Masatake YAMATO [Fri, 10 Oct 2025 21:51:51 +0000 (06:51 +0900)]
ts/kill/decode: compare decoded output in an architecture-independent way
Reported-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Fixed: https://github.com/util-linux/util-linux/issues/3774 Signed-off-by: Masatake YAMATO <yamato@redhat.com>
Masatake YAMATO [Fri, 10 Oct 2025 20:41:25 +0000 (05:41 +0900)]
ts/kill/decode: consider arch dependent signum/name association
In the original code, the signal bits passed to the decode test were
sampled on x86_64. As reported in #3774, the association between signal
number and its name is architecture dependent.
With this change, the test case calculates the signal bits from signal
names at runtime instead of hardcoding.
Karel Zak [Thu, 9 Oct 2025 08:17:45 +0000 (10:17 +0200)]
Merge branch 'PR/sprintf-replace' of https://github.com/karelzak/util-linux-work
* 'PR/sprintf-replace' of https://github.com/karelzak/util-linux-work:
hwclock: use snprintf() instead of sprintf()
lib/mbsalign: use snprintf() instead of sprintf()
col: use snprintf() instead of sprintf()
libfdisk: use snprintf() instead of sprintf()
lsfd: use snprintf() instead of sprintf()
pipesz: use snprintf() instead of sprintf()
dmesg: use snprintf() instead of sprintf()
lsipc: use snprintf() instead of sprintf()
libblkid: use snprintf() instead of sprintf()
If there are no IP addresses, ul_netaddrq_bestaddr() returns threshold
ULNETLINK_RATING_BAD, but there were no addresses in the best array, and
best_ifaceq remains unset, which caused crash. Setting the initial
threshold to __ULNETLINK_RATING_MAX and checking for that value fixes that.
And more, it also allows to accept IP addresses with ULNETLINK_RATING_BAD
rating.
Karel Zak [Tue, 7 Oct 2025 10:48:41 +0000 (12:48 +0200)]
lib/mbsalign: use snprintf() instead of sprintf()
Add bufsiz parameter to mbs_safe_encode_to_buffer() and
mbs_invalid_encode_to_buffer() functions to enable safe buffer
operations. Track remaining buffer size internally using int
variable and convert all sprintf() calls to snprintf().
Add buffer overflow protection by checking snprintf() return
values and verifying sufficient space before memcpy() and
direct writes. Break encoding loop early if buffer space
is exhausted.
This change improves code safety by preventing potential buffer
overflows and makes buffer size limits explicit in the API.
Karel Zak [Mon, 6 Oct 2025 12:31:29 +0000 (14:31 +0200)]
Merge branch 'fix/rename_annotation_opt_to_annotate' of https://github.com/cgoesche/util-linux-fork
* 'fix/rename_annotation_opt_to_annotate' of https://github.com/cgoesche/util-linux-fork:
man-common: rename annotation.adoc to annotate.adoc
swapon: use ANNOTATE_OPTION as enum name for --annotate
swapon: rename the new --annotation option to --annotate
swapon: (usage) make the help text fit within 80 columns again
Karel Zak [Mon, 6 Oct 2025 12:29:50 +0000 (14:29 +0200)]
Merge branch 'kill--line-oriented-list' of https://github.com/masatake/util-linux
* 'kill--line-oriented-list' of https://github.com/masatake/util-linux:
kill (-l/-L): print one signal per line when stdout is not a TTY
kill: (refactor) rename parameter to better reflect its purpose
swapon: rename the new --annotation option to --annotate
The form --annotation gave the impression that after the subsequent
"=" comes the text that each column name should be annotated with.
Furthermore, the other two options with optional arguments (--show
and --discard) are in the imperative, so follow that style and use
--annotate instead.
(In the bargain, this allows aligning the descriptions in the help
text perfectly again.)
Also, improve the description of --annotate in the man page somewhat.
swapon: (usage) make the help text fit within 80 columns again
Recent commit 72829b08fe added four extra spaces to each description
in the usage text, causing some lines to be wider than 80 characters.
Undo that addition, to make the help text fit again in 80 columns.
Also, put the equals sign for the new --annotation option inside the
square brackets instead of mistakenly before them.
UL_NL_SOFT_ERROR can be issued if kernel sends unpaired RTM_DELADDR. It
should not happen, but it can happen due to race condition. And it happened
in some kernel versions. It is not reason to exit the processing loop.
When duplicating struct ul_nl_addr, set address to ifa_local, if it is set
to ifa_local in the source. This fixes the address for PtP IPv4 network
interfaces and avoids UL_NL_SOFT_ERROR during address removal.
Karel Zak [Thu, 2 Oct 2025 12:44:10 +0000 (14:44 +0200)]
ci: add bash-completion consistency check to CODECHECK
Add 'make checkcompletion' to the GitHub CI CODECHECK phase to verify
bash-completion files consistency for all future pull requests. This
ensures that all user-facing programs have proper bash-completion
scripts registered in both autotools and meson build systems.
Karel Zak [Thu, 2 Oct 2025 12:07:53 +0000 (14:07 +0200)]
tools: rewrite checkcompletion.sh to be source-based
The original checkcompletion.sh checked compiled binaries in the build
directory, which was unreliable and build-dependent. The new version
performs comprehensive source-based verification by:
1. Extracting program names from */Makemodule.am files (bin_PROGRAMS,
sbin_PROGRAMS, usrbin_exec_PROGRAMS, usrsbin_exec_PROGRAMS)
2. Verifying three-way consistency:
- Programs defined in Makemodule.am files
- bash-completion files in bash-completion/ directory
- Entries in bash-completion/Makemodule.am
- Entries in meson.build
3. Properly handling special cases:
- Programs that don't need completion (system tools like login, agetty)
- Programs with special install hooks (runuser, lastb symlinks)
The script now provides clear categorized output:
- Programs missing bash-completion files
- bash-completion files not registered in Makemodule.am
- Orphaned bash-completion files without programs
This makes it easier to maintain bash-completion consistency and catch
issues before release.
Masatake YAMATO [Thu, 2 Oct 2025 16:21:21 +0000 (01:21 +0900)]
lsns: don't abort if /proc/self/ns/user is absent; probe other ns entries
In 7d5036fdafe0 ("lsns: show namespaces only kept alive by open file
descriptors"), I added code that calls stat(2) on /proc/self/ns/user
and made lsns exit after reporting an error if the call failed. I
assumed /proc/self/ns/user would be available on all platforms.
As Axel Karjalainen reported (link below), that assumption was
wrong: on some platforms, the file is absent. Exiting for this reason
is undesirable.
The stat(2) call is used to obtain the dev_t of the backing device of
nsfs. However, /proc/self/ns/user is not the only source; calling
stat(2) on other namespace files under /proc/self/ns yields the same
dev_t. This change iterates over entries under /proc/self/ns and uses
the first one whose stat(2) succeeds.
Karel Zak [Thu, 2 Oct 2025 11:38:13 +0000 (13:38 +0200)]
bash-completion: add bits to dist tarball
The bits command was added in commit 6e1301d59 (text-utils: add bits
command) but the bash-completion file was not added to the autotools
Makemodule.am, causing it to be missing from release tarballs.
This resulted in meson build failures when using the tarball with
bash-completion installed, as meson.build references the file but
it doesn't exist in the tarball.
Fixes: https://github.com/util-linux/util-linux/issues/3766 Signed-off-by: Karel Zak <kzak@redhat.com>
Karel Zak [Thu, 2 Oct 2025 09:52:52 +0000 (11:52 +0200)]
Merge branch 'PR/agetty-issue-read' of https://github.com/karelzak/util-linux-work
* 'PR/agetty-issue-read' of https://github.com/karelzak/util-linux-work:
lib/configs: improve readability
agetty: use standard path macros
build-sys: make sure _PATH_SYSCONFDIR is defined
The current \4 and \6 issue file escapes implementation is inferior. It
uses get getifaddrs() to get a list of IP addresses. This function does not
provide enough information to discriminate between stable IP addresses and
ephemeral addresses. As a result, especially \6 often gives unreliable
results.
The code is actually unable to get list of all interfaces, so a proper out
of the box IP address reporting depends on external tools that generate
issue file with the interfaces list.
The netlink messages are already used, but only as a change notifier. The
contents is not used, even if it contains exact information about the
change. As a result, change processing is triggered even for unrelated
network changes like IPv6 router advertisement.
The new implementation uses the new netaddrq library. It reports more
reliable results especially for IPv6.
Additionally, two new escapes are implemented:
\a Report all interfaces and assigned addresses that are considered as
reliable.
\A Report all interfaces and all assigned addresses.
TODO:
To prevent overflooding of the console, the list is currently limited to 12
interfaces. It would be nice to make it configurable.
Two pass processing of issue files. First pass just collects IP protocols
and list of interfaces (in future interface patterns). Now it always
processes both IPv4 and IPv6 on all interfaces. Not so bad, as \a is smart
enough to display just the useful part.
Maybe implement more options and formatting support for \a and \A.
Maybe implement interface filter globs or regexps for \a and \A. Still not
so bad, as \a automatically skips interfaces without reliable addresses
(e. g. lo or TUN).
To support netlink and IP address processing, two new library files were
added:
netlink: Generic netlink message processing code converting netlink
messages to calls of callbacks with a pre-processed data.
netaddrq: A code that gets and maintains linked list of the current
interfaces and assigned IP addresses. It also provides a rating of IP
addresses based on its "quality", i. e. type of address, validity, lifetime
etc.
Karel Zak [Wed, 1 Oct 2025 13:13:02 +0000 (15:13 +0200)]
Merge branch 'feat/swapon_translate_col_names' of https://github.com/cgoesche/util-linux-fork
* 'feat/swapon_translate_col_names' of https://github.com/cgoesche/util-linux-fork:
bash-completion: (swapon) add --annotation option
swapon: document --annotation option on the man page
swapon: control column header tooltips with --annotation
swapon: annotate column header names with OSC8 hyperlinks
lib/strutils: add helper function for --annotation option
libsmartcols: new scols_column_{refer,get}_annotation routines
Lukas Herbolt [Wed, 1 Oct 2025 11:54:50 +0000 (13:54 +0200)]
fallocate: allow O_CREATE if mode is FALLOC_FL_WRITE_ZEROES
With the new flag we can create the ZEROED allcoated files directly and not
in the two steps. Removing FALLOC_FL_WRITE_ZEROES from the flags of not using
O_CREATE on open().
Karel Zak [Wed, 1 Oct 2025 11:54:08 +0000 (13:54 +0200)]
Merge branch 'configs3' of https://github.com/schubi2/util-linux
* 'configs3' of https://github.com/schubi2/util-linux:
Using fix issue dir path "/usr/lib" for agetty
agetty: using configs lib for parsing issue files
libcommon: added lib "configs" for parsing configuration files in the correct order
Masatake YAMATO [Sun, 31 Aug 2025 01:03:03 +0000 (10:03 +0900)]
lsfd: refer to /proc/$pid/map_files if a mapped file is masked
To extract file information, lsfd called stat(2) on the the file name
found in /proc/$pid/maps. However, lsfd may retrieve information for
the wrong file if the file is bind-mount'ed after the process maps the
file.
To obtain the information about the originally mapped file, this
change makes lsfd call fstat(2) on the corresponding entry under
/proc/$pid/map_files avoiding the bind-mount ambiguity.
swapon: control column header tooltips with --annotation
Whether column header names should be annotated can be specified
with the --annotation option which takes the following three
arguments: 'auto', 'never' and 'always', where the former is the
default and makes it so that annotations are only added when the
output device is a terminal. If the option is omitted the 'auto'
behavior should be expected.
Signed-off-by: Christian Goeschel Ndjomouo <cgoesc2@wgu.edu>
swapon: annotate column header names with OSC8 hyperlinks
Column header names are annotated with OSC8 hyperlinks to
provide tooltips with a description of the column.
This is mainly motivated by the fact that column headers
are currently not translated, so as an alternative we will
provide a column description translated according to the
environment locale setting.
Addresses: #1291 Signed-off-by: Christian Goeschel Ndjomouo <cgoesc2@wgu.edu>
lib/strutils: add helper function for --annotation option
This function helps evaluates the --annotation option argument
and determines whether annotations are enabled or not.
It is identical to the hyperlinkwanted() routine, and although
that forms some form of redundancy, it is intentionally kept
seperate for potential future changes.
Signed-off-by: Christian Goeschel Ndjomouo <cgoesc2@wgu.edu>
Karel Zak [Wed, 24 Sep 2025 08:06:04 +0000 (10:06 +0200)]
Merge branch 'PR/Benno-23Sep' of https://github.com/karelzak/util-linux-work
* 'PR/Benno-23Sep' of https://github.com/karelzak/util-linux-work:
fallocate: (man,usage) correct the alphabetical sorting of the options
fallocate: (man) slightly improve the grammar of two sentences
fallocate: redo four tweaks that were accidentally undone
column: (usage) wrap two descriptions, to make them fit within 80 columns
column: (usage) correct the description of --wrap-separator
getopt: (usage) make the description of -U fit within 80 columns
Karel Zak [Wed, 24 Sep 2025 08:05:26 +0000 (10:05 +0200)]
Merge branch 'master' of https://github.com/wuguanghao3/util-linux
* 'master' of https://github.com/wuguanghao3/util-linux:
losetup: Add the --remove/-R parameter to remove loop devices
losetup: rename function *_delete_* to *_detach_*
libsmartcols: new scols_column_{refer,get}_annotation routines
The new routines can be used to add an annotation to columns
which are intended to serve as descriptions for column headers.
These annotations are formatted with OSC8 escape sequences and
then sent to the output stream. A supporting terminal can then
display the formatted string as a tooltip.
This can be fairly useful when an application wants to provide
a translated description for a column header name, since the
latter is sometimes a mere abbreviation and not necessarily
telling enough.
Addresses: #1291 Signed-off-by: Christian Goeschel Ndjomouo <cgoesc2@wgu.edu>
Karel Zak [Tue, 23 Sep 2025 09:04:15 +0000 (11:04 +0200)]
liblastlog2: refactor conditional assignments for better readability
Split conditional assignments like 'if ((ret = func()) != 0)' into
separate assignment and comparison statements for improved code clarity,
maintainability, and debugging.
Karel Zak [Tue, 23 Sep 2025 08:46:59 +0000 (10:46 +0200)]
liblastlog2: fix operator precedence in conditional assignments
Fix operator precedence in ll2_rename_user() where != comparison was
taking precedence over assignment, causing retval to be assigned 0 or 1
instead of the actual function return value.
Fixes: https://github.com/util-linux/util-linux/issues/3756 Signed-off-by: Karel Zak <kzak@redhat.com>
Karel Zak [Mon, 22 Sep 2025 13:51:06 +0000 (15:51 +0200)]
tools: improve git-version-bump with validation and help
- Add --help option with usage information
- Validate script runs only on stable/vX.Y branches
- Validate version format (X.Y, X.Y.Z, X.Y-rcN, X.Y.Z-rcN)
Karel Zak [Mon, 22 Sep 2025 13:21:18 +0000 (15:21 +0200)]
tools: improve git-version-next to always output version variables
- Always output UL_NEXT_FINAL_RELEASE (previously only for RC releases)
- Add UL_LAST_FINAL_RELEASE for last stable (non-RC) release
- Remove _RAW variables from output for cleaner interface
The UL_NEXT_FINAL_RELEASE now represents the version that will be
released after code stabilization, while UL_LAST_FINAL_RELEASE
represents the last stable release without -rc suffix.
- Only works on stable/* branches (releases are generated from stable branches)
- Supports util-linux versioning: vX.Y (major), vX.Y.Z (maintenance)
- Command-line options: --rc, --release-master, --release-update
- Safety check prevents --release-master on branches with existing releases
- Testing support via --last-release and --last-xy-release overrides
Version Schema:
Major releases: vX.Y-rc1, vX.Y-rc2, [vX.Y-rc3], vX.Y
Maintenance releases: vX.Y.Z-rc1, vX.Y.Z
Sam Fink [Wed, 17 Sep 2025 14:04:51 +0000 (10:04 -0400)]
libblkid: Fix probe_ioctl_tp assigning BLKGETDISKSEQ as physical sector size
Fix issue introduced by PR#2908 in probe_ioctl_tp where the BLKGETDISKSEQ ioctl
result is incorrectly assigned to the topology with
blkid_topology_set_physical_sector_size instead of blkid_topology_set_diskseq.
This issue was observed while using a Debian 13 container on a RHEL 9.4 host
attempting to format a volume. The physical sector size was incorrectly
reported as 3. This issue also presents with the fdisk command,
which also uses this library to resolve physical sector size of devices.
Karel Zak [Wed, 17 Sep 2025 10:58:59 +0000 (12:58 +0200)]
agetty: fix erasure of escape sequences and tab characters
When escape sequences (like arrow keys) or tab characters are entered
at the login prompt, they are properly visualized but only partially
erasable with backspace. This is because the erase logic assumes each
stored character corresponds to one visual character, but escape
sequences display as "^[" (2 chars) and tabs expand to multiple spaces.
Track visual character width for each stored byte in a parallel array.
When erasing, use the stored visual width to properly erase all
displayed characters for that input byte.
Fixes: https://github.com/util-linux/util-linux/issues/3624 Signed-off-by: Karel Zak <kzak@redhat.com>
(Mentioning -u instead of -n, and falsely claiming emptiness.)
Now the above command will report:
blkid: invalid argument: no
(I don't think option -n needs to be mentioned, as it will be obvious
when inspecting the command line. Not mentioning the option allows
reusing an existing error message, avoiding needless ballooning and
allowing the patch to be applied to the stable branch.)
Karel Zak [Mon, 15 Sep 2025 14:14:08 +0000 (16:14 +0200)]
column: add --wrap-separator option for custom text wrapping
Add a new --wrap-separator option that allows users to specify custom
separator characters for text wrapping within table columns. When used
with --table-wrap, the separator is replaced with newlines to enable
wrapping at specific points rather than just by column width.
The implementation:
- Processes wrap separator replacement in modify_table() after column
wrap flags are set
- Uses scols_cell_refer_memory() for data containing embedded nulls
- Enables scols_wrapzero_nextchunk for columns with wrapping enabled
- Only applies to columns that have the wrap flag set via --table-wrap
Example usage:
$ echo -e 'Name:Desc\nJohn:A|software|dev' | \
column --table --separator ':' \
--table-wrap 2 --wrap-separator '|'
Name Desc
John A
software
dev
Fixes: https://github.com/util-linux/util-linux/issues/3739 Signed-off-by: Karel Zak <kzak@redhat.com>
Karel Zak [Thu, 11 Sep 2025 07:35:58 +0000 (09:35 +0200)]
libmount: don't report fsconfig errors with "nofail"
The new kernel API returns EINVAL on FSCONFIG_CMD_CREATE if the mount
source is inaccessible. We do not want to report this as an error when
the "nofail" mount option is specified.
Note that EINVAL may also be returned by other fsconfig() settings, so
we need to check whether a source is specified and whether it is really
inaccessible. This is just a heuristic (as with the old mount(2)).
Fixes: https://github.com/util-linux/util-linux/issues/3741 Signed-off-by: Karel Zak <kzak@redhat.com>
projects / thirdparty / util-linux.git / log
