So far we manually hardcoded $LISTEN_FDNAMES to "varlink" in various
varlink service units we ship, even though FileDescriptorName=varlink
is specified in associated socket units already, because
FileDescriptorName= is currently silently ignored when combined with
Accept=yes. Let's step away from this, which seems saner.
Note that this is technically a compat break, but a mostly negligible
one as there shall be few users setting FileDescriptorName= but
still expecting LISTEN_FDNAMES=connection in the actual executable.
Preparation for #34080
We call similar other fields in main.c (notably: rlimit stuff, env vars) "saved",
rather than "original". Hence stick to that kind of naming here too.
Follow-up for: #32937
In apply_one_mount(), in the MOUNT_EXTENSION_DIRECTORY case,
char **extension_release was used as a return pointer twice but only
cleaned up once in the end. Fix it by removing duplicate code that
was causing this issue.
Fixes issue introduced in 55ea4ef096.
Follow-up for 6d2984d21b
The current semantics of "filtered" in unit_is_filtered()
are actually the contrary of ListUnitsFiltered(). Let's
make things consistent, i.e. return true when the unit
shall be included.
The credential mounts should be managed singlehandedly by pid1.
Preparation for the future introduction of RefreshOnReload=credential,
where refreshing creds will be properly supported on reload.
Perform some checks earlier to avoid pointless polkit auth.
Plus, the missing unit_get_exec_context() shall not be
a formalized error. As it's our internal representation
and in the normal operation should never happen.
In the next commit, we'll introduce a varlink server for the user
manager. As preparation for that, let's introduce a new function to
initialize only the managed OOM connection whenever we send a managed
OOM update.
The nice value is part of struct sched_attr, and consequently invoking
sched_setattr() after setpriority() would clobber the nice value with
the default (as we are not setting it in struct sched_attr).
It would be best to combine both calls, but for now simply invoke
setpriority() after sched_setattr() to make sure Nice= remains effective
when used together with CPUSchedulingPolicy=.
When unit_need_daemon_reload() calls unit_find_dropin_paths() to check
for new drop-in configs, the manager's unit path cache is used to limit
which directories are considered. If a new drop-in directory is created,
it may not be in the unit path cache, and hence unit_need_daemon_reload()
may return false, despite a new drop-in being present. However, if a
unit path cache is not given to unit_file_find_dropin_paths() at all,
then it behaves as if the target path was found in the unit path cache.
So, to fix this, adapt unit_find_dropin_paths() to take a boolean
argument indicating whether or not to pass along the unit path cache.
Set this to false in unit_need_daemon_reload().
Fixes#31752
When pid 1 crashes, the getty unit for the console will happily keep
running which means we end up with two shells competing for the same
tty. Let's call vhangup on /dev/console to kill every other process
attached to the console before we spawn the crash shell. The getty
units have Restart=always but lucky for us, pid 1 just crashed in fire
and flames so it isn't actually able to restart the getty unit.
Now that we track auto-restarts with a dedicated state,
there's no need for a separate variable for this.
I also took the chance to reorder some struct members.
unit_start() advertises that start requests don't get suppressed,
so that it could be used to manually speed up auto restarts.
However, service_start() so far rejected this, stating that
clients should issue restart request in order to trigger
BindsTo=/OnFailure=.
That seems to be a red herring though, because for a long time
the service states between auto-restarts were buggy (#27594).
With the introduction of RestartMode=direct, the behavior
is sane again and customizable, hence I see no reason to refuse
this anymore. Whether those deps are triggered solely depends
on RestartMode= now.
Plus, filter out some intermediate states that should never
be seen in service_start().
Fixes#33890
Even if the glob pattern is valid, the pattern may match credentials
with invalid names. So, we need to check the names of the found
credentials.
Follow-up for 947c4d3952.
History (c068650fcf,
941a12dcba) has proven
that we're not good at keeping socket and service states
in sync. Instead, let's query the high-level unit_active_state()
first, and only hardcode the two special auto-restart
service states.
Additionally, allow returning to listening state on SERVICE_CLEANING.
Daan De Meyer
Mike Yuan
Lennart Poettering
Ivan Shapovalov
maia x.
Yu Watanabe
Nick Rosbrook