If chase_symlinks() encouters an absolute symlink, it resets the todo
buffer to just the newly discovered symlink and discards any of the
remaining previous symlink path. Regardless of whether or not the
symlink is absolute or relative, we need to preserve the remainder of
the path that has not yet been resolved.
Instead of using a temp buffer to replace whitespace in variable
substitutions, just allow util_replace_whitespace to replace in-place.
Add a comment to util_replace_whitespace indicating it is used to replace
in-place, to prevent accidental future breakage.
Fixes:
```sh
systemd-run --property EnvironmentFile=/some/environment/file /bin/sleep 30
```
```
23 bytes in 1 blocks are definitely lost in loss record 1 of 7
at 0x4C2DB9D: malloc (vg_replace_malloc.c:299)
by 0x4E85488: malloc_multiply (alloc-util.h:70)
by 0x4E85F19: strjoin_real (string-util.c:252)
by 0x1AF741: bus_exec_context_set_transient_property (dbus-execute.c:1418)
by 0x1A907C: bus_service_set_property (dbus-service.c:330)
by 0x1A66DD: bus_unit_set_properties (dbus-unit.c:1456)
by 0x19CF93: transient_unit_from_message (dbus-manager.c:892)
by 0x19D388: method_start_transient_unit (dbus-manager.c:980)
by 0x4F60544: method_callbacks_run (bus-objects.c:418)
by 0x4F62D9D: object_find_and_run (bus-objects.c:1255)
by 0x4F633CE: bus_process_object (bus-objects.c:1371)
by 0x4F2CE1D: process_message (sd-bus.c:2563)
```
Closes: #5142
When an alias is loaded, we resolve this alias to its final unit first to load
the dropin data.
However if the final unit was already loaded, there's no point in reloading the
dropin data a second time.
This patch optimizes this case.
Also this allows the dropin loading code to assume that only units not yet
loaded are passed down. This assumption is not yet used but might be in the
future.
[zj: invert the condition in the if]
This substantially reworks the seccomp code, to ensure better
compatibility with some architectures, including i386.
So far we relied on libseccomp's internal handling of the multiple
syscall ABIs supported on Linux. This is problematic however, as it does
not define clear semantics if an ABI is not able to support specific
seccomp rules we install.
This rework hence changes a couple of things:
- We no longer use seccomp_rule_add(), but only
seccomp_rule_add_exact(), and fail the installation of a filter if the
architecture doesn't support it.
- We no longer rely on adding multiple syscall architectures to a single filter,
but instead install a separate filter for each syscall architecture
supported. This way, we can install a strict filter for x86-64, while
permitting a less strict filter for i386.
- All high-level filter additions are now moved from execute.c to
seccomp-util.c, so that we can test them independently of the service
execution logic.
- Tests have been added for all types of our seccomp filters.
- SystemCallFilters= and SystemCallArchitectures= are now implemented in
independent filters and installation logic, as they semantically are
very much independent of each other.
Fixes: #4575
When a new entry appears in /proc/self/mountinfo, mount_setup_unit()
allocated a new mount unit for it and starts initializing it.
mount_setup_unit() is also used to update a mount unit when a change happens in
/proc/self/mountinfo, for example a mountpoint can be remounted with additional
mount options.
This patch introduces 2 separate functions to deal with those 2 cases instead
of mount_setup_unit() dealing with both of them. The common code is small and
doing the split makes the code easier to read and less error prone if extended
later.
It also makes sure to initialize in both functions the mount parameters of the
mount unit before calling mount_is_extrinsic() since this function relies on
them.
Fixes: #4902
RDNSS and DNSLL options received in RA are always used, possibly breaking the resolution of private domains hosted on a local DNS server. When setting UseDNS=no in a [IPv6AcceptRA] section of a .network file, both RDNSS and DNSLL options in received RA should be ignored.
Fixes: #5040
A fixed value (6 and later 5) was added back in 4deb3b9391, and
not updated when load_len was added later on.
Also the other 5 with 1 + 1 + 1 + 1 + 1 to make it easier to see
that this is about the column separators.
After parsing the --verify-key argument, overwrite it with null bytes.
This minimizes (but does not completely eliminate) the time frame within
which another process on the system can extract the verification key
from the journalctl command line.
If we are running in a chroot/container/..., we would print a useless warning about
not being able to communicate with systemd. Trying to do daemon-reload is pointless
in those cases, so let's just skip all actions in that case.
The check uses /run/system/system, as recommended by sd_booted(3).
https://bugzilla.redhat.com/show_bug.cgi?id=1411299
When a unit is part of several dependency constraints of another
unit, list-dependencies repeats the name of the dependency for each
dep constraint the unit is encountered.
For example:
$ systemctl cat test-main.target
# /etc/systemd/system/test-main.target
[Unit]
Description=Main Target
$ systemctl cat test.target
[Unit]
Description=Sub target
PartOf=test-main.target
[Install]
WantedBy=test-main.target
$ systemctl enable test.target
Created symlink from /etc/systemd/system/test-main.target.wants/test.target to /etc/systemd/system/test.target.
$ systemctl show test-main.target | grep test.target
Wants=test.target
ConsistsOf=test.target
[...]
$ systemctl list-dependencies test-main.target
test-main.target
● ├─test.target
● └─test.target
With this patch applied, dependencies are shown only once.
