2 * Background jobs (long-running operations)
4 * Copyright (c) 2011 IBM Corp.
5 * Copyright (c) 2012, 2018 Red Hat, Inc.
7 * Permission is hereby granted, free of charge, to any person obtaining a copy
8 * of this software and associated documentation files (the "Software"), to deal
9 * in the Software without restriction, including without limitation the rights
10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11 * copies of the Software, and to permit persons to whom the Software is
12 * furnished to do so, subject to the following conditions:
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
26 #include "qemu/osdep.h"
27 #include "qemu-common.h"
28 #include "qapi/error.h"
31 #include "qemu/main-loop.h"
32 #include "trace-root.h"
34 static QLIST_HEAD(, Job) jobs = QLIST_HEAD_INITIALIZER(jobs);
36 /* Job State Transition Table */
37 bool JobSTT[JOB_STATUS__MAX][JOB_STATUS__MAX] = {
38 /* U, C, R, P, Y, S, W, D, X, E, N */
39 /* U: */ [JOB_STATUS_UNDEFINED] = {0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0},
40 /* C: */ [JOB_STATUS_CREATED] = {0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1},
41 /* R: */ [JOB_STATUS_RUNNING] = {0, 0, 0, 1, 1, 0, 1, 0, 1, 0, 0},
42 /* P: */ [JOB_STATUS_PAUSED] = {0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0},
43 /* Y: */ [JOB_STATUS_READY] = {0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0},
44 /* S: */ [JOB_STATUS_STANDBY] = {0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0},
45 /* W: */ [JOB_STATUS_WAITING] = {0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0},
46 /* D: */ [JOB_STATUS_PENDING] = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0},
47 /* X: */ [JOB_STATUS_ABORTING] = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0},
48 /* E: */ [JOB_STATUS_CONCLUDED] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1},
49 /* N: */ [JOB_STATUS_NULL] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
52 bool JobVerbTable[JOB_VERB__MAX][JOB_STATUS__MAX] = {
53 /* U, C, R, P, Y, S, W, D, X, E, N */
54 [JOB_VERB_CANCEL] = {0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0},
55 [JOB_VERB_PAUSE] = {0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0},
56 [JOB_VERB_RESUME] = {0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0},
57 [JOB_VERB_SET_SPEED] = {0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0},
58 [JOB_VERB_COMPLETE] = {0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0},
59 [JOB_VERB_FINALIZE] = {0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0},
60 [JOB_VERB_DISMISS] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0},
63 /* Right now, this mutex is only needed to synchronize accesses to job->busy
64 * and job->sleep_timer, such as concurrent calls to job_do_yield and
66 static QemuMutex job_mutex;
68 static void job_lock(void)
70 qemu_mutex_lock(&job_mutex);
73 static void job_unlock(void)
75 qemu_mutex_unlock(&job_mutex);
78 static void __attribute__((__constructor__)) job_init(void)
80 qemu_mutex_init(&job_mutex);
83 /* TODO Make static once the whole state machine is in job.c */
84 void job_state_transition(Job *job, JobStatus s1)
86 JobStatus s0 = job->status;
87 assert(s1 >= 0 && s1 <= JOB_STATUS__MAX);
88 trace_job_state_transition(job, job->ret,
89 JobSTT[s0][s1] ? "allowed" : "disallowed",
90 JobStatus_str(s0), JobStatus_str(s1));
91 assert(JobSTT[s0][s1]);
95 int job_apply_verb(Job *job, JobVerb verb, Error **errp)
97 JobStatus s0 = job->status;
98 assert(verb >= 0 && verb <= JOB_VERB__MAX);
99 trace_job_apply_verb(job, JobStatus_str(s0), JobVerb_str(verb),
100 JobVerbTable[verb][s0] ? "allowed" : "prohibited");
101 if (JobVerbTable[verb][s0]) {
104 error_setg(errp, "Job '%s' in state '%s' cannot accept command verb '%s'",
105 job->id, JobStatus_str(s0), JobVerb_str(verb));
109 JobType job_type(const Job *job)
111 return job->driver->job_type;
114 const char *job_type_str(const Job *job)
116 return JobType_str(job_type(job));
119 bool job_is_cancelled(Job *job)
121 return job->cancelled;
124 bool job_is_completed(Job *job)
126 switch (job->status) {
127 case JOB_STATUS_UNDEFINED:
128 case JOB_STATUS_CREATED:
129 case JOB_STATUS_RUNNING:
130 case JOB_STATUS_PAUSED:
131 case JOB_STATUS_READY:
132 case JOB_STATUS_STANDBY:
134 case JOB_STATUS_WAITING:
135 case JOB_STATUS_PENDING:
136 case JOB_STATUS_ABORTING:
137 case JOB_STATUS_CONCLUDED:
138 case JOB_STATUS_NULL:
141 g_assert_not_reached();
146 bool job_started(Job *job)
151 bool job_should_pause(Job *job)
153 return job->pause_count > 0;
156 Job *job_next(Job *job)
159 return QLIST_FIRST(&jobs);
161 return QLIST_NEXT(job, job_list);
164 Job *job_get(const char *id)
168 QLIST_FOREACH(job, &jobs, job_list) {
169 if (job->id && !strcmp(id, job->id)) {
177 static void job_sleep_timer_cb(void *opaque)
184 void *job_create(const char *job_id, const JobDriver *driver, AioContext *ctx,
185 int flags, BlockCompletionFunc *cb, void *opaque, Error **errp)
190 if (flags & JOB_INTERNAL) {
191 error_setg(errp, "Cannot specify job ID for internal job");
194 if (!id_wellformed(job_id)) {
195 error_setg(errp, "Invalid job ID '%s'", job_id);
198 if (job_get(job_id)) {
199 error_setg(errp, "Job ID '%s' already in use", job_id);
202 } else if (!(flags & JOB_INTERNAL)) {
203 error_setg(errp, "An explicit job ID is required");
207 job = g_malloc0(driver->instance_size);
208 job->driver = driver;
209 job->id = g_strdup(job_id);
211 job->aio_context = ctx;
214 job->pause_count = 1;
215 job->auto_finalize = !(flags & JOB_MANUAL_FINALIZE);
216 job->auto_dismiss = !(flags & JOB_MANUAL_DISMISS);
218 job->opaque = opaque;
220 notifier_list_init(&job->on_finalize_cancelled);
221 notifier_list_init(&job->on_finalize_completed);
222 notifier_list_init(&job->on_pending);
224 job_state_transition(job, JOB_STATUS_CREATED);
225 aio_timer_init(qemu_get_aio_context(), &job->sleep_timer,
226 QEMU_CLOCK_REALTIME, SCALE_NS,
227 job_sleep_timer_cb, job);
229 QLIST_INSERT_HEAD(&jobs, job, job_list);
234 void job_ref(Job *job)
239 void job_unref(Job *job)
241 if (--job->refcnt == 0) {
242 assert(job->status == JOB_STATUS_NULL);
243 assert(!timer_pending(&job->sleep_timer));
245 if (job->driver->free) {
246 job->driver->free(job);
249 QLIST_REMOVE(job, job_list);
256 void job_event_cancelled(Job *job)
258 notifier_list_notify(&job->on_finalize_cancelled, job);
261 void job_event_completed(Job *job)
263 notifier_list_notify(&job->on_finalize_completed, job);
266 void job_event_pending(Job *job)
268 notifier_list_notify(&job->on_pending, job);
271 void job_enter_cond(Job *job, bool(*fn)(Job *job))
273 if (!job_started(job)) {
276 if (job->deferred_to_main_loop) {
286 if (fn && !fn(job)) {
291 assert(!job->deferred_to_main_loop);
292 timer_del(&job->sleep_timer);
295 aio_co_wake(job->co);
298 void job_enter(Job *job)
300 job_enter_cond(job, NULL);
303 /* Yield, and schedule a timer to reenter the coroutine after @ns nanoseconds.
304 * Reentering the job coroutine with block_job_enter() before the timer has
305 * expired is allowed and cancels the timer.
307 * If @ns is (uint64_t) -1, no timer is scheduled and block_job_enter() must be
308 * called explicitly. */
309 void coroutine_fn job_do_yield(Job *job, uint64_t ns)
313 timer_mod(&job->sleep_timer, ns);
317 qemu_coroutine_yield();
319 /* Set by job_enter_cond() before re-entering the coroutine. */
323 void coroutine_fn job_pause_point(Job *job)
325 assert(job && job_started(job));
327 if (!job_should_pause(job)) {
330 if (job_is_cancelled(job)) {
334 if (job->driver->pause) {
335 job->driver->pause(job);
338 if (job_should_pause(job) && !job_is_cancelled(job)) {
339 JobStatus status = job->status;
340 job_state_transition(job, status == JOB_STATUS_READY
342 : JOB_STATUS_PAUSED);
344 job_do_yield(job, -1);
346 job_state_transition(job, status);
349 if (job->driver->resume) {
350 job->driver->resume(job);
354 void coroutine_fn job_sleep_ns(Job *job, int64_t ns)
358 /* Check cancellation *before* setting busy = false, too! */
359 if (job_is_cancelled(job)) {
363 if (!job_should_pause(job)) {
364 job_do_yield(job, qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + ns);
367 job_pause_point(job);
370 void job_drain(Job *job)
372 /* If job is !busy this kicks it into the next pause point. */
375 if (job->driver->drain) {
376 job->driver->drain(job);
382 * All jobs must allow a pause point before entering their job proper. This
383 * ensures that jobs can be paused prior to being started, then resumed later.
385 static void coroutine_fn job_co_entry(void *opaque)
389 assert(job && job->driver && job->driver->start);
390 job_pause_point(job);
391 job->driver->start(job);
395 void job_start(Job *job)
397 assert(job && !job_started(job) && job->paused &&
398 job->driver && job->driver->start);
399 job->co = qemu_coroutine_create(job_co_entry, job);
403 job_state_transition(job, JOB_STATUS_RUNNING);
404 aio_co_enter(job->aio_context, job->co);
407 /* Assumes the block_job_mutex is held */
408 static bool job_timer_not_pending(Job *job)
410 return !timer_pending(&job->sleep_timer);
413 void job_pause(Job *job)
418 void job_resume(Job *job)
420 assert(job->pause_count > 0);
422 if (job->pause_count) {
426 /* kick only if no timer is pending */
427 job_enter_cond(job, job_timer_not_pending);
430 void job_user_pause(Job *job, Error **errp)
432 if (job_apply_verb(job, JOB_VERB_PAUSE, errp)) {
435 if (job->user_paused) {
436 error_setg(errp, "Job is already paused");
439 job->user_paused = true;
443 bool job_user_paused(Job *job)
445 return job->user_paused;
448 void job_user_resume(Job *job, Error **errp)
451 if (!job->user_paused || job->pause_count <= 0) {
452 error_setg(errp, "Can't resume a job that was not paused");
455 if (job_apply_verb(job, JOB_VERB_RESUME, errp)) {
458 if (job->driver->user_resume) {
459 job->driver->user_resume(job);
461 job->user_paused = false;
465 void job_do_dismiss(Job *job)
470 job->deferred_to_main_loop = true;
472 /* TODO Don't assume it's a BlockJob */
473 block_job_txn_del_job((BlockJob*) job);
475 job_state_transition(job, JOB_STATUS_NULL);
479 void job_early_fail(Job *job)
481 assert(job->status == JOB_STATUS_CREATED);
485 static void job_conclude(Job *job)
487 job_state_transition(job, JOB_STATUS_CONCLUDED);
488 if (job->auto_dismiss || !job_started(job)) {
493 void job_update_rc(Job *job)
495 if (!job->ret && job_is_cancelled(job)) {
496 job->ret = -ECANCELED;
499 job_state_transition(job, JOB_STATUS_ABORTING);
503 static void job_commit(Job *job)
506 if (job->driver->commit) {
507 job->driver->commit(job);
511 static void job_abort(Job *job)
514 if (job->driver->abort) {
515 job->driver->abort(job);
519 static void job_clean(Job *job)
521 if (job->driver->clean) {
522 job->driver->clean(job);
526 int job_finalize_single(Job *job)
528 assert(job_is_completed(job));
530 /* Ensure abort is called for late-transactional failures */
541 job->cb(job->opaque, job->ret);
544 /* Emit events only if we actually started */
545 if (job_started(job)) {
546 if (job_is_cancelled(job)) {
547 job_event_cancelled(job);
549 job_event_completed(job);
553 /* TODO Don't assume it's a BlockJob */
554 block_job_txn_del_job((BlockJob*) job);
559 void job_complete(Job *job, Error **errp)
561 /* Should not be reachable via external interface for internal jobs */
563 if (job_apply_verb(job, JOB_VERB_COMPLETE, errp)) {
566 if (job->pause_count || job_is_cancelled(job) || !job->driver->complete) {
567 error_setg(errp, "The active block job '%s' cannot be completed",
572 job->driver->complete(job, errp);
578 JobDeferToMainLoopFn *fn;
580 } JobDeferToMainLoopData;
582 static void job_defer_to_main_loop_bh(void *opaque)
584 JobDeferToMainLoopData *data = opaque;
585 Job *job = data->job;
586 AioContext *aio_context = job->aio_context;
588 aio_context_acquire(aio_context);
589 data->fn(data->job, data->opaque);
590 aio_context_release(aio_context);
595 void job_defer_to_main_loop(Job *job, JobDeferToMainLoopFn *fn, void *opaque)
597 JobDeferToMainLoopData *data = g_malloc(sizeof(*data));
600 data->opaque = opaque;
601 job->deferred_to_main_loop = true;
603 aio_bh_schedule_oneshot(qemu_get_aio_context(),
604 job_defer_to_main_loop_bh, data);
607 int job_finish_sync(Job *job, void (*finish)(Job *, Error **errp), Error **errp)
609 Error *local_err = NULL;
615 finish(job, &local_err);
618 error_propagate(errp, local_err);
622 /* job_drain calls job_enter, and it should be enough to induce progress
623 * until the job completes or moves to the main thread. */
624 while (!job->deferred_to_main_loop && !job_is_completed(job)) {
627 while (!job_is_completed(job)) {
628 aio_poll(qemu_get_aio_context(), true);
630 ret = (job_is_cancelled(job) && job->ret == 0) ? -ECANCELED : job->ret;