dumpe2fs, debugfs, e2image: Add support for orphan file

Print inode number of orphan file in outputs, dump e2image file to
filesystem image.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

mke2fs: Add orphan_file feature into mke2fs.conf

Enable orphan_file feature by default in larger filesystems. Since the
feature is COMPAT, older kernels will just ignore it and happily work
with the filesystem as well.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

tune2fs: Add support for orphan_file feature

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

e2fsck: Add support for handling orphan file

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

mke2fs: Add support for orphan_file feature

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

libext2fs: Support for orphan file feature

Add support for creating and deleting orphan file and a couple of
utility functions that will be used in other tools.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

e2image: Dump quota files

Dump quota files to resulting filesystem image. They are fs metadata.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

quota: Do not account space used by project quota file to quota

Project quota files have high inode numbers but are not accounted in
quota usage. Do not track them when computing quota usage.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

ss_create_invocation: fix potential unititalized reference in error path

Fixes: eccdde1ff381 ("ss_create_invocation: fix error handling when ...")
Addresses-Coverity-Bug: 1489771
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

lsattr: check whether path is NULL in lsattr_dir_proc()

In lsattr_dir_proc(), if malloc() return NULL, it will cause
a segmentation fault problem.

Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

hashmap: change return value type of ext2fs_hashmap_add()

In ext2fs_hashmap_add(), new entry is allocated by calling
malloc(). If malloc() return NULL, it will cause a
segmentation fault problem.

Here, we change return value type of ext2fs_hashmap_add()
from void to int. If allocating new entry fails, we will
return -1, and the callers should also verify the return
value of ext2fs_hashmap_add().

Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

ss_create_invocation: fix error handling when memory allocation fails

In ss_create_invocation(), it is necessary to check whether
returned by malloc is a null pointer.

Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

ss_add_info_dir: fix error handling when memory allocation fails

If the realloc() and malloc() calls fail, avoid a memory leak as well
as a potential seg fault.

[ Fix error code setting to avoid depending on malloc() and realloc()
  setting errno. -- TYT ]

Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Reviewed-by: Wu Bo <wubo40@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

mke2fs: fix creating a file system image w/o a pre-existing file

The mke2fs program should allow creating a file system image when an
explicit file system size is specified, even if the file doesn't yet
exist.  By deferring the call to check_plausible() in commit
942b00cb9d2f ("mke2fs: do not warn about a pre-existing partition
table when using a non-zero offset") this behaviour was broken.

Fix this regression by explicitly creating the file if the file system
size is specified.

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

libext2fs: improve error handling in POSIX ACL conversions

When encoding a POSIX ACL to the EXT4 ACL format, if an unknown tag
is encountered, that entry is silently ignored. It would be better
to return an error to inform the user that the ACL is incompatible.

Also fix the mismatched indentation in the opposite function.

Signed-off-by: Samuel Holland <samuel@sholland.org>

setup-schroot: install the acl and libreadline-dev packages

The acl package is needed to run the m_rootdir_acl test.  The
libreadline-dev package will drag in the shared library package for
libreadline (libreadlineN) so that the developer running debugfs will
be able to use line editing.

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

tests: skip m_rootdir_acl on GNU Hurd

The GNU Hurd doesn't support Posix ACL's, so even if the acl package
is installed, the setfattr command will fail.  So just skip the test
on Hurd.

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

libext2fs: fix translation of Posix ACL's on big-endian systems

The ACL returned by the kernel in lgetxattr(2) is returned in Little
Endian, even on Big Endian systems.  Fix the functions
convert_posix_acl_to_disk_buffer() and convert_disk_buffer_to_posix_acl()
to work correctly on Big Endian systems.  This fixes a failure of
the test m_rootdir_acl.

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

contrib: add setup-schroot command for use on Debian porter boxes

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

tests: add description for j_recover_fast_commit

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

tests: force test file systems to be built for the Linux OS

This is needed to fix a large number of test failures on GNU Hurd.

Also skip a number of tests that require creating very large test file
systems,since Hurd does not support files greater than 4GB.

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

tests: try using truncate command before falling back to using dd

On the GNU Hurd, dd seems to be buggy and hangs if seeking beyond 4G.
Fortunately the GNU Hurd does have the truncate command, which does
work correctly.  So try using the truncate command first, and fall
back to using dd only if the truncate command doesn't work.

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

libsupport: fix sort_r.h to work on the GNU Hurd

On the GNU Hurd both __MACH__ and __GNU__ are defined.  So rearrange
the #ifdef to prioritize checking for GLIBC compatibility over BSD
compatibility.

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

e2fsck: fix f_baddotdir failure on big-endian systems

Commit 63f44aafb1f2 ("e2fsck: fix ".." more gracefully if possible")
changed the check_dot() function to try to avoid resetting the '..'
entry when the '.' entry is too large..  But if we do that, then on
big-endian systems, we need to try byte swapping the rest of the
directory entries, or else the f_baddotdir test will fail on
big-endian systems.

Also add a check to avoid UBSAN warning when there is not enough space
at the end of the directory block for a directory entry, and so we can
potentially overflow some pointer arithmetic when trying to byte swap
the remainder of the (negative) space in the directory block.

Fixes: 63f44aafb1f2 ("e2fsck: fix ".." more gracefully if possible")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

Update release notes, etc., for the 1.46.3 release

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

AOSP: Move system_shared_libs into target.bionic clause

Use target.bionic.system_shared_libs when it is used to limit the
default shared libraries (as opposed to remove them completely).
This avoids attempting to add a host dependency on libc when
system_shared_libs is modified to apply to all variants.

Also remove system_shared_libs from static binaries where it has
no effect, and consolidate it into e2fsprogs-defaults.

Bug: 193559105
Test: m checkbuild
Change-Id: I2d447b006afc783f4acd6c1acd93f338a68a01ed
From AOSP commit: 48fa7248112701c30d3cabfb8d3360b2408d6491

AOSP: Use -Wno-pointer-arith in Android build

Some "arithmetic on a void pointer is a GNU extension" warnings were
introduced into upstream e2fsprogs, but they are of questionable value,
so disable them for now.

Change-Id: I19e99382e9434828927c1b8287b91f3333110151
From AOSP commit: f203c38fd43dc384cedfcaeb155771a15906022f

AOSP: Add a new upstream source file to lib/e2p/Android.bp

Change-Id: Ibc305759a9046996a397d652d799e930629f9f71
From AOSP commit: a02aa2b8e145449711e0dcdae5573fbaa888bf33

AOSP: [LSC] Add LOCAL_LICENSE_KINDS to external/e2fsprogs

Added SPDX-license-identifier-0BSD SPDX-license-identifier-Apache-2.0
    SPDX-license-identifier-BSD SPDX-license-identifier-GPL
    SPDX-license-identifier-GPL-2.0 SPDX-license-identifier-LGPL
    SPDX-license-identifier-LGPL-2.0 SPDX-license-identifier-LGPL-2.1
    SPDX-license-identifier-LGPL-3.0 SPDX-license-identifier-MIT
    legacy_notice legacy_unencumbered
to:
  Android.bp

Added SPDX-license-identifier-0BSD SPDX-license-identifier-BSD
    SPDX-license-identifier-GPL SPDX-license-identifier-GPL-2.0
    SPDX-license-identifier-LGPL SPDX-license-identifier-LGPL-2.1
    SPDX-license-identifier-LGPL-3.0 SPDX-license-identifier-MIT
    legacy_unencumbered
to:
  lib/Android.bp

Added SPDX-license-identifier-0BSD SPDX-license-identifier-BSD
    SPDX-license-identifier-GPL-2.0 SPDX-license-identifier-MIT
to:
  lib/et/Android.bp

Added SPDX-license-identifier-0BSD SPDX-license-identifier-MIT
to:
  lib/ss/Android.bp

Added SPDX-license-identifier-Apache-2.0
to:
  contrib/android/Android.bp

Added SPDX-license-identifier-Apache-2.0 SPDX-license-identifier-GPL
to:
  contrib/Android.bp

Added SPDX-license-identifier-BSD
to:
  lib/uuid/Android.bp

Added SPDX-license-identifier-GPL
to:
  resize/Android.bp

Added SPDX-license-identifier-GPL SPDX-license-identifier-GPL-2.0
to:
  debugfs/Android.bp

Added SPDX-license-identifier-GPL SPDX-license-identifier-GPL-2.0
    SPDX-license-identifier-LGPL
to:
  e2fsck/Android.bp

Added SPDX-license-identifier-GPL SPDX-license-identifier-GPL-2.0
    SPDX-license-identifier-LGPL SPDX-license-identifier-LGPL-2.1
    SPDX-license-identifier-LGPL-3.0 legacy_unencumbered
to:
  lib/ext2fs/Android.bp

Added SPDX-license-identifier-GPL SPDX-license-identifier-LGPL
to:
  lib/e2p/Android.bp

Added SPDX-license-identifier-GPL SPDX-license-identifier-LGPL
    SPDX-license-identifier-LGPL-2.1 SPDX-license-identifier-LGPL-3.0
to:
  lib/blkid/Android.bp
  misc/Android.bp

Added SPDX-license-identifier-GPL SPDX-license-identifier-MIT
to:
  lib/support/Android.bp

Bug: 68860345
Bug: 151177513
Bug: 151953481

Test: m all

Exempt-From-Owner-Approval: janitorial work
Change-Id: I239a04a83f12ba051be911d18f6df4ae77fb3368
From AOSP commit: e86522c572b5715b85889cf8ca1c52a5cc350ca7

po: update sr.po (from translationproject.org)

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

po: update nl.po (from translationproject.org)

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

po: update ms.po (from translationproject.org)

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

Fix miscellaneous compiler warnings using "make gcc-wall"

Address a number of signed vs. unsigned comparison errors, unused
function parameters, casts which drop const, etc.

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

libext2fs: avoid unnecessary stat(2) calls on mountpoints

If the device name in the mtab or /proc/mounts file does not match
with the device passed into ext2fs_check_if_mounted() or
ext2fs_check_mount_point(), skip the stat(2) call on the mountpoint,
since we never use the results of the stat(2) in that case.  Not only
does this provide a slight performance win, but it the stat calls on
the mountpoints could potentially trigger some SELinux denials that
could stress some sysadmins out.

Google-Bug-Id:  193137337
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

debian/changelog: fix missing Closed: annotation

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

AOSP: e2fsdroid: Don't over-reserve blocks for files in Base FS.

If a large number of blocks move from one file to another file,
e2fsdroid will inadvertently reserve the moved blocks for the source
file. If the larger file is visited first, it can fail to acquire blocks
because they're reserved by the smaller file.

This patch only reserves the first N blocks of a file in Base FS, where N
is large enough to satisfy the |st_size| property from lstat(2). We only
consider "owned" blocks rather than deduplicated blocks, in case the new
file cannot be deduplicated.

Google-Bug-Id: 145316683
Test: e2fsdroid with dynamic partitions
Change-Id: I32e255a19550d52d90342c21d7218981108a71b1
From AOSP commit: 2985b26cdbe674084fa02b211503eef9bf970023

mke2fs: do not warn about a pre-existing partition table when using a non-zero offset

The existing code attempted to avoid warning about a pre-existing file
system with a non-zero offset, but because the offset was not set at
the time of the check, this intention was not actually working.  So
this commit will suppress warnings about pre-existing a partition
table as well as pre-existing file system when there is a non-zero
offset.

Addresses-Debian-Bug: #989612
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

mke2fs: only try discarding a single block to test if discard works

Commit d2bfdc7ff15c ("Use punch hole as "discard" on regular files")
added a test to see if the storage device actually supports discard.
The intent was to try discarding the first block but since
io_channel_discard() interprets the offset and count arguments in
blocks, and not bytes, mke2fs was actually discarding the first 16
megabytes (when the block size is 4k).  This is normally not a
problem, since most file systems are larger than that, and requests to
discard beyond the end of the block device are ignored.

However, when creating a small file system as part of a image
containing multiple partitions, the initial test discard can end up
discarding data beyond the file system being created.

Addresses-Debian-Bug: #989630
Reported-by: Josh Triplett <josh@joshtriplett.org>
Fixes: d2bfdc7ff15c ("Use punch hole as "discard" on regular files")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

libext2fs: avoid forking a single thread in ext2fs_rw_bitmaps()

https://github.com/tytso/e2fsprogs/issues/67

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

libe2p: use stat to prevent calling EXT2_IOC_[GS]ETFLAGS on devices

Some devices can react badly to the EXT2_IOC_[GS]ETFLAGS ioctls, since
ioctl codes are not guaranteed to be unique across different device
drivers and file systems.

Addresses-Debian-Bug: #986332
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

ext2ed: fix potential NULL pointer dereference in dupstr()

In dupstr(), we should check return value of malloc().

Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Reviewed-by: Wu Bo <wubo40@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

lib/ss/error.c: check return value malloc in ss_name()

In ss_name(), we should check return value of malloc(),
otherwise, it may cause a segmentation fault problem.

Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

misc: fix potential segmentation fault problem in scandir()

In scandir(), temp_list[num_dent] is allocated by calling
malloc(), we should check whether malloc() returns NULL before
accessing temp_list[num_dent].

Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

argv_parse: check return value of malloc in argv_parse()

In argv_parse(), return value of malloc should be checked
whether it is NULL, otherwise, it may cause a segfault error.

Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

append_pathname: check the value returned by realloc

In append_pathname(), we need to add a new path to save the value
returned by realloc, otherwise the name->path may be NULL, causing
a segfault.

Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

zap_sector: fix memory leak

In zap_sector(), need free buf before return,
otherwise it will cause memory leak.

Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Reviewed-by: Wu Bo <wubo40@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

tdb_transaction_recover: fix memory leak

In tdb_transaction_recover(), need free data before return,
otherwise it will cause memory leak.

Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Reviewed-by: Wu Bo <wubo40@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

profile_create_node: set magic before strdup(name) to avoid memory leak

If new->magic != PROF_MAGIC_NODE, profile_free_node() don't free node.
This will cause the node to be unable to be released correctly and
a memory leak will occur.

Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com>
Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Reviewed-by: Wu Bo <wubo40@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

libext2fs: fix a -Wunused-label warning

Remove a label which isn't used.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

Fix -Wunused-variable warnings

Fix all warnings about unused variables that were introduced since
e2fsprogs v1.45.4.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

Fix -Wunused-parameter warnings

Fix all warnings about unused function parameters that were introduced
since e2fsprogs v1.45.4, by adding EXT2FS_ATTR((unused)) or removing
parameters as appropriate.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

mke2fs: use ext2fs_get_device_size2() on all platforms

Since commit e8c858047be6 ("libext2fs: fix build issue for on
Windows/Cygwin systems"), ext2fs_get_device_size2() is available in
Windows builds of libext2fs.  So there is no need for mke2fs to call
ext2fs_get_device_size() instead.

This fixes a -Wincompatible-pointer-types warning because
ext2fs_get_device_size() was being passed a 'blk64_t *', but it expected
a 'blk_t *'.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

e2fsck: sync fc_do_one_pass() changes from kernel

Sync the changes to fc_do_one_pass() from the kernel's recovery.c so
that e2fsck picks up the fixes to the jbd_debug() statements.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

libext2fs: improve jbd_debug() implementation

Make jbd_debug() do format string checking (but still get compiled away
to nothing) when --enable-jbd-debug isn't specified, similar to
commit d556435156b7 ("jbd2: avoid -Wempty-body warnings") on the kernel
side.  This should prevent --enable-jbd-debug from getting broken due to
bad jbd_debug() statements.  It also eliminates a -Wunused-variable
warning where a variable was only used in a jbd_debug() statement.

Also remove an alternative definition of jbd_debug() that was
conditional on CONFIG_JBD_DEBUG && !CONFIG_JBD_DEBUG, so was dead code.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

tune2fs: update overhead when toggling journal feature

When adding or removing journal from a filesystem, we also need to add /
remove journal blocks from overhead stored in the superblock.  Otherwise
total number of blocks in the filesystem as reported by statfs(2) need
not match reality and could lead to odd results like negative number of
used blocks reported by df(1).

Fixes: 9046b4dfd0ce ("mke2fs: set overhead in super block")
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

e2fsck: fix last mount/write time when e2fsck is forced

With commit c52d930f e2fsck is no longer able to fix bad last
mount/write time by default because it is conditioned on s_checkinterval
not being zero, which it is by default.

One place where it matters is when other e2fsprogs tools require to run
full file system check before a certain operation. If the last mount
time is for any reason in future, it will not allow it to run even if
full e2fsck is ran.

Fix it by checking the last mount/write time when the e2fsck is forced,
except for the case where we know the system clock is broken.

[ Reworked the conditionals so error messages claiming that the last
  write/mount time were corrupted wouldn't be always printed when the
  e2fsck was run with the -f option, thus causing 299 out of 372
  regression tests to fail.  -- TYT ]

Fixes: c52d930f ("e2fsck: don't check for future superblock times if checkinterval == 0")
Reported-by: Dusty Mabe <dustymabe@redhat.com>
Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

e2fsck: fix ".." more gracefully if possible

If the "." entry is corrupted, it will be reset in check_dot().
It is possible that the ".." entry can be recovered from the
directory block instead of also resetting it immediately.  If
it appears that there is a valid ".." entry in the block, allow
that to be used, and let check_dotdot() verify the dirent itself.

When resetting the "." and ".." entries, use EXT2_FT_DIR as the
file type instead of EXT2_FT_UNKNOWN for the very common case of
filesystems with the "filetype" feature, to avoid later problems
that can be easily avoided.  This can't always be done, even if
filesystems without "filetype" are totally obsolete, because many
old test images do not have this feature enabled.

Fixup affected tests using the new "repair-test" script that
updates the expect.[12] files from $test.[12].log for the given
tests and re-runs the test to ensure it now passes.

Signed-off-by: Andreas dilger <adilger@whamcloud.com>
Reviewed-by: Artem Blagodarenko <artem.blagodarenko@hpe.com>
Lustre-bug-Id: https://jira.whamcloud.com/browse/LU-14710
Change-Id: Ia5e579bcf31a9d9ee260d5640de6dbdb60514823
Reviewed-on: https://review.whamcloud.com/43858
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

libext2fs: use offsetof() from stddef.h

offsetof is a standard C feature available from stddef.h, going
back all the way to ANSI C.

Signed-off-by: Michael Forney <mforney@mforney.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

libext2fs: use statement-expression for container_of only on GNU-compatible compilers

Functionally, the statement-expression is not necessary here; it
just gives a bit of type-safety to make sure the pointer really
does have a compatible type with the specified member of the struct.

When statement expressions are not available, we can just use a
portable fallback macro that skips this member type check.

Signed-off-by: Michael Forney <mforney@mforney.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

e2image: fix overflow in l2 table processing

For a large partition during e2image capture process
it is possible to overflow offset at multiply operation.
This leads to the situation when data is written to the
position at the start of the image instead of the image end.

Let's use the right cast to avoid integer overflow.

Signed-off-by: Alexey Lyashkov <c17817@cray.com>
Signed-off-by: Artem Blagodarenko <c17828@cray.com>
HPE-bug-id: LUS-9368
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

tests: add test of Direct I/O support in the Unix I/O manager

This test requires root, and only works on Linux systems.

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

libext2fs: fix Direct I/O support in the Unix I/O manager

Commit d557b9659ba9 ("libext2fs: fix potential races in unix_io")
contained some additional changes that were supposed to clarify the
management of the length calculations when reading and writing
non-page aligned data using Direct I/O; unfortunately, those changes
were incorrect, and actually broke the Direct I/O support.

Fix these problems by reverting those portions of that commit.

[ commit description rewritten by tytso ]

Originally from https://github.com/tytso/e2fsprogs/pull/68

Fixes: d557b9659ba9 ("libext2fs: fix potential races in unix_io")
Signed-off-by: Alexander Kanavin <alex.kanavin@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

libext2fs: fix missing mutex unlock in an error path of the Unix I/O manager

Originally from https://github.com/tytso/e2fsprogs/pull/68

Signed-off-by: Alexander Kanavin <alex.kanavin@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

e2fsck: fix unaligned accesses to ext4_fc_add_range and fc_raw_inode

These fast commit related structures can be unaligned on disk.  So we
need to avoid accessing these structures directly, and first copy
them to memory which we know is appropriately aligned.

This fixes an e2fsck crash while running the j_recovery_fast_commit
regression test on a sparc64 system.

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

e2fsck: fix unaligned accesses to ext4_fc_tl struct

Fast commit related struct ext4_fc_tl can be unaligned on disk. So,
while accessing that we should ensure that the pointers are
aligned. This patch fixes unaligned accesses to ext4_fc_tl and also
gets rid of macros fc_for_each_tl and ext4_fc_tag_val that may result
in unaligned accesses to struct ext4_fc_tl.

Signed-off-by: Harshad Shirwadkar <harshadshirwadkar@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

e2fsck: fix portability problems caused by unaligned accesses

The on-disk format for the ext4 journal can have unaigned 32-bit
integers.  This can happen when replaying a journal using a obsolete
checksum format (which was never popularly used, since the v3 format
replaced v2 while the metadata checksum feature was being stablized),
and in the fast commit feature (which landed in the 5.10 kernel,
although it is not enabled by default).

This commit fixes the following regression tests on some platforms
(such as running 32-bit arm architectures on a 64-bit arm kernel):
j_recover_csum2_32bit, j_recover_csum2_64bit, j_recover_fast_commit.

https://github.com/tytso/e2fsprogs/issues/65

Addresses-Debian-Bug: #987641
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

blkid: include time.h to provide explicit declaration for time()

This was originally reported to the MacPorts of e2fsprogs at:

https://github.com/macports/macports-ports/pull/9137

Reported-by: Ryan Schmidt <ryandesign@macports.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

filefrag: minor usability improvements

Add '-V' to filefrag to print the installed version of the tool.

If '-V' is used twice, print out the list of supported FIEMAP flags.
This can be used to check if filefrag understands a specific feature.

Include FIEMAP in the error message printed when filefrag cannot
get the file layout. Since FIEMAP is commonly available and tried
first, it should also be mentioned in the error message unless it
was requested to only run FIBMAP.

Update filefrag.1.in man page to cover the new -V option.

Fix a formatting error with the recently added '-P' options, and
include '-E' and '-P' in the SYNOPSIS section.

Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Lustre-bug-id: https://jira.whamcloud.com/browse/LU-11848
Reviewed-by: Wang Shilong <wshilong@whamcloud.com>
Change-Id: Ib126bdd70efa1775aef6db761f54e27a593ebbe5
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

e2image: add OPTIONS section to man page

Reorganize the e2image.8 man page so that the command-line options
are listed in a dedicated OPTIONS section, rather than being
interspersed among the text in the DESCRIPTION section.  Otherwise,
it is difficult to determine which options are available, and to
find where each option is described.

Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

misc: remove useless code in set_inode_xattr()

In set_inode_xattr(), there are two returns as follows,
-
  return retval;
  return 0;
-
Here, we remove useless 'return 0;' code.

Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

ext2fs: compile the io implementation according to os

In mingw and cygwin, compile the windows_io manager, compile the unix_io
everywhere else.

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

libext2fs: code adaptation to use the Windows IO manager

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

libext2fs: add a Windows implementation of the IO manager

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

resize2fs: close the file system on errors or early exits

When resize2fs exits early, perhaps because of an error, we should
free the file system so that if MMP is in use, the MMP block is reset.
This also releases the memory to avoid memory leak reports.

Addresses-Debian-Bug: #984472
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

resize2fs: avoid allocating over the MMP block

When resizing past the point where the reserve inode has reserved
space for the block group descriptors to expand, and resize2fs (in an
offline resize) needs to move the allocation bitmaps and/or inode
table around, it's possible for resize2fs to allocate over the MMP
block, which would be bad.

Prevent this from happening by reserving the MMP block as a file
system metadata block (which it is) in resize2fs's accounting.

Addresses-Debian-Bug: #984472
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

Update release notes, etc., for the 1.46.2 release

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

libext2fs: fix potential races in unix_io

When unix_io does not use pread/pread64 (which is the case the bounce
buffer is in use, either when Direct I/O is in use or the
IO_FLAG_FORCE_BOUNCE in enabled), there are races between the llseek
and and read or write system calls.  Fix this by using the BOUNCE_MTX
so only one thread is using the file descriptor at a time.

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

libext2s: fix unix_io with IO_FLAG_FORCE_BOUNCE flag set

The bounce read/write code would crash with a floating point exception
if alignment is set to 0.

Fixes: c001596110e8 ("libext2fs: fix unix_io's Direct I/O support")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

po: update es.po (from translationproject.org)

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

libext2fs: fix unix_io's Direct I/O support

The previous Direct I/O support worked on HDD's with 512 byte logical
sector sizes, and on FreeBSD which required 4k aligned memory buffers.
However, it was incomplete and was not correctly working on HDD's with
4k logical sector sizes (aka Advanced Format Disks).

Based on a patch from Alexey Lyashkov <alexey.lyashkov@hpe.com> but
rewritten to work with the latest e2fsprogs and to minimize changes to
make it easier to review.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reported-by: Artem Blagodarenko <artem.blagodarenko@gmail.com>

iscan: fix the test program iscan so it builds again

The iscan program program isn't built by default, and was relying on
e2fsck's util.c, so it had suffered bitrot as e2fsck/util.c had
evolved.  Fix it so that iscan builds correct.

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

mmp: do not use O_DIRECT when working with regular file

Currently the mmp block is read using O_DIRECT to avoid any caching that
may be done by the VM. However when working with regular files this
creates alignment issues when the device of the host file system has
sector size larger than the blocksize of the file system in the file
we're working with.

This can be reproduced with t_mmp_fail test when run on the device with
4k sector size because the mke2fs fails when trying to read the mmp
block.

Fix it by disabling O_DIRECT when working with regular files. I don't
think there is any risk of doing so since the file system layer, unlike
shared block device, should guarantee cache consistency.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

resize2fs: prevent block bitmap warnings when doing extreme fs expansions

This commit fixes a bug where if a small file system is resized to
ridiculous sizes, such that the size of the resized block group
descriptor blocks exceed the original file system, and this would
result in resize2fs triggering a large number of scary warning
messages:

   Illegal block number passed to ext2fs_test_block_bitmap #12440
   for block bitmap for broken.img

This can be replicated via:

   ./misc/mke2fs broken.img -b 1024 -T default
   truncate -s +1500G broken.img
   ./resize/resize2fs broken.img

Fortunately, aside from triggering these warning messages, the bug had
no other bad effects.

https://github.com/tytso/e2fsprogs/issues/60

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

debugfs: fix rdump and ls to handle uids and gids > 65536 correctly

https://github.com/tytso/e2fsprogs/issues/63

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

Remove superfluous break statements

https://github.com/tytso/e2fsprogs/pull/45
https://github.com/tytso/e2fsprogs/pull/46

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

chattr/lsattr: expose FS_NOCOMP_FL (aka EXT2_NOCOMPR_FL)

This adds support for setting/querying the FS_NOCOMP_FL/EXT2_NOCOMPR_FL
file flag to chattr/lsattr. I picked the character "m" because it was
so far unused and all other characters that were more obvious candidates
were already taken.

The flag is available on btrfs, and with this patch it is possible to
manage it correctly.

Signed-off-by: Lennart Poettering <lennart@poettering.net>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

debugfs: fix memory leak problem in read_list()

In read_list func, if strtoull() fails in while loop,
we will return the error code directly. Then, memory of
variable lst will be leaked without setting to *list.

Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Signed-off-by: linfeilong <linfeilong@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

libe2p: remove nested #ifdefs in fgetflags.c and fsetflags.c

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

debugfs: fix file descriptor leak on error path in do_logdump()

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

misc/Makefile.in: add rules to build static versions of lsattr and chattr

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

libe2p: skip unneeded lstat(2) in fgetflags() and fsetflags()

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

debugfs: fix uninitialized variable es in do_logdump()

Fixes: d96064e9821a ("debugfs: simplify the do_logdump() function")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

debugfs: simplify the do_logdump() function

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

Add checks for fs->blocksize == 0 which could cause some crashes

This should never happeb, but some checks is useful, and also fixes
some Coverity warnings.

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

e2fsck: fix miscellaneous clang warnings

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

e2fsck: initialize variable before first use in fast commit replay

Initialize ext2fs_ex variable in ext4_fc_replay_scan() before first
use.  Also make sure ext2fs_decode_extent() completely overwrites the
extent structure passed to it as argument to prevent potential future
bugs for the users of the function.

Signed-off-by: Harshad Shirwadkar <harshadshirwadkar@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

e2fsck: add fallthrough comment in fc replay switch case

During fast commit replay scan phase, in ext4_fc_replay_scan(), we
want to fallthrough in switch case for EXT4_FC_TAG_ADD_RANGE case. Add
a comment for that.

Signed-off-by: Harshad Shirwadkar <harshadshirwadkar@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

ext2fs: don't ignore return value in ext2fs_count_blocks

Don't ignore return value of ext2fs_get_array() in
ext2fs_count_blocks().

Signed-off-by: Harshad Shirwadkar <harshadshirwadkar@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

e2fsck: don't ignore return values in e2fsck_rewrite_extent_tree

Don't ignore return values of library function calls in
e2fsck_rewrite_extent_tree.

Signed-off-by: Harshad Shirwadkar <harshadshirwadkar@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

create_inode: fix potential file descriptor leak on an error path

Addresses-Coverity-Bug: 1473307
Signed-off-by: Theodore Ts'o <tytso@mit.edu>