Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoconfigure: control whether e4defrag is built/installed via --disable-defrag
Theodore Ts'o [Fri, 24 Dec 2010 20:19:44 +0000 (15:19 -0500)]
configure: control whether e4defrag is built/installed via --disable-defrag

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe4defrag: fix segfault when e4defrag races with unlink/truncate
Kazuya Mio [Thu, 18 Nov 2010 21:51:25 +0000 (21:51 +0000)]
e4defrag: fix segfault when e4defrag races with unlink/truncate

If a file gets deleted or truncated while e4defrag is trying to
operate on it, it's possible for it seg fault.

Addresses-Red-Hat-Bugzilla: #641926

Reported-by: Michal Piotrowski <mkkp4x4@gmail.com>
Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoMerge branch 'maint' into next
Theodore Ts'o [Thu, 23 Dec 2010 00:00:21 +0000 (19:00 -0500)]
Merge branch 'maint' into next

Conflicts:
resize/resize2fs.c

13 years agoUpdate Release Notes, Changelogs, version.h, etc. for 1.41.14 release
Theodore Ts'o [Wed, 22 Dec 2010 23:55:15 +0000 (18:55 -0500)]
Update Release Notes, Changelogs, version.h, etc. for 1.41.14 release

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: don't complain if the fs type "default" is not defined in mke2fs.conf
Theodore Ts'o [Wed, 22 Dec 2010 23:31:36 +0000 (18:31 -0500)]
mke2fs: don't complain if the fs type "default" is not defined in mke2fs.conf

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: take the device size into account when determining the size type
Theodore Ts'o [Wed, 22 Dec 2010 23:22:40 +0000 (18:22 -0500)]
mke2fs: take the device size into account when determining the size type

If the file system size was not specified on the command line, we were
always using the usage type "floppy" since we didn't determine the
device size until after calling parse_fs_types().  Doh!

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoresize2fs: do not clear resize inode for 0 resvd blocks
Eric Sandeen [Tue, 21 Dec 2010 21:32:05 +0000 (15:32 -0600)]
resize2fs: do not clear resize inode for 0 resvd blocks

I ran into odd behavior where mkfs.ext4 of a 16T filesystem would
create a resize inode with 0 reserved blocks, and mark the resize_inode
feature.

A subsequent slight downward resize of the filesystem would remove
the resize inode, making any further offline resizing impossible.

This is especially odd in light of the fact that a large downward
resize (say, to 8T) will actually add blocks to the resize inode -
so a small resize removes it, a large resize expands it ...

commit 8ade268cf2fde8629b51bfd1c044a83db88234cd had added this:

  If the filesystem is grown to the point where the resize_inode is no
  longer needed, clean it up properly so e2fsck doesn't have to.

but, it seems e2fsck does not care about this situation, either.

So, simply leave the resize_inode intact in this case, and everything
seems to be happy.

Note, this is for the 1.41.xx branch.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agolibext2fs: Don't use the extended rec_len encoding for standard file systems
Theodore Ts'o [Wed, 22 Dec 2010 18:53:02 +0000 (13:53 -0500)]
libext2fs: Don't use the extended rec_len encoding for standard file systems

If the file system has a blocksize less than 64k, then don't use the
extended rec_len encoding, to be consistent with what the kernel will
do.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs.8.in: add ENVIRONMENT section
Namhyung Kim [Mon, 29 Nov 2010 08:55:17 +0000 (17:55 +0900)]
mke2fs.8.in: add ENVIRONMENT section

Add ENVIRONMENT section and describe behavior of MKE2FS_SYNC,
MKE2FS_CONFIG, MKE2FS_FIRST_META_BG, MKE2FS_DEVICE_SECTSIZE
and MKE2FS_SKIP_CHECK_MSG.

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agolibext2fs: fix possible memory leak in write_journal_inode()
Namhyung Kim [Mon, 29 Nov 2010 08:55:16 +0000 (17:55 +0900)]
libext2fs: fix possible memory leak in write_journal_inode()

ext2fs_zero_block2() allocates static buffer if needed so it
should be freed at last (call it again with 0 args).

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agomke2fs: clean up error handling in mke2fs_setup_tdb()
Theodore Ts'o [Wed, 22 Dec 2010 02:57:02 +0000 (21:57 -0500)]
mke2fs: clean up error handling in mke2fs_setup_tdb()

Avoid memory leaks on error paths, and make sure we issue the correct
error messages in the case of (highly) unlikely errors.

Original patch submitted by Namhyung Kim <namhyung@gmail.com>, but
highly rewritten since then.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: add some error checks into PRS()
Namhyung Kim [Thu, 16 Dec 2010 09:40:40 +0000 (18:40 +0900)]
mke2fs: add some error checks into PRS()

Check return value of some functions and exit if unhandled error occurred.

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agomke2fs: fix determination of size_type
Namhyung Kim [Mon, 29 Nov 2010 08:55:13 +0000 (17:55 +0900)]
mke2fs: fix determination of size_type

In original code, 'huge' type could not be selected because it
always be caught for 'big' type. Change the ordering.

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agomke2fs.8.in: add missing "big" and "huge" usage-type description
Namhyung Kim [Mon, 29 Nov 2010 08:55:12 +0000 (17:55 +0900)]
mke2fs.8.in: add missing "big" and "huge" usage-type description

The commit 493024ea1d74e4cb48aac3a24111f5c8da343e9f ("mke2fs: Fix up the
fs type and feature selection for 64-bit blocks") added 'big' and 'huge'
usage-type but was missing description in man page. Add it.

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agomke2fs: Avoid potential NULL dereference
Theodore Ts'o [Tue, 21 Dec 2010 23:12:12 +0000 (18:12 -0500)]
mke2fs: Avoid potential NULL dereference

... in the very unlikely case that e2p_os2string fails to allocate
memory.

Reported-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoext2fs: fix memory leak in an error case in inode_open()
Theodore Ts'o [Tue, 21 Dec 2010 23:05:07 +0000 (18:05 -0500)]
ext2fs: fix memory leak in an error case in inode_open()

Reported-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agolibext2fs: fix potential free() of garbage in ext2fs_update_bb_inode()
Theodore Ts'o [Mon, 20 Dec 2010 15:57:29 +0000 (10:57 -0500)]
libext2fs: fix potential free() of garbage in ext2fs_update_bb_inode()

There was a potential of freeing an uninitialized pointer in
rec.block_buf, which was pointed out by Namhyung Kim <namhyung@gmail.com>

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agolibext2fs: fix dubious code in ext2fs_unmark_generic_bitmap()
Namhyung Kim [Mon, 29 Nov 2010 08:55:07 +0000 (17:55 +0900)]
libext2fs: fix dubious code in ext2fs_unmark_generic_bitmap()

Looks like a copy & paste problem to me.

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agolibext2fs: remove unnecessary casts to ext2fs_generic_bitmap
Namhyung Kim [Mon, 29 Nov 2010 08:55:06 +0000 (17:55 +0900)]
libext2fs: remove unnecessary casts to ext2fs_generic_bitmap

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agomke2fs: Use ext2fs_div_ceil to simplify write_inode_tables()
Theodore Ts'o [Mon, 20 Dec 2010 15:42:57 +0000 (10:42 -0500)]
mke2fs: Use ext2fs_div_ceil to simplify write_inode_tables()

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agolibext2fs: fix dubious code in ext2fs_numeric_progress_init()
Namhyung Kim [Mon, 20 Dec 2010 15:07:10 +0000 (10:07 -0500)]
libext2fs: fix dubious code in ext2fs_numeric_progress_init()

Looks like a copy & paste problem to me.

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agolibext2fs: fix potential build failure with OMIT_COM_ERR
Namhyung Kim [Mon, 20 Dec 2010 15:07:10 +0000 (10:07 -0500)]
libext2fs: fix potential build failure with OMIT_COM_ERR

This fixes following build failure when OMIT_COM_ERR is defined:

 lib/ext2fs/gen_bitmap.c: In function ‘ext2fs_clear_generic_bitmap’:
 lib/ext2fs/gen_bitmap.c:437: error: invalid storage class for function ‘ext2fs_test_clear_generic_bitmap_range’
 lib/ext2fs/gen_bitmap.c:559: error: expected declaration or statement at end of input
 lib/ext2fs/gen_bitmap.c: In function ‘ext2fs_get_generic_bitmap_end’:
 lib/ext2fs/gen_bitmap.c:559: error: expected declaration or statement at end of input
 lib/ext2fs/gen_bitmap.c: In function ‘ext2fs_get_generic_bitmap_start’:
 lib/ext2fs/gen_bitmap.c:559: error: expected declaration or statement at end of input
 lib/ext2fs/gen_bitmap.c: In function ‘ext2fs_unmark_generic_bitmap’:
 lib/ext2fs/gen_bitmap.c:559: error: expected declaration or statement at end of input
 lib/ext2fs/gen_bitmap.c: In function ‘ext2fs_mark_generic_bitmap’:
 lib/ext2fs/gen_bitmap.c:559: error: expected declaration or statement at end of input
 lib/ext2fs/gen_bitmap.c: In function ‘ext2fs_test_generic_bitmap’:
 lib/ext2fs/gen_bitmap.c:559: error: expected declaration or statement at end of input
make[2]: *** [gen_bitmap.o] Error 1
make[2]: Leaving directory e2fsprogs/lib/ext2fs'
make[1]: *** [all-libs-recursive] Error 1
make[1]: Leaving directory e2fsprogs'
make: *** [all] Error 2

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoMerge branch 'maint' into next
Theodore Ts'o [Mon, 20 Dec 2010 15:06:58 +0000 (10:06 -0500)]
Merge branch 'maint' into next

Conflicts:
lib/ext2fs/block.c
lib/ext2fs/csum.c
resize/main.c

13 years agoMerge branch 'master' into next
Theodore Ts'o [Fri, 17 Dec 2010 04:20:11 +0000 (23:20 -0500)]
Merge branch 'master' into next

Conflicts:
e2fsck/e2fsck.h
e2fsck/unix.c

13 years agoe4defrag: Fix the overflow in e4defrag with > 2GB files
Theodore Ts'o [Fri, 17 Dec 2010 03:53:34 +0000 (22:53 -0500)]
e4defrag: Fix the overflow in e4defrag with > 2GB files

The fallocate() interface on 32-bit machines is defined to use off_t,
not loff_t (even though the system call interface is 64-bit clean).
This causes e4defrag to fail on files greater than 2GB.  Fix this by
trying to use fallocate64(), and using the hard-coded syscall if it
does not exist.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agolibext2fs: Fix compile bug on big-endian architectures
Theodore Ts'o [Fri, 17 Dec 2010 03:11:43 +0000 (22:11 -0500)]
libext2fs: Fix compile bug on big-endian architectures

Addresses-Sourceforge-Bug: #3138115

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agolibext2fs: fix ref_offset callback with sparse files in ext2fs_block_iterate2()
Theodore Ts'o [Fri, 17 Dec 2010 00:34:24 +0000 (19:34 -0500)]
libext2fs: fix ref_offset callback with sparse files in ext2fs_block_iterate2()

Addresses-Sourceforge-Bug: 3081087

Reported-by: vmo@users.sourceforge.net
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoresize2fs: handle exactly-16T filesystems in resize2fs
Eric Sandeen [Thu, 16 Dec 2010 04:37:34 +0000 (22:37 -0600)]
resize2fs: handle exactly-16T filesystems in resize2fs

Before we go whole-hog on 64-bit e2fsprogs, I wonder if this
is worth considering as a last-minute addition to the 1.41
stream.  Currently, mke2fs will shave a block off an exactly-16T
device to fit*, but resize2fs does not do the same, leading
to some asymmetry.  This patch fixes that up, and allows 16T
devices to be handled more gracefully in offline resize.
(in fact resize2fs will not even open a 16T device, today).

*commit 37d17a65ecb4615546b417038190a41bafca7c51

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe2fsprogs: fix type-punning warnings
Eric Sandeen [Tue, 14 Dec 2010 19:00:01 +0000 (13:00 -0600)]
e2fsprogs: fix type-punning warnings

Flags used during RHEL/Fedora builds lead to a couple type-punning
warnings:

  recovery.c: In function 'do_one_pass':
  recovery.c:539: warning: dereferencing type-punned pointer will break strict-aliasing rules
  ./csum.c: In function 'print_csum':
  ./csum.c:170: warning: dereferencing type-punned pointer will break strict-aliasing rules

The two changes below fix this up.

Note that the csum test binary output changes slightly, but this does
not break any tests.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe4defrag: update man page about -c option
Kazuya Mio [Mon, 13 Dec 2010 14:59:07 +0000 (09:59 -0500)]
e4defrag: update man page about -c option

Add the description of the size per one extent and the maximum extent size
in ext4 into e4defrag man page.

Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe4defrag: avoid unsuccessful return for an non-privileged user
Kazuya Mio [Mon, 13 Dec 2010 14:59:07 +0000 (09:59 -0500)]
e4defrag: avoid unsuccessful return for an non-privileged user

If non-privileged user runs e4defrag, e4defrag returns an exit status
of 1 despite its success. This patch fixes this problem.

Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe4defrag: fix file blocks calculation
Kazuya Mio [Mon, 13 Dec 2010 14:59:06 +0000 (09:59 -0500)]
e4defrag: fix file blocks calculation

e4defrag uses st_blocks (struct stat) to calculate file blocks. However,
st_blocks also has meta data blocks in addition to file blocks. So, we
calculate file blocks by sum of the extent length.

Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe4defrag: output size per extent by -c option
Kazuya Mio [Mon, 13 Dec 2010 14:57:26 +0000 (09:57 -0500)]
e4defrag: output size per extent by -c option

e4defrag with -c option outputs "ratio" that means the levels of
fragmentation. However, it's difficult for users to understand, so we will
use size per extent instead of ratio.

Before:
# e4defrag -c /mnt/mp1/file
<File>                                         now/best          ratio
/mnt/mp1/file                                    6/1             0.00%

 Total/best extents                             6/1
 Fragmentation ratio                            0.00%
 Fragmentation score                            0.04
 [0-30 no problem: 31-55 a little bit fragmented: 55- needs defrag]
 This file(/mnt/mp1/file) does not need defragmentation.
 Done.

After:
# e4defrag -c /mnt/mp1/file
<File>                                         now/best       size/ext
/mnt/mp1/file                                    6/1          16666 KB

 Total/best extents                             6/1
 Average size per extent                        16666 KB
 Fragmentation score                            0
 [0-30 no problem: 31-55 a little bit fragmented: 56- needs defrag]
 This file (/mnt/mp1/file) does not need defragmentation.
 Done.

Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe4defrag: return more specific error message on ioctl failure
Peng Tao [Mon, 13 Dec 2010 14:44:57 +0000 (09:44 -0500)]
e4defrag: return more specific error message on ioctl failure

Currently e4defrag relies on the EXT4_IOC_MOVE_EXT ioctl to perform online
defragmentation. However, this iotcl kernel patch is not available before
2.6.30-rc1. e4defrag shall fail without obvious reasons on systems running
older kernels. The patch adds more detailed error message addressing this
issue and prompts users with the minimal kernel version that is needed to
run e4defrag.

Signed-off-by: Peng Tao <bergwolf@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe4defrag: skip "rootfs" entry when checking for ext4 filesystem
Eric Sandeen [Tue, 8 Sep 2009 06:13:50 +0000 (06:13 +0000)]
e4defrag: skip "rootfs" entry when checking for ext4 filesystem

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoe4defrag: open the file to be defragged in read/write mode
Darrick J. Wong [Mon, 13 Dec 2010 14:23:45 +0000 (09:23 -0500)]
e4defrag: open the file to be defragged in read/write mode

Akira Fujita merged a patch into 2.6.33 that adds a requirement that a
file being defragged must be opened with read and write access, so
e2fsprogs needs to satisfy that.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoMerge branch 'maint' into next
Theodore Ts'o [Mon, 13 Dec 2010 14:16:09 +0000 (09:16 -0500)]
Merge branch 'maint' into next

Conflicts:
e2fsck/pass4.c
misc/dumpe2fs.c
resize/online.c

13 years agoUpdate Release Notes, Changelogs, version.h, etc. for 1.41.13 release
Theodore Ts'o [Mon, 13 Dec 2010 13:57:14 +0000 (08:57 -0500)]
Update Release Notes, Changelogs, version.h, etc. for 1.41.13 release

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoe2fsck: add an option which causes it to only do a journal replay
Bernd Schubert [Fri, 12 Nov 2010 23:09:07 +0000 (00:09 +0100)]
e2fsck: add an option which causes it to only do a journal replay

As recently discussed on linux-ext4@vger.kernel.org add an option to e2fsck
to allow to replay the journal only. That will allow scripts, such as
pacemakers 'Filesystem' RA to first replay the journal and if that sets
an error state from the journal replay, further check for that error
(dumpe2fh -h | grep "Filesystem state:") and if that shows and error
to refuse to mount. It also allows automatic e2fsck scripts to first
replay the journal and on a second run after the real pass1 to passX checks
to test for the return code.

Signed-off-by: Bernd Schubert <bschubert@ddn.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe2fsck: Do blkid interpretation on the external journal specifier
Theodore Ts'o [Mon, 6 Dec 2010 22:07:27 +0000 (17:07 -0500)]
e2fsck: Do blkid interpretation on the external journal specifier

If the user specifies "e2fsck -j UUID=XXX", e2fsck should do blkid
interpretation, since e2fsck does it with the base file system name.
So from the sake of consistency and user convenience, we should do it
here too.

Addresses-Debian-Bug: #559315

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoe2fsck: Add the ability to force a problem to not be fixed
Theodore Ts'o [Mon, 6 Dec 2010 15:10:33 +0000 (10:10 -0500)]
e2fsck: Add the ability to force a problem to not be fixed

The boolean options "force_no" in the problems stanza of e2fsck.conf
allows a particular problem code be treated as if the user will answer
"no" to the question of whether a particular problem should be fixed
--- even if e2fsck is run with the -y option.

As an example use case, suppose a distribution had widely deployed a
version of the kernel where under some circumstances, the EOFBLOCKS_FL
flag would be left set even though it should not be left set, and a
customer had a workload which exercised the fencepost error all the
time, resulting in many large number of inodes that had EOFBLOCKS_FL
set erroneously.  Enough, in fact, the e2fsck runs were taking too
long.  (There was such a bug in the kernel, which was fixed by commit
58590b06d in 2.6.36).

Leaving EOFBLOCKS_FL set when it should not be isn't a huge deal, and
is certainly than having high availability timeout alerts going off
left and right.  So in this case, the best fix might be to put the
following in /etc/e2fsck.conf:

[problems]
0x010060 = { # PR_1_EOFBLOCKS_FL_SET
 force_no = true
 no_ok = true
 no_nomsg = true
}

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agodumpe2fs: fix output for flex_bg bitmap offsets
Andreas Dilger [Mon, 6 Dec 2010 03:20:19 +0000 (22:20 -0500)]
dumpe2fs: fix output for flex_bg bitmap offsets

When running dumpe2fs on a filesystem formatted with flex_bg, it
prints out the relative offsets for the bitmaps and inode table
badly on 64-bit systems, because the offset is computed as a
large positive number instead of being a negative numer (which
will not be printed at all):

Group 1: (Blocks 0x8000-0xffff) [INODE_UNINIT, ITABLE_ZEROED]
 Block bitmap at 0x0102 (+4294934786), Inode bitmap at 0x0202 (+4294935042)
 Inode table at 0x037e-0x03fa (+4294935422)

This commit prints out the relative offsets for flex_bg
groups as the offset within the reported group.  This makes it
more clear where the metadata is located, rather than simply
printing some large negative number.

Group 1: (Blocks 0x8000-0xffff) [INODE_UNINIT, ITABLE_ZEROED]
 Block bitmap at 0x0102 (bg #0 +258), Inode bitmap at 0x0202 (bg #0 +514)
 Inode table at 0x037e-0x03fa (bg #0 +894)

Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: Fix (minor) memory leaks
Theodore Ts'o [Wed, 1 Dec 2010 23:49:26 +0000 (18:49 -0500)]
mke2fs: Fix (minor) memory leaks

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: Fail if the requested file system type is not defined in mke2fs.conf
Theodore Ts'o [Wed, 1 Dec 2010 23:28:35 +0000 (18:28 -0500)]
mke2fs: Fail if the requested file system type is not defined in mke2fs.conf

If the user passes a file system type which is not defined in
mke2fs.conf (i.e., mke2fs -t xfs ...) change mke2fs so that it prints
a warning and aborts the run.  (There is an exception for ext2, since
that file system does not need a special definition in the fs_types
section of the /etc/mke2fs.conf file.)

In addition, print a warning if there are usage types (specified using
the -T option) which are not defined in /etc/mke2fs.conf.

Addresses-Debian-Bug: #594609

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoresize2fs: Clarify error explaining on-line shrinking is not supported at all
Theodore Ts'o [Sat, 27 Nov 2010 00:42:15 +0000 (19:42 -0500)]
resize2fs: Clarify error explaining on-line shrinking is not supported at all

Addresses-Debian-Bug: #599786

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoresize2fs.8.in: Document that resize2fs -p only works on offline resizes
Theodore Ts'o [Sat, 27 Nov 2010 00:35:30 +0000 (19:35 -0500)]
resize2fs.8.in: Document that resize2fs -p only works on offline resizes

Addresses-Launchpad-Bug: #505719

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoMCONFIG: Fix dependency definitions for the static and profiled blkid library
Theodore Ts'o [Sat, 27 Nov 2010 00:25:26 +0000 (19:25 -0500)]
MCONFIG: Fix dependency definitions for the static and profiled blkid library

The dependency definitions for DEPSTATIC_LIBBLKID and
DEPPROFILED_LIBBLKID incorrectly referenced the non-dependency macros
(i.e., STATIC_LIBUUID instead of DEPSTATIC_LIBUUID).  This resulted in
-luuid showing up as a Makefile dependency, which is of course wrong.

Addresses-Debian-Bug: #604629

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoe2fsck: Fix inode nlink accounting that could cause PROGRAMMING BUG errors
Theodore Ts'o [Sat, 27 Nov 2010 00:09:43 +0000 (19:09 -0500)]
e2fsck: Fix inode nlink accounting that could cause PROGRAMMING BUG errors

This fixes two possible causes for the error message:

WARNING: PROGRAMMING BUG IN E2FSCK!
        OR SOME BONEHEAD (YOU) IS CHECKING A MOUNTED (LIVE) FILESYSTEM.
inode_link_info[X] is Y, inode.i_links_count is Z.  They should be the same!

One cause which can trigger this message is when an inode has an
illegal link count > 65500 --- for example, 65535.  This was the case
in the Debian Bug report #555456.

Another cause which could trigger this message is if an ext4 directory
previously had more than 65000 subdirectories (thus causing
i_link_count to be set to 1), but then some of the subdirectories were
deleted, such that i_link_count should now be the actual number of
subdirectories.

Addresses-Debian-Bug: #555456

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoe2fsck.conf.5: Clarify man page
Theodore Ts'o [Fri, 26 Nov 2010 23:07:32 +0000 (18:07 -0500)]
e2fsck.conf.5: Clarify man page

Clarify the e2fsck.conf(5) man page to make it clear that it applies
for ext4 file systems.

Addresses-Debian-Bug: #591083

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: Use unix_discard() for discards
Lukas Czerner [Thu, 18 Nov 2010 03:38:40 +0000 (03:38 +0000)]
mke2fs: Use unix_discard() for discards

There is generic discard function in struct_io_manager, or in
unix_io_manager to be specific. So use this instead of
mke2fs_discard_blocks().

Since mke2fs_discard_blocks() is not used anymore (and should not be)
remove it.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe2fsck: Discard free data and inode blocks.
Lukas Czerner [Thu, 18 Nov 2010 03:38:38 +0000 (03:38 +0000)]
e2fsck: Discard free data and inode blocks.

In Pass 5 when we are checking block and inode bitmaps we have great
opportunity to discard free space and unused inodes on the device,
because bitmaps has just been verified as valid. This commit takes
advantage of this opportunity and discards both, all free space and
unused inodes.

I have added new set of options, 'nodiscard' and 'discard'. When the
underlying devices does not support discard, or discard ends with an
error, or when any kind of error occurs on the filesystem, no further
discard attempt will be made and the e2fsck will behave as it would
with nodiscard option provided.

As an addition, when there is any not-yet-zeroed inode table and
discard zeroes data, then inode table is marked as zeroed.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe2fsprogs: Add CHANNEL_FLAGS_DISCARD_ZEROES flag for io_manager
Lukas Czerner [Thu, 18 Nov 2010 03:38:37 +0000 (03:38 +0000)]
e2fsprogs: Add CHANNEL_FLAGS_DISCARD_ZEROES flag for io_manager

When the device have discard support and simultaneously discard zeroes
data (and it is properly advertised), then we can take advantage of such
behavior in several e2fsprogs tools.

Add new flag CHANNEL_FLAGS_DISCARD_ZEROES for struct_io_channel so
each io_manager can take advantage of this. The flag is properly set
according to BLKDISCARDZEROES ioctl in unix_open.

Also remove old mke2fs_discard_zeroes_data() function and substitute it
with helper which test this flag.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoe2fsprogs: Add discard function into struct_io_manager
Lukas Czerner [Thu, 18 Nov 2010 03:38:36 +0000 (03:38 +0000)]
e2fsprogs: Add discard function into struct_io_manager

In order to provide generic "discard" function for all e2fsprogs tools
add a discard function prototype into struct_io_manager. Specific
function for specific io managers can be crated that way.

This commit also creates unix_discard function which uses BLKDISCARD
ioctl to discard data blocks on the block device and bind it into
unit_io_manager structure to be available for all e2fsprogs tools.
Note that BLKDISCARD is still Linux specific ioctl, however other
unix systems may provide similar functionality. So far the
unix_discard() remains linux specific hence is embedded in #ifdef
__linux__ macro.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoFix EXT4_FEATURE_RO_COMPAT_HUGE_FILE check
Justin Maggard [Mon, 22 Nov 2010 22:32:28 +0000 (17:32 -0500)]
Fix EXT4_FEATURE_RO_COMPAT_HUGE_FILE check

Creating a 4TB file on a filesystem with the 64bit flag set results in
e2fsck consistently complaining about i_blocks being wrong, with
confusing messages like this:

Inode 29818882, i_blocks is 8388608816, should be 8388608816.  Fix? no

That appears to be caused by ext2fs_inode_i_blocks() checking for the
EXT4_FEATURE_RO_COMPAT_HUGE_FILE in the wrong place.  Fix it.

Signed-off-by: Justin Maggard <jmaggard10@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 years agoMerge branch 'maint'
Theodore Ts'o [Mon, 22 Nov 2010 21:33:39 +0000 (16:33 -0500)]
Merge branch 'maint'

13 years agoUpdate Release Notes, Changelogs, version.h, etc. for 1.41.13 release
Theodore Ts'o [Mon, 22 Nov 2010 21:10:23 +0000 (16:10 -0500)]
Update Release Notes, Changelogs, version.h, etc. for 1.41.13 release

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoDisable lazy inode table initialization when running regression tests
Theodore Ts'o [Mon, 22 Nov 2010 20:20:36 +0000 (15:20 -0500)]
Disable lazy inode table initialization when running regression tests

This avoids test failures when running on new kernels that allow lazy
itable initialization.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: Add discard option into mke2fs.conf
Lukas Czerner [Thu, 18 Nov 2010 13:38:41 +0000 (14:38 +0100)]
mke2fs: Add discard option into mke2fs.conf

Allow to specify discard in mke2fs.conf. Also change the way how to
specify default value for lazy_itable_init. It is better to have all
this defaulting done in the same place so do it in definition (as we do
with discard).

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: Deprecate -K option, introduce discard/nodiscard
Lukas Czerner [Thu, 18 Nov 2010 13:38:39 +0000 (14:38 +0100)]
mke2fs: Deprecate -K option, introduce discard/nodiscard

It would be nice to have consistent "discard" options in every system
tool (mount, fsck, mkfs) taking advantage of discards. Also "discard"
and "nodiscard" is more descriptive instead of just "-K" and can be
easily defaulted and it is something we can not do with "-K".

With this commit you need to specify extended option like this:

./mke2fs -T <fstype> -E nodiscard <device>

in order make a filesystem without discarding the device first. And

./mke2fs -T <fstype> -E discard <device>

respectively.

-K option is with this commit deprecated and should not be used anymore.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: Force the default blocksize to be at least the logical sector size
Theodore Ts'o [Mon, 22 Nov 2010 16:14:35 +0000 (11:14 -0500)]
mke2fs: Force the default blocksize to be at least the logical sector size

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: Set logical/physical sector size from environment for debugging
Theodore Ts'o [Mon, 22 Nov 2010 16:09:00 +0000 (11:09 -0500)]
mke2fs: Set logical/physical sector size from environment for debugging

If MKE2FS_DEVICE_SECTSIZE is set, then this will override the logical
sector size, which is the smallest sector size that can be written
atomically by the device.  (Previously MKE2FS_DEVICE_SECTSIZE set the
physical sector size, which was incorrect given its historical usage.)
The environment variable MKE2FS_DEVICE_PHYS_SECTSIZE will set the
physical sector size, which is the actual sector size used by the
device in reality.

The logical sector size is always less than or equal to the physical
sector size; and writes smaller than the physical sector size but
greather than or equal to the logical sector size will cause a
read-modify-write cycle within the device firmware (or in some
abstract layer lower than the Linux block I/O subsystem, at any rate).

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: Fill in min_io and opt_io with physical sector size
Theodore Ts'o [Mon, 22 Nov 2010 15:50:42 +0000 (10:50 -0500)]
mke2fs: Fill in min_io and opt_io with physical sector size

If the device does not have an explicitly specified minimum io_size or
optimal io_size, and the physical sector size is greater than the
block size, then use the physical sector size as a better-than-nothing
hint.

This should help for SSD's that have a physical sector size of 8k or
16k (which are reportedly will be coming soon).

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: Do not require -F for block size < physical size
Theodore Ts'o [Sun, 21 Nov 2010 14:56:53 +0000 (09:56 -0500)]
mke2fs: Do not require -F for block size < physical size

There will be SSD's out soon that have 8k or 16k phyiscal block sizes.
So don't enforce a requirement that the block size be less than the
physical block size unless the force option is given, and don't give a
warning if the user can't do anything about it (i.e., if the physical
block size is > than the page size).

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agomke2fs: Enable lazy_itable_init if the kernel supports this feature
Theodore Ts'o [Fri, 1 Oct 2010 14:47:38 +0000 (10:47 -0400)]
mke2fs: Enable lazy_itable_init if the kernel supports this feature

Add check for /sys/fs/ext4/features/lazy_itable_init.  If this file
exists, it should be OK to skip initializing the inode table since the
kernel will do it at mount time.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agotune2fs.8: Document that the device can be specified via LABEL= or UUID=
Theodore Ts'o [Sun, 26 Sep 2010 01:41:21 +0000 (21:41 -0400)]
tune2fs.8: Document that the device can be specified via LABEL= or UUID=

Addresses-Debian-Bug: #580236

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoe2fsck: Open the external journal in exclusive mode
Theodore Ts'o [Sun, 26 Sep 2010 01:14:06 +0000 (21:14 -0400)]
e2fsck: Open the external journal in exclusive mode

This prevents accidentally replaying and resetting the journal while
it is mounted, due to an accidental attempt to run e2fsck on an LVM
snapshot of a file system with an external journal.

Addresses-Debian-Bug: #587531

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoresize2fs.8: Make it clear that power-of-2 units are meant by kilobytes
Theodore Ts'o [Sun, 26 Sep 2010 00:23:33 +0000 (20:23 -0400)]
resize2fs.8: Make it clear that power-of-2 units are meant by kilobytes

It's sad that this needs to be made clear....

Addresses-Debian-Bug: #594004

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoe2fsck: Set i_blocks_hi when correcting the i_blocks field in pass #1
Theodore Ts'o [Sat, 25 Sep 2010 02:57:06 +0000 (22:57 -0400)]
e2fsck: Set i_blocks_hi when correcting the i_blocks field in pass #1

For file systems with 64-bit block numbers, we need to make sure we
correct the i_blocks_hi field as well as the i_blocks field when
setting it to the correct value.

Thanks to Justin Maggard for pointing this out.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
13 years agoMerge branch 'maint' into next
Theodore Ts'o [Sat, 25 Sep 2010 02:40:21 +0000 (22:40 -0400)]
Merge branch 'maint' into next

Conflicts:
configure
configure.in
lib/ext2fs/ext2fs.h
misc/mke2fs.c

13 years agolibext2fs: Change EXT2_FLAG_DIRECT_IO to avoid conflict with devel branch
Theodore Ts'o [Sat, 25 Sep 2010 02:22:09 +0000 (22:22 -0400)]
libext2fs: Change EXT2_FLAG_DIRECT_IO to avoid conflict with devel branch

The development branch of e2fsprogs already has a code point assigned
in conflict with EXT2_FLAG_DIRECT_IO.  Fix this.

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