block/cfq-iosched.c

   1 /*
   2  *  CFQ, or complete fairness queueing, disk scheduler.
   3  *
   4  *  Based on ideas from a previously unfinished io
   5  *  scheduler (round robin per-process disk scheduling) and Andrea Arcangeli.
   6  *
   7  *  Copyright (C) 2003 Jens Axboe <axboe@suse.de>
   8  */
   9 #include <linux/kernel.h>
  10 #include <linux/fs.h>
  11 #include <linux/blkdev.h>
  12 #include <linux/elevator.h>
  13 #include <linux/bio.h>
  14 #include <linux/config.h>
  15 #include <linux/module.h>
  16 #include <linux/slab.h>
  17 #include <linux/init.h>
  18 #include <linux/compiler.h>
  19 #include <linux/hash.h>
  20 #include <linux/rbtree.h>
  21 #include <linux/mempool.h>
  22 #include <linux/ioprio.h>
  23 #include <linux/writeback.h>
  24
  25 /*
  26  * tunables
  27  */
  28 static const int cfq_quantum = 4;               /* max queue in one round of service */
  29 static const int cfq_queued = 8;                /* minimum rq allocate limit per-queue*/
  30 static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 };
  31 static const int cfq_back_max = 16 * 1024;      /* maximum backwards seek, in KiB */
  32 static const int cfq_back_penalty = 2;          /* penalty of a backwards seek */
  33
  34 static const int cfq_slice_sync = HZ / 10;
  35 static int cfq_slice_async = HZ / 25;
  36 static const int cfq_slice_async_rq = 2;
  37 static int cfq_slice_idle = HZ / 100;
  38
  39 #define CFQ_IDLE_GRACE          (HZ / 10)
  40 #define CFQ_SLICE_SCALE         (5)
  41
  42 #define CFQ_KEY_ASYNC           (0)
  43 #define CFQ_KEY_ANY             (0xffff)
  44
  45 /*
  46  * disable queueing at the driver/hardware level
  47  */
  48 static const int cfq_max_depth = 2;
  49
  50 static DEFINE_RWLOCK(cfq_exit_lock);
  51
  52 /*
  53  * for the hash of cfqq inside the cfqd
  54  */
  55 #define CFQ_QHASH_SHIFT         6
  56 #define CFQ_QHASH_ENTRIES       (1 << CFQ_QHASH_SHIFT)
  57 #define list_entry_qhash(entry) hlist_entry((entry), struct cfq_queue, cfq_hash)
  58
  59 /*
  60  * for the hash of crq inside the cfqq
  61  */
  62 #define CFQ_MHASH_SHIFT         6
  63 #define CFQ_MHASH_BLOCK(sec)    ((sec) >> 3)
  64 #define CFQ_MHASH_ENTRIES       (1 << CFQ_MHASH_SHIFT)
  65 #define CFQ_MHASH_FN(sec)       hash_long(CFQ_MHASH_BLOCK(sec), CFQ_MHASH_SHIFT)
  66 #define rq_hash_key(rq)         ((rq)->sector + (rq)->nr_sectors)
  67 #define list_entry_hash(ptr)    hlist_entry((ptr), struct cfq_rq, hash)
  68
  69 #define list_entry_cfqq(ptr)    list_entry((ptr), struct cfq_queue, cfq_list)
  70 #define list_entry_fifo(ptr)    list_entry((ptr), struct request, queuelist)
  71
  72 #define RQ_DATA(rq)             (rq)->elevator_private
  73
  74 /*
  75  * rb-tree defines
  76  */
  77 #define RB_NONE                 (2)
  78 #define RB_EMPTY(node)          ((node)->rb_node == NULL)
  79 #define RB_CLEAR_COLOR(node)    (node)->rb_color = RB_NONE
  80 #define RB_CLEAR(node)          do {    \
  81         (node)->rb_parent = NULL;       \
  82         RB_CLEAR_COLOR((node));         \
  83         (node)->rb_right = NULL;        \
  84         (node)->rb_left = NULL;         \
  85 } while (0)
  86 #define RB_CLEAR_ROOT(root)     ((root)->rb_node = NULL)
  87 #define rb_entry_crq(node)      rb_entry((node), struct cfq_rq, rb_node)
  88 #define rq_rb_key(rq)           (rq)->sector
  89
  90 static kmem_cache_t *crq_pool;
  91 static kmem_cache_t *cfq_pool;
  92 static kmem_cache_t *cfq_ioc_pool;
  93
  94 #define CFQ_PRIO_LISTS          IOPRIO_BE_NR
  95 #define cfq_class_idle(cfqq)    ((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
  96 #define cfq_class_be(cfqq)      ((cfqq)->ioprio_class == IOPRIO_CLASS_BE)
  97 #define cfq_class_rt(cfqq)      ((cfqq)->ioprio_class == IOPRIO_CLASS_RT)
  98
  99 #define ASYNC                   (0)
 100 #define SYNC                    (1)
 101
 102 #define cfq_cfqq_dispatched(cfqq)       \
 103         ((cfqq)->on_dispatch[ASYNC] + (cfqq)->on_dispatch[SYNC])
 104
 105 #define cfq_cfqq_class_sync(cfqq)       ((cfqq)->key != CFQ_KEY_ASYNC)
 106
 107 #define cfq_cfqq_sync(cfqq)             \
 108         (cfq_cfqq_class_sync(cfqq) || (cfqq)->on_dispatch[SYNC])
 109
 110 /*
 111  * Per block device queue structure
 112  */
 113 struct cfq_data {
 114         atomic_t ref;
 115         request_queue_t *queue;
 116
 117         /*
 118          * rr list of queues with requests and the count of them
 119          */
 120         struct list_head rr_list[CFQ_PRIO_LISTS];
 121         struct list_head busy_rr;
 122         struct list_head cur_rr;
 123         struct list_head idle_rr;
 124         unsigned int busy_queues;
 125
 126         /*
 127          * non-ordered list of empty cfqq's
 128          */
 129         struct list_head empty_list;
 130
 131         /*
 132          * cfqq lookup hash
 133          */
 134         struct hlist_head *cfq_hash;
 135
 136         /*
 137          * global crq hash for all queues
 138          */
 139         struct hlist_head *crq_hash;
 140
 141         unsigned int max_queued;
 142
 143         mempool_t *crq_pool;
 144
 145         int rq_in_driver;
 146
 147         /*
 148          * schedule slice state info
 149          */
 150         /*
 151          * idle window management
 152          */
 153         struct timer_list idle_slice_timer;
 154         struct work_struct unplug_work;
 155
 156         struct cfq_queue *active_queue;
 157         struct cfq_io_context *active_cic;
 158         int cur_prio, cur_end_prio;
 159         unsigned int dispatch_slice;
 160
 161         struct timer_list idle_class_timer;
 162
 163         sector_t last_sector;
 164         unsigned long last_end_request;
 165
 166         unsigned int rq_starved;
 167
 168         /*
 169          * tunables, see top of file
 170          */
 171         unsigned int cfq_quantum;
 172         unsigned int cfq_queued;
 173         unsigned int cfq_fifo_expire[2];
 174         unsigned int cfq_back_penalty;
 175         unsigned int cfq_back_max;
 176         unsigned int cfq_slice[2];
 177         unsigned int cfq_slice_async_rq;
 178         unsigned int cfq_slice_idle;
 179         unsigned int cfq_max_depth;
 180
 181         struct list_head cic_list;
 182 };
 183
 184 /*
 185  * Per process-grouping structure
 186  */
 187 struct cfq_queue {
 188         /* reference count */
 189         atomic_t ref;
 190         /* parent cfq_data */
 191         struct cfq_data *cfqd;
 192         /* cfqq lookup hash */
 193         struct hlist_node cfq_hash;
 194         /* hash key */
 195         unsigned int key;
 196         /* on either rr or empty list of cfqd */
 197         struct list_head cfq_list;
 198         /* sorted list of pending requests */
 199         struct rb_root sort_list;
 200         /* if fifo isn't expired, next request to serve */
 201         struct cfq_rq *next_crq;
 202         /* requests queued in sort_list */
 203         int queued[2];
 204         /* currently allocated requests */
 205         int allocated[2];
 206         /* fifo list of requests in sort_list */
 207         struct list_head fifo;
 208
 209         unsigned long slice_start;
 210         unsigned long slice_end;
 211         unsigned long slice_left;
 212         unsigned long service_last;
 213
 214         /* number of requests that are on the dispatch list */
 215         int on_dispatch[2];
 216
 217         /* io prio of this group */
 218         unsigned short ioprio, org_ioprio;
 219         unsigned short ioprio_class, org_ioprio_class;
 220
 221         /* various state flags, see below */
 222         unsigned int flags;
 223 };
 224
 225 struct cfq_rq {
 226         struct rb_node rb_node;
 227         sector_t rb_key;
 228         struct request *request;
 229         struct hlist_node hash;
 230
 231         struct cfq_queue *cfq_queue;
 232         struct cfq_io_context *io_context;
 233
 234         unsigned int crq_flags;
 235 };
 236
 237 enum cfqq_state_flags {
 238         CFQ_CFQQ_FLAG_on_rr = 0,
 239         CFQ_CFQQ_FLAG_wait_request,
 240         CFQ_CFQQ_FLAG_must_alloc,
 241         CFQ_CFQQ_FLAG_must_alloc_slice,
 242         CFQ_CFQQ_FLAG_must_dispatch,
 243         CFQ_CFQQ_FLAG_fifo_expire,
 244         CFQ_CFQQ_FLAG_idle_window,
 245         CFQ_CFQQ_FLAG_prio_changed,
 246 };
 247
 248 #define CFQ_CFQQ_FNS(name)                                              \
 249 static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)         \
 250 {                                                                       \
 251         cfqq->flags |= (1 << CFQ_CFQQ_FLAG_##name);                     \
 252 }                                                                       \
 253 static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)        \
 254 {                                                                       \
 255         cfqq->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);                    \
 256 }                                                                       \
 257 static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)         \
 258 {                                                                       \
 259         return (cfqq->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;        \
 260 }
 261
 262 CFQ_CFQQ_FNS(on_rr);
 263 CFQ_CFQQ_FNS(wait_request);
 264 CFQ_CFQQ_FNS(must_alloc);
 265 CFQ_CFQQ_FNS(must_alloc_slice);
 266 CFQ_CFQQ_FNS(must_dispatch);
 267 CFQ_CFQQ_FNS(fifo_expire);
 268 CFQ_CFQQ_FNS(idle_window);
 269 CFQ_CFQQ_FNS(prio_changed);
 270 #undef CFQ_CFQQ_FNS
 271
 272 enum cfq_rq_state_flags {
 273         CFQ_CRQ_FLAG_is_sync = 0,
 274 };
 275
 276 #define CFQ_CRQ_FNS(name)                                               \
 277 static inline void cfq_mark_crq_##name(struct cfq_rq *crq)              \
 278 {                                                                       \
 279         crq->crq_flags |= (1 << CFQ_CRQ_FLAG_##name);                   \
 280 }                                                                       \
 281 static inline void cfq_clear_crq_##name(struct cfq_rq *crq)             \
 282 {                                                                       \
 283         crq->crq_flags &= ~(1 << CFQ_CRQ_FLAG_##name);                  \
 284 }                                                                       \
 285 static inline int cfq_crq_##name(const struct cfq_rq *crq)              \
 286 {                                                                       \
 287         return (crq->crq_flags & (1 << CFQ_CRQ_FLAG_##name)) != 0;      \
 288 }
 289
 290 CFQ_CRQ_FNS(is_sync);
 291 #undef CFQ_CRQ_FNS
 292
 293 static struct cfq_queue *cfq_find_cfq_hash(struct cfq_data *, unsigned int, unsigned short);
 294 static void cfq_dispatch_insert(request_queue_t *, struct cfq_rq *);
 295 static void cfq_put_cfqd(struct cfq_data *cfqd);
 296
 297 #define process_sync(tsk)       ((tsk)->flags & PF_SYNCWRITE)
 298
 299 /*
 300  * lots of deadline iosched dupes, can be abstracted later...
 301  */
 302 static inline void cfq_del_crq_hash(struct cfq_rq *crq)
 303 {
 304         hlist_del_init(&crq->hash);
 305 }
 306
 307 static inline void cfq_add_crq_hash(struct cfq_data *cfqd, struct cfq_rq *crq)
 308 {
 309         const int hash_idx = CFQ_MHASH_FN(rq_hash_key(crq->request));
 310
 311         hlist_add_head(&crq->hash, &cfqd->crq_hash[hash_idx]);
 312 }
 313
 314 static struct request *cfq_find_rq_hash(struct cfq_data *cfqd, sector_t offset)
 315 {
 316         struct hlist_head *hash_list = &cfqd->crq_hash[CFQ_MHASH_FN(offset)];
 317         struct hlist_node *entry, *next;
 318
 319         hlist_for_each_safe(entry, next, hash_list) {
 320                 struct cfq_rq *crq = list_entry_hash(entry);
 321                 struct request *__rq = crq->request;
 322
 323                 if (!rq_mergeable(__rq)) {
 324                         cfq_del_crq_hash(crq);
 325                         continue;
 326                 }
 327
 328                 if (rq_hash_key(__rq) == offset)
 329                         return __rq;
 330         }
 331
 332         return NULL;
 333 }
 334
 335 /*
 336  * scheduler run of queue, if there are requests pending and no one in the
 337  * driver that will restart queueing
 338  */
 339 static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
 340 {
 341         if (cfqd->busy_queues)
 342                 kblockd_schedule_work(&cfqd->unplug_work);
 343 }
 344
 345 static int cfq_queue_empty(request_queue_t *q)
 346 {
 347         struct cfq_data *cfqd = q->elevator->elevator_data;
 348
 349         return !cfqd->busy_queues;
 350 }
 351
 352 /*
 353  * Lifted from AS - choose which of crq1 and crq2 that is best served now.
 354  * We choose the request that is closest to the head right now. Distance
 355  * behind the head are penalized and only allowed to a certain extent.
 356  */
 357 static struct cfq_rq *
 358 cfq_choose_req(struct cfq_data *cfqd, struct cfq_rq *crq1, struct cfq_rq *crq2)
 359 {
 360         sector_t last, s1, s2, d1 = 0, d2 = 0;
 361         int r1_wrap = 0, r2_wrap = 0;   /* requests are behind the disk head */
 362         unsigned long back_max;
 363
 364         if (crq1 == NULL || crq1 == crq2)
 365                 return crq2;
 366         if (crq2 == NULL)
 367                 return crq1;
 368
 369         if (cfq_crq_is_sync(crq1) && !cfq_crq_is_sync(crq2))
 370                 return crq1;
 371         else if (cfq_crq_is_sync(crq2) && !cfq_crq_is_sync(crq1))
 372                 return crq2;
 373
 374         s1 = crq1->request->sector;
 375         s2 = crq2->request->sector;
 376
 377         last = cfqd->last_sector;
 378
 379         /*
 380          * by definition, 1KiB is 2 sectors
 381          */
 382         back_max = cfqd->cfq_back_max * 2;
 383
 384         /*
 385          * Strict one way elevator _except_ in the case where we allow
 386          * short backward seeks which are biased as twice the cost of a
 387          * similar forward seek.
 388          */
 389         if (s1 >= last)
 390                 d1 = s1 - last;
 391         else if (s1 + back_max >= last)
 392                 d1 = (last - s1) * cfqd->cfq_back_penalty;
 393         else
 394                 r1_wrap = 1;
 395
 396         if (s2 >= last)
 397                 d2 = s2 - last;
 398         else if (s2 + back_max >= last)
 399                 d2 = (last - s2) * cfqd->cfq_back_penalty;
 400         else
 401                 r2_wrap = 1;
 402
 403         /* Found required data */
 404         if (!r1_wrap && r2_wrap)
 405                 return crq1;
 406         else if (!r2_wrap && r1_wrap)
 407                 return crq2;
 408         else if (r1_wrap && r2_wrap) {
 409                 /* both behind the head */
 410                 if (s1 <= s2)
 411                         return crq1;
 412                 else
 413                         return crq2;
 414         }
 415
 416         /* Both requests in front of the head */
 417         if (d1 < d2)
 418                 return crq1;
 419         else if (d2 < d1)
 420                 return crq2;
 421         else {
 422                 if (s1 >= s2)
 423                         return crq1;
 424                 else
 425                         return crq2;
 426         }
 427 }
 428
 429 /*
 430  * would be nice to take fifo expire time into account as well
 431  */
 432 static struct cfq_rq *
 433 cfq_find_next_crq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
 434                   struct cfq_rq *last)
 435 {
 436         struct cfq_rq *crq_next = NULL, *crq_prev = NULL;
 437         struct rb_node *rbnext, *rbprev;
 438
 439         if (!(rbnext = rb_next(&last->rb_node))) {
 440                 rbnext = rb_first(&cfqq->sort_list);
 441                 if (rbnext == &last->rb_node)
 442                         rbnext = NULL;
 443         }
 444
 445         rbprev = rb_prev(&last->rb_node);
 446
 447         if (rbprev)
 448                 crq_prev = rb_entry_crq(rbprev);
 449         if (rbnext)
 450                 crq_next = rb_entry_crq(rbnext);
 451
 452         return cfq_choose_req(cfqd, crq_next, crq_prev);
 453 }
 454
 455 static void cfq_update_next_crq(struct cfq_rq *crq)
 456 {
 457         struct cfq_queue *cfqq = crq->cfq_queue;
 458
 459         if (cfqq->next_crq == crq)
 460                 cfqq->next_crq = cfq_find_next_crq(cfqq->cfqd, cfqq, crq);
 461 }
 462
 463 static void cfq_resort_rr_list(struct cfq_queue *cfqq, int preempted)
 464 {
 465         struct cfq_data *cfqd = cfqq->cfqd;
 466         struct list_head *list, *entry;
 467
 468         BUG_ON(!cfq_cfqq_on_rr(cfqq));
 469
 470         list_del(&cfqq->cfq_list);
 471
 472         if (cfq_class_rt(cfqq))
 473                 list = &cfqd->cur_rr;
 474         else if (cfq_class_idle(cfqq))
 475                 list = &cfqd->idle_rr;
 476         else {
 477                 /*
 478                  * if cfqq has requests in flight, don't allow it to be
 479                  * found in cfq_set_active_queue before it has finished them.
 480                  * this is done to increase fairness between a process that
 481                  * has lots of io pending vs one that only generates one
 482                  * sporadically or synchronously
 483                  */
 484                 if (cfq_cfqq_dispatched(cfqq))
 485                         list = &cfqd->busy_rr;
 486                 else
 487                         list = &cfqd->rr_list[cfqq->ioprio];
 488         }
 489
 490         /*
 491          * if queue was preempted, just add to front to be fair. busy_rr
 492          * isn't sorted.
 493          */
 494         if (preempted || list == &cfqd->busy_rr) {
 495                 list_add(&cfqq->cfq_list, list);
 496                 return;
 497         }
 498
 499         /*
 500          * sort by when queue was last serviced
 501          */
 502         entry = list;
 503         while ((entry = entry->prev) != list) {
 504                 struct cfq_queue *__cfqq = list_entry_cfqq(entry);
 505
 506                 if (!__cfqq->service_last)
 507                         break;
 508                 if (time_before(__cfqq->service_last, cfqq->service_last))
 509                         break;
 510         }
 511
 512         list_add(&cfqq->cfq_list, entry);
 513 }
 514
 515 /*
 516  * add to busy list of queues for service, trying to be fair in ordering
 517  * the pending list according to last request service
 518  */
 519 static inline void
 520 cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 521 {
 522         BUG_ON(cfq_cfqq_on_rr(cfqq));
 523         cfq_mark_cfqq_on_rr(cfqq);
 524         cfqd->busy_queues++;
 525
 526         cfq_resort_rr_list(cfqq, 0);
 527 }
 528
 529 static inline void
 530 cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 531 {
 532         BUG_ON(!cfq_cfqq_on_rr(cfqq));
 533         cfq_clear_cfqq_on_rr(cfqq);
 534         list_move(&cfqq->cfq_list, &cfqd->empty_list);
 535
 536         BUG_ON(!cfqd->busy_queues);
 537         cfqd->busy_queues--;
 538 }
 539
 540 /*
 541  * rb tree support functions
 542  */
 543 static inline void cfq_del_crq_rb(struct cfq_rq *crq)
 544 {
 545         struct cfq_queue *cfqq = crq->cfq_queue;
 546         struct cfq_data *cfqd = cfqq->cfqd;
 547         const int sync = cfq_crq_is_sync(crq);
 548
 549         BUG_ON(!cfqq->queued[sync]);
 550         cfqq->queued[sync]--;
 551
 552         cfq_update_next_crq(crq);
 553
 554         rb_erase(&crq->rb_node, &cfqq->sort_list);
 555         RB_CLEAR_COLOR(&crq->rb_node);
 556
 557         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY(&cfqq->sort_list))
 558                 cfq_del_cfqq_rr(cfqd, cfqq);
 559 }
 560
 561 static struct cfq_rq *
 562 __cfq_add_crq_rb(struct cfq_rq *crq)
 563 {
 564         struct rb_node **p = &crq->cfq_queue->sort_list.rb_node;
 565         struct rb_node *parent = NULL;
 566         struct cfq_rq *__crq;
 567
 568         while (*p) {
 569                 parent = *p;
 570                 __crq = rb_entry_crq(parent);
 571
 572                 if (crq->rb_key < __crq->rb_key)
 573                         p = &(*p)->rb_left;
 574                 else if (crq->rb_key > __crq->rb_key)
 575                         p = &(*p)->rb_right;
 576                 else
 577                         return __crq;
 578         }
 579
 580         rb_link_node(&crq->rb_node, parent, p);
 581         return NULL;
 582 }
 583
 584 static void cfq_add_crq_rb(struct cfq_rq *crq)
 585 {
 586         struct cfq_queue *cfqq = crq->cfq_queue;
 587         struct cfq_data *cfqd = cfqq->cfqd;
 588         struct request *rq = crq->request;
 589         struct cfq_rq *__alias;
 590
 591         crq->rb_key = rq_rb_key(rq);
 592         cfqq->queued[cfq_crq_is_sync(crq)]++;
 593
 594         /*
 595          * looks a little odd, but the first insert might return an alias.
 596          * if that happens, put the alias on the dispatch list
 597          */
 598         while ((__alias = __cfq_add_crq_rb(crq)) != NULL)
 599                 cfq_dispatch_insert(cfqd->queue, __alias);
 600
 601         rb_insert_color(&crq->rb_node, &cfqq->sort_list);
 602
 603         if (!cfq_cfqq_on_rr(cfqq))
 604                 cfq_add_cfqq_rr(cfqd, cfqq);
 605
 606         /*
 607          * check if this request is a better next-serve candidate
 608          */
 609         cfqq->next_crq = cfq_choose_req(cfqd, cfqq->next_crq, crq);
 610 }
 611
 612 static inline void
 613 cfq_reposition_crq_rb(struct cfq_queue *cfqq, struct cfq_rq *crq)
 614 {
 615         rb_erase(&crq->rb_node, &cfqq->sort_list);
 616         cfqq->queued[cfq_crq_is_sync(crq)]--;
 617
 618         cfq_add_crq_rb(crq);
 619 }
 620
 621 static struct request *cfq_find_rq_rb(struct cfq_data *cfqd, sector_t sector)
 622
 623 {
 624         struct cfq_queue *cfqq = cfq_find_cfq_hash(cfqd, current->pid, CFQ_KEY_ANY);
 625         struct rb_node *n;
 626
 627         if (!cfqq)
 628                 goto out;
 629
 630         n = cfqq->sort_list.rb_node;
 631         while (n) {
 632                 struct cfq_rq *crq = rb_entry_crq(n);
 633
 634                 if (sector < crq->rb_key)
 635                         n = n->rb_left;
 636                 else if (sector > crq->rb_key)
 637                         n = n->rb_right;
 638                 else
 639                         return crq->request;
 640         }
 641
 642 out:
 643         return NULL;
 644 }
 645
 646 static void cfq_activate_request(request_queue_t *q, struct request *rq)
 647 {
 648         struct cfq_data *cfqd = q->elevator->elevator_data;
 649
 650         cfqd->rq_in_driver++;
 651 }
 652
 653 static void cfq_deactivate_request(request_queue_t *q, struct request *rq)
 654 {
 655         struct cfq_data *cfqd = q->elevator->elevator_data;
 656
 657         WARN_ON(!cfqd->rq_in_driver);
 658         cfqd->rq_in_driver--;
 659 }
 660
 661 static void cfq_remove_request(struct request *rq)
 662 {
 663         struct cfq_rq *crq = RQ_DATA(rq);
 664
 665         list_del_init(&rq->queuelist);
 666         cfq_del_crq_rb(crq);
 667         cfq_del_crq_hash(crq);
 668 }
 669
 670 static int
 671 cfq_merge(request_queue_t *q, struct request **req, struct bio *bio)
 672 {
 673         struct cfq_data *cfqd = q->elevator->elevator_data;
 674         struct request *__rq;
 675         int ret;
 676
 677         __rq = cfq_find_rq_hash(cfqd, bio->bi_sector);
 678         if (__rq && elv_rq_merge_ok(__rq, bio)) {
 679                 ret = ELEVATOR_BACK_MERGE;
 680                 goto out;
 681         }
 682
 683         __rq = cfq_find_rq_rb(cfqd, bio->bi_sector + bio_sectors(bio));
 684         if (__rq && elv_rq_merge_ok(__rq, bio)) {
 685                 ret = ELEVATOR_FRONT_MERGE;
 686                 goto out;
 687         }
 688
 689         return ELEVATOR_NO_MERGE;
 690 out:
 691         *req = __rq;
 692         return ret;
 693 }
 694
 695 static void cfq_merged_request(request_queue_t *q, struct request *req)
 696 {
 697         struct cfq_data *cfqd = q->elevator->elevator_data;
 698         struct cfq_rq *crq = RQ_DATA(req);
 699
 700         cfq_del_crq_hash(crq);
 701         cfq_add_crq_hash(cfqd, crq);
 702
 703         if (rq_rb_key(req) != crq->rb_key) {
 704                 struct cfq_queue *cfqq = crq->cfq_queue;
 705
 706                 cfq_update_next_crq(crq);
 707                 cfq_reposition_crq_rb(cfqq, crq);
 708         }
 709 }
 710
 711 static void
 712 cfq_merged_requests(request_queue_t *q, struct request *rq,
 713                     struct request *next)
 714 {
 715         cfq_merged_request(q, rq);
 716
 717         /*
 718          * reposition in fifo if next is older than rq
 719          */
 720         if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
 721             time_before(next->start_time, rq->start_time))
 722                 list_move(&rq->queuelist, &next->queuelist);
 723
 724         cfq_remove_request(next);
 725 }
 726
 727 static inline void
 728 __cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 729 {
 730         if (cfqq) {
 731                 /*
 732                  * stop potential idle class queues waiting service
 733                  */
 734                 del_timer(&cfqd->idle_class_timer);
 735
 736                 cfqq->slice_start = jiffies;
 737                 cfqq->slice_end = 0;
 738                 cfqq->slice_left = 0;
 739                 cfq_clear_cfqq_must_alloc_slice(cfqq);
 740                 cfq_clear_cfqq_fifo_expire(cfqq);
 741         }
 742
 743         cfqd->active_queue = cfqq;
 744 }
 745
 746 /*
 747  * current cfqq expired its slice (or was too idle), select new one
 748  */
 749 static void
 750 __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
 751                     int preempted)
 752 {
 753         unsigned long now = jiffies;
 754
 755         if (cfq_cfqq_wait_request(cfqq))
 756                 del_timer(&cfqd->idle_slice_timer);
 757
 758         if (!preempted && !cfq_cfqq_dispatched(cfqq)) {
 759                 cfqq->service_last = now;
 760                 cfq_schedule_dispatch(cfqd);
 761         }
 762
 763         cfq_clear_cfqq_must_dispatch(cfqq);
 764         cfq_clear_cfqq_wait_request(cfqq);
 765
 766         /*
 767          * store what was left of this slice, if the queue idled out
 768          * or was preempted
 769          */
 770         if (time_after(cfqq->slice_end, now))
 771                 cfqq->slice_left = cfqq->slice_end - now;
 772         else
 773                 cfqq->slice_left = 0;
 774
 775         if (cfq_cfqq_on_rr(cfqq))
 776                 cfq_resort_rr_list(cfqq, preempted);
 777
 778         if (cfqq == cfqd->active_queue)
 779                 cfqd->active_queue = NULL;
 780
 781         if (cfqd->active_cic) {
 782                 put_io_context(cfqd->active_cic->ioc);
 783                 cfqd->active_cic = NULL;
 784         }
 785
 786         cfqd->dispatch_slice = 0;
 787 }
 788
 789 static inline void cfq_slice_expired(struct cfq_data *cfqd, int preempted)
 790 {
 791         struct cfq_queue *cfqq = cfqd->active_queue;
 792
 793         if (cfqq)
 794                 __cfq_slice_expired(cfqd, cfqq, preempted);
 795 }
 796
 797 /*
 798  * 0
 799  * 0,1
 800  * 0,1,2
 801  * 0,1,2,3
 802  * 0,1,2,3,4
 803  * 0,1,2,3,4,5
 804  * 0,1,2,3,4,5,6
 805  * 0,1,2,3,4,5,6,7
 806  */
 807 static int cfq_get_next_prio_level(struct cfq_data *cfqd)
 808 {
 809         int prio, wrap;
 810
 811         prio = -1;
 812         wrap = 0;
 813         do {
 814                 int p;
 815
 816                 for (p = cfqd->cur_prio; p <= cfqd->cur_end_prio; p++) {
 817                         if (!list_empty(&cfqd->rr_list[p])) {
 818                                 prio = p;
 819                                 break;
 820                         }
 821                 }
 822
 823                 if (prio != -1)
 824                         break;
 825                 cfqd->cur_prio = 0;
 826                 if (++cfqd->cur_end_prio == CFQ_PRIO_LISTS) {
 827                         cfqd->cur_end_prio = 0;
 828                         if (wrap)
 829                                 break;
 830                         wrap = 1;
 831                 }
 832         } while (1);
 833
 834         if (unlikely(prio == -1))
 835                 return -1;
 836
 837         BUG_ON(prio >= CFQ_PRIO_LISTS);
 838
 839         list_splice_init(&cfqd->rr_list[prio], &cfqd->cur_rr);
 840
 841         cfqd->cur_prio = prio + 1;
 842         if (cfqd->cur_prio > cfqd->cur_end_prio) {
 843                 cfqd->cur_end_prio = cfqd->cur_prio;
 844                 cfqd->cur_prio = 0;
 845         }
 846         if (cfqd->cur_end_prio == CFQ_PRIO_LISTS) {
 847                 cfqd->cur_prio = 0;
 848                 cfqd->cur_end_prio = 0;
 849         }
 850
 851         return prio;
 852 }
 853
 854 static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd)
 855 {
 856         struct cfq_queue *cfqq = NULL;
 857
 858         /*
 859          * if current list is non-empty, grab first entry. if it is empty,
 860          * get next prio level and grab first entry then if any are spliced
 861          */
 862         if (!list_empty(&cfqd->cur_rr) || cfq_get_next_prio_level(cfqd) != -1)
 863                 cfqq = list_entry_cfqq(cfqd->cur_rr.next);
 864
 865         /*
 866          * if we have idle queues and no rt or be queues had pending
 867          * requests, either allow immediate service if the grace period
 868          * has passed or arm the idle grace timer
 869          */
 870         if (!cfqq && !list_empty(&cfqd->idle_rr)) {
 871                 unsigned long end = cfqd->last_end_request + CFQ_IDLE_GRACE;
 872
 873                 if (time_after_eq(jiffies, end))
 874                         cfqq = list_entry_cfqq(cfqd->idle_rr.next);
 875                 else
 876                         mod_timer(&cfqd->idle_class_timer, end);
 877         }
 878
 879         __cfq_set_active_queue(cfqd, cfqq);
 880         return cfqq;
 881 }
 882
 883 static int cfq_arm_slice_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 884
 885 {
 886         unsigned long sl;
 887
 888         WARN_ON(!RB_EMPTY(&cfqq->sort_list));
 889         WARN_ON(cfqq != cfqd->active_queue);
 890
 891         /*
 892          * idle is disabled, either manually or by past process history
 893          */
 894         if (!cfqd->cfq_slice_idle)
 895                 return 0;
 896         if (!cfq_cfqq_idle_window(cfqq))
 897                 return 0;
 898         /*
 899          * task has exited, don't wait
 900          */
 901         if (cfqd->active_cic && !cfqd->active_cic->ioc->task)
 902                 return 0;
 903
 904         cfq_mark_cfqq_must_dispatch(cfqq);
 905         cfq_mark_cfqq_wait_request(cfqq);
 906
 907         sl = min(cfqq->slice_end - 1, (unsigned long) cfqd->cfq_slice_idle);
 908         mod_timer(&cfqd->idle_slice_timer, jiffies + sl);
 909         return 1;
 910 }
 911
 912 static void cfq_dispatch_insert(request_queue_t *q, struct cfq_rq *crq)
 913 {
 914         struct cfq_data *cfqd = q->elevator->elevator_data;
 915         struct cfq_queue *cfqq = crq->cfq_queue;
 916
 917         cfqq->next_crq = cfq_find_next_crq(cfqd, cfqq, crq);
 918         cfq_remove_request(crq->request);
 919         cfqq->on_dispatch[cfq_crq_is_sync(crq)]++;
 920         elv_dispatch_sort(q, crq->request);
 921 }
 922
 923 /*
 924  * return expired entry, or NULL to just start from scratch in rbtree
 925  */
 926 static inline struct cfq_rq *cfq_check_fifo(struct cfq_queue *cfqq)
 927 {
 928         struct cfq_data *cfqd = cfqq->cfqd;
 929         struct request *rq;
 930         struct cfq_rq *crq;
 931
 932         if (cfq_cfqq_fifo_expire(cfqq))
 933                 return NULL;
 934
 935         if (!list_empty(&cfqq->fifo)) {
 936                 int fifo = cfq_cfqq_class_sync(cfqq);
 937
 938                 crq = RQ_DATA(list_entry_fifo(cfqq->fifo.next));
 939                 rq = crq->request;
 940                 if (time_after(jiffies, rq->start_time + cfqd->cfq_fifo_expire[fifo])) {
 941                         cfq_mark_cfqq_fifo_expire(cfqq);
 942                         return crq;
 943                 }
 944         }
 945
 946         return NULL;
 947 }
 948
 949 /*
 950  * Scale schedule slice based on io priority. Use the sync time slice only
 951  * if a queue is marked sync and has sync io queued. A sync queue with async
 952  * io only, should not get full sync slice length.
 953  */
 954 static inline int
 955 cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 956 {
 957         const int base_slice = cfqd->cfq_slice[cfq_cfqq_sync(cfqq)];
 958
 959         WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
 960
 961         return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - cfqq->ioprio));
 962 }
 963
 964 static inline void
 965 cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 966 {
 967         cfqq->slice_end = cfq_prio_to_slice(cfqd, cfqq) + jiffies;
 968 }
 969
 970 static inline int
 971 cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 972 {
 973         const int base_rq = cfqd->cfq_slice_async_rq;
 974
 975         WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
 976
 977         return 2 * (base_rq + base_rq * (CFQ_PRIO_LISTS - 1 - cfqq->ioprio));
 978 }
 979
 980 /*
 981  * get next queue for service
 982  */
 983 static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
 984 {
 985         unsigned long now = jiffies;
 986         struct cfq_queue *cfqq;
 987
 988         cfqq = cfqd->active_queue;
 989         if (!cfqq)
 990                 goto new_queue;
 991
 992         /*
 993          * slice has expired
 994          */
 995         if (!cfq_cfqq_must_dispatch(cfqq) && time_after(now, cfqq->slice_end))
 996                 goto expire;
 997
 998         /*
 999          * if queue has requests, dispatch one. if not, check if
1000          * enough slice is left to wait for one
1001          */
1002         if (!RB_EMPTY(&cfqq->sort_list))
1003                 goto keep_queue;
1004         else if (cfq_cfqq_class_sync(cfqq) &&
1005                  time_before(now, cfqq->slice_end)) {
1006                 if (cfq_arm_slice_timer(cfqd, cfqq))
1007                         return NULL;
1008         }
1009
1010 expire:
1011         cfq_slice_expired(cfqd, 0);
1012 new_queue:
1013         cfqq = cfq_set_active_queue(cfqd);
1014 keep_queue:
1015         return cfqq;
1016 }
1017
1018 static int
1019 __cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1020                         int max_dispatch)
1021 {
1022         int dispatched = 0;
1023
1024         BUG_ON(RB_EMPTY(&cfqq->sort_list));
1025
1026         do {
1027                 struct cfq_rq *crq;
1028
1029                 /*
1030                  * follow expired path, else get first next available
1031                  */
1032                 if ((crq = cfq_check_fifo(cfqq)) == NULL)
1033                         crq = cfqq->next_crq;
1034
1035                 /*
1036                  * finally, insert request into driver dispatch list
1037                  */
1038                 cfq_dispatch_insert(cfqd->queue, crq);
1039
1040                 cfqd->dispatch_slice++;
1041                 dispatched++;
1042
1043                 if (!cfqd->active_cic) {
1044                         atomic_inc(&crq->io_context->ioc->refcount);
1045                         cfqd->active_cic = crq->io_context;
1046                 }
1047
1048                 if (RB_EMPTY(&cfqq->sort_list))
1049                         break;
1050
1051         } while (dispatched < max_dispatch);
1052
1053         /*
1054          * if slice end isn't set yet, set it. if at least one request was
1055          * sync, use the sync time slice value
1056          */
1057         if (!cfqq->slice_end)
1058                 cfq_set_prio_slice(cfqd, cfqq);
1059
1060         /*
1061          * expire an async queue immediately if it has used up its slice. idle
1062          * queue always expire after 1 dispatch round.
1063          */
1064         if ((!cfq_cfqq_sync(cfqq) &&
1065             cfqd->dispatch_slice >= cfq_prio_to_maxrq(cfqd, cfqq)) ||
1066             cfq_class_idle(cfqq))
1067                 cfq_slice_expired(cfqd, 0);
1068
1069         return dispatched;
1070 }
1071
1072 static int
1073 cfq_forced_dispatch_cfqqs(struct list_head *list)
1074 {
1075         int dispatched = 0;
1076         struct cfq_queue *cfqq, *next;
1077         struct cfq_rq *crq;
1078
1079         list_for_each_entry_safe(cfqq, next, list, cfq_list) {
1080                 while ((crq = cfqq->next_crq)) {
1081                         cfq_dispatch_insert(cfqq->cfqd->queue, crq);
1082                         dispatched++;
1083                 }
1084                 BUG_ON(!list_empty(&cfqq->fifo));
1085         }
1086         return dispatched;
1087 }
1088
1089 static int
1090 cfq_forced_dispatch(struct cfq_data *cfqd)
1091 {
1092         int i, dispatched = 0;
1093
1094         for (i = 0; i < CFQ_PRIO_LISTS; i++)
1095                 dispatched += cfq_forced_dispatch_cfqqs(&cfqd->rr_list[i]);
1096
1097         dispatched += cfq_forced_dispatch_cfqqs(&cfqd->busy_rr);
1098         dispatched += cfq_forced_dispatch_cfqqs(&cfqd->cur_rr);
1099         dispatched += cfq_forced_dispatch_cfqqs(&cfqd->idle_rr);
1100
1101         cfq_slice_expired(cfqd, 0);
1102
1103         BUG_ON(cfqd->busy_queues);
1104
1105         return dispatched;
1106 }
1107
1108 static int
1109 cfq_dispatch_requests(request_queue_t *q, int force)
1110 {
1111         struct cfq_data *cfqd = q->elevator->elevator_data;
1112         struct cfq_queue *cfqq;
1113
1114         if (!cfqd->busy_queues)
1115                 return 0;
1116
1117         if (unlikely(force))
1118                 return cfq_forced_dispatch(cfqd);
1119
1120         cfqq = cfq_select_queue(cfqd);
1121         if (cfqq) {
1122                 int max_dispatch;
1123
1124                 /*
1125                  * if idle window is disabled, allow queue buildup
1126                  */
1127                 if (!cfq_cfqq_idle_window(cfqq) &&
1128                     cfqd->rq_in_driver >= cfqd->cfq_max_depth)
1129                         return 0;
1130
1131                 cfq_clear_cfqq_must_dispatch(cfqq);
1132                 cfq_clear_cfqq_wait_request(cfqq);
1133                 del_timer(&cfqd->idle_slice_timer);
1134
1135                 max_dispatch = cfqd->cfq_quantum;
1136                 if (cfq_class_idle(cfqq))
1137                         max_dispatch = 1;
1138
1139                 return __cfq_dispatch_requests(cfqd, cfqq, max_dispatch);
1140         }
1141
1142         return 0;
1143 }
1144
1145 /*
1146  * task holds one reference to the queue, dropped when task exits. each crq
1147  * in-flight on this queue also holds a reference, dropped when crq is freed.
1148  *
1149  * queue lock must be held here.
1150  */
1151 static void cfq_put_queue(struct cfq_queue *cfqq)
1152 {
1153         struct cfq_data *cfqd = cfqq->cfqd;
1154
1155         BUG_ON(atomic_read(&cfqq->ref) <= 0);
1156
1157         if (!atomic_dec_and_test(&cfqq->ref))
1158                 return;
1159
1160         BUG_ON(rb_first(&cfqq->sort_list));
1161         BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
1162         BUG_ON(cfq_cfqq_on_rr(cfqq));
1163
1164         if (unlikely(cfqd->active_queue == cfqq))
1165                 __cfq_slice_expired(cfqd, cfqq, 0);
1166
1167         cfq_put_cfqd(cfqq->cfqd);
1168
1169         /*
1170          * it's on the empty list and still hashed
1171          */
1172         list_del(&cfqq->cfq_list);
1173         hlist_del(&cfqq->cfq_hash);
1174         kmem_cache_free(cfq_pool, cfqq);
1175 }
1176
1177 static inline struct cfq_queue *
1178 __cfq_find_cfq_hash(struct cfq_data *cfqd, unsigned int key, unsigned int prio,
1179                     const int hashval)
1180 {
1181         struct hlist_head *hash_list = &cfqd->cfq_hash[hashval];
1182         struct hlist_node *entry, *next;
1183
1184         hlist_for_each_safe(entry, next, hash_list) {
1185                 struct cfq_queue *__cfqq = list_entry_qhash(entry);
1186                 const unsigned short __p = IOPRIO_PRIO_VALUE(__cfqq->org_ioprio_class, __cfqq->org_ioprio);
1187
1188                 if (__cfqq->key == key && (__p == prio || prio == CFQ_KEY_ANY))
1189                         return __cfqq;
1190         }
1191
1192         return NULL;
1193 }
1194
1195 static struct cfq_queue *
1196 cfq_find_cfq_hash(struct cfq_data *cfqd, unsigned int key, unsigned short prio)
1197 {
1198         return __cfq_find_cfq_hash(cfqd, key, prio, hash_long(key, CFQ_QHASH_SHIFT));
1199 }
1200
1201 static void cfq_free_io_context(struct cfq_io_context *cic)
1202 {
1203         struct cfq_io_context *__cic;
1204         struct list_head *entry, *next;
1205
1206         list_for_each_safe(entry, next, &cic->list) {
1207                 __cic = list_entry(entry, struct cfq_io_context, list);
1208                 kmem_cache_free(cfq_ioc_pool, __cic);
1209         }
1210
1211         kmem_cache_free(cfq_ioc_pool, cic);
1212 }
1213
1214 /*
1215  * Called with interrupts disabled
1216  */
1217 static void cfq_exit_single_io_context(struct cfq_io_context *cic)
1218 {
1219         struct cfq_data *cfqd = cic->key;
1220         request_queue_t *q;
1221
1222         if (!cfqd)
1223                 return;
1224
1225         q = cfqd->queue;
1226
1227         WARN_ON(!irqs_disabled());
1228
1229         spin_lock(q->queue_lock);
1230
1231         if (cic->cfqq[ASYNC]) {
1232                 if (unlikely(cic->cfqq[ASYNC] == cfqd->active_queue))
1233                         __cfq_slice_expired(cfqd, cic->cfqq[ASYNC], 0);
1234                 cfq_put_queue(cic->cfqq[ASYNC]);
1235                 cic->cfqq[ASYNC] = NULL;
1236         }
1237
1238         if (cic->cfqq[SYNC]) {
1239                 if (unlikely(cic->cfqq[SYNC] == cfqd->active_queue))
1240                         __cfq_slice_expired(cfqd, cic->cfqq[SYNC], 0);
1241                 cfq_put_queue(cic->cfqq[SYNC]);
1242                 cic->cfqq[SYNC] = NULL;
1243         }
1244
1245         cic->key = NULL;
1246         list_del_init(&cic->queue_list);
1247         spin_unlock(q->queue_lock);
1248 }
1249
1250 /*
1251  * Another task may update the task cic list, if it is doing a queue lookup
1252  * on its behalf. cfq_cic_lock excludes such concurrent updates
1253  */
1254 static void cfq_exit_io_context(struct cfq_io_context *cic)
1255 {
1256         struct cfq_io_context *__cic;
1257         struct list_head *entry;
1258         unsigned long flags;
1259
1260         local_irq_save(flags);
1261
1262         /*
1263          * put the reference this task is holding to the various queues
1264          */
1265         read_lock(&cfq_exit_lock);
1266         list_for_each(entry, &cic->list) {
1267                 __cic = list_entry(entry, struct cfq_io_context, list);
1268                 cfq_exit_single_io_context(__cic);
1269         }
1270
1271         cfq_exit_single_io_context(cic);
1272         read_unlock(&cfq_exit_lock);
1273         local_irq_restore(flags);
1274 }
1275
1276 static struct cfq_io_context *
1277 cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
1278 {
1279         struct cfq_io_context *cic = kmem_cache_alloc(cfq_ioc_pool, gfp_mask);
1280
1281         if (cic) {
1282                 INIT_LIST_HEAD(&cic->list);
1283                 cic->cfqq[ASYNC] = NULL;
1284                 cic->cfqq[SYNC] = NULL;
1285                 cic->key = NULL;
1286                 cic->last_end_request = jiffies;
1287                 cic->ttime_total = 0;
1288                 cic->ttime_samples = 0;
1289                 cic->ttime_mean = 0;
1290                 cic->dtor = cfq_free_io_context;
1291                 cic->exit = cfq_exit_io_context;
1292                 INIT_LIST_HEAD(&cic->queue_list);
1293         }
1294
1295         return cic;
1296 }
1297
1298 static void cfq_init_prio_data(struct cfq_queue *cfqq)
1299 {
1300         struct task_struct *tsk = current;
1301         int ioprio_class;
1302
1303         if (!cfq_cfqq_prio_changed(cfqq))
1304                 return;
1305
1306         ioprio_class = IOPRIO_PRIO_CLASS(tsk->ioprio);
1307         switch (ioprio_class) {
1308                 default:
1309                         printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
1310                 case IOPRIO_CLASS_NONE:
1311                         /*
1312                          * no prio set, place us in the middle of the BE classes
1313                          */
1314                         cfqq->ioprio = task_nice_ioprio(tsk);
1315                         cfqq->ioprio_class = IOPRIO_CLASS_BE;
1316                         break;
1317                 case IOPRIO_CLASS_RT:
1318                         cfqq->ioprio = task_ioprio(tsk);
1319                         cfqq->ioprio_class = IOPRIO_CLASS_RT;
1320                         break;
1321                 case IOPRIO_CLASS_BE:
1322                         cfqq->ioprio = task_ioprio(tsk);
1323                         cfqq->ioprio_class = IOPRIO_CLASS_BE;
1324                         break;
1325                 case IOPRIO_CLASS_IDLE:
1326                         cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
1327                         cfqq->ioprio = 7;
1328                         cfq_clear_cfqq_idle_window(cfqq);
1329                         break;
1330         }
1331
1332         /*
1333          * keep track of original prio settings in case we have to temporarily
1334          * elevate the priority of this queue
1335          */
1336         cfqq->org_ioprio = cfqq->ioprio;
1337         cfqq->org_ioprio_class = cfqq->ioprio_class;
1338
1339         if (cfq_cfqq_on_rr(cfqq))
1340                 cfq_resort_rr_list(cfqq, 0);
1341
1342         cfq_clear_cfqq_prio_changed(cfqq);
1343 }
1344
1345 static inline void changed_ioprio(struct cfq_io_context *cic)
1346 {
1347         struct cfq_data *cfqd = cic->key;
1348         struct cfq_queue *cfqq;
1349         if (cfqd) {
1350                 spin_lock(cfqd->queue->queue_lock);
1351                 cfqq = cic->cfqq[ASYNC];
1352                 if (cfqq) {
1353                         cfq_mark_cfqq_prio_changed(cfqq);
1354                         cfq_init_prio_data(cfqq);
1355                 }
1356                 cfqq = cic->cfqq[SYNC];
1357                 if (cfqq) {
1358                         cfq_mark_cfqq_prio_changed(cfqq);
1359                         cfq_init_prio_data(cfqq);
1360                 }
1361                 spin_unlock(cfqd->queue->queue_lock);
1362         }
1363 }
1364
1365 /*
1366  * callback from sys_ioprio_set, irqs are disabled
1367  */
1368 static int cfq_ioc_set_ioprio(struct io_context *ioc, unsigned int ioprio)
1369 {
1370         struct cfq_io_context *cic;
1371
1372         write_lock(&cfq_exit_lock);
1373
1374         cic = ioc->cic;
1375
1376         changed_ioprio(cic);
1377
1378         list_for_each_entry(cic, &cic->list, list)
1379                 changed_ioprio(cic);
1380
1381         write_unlock(&cfq_exit_lock);
1382
1383         return 0;
1384 }
1385
1386 static struct cfq_queue *
1387 cfq_get_queue(struct cfq_data *cfqd, unsigned int key, unsigned short ioprio,
1388               gfp_t gfp_mask)
1389 {
1390         const int hashval = hash_long(key, CFQ_QHASH_SHIFT);
1391         struct cfq_queue *cfqq, *new_cfqq = NULL;
1392
1393 retry:
1394         cfqq = __cfq_find_cfq_hash(cfqd, key, ioprio, hashval);
1395
1396         if (!cfqq) {
1397                 if (new_cfqq) {
1398                         cfqq = new_cfqq;
1399                         new_cfqq = NULL;
1400                 } else if (gfp_mask & __GFP_WAIT) {
1401                         spin_unlock_irq(cfqd->queue->queue_lock);
1402                         new_cfqq = kmem_cache_alloc(cfq_pool, gfp_mask);
1403                         spin_lock_irq(cfqd->queue->queue_lock);
1404                         goto retry;
1405                 } else {
1406                         cfqq = kmem_cache_alloc(cfq_pool, gfp_mask);
1407                         if (!cfqq)
1408                                 goto out;
1409                 }
1410
1411                 memset(cfqq, 0, sizeof(*cfqq));
1412
1413                 INIT_HLIST_NODE(&cfqq->cfq_hash);
1414                 INIT_LIST_HEAD(&cfqq->cfq_list);
1415                 RB_CLEAR_ROOT(&cfqq->sort_list);
1416                 INIT_LIST_HEAD(&cfqq->fifo);
1417
1418                 cfqq->key = key;
1419                 hlist_add_head(&cfqq->cfq_hash, &cfqd->cfq_hash[hashval]);
1420                 atomic_set(&cfqq->ref, 0);
1421                 cfqq->cfqd = cfqd;
1422                 atomic_inc(&cfqd->ref);
1423                 cfqq->service_last = 0;
1424                 /*
1425                  * set ->slice_left to allow preemption for a new process
1426                  */
1427                 cfqq->slice_left = 2 * cfqd->cfq_slice_idle;
1428                 cfq_mark_cfqq_idle_window(cfqq);
1429                 cfq_mark_cfqq_prio_changed(cfqq);
1430                 cfq_init_prio_data(cfqq);
1431         }
1432
1433         if (new_cfqq)
1434                 kmem_cache_free(cfq_pool, new_cfqq);
1435
1436         atomic_inc(&cfqq->ref);
1437 out:
1438         WARN_ON((gfp_mask & __GFP_WAIT) && !cfqq);
1439         return cfqq;
1440 }
1441
1442 /*
1443  * Setup general io context and cfq io context. There can be several cfq
1444  * io contexts per general io context, if this process is doing io to more
1445  * than one device managed by cfq. Note that caller is holding a reference to
1446  * cfqq, so we don't need to worry about it disappearing
1447  */
1448 static struct cfq_io_context *
1449 cfq_get_io_context(struct cfq_data *cfqd, pid_t pid, gfp_t gfp_mask)
1450 {
1451         struct io_context *ioc = NULL;
1452         struct cfq_io_context *cic;
1453
1454         might_sleep_if(gfp_mask & __GFP_WAIT);
1455
1456         ioc = get_io_context(gfp_mask);
1457         if (!ioc)
1458                 return NULL;
1459
1460 restart:
1461         if ((cic = ioc->cic) == NULL) {
1462                 cic = cfq_alloc_io_context(cfqd, gfp_mask);
1463
1464                 if (cic == NULL)
1465                         goto err;
1466
1467                 /*
1468                  * manually increment generic io_context usage count, it
1469                  * cannot go away since we are already holding one ref to it
1470                  */
1471                 cic->ioc = ioc;
1472                 cic->key = cfqd;
1473                 read_lock(&cfq_exit_lock);
1474                 ioc->set_ioprio = cfq_ioc_set_ioprio;
1475                 ioc->cic = cic;
1476                 list_add(&cic->queue_list, &cfqd->cic_list);
1477                 read_unlock(&cfq_exit_lock);
1478         } else {
1479                 struct cfq_io_context *__cic;
1480
1481                 /*
1482                  * the first cic on the list is actually the head itself
1483                  */
1484                 if (cic->key == cfqd)
1485                         goto out;
1486
1487                 if (unlikely(!cic->key)) {
1488                         read_lock(&cfq_exit_lock);
1489                         if (list_empty(&cic->list))
1490                                 ioc->cic = NULL;
1491                         else
1492                                 ioc->cic = list_entry(cic->list.next,
1493                                                       struct cfq_io_context,
1494                                                       list);
1495                         read_unlock(&cfq_exit_lock);
1496                         kmem_cache_free(cfq_ioc_pool, cic);
1497                         goto restart;
1498                 }
1499
1500                 /*
1501                  * cic exists, check if we already are there. linear search
1502                  * should be ok here, the list will usually not be more than
1503                  * 1 or a few entries long
1504                  */
1505                 list_for_each_entry(__cic, &cic->list, list) {
1506                         /*
1507                          * this process is already holding a reference to
1508                          * this queue, so no need to get one more
1509                          */
1510                         if (__cic->key == cfqd) {
1511                                 cic = __cic;
1512                                 goto out;
1513                         }
1514                         if (unlikely(!__cic->key)) {
1515                                 read_lock(&cfq_exit_lock);
1516                                 list_del(&__cic->list);
1517                                 read_unlock(&cfq_exit_lock);
1518                                 kmem_cache_free(cfq_ioc_pool, __cic);
1519                                 goto restart;
1520                         }
1521                 }
1522
1523                 /*
1524                  * nope, process doesn't have a cic assoicated with this
1525                  * cfqq yet. get a new one and add to list
1526                  */
1527                 __cic = cfq_alloc_io_context(cfqd, gfp_mask);
1528                 if (__cic == NULL)
1529                         goto err;
1530
1531                 __cic->ioc = ioc;
1532                 __cic->key = cfqd;
1533                 read_lock(&cfq_exit_lock);
1534                 list_add(&__cic->list, &cic->list);
1535                 list_add(&__cic->queue_list, &cfqd->cic_list);
1536                 read_unlock(&cfq_exit_lock);
1537                 cic = __cic;
1538         }
1539
1540 out:
1541         return cic;
1542 err:
1543         put_io_context(ioc);
1544         return NULL;
1545 }
1546
1547 static void
1548 cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_io_context *cic)
1549 {
1550         unsigned long elapsed, ttime;
1551
1552         /*
1553          * if this context already has stuff queued, thinktime is from
1554          * last queue not last end
1555          */
1556 #if 0
1557         if (time_after(cic->last_end_request, cic->last_queue))
1558                 elapsed = jiffies - cic->last_end_request;
1559         else
1560                 elapsed = jiffies - cic->last_queue;
1561 #else
1562                 elapsed = jiffies - cic->last_end_request;
1563 #endif
1564
1565         ttime = min(elapsed, 2UL * cfqd->cfq_slice_idle);
1566
1567         cic->ttime_samples = (7*cic->ttime_samples + 256) / 8;
1568         cic->ttime_total = (7*cic->ttime_total + 256*ttime) / 8;
1569         cic->ttime_mean = (cic->ttime_total + 128) / cic->ttime_samples;
1570 }
1571
1572 #define sample_valid(samples)   ((samples) > 80)
1573
1574 /*
1575  * Disable idle window if the process thinks too long or seeks so much that
1576  * it doesn't matter
1577  */
1578 static void
1579 cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1580                        struct cfq_io_context *cic)
1581 {
1582         int enable_idle = cfq_cfqq_idle_window(cfqq);
1583
1584         if (!cic->ioc->task || !cfqd->cfq_slice_idle)
1585                 enable_idle = 0;
1586         else if (sample_valid(cic->ttime_samples)) {
1587                 if (cic->ttime_mean > cfqd->cfq_slice_idle)
1588                         enable_idle = 0;
1589                 else
1590                         enable_idle = 1;
1591         }
1592
1593         if (enable_idle)
1594                 cfq_mark_cfqq_idle_window(cfqq);
1595         else
1596                 cfq_clear_cfqq_idle_window(cfqq);
1597 }
1598
1599
1600 /*
1601  * Check if new_cfqq should preempt the currently active queue. Return 0 for
1602  * no or if we aren't sure, a 1 will cause a preempt.
1603  */
1604 static int
1605 cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
1606                    struct cfq_rq *crq)
1607 {
1608         struct cfq_queue *cfqq = cfqd->active_queue;
1609
1610         if (cfq_class_idle(new_cfqq))
1611                 return 0;
1612
1613         if (!cfqq)
1614                 return 1;
1615
1616         if (cfq_class_idle(cfqq))
1617                 return 1;
1618         if (!cfq_cfqq_wait_request(new_cfqq))
1619                 return 0;
1620         /*
1621          * if it doesn't have slice left, forget it
1622          */
1623         if (new_cfqq->slice_left < cfqd->cfq_slice_idle)
1624                 return 0;
1625         if (cfq_crq_is_sync(crq) && !cfq_cfqq_sync(cfqq))
1626                 return 1;
1627
1628         return 0;
1629 }
1630
1631 /*
1632  * cfqq preempts the active queue. if we allowed preempt with no slice left,
1633  * let it have half of its nominal slice.
1634  */
1635 static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1636 {
1637         struct cfq_queue *__cfqq, *next;
1638
1639         list_for_each_entry_safe(__cfqq, next, &cfqd->cur_rr, cfq_list)
1640                 cfq_resort_rr_list(__cfqq, 1);
1641
1642         if (!cfqq->slice_left)
1643                 cfqq->slice_left = cfq_prio_to_slice(cfqd, cfqq) / 2;
1644
1645         cfqq->slice_end = cfqq->slice_left + jiffies;
1646         __cfq_slice_expired(cfqd, cfqq, 1);
1647         __cfq_set_active_queue(cfqd, cfqq);
1648 }
1649
1650 /*
1651  * should really be a ll_rw_blk.c helper
1652  */
1653 static void cfq_start_queueing(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1654 {
1655         request_queue_t *q = cfqd->queue;
1656
1657         if (!blk_queue_plugged(q))
1658                 q->request_fn(q);
1659         else
1660                 __generic_unplug_device(q);
1661 }
1662
1663 /*
1664  * Called when a new fs request (crq) is added (to cfqq). Check if there's
1665  * something we should do about it
1666  */
1667 static void
1668 cfq_crq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1669                  struct cfq_rq *crq)
1670 {
1671         struct cfq_io_context *cic;
1672
1673         cfqq->next_crq = cfq_choose_req(cfqd, cfqq->next_crq, crq);
1674
1675         /*
1676          * we never wait for an async request and we don't allow preemption
1677          * of an async request. so just return early
1678          */
1679         if (!cfq_crq_is_sync(crq))
1680                 return;
1681
1682         cic = crq->io_context;
1683
1684         cfq_update_io_thinktime(cfqd, cic);
1685         cfq_update_idle_window(cfqd, cfqq, cic);
1686
1687         cic->last_queue = jiffies;
1688
1689         if (cfqq == cfqd->active_queue) {
1690                 /*
1691                  * if we are waiting for a request for this queue, let it rip
1692                  * immediately and flag that we must not expire this queue
1693                  * just now
1694                  */
1695                 if (cfq_cfqq_wait_request(cfqq)) {
1696                         cfq_mark_cfqq_must_dispatch(cfqq);
1697                         del_timer(&cfqd->idle_slice_timer);
1698                         cfq_start_queueing(cfqd, cfqq);
1699                 }
1700         } else if (cfq_should_preempt(cfqd, cfqq, crq)) {
1701                 /*
1702                  * not the active queue - expire current slice if it is
1703                  * idle and has expired it's mean thinktime or this new queue
1704                  * has some old slice time left and is of higher priority
1705                  */
1706                 cfq_preempt_queue(cfqd, cfqq);
1707                 cfq_mark_cfqq_must_dispatch(cfqq);
1708                 cfq_start_queueing(cfqd, cfqq);
1709         }
1710 }
1711
1712 static void cfq_insert_request(request_queue_t *q, struct request *rq)
1713 {
1714         struct cfq_data *cfqd = q->elevator->elevator_data;
1715         struct cfq_rq *crq = RQ_DATA(rq);
1716         struct cfq_queue *cfqq = crq->cfq_queue;
1717
1718         cfq_init_prio_data(cfqq);
1719
1720         cfq_add_crq_rb(crq);
1721
1722         list_add_tail(&rq->queuelist, &cfqq->fifo);
1723
1724         if (rq_mergeable(rq))
1725                 cfq_add_crq_hash(cfqd, crq);
1726
1727         cfq_crq_enqueued(cfqd, cfqq, crq);
1728 }
1729
1730 static void cfq_completed_request(request_queue_t *q, struct request *rq)
1731 {
1732         struct cfq_rq *crq = RQ_DATA(rq);
1733         struct cfq_queue *cfqq = crq->cfq_queue;
1734         struct cfq_data *cfqd = cfqq->cfqd;
1735         const int sync = cfq_crq_is_sync(crq);
1736         unsigned long now;
1737
1738         now = jiffies;
1739
1740         WARN_ON(!cfqd->rq_in_driver);
1741         WARN_ON(!cfqq->on_dispatch[sync]);
1742         cfqd->rq_in_driver--;
1743         cfqq->on_dispatch[sync]--;
1744
1745         if (!cfq_class_idle(cfqq))
1746                 cfqd->last_end_request = now;
1747
1748         if (!cfq_cfqq_dispatched(cfqq)) {
1749                 if (cfq_cfqq_on_rr(cfqq)) {
1750                         cfqq->service_last = now;
1751                         cfq_resort_rr_list(cfqq, 0);
1752                 }
1753                 cfq_schedule_dispatch(cfqd);
1754         }
1755
1756         if (cfq_crq_is_sync(crq))
1757                 crq->io_context->last_end_request = now;
1758 }
1759
1760 static struct request *
1761 cfq_former_request(request_queue_t *q, struct request *rq)
1762 {
1763         struct cfq_rq *crq = RQ_DATA(rq);
1764         struct rb_node *rbprev = rb_prev(&crq->rb_node);
1765
1766         if (rbprev)
1767                 return rb_entry_crq(rbprev)->request;
1768
1769         return NULL;
1770 }
1771
1772 static struct request *
1773 cfq_latter_request(request_queue_t *q, struct request *rq)
1774 {
1775         struct cfq_rq *crq = RQ_DATA(rq);
1776         struct rb_node *rbnext = rb_next(&crq->rb_node);
1777
1778         if (rbnext)
1779                 return rb_entry_crq(rbnext)->request;
1780
1781         return NULL;
1782 }
1783
1784 /*
1785  * we temporarily boost lower priority queues if they are holding fs exclusive
1786  * resources. they are boosted to normal prio (CLASS_BE/4)
1787  */
1788 static void cfq_prio_boost(struct cfq_queue *cfqq)
1789 {
1790         const int ioprio_class = cfqq->ioprio_class;
1791         const int ioprio = cfqq->ioprio;
1792
1793         if (has_fs_excl()) {
1794                 /*
1795                  * boost idle prio on transactions that would lock out other
1796                  * users of the filesystem
1797                  */
1798                 if (cfq_class_idle(cfqq))
1799                         cfqq->ioprio_class = IOPRIO_CLASS_BE;
1800                 if (cfqq->ioprio > IOPRIO_NORM)
1801                         cfqq->ioprio = IOPRIO_NORM;
1802         } else {
1803                 /*
1804                  * check if we need to unboost the queue
1805                  */
1806                 if (cfqq->ioprio_class != cfqq->org_ioprio_class)
1807                         cfqq->ioprio_class = cfqq->org_ioprio_class;
1808                 if (cfqq->ioprio != cfqq->org_ioprio)
1809                         cfqq->ioprio = cfqq->org_ioprio;
1810         }
1811
1812         /*
1813          * refile between round-robin lists if we moved the priority class
1814          */
1815         if ((ioprio_class != cfqq->ioprio_class || ioprio != cfqq->ioprio) &&
1816             cfq_cfqq_on_rr(cfqq))
1817                 cfq_resort_rr_list(cfqq, 0);
1818 }
1819
1820 static inline pid_t cfq_queue_pid(struct task_struct *task, int rw)
1821 {
1822         if (rw == READ || process_sync(task))
1823                 return task->pid;
1824
1825         return CFQ_KEY_ASYNC;
1826 }
1827
1828 static inline int
1829 __cfq_may_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1830                 struct task_struct *task, int rw)
1831 {
1832 #if 1
1833         if ((cfq_cfqq_wait_request(cfqq) || cfq_cfqq_must_alloc(cfqq)) &&
1834             !cfq_cfqq_must_alloc_slice(cfqq)) {
1835                 cfq_mark_cfqq_must_alloc_slice(cfqq);
1836                 return ELV_MQUEUE_MUST;
1837         }
1838
1839         return ELV_MQUEUE_MAY;
1840 #else
1841         if (!cfqq || task->flags & PF_MEMALLOC)
1842                 return ELV_MQUEUE_MAY;
1843         if (!cfqq->allocated[rw] || cfq_cfqq_must_alloc(cfqq)) {
1844                 if (cfq_cfqq_wait_request(cfqq))
1845                         return ELV_MQUEUE_MUST;
1846
1847                 /*
1848                  * only allow 1 ELV_MQUEUE_MUST per slice, otherwise we
1849                  * can quickly flood the queue with writes from a single task
1850                  */
1851                 if (rw == READ || !cfq_cfqq_must_alloc_slice(cfqq)) {
1852                         cfq_mark_cfqq_must_alloc_slice(cfqq);
1853                         return ELV_MQUEUE_MUST;
1854                 }
1855
1856                 return ELV_MQUEUE_MAY;
1857         }
1858         if (cfq_class_idle(cfqq))
1859                 return ELV_MQUEUE_NO;
1860         if (cfqq->allocated[rw] >= cfqd->max_queued) {
1861                 struct io_context *ioc = get_io_context(GFP_ATOMIC);
1862                 int ret = ELV_MQUEUE_NO;
1863
1864                 if (ioc && ioc->nr_batch_requests)
1865                         ret = ELV_MQUEUE_MAY;
1866
1867                 put_io_context(ioc);
1868                 return ret;
1869         }
1870
1871         return ELV_MQUEUE_MAY;
1872 #endif
1873 }
1874
1875 static int cfq_may_queue(request_queue_t *q, int rw, struct bio *bio)
1876 {
1877         struct cfq_data *cfqd = q->elevator->elevator_data;
1878         struct task_struct *tsk = current;
1879         struct cfq_queue *cfqq;
1880
1881         /*
1882          * don't force setup of a queue from here, as a call to may_queue
1883          * does not necessarily imply that a request actually will be queued.
1884          * so just lookup a possibly existing queue, or return 'may queue'
1885          * if that fails
1886          */
1887         cfqq = cfq_find_cfq_hash(cfqd, cfq_queue_pid(tsk, rw), tsk->ioprio);
1888         if (cfqq) {
1889                 cfq_init_prio_data(cfqq);
1890                 cfq_prio_boost(cfqq);
1891
1892                 return __cfq_may_queue(cfqd, cfqq, tsk, rw);
1893         }
1894
1895         return ELV_MQUEUE_MAY;
1896 }
1897
1898 static void cfq_check_waiters(request_queue_t *q, struct cfq_queue *cfqq)
1899 {
1900         struct cfq_data *cfqd = q->elevator->elevator_data;
1901         struct request_list *rl = &q->rq;
1902
1903         if (cfqq->allocated[READ] <= cfqd->max_queued || cfqd->rq_starved) {
1904                 smp_mb();
1905                 if (waitqueue_active(&rl->wait[READ]))
1906                         wake_up(&rl->wait[READ]);
1907         }
1908
1909         if (cfqq->allocated[WRITE] <= cfqd->max_queued || cfqd->rq_starved) {
1910                 smp_mb();
1911                 if (waitqueue_active(&rl->wait[WRITE]))
1912                         wake_up(&rl->wait[WRITE]);
1913         }
1914 }
1915
1916 /*
1917  * queue lock held here
1918  */
1919 static void cfq_put_request(request_queue_t *q, struct request *rq)
1920 {
1921         struct cfq_data *cfqd = q->elevator->elevator_data;
1922         struct cfq_rq *crq = RQ_DATA(rq);
1923
1924         if (crq) {
1925                 struct cfq_queue *cfqq = crq->cfq_queue;
1926                 const int rw = rq_data_dir(rq);
1927
1928                 BUG_ON(!cfqq->allocated[rw]);
1929                 cfqq->allocated[rw]--;
1930
1931                 put_io_context(crq->io_context->ioc);
1932
1933                 mempool_free(crq, cfqd->crq_pool);
1934                 rq->elevator_private = NULL;
1935
1936                 cfq_check_waiters(q, cfqq);
1937                 cfq_put_queue(cfqq);
1938         }
1939 }
1940
1941 /*
1942  * Allocate cfq data structures associated with this request.
1943  */
1944 static int
1945 cfq_set_request(request_queue_t *q, struct request *rq, struct bio *bio,
1946                 gfp_t gfp_mask)
1947 {
1948         struct cfq_data *cfqd = q->elevator->elevator_data;
1949         struct task_struct *tsk = current;
1950         struct cfq_io_context *cic;
1951         const int rw = rq_data_dir(rq);
1952         pid_t key = cfq_queue_pid(tsk, rw);
1953         struct cfq_queue *cfqq;
1954         struct cfq_rq *crq;
1955         unsigned long flags;
1956         int is_sync = key != CFQ_KEY_ASYNC;
1957
1958         might_sleep_if(gfp_mask & __GFP_WAIT);
1959
1960         cic = cfq_get_io_context(cfqd, key, gfp_mask);
1961
1962         spin_lock_irqsave(q->queue_lock, flags);
1963
1964         if (!cic)
1965                 goto queue_fail;
1966
1967         if (!cic->cfqq[is_sync]) {
1968                 cfqq = cfq_get_queue(cfqd, key, tsk->ioprio, gfp_mask);
1969                 if (!cfqq)
1970                         goto queue_fail;
1971
1972                 cic->cfqq[is_sync] = cfqq;
1973         } else
1974                 cfqq = cic->cfqq[is_sync];
1975
1976         cfqq->allocated[rw]++;
1977         cfq_clear_cfqq_must_alloc(cfqq);
1978         cfqd->rq_starved = 0;
1979         atomic_inc(&cfqq->ref);
1980         spin_unlock_irqrestore(q->queue_lock, flags);
1981
1982         crq = mempool_alloc(cfqd->crq_pool, gfp_mask);
1983         if (crq) {
1984                 RB_CLEAR(&crq->rb_node);
1985                 crq->rb_key = 0;
1986                 crq->request = rq;
1987                 INIT_HLIST_NODE(&crq->hash);
1988                 crq->cfq_queue = cfqq;
1989                 crq->io_context = cic;
1990
1991                 if (is_sync)
1992                         cfq_mark_crq_is_sync(crq);
1993                 else
1994                         cfq_clear_crq_is_sync(crq);
1995
1996                 rq->elevator_private = crq;
1997                 return 0;
1998         }
1999
2000         spin_lock_irqsave(q->queue_lock, flags);
2001         cfqq->allocated[rw]--;
2002         if (!(cfqq->allocated[0] + cfqq->allocated[1]))
2003                 cfq_mark_cfqq_must_alloc(cfqq);
2004         cfq_put_queue(cfqq);
2005 queue_fail:
2006         if (cic)
2007                 put_io_context(cic->ioc);
2008         /*
2009          * mark us rq allocation starved. we need to kickstart the process
2010          * ourselves if there are no pending requests that can do it for us.
2011          * that would be an extremely rare OOM situation
2012          */
2013         cfqd->rq_starved = 1;
2014         cfq_schedule_dispatch(cfqd);
2015         spin_unlock_irqrestore(q->queue_lock, flags);
2016         return 1;
2017 }
2018
2019 static void cfq_kick_queue(void *data)
2020 {
2021         request_queue_t *q = data;
2022         struct cfq_data *cfqd = q->elevator->elevator_data;
2023         unsigned long flags;
2024
2025         spin_lock_irqsave(q->queue_lock, flags);
2026
2027         if (cfqd->rq_starved) {
2028                 struct request_list *rl = &q->rq;
2029
2030                 /*
2031                  * we aren't guaranteed to get a request after this, but we
2032                  * have to be opportunistic
2033                  */
2034                 smp_mb();
2035                 if (waitqueue_active(&rl->wait[READ]))
2036                         wake_up(&rl->wait[READ]);
2037                 if (waitqueue_active(&rl->wait[WRITE]))
2038                         wake_up(&rl->wait[WRITE]);
2039         }
2040
2041         blk_remove_plug(q);
2042         q->request_fn(q);
2043         spin_unlock_irqrestore(q->queue_lock, flags);
2044 }
2045
2046 /*
2047  * Timer running if the active_queue is currently idling inside its time slice
2048  */
2049 static void cfq_idle_slice_timer(unsigned long data)
2050 {
2051         struct cfq_data *cfqd = (struct cfq_data *) data;
2052         struct cfq_queue *cfqq;
2053         unsigned long flags;
2054
2055         spin_lock_irqsave(cfqd->queue->queue_lock, flags);
2056
2057         if ((cfqq = cfqd->active_queue) != NULL) {
2058                 unsigned long now = jiffies;
2059
2060                 /*
2061                  * expired
2062                  */
2063                 if (time_after(now, cfqq->slice_end))
2064                         goto expire;
2065
2066                 /*
2067                  * only expire and reinvoke request handler, if there are
2068                  * other queues with pending requests
2069                  */
2070                 if (!cfqd->busy_queues) {
2071                         cfqd->idle_slice_timer.expires = min(now + cfqd->cfq_slice_idle, cfqq->slice_end);
2072                         add_timer(&cfqd->idle_slice_timer);
2073                         goto out_cont;
2074                 }
2075
2076                 /*
2077                  * not expired and it has a request pending, let it dispatch
2078                  */
2079                 if (!RB_EMPTY(&cfqq->sort_list)) {
2080                         cfq_mark_cfqq_must_dispatch(cfqq);
2081                         goto out_kick;
2082                 }
2083         }
2084 expire:
2085         cfq_slice_expired(cfqd, 0);
2086 out_kick:
2087         cfq_schedule_dispatch(cfqd);
2088 out_cont:
2089         spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
2090 }
2091
2092 /*
2093  * Timer running if an idle class queue is waiting for service
2094  */
2095 static void cfq_idle_class_timer(unsigned long data)
2096 {
2097         struct cfq_data *cfqd = (struct cfq_data *) data;
2098         unsigned long flags, end;
2099
2100         spin_lock_irqsave(cfqd->queue->queue_lock, flags);
2101
2102         /*
2103          * race with a non-idle queue, reset timer
2104          */
2105         end = cfqd->last_end_request + CFQ_IDLE_GRACE;
2106         if (!time_after_eq(jiffies, end)) {
2107                 cfqd->idle_class_timer.expires = end;
2108                 add_timer(&cfqd->idle_class_timer);
2109         } else
2110                 cfq_schedule_dispatch(cfqd);
2111
2112         spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
2113 }
2114
2115 static void cfq_shutdown_timer_wq(struct cfq_data *cfqd)
2116 {
2117         del_timer_sync(&cfqd->idle_slice_timer);
2118         del_timer_sync(&cfqd->idle_class_timer);
2119         blk_sync_queue(cfqd->queue);
2120 }
2121
2122 static void cfq_put_cfqd(struct cfq_data *cfqd)
2123 {
2124         if (!atomic_dec_and_test(&cfqd->ref))
2125                 return;
2126
2127         cfq_shutdown_timer_wq(cfqd);
2128
2129         mempool_destroy(cfqd->crq_pool);
2130         kfree(cfqd->crq_hash);
2131         kfree(cfqd->cfq_hash);
2132         kfree(cfqd);
2133 }
2134
2135 static void cfq_exit_queue(elevator_t *e)
2136 {
2137         struct cfq_data *cfqd = e->elevator_data;
2138         request_queue_t *q = cfqd->queue;
2139
2140         cfq_shutdown_timer_wq(cfqd);
2141         write_lock(&cfq_exit_lock);
2142         spin_lock_irq(q->queue_lock);
2143         if (cfqd->active_queue)
2144                 __cfq_slice_expired(cfqd, cfqd->active_queue, 0);
2145         while(!list_empty(&cfqd->cic_list)) {
2146                 struct cfq_io_context *cic = list_entry(cfqd->cic_list.next,
2147                                                         struct cfq_io_context,
2148                                                         queue_list);
2149                 if (cic->cfqq[ASYNC]) {
2150                         cfq_put_queue(cic->cfqq[ASYNC]);
2151                         cic->cfqq[ASYNC] = NULL;
2152                 }
2153                 if (cic->cfqq[SYNC]) {
2154                         cfq_put_queue(cic->cfqq[SYNC]);
2155                         cic->cfqq[SYNC] = NULL;
2156                 }
2157                 cic->key = NULL;
2158                 list_del_init(&cic->queue_list);
2159         }
2160         spin_unlock_irq(q->queue_lock);
2161         write_unlock(&cfq_exit_lock);
2162         cfq_put_cfqd(cfqd);
2163 }
2164
2165 static int cfq_init_queue(request_queue_t *q, elevator_t *e)
2166 {
2167         struct cfq_data *cfqd;
2168         int i;
2169
2170         cfqd = kmalloc(sizeof(*cfqd), GFP_KERNEL);
2171         if (!cfqd)
2172                 return -ENOMEM;
2173
2174         memset(cfqd, 0, sizeof(*cfqd));
2175
2176         for (i = 0; i < CFQ_PRIO_LISTS; i++)
2177                 INIT_LIST_HEAD(&cfqd->rr_list[i]);
2178
2179         INIT_LIST_HEAD(&cfqd->busy_rr);
2180         INIT_LIST_HEAD(&cfqd->cur_rr);
2181         INIT_LIST_HEAD(&cfqd->idle_rr);
2182         INIT_LIST_HEAD(&cfqd->empty_list);
2183         INIT_LIST_HEAD(&cfqd->cic_list);
2184
2185         cfqd->crq_hash = kmalloc(sizeof(struct hlist_head) * CFQ_MHASH_ENTRIES, GFP_KERNEL);
2186         if (!cfqd->crq_hash)
2187                 goto out_crqhash;
2188
2189         cfqd->cfq_hash = kmalloc(sizeof(struct hlist_head) * CFQ_QHASH_ENTRIES, GFP_KERNEL);
2190         if (!cfqd->cfq_hash)
2191                 goto out_cfqhash;
2192
2193         cfqd->crq_pool = mempool_create(BLKDEV_MIN_RQ, mempool_alloc_slab, mempool_free_slab, crq_pool);
2194         if (!cfqd->crq_pool)
2195                 goto out_crqpool;
2196
2197         for (i = 0; i < CFQ_MHASH_ENTRIES; i++)
2198                 INIT_HLIST_HEAD(&cfqd->crq_hash[i]);
2199         for (i = 0; i < CFQ_QHASH_ENTRIES; i++)
2200                 INIT_HLIST_HEAD(&cfqd->cfq_hash[i]);
2201
2202         e->elevator_data = cfqd;
2203
2204         cfqd->queue = q;
2205
2206         cfqd->max_queued = q->nr_requests / 4;
2207         q->nr_batching = cfq_queued;
2208
2209         init_timer(&cfqd->idle_slice_timer);
2210         cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
2211         cfqd->idle_slice_timer.data = (unsigned long) cfqd;
2212
2213         init_timer(&cfqd->idle_class_timer);
2214         cfqd->idle_class_timer.function = cfq_idle_class_timer;
2215         cfqd->idle_class_timer.data = (unsigned long) cfqd;
2216
2217         INIT_WORK(&cfqd->unplug_work, cfq_kick_queue, q);
2218
2219         atomic_set(&cfqd->ref, 1);
2220
2221         cfqd->cfq_queued = cfq_queued;
2222         cfqd->cfq_quantum = cfq_quantum;
2223         cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0];
2224         cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1];
2225         cfqd->cfq_back_max = cfq_back_max;
2226         cfqd->cfq_back_penalty = cfq_back_penalty;
2227         cfqd->cfq_slice[0] = cfq_slice_async;
2228         cfqd->cfq_slice[1] = cfq_slice_sync;
2229         cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
2230         cfqd->cfq_slice_idle = cfq_slice_idle;
2231         cfqd->cfq_max_depth = cfq_max_depth;
2232
2233         return 0;
2234 out_crqpool:
2235         kfree(cfqd->cfq_hash);
2236 out_cfqhash:
2237         kfree(cfqd->crq_hash);
2238 out_crqhash:
2239         kfree(cfqd);
2240         return -ENOMEM;
2241 }
2242
2243 static void cfq_slab_kill(void)
2244 {
2245         if (crq_pool)
2246                 kmem_cache_destroy(crq_pool);
2247         if (cfq_pool)
2248                 kmem_cache_destroy(cfq_pool);
2249         if (cfq_ioc_pool)
2250                 kmem_cache_destroy(cfq_ioc_pool);
2251 }
2252
2253 static int __init cfq_slab_setup(void)
2254 {
2255         crq_pool = kmem_cache_create("crq_pool", sizeof(struct cfq_rq), 0, 0,
2256                                         NULL, NULL);
2257         if (!crq_pool)
2258                 goto fail;
2259
2260         cfq_pool = kmem_cache_create("cfq_pool", sizeof(struct cfq_queue), 0, 0,
2261                                         NULL, NULL);
2262         if (!cfq_pool)
2263                 goto fail;
2264
2265         cfq_ioc_pool = kmem_cache_create("cfq_ioc_pool",
2266                         sizeof(struct cfq_io_context), 0, 0, NULL, NULL);
2267         if (!cfq_ioc_pool)
2268                 goto fail;
2269
2270         return 0;
2271 fail:
2272         cfq_slab_kill();
2273         return -ENOMEM;
2274 }
2275
2276 /*
2277  * sysfs parts below -->
2278  */
2279 struct cfq_fs_entry {
2280         struct attribute attr;
2281         ssize_t (*show)(struct cfq_data *, char *);
2282         ssize_t (*store)(struct cfq_data *, const char *, size_t);
2283 };
2284
2285 static ssize_t
2286 cfq_var_show(unsigned int var, char *page)
2287 {
2288         return sprintf(page, "%d\n", var);
2289 }
2290
2291 static ssize_t
2292 cfq_var_store(unsigned int *var, const char *page, size_t count)
2293 {
2294         char *p = (char *) page;
2295
2296         *var = simple_strtoul(p, &p, 10);
2297         return count;
2298 }
2299
2300 #define SHOW_FUNCTION(__FUNC, __VAR, __CONV)                            \
2301 static ssize_t __FUNC(struct cfq_data *cfqd, char *page)                \
2302 {                                                                       \
2303         unsigned int __data = __VAR;                                    \
2304         if (__CONV)                                                     \
2305                 __data = jiffies_to_msecs(__data);                      \
2306         return cfq_var_show(__data, (page));                            \
2307 }
2308 SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0);
2309 SHOW_FUNCTION(cfq_queued_show, cfqd->cfq_queued, 0);
2310 SHOW_FUNCTION(cfq_fifo_expire_sync_show, cfqd->cfq_fifo_expire[1], 1);
2311 SHOW_FUNCTION(cfq_fifo_expire_async_show, cfqd->cfq_fifo_expire[0], 1);
2312 SHOW_FUNCTION(cfq_back_max_show, cfqd->cfq_back_max, 0);
2313 SHOW_FUNCTION(cfq_back_penalty_show, cfqd->cfq_back_penalty, 0);
2314 SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1);
2315 SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1);
2316 SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
2317 SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
2318 SHOW_FUNCTION(cfq_max_depth_show, cfqd->cfq_max_depth, 0);
2319 #undef SHOW_FUNCTION
2320
2321 #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV)                 \
2322 static ssize_t __FUNC(struct cfq_data *cfqd, const char *page, size_t count)    \
2323 {                                                                       \
2324         unsigned int __data;                                            \
2325         int ret = cfq_var_store(&__data, (page), count);                \
2326         if (__data < (MIN))                                             \
2327                 __data = (MIN);                                         \
2328         else if (__data > (MAX))                                        \
2329                 __data = (MAX);                                         \
2330         if (__CONV)                                                     \
2331                 *(__PTR) = msecs_to_jiffies(__data);                    \
2332         else                                                            \
2333                 *(__PTR) = __data;                                      \
2334         return ret;                                                     \
2335 }
2336 STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
2337 STORE_FUNCTION(cfq_queued_store, &cfqd->cfq_queued, 1, UINT_MAX, 0);
2338 STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1, UINT_MAX, 1);
2339 STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1, UINT_MAX, 1);
2340 STORE_FUNCTION(cfq_back_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
2341 STORE_FUNCTION(cfq_back_penalty_store, &cfqd->cfq_back_penalty, 1, UINT_MAX, 0);
2342 STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1);
2343 STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1);
2344 STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1);
2345 STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1, UINT_MAX, 0);
2346 STORE_FUNCTION(cfq_max_depth_store, &cfqd->cfq_max_depth, 1, UINT_MAX, 0);
2347 #undef STORE_FUNCTION
2348
2349 static struct cfq_fs_entry cfq_quantum_entry = {
2350         .attr = {.name = "quantum", .mode = S_IRUGO | S_IWUSR },
2351         .show = cfq_quantum_show,
2352         .store = cfq_quantum_store,
2353 };
2354 static struct cfq_fs_entry cfq_queued_entry = {
2355         .attr = {.name = "queued", .mode = S_IRUGO | S_IWUSR },
2356         .show = cfq_queued_show,
2357         .store = cfq_queued_store,
2358 };
2359 static struct cfq_fs_entry cfq_fifo_expire_sync_entry = {
2360         .attr = {.name = "fifo_expire_sync", .mode = S_IRUGO | S_IWUSR },
2361         .show = cfq_fifo_expire_sync_show,
2362         .store = cfq_fifo_expire_sync_store,
2363 };
2364 static struct cfq_fs_entry cfq_fifo_expire_async_entry = {
2365         .attr = {.name = "fifo_expire_async", .mode = S_IRUGO | S_IWUSR },
2366         .show = cfq_fifo_expire_async_show,
2367         .store = cfq_fifo_expire_async_store,
2368 };
2369 static struct cfq_fs_entry cfq_back_max_entry = {
2370         .attr = {.name = "back_seek_max", .mode = S_IRUGO | S_IWUSR },
2371         .show = cfq_back_max_show,
2372         .store = cfq_back_max_store,
2373 };
2374 static struct cfq_fs_entry cfq_back_penalty_entry = {
2375         .attr = {.name = "back_seek_penalty", .mode = S_IRUGO | S_IWUSR },
2376         .show = cfq_back_penalty_show,
2377         .store = cfq_back_penalty_store,
2378 };
2379 static struct cfq_fs_entry cfq_slice_sync_entry = {
2380         .attr = {.name = "slice_sync", .mode = S_IRUGO | S_IWUSR },
2381         .show = cfq_slice_sync_show,
2382         .store = cfq_slice_sync_store,
2383 };
2384 static struct cfq_fs_entry cfq_slice_async_entry = {
2385         .attr = {.name = "slice_async", .mode = S_IRUGO | S_IWUSR },
2386         .show = cfq_slice_async_show,
2387         .store = cfq_slice_async_store,
2388 };
2389 static struct cfq_fs_entry cfq_slice_async_rq_entry = {
2390         .attr = {.name = "slice_async_rq", .mode = S_IRUGO | S_IWUSR },
2391         .show = cfq_slice_async_rq_show,
2392         .store = cfq_slice_async_rq_store,
2393 };
2394 static struct cfq_fs_entry cfq_slice_idle_entry = {
2395         .attr = {.name = "slice_idle", .mode = S_IRUGO | S_IWUSR },
2396         .show = cfq_slice_idle_show,
2397         .store = cfq_slice_idle_store,
2398 };
2399 static struct cfq_fs_entry cfq_max_depth_entry = {
2400         .attr = {.name = "max_depth", .mode = S_IRUGO | S_IWUSR },
2401         .show = cfq_max_depth_show,
2402         .store = cfq_max_depth_store,
2403 };
2404
2405 static struct attribute *default_attrs[] = {
2406         &cfq_quantum_entry.attr,
2407         &cfq_queued_entry.attr,
2408         &cfq_fifo_expire_sync_entry.attr,
2409         &cfq_fifo_expire_async_entry.attr,
2410         &cfq_back_max_entry.attr,
2411         &cfq_back_penalty_entry.attr,
2412         &cfq_slice_sync_entry.attr,
2413         &cfq_slice_async_entry.attr,
2414         &cfq_slice_async_rq_entry.attr,
2415         &cfq_slice_idle_entry.attr,
2416         &cfq_max_depth_entry.attr,
2417         NULL,
2418 };
2419
2420 #define to_cfq(atr) container_of((atr), struct cfq_fs_entry, attr)
2421
2422 static ssize_t
2423 cfq_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
2424 {
2425         elevator_t *e = container_of(kobj, elevator_t, kobj);
2426         struct cfq_fs_entry *entry = to_cfq(attr);
2427
2428         if (!entry->show)
2429                 return -EIO;
2430
2431         return entry->show(e->elevator_data, page);
2432 }
2433
2434 static ssize_t
2435 cfq_attr_store(struct kobject *kobj, struct attribute *attr,
2436                const char *page, size_t length)
2437 {
2438         elevator_t *e = container_of(kobj, elevator_t, kobj);
2439         struct cfq_fs_entry *entry = to_cfq(attr);
2440
2441         if (!entry->store)
2442                 return -EIO;
2443
2444         return entry->store(e->elevator_data, page, length);
2445 }
2446
2447 static struct sysfs_ops cfq_sysfs_ops = {
2448         .show   = cfq_attr_show,
2449         .store  = cfq_attr_store,
2450 };
2451
2452 static struct kobj_type cfq_ktype = {
2453         .sysfs_ops      = &cfq_sysfs_ops,
2454         .default_attrs  = default_attrs,
2455 };
2456
2457 static struct elevator_type iosched_cfq = {
2458         .ops = {
2459                 .elevator_merge_fn =            cfq_merge,
2460                 .elevator_merged_fn =           cfq_merged_request,
2461                 .elevator_merge_req_fn =        cfq_merged_requests,
2462                 .elevator_dispatch_fn =         cfq_dispatch_requests,
2463                 .elevator_add_req_fn =          cfq_insert_request,
2464                 .elevator_activate_req_fn =     cfq_activate_request,
2465                 .elevator_deactivate_req_fn =   cfq_deactivate_request,
2466                 .elevator_queue_empty_fn =      cfq_queue_empty,
2467                 .elevator_completed_req_fn =    cfq_completed_request,
2468                 .elevator_former_req_fn =       cfq_former_request,
2469                 .elevator_latter_req_fn =       cfq_latter_request,
2470                 .elevator_set_req_fn =          cfq_set_request,
2471                 .elevator_put_req_fn =          cfq_put_request,
2472                 .elevator_may_queue_fn =        cfq_may_queue,
2473                 .elevator_init_fn =             cfq_init_queue,
2474                 .elevator_exit_fn =             cfq_exit_queue,
2475         },
2476         .elevator_ktype =       &cfq_ktype,
2477         .elevator_name =        "cfq",
2478         .elevator_owner =       THIS_MODULE,
2479 };
2480
2481 static int __init cfq_init(void)
2482 {
2483         int ret;
2484
2485         /*
2486          * could be 0 on HZ < 1000 setups
2487          */
2488         if (!cfq_slice_async)
2489                 cfq_slice_async = 1;
2490         if (!cfq_slice_idle)
2491                 cfq_slice_idle = 1;
2492
2493         if (cfq_slab_setup())
2494                 return -ENOMEM;
2495
2496         ret = elv_register(&iosched_cfq);
2497         if (ret)
2498                 cfq_slab_kill();
2499
2500         return ret;
2501 }
2502
2503 static void __exit cfq_exit(void)
2504 {
2505         elv_unregister(&iosched_cfq);
2506         cfq_slab_kill();
2507 }
2508
2509 module_init(cfq_init);
2510 module_exit(cfq_exit);
2511
2512 MODULE_AUTHOR("Jens Axboe");
2513 MODULE_LICENSE("GPL");
2514 MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler");