block/elevator.c

   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);