Whamcloud - gitweb
tools/e2fsprogs.git
4 years agoMerge branch 'maint' into next
Theodore Ts'o [Sat, 23 Nov 2019 04:30:37 +0000 (23:30 -0500)]
Merge branch 'maint' into next

4 years agoAdd a program to test images provided by UBSAN fuzzing reports
Theodore Ts'o [Sat, 23 Nov 2019 04:27:47 +0000 (23:27 -0500)]
Add a program to test images provided by UBSAN fuzzing reports

This program calls a few ext2fs library functions used by the current
generation of libext2fs fuzzers, and is helpful in reproducing UBSAN
failures reported externally.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agolibext2fs: define PATH_MAX if not provided by the system headers
Theodore Ts'o [Sat, 23 Nov 2019 04:26:39 +0000 (23:26 -0500)]
libext2fs: define PATH_MAX if not provided by the system headers

This is needed to compile on Illumos and its derivatives.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agotune2fs: prohibit toggling uninit_bg on live filesystems
Darrick J. Wong [Wed, 20 Nov 2019 19:32:55 +0000 (11:32 -0800)]
tune2fs: prohibit toggling uninit_bg on live filesystems

An internal customer followed an erroneous AskUbuntu article[1] to try to
change the UUID of a live ext4 filesystem.  The article claims that you
can work around tune2fs' "cannot change UUID on live fs" error by
disabling uninit_bg, changing the UUID, and re-enabling the feature.

This led to metadata corruption because tune2fs' journal descriptor
rewrite races with regular filesystem writes.  Therefore, prevent
administrators from turning on or off uninit_bg on a mounted fs.

[1] https://askubuntu.com/questions/132079/how-do-i-change-uuid-of-a-disk-to-whatever-i-want/195839#459097

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agochattr.1: say "cleared" instead of "reset"
Eric Biggers [Mon, 18 Nov 2019 01:48:52 +0000 (17:48 -0800)]
chattr.1: say "cleared" instead of "reset"

Setting a bit to 0 is normally called "clearing", not "resetting"; and
chattr.1 already says "clear" in some places.  Use "clear" consistently.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agochattr.1: clarify that ext4 doesn't support tail-merging either
Eric Biggers [Mon, 18 Nov 2019 01:48:51 +0000 (17:48 -0800)]
chattr.1: clarify that ext4 doesn't support tail-merging either

This old text was never updated to mention ext4 in addition to ext2 and
ext3.  Do so now.  Also don't bother to mention old unmerged patches.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agochattr.1: fix some grammatical errors
Eric Biggers [Mon, 18 Nov 2019 01:48:50 +0000 (17:48 -0800)]
chattr.1: fix some grammatical errors

- "can only be open" => "can only be opened"
- "is not candidate" => "is not a candidate"
- "written ... on the disk" => "written ... to the disk"

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agochattr.1: add casefold attribute to mode string
Eric Biggers [Mon, 18 Nov 2019 01:48:49 +0000 (17:48 -0800)]
chattr.1: add casefold attribute to mode string

When the casefold attribute ('F') was added to the chattr man page, it
was forgotten to add it to the mode string.  Add it.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agochattr.1: adjust documentation for encryption attribute
Eric Biggers [Mon, 18 Nov 2019 01:48:48 +0000 (17:48 -0800)]
chattr.1: adjust documentation for encryption attribute

Adjust the documentation for the encryption attribute ('E') to clarify
that encryption isn't experimental anymore and isn't restricted to
regular files, and that the encryption is done by the filesystem.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agochattr.1: document the verity attribute
Eric Biggers [Mon, 18 Nov 2019 01:48:47 +0000 (17:48 -0800)]
chattr.1: document the verity attribute

Document the verity file attribute ('V').

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoMerge branch 'maint' into next
Theodore Ts'o [Thu, 14 Nov 2019 16:46:13 +0000 (11:46 -0500)]
Merge branch 'maint' into next

4 years agoconfigure: don't fail if /usr/include/sys/mount.h does not exist
Theodore Ts'o [Wed, 13 Nov 2019 18:55:29 +0000 (13:55 -0500)]
configure: don't fail if /usr/include/sys/mount.h does not exist

The AX_CHECK_MOUNT_OPT macro fails if /usr/include/sys/mount.h.
Unfortunately, the GNU HURD doesn't have this header file.  Drop the
requirement, since if it doesn't exist, the macro will assume that
given mount options don't exist, which for e2fsprogs is used to test
to for the existence of the mount options nosuid and nodev.  This is
only used for fuse2fs, and HURD doesn't support fuse2fs anyway.

Addresses-Debian-Bug: #944649
Reported-by: Svante Signell <svante.signell@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoRevert "e2fsck: Change kmem_cache_t to lkmem_cache_t for Solaris"
Theodore Ts'o [Tue, 12 Nov 2019 15:47:37 +0000 (10:47 -0500)]
Revert "e2fsck: Change kmem_cache_t to lkmem_cache_t for Solaris"

This is a logical revert of commit 1911bf113ef0, for which the
description reads:

    Solaris polutes the C namespace with kmem_cache_t when
    you include in/netinet.h is included, so rename kmem_cache_t
    to lkmem_cache_t.

Reverting this change allows us to keep e2fsck/revoke.c in sync with
its upstream kernel source of fs/jbd2/revoke.c, and was the last
change required to make the e2fsprogs and kernel versions of revoke.c
to be bit identical.

I've confirmed that this is no longer a problem with OmniOS (an
Illumos / Open Solaris derivative).  It may be a problem with Solaris,
but since I don't have easy access to Solaris, ¯\_(ツ)_/¯

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agolibext2fs/ismounted.c: check device id in advance to skip false device names
Theodore Ts'o [Sun, 10 Nov 2019 17:11:49 +0000 (12:11 -0500)]
libext2fs/ismounted.c: check device id in advance to skip false device names

If there is a trickster which tries to use device names as the mount
device for pseudo-file systems, the resulting /proc/mounts can confuse
ext2fs_check_mount_point().  (So far as I can tell, there's no good
reason to do this, but sysadmins do the darnest things.)

An example of this might be the following /proc/mounts excerpt:

/dev/sdb /mnt2 tmpfs rw,relatime 0 0
/dev/sdb /mnt ext4 rw,relatime 0 0

This is created via "mount -t tmpfs /dev/sdb /mnt2" followed via
"mount -t ext4 /dev/sdb /mnt".  (Normally, a sane mount of tmpfs would
use something like "mount -t tmpfs tmpfs /mnt2".)

Fix this by double checking the st_rdev of the claimed mountpoint and
match it with the dev_t of the device.  (Note that the GNU HURD
doesn't support st_rdev, so we can't solve this problem for the HURD.)

Reported-by: GuiYao <guiyao@huawei.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoresize2fs: make minimum size estimates more reliable for mounted fs
Jan Kara [Fri, 18 Oct 2019 12:50:59 +0000 (14:50 +0200)]
resize2fs: make minimum size estimates more reliable for mounted fs

Currently, the estimate of minimum filesystem size is using free blocks
counter in the superblock. The counter generally doesn't get updated
while the filesystem is mounted and thus the estimate is very unreliable
for a mounted filesystem. For some usecases such as automated
partitioning proposal to the user it is desirable that the estimate of
minimum filesystem size is reasonably accurate even for a mounted
filesystem. So use group descriptor counters of free blocks for the
estimate of minimum filesystem size. These get updated together with
block being allocated and so the resulting estimate is more accurate.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoDefine the beXX_to_cpu and cpu_to_beXX macros for e2fsck/{recovery.c,revoke.c}
Theodore Ts'o [Sun, 10 Nov 2019 04:28:43 +0000 (23:28 -0500)]
Define the beXX_to_cpu and cpu_to_beXX macros for e2fsck/{recovery.c,revoke.c}

We were previously using contrib/jbd2-resync.sh to transmogrify the
beXX_to_cpu and cpu_to_beXX macros to ext2fs_beXX_to_cpu and
ext2fs_cpu_to_beXX.  Define them in lib/ext2fs/jfs_compat.h so we can
more easily keep them in sync with the kernel version of those files.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2scrub_all: fix broken stdin redirection
Darrick J. Wong [Tue, 5 Nov 2019 01:54:20 +0000 (17:54 -0800)]
e2scrub_all: fix broken stdin redirection

gregor herrmann reports that the weekly e2scrub cronjob emits these
errors:

/sbin/e2scrub_all: line 173: /proc/8234/fd/pipe:[90083173]: No such file or directory

The root cause of this is that the ls_targets stdout is piped to stdin
to the entire ls_targets loop body to prevent the loop body from reading
the loop iteration items.  Remove all the broken hackery by reading the
target list into a bash array and iterating the bash array.

Addresses-Debian-Bug: #944033

Reported-by: gregor herrmann <gregoa@debian.org>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2scrub_all: don't even reap if the config file doesn't allow it
Darrick J. Wong [Tue, 5 Nov 2019 01:54:14 +0000 (17:54 -0800)]
e2scrub_all: don't even reap if the config file doesn't allow it

Dave Chinner complains that the automated on-boot e2scrub reaping takes
a long time (because the lvs command can take a while to run) even
though the automated e2scrub is disabled via e2scrub.conf on his
systems.

We still need the reaping service to kill off stale e2scrub snapshots
after a crash, but it's unnecessary to annoy everyone with slow bootup.
Because we can look for the e2scrub snapshots in /dev/mapper, let's
skip reaping if periodic e2scrub is disabled unless we find evidence of
e2scrub snapshots in /dev.

Reported-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoAlign the types used by jbd2_journal_bmap and getblk with the kernel
Theodore Ts'o [Sat, 9 Nov 2019 06:47:24 +0000 (01:47 -0500)]
Align the types used by jbd2_journal_bmap and getblk with the kernel

This avoids some 32-bit vs 64-bit discrepancies in the function
signatures and the types used by their callers.

This cleans up some sparse warnings in recovery.c.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2fsck/recovery.c: sync up with kernel's use of __be32
Theodore Ts'o [Sat, 9 Nov 2019 06:42:59 +0000 (01:42 -0500)]
e2fsck/recovery.c: sync up with kernel's use of __be32

E2fsprogs as a whole is not sparse-clean, but it does have and
understand the __beXX and __leXX types from the kernel.  The structure
definitions in kernel-jbd.h have been updated to use the __beXX types,
so that recovery.c and revoke.c are more sparse-clean.

This removes a few more unneeded deltas from the kernel's recovery.c.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoSync kernel's removal of open-coded allocation retry loop in revoke.c
Theodore Ts'o [Sat, 9 Nov 2019 03:47:31 +0000 (22:47 -0500)]
Sync kernel's removal of open-coded allocation retry loop in revoke.c

Apply the kernel's changes for commit 7b506b103532 ("jbd2: get rid of
open coded allocation retry loop") for revoke.c.

This required adjusting some of kernel compatibility defines.

Note that retrying allocations in user space never makes any sense!

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoSync kernel's fix for potential double free in jbd2
Theodore Ts'o [Sat, 9 Nov 2019 02:53:42 +0000 (21:53 -0500)]
Sync kernel's fix for potential double free in jbd2

Commit 0d52154bb0a7 ("jbd2: fix potential double free") changes the
interface exported by revoke.c to initialize and destroy the slab
caches.  Make the necessary changes to the code in e2fsck and debugfs
which calls revoke.c

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2fsck/revoke.c: sync kernel's adoption of kmalloc_array()
Theodore Ts'o [Sat, 9 Nov 2019 02:25:59 +0000 (21:25 -0500)]
e2fsck/revoke.c: sync kernel's adoption of kmalloc_array()

Sync the changes to e2fsck/revoke.c from commit 6da2ec56059c
("treewide: kmalloc() -> kmalloc_array()"), and add the emulation of
kmalloc_array() to e2fsck/jfs_user.h

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoSync kernel's unification of jbd2 revoke and tag block checksum handling
Theodore Ts'o [Sat, 9 Nov 2019 01:26:05 +0000 (20:26 -0500)]
Sync kernel's unification of jbd2 revoke and tag block checksum handling

Commit 1101cd4d13ba ("jbd2: unify revoke and tag block checksum
handling") cleans up the fact that the jbd2_journal_revoke_tail and
jbd2_journal_block_tail structures are basically the same.  So it
drops the definition of struct jbd2_journal_revoke_tail and unifies
the functions which calculates and verifies the checksums for revoke
blocks and tag blocks.

Make the same changes in e2fsprogs so eliminate unnecessary
differences in e2fsck/recovery.c and e2fsck/revoke.c.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2fsck/revoke.c: sync changes from kernel
Theodore Ts'o [Sat, 9 Nov 2019 00:00:30 +0000 (19:00 -0500)]
e2fsck/revoke.c: sync changes from kernel

Sync up the revoke.c specific changes from kernel commits 9bcf976cb8b8
("jbd2: remove unnecessary arguments of jbd2_journal_write_revoke_records"),
32ab671599a8 ("jbd2: factor out common descriptor block initialization"),
70fd76140a6c ("block,fs: use REQ_* flags directly"), cd9cb405e0b9
("jbd2: don't leak memory if setting up journal fails"), 8bdd5b60e027
("jbd2: remove NULL check before calling kmem_cache_destroy()"),
547b9ad698b4 ("jbd2: flush_descriptor(): Do not decrease buffer head's
ref count"), and fdc3ef882a5d ("jbd2: Reserve space for revoke descriptor
blocks").

Nearly all of the changes is in code under an #ifdef __KERNEL__.  The
changes that will actually affect e2fprogs compilation are trivial and
easy to hand verify.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoRename functions, types, constants to reflect jbd2 usage
Theodore Ts'o [Fri, 8 Nov 2019 22:17:35 +0000 (17:17 -0500)]
Rename functions, types, constants to reflect jbd2 usage

We had previously stuck to using the names from ext3/jbd kernel files,
and used a script in contrib/jbd2-resync.sh to convert the kernel
files to use the ext3/jbd conventions so we could keep the files
e2fsck/recovery.c and e2fsck/revoke.c in sync with jbd2/recovery.c and
jbd2/revoke.c, respectively.

This has been getting harder and harder, so let's make a global sweep
through e2fsprogs to use the jbd2 names.  Fortunately none of the
ext3/jbd names had leaked out into publically exported header files,
so this is only an internal change.  Which looks scary, but it's
basically a search and replace, so if it compiles it's going to be
correct.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoConvert kernel compat functions to use new ll_rw_block() function signature
Theodore Ts'o [Fri, 8 Nov 2019 21:03:38 +0000 (16:03 -0500)]
Convert kernel compat functions to use new ll_rw_block() function signature

In newer kernels, ll_rw_block() separated the request operation and
the operational flags arguments.  This means adding a new parameter to
ll_rw_block() (which is ignored in our compat layer) and changing READ
and WRITE to REQ_OP_READ and REQ_OP_WRITE, respectively.

This makes it easier to keep us in sync with the kernel tree.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoClean up minor differences between kernel and e2fsck's jbd2 source files
Theodore Ts'o [Fri, 8 Nov 2019 20:02:42 +0000 (15:02 -0500)]
Clean up minor differences between kernel and e2fsck's jbd2 source files

Historically e2fsprogs's e2fsck/recovery.c and e2fsck/revoke.c was
sync'ed against the ext3 version of jbd/recovery.c and jbd/revoke.c.
Remove minor differences so we can better sync up between the two
versions.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoMerge branch 'maint' into next
Theodore Ts'o [Sat, 9 Nov 2019 06:54:31 +0000 (01:54 -0500)]
Merge branch 'maint' into next

4 years agolibext2fs: verify the block numbers for the allocation bitmaps are valid
Theodore Ts'o [Fri, 8 Nov 2019 16:58:10 +0000 (11:58 -0500)]
libext2fs: verify the block numbers for the allocation bitmaps are valid

This avoids a potential UBsan failure when we multiply an insanely
high block number with the block size and we get a multiplication
overflow.

Google-Bug-Id: 128130353
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agolibext2fs: properly free the bitmaps in read_bitmap()'s error/cleanup path
Theodore Ts'o [Fri, 8 Nov 2019 14:49:51 +0000 (09:49 -0500)]
libext2fs: properly free the bitmaps in read_bitmap()'s error/cleanup path

Use ext2fs_free_{block,inode}_bitmaps() instead of ext2fs_free_mem()
to avoid some memory leaks if there is an error while calling
read_bitmaps().

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebugfs: teach the htree command to check and display the dx block checksum
Theodore Ts'o [Thu, 7 Nov 2019 14:01:23 +0000 (09:01 -0500)]
debugfs: teach the htree command to check and display the dx block checksum

To do this we need to export the ext2fs_dx_csum() function from
libext2fs.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agolibext2fs: refactor common code into ext2fs_dx_csum()
Theodore Ts'o [Thu, 7 Nov 2019 12:17:13 +0000 (07:17 -0500)]
libext2fs: refactor common code into ext2fs_dx_csum()

ext2fs_dx_csum() is currently a static function, so this has no
visible changes except shrinking the text size of csum.o by
approximately 100 bytes.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoMerge branch 'maint' into next
Theodore Ts'o [Thu, 7 Nov 2019 03:33:11 +0000 (22:33 -0500)]
Merge branch 'maint' into next

4 years agoFix UBSan when shifting (1LL << 63)
Theodore Ts'o [Tue, 5 Nov 2019 02:22:54 +0000 (21:22 -0500)]
Fix UBSan when shifting (1LL << 63)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agolibext2fs: fix UBSan when updating an inline_data file
Theodore Ts'o [Mon, 4 Nov 2019 23:43:49 +0000 (18:43 -0500)]
libext2fs: fix UBSan when updating an inline_data file

What memcpy does when the length is zero is not well-defined.  So
avoid it.

Bug: https://github.com/tytso/e2fsprogs/issues/25
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agolibext2fs: fix bug when reading or writing more than 2GB in unix_io
Theodore Ts'o [Mon, 4 Nov 2019 21:43:41 +0000 (16:43 -0500)]
libext2fs: fix bug when reading or writing more than 2GB in unix_io

If count * block_size exceeds 2GB, we will overflow a 32-bit signed
integer value.  This shouldn't happen in practice except for
fuzz-corrupted file systems, but let's fix the code so it's correct.

Bug: https://github.com/tytso/e2fsprogs/issues/24
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoAOSP: support the stable_inodes feature
Eric Biggers [Mon, 21 Oct 2019 23:30:43 +0000 (16:30 -0700)]
AOSP: support the stable_inodes feature

Reserve the codepoint for EXT4_FEATURE_COMPAT_STABLE_INODES, allow it to
be set and cleared, and teach resize2fs to forbid shrinking the
filesystem if it is set.

This feature will allow the use of encryption policies where the inode
number is included in the IVs (initialization vectors) for encryption,
so data would be corrupted if the inodes were to be renumbered.

For more details, see the kernel patchset:
https://lkml.kernel.org/linux-fsdevel/20191021230355.23136-1-ebiggers@kernel.org/T/#u

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
From AOSP commit: 9aa30c254dd57df54f00c5d520b7ac867ad7ca68

4 years agoMerge branch 'maint' into next
Theodore Ts'o [Thu, 31 Oct 2019 23:30:54 +0000 (19:30 -0400)]
Merge branch 'maint' into next

4 years agoAOSP: Link to production releases, not testing releases
Eric Biggers [Tue, 24 Sep 2019 21:59:10 +0000 (14:59 -0700)]
AOSP: Link to production releases, not testing releases

The testing links are not stable and are not guaranteed to exist.

Test: After re-generating the generated files, the link works now.
Change-Id: I36e73ef74571b3246f470280a75ae1098245eff5
From AOSP commit: 5971f6a5113b12b33d9454229bf27621853e1da7

4 years agoAOSP: Add new source files to Android.bp files
Eric Biggers [Tue, 24 Sep 2019 21:59:10 +0000 (14:59 -0700)]
AOSP: Add new source files to Android.bp files

Test: see I3781b6d1e55923e9410644e8a7ba834b4d13b733
Change-Id: Ie071b30937dcf73f34df5dc3aa08d23ae30d9c63
From AOSP commit: d2a16e1616b807a6143b9c6232a1c54f90dc06f6

4 years agoTeach ext2fs_open2() to honor the EXT2_FLAG_SUPER_ONLY flag
Theodore Ts'o [Tue, 22 Oct 2019 22:42:25 +0000 (18:42 -0400)]
Teach ext2fs_open2() to honor the EXT2_FLAG_SUPER_ONLY flag

Opening the file system with EXT2_FLAG_SUPER_ONLY will leave
fs->group_desc to be NULL and modify "dumpe2fs -h" and tune2fs when it
is emulating e2label to use this flag.  This speeds up "dumpe2fs -h"
and "e2label" when operating on very large file systems.

To allow other libext2fs functions to work without too many surprises,
ext2fs_group_desc() will read in the block group descriptors on
demand.  This allows "dumpe2fs -h" to be able to read the journal
inode, for example.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cray-bug-id: LUS-5777

4 years agoUpdate debian changelog and version.h for 1.46~WIP.2019.10.09-1
Theodore Ts'o [Thu, 10 Oct 2019 00:25:01 +0000 (20:25 -0400)]
Update debian changelog and version.h for 1.46~WIP.2019.10.09-1

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2fsck: check for consistent encryption policies
Eric Biggers [Wed, 18 Sep 2019 01:07:34 +0000 (18:07 -0700)]
e2fsck: check for consistent encryption policies

By design, the kernel enforces that all files in an encrypted directory
use the same encryption policy as the directory.  It's not possible to
violate this constraint using syscalls.  Lookups of files that violate
this constraint also fail, in case the disk was manipulated.

But this constraint can also be violated by accidental filesystem
corruption.  E.g., a power cut when using ext4 without a journal might
leave new files without the encryption bit and/or xattr.  Thus, it's
important that e2fsck correct this condition.

Therefore, this patch makes the following changes to e2fsck:

- During pass 1 (inode table scan), create a map from inode number to
  encryption policy for all encrypted inodes.  But it's optimized so
  that the full xattrs aren't saved but rather only 32-bit "policy IDs",
  since usually many inodes share the same encryption policy.  Also, if
  an encryption xattr is missing, offer to clear the encrypt flag.  If
  an encryption xattr is clearly corrupt, offer to clear the inode.

- During pass 2 (directory structure check), use the map to verify that
  all regular files, directories, and symlinks in encrypted directories
  use the directory's encryption policy.  Offer to clear any directory
  entries for which this isn't the case.

Add a new test "f_bad_encryption" to test the new behavior.

Due to the new checks, it was also necessary to update the existing test
"f_short_encrypted_dirent" to add an encryption xattr to the test file,
since it was missing one before, which is now considered invalid.

Google-Bug-Id: 135138675
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
4 years agodebian: update changelog for 1.46~WIP.2019.10.03-1
Theodore Ts'o [Fri, 4 Oct 2019 01:51:14 +0000 (21:51 -0400)]
debian: update changelog for 1.46~WIP.2019.10.03-1

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoMark version as 1.46-WIP
Theodore Ts'o [Fri, 4 Oct 2019 01:47:52 +0000 (21:47 -0400)]
Mark version as 1.46-WIP

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoEnable support for the fast_commit compat feature
Theodore Ts'o [Fri, 4 Oct 2019 01:45:50 +0000 (21:45 -0400)]
Enable support for the fast_commit compat feature

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoMerge branch 'maint' into next
Theodore Ts'o [Fri, 4 Oct 2019 01:16:26 +0000 (21:16 -0400)]
Merge branch 'maint' into next

4 years agoDefine the codepoint for the fast_commit compat feature
Theodore Ts'o [Fri, 4 Oct 2019 01:14:31 +0000 (21:14 -0400)]
Define the codepoint for the fast_commit compat feature

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoMerge tag 'v1.45.4' into next
Theodore Ts'o [Mon, 23 Sep 2019 22:29:14 +0000 (18:29 -0400)]
Merge tag 'v1.45.4' into next

v1.45.4

4 years agoUpdate release notes, etc., for the 1.45.4 release v1.45.4
Theodore Ts'o [Mon, 23 Sep 2019 20:04:59 +0000 (16:04 -0400)]
Update release notes, etc., for the 1.45.4 release

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agotune2fs.8: tweak the documentation for the encrypt feature
Eric Biggers [Fri, 20 Sep 2019 21:29:54 +0000 (14:29 -0700)]
tune2fs.8: tweak the documentation for the encrypt feature

Try to make it clearer that enabling 'encrypt' just enables *support*
for encryption; it doesn't actually encrypt anything by itself.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agotune2fs.8: document the verity feature
Eric Biggers [Fri, 20 Sep 2019 21:29:53 +0000 (14:29 -0700)]
tune2fs.8: document the verity feature

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoext4.5: tweak the documentation for the encrypt feature
Eric Biggers [Fri, 20 Sep 2019 21:29:52 +0000 (14:29 -0700)]
ext4.5: tweak the documentation for the encrypt feature

Try to make it clearer that enabling 'encrypt' just enables *support*
for encryption; it doesn't actually encrypt anything by itself.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoext4.5: document the verity feature
Eric Biggers [Fri, 20 Sep 2019 21:29:51 +0000 (14:29 -0700)]
ext4.5: document the verity feature

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoext4.5: document first kernel version to support casefold feature
Eric Biggers [Fri, 20 Sep 2019 21:29:50 +0000 (14:29 -0700)]
ext4.5: document first kernel version to support casefold feature

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoext4.5: move casefold feature to correct position
Eric Biggers [Fri, 20 Sep 2019 21:29:49 +0000 (14:29 -0700)]
ext4.5: move casefold feature to correct position

The features are listed in alphabetic order, so put the casefold feature
in the right place.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2scrub_all: make sure fd 3 is closed before running lvm commands
Theodore Ts'o [Mon, 23 Sep 2019 17:17:13 +0000 (13:17 -0400)]
e2scrub_all: make sure fd 3 is closed before running lvm commands

Some versions of cron leave fd 3 open for some unknown reason.  So
when e2scrub_all is run by cron (on non-systemd systems) this results
in an annoying message from the Cron Daemon because lvm will print
warning messages about "leaked file descriptors.  So force close fd 3
at the beginning of e2scrub and e2scrub_all.

Addresses-Debian-Bug: #940240
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agopo: update vi.po (from translationproject.org)
Trần Ngọc Quân [Mon, 23 Sep 2019 17:10:37 +0000 (13:10 -0400)]
po: update vi.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agopo: update nl.po (from translationproject.org)
Benno Schulenberg [Mon, 23 Sep 2019 17:10:36 +0000 (13:10 -0400)]
po: update nl.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agopo: update ms.po (from translationproject.org)
Sharuzzaman Ahmat Raslan [Mon, 23 Sep 2019 17:10:36 +0000 (13:10 -0400)]
po: update ms.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agopo: update es.po (from translationproject.org)
Antonio Ceballos [Mon, 23 Sep 2019 17:10:36 +0000 (13:10 -0400)]
po: update es.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agopo: update de.po (from translationproject.org)
Mario Blättermann [Mon, 23 Sep 2019 17:10:36 +0000 (13:10 -0400)]
po: update de.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agolibsupport: add checks to prevent buffer overrun bugs in quota code
Theodore Ts'o [Sun, 1 Sep 2019 04:59:16 +0000 (00:59 -0400)]
libsupport: add checks to prevent buffer overrun bugs in quota code

A maliciously corrupted file systems can trigger buffer overruns in
the quota code used by e2fsck.  To fix this, add sanity checks to the
quota header fields as well as to block number references in the quota
tree.

Addresses: CVE-2019-5094
Addresses: TALOS-2019-0887
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoMerge branch 'maint' into next
Theodore Ts'o [Tue, 3 Sep 2019 14:39:23 +0000 (10:39 -0400)]
Merge branch 'maint' into next

4 years agoe2fsck: check the validity of the casefold flag
Theodore Ts'o [Tue, 3 Sep 2019 00:49:09 +0000 (20:49 -0400)]
e2fsck: check the validity of the casefold flag

The casefold flag is only allowed on directories and when the casefold
feature is enabled.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2fsck: make the low dtime check consistent when using the backup superblock
Theodore Ts'o [Mon, 2 Sep 2019 21:44:14 +0000 (17:44 -0400)]
e2fsck: make the low dtime check consistent when using the backup superblock

The backup superblock may have a last mounted time of zero, if it has
never been updated since the file system was created.  In that case,
the low dtime check may get disabled when using the backup superblock,
even though subsequent e2fsck runs will end up using the low dtime
check.  This can cause a failure of ext4/007, since since when e2fsck
is run a second time after the file system is mounted, the low dtime
check will trigger the e2fsck complaint:

Inode NNNN was part of the orphaned inode list.  IGNORED.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2fsck: update the quota records when the root directory is recreated
Theodore Ts'o [Sun, 1 Sep 2019 05:19:24 +0000 (01:19 -0400)]
e2fsck: update the quota records when the root directory is recreated

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agolibsupport: remove unnecessary code to set the quota inode size
Theodore Ts'o [Sun, 1 Sep 2019 05:17:42 +0000 (01:17 -0400)]
libsupport: remove unnecessary code to set the quota inode size

This is done correctly by the fileio routines; and it forces the quota
inode's file size to be unnecessarily rounded up to the block size.

Also remove the unnecessary double colon in the quota's error messages.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agotests: add tests for uninitialized bitmaps
Andreas Dilger [Thu, 12 Apr 2012 23:52:44 +0000 (17:52 -0600)]
tests: add tests for uninitialized bitmaps

Various tests for handing uninitialized block and inode bitmaps,
and inodes beyond the in-use high watermark.

Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agotests: verify > 65000 subdirectories
Andreas Dilger [Thu, 12 Apr 2012 22:02:12 +0000 (16:02 -0600)]
tests: verify > 65000 subdirectories

Add test case to verify nlink handling of large directories.

Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoMerge branch 'maint' into next
Theodore Ts'o [Mon, 19 Aug 2019 00:58:18 +0000 (20:58 -0400)]
Merge branch 'maint' into next

4 years agofuse2fs: add a norecovery option which suppresses journal replay
Theodore Ts'o [Mon, 19 Aug 2019 00:25:53 +0000 (20:25 -0400)]
fuse2fs: add a norecovery option which suppresses journal replay

Teach fuse2fs the "-o norecovery" option, which will suppress any
journal replay that might be necessary, and mounts the file system
read-only.

Addresses-Debian-Bug: #878927

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2scrub_all: allow scrubbing in vg's whose free space == snapshot size
Theodore Ts'o [Sun, 18 Aug 2019 23:23:07 +0000 (19:23 -0400)]
e2scrub_all: allow scrubbing in vg's whose free space == snapshot size

If the volume group's free space is exactly the same as snapshot size,
e2scrub_all will skip those logical volumes in those volume groups.
Fix this by changing the test from '>' to '>='.

Fixes: c120312253 ("e2scrub_all: make sure there's enough free space...")
Addresses-Debian-Bug: #935009
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agolibuuid: signedness/type fixes
Matthias Andree [Sat, 17 Aug 2019 11:16:02 +0000 (13:16 +0200)]
libuuid: signedness/type fixes

Signed-off-by: Matthias Andree <matthias.andree@gmx.de>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agocontrib: add ext4-ioc.c
Theodore Ts'o [Fri, 9 Aug 2019 18:54:36 +0000 (14:54 -0400)]
contrib: add ext4-ioc.c

Add a debugging program which allows calling ext4-specific ioctls on
files.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2fsck: add a developer-only extended option: clear_all_uninit_bits
Theodore Ts'o [Mon, 5 Aug 2019 16:47:18 +0000 (12:47 -0400)]
e2fsck: add a developer-only extended option: clear_all_uninit_bits

This option clears the uninitialized bit on all extents of all inodes.
Note that this can end up exposing uninitialized data to userspace.
It should only used in very specialized situations.

This option is only enabled via a new configure flag,
--enable-developer-features.  It should *not* be enabled by
distributions, as it enables features thare only designed for use by
ext4 developers.  These features have no documentation in the man
page, or regression tests, and if it breaks, you get to keep both
pieces.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: add changelog for 1.45.3-4 release
Theodore Ts'o [Wed, 7 Aug 2019 00:19:26 +0000 (20:19 -0400)]
debian: add changelog for 1.45.3-4 release

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: don't gitignore debian/patches
Theodore Ts'o [Sat, 3 Aug 2019 03:53:38 +0000 (23:53 -0400)]
debian: don't gitignore debian/patches

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agolibblkid: fix gcc -Wall warnings
Theodore Ts'o [Wed, 31 Jul 2019 01:13:44 +0000 (21:13 -0400)]
libblkid: fix gcc -Wall warnings

Google-Bug-Id: 118836063

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: stop using LTO (link-time optimization)
Theodore Ts'o [Mon, 29 Jul 2019 15:40:25 +0000 (11:40 -0400)]
debian: stop using LTO (link-time optimization)

LTO breaks reproducible builds, and there is some question as to how
reliable LTO's code generator is --- there are some scary stories that
it doesn't work well, and the GCC maintainers aren't super-interested
in fixing the bugs:

https://lists.debian.org/debian-devel/2019/07/msg00610.html

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: skip running "make check" if DEB_BUILD_OPTIONS contains nocheck
Theodore Ts'o [Sat, 27 Jul 2019 16:17:06 +0000 (12:17 -0400)]
debian: skip running "make check" if DEB_BUILD_OPTIONS contains nocheck

This was done automatically by debhelper, but it got dropped when
override_dh_auto_test was added by commit 7f4c3bb120 ("debian: run
"make check" with V=1 to keep blhc happy").

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: mark more files in e2fsprogs.install as only being present on Linux
Theodore Ts'o [Fri, 26 Jul 2019 00:07:45 +0000 (20:07 -0400)]
debian: mark more files in e2fsprogs.install as only being present on Linux

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: update changelog for 1.45.3-3 release
Theodore Ts'o [Thu, 25 Jul 2019 19:38:59 +0000 (15:38 -0400)]
debian: update changelog for 1.45.3-3 release

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2fsck: set E2FSCK_TIME correctly on a 32-bit arch with a 64-bit time_t
Theodore Ts'o [Thu, 25 Jul 2019 02:25:11 +0000 (22:25 -0400)]
e2fsck: set E2FSCK_TIME correctly on a 32-bit arch with a 64-bit time_t

Addresses-Debian-Bug: #932906

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: mark logsave as Multi-Arch: foreign
Theodore Ts'o [Thu, 25 Jul 2019 01:54:23 +0000 (21:54 -0400)]
debian: mark logsave as Multi-Arch: foreign

Also fix logsave so it is Architecure: any.  It was previously
"linux-any kfreebsd-any" due to a cut-and-paste error.

Addresses-Debian-Bug: #932876

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: fix Breaks/Replace dependency for logsave
Theodore Ts'o [Thu, 25 Jul 2019 01:50:24 +0000 (21:50 -0400)]
debian: fix Breaks/Replace dependency for logsave

Addresses-Debian-Bug: #932874

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: only install the udev and systemd files to e2fsprogs on Linux systems
Theodore Ts'o [Thu, 25 Jul 2019 01:31:20 +0000 (21:31 -0400)]
debian: only install the udev and systemd files to e2fsprogs on Linux systems

Otherwise it will fail to build on hurd and kfreebsd.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: run "make check" with V=1 to keep blhc happy
Theodore Ts'o [Thu, 25 Jul 2019 00:40:19 +0000 (20:40 -0400)]
debian: run "make check" with V=1 to keep blhc happy

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: update changelog for 1.45.3-2 release
Theodore Ts'o [Wed, 24 Jul 2019 17:19:10 +0000 (13:19 -0400)]
debian: update changelog for 1.45.3-2 release

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: add a hard dependency on logsave to e2fsprogs
Theodore Ts'o [Wed, 24 Jul 2019 16:20:11 +0000 (12:20 -0400)]
debian: add a hard dependency on logsave to e2fsprogs

The initramfs created by the initramfs-tools package needs logsave and
assumes it comes along with e2fsprogs.  If it is not present, the
result systems which will fail to boot.  Fix this by adding the
dependency.

In the future initramfs-tools will explicitly ask for logsave (tracked
in Debian Bug: #932854), but we'll need to keep this dependency until
the next stable release of Debian.

Addresses-Debian-Bug: #932855
Addresses-Debian-Bug: #932859
Addresses-Debian-Bug: #932861
Addresses-Debian-Bug: #932881
Addresses-Debian-Bug: #932888

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agopo: update fr.po (from translationproject.org)
Samuel Thibault [Sun, 21 Jul 2019 23:33:33 +0000 (19:33 -0400)]
po: update fr.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agodebian: update to standards version 4.4.0
Theodore Ts'o [Sun, 21 Jul 2019 19:08:14 +0000 (15:08 -0400)]
debian: update to standards version 4.4.0

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agoe2scrub_all_cron: check to make sure e2scrub_all exists
Theodore Ts'o [Sun, 21 Jul 2019 17:13:24 +0000 (13:13 -0400)]
e2scrub_all_cron: check to make sure e2scrub_all exists

Since e2scrub_all.cron is marked as a config file, it can hang around
after the package is removed, in which case e2scrub_all might not be
present.  So check to make sure e2scrub_all exists before trying to
execute it.

Addresses-Debian-Bug: #932622

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reported-by: Laurent Bigonville <bigon@debian.org>
4 years agopo: update pt.po (from translationproject.org)
Pedro Albuquerque [Sun, 21 Jul 2019 16:53:10 +0000 (12:53 -0400)]
po: update pt.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agopo: update uk.po (from translationproject.org)
Yuri Chornoivan [Sat, 20 Jul 2019 20:32:18 +0000 (16:32 -0400)]
po: update uk.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agopo: update sv.po (from translationproject.org)
Göran Uddeborg [Sat, 20 Jul 2019 20:32:18 +0000 (16:32 -0400)]
po: update sv.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agopo: update pl.po (from translationproject.org)
Jakub Bogusz [Sat, 20 Jul 2019 20:32:18 +0000 (16:32 -0400)]
po: update pl.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agopo: update de.po (from translationproject.org)
Mario Blättermann [Sat, 20 Jul 2019 20:32:17 +0000 (16:32 -0400)]
po: update de.po (from translationproject.org)

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agopo: update cs.po (from translationproject.org)
Petr Pisar [Sat, 20 Jul 2019 20:32:17 +0000 (16:32 -0400)]
po: update cs.po (from translationproject.org)

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