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/delay.h>
  35 #include <linux/blktrace_api.h>
  36 #include <linux/hash.h>
  37
  38 #include <asm/uaccess.h>
  39
  40 static DEFINE_SPINLOCK(elv_list_lock);
  41 static LIST_HEAD(elv_list);
  42
  43 /*
  44  * Merge hash stuff.
  45  */
  46 static const int elv_hash_shift = 6;
  47 #define ELV_HASH_BLOCK(sec)     ((sec) >> 3)
  48 #define ELV_HASH_FN(sec)        (hash_long(ELV_HASH_BLOCK((sec)), elv_hash_shift))
  49 #define ELV_HASH_ENTRIES        (1 << elv_hash_shift)
  50 #define rq_hash_key(rq)         ((rq)->sector + (rq)->nr_sectors)
  51 #define ELV_ON_HASH(rq)         (!hlist_unhashed(&(rq)->hash))
  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_merge(struct request *rq, struct bio *bio)
  58 {
  59         request_queue_t *q = rq->q;
  60         elevator_t *e = q->elevator;
  61
  62         if (e->ops->elevator_allow_merge_fn)
  63                 return e->ops->elevator_allow_merge_fn(q, rq, bio);
  64
  65         return 1;
  66 }
  67
  68 /*
  69  * can we safely merge with this request?
  70  */
  71 inline int elv_rq_merge_ok(struct request *rq, struct bio *bio)
  72 {
  73         if (!rq_mergeable(rq))
  74                 return 0;
  75
  76         /*
  77          * different data direction or already started, don't merge
  78          */
  79         if (bio_data_dir(bio) != rq_data_dir(rq))
  80                 return 0;
  81
  82         /*
  83          * must be same device and not a special request
  84          */
  85         if (rq->rq_disk != bio->bi_bdev->bd_disk || rq->special)
  86                 return 0;
  87
  88         if (!elv_iosched_allow_merge(rq, bio))
  89                 return 0;
  90
  91         return 1;
  92 }
  93 EXPORT_SYMBOL(elv_rq_merge_ok);
  94
  95 static inline int elv_try_merge(struct request *__rq, struct bio *bio)
  96 {
  97         int ret = ELEVATOR_NO_MERGE;
  98
  99         /*
 100          * we can merge and sequence is ok, check if it's possible
 101          */
 102         if (elv_rq_merge_ok(__rq, bio)) {
 103                 if (__rq->sector + __rq->nr_sectors == bio->bi_sector)
 104                         ret = ELEVATOR_BACK_MERGE;
 105                 else if (__rq->sector - bio_sectors(bio) == bio->bi_sector)
 106                         ret = ELEVATOR_FRONT_MERGE;
 107         }
 108
 109         return ret;
 110 }
 111
 112 static struct elevator_type *elevator_find(const char *name)
 113 {
 114         struct elevator_type *e;
 115
 116         list_for_each_entry(e, &elv_list, list) {
 117                 if (!strcmp(e->elevator_name, name))
 118                         return e;
 119         }
 120
 121         return NULL;
 122 }
 123
 124 static void elevator_put(struct elevator_type *e)
 125 {
 126         module_put(e->elevator_owner);
 127 }
 128
 129 static struct elevator_type *elevator_get(const char *name)
 130 {
 131         struct elevator_type *e;
 132
 133         spin_lock(&elv_list_lock);
 134
 135         e = elevator_find(name);
 136         if (e && !try_module_get(e->elevator_owner))
 137                 e = NULL;
 138
 139         spin_unlock(&elv_list_lock);
 140
 141         return e;
 142 }
 143
 144 static void *elevator_init_queue(request_queue_t *q, struct elevator_queue *eq)
 145 {
 146         return eq->ops->elevator_init_fn(q);
 147 }
 148
 149 static void elevator_attach(request_queue_t *q, struct elevator_queue *eq,
 150                            void *data)
 151 {
 152         q->elevator = eq;
 153         eq->elevator_data = data;
 154 }
 155
 156 static char chosen_elevator[16];
 157
 158 static int __init elevator_setup(char *str)
 159 {
 160         /*
 161          * Be backwards-compatible with previous kernels, so users
 162          * won't get the wrong elevator.
 163          */
 164         if (!strcmp(str, "as"))
 165                 strcpy(chosen_elevator, "anticipatory");
 166         else
 167                 strncpy(chosen_elevator, str, sizeof(chosen_elevator) - 1);
 168         return 1;
 169 }
 170
 171 __setup("elevator=", elevator_setup);
 172
 173 static struct kobj_type elv_ktype;
 174
 175 static elevator_t *elevator_alloc(request_queue_t *q, struct elevator_type *e)
 176 {
 177         elevator_t *eq;
 178         int i;
 179
 180         eq = kmalloc_node(sizeof(elevator_t), GFP_KERNEL, q->node);
 181         if (unlikely(!eq))
 182                 goto err;
 183
 184         memset(eq, 0, sizeof(*eq));
 185         eq->ops = &e->ops;
 186         eq->elevator_type = e;
 187         kobject_init(&eq->kobj);
 188         snprintf(eq->kobj.name, KOBJ_NAME_LEN, "%s", "iosched");
 189         eq->kobj.ktype = &elv_ktype;
 190         mutex_init(&eq->sysfs_lock);
 191
 192         eq->hash = kmalloc_node(sizeof(struct hlist_head) * ELV_HASH_ENTRIES,
 193                                         GFP_KERNEL, q->node);
 194         if (!eq->hash)
 195                 goto err;
 196
 197         for (i = 0; i < ELV_HASH_ENTRIES; i++)
 198                 INIT_HLIST_HEAD(&eq->hash[i]);
 199
 200         return eq;
 201 err:
 202         kfree(eq);
 203         elevator_put(e);
 204         return NULL;
 205 }
 206
 207 static void elevator_release(struct kobject *kobj)
 208 {
 209         elevator_t *e = container_of(kobj, elevator_t, kobj);
 210
 211         elevator_put(e->elevator_type);
 212         kfree(e->hash);
 213         kfree(e);
 214 }
 215
 216 int elevator_init(request_queue_t *q, char *name)
 217 {
 218         struct elevator_type *e = NULL;
 219         struct elevator_queue *eq;
 220         int ret = 0;
 221         void *data;
 222
 223         INIT_LIST_HEAD(&q->queue_head);
 224         q->last_merge = NULL;
 225         q->end_sector = 0;
 226         q->boundary_rq = NULL;
 227
 228         if (name && !(e = elevator_get(name)))
 229                 return -EINVAL;
 230
 231         if (!e && *chosen_elevator && !(e = elevator_get(chosen_elevator)))
 232                 printk("I/O scheduler %s not found\n", chosen_elevator);
 233
 234         if (!e && !(e = elevator_get(CONFIG_DEFAULT_IOSCHED))) {
 235                 printk("Default I/O scheduler not found, using no-op\n");
 236                 e = elevator_get("noop");
 237         }
 238
 239         eq = elevator_alloc(q, e);
 240         if (!eq)
 241                 return -ENOMEM;
 242
 243         data = elevator_init_queue(q, eq);
 244         if (!data) {
 245                 kobject_put(&eq->kobj);
 246                 return -ENOMEM;
 247         }
 248
 249         elevator_attach(q, eq, data);
 250         return ret;
 251 }
 252
 253 EXPORT_SYMBOL(elevator_init);
 254
 255 void elevator_exit(elevator_t *e)
 256 {
 257         mutex_lock(&e->sysfs_lock);
 258         if (e->ops->elevator_exit_fn)
 259                 e->ops->elevator_exit_fn(e);
 260         e->ops = NULL;
 261         mutex_unlock(&e->sysfs_lock);
 262
 263         kobject_put(&e->kobj);
 264 }
 265
 266 EXPORT_SYMBOL(elevator_exit);
 267
 268 static void elv_activate_rq(request_queue_t *q, struct request *rq)
 269 {
 270         elevator_t *e = q->elevator;
 271
 272         if (e->ops->elevator_activate_req_fn)
 273                 e->ops->elevator_activate_req_fn(q, rq);
 274 }
 275
 276 static void elv_deactivate_rq(request_queue_t *q, struct request *rq)
 277 {
 278         elevator_t *e = q->elevator;
 279
 280         if (e->ops->elevator_deactivate_req_fn)
 281                 e->ops->elevator_deactivate_req_fn(q, rq);
 282 }
 283
 284 static inline void __elv_rqhash_del(struct request *rq)
 285 {
 286         hlist_del_init(&rq->hash);
 287 }
 288
 289 static void elv_rqhash_del(request_queue_t *q, struct request *rq)
 290 {
 291         if (ELV_ON_HASH(rq))
 292                 __elv_rqhash_del(rq);
 293 }
 294
 295 static void elv_rqhash_add(request_queue_t *q, struct request *rq)
 296 {
 297         elevator_t *e = q->elevator;
 298
 299         BUG_ON(ELV_ON_HASH(rq));
 300         hlist_add_head(&rq->hash, &e->hash[ELV_HASH_FN(rq_hash_key(rq))]);
 301 }
 302
 303 static void elv_rqhash_reposition(request_queue_t *q, struct request *rq)
 304 {
 305         __elv_rqhash_del(rq);
 306         elv_rqhash_add(q, rq);
 307 }
 308
 309 static struct request *elv_rqhash_find(request_queue_t *q, sector_t offset)
 310 {
 311         elevator_t *e = q->elevator;
 312         struct hlist_head *hash_list = &e->hash[ELV_HASH_FN(offset)];
 313         struct hlist_node *entry, *next;
 314         struct request *rq;
 315
 316         hlist_for_each_entry_safe(rq, entry, next, hash_list, hash) {
 317                 BUG_ON(!ELV_ON_HASH(rq));
 318
 319                 if (unlikely(!rq_mergeable(rq))) {
 320                         __elv_rqhash_del(rq);
 321                         continue;
 322                 }
 323
 324                 if (rq_hash_key(rq) == offset)
 325                         return rq;
 326         }
 327
 328         return NULL;
 329 }
 330
 331 /*
 332  * RB-tree support functions for inserting/lookup/removal of requests
 333  * in a sorted RB tree.
 334  */
 335 struct request *elv_rb_add(struct rb_root *root, struct request *rq)
 336 {
 337         struct rb_node **p = &root->rb_node;
 338         struct rb_node *parent = NULL;
 339         struct request *__rq;
 340
 341         while (*p) {
 342                 parent = *p;
 343                 __rq = rb_entry(parent, struct request, rb_node);
 344
 345                 if (rq->sector < __rq->sector)
 346                         p = &(*p)->rb_left;
 347                 else if (rq->sector > __rq->sector)
 348                         p = &(*p)->rb_right;
 349                 else
 350                         return __rq;
 351         }
 352
 353         rb_link_node(&rq->rb_node, parent, p);
 354         rb_insert_color(&rq->rb_node, root);
 355         return NULL;
 356 }
 357
 358 EXPORT_SYMBOL(elv_rb_add);
 359
 360 void elv_rb_del(struct rb_root *root, struct request *rq)
 361 {
 362         BUG_ON(RB_EMPTY_NODE(&rq->rb_node));
 363         rb_erase(&rq->rb_node, root);
 364         RB_CLEAR_NODE(&rq->rb_node);
 365 }
 366
 367 EXPORT_SYMBOL(elv_rb_del);
 368
 369 struct request *elv_rb_find(struct rb_root *root, sector_t sector)
 370 {
 371         struct rb_node *n = root->rb_node;
 372         struct request *rq;
 373
 374         while (n) {
 375                 rq = rb_entry(n, struct request, rb_node);
 376
 377                 if (sector < rq->sector)
 378                         n = n->rb_left;
 379                 else if (sector > rq->sector)
 380                         n = n->rb_right;
 381                 else
 382                         return rq;
 383         }
 384
 385         return NULL;
 386 }
 387
 388 EXPORT_SYMBOL(elv_rb_find);
 389
 390 /*
 391  * Insert rq into dispatch queue of q.  Queue lock must be held on
 392  * entry.  rq is sort insted into the dispatch queue. To be used by
 393  * specific elevators.
 394  */
 395 void elv_dispatch_sort(request_queue_t *q, struct request *rq)
 396 {
 397         sector_t boundary;
 398         struct list_head *entry;
 399
 400         if (q->last_merge == rq)
 401                 q->last_merge = NULL;
 402
 403         elv_rqhash_del(q, rq);
 404
 405         q->nr_sorted--;
 406
 407         boundary = q->end_sector;
 408
 409         list_for_each_prev(entry, &q->queue_head) {
 410                 struct request *pos = list_entry_rq(entry);
 411
 412                 if (rq_data_dir(rq) != rq_data_dir(pos))
 413                         break;
 414                 if (pos->cmd_flags & (REQ_SOFTBARRIER|REQ_HARDBARRIER|REQ_STARTED))
 415                         break;
 416                 if (rq->sector >= boundary) {
 417                         if (pos->sector < boundary)
 418                                 continue;
 419                 } else {
 420                         if (pos->sector >= boundary)
 421                                 break;
 422                 }
 423                 if (rq->sector >= pos->sector)
 424                         break;
 425         }
 426
 427         list_add(&rq->queuelist, entry);
 428 }
 429
 430 EXPORT_SYMBOL(elv_dispatch_sort);
 431
 432 /*
 433  * Insert rq into dispatch queue of q.  Queue lock must be held on
 434  * entry.  rq is added to the back of the dispatch queue. To be used by
 435  * specific elevators.
 436  */
 437 void elv_dispatch_add_tail(struct request_queue *q, struct request *rq)
 438 {
 439         if (q->last_merge == rq)
 440                 q->last_merge = NULL;
 441
 442         elv_rqhash_del(q, rq);
 443
 444         q->nr_sorted--;
 445
 446         q->end_sector = rq_end_sector(rq);
 447         q->boundary_rq = rq;
 448         list_add_tail(&rq->queuelist, &q->queue_head);
 449 }
 450
 451 EXPORT_SYMBOL(elv_dispatch_add_tail);
 452
 453 int elv_merge(request_queue_t *q, struct request **req, struct bio *bio)
 454 {
 455         elevator_t *e = q->elevator;
 456         struct request *__rq;
 457         int ret;
 458
 459         /*
 460          * First try one-hit cache.
 461          */
 462         if (q->last_merge) {
 463                 ret = elv_try_merge(q->last_merge, bio);
 464                 if (ret != ELEVATOR_NO_MERGE) {
 465                         *req = q->last_merge;
 466                         return ret;
 467                 }
 468         }
 469
 470         /*
 471          * See if our hash lookup can find a potential backmerge.
 472          */
 473         __rq = elv_rqhash_find(q, bio->bi_sector);
 474         if (__rq && elv_rq_merge_ok(__rq, bio)) {
 475                 *req = __rq;
 476                 return ELEVATOR_BACK_MERGE;
 477         }
 478
 479         if (e->ops->elevator_merge_fn)
 480                 return e->ops->elevator_merge_fn(q, req, bio);
 481
 482         return ELEVATOR_NO_MERGE;
 483 }
 484
 485 void elv_merged_request(request_queue_t *q, struct request *rq, int type)
 486 {
 487         elevator_t *e = q->elevator;
 488
 489         if (e->ops->elevator_merged_fn)
 490                 e->ops->elevator_merged_fn(q, rq, type);
 491
 492         if (type == ELEVATOR_BACK_MERGE)
 493                 elv_rqhash_reposition(q, rq);
 494
 495         q->last_merge = rq;
 496 }
 497
 498 void elv_merge_requests(request_queue_t *q, struct request *rq,
 499                              struct request *next)
 500 {
 501         elevator_t *e = q->elevator;
 502
 503         if (e->ops->elevator_merge_req_fn)
 504                 e->ops->elevator_merge_req_fn(q, rq, next);
 505
 506         elv_rqhash_reposition(q, rq);
 507         elv_rqhash_del(q, next);
 508
 509         q->nr_sorted--;
 510         q->last_merge = rq;
 511 }
 512
 513 void elv_requeue_request(request_queue_t *q, struct request *rq)
 514 {
 515         /*
 516          * it already went through dequeue, we need to decrement the
 517          * in_flight count again
 518          */
 519         if (blk_account_rq(rq)) {
 520                 q->in_flight--;
 521                 if (blk_sorted_rq(rq))
 522                         elv_deactivate_rq(q, rq);
 523         }
 524
 525         rq->cmd_flags &= ~REQ_STARTED;
 526
 527         elv_insert(q, rq, ELEVATOR_INSERT_REQUEUE);
 528 }
 529
 530 static void elv_drain_elevator(request_queue_t *q)
 531 {
 532         static int printed;
 533         while (q->elevator->ops->elevator_dispatch_fn(q, 1))
 534                 ;
 535         if (q->nr_sorted == 0)
 536                 return;
 537         if (printed++ < 10) {
 538                 printk(KERN_ERR "%s: forced dispatching is broken "
 539                        "(nr_sorted=%u), please report this\n",
 540                        q->elevator->elevator_type->elevator_name, q->nr_sorted);
 541         }
 542 }
 543
 544 void elv_insert(request_queue_t *q, struct request *rq, int where)
 545 {
 546         struct list_head *pos;
 547         unsigned ordseq;
 548         int unplug_it = 1;
 549
 550         blk_add_trace_rq(q, rq, BLK_TA_INSERT);
 551
 552         rq->q = q;
 553
 554         switch (where) {
 555         case ELEVATOR_INSERT_FRONT:
 556                 rq->cmd_flags |= REQ_SOFTBARRIER;
 557
 558                 list_add(&rq->queuelist, &q->queue_head);
 559                 break;
 560
 561         case ELEVATOR_INSERT_BACK:
 562                 rq->cmd_flags |= REQ_SOFTBARRIER;
 563                 elv_drain_elevator(q);
 564                 list_add_tail(&rq->queuelist, &q->queue_head);
 565                 /*
 566                  * We kick the queue here for the following reasons.
 567                  * - The elevator might have returned NULL previously
 568                  *   to delay requests and returned them now.  As the
 569                  *   queue wasn't empty before this request, ll_rw_blk
 570                  *   won't run the queue on return, resulting in hang.
 571                  * - Usually, back inserted requests won't be merged
 572                  *   with anything.  There's no point in delaying queue
 573                  *   processing.
 574                  */
 575                 blk_remove_plug(q);
 576                 q->request_fn(q);
 577                 break;
 578
 579         case ELEVATOR_INSERT_SORT:
 580                 BUG_ON(!blk_fs_request(rq));
 581                 rq->cmd_flags |= REQ_SORTED;
 582                 q->nr_sorted++;
 583                 if (rq_mergeable(rq)) {
 584                         elv_rqhash_add(q, rq);
 585                         if (!q->last_merge)
 586                                 q->last_merge = rq;
 587                 }
 588
 589                 /*
 590                  * Some ioscheds (cfq) run q->request_fn directly, so
 591                  * rq cannot be accessed after calling
 592                  * elevator_add_req_fn.
 593                  */
 594                 q->elevator->ops->elevator_add_req_fn(q, rq);
 595                 break;
 596
 597         case ELEVATOR_INSERT_REQUEUE:
 598                 /*
 599                  * If ordered flush isn't in progress, we do front
 600                  * insertion; otherwise, requests should be requeued
 601                  * in ordseq order.
 602                  */
 603                 rq->cmd_flags |= REQ_SOFTBARRIER;
 604
 605                 /*
 606                  * Most requeues happen because of a busy condition,
 607                  * don't force unplug of the queue for that case.
 608                  */
 609                 unplug_it = 0;
 610
 611                 if (q->ordseq == 0) {
 612                         list_add(&rq->queuelist, &q->queue_head);
 613                         break;
 614                 }
 615
 616                 ordseq = blk_ordered_req_seq(rq);
 617
 618                 list_for_each(pos, &q->queue_head) {
 619                         struct request *pos_rq = list_entry_rq(pos);
 620                         if (ordseq <= blk_ordered_req_seq(pos_rq))
 621                                 break;
 622                 }
 623
 624                 list_add_tail(&rq->queuelist, pos);
 625                 break;
 626
 627         default:
 628                 printk(KERN_ERR "%s: bad insertion point %d\n",
 629                        __FUNCTION__, where);
 630                 BUG();
 631         }
 632
 633         if (unplug_it && blk_queue_plugged(q)) {
 634                 int nrq = q->rq.count[READ] + q->rq.count[WRITE]
 635                         - q->in_flight;
 636
 637                 if (nrq >= q->unplug_thresh)
 638                         __generic_unplug_device(q);
 639         }
 640 }
 641
 642 void __elv_add_request(request_queue_t *q, struct request *rq, int where,
 643                        int plug)
 644 {
 645         if (q->ordcolor)
 646                 rq->cmd_flags |= REQ_ORDERED_COLOR;
 647
 648         if (rq->cmd_flags & (REQ_SOFTBARRIER | REQ_HARDBARRIER)) {
 649                 /*
 650                  * toggle ordered color
 651                  */
 652                 if (blk_barrier_rq(rq))
 653                         q->ordcolor ^= 1;
 654
 655                 /*
 656                  * barriers implicitly indicate back insertion
 657                  */
 658                 if (where == ELEVATOR_INSERT_SORT)
 659                         where = ELEVATOR_INSERT_BACK;
 660
 661                 /*
 662                  * this request is scheduling boundary, update
 663                  * end_sector
 664                  */
 665                 if (blk_fs_request(rq)) {
 666                         q->end_sector = rq_end_sector(rq);
 667                         q->boundary_rq = rq;
 668                 }
 669         } else if (!(rq->cmd_flags & REQ_ELVPRIV) && where == ELEVATOR_INSERT_SORT)
 670                 where = ELEVATOR_INSERT_BACK;
 671
 672         if (plug)
 673                 blk_plug_device(q);
 674
 675         elv_insert(q, rq, where);
 676 }
 677
 678 EXPORT_SYMBOL(__elv_add_request);
 679
 680 void elv_add_request(request_queue_t *q, struct request *rq, int where,
 681                      int plug)
 682 {
 683         unsigned long flags;
 684
 685         spin_lock_irqsave(q->queue_lock, flags);
 686         __elv_add_request(q, rq, where, plug);
 687         spin_unlock_irqrestore(q->queue_lock, flags);
 688 }
 689
 690 EXPORT_SYMBOL(elv_add_request);
 691
 692 static inline struct request *__elv_next_request(request_queue_t *q)
 693 {
 694         struct request *rq;
 695
 696         while (1) {
 697                 while (!list_empty(&q->queue_head)) {
 698                         rq = list_entry_rq(q->queue_head.next);
 699                         if (blk_do_ordered(q, &rq))
 700                                 return rq;
 701                 }
 702
 703                 if (!q->elevator->ops->elevator_dispatch_fn(q, 0))
 704                         return NULL;
 705         }
 706 }
 707
 708 struct request *elv_next_request(request_queue_t *q)
 709 {
 710         struct request *rq;
 711         int ret;
 712
 713         while ((rq = __elv_next_request(q)) != NULL) {
 714                 if (!(rq->cmd_flags & REQ_STARTED)) {
 715                         /*
 716                          * This is the first time the device driver
 717                          * sees this request (possibly after
 718                          * requeueing).  Notify IO scheduler.
 719                          */
 720                         if (blk_sorted_rq(rq))
 721                                 elv_activate_rq(q, rq);
 722
 723                         /*
 724                          * just mark as started even if we don't start
 725                          * it, a request that has been delayed should
 726                          * not be passed by new incoming requests
 727                          */
 728                         rq->cmd_flags |= REQ_STARTED;
 729                         blk_add_trace_rq(q, rq, BLK_TA_ISSUE);
 730                 }
 731
 732                 if (!q->boundary_rq || q->boundary_rq == rq) {
 733                         q->end_sector = rq_end_sector(rq);
 734                         q->boundary_rq = NULL;
 735                 }
 736
 737                 if ((rq->cmd_flags & REQ_DONTPREP) || !q->prep_rq_fn)
 738                         break;
 739
 740                 ret = q->prep_rq_fn(q, rq);
 741                 if (ret == BLKPREP_OK) {
 742                         break;
 743                 } else if (ret == BLKPREP_DEFER) {
 744                         /*
 745                          * the request may have been (partially) prepped.
 746                          * we need to keep this request in the front to
 747                          * avoid resource deadlock.  REQ_STARTED will
 748                          * prevent other fs requests from passing this one.
 749                          */
 750                         rq = NULL;
 751                         break;
 752                 } else if (ret == BLKPREP_KILL) {
 753                         int nr_bytes = rq->hard_nr_sectors << 9;
 754
 755                         if (!nr_bytes)
 756                                 nr_bytes = rq->data_len;
 757
 758                         blkdev_dequeue_request(rq);
 759                         rq->cmd_flags |= REQ_QUIET;
 760                         end_that_request_chunk(rq, 0, nr_bytes);
 761                         end_that_request_last(rq, 0);
 762                 } else {
 763                         printk(KERN_ERR "%s: bad return=%d\n", __FUNCTION__,
 764                                                                 ret);
 765                         break;
 766                 }
 767         }
 768
 769         return rq;
 770 }
 771
 772 EXPORT_SYMBOL(elv_next_request);
 773
 774 void elv_dequeue_request(request_queue_t *q, struct request *rq)
 775 {
 776         BUG_ON(list_empty(&rq->queuelist));
 777         BUG_ON(ELV_ON_HASH(rq));
 778
 779         list_del_init(&rq->queuelist);
 780
 781         /*
 782          * the time frame between a request being removed from the lists
 783          * and to it is freed is accounted as io that is in progress at
 784          * the driver side.
 785          */
 786         if (blk_account_rq(rq))
 787                 q->in_flight++;
 788 }
 789
 790 EXPORT_SYMBOL(elv_dequeue_request);
 791
 792 int elv_queue_empty(request_queue_t *q)
 793 {
 794         elevator_t *e = q->elevator;
 795
 796         if (!list_empty(&q->queue_head))
 797                 return 0;
 798
 799         if (e->ops->elevator_queue_empty_fn)
 800                 return e->ops->elevator_queue_empty_fn(q);
 801
 802         return 1;
 803 }
 804
 805 EXPORT_SYMBOL(elv_queue_empty);
 806
 807 struct request *elv_latter_request(request_queue_t *q, struct request *rq)
 808 {
 809         elevator_t *e = q->elevator;
 810
 811         if (e->ops->elevator_latter_req_fn)
 812                 return e->ops->elevator_latter_req_fn(q, rq);
 813         return NULL;
 814 }
 815
 816 struct request *elv_former_request(request_queue_t *q, struct request *rq)
 817 {
 818         elevator_t *e = q->elevator;
 819
 820         if (e->ops->elevator_former_req_fn)
 821                 return e->ops->elevator_former_req_fn(q, rq);
 822         return NULL;
 823 }
 824
 825 int elv_set_request(request_queue_t *q, struct request *rq, gfp_t gfp_mask)
 826 {
 827         elevator_t *e = q->elevator;
 828
 829         if (e->ops->elevator_set_req_fn)
 830                 return e->ops->elevator_set_req_fn(q, rq, gfp_mask);
 831
 832         rq->elevator_private = NULL;
 833         return 0;
 834 }
 835
 836 void elv_put_request(request_queue_t *q, struct request *rq)
 837 {
 838         elevator_t *e = q->elevator;
 839
 840         if (e->ops->elevator_put_req_fn)
 841                 e->ops->elevator_put_req_fn(rq);
 842 }
 843
 844 int elv_may_queue(request_queue_t *q, int rw)
 845 {
 846         elevator_t *e = q->elevator;
 847
 848         if (e->ops->elevator_may_queue_fn)
 849                 return e->ops->elevator_may_queue_fn(q, rw);
 850
 851         return ELV_MQUEUE_MAY;
 852 }
 853
 854 void elv_completed_request(request_queue_t *q, struct request *rq)
 855 {
 856         elevator_t *e = q->elevator;
 857
 858         /*
 859          * request is released from the driver, io must be done
 860          */
 861         if (blk_account_rq(rq)) {
 862                 q->in_flight--;
 863                 if (blk_sorted_rq(rq) && e->ops->elevator_completed_req_fn)
 864                         e->ops->elevator_completed_req_fn(q, rq);
 865         }
 866
 867         /*
 868          * Check if the queue is waiting for fs requests to be
 869          * drained for flush sequence.
 870          */
 871         if (unlikely(q->ordseq)) {
 872                 struct request *first_rq = list_entry_rq(q->queue_head.next);
 873                 if (q->in_flight == 0 &&
 874                     blk_ordered_cur_seq(q) == QUEUE_ORDSEQ_DRAIN &&
 875                     blk_ordered_req_seq(first_rq) > QUEUE_ORDSEQ_DRAIN) {
 876                         blk_ordered_complete_seq(q, QUEUE_ORDSEQ_DRAIN, 0);
 877                         q->request_fn(q);
 878                 }
 879         }
 880 }
 881
 882 #define to_elv(atr) container_of((atr), struct elv_fs_entry, attr)
 883
 884 static ssize_t
 885 elv_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
 886 {
 887         elevator_t *e = container_of(kobj, elevator_t, kobj);
 888         struct elv_fs_entry *entry = to_elv(attr);
 889         ssize_t error;
 890
 891         if (!entry->show)
 892                 return -EIO;
 893
 894         mutex_lock(&e->sysfs_lock);
 895         error = e->ops ? entry->show(e, page) : -ENOENT;
 896         mutex_unlock(&e->sysfs_lock);
 897         return error;
 898 }
 899
 900 static ssize_t
 901 elv_attr_store(struct kobject *kobj, struct attribute *attr,
 902                const char *page, size_t length)
 903 {
 904         elevator_t *e = container_of(kobj, elevator_t, kobj);
 905         struct elv_fs_entry *entry = to_elv(attr);
 906         ssize_t error;
 907
 908         if (!entry->store)
 909                 return -EIO;
 910
 911         mutex_lock(&e->sysfs_lock);
 912         error = e->ops ? entry->store(e, page, length) : -ENOENT;
 913         mutex_unlock(&e->sysfs_lock);
 914         return error;
 915 }
 916
 917 static struct sysfs_ops elv_sysfs_ops = {
 918         .show   = elv_attr_show,
 919         .store  = elv_attr_store,
 920 };
 921
 922 static struct kobj_type elv_ktype = {
 923         .sysfs_ops      = &elv_sysfs_ops,
 924         .release        = elevator_release,
 925 };
 926
 927 int elv_register_queue(struct request_queue *q)
 928 {
 929         elevator_t *e = q->elevator;
 930         int error;
 931
 932         e->kobj.parent = &q->kobj;
 933
 934         error = kobject_add(&e->kobj);
 935         if (!error) {
 936                 struct elv_fs_entry *attr = e->elevator_type->elevator_attrs;
 937                 if (attr) {
 938                         while (attr->attr.name) {
 939                                 if (sysfs_create_file(&e->kobj, &attr->attr))
 940                                         break;
 941                                 attr++;
 942                         }
 943                 }
 944                 kobject_uevent(&e->kobj, KOBJ_ADD);
 945         }
 946         return error;
 947 }
 948
 949 static void __elv_unregister_queue(elevator_t *e)
 950 {
 951         kobject_uevent(&e->kobj, KOBJ_REMOVE);
 952         kobject_del(&e->kobj);
 953 }
 954
 955 void elv_unregister_queue(struct request_queue *q)
 956 {
 957         if (q)
 958                 __elv_unregister_queue(q->elevator);
 959 }
 960
 961 int elv_register(struct elevator_type *e)
 962 {
 963         char *def = "";
 964
 965         spin_lock(&elv_list_lock);
 966         BUG_ON(elevator_find(e->elevator_name));
 967         list_add_tail(&e->list, &elv_list);
 968         spin_unlock(&elv_list_lock);
 969
 970         if (!strcmp(e->elevator_name, chosen_elevator) ||
 971                         (!*chosen_elevator &&
 972                          !strcmp(e->elevator_name, CONFIG_DEFAULT_IOSCHED)))
 973                                 def = " (default)";
 974
 975         printk(KERN_INFO "io scheduler %s registered%s\n", e->elevator_name, def);
 976         return 0;
 977 }
 978 EXPORT_SYMBOL_GPL(elv_register);
 979
 980 void elv_unregister(struct elevator_type *e)
 981 {
 982         struct task_struct *g, *p;
 983
 984         /*
 985          * Iterate every thread in the process to remove the io contexts.
 986          */
 987         if (e->ops.trim) {
 988                 read_lock(&tasklist_lock);
 989                 do_each_thread(g, p) {
 990                         task_lock(p);
 991                         if (p->io_context)
 992                                 e->ops.trim(p->io_context);
 993                         task_unlock(p);
 994                 } while_each_thread(g, p);
 995                 read_unlock(&tasklist_lock);
 996         }
 997
 998         spin_lock(&elv_list_lock);
 999         list_del_init(&e->list);
1000         spin_unlock(&elv_list_lock);
1001 }
1002 EXPORT_SYMBOL_GPL(elv_unregister);
1003
1004 /*
1005  * switch to new_e io scheduler. be careful not to introduce deadlocks -
1006  * we don't free the old io scheduler, before we have allocated what we
1007  * need for the new one. this way we have a chance of going back to the old
1008  * one, if the new one fails init for some reason.
1009  */
1010 static int elevator_switch(request_queue_t *q, struct elevator_type *new_e)
1011 {
1012         elevator_t *old_elevator, *e;
1013         void *data;
1014
1015         /*
1016          * Allocate new elevator
1017          */
1018         e = elevator_alloc(q, new_e);
1019         if (!e)
1020                 return 0;
1021
1022         data = elevator_init_queue(q, e);
1023         if (!data) {
1024                 kobject_put(&e->kobj);
1025                 return 0;
1026         }
1027
1028         /*
1029          * Turn on BYPASS and drain all requests w/ elevator private data
1030          */
1031         spin_lock_irq(q->queue_lock);
1032
1033         set_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
1034
1035         elv_drain_elevator(q);
1036
1037         while (q->rq.elvpriv) {
1038                 blk_remove_plug(q);
1039                 q->request_fn(q);
1040                 spin_unlock_irq(q->queue_lock);
1041                 msleep(10);
1042                 spin_lock_irq(q->queue_lock);
1043                 elv_drain_elevator(q);
1044         }
1045
1046         /*
1047          * Remember old elevator.
1048          */
1049         old_elevator = q->elevator;
1050
1051         /*
1052          * attach and start new elevator
1053          */
1054         elevator_attach(q, e, data);
1055
1056         spin_unlock_irq(q->queue_lock);
1057
1058         __elv_unregister_queue(old_elevator);
1059
1060         if (elv_register_queue(q))
1061                 goto fail_register;
1062
1063         /*
1064          * finally exit old elevator and turn off BYPASS.
1065          */
1066         elevator_exit(old_elevator);
1067         clear_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
1068         return 1;
1069
1070 fail_register:
1071         /*
1072          * switch failed, exit the new io scheduler and reattach the old
1073          * one again (along with re-adding the sysfs dir)
1074          */
1075         elevator_exit(e);
1076         q->elevator = old_elevator;
1077         elv_register_queue(q);
1078         clear_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
1079         return 0;
1080 }
1081
1082 ssize_t elv_iosched_store(request_queue_t *q, const char *name, size_t count)
1083 {
1084         char elevator_name[ELV_NAME_MAX];
1085         size_t len;
1086         struct elevator_type *e;
1087
1088         elevator_name[sizeof(elevator_name) - 1] = '\0';
1089         strncpy(elevator_name, name, sizeof(elevator_name) - 1);
1090         len = strlen(elevator_name);
1091
1092         if (len && elevator_name[len - 1] == '\n')
1093                 elevator_name[len - 1] = '\0';
1094
1095         e = elevator_get(elevator_name);
1096         if (!e) {
1097                 printk(KERN_ERR "elevator: type %s not found\n", elevator_name);
1098                 return -EINVAL;
1099         }
1100
1101         if (!strcmp(elevator_name, q->elevator->elevator_type->elevator_name)) {
1102                 elevator_put(e);
1103                 return count;
1104         }
1105
1106         if (!elevator_switch(q, e))
1107                 printk(KERN_ERR "elevator: switch to %s failed\n",elevator_name);
1108         return count;
1109 }
1110
1111 ssize_t elv_iosched_show(request_queue_t *q, char *name)
1112 {
1113         elevator_t *e = q->elevator;
1114         struct elevator_type *elv = e->elevator_type;
1115         struct elevator_type *__e;
1116         int len = 0;
1117
1118         spin_lock(&elv_list_lock);
1119         list_for_each_entry(__e, &elv_list, list) {
1120                 if (!strcmp(elv->elevator_name, __e->elevator_name))
1121                         len += sprintf(name+len, "[%s] ", elv->elevator_name);
1122                 else
1123                         len += sprintf(name+len, "%s ", __e->elevator_name);
1124         }
1125         spin_unlock(&elv_list_lock);
1126
1127         len += sprintf(len+name, "\n");
1128         return len;
1129 }
1130
1131 struct request *elv_rb_former_request(request_queue_t *q, struct request *rq)
1132 {
1133         struct rb_node *rbprev = rb_prev(&rq->rb_node);
1134
1135         if (rbprev)
1136                 return rb_entry_rq(rbprev);
1137
1138         return NULL;
1139 }
1140
1141 EXPORT_SYMBOL(elv_rb_former_request);
1142
1143 struct request *elv_rb_latter_request(request_queue_t *q, struct request *rq)
1144 {
1145         struct rb_node *rbnext = rb_next(&rq->rb_node);
1146
1147         if (rbnext)
1148                 return rb_entry_rq(rbnext);
1149
1150         return NULL;
1151 }
1152
1153 EXPORT_SYMBOL(elv_rb_latter_request);