Whamcloud - gitweb
tools/e2fsprogs.git
21 hours agodebugfs: fix printing post-2038 inode timestamps next
Theodore Ts'o [Thu, 18 Apr 2024 18:16:55 +0000 (14:16 -0400)]
debugfs: fix printing post-2038 inode timestamps

Debugfs's stat command called ext2fs_inode_xtime_get() with a struct
inode * instead of a struct large_inode *.  As a result, printing
inode timestamps will be incorrect if the time value is larger than
2**32.

Fixes: ca8bc9240a00 ("Add post-2038 timestamp support to e2fsprogs")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
21 hours agomisc: update mke2fs's man page regarding the default inode size
Pascal Hambourg [Wed, 17 Apr 2024 11:48:22 +0000 (13:48 +0200)]
misc: update mke2fs's man page regarding the default inode size

Since a23b50cd ("mke2fs: warn about missing y2038 support when
formatting fresh ext4 fs"), the default inode size is 256 bytes
for all filesystems, including small and floppy, except for the
Hurd since it currently only supports 128-byte inodes.

Signed-off-by: Pascal Hambourg <pascal@plouf.fr.eu.org>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/E1rx4t4-00073d-1e@zenith
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
24 hours agodebugfs: fix set_field's handling of timestamps
Theodore Ts'o [Thu, 18 Apr 2024 14:04:27 +0000 (10:04 -0400)]
debugfs: fix set_field's handling of timestamps

How timestamps are encoded in inodes and superblocks are different.
Unfortunately, commit ca8bc9240a00 which added post-2038 timestamps
was (a) overwriting adjacent superblock fields and/or attempting
unaligned writes to a 8-bit field from a 32-bit pointer, and (b) using
the incorrect encoding for timestamps stored in inodes.  Fix both of
these issues, which were found thanks to UBSAN.

Fixes: ca8bc9240a00 ("Add post-2038 timestamp support to e2fsprogs")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
34 hours agoci.yml: upload *.failed test failure artifacts
Theodore Ts'o [Thu, 18 Apr 2024 04:15:54 +0000 (00:15 -0400)]
ci.yml: upload *.failed test failure artifacts

Also upgrade to v4 of the upload-artifact and checkout actions.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
35 hours agoFix 32-bit build and test failures
Theodore Ts'o [Thu, 18 Apr 2024 03:47:02 +0000 (23:47 -0400)]
Fix 32-bit build and test failures

Commit ca8bc9240a00 ("Add post-2038 timestamp support to e2fsprogs")
was never built or tested on a 32-bit.  It introduced some build
problems when time_t is a 32-bit integer, and it exposed some test
bugs.  Fix them.

Fixes: ca8bc9240a00 ("Add post-2038 timestamp support to e2fsprogs")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
42 hours agoe2fsck: don't try backup superblocks beyond the size of the device
Theodore Ts'o [Wed, 17 Apr 2024 20:52:00 +0000 (16:52 -0400)]
e2fsck: don't try backup superblocks beyond the size of the device

Commit f7ef5f3e356d ("e2fsck: check all sparse_super backups") tries
to limit the number of block groups to search for backup superblocks
based on ctx->num_blocks.  Unfortunately, get_backup_sb() gets called
before ctx->num_blocks is set, so we try all block groups up to 2**32
- 1.  Not only does this waste time trying to read from blocks that
don't exist, it triggers the UBSAN checker when multiplying a very
large number by the block size.

Fix this by using ext2fs_get_Device_size(), and if that isn't
available, arbitrarily cap things so that we search block groups up to
128.

Fixes: f7ef5f3e356d ("e2fsck: check all sparse_super backups")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
45 hours agodebugfs: fix clang warnings
Theodore Ts'o [Wed, 17 Apr 2024 17:58:49 +0000 (13:58 -0400)]
debugfs: fix clang warnings

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 days agoext2fs: fix -Walloc-size
Sam James [Tue, 7 Nov 2023 23:31:20 +0000 (23:31 +0000)]
ext2fs: fix -Walloc-size

GCC 14 introduces a new -Walloc-size included in -Wextra which gives:
```
lib/ext2fs/hashmap.c:37:36: warning: allocation of insufficient size ‘1’ for type ‘struct ext2fs_hashmap’ with size ‘20’ [-Walloc-size]
```

The calloc prototype is:
```
void *calloc(size_t nmemb, size_t size);
```

So, just swap the number of members and size arguments to match the prototype, as
we're initialising 1 struct of size `sizeof(...)`. GCC then sees we're not
doing anything wrong.

Signed-off-by: Sam James <sam@gentoo.org>
Link: https://lore.kernel.org/r/20231107233122.2013191-1-sam@gentoo.org
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 days agoe4crypt: fix spurious "Success" error message
Jakub Wilk [Wed, 27 Dec 2023 08:08:05 +0000 (09:08 +0100)]
e4crypt: fix spurious "Success" error message

Before:

    # e4crypt set_policy 0000000000000000 /dev/null
    /dev/null is not a directory
    /dev/null: Success

After:

    # e4crypt set_policy 0000000000000000 /dev/null
    /dev/null: Not a directory

Signed-off-by: Jakub Wilk <jwilk@jwilk.net>
Link: https://lore.kernel.org/r/20231227080805.6801-1-jwilk@jwilk.net
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 days agodebugfs: fix infinite loop while dumping the journal
Wenchao Hao [Fri, 17 Nov 2023 10:23:15 +0000 (18:23 +0800)]
debugfs: fix infinite loop while dumping the journal

There are 2 scenarios which would trigger infinite loop:

1. None log is recorded, then dumplog with "-n", for example:
   debugfs -R "logdump -O -n 10" /dev/xxx
   while /dev/xxx has no valid log recorded.
2. The log area is full and cycle write is triggered, then dumplog with
   debugfs -R "logdump -aOS" /dev/xxx

This patch add a new flag "wrapped_flag" to mark if logdump has
reached to tail of logarea set in macro WRAP().

If wrapped_flag is true, and we comes to first_transaction_blocknr
again, just break the logdump loop.

[ Renamed reverse_flag to wrapped_flag to make it clearer what it is. -- TYT ]

Signed-off-by: Wenchao Hao <haowenchao2@huawei.com>
Link: https://lore.kernel.org/r/20231117102315.2431846-1-haowenchao2@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 days agoresize2fs: avoid constantly flushing while moving blocks
Anssi Hannula [Tue, 7 Nov 2023 09:46:53 +0000 (11:46 +0200)]
resize2fs: avoid constantly flushing while moving blocks

resize2fs block_mover() flushes data after each extent and, curiously,
only if progress indicator is enabled, every inode_blocks_per_group
blocks.

This significantly affects performance, e.g. on a tested large
filesystem on top of MD-RAID6+LVM+dm-crypt these flush calls reduce the
operation rate from approx. 500MB/s to 5MB/s, causing extremely long
shrinking times for large size deltas (70TB in my case).

Since this step performs just plain data copying and does not e.g. save
any progress/checkpoint information or similar metadata, it seems like
this flushing is of very limited usefulness, especially when considering
the (in some cases) 100x performance impact.

Remove the mid-operation flushes and only flush after all blocks have
been moved.

Signed-off-by: Anssi Hannula <anssi.hannula@iki.fi>
Link: https://lore.kernel.org/r/20231107094920.4056281-1-anssi.hannula@iki.fi
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 days agoMerge branch 'maint' into next
Theodore Ts'o [Tue, 16 Apr 2024 16:50:46 +0000 (12:50 -0400)]
Merge branch 'maint' into next

3 days agotests: new test to check quota after a bad inode deallocation
Luis Henriques (SUSE) [Fri, 5 Apr 2024 14:24:05 +0000 (15:24 +0100)]
tests: new test to check quota after a bad inode deallocation

This new test validates e2fsck by verifying that quota is updated after a bad
inode is deallocated.  It mimics fstest ext4/019 by including a filesystem image
where a symbolic link was created to an existing file, using a long symlink
name.  This symbolic link was then wiped with:

  # debugfs -w -R 'zap -f /testlink 0' f_testnew/image

Signed-off-by: Luis Henriques (SUSE) <luis.henriques@linux.dev>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/20240405142405.12312-5-luis.henriques@linux.dev
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
3 days agotests: new test to check quota after directory optimization
Luis Henriques (SUSE) [Fri, 5 Apr 2024 14:24:04 +0000 (15:24 +0100)]
tests: new test to check quota after directory optimization

This new test validates e2fsck by verifying that quota data is updated after a
directory optimization is performed.  This issue was initially found by fstest
ext4/014, and this test was based on it.  It includes a filesystem image where
the lost+found directory is unlinked after a new link to it is created:

  # debugfs -w -R "ln lost+found foo" f_testnew/image
  # debugfs -w -R "unlink lost+found" f_testnew/image

Signed-off-by: Luis Henriques (SUSE) <luis.henriques@linux.dev>
Link: https://lore.kernel.org/r/20240405142405.12312-4-luis.henriques@linux.dev
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
3 days agoe2fsck: update quota when deallocating a bad inode
Luis Henriques (SUSE) [Fri, 5 Apr 2024 14:24:03 +0000 (15:24 +0100)]
e2fsck: update quota when deallocating a bad inode

If a bad inode is found it will be deallocated.  However, if the filesystem has
quota enabled, the quota information isn't being updated accordingly.  This
issue was detected by running fstest ext4/019.

This patch fixes the issue by decreasing the inode count from the
quota and, if blocks are also being released, also subtract them as well.

While there, and as suggested by Andreas Dilger, the deallocate_inode()
function documentation is also updated by this patch to make it clear what
that function really does.

Signed-off-by: Luis Henriques (SUSE) <luis.henriques@linux.dev>
Link: https://lore.kernel.org/r/20240405142405.12312-3-luis.henriques@linux.dev
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
3 days agoe2fsck: update quota accounting after directory optimization
Luis Henriques (SUSE) [Fri, 5 Apr 2024 14:24:02 +0000 (15:24 +0100)]
e2fsck: update quota accounting after directory optimization

In "Pass 3A: Optimizing directories", a directory may have it's size reduced.
If that happens and quota is enabled in the filesystem, the quota information
will be incorrect because it doesn't take the rehash into account.  This issue
was detected by running fstest ext4/014.

This patch simply updates the quota data accordingly, after the directory is
written and it's size has been updated.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=218626
Signed-off-by: Luis Henriques (SUSE) <luis.henriques@linux.dev>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/20240405142405.12312-2-luis.henriques@linux.dev
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
3 days agoe2fsprogs: misc/mke2fs.8.in: Correct valid cluster-size values
Srivathsa Dara [Wed, 3 Apr 2024 04:30:37 +0000 (04:30 +0000)]
e2fsprogs: misc/mke2fs.8.in: Correct valid cluster-size values

According to the mke2fs man page, the supported cluster-size values
for an ext4 filesystem are 2048 to 256M bytes. However, this is not
the case.

When mkfs is run to create a filesystem with following specifications:
* 1k blocksize and cluster-size greater than 32M
* 2k blocksize and cluster-size greater than 64M
* 4k blocksize and cluster-size greater than 128M
mkfs fails with "Invalid argument passed to ext2 library while trying
to create journal" error. In general, when the cluster-size to blocksize
ratio is greater than 32k, mkfs fails with this error.

Went through the code and found out that the function
`ext2fs_new_range()` is the source of this error. This is because when
the cluster-size to blocksize ratio exceeds 32k, the length argument
to the function `ext2fs_new_range()` results in 0. Hence, the error.

This patch corrects the valid cluster-size values.

Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Link: https://lore.kernel.org/r/20240403043037.3992724-1-srivathsa.d.dara@oracle.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agolib/ext2fs: llseek: simplify linux section
Mike Gilbert [Tue, 7 Nov 2023 23:33:22 +0000 (23:33 +0000)]
lib/ext2fs: llseek: simplify linux section

On 32-bit musl systems, off_t is always 8 bytes regardless of
_FILE_OFFSET_BITS. The previous code did not cover this case.

The previous #ifdef logic was rather confusing, so I reworked it into a
more understandable form.

Bug: https://bugs.gentoo.org/908892
Signed-off-by: Mike Gilbert <floppym@gentoo.org>
Closes: https://github.com/tytso/e2fsprogs/pull/150
Signed-off-by: Sam James <sam@gentoo.org>
Link: https://lore.kernel.org/r/20231107233323.2013334-2-sam@gentoo.org
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agoconfigure.ac: call AC_SYS_LARGEFILE before checking the size of off_t
Mike Gilbert [Tue, 7 Nov 2023 23:33:21 +0000 (23:33 +0000)]
configure.ac: call AC_SYS_LARGEFILE before checking the size of off_t

Signed-off-by: Mike Gilbert <floppym@gentoo.org>
Signed-off-by: Sam James <sam@gentoo.org>
Link: https://lore.kernel.org/r/20231107233323.2013334-1-sam@gentoo.org
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agoconfigure: update configure so it is generated using autoconf 2.71
Theodore Ts'o [Thu, 4 Apr 2024 14:59:54 +0000 (10:59 -0400)]
configure: update configure so it is generated using autoconf 2.71

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agomke2fs: do not set the BLOCK_UNINIT on groups has GDT
Li Dongyang [Mon, 25 Sep 2023 06:08:01 +0000 (16:08 +1000)]
mke2fs: do not set the BLOCK_UNINIT on groups has GDT

This patch prepares the expansion of GDT blocks beyond a
single group, by make mke2fs to not set BLOCK_UNINIT on
groups with GDT blocks, block/inode bitmaps, or inode table
blocks allocated.

Otherwise, we still rely on kernel side to initialize the
block bitmap if the groups has BLOCK_UNINIT set, and the
kernel doesn't know a group could have GDT blocks allocated,
so it would make an bad block bitmap.

As a result, expect output of several tests needs to be changed,
especially if the test uses dumpe2fs to print the group summary.

Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/20230925060801.1397581-2-dongyangli@ddn.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agomke2fs: set free blocks accurately for groups has GDT
Li Dongyang [Mon, 25 Sep 2023 06:08:00 +0000 (16:08 +1000)]
mke2fs: set free blocks accurately for groups has GDT

This patch is part of the preparation required to allow
GDT blocks expand beyond a single group,
it introduces 2 new interfaces:
- ext2fs_count_used_blocks(), to return the blocks used
in the bitmap range.
- ext2fs_reserve_super_and_bgd2() to return blocks used by
superblock/GDT blocks for every group, by looking up blocks used.

Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/20230925060801.1397581-1-dongyangli@ddn.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agoe2fsck: check all sparse_super backups
Andreas Dilger [Mon, 4 Sep 2023 04:57:42 +0000 (14:57 +1000)]
e2fsck: check all sparse_super backups

Teach e2fsck to look for backup super blocks in the "sparse_super"
groups, by checking group #1 first and then powers of 3^n, 5^n,
and 7^n, up to the limit of available block groups.

Export ext2fs_list_backups() function to efficiently iterate groups
for backup sb/GDT instead of checking every group.  Ensure that the
group counters do not try to overflow the 2^32-1 group limit, and
try to limit scanning to the size of the block device (if available).

Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Link: https://lore.kernel.org/r/20230904045742.827584-1-dongyangli@ddn.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agomke2fs: batch zeroing inode table
Li Dongyang [Mon, 4 Sep 2023 04:58:06 +0000 (14:58 +1000)]
mke2fs: batch zeroing inode table

For flex_bg enabled fs, we could merge the
inode table blocks into a contiguous range,
this improves mke2fs time on large devices
when lazy_itable_init is disabled.

On a 977TB device, unpatched mke2fs was running
for 449m10s before getting terminated manually.
strace shows huge number of fallocate, given the
offset from fallocate it has done 41% of the inode
tables, the estimated time needed would be 1082m.

unpatched  patched
real 449m10.954s 4m20.531s
user 0m18.217s 0m16.147s
sys 0m20.311s 0m8.944s

Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Link: https://lore.kernel.org/r/20230904045806.827621-1-dongyangli@ddn.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agolibext2fs: always refuse to open a file system with a zero s_desc_size maint
Theodore Ts'o [Wed, 3 Apr 2024 21:30:37 +0000 (17:30 -0400)]
libext2fs: always refuse to open a file system with a zero s_desc_size

Commit 42c11edd0863 ("ext2fs_open[2](), return an error if s_desc_size
is too large") added a check for an insanely large s_desc_size to
prevent some failures triggered by fuzz testing.  However, it would
allow e2fsck to fall back to recover the file system by using the
backup superblocks by having e2fsck pass the flag
EXT2_FLAG_IGNORE_SB_ERRORS.  But by allowing an s_desc_Size of zero,
it's possible that e2fsck will die with a division of zero error.
With this fix, e2fsck will now print an error message and exit
instead.

https://github.com/tytso/e2fsprogs/issues/183

Fixes: 42c11edd0863 ("ext2fs_open[2](), return an error if s_desc_size is too large")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agoAdd post-2038 timestamp support to e2fsprogs
Andreas Dilger [Wed, 27 Sep 2023 05:40:16 +0000 (23:40 -0600)]
Add post-2038 timestamp support to e2fsprogs

The ext4 kernel code implemented support for s_mtime_hi,
s_wtime_hi, and related timestamp fields to avoid timestamp
overflow in 2038, but similar handling is not in e2fsprogs.

Add helper macros for the superblock _hi timestamp fields
ext2fs_super_tstamp_get() and ext2fs_super_tstamp_set().

Add helper macro for inode _extra timestamp fields
ext2fs_inode_xtime_get() and ext2fs_inode_xtime_set().

Add helper macro ext2fs_actual_inode_size() to avoid open
coding the i_extra_isize check in multiple places.

Remove inode_time_to_string() since this is unused once callers
change to time_to_string(ext2fs_inode_xtime_get()) directly.

Fix inode_includes() macro to properly wrap "inode" parameter,
and rename to ext2fs_inode_includes() to avoid potential name
clashes.  Use this to check inode field inclusion in debugfs
instead of bare constants for inode field offsets.

Use these interfaces to access timestamps in debugfs, e2fsck,
libext2fs, fuse2fs, tune2fs, and e2undo.

Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/20230927054016.16645-1-adilger@dilger.ca
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agolibext2fs: fix ext2fs_get_device_size2() return value on Windows
Eric Biggers [Wed, 1 Mar 2023 03:45:18 +0000 (19:45 -0800)]
libext2fs: fix ext2fs_get_device_size2() return value on Windows

Creating a file system on Windows without a pre-existing file stopped
working because the Windows version of ext2fs_get_device_size2() doesn't
return ENOENT if the file doesn't exist.  Fix this.

Fixes: 53464654bd33 ("mke2fs: fix creating a file system image w/o a pre-existing file")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Link: https://lore.kernel.org/r/20230301034518.373859-1-ebiggers@kernel.org
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agodebian: drop "package-supports-alternative-init-but-no-init.d-script" override
Theodore Ts'o [Sun, 31 Mar 2024 03:55:44 +0000 (23:55 -0400)]
debian: drop "package-supports-alternative-init-but-no-init.d-script" override

The Lintian tag package-supports-alternative-init-but-no-init.d-script
has been removed, as it's obsolete.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 weeks agodebian: acknowledge NMU'ed changes to 1.47.0-2.4
Theodore Ts'o [Sat, 30 Mar 2024 23:14:36 +0000 (19:14 -0400)]
debian: acknowledge NMU'ed changes to 1.47.0-2.4

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
3 weeks agoe2scrub: skip filesystems that don't have journals
Darrick J. Wong [Wed, 10 Jan 2024 05:58:09 +0000 (21:58 -0800)]
e2scrub: skip filesystems that don't have journals

Brian J. Murrell reported that e2scrub reports failures with one of his
filesystems.  From the email discussion after he supplied a metadump:

AHA!  This is an ext2 filesystem, since it doesn't have the
"has_journal" or "extents" features turned on:

# e2image -r /tmp/disk.qcow2 /dev/sda
# dumpe2fs /dev/sda -h
dumpe2fs 1.47.1~WIP-2023-12-27 (27-Dec-2023)
Filesystem volume name:   <none>
Last mounted on:          /opt
Filesystem UUID:          2c70368a-0d54-4805-8620-fda19466d819
Filesystem magic number:  0xEF53
Filesystem revision #:    1 (dynamic)
Filesystem features:      ext_attr resize_inode dir_index filetype sparse_super large_file
Filesystem flags:         signed_directory_hash
Default mount options:    user_xattr acl
Filesystem state:         not clean with errors

(Note: Filesystem state == "clean" means that EXT2_VALID_FS is set in
the superblock s_state field; "not clean with errors" means that the
flag is not set.)

I bet the "journal only" preen doesn't actually reset the filesystem
state either:

# e2fsck -E journal_only -p  /dev/sda
# dumpe2fs /dev/sda -h | grep state
dumpe2fs 1.47.1~WIP-2023-12-27 (27-Dec-2023)
Filesystem state:         not clean with errors

Nope.

So now I know what happened -- when mounting an ext* filesystem that
doesn't have a journal, the driver clears EXT2_VALID_FS from the primary
superblock.  This forces the system to run e2fsck after a crash, because
that's what you have to do for unjournalled filesystems.

The "e2fsck -E journal_only -p" call in e2scrub only replays the
journal.  Since there is no journal, it exits almost immediately.
That's the intended behavior, but then it means that the "e2fsck -fy"
call immediately after sees that the superblock doesn't have
EXT2_VALID_FS set, sets it, and makes e2fsck return 1.

So that's why you're getting the e2scrub failures.

Contrast this to what you get when the filesystem has a journal:

# dumpe2fs -h /dev/sdb
dumpe2fs 1.47.0 (5-Feb-2023)
Filesystem volume name:   <none>
Last mounted on:          <not available>
Filesystem UUID:          e18b8b57-a75e-4316-87ce-6a08969476c3
Filesystem magic number:  0xEF53
Filesystem revision #:    1 (dynamic)
Filesystem features:      has_journal ext_attr resize_inode dir_index filetype needs_recovery sparse_super large_file
Filesystem flags:         signed_directory_hash
Default mount options:    user_xattr acl
Filesystem state:         clean

Filesystems with journals retain their EXT4_VALID_FS state when they're
mounted.

Hmm.  What e2scrub should do about unjournalled filesystems is a thorny
question.  My initial thought is that it should skip them, because a
mounted unjournalled filesystem cannot by definition be kept consistent.
Therefore, teach e2scrub_all to avoid them and e2scrub to fail them at
the onset.

Restricting the scope of e2scrub sucks, but in the meantime at least it
means that your filesystem isn't massively corrupt.  Thanks for the
metadump, it was very useful for root cause analysis.

Reported-by: "Brian J. Murrell" <brian@interlinx.bc.ca>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Link: https://lore.kernel.org/r/20240110055809.GD722946@frogsfrogsfrogs
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
3 weeks agodebian: don't restart e2scrub_all when upgrading package
Darrick J. Wong [Wed, 10 Jan 2024 05:57:24 +0000 (21:57 -0800)]
debian: don't restart e2scrub_all when upgrading package

When installing or upgrading the e2fsprogs package, only start the
e2scrub_all timer and the reaping service.  Don't restart e2scrub_all
itself, because that will kill any scrubs in progress, which will
trigger the failure reporting.

Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Link: https://lore.kernel.org/r/20240110055724.GC722946@frogsfrogsfrogs
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
3 weeks agoe2scrub_fail: move executable script to /usr/libexec
Darrick J. Wong [Sun, 31 Dec 2023 20:39:03 +0000 (12:39 -0800)]
e2scrub_fail: move executable script to /usr/libexec

Per FHS 3.0, non-PATH executable binaries are supposed to live under
/usr/libexec, not /usr/lib.  e2scrub_fail is an executable script, so
move it to libexec in case some distro some day tries to mount /usr/lib
as noexec or something.  Also, there's no reason why these scripts need
to be put under an arch-dependent path.

Cc: Neal Gompa <neal@gompa.dev>
Link: https://refspecs.linuxfoundation.org/FHS_3.0/fhs/ch04s07.html
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Neal Gompa <neal@gompa.dev>
Link: https://lore.kernel.org/r/20231231203903.GC36164@frogsfrogsfrogs
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
3 weeks agoMerge branch 'e2scrub-fixes' of https://git.kernel.org/pub/scm/linux/kernel/git/djwon...
Theodore Ts'o [Thu, 28 Mar 2024 18:39:24 +0000 (14:39 -0400)]
Merge branch 'e2scrub-fixes' of https://git./linux/kernel/git/djwong/e2fsprogs into next

3 weeks agoMerge branch 'maint' into next
Theodore Ts'o [Thu, 28 Mar 2024 17:51:24 +0000 (13:51 -0400)]
Merge branch 'maint' into next

3 weeks agoe2fsck: fix various -Wall nits picked up by clang
Theodore Ts'o [Thu, 28 Mar 2024 17:46:44 +0000 (13:46 -0400)]
e2fsck: fix various -Wall nits picked up by clang

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
3 weeks agoblkid: fix -Wunused-but-set-variable warning in blkid_read_cache()
Theodore Ts'o [Thu, 28 Mar 2024 17:33:04 +0000 (13:33 -0400)]
blkid: fix -Wunused-but-set-variable warning in blkid_read_cache()

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 months agotune2fs: check whether filesystem is in use for I_flag and Q_flag test
Zhiqiang Liu [Mon, 20 Mar 2023 05:04:34 +0000 (13:04 +0800)]
tune2fs: check whether filesystem is in use for I_flag and Q_flag test

For changing inode size (-I) and setting quota fearture (-Q), tune2fs
only check whether the filesystem is umounted. Considering mount
namepspaces, the filesystem is umounted, however it already be left
in other mount namespace.
So we add one check whether the filesystem is not in use with using
EXT2_MF_BUSY flag, which can indicate the device is already opened
with O_EXCL, as suggested by Ted.

Reported-by: Baokun Li <libaokun1@huawei.com>
Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Signed-off-by: zhanchengbin <zhanchengbin1@huawei.com>
Link: https://lore.kernel.org/r/28455341-ca26-d203-8b54-792bae002251@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 months agoext2fs: make sure we have at least EXT2_FIRST_INO + 1 inodes
Li Dongyang [Thu, 20 Jul 2023 12:50:12 +0000 (22:50 +1000)]
ext2fs: make sure we have at least EXT2_FIRST_INO + 1 inodes

When creating a small fs with 100 1k blocks, mke2fs fails with:

Creating filesystem with 100 1k blocks and 8 inodes

Allocating group tables: done
Writing inode tables: done
ext2fs_mkdir: Could not allocate inode in ext2 filesystem while creating /lost+found

Increase s_inodes_per_group with a step of 8 to make
sure we have at least EXT2_FIRST_INO + 1 inodes.

Change-Id: Ib885735641dfa0ed9c6f6a4a1f9afec291673126
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/20230720125012.641504-1-dongyangli@ddn.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 months agoe2image: correct group descriptors size in ext2fs_image_super_read()
Emoly Liu [Fri, 14 Jul 2023 00:59:58 +0000 (10:59 +1000)]
e2image: correct group descriptors size in ext2fs_image_super_read()

In function ext2fs_image_super_read(), the size of block group
descriptors should be (fs->blocksize * fs->desc_blocks), but not
(fs->blocksize * fs->group_desc_count).

Signed-off-by: Emoly Liu <emoly@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/20230714005958.442487-1-dongyangli@ddn.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
3 months agodebugfs: dx_hash: honor the unsigned hash flag if a file system is opened
Theodore Ts'o [Tue, 16 Jan 2024 02:59:49 +0000 (21:59 -0500)]
debugfs: dx_hash: honor the unsigned hash flag if a file system is opened

If we are using the hash seed and hash version from an open file
systenm, then we should also use the unsigned version of the hash
algorithm if the superblock as the unsigned hash flag set.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
3 months agodebugfs: teach the dx_hash command the -v option
Theodore Ts'o [Tue, 16 Jan 2024 02:49:29 +0000 (21:49 -0500)]
debugfs: teach the dx_hash command the -v option

Add an option for dx_hash to print more details about the hash
algorithm and hash seed to calculate the directory hash value.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 months agoe2fsprogs: don't allow udisks to automount ext4 filesystems with no prompt
Darrick J. Wong [Thu, 24 Aug 2023 22:26:31 +0000 (15:26 -0700)]
e2fsprogs: don't allow udisks to automount ext4 filesystems with no prompt

The unending stream of syzbot bug reports and overwrought filing of CVEs
for corner case handling (i.e. things that distract from actual user
complaints) in ext4 has generated all sorts of of overheated rhetoric
about how every bug is a Serious Security Issue(tm) because anyone can
craft a malicious filesystem on a USB stick, insert the stick into a
victim machine, and mount will trigger a bug in the kernel driver that
leads to some compromise or DoS or something.

I thought that nobody would be foolish enough to automount an ext4
filesystem.  What a fool I was!  It turns out that udisks can be told
that it's okay to automount things, and then GNOME will do exactly that.
Including mounting mangled ext4 filesystems!

<delete angry rant about poor decisionmaking and armchair fs developers
blasting us on X while not actually doing any of the work>

Turn off /this/ idiocy by adding a udev rule to tell udisks not to
automount ext4 filesystems.

This will not stop a logged in user from unwittingly inserting a
malicious storage device and pressing [mount] and getting breached.
This is not a substitute for a thorough audit of all codebases.  This is
not a substitute for lklfuse.  This does not solve the general problem
of in-kernel fs drivers being a huge attack surface.  I just want a
vacation from the sh*tstorm of bad ideas and threat models that I never
agreed to support.

Signed-off-by: Darrick J. Wong <djwong@kernel.org>
4 months agoe2scrub: fix pathname escaping across all service definitions
Darrick J. Wong [Fri, 17 Mar 2023 18:19:11 +0000 (11:19 -0700)]
e2scrub: fix pathname escaping across all service definitions

systemd services provide an "instance name" that can be associated with
a particular invocation of a service.  This allows service users to
invoke multiple copies of a service, each with a unique string.  For
e2scrub, we pass the mountpoint of the filesystem as the instance name.
However, systemd services aren't supposed to have slashes in them, so
we're supposed to escape them.

The canonical escaping scheme for pathnames is defined by the
systemd-escape --path command.  Unfortunately, we've been adding our own
opinionated sauce for years, to work around the fact that --path didn't
quite work right in systemd before January 2017.  The special sauce is
incorrect, and we no longer care about systemd of 7 years past.

Clean up this mess by following the systemd escaping scheme throughout
the service units.  Now we can use the '%f' specifier in them, which
makes things a lot less complicated.

Signed-off-by: Darrick J. Wong <djwong@kernel.org>
4 months agoMerge branch 'maint' into next master
Theodore Ts'o [Thu, 7 Dec 2023 16:02:53 +0000 (11:02 -0500)]
Merge branch 'maint' into next

4 months agodebugfs: Use the hash_version from superblock if a file system is opened
Srivathsa Dara [Thu, 24 Aug 2023 06:56:34 +0000 (06:56 +0000)]
debugfs: Use the hash_version from superblock if a file system is opened

The debugfs program's dx_hash command computes the hash for the given
filename, taking the hash_seed and hash_version (i.e hash algorithm)
as arguments.  So the user has to refer to the superblock to get these
values used by the filesystem.  So if debugfs has an opened file
system, use those values from the current file system.

[ Fixed patch to avoid crashing when a file system is not opened. --TYT ]

Signed-off-by: Srivathsa Dara <srivathsa.d.dara@oracle.com>
Link: https://lore.kernel.org/r/20230824065634.2662858-1-srivathsa.d.dara@oracle.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 months agotune2fs: fall back to old get/set fs label on error
Li Dongyang [Sat, 20 May 2023 10:43:29 +0000 (20:43 +1000)]
tune2fs: fall back to old get/set fs label on error

If we fail to get/open the mount point for get/set
fs label ioctl, just fall back to old method and
silence the error messages.

Fixes: f85b4526f ("tune2fs: implement support for set/get label iocts")
Signed-off-by: Li Dongyang <dongyangli@ddn.com>
Link: https://lore.kernel.org/r/20230520104329.2402182-1-dongyangli@ddn.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 months agoext2fs: don't retry discard/zeroout repeatedly
Andreas Dilger [Wed, 10 May 2023 05:18:49 +0000 (23:18 -0600)]
ext2fs: don't retry discard/zeroout repeatedly

Call safe_getenv(UNIX_IO_NOZEROOUT) once when the device is
opened and set CHANNEL_FLAG_NOZEROOUT if present instead of
getting uid/euid/getenv every time unix_zeroout() is called.

For unix_discard() and unix_zeroout() don't continue to call
them if the block device doesn't support these operations.

Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/1683695929-26972-1-git-send-email-adilger@dilger.ca
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 months agotune2fs/fuse2fs/debugfs: save error information during journal replay
Baokun Li [Fri, 17 Feb 2023 10:09:22 +0000 (18:09 +0800)]
tune2fs/fuse2fs/debugfs: save error information during journal replay

Saving error information during journal replay, as in the kernel,
prevents information loss from making problems difficult to locate.
We save these error information until someone uses e2fsck to check
for and fix possible errors.

Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: zhanchengbin <zhanchengbin1@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230217100922.588961-3-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 months agoe2fsck: save EXT2_ERROR_FS flag during journal replay
Baokun Li [Fri, 17 Feb 2023 10:09:21 +0000 (18:09 +0800)]
e2fsck: save EXT2_ERROR_FS flag during journal replay

When repairing a file system with s_errno missing from the journal
superblock but the file system superblock contains the ERROR_FS flag,
the ERROR_FS flag on the file system image is overwritten after the
journal replay, followed by a reload of the file system data from disk
and the ERROR_FS flag in memory is overwritten. Also s_errno is not set
and the ERROR_FS flag is not reset. Therefore, when checked later, no
forced check is performed, which makes it possible to have some errors
hidden in the disk image, which may make it read-only when using the
file system. So we save the ERROR_FS flag to the superblock after the
journal replay, instead of just relying on the jsb->s_errno to do this.

Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: zhanchengbin <zhanchengbin1@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230217100922.588961-2-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 months agobuild: split version and release in configure
Andreas Dilger [Wed, 10 May 2023 04:57:57 +0000 (22:57 -0600)]
build: split version and release in configure

Update configure.ac to separate Version from Release if there is
a '-' in version.h::E2FSPROGS_VERSION (e.g. "1.46.6-rc1").
Otherwise, the '-' in the version can make RPM building unhappy.

Simplify the generation of E2FSPROGS_VERESION, E2FSPROGS_DATE and
E2FSPROGS_DAY to avoid multiple grep/awk/sed/tr stages.

Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/1683694677-9366-1-git-send-email-adilger@dilger.ca
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 months agotune2fs.c: define PATH_MAX if it is not defined by the system headers
Theodore Ts'o [Mon, 20 Nov 2023 02:06:12 +0000 (21:06 -0500)]
tune2fs.c: define PATH_MAX if it is not defined by the system headers

This is needed to compile on GNU/Hurd.

Addresses-Debian-Bug: #1056145
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
7 months agolibext2fs: don't truncate the orphan file inode if it is newly allocated
Theodore Ts'o [Fri, 25 Aug 2023 21:28:01 +0000 (17:28 -0400)]
libext2fs: don't truncate the orphan file inode if it is newly allocated

In ext2fs_create_orphan_file(), don't try truncating inode for the
orphan file if ext2fs_create_orphan_file() allocated the inode.  This
avoids problems where the newly allocated inode in the inode table
might contain garbage; if the metadata checksum feature is enabled,
this will generally result in the function failing with a checksum
invalid error, but this can cause mke2fs (which calls
ext2fs_create_orphan_file) to fail.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
8 months agoutil: change git-ver to only use a version tag to describe git version
Theodore Ts'o [Thu, 10 Aug 2023 01:56:07 +0000 (21:56 -0400)]
util: change git-ver to only use a version tag to describe git version

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
8 months agoMerge branch 'maint' into next
Theodore Ts'o [Tue, 8 Aug 2023 20:01:58 +0000 (16:01 -0400)]
Merge branch 'maint' into next

8 months agoe2fsprogs: modify dumpe2fs to report free block ranges for bigalloc
Eric Whitney [Fri, 21 Jul 2023 18:55:06 +0000 (14:55 -0400)]
e2fsprogs: modify dumpe2fs to report free block ranges for bigalloc

dumpe2fs has never been modified to correctly report block ranges
corresponding to free clusters in block allocation bitmaps from bigalloc
file systems.  Rather than reporting block ranges covering all the
blocks in free clusters found in a block bitmap, it either reports just
the first block number in a cluster for a single free cluster, or a
range beginning with the first block number in the first cluster in a
series of free clusters, and ending with the first block number in the
last cluster in that series.

This behavior causes xfstest shared/298 to fail when run on a bigalloc
file system with a 1k block size.  The test uses dumpe2fs to collect
a list of the blocks freed when files are deleted from a file system.
When the test deletes a file containing blocks located after the first
block in the last cluster in a series of clusters, dumpe2fs does not
report those blocks as free per the test's expectations.

Modify dumpe2fs to report full block ranges for free clusters.  At the
same time, fix a small bug causing unnecessary !in_use() retests while
iterating over a block bitmap.

Signed-off-by: Eric Whitney <enwlinux@gmail.com>
Link: https://lore.kernel.org/r/20230721185506.1020225-1-enwlinux@gmail.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
8 months agodebian: add missing copyright information
Theodore Ts'o [Tue, 8 Aug 2023 19:50:42 +0000 (15:50 -0400)]
debian: add missing copyright information

When the package-specific copyright information was removed, it
resulted in some information being lost.  It probably makes sence to
have all of the licensing information in a single file, so add it back
to the debian/copyright file.

Fixes: 76f2e8d11582 ("debian: remove package-specific copyright notices")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 months agoMerge branch 'maint' into next
Theodore Ts'o [Thu, 15 Jun 2023 04:29:41 +0000 (00:29 -0400)]
Merge branch 'maint' into next

10 months agoresize2fs: use Direct I/O when reading the superblock for online resizes
Theodore Ts'o [Thu, 15 Jun 2023 04:17:01 +0000 (00:17 -0400)]
resize2fs: use Direct I/O when reading the superblock for online resizes

If the file system is mounted, the superblock can be changing while
resize2fs is trying to read the superblock, resulting in checksum
failures.  One way of avoiding this problem is read the superblock
using Direct I/O, since the kernel makes sure that what gets written
to disk is self-consistent.

Suggested-by: Krister Johansen <kjlx@templeofstupid.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 months agotests: add test for handling an invalid symlink in an inline directory
Theodore Ts'o [Wed, 14 Jun 2023 19:15:48 +0000 (15:15 -0400)]
tests: add test for handling an invalid symlink in an inline directory

Add a test for the commit "e2fsck: fix handling of a invalid symlink
in an inline_data directory"

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 months agoe2fsck: fix handling of a invalid symlink in an inline_data directory
Theodore Ts'o [Wed, 14 Jun 2023 18:44:19 +0000 (14:44 -0400)]
e2fsck: fix handling of a invalid symlink in an inline_data directory

If there is an inline directory that contains a directory entry to an
invalid symlink, and that invalid symlink is the portion of the inline
directory stored in an xattr portion of the inode, this can result in
a buffer overrun.

When check_dir_block() is handling the in-xattr portion of the inline
directory, it sets the buf pointer to the beginning of that part of
the inline directory.  This results in the scratch buffer passed to
e2fsck_process_bad_inode() to incorrect, resulting in a buffer overrun
if e2fsck_pass1_check_symlink() needs to read the symlink target (when
the symlink is too long to fit in the i_blocks[] space).

This commit fixes this by using the original cd->buf instead of buf,
since it can get modified when handling inline directories.

Fixes: 0ac4b3973f31 ("e2fsck: inspect inline dir data as two directory blocks")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 months agoMerge branch 'maint' into next
Theodore Ts'o [Wed, 31 May 2023 16:45:01 +0000 (12:45 -0400)]
Merge branch 'maint' into next

10 months agolibsupport: fix function prototype for quota_write_inode()
Theodore Ts'o [Wed, 31 May 2023 16:11:53 +0000 (12:11 -0400)]
libsupport: fix function prototype for quota_write_inode()

Commit 2d2d799c7261 ("Clean up codes for adding new quota type")
changed the second paramter of quota_write_inode() from taking a
single quota type to taking a logical OR of (1 << quota_types).

The one thing this commit didn't change was the function prototype for
quota_write_inode() in the header file from an enum to an unsigned
int.  Most C compilers don't seem to mind, and omission is mostly
harmless.  However, mingw64 does issue a warning which gets promoted
to an error.

Fixes: 2d2d799c7261 ("Clean up codes for adding new quota type")
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 months agoci.yml: test cross-compiling for Android
Eric Biggers [Wed, 22 Mar 2023 03:29:45 +0000 (20:29 -0700)]
ci.yml: test cross-compiling for Android

Add jobs that cross-compile e2fsprogs for Android using the Android NDK.
These use the autotools-based build system, so they're a bit different
from the actual Android builds, but they should still be useful.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 months agoe2freefrag: don't use linux/fsmap.h when fsmap_sizeof() is missing
Eric Biggers [Wed, 22 Mar 2023 03:29:44 +0000 (20:29 -0700)]
e2freefrag: don't use linux/fsmap.h when fsmap_sizeof() is missing

Work around an issue with the Android NDK where its copy of
linux/fsmap.h is missing the inline functions fsmap_sizeof() and
fsmap_advance().  This was causing an error when building e2fsprogs
using the Android NDK, using the autotools-based build system.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 months agoe2fsck: avoid -Wtautological-constant-out-of-range-compare warnings
Eric Biggers [Wed, 22 Mar 2023 03:29:43 +0000 (20:29 -0700)]
e2fsck: avoid -Wtautological-constant-out-of-range-compare warnings

Fix two compiler warnings on 32-bit platforms that have mallinfo() but
not mallinfo2().  These showed up when building e2fsprogs for armv7a or
i686 Android using the Android NDK, targeting Android API level 32 or
lower and using the autotools-based build system.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 months agoe2fsck: Suppress "orphan file is clean" message in preen mode
Theodore Ts'o [Tue, 30 May 2023 17:50:52 +0000 (13:50 -0400)]
e2fsck: Suppress "orphan file is clean" message in preen mode

The e2fsck report, "Feature orphan_present is set but orphan file is
clean" is intended to request permission before removing the r/o
compat feature, orphan_present.  However, it is normal if the orphan
file is empty, and removing the r/o compat feature is a good thing so
that the file system can be mounted on older kernels.

When a file system with an orphan_file feature is mounted, the
orphan_present feature is set, and it is cleared when the file system
is cleanly unmounted.  IF the sytstem crashes when there are no inodes
in the orphan file, e2fsck should just silently clear the flag in
preen mode.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
12 months agoblkidP.h: add missing extern "C" declaration
Theodore Ts'o [Fri, 14 Apr 2023 15:41:25 +0000 (11:41 -0400)]
blkidP.h: add missing extern "C" declaration

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 months agotests: add test for "e2fsck: fix bad htree checksums in preen mode"
Theodore Ts'o [Fri, 17 Mar 2023 03:29:28 +0000 (23:29 -0400)]
tests: add test for "e2fsck: fix bad htree checksums in preen mode"

Add a test for commit bbe08adac044 ("e2fsck: fix bad htree checksums
in preen mode").

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 months agoe2fsck: restructure code to reduce indentation level in check_dir_block()
Theodore Ts'o [Fri, 17 Mar 2023 03:17:24 +0000 (23:17 -0400)]
e2fsck: restructure code to reduce indentation level in check_dir_block()

No functional changes; just move things around so we can avoid
indenting the code quite so much.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 months agoe2fsck: fix bad htree checksums in preen mode
Theodore Ts'o [Fri, 17 Mar 2023 02:57:10 +0000 (22:57 -0400)]
e2fsck: fix bad htree checksums in preen mode

We attempt to fix directories which have a bad/corrupted htree index
node by completely rebuilding the directory htree nodes.  Since this
is a very safe thing to do and has no risk of losing directory
entries, we've enabled this for preen mode.  Unfortunately, subsequent
index nodes look like empty directory entries that fill the entire
block --- without a checksum at the end of the directory.  So these
nodes will be treated as a completely corrupted directory block, and
this will *not* be fixed while in preen mode.

So add code to treat an empty directory entry which covers the entire
block as valid if the directory is already on the list of inodes to be
rebuilt.

Addresses-Gooogle-Bug: 178607853
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 months agoci.yml: store the config.h files as workflow artifacts
Eric Biggers [Wed, 8 Feb 2023 06:58:58 +0000 (22:58 -0800)]
ci.yml: store the config.h files as workflow artifacts

Store the config.h file for each platform as a workflow artifact, so
that it will be possible to download them and compare them to
util/android_config.h.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 months agotests: fix r_move_itable_realloc to run on Linux
Andreas Dilger [Tue, 10 Jan 2023 04:02:52 +0000 (21:02 -0700)]
tests: fix r_move_itable_realloc to run on Linux

The check for the various unsupported OSes incorrectly checked if
the string "FreeBSD" was true, which it always was.  Fix this.

Update the expect file as commit v1.46.4-17-g4ea80d031c7e did to
adjust the total number of blocks requested during resize.

Change-Id: I272dbec67ab30bac6413eb4cba0e3ab00183b893
Fixes: 5a3ea3905f ("tests: force test file systems to be built for Linux OS")
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
13 months agotests: fix u_direct_io to work with older losetup
Andreas Dilger [Tue, 10 Jan 2023 04:05:35 +0000 (21:05 -0700)]
tests: fix u_direct_io to work with older losetup

Older losetup does not have --sector-size, but this isn't really
needed for the test to work.  Instead specify the filesystem block
size directly to mke2fs, so that it works on all distros instead
of being skipped.

Change-Id: I5a0c82a9efdefd1b48f4d4288998c7725c9ae71e
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agoUpdate release notes, etc., for the 1.47.0 release
Theodore Ts'o [Sun, 5 Feb 2023 20:09:24 +0000 (15:09 -0500)]
Update release notes, etc., for the 1.47.0 release

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agodebian: update to standards version 4.6.2
Theodore Ts'o [Tue, 7 Feb 2023 03:06:22 +0000 (22:06 -0500)]
debian: update to standards version 4.6.2

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agolib/ext2fs: fix unbalanced mutex unlock for BOUNCE_MTX in unix_io
Ritesh Harjani (IBM) [Mon, 7 Nov 2022 12:20:49 +0000 (17:50 +0530)]
lib/ext2fs: fix unbalanced mutex unlock for BOUNCE_MTX in unix_io

f_crashdisk test failed with UNIX_IO_FORCE_BOUNCE=yes due to unbalanced
mutex unlock in below path.

This patch fixes it.

Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agoMerge branch 'maint' into next
Theodore Ts'o [Sun, 5 Feb 2023 14:31:36 +0000 (09:31 -0500)]
Merge branch 'maint' into next

14 months agoFix date in the release notes for v1.46.6
Theodore Ts'o [Thu, 2 Feb 2023 20:38:07 +0000 (15:38 -0500)]
Fix date in the release notes for v1.46.6

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agoUpdate e2fsprogs.lsm for 1.46.6 release
Theodore Ts'o [Thu, 2 Feb 2023 15:57:32 +0000 (10:57 -0500)]
Update e2fsprogs.lsm for 1.46.6 release

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agodebian: remove package-specific copyright notices
Theodore Ts'o [Thu, 2 Feb 2023 15:54:20 +0000 (10:54 -0500)]
debian: remove package-specific copyright notices

These are causing a large number of Lintian warnings
"file-without-copyright-information".

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agoMerge branch 'maint' into next
Theodore Ts'o [Thu, 2 Feb 2023 06:11:51 +0000 (01:11 -0500)]
Merge branch 'maint' into next

14 months agoUpdate release notes, etc., for the 1.46.6 release
Theodore Ts'o [Thu, 2 Feb 2023 05:39:32 +0000 (00:39 -0500)]
Update release notes, etc., for the 1.46.6 release

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agoconfig: update config.{guess,sub}
Theodore Ts'o [Thu, 2 Feb 2023 06:03:08 +0000 (01:03 -0500)]
config: update config.{guess,sub}

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agopo: update zh_CN.po (from translationproject.org)
Wenbin Lv [Thu, 2 Feb 2023 05:40:49 +0000 (00:40 -0500)]
po: update zh_CN.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agopo: update uk.po (from translationproject.org)
Yuri Chornoivan [Thu, 2 Feb 2023 05:40:49 +0000 (00:40 -0500)]
po: update uk.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agopo: update sv.po (from translationproject.org)
Göran Uddeborg [Thu, 2 Feb 2023 05:40:49 +0000 (00:40 -0500)]
po: update sv.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agopo: update sr.po (from translationproject.org)
Мирослав Николић [Thu, 2 Feb 2023 05:40:49 +0000 (00:40 -0500)]
po: update sr.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agopo: update pl.po (from translationproject.org)
Jakub Bogusz [Thu, 2 Feb 2023 05:40:49 +0000 (00:40 -0500)]
po: update pl.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agopo: update nl.po (from translationproject.org)
Benno Schulenberg [Thu, 2 Feb 2023 05:40:48 +0000 (00:40 -0500)]
po: update nl.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agopo: update ms.po (from translationproject.org)
Sharuzzaman Ahmat Raslan [Thu, 2 Feb 2023 05:40:48 +0000 (00:40 -0500)]
po: update ms.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agopo: update fr.po (from translationproject.org)
Samuel Thibault [Thu, 2 Feb 2023 05:40:48 +0000 (00:40 -0500)]
po: update fr.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agopo: update es.po (from translationproject.org)
Antonio Ceballos [Thu, 2 Feb 2023 05:40:48 +0000 (00:40 -0500)]
po: update es.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agopo: update de.po (from translationproject.org)
Mario Blättermann [Thu, 2 Feb 2023 05:40:48 +0000 (00:40 -0500)]
po: update de.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agopo: update cs.po (from translationproject.org)
Petr Pisar [Thu, 2 Feb 2023 05:40:48 +0000 (00:40 -0500)]
po: update cs.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agobadblocks: fix operation with large-ish block sizes and/or counts
Corey Hickey [Mon, 24 Jan 2022 02:53:13 +0000 (18:53 -0800)]
badblocks: fix operation with large-ish block sizes and/or counts

test_rw() and test_nd() need to allocate two or three times the product
of the block size and the block counts. This can overflow the signed int
type of block_size and result in allocate_buffer() being called with a
value smaller than intended. Once that buffer is written to, badblocks
segfaults.

Since allocate_buffer() accepts a size_t, change the input validation to
use SIZE_MAX and cast accordingly when calculating the argument.

Fixing the segfault allows larger values to be passed to read() and
write(); these need to be cast to size_t as well in order to avoid a
signed integer overflow causing failure, in which case badblocks would
fall back to testing a single block at once.

Before:
$ misc/badblocks -w -b 4096 -c 524288 -e 1 -s -v /tmp/testfile.bin
Checking for bad blocks in read-write mode
From block 0 to 524287
Segmentation fault

$ misc/badblocks -n -b 4096 -c 524288 -e 1 -s -v /tmp/testfile.bin
Checking for bad blocks in non-destructive read-write mode
From block 0 to 524287
Checking for bad blocks (non-destructive read-write test)
Segmentation fault

After:
$ misc/badblocks -w -b 4096 -c 524288 -e 1 -s -v /tmp/testfile.bin
Checking for bad blocks in read-write mode
From block 0 to 524287
Testing with pattern 0xaa: done
Reading and comparing: done
Testing with pattern 0x55: done
Reading and comparing: done
Testing with pattern 0xff: done
Reading and comparing: done
Testing with pattern 0x00: done
Reading and comparing: done
Pass completed, 0 bad blocks found. (0/0/0 errors)

$ misc/badblocks -n -b 4096 -c 524288 -e 1 -s -v /tmp/testfile.bin
Checking for bad blocks in non-destructive read-write mode
From block 0 to 524287
Checking for bad blocks (non-destructive read-write test)
Testing with random pattern: done
Pass completed, 0 bad blocks found. (0/0/0 errors)

Signed-off-by: Corey Hickey <bugfood-c@fatooh.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agobadblocks: separate and improve error messages for blocks_at_once
Corey Hickey [Mon, 24 Jan 2022 02:33:22 +0000 (18:33 -0800)]
badblocks: separate and improve error messages for blocks_at_once

Since the conditional checks the product of block_size and
blocks_at_once, reporting that the problem is solely with
blocks_at_once is misleading.

Also change the error to use the name of the parameter listed in the
manual rather than the variable name.

Since blocks_at_once is unsigned, change the test to == rather than <=.

Before:
$ misc/badblocks -w -b 16777216 -c 524288 -e 1 -s -v /tmp/testfile.bin
misc/badblocks: Invalid blocks_at_once: 524288

After:
$ misc/badblocks -w -b 16777216 -c 524288 -e 1 -s -v /tmp/testfile.bin
misc/badblocks: For block size 16777216, blocks_at_once too large: 524288

$ misc/badblocks -w -b 16777216 -c 0 -e 1 -s -v /tmp/testfile.bin
misc/badblocks: Invalid number of blocks: 0

Signed-off-by: Corey Hickey <bugfood-c@fatooh.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agobadblocks: fix mis-printed error from block size check
Corey Hickey [Mon, 24 Jan 2022 01:39:33 +0000 (17:39 -0800)]
badblocks: fix mis-printed error from block size check

block_size is parsed as an unsigned int from parse_uint(), so retain it
as such until _after_ it has been constrained to a size within INT_MAX.

Lower level code still requires this to be an int, so cast to int for
anything below main().

Before:
$ misc/badblocks -w -b 4294967295 -c 1 /tmp/testfile.bin
misc/badblocks: Invalid block size: -1

After:
$ misc/badblocks -w -b 4294967295 -c 1 /tmp/testfile.bin
misc/badblocks: Invalid block size: 4294967295

Signed-off-by: Corey Hickey <bugfood-c@fatooh.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agobadblocks: print a more explanatory message when a parameter is too large
Corey Hickey [Mon, 24 Jan 2022 01:23:39 +0000 (17:23 -0800)]
badblocks: print a more explanatory message when a parameter is too large

Before:
$ misc/badblocks -w -b 4294967296 -c 1 /tmp/testfile.bin
misc/badblocks: invalid block size - 4294967296

After:
$ misc/badblocks -w -b 4294967296 -c 1 /tmp/testfile.bin
misc/badblocks: block size too large - 4294967296

The original error is retained for invalid arguments, e.g.:
$ misc/badblocks -w -b foo -c 1 /tmp/testfile.bin
misc/badblocks: invalid block size - foo

Signed-off-by: Corey Hickey <bugfood-c@fatooh.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agofuse2fs: support "fuse2fs -o offset=<bytes>"
Matt Stark [Mon, 17 Oct 2022 05:41:57 +0000 (16:41 +1100)]
fuse2fs: support "fuse2fs -o offset=<bytes>"

This works the same way that mount -o offset=<bytes> works, and can be
used to mount particular partitions from a whole disk image.

Signed-off-by: Matt Stark <msta@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
14 months agoext2fs: Use 64bit lseek when _FILE_OFFSET_BITS is 64
Khem Raj [Thu, 15 Dec 2022 04:56:44 +0000 (20:56 -0800)]
ext2fs: Use 64bit lseek when _FILE_OFFSET_BITS is 64

Use lseek() with 64bit off_t when _FILE_OFFSET_BITS is 64
this fixes build with musl where there is no _llseek but lseek
is using off_t which is 64bit on musl

Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>