Add support for configuring cpuset partition type via the
CPUSetPartition= unit file setting. This controls the kernel's
cpuset.cpus.partition cgroup attribute.
The setting takes one of "member", "root", or "isolated". This is
useful for real-time workloads that require dedicated CPU resources
without interference from other processes.
When set, systemd will write the partition type to the
cpuset.cpus.partition cgroup file. If the kernel rejects the value
(e.g., due to partition hierarchy rules), a warning is logged and the
unit continues with the kernel's default partition type.
Co-developed-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Signed-off-by: Gabriele Monaco <gmonaco@redhat.com>
The updated parsing_hacks.h file uncovered a bunch of extra
things that the check-pointer-deref coccinelle script flags.
This commit fixes them to make the tree check-pointer-deref clean.
This allows controlling resources to be refreshed before performing
reload, with one extra benefit that in the future we can permit
"seemless reload"s, i.e. no active signaling is done to the main process
after refreshing get updated. This could come in handy for programs
that loads stuff on demand or watches changes via inotify.
Unlike other settings that takes multiple values, the setting takes
colon separated list of paths, but when specified as a DBus property,
it previously accepted space separated list of paths.
Let's also make the DBus property accepts colon separated lists.
Fixes#40513.
Transparent Hugepages (THP) is a Linux kernel feature that manages
memory using larger pages (2MB on x86, compared to the default 4KB).
The main goal is to improve memory management efficiency and system
performance, especially for memory-intensive applications.
However, it can cause drawbacks in some scenarios, such as memory
regression and latency spikes. THP policy is governed for the entire
system via /sys/kernel/mm/transparent_hugepage/enabled.
However, it can be overridden for individual workloads via prctl(2)
call.
MemoryTHP= is used to disable THPs at exec-invoke to stop
providing THPs for workloads where the drawbacks outweigh the advantages.
When set to "disable", MemoryTHP= disables THPs completely for the
process, irrespecitive of global THP controls.
When set to "madvise", MemoryTHP= disables THPs for the process except
when specifically madvised by the process with MADV_HUGEPAGE or MADV_COLLAPSE.
Latest glibc uses _Generic to have strstr() and other functions return
const char* or char* based on whether the input is a const char* or a
char*. This causes build failures as we previously always expected a char*.
Let's fix the compilation failures and add our own macros similar to glibc's
to have string functions that return a mutable or const pointer depending on
the input.
When Type=notify-reload got introduced, it wasn't intended to be
mutually exclusive with ExecReload=. However, currently ExecReload=
is immediately forked off after the service main process is signaled,
leaving states in between essentially undefined. Given so broken
it is I doubt any sane user is using this setup, hence I took a stab
to rework everything:
1. Extensions are refreshed (unchanged)
2. ExecReload= is forked off without signaling the process
3a. If RELOADING=1 is sent during the ExecReload= invocation,
we'd refrain from signaling the process again, instead
just transition to SERVICE_RELOAD_NOTIFY directly and
wait for READY=1
3b. If not, signal the process after ExecReload= finishes
(from now on the same as Type=notify-reload w/o ExecReload=)
4. To accomodate the use case of performing post-reload tasks,
ExecReloadPost= is introduced which executes after READY=1
The new model greatly simplifies things, as no control processes
will be around in SERVICE_RELOAD_SIGNAL and SERVICE_RELOAD_NOTIFY
states.
See also: https://github.com/systemd/systemd/issues/37515#issuecomment-2891229652
This allows a service to reuse the user namespace created for an
existing service, similarly to NetworkNamespacePath=. The configuration
is the initial user namespace (e.g. ID mapping) is preserved.
Add a new option PrivateBPF= to mount a new instance of bpffs within a
namespace.
PrivateBPF= can be set to "no" to use the host bpffs in readonly mode
and "yes" to do a new mount.
The mount is done with the new fsopen()/fsmount() API because in future
we'll hook some commands between the two calls.
For some fields, we perform careful parsing and verification on the sender
side. For other fields, we accept any string or strv. I think that actually
this is fine: we should optimize for the correct case, i.e. the user runs a
command that is valid. The server must perform parsing in all cases, so doing
the verification on the sender side doesn't add value. When doing parsing
locally, in case of invalid or unsupported input, we would generate the error
message locally, so we would avoid the D-Bus call, but the message itself is
not better and from the user's point of view, the result is the same. And by
doing the parsing only on the server side, we deal better with the case where
the sender has an older version of the software. By not doing verification, we
implicitly "support" new values. And when the sender has a newer version that
supports additional fields, that does not help as long as the server uses an
older version. So in case of version mismatches, parsing on the server side is
as good or better.
Resolves https://github.com/systemd/systemd/issues/37174.
This changes little in behaviour, the conceptual part is more important. The
non-Ex variant is the actual name on the command line, and we should use the
non-Ex D-Bus property too, if it works. This increases compatibility with old
versions. But the code was mostly doing the right thing. Even the tests tested
the right thing.
Follow-up for b3d593673c and
898fc00e79.
The test is simplified by taking advantage of the fact that both names
on the commandline are supposed to behave identically.
Partially resolves https://github.com/systemd/systemd/issues/37174.
We generally want to have error messages with a fixed structure that convey the
important information, i.e. field name, error value, and the offending text for
options that take short values. (The text is not printed for strings encoded with
base64 and hexmem or for credentials.)
Let's use a helper that prints the message in a fixed format in the majority of
cases. In the few places where a custom message is useful, the helper is not
used. The helper:
- prints the field name, value, and error info,
- quotes the value,
- handles -ENOMEM, so we don't need to handle it separately everywhere.
When this code was originally written, parse functions would return -1
as error. Nowadays day all return a good errno, so it is fine if we print
the corresponding strerror.
exec_command_flags_to_strv() should not fail, unless we screwed up, so assert
instead of returning an error. Also, no need to strdup constant _PATH_BSHELL;
drop that so that we can get rid of the oom error handling. Finally, rename
l → cmdline for clarity.
Before, for empty input, we'd send an array with one item with an empty
pattern. Use the helper which sends an empty array instead.
bus_exec_context_set_transient_property() ignores items with an empty
pattern, so the result should be the same.
Request in review:
https://github.com/systemd/systemd/pull/37665#discussion_r2182375988.
Those are user-controlled strings, so let's use heap allocations in the usual
fashion. (Though, with strndupa_safe, the allocations were bounded anyway, so
ultimately this doesn't matter.)
bus_append_unit_property() and associated functions accept a long list of
properties. But the specific names are only available through code. But it is
useful to be able to know the specific list of properties that is supported, in
particular for shell completions. Thus, add a way to list the properties that
are supported by the code.
In the future we could also turn this into a test for the documentation. For
various reasons, the list of properties listed in the docs is a partially
overlapping set. E.g. for service type, the pull request
https://github.com/systemd/systemd/pull/37661 creates a list with 212 entries,
and this code generates 7 entries less and 184 more. I didn't check all the
differences, but in the few cases I did, the list generated here was actually
correctly supported by 'systemd-run -p'.
A smoke test is added.
As with grandparent commit for ImportCredentialEx=, the whole series of commits
that extended ProtectHostname was confused (6746f28854,
cf48bde7ae, e76fcd0e40),
because it added ProtectHostnameEx in places where parsing of ProtectHostname
should be have been extended.
Accept ProtectHostname=… with the new extended syntax, keep accepting
ProtectHostnameEx=… for compat with release v257. Prefer sending ProtectHostname.
Partially resolves https://github.com/systemd/systemd/issues/37174.
