E2fsprogs 1.41.14 (December 22, 2010)
=====================================
Fix spurious complaint in mke2fs where it would complain if the file
system type "default" is not defined in mke2fs.conf.
The resize2fs program will no longer clear the resize_inode feature
when the number reserved group descriptor blocks reaches zero. This
allows for subsequent shrinks of the file system to work cleanly for
flex_bg file systems.
The resize2fs program now handles devices which are exactly 16T;
previously it would give an error saying that the file system was too
big.
E2fsck (and the libext2fs library) will not use the extended rec_len
encoding for file systems whose block size is less than 64k, for
consistency with the kernel.
Programming notes
-----------------
E2fsprogs 1.41.13 would not compile on big-endian systems. This has
been fixed. (Addresses Sourceforge Bug: #3138115)
The ext2fs_block_iterator2() function passed an incorrect ref_offset
to its callback function in the case of sparse files. (Addresses
Sourceforge Bug: #3081087)
Fix some type-punning warnings generated by newer versions of gcc.
E2fsprogs 1.41.13 (December 13, 2010)
=====================================
E2fsck now supports the extended option "-E journal_only", which
causes it to only do a journal replay. This is useful for scripts
that want to first replay the journal and then check to see if it
contains errors.
E2fsck will now support UUID= and LABEL= specifiers for the -j option
(which specifies where to find the external journal). (Addresses
Debian Bug #559315)
E2fsck now has support for the problems/<problem code>/force_no
configuration option in e2fsck.conf, which forces a problem to not be
fixed.
Dumpe2fs will now avoid printing large negative offsets for the bitmap
blocks and instead print a message which is a bit more helpful for
flex_bg file systems.
Mke2fs will now check the file system type (specified with the -t
option) to make sure it is defined in the mke2fs.conf file; if it is
not, it will print an error and abort. If the usage type (specified
with the -T option) is not defined in mke2fs.conf, mke2fs will print a
warning but will continue. (Addresses Debian Bug #594609)
Clarified error message from resize2fs clarifying that on-line
shrinking is not supported at all. (Addresses Debian Bug #599786)
Fix an e2fsck bug that could cause a PROGRAMMING BUG error to be
displayed. (Addresses Debian Bug #555456)
E2fsck will open the external journal in exclusive mode, to prevent
the journal from getting truncated while it is in use due to a user
accidentally trying to run e2fsck on a snapshotted file system volume.
(Addresses Debian Bug #587531)
Fix a bug in e2fsck so it has the correct test for the EOFBLOCKS_FL
flag.
The tune2fs program can now set the uninit_bg feature without
requiring an fsck.
The tune2fs, dumpe2fs, and debugfs programs now support the new ext4
default mount options settings which were added in 2.6.35.
The e2fsck and dumpe2fs programs now support the new ext4 superblock
fields which track where and when the first and most recent file
system errors occurred. These fields are displayed by dumpe2fs and
cleared by e2fsck. These new superblock fields were added in 2.6.36.
Debugfs now uses a more concise format for listing extents in its
stat command. This format also includes the interior extent tree
blocks, which previously was missing from stat's output for
extent-based files.
Debugfs has a new option, -D, which will request Direct I/O access of
the file system.
Mke2fs will skip initializing the inode table if a device supports
discard and the discard operation will result in zero'ed blocks.
Badblocks will now correctly backspace over UTF-8 characters when
displaying its progress bar. (Addresses Gentoo Bug #309909; Addresses
Debian Bugs #583782 and #587834)
E2freefrag will now display the total number of free extents.
Resize2fs -P no longer requires a freshly checked filesystem before
printing the minimum resize size.
Fixed a floating point precision error in a binary tree search routine
that can lead to seg fault in e2fsck and resize2fs.
Fixed a bug in e2fsck where if both the original and backup superblock
are invalid in some way, e2fsck will fail going back to the original
superblock because it didn't close the backup superblock first, and
the exclusive open prevented the file system from being reopened.
Fixed a big in e2freefrag which caused getopt parsing to fail on
architectures with unsigned chars. (Addresses Gentoo Bug: #299386)
Clarified an mke2fs error message so a missed common in an -E option
(i.e., mke2fs -t ext4 -E stride=128 stripe-width=512 /dev/sda1")
results in a more understandable explanation to the user.
Mke2fs now displays the correct valid inode ratio range when
complaining about an invalid inode ratio specified by the user.
Mke2fs now understands the extended option "discard" and "nodiscard",
and the older option -K is deprecated. The default of whether
discards are enabled by default can be controlled by the mke2fs.conf
file.
Mke2fs's handling of logical and physical sector sizes has been
improved to reflect the fact that there will be some SSD's with 8k and
16k physical sectors arriving soon. Mke2fs will no longer force block
size to be the physical sector size, since there will be devices where
the physical sector size is larger than the system's page size, and
hence larger than the maximal supported block size. In addition, if
the minimal and optimal io size are not exported by the device, and
the physical sector size is larger than the block size, the physical
sector size will be used to set the Raid I/O optimization hints in the
superblock.
E2fsck will now display a better, more specific error message when the
user enters a typo'ed device name, instead of blathering on about
alternate superblocks.
Fixed various Debian Packaging Issues
Updated/clarified man pages (Addresses Debian Bugs: #580236, #594004,
#589345, #591083; Addresses Launchpad Bug: #505719)
Update the Chinese, Chzech, Dutch, French, Germany, Indonesian,
Polish, Swedish, and Vietnamese translations.
Programmer's Notes
------------------
Fix a dependency definition for the static and profiled blkid
libraries which could cause compile failures in some configurations.
(Addresses Debian Bug: #604629)
Add support for Direct I/O in the Unix I/O access layer.
Fixed a memory leak in the Unix I/O layer when changing block sizes.
Fixed minor memory leaks in mke2fs.
Added a new function to the ext2fs library, ext2fs_get_memalign().
The tst_super_size test program will check to make sure the superblock
fields are correctly aligned and will print them out so they can be
manually checked for correctness.
Fixed some makefile dependencies for test programs in lib/ext2fs.
Reserved the feature flags and superblock fields needed for the Next3
snapshot feature.
Reserved the feature flags for EXT4_FEATURE_INCOMPAT_DIRDATA and
EXT4_INCOMPAT_EA_INODE.
E2fsprogs 1.41.12 (May 17, 2010)
================================
Mke2fs now gives a correct error message if the external journal
device is not found. (Addresses Red Hat Bug #572935)
Resize2fs -P will now refuse to print a minimum size if the file
system is not clean. Previously it would go ahead and print a minimum
size anyway, which might not be correct, leading to user confusion.
E2fsck now tests for extents that begin at physical block 0 and
rejects them as invalid. (Addresses Google Bug: #2573806)
Fixed a bug in e2fsck which could cause it to crash when trying to
remove an invalid extent and the block bitmaps hadn't yet been loaded.
(Addresses SourceForge Bug: #2971800)
E2fsck now will completely skip time-based checks if the system clock
looks insane or the option broken_system_clock is set in
/etc/e2fsck.conf.
E2fsck would previously report an i_blocks corruption for a 4T file
created using posix_fallocate; this bug has been fixed.
E2fsck will now correctly mark a sparse journal as invalid and will
delete and recreate the journal to address the problem.
E2fsck would previously incorrectly ask the user whether she would
like to abort the file system check after finding a problem --- and
then abort regardless of the user's answer. This is annoying, and
has been fixed.
E2fsck can now continue even if it fails to recreate the resize
inode; previously it would just abort the file system check
altogether.
E2fsck could potentially remove directory entries for inodes found in
the unused region of the inode table; this would generally happen on
ext4 file systems that do not use journalling. This bug has been fixed
by not clearing these directory entries once it has been established
that bg_unused_inodes may not be trustworthy; once pass #2 has been
completed, e2fsck will restart the file system check from the
beginning, and then it will be safe to delete any directory entries
pointing to inodes that appear to be deleted. (Addresses Google Bug:
#2642165)
E2fsck will not try to set the block group checksums if the user
cancels the fsck with a control-C. It's a bad idea to set the
checksums if e2fsck hasn't been completed, and it often results an the
error message, "Inode bitmap not loaded while setting block group
checksum info". (Addresses Launchpad Bug: #582035)
The mke2fs program now queries the kernel for the physical as well as
the logical sector size, and will not allow a blocksize below the
logical, and will strongly encourage a blocksize at least as big as
the physical blocksize. This is needed for 4k sector drives that
emulate 512 byte sector sizes.
Mke2fs will now allow a flex_bg size of 1. This is unusual, and
rarely needed, but it is a legal value.
E2fsck will check for cases where the EOFBLOCKS_FL is set when it is
not needed, and offer to clear it; this is a sign of a kernel bug, but
more importantly, some released kernels may crash when this situation
is encountered on ext4 file systems. (Addresses Google Bug: #2604224)
E2fsck will use the EOFBLOCKS_FL flag exclusively to check whether
i_size is correct. (Kernels starting with 2.6.34 will set
EOFBLOCKS_FL.)
The com_err library will now only output ^M (a CR character) when the
tty is in raw mode.
Update the Czech, Chinese, Dutch, French, Germany, Indonesian, Polish,
and Vietnamese translations.
Fixed various Debian packaging issues --- see debian/changelog for
details. (Addresses Debian Bugs: #571247, #563487)
Programmer's Notes
------------------
The regression test suite now uses its own mke2fs.conf file, so that
downstream distributions want change the mke2fs.conf file which is
distributed in the RPM or dpkg file, without worrying about screwing
up the regression test results.
Always build namei.o so that building with configure --disable-debugfs
works correctly. Long-term, if we care about reduced e2fsprogs
builds, we need a more general solution for deciding what .o files are
needed for a particular build. Given that install floppies are going
(gone?) the way the dodo bird, we probably don't care, though.
(Addresses Sourceforge Bug: #2911433)
Add configure options --enable-symlink-build and
--enable-symlink-install, which allow e2fsprogs be built using
symlinks instead of hard links, and to be installed using symlinks
instead of hard links, respectively. It is useful when the file
system where the build is taking place, or the file system where
e2fsprogs is installed, can't handle hard links for some reason.
(Addresses Sourceforge Bug: #1436294)
Fixed compile warning in mke2fs.c.
E2fsprogs 1.41.11 (March 14, 2010)
==================================
E2fsck will no longer give a fatal error and abort if the physical
device has been resized beyond 2**32 blocks. (Addresses Launchpad
Bug: #521648)
Debugfs has a bug fixed so that "logdump -b <blk>" now properly shows
the allocation status of the block <blk>. (Addresses Debian Bug:
#564084)
E2fsck now prints a much more emphatic and hopefully scary message
when a file system is detected as mounted while doing a read/write
check of the filesystem. Hopefully this will dissuade users from
thinking, "surely that message doesn't apply to *me*" :-(
E2fsck -n will now always open the file system read-only. We now
disallow certain combination of options which previously were manual
exceptions; this is bad because it causes users to think they are
smarter than they really are. So "-n -c", "-n -l", "-n -L", and "-n
-D" are no longer supported. (Addresses Launchpad Bug: #537483)
In e2fsprogs 1.41.10, mke2fs would ask for confirmation to proceed if
it detected a badly aligned partition. Unfortunately, this broke some
distribution installation scripts, so it now just prints the warning
message and proceeds. (Addresses Red Hat Bug: #569021. Addresses
Launchpad Bug: #530071)
Mke2fs would take a long time to create very large journal files for
ext4. This was caused by a bug in ext2fs_block_iterate2(), which is
now fixed.
E2fsck now understands the EOFBLOCKS_FL flag which will be used in
2.6.34 kernels to make e2fsck not complain about blocks deliberately
fallocated() beyond an inode's i_size.
E2fsprogs 1.41.10 introduced a regression (in commit b71e018) where
e2fsck -fD can corrupt non-indexed directories when are exists one or
more file names which alphabetically sort before ".". This can happen
with ext2 filesystems or for small directories (take less than a lock)
which contain filenames that begin with a space or some other
punctuation mark. (Addresses Debian Bug: #573923, Addresses Launchpad
Bug: #525114)
Programmer's Notes
------------------
Add new test, f_rehash_dir, which checks to make sure e2fsck -D works
correctly.
The libcom_err function now has support for Heimdal's com_right_r
function(). (Addresses Sourceforge Bug: #2963865, Addresses Debian
Bug: #558910)
E2fsprogs 1.41.10 (February 7, 2010)
====================================
Fix resize2fs bug which causes it to access invalid memory.
Add libss support for libreadline.so.6.
Fix e2fsck's check for extent-mapped directory with an incorrect file
type.
Add new e2fsck.conf configuration option, default/broken_system_clock
to support systems with broken CMOS hardware clocks. Also, since too
many distributions seem to have broken virtualization scripts now,
e2fsck will by default accept dates which are off by up to 24 hours by
default. (Addresses Debian Bugs: #559776, #557636)
Fix a bug where mke2fs may not use the best placement of the inode
table when there is only room for a single block group in the last
flex_bg.
E2fsck is now smarter when it needs to allocate blocks in the course
of fixing file system problems. This reduces the number of spurious
differences found in pass #5.
E2fsck will no longer rehash directories which fit in a single
directory block.
E2fsck now correctly handles holes in extent-mapped directories (i.e.,
sparse directories which use extents).
Fix big-endian problems with ext2fs_bmap() and ext2fs_bmap2().
Fix a bug in filefrag where on platforms which can allow file systems
with 8k blocks, that it doesn't core dump when it sees a file system
with 8k block sizes. (Thanks to Mikulas Patocka for pointing this
out.)
E2fsck will correctly fix directories that are have an inaccurate
i_size as well as other problems in a single pass, instead of
requiring two e2fsck runs before the file system is fully fixed.
Fix e2fsck so it will correctly find and detect duplicate directory
entries for non-indexed directories. ( Addresses Sourceforge Bug:
#2862551)
Mke2fs will use BLKDISCARD to pre-discard all blocks on an SSD or
thinly-provisioned storage device. This can be disabled using the -K
option.
Enhance libext2fs so it works around bug in Linux version 2.6.19
and earlier where the /proc/swaps file was missing the header on
the first line.
Fix bug in Linux version 2.6.19 and earlier where the /proc/swaps file
was missing the header on the first line.
Fix some big-endian bugs in e2fsck and libext2fs.
Fix resize2fs so it works correctly on file systems with external
journals instead of failing early with the error "Illegal inode
number".
Fix libss so that it does not seg fault when using a readline library
which does not supply a readline_shutdown() function.
Updated dumpe2fs's usage message so it correctly gives the right
arguments summary for "-o superblock=<num>" and "-o blocksize=<num>".
(Addresses Launchpad Bug: #448099)
Teach libext2fs to ignore the high 32 bits of the i_blocks field
when huge_file file system feature is set, but the inode does not
have the HUGE_FILE_FL flag set.
Fix e2fsck's handling of 64-bit i_blocks fields.
E2fsck will now print "Illegal indirect block", "Illegal
double-indirect block", etc., instead of "Illegal block #-1" or
"Illegal block #-2", etc. This makes it easier for users to
understand what has gone wrong. (Addresses SourceForge Bug: #2871782)
Mke2fs now will obtain get device topology information from blkid and
use it to populate the superblock stride and stripe sizes. It will
also warn if the block device is misaligned
Fix file descriptor leaks in fsck and debugfs. (Addresses Novell Bug:
##524526)
Fix the libext2fs library code to round up the bitmap size to a 4-byte
boundary, to prevent spurious seg faults caused by the x86
architecture. This doesn't affect Linux systems, but was a major
problem on a number of *BSD systems. (Addresses Sourceforge Bug:
#2328708)
Fix resize2fs's minimum size required for a file system so it doesn't
fail when "resize2fs -M" is run. (Addresses RedHat Bugzilla: #519131)
Dumpe2fs now prints summary information about the contents of the
journal.
Avoid printing scary error messages when e2fsck starts running
problems on low-memory systems, as it tends to panic and mislead the
user. (Addresses Debian Bug: #509529)
Fix blkid's modules.dep parser so it handles compressed (.ko.gz)
modules files. (Address Red Hat Bug: #518572)
Fix tune2fs so it can add a journal when an extent-enabled file system
is mounted. (Addresses Launchpad bug: #416648)
Update Czech, Indonesian, Polish and Vietnamese translations (from the
Translation Project).
Update/clarify man pages. (Addresses Sourceforge Bug: #2822186)
Fixed various Debian packaging issues --- see debian/changelog for
details. (Addresses Debian Bugs: #540111)
Programmer's Notes
------------------
The configure script supports the --with-cc, --with-ccopts, and
--with-ldopts options. Instead, the more standard use of CC=,
CCFLAGS=, and LDFLAGS= in the configure command line is used instead.
Also, --with-ld, which never worked, was also removed.
(Addresses Sourceforge Bug: #2843248)
The in-tree header files are only used if the in-tree uuid or blkid
libraries are used. Otherwise, use the system-provided uuid or blkid
header files if using the system-provided libraries.
Fix some build failures caused by --disable-* configure options.
Work around a bug in autoconf 2.64.
E2fsprogs 1.41.9 (August 22, 2009)
==================================
Fix a bug in e2fsck routines for reallocating an inode table which
could cause it to loop forever on an ext4 filesystem with the FLEX_BG
filesystem feature with a relatively rare (and specific) filesystem
corruption. This fix causes e2fsck to try to find space for a new
portion of the inode table in the containing flex_bg, and if that
fails, the new portion of the inode table will be allocated in any
free space available in the filesystem.
Make e2fsck less annoying by only asking for permission to relocate a
block group's inode table once, instead of for every overlapping
block. Similarly, only ask once to recompute the block group
checksums, instead of once for each corrupted block group's checksum.
Fix filefrag to avoid print the extent header if the FIEMAP ioctl is
not present, and it needs to fall back to using the FIBMAP ioctl.
Fix filefrag to correctly print the number of extents for zero-length
files. (Addresses Debian Bug: #540376)
Filefrag now has a -B option which forces the use of the FIBMAP ioctl
to more easily debug the FIBMAP code.
Fixed filefrag for non-extent based files.
Add a new program, e2freefrag, which displays information about the
free space fragmentation in an ext2/3/4 filesystem.
Fix inode resizing via tune2fs -I so that it works correctly in the
face of non-empty bad blocks inodes, and if the filesystem was
formatted using the "mke2fs -E stride=N" option for RAID arrays.
Fix regression in ext2fs_extent_set_bmap() caused e2fsck -fD to fail
and corrupt large directories if the directory needs to shrink by more
than one block. (Addresses Debian Bug: #537510)
Fix e2fsck's buggy_init_scritps=1 so that the if the last write and/or
last mount times are in the future, they are corrected even if
buggy_init_scripts is set. This is needed because otherwise resize2fs
will refuse to resize the filesystem, even after running "e2fsck -f".
(Addresses Launchpad bug: #373409)
E2fsck will now print much fuller information when the last mount time
or last written time is in the future, since most people can't seem to
believe their distribution has buggy init scripts, or they have a
failed CMOS/RTS clock battery.
Enhance dumpe2fs to dump the extent information via the 'stat'
command, and more detailed extent information via the new command
'dump_extents'.
Update French, Polish, Czech, and Swedish translations from the
Translation Project.
Fixed various Debian packaging issues --- see debian/changelog for
details.
Programmer's Notes
------------------
Fixed miscellaneous gcc -Wall warnings.
Fixed memory leak in error path in ext2fs_block_iterate2()
Fixed non-Linux build of the intl directory by adding support for the
E/Q/V macros.
The bitmap read/write functions now treat uninitialized bitmaps as
unallocated; this fixes a number of problems in all e2fsprogs for ext4
filesystems when there is a need to allocate new blocks or inodes, and
there aren't any free blocks or inodes in the already-used block
groups.
Improve ext2fs_extent_set_bmap() to avoid creating new extents which
get inserted into the extent tree when they are not needed.
E2fsprogs 1.41.8 (July 11, 2009)
================================
Fix resize2fs's online resizing, fixing a regression which in
e2fpsrogs 1.41.7. (Addresses Debian Bug: #535452)
Fix potential filesystem corruptions caused by using resize2fs to
shrinking ext4 filesystems with extents enabled. (Addresses Red Hat
Bug: #510379)
Optimize uuid_generate() to avoid running uuidd if it is not setuid or
setgid and the currently running program doesn't have write access to
the uuidd work directory.
Add safety checks (for non-Linux systems) so that uuidd isn't run with
file descriptors 0, 1, and 2 closed; and if they are closed, uuidd
will be careful not to close the file descriptor for its unix domain
socket when it detaches itself from the controlling tty. Also add
safety checks so that if the unix domain socket between the uuid
library and uuidd program is closed for any reason, both the library
and the uuidd will return an appropriate error code instead of looping
in an infinite loop.
The e2croncheck script, which creates an LVM snapshot and then checks
the ext3/4 filesystem via the LVM snapshot, has been added to the
contrib directory.
Fix the filefrag program for files that have more than 144 extents.
Update French, Polish, Czech, Indonesian, and Swedish translations from
the Translation Project.
Fixed various Debian packaging issues --- see debian/changelog for
details. (Addresses Debian Bug #535530)
Update/clarify man pages.
Programmer's Notes
------------------
Fix compilation problem when configured with --disable-uuid.
Don't build uuidgen if configured with --disable-uuid.
Add the new library function ext2fs_test_inode_bitmap_range(), and
optimized ext2fs_test_block_bitmap_range(), which will be needed for
future optimizations for e2fsck.
Fix makefile dependencies for libcom_err so that the Makefiles work
well on non-GNU make program. (Addresses Sourceforge Patches: #2813809)
Enhance the build system so that the full set of commands executed by
the Makefiles are displayed, instead of the Linux kernel summary
output, if the build was configured with --enable-verbose-makecmds, or
if GNU make is in use and the V variable is non-null, i.e., via "make
V=1".
E2fsprogs 1.41.7 (June 29, 2009)
================================
Fix a bug in libext2fs which can cause e2fsck and resize2fs to write
uninitialized data into the portion of the inode beyond the first 128
bytes when operating on inodes mapped via extents; potentially
corrupting filesystems.
Fix memory leaks in e2fsprogs, including a very large memory leak
which can cause e2fsck to run out of memory when checking very large
filesystems using extents.
The logsave program will now filter out the ^A and ^B characters when
writing to the console.
Harden ext2fs_validate_entry() so that lsdel will not read beyond the
end of the block even if the directory block is corrupted.
Fix debugfs from core dumping if the logdump command fails to open the
output file.
Enhance badblocks to print the currently tested block number when
interrupted with ^C.
Fix lsattr to exit with a non-zero status when it encounters errors.
(Addresses RedHat Bugzilla #489841)
Fix e2fsprogs to use the same encoding for rec_len in 64k (and larger)
filesystems as the kernel when encoding rec_len >= 64k. For 64k
filesystems (currently all that is supported by e2fsprogs, this is
only a minor corner case).
Resize2fs will now update the journal backup fields in the superblock
if the journal is moved; this avoids an unnecessary full fsck after
resizing the filesystem. (Addresses RedHat Bugzilla: #505339)
Fix libext2fs to properly initialize i_extra_size when creating the
journal and resize inodes.
Change badblocks to allow block sizes larger than 4k.
Fix the filefrag program so it correctly checks for errors from the
fiemap ioctl.
Update Chinese and Czech translation from the Translation Project.
Clean up various man pages. (Addresses Debian Bug #531385 and #523063)
Programmer's Notes
------------------
Add --disable-libuuid option to configure which uses an in-system
installed version of libuuid; the private version is enabled by
default.
Add --valgrind-leakcheck option to the test_script program which runs
valgrind with the appropriate options to find memory leaks.
E2fsprogs 1.41.6 (May 30, 2009)
===============================
Fix a critical regression in e2fsck introduced version 1.41.5 which
can undo updates to the block group descriptors after a journal replay.
If e2fsck receives an I/O error while replaying the journal, prompt
the user whether they want to ignore the error or not; if e2fsck is
run in preen mode, force fsck to abort the preen pass.
Fix a bug which would cause e2fsck to core dump if the filesystem
contains a corrupt extent header, and the user declines to clear the
inode. (Addresses Sourceforge Bug: #2791794)
Fix e2fsck to restart only once in the case of multiple block groups
which inodes in use that are in the uninitialized part of the block
group's inode table.
To reduce user confusion, if the /etc/mtab file is missing
ext2fs_check_mount_point and ext2fs_check_if_mounted will return a
new, explicit error code to indicate this case. This will cause
e2fsck to give a clearer error message when the user is using buggy
rescue CD's that don't properly set up /etc/mtab. (Addresses Debian
Bug: #527859)
Fix e2fsck so that if the primary group descriptors are corrupted, but
the backup superblock is entirely invalid, to go back to using (and
fixing) the primary group descriptors instead of completely giving up
on the filesystem. (Addresses Debian Bug: #516820)
Change e2fsck to not abort a preen pass if an inode is found to have
i_file_acl_hi non-zero. Ext3 filesystems don't care, and newer
kernels (post 2.6.29) will ignore this field. So let's fix it
automatically during the preen pass instead of forcing the user to fix
this up manually. (Addresses Debian Bug: #526524)
Add resource tracking for e2fsck passes 1b through 1d.
Speed up e2fsck by eliminating unnecessary journal checks if the
filesystem is already mounted and doesn't need recovery (since the
kernel would have run the journal when the filesystem was mounted.)
Also speed up e2fsck by avoiding unnecessary block group descriptor
updates in ext2fs_close().
Add support to chattr to migrate inodes from using direct/indirect
blocks to extents.
Avoid corrupting the filesystem if there is an attempt to shrink a
filesystem using resize2fs smaller than possible by making
ext2fs_set_bmap() more careful not to delete the old block until the
new block can be inserted. In addition, fix a bug in how the minimum
size of the filesystem (plus a safety margin) is calculated, and
modify resize2fs to refuse to shrink the filesystem below that minimum
size without the force flag.
Teach blkid to try to figure out DM device names using take advantage
of information in sysfs when running on kernels (2.6.29 and later)
which provide this information in /sys/block/dm-<N>/dm/name; this is
much faster than scanning for the device number in /dev/mapper.
Fix blkid to prefer /dev/mapper/<name> device names instead of the
private /dev/dm-X names.
Add an -a option to debugfs's close_filesys command which writes any
changes to the superblock or block group descriptors to all of the
backup superblock locations.
Add support to the filefrag program to use the FIEMAP ioctl.
Update Chinese translation from the Translation Project.
Clean up various man pages. (Addresses Red Hat Bugzilla: #502971 and
Launchpad Bug: #381854)
Fixed various Debian packaging issues --- see debian/changelog for
details. (Addresses Debian Bug #506064)
Programmer's Notes
------------------
Add test code to make sure e2fsck's problem.c doesn't have two problem
codes assigned to duplicate values.
Avoid using a hard-coded path for /bin/rm in block's test_probe.in.
(Addresses Sourceforge Bug: #2780205)
Clean up e2fsck by removing #ifdef RESOURCE_TRACK by adding an empty
function declaration for init_resource_track() and
print_resource_track() when RESOURCE_TRACK is not defined.
The test code which is used to build the tst_csum program has been
moved from from tst_csum.c into csum.c under an #ifdef DEBUG to to
avoid compile problems caused by not having a prototype for
ext2fs_group_desc_csum(). (Addresses Sourceforge Bug #2484331)
Update the config.guess and config.sub file to the latest from the
FSF, to allow e2fsprogs to build on the avr32 platform. (Addresses
Debian Bug: #528330)
Add a new function, ext2fs_extent_open2(), which behaves like
pext2fs_extent_open(), but uses the user-supplied inode structure
when opening an extent instead of reading the inode from disk.
E2fsprogs 1.41.5 (April 23, 2009)
=================================
Fix a number of filesystem corruption bugs in resize2fs when growing
or shrinking ext4 filesystems off-line (i.e., when the ext4 filesystem
is not mounted).
Debugfs can now set i_file_acl_high via the set_inodes_field command,
and print a 64-bit file acl. This is useful for debugging filesystem
corruptions where the high bits of i_file_acl_high are set. E2fsck
will detect and fix non-zero i_file_acl_high on 32-bit filesystems
since some Linux kernel versions pay attention to this field even when
they shouldn't.
Speed up e2fsck when checking clean filesystems by avoiding
unnecessary block reads, and coalescing the block group descriptor
blocks so they are read using a single read operation.
The libuuid library will now close all file descriptors before running
uuidd. This avoids problems when the calling program has open sockets
which then never get closed because uuidd is a long-running helper
daemon. (Addresses Launchpad bug: #305057)
In order to avoid unnecessary full filesystem checks by e2fsck after
an on-line resize, e2fsck will ignore the NEEDS_RECOVERY flag set on
the backup superblocks. (Addresses Red Hat Bugzilla: #471925)
Mke2fs will avoid trying to create the journal when run in
superblock-only mode (mke2fs -S), since the left-over journal in the
inode table will cause mke2fs to fail.
Fix a bug in libext2fs functions that check to see if a particular
device or filesystem image is mounted, which would cause these
functions to report that a file identical to the (relative) pathname
used by a pseudo-filesystem was mounted when in fact it was not.
Update Czech translation from the Translation Project.
Add Chinese (simplified) translation from the Translation Project.
Fix support for external journals (which was broken in e2fsprogs
1.41.4).
Fix a regression in debugfs where the "stat" command when no
filesystem was open would cause debugfs to crash with a segmentation
violation.
Starting in the 2.6.29 linux kernel, the ext4 filesystem driver can be
used to support filesystems without a journal. Update the blkid
library so it understands this.
The blkid library will remove an entry from the blkid cache
representing the entire disk if partitions are found, since presumably
the device previously had no partition table, but has now transitioned
to using a partition table.
Add a check to mke2fs and tune2fs that the argument to the -m option
(which specifies the reserved ratio) must be greater than zero.
(Addresses Debian Bug: #517015)
Add support for tracking the number kilobytes written to the
filesystem via the superblock field s_kbytes_written. It will be
updated by the kernel as well as by e2fsprogs programs which write to
the filesystem. This is useful for tracking the wear to filesystems
on Solid State Drives.
Fix compatibility issue in the libext2fs info file and makeinfo
version 4.12. (Addresses Red Hat Bugzilla: #481620)
Update/clarify man pages. (Addresses Debian Bug: #515693, #365619)
Fixed various Debian packaging issues --- see debian/changelog for
details. (Addresses Debian Bug: #506279)
Programmer's Notes
------------------
Fix Hurd compilation problem in e2fsck and tune2fs (Addresses Debian
Bug: #521602)
Fix various gcc compilation warnings and other programming cleanups.
(Addresses Red Hat Bugzilla: #486997)
Add support for building the blkid command statically.
Add support for disabling the built-in blkid library in favor of a
system-installed blkid implementation via the configure option
--disable-libblkid.
E2fsprogs 1.41.4 (January 27, 2009)
===================================
Fixed a bug which could sometimes cause blkid to return an exit value
of zero for a non-existent device (Addresses Debian Bug: #502541)
Blkid will now recognize ext3 filesystems that have the test_fs flag
set as ext3 filesystems.
The blkid library will now recognize btrfs filesystems and swap
devices currently used by user-level software suspend.
Tune2fs now updates the block group checksums when changing the UUID
to avoid causing e2fsck to complain vociferously at the next reboot.
Tune2fs's inode size resizing algorithms have been fixed so it is not
vastly inefficient for moderate-to-large filesystems, due to some
O(n**2) and O(n*m) algorithms that didn't scale well at all.
Fix tune2fs's inode resizing algorithm so it will not corrupt
filesystems laid out for RAID filesystems; in addition, tune2fs will
refuse to change the inode size for filesystems that have the flex_bg
feature enabled. (This is a limitation in the current implementation
of tune2fs -I.)
E2fsprogs 1.41 broke debugfs's logdump command for normal ext3/4
filesystems with 32-bit block numbers, when the headers for 64-bit
block numbers was added. This regression has been fixed.
Debugfs's ncheck command has been fixed to avoid printing garbage
characters at the end of file names.
Fix resize2fs for ext4 filesystems. Some blocks that that need moving
when shrinking filesystems with uninit_bg feature would not be moved.
In addition, blocks and inode table blocks were not being correctly
freed when shrinking filesystems with the flex_bg feable, which caused
resize2fs -M to fail. Finally, when blocks are moved, make sure the
uninitialized flag in extents is preserved.
Fix bug which caused dumpe2fs to abort with an error if run on a
filesystem that contained an external journal.
Some distributions used "mke3fs" as an alias for "mkfs.ext3"; check
for this in argv[0] to provide better legacy support for these
distributions. This is a practice that should NOT be continued,
however.
Mke2fs now has a new option -U, which allows the user to specify the
UUID that should be used for the new filesystem.
Mke2fs will treat devices that are exactly 16TB as if they were 16TB
minus one block. This allows users who have read that ext3 supports
up to 16TB filesystems and who create a 16TB LVM to not get confused,
since the true limit is really 16TB minus one block.
E2fsck will no longer abort an fsck run if block group has an errant
INODE_UNINIT flag.
E2fsck now distinguishes between fragmented directories and fragmented
files in verbose mode statistics and in the fragcheck report.
Fix a bug in e2fsck which caused it double count non-contiguous
extent-based inodes.
E2fsck will leave some slack space when repacking directories to allow
room for a few directory entries to be added without causing leaf
nodes to be split right away.
Fix a bug which caused e2fsck to crash when it comes across a
corrupted interior node in an extent tree with the error message:
"Error1: Corrupt extent header on inode XXXXXX"
E2fsck problem descriptions involving the journal are no longer
referred to as "ext3" problems, since ext4 filesystems also have
journals.
Fix a long-standing bug in e2fsck which would cause it to crash when
replying journals for filesystems with block sizes greater than 8k.
Update Catalan translation from the Translation Project.
Fixed various Debian packaging issues --- see debian/changelog for
details. (Addresses Debian Bugs: #503057, #502323, #511207)
Programmer's Notes
------------------
Fix build of e2fsck.profiled, and add support for building profiled
binaries in the misc directory if configured with --enable-profile.
The ext2fs_open() function now performs more sanity checks on the
superblock to avoid potential divide by zero errors by other parts of
library.
The ext2fs_read_inode_full() function now has a safety check to avoid
a segmentation fault on corrupted filesystems.
The ext2fs_new_inode() function now has a sanity check so that if the
s_first_inode field in the superblock is insane, it will return
EXT2_ET_INODE_ALLOC_FAIL instead of returning an invalid inode number.
To avoid segmentation faults, ext2fs_block_alloc_stats() and
ext2fs_inode_alloc_stats() now validates the passed inode or block
number to avoid overrunning an array boundary.
Various signed/unsigned errors for variables containing block numbers
have been fixed.
Accommodations for gcc's stupidity in not realizing that constant
strings that do not contain a '%' character are safe to use in format
strings have been made so that distributions that want to compile
e2fsprogs with -Werror=format-security have an easier time doing so.
Added a new 64-bit getsize interface, ext2fs_get_device_size2().
Added the utility make-sparse.c to the contrib directory.
The ext2fs_block_iterate2() function now reflects errors from
ext2fs_extent_set_bmap() to the caller, if the callback tries to
change a block on an extent-based file, and ext2fs_extent_set_bmap()
fails for some reason (for example, there isn't enough disk space to
split a node and expand the extent tree.
The ext2fs_block_iterate2() function will preserve the uninit flag in
extents when the callback function modifies a block in an extent-based
file.
E2fsck will now flag filesystems that have an insane s_first_ino field
in their superblock, and attempt to use a backup superblock to repair
the filesystem.
E2fsprogs 1.41.3 (October 12, 2008)
===================================
E2fsck has been fixed so it prints the correct inode number for
uinit_bg related problems.
E2fsck will now offer to clear the test_fs flag if the ext4 filesystem
is available on linux. This can be disabled via a configuration
option in /etc/e2fsck.conf.
Fix a file descriptor leak in libblkid when checking to see if an ext4
or ext4dev module exists.
Fix a bug in e2fsck where in preen mode, if there are disk I/O errors
while trying to close a filesystem can lead to infinite loops.
(Addresses Red Hat Bugzilla #465679)
Fix a bug in resize2fs where passing in a bogus new size of 0 blocks
will cause resize2fs to drop into an infinite loop. (Addresses Red
Hat Bugzilla: #465984)
Add a check in the Unix I/O functions in libext2fs so that when a
device is opened read/write, return an error if the device is
read-only using the BLKROGET ioctl.
Fix debugfs's ncheck command so that it prints all of the names of
hardlinks in the same directory.
Fix a bug in libblkid so it correctly detects whether the ext4 and
ext4dev filesystems are available, so that the ext4dev->ext4
fallback code works correctly.
Programmer's Notes
------------------
Fix a parallel build problem by making sure util/subst is built before
trying to build the lib/et directory. (Addresses Sourceforge Bug:
#2143281)
Updated "make depend" information for crc16.o
E2fsprogs 1.41.2 (October 2, 2008)
==================================
Fix e2fsck's automatic blocksize detection. This fixes a regression
from e2fsprogs 1.40.7 which caused e2fsck to fail if the user
specifies a block number using the -b option if the blocksize option
isn't also specified using -B. Unfortunately, users very commonly
invoke e2fsck using "e2fsck -b 32768 /dev/hdXXX" to use the backup
superblock; in fact e2fsck will often suggest this kind of command
line. Oops.
Enhance the debugfs's "ncheck" command so it will print all of the
pathnames for the specified inodes. (Previously, in some cases ncheck
might not print a pathname for an inode at all if some of the other
inodes had multiple hard links.)
Enhance debugfs's "hash" command so the hash seed can be specified via
a command-line option. In addition, allow the hash algorithm to be
specified by name instead of just by number.
Fix e2fsck so that we don't accidentally print the translation file's
header when asking the user a custom question so there is no prompt
defined for a particular problem record. For example, the question
"Run journal anyway" will get the PO header tacked on because e2fsck
erroneously passed the null string to _(). (Addresses Launchpad Bug:
#246892)
Enhance badblocks so that it can test a normal file which is greater
than 2GB.
Enhance the badblocks command so that it displays the time and
percentage complete when in verbose mode. (Addresses Debian Bug:
#429739)
Fix a potential memory leak in a error handling path in debugfs's
ncheck function.
Fix a potential memory corruption problem if a memory allocation fails
in resize2fs.
Fix the usage message for debugfs's logdump command to be consistent
with its man manpage.
Update Polish, French, Vietnamese, Dutch, Indonesian, German, Czech,
and Swedish translations from the Translation Project.
Add documentation for the file I/O functions to the libext2fs.texinfo
file. (Addresses Debian Bug: #484877)
Update and clarified various man pages. (Addresses Launchpad Bug
#275272; Addresses Debian Bugs: #498100, #498101, #498102, #498103)
Fixed various Debian packaging issues --- see debian/changelog for
details. (Addresses Debian Bug: #497619)
Programmer's Notes
------------------
Fix a potential file descriptor leak in libcom_err by setting the
close-on-exec flag for a fd used for debugging. (Addresses Red Hat
Bugzilla #464689)
Fix a potential race in libcom_err by using sem_post/sem_init. SuSE
has been carrying a patch for a long time to prevent a largely
theoretical race condition if a multi-threaded application adds and
removes error tables in multiple threads. Unfortunately SuSE's
approach breaks compatibility by forcing applications to link and
compile with the -pthread option; using pthread mutexes has
historically been problematic. We fix this by using sem_post/sem_init
instead.
Fix e2fsprogs-libs build failure due to 'subs' target. (Addresses
Sourceforge Bug: #2087502)
Avoid linking e2initrd_helper, debugfs, blkid, and fsck with unneeded
libraries when using ELF shared libraries.
Fix ELF shared library when building on systems that don't already
have the e2fsprogs shared libraries already installed. (Addresses
Sourceforge Bug: #2088537)
Fix the pkg-config files so they work correctly when linking with
static libraries and fix the include directory so programs don't have
to use #include <ext2fs/ext2fs.h>, but can use #include <ext2fs.h>
instead. (Addresses Sourceforge Bug: #2089537)
Make sure ext2fs_swab64() is compiled for all platforms, and not just
for x86. (Addresses Debian Bug: #497515)
Remove the unused ext2fs_find_{first,ext}_bit_set() functions for all
non-x86 platforms. (They had been removed for x86 earlier.)
Fix diet libc compilation support, which had bitrotted due to lack of
TLC. Fixing this improves general portability.
When installing the link library when using ELF shared libraries,
avoid using absolute pathnames if the link library and the shared
library are installed in the same directory. (Addresses Sourceforge
Bug: #1782913)
Fix gen-tarball so it will work even if the top-level directory has
been renamed to something other than "e2fsprogs". Also make
gen-tarball print the size of the resulting tar.gz file.
E2fsprogs 1.41.1 (September 1, 2008)
====================================
Many people are forgetting to update their mke2fs.conf file, and this
causes ext3, ext4, and ext4dev filesystems won't get created with the
proper features enabled. We address this in two ways. First, mke2fs
will issue a warning if there is not definition for an ext3, ext4, or
ext4dev filesystem and the user is trying to create such a filesystem
type. Secondly, when installing from a source build, "make install"
will provide basic configuration file handling for /etc/mke2fs.conf.
If it exists, and does not mention ext4dev, it will be moved aside to
/etc/mke2fs.conf.e2fpsrogs-old and the new /etc/mke2fs.conf file will
be installed. If the existing /etc/mke2fs.conf file does mention
ext4dev, then "make install" will install official mke2fs.conf file as
/etc/mke2fs.conf.e2fsprogs-new and issue a message to the user that
they should look to see if any changes need to be merged.
The mke2fs program will now create the journal in the middle of the
filesystem, since this minimizes seek times on average for fsync-heavy
workloads. In addition, mke2fs will now create journals using extents
for filesystems that support them. This results in a more efficient
encoding for the journal since it eliminates the need for using
indirect blocks.
The mke2fs program will avoid allocating an extra block to the
journal. (Addresses Sourceforge Bug: #1483791)
Mke2fs will correctly enforce the prohibition against features
(specifically read-only features) in revision 0 filesystems. (Thanks
to Benno Schulenberg for noticing this problem.)
Mke2fs previously would occasionally create some slightly non-optimally
placed inode tables; this bug has been fixed.
The mke2fs and tune2fs programs now print the correct usage message
describing the maximum journal size. (Addresses Debian Bug: #491620)
Add support for setting the default hash algorithm used in b-tree
directories in tune2fs (from a command-line option) or mke2fs (via
mke2fs.conf). In addition, change the default hash algorithm to
half_md4, since it is faster and better.
The blkid library will now recognize MacOS hfsx filesystems, and
correctly extract the label and uuid for hfs, hfsx, and hfsplus
filesystems. (Addresses Sourceforge Feature Requests: #2060292)
The blkid library has improved detection of JFS and HPFS filesystems.
(Addresses Launchpad Bug: #255255)
The blkid library is now much more efficiently handling devicemapper
devices, mainly by no longer using the devicemapper library. This can
speed up access for systems with a large number of device mapper
devices.
Blkid had a number of cache validation bugs in libblkid that have been
fixed. (Addresses Debian Bug: #493216)
Resize2fs will now properly close out the "updating inode references"
progress bar so there is a newline printed before printing the final
"resize is successful" message.
Resize2fs will now correctly handle filesystems with extents and/o
uninitialized block groups correctly when file/directory blocks need
to relocated (i.e., when shrinking a filesystem or if the resize_inode
is not present). To support this, the ext2fs library now supports
initializing inode and block bitmaps that are not yet initialized when
allocating them using ext2fs_new_block() and ext2fs_new_inode(). In
addition, e2fs_block_iterate2() can now support changing the location
of interior nodes of an extent tree, and ext2fs_extent_set_bmap() has
been optimized to avoid creating unnecessary new extents when updating
the location of blocks in the extent tree. This will also help out
e2fsck's recovery of obscurely corrupted filesystems with extents,
when blocks are claimed by multiple inodes.
Add support for on-line resizing ext4 filesystem with the flex_bg
filesystem feature. The method for doing so is not optimal, but to do
a better job will require kernel support.
E2fsprogs 1.41.0 introduced a bug in libext2fs which caused e2image and
debugfs programs to not be able to read e2image files; the signed
vs. unsigned bug in the code which read bitmaps from the e2image has
been fixed. (Addresses Debian Bug: #495830)
Resize2fs is now correctly managing the directory in-use counts when
shrinking filesystems and directory inodes needed to be moved from one
block group to another. This bug has been around since e2fsprogs
1.26, and is largely harmless, but does cause a filesystem corruption
which will be flagged by e2fsck after the filesystem has been shrunk.
E2fsck will no longer issue spurious complaints about the inode size
caused by very large extent-based files, and by blocks reallocated
using fallocate() with the FALLOC_FL_KEEP_SIZE option. (Addresses
Kernel Bugzilla: #11341)
Mke2fs will now set the creation timestamp on the lost+found directory
and the root directory. (More generally, all new inodes created using
the ext2fs library will correctly set the creation timestamp.)
E2fsck now correctly calculates ind/dind/tind statistics in the
presence of extent-based files. In addition, "e2fsck -v" will report
statistics of the depth of extent trees in the filesystem. E2fsck can
also give an inode fragmentation report using "e2fsck -E fragcheck"
which can be useful when debugging the kernel block allocation
routines.
Fix support for empty directory blocks in ext4 filesystems with 64k
blocksize filesystems.
E2fsck will now print the depth of corrupt htree directories.
Debugfs's htree command now correctly understands extent-based
directories. It will also print out the minor hash as well as the
major hash.
Debugfs has a new command which will print the supported features of
e2fsprogs, to enable scripts to know whether the installed version of
e2fsprogs can support a specific feature.
Debugfs will now write files using extents for filesystems that
support them.
The error message printed by "tune2fs -I" if the inode size was too
small was rather confusing, so it has been improved. Also, we won't
try to create an undo log until we know that command-line-specified
parameters such as "tune2fs -I <inode size>" are valid.
Given some filesystems found "in the wild" that had non-zero block
group checksums even though the uninit_bg/gdt_sum feature was not
enabled, e2fsck would issue spurious error messages. Teach
ext2fs_group_desc_csum_verify() to ignore the block group checksum
entirely if the feature flag is not set. (Addresses Debian Bug:
#490637)
The blkid program will now print out a user-friendly listing of all of
the block devices in the system and what they contain when given the
-L option. (Addresses Debian Bug: #490527)
The filefrag program now has a more accurate calculation for the
number of ideal extents. (Addresses Debian Bug: #458306)
The test I/O manager is now enabled by default, but its overhead is
only incurred when it would be enabled via the TEST_IO_FLAGS or
TEST_IO_BLOCK environment variables.
Typographical errors in various program strings and usage messages
have been fixed; most of these were pointed out by the e2fsprogs
message catalog translators. (Thanks, translators!)
Update and clarified various man pages, as well as some typographical
errors in the libext2fs texinfo file.
Fixed various Debian packaging issues --- see debian/changelog for
details.
Add Indonesian and update French, Polish, Dutch, German, Swedish,
Czech, and Vietnamese Translations. (Addresses Debian Bugs: #313697,
#401092)
Programmer's Notes
------------------
Fix portability problem with the badblocks group; for systems that
don't have nanosleep(), try using usleep() instead.
The "make check" target in the e2fsck directory now sets
LD_LIBRARY_PATH before running the various e2fsck internal library
regression tests.
The crc32 regression test in the e2fsck library is now portable to
greater variety of environments, including big-endian systems and
when cross-building e2fsprogs for embedded systems. (Addresses
Sourceforge Bug: #2019287)
The ext2fs_extent_set_bmap() had some bugs when setting the first
block in a file, or when replacing a single block extent. Those cases
fortunately were came up relatively rarely when e2fsck was checking
files, but caused some problems when resize2fs was shrinking
extent-based files.
Fix a potential core-dumping bug in libe2p's iterate_on_dir()
function.
Various ext2fs library functions --- ext2fs_block_iterate2(),
ext2fs_initialize() and ext2fs_extent_open() --- now correctly free
allocated memory to avoid memory leaks in all of their error return
paths.
Ext2ed was failing to build because masix support had been removed in
the rest of e2fsprogs, so ext2ed no longer has masix support, either.
The configure script now respects the LDFLAGS environment variable if
it is set when configure is called. (Addresses Sourceforge Feature
Request: #1937287)
Libuuid is now more portable to the Windows platform. (Addresses
Sourceforge Feature Request: #1937287)
The configure script now uses AC_MSG_{RESULT,WARN,ERROR} instead of
bare echo commands so that configure flags such as --quiet work
correctly. (Addresses Sourceforge Patches: #2058794)
A few uses of sprintf have been removed from the ext2fs library to
make life easier for bootloaders with a limited libc environment.
(Addresses Sourceforge Bug: #2049120)
The ext2fs_read_inode() checks the validity of the inode number passed
to it earlier, to avoid doing some needless work when it would fail
anyway.
The ext2fs_open() checks the validity of the blocksize parameter
passed to it earlier, to avoid doing some needless work when it would
fail anyway.
Disable a very annoying automatic "%.sh -> %" GNU make rule in the
top-level Makefile. That automatic rule is used to better support
SCCS, but it caused problems for a particular niche distribution which
likes to use configure.sh files to store the configure options used to
build a package. Unfortunately GNU make will use the configure.sh to
replace the configure script, resulting in a self-inflicted fork bomb
leading to an out-of-memory crash.
To support old GNU C compilers don't use C99/C++ comments, but only
K&R style comments, and don't try to use __builtin_expect if __GNUC__
is less than 3. (__builtin_expect is only supported for gcc versions
2.96 and up, and it's tricky to check for gcc 2.95 vs gcc 2.96; since
this is an optimization, we only try to use __builtin_expect for gcc 3
and up.)
In e2fsck's crc routines, make sure we use WORDS_BIGENDIAN instead of
__LITTLE_ENDIAN, which are only defined by glibc's header files and
hence isn't portable.
For the convenience for some distributions that need a static tune2fs,
the Makefile for misc/ now has a tune2fs.static target.
The ext2fs_block_iterate2() function now supports BLOCK_FLAG_APPEND
for extent-based files
The ext2fs_bmap() function now supports BMAP_ALLOC for extent-based
files.
All source files no longer have any trailing white space.
The io_channel_read_blk64() and io_channel_write_blk64() functions are
now functions instead of C preprocessor macros to provide better
forward compatibility.
The e2fpsrogs translation template now expands the @x abbreviation.
Various namespace leakages in libblkid, libe2p, and libext2fs have
been fixed.
Fix a parallel build problem in e2fsprogs.
E2fsprogs is now more portable to Solaris.
* blkid no longer assumes that the TIOCGSIZE and TIOCGWINSZ ioctl's
are always present.
* Scripts do not assume that /bin/true is always in /bin
* Don't use __FUNCTION__ since Solaris's C99 doesn't support it.
* Flush stdio handles before calling setbuf(), since Solaris will
discard any pending output to the stream.
* Define _XOPEN_SOURCE to 600 since Solaris's header files are very
picky about which C compiler can be used for SUSv3 conformance.
Use of C99 is not compatible with SUSv2 (_XOPEN_SOURCE=500),
and C89 is not compatible with SUSv3 (_XOPEN_SOURCE=600).
Since we need some SUSv3 functions, consistently use SUSv3 so
that e2fsprogs will build on Solaris using c99.
* Solaris C99 does not support varargs C preprocessor macros
* Solaris header files pollute the C namespace if in/netinet.h
is included, which conflicts with e2fsprogs' use of the kmem_cache_t
typedef.
* Solaris ships with a pathetically ancient shell in /bin/sh so we
avoid the use of various more advanced shell constructs such as $().
The "make rpm" command will now take some extra configure options from
the build environment without needing to patch the source tree.
The ext2fs_add_dir_block() function will now grow the dblist more
aggressively as an optimization to avoid copying the array too often.
The e2fsck_write_bitmaps() will write the block and inode bitmaps
together instead of in two passes.
E2fsprogs 1.41 (July 10, 2008)
==============================
Add support for ext4 filesystem features, in particular extents,
uninit_bg, flex_bg, huge_file, and dir_nlink features. Also add
support for checking journal checksums. Debugfs will print new
superblock and inode fields that were defined for ext4. For example,
the nanosecond and i_version fields of an inode, and the
s_min_extra_isize and s_wanted_extra_isize fields from the superblock.
Note: Resize2fs doesn't currently support the combination of flex_bg
and !resize_inode. (Addresses Debian Bug: #388452, #425477)
Tune2fs can support migrating a filesystem from using 128 byte inodes
to 256 byte inodes, so it can take advantage of the full features of
ext4.
Add support for "undo" support. E2fsck and mke2fs can optionally
record an undo log which can replayed by the program e2undo.
E2fsck could damage a filesystem by trying to relocate inode tables
due to corrupted block group descriptors, where the attempted inode
table relocation would do far more harm than good. E2fsck has been
fixed to detect this these sorts of corrupted block group descriptors
much earlier in e2fsck processing, so it can try to use the backup
superblock and block group descriptors first. This should be a much
better strategy for recovering these types of corrupted filesystems.
(Addresses Sourceforge Bug: #1840291)
E2fsck will display a more understandable message when the last check
field in the superblock is in the future. (Addresses Debian Bug:
#446005).
E2fsck now performs more extensive and careful checks of extended
attributes stored in the inode.
Enhance mke2fs to print a more explanatory error message when
ext2fs_get_device_size() returns EFBIG. (Addresses Debian Bug:
#488663)
Fix mke2fs to use a default block size of 4k when formatting an
external journal device. This is done by using a fixed filesystem
type list that consists only of the single filesystem type "journal"
when looking up configuration keys in /etc/mke2fs.conf. (Addresses
Debian Bug: #488663)
Speed up how mke2fs writes the journal data blocks by writing the disk
blocks in larger chunks.
Fix blkid handling of stale devices. Fix a bug which could cause a
core dump while garbage collecting the blkid cache, and assure that
blkid_find_dev_with_tag() never returns a non-existent device. Also,
if a filesystem is found at a new /dev location, eliminate any
duplicate stale entries which can not be verified.
(Addresses Debian Bugs: #487758, #487783)
Add more paranoid checks for LVM volumes and swap partitions in
blkid's probe function, to reduce the chances of false positives.
The mke2fs program now has 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. The filesystem type information and type
information is used to extract configuration parameters from the
/etc/mke2fs.conf file.
The mke2fs program will no longer complain and request the -f option
when the user tries to create a filesystem with greater than 2**31
blocks.
When creating a filesystem for the GNU Hurd use a fs-type of Hurd and
adjust the mke2fs.conf file so filesystems for the Hurd are created
with a blocksize of 4096 and inode size of 128, which is all it knows
how to handle. (Addresses Debian Bug: #471977)
Mke2fs will always make sure that lost+found always has at least 2
blocks, even for filesystems with very large blocksizes (i.e., 64kb).
Resize2fs will now print the minimum needed filesystem size if given
the -P option, and will resize the filesystem to the smallest possible
size if given the -M option.
Fix resize2fs to clean up the resize_inode if all of the reserved gdt
blocks are consumed during an off-line resize.
The "ls" command in debugfs now supports the -p option, which causes
it to quote the filenames so that spaces or tabs in directory entries
are easily visible. (Addresses Red Hat Bugzilla: #149480; Addresses
Sourceforge Feature Request: #1201667)
Fix a potential off-by-one buffer overflow in the fs_device_name in an
e2image file.
The chattr program will return a non-zero exit code in case of
failures, and error messages can be suppressed with the -f option.
(Addresses Red Hat Bugzilla: #180596)
Fix a bug in badblocks which caused it to overrun an array and likely
crash if more than 8 test patterns are specified using the -t option.
(Addresses Debian Bug: #487298)
Add support to badblocks to limit how quickly it reads from the disk
drive (so it can be used for background scrubbing), and so it will
abort after finding a given number of errors.
Remove support for the legacy big-endian filesystem format which only
existed on extremely long-dead PowerPC kernels almost a decade ago.
Remove MASIX support from e2fsprogs.
Add I/O statistics reporting to e2fsck.
Update Vietnamese, Polish, French, Spanish, German, Catalan, Dutch,
Czech translations.
Fixed various Debian packaging issues --- see debian/changelog for
details. (Addresses Debian Bugs: #487443, #487675, #490003)
Fixed spelling mistakes, typos, and otherwise clarified man pages.
(Addresses Debian Bugs: #393313, #487849, #440983, #440981)
Programmer's Notes
------------------
Factor out bitmap code in preparation for adding 64-bit new-style
bitmaps.
Fix gcc -Wall warnings
Fix the pkg-config files so that private libraries are specified in
"Libs.private:".
Fix the libext2fs.texinfo manual so it builds with modern versions of
texinfo.
Silence the makefile from showing the awk command used to build the
Clean up the badblocks group so to make it more portable and robust.
Avoid using predictable filenames in /tmp in blkid's regression test
suite. Also remove bashism's in the regression test script.
If the configure script is given --with-diet-libc, don't use thread
local storage, since diet libc doesn't support TLS. (Addresses
Sourceforge Bug: #2000654)
Fix the blkid regression test suite to tolerate older versions of
mkswap that don't support the -U option.
A few library routines have been converted to support 64-bit block
numbers; in particular, the I/O manager functions, the test_io,
inode_io, and unix_io managers have all be converted to support 64-bit
operation.
Debugfs can now be extended for use by test programs. See
lib/ext2fs/extents.c for an example for how it can be used. The test
program links against the debugfs object files, and provides
additional commands by defining an auxiliary libss command table.
The lazy_bg filesystem feature, which was only used by developer's
testing, has been removed since it has been largely supplanted by
uninit_bg. This also simplifies the code.