Pileus Git - ~andy/linux/blob - 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 seq_file *sf, void *v)
1636 {
1637         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1638                           cfqg_prfill_weight_device, &blkcg_policy_cfq,
1639                           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 seq_file *sf, void *v)
1654 {
1655         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1656                           cfqg_prfill_leaf_weight_device, &blkcg_policy_cfq,
1657                           0, false);
1658         return 0;
1659 }
1660
1661 static int cfq_print_weight(struct seq_file *sf, void *v)
1662 {
1663         seq_printf(sf, "%u\n", css_to_blkcg(seq_css(sf))->cfq_weight);
1664         return 0;
1665 }
1666
1667 static int cfq_print_leaf_weight(struct seq_file *sf, void *v)
1668 {
1669         seq_printf(sf, "%u\n", css_to_blkcg(seq_css(sf))->cfq_leaf_weight);
1670         return 0;
1671 }
1672
1673 static int __cfqg_set_weight_device(struct cgroup_subsys_state *css,
1674                                     struct cftype *cft, const char *buf,
1675                                     bool is_leaf_weight)
1676 {
1677         struct blkcg *blkcg = css_to_blkcg(css);
1678         struct blkg_conf_ctx ctx;
1679         struct cfq_group *cfqg;
1680         int ret;
1681
1682         ret = blkg_conf_prep(blkcg, &blkcg_policy_cfq, buf, &ctx);
1683         if (ret)
1684                 return ret;
1685
1686         ret = -EINVAL;
1687         cfqg = blkg_to_cfqg(ctx.blkg);
1688         if (!ctx.v || (ctx.v >= CFQ_WEIGHT_MIN && ctx.v <= CFQ_WEIGHT_MAX)) {
1689                 if (!is_leaf_weight) {
1690                         cfqg->dev_weight = ctx.v;
1691                         cfqg->new_weight = ctx.v ?: blkcg->cfq_weight;
1692                 } else {
1693                         cfqg->dev_leaf_weight = ctx.v;
1694                         cfqg->new_leaf_weight = ctx.v ?: blkcg->cfq_leaf_weight;
1695                 }
1696                 ret = 0;
1697         }
1698
1699         blkg_conf_finish(&ctx);
1700         return ret;
1701 }
1702
1703 static int cfqg_set_weight_device(struct cgroup_subsys_state *css,
1704                                   struct cftype *cft, const char *buf)
1705 {
1706         return __cfqg_set_weight_device(css, cft, buf, false);
1707 }
1708
1709 static int cfqg_set_leaf_weight_device(struct cgroup_subsys_state *css,
1710                                        struct cftype *cft, const char *buf)
1711 {
1712         return __cfqg_set_weight_device(css, cft, buf, true);
1713 }
1714
1715 static int __cfq_set_weight(struct cgroup_subsys_state *css, struct cftype *cft,
1716                             u64 val, bool is_leaf_weight)
1717 {
1718         struct blkcg *blkcg = css_to_blkcg(css);
1719         struct blkcg_gq *blkg;
1720
1721         if (val < CFQ_WEIGHT_MIN || val > CFQ_WEIGHT_MAX)
1722                 return -EINVAL;
1723
1724         spin_lock_irq(&blkcg->lock);
1725
1726         if (!is_leaf_weight)
1727                 blkcg->cfq_weight = val;
1728         else
1729                 blkcg->cfq_leaf_weight = val;
1730
1731         hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
1732                 struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1733
1734                 if (!cfqg)
1735                         continue;
1736
1737                 if (!is_leaf_weight) {
1738                         if (!cfqg->dev_weight)
1739                                 cfqg->new_weight = blkcg->cfq_weight;
1740                 } else {
1741                         if (!cfqg->dev_leaf_weight)
1742                                 cfqg->new_leaf_weight = blkcg->cfq_leaf_weight;
1743                 }
1744         }
1745
1746         spin_unlock_irq(&blkcg->lock);
1747         return 0;
1748 }
1749
1750 static int cfq_set_weight(struct cgroup_subsys_state *css, struct cftype *cft,
1751                           u64 val)
1752 {
1753         return __cfq_set_weight(css, cft, val, false);
1754 }
1755
1756 static int cfq_set_leaf_weight(struct cgroup_subsys_state *css,
1757                                struct cftype *cft, u64 val)
1758 {
1759         return __cfq_set_weight(css, cft, val, true);
1760 }
1761
1762 static int cfqg_print_stat(struct seq_file *sf, void *v)
1763 {
1764         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), blkg_prfill_stat,
1765                           &blkcg_policy_cfq, seq_cft(sf)->private, false);
1766         return 0;
1767 }
1768
1769 static int cfqg_print_rwstat(struct seq_file *sf, void *v)
1770 {
1771         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), blkg_prfill_rwstat,
1772                           &blkcg_policy_cfq, seq_cft(sf)->private, true);
1773         return 0;
1774 }
1775
1776 static u64 cfqg_prfill_stat_recursive(struct seq_file *sf,
1777                                       struct blkg_policy_data *pd, int off)
1778 {
1779         u64 sum = cfqg_stat_pd_recursive_sum(pd, off);
1780
1781         return __blkg_prfill_u64(sf, pd, sum);
1782 }
1783
1784 static u64 cfqg_prfill_rwstat_recursive(struct seq_file *sf,
1785                                         struct blkg_policy_data *pd, int off)
1786 {
1787         struct blkg_rwstat sum = cfqg_rwstat_pd_recursive_sum(pd, off);
1788
1789         return __blkg_prfill_rwstat(sf, pd, &sum);
1790 }
1791
1792 static int cfqg_print_stat_recursive(struct seq_file *sf, void *v)
1793 {
1794         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1795                           cfqg_prfill_stat_recursive, &blkcg_policy_cfq,
1796                           seq_cft(sf)->private, false);
1797         return 0;
1798 }
1799
1800 static int cfqg_print_rwstat_recursive(struct seq_file *sf, void *v)
1801 {
1802         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1803                           cfqg_prfill_rwstat_recursive, &blkcg_policy_cfq,
1804                           seq_cft(sf)->private, true);
1805         return 0;
1806 }
1807
1808 #ifdef CONFIG_DEBUG_BLK_CGROUP
1809 static u64 cfqg_prfill_avg_queue_size(struct seq_file *sf,
1810                                       struct blkg_policy_data *pd, int off)
1811 {
1812         struct cfq_group *cfqg = pd_to_cfqg(pd);
1813         u64 samples = blkg_stat_read(&cfqg->stats.avg_queue_size_samples);
1814         u64 v = 0;
1815
1816         if (samples) {
1817                 v = blkg_stat_read(&cfqg->stats.avg_queue_size_sum);
1818                 v = div64_u64(v, samples);
1819         }
1820         __blkg_prfill_u64(sf, pd, v);
1821         return 0;
1822 }
1823
1824 /* print avg_queue_size */
1825 static int cfqg_print_avg_queue_size(struct seq_file *sf, void *v)
1826 {
1827         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1828                           cfqg_prfill_avg_queue_size, &blkcg_policy_cfq,
1829                           0, false);
1830         return 0;
1831 }
1832 #endif  /* CONFIG_DEBUG_BLK_CGROUP */
1833
1834 static struct cftype cfq_blkcg_files[] = {
1835         /* on root, weight is mapped to leaf_weight */
1836         {
1837                 .name = "weight_device",
1838                 .flags = CFTYPE_ONLY_ON_ROOT,
1839                 .seq_show = cfqg_print_leaf_weight_device,
1840                 .write_string = cfqg_set_leaf_weight_device,
1841                 .max_write_len = 256,
1842         },
1843         {
1844                 .name = "weight",
1845                 .flags = CFTYPE_ONLY_ON_ROOT,
1846                 .seq_show = cfq_print_leaf_weight,
1847                 .write_u64 = cfq_set_leaf_weight,
1848         },
1849
1850         /* no such mapping necessary for !roots */
1851         {
1852                 .name = "weight_device",
1853                 .flags = CFTYPE_NOT_ON_ROOT,
1854                 .seq_show = cfqg_print_weight_device,
1855                 .write_string = cfqg_set_weight_device,
1856                 .max_write_len = 256,
1857         },
1858         {
1859                 .name = "weight",
1860                 .flags = CFTYPE_NOT_ON_ROOT,
1861                 .seq_show = cfq_print_weight,
1862                 .write_u64 = cfq_set_weight,
1863         },
1864
1865         {
1866                 .name = "leaf_weight_device",
1867                 .seq_show = cfqg_print_leaf_weight_device,
1868                 .write_string = cfqg_set_leaf_weight_device,
1869                 .max_write_len = 256,
1870         },
1871         {
1872                 .name = "leaf_weight",
1873                 .seq_show = cfq_print_leaf_weight,
1874                 .write_u64 = cfq_set_leaf_weight,
1875         },
1876
1877         /* statistics, covers only the tasks in the cfqg */
1878         {
1879                 .name = "time",
1880                 .private = offsetof(struct cfq_group, stats.time),
1881                 .seq_show = cfqg_print_stat,
1882         },
1883         {
1884                 .name = "sectors",
1885                 .private = offsetof(struct cfq_group, stats.sectors),
1886                 .seq_show = cfqg_print_stat,
1887         },
1888         {
1889                 .name = "io_service_bytes",
1890                 .private = offsetof(struct cfq_group, stats.service_bytes),
1891                 .seq_show = cfqg_print_rwstat,
1892         },
1893         {
1894                 .name = "io_serviced",
1895                 .private = offsetof(struct cfq_group, stats.serviced),
1896                 .seq_show = cfqg_print_rwstat,
1897         },
1898         {
1899                 .name = "io_service_time",
1900                 .private = offsetof(struct cfq_group, stats.service_time),
1901                 .seq_show = cfqg_print_rwstat,
1902         },
1903         {
1904                 .name = "io_wait_time",
1905                 .private = offsetof(struct cfq_group, stats.wait_time),
1906                 .seq_show = cfqg_print_rwstat,
1907         },
1908         {
1909                 .name = "io_merged",
1910                 .private = offsetof(struct cfq_group, stats.merged),
1911                 .seq_show = cfqg_print_rwstat,
1912         },
1913         {
1914                 .name = "io_queued",
1915                 .private = offsetof(struct cfq_group, stats.queued),
1916                 .seq_show = cfqg_print_rwstat,
1917         },
1918
1919         /* the same statictics which cover the cfqg and its descendants */
1920         {
1921                 .name = "time_recursive",
1922                 .private = offsetof(struct cfq_group, stats.time),
1923                 .seq_show = cfqg_print_stat_recursive,
1924         },
1925         {
1926                 .name = "sectors_recursive",
1927                 .private = offsetof(struct cfq_group, stats.sectors),
1928                 .seq_show = cfqg_print_stat_recursive,
1929         },
1930         {
1931                 .name = "io_service_bytes_recursive",
1932                 .private = offsetof(struct cfq_group, stats.service_bytes),
1933                 .seq_show = cfqg_print_rwstat_recursive,
1934         },
1935         {
1936                 .name = "io_serviced_recursive",
1937                 .private = offsetof(struct cfq_group, stats.serviced),
1938                 .seq_show = cfqg_print_rwstat_recursive,
1939         },
1940         {
1941                 .name = "io_service_time_recursive",
1942                 .private = offsetof(struct cfq_group, stats.service_time),
1943                 .seq_show = cfqg_print_rwstat_recursive,
1944         },
1945         {
1946                 .name = "io_wait_time_recursive",
1947                 .private = offsetof(struct cfq_group, stats.wait_time),
1948                 .seq_show = cfqg_print_rwstat_recursive,
1949         },
1950         {
1951                 .name = "io_merged_recursive",
1952                 .private = offsetof(struct cfq_group, stats.merged),
1953                 .seq_show = cfqg_print_rwstat_recursive,
1954         },
1955         {
1956                 .name = "io_queued_recursive",
1957                 .private = offsetof(struct cfq_group, stats.queued),
1958                 .seq_show = cfqg_print_rwstat_recursive,
1959         },
1960 #ifdef CONFIG_DEBUG_BLK_CGROUP
1961         {
1962                 .name = "avg_queue_size",
1963                 .seq_show = cfqg_print_avg_queue_size,
1964         },
1965         {
1966                 .name = "group_wait_time",
1967                 .private = offsetof(struct cfq_group, stats.group_wait_time),
1968                 .seq_show = cfqg_print_stat,
1969         },
1970         {
1971                 .name = "idle_time",
1972                 .private = offsetof(struct cfq_group, stats.idle_time),
1973                 .seq_show = cfqg_print_stat,
1974         },
1975         {
1976                 .name = "empty_time",
1977                 .private = offsetof(struct cfq_group, stats.empty_time),
1978                 .seq_show = cfqg_print_stat,
1979         },
1980         {
1981                 .name = "dequeue",
1982                 .private = offsetof(struct cfq_group, stats.dequeue),
1983                 .seq_show = cfqg_print_stat,
1984         },
1985         {
1986                 .name = "unaccounted_time",
1987                 .private = offsetof(struct cfq_group, stats.unaccounted_time),
1988                 .seq_show = cfqg_print_stat,
1989         },
1990 #endif  /* CONFIG_DEBUG_BLK_CGROUP */
1991         { }     /* terminate */
1992 };
1993 #else /* GROUP_IOSCHED */
1994 static struct cfq_group *cfq_lookup_create_cfqg(struct cfq_data *cfqd,
1995                                                 struct blkcg *blkcg)
1996 {
1997         return cfqd->root_group;
1998 }
1999
2000 static inline void
2001 cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg) {
2002         cfqq->cfqg = cfqg;
2003 }
2004
2005 #endif /* GROUP_IOSCHED */
2006
2007 /*
2008  * The cfqd->service_trees holds all pending cfq_queue's that have
2009  * requests waiting to be processed. It is sorted in the order that
2010  * we will service the queues.
2011  */
2012 static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2013                                  bool add_front)
2014 {
2015         struct rb_node **p, *parent;
2016         struct cfq_queue *__cfqq;
2017         unsigned long rb_key;
2018         struct cfq_rb_root *st;
2019         int left;
2020         int new_cfqq = 1;
2021
2022         st = st_for(cfqq->cfqg, cfqq_class(cfqq), cfqq_type(cfqq));
2023         if (cfq_class_idle(cfqq)) {
2024                 rb_key = CFQ_IDLE_DELAY;
2025                 parent = rb_last(&st->rb);
2026                 if (parent && parent != &cfqq->rb_node) {
2027                         __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2028                         rb_key += __cfqq->rb_key;
2029                 } else
2030                         rb_key += jiffies;
2031         } else if (!add_front) {
2032                 /*
2033                  * Get our rb key offset. Subtract any residual slice
2034                  * value carried from last service. A negative resid
2035                  * count indicates slice overrun, and this should position
2036                  * the next service time further away in the tree.
2037                  */
2038                 rb_key = cfq_slice_offset(cfqd, cfqq) + jiffies;
2039                 rb_key -= cfqq->slice_resid;
2040                 cfqq->slice_resid = 0;
2041         } else {
2042                 rb_key = -HZ;
2043                 __cfqq = cfq_rb_first(st);
2044                 rb_key += __cfqq ? __cfqq->rb_key : jiffies;
2045         }
2046
2047         if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2048                 new_cfqq = 0;
2049                 /*
2050                  * same position, nothing more to do
2051                  */
2052                 if (rb_key == cfqq->rb_key && cfqq->service_tree == st)
2053                         return;
2054
2055                 cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2056                 cfqq->service_tree = NULL;
2057         }
2058
2059         left = 1;
2060         parent = NULL;
2061         cfqq->service_tree = st;
2062         p = &st->rb.rb_node;
2063         while (*p) {
2064                 parent = *p;
2065                 __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2066
2067                 /*
2068                  * sort by key, that represents service time.
2069                  */
2070                 if (time_before(rb_key, __cfqq->rb_key))
2071                         p = &parent->rb_left;
2072                 else {
2073                         p = &parent->rb_right;
2074                         left = 0;
2075                 }
2076         }
2077
2078         if (left)
2079                 st->left = &cfqq->rb_node;
2080
2081         cfqq->rb_key = rb_key;
2082         rb_link_node(&cfqq->rb_node, parent, p);
2083         rb_insert_color(&cfqq->rb_node, &st->rb);
2084         st->count++;
2085         if (add_front || !new_cfqq)
2086                 return;
2087         cfq_group_notify_queue_add(cfqd, cfqq->cfqg);
2088 }
2089
2090 static struct cfq_queue *
2091 cfq_prio_tree_lookup(struct cfq_data *cfqd, struct rb_root *root,
2092                      sector_t sector, struct rb_node **ret_parent,
2093                      struct rb_node ***rb_link)
2094 {
2095         struct rb_node **p, *parent;
2096         struct cfq_queue *cfqq = NULL;
2097
2098         parent = NULL;
2099         p = &root->rb_node;
2100         while (*p) {
2101                 struct rb_node **n;
2102
2103                 parent = *p;
2104                 cfqq = rb_entry(parent, struct cfq_queue, p_node);
2105
2106                 /*
2107                  * Sort strictly based on sector.  Smallest to the left,
2108                  * largest to the right.
2109                  */
2110                 if (sector > blk_rq_pos(cfqq->next_rq))
2111                         n = &(*p)->rb_right;
2112                 else if (sector < blk_rq_pos(cfqq->next_rq))
2113                         n = &(*p)->rb_left;
2114                 else
2115                         break;
2116                 p = n;
2117                 cfqq = NULL;
2118         }
2119
2120         *ret_parent = parent;
2121         if (rb_link)
2122                 *rb_link = p;
2123         return cfqq;
2124 }
2125
2126 static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2127 {
2128         struct rb_node **p, *parent;
2129         struct cfq_queue *__cfqq;
2130
2131         if (cfqq->p_root) {
2132                 rb_erase(&cfqq->p_node, cfqq->p_root);
2133                 cfqq->p_root = NULL;
2134         }
2135
2136         if (cfq_class_idle(cfqq))
2137                 return;
2138         if (!cfqq->next_rq)
2139                 return;
2140
2141         cfqq->p_root = &cfqd->prio_trees[cfqq->org_ioprio];
2142         __cfqq = cfq_prio_tree_lookup(cfqd, cfqq->p_root,
2143                                       blk_rq_pos(cfqq->next_rq), &parent, &p);
2144         if (!__cfqq) {
2145                 rb_link_node(&cfqq->p_node, parent, p);
2146                 rb_insert_color(&cfqq->p_node, cfqq->p_root);
2147         } else
2148                 cfqq->p_root = NULL;
2149 }
2150
2151 /*
2152  * Update cfqq's position in the service tree.
2153  */
2154 static void cfq_resort_rr_list(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2155 {
2156         /*
2157          * Resorting requires the cfqq to be on the RR list already.
2158          */
2159         if (cfq_cfqq_on_rr(cfqq)) {
2160                 cfq_service_tree_add(cfqd, cfqq, 0);
2161                 cfq_prio_tree_add(cfqd, cfqq);
2162         }
2163 }
2164
2165 /*
2166  * add to busy list of queues for service, trying to be fair in ordering
2167  * the pending list according to last request service
2168  */
2169 static void cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2170 {
2171         cfq_log_cfqq(cfqd, cfqq, "add_to_rr");
2172         BUG_ON(cfq_cfqq_on_rr(cfqq));
2173         cfq_mark_cfqq_on_rr(cfqq);
2174         cfqd->busy_queues++;
2175         if (cfq_cfqq_sync(cfqq))
2176                 cfqd->busy_sync_queues++;
2177
2178         cfq_resort_rr_list(cfqd, cfqq);
2179 }
2180
2181 /*
2182  * Called when the cfqq no longer has requests pending, remove it from
2183  * the service tree.
2184  */
2185 static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2186 {
2187         cfq_log_cfqq(cfqd, cfqq, "del_from_rr");
2188         BUG_ON(!cfq_cfqq_on_rr(cfqq));
2189         cfq_clear_cfqq_on_rr(cfqq);
2190
2191         if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2192                 cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2193                 cfqq->service_tree = NULL;
2194         }
2195         if (cfqq->p_root) {
2196                 rb_erase(&cfqq->p_node, cfqq->p_root);
2197                 cfqq->p_root = NULL;
2198         }
2199
2200         cfq_group_notify_queue_del(cfqd, cfqq->cfqg);
2201         BUG_ON(!cfqd->busy_queues);
2202         cfqd->busy_queues--;
2203         if (cfq_cfqq_sync(cfqq))
2204                 cfqd->busy_sync_queues--;
2205 }
2206
2207 /*
2208  * rb tree support functions
2209  */
2210 static void cfq_del_rq_rb(struct request *rq)
2211 {
2212         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2213         const int sync = rq_is_sync(rq);
2214
2215         BUG_ON(!cfqq->queued[sync]);
2216         cfqq->queued[sync]--;
2217
2218         elv_rb_del(&cfqq->sort_list, rq);
2219
2220         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list)) {
2221                 /*
2222                  * Queue will be deleted from service tree when we actually
2223                  * expire it later. Right now just remove it from prio tree
2224                  * as it is empty.
2225                  */
2226                 if (cfqq->p_root) {
2227                         rb_erase(&cfqq->p_node, cfqq->p_root);
2228                         cfqq->p_root = NULL;
2229                 }
2230         }
2231 }
2232
2233 static void cfq_add_rq_rb(struct request *rq)
2234 {
2235         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2236         struct cfq_data *cfqd = cfqq->cfqd;
2237         struct request *prev;
2238
2239         cfqq->queued[rq_is_sync(rq)]++;
2240
2241         elv_rb_add(&cfqq->sort_list, rq);
2242
2243         if (!cfq_cfqq_on_rr(cfqq))
2244                 cfq_add_cfqq_rr(cfqd, cfqq);
2245
2246         /*
2247          * check if this request is a better next-serve candidate
2248          */
2249         prev = cfqq->next_rq;
2250         cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq, cfqd->last_position);
2251
2252         /*
2253          * adjust priority tree position, if ->next_rq changes
2254          */
2255         if (prev != cfqq->next_rq)
2256                 cfq_prio_tree_add(cfqd, cfqq);
2257
2258         BUG_ON(!cfqq->next_rq);
2259 }
2260
2261 static void cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
2262 {
2263         elv_rb_del(&cfqq->sort_list, rq);
2264         cfqq->queued[rq_is_sync(rq)]--;
2265         cfqg_stats_update_io_remove(RQ_CFQG(rq), rq->cmd_flags);
2266         cfq_add_rq_rb(rq);
2267         cfqg_stats_update_io_add(RQ_CFQG(rq), cfqq->cfqd->serving_group,
2268                                  rq->cmd_flags);
2269 }
2270
2271 static struct request *
2272 cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio)
2273 {
2274         struct task_struct *tsk = current;
2275         struct cfq_io_cq *cic;
2276         struct cfq_queue *cfqq;
2277
2278         cic = cfq_cic_lookup(cfqd, tsk->io_context);
2279         if (!cic)
2280                 return NULL;
2281
2282         cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
2283         if (cfqq)
2284                 return elv_rb_find(&cfqq->sort_list, bio_end_sector(bio));
2285
2286         return NULL;
2287 }
2288
2289 static void cfq_activate_request(struct request_queue *q, struct request *rq)
2290 {
2291         struct cfq_data *cfqd = q->elevator->elevator_data;
2292
2293         cfqd->rq_in_driver++;
2294         cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "activate rq, drv=%d",
2295                                                 cfqd->rq_in_driver);
2296
2297         cfqd->last_position = blk_rq_pos(rq) + blk_rq_sectors(rq);
2298 }
2299
2300 static void cfq_deactivate_request(struct request_queue *q, struct request *rq)
2301 {
2302         struct cfq_data *cfqd = q->elevator->elevator_data;
2303
2304         WARN_ON(!cfqd->rq_in_driver);
2305         cfqd->rq_in_driver--;
2306         cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "deactivate rq, drv=%d",
2307                                                 cfqd->rq_in_driver);
2308 }
2309
2310 static void cfq_remove_request(struct request *rq)
2311 {
2312         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2313
2314         if (cfqq->next_rq == rq)
2315                 cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq);
2316
2317         list_del_init(&rq->queuelist);
2318         cfq_del_rq_rb(rq);
2319
2320         cfqq->cfqd->rq_queued--;
2321         cfqg_stats_update_io_remove(RQ_CFQG(rq), rq->cmd_flags);
2322         if (rq->cmd_flags & REQ_PRIO) {
2323                 WARN_ON(!cfqq->prio_pending);
2324                 cfqq->prio_pending--;
2325         }
2326 }
2327
2328 static int cfq_merge(struct request_queue *q, struct request **req,
2329                      struct bio *bio)
2330 {
2331         struct cfq_data *cfqd = q->elevator->elevator_data;
2332         struct request *__rq;
2333
2334         __rq = cfq_find_rq_fmerge(cfqd, bio);
2335         if (__rq && elv_rq_merge_ok(__rq, bio)) {
2336                 *req = __rq;
2337                 return ELEVATOR_FRONT_MERGE;
2338         }
2339
2340         return ELEVATOR_NO_MERGE;
2341 }
2342
2343 static void cfq_merged_request(struct request_queue *q, struct request *req,
2344                                int type)
2345 {
2346         if (type == ELEVATOR_FRONT_MERGE) {
2347                 struct cfq_queue *cfqq = RQ_CFQQ(req);
2348
2349                 cfq_reposition_rq_rb(cfqq, req);
2350         }
2351 }
2352
2353 static void cfq_bio_merged(struct request_queue *q, struct request *req,
2354                                 struct bio *bio)
2355 {
2356         cfqg_stats_update_io_merged(RQ_CFQG(req), bio->bi_rw);
2357 }
2358
2359 static void
2360 cfq_merged_requests(struct request_queue *q, struct request *rq,
2361                     struct request *next)
2362 {
2363         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2364         struct cfq_data *cfqd = q->elevator->elevator_data;
2365
2366         /*
2367          * reposition in fifo if next is older than rq
2368          */
2369         if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
2370             time_before(rq_fifo_time(next), rq_fifo_time(rq)) &&
2371             cfqq == RQ_CFQQ(next)) {
2372                 list_move(&rq->queuelist, &next->queuelist);
2373                 rq_set_fifo_time(rq, rq_fifo_time(next));
2374         }
2375
2376         if (cfqq->next_rq == next)
2377                 cfqq->next_rq = rq;
2378         cfq_remove_request(next);
2379         cfqg_stats_update_io_merged(RQ_CFQG(rq), next->cmd_flags);
2380
2381         cfqq = RQ_CFQQ(next);
2382         /*
2383          * all requests of this queue are merged to other queues, delete it
2384          * from the service tree. If it's the active_queue,
2385          * cfq_dispatch_requests() will choose to expire it or do idle
2386          */
2387         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list) &&
2388             cfqq != cfqd->active_queue)
2389                 cfq_del_cfqq_rr(cfqd, cfqq);
2390 }
2391
2392 static int cfq_allow_merge(struct request_queue *q, struct request *rq,
2393                            struct bio *bio)
2394 {
2395         struct cfq_data *cfqd = q->elevator->elevator_data;
2396         struct cfq_io_cq *cic;
2397         struct cfq_queue *cfqq;
2398
2399         /*
2400          * Disallow merge of a sync bio into an async request.
2401          */
2402         if (cfq_bio_sync(bio) && !rq_is_sync(rq))
2403                 return false;
2404
2405         /*
2406          * Lookup the cfqq that this bio will be queued with and allow
2407          * merge only if rq is queued there.
2408          */
2409         cic = cfq_cic_lookup(cfqd, current->io_context);
2410         if (!cic)
2411                 return false;
2412
2413         cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
2414         return cfqq == RQ_CFQQ(rq);
2415 }
2416
2417 static inline void cfq_del_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2418 {
2419         del_timer(&cfqd->idle_slice_timer);
2420         cfqg_stats_update_idle_time(cfqq->cfqg);
2421 }
2422
2423 static void __cfq_set_active_queue(struct cfq_data *cfqd,
2424                                    struct cfq_queue *cfqq)
2425 {
2426         if (cfqq) {
2427                 cfq_log_cfqq(cfqd, cfqq, "set_active wl_class:%d wl_type:%d",
2428                                 cfqd->serving_wl_class, cfqd->serving_wl_type);
2429                 cfqg_stats_update_avg_queue_size(cfqq->cfqg);
2430                 cfqq->slice_start = 0;
2431                 cfqq->dispatch_start = jiffies;
2432                 cfqq->allocated_slice = 0;
2433                 cfqq->slice_end = 0;
2434                 cfqq->slice_dispatch = 0;
2435                 cfqq->nr_sectors = 0;
2436
2437                 cfq_clear_cfqq_wait_request(cfqq);
2438                 cfq_clear_cfqq_must_dispatch(cfqq);
2439                 cfq_clear_cfqq_must_alloc_slice(cfqq);
2440                 cfq_clear_cfqq_fifo_expire(cfqq);
2441                 cfq_mark_cfqq_slice_new(cfqq);
2442
2443                 cfq_del_timer(cfqd, cfqq);
2444         }
2445
2446         cfqd->active_queue = cfqq;
2447 }
2448
2449 /*
2450  * current cfqq expired its slice (or was too idle), select new one
2451  */
2452 static void
2453 __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2454                     bool timed_out)
2455 {
2456         cfq_log_cfqq(cfqd, cfqq, "slice expired t=%d", timed_out);
2457
2458         if (cfq_cfqq_wait_request(cfqq))
2459                 cfq_del_timer(cfqd, cfqq);
2460
2461         cfq_clear_cfqq_wait_request(cfqq);
2462         cfq_clear_cfqq_wait_busy(cfqq);
2463
2464         /*
2465          * If this cfqq is shared between multiple processes, check to
2466          * make sure that those processes are still issuing I/Os within
2467          * the mean seek distance.  If not, it may be time to break the
2468          * queues apart again.
2469          */
2470         if (cfq_cfqq_coop(cfqq) && CFQQ_SEEKY(cfqq))
2471                 cfq_mark_cfqq_split_coop(cfqq);
2472
2473         /*
2474          * store what was left of this slice, if the queue idled/timed out
2475          */
2476         if (timed_out) {
2477                 if (cfq_cfqq_slice_new(cfqq))
2478                         cfqq->slice_resid = cfq_scaled_cfqq_slice(cfqd, cfqq);
2479                 else
2480                         cfqq->slice_resid = cfqq->slice_end - jiffies;
2481                 cfq_log_cfqq(cfqd, cfqq, "resid=%ld", cfqq->slice_resid);
2482         }
2483
2484         cfq_group_served(cfqd, cfqq->cfqg, cfqq);
2485
2486         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list))
2487                 cfq_del_cfqq_rr(cfqd, cfqq);
2488
2489         cfq_resort_rr_list(cfqd, cfqq);
2490
2491         if (cfqq == cfqd->active_queue)
2492                 cfqd->active_queue = NULL;
2493
2494         if (cfqd->active_cic) {
2495                 put_io_context(cfqd->active_cic->icq.ioc);
2496                 cfqd->active_cic = NULL;
2497         }
2498 }
2499
2500 static inline void cfq_slice_expired(struct cfq_data *cfqd, bool timed_out)
2501 {
2502         struct cfq_queue *cfqq = cfqd->active_queue;
2503
2504         if (cfqq)
2505                 __cfq_slice_expired(cfqd, cfqq, timed_out);
2506 }
2507
2508 /*
2509  * Get next queue for service. Unless we have a queue preemption,
2510  * we'll simply select the first cfqq in the service tree.
2511  */
2512 static struct cfq_queue *cfq_get_next_queue(struct cfq_data *cfqd)
2513 {
2514         struct cfq_rb_root *st = st_for(cfqd->serving_group,
2515                         cfqd->serving_wl_class, cfqd->serving_wl_type);
2516
2517         if (!cfqd->rq_queued)
2518                 return NULL;
2519
2520         /* There is nothing to dispatch */
2521         if (!st)
2522                 return NULL;
2523         if (RB_EMPTY_ROOT(&st->rb))
2524                 return NULL;
2525         return cfq_rb_first(st);
2526 }
2527
2528 static struct cfq_queue *cfq_get_next_queue_forced(struct cfq_data *cfqd)
2529 {
2530         struct cfq_group *cfqg;
2531         struct cfq_queue *cfqq;
2532         int i, j;
2533         struct cfq_rb_root *st;
2534
2535         if (!cfqd->rq_queued)
2536                 return NULL;
2537
2538         cfqg = cfq_get_next_cfqg(cfqd);
2539         if (!cfqg)
2540                 return NULL;
2541
2542         for_each_cfqg_st(cfqg, i, j, st)
2543                 if ((cfqq = cfq_rb_first(st)) != NULL)
2544                         return cfqq;
2545         return NULL;
2546 }
2547
2548 /*
2549  * Get and set a new active queue for service.
2550  */
2551 static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd,
2552                                               struct cfq_queue *cfqq)
2553 {
2554         if (!cfqq)
2555                 cfqq = cfq_get_next_queue(cfqd);
2556
2557         __cfq_set_active_queue(cfqd, cfqq);
2558         return cfqq;
2559 }
2560
2561 static inline sector_t cfq_dist_from_last(struct cfq_data *cfqd,
2562                                           struct request *rq)
2563 {
2564         if (blk_rq_pos(rq) >= cfqd->last_position)
2565                 return blk_rq_pos(rq) - cfqd->last_position;
2566         else
2567                 return cfqd->last_position - blk_rq_pos(rq);
2568 }
2569
2570 static inline int cfq_rq_close(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2571                                struct request *rq)
2572 {
2573         return cfq_dist_from_last(cfqd, rq) <= CFQQ_CLOSE_THR;
2574 }
2575
2576 static struct cfq_queue *cfqq_close(struct cfq_data *cfqd,
2577                                     struct cfq_queue *cur_cfqq)
2578 {
2579         struct rb_root *root = &cfqd->prio_trees[cur_cfqq->org_ioprio];
2580         struct rb_node *parent, *node;
2581         struct cfq_queue *__cfqq;
2582         sector_t sector = cfqd->last_position;
2583
2584         if (RB_EMPTY_ROOT(root))
2585                 return NULL;
2586
2587         /*
2588          * First, if we find a request starting at the end of the last
2589          * request, choose it.
2590          */
2591         __cfqq = cfq_prio_tree_lookup(cfqd, root, sector, &parent, NULL);
2592         if (__cfqq)
2593                 return __cfqq;
2594
2595         /*
2596          * If the exact sector wasn't found, the parent of the NULL leaf
2597          * will contain the closest sector.
2598          */
2599         __cfqq = rb_entry(parent, struct cfq_queue, p_node);
2600         if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2601                 return __cfqq;
2602
2603         if (blk_rq_pos(__cfqq->next_rq) < sector)
2604                 node = rb_next(&__cfqq->p_node);
2605         else
2606                 node = rb_prev(&__cfqq->p_node);
2607         if (!node)
2608                 return NULL;
2609
2610         __cfqq = rb_entry(node, struct cfq_queue, p_node);
2611         if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2612                 return __cfqq;
2613
2614         return NULL;
2615 }
2616
2617 /*
2618  * cfqd - obvious
2619  * cur_cfqq - passed in so that we don't decide that the current queue is
2620  *            closely cooperating with itself.
2621  *
2622  * So, basically we're assuming that that cur_cfqq has dispatched at least
2623  * one request, and that cfqd->last_position reflects a position on the disk
2624  * associated with the I/O issued by cur_cfqq.  I'm not sure this is a valid
2625  * assumption.
2626  */
2627 static struct cfq_queue *cfq_close_cooperator(struct cfq_data *cfqd,
2628                                               struct cfq_queue *cur_cfqq)
2629 {
2630         struct cfq_queue *cfqq;
2631
2632         if (cfq_class_idle(cur_cfqq))
2633                 return NULL;
2634         if (!cfq_cfqq_sync(cur_cfqq))
2635                 return NULL;
2636         if (CFQQ_SEEKY(cur_cfqq))
2637                 return NULL;
2638
2639         /*
2640          * Don't search priority tree if it's the only queue in the group.
2641          */
2642         if (cur_cfqq->cfqg->nr_cfqq == 1)
2643                 return NULL;
2644
2645         /*
2646          * We should notice if some of the queues are cooperating, eg
2647          * working closely on the same area of the disk. In that case,
2648          * we can group them together and don't waste time idling.
2649          */
2650         cfqq = cfqq_close(cfqd, cur_cfqq);
2651         if (!cfqq)
2652                 return NULL;
2653
2654         /* If new queue belongs to different cfq_group, don't choose it */
2655         if (cur_cfqq->cfqg != cfqq->cfqg)
2656                 return NULL;
2657
2658         /*
2659          * It only makes sense to merge sync queues.
2660          */
2661         if (!cfq_cfqq_sync(cfqq))
2662                 return NULL;
2663         if (CFQQ_SEEKY(cfqq))
2664                 return NULL;
2665
2666         /*
2667          * Do not merge queues of different priority classes
2668          */
2669         if (cfq_class_rt(cfqq) != cfq_class_rt(cur_cfqq))
2670                 return NULL;
2671
2672         return cfqq;
2673 }
2674
2675 /*
2676  * Determine whether we should enforce idle window for this queue.
2677  */
2678
2679 static bool cfq_should_idle(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2680 {
2681         enum wl_class_t wl_class = cfqq_class(cfqq);
2682         struct cfq_rb_root *st = cfqq->service_tree;
2683
2684         BUG_ON(!st);
2685         BUG_ON(!st->count);
2686
2687         if (!cfqd->cfq_slice_idle)
2688                 return false;
2689
2690         /* We never do for idle class queues. */
2691         if (wl_class == IDLE_WORKLOAD)
2692                 return false;
2693
2694         /* We do for queues that were marked with idle window flag. */
2695         if (cfq_cfqq_idle_window(cfqq) &&
2696            !(blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag))
2697                 return true;
2698
2699         /*
2700          * Otherwise, we do only if they are the last ones
2701          * in their service tree.
2702          */
2703         if (st->count == 1 && cfq_cfqq_sync(cfqq) &&
2704            !cfq_io_thinktime_big(cfqd, &st->ttime, false))
2705                 return true;
2706         cfq_log_cfqq(cfqd, cfqq, "Not idling. st->count:%d", st->count);
2707         return false;
2708 }
2709
2710 static void cfq_arm_slice_timer(struct cfq_data *cfqd)
2711 {
2712         struct cfq_queue *cfqq = cfqd->active_queue;
2713         struct cfq_io_cq *cic;
2714         unsigned long sl, group_idle = 0;
2715
2716         /*
2717          * SSD device without seek penalty, disable idling. But only do so
2718          * for devices that support queuing, otherwise we still have a problem
2719          * with sync vs async workloads.
2720          */
2721         if (blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag)
2722                 return;
2723
2724         WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list));
2725         WARN_ON(cfq_cfqq_slice_new(cfqq));
2726
2727         /*
2728          * idle is disabled, either manually or by past process history
2729          */
2730         if (!cfq_should_idle(cfqd, cfqq)) {
2731                 /* no queue idling. Check for group idling */
2732                 if (cfqd->cfq_group_idle)
2733                         group_idle = cfqd->cfq_group_idle;
2734                 else
2735                         return;
2736         }
2737
2738         /*
2739          * still active requests from this queue, don't idle
2740          */
2741         if (cfqq->dispatched)
2742                 return;
2743
2744         /*
2745          * task has exited, don't wait
2746          */
2747         cic = cfqd->active_cic;
2748         if (!cic || !atomic_read(&cic->icq.ioc->active_ref))
2749                 return;
2750
2751         /*
2752          * If our average think time is larger than the remaining time
2753          * slice, then don't idle. This avoids overrunning the allotted
2754          * time slice.
2755          */
2756         if (sample_valid(cic->ttime.ttime_samples) &&
2757             (cfqq->slice_end - jiffies < cic->ttime.ttime_mean)) {
2758                 cfq_log_cfqq(cfqd, cfqq, "Not idling. think_time:%lu",
2759                              cic->ttime.ttime_mean);
2760                 return;
2761         }
2762
2763         /* There are other queues in the group, don't do group idle */
2764         if (group_idle && cfqq->cfqg->nr_cfqq > 1)
2765                 return;
2766
2767         cfq_mark_cfqq_wait_request(cfqq);
2768
2769         if (group_idle)
2770                 sl = cfqd->cfq_group_idle;
2771         else
2772                 sl = cfqd->cfq_slice_idle;
2773
2774         mod_timer(&cfqd->idle_slice_timer, jiffies + sl);
2775         cfqg_stats_set_start_idle_time(cfqq->cfqg);
2776         cfq_log_cfqq(cfqd, cfqq, "arm_idle: %lu group_idle: %d", sl,
2777                         group_idle ? 1 : 0);
2778 }
2779
2780 /*
2781  * Move request from internal lists to the request queue dispatch list.
2782  */
2783 static void cfq_dispatch_insert(struct request_queue *q, struct request *rq)
2784 {
2785         struct cfq_data *cfqd = q->elevator->elevator_data;
2786         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2787
2788         cfq_log_cfqq(cfqd, cfqq, "dispatch_insert");
2789
2790         cfqq->next_rq = cfq_find_next_rq(cfqd, cfqq, rq);
2791         cfq_remove_request(rq);
2792         cfqq->dispatched++;
2793         (RQ_CFQG(rq))->dispatched++;
2794         elv_dispatch_sort(q, rq);
2795
2796         cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]++;
2797         cfqq->nr_sectors += blk_rq_sectors(rq);
2798         cfqg_stats_update_dispatch(cfqq->cfqg, blk_rq_bytes(rq), rq->cmd_flags);
2799 }
2800
2801 /*
2802  * return expired entry, or NULL to just start from scratch in rbtree
2803  */
2804 static struct request *cfq_check_fifo(struct cfq_queue *cfqq)
2805 {
2806         struct request *rq = NULL;
2807
2808         if (cfq_cfqq_fifo_expire(cfqq))
2809                 return NULL;
2810
2811         cfq_mark_cfqq_fifo_expire(cfqq);
2812
2813         if (list_empty(&cfqq->fifo))
2814                 return NULL;
2815
2816         rq = rq_entry_fifo(cfqq->fifo.next);
2817         if (time_before(jiffies, rq_fifo_time(rq)))
2818                 rq = NULL;
2819
2820         cfq_log_cfqq(cfqq->cfqd, cfqq, "fifo=%p", rq);
2821         return rq;
2822 }
2823
2824 static inline int
2825 cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2826 {
2827         const int base_rq = cfqd->cfq_slice_async_rq;
2828
2829         WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
2830
2831         return 2 * base_rq * (IOPRIO_BE_NR - cfqq->ioprio);
2832 }
2833
2834 /*
2835  * Must be called with the queue_lock held.
2836  */
2837 static int cfqq_process_refs(struct cfq_queue *cfqq)
2838 {
2839         int process_refs, io_refs;
2840
2841         io_refs = cfqq->allocated[READ] + cfqq->allocated[WRITE];
2842         process_refs = cfqq->ref - io_refs;
2843         BUG_ON(process_refs < 0);
2844         return process_refs;
2845 }
2846
2847 static void cfq_setup_merge(struct cfq_queue *cfqq, struct cfq_queue *new_cfqq)
2848 {
2849         int process_refs, new_process_refs;
2850         struct cfq_queue *__cfqq;
2851
2852         /*
2853          * If there are no process references on the new_cfqq, then it is
2854          * unsafe to follow the ->new_cfqq chain as other cfqq's in the
2855          * chain may have dropped their last reference (not just their
2856          * last process reference).
2857          */
2858         if (!cfqq_process_refs(new_cfqq))
2859                 return;
2860
2861         /* Avoid a circular list and skip interim queue merges */
2862         while ((__cfqq = new_cfqq->new_cfqq)) {
2863                 if (__cfqq == cfqq)
2864                         return;
2865                 new_cfqq = __cfqq;
2866         }
2867
2868         process_refs = cfqq_process_refs(cfqq);
2869         new_process_refs = cfqq_process_refs(new_cfqq);
2870         /*
2871          * If the process for the cfqq has gone away, there is no
2872          * sense in merging the queues.
2873          */
2874         if (process_refs == 0 || new_process_refs == 0)
2875                 return;
2876
2877         /*
2878          * Merge in the direction of the lesser amount of work.
2879          */
2880         if (new_process_refs >= process_refs) {
2881                 cfqq->new_cfqq = new_cfqq;
2882                 new_cfqq->ref += process_refs;
2883         } else {
2884                 new_cfqq->new_cfqq = cfqq;
2885                 cfqq->ref += new_process_refs;
2886         }
2887 }
2888
2889 static enum wl_type_t cfq_choose_wl_type(struct cfq_data *cfqd,
2890                         struct cfq_group *cfqg, enum wl_class_t wl_class)
2891 {
2892         struct cfq_queue *queue;
2893         int i;
2894         bool key_valid = false;
2895         unsigned long lowest_key = 0;
2896         enum wl_type_t cur_best = SYNC_NOIDLE_WORKLOAD;
2897
2898         for (i = 0; i <= SYNC_WORKLOAD; ++i) {
2899                 /* select the one with lowest rb_key */
2900                 queue = cfq_rb_first(st_for(cfqg, wl_class, i));
2901                 if (queue &&
2902                     (!key_valid || time_before(queue->rb_key, lowest_key))) {
2903                         lowest_key = queue->rb_key;
2904                         cur_best = i;
2905                         key_valid = true;
2906                 }
2907         }
2908
2909         return cur_best;
2910 }
2911
2912 static void
2913 choose_wl_class_and_type(struct cfq_data *cfqd, struct cfq_group *cfqg)
2914 {
2915         unsigned slice;
2916         unsigned count;
2917         struct cfq_rb_root *st;
2918         unsigned group_slice;
2919         enum wl_class_t original_class = cfqd->serving_wl_class;
2920
2921         /* Choose next priority. RT > BE > IDLE */
2922         if (cfq_group_busy_queues_wl(RT_WORKLOAD, cfqd, cfqg))
2923                 cfqd->serving_wl_class = RT_WORKLOAD;
2924         else if (cfq_group_busy_queues_wl(BE_WORKLOAD, cfqd, cfqg))
2925                 cfqd->serving_wl_class = BE_WORKLOAD;
2926         else {
2927                 cfqd->serving_wl_class = IDLE_WORKLOAD;
2928                 cfqd->workload_expires = jiffies + 1;
2929                 return;
2930         }
2931
2932         if (original_class != cfqd->serving_wl_class)
2933                 goto new_workload;
2934
2935         /*
2936          * For RT and BE, we have to choose also the type
2937          * (SYNC, SYNC_NOIDLE, ASYNC), and to compute a workload
2938          * expiration time
2939          */
2940         st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
2941         count = st->count;
2942
2943         /*
2944          * check workload expiration, and that we still have other queues ready
2945          */
2946         if (count && !time_after(jiffies, cfqd->workload_expires))
2947                 return;
2948
2949 new_workload:
2950         /* otherwise select new workload type */
2951         cfqd->serving_wl_type = cfq_choose_wl_type(cfqd, cfqg,
2952                                         cfqd->serving_wl_class);
2953         st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
2954         count = st->count;
2955
2956         /*
2957          * the workload slice is computed as a fraction of target latency
2958          * proportional to the number of queues in that workload, over
2959          * all the queues in the same priority class
2960          */
2961         group_slice = cfq_group_slice(cfqd, cfqg);
2962
2963         slice = group_slice * count /
2964                 max_t(unsigned, cfqg->busy_queues_avg[cfqd->serving_wl_class],
2965                       cfq_group_busy_queues_wl(cfqd->serving_wl_class, cfqd,
2966                                         cfqg));
2967
2968         if (cfqd->serving_wl_type == ASYNC_WORKLOAD) {
2969                 unsigned int tmp;
2970
2971                 /*
2972                  * Async queues are currently system wide. Just taking
2973                  * proportion of queues with-in same group will lead to higher
2974                  * async ratio system wide as generally root group is going
2975                  * to have higher weight. A more accurate thing would be to
2976                  * calculate system wide asnc/sync ratio.
2977                  */
2978                 tmp = cfqd->cfq_target_latency *
2979                         cfqg_busy_async_queues(cfqd, cfqg);
2980                 tmp = tmp/cfqd->busy_queues;
2981                 slice = min_t(unsigned, slice, tmp);
2982
2983                 /* async workload slice is scaled down according to
2984                  * the sync/async slice ratio. */
2985                 slice = slice * cfqd->cfq_slice[0] / cfqd->cfq_slice[1];
2986         } else
2987                 /* sync workload slice is at least 2 * cfq_slice_idle */
2988                 slice = max(slice, 2 * cfqd->cfq_slice_idle);
2989
2990         slice = max_t(unsigned, slice, CFQ_MIN_TT);
2991         cfq_log(cfqd, "workload slice:%d", slice);
2992         cfqd->workload_expires = jiffies + slice;
2993 }
2994
2995 static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd)
2996 {
2997         struct cfq_rb_root *st = &cfqd->grp_service_tree;
2998         struct cfq_group *cfqg;
2999
3000         if (RB_EMPTY_ROOT(&st->rb))
3001                 return NULL;
3002         cfqg = cfq_rb_first_group(st);
3003         update_min_vdisktime(st);
3004         return cfqg;
3005 }
3006
3007 static void cfq_choose_cfqg(struct cfq_data *cfqd)
3008 {
3009         struct cfq_group *cfqg = cfq_get_next_cfqg(cfqd);
3010
3011         cfqd->serving_group = cfqg;
3012
3013         /* Restore the workload type data */
3014         if (cfqg->saved_wl_slice) {
3015                 cfqd->workload_expires = jiffies + cfqg->saved_wl_slice;
3016                 cfqd->serving_wl_type = cfqg->saved_wl_type;
3017                 cfqd->serving_wl_class = cfqg->saved_wl_class;
3018         } else
3019                 cfqd->workload_expires = jiffies - 1;
3020
3021         choose_wl_class_and_type(cfqd, cfqg);
3022 }
3023
3024 /*
3025  * Select a queue for service. If we have a current active queue,
3026  * check whether to continue servicing it, or retrieve and set a new one.
3027  */
3028 static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
3029 {
3030         struct cfq_queue *cfqq, *new_cfqq = NULL;
3031
3032         cfqq = cfqd->active_queue;
3033         if (!cfqq)
3034                 goto new_queue;
3035
3036         if (!cfqd->rq_queued)
3037                 return NULL;
3038
3039         /*
3040          * We were waiting for group to get backlogged. Expire the queue
3041          */
3042         if (cfq_cfqq_wait_busy(cfqq) && !RB_EMPTY_ROOT(&cfqq->sort_list))
3043                 goto expire;
3044
3045         /*
3046          * The active queue has run out of time, expire it and select new.
3047          */
3048         if (cfq_slice_used(cfqq) && !cfq_cfqq_must_dispatch(cfqq)) {
3049                 /*
3050                  * If slice had not expired at the completion of last request
3051                  * we might not have turned on wait_busy flag. Don't expire
3052                  * the queue yet. Allow the group to get backlogged.
3053                  *
3054                  * The very fact that we have used the slice, that means we
3055                  * have been idling all along on this queue and it should be
3056                  * ok to wait for this request to complete.
3057                  */
3058                 if (cfqq->cfqg->nr_cfqq == 1 && RB_EMPTY_ROOT(&cfqq->sort_list)
3059                     && cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3060                         cfqq = NULL;
3061                         goto keep_queue;
3062                 } else
3063                         goto check_group_idle;
3064         }
3065
3066         /*
3067          * The active queue has requests and isn't expired, allow it to
3068          * dispatch.
3069          */
3070         if (!RB_EMPTY_ROOT(&cfqq->sort_list))
3071                 goto keep_queue;
3072
3073         /*
3074          * If another queue has a request waiting within our mean seek
3075          * distance, let it run.  The expire code will check for close
3076          * cooperators and put the close queue at the front of the service
3077          * tree.  If possible, merge the expiring queue with the new cfqq.
3078          */
3079         new_cfqq = cfq_close_cooperator(cfqd, cfqq);
3080         if (new_cfqq) {
3081                 if (!cfqq->new_cfqq)
3082                         cfq_setup_merge(cfqq, new_cfqq);
3083                 goto expire;
3084         }
3085
3086         /*
3087          * No requests pending. If the active queue still has requests in
3088          * flight or is idling for a new request, allow either of these
3089          * conditions to happen (or time out) before selecting a new queue.
3090          */
3091         if (timer_pending(&cfqd->idle_slice_timer)) {
3092                 cfqq = NULL;
3093                 goto keep_queue;
3094         }
3095
3096         /*
3097          * This is a deep seek queue, but the device is much faster than
3098          * the queue can deliver, don't idle
3099          **/
3100         if (CFQQ_SEEKY(cfqq) && cfq_cfqq_idle_window(cfqq) &&
3101             (cfq_cfqq_slice_new(cfqq) ||
3102             (cfqq->slice_end - jiffies > jiffies - cfqq->slice_start))) {
3103                 cfq_clear_cfqq_deep(cfqq);
3104                 cfq_clear_cfqq_idle_window(cfqq);
3105         }
3106
3107         if (cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3108                 cfqq = NULL;
3109                 goto keep_queue;
3110         }
3111
3112         /*
3113          * If group idle is enabled and there are requests dispatched from
3114          * this group, wait for requests to complete.
3115          */
3116 check_group_idle:
3117         if (cfqd->cfq_group_idle && cfqq->cfqg->nr_cfqq == 1 &&
3118             cfqq->cfqg->dispatched &&
3119             !cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true)) {
3120                 cfqq = NULL;
3121                 goto keep_queue;
3122         }
3123
3124 expire:
3125         cfq_slice_expired(cfqd, 0);
3126 new_queue:
3127         /*
3128          * Current queue expired. Check if we have to switch to a new
3129          * service tree
3130          */
3131         if (!new_cfqq)
3132                 cfq_choose_cfqg(cfqd);
3133
3134         cfqq = cfq_set_active_queue(cfqd, new_cfqq);
3135 keep_queue:
3136         return cfqq;
3137 }
3138
3139 static int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq)
3140 {
3141         int dispatched = 0;
3142
3143         while (cfqq->next_rq) {
3144                 cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq);
3145                 dispatched++;
3146         }
3147
3148         BUG_ON(!list_empty(&cfqq->fifo));
3149
3150         /* By default cfqq is not expired if it is empty. Do it explicitly */
3151         __cfq_slice_expired(cfqq->cfqd, cfqq, 0);
3152         return dispatched;
3153 }
3154
3155 /*
3156  * Drain our current requests. Used for barriers and when switching
3157  * io schedulers on-the-fly.
3158  */
3159 static int cfq_forced_dispatch(struct cfq_data *cfqd)
3160 {
3161         struct cfq_queue *cfqq;
3162         int dispatched = 0;
3163
3164         /* Expire the timeslice of the current active queue first */
3165         cfq_slice_expired(cfqd, 0);
3166         while ((cfqq = cfq_get_next_queue_forced(cfqd)) != NULL) {
3167                 __cfq_set_active_queue(cfqd, cfqq);
3168                 dispatched += __cfq_forced_dispatch_cfqq(cfqq);
3169         }
3170
3171         BUG_ON(cfqd->busy_queues);
3172
3173         cfq_log(cfqd, "forced_dispatch=%d", dispatched);
3174         return dispatched;
3175 }
3176
3177 static inline bool cfq_slice_used_soon(struct cfq_data *cfqd,
3178         struct cfq_queue *cfqq)
3179 {
3180         /* the queue hasn't finished any request, can't estimate */
3181         if (cfq_cfqq_slice_new(cfqq))
3182                 return true;
3183         if (time_after(jiffies + cfqd->cfq_slice_idle * cfqq->dispatched,
3184                 cfqq->slice_end))
3185                 return true;
3186
3187         return false;
3188 }
3189
3190 static bool cfq_may_dispatch(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3191 {
3192         unsigned int max_dispatch;
3193
3194         /*
3195          * Drain async requests before we start sync IO
3196          */
3197         if (cfq_should_idle(cfqd, cfqq) && cfqd->rq_in_flight[BLK_RW_ASYNC])
3198                 return false;
3199
3200         /*
3201          * If this is an async queue and we have sync IO in flight, let it wait
3202          */
3203         if (cfqd->rq_in_flight[BLK_RW_SYNC] && !cfq_cfqq_sync(cfqq))
3204                 return false;
3205
3206         max_dispatch = max_t(unsigned int, cfqd->cfq_quantum / 2, 1);
3207         if (cfq_class_idle(cfqq))
3208                 max_dispatch = 1;
3209
3210         /*
3211          * Does this cfqq already have too much IO in flight?
3212          */
3213         if (cfqq->dispatched >= max_dispatch) {
3214                 bool promote_sync = false;
3215                 /*
3216                  * idle queue must always only have a single IO in flight
3217                  */
3218                 if (cfq_class_idle(cfqq))
3219                         return false;
3220
3221                 /*
3222                  * If there is only one sync queue
3223                  * we can ignore async queue here and give the sync
3224                  * queue no dispatch limit. The reason is a sync queue can
3225                  * preempt async queue, limiting the sync queue doesn't make
3226                  * sense. This is useful for aiostress test.
3227                  */
3228                 if (cfq_cfqq_sync(cfqq) && cfqd->busy_sync_queues == 1)
3229                         promote_sync = true;
3230
3231                 /*
3232                  * We have other queues, don't allow more IO from this one
3233                  */
3234                 if (cfqd->busy_queues > 1 && cfq_slice_used_soon(cfqd, cfqq) &&
3235                                 !promote_sync)
3236                         return false;
3237
3238                 /*
3239                  * Sole queue user, no limit
3240                  */
3241                 if (cfqd->busy_queues == 1 || promote_sync)
3242                         max_dispatch = -1;
3243                 else
3244                         /*
3245                          * Normally we start throttling cfqq when cfq_quantum/2
3246                          * requests have been dispatched. But we can drive
3247                          * deeper queue depths at the beginning of slice
3248                          * subjected to upper limit of cfq_quantum.
3249                          * */
3250                         max_dispatch = cfqd->cfq_quantum;
3251         }
3252
3253         /*
3254          * Async queues must wait a bit before being allowed dispatch.
3255          * We also ramp up the dispatch depth gradually for async IO,
3256          * based on the last sync IO we serviced
3257          */
3258         if (!cfq_cfqq_sync(cfqq) && cfqd->cfq_latency) {
3259                 unsigned long last_sync = jiffies - cfqd->last_delayed_sync;
3260                 unsigned int depth;
3261
3262                 depth = last_sync / cfqd->cfq_slice[1];
3263                 if (!depth && !cfqq->dispatched)
3264                         depth = 1;
3265                 if (depth < max_dispatch)
3266                         max_dispatch = depth;
3267         }
3268
3269         /*
3270          * If we're below the current max, allow a dispatch
3271          */
3272         return cfqq->dispatched < max_dispatch;
3273 }
3274
3275 /*
3276  * Dispatch a request from cfqq, moving them to the request queue
3277  * dispatch list.
3278  */
3279 static bool cfq_dispatch_request(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3280 {
3281         struct request *rq;
3282
3283         BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list));
3284
3285         if (!cfq_may_dispatch(cfqd, cfqq))
3286                 return false;
3287
3288         /*
3289          * follow expired path, else get first next available
3290          */
3291         rq = cfq_check_fifo(cfqq);
3292         if (!rq)
3293                 rq = cfqq->next_rq;
3294
3295         /*
3296          * insert request into driver dispatch list
3297          */
3298         cfq_dispatch_insert(cfqd->queue, rq);
3299
3300         if (!cfqd->active_cic) {
3301                 struct cfq_io_cq *cic = RQ_CIC(rq);
3302
3303                 atomic_long_inc(&cic->icq.ioc->refcount);
3304                 cfqd->active_cic = cic;
3305         }
3306
3307         return true;
3308 }
3309
3310 /*
3311  * Find the cfqq that we need to service and move a request from that to the
3312  * dispatch list
3313  */
3314 static int cfq_dispatch_requests(struct request_queue *q, int force)
3315 {
3316         struct cfq_data *cfqd = q->elevator->elevator_data;
3317         struct cfq_queue *cfqq;
3318
3319         if (!cfqd->busy_queues)
3320                 return 0;
3321
3322         if (unlikely(force))
3323                 return cfq_forced_dispatch(cfqd);
3324
3325         cfqq = cfq_select_queue(cfqd);
3326         if (!cfqq)
3327                 return 0;
3328
3329         /*
3330          * Dispatch a request from this cfqq, if it is allowed
3331          */
3332         if (!cfq_dispatch_request(cfqd, cfqq))
3333                 return 0;
3334
3335         cfqq->slice_dispatch++;
3336         cfq_clear_cfqq_must_dispatch(cfqq);
3337
3338         /*
3339          * expire an async queue immediately if it has used up its slice. idle
3340          * queue always expire after 1 dispatch round.
3341          */
3342         if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) &&
3343             cfqq->slice_dispatch >= cfq_prio_to_maxrq(cfqd, cfqq)) ||
3344             cfq_class_idle(cfqq))) {
3345                 cfqq->slice_end = jiffies + 1;
3346                 cfq_slice_expired(cfqd, 0);
3347         }
3348
3349         cfq_log_cfqq(cfqd, cfqq, "dispatched a request");
3350         return 1;
3351 }
3352
3353 /*
3354  * task holds one reference to the queue, dropped when task exits. each rq
3355  * in-flight on this queue also holds a reference, dropped when rq is freed.
3356  *
3357  * Each cfq queue took a reference on the parent group. Drop it now.
3358  * queue lock must be held here.
3359  */
3360 static void cfq_put_queue(struct cfq_queue *cfqq)
3361 {
3362         struct cfq_data *cfqd = cfqq->cfqd;
3363         struct cfq_group *cfqg;
3364
3365         BUG_ON(cfqq->ref <= 0);
3366
3367         cfqq->ref--;
3368         if (cfqq->ref)
3369                 return;
3370
3371         cfq_log_cfqq(cfqd, cfqq, "put_queue");
3372         BUG_ON(rb_first(&cfqq->sort_list));
3373         BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
3374         cfqg = cfqq->cfqg;
3375
3376         if (unlikely(cfqd->active_queue == cfqq)) {
3377                 __cfq_slice_expired(cfqd, cfqq, 0);
3378                 cfq_schedule_dispatch(cfqd);
3379         }
3380
3381         BUG_ON(cfq_cfqq_on_rr(cfqq));
3382         kmem_cache_free(cfq_pool, cfqq);
3383         cfqg_put(cfqg);
3384 }
3385
3386 static void cfq_put_cooperator(struct cfq_queue *cfqq)
3387 {
3388         struct cfq_queue *__cfqq, *next;
3389
3390         /*
3391          * If this queue was scheduled to merge with another queue, be
3392          * sure to drop the reference taken on that queue (and others in
3393          * the merge chain).  See cfq_setup_merge and cfq_merge_cfqqs.
3394          */
3395         __cfqq = cfqq->new_cfqq;
3396         while (__cfqq) {
3397                 if (__cfqq == cfqq) {
3398                         WARN(1, "cfqq->new_cfqq loop detected\n");
3399                         break;
3400                 }
3401                 next = __cfqq->new_cfqq;
3402                 cfq_put_queue(__cfqq);
3403                 __cfqq = next;
3404         }
3405 }
3406
3407 static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3408 {
3409         if (unlikely(cfqq == cfqd->active_queue)) {
3410                 __cfq_slice_expired(cfqd, cfqq, 0);
3411                 cfq_schedule_dispatch(cfqd);
3412         }
3413
3414         cfq_put_cooperator(cfqq);
3415
3416         cfq_put_queue(cfqq);
3417 }
3418
3419 static void cfq_init_icq(struct io_cq *icq)
3420 {
3421         struct cfq_io_cq *cic = icq_to_cic(icq);
3422
3423         cic->ttime.last_end_request = jiffies;
3424 }
3425
3426 static void cfq_exit_icq(struct io_cq *icq)
3427 {
3428         struct cfq_io_cq *cic = icq_to_cic(icq);
3429         struct cfq_data *cfqd = cic_to_cfqd(cic);
3430
3431         if (cic->cfqq[BLK_RW_ASYNC]) {
3432                 cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_ASYNC]);
3433                 cic->cfqq[BLK_RW_ASYNC] = NULL;
3434         }
3435
3436         if (cic->cfqq[BLK_RW_SYNC]) {
3437                 cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_SYNC]);
3438                 cic->cfqq[BLK_RW_SYNC] = NULL;
3439         }
3440 }
3441
3442 static void cfq_init_prio_data(struct cfq_queue *cfqq, struct cfq_io_cq *cic)
3443 {
3444         struct task_struct *tsk = current;
3445         int ioprio_class;
3446
3447         if (!cfq_cfqq_prio_changed(cfqq))
3448                 return;
3449
3450         ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3451         switch (ioprio_class) {
3452         default:
3453                 printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
3454         case IOPRIO_CLASS_NONE:
3455                 /*
3456                  * no prio set, inherit CPU scheduling settings
3457                  */
3458                 cfqq->ioprio = task_nice_ioprio(tsk);
3459                 cfqq->ioprio_class = task_nice_ioclass(tsk);
3460                 break;
3461         case IOPRIO_CLASS_RT:
3462                 cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3463                 cfqq->ioprio_class = IOPRIO_CLASS_RT;
3464                 break;
3465         case IOPRIO_CLASS_BE:
3466                 cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3467                 cfqq->ioprio_class = IOPRIO_CLASS_BE;
3468                 break;
3469         case IOPRIO_CLASS_IDLE:
3470                 cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
3471                 cfqq->ioprio = 7;
3472                 cfq_clear_cfqq_idle_window(cfqq);
3473                 break;
3474         }
3475
3476         /*
3477          * keep track of original prio settings in case we have to temporarily
3478          * elevate the priority of this queue
3479          */
3480         cfqq->org_ioprio = cfqq->ioprio;
3481         cfq_clear_cfqq_prio_changed(cfqq);
3482 }
3483
3484 static void check_ioprio_changed(struct cfq_io_cq *cic, struct bio *bio)
3485 {
3486         int ioprio = cic->icq.ioc->ioprio;
3487         struct cfq_data *cfqd = cic_to_cfqd(cic);
3488         struct cfq_queue *cfqq;
3489
3490         /*
3491          * Check whether ioprio has changed.  The condition may trigger
3492          * spuriously on a newly created cic but there's no harm.
3493          */
3494         if (unlikely(!cfqd) || likely(cic->ioprio == ioprio))
3495                 return;
3496
3497         cfqq = cic->cfqq[BLK_RW_ASYNC];
3498         if (cfqq) {
3499                 struct cfq_queue *new_cfqq;
3500                 new_cfqq = cfq_get_queue(cfqd, BLK_RW_ASYNC, cic, bio,
3501                                          GFP_ATOMIC);
3502                 if (new_cfqq) {
3503                         cic->cfqq[BLK_RW_ASYNC] = new_cfqq;
3504                         cfq_put_queue(cfqq);
3505                 }
3506         }
3507
3508         cfqq = cic->cfqq[BLK_RW_SYNC];
3509         if (cfqq)
3510                 cfq_mark_cfqq_prio_changed(cfqq);
3511
3512         cic->ioprio = ioprio;
3513 }
3514
3515 static void cfq_init_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3516                           pid_t pid, bool is_sync)
3517 {
3518         RB_CLEAR_NODE(&cfqq->rb_node);
3519         RB_CLEAR_NODE(&cfqq->p_node);
3520         INIT_LIST_HEAD(&cfqq->fifo);
3521
3522         cfqq->ref = 0;
3523         cfqq->cfqd = cfqd;
3524
3525         cfq_mark_cfqq_prio_changed(cfqq);
3526
3527         if (is_sync) {
3528                 if (!cfq_class_idle(cfqq))
3529                         cfq_mark_cfqq_idle_window(cfqq);
3530                 cfq_mark_cfqq_sync(cfqq);
3531         }
3532         cfqq->pid = pid;
3533 }
3534
3535 #ifdef CONFIG_CFQ_GROUP_IOSCHED
3536 static void check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio)
3537 {
3538         struct cfq_data *cfqd = cic_to_cfqd(cic);
3539         struct cfq_queue *sync_cfqq;
3540         uint64_t id;
3541
3542         rcu_read_lock();
3543         id = bio_blkcg(bio)->id;
3544         rcu_read_unlock();
3545
3546         /*
3547          * Check whether blkcg has changed.  The condition may trigger
3548          * spuriously on a newly created cic but there's no harm.
3549          */
3550         if (unlikely(!cfqd) || likely(cic->blkcg_id == id))
3551                 return;
3552
3553         sync_cfqq = cic_to_cfqq(cic, 1);
3554         if (sync_cfqq) {
3555                 /*
3556                  * Drop reference to sync queue. A new sync queue will be
3557                  * assigned in new group upon arrival of a fresh request.
3558                  */
3559                 cfq_log_cfqq(cfqd, sync_cfqq, "changed cgroup");
3560                 cic_set_cfqq(cic, NULL, 1);
3561                 cfq_put_queue(sync_cfqq);
3562         }
3563
3564         cic->blkcg_id = id;
3565 }
3566 #else
3567 static inline void check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio) { }
3568 #endif  /* CONFIG_CFQ_GROUP_IOSCHED */
3569
3570 static struct cfq_queue *
3571 cfq_find_alloc_queue(struct cfq_data *cfqd, bool is_sync, struct cfq_io_cq *cic,
3572                      struct bio *bio, gfp_t gfp_mask)
3573 {
3574         struct blkcg *blkcg;
3575         struct cfq_queue *cfqq, *new_cfqq = NULL;
3576         struct cfq_group *cfqg;
3577
3578 retry:
3579         rcu_read_lock();
3580
3581         blkcg = bio_blkcg(bio);
3582         cfqg = cfq_lookup_create_cfqg(cfqd, blkcg);
3583         cfqq = cic_to_cfqq(cic, is_sync);
3584
3585         /*
3586          * Always try a new alloc if we fell back to the OOM cfqq
3587          * originally, since it should just be a temporary situation.
3588          */
3589         if (!cfqq || cfqq == &cfqd->oom_cfqq) {
3590                 cfqq = NULL;
3591                 if (new_cfqq) {
3592                         cfqq = new_cfqq;
3593                         new_cfqq = NULL;
3594                 } else if (gfp_mask & __GFP_WAIT) {
3595                         rcu_read_unlock();
3596                         spin_unlock_irq(cfqd->queue->queue_lock);
3597                         new_cfqq = kmem_cache_alloc_node(cfq_pool,
3598                                         gfp_mask | __GFP_ZERO,
3599                                         cfqd->queue->node);
3600                         spin_lock_irq(cfqd->queue->queue_lock);
3601                         if (new_cfqq)
3602                                 goto retry;
3603                         else
3604                                 return &cfqd->oom_cfqq;
3605                 } else {
3606                         cfqq = kmem_cache_alloc_node(cfq_pool,
3607                                         gfp_mask | __GFP_ZERO,
3608                                         cfqd->queue->node);
3609                 }
3610
3611                 if (cfqq) {
3612                         cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync);
3613                         cfq_init_prio_data(cfqq, cic);
3614                         cfq_link_cfqq_cfqg(cfqq, cfqg);
3615                         cfq_log_cfqq(cfqd, cfqq, "alloced");
3616                 } else
3617                         cfqq = &cfqd->oom_cfqq;
3618         }
3619
3620         if (new_cfqq)
3621                 kmem_cache_free(cfq_pool, new_cfqq);
3622
3623         rcu_read_unlock();
3624         return cfqq;
3625 }
3626
3627 static struct cfq_queue **
3628 cfq_async_queue_prio(struct cfq_data *cfqd, int ioprio_class, int ioprio)
3629 {
3630         switch (ioprio_class) {
3631         case IOPRIO_CLASS_RT:
3632                 return &cfqd->async_cfqq[0][ioprio];
3633         case IOPRIO_CLASS_NONE:
3634                 ioprio = IOPRIO_NORM;
3635                 /* fall through */
3636         case IOPRIO_CLASS_BE:
3637                 return &cfqd->async_cfqq[1][ioprio];
3638         case IOPRIO_CLASS_IDLE:
3639                 return &cfqd->async_idle_cfqq;
3640         default:
3641                 BUG();
3642         }
3643 }
3644
3645 static struct cfq_queue *
3646 cfq_get_queue(struct cfq_data *cfqd, bool is_sync, struct cfq_io_cq *cic,
3647               struct bio *bio, gfp_t gfp_mask)
3648 {
3649         const int ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3650         const int ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3651         struct cfq_queue **async_cfqq = NULL;
3652         struct cfq_queue *cfqq = NULL;
3653
3654         if (!is_sync) {
3655                 async_cfqq = cfq_async_queue_prio(cfqd, ioprio_class, ioprio);
3656                 cfqq = *async_cfqq;
3657         }
3658
3659         if (!cfqq)
3660                 cfqq = cfq_find_alloc_queue(cfqd, is_sync, cic, bio, gfp_mask);
3661
3662         /*
3663          * pin the queue now that it's allocated, scheduler exit will prune it
3664          */
3665         if (!is_sync && !(*async_cfqq)) {
3666                 cfqq->ref++;
3667                 *async_cfqq = cfqq;
3668         }
3669
3670         cfqq->ref++;
3671         return cfqq;
3672 }
3673
3674 static void
3675 __cfq_update_io_thinktime(struct cfq_ttime *ttime, unsigned long slice_idle)
3676 {
3677         unsigned long elapsed = jiffies - ttime->last_end_request;
3678         elapsed = min(elapsed, 2UL * slice_idle);
3679
3680         ttime->ttime_samples = (7*ttime->ttime_samples + 256) / 8;
3681         ttime->ttime_total = (7*ttime->ttime_total + 256*elapsed) / 8;
3682         ttime->ttime_mean = (ttime->ttime_total + 128) / ttime->ttime_samples;
3683 }
3684
3685 static void
3686 cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3687                         struct cfq_io_cq *cic)
3688 {
3689         if (cfq_cfqq_sync(cfqq)) {
3690                 __cfq_update_io_thinktime(&cic->ttime, cfqd->cfq_slice_idle);
3691                 __cfq_update_io_thinktime(&cfqq->service_tree->ttime,
3692                         cfqd->cfq_slice_idle);
3693         }
3694 #ifdef CONFIG_CFQ_GROUP_IOSCHED
3695         __cfq_update_io_thinktime(&cfqq->cfqg->ttime, cfqd->cfq_group_idle);
3696 #endif
3697 }
3698
3699 static void
3700 cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3701                        struct request *rq)
3702 {
3703         sector_t sdist = 0;
3704         sector_t n_sec = blk_rq_sectors(rq);
3705         if (cfqq->last_request_pos) {
3706                 if (cfqq->last_request_pos < blk_rq_pos(rq))
3707                         sdist = blk_rq_pos(rq) - cfqq->last_request_pos;
3708                 else
3709                         sdist = cfqq->last_request_pos - blk_rq_pos(rq);
3710         }
3711
3712         cfqq->seek_history <<= 1;
3713         if (blk_queue_nonrot(cfqd->queue))
3714                 cfqq->seek_history |= (n_sec < CFQQ_SECT_THR_NONROT);
3715         else
3716                 cfqq->seek_history |= (sdist > CFQQ_SEEK_THR);
3717 }
3718
3719 /*
3720  * Disable idle window if the process thinks too long or seeks so much that
3721  * it doesn't matter
3722  */
3723 static void
3724 cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3725                        struct cfq_io_cq *cic)
3726 {
3727         int old_idle, enable_idle;
3728
3729         /*
3730          * Don't idle for async or idle io prio class
3731          */
3732         if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq))
3733                 return;
3734
3735         enable_idle = old_idle = cfq_cfqq_idle_window(cfqq);
3736
3737         if (cfqq->queued[0] + cfqq->queued[1] >= 4)
3738                 cfq_mark_cfqq_deep(cfqq);
3739
3740         if (cfqq->next_rq && (cfqq->next_rq->cmd_flags & REQ_NOIDLE))
3741                 enable_idle = 0;
3742         else if (!atomic_read(&cic->icq.ioc->active_ref) ||
3743                  !cfqd->cfq_slice_idle ||
3744                  (!cfq_cfqq_deep(cfqq) && CFQQ_SEEKY(cfqq)))
3745                 enable_idle = 0;
3746         else if (sample_valid(cic->ttime.ttime_samples)) {
3747                 if (cic->ttime.ttime_mean > cfqd->cfq_slice_idle)
3748                         enable_idle = 0;
3749                 else
3750                         enable_idle = 1;
3751         }
3752
3753         if (old_idle != enable_idle) {
3754                 cfq_log_cfqq(cfqd, cfqq, "idle=%d", enable_idle);
3755                 if (enable_idle)
3756                         cfq_mark_cfqq_idle_window(cfqq);
3757                 else
3758                         cfq_clear_cfqq_idle_window(cfqq);
3759         }
3760 }
3761
3762 /*
3763  * Check if new_cfqq should preempt the currently active queue. Return 0 for
3764  * no or if we aren't sure, a 1 will cause a preempt.
3765  */
3766 static bool
3767 cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
3768                    struct request *rq)
3769 {
3770         struct cfq_queue *cfqq;
3771
3772         cfqq = cfqd->active_queue;
3773         if (!cfqq)
3774                 return false;
3775
3776         if (cfq_class_idle(new_cfqq))
3777                 return false;
3778
3779         if (cfq_class_idle(cfqq))
3780                 return true;
3781
3782         /*
3783          * Don't allow a non-RT request to preempt an ongoing RT cfqq timeslice.
3784          */
3785         if (cfq_class_rt(cfqq) && !cfq_class_rt(new_cfqq))
3786                 return false;
3787
3788         /*
3789          * if the new request is sync, but the currently running queue is
3790          * not, let the sync request have priority.
3791          */
3792         if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq))
3793                 return true;
3794
3795         if (new_cfqq->cfqg != cfqq->cfqg)
3796                 return false;
3797
3798         if (cfq_slice_used(cfqq))
3799                 return true;
3800
3801         /* Allow preemption only if we are idling on sync-noidle tree */
3802         if (cfqd->serving_wl_type == SYNC_NOIDLE_WORKLOAD &&
3803             cfqq_type(new_cfqq) == SYNC_NOIDLE_WORKLOAD &&
3804             new_cfqq->service_tree->count == 2 &&
3805             RB_EMPTY_ROOT(&cfqq->sort_list))
3806                 return true;
3807
3808         /*
3809          * So both queues are sync. Let the new request get disk time if
3810          * it's a metadata request and the current queue is doing regular IO.
3811          */
3812         if ((rq->cmd_flags & REQ_PRIO) && !cfqq->prio_pending)
3813                 return true;
3814
3815         /*
3816          * Allow an RT request to pre-empt an ongoing non-RT cfqq timeslice.
3817          */
3818         if (cfq_class_rt(new_cfqq) && !cfq_class_rt(cfqq))
3819                 return true;
3820
3821         /* An idle queue should not be idle now for some reason */
3822         if (RB_EMPTY_ROOT(&cfqq->sort_list) && !cfq_should_idle(cfqd, cfqq))
3823                 return true;
3824
3825         if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq))
3826                 return false;
3827
3828         /*
3829          * if this request is as-good as one we would expect from the
3830          * current cfqq, let it preempt
3831          */
3832         if (cfq_rq_close(cfqd, cfqq, rq))
3833                 return true;
3834
3835         return false;
3836 }
3837
3838 /*
3839  * cfqq preempts the active queue. if we allowed preempt with no slice left,
3840  * let it have half of its nominal slice.
3841  */
3842 static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3843 {
3844         enum wl_type_t old_type = cfqq_type(cfqd->active_queue);
3845
3846         cfq_log_cfqq(cfqd, cfqq, "preempt");
3847         cfq_slice_expired(cfqd, 1);
3848
3849         /*
3850          * workload type is changed, don't save slice, otherwise preempt
3851          * doesn't happen
3852          */
3853         if (old_type != cfqq_type(cfqq))
3854                 cfqq->cfqg->saved_wl_slice = 0;
3855
3856         /*
3857          * Put the new queue at the front of the of the current list,
3858          * so we know that it will be selected next.
3859          */
3860         BUG_ON(!cfq_cfqq_on_rr(cfqq));
3861
3862         cfq_service_tree_add(cfqd, cfqq, 1);
3863
3864         cfqq->slice_end = 0;
3865         cfq_mark_cfqq_slice_new(cfqq);
3866 }
3867
3868 /*
3869  * Called when a new fs request (rq) is added (to cfqq). Check if there's
3870  * something we should do about it
3871  */
3872 static void
3873 cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3874                 struct request *rq)
3875 {
3876         struct cfq_io_cq *cic = RQ_CIC(rq);
3877
3878         cfqd->rq_queued++;
3879         if (rq->cmd_flags & REQ_PRIO)
3880                 cfqq->prio_pending++;
3881
3882         cfq_update_io_thinktime(cfqd, cfqq, cic);
3883         cfq_update_io_seektime(cfqd, cfqq, rq);
3884         cfq_update_idle_window(cfqd, cfqq, cic);
3885
3886         cfqq->last_request_pos = blk_rq_pos(rq) + blk_rq_sectors(rq);
3887
3888         if (cfqq == cfqd->active_queue) {
3889                 /*
3890                  * Remember that we saw a request from this process, but
3891                  * don't start queuing just yet. Otherwise we risk seeing lots
3892                  * of tiny requests, because we disrupt the normal plugging
3893                  * and merging. If the request is already larger than a single
3894                  * page, let it rip immediately. For that case we assume that
3895                  * merging is already done. Ditto for a busy system that
3896                  * has other work pending, don't risk delaying until the
3897                  * idle timer unplug to continue working.
3898                  */
3899                 if (cfq_cfqq_wait_request(cfqq)) {
3900                         if (blk_rq_bytes(rq) > PAGE_CACHE_SIZE ||
3901                             cfqd->busy_queues > 1) {
3902                                 cfq_del_timer(cfqd, cfqq);
3903                                 cfq_clear_cfqq_wait_request(cfqq);
3904                                 __blk_run_queue(cfqd->queue);
3905                         } else {
3906                                 cfqg_stats_update_idle_time(cfqq->cfqg);
3907                                 cfq_mark_cfqq_must_dispatch(cfqq);
3908                         }
3909                 }
3910         } else if (cfq_should_preempt(cfqd, cfqq, rq)) {
3911                 /*
3912                  * not the active queue - expire current slice if it is
3913                  * idle and has expired it's mean thinktime or this new queue
3914                  * has some old slice time left and is of higher priority or
3915                  * this new queue is RT and the current one is BE
3916                  */
3917                 cfq_preempt_queue(cfqd, cfqq);
3918                 __blk_run_queue(cfqd->queue);
3919         }
3920 }
3921
3922 static void cfq_insert_request(struct request_queue *q, struct request *rq)
3923 {
3924         struct cfq_data *cfqd = q->elevator->elevator_data;
3925         struct cfq_queue *cfqq = RQ_CFQQ(rq);
3926
3927         cfq_log_cfqq(cfqd, cfqq, "insert_request");
3928         cfq_init_prio_data(cfqq, RQ_CIC(rq));
3929
3930         rq_set_fifo_time(rq, jiffies + cfqd->cfq_fifo_expire[rq_is_sync(rq)]);
3931         list_add_tail(&rq->queuelist, &cfqq->fifo);
3932         cfq_add_rq_rb(rq);
3933         cfqg_stats_update_io_add(RQ_CFQG(rq), cfqd->serving_group,
3934                                  rq->cmd_flags);
3935         cfq_rq_enqueued(cfqd, cfqq, rq);
3936 }
3937
3938 /*
3939  * Update hw_tag based on peak queue depth over 50 samples under
3940  * sufficient load.
3941  */
3942 static void cfq_update_hw_tag(struct cfq_data *cfqd)
3943 {
3944         struct cfq_queue *cfqq = cfqd->active_queue;
3945
3946         if (cfqd->rq_in_driver > cfqd->hw_tag_est_depth)
3947                 cfqd->hw_tag_est_depth = cfqd->rq_in_driver;
3948
3949         if (cfqd->hw_tag == 1)
3950                 return;
3951
3952         if (cfqd->rq_queued <= CFQ_HW_QUEUE_MIN &&
3953             cfqd->rq_in_driver <= CFQ_HW_QUEUE_MIN)
3954                 return;
3955
3956         /*
3957          * If active queue hasn't enough requests and can idle, cfq might not
3958          * dispatch sufficient requests to hardware. Don't zero hw_tag in this
3959          * case
3960          */
3961         if (cfqq && cfq_cfqq_idle_window(cfqq) &&
3962             cfqq->dispatched + cfqq->queued[0] + cfqq->queued[1] <
3963             CFQ_HW_QUEUE_MIN && cfqd->rq_in_driver < CFQ_HW_QUEUE_MIN)
3964                 return;
3965
3966         if (cfqd->hw_tag_samples++ < 50)
3967                 return;
3968
3969         if (cfqd->hw_tag_est_depth >= CFQ_HW_QUEUE_MIN)
3970                 cfqd->hw_tag = 1;
3971         else
3972                 cfqd->hw_tag = 0;
3973 }
3974
3975 static bool cfq_should_wait_busy(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3976 {
3977         struct cfq_io_cq *cic = cfqd->active_cic;
3978
3979         /* If the queue already has requests, don't wait */
3980         if (!RB_EMPTY_ROOT(&cfqq->sort_list))
3981                 return false;
3982
3983         /* If there are other queues in the group, don't wait */
3984         if (cfqq->cfqg->nr_cfqq > 1)
3985                 return false;
3986
3987         /* the only queue in the group, but think time is big */
3988         if (cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true))
3989                 return false;
3990
3991         if (cfq_slice_used(cfqq))
3992                 return true;
3993
3994         /* if slice left is less than think time, wait busy */
3995         if (cic && sample_valid(cic->ttime.ttime_samples)
3996             && (cfqq->slice_end - jiffies < cic->ttime.ttime_mean))
3997                 return true;
3998
3999         /*
4000          * If think times is less than a jiffy than ttime_mean=0 and above
4001          * will not be true. It might happen that slice has not expired yet
4002          * but will expire soon (4-5 ns) during select_queue(). To cover the
4003          * case where think time is less than a jiffy, mark the queue wait
4004          * busy if only 1 jiffy is left in the slice.
4005          */
4006         if (cfqq->slice_end - jiffies == 1)
4007                 return true;
4008
4009         return false;
4010 }
4011
4012 static void cfq_completed_request(struct request_queue *q, struct request *rq)
4013 {
4014         struct cfq_queue *cfqq = RQ_CFQQ(rq);
4015         struct cfq_data *cfqd = cfqq->cfqd;
4016         const int sync = rq_is_sync(rq);
4017         unsigned long now;
4018
4019         now = jiffies;
4020         cfq_log_cfqq(cfqd, cfqq, "complete rqnoidle %d",
4021                      !!(rq->cmd_flags & REQ_NOIDLE));
4022
4023         cfq_update_hw_tag(cfqd);
4024
4025         WARN_ON(!cfqd->rq_in_driver);
4026         WARN_ON(!cfqq->dispatched);
4027         cfqd->rq_in_driver--;
4028         cfqq->dispatched--;
4029         (RQ_CFQG(rq))->dispatched--;
4030         cfqg_stats_update_completion(cfqq->cfqg, rq_start_time_ns(rq),
4031                                      rq_io_start_time_ns(rq), rq->cmd_flags);
4032
4033         cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]--;
4034
4035         if (sync) {
4036                 struct cfq_rb_root *st;
4037
4038                 RQ_CIC(rq)->ttime.last_end_request = now;
4039
4040                 if (cfq_cfqq_on_rr(cfqq))
4041                         st = cfqq->service_tree;
4042                 else
4043                         st = st_for(cfqq->cfqg, cfqq_class(cfqq),
4044                                         cfqq_type(cfqq));
4045
4046                 st->ttime.last_end_request = now;
4047                 if (!time_after(rq->start_time + cfqd->cfq_fifo_expire[1], now))
4048                         cfqd->last_delayed_sync = now;
4049         }
4050
4051 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4052         cfqq->cfqg->ttime.last_end_request = now;
4053 #endif
4054
4055         /*
4056          * If this is the active queue, check if it needs to be expired,
4057          * or if we want to idle in case it has no pending requests.
4058          */
4059         if (cfqd->active_queue == cfqq) {
4060                 const bool cfqq_empty = RB_EMPTY_ROOT(&cfqq->sort_list);
4061
4062                 if (cfq_cfqq_slice_new(cfqq)) {
4063                         cfq_set_prio_slice(cfqd, cfqq);
4064                         cfq_clear_cfqq_slice_new(cfqq);
4065                 }
4066
4067                 /*
4068                  * Should we wait for next request to come in before we expire
4069                  * the queue.
4070                  */
4071                 if (cfq_should_wait_busy(cfqd, cfqq)) {
4072                         unsigned long extend_sl = cfqd->cfq_slice_idle;
4073                         if (!cfqd->cfq_slice_idle)
4074                                 extend_sl = cfqd->cfq_group_idle;
4075                         cfqq->slice_end = jiffies + extend_sl;
4076                         cfq_mark_cfqq_wait_busy(cfqq);
4077                         cfq_log_cfqq(cfqd, cfqq, "will busy wait");
4078                 }
4079
4080                 /*
4081                  * Idling is not enabled on:
4082                  * - expired queues
4083                  * - idle-priority queues
4084                  * - async queues
4085                  * - queues with still some requests queued
4086                  * - when there is a close cooperator
4087                  */
4088                 if (cfq_slice_used(cfqq) || cfq_class_idle(cfqq))
4089                         cfq_slice_expired(cfqd, 1);
4090                 else if (sync && cfqq_empty &&
4091                          !cfq_close_cooperator(cfqd, cfqq)) {
4092                         cfq_arm_slice_timer(cfqd);
4093                 }
4094         }
4095
4096         if (!cfqd->rq_in_driver)
4097                 cfq_schedule_dispatch(cfqd);
4098 }
4099
4100 static inline int __cfq_may_queue(struct cfq_queue *cfqq)
4101 {
4102         if (cfq_cfqq_wait_request(cfqq) && !cfq_cfqq_must_alloc_slice(cfqq)) {
4103                 cfq_mark_cfqq_must_alloc_slice(cfqq);
4104                 return ELV_MQUEUE_MUST;
4105         }
4106
4107         return ELV_MQUEUE_MAY;
4108 }
4109
4110 static int cfq_may_queue(struct request_queue *q, int rw)
4111 {
4112         struct cfq_data *cfqd = q->elevator->elevator_data;
4113         struct task_struct *tsk = current;
4114         struct cfq_io_cq *cic;
4115         struct cfq_queue *cfqq;
4116
4117         /*
4118          * don't force setup of a queue from here, as a call to may_queue
4119          * does not necessarily imply that a request actually will be queued.
4120          * so just lookup a possibly existing queue, or return 'may queue'
4121          * if that fails
4122          */
4123         cic = cfq_cic_lookup(cfqd, tsk->io_context);
4124         if (!cic)
4125                 return ELV_MQUEUE_MAY;
4126
4127         cfqq = cic_to_cfqq(cic, rw_is_sync(rw));
4128         if (cfqq) {
4129                 cfq_init_prio_data(cfqq, cic);
4130
4131                 return __cfq_may_queue(cfqq);
4132         }
4133
4134         return ELV_MQUEUE_MAY;
4135 }
4136
4137 /*
4138  * queue lock held here
4139  */
4140 static void cfq_put_request(struct request *rq)
4141 {
4142         struct cfq_queue *cfqq = RQ_CFQQ(rq);
4143
4144         if (cfqq) {
4145                 const int rw = rq_data_dir(rq);
4146
4147                 BUG_ON(!cfqq->allocated[rw]);
4148                 cfqq->allocated[rw]--;
4149
4150                 /* Put down rq reference on cfqg */
4151                 cfqg_put(RQ_CFQG(rq));
4152                 rq->elv.priv[0] = NULL;
4153                 rq->elv.priv[1] = NULL;
4154
4155                 cfq_put_queue(cfqq);
4156         }
4157 }
4158
4159 static struct cfq_queue *
4160 cfq_merge_cfqqs(struct cfq_data *cfqd, struct cfq_io_cq *cic,
4161                 struct cfq_queue *cfqq)
4162 {
4163         cfq_log_cfqq(cfqd, cfqq, "merging with queue %p", cfqq->new_cfqq);
4164         cic_set_cfqq(cic, cfqq->new_cfqq, 1);
4165         cfq_mark_cfqq_coop(cfqq->new_cfqq);
4166         cfq_put_queue(cfqq);
4167         return cic_to_cfqq(cic, 1);
4168 }
4169
4170 /*
4171  * Returns NULL if a new cfqq should be allocated, or the old cfqq if this
4172  * was the last process referring to said cfqq.
4173  */
4174 static struct cfq_queue *
4175 split_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq)
4176 {
4177         if (cfqq_process_refs(cfqq) == 1) {
4178                 cfqq->pid = current->pid;
4179                 cfq_clear_cfqq_coop(cfqq);
4180                 cfq_clear_cfqq_split_coop(cfqq);
4181                 return cfqq;
4182         }
4183
4184         cic_set_cfqq(cic, NULL, 1);
4185
4186         cfq_put_cooperator(cfqq);
4187
4188         cfq_put_queue(cfqq);
4189         return NULL;
4190 }
4191 /*
4192  * Allocate cfq data structures associated with this request.
4193  */
4194 static int
4195 cfq_set_request(struct request_queue *q, struct request *rq, struct bio *bio,
4196                 gfp_t gfp_mask)
4197 {
4198         struct cfq_data *cfqd = q->elevator->elevator_data;
4199         struct cfq_io_cq *cic = icq_to_cic(rq->elv.icq);
4200         const int rw = rq_data_dir(rq);
4201         const bool is_sync = rq_is_sync(rq);
4202         struct cfq_queue *cfqq;
4203
4204         might_sleep_if(gfp_mask & __GFP_WAIT);
4205
4206         spin_lock_irq(q->queue_lock);
4207
4208         check_ioprio_changed(cic, bio);
4209         check_blkcg_changed(cic, bio);
4210 new_queue:
4211         cfqq = cic_to_cfqq(cic, is_sync);
4212         if (!cfqq || cfqq == &cfqd->oom_cfqq) {
4213                 cfqq = cfq_get_queue(cfqd, is_sync, cic, bio, gfp_mask);
4214                 cic_set_cfqq(cic, cfqq, is_sync);
4215         } else {
4216                 /*
4217                  * If the queue was seeky for too long, break it apart.
4218                  */
4219                 if (cfq_cfqq_coop(cfqq) && cfq_cfqq_split_coop(cfqq)) {
4220                         cfq_log_cfqq(cfqd, cfqq, "breaking apart cfqq");
4221                         cfqq = split_cfqq(cic, cfqq);
4222                         if (!cfqq)
4223                                 goto new_queue;
4224                 }
4225
4226                 /*
4227                  * Check to see if this queue is scheduled to merge with
4228                  * another, closely cooperating queue.  The merging of
4229                  * queues happens here as it must be done in process context.
4230                  * The reference on new_cfqq was taken in merge_cfqqs.
4231                  */
4232                 if (cfqq->new_cfqq)
4233                         cfqq = cfq_merge_cfqqs(cfqd, cic, cfqq);
4234         }
4235
4236         cfqq->allocated[rw]++;
4237
4238         cfqq->ref++;
4239         cfqg_get(cfqq->cfqg);
4240         rq->elv.priv[0] = cfqq;
4241         rq->elv.priv[1] = cfqq->cfqg;
4242         spin_unlock_irq(q->queue_lock);
4243         return 0;
4244 }
4245
4246 static void cfq_kick_queue(struct work_struct *work)
4247 {
4248         struct cfq_data *cfqd =
4249                 container_of(work, struct cfq_data, unplug_work);
4250         struct request_queue *q = cfqd->queue;
4251
4252         spin_lock_irq(q->queue_lock);
4253         __blk_run_queue(cfqd->queue);
4254         spin_unlock_irq(q->queue_lock);
4255 }
4256
4257 /*
4258  * Timer running if the active_queue is currently idling inside its time slice
4259  */
4260 static void cfq_idle_slice_timer(unsigned long data)
4261 {
4262         struct cfq_data *cfqd = (struct cfq_data *) data;
4263         struct cfq_queue *cfqq;
4264         unsigned long flags;
4265         int timed_out = 1;
4266
4267         cfq_log(cfqd, "idle timer fired");
4268
4269         spin_lock_irqsave(cfqd->queue->queue_lock, flags);
4270
4271         cfqq = cfqd->active_queue;
4272         if (cfqq) {
4273                 timed_out = 0;
4274
4275                 /*
4276                  * We saw a request before the queue expired, let it through
4277                  */
4278                 if (cfq_cfqq_must_dispatch(cfqq))
4279                         goto out_kick;
4280
4281                 /*
4282                  * expired
4283                  */
4284                 if (cfq_slice_used(cfqq))
4285                         goto expire;
4286
4287                 /*
4288                  * only expire and reinvoke request handler, if there are
4289                  * other queues with pending requests
4290                  */
4291                 if (!cfqd->busy_queues)
4292                         goto out_cont;
4293
4294                 /*
4295                  * not expired and it has a request pending, let it dispatch
4296                  */
4297                 if (!RB_EMPTY_ROOT(&cfqq->sort_list))
4298                         goto out_kick;
4299
4300                 /*
4301                  * Queue depth flag is reset only when the idle didn't succeed
4302                  */
4303                 cfq_clear_cfqq_deep(cfqq);
4304         }
4305 expire:
4306         cfq_slice_expired(cfqd, timed_out);
4307 out_kick:
4308         cfq_schedule_dispatch(cfqd);
4309 out_cont:
4310         spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
4311 }
4312
4313 static void cfq_shutdown_timer_wq(struct cfq_data *cfqd)
4314 {
4315         del_timer_sync(&cfqd->idle_slice_timer);
4316         cancel_work_sync(&cfqd->unplug_work);
4317 }
4318
4319 static void cfq_put_async_queues(struct cfq_data *cfqd)
4320 {
4321         int i;
4322
4323         for (i = 0; i < IOPRIO_BE_NR; i++) {
4324                 if (cfqd->async_cfqq[0][i])
4325                         cfq_put_queue(cfqd->async_cfqq[0][i]);
4326                 if (cfqd->async_cfqq[1][i])
4327                         cfq_put_queue(cfqd->async_cfqq[1][i]);
4328         }
4329
4330         if (cfqd->async_idle_cfqq)
4331                 cfq_put_queue(cfqd->async_idle_cfqq);
4332 }
4333
4334 static void cfq_exit_queue(struct elevator_queue *e)
4335 {
4336         struct cfq_data *cfqd = e->elevator_data;
4337         struct request_queue *q = cfqd->queue;
4338
4339         cfq_shutdown_timer_wq(cfqd);
4340
4341         spin_lock_irq(q->queue_lock);
4342
4343         if (cfqd->active_queue)
4344                 __cfq_slice_expired(cfqd, cfqd->active_queue, 0);
4345
4346         cfq_put_async_queues(cfqd);
4347
4348         spin_unlock_irq(q->queue_lock);
4349
4350         cfq_shutdown_timer_wq(cfqd);
4351
4352 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4353         blkcg_deactivate_policy(q, &blkcg_policy_cfq);
4354 #else
4355         kfree(cfqd->root_group);
4356 #endif
4357         kfree(cfqd);
4358 }
4359
4360 static int cfq_init_queue(struct request_queue *q, struct elevator_type *e)
4361 {
4362         struct cfq_data *cfqd;
4363         struct blkcg_gq *blkg __maybe_unused;
4364         int i, ret;
4365         struct elevator_queue *eq;
4366
4367         eq = elevator_alloc(q, e);
4368         if (!eq)
4369                 return -ENOMEM;
4370
4371         cfqd = kzalloc_node(sizeof(*cfqd), GFP_KERNEL, q->node);
4372         if (!cfqd) {
4373                 kobject_put(&eq->kobj);
4374                 return -ENOMEM;
4375         }
4376         eq->elevator_data = cfqd;
4377
4378         cfqd->queue = q;
4379         spin_lock_irq(q->queue_lock);
4380         q->elevator = eq;
4381         spin_unlock_irq(q->queue_lock);
4382
4383         /* Init root service tree */
4384         cfqd->grp_service_tree = CFQ_RB_ROOT;
4385
4386         /* Init root group and prefer root group over other groups by default */
4387 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4388         ret = blkcg_activate_policy(q, &blkcg_policy_cfq);
4389         if (ret)
4390                 goto out_free;
4391
4392         cfqd->root_group = blkg_to_cfqg(q->root_blkg);
4393 #else
4394         ret = -ENOMEM;
4395         cfqd->root_group = kzalloc_node(sizeof(*cfqd->root_group),
4396                                         GFP_KERNEL, cfqd->queue->node);
4397         if (!cfqd->root_group)
4398                 goto out_free;
4399
4400         cfq_init_cfqg_base(cfqd->root_group);
4401 #endif
4402         cfqd->root_group->weight = 2 * CFQ_WEIGHT_DEFAULT;
4403         cfqd->root_group->leaf_weight = 2 * CFQ_WEIGHT_DEFAULT;
4404
4405         /*
4406          * Not strictly needed (since RB_ROOT just clears the node and we
4407          * zeroed cfqd on alloc), but better be safe in case someone decides
4408          * to add magic to the rb code
4409          */
4410         for (i = 0; i < CFQ_PRIO_LISTS; i++)
4411                 cfqd->prio_trees[i] = RB_ROOT;
4412
4413         /*
4414          * Our fallback cfqq if cfq_find_alloc_queue() runs into OOM issues.
4415          * Grab a permanent reference to it, so that the normal code flow
4416          * will not attempt to free it.  oom_cfqq is linked to root_group
4417          * but shouldn't hold a reference as it'll never be unlinked.  Lose
4418          * the reference from linking right away.
4419          */
4420         cfq_init_cfqq(cfqd, &cfqd->oom_cfqq, 1, 0);
4421         cfqd->oom_cfqq.ref++;
4422
4423         spin_lock_irq(q->queue_lock);
4424         cfq_link_cfqq_cfqg(&cfqd->oom_cfqq, cfqd->root_group);
4425         cfqg_put(cfqd->root_group);
4426         spin_unlock_irq(q->queue_lock);
4427
4428         init_timer(&cfqd->idle_slice_timer);
4429         cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
4430         cfqd->idle_slice_timer.data = (unsigned long) cfqd;
4431
4432         INIT_WORK(&cfqd->unplug_work, cfq_kick_queue);
4433
4434         cfqd->cfq_quantum = cfq_quantum;
4435         cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0];
4436         cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1];
4437         cfqd->cfq_back_max = cfq_back_max;
4438         cfqd->cfq_back_penalty = cfq_back_penalty;
4439         cfqd->cfq_slice[0] = cfq_slice_async;
4440         cfqd->cfq_slice[1] = cfq_slice_sync;
4441         cfqd->cfq_target_latency = cfq_target_latency;
4442         cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
4443         cfqd->cfq_slice_idle = cfq_slice_idle;
4444         cfqd->cfq_group_idle = cfq_group_idle;
4445         cfqd->cfq_latency = 1;
4446         cfqd->hw_tag = -1;
4447         /*
4448          * we optimistically start assuming sync ops weren't delayed in last
4449          * second, in order to have larger depth for async operations.
4450          */
4451         cfqd->last_delayed_sync = jiffies - HZ;
4452         return 0;
4453
4454 out_free:
4455         kfree(cfqd);
4456         kobject_put(&eq->kobj);
4457         return ret;
4458 }
4459
4460 /*
4461  * sysfs parts below -->
4462  */
4463 static ssize_t
4464 cfq_var_show(unsigned int var, char *page)
4465 {
4466         return sprintf(page, "%d\n", var);
4467 }
4468
4469 static ssize_t
4470 cfq_var_store(unsigned int *var, const char *page, size_t count)
4471 {
4472         char *p = (char *) page;
4473
4474         *var = simple_strtoul(p, &p, 10);
4475         return count;
4476 }
4477
4478 #define SHOW_FUNCTION(__FUNC, __VAR, __CONV)                            \
4479 static ssize_t __FUNC(struct elevator_queue *e, char *page)             \
4480 {                                                                       \
4481         struct cfq_data *cfqd = e->elevator_data;                       \
4482         unsigned int __data = __VAR;                                    \
4483         if (__CONV)                                                     \
4484                 __data = jiffies_to_msecs(__data);                      \
4485         return cfq_var_show(__data, (page));                            \
4486 }
4487 SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0);
4488 SHOW_FUNCTION(cfq_fifo_expire_sync_show, cfqd->cfq_fifo_expire[1], 1);
4489 SHOW_FUNCTION(cfq_fifo_expire_async_show, cfqd->cfq_fifo_expire[0], 1);
4490 SHOW_FUNCTION(cfq_back_seek_max_show, cfqd->cfq_back_max, 0);
4491 SHOW_FUNCTION(cfq_back_seek_penalty_show, cfqd->cfq_back_penalty, 0);
4492 SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1);
4493 SHOW_FUNCTION(cfq_group_idle_show, cfqd->cfq_group_idle, 1);
4494 SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1);
4495 SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
4496 SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
4497 SHOW_FUNCTION(cfq_low_latency_show, cfqd->cfq_latency, 0);
4498 SHOW_FUNCTION(cfq_target_latency_show, cfqd->cfq_target_latency, 1);
4499 #undef SHOW_FUNCTION
4500
4501 #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV)                 \
4502 static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count) \
4503 {                                                                       \
4504         struct cfq_data *cfqd = e->elevator_data;                       \
4505         unsigned int __data;                                            \
4506         int ret = cfq_var_store(&__data, (page), count);                \
4507         if (__data < (MIN))                                             \
4508                 __data = (MIN);                                         \
4509         else if (__data > (MAX))                                        \
4510                 __data = (MAX);                                         \
4511         if (__CONV)                                                     \
4512                 *(__PTR) = msecs_to_jiffies(__data);                    \
4513         else                                                            \
4514                 *(__PTR) = __data;                                      \
4515         return ret;                                                     \
4516 }
4517 STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
4518 STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1,
4519                 UINT_MAX, 1);
4520 STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1,
4521                 UINT_MAX, 1);
4522 STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
4523 STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1,
4524                 UINT_MAX, 0);
4525 STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1);
4526 STORE_FUNCTION(cfq_group_idle_store, &cfqd->cfq_group_idle, 0, UINT_MAX, 1);
4527 STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1);
4528 STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1);
4529 STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1,
4530                 UINT_MAX, 0);
4531 STORE_FUNCTION(cfq_low_latency_store, &cfqd->cfq_latency, 0, 1, 0);
4532 STORE_FUNCTION(cfq_target_latency_store, &cfqd->cfq_target_latency, 1, UINT_MAX, 1);
4533 #undef STORE_FUNCTION
4534
4535 #define CFQ_ATTR(name) \
4536         __ATTR(name, S_IRUGO|S_IWUSR, cfq_##name##_show, cfq_##name##_store)
4537
4538 static struct elv_fs_entry cfq_attrs[] = {
4539         CFQ_ATTR(quantum),
4540         CFQ_ATTR(fifo_expire_sync),
4541         CFQ_ATTR(fifo_expire_async),
4542         CFQ_ATTR(back_seek_max),
4543         CFQ_ATTR(back_seek_penalty),
4544         CFQ_ATTR(slice_sync),
4545         CFQ_ATTR(slice_async),
4546         CFQ_ATTR(slice_async_rq),
4547         CFQ_ATTR(slice_idle),
4548         CFQ_ATTR(group_idle),
4549         CFQ_ATTR(low_latency),
4550         CFQ_ATTR(target_latency),
4551         __ATTR_NULL
4552 };
4553
4554 static struct elevator_type iosched_cfq = {
4555         .ops = {
4556                 .elevator_merge_fn =            cfq_merge,
4557                 .elevator_merged_fn =           cfq_merged_request,
4558                 .elevator_merge_req_fn =        cfq_merged_requests,
4559                 .elevator_allow_merge_fn =      cfq_allow_merge,
4560                 .elevator_bio_merged_fn =       cfq_bio_merged,
4561                 .elevator_dispatch_fn =         cfq_dispatch_requests,
4562                 .elevator_add_req_fn =          cfq_insert_request,
4563                 .elevator_activate_req_fn =     cfq_activate_request,
4564                 .elevator_deactivate_req_fn =   cfq_deactivate_request,
4565                 .elevator_completed_req_fn =    cfq_completed_request,
4566                 .elevator_former_req_fn =       elv_rb_former_request,
4567                 .elevator_latter_req_fn =       elv_rb_latter_request,
4568                 .elevator_init_icq_fn =         cfq_init_icq,
4569                 .elevator_exit_icq_fn =         cfq_exit_icq,
4570                 .elevator_set_req_fn =          cfq_set_request,
4571                 .elevator_put_req_fn =          cfq_put_request,
4572                 .elevator_may_queue_fn =        cfq_may_queue,
4573                 .elevator_init_fn =             cfq_init_queue,
4574                 .elevator_exit_fn =             cfq_exit_queue,
4575         },
4576         .icq_size       =       sizeof(struct cfq_io_cq),
4577         .icq_align      =       __alignof__(struct cfq_io_cq),
4578         .elevator_attrs =       cfq_attrs,
4579         .elevator_name  =       "cfq",
4580         .elevator_owner =       THIS_MODULE,
4581 };
4582
4583 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4584 static struct blkcg_policy blkcg_policy_cfq = {
4585         .pd_size                = sizeof(struct cfq_group),
4586         .cftypes                = cfq_blkcg_files,
4587
4588         .pd_init_fn             = cfq_pd_init,
4589         .pd_offline_fn          = cfq_pd_offline,
4590         .pd_reset_stats_fn      = cfq_pd_reset_stats,
4591 };
4592 #endif
4593
4594 static int __init cfq_init(void)
4595 {
4596         int ret;
4597
4598         /*
4599          * could be 0 on HZ < 1000 setups
4600          */
4601         if (!cfq_slice_async)
4602                 cfq_slice_async = 1;
4603         if (!cfq_slice_idle)
4604                 cfq_slice_idle = 1;
4605
4606 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4607         if (!cfq_group_idle)
4608                 cfq_group_idle = 1;
4609
4610         ret = blkcg_policy_register(&blkcg_policy_cfq);
4611         if (ret)
4612                 return ret;
4613 #else
4614         cfq_group_idle = 0;
4615 #endif
4616
4617         ret = -ENOMEM;
4618         cfq_pool = KMEM_CACHE(cfq_queue, 0);
4619         if (!cfq_pool)
4620                 goto err_pol_unreg;
4621
4622         ret = elv_register(&iosched_cfq);
4623         if (ret)
4624                 goto err_free_pool;
4625
4626         return 0;
4627
4628 err_free_pool:
4629         kmem_cache_destroy(cfq_pool);
4630 err_pol_unreg:
4631 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4632         blkcg_policy_unregister(&blkcg_policy_cfq);
4633 #endif
4634         return ret;
4635 }
4636
4637 static void __exit cfq_exit(void)
4638 {
4639 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4640         blkcg_policy_unregister(&blkcg_policy_cfq);
4641 #endif
4642         elv_unregister(&iosched_cfq);
4643         kmem_cache_destroy(cfq_pool);
4644 }
4645
4646 module_init(cfq_init);
4647 module_exit(cfq_exit);
4648
4649 MODULE_AUTHOR("Jens Axboe");
4650 MODULE_LICENSE("GPL");
4651 MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler");