]> git.ipfire.org Git - thirdparty/systemd.git/log
thirdparty/systemd.git
2 years agobootctl: split out setting of system token into function of its own 25802/head
Lennart Poettering [Tue, 20 Dec 2022 17:03:06 +0000 (18:03 +0100)] 
bootctl: split out setting of system token into function of its own

Let's break a huge function in two. No code change, just some
refactoring.

2 years agounits: rename/rework systemd-boot-system-token.service → systemd-boot-random-seed...
Lennart Poettering [Tue, 20 Dec 2022 16:16:47 +0000 (17:16 +0100)] 
units: rename/rework systemd-boot-system-token.service → systemd-boot-random-seed.service

This renames systemd-boot-system-token.service to
systemd-boot-random-seed.service and conditions it less strictly.

Previously, the job of the service was to write a "system token" EFI
variable if it was missing. It called "bootctl --graceful random-seed"
for that. With this change we condition it more liberally: instead of
calling it only when the "system token" EFI variable isn't set, we call
it whenever a boot loader interface compatible boot loader is used. This
means, previously it was invoked on the first boot only: now it is
invoked at every boot.

This doesn#t change the command that is invoked. That's because
previously already the "bootctl --graceful random-seed" did two things:
set the system token if not set yet *and* refresh the random seed in the
ESP. Previousy we put the focus on the former, now we shift the focus to
the latter.

With this simple change we can replace the logic
f913c784ad4c93894fd6cb2590738113dff5a694 added, but from a service that
can run much later and doesn't keep the ESP pinned.

2 years agobootctl: downgrade graceful messages to LOG_NOTICE
Lennart Poettering [Tue, 20 Dec 2022 15:34:36 +0000 (16:34 +0100)] 
bootctl: downgrade graceful messages to LOG_NOTICE

2 years agorandom-seed: don't refresh EFI random seed from random-seed.c anymore
Lennart Poettering [Tue, 20 Dec 2022 15:18:11 +0000 (16:18 +0100)] 
random-seed: don't refresh EFI random seed from random-seed.c anymore

The ESP is simply not mounted early enough for this. We want that the
regular random seed handling runs as early as we possibly could, but we
don't want to delay this until the ESP is actually mounted.

Hence, let's remove this from random-seed.c here. A follow-up commit
will then add this back in, in a separate service which just calls
"bootctl random-seed".

Effectively reverts: f913c784ad4c93894fd6cb2590738113dff5a694

Fixes: #25769
2 years agosha256: add helper than hashes a buffer *and* its size
Lennart Poettering [Tue, 20 Dec 2022 10:53:37 +0000 (11:53 +0100)] 
sha256: add helper than hashes a buffer *and* its size

We use this pattern all the time in order to thward extension attacks,
add a helper to make it shorter.

2 years agobootctl: when updating EFI random seed file, hash old seed with new one
Lennart Poettering [Tue, 20 Dec 2022 10:48:21 +0000 (11:48 +0100)] 
bootctl: when updating EFI random seed file, hash old seed with new one

Let's not regress in entropy in any case.

This does what f913c784ad4c93894fd6cb2590738113dff5a694 also does.

2 years agobootctl: properly sync fs before/after moving random seed file into place
Lennart Poettering [Tue, 20 Dec 2022 10:15:51 +0000 (11:15 +0100)] 
bootctl: properly sync fs before/after moving random seed file into place

Let's do a careful, focussed sync at the right places instead of a
blanket sync at the end. After all we want to run this on every boot
soon.

2 years agobootctl: rework random seed logic to use open_mkdir_at() and openat()
Lennart Poettering [Mon, 19 Dec 2022 21:26:30 +0000 (22:26 +0100)] 
bootctl: rework random seed logic to use open_mkdir_at() and openat()

This doesn't really fix anything, but in general we should put stronger
emphasis on operating via dir fds rather than paths more (in particular
when writing files as opposed to consuming them).

No real change in behaviour.

2 years agobootctl: fix errno logging
Lennart Poettering [Mon, 19 Dec 2022 21:25:28 +0000 (22:25 +0100)] 
bootctl: fix errno logging

2 years agohwdb: remove spurious whitespace
Luca Boccassi [Tue, 3 Jan 2023 22:18:43 +0000 (23:18 +0100)] 
hwdb: remove spurious whitespace

Fixes test failures.
Follow-up for c3950f0a27fb7d2298a2182f21d645c7017f9c0c and
https://github.com/systemd/systemd/pull/25908

2 years agoefi: skip Read() calls with zero sizes
Lennart Poettering [Tue, 3 Jan 2023 14:58:46 +0000 (15:58 +0100)] 
efi: skip Read() calls with zero sizes

Let's avoid calling Read() with zero-sized buffer, to avoid needless firmware
quirkiness.

See: #25911

2 years agoresolved: downgrade inability to send varlink reply error message
Lennart Poettering [Tue, 3 Jan 2023 15:26:55 +0000 (16:26 +0100)] 
resolved: downgrade inability to send varlink reply error message

Previously, if a client disconnected after sending a lookup request but
before waiting for the reply we'd log at LOG_ERR level. That's
confusing, since it's entirely OK for the client to lose interest.

Hence, let's downgrade to debug level.

Fixes: #25892
2 years agoMerge pull request #25910 from crrodriguez/wth_c2x
Luca Boccassi [Tue, 3 Jan 2023 22:10:36 +0000 (23:10 +0100)] 
Merge pull request #25910 from crrodriguez/wth_c2x

Fix FTBS when -std=gnu2x

2 years agoudev: match device tags in rules using current device tags
msizanoen1 [Tue, 27 Dec 2022 04:57:12 +0000 (11:57 +0700)] 
udev: match device tags in rules using current device tags

This ensures that udev scripts using `TAG-="..."` and expecting later
udev rules to honor it will work properly. An use case is removing the
`uaccess` tag from a device without overriding the original file and
ensuring that `73-seat-uaccess.rules` won't run the uaccess builtin later.

2 years agoAdded Tablet Teclast X98 Air 3G (C5J6)
marcel151 [Sun, 1 Jan 2023 18:19:19 +0000 (19:19 +0100)] 
Added Tablet Teclast X98 Air 3G (C5J6)

Latest BIOS of X98 Air 3G has later date, that's why it wasn't recognized. Uses the same Matrix as the X89 (E7ED).

2 years agoefi-api: check the EFI_TCG2_FINAL_EVENTS_TABLE in efi_has_tpm2()
Jia Zhang [Sat, 24 Dec 2022 09:46:01 +0000 (17:46 +0800)] 
efi-api: check the EFI_TCG2_FINAL_EVENTS_TABLE in efi_has_tpm2()

The EFI firmware may provide the TPM2 event log using
EFI_TCG2_FINAL_EVENTS_TABLE stored in EFI configuration table,
instead of the ACPI Table TPM2.

If the ACPI Table TPM2 doesn't exist, try to check whether
EFI_TCG2_FINAL_EVENTS_TABLE is available or not.

Signed-off-by: Jia Zhang <zhang.jia@linux.alibaba.com>
2 years agoRevert "boot: Use EFI_BOOT_MANAGER_POLICY_PROTOCOL to connect console devices"
Jan Janssen [Fri, 23 Dec 2022 13:14:53 +0000 (14:14 +0100)] 
Revert "boot: Use EFI_BOOT_MANAGER_POLICY_PROTOCOL to connect console devices"

This reverts commit b99bf5811850afdb2502ba37251c48348da63c82.

It seems that using this protocol on some firmwares to forcibly
initialize console devices may break handles (already opened file
handles and the device handle itself) that we rely on to access the
boot filesystem, making it impossible to load the selected entry.

It might be possible to get a new handle by querying for the device
handle by using its device path after calling into this protocol, but
this is untested. The firmware might also be so buggy that accessing
devices after using this protocol is impossible.

It seems prudent to revert this for now until some reliable way is found
to initialize console devices without introducing huge boot delays. Any
users on firmware where console devices cannot be accessed would have to
rely on disabling fastboot.

Fixes: #25737, #25846
2 years agolibsystemd-network: FTBS in c2x mode 25910/head
Cristian Rodríguez [Mon, 2 Jan 2023 00:08:52 +0000 (00:08 +0000)] 
libsystemd-network: FTBS in c2x mode

cannot pass false as argument because function wants a pointer to bool
instead, use NULL instead

2 years agobasic: Fix incompatible type for arguments errors in C2X
Cristian Rodríguez [Mon, 2 Jan 2023 00:06:57 +0000 (00:06 +0000)] 
basic: Fix incompatible type for arguments errors in C2X

GCC-13 -std=gnu2x FTBS with:

error: incompatible type for argument 3 of ‘_hashmap_free’

2 years agobasic: fix hosed return value in skip_session()
Cristian Rodríguez [Sun, 1 Jan 2023 22:40:26 +0000 (22:40 +0000)] 
basic: fix hosed return value in skip_session()

../src/basic/cgroup-util.c: In function ‘skip_session’:
../src/basic/cgroup-util.c:1241:32: error: incompatible types when returning type ‘_Bool’ but ‘const char *’ was expected
 1241 |                         return false;

2 years agoman: Fix systemd-boot man page wrong secure keys location
igo95862 [Tue, 3 Jan 2023 07:18:58 +0000 (10:18 +0300)] 
man: Fix systemd-boot man page wrong secure keys location

The keys should be placed at `/loader/keys/` not `/keys/`.

See `src/boot/efi/boot.c` file and
function `secure_boot_discover_keys`.

2 years agofundamental: rework IN_SET() to require at least three arguments
Lennart Poettering [Mon, 2 Jan 2023 14:22:15 +0000 (15:22 +0100)] 
fundamental: rework IN_SET() to require at least three arguments

If less than three parameters are passed a simple comparison is the
better choice.

Lo and behold this found two pretty bad typos.

2 years agocore: rework operation_to_signal() to rename return parameter ret_xyz
Lennart Poettering [Mon, 2 Jan 2023 12:32:33 +0000 (13:32 +0100)] 
core: rework operation_to_signal() to rename return parameter ret_xyz

2 years agoMerge pull request #25912 from DaanDeMeyer/ukify
Daan De Meyer [Mon, 2 Jan 2023 15:07:58 +0000 (16:07 +0100)] 
Merge pull request #25912 from DaanDeMeyer/ukify

ukify fixes

2 years agomanager: perform objective->shutdown_verb mapping locally
Vito Caputo [Thu, 29 Dec 2022 01:53:22 +0000 (17:53 -0800)] 
manager: perform objective->shutdown_verb mapping locally

This is a small cleanup removing the need for the spurious
*ret_shutdown_verb argument on invoke_main_loop() while moving
the MANAGER_OBJECTIVE::shutdown_verb string mapping local to
where it actually gets added to the shutdown argv in
become_shutdown().

This also eliminates the need for the several clearings of
*ret_shutdown_argv, and the streq() branches of shutdown_verb in
favor of plain equality tests against the objective value.

Nothing functionally has been changed.

2 years agoukify: Fix section offset calculation
Jan Janssen [Sun, 1 Jan 2023 10:32:55 +0000 (11:32 +0100)] 
ukify: Fix section offset calculation

objcopy seems to expect that the offset passed to --change-section-vma
is absolute instead of relative to ImageBase. If this is not accounted
for an invalid image is created that cannot be loaded:

  0 .osrel        0000016b  0000000200016000  0000000200016000  00000400  2**2
  …
  6 .text         0000d242  0000000140001000  0000000140001000  00c6e800  2**4

This isn't an issue with gnu-efi based PE images, but natively created
ones will have a non-zero ImageBase.

2 years agoukify: Require specifying --tools for each tools directory 25912/head
Daan De Meyer [Mon, 2 Jan 2023 13:53:47 +0000 (14:53 +0100)] 
ukify: Require specifying --tools for each tools directory

Instead of consuming N arguments as tools directories, let's always
only consume one argument per specification of --tools. This avoids
issues where the linux image and initrd are interpreted as tools
directories.

2 years agoukify: Handle directories in path_is_readable()
Daan De Meyer [Mon, 2 Jan 2023 13:52:49 +0000 (14:52 +0100)] 
ukify: Handle directories in path_is_readable()

2 years agoUpdate 60-sensor.hwdb
Gio [Fri, 30 Dec 2022 22:23:50 +0000 (17:23 -0500)] 
Update 60-sensor.hwdb

Updated DBook D10 / Juno Tablet. Before the bvr (bios version) was hard coded to v1.05. This will read all versions

2 years agomachined: pass shell command line to polkit
Ludwig Nussel [Wed, 28 Dec 2022 13:00:35 +0000 (14:00 +0100)] 
machined: pass shell command line to polkit

Pass the joined arguments to polkit as command_line property like
pkexec does. Obviously not exactly perfect as it lacks quoting. Good
enough for polkit rules to check whether a program was called with
or without arguments though.

2 years agoMerge pull request #25864 from yuwata/dissect-fix-mtree
Luca Boccassi [Mon, 26 Dec 2022 16:43:27 +0000 (17:43 +0100)] 
Merge pull request #25864 from yuwata/dissect-fix-mtree

dissect: fix mtree

2 years agoPrevents airplane mode toggle for HP Spectre 16
Hannoskaj [Mon, 26 Dec 2022 08:43:36 +0000 (09:43 +0100)] 
Prevents airplane mode toggle for HP Spectre 16

2 years agoboot: don't convert the trailing newline in mangle_stub_cmdline()
Jia Zhang [Sun, 25 Dec 2022 04:29:11 +0000 (12:29 +0800)] 
boot: don't convert the trailing newline in mangle_stub_cmdline()

It is pretty convenient to add .cmdline using /proc/cmdline like
this:
  --add-section .cmdline=/proc/cmdline --change-section-vma .cmdline=0x25000

However, it always returns a trailing newline, and stub will
convert it to a whitespace by mangle_stub_cmdline() in next boot.
Thus the resulting /proc/cmdline would contain a trailing
whitespace. When /proc/cmdline is used to generate .cmdline again,
the resulting UKI is mangled.

To address this kind of inconvenience, mangle_stub_cmdline() should
skip converting the trailing newline, and try to chomp all the
trailing whitespaces.

Signed-off-by: Jia Zhang <zhang.jia@linux.alibaba.com>
2 years agodissect: reduce indent in mtree_print_item() 25864/head
Yu Watanabe [Mon, 26 Dec 2022 07:38:40 +0000 (16:38 +0900)] 
dissect: reduce indent in mtree_print_item()

2 years agodissect: do not trigger assertion on error
Yu Watanabe [Mon, 26 Dec 2022 07:41:17 +0000 (16:41 +0900)] 
dissect: do not trigger assertion on error

recurse_dir() does not provide 'sx' on error.

Fixes #25862.

2 years agodissect: use assert() when no side effect
Yu Watanabe [Mon, 26 Dec 2022 07:35:24 +0000 (16:35 +0900)] 
dissect: use assert() when no side effect

2 years agoMerge pull request #25850 from poettering/switch-root-tweaks-minor
Lennart Poettering [Sat, 24 Dec 2022 10:20:02 +0000 (11:20 +0100)] 
Merge pull request #25850 from poettering/switch-root-tweaks-minor

pid1: minor tweaks to the SwitchRoot() logic

2 years agoMerge pull request #25828 from poettering/no-more-basename
Yu Watanabe [Fri, 23 Dec 2022 23:13:00 +0000 (08:13 +0900)] 
Merge pull request #25828 from poettering/no-more-basename

convert a number of basename() invocations to path_extract_filename()

2 years agoprioq: make prioq_reshuffle() void
Lennart Poettering [Fri, 23 Dec 2022 17:49:02 +0000 (18:49 +0100)] 
prioq: make prioq_reshuffle() void

We never use the return value, and it's confusing and kinda pointless
what we return there.

Hence drop it.

Originally noticed by: tristone13th <tristone13th@outlook.com>

Alternative to: #25810

2 years agoMerge pull request #25849 from poettering/repart-loop
Lennart Poettering [Fri, 23 Dec 2022 22:26:00 +0000 (23:26 +0100)] 
Merge pull request #25849 from poettering/repart-loop

pull in loop.ko from repart unit

2 years agodocs: remove /dev/tty* confusion
Lennart Poettering [Fri, 23 Dec 2022 17:39:54 +0000 (18:39 +0100)] 
docs: remove /dev/tty* confusion

The text said /dev/tty* as a whole was the VT subsystem and that VT is
not supported in containers.

But that's not accurate as /dev/tty* will match /dev/tty too and that
one device node is special and is not related to VT: it always points to
the current process own controlling tty, regardless what that is.

hence, rewrite /dev/tty* as /dev/tty[0-9]*.

2 years agosysroot: Order systemd-fsck-root after systemd-makefs
Will Fancher [Sat, 17 Dec 2022 20:06:21 +0000 (15:06 -0500)] 
sysroot: Order systemd-fsck-root after systemd-makefs

2 years agosystemctl: rework how we detect if init is systemd 25850/head
Lennart Poettering [Fri, 23 Dec 2022 17:27:33 +0000 (18:27 +0100)] 
systemctl: rework how we detect if init is systemd

2 years agocore: use chase_symlinks_and_access() where appropriate
Lennart Poettering [Thu, 22 Dec 2022 09:30:02 +0000 (10:30 +0100)] 
core: use chase_symlinks_and_access() where appropriate

2 years agocore: tighten validation checks in SwitchRoot() dbus call
Lennart Poettering [Thu, 22 Dec 2022 09:28:05 +0000 (10:28 +0100)] 
core: tighten validation checks in SwitchRoot() dbus call

2 years agolock-file: various updates 25828/head
Lennart Poettering [Thu, 22 Dec 2022 09:56:56 +0000 (10:56 +0100)] 
lock-file: various updates

2 years agoanalyze: basename() → path_extract_filename()
Lennart Poettering [Wed, 21 Dec 2022 16:12:13 +0000 (17:12 +0100)] 
analyze: basename() → path_extract_filename()

2 years agoformat-table: add field type TABLE_PATH_BASENAME
Lennart Poettering [Wed, 21 Dec 2022 16:06:14 +0000 (17:06 +0100)] 
format-table: add field type TABLE_PATH_BASENAME

This is just like TABLE_PATH, but only shows the basename in regular
tabular output.

This is useful in systemd-repart for example

2 years agotmpfiles: port basename() → path_extract_filename()
Lennart Poettering [Wed, 21 Dec 2022 15:52:07 +0000 (16:52 +0100)] 
tmpfiles: port basename() → path_extract_filename()

2 years agosysv-generator: port basename() → path_extract_filename()
Lennart Poettering [Wed, 21 Dec 2022 15:51:48 +0000 (16:51 +0100)] 
sysv-generator: port basename() → path_extract_filename()

2 years agoshared: port various shared helpers basename() → path_extract_filename()
Lennart Poettering [Wed, 21 Dec 2022 15:51:05 +0000 (16:51 +0100)] 
shared: port various shared helpers basename() → path_extract_filename()

2 years agounits: pull in loop.ko and dm-mod.ko before repart 25849/head
Lennart Poettering [Fri, 23 Dec 2022 16:23:14 +0000 (17:23 +0100)] 
units: pull in loop.ko and dm-mod.ko before repart

We want to make use of that when formatting file systems, hence let's
pull in these modules explicitly.

(This is necessary because we are an early boot service that might run
before systemd-tmpfiles-dev.service, which creates /dev/loop-control and
/dev/mapper/control.)

Alternatively we could just order ourselves after
systemd-tmpfiles-dev.service, but I think there's value in adding an
explicit minimal ordering here, since we know what we'll need.

Fixes: #25775
2 years agounits: change modprobe@dm-mod.service → modprobe@dm_mod.service
Lennart Poettering [Fri, 23 Dec 2022 16:23:00 +0000 (17:23 +0100)] 
units: change modprobe@dm-mod.service → modprobe@dm_mod.service

Follow-up for 8f1359bf854e9683e4e0b89fd3a537e0d82d4b95

2 years agorepart/makefs: port over to path_extract_filename()
Lennart Poettering [Wed, 21 Dec 2022 15:50:18 +0000 (16:50 +0100)] 
repart/makefs: port over to path_extract_filename()

2 years agoportable: port basename() → path_extract_filename()
Lennart Poettering [Wed, 21 Dec 2022 15:49:30 +0000 (16:49 +0100)] 
portable: port basename() → path_extract_filename()

2 years agonspawn: port over basename() → path_extract_filename()
Lennart Poettering [Wed, 21 Dec 2022 15:49:03 +0000 (16:49 +0100)] 
nspawn: port over basename() → path_extract_filename()

2 years agomount-tool: port over basename() → path_extract_filename()
Lennart Poettering [Wed, 21 Dec 2022 15:48:25 +0000 (16:48 +0100)] 
mount-tool: port over basename() → path_extract_filename()

2 years agojournalctl: port COMM= matching to path_extract_filename()
Lennart Poettering [Wed, 21 Dec 2022 15:47:41 +0000 (16:47 +0100)] 
journalctl: port COMM= matching to path_extract_filename()

2 years agodissect: remove one more use of basename()
Lennart Poettering [Wed, 21 Dec 2022 15:46:28 +0000 (16:46 +0100)] 
dissect: remove one more use of basename()

And while we are at it, make use of path_extract_filename() return value
that indicates whether the path was suffixed with a slash.

2 years agocore: move some basename() use → path_extract_filename()
Lennart Poettering [Thu, 22 Dec 2022 17:45:27 +0000 (18:45 +0100)] 
core: move some basename() use → path_extract_filename()

2 years agotests: we don't actually use basename() anymore
Lennart Poettering [Wed, 21 Dec 2022 15:13:23 +0000 (16:13 +0100)] 
tests: we don't actually use basename() anymore

2 years agohomed: move away from basename()
Lennart Poettering [Wed, 21 Dec 2022 13:50:04 +0000 (14:50 +0100)] 
homed: move away from basename()

2 years agomeasure: fix the failures of compare_reported_pcr_nr()
Jia Zhang [Fri, 23 Dec 2022 09:23:36 +0000 (17:23 +0800)] 
measure: fix the failures of compare_reported_pcr_nr()

EFI_LOADER_VARIABLE() already applies to the stringify to construct the
path of EFI variable in efivars, so it is wrong to enclose the name of
EFI variable with the quotes. Otherwise, the following errors are
reported.

Reading EFI variable /sys/firmware/efi/efivars/StubFeatures-4a67b082-0a4c-41cf-b6c7-440b29bb8c4f.
Reading EFI variable /sys/firmware/efi/efivars/"StubPcrKernelImage"-4a67b082-0a4c-41cf-b6c7-440b29bb8c4f.
open("/sys/firmware/efi/efivars/"StubPcrKernelImage"-4a67b082-0a4c-41cf-b6c7-440b29bb8c4f") failed: No such file or directory
Reading EFI variable /sys/firmware/efi/efivars/"StubPcrKernelParameters"-4a67b082-0a4c-41cf-b6c7-440b29bb8c4f.
open("/sys/firmware/efi/efivars/"StubPcrKernelParameters"-4a67b082-0a4c-41cf-b6c7-440b29bb8c4f") failed: No such file or directory
Reading EFI variable /sys/firmware/efi/efivars/"StubPcrInitRDSysExts"-4a67b082-0a4c-41cf-b6c7-440b29bb8c4f.
open("/sys/firmware/efi/efivars/"StubPcrInitRDSysExts"-4a67b082-0a4c-41cf-b6c7-440b29bb8c4f") failed: No such file or directory

Signed-off-by: Jia Zhang <zhang.jia@linux.alibaba.com>
2 years agoMerge pull request #25823 from poettering/dissect-diskseq
Yu Watanabe [Fri, 23 Dec 2022 13:11:59 +0000 (22:11 +0900)] 
Merge pull request #25823 from poettering/dissect-diskseq

gpt-auto: reference dissected partitions to mount via diskseq block device symlinks

2 years agotree-wide: have_effective_cap() may return negative errno
Yu Watanabe [Fri, 23 Dec 2022 08:29:40 +0000 (17:29 +0900)] 
tree-wide: have_effective_cap() may return negative errno

2 years agoMerge pull request #25830 from yuwata/wait-online-unmanaged
Yu Watanabe [Fri, 23 Dec 2022 13:10:38 +0000 (22:10 +0900)] 
Merge pull request #25830 from yuwata/wait-online-unmanaged

wait-online: exit with success when all interfaces are ready or unmanaged

2 years agoMerge pull request #25777 from PeterCxy/enroll-fido2
Lennart Poettering [Fri, 23 Dec 2022 12:14:14 +0000 (13:14 +0100)] 
Merge pull request #25777 from PeterCxy/enroll-fido2

cryptenroll: Implement support for unlocking via FIDO2 tokens

2 years agoupdate TODO 25823/head
Lennart Poettering [Wed, 21 Dec 2022 13:32:22 +0000 (14:32 +0100)] 
update TODO

2 years agogpt-auto-generator: enable referencing partitions via diskseq symlinks
Lennart Poettering [Wed, 21 Dec 2022 11:28:31 +0000 (12:28 +0100)] 
gpt-auto-generator: enable referencing partitions via diskseq symlinks

2 years agodissect-image: optionally, reference dissected partition device nodes by diskseq
Lennart Poettering [Wed, 21 Dec 2022 11:26:21 +0000 (12:26 +0100)] 
dissect-image: optionally, reference dissected partition device nodes by diskseq

This is useful to make the dissection logic at boot a bit safer, as we
can reference device nodes by diskseq.

This locks down dissection a bit, since it makes it harder to swap out
the backing device between the time we dissected and validated it, until
we actually mounted it.

This is not complete though, as /bin/mount would have to verify the
diskseq after opening the diskseq symlink again.

See: https://github.com/util-linux/util-linux/issues/1786

2 years agodissect-image: simplify things by avoiding one string copy
Lennart Poettering [Wed, 21 Dec 2022 11:25:32 +0000 (12:25 +0100)] 
dissect-image: simplify things by avoiding one string copy

We don't need the node in its original variable anymore, hence let's
just move it over instead of allocating a copy.

2 years agoudev: also create partition /dev/disk/by-diskseq/ symlinks
Lennart Poettering [Wed, 21 Dec 2022 11:24:41 +0000 (12:24 +0100)] 
udev: also create partition /dev/disk/by-diskseq/ symlinks

2 years agoimport-creds: don't try to import VM creds in a container
Lennart Poettering [Thu, 22 Dec 2022 16:57:56 +0000 (17:57 +0100)] 
import-creds: don't try to import VM creds in a container

If we run in a container we shouldn#t try to import creds passed in from
a VMM, as they are not for us, but for the VM itself.

2 years agocryptenroll: Implement support for unlocking via FIDO2 tokens 25777/head
Peter Cai [Sun, 18 Dec 2022 01:33:05 +0000 (20:33 -0500)] 
cryptenroll: Implement support for unlocking via FIDO2 tokens

This allows FIDO2 users to wipe out password slots and still be able to
enroll new key slots via systemd-cryptenroll. Note that when the user
wants to both unlock with a FIDO2 token and enroll a new FIDO2 token,
they cannot be set to automatic discovery. This is to safeguard against
confusion, because there will be multiple tokens connected to the system
when doing so -- and we require users to explicitly confirm which one to
use for unlocking and which one to use for enrollment.

Addresses #20230 for the FIDO2 case.

2 years agoMerge pull request #25814 from DaanDeMeyer/ukify
Luca Boccassi [Thu, 22 Dec 2022 21:19:05 +0000 (22:19 +0100)] 
Merge pull request #25814 from DaanDeMeyer/ukify

ukify: Prefer using llvm-objcopy instead of objcopy

2 years agocryptsetup-fido2: Remove plain mode parameters from `acquire_fido2_key_auto()`
Peter Cai [Sun, 18 Dec 2022 01:28:51 +0000 (20:28 -0500)] 
cryptsetup-fido2: Remove plain mode parameters from `acquire_fido2_key_auto()`

`acquire_fido2_key_auto()` will not be used in PLAIN mode, and
parameters such as the salt will be acquired from the LUKS header.
Parameters intended for PLAIN mode are useless in
`acquire_fido2_key_auto()`.

2 years agocryptsetup-fido2: Relocate to libsystemd-shared
Peter Cai [Sat, 17 Dec 2022 20:46:36 +0000 (15:46 -0500)] 
cryptsetup-fido2: Relocate to libsystemd-shared

2 years agoukify: Validate that there are no overlapping sections 25814/head
Daan De Meyer [Wed, 21 Dec 2022 14:18:13 +0000 (15:18 +0100)] 
ukify: Validate that there are no overlapping sections

Let's make sure that after calling objcopy we have no overlapping
sections in the UKI

2 years agoukify: Prefer using llvm-objcopy instead of objcopy
Daan De Meyer [Mon, 19 Dec 2022 14:29:43 +0000 (15:29 +0100)] 
ukify: Prefer using llvm-objcopy instead of objcopy

llvm-objcopy works on stubs built for foreign architectures whereas
objcopy doesn't so let's prefer using llvm-objcopy instead of objcopy.

llvm-objcopy automatically sets the virtual address and doesn't provide
an option to set it manually so we only add --change-section-vma when
using objcopy

The default section flags differ between llvm-objcopy and objcopy
so we add a default for the section flags so we make sure all sections
are read-only data unless specified otherwise.

2 years agoukify: Allow passing multiple directories to --tools
Daan De Meyer [Mon, 19 Dec 2022 14:36:07 +0000 (15:36 +0100)] 
ukify: Allow passing multiple directories to --tools

2 years agoMerge pull request #25791 from keszybz/ukify-check-inputs
Luca Boccassi [Thu, 22 Dec 2022 11:10:32 +0000 (12:10 +0100)] 
Merge pull request #25791 from keszybz/ukify-check-inputs

ukify: check inputs

2 years agomacro: check existence of cleanup function before call it
Yu Watanabe [Mon, 19 Dec 2022 12:07:39 +0000 (21:07 +0900)] 
macro: check existence of cleanup function before call it

The free function specified in the macro may be provided by a
dynamically loaded library.

Replaces #25781.

2 years agotest-network: add testcase for no managed interface 25830/head
Yu Watanabe [Thu, 22 Dec 2022 07:42:55 +0000 (16:42 +0900)] 
test-network: add testcase for no managed interface

2 years agotest-network: split out wait-online related test case
Yu Watanabe [Thu, 22 Dec 2022 07:29:20 +0000 (16:29 +0900)] 
test-network: split out wait-online related test case

2 years agotest-network: move one test case
Yu Watanabe [Thu, 22 Dec 2022 07:27:58 +0000 (16:27 +0900)] 
test-network: move one test case

2 years agodissect-image: let's lock down fstypes a bit
Lennart Poettering [Wed, 21 Dec 2022 09:00:06 +0000 (10:00 +0100)] 
dissect-image: let's lock down fstypes a bit

When we dissect images automatically, let's be a bit more conservative
with the file system types we are willing to mount: only mount common
file systems automatically.

Explicit mounts requested by admins should always be OK, but when we do
automatic mounts, let's not permit barely maintained, possibly legacy
file systems.

The list for now covers the four common writable and two common
read-only file systems. Sooner or later we might want to add more to the
list.

Also, it might make sense to eventually make this configurable via the
image dissection policy logic.

2 years agoMerge pull request #25829 from poettering/empty-to-null-const-fix
Yu Watanabe [Thu, 22 Dec 2022 01:29:59 +0000 (10:29 +0900)] 
Merge pull request #25829 from poettering/empty-to-null-const-fix

string-util: don't add `const` to return pointer of empty_to_null() if input didn't have it

2 years agowait-online: exit with success when all interfaces are ready or unmanaged
Yu Watanabe [Thu, 22 Dec 2022 00:40:37 +0000 (09:40 +0900)] 
wait-online: exit with success when all interfaces are ready or unmanaged

Previously, when neither '--any' nor '--interface' options specified,
at least one of the interfaces must be in configured state.
This patch makes wait-online exit with success even if all interfaces
are in unmanaged state.

This may break backward compatibility in a rare situation. But in most
cases, this typically not change anything, as at least one interface
is managed by networkd.service when it is enabled.

This is mostly for making wait-online gracefully handle the case that
networkd.service and wait-online.service are enabled by mistake. In such
situation, all interfaces are typically not managed.

Fixes #25813.

2 years agocreds: use empty_or_dash() where appropriate 25829/head
Lennart Poettering [Wed, 21 Dec 2022 21:47:47 +0000 (22:47 +0100)] 
creds: use empty_or_dash() where appropriate

2 years agostring-util: rework empty_to_null() to not change "const" qualifier of input
Lennart Poettering [Wed, 21 Dec 2022 21:35:51 +0000 (22:35 +0100)] 
string-util: rework empty_to_null() to not change "const" qualifier of input

This changes the definition from enpty_to_null() so that we are still
typesafe (i.e. only accept strings) but do not drop (or add) any const
to the returned string that wasn't also on the input.

Inspired by: https://github.com/systemd/systemd/pull/25805/commits/3196e2996f613a2e3568a791c503306b7c58d593

2 years agogpt-auto-generator: honour rootfstype= and rootflags= kernel cmdline option
Lennart Poettering [Wed, 21 Dec 2022 08:47:46 +0000 (09:47 +0100)] 
gpt-auto-generator: honour rootfstype= and rootflags= kernel cmdline option

Even if root= is not specified on the kernel cmdline, we should honour
the other rootXYZ= options.

Fixes: #8411
See: #17034

2 years agohwdb: Add mic-mute, control-center and screen-rotation mappings for MSI laptops
Hans de Goede [Wed, 21 Dec 2022 11:18:52 +0000 (12:18 +0100)] 
hwdb: Add mic-mute, control-center and screen-rotation mappings for MSI laptops

The MSI Summit E16 Flip A12UCT laptop sends the following unmapped
atkbd scancodes:

0x91: Launch MSI Control Center
0xf1: Toggle mic mute
0xf2: Rotate screen

The 0x91, 0xf1 and 0xf2 codes are already present in the MSI Prestige/Modern
series specific keymappings and the 0xf1 mapping is also already present in
the MSI Bravo 15-B5DX FnKeys entry.

This shows that these are generic to many MSI models, so add mappings for
these to the generic MSI mappings.

Since the MSI Bravo 15-B5DX FnKeys entry only contains the 0xf1 mapping and
that is covered by the generic MSI mappings now, that entry is removed.

Link: https://gitlab.freedesktop.org/libinput/libinput/-/issues/822
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216824
2 years agohwdb: change definition of PROXIMITY_NEAR_LEVEL for sensors
Philipp Jungkamp [Wed, 21 Dec 2022 11:43:49 +0000 (12:43 +0100)] 
hwdb: change definition of PROXIMITY_NEAR_LEVEL for sensors

The [kernel documentation][0] for the in_proximity_nearlevel sysfs
attribute on iio proximity devices states:

    If the value read from the sensor is above or equal to the value in
    this file an object should typically be considered near.

Meaning a 'greater than or equal to' comparison.

Make the documentation comment in 60-sensors.hwdb suggest a
greater-or-equal rather than a strict greater-than comparison.

[0]: https://www.kernel.org/doc/Documentation/ABI/testing/sysfs-bus-iio-proximity

Fixes #25793

2 years agosystemctl: fix typo
Antonio Alvarez Feijoo [Wed, 21 Dec 2022 10:25:03 +0000 (11:25 +0100)] 
systemctl: fix typo

2 years agorules: add missing line continuation
Yu Watanabe [Wed, 21 Dec 2022 08:32:56 +0000 (17:32 +0900)] 
rules: add missing line continuation

Fixes a bug introduced by 953c928c24455744d5534679998d129b947a5e04.

Fixes #25811.

2 years agoman: create a new section for nspawn files in systemd.syntax man page (#25807)
Steve Ramage [Wed, 21 Dec 2022 04:31:09 +0000 (20:31 -0800)] 
man: create a new section for nspawn files in systemd.syntax man page (#25807)

Closes #25806.

2 years agoresolve: fix enumerator name for DNS search domain
Jacek Migacz [Tue, 20 Dec 2022 14:26:12 +0000 (15:26 +0100)] 
resolve: fix enumerator name for DNS search domain

2 years agoMerge pull request #25789 from yuwata/EBADF
Lennart Poettering [Tue, 20 Dec 2022 18:01:54 +0000 (19:01 +0100)] 
Merge pull request #25789 from yuwata/EBADF

tree-wide: use -EBADF more

2 years agoMerge branch 'systemd-security/coredump-capabilities'
Zbigniew Jędrzejewski-Szmek [Tue, 20 Dec 2022 16:56:28 +0000 (17:56 +0100)] 
Merge branch 'systemd-security/coredump-capabilities'

CVE-2022-4415: systemd: coredump not respecting fs.suid_dumpable kernel setting
Affects systemd >= 247 with libacl support enabled.

This is a merge of https://github.com/systemd/systemd-security/pull/12/.
I'm doing the merge locally because github doesn't support merging directly
from systemd/systemd-security to systemd/systemd.

2 years agotree-wide: use -EBADF more 25789/head
Yu Watanabe [Tue, 20 Dec 2022 02:09:08 +0000 (11:09 +0900)] 
tree-wide: use -EBADF more