job.c

   1 /*
   2  * Background jobs (long-running operations)
   3  *
   4  * Copyright (c) 2011 IBM Corp.
   5  * Copyright (c) 2012, 2018 Red Hat, Inc.
   6  *
   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:
  13  *
  14  * The above copyright notice and this permission notice shall be included in
  15  * all copies or substantial portions of the Software.
  16  *
  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
  23  * THE SOFTWARE.
  24  */
  25
  26 #include "qemu/osdep.h"
  27 #include "qapi/error.h"
  28 #include "qemu/job.h"
  29 #include "qemu/id.h"
  30 #include "qemu/main-loop.h"
  31 #include "block/aio-wait.h"
  32 #include "trace/trace-root.h"
  33 #include "qapi/qapi-events-job.h"
  34
  35 /*
  36  * The job API is composed of two categories of functions.
  37  *
  38  * The first includes functions used by the monitor.  The monitor is
  39  * peculiar in that it accesses the job list with job_get, and
  40  * therefore needs consistency across job_get and the actual operation
  41  * (e.g. job_user_cancel). To achieve this consistency, the caller
  42  * calls job_lock/job_unlock itself around the whole operation.
  43  *
  44  *
  45  * The second includes functions used by the job drivers and sometimes
  46  * by the core block layer. These delegate the locking to the callee instead.
  47  */
  48
  49 /*
  50  * job_mutex protects the jobs list, but also makes the
  51  * struct job fields thread-safe.
  52  */
  53 QemuMutex job_mutex;
  54
  55 /* Protected by job_mutex */
  56 static QLIST_HEAD(, Job) jobs = QLIST_HEAD_INITIALIZER(jobs);
  57
  58 /* Job State Transition Table */
  59 bool JobSTT[JOB_STATUS__MAX][JOB_STATUS__MAX] = {
  60                                     /* U, C, R, P, Y, S, W, D, X, E, N */
  61     /* U: */ [JOB_STATUS_UNDEFINED] = {0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0},
  62     /* C: */ [JOB_STATUS_CREATED]   = {0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1},
  63     /* R: */ [JOB_STATUS_RUNNING]   = {0, 0, 0, 1, 1, 0, 1, 0, 1, 0, 0},
  64     /* P: */ [JOB_STATUS_PAUSED]    = {0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0},
  65     /* Y: */ [JOB_STATUS_READY]     = {0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0},
  66     /* S: */ [JOB_STATUS_STANDBY]   = {0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0},
  67     /* W: */ [JOB_STATUS_WAITING]   = {0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0},
  68     /* D: */ [JOB_STATUS_PENDING]   = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0},
  69     /* X: */ [JOB_STATUS_ABORTING]  = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0},
  70     /* E: */ [JOB_STATUS_CONCLUDED] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1},
  71     /* N: */ [JOB_STATUS_NULL]      = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
  72 };
  73
  74 bool JobVerbTable[JOB_VERB__MAX][JOB_STATUS__MAX] = {
  75                                     /* U, C, R, P, Y, S, W, D, X, E, N */
  76     [JOB_VERB_CANCEL]               = {0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0},
  77     [JOB_VERB_PAUSE]                = {0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0},
  78     [JOB_VERB_RESUME]               = {0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0},
  79     [JOB_VERB_SET_SPEED]            = {0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0},
  80     [JOB_VERB_COMPLETE]             = {0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0},
  81     [JOB_VERB_FINALIZE]             = {0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0},
  82     [JOB_VERB_DISMISS]              = {0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0},
  83     [JOB_VERB_CHANGE]               = {0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0},
  84 };
  85
  86 /* Transactional group of jobs */
  87 struct JobTxn {
  88
  89     /* Is this txn being cancelled? */
  90     bool aborting;
  91
  92     /* List of jobs */
  93     QLIST_HEAD(, Job) jobs;
  94
  95     /* Reference count */
  96     int refcnt;
  97 };
  98
  99 void job_lock(void)
 100 {
 101     qemu_mutex_lock(&job_mutex);
 102 }
 103
 104 void job_unlock(void)
 105 {
 106     qemu_mutex_unlock(&job_mutex);
 107 }
 108
 109 static void __attribute__((__constructor__)) job_init(void)
 110 {
 111     qemu_mutex_init(&job_mutex);
 112 }
 113
 114 JobTxn *job_txn_new(void)
 115 {
 116     JobTxn *txn = g_new0(JobTxn, 1);
 117     QLIST_INIT(&txn->jobs);
 118     txn->refcnt = 1;
 119     return txn;
 120 }
 121
 122 /* Called with job_mutex held. */
 123 static void job_txn_ref_locked(JobTxn *txn)
 124 {
 125     txn->refcnt++;
 126 }
 127
 128 void job_txn_unref_locked(JobTxn *txn)
 129 {
 130     if (txn && --txn->refcnt == 0) {
 131         g_free(txn);
 132     }
 133 }
 134
 135 void job_txn_unref(JobTxn *txn)
 136 {
 137     JOB_LOCK_GUARD();
 138     job_txn_unref_locked(txn);
 139 }
 140
 141 /**
 142  * @txn: The transaction (may be NULL)
 143  * @job: Job to add to the transaction
 144  *
 145  * Add @job to the transaction.  The @job must not already be in a transaction.
 146  * The caller must call either job_txn_unref() or job_completed() to release
 147  * the reference that is automatically grabbed here.
 148  *
 149  * If @txn is NULL, the function does nothing.
 150  *
 151  * Called with job_mutex held.
 152  */
 153 static void job_txn_add_job_locked(JobTxn *txn, Job *job)
 154 {
 155     if (!txn) {
 156         return;
 157     }
 158
 159     assert(!job->txn);
 160     job->txn = txn;
 161
 162     QLIST_INSERT_HEAD(&txn->jobs, job, txn_list);
 163     job_txn_ref_locked(txn);
 164 }
 165
 166 /* Called with job_mutex held. */
 167 static void job_txn_del_job_locked(Job *job)
 168 {
 169     if (job->txn) {
 170         QLIST_REMOVE(job, txn_list);
 171         job_txn_unref_locked(job->txn);
 172         job->txn = NULL;
 173     }
 174 }
 175
 176 /* Called with job_mutex held, but releases it temporarily. */
 177 static int job_txn_apply_locked(Job *job, int fn(Job *))
 178 {
 179     Job *other_job, *next;
 180     JobTxn *txn = job->txn;
 181     int rc = 0;
 182
 183     /*
 184      * Similar to job_completed_txn_abort, we take each job's lock before
 185      * applying fn, but since we assume that outer_ctx is held by the caller,
 186      * we need to release it here to avoid holding the lock twice - which would
 187      * break AIO_WAIT_WHILE from within fn.
 188      */
 189     job_ref_locked(job);
 190
 191     QLIST_FOREACH_SAFE(other_job, &txn->jobs, txn_list, next) {
 192         rc = fn(other_job);
 193         if (rc) {
 194             break;
 195         }
 196     }
 197
 198     job_unref_locked(job);
 199     return rc;
 200 }
 201
 202 bool job_is_internal(Job *job)
 203 {
 204     return (job->id == NULL);
 205 }
 206
 207 /* Called with job_mutex held. */
 208 static void job_state_transition_locked(Job *job, JobStatus s1)
 209 {
 210     JobStatus s0 = job->status;
 211     assert(s1 >= 0 && s1 < JOB_STATUS__MAX);
 212     trace_job_state_transition(job, job->ret,
 213                                JobSTT[s0][s1] ? "allowed" : "disallowed",
 214                                JobStatus_str(s0), JobStatus_str(s1));
 215     assert(JobSTT[s0][s1]);
 216     job->status = s1;
 217
 218     if (!job_is_internal(job) && s1 != s0) {
 219         qapi_event_send_job_status_change(job->id, job->status);
 220     }
 221 }
 222
 223 int job_apply_verb_locked(Job *job, JobVerb verb, Error **errp)
 224 {
 225     JobStatus s0 = job->status;
 226     assert(verb >= 0 && verb < JOB_VERB__MAX);
 227     trace_job_apply_verb(job, JobStatus_str(s0), JobVerb_str(verb),
 228                          JobVerbTable[verb][s0] ? "allowed" : "prohibited");
 229     if (JobVerbTable[verb][s0]) {
 230         return 0;
 231     }
 232     error_setg(errp, "Job '%s' in state '%s' cannot accept command verb '%s'",
 233                job->id, JobStatus_str(s0), JobVerb_str(verb));
 234     return -EPERM;
 235 }
 236
 237 JobType job_type(const Job *job)
 238 {
 239     return job->driver->job_type;
 240 }
 241
 242 const char *job_type_str(const Job *job)
 243 {
 244     return JobType_str(job_type(job));
 245 }
 246
 247 bool job_is_cancelled_locked(Job *job)
 248 {
 249     /* force_cancel may be true only if cancelled is true, too */
 250     assert(job->cancelled || !job->force_cancel);
 251     return job->force_cancel;
 252 }
 253
 254 bool job_is_cancelled(Job *job)
 255 {
 256     JOB_LOCK_GUARD();
 257     return job_is_cancelled_locked(job);
 258 }
 259
 260 /* Called with job_mutex held. */
 261 static bool job_cancel_requested_locked(Job *job)
 262 {
 263     return job->cancelled;
 264 }
 265
 266 bool job_cancel_requested(Job *job)
 267 {
 268     JOB_LOCK_GUARD();
 269     return job_cancel_requested_locked(job);
 270 }
 271
 272 bool job_is_ready_locked(Job *job)
 273 {
 274     switch (job->status) {
 275     case JOB_STATUS_UNDEFINED:
 276     case JOB_STATUS_CREATED:
 277     case JOB_STATUS_RUNNING:
 278     case JOB_STATUS_PAUSED:
 279     case JOB_STATUS_WAITING:
 280     case JOB_STATUS_PENDING:
 281     case JOB_STATUS_ABORTING:
 282     case JOB_STATUS_CONCLUDED:
 283     case JOB_STATUS_NULL:
 284         return false;
 285     case JOB_STATUS_READY:
 286     case JOB_STATUS_STANDBY:
 287         return true;
 288     default:
 289         g_assert_not_reached();
 290     }
 291     return false;
 292 }
 293
 294 bool job_is_ready(Job *job)
 295 {
 296     JOB_LOCK_GUARD();
 297     return job_is_ready_locked(job);
 298 }
 299
 300 bool job_is_completed_locked(Job *job)
 301 {
 302     switch (job->status) {
 303     case JOB_STATUS_UNDEFINED:
 304     case JOB_STATUS_CREATED:
 305     case JOB_STATUS_RUNNING:
 306     case JOB_STATUS_PAUSED:
 307     case JOB_STATUS_READY:
 308     case JOB_STATUS_STANDBY:
 309         return false;
 310     case JOB_STATUS_WAITING:
 311     case JOB_STATUS_PENDING:
 312     case JOB_STATUS_ABORTING:
 313     case JOB_STATUS_CONCLUDED:
 314     case JOB_STATUS_NULL:
 315         return true;
 316     default:
 317         g_assert_not_reached();
 318     }
 319     return false;
 320 }
 321
 322 static bool job_is_completed(Job *job)
 323 {
 324     JOB_LOCK_GUARD();
 325     return job_is_completed_locked(job);
 326 }
 327
 328 static bool job_started_locked(Job *job)
 329 {
 330     return job->co;
 331 }
 332
 333 /* Called with job_mutex held. */
 334 static bool job_should_pause_locked(Job *job)
 335 {
 336     return job->pause_count > 0;
 337 }
 338
 339 Job *job_next_locked(Job *job)
 340 {
 341     if (!job) {
 342         return QLIST_FIRST(&jobs);
 343     }
 344     return QLIST_NEXT(job, job_list);
 345 }
 346
 347 Job *job_next(Job *job)
 348 {
 349     JOB_LOCK_GUARD();
 350     return job_next_locked(job);
 351 }
 352
 353 Job *job_get_locked(const char *id)
 354 {
 355     Job *job;
 356
 357     QLIST_FOREACH(job, &jobs, job_list) {
 358         if (job->id && !strcmp(id, job->id)) {
 359             return job;
 360         }
 361     }
 362
 363     return NULL;
 364 }
 365
 366 void job_set_aio_context(Job *job, AioContext *ctx)
 367 {
 368     /* protect against read in job_finish_sync_locked and job_start */
 369     GLOBAL_STATE_CODE();
 370     /* protect against read in job_do_yield_locked */
 371     JOB_LOCK_GUARD();
 372     /* ensure the job is quiescent while the AioContext is changed */
 373     assert(job->paused || job_is_completed_locked(job));
 374     job->aio_context = ctx;
 375 }
 376
 377 /* Called with job_mutex *not* held. */
 378 static void job_sleep_timer_cb(void *opaque)
 379 {
 380     Job *job = opaque;
 381
 382     job_enter(job);
 383 }
 384
 385 void *job_create(const char *job_id, const JobDriver *driver, JobTxn *txn,
 386                  AioContext *ctx, int flags, BlockCompletionFunc *cb,
 387                  void *opaque, Error **errp)
 388 {
 389     Job *job;
 390
 391     JOB_LOCK_GUARD();
 392
 393     if (job_id) {
 394         if (flags & JOB_INTERNAL) {
 395             error_setg(errp, "Cannot specify job ID for internal job");
 396             return NULL;
 397         }
 398         if (!id_wellformed(job_id)) {
 399             error_setg(errp, "Invalid job ID '%s'", job_id);
 400             return NULL;
 401         }
 402         if (job_get_locked(job_id)) {
 403             error_setg(errp, "Job ID '%s' already in use", job_id);
 404             return NULL;
 405         }
 406     } else if (!(flags & JOB_INTERNAL)) {
 407         error_setg(errp, "An explicit job ID is required");
 408         return NULL;
 409     }
 410
 411     job = g_malloc0(driver->instance_size);
 412     job->driver        = driver;
 413     job->id            = g_strdup(job_id);
 414     job->refcnt        = 1;
 415     job->aio_context   = ctx;
 416     job->busy          = false;
 417     job->paused        = true;
 418     job->pause_count   = 1;
 419     job->auto_finalize = !(flags & JOB_MANUAL_FINALIZE);
 420     job->auto_dismiss  = !(flags & JOB_MANUAL_DISMISS);
 421     job->cb            = cb;
 422     job->opaque        = opaque;
 423
 424     progress_init(&job->progress);
 425
 426     notifier_list_init(&job->on_finalize_cancelled);
 427     notifier_list_init(&job->on_finalize_completed);
 428     notifier_list_init(&job->on_pending);
 429     notifier_list_init(&job->on_ready);
 430     notifier_list_init(&job->on_idle);
 431
 432     job_state_transition_locked(job, JOB_STATUS_CREATED);
 433     aio_timer_init(qemu_get_aio_context(), &job->sleep_timer,
 434                    QEMU_CLOCK_REALTIME, SCALE_NS,
 435                    job_sleep_timer_cb, job);
 436
 437     QLIST_INSERT_HEAD(&jobs, job, job_list);
 438
 439     /* Single jobs are modeled as single-job transactions for sake of
 440      * consolidating the job management logic */
 441     if (!txn) {
 442         txn = job_txn_new();
 443         job_txn_add_job_locked(txn, job);
 444         job_txn_unref_locked(txn);
 445     } else {
 446         job_txn_add_job_locked(txn, job);
 447     }
 448
 449     return job;
 450 }
 451
 452 void job_ref_locked(Job *job)
 453 {
 454     ++job->refcnt;
 455 }
 456
 457 void job_unref_locked(Job *job)
 458 {
 459     GLOBAL_STATE_CODE();
 460
 461     if (--job->refcnt == 0) {
 462         assert(job->status == JOB_STATUS_NULL);
 463         assert(!timer_pending(&job->sleep_timer));
 464         assert(!job->txn);
 465
 466         if (job->driver->free) {
 467             job_unlock();
 468             job->driver->free(job);
 469             job_lock();
 470         }
 471
 472         QLIST_REMOVE(job, job_list);
 473
 474         progress_destroy(&job->progress);
 475         error_free(job->err);
 476         g_free(job->id);
 477         g_free(job);
 478     }
 479 }
 480
 481 void job_progress_update(Job *job, uint64_t done)
 482 {
 483     progress_work_done(&job->progress, done);
 484 }
 485
 486 void job_progress_set_remaining(Job *job, uint64_t remaining)
 487 {
 488     progress_set_remaining(&job->progress, remaining);
 489 }
 490
 491 void job_progress_increase_remaining(Job *job, uint64_t delta)
 492 {
 493     progress_increase_remaining(&job->progress, delta);
 494 }
 495
 496 /**
 497  * To be called when a cancelled job is finalised.
 498  * Called with job_mutex held.
 499  */
 500 static void job_event_cancelled_locked(Job *job)
 501 {
 502     notifier_list_notify(&job->on_finalize_cancelled, job);
 503 }
 504
 505 /**
 506  * To be called when a successfully completed job is finalised.
 507  * Called with job_mutex held.
 508  */
 509 static void job_event_completed_locked(Job *job)
 510 {
 511     notifier_list_notify(&job->on_finalize_completed, job);
 512 }
 513
 514 /* Called with job_mutex held. */
 515 static void job_event_pending_locked(Job *job)
 516 {
 517     notifier_list_notify(&job->on_pending, job);
 518 }
 519
 520 /* Called with job_mutex held. */
 521 static void job_event_ready_locked(Job *job)
 522 {
 523     notifier_list_notify(&job->on_ready, job);
 524 }
 525
 526 /* Called with job_mutex held. */
 527 static void job_event_idle_locked(Job *job)
 528 {
 529     notifier_list_notify(&job->on_idle, job);
 530 }
 531
 532 void job_enter_cond_locked(Job *job, bool(*fn)(Job *job))
 533 {
 534     if (!job_started_locked(job)) {
 535         return;
 536     }
 537     if (job->deferred_to_main_loop) {
 538         return;
 539     }
 540
 541     if (job->busy) {
 542         return;
 543     }
 544
 545     if (fn && !fn(job)) {
 546         return;
 547     }
 548
 549     assert(!job->deferred_to_main_loop);
 550     timer_del(&job->sleep_timer);
 551     job->busy = true;
 552     job_unlock();
 553     aio_co_wake(job->co);
 554     job_lock();
 555 }
 556
 557 void job_enter(Job *job)
 558 {
 559     JOB_LOCK_GUARD();
 560     job_enter_cond_locked(job, NULL);
 561 }
 562
 563 /* Yield, and schedule a timer to reenter the coroutine after @ns nanoseconds.
 564  * Reentering the job coroutine with job_enter() before the timer has expired
 565  * is allowed and cancels the timer.
 566  *
 567  * If @ns is (uint64_t) -1, no timer is scheduled and job_enter() must be
 568  * called explicitly.
 569  *
 570  * Called with job_mutex held, but releases it temporarily.
 571  */
 572 static void coroutine_fn job_do_yield_locked(Job *job, uint64_t ns)
 573 {
 574     AioContext *next_aio_context;
 575
 576     if (ns != -1) {
 577         timer_mod(&job->sleep_timer, ns);
 578     }
 579     job->busy = false;
 580     job_event_idle_locked(job);
 581     job_unlock();
 582     qemu_coroutine_yield();
 583     job_lock();
 584
 585     next_aio_context = job->aio_context;
 586     /*
 587      * Coroutine has resumed, but in the meanwhile the job AioContext
 588      * might have changed via bdrv_try_change_aio_context(), so we need to move
 589      * the coroutine too in the new aiocontext.
 590      */
 591     while (qemu_get_current_aio_context() != next_aio_context) {
 592         job_unlock();
 593         aio_co_reschedule_self(next_aio_context);
 594         job_lock();
 595         next_aio_context = job->aio_context;
 596     }
 597
 598     /* Set by job_enter_cond_locked() before re-entering the coroutine.  */
 599     assert(job->busy);
 600 }
 601
 602 /* Called with job_mutex held, but releases it temporarily. */
 603 static void coroutine_fn job_pause_point_locked(Job *job)
 604 {
 605     assert(job && job_started_locked(job));
 606
 607     if (!job_should_pause_locked(job)) {
 608         return;
 609     }
 610     if (job_is_cancelled_locked(job)) {
 611         return;
 612     }
 613
 614     if (job->driver->pause) {
 615         job_unlock();
 616         job->driver->pause(job);
 617         job_lock();
 618     }
 619
 620     if (job_should_pause_locked(job) && !job_is_cancelled_locked(job)) {
 621         JobStatus status = job->status;
 622         job_state_transition_locked(job, status == JOB_STATUS_READY
 623                                     ? JOB_STATUS_STANDBY
 624                                     : JOB_STATUS_PAUSED);
 625         job->paused = true;
 626         job_do_yield_locked(job, -1);
 627         job->paused = false;
 628         job_state_transition_locked(job, status);
 629     }
 630
 631     if (job->driver->resume) {
 632         job_unlock();
 633         job->driver->resume(job);
 634         job_lock();
 635     }
 636 }
 637
 638 void coroutine_fn job_pause_point(Job *job)
 639 {
 640     JOB_LOCK_GUARD();
 641     job_pause_point_locked(job);
 642 }
 643
 644 void coroutine_fn job_yield(Job *job)
 645 {
 646     JOB_LOCK_GUARD();
 647     assert(job->busy);
 648
 649     /* Check cancellation *before* setting busy = false, too!  */
 650     if (job_is_cancelled_locked(job)) {
 651         return;
 652     }
 653
 654     if (!job_should_pause_locked(job)) {
 655         job_do_yield_locked(job, -1);
 656     }
 657
 658     job_pause_point_locked(job);
 659 }
 660
 661 void coroutine_fn job_sleep_ns(Job *job, int64_t ns)
 662 {
 663     JOB_LOCK_GUARD();
 664     assert(job->busy);
 665
 666     /* Check cancellation *before* setting busy = false, too!  */
 667     if (job_is_cancelled_locked(job)) {
 668         return;
 669     }
 670
 671     if (!job_should_pause_locked(job)) {
 672         job_do_yield_locked(job, qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + ns);
 673     }
 674
 675     job_pause_point_locked(job);
 676 }
 677
 678 /* Assumes the job_mutex is held */
 679 static bool job_timer_not_pending_locked(Job *job)
 680 {
 681     return !timer_pending(&job->sleep_timer);
 682 }
 683
 684 void job_pause_locked(Job *job)
 685 {
 686     job->pause_count++;
 687     if (!job->paused) {
 688         job_enter_cond_locked(job, NULL);
 689     }
 690 }
 691
 692 void job_pause(Job *job)
 693 {
 694     JOB_LOCK_GUARD();
 695     job_pause_locked(job);
 696 }
 697
 698 void job_resume_locked(Job *job)
 699 {
 700     assert(job->pause_count > 0);
 701     job->pause_count--;
 702     if (job->pause_count) {
 703         return;
 704     }
 705
 706     /* kick only if no timer is pending */
 707     job_enter_cond_locked(job, job_timer_not_pending_locked);
 708 }
 709
 710 void job_resume(Job *job)
 711 {
 712     JOB_LOCK_GUARD();
 713     job_resume_locked(job);
 714 }
 715
 716 void job_user_pause_locked(Job *job, Error **errp)
 717 {
 718     if (job_apply_verb_locked(job, JOB_VERB_PAUSE, errp)) {
 719         return;
 720     }
 721     if (job->user_paused) {
 722         error_setg(errp, "Job is already paused");
 723         return;
 724     }
 725     job->user_paused = true;
 726     job_pause_locked(job);
 727 }
 728
 729 bool job_user_paused_locked(Job *job)
 730 {
 731     return job->user_paused;
 732 }
 733
 734 void job_user_resume_locked(Job *job, Error **errp)
 735 {
 736     assert(job);
 737     GLOBAL_STATE_CODE();
 738     if (!job->user_paused || job->pause_count <= 0) {
 739         error_setg(errp, "Can't resume a job that was not paused");
 740         return;
 741     }
 742     if (job_apply_verb_locked(job, JOB_VERB_RESUME, errp)) {
 743         return;
 744     }
 745     if (job->driver->user_resume) {
 746         job_unlock();
 747         job->driver->user_resume(job);
 748         job_lock();
 749     }
 750     job->user_paused = false;
 751     job_resume_locked(job);
 752 }
 753
 754 /* Called with job_mutex held, but releases it temporarily. */
 755 static void job_do_dismiss_locked(Job *job)
 756 {
 757     assert(job);
 758     job->busy = false;
 759     job->paused = false;
 760     job->deferred_to_main_loop = true;
 761
 762     job_txn_del_job_locked(job);
 763
 764     job_state_transition_locked(job, JOB_STATUS_NULL);
 765     job_unref_locked(job);
 766 }
 767
 768 void job_dismiss_locked(Job **jobptr, Error **errp)
 769 {
 770     Job *job = *jobptr;
 771     /* similarly to _complete, this is QMP-interface only. */
 772     assert(job->id);
 773     if (job_apply_verb_locked(job, JOB_VERB_DISMISS, errp)) {
 774         return;
 775     }
 776
 777     job_do_dismiss_locked(job);
 778     *jobptr = NULL;
 779 }
 780
 781 void job_early_fail(Job *job)
 782 {
 783     JOB_LOCK_GUARD();
 784     assert(job->status == JOB_STATUS_CREATED);
 785     job_do_dismiss_locked(job);
 786 }
 787
 788 /* Called with job_mutex held. */
 789 static void job_conclude_locked(Job *job)
 790 {
 791     job_state_transition_locked(job, JOB_STATUS_CONCLUDED);
 792     if (job->auto_dismiss || !job_started_locked(job)) {
 793         job_do_dismiss_locked(job);
 794     }
 795 }
 796
 797 /* Called with job_mutex held. */
 798 static void job_update_rc_locked(Job *job)
 799 {
 800     if (!job->ret && job_is_cancelled_locked(job)) {
 801         job->ret = -ECANCELED;
 802     }
 803     if (job->ret) {
 804         if (!job->err) {
 805             error_setg(&job->err, "%s", strerror(-job->ret));
 806         }
 807         job_state_transition_locked(job, JOB_STATUS_ABORTING);
 808     }
 809 }
 810
 811 static void job_commit(Job *job)
 812 {
 813     assert(!job->ret);
 814     GLOBAL_STATE_CODE();
 815     if (job->driver->commit) {
 816         job->driver->commit(job);
 817     }
 818 }
 819
 820 static void job_abort(Job *job)
 821 {
 822     assert(job->ret);
 823     GLOBAL_STATE_CODE();
 824     if (job->driver->abort) {
 825         job->driver->abort(job);
 826     }
 827 }
 828
 829 static void job_clean(Job *job)
 830 {
 831     GLOBAL_STATE_CODE();
 832     if (job->driver->clean) {
 833         job->driver->clean(job);
 834     }
 835 }
 836
 837 /*
 838  * Called with job_mutex held, but releases it temporarily.
 839  */
 840 static int job_finalize_single_locked(Job *job)
 841 {
 842     int job_ret;
 843
 844     assert(job_is_completed_locked(job));
 845
 846     /* Ensure abort is called for late-transactional failures */
 847     job_update_rc_locked(job);
 848
 849     job_ret = job->ret;
 850     job_unlock();
 851
 852     if (!job_ret) {
 853         job_commit(job);
 854     } else {
 855         job_abort(job);
 856     }
 857     job_clean(job);
 858
 859     if (job->cb) {
 860         job->cb(job->opaque, job_ret);
 861     }
 862
 863     job_lock();
 864
 865     /* Emit events only if we actually started */
 866     if (job_started_locked(job)) {
 867         if (job_is_cancelled_locked(job)) {
 868             job_event_cancelled_locked(job);
 869         } else {
 870             job_event_completed_locked(job);
 871         }
 872     }
 873
 874     job_txn_del_job_locked(job);
 875     job_conclude_locked(job);
 876     return 0;
 877 }
 878
 879 /*
 880  * Called with job_mutex held, but releases it temporarily.
 881  */
 882 static void job_cancel_async_locked(Job *job, bool force)
 883 {
 884     GLOBAL_STATE_CODE();
 885     if (job->driver->cancel) {
 886         job_unlock();
 887         force = job->driver->cancel(job, force);
 888         job_lock();
 889     } else {
 890         /* No .cancel() means the job will behave as if force-cancelled */
 891         force = true;
 892     }
 893
 894     if (job->user_paused) {
 895         /* Do not call job_enter here, the caller will handle it.  */
 896         if (job->driver->user_resume) {
 897             job_unlock();
 898             job->driver->user_resume(job);
 899             job_lock();
 900         }
 901         job->user_paused = false;
 902         assert(job->pause_count > 0);
 903         job->pause_count--;
 904     }
 905
 906     /*
 907      * Ignore soft cancel requests after the job is already done
 908      * (We will still invoke job->driver->cancel() above, but if the
 909      * job driver supports soft cancelling and the job is done, that
 910      * should be a no-op, too.  We still call it so it can override
 911      * @force.)
 912      */
 913     if (force || !job->deferred_to_main_loop) {
 914         job->cancelled = true;
 915         /* To prevent 'force == false' overriding a previous 'force == true' */
 916         job->force_cancel |= force;
 917     }
 918 }
 919
 920 /*
 921  * Called with job_mutex held, but releases it temporarily.
 922  */
 923 static void job_completed_txn_abort_locked(Job *job)
 924 {
 925     JobTxn *txn = job->txn;
 926     Job *other_job;
 927
 928     if (txn->aborting) {
 929         /*
 930          * We are cancelled by another job, which will handle everything.
 931          */
 932         return;
 933     }
 934     txn->aborting = true;
 935     job_txn_ref_locked(txn);
 936
 937     job_ref_locked(job);
 938
 939     /* Other jobs are effectively cancelled by us, set the status for
 940      * them; this job, however, may or may not be cancelled, depending
 941      * on the caller, so leave it. */
 942     QLIST_FOREACH(other_job, &txn->jobs, txn_list) {
 943         if (other_job != job) {
 944             /*
 945              * This is a transaction: If one job failed, no result will matter.
 946              * Therefore, pass force=true to terminate all other jobs as quickly
 947              * as possible.
 948              */
 949             job_cancel_async_locked(other_job, true);
 950         }
 951     }
 952     while (!QLIST_EMPTY(&txn->jobs)) {
 953         other_job = QLIST_FIRST(&txn->jobs);
 954         if (!job_is_completed_locked(other_job)) {
 955             assert(job_cancel_requested_locked(other_job));
 956             job_finish_sync_locked(other_job, NULL, NULL);
 957         }
 958         job_finalize_single_locked(other_job);
 959     }
 960
 961     job_unref_locked(job);
 962     job_txn_unref_locked(txn);
 963 }
 964
 965 /* Called with job_mutex held, but releases it temporarily */
 966 static int job_prepare_locked(Job *job)
 967 {
 968     int ret;
 969
 970     GLOBAL_STATE_CODE();
 971
 972     if (job->ret == 0 && job->driver->prepare) {
 973         job_unlock();
 974         ret = job->driver->prepare(job);
 975         job_lock();
 976         job->ret = ret;
 977         job_update_rc_locked(job);
 978     }
 979
 980     return job->ret;
 981 }
 982
 983 /* Called with job_mutex held */
 984 static int job_needs_finalize_locked(Job *job)
 985 {
 986     return !job->auto_finalize;
 987 }
 988
 989 /* Called with job_mutex held */
 990 static void job_do_finalize_locked(Job *job)
 991 {
 992     int rc;
 993     assert(job && job->txn);
 994
 995     /* prepare the transaction to complete */
 996     rc = job_txn_apply_locked(job, job_prepare_locked);
 997     if (rc) {
 998         job_completed_txn_abort_locked(job);
 999     } else {
1000         job_txn_apply_locked(job, job_finalize_single_locked);
1001     }
1002 }
1003
1004 void job_finalize_locked(Job *job, Error **errp)
1005 {
1006     assert(job && job->id);
1007     if (job_apply_verb_locked(job, JOB_VERB_FINALIZE, errp)) {
1008         return;
1009     }
1010     job_do_finalize_locked(job);
1011 }
1012
1013 /* Called with job_mutex held. */
1014 static int job_transition_to_pending_locked(Job *job)
1015 {
1016     job_state_transition_locked(job, JOB_STATUS_PENDING);
1017     if (!job->auto_finalize) {
1018         job_event_pending_locked(job);
1019     }
1020     return 0;
1021 }
1022
1023 void job_transition_to_ready(Job *job)
1024 {
1025     JOB_LOCK_GUARD();
1026     job_state_transition_locked(job, JOB_STATUS_READY);
1027     job_event_ready_locked(job);
1028 }
1029
1030 /* Called with job_mutex held. */
1031 static void job_completed_txn_success_locked(Job *job)
1032 {
1033     JobTxn *txn = job->txn;
1034     Job *other_job;
1035
1036     job_state_transition_locked(job, JOB_STATUS_WAITING);
1037
1038     /*
1039      * Successful completion, see if there are other running jobs in this
1040      * txn.
1041      */
1042     QLIST_FOREACH(other_job, &txn->jobs, txn_list) {
1043         if (!job_is_completed_locked(other_job)) {
1044             return;
1045         }
1046         assert(other_job->ret == 0);
1047     }
1048
1049     job_txn_apply_locked(job, job_transition_to_pending_locked);
1050
1051     /* If no jobs need manual finalization, automatically do so */
1052     if (job_txn_apply_locked(job, job_needs_finalize_locked) == 0) {
1053         job_do_finalize_locked(job);
1054     }
1055 }
1056
1057 /* Called with job_mutex held. */
1058 static void job_completed_locked(Job *job)
1059 {
1060     assert(job && job->txn && !job_is_completed_locked(job));
1061
1062     job_update_rc_locked(job);
1063     trace_job_completed(job, job->ret);
1064     if (job->ret) {
1065         job_completed_txn_abort_locked(job);
1066     } else {
1067         job_completed_txn_success_locked(job);
1068     }
1069 }
1070
1071 /**
1072  * Useful only as a type shim for aio_bh_schedule_oneshot.
1073  * Called with job_mutex *not* held.
1074  */
1075 static void job_exit(void *opaque)
1076 {
1077     Job *job = (Job *)opaque;
1078     JOB_LOCK_GUARD();
1079     job_ref_locked(job);
1080
1081     /* This is a lie, we're not quiescent, but still doing the completion
1082      * callbacks. However, completion callbacks tend to involve operations that
1083      * drain block nodes, and if .drained_poll still returned true, we would
1084      * deadlock. */
1085     job->busy = false;
1086     job_event_idle_locked(job);
1087
1088     job_completed_locked(job);
1089     job_unref_locked(job);
1090 }
1091
1092 /**
1093  * All jobs must allow a pause point before entering their job proper. This
1094  * ensures that jobs can be paused prior to being started, then resumed later.
1095  */
1096 static void coroutine_fn job_co_entry(void *opaque)
1097 {
1098     Job *job = opaque;
1099     int ret;
1100
1101     assert(job && job->driver && job->driver->run);
1102     WITH_JOB_LOCK_GUARD() {
1103         assert(job->aio_context == qemu_get_current_aio_context());
1104         job_pause_point_locked(job);
1105     }
1106     ret = job->driver->run(job, &job->err);
1107     WITH_JOB_LOCK_GUARD() {
1108         job->ret = ret;
1109         job->deferred_to_main_loop = true;
1110         job->busy = true;
1111     }
1112     aio_bh_schedule_oneshot(qemu_get_aio_context(), job_exit, job);
1113 }
1114
1115 void job_start(Job *job)
1116 {
1117     assert(qemu_in_main_thread());
1118
1119     WITH_JOB_LOCK_GUARD() {
1120         assert(job && !job_started_locked(job) && job->paused &&
1121             job->driver && job->driver->run);
1122         job->co = qemu_coroutine_create(job_co_entry, job);
1123         job->pause_count--;
1124         job->busy = true;
1125         job->paused = false;
1126         job_state_transition_locked(job, JOB_STATUS_RUNNING);
1127     }
1128     aio_co_enter(job->aio_context, job->co);
1129 }
1130
1131 void job_cancel_locked(Job *job, bool force)
1132 {
1133     if (job->status == JOB_STATUS_CONCLUDED) {
1134         job_do_dismiss_locked(job);
1135         return;
1136     }
1137     job_cancel_async_locked(job, force);
1138     if (!job_started_locked(job)) {
1139         job_completed_locked(job);
1140     } else if (job->deferred_to_main_loop) {
1141         /*
1142          * job_cancel_async() ignores soft-cancel requests for jobs
1143          * that are already done (i.e. deferred to the main loop).  We
1144          * have to check again whether the job is really cancelled.
1145          * (job_cancel_requested() and job_is_cancelled() are equivalent
1146          * here, because job_cancel_async() will make soft-cancel
1147          * requests no-ops when deferred_to_main_loop is true.  We
1148          * choose to call job_is_cancelled() to show that we invoke
1149          * job_completed_txn_abort() only for force-cancelled jobs.)
1150          */
1151         if (job_is_cancelled_locked(job)) {
1152             job_completed_txn_abort_locked(job);
1153         }
1154     } else {
1155         job_enter_cond_locked(job, NULL);
1156     }
1157 }
1158
1159 void job_user_cancel_locked(Job *job, bool force, Error **errp)
1160 {
1161     if (job_apply_verb_locked(job, JOB_VERB_CANCEL, errp)) {
1162         return;
1163     }
1164     job_cancel_locked(job, force);
1165 }
1166
1167 /* A wrapper around job_cancel_locked() taking an Error ** parameter so it may
1168  * be used with job_finish_sync_locked() without the need for (rather nasty)
1169  * function pointer casts there.
1170  *
1171  * Called with job_mutex held.
1172  */
1173 static void job_cancel_err_locked(Job *job, Error **errp)
1174 {
1175     job_cancel_locked(job, false);
1176 }
1177
1178 /**
1179  * Same as job_cancel_err(), but force-cancel.
1180  * Called with job_mutex held.
1181  */
1182 static void job_force_cancel_err_locked(Job *job, Error **errp)
1183 {
1184     job_cancel_locked(job, true);
1185 }
1186
1187 int job_cancel_sync_locked(Job *job, bool force)
1188 {
1189     if (force) {
1190         return job_finish_sync_locked(job, &job_force_cancel_err_locked, NULL);
1191     } else {
1192         return job_finish_sync_locked(job, &job_cancel_err_locked, NULL);
1193     }
1194 }
1195
1196 int job_cancel_sync(Job *job, bool force)
1197 {
1198     JOB_LOCK_GUARD();
1199     return job_cancel_sync_locked(job, force);
1200 }
1201
1202 void job_cancel_sync_all(void)
1203 {
1204     Job *job;
1205     JOB_LOCK_GUARD();
1206
1207     while ((job = job_next_locked(NULL))) {
1208         job_cancel_sync_locked(job, true);
1209     }
1210 }
1211
1212 int job_complete_sync_locked(Job *job, Error **errp)
1213 {
1214     return job_finish_sync_locked(job, job_complete_locked, errp);
1215 }
1216
1217 void job_complete_locked(Job *job, Error **errp)
1218 {
1219     /* Should not be reachable via external interface for internal jobs */
1220     assert(job->id);
1221     GLOBAL_STATE_CODE();
1222     if (job_apply_verb_locked(job, JOB_VERB_COMPLETE, errp)) {
1223         return;
1224     }
1225     if (job_cancel_requested_locked(job) || !job->driver->complete) {
1226         error_setg(errp, "The active block job '%s' cannot be completed",
1227                    job->id);
1228         return;
1229     }
1230
1231     job_unlock();
1232     job->driver->complete(job, errp);
1233     job_lock();
1234 }
1235
1236 int job_finish_sync_locked(Job *job,
1237                            void (*finish)(Job *, Error **errp),
1238                            Error **errp)
1239 {
1240     Error *local_err = NULL;
1241     int ret;
1242     GLOBAL_STATE_CODE();
1243
1244     job_ref_locked(job);
1245
1246     if (finish) {
1247         finish(job, &local_err);
1248     }
1249     if (local_err) {
1250         error_propagate(errp, local_err);
1251         job_unref_locked(job);
1252         return -EBUSY;
1253     }
1254
1255     job_unlock();
1256     AIO_WAIT_WHILE_UNLOCKED(job->aio_context,
1257                             (job_enter(job), !job_is_completed(job)));
1258     job_lock();
1259
1260     ret = (job_is_cancelled_locked(job) && job->ret == 0)
1261           ? -ECANCELED : job->ret;
1262     job_unref_locked(job);
1263     return ret;
1264 }