1 /* 2 * Block device elevator/IO-scheduler. 3 * 4 * Copyright (C) 2000 Andrea Arcangeli <andrea@suse.de> SuSE 5 * 6 * 30042000 Jens Axboe <axboe@kernel.dk> : 7 * 8 * Split the elevator a bit so that it is possible to choose a different 9 * one or even write a new "plug in". There are three pieces: 10 * - elevator_fn, inserts a new request in the queue list 11 * - elevator_merge_fn, decides whether a new buffer can be merged with 12 * an existing request 13 * - elevator_dequeue_fn, called when a request is taken off the active list 14 * 15 * 20082000 Dave Jones <davej@suse.de> : 16 * Removed tests for max-bomb-segments, which was breaking elvtune 17 * when run without -bN 18 * 19 * Jens: 20 * - Rework again to work with bio instead of buffer_heads 21 * - loose bi_dev comparisons, partition handling is right now 22 * - completely modularize elevator setup and teardown 23 * 24 */ 25 #include <linux/kernel.h> 26 #include <linux/fs.h> 27 #include <linux/blkdev.h> 28 #include <linux/elevator.h> 29 #include <linux/bio.h> 30 #include <linux/module.h> 31 #include <linux/slab.h> 32 #include <linux/init.h> 33 #include <linux/compiler.h> 34 #include <linux/blktrace_api.h> 35 #include <linux/hash.h> 36 #include <linux/uaccess.h> 37 #include <linux/pm_runtime.h> 38 #include <linux/blk-cgroup.h> 39 40 #include <trace/events/block.h> 41 42 #include "blk.h" 43 #include "blk-mq-sched.h" 44 45 static DEFINE_SPINLOCK(elv_list_lock); 46 static LIST_HEAD(elv_list); 47 48 /* 49 * Merge hash stuff. 50 */ 51 #define rq_hash_key(rq) (blk_rq_pos(rq) + blk_rq_sectors(rq)) 52 53 /* 54 * Query io scheduler to see if the current process issuing bio may be 55 * merged with rq. 56 */ 57 static int elv_iosched_allow_bio_merge(struct request *rq, struct bio *bio) 58 { 59 struct request_queue *q = rq->q; 60 struct elevator_queue *e = q->elevator; 61 62 if (e->uses_mq && e->type->ops.mq.allow_merge) 63 return e->type->ops.mq.allow_merge(q, rq, bio); 64 else if (!e->uses_mq && e->type->ops.sq.elevator_allow_bio_merge_fn) 65 return e->type->ops.sq.elevator_allow_bio_merge_fn(q, rq, bio); 66 67 return 1; 68 } 69 70 /* 71 * can we safely merge with this request? 72 */ 73 bool elv_bio_merge_ok(struct request *rq, struct bio *bio) 74 { 75 if (!blk_rq_merge_ok(rq, bio)) 76 return false; 77 78 if (!elv_iosched_allow_bio_merge(rq, bio)) 79 return false; 80 81 return true; 82 } 83 EXPORT_SYMBOL(elv_bio_merge_ok); 84 85 static struct elevator_type *elevator_find(const char *name) 86 { 87 struct elevator_type *e; 88 89 list_for_each_entry(e, &elv_list, list) { 90 if (!strcmp(e->elevator_name, name)) 91 return e; 92 } 93 94 return NULL; 95 } 96 97 static void elevator_put(struct elevator_type *e) 98 { 99 module_put(e->elevator_owner); 100 } 101 102 static struct elevator_type *elevator_get(const char *name, bool try_loading) 103 { 104 struct elevator_type *e; 105 106 spin_lock(&elv_list_lock); 107 108 e = elevator_find(name); 109 if (!e && try_loading) { 110 spin_unlock(&elv_list_lock); 111 request_module("%s-iosched", name); 112 spin_lock(&elv_list_lock); 113 e = elevator_find(name); 114 } 115 116 if (e && !try_module_get(e->elevator_owner)) 117 e = NULL; 118 119 spin_unlock(&elv_list_lock); 120 121 return e; 122 } 123 124 static char chosen_elevator[ELV_NAME_MAX]; 125 126 static int __init elevator_setup(char *str) 127 { 128 /* 129 * Be backwards-compatible with previous kernels, so users 130 * won't get the wrong elevator. 131 */ 132 strncpy(chosen_elevator, str, sizeof(chosen_elevator) - 1); 133 return 1; 134 } 135 136 __setup("elevator=", elevator_setup); 137 138 /* called during boot to load the elevator chosen by the elevator param */ 139 void __init load_default_elevator_module(void) 140 { 141 struct elevator_type *e; 142 143 if (!chosen_elevator[0]) 144 return; 145 146 spin_lock(&elv_list_lock); 147 e = elevator_find(chosen_elevator); 148 spin_unlock(&elv_list_lock); 149 150 if (!e) 151 request_module("%s-iosched", chosen_elevator); 152 } 153 154 static struct kobj_type elv_ktype; 155 156 struct elevator_queue *elevator_alloc(struct request_queue *q, 157 struct elevator_type *e) 158 { 159 struct elevator_queue *eq; 160 161 eq = kzalloc_node(sizeof(*eq), GFP_KERNEL, q->node); 162 if (unlikely(!eq)) 163 return NULL; 164 165 eq->type = e; 166 kobject_init(&eq->kobj, &elv_ktype); 167 mutex_init(&eq->sysfs_lock); 168 hash_init(eq->hash); 169 eq->uses_mq = e->uses_mq; 170 171 return eq; 172 } 173 EXPORT_SYMBOL(elevator_alloc); 174 175 static void elevator_release(struct kobject *kobj) 176 { 177 struct elevator_queue *e; 178 179 e = container_of(kobj, struct elevator_queue, kobj); 180 elevator_put(e->type); 181 kfree(e); 182 } 183 184 int elevator_init(struct request_queue *q, char *name) 185 { 186 struct elevator_type *e = NULL; 187 int err; 188 189 /* 190 * q->sysfs_lock must be held to provide mutual exclusion between 191 * elevator_switch() and here. 192 */ 193 lockdep_assert_held(&q->sysfs_lock); 194 195 if (unlikely(q->elevator)) 196 return 0; 197 198 INIT_LIST_HEAD(&q->queue_head); 199 q->last_merge = NULL; 200 q->end_sector = 0; 201 q->boundary_rq = NULL; 202 203 if (name) { 204 e = elevator_get(name, true); 205 if (!e) 206 return -EINVAL; 207 } 208 209 /* 210 * Use the default elevator specified by config boot param for 211 * non-mq devices, or by config option. Don't try to load modules 212 * as we could be running off async and request_module() isn't 213 * allowed from async. 214 */ 215 if (!e && !q->mq_ops && *chosen_elevator) { 216 e = elevator_get(chosen_elevator, false); 217 if (!e) 218 printk(KERN_ERR "I/O scheduler %s not found\n", 219 chosen_elevator); 220 } 221 222 if (!e) { 223 /* 224 * For blk-mq devices, we default to using mq-deadline, 225 * if available, for single queue devices. If deadline 226 * isn't available OR we have multiple queues, default 227 * to "none". 228 */ 229 if (q->mq_ops) { 230 if (q->nr_hw_queues == 1) 231 e = elevator_get("mq-deadline", false); 232 if (!e) 233 return 0; 234 } else 235 e = elevator_get(CONFIG_DEFAULT_IOSCHED, false); 236 237 if (!e) { 238 printk(KERN_ERR 239 "Default I/O scheduler not found. " \ 240 "Using noop.\n"); 241 e = elevator_get("noop", false); 242 } 243 } 244 245 if (e->uses_mq) { 246 err = blk_mq_sched_setup(q); 247 if (!err) 248 err = e->ops.mq.init_sched(q, e); 249 } else 250 err = e->ops.sq.elevator_init_fn(q, e); 251 if (err) { 252 if (e->uses_mq) 253 blk_mq_sched_teardown(q); 254 elevator_put(e); 255 } 256 return err; 257 } 258 EXPORT_SYMBOL(elevator_init); 259 260 void elevator_exit(struct elevator_queue *e) 261 { 262 mutex_lock(&e->sysfs_lock); 263 if (e->uses_mq && e->type->ops.mq.exit_sched) 264 e->type->ops.mq.exit_sched(e); 265 else if (!e->uses_mq && e->type->ops.sq.elevator_exit_fn) 266 e->type->ops.sq.elevator_exit_fn(e); 267 mutex_unlock(&e->sysfs_lock); 268 269 kobject_put(&e->kobj); 270 } 271 EXPORT_SYMBOL(elevator_exit); 272 273 static inline void __elv_rqhash_del(struct request *rq) 274 { 275 hash_del(&rq->hash); 276 rq->rq_flags &= ~RQF_HASHED; 277 } 278 279 void elv_rqhash_del(struct request_queue *q, struct request *rq) 280 { 281 if (ELV_ON_HASH(rq)) 282 __elv_rqhash_del(rq); 283 } 284 EXPORT_SYMBOL_GPL(elv_rqhash_del); 285 286 void elv_rqhash_add(struct request_queue *q, struct request *rq) 287 { 288 struct elevator_queue *e = q->elevator; 289 290 BUG_ON(ELV_ON_HASH(rq)); 291 hash_add(e->hash, &rq->hash, rq_hash_key(rq)); 292 rq->rq_flags |= RQF_HASHED; 293 } 294 EXPORT_SYMBOL_GPL(elv_rqhash_add); 295 296 void elv_rqhash_reposition(struct request_queue *q, struct request *rq) 297 { 298 __elv_rqhash_del(rq); 299 elv_rqhash_add(q, rq); 300 } 301 302 struct request *elv_rqhash_find(struct request_queue *q, sector_t offset) 303 { 304 struct elevator_queue *e = q->elevator; 305 struct hlist_node *next; 306 struct request *rq; 307 308 hash_for_each_possible_safe(e->hash, rq, next, hash, offset) { 309 BUG_ON(!ELV_ON_HASH(rq)); 310 311 if (unlikely(!rq_mergeable(rq))) { 312 __elv_rqhash_del(rq); 313 continue; 314 } 315 316 if (rq_hash_key(rq) == offset) 317 return rq; 318 } 319 320 return NULL; 321 } 322 323 /* 324 * RB-tree support functions for inserting/lookup/removal of requests 325 * in a sorted RB tree. 326 */ 327 void elv_rb_add(struct rb_root *root, struct request *rq) 328 { 329 struct rb_node **p = &root->rb_node; 330 struct rb_node *parent = NULL; 331 struct request *__rq; 332 333 while (*p) { 334 parent = *p; 335 __rq = rb_entry(parent, struct request, rb_node); 336 337 if (blk_rq_pos(rq) < blk_rq_pos(__rq)) 338 p = &(*p)->rb_left; 339 else if (blk_rq_pos(rq) >= blk_rq_pos(__rq)) 340 p = &(*p)->rb_right; 341 } 342 343 rb_link_node(&rq->rb_node, parent, p); 344 rb_insert_color(&rq->rb_node, root); 345 } 346 EXPORT_SYMBOL(elv_rb_add); 347 348 void elv_rb_del(struct rb_root *root, struct request *rq) 349 { 350 BUG_ON(RB_EMPTY_NODE(&rq->rb_node)); 351 rb_erase(&rq->rb_node, root); 352 RB_CLEAR_NODE(&rq->rb_node); 353 } 354 EXPORT_SYMBOL(elv_rb_del); 355 356 struct request *elv_rb_find(struct rb_root *root, sector_t sector) 357 { 358 struct rb_node *n = root->rb_node; 359 struct request *rq; 360 361 while (n) { 362 rq = rb_entry(n, struct request, rb_node); 363 364 if (sector < blk_rq_pos(rq)) 365 n = n->rb_left; 366 else if (sector > blk_rq_pos(rq)) 367 n = n->rb_right; 368 else 369 return rq; 370 } 371 372 return NULL; 373 } 374 EXPORT_SYMBOL(elv_rb_find); 375 376 /* 377 * Insert rq into dispatch queue of q. Queue lock must be held on 378 * entry. rq is sort instead into the dispatch queue. To be used by 379 * specific elevators. 380 */ 381 void elv_dispatch_sort(struct request_queue *q, struct request *rq) 382 { 383 sector_t boundary; 384 struct list_head *entry; 385 386 if (q->last_merge == rq) 387 q->last_merge = NULL; 388 389 elv_rqhash_del(q, rq); 390 391 q->nr_sorted--; 392 393 boundary = q->end_sector; 394 list_for_each_prev(entry, &q->queue_head) { 395 struct request *pos = list_entry_rq(entry); 396 397 if (req_op(rq) != req_op(pos)) 398 break; 399 if (rq_data_dir(rq) != rq_data_dir(pos)) 400 break; 401 if (pos->rq_flags & (RQF_STARTED | RQF_SOFTBARRIER)) 402 break; 403 if (blk_rq_pos(rq) >= boundary) { 404 if (blk_rq_pos(pos) < boundary) 405 continue; 406 } else { 407 if (blk_rq_pos(pos) >= boundary) 408 break; 409 } 410 if (blk_rq_pos(rq) >= blk_rq_pos(pos)) 411 break; 412 } 413 414 list_add(&rq->queuelist, entry); 415 } 416 EXPORT_SYMBOL(elv_dispatch_sort); 417 418 /* 419 * Insert rq into dispatch queue of q. Queue lock must be held on 420 * entry. rq is added to the back of the dispatch queue. To be used by 421 * specific elevators. 422 */ 423 void elv_dispatch_add_tail(struct request_queue *q, struct request *rq) 424 { 425 if (q->last_merge == rq) 426 q->last_merge = NULL; 427 428 elv_rqhash_del(q, rq); 429 430 q->nr_sorted--; 431 432 q->end_sector = rq_end_sector(rq); 433 q->boundary_rq = rq; 434 list_add_tail(&rq->queuelist, &q->queue_head); 435 } 436 EXPORT_SYMBOL(elv_dispatch_add_tail); 437 438 enum elv_merge elv_merge(struct request_queue *q, struct request **req, 439 struct bio *bio) 440 { 441 struct elevator_queue *e = q->elevator; 442 struct request *__rq; 443 444 /* 445 * Levels of merges: 446 * nomerges: No merges at all attempted 447 * noxmerges: Only simple one-hit cache try 448 * merges: All merge tries attempted 449 */ 450 if (blk_queue_nomerges(q) || !bio_mergeable(bio)) 451 return ELEVATOR_NO_MERGE; 452 453 /* 454 * First try one-hit cache. 455 */ 456 if (q->last_merge && elv_bio_merge_ok(q->last_merge, bio)) { 457 enum elv_merge ret = blk_try_merge(q->last_merge, bio); 458 459 if (ret != ELEVATOR_NO_MERGE) { 460 *req = q->last_merge; 461 return ret; 462 } 463 } 464 465 if (blk_queue_noxmerges(q)) 466 return ELEVATOR_NO_MERGE; 467 468 /* 469 * See if our hash lookup can find a potential backmerge. 470 */ 471 __rq = elv_rqhash_find(q, bio->bi_iter.bi_sector); 472 if (__rq && elv_bio_merge_ok(__rq, bio)) { 473 *req = __rq; 474 return ELEVATOR_BACK_MERGE; 475 } 476 477 if (e->uses_mq && e->type->ops.mq.request_merge) 478 return e->type->ops.mq.request_merge(q, req, bio); 479 else if (!e->uses_mq && e->type->ops.sq.elevator_merge_fn) 480 return e->type->ops.sq.elevator_merge_fn(q, req, bio); 481 482 return ELEVATOR_NO_MERGE; 483 } 484 485 /* 486 * Attempt to do an insertion back merge. Only check for the case where 487 * we can append 'rq' to an existing request, so we can throw 'rq' away 488 * afterwards. 489 * 490 * Returns true if we merged, false otherwise 491 */ 492 bool elv_attempt_insert_merge(struct request_queue *q, struct request *rq) 493 { 494 struct request *__rq; 495 bool ret; 496 497 if (blk_queue_nomerges(q)) 498 return false; 499 500 /* 501 * First try one-hit cache. 502 */ 503 if (q->last_merge && blk_attempt_req_merge(q, q->last_merge, rq)) 504 return true; 505 506 if (blk_queue_noxmerges(q)) 507 return false; 508 509 ret = false; 510 /* 511 * See if our hash lookup can find a potential backmerge. 512 */ 513 while (1) { 514 __rq = elv_rqhash_find(q, blk_rq_pos(rq)); 515 if (!__rq || !blk_attempt_req_merge(q, __rq, rq)) 516 break; 517 518 /* The merged request could be merged with others, try again */ 519 ret = true; 520 rq = __rq; 521 } 522 523 return ret; 524 } 525 526 void elv_merged_request(struct request_queue *q, struct request *rq, 527 enum elv_merge type) 528 { 529 struct elevator_queue *e = q->elevator; 530 531 if (e->uses_mq && e->type->ops.mq.request_merged) 532 e->type->ops.mq.request_merged(q, rq, type); 533 else if (!e->uses_mq && e->type->ops.sq.elevator_merged_fn) 534 e->type->ops.sq.elevator_merged_fn(q, rq, type); 535 536 if (type == ELEVATOR_BACK_MERGE) 537 elv_rqhash_reposition(q, rq); 538 539 q->last_merge = rq; 540 } 541 542 void elv_merge_requests(struct request_queue *q, struct request *rq, 543 struct request *next) 544 { 545 struct elevator_queue *e = q->elevator; 546 bool next_sorted = false; 547 548 if (e->uses_mq && e->type->ops.mq.requests_merged) 549 e->type->ops.mq.requests_merged(q, rq, next); 550 else if (e->type->ops.sq.elevator_merge_req_fn) { 551 next_sorted = (__force bool)(next->rq_flags & RQF_SORTED); 552 if (next_sorted) 553 e->type->ops.sq.elevator_merge_req_fn(q, rq, next); 554 } 555 556 elv_rqhash_reposition(q, rq); 557 558 if (next_sorted) { 559 elv_rqhash_del(q, next); 560 q->nr_sorted--; 561 } 562 563 q->last_merge = rq; 564 } 565 566 void elv_bio_merged(struct request_queue *q, struct request *rq, 567 struct bio *bio) 568 { 569 struct elevator_queue *e = q->elevator; 570 571 if (WARN_ON_ONCE(e->uses_mq)) 572 return; 573 574 if (e->type->ops.sq.elevator_bio_merged_fn) 575 e->type->ops.sq.elevator_bio_merged_fn(q, rq, bio); 576 } 577 578 #ifdef CONFIG_PM 579 static void blk_pm_requeue_request(struct request *rq) 580 { 581 if (rq->q->dev && !(rq->rq_flags & RQF_PM)) 582 rq->q->nr_pending--; 583 } 584 585 static void blk_pm_add_request(struct request_queue *q, struct request *rq) 586 { 587 if (q->dev && !(rq->rq_flags & RQF_PM) && q->nr_pending++ == 0 && 588 (q->rpm_status == RPM_SUSPENDED || q->rpm_status == RPM_SUSPENDING)) 589 pm_request_resume(q->dev); 590 } 591 #else 592 static inline void blk_pm_requeue_request(struct request *rq) {} 593 static inline void blk_pm_add_request(struct request_queue *q, 594 struct request *rq) 595 { 596 } 597 #endif 598 599 void elv_requeue_request(struct request_queue *q, struct request *rq) 600 { 601 /* 602 * it already went through dequeue, we need to decrement the 603 * in_flight count again 604 */ 605 if (blk_account_rq(rq)) { 606 q->in_flight[rq_is_sync(rq)]--; 607 if (rq->rq_flags & RQF_SORTED) 608 elv_deactivate_rq(q, rq); 609 } 610 611 rq->rq_flags &= ~RQF_STARTED; 612 613 blk_pm_requeue_request(rq); 614 615 __elv_add_request(q, rq, ELEVATOR_INSERT_REQUEUE); 616 } 617 618 void elv_drain_elevator(struct request_queue *q) 619 { 620 struct elevator_queue *e = q->elevator; 621 static int printed; 622 623 if (WARN_ON_ONCE(e->uses_mq)) 624 return; 625 626 lockdep_assert_held(q->queue_lock); 627 628 while (e->type->ops.sq.elevator_dispatch_fn(q, 1)) 629 ; 630 if (q->nr_sorted && printed++ < 10) { 631 printk(KERN_ERR "%s: forced dispatching is broken " 632 "(nr_sorted=%u), please report this\n", 633 q->elevator->type->elevator_name, q->nr_sorted); 634 } 635 } 636 637 void __elv_add_request(struct request_queue *q, struct request *rq, int where) 638 { 639 trace_block_rq_insert(q, rq); 640 641 blk_pm_add_request(q, rq); 642 643 rq->q = q; 644 645 if (rq->rq_flags & RQF_SOFTBARRIER) { 646 /* barriers are scheduling boundary, update end_sector */ 647 if (!blk_rq_is_passthrough(rq)) { 648 q->end_sector = rq_end_sector(rq); 649 q->boundary_rq = rq; 650 } 651 } else if (!(rq->rq_flags & RQF_ELVPRIV) && 652 (where == ELEVATOR_INSERT_SORT || 653 where == ELEVATOR_INSERT_SORT_MERGE)) 654 where = ELEVATOR_INSERT_BACK; 655 656 switch (where) { 657 case ELEVATOR_INSERT_REQUEUE: 658 case ELEVATOR_INSERT_FRONT: 659 rq->rq_flags |= RQF_SOFTBARRIER; 660 list_add(&rq->queuelist, &q->queue_head); 661 break; 662 663 case ELEVATOR_INSERT_BACK: 664 rq->rq_flags |= RQF_SOFTBARRIER; 665 elv_drain_elevator(q); 666 list_add_tail(&rq->queuelist, &q->queue_head); 667 /* 668 * We kick the queue here for the following reasons. 669 * - The elevator might have returned NULL previously 670 * to delay requests and returned them now. As the 671 * queue wasn't empty before this request, ll_rw_blk 672 * won't run the queue on return, resulting in hang. 673 * - Usually, back inserted requests won't be merged 674 * with anything. There's no point in delaying queue 675 * processing. 676 */ 677 __blk_run_queue(q); 678 break; 679 680 case ELEVATOR_INSERT_SORT_MERGE: 681 /* 682 * If we succeed in merging this request with one in the 683 * queue already, we are done - rq has now been freed, 684 * so no need to do anything further. 685 */ 686 if (elv_attempt_insert_merge(q, rq)) 687 break; 688 case ELEVATOR_INSERT_SORT: 689 BUG_ON(blk_rq_is_passthrough(rq)); 690 rq->rq_flags |= RQF_SORTED; 691 q->nr_sorted++; 692 if (rq_mergeable(rq)) { 693 elv_rqhash_add(q, rq); 694 if (!q->last_merge) 695 q->last_merge = rq; 696 } 697 698 /* 699 * Some ioscheds (cfq) run q->request_fn directly, so 700 * rq cannot be accessed after calling 701 * elevator_add_req_fn. 702 */ 703 q->elevator->type->ops.sq.elevator_add_req_fn(q, rq); 704 break; 705 706 case ELEVATOR_INSERT_FLUSH: 707 rq->rq_flags |= RQF_SOFTBARRIER; 708 blk_insert_flush(rq); 709 break; 710 default: 711 printk(KERN_ERR "%s: bad insertion point %d\n", 712 __func__, where); 713 BUG(); 714 } 715 } 716 EXPORT_SYMBOL(__elv_add_request); 717 718 void elv_add_request(struct request_queue *q, struct request *rq, int where) 719 { 720 unsigned long flags; 721 722 spin_lock_irqsave(q->queue_lock, flags); 723 __elv_add_request(q, rq, where); 724 spin_unlock_irqrestore(q->queue_lock, flags); 725 } 726 EXPORT_SYMBOL(elv_add_request); 727 728 struct request *elv_latter_request(struct request_queue *q, struct request *rq) 729 { 730 struct elevator_queue *e = q->elevator; 731 732 if (e->uses_mq && e->type->ops.mq.next_request) 733 return e->type->ops.mq.next_request(q, rq); 734 else if (!e->uses_mq && e->type->ops.sq.elevator_latter_req_fn) 735 return e->type->ops.sq.elevator_latter_req_fn(q, rq); 736 737 return NULL; 738 } 739 740 struct request *elv_former_request(struct request_queue *q, struct request *rq) 741 { 742 struct elevator_queue *e = q->elevator; 743 744 if (e->uses_mq && e->type->ops.mq.former_request) 745 return e->type->ops.mq.former_request(q, rq); 746 if (!e->uses_mq && e->type->ops.sq.elevator_former_req_fn) 747 return e->type->ops.sq.elevator_former_req_fn(q, rq); 748 return NULL; 749 } 750 751 int elv_set_request(struct request_queue *q, struct request *rq, 752 struct bio *bio, gfp_t gfp_mask) 753 { 754 struct elevator_queue *e = q->elevator; 755 756 if (WARN_ON_ONCE(e->uses_mq)) 757 return 0; 758 759 if (e->type->ops.sq.elevator_set_req_fn) 760 return e->type->ops.sq.elevator_set_req_fn(q, rq, bio, gfp_mask); 761 return 0; 762 } 763 764 void elv_put_request(struct request_queue *q, struct request *rq) 765 { 766 struct elevator_queue *e = q->elevator; 767 768 if (WARN_ON_ONCE(e->uses_mq)) 769 return; 770 771 if (e->type->ops.sq.elevator_put_req_fn) 772 e->type->ops.sq.elevator_put_req_fn(rq); 773 } 774 775 int elv_may_queue(struct request_queue *q, unsigned int op) 776 { 777 struct elevator_queue *e = q->elevator; 778 779 if (WARN_ON_ONCE(e->uses_mq)) 780 return 0; 781 782 if (e->type->ops.sq.elevator_may_queue_fn) 783 return e->type->ops.sq.elevator_may_queue_fn(q, op); 784 785 return ELV_MQUEUE_MAY; 786 } 787 788 void elv_completed_request(struct request_queue *q, struct request *rq) 789 { 790 struct elevator_queue *e = q->elevator; 791 792 if (WARN_ON_ONCE(e->uses_mq)) 793 return; 794 795 /* 796 * request is released from the driver, io must be done 797 */ 798 if (blk_account_rq(rq)) { 799 q->in_flight[rq_is_sync(rq)]--; 800 if ((rq->rq_flags & RQF_SORTED) && 801 e->type->ops.sq.elevator_completed_req_fn) 802 e->type->ops.sq.elevator_completed_req_fn(q, rq); 803 } 804 } 805 806 #define to_elv(atr) container_of((atr), struct elv_fs_entry, attr) 807 808 static ssize_t 809 elv_attr_show(struct kobject *kobj, struct attribute *attr, char *page) 810 { 811 struct elv_fs_entry *entry = to_elv(attr); 812 struct elevator_queue *e; 813 ssize_t error; 814 815 if (!entry->show) 816 return -EIO; 817 818 e = container_of(kobj, struct elevator_queue, kobj); 819 mutex_lock(&e->sysfs_lock); 820 error = e->type ? entry->show(e, page) : -ENOENT; 821 mutex_unlock(&e->sysfs_lock); 822 return error; 823 } 824 825 static ssize_t 826 elv_attr_store(struct kobject *kobj, struct attribute *attr, 827 const char *page, size_t length) 828 { 829 struct elv_fs_entry *entry = to_elv(attr); 830 struct elevator_queue *e; 831 ssize_t error; 832 833 if (!entry->store) 834 return -EIO; 835 836 e = container_of(kobj, struct elevator_queue, kobj); 837 mutex_lock(&e->sysfs_lock); 838 error = e->type ? entry->store(e, page, length) : -ENOENT; 839 mutex_unlock(&e->sysfs_lock); 840 return error; 841 } 842 843 static const struct sysfs_ops elv_sysfs_ops = { 844 .show = elv_attr_show, 845 .store = elv_attr_store, 846 }; 847 848 static struct kobj_type elv_ktype = { 849 .sysfs_ops = &elv_sysfs_ops, 850 .release = elevator_release, 851 }; 852 853 int elv_register_queue(struct request_queue *q) 854 { 855 struct elevator_queue *e = q->elevator; 856 int error; 857 858 error = kobject_add(&e->kobj, &q->kobj, "%s", "iosched"); 859 if (!error) { 860 struct elv_fs_entry *attr = e->type->elevator_attrs; 861 if (attr) { 862 while (attr->attr.name) { 863 if (sysfs_create_file(&e->kobj, &attr->attr)) 864 break; 865 attr++; 866 } 867 } 868 kobject_uevent(&e->kobj, KOBJ_ADD); 869 e->registered = 1; 870 if (!e->uses_mq && e->type->ops.sq.elevator_registered_fn) 871 e->type->ops.sq.elevator_registered_fn(q); 872 } 873 return error; 874 } 875 EXPORT_SYMBOL(elv_register_queue); 876 877 void elv_unregister_queue(struct request_queue *q) 878 { 879 if (q) { 880 struct elevator_queue *e = q->elevator; 881 882 kobject_uevent(&e->kobj, KOBJ_REMOVE); 883 kobject_del(&e->kobj); 884 e->registered = 0; 885 } 886 } 887 EXPORT_SYMBOL(elv_unregister_queue); 888 889 int elv_register(struct elevator_type *e) 890 { 891 char *def = ""; 892 893 /* create icq_cache if requested */ 894 if (e->icq_size) { 895 if (WARN_ON(e->icq_size < sizeof(struct io_cq)) || 896 WARN_ON(e->icq_align < __alignof__(struct io_cq))) 897 return -EINVAL; 898 899 snprintf(e->icq_cache_name, sizeof(e->icq_cache_name), 900 "%s_io_cq", e->elevator_name); 901 e->icq_cache = kmem_cache_create(e->icq_cache_name, e->icq_size, 902 e->icq_align, 0, NULL); 903 if (!e->icq_cache) 904 return -ENOMEM; 905 } 906 907 /* register, don't allow duplicate names */ 908 spin_lock(&elv_list_lock); 909 if (elevator_find(e->elevator_name)) { 910 spin_unlock(&elv_list_lock); 911 if (e->icq_cache) 912 kmem_cache_destroy(e->icq_cache); 913 return -EBUSY; 914 } 915 list_add_tail(&e->list, &elv_list); 916 spin_unlock(&elv_list_lock); 917 918 /* print pretty message */ 919 if (!strcmp(e->elevator_name, chosen_elevator) || 920 (!*chosen_elevator && 921 !strcmp(e->elevator_name, CONFIG_DEFAULT_IOSCHED))) 922 def = " (default)"; 923 924 printk(KERN_INFO "io scheduler %s registered%s\n", e->elevator_name, 925 def); 926 return 0; 927 } 928 EXPORT_SYMBOL_GPL(elv_register); 929 930 void elv_unregister(struct elevator_type *e) 931 { 932 /* unregister */ 933 spin_lock(&elv_list_lock); 934 list_del_init(&e->list); 935 spin_unlock(&elv_list_lock); 936 937 /* 938 * Destroy icq_cache if it exists. icq's are RCU managed. Make 939 * sure all RCU operations are complete before proceeding. 940 */ 941 if (e->icq_cache) { 942 rcu_barrier(); 943 kmem_cache_destroy(e->icq_cache); 944 e->icq_cache = NULL; 945 } 946 } 947 EXPORT_SYMBOL_GPL(elv_unregister); 948 949 /* 950 * switch to new_e io scheduler. be careful not to introduce deadlocks - 951 * we don't free the old io scheduler, before we have allocated what we 952 * need for the new one. this way we have a chance of going back to the old 953 * one, if the new one fails init for some reason. 954 */ 955 static int elevator_switch(struct request_queue *q, struct elevator_type *new_e) 956 { 957 struct elevator_queue *old = q->elevator; 958 bool old_registered = false; 959 int err; 960 961 if (q->mq_ops) { 962 blk_mq_freeze_queue(q); 963 blk_mq_quiesce_queue(q); 964 } 965 966 /* 967 * Turn on BYPASS and drain all requests w/ elevator private data. 968 * Block layer doesn't call into a quiesced elevator - all requests 969 * are directly put on the dispatch list without elevator data 970 * using INSERT_BACK. All requests have SOFTBARRIER set and no 971 * merge happens either. 972 */ 973 if (old) { 974 old_registered = old->registered; 975 976 if (old->uses_mq) 977 blk_mq_sched_teardown(q); 978 979 if (!q->mq_ops) 980 blk_queue_bypass_start(q); 981 982 /* unregister and clear all auxiliary data of the old elevator */ 983 if (old_registered) 984 elv_unregister_queue(q); 985 986 ioc_clear_queue(q); 987 } 988 989 /* allocate, init and register new elevator */ 990 if (new_e) { 991 if (new_e->uses_mq) { 992 err = blk_mq_sched_setup(q); 993 if (!err) 994 err = new_e->ops.mq.init_sched(q, new_e); 995 } else 996 err = new_e->ops.sq.elevator_init_fn(q, new_e); 997 if (err) 998 goto fail_init; 999 1000 err = elv_register_queue(q); 1001 if (err) 1002 goto fail_register; 1003 } else 1004 q->elevator = NULL; 1005 1006 /* done, kill the old one and finish */ 1007 if (old) { 1008 elevator_exit(old); 1009 if (!q->mq_ops) 1010 blk_queue_bypass_end(q); 1011 } 1012 1013 if (q->mq_ops) { 1014 blk_mq_unfreeze_queue(q); 1015 blk_mq_start_stopped_hw_queues(q, true); 1016 } 1017 1018 if (new_e) 1019 blk_add_trace_msg(q, "elv switch: %s", new_e->elevator_name); 1020 else 1021 blk_add_trace_msg(q, "elv switch: none"); 1022 1023 return 0; 1024 1025 fail_register: 1026 if (q->mq_ops) 1027 blk_mq_sched_teardown(q); 1028 elevator_exit(q->elevator); 1029 fail_init: 1030 /* switch failed, restore and re-register old elevator */ 1031 if (old) { 1032 q->elevator = old; 1033 elv_register_queue(q); 1034 if (!q->mq_ops) 1035 blk_queue_bypass_end(q); 1036 } 1037 if (q->mq_ops) { 1038 blk_mq_unfreeze_queue(q); 1039 blk_mq_start_stopped_hw_queues(q, true); 1040 } 1041 1042 return err; 1043 } 1044 1045 /* 1046 * Switch this queue to the given IO scheduler. 1047 */ 1048 static int __elevator_change(struct request_queue *q, const char *name) 1049 { 1050 char elevator_name[ELV_NAME_MAX]; 1051 struct elevator_type *e; 1052 1053 /* 1054 * Special case for mq, turn off scheduling 1055 */ 1056 if (q->mq_ops && !strncmp(name, "none", 4)) 1057 return elevator_switch(q, NULL); 1058 1059 strlcpy(elevator_name, name, sizeof(elevator_name)); 1060 e = elevator_get(strstrip(elevator_name), true); 1061 if (!e) { 1062 printk(KERN_ERR "elevator: type %s not found\n", elevator_name); 1063 return -EINVAL; 1064 } 1065 1066 if (q->elevator && 1067 !strcmp(elevator_name, q->elevator->type->elevator_name)) { 1068 elevator_put(e); 1069 return 0; 1070 } 1071 1072 if (!e->uses_mq && q->mq_ops) { 1073 elevator_put(e); 1074 return -EINVAL; 1075 } 1076 if (e->uses_mq && !q->mq_ops) { 1077 elevator_put(e); 1078 return -EINVAL; 1079 } 1080 1081 return elevator_switch(q, e); 1082 } 1083 1084 int elevator_change(struct request_queue *q, const char *name) 1085 { 1086 int ret; 1087 1088 /* Protect q->elevator from elevator_init() */ 1089 mutex_lock(&q->sysfs_lock); 1090 ret = __elevator_change(q, name); 1091 mutex_unlock(&q->sysfs_lock); 1092 1093 return ret; 1094 } 1095 EXPORT_SYMBOL(elevator_change); 1096 1097 ssize_t elv_iosched_store(struct request_queue *q, const char *name, 1098 size_t count) 1099 { 1100 int ret; 1101 1102 if (!(q->mq_ops || q->request_fn)) 1103 return count; 1104 1105 ret = __elevator_change(q, name); 1106 if (!ret) 1107 return count; 1108 1109 printk(KERN_ERR "elevator: switch to %s failed\n", name); 1110 return ret; 1111 } 1112 1113 ssize_t elv_iosched_show(struct request_queue *q, char *name) 1114 { 1115 struct elevator_queue *e = q->elevator; 1116 struct elevator_type *elv = NULL; 1117 struct elevator_type *__e; 1118 int len = 0; 1119 1120 if (!blk_queue_stackable(q)) 1121 return sprintf(name, "none\n"); 1122 1123 if (!q->elevator) 1124 len += sprintf(name+len, "[none] "); 1125 else 1126 elv = e->type; 1127 1128 spin_lock(&elv_list_lock); 1129 list_for_each_entry(__e, &elv_list, list) { 1130 if (elv && !strcmp(elv->elevator_name, __e->elevator_name)) { 1131 len += sprintf(name+len, "[%s] ", elv->elevator_name); 1132 continue; 1133 } 1134 if (__e->uses_mq && q->mq_ops) 1135 len += sprintf(name+len, "%s ", __e->elevator_name); 1136 else if (!__e->uses_mq && !q->mq_ops) 1137 len += sprintf(name+len, "%s ", __e->elevator_name); 1138 } 1139 spin_unlock(&elv_list_lock); 1140 1141 if (q->mq_ops && q->elevator) 1142 len += sprintf(name+len, "none"); 1143 1144 len += sprintf(len+name, "\n"); 1145 return len; 1146 } 1147 1148 struct request *elv_rb_former_request(struct request_queue *q, 1149 struct request *rq) 1150 { 1151 struct rb_node *rbprev = rb_prev(&rq->rb_node); 1152 1153 if (rbprev) 1154 return rb_entry_rq(rbprev); 1155 1156 return NULL; 1157 } 1158 EXPORT_SYMBOL(elv_rb_former_request); 1159 1160 struct request *elv_rb_latter_request(struct request_queue *q, 1161 struct request *rq) 1162 { 1163 struct rb_node *rbnext = rb_next(&rq->rb_node); 1164 1165 if (rbnext) 1166 return rb_entry_rq(rbnext); 1167 1168 return NULL; 1169 } 1170 EXPORT_SYMBOL(elv_rb_latter_request); 1171