This reverts commit 1f3f267021.
Symlink will remain in place to ensure the reverted version is always
shipped to our users, including those that have installed Core Update
174 (testing).
Fixes: #13073
Reported-by: Arne Fitzenreiter <arne.fitzenreiter@ipfire.org>
Signed-off-by: Peter Müller <peter.mueller@ipfire.org>
- Update from version 1.46.5 to 1.47.0
- Update of rootfile not required
- Changelog
E2fsprogs 1.47.0 (February 5, 2023)
Updates/Fixes since v1.46.6:
UI and Features
Add support for the orphan_file feature, which speeds up workloads that
are deleting or truncating a large number files in parallel. This
compat feature was first supported in the v5.15 Linux kernel.
The mke2fs program (via the mke2fs.conf file) now enables the
metadata_csum_seed and orphan_file features by default. The
metadata_csum_seed feature is an incompat feature which is first
supported in the Linux kernel starting in the 4.4 kernel and e2fsprogs
1.43.
Mke2fs now supports the extended option "assume_storage_prezeroed" which
causes mke2fs to skip zeroing the journal and inode tables and to mark
the inode tables as zeroed.
Add support to tune2fs and e2label to set the label and UUID for a
mounted file system using a ioctl, which is more reliable than modifying
the superblock via writing to the block device. The kernel support for
setting the label landed in v5.17, while the support for adding the UUID
landed in v6.0. If the ioctls are not supported, tune2fs and e2label
will fall back old strategy of directly modifying the superblock.
Allow tune2fs to disable the casefold feature after scanning all of the
directories do not have the Casefold flag set.
Fixes
Fix a potential unbalanced mutex unlock when there is a short read while
using the bounce buffer when using direct I/O.
Performance, Internal Implementation, Development Support etc.
Fix various Coverity and compiler warnings.
Add the new function ext2fs_xattrs_read_inode() which takes an in-memory
inode to avoid needing to reread an inode that was already read into
memory.
Teach debugfs logdump command the -n option which forces printing a
specified number of transactions, even when a block missing a magic
number would have stopped the logdump. (This is for debugging
journalling problems.)
E2fsprogs 1.46.6 (February 1, 2023)
Updates/Fixes since v1.46.5:
UI and Features
Debugfs's ncheck command now allows the inode number to be surrounded by
angle brackets, to be consistent with other debugfs commands.
Debugfs no longer prints a scary message when debugfs -c (which enables
"catastrophic mode") is used. This was intended to allow debugfs to
operate on very badly corrupted file systems, but it is now sometimes
used to suppress reading the block and inode bitmaps when they are not
needed.
Resize2fs will round down the requested new file system size to the
nearest cluster boundary when resizing bigalloc file systems.
Improve error messages issued by badblocks.
Fuse2fs now supports an offset=<bytes> option which allows operating on
a file system image which is located starting at the specified offset
from the beginning of the image.
Fixes
Pre-v6.2 Linux kernels had long-standing bug in how the extended
attribute hash was calculated when there were non-ASCII characters in
the xattr name, when the hash would be different depending on whether
the C 'char' type was signed or unsigned. To address this bug, starting
with e2fsprogs 1.46.6+ and Linux 6.2+, we will accept either the signed
or unsigned hash variant, but only set the unsigned hash variant. Since
extended attribute names are in practice composed of ASCII characters,
other than various tests (such as generic/454), most users will
hopefully not notice this change.
Avoid triggering udev in dumpe2fs and "resize2fs -P" for file systems
with MMP enabled by opening the device read-only when reading the MMP
block.
Fix MMP handling so it can notice when another writer has modify the MMP
block out from under it when stopping a MMP sessions.
Fix tune2fs so it will detect another device stealing the MMP sessions
while rewriting metadata checksums.
E2fsck will now check to make sure the journal inode does not have the
encrypt flag set.
Fix a deadlock bug in e2fsck's error handler when there are errors
trying to write to the file system.
Fix a bug where e2fsck could fail when specifying an undo file and an
explicit superblock number.
Fix e2image so it won't potentially loop forever for certain invalid
file systems.
Fix resize2fs to honor the E2FSPROGS_FAKE_TIME environment variable.
This allows embedded system builders who use resize2fs as part of their
image build process to create reproducible images.
Fix tune2fs to avoid a crash if the journal replay fails and to make
sure its exit status is non-zero if there is some failure.
Fix tune2fs, fuse2fs, and debugsfs to update j_tail_sequence when
replaying the journal.
Add additional bullet-proofing for very badly corrupted file systems.
Try avoid UBSAN warnings, null pointer derferences, and other memory
bugs. (Addresses CVE-2022-1304)
Don't fail when the source directory for mke2fs -d doesn't support
extended attributese.
Check for and handle malloc() failures when computing the log filename
in e2fsck and in the libss library.
Fix tune2fs and e2fsck to accept pathames which include '=' characters.
Previously arguments to tune2fs and e2fsck which included '=' characters
are presumed to be blkid specifiers such as UUID=xxx or LABEL=yyy. If a
specifier is both a valid pathname name and blkid tag name specifier,
priority is given to a blkid resolved pathname.
Improve tune2fs's error messages.
Fix a bug in tune2fs which could cause it to crash if device goes
off-line just as it being opened.
Fix the fsck driver so if it is interrupted while running fsck -N it
doesn't end up kllling all processes on the system.
Fix a crash in badblocks when the user specifies an overly large
number of blocks tested at a time in read/write or nondestructive
mode.
Update and clarify's chattr's man page and usage message. Fix spelling
typo's in a variety of different man pages and comments.
Performance, Internal Implementation, Development Support etc.
Update to autoconf 2.71.
Update flags used to create shared library on Darwin/MacOS.
Speed up e2fsck's clonning of multiply-claimed blocks so it is
substantially faster on very large file systems.
Add tests/fuzz directory with fuzzers from oss-fuzz.
Add a Github Actions configuration file so that Github will run CI tests
on Linux, Windows and MacOS on a push to the e2fsprogs github repo.
Make the mtab parsing in ext2fs_check_mount_point() more careful so it
won't get confused when a block device shows up in the mnt_name field
for a virtual file system.
Fix the libss's Makefile to create the man page directory before trying
to install its man page.
Fix various Coverity and compiler warnings.
Make tests more portable on various different OS's and system
configurations (e.g., with SELinux enabled, MacOS, and Windows)
Use mallinfo2() instead of mallinfo() where avilable, since mallinfo()
is deprecated on newer glibc versions.
E2fsck will no longer do a full scan of disconnected directory when
trying to print the parent directory, which is pointless and can slow
down e2fsck if there are a large number of disconnected directories.
Debugfs will now print the extended attribute's e_hash field.
Fix the setup-schroot script to work on non-Linux platforms.
Fix ext2fs_compare_generic_bmap() so it correctly compares all of the
bits in the bitmap, and so that it works correctly when comparing a
bitarray bitmap with a rbtree-based bitmap. (Fortunately, none of the
programs in e2fsprogs uses bitmap comparison functions.)
Fix memory leaks on error paths.
Add support for the configure option --enable-largefile so that
e2fsprogs can utilize largefile support for the MUSL C library.
Add an note that the dict library code has been modified, as required by
the Kazlib license.
Synchronized changes from Android's AOSP e2fsprogs tree.
Updated config.guess and config.sub with newer versions from the FSF.
Add Friulian translation.
Update Chinese, Czech, Dutch, French, German, Malay, Polish, Serbian,
Spanish, Swedish, and Ukrainian translations.
Signed-off-by: Adolf Belka <adolf.belka@ipfire.org>
Historically, the MD5 checksums in our LFS files serve as a protection
against broken downloads, or accidentally corrupted source files.
While the sources are nowadays downloaded via HTTPS, it make sense to
beef up integrity protection for them, since transparently intercepting
TLS is believed to be feasible for more powerful actors, and the state
of the public PKI ecosystem is clearly not helping.
Therefore, this patch switches from MD5 to BLAKE2, updating all LFS
files as well as make.sh to deal with this checksum algorithm. BLAKE2 is
notably faster (and more secure) than SHA2, so the performance penalty
introduced by this patch is negligible, if noticeable at all.
In preparation of this patch, the toolchain files currently used have
been supplied with BLAKE2 checksums as well on
https://source.ipfire.org/.
Cc: Michael Tremer <michael.tremer@ipfire.org>
Signed-off-by: Peter Müller <peter.mueller@ipfire.org>
Acked-by: Michael Tremer <michael.tremeripfire.org>
- Update from 1.46.3 to 1.46.5
- Update of rootfile not required
- Changelog
1.46.5 (December 30, 2021)
Updates/Fixes since v1.46.4:
UI and Features
When resizing a file system and the inode count exceeds the 2**32
maximum, if resize2fs can successfully perform the resize by dropping
the last block group, resize2fs will do that in order to allow the file
system grow operation to succeed. For example, using the default inode
ratio size of 16k, this will allow a successful resize to 64TB - 128MB
when the storage device is 64TB.
Fixes
Avoid a potential infinite loop in resize2fs -P when the file system is
corrupted (introduced in e2fsprogs 1.45.5). (Addresses github issue
https://github.com/tytso/e2fsprogs/issues/94)
E2fsck now updates the bg_checksum after fixing problems in the block
group descriptor, which eliminates some unnecessary messages printed or
asked of the system administrator.
Fixed some potential deadlock problems in the unix_io handler in the case
of I/O errors. The fix should also improve the performance of parallel
bitmap loading.
Fixed e2fsck's fast commit handling which could result it in crashing
when trying to merge extents when there were none available to be
merged.
Fix e2fsck's support of quota limit data, which could sometimes get
dropped when the quota data needs to be regenerated, or when processing
the orphan list.
Fix tune2fs to correctly transfer the quota limits when converting quota
files to the internal quota inodes. Also add support for tune2fs to
properly handle the older version 0 quota files.
Fix debugfs's get_quota and list_quota commands so that the header of
the report printed by these commands correctly reflect that the units of
used space is in bytes instead of blocks.
Performance, Internal Implementation, Development Support etc.
Add some additional packages to the setup-schroot script to account for
the fact that the script can be run on older Debian distributions and so
the build dependencies might omit some packages needed to build
e2fsprogs on unstable version of Debian.
Reduce resize2fs's CPU overhead when counting the number of blocks in
use which can reduce the wall clock time for very large file systems
by substantial amount.
Teach libuuid to use getrandom() or getentropy() if available in favor
of reading from /dev/[u]random.
Teach libss to use libreadline.so.8 if it is available.
Update some test expect files to fix some regression tests that were
broken in e2fsprogs 1.46.4.
If the PRINT_FAILED environment variable is set, failed tests will
display the diff output to make it easier to debug test failures on
autobuilders.
Fix various compiler warnings.
Update tst_getsize to use ext2fs_get_size2() to support testing devices
which are larger than 2**32 sectors.
Fixed spelling mistakes in the mke2fs.conf man page.
Update Chinese, Malay, Serbian, Spanish, Swedish, and Ukrainian
translations.
1.46.4 (August 18, 2021)
Updates/Fixes since v1.46.3:
UI and Features
The defaults for mke2fs now call for 256 byte inodes for all file
systems (with the exception of file systems for the GNU Hurd, which only
supports 128 byte inodes). Creating non-Hurd file systems with 128 byte
inodes will trigger a warning message to make sure users are aware of
the potential problems of using small/legacy inode sizes.
The bigalloc feature is now considered supported if the cluster size no
more than 16 times the block size. So the mke2fs program has been
changes to only warn if the cluster size is larger than that.
Fixes
E2fsck now checks to make sure directory entries do not reference
internal quota inodes.
E2image now includes the quota inodes when creating file system image,
since they are part of the file system metadata.
E2fsck now properly accounts the quota usage of the project quota file.
Fix a regression introduced in 1.64.3 where attempting to create a file
system image using mke2fs into a non-existent file would fail.
(Addresses Debian Bug: #992094)
Fix mke2fs to correctly create Posix ACL's on big-endian systems when
copying files from a directory hierarchy.
Updated and clarified the resize2fs man page. (Addresses Debian Bug:
#979411)
Performance, Internal Implementation, Development Support etc.
Improve various regression tests to be more portable and to reflect the
new default inode size of 256 byte inodes, even for small file systems.
Fixed a GNU Hurd portability problem which was causing tests to fail.
Fixed a test failure in f_baddotdir on big-endian systems. This wasn't
necessarily a bug per se in e2fsck, but rather e2fsck having different
behaviour on big-endian systems. (Addresses Debian Bug: #991922)
Use WantedBy=multi-user.target in e2scrub_reap.service. (Addresses
Debian Bug: #991349)
Synchronize e2fsck/recovery.c with the kernel's fs/jbd2/recovery.c
Fix various Coverity and compiler warnings.
Fix various error pathes to make sure we don't leak resources or
potentially use or try to free uninitialized pointers.
Added a setup-schroot command for use on Debian porter boxes.
Updated config.guess and config.sub with newer versions from the FSF.
Update Czech, Dutch, French, Polish, Portuguese, and Swedish translations.
Signed-off-by: Adolf Belka <adolf.belka@ipfire.org>
Reviewed-by: Michael Tremer <michael.tremer@ipfire.org>
Most of these files still used old dates and/or domain names for contact
mail addresses. This is now replaced by an up-to-date copyright line.
Just some housekeeping... :-)
Signed-off-by: Peter Müller <peter.mueller@link38.eu>
Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>
Zurueck zu ReiserFS, da wir mit Reiser4 noch ein paar Probleme haben und wir die Prioritaeten noch auf andere Sachen legen muessen. :'(
git-svn-id: http://svn.ipfire.org/svn/ipfire/trunk@445 ea5c0bd1-69bd-2848-81d8-4f18e57aeed8
