1 /* SPDX-License-Identifier: LGPL-2.1+ */
6 #include "sd-messages.h"
8 #include "alloc-util.h"
17 #include "parse-util.h"
18 #include "serialize.h"
21 #include "stdio-util.h"
22 #include "string-table.h"
23 #include "string-util.h"
25 #include "terminal-util.h"
29 Job
* job_new_raw(Unit
*unit
) {
32 /* used for deserialization */
41 .manager
= unit
->manager
,
43 .type
= _JOB_TYPE_INVALID
,
49 Job
* job_new(Unit
*unit
, JobType type
) {
52 assert(type
< _JOB_TYPE_MAX
);
54 j
= job_new_raw(unit
);
58 j
->id
= j
->manager
->current_job_id
++;
61 /* We don't link it here, that's what job_dependency() is for */
66 void job_unlink(Job
*j
) {
68 assert(!j
->installed
);
69 assert(!j
->transaction_prev
);
70 assert(!j
->transaction_next
);
71 assert(!j
->subject_list
);
72 assert(!j
->object_list
);
74 if (j
->in_run_queue
) {
75 LIST_REMOVE(run_queue
, j
->manager
->run_queue
, j
);
76 j
->in_run_queue
= false;
79 if (j
->in_dbus_queue
) {
80 LIST_REMOVE(dbus_queue
, j
->manager
->dbus_job_queue
, j
);
81 j
->in_dbus_queue
= false;
85 LIST_REMOVE(gc_queue
, j
->manager
->gc_job_queue
, j
);
86 j
->in_gc_queue
= false;
89 j
->timer_event_source
= sd_event_source_unref(j
->timer_event_source
);
92 Job
* job_free(Job
*j
) {
94 assert(!j
->installed
);
95 assert(!j
->transaction_prev
);
96 assert(!j
->transaction_next
);
97 assert(!j
->subject_list
);
98 assert(!j
->object_list
);
102 sd_bus_track_unref(j
->bus_track
);
103 strv_free(j
->deserialized_clients
);
108 static void job_set_state(Job
*j
, JobState state
) {
111 assert(state
< _JOB_STATE_MAX
);
113 if (j
->state
== state
)
121 if (j
->state
== JOB_RUNNING
)
122 j
->unit
->manager
->n_running_jobs
++;
124 assert(j
->state
== JOB_WAITING
);
125 assert(j
->unit
->manager
->n_running_jobs
> 0);
127 j
->unit
->manager
->n_running_jobs
--;
129 if (j
->unit
->manager
->n_running_jobs
<= 0)
130 j
->unit
->manager
->jobs_in_progress_event_source
= sd_event_source_unref(j
->unit
->manager
->jobs_in_progress_event_source
);
134 void job_uninstall(Job
*j
) {
137 assert(j
->installed
);
139 job_set_state(j
, JOB_WAITING
);
141 pj
= (j
->type
== JOB_NOP
) ? &j
->unit
->nop_job
: &j
->unit
->job
;
144 /* Detach from next 'bigger' objects */
146 /* daemon-reload should be transparent to job observers */
147 if (!MANAGER_IS_RELOADING(j
->manager
))
148 bus_job_send_removed_signal(j
);
152 unit_add_to_gc_queue(j
->unit
);
154 unit_add_to_dbus_queue(j
->unit
); /* The Job property of the unit has changed now */
156 hashmap_remove_value(j
->manager
->jobs
, UINT32_TO_PTR(j
->id
), j
);
157 j
->installed
= false;
160 static bool job_type_allows_late_merge(JobType t
) {
161 /* Tells whether it is OK to merge a job of type 't' with an already
163 * Reloads cannot be merged this way. Think of the sequence:
164 * 1. Reload of a daemon is in progress; the daemon has already loaded
165 * its config file, but hasn't completed the reload operation yet.
166 * 2. Edit foo's config file.
167 * 3. Trigger another reload to have the daemon use the new config.
168 * Should the second reload job be merged into the first one, the daemon
169 * would not know about the new config.
170 * JOB_RESTART jobs on the other hand can be merged, because they get
171 * patched into JOB_START after stopping the unit. So if we see a
172 * JOB_RESTART running, it means the unit hasn't stopped yet and at
173 * this time the merge is still allowed. */
174 return t
!= JOB_RELOAD
;
177 static void job_merge_into_installed(Job
*j
, Job
*other
) {
178 assert(j
->installed
);
179 assert(j
->unit
== other
->unit
);
181 if (j
->type
!= JOB_NOP
)
182 assert_se(job_type_merge_and_collapse(&j
->type
, other
->type
, j
->unit
) == 0);
184 assert(other
->type
== JOB_NOP
);
186 j
->irreversible
= j
->irreversible
|| other
->irreversible
;
187 j
->ignore_order
= j
->ignore_order
|| other
->ignore_order
;
190 Job
* job_install(Job
*j
) {
194 assert(!j
->installed
);
195 assert(j
->type
< _JOB_TYPE_MAX_IN_TRANSACTION
);
196 assert(j
->state
== JOB_WAITING
);
198 pj
= (j
->type
== JOB_NOP
) ? &j
->unit
->nop_job
: &j
->unit
->job
;
202 if (job_type_is_conflicting(uj
->type
, j
->type
))
203 job_finish_and_invalidate(uj
, JOB_CANCELED
, false, false);
205 /* not conflicting, i.e. mergeable */
207 if (uj
->state
== JOB_WAITING
||
208 (job_type_allows_late_merge(j
->type
) && job_type_is_superset(uj
->type
, j
->type
))) {
209 job_merge_into_installed(uj
, j
);
210 log_unit_debug(uj
->unit
,
211 "Merged %s/%s into installed job %s/%s as %"PRIu32
,
212 j
->unit
->id
, job_type_to_string(j
->type
), uj
->unit
->id
,
213 job_type_to_string(uj
->type
), uj
->id
);
216 /* already running and not safe to merge into */
217 /* Patch uj to become a merged job and re-run it. */
218 /* XXX It should be safer to queue j to run after uj finishes, but it is
219 * not currently possible to have more than one installed job per unit. */
220 job_merge_into_installed(uj
, j
);
221 log_unit_debug(uj
->unit
,
222 "Merged into running job, re-running: %s/%s as %"PRIu32
,
223 uj
->unit
->id
, job_type_to_string(uj
->type
), uj
->id
);
225 job_set_state(uj
, JOB_WAITING
);
231 /* Install the job */
235 j
->manager
->n_installed_jobs
++;
236 log_unit_debug(j
->unit
,
237 "Installed new job %s/%s as %u",
238 j
->unit
->id
, job_type_to_string(j
->type
), (unsigned) j
->id
);
240 job_add_to_gc_queue(j
);
242 job_add_to_dbus_queue(j
); /* announce this job to clients */
243 unit_add_to_dbus_queue(j
->unit
); /* The Job property of the unit has changed now */
248 int job_install_deserialized(Job
*j
) {
252 assert(!j
->installed
);
254 if (j
->type
< 0 || j
->type
>= _JOB_TYPE_MAX_IN_TRANSACTION
)
255 return log_unit_debug_errno(j
->unit
, SYNTHETIC_ERRNO(EINVAL
),
256 "Invalid job type %s in deserialization.",
257 strna(job_type_to_string(j
->type
)));
259 pj
= (j
->type
== JOB_NOP
) ? &j
->unit
->nop_job
: &j
->unit
->job
;
261 return log_unit_debug_errno(j
->unit
, SYNTHETIC_ERRNO(EEXIST
),
262 "Unit already has a job installed. Not installing deserialized job.");
264 r
= hashmap_put(j
->manager
->jobs
, UINT32_TO_PTR(j
->id
), j
);
266 return log_unit_debug_errno(j
->unit
, r
, "Job ID %" PRIu32
" already used, cannot deserialize job.", j
->id
);
268 return log_unit_debug_errno(j
->unit
, r
, "Failed to insert job into jobs hash table: %m");
273 if (j
->state
== JOB_RUNNING
)
274 j
->unit
->manager
->n_running_jobs
++;
276 log_unit_debug(j
->unit
,
277 "Reinstalled deserialized job %s/%s as %u",
278 j
->unit
->id
, job_type_to_string(j
->type
), (unsigned) j
->id
);
282 JobDependency
* job_dependency_new(Job
*subject
, Job
*object
, bool matters
, bool conflicts
) {
287 /* Adds a new job link, which encodes that the 'subject' job
288 * needs the 'object' job in some way. If 'subject' is NULL
289 * this means the 'anchor' job (i.e. the one the user
290 * explicitly asked for) is the requester. */
292 l
= new0(JobDependency
, 1);
296 l
->subject
= subject
;
298 l
->matters
= matters
;
299 l
->conflicts
= conflicts
;
302 LIST_PREPEND(subject
, subject
->subject_list
, l
);
304 LIST_PREPEND(object
, object
->object_list
, l
);
309 void job_dependency_free(JobDependency
*l
) {
313 LIST_REMOVE(subject
, l
->subject
->subject_list
, l
);
315 LIST_REMOVE(object
, l
->object
->object_list
, l
);
320 void job_dump(Job
*j
, FILE *f
, const char *prefix
) {
324 prefix
= strempty(prefix
);
328 "%s\tAction: %s -> %s\n"
330 "%s\tIrreversible: %s\n"
333 prefix
, j
->unit
->id
, job_type_to_string(j
->type
),
334 prefix
, job_state_to_string(j
->state
),
335 prefix
, yes_no(j
->irreversible
),
336 prefix
, yes_no(job_may_gc(j
)));
340 * Merging is commutative, so imagine the matrix as symmetric. We store only
341 * its lower triangle to avoid duplication. We don't store the main diagonal,
342 * because A merged with A is simply A.
344 * If the resulting type is collapsed immediately afterwards (to get rid of
345 * the JOB_RELOAD_OR_START, which lies outside the lookup function's domain),
346 * the following properties hold:
348 * Merging is associative! A merged with B, and then merged with C is the same
349 * as A merged with the result of B merged with C.
351 * Mergeability is transitive! If A can be merged with B and B with C then
354 * Also, if A merged with B cannot be merged with C, then either A or B cannot
355 * be merged with C either.
357 static const JobType job_merging_table
[] = {
358 /* What \ With * JOB_START JOB_VERIFY_ACTIVE JOB_STOP JOB_RELOAD */
359 /*********************************************************************************/
361 /*JOB_VERIFY_ACTIVE */ JOB_START
,
362 /*JOB_STOP */ -1, -1,
363 /*JOB_RELOAD */ JOB_RELOAD_OR_START
, JOB_RELOAD
, -1,
364 /*JOB_RESTART */ JOB_RESTART
, JOB_RESTART
, -1, JOB_RESTART
,
367 JobType
job_type_lookup_merge(JobType a
, JobType b
) {
368 assert_cc(ELEMENTSOF(job_merging_table
) == _JOB_TYPE_MAX_MERGING
* (_JOB_TYPE_MAX_MERGING
- 1) / 2);
369 assert(a
>= 0 && a
< _JOB_TYPE_MAX_MERGING
);
370 assert(b
>= 0 && b
< _JOB_TYPE_MAX_MERGING
);
381 return job_merging_table
[(a
- 1) * a
/ 2 + b
];
384 bool job_type_is_redundant(JobType a
, UnitActiveState b
) {
388 return IN_SET(b
, UNIT_ACTIVE
, UNIT_RELOADING
);
391 return IN_SET(b
, UNIT_INACTIVE
, UNIT_FAILED
);
393 case JOB_VERIFY_ACTIVE
:
394 return IN_SET(b
, UNIT_ACTIVE
, UNIT_RELOADING
);
402 b
== UNIT_ACTIVATING
;
408 assert_not_reached("Invalid job type");
412 JobType
job_type_collapse(JobType t
, Unit
*u
) {
417 case JOB_TRY_RESTART
:
418 s
= unit_active_state(u
);
419 if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s
))
425 s
= unit_active_state(u
);
426 if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s
))
431 case JOB_RELOAD_OR_START
:
432 s
= unit_active_state(u
);
433 if (UNIT_IS_INACTIVE_OR_DEACTIVATING(s
))
443 int job_type_merge_and_collapse(JobType
*a
, JobType b
, Unit
*u
) {
446 t
= job_type_lookup_merge(*a
, b
);
450 *a
= job_type_collapse(t
, u
);
454 static bool job_is_runnable(Job
*j
) {
460 assert(j
->installed
);
462 /* Checks whether there is any job running for the units this
463 * job needs to be running after (in the case of a 'positive'
464 * job type) or before (in the case of a 'negative' job
467 /* Note that unit types have a say in what is runnable,
468 * too. For example, if they return -EAGAIN from
469 * unit_start() they can indicate they are not
472 /* First check if there is an override */
476 if (j
->type
== JOB_NOP
)
479 if (IN_SET(j
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
)) {
480 /* Immediate result is that the job is or might be
481 * started. In this case let's wait for the
482 * dependencies, regardless whether they are
483 * starting or stopping something. */
485 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
)
490 /* Also, if something else is being stopped and we should
491 * change state after it, then let's wait. */
493 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
)
495 IN_SET(other
->job
->type
, JOB_STOP
, JOB_RESTART
))
498 /* This means that for a service a and a service b where b
499 * shall be started after a:
501 * start a + start b → 1st step start a, 2nd step start b
502 * start a + stop b → 1st step stop b, 2nd step start a
503 * stop a + start b → 1st step stop a, 2nd step start b
504 * stop a + stop b → 1st step stop b, 2nd step stop a
506 * This has the side effect that restarts are properly
507 * synchronized too. */
512 static void job_change_type(Job
*j
, JobType newtype
) {
515 log_unit_debug(j
->unit
,
516 "Converting job %s/%s -> %s/%s",
517 j
->unit
->id
, job_type_to_string(j
->type
),
518 j
->unit
->id
, job_type_to_string(newtype
));
523 _pure_
static const char* job_get_begin_status_message_format(Unit
*u
, JobType t
) {
529 return "Reloading %s.";
531 assert(IN_SET(t
, JOB_START
, JOB_STOP
));
533 format
= UNIT_VTABLE(u
)->status_message_formats
.starting_stopping
[t
== JOB_STOP
];
537 /* Return generic strings */
539 return "Starting %s.";
541 assert(t
== JOB_STOP
);
542 return "Stopping %s.";
546 static void job_print_begin_status_message(Unit
*u
, JobType t
) {
551 /* Reload status messages have traditionally not been printed to console. */
552 if (!IN_SET(t
, JOB_START
, JOB_STOP
))
555 format
= job_get_begin_status_message_format(u
, t
);
557 DISABLE_WARNING_FORMAT_NONLITERAL
;
558 unit_status_printf(u
, "", format
);
562 static void job_log_begin_status_message(Unit
*u
, uint32_t job_id
, JobType t
) {
563 const char *format
, *mid
;
568 assert(t
< _JOB_TYPE_MAX
);
570 if (!IN_SET(t
, JOB_START
, JOB_STOP
, JOB_RELOAD
))
573 if (log_on_console()) /* Skip this if it would only go on the console anyway */
576 /* We log status messages for all units and all operations. */
578 format
= job_get_begin_status_message_format(u
, t
);
580 DISABLE_WARNING_FORMAT_NONLITERAL
;
581 (void) snprintf(buf
, sizeof buf
, format
, unit_description(u
));
584 mid
= t
== JOB_START
? "MESSAGE_ID=" SD_MESSAGE_UNIT_STARTING_STR
:
585 t
== JOB_STOP
? "MESSAGE_ID=" SD_MESSAGE_UNIT_STOPPING_STR
:
586 "MESSAGE_ID=" SD_MESSAGE_UNIT_RELOADING_STR
;
588 /* Note that we deliberately use LOG_MESSAGE() instead of
589 * LOG_UNIT_MESSAGE() here, since this is supposed to mimic
590 * closely what is written to screen using the status output,
591 * which is supposed the highest level, friendliest output
592 * possible, which means we should avoid the low-level unit
595 LOG_MESSAGE("%s", buf
),
596 "JOB_ID=%" PRIu32
, job_id
,
597 "JOB_TYPE=%s", job_type_to_string(t
),
599 LOG_UNIT_INVOCATION_ID(u
),
603 static void job_emit_begin_status_message(Unit
*u
, uint32_t job_id
, JobType t
) {
606 assert(t
< _JOB_TYPE_MAX
);
608 job_log_begin_status_message(u
, job_id
, t
);
609 job_print_begin_status_message(u
, t
);
612 static int job_perform_on_unit(Job
**j
) {
619 /* While we execute this operation the job might go away (for
620 * example: because it finishes immediately or is replaced by
621 * a new, conflicting job.) To make sure we don't access a
622 * freed job later on we store the id here, so that we can
623 * verify the job is still valid. */
650 assert_not_reached("Invalid job type");
653 /* Log if the job still exists and the start/stop/reload function actually did something. Note that this means
654 * for units for which there's no 'activating' phase (i.e. because we transition directly from 'inactive' to
655 * 'active') we'll possibly skip the "Starting..." message. */
656 *j
= manager_get_job(m
, id
);
658 job_emit_begin_status_message(u
, id
, t
);
663 int job_run_and_invalidate(Job
*j
) {
667 assert(j
->installed
);
668 assert(j
->type
< _JOB_TYPE_MAX_IN_TRANSACTION
);
669 assert(j
->in_run_queue
);
671 LIST_REMOVE(run_queue
, j
->manager
->run_queue
, j
);
672 j
->in_run_queue
= false;
674 if (j
->state
!= JOB_WAITING
)
677 if (!job_is_runnable(j
))
680 job_start_timer(j
, true);
681 job_set_state(j
, JOB_RUNNING
);
682 job_add_to_dbus_queue(j
);
686 case JOB_VERIFY_ACTIVE
: {
689 t
= unit_active_state(j
->unit
);
690 if (UNIT_IS_ACTIVE_OR_RELOADING(t
))
692 else if (t
== UNIT_ACTIVATING
)
702 r
= job_perform_on_unit(&j
);
704 /* If the unit type does not support starting/stopping, then simply wait. */
710 r
= job_perform_on_unit(&j
);
718 assert_not_reached("Unknown job type");
723 job_set_state(j
, JOB_WAITING
); /* Hmm, not ready after all, let's return to JOB_WAITING state */
724 else if (r
== -EALREADY
) /* already being executed */
725 r
= job_finish_and_invalidate(j
, JOB_DONE
, true, true);
726 else if (r
== -ECOMM
) /* condition failed, but all is good */
727 r
= job_finish_and_invalidate(j
, JOB_DONE
, true, false);
728 else if (r
== -EBADR
)
729 r
= job_finish_and_invalidate(j
, JOB_SKIPPED
, true, false);
730 else if (r
== -ENOEXEC
)
731 r
= job_finish_and_invalidate(j
, JOB_INVALID
, true, false);
732 else if (r
== -EPROTO
)
733 r
= job_finish_and_invalidate(j
, JOB_ASSERT
, true, false);
734 else if (r
== -EOPNOTSUPP
)
735 r
= job_finish_and_invalidate(j
, JOB_UNSUPPORTED
, true, false);
736 else if (r
== -ENOLINK
)
737 r
= job_finish_and_invalidate(j
, JOB_DEPENDENCY
, true, false);
738 else if (r
== -ESTALE
)
739 r
= job_finish_and_invalidate(j
, JOB_ONCE
, true, false);
741 r
= job_finish_and_invalidate(j
, JOB_FAILED
, true, false);
747 _pure_
static const char *job_get_done_status_message_format(Unit
*u
, JobType t
, JobResult result
) {
749 static const char *const generic_finished_start_job
[_JOB_RESULT_MAX
] = {
750 [JOB_DONE
] = "Started %s.",
751 [JOB_TIMEOUT
] = "Timed out starting %s.",
752 [JOB_FAILED
] = "Failed to start %s.",
753 [JOB_DEPENDENCY
] = "Dependency failed for %s.",
754 [JOB_ASSERT
] = "Assertion failed for %s.",
755 [JOB_UNSUPPORTED
] = "Starting of %s not supported.",
756 [JOB_COLLECTED
] = "Unnecessary job for %s was removed.",
757 [JOB_ONCE
] = "Unit %s has been started before and cannot be started again."
759 static const char *const generic_finished_stop_job
[_JOB_RESULT_MAX
] = {
760 [JOB_DONE
] = "Stopped %s.",
761 [JOB_FAILED
] = "Stopped (with error) %s.",
762 [JOB_TIMEOUT
] = "Timed out stopping %s.",
764 static const char *const generic_finished_reload_job
[_JOB_RESULT_MAX
] = {
765 [JOB_DONE
] = "Reloaded %s.",
766 [JOB_FAILED
] = "Reload failed for %s.",
767 [JOB_TIMEOUT
] = "Timed out reloading %s.",
769 /* When verify-active detects the unit is inactive, report it.
770 * Most likely a DEPEND warning from a requisiting unit will
771 * occur next and it's nice to see what was requisited. */
772 static const char *const generic_finished_verify_active_job
[_JOB_RESULT_MAX
] = {
773 [JOB_SKIPPED
] = "%s is not active.",
780 assert(t
< _JOB_TYPE_MAX
);
782 if (IN_SET(t
, JOB_START
, JOB_STOP
, JOB_RESTART
)) {
783 format
= t
== JOB_START
?
784 UNIT_VTABLE(u
)->status_message_formats
.finished_start_job
[result
] :
785 UNIT_VTABLE(u
)->status_message_formats
.finished_stop_job
[result
];
790 /* Return generic strings */
792 return generic_finished_start_job
[result
];
793 else if (IN_SET(t
, JOB_STOP
, JOB_RESTART
))
794 return generic_finished_stop_job
[result
];
795 else if (t
== JOB_RELOAD
)
796 return generic_finished_reload_job
[result
];
797 else if (t
== JOB_VERIFY_ACTIVE
)
798 return generic_finished_verify_active_job
[result
];
803 static const struct {
804 const char *color
, *word
;
805 } job_print_done_status_messages
[_JOB_RESULT_MAX
] = {
806 [JOB_DONE
] = { ANSI_OK_COLOR
, " OK " },
807 [JOB_TIMEOUT
] = { ANSI_HIGHLIGHT_RED
, " TIME " },
808 [JOB_FAILED
] = { ANSI_HIGHLIGHT_RED
, "FAILED" },
809 [JOB_DEPENDENCY
] = { ANSI_HIGHLIGHT_YELLOW
, "DEPEND" },
810 [JOB_SKIPPED
] = { ANSI_HIGHLIGHT
, " INFO " },
811 [JOB_ASSERT
] = { ANSI_HIGHLIGHT_YELLOW
, "ASSERT" },
812 [JOB_UNSUPPORTED
] = { ANSI_HIGHLIGHT_YELLOW
, "UNSUPP" },
814 [JOB_ONCE
] = { ANSI_HIGHLIGHT_RED
, " ONCE " },
817 static void job_print_done_status_message(Unit
*u
, JobType t
, JobResult result
) {
823 assert(t
< _JOB_TYPE_MAX
);
825 /* Reload status messages have traditionally not been printed to console. */
829 /* No message if the job did not actually do anything due to failed condition. */
830 if (t
== JOB_START
&& result
== JOB_DONE
&& !u
->condition_result
)
833 if (!job_print_done_status_messages
[result
].word
)
836 format
= job_get_done_status_message_format(u
, t
, result
);
840 if (log_get_show_color())
841 status
= strjoina(job_print_done_status_messages
[result
].color
,
842 job_print_done_status_messages
[result
].word
,
845 status
= job_print_done_status_messages
[result
].word
;
847 if (result
!= JOB_DONE
)
848 manager_flip_auto_status(u
->manager
, true);
850 DISABLE_WARNING_FORMAT_NONLITERAL
;
851 unit_status_printf(u
, status
, format
);
854 if (t
== JOB_START
&& result
== JOB_FAILED
) {
855 _cleanup_free_
char *quoted
;
857 quoted
= shell_maybe_quote(u
->id
, ESCAPE_BACKSLASH
);
858 manager_status_printf(u
->manager
, STATUS_TYPE_NORMAL
, NULL
, "See 'systemctl status %s' for details.", strna(quoted
));
862 static void job_log_done_status_message(Unit
*u
, uint32_t job_id
, JobType t
, JobResult result
) {
863 const char *format
, *mid
;
865 static const int job_result_log_level
[_JOB_RESULT_MAX
] = {
866 [JOB_DONE
] = LOG_INFO
,
867 [JOB_CANCELED
] = LOG_INFO
,
868 [JOB_TIMEOUT
] = LOG_ERR
,
869 [JOB_FAILED
] = LOG_ERR
,
870 [JOB_DEPENDENCY
] = LOG_WARNING
,
871 [JOB_SKIPPED
] = LOG_NOTICE
,
872 [JOB_INVALID
] = LOG_INFO
,
873 [JOB_ASSERT
] = LOG_WARNING
,
874 [JOB_UNSUPPORTED
] = LOG_WARNING
,
875 [JOB_COLLECTED
] = LOG_INFO
,
876 [JOB_ONCE
] = LOG_ERR
,
881 assert(t
< _JOB_TYPE_MAX
);
883 /* Skip printing if output goes to the console, and job_print_status_message()
884 will actually print something to the console. */
885 if (log_on_console() && job_print_done_status_messages
[result
].word
)
888 /* Show condition check message if the job did not actually do anything due to failed condition. */
889 if (t
== JOB_START
&& result
== JOB_DONE
&& !u
->condition_result
) {
891 "MESSAGE=Condition check resulted in %s being skipped.", unit_description(u
),
892 "JOB_ID=%" PRIu32
, job_id
,
893 "JOB_TYPE=%s", job_type_to_string(t
),
894 "JOB_RESULT=%s", job_result_to_string(result
),
896 LOG_UNIT_INVOCATION_ID(u
),
897 "MESSAGE_ID=" SD_MESSAGE_UNIT_STARTED_STR
);
902 format
= job_get_done_status_message_format(u
, t
, result
);
906 /* The description might be longer than the buffer, but that's OK,
907 * we'll just truncate it here. Note that we use snprintf() rather than
908 * xsprintf() on purpose here: we are fine with truncation and don't
909 * consider that an error. */
910 DISABLE_WARNING_FORMAT_NONLITERAL
;
911 (void) snprintf(buf
, sizeof(buf
), format
, unit_description(u
));
917 if (result
== JOB_DONE
)
918 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_STARTED_STR
;
920 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_FAILED_STR
;
924 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_RELOADED_STR
;
929 mid
= "MESSAGE_ID=" SD_MESSAGE_UNIT_STOPPED_STR
;
933 log_struct(job_result_log_level
[result
],
934 LOG_MESSAGE("%s", buf
),
935 "JOB_ID=%" PRIu32
, job_id
,
936 "JOB_TYPE=%s", job_type_to_string(t
),
937 "JOB_RESULT=%s", job_result_to_string(result
),
939 LOG_UNIT_INVOCATION_ID(u
));
943 log_struct(job_result_log_level
[result
],
944 LOG_MESSAGE("%s", buf
),
945 "JOB_ID=%" PRIu32
, job_id
,
946 "JOB_TYPE=%s", job_type_to_string(t
),
947 "JOB_RESULT=%s", job_result_to_string(result
),
949 LOG_UNIT_INVOCATION_ID(u
),
953 static void job_emit_done_status_message(Unit
*u
, uint32_t job_id
, JobType t
, JobResult result
) {
956 job_log_done_status_message(u
, job_id
, t
, result
);
957 job_print_done_status_message(u
, t
, result
);
960 static void job_fail_dependencies(Unit
*u
, UnitDependency d
) {
967 HASHMAP_FOREACH_KEY(v
, other
, u
->dependencies
[d
], i
) {
972 if (!IN_SET(j
->type
, JOB_START
, JOB_VERIFY_ACTIVE
))
975 job_finish_and_invalidate(j
, JOB_DEPENDENCY
, true, false);
979 int job_finish_and_invalidate(Job
*j
, JobResult result
, bool recursive
, bool already
) {
987 assert(j
->installed
);
988 assert(j
->type
< _JOB_TYPE_MAX_IN_TRANSACTION
);
995 log_unit_debug(u
, "Job %" PRIu32
" %s/%s finished, result=%s", j
->id
, u
->id
, job_type_to_string(t
), job_result_to_string(result
));
997 /* If this job did nothing to respective unit we don't log the status message */
999 job_emit_done_status_message(u
, j
->id
, t
, result
);
1001 /* Patch restart jobs so that they become normal start jobs */
1002 if (result
== JOB_DONE
&& t
== JOB_RESTART
) {
1004 job_change_type(j
, JOB_START
);
1005 job_set_state(j
, JOB_WAITING
);
1007 job_add_to_dbus_queue(j
);
1008 job_add_to_run_queue(j
);
1009 job_add_to_gc_queue(j
);
1014 if (IN_SET(result
, JOB_FAILED
, JOB_INVALID
))
1015 j
->manager
->n_failed_jobs
++;
1020 /* Fail depending jobs on failure */
1021 if (result
!= JOB_DONE
&& recursive
) {
1022 if (IN_SET(t
, JOB_START
, JOB_VERIFY_ACTIVE
)) {
1023 job_fail_dependencies(u
, UNIT_REQUIRED_BY
);
1024 job_fail_dependencies(u
, UNIT_REQUISITE_OF
);
1025 job_fail_dependencies(u
, UNIT_BOUND_BY
);
1026 } else if (t
== JOB_STOP
)
1027 job_fail_dependencies(u
, UNIT_CONFLICTED_BY
);
1030 /* Trigger OnFailure dependencies that are not generated by
1031 * the unit itself. We don't treat JOB_CANCELED as failure in
1032 * this context. And JOB_FAILURE is already handled by the
1034 if (IN_SET(result
, JOB_TIMEOUT
, JOB_DEPENDENCY
)) {
1035 log_struct(LOG_NOTICE
,
1036 "JOB_TYPE=%s", job_type_to_string(t
),
1037 "JOB_RESULT=%s", job_result_to_string(result
),
1039 LOG_UNIT_MESSAGE(u
, "Job %s/%s failed with result '%s'.",
1041 job_type_to_string(t
),
1042 job_result_to_string(result
)));
1044 unit_start_on_failure(u
);
1047 unit_trigger_notify(u
);
1050 /* Try to start the next jobs that can be started */
1051 HASHMAP_FOREACH_KEY(v
, other
, u
->dependencies
[UNIT_AFTER
], i
)
1053 job_add_to_run_queue(other
->job
);
1054 job_add_to_gc_queue(other
->job
);
1056 HASHMAP_FOREACH_KEY(v
, other
, u
->dependencies
[UNIT_BEFORE
], i
)
1058 job_add_to_run_queue(other
->job
);
1059 job_add_to_gc_queue(other
->job
);
1062 manager_check_finished(u
->manager
);
1067 static int job_dispatch_timer(sd_event_source
*s
, uint64_t monotonic
, void *userdata
) {
1072 assert(s
== j
->timer_event_source
);
1074 log_unit_warning(j
->unit
, "Job %s/%s timed out.", j
->unit
->id
, job_type_to_string(j
->type
));
1077 job_finish_and_invalidate(j
, JOB_TIMEOUT
, true, false);
1079 emergency_action(u
->manager
, u
->job_timeout_action
,
1080 EMERGENCY_ACTION_IS_WATCHDOG
|EMERGENCY_ACTION_WARN
,
1081 u
->job_timeout_reboot_arg
, -1, "job timed out");
1086 int job_start_timer(Job
*j
, bool job_running
) {
1088 usec_t timeout_time
, old_timeout_time
;
1091 j
->begin_running_usec
= now(CLOCK_MONOTONIC
);
1093 if (j
->unit
->job_running_timeout
== USEC_INFINITY
)
1096 timeout_time
= usec_add(j
->begin_running_usec
, j
->unit
->job_running_timeout
);
1098 if (j
->timer_event_source
) {
1099 /* Update only if JobRunningTimeoutSec= results in earlier timeout */
1100 r
= sd_event_source_get_time(j
->timer_event_source
, &old_timeout_time
);
1104 if (old_timeout_time
<= timeout_time
)
1107 return sd_event_source_set_time(j
->timer_event_source
, timeout_time
);
1110 if (j
->timer_event_source
)
1113 j
->begin_usec
= now(CLOCK_MONOTONIC
);
1115 if (j
->unit
->job_timeout
== USEC_INFINITY
)
1118 timeout_time
= usec_add(j
->begin_usec
, j
->unit
->job_timeout
);
1121 r
= sd_event_add_time(
1123 &j
->timer_event_source
,
1126 job_dispatch_timer
, j
);
1130 (void) sd_event_source_set_description(j
->timer_event_source
, "job-start");
1135 void job_add_to_run_queue(Job
*j
) {
1139 assert(j
->installed
);
1141 if (j
->in_run_queue
)
1144 if (!j
->manager
->run_queue
) {
1145 r
= sd_event_source_set_enabled(j
->manager
->run_queue_event_source
, SD_EVENT_ONESHOT
);
1147 log_warning_errno(r
, "Failed to enable job run queue event source, ignoring: %m");
1150 LIST_PREPEND(run_queue
, j
->manager
->run_queue
, j
);
1151 j
->in_run_queue
= true;
1154 void job_add_to_dbus_queue(Job
*j
) {
1156 assert(j
->installed
);
1158 if (j
->in_dbus_queue
)
1161 /* We don't check if anybody is subscribed here, since this
1162 * job might just have been created and not yet assigned to a
1163 * connection/client. */
1165 LIST_PREPEND(dbus_queue
, j
->manager
->dbus_job_queue
, j
);
1166 j
->in_dbus_queue
= true;
1169 char *job_dbus_path(Job
*j
) {
1174 if (asprintf(&p
, "/org/freedesktop/systemd1/job/%"PRIu32
, j
->id
) < 0)
1180 int job_serialize(Job
*j
, FILE *f
) {
1184 (void) serialize_item_format(f
, "job-id", "%u", j
->id
);
1185 (void) serialize_item(f
, "job-type", job_type_to_string(j
->type
));
1186 (void) serialize_item(f
, "job-state", job_state_to_string(j
->state
));
1187 (void) serialize_bool(f
, "job-irreversible", j
->irreversible
);
1188 (void) serialize_bool(f
, "job-sent-dbus-new-signal", j
->sent_dbus_new_signal
);
1189 (void) serialize_bool(f
, "job-ignore-order", j
->ignore_order
);
1191 if (j
->begin_usec
> 0)
1192 (void) serialize_usec(f
, "job-begin", j
->begin_usec
);
1193 if (j
->begin_running_usec
> 0)
1194 (void) serialize_usec(f
, "job-begin-running", j
->begin_running_usec
);
1196 bus_track_serialize(j
->bus_track
, f
, "subscribed");
1203 int job_deserialize(Job
*j
, FILE *f
) {
1210 _cleanup_free_
char *line
= NULL
;
1214 r
= read_line(f
, LONG_LINE_MAX
, &line
);
1216 return log_error_errno(r
, "Failed to read serialization line: %m");
1226 k
= strcspn(l
, "=");
1234 if (streq(l
, "job-id")) {
1236 if (safe_atou32(v
, &j
->id
) < 0)
1237 log_debug("Failed to parse job id value: %s", v
);
1239 } else if (streq(l
, "job-type")) {
1242 t
= job_type_from_string(v
);
1244 log_debug("Failed to parse job type: %s", v
);
1245 else if (t
>= _JOB_TYPE_MAX_IN_TRANSACTION
)
1246 log_debug("Cannot deserialize job of type: %s", v
);
1250 } else if (streq(l
, "job-state")) {
1253 s
= job_state_from_string(v
);
1255 log_debug("Failed to parse job state: %s", v
);
1257 job_set_state(j
, s
);
1259 } else if (streq(l
, "job-irreversible")) {
1262 b
= parse_boolean(v
);
1264 log_debug("Failed to parse job irreversible flag: %s", v
);
1266 j
->irreversible
= j
->irreversible
|| b
;
1268 } else if (streq(l
, "job-sent-dbus-new-signal")) {
1271 b
= parse_boolean(v
);
1273 log_debug("Failed to parse job sent_dbus_new_signal flag: %s", v
);
1275 j
->sent_dbus_new_signal
= j
->sent_dbus_new_signal
|| b
;
1277 } else if (streq(l
, "job-ignore-order")) {
1280 b
= parse_boolean(v
);
1282 log_debug("Failed to parse job ignore_order flag: %s", v
);
1284 j
->ignore_order
= j
->ignore_order
|| b
;
1286 } else if (streq(l
, "job-begin"))
1287 (void) deserialize_usec(v
, &j
->begin_usec
);
1289 else if (streq(l
, "job-begin-running"))
1290 (void) deserialize_usec(v
, &j
->begin_running_usec
);
1292 else if (streq(l
, "subscribed")) {
1293 if (strv_extend(&j
->deserialized_clients
, v
) < 0)
1296 log_debug("Unknown job serialization key: %s", l
);
1300 int job_coldplug(Job
*j
) {
1302 usec_t timeout_time
= USEC_INFINITY
;
1306 /* After deserialization is complete and the bus connection
1307 * set up again, let's start watching our subscribers again */
1308 (void) bus_job_coldplug_bus_track(j
);
1310 if (j
->state
== JOB_WAITING
)
1311 job_add_to_run_queue(j
);
1313 /* Maybe due to new dependencies we don't actually need this job anymore? */
1314 job_add_to_gc_queue(j
);
1316 /* Create timer only when job began or began running and the respective timeout is finite.
1317 * Follow logic of job_start_timer() if both timeouts are finite */
1318 if (j
->begin_usec
== 0)
1321 if (j
->unit
->job_timeout
!= USEC_INFINITY
)
1322 timeout_time
= usec_add(j
->begin_usec
, j
->unit
->job_timeout
);
1324 if (j
->begin_running_usec
> 0 && j
->unit
->job_running_timeout
!= USEC_INFINITY
)
1325 timeout_time
= MIN(timeout_time
, usec_add(j
->begin_running_usec
, j
->unit
->job_running_timeout
));
1327 if (timeout_time
== USEC_INFINITY
)
1330 j
->timer_event_source
= sd_event_source_unref(j
->timer_event_source
);
1332 r
= sd_event_add_time(
1334 &j
->timer_event_source
,
1337 job_dispatch_timer
, j
);
1339 log_debug_errno(r
, "Failed to restart timeout for job: %m");
1341 (void) sd_event_source_set_description(j
->timer_event_source
, "job-timeout");
1346 void job_shutdown_magic(Job
*j
) {
1349 /* The shutdown target gets some special treatment here: we
1350 * tell the kernel to begin with flushing its disk caches, to
1351 * optimize shutdown time a bit. Ideally we wouldn't hardcode
1352 * this magic into PID 1. However all other processes aren't
1353 * options either since they'd exit much sooner than PID 1 and
1354 * asynchronous sync() would cause their exit to be
1357 if (j
->type
!= JOB_START
)
1360 if (!MANAGER_IS_SYSTEM(j
->unit
->manager
))
1363 if (!unit_has_name(j
->unit
, SPECIAL_SHUTDOWN_TARGET
))
1366 /* In case messages on console has been disabled on boot */
1367 j
->unit
->manager
->no_console_output
= false;
1369 if (detect_container() > 0)
1372 (void) asynchronous_sync(NULL
);
1375 int job_get_timeout(Job
*j
, usec_t
*timeout
) {
1376 usec_t x
= USEC_INFINITY
, y
= USEC_INFINITY
;
1382 if (j
->timer_event_source
) {
1383 r
= sd_event_source_get_time(j
->timer_event_source
, &x
);
1388 if (UNIT_VTABLE(u
)->get_timeout
) {
1389 r
= UNIT_VTABLE(u
)->get_timeout(u
, &y
);
1394 if (x
== USEC_INFINITY
&& y
== USEC_INFINITY
)
1397 *timeout
= MIN(x
, y
);
1401 bool job_may_gc(Job
*j
) {
1408 /* Checks whether this job should be GC'ed away. We only do this for jobs of units that have no effect on their
1409 * own and just track external state. For now the only unit type that qualifies for this are .device units.
1410 * Returns true if the job can be collected. */
1412 if (!UNIT_VTABLE(j
->unit
)->gc_jobs
)
1415 if (sd_bus_track_count(j
->bus_track
) > 0)
1418 /* FIXME: So this is a bit ugly: for now we don't properly track references made via private bus connections
1419 * (because it's nasty, as sd_bus_track doesn't apply to it). We simply remember that the job was once
1420 * referenced by one, and reset this whenever we notice that no private bus connections are around. This means
1421 * the GC is a bit too conservative when it comes to jobs created by private bus connections. */
1422 if (j
->ref_by_private_bus
) {
1423 if (set_isempty(j
->unit
->manager
->private_buses
))
1424 j
->ref_by_private_bus
= false;
1429 if (j
->type
== JOB_NOP
)
1432 /* If a job is ordered after ours, and is to be started, then it needs to wait for us, regardless if we stop or
1433 * start, hence let's not GC in that case. */
1434 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
) {
1438 if (other
->job
->ignore_order
)
1441 if (IN_SET(other
->job
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
))
1445 /* If we are going down, but something else is ordered After= us, then it needs to wait for us */
1446 if (IN_SET(j
->type
, JOB_STOP
, JOB_RESTART
))
1447 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
) {
1451 if (other
->job
->ignore_order
)
1457 /* The logic above is kinda the inverse of the job_is_runnable() logic. Specifically, if the job "we" is
1458 * ordered before the job "other":
1460 * we start + other start → stay
1461 * we start + other stop → gc
1462 * we stop + other start → stay
1463 * we stop + other stop → gc
1465 * "we" are ordered after "other":
1467 * we start + other start → gc
1468 * we start + other stop → gc
1469 * we stop + other start → stay
1470 * we stop + other stop → stay
1476 void job_add_to_gc_queue(Job
*j
) {
1485 LIST_PREPEND(gc_queue
, j
->unit
->manager
->gc_job_queue
, j
);
1486 j
->in_gc_queue
= true;
1489 static int job_compare(Job
* const *a
, Job
* const *b
) {
1490 return CMP((*a
)->id
, (*b
)->id
);
1493 static size_t sort_job_list(Job
**list
, size_t n
) {
1494 Job
*previous
= NULL
;
1497 /* Order by numeric IDs */
1498 typesafe_qsort(list
, n
, job_compare
);
1500 /* Filter out duplicates */
1501 for (a
= 0, b
= 0; a
< n
; a
++) {
1503 if (previous
== list
[a
])
1506 previous
= list
[b
++] = list
[a
];
1512 int job_get_before(Job
*j
, Job
*** ret
) {
1513 _cleanup_free_ Job
** list
= NULL
;
1514 size_t n
= 0, n_allocated
= 0;
1519 /* Returns a list of all pending jobs that need to finish before this job may be started. */
1524 if (j
->ignore_order
) {
1529 if (IN_SET(j
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
)) {
1531 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
) {
1535 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1537 list
[n
++] = other
->job
;
1541 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
) {
1545 if (!IN_SET(other
->job
->type
, JOB_STOP
, JOB_RESTART
))
1548 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1550 list
[n
++] = other
->job
;
1553 n
= sort_job_list(list
, n
);
1555 *ret
= TAKE_PTR(list
);
1560 int job_get_after(Job
*j
, Job
*** ret
) {
1561 _cleanup_free_ Job
** list
= NULL
;
1562 size_t n
= 0, n_allocated
= 0;
1570 /* Returns a list of all pending jobs that are waiting for this job to finish. */
1572 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_BEFORE
], i
) {
1576 if (other
->job
->ignore_order
)
1579 if (!IN_SET(other
->job
->type
, JOB_START
, JOB_VERIFY_ACTIVE
, JOB_RELOAD
))
1582 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1584 list
[n
++] = other
->job
;
1587 if (IN_SET(j
->type
, JOB_STOP
, JOB_RESTART
)) {
1589 HASHMAP_FOREACH_KEY(v
, other
, j
->unit
->dependencies
[UNIT_AFTER
], i
) {
1593 if (other
->job
->ignore_order
)
1596 if (!GREEDY_REALLOC(list
, n_allocated
, n
+1))
1598 list
[n
++] = other
->job
;
1602 n
= sort_job_list(list
, n
);
1604 *ret
= TAKE_PTR(list
);
1609 static const char* const job_state_table
[_JOB_STATE_MAX
] = {
1610 [JOB_WAITING
] = "waiting",
1611 [JOB_RUNNING
] = "running",
1614 DEFINE_STRING_TABLE_LOOKUP(job_state
, JobState
);
1616 static const char* const job_type_table
[_JOB_TYPE_MAX
] = {
1617 [JOB_START
] = "start",
1618 [JOB_VERIFY_ACTIVE
] = "verify-active",
1619 [JOB_STOP
] = "stop",
1620 [JOB_RELOAD
] = "reload",
1621 [JOB_RELOAD_OR_START
] = "reload-or-start",
1622 [JOB_RESTART
] = "restart",
1623 [JOB_TRY_RESTART
] = "try-restart",
1624 [JOB_TRY_RELOAD
] = "try-reload",
1628 DEFINE_STRING_TABLE_LOOKUP(job_type
, JobType
);
1630 static const char* const job_mode_table
[_JOB_MODE_MAX
] = {
1631 [JOB_FAIL
] = "fail",
1632 [JOB_REPLACE
] = "replace",
1633 [JOB_REPLACE_IRREVERSIBLY
] = "replace-irreversibly",
1634 [JOB_ISOLATE
] = "isolate",
1635 [JOB_FLUSH
] = "flush",
1636 [JOB_IGNORE_DEPENDENCIES
] = "ignore-dependencies",
1637 [JOB_IGNORE_REQUIREMENTS
] = "ignore-requirements",
1640 DEFINE_STRING_TABLE_LOOKUP(job_mode
, JobMode
);
1642 static const char* const job_result_table
[_JOB_RESULT_MAX
] = {
1643 [JOB_DONE
] = "done",
1644 [JOB_CANCELED
] = "canceled",
1645 [JOB_TIMEOUT
] = "timeout",
1646 [JOB_FAILED
] = "failed",
1647 [JOB_DEPENDENCY
] = "dependency",
1648 [JOB_SKIPPED
] = "skipped",
1649 [JOB_INVALID
] = "invalid",
1650 [JOB_ASSERT
] = "assert",
1651 [JOB_UNSUPPORTED
] = "unsupported",
1652 [JOB_COLLECTED
] = "collected",
1653 [JOB_ONCE
] = "once",
1656 DEFINE_STRING_TABLE_LOOKUP(job_result
, JobResult
);
1658 const char* job_type_to_access_method(JobType t
) {
1660 assert(t
< _JOB_TYPE_MAX
);
1662 if (IN_SET(t
, JOB_START
, JOB_RESTART
, JOB_TRY_RESTART
))
1664 else if (t
== JOB_STOP
)