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@kernel.dk>
   8  */
   9 #include <linux/module.h>
  10 #include <linux/blkdev.h>
  11 #include <linux/elevator.h>
  12 #include <linux/rbtree.h>
  13 #include <linux/ioprio.h>
  14
  15 /*
  16  * tunables
  17  */
  18 static const int cfq_quantum = 4;               /* max queue in one round of service */
  19 static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 };
  20 static const int cfq_back_max = 16 * 1024;      /* maximum backwards seek, in KiB */
  21 static const int cfq_back_penalty = 2;          /* penalty of a backwards seek */
  22
  23 static const int cfq_slice_sync = HZ / 10;
  24 static int cfq_slice_async = HZ / 25;
  25 static const int cfq_slice_async_rq = 2;
  26 static int cfq_slice_idle = HZ / 125;
  27
  28 /*
  29  * offset from end of service tree
  30  */
  31 #define CFQ_IDLE_DELAY          (HZ / 5)
  32
  33 /*
  34  * below this threshold, we consider thinktime immediate
  35  */
  36 #define CFQ_MIN_TT              (2)
  37
  38 #define CFQ_SLICE_SCALE         (5)
  39
  40 #define RQ_CIC(rq)              ((struct cfq_io_context*)(rq)->elevator_private)
  41 #define RQ_CFQQ(rq)             ((rq)->elevator_private2)
  42
  43 static struct kmem_cache *cfq_pool;
  44 static struct kmem_cache *cfq_ioc_pool;
  45
  46 static DEFINE_PER_CPU(unsigned long, ioc_count);
  47 static struct completion *ioc_gone;
  48
  49 #define CFQ_PRIO_LISTS          IOPRIO_BE_NR
  50 #define cfq_class_idle(cfqq)    ((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
  51 #define cfq_class_rt(cfqq)      ((cfqq)->ioprio_class == IOPRIO_CLASS_RT)
  52
  53 #define ASYNC                   (0)
  54 #define SYNC                    (1)
  55
  56 #define sample_valid(samples)   ((samples) > 80)
  57
  58 /*
  59  * Most of our rbtree usage is for sorting with min extraction, so
  60  * if we cache the leftmost node we don't have to walk down the tree
  61  * to find it. Idea borrowed from Ingo Molnars CFS scheduler. We should
  62  * move this into the elevator for the rq sorting as well.
  63  */
  64 struct cfq_rb_root {
  65         struct rb_root rb;
  66         struct rb_node *left;
  67 };
  68 #define CFQ_RB_ROOT     (struct cfq_rb_root) { RB_ROOT, NULL, }
  69
  70 /*
  71  * Per block device queue structure
  72  */
  73 struct cfq_data {
  74         struct request_queue *queue;
  75
  76         /*
  77          * rr list of queues with requests and the count of them
  78          */
  79         struct cfq_rb_root service_tree;
  80         unsigned int busy_queues;
  81
  82         int rq_in_driver;
  83         int sync_flight;
  84         int hw_tag;
  85
  86         /*
  87          * idle window management
  88          */
  89         struct timer_list idle_slice_timer;
  90         struct work_struct unplug_work;
  91
  92         struct cfq_queue *active_queue;
  93         struct cfq_io_context *active_cic;
  94
  95         /*
  96          * async queue for each priority case
  97          */
  98         struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
  99         struct cfq_queue *async_idle_cfqq;
 100
 101         sector_t last_position;
 102         unsigned long last_end_request;
 103
 104         /*
 105          * tunables, see top of file
 106          */
 107         unsigned int cfq_quantum;
 108         unsigned int cfq_fifo_expire[2];
 109         unsigned int cfq_back_penalty;
 110         unsigned int cfq_back_max;
 111         unsigned int cfq_slice[2];
 112         unsigned int cfq_slice_async_rq;
 113         unsigned int cfq_slice_idle;
 114
 115         struct list_head cic_list;
 116 };
 117
 118 /*
 119  * Per process-grouping structure
 120  */
 121 struct cfq_queue {
 122         /* reference count */
 123         atomic_t ref;
 124         /* parent cfq_data */
 125         struct cfq_data *cfqd;
 126         /* service_tree member */
 127         struct rb_node rb_node;
 128         /* service_tree key */
 129         unsigned long rb_key;
 130         /* sorted list of pending requests */
 131         struct rb_root sort_list;
 132         /* if fifo isn't expired, next request to serve */
 133         struct request *next_rq;
 134         /* requests queued in sort_list */
 135         int queued[2];
 136         /* currently allocated requests */
 137         int allocated[2];
 138         /* pending metadata requests */
 139         int meta_pending;
 140         /* fifo list of requests in sort_list */
 141         struct list_head fifo;
 142
 143         unsigned long slice_end;
 144         long slice_resid;
 145
 146         /* number of requests that are on the dispatch list or inside driver */
 147         int dispatched;
 148
 149         /* io prio of this group */
 150         unsigned short ioprio, org_ioprio;
 151         unsigned short ioprio_class, org_ioprio_class;
 152
 153         /* various state flags, see below */
 154         unsigned int flags;
 155 };
 156
 157 enum cfqq_state_flags {
 158         CFQ_CFQQ_FLAG_on_rr = 0,        /* on round-robin busy list */
 159         CFQ_CFQQ_FLAG_wait_request,     /* waiting for a request */
 160         CFQ_CFQQ_FLAG_must_alloc,       /* must be allowed rq alloc */
 161         CFQ_CFQQ_FLAG_must_alloc_slice, /* per-slice must_alloc flag */
 162         CFQ_CFQQ_FLAG_must_dispatch,    /* must dispatch, even if expired */
 163         CFQ_CFQQ_FLAG_fifo_expire,      /* FIFO checked in this slice */
 164         CFQ_CFQQ_FLAG_idle_window,      /* slice idling enabled */
 165         CFQ_CFQQ_FLAG_prio_changed,     /* task priority has changed */
 166         CFQ_CFQQ_FLAG_queue_new,        /* queue never been serviced */
 167         CFQ_CFQQ_FLAG_slice_new,        /* no requests dispatched in slice */
 168         CFQ_CFQQ_FLAG_sync,             /* synchronous queue */
 169 };
 170
 171 #define CFQ_CFQQ_FNS(name)                                              \
 172 static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)         \
 173 {                                                                       \
 174         cfqq->flags |= (1 << CFQ_CFQQ_FLAG_##name);                     \
 175 }                                                                       \
 176 static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)        \
 177 {                                                                       \
 178         cfqq->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);                    \
 179 }                                                                       \
 180 static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)         \
 181 {                                                                       \
 182         return (cfqq->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;        \
 183 }
 184
 185 CFQ_CFQQ_FNS(on_rr);
 186 CFQ_CFQQ_FNS(wait_request);
 187 CFQ_CFQQ_FNS(must_alloc);
 188 CFQ_CFQQ_FNS(must_alloc_slice);
 189 CFQ_CFQQ_FNS(must_dispatch);
 190 CFQ_CFQQ_FNS(fifo_expire);
 191 CFQ_CFQQ_FNS(idle_window);
 192 CFQ_CFQQ_FNS(prio_changed);
 193 CFQ_CFQQ_FNS(queue_new);
 194 CFQ_CFQQ_FNS(slice_new);
 195 CFQ_CFQQ_FNS(sync);
 196 #undef CFQ_CFQQ_FNS
 197
 198 static void cfq_dispatch_insert(struct request_queue *, struct request *);
 199 static struct cfq_queue *cfq_get_queue(struct cfq_data *, int,
 200                                        struct io_context *, gfp_t);
 201 static struct cfq_io_context *cfq_cic_lookup(struct cfq_data *,
 202                                                 struct io_context *);
 203
 204 static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_context *cic,
 205                                             int is_sync)
 206 {
 207         return cic->cfqq[!!is_sync];
 208 }
 209
 210 static inline void cic_set_cfqq(struct cfq_io_context *cic,
 211                                 struct cfq_queue *cfqq, int is_sync)
 212 {
 213         cic->cfqq[!!is_sync] = cfqq;
 214 }
 215
 216 /*
 217  * We regard a request as SYNC, if it's either a read or has the SYNC bit
 218  * set (in which case it could also be direct WRITE).
 219  */
 220 static inline int cfq_bio_sync(struct bio *bio)
 221 {
 222         if (bio_data_dir(bio) == READ || bio_sync(bio))
 223                 return 1;
 224
 225         return 0;
 226 }
 227
 228 /*
 229  * scheduler run of queue, if there are requests pending and no one in the
 230  * driver that will restart queueing
 231  */
 232 static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
 233 {
 234         if (cfqd->busy_queues)
 235                 kblockd_schedule_work(&cfqd->unplug_work);
 236 }
 237
 238 static int cfq_queue_empty(struct request_queue *q)
 239 {
 240         struct cfq_data *cfqd = q->elevator->elevator_data;
 241
 242         return !cfqd->busy_queues;
 243 }
 244
 245 /*
 246  * Scale schedule slice based on io priority. Use the sync time slice only
 247  * if a queue is marked sync and has sync io queued. A sync queue with async
 248  * io only, should not get full sync slice length.
 249  */
 250 static inline int cfq_prio_slice(struct cfq_data *cfqd, int sync,
 251                                  unsigned short prio)
 252 {
 253         const int base_slice = cfqd->cfq_slice[sync];
 254
 255         WARN_ON(prio >= IOPRIO_BE_NR);
 256
 257         return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio));
 258 }
 259
 260 static inline int
 261 cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 262 {
 263         return cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio);
 264 }
 265
 266 static inline void
 267 cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 268 {
 269         cfqq->slice_end = cfq_prio_to_slice(cfqd, cfqq) + jiffies;
 270 }
 271
 272 /*
 273  * We need to wrap this check in cfq_cfqq_slice_new(), since ->slice_end
 274  * isn't valid until the first request from the dispatch is activated
 275  * and the slice time set.
 276  */
 277 static inline int cfq_slice_used(struct cfq_queue *cfqq)
 278 {
 279         if (cfq_cfqq_slice_new(cfqq))
 280                 return 0;
 281         if (time_before(jiffies, cfqq->slice_end))
 282                 return 0;
 283
 284         return 1;
 285 }
 286
 287 /*
 288  * Lifted from AS - choose which of rq1 and rq2 that is best served now.
 289  * We choose the request that is closest to the head right now. Distance
 290  * behind the head is penalized and only allowed to a certain extent.
 291  */
 292 static struct request *
 293 cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2)
 294 {
 295         sector_t last, s1, s2, d1 = 0, d2 = 0;
 296         unsigned long back_max;
 297 #define CFQ_RQ1_WRAP    0x01 /* request 1 wraps */
 298 #define CFQ_RQ2_WRAP    0x02 /* request 2 wraps */
 299         unsigned wrap = 0; /* bit mask: requests behind the disk head? */
 300
 301         if (rq1 == NULL || rq1 == rq2)
 302                 return rq2;
 303         if (rq2 == NULL)
 304                 return rq1;
 305
 306         if (rq_is_sync(rq1) && !rq_is_sync(rq2))
 307                 return rq1;
 308         else if (rq_is_sync(rq2) && !rq_is_sync(rq1))
 309                 return rq2;
 310         if (rq_is_meta(rq1) && !rq_is_meta(rq2))
 311                 return rq1;
 312         else if (rq_is_meta(rq2) && !rq_is_meta(rq1))
 313                 return rq2;
 314
 315         s1 = rq1->sector;
 316         s2 = rq2->sector;
 317
 318         last = cfqd->last_position;
 319
 320         /*
 321          * by definition, 1KiB is 2 sectors
 322          */
 323         back_max = cfqd->cfq_back_max * 2;
 324
 325         /*
 326          * Strict one way elevator _except_ in the case where we allow
 327          * short backward seeks which are biased as twice the cost of a
 328          * similar forward seek.
 329          */
 330         if (s1 >= last)
 331                 d1 = s1 - last;
 332         else if (s1 + back_max >= last)
 333                 d1 = (last - s1) * cfqd->cfq_back_penalty;
 334         else
 335                 wrap |= CFQ_RQ1_WRAP;
 336
 337         if (s2 >= last)
 338                 d2 = s2 - last;
 339         else if (s2 + back_max >= last)
 340                 d2 = (last - s2) * cfqd->cfq_back_penalty;
 341         else
 342                 wrap |= CFQ_RQ2_WRAP;
 343
 344         /* Found required data */
 345
 346         /*
 347          * By doing switch() on the bit mask "wrap" we avoid having to
 348          * check two variables for all permutations: --> faster!
 349          */
 350         switch (wrap) {
 351         case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
 352                 if (d1 < d2)
 353                         return rq1;
 354                 else if (d2 < d1)
 355                         return rq2;
 356                 else {
 357                         if (s1 >= s2)
 358                                 return rq1;
 359                         else
 360                                 return rq2;
 361                 }
 362
 363         case CFQ_RQ2_WRAP:
 364                 return rq1;
 365         case CFQ_RQ1_WRAP:
 366                 return rq2;
 367         case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */
 368         default:
 369                 /*
 370                  * Since both rqs are wrapped,
 371                  * start with the one that's further behind head
 372                  * (--> only *one* back seek required),
 373                  * since back seek takes more time than forward.
 374                  */
 375                 if (s1 <= s2)
 376                         return rq1;
 377                 else
 378                         return rq2;
 379         }
 380 }
 381
 382 /*
 383  * The below is leftmost cache rbtree addon
 384  */
 385 static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root)
 386 {
 387         if (!root->left)
 388                 root->left = rb_first(&root->rb);
 389
 390         if (root->left)
 391                 return rb_entry(root->left, struct cfq_queue, rb_node);
 392
 393         return NULL;
 394 }
 395
 396 static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root)
 397 {
 398         if (root->left == n)
 399                 root->left = NULL;
 400
 401         rb_erase(n, &root->rb);
 402         RB_CLEAR_NODE(n);
 403 }
 404
 405 /*
 406  * would be nice to take fifo expire time into account as well
 407  */
 408 static struct request *
 409 cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
 410                   struct request *last)
 411 {
 412         struct rb_node *rbnext = rb_next(&last->rb_node);
 413         struct rb_node *rbprev = rb_prev(&last->rb_node);
 414         struct request *next = NULL, *prev = NULL;
 415
 416         BUG_ON(RB_EMPTY_NODE(&last->rb_node));
 417
 418         if (rbprev)
 419                 prev = rb_entry_rq(rbprev);
 420
 421         if (rbnext)
 422                 next = rb_entry_rq(rbnext);
 423         else {
 424                 rbnext = rb_first(&cfqq->sort_list);
 425                 if (rbnext && rbnext != &last->rb_node)
 426                         next = rb_entry_rq(rbnext);
 427         }
 428
 429         return cfq_choose_req(cfqd, next, prev);
 430 }
 431
 432 static unsigned long cfq_slice_offset(struct cfq_data *cfqd,
 433                                       struct cfq_queue *cfqq)
 434 {
 435         /*
 436          * just an approximation, should be ok.
 437          */
 438         return (cfqd->busy_queues - 1) * (cfq_prio_slice(cfqd, 1, 0) -
 439                        cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio));
 440 }
 441
 442 /*
 443  * The cfqd->service_tree holds all pending cfq_queue's that have
 444  * requests waiting to be processed. It is sorted in the order that
 445  * we will service the queues.
 446  */
 447 static void cfq_service_tree_add(struct cfq_data *cfqd,
 448                                     struct cfq_queue *cfqq, int add_front)
 449 {
 450         struct rb_node **p, *parent;
 451         struct cfq_queue *__cfqq;
 452         unsigned long rb_key;
 453         int left;
 454
 455         if (cfq_class_idle(cfqq)) {
 456                 rb_key = CFQ_IDLE_DELAY;
 457                 parent = rb_last(&cfqd->service_tree.rb);
 458                 if (parent && parent != &cfqq->rb_node) {
 459                         __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
 460                         rb_key += __cfqq->rb_key;
 461                 } else
 462                         rb_key += jiffies;
 463         } else if (!add_front) {
 464                 rb_key = cfq_slice_offset(cfqd, cfqq) + jiffies;
 465                 rb_key += cfqq->slice_resid;
 466                 cfqq->slice_resid = 0;
 467         } else
 468                 rb_key = 0;
 469
 470         if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
 471                 /*
 472                  * same position, nothing more to do
 473                  */
 474                 if (rb_key == cfqq->rb_key)
 475                         return;
 476
 477                 cfq_rb_erase(&cfqq->rb_node, &cfqd->service_tree);
 478         }
 479
 480         left = 1;
 481         parent = NULL;
 482         p = &cfqd->service_tree.rb.rb_node;
 483         while (*p) {
 484                 struct rb_node **n;
 485
 486                 parent = *p;
 487                 __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
 488
 489                 /*
 490                  * sort RT queues first, we always want to give
 491                  * preference to them. IDLE queues goes to the back.
 492                  * after that, sort on the next service time.
 493                  */
 494                 if (cfq_class_rt(cfqq) > cfq_class_rt(__cfqq))
 495                         n = &(*p)->rb_left;
 496                 else if (cfq_class_rt(cfqq) < cfq_class_rt(__cfqq))
 497                         n = &(*p)->rb_right;
 498                 else if (cfq_class_idle(cfqq) < cfq_class_idle(__cfqq))
 499                         n = &(*p)->rb_left;
 500                 else if (cfq_class_idle(cfqq) > cfq_class_idle(__cfqq))
 501                         n = &(*p)->rb_right;
 502                 else if (rb_key < __cfqq->rb_key)
 503                         n = &(*p)->rb_left;
 504                 else
 505                         n = &(*p)->rb_right;
 506
 507                 if (n == &(*p)->rb_right)
 508                         left = 0;
 509
 510                 p = n;
 511         }
 512
 513         if (left)
 514                 cfqd->service_tree.left = &cfqq->rb_node;
 515
 516         cfqq->rb_key = rb_key;
 517         rb_link_node(&cfqq->rb_node, parent, p);
 518         rb_insert_color(&cfqq->rb_node, &cfqd->service_tree.rb);
 519 }
 520
 521 /*
 522  * Update cfqq's position in the service tree.
 523  */
 524 static void cfq_resort_rr_list(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 525 {
 526         /*
 527          * Resorting requires the cfqq to be on the RR list already.
 528          */
 529         if (cfq_cfqq_on_rr(cfqq))
 530                 cfq_service_tree_add(cfqd, cfqq, 0);
 531 }
 532
 533 /*
 534  * add to busy list of queues for service, trying to be fair in ordering
 535  * the pending list according to last request service
 536  */
 537 static inline void
 538 cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 539 {
 540         BUG_ON(cfq_cfqq_on_rr(cfqq));
 541         cfq_mark_cfqq_on_rr(cfqq);
 542         cfqd->busy_queues++;
 543
 544         cfq_resort_rr_list(cfqd, cfqq);
 545 }
 546
 547 /*
 548  * Called when the cfqq no longer has requests pending, remove it from
 549  * the service tree.
 550  */
 551 static inline void
 552 cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 553 {
 554         BUG_ON(!cfq_cfqq_on_rr(cfqq));
 555         cfq_clear_cfqq_on_rr(cfqq);
 556
 557         if (!RB_EMPTY_NODE(&cfqq->rb_node))
 558                 cfq_rb_erase(&cfqq->rb_node, &cfqd->service_tree);
 559
 560         BUG_ON(!cfqd->busy_queues);
 561         cfqd->busy_queues--;
 562 }
 563
 564 /*
 565  * rb tree support functions
 566  */
 567 static inline void cfq_del_rq_rb(struct request *rq)
 568 {
 569         struct cfq_queue *cfqq = RQ_CFQQ(rq);
 570         struct cfq_data *cfqd = cfqq->cfqd;
 571         const int sync = rq_is_sync(rq);
 572
 573         BUG_ON(!cfqq->queued[sync]);
 574         cfqq->queued[sync]--;
 575
 576         elv_rb_del(&cfqq->sort_list, rq);
 577
 578         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list))
 579                 cfq_del_cfqq_rr(cfqd, cfqq);
 580 }
 581
 582 static void cfq_add_rq_rb(struct request *rq)
 583 {
 584         struct cfq_queue *cfqq = RQ_CFQQ(rq);
 585         struct cfq_data *cfqd = cfqq->cfqd;
 586         struct request *__alias;
 587
 588         cfqq->queued[rq_is_sync(rq)]++;
 589
 590         /*
 591          * looks a little odd, but the first insert might return an alias.
 592          * if that happens, put the alias on the dispatch list
 593          */
 594         while ((__alias = elv_rb_add(&cfqq->sort_list, rq)) != NULL)
 595                 cfq_dispatch_insert(cfqd->queue, __alias);
 596
 597         if (!cfq_cfqq_on_rr(cfqq))
 598                 cfq_add_cfqq_rr(cfqd, cfqq);
 599
 600         /*
 601          * check if this request is a better next-serve candidate
 602          */
 603         cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq);
 604         BUG_ON(!cfqq->next_rq);
 605 }
 606
 607 static inline void
 608 cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
 609 {
 610         elv_rb_del(&cfqq->sort_list, rq);
 611         cfqq->queued[rq_is_sync(rq)]--;
 612         cfq_add_rq_rb(rq);
 613 }
 614
 615 static struct request *
 616 cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio)
 617 {
 618         struct task_struct *tsk = current;
 619         struct cfq_io_context *cic;
 620         struct cfq_queue *cfqq;
 621
 622         cic = cfq_cic_lookup(cfqd, tsk->io_context);
 623         if (!cic)
 624                 return NULL;
 625
 626         cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
 627         if (cfqq) {
 628                 sector_t sector = bio->bi_sector + bio_sectors(bio);
 629
 630                 return elv_rb_find(&cfqq->sort_list, sector);
 631         }
 632
 633         return NULL;
 634 }
 635
 636 static void cfq_activate_request(struct request_queue *q, struct request *rq)
 637 {
 638         struct cfq_data *cfqd = q->elevator->elevator_data;
 639
 640         cfqd->rq_in_driver++;
 641
 642         /*
 643          * If the depth is larger 1, it really could be queueing. But lets
 644          * make the mark a little higher - idling could still be good for
 645          * low queueing, and a low queueing number could also just indicate
 646          * a SCSI mid layer like behaviour where limit+1 is often seen.
 647          */
 648         if (!cfqd->hw_tag && cfqd->rq_in_driver > 4)
 649                 cfqd->hw_tag = 1;
 650
 651         cfqd->last_position = rq->hard_sector + rq->hard_nr_sectors;
 652 }
 653
 654 static void cfq_deactivate_request(struct request_queue *q, struct request *rq)
 655 {
 656         struct cfq_data *cfqd = q->elevator->elevator_data;
 657
 658         WARN_ON(!cfqd->rq_in_driver);
 659         cfqd->rq_in_driver--;
 660 }
 661
 662 static void cfq_remove_request(struct request *rq)
 663 {
 664         struct cfq_queue *cfqq = RQ_CFQQ(rq);
 665
 666         if (cfqq->next_rq == rq)
 667                 cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq);
 668
 669         list_del_init(&rq->queuelist);
 670         cfq_del_rq_rb(rq);
 671
 672         if (rq_is_meta(rq)) {
 673                 WARN_ON(!cfqq->meta_pending);
 674                 cfqq->meta_pending--;
 675         }
 676 }
 677
 678 static int cfq_merge(struct request_queue *q, struct request **req,
 679                      struct bio *bio)
 680 {
 681         struct cfq_data *cfqd = q->elevator->elevator_data;
 682         struct request *__rq;
 683
 684         __rq = cfq_find_rq_fmerge(cfqd, bio);
 685         if (__rq && elv_rq_merge_ok(__rq, bio)) {
 686                 *req = __rq;
 687                 return ELEVATOR_FRONT_MERGE;
 688         }
 689
 690         return ELEVATOR_NO_MERGE;
 691 }
 692
 693 static void cfq_merged_request(struct request_queue *q, struct request *req,
 694                                int type)
 695 {
 696         if (type == ELEVATOR_FRONT_MERGE) {
 697                 struct cfq_queue *cfqq = RQ_CFQQ(req);
 698
 699                 cfq_reposition_rq_rb(cfqq, req);
 700         }
 701 }
 702
 703 static void
 704 cfq_merged_requests(struct request_queue *q, struct request *rq,
 705                     struct request *next)
 706 {
 707         /*
 708          * reposition in fifo if next is older than rq
 709          */
 710         if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
 711             time_before(next->start_time, rq->start_time))
 712                 list_move(&rq->queuelist, &next->queuelist);
 713
 714         cfq_remove_request(next);
 715 }
 716
 717 static int cfq_allow_merge(struct request_queue *q, struct request *rq,
 718                            struct bio *bio)
 719 {
 720         struct cfq_data *cfqd = q->elevator->elevator_data;
 721         struct cfq_io_context *cic;
 722         struct cfq_queue *cfqq;
 723
 724         /*
 725          * Disallow merge of a sync bio into an async request.
 726          */
 727         if (cfq_bio_sync(bio) && !rq_is_sync(rq))
 728                 return 0;
 729
 730         /*
 731          * Lookup the cfqq that this bio will be queued with. Allow
 732          * merge only if rq is queued there.
 733          */
 734         cic = cfq_cic_lookup(cfqd, current->io_context);
 735         if (!cic)
 736                 return 0;
 737
 738         cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
 739         if (cfqq == RQ_CFQQ(rq))
 740                 return 1;
 741
 742         return 0;
 743 }
 744
 745 static inline void
 746 __cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 747 {
 748         if (cfqq) {
 749                 cfqq->slice_end = 0;
 750                 cfq_clear_cfqq_must_alloc_slice(cfqq);
 751                 cfq_clear_cfqq_fifo_expire(cfqq);
 752                 cfq_mark_cfqq_slice_new(cfqq);
 753                 cfq_clear_cfqq_queue_new(cfqq);
 754         }
 755
 756         cfqd->active_queue = cfqq;
 757 }
 758
 759 /*
 760  * current cfqq expired its slice (or was too idle), select new one
 761  */
 762 static void
 763 __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
 764                     int timed_out)
 765 {
 766         if (cfq_cfqq_wait_request(cfqq))
 767                 del_timer(&cfqd->idle_slice_timer);
 768
 769         cfq_clear_cfqq_must_dispatch(cfqq);
 770         cfq_clear_cfqq_wait_request(cfqq);
 771
 772         /*
 773          * store what was left of this slice, if the queue idled/timed out
 774          */
 775         if (timed_out && !cfq_cfqq_slice_new(cfqq))
 776                 cfqq->slice_resid = cfqq->slice_end - jiffies;
 777
 778         cfq_resort_rr_list(cfqd, cfqq);
 779
 780         if (cfqq == cfqd->active_queue)
 781                 cfqd->active_queue = NULL;
 782
 783         if (cfqd->active_cic) {
 784                 put_io_context(cfqd->active_cic->ioc);
 785                 cfqd->active_cic = NULL;
 786         }
 787 }
 788
 789 static inline void cfq_slice_expired(struct cfq_data *cfqd, int timed_out)
 790 {
 791         struct cfq_queue *cfqq = cfqd->active_queue;
 792
 793         if (cfqq)
 794                 __cfq_slice_expired(cfqd, cfqq, timed_out);
 795 }
 796
 797 /*
 798  * Get next queue for service. Unless we have a queue preemption,
 799  * we'll simply select the first cfqq in the service tree.
 800  */
 801 static struct cfq_queue *cfq_get_next_queue(struct cfq_data *cfqd)
 802 {
 803         if (RB_EMPTY_ROOT(&cfqd->service_tree.rb))
 804                 return NULL;
 805
 806         return cfq_rb_first(&cfqd->service_tree);
 807 }
 808
 809 /*
 810  * Get and set a new active queue for service.
 811  */
 812 static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd)
 813 {
 814         struct cfq_queue *cfqq;
 815
 816         cfqq = cfq_get_next_queue(cfqd);
 817         __cfq_set_active_queue(cfqd, cfqq);
 818         return cfqq;
 819 }
 820
 821 static inline sector_t cfq_dist_from_last(struct cfq_data *cfqd,
 822                                           struct request *rq)
 823 {
 824         if (rq->sector >= cfqd->last_position)
 825                 return rq->sector - cfqd->last_position;
 826         else
 827                 return cfqd->last_position - rq->sector;
 828 }
 829
 830 static inline int cfq_rq_close(struct cfq_data *cfqd, struct request *rq)
 831 {
 832         struct cfq_io_context *cic = cfqd->active_cic;
 833
 834         if (!sample_valid(cic->seek_samples))
 835                 return 0;
 836
 837         return cfq_dist_from_last(cfqd, rq) <= cic->seek_mean;
 838 }
 839
 840 static int cfq_close_cooperator(struct cfq_data *cfq_data,
 841                                 struct cfq_queue *cfqq)
 842 {
 843         /*
 844          * We should notice if some of the queues are cooperating, eg
 845          * working closely on the same area of the disk. In that case,
 846          * we can group them together and don't waste time idling.
 847          */
 848         return 0;
 849 }
 850
 851 #define CIC_SEEKY(cic) ((cic)->seek_mean > (8 * 1024))
 852
 853 static void cfq_arm_slice_timer(struct cfq_data *cfqd)
 854 {
 855         struct cfq_queue *cfqq = cfqd->active_queue;
 856         struct cfq_io_context *cic;
 857         unsigned long sl;
 858
 859         WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list));
 860         WARN_ON(cfq_cfqq_slice_new(cfqq));
 861
 862         /*
 863          * idle is disabled, either manually or by past process history
 864          */
 865         if (!cfqd->cfq_slice_idle || !cfq_cfqq_idle_window(cfqq))
 866                 return;
 867
 868         /*
 869          * task has exited, don't wait
 870          */
 871         cic = cfqd->active_cic;
 872         if (!cic || !atomic_read(&cic->ioc->nr_tasks))
 873                 return;
 874
 875         /*
 876          * See if this prio level has a good candidate
 877          */
 878         if (cfq_close_cooperator(cfqd, cfqq) &&
 879             (sample_valid(cic->ttime_samples) && cic->ttime_mean > 2))
 880                 return;
 881
 882         cfq_mark_cfqq_must_dispatch(cfqq);
 883         cfq_mark_cfqq_wait_request(cfqq);
 884
 885         /*
 886          * we don't want to idle for seeks, but we do want to allow
 887          * fair distribution of slice time for a process doing back-to-back
 888          * seeks. so allow a little bit of time for him to submit a new rq
 889          */
 890         sl = cfqd->cfq_slice_idle;
 891         if (sample_valid(cic->seek_samples) && CIC_SEEKY(cic))
 892                 sl = min(sl, msecs_to_jiffies(CFQ_MIN_TT));
 893
 894         mod_timer(&cfqd->idle_slice_timer, jiffies + sl);
 895 }
 896
 897 /*
 898  * Move request from internal lists to the request queue dispatch list.
 899  */
 900 static void cfq_dispatch_insert(struct request_queue *q, struct request *rq)
 901 {
 902         struct cfq_data *cfqd = q->elevator->elevator_data;
 903         struct cfq_queue *cfqq = RQ_CFQQ(rq);
 904
 905         cfq_remove_request(rq);
 906         cfqq->dispatched++;
 907         elv_dispatch_sort(q, rq);
 908
 909         if (cfq_cfqq_sync(cfqq))
 910                 cfqd->sync_flight++;
 911 }
 912
 913 /*
 914  * return expired entry, or NULL to just start from scratch in rbtree
 915  */
 916 static inline struct request *cfq_check_fifo(struct cfq_queue *cfqq)
 917 {
 918         struct cfq_data *cfqd = cfqq->cfqd;
 919         struct request *rq;
 920         int fifo;
 921
 922         if (cfq_cfqq_fifo_expire(cfqq))
 923                 return NULL;
 924
 925         cfq_mark_cfqq_fifo_expire(cfqq);
 926
 927         if (list_empty(&cfqq->fifo))
 928                 return NULL;
 929
 930         fifo = cfq_cfqq_sync(cfqq);
 931         rq = rq_entry_fifo(cfqq->fifo.next);
 932
 933         if (time_before(jiffies, rq->start_time + cfqd->cfq_fifo_expire[fifo]))
 934                 return NULL;
 935
 936         return rq;
 937 }
 938
 939 static inline int
 940 cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 941 {
 942         const int base_rq = cfqd->cfq_slice_async_rq;
 943
 944         WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
 945
 946         return 2 * (base_rq + base_rq * (CFQ_PRIO_LISTS - 1 - cfqq->ioprio));
 947 }
 948
 949 /*
 950  * Select a queue for service. If we have a current active queue,
 951  * check whether to continue servicing it, or retrieve and set a new one.
 952  */
 953 static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
 954 {
 955         struct cfq_queue *cfqq;
 956
 957         cfqq = cfqd->active_queue;
 958         if (!cfqq)
 959                 goto new_queue;
 960
 961         /*
 962          * The active queue has run out of time, expire it and select new.
 963          */
 964         if (cfq_slice_used(cfqq))
 965                 goto expire;
 966
 967         /*
 968          * The active queue has requests and isn't expired, allow it to
 969          * dispatch.
 970          */
 971         if (!RB_EMPTY_ROOT(&cfqq->sort_list))
 972                 goto keep_queue;
 973
 974         /*
 975          * No requests pending. If the active queue still has requests in
 976          * flight or is idling for a new request, allow either of these
 977          * conditions to happen (or time out) before selecting a new queue.
 978          */
 979         if (timer_pending(&cfqd->idle_slice_timer) ||
 980             (cfqq->dispatched && cfq_cfqq_idle_window(cfqq))) {
 981                 cfqq = NULL;
 982                 goto keep_queue;
 983         }
 984
 985 expire:
 986         cfq_slice_expired(cfqd, 0);
 987 new_queue:
 988         cfqq = cfq_set_active_queue(cfqd);
 989 keep_queue:
 990         return cfqq;
 991 }
 992
 993 /*
 994  * Dispatch some requests from cfqq, moving them to the request queue
 995  * dispatch list.
 996  */
 997 static int
 998 __cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq,
 999                         int max_dispatch)
1000 {
1001         int dispatched = 0;
1002
1003         BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list));
1004
1005         do {
1006                 struct request *rq;
1007
1008                 /*
1009                  * follow expired path, else get first next available
1010                  */
1011                 if ((rq = cfq_check_fifo(cfqq)) == NULL)
1012                         rq = cfqq->next_rq;
1013
1014                 /*
1015                  * finally, insert request into driver dispatch list
1016                  */
1017                 cfq_dispatch_insert(cfqd->queue, rq);
1018
1019                 dispatched++;
1020
1021                 if (!cfqd->active_cic) {
1022                         atomic_inc(&RQ_CIC(rq)->ioc->refcount);
1023                         cfqd->active_cic = RQ_CIC(rq);
1024                 }
1025
1026                 if (RB_EMPTY_ROOT(&cfqq->sort_list))
1027                         break;
1028
1029         } while (dispatched < max_dispatch);
1030
1031         /*
1032          * expire an async queue immediately if it has used up its slice. idle
1033          * queue always expire after 1 dispatch round.
1034          */
1035         if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) &&
1036             dispatched >= cfq_prio_to_maxrq(cfqd, cfqq)) ||
1037             cfq_class_idle(cfqq))) {
1038                 cfqq->slice_end = jiffies + 1;
1039                 cfq_slice_expired(cfqd, 0);
1040         }
1041
1042         return dispatched;
1043 }
1044
1045 static inline int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq)
1046 {
1047         int dispatched = 0;
1048
1049         while (cfqq->next_rq) {
1050                 cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq);
1051                 dispatched++;
1052         }
1053
1054         BUG_ON(!list_empty(&cfqq->fifo));
1055         return dispatched;
1056 }
1057
1058 /*
1059  * Drain our current requests. Used for barriers and when switching
1060  * io schedulers on-the-fly.
1061  */
1062 static int cfq_forced_dispatch(struct cfq_data *cfqd)
1063 {
1064         struct cfq_queue *cfqq;
1065         int dispatched = 0;
1066
1067         while ((cfqq = cfq_rb_first(&cfqd->service_tree)) != NULL)
1068                 dispatched += __cfq_forced_dispatch_cfqq(cfqq);
1069
1070         cfq_slice_expired(cfqd, 0);
1071
1072         BUG_ON(cfqd->busy_queues);
1073
1074         return dispatched;
1075 }
1076
1077 static int cfq_dispatch_requests(struct request_queue *q, int force)
1078 {
1079         struct cfq_data *cfqd = q->elevator->elevator_data;
1080         struct cfq_queue *cfqq;
1081         int dispatched;
1082
1083         if (!cfqd->busy_queues)
1084                 return 0;
1085
1086         if (unlikely(force))
1087                 return cfq_forced_dispatch(cfqd);
1088
1089         dispatched = 0;
1090         while ((cfqq = cfq_select_queue(cfqd)) != NULL) {
1091                 int max_dispatch;
1092
1093                 max_dispatch = cfqd->cfq_quantum;
1094                 if (cfq_class_idle(cfqq))
1095                         max_dispatch = 1;
1096
1097                 if (cfqq->dispatched >= max_dispatch) {
1098                         if (cfqd->busy_queues > 1)
1099                                 break;
1100                         if (cfqq->dispatched >= 4 * max_dispatch)
1101                                 break;
1102                 }
1103
1104                 if (cfqd->sync_flight && !cfq_cfqq_sync(cfqq))
1105                         break;
1106
1107                 cfq_clear_cfqq_must_dispatch(cfqq);
1108                 cfq_clear_cfqq_wait_request(cfqq);
1109                 del_timer(&cfqd->idle_slice_timer);
1110
1111                 dispatched += __cfq_dispatch_requests(cfqd, cfqq, max_dispatch);
1112         }
1113
1114         return dispatched;
1115 }
1116
1117 /*
1118  * task holds one reference to the queue, dropped when task exits. each rq
1119  * in-flight on this queue also holds a reference, dropped when rq is freed.
1120  *
1121  * queue lock must be held here.
1122  */
1123 static void cfq_put_queue(struct cfq_queue *cfqq)
1124 {
1125         struct cfq_data *cfqd = cfqq->cfqd;
1126
1127         BUG_ON(atomic_read(&cfqq->ref) <= 0);
1128
1129         if (!atomic_dec_and_test(&cfqq->ref))
1130                 return;
1131
1132         BUG_ON(rb_first(&cfqq->sort_list));
1133         BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
1134         BUG_ON(cfq_cfqq_on_rr(cfqq));
1135
1136         if (unlikely(cfqd->active_queue == cfqq)) {
1137                 __cfq_slice_expired(cfqd, cfqq, 0);
1138                 cfq_schedule_dispatch(cfqd);
1139         }
1140
1141         kmem_cache_free(cfq_pool, cfqq);
1142 }
1143
1144 /*
1145  * Call func for each cic attached to this ioc. Returns number of cic's seen.
1146  */
1147 #define CIC_GANG_NR     16
1148 static unsigned int
1149 call_for_each_cic(struct io_context *ioc,
1150                   void (*func)(struct io_context *, struct cfq_io_context *))
1151 {
1152         struct cfq_io_context *cics[CIC_GANG_NR];
1153         unsigned long index = 0;
1154         unsigned int called = 0;
1155         int nr;
1156
1157         rcu_read_lock();
1158
1159         do {
1160                 int i;
1161
1162                 /*
1163                  * Perhaps there's a better way - this just gang lookups from
1164                  * 0 to the end, restarting after each CIC_GANG_NR from the
1165                  * last key + 1.
1166                  */
1167                 nr = radix_tree_gang_lookup(&ioc->radix_root, (void **) cics,
1168                                                 index, CIC_GANG_NR);
1169                 if (!nr)
1170                         break;
1171
1172                 called += nr;
1173                 index = 1 + (unsigned long) cics[nr - 1]->key;
1174
1175                 for (i = 0; i < nr; i++)
1176                         func(ioc, cics[i]);
1177         } while (nr == CIC_GANG_NR);
1178
1179         rcu_read_unlock();
1180
1181         return called;
1182 }
1183
1184 static void cic_free_func(struct io_context *ioc, struct cfq_io_context *cic)
1185 {
1186         unsigned long flags;
1187
1188         BUG_ON(!cic->dead_key);
1189
1190         spin_lock_irqsave(&ioc->lock, flags);
1191         radix_tree_delete(&ioc->radix_root, cic->dead_key);
1192         spin_unlock_irqrestore(&ioc->lock, flags);
1193
1194         kmem_cache_free(cfq_ioc_pool, cic);
1195 }
1196
1197 static void cfq_free_io_context(struct io_context *ioc)
1198 {
1199         int freed;
1200
1201         /*
1202          * ioc->refcount is zero here, so no more cic's are allowed to be
1203          * linked into this ioc. So it should be ok to iterate over the known
1204          * list, we will see all cic's since no new ones are added.
1205          */
1206         freed = call_for_each_cic(ioc, cic_free_func);
1207
1208         elv_ioc_count_mod(ioc_count, -freed);
1209
1210         if (ioc_gone && !elv_ioc_count_read(ioc_count))
1211                 complete(ioc_gone);
1212 }
1213
1214 static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1215 {
1216         if (unlikely(cfqq == cfqd->active_queue)) {
1217                 __cfq_slice_expired(cfqd, cfqq, 0);
1218                 cfq_schedule_dispatch(cfqd);
1219         }
1220
1221         cfq_put_queue(cfqq);
1222 }
1223
1224 static void __cfq_exit_single_io_context(struct cfq_data *cfqd,
1225                                          struct cfq_io_context *cic)
1226 {
1227         list_del_init(&cic->queue_list);
1228
1229         /*
1230          * Make sure key == NULL is seen for dead queues
1231          */
1232         smp_wmb();
1233         cic->dead_key = (unsigned long) cic->key;
1234         cic->key = NULL;
1235
1236         if (cic->cfqq[ASYNC]) {
1237                 cfq_exit_cfqq(cfqd, cic->cfqq[ASYNC]);
1238                 cic->cfqq[ASYNC] = NULL;
1239         }
1240
1241         if (cic->cfqq[SYNC]) {
1242                 cfq_exit_cfqq(cfqd, cic->cfqq[SYNC]);
1243                 cic->cfqq[SYNC] = NULL;
1244         }
1245 }
1246
1247 static void cfq_exit_single_io_context(struct io_context *ioc,
1248                                        struct cfq_io_context *cic)
1249 {
1250         struct cfq_data *cfqd = cic->key;
1251
1252         if (cfqd) {
1253                 struct request_queue *q = cfqd->queue;
1254                 unsigned long flags;
1255
1256                 spin_lock_irqsave(q->queue_lock, flags);
1257                 __cfq_exit_single_io_context(cfqd, cic);
1258                 spin_unlock_irqrestore(q->queue_lock, flags);
1259         }
1260 }
1261
1262 /*
1263  * The process that ioc belongs to has exited, we need to clean up
1264  * and put the internal structures we have that belongs to that process.
1265  */
1266 static void cfq_exit_io_context(struct io_context *ioc)
1267 {
1268         rcu_assign_pointer(ioc->ioc_data, NULL);
1269         call_for_each_cic(ioc, cfq_exit_single_io_context);
1270 }
1271
1272 static struct cfq_io_context *
1273 cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
1274 {
1275         struct cfq_io_context *cic;
1276
1277         cic = kmem_cache_alloc_node(cfq_ioc_pool, gfp_mask | __GFP_ZERO,
1278                                                         cfqd->queue->node);
1279         if (cic) {
1280                 cic->last_end_request = jiffies;
1281                 INIT_LIST_HEAD(&cic->queue_list);
1282                 cic->dtor = cfq_free_io_context;
1283                 cic->exit = cfq_exit_io_context;
1284                 elv_ioc_count_inc(ioc_count);
1285         }
1286
1287         return cic;
1288 }
1289
1290 static void cfq_init_prio_data(struct cfq_queue *cfqq, struct io_context *ioc)
1291 {
1292         struct task_struct *tsk = current;
1293         int ioprio_class;
1294
1295         if (!cfq_cfqq_prio_changed(cfqq))
1296                 return;
1297
1298         ioprio_class = IOPRIO_PRIO_CLASS(ioc->ioprio);
1299         switch (ioprio_class) {
1300                 default:
1301                         printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
1302                 case IOPRIO_CLASS_NONE:
1303                         /*
1304                          * no prio set, place us in the middle of the BE classes
1305                          */
1306                         cfqq->ioprio = task_nice_ioprio(tsk);
1307                         cfqq->ioprio_class = IOPRIO_CLASS_BE;
1308                         break;
1309                 case IOPRIO_CLASS_RT:
1310                         cfqq->ioprio = task_ioprio(ioc);
1311                         cfqq->ioprio_class = IOPRIO_CLASS_RT;
1312                         break;
1313                 case IOPRIO_CLASS_BE:
1314                         cfqq->ioprio = task_ioprio(ioc);
1315                         cfqq->ioprio_class = IOPRIO_CLASS_BE;
1316                         break;
1317                 case IOPRIO_CLASS_IDLE:
1318                         cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
1319                         cfqq->ioprio = 7;
1320                         cfq_clear_cfqq_idle_window(cfqq);
1321                         break;
1322         }
1323
1324         /*
1325          * keep track of original prio settings in case we have to temporarily
1326          * elevate the priority of this queue
1327          */
1328         cfqq->org_ioprio = cfqq->ioprio;
1329         cfqq->org_ioprio_class = cfqq->ioprio_class;
1330         cfq_clear_cfqq_prio_changed(cfqq);
1331 }
1332
1333 static inline void changed_ioprio(struct io_context *ioc,
1334                                   struct cfq_io_context *cic)
1335 {
1336         struct cfq_data *cfqd = cic->key;
1337         struct cfq_queue *cfqq;
1338         unsigned long flags;
1339
1340         if (unlikely(!cfqd))
1341                 return;
1342
1343         spin_lock_irqsave(cfqd->queue->queue_lock, flags);
1344
1345         cfqq = cic->cfqq[ASYNC];
1346         if (cfqq) {
1347                 struct cfq_queue *new_cfqq;
1348                 new_cfqq = cfq_get_queue(cfqd, ASYNC, cic->ioc, GFP_ATOMIC);
1349                 if (new_cfqq) {
1350                         cic->cfqq[ASYNC] = new_cfqq;
1351                         cfq_put_queue(cfqq);
1352                 }
1353         }
1354
1355         cfqq = cic->cfqq[SYNC];
1356         if (cfqq)
1357                 cfq_mark_cfqq_prio_changed(cfqq);
1358
1359         spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
1360 }
1361
1362 static void cfq_ioc_set_ioprio(struct io_context *ioc)
1363 {
1364         call_for_each_cic(ioc, changed_ioprio);
1365         ioc->ioprio_changed = 0;
1366 }
1367
1368 static struct cfq_queue *
1369 cfq_find_alloc_queue(struct cfq_data *cfqd, int is_sync,
1370                      struct io_context *ioc, gfp_t gfp_mask)
1371 {
1372         struct cfq_queue *cfqq, *new_cfqq = NULL;
1373         struct cfq_io_context *cic;
1374
1375 retry:
1376         cic = cfq_cic_lookup(cfqd, ioc);
1377         /* cic always exists here */
1378         cfqq = cic_to_cfqq(cic, is_sync);
1379
1380         if (!cfqq) {
1381                 if (new_cfqq) {
1382                         cfqq = new_cfqq;
1383                         new_cfqq = NULL;
1384                 } else if (gfp_mask & __GFP_WAIT) {
1385                         /*
1386                          * Inform the allocator of the fact that we will
1387                          * just repeat this allocation if it fails, to allow
1388                          * the allocator to do whatever it needs to attempt to
1389                          * free memory.
1390                          */
1391                         spin_unlock_irq(cfqd->queue->queue_lock);
1392                         new_cfqq = kmem_cache_alloc_node(cfq_pool,
1393                                         gfp_mask | __GFP_NOFAIL | __GFP_ZERO,
1394                                         cfqd->queue->node);
1395                         spin_lock_irq(cfqd->queue->queue_lock);
1396                         goto retry;
1397                 } else {
1398                         cfqq = kmem_cache_alloc_node(cfq_pool,
1399                                         gfp_mask | __GFP_ZERO,
1400                                         cfqd->queue->node);
1401                         if (!cfqq)
1402                                 goto out;
1403                 }
1404
1405                 RB_CLEAR_NODE(&cfqq->rb_node);
1406                 INIT_LIST_HEAD(&cfqq->fifo);
1407
1408                 atomic_set(&cfqq->ref, 0);
1409                 cfqq->cfqd = cfqd;
1410
1411                 cfq_mark_cfqq_prio_changed(cfqq);
1412                 cfq_mark_cfqq_queue_new(cfqq);
1413
1414                 cfq_init_prio_data(cfqq, ioc);
1415
1416                 if (is_sync) {
1417                         if (!cfq_class_idle(cfqq))
1418                                 cfq_mark_cfqq_idle_window(cfqq);
1419                         cfq_mark_cfqq_sync(cfqq);
1420                 }
1421         }
1422
1423         if (new_cfqq)
1424                 kmem_cache_free(cfq_pool, new_cfqq);
1425
1426 out:
1427         WARN_ON((gfp_mask & __GFP_WAIT) && !cfqq);
1428         return cfqq;
1429 }
1430
1431 static struct cfq_queue **
1432 cfq_async_queue_prio(struct cfq_data *cfqd, int ioprio_class, int ioprio)
1433 {
1434         switch(ioprio_class) {
1435         case IOPRIO_CLASS_RT:
1436                 return &cfqd->async_cfqq[0][ioprio];
1437         case IOPRIO_CLASS_BE:
1438                 return &cfqd->async_cfqq[1][ioprio];
1439         case IOPRIO_CLASS_IDLE:
1440                 return &cfqd->async_idle_cfqq;
1441         default:
1442                 BUG();
1443         }
1444 }
1445
1446 static struct cfq_queue *
1447 cfq_get_queue(struct cfq_data *cfqd, int is_sync, struct io_context *ioc,
1448               gfp_t gfp_mask)
1449 {
1450         const int ioprio = task_ioprio(ioc);
1451         const int ioprio_class = task_ioprio_class(ioc);
1452         struct cfq_queue **async_cfqq = NULL;
1453         struct cfq_queue *cfqq = NULL;
1454
1455         if (!is_sync) {
1456                 async_cfqq = cfq_async_queue_prio(cfqd, ioprio_class, ioprio);
1457                 cfqq = *async_cfqq;
1458         }
1459
1460         if (!cfqq) {
1461                 cfqq = cfq_find_alloc_queue(cfqd, is_sync, ioc, gfp_mask);
1462                 if (!cfqq)
1463                         return NULL;
1464         }
1465
1466         /*
1467          * pin the queue now that it's allocated, scheduler exit will prune it
1468          */
1469         if (!is_sync && !(*async_cfqq)) {
1470                 atomic_inc(&cfqq->ref);
1471                 *async_cfqq = cfqq;
1472         }
1473
1474         atomic_inc(&cfqq->ref);
1475         return cfqq;
1476 }
1477
1478 static void cfq_cic_free(struct cfq_io_context *cic)
1479 {
1480         kmem_cache_free(cfq_ioc_pool, cic);
1481         elv_ioc_count_dec(ioc_count);
1482
1483         if (ioc_gone && !elv_ioc_count_read(ioc_count))
1484                 complete(ioc_gone);
1485 }
1486
1487 /*
1488  * We drop cfq io contexts lazily, so we may find a dead one.
1489  */
1490 static void
1491 cfq_drop_dead_cic(struct cfq_data *cfqd, struct io_context *ioc,
1492                   struct cfq_io_context *cic)
1493 {
1494         unsigned long flags;
1495
1496         WARN_ON(!list_empty(&cic->queue_list));
1497
1498         spin_lock_irqsave(&ioc->lock, flags);
1499
1500         if (ioc->ioc_data == cic)
1501                 rcu_assign_pointer(ioc->ioc_data, NULL);
1502
1503         radix_tree_delete(&ioc->radix_root, (unsigned long) cfqd);
1504         spin_unlock_irqrestore(&ioc->lock, flags);
1505
1506         cfq_cic_free(cic);
1507 }
1508
1509 static struct cfq_io_context *
1510 cfq_cic_lookup(struct cfq_data *cfqd, struct io_context *ioc)
1511 {
1512         struct cfq_io_context *cic;
1513         void *k;
1514
1515         if (unlikely(!ioc))
1516                 return NULL;
1517
1518         /*
1519          * we maintain a last-hit cache, to avoid browsing over the tree
1520          */
1521         cic = rcu_dereference(ioc->ioc_data);
1522         if (cic && cic->key == cfqd)
1523                 return cic;
1524
1525         do {
1526                 rcu_read_lock();
1527                 cic = radix_tree_lookup(&ioc->radix_root, (unsigned long) cfqd);
1528                 rcu_read_unlock();
1529                 if (!cic)
1530                         break;
1531                 /* ->key must be copied to avoid race with cfq_exit_queue() */
1532                 k = cic->key;
1533                 if (unlikely(!k)) {
1534                         cfq_drop_dead_cic(cfqd, ioc, cic);
1535                         continue;
1536                 }
1537
1538                 rcu_assign_pointer(ioc->ioc_data, cic);
1539                 break;
1540         } while (1);
1541
1542         return cic;
1543 }
1544
1545 /*
1546  * Add cic into ioc, using cfqd as the search key. This enables us to lookup
1547  * the process specific cfq io context when entered from the block layer.
1548  * Also adds the cic to a per-cfqd list, used when this queue is removed.
1549  */
1550 static inline int
1551 cfq_cic_link(struct cfq_data *cfqd, struct io_context *ioc,
1552              struct cfq_io_context *cic, gfp_t gfp_mask)
1553 {
1554         unsigned long flags;
1555         int ret;
1556
1557         ret = radix_tree_preload(gfp_mask);
1558         if (!ret) {
1559                 cic->ioc = ioc;
1560                 cic->key = cfqd;
1561
1562                 spin_lock_irqsave(&ioc->lock, flags);
1563                 ret = radix_tree_insert(&ioc->radix_root,
1564                                                 (unsigned long) cfqd, cic);
1565                 spin_unlock_irqrestore(&ioc->lock, flags);
1566
1567                 radix_tree_preload_end();
1568
1569                 if (!ret) {
1570                         spin_lock_irqsave(cfqd->queue->queue_lock, flags);
1571                         list_add(&cic->queue_list, &cfqd->cic_list);
1572                         spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
1573                 }
1574         }
1575
1576         if (ret)
1577                 printk(KERN_ERR "cfq: cic link failed!\n");
1578
1579         return ret;
1580 }
1581
1582 /*
1583  * Setup general io context and cfq io context. There can be several cfq
1584  * io contexts per general io context, if this process is doing io to more
1585  * than one device managed by cfq.
1586  */
1587 static struct cfq_io_context *
1588 cfq_get_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
1589 {
1590         struct io_context *ioc = NULL;
1591         struct cfq_io_context *cic;
1592
1593         might_sleep_if(gfp_mask & __GFP_WAIT);
1594
1595         ioc = get_io_context(gfp_mask, cfqd->queue->node);
1596         if (!ioc)
1597                 return NULL;
1598
1599         cic = cfq_cic_lookup(cfqd, ioc);
1600         if (cic)
1601                 goto out;
1602
1603         cic = cfq_alloc_io_context(cfqd, gfp_mask);
1604         if (cic == NULL)
1605                 goto err;
1606
1607         if (cfq_cic_link(cfqd, ioc, cic, gfp_mask))
1608                 goto err_free;
1609
1610 out:
1611         smp_read_barrier_depends();
1612         if (unlikely(ioc->ioprio_changed))
1613                 cfq_ioc_set_ioprio(ioc);
1614
1615         return cic;
1616 err_free:
1617         cfq_cic_free(cic);
1618 err:
1619         put_io_context(ioc);
1620         return NULL;
1621 }
1622
1623 static void
1624 cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_io_context *cic)
1625 {
1626         unsigned long elapsed = jiffies - cic->last_end_request;
1627         unsigned long ttime = min(elapsed, 2UL * cfqd->cfq_slice_idle);
1628
1629         cic->ttime_samples = (7*cic->ttime_samples + 256) / 8;
1630         cic->ttime_total = (7*cic->ttime_total + 256*ttime) / 8;
1631         cic->ttime_mean = (cic->ttime_total + 128) / cic->ttime_samples;
1632 }
1633
1634 static void
1635 cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_io_context *cic,
1636                        struct request *rq)
1637 {
1638         sector_t sdist;
1639         u64 total;
1640
1641         if (cic->last_request_pos < rq->sector)
1642                 sdist = rq->sector - cic->last_request_pos;
1643         else
1644                 sdist = cic->last_request_pos - rq->sector;
1645
1646         /*
1647          * Don't allow the seek distance to get too large from the
1648          * odd fragment, pagein, etc
1649          */
1650         if (cic->seek_samples <= 60) /* second&third seek */
1651                 sdist = min(sdist, (cic->seek_mean * 4) + 2*1024*1024);
1652         else
1653                 sdist = min(sdist, (cic->seek_mean * 4) + 2*1024*64);
1654
1655         cic->seek_samples = (7*cic->seek_samples + 256) / 8;
1656         cic->seek_total = (7*cic->seek_total + (u64)256*sdist) / 8;
1657         total = cic->seek_total + (cic->seek_samples/2);
1658         do_div(total, cic->seek_samples);
1659         cic->seek_mean = (sector_t)total;
1660 }
1661
1662 /*
1663  * Disable idle window if the process thinks too long or seeks so much that
1664  * it doesn't matter
1665  */
1666 static void
1667 cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1668                        struct cfq_io_context *cic)
1669 {
1670         int enable_idle;
1671
1672         /*
1673          * Don't idle for async or idle io prio class
1674          */
1675         if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq))
1676                 return;
1677
1678         enable_idle = cfq_cfqq_idle_window(cfqq);
1679
1680         if (!atomic_read(&cic->ioc->nr_tasks) || !cfqd->cfq_slice_idle ||
1681             (cfqd->hw_tag && CIC_SEEKY(cic)))
1682                 enable_idle = 0;
1683         else if (sample_valid(cic->ttime_samples)) {
1684                 if (cic->ttime_mean > cfqd->cfq_slice_idle)
1685                         enable_idle = 0;
1686                 else
1687                         enable_idle = 1;
1688         }
1689
1690         if (enable_idle)
1691                 cfq_mark_cfqq_idle_window(cfqq);
1692         else
1693                 cfq_clear_cfqq_idle_window(cfqq);
1694 }
1695
1696 /*
1697  * Check if new_cfqq should preempt the currently active queue. Return 0 for
1698  * no or if we aren't sure, a 1 will cause a preempt.
1699  */
1700 static int
1701 cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
1702                    struct request *rq)
1703 {
1704         struct cfq_queue *cfqq;
1705
1706         cfqq = cfqd->active_queue;
1707         if (!cfqq)
1708                 return 0;
1709
1710         if (cfq_slice_used(cfqq))
1711                 return 1;
1712
1713         if (cfq_class_idle(new_cfqq))
1714                 return 0;
1715
1716         if (cfq_class_idle(cfqq))
1717                 return 1;
1718
1719         /*
1720          * if the new request is sync, but the currently running queue is
1721          * not, let the sync request have priority.
1722          */
1723         if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq))
1724                 return 1;
1725
1726         /*
1727          * So both queues are sync. Let the new request get disk time if
1728          * it's a metadata request and the current queue is doing regular IO.
1729          */
1730         if (rq_is_meta(rq) && !cfqq->meta_pending)
1731                 return 1;
1732
1733         if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq))
1734                 return 0;
1735
1736         /*
1737          * if this request is as-good as one we would expect from the
1738          * current cfqq, let it preempt
1739          */
1740         if (cfq_rq_close(cfqd, rq))
1741                 return 1;
1742
1743         return 0;
1744 }
1745
1746 /*
1747  * cfqq preempts the active queue. if we allowed preempt with no slice left,
1748  * let it have half of its nominal slice.
1749  */
1750 static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1751 {
1752         cfq_slice_expired(cfqd, 1);
1753
1754         /*
1755          * Put the new queue at the front of the of the current list,
1756          * so we know that it will be selected next.
1757          */
1758         BUG_ON(!cfq_cfqq_on_rr(cfqq));
1759
1760         cfq_service_tree_add(cfqd, cfqq, 1);
1761
1762         cfqq->slice_end = 0;
1763         cfq_mark_cfqq_slice_new(cfqq);
1764 }
1765
1766 /*
1767  * Called when a new fs request (rq) is added (to cfqq). Check if there's
1768  * something we should do about it
1769  */
1770 static void
1771 cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1772                 struct request *rq)
1773 {
1774         struct cfq_io_context *cic = RQ_CIC(rq);
1775
1776         if (rq_is_meta(rq))
1777                 cfqq->meta_pending++;
1778
1779         cfq_update_io_thinktime(cfqd, cic);
1780         cfq_update_io_seektime(cfqd, cic, rq);
1781         cfq_update_idle_window(cfqd, cfqq, cic);
1782
1783         cic->last_request_pos = rq->sector + rq->nr_sectors;
1784
1785         if (cfqq == cfqd->active_queue) {
1786                 /*
1787                  * if we are waiting for a request for this queue, let it rip
1788                  * immediately and flag that we must not expire this queue
1789                  * just now
1790                  */
1791                 if (cfq_cfqq_wait_request(cfqq)) {
1792                         cfq_mark_cfqq_must_dispatch(cfqq);
1793                         del_timer(&cfqd->idle_slice_timer);
1794                         blk_start_queueing(cfqd->queue);
1795                 }
1796         } else if (cfq_should_preempt(cfqd, cfqq, rq)) {
1797                 /*
1798                  * not the active queue - expire current slice if it is
1799                  * idle and has expired it's mean thinktime or this new queue
1800                  * has some old slice time left and is of higher priority
1801                  */
1802                 cfq_preempt_queue(cfqd, cfqq);
1803                 cfq_mark_cfqq_must_dispatch(cfqq);
1804                 blk_start_queueing(cfqd->queue);
1805         }
1806 }
1807
1808 static void cfq_insert_request(struct request_queue *q, struct request *rq)
1809 {
1810         struct cfq_data *cfqd = q->elevator->elevator_data;
1811         struct cfq_queue *cfqq = RQ_CFQQ(rq);
1812
1813         cfq_init_prio_data(cfqq, RQ_CIC(rq)->ioc);
1814
1815         cfq_add_rq_rb(rq);
1816
1817         list_add_tail(&rq->queuelist, &cfqq->fifo);
1818
1819         cfq_rq_enqueued(cfqd, cfqq, rq);
1820 }
1821
1822 static void cfq_completed_request(struct request_queue *q, struct request *rq)
1823 {
1824         struct cfq_queue *cfqq = RQ_CFQQ(rq);
1825         struct cfq_data *cfqd = cfqq->cfqd;
1826         const int sync = rq_is_sync(rq);
1827         unsigned long now;
1828
1829         now = jiffies;
1830
1831         WARN_ON(!cfqd->rq_in_driver);
1832         WARN_ON(!cfqq->dispatched);
1833         cfqd->rq_in_driver--;
1834         cfqq->dispatched--;
1835
1836         if (cfq_cfqq_sync(cfqq))
1837                 cfqd->sync_flight--;
1838
1839         if (!cfq_class_idle(cfqq))
1840                 cfqd->last_end_request = now;
1841
1842         if (sync)
1843                 RQ_CIC(rq)->last_end_request = now;
1844
1845         /*
1846          * If this is the active queue, check if it needs to be expired,
1847          * or if we want to idle in case it has no pending requests.
1848          */
1849         if (cfqd->active_queue == cfqq) {
1850                 if (cfq_cfqq_slice_new(cfqq)) {
1851                         cfq_set_prio_slice(cfqd, cfqq);
1852                         cfq_clear_cfqq_slice_new(cfqq);
1853                 }
1854                 if (cfq_slice_used(cfqq) || cfq_class_idle(cfqq))
1855                         cfq_slice_expired(cfqd, 1);
1856                 else if (sync && RB_EMPTY_ROOT(&cfqq->sort_list))
1857                         cfq_arm_slice_timer(cfqd);
1858         }
1859
1860         if (!cfqd->rq_in_driver)
1861                 cfq_schedule_dispatch(cfqd);
1862 }
1863
1864 /*
1865  * we temporarily boost lower priority queues if they are holding fs exclusive
1866  * resources. they are boosted to normal prio (CLASS_BE/4)
1867  */
1868 static void cfq_prio_boost(struct cfq_queue *cfqq)
1869 {
1870         if (has_fs_excl()) {
1871                 /*
1872                  * boost idle prio on transactions that would lock out other
1873                  * users of the filesystem
1874                  */
1875                 if (cfq_class_idle(cfqq))
1876                         cfqq->ioprio_class = IOPRIO_CLASS_BE;
1877                 if (cfqq->ioprio > IOPRIO_NORM)
1878                         cfqq->ioprio = IOPRIO_NORM;
1879         } else {
1880                 /*
1881                  * check if we need to unboost the queue
1882                  */
1883                 if (cfqq->ioprio_class != cfqq->org_ioprio_class)
1884                         cfqq->ioprio_class = cfqq->org_ioprio_class;
1885                 if (cfqq->ioprio != cfqq->org_ioprio)
1886                         cfqq->ioprio = cfqq->org_ioprio;
1887         }
1888 }
1889
1890 static inline int __cfq_may_queue(struct cfq_queue *cfqq)
1891 {
1892         if ((cfq_cfqq_wait_request(cfqq) || cfq_cfqq_must_alloc(cfqq)) &&
1893             !cfq_cfqq_must_alloc_slice(cfqq)) {
1894                 cfq_mark_cfqq_must_alloc_slice(cfqq);
1895                 return ELV_MQUEUE_MUST;
1896         }
1897
1898         return ELV_MQUEUE_MAY;
1899 }
1900
1901 static int cfq_may_queue(struct request_queue *q, int rw)
1902 {
1903         struct cfq_data *cfqd = q->elevator->elevator_data;
1904         struct task_struct *tsk = current;
1905         struct cfq_io_context *cic;
1906         struct cfq_queue *cfqq;
1907
1908         /*
1909          * don't force setup of a queue from here, as a call to may_queue
1910          * does not necessarily imply that a request actually will be queued.
1911          * so just lookup a possibly existing queue, or return 'may queue'
1912          * if that fails
1913          */
1914         cic = cfq_cic_lookup(cfqd, tsk->io_context);
1915         if (!cic)
1916                 return ELV_MQUEUE_MAY;
1917
1918         cfqq = cic_to_cfqq(cic, rw & REQ_RW_SYNC);
1919         if (cfqq) {
1920                 cfq_init_prio_data(cfqq, cic->ioc);
1921                 cfq_prio_boost(cfqq);
1922
1923                 return __cfq_may_queue(cfqq);
1924         }
1925
1926         return ELV_MQUEUE_MAY;
1927 }
1928
1929 /*
1930  * queue lock held here
1931  */
1932 static void cfq_put_request(struct request *rq)
1933 {
1934         struct cfq_queue *cfqq = RQ_CFQQ(rq);
1935
1936         if (cfqq) {
1937                 const int rw = rq_data_dir(rq);
1938
1939                 BUG_ON(!cfqq->allocated[rw]);
1940                 cfqq->allocated[rw]--;
1941
1942                 put_io_context(RQ_CIC(rq)->ioc);
1943
1944                 rq->elevator_private = NULL;
1945                 rq->elevator_private2 = NULL;
1946
1947                 cfq_put_queue(cfqq);
1948         }
1949 }
1950
1951 /*
1952  * Allocate cfq data structures associated with this request.
1953  */
1954 static int
1955 cfq_set_request(struct request_queue *q, struct request *rq, gfp_t gfp_mask)
1956 {
1957         struct cfq_data *cfqd = q->elevator->elevator_data;
1958         struct cfq_io_context *cic;
1959         const int rw = rq_data_dir(rq);
1960         const int is_sync = rq_is_sync(rq);
1961         struct cfq_queue *cfqq;
1962         unsigned long flags;
1963
1964         might_sleep_if(gfp_mask & __GFP_WAIT);
1965
1966         cic = cfq_get_io_context(cfqd, gfp_mask);
1967
1968         spin_lock_irqsave(q->queue_lock, flags);
1969
1970         if (!cic)
1971                 goto queue_fail;
1972
1973         cfqq = cic_to_cfqq(cic, is_sync);
1974         if (!cfqq) {
1975                 cfqq = cfq_get_queue(cfqd, is_sync, cic->ioc, gfp_mask);
1976
1977                 if (!cfqq)
1978                         goto queue_fail;
1979
1980                 cic_set_cfqq(cic, cfqq, is_sync);
1981         }
1982
1983         cfqq->allocated[rw]++;
1984         cfq_clear_cfqq_must_alloc(cfqq);
1985         atomic_inc(&cfqq->ref);
1986
1987         spin_unlock_irqrestore(q->queue_lock, flags);
1988
1989         rq->elevator_private = cic;
1990         rq->elevator_private2 = cfqq;
1991         return 0;
1992
1993 queue_fail:
1994         if (cic)
1995                 put_io_context(cic->ioc);
1996
1997         cfq_schedule_dispatch(cfqd);
1998         spin_unlock_irqrestore(q->queue_lock, flags);
1999         return 1;
2000 }
2001
2002 static void cfq_kick_queue(struct work_struct *work)
2003 {
2004         struct cfq_data *cfqd =
2005                 container_of(work, struct cfq_data, unplug_work);
2006         struct request_queue *q = cfqd->queue;
2007         unsigned long flags;
2008
2009         spin_lock_irqsave(q->queue_lock, flags);
2010         blk_start_queueing(q);
2011         spin_unlock_irqrestore(q->queue_lock, flags);
2012 }
2013
2014 /*
2015  * Timer running if the active_queue is currently idling inside its time slice
2016  */
2017 static void cfq_idle_slice_timer(unsigned long data)
2018 {
2019         struct cfq_data *cfqd = (struct cfq_data *) data;
2020         struct cfq_queue *cfqq;
2021         unsigned long flags;
2022         int timed_out = 1;
2023
2024         spin_lock_irqsave(cfqd->queue->queue_lock, flags);
2025
2026         if ((cfqq = cfqd->active_queue) != NULL) {
2027                 timed_out = 0;
2028
2029                 /*
2030                  * expired
2031                  */
2032                 if (cfq_slice_used(cfqq))
2033                         goto expire;
2034
2035                 /*
2036                  * only expire and reinvoke request handler, if there are
2037                  * other queues with pending requests
2038                  */
2039                 if (!cfqd->busy_queues)
2040                         goto out_cont;
2041
2042                 /*
2043                  * not expired and it has a request pending, let it dispatch
2044                  */
2045                 if (!RB_EMPTY_ROOT(&cfqq->sort_list)) {
2046                         cfq_mark_cfqq_must_dispatch(cfqq);
2047                         goto out_kick;
2048                 }
2049         }
2050 expire:
2051         cfq_slice_expired(cfqd, timed_out);
2052 out_kick:
2053         cfq_schedule_dispatch(cfqd);
2054 out_cont:
2055         spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
2056 }
2057
2058 static void cfq_shutdown_timer_wq(struct cfq_data *cfqd)
2059 {
2060         del_timer_sync(&cfqd->idle_slice_timer);
2061         kblockd_flush_work(&cfqd->unplug_work);
2062 }
2063
2064 static void cfq_put_async_queues(struct cfq_data *cfqd)
2065 {
2066         int i;
2067
2068         for (i = 0; i < IOPRIO_BE_NR; i++) {
2069                 if (cfqd->async_cfqq[0][i])
2070                         cfq_put_queue(cfqd->async_cfqq[0][i]);
2071                 if (cfqd->async_cfqq[1][i])
2072                         cfq_put_queue(cfqd->async_cfqq[1][i]);
2073         }
2074
2075         if (cfqd->async_idle_cfqq)
2076                 cfq_put_queue(cfqd->async_idle_cfqq);
2077 }
2078
2079 static void cfq_exit_queue(elevator_t *e)
2080 {
2081         struct cfq_data *cfqd = e->elevator_data;
2082         struct request_queue *q = cfqd->queue;
2083
2084         cfq_shutdown_timer_wq(cfqd);
2085
2086         spin_lock_irq(q->queue_lock);
2087
2088         if (cfqd->active_queue)
2089                 __cfq_slice_expired(cfqd, cfqd->active_queue, 0);
2090
2091         while (!list_empty(&cfqd->cic_list)) {
2092                 struct cfq_io_context *cic = list_entry(cfqd->cic_list.next,
2093                                                         struct cfq_io_context,
2094                                                         queue_list);
2095
2096                 __cfq_exit_single_io_context(cfqd, cic);
2097         }
2098
2099         cfq_put_async_queues(cfqd);
2100
2101         spin_unlock_irq(q->queue_lock);
2102
2103         cfq_shutdown_timer_wq(cfqd);
2104
2105         kfree(cfqd);
2106 }
2107
2108 static void *cfq_init_queue(struct request_queue *q)
2109 {
2110         struct cfq_data *cfqd;
2111
2112         cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL | __GFP_ZERO, q->node);
2113         if (!cfqd)
2114                 return NULL;
2115
2116         cfqd->service_tree = CFQ_RB_ROOT;
2117         INIT_LIST_HEAD(&cfqd->cic_list);
2118
2119         cfqd->queue = q;
2120
2121         init_timer(&cfqd->idle_slice_timer);
2122         cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
2123         cfqd->idle_slice_timer.data = (unsigned long) cfqd;
2124
2125         INIT_WORK(&cfqd->unplug_work, cfq_kick_queue);
2126
2127         cfqd->last_end_request = jiffies;
2128         cfqd->cfq_quantum = cfq_quantum;
2129         cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0];
2130         cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1];
2131         cfqd->cfq_back_max = cfq_back_max;
2132         cfqd->cfq_back_penalty = cfq_back_penalty;
2133         cfqd->cfq_slice[0] = cfq_slice_async;
2134         cfqd->cfq_slice[1] = cfq_slice_sync;
2135         cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
2136         cfqd->cfq_slice_idle = cfq_slice_idle;
2137
2138         return cfqd;
2139 }
2140
2141 static void cfq_slab_kill(void)
2142 {
2143         if (cfq_pool)
2144                 kmem_cache_destroy(cfq_pool);
2145         if (cfq_ioc_pool)
2146                 kmem_cache_destroy(cfq_ioc_pool);
2147 }
2148
2149 static int __init cfq_slab_setup(void)
2150 {
2151         cfq_pool = KMEM_CACHE(cfq_queue, 0);
2152         if (!cfq_pool)
2153                 goto fail;
2154
2155         cfq_ioc_pool = KMEM_CACHE(cfq_io_context, SLAB_DESTROY_BY_RCU);
2156         if (!cfq_ioc_pool)
2157                 goto fail;
2158
2159         return 0;
2160 fail:
2161         cfq_slab_kill();
2162         return -ENOMEM;
2163 }
2164
2165 /*
2166  * sysfs parts below -->
2167  */
2168 static ssize_t
2169 cfq_var_show(unsigned int var, char *page)
2170 {
2171         return sprintf(page, "%d\n", var);
2172 }
2173
2174 static ssize_t
2175 cfq_var_store(unsigned int *var, const char *page, size_t count)
2176 {
2177         char *p = (char *) page;
2178
2179         *var = simple_strtoul(p, &p, 10);
2180         return count;
2181 }
2182
2183 #define SHOW_FUNCTION(__FUNC, __VAR, __CONV)                            \
2184 static ssize_t __FUNC(elevator_t *e, char *page)                        \
2185 {                                                                       \
2186         struct cfq_data *cfqd = e->elevator_data;                       \
2187         unsigned int __data = __VAR;                                    \
2188         if (__CONV)                                                     \
2189                 __data = jiffies_to_msecs(__data);                      \
2190         return cfq_var_show(__data, (page));                            \
2191 }
2192 SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0);
2193 SHOW_FUNCTION(cfq_fifo_expire_sync_show, cfqd->cfq_fifo_expire[1], 1);
2194 SHOW_FUNCTION(cfq_fifo_expire_async_show, cfqd->cfq_fifo_expire[0], 1);
2195 SHOW_FUNCTION(cfq_back_seek_max_show, cfqd->cfq_back_max, 0);
2196 SHOW_FUNCTION(cfq_back_seek_penalty_show, cfqd->cfq_back_penalty, 0);
2197 SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1);
2198 SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1);
2199 SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
2200 SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
2201 #undef SHOW_FUNCTION
2202
2203 #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV)                 \
2204 static ssize_t __FUNC(elevator_t *e, const char *page, size_t count)    \
2205 {                                                                       \
2206         struct cfq_data *cfqd = e->elevator_data;                       \
2207         unsigned int __data;                                            \
2208         int ret = cfq_var_store(&__data, (page), count);                \
2209         if (__data < (MIN))                                             \
2210                 __data = (MIN);                                         \
2211         else if (__data > (MAX))                                        \
2212                 __data = (MAX);                                         \
2213         if (__CONV)                                                     \
2214                 *(__PTR) = msecs_to_jiffies(__data);                    \
2215         else                                                            \
2216                 *(__PTR) = __data;                                      \
2217         return ret;                                                     \
2218 }
2219 STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
2220 STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1, UINT_MAX, 1);
2221 STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1, UINT_MAX, 1);
2222 STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
2223 STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1, UINT_MAX, 0);
2224 STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1);
2225 STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1);
2226 STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1);
2227 STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1, UINT_MAX, 0);
2228 #undef STORE_FUNCTION
2229
2230 #define CFQ_ATTR(name) \
2231         __ATTR(name, S_IRUGO|S_IWUSR, cfq_##name##_show, cfq_##name##_store)
2232
2233 static struct elv_fs_entry cfq_attrs[] = {
2234         CFQ_ATTR(quantum),
2235         CFQ_ATTR(fifo_expire_sync),
2236         CFQ_ATTR(fifo_expire_async),
2237         CFQ_ATTR(back_seek_max),
2238         CFQ_ATTR(back_seek_penalty),
2239         CFQ_ATTR(slice_sync),
2240         CFQ_ATTR(slice_async),
2241         CFQ_ATTR(slice_async_rq),
2242         CFQ_ATTR(slice_idle),
2243         __ATTR_NULL
2244 };
2245
2246 static struct elevator_type iosched_cfq = {
2247         .ops = {
2248                 .elevator_merge_fn =            cfq_merge,
2249                 .elevator_merged_fn =           cfq_merged_request,
2250                 .elevator_merge_req_fn =        cfq_merged_requests,
2251                 .elevator_allow_merge_fn =      cfq_allow_merge,
2252                 .elevator_dispatch_fn =         cfq_dispatch_requests,
2253                 .elevator_add_req_fn =          cfq_insert_request,
2254                 .elevator_activate_req_fn =     cfq_activate_request,
2255                 .elevator_deactivate_req_fn =   cfq_deactivate_request,
2256                 .elevator_queue_empty_fn =      cfq_queue_empty,
2257                 .elevator_completed_req_fn =    cfq_completed_request,
2258                 .elevator_former_req_fn =       elv_rb_former_request,
2259                 .elevator_latter_req_fn =       elv_rb_latter_request,
2260                 .elevator_set_req_fn =          cfq_set_request,
2261                 .elevator_put_req_fn =          cfq_put_request,
2262                 .elevator_may_queue_fn =        cfq_may_queue,
2263                 .elevator_init_fn =             cfq_init_queue,
2264                 .elevator_exit_fn =             cfq_exit_queue,
2265                 .trim =                         cfq_free_io_context,
2266         },
2267         .elevator_attrs =       cfq_attrs,
2268         .elevator_name =        "cfq",
2269         .elevator_owner =       THIS_MODULE,
2270 };
2271
2272 static int __init cfq_init(void)
2273 {
2274         /*
2275          * could be 0 on HZ < 1000 setups
2276          */
2277         if (!cfq_slice_async)
2278                 cfq_slice_async = 1;
2279         if (!cfq_slice_idle)
2280                 cfq_slice_idle = 1;
2281
2282         if (cfq_slab_setup())
2283                 return -ENOMEM;
2284
2285         elv_register(&iosched_cfq);
2286
2287         return 0;
2288 }
2289
2290 static void __exit cfq_exit(void)
2291 {
2292         DECLARE_COMPLETION_ONSTACK(all_gone);
2293         elv_unregister(&iosched_cfq);
2294         ioc_gone = &all_gone;
2295         /* ioc_gone's update must be visible before reading ioc_count */
2296         smp_wmb();
2297         if (elv_ioc_count_read(ioc_count))
2298                 wait_for_completion(ioc_gone);
2299         synchronize_rcu();
2300         cfq_slab_kill();
2301 }
2302
2303 module_init(cfq_init);
2304 module_exit(cfq_exit);
2305
2306 MODULE_AUTHOR("Jens Axboe");
2307 MODULE_LICENSE("GPL");
2308 MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler");