userdb: suppress userdb queries for backends indicating uid/gid/name range info via xattrs on entrypoint sockets (#42961)
Let's optimize userdb queries a bit: by encoding the covered UID/GID
ranges and user/group name patterns on the varlink entrypoint sockets
for userdb backends we can make them wake up less and reduce the work
triggered by queries.
repart: allow empty EncryptedVolume= volume name (#42889)
Treat an empty volume name alongside other fields as unset instead of
rejecting it as invalid.
Example use case:
```
EncryptedVolume=:none:discard
```
In this case, the volume name is not specified so it can be generated as
luks-UUID.
From the docs:
> EncryptedVolume=
> Specifies how the encrypted partition should be set up. Takes at least
one and at most four fields separated with a colon (":"). The first
field specifies the encrypted volume name under /dev/mapper/. If not
specified, "luks-UUID" will be used where "UUID" is the LUKS UUID.
process-util: introduce prctl_safe() and port everything over to it (#43006)
prctl() is an API full of pitfalls: it is variadic, and some interfaces
don't expect zero-initialization of excess arguments, and others do.
Moreover, the parameters are "long", and nonetheless we usually pass
"int" to them. If we are too dumb to call it properly, let's just not
call it directly anymore, but let's add a wrapper around it that makes
the function non-variadic and declares the right types. Then, let's port
over everything to it.
This is inspired by #42996, but we had issues with this many times,
before and looking at this PR one can see that we otherwise still are
having the issue at numerous other places.
cryptsetup: add Argon2id-based PIN mode for TPM2 enrollment (#41859)
The current TPM2 PIN mode is flawed as a compromised TPM directly
exposes
the sealed secret which is the LUKS volume key itself
(https://github.com/systemd/systemd/pull/27502 and
https://github.com/systemd/systemd/issues/37386).
Goal: add Argon2id-based PIN hardening to TPM2 enrollment, making
the TPM a second factor rather than a single point of failure:
1. Password + salt → Argon2id → 512-bit key split into Key1 + Key2
2. Key2 (base64-encoded) is used as the PIN to seal a random secret
in the TPM
3. Key1 + unsealed secret → HKDF-SHA256 → final LUKS volume key
This implementation ensures that if the TPM is compromised, an attacker
still needs the password to derive Key1 and combine it with the unsealed
secret.
The --tpm2-with-pin= option now accepts three values:
- false (no PIN used)
- true (PIN hardened with Argon2id - default)
- "direct" (legacy PIN without Argon2id for backward compatibility)
These default to a function of available CPUs and physical memory, with
a benchmark that scales iterations to the target time (default: 2s) and
falls back to ARGON2ID_PARAMETERS_DEFAULT (64 MiB, 8 iter, 4 lanes) when
auto detection fails.
Also if the runtime OpenSSL lacks Argon2id support (< 3.2), the feature
silently falls back to direct PIN mode with a warning.
Added includes:
- src/cryptenroll/cryptenroll.c: cpu-set-util.h, limits-util.h,
time-util.h
for Argon2id benchmark auto-tuning (cpus_online, physical_memory_scale,
now/usec_t)
- src/cryptenroll/cryptenroll-tpm2.c: crypto-util.h for
Argon2IdParameters
struct in load_volume_key_tpm2()
- src/shared/tpm2-util.h: crypto-util.h for Argon2IdParameters in
tpm2_make_luks2_json() API
- src/cryptsetup/cryptsetup-tokens/luks2-tpm2.c: crypto-util.h for
kdf_argon2id_derive()/kdf_hkdf_sha256() on the token unlock path
sysupdate: add support for --component-all + --feature-all + --feature-suggested to enable-feature verb
This adds similar concepts as we already have for enable-component:
let's add a way to enable all features or the suggeste dones, and
possibly on all components.
Or in other words, with this:
systemd-sysupdate enable-component -S
systemd-sysupdate enable-feature -A -s
We'll automatically enable all suggested components, and all features of
them.
sysupdate: implement --component-all and --component-suggested for enable-component and disable-component
Let's now introduce "systemd-sysupdate enable-component
--component-suggested" and systemd-sysupdate enable-component
--component-all" for enabling all or all suggested components at once.
Similar, if used for disable-component will disable all components or
those not suggested.
sysupdate: add a "suggests" concept to features and components
Let's make it possible to "suggest" that certain features or components
are enabled under some conditions.
For this, both features and components gain two things:
1. A Suggested= field which takes a boolean. If true the
feature/component will be suggested for installation, if false it
will not.
2. A set of SuggestedOnXYZ= settings are modelled after ConditionXYZ= in
unit files (and implement a subset of them), will suggest some
component/feature under specific conditions.
The result of the condition is shown in the various output tools.
3) timer_time_change() fires and calls timer_enter_waiting(t, true)
Because time_change=true, v->next_elapse is not recalculated and
keeps the original M0 + 60s value. The timer then correctly computes
the remaining time as ~12.8 seconds:
Jul 08 06:58:13 ...: Monotonic timer elapses in 12.785674s.
4) Timer elapses at 06:58:26
Jul 08 06:58:26 ...: Timer elapsed.
timer_enter_running() sets last_trigger.monotonic to CLOCK_MONOTONIC
(which equals to M0 + 12s before suspend + 13s after suspend, thus
+ 25s)
5) Unit deactivates at 06:58:29
timer_trigger_notify() calls timer_enter_waiting(t, false) -
time_change=false; that means that this time v->next_elapse gets
recalculated:
base = inactive_exit_timestamp.monotonic = M0 (i.e. when the timer was originally armed)
usec_shift_clock(M0, CLOCK_MONOTONIC, CLOCK_BOOTTIME_ALARM)
a = now(CLOCK_MONOTONIC) = M0 + 28s
b = now(CLOCK_BOOTTIME_ALARM) = M0 + 64s
result = b - (a - M0) = (M0 + 64s) - 28s = M0 + 36s => the time spent in suspend
is false, because the v->next_elapse (M0 + 96s) is not less than the
current time (M0 + 64s), so the timer is re-armed again:
Jul 08 06:58:29 ...: Monotonic timer elapses in 33.544681s
Let's mitigate this by skipping the next elapse timestamp recalculations
for one-shot timers for which we've already calculated the value in this
activation cycle.
Let's the common function for querying the local thread name.
This changes the escaping rules when the therad name is not quite
kosher: previously we'd just escape invalid UTF-8 charcaters, now we do
what we usually do: also escape control characters and such, and limit
us to ASCII.
The description generated here is mostly for debug purposes, and process
names should normally not require this escaping anyway (it's mostly
paranoia), hence I think this change in behaviour should be fine, it's
not part of the API in any form.
With 2c6f9af8e5425c2086fbc8ca496843f162e4af9b sd-varlink gained protocol
upgrade support, however only in a synchronous fashion. This adds
asynchronous protocol upgrade for the server side, thus enabling
multiplexing daemons (i.e. those that handle multiple connections from
the same event loop) to support protocol upgrades too.
I plan to use this in the upcoming "ptybroker" component that allows
acquiring a pty through varlink.
Jonas Dreßler [Thu, 9 Jul 2026 13:22:13 +0000 (15:22 +0200)]
repart: Don't copy trailing padding when using --copy-from=
Currently, --copy-from= copies the paddings in between the source partitions,
as well as the trailing padding that is at the end of the source partition table.
It doesn't copy the leading padding at the start of the source partition table
though.
This seems inconsistent, and likely it was an oversight that the trailing padding
is copied.
Fix that, and add a test to ensure we don't regress.
Jonas Dreßler [Tue, 7 Jul 2026 11:15:36 +0000 (13:15 +0200)]
repart: Clarify and test that --copy-from= argument respects grain size
The --copy-from= argument currently is documented as "copied partitions will have
the same size". This doesn't hold true in the case where a different grain-size is
passed to repart. Because `partition_min/max_size()` currently do rounding, the
size is implicitly rounded to grain size, and therefore partitions are enlarged
to align to grain size whenever possible.
Clarify this behavior and change the manpage, and also add a test for it.
Jonas Dreßler [Tue, 7 Jul 2026 12:41:16 +0000 (14:41 +0200)]
repart: Don't get old grain size from fdisk for --copy-from=
This is a little bit confusing, but grain size is not actually stored in the gpt
metadata. Rather, fdisk's `get_grain_size()` returns an autodiscovered "optimal io
size" value as grain size. This might not actually be the grain size that the
disk we're copying is using.
Since we're setting the padding of the copied partitions using that value from
fdisk, we're rounding the new paddings by fdisk's optimal grain size, which is
usually 1MiB (a lot more then the default 4KiB that we're using otherwise).
Set the grain size here to 1 byte instead, ensuring that the min/max padding set
is exactly the padding that was present before.
Also add a test to confirm the behavior is fixed: The test calls --copy-from= on
an existing disk with 4MiB grain size, and because we pass --grain-size=512, now
no rounding should happen and the paddings should be transferred to exactly the
same size.
tree-wide: add missing libopenssl_cflags to targets using tpm2-util.h
tpm2-util.h includes crypto-util.h, which requires OpenSSL compiler
flags when HAVE_OPENSSL is defined. Add libopenssl_cflags to every
build target that uses tpm2_cflags but was missing the transitive
OpenSSL dependency.
fuldeka [Sun, 28 Jun 2026 14:42:47 +0000 (16:42 +0200)]
cryptsetup: add Argon2id-based PIN mode for TPM2 enrollment
The current TPM2 PIN mode is flawed as a compromised TPM directly exposes
the sealed secret which is the LUKS volume key itself (#27502 and #37386).
Goal: add Argon2id-based PIN hardening to TPM2 enrollment, making
the TPM a second factor rather than a single point of failure:
1. Password + salt -> Argon2id -> 512-bit key split into Key1 + Key2
2. Key2 (base64-encoded) is used as the PIN to seal a random secret
in the TPM
3. Key1 + unsealed secret -> HKDF-SHA256 -> final LUKS volume key
This implementation ensures that if the TPM is compromised, an attacker
still needs the password to derive Key1 and combine it with the unsealed
secret.
The --tpm2-with-pin= option now accepts three values:
- false (no PIN used)
- true (PIN hardened with Argon2id - default)
- "direct" (legacy PIN without Argon2id for backward compatibility)
These default to a function of available CPUs and physical memory, with
a benchmark that scales iterations to the target time (default: 2s) and
falls back to ARGON2ID_PARAMETERS_DEFAULT (64 MiB, 8 iter, 4 lanes) when
auto detection fails.
Also if the runtime OpenSSL lacks Argon2id support (< 3.2), the feature
silently falls back to direct PIN mode with a warning.
Added includes:
- src/cryptenroll/cryptenroll.c: cpu-set-util.h, limits-util.h, time-util.h
for Argon2id benchmark auto-tuning (cpus_online, physical_memory_scale,
now/usec_t)
- src/cryptenroll/cryptenroll-tpm2.c: crypto-util.h for Argon2IdParameters
struct in load_volume_key_tpm2()
- src/shared/tpm2-util.h: crypto-util.h for Argon2IdParameters in
tpm2_make_luks2_json() API
- src/shared/tpm2-util.c: limits-util.h, tpm2-util.h for physical_memory() validation
of Argon2id memory cost and function prototypes
- src/cryptsetup/cryptsetup-tokens/luks2-tpm2.c: crypto-util.h for
kdf_argon2id_derive()/kdf_hkdf_sha256() on the token unlock path
homed: set user.userdb xattrs on homed's userdb socket
We can't make a lot of restrictions here, since users are allowed to
basically freely pick their user names and UIDs too. But let's at least
exclude system UID ranges and those beyond the 16bit range.
userdb: suppress lookups outside of indicated ranges
When doing a userdb lookup we might end up issuing a lot of IPC calls in
parallel to backends and cause them all to do work, with us waiting for
it. Let's optimize this a bit, and indicate on the socket inodes via
xattrs hints which kind of records are provided by a backend. That way
we can suppress lookups to them and optimize runtime behaviour.
This only works on Linux 7.1 and newer where socket inodes gained
support for extended attributes.
This are to the existing sd_varlink_call_and_upgrade() what
sd_varlink_callb() and sd_varlink_callbo() are to sd_varlink_call():
they put together an object on the fly, via the usual JSON builder
logic.
ci: check 'update-man-rules' to ensure it is not forgotten
We often forget to run this command when updating manpages, and notice
only much later. Add a step in the CI to flag it, as we already do for
the D-Bus docs.
Cache the IPv6 enabled state while sizing and filling NSS result
buffers, so a transient sysctl read result cannot change the tuple or
address layout after the scratch buffer size has been computed. Also
zero-initialize gaih_addrtuple records before filling IPv4 addresses.
Automatic key modes tolerate a failed TPM2 sealing attempt and
fall back to a host or null key. Do not consume TPM2 blob output in
that case, tpm2_seal() leaves it empty on failure, so the fallback
path should continue without TPM2 metadata.
NDisc packets received through the socket path are filtered before
parser dispatch, but parser entry points should still reject malformed
packet bytes instead of asserting on them. Return EBADMSG for non-zero
ICMPv6 code values in RA, NA, and Redirect messages.
An OCI index redirect already carries the digest of the selected
manifest. Store it in the expected checksum field so pull-job
verifies the downloaded manifest instead of overwriting the digest
with the computed checksum before comparison.
efi_get_variable() allocates one byte for probing whether efivarfs
has grown since fstat(), then three more bytes for NUL termination.
Account for both sizes separately so a full readv() result is treated
as concurrent growth and retried before the terminators are written.
network: do not regenerate MAC address if already set by userspace
When MACAddress= is unspecified, a stable MAC is generated to seed a newly
created netdev. Since existing netdevs are reconfigured on reload/restart,
this seed got re-applied to an already existing interface, conflicting with
an explicit MACAddress= from the matching .network file and flipping MAC.
Now, when MACAddress= is unspecified, only generate a new address if
addr_assign_type is not NET_ADDR_SET (i.e. not already set by userspace). On
first creation the interface does not exist yet, so this is a no-op and the
address is generated as before. Mirrors src/udev/net/link-config.c.
Popax21 [Thu, 19 Mar 2026 02:33:26 +0000 (03:33 +0100)]
creds-util: implement TPM2 SRK pinning
Stores the TPM2 SRK within the credential header, allowing for parameter decryption to be utilized when decrypting the credential.
A new dimension is added to the credential ID matrix to encode this capability.
This also allows for usage of TPM2-bound credentials when a TPM owner password is set since `Esys_CreatePrimary` is no longer used for sealing credentials.
Michael Vogt [Fri, 10 Jul 2026 15:27:11 +0000 (17:27 +0200)]
creds: allow normal users to encrypt with`--with-key=null`
When encrypting with the `--with-key=null` option systemd-creds
is currently doing the encryption via IPC. This is not needed
for the null key, no privs are required so we can just do the
in-process operation. So instead simply check for the null-key and
if its requested use the in-process path.
* b322b8d98e Install new files for upstream build
* 3fd1b81c94 Update changelog for 261.1-3 release
* 8f95f1370c Move modules-load from systemd package to udev package
* bcdf90c670 debian/libpam-systemd.postinst: pam-auth-update does not use getopt
* 743e3399ac d/README.source: document policy for adding new binary packages
* cbea74783c Install new files for upstream build
* ef267b3ad6 debian/libpam-systemd.postinst: run pam-auth-update with --root=$DPKG_ROOT
* 54df2859b4 d/t/upstream: use mkosi from archive for Ubuntu autopkgtest
* 21959e8a59 Fix zsh installation path, again
* f01dddb047 Fix zsh installation path
* e31737edc4 Install new files for upstream build
* ca0630a51e Update changelog for 261.1-2 release
* 1ebb987599 d/t/control: pull in cpio for upstream suite
* fcf5a24f47 Two more fixups for d/copyright
* 6ad198086d d/t/control: pull new packages in upstream test suite
* bb9fd757fd Make systemd actually temporarily depend on systemd-tpm
credentials: add policy that can allow key=null creds from the ESP (#42555)
This PR only sets the default to "relaxed" - I can change the default
to "tofu" if desired. But for that we will also need to update the NEWS
file to ensure everyone is aware of this new default.
---
This PR adds a new `systemd.credentials-boot=` kernel
commandline that allows to control if credentials with
a `null` key are accepted.
The possible options are:
* strict: always insist on tpm encryption
* tofu: allow null encryption in firstboot mode and when no tpm is
available
* relaxed: allow null encryption when sb is off, or no tpm is available
* off: allow null encryption always
The default is `relaxed` which is exactly the behavior we had before.
This replaces the initial idea of using plaintext credentials
at firstboot (thanks to Lennart for this nicer and simpler design).
---
With that we can drop `- firstboot: optionally accept credentials at
firstboot without authentication` from TODO.md
These are philosophically similar concepts: one deletes old versions
based on whether they are old, and the other deletes old files based on
whether they are orphaned, let's list them together.
Let's switch to string_is_safe(), and make this available to the rest of
the sysupdate sources too.
This both relaxes and tighten the rules slightly. i.e. control character
and stuff are no longer allowed, but valid UTF-8 (as opposed to ASCII)
now is.
sysupdate: explicitly refuse --root=/--image= in conjunction with --definitions=
This is currently effectively not supported (as we sometimes prefix the
definitions path with root and sometimes not), let's make this official
for now and refuse it. We could in theory support the combination but I
don't see the big benefit of it for now, hence let's just refuse it.
Kai Lüke [Fri, 3 Jul 2026 13:39:22 +0000 (22:39 +0900)]
sysupdate: Evaluate all patterns before descending into a subdir
When we have two match patterns, e.g., because the repository layout
changed and we look for old and new style files, we can have the case
that one pattern suggests a descend into a subdir but the other pattern
would be a direct match. This was missed and only the descent was done.
Yet the other way round also has to be covered: We can't just use the
direct match because it might be that it's rejected if the type doesn't
fit (directory vs. regular file). In this case we should still descent.
Remember that we want to descend but continue checking the other
pattern first to maybe get a direct match. Introduce a new combined
return code YES_AND_RETRY to give the full information to the caller
which now can decide to fallback to a descent. While at it, use the
stat info for the directory check. Also add tests to ensure that the
order of patterns doesn't matter and we handle the above corner cases.
Kai Lüke [Thu, 14 May 2026 14:28:38 +0000 (23:28 +0900)]
sysupdate: Support matching for filenames in subdirectories
While for sysupdate it's fine to consume a large set of all possible
update payloads in a single directory this is not so handy for managing
and serving the update payloads. Since this large update folder is not
where the build output directly gets written to one has to create
copies and later possibly delete this added set of files.
Support matching for filenames in subdirectories by having a new **/
match pattern prefix which matches any number of nested subdirectories
or no subdirectory at all. For simplicity it's only allowed at the start
of a pattern and not a regular wildcard as the rest because the main
use case is to descend into subdirectories and only do the pattern
matching for the basenames. This way one can create a SHA256SUMS file
in the top folder and have it include all update payloads from the
release-specific (or arch-specific) subdirectories. Something similar
was already supported for directory sources where the match pattern can
start with a subdirectory path. Do also support this for SHA256SUMS for
parity while we are at it. Having the new wildcard makes mirroring also
easier because one does not have to follow the exact subdirectory layout
and one can filter by folder instead of by filename. It also makes it
possible to point the same transfer files with the new wildcard to
either a SHA256SUMS file that uses release-specific (or arch-specific)
subdirectories and includes all versions or to a SHA256SUMS file as
generated from mkosi that does not use subdirectories because it only
has files for a single version.
With the upcoming UAPI.16 JSON format we will also be able to encode
subdirectories and it makes sense to add this to SHA256SUMS for being
able to convert them. It also supports custom URLs for each entry which
is more powerful than the (arbitrary) subdirectory feature used here but
subdirectories have the advantage that they don't break mirroring.
This change also fixes the existing subdirectory handling bugs where
everything greater than two subdirectory levels failed to work because
rel_joined instead of de->d_name got used, symlinks were followed, and
we would continue silently on non-ENOENT errors.
Simran Singh [Wed, 10 Dec 2025 00:44:53 +0000 (06:14 +0530)]
clonesetup: add support to clone devices via /etc/clonetab
Adds dm-clone device setup at boot via a new /etc/clonetab config file,
following the crypttab/veritytab pattern.
- Add systemd-clonesetup-generator to parse /etc/clonetab and generate units.
- Add systemd-clonesetup binary to create/remove dm-clone devices via ioctl.
- Add clonesetup.target for ordering dm-clone activation at boot.
- Add region_size= option in clonetab to configure dm-clone hydration granularity.
- Add clonetab(5) and systemd-clonesetup-generator(8) man pages.
Commit 9b917abe02505d4ad09c4963ec5a4b2744eb2fee (udev-builtin:
simplify code a bit) changed builtin lookup from copying the first
command word and comparing it with streq() to comparing only the first
command word length with strneq().
That made prefixes such as "k" or "key" resolve to the "keyboard"
builtin, depending on the builtins table order. This was not the
original behavior.
Require the matched builtin name to end at the first command word
boundary so only complete builtin names are accepted. Arguments after
the builtin name continue to work as before.
Extend ConditionCPUFeature to support architecture prefixes in feature names.
The goal is avoiding ambiguities on mixed-arch fleets, where CPU feature names
may potentially conflict.
With this patch, ConditionCPUFeature=arm64.bti is now equivalent to
ConditionCPUFeature=bti on arm64 systems.
Add an override file for all capabilities missing in glibc v2.34 and musl
1.2.6. The constant AT_HWCAP3 is also not in glibc v2.34, ship a glibc-specific
elf.h in order to provide it.
Optimize dlopen ELF notes via anchoring and explicitly embed them into executables (#42908)
This PR optimizes the handling of `dlopen` ELF notes and reduces binary
footprints across the tree.
By enabling compiler section-splitting
(`-ffunction-sections`/`-fdata-sections`), the linker can now accurately
garbage-collect unused code and data. To align with this,
`SD_ELF_NOTE_DLOPEN_ANCHORED()` macro is introduced, which ties `dlopen`
notes to their calling functions so that unused notes are automatically
removed by `--gc-sections`.
Additionally, this explicitly embeds required `dlopen` notes into
individual executables to fix a visibility issue where package managers
missed runtime dependencies invoked indirectly through
`libsystemd-shared.so`.
--on-clock-change and --on-timezone-change are documented as shortcuts for
setting the corresponding timer properties with --timer-property=.
However, --timer-property= only treated the monotonic and calendar timer
settings as actual timer triggers. OnClockChange= and OnTimezoneChange=
were stored as timer properties, but arg_with_timer remained false and the
command was rejected as having no timer options.
Treat both properties as timer triggers too, matching the shortcut options
and the documented equivalent command line.
nspawn-oci: match the spec-correct "swappiness" memory field key
The OCI runtime specification names the memory swappiness knob
"swappiness" (memory.swappiness), but the dispatch table
in oci_cgroup_memory() registered it as "swapiness".
Since sd_json_dispatch_full() matches object keys by exact string
comparison, a spec-correct config carrying "swappiness" never hit the
intended oci_unsupported() handler and was instead routed to the
bad-field callback oci_unexpected(), which returns -EINVAL and thus
fails the whole memory cgroup section.
Register the correct key so that real OCI bundles parse, keep the
misspelt one around for compatibility, mark both as such, and drop the
now-resolved reminder from the file header TODO list.
Configure an IMPORT{program} rule whose output exceeds UDEV_LINE_SIZE
and ends with an incomplete property line, for example:
TRUNCATED_OK=yes
TRUNCATED_BAD=<very long value without terminating newline>
Only TRUNCATED_OK should be imported. Before this fix, the truncation
handler edited the command buffer instead of the output buffer, so
TRUNCATED_BAD could still be parsed from the truncated result.
condition: split condition_test_list() to minimize dlopen dependencies
Even though systemd-networkd and systemd-udevd do not support
ConditionSecurity= in .network, .netdev, and .link files, the unified
condition_test() function maintained a function pointer table that
unconditionally referenced condition_test_security(). Consequently,
linker garbage collection (--gc-sections) could not drop the security
test code, inadvertently pulling in dlopen dependencies and ELF notes
for apparmor, audit, and tpm2 libraries into these binaries.
To resolve this, introduce condition_test_net() and condition_test_list_net(),
which utilize a trimmed-down function pointer table that excludes
security-related condition evaluators.
By migrating networkd and udevd to these new network-specific variants,
the reference to condition_test_security() is completely severed in
their dependency chains. This allows the compiler and linker to
successfully garbage-collect the unused security logic and safely drops
the unnecessary dlopen notes from those binaries.
compress: split compress_blob() and friends to minimize dlopen dependencies
Previously, even though sd-journal does not support bzip2 and gzip
compression, the dlopen ELF notes for those libraries were still being
attached to libsystemd.so and various journal-handling executables.
This occurred because the unified compression/decompression interfaces
handled all formats unconditionally, causing the compiler and linker to
pull in all associated dlopen notes across the board.
To resolve this, split these functions into generic and journal-specific
variants (e.g., introducing compress_blob_journal() and decompress_blob_journal()).
The journal-specific variants only handle formats actually supported by
the journal (LZ4, XZ, and ZSTD).
By updating sd-journal, journald, journalctl, and related utilities to
use these new `_journal` interfaces and switching them to the narrower
COMPRESS_JOURNAL_NOTE macro, we ensure that unnecessary dlopen notes
(for bzip2 and gzip) are no longer embedded into these binaries.
Enable the `dlopen` unit test suite in GitHub Actions, except for the
following configurations where linker garbage collection (`--gc-sections`)
or dead-code elimination fails to drop unused dlopen symbols:
This introduces a test utility to validate the presence and correctness
of dlopen ELF notes.
The test extracts and compares dlopen ELF notes between the dynamic and
static versions of target binaries to catch missing entries. Furthermore,
it scans the code for dlopen_foo() invocations to ensure that no stale or
unused notes remain in the final binary, thereby verifying the linker's
garbage-collection integration.
sd-dlopen: introduce _dlopen_loader_ macro to prevent inlining and cloning
When compiler optimization or LTO (Link Time Optimization) is enabled,
dlopen helper functions (such as dlopen_libfoo()) can be aggressively
inlined into their callers or cloned for specific call sites.
While this is beneficial for production builds, it alters or completely
removes the original function symbols from the final executable.
Consequently, this breaks the test (to be added in this series) that
checks whether the dlopen_libfoo() function corresponding to a dlopen
ELF note is actually called.
To resolve this, introduce the `_dlopen_loader_` macro. Under developer
builds, it applies the `_noclone_` and `_noinline_` attributes to guarantee
that helper symbols remain intact and discoverable by the test suite. In
production builds, the macro evaluates to nothing, allowing full compiler
optimization to proceed unimpeded.
tree-wide: embed dlopen notes into individual executables
Previously, when a library was dynamically loaded via a helper function
inside libsystemd-shared.so, the resulting dlopen ELF note was not
propagated to the invoking executable's ELF metadata.
This commit explicitly annotates each executable with the relevant
dlopen ELF notes for any optional dependencies it might potentially
load. This ensures that package managers and build systems can properly
discover runtime dependencies that are triggered indirectly.
The new SD_ELF_NOTE_DLOPEN_ANCHORED() macro utilizes the 'o' assembler
flag (SHF_LINK_ORDER), which requires binutils >= 2.35 or LLVM >= 18.
If an LLVM version older than 18 is encountered, the macro automatically
falls back to the non-anchored variant.
This non-anchored fallback relies on the 'R' (SHF_GNU_RETAIN) flag,
which requires binutils >= 2.36 or LLVM >= 13.
Since the codebase now unconditionally adopts SD_ELF_NOTE_DLOPEN_ANCHORED()
tree-wide, the effective minimum toolchain requirements become:
- binutils >= 2.35 (to support the 'o' flag)
- LLVM/Clang >= 13 (to support the 'R' flag fallback for versions < 18)
Update the minimal toolchain versions in the README to reflect these
requirements for building systemd.
projects / thirdparty / systemd.git / log
