On Wed, Feb 28, 2018 at 03:03:09AM -0500, Theodore Ts'o wrote:
> Since e2fsprogs 1.44.0 will have a lot of new features over the 1.43.x
> releases, I've released e2fsprogs 1.44.0-rc1 on kernel.org and
Uhoh, I wasn't expecting this out so soon. I was going to send you a
second round of e2scrub patches, and a few others to enable fancy debug
tools like {ub,addr}san, LTO, and retpolines. Is it too late for that?
TBH I've been meaning to ask about whether it makes sense to build
e2fsprogs (or at least e2fsck) with retpolines but haven't had a chance
to converse about it for a couple of weeks.
> sourceforge, and of course it's tagged on the git trees in
> git.kernel.org, github, and sourceforge. The compressed tar files can
> be found in:
>
> https://www.kernel.org/pub/linux/kernel/people/tytso/e2fsprogs-test/v1.44.0-rc1
>
> and
>
> https://sourceforge.net/projects/e2fsprogs/files/e2fsprogs-TEST/1.44.0-rc1
>
> The release notes for 1.44.0 can be found below.
>
> - Ted
>
> E2fsprogs 1.44.0 (February 27, 2018)
> ====================================
>
> Updates/Fixes since v1.43.9:
>
> UI and Features
> ---------------
>
> E2fsprogs now supports the large_dir (INCOMPAT_LARGEDIR) feature. This
> feature allows larger directories to be created, both with directory
> sizes over 2GB and and a maximum htree depth of 3 instead of the
> current limit of 2. These features are needed in order to exceed
> the currently limit of approximately 10M entries in a single
> directory for 4KB blocksize (~100k for 1KB).
>
> E2fsprogs now supports the ea_inode (INCOMPAT_EA_INODE) feature, which
> stores large extended attributes in an external inode instead of a
> single data block. This is backwards compatible with a Lustre file
> system feature, but it has been extended to support deduplicating
> extended attribute values which are used in multiple inodes.
>
> The metadata_csum feature (RO_COMPAT_METADATA_CSUM) is now enabled by
> default by mke2fs.
Yay. :)
--D
> The resize2fs program will no longer complain about resizing bigalloc
> file systems as being dangerous when doing an on-line resize (since this
> is purely a kernel feature).
>
> Debugfs's ls command will now print the high bits of the inode's mode
> bits.
>
> Fixes
> -----
>
> Fixed some potential buffer overrun bugs in the blkid library and in the
> fsck program.
>
> There were some corner cases which tune2fs wouldn't correctly handle
> when operating on a file system which was not cleanly mounted and thus
> required running the journal because it could be modified. Tune2fs will
> now run the journal before trying to change the superblock, and if the
> file system needs to be checked after the journal has been replayed, it
> will require it before proceeding. Also if the file system is busy (for
> example, because it is in use by Lustre), but is not present in the
> /proc/mounts or /etc/mtab, tune2fs will not try to replay the journal,
> since this could do real damage if it the file system is being actively
> modified by the kernel.
>
> The libss library (used by debugfs) now supports the version 7 of the
> libreadline shared library.
>
> The e2freefrag program will use the GETFSMAP ioctl if it is available
> and the file system is mounted, so that the free space statistics will
> be more accurate.
>
>
> Performance, Internal Implementation, Development Support etc.
> --------------------------------------------------------------
>
> Mke2fs now uses io_channel_flush() instead of sync() for the sync_kludge
> feature.
>
> E2fsprogs now uses i_size as the primary way of determining whether a
> symlink is a fast symlink.
>
> Fix various compiler and UBSAN warnings.
>
> Long running test are now skipped with "make check" and only run when
> "make fullcheck". The test runner will warn when potentially long
> running tests are being run via "make check".
>
> E2fsck is now much faster for bigalloc file systems when scanning
> extents for bigalloc file systems. (Addresses Google Bug #36886699)
