1 /* 2 * QEMU System Emulator block driver 3 * 4 * Copyright (c) 2011 IBM Corp. 5 * Copyright (c) 2012 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 "qemu-common.h" 28 #include "block/block.h" 29 #include "block/blockjob_int.h" 30 #include "block/block_int.h" 31 #include "sysemu/block-backend.h" 32 #include "qapi/qmp/qerror.h" 33 #include "qapi/qmp/qjson.h" 34 #include "qemu/coroutine.h" 35 #include "qemu/id.h" 36 #include "qmp-commands.h" 37 #include "qemu/timer.h" 38 #include "qapi-event.h" 39 40 static void block_job_event_cancelled(BlockJob *job); 41 static void block_job_event_completed(BlockJob *job, const char *msg); 42 43 /* Transactional group of block jobs */ 44 struct BlockJobTxn { 45 46 /* Is this txn being cancelled? */ 47 bool aborting; 48 49 /* List of jobs */ 50 QLIST_HEAD(, BlockJob) jobs; 51 52 /* Reference count */ 53 int refcnt; 54 }; 55 56 static QLIST_HEAD(, BlockJob) block_jobs = QLIST_HEAD_INITIALIZER(block_jobs); 57 58 BlockJob *block_job_next(BlockJob *job) 59 { 60 if (!job) { 61 return QLIST_FIRST(&block_jobs); 62 } 63 return QLIST_NEXT(job, job_list); 64 } 65 66 BlockJob *block_job_get(const char *id) 67 { 68 BlockJob *job; 69 70 QLIST_FOREACH(job, &block_jobs, job_list) { 71 if (job->id && !strcmp(id, job->id)) { 72 return job; 73 } 74 } 75 76 return NULL; 77 } 78 79 static void block_job_attached_aio_context(AioContext *new_context, 80 void *opaque) 81 { 82 BlockJob *job = opaque; 83 84 if (job->driver->attached_aio_context) { 85 job->driver->attached_aio_context(job, new_context); 86 } 87 88 block_job_resume(job); 89 } 90 91 static void block_job_drain(BlockJob *job) 92 { 93 /* If job is !job->busy this kicks it into the next pause point. */ 94 block_job_enter(job); 95 96 blk_drain(job->blk); 97 if (job->driver->drain) { 98 job->driver->drain(job); 99 } 100 } 101 102 static void block_job_detach_aio_context(void *opaque) 103 { 104 BlockJob *job = opaque; 105 106 /* In case the job terminates during aio_poll()... */ 107 block_job_ref(job); 108 109 block_job_pause(job); 110 111 while (!job->paused && !job->completed) { 112 block_job_drain(job); 113 } 114 115 block_job_unref(job); 116 } 117 118 void block_job_add_bdrv(BlockJob *job, BlockDriverState *bs) 119 { 120 job->nodes = g_slist_prepend(job->nodes, bs); 121 bdrv_ref(bs); 122 bdrv_op_block_all(bs, job->blocker); 123 } 124 125 void *block_job_create(const char *job_id, const BlockJobDriver *driver, 126 BlockDriverState *bs, int64_t speed, int flags, 127 BlockCompletionFunc *cb, void *opaque, Error **errp) 128 { 129 BlockBackend *blk; 130 BlockJob *job; 131 132 if (bs->job) { 133 error_setg(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs)); 134 return NULL; 135 } 136 137 if (job_id == NULL && !(flags & BLOCK_JOB_INTERNAL)) { 138 job_id = bdrv_get_device_name(bs); 139 if (!*job_id) { 140 error_setg(errp, "An explicit job ID is required for this node"); 141 return NULL; 142 } 143 } 144 145 if (job_id) { 146 if (flags & BLOCK_JOB_INTERNAL) { 147 error_setg(errp, "Cannot specify job ID for internal block job"); 148 return NULL; 149 } 150 151 if (!id_wellformed(job_id)) { 152 error_setg(errp, "Invalid job ID '%s'", job_id); 153 return NULL; 154 } 155 156 if (block_job_get(job_id)) { 157 error_setg(errp, "Job ID '%s' already in use", job_id); 158 return NULL; 159 } 160 } 161 162 blk = blk_new(); 163 blk_insert_bs(blk, bs); 164 165 job = g_malloc0(driver->instance_size); 166 error_setg(&job->blocker, "block device is in use by block job: %s", 167 BlockJobType_lookup[driver->job_type]); 168 block_job_add_bdrv(job, bs); 169 bdrv_op_unblock(bs, BLOCK_OP_TYPE_DATAPLANE, job->blocker); 170 171 job->driver = driver; 172 job->id = g_strdup(job_id); 173 job->blk = blk; 174 job->cb = cb; 175 job->opaque = opaque; 176 job->busy = false; 177 job->paused = true; 178 job->pause_count = 1; 179 job->refcnt = 1; 180 bs->job = job; 181 182 QLIST_INSERT_HEAD(&block_jobs, job, job_list); 183 184 blk_add_aio_context_notifier(blk, block_job_attached_aio_context, 185 block_job_detach_aio_context, job); 186 187 /* Only set speed when necessary to avoid NotSupported error */ 188 if (speed != 0) { 189 Error *local_err = NULL; 190 191 block_job_set_speed(job, speed, &local_err); 192 if (local_err) { 193 block_job_unref(job); 194 error_propagate(errp, local_err); 195 return NULL; 196 } 197 } 198 return job; 199 } 200 201 bool block_job_is_internal(BlockJob *job) 202 { 203 return (job->id == NULL); 204 } 205 206 static bool block_job_started(BlockJob *job) 207 { 208 return job->co; 209 } 210 211 void block_job_start(BlockJob *job) 212 { 213 assert(job && !block_job_started(job) && job->paused && 214 !job->busy && job->driver->start); 215 job->co = qemu_coroutine_create(job->driver->start, job); 216 if (--job->pause_count == 0) { 217 job->paused = false; 218 job->busy = true; 219 qemu_coroutine_enter(job->co); 220 } 221 } 222 223 void block_job_ref(BlockJob *job) 224 { 225 ++job->refcnt; 226 } 227 228 void block_job_unref(BlockJob *job) 229 { 230 if (--job->refcnt == 0) { 231 GSList *l; 232 BlockDriverState *bs = blk_bs(job->blk); 233 bs->job = NULL; 234 for (l = job->nodes; l; l = l->next) { 235 bs = l->data; 236 bdrv_op_unblock_all(bs, job->blocker); 237 bdrv_unref(bs); 238 } 239 g_slist_free(job->nodes); 240 blk_remove_aio_context_notifier(job->blk, 241 block_job_attached_aio_context, 242 block_job_detach_aio_context, job); 243 blk_unref(job->blk); 244 error_free(job->blocker); 245 g_free(job->id); 246 QLIST_REMOVE(job, job_list); 247 g_free(job); 248 } 249 } 250 251 static void block_job_completed_single(BlockJob *job) 252 { 253 if (!job->ret) { 254 if (job->driver->commit) { 255 job->driver->commit(job); 256 } 257 } else { 258 if (job->driver->abort) { 259 job->driver->abort(job); 260 } 261 } 262 if (job->driver->clean) { 263 job->driver->clean(job); 264 } 265 266 if (job->cb) { 267 job->cb(job->opaque, job->ret); 268 } 269 270 /* Emit events only if we actually started */ 271 if (block_job_started(job)) { 272 if (block_job_is_cancelled(job)) { 273 block_job_event_cancelled(job); 274 } else { 275 const char *msg = NULL; 276 if (job->ret < 0) { 277 msg = strerror(-job->ret); 278 } 279 block_job_event_completed(job, msg); 280 } 281 } 282 283 if (job->txn) { 284 QLIST_REMOVE(job, txn_list); 285 block_job_txn_unref(job->txn); 286 } 287 block_job_unref(job); 288 } 289 290 static void block_job_completed_txn_abort(BlockJob *job) 291 { 292 AioContext *ctx; 293 BlockJobTxn *txn = job->txn; 294 BlockJob *other_job, *next; 295 296 if (txn->aborting) { 297 /* 298 * We are cancelled by another job, which will handle everything. 299 */ 300 return; 301 } 302 txn->aborting = true; 303 /* We are the first failed job. Cancel other jobs. */ 304 QLIST_FOREACH(other_job, &txn->jobs, txn_list) { 305 ctx = blk_get_aio_context(other_job->blk); 306 aio_context_acquire(ctx); 307 } 308 QLIST_FOREACH(other_job, &txn->jobs, txn_list) { 309 if (other_job == job || other_job->completed) { 310 /* Other jobs are "effectively" cancelled by us, set the status for 311 * them; this job, however, may or may not be cancelled, depending 312 * on the caller, so leave it. */ 313 if (other_job != job) { 314 other_job->cancelled = true; 315 } 316 continue; 317 } 318 block_job_cancel_sync(other_job); 319 assert(other_job->completed); 320 } 321 QLIST_FOREACH_SAFE(other_job, &txn->jobs, txn_list, next) { 322 ctx = blk_get_aio_context(other_job->blk); 323 block_job_completed_single(other_job); 324 aio_context_release(ctx); 325 } 326 } 327 328 static void block_job_completed_txn_success(BlockJob *job) 329 { 330 AioContext *ctx; 331 BlockJobTxn *txn = job->txn; 332 BlockJob *other_job, *next; 333 /* 334 * Successful completion, see if there are other running jobs in this 335 * txn. 336 */ 337 QLIST_FOREACH(other_job, &txn->jobs, txn_list) { 338 if (!other_job->completed) { 339 return; 340 } 341 } 342 /* We are the last completed job, commit the transaction. */ 343 QLIST_FOREACH_SAFE(other_job, &txn->jobs, txn_list, next) { 344 ctx = blk_get_aio_context(other_job->blk); 345 aio_context_acquire(ctx); 346 assert(other_job->ret == 0); 347 block_job_completed_single(other_job); 348 aio_context_release(ctx); 349 } 350 } 351 352 void block_job_completed(BlockJob *job, int ret) 353 { 354 assert(blk_bs(job->blk)->job == job); 355 assert(!job->completed); 356 job->completed = true; 357 job->ret = ret; 358 if (!job->txn) { 359 block_job_completed_single(job); 360 } else if (ret < 0 || block_job_is_cancelled(job)) { 361 block_job_completed_txn_abort(job); 362 } else { 363 block_job_completed_txn_success(job); 364 } 365 } 366 367 void block_job_set_speed(BlockJob *job, int64_t speed, Error **errp) 368 { 369 Error *local_err = NULL; 370 371 if (!job->driver->set_speed) { 372 error_setg(errp, QERR_UNSUPPORTED); 373 return; 374 } 375 job->driver->set_speed(job, speed, &local_err); 376 if (local_err) { 377 error_propagate(errp, local_err); 378 return; 379 } 380 381 job->speed = speed; 382 } 383 384 void block_job_complete(BlockJob *job, Error **errp) 385 { 386 /* Should not be reachable via external interface for internal jobs */ 387 assert(job->id); 388 if (job->pause_count || job->cancelled || 389 !block_job_started(job) || !job->driver->complete) { 390 error_setg(errp, "The active block job '%s' cannot be completed", 391 job->id); 392 return; 393 } 394 395 job->driver->complete(job, errp); 396 } 397 398 void block_job_pause(BlockJob *job) 399 { 400 job->pause_count++; 401 } 402 403 void block_job_user_pause(BlockJob *job) 404 { 405 job->user_paused = true; 406 block_job_pause(job); 407 } 408 409 static bool block_job_should_pause(BlockJob *job) 410 { 411 return job->pause_count > 0; 412 } 413 414 bool block_job_user_paused(BlockJob *job) 415 { 416 return job ? job->user_paused : 0; 417 } 418 419 void coroutine_fn block_job_pause_point(BlockJob *job) 420 { 421 assert(job && block_job_started(job)); 422 423 if (!block_job_should_pause(job)) { 424 return; 425 } 426 if (block_job_is_cancelled(job)) { 427 return; 428 } 429 430 if (job->driver->pause) { 431 job->driver->pause(job); 432 } 433 434 if (block_job_should_pause(job) && !block_job_is_cancelled(job)) { 435 job->paused = true; 436 job->busy = false; 437 qemu_coroutine_yield(); /* wait for block_job_resume() */ 438 job->busy = true; 439 job->paused = false; 440 } 441 442 if (job->driver->resume) { 443 job->driver->resume(job); 444 } 445 } 446 447 void block_job_resume(BlockJob *job) 448 { 449 assert(job->pause_count > 0); 450 job->pause_count--; 451 if (job->pause_count) { 452 return; 453 } 454 block_job_enter(job); 455 } 456 457 void block_job_user_resume(BlockJob *job) 458 { 459 if (job && job->user_paused && job->pause_count > 0) { 460 job->user_paused = false; 461 block_job_resume(job); 462 } 463 } 464 465 void block_job_enter(BlockJob *job) 466 { 467 if (job->co && !job->busy) { 468 qemu_coroutine_enter(job->co); 469 } 470 } 471 472 void block_job_cancel(BlockJob *job) 473 { 474 if (block_job_started(job)) { 475 job->cancelled = true; 476 block_job_iostatus_reset(job); 477 block_job_enter(job); 478 } else { 479 block_job_completed(job, -ECANCELED); 480 } 481 } 482 483 bool block_job_is_cancelled(BlockJob *job) 484 { 485 return job->cancelled; 486 } 487 488 void block_job_iostatus_reset(BlockJob *job) 489 { 490 job->iostatus = BLOCK_DEVICE_IO_STATUS_OK; 491 if (job->driver->iostatus_reset) { 492 job->driver->iostatus_reset(job); 493 } 494 } 495 496 static int block_job_finish_sync(BlockJob *job, 497 void (*finish)(BlockJob *, Error **errp), 498 Error **errp) 499 { 500 Error *local_err = NULL; 501 int ret; 502 503 assert(blk_bs(job->blk)->job == job); 504 505 block_job_ref(job); 506 507 finish(job, &local_err); 508 if (local_err) { 509 error_propagate(errp, local_err); 510 block_job_unref(job); 511 return -EBUSY; 512 } 513 /* block_job_drain calls block_job_enter, and it should be enough to 514 * induce progress until the job completes or moves to the main thread. 515 */ 516 while (!job->deferred_to_main_loop && !job->completed) { 517 block_job_drain(job); 518 } 519 while (!job->completed) { 520 aio_poll(qemu_get_aio_context(), true); 521 } 522 ret = (job->cancelled && job->ret == 0) ? -ECANCELED : job->ret; 523 block_job_unref(job); 524 return ret; 525 } 526 527 /* A wrapper around block_job_cancel() taking an Error ** parameter so it may be 528 * used with block_job_finish_sync() without the need for (rather nasty) 529 * function pointer casts there. */ 530 static void block_job_cancel_err(BlockJob *job, Error **errp) 531 { 532 block_job_cancel(job); 533 } 534 535 int block_job_cancel_sync(BlockJob *job) 536 { 537 return block_job_finish_sync(job, &block_job_cancel_err, NULL); 538 } 539 540 void block_job_cancel_sync_all(void) 541 { 542 BlockJob *job; 543 AioContext *aio_context; 544 545 while ((job = QLIST_FIRST(&block_jobs))) { 546 aio_context = blk_get_aio_context(job->blk); 547 aio_context_acquire(aio_context); 548 block_job_cancel_sync(job); 549 aio_context_release(aio_context); 550 } 551 } 552 553 int block_job_complete_sync(BlockJob *job, Error **errp) 554 { 555 return block_job_finish_sync(job, &block_job_complete, errp); 556 } 557 558 void block_job_sleep_ns(BlockJob *job, QEMUClockType type, int64_t ns) 559 { 560 assert(job->busy); 561 562 /* Check cancellation *before* setting busy = false, too! */ 563 if (block_job_is_cancelled(job)) { 564 return; 565 } 566 567 job->busy = false; 568 if (!block_job_should_pause(job)) { 569 co_aio_sleep_ns(blk_get_aio_context(job->blk), type, ns); 570 } 571 job->busy = true; 572 573 block_job_pause_point(job); 574 } 575 576 void block_job_yield(BlockJob *job) 577 { 578 assert(job->busy); 579 580 /* Check cancellation *before* setting busy = false, too! */ 581 if (block_job_is_cancelled(job)) { 582 return; 583 } 584 585 job->busy = false; 586 if (!block_job_should_pause(job)) { 587 qemu_coroutine_yield(); 588 } 589 job->busy = true; 590 591 block_job_pause_point(job); 592 } 593 594 BlockJobInfo *block_job_query(BlockJob *job, Error **errp) 595 { 596 BlockJobInfo *info; 597 598 if (block_job_is_internal(job)) { 599 error_setg(errp, "Cannot query QEMU internal jobs"); 600 return NULL; 601 } 602 info = g_new0(BlockJobInfo, 1); 603 info->type = g_strdup(BlockJobType_lookup[job->driver->job_type]); 604 info->device = g_strdup(job->id); 605 info->len = job->len; 606 info->busy = job->busy; 607 info->paused = job->pause_count > 0; 608 info->offset = job->offset; 609 info->speed = job->speed; 610 info->io_status = job->iostatus; 611 info->ready = job->ready; 612 return info; 613 } 614 615 static void block_job_iostatus_set_err(BlockJob *job, int error) 616 { 617 if (job->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { 618 job->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : 619 BLOCK_DEVICE_IO_STATUS_FAILED; 620 } 621 } 622 623 static void block_job_event_cancelled(BlockJob *job) 624 { 625 if (block_job_is_internal(job)) { 626 return; 627 } 628 629 qapi_event_send_block_job_cancelled(job->driver->job_type, 630 job->id, 631 job->len, 632 job->offset, 633 job->speed, 634 &error_abort); 635 } 636 637 static void block_job_event_completed(BlockJob *job, const char *msg) 638 { 639 if (block_job_is_internal(job)) { 640 return; 641 } 642 643 qapi_event_send_block_job_completed(job->driver->job_type, 644 job->id, 645 job->len, 646 job->offset, 647 job->speed, 648 !!msg, 649 msg, 650 &error_abort); 651 } 652 653 void block_job_event_ready(BlockJob *job) 654 { 655 job->ready = true; 656 657 if (block_job_is_internal(job)) { 658 return; 659 } 660 661 qapi_event_send_block_job_ready(job->driver->job_type, 662 job->id, 663 job->len, 664 job->offset, 665 job->speed, &error_abort); 666 } 667 668 BlockErrorAction block_job_error_action(BlockJob *job, BlockdevOnError on_err, 669 int is_read, int error) 670 { 671 BlockErrorAction action; 672 673 switch (on_err) { 674 case BLOCKDEV_ON_ERROR_ENOSPC: 675 case BLOCKDEV_ON_ERROR_AUTO: 676 action = (error == ENOSPC) ? 677 BLOCK_ERROR_ACTION_STOP : BLOCK_ERROR_ACTION_REPORT; 678 break; 679 case BLOCKDEV_ON_ERROR_STOP: 680 action = BLOCK_ERROR_ACTION_STOP; 681 break; 682 case BLOCKDEV_ON_ERROR_REPORT: 683 action = BLOCK_ERROR_ACTION_REPORT; 684 break; 685 case BLOCKDEV_ON_ERROR_IGNORE: 686 action = BLOCK_ERROR_ACTION_IGNORE; 687 break; 688 default: 689 abort(); 690 } 691 if (!block_job_is_internal(job)) { 692 qapi_event_send_block_job_error(job->id, 693 is_read ? IO_OPERATION_TYPE_READ : 694 IO_OPERATION_TYPE_WRITE, 695 action, &error_abort); 696 } 697 if (action == BLOCK_ERROR_ACTION_STOP) { 698 /* make the pause user visible, which will be resumed from QMP. */ 699 block_job_user_pause(job); 700 block_job_iostatus_set_err(job, error); 701 } 702 return action; 703 } 704 705 typedef struct { 706 BlockJob *job; 707 AioContext *aio_context; 708 BlockJobDeferToMainLoopFn *fn; 709 void *opaque; 710 } BlockJobDeferToMainLoopData; 711 712 static void block_job_defer_to_main_loop_bh(void *opaque) 713 { 714 BlockJobDeferToMainLoopData *data = opaque; 715 AioContext *aio_context; 716 717 /* Prevent race with block_job_defer_to_main_loop() */ 718 aio_context_acquire(data->aio_context); 719 720 /* Fetch BDS AioContext again, in case it has changed */ 721 aio_context = blk_get_aio_context(data->job->blk); 722 aio_context_acquire(aio_context); 723 724 data->job->deferred_to_main_loop = false; 725 data->fn(data->job, data->opaque); 726 727 aio_context_release(aio_context); 728 729 aio_context_release(data->aio_context); 730 731 g_free(data); 732 } 733 734 void block_job_defer_to_main_loop(BlockJob *job, 735 BlockJobDeferToMainLoopFn *fn, 736 void *opaque) 737 { 738 BlockJobDeferToMainLoopData *data = g_malloc(sizeof(*data)); 739 data->job = job; 740 data->aio_context = blk_get_aio_context(job->blk); 741 data->fn = fn; 742 data->opaque = opaque; 743 job->deferred_to_main_loop = true; 744 745 aio_bh_schedule_oneshot(qemu_get_aio_context(), 746 block_job_defer_to_main_loop_bh, data); 747 } 748 749 BlockJobTxn *block_job_txn_new(void) 750 { 751 BlockJobTxn *txn = g_new0(BlockJobTxn, 1); 752 QLIST_INIT(&txn->jobs); 753 txn->refcnt = 1; 754 return txn; 755 } 756 757 static void block_job_txn_ref(BlockJobTxn *txn) 758 { 759 txn->refcnt++; 760 } 761 762 void block_job_txn_unref(BlockJobTxn *txn) 763 { 764 if (txn && --txn->refcnt == 0) { 765 g_free(txn); 766 } 767 } 768 769 void block_job_txn_add_job(BlockJobTxn *txn, BlockJob *job) 770 { 771 if (!txn) { 772 return; 773 } 774 775 assert(!job->txn); 776 job->txn = txn; 777 778 QLIST_INSERT_HEAD(&txn->jobs, job, txn_list); 779 block_job_txn_ref(txn); 780 } 781