]> git.ipfire.org Git - thirdparty/systemd.git/log
thirdparty/systemd.git
18 months agoexecute: Rename ExecRuntime to ExecSharedRuntime
Daan De Meyer [Sun, 26 Mar 2023 18:07:36 +0000 (20:07 +0200)] 
execute: Rename ExecRuntime to ExecSharedRuntime

Preparation for next commit

18 months agomkosi: Narrow glob used to install python packages
Daan De Meyer [Mon, 27 Mar 2023 11:23:40 +0000 (13:23 +0200)] 
mkosi: Narrow glob used to install python packages

Let's make sure we only install the python3 and python39 python
packages, instead of all the packages of all versions that are
packaged.

This also fixes the CentOS 8 CI because python3.11-pytest was failing
to install.

We have to ship our own powertools repo definition because we need to
enable module_hotfixes for powertools to coerce dnf into installing
some of the python packages.

18 months agoupdate TODO
Lennart Poettering [Mon, 27 Mar 2023 11:20:11 +0000 (13:20 +0200)] 
update TODO

18 months agotest: verify the journal with and without a sealing key
Frantisek Sumsal [Sun, 26 Mar 2023 15:04:57 +0000 (17:04 +0200)] 
test: verify the journal with and without a sealing key

The bit flips during journal verification cause various types of journal
corruptions, so it's useful to go through it even without a sealing key
to see how we handle corrupted stuff.

Also, provide a sealing key if running in "CI mode" (i.e. arguments),
to check the FSS-related codepaths in CIs as well.

18 months agosd-id128: make id128_read() optionally take root directory
Yu Watanabe [Sun, 26 Mar 2023 05:03:10 +0000 (14:03 +0900)] 
sd-id128: make id128_read() optionally take root directory

18 months agoMerge pull request #26961 from DaanDeMeyer/chase-and-pin
Daan De Meyer [Mon, 27 Mar 2023 08:41:08 +0000 (10:41 +0200)] 
Merge pull request #26961 from DaanDeMeyer/chase-and-pin

chase-symlinks: Add chase_and_open_parent() + at() variant

18 months agoMerge pull request #26870 from fbuihuu/move-unit-tests-in-a-dedicated-subdir
Luca Boccassi [Mon, 27 Mar 2023 08:38:15 +0000 (09:38 +0100)] 
Merge pull request #26870 from fbuihuu/move-unit-tests-in-a-dedicated-subdir

Move unit tests in a dedicated subdir

18 months agojournalctl: fix when --grep is used with --follow
Mike Yuan [Thu, 23 Mar 2023 14:17:40 +0000 (22:17 +0800)] 
journalctl: fix when --grep is used with --follow

Follow-up for #25147 (db4691961ca52759fe6645d0fddb659ee4299ac2)

--follow sets arg_lines to 10, which breaks
--grep as the latter implies --reverse.
So let's not set --reverse if --follow is used.

18 months agoMerge pull request #26294 from yuwata/locale-verify-conf
Zbigniew Jędrzejewski-Szmek [Mon, 27 Mar 2023 08:26:24 +0000 (10:26 +0200)] 
Merge pull request #26294 from yuwata/locale-verify-conf

locale: verify loaded keymap or friends

18 months agobootctl: split-out entry token related definitions into boot-entry.[ch]
Yu Watanabe [Sun, 26 Mar 2023 07:33:43 +0000 (16:33 +0900)] 
bootctl: split-out entry token related definitions into boot-entry.[ch]

No functional change, just preparation for later commits.
These can be used in kernel-install later.

Note, unlike the our usual coding style, the arguments for
boot_entry_token_ensure() and parse_boot_entry_token_type() are
referenced, updated, and may freed, hence, always pass initialized
values. That's why they are not named as 'ret_xyz'.

18 months agochase-symlinks: Add chase_and_open_parent() + at() variant 26961/head
Daan De Meyer [Fri, 24 Mar 2023 11:10:16 +0000 (12:10 +0100)] 
chase-symlinks: Add chase_and_open_parent() + at() variant

Helper to chase a path, pin its parent directory and return the filename
of the resolved path in its parent directory.

18 months agoDefine $KERNEL_INSTALL_UKI_GENERATOR
Gertalitec [Wed, 22 Mar 2023 21:09:10 +0000 (01:09 +0400)] 
Define $KERNEL_INSTALL_UKI_GENERATOR

Define $KERNEL_INSTALL_UKI_GENERATOR in case one wants it to be different from $KERNEL_INSTALL_INITRD_GENERATOR. This can be useful if one wants to use mkinitcpio / Dracut to generate the initrd, but without creating the UKI so this can be left for e.g. ukify or something else. Right now these initrd generators will read /etc/kernel/install.conf and generate the UKI

18 months agohwdb: Fixed thumb buttons reversed on CHERRY MW 2310 (#26992)
taniishkaaa [Sun, 26 Mar 2023 14:52:31 +0000 (20:22 +0530)] 
hwdb: Fixed thumb buttons reversed on CHERRY MW 2310 (#26992)

Fixes #12914.

18 months agohwdb: drop redundant entry
Yu Watanabe [Sun, 26 Mar 2023 00:44:48 +0000 (09:44 +0900)] 
hwdb: drop redundant entry

Follow-up for ff155c81620bc4e1ccce27d4d5194707048d8143.

The entry already matches with the generic rule defined at the beginning
of the hwdb file.

18 months agotest: install symlinks with valid targets on SUSE and Debian 26870/head
Franck Bui [Fri, 24 Mar 2023 12:46:24 +0000 (13:46 +0100)] 
test: install symlinks with valid targets on SUSE and Debian

18 months agoMerge pull request #26973 from mrc0mmand/userdbctl-tests
Yu Watanabe [Sun, 26 Mar 2023 00:49:27 +0000 (09:49 +0900)] 
Merge pull request #26973 from mrc0mmand/userdbctl-tests

test: add a couple of tests for userdbctl

18 months agocore/main: restore the correct assert about array position
Zbigniew Jędrzejewski-Szmek [Sat, 25 Mar 2023 10:58:49 +0000 (11:58 +0100)] 
core/main: restore the correct assert about array position

'pos' is incremented after each assignment. If we use the maximum number
of arguments, we end up with pos==9 after all the assignments, and it
points to where the next value would be assigned. This position must remain
NULL.

The assert I "fixed" was intentionally introduced in
26abdc73a212b90f7c4b71808a1028d2e87ab09f as a bugfix. So my "fix" repeated
the same error that was fixed back then.

18 months agoMerge pull request #26983 from mrc0mmand/coredump-test-followup
Yu Watanabe [Sun, 26 Mar 2023 00:47:49 +0000 (09:47 +0900)] 
Merge pull request #26983 from mrc0mmand/coredump-test-followup

test: don't expand the subshell expression prematurely

18 months agotest: filter the merged coverage report instead
Frantisek Sumsal [Sat, 25 Mar 2023 13:39:13 +0000 (14:39 +0100)] 
test: filter the merged coverage report instead

So we don't have to do this twice - once for the base report and then
for each "real" one.

Follow-up to 7fdd6e157a.

18 months agotest: add a couple of tests for userdbctl 26973/head
Frantisek Sumsal [Fri, 24 Mar 2023 16:26:48 +0000 (17:26 +0100)] 
test: add a couple of tests for userdbctl

18 months agotest: do an initial coverage capture
Frantisek Sumsal [Fri, 24 Mar 2023 20:15:28 +0000 (21:15 +0100)] 
test: do an initial coverage capture

I noticed that our coverage reports miss some files completely - this
happens when the test doesn't touch the code in them at all, so the
generated coverage data (and resulting reports) have no information
about them. Let's fix this by doing an initial zero coverage capture
that contains a zeroed counter for every instrumented line in every
file, so when we later merge it with a capture from the test, it shows up
with a missing coverage instead of not showing at all.

18 months agotest: drop unnecessary return 26983/head
Frantisek Sumsal [Sat, 25 Mar 2023 11:17:11 +0000 (12:17 +0100)] 
test: drop unnecessary return

Return code of the EXIT trap handler is ignored in bash.

18 months agotest: don't touch the /failed marker in the subtest scripts
Frantisek Sumsal [Sat, 25 Mar 2023 11:15:33 +0000 (12:15 +0100)] 
test: don't touch the /failed marker in the subtest scripts

As it is already handled by the "main" script (testsuite-74.sh).

18 months agotest: don't expand the subshell expression prematurely
Frantisek Sumsal [Sat, 25 Mar 2023 11:02:15 +0000 (12:02 +0100)] 
test: don't expand the subshell expression prematurely

We need to expand the subshell expression during the `bash -c`
invocation, not before, to take the desired effect, as now it expands to:

timeout 30 bash -c 'while [[ 0 -eq 0 ]]; do sleep 1; done'

instead of the expected:

timeout 30 bash -c 'while [[ $(coredumpctl list -q --no-legend 770 | wc -l) -eq 0 ]]; do sleep 1; done'

Follow-up to aadbd81f7f.

18 months agohwdb: Fix incorrect touchpad dimensions on Thinkpad L14 Gen1 (#26937)
taniishkaaa [Sat, 25 Mar 2023 10:41:07 +0000 (16:11 +0530)] 
hwdb: Fix incorrect touchpad dimensions on Thinkpad L14 Gen1 (#26937)

Closes #22793.

18 months agotest: add test for invalid keymap setting 26294/head
Yu Watanabe [Wed, 1 Feb 2023 19:29:43 +0000 (04:29 +0900)] 
test: add test for invalid keymap setting

18 months agolocale: also verify keymaps loaded from config file and converted keymaps
Yu Watanabe [Wed, 1 Feb 2023 18:23:50 +0000 (03:23 +0900)] 
locale: also verify keymaps loaded from config file and converted keymaps

If vconsole.conf or friends contain an invalid keymap, then let's
ignore the settings.

18 months agolocale: introduce x11_context_verify()
Yu Watanabe [Wed, 1 Feb 2023 17:53:35 +0000 (02:53 +0900)] 
locale: introduce x11_context_verify()

No functional change, preparation for later commits.

18 months agolocale: split out xkbcommon related functions to xkbcommon-util.c
Yu Watanabe [Wed, 1 Feb 2023 17:47:47 +0000 (02:47 +0900)] 
locale: split out xkbcommon related functions to xkbcommon-util.c

Then, use dlopen_many_sym_or_warn() with DLSYM_ARG() macro.

18 months agolocale: split out checking existence of keymap
Yu Watanabe [Wed, 1 Feb 2023 16:50:37 +0000 (01:50 +0900)] 
locale: split out checking existence of keymap

No functional change, preparation for later commits.

18 months agolocale: move x11_convert_to_vconsole() near the relevant functions
Yu Watanabe [Wed, 1 Feb 2023 16:34:24 +0000 (01:34 +0900)] 
locale: move x11_convert_to_vconsole() near the relevant functions

No functional changes, just refactoring.

18 months agoMerge pull request #26960 from poettering/syscall-catchup
Yu Watanabe [Sat, 25 Mar 2023 04:39:27 +0000 (13:39 +0900)] 
Merge pull request #26960 from poettering/syscall-catchup

syscall filter group updates

18 months agoxdg-autostart-generator: do not warn about unknown fields
Zbigniew Jędrzejewski-Szmek [Fri, 24 Mar 2023 16:51:52 +0000 (17:51 +0100)] 
xdg-autostart-generator: do not warn about unknown fields

My user manager says:
systemd-xdg-autostart-generator[2933]: /home/zbyszek/.config/autostart/org.gnome.Terminal.desktop:256: Unknown key name 'Actions' in section 'Desktop Entry', ignoring.
systemd-xdg-autostart-generator[2933]: /home/zbyszek/.config/autostart/org.gnome.Terminal.desktop:258: Unknown section 'Desktop Action new-window'. Ignoring.
systemd-xdg-autostart-generator[2933]: /home/zbyszek/.config/autostart/org.gnome.Terminal.desktop:343: Unknown section 'Desktop Action preferences'. Ignoring.
systemd-xdg-autostart-generator[2933]: /home/zbyszek/.config/autostart/org.telegram.desktop.desktop:12: Unknown key name 'Actions' in section 'Desktop Entry', ignoring.
systemd-xdg-autostart-generator[2933]: /home/zbyszek/.config/autostart/org.telegram.desktop.desktop:13: Unknown key name 'SingleMainWindow' in section 'Desktop Entry', ignoring.
systemd-xdg-autostart-generator[2933]: /home/zbyszek/.config/autostart/org.telegram.desktop.desktop:19: Unknown section 'Desktop Action Quit'. Ignoring.

This is not useful. Those are externally-provided files, and they are likely to
have entries which we know nothing about.

18 months agoMerge pull request #26977 from poettering/find-line-startswith
Yu Watanabe [Sat, 25 Mar 2023 04:38:24 +0000 (13:38 +0900)] 
Merge pull request #26977 from poettering/find-line-startswith

add new find_line_startswith() helper

18 months agobootctl: enable colored logging
Yu Watanabe [Fri, 24 Mar 2023 21:58:35 +0000 (06:58 +0900)] 
bootctl: enable colored logging

18 months agocore/main: fix setting of arguments for shutdown
Zbigniew Jędrzejewski-Szmek [Fri, 24 Mar 2023 13:11:48 +0000 (14:11 +0100)] 
core/main: fix setting of arguments for shutdown

Fixup for d2ebd50d7f9740dcf30e84efc75610af173967d2
and 6920049fad4fa39db5fec712f82f7f75b98fd4b9:
- add a comment that the last arg must be NULL and adjust the assert.
- move initialization around so that fields are declared,
  initialized, and consumed in the same order.
- move declaration of pos adjacent do declaration of command_line.
  This makes it easy to see that it was not initialized correctly.
- initialize buffers before writing the pointer into the args array.
  This makes no difference for the compiler, but it just feels "wrong"
  to do it in opposite order.

Because pos was off, we would ignore args after the timeout, and also
overwrite the buffer if enough args were used.

I think this is case shows clearly that declaring all variables at the
top of the function, with some initialized and other not, is very
error-prone. The compiler has no issue with declaring variables whereever,
and we should take advantage of this to make it keep declaration,
initialization, and use close. (Within reason of course.)

18 months agotools: add dump-auxv.py
Zbigniew Jędrzejewski-Szmek [Fri, 24 Mar 2023 13:54:34 +0000 (14:54 +0100)] 
tools: add dump-auxv.py

This is a little helper I used when preparing the tests for auxv
parsing. Just looking at hexdump output is pretty hard. We could
enhance it to display some specific data types better.

18 months agohwdb: drop boilerplate about match patterns in two more cases
Zbigniew Jędrzejewski-Szmek [Fri, 24 Mar 2023 16:25:59 +0000 (17:25 +0100)] 
hwdb: drop boilerplate about match patterns in two more cases

Follow-up for adbe000e3df5f91572bf4a77a2acd84ff5b87320.
Somehow I missed those two files.

18 months agoimport: use truncate_nl() where appropriate 26977/head
Lennart Poettering [Fri, 24 Mar 2023 17:27:03 +0000 (18:27 +0100)] 
import: use truncate_nl() where appropriate

18 months agotree-wide: port various places to find_line_startswith()
Lennart Poettering [Fri, 24 Mar 2023 17:25:13 +0000 (18:25 +0100)] 
tree-wide: port various places to find_line_startswith()

18 months agostring-util: add new helper for finding line starting with specific string in a text...
Lennart Poettering [Fri, 24 Mar 2023 17:02:32 +0000 (18:02 +0100)] 
string-util: add new helper for finding line starting with specific string in a text buffer

We have implemented this manually a couple of times, and always wrong.
Hence let's implement this correctly for once and use everywhere.

18 months agouserdbctl: flush stdout before running the chain command
Frantisek Sumsal [Fri, 24 Mar 2023 16:17:31 +0000 (17:17 +0100)] 
userdbctl: flush stdout before running the chain command

Otherwise it's quite difficult to capture the entire output:

$ userdbctl ssh-authorized-keys dropinuser --chain /bin/echo hello
ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIA//dxI2xLg4MgxIKKZv1nqwTEIlE/fdakii2Fb75pG+ foo@bar.tld
ecdsa-sha2-nistp256 AAAAE2VjZHNhLXNoYTItbmlzdHAyNTYAAAAIbmlzdHAyNTYAAABBBMlaqG2rTMje5CQnfjXJKmoSpEVJ2gWtx4jBvsQbmee2XbU/Qdq5+SRisssR9zVuxgg5NA5fv08MgjwJQMm+csc= hello@world.tld
hello
$ userdbctl ssh-authorized-keys dropinuser --chain /bin/echo hello | tee
hello

18 months agoMerge pull request #26920 from medhefgo/ukify
Zbigniew Jędrzejewski-Szmek [Fri, 24 Mar 2023 16:00:31 +0000 (17:00 +0100)] 
Merge pull request #26920 from medhefgo/ukify

ukify: Use pefile to add sections to EFI stub

18 months agodocs: s/authorized_key/authorized_keys/
Frantisek Sumsal [Fri, 24 Mar 2023 15:16:50 +0000 (16:16 +0100)] 
docs: s/authorized_key/authorized_keys/

18 months agoman: s/io.systemd.Dropin/io.systemd.DropIn/
Frantisek Sumsal [Fri, 24 Mar 2023 14:43:04 +0000 (15:43 +0100)] 
man: s/io.systemd.Dropin/io.systemd.DropIn/

18 months agouserdbctl: don't show legend when dumping JSON with services
Frantisek Sumsal [Fri, 24 Mar 2023 14:19:21 +0000 (15:19 +0100)] 
userdbctl: don't show legend when dumping JSON with services

18 months agoman: update syscal filter group list 26960/head
Lennart Poettering [Fri, 24 Mar 2023 13:52:34 +0000 (14:52 +0100)] 
man: update syscal filter group list

18 months agotest: on openSUSE install the collection of unit test binaries in the target only...
Franck Bui [Mon, 20 Mar 2023 17:23:33 +0000 (18:23 +0100)] 
test: on openSUSE install the collection of unit test binaries in the target only for TEST-02-UNITTESTS

18 months agomeson: make sure the unit test scripts find testdata/ even if they are not installed...
Franck Bui [Fri, 17 Mar 2023 13:18:50 +0000 (14:18 +0100)] 
meson: make sure the unit test scripts find testdata/ even if they are not installed in the same directory

testdata/ is shared by both the unit tests and the integration tests. Hence it
makes sense to place them right under /usr/lib/systemd/tests/ while the tests
themself are located under dedicated sub-directories:

     /usr/lib/systemd/tests/
     ├── integration-tests
     ├── testdata
     └── unit-tests
         ├── testdata -> ../testdata

However the unit tests implemented as shell scripts (such as
test-fstab-generator.sh) expect to find testdata/ where the scripts are
installed, ie they do something like:

  src="$(dirname "$0")/testdata/test-fstab-generator"

This patch installs a symlink in 'unitestsdir' named "testdata" and that points
to whatever value 'testdata_dir' is set.

18 months agomeson: define testdata_dir globally
Franck Bui [Fri, 17 Mar 2023 13:14:16 +0000 (14:14 +0100)] 
meson: define testdata_dir globally

18 months agotest: install unit tests in a dedicated subdirectory below '$testsdir'
Franck Bui [Thu, 26 Jan 2023 15:07:17 +0000 (16:07 +0100)] 
test: install unit tests in a dedicated subdirectory below '$testsdir'

/usr/lib/systemd/tests may contain more than the unit tests. For example on
SUSE we also install the integration tests there.

Putting the unit tests in a dedicated directory named 'unit-tests' makes the
layout cleaner.

Note that `run-unit-tests.py` has not been moved so we don't need to adjust
(Fedora) packaging and users also don't need to descend into the subdirectory.

18 months agochase-symlinks: Allow optional path with CHASE_EXTRACT_FILENAME
Daan De Meyer [Fri, 24 Mar 2023 11:21:45 +0000 (12:21 +0100)] 
chase-symlinks: Allow optional path with CHASE_EXTRACT_FILENAME

This allows helpers to set it unconditionally even if the caller
is not interested in the path.

18 months agochase-symlinks: Rename chase_symlinks() to chase()
Daan De Meyer [Tue, 21 Mar 2023 20:17:22 +0000 (21:17 +0100)] 
chase-symlinks: Rename chase_symlinks() to chase()

Chasing symlinks is a core function that's used in a lot of places
so it deservers a less verbose names so let's rename it to chase()
and chaseat().

We also slightly change the pattern used for the chaseat() helpers
so we get chase_and_openat() and similar.

18 months agoMerge pull request #26785 from keszybz/udev-distcheck
Zbigniew Jędrzejewski-Szmek [Fri, 24 Mar 2023 12:38:54 +0000 (13:38 +0100)] 
Merge pull request #26785 from keszybz/udev-distcheck

Implement --help/--version in all udev builtins

18 months agoukify: Add riscv32 and loongarch support 26920/head
Jan Janssen [Tue, 21 Mar 2023 16:10:19 +0000 (17:10 +0100)] 
ukify: Add riscv32 and loongarch support

18 months agoukify: Use pefile to add sections to EFI stub
Jan Janssen [Tue, 21 Mar 2023 16:07:06 +0000 (17:07 +0100)] 
ukify: Use pefile to add sections to EFI stub

18 months agoanalyze: don't claim "@known" was an unlisted syscall
Lennart Poettering [Fri, 24 Mar 2023 10:35:55 +0000 (11:35 +0100)] 
analyze: don't claim "@known" was an unlisted syscall

It's a sycall group of our own definition, and the output is erroneous
to claim otherwise. Let's hide it.

This adds syscall_set_add() which is nicely symmetric to the existing
syscall_set_remove().

Follow-up for: 6d6a08547c03f96dc798cda1ef4a8d3013d292d5

18 months agoseccomp-util: enforce group ordering
Lennart Poettering [Fri, 24 Mar 2023 10:17:08 +0000 (11:17 +0100)] 
seccomp-util: enforce group ordering

So far we asked via a comment that @default should stay the first group
and @known the last group in the list. Let's enforce that statically, in
code, too.

18 months agoseccomp-util: add some newer syscalls to existing groups
Lennart Poettering [Fri, 24 Mar 2023 10:16:38 +0000 (11:16 +0100)] 
seccomp-util: add some newer syscalls to existing groups

These three new syscalls are mostly just new flavours or extensions of
existing syscalls, hence add them to the same groups.

18 months agoseccomp-util: add new @sandbox syscall group with landlock/seccomp
Lennart Poettering [Fri, 24 Mar 2023 10:15:21 +0000 (11:15 +0100)] 
seccomp-util: add new @sandbox syscall group with landlock/seccomp

Let's group these 4 syscalls, as they offer similar things and I guess
might be used in conjunction quite often, as they offer unprivileged
sandboxing.

Fixes: #26913
18 months agotest: add a couple of tests for systemd-coredump
Frantisek Sumsal [Tue, 21 Mar 2023 22:19:41 +0000 (23:19 +0100)] 
test: add a couple of tests for systemd-coredump

18 months agoMerge pull request #26958 from yuwata/nulstr-optionally-drop-trailing-nulstr
Daan De Meyer [Fri, 24 Mar 2023 10:33:58 +0000 (11:33 +0100)] 
Merge pull request #26958 from yuwata/nulstr-optionally-drop-trailing-nulstr

nulstr: make strv_parse_nulstr() optionally drop trailing NULs

18 months agoMerge pull request #26957 from yuwata/proc-cmdline-cleanups
Daan De Meyer [Fri, 24 Mar 2023 10:29:54 +0000 (11:29 +0100)] 
Merge pull request #26957 from yuwata/proc-cmdline-cleanups

proc-cmdline: several cleanups

18 months agoMerge pull request #26948 from yuwata/vconsole-fix-memleak
Daan De Meyer [Fri, 24 Mar 2023 10:28:00 +0000 (11:28 +0100)] 
Merge pull request #26948 from yuwata/vconsole-fix-memleak

vconsole: introduce Context and its helper functions

18 months agoMerge pull request #26956 from yuwata/core-main-arguments-followed-by-equal
Daan De Meyer [Fri, 24 Mar 2023 10:27:45 +0000 (11:27 +0100)] 
Merge pull request #26956 from yuwata/core-main-arguments-followed-by-equal

core/main: make arguments followed by '='

18 months agoprocess-util: drop trailing NUls before parsing the nulstr 26958/head
Yu Watanabe [Thu, 23 Mar 2023 02:59:44 +0000 (11:59 +0900)] 
process-util: drop trailing NUls before parsing the nulstr

No functional changes, just refactoring.

18 months agotest: add tests from strv_parse_nulstr_full()
Yu Watanabe [Thu, 23 Mar 2023 03:40:03 +0000 (12:40 +0900)] 
test: add tests from strv_parse_nulstr_full()

18 months agonulstr-util: introduce strv_parse_nulstr_full() that optionally drop trailing empty...
Yu Watanabe [Thu, 23 Mar 2023 03:39:36 +0000 (12:39 +0900)] 
nulstr-util: introduce strv_parse_nulstr_full() that optionally drop trailing empty strings

18 months agotest-proc-cmdline: test proc_cmdline_get_key() actually parses EFI options 26957/head
Yu Watanabe [Wed, 22 Mar 2023 20:52:32 +0000 (05:52 +0900)] 
test-proc-cmdline: test proc_cmdline_get_key() actually parses EFI options

Follow-up for 53aa0d02add93d8c0afa2772609372a2040c162a.

18 months agoproc-cmdline: make proc_cmdline_parse_given() static
Yu Watanabe [Wed, 22 Mar 2023 19:55:59 +0000 (04:55 +0900)] 
proc-cmdline: make proc_cmdline_parse_given() static

It is used only in proc-cmdline.c and its test. And the test can be
covered by proc_cmdline_parse().

18 months agoproc-cmdline: use proc_cmdline_key_string() when we search for key
Yu Watanabe [Wed, 22 Mar 2023 18:06:16 +0000 (03:06 +0900)] 
proc-cmdline: use proc_cmdline_key_string() when we search for key

18 months agoproc-cmdline: rename variable
Yu Watanabe [Wed, 22 Mar 2023 19:32:23 +0000 (04:32 +0900)] 
proc-cmdline: rename variable

18 months agoproc-cmdline: insert an empty line between variable declaration and assertion
Yu Watanabe [Thu, 23 Mar 2023 00:21:28 +0000 (09:21 +0900)] 
proc-cmdline: insert an empty line between variable declaration and assertion

18 months agocore/main: make positional arguments followed by '=', then by value 26956/head
Yu Watanabe [Thu, 23 Mar 2023 05:10:30 +0000 (14:10 +0900)] 
core/main: make positional arguments followed by '=', then by value

To make ConditionKernelCommandLine= or friend not confused when we are
running in a container.

Addresses https://github.com/systemd/systemd/pull/26887#discussion_r1143358884.

18 months agocore/main: fix maximum number of arguments for shutdown command
Yu Watanabe [Mon, 20 Mar 2023 23:02:16 +0000 (08:02 +0900)] 
core/main: fix maximum number of arguments for shutdown command

Follow-up for c5673ed0de3bec38f68d8113d253842b47766e27.

18 months agovconsole: introduce context_get_config() helper function 26948/head
Yu Watanabe [Fri, 24 Mar 2023 05:54:15 +0000 (14:54 +0900)] 
vconsole: introduce context_get_config() helper function

And make keyboard_load_and_wait() and font_load_and_wait() take Context.

18 months agovconsole: introduce Context and its helper functions
Yu Watanabe [Thu, 23 Mar 2023 13:30:30 +0000 (22:30 +0900)] 
vconsole: introduce Context and its helper functions

Fixes memleaks introduced by 01771226c202183ff447da712f43d2fad8874484
and ea575e176aac9fa8f430bb30a3e8abd8da767a10

Fixes #26945.

18 months agochase-symlinks: Use xopenat() instead of open_mkdir_at()
Daan De Meyer [Thu, 23 Mar 2023 19:05:16 +0000 (20:05 +0100)] 
chase-symlinks: Use xopenat() instead of open_mkdir_at()

We don't rely on any of the extra functionality of open_mkdir_at()
so let's just use xopenat() directly.

18 months agoMerge pull request #26935 from keszybz/test-parse_aux
Luca Boccassi [Thu, 23 Mar 2023 23:13:42 +0000 (23:13 +0000)] 
Merge pull request #26935 from keszybz/test-parse_aux

Add test for auxv parsing

18 months agoMerge pull request #26949 from DaanDeMeyer/xopenat-reopen
Luca Boccassi [Thu, 23 Mar 2023 23:08:21 +0000 (23:08 +0000)] 
Merge pull request #26949 from DaanDeMeyer/xopenat-reopen

loop-util: Add loop_device_make_by_path_at()

18 months agofileio: add new helper fdopen_independent()
Lennart Poettering [Thu, 23 Mar 2023 11:33:41 +0000 (12:33 +0100)] 
fileio: add new helper fdopen_independent()

This is a combination of fdopen() and fd_reopen(). i.e. it first reopens
the fd, and then converts that into a FILE*.

We do this at various places already manually. let's move this into a
helper call of its own.

18 months agoMerge pull request #26953 from poettering/encrypted-cred-mini-refactor
Lennart Poettering [Thu, 23 Mar 2023 20:21:45 +0000 (21:21 +0100)] 
Merge pull request #26953 from poettering/encrypted-cred-mini-refactor

core: minor refactoring of allowlisting TPM devices when encrypted creds are loaded

18 months agopid1: allowlist all tpm devices for a unit when encrypted creds are needed 26953/head
Lennart Poettering [Thu, 23 Mar 2023 17:22:43 +0000 (18:22 +0100)] 
pid1: allowlist all tpm devices for a unit when encrypted creds are needed

We might be configured to use some ther device than /dev/tpmrm0, hence
allow them all by allowlisting the tpm char device class as a whole.

18 months agocore: move encrypted credential check to execute.c
Lennart Poettering [Thu, 23 Mar 2023 12:58:34 +0000 (13:58 +0100)] 
core: move encrypted credential check to execute.c

This is an operation on an ExecContext, hence it probably should be
placed there.

18 months agotest-coredump-util: also test parse_auxv() with unaligned data 26935/head
Zbigniew Jędrzejewski-Szmek [Wed, 22 Mar 2023 13:06:31 +0000 (14:06 +0100)] 
test-coredump-util: also test parse_auxv() with unaligned data

18 months agotest-coredump-util: add tests for parse_aux()
Zbigniew Jędrzejewski-Szmek [Wed, 22 Mar 2023 12:42:40 +0000 (13:42 +0100)] 
test-coredump-util: add tests for parse_aux()

The test files are /proc//auxv files copies from various architecutres
signified by the file name suffix.

Those tests are fairly simple, but when we run them on n architectures, we do
~n² cross-arch tests.

18 months agocoredump: split out parse_auxv() to src/shared/
Zbigniew Jędrzejewski-Szmek [Wed, 22 Mar 2023 08:53:16 +0000 (09:53 +0100)] 
coredump: split out parse_auxv() to src/shared/

No functional change. (We already checked for ELFCLASS32 or ELFCLASS64 before,
so even though there's a new check for other architectures, the only caller
only passes ELFCLASS32 or ELFCLASS64.)

18 months agoloop-util: Add loop_device_make_by_path_at() 26949/head
Daan De Meyer [Thu, 23 Mar 2023 13:42:35 +0000 (14:42 +0100)] 
loop-util: Add loop_device_make_by_path_at()

On top of taking a directory file descriptor, we use xopenat() so
that the function can also be used to work on existing file
descriptors to image files including all the logic to use O_DIRECT
and fallback to O_RDONLY if needed.

18 months agofs-util: Allow xopenat() to reopen existing file descriptors
Daan De Meyer [Thu, 23 Mar 2023 13:30:43 +0000 (14:30 +0100)] 
fs-util: Allow xopenat() to reopen existing file descriptors

18 months agoMerge pull request #26214 from YHNdnzj/sd-notify-change-notifyaccess
Mike Yuan [Thu, 23 Mar 2023 16:11:29 +0000 (00:11 +0800)] 
Merge pull request #26214 from YHNdnzj/sd-notify-change-notifyaccess

core: support overriding NOTIFYACCESS= through sd-notify during runtime

18 months agodocs: Explicitly tell developers to enable mkosi required meson options
Daan De Meyer [Thu, 23 Mar 2023 10:15:29 +0000 (11:15 +0100)] 
docs: Explicitly tell developers to enable mkosi required meson options

We need repart, bootctl, analyze and ukify for mkosi so let's make
sure those get built in the HACKING guide.

18 months agofind-esp: don't silently error bootctl install if presumed XBOOTLDR part is stx_dev_m...
наб [Fri, 3 Mar 2023 14:57:42 +0000 (15:57 +0100)] 
find-esp: don't silently error bootctl install if presumed XBOOTLDR part is stx_dev_major=0 but not btrfs

btrfs_get_block_device_fd() returns -ENOTTY if fstatfs().f_type !=
BTRFS_SUPER_MAGIC

btrfs_get_block_device_fd() is run by verify_fsroot_dir() by
verify_xbootldr() by find_xbootldr_and_warn() if
statx($presumed-XBOOTLDR).stx_dev_major == 0 ("maybe a btrfs device")

Every bootctl verb_install() runs find_xbootldr_and_warn(), by default
with /boot

If your /boot .stx_dev_major=0 but /not/ btrfs, bootctl install/update
quietly exits 1 with no note so as to what exactly failed (debug also
empty, and the strace isn't exactly clear since no syscall actually
failed)

This is the case on ZFS and the Debian filesystem layout: /boot/efi is
the ESP, and everything else under / is ZFS:
  $ sudo env SYSTEMD_LOG_LEVEL=debug bootctl update
  Found cgroup2 on /sys/fs/cgroup/, full unified hierarchy
  Found container virtualization none.
  File system "/boot" is not a FAT EFI System Partition (ESP) file system.
  Using EFI System Partition at /boot/efi.
  Checking whether /boot/efi/EFI/systemd/ contains any files…
  $ echo $?
  1
and funnier still:
  $ sudo bootctl update --graceful
  $ echo $?
  1

Which is great, and also breaks postinst, which runs precisely the
latter, with no feedback at all

By checking for -ENOTTY we accept that the path being investigated
"is not it" if it's on ZFS (and any other filesystem that returns
.stx_dev_major == 0 but isn't btrfs)

18 months agotest-fileio: add test for return value of read_one_line_file()
Zbigniew Jędrzejewski-Szmek [Tue, 21 Mar 2023 08:32:12 +0000 (09:32 +0100)] 
test-fileio: add test for return value of read_one_line_file()

Prompted by https://github.com/systemd/systemd/pull/26904#pullrequestreview-1349890241.

18 months agoMerge pull request #26936 from DaanDeMeyer/xopenat-lock
Daan De Meyer [Thu, 23 Mar 2023 06:49:12 +0000 (07:49 +0100)] 
Merge pull request #26936 from DaanDeMeyer/xopenat-lock

fs-util: Add xopenat_lock()

18 months agoMerge pull request #26916 from DaanDeMeyer/log-context-ref
Daan De Meyer [Wed, 22 Mar 2023 21:07:45 +0000 (22:07 +0100)] 
Merge pull request #26916 from DaanDeMeyer/log-context-ref

log: Avoid pushing the same fields more than once on the log context

18 months agofs-util: Add xopenat_lock() 26936/head
Daan De Meyer [Wed, 22 Mar 2023 16:04:36 +0000 (17:04 +0100)] 
fs-util: Add xopenat_lock()

open/create a file/directory and lock it using the given lock type.

18 months agoedit-util: alloc correct amount of memory
Mike Yuan [Wed, 22 Mar 2023 12:15:19 +0000 (20:15 +0800)] 
edit-util: alloc correct amount of memory

18 months agocoredump: use unaligned_read_ne{32,64}() to parse auxv
Yu Watanabe [Tue, 21 Mar 2023 23:49:49 +0000 (08:49 +0900)] 
coredump: use unaligned_read_ne{32,64}() to parse auxv

Fixes a bug introduced by 3e4d0f6cf99f8677edd6a237382a65bfe758de03.

The auxv metadata is unaligned, as the length of the prefix
"COREDUMP_PROC_AUXV=" is 19. Hence, parse_auxv{32,64}() may triger
an undefined behavior (or at least cause slow down), which can be
detected when running on an undefined behavior sanitizer.

This also introduces a macro to define `parse_auxv{32,64}()`.

Fixes #26912.

18 months agoukify: allow uncompressed kernel images for UNAME detection on aarch64 and riscv64...
Tobias Powalowski [Wed, 22 Mar 2023 12:24:57 +0000 (13:24 +0100)] 
ukify: allow uncompressed kernel images for UNAME detection on aarch64 and riscv64 (#26929)

Uncompressed aarch64 and riscv64 kernels have a different startpoint than x86.
Example output from ukify:
aarch64: NotImplementedError: unknown file format (starts with b'MZ@\xfa')
riscv64: NotImplementedError: unknown file format (starts with b'MZo\x10')
Add check for (b'MZ') to catch both in one call.

Fix:
https://github.com/systemd/systemd/issues/26923

18 months agocore: Settle log target if we're going to be closing all fds
Daan De Meyer [Mon, 20 Mar 2023 12:26:57 +0000 (13:26 +0100)] 
core: Settle log target if we're going to be closing all fds

Whenever we're going to close all file descriptors, we tend to close
the log and set it into open when needed mode. When this is done with
the logging target set to LOG_TARGET_AUTO, we run into issues because
for every logging call, we'll check if stderr is connected to the
journal to determine where to send the logging message. This check
obviously stops working when we close stderr, so we settle the log
target before we do that so that we keep using the same logging
target even after stderr is closed.