1 /* SPDX-License-Identifier: LGPL-2.1+ */
3 This file is part of systemd.
5 Copyright 2010 Lennart Poettering
11 #include "sd-messages.h"
13 #include "alloc-util.h"
21 #include "parse-util.h"
24 #include "stdio-util.h"
25 #include "string-table.h"
26 #include "string-util.h"
28 #include "terminal-util.h"
32 Job
* job_new_raw(Unit
*unit
) {
35 /* used for deserialization */
43 j
->manager
= unit
->manager
;
45 j
->type
= _JOB_TYPE_INVALID
;
50 Job
* job_new(Unit
*unit
, JobType type
) {
53 assert(type
< _JOB_TYPE_MAX
);
55 j
= job_new_raw(unit
);
59 j
->id
= j
->manager
->current_job_id
++;
62 /* We don't link it here, that's what job_dependency() is for */
67 void job_free(Job
*j
) {
69 assert(!j
->installed
);
70 assert(!j
->transaction_prev
);
71 assert(!j
->transaction_next
);
72 assert(!j
->subject_list
);
73 assert(!j
->object_list
);
76 LIST_REMOVE(run_queue
, j
->manager
->run_queue
, j
);
79 LIST_REMOVE(dbus_queue
, j
->manager
->dbus_job_queue
, j
);
82 LIST_REMOVE(gc_queue
, j
->manager
->gc_job_queue
, j
);
84 sd_event_source_unref(j
->timer_event_source
);
86 sd_bus_track_unref(j
->bus_track
);
87 strv_free(j
->deserialized_clients
);
92 static void job_set_state(Job
*j
, JobState state
) {
95 assert(state
< _JOB_STATE_MAX
);
97 if (j
->state
== state
)
105 if (j
->state
== JOB_RUNNING
)
106 j
->unit
->manager
->n_running_jobs
++;
108 assert(j
->state
== JOB_WAITING
);
109 assert(j
->unit
->manager
->n_running_jobs
> 0);
111 j
->unit
->manager
->n_running_jobs
--;
113 if (j
->unit
->manager
->n_running_jobs
<= 0)
114 j
->unit
->manager
->jobs_in_progress_event_source
= sd_event_source_unref(j
->unit
->manager
->jobs_in_progress_event_source
);
118 void job_uninstall(Job
*j
) {
121 assert(j
->installed
);
123 job_set_state(j
, JOB_WAITING
);
125 pj
= (j
->type
== JOB_NOP
) ? &j
->unit
->nop_job
: &j
->unit
->job
;
128 /* Detach from next 'bigger' objects */
130 /* daemon-reload should be transparent to job observers */
131 if (!MANAGER_IS_RELOADING(j
->manager
))
132 bus_job_send_removed_signal(j
);
136 unit_add_to_gc_queue(j
->unit
);
138 hashmap_remove(j
->manager
->jobs
, UINT32_TO_PTR(j
->id
));
139 j
->installed
= false;
142 static bool job_type_allows_late_merge(JobType t
) {
143 /* Tells whether it is OK to merge a job of type 't' with an already
145 * Reloads cannot be merged this way. Think of the sequence:
146 * 1. Reload of a daemon is in progress; the daemon has already loaded
147 * its config file, but hasn't completed the reload operation yet.
148 * 2. Edit foo's config file.
149 * 3. Trigger another reload to have the daemon use the new config.
150 * Should the second reload job be merged into the first one, the daemon
151 * would not know about the new config.
152 * JOB_RESTART jobs on the other hand can be merged, because they get
153 * patched into JOB_START after stopping the unit. So if we see a
154 * JOB_RESTART running, it means the unit hasn't stopped yet and at
155 * this time the merge is still allowed. */
156 return t
!= JOB_RELOAD
;
159 static void job_merge_into_installed(Job
*j
, Job
*other
) {
160 assert(j
->installed
);
161 assert(j
->unit
== other
->unit
);
163 if (j
->type
!= JOB_NOP
)
164 job_type_merge_and_collapse(&j
->type
, other
->type
, j
->unit
);
166 assert(other
->type
== JOB_NOP
);
168 j
->irreversible
= j
->irreversible
|| other
->irreversible
;
169 j
->ignore_order
= j
->ignore_order
|| other
->ignore_order
;
172 Job
* job_install(Job
*j
) {
176 assert(!j
->installed
);
177 assert(j
->type
< _JOB_TYPE_MAX_IN_TRANSACTION
);
178 assert(j
->state
== JOB_WAITING
);
180 pj
= (j
->type
== JOB_NOP
) ? &j
->unit
->nop_job
: &j
->unit
->job
;
184 if (job_type_is_conflicting(uj
->type
, j
->type
))
185 job_finish_and_invalidate(uj
, JOB_CANCELED
, false, false);
187 /* not conflicting, i.e. mergeable */
189 if (uj
->state
== JOB_WAITING
||
190 (job_type_allows_late_merge(j
->type
) && job_type_is_superset(uj
->type
, j
->type
))) {
191 job_merge_into_installed(uj
, j
);
192 log_unit_debug(uj
->unit
,
193 "Merged into installed job %s/%s as %u",
194 uj
->unit
->id
, job_type_to_string(uj
->type
), (unsigned) uj
->id
);
197 /* already running and not safe to merge into */
198 /* Patch uj to become a merged job and re-run it. */
199 /* XXX It should be safer to queue j to run after uj finishes, but it is
200 * not currently possible to have more than one installed job per unit. */
201 job_merge_into_installed(uj
, j
);
202 log_unit_debug(uj
->unit
,
203 "Merged into running job, re-running: %s/%s as %u",
204 uj
->unit
->id
, job_type_to_string(uj
->type
), (unsigned) uj
->id
);
206 job_set_state(uj
, JOB_WAITING
);
212 /* Install the job */
216 j
->manager
->n_installed_jobs
++;
217 log_unit_debug(j
->unit
,
218 "Installed new job %s/%s as %u",
219 j
->unit
->id
, job_type_to_string(j
->type
), (unsigned) j
->id
);
221 job_add_to_gc_queue(j
);
226 int job_install_deserialized(Job
*j
) {
229 assert(!j
->installed
);
231 if (j
->type
< 0 || j
->type
>= _JOB_TYPE_MAX_IN_TRANSACTION
) {
232 log_debug("Invalid job type %s in deserialization.", strna(job_type_to_string(j
->type
)));
236 pj
= (j
->type
== JOB_NOP
) ? &j
->unit
->nop_job
: &j
->unit
->job
;
238 log_unit_debug(j
->unit
, "Unit already has a job installed. Not installing deserialized job.");
245 if (j
->state
== JOB_RUNNING
)
246 j
->unit
->manager
->n_running_jobs
++;
248 log_unit_debug(j
->unit
,
249 "Reinstalled deserialized job %s/%s as %u",
250 j
->unit
->id
, job_type_to_string(j
->type
), (unsigned) j
->id
);
254 JobDependency
* job_dependency_new(Job
*subject
, Job
*object
, bool matters
, bool conflicts
) {
259 /* Adds a new job link, which encodes that the 'subject' job
260 * needs the 'object' job in some way. If 'subject' is NULL
261 * this means the 'anchor' job (i.e. the one the user
262 * explicitly asked for) is the requester. */
264 l
= new0(JobDependency
, 1);
268 l
->subject
= subject
;
270 l
->matters
= matters
;
271 l
->conflicts
= conflicts
;
274 LIST_PREPEND(subject
, subject
->subject_list
, l
);
276 LIST_PREPEND(object
, object
->object_list
, l
);
281 void job_dependency_free(JobDependency
*l
) {
285 LIST_REMOVE(subject
, l
->subject
->subject_list
, l
);
287 LIST_REMOVE(object
, l
->object
->object_list
, l
);
292 void job_dump(Job
*j
, FILE*f
, const char *prefix
) {
296 prefix
= strempty(prefix
);
300 "%s\tAction: %s -> %s\n"
302 "%s\tIrreversible: %s\n"
305 prefix
, j
->unit
->id
, job_type_to_string(j
->type
),
306 prefix
, job_state_to_string(j
->state
),
307 prefix
, yes_no(j
->irreversible
),
308 prefix
, yes_no(job_may_gc(j
)));
312 * Merging is commutative, so imagine the matrix as symmetric. We store only
313 * its lower triangle to avoid duplication. We don't store the main diagonal,
314 * because A merged with A is simply A.
316 * If the resulting type is collapsed immediately afterwards (to get rid of
317 * the JOB_RELOAD_OR_START, which lies outside the lookup function's domain),
318 * the following properties hold:
320 * Merging is associative! A merged with B, and then merged with C is the same
321 * as A merged with the result of B merged with C.
323 * Mergeability is transitive! If A can be merged with B and B with C then
326 * Also, if A merged with B cannot be merged with C, then either A or B cannot
327 * be merged with C either.
329 static const JobType job_merging_table
[] = {
330 /* What \ With * JOB_START JOB_VERIFY_ACTIVE JOB_STOP JOB_RELOAD */
331 /*********************************************************************************/
333 /*JOB_VERIFY_ACTIVE */ JOB_START
,
334 /*JOB_STOP */ -1, -1,
335 /*JOB_RELOAD */ JOB_RELOAD_OR_START
, JOB_RELOAD
, -1,
336 /*JOB_RESTART */ JOB_RESTART
, JOB_RESTART
, -1, JOB_RESTART
,
339 JobType
job_type_lookup_merge(JobType a
, JobType b
) {
340 assert_cc(ELEMENTSOF(job_merging_table
) == _JOB_TYPE_MAX_MERGING
* (_JOB_TYPE_MAX_MERGING
- 1) / 2);
341 assert(a
>= 0 && a
< _JOB_TYPE_MAX_MERGING
);
342 assert(b
>= 0 && b
< _JOB_TYPE_MAX_MERGING
);
353 return job_merging_table
[(a
- 1) * a
/ 2 + b
];
356 bool job_type_is_redundant(JobType a
, UnitActiveState b
) {
360 return IN_SET(b
, UNIT_ACTIVE
, UNIT_RELOADING
);
363 return IN_SET(b
, UNIT_INACTIVE
, UNIT_FAILED
);
365 case JOB_VERIFY_ACTIVE
:
366 return IN_SET(b
, UNIT_ACTIVE
, UNIT_RELOADING
);
374 b
== UNIT_ACTIVATING
;
380 assert_not_reached("Invalid job type");
384 JobType
job_type_collapse(JobType t
, Unit
*u
) {
389 case JOB_TRY_RESTART
:
390 s
= unit_active_state(u
);
391 if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s
))
397 s
= unit_active_state(u
);
398 if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s
))
403 case JOB_RELOAD_OR_START
:
404 s
= unit_active_state(u
);
405 if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s
))
415 int job_type_merge_and_collapse(JobType
*a
, JobType b
, Unit
*u
) {
418 t
= job_type_lookup_merge(*a
, b
);
422 *a
= job_type_collapse(t
, u
);
426 static bool job_is_runnable(Job
*j
) {
432 assert(j
->installed
);
434 /* Checks whether there is any job running for the units this
435 * job needs to be running after (in the case of a 'positive'
436 * job type) or before (in the case of a 'negative' job
439 /* Note that unit types have a say in what is runnable,
440 * too. For example, if they return -EAGAIN from
441 * unit_start() they can indicate they are not
444 /* First check if there is an override */
448 if (j
->type
== JOB_NOP
)
451 if (IN_SET(j
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
)) {
452 /* Immediate result is that the job is or might be
453 * started. In this case let's wait for the
454 * dependencies, regardless whether they are
455 * starting or stopping something. */
457 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
)
462 /* Also, if something else is being stopped and we should
463 * change state after it, then let's wait. */
465 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
)
467 IN_SET(other
->job
->type
, JOB_STOP
, JOB_RESTART
))
470 /* This means that for a service a and a service b where b
471 * shall be started after a:
473 * start a + start b → 1st step start a, 2nd step start b
474 * start a + stop b → 1st step stop b, 2nd step start a
475 * stop a + start b → 1st step stop a, 2nd step start b
476 * stop a + stop b → 1st step stop b, 2nd step stop a
478 * This has the side effect that restarts are properly
479 * synchronized too. */
484 static void job_change_type(Job
*j
, JobType newtype
) {
487 log_unit_debug(j
->unit
,
488 "Converting job %s/%s -> %s/%s",
489 j
->unit
->id
, job_type_to_string(j
->type
),
490 j
->unit
->id
, job_type_to_string(newtype
));
495 static int job_perform_on_unit(Job
**j
) {
502 /* While we execute this operation the job might go away (for
503 * example: because it finishes immediately or is replaced by
504 * a new, conflicting job.) To make sure we don't access a
505 * freed job later on we store the id here, so that we can
506 * verify the job is still valid. */
533 assert_not_reached("Invalid job type");
536 /* Log if the job still exists and the start/stop/reload function
537 * actually did something. */
538 *j
= manager_get_job(m
, id
);
540 unit_status_emit_starting_stopping_reloading(u
, t
);
545 int job_run_and_invalidate(Job
*j
) {
549 assert(j
->installed
);
550 assert(j
->type
< _JOB_TYPE_MAX_IN_TRANSACTION
);
551 assert(j
->in_run_queue
);
553 LIST_REMOVE(run_queue
, j
->manager
->run_queue
, j
);
554 j
->in_run_queue
= false;
556 if (j
->state
!= JOB_WAITING
)
559 if (!job_is_runnable(j
))
562 job_start_timer(j
, true);
563 job_set_state(j
, JOB_RUNNING
);
564 job_add_to_dbus_queue(j
);
568 case JOB_VERIFY_ACTIVE
: {
569 UnitActiveState t
= unit_active_state(j
->unit
);
570 if (UNIT_IS_ACTIVE_OR_RELOADING(t
))
572 else if (t
== UNIT_ACTIVATING
)
582 r
= job_perform_on_unit(&j
);
584 /* If the unit type does not support starting/stopping,
585 * then simply wait. */
591 r
= job_perform_on_unit(&j
);
599 assert_not_reached("Unknown job type");
604 r
= job_finish_and_invalidate(j
, JOB_DONE
, true, true);
605 else if (r
== -EBADR
)
606 r
= job_finish_and_invalidate(j
, JOB_SKIPPED
, true, false);
607 else if (r
== -ENOEXEC
)
608 r
= job_finish_and_invalidate(j
, JOB_INVALID
, true, false);
609 else if (r
== -EPROTO
)
610 r
= job_finish_and_invalidate(j
, JOB_ASSERT
, true, false);
611 else if (r
== -EOPNOTSUPP
)
612 r
= job_finish_and_invalidate(j
, JOB_UNSUPPORTED
, true, false);
613 else if (r
== -ENOLINK
)
614 r
= job_finish_and_invalidate(j
, JOB_DEPENDENCY
, true, false);
615 else if (r
== -EAGAIN
)
616 job_set_state(j
, JOB_WAITING
);
618 r
= job_finish_and_invalidate(j
, JOB_FAILED
, true, false);
624 _pure_
static const char *job_get_status_message_format(Unit
*u
, JobType t
, JobResult result
) {
626 static const char *const generic_finished_start_job
[_JOB_RESULT_MAX
] = {
627 [JOB_DONE
] = "Started %s.",
628 [JOB_TIMEOUT
] = "Timed out starting %s.",
629 [JOB_FAILED
] = "Failed to start %s.",
630 [JOB_DEPENDENCY
] = "Dependency failed for %s.",
631 [JOB_ASSERT
] = "Assertion failed for %s.",
632 [JOB_UNSUPPORTED
] = "Starting of %s not supported.",
633 [JOB_COLLECTED
] = "Unnecessary job for %s was removed.",
635 static const char *const generic_finished_stop_job
[_JOB_RESULT_MAX
] = {
636 [JOB_DONE
] = "Stopped %s.",
637 [JOB_FAILED
] = "Stopped (with error) %s.",
638 [JOB_TIMEOUT
] = "Timed out stopping %s.",
640 static const char *const generic_finished_reload_job
[_JOB_RESULT_MAX
] = {
641 [JOB_DONE
] = "Reloaded %s.",
642 [JOB_FAILED
] = "Reload failed for %s.",
643 [JOB_TIMEOUT
] = "Timed out reloading %s.",
645 /* When verify-active detects the unit is inactive, report it.
646 * Most likely a DEPEND warning from a requisiting unit will
647 * occur next and it's nice to see what was requisited. */
648 static const char *const generic_finished_verify_active_job
[_JOB_RESULT_MAX
] = {
649 [JOB_SKIPPED
] = "%s is not active.",
652 const UnitStatusMessageFormats
*format_table
;
657 assert(t
< _JOB_TYPE_MAX
);
659 if (IN_SET(t
, JOB_START
, JOB_STOP
, JOB_RESTART
)) {
660 format_table
= &UNIT_VTABLE(u
)->status_message_formats
;
662 format
= t
== JOB_START
? format_table
->finished_start_job
[result
] :
663 format_table
->finished_stop_job
[result
];
669 /* Return generic strings */
671 return generic_finished_start_job
[result
];
672 else if (IN_SET(t
, JOB_STOP
, JOB_RESTART
))
673 return generic_finished_stop_job
[result
];
674 else if (t
== JOB_RELOAD
)
675 return generic_finished_reload_job
[result
];
676 else if (t
== JOB_VERIFY_ACTIVE
)
677 return generic_finished_verify_active_job
[result
];
682 static const struct {
683 const char *color
, *word
;
684 } job_print_status_messages
[_JOB_RESULT_MAX
] = {
685 [JOB_DONE
] = { ANSI_OK_COLOR
, " OK " },
686 [JOB_TIMEOUT
] = { ANSI_HIGHLIGHT_RED
, " TIME " },
687 [JOB_FAILED
] = { ANSI_HIGHLIGHT_RED
, "FAILED" },
688 [JOB_DEPENDENCY
] = { ANSI_HIGHLIGHT_YELLOW
, "DEPEND" },
689 [JOB_SKIPPED
] = { ANSI_HIGHLIGHT
, " INFO " },
690 [JOB_ASSERT
] = { ANSI_HIGHLIGHT_YELLOW
, "ASSERT" },
691 [JOB_UNSUPPORTED
] = { ANSI_HIGHLIGHT_YELLOW
, "UNSUPP" },
695 static void job_print_status_message(Unit
*u
, JobType t
, JobResult result
) {
701 assert(t
< _JOB_TYPE_MAX
);
703 /* Reload status messages have traditionally not been printed to console. */
707 if (!job_print_status_messages
[result
].word
)
710 format
= job_get_status_message_format(u
, t
, result
);
714 if (log_get_show_color())
715 status
= strjoina(job_print_status_messages
[result
].color
,
716 job_print_status_messages
[result
].word
,
719 status
= job_print_status_messages
[result
].word
;
721 if (result
!= JOB_DONE
)
722 manager_flip_auto_status(u
->manager
, true);
724 DISABLE_WARNING_FORMAT_NONLITERAL
;
725 unit_status_printf(u
, status
, format
);
728 if (t
== JOB_START
&& result
== JOB_FAILED
) {
729 _cleanup_free_
char *quoted
;
731 quoted
= shell_maybe_quote(u
->id
, ESCAPE_BACKSLASH
);
732 manager_status_printf(u
->manager
, STATUS_TYPE_NORMAL
, NULL
, "See 'systemctl status %s' for details.", strna(quoted
));
736 static void job_log_status_message(Unit
*u
, JobType t
, JobResult result
) {
737 const char *format
, *mid
;
739 static const int job_result_log_level
[_JOB_RESULT_MAX
] = {
740 [JOB_DONE
] = LOG_INFO
,
741 [JOB_CANCELED
] = LOG_INFO
,
742 [JOB_TIMEOUT
] = LOG_ERR
,
743 [JOB_FAILED
] = LOG_ERR
,
744 [JOB_DEPENDENCY
] = LOG_WARNING
,
745 [JOB_SKIPPED
] = LOG_NOTICE
,
746 [JOB_INVALID
] = LOG_INFO
,
747 [JOB_ASSERT
] = LOG_WARNING
,
748 [JOB_UNSUPPORTED
] = LOG_WARNING
,
749 [JOB_COLLECTED
] = LOG_INFO
,
754 assert(t
< _JOB_TYPE_MAX
);
756 /* Skip printing if output goes to the console, and job_print_status_message()
757 will actually print something to the console. */
758 if (log_on_console() && job_print_status_messages
[result
].word
)
761 format
= job_get_status_message_format(u
, t
, result
);
765 /* The description might be longer than the buffer, but that's OK,
766 * we'll just truncate it here. Note that we use snprintf() rather than
767 * xsprintf() on purpose here: we are fine with truncation and don't
768 * consider that an error. */
769 DISABLE_WARNING_FORMAT_NONLITERAL
;
770 (void) snprintf(buf
, sizeof(buf
), format
, unit_description(u
));
776 if (result
== JOB_DONE
)
777 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_STARTED_STR
;
779 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_FAILED_STR
;
783 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_RELOADED_STR
;
788 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_STOPPED_STR
;
792 log_struct(job_result_log_level
[result
],
793 LOG_MESSAGE("%s", buf
),
794 "JOB_TYPE=%s", job_type_to_string(t
),
795 "JOB_RESULT=%s", job_result_to_string(result
),
797 LOG_UNIT_INVOCATION_ID(u
),
802 log_struct(job_result_log_level
[result
],
803 LOG_MESSAGE("%s", buf
),
804 "JOB_TYPE=%s", job_type_to_string(t
),
805 "JOB_RESULT=%s", job_result_to_string(result
),
807 LOG_UNIT_INVOCATION_ID(u
),
812 static void job_emit_status_message(Unit
*u
, JobType t
, JobResult result
) {
815 /* No message if the job did not actually do anything due to failed condition. */
816 if (t
== JOB_START
&& result
== JOB_DONE
&& !u
->condition_result
)
819 job_log_status_message(u
, t
, result
);
820 job_print_status_message(u
, t
, result
);
823 static void job_fail_dependencies(Unit
*u
, UnitDependency d
) {
830 HASHMAP_FOREACH_KEY(v
, other
, u
->dependencies
[d
], i
) {
835 if (!IN_SET(j
->type
, JOB_START
, JOB_VERIFY_ACTIVE
))
838 job_finish_and_invalidate(j
, JOB_DEPENDENCY
, true, false);
842 int job_finish_and_invalidate(Job
*j
, JobResult result
, bool recursive
, bool already
) {
850 assert(j
->installed
);
851 assert(j
->type
< _JOB_TYPE_MAX_IN_TRANSACTION
);
858 log_unit_debug(u
, "Job %s/%s finished, result=%s", u
->id
, job_type_to_string(t
), job_result_to_string(result
));
860 /* If this job did nothing to respective unit we don't log the status message */
862 job_emit_status_message(u
, t
, result
);
864 /* Patch restart jobs so that they become normal start jobs */
865 if (result
== JOB_DONE
&& t
== JOB_RESTART
) {
867 job_change_type(j
, JOB_START
);
868 job_set_state(j
, JOB_WAITING
);
870 job_add_to_dbus_queue(j
);
871 job_add_to_run_queue(j
);
872 job_add_to_gc_queue(j
);
877 if (IN_SET(result
, JOB_FAILED
, JOB_INVALID
))
878 j
->manager
->n_failed_jobs
++;
883 /* Fail depending jobs on failure */
884 if (result
!= JOB_DONE
&& recursive
) {
885 if (IN_SET(t
, JOB_START
, JOB_VERIFY_ACTIVE
)) {
886 job_fail_dependencies(u
, UNIT_REQUIRED_BY
);
887 job_fail_dependencies(u
, UNIT_REQUISITE_OF
);
888 job_fail_dependencies(u
, UNIT_BOUND_BY
);
889 } else if (t
== JOB_STOP
)
890 job_fail_dependencies(u
, UNIT_CONFLICTED_BY
);
893 /* Trigger OnFailure dependencies that are not generated by
894 * the unit itself. We don't treat JOB_CANCELED as failure in
895 * this context. And JOB_FAILURE is already handled by the
897 if (IN_SET(result
, JOB_TIMEOUT
, JOB_DEPENDENCY
)) {
898 log_struct(LOG_NOTICE
,
899 "JOB_TYPE=%s", job_type_to_string(t
),
900 "JOB_RESULT=%s", job_result_to_string(result
),
902 LOG_UNIT_MESSAGE(u
, "Job %s/%s failed with result '%s'.",
904 job_type_to_string(t
),
905 job_result_to_string(result
)),
908 unit_start_on_failure(u
);
911 unit_trigger_notify(u
);
914 /* Try to start the next jobs that can be started */
915 HASHMAP_FOREACH_KEY(v
, other
, u
->dependencies
[UNIT_AFTER
], i
)
917 job_add_to_run_queue(other
->job
);
918 job_add_to_gc_queue(other
->job
);
920 HASHMAP_FOREACH_KEY(v
, other
, u
->dependencies
[UNIT_BEFORE
], i
)
922 job_add_to_run_queue(other
->job
);
923 job_add_to_gc_queue(other
->job
);
926 manager_check_finished(u
->manager
);
931 static int job_dispatch_timer(sd_event_source
*s
, uint64_t monotonic
, void *userdata
) {
936 assert(s
== j
->timer_event_source
);
938 log_unit_warning(j
->unit
, "Job %s/%s timed out.", j
->unit
->id
, job_type_to_string(j
->type
));
941 job_finish_and_invalidate(j
, JOB_TIMEOUT
, true, false);
943 emergency_action(u
->manager
, u
->job_timeout_action
, u
->job_timeout_reboot_arg
, "job timed out");
948 int job_start_timer(Job
*j
, bool job_running
) {
950 usec_t timeout_time
, old_timeout_time
;
953 j
->begin_running_usec
= now(CLOCK_MONOTONIC
);
955 if (j
->unit
->job_running_timeout
== USEC_INFINITY
)
958 timeout_time
= usec_add(j
->begin_running_usec
, j
->unit
->job_running_timeout
);
960 if (j
->timer_event_source
) {
961 /* Update only if JobRunningTimeoutSec= results in earlier timeout */
962 r
= sd_event_source_get_time(j
->timer_event_source
, &old_timeout_time
);
966 if (old_timeout_time
<= timeout_time
)
969 return sd_event_source_set_time(j
->timer_event_source
, timeout_time
);
972 if (j
->timer_event_source
)
975 j
->begin_usec
= now(CLOCK_MONOTONIC
);
977 if (j
->unit
->job_timeout
== USEC_INFINITY
)
980 timeout_time
= usec_add(j
->begin_usec
, j
->unit
->job_timeout
);
983 r
= sd_event_add_time(
985 &j
->timer_event_source
,
988 job_dispatch_timer
, j
);
992 (void) sd_event_source_set_description(j
->timer_event_source
, "job-start");
997 void job_add_to_run_queue(Job
*j
) {
999 assert(j
->installed
);
1001 if (j
->in_run_queue
)
1004 if (!j
->manager
->run_queue
)
1005 sd_event_source_set_enabled(j
->manager
->run_queue_event_source
, SD_EVENT_ONESHOT
);
1007 LIST_PREPEND(run_queue
, j
->manager
->run_queue
, j
);
1008 j
->in_run_queue
= true;
1011 void job_add_to_dbus_queue(Job
*j
) {
1013 assert(j
->installed
);
1015 if (j
->in_dbus_queue
)
1018 /* We don't check if anybody is subscribed here, since this
1019 * job might just have been created and not yet assigned to a
1020 * connection/client. */
1022 LIST_PREPEND(dbus_queue
, j
->manager
->dbus_job_queue
, j
);
1023 j
->in_dbus_queue
= true;
1026 char *job_dbus_path(Job
*j
) {
1031 if (asprintf(&p
, "/org/freedesktop/systemd1/job/%"PRIu32
, j
->id
) < 0)
1037 int job_serialize(Job
*j
, FILE *f
) {
1041 fprintf(f
, "job-id=%u\n", j
->id
);
1042 fprintf(f
, "job-type=%s\n", job_type_to_string(j
->type
));
1043 fprintf(f
, "job-state=%s\n", job_state_to_string(j
->state
));
1044 fprintf(f
, "job-irreversible=%s\n", yes_no(j
->irreversible
));
1045 fprintf(f
, "job-sent-dbus-new-signal=%s\n", yes_no(j
->sent_dbus_new_signal
));
1046 fprintf(f
, "job-ignore-order=%s\n", yes_no(j
->ignore_order
));
1048 if (j
->begin_usec
> 0)
1049 fprintf(f
, "job-begin="USEC_FMT
"\n", j
->begin_usec
);
1050 if (j
->begin_running_usec
> 0)
1051 fprintf(f
, "job-begin-running="USEC_FMT
"\n", j
->begin_running_usec
);
1053 bus_track_serialize(j
->bus_track
, f
, "subscribed");
1060 int job_deserialize(Job
*j
, FILE *f
) {
1065 char line
[LINE_MAX
], *l
, *v
;
1068 if (!fgets(line
, sizeof(line
), f
)) {
1081 k
= strcspn(l
, "=");
1089 if (streq(l
, "job-id")) {
1091 if (safe_atou32(v
, &j
->id
) < 0)
1092 log_debug("Failed to parse job id value %s", v
);
1094 } else if (streq(l
, "job-type")) {
1097 t
= job_type_from_string(v
);
1099 log_debug("Failed to parse job type %s", v
);
1100 else if (t
>= _JOB_TYPE_MAX_IN_TRANSACTION
)
1101 log_debug("Cannot deserialize job of type %s", v
);
1105 } else if (streq(l
, "job-state")) {
1108 s
= job_state_from_string(v
);
1110 log_debug("Failed to parse job state %s", v
);
1112 job_set_state(j
, s
);
1114 } else if (streq(l
, "job-irreversible")) {
1117 b
= parse_boolean(v
);
1119 log_debug("Failed to parse job irreversible flag %s", v
);
1121 j
->irreversible
= j
->irreversible
|| b
;
1123 } else if (streq(l
, "job-sent-dbus-new-signal")) {
1126 b
= parse_boolean(v
);
1128 log_debug("Failed to parse job sent_dbus_new_signal flag %s", v
);
1130 j
->sent_dbus_new_signal
= j
->sent_dbus_new_signal
|| b
;
1132 } else if (streq(l
, "job-ignore-order")) {
1135 b
= parse_boolean(v
);
1137 log_debug("Failed to parse job ignore_order flag %s", v
);
1139 j
->ignore_order
= j
->ignore_order
|| b
;
1141 } else if (streq(l
, "job-begin")) {
1142 unsigned long long ull
;
1144 if (sscanf(v
, "%llu", &ull
) != 1)
1145 log_debug("Failed to parse job-begin value %s", v
);
1147 j
->begin_usec
= ull
;
1149 } else if (streq(l
, "job-begin-running")) {
1150 unsigned long long ull
;
1152 if (sscanf(v
, "%llu", &ull
) != 1)
1153 log_debug("Failed to parse job-begin-running value %s", v
);
1155 j
->begin_running_usec
= ull
;
1157 } else if (streq(l
, "subscribed")) {
1159 if (strv_extend(&j
->deserialized_clients
, v
) < 0)
1165 int job_coldplug(Job
*j
) {
1167 usec_t timeout_time
= USEC_INFINITY
;
1171 /* After deserialization is complete and the bus connection
1172 * set up again, let's start watching our subscribers again */
1173 (void) bus_job_coldplug_bus_track(j
);
1175 if (j
->state
== JOB_WAITING
)
1176 job_add_to_run_queue(j
);
1178 /* Maybe due to new dependencies we don't actually need this job anymore? */
1179 job_add_to_gc_queue(j
);
1181 /* Create timer only when job began or began running and the respective timeout is finite.
1182 * Follow logic of job_start_timer() if both timeouts are finite */
1183 if (j
->begin_usec
== 0)
1186 if (j
->unit
->job_timeout
!= USEC_INFINITY
)
1187 timeout_time
= usec_add(j
->begin_usec
, j
->unit
->job_timeout
);
1189 if (j
->begin_running_usec
> 0 && j
->unit
->job_running_timeout
!= USEC_INFINITY
)
1190 timeout_time
= MIN(timeout_time
, usec_add(j
->begin_running_usec
, j
->unit
->job_running_timeout
));
1192 if (timeout_time
== USEC_INFINITY
)
1195 j
->timer_event_source
= sd_event_source_unref(j
->timer_event_source
);
1197 r
= sd_event_add_time(
1199 &j
->timer_event_source
,
1202 job_dispatch_timer
, j
);
1204 log_debug_errno(r
, "Failed to restart timeout for job: %m");
1206 (void) sd_event_source_set_description(j
->timer_event_source
, "job-timeout");
1211 void job_shutdown_magic(Job
*j
) {
1214 /* The shutdown target gets some special treatment here: we
1215 * tell the kernel to begin with flushing its disk caches, to
1216 * optimize shutdown time a bit. Ideally we wouldn't hardcode
1217 * this magic into PID 1. However all other processes aren't
1218 * options either since they'd exit much sooner than PID 1 and
1219 * asynchronous sync() would cause their exit to be
1222 if (j
->type
!= JOB_START
)
1225 if (!MANAGER_IS_SYSTEM(j
->unit
->manager
))
1228 if (!unit_has_name(j
->unit
, SPECIAL_SHUTDOWN_TARGET
))
1231 /* In case messages on console has been disabled on boot */
1232 j
->unit
->manager
->no_console_output
= false;
1234 if (detect_container() > 0)
1237 (void) asynchronous_sync(NULL
);
1240 int job_get_timeout(Job
*j
, usec_t
*timeout
) {
1241 usec_t x
= USEC_INFINITY
, y
= USEC_INFINITY
;
1247 if (j
->timer_event_source
) {
1248 r
= sd_event_source_get_time(j
->timer_event_source
, &x
);
1253 if (UNIT_VTABLE(u
)->get_timeout
) {
1254 r
= UNIT_VTABLE(u
)->get_timeout(u
, &y
);
1259 if (x
== USEC_INFINITY
&& y
== USEC_INFINITY
)
1262 *timeout
= MIN(x
, y
);
1266 bool job_may_gc(Job
*j
) {
1273 /* Checks whether this job should be GC'ed away. We only do this for jobs of units that have no effect on their
1274 * own and just track external state. For now the only unit type that qualifies for this are .device units.
1275 * Returns true if the job can be collected. */
1277 if (!UNIT_VTABLE(j
->unit
)->gc_jobs
)
1280 if (sd_bus_track_count(j
->bus_track
) > 0)
1283 /* FIXME: So this is a bit ugly: for now we don't properly track references made via private bus connections
1284 * (because it's nasty, as sd_bus_track doesn't apply to it). We simply remember that the job was once
1285 * referenced by one, and reset this whenever we notice that no private bus connections are around. This means
1286 * the GC is a bit too conservative when it comes to jobs created by private bus connections. */
1287 if (j
->ref_by_private_bus
) {
1288 if (set_isempty(j
->unit
->manager
->private_buses
))
1289 j
->ref_by_private_bus
= false;
1294 if (j
->type
== JOB_NOP
)
1297 /* If a job is ordered after ours, and is to be started, then it needs to wait for us, regardless if we stop or
1298 * start, hence let's not GC in that case. */
1299 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
) {
1303 if (other
->job
->ignore_order
)
1306 if (IN_SET(other
->job
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
))
1310 /* If we are going down, but something else is ordered After= us, then it needs to wait for us */
1311 if (IN_SET(j
->type
, JOB_STOP
, JOB_RESTART
))
1312 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
) {
1316 if (other
->job
->ignore_order
)
1322 /* The logic above is kinda the inverse of the job_is_runnable() logic. Specifically, if the job "we" is
1323 * ordered before the job "other":
1325 * we start + other start → stay
1326 * we start + other stop → gc
1327 * we stop + other start → stay
1328 * we stop + other stop → gc
1330 * "we" are ordered after "other":
1332 * we start + other start → gc
1333 * we start + other stop → gc
1334 * we stop + other start → stay
1335 * we stop + other stop → stay
1342 void job_add_to_gc_queue(Job
*j
) {
1351 LIST_PREPEND(gc_queue
, j
->unit
->manager
->gc_job_queue
, j
);
1352 j
->in_gc_queue
= true;
1355 static int job_compare(const void *a
, const void *b
) {
1356 Job
*x
= *(Job
**) a
, *y
= *(Job
**) b
;
1366 static size_t sort_job_list(Job
**list
, size_t n
) {
1367 Job
*previous
= NULL
;
1370 /* Order by numeric IDs */
1371 qsort_safe(list
, n
, sizeof(Job
*), job_compare
);
1373 /* Filter out duplicates */
1374 for (a
= 0, b
= 0; a
< n
; a
++) {
1376 if (previous
== list
[a
])
1379 previous
= list
[b
++] = list
[a
];
1385 int job_get_before(Job
*j
, Job
*** ret
) {
1386 _cleanup_free_ Job
** list
= NULL
;
1387 size_t n
= 0, n_allocated
= 0;
1392 /* Returns a list of all pending jobs that need to finish before this job may be started. */
1397 if (j
->ignore_order
) {
1402 if (IN_SET(j
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
)) {
1404 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
) {
1408 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1410 list
[n
++] = other
->job
;
1414 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
) {
1418 if (!IN_SET(other
->job
->type
, JOB_STOP
, JOB_RESTART
))
1421 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1423 list
[n
++] = other
->job
;
1426 n
= sort_job_list(list
, n
);
1428 *ret
= TAKE_PTR(list
);
1433 int job_get_after(Job
*j
, Job
*** ret
) {
1434 _cleanup_free_ Job
** list
= NULL
;
1435 size_t n
= 0, n_allocated
= 0;
1443 /* Returns a list of all pending jobs that are waiting for this job to finish. */
1445 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
) {
1449 if (other
->job
->ignore_order
)
1452 if (!IN_SET(other
->job
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
))
1455 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1457 list
[n
++] = other
->job
;
1460 if (IN_SET(j
->type
, JOB_STOP
, JOB_RESTART
)) {
1462 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
) {
1466 if (other
->job
->ignore_order
)
1469 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1471 list
[n
++] = other
->job
;
1475 n
= sort_job_list(list
, n
);
1477 *ret
= TAKE_PTR(list
);
1482 static const char* const job_state_table
[_JOB_STATE_MAX
] = {
1483 [JOB_WAITING
] = "waiting",
1484 [JOB_RUNNING
] = "running",
1487 DEFINE_STRING_TABLE_LOOKUP(job_state
, JobState
);
1489 static const char* const job_type_table
[_JOB_TYPE_MAX
] = {
1490 [JOB_START
] = "start",
1491 [JOB_VERIFY_ACTIVE
] = "verify-active",
1492 [JOB_STOP
] = "stop",
1493 [JOB_RELOAD
] = "reload",
1494 [JOB_RELOAD_OR_START
] = "reload-or-start",
1495 [JOB_RESTART
] = "restart",
1496 [JOB_TRY_RESTART
] = "try-restart",
1497 [JOB_TRY_RELOAD
] = "try-reload",
1501 DEFINE_STRING_TABLE_LOOKUP(job_type
, JobType
);
1503 static const char* const job_mode_table
[_JOB_MODE_MAX
] = {
1504 [JOB_FAIL
] = "fail",
1505 [JOB_REPLACE
] = "replace",
1506 [JOB_REPLACE_IRREVERSIBLY
] = "replace-irreversibly",
1507 [JOB_ISOLATE
] = "isolate",
1508 [JOB_FLUSH
] = "flush",
1509 [JOB_IGNORE_DEPENDENCIES
] = "ignore-dependencies",
1510 [JOB_IGNORE_REQUIREMENTS
] = "ignore-requirements",
1513 DEFINE_STRING_TABLE_LOOKUP(job_mode
, JobMode
);
1515 static const char* const job_result_table
[_JOB_RESULT_MAX
] = {
1516 [JOB_DONE
] = "done",
1517 [JOB_CANCELED
] = "canceled",
1518 [JOB_TIMEOUT
] = "timeout",
1519 [JOB_FAILED
] = "failed",
1520 [JOB_DEPENDENCY
] = "dependency",
1521 [JOB_SKIPPED
] = "skipped",
1522 [JOB_INVALID
] = "invalid",
1523 [JOB_ASSERT
] = "assert",
1524 [JOB_UNSUPPORTED
] = "unsupported",
1525 [JOB_COLLECTED
] = "collected",
1528 DEFINE_STRING_TABLE_LOOKUP(job_result
, JobResult
);
1530 const char* job_type_to_access_method(JobType t
) {
1532 assert(t
< _JOB_TYPE_MAX
);
1534 if (IN_SET(t
, JOB_START
, JOB_RESTART
, JOB_TRY_RESTART
))
1536 else if (t
== JOB_STOP
)