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/slab.h>
  11 #include <linux/blkdev.h>
  12 #include <linux/elevator.h>
  13 #include <linux/jiffies.h>
  14 #include <linux/rbtree.h>
  15 #include <linux/ioprio.h>
  16 #include <linux/blktrace_api.h>
  17 #include "blk.h"
  18 #include "blk-cgroup.h"
  19
  20 /*
  21  * tunables
  22  */
  23 /* max queue in one round of service */
  24 static const int cfq_quantum = 8;
  25 static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 };
  26 /* maximum backwards seek, in KiB */
  27 static const int cfq_back_max = 16 * 1024;
  28 /* penalty of a backwards seek */
  29 static const int cfq_back_penalty = 2;
  30 static const int cfq_slice_sync = HZ / 10;
  31 static int cfq_slice_async = HZ / 25;
  32 static const int cfq_slice_async_rq = 2;
  33 static int cfq_slice_idle = HZ / 125;
  34 static int cfq_group_idle = HZ / 125;
  35 static const int cfq_target_latency = HZ * 3/10; /* 300 ms */
  36 static const int cfq_hist_divisor = 4;
  37
  38 /*
  39  * offset from end of service tree
  40  */
  41 #define CFQ_IDLE_DELAY          (HZ / 5)
  42
  43 /*
  44  * below this threshold, we consider thinktime immediate
  45  */
  46 #define CFQ_MIN_TT              (2)
  47
  48 #define CFQ_SLICE_SCALE         (5)
  49 #define CFQ_HW_QUEUE_MIN        (5)
  50 #define CFQ_SERVICE_SHIFT       12
  51
  52 #define CFQQ_SEEK_THR           (sector_t)(8 * 100)
  53 #define CFQQ_CLOSE_THR          (sector_t)(8 * 1024)
  54 #define CFQQ_SECT_THR_NONROT    (sector_t)(2 * 32)
  55 #define CFQQ_SEEKY(cfqq)        (hweight32(cfqq->seek_history) > 32/8)
  56
  57 #define RQ_CIC(rq)              icq_to_cic((rq)->elv.icq)
  58 #define RQ_CFQQ(rq)             (struct cfq_queue *) ((rq)->elv.priv[0])
  59 #define RQ_CFQG(rq)             (struct cfq_group *) ((rq)->elv.priv[1])
  60
  61 static struct kmem_cache *cfq_pool;
  62
  63 #define CFQ_PRIO_LISTS          IOPRIO_BE_NR
  64 #define cfq_class_idle(cfqq)    ((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
  65 #define cfq_class_rt(cfqq)      ((cfqq)->ioprio_class == IOPRIO_CLASS_RT)
  66
  67 #define sample_valid(samples)   ((samples) > 80)
  68 #define rb_entry_cfqg(node)     rb_entry((node), struct cfq_group, rb_node)
  69
  70 struct cfq_ttime {
  71         unsigned long last_end_request;
  72
  73         unsigned long ttime_total;
  74         unsigned long ttime_samples;
  75         unsigned long ttime_mean;
  76 };
  77
  78 /*
  79  * Most of our rbtree usage is for sorting with min extraction, so
  80  * if we cache the leftmost node we don't have to walk down the tree
  81  * to find it. Idea borrowed from Ingo Molnars CFS scheduler. We should
  82  * move this into the elevator for the rq sorting as well.
  83  */
  84 struct cfq_rb_root {
  85         struct rb_root rb;
  86         struct rb_node *left;
  87         unsigned count;
  88         u64 min_vdisktime;
  89         struct cfq_ttime ttime;
  90 };
  91 #define CFQ_RB_ROOT     (struct cfq_rb_root) { .rb = RB_ROOT, \
  92                         .ttime = {.last_end_request = jiffies,},}
  93
  94 /*
  95  * Per process-grouping structure
  96  */
  97 struct cfq_queue {
  98         /* reference count */
  99         int ref;
 100         /* various state flags, see below */
 101         unsigned int flags;
 102         /* parent cfq_data */
 103         struct cfq_data *cfqd;
 104         /* service_tree member */
 105         struct rb_node rb_node;
 106         /* service_tree key */
 107         unsigned long rb_key;
 108         /* prio tree member */
 109         struct rb_node p_node;
 110         /* prio tree root we belong to, if any */
 111         struct rb_root *p_root;
 112         /* sorted list of pending requests */
 113         struct rb_root sort_list;
 114         /* if fifo isn't expired, next request to serve */
 115         struct request *next_rq;
 116         /* requests queued in sort_list */
 117         int queued[2];
 118         /* currently allocated requests */
 119         int allocated[2];
 120         /* fifo list of requests in sort_list */
 121         struct list_head fifo;
 122
 123         /* time when queue got scheduled in to dispatch first request. */
 124         unsigned long dispatch_start;
 125         unsigned int allocated_slice;
 126         unsigned int slice_dispatch;
 127         /* time when first request from queue completed and slice started. */
 128         unsigned long slice_start;
 129         unsigned long slice_end;
 130         long slice_resid;
 131
 132         /* pending priority requests */
 133         int prio_pending;
 134         /* number of requests that are on the dispatch list or inside driver */
 135         int dispatched;
 136
 137         /* io prio of this group */
 138         unsigned short ioprio, org_ioprio;
 139         unsigned short ioprio_class;
 140
 141         pid_t pid;
 142
 143         u32 seek_history;
 144         sector_t last_request_pos;
 145
 146         struct cfq_rb_root *service_tree;
 147         struct cfq_queue *new_cfqq;
 148         struct cfq_group *cfqg;
 149         /* Number of sectors dispatched from queue in single dispatch round */
 150         unsigned long nr_sectors;
 151 };
 152
 153 /*
 154  * First index in the service_trees.
 155  * IDLE is handled separately, so it has negative index
 156  */
 157 enum wl_class_t {
 158         BE_WORKLOAD = 0,
 159         RT_WORKLOAD = 1,
 160         IDLE_WORKLOAD = 2,
 161         CFQ_PRIO_NR,
 162 };
 163
 164 /*
 165  * Second index in the service_trees.
 166  */
 167 enum wl_type_t {
 168         ASYNC_WORKLOAD = 0,
 169         SYNC_NOIDLE_WORKLOAD = 1,
 170         SYNC_WORKLOAD = 2
 171 };
 172
 173 struct cfqg_stats {
 174 #ifdef CONFIG_CFQ_GROUP_IOSCHED
 175         /* total bytes transferred */
 176         struct blkg_rwstat              service_bytes;
 177         /* total IOs serviced, post merge */
 178         struct blkg_rwstat              serviced;
 179         /* number of ios merged */
 180         struct blkg_rwstat              merged;
 181         /* total time spent on device in ns, may not be accurate w/ queueing */
 182         struct blkg_rwstat              service_time;
 183         /* total time spent waiting in scheduler queue in ns */
 184         struct blkg_rwstat              wait_time;
 185         /* number of IOs queued up */
 186         struct blkg_rwstat              queued;
 187         /* total sectors transferred */
 188         struct blkg_stat                sectors;
 189         /* total disk time and nr sectors dispatched by this group */
 190         struct blkg_stat                time;
 191 #ifdef CONFIG_DEBUG_BLK_CGROUP
 192         /* time not charged to this cgroup */
 193         struct blkg_stat                unaccounted_time;
 194         /* sum of number of ios queued across all samples */
 195         struct blkg_stat                avg_queue_size_sum;
 196         /* count of samples taken for average */
 197         struct blkg_stat                avg_queue_size_samples;
 198         /* how many times this group has been removed from service tree */
 199         struct blkg_stat                dequeue;
 200         /* total time spent waiting for it to be assigned a timeslice. */
 201         struct blkg_stat                group_wait_time;
 202         /* time spent idling for this blkcg_gq */
 203         struct blkg_stat                idle_time;
 204         /* total time with empty current active q with other requests queued */
 205         struct blkg_stat                empty_time;
 206         /* fields after this shouldn't be cleared on stat reset */
 207         uint64_t                        start_group_wait_time;
 208         uint64_t                        start_idle_time;
 209         uint64_t                        start_empty_time;
 210         uint16_t                        flags;
 211 #endif  /* CONFIG_DEBUG_BLK_CGROUP */
 212 #endif  /* CONFIG_CFQ_GROUP_IOSCHED */
 213 };
 214
 215 /* This is per cgroup per device grouping structure */
 216 struct cfq_group {
 217         /* must be the first member */
 218         struct blkg_policy_data pd;
 219
 220         /* group service_tree member */
 221         struct rb_node rb_node;
 222
 223         /* group service_tree key */
 224         u64 vdisktime;
 225
 226         /*
 227          * The number of active cfqgs and sum of their weights under this
 228          * cfqg.  This covers this cfqg's leaf_weight and all children's
 229          * weights, but does not cover weights of further descendants.
 230          *
 231          * If a cfqg is on the service tree, it's active.  An active cfqg
 232          * also activates its parent and contributes to the children_weight
 233          * of the parent.
 234          */
 235         int nr_active;
 236         unsigned int children_weight;
 237
 238         /*
 239          * vfraction is the fraction of vdisktime that the tasks in this
 240          * cfqg are entitled to.  This is determined by compounding the
 241          * ratios walking up from this cfqg to the root.
 242          *
 243          * It is in fixed point w/ CFQ_SERVICE_SHIFT and the sum of all
 244          * vfractions on a service tree is approximately 1.  The sum may
 245          * deviate a bit due to rounding errors and fluctuations caused by
 246          * cfqgs entering and leaving the service tree.
 247          */
 248         unsigned int vfraction;
 249
 250         /*
 251          * There are two weights - (internal) weight is the weight of this
 252          * cfqg against the sibling cfqgs.  leaf_weight is the wight of
 253          * this cfqg against the child cfqgs.  For the root cfqg, both
 254          * weights are kept in sync for backward compatibility.
 255          */
 256         unsigned int weight;
 257         unsigned int new_weight;
 258         unsigned int dev_weight;
 259
 260         unsigned int leaf_weight;
 261         unsigned int new_leaf_weight;
 262         unsigned int dev_leaf_weight;
 263
 264         /* number of cfqq currently on this group */
 265         int nr_cfqq;
 266
 267         /*
 268          * Per group busy queues average. Useful for workload slice calc. We
 269          * create the array for each prio class but at run time it is used
 270          * only for RT and BE class and slot for IDLE class remains unused.
 271          * This is primarily done to avoid confusion and a gcc warning.
 272          */
 273         unsigned int busy_queues_avg[CFQ_PRIO_NR];
 274         /*
 275          * rr lists of queues with requests. We maintain service trees for
 276          * RT and BE classes. These trees are subdivided in subclasses
 277          * of SYNC, SYNC_NOIDLE and ASYNC based on workload type. For IDLE
 278          * class there is no subclassification and all the cfq queues go on
 279          * a single tree service_tree_idle.
 280          * Counts are embedded in the cfq_rb_root
 281          */
 282         struct cfq_rb_root service_trees[2][3];
 283         struct cfq_rb_root service_tree_idle;
 284
 285         unsigned long saved_wl_slice;
 286         enum wl_type_t saved_wl_type;
 287         enum wl_class_t saved_wl_class;
 288
 289         /* number of requests that are on the dispatch list or inside driver */
 290         int dispatched;
 291         struct cfq_ttime ttime;
 292         struct cfqg_stats stats;        /* stats for this cfqg */
 293         struct cfqg_stats dead_stats;   /* stats pushed from dead children */
 294 };
 295
 296 struct cfq_io_cq {
 297         struct io_cq            icq;            /* must be the first member */
 298         struct cfq_queue        *cfqq[2];
 299         struct cfq_ttime        ttime;
 300         int                     ioprio;         /* the current ioprio */
 301 #ifdef CONFIG_CFQ_GROUP_IOSCHED
 302         uint64_t                blkcg_id;       /* the current blkcg ID */
 303 #endif
 304 };
 305
 306 /*
 307  * Per block device queue structure
 308  */
 309 struct cfq_data {
 310         struct request_queue *queue;
 311         /* Root service tree for cfq_groups */
 312         struct cfq_rb_root grp_service_tree;
 313         struct cfq_group *root_group;
 314
 315         /*
 316          * The priority currently being served
 317          */
 318         enum wl_class_t serving_wl_class;
 319         enum wl_type_t serving_wl_type;
 320         unsigned long workload_expires;
 321         struct cfq_group *serving_group;
 322
 323         /*
 324          * Each priority tree is sorted by next_request position.  These
 325          * trees are used when determining if two or more queues are
 326          * interleaving requests (see cfq_close_cooperator).
 327          */
 328         struct rb_root prio_trees[CFQ_PRIO_LISTS];
 329
 330         unsigned int busy_queues;
 331         unsigned int busy_sync_queues;
 332
 333         int rq_in_driver;
 334         int rq_in_flight[2];
 335
 336         /*
 337          * queue-depth detection
 338          */
 339         int rq_queued;
 340         int hw_tag;
 341         /*
 342          * hw_tag can be
 343          * -1 => indeterminate, (cfq will behave as if NCQ is present, to allow better detection)
 344          *  1 => NCQ is present (hw_tag_est_depth is the estimated max depth)
 345          *  0 => no NCQ
 346          */
 347         int hw_tag_est_depth;
 348         unsigned int hw_tag_samples;
 349
 350         /*
 351          * idle window management
 352          */
 353         struct timer_list idle_slice_timer;
 354         struct work_struct unplug_work;
 355
 356         struct cfq_queue *active_queue;
 357         struct cfq_io_cq *active_cic;
 358
 359         /*
 360          * async queue for each priority case
 361          */
 362         struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
 363         struct cfq_queue *async_idle_cfqq;
 364
 365         sector_t last_position;
 366
 367         /*
 368          * tunables, see top of file
 369          */
 370         unsigned int cfq_quantum;
 371         unsigned int cfq_fifo_expire[2];
 372         unsigned int cfq_back_penalty;
 373         unsigned int cfq_back_max;
 374         unsigned int cfq_slice[2];
 375         unsigned int cfq_slice_async_rq;
 376         unsigned int cfq_slice_idle;
 377         unsigned int cfq_group_idle;
 378         unsigned int cfq_latency;
 379         unsigned int cfq_target_latency;
 380
 381         /*
 382          * Fallback dummy cfqq for extreme OOM conditions
 383          */
 384         struct cfq_queue oom_cfqq;
 385
 386         unsigned long last_delayed_sync;
 387 };
 388
 389 static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd);
 390
 391 static struct cfq_rb_root *st_for(struct cfq_group *cfqg,
 392                                             enum wl_class_t class,
 393                                             enum wl_type_t type)
 394 {
 395         if (!cfqg)
 396                 return NULL;
 397
 398         if (class == IDLE_WORKLOAD)
 399                 return &cfqg->service_tree_idle;
 400
 401         return &cfqg->service_trees[class][type];
 402 }
 403
 404 enum cfqq_state_flags {
 405         CFQ_CFQQ_FLAG_on_rr = 0,        /* on round-robin busy list */
 406         CFQ_CFQQ_FLAG_wait_request,     /* waiting for a request */
 407         CFQ_CFQQ_FLAG_must_dispatch,    /* must be allowed a dispatch */
 408         CFQ_CFQQ_FLAG_must_alloc_slice, /* per-slice must_alloc flag */
 409         CFQ_CFQQ_FLAG_fifo_expire,      /* FIFO checked in this slice */
 410         CFQ_CFQQ_FLAG_idle_window,      /* slice idling enabled */
 411         CFQ_CFQQ_FLAG_prio_changed,     /* task priority has changed */
 412         CFQ_CFQQ_FLAG_slice_new,        /* no requests dispatched in slice */
 413         CFQ_CFQQ_FLAG_sync,             /* synchronous queue */
 414         CFQ_CFQQ_FLAG_coop,             /* cfqq is shared */
 415         CFQ_CFQQ_FLAG_split_coop,       /* shared cfqq will be splitted */
 416         CFQ_CFQQ_FLAG_deep,             /* sync cfqq experienced large depth */
 417         CFQ_CFQQ_FLAG_wait_busy,        /* Waiting for next request */
 418 };
 419
 420 #define CFQ_CFQQ_FNS(name)                                              \
 421 static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)         \
 422 {                                                                       \
 423         (cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name);                   \
 424 }                                                                       \
 425 static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)        \
 426 {                                                                       \
 427         (cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);                  \
 428 }                                                                       \
 429 static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)         \
 430 {                                                                       \
 431         return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;      \
 432 }
 433
 434 CFQ_CFQQ_FNS(on_rr);
 435 CFQ_CFQQ_FNS(wait_request);
 436 CFQ_CFQQ_FNS(must_dispatch);
 437 CFQ_CFQQ_FNS(must_alloc_slice);
 438 CFQ_CFQQ_FNS(fifo_expire);
 439 CFQ_CFQQ_FNS(idle_window);
 440 CFQ_CFQQ_FNS(prio_changed);
 441 CFQ_CFQQ_FNS(slice_new);
 442 CFQ_CFQQ_FNS(sync);
 443 CFQ_CFQQ_FNS(coop);
 444 CFQ_CFQQ_FNS(split_coop);
 445 CFQ_CFQQ_FNS(deep);
 446 CFQ_CFQQ_FNS(wait_busy);
 447 #undef CFQ_CFQQ_FNS
 448
 449 static inline struct cfq_group *pd_to_cfqg(struct blkg_policy_data *pd)
 450 {
 451         return pd ? container_of(pd, struct cfq_group, pd) : NULL;
 452 }
 453
 454 static inline struct blkcg_gq *cfqg_to_blkg(struct cfq_group *cfqg)
 455 {
 456         return pd_to_blkg(&cfqg->pd);
 457 }
 458
 459 #if defined(CONFIG_CFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP)
 460
 461 /* cfqg stats flags */
 462 enum cfqg_stats_flags {
 463         CFQG_stats_waiting = 0,
 464         CFQG_stats_idling,
 465         CFQG_stats_empty,
 466 };
 467
 468 #define CFQG_FLAG_FNS(name)                                             \
 469 static inline void cfqg_stats_mark_##name(struct cfqg_stats *stats)     \
 470 {                                                                       \
 471         stats->flags |= (1 << CFQG_stats_##name);                       \
 472 }                                                                       \
 473 static inline void cfqg_stats_clear_##name(struct cfqg_stats *stats)    \
 474 {                                                                       \
 475         stats->flags &= ~(1 << CFQG_stats_##name);                      \
 476 }                                                                       \
 477 static inline int cfqg_stats_##name(struct cfqg_stats *stats)           \
 478 {                                                                       \
 479         return (stats->flags & (1 << CFQG_stats_##name)) != 0;          \
 480 }                                                                       \
 481
 482 CFQG_FLAG_FNS(waiting)
 483 CFQG_FLAG_FNS(idling)
 484 CFQG_FLAG_FNS(empty)
 485 #undef CFQG_FLAG_FNS
 486
 487 /* This should be called with the queue_lock held. */
 488 static void cfqg_stats_update_group_wait_time(struct cfqg_stats *stats)
 489 {
 490         unsigned long long now;
 491
 492         if (!cfqg_stats_waiting(stats))
 493                 return;
 494
 495         now = sched_clock();
 496         if (time_after64(now, stats->start_group_wait_time))
 497                 blkg_stat_add(&stats->group_wait_time,
 498                               now - stats->start_group_wait_time);
 499         cfqg_stats_clear_waiting(stats);
 500 }
 501
 502 /* This should be called with the queue_lock held. */
 503 static void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg,
 504                                                  struct cfq_group *curr_cfqg)
 505 {
 506         struct cfqg_stats *stats = &cfqg->stats;
 507
 508         if (cfqg_stats_waiting(stats))
 509                 return;
 510         if (cfqg == curr_cfqg)
 511                 return;
 512         stats->start_group_wait_time = sched_clock();
 513         cfqg_stats_mark_waiting(stats);
 514 }
 515
 516 /* This should be called with the queue_lock held. */
 517 static void cfqg_stats_end_empty_time(struct cfqg_stats *stats)
 518 {
 519         unsigned long long now;
 520
 521         if (!cfqg_stats_empty(stats))
 522                 return;
 523
 524         now = sched_clock();
 525         if (time_after64(now, stats->start_empty_time))
 526                 blkg_stat_add(&stats->empty_time,
 527                               now - stats->start_empty_time);
 528         cfqg_stats_clear_empty(stats);
 529 }
 530
 531 static void cfqg_stats_update_dequeue(struct cfq_group *cfqg)
 532 {
 533         blkg_stat_add(&cfqg->stats.dequeue, 1);
 534 }
 535
 536 static void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg)
 537 {
 538         struct cfqg_stats *stats = &cfqg->stats;
 539
 540         if (blkg_rwstat_total(&stats->queued))
 541                 return;
 542
 543         /*
 544          * group is already marked empty. This can happen if cfqq got new
 545          * request in parent group and moved to this group while being added
 546          * to service tree. Just ignore the event and move on.
 547          */
 548         if (cfqg_stats_empty(stats))
 549                 return;
 550
 551         stats->start_empty_time = sched_clock();
 552         cfqg_stats_mark_empty(stats);
 553 }
 554
 555 static void cfqg_stats_update_idle_time(struct cfq_group *cfqg)
 556 {
 557         struct cfqg_stats *stats = &cfqg->stats;
 558
 559         if (cfqg_stats_idling(stats)) {
 560                 unsigned long long now = sched_clock();
 561
 562                 if (time_after64(now, stats->start_idle_time))
 563                         blkg_stat_add(&stats->idle_time,
 564                                       now - stats->start_idle_time);
 565                 cfqg_stats_clear_idling(stats);
 566         }
 567 }
 568
 569 static void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg)
 570 {
 571         struct cfqg_stats *stats = &cfqg->stats;
 572
 573         BUG_ON(cfqg_stats_idling(stats));
 574
 575         stats->start_idle_time = sched_clock();
 576         cfqg_stats_mark_idling(stats);
 577 }
 578
 579 static void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg)
 580 {
 581         struct cfqg_stats *stats = &cfqg->stats;
 582
 583         blkg_stat_add(&stats->avg_queue_size_sum,
 584                       blkg_rwstat_total(&stats->queued));
 585         blkg_stat_add(&stats->avg_queue_size_samples, 1);
 586         cfqg_stats_update_group_wait_time(stats);
 587 }
 588
 589 #else   /* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */
 590
 591 static inline void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg, struct cfq_group *curr_cfqg) { }
 592 static inline void cfqg_stats_end_empty_time(struct cfqg_stats *stats) { }
 593 static inline void cfqg_stats_update_dequeue(struct cfq_group *cfqg) { }
 594 static inline void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg) { }
 595 static inline void cfqg_stats_update_idle_time(struct cfq_group *cfqg) { }
 596 static inline void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg) { }
 597 static inline void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg) { }
 598
 599 #endif  /* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */
 600
 601 #ifdef CONFIG_CFQ_GROUP_IOSCHED
 602
 603 static struct blkcg_policy blkcg_policy_cfq;
 604
 605 static inline struct cfq_group *blkg_to_cfqg(struct blkcg_gq *blkg)
 606 {
 607         return pd_to_cfqg(blkg_to_pd(blkg, &blkcg_policy_cfq));
 608 }
 609
 610 static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg)
 611 {
 612         struct blkcg_gq *pblkg = cfqg_to_blkg(cfqg)->parent;
 613
 614         return pblkg ? blkg_to_cfqg(pblkg) : NULL;
 615 }
 616
 617 static inline void cfqg_get(struct cfq_group *cfqg)
 618 {
 619         return blkg_get(cfqg_to_blkg(cfqg));
 620 }
 621
 622 static inline void cfqg_put(struct cfq_group *cfqg)
 623 {
 624         return blkg_put(cfqg_to_blkg(cfqg));
 625 }
 626
 627 #define cfq_log_cfqq(cfqd, cfqq, fmt, args...)  do {                    \
 628         char __pbuf[128];                                               \
 629                                                                         \
 630         blkg_path(cfqg_to_blkg((cfqq)->cfqg), __pbuf, sizeof(__pbuf));  \
 631         blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c %s " fmt, (cfqq)->pid, \
 632                         cfq_cfqq_sync((cfqq)) ? 'S' : 'A',              \
 633                         cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
 634                           __pbuf, ##args);                              \
 635 } while (0)
 636
 637 #define cfq_log_cfqg(cfqd, cfqg, fmt, args...)  do {                    \
 638         char __pbuf[128];                                               \
 639                                                                         \
 640         blkg_path(cfqg_to_blkg(cfqg), __pbuf, sizeof(__pbuf));          \
 641         blk_add_trace_msg((cfqd)->queue, "%s " fmt, __pbuf, ##args);    \
 642 } while (0)
 643
 644 static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
 645                                             struct cfq_group *curr_cfqg, int rw)
 646 {
 647         blkg_rwstat_add(&cfqg->stats.queued, rw, 1);
 648         cfqg_stats_end_empty_time(&cfqg->stats);
 649         cfqg_stats_set_start_group_wait_time(cfqg, curr_cfqg);
 650 }
 651
 652 static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
 653                         unsigned long time, unsigned long unaccounted_time)
 654 {
 655         blkg_stat_add(&cfqg->stats.time, time);
 656 #ifdef CONFIG_DEBUG_BLK_CGROUP
 657         blkg_stat_add(&cfqg->stats.unaccounted_time, unaccounted_time);
 658 #endif
 659 }
 660
 661 static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int rw)
 662 {
 663         blkg_rwstat_add(&cfqg->stats.queued, rw, -1);
 664 }
 665
 666 static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int rw)
 667 {
 668         blkg_rwstat_add(&cfqg->stats.merged, rw, 1);
 669 }
 670
 671 static inline void cfqg_stats_update_dispatch(struct cfq_group *cfqg,
 672                                               uint64_t bytes, int rw)
 673 {
 674         blkg_stat_add(&cfqg->stats.sectors, bytes >> 9);
 675         blkg_rwstat_add(&cfqg->stats.serviced, rw, 1);
 676         blkg_rwstat_add(&cfqg->stats.service_bytes, rw, bytes);
 677 }
 678
 679 static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
 680                         uint64_t start_time, uint64_t io_start_time, int rw)
 681 {
 682         struct cfqg_stats *stats = &cfqg->stats;
 683         unsigned long long now = sched_clock();
 684
 685         if (time_after64(now, io_start_time))
 686                 blkg_rwstat_add(&stats->service_time, rw, now - io_start_time);
 687         if (time_after64(io_start_time, start_time))
 688                 blkg_rwstat_add(&stats->wait_time, rw,
 689                                 io_start_time - start_time);
 690 }
 691
 692 /* @stats = 0 */
 693 static void cfqg_stats_reset(struct cfqg_stats *stats)
 694 {
 695         /* queued stats shouldn't be cleared */
 696         blkg_rwstat_reset(&stats->service_bytes);
 697         blkg_rwstat_reset(&stats->serviced);
 698         blkg_rwstat_reset(&stats->merged);
 699         blkg_rwstat_reset(&stats->service_time);
 700         blkg_rwstat_reset(&stats->wait_time);
 701         blkg_stat_reset(&stats->time);
 702 #ifdef CONFIG_DEBUG_BLK_CGROUP
 703         blkg_stat_reset(&stats->unaccounted_time);
 704         blkg_stat_reset(&stats->avg_queue_size_sum);
 705         blkg_stat_reset(&stats->avg_queue_size_samples);
 706         blkg_stat_reset(&stats->dequeue);
 707         blkg_stat_reset(&stats->group_wait_time);
 708         blkg_stat_reset(&stats->idle_time);
 709         blkg_stat_reset(&stats->empty_time);
 710 #endif
 711 }
 712
 713 /* @to += @from */
 714 static void cfqg_stats_merge(struct cfqg_stats *to, struct cfqg_stats *from)
 715 {
 716         /* queued stats shouldn't be cleared */
 717         blkg_rwstat_merge(&to->service_bytes, &from->service_bytes);
 718         blkg_rwstat_merge(&to->serviced, &from->serviced);
 719         blkg_rwstat_merge(&to->merged, &from->merged);
 720         blkg_rwstat_merge(&to->service_time, &from->service_time);
 721         blkg_rwstat_merge(&to->wait_time, &from->wait_time);
 722         blkg_stat_merge(&from->time, &from->time);
 723 #ifdef CONFIG_DEBUG_BLK_CGROUP
 724         blkg_stat_merge(&to->unaccounted_time, &from->unaccounted_time);
 725         blkg_stat_merge(&to->avg_queue_size_sum, &from->avg_queue_size_sum);
 726         blkg_stat_merge(&to->avg_queue_size_samples, &from->avg_queue_size_samples);
 727         blkg_stat_merge(&to->dequeue, &from->dequeue);
 728         blkg_stat_merge(&to->group_wait_time, &from->group_wait_time);
 729         blkg_stat_merge(&to->idle_time, &from->idle_time);
 730         blkg_stat_merge(&to->empty_time, &from->empty_time);
 731 #endif
 732 }
 733
 734 /*
 735  * Transfer @cfqg's stats to its parent's dead_stats so that the ancestors'
 736  * recursive stats can still account for the amount used by this cfqg after
 737  * it's gone.
 738  */
 739 static void cfqg_stats_xfer_dead(struct cfq_group *cfqg)
 740 {
 741         struct cfq_group *parent = cfqg_parent(cfqg);
 742
 743         lockdep_assert_held(cfqg_to_blkg(cfqg)->q->queue_lock);
 744
 745         if (unlikely(!parent))
 746                 return;
 747
 748         cfqg_stats_merge(&parent->dead_stats, &cfqg->stats);
 749         cfqg_stats_merge(&parent->dead_stats, &cfqg->dead_stats);
 750         cfqg_stats_reset(&cfqg->stats);
 751         cfqg_stats_reset(&cfqg->dead_stats);
 752 }
 753
 754 #else   /* CONFIG_CFQ_GROUP_IOSCHED */
 755
 756 static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg) { return NULL; }
 757 static inline void cfqg_get(struct cfq_group *cfqg) { }
 758 static inline void cfqg_put(struct cfq_group *cfqg) { }
 759
 760 #define cfq_log_cfqq(cfqd, cfqq, fmt, args...)  \
 761         blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c " fmt, (cfqq)->pid, \
 762                         cfq_cfqq_sync((cfqq)) ? 'S' : 'A',              \
 763                         cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
 764                                 ##args)
 765 #define cfq_log_cfqg(cfqd, cfqg, fmt, args...)          do {} while (0)
 766
 767 static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
 768                         struct cfq_group *curr_cfqg, int rw) { }
 769 static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
 770                         unsigned long time, unsigned long unaccounted_time) { }
 771 static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int rw) { }
 772 static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int rw) { }
 773 static inline void cfqg_stats_update_dispatch(struct cfq_group *cfqg,
 774                                               uint64_t bytes, int rw) { }
 775 static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
 776                         uint64_t start_time, uint64_t io_start_time, int rw) { }
 777
 778 #endif  /* CONFIG_CFQ_GROUP_IOSCHED */
 779
 780 #define cfq_log(cfqd, fmt, args...)     \
 781         blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)
 782
 783 /* Traverses through cfq group service trees */
 784 #define for_each_cfqg_st(cfqg, i, j, st) \
 785         for (i = 0; i <= IDLE_WORKLOAD; i++) \
 786                 for (j = 0, st = i < IDLE_WORKLOAD ? &cfqg->service_trees[i][j]\
 787                         : &cfqg->service_tree_idle; \
 788                         (i < IDLE_WORKLOAD && j <= SYNC_WORKLOAD) || \
 789                         (i == IDLE_WORKLOAD && j == 0); \
 790                         j++, st = i < IDLE_WORKLOAD ? \
 791                         &cfqg->service_trees[i][j]: NULL) \
 792
 793 static inline bool cfq_io_thinktime_big(struct cfq_data *cfqd,
 794         struct cfq_ttime *ttime, bool group_idle)
 795 {
 796         unsigned long slice;
 797         if (!sample_valid(ttime->ttime_samples))
 798                 return false;
 799         if (group_idle)
 800                 slice = cfqd->cfq_group_idle;
 801         else
 802                 slice = cfqd->cfq_slice_idle;
 803         return ttime->ttime_mean > slice;
 804 }
 805
 806 static inline bool iops_mode(struct cfq_data *cfqd)
 807 {
 808         /*
 809          * If we are not idling on queues and it is a NCQ drive, parallel
 810          * execution of requests is on and measuring time is not possible
 811          * in most of the cases until and unless we drive shallower queue
 812          * depths and that becomes a performance bottleneck. In such cases
 813          * switch to start providing fairness in terms of number of IOs.
 814          */
 815         if (!cfqd->cfq_slice_idle && cfqd->hw_tag)
 816                 return true;
 817         else
 818                 return false;
 819 }
 820
 821 static inline enum wl_class_t cfqq_class(struct cfq_queue *cfqq)
 822 {
 823         if (cfq_class_idle(cfqq))
 824                 return IDLE_WORKLOAD;
 825         if (cfq_class_rt(cfqq))
 826                 return RT_WORKLOAD;
 827         return BE_WORKLOAD;
 828 }
 829
 830
 831 static enum wl_type_t cfqq_type(struct cfq_queue *cfqq)
 832 {
 833         if (!cfq_cfqq_sync(cfqq))
 834                 return ASYNC_WORKLOAD;
 835         if (!cfq_cfqq_idle_window(cfqq))
 836                 return SYNC_NOIDLE_WORKLOAD;
 837         return SYNC_WORKLOAD;
 838 }
 839
 840 static inline int cfq_group_busy_queues_wl(enum wl_class_t wl_class,
 841                                         struct cfq_data *cfqd,
 842                                         struct cfq_group *cfqg)
 843 {
 844         if (wl_class == IDLE_WORKLOAD)
 845                 return cfqg->service_tree_idle.count;
 846
 847         return cfqg->service_trees[wl_class][ASYNC_WORKLOAD].count +
 848                 cfqg->service_trees[wl_class][SYNC_NOIDLE_WORKLOAD].count +
 849                 cfqg->service_trees[wl_class][SYNC_WORKLOAD].count;
 850 }
 851
 852 static inline int cfqg_busy_async_queues(struct cfq_data *cfqd,
 853                                         struct cfq_group *cfqg)
 854 {
 855         return cfqg->service_trees[RT_WORKLOAD][ASYNC_WORKLOAD].count +
 856                 cfqg->service_trees[BE_WORKLOAD][ASYNC_WORKLOAD].count;
 857 }
 858
 859 static void cfq_dispatch_insert(struct request_queue *, struct request *);
 860 static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, bool is_sync,
 861                                        struct cfq_io_cq *cic, struct bio *bio,
 862                                        gfp_t gfp_mask);
 863
 864 static inline struct cfq_io_cq *icq_to_cic(struct io_cq *icq)
 865 {
 866         /* cic->icq is the first member, %NULL will convert to %NULL */
 867         return container_of(icq, struct cfq_io_cq, icq);
 868 }
 869
 870 static inline struct cfq_io_cq *cfq_cic_lookup(struct cfq_data *cfqd,
 871                                                struct io_context *ioc)
 872 {
 873         if (ioc)
 874                 return icq_to_cic(ioc_lookup_icq(ioc, cfqd->queue));
 875         return NULL;
 876 }
 877
 878 static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_cq *cic, bool is_sync)
 879 {
 880         return cic->cfqq[is_sync];
 881 }
 882
 883 static inline void cic_set_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq,
 884                                 bool is_sync)
 885 {
 886         cic->cfqq[is_sync] = cfqq;
 887 }
 888
 889 static inline struct cfq_data *cic_to_cfqd(struct cfq_io_cq *cic)
 890 {
 891         return cic->icq.q->elevator->elevator_data;
 892 }
 893
 894 /*
 895  * We regard a request as SYNC, if it's either a read or has the SYNC bit
 896  * set (in which case it could also be direct WRITE).
 897  */
 898 static inline bool cfq_bio_sync(struct bio *bio)
 899 {
 900         return bio_data_dir(bio) == READ || (bio->bi_rw & REQ_SYNC);
 901 }
 902
 903 /*
 904  * scheduler run of queue, if there are requests pending and no one in the
 905  * driver that will restart queueing
 906  */
 907 static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
 908 {
 909         if (cfqd->busy_queues) {
 910                 cfq_log(cfqd, "schedule dispatch");
 911                 kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work);
 912         }
 913 }
 914
 915 /*
 916  * Scale schedule slice based on io priority. Use the sync time slice only
 917  * if a queue is marked sync and has sync io queued. A sync queue with async
 918  * io only, should not get full sync slice length.
 919  */
 920 static inline int cfq_prio_slice(struct cfq_data *cfqd, bool sync,
 921                                  unsigned short prio)
 922 {
 923         const int base_slice = cfqd->cfq_slice[sync];
 924
 925         WARN_ON(prio >= IOPRIO_BE_NR);
 926
 927         return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio));
 928 }
 929
 930 static inline int
 931 cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 932 {
 933         return cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio);
 934 }
 935
 936 /**
 937  * cfqg_scale_charge - scale disk time charge according to cfqg weight
 938  * @charge: disk time being charged
 939  * @vfraction: vfraction of the cfqg, fixed point w/ CFQ_SERVICE_SHIFT
 940  *
 941  * Scale @charge according to @vfraction, which is in range (0, 1].  The
 942  * scaling is inversely proportional.
 943  *
 944  * scaled = charge / vfraction
 945  *
 946  * The result is also in fixed point w/ CFQ_SERVICE_SHIFT.
 947  */
 948 static inline u64 cfqg_scale_charge(unsigned long charge,
 949                                     unsigned int vfraction)
 950 {
 951         u64 c = charge << CFQ_SERVICE_SHIFT;    /* make it fixed point */
 952
 953         /* charge / vfraction */
 954         c <<= CFQ_SERVICE_SHIFT;
 955         do_div(c, vfraction);
 956         return c;
 957 }
 958
 959 static inline u64 max_vdisktime(u64 min_vdisktime, u64 vdisktime)
 960 {
 961         s64 delta = (s64)(vdisktime - min_vdisktime);
 962         if (delta > 0)
 963                 min_vdisktime = vdisktime;
 964
 965         return min_vdisktime;
 966 }
 967
 968 static inline u64 min_vdisktime(u64 min_vdisktime, u64 vdisktime)
 969 {
 970         s64 delta = (s64)(vdisktime - min_vdisktime);
 971         if (delta < 0)
 972                 min_vdisktime = vdisktime;
 973
 974         return min_vdisktime;
 975 }
 976
 977 static void update_min_vdisktime(struct cfq_rb_root *st)
 978 {
 979         struct cfq_group *cfqg;
 980
 981         if (st->left) {
 982                 cfqg = rb_entry_cfqg(st->left);
 983                 st->min_vdisktime = max_vdisktime(st->min_vdisktime,
 984                                                   cfqg->vdisktime);
 985         }
 986 }
 987
 988 /*
 989  * get averaged number of queues of RT/BE priority.
 990  * average is updated, with a formula that gives more weight to higher numbers,
 991  * to quickly follows sudden increases and decrease slowly
 992  */
 993
 994 static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd,
 995                                         struct cfq_group *cfqg, bool rt)
 996 {
 997         unsigned min_q, max_q;
 998         unsigned mult  = cfq_hist_divisor - 1;
 999         unsigned round = cfq_hist_divisor / 2;
1000         unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg);
1001
1002         min_q = min(cfqg->busy_queues_avg[rt], busy);
1003         max_q = max(cfqg->busy_queues_avg[rt], busy);
1004         cfqg->busy_queues_avg[rt] = (mult * max_q + min_q + round) /
1005                 cfq_hist_divisor;
1006         return cfqg->busy_queues_avg[rt];
1007 }
1008
1009 static inline unsigned
1010 cfq_group_slice(struct cfq_data *cfqd, struct cfq_group *cfqg)
1011 {
1012         return cfqd->cfq_target_latency * cfqg->vfraction >> CFQ_SERVICE_SHIFT;
1013 }
1014
1015 static inline unsigned
1016 cfq_scaled_cfqq_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1017 {
1018         unsigned slice = cfq_prio_to_slice(cfqd, cfqq);
1019         if (cfqd->cfq_latency) {
1020                 /*
1021                  * interested queues (we consider only the ones with the same
1022                  * priority class in the cfq group)
1023                  */
1024                 unsigned iq = cfq_group_get_avg_queues(cfqd, cfqq->cfqg,
1025                                                 cfq_class_rt(cfqq));
1026                 unsigned sync_slice = cfqd->cfq_slice[1];
1027                 unsigned expect_latency = sync_slice * iq;
1028                 unsigned group_slice = cfq_group_slice(cfqd, cfqq->cfqg);
1029
1030                 if (expect_latency > group_slice) {
1031                         unsigned base_low_slice = 2 * cfqd->cfq_slice_idle;
1032                         /* scale low_slice according to IO priority
1033                          * and sync vs async */
1034                         unsigned low_slice =
1035                                 min(slice, base_low_slice * slice / sync_slice);
1036                         /* the adapted slice value is scaled to fit all iqs
1037                          * into the target latency */
1038                         slice = max(slice * group_slice / expect_latency,
1039                                     low_slice);
1040                 }
1041         }
1042         return slice;
1043 }
1044
1045 static inline void
1046 cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1047 {
1048         unsigned slice = cfq_scaled_cfqq_slice(cfqd, cfqq);
1049
1050         cfqq->slice_start = jiffies;
1051         cfqq->slice_end = jiffies + slice;
1052         cfqq->allocated_slice = slice;
1053         cfq_log_cfqq(cfqd, cfqq, "set_slice=%lu", cfqq->slice_end - jiffies);
1054 }
1055
1056 /*
1057  * We need to wrap this check in cfq_cfqq_slice_new(), since ->slice_end
1058  * isn't valid until the first request from the dispatch is activated
1059  * and the slice time set.
1060  */
1061 static inline bool cfq_slice_used(struct cfq_queue *cfqq)
1062 {
1063         if (cfq_cfqq_slice_new(cfqq))
1064                 return false;
1065         if (time_before(jiffies, cfqq->slice_end))
1066                 return false;
1067
1068         return true;
1069 }
1070
1071 /*
1072  * Lifted from AS - choose which of rq1 and rq2 that is best served now.
1073  * We choose the request that is closest to the head right now. Distance
1074  * behind the head is penalized and only allowed to a certain extent.
1075  */
1076 static struct request *
1077 cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2, sector_t last)
1078 {
1079         sector_t s1, s2, d1 = 0, d2 = 0;
1080         unsigned long back_max;
1081 #define CFQ_RQ1_WRAP    0x01 /* request 1 wraps */
1082 #define CFQ_RQ2_WRAP    0x02 /* request 2 wraps */
1083         unsigned wrap = 0; /* bit mask: requests behind the disk head? */
1084
1085         if (rq1 == NULL || rq1 == rq2)
1086                 return rq2;
1087         if (rq2 == NULL)
1088                 return rq1;
1089
1090         if (rq_is_sync(rq1) != rq_is_sync(rq2))
1091                 return rq_is_sync(rq1) ? rq1 : rq2;
1092
1093         if ((rq1->cmd_flags ^ rq2->cmd_flags) & REQ_PRIO)
1094                 return rq1->cmd_flags & REQ_PRIO ? rq1 : rq2;
1095
1096         s1 = blk_rq_pos(rq1);
1097         s2 = blk_rq_pos(rq2);
1098
1099         /*
1100          * by definition, 1KiB is 2 sectors
1101          */
1102         back_max = cfqd->cfq_back_max * 2;
1103
1104         /*
1105          * Strict one way elevator _except_ in the case where we allow
1106          * short backward seeks which are biased as twice the cost of a
1107          * similar forward seek.
1108          */
1109         if (s1 >= last)
1110                 d1 = s1 - last;
1111         else if (s1 + back_max >= last)
1112                 d1 = (last - s1) * cfqd->cfq_back_penalty;
1113         else
1114                 wrap |= CFQ_RQ1_WRAP;
1115
1116         if (s2 >= last)
1117                 d2 = s2 - last;
1118         else if (s2 + back_max >= last)
1119                 d2 = (last - s2) * cfqd->cfq_back_penalty;
1120         else
1121                 wrap |= CFQ_RQ2_WRAP;
1122
1123         /* Found required data */
1124
1125         /*
1126          * By doing switch() on the bit mask "wrap" we avoid having to
1127          * check two variables for all permutations: --> faster!
1128          */
1129         switch (wrap) {
1130         case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
1131                 if (d1 < d2)
1132                         return rq1;
1133                 else if (d2 < d1)
1134                         return rq2;
1135                 else {
1136                         if (s1 >= s2)
1137                                 return rq1;
1138                         else
1139                                 return rq2;
1140                 }
1141
1142         case CFQ_RQ2_WRAP:
1143                 return rq1;
1144         case CFQ_RQ1_WRAP:
1145                 return rq2;
1146         case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */
1147         default:
1148                 /*
1149                  * Since both rqs are wrapped,
1150                  * start with the one that's further behind head
1151                  * (--> only *one* back seek required),
1152                  * since back seek takes more time than forward.
1153                  */
1154                 if (s1 <= s2)
1155                         return rq1;
1156                 else
1157                         return rq2;
1158         }
1159 }
1160
1161 /*
1162  * The below is leftmost cache rbtree addon
1163  */
1164 static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root)
1165 {
1166         /* Service tree is empty */
1167         if (!root->count)
1168                 return NULL;
1169
1170         if (!root->left)
1171                 root->left = rb_first(&root->rb);
1172
1173         if (root->left)
1174                 return rb_entry(root->left, struct cfq_queue, rb_node);
1175
1176         return NULL;
1177 }
1178
1179 static struct cfq_group *cfq_rb_first_group(struct cfq_rb_root *root)
1180 {
1181         if (!root->left)
1182                 root->left = rb_first(&root->rb);
1183
1184         if (root->left)
1185                 return rb_entry_cfqg(root->left);
1186
1187         return NULL;
1188 }
1189
1190 static void rb_erase_init(struct rb_node *n, struct rb_root *root)
1191 {
1192         rb_erase(n, root);
1193         RB_CLEAR_NODE(n);
1194 }
1195
1196 static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root)
1197 {
1198         if (root->left == n)
1199                 root->left = NULL;
1200         rb_erase_init(n, &root->rb);
1201         --root->count;
1202 }
1203
1204 /*
1205  * would be nice to take fifo expire time into account as well
1206  */
1207 static struct request *
1208 cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1209                   struct request *last)
1210 {
1211         struct rb_node *rbnext = rb_next(&last->rb_node);
1212         struct rb_node *rbprev = rb_prev(&last->rb_node);
1213         struct request *next = NULL, *prev = NULL;
1214
1215         BUG_ON(RB_EMPTY_NODE(&last->rb_node));
1216
1217         if (rbprev)
1218                 prev = rb_entry_rq(rbprev);
1219
1220         if (rbnext)
1221                 next = rb_entry_rq(rbnext);
1222         else {
1223                 rbnext = rb_first(&cfqq->sort_list);
1224                 if (rbnext && rbnext != &last->rb_node)
1225                         next = rb_entry_rq(rbnext);
1226         }
1227
1228         return cfq_choose_req(cfqd, next, prev, blk_rq_pos(last));
1229 }
1230
1231 static unsigned long cfq_slice_offset(struct cfq_data *cfqd,
1232                                       struct cfq_queue *cfqq)
1233 {
1234         /*
1235          * just an approximation, should be ok.
1236          */
1237         return (cfqq->cfqg->nr_cfqq - 1) * (cfq_prio_slice(cfqd, 1, 0) -
1238                        cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio));
1239 }
1240
1241 static inline s64
1242 cfqg_key(struct cfq_rb_root *st, struct cfq_group *cfqg)
1243 {
1244         return cfqg->vdisktime - st->min_vdisktime;
1245 }
1246
1247 static void
1248 __cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
1249 {
1250         struct rb_node **node = &st->rb.rb_node;
1251         struct rb_node *parent = NULL;
1252         struct cfq_group *__cfqg;
1253         s64 key = cfqg_key(st, cfqg);
1254         int left = 1;
1255
1256         while (*node != NULL) {
1257                 parent = *node;
1258                 __cfqg = rb_entry_cfqg(parent);
1259
1260                 if (key < cfqg_key(st, __cfqg))
1261                         node = &parent->rb_left;
1262                 else {
1263                         node = &parent->rb_right;
1264                         left = 0;
1265                 }
1266         }
1267
1268         if (left)
1269                 st->left = &cfqg->rb_node;
1270
1271         rb_link_node(&cfqg->rb_node, parent, node);
1272         rb_insert_color(&cfqg->rb_node, &st->rb);
1273 }
1274
1275 static void
1276 cfq_update_group_weight(struct cfq_group *cfqg)
1277 {
1278         BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
1279
1280         if (cfqg->new_weight) {
1281                 cfqg->weight = cfqg->new_weight;
1282                 cfqg->new_weight = 0;
1283         }
1284
1285         if (cfqg->new_leaf_weight) {
1286                 cfqg->leaf_weight = cfqg->new_leaf_weight;
1287                 cfqg->new_leaf_weight = 0;
1288         }
1289 }
1290
1291 static void
1292 cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
1293 {
1294         unsigned int vfr = 1 << CFQ_SERVICE_SHIFT;      /* start with 1 */
1295         struct cfq_group *pos = cfqg;
1296         struct cfq_group *parent;
1297         bool propagate;
1298
1299         /* add to the service tree */
1300         BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
1301
1302         cfq_update_group_weight(cfqg);
1303         __cfq_group_service_tree_add(st, cfqg);
1304
1305         /*
1306          * Activate @cfqg and calculate the portion of vfraction @cfqg is
1307          * entitled to.  vfraction is calculated by walking the tree
1308          * towards the root calculating the fraction it has at each level.
1309          * The compounded ratio is how much vfraction @cfqg owns.
1310          *
1311          * Start with the proportion tasks in this cfqg has against active
1312          * children cfqgs - its leaf_weight against children_weight.
1313          */
1314         propagate = !pos->nr_active++;
1315         pos->children_weight += pos->leaf_weight;
1316         vfr = vfr * pos->leaf_weight / pos->children_weight;
1317
1318         /*
1319          * Compound ->weight walking up the tree.  Both activation and
1320          * vfraction calculation are done in the same loop.  Propagation
1321          * stops once an already activated node is met.  vfraction
1322          * calculation should always continue to the root.
1323          */
1324         while ((parent = cfqg_parent(pos))) {
1325                 if (propagate) {
1326                         propagate = !parent->nr_active++;
1327                         parent->children_weight += pos->weight;
1328                 }
1329                 vfr = vfr * pos->weight / parent->children_weight;
1330                 pos = parent;
1331         }
1332
1333         cfqg->vfraction = max_t(unsigned, vfr, 1);
1334 }
1335
1336 static void
1337 cfq_group_notify_queue_add(struct cfq_data *cfqd, struct cfq_group *cfqg)
1338 {
1339         struct cfq_rb_root *st = &cfqd->grp_service_tree;
1340         struct cfq_group *__cfqg;
1341         struct rb_node *n;
1342
1343         cfqg->nr_cfqq++;
1344         if (!RB_EMPTY_NODE(&cfqg->rb_node))
1345                 return;
1346
1347         /*
1348          * Currently put the group at the end. Later implement something
1349          * so that groups get lesser vtime based on their weights, so that
1350          * if group does not loose all if it was not continuously backlogged.
1351          */
1352         n = rb_last(&st->rb);
1353         if (n) {
1354                 __cfqg = rb_entry_cfqg(n);
1355                 cfqg->vdisktime = __cfqg->vdisktime + CFQ_IDLE_DELAY;
1356         } else
1357                 cfqg->vdisktime = st->min_vdisktime;
1358         cfq_group_service_tree_add(st, cfqg);
1359 }
1360
1361 static void
1362 cfq_group_service_tree_del(struct cfq_rb_root *st, struct cfq_group *cfqg)
1363 {
1364         struct cfq_group *pos = cfqg;
1365         bool propagate;
1366
1367         /*
1368          * Undo activation from cfq_group_service_tree_add().  Deactivate
1369          * @cfqg and propagate deactivation upwards.
1370          */
1371         propagate = !--pos->nr_active;
1372         pos->children_weight -= pos->leaf_weight;
1373
1374         while (propagate) {
1375                 struct cfq_group *parent = cfqg_parent(pos);
1376
1377                 /* @pos has 0 nr_active at this point */
1378                 WARN_ON_ONCE(pos->children_weight);
1379                 pos->vfraction = 0;
1380
1381                 if (!parent)
1382                         break;
1383
1384                 propagate = !--parent->nr_active;
1385                 parent->children_weight -= pos->weight;
1386                 pos = parent;
1387         }
1388
1389         /* remove from the service tree */
1390         if (!RB_EMPTY_NODE(&cfqg->rb_node))
1391                 cfq_rb_erase(&cfqg->rb_node, st);
1392 }
1393
1394 static void
1395 cfq_group_notify_queue_del(struct cfq_data *cfqd, struct cfq_group *cfqg)
1396 {
1397         struct cfq_rb_root *st = &cfqd->grp_service_tree;
1398
1399         BUG_ON(cfqg->nr_cfqq < 1);
1400         cfqg->nr_cfqq--;
1401
1402         /* If there are other cfq queues under this group, don't delete it */
1403         if (cfqg->nr_cfqq)
1404                 return;
1405
1406         cfq_log_cfqg(cfqd, cfqg, "del_from_rr group");
1407         cfq_group_service_tree_del(st, cfqg);
1408         cfqg->saved_wl_slice = 0;
1409         cfqg_stats_update_dequeue(cfqg);
1410 }
1411
1412 static inline unsigned int cfq_cfqq_slice_usage(struct cfq_queue *cfqq,
1413                                                 unsigned int *unaccounted_time)
1414 {
1415         unsigned int slice_used;
1416
1417         /*
1418          * Queue got expired before even a single request completed or
1419          * got expired immediately after first request completion.
1420          */
1421         if (!cfqq->slice_start || cfqq->slice_start == jiffies) {
1422                 /*
1423                  * Also charge the seek time incurred to the group, otherwise
1424                  * if there are mutiple queues in the group, each can dispatch
1425                  * a single request on seeky media and cause lots of seek time
1426                  * and group will never know it.
1427                  */
1428                 slice_used = max_t(unsigned, (jiffies - cfqq->dispatch_start),
1429                                         1);
1430         } else {
1431                 slice_used = jiffies - cfqq->slice_start;
1432                 if (slice_used > cfqq->allocated_slice) {
1433                         *unaccounted_time = slice_used - cfqq->allocated_slice;
1434                         slice_used = cfqq->allocated_slice;
1435                 }
1436                 if (time_after(cfqq->slice_start, cfqq->dispatch_start))
1437                         *unaccounted_time += cfqq->slice_start -
1438                                         cfqq->dispatch_start;
1439         }
1440
1441         return slice_used;
1442 }
1443
1444 static void cfq_group_served(struct cfq_data *cfqd, struct cfq_group *cfqg,
1445                                 struct cfq_queue *cfqq)
1446 {
1447         struct cfq_rb_root *st = &cfqd->grp_service_tree;
1448         unsigned int used_sl, charge, unaccounted_sl = 0;
1449         int nr_sync = cfqg->nr_cfqq - cfqg_busy_async_queues(cfqd, cfqg)
1450                         - cfqg->service_tree_idle.count;
1451         unsigned int vfr;
1452
1453         BUG_ON(nr_sync < 0);
1454         used_sl = charge = cfq_cfqq_slice_usage(cfqq, &unaccounted_sl);
1455
1456         if (iops_mode(cfqd))
1457                 charge = cfqq->slice_dispatch;
1458         else if (!cfq_cfqq_sync(cfqq) && !nr_sync)
1459                 charge = cfqq->allocated_slice;
1460
1461         /*
1462          * Can't update vdisktime while on service tree and cfqg->vfraction
1463          * is valid only while on it.  Cache vfr, leave the service tree,
1464          * update vdisktime and go back on.  The re-addition to the tree
1465          * will also update the weights as necessary.
1466          */
1467         vfr = cfqg->vfraction;
1468         cfq_group_service_tree_del(st, cfqg);
1469         cfqg->vdisktime += cfqg_scale_charge(charge, vfr);
1470         cfq_group_service_tree_add(st, cfqg);
1471
1472         /* This group is being expired. Save the context */
1473         if (time_after(cfqd->workload_expires, jiffies)) {
1474                 cfqg->saved_wl_slice = cfqd->workload_expires
1475                                                 - jiffies;
1476                 cfqg->saved_wl_type = cfqd->serving_wl_type;
1477                 cfqg->saved_wl_class = cfqd->serving_wl_class;
1478         } else
1479                 cfqg->saved_wl_slice = 0;
1480
1481         cfq_log_cfqg(cfqd, cfqg, "served: vt=%llu min_vt=%llu", cfqg->vdisktime,
1482                                         st->min_vdisktime);
1483         cfq_log_cfqq(cfqq->cfqd, cfqq,
1484                      "sl_used=%u disp=%u charge=%u iops=%u sect=%lu",
1485                      used_sl, cfqq->slice_dispatch, charge,
1486                      iops_mode(cfqd), cfqq->nr_sectors);
1487         cfqg_stats_update_timeslice_used(cfqg, used_sl, unaccounted_sl);
1488         cfqg_stats_set_start_empty_time(cfqg);
1489 }
1490
1491 /**
1492  * cfq_init_cfqg_base - initialize base part of a cfq_group
1493  * @cfqg: cfq_group to initialize
1494  *
1495  * Initialize the base part which is used whether %CONFIG_CFQ_GROUP_IOSCHED
1496  * is enabled or not.
1497  */
1498 static void cfq_init_cfqg_base(struct cfq_group *cfqg)
1499 {
1500         struct cfq_rb_root *st;
1501         int i, j;
1502
1503         for_each_cfqg_st(cfqg, i, j, st)
1504                 *st = CFQ_RB_ROOT;
1505         RB_CLEAR_NODE(&cfqg->rb_node);
1506
1507         cfqg->ttime.last_end_request = jiffies;
1508 }
1509
1510 #ifdef CONFIG_CFQ_GROUP_IOSCHED
1511 static void cfqg_stats_init(struct cfqg_stats *stats)
1512 {
1513         blkg_rwstat_init(&stats->service_bytes);
1514         blkg_rwstat_init(&stats->serviced);
1515         blkg_rwstat_init(&stats->merged);
1516         blkg_rwstat_init(&stats->service_time);
1517         blkg_rwstat_init(&stats->wait_time);
1518         blkg_rwstat_init(&stats->queued);
1519
1520         blkg_stat_init(&stats->sectors);
1521         blkg_stat_init(&stats->time);
1522
1523 #ifdef CONFIG_DEBUG_BLK_CGROUP
1524         blkg_stat_init(&stats->unaccounted_time);
1525         blkg_stat_init(&stats->avg_queue_size_sum);
1526         blkg_stat_init(&stats->avg_queue_size_samples);
1527         blkg_stat_init(&stats->dequeue);
1528         blkg_stat_init(&stats->group_wait_time);
1529         blkg_stat_init(&stats->idle_time);
1530         blkg_stat_init(&stats->empty_time);
1531 #endif
1532 }
1533
1534 static void cfq_pd_init(struct blkcg_gq *blkg)
1535 {
1536         struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1537
1538         cfq_init_cfqg_base(cfqg);
1539         cfqg->weight = blkg->blkcg->cfq_weight;
1540         cfqg->leaf_weight = blkg->blkcg->cfq_leaf_weight;
1541         cfqg_stats_init(&cfqg->stats);
1542         cfqg_stats_init(&cfqg->dead_stats);
1543 }
1544
1545 static void cfq_pd_offline(struct blkcg_gq *blkg)
1546 {
1547         /*
1548          * @blkg is going offline and will be ignored by
1549          * blkg_[rw]stat_recursive_sum().  Transfer stats to the parent so
1550          * that they don't get lost.  If IOs complete after this point, the
1551          * stats for them will be lost.  Oh well...
1552          */
1553         cfqg_stats_xfer_dead(blkg_to_cfqg(blkg));
1554 }
1555
1556 /* offset delta from cfqg->stats to cfqg->dead_stats */
1557 static const int dead_stats_off_delta = offsetof(struct cfq_group, dead_stats) -
1558                                         offsetof(struct cfq_group, stats);
1559
1560 /* to be used by recursive prfill, sums live and dead stats recursively */
1561 static u64 cfqg_stat_pd_recursive_sum(struct blkg_policy_data *pd, int off)
1562 {
1563         u64 sum = 0;
1564
1565         sum += blkg_stat_recursive_sum(pd, off);
1566         sum += blkg_stat_recursive_sum(pd, off + dead_stats_off_delta);
1567         return sum;
1568 }
1569
1570 /* to be used by recursive prfill, sums live and dead rwstats recursively */
1571 static struct blkg_rwstat cfqg_rwstat_pd_recursive_sum(struct blkg_policy_data *pd,
1572                                                        int off)
1573 {
1574         struct blkg_rwstat a, b;
1575
1576         a = blkg_rwstat_recursive_sum(pd, off);
1577         b = blkg_rwstat_recursive_sum(pd, off + dead_stats_off_delta);
1578         blkg_rwstat_merge(&a, &b);
1579         return a;
1580 }
1581
1582 static void cfq_pd_reset_stats(struct blkcg_gq *blkg)
1583 {
1584         struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1585
1586         cfqg_stats_reset(&cfqg->stats);
1587         cfqg_stats_reset(&cfqg->dead_stats);
1588 }
1589
1590 /*
1591  * Search for the cfq group current task belongs to. request_queue lock must
1592  * be held.
1593  */
1594 static struct cfq_group *cfq_lookup_create_cfqg(struct cfq_data *cfqd,
1595                                                 struct blkcg *blkcg)
1596 {
1597         struct request_queue *q = cfqd->queue;
1598         struct cfq_group *cfqg = NULL;
1599
1600         /* avoid lookup for the common case where there's no blkcg */
1601         if (blkcg == &blkcg_root) {
1602                 cfqg = cfqd->root_group;
1603         } else {
1604                 struct blkcg_gq *blkg;
1605
1606                 blkg = blkg_lookup_create(blkcg, q);
1607                 if (!IS_ERR(blkg))
1608                         cfqg = blkg_to_cfqg(blkg);
1609         }
1610
1611         return cfqg;
1612 }
1613
1614 static void cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg)
1615 {
1616         /* Currently, all async queues are mapped to root group */
1617         if (!cfq_cfqq_sync(cfqq))
1618                 cfqg = cfqq->cfqd->root_group;
1619
1620         cfqq->cfqg = cfqg;
1621         /* cfqq reference on cfqg */
1622         cfqg_get(cfqg);
1623 }
1624
1625 static u64 cfqg_prfill_weight_device(struct seq_file *sf,
1626                                      struct blkg_policy_data *pd, int off)
1627 {
1628         struct cfq_group *cfqg = pd_to_cfqg(pd);
1629
1630         if (!cfqg->dev_weight)
1631                 return 0;
1632         return __blkg_prfill_u64(sf, pd, cfqg->dev_weight);
1633 }
1634
1635 static int cfqg_print_weight_device(struct cgroup_subsys_state *css,
1636                                     struct cftype *cft, struct seq_file *sf)
1637 {
1638         blkcg_print_blkgs(sf, css_to_blkcg(css), cfqg_prfill_weight_device,
1639                           &blkcg_policy_cfq, 0, false);
1640         return 0;
1641 }
1642
1643 static u64 cfqg_prfill_leaf_weight_device(struct seq_file *sf,
1644                                           struct blkg_policy_data *pd, int off)
1645 {
1646         struct cfq_group *cfqg = pd_to_cfqg(pd);
1647
1648         if (!cfqg->dev_leaf_weight)
1649                 return 0;
1650         return __blkg_prfill_u64(sf, pd, cfqg->dev_leaf_weight);
1651 }
1652
1653 static int cfqg_print_leaf_weight_device(struct cgroup_subsys_state *css,
1654                                          struct cftype *cft,
1655                                          struct seq_file *sf)
1656 {
1657         blkcg_print_blkgs(sf, css_to_blkcg(css), cfqg_prfill_leaf_weight_device,
1658                           &blkcg_policy_cfq, 0, false);
1659         return 0;
1660 }
1661
1662 static int cfq_print_weight(struct cgroup_subsys_state *css, struct cftype *cft,
1663                             struct seq_file *sf)
1664 {
1665         seq_printf(sf, "%u\n", css_to_blkcg(css)->cfq_weight);
1666         return 0;
1667 }
1668
1669 static int cfq_print_leaf_weight(struct cgroup_subsys_state *css,
1670                                  struct cftype *cft, struct seq_file *sf)
1671 {
1672         seq_printf(sf, "%u\n", css_to_blkcg(css)->cfq_leaf_weight);
1673         return 0;
1674 }
1675
1676 static int __cfqg_set_weight_device(struct cgroup_subsys_state *css,
1677                                     struct cftype *cft, const char *buf,
1678                                     bool is_leaf_weight)
1679 {
1680         struct blkcg *blkcg = css_to_blkcg(css);
1681         struct blkg_conf_ctx ctx;
1682         struct cfq_group *cfqg;
1683         int ret;
1684
1685         ret = blkg_conf_prep(blkcg, &blkcg_policy_cfq, buf, &ctx);
1686         if (ret)
1687                 return ret;
1688
1689         ret = -EINVAL;
1690         cfqg = blkg_to_cfqg(ctx.blkg);
1691         if (!ctx.v || (ctx.v >= CFQ_WEIGHT_MIN && ctx.v <= CFQ_WEIGHT_MAX)) {
1692                 if (!is_leaf_weight) {
1693                         cfqg->dev_weight = ctx.v;
1694                         cfqg->new_weight = ctx.v ?: blkcg->cfq_weight;
1695                 } else {
1696                         cfqg->dev_leaf_weight = ctx.v;
1697                         cfqg->new_leaf_weight = ctx.v ?: blkcg->cfq_leaf_weight;
1698                 }
1699                 ret = 0;
1700         }
1701
1702         blkg_conf_finish(&ctx);
1703         return ret;
1704 }
1705
1706 static int cfqg_set_weight_device(struct cgroup_subsys_state *css,
1707                                   struct cftype *cft, const char *buf)
1708 {
1709         return __cfqg_set_weight_device(css, cft, buf, false);
1710 }
1711
1712 static int cfqg_set_leaf_weight_device(struct cgroup_subsys_state *css,
1713                                        struct cftype *cft, const char *buf)
1714 {
1715         return __cfqg_set_weight_device(css, cft, buf, true);
1716 }
1717
1718 static int __cfq_set_weight(struct cgroup_subsys_state *css, struct cftype *cft,
1719                             u64 val, bool is_leaf_weight)
1720 {
1721         struct blkcg *blkcg = css_to_blkcg(css);
1722         struct blkcg_gq *blkg;
1723
1724         if (val < CFQ_WEIGHT_MIN || val > CFQ_WEIGHT_MAX)
1725                 return -EINVAL;
1726
1727         spin_lock_irq(&blkcg->lock);
1728
1729         if (!is_leaf_weight)
1730                 blkcg->cfq_weight = val;
1731         else
1732                 blkcg->cfq_leaf_weight = val;
1733
1734         hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
1735                 struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1736
1737                 if (!cfqg)
1738                         continue;
1739
1740                 if (!is_leaf_weight) {
1741                         if (!cfqg->dev_weight)
1742                                 cfqg->new_weight = blkcg->cfq_weight;
1743                 } else {
1744                         if (!cfqg->dev_leaf_weight)
1745                                 cfqg->new_leaf_weight = blkcg->cfq_leaf_weight;
1746                 }
1747         }
1748
1749         spin_unlock_irq(&blkcg->lock);
1750         return 0;
1751 }
1752
1753 static int cfq_set_weight(struct cgroup_subsys_state *css, struct cftype *cft,
1754                           u64 val)
1755 {
1756         return __cfq_set_weight(css, cft, val, false);
1757 }
1758
1759 static int cfq_set_leaf_weight(struct cgroup_subsys_state *css,
1760                                struct cftype *cft, u64 val)
1761 {
1762         return __cfq_set_weight(css, cft, val, true);
1763 }
1764
1765 static int cfqg_print_stat(struct cgroup_subsys_state *css, struct cftype *cft,
1766                            struct seq_file *sf)
1767 {
1768         struct blkcg *blkcg = css_to_blkcg(css);
1769
1770         blkcg_print_blkgs(sf, blkcg, blkg_prfill_stat, &blkcg_policy_cfq,
1771                           cft->private, false);
1772         return 0;
1773 }
1774
1775 static int cfqg_print_rwstat(struct cgroup_subsys_state *css,
1776                              struct cftype *cft, struct seq_file *sf)
1777 {
1778         struct blkcg *blkcg = css_to_blkcg(css);
1779
1780         blkcg_print_blkgs(sf, blkcg, blkg_prfill_rwstat, &blkcg_policy_cfq,
1781                           cft->private, true);
1782         return 0;
1783 }
1784
1785 static u64 cfqg_prfill_stat_recursive(struct seq_file *sf,
1786                                       struct blkg_policy_data *pd, int off)
1787 {
1788         u64 sum = cfqg_stat_pd_recursive_sum(pd, off);
1789
1790         return __blkg_prfill_u64(sf, pd, sum);
1791 }
1792
1793 static u64 cfqg_prfill_rwstat_recursive(struct seq_file *sf,
1794                                         struct blkg_policy_data *pd, int off)
1795 {
1796         struct blkg_rwstat sum = cfqg_rwstat_pd_recursive_sum(pd, off);
1797
1798         return __blkg_prfill_rwstat(sf, pd, &sum);
1799 }
1800
1801 static int cfqg_print_stat_recursive(struct cgroup_subsys_state *css,
1802                                      struct cftype *cft, struct seq_file *sf)
1803 {
1804         struct blkcg *blkcg = css_to_blkcg(css);
1805
1806         blkcg_print_blkgs(sf, blkcg, cfqg_prfill_stat_recursive,
1807                           &blkcg_policy_cfq, cft->private, false);
1808         return 0;
1809 }
1810
1811 static int cfqg_print_rwstat_recursive(struct cgroup_subsys_state *css,
1812                                        struct cftype *cft, struct seq_file *sf)
1813 {
1814         struct blkcg *blkcg = css_to_blkcg(css);
1815
1816         blkcg_print_blkgs(sf, blkcg, cfqg_prfill_rwstat_recursive,
1817                           &blkcg_policy_cfq, cft->private, true);
1818         return 0;
1819 }
1820
1821 #ifdef CONFIG_DEBUG_BLK_CGROUP
1822 static u64 cfqg_prfill_avg_queue_size(struct seq_file *sf,
1823                                       struct blkg_policy_data *pd, int off)
1824 {
1825         struct cfq_group *cfqg = pd_to_cfqg(pd);
1826         u64 samples = blkg_stat_read(&cfqg->stats.avg_queue_size_samples);
1827         u64 v = 0;
1828
1829         if (samples) {
1830                 v = blkg_stat_read(&cfqg->stats.avg_queue_size_sum);
1831                 v = div64_u64(v, samples);
1832         }
1833         __blkg_prfill_u64(sf, pd, v);
1834         return 0;
1835 }
1836
1837 /* print avg_queue_size */
1838 static int cfqg_print_avg_queue_size(struct cgroup_subsys_state *css,
1839                                      struct cftype *cft, struct seq_file *sf)
1840 {
1841         struct blkcg *blkcg = css_to_blkcg(css);
1842
1843         blkcg_print_blkgs(sf, blkcg, cfqg_prfill_avg_queue_size,
1844                           &blkcg_policy_cfq, 0, false);
1845         return 0;
1846 }
1847 #endif  /* CONFIG_DEBUG_BLK_CGROUP */
1848
1849 static struct cftype cfq_blkcg_files[] = {
1850         /* on root, weight is mapped to leaf_weight */
1851         {
1852                 .name = "weight_device",
1853                 .flags = CFTYPE_ONLY_ON_ROOT,
1854                 .read_seq_string = cfqg_print_leaf_weight_device,
1855                 .write_string = cfqg_set_leaf_weight_device,
1856                 .max_write_len = 256,
1857         },
1858         {
1859                 .name = "weight",
1860                 .flags = CFTYPE_ONLY_ON_ROOT,
1861                 .read_seq_string = cfq_print_leaf_weight,
1862                 .write_u64 = cfq_set_leaf_weight,
1863         },
1864
1865         /* no such mapping necessary for !roots */
1866         {
1867                 .name = "weight_device",
1868                 .flags = CFTYPE_NOT_ON_ROOT,
1869                 .read_seq_string = cfqg_print_weight_device,
1870                 .write_string = cfqg_set_weight_device,
1871                 .max_write_len = 256,
1872         },
1873         {
1874                 .name = "weight",
1875                 .flags = CFTYPE_NOT_ON_ROOT,
1876                 .read_seq_string = cfq_print_weight,
1877                 .write_u64 = cfq_set_weight,
1878         },
1879
1880         {
1881                 .name = "leaf_weight_device",
1882                 .read_seq_string = cfqg_print_leaf_weight_device,
1883                 .write_string = cfqg_set_leaf_weight_device,
1884                 .max_write_len = 256,
1885         },
1886         {
1887                 .name = "leaf_weight",
1888                 .read_seq_string = cfq_print_leaf_weight,
1889                 .write_u64 = cfq_set_leaf_weight,
1890         },
1891
1892         /* statistics, covers only the tasks in the cfqg */
1893         {
1894                 .name = "time",
1895                 .private = offsetof(struct cfq_group, stats.time),
1896                 .read_seq_string = cfqg_print_stat,
1897         },
1898         {
1899                 .name = "sectors",
1900                 .private = offsetof(struct cfq_group, stats.sectors),
1901                 .read_seq_string = cfqg_print_stat,
1902         },
1903         {
1904                 .name = "io_service_bytes",
1905                 .private = offsetof(struct cfq_group, stats.service_bytes),
1906                 .read_seq_string = cfqg_print_rwstat,
1907         },
1908         {
1909                 .name = "io_serviced",
1910                 .private = offsetof(struct cfq_group, stats.serviced),
1911                 .read_seq_string = cfqg_print_rwstat,
1912         },
1913         {
1914                 .name = "io_service_time",
1915                 .private = offsetof(struct cfq_group, stats.service_time),
1916                 .read_seq_string = cfqg_print_rwstat,
1917         },
1918         {
1919                 .name = "io_wait_time",
1920                 .private = offsetof(struct cfq_group, stats.wait_time),
1921                 .read_seq_string = cfqg_print_rwstat,
1922         },
1923         {
1924                 .name = "io_merged",
1925                 .private = offsetof(struct cfq_group, stats.merged),
1926                 .read_seq_string = cfqg_print_rwstat,
1927         },
1928         {
1929                 .name = "io_queued",
1930                 .private = offsetof(struct cfq_group, stats.queued),
1931                 .read_seq_string = cfqg_print_rwstat,
1932         },
1933
1934         /* the same statictics which cover the cfqg and its descendants */
1935         {
1936                 .name = "time_recursive",
1937                 .private = offsetof(struct cfq_group, stats.time),
1938                 .read_seq_string = cfqg_print_stat_recursive,
1939         },
1940         {
1941                 .name = "sectors_recursive",
1942                 .private = offsetof(struct cfq_group, stats.sectors),
1943                 .read_seq_string = cfqg_print_stat_recursive,
1944         },
1945         {
1946                 .name = "io_service_bytes_recursive",
1947                 .private = offsetof(struct cfq_group, stats.service_bytes),
1948                 .read_seq_string = cfqg_print_rwstat_recursive,
1949         },
1950         {
1951                 .name = "io_serviced_recursive",
1952                 .private = offsetof(struct cfq_group, stats.serviced),
1953                 .read_seq_string = cfqg_print_rwstat_recursive,
1954         },
1955         {
1956                 .name = "io_service_time_recursive",
1957                 .private = offsetof(struct cfq_group, stats.service_time),
1958                 .read_seq_string = cfqg_print_rwstat_recursive,
1959         },
1960         {
1961                 .name = "io_wait_time_recursive",
1962                 .private = offsetof(struct cfq_group, stats.wait_time),
1963                 .read_seq_string = cfqg_print_rwstat_recursive,
1964         },
1965         {
1966                 .name = "io_merged_recursive",
1967                 .private = offsetof(struct cfq_group, stats.merged),
1968                 .read_seq_string = cfqg_print_rwstat_recursive,
1969         },
1970         {
1971                 .name = "io_queued_recursive",
1972                 .private = offsetof(struct cfq_group, stats.queued),
1973                 .read_seq_string = cfqg_print_rwstat_recursive,
1974         },
1975 #ifdef CONFIG_DEBUG_BLK_CGROUP
1976         {
1977                 .name = "avg_queue_size",
1978                 .read_seq_string = cfqg_print_avg_queue_size,
1979         },
1980         {
1981                 .name = "group_wait_time",
1982                 .private = offsetof(struct cfq_group, stats.group_wait_time),
1983                 .read_seq_string = cfqg_print_stat,
1984         },
1985         {
1986                 .name = "idle_time",
1987                 .private = offsetof(struct cfq_group, stats.idle_time),
1988                 .read_seq_string = cfqg_print_stat,
1989         },
1990         {
1991                 .name = "empty_time",
1992                 .private = offsetof(struct cfq_group, stats.empty_time),
1993                 .read_seq_string = cfqg_print_stat,
1994         },
1995         {
1996                 .name = "dequeue",
1997                 .private = offsetof(struct cfq_group, stats.dequeue),
1998                 .read_seq_string = cfqg_print_stat,
1999         },
2000         {
2001                 .name = "unaccounted_time",
2002                 .private = offsetof(struct cfq_group, stats.unaccounted_time),
2003                 .read_seq_string = cfqg_print_stat,
2004         },
2005 #endif  /* CONFIG_DEBUG_BLK_CGROUP */
2006         { }     /* terminate */
2007 };
2008 #else /* GROUP_IOSCHED */
2009 static struct cfq_group *cfq_lookup_create_cfqg(struct cfq_data *cfqd,
2010                                                 struct blkcg *blkcg)
2011 {
2012         return cfqd->root_group;
2013 }
2014
2015 static inline void
2016 cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg) {
2017         cfqq->cfqg = cfqg;
2018 }
2019
2020 #endif /* GROUP_IOSCHED */
2021
2022 /*
2023  * The cfqd->service_trees holds all pending cfq_queue's that have
2024  * requests waiting to be processed. It is sorted in the order that
2025  * we will service the queues.
2026  */
2027 static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2028                                  bool add_front)
2029 {
2030         struct rb_node **p, *parent;
2031         struct cfq_queue *__cfqq;
2032         unsigned long rb_key;
2033         struct cfq_rb_root *st;
2034         int left;
2035         int new_cfqq = 1;
2036
2037         st = st_for(cfqq->cfqg, cfqq_class(cfqq), cfqq_type(cfqq));
2038         if (cfq_class_idle(cfqq)) {
2039                 rb_key = CFQ_IDLE_DELAY;
2040                 parent = rb_last(&st->rb);
2041                 if (parent && parent != &cfqq->rb_node) {
2042                         __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2043                         rb_key += __cfqq->rb_key;
2044                 } else
2045                         rb_key += jiffies;
2046         } else if (!add_front) {
2047                 /*
2048                  * Get our rb key offset. Subtract any residual slice
2049                  * value carried from last service. A negative resid
2050                  * count indicates slice overrun, and this should position
2051                  * the next service time further away in the tree.
2052                  */
2053                 rb_key = cfq_slice_offset(cfqd, cfqq) + jiffies;
2054                 rb_key -= cfqq->slice_resid;
2055                 cfqq->slice_resid = 0;
2056         } else {
2057                 rb_key = -HZ;
2058                 __cfqq = cfq_rb_first(st);
2059                 rb_key += __cfqq ? __cfqq->rb_key : jiffies;
2060         }
2061
2062         if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2063                 new_cfqq = 0;
2064                 /*
2065                  * same position, nothing more to do
2066                  */
2067                 if (rb_key == cfqq->rb_key && cfqq->service_tree == st)
2068                         return;
2069
2070                 cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2071                 cfqq->service_tree = NULL;
2072         }
2073
2074         left = 1;
2075         parent = NULL;
2076         cfqq->service_tree = st;
2077         p = &st->rb.rb_node;
2078         while (*p) {
2079                 parent = *p;
2080                 __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2081
2082                 /*
2083                  * sort by key, that represents service time.
2084                  */
2085                 if (time_before(rb_key, __cfqq->rb_key))
2086                         p = &parent->rb_left;
2087                 else {
2088                         p = &parent->rb_right;
2089                         left = 0;
2090                 }
2091         }
2092
2093         if (left)
2094                 st->left = &cfqq->rb_node;
2095
2096         cfqq->rb_key = rb_key;
2097         rb_link_node(&cfqq->rb_node, parent, p);
2098         rb_insert_color(&cfqq->rb_node, &st->rb);
2099         st->count++;
2100         if (add_front || !new_cfqq)
2101                 return;
2102         cfq_group_notify_queue_add(cfqd, cfqq->cfqg);
2103 }
2104
2105 static struct cfq_queue *
2106 cfq_prio_tree_lookup(struct cfq_data *cfqd, struct rb_root *root,
2107                      sector_t sector, struct rb_node **ret_parent,
2108                      struct rb_node ***rb_link)
2109 {
2110         struct rb_node **p, *parent;
2111         struct cfq_queue *cfqq = NULL;
2112
2113         parent = NULL;
2114         p = &root->rb_node;
2115         while (*p) {
2116                 struct rb_node **n;
2117
2118                 parent = *p;
2119                 cfqq = rb_entry(parent, struct cfq_queue, p_node);
2120
2121                 /*
2122                  * Sort strictly based on sector.  Smallest to the left,
2123                  * largest to the right.
2124                  */
2125                 if (sector > blk_rq_pos(cfqq->next_rq))
2126                         n = &(*p)->rb_right;
2127                 else if (sector < blk_rq_pos(cfqq->next_rq))
2128                         n = &(*p)->rb_left;
2129                 else
2130                         break;
2131                 p = n;
2132                 cfqq = NULL;
2133         }
2134
2135         *ret_parent = parent;
2136         if (rb_link)
2137                 *rb_link = p;
2138         return cfqq;
2139 }
2140
2141 static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2142 {
2143         struct rb_node **p, *parent;
2144         struct cfq_queue *__cfqq;
2145
2146         if (cfqq->p_root) {
2147                 rb_erase(&cfqq->p_node, cfqq->p_root);
2148                 cfqq->p_root = NULL;
2149         }
2150
2151         if (cfq_class_idle(cfqq))
2152                 return;
2153         if (!cfqq->next_rq)
2154                 return;
2155
2156         cfqq->p_root = &cfqd->prio_trees[cfqq->org_ioprio];
2157         __cfqq = cfq_prio_tree_lookup(cfqd, cfqq->p_root,
2158                                       blk_rq_pos(cfqq->next_rq), &parent, &p);
2159         if (!__cfqq) {
2160                 rb_link_node(&cfqq->p_node, parent, p);
2161                 rb_insert_color(&cfqq->p_node, cfqq->p_root);
2162         } else
2163                 cfqq->p_root = NULL;
2164 }
2165
2166 /*
2167  * Update cfqq's position in the service tree.
2168  */
2169 static void cfq_resort_rr_list(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2170 {
2171         /*
2172          * Resorting requires the cfqq to be on the RR list already.
2173          */
2174         if (cfq_cfqq_on_rr(cfqq)) {
2175                 cfq_service_tree_add(cfqd, cfqq, 0);
2176                 cfq_prio_tree_add(cfqd, cfqq);
2177         }
2178 }
2179
2180 /*
2181  * add to busy list of queues for service, trying to be fair in ordering
2182  * the pending list according to last request service
2183  */
2184 static void cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2185 {
2186         cfq_log_cfqq(cfqd, cfqq, "add_to_rr");
2187         BUG_ON(cfq_cfqq_on_rr(cfqq));
2188         cfq_mark_cfqq_on_rr(cfqq);
2189         cfqd->busy_queues++;
2190         if (cfq_cfqq_sync(cfqq))
2191                 cfqd->busy_sync_queues++;
2192
2193         cfq_resort_rr_list(cfqd, cfqq);
2194 }
2195
2196 /*
2197  * Called when the cfqq no longer has requests pending, remove it from
2198  * the service tree.
2199  */
2200 static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2201 {
2202         cfq_log_cfqq(cfqd, cfqq, "del_from_rr");
2203         BUG_ON(!cfq_cfqq_on_rr(cfqq));
2204         cfq_clear_cfqq_on_rr(cfqq);
2205
2206         if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2207                 cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2208                 cfqq->service_tree = NULL;
2209         }
2210         if (cfqq->p_root) {
2211                 rb_erase(&cfqq->p_node, cfqq->p_root);
2212                 cfqq->p_root = NULL;
2213         }
2214
2215         cfq_group_notify_queue_del(cfqd, cfqq->cfqg);
2216         BUG_ON(!cfqd->busy_queues);
2217         cfqd->busy_queues--;
2218         if (cfq_cfqq_sync(cfqq))
2219                 cfqd->busy_sync_queues--;
2220 }
2221
2222 /*
2223  * rb tree support functions
2224  */
2225 static void cfq_del_rq_rb(struct request *rq)
2226 {
2227         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2228         const int sync = rq_is_sync(rq);
2229
2230         BUG_ON(!cfqq->queued[sync]);
2231         cfqq->queued[sync]--;
2232
2233         elv_rb_del(&cfqq->sort_list, rq);
2234
2235         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list)) {
2236                 /*
2237                  * Queue will be deleted from service tree when we actually
2238                  * expire it later. Right now just remove it from prio tree
2239                  * as it is empty.
2240                  */
2241                 if (cfqq->p_root) {
2242                         rb_erase(&cfqq->p_node, cfqq->p_root);
2243                         cfqq->p_root = NULL;
2244                 }
2245         }
2246 }
2247
2248 static void cfq_add_rq_rb(struct request *rq)
2249 {
2250         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2251         struct cfq_data *cfqd = cfqq->cfqd;
2252         struct request *prev;
2253
2254         cfqq->queued[rq_is_sync(rq)]++;
2255
2256         elv_rb_add(&cfqq->sort_list, rq);
2257
2258         if (!cfq_cfqq_on_rr(cfqq))
2259                 cfq_add_cfqq_rr(cfqd, cfqq);
2260
2261         /*
2262          * check if this request is a better next-serve candidate
2263          */
2264         prev = cfqq->next_rq;
2265         cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq, cfqd->last_position);
2266
2267         /*
2268          * adjust priority tree position, if ->next_rq changes
2269          */
2270         if (prev != cfqq->next_rq)
2271                 cfq_prio_tree_add(cfqd, cfqq);
2272
2273         BUG_ON(!cfqq->next_rq);
2274 }
2275
2276 static void cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
2277 {
2278         elv_rb_del(&cfqq->sort_list, rq);
2279         cfqq->queued[rq_is_sync(rq)]--;
2280         cfqg_stats_update_io_remove(RQ_CFQG(rq), rq->cmd_flags);
2281         cfq_add_rq_rb(rq);
2282         cfqg_stats_update_io_add(RQ_CFQG(rq), cfqq->cfqd->serving_group,
2283                                  rq->cmd_flags);
2284 }
2285
2286 static struct request *
2287 cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio)
2288 {
2289         struct task_struct *tsk = current;
2290         struct cfq_io_cq *cic;
2291         struct cfq_queue *cfqq;
2292
2293         cic = cfq_cic_lookup(cfqd, tsk->io_context);
2294         if (!cic)
2295                 return NULL;
2296
2297         cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
2298         if (cfqq)
2299                 return elv_rb_find(&cfqq->sort_list, bio_end_sector(bio));
2300
2301         return NULL;
2302 }
2303
2304 static void cfq_activate_request(struct request_queue *q, struct request *rq)
2305 {
2306         struct cfq_data *cfqd = q->elevator->elevator_data;
2307
2308         cfqd->rq_in_driver++;
2309         cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "activate rq, drv=%d",
2310                                                 cfqd->rq_in_driver);
2311
2312         cfqd->last_position = blk_rq_pos(rq) + blk_rq_sectors(rq);
2313 }
2314
2315 static void cfq_deactivate_request(struct request_queue *q, struct request *rq)
2316 {
2317         struct cfq_data *cfqd = q->elevator->elevator_data;
2318
2319         WARN_ON(!cfqd->rq_in_driver);
2320         cfqd->rq_in_driver--;
2321         cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "deactivate rq, drv=%d",
2322                                                 cfqd->rq_in_driver);
2323 }
2324
2325 static void cfq_remove_request(struct request *rq)
2326 {
2327         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2328
2329         if (cfqq->next_rq == rq)
2330                 cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq);
2331
2332         list_del_init(&rq->queuelist);
2333         cfq_del_rq_rb(rq);
2334
2335         cfqq->cfqd->rq_queued--;
2336         cfqg_stats_update_io_remove(RQ_CFQG(rq), rq->cmd_flags);
2337         if (rq->cmd_flags & REQ_PRIO) {
2338                 WARN_ON(!cfqq->prio_pending);
2339                 cfqq->prio_pending--;
2340         }
2341 }
2342
2343 static int cfq_merge(struct request_queue *q, struct request **req,
2344                      struct bio *bio)
2345 {
2346         struct cfq_data *cfqd = q->elevator->elevator_data;
2347         struct request *__rq;
2348
2349         __rq = cfq_find_rq_fmerge(cfqd, bio);
2350         if (__rq && elv_rq_merge_ok(__rq, bio)) {
2351                 *req = __rq;
2352                 return ELEVATOR_FRONT_MERGE;
2353         }
2354
2355         return ELEVATOR_NO_MERGE;
2356 }
2357
2358 static void cfq_merged_request(struct request_queue *q, struct request *req,
2359                                int type)
2360 {
2361         if (type == ELEVATOR_FRONT_MERGE) {
2362                 struct cfq_queue *cfqq = RQ_CFQQ(req);
2363
2364                 cfq_reposition_rq_rb(cfqq, req);
2365         }
2366 }
2367
2368 static void cfq_bio_merged(struct request_queue *q, struct request *req,
2369                                 struct bio *bio)
2370 {
2371         cfqg_stats_update_io_merged(RQ_CFQG(req), bio->bi_rw);
2372 }
2373
2374 static void
2375 cfq_merged_requests(struct request_queue *q, struct request *rq,
2376                     struct request *next)
2377 {
2378         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2379         struct cfq_data *cfqd = q->elevator->elevator_data;
2380
2381         /*
2382          * reposition in fifo if next is older than rq
2383          */
2384         if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
2385             time_before(rq_fifo_time(next), rq_fifo_time(rq)) &&
2386             cfqq == RQ_CFQQ(next)) {
2387                 list_move(&rq->queuelist, &next->queuelist);
2388                 rq_set_fifo_time(rq, rq_fifo_time(next));
2389         }
2390
2391         if (cfqq->next_rq == next)
2392                 cfqq->next_rq = rq;
2393         cfq_remove_request(next);
2394         cfqg_stats_update_io_merged(RQ_CFQG(rq), next->cmd_flags);
2395
2396         cfqq = RQ_CFQQ(next);
2397         /*
2398          * all requests of this queue are merged to other queues, delete it
2399          * from the service tree. If it's the active_queue,
2400          * cfq_dispatch_requests() will choose to expire it or do idle
2401          */
2402         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list) &&
2403             cfqq != cfqd->active_queue)
2404                 cfq_del_cfqq_rr(cfqd, cfqq);
2405 }
2406
2407 static int cfq_allow_merge(struct request_queue *q, struct request *rq,
2408                            struct bio *bio)
2409 {
2410         struct cfq_data *cfqd = q->elevator->elevator_data;
2411         struct cfq_io_cq *cic;
2412         struct cfq_queue *cfqq;
2413
2414         /*
2415          * Disallow merge of a sync bio into an async request.
2416          */
2417         if (cfq_bio_sync(bio) && !rq_is_sync(rq))
2418                 return false;
2419
2420         /*
2421          * Lookup the cfqq that this bio will be queued with and allow
2422          * merge only if rq is queued there.
2423          */
2424         cic = cfq_cic_lookup(cfqd, current->io_context);
2425         if (!cic)
2426                 return false;
2427
2428         cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
2429         return cfqq == RQ_CFQQ(rq);
2430 }
2431
2432 static inline void cfq_del_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2433 {
2434         del_timer(&cfqd->idle_slice_timer);
2435         cfqg_stats_update_idle_time(cfqq->cfqg);
2436 }
2437
2438 static void __cfq_set_active_queue(struct cfq_data *cfqd,
2439                                    struct cfq_queue *cfqq)
2440 {
2441         if (cfqq) {
2442                 cfq_log_cfqq(cfqd, cfqq, "set_active wl_class:%d wl_type:%d",
2443                                 cfqd->serving_wl_class, cfqd->serving_wl_type);
2444                 cfqg_stats_update_avg_queue_size(cfqq->cfqg);
2445                 cfqq->slice_start = 0;
2446                 cfqq->dispatch_start = jiffies;
2447                 cfqq->allocated_slice = 0;
2448                 cfqq->slice_end = 0;
2449                 cfqq->slice_dispatch = 0;
2450                 cfqq->nr_sectors = 0;
2451
2452                 cfq_clear_cfqq_wait_request(cfqq);
2453                 cfq_clear_cfqq_must_dispatch(cfqq);
2454                 cfq_clear_cfqq_must_alloc_slice(cfqq);
2455                 cfq_clear_cfqq_fifo_expire(cfqq);
2456                 cfq_mark_cfqq_slice_new(cfqq);
2457
2458                 cfq_del_timer(cfqd, cfqq);
2459         }
2460
2461         cfqd->active_queue = cfqq;
2462 }
2463
2464 /*
2465  * current cfqq expired its slice (or was too idle), select new one
2466  */
2467 static void
2468 __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2469                     bool timed_out)
2470 {
2471         cfq_log_cfqq(cfqd, cfqq, "slice expired t=%d", timed_out);
2472
2473         if (cfq_cfqq_wait_request(cfqq))
2474                 cfq_del_timer(cfqd, cfqq);
2475
2476         cfq_clear_cfqq_wait_request(cfqq);
2477         cfq_clear_cfqq_wait_busy(cfqq);
2478
2479         /*
2480          * If this cfqq is shared between multiple processes, check to
2481          * make sure that those processes are still issuing I/Os within
2482          * the mean seek distance.  If not, it may be time to break the
2483          * queues apart again.
2484          */
2485         if (cfq_cfqq_coop(cfqq) && CFQQ_SEEKY(cfqq))
2486                 cfq_mark_cfqq_split_coop(cfqq);
2487
2488         /*
2489          * store what was left of this slice, if the queue idled/timed out
2490          */
2491         if (timed_out) {
2492                 if (cfq_cfqq_slice_new(cfqq))
2493                         cfqq->slice_resid = cfq_scaled_cfqq_slice(cfqd, cfqq);
2494                 else
2495                         cfqq->slice_resid = cfqq->slice_end - jiffies;
2496                 cfq_log_cfqq(cfqd, cfqq, "resid=%ld", cfqq->slice_resid);
2497         }
2498
2499         cfq_group_served(cfqd, cfqq->cfqg, cfqq);
2500
2501         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list))
2502                 cfq_del_cfqq_rr(cfqd, cfqq);
2503
2504         cfq_resort_rr_list(cfqd, cfqq);
2505
2506         if (cfqq == cfqd->active_queue)
2507                 cfqd->active_queue = NULL;
2508
2509         if (cfqd->active_cic) {
2510                 put_io_context(cfqd->active_cic->icq.ioc);
2511                 cfqd->active_cic = NULL;
2512         }
2513 }
2514
2515 static inline void cfq_slice_expired(struct cfq_data *cfqd, bool timed_out)
2516 {
2517         struct cfq_queue *cfqq = cfqd->active_queue;
2518
2519         if (cfqq)
2520                 __cfq_slice_expired(cfqd, cfqq, timed_out);
2521 }
2522
2523 /*
2524  * Get next queue for service. Unless we have a queue preemption,
2525  * we'll simply select the first cfqq in the service tree.
2526  */
2527 static struct cfq_queue *cfq_get_next_queue(struct cfq_data *cfqd)
2528 {
2529         struct cfq_rb_root *st = st_for(cfqd->serving_group,
2530                         cfqd->serving_wl_class, cfqd->serving_wl_type);
2531
2532         if (!cfqd->rq_queued)
2533                 return NULL;
2534
2535         /* There is nothing to dispatch */
2536         if (!st)
2537                 return NULL;
2538         if (RB_EMPTY_ROOT(&st->rb))
2539                 return NULL;
2540         return cfq_rb_first(st);
2541 }
2542
2543 static struct cfq_queue *cfq_get_next_queue_forced(struct cfq_data *cfqd)
2544 {
2545         struct cfq_group *cfqg;
2546         struct cfq_queue *cfqq;
2547         int i, j;
2548         struct cfq_rb_root *st;
2549
2550         if (!cfqd->rq_queued)
2551                 return NULL;
2552
2553         cfqg = cfq_get_next_cfqg(cfqd);
2554         if (!cfqg)
2555                 return NULL;
2556
2557         for_each_cfqg_st(cfqg, i, j, st)
2558                 if ((cfqq = cfq_rb_first(st)) != NULL)
2559                         return cfqq;
2560         return NULL;
2561 }
2562
2563 /*
2564  * Get and set a new active queue for service.
2565  */
2566 static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd,
2567                                               struct cfq_queue *cfqq)
2568 {
2569         if (!cfqq)
2570                 cfqq = cfq_get_next_queue(cfqd);
2571
2572         __cfq_set_active_queue(cfqd, cfqq);
2573         return cfqq;
2574 }
2575
2576 static inline sector_t cfq_dist_from_last(struct cfq_data *cfqd,
2577                                           struct request *rq)
2578 {
2579         if (blk_rq_pos(rq) >= cfqd->last_position)
2580                 return blk_rq_pos(rq) - cfqd->last_position;
2581         else
2582                 return cfqd->last_position - blk_rq_pos(rq);
2583 }
2584
2585 static inline int cfq_rq_close(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2586                                struct request *rq)
2587 {
2588         return cfq_dist_from_last(cfqd, rq) <= CFQQ_CLOSE_THR;
2589 }
2590
2591 static struct cfq_queue *cfqq_close(struct cfq_data *cfqd,
2592                                     struct cfq_queue *cur_cfqq)
2593 {
2594         struct rb_root *root = &cfqd->prio_trees[cur_cfqq->org_ioprio];
2595         struct rb_node *parent, *node;
2596         struct cfq_queue *__cfqq;
2597         sector_t sector = cfqd->last_position;
2598
2599         if (RB_EMPTY_ROOT(root))
2600                 return NULL;
2601
2602         /*
2603          * First, if we find a request starting at the end of the last
2604          * request, choose it.
2605          */
2606         __cfqq = cfq_prio_tree_lookup(cfqd, root, sector, &parent, NULL);
2607         if (__cfqq)
2608                 return __cfqq;
2609
2610         /*
2611          * If the exact sector wasn't found, the parent of the NULL leaf
2612          * will contain the closest sector.
2613          */
2614         __cfqq = rb_entry(parent, struct cfq_queue, p_node);
2615         if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2616                 return __cfqq;
2617
2618         if (blk_rq_pos(__cfqq->next_rq) < sector)
2619                 node = rb_next(&__cfqq->p_node);
2620         else
2621                 node = rb_prev(&__cfqq->p_node);
2622         if (!node)
2623                 return NULL;
2624
2625         __cfqq = rb_entry(node, struct cfq_queue, p_node);
2626         if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2627                 return __cfqq;
2628
2629         return NULL;
2630 }
2631
2632 /*
2633  * cfqd - obvious
2634  * cur_cfqq - passed in so that we don't decide that the current queue is
2635  *            closely cooperating with itself.
2636  *
2637  * So, basically we're assuming that that cur_cfqq has dispatched at least
2638  * one request, and that cfqd->last_position reflects a position on the disk
2639  * associated with the I/O issued by cur_cfqq.  I'm not sure this is a valid
2640  * assumption.
2641  */
2642 static struct cfq_queue *cfq_close_cooperator(struct cfq_data *cfqd,
2643                                               struct cfq_queue *cur_cfqq)
2644 {
2645         struct cfq_queue *cfqq;
2646
2647         if (cfq_class_idle(cur_cfqq))
2648                 return NULL;
2649         if (!cfq_cfqq_sync(cur_cfqq))
2650                 return NULL;
2651         if (CFQQ_SEEKY(cur_cfqq))
2652                 return NULL;
2653
2654         /*
2655          * Don't search priority tree if it's the only queue in the group.
2656          */
2657         if (cur_cfqq->cfqg->nr_cfqq == 1)
2658                 return NULL;
2659
2660         /*
2661          * We should notice if some of the queues are cooperating, eg
2662          * working closely on the same area of the disk. In that case,
2663          * we can group them together and don't waste time idling.
2664          */
2665         cfqq = cfqq_close(cfqd, cur_cfqq);
2666         if (!cfqq)
2667                 return NULL;
2668
2669         /* If new queue belongs to different cfq_group, don't choose it */
2670         if (cur_cfqq->cfqg != cfqq->cfqg)
2671                 return NULL;
2672
2673         /*
2674          * It only makes sense to merge sync queues.
2675          */
2676         if (!cfq_cfqq_sync(cfqq))
2677                 return NULL;
2678         if (CFQQ_SEEKY(cfqq))
2679                 return NULL;
2680
2681         /*
2682          * Do not merge queues of different priority classes
2683          */
2684         if (cfq_class_rt(cfqq) != cfq_class_rt(cur_cfqq))
2685                 return NULL;
2686
2687         return cfqq;
2688 }
2689
2690 /*
2691  * Determine whether we should enforce idle window for this queue.
2692  */
2693
2694 static bool cfq_should_idle(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2695 {
2696         enum wl_class_t wl_class = cfqq_class(cfqq);
2697         struct cfq_rb_root *st = cfqq->service_tree;
2698
2699         BUG_ON(!st);
2700         BUG_ON(!st->count);
2701
2702         if (!cfqd->cfq_slice_idle)
2703                 return false;
2704
2705         /* We never do for idle class queues. */
2706         if (wl_class == IDLE_WORKLOAD)
2707                 return false;
2708
2709         /* We do for queues that were marked with idle window flag. */
2710         if (cfq_cfqq_idle_window(cfqq) &&
2711            !(blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag))
2712                 return true;
2713
2714         /*
2715          * Otherwise, we do only if they are the last ones
2716          * in their service tree.
2717          */
2718         if (st->count == 1 && cfq_cfqq_sync(cfqq) &&
2719            !cfq_io_thinktime_big(cfqd, &st->ttime, false))
2720                 return true;
2721         cfq_log_cfqq(cfqd, cfqq, "Not idling. st->count:%d", st->count);
2722         return false;
2723 }
2724
2725 static void cfq_arm_slice_timer(struct cfq_data *cfqd)
2726 {
2727         struct cfq_queue *cfqq = cfqd->active_queue;
2728         struct cfq_io_cq *cic;
2729         unsigned long sl, group_idle = 0;
2730
2731         /*
2732          * SSD device without seek penalty, disable idling. But only do so
2733          * for devices that support queuing, otherwise we still have a problem
2734          * with sync vs async workloads.
2735          */
2736         if (blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag)
2737                 return;
2738
2739         WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list));
2740         WARN_ON(cfq_cfqq_slice_new(cfqq));
2741
2742         /*
2743          * idle is disabled, either manually or by past process history
2744          */
2745         if (!cfq_should_idle(cfqd, cfqq)) {
2746                 /* no queue idling. Check for group idling */
2747                 if (cfqd->cfq_group_idle)
2748                         group_idle = cfqd->cfq_group_idle;
2749                 else
2750                         return;
2751         }
2752
2753         /*
2754          * still active requests from this queue, don't idle
2755          */
2756         if (cfqq->dispatched)
2757                 return;
2758
2759         /*
2760          * task has exited, don't wait
2761          */
2762         cic = cfqd->active_cic;
2763         if (!cic || !atomic_read(&cic->icq.ioc->active_ref))
2764                 return;
2765
2766         /*
2767          * If our average think time is larger than the remaining time
2768          * slice, then don't idle. This avoids overrunning the allotted
2769          * time slice.
2770          */
2771         if (sample_valid(cic->ttime.ttime_samples) &&
2772             (cfqq->slice_end - jiffies < cic->ttime.ttime_mean)) {
2773                 cfq_log_cfqq(cfqd, cfqq, "Not idling. think_time:%lu",
2774                              cic->ttime.ttime_mean);
2775                 return;
2776         }
2777
2778         /* There are other queues in the group, don't do group idle */
2779         if (group_idle && cfqq->cfqg->nr_cfqq > 1)
2780                 return;
2781
2782         cfq_mark_cfqq_wait_request(cfqq);
2783
2784         if (group_idle)
2785                 sl = cfqd->cfq_group_idle;
2786         else
2787                 sl = cfqd->cfq_slice_idle;
2788
2789         mod_timer(&cfqd->idle_slice_timer, jiffies + sl);
2790         cfqg_stats_set_start_idle_time(cfqq->cfqg);
2791         cfq_log_cfqq(cfqd, cfqq, "arm_idle: %lu group_idle: %d", sl,
2792                         group_idle ? 1 : 0);
2793 }
2794
2795 /*
2796  * Move request from internal lists to the request queue dispatch list.
2797  */
2798 static void cfq_dispatch_insert(struct request_queue *q, struct request *rq)
2799 {
2800         struct cfq_data *cfqd = q->elevator->elevator_data;
2801         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2802
2803         cfq_log_cfqq(cfqd, cfqq, "dispatch_insert");
2804
2805         cfqq->next_rq = cfq_find_next_rq(cfqd, cfqq, rq);
2806         cfq_remove_request(rq);
2807         cfqq->dispatched++;
2808         (RQ_CFQG(rq))->dispatched++;
2809         elv_dispatch_sort(q, rq);
2810
2811         cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]++;
2812         cfqq->nr_sectors += blk_rq_sectors(rq);
2813         cfqg_stats_update_dispatch(cfqq->cfqg, blk_rq_bytes(rq), rq->cmd_flags);
2814 }
2815
2816 /*
2817  * return expired entry, or NULL to just start from scratch in rbtree
2818  */
2819 static struct request *cfq_check_fifo(struct cfq_queue *cfqq)
2820 {
2821         struct request *rq = NULL;
2822
2823         if (cfq_cfqq_fifo_expire(cfqq))
2824                 return NULL;
2825
2826         cfq_mark_cfqq_fifo_expire(cfqq);
2827
2828         if (list_empty(&cfqq->fifo))
2829                 return NULL;
2830
2831         rq = rq_entry_fifo(cfqq->fifo.next);
2832         if (time_before(jiffies, rq_fifo_time(rq)))
2833                 rq = NULL;
2834
2835         cfq_log_cfqq(cfqq->cfqd, cfqq, "fifo=%p", rq);
2836         return rq;
2837 }
2838
2839 static inline int
2840 cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2841 {
2842         const int base_rq = cfqd->cfq_slice_async_rq;
2843
2844         WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
2845
2846         return 2 * base_rq * (IOPRIO_BE_NR - cfqq->ioprio);
2847 }
2848
2849 /*
2850  * Must be called with the queue_lock held.
2851  */
2852 static int cfqq_process_refs(struct cfq_queue *cfqq)
2853 {
2854         int process_refs, io_refs;
2855
2856         io_refs = cfqq->allocated[READ] + cfqq->allocated[WRITE];
2857         process_refs = cfqq->ref - io_refs;
2858         BUG_ON(process_refs < 0);
2859         return process_refs;
2860 }
2861
2862 static void cfq_setup_merge(struct cfq_queue *cfqq, struct cfq_queue *new_cfqq)
2863 {
2864         int process_refs, new_process_refs;
2865         struct cfq_queue *__cfqq;
2866
2867         /*
2868          * If there are no process references on the new_cfqq, then it is
2869          * unsafe to follow the ->new_cfqq chain as other cfqq's in the
2870          * chain may have dropped their last reference (not just their
2871          * last process reference).
2872          */
2873         if (!cfqq_process_refs(new_cfqq))
2874                 return;
2875
2876         /* Avoid a circular list and skip interim queue merges */
2877         while ((__cfqq = new_cfqq->new_cfqq)) {
2878                 if (__cfqq == cfqq)
2879                         return;
2880                 new_cfqq = __cfqq;
2881         }
2882
2883         process_refs = cfqq_process_refs(cfqq);
2884         new_process_refs = cfqq_process_refs(new_cfqq);
2885         /*
2886          * If the process for the cfqq has gone away, there is no
2887          * sense in merging the queues.
2888          */
2889         if (process_refs == 0 || new_process_refs == 0)
2890                 return;
2891
2892         /*
2893          * Merge in the direction of the lesser amount of work.
2894          */
2895         if (new_process_refs >= process_refs) {
2896                 cfqq->new_cfqq = new_cfqq;
2897                 new_cfqq->ref += process_refs;
2898         } else {
2899                 new_cfqq->new_cfqq = cfqq;
2900                 cfqq->ref += new_process_refs;
2901         }
2902 }
2903
2904 static enum wl_type_t cfq_choose_wl_type(struct cfq_data *cfqd,
2905                         struct cfq_group *cfqg, enum wl_class_t wl_class)
2906 {
2907         struct cfq_queue *queue;
2908         int i;
2909         bool key_valid = false;
2910         unsigned long lowest_key = 0;
2911         enum wl_type_t cur_best = SYNC_NOIDLE_WORKLOAD;
2912
2913         for (i = 0; i <= SYNC_WORKLOAD; ++i) {
2914                 /* select the one with lowest rb_key */
2915                 queue = cfq_rb_first(st_for(cfqg, wl_class, i));
2916                 if (queue &&
2917                     (!key_valid || time_before(queue->rb_key, lowest_key))) {
2918                         lowest_key = queue->rb_key;
2919                         cur_best = i;
2920                         key_valid = true;
2921                 }
2922         }
2923
2924         return cur_best;
2925 }
2926
2927 static void
2928 choose_wl_class_and_type(struct cfq_data *cfqd, struct cfq_group *cfqg)
2929 {
2930         unsigned slice;
2931         unsigned count;
2932         struct cfq_rb_root *st;
2933         unsigned group_slice;
2934         enum wl_class_t original_class = cfqd->serving_wl_class;
2935
2936         /* Choose next priority. RT > BE > IDLE */
2937         if (cfq_group_busy_queues_wl(RT_WORKLOAD, cfqd, cfqg))
2938                 cfqd->serving_wl_class = RT_WORKLOAD;
2939         else if (cfq_group_busy_queues_wl(BE_WORKLOAD, cfqd, cfqg))
2940                 cfqd->serving_wl_class = BE_WORKLOAD;
2941         else {
2942                 cfqd->serving_wl_class = IDLE_WORKLOAD;
2943                 cfqd->workload_expires = jiffies + 1;
2944                 return;
2945         }
2946
2947         if (original_class != cfqd->serving_wl_class)
2948                 goto new_workload;
2949
2950         /*
2951          * For RT and BE, we have to choose also the type
2952          * (SYNC, SYNC_NOIDLE, ASYNC), and to compute a workload
2953          * expiration time
2954          */
2955         st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
2956         count = st->count;
2957
2958         /*
2959          * check workload expiration, and that we still have other queues ready
2960          */
2961         if (count && !time_after(jiffies, cfqd->workload_expires))
2962                 return;
2963
2964 new_workload:
2965         /* otherwise select new workload type */
2966         cfqd->serving_wl_type = cfq_choose_wl_type(cfqd, cfqg,
2967                                         cfqd->serving_wl_class);
2968         st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
2969         count = st->count;
2970
2971         /*
2972          * the workload slice is computed as a fraction of target latency
2973          * proportional to the number of queues in that workload, over
2974          * all the queues in the same priority class
2975          */
2976         group_slice = cfq_group_slice(cfqd, cfqg);
2977
2978         slice = group_slice * count /
2979                 max_t(unsigned, cfqg->busy_queues_avg[cfqd->serving_wl_class],
2980                       cfq_group_busy_queues_wl(cfqd->serving_wl_class, cfqd,
2981                                         cfqg));
2982
2983         if (cfqd->serving_wl_type == ASYNC_WORKLOAD) {
2984                 unsigned int tmp;
2985
2986                 /*
2987                  * Async queues are currently system wide. Just taking
2988                  * proportion of queues with-in same group will lead to higher
2989                  * async ratio system wide as generally root group is going
2990                  * to have higher weight. A more accurate thing would be to
2991                  * calculate system wide asnc/sync ratio.
2992                  */
2993                 tmp = cfqd->cfq_target_latency *
2994                         cfqg_busy_async_queues(cfqd, cfqg);
2995                 tmp = tmp/cfqd->busy_queues;
2996                 slice = min_t(unsigned, slice, tmp);
2997
2998                 /* async workload slice is scaled down according to
2999                  * the sync/async slice ratio. */
3000                 slice = slice * cfqd->cfq_slice[0] / cfqd->cfq_slice[1];
3001         } else
3002                 /* sync workload slice is at least 2 * cfq_slice_idle */
3003                 slice = max(slice, 2 * cfqd->cfq_slice_idle);
3004
3005         slice = max_t(unsigned, slice, CFQ_MIN_TT);
3006         cfq_log(cfqd, "workload slice:%d", slice);
3007         cfqd->workload_expires = jiffies + slice;
3008 }
3009
3010 static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd)
3011 {
3012         struct cfq_rb_root *st = &cfqd->grp_service_tree;
3013         struct cfq_group *cfqg;
3014
3015         if (RB_EMPTY_ROOT(&st->rb))
3016                 return NULL;
3017         cfqg = cfq_rb_first_group(st);
3018         update_min_vdisktime(st);
3019         return cfqg;
3020 }
3021
3022 static void cfq_choose_cfqg(struct cfq_data *cfqd)
3023 {
3024         struct cfq_group *cfqg = cfq_get_next_cfqg(cfqd);
3025
3026         cfqd->serving_group = cfqg;
3027
3028         /* Restore the workload type data */
3029         if (cfqg->saved_wl_slice) {
3030                 cfqd->workload_expires = jiffies + cfqg->saved_wl_slice;
3031                 cfqd->serving_wl_type = cfqg->saved_wl_type;
3032                 cfqd->serving_wl_class = cfqg->saved_wl_class;
3033         } else
3034                 cfqd->workload_expires = jiffies - 1;
3035
3036         choose_wl_class_and_type(cfqd, cfqg);
3037 }
3038
3039 /*
3040  * Select a queue for service. If we have a current active queue,
3041  * check whether to continue servicing it, or retrieve and set a new one.
3042  */
3043 static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
3044 {
3045         struct cfq_queue *cfqq, *new_cfqq = NULL;
3046
3047         cfqq = cfqd->active_queue;
3048         if (!cfqq)
3049                 goto new_queue;
3050
3051         if (!cfqd->rq_queued)
3052                 return NULL;
3053
3054         /*
3055          * We were waiting for group to get backlogged. Expire the queue
3056          */
3057         if (cfq_cfqq_wait_busy(cfqq) && !RB_EMPTY_ROOT(&cfqq->sort_list))
3058                 goto expire;
3059
3060         /*
3061          * The active queue has run out of time, expire it and select new.
3062          */
3063         if (cfq_slice_used(cfqq) && !cfq_cfqq_must_dispatch(cfqq)) {
3064                 /*
3065                  * If slice had not expired at the completion of last request
3066                  * we might not have turned on wait_busy flag. Don't expire
3067                  * the queue yet. Allow the group to get backlogged.
3068                  *
3069                  * The very fact that we have used the slice, that means we
3070                  * have been idling all along on this queue and it should be
3071                  * ok to wait for this request to complete.
3072                  */
3073                 if (cfqq->cfqg->nr_cfqq == 1 && RB_EMPTY_ROOT(&cfqq->sort_list)
3074                     && cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3075                         cfqq = NULL;
3076                         goto keep_queue;
3077                 } else
3078                         goto check_group_idle;
3079         }
3080
3081         /*
3082          * The active queue has requests and isn't expired, allow it to
3083          * dispatch.
3084          */
3085         if (!RB_EMPTY_ROOT(&cfqq->sort_list))
3086                 goto keep_queue;
3087
3088         /*
3089          * If another queue has a request waiting within our mean seek
3090          * distance, let it run.  The expire code will check for close
3091          * cooperators and put the close queue at the front of the service
3092          * tree.  If possible, merge the expiring queue with the new cfqq.
3093          */
3094         new_cfqq = cfq_close_cooperator(cfqd, cfqq);
3095         if (new_cfqq) {
3096                 if (!cfqq->new_cfqq)
3097                         cfq_setup_merge(cfqq, new_cfqq);
3098                 goto expire;
3099         }
3100
3101         /*
3102          * No requests pending. If the active queue still has requests in
3103          * flight or is idling for a new request, allow either of these
3104          * conditions to happen (or time out) before selecting a new queue.
3105          */
3106         if (timer_pending(&cfqd->idle_slice_timer)) {
3107                 cfqq = NULL;
3108                 goto keep_queue;
3109         }
3110
3111         /*
3112          * This is a deep seek queue, but the device is much faster than
3113          * the queue can deliver, don't idle
3114          **/
3115         if (CFQQ_SEEKY(cfqq) && cfq_cfqq_idle_window(cfqq) &&
3116             (cfq_cfqq_slice_new(cfqq) ||
3117             (cfqq->slice_end - jiffies > jiffies - cfqq->slice_start))) {
3118                 cfq_clear_cfqq_deep(cfqq);
3119                 cfq_clear_cfqq_idle_window(cfqq);
3120         }
3121
3122         if (cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3123                 cfqq = NULL;
3124                 goto keep_queue;
3125         }
3126
3127         /*
3128          * If group idle is enabled and there are requests dispatched from
3129          * this group, wait for requests to complete.
3130          */
3131 check_group_idle:
3132         if (cfqd->cfq_group_idle && cfqq->cfqg->nr_cfqq == 1 &&
3133             cfqq->cfqg->dispatched &&
3134             !cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true)) {
3135                 cfqq = NULL;
3136                 goto keep_queue;
3137         }
3138
3139 expire:
3140         cfq_slice_expired(cfqd, 0);
3141 new_queue:
3142         /*
3143          * Current queue expired. Check if we have to switch to a new
3144          * service tree
3145          */
3146         if (!new_cfqq)
3147                 cfq_choose_cfqg(cfqd);
3148
3149         cfqq = cfq_set_active_queue(cfqd, new_cfqq);
3150 keep_queue:
3151         return cfqq;
3152 }
3153
3154 static int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq)
3155 {
3156         int dispatched = 0;
3157
3158         while (cfqq->next_rq) {
3159                 cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq);
3160                 dispatched++;
3161         }
3162
3163         BUG_ON(!list_empty(&cfqq->fifo));
3164
3165         /* By default cfqq is not expired if it is empty. Do it explicitly */
3166         __cfq_slice_expired(cfqq->cfqd, cfqq, 0);
3167         return dispatched;
3168 }
3169
3170 /*
3171  * Drain our current requests. Used for barriers and when switching
3172  * io schedulers on-the-fly.
3173  */
3174 static int cfq_forced_dispatch(struct cfq_data *cfqd)
3175 {
3176         struct cfq_queue *cfqq;
3177         int dispatched = 0;
3178
3179         /* Expire the timeslice of the current active queue first */
3180         cfq_slice_expired(cfqd, 0);
3181         while ((cfqq = cfq_get_next_queue_forced(cfqd)) != NULL) {
3182                 __cfq_set_active_queue(cfqd, cfqq);
3183                 dispatched += __cfq_forced_dispatch_cfqq(cfqq);
3184         }
3185
3186         BUG_ON(cfqd->busy_queues);
3187
3188         cfq_log(cfqd, "forced_dispatch=%d", dispatched);
3189         return dispatched;
3190 }
3191
3192 static inline bool cfq_slice_used_soon(struct cfq_data *cfqd,
3193         struct cfq_queue *cfqq)
3194 {
3195         /* the queue hasn't finished any request, can't estimate */
3196         if (cfq_cfqq_slice_new(cfqq))
3197                 return true;
3198         if (time_after(jiffies + cfqd->cfq_slice_idle * cfqq->dispatched,
3199                 cfqq->slice_end))
3200                 return true;
3201
3202         return false;
3203 }
3204
3205 static bool cfq_may_dispatch(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3206 {
3207         unsigned int max_dispatch;
3208
3209         /*
3210          * Drain async requests before we start sync IO
3211          */
3212         if (cfq_should_idle(cfqd, cfqq) && cfqd->rq_in_flight[BLK_RW_ASYNC])
3213                 return false;
3214
3215         /*
3216          * If this is an async queue and we have sync IO in flight, let it wait
3217          */
3218         if (cfqd->rq_in_flight[BLK_RW_SYNC] && !cfq_cfqq_sync(cfqq))
3219                 return false;
3220
3221         max_dispatch = max_t(unsigned int, cfqd->cfq_quantum / 2, 1);
3222         if (cfq_class_idle(cfqq))
3223                 max_dispatch = 1;
3224
3225         /*
3226          * Does this cfqq already have too much IO in flight?
3227          */
3228         if (cfqq->dispatched >= max_dispatch) {
3229                 bool promote_sync = false;
3230                 /*
3231                  * idle queue must always only have a single IO in flight
3232                  */
3233                 if (cfq_class_idle(cfqq))
3234                         return false;
3235
3236                 /*
3237                  * If there is only one sync queue
3238                  * we can ignore async queue here and give the sync
3239                  * queue no dispatch limit. The reason is a sync queue can
3240                  * preempt async queue, limiting the sync queue doesn't make
3241                  * sense. This is useful for aiostress test.
3242                  */
3243                 if (cfq_cfqq_sync(cfqq) && cfqd->busy_sync_queues == 1)
3244                         promote_sync = true;
3245
3246                 /*
3247                  * We have other queues, don't allow more IO from this one
3248                  */
3249                 if (cfqd->busy_queues > 1 && cfq_slice_used_soon(cfqd, cfqq) &&
3250                                 !promote_sync)
3251                         return false;
3252
3253                 /*
3254                  * Sole queue user, no limit
3255                  */
3256                 if (cfqd->busy_queues == 1 || promote_sync)
3257                         max_dispatch = -1;
3258                 else
3259                         /*
3260                          * Normally we start throttling cfqq when cfq_quantum/2
3261                          * requests have been dispatched. But we can drive
3262                          * deeper queue depths at the beginning of slice
3263                          * subjected to upper limit of cfq_quantum.
3264                          * */
3265                         max_dispatch = cfqd->cfq_quantum;
3266         }
3267
3268         /*
3269          * Async queues must wait a bit before being allowed dispatch.
3270          * We also ramp up the dispatch depth gradually for async IO,
3271          * based on the last sync IO we serviced
3272          */
3273         if (!cfq_cfqq_sync(cfqq) && cfqd->cfq_latency) {
3274                 unsigned long last_sync = jiffies - cfqd->last_delayed_sync;
3275                 unsigned int depth;
3276
3277                 depth = last_sync / cfqd->cfq_slice[1];
3278                 if (!depth && !cfqq->dispatched)
3279                         depth = 1;
3280                 if (depth < max_dispatch)
3281                         max_dispatch = depth;
3282         }
3283
3284         /*
3285          * If we're below the current max, allow a dispatch
3286          */
3287         return cfqq->dispatched < max_dispatch;
3288 }
3289
3290 /*
3291  * Dispatch a request from cfqq, moving them to the request queue
3292  * dispatch list.
3293  */
3294 static bool cfq_dispatch_request(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3295 {
3296         struct request *rq;
3297
3298         BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list));
3299
3300         if (!cfq_may_dispatch(cfqd, cfqq))
3301                 return false;
3302
3303         /*
3304          * follow expired path, else get first next available
3305          */
3306         rq = cfq_check_fifo(cfqq);
3307         if (!rq)
3308                 rq = cfqq->next_rq;
3309
3310         /*
3311          * insert request into driver dispatch list
3312          */
3313         cfq_dispatch_insert(cfqd->queue, rq);
3314
3315         if (!cfqd->active_cic) {
3316                 struct cfq_io_cq *cic = RQ_CIC(rq);
3317
3318                 atomic_long_inc(&cic->icq.ioc->refcount);
3319                 cfqd->active_cic = cic;
3320         }
3321
3322         return true;
3323 }
3324
3325 /*
3326  * Find the cfqq that we need to service and move a request from that to the
3327  * dispatch list
3328  */
3329 static int cfq_dispatch_requests(struct request_queue *q, int force)
3330 {
3331         struct cfq_data *cfqd = q->elevator->elevator_data;
3332         struct cfq_queue *cfqq;
3333
3334         if (!cfqd->busy_queues)
3335                 return 0;
3336
3337         if (unlikely(force))
3338                 return cfq_forced_dispatch(cfqd);
3339
3340         cfqq = cfq_select_queue(cfqd);
3341         if (!cfqq)
3342                 return 0;
3343
3344         /*
3345          * Dispatch a request from this cfqq, if it is allowed
3346          */
3347         if (!cfq_dispatch_request(cfqd, cfqq))
3348                 return 0;
3349
3350         cfqq->slice_dispatch++;
3351         cfq_clear_cfqq_must_dispatch(cfqq);
3352
3353         /*
3354          * expire an async queue immediately if it has used up its slice. idle
3355          * queue always expire after 1 dispatch round.
3356          */
3357         if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) &&
3358             cfqq->slice_dispatch >= cfq_prio_to_maxrq(cfqd, cfqq)) ||
3359             cfq_class_idle(cfqq))) {
3360                 cfqq->slice_end = jiffies + 1;
3361                 cfq_slice_expired(cfqd, 0);
3362         }
3363
3364         cfq_log_cfqq(cfqd, cfqq, "dispatched a request");
3365         return 1;
3366 }
3367
3368 /*
3369  * task holds one reference to the queue, dropped when task exits. each rq
3370  * in-flight on this queue also holds a reference, dropped when rq is freed.
3371  *
3372  * Each cfq queue took a reference on the parent group. Drop it now.
3373  * queue lock must be held here.
3374  */
3375 static void cfq_put_queue(struct cfq_queue *cfqq)
3376 {
3377         struct cfq_data *cfqd = cfqq->cfqd;
3378         struct cfq_group *cfqg;
3379
3380         BUG_ON(cfqq->ref <= 0);
3381
3382         cfqq->ref--;
3383         if (cfqq->ref)
3384                 return;
3385
3386         cfq_log_cfqq(cfqd, cfqq, "put_queue");
3387         BUG_ON(rb_first(&cfqq->sort_list));
3388         BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
3389         cfqg = cfqq->cfqg;
3390
3391         if (unlikely(cfqd->active_queue == cfqq)) {
3392                 __cfq_slice_expired(cfqd, cfqq, 0);
3393                 cfq_schedule_dispatch(cfqd);
3394         }
3395
3396         BUG_ON(cfq_cfqq_on_rr(cfqq));
3397         kmem_cache_free(cfq_pool, cfqq);
3398         cfqg_put(cfqg);
3399 }
3400
3401 static void cfq_put_cooperator(struct cfq_queue *cfqq)
3402 {
3403         struct cfq_queue *__cfqq, *next;
3404
3405         /*
3406          * If this queue was scheduled to merge with another queue, be
3407          * sure to drop the reference taken on that queue (and others in
3408          * the merge chain).  See cfq_setup_merge and cfq_merge_cfqqs.
3409          */
3410         __cfqq = cfqq->new_cfqq;
3411         while (__cfqq) {
3412                 if (__cfqq == cfqq) {
3413                         WARN(1, "cfqq->new_cfqq loop detected\n");
3414                         break;
3415                 }
3416                 next = __cfqq->new_cfqq;
3417                 cfq_put_queue(__cfqq);
3418                 __cfqq = next;
3419         }
3420 }
3421
3422 static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3423 {
3424         if (unlikely(cfqq == cfqd->active_queue)) {
3425                 __cfq_slice_expired(cfqd, cfqq, 0);
3426                 cfq_schedule_dispatch(cfqd);
3427         }
3428
3429         cfq_put_cooperator(cfqq);
3430
3431         cfq_put_queue(cfqq);
3432 }
3433
3434 static void cfq_init_icq(struct io_cq *icq)
3435 {
3436         struct cfq_io_cq *cic = icq_to_cic(icq);
3437
3438         cic->ttime.last_end_request = jiffies;
3439 }
3440
3441 static void cfq_exit_icq(struct io_cq *icq)
3442 {
3443         struct cfq_io_cq *cic = icq_to_cic(icq);
3444         struct cfq_data *cfqd = cic_to_cfqd(cic);
3445
3446         if (cic->cfqq[BLK_RW_ASYNC]) {
3447                 cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_ASYNC]);
3448                 cic->cfqq[BLK_RW_ASYNC] = NULL;
3449         }
3450
3451         if (cic->cfqq[BLK_RW_SYNC]) {
3452                 cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_SYNC]);
3453                 cic->cfqq[BLK_RW_SYNC] = NULL;
3454         }
3455 }
3456
3457 static void cfq_init_prio_data(struct cfq_queue *cfqq, struct cfq_io_cq *cic)
3458 {
3459         struct task_struct *tsk = current;
3460         int ioprio_class;
3461
3462         if (!cfq_cfqq_prio_changed(cfqq))
3463                 return;
3464
3465         ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3466         switch (ioprio_class) {
3467         default:
3468                 printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
3469         case IOPRIO_CLASS_NONE:
3470                 /*
3471                  * no prio set, inherit CPU scheduling settings
3472                  */
3473                 cfqq->ioprio = task_nice_ioprio(tsk);
3474                 cfqq->ioprio_class = task_nice_ioclass(tsk);
3475                 break;
3476         case IOPRIO_CLASS_RT:
3477                 cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3478                 cfqq->ioprio_class = IOPRIO_CLASS_RT;
3479                 break;
3480         case IOPRIO_CLASS_BE:
3481                 cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3482                 cfqq->ioprio_class = IOPRIO_CLASS_BE;
3483                 break;
3484         case IOPRIO_CLASS_IDLE:
3485                 cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
3486                 cfqq->ioprio = 7;
3487                 cfq_clear_cfqq_idle_window(cfqq);
3488                 break;
3489         }
3490
3491         /*
3492          * keep track of original prio settings in case we have to temporarily
3493          * elevate the priority of this queue
3494          */
3495         cfqq->org_ioprio = cfqq->ioprio;
3496         cfq_clear_cfqq_prio_changed(cfqq);
3497 }
3498
3499 static void check_ioprio_changed(struct cfq_io_cq *cic, struct bio *bio)
3500 {
3501         int ioprio = cic->icq.ioc->ioprio;
3502         struct cfq_data *cfqd = cic_to_cfqd(cic);
3503         struct cfq_queue *cfqq;
3504
3505         /*
3506          * Check whether ioprio has changed.  The condition may trigger
3507          * spuriously on a newly created cic but there's no harm.
3508          */
3509         if (unlikely(!cfqd) || likely(cic->ioprio == ioprio))
3510                 return;
3511
3512         cfqq = cic->cfqq[BLK_RW_ASYNC];
3513         if (cfqq) {
3514                 struct cfq_queue *new_cfqq;
3515                 new_cfqq = cfq_get_queue(cfqd, BLK_RW_ASYNC, cic, bio,
3516                                          GFP_ATOMIC);
3517                 if (new_cfqq) {
3518                         cic->cfqq[BLK_RW_ASYNC] = new_cfqq;
3519                         cfq_put_queue(cfqq);
3520                 }
3521         }
3522
3523         cfqq = cic->cfqq[BLK_RW_SYNC];
3524         if (cfqq)
3525                 cfq_mark_cfqq_prio_changed(cfqq);
3526
3527         cic->ioprio = ioprio;
3528 }
3529
3530 static void cfq_init_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3531                           pid_t pid, bool is_sync)
3532 {
3533         RB_CLEAR_NODE(&cfqq->rb_node);
3534         RB_CLEAR_NODE(&cfqq->p_node);
3535         INIT_LIST_HEAD(&cfqq->fifo);
3536
3537         cfqq->ref = 0;
3538         cfqq->cfqd = cfqd;
3539
3540         cfq_mark_cfqq_prio_changed(cfqq);
3541
3542         if (is_sync) {
3543                 if (!cfq_class_idle(cfqq))
3544                         cfq_mark_cfqq_idle_window(cfqq);
3545                 cfq_mark_cfqq_sync(cfqq);
3546         }
3547         cfqq->pid = pid;
3548 }
3549
3550 #ifdef CONFIG_CFQ_GROUP_IOSCHED
3551 static void check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio)
3552 {
3553         struct cfq_data *cfqd = cic_to_cfqd(cic);
3554         struct cfq_queue *sync_cfqq;
3555         uint64_t id;
3556
3557         rcu_read_lock();
3558         id = bio_blkcg(bio)->id;
3559         rcu_read_unlock();
3560
3561         /*
3562          * Check whether blkcg has changed.  The condition may trigger
3563          * spuriously on a newly created cic but there's no harm.
3564          */
3565         if (unlikely(!cfqd) || likely(cic->blkcg_id == id))
3566                 return;
3567
3568         sync_cfqq = cic_to_cfqq(cic, 1);
3569         if (sync_cfqq) {
3570                 /*
3571                  * Drop reference to sync queue. A new sync queue will be
3572                  * assigned in new group upon arrival of a fresh request.
3573                  */
3574                 cfq_log_cfqq(cfqd, sync_cfqq, "changed cgroup");
3575                 cic_set_cfqq(cic, NULL, 1);
3576                 cfq_put_queue(sync_cfqq);
3577         }
3578
3579         cic->blkcg_id = id;
3580 }
3581 #else
3582 static inline void check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio) { }
3583 #endif  /* CONFIG_CFQ_GROUP_IOSCHED */
3584
3585 static struct cfq_queue *
3586 cfq_find_alloc_queue(struct cfq_data *cfqd, bool is_sync, struct cfq_io_cq *cic,
3587                      struct bio *bio, gfp_t gfp_mask)
3588 {
3589         struct blkcg *blkcg;
3590         struct cfq_queue *cfqq, *new_cfqq = NULL;
3591         struct cfq_group *cfqg;
3592
3593 retry:
3594         rcu_read_lock();
3595
3596         blkcg = bio_blkcg(bio);
3597         cfqg = cfq_lookup_create_cfqg(cfqd, blkcg);
3598         cfqq = cic_to_cfqq(cic, is_sync);
3599
3600         /*
3601          * Always try a new alloc if we fell back to the OOM cfqq
3602          * originally, since it should just be a temporary situation.
3603          */
3604         if (!cfqq || cfqq == &cfqd->oom_cfqq) {
3605                 cfqq = NULL;
3606                 if (new_cfqq) {
3607                         cfqq = new_cfqq;
3608                         new_cfqq = NULL;
3609                 } else if (gfp_mask & __GFP_WAIT) {
3610                         rcu_read_unlock();
3611                         spin_unlock_irq(cfqd->queue->queue_lock);
3612                         new_cfqq = kmem_cache_alloc_node(cfq_pool,
3613                                         gfp_mask | __GFP_ZERO,
3614                                         cfqd->queue->node);
3615                         spin_lock_irq(cfqd->queue->queue_lock);
3616                         if (new_cfqq)
3617                                 goto retry;
3618                         else
3619                                 return &cfqd->oom_cfqq;
3620                 } else {
3621                         cfqq = kmem_cache_alloc_node(cfq_pool,
3622                                         gfp_mask | __GFP_ZERO,
3623                                         cfqd->queue->node);
3624                 }
3625
3626                 if (cfqq) {
3627                         cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync);
3628                         cfq_init_prio_data(cfqq, cic);
3629                         cfq_link_cfqq_cfqg(cfqq, cfqg);
3630                         cfq_log_cfqq(cfqd, cfqq, "alloced");
3631                 } else
3632                         cfqq = &cfqd->oom_cfqq;
3633         }
3634
3635         if (new_cfqq)
3636                 kmem_cache_free(cfq_pool, new_cfqq);
3637
3638         rcu_read_unlock();
3639         return cfqq;
3640 }
3641
3642 static struct cfq_queue **
3643 cfq_async_queue_prio(struct cfq_data *cfqd, int ioprio_class, int ioprio)
3644 {
3645         switch (ioprio_class) {
3646         case IOPRIO_CLASS_RT:
3647                 return &cfqd->async_cfqq[0][ioprio];
3648         case IOPRIO_CLASS_NONE:
3649                 ioprio = IOPRIO_NORM;
3650                 /* fall through */
3651         case IOPRIO_CLASS_BE:
3652                 return &cfqd->async_cfqq[1][ioprio];
3653         case IOPRIO_CLASS_IDLE:
3654                 return &cfqd->async_idle_cfqq;
3655         default:
3656                 BUG();
3657         }
3658 }
3659
3660 static struct cfq_queue *
3661 cfq_get_queue(struct cfq_data *cfqd, bool is_sync, struct cfq_io_cq *cic,
3662               struct bio *bio, gfp_t gfp_mask)
3663 {
3664         const int ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3665         const int ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3666         struct cfq_queue **async_cfqq = NULL;
3667         struct cfq_queue *cfqq = NULL;
3668
3669         if (!is_sync) {
3670                 async_cfqq = cfq_async_queue_prio(cfqd, ioprio_class, ioprio);
3671                 cfqq = *async_cfqq;
3672         }
3673
3674         if (!cfqq)
3675                 cfqq = cfq_find_alloc_queue(cfqd, is_sync, cic, bio, gfp_mask);
3676
3677         /*
3678          * pin the queue now that it's allocated, scheduler exit will prune it
3679          */
3680         if (!is_sync && !(*async_cfqq)) {
3681                 cfqq->ref++;
3682                 *async_cfqq = cfqq;
3683         }
3684
3685         cfqq->ref++;
3686         return cfqq;
3687 }
3688
3689 static void
3690 __cfq_update_io_thinktime(struct cfq_ttime *ttime, unsigned long slice_idle)
3691 {
3692         unsigned long elapsed = jiffies - ttime->last_end_request;
3693         elapsed = min(elapsed, 2UL * slice_idle);
3694
3695         ttime->ttime_samples = (7*ttime->ttime_samples + 256) / 8;
3696         ttime->ttime_total = (7*ttime->ttime_total + 256*elapsed) / 8;
3697         ttime->ttime_mean = (ttime->ttime_total + 128) / ttime->ttime_samples;
3698 }
3699
3700 static void
3701 cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3702                         struct cfq_io_cq *cic)
3703 {
3704         if (cfq_cfqq_sync(cfqq)) {
3705                 __cfq_update_io_thinktime(&cic->ttime, cfqd->cfq_slice_idle);
3706                 __cfq_update_io_thinktime(&cfqq->service_tree->ttime,
3707                         cfqd->cfq_slice_idle);
3708         }
3709 #ifdef CONFIG_CFQ_GROUP_IOSCHED
3710         __cfq_update_io_thinktime(&cfqq->cfqg->ttime, cfqd->cfq_group_idle);
3711 #endif
3712 }
3713
3714 static void
3715 cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3716                        struct request *rq)
3717 {
3718         sector_t sdist = 0;
3719         sector_t n_sec = blk_rq_sectors(rq);
3720         if (cfqq->last_request_pos) {
3721                 if (cfqq->last_request_pos < blk_rq_pos(rq))
3722                         sdist = blk_rq_pos(rq) - cfqq->last_request_pos;
3723                 else
3724                         sdist = cfqq->last_request_pos - blk_rq_pos(rq);
3725         }
3726
3727         cfqq->seek_history <<= 1;
3728         if (blk_queue_nonrot(cfqd->queue))
3729                 cfqq->seek_history |= (n_sec < CFQQ_SECT_THR_NONROT);
3730         else
3731                 cfqq->seek_history |= (sdist > CFQQ_SEEK_THR);
3732 }
3733
3734 /*
3735  * Disable idle window if the process thinks too long or seeks so much that
3736  * it doesn't matter
3737  */
3738 static void
3739 cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3740                        struct cfq_io_cq *cic)
3741 {
3742         int old_idle, enable_idle;
3743
3744         /*
3745          * Don't idle for async or idle io prio class
3746          */
3747         if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq))
3748                 return;
3749
3750         enable_idle = old_idle = cfq_cfqq_idle_window(cfqq);
3751
3752         if (cfqq->queued[0] + cfqq->queued[1] >= 4)
3753                 cfq_mark_cfqq_deep(cfqq);
3754
3755         if (cfqq->next_rq && (cfqq->next_rq->cmd_flags & REQ_NOIDLE))
3756                 enable_idle = 0;
3757         else if (!atomic_read(&cic->icq.ioc->active_ref) ||
3758                  !cfqd->cfq_slice_idle ||
3759                  (!cfq_cfqq_deep(cfqq) && CFQQ_SEEKY(cfqq)))
3760                 enable_idle = 0;
3761         else if (sample_valid(cic->ttime.ttime_samples)) {
3762                 if (cic->ttime.ttime_mean > cfqd->cfq_slice_idle)
3763                         enable_idle = 0;
3764                 else
3765                         enable_idle = 1;
3766         }
3767
3768         if (old_idle != enable_idle) {
3769                 cfq_log_cfqq(cfqd, cfqq, "idle=%d", enable_idle);
3770                 if (enable_idle)
3771                         cfq_mark_cfqq_idle_window(cfqq);
3772                 else
3773                         cfq_clear_cfqq_idle_window(cfqq);
3774         }
3775 }
3776
3777 /*
3778  * Check if new_cfqq should preempt the currently active queue. Return 0 for
3779  * no or if we aren't sure, a 1 will cause a preempt.
3780  */
3781 static bool
3782 cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
3783                    struct request *rq)
3784 {
3785         struct cfq_queue *cfqq;
3786
3787         cfqq = cfqd->active_queue;
3788         if (!cfqq)
3789                 return false;
3790
3791         if (cfq_class_idle(new_cfqq))
3792                 return false;
3793
3794         if (cfq_class_idle(cfqq))
3795                 return true;
3796
3797         /*
3798          * Don't allow a non-RT request to preempt an ongoing RT cfqq timeslice.
3799          */
3800         if (cfq_class_rt(cfqq) && !cfq_class_rt(new_cfqq))
3801                 return false;
3802
3803         /*
3804          * if the new request is sync, but the currently running queue is
3805          * not, let the sync request have priority.
3806          */
3807         if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq))
3808                 return true;
3809
3810         if (new_cfqq->cfqg != cfqq->cfqg)
3811                 return false;
3812
3813         if (cfq_slice_used(cfqq))
3814                 return true;
3815
3816         /* Allow preemption only if we are idling on sync-noidle tree */
3817         if (cfqd->serving_wl_type == SYNC_NOIDLE_WORKLOAD &&
3818             cfqq_type(new_cfqq) == SYNC_NOIDLE_WORKLOAD &&
3819             new_cfqq->service_tree->count == 2 &&
3820             RB_EMPTY_ROOT(&cfqq->sort_list))
3821                 return true;
3822
3823         /*
3824          * So both queues are sync. Let the new request get disk time if
3825          * it's a metadata request and the current queue is doing regular IO.
3826          */
3827         if ((rq->cmd_flags & REQ_PRIO) && !cfqq->prio_pending)
3828                 return true;
3829
3830         /*
3831          * Allow an RT request to pre-empt an ongoing non-RT cfqq timeslice.
3832          */
3833         if (cfq_class_rt(new_cfqq) && !cfq_class_rt(cfqq))
3834                 return true;
3835
3836         /* An idle queue should not be idle now for some reason */
3837         if (RB_EMPTY_ROOT(&cfqq->sort_list) && !cfq_should_idle(cfqd, cfqq))
3838                 return true;
3839
3840         if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq))
3841                 return false;
3842
3843         /*
3844          * if this request is as-good as one we would expect from the
3845          * current cfqq, let it preempt
3846          */
3847         if (cfq_rq_close(cfqd, cfqq, rq))
3848                 return true;
3849
3850         return false;
3851 }
3852
3853 /*
3854  * cfqq preempts the active queue. if we allowed preempt with no slice left,
3855  * let it have half of its nominal slice.
3856  */
3857 static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3858 {
3859         enum wl_type_t old_type = cfqq_type(cfqd->active_queue);
3860
3861         cfq_log_cfqq(cfqd, cfqq, "preempt");
3862         cfq_slice_expired(cfqd, 1);
3863
3864         /*
3865          * workload type is changed, don't save slice, otherwise preempt
3866          * doesn't happen
3867          */
3868         if (old_type != cfqq_type(cfqq))
3869                 cfqq->cfqg->saved_wl_slice = 0;
3870
3871         /*
3872          * Put the new queue at the front of the of the current list,
3873          * so we know that it will be selected next.
3874          */
3875         BUG_ON(!cfq_cfqq_on_rr(cfqq));
3876
3877         cfq_service_tree_add(cfqd, cfqq, 1);
3878
3879         cfqq->slice_end = 0;
3880         cfq_mark_cfqq_slice_new(cfqq);
3881 }
3882
3883 /*
3884  * Called when a new fs request (rq) is added (to cfqq). Check if there's
3885  * something we should do about it
3886  */
3887 static void
3888 cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3889                 struct request *rq)
3890 {
3891         struct cfq_io_cq *cic = RQ_CIC(rq);
3892
3893         cfqd->rq_queued++;
3894         if (rq->cmd_flags & REQ_PRIO)
3895                 cfqq->prio_pending++;
3896
3897         cfq_update_io_thinktime(cfqd, cfqq, cic);
3898         cfq_update_io_seektime(cfqd, cfqq, rq);
3899         cfq_update_idle_window(cfqd, cfqq, cic);
3900
3901         cfqq->last_request_pos = blk_rq_pos(rq) + blk_rq_sectors(rq);
3902
3903         if (cfqq == cfqd->active_queue) {
3904                 /*
3905                  * Remember that we saw a request from this process, but
3906                  * don't start queuing just yet. Otherwise we risk seeing lots
3907                  * of tiny requests, because we disrupt the normal plugging
3908                  * and merging. If the request is already larger than a single
3909                  * page, let it rip immediately. For that case we assume that
3910                  * merging is already done. Ditto for a busy system that
3911                  * has other work pending, don't risk delaying until the
3912                  * idle timer unplug to continue working.
3913                  */
3914                 if (cfq_cfqq_wait_request(cfqq)) {
3915                         if (blk_rq_bytes(rq) > PAGE_CACHE_SIZE ||
3916                             cfqd->busy_queues > 1) {
3917                                 cfq_del_timer(cfqd, cfqq);
3918                                 cfq_clear_cfqq_wait_request(cfqq);
3919                                 __blk_run_queue(cfqd->queue);
3920                         } else {
3921                                 cfqg_stats_update_idle_time(cfqq->cfqg);
3922                                 cfq_mark_cfqq_must_dispatch(cfqq);
3923                         }
3924                 }
3925         } else if (cfq_should_preempt(cfqd, cfqq, rq)) {
3926                 /*
3927                  * not the active queue - expire current slice if it is
3928                  * idle and has expired it's mean thinktime or this new queue
3929                  * has some old slice time left and is of higher priority or
3930                  * this new queue is RT and the current one is BE
3931                  */
3932                 cfq_preempt_queue(cfqd, cfqq);
3933                 __blk_run_queue(cfqd->queue);
3934         }
3935 }
3936
3937 static void cfq_insert_request(struct request_queue *q, struct request *rq)
3938 {
3939         struct cfq_data *cfqd = q->elevator->elevator_data;
3940         struct cfq_queue *cfqq = RQ_CFQQ(rq);
3941
3942         cfq_log_cfqq(cfqd, cfqq, "insert_request");
3943         cfq_init_prio_data(cfqq, RQ_CIC(rq));
3944
3945         rq_set_fifo_time(rq, jiffies + cfqd->cfq_fifo_expire[rq_is_sync(rq)]);
3946         list_add_tail(&rq->queuelist, &cfqq->fifo);
3947         cfq_add_rq_rb(rq);
3948         cfqg_stats_update_io_add(RQ_CFQG(rq), cfqd->serving_group,
3949                                  rq->cmd_flags);
3950         cfq_rq_enqueued(cfqd, cfqq, rq);
3951 }
3952
3953 /*
3954  * Update hw_tag based on peak queue depth over 50 samples under
3955  * sufficient load.
3956  */
3957 static void cfq_update_hw_tag(struct cfq_data *cfqd)
3958 {
3959         struct cfq_queue *cfqq = cfqd->active_queue;
3960
3961         if (cfqd->rq_in_driver > cfqd->hw_tag_est_depth)
3962                 cfqd->hw_tag_est_depth = cfqd->rq_in_driver;
3963
3964         if (cfqd->hw_tag == 1)
3965                 return;
3966
3967         if (cfqd->rq_queued <= CFQ_HW_QUEUE_MIN &&
3968             cfqd->rq_in_driver <= CFQ_HW_QUEUE_MIN)
3969                 return;
3970
3971         /*
3972          * If active queue hasn't enough requests and can idle, cfq might not
3973          * dispatch sufficient requests to hardware. Don't zero hw_tag in this
3974          * case
3975          */
3976         if (cfqq && cfq_cfqq_idle_window(cfqq) &&
3977             cfqq->dispatched + cfqq->queued[0] + cfqq->queued[1] <
3978             CFQ_HW_QUEUE_MIN && cfqd->rq_in_driver < CFQ_HW_QUEUE_MIN)
3979                 return;
3980
3981         if (cfqd->hw_tag_samples++ < 50)
3982                 return;
3983
3984         if (cfqd->hw_tag_est_depth >= CFQ_HW_QUEUE_MIN)
3985                 cfqd->hw_tag = 1;
3986         else
3987                 cfqd->hw_tag = 0;
3988 }
3989
3990 static bool cfq_should_wait_busy(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3991 {
3992         struct cfq_io_cq *cic = cfqd->active_cic;
3993
3994         /* If the queue already has requests, don't wait */
3995         if (!RB_EMPTY_ROOT(&cfqq->sort_list))
3996                 return false;
3997
3998         /* If there are other queues in the group, don't wait */
3999         if (cfqq->cfqg->nr_cfqq > 1)
4000                 return false;
4001
4002         /* the only queue in the group, but think time is big */
4003         if (cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true))
4004                 return false;
4005
4006         if (cfq_slice_used(cfqq))
4007                 return true;
4008
4009         /* if slice left is less than think time, wait busy */
4010         if (cic && sample_valid(cic->ttime.ttime_samples)
4011             && (cfqq->slice_end - jiffies < cic->ttime.ttime_mean))
4012                 return true;
4013
4014         /*
4015          * If think times is less than a jiffy than ttime_mean=0 and above
4016          * will not be true. It might happen that slice has not expired yet
4017          * but will expire soon (4-5 ns) during select_queue(). To cover the
4018          * case where think time is less than a jiffy, mark the queue wait
4019          * busy if only 1 jiffy is left in the slice.
4020          */
4021         if (cfqq->slice_end - jiffies == 1)
4022                 return true;
4023
4024         return false;
4025 }
4026
4027 static void cfq_completed_request(struct request_queue *q, struct request *rq)
4028 {
4029         struct cfq_queue *cfqq = RQ_CFQQ(rq);
4030         struct cfq_data *cfqd = cfqq->cfqd;
4031         const int sync = rq_is_sync(rq);
4032         unsigned long now;
4033
4034         now = jiffies;
4035         cfq_log_cfqq(cfqd, cfqq, "complete rqnoidle %d",
4036                      !!(rq->cmd_flags & REQ_NOIDLE));
4037
4038         cfq_update_hw_tag(cfqd);
4039
4040         WARN_ON(!cfqd->rq_in_driver);
4041         WARN_ON(!cfqq->dispatched);
4042         cfqd->rq_in_driver--;
4043         cfqq->dispatched--;
4044         (RQ_CFQG(rq))->dispatched--;
4045         cfqg_stats_update_completion(cfqq->cfqg, rq_start_time_ns(rq),
4046                                      rq_io_start_time_ns(rq), rq->cmd_flags);
4047
4048         cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]--;
4049
4050         if (sync) {
4051                 struct cfq_rb_root *st;
4052
4053                 RQ_CIC(rq)->ttime.last_end_request = now;
4054
4055                 if (cfq_cfqq_on_rr(cfqq))
4056                         st = cfqq->service_tree;
4057                 else
4058                         st = st_for(cfqq->cfqg, cfqq_class(cfqq),
4059                                         cfqq_type(cfqq));
4060
4061                 st->ttime.last_end_request = now;
4062                 if (!time_after(rq->start_time + cfqd->cfq_fifo_expire[1], now))
4063                         cfqd->last_delayed_sync = now;
4064         }
4065
4066 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4067         cfqq->cfqg->ttime.last_end_request = now;
4068 #endif
4069
4070         /*
4071          * If this is the active queue, check if it needs to be expired,
4072          * or if we want to idle in case it has no pending requests.
4073          */
4074         if (cfqd->active_queue == cfqq) {
4075                 const bool cfqq_empty = RB_EMPTY_ROOT(&cfqq->sort_list);
4076
4077                 if (cfq_cfqq_slice_new(cfqq)) {
4078                         cfq_set_prio_slice(cfqd, cfqq);
4079                         cfq_clear_cfqq_slice_new(cfqq);
4080                 }
4081
4082                 /*
4083                  * Should we wait for next request to come in before we expire
4084                  * the queue.
4085                  */
4086                 if (cfq_should_wait_busy(cfqd, cfqq)) {
4087                         unsigned long extend_sl = cfqd->cfq_slice_idle;
4088                         if (!cfqd->cfq_slice_idle)
4089                                 extend_sl = cfqd->cfq_group_idle;
4090                         cfqq->slice_end = jiffies + extend_sl;
4091                         cfq_mark_cfqq_wait_busy(cfqq);
4092                         cfq_log_cfqq(cfqd, cfqq, "will busy wait");
4093                 }
4094
4095                 /*
4096                  * Idling is not enabled on:
4097                  * - expired queues
4098                  * - idle-priority queues
4099                  * - async queues
4100                  * - queues with still some requests queued
4101                  * - when there is a close cooperator
4102                  */
4103                 if (cfq_slice_used(cfqq) || cfq_class_idle(cfqq))
4104                         cfq_slice_expired(cfqd, 1);
4105                 else if (sync && cfqq_empty &&
4106                          !cfq_close_cooperator(cfqd, cfqq)) {
4107                         cfq_arm_slice_timer(cfqd);
4108                 }
4109         }
4110
4111         if (!cfqd->rq_in_driver)
4112                 cfq_schedule_dispatch(cfqd);
4113 }
4114
4115 static inline int __cfq_may_queue(struct cfq_queue *cfqq)
4116 {
4117         if (cfq_cfqq_wait_request(cfqq) && !cfq_cfqq_must_alloc_slice(cfqq)) {
4118                 cfq_mark_cfqq_must_alloc_slice(cfqq);
4119                 return ELV_MQUEUE_MUST;
4120         }
4121
4122         return ELV_MQUEUE_MAY;
4123 }
4124
4125 static int cfq_may_queue(struct request_queue *q, int rw)
4126 {
4127         struct cfq_data *cfqd = q->elevator->elevator_data;
4128         struct task_struct *tsk = current;
4129         struct cfq_io_cq *cic;
4130         struct cfq_queue *cfqq;
4131
4132         /*
4133          * don't force setup of a queue from here, as a call to may_queue
4134          * does not necessarily imply that a request actually will be queued.
4135          * so just lookup a possibly existing queue, or return 'may queue'
4136          * if that fails
4137          */
4138         cic = cfq_cic_lookup(cfqd, tsk->io_context);
4139         if (!cic)
4140                 return ELV_MQUEUE_MAY;
4141
4142         cfqq = cic_to_cfqq(cic, rw_is_sync(rw));
4143         if (cfqq) {
4144                 cfq_init_prio_data(cfqq, cic);
4145
4146                 return __cfq_may_queue(cfqq);
4147         }
4148
4149         return ELV_MQUEUE_MAY;
4150 }
4151
4152 /*
4153  * queue lock held here
4154  */
4155 static void cfq_put_request(struct request *rq)
4156 {
4157         struct cfq_queue *cfqq = RQ_CFQQ(rq);
4158
4159         if (cfqq) {
4160                 const int rw = rq_data_dir(rq);
4161
4162                 BUG_ON(!cfqq->allocated[rw]);
4163                 cfqq->allocated[rw]--;
4164
4165                 /* Put down rq reference on cfqg */
4166                 cfqg_put(RQ_CFQG(rq));
4167                 rq->elv.priv[0] = NULL;
4168                 rq->elv.priv[1] = NULL;
4169
4170                 cfq_put_queue(cfqq);
4171         }
4172 }
4173
4174 static struct cfq_queue *
4175 cfq_merge_cfqqs(struct cfq_data *cfqd, struct cfq_io_cq *cic,
4176                 struct cfq_queue *cfqq)
4177 {
4178         cfq_log_cfqq(cfqd, cfqq, "merging with queue %p", cfqq->new_cfqq);
4179         cic_set_cfqq(cic, cfqq->new_cfqq, 1);
4180         cfq_mark_cfqq_coop(cfqq->new_cfqq);
4181         cfq_put_queue(cfqq);
4182         return cic_to_cfqq(cic, 1);
4183 }
4184
4185 /*
4186  * Returns NULL if a new cfqq should be allocated, or the old cfqq if this
4187  * was the last process referring to said cfqq.
4188  */
4189 static struct cfq_queue *
4190 split_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq)
4191 {
4192         if (cfqq_process_refs(cfqq) == 1) {
4193                 cfqq->pid = current->pid;
4194                 cfq_clear_cfqq_coop(cfqq);
4195                 cfq_clear_cfqq_split_coop(cfqq);
4196                 return cfqq;
4197         }
4198
4199         cic_set_cfqq(cic, NULL, 1);
4200
4201         cfq_put_cooperator(cfqq);
4202
4203         cfq_put_queue(cfqq);
4204         return NULL;
4205 }
4206 /*
4207  * Allocate cfq data structures associated with this request.
4208  */
4209 static int
4210 cfq_set_request(struct request_queue *q, struct request *rq, struct bio *bio,
4211                 gfp_t gfp_mask)
4212 {
4213         struct cfq_data *cfqd = q->elevator->elevator_data;
4214         struct cfq_io_cq *cic = icq_to_cic(rq->elv.icq);
4215         const int rw = rq_data_dir(rq);
4216         const bool is_sync = rq_is_sync(rq);
4217         struct cfq_queue *cfqq;
4218
4219         might_sleep_if(gfp_mask & __GFP_WAIT);
4220
4221         spin_lock_irq(q->queue_lock);
4222
4223         check_ioprio_changed(cic, bio);
4224         check_blkcg_changed(cic, bio);
4225 new_queue:
4226         cfqq = cic_to_cfqq(cic, is_sync);
4227         if (!cfqq || cfqq == &cfqd->oom_cfqq) {
4228                 cfqq = cfq_get_queue(cfqd, is_sync, cic, bio, gfp_mask);
4229                 cic_set_cfqq(cic, cfqq, is_sync);
4230         } else {
4231                 /*
4232                  * If the queue was seeky for too long, break it apart.
4233                  */
4234                 if (cfq_cfqq_coop(cfqq) && cfq_cfqq_split_coop(cfqq)) {
4235                         cfq_log_cfqq(cfqd, cfqq, "breaking apart cfqq");
4236                         cfqq = split_cfqq(cic, cfqq);
4237                         if (!cfqq)
4238                                 goto new_queue;
4239                 }
4240
4241                 /*
4242                  * Check to see if this queue is scheduled to merge with
4243                  * another, closely cooperating queue.  The merging of
4244                  * queues happens here as it must be done in process context.
4245                  * The reference on new_cfqq was taken in merge_cfqqs.
4246                  */
4247                 if (cfqq->new_cfqq)
4248                         cfqq = cfq_merge_cfqqs(cfqd, cic, cfqq);
4249         }
4250
4251         cfqq->allocated[rw]++;
4252
4253         cfqq->ref++;
4254         cfqg_get(cfqq->cfqg);
4255         rq->elv.priv[0] = cfqq;
4256         rq->elv.priv[1] = cfqq->cfqg;
4257         spin_unlock_irq(q->queue_lock);
4258         return 0;
4259 }
4260
4261 static void cfq_kick_queue(struct work_struct *work)
4262 {
4263         struct cfq_data *cfqd =
4264                 container_of(work, struct cfq_data, unplug_work);
4265         struct request_queue *q = cfqd->queue;
4266
4267         spin_lock_irq(q->queue_lock);
4268         __blk_run_queue(cfqd->queue);
4269         spin_unlock_irq(q->queue_lock);
4270 }
4271
4272 /*
4273  * Timer running if the active_queue is currently idling inside its time slice
4274  */
4275 static void cfq_idle_slice_timer(unsigned long data)
4276 {
4277         struct cfq_data *cfqd = (struct cfq_data *) data;
4278         struct cfq_queue *cfqq;
4279         unsigned long flags;
4280         int timed_out = 1;
4281
4282         cfq_log(cfqd, "idle timer fired");
4283
4284         spin_lock_irqsave(cfqd->queue->queue_lock, flags);
4285
4286         cfqq = cfqd->active_queue;
4287         if (cfqq) {
4288                 timed_out = 0;
4289
4290                 /*
4291                  * We saw a request before the queue expired, let it through
4292                  */
4293                 if (cfq_cfqq_must_dispatch(cfqq))
4294                         goto out_kick;
4295
4296                 /*
4297                  * expired
4298                  */
4299                 if (cfq_slice_used(cfqq))
4300                         goto expire;
4301
4302                 /*
4303                  * only expire and reinvoke request handler, if there are
4304                  * other queues with pending requests
4305                  */
4306                 if (!cfqd->busy_queues)
4307                         goto out_cont;
4308
4309                 /*
4310                  * not expired and it has a request pending, let it dispatch
4311                  */
4312                 if (!RB_EMPTY_ROOT(&cfqq->sort_list))
4313                         goto out_kick;
4314
4315                 /*
4316                  * Queue depth flag is reset only when the idle didn't succeed
4317                  */
4318                 cfq_clear_cfqq_deep(cfqq);
4319         }
4320 expire:
4321         cfq_slice_expired(cfqd, timed_out);
4322 out_kick:
4323         cfq_schedule_dispatch(cfqd);
4324 out_cont:
4325         spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
4326 }
4327
4328 static void cfq_shutdown_timer_wq(struct cfq_data *cfqd)
4329 {
4330         del_timer_sync(&cfqd->idle_slice_timer);
4331         cancel_work_sync(&cfqd->unplug_work);
4332 }
4333
4334 static void cfq_put_async_queues(struct cfq_data *cfqd)
4335 {
4336         int i;
4337
4338         for (i = 0; i < IOPRIO_BE_NR; i++) {
4339                 if (cfqd->async_cfqq[0][i])
4340                         cfq_put_queue(cfqd->async_cfqq[0][i]);
4341                 if (cfqd->async_cfqq[1][i])
4342                         cfq_put_queue(cfqd->async_cfqq[1][i]);
4343         }
4344
4345         if (cfqd->async_idle_cfqq)
4346                 cfq_put_queue(cfqd->async_idle_cfqq);
4347 }
4348
4349 static void cfq_exit_queue(struct elevator_queue *e)
4350 {
4351         struct cfq_data *cfqd = e->elevator_data;
4352         struct request_queue *q = cfqd->queue;
4353
4354         cfq_shutdown_timer_wq(cfqd);
4355
4356         spin_lock_irq(q->queue_lock);
4357
4358         if (cfqd->active_queue)
4359                 __cfq_slice_expired(cfqd, cfqd->active_queue, 0);
4360
4361         cfq_put_async_queues(cfqd);
4362
4363         spin_unlock_irq(q->queue_lock);
4364
4365         cfq_shutdown_timer_wq(cfqd);
4366
4367 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4368         blkcg_deactivate_policy(q, &blkcg_policy_cfq);
4369 #else
4370         kfree(cfqd->root_group);
4371 #endif
4372         kfree(cfqd);
4373 }
4374
4375 static int cfq_init_queue(struct request_queue *q, struct elevator_type *e)
4376 {
4377         struct cfq_data *cfqd;
4378         struct blkcg_gq *blkg __maybe_unused;
4379         int i, ret;
4380         struct elevator_queue *eq;
4381
4382         eq = elevator_alloc(q, e);
4383         if (!eq)
4384                 return -ENOMEM;
4385
4386         cfqd = kzalloc_node(sizeof(*cfqd), GFP_KERNEL, q->node);
4387         if (!cfqd) {
4388                 kobject_put(&eq->kobj);
4389                 return -ENOMEM;
4390         }
4391         eq->elevator_data = cfqd;
4392
4393         cfqd->queue = q;
4394         spin_lock_irq(q->queue_lock);
4395         q->elevator = eq;
4396         spin_unlock_irq(q->queue_lock);
4397
4398         /* Init root service tree */
4399         cfqd->grp_service_tree = CFQ_RB_ROOT;
4400
4401         /* Init root group and prefer root group over other groups by default */
4402 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4403         ret = blkcg_activate_policy(q, &blkcg_policy_cfq);
4404         if (ret)
4405                 goto out_free;
4406
4407         cfqd->root_group = blkg_to_cfqg(q->root_blkg);
4408 #else
4409         ret = -ENOMEM;
4410         cfqd->root_group = kzalloc_node(sizeof(*cfqd->root_group),
4411                                         GFP_KERNEL, cfqd->queue->node);
4412         if (!cfqd->root_group)
4413                 goto out_free;
4414
4415         cfq_init_cfqg_base(cfqd->root_group);
4416 #endif
4417         cfqd->root_group->weight = 2 * CFQ_WEIGHT_DEFAULT;
4418         cfqd->root_group->leaf_weight = 2 * CFQ_WEIGHT_DEFAULT;
4419
4420         /*
4421          * Not strictly needed (since RB_ROOT just clears the node and we
4422          * zeroed cfqd on alloc), but better be safe in case someone decides
4423          * to add magic to the rb code
4424          */
4425         for (i = 0; i < CFQ_PRIO_LISTS; i++)
4426                 cfqd->prio_trees[i] = RB_ROOT;
4427
4428         /*
4429          * Our fallback cfqq if cfq_find_alloc_queue() runs into OOM issues.
4430          * Grab a permanent reference to it, so that the normal code flow
4431          * will not attempt to free it.  oom_cfqq is linked to root_group
4432          * but shouldn't hold a reference as it'll never be unlinked.  Lose
4433          * the reference from linking right away.
4434          */
4435         cfq_init_cfqq(cfqd, &cfqd->oom_cfqq, 1, 0);
4436         cfqd->oom_cfqq.ref++;
4437
4438         spin_lock_irq(q->queue_lock);
4439         cfq_link_cfqq_cfqg(&cfqd->oom_cfqq, cfqd->root_group);
4440         cfqg_put(cfqd->root_group);
4441         spin_unlock_irq(q->queue_lock);
4442
4443         init_timer(&cfqd->idle_slice_timer);
4444         cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
4445         cfqd->idle_slice_timer.data = (unsigned long) cfqd;
4446
4447         INIT_WORK(&cfqd->unplug_work, cfq_kick_queue);
4448
4449         cfqd->cfq_quantum = cfq_quantum;
4450         cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0];
4451         cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1];
4452         cfqd->cfq_back_max = cfq_back_max;
4453         cfqd->cfq_back_penalty = cfq_back_penalty;
4454         cfqd->cfq_slice[0] = cfq_slice_async;
4455         cfqd->cfq_slice[1] = cfq_slice_sync;
4456         cfqd->cfq_target_latency = cfq_target_latency;
4457         cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
4458         cfqd->cfq_slice_idle = cfq_slice_idle;
4459         cfqd->cfq_group_idle = cfq_group_idle;
4460         cfqd->cfq_latency = 1;
4461         cfqd->hw_tag = -1;
4462         /*
4463          * we optimistically start assuming sync ops weren't delayed in last
4464          * second, in order to have larger depth for async operations.
4465          */
4466         cfqd->last_delayed_sync = jiffies - HZ;
4467         return 0;
4468
4469 out_free:
4470         kfree(cfqd);
4471         kobject_put(&eq->kobj);
4472         return ret;
4473 }
4474
4475 /*
4476  * sysfs parts below -->
4477  */
4478 static ssize_t
4479 cfq_var_show(unsigned int var, char *page)
4480 {
4481         return sprintf(page, "%d\n", var);
4482 }
4483
4484 static ssize_t
4485 cfq_var_store(unsigned int *var, const char *page, size_t count)
4486 {
4487         char *p = (char *) page;
4488
4489         *var = simple_strtoul(p, &p, 10);
4490         return count;
4491 }
4492
4493 #define SHOW_FUNCTION(__FUNC, __VAR, __CONV)                            \
4494 static ssize_t __FUNC(struct elevator_queue *e, char *page)             \
4495 {                                                                       \
4496         struct cfq_data *cfqd = e->elevator_data;                       \
4497         unsigned int __data = __VAR;                                    \
4498         if (__CONV)                                                     \
4499                 __data = jiffies_to_msecs(__data);                      \
4500         return cfq_var_show(__data, (page));                            \
4501 }
4502 SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0);
4503 SHOW_FUNCTION(cfq_fifo_expire_sync_show, cfqd->cfq_fifo_expire[1], 1);
4504 SHOW_FUNCTION(cfq_fifo_expire_async_show, cfqd->cfq_fifo_expire[0], 1);
4505 SHOW_FUNCTION(cfq_back_seek_max_show, cfqd->cfq_back_max, 0);
4506 SHOW_FUNCTION(cfq_back_seek_penalty_show, cfqd->cfq_back_penalty, 0);
4507 SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1);
4508 SHOW_FUNCTION(cfq_group_idle_show, cfqd->cfq_group_idle, 1);
4509 SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1);
4510 SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
4511 SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
4512 SHOW_FUNCTION(cfq_low_latency_show, cfqd->cfq_latency, 0);
4513 SHOW_FUNCTION(cfq_target_latency_show, cfqd->cfq_target_latency, 1);
4514 #undef SHOW_FUNCTION
4515
4516 #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV)                 \
4517 static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count) \
4518 {                                                                       \
4519         struct cfq_data *cfqd = e->elevator_data;                       \
4520         unsigned int __data;                                            \
4521         int ret = cfq_var_store(&__data, (page), count);                \
4522         if (__data < (MIN))                                             \
4523                 __data = (MIN);                                         \
4524         else if (__data > (MAX))                                        \
4525                 __data = (MAX);                                         \
4526         if (__CONV)                                                     \
4527                 *(__PTR) = msecs_to_jiffies(__data);                    \
4528         else                                                            \
4529                 *(__PTR) = __data;                                      \
4530         return ret;                                                     \
4531 }
4532 STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
4533 STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1,
4534                 UINT_MAX, 1);
4535 STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1,
4536                 UINT_MAX, 1);
4537 STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
4538 STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1,
4539                 UINT_MAX, 0);
4540 STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1);
4541 STORE_FUNCTION(cfq_group_idle_store, &cfqd->cfq_group_idle, 0, UINT_MAX, 1);
4542 STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1);
4543 STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1);
4544 STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1,
4545                 UINT_MAX, 0);
4546 STORE_FUNCTION(cfq_low_latency_store, &cfqd->cfq_latency, 0, 1, 0);
4547 STORE_FUNCTION(cfq_target_latency_store, &cfqd->cfq_target_latency, 1, UINT_MAX, 1);
4548 #undef STORE_FUNCTION
4549
4550 #define CFQ_ATTR(name) \
4551         __ATTR(name, S_IRUGO|S_IWUSR, cfq_##name##_show, cfq_##name##_store)
4552
4553 static struct elv_fs_entry cfq_attrs[] = {
4554         CFQ_ATTR(quantum),
4555         CFQ_ATTR(fifo_expire_sync),
4556         CFQ_ATTR(fifo_expire_async),
4557         CFQ_ATTR(back_seek_max),
4558         CFQ_ATTR(back_seek_penalty),
4559         CFQ_ATTR(slice_sync),
4560         CFQ_ATTR(slice_async),
4561         CFQ_ATTR(slice_async_rq),
4562         CFQ_ATTR(slice_idle),
4563         CFQ_ATTR(group_idle),
4564         CFQ_ATTR(low_latency),
4565         CFQ_ATTR(target_latency),
4566         __ATTR_NULL
4567 };
4568
4569 static struct elevator_type iosched_cfq = {
4570         .ops = {
4571                 .elevator_merge_fn =            cfq_merge,
4572                 .elevator_merged_fn =           cfq_merged_request,
4573                 .elevator_merge_req_fn =        cfq_merged_requests,
4574                 .elevator_allow_merge_fn =      cfq_allow_merge,
4575                 .elevator_bio_merged_fn =       cfq_bio_merged,
4576                 .elevator_dispatch_fn =         cfq_dispatch_requests,
4577                 .elevator_add_req_fn =          cfq_insert_request,
4578                 .elevator_activate_req_fn =     cfq_activate_request,
4579                 .elevator_deactivate_req_fn =   cfq_deactivate_request,
4580                 .elevator_completed_req_fn =    cfq_completed_request,
4581                 .elevator_former_req_fn =       elv_rb_former_request,
4582                 .elevator_latter_req_fn =       elv_rb_latter_request,
4583                 .elevator_init_icq_fn =         cfq_init_icq,
4584                 .elevator_exit_icq_fn =         cfq_exit_icq,
4585                 .elevator_set_req_fn =          cfq_set_request,
4586                 .elevator_put_req_fn =          cfq_put_request,
4587                 .elevator_may_queue_fn =        cfq_may_queue,
4588                 .elevator_init_fn =             cfq_init_queue,
4589                 .elevator_exit_fn =             cfq_exit_queue,
4590         },
4591         .icq_size       =       sizeof(struct cfq_io_cq),
4592         .icq_align      =       __alignof__(struct cfq_io_cq),
4593         .elevator_attrs =       cfq_attrs,
4594         .elevator_name  =       "cfq",
4595         .elevator_owner =       THIS_MODULE,
4596 };
4597
4598 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4599 static struct blkcg_policy blkcg_policy_cfq = {
4600         .pd_size                = sizeof(struct cfq_group),
4601         .cftypes                = cfq_blkcg_files,
4602
4603         .pd_init_fn             = cfq_pd_init,
4604         .pd_offline_fn          = cfq_pd_offline,
4605         .pd_reset_stats_fn      = cfq_pd_reset_stats,
4606 };
4607 #endif
4608
4609 static int __init cfq_init(void)
4610 {
4611         int ret;
4612
4613         /*
4614          * could be 0 on HZ < 1000 setups
4615          */
4616         if (!cfq_slice_async)
4617                 cfq_slice_async = 1;
4618         if (!cfq_slice_idle)
4619                 cfq_slice_idle = 1;
4620
4621 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4622         if (!cfq_group_idle)
4623                 cfq_group_idle = 1;
4624
4625         ret = blkcg_policy_register(&blkcg_policy_cfq);
4626         if (ret)
4627                 return ret;
4628 #else
4629         cfq_group_idle = 0;
4630 #endif
4631
4632         ret = -ENOMEM;
4633         cfq_pool = KMEM_CACHE(cfq_queue, 0);
4634         if (!cfq_pool)
4635                 goto err_pol_unreg;
4636
4637         ret = elv_register(&iosched_cfq);
4638         if (ret)
4639                 goto err_free_pool;
4640
4641         return 0;
4642
4643 err_free_pool:
4644         kmem_cache_destroy(cfq_pool);
4645 err_pol_unreg:
4646 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4647         blkcg_policy_unregister(&blkcg_policy_cfq);
4648 #endif
4649         return ret;
4650 }
4651
4652 static void __exit cfq_exit(void)
4653 {
4654 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4655         blkcg_policy_unregister(&blkcg_policy_cfq);
4656 #endif
4657         elv_unregister(&iosched_cfq);
4658         kmem_cache_destroy(cfq_pool);
4659 }
4660
4661 module_init(cfq_init);
4662 module_exit(cfq_exit);
4663
4664 MODULE_AUTHOR("Jens Axboe");
4665 MODULE_LICENSE("GPL");
4666 MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler");