1 /* SPDX-License-Identifier: LGPL-2.1+ */
6 #include "sd-messages.h"
8 #include "alloc-util.h"
16 #include "parse-util.h"
19 #include "stdio-util.h"
20 #include "string-table.h"
21 #include "string-util.h"
23 #include "terminal-util.h"
27 Job
* job_new_raw(Unit
*unit
) {
30 /* used for deserialization */
39 .manager
= unit
->manager
,
41 .type
= _JOB_TYPE_INVALID
,
47 Job
* job_new(Unit
*unit
, JobType type
) {
50 assert(type
< _JOB_TYPE_MAX
);
52 j
= job_new_raw(unit
);
56 j
->id
= j
->manager
->current_job_id
++;
59 /* We don't link it here, that's what job_dependency() is for */
64 void job_unlink(Job
*j
) {
66 assert(!j
->installed
);
67 assert(!j
->transaction_prev
);
68 assert(!j
->transaction_next
);
69 assert(!j
->subject_list
);
70 assert(!j
->object_list
);
72 if (j
->in_run_queue
) {
73 LIST_REMOVE(run_queue
, j
->manager
->run_queue
, j
);
74 j
->in_run_queue
= false;
77 if (j
->in_dbus_queue
) {
78 LIST_REMOVE(dbus_queue
, j
->manager
->dbus_job_queue
, j
);
79 j
->in_dbus_queue
= false;
83 LIST_REMOVE(gc_queue
, j
->manager
->gc_job_queue
, j
);
84 j
->in_gc_queue
= false;
87 j
->timer_event_source
= sd_event_source_unref(j
->timer_event_source
);
90 void job_free(Job
*j
) {
92 assert(!j
->installed
);
93 assert(!j
->transaction_prev
);
94 assert(!j
->transaction_next
);
95 assert(!j
->subject_list
);
96 assert(!j
->object_list
);
100 sd_bus_track_unref(j
->bus_track
);
101 strv_free(j
->deserialized_clients
);
106 static void job_set_state(Job
*j
, JobState state
) {
109 assert(state
< _JOB_STATE_MAX
);
111 if (j
->state
== state
)
119 if (j
->state
== JOB_RUNNING
)
120 j
->unit
->manager
->n_running_jobs
++;
122 assert(j
->state
== JOB_WAITING
);
123 assert(j
->unit
->manager
->n_running_jobs
> 0);
125 j
->unit
->manager
->n_running_jobs
--;
127 if (j
->unit
->manager
->n_running_jobs
<= 0)
128 j
->unit
->manager
->jobs_in_progress_event_source
= sd_event_source_unref(j
->unit
->manager
->jobs_in_progress_event_source
);
132 void job_uninstall(Job
*j
) {
135 assert(j
->installed
);
137 job_set_state(j
, JOB_WAITING
);
139 pj
= (j
->type
== JOB_NOP
) ? &j
->unit
->nop_job
: &j
->unit
->job
;
142 /* Detach from next 'bigger' objects */
144 /* daemon-reload should be transparent to job observers */
145 if (!MANAGER_IS_RELOADING(j
->manager
))
146 bus_job_send_removed_signal(j
);
150 unit_add_to_gc_queue(j
->unit
);
152 hashmap_remove(j
->manager
->jobs
, UINT32_TO_PTR(j
->id
));
153 j
->installed
= false;
156 static bool job_type_allows_late_merge(JobType t
) {
157 /* Tells whether it is OK to merge a job of type 't' with an already
159 * Reloads cannot be merged this way. Think of the sequence:
160 * 1. Reload of a daemon is in progress; the daemon has already loaded
161 * its config file, but hasn't completed the reload operation yet.
162 * 2. Edit foo's config file.
163 * 3. Trigger another reload to have the daemon use the new config.
164 * Should the second reload job be merged into the first one, the daemon
165 * would not know about the new config.
166 * JOB_RESTART jobs on the other hand can be merged, because they get
167 * patched into JOB_START after stopping the unit. So if we see a
168 * JOB_RESTART running, it means the unit hasn't stopped yet and at
169 * this time the merge is still allowed. */
170 return t
!= JOB_RELOAD
;
173 static void job_merge_into_installed(Job
*j
, Job
*other
) {
174 assert(j
->installed
);
175 assert(j
->unit
== other
->unit
);
177 if (j
->type
!= JOB_NOP
)
178 assert_se(job_type_merge_and_collapse(&j
->type
, other
->type
, j
->unit
) == 0);
180 assert(other
->type
== JOB_NOP
);
182 j
->irreversible
= j
->irreversible
|| other
->irreversible
;
183 j
->ignore_order
= j
->ignore_order
|| other
->ignore_order
;
186 Job
* job_install(Job
*j
) {
190 assert(!j
->installed
);
191 assert(j
->type
< _JOB_TYPE_MAX_IN_TRANSACTION
);
192 assert(j
->state
== JOB_WAITING
);
194 pj
= (j
->type
== JOB_NOP
) ? &j
->unit
->nop_job
: &j
->unit
->job
;
198 if (job_type_is_conflicting(uj
->type
, j
->type
))
199 job_finish_and_invalidate(uj
, JOB_CANCELED
, false, false);
201 /* not conflicting, i.e. mergeable */
203 if (uj
->state
== JOB_WAITING
||
204 (job_type_allows_late_merge(j
->type
) && job_type_is_superset(uj
->type
, j
->type
))) {
205 job_merge_into_installed(uj
, j
);
206 log_unit_debug(uj
->unit
,
207 "Merged into installed job %s/%s as %u",
208 uj
->unit
->id
, job_type_to_string(uj
->type
), (unsigned) uj
->id
);
211 /* already running and not safe to merge into */
212 /* Patch uj to become a merged job and re-run it. */
213 /* XXX It should be safer to queue j to run after uj finishes, but it is
214 * not currently possible to have more than one installed job per unit. */
215 job_merge_into_installed(uj
, j
);
216 log_unit_debug(uj
->unit
,
217 "Merged into running job, re-running: %s/%s as %u",
218 uj
->unit
->id
, job_type_to_string(uj
->type
), (unsigned) uj
->id
);
220 job_set_state(uj
, JOB_WAITING
);
226 /* Install the job */
230 j
->manager
->n_installed_jobs
++;
231 log_unit_debug(j
->unit
,
232 "Installed new job %s/%s as %u",
233 j
->unit
->id
, job_type_to_string(j
->type
), (unsigned) j
->id
);
235 job_add_to_gc_queue(j
);
240 int job_install_deserialized(Job
*j
) {
243 assert(!j
->installed
);
245 if (j
->type
< 0 || j
->type
>= _JOB_TYPE_MAX_IN_TRANSACTION
) {
246 log_debug("Invalid job type %s in deserialization.", strna(job_type_to_string(j
->type
)));
250 pj
= (j
->type
== JOB_NOP
) ? &j
->unit
->nop_job
: &j
->unit
->job
;
252 log_unit_debug(j
->unit
, "Unit already has a job installed. Not installing deserialized job.");
260 if (j
->state
== JOB_RUNNING
)
261 j
->unit
->manager
->n_running_jobs
++;
263 log_unit_debug(j
->unit
,
264 "Reinstalled deserialized job %s/%s as %u",
265 j
->unit
->id
, job_type_to_string(j
->type
), (unsigned) j
->id
);
269 JobDependency
* job_dependency_new(Job
*subject
, Job
*object
, bool matters
, bool conflicts
) {
274 /* Adds a new job link, which encodes that the 'subject' job
275 * needs the 'object' job in some way. If 'subject' is NULL
276 * this means the 'anchor' job (i.e. the one the user
277 * explicitly asked for) is the requester. */
279 l
= new0(JobDependency
, 1);
283 l
->subject
= subject
;
285 l
->matters
= matters
;
286 l
->conflicts
= conflicts
;
289 LIST_PREPEND(subject
, subject
->subject_list
, l
);
291 LIST_PREPEND(object
, object
->object_list
, l
);
296 void job_dependency_free(JobDependency
*l
) {
300 LIST_REMOVE(subject
, l
->subject
->subject_list
, l
);
302 LIST_REMOVE(object
, l
->object
->object_list
, l
);
307 void job_dump(Job
*j
, FILE*f
, const char *prefix
) {
311 prefix
= strempty(prefix
);
315 "%s\tAction: %s -> %s\n"
317 "%s\tIrreversible: %s\n"
320 prefix
, j
->unit
->id
, job_type_to_string(j
->type
),
321 prefix
, job_state_to_string(j
->state
),
322 prefix
, yes_no(j
->irreversible
),
323 prefix
, yes_no(job_may_gc(j
)));
327 * Merging is commutative, so imagine the matrix as symmetric. We store only
328 * its lower triangle to avoid duplication. We don't store the main diagonal,
329 * because A merged with A is simply A.
331 * If the resulting type is collapsed immediately afterwards (to get rid of
332 * the JOB_RELOAD_OR_START, which lies outside the lookup function's domain),
333 * the following properties hold:
335 * Merging is associative! A merged with B, and then merged with C is the same
336 * as A merged with the result of B merged with C.
338 * Mergeability is transitive! If A can be merged with B and B with C then
341 * Also, if A merged with B cannot be merged with C, then either A or B cannot
342 * be merged with C either.
344 static const JobType job_merging_table
[] = {
345 /* What \ With * JOB_START JOB_VERIFY_ACTIVE JOB_STOP JOB_RELOAD */
346 /*********************************************************************************/
348 /*JOB_VERIFY_ACTIVE */ JOB_START
,
349 /*JOB_STOP */ -1, -1,
350 /*JOB_RELOAD */ JOB_RELOAD_OR_START
, JOB_RELOAD
, -1,
351 /*JOB_RESTART */ JOB_RESTART
, JOB_RESTART
, -1, JOB_RESTART
,
354 JobType
job_type_lookup_merge(JobType a
, JobType b
) {
355 assert_cc(ELEMENTSOF(job_merging_table
) == _JOB_TYPE_MAX_MERGING
* (_JOB_TYPE_MAX_MERGING
- 1) / 2);
356 assert(a
>= 0 && a
< _JOB_TYPE_MAX_MERGING
);
357 assert(b
>= 0 && b
< _JOB_TYPE_MAX_MERGING
);
368 return job_merging_table
[(a
- 1) * a
/ 2 + b
];
371 bool job_type_is_redundant(JobType a
, UnitActiveState b
) {
375 return IN_SET(b
, UNIT_ACTIVE
, UNIT_RELOADING
);
378 return IN_SET(b
, UNIT_INACTIVE
, UNIT_FAILED
);
380 case JOB_VERIFY_ACTIVE
:
381 return IN_SET(b
, UNIT_ACTIVE
, UNIT_RELOADING
);
389 b
== UNIT_ACTIVATING
;
395 assert_not_reached("Invalid job type");
399 JobType
job_type_collapse(JobType t
, Unit
*u
) {
404 case JOB_TRY_RESTART
:
405 s
= unit_active_state(u
);
406 if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s
))
412 s
= unit_active_state(u
);
413 if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s
))
418 case JOB_RELOAD_OR_START
:
419 s
= unit_active_state(u
);
420 if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s
))
430 int job_type_merge_and_collapse(JobType
*a
, JobType b
, Unit
*u
) {
433 t
= job_type_lookup_merge(*a
, b
);
437 *a
= job_type_collapse(t
, u
);
441 static bool job_is_runnable(Job
*j
) {
447 assert(j
->installed
);
449 /* Checks whether there is any job running for the units this
450 * job needs to be running after (in the case of a 'positive'
451 * job type) or before (in the case of a 'negative' job
454 /* Note that unit types have a say in what is runnable,
455 * too. For example, if they return -EAGAIN from
456 * unit_start() they can indicate they are not
459 /* First check if there is an override */
463 if (j
->type
== JOB_NOP
)
466 if (IN_SET(j
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
)) {
467 /* Immediate result is that the job is or might be
468 * started. In this case let's wait for the
469 * dependencies, regardless whether they are
470 * starting or stopping something. */
472 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
)
477 /* Also, if something else is being stopped and we should
478 * change state after it, then let's wait. */
480 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
)
482 IN_SET(other
->job
->type
, JOB_STOP
, JOB_RESTART
))
485 /* This means that for a service a and a service b where b
486 * shall be started after a:
488 * start a + start b → 1st step start a, 2nd step start b
489 * start a + stop b → 1st step stop b, 2nd step start a
490 * stop a + start b → 1st step stop a, 2nd step start b
491 * stop a + stop b → 1st step stop b, 2nd step stop a
493 * This has the side effect that restarts are properly
494 * synchronized too. */
499 static void job_change_type(Job
*j
, JobType newtype
) {
502 log_unit_debug(j
->unit
,
503 "Converting job %s/%s -> %s/%s",
504 j
->unit
->id
, job_type_to_string(j
->type
),
505 j
->unit
->id
, job_type_to_string(newtype
));
510 static int job_perform_on_unit(Job
**j
) {
517 /* While we execute this operation the job might go away (for
518 * example: because it finishes immediately or is replaced by
519 * a new, conflicting job.) To make sure we don't access a
520 * freed job later on we store the id here, so that we can
521 * verify the job is still valid. */
548 assert_not_reached("Invalid job type");
551 /* Log if the job still exists and the start/stop/reload function
552 * actually did something. */
553 *j
= manager_get_job(m
, id
);
555 unit_status_emit_starting_stopping_reloading(u
, t
);
560 int job_run_and_invalidate(Job
*j
) {
564 assert(j
->installed
);
565 assert(j
->type
< _JOB_TYPE_MAX_IN_TRANSACTION
);
566 assert(j
->in_run_queue
);
568 LIST_REMOVE(run_queue
, j
->manager
->run_queue
, j
);
569 j
->in_run_queue
= false;
571 if (j
->state
!= JOB_WAITING
)
574 if (!job_is_runnable(j
))
577 job_start_timer(j
, true);
578 job_set_state(j
, JOB_RUNNING
);
579 job_add_to_dbus_queue(j
);
583 case JOB_VERIFY_ACTIVE
: {
584 UnitActiveState t
= unit_active_state(j
->unit
);
585 if (UNIT_IS_ACTIVE_OR_RELOADING(t
))
587 else if (t
== UNIT_ACTIVATING
)
597 r
= job_perform_on_unit(&j
);
599 /* If the unit type does not support starting/stopping,
600 * then simply wait. */
606 r
= job_perform_on_unit(&j
);
614 assert_not_reached("Unknown job type");
619 r
= job_finish_and_invalidate(j
, JOB_DONE
, true, true);
620 else if (r
== -EBADR
)
621 r
= job_finish_and_invalidate(j
, JOB_SKIPPED
, true, false);
622 else if (r
== -ENOEXEC
)
623 r
= job_finish_and_invalidate(j
, JOB_INVALID
, true, false);
624 else if (r
== -EPROTO
)
625 r
= job_finish_and_invalidate(j
, JOB_ASSERT
, true, false);
626 else if (r
== -EOPNOTSUPP
)
627 r
= job_finish_and_invalidate(j
, JOB_UNSUPPORTED
, true, false);
628 else if (r
== -ENOLINK
)
629 r
= job_finish_and_invalidate(j
, JOB_DEPENDENCY
, true, false);
630 else if (r
== -ESTALE
)
631 r
= job_finish_and_invalidate(j
, JOB_ONCE
, true, false);
632 else if (r
== -EAGAIN
)
633 job_set_state(j
, JOB_WAITING
);
635 r
= job_finish_and_invalidate(j
, JOB_FAILED
, true, false);
641 _pure_
static const char *job_get_status_message_format(Unit
*u
, JobType t
, JobResult result
) {
643 static const char *const generic_finished_start_job
[_JOB_RESULT_MAX
] = {
644 [JOB_DONE
] = "Started %s.",
645 [JOB_TIMEOUT
] = "Timed out starting %s.",
646 [JOB_FAILED
] = "Failed to start %s.",
647 [JOB_DEPENDENCY
] = "Dependency failed for %s.",
648 [JOB_ASSERT
] = "Assertion failed for %s.",
649 [JOB_UNSUPPORTED
] = "Starting of %s not supported.",
650 [JOB_COLLECTED
] = "Unnecessary job for %s was removed.",
651 [JOB_ONCE
] = "Unit %s has been started before and cannot be started again."
653 static const char *const generic_finished_stop_job
[_JOB_RESULT_MAX
] = {
654 [JOB_DONE
] = "Stopped %s.",
655 [JOB_FAILED
] = "Stopped (with error) %s.",
656 [JOB_TIMEOUT
] = "Timed out stopping %s.",
658 static const char *const generic_finished_reload_job
[_JOB_RESULT_MAX
] = {
659 [JOB_DONE
] = "Reloaded %s.",
660 [JOB_FAILED
] = "Reload failed for %s.",
661 [JOB_TIMEOUT
] = "Timed out reloading %s.",
663 /* When verify-active detects the unit is inactive, report it.
664 * Most likely a DEPEND warning from a requisiting unit will
665 * occur next and it's nice to see what was requisited. */
666 static const char *const generic_finished_verify_active_job
[_JOB_RESULT_MAX
] = {
667 [JOB_SKIPPED
] = "%s is not active.",
670 const UnitStatusMessageFormats
*format_table
;
675 assert(t
< _JOB_TYPE_MAX
);
677 if (IN_SET(t
, JOB_START
, JOB_STOP
, JOB_RESTART
)) {
678 format_table
= &UNIT_VTABLE(u
)->status_message_formats
;
680 format
= t
== JOB_START
? format_table
->finished_start_job
[result
] :
681 format_table
->finished_stop_job
[result
];
687 /* Return generic strings */
689 return generic_finished_start_job
[result
];
690 else if (IN_SET(t
, JOB_STOP
, JOB_RESTART
))
691 return generic_finished_stop_job
[result
];
692 else if (t
== JOB_RELOAD
)
693 return generic_finished_reload_job
[result
];
694 else if (t
== JOB_VERIFY_ACTIVE
)
695 return generic_finished_verify_active_job
[result
];
700 static const struct {
701 const char *color
, *word
;
702 } job_print_status_messages
[_JOB_RESULT_MAX
] = {
703 [JOB_DONE
] = { ANSI_OK_COLOR
, " OK " },
704 [JOB_TIMEOUT
] = { ANSI_HIGHLIGHT_RED
, " TIME " },
705 [JOB_FAILED
] = { ANSI_HIGHLIGHT_RED
, "FAILED" },
706 [JOB_DEPENDENCY
] = { ANSI_HIGHLIGHT_YELLOW
, "DEPEND" },
707 [JOB_SKIPPED
] = { ANSI_HIGHLIGHT
, " INFO " },
708 [JOB_ASSERT
] = { ANSI_HIGHLIGHT_YELLOW
, "ASSERT" },
709 [JOB_UNSUPPORTED
] = { ANSI_HIGHLIGHT_YELLOW
, "UNSUPP" },
711 [JOB_ONCE
] = { ANSI_HIGHLIGHT_RED
, " ONCE " },
714 static void job_print_status_message(Unit
*u
, JobType t
, JobResult result
) {
720 assert(t
< _JOB_TYPE_MAX
);
722 /* Reload status messages have traditionally not been printed to console. */
726 if (!job_print_status_messages
[result
].word
)
729 format
= job_get_status_message_format(u
, t
, result
);
733 if (log_get_show_color())
734 status
= strjoina(job_print_status_messages
[result
].color
,
735 job_print_status_messages
[result
].word
,
738 status
= job_print_status_messages
[result
].word
;
740 if (result
!= JOB_DONE
)
741 manager_flip_auto_status(u
->manager
, true);
743 DISABLE_WARNING_FORMAT_NONLITERAL
;
744 unit_status_printf(u
, status
, format
);
747 if (t
== JOB_START
&& result
== JOB_FAILED
) {
748 _cleanup_free_
char *quoted
;
750 quoted
= shell_maybe_quote(u
->id
, ESCAPE_BACKSLASH
);
751 manager_status_printf(u
->manager
, STATUS_TYPE_NORMAL
, NULL
, "See 'systemctl status %s' for details.", strna(quoted
));
755 static void job_log_status_message(Unit
*u
, JobType t
, JobResult result
) {
756 const char *format
, *mid
;
758 static const int job_result_log_level
[_JOB_RESULT_MAX
] = {
759 [JOB_DONE
] = LOG_INFO
,
760 [JOB_CANCELED
] = LOG_INFO
,
761 [JOB_TIMEOUT
] = LOG_ERR
,
762 [JOB_FAILED
] = LOG_ERR
,
763 [JOB_DEPENDENCY
] = LOG_WARNING
,
764 [JOB_SKIPPED
] = LOG_NOTICE
,
765 [JOB_INVALID
] = LOG_INFO
,
766 [JOB_ASSERT
] = LOG_WARNING
,
767 [JOB_UNSUPPORTED
] = LOG_WARNING
,
768 [JOB_COLLECTED
] = LOG_INFO
,
769 [JOB_ONCE
] = LOG_ERR
,
774 assert(t
< _JOB_TYPE_MAX
);
776 /* Skip printing if output goes to the console, and job_print_status_message()
777 will actually print something to the console. */
778 if (log_on_console() && job_print_status_messages
[result
].word
)
781 format
= job_get_status_message_format(u
, t
, result
);
785 /* The description might be longer than the buffer, but that's OK,
786 * we'll just truncate it here. Note that we use snprintf() rather than
787 * xsprintf() on purpose here: we are fine with truncation and don't
788 * consider that an error. */
789 DISABLE_WARNING_FORMAT_NONLITERAL
;
790 (void) snprintf(buf
, sizeof(buf
), format
, unit_description(u
));
796 if (result
== JOB_DONE
)
797 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_STARTED_STR
;
799 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_FAILED_STR
;
803 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_RELOADED_STR
;
808 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_STOPPED_STR
;
812 log_struct(job_result_log_level
[result
],
813 LOG_MESSAGE("%s", buf
),
814 "JOB_TYPE=%s", job_type_to_string(t
),
815 "JOB_RESULT=%s", job_result_to_string(result
),
817 LOG_UNIT_INVOCATION_ID(u
));
821 log_struct(job_result_log_level
[result
],
822 LOG_MESSAGE("%s", buf
),
823 "JOB_TYPE=%s", job_type_to_string(t
),
824 "JOB_RESULT=%s", job_result_to_string(result
),
826 LOG_UNIT_INVOCATION_ID(u
),
830 static void job_emit_status_message(Unit
*u
, JobType t
, JobResult result
) {
833 /* No message if the job did not actually do anything due to failed condition. */
834 if (t
== JOB_START
&& result
== JOB_DONE
&& !u
->condition_result
)
837 job_log_status_message(u
, t
, result
);
838 job_print_status_message(u
, t
, result
);
841 static void job_fail_dependencies(Unit
*u
, UnitDependency d
) {
848 HASHMAP_FOREACH_KEY(v
, other
, u
->dependencies
[d
], i
) {
853 if (!IN_SET(j
->type
, JOB_START
, JOB_VERIFY_ACTIVE
))
856 job_finish_and_invalidate(j
, JOB_DEPENDENCY
, true, false);
860 static int job_save_pending_finished_job(Job
*j
) {
865 r
= set_ensure_allocated(&j
->manager
->pending_finished_jobs
, NULL
);
870 return set_put(j
->manager
->pending_finished_jobs
, j
);
873 int job_finish_and_invalidate(Job
*j
, JobResult result
, bool recursive
, bool already
) {
881 assert(j
->installed
);
882 assert(j
->type
< _JOB_TYPE_MAX_IN_TRANSACTION
);
889 log_unit_debug(u
, "Job %s/%s finished, result=%s", u
->id
, job_type_to_string(t
), job_result_to_string(result
));
891 /* If this job did nothing to respective unit we don't log the status message */
893 job_emit_status_message(u
, t
, result
);
895 /* Patch restart jobs so that they become normal start jobs */
896 if (result
== JOB_DONE
&& t
== JOB_RESTART
) {
898 job_change_type(j
, JOB_START
);
899 job_set_state(j
, JOB_WAITING
);
901 job_add_to_dbus_queue(j
);
902 job_add_to_run_queue(j
);
903 job_add_to_gc_queue(j
);
908 if (IN_SET(result
, JOB_FAILED
, JOB_INVALID
))
909 j
->manager
->n_failed_jobs
++;
912 /* Keep jobs started before the reload to send singal later, free all others */
913 if (!MANAGER_IS_RELOADING(j
->manager
) ||
915 job_save_pending_finished_job(j
) < 0)
918 /* Fail depending jobs on failure */
919 if (result
!= JOB_DONE
&& recursive
) {
920 if (IN_SET(t
, JOB_START
, JOB_VERIFY_ACTIVE
)) {
921 job_fail_dependencies(u
, UNIT_REQUIRED_BY
);
922 job_fail_dependencies(u
, UNIT_REQUISITE_OF
);
923 job_fail_dependencies(u
, UNIT_BOUND_BY
);
924 } else if (t
== JOB_STOP
)
925 job_fail_dependencies(u
, UNIT_CONFLICTED_BY
);
928 /* Trigger OnFailure dependencies that are not generated by
929 * the unit itself. We don't treat JOB_CANCELED as failure in
930 * this context. And JOB_FAILURE is already handled by the
932 if (IN_SET(result
, JOB_TIMEOUT
, JOB_DEPENDENCY
)) {
933 log_struct(LOG_NOTICE
,
934 "JOB_TYPE=%s", job_type_to_string(t
),
935 "JOB_RESULT=%s", job_result_to_string(result
),
937 LOG_UNIT_MESSAGE(u
, "Job %s/%s failed with result '%s'.",
939 job_type_to_string(t
),
940 job_result_to_string(result
)));
942 unit_start_on_failure(u
);
945 unit_trigger_notify(u
);
948 /* Try to start the next jobs that can be started */
949 HASHMAP_FOREACH_KEY(v
, other
, u
->dependencies
[UNIT_AFTER
], i
)
951 job_add_to_run_queue(other
->job
);
952 job_add_to_gc_queue(other
->job
);
954 HASHMAP_FOREACH_KEY(v
, other
, u
->dependencies
[UNIT_BEFORE
], i
)
956 job_add_to_run_queue(other
->job
);
957 job_add_to_gc_queue(other
->job
);
960 manager_check_finished(u
->manager
);
965 static int job_dispatch_timer(sd_event_source
*s
, uint64_t monotonic
, void *userdata
) {
970 assert(s
== j
->timer_event_source
);
972 log_unit_warning(j
->unit
, "Job %s/%s timed out.", j
->unit
->id
, job_type_to_string(j
->type
));
975 job_finish_and_invalidate(j
, JOB_TIMEOUT
, true, false);
977 emergency_action(u
->manager
, u
->job_timeout_action
,
978 EMERGENCY_ACTION_IS_WATCHDOG
|EMERGENCY_ACTION_WARN
,
979 u
->job_timeout_reboot_arg
, "job timed out");
984 int job_start_timer(Job
*j
, bool job_running
) {
986 usec_t timeout_time
, old_timeout_time
;
989 j
->begin_running_usec
= now(CLOCK_MONOTONIC
);
991 if (j
->unit
->job_running_timeout
== USEC_INFINITY
)
994 timeout_time
= usec_add(j
->begin_running_usec
, j
->unit
->job_running_timeout
);
996 if (j
->timer_event_source
) {
997 /* Update only if JobRunningTimeoutSec= results in earlier timeout */
998 r
= sd_event_source_get_time(j
->timer_event_source
, &old_timeout_time
);
1002 if (old_timeout_time
<= timeout_time
)
1005 return sd_event_source_set_time(j
->timer_event_source
, timeout_time
);
1008 if (j
->timer_event_source
)
1011 j
->begin_usec
= now(CLOCK_MONOTONIC
);
1013 if (j
->unit
->job_timeout
== USEC_INFINITY
)
1016 timeout_time
= usec_add(j
->begin_usec
, j
->unit
->job_timeout
);
1019 r
= sd_event_add_time(
1021 &j
->timer_event_source
,
1024 job_dispatch_timer
, j
);
1028 (void) sd_event_source_set_description(j
->timer_event_source
, "job-start");
1033 void job_add_to_run_queue(Job
*j
) {
1035 assert(j
->installed
);
1037 if (j
->in_run_queue
)
1040 if (!j
->manager
->run_queue
)
1041 sd_event_source_set_enabled(j
->manager
->run_queue_event_source
, SD_EVENT_ONESHOT
);
1043 LIST_PREPEND(run_queue
, j
->manager
->run_queue
, j
);
1044 j
->in_run_queue
= true;
1047 void job_add_to_dbus_queue(Job
*j
) {
1049 assert(j
->installed
);
1051 if (j
->in_dbus_queue
)
1054 /* We don't check if anybody is subscribed here, since this
1055 * job might just have been created and not yet assigned to a
1056 * connection/client. */
1058 LIST_PREPEND(dbus_queue
, j
->manager
->dbus_job_queue
, j
);
1059 j
->in_dbus_queue
= true;
1062 char *job_dbus_path(Job
*j
) {
1067 if (asprintf(&p
, "/org/freedesktop/systemd1/job/%"PRIu32
, j
->id
) < 0)
1073 int job_serialize(Job
*j
, FILE *f
) {
1077 fprintf(f
, "job-id=%u\n", j
->id
);
1078 fprintf(f
, "job-type=%s\n", job_type_to_string(j
->type
));
1079 fprintf(f
, "job-state=%s\n", job_state_to_string(j
->state
));
1080 fprintf(f
, "job-irreversible=%s\n", yes_no(j
->irreversible
));
1081 fprintf(f
, "job-sent-dbus-new-signal=%s\n", yes_no(j
->sent_dbus_new_signal
));
1082 fprintf(f
, "job-ignore-order=%s\n", yes_no(j
->ignore_order
));
1084 if (j
->begin_usec
> 0)
1085 fprintf(f
, "job-begin="USEC_FMT
"\n", j
->begin_usec
);
1086 if (j
->begin_running_usec
> 0)
1087 fprintf(f
, "job-begin-running="USEC_FMT
"\n", j
->begin_running_usec
);
1089 bus_track_serialize(j
->bus_track
, f
, "subscribed");
1096 int job_deserialize(Job
*j
, FILE *f
) {
1101 char line
[LINE_MAX
], *l
, *v
;
1104 if (!fgets(line
, sizeof(line
), f
)) {
1117 k
= strcspn(l
, "=");
1125 if (streq(l
, "job-id")) {
1127 if (safe_atou32(v
, &j
->id
) < 0)
1128 log_debug("Failed to parse job id value: %s", v
);
1130 } else if (streq(l
, "job-type")) {
1133 t
= job_type_from_string(v
);
1135 log_debug("Failed to parse job type: %s", v
);
1136 else if (t
>= _JOB_TYPE_MAX_IN_TRANSACTION
)
1137 log_debug("Cannot deserialize job of type: %s", v
);
1141 } else if (streq(l
, "job-state")) {
1144 s
= job_state_from_string(v
);
1146 log_debug("Failed to parse job state: %s", v
);
1148 job_set_state(j
, s
);
1150 } else if (streq(l
, "job-irreversible")) {
1153 b
= parse_boolean(v
);
1155 log_debug("Failed to parse job irreversible flag: %s", v
);
1157 j
->irreversible
= j
->irreversible
|| b
;
1159 } else if (streq(l
, "job-sent-dbus-new-signal")) {
1162 b
= parse_boolean(v
);
1164 log_debug("Failed to parse job sent_dbus_new_signal flag: %s", v
);
1166 j
->sent_dbus_new_signal
= j
->sent_dbus_new_signal
|| b
;
1168 } else if (streq(l
, "job-ignore-order")) {
1171 b
= parse_boolean(v
);
1173 log_debug("Failed to parse job ignore_order flag: %s", v
);
1175 j
->ignore_order
= j
->ignore_order
|| b
;
1177 } else if (streq(l
, "job-begin")) {
1178 unsigned long long ull
;
1180 if (sscanf(v
, "%llu", &ull
) != 1)
1181 log_debug("Failed to parse job-begin value %s", v
);
1183 j
->begin_usec
= ull
;
1185 } else if (streq(l
, "job-begin-running")) {
1186 unsigned long long ull
;
1188 if (sscanf(v
, "%llu", &ull
) != 1)
1189 log_debug("Failed to parse job-begin-running value %s", v
);
1191 j
->begin_running_usec
= ull
;
1193 } else if (streq(l
, "subscribed")) {
1195 if (strv_extend(&j
->deserialized_clients
, v
) < 0)
1201 int job_coldplug(Job
*j
) {
1203 usec_t timeout_time
= USEC_INFINITY
;
1207 /* After deserialization is complete and the bus connection
1208 * set up again, let's start watching our subscribers again */
1209 (void) bus_job_coldplug_bus_track(j
);
1211 if (j
->state
== JOB_WAITING
)
1212 job_add_to_run_queue(j
);
1214 /* Maybe due to new dependencies we don't actually need this job anymore? */
1215 job_add_to_gc_queue(j
);
1217 /* Create timer only when job began or began running and the respective timeout is finite.
1218 * Follow logic of job_start_timer() if both timeouts are finite */
1219 if (j
->begin_usec
== 0)
1222 if (j
->unit
->job_timeout
!= USEC_INFINITY
)
1223 timeout_time
= usec_add(j
->begin_usec
, j
->unit
->job_timeout
);
1225 if (j
->begin_running_usec
> 0 && j
->unit
->job_running_timeout
!= USEC_INFINITY
)
1226 timeout_time
= MIN(timeout_time
, usec_add(j
->begin_running_usec
, j
->unit
->job_running_timeout
));
1228 if (timeout_time
== USEC_INFINITY
)
1231 j
->timer_event_source
= sd_event_source_unref(j
->timer_event_source
);
1233 r
= sd_event_add_time(
1235 &j
->timer_event_source
,
1238 job_dispatch_timer
, j
);
1240 log_debug_errno(r
, "Failed to restart timeout for job: %m");
1242 (void) sd_event_source_set_description(j
->timer_event_source
, "job-timeout");
1247 void job_shutdown_magic(Job
*j
) {
1250 /* The shutdown target gets some special treatment here: we
1251 * tell the kernel to begin with flushing its disk caches, to
1252 * optimize shutdown time a bit. Ideally we wouldn't hardcode
1253 * this magic into PID 1. However all other processes aren't
1254 * options either since they'd exit much sooner than PID 1 and
1255 * asynchronous sync() would cause their exit to be
1258 if (j
->type
!= JOB_START
)
1261 if (!MANAGER_IS_SYSTEM(j
->unit
->manager
))
1264 if (!unit_has_name(j
->unit
, SPECIAL_SHUTDOWN_TARGET
))
1267 /* In case messages on console has been disabled on boot */
1268 j
->unit
->manager
->no_console_output
= false;
1270 if (detect_container() > 0)
1273 (void) asynchronous_sync(NULL
);
1276 int job_get_timeout(Job
*j
, usec_t
*timeout
) {
1277 usec_t x
= USEC_INFINITY
, y
= USEC_INFINITY
;
1283 if (j
->timer_event_source
) {
1284 r
= sd_event_source_get_time(j
->timer_event_source
, &x
);
1289 if (UNIT_VTABLE(u
)->get_timeout
) {
1290 r
= UNIT_VTABLE(u
)->get_timeout(u
, &y
);
1295 if (x
== USEC_INFINITY
&& y
== USEC_INFINITY
)
1298 *timeout
= MIN(x
, y
);
1302 bool job_may_gc(Job
*j
) {
1309 /* Checks whether this job should be GC'ed away. We only do this for jobs of units that have no effect on their
1310 * own and just track external state. For now the only unit type that qualifies for this are .device units.
1311 * Returns true if the job can be collected. */
1313 if (!UNIT_VTABLE(j
->unit
)->gc_jobs
)
1316 if (sd_bus_track_count(j
->bus_track
) > 0)
1319 /* FIXME: So this is a bit ugly: for now we don't properly track references made via private bus connections
1320 * (because it's nasty, as sd_bus_track doesn't apply to it). We simply remember that the job was once
1321 * referenced by one, and reset this whenever we notice that no private bus connections are around. This means
1322 * the GC is a bit too conservative when it comes to jobs created by private bus connections. */
1323 if (j
->ref_by_private_bus
) {
1324 if (set_isempty(j
->unit
->manager
->private_buses
))
1325 j
->ref_by_private_bus
= false;
1330 if (j
->type
== JOB_NOP
)
1333 /* If a job is ordered after ours, and is to be started, then it needs to wait for us, regardless if we stop or
1334 * start, hence let's not GC in that case. */
1335 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
) {
1339 if (other
->job
->ignore_order
)
1342 if (IN_SET(other
->job
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
))
1346 /* If we are going down, but something else is ordered After= us, then it needs to wait for us */
1347 if (IN_SET(j
->type
, JOB_STOP
, JOB_RESTART
))
1348 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
) {
1352 if (other
->job
->ignore_order
)
1358 /* The logic above is kinda the inverse of the job_is_runnable() logic. Specifically, if the job "we" is
1359 * ordered before the job "other":
1361 * we start + other start → stay
1362 * we start + other stop → gc
1363 * we stop + other start → stay
1364 * we stop + other stop → gc
1366 * "we" are ordered after "other":
1368 * we start + other start → gc
1369 * we start + other stop → gc
1370 * we stop + other start → stay
1371 * we stop + other stop → stay
1377 void job_add_to_gc_queue(Job
*j
) {
1386 LIST_PREPEND(gc_queue
, j
->unit
->manager
->gc_job_queue
, j
);
1387 j
->in_gc_queue
= true;
1390 static int job_compare(Job
* const *a
, Job
* const *b
) {
1391 return CMP((*a
)->id
, (*b
)->id
);
1394 static size_t sort_job_list(Job
**list
, size_t n
) {
1395 Job
*previous
= NULL
;
1398 /* Order by numeric IDs */
1399 typesafe_qsort(list
, n
, job_compare
);
1401 /* Filter out duplicates */
1402 for (a
= 0, b
= 0; a
< n
; a
++) {
1404 if (previous
== list
[a
])
1407 previous
= list
[b
++] = list
[a
];
1413 int job_get_before(Job
*j
, Job
*** ret
) {
1414 _cleanup_free_ Job
** list
= NULL
;
1415 size_t n
= 0, n_allocated
= 0;
1420 /* Returns a list of all pending jobs that need to finish before this job may be started. */
1425 if (j
->ignore_order
) {
1430 if (IN_SET(j
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
)) {
1432 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
) {
1436 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1438 list
[n
++] = other
->job
;
1442 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
) {
1446 if (!IN_SET(other
->job
->type
, JOB_STOP
, JOB_RESTART
))
1449 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1451 list
[n
++] = other
->job
;
1454 n
= sort_job_list(list
, n
);
1456 *ret
= TAKE_PTR(list
);
1461 int job_get_after(Job
*j
, Job
*** ret
) {
1462 _cleanup_free_ Job
** list
= NULL
;
1463 size_t n
= 0, n_allocated
= 0;
1471 /* Returns a list of all pending jobs that are waiting for this job to finish. */
1473 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
) {
1477 if (other
->job
->ignore_order
)
1480 if (!IN_SET(other
->job
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
))
1483 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1485 list
[n
++] = other
->job
;
1488 if (IN_SET(j
->type
, JOB_STOP
, JOB_RESTART
)) {
1490 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
) {
1494 if (other
->job
->ignore_order
)
1497 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1499 list
[n
++] = other
->job
;
1503 n
= sort_job_list(list
, n
);
1505 *ret
= TAKE_PTR(list
);
1510 static const char* const job_state_table
[_JOB_STATE_MAX
] = {
1511 [JOB_WAITING
] = "waiting",
1512 [JOB_RUNNING
] = "running",
1515 DEFINE_STRING_TABLE_LOOKUP(job_state
, JobState
);
1517 static const char* const job_type_table
[_JOB_TYPE_MAX
] = {
1518 [JOB_START
] = "start",
1519 [JOB_VERIFY_ACTIVE
] = "verify-active",
1520 [JOB_STOP
] = "stop",
1521 [JOB_RELOAD
] = "reload",
1522 [JOB_RELOAD_OR_START
] = "reload-or-start",
1523 [JOB_RESTART
] = "restart",
1524 [JOB_TRY_RESTART
] = "try-restart",
1525 [JOB_TRY_RELOAD
] = "try-reload",
1529 DEFINE_STRING_TABLE_LOOKUP(job_type
, JobType
);
1531 static const char* const job_mode_table
[_JOB_MODE_MAX
] = {
1532 [JOB_FAIL
] = "fail",
1533 [JOB_REPLACE
] = "replace",
1534 [JOB_REPLACE_IRREVERSIBLY
] = "replace-irreversibly",
1535 [JOB_ISOLATE
] = "isolate",
1536 [JOB_FLUSH
] = "flush",
1537 [JOB_IGNORE_DEPENDENCIES
] = "ignore-dependencies",
1538 [JOB_IGNORE_REQUIREMENTS
] = "ignore-requirements",
1541 DEFINE_STRING_TABLE_LOOKUP(job_mode
, JobMode
);
1543 static const char* const job_result_table
[_JOB_RESULT_MAX
] = {
1544 [JOB_DONE
] = "done",
1545 [JOB_CANCELED
] = "canceled",
1546 [JOB_TIMEOUT
] = "timeout",
1547 [JOB_FAILED
] = "failed",
1548 [JOB_DEPENDENCY
] = "dependency",
1549 [JOB_SKIPPED
] = "skipped",
1550 [JOB_INVALID
] = "invalid",
1551 [JOB_ASSERT
] = "assert",
1552 [JOB_UNSUPPORTED
] = "unsupported",
1553 [JOB_COLLECTED
] = "collected",
1554 [JOB_ONCE
] = "once",
1557 DEFINE_STRING_TABLE_LOOKUP(job_result
, JobResult
);
1559 const char* job_type_to_access_method(JobType t
) {
1561 assert(t
< _JOB_TYPE_MAX
);
1563 if (IN_SET(t
, JOB_START
, JOB_RESTART
, JOB_TRY_RESTART
))
1565 else if (t
== JOB_STOP
)