]> git.ipfire.org Git - thirdparty/systemd.git/log
thirdparty/systemd.git
2 weeks agopcrextend: measure another separator at boot
Lennart Poettering [Tue, 10 Mar 2026 07:09:52 +0000 (08:09 +0100)] 
pcrextend: measure another separator at boot

This has been requested previously for PCR 7 (#40567), but let's do that
for all firmware owned PCRs, since some firmwares forget to measure
their own separator. Let's hence measure our own guranteed one.

Fixes: #40567
2 weeks agocreds-util: only lock against public key PCR stuff if we are booted with UEFI support...
Lennart Poettering [Tue, 10 Mar 2026 17:17:04 +0000 (18:17 +0100)] 
creds-util: only lock against public key PCR stuff if we are booted with UEFI supporting TPMs

The UKI public key PCR stuff only works if we get PCR measurements from
the pre-boot environment, hence automatically disable the logic by
default if we don't have that.

2 weeks agopcrlock: deal with firmwares which understand TPM but where no TPM is available
Lennart Poettering [Tue, 10 Mar 2026 07:07:25 +0000 (08:07 +0100)] 
pcrlock: deal with firmwares which understand TPM but where no TPM is available

This is a potentially common case in VMs: firmwares might know the
concept of TPMs, but the hardware is not enabled in the specific VM.
Let's handle this case nicely.

2 weeks agopcrlock: don't fail if firmware measurements aren't available
Lennart Poettering [Tue, 10 Mar 2026 07:04:52 +0000 (08:04 +0100)] 
pcrlock: don't fail if firmware measurements aren't available

With swtpm in place we now commonly have systems where TPM is available
during runtime, but not in the firmware. Handle that nicely.

2 weeks agocore: introduce ConditionSecurity=measured-os
Lennart Poettering [Mon, 9 Mar 2026 17:53:09 +0000 (18:53 +0100)] 
core: introduce ConditionSecurity=measured-os

So far we always conditioned our TPM magic on the UKI having detected
TPM support in the firmware. This is a bit limiting when we want to
support a software TPM that is not visible to the firmware. Hence let's
split this up, and add a separate control that can be set via the kernel
command line.  However, as before, let's by default inherit the firmare
TPM discovery state into it, to retain the current behaviour unless
overriden.

With this in place, boot with "systemd.tpm2_measured_os=1
systemd.tpm2_software_fallback=1" on the kernel cmdline to get the swtpm
fallback and then a measured OS based on it.

2 weeks agotree-wide: relax TPM available checks for many cases
Lennart Poettering [Mon, 9 Mar 2026 17:51:49 +0000 (18:51 +0100)] 
tree-wide: relax TPM available checks for many cases

In many cases it's essential to know if the firmware supports a TPM, but
in others we should accept it if the firmware doesn't have TPM support,
in particular if we want to run the OS with a software TPM.

Hence, add tpm2_is_mostly_supported() as function similar to
tpm2_is_fully_supported(), with the only difference that the former
doesn't insist on a firmware supported TPM. Then, change a number of
users over to this (but not all).

2 weeks agotpm2-generator: if requested run things with an swtpm
Lennart Poettering [Mon, 9 Mar 2026 12:08:28 +0000 (13:08 +0100)] 
tpm2-generator: if requested run things with an swtpm

We want to start the software TPM fallback only if no real hw is
evailable and if the user opts-in to this behaviour. Add a generator
that drives all this, based on kernel command line configuration.

2 weeks agotpm2: add "systemd-tpm2-swtpm" wrapper for "swtpm"
Lennart Poettering [Mon, 9 Mar 2026 12:06:58 +0000 (13:06 +0100)] 
tpm2: add "systemd-tpm2-swtpm" wrapper for "swtpm"

For TPM-less systems it's sometimes valuable to have a fill-in software
TPM running from early boot on, so that TPM-based functionality can
"just work" and rely on TPM semantics, even if it's at a substantially
weaker security level.

This adds a wrapper around swtpm. It's a binary that chainloads swtpm
but does a few preparatory steps and integrates into systemd's logic
otherwise.

All this is then exposed as systemd-tpm2-swtpm.service.

The service is not hooked into much yet, that is added in later commits.

2 weeks agolabeler: limit file-based label to 5 41358/head
Luca Boccassi [Thu, 26 Mar 2026 15:07:56 +0000 (15:07 +0000)] 
labeler: limit file-based label to 5

When doing large refactors or large changes the bot spams
labels left and right, making the PR unreadable. Use the
new option to limit the bot to a max of 5 file-based
labels. If more than 5 would be set, all file-based labels
are skipped.

2 weeks agolabeler: update to latest commit
Luca Boccassi [Thu, 26 Mar 2026 15:06:16 +0000 (15:06 +0000)] 
labeler: update to latest commit

Adds 'changed-files-labels-limit' and 'max-files-changed' configs

2 weeks agocreds: minor tweak for fputc() error handling 41329/head
Mike Yuan [Wed, 25 Mar 2026 15:03:07 +0000 (16:03 +0100)] 
creds: minor tweak for fputc() error handling

Let's do not assume errno is set for return values other than
EOF, following what we do in fileio.c.

2 weeks agocreds: if newline is explicitly requested, skip tty check
Mike Yuan [Wed, 25 Mar 2026 15:00:14 +0000 (16:00 +0100)] 
creds: if newline is explicitly requested, skip tty check

Before this commit, the > 0 state of arg_newline tristate is
simply ignored.

Yes, this is a minor compat break, but I'd argue the previous
behavior was not useful as "yes" is treated the same as "auto".
An issue also reported that it was quite surprising.

Fixes #41348

2 weeks agocreds: use parse_tristate_argument_with_auto() where appropriate
Mike Yuan [Wed, 25 Mar 2026 14:39:26 +0000 (15:39 +0100)] 
creds: use parse_tristate_argument_with_auto() where appropriate

2 weeks agocleanup-util: include assert-fundamental.h instead of -util
Mike Yuan [Wed, 25 Mar 2026 15:45:27 +0000 (16:45 +0100)] 
cleanup-util: include assert-fundamental.h instead of -util

The split was initially done to reduce transitive includes,
and for assert() only the former is needed.

2 weeks agocore: always call manager_log_caller() even without pidref 41354/head
Michael Vogt [Thu, 26 Mar 2026 11:51:23 +0000 (12:51 +0100)] 
core: always call manager_log_caller() even without pidref

Its fine if `manager_log_caller()` with an empty pidref,
it will log an unknown caller.

Thanks to @YHNdnzj for suggesting this.

2 weeks agocore: drop incorrect comment about SD_BUS_VTABLE_CAPABILITY(CAP_SYS_BOOT)
Michael Vogt [Thu, 26 Mar 2026 12:25:02 +0000 (13:25 +0100)] 
core: drop incorrect comment about SD_BUS_VTABLE_CAPABILITY(CAP_SYS_BOOT)

The comment about `SD_BUS_VTABLE_CAPABILITY(CAP_SYS_BOOT)` is
incorrect. To quote @YHNdnzj:
```
nah, the capability-based permission model is a legacy from kdbus. We cannot do it race-freely without it. Please simply drop the comment.
```

Thanks to @YHNdnzj for suggesting this.

2 weeks agocore: add assert(selinux_permission) to manager_do_set_objective
Michael Vogt [Thu, 26 Mar 2026 11:55:54 +0000 (12:55 +0100)] 
core: add assert(selinux_permission) to manager_do_set_objective

Thanks to @YHNdnzj for suggesting this.

2 weeks agocore: simplify manager_do_set_objective `root` path handling
Michael Vogt [Thu, 26 Mar 2026 12:16:57 +0000 (13:16 +0100)] 
core: simplify manager_do_set_objective `root` path handling

The manager_do_set_objective() was doing a bunch of work to
check the `root` path that can already be done via
`json_dispatch_path` so instead of duplicating use the helper.

Thanks to @YHNdnzj for suggesting this.

2 weeks agocore: drop `_manager()` from new vl_method to shutdown
Michael Vogt [Thu, 26 Mar 2026 11:46:56 +0000 (12:46 +0100)] 
core: drop `_manager()` from new vl_method to shutdown

Drop the _manager suffix for vl_method_{poweroff,{soft,}reboot,
halt,kexec}_manager.

They're unambiguous enough and the power mgmt operations are
not manager-wide but system-wide.

Thanks to @YHNdnzj for suggesting this.

2 weeks agogpt-auto-generator: generate an initrd ESP mount if it makes sense
Lennart Poettering [Mon, 9 Mar 2026 17:52:20 +0000 (18:52 +0100)] 
gpt-auto-generator: generate an initrd ESP mount if it makes sense

We need to store state persistently for the software TPM (i.e. the root
key). But given that TPMs are generally used to unlock the rootfs, this
storage cannot be on the rootfs. Hence let's use the ESP instead, as the
next best thing, that is guaranteed to exist during early boot, given we
just were booted from it.

This defines automatic logic for this, but does not cause the ESP mount
job to be enqueued (since typically we don't actually want that
mounted), this is left for the actual services that needs to be done.

Note that the mount here is set up quite differently from the one from
the host: since initrds are short-lived anyway, it seemed pointless to
use autofs. Moreover this uses a fixed place to mount the ESP, inspired
by the /sysroot/ + /sysusr/ mount naming. All that to simplify things a
bit for the consumers (which is mostly swtpm)

2 weeks agofileio: add WRITE_DATA_FILE_MODE_0400 mode
Lennart Poettering [Mon, 9 Mar 2026 12:06:42 +0000 (13:06 +0100)] 
fileio: add WRITE_DATA_FILE_MODE_0400 mode

2 weeks agokernel-install(uki): filter comments from cmdline
Cynthia [Tue, 17 Mar 2026 22:30:31 +0000 (23:30 +0100)] 
kernel-install(uki): filter comments from cmdline

This change aligns the behaviour of UKI generation with the behaviour
of BLS. The latter filters out lines starting with a #, allowing users
to add comments and/or temporarily remove some flags from the kernel
command line.

The kernel-install test have been adjusted to use a multiline cmdline
with a comment in it. Without this patch, the test fails.

2 weeks agoBasic IMDS support (#40980)
Zbigniew Jędrzejewski-Szmek [Thu, 26 Mar 2026 13:12:31 +0000 (14:12 +0100)] 
Basic IMDS support (#40980)

This adds a new cloud IMDS client, that can cover AWS, Azure, GCP,
Hetzner imds to varying degrees. Each cloud has this and it's very basic
functionality, hence I think it makes sense to add this to systemd.

Since the clouds are all different this tries hard to do the abstract
common logic in code, but encode the endpoint details, and well-known
keys in hwdb, attached to the DMI id device.

Why all this?

* Efficiency: we can schedule this in the initrd, at the earliest points
possible, without unnecessary delays
* Robustness: imds is typically slow and/or heavily rate-limited:
systemd-imdsd as single entrypoint can deal with that, and provide a
reliable, cached interface
* Security: the idea is that systemd-imdsd is the only service behing
able to access the IMDS HTTP, and the host carries a blackhole route for
it otherwise. That way sensitive info can be kept away from clients, and
requires polkit auth for access
* Simplicity: extraction of systemd's system credentials from IMDS
userdata happens with systemd's own infra, and for many usecases that
should already be enough.
* Measurements: before accepting the IDMS userdata, it can be measured
into a PCR, as any other configuration input for the system

Fixes: #40763
2 weeks agoci: Support multi-line review comments in claude-review
Daan De Meyer [Thu, 26 Mar 2026 12:33:38 +0000 (12:33 +0000)] 
ci: Support multi-line review comments in claude-review

Pass side, start_line, and start_side through to createReviewComment()
when present, enabling multi-line review comments. Update the prompt to
document all positioning fields using JSON Schema and make line required.

2 weeks agoRevert "ci: Add subject_type to createReviewComment()"
Daan De Meyer [Thu, 26 Mar 2026 12:19:15 +0000 (13:19 +0100)] 
Revert "ci: Add subject_type to createReviewComment()"

This reverts commit 211cd6e9a34d957dfa3b7616f0e618b6d17a51c2.

They document it here:
https://octokit.github.io/rest.js/v22/#pulls-create-review-comment
but apparently that's out of date and this doesn't work anymore.

2 weeks agoboot: add checks for invalid splash images in UKI
Luca Boccassi [Fri, 13 Mar 2026 01:52:12 +0000 (01:52 +0000)] 
boot: add checks for invalid splash images in UKI

A malformed bmp with 8bits depth but smaller color
map would cause out of bounds reads. This is not a real
problem as the image is signed, but better to be safe.

Reported on yeswehack.com as:
YWH-PGM9780-135

Follow-up for 0fa2cac4f0cdefaf1addd7f1fe0fd8113db9360b

2 weeks agoboot: fix typo in function name
Luca Boccassi [Fri, 13 Mar 2026 01:51:21 +0000 (01:51 +0000)] 
boot: fix typo in function name

Follow-up for dde03dd2a843b05d65885ce1242e43c8cabb9924

2 weeks agomkosi: Install clang-tidy package instead of clang-tools
Daan De Meyer [Thu, 26 Mar 2026 08:25:02 +0000 (09:25 +0100)] 
mkosi: Install clang-tidy package instead of clang-tools

clang-tools surprisingly enough doesn't provide clang-tidy

2 weeks agoupdate TODO 40980/head
Lennart Poettering [Thu, 5 Mar 2026 10:36:03 +0000 (11:36 +0100)] 
update TODO

2 weeks agotest: add simple integration test for systemd-imdsd
Lennart Poettering [Fri, 6 Mar 2026 16:31:10 +0000 (17:31 +0100)] 
test: add simple integration test for systemd-imdsd

2 weeks agoimds: add generator that hooks in IMDS logic on cloud guests
Lennart Poettering [Wed, 4 Mar 2026 14:16:14 +0000 (15:16 +0100)] 
imds: add generator that hooks in IMDS logic on cloud guests

The infrastructure added in the previous commits added support for IMDS
client functionality, but didn't really to enable the logic by default
on suitable hosts.

This commit adds a generator that automatically hooks the IMDS
functionality into the boot process if it detects that the system is
running on a compliant cloud system. it enables both the imds daemon and
the client.

2 weeks agoimds: add TPM measurements to imds tool
Lennart Poettering [Tue, 24 Mar 2026 08:41:03 +0000 (09:41 +0100)] 
imds: add TPM measurements to imds tool

This automatically measures the IMDS 'userdata' into PCR 12, i.e. where
we measure the other owner-supplied configuration, such as confexts and
credentials and similar.

(Why 12? It's really about who owns the data and what it is for.
PCRs/NvPCRs are scarce hence there's a strong incentive to not go
overboard with new allocations, and IMDS userdata in purpose and owner
is very very similar to confexts and credentials, hence let's reuse the
PCR for this purpose.)

2 weeks agoimds: add "systemd-imds" tool that is a simple client to "systemd-imdsd"
Lennart Poettering [Wed, 4 Mar 2026 14:13:25 +0000 (15:13 +0100)] 
imds: add "systemd-imds" tool that is a simple client to "systemd-imdsd"

This is a client tool to the systemd-imdsd@.service added in the
previous commit. It's mostly just a 1:1 IPC client via Varlink. It can
be used to query any IMDS key, but it's primary usecase is to acquire
the "userdata" from IMDS. Moreover, if invoked with the --import switch
it will check if the userdata contains a list of system credentials. If
so, it will import them into the local credstore. If the userdata does
not look like a list of system credentials no operation is executed,
under the assumption the data is intended for cloud-init instead.

It also imports a couple of other fields, if available and recogniuzed,
such as SSH keys and the hostname.

2 weeks agoimds: add new systemd-imdsd.service that makes IMDS data accessible locally
Lennart Poettering [Wed, 4 Mar 2026 14:10:37 +0000 (15:10 +0100)] 
imds: add new systemd-imdsd.service that makes IMDS data accessible locally

This service's job is to talk to a VM associated IMDS service provided
by the local Cloud. It tries to abstract the protocol differences
various IMDS implementations implement, but does *not* really try to
abstract more than a few basic fields of the actual IMDS metadata.

IMDS access is wrapped in a Varlink API that local clients can talk to.

If possible this makes use of the IMDS endpoint information that has
been added to hwdb in the preceeding commit. However, endpoint info can
also be provided via kernel command line and credentials. For debugging
purposes we also accept them via environment variables and command line
arguments.

This adds a concept of early-boot networking, just enough to be able to
talk to the IMDS service. It is minimally configurable via a kernel
cmdline option (and a build-time option): the user may choose between
"locked" and "unlocked" mode. In the former mode direct access to IMDS via
HTTPS is blocked via a prohibit route (and thus all IMDS communication
has to be done via systemd-imdsd@.service). In the latter case no such
lockdown takes place, and IMDS may be acquired both via this new service
and directly. The latter is typically a good idea for compatibility with
current systems, the former is preferable for secure installations.

Access to IMDS fields is controlled via PK.

2 weeks agohwdb: add database for basic IMDS properties
Lennart Poettering [Wed, 4 Mar 2026 14:07:09 +0000 (15:07 +0100)] 
hwdb: add database for basic IMDS properties

This adds a hardware database that contains information about IDMS
functionality of various clouds, keyed off the SMBIOS identification of
each. Currently this contains information about 6 major clouds, but the
idea is that this grows to include more and more major clouds.

Nothing uses this data yet, that's added in a later commit.

2 weeks agoshared/table-format: generalize table_sync_column_width to more columns
Zbigniew Jędrzejewski-Szmek [Wed, 25 Mar 2026 13:49:55 +0000 (14:49 +0100)] 
shared/table-format: generalize table_sync_column_width to more columns

The column index is moved to the first position. I think we're unlikely
to want to synchronize widths of *different* columns, and having just
one column argument makes the callers simpler. Also, the type is changed
to size_t to match other functions, and this avoids the need to cast
to size_t in the callers.

2 weeks agooptions: add common option macros for --no-ask-password, --host, --machine
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 08:49:24 +0000 (09:49 +0100)] 
options: add common option macros for --no-ask-password, --host, --machine

Co-developed-by: Claude <claude@anthropic.com>
2 weeks agoshared/verbs: introduce verb groups
Zbigniew Jędrzejewski-Szmek [Wed, 25 Mar 2026 12:57:54 +0000 (13:57 +0100)] 
shared/verbs: introduce verb groups

This mirrors the idea and implementation for options.

Previously, the Verb struct was named as verb_func_data_nn, but
with the group marker entry, we don't have 'verb_func', so let's
just call the item verb_data_nn. Using the verb here is a complication
that is not needed.

2 weeks agoshared/options: store argc+argv in the OptionParser state struct
Zbigniew Jędrzejewski-Szmek [Wed, 25 Mar 2026 15:35:12 +0000 (16:35 +0100)] 
shared/options: store argc+argv in the OptionParser state struct

After writing the code to parse options in a bunch of places, I think
passing the argc+argv to various functions to query state is annoying.
It seems nicer to just stash them in the state struct once.

2 weeks agoshared/options: stop removing items from argv array
Zbigniew Jędrzejewski-Szmek [Wed, 25 Mar 2026 07:15:21 +0000 (08:15 +0100)] 
shared/options: stop removing items from argv array

If we remove "--" from argv, the argc parameter stops being valid. But
that state is effectively global, albeit readonly, and somebody looking
at argv+argc after that will see an inconsistent state. Let's behave
like the libc parsing code and instead just shuffle things around in
argv, so that the argv+argc pair remains consistent.

This allows the code to calculate how many options are remaining to
be simplified.

2 weeks agoci: Add subject_type to createReviewComment()
Daan De Meyer [Thu, 26 Mar 2026 08:38:03 +0000 (09:38 +0100)] 
ci: Add subject_type to createReviewComment()

Apparently this is required by the createReviewComment() API.

2 weeks agoci: Use path instead of file in claude-review prompt as JSON key
Daan De Meyer [Thu, 26 Mar 2026 08:36:51 +0000 (09:36 +0100)] 
ci: Use path instead of file in claude-review prompt as JSON key

In https://github.com/systemd/systemd/pull/40980 claude hallucinated
and used "path" instead of "file" as the JSON key. Since "path" is
arguably more correct than "file" anyway, let's switch to that.

2 weeks agohwdb: Add PXN HB S handbrake
Luan Vitor Simião Oliveira [Wed, 25 Mar 2026 22:11:00 +0000 (19:11 -0300)] 
hwdb: Add PXN HB S handbrake

otherwise, it is not classified.
reports 2 axes and 2 buttons although only 1 is actually used.

2 weeks agotwo small resolved fixups (#41337)
Lennart Poettering [Thu, 26 Mar 2026 07:33:49 +0000 (08:33 +0100)] 
two small resolved fixups (#41337)

2 weeks agotpm2-util: fix PCR bank guessing without EFI
Patrick Wicki [Fri, 20 Mar 2026 14:56:56 +0000 (15:56 +0100)] 
tpm2-util: fix PCR bank guessing without EFI

Since 7643e4a89 efi_get_active_pcr_banks() is used to determine the
active PCR banks. Without EFI support, this returns -EOPNOTSUPP. This in
turns leads to cryptenroll and cryptsetup attach failures unless the PCR
bank is explicitly set, i.e.

$ systemd-cryptenroll $LUKS_PART --tpm2-device=auto --tpm2-pcrs='7'
[...]
Could not read pcr values: Operation not supported

But it works fine with --tpm2-pcrs='7:sha256'.

Similarly, unsealing during cryptsetup attach also fails if the bank
needs to be determined:

Failed to unseal secret using TPM2: Operation not supported

Catch the -EOPNOTSUPP and fallback to the guessing strategy.

Signed-off-by: Patrick Wicki <patrick.wicki@subset.ch>
2 weeks agopo: Translated using Weblate (Kabyle)
Massii Aqvayli [Wed, 25 Mar 2026 21:58:45 +0000 (21:58 +0000)] 
po: Translated using Weblate (Kabyle)

Currently translated at 36.4% (97 of 266 strings)

Co-authored-by: Massii Aqvayli <massiin@proton.me>
Translate-URL: https://translate.fedoraproject.org/projects/systemd/main/kab/
Translation: systemd/main

2 weeks agoresolved: move resetting of {etc_hosts|static_records}_last to manager_dispatch_reloa... 41337/head
Lennart Poettering [Wed, 25 Mar 2026 21:18:12 +0000 (22:18 +0100)] 
resolved: move resetting of {etc_hosts|static_records}_last to manager_dispatch_reload_signal()

This addresses
https://github.com/systemd/systemd/pull/41213#pullrequestreview-4002247053
which I somehow missed earlier.

Claude found a real issue for the case of manager_etc_hosts_flush().
We'll do the equivalent change in manager_static_records_flush() too,
even though it's not really necessary there, simply to keep things
nicely mirrored.

2 weeks agoresolved: fix typo in comment
Lennart Poettering [Wed, 25 Mar 2026 21:18:58 +0000 (22:18 +0100)] 
resolved: fix typo in comment

2 weeks agovmspawn: Use qemu config files
Daan De Meyer [Wed, 25 Mar 2026 13:20:03 +0000 (14:20 +0100)] 
vmspawn: Use qemu config files

Let's avoid generating giant qemu command lines by using qemu config
files instead.

2 weeks agovmspawn: Create journal parent directories if needed
Daan De Meyer [Wed, 25 Mar 2026 09:22:30 +0000 (10:22 +0100)] 
vmspawn: Create journal parent directories if needed

2 weeks agovmspawn: Fix --tpm-state= parsing
Daan De Meyer [Wed, 25 Mar 2026 14:16:55 +0000 (15:16 +0100)] 
vmspawn: Fix --tpm-state= parsing

path_startswith() considers "no" and "./no" equal. Use startswith()
to avoid that.

2 weeks agomeson: detect availability of attributes using cc.has_function_attribute()
Mike Yuan [Wed, 25 Mar 2026 17:13:11 +0000 (18:13 +0100)] 
meson: detect availability of attributes using cc.has_function_attribute()

Alternative to fabc22f5998e610eb7ba70a963cab9f94dca5c0a

As suggested in https://github.com/systemd/systemd/pull/41174#discussion_r2966411375

2 weeks agomemory-util: avoid passing invalid pointer to memcmp() when length == 16
Mike Yuan [Wed, 25 Mar 2026 17:30:14 +0000 (18:30 +0100)] 
memory-util: avoid passing invalid pointer to memcmp() when length == 16

If length is exactly 16, the loop would finish with length == 0,
but we'd carry on to the memcmp() check, where the 'p + 16' passed
would be invalid memory. memcmp() demands valid pointers even if
size is specified to 0, hence let's catch this ourselves.

2 weeks agoswtpm: Properly configure state directory (#41322)
Daan De Meyer [Wed, 25 Mar 2026 19:21:16 +0000 (20:21 +0100)] 
swtpm: Properly configure state directory (#41322)

2 weeks agoAGENTS: Tell agents to not use mkosi box
Daan De Meyer [Wed, 25 Mar 2026 14:43:35 +0000 (15:43 +0100)] 
AGENTS: Tell agents to not use mkosi box

It's easier to run the AI tool within mkosi box rather than telling
it to use mkosi box and forgetting to use it half the time.

2 weeks agosd-stub: support for a "boot secret" stored in a EFI variable, as a fallback for...
Lennart Poettering [Wed, 25 Mar 2026 17:24:07 +0000 (18:24 +0100)] 
sd-stub: support for a "boot secret" stored in a EFI variable, as a fallback for TPM-less systems (#41217)

Split out of #41016

2 weeks agostub: introduce "boot secret" stored in an EFI variable inaccessible to the OS 41217/head
Lennart Poettering [Sat, 7 Mar 2026 22:44:37 +0000 (23:44 +0100)] 
stub: introduce "boot secret" stored in an EFI variable inaccessible to the OS

2 weeks agorandom-seed: when we have a reasonable RNG then create random seed file if missing
Lennart Poettering [Sat, 7 Mar 2026 21:11:44 +0000 (22:11 +0100)] 
random-seed: when we have a reasonable RNG then create random seed file if missing

Previously we'd never write the ESP random seed file (or initialize the
random seed EFI table) if it didn't already exist. Let's adjust this a
bit, and also create it fresh if we have a "good" random source, i.e. if
the EFI table already existed or if the RNG protocol is implemented by
EFI.

This is useful as it increases the chance the random seed table is
valid, and we can use it as source for randomness in later stages.

2 weeks agoswtpm-util: Write our own CA options rather than using the distro ones 41322/head
Daan De Meyer [Wed, 25 Mar 2026 14:34:58 +0000 (15:34 +0100)] 
swtpm-util: Write our own CA options rather than using the distro ones

2 weeks agoswtpm: Properly configure state directory
Daan De Meyer [Wed, 25 Mar 2026 13:23:11 +0000 (14:23 +0100)] 
swtpm: Properly configure state directory

Otherwise it will unconditionally try to use /var/lib/xxx which won't
work when running unprivileged.

2 weeks agorandom-seed: move seed efi table definitions to header
Lennart Poettering [Sat, 7 Mar 2026 21:11:31 +0000 (22:11 +0100)] 
random-seed: move seed efi table definitions to header

2 weeks agoefi: add efivar_get_raw_full() flavour that returns the variable attributes too
Lennart Poettering [Fri, 6 Mar 2026 16:53:09 +0000 (17:53 +0100)] 
efi: add efivar_get_raw_full() flavour that returns the variable attributes too

2 weeks agomemory-util: move memeqbyte() & friends to src/fundamental/
Lennart Poettering [Fri, 6 Mar 2026 20:17:53 +0000 (21:17 +0100)] 
memory-util: move memeqbyte() & friends to src/fundamental/

2 weeks agotwo small prep commits from the imds PR (#41316)
Luca Boccassi [Wed, 25 Mar 2026 12:54:23 +0000 (12:54 +0000)] 
two small prep commits from the imds PR (#41316)

They are not directly IMDS stuff, and should be easy to merge ahead.

2 weeks agoAdd condition for mutable extensions directory
Hadi Chokr [Wed, 25 Feb 2026 14:27:24 +0000 (15:27 +0100)] 
Add condition for mutable extensions directory

Signed-off-by: Hadi Chokr <hadichokr@icloud.com>
2 weeks agovarlink: comment that "more" flag IDL comment is API
Michael Vogt [Wed, 25 Mar 2026 10:53:36 +0000 (11:53 +0100)] 
varlink: comment that "more" flag IDL comment is API

External tools that use the systemd varlink ecosystem require
to know if a specific varlink method supports/requires the
"more" flag from the IDL. This is tracked upstream in
https://github.com/varlink/varlink.github.io/issues/26

As an intermediate step systemd adds the (very nice) comments
```
 # [Requires 'more' flag]
or
 # [Supports 'more' flag]
```
to the various methods.

This commit extends the comment around the code that adds the
comment to clarify that this should be considered API and that
the comment should not be changed as external tools (like e.g.
the varlink-http-bridge) rely on it.

2 weeks agomacro.h: move DEFER_VOID_CALL() to cleanup-util.h 41316/head
Lennart Poettering [Tue, 24 Mar 2026 08:03:49 +0000 (09:03 +0100)] 
macro.h: move DEFER_VOID_CALL() to cleanup-util.h

For some reason the IMDS PR for the first time triggers an issue with
the DEFER_VOID_CALL() logic relying on assert() and being places in
macro.h, let's hence move this elsewhere.

2 weeks agotest-iovec: add unit test for IOVEC_MAKE_BYTE()
Lennart Poettering [Tue, 24 Mar 2026 07:57:26 +0000 (08:57 +0100)] 
test-iovec: add unit test for IOVEC_MAKE_BYTE()

As requested here: https://github.com/systemd/systemd/pull/40980#discussion_r2964650885

2 weeks agoA batch of conversions to the new options parsers (#41302)
Zbigniew Jędrzejewski-Szmek [Wed, 25 Mar 2026 10:39:36 +0000 (11:39 +0100)] 
A batch of conversions to the new options parsers (#41302)

This was partially done with Claude, but it needed an insane amount of prodding to
not do stupid things.

2 weeks agonetwork: increase transmit/receive queues size to 16384 (#41289)
Walter McKelvie [Wed, 25 Mar 2026 10:37:31 +0000 (06:37 -0400)] 
network: increase transmit/receive queues size to 16384 (#41289)

A 10G Marvell AQC113 included in an ASRock TRX50WS motherboard NIC claims to
support tx/rx queues as large as 8184.

After boot 'ethtool -g eth0' outputs:
  Ring parameters for eth0:
  RX: 8184
  RX Mini: n/a
  RX Jumbo: n/a
  TX: 8184
  TX push buff len: n/a
  HDS thresh: n/a
  RX: 2048
  RX Mini: n/a
  RX Jumbo: n/a
  TX: 4096
  RX Buf Len: n/a
  CQE Size: n/a
  TX Push: off
  RX Push: off
  TX push buff len: n/a
  TCP data split: n/a
  HDS thresh: n/a

'ethtool --set-ring eth0 rx 8184 tx 8184 && ethtool -g eth0' yields:
  Ring parameters for eth0:
  RX: 8184
  RX Mini: n/a
  RX Jumbo: n/a
  TX: 8184
  TX push buff len: n/a
  HDS thresh: n/a
  RX: 8184
  RX Mini: n/a
  RX Jumbo: n/a
  TX: 8184
  RX Buf Len: n/a
  CQE Size: n/a
  TX Push: off
  RX Push: off
  TX push buff len: n/a
  TCP data split: n/a
  HDS thresh: n/a

I can measure a throughput difference between using using buffer sizes
4096 and 8184 on my hardware, so it really seems that this is doing
something beyond buggy firmware.

Original PR https://github.com/systemd/systemd/pull/17635 didn't give any
explanation for the limit of 4096, but that's probably what was supported by
the kernel drivers at the time.

A web search shows that CISCO VIC 15000 supports 16k, so allow up to that.

[zjs: edited the message]

2 weeks agoresolved: use the SOA to find chain of trust quicker
Ronan Pigott [Wed, 11 Mar 2026 17:52:49 +0000 (10:52 -0700)] 
resolved: use the SOA to find chain of trust quicker

sd-resolved does dnssec "backwards" compared to most resolvers.

A typical strategy is to start from the DNS root and gather the
requisite keys on the way down, but sd-resolved requests the final
answer it wants and then goes searching for the requisite keys later.

We don't know in advance under which names we should expect to find
those keys, because we don't know the zone cuts a priori, but we can use
what we have found in prior responses to make an educated guess. This
was more or less the intent of 47690634f157, but it was partially
regressed in d840783db520 while fixing a bug handling totally empty
responses.

Fixes #37472

Ref: 47690634f157 ("resolved: don't request the SOA for every dns label")
Fixes: d840783db520 ("resolved: always progress DS queries")
2 weeks agohomectl: apply all --member-of= groups from a comma-separated list
Frantisek Sumsal [Tue, 24 Mar 2026 13:29:27 +0000 (14:29 +0100)] 
homectl: apply all --member-of= groups from a comma-separated list

Commit 0e1ede4b4b6d1ce6b5b6cda5f803e4f1b5aa4a03 introduced a bug where
we'd always fetch the "original" (empty) list of groups when processing
a comma-separated list of groups from the --member-of= option, so only
the last group from the list would get applied. This bug was then later
(in 316e9887f2a48bd1c4efa3e31b4bfbaeb22de3a3) refactored into a separate
function.

Follow-up for 0e1ede4b4b6d1ce6b5b6cda5f803e4f1b5aa4a03.
Resolves: #41286

2 weeks agoresolved: resolve insecure answers with unsupported sig algorithms (#40778)
Zbigniew Jędrzejewski-Szmek [Wed, 25 Mar 2026 09:38:08 +0000 (10:38 +0100)] 
resolved: resolve insecure answers with unsupported sig algorithms (#40778)

sd-resolved does not support all the permissible DNSSEC signature
algorithms, and some are intentionally unsupported as a matter of
policy. Answers that can only be validated via unsupported algorithms
should be treated as if they were unsigned, per RFC4035 § 5.2.

Previously, sd-resolved tried to properly record insecure answers for
unsupported algortihms, but did not record this status for each of the
auxilliary DNSSEC transactions, so the primary transaction had no way to
know if there was a plausible DNSKEY with an unsupported signature
algorithm in the chain of trust.

This commit adds the insecure DNSKEYs that use unsupported algorithms to
the list of validated keys for each transaction, so that dependent
transactions can learn that a plausible chain of trust exists, even if
no authenticated one does, and report the insecure answer.

This should improve the situation in #35126.

2 weeks agovarious: fix typos 41302/head
Zbigniew Jędrzejewski-Szmek [Wed, 25 Mar 2026 00:43:29 +0000 (01:43 +0100)] 
various: fix typos

2 weeks agobless-boot: use the new option parser and verb macros
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 09:20:52 +0000 (10:20 +0100)] 
bless-boot: use the new option parser and verb macros

Co-developed-by: Claude <claude@anthropic.com>
2 weeks agoshared/verbs: add VERB_COMMON_HELP_HIDDEN macro and skip verbs with NULL help
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 11:54:40 +0000 (12:54 +0100)] 
shared/verbs: add VERB_COMMON_HELP_HIDDEN macro and skip verbs with NULL help

Co-developed-by: Claude <claude@anthropic.com>
2 weeks agoshared/verbs: allow multiple verbs to be handled by a single function
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 22:33:49 +0000 (23:33 +0100)] 
shared/verbs: allow multiple verbs to be handled by a single function

With the uintptr_t data parameter, it is actually quite nice to have
VERB(do_impl, "name-a", …)
VERB(do_impl, "name-b", …)
int do_impl(…) { … }

To make this work, the do_impl_data struct needs to have a unique name and
we also need to suppress the warning about the forward declaration for
do_impl being repeated. I think it's fine to suppress the warning, it's
not needed for anything. If somebody declares the function with the same
name by mistake, the implementations are going to conflict too.

2 weeks agovalidatefs: shorten and reindent code
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 10:40:12 +0000 (11:40 +0100)] 
validatefs: shorten and reindent code

2 weeks agovalidatefs: use the new option parser
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 10:47:09 +0000 (11:47 +0100)] 
validatefs: use the new option parser

Co-developed-by: Claude <claude@anthropic.com>
2 weeks agoupdate-done: use the new option parser
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 10:07:06 +0000 (11:07 +0100)] 
update-done: use the new option parser

While at it, add --version.

Co-developed-by: Claude <claude@anthropic.com>
2 weeks agoask-password: use the new option parser
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 09:12:13 +0000 (10:12 +0100)] 
ask-password: use the new option parser

--version was missing from the old help text and is now included.

Co-developed-by: Claude <claude@anthropic.com>
2 weeks agoshared/options: allow option help to be extended with fake lines
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 10:01:37 +0000 (11:01 +0100)] 
shared/options: allow option help to be extended with fake lines

Useful when we want to enumerate options rvalues with custom help
texts.

2 weeks agobinfmt: use the new option parser
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 08:46:16 +0000 (09:46 +0100)] 
binfmt: use the new option parser

Co-developed-by: Claude <claude@anthropic.com>
2 weeks agoshared/options: add common option macros for --cat-config and --tldr
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 08:40:17 +0000 (09:40 +0100)] 
shared/options: add common option macros for --cat-config and --tldr

Co-developed-by: Claude <claude@anthropic.com>
2 weeks agoac-power: use the new option parser
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 08:34:15 +0000 (09:34 +0100)] 
ac-power: use the new option parser

Co-developed-by: Claude <claude@anthropic.com>
2 weeks agoshared/options: add helper function to count positional args
Zbigniew Jędrzejewski-Szmek [Tue, 24 Mar 2026 23:32:20 +0000 (00:32 +0100)] 
shared/options: add helper function to count positional args

2 weeks agoresolved: add "static RRs" concept (#41213)
Lennart Poettering [Tue, 24 Mar 2026 21:45:22 +0000 (22:45 +0100)] 
resolved: add "static RRs" concept (#41213)

split out of #40980

2 weeks agovmspawn: Fix --help width
Daan De Meyer [Tue, 24 Mar 2026 21:01:40 +0000 (22:01 +0100)] 
vmspawn: Fix --help width

2 weeks agovmspawn: add disk type selection for root and extra drives (#41301)
Lennart Poettering [Tue, 24 Mar 2026 20:48:00 +0000 (21:48 +0100)] 
vmspawn: add disk type selection for root and extra drives (#41301)

  vmspawn previously hardcoded virtio-blk for all drives. This adds
  --image-disk-type= to select the root disk type (virtio-blk,
  virtio-scsi, or nvme) and allows per-drive overrides via a
  colon-separated prefix on --extra-drive=. The format and disk type
  prefixes can appear in any order since their value sets don't overlap.

  For virtio-scsi, a single shared controller is created with drives
  attached as scsi-hd devices. For nvme, each drive gets its own
controller. Both have serial number length limits (30 and 20 characters
  respectively), so long filenames are replaced with a truncated SHA-256
  hex digest.

2 weeks agovmspawn: Add headless console support
Daan De Meyer [Mon, 23 Mar 2026 21:00:03 +0000 (22:00 +0100)] 
vmspawn: Add headless console support

2 weeks agopo: Translated using Weblate (Kabyle)
Massii Aqvayli [Tue, 24 Mar 2026 19:58:45 +0000 (19:58 +0000)] 
po: Translated using Weblate (Kabyle)

Currently translated at 22.5% (60 of 266 strings)

Co-authored-by: Massii Aqvayli <massiin@proton.me>
Translate-URL: https://translate.fedoraproject.org/projects/systemd/main/kab/
Translation: systemd/main

2 weeks agoresolved: add ability to define additional local RRs via drop-ins 41213/head
Lennart Poettering [Thu, 26 Feb 2026 14:54:14 +0000 (15:54 +0100)] 
resolved: add ability to define additional local RRs via drop-ins

This is an extension of the /etc/hosts concept, but can provide any kind
of RRs (well, actually, we only parse A/AAAA/PTR for now, but the
concept is open for more).

Fixes: #17791
2 weeks agoresolved: also flush /etc/hosts on reload
Lennart Poettering [Wed, 4 Mar 2026 16:08:03 +0000 (17:08 +0100)] 
resolved: also flush /etc/hosts on reload

When we are told to reload our configuration also flush out /etc/hosts
explicitly. This is particularly relevant since we suppress too frequent
reloads, and hence a synchronous way to force a reload is very useful.

2 weeks agodns-rr: add dns_resource_record_from_json()
Lennart Poettering [Thu, 26 Feb 2026 14:50:26 +0000 (15:50 +0100)] 
dns-rr: add dns_resource_record_from_json()

This only parses a small subset of RR types for now, but we can add more
later.

Covered are the most important RR types: A, AAAA, PTR.

2 weeks agodns-rr: tighten rules on parsing RR keys from JSON
Lennart Poettering [Thu, 26 Feb 2026 14:51:09 +0000 (15:51 +0100)] 
dns-rr: tighten rules on parsing RR keys from JSON

let's ensure the name is actually a valid DNS name.

2 weeks agojson-util: add json_dispatch_in6_addr()
Lennart Poettering [Fri, 20 Mar 2026 14:16:19 +0000 (15:16 +0100)] 
json-util: add json_dispatch_in6_addr()

2 weeks agojson-util: optionally accept string-based serialization for IPv4 addresses
Lennart Poettering [Fri, 20 Mar 2026 13:59:27 +0000 (14:59 +0100)] 
json-util: optionally accept string-based serialization for IPv4 addresses

2 weeks agostat-util: introduce inode_unmodified_hash_ops
Lennart Poettering [Tue, 24 Mar 2026 16:06:09 +0000 (17:06 +0100)] 
stat-util: introduce inode_unmodified_hash_ops

This is almost the same as inode_hash_ops, but also hashes + compares
all attributes that could affect the contents of a file. It ignores
"superficial"/"external" attributes such as ownership or access mode
however.

2 weeks agostat-util: also include inode type in hash ops
Lennart Poettering [Tue, 24 Mar 2026 15:56:40 +0000 (16:56 +0100)] 
stat-util: also include inode type in hash ops

This doesn't really have any major benefit, but it does make this nicely
mirror stat_inode_same() which also checks this triplet for identifying
identical inodes.

2 weeks agoci: Drop codeql workflow
Daan De Meyer [Mon, 23 Mar 2026 10:31:56 +0000 (11:31 +0100)] 
ci: Drop codeql workflow

After analyzing all 218 CodeQL alerts across the project's history, the
workflow has not justified its CI cost:

- The most impactful query (PotentiallyDangerousFunction) was a custom
  systemd-specific query that has already been replaced by clang-tidy's
  bugprone-unsafe-functions check (6fb5ec3dd1).

- Of the remaining C++ queries, 6 never triggered at all
  (bad-strncpy-size, unsafe-strcat, unsafe-strncat,
  suspicious-pointer-scaling, suspicious-pointer-scaling-void,
  inconsistent-null-check).

- Several high-value-sounding queries had extreme false positive rates:
  toctou-race-condition (95% FP), use-after-free (88% FP),
  cleartext-transmission (100% FP).

- Many queries that did trigger are already covered by compiler warnings
  (-Wshadow, -Wformat, -Wunused-variable, -Wreturn-type,
  -Wtautological-compare) or existing clang-tidy checks
  (bugprone-sizeof-expression).

- Across all alerts, only 3 genuinely useful C++ fixes can be
  attributed to CodeQL: 1 tainted-format-string, 2
  incorrectly-checked-scanf. The rest were either false positives or
  incidental fixes during refactoring that weren't prompted by CodeQL.

- The Python queries are largely superseded by ruff (already in CI) and
  had an 89% false positive rate on the security-focused checks.

The workflow consumed significant CI resources (40+ minutes per run) and
the ongoing maintenance burden of triaging false positives outweighs the
marginal value of the 2-3 real findings it produced across its entire
lifetime.