This also switches all these notes to be defined via SD_ELF_NOTE_DLOPEN_ANCHORED().
Consequently, any notes added within unreachable or unused functions will be
automatically garbage-collected by the linker (--gc-sections) instead of
bloating the final binary.
E.g. unused p11-kit library dependency is now dropped from
systemd-repart.standalone binary.
Before:
```
$ systemd-analyze dlopen-metadata build/systemd-repart.standalone
FEATURE DESCRIPTION SONAME PRIORITY
cryptsetup Support for disk encryption, integrity, and authentication libcryptsetup.so.12 recommended
blkid Support for block device identification libblkid.so.1 required
libcrypto Support for cryptographic operations libcrypto.so.4 libcrypto.so.3 recommended
mount Support for mount enumeration libmount.so.1 required
fdisk Support for reading and writing partition tables libfdisk.so.1 required
blkid Support for block device identification libblkid.so.1 recommended
libcrypto Support for cryptographic operations libcrypto.so.4 libcrypto.so.3 suggested
cryptsetup Support for disk encryption, integrity, and authentication libcryptsetup.so.12 suggested
fdisk Support for reading and writing partition tables libfdisk.so.1 suggested
idn Support for internationalized domain names libidn2.so.0 suggested
mount Support for mount enumeration libmount.so.1 recommended
selinux Support for SELinux libselinux.so.1 recommended
tpm Support for TPM libtss2-esys.so.0 suggested
tpm Support for TPM libtss2-rc.so.0 suggested
tpm Support for TPM libtss2-mu.so.0 suggested
tpm Support for TPM libtss2-tcti-device.so.0 suggested
p11-kit Support for PKCS11 hardware tokens libp11-kit.so.0 suggested
```
After:
```
$ systemd-analyze dlopen-metadata build/systemd-repart.standalone
FEATURE DESCRIPTION SONAME PRIORITY
cryptsetup Support for disk encryption, integrity, and authentication libcryptsetup.so.12 recommended
blkid Support for block device identification libblkid.so.1 required
libcrypto Support for cryptographic operations libcrypto.so.4 libcrypto.so.3 recommended
mount Support for mount enumeration libmount.so.1 required
fdisk Support for reading and writing partition tables libfdisk.so.1 required
blkid Support for block device identification libblkid.so.1 recommended
libcrypto Support for cryptographic operations libcrypto.so.4 libcrypto.so.3 suggested
cryptsetup Support for disk encryption, integrity, and authentication libcryptsetup.so.12 suggested
fdisk Support for reading and writing partition tables libfdisk.so.1 suggested
mount Support for mount enumeration libmount.so.1 recommended
selinux Support for SELinux libselinux.so.1 recommended
tpm Support for TPM libtss2-esys.so.0 suggested
tpm Support for TPM libtss2-rc.so.0 suggested
tpm Support for TPM libtss2-mu.so.0 suggested
tpm Support for TPM libtss2-tcti-device.so.0 suggested
```
Extend the 'install' keyword for executable definitions to accept four
modes—'yes', 'no', 'both', and 'static'—to elegantly manage the creation
and installation of both shared and statically-linked (.standalone) binaries.
- 'yes' / 'no': Standard behavior (mapped to true/false).
- 'both': Installs both the shared and static variants.
- 'static': Installs the static variant under the original name, while
suffixing the uninstalled shared variant with '.shared'.
With this change, any arbitrary executable can now have its `.standalone`
variant built on demand simply by invoking `ninja <target>.standalone`.
For example, `varlinkctl` did not previously support a standalone variant,
but it can now be built explicitly via `ninja varlinkctl.standalone`.
These `.standalone` binaries are not built by default unless explicitly
specified as a ninja target or enabled via `-Dstandalone-binaries=true`.
Thus, the default build time should remain unaffected.
This centralisation eliminates a massive amount of boilerplate and duplicated
target declarations across almost all subdirectories (e.g., systemd-repart,
systemd-tmpfiles, systemd-shutdown, and systemd-report tools).
By chance (because I had this mistake in my own commit) I noticed
that there are a bunch of duplicated works like "the the" in the
code and man-pages.
This commit fixes them and some similar issues with "and and" etc.
Recursive credential directory loading intentionally
includes socket entries, but load_credential()
only enabled AF_UNIX socket handling for absolute paths.
Let the recursive directory path request socket
connections explicitly, and update the stale comment for directory-fd reads.
Signed-off-by: dongshengyuan <dongshengyuan@uniontech.com>
When a rate limit is armed, for example via StartLimitInterval=, and the
machine is suspended, the rate limit's "clock" is suspended as well,
since the elapse checks use CLOCK_MONOTONIC. This then causes unexpected
situations where a rate limit armed via StartLimitInterval=1h, followed
by a 10 hour suspend, would elapse after 11 hours total.
Let's avoid this by switching the rate limits to CLOCK_BOOTTIME, which
works the same as CLOCK_MONOTONIC, but accounts for the time spent in
suspend as well.
There's one slight concern when it comes to upgrade path - the old
"begin" value of the rate limit is stored as CLOCK_MONOTONIC, but after
upgrading systemd and serializing/deserializing the state it will be
suddenly compared against now(CLOCK_BOOTTIME), which might cause some
rate limits to elapse "prematurely". But this is just a one-time thing,
after which the rate limit timers should re-assess themselves.
Resolves: #42912
Add StandardInputFileDescriptor, StandardOutputFileDescriptor and
StandardErrorFileDescriptor to the Service context of the Varlink
StartTransient() method. Each carries the push-order index of a file descriptor
passed along with the method call (via SCM_RIGHTS), and connects it to the
transient service's standard input/output/error. This is the Varlink equivalent
of the StandardInputFileDescriptor= / StandardOutputFileDescriptor= /
StandardErrorFileDescriptor= D-Bus transient properties behind
"systemd-run --pipe", which had no Varlink counterpart.
The manager Varlink server now enables SD_VARLINK_SERVER_ALLOW_FD_PASSING_INPUT
so clients may attach descriptors to their method calls, matching other
fd-accepting Varlink services (mountfsd, networkd, vmspawn, ...). The indices are
resolved with sd_varlink_peek_dup_fd() after the polkit authorization check and
stored on the Service exactly like the D-Bus path
(bus_set_transient_exec_context_fd()): the fds land in stdin_fd/stdout_fd/
stderr_fd and exec_context.stdio_as_fds is set, so the existing exec-invoke
plumbing wires them to the spawned process unchanged. Unsuitable fds (bad
index, wrong access mode) are rejected as InvalidParameter, other resolution
failures propagate as raw errnos.
Add coverage to TEST-74-AUX-UTILS.varlinkctl-unit.sh, passing regular files as
the stdout/stderr fds and asserting the unit's output lands on them.
This makes each `foo_cflags` dependency define a `SYSTEMD_CFLAGS_MARKER_FOO`
macro, and checks if the macro is set when headers provided by external
libraries are included.
With this, we can fail fast at compile time if necessary `_cflags`
dependencies are omitted in meson.build. Missing dependencies found by this
mechanism have been added across the tree.
Overriding <linux/audit.h> to implicitly include <libaudit.h> turns out to
be problematic. Because <linux/audit.h> is pulled in by various other kernel
headers, any source file including those kernel headers indirectly ends up
depending on <libaudit.h>, even if the executable or library being built
does not require libaudit at all.
Let's drop the override header and include <libaudit.h> only where explicitly
needed. This also moves the fallback definitions of AUDIT_SERVICE_* and
MAX_AUDIT_MESSAGE_LENGTH to libaudit-util.h.
When a fixed date (e.g. 2027-01-01) is paired with a weekday constraint
(e.g. Thu) that does not match, the timer silently never elapses.
Add calendar_spec_from_string_full(..., warn_on_weekday_mismatch) so
user-facing tools can opt in to a log_warning() at parse time:
- systemd-analyze calendar: uses _full(true)
- systemd-run --on-calendar: uses _full(true)
- .timer OnCalendar=: uses log_syntax() with file/line context
Add test_calendar_spec_weekday_conflict(): forks a child with stderr
captured in a memfd via pidref_safe_fork_full(), verifies the warning
is emitted for conflicting specs and suppressed for valid ones.
Fixes: #40350
Signed-off-by: dongshengyuan <dongshengyuan@uniontech.com>
When PID 1 reaches the shutdown phase, if systemd-shutdown was missing,
it'd complain loudly and freeze. Implement an internal fallback to
handle this case gracefully. The fallback is always in place, since it
makes things generally more robust and is only a little bit of code.
This shutdown is simplified, systemd issues the reboot() matching the
requested objective directly, mirroring what systemd-shutdown would do
as its final step. This skips all the cleanup systemd-shutdown normally
performs (unmounting file systems, detaching storage, killing remaining
processes, switching into the exitrd).
With SYSTEMD_LOG_LEVEL=debug:
...
[ OK ] Reached target poweroff.target.
Notify message sent to '/run/host/notify': "X_SYSTEMD_UNIT_ACTIVE=poweroff.target"
Shutting down.
OSC sequence for shutdown successfully written: ...
Failed to execute shutdown binary: No such file or directory
Powering off.
Container container has been shut down.
systemd tries to start ctrl-alt-del.target after receiving a request
to reboot in a container (or when ctrl-alt-del is pressed on a real
machine). If the alias is missing, this will fail. Let's provide a
fallback for this too. I don't think it makes sense to make it
configurable: it seems almost everybody is fine with the default
mapping. If not, people can always just create the symlink in the
filesystem.
The "and" in the name was a misnomer. We do one *or* the other.
So rename 's/_and_/_or_/', and extend with the ability to load a second
unit. No functional change.
If we try to reexecute, and the binary was started from a non-standard
path, this would fail:
src/core/main.c:2396: Reexecuting.
src/core/main.c:2240: Failed to execute our own binary /usr/lib/systemd/systemd: No such file or directory
Let's make things more robust by trying /proc/self/exe too.
Add a log_level parameter so callers can control the severity of load
failures. During boot, when no unit was explicitly requested on the
kernel command line (arg_default_unit is unset), failing to load the
default unit is not fatal since we fall back to other targets, so log
those attempts at LOG_INFO instead of LOG_ERR. All other callers pass
LOG_ERR to keep their existing behaviour.
Compile the contents of a few essential target unit files
(graphical.target, multi-user.target, rescue.target, etc.) into the
manager binary, with comments stripped, and fall back to those built-in
copies when no fragment is found on disk. This lets the manager reach a
usable state even on a system that ships none of these unit files, e.g.
in minimal environments or one-off chroots.
On-disk files always take precedence (masks included), as the fallback is
only consulted when nothing is found in the lookup paths. The unit file
list is defined in units/meson.build and embedded via a small generator
script; 'units' is now processed before 'src/core' so the list is
available there.
The set of units and services is selected that should allow for a normal
operation of starting/stopping/restarting, of the machine. The services
all use SuccessAction=…, so they don't require any binaries to be
installed.
sigpwr.target is included, even though it doesn't do anything useful.
The same is true in normal installations. We might want to change it
do something useful there too.
Add a 'fallback-default-target' meson option that configures which unit
to activate when default.target is not installed. This makes things more
resilient in general and is also useful for minimal/statically-linked
installations that may not ship a default.target symlink.
'graphical.target' is used as the default value of the setting, because
that's what we symlink as default.target in units/meson.build.
In the initrd, we had a fallback to start default.target if initrd.target
cannot be started. This fallback is changed to only do that if it is
not found, not on other errors. This seems more correct (and makes
the two fallbacks symmetrical.)
After stripping, in an unoptimized build with glibc, we get a 8.7 MB binary:
$ file build-static/systemd
build-static/systemd: ELF 64-bit LSB executable, x86-64, version 1 (GNU/Linux), statically linked, BuildID[sha1]=1ec004bf61b775d284d909a818a3c4f63f83f055, for GNU/Linux 3.2.0, stripped
$ ls -lh build-static/systemd
-rwxr-xr-x 1 zbyszek zbyszek 8.7M Jun 30 14:07 build-static/systemd
The libc API is rather ugly: it requires the caller to pass an
array of the maximum possible size (or to repeatedly query the group
list to discover the actual size). With our "static-friendly" helpers,
we reallocate the array on the fly.
This effectively reverts the meson config changes introduced in
0c6186695a. I wasn't happy with that form
back when it was commited, and in hindsight, I like it even less. Let's
just add a new option that hopefully is easier to understand and will
work better when new functionality is added in the future.
Now there are two options:
link-executor-shared (boolean, defaults to true)
— "link (separate) systemd-executor to libsystemd-shared.so and libsystemd-core.so"
systemd-multicall-binary (also boolean, defaults to false)
— "link systemd+systemd-executor as a single binary"
Build directories will need to be recreated again, sorry!
The inner map is created with a uint32_t key, but the update passed
&magic[i] where magic is a (possibly 64-bit) statfs_f_type_t. On
little-endian the low 32 bits happen to be read; on big-endian 64-bit
(s390x, ppc64 BE) the zero high word is read instead, so every
filesystem collides on key 0 (the allow/deny selector) and
RestrictFileSystems= is silently broken. Pass a truncated copy.
Follow-up for 184b4f78cf
While a delegated scope waits for the async cgroup chown helper it sits
in SCOPE_START_CHOWN (UNIT_ACTIVATING). unit_start() forwards to
->start() while activating, so scope_start() could be re-entered in
that state and trip assert(s->state == SCOPE_DEAD), aborting PID 1.
Treat SCOPE_START_CHOWN as already-starting instead.
Follow-up for 03860190fe
do_reexecute() freed the FDSet on the switch-root/soft-reboot fallback
but the caller's copy stayed live, so main() freed it again if every
fallback exec then failed. Donate the fdset instead: pass it with
TAKE_PTR() and take ownership via a _cleanup_ local, freeing it exactly
once on every exit path.
Follow-up for 3c7878f94b
unbase64mem() leaves p/sz uninitialized on failure, and the
deserialization doesn't bail out in that case, unlike other cases.
Follow-up for d247f232a8
Similar to PR#42360 this commit adds missing properties for
absolute path handling in io.systemd.Unit.StartTransient for
the `Exec` context and a macro helper to share the common code.
The properties added are:
IPCNamespacePath, NetworkNamespacePath, RootDirectory, RootHashPath,
RootHashSignaturePath, RootImage, RootVerity, UserNamespacePath,
RootMStack.
Note that RootHashPath, RootHashSignaturePath need custom apply
functions because the varlink name "RootHashPath" differs from the
name that needs to be written into the unit file ("RootHash=")
and the iovec must be cleared.
SMBIOS OEM strings are host-controlled and normally distrusted by PID1
in confidential VMs. Under TDX, however, the SMBIOS table (including
Type 11) is measured into RTMR0 by the firmware (TDVF), so a remote
verifier can detect host tampering with credentials delivered this
way. Accept them there, while keeping fw_cfg distrusted as those
items are not measured even on TDX.
This lets systemd-vmspawn deliver credentials to TDX guests via the
normal SMBIOS path, unlike SNP which requires the initrd cpio channel.
Signed-off-by: Paul Meyer <katexochen0@gmail.com>
Add more shutdown timestamps, export them via D-Bus/Varlink, use them in
`systemd-analyze time`, and preserve them across kexec via LUO. This is
useful to measure reboot performance with added measurement points and
more granular intervals.
Until now the SHUTDOWN_START/FINISH timestamps were carried across a
soft-reboot in the same fields, so afterwards they described the
previous boot rather than the current one, mixing current- and
previous-boot data in the same set. Mirror what is now done on kexec
and move them into the PREVIOUS_SHUTDOWN_* fields on deserialization.
Also reset local timestamps for events that rerun (eg: generators,
units loading, etc) on soft-reboot and switch-root.
We already record when shutdown.target is initiated (ShutdownStart).
This adds a few more measuring points during shutdown: when
shutdown.target completes (i.e., all units stopped), when the shutdown
binary starts executing, and immediately before it hands off to the
kernel.
Obviously these are immediately lost upon reboot. But they will later
be wired in LUO so they get preserved on kexec, and can be inspected,
which is very useful for performance measurements.
When setenv("CREDENTIALS_DIRECTORY") fails, log_error_errno() was
called with the stale return value of exec_context_get_credential_directory()
(which is >= 0 on success) instead of errno.
The %m in the format string correctly expands from libc's errno, so
the human-readable log message was unaffected. However, the structured
journal field ERRNO= received an incorrect value (0 or positive),
making automated log analysis and alerting on this failure unreliable.
Signed-off-by: dongshengyuan <dongshengyuan@uniontech.com>
Add support for two newer NUMA memory policies:
- MPOL_PREFERRED_MANY (Linux 5.15): like MPOL_PREFERRED but accepts
a set of nodes instead of a single node, falling back to all nodes
if preferred nodes cannot satisfy the allocation.
- MPOL_WEIGHTED_INTERLEAVE (Linux 6.9): like MPOL_INTERLEAVE but
distributes pages across nodes proportionally to per-node weights
configured via /sys/kernel/mm/mempolicy/weighted_interleave/.
On kernels that do not support the requested policy, set_mempolicy()
returns EINVAL. We convert EINVAL to EOPNOTSUPP only for the two new
policies (MPOL_PREFERRED_MANY, MPOL_WEIGHTED_INTERLEAVE), so that a
bad NUMAMask= for already-supported policies still fails the service
rather than being silently ignored.
The NUMA subsystem being absent (ENOSYS) continues to be handled
silently at debug level, as before.
Varlink serialization uses json_underscorify() on an owned copy of
the policy name string to convert hyphenated names to the underscore
form declared in the IDL enum, avoiding mutation of the read-only
static string table.
Signed-off-by: dongshengyuan <dongshengyuan@uniontech.com>
Re-enables `--set-credential=` / `--load-credential=` under
`--coco=sev-snp` by packaging credentials into a cpio appended to the
initrd, mirroring what `systemd-stub` does for ESP-sourced credentials.
The initrd is covered by the launch measurement via `kernel-hashes=on`,
so the credentials are too.
Tested end-to-end on an SNP-capable host: credentials passed via
`--set-credential=` land in `/run/credentials/@encrypted/` inside the
guest.
The userns/netns/ipcns fdpairs were declared as plain int arrays without
_cleanup_close_pair_. If exec_shared_runtime_add() fails (e.g. OOM on
hashmap_ensure_put), the already-opened fds are leaked.
Since exec_shared_runtime_add() uses TAKE_FD on success, the array
entries are reset to -1 after ownership transfer, so adding
_cleanup_close_pair_ is safe and closes the fds only when they were
never consumed.
Signed-off-by: dongshengyuan <dongshengyuan@uniontech.com>
The strv_sort() call sat after a for (;;) loop whose only exits are
return statements inside the loop, so it never ran.
CID#1660125
Follow-up for 82b8615463
Mirror how chaseat() works these days: instead of a single toplevel_fd that
serves as both the root (chroot) boundary and the directory that resolution
starts from, path_pick() now takes a separate root_fd and dir_fd. This lets
callers resolve a path relative to a specific directory fd while confining
symlink and absolute-path resolution to a root directory fd.
All existing callers are updated to pass the same fd for both, preserving
their current behaviour.
PID1's import_credentials_boot() so far always treated initrd-delivered
credentials as untrusted: anything found under /.extra/credentials/
was routed into ENCRYPTED_CREDENTIALS_DIRECTORY, forcing consumers to
use LoadCredentialEncrypted= and provide credentials in systemd-creds
encrypted form. That matches the trust model for stub which sources
credentials from the EFI System Partition (mountable and editable
offline).
Host-side producers that take responsibility for the trust of the cpio
they hand to the kernel have a different model: e.g. systemd-vmspawn
builds an initrd-credentials cpio whose bytes are covered by the SEV-SNP
launch measurement via QEMU's kernel-hashes=on (or, in non-confidential
setups, by the host itself being the trust root). Forcing those through
the @encrypted bucket would require null-key wrapping on the host and
LoadCredentialEncrypted= on the consumer side, a needless API split for
unit files that should otherwise be portable between confidential and
non-confidential boots.
Extend import_credentials_boot() to import credentials from
/.extra/system_credentials/, routed into SYSTEM_CREDENTIALS_DIRECTORY.
Consumers access these via LoadCredential= directly, with no
encrypted-credential ceremony.
The per-directory walk is hoisted into a small static helper so the new
target bucket can share the existing copy/validation logic; behavior
for the existing untrusted paths is unchanged.
Also set the new RECURSE_DIR_MUST_BE_REGULAR flag as a drive-by.
Signed-off-by: Paul Meyer <katexochen0@gmail.com>
The clear-store in restrict_fsaccess_clear_initramfs_trust() writes a fixed
4 bytes (*(uint32_t *)(p + INITRAMFS_S_DEV_OFF) = 0). INITRAMFS_S_DEV_OFF is
derived from the skeleton, so the offset tracks any field widening, but the
store width does not: were initramfs_s_dev widened (e.g. __u32 -> __u64) in
the BPF program, the store would clear only the low 4 bytes and silently
leave the initramfs trust window partially open. That is exactly the class
of bug the mirror-struct asserts (removed earlier in this branch) guarded
against.
Add a compile-time assert pinning the store width to the skeleton field
width (sizeof_field(typeof_field(struct restrict_fsaccess_bpf, bss[0]),
initramfs_s_dev) == sizeof(uint32_t)), so widening the field fails the build
instead of clearing half of it.
Building with -Dbpf=enabled -Dbpf_compiler=gcc (GCC's BPF backend) fails on
the static assertions in bpf-restrict-fsaccess.c, introduced in 68fe7fa4d6:
error: static assertion failed:
"offsetof(struct restrict_fsaccess_bss, initramfs_s_dev) ==
offsetof(typeof_field(struct restrict_fsaccess_bpf, bss[0]), initramfs_s_dev)"
The hand-written struct restrict_fsaccess_bss lists the BPF .bss globals in
source declaration order and asserts that its layout matches the skeleton's
generated bss struct. bpftool gen skeleton emits that struct from the BTF
.bss DATASEC, whose member order reflects the physical order the compiler
placed the variables, not the source order. clang preserves declaration
order, so the asserts pass; gcc reorders .bss globals, so initramfs_s_dev no
longer sits at offset 0 and the asserts fail.
This is more than a build break: restrict_fsaccess_clear_initramfs_trust()
clears initramfs_s_dev by mmap()ing the .bss map and storing 0 at a hardcoded
offset 0. Under the gcc layout that store would clobber the wrong global,
silently leaving the initramfs trust window open after switch_root instead of
closing it. The asserts were correctly catching this.
Fix it by deriving the offset from the generated skeleton instead of a mirror
struct: drop struct restrict_fsaccess_bss and the four field-order
assert_cc()s, take INITRAMFS_S_DEV_OFF from the skeleton's bss struct
(offsetof(typeof_field(struct restrict_fsaccess_bpf, bss[0]),
initramfs_s_dev)), and store at p + INITRAMFS_S_DEV_OFF. The offset is a
compile-time constant, so clang (offset 0) is unchanged while gcc tracks the
real layout. A retained assert_cc() documents the 4-byte alignment the
single-store atomicity relies on.
Fixes: #42689
Yu Watanabe
Michael Vogt
dongshengyuan
Frantisek Sumsal
Lars Sjöstrom
Zbigniew Jędrzejewski-Szmek
Luca Boccassi
Lennart Poettering
Paul Meyer
tunaichao