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 /* blkio-related constants */
  71 #define CFQ_WEIGHT_MIN          10
  72 #define CFQ_WEIGHT_MAX          1000
  73 #define CFQ_WEIGHT_DEFAULT      500
  74
  75 struct cfq_ttime {
  76         unsigned long last_end_request;
  77
  78         unsigned long ttime_total;
  79         unsigned long ttime_samples;
  80         unsigned long ttime_mean;
  81 };
  82
  83 /*
  84  * Most of our rbtree usage is for sorting with min extraction, so
  85  * if we cache the leftmost node we don't have to walk down the tree
  86  * to find it. Idea borrowed from Ingo Molnars CFS scheduler. We should
  87  * move this into the elevator for the rq sorting as well.
  88  */
  89 struct cfq_rb_root {
  90         struct rb_root rb;
  91         struct rb_node *left;
  92         unsigned count;
  93         u64 min_vdisktime;
  94         struct cfq_ttime ttime;
  95 };
  96 #define CFQ_RB_ROOT     (struct cfq_rb_root) { .rb = RB_ROOT, \
  97                         .ttime = {.last_end_request = jiffies,},}
  98
  99 /*
 100  * Per process-grouping structure
 101  */
 102 struct cfq_queue {
 103         /* reference count */
 104         int ref;
 105         /* various state flags, see below */
 106         unsigned int flags;
 107         /* parent cfq_data */
 108         struct cfq_data *cfqd;
 109         /* service_tree member */
 110         struct rb_node rb_node;
 111         /* service_tree key */
 112         unsigned long rb_key;
 113         /* prio tree member */
 114         struct rb_node p_node;
 115         /* prio tree root we belong to, if any */
 116         struct rb_root *p_root;
 117         /* sorted list of pending requests */
 118         struct rb_root sort_list;
 119         /* if fifo isn't expired, next request to serve */
 120         struct request *next_rq;
 121         /* requests queued in sort_list */
 122         int queued[2];
 123         /* currently allocated requests */
 124         int allocated[2];
 125         /* fifo list of requests in sort_list */
 126         struct list_head fifo;
 127
 128         /* time when queue got scheduled in to dispatch first request. */
 129         unsigned long dispatch_start;
 130         unsigned int allocated_slice;
 131         unsigned int slice_dispatch;
 132         /* time when first request from queue completed and slice started. */
 133         unsigned long slice_start;
 134         unsigned long slice_end;
 135         long slice_resid;
 136
 137         /* pending priority requests */
 138         int prio_pending;
 139         /* number of requests that are on the dispatch list or inside driver */
 140         int dispatched;
 141
 142         /* io prio of this group */
 143         unsigned short ioprio, org_ioprio;
 144         unsigned short ioprio_class;
 145
 146         pid_t pid;
 147
 148         u32 seek_history;
 149         sector_t last_request_pos;
 150
 151         struct cfq_rb_root *service_tree;
 152         struct cfq_queue *new_cfqq;
 153         struct cfq_group *cfqg;
 154         /* Number of sectors dispatched from queue in single dispatch round */
 155         unsigned long nr_sectors;
 156 };
 157
 158 /*
 159  * First index in the service_trees.
 160  * IDLE is handled separately, so it has negative index
 161  */
 162 enum wl_class_t {
 163         BE_WORKLOAD = 0,
 164         RT_WORKLOAD = 1,
 165         IDLE_WORKLOAD = 2,
 166         CFQ_PRIO_NR,
 167 };
 168
 169 /*
 170  * Second index in the service_trees.
 171  */
 172 enum wl_type_t {
 173         ASYNC_WORKLOAD = 0,
 174         SYNC_NOIDLE_WORKLOAD = 1,
 175         SYNC_WORKLOAD = 2
 176 };
 177
 178 struct cfqg_stats {
 179 #ifdef CONFIG_CFQ_GROUP_IOSCHED
 180         /* total bytes transferred */
 181         struct blkg_rwstat              service_bytes;
 182         /* total IOs serviced, post merge */
 183         struct blkg_rwstat              serviced;
 184         /* number of ios merged */
 185         struct blkg_rwstat              merged;
 186         /* total time spent on device in ns, may not be accurate w/ queueing */
 187         struct blkg_rwstat              service_time;
 188         /* total time spent waiting in scheduler queue in ns */
 189         struct blkg_rwstat              wait_time;
 190         /* number of IOs queued up */
 191         struct blkg_rwstat              queued;
 192         /* total sectors transferred */
 193         struct blkg_stat                sectors;
 194         /* total disk time and nr sectors dispatched by this group */
 195         struct blkg_stat                time;
 196 #ifdef CONFIG_DEBUG_BLK_CGROUP
 197         /* time not charged to this cgroup */
 198         struct blkg_stat                unaccounted_time;
 199         /* sum of number of ios queued across all samples */
 200         struct blkg_stat                avg_queue_size_sum;
 201         /* count of samples taken for average */
 202         struct blkg_stat                avg_queue_size_samples;
 203         /* how many times this group has been removed from service tree */
 204         struct blkg_stat                dequeue;
 205         /* total time spent waiting for it to be assigned a timeslice. */
 206         struct blkg_stat                group_wait_time;
 207         /* time spent idling for this blkcg_gq */
 208         struct blkg_stat                idle_time;
 209         /* total time with empty current active q with other requests queued */
 210         struct blkg_stat                empty_time;
 211         /* fields after this shouldn't be cleared on stat reset */
 212         uint64_t                        start_group_wait_time;
 213         uint64_t                        start_idle_time;
 214         uint64_t                        start_empty_time;
 215         uint16_t                        flags;
 216 #endif  /* CONFIG_DEBUG_BLK_CGROUP */
 217 #endif  /* CONFIG_CFQ_GROUP_IOSCHED */
 218 };
 219
 220 /* Per-cgroup data */
 221 struct cfq_group_data {
 222         /* must be the first member */
 223         struct blkcg_policy_data pd;
 224
 225         unsigned int weight;
 226         unsigned int leaf_weight;
 227 };
 228
 229 /* This is per cgroup per device grouping structure */
 230 struct cfq_group {
 231         /* must be the first member */
 232         struct blkg_policy_data pd;
 233
 234         /* group service_tree member */
 235         struct rb_node rb_node;
 236
 237         /* group service_tree key */
 238         u64 vdisktime;
 239
 240         /*
 241          * The number of active cfqgs and sum of their weights under this
 242          * cfqg.  This covers this cfqg's leaf_weight and all children's
 243          * weights, but does not cover weights of further descendants.
 244          *
 245          * If a cfqg is on the service tree, it's active.  An active cfqg
 246          * also activates its parent and contributes to the children_weight
 247          * of the parent.
 248          */
 249         int nr_active;
 250         unsigned int children_weight;
 251
 252         /*
 253          * vfraction is the fraction of vdisktime that the tasks in this
 254          * cfqg are entitled to.  This is determined by compounding the
 255          * ratios walking up from this cfqg to the root.
 256          *
 257          * It is in fixed point w/ CFQ_SERVICE_SHIFT and the sum of all
 258          * vfractions on a service tree is approximately 1.  The sum may
 259          * deviate a bit due to rounding errors and fluctuations caused by
 260          * cfqgs entering and leaving the service tree.
 261          */
 262         unsigned int vfraction;
 263
 264         /*
 265          * There are two weights - (internal) weight is the weight of this
 266          * cfqg against the sibling cfqgs.  leaf_weight is the wight of
 267          * this cfqg against the child cfqgs.  For the root cfqg, both
 268          * weights are kept in sync for backward compatibility.
 269          */
 270         unsigned int weight;
 271         unsigned int new_weight;
 272         unsigned int dev_weight;
 273
 274         unsigned int leaf_weight;
 275         unsigned int new_leaf_weight;
 276         unsigned int dev_leaf_weight;
 277
 278         /* number of cfqq currently on this group */
 279         int nr_cfqq;
 280
 281         /*
 282          * Per group busy queues average. Useful for workload slice calc. We
 283          * create the array for each prio class but at run time it is used
 284          * only for RT and BE class and slot for IDLE class remains unused.
 285          * This is primarily done to avoid confusion and a gcc warning.
 286          */
 287         unsigned int busy_queues_avg[CFQ_PRIO_NR];
 288         /*
 289          * rr lists of queues with requests. We maintain service trees for
 290          * RT and BE classes. These trees are subdivided in subclasses
 291          * of SYNC, SYNC_NOIDLE and ASYNC based on workload type. For IDLE
 292          * class there is no subclassification and all the cfq queues go on
 293          * a single tree service_tree_idle.
 294          * Counts are embedded in the cfq_rb_root
 295          */
 296         struct cfq_rb_root service_trees[2][3];
 297         struct cfq_rb_root service_tree_idle;
 298
 299         unsigned long saved_wl_slice;
 300         enum wl_type_t saved_wl_type;
 301         enum wl_class_t saved_wl_class;
 302
 303         /* number of requests that are on the dispatch list or inside driver */
 304         int dispatched;
 305         struct cfq_ttime ttime;
 306         struct cfqg_stats stats;        /* stats for this cfqg */
 307         struct cfqg_stats dead_stats;   /* stats pushed from dead children */
 308 };
 309
 310 struct cfq_io_cq {
 311         struct io_cq            icq;            /* must be the first member */
 312         struct cfq_queue        *cfqq[2];
 313         struct cfq_ttime        ttime;
 314         int                     ioprio;         /* the current ioprio */
 315 #ifdef CONFIG_CFQ_GROUP_IOSCHED
 316         uint64_t                blkcg_serial_nr; /* the current blkcg serial */
 317 #endif
 318 };
 319
 320 /*
 321  * Per block device queue structure
 322  */
 323 struct cfq_data {
 324         struct request_queue *queue;
 325         /* Root service tree for cfq_groups */
 326         struct cfq_rb_root grp_service_tree;
 327         struct cfq_group *root_group;
 328
 329         /*
 330          * The priority currently being served
 331          */
 332         enum wl_class_t serving_wl_class;
 333         enum wl_type_t serving_wl_type;
 334         unsigned long workload_expires;
 335         struct cfq_group *serving_group;
 336
 337         /*
 338          * Each priority tree is sorted by next_request position.  These
 339          * trees are used when determining if two or more queues are
 340          * interleaving requests (see cfq_close_cooperator).
 341          */
 342         struct rb_root prio_trees[CFQ_PRIO_LISTS];
 343
 344         unsigned int busy_queues;
 345         unsigned int busy_sync_queues;
 346
 347         int rq_in_driver;
 348         int rq_in_flight[2];
 349
 350         /*
 351          * queue-depth detection
 352          */
 353         int rq_queued;
 354         int hw_tag;
 355         /*
 356          * hw_tag can be
 357          * -1 => indeterminate, (cfq will behave as if NCQ is present, to allow better detection)
 358          *  1 => NCQ is present (hw_tag_est_depth is the estimated max depth)
 359          *  0 => no NCQ
 360          */
 361         int hw_tag_est_depth;
 362         unsigned int hw_tag_samples;
 363
 364         /*
 365          * idle window management
 366          */
 367         struct timer_list idle_slice_timer;
 368         struct work_struct unplug_work;
 369
 370         struct cfq_queue *active_queue;
 371         struct cfq_io_cq *active_cic;
 372
 373         /*
 374          * async queue for each priority case
 375          */
 376         struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
 377         struct cfq_queue *async_idle_cfqq;
 378
 379         sector_t last_position;
 380
 381         /*
 382          * tunables, see top of file
 383          */
 384         unsigned int cfq_quantum;
 385         unsigned int cfq_fifo_expire[2];
 386         unsigned int cfq_back_penalty;
 387         unsigned int cfq_back_max;
 388         unsigned int cfq_slice[2];
 389         unsigned int cfq_slice_async_rq;
 390         unsigned int cfq_slice_idle;
 391         unsigned int cfq_group_idle;
 392         unsigned int cfq_latency;
 393         unsigned int cfq_target_latency;
 394
 395         /*
 396          * Fallback dummy cfqq for extreme OOM conditions
 397          */
 398         struct cfq_queue oom_cfqq;
 399
 400         unsigned long last_delayed_sync;
 401 };
 402
 403 static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd);
 404
 405 static struct cfq_rb_root *st_for(struct cfq_group *cfqg,
 406                                             enum wl_class_t class,
 407                                             enum wl_type_t type)
 408 {
 409         if (!cfqg)
 410                 return NULL;
 411
 412         if (class == IDLE_WORKLOAD)
 413                 return &cfqg->service_tree_idle;
 414
 415         return &cfqg->service_trees[class][type];
 416 }
 417
 418 enum cfqq_state_flags {
 419         CFQ_CFQQ_FLAG_on_rr = 0,        /* on round-robin busy list */
 420         CFQ_CFQQ_FLAG_wait_request,     /* waiting for a request */
 421         CFQ_CFQQ_FLAG_must_dispatch,    /* must be allowed a dispatch */
 422         CFQ_CFQQ_FLAG_must_alloc_slice, /* per-slice must_alloc flag */
 423         CFQ_CFQQ_FLAG_fifo_expire,      /* FIFO checked in this slice */
 424         CFQ_CFQQ_FLAG_idle_window,      /* slice idling enabled */
 425         CFQ_CFQQ_FLAG_prio_changed,     /* task priority has changed */
 426         CFQ_CFQQ_FLAG_slice_new,        /* no requests dispatched in slice */
 427         CFQ_CFQQ_FLAG_sync,             /* synchronous queue */
 428         CFQ_CFQQ_FLAG_coop,             /* cfqq is shared */
 429         CFQ_CFQQ_FLAG_split_coop,       /* shared cfqq will be splitted */
 430         CFQ_CFQQ_FLAG_deep,             /* sync cfqq experienced large depth */
 431         CFQ_CFQQ_FLAG_wait_busy,        /* Waiting for next request */
 432 };
 433
 434 #define CFQ_CFQQ_FNS(name)                                              \
 435 static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)         \
 436 {                                                                       \
 437         (cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name);                   \
 438 }                                                                       \
 439 static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)        \
 440 {                                                                       \
 441         (cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);                  \
 442 }                                                                       \
 443 static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)         \
 444 {                                                                       \
 445         return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;      \
 446 }
 447
 448 CFQ_CFQQ_FNS(on_rr);
 449 CFQ_CFQQ_FNS(wait_request);
 450 CFQ_CFQQ_FNS(must_dispatch);
 451 CFQ_CFQQ_FNS(must_alloc_slice);
 452 CFQ_CFQQ_FNS(fifo_expire);
 453 CFQ_CFQQ_FNS(idle_window);
 454 CFQ_CFQQ_FNS(prio_changed);
 455 CFQ_CFQQ_FNS(slice_new);
 456 CFQ_CFQQ_FNS(sync);
 457 CFQ_CFQQ_FNS(coop);
 458 CFQ_CFQQ_FNS(split_coop);
 459 CFQ_CFQQ_FNS(deep);
 460 CFQ_CFQQ_FNS(wait_busy);
 461 #undef CFQ_CFQQ_FNS
 462
 463 #if defined(CONFIG_CFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP)
 464
 465 /* cfqg stats flags */
 466 enum cfqg_stats_flags {
 467         CFQG_stats_waiting = 0,
 468         CFQG_stats_idling,
 469         CFQG_stats_empty,
 470 };
 471
 472 #define CFQG_FLAG_FNS(name)                                             \
 473 static inline void cfqg_stats_mark_##name(struct cfqg_stats *stats)     \
 474 {                                                                       \
 475         stats->flags |= (1 << CFQG_stats_##name);                       \
 476 }                                                                       \
 477 static inline void cfqg_stats_clear_##name(struct cfqg_stats *stats)    \
 478 {                                                                       \
 479         stats->flags &= ~(1 << CFQG_stats_##name);                      \
 480 }                                                                       \
 481 static inline int cfqg_stats_##name(struct cfqg_stats *stats)           \
 482 {                                                                       \
 483         return (stats->flags & (1 << CFQG_stats_##name)) != 0;          \
 484 }                                                                       \
 485
 486 CFQG_FLAG_FNS(waiting)
 487 CFQG_FLAG_FNS(idling)
 488 CFQG_FLAG_FNS(empty)
 489 #undef CFQG_FLAG_FNS
 490
 491 /* This should be called with the queue_lock held. */
 492 static void cfqg_stats_update_group_wait_time(struct cfqg_stats *stats)
 493 {
 494         unsigned long long now;
 495
 496         if (!cfqg_stats_waiting(stats))
 497                 return;
 498
 499         now = sched_clock();
 500         if (time_after64(now, stats->start_group_wait_time))
 501                 blkg_stat_add(&stats->group_wait_time,
 502                               now - stats->start_group_wait_time);
 503         cfqg_stats_clear_waiting(stats);
 504 }
 505
 506 /* This should be called with the queue_lock held. */
 507 static void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg,
 508                                                  struct cfq_group *curr_cfqg)
 509 {
 510         struct cfqg_stats *stats = &cfqg->stats;
 511
 512         if (cfqg_stats_waiting(stats))
 513                 return;
 514         if (cfqg == curr_cfqg)
 515                 return;
 516         stats->start_group_wait_time = sched_clock();
 517         cfqg_stats_mark_waiting(stats);
 518 }
 519
 520 /* This should be called with the queue_lock held. */
 521 static void cfqg_stats_end_empty_time(struct cfqg_stats *stats)
 522 {
 523         unsigned long long now;
 524
 525         if (!cfqg_stats_empty(stats))
 526                 return;
 527
 528         now = sched_clock();
 529         if (time_after64(now, stats->start_empty_time))
 530                 blkg_stat_add(&stats->empty_time,
 531                               now - stats->start_empty_time);
 532         cfqg_stats_clear_empty(stats);
 533 }
 534
 535 static void cfqg_stats_update_dequeue(struct cfq_group *cfqg)
 536 {
 537         blkg_stat_add(&cfqg->stats.dequeue, 1);
 538 }
 539
 540 static void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg)
 541 {
 542         struct cfqg_stats *stats = &cfqg->stats;
 543
 544         if (blkg_rwstat_total(&stats->queued))
 545                 return;
 546
 547         /*
 548          * group is already marked empty. This can happen if cfqq got new
 549          * request in parent group and moved to this group while being added
 550          * to service tree. Just ignore the event and move on.
 551          */
 552         if (cfqg_stats_empty(stats))
 553                 return;
 554
 555         stats->start_empty_time = sched_clock();
 556         cfqg_stats_mark_empty(stats);
 557 }
 558
 559 static void cfqg_stats_update_idle_time(struct cfq_group *cfqg)
 560 {
 561         struct cfqg_stats *stats = &cfqg->stats;
 562
 563         if (cfqg_stats_idling(stats)) {
 564                 unsigned long long now = sched_clock();
 565
 566                 if (time_after64(now, stats->start_idle_time))
 567                         blkg_stat_add(&stats->idle_time,
 568                                       now - stats->start_idle_time);
 569                 cfqg_stats_clear_idling(stats);
 570         }
 571 }
 572
 573 static void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg)
 574 {
 575         struct cfqg_stats *stats = &cfqg->stats;
 576
 577         BUG_ON(cfqg_stats_idling(stats));
 578
 579         stats->start_idle_time = sched_clock();
 580         cfqg_stats_mark_idling(stats);
 581 }
 582
 583 static void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg)
 584 {
 585         struct cfqg_stats *stats = &cfqg->stats;
 586
 587         blkg_stat_add(&stats->avg_queue_size_sum,
 588                       blkg_rwstat_total(&stats->queued));
 589         blkg_stat_add(&stats->avg_queue_size_samples, 1);
 590         cfqg_stats_update_group_wait_time(stats);
 591 }
 592
 593 #else   /* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */
 594
 595 static inline void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg, struct cfq_group *curr_cfqg) { }
 596 static inline void cfqg_stats_end_empty_time(struct cfqg_stats *stats) { }
 597 static inline void cfqg_stats_update_dequeue(struct cfq_group *cfqg) { }
 598 static inline void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg) { }
 599 static inline void cfqg_stats_update_idle_time(struct cfq_group *cfqg) { }
 600 static inline void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg) { }
 601 static inline void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg) { }
 602
 603 #endif  /* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */
 604
 605 #ifdef CONFIG_CFQ_GROUP_IOSCHED
 606
 607 static inline struct cfq_group *pd_to_cfqg(struct blkg_policy_data *pd)
 608 {
 609         return pd ? container_of(pd, struct cfq_group, pd) : NULL;
 610 }
 611
 612 static struct cfq_group_data
 613 *cpd_to_cfqgd(struct blkcg_policy_data *cpd)
 614 {
 615         return cpd ? container_of(cpd, struct cfq_group_data, pd) : NULL;
 616 }
 617
 618 static inline struct blkcg_gq *cfqg_to_blkg(struct cfq_group *cfqg)
 619 {
 620         return pd_to_blkg(&cfqg->pd);
 621 }
 622
 623 static struct blkcg_policy blkcg_policy_cfq;
 624
 625 static inline struct cfq_group *blkg_to_cfqg(struct blkcg_gq *blkg)
 626 {
 627         return pd_to_cfqg(blkg_to_pd(blkg, &blkcg_policy_cfq));
 628 }
 629
 630 static struct cfq_group_data *blkcg_to_cfqgd(struct blkcg *blkcg)
 631 {
 632         return cpd_to_cfqgd(blkcg_to_cpd(blkcg, &blkcg_policy_cfq));
 633 }
 634
 635 static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg)
 636 {
 637         struct blkcg_gq *pblkg = cfqg_to_blkg(cfqg)->parent;
 638
 639         return pblkg ? blkg_to_cfqg(pblkg) : NULL;
 640 }
 641
 642 static inline void cfqg_get(struct cfq_group *cfqg)
 643 {
 644         return blkg_get(cfqg_to_blkg(cfqg));
 645 }
 646
 647 static inline void cfqg_put(struct cfq_group *cfqg)
 648 {
 649         return blkg_put(cfqg_to_blkg(cfqg));
 650 }
 651
 652 #define cfq_log_cfqq(cfqd, cfqq, fmt, args...)  do {                    \
 653         char __pbuf[128];                                               \
 654                                                                         \
 655         blkg_path(cfqg_to_blkg((cfqq)->cfqg), __pbuf, sizeof(__pbuf));  \
 656         blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c %s " fmt, (cfqq)->pid, \
 657                         cfq_cfqq_sync((cfqq)) ? 'S' : 'A',              \
 658                         cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
 659                           __pbuf, ##args);                              \
 660 } while (0)
 661
 662 #define cfq_log_cfqg(cfqd, cfqg, fmt, args...)  do {                    \
 663         char __pbuf[128];                                               \
 664                                                                         \
 665         blkg_path(cfqg_to_blkg(cfqg), __pbuf, sizeof(__pbuf));          \
 666         blk_add_trace_msg((cfqd)->queue, "%s " fmt, __pbuf, ##args);    \
 667 } while (0)
 668
 669 static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
 670                                             struct cfq_group *curr_cfqg, int rw)
 671 {
 672         blkg_rwstat_add(&cfqg->stats.queued, rw, 1);
 673         cfqg_stats_end_empty_time(&cfqg->stats);
 674         cfqg_stats_set_start_group_wait_time(cfqg, curr_cfqg);
 675 }
 676
 677 static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
 678                         unsigned long time, unsigned long unaccounted_time)
 679 {
 680         blkg_stat_add(&cfqg->stats.time, time);
 681 #ifdef CONFIG_DEBUG_BLK_CGROUP
 682         blkg_stat_add(&cfqg->stats.unaccounted_time, unaccounted_time);
 683 #endif
 684 }
 685
 686 static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int rw)
 687 {
 688         blkg_rwstat_add(&cfqg->stats.queued, rw, -1);
 689 }
 690
 691 static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int rw)
 692 {
 693         blkg_rwstat_add(&cfqg->stats.merged, rw, 1);
 694 }
 695
 696 static inline void cfqg_stats_update_dispatch(struct cfq_group *cfqg,
 697                                               uint64_t bytes, int rw)
 698 {
 699         blkg_stat_add(&cfqg->stats.sectors, bytes >> 9);
 700         blkg_rwstat_add(&cfqg->stats.serviced, rw, 1);
 701         blkg_rwstat_add(&cfqg->stats.service_bytes, rw, bytes);
 702 }
 703
 704 static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
 705                         uint64_t start_time, uint64_t io_start_time, int rw)
 706 {
 707         struct cfqg_stats *stats = &cfqg->stats;
 708         unsigned long long now = sched_clock();
 709
 710         if (time_after64(now, io_start_time))
 711                 blkg_rwstat_add(&stats->service_time, rw, now - io_start_time);
 712         if (time_after64(io_start_time, start_time))
 713                 blkg_rwstat_add(&stats->wait_time, rw,
 714                                 io_start_time - start_time);
 715 }
 716
 717 /* @stats = 0 */
 718 static void cfqg_stats_reset(struct cfqg_stats *stats)
 719 {
 720         /* queued stats shouldn't be cleared */
 721         blkg_rwstat_reset(&stats->service_bytes);
 722         blkg_rwstat_reset(&stats->serviced);
 723         blkg_rwstat_reset(&stats->merged);
 724         blkg_rwstat_reset(&stats->service_time);
 725         blkg_rwstat_reset(&stats->wait_time);
 726         blkg_stat_reset(&stats->time);
 727 #ifdef CONFIG_DEBUG_BLK_CGROUP
 728         blkg_stat_reset(&stats->unaccounted_time);
 729         blkg_stat_reset(&stats->avg_queue_size_sum);
 730         blkg_stat_reset(&stats->avg_queue_size_samples);
 731         blkg_stat_reset(&stats->dequeue);
 732         blkg_stat_reset(&stats->group_wait_time);
 733         blkg_stat_reset(&stats->idle_time);
 734         blkg_stat_reset(&stats->empty_time);
 735 #endif
 736 }
 737
 738 /* @to += @from */
 739 static void cfqg_stats_merge(struct cfqg_stats *to, struct cfqg_stats *from)
 740 {
 741         /* queued stats shouldn't be cleared */
 742         blkg_rwstat_merge(&to->service_bytes, &from->service_bytes);
 743         blkg_rwstat_merge(&to->serviced, &from->serviced);
 744         blkg_rwstat_merge(&to->merged, &from->merged);
 745         blkg_rwstat_merge(&to->service_time, &from->service_time);
 746         blkg_rwstat_merge(&to->wait_time, &from->wait_time);
 747         blkg_stat_merge(&from->time, &from->time);
 748 #ifdef CONFIG_DEBUG_BLK_CGROUP
 749         blkg_stat_merge(&to->unaccounted_time, &from->unaccounted_time);
 750         blkg_stat_merge(&to->avg_queue_size_sum, &from->avg_queue_size_sum);
 751         blkg_stat_merge(&to->avg_queue_size_samples, &from->avg_queue_size_samples);
 752         blkg_stat_merge(&to->dequeue, &from->dequeue);
 753         blkg_stat_merge(&to->group_wait_time, &from->group_wait_time);
 754         blkg_stat_merge(&to->idle_time, &from->idle_time);
 755         blkg_stat_merge(&to->empty_time, &from->empty_time);
 756 #endif
 757 }
 758
 759 /*
 760  * Transfer @cfqg's stats to its parent's dead_stats so that the ancestors'
 761  * recursive stats can still account for the amount used by this cfqg after
 762  * it's gone.
 763  */
 764 static void cfqg_stats_xfer_dead(struct cfq_group *cfqg)
 765 {
 766         struct cfq_group *parent = cfqg_parent(cfqg);
 767
 768         lockdep_assert_held(cfqg_to_blkg(cfqg)->q->queue_lock);
 769
 770         if (unlikely(!parent))
 771                 return;
 772
 773         cfqg_stats_merge(&parent->dead_stats, &cfqg->stats);
 774         cfqg_stats_merge(&parent->dead_stats, &cfqg->dead_stats);
 775         cfqg_stats_reset(&cfqg->stats);
 776         cfqg_stats_reset(&cfqg->dead_stats);
 777 }
 778
 779 #else   /* CONFIG_CFQ_GROUP_IOSCHED */
 780
 781 static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg) { return NULL; }
 782 static inline void cfqg_get(struct cfq_group *cfqg) { }
 783 static inline void cfqg_put(struct cfq_group *cfqg) { }
 784
 785 #define cfq_log_cfqq(cfqd, cfqq, fmt, args...)  \
 786         blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c " fmt, (cfqq)->pid, \
 787                         cfq_cfqq_sync((cfqq)) ? 'S' : 'A',              \
 788                         cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
 789                                 ##args)
 790 #define cfq_log_cfqg(cfqd, cfqg, fmt, args...)          do {} while (0)
 791
 792 static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
 793                         struct cfq_group *curr_cfqg, int rw) { }
 794 static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
 795                         unsigned long time, unsigned long unaccounted_time) { }
 796 static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int rw) { }
 797 static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int rw) { }
 798 static inline void cfqg_stats_update_dispatch(struct cfq_group *cfqg,
 799                                               uint64_t bytes, int rw) { }
 800 static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
 801                         uint64_t start_time, uint64_t io_start_time, int rw) { }
 802
 803 #endif  /* CONFIG_CFQ_GROUP_IOSCHED */
 804
 805 #define cfq_log(cfqd, fmt, args...)     \
 806         blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)
 807
 808 /* Traverses through cfq group service trees */
 809 #define for_each_cfqg_st(cfqg, i, j, st) \
 810         for (i = 0; i <= IDLE_WORKLOAD; i++) \
 811                 for (j = 0, st = i < IDLE_WORKLOAD ? &cfqg->service_trees[i][j]\
 812                         : &cfqg->service_tree_idle; \
 813                         (i < IDLE_WORKLOAD && j <= SYNC_WORKLOAD) || \
 814                         (i == IDLE_WORKLOAD && j == 0); \
 815                         j++, st = i < IDLE_WORKLOAD ? \
 816                         &cfqg->service_trees[i][j]: NULL) \
 817
 818 static inline bool cfq_io_thinktime_big(struct cfq_data *cfqd,
 819         struct cfq_ttime *ttime, bool group_idle)
 820 {
 821         unsigned long slice;
 822         if (!sample_valid(ttime->ttime_samples))
 823                 return false;
 824         if (group_idle)
 825                 slice = cfqd->cfq_group_idle;
 826         else
 827                 slice = cfqd->cfq_slice_idle;
 828         return ttime->ttime_mean > slice;
 829 }
 830
 831 static inline bool iops_mode(struct cfq_data *cfqd)
 832 {
 833         /*
 834          * If we are not idling on queues and it is a NCQ drive, parallel
 835          * execution of requests is on and measuring time is not possible
 836          * in most of the cases until and unless we drive shallower queue
 837          * depths and that becomes a performance bottleneck. In such cases
 838          * switch to start providing fairness in terms of number of IOs.
 839          */
 840         if (!cfqd->cfq_slice_idle && cfqd->hw_tag)
 841                 return true;
 842         else
 843                 return false;
 844 }
 845
 846 static inline enum wl_class_t cfqq_class(struct cfq_queue *cfqq)
 847 {
 848         if (cfq_class_idle(cfqq))
 849                 return IDLE_WORKLOAD;
 850         if (cfq_class_rt(cfqq))
 851                 return RT_WORKLOAD;
 852         return BE_WORKLOAD;
 853 }
 854
 855
 856 static enum wl_type_t cfqq_type(struct cfq_queue *cfqq)
 857 {
 858         if (!cfq_cfqq_sync(cfqq))
 859                 return ASYNC_WORKLOAD;
 860         if (!cfq_cfqq_idle_window(cfqq))
 861                 return SYNC_NOIDLE_WORKLOAD;
 862         return SYNC_WORKLOAD;
 863 }
 864
 865 static inline int cfq_group_busy_queues_wl(enum wl_class_t wl_class,
 866                                         struct cfq_data *cfqd,
 867                                         struct cfq_group *cfqg)
 868 {
 869         if (wl_class == IDLE_WORKLOAD)
 870                 return cfqg->service_tree_idle.count;
 871
 872         return cfqg->service_trees[wl_class][ASYNC_WORKLOAD].count +
 873                 cfqg->service_trees[wl_class][SYNC_NOIDLE_WORKLOAD].count +
 874                 cfqg->service_trees[wl_class][SYNC_WORKLOAD].count;
 875 }
 876
 877 static inline int cfqg_busy_async_queues(struct cfq_data *cfqd,
 878                                         struct cfq_group *cfqg)
 879 {
 880         return cfqg->service_trees[RT_WORKLOAD][ASYNC_WORKLOAD].count +
 881                 cfqg->service_trees[BE_WORKLOAD][ASYNC_WORKLOAD].count;
 882 }
 883
 884 static void cfq_dispatch_insert(struct request_queue *, struct request *);
 885 static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, bool is_sync,
 886                                        struct cfq_io_cq *cic, struct bio *bio,
 887                                        gfp_t gfp_mask);
 888
 889 static inline struct cfq_io_cq *icq_to_cic(struct io_cq *icq)
 890 {
 891         /* cic->icq is the first member, %NULL will convert to %NULL */
 892         return container_of(icq, struct cfq_io_cq, icq);
 893 }
 894
 895 static inline struct cfq_io_cq *cfq_cic_lookup(struct cfq_data *cfqd,
 896                                                struct io_context *ioc)
 897 {
 898         if (ioc)
 899                 return icq_to_cic(ioc_lookup_icq(ioc, cfqd->queue));
 900         return NULL;
 901 }
 902
 903 static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_cq *cic, bool is_sync)
 904 {
 905         return cic->cfqq[is_sync];
 906 }
 907
 908 static inline void cic_set_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq,
 909                                 bool is_sync)
 910 {
 911         cic->cfqq[is_sync] = cfqq;
 912 }
 913
 914 static inline struct cfq_data *cic_to_cfqd(struct cfq_io_cq *cic)
 915 {
 916         return cic->icq.q->elevator->elevator_data;
 917 }
 918
 919 /*
 920  * We regard a request as SYNC, if it's either a read or has the SYNC bit
 921  * set (in which case it could also be direct WRITE).
 922  */
 923 static inline bool cfq_bio_sync(struct bio *bio)
 924 {
 925         return bio_data_dir(bio) == READ || (bio->bi_rw & REQ_SYNC);
 926 }
 927
 928 /*
 929  * scheduler run of queue, if there are requests pending and no one in the
 930  * driver that will restart queueing
 931  */
 932 static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
 933 {
 934         if (cfqd->busy_queues) {
 935                 cfq_log(cfqd, "schedule dispatch");
 936                 kblockd_schedule_work(&cfqd->unplug_work);
 937         }
 938 }
 939
 940 /*
 941  * Scale schedule slice based on io priority. Use the sync time slice only
 942  * if a queue is marked sync and has sync io queued. A sync queue with async
 943  * io only, should not get full sync slice length.
 944  */
 945 static inline int cfq_prio_slice(struct cfq_data *cfqd, bool sync,
 946                                  unsigned short prio)
 947 {
 948         const int base_slice = cfqd->cfq_slice[sync];
 949
 950         WARN_ON(prio >= IOPRIO_BE_NR);
 951
 952         return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio));
 953 }
 954
 955 static inline int
 956 cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 957 {
 958         return cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio);
 959 }
 960
 961 /**
 962  * cfqg_scale_charge - scale disk time charge according to cfqg weight
 963  * @charge: disk time being charged
 964  * @vfraction: vfraction of the cfqg, fixed point w/ CFQ_SERVICE_SHIFT
 965  *
 966  * Scale @charge according to @vfraction, which is in range (0, 1].  The
 967  * scaling is inversely proportional.
 968  *
 969  * scaled = charge / vfraction
 970  *
 971  * The result is also in fixed point w/ CFQ_SERVICE_SHIFT.
 972  */
 973 static inline u64 cfqg_scale_charge(unsigned long charge,
 974                                     unsigned int vfraction)
 975 {
 976         u64 c = charge << CFQ_SERVICE_SHIFT;    /* make it fixed point */
 977
 978         /* charge / vfraction */
 979         c <<= CFQ_SERVICE_SHIFT;
 980         do_div(c, vfraction);
 981         return c;
 982 }
 983
 984 static inline u64 max_vdisktime(u64 min_vdisktime, u64 vdisktime)
 985 {
 986         s64 delta = (s64)(vdisktime - min_vdisktime);
 987         if (delta > 0)
 988                 min_vdisktime = vdisktime;
 989
 990         return min_vdisktime;
 991 }
 992
 993 static inline u64 min_vdisktime(u64 min_vdisktime, u64 vdisktime)
 994 {
 995         s64 delta = (s64)(vdisktime - min_vdisktime);
 996         if (delta < 0)
 997                 min_vdisktime = vdisktime;
 998
 999         return min_vdisktime;
1000 }
1001
1002 static void update_min_vdisktime(struct cfq_rb_root *st)
1003 {
1004         struct cfq_group *cfqg;
1005
1006         if (st->left) {
1007                 cfqg = rb_entry_cfqg(st->left);
1008                 st->min_vdisktime = max_vdisktime(st->min_vdisktime,
1009                                                   cfqg->vdisktime);
1010         }
1011 }
1012
1013 /*
1014  * get averaged number of queues of RT/BE priority.
1015  * average is updated, with a formula that gives more weight to higher numbers,
1016  * to quickly follows sudden increases and decrease slowly
1017  */
1018
1019 static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd,
1020                                         struct cfq_group *cfqg, bool rt)
1021 {
1022         unsigned min_q, max_q;
1023         unsigned mult  = cfq_hist_divisor - 1;
1024         unsigned round = cfq_hist_divisor / 2;
1025         unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg);
1026
1027         min_q = min(cfqg->busy_queues_avg[rt], busy);
1028         max_q = max(cfqg->busy_queues_avg[rt], busy);
1029         cfqg->busy_queues_avg[rt] = (mult * max_q + min_q + round) /
1030                 cfq_hist_divisor;
1031         return cfqg->busy_queues_avg[rt];
1032 }
1033
1034 static inline unsigned
1035 cfq_group_slice(struct cfq_data *cfqd, struct cfq_group *cfqg)
1036 {
1037         return cfqd->cfq_target_latency * cfqg->vfraction >> CFQ_SERVICE_SHIFT;
1038 }
1039
1040 static inline unsigned
1041 cfq_scaled_cfqq_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1042 {
1043         unsigned slice = cfq_prio_to_slice(cfqd, cfqq);
1044         if (cfqd->cfq_latency) {
1045                 /*
1046                  * interested queues (we consider only the ones with the same
1047                  * priority class in the cfq group)
1048                  */
1049                 unsigned iq = cfq_group_get_avg_queues(cfqd, cfqq->cfqg,
1050                                                 cfq_class_rt(cfqq));
1051                 unsigned sync_slice = cfqd->cfq_slice[1];
1052                 unsigned expect_latency = sync_slice * iq;
1053                 unsigned group_slice = cfq_group_slice(cfqd, cfqq->cfqg);
1054
1055                 if (expect_latency > group_slice) {
1056                         unsigned base_low_slice = 2 * cfqd->cfq_slice_idle;
1057                         /* scale low_slice according to IO priority
1058                          * and sync vs async */
1059                         unsigned low_slice =
1060                                 min(slice, base_low_slice * slice / sync_slice);
1061                         /* the adapted slice value is scaled to fit all iqs
1062                          * into the target latency */
1063                         slice = max(slice * group_slice / expect_latency,
1064                                     low_slice);
1065                 }
1066         }
1067         return slice;
1068 }
1069
1070 static inline void
1071 cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1072 {
1073         unsigned slice = cfq_scaled_cfqq_slice(cfqd, cfqq);
1074
1075         cfqq->slice_start = jiffies;
1076         cfqq->slice_end = jiffies + slice;
1077         cfqq->allocated_slice = slice;
1078         cfq_log_cfqq(cfqd, cfqq, "set_slice=%lu", cfqq->slice_end - jiffies);
1079 }
1080
1081 /*
1082  * We need to wrap this check in cfq_cfqq_slice_new(), since ->slice_end
1083  * isn't valid until the first request from the dispatch is activated
1084  * and the slice time set.
1085  */
1086 static inline bool cfq_slice_used(struct cfq_queue *cfqq)
1087 {
1088         if (cfq_cfqq_slice_new(cfqq))
1089                 return false;
1090         if (time_before(jiffies, cfqq->slice_end))
1091                 return false;
1092
1093         return true;
1094 }
1095
1096 /*
1097  * Lifted from AS - choose which of rq1 and rq2 that is best served now.
1098  * We choose the request that is closest to the head right now. Distance
1099  * behind the head is penalized and only allowed to a certain extent.
1100  */
1101 static struct request *
1102 cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2, sector_t last)
1103 {
1104         sector_t s1, s2, d1 = 0, d2 = 0;
1105         unsigned long back_max;
1106 #define CFQ_RQ1_WRAP    0x01 /* request 1 wraps */
1107 #define CFQ_RQ2_WRAP    0x02 /* request 2 wraps */
1108         unsigned wrap = 0; /* bit mask: requests behind the disk head? */
1109
1110         if (rq1 == NULL || rq1 == rq2)
1111                 return rq2;
1112         if (rq2 == NULL)
1113                 return rq1;
1114
1115         if (rq_is_sync(rq1) != rq_is_sync(rq2))
1116                 return rq_is_sync(rq1) ? rq1 : rq2;
1117
1118         if ((rq1->cmd_flags ^ rq2->cmd_flags) & REQ_PRIO)
1119                 return rq1->cmd_flags & REQ_PRIO ? rq1 : rq2;
1120
1121         s1 = blk_rq_pos(rq1);
1122         s2 = blk_rq_pos(rq2);
1123
1124         /*
1125          * by definition, 1KiB is 2 sectors
1126          */
1127         back_max = cfqd->cfq_back_max * 2;
1128
1129         /*
1130          * Strict one way elevator _except_ in the case where we allow
1131          * short backward seeks which are biased as twice the cost of a
1132          * similar forward seek.
1133          */
1134         if (s1 >= last)
1135                 d1 = s1 - last;
1136         else if (s1 + back_max >= last)
1137                 d1 = (last - s1) * cfqd->cfq_back_penalty;
1138         else
1139                 wrap |= CFQ_RQ1_WRAP;
1140
1141         if (s2 >= last)
1142                 d2 = s2 - last;
1143         else if (s2 + back_max >= last)
1144                 d2 = (last - s2) * cfqd->cfq_back_penalty;
1145         else
1146                 wrap |= CFQ_RQ2_WRAP;
1147
1148         /* Found required data */
1149
1150         /*
1151          * By doing switch() on the bit mask "wrap" we avoid having to
1152          * check two variables for all permutations: --> faster!
1153          */
1154         switch (wrap) {
1155         case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
1156                 if (d1 < d2)
1157                         return rq1;
1158                 else if (d2 < d1)
1159                         return rq2;
1160                 else {
1161                         if (s1 >= s2)
1162                                 return rq1;
1163                         else
1164                                 return rq2;
1165                 }
1166
1167         case CFQ_RQ2_WRAP:
1168                 return rq1;
1169         case CFQ_RQ1_WRAP:
1170                 return rq2;
1171         case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */
1172         default:
1173                 /*
1174                  * Since both rqs are wrapped,
1175                  * start with the one that's further behind head
1176                  * (--> only *one* back seek required),
1177                  * since back seek takes more time than forward.
1178                  */
1179                 if (s1 <= s2)
1180                         return rq1;
1181                 else
1182                         return rq2;
1183         }
1184 }
1185
1186 /*
1187  * The below is leftmost cache rbtree addon
1188  */
1189 static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root)
1190 {
1191         /* Service tree is empty */
1192         if (!root->count)
1193                 return NULL;
1194
1195         if (!root->left)
1196                 root->left = rb_first(&root->rb);
1197
1198         if (root->left)
1199                 return rb_entry(root->left, struct cfq_queue, rb_node);
1200
1201         return NULL;
1202 }
1203
1204 static struct cfq_group *cfq_rb_first_group(struct cfq_rb_root *root)
1205 {
1206         if (!root->left)
1207                 root->left = rb_first(&root->rb);
1208
1209         if (root->left)
1210                 return rb_entry_cfqg(root->left);
1211
1212         return NULL;
1213 }
1214
1215 static void rb_erase_init(struct rb_node *n, struct rb_root *root)
1216 {
1217         rb_erase(n, root);
1218         RB_CLEAR_NODE(n);
1219 }
1220
1221 static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root)
1222 {
1223         if (root->left == n)
1224                 root->left = NULL;
1225         rb_erase_init(n, &root->rb);
1226         --root->count;
1227 }
1228
1229 /*
1230  * would be nice to take fifo expire time into account as well
1231  */
1232 static struct request *
1233 cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1234                   struct request *last)
1235 {
1236         struct rb_node *rbnext = rb_next(&last->rb_node);
1237         struct rb_node *rbprev = rb_prev(&last->rb_node);
1238         struct request *next = NULL, *prev = NULL;
1239
1240         BUG_ON(RB_EMPTY_NODE(&last->rb_node));
1241
1242         if (rbprev)
1243                 prev = rb_entry_rq(rbprev);
1244
1245         if (rbnext)
1246                 next = rb_entry_rq(rbnext);
1247         else {
1248                 rbnext = rb_first(&cfqq->sort_list);
1249                 if (rbnext && rbnext != &last->rb_node)
1250                         next = rb_entry_rq(rbnext);
1251         }
1252
1253         return cfq_choose_req(cfqd, next, prev, blk_rq_pos(last));
1254 }
1255
1256 static unsigned long cfq_slice_offset(struct cfq_data *cfqd,
1257                                       struct cfq_queue *cfqq)
1258 {
1259         /*
1260          * just an approximation, should be ok.
1261          */
1262         return (cfqq->cfqg->nr_cfqq - 1) * (cfq_prio_slice(cfqd, 1, 0) -
1263                        cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio));
1264 }
1265
1266 static inline s64
1267 cfqg_key(struct cfq_rb_root *st, struct cfq_group *cfqg)
1268 {
1269         return cfqg->vdisktime - st->min_vdisktime;
1270 }
1271
1272 static void
1273 __cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
1274 {
1275         struct rb_node **node = &st->rb.rb_node;
1276         struct rb_node *parent = NULL;
1277         struct cfq_group *__cfqg;
1278         s64 key = cfqg_key(st, cfqg);
1279         int left = 1;
1280
1281         while (*node != NULL) {
1282                 parent = *node;
1283                 __cfqg = rb_entry_cfqg(parent);
1284
1285                 if (key < cfqg_key(st, __cfqg))
1286                         node = &parent->rb_left;
1287                 else {
1288                         node = &parent->rb_right;
1289                         left = 0;
1290                 }
1291         }
1292
1293         if (left)
1294                 st->left = &cfqg->rb_node;
1295
1296         rb_link_node(&cfqg->rb_node, parent, node);
1297         rb_insert_color(&cfqg->rb_node, &st->rb);
1298 }
1299
1300 /*
1301  * This has to be called only on activation of cfqg
1302  */
1303 static void
1304 cfq_update_group_weight(struct cfq_group *cfqg)
1305 {
1306         if (cfqg->new_weight) {
1307                 cfqg->weight = cfqg->new_weight;
1308                 cfqg->new_weight = 0;
1309         }
1310 }
1311
1312 static void
1313 cfq_update_group_leaf_weight(struct cfq_group *cfqg)
1314 {
1315         BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
1316
1317         if (cfqg->new_leaf_weight) {
1318                 cfqg->leaf_weight = cfqg->new_leaf_weight;
1319                 cfqg->new_leaf_weight = 0;
1320         }
1321 }
1322
1323 static void
1324 cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
1325 {
1326         unsigned int vfr = 1 << CFQ_SERVICE_SHIFT;      /* start with 1 */
1327         struct cfq_group *pos = cfqg;
1328         struct cfq_group *parent;
1329         bool propagate;
1330
1331         /* add to the service tree */
1332         BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
1333
1334         /*
1335          * Update leaf_weight.  We cannot update weight at this point
1336          * because cfqg might already have been activated and is
1337          * contributing its current weight to the parent's child_weight.
1338          */
1339         cfq_update_group_leaf_weight(cfqg);
1340         __cfq_group_service_tree_add(st, cfqg);
1341
1342         /*
1343          * Activate @cfqg and calculate the portion of vfraction @cfqg is
1344          * entitled to.  vfraction is calculated by walking the tree
1345          * towards the root calculating the fraction it has at each level.
1346          * The compounded ratio is how much vfraction @cfqg owns.
1347          *
1348          * Start with the proportion tasks in this cfqg has against active
1349          * children cfqgs - its leaf_weight against children_weight.
1350          */
1351         propagate = !pos->nr_active++;
1352         pos->children_weight += pos->leaf_weight;
1353         vfr = vfr * pos->leaf_weight / pos->children_weight;
1354
1355         /*
1356          * Compound ->weight walking up the tree.  Both activation and
1357          * vfraction calculation are done in the same loop.  Propagation
1358          * stops once an already activated node is met.  vfraction
1359          * calculation should always continue to the root.
1360          */
1361         while ((parent = cfqg_parent(pos))) {
1362                 if (propagate) {
1363                         cfq_update_group_weight(pos);
1364                         propagate = !parent->nr_active++;
1365                         parent->children_weight += pos->weight;
1366                 }
1367                 vfr = vfr * pos->weight / parent->children_weight;
1368                 pos = parent;
1369         }
1370
1371         cfqg->vfraction = max_t(unsigned, vfr, 1);
1372 }
1373
1374 static void
1375 cfq_group_notify_queue_add(struct cfq_data *cfqd, struct cfq_group *cfqg)
1376 {
1377         struct cfq_rb_root *st = &cfqd->grp_service_tree;
1378         struct cfq_group *__cfqg;
1379         struct rb_node *n;
1380
1381         cfqg->nr_cfqq++;
1382         if (!RB_EMPTY_NODE(&cfqg->rb_node))
1383                 return;
1384
1385         /*
1386          * Currently put the group at the end. Later implement something
1387          * so that groups get lesser vtime based on their weights, so that
1388          * if group does not loose all if it was not continuously backlogged.
1389          */
1390         n = rb_last(&st->rb);
1391         if (n) {
1392                 __cfqg = rb_entry_cfqg(n);
1393                 cfqg->vdisktime = __cfqg->vdisktime + CFQ_IDLE_DELAY;
1394         } else
1395                 cfqg->vdisktime = st->min_vdisktime;
1396         cfq_group_service_tree_add(st, cfqg);
1397 }
1398
1399 static void
1400 cfq_group_service_tree_del(struct cfq_rb_root *st, struct cfq_group *cfqg)
1401 {
1402         struct cfq_group *pos = cfqg;
1403         bool propagate;
1404
1405         /*
1406          * Undo activation from cfq_group_service_tree_add().  Deactivate
1407          * @cfqg and propagate deactivation upwards.
1408          */
1409         propagate = !--pos->nr_active;
1410         pos->children_weight -= pos->leaf_weight;
1411
1412         while (propagate) {
1413                 struct cfq_group *parent = cfqg_parent(pos);
1414
1415                 /* @pos has 0 nr_active at this point */
1416                 WARN_ON_ONCE(pos->children_weight);
1417                 pos->vfraction = 0;
1418
1419                 if (!parent)
1420                         break;
1421
1422                 propagate = !--parent->nr_active;
1423                 parent->children_weight -= pos->weight;
1424                 pos = parent;
1425         }
1426
1427         /* remove from the service tree */
1428         if (!RB_EMPTY_NODE(&cfqg->rb_node))
1429                 cfq_rb_erase(&cfqg->rb_node, st);
1430 }
1431
1432 static void
1433 cfq_group_notify_queue_del(struct cfq_data *cfqd, struct cfq_group *cfqg)
1434 {
1435         struct cfq_rb_root *st = &cfqd->grp_service_tree;
1436
1437         BUG_ON(cfqg->nr_cfqq < 1);
1438         cfqg->nr_cfqq--;
1439
1440         /* If there are other cfq queues under this group, don't delete it */
1441         if (cfqg->nr_cfqq)
1442                 return;
1443
1444         cfq_log_cfqg(cfqd, cfqg, "del_from_rr group");
1445         cfq_group_service_tree_del(st, cfqg);
1446         cfqg->saved_wl_slice = 0;
1447         cfqg_stats_update_dequeue(cfqg);
1448 }
1449
1450 static inline unsigned int cfq_cfqq_slice_usage(struct cfq_queue *cfqq,
1451                                                 unsigned int *unaccounted_time)
1452 {
1453         unsigned int slice_used;
1454
1455         /*
1456          * Queue got expired before even a single request completed or
1457          * got expired immediately after first request completion.
1458          */
1459         if (!cfqq->slice_start || cfqq->slice_start == jiffies) {
1460                 /*
1461                  * Also charge the seek time incurred to the group, otherwise
1462                  * if there are mutiple queues in the group, each can dispatch
1463                  * a single request on seeky media and cause lots of seek time
1464                  * and group will never know it.
1465                  */
1466                 slice_used = max_t(unsigned, (jiffies - cfqq->dispatch_start),
1467                                         1);
1468         } else {
1469                 slice_used = jiffies - cfqq->slice_start;
1470                 if (slice_used > cfqq->allocated_slice) {
1471                         *unaccounted_time = slice_used - cfqq->allocated_slice;
1472                         slice_used = cfqq->allocated_slice;
1473                 }
1474                 if (time_after(cfqq->slice_start, cfqq->dispatch_start))
1475                         *unaccounted_time += cfqq->slice_start -
1476                                         cfqq->dispatch_start;
1477         }
1478
1479         return slice_used;
1480 }
1481
1482 static void cfq_group_served(struct cfq_data *cfqd, struct cfq_group *cfqg,
1483                                 struct cfq_queue *cfqq)
1484 {
1485         struct cfq_rb_root *st = &cfqd->grp_service_tree;
1486         unsigned int used_sl, charge, unaccounted_sl = 0;
1487         int nr_sync = cfqg->nr_cfqq - cfqg_busy_async_queues(cfqd, cfqg)
1488                         - cfqg->service_tree_idle.count;
1489         unsigned int vfr;
1490
1491         BUG_ON(nr_sync < 0);
1492         used_sl = charge = cfq_cfqq_slice_usage(cfqq, &unaccounted_sl);
1493
1494         if (iops_mode(cfqd))
1495                 charge = cfqq->slice_dispatch;
1496         else if (!cfq_cfqq_sync(cfqq) && !nr_sync)
1497                 charge = cfqq->allocated_slice;
1498
1499         /*
1500          * Can't update vdisktime while on service tree and cfqg->vfraction
1501          * is valid only while on it.  Cache vfr, leave the service tree,
1502          * update vdisktime and go back on.  The re-addition to the tree
1503          * will also update the weights as necessary.
1504          */
1505         vfr = cfqg->vfraction;
1506         cfq_group_service_tree_del(st, cfqg);
1507         cfqg->vdisktime += cfqg_scale_charge(charge, vfr);
1508         cfq_group_service_tree_add(st, cfqg);
1509
1510         /* This group is being expired. Save the context */
1511         if (time_after(cfqd->workload_expires, jiffies)) {
1512                 cfqg->saved_wl_slice = cfqd->workload_expires
1513                                                 - jiffies;
1514                 cfqg->saved_wl_type = cfqd->serving_wl_type;
1515                 cfqg->saved_wl_class = cfqd->serving_wl_class;
1516         } else
1517                 cfqg->saved_wl_slice = 0;
1518
1519         cfq_log_cfqg(cfqd, cfqg, "served: vt=%llu min_vt=%llu", cfqg->vdisktime,
1520                                         st->min_vdisktime);
1521         cfq_log_cfqq(cfqq->cfqd, cfqq,
1522                      "sl_used=%u disp=%u charge=%u iops=%u sect=%lu",
1523                      used_sl, cfqq->slice_dispatch, charge,
1524                      iops_mode(cfqd), cfqq->nr_sectors);
1525         cfqg_stats_update_timeslice_used(cfqg, used_sl, unaccounted_sl);
1526         cfqg_stats_set_start_empty_time(cfqg);
1527 }
1528
1529 /**
1530  * cfq_init_cfqg_base - initialize base part of a cfq_group
1531  * @cfqg: cfq_group to initialize
1532  *
1533  * Initialize the base part which is used whether %CONFIG_CFQ_GROUP_IOSCHED
1534  * is enabled or not.
1535  */
1536 static void cfq_init_cfqg_base(struct cfq_group *cfqg)
1537 {
1538         struct cfq_rb_root *st;
1539         int i, j;
1540
1541         for_each_cfqg_st(cfqg, i, j, st)
1542                 *st = CFQ_RB_ROOT;
1543         RB_CLEAR_NODE(&cfqg->rb_node);
1544
1545         cfqg->ttime.last_end_request = jiffies;
1546 }
1547
1548 #ifdef CONFIG_CFQ_GROUP_IOSCHED
1549 static void cfqg_stats_init(struct cfqg_stats *stats)
1550 {
1551         blkg_rwstat_init(&stats->service_bytes);
1552         blkg_rwstat_init(&stats->serviced);
1553         blkg_rwstat_init(&stats->merged);
1554         blkg_rwstat_init(&stats->service_time);
1555         blkg_rwstat_init(&stats->wait_time);
1556         blkg_rwstat_init(&stats->queued);
1557
1558         blkg_stat_init(&stats->sectors);
1559         blkg_stat_init(&stats->time);
1560
1561 #ifdef CONFIG_DEBUG_BLK_CGROUP
1562         blkg_stat_init(&stats->unaccounted_time);
1563         blkg_stat_init(&stats->avg_queue_size_sum);
1564         blkg_stat_init(&stats->avg_queue_size_samples);
1565         blkg_stat_init(&stats->dequeue);
1566         blkg_stat_init(&stats->group_wait_time);
1567         blkg_stat_init(&stats->idle_time);
1568         blkg_stat_init(&stats->empty_time);
1569 #endif
1570 }
1571
1572 static void cfq_cpd_init(const struct blkcg *blkcg)
1573 {
1574         struct cfq_group_data *cgd =
1575                 cpd_to_cfqgd(blkcg->pd[blkcg_policy_cfq.plid]);
1576
1577         if (blkcg == &blkcg_root) {
1578                 cgd->weight = 2 * CFQ_WEIGHT_DEFAULT;
1579                 cgd->leaf_weight = 2 * CFQ_WEIGHT_DEFAULT;
1580         } else {
1581                 cgd->weight = CFQ_WEIGHT_DEFAULT;
1582                 cgd->leaf_weight = CFQ_WEIGHT_DEFAULT;
1583         }
1584 }
1585
1586 static void cfq_pd_init(struct blkcg_gq *blkg)
1587 {
1588         struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1589         struct cfq_group_data *cgd = blkcg_to_cfqgd(blkg->blkcg);
1590
1591         cfq_init_cfqg_base(cfqg);
1592         cfqg->weight = cgd->weight;
1593         cfqg->leaf_weight = cgd->leaf_weight;
1594         cfqg_stats_init(&cfqg->stats);
1595         cfqg_stats_init(&cfqg->dead_stats);
1596 }
1597
1598 static void cfq_pd_offline(struct blkcg_gq *blkg)
1599 {
1600         /*
1601          * @blkg is going offline and will be ignored by
1602          * blkg_[rw]stat_recursive_sum().  Transfer stats to the parent so
1603          * that they don't get lost.  If IOs complete after this point, the
1604          * stats for them will be lost.  Oh well...
1605          */
1606         cfqg_stats_xfer_dead(blkg_to_cfqg(blkg));
1607 }
1608
1609 /* offset delta from cfqg->stats to cfqg->dead_stats */
1610 static const int dead_stats_off_delta = offsetof(struct cfq_group, dead_stats) -
1611                                         offsetof(struct cfq_group, stats);
1612
1613 /* to be used by recursive prfill, sums live and dead stats recursively */
1614 static u64 cfqg_stat_pd_recursive_sum(struct blkg_policy_data *pd, int off)
1615 {
1616         u64 sum = 0;
1617
1618         sum += blkg_stat_recursive_sum(pd, off);
1619         sum += blkg_stat_recursive_sum(pd, off + dead_stats_off_delta);
1620         return sum;
1621 }
1622
1623 /* to be used by recursive prfill, sums live and dead rwstats recursively */
1624 static struct blkg_rwstat cfqg_rwstat_pd_recursive_sum(struct blkg_policy_data *pd,
1625                                                        int off)
1626 {
1627         struct blkg_rwstat a, b;
1628
1629         a = blkg_rwstat_recursive_sum(pd, off);
1630         b = blkg_rwstat_recursive_sum(pd, off + dead_stats_off_delta);
1631         blkg_rwstat_merge(&a, &b);
1632         return a;
1633 }
1634
1635 static void cfq_pd_reset_stats(struct blkcg_gq *blkg)
1636 {
1637         struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1638
1639         cfqg_stats_reset(&cfqg->stats);
1640         cfqg_stats_reset(&cfqg->dead_stats);
1641 }
1642
1643 /*
1644  * Search for the cfq group current task belongs to. request_queue lock must
1645  * be held.
1646  */
1647 static struct cfq_group *cfq_lookup_create_cfqg(struct cfq_data *cfqd,
1648                                                 struct blkcg *blkcg)
1649 {
1650         struct request_queue *q = cfqd->queue;
1651         struct cfq_group *cfqg = NULL;
1652
1653         /* avoid lookup for the common case where there's no blkcg */
1654         if (blkcg == &blkcg_root) {
1655                 cfqg = cfqd->root_group;
1656         } else {
1657                 struct blkcg_gq *blkg;
1658
1659                 blkg = blkg_lookup_create(blkcg, q);
1660                 if (!IS_ERR(blkg))
1661                         cfqg = blkg_to_cfqg(blkg);
1662         }
1663
1664         return cfqg;
1665 }
1666
1667 static void cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg)
1668 {
1669         /* Currently, all async queues are mapped to root group */
1670         if (!cfq_cfqq_sync(cfqq))
1671                 cfqg = cfqq->cfqd->root_group;
1672
1673         cfqq->cfqg = cfqg;
1674         /* cfqq reference on cfqg */
1675         cfqg_get(cfqg);
1676 }
1677
1678 static u64 cfqg_prfill_weight_device(struct seq_file *sf,
1679                                      struct blkg_policy_data *pd, int off)
1680 {
1681         struct cfq_group *cfqg = pd_to_cfqg(pd);
1682
1683         if (!cfqg->dev_weight)
1684                 return 0;
1685         return __blkg_prfill_u64(sf, pd, cfqg->dev_weight);
1686 }
1687
1688 static int cfqg_print_weight_device(struct seq_file *sf, void *v)
1689 {
1690         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1691                           cfqg_prfill_weight_device, &blkcg_policy_cfq,
1692                           0, false);
1693         return 0;
1694 }
1695
1696 static u64 cfqg_prfill_leaf_weight_device(struct seq_file *sf,
1697                                           struct blkg_policy_data *pd, int off)
1698 {
1699         struct cfq_group *cfqg = pd_to_cfqg(pd);
1700
1701         if (!cfqg->dev_leaf_weight)
1702                 return 0;
1703         return __blkg_prfill_u64(sf, pd, cfqg->dev_leaf_weight);
1704 }
1705
1706 static int cfqg_print_leaf_weight_device(struct seq_file *sf, void *v)
1707 {
1708         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1709                           cfqg_prfill_leaf_weight_device, &blkcg_policy_cfq,
1710                           0, false);
1711         return 0;
1712 }
1713
1714 static int cfq_print_weight(struct seq_file *sf, void *v)
1715 {
1716         struct blkcg *blkcg = css_to_blkcg(seq_css(sf));
1717         struct cfq_group_data *cgd = blkcg_to_cfqgd(blkcg);
1718         unsigned int val = 0;
1719
1720         if (cgd)
1721                 val = cgd->weight;
1722
1723         seq_printf(sf, "%u\n", val);
1724         return 0;
1725 }
1726
1727 static int cfq_print_leaf_weight(struct seq_file *sf, void *v)
1728 {
1729         struct blkcg *blkcg = css_to_blkcg(seq_css(sf));
1730         struct cfq_group_data *cgd = blkcg_to_cfqgd(blkcg);
1731         unsigned int val = 0;
1732
1733         if (cgd)
1734                 val = cgd->leaf_weight;
1735
1736         seq_printf(sf, "%u\n", val);
1737         return 0;
1738 }
1739
1740 static ssize_t __cfqg_set_weight_device(struct kernfs_open_file *of,
1741                                         char *buf, size_t nbytes, loff_t off,
1742                                         bool is_leaf_weight)
1743 {
1744         struct blkcg *blkcg = css_to_blkcg(of_css(of));
1745         struct blkg_conf_ctx ctx;
1746         struct cfq_group *cfqg;
1747         struct cfq_group_data *cfqgd;
1748         int ret;
1749
1750         ret = blkg_conf_prep(blkcg, &blkcg_policy_cfq, buf, &ctx);
1751         if (ret)
1752                 return ret;
1753
1754         ret = -EINVAL;
1755         cfqg = blkg_to_cfqg(ctx.blkg);
1756         cfqgd = blkcg_to_cfqgd(blkcg);
1757         if (!cfqg || !cfqgd)
1758                 goto err;
1759
1760         if (!ctx.v || (ctx.v >= CFQ_WEIGHT_MIN && ctx.v <= CFQ_WEIGHT_MAX)) {
1761                 if (!is_leaf_weight) {
1762                         cfqg->dev_weight = ctx.v;
1763                         cfqg->new_weight = ctx.v ?: cfqgd->weight;
1764                 } else {
1765                         cfqg->dev_leaf_weight = ctx.v;
1766                         cfqg->new_leaf_weight = ctx.v ?: cfqgd->leaf_weight;
1767                 }
1768                 ret = 0;
1769         }
1770
1771 err:
1772         blkg_conf_finish(&ctx);
1773         return ret ?: nbytes;
1774 }
1775
1776 static ssize_t cfqg_set_weight_device(struct kernfs_open_file *of,
1777                                       char *buf, size_t nbytes, loff_t off)
1778 {
1779         return __cfqg_set_weight_device(of, buf, nbytes, off, false);
1780 }
1781
1782 static ssize_t cfqg_set_leaf_weight_device(struct kernfs_open_file *of,
1783                                            char *buf, size_t nbytes, loff_t off)
1784 {
1785         return __cfqg_set_weight_device(of, buf, nbytes, off, true);
1786 }
1787
1788 static int __cfq_set_weight(struct cgroup_subsys_state *css, struct cftype *cft,
1789                             u64 val, bool is_leaf_weight)
1790 {
1791         struct blkcg *blkcg = css_to_blkcg(css);
1792         struct blkcg_gq *blkg;
1793         struct cfq_group_data *cfqgd;
1794         int ret = 0;
1795
1796         if (val < CFQ_WEIGHT_MIN || val > CFQ_WEIGHT_MAX)
1797                 return -EINVAL;
1798
1799         spin_lock_irq(&blkcg->lock);
1800         cfqgd = blkcg_to_cfqgd(blkcg);
1801         if (!cfqgd) {
1802                 ret = -EINVAL;
1803                 goto out;
1804         }
1805
1806         if (!is_leaf_weight)
1807                 cfqgd->weight = val;
1808         else
1809                 cfqgd->leaf_weight = val;
1810
1811         hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
1812                 struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1813
1814                 if (!cfqg)
1815                         continue;
1816
1817                 if (!is_leaf_weight) {
1818                         if (!cfqg->dev_weight)
1819                                 cfqg->new_weight = cfqgd->weight;
1820                 } else {
1821                         if (!cfqg->dev_leaf_weight)
1822                                 cfqg->new_leaf_weight = cfqgd->leaf_weight;
1823                 }
1824         }
1825
1826 out:
1827         spin_unlock_irq(&blkcg->lock);
1828         return ret;
1829 }
1830
1831 static int cfq_set_weight(struct cgroup_subsys_state *css, struct cftype *cft,
1832                           u64 val)
1833 {
1834         return __cfq_set_weight(css, cft, val, false);
1835 }
1836
1837 static int cfq_set_leaf_weight(struct cgroup_subsys_state *css,
1838                                struct cftype *cft, u64 val)
1839 {
1840         return __cfq_set_weight(css, cft, val, true);
1841 }
1842
1843 static int cfqg_print_stat(struct seq_file *sf, void *v)
1844 {
1845         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), blkg_prfill_stat,
1846                           &blkcg_policy_cfq, seq_cft(sf)->private, false);
1847         return 0;
1848 }
1849
1850 static int cfqg_print_rwstat(struct seq_file *sf, void *v)
1851 {
1852         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), blkg_prfill_rwstat,
1853                           &blkcg_policy_cfq, seq_cft(sf)->private, true);
1854         return 0;
1855 }
1856
1857 static u64 cfqg_prfill_stat_recursive(struct seq_file *sf,
1858                                       struct blkg_policy_data *pd, int off)
1859 {
1860         u64 sum = cfqg_stat_pd_recursive_sum(pd, off);
1861
1862         return __blkg_prfill_u64(sf, pd, sum);
1863 }
1864
1865 static u64 cfqg_prfill_rwstat_recursive(struct seq_file *sf,
1866                                         struct blkg_policy_data *pd, int off)
1867 {
1868         struct blkg_rwstat sum = cfqg_rwstat_pd_recursive_sum(pd, off);
1869
1870         return __blkg_prfill_rwstat(sf, pd, &sum);
1871 }
1872
1873 static int cfqg_print_stat_recursive(struct seq_file *sf, void *v)
1874 {
1875         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1876                           cfqg_prfill_stat_recursive, &blkcg_policy_cfq,
1877                           seq_cft(sf)->private, false);
1878         return 0;
1879 }
1880
1881 static int cfqg_print_rwstat_recursive(struct seq_file *sf, void *v)
1882 {
1883         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1884                           cfqg_prfill_rwstat_recursive, &blkcg_policy_cfq,
1885                           seq_cft(sf)->private, true);
1886         return 0;
1887 }
1888
1889 #ifdef CONFIG_DEBUG_BLK_CGROUP
1890 static u64 cfqg_prfill_avg_queue_size(struct seq_file *sf,
1891                                       struct blkg_policy_data *pd, int off)
1892 {
1893         struct cfq_group *cfqg = pd_to_cfqg(pd);
1894         u64 samples = blkg_stat_read(&cfqg->stats.avg_queue_size_samples);
1895         u64 v = 0;
1896
1897         if (samples) {
1898                 v = blkg_stat_read(&cfqg->stats.avg_queue_size_sum);
1899                 v = div64_u64(v, samples);
1900         }
1901         __blkg_prfill_u64(sf, pd, v);
1902         return 0;
1903 }
1904
1905 /* print avg_queue_size */
1906 static int cfqg_print_avg_queue_size(struct seq_file *sf, void *v)
1907 {
1908         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1909                           cfqg_prfill_avg_queue_size, &blkcg_policy_cfq,
1910                           0, false);
1911         return 0;
1912 }
1913 #endif  /* CONFIG_DEBUG_BLK_CGROUP */
1914
1915 static struct cftype cfq_blkcg_files[] = {
1916         /* on root, weight is mapped to leaf_weight */
1917         {
1918                 .name = "weight_device",
1919                 .flags = CFTYPE_ONLY_ON_ROOT,
1920                 .seq_show = cfqg_print_leaf_weight_device,
1921                 .write = cfqg_set_leaf_weight_device,
1922         },
1923         {
1924                 .name = "weight",
1925                 .flags = CFTYPE_ONLY_ON_ROOT,
1926                 .seq_show = cfq_print_leaf_weight,
1927                 .write_u64 = cfq_set_leaf_weight,
1928         },
1929
1930         /* no such mapping necessary for !roots */
1931         {
1932                 .name = "weight_device",
1933                 .flags = CFTYPE_NOT_ON_ROOT,
1934                 .seq_show = cfqg_print_weight_device,
1935                 .write = cfqg_set_weight_device,
1936         },
1937         {
1938                 .name = "weight",
1939                 .flags = CFTYPE_NOT_ON_ROOT,
1940                 .seq_show = cfq_print_weight,
1941                 .write_u64 = cfq_set_weight,
1942         },
1943
1944         {
1945                 .name = "leaf_weight_device",
1946                 .seq_show = cfqg_print_leaf_weight_device,
1947                 .write = cfqg_set_leaf_weight_device,
1948         },
1949         {
1950                 .name = "leaf_weight",
1951                 .seq_show = cfq_print_leaf_weight,
1952                 .write_u64 = cfq_set_leaf_weight,
1953         },
1954
1955         /* statistics, covers only the tasks in the cfqg */
1956         {
1957                 .name = "time",
1958                 .private = offsetof(struct cfq_group, stats.time),
1959                 .seq_show = cfqg_print_stat,
1960         },
1961         {
1962                 .name = "sectors",
1963                 .private = offsetof(struct cfq_group, stats.sectors),
1964                 .seq_show = cfqg_print_stat,
1965         },
1966         {
1967                 .name = "io_service_bytes",
1968                 .private = offsetof(struct cfq_group, stats.service_bytes),
1969                 .seq_show = cfqg_print_rwstat,
1970         },
1971         {
1972                 .name = "io_serviced",
1973                 .private = offsetof(struct cfq_group, stats.serviced),
1974                 .seq_show = cfqg_print_rwstat,
1975         },
1976         {
1977                 .name = "io_service_time",
1978                 .private = offsetof(struct cfq_group, stats.service_time),
1979                 .seq_show = cfqg_print_rwstat,
1980         },
1981         {
1982                 .name = "io_wait_time",
1983                 .private = offsetof(struct cfq_group, stats.wait_time),
1984                 .seq_show = cfqg_print_rwstat,
1985         },
1986         {
1987                 .name = "io_merged",
1988                 .private = offsetof(struct cfq_group, stats.merged),
1989                 .seq_show = cfqg_print_rwstat,
1990         },
1991         {
1992                 .name = "io_queued",
1993                 .private = offsetof(struct cfq_group, stats.queued),
1994                 .seq_show = cfqg_print_rwstat,
1995         },
1996
1997         /* the same statictics which cover the cfqg and its descendants */
1998         {
1999                 .name = "time_recursive",
2000                 .private = offsetof(struct cfq_group, stats.time),
2001                 .seq_show = cfqg_print_stat_recursive,
2002         },
2003         {
2004                 .name = "sectors_recursive",
2005                 .private = offsetof(struct cfq_group, stats.sectors),
2006                 .seq_show = cfqg_print_stat_recursive,
2007         },
2008         {
2009                 .name = "io_service_bytes_recursive",
2010                 .private = offsetof(struct cfq_group, stats.service_bytes),
2011                 .seq_show = cfqg_print_rwstat_recursive,
2012         },
2013         {
2014                 .name = "io_serviced_recursive",
2015                 .private = offsetof(struct cfq_group, stats.serviced),
2016                 .seq_show = cfqg_print_rwstat_recursive,
2017         },
2018         {
2019                 .name = "io_service_time_recursive",
2020                 .private = offsetof(struct cfq_group, stats.service_time),
2021                 .seq_show = cfqg_print_rwstat_recursive,
2022         },
2023         {
2024                 .name = "io_wait_time_recursive",
2025                 .private = offsetof(struct cfq_group, stats.wait_time),
2026                 .seq_show = cfqg_print_rwstat_recursive,
2027         },
2028         {
2029                 .name = "io_merged_recursive",
2030                 .private = offsetof(struct cfq_group, stats.merged),
2031                 .seq_show = cfqg_print_rwstat_recursive,
2032         },
2033         {
2034                 .name = "io_queued_recursive",
2035                 .private = offsetof(struct cfq_group, stats.queued),
2036                 .seq_show = cfqg_print_rwstat_recursive,
2037         },
2038 #ifdef CONFIG_DEBUG_BLK_CGROUP
2039         {
2040                 .name = "avg_queue_size",
2041                 .seq_show = cfqg_print_avg_queue_size,
2042         },
2043         {
2044                 .name = "group_wait_time",
2045                 .private = offsetof(struct cfq_group, stats.group_wait_time),
2046                 .seq_show = cfqg_print_stat,
2047         },
2048         {
2049                 .name = "idle_time",
2050                 .private = offsetof(struct cfq_group, stats.idle_time),
2051                 .seq_show = cfqg_print_stat,
2052         },
2053         {
2054                 .name = "empty_time",
2055                 .private = offsetof(struct cfq_group, stats.empty_time),
2056                 .seq_show = cfqg_print_stat,
2057         },
2058         {
2059                 .name = "dequeue",
2060                 .private = offsetof(struct cfq_group, stats.dequeue),
2061                 .seq_show = cfqg_print_stat,
2062         },
2063         {
2064                 .name = "unaccounted_time",
2065                 .private = offsetof(struct cfq_group, stats.unaccounted_time),
2066                 .seq_show = cfqg_print_stat,
2067         },
2068 #endif  /* CONFIG_DEBUG_BLK_CGROUP */
2069         { }     /* terminate */
2070 };
2071 #else /* GROUP_IOSCHED */
2072 static struct cfq_group *cfq_lookup_create_cfqg(struct cfq_data *cfqd,
2073                                                 struct blkcg *blkcg)
2074 {
2075         return cfqd->root_group;
2076 }
2077
2078 static inline void
2079 cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg) {
2080         cfqq->cfqg = cfqg;
2081 }
2082
2083 #endif /* GROUP_IOSCHED */
2084
2085 /*
2086  * The cfqd->service_trees holds all pending cfq_queue's that have
2087  * requests waiting to be processed. It is sorted in the order that
2088  * we will service the queues.
2089  */
2090 static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2091                                  bool add_front)
2092 {
2093         struct rb_node **p, *parent;
2094         struct cfq_queue *__cfqq;
2095         unsigned long rb_key;
2096         struct cfq_rb_root *st;
2097         int left;
2098         int new_cfqq = 1;
2099
2100         st = st_for(cfqq->cfqg, cfqq_class(cfqq), cfqq_type(cfqq));
2101         if (cfq_class_idle(cfqq)) {
2102                 rb_key = CFQ_IDLE_DELAY;
2103                 parent = rb_last(&st->rb);
2104                 if (parent && parent != &cfqq->rb_node) {
2105                         __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2106                         rb_key += __cfqq->rb_key;
2107                 } else
2108                         rb_key += jiffies;
2109         } else if (!add_front) {
2110                 /*
2111                  * Get our rb key offset. Subtract any residual slice
2112                  * value carried from last service. A negative resid
2113                  * count indicates slice overrun, and this should position
2114                  * the next service time further away in the tree.
2115                  */
2116                 rb_key = cfq_slice_offset(cfqd, cfqq) + jiffies;
2117                 rb_key -= cfqq->slice_resid;
2118                 cfqq->slice_resid = 0;
2119         } else {
2120                 rb_key = -HZ;
2121                 __cfqq = cfq_rb_first(st);
2122                 rb_key += __cfqq ? __cfqq->rb_key : jiffies;
2123         }
2124
2125         if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2126                 new_cfqq = 0;
2127                 /*
2128                  * same position, nothing more to do
2129                  */
2130                 if (rb_key == cfqq->rb_key && cfqq->service_tree == st)
2131                         return;
2132
2133                 cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2134                 cfqq->service_tree = NULL;
2135         }
2136
2137         left = 1;
2138         parent = NULL;
2139         cfqq->service_tree = st;
2140         p = &st->rb.rb_node;
2141         while (*p) {
2142                 parent = *p;
2143                 __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2144
2145                 /*
2146                  * sort by key, that represents service time.
2147                  */
2148                 if (time_before(rb_key, __cfqq->rb_key))
2149                         p = &parent->rb_left;
2150                 else {
2151                         p = &parent->rb_right;
2152                         left = 0;
2153                 }
2154         }
2155
2156         if (left)
2157                 st->left = &cfqq->rb_node;
2158
2159         cfqq->rb_key = rb_key;
2160         rb_link_node(&cfqq->rb_node, parent, p);
2161         rb_insert_color(&cfqq->rb_node, &st->rb);
2162         st->count++;
2163         if (add_front || !new_cfqq)
2164                 return;
2165         cfq_group_notify_queue_add(cfqd, cfqq->cfqg);
2166 }
2167
2168 static struct cfq_queue *
2169 cfq_prio_tree_lookup(struct cfq_data *cfqd, struct rb_root *root,
2170                      sector_t sector, struct rb_node **ret_parent,
2171                      struct rb_node ***rb_link)
2172 {
2173         struct rb_node **p, *parent;
2174         struct cfq_queue *cfqq = NULL;
2175
2176         parent = NULL;
2177         p = &root->rb_node;
2178         while (*p) {
2179                 struct rb_node **n;
2180
2181                 parent = *p;
2182                 cfqq = rb_entry(parent, struct cfq_queue, p_node);
2183
2184                 /*
2185                  * Sort strictly based on sector.  Smallest to the left,
2186                  * largest to the right.
2187                  */
2188                 if (sector > blk_rq_pos(cfqq->next_rq))
2189                         n = &(*p)->rb_right;
2190                 else if (sector < blk_rq_pos(cfqq->next_rq))
2191                         n = &(*p)->rb_left;
2192                 else
2193                         break;
2194                 p = n;
2195                 cfqq = NULL;
2196         }
2197
2198         *ret_parent = parent;
2199         if (rb_link)
2200                 *rb_link = p;
2201         return cfqq;
2202 }
2203
2204 static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2205 {
2206         struct rb_node **p, *parent;
2207         struct cfq_queue *__cfqq;
2208
2209         if (cfqq->p_root) {
2210                 rb_erase(&cfqq->p_node, cfqq->p_root);
2211                 cfqq->p_root = NULL;
2212         }
2213
2214         if (cfq_class_idle(cfqq))
2215                 return;
2216         if (!cfqq->next_rq)
2217                 return;
2218
2219         cfqq->p_root = &cfqd->prio_trees[cfqq->org_ioprio];
2220         __cfqq = cfq_prio_tree_lookup(cfqd, cfqq->p_root,
2221                                       blk_rq_pos(cfqq->next_rq), &parent, &p);
2222         if (!__cfqq) {
2223                 rb_link_node(&cfqq->p_node, parent, p);
2224                 rb_insert_color(&cfqq->p_node, cfqq->p_root);
2225         } else
2226                 cfqq->p_root = NULL;
2227 }
2228
2229 /*
2230  * Update cfqq's position in the service tree.
2231  */
2232 static void cfq_resort_rr_list(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2233 {
2234         /*
2235          * Resorting requires the cfqq to be on the RR list already.
2236          */
2237         if (cfq_cfqq_on_rr(cfqq)) {
2238                 cfq_service_tree_add(cfqd, cfqq, 0);
2239                 cfq_prio_tree_add(cfqd, cfqq);
2240         }
2241 }
2242
2243 /*
2244  * add to busy list of queues for service, trying to be fair in ordering
2245  * the pending list according to last request service
2246  */
2247 static void cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2248 {
2249         cfq_log_cfqq(cfqd, cfqq, "add_to_rr");
2250         BUG_ON(cfq_cfqq_on_rr(cfqq));
2251         cfq_mark_cfqq_on_rr(cfqq);
2252         cfqd->busy_queues++;
2253         if (cfq_cfqq_sync(cfqq))
2254                 cfqd->busy_sync_queues++;
2255
2256         cfq_resort_rr_list(cfqd, cfqq);
2257 }
2258
2259 /*
2260  * Called when the cfqq no longer has requests pending, remove it from
2261  * the service tree.
2262  */
2263 static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2264 {
2265         cfq_log_cfqq(cfqd, cfqq, "del_from_rr");
2266         BUG_ON(!cfq_cfqq_on_rr(cfqq));
2267         cfq_clear_cfqq_on_rr(cfqq);
2268
2269         if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2270                 cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2271                 cfqq->service_tree = NULL;
2272         }
2273         if (cfqq->p_root) {
2274                 rb_erase(&cfqq->p_node, cfqq->p_root);
2275                 cfqq->p_root = NULL;
2276         }
2277
2278         cfq_group_notify_queue_del(cfqd, cfqq->cfqg);
2279         BUG_ON(!cfqd->busy_queues);
2280         cfqd->busy_queues--;
2281         if (cfq_cfqq_sync(cfqq))
2282                 cfqd->busy_sync_queues--;
2283 }
2284
2285 /*
2286  * rb tree support functions
2287  */
2288 static void cfq_del_rq_rb(struct request *rq)
2289 {
2290         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2291         const int sync = rq_is_sync(rq);
2292
2293         BUG_ON(!cfqq->queued[sync]);
2294         cfqq->queued[sync]--;
2295
2296         elv_rb_del(&cfqq->sort_list, rq);
2297
2298         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list)) {
2299                 /*
2300                  * Queue will be deleted from service tree when we actually
2301                  * expire it later. Right now just remove it from prio tree
2302                  * as it is empty.
2303                  */
2304                 if (cfqq->p_root) {
2305                         rb_erase(&cfqq->p_node, cfqq->p_root);
2306                         cfqq->p_root = NULL;
2307                 }
2308         }
2309 }
2310
2311 static void cfq_add_rq_rb(struct request *rq)
2312 {
2313         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2314         struct cfq_data *cfqd = cfqq->cfqd;
2315         struct request *prev;
2316
2317         cfqq->queued[rq_is_sync(rq)]++;
2318
2319         elv_rb_add(&cfqq->sort_list, rq);
2320
2321         if (!cfq_cfqq_on_rr(cfqq))
2322                 cfq_add_cfqq_rr(cfqd, cfqq);
2323
2324         /*
2325          * check if this request is a better next-serve candidate
2326          */
2327         prev = cfqq->next_rq;
2328         cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq, cfqd->last_position);
2329
2330         /*
2331          * adjust priority tree position, if ->next_rq changes
2332          */
2333         if (prev != cfqq->next_rq)
2334                 cfq_prio_tree_add(cfqd, cfqq);
2335
2336         BUG_ON(!cfqq->next_rq);
2337 }
2338
2339 static void cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
2340 {
2341         elv_rb_del(&cfqq->sort_list, rq);
2342         cfqq->queued[rq_is_sync(rq)]--;
2343         cfqg_stats_update_io_remove(RQ_CFQG(rq), rq->cmd_flags);
2344         cfq_add_rq_rb(rq);
2345         cfqg_stats_update_io_add(RQ_CFQG(rq), cfqq->cfqd->serving_group,
2346                                  rq->cmd_flags);
2347 }
2348
2349 static struct request *
2350 cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio)
2351 {
2352         struct task_struct *tsk = current;
2353         struct cfq_io_cq *cic;
2354         struct cfq_queue *cfqq;
2355
2356         cic = cfq_cic_lookup(cfqd, tsk->io_context);
2357         if (!cic)
2358                 return NULL;
2359
2360         cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
2361         if (cfqq)
2362                 return elv_rb_find(&cfqq->sort_list, bio_end_sector(bio));
2363
2364         return NULL;
2365 }
2366
2367 static void cfq_activate_request(struct request_queue *q, struct request *rq)
2368 {
2369         struct cfq_data *cfqd = q->elevator->elevator_data;
2370
2371         cfqd->rq_in_driver++;
2372         cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "activate rq, drv=%d",
2373                                                 cfqd->rq_in_driver);
2374
2375         cfqd->last_position = blk_rq_pos(rq) + blk_rq_sectors(rq);
2376 }
2377
2378 static void cfq_deactivate_request(struct request_queue *q, struct request *rq)
2379 {
2380         struct cfq_data *cfqd = q->elevator->elevator_data;
2381
2382         WARN_ON(!cfqd->rq_in_driver);
2383         cfqd->rq_in_driver--;
2384         cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "deactivate rq, drv=%d",
2385                                                 cfqd->rq_in_driver);
2386 }
2387
2388 static void cfq_remove_request(struct request *rq)
2389 {
2390         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2391
2392         if (cfqq->next_rq == rq)
2393                 cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq);
2394
2395         list_del_init(&rq->queuelist);
2396         cfq_del_rq_rb(rq);
2397
2398         cfqq->cfqd->rq_queued--;
2399         cfqg_stats_update_io_remove(RQ_CFQG(rq), rq->cmd_flags);
2400         if (rq->cmd_flags & REQ_PRIO) {
2401                 WARN_ON(!cfqq->prio_pending);
2402                 cfqq->prio_pending--;
2403         }
2404 }
2405
2406 static int cfq_merge(struct request_queue *q, struct request **req,
2407                      struct bio *bio)
2408 {
2409         struct cfq_data *cfqd = q->elevator->elevator_data;
2410         struct request *__rq;
2411
2412         __rq = cfq_find_rq_fmerge(cfqd, bio);
2413         if (__rq && elv_rq_merge_ok(__rq, bio)) {
2414                 *req = __rq;
2415                 return ELEVATOR_FRONT_MERGE;
2416         }
2417
2418         return ELEVATOR_NO_MERGE;
2419 }
2420
2421 static void cfq_merged_request(struct request_queue *q, struct request *req,
2422                                int type)
2423 {
2424         if (type == ELEVATOR_FRONT_MERGE) {
2425                 struct cfq_queue *cfqq = RQ_CFQQ(req);
2426
2427                 cfq_reposition_rq_rb(cfqq, req);
2428         }
2429 }
2430
2431 static void cfq_bio_merged(struct request_queue *q, struct request *req,
2432                                 struct bio *bio)
2433 {
2434         cfqg_stats_update_io_merged(RQ_CFQG(req), bio->bi_rw);
2435 }
2436
2437 static void
2438 cfq_merged_requests(struct request_queue *q, struct request *rq,
2439                     struct request *next)
2440 {
2441         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2442         struct cfq_data *cfqd = q->elevator->elevator_data;
2443
2444         /*
2445          * reposition in fifo if next is older than rq
2446          */
2447         if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
2448             time_before(next->fifo_time, rq->fifo_time) &&
2449             cfqq == RQ_CFQQ(next)) {
2450                 list_move(&rq->queuelist, &next->queuelist);
2451                 rq->fifo_time = next->fifo_time;
2452         }
2453
2454         if (cfqq->next_rq == next)
2455                 cfqq->next_rq = rq;
2456         cfq_remove_request(next);
2457         cfqg_stats_update_io_merged(RQ_CFQG(rq), next->cmd_flags);
2458
2459         cfqq = RQ_CFQQ(next);
2460         /*
2461          * all requests of this queue are merged to other queues, delete it
2462          * from the service tree. If it's the active_queue,
2463          * cfq_dispatch_requests() will choose to expire it or do idle
2464          */
2465         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list) &&
2466             cfqq != cfqd->active_queue)
2467                 cfq_del_cfqq_rr(cfqd, cfqq);
2468 }
2469
2470 static int cfq_allow_merge(struct request_queue *q, struct request *rq,
2471                            struct bio *bio)
2472 {
2473         struct cfq_data *cfqd = q->elevator->elevator_data;
2474         struct cfq_io_cq *cic;
2475         struct cfq_queue *cfqq;
2476
2477         /*
2478          * Disallow merge of a sync bio into an async request.
2479          */
2480         if (cfq_bio_sync(bio) && !rq_is_sync(rq))
2481                 return false;
2482
2483         /*
2484          * Lookup the cfqq that this bio will be queued with and allow
2485          * merge only if rq is queued there.
2486          */
2487         cic = cfq_cic_lookup(cfqd, current->io_context);
2488         if (!cic)
2489                 return false;
2490
2491         cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
2492         return cfqq == RQ_CFQQ(rq);
2493 }
2494
2495 static inline void cfq_del_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2496 {
2497         del_timer(&cfqd->idle_slice_timer);
2498         cfqg_stats_update_idle_time(cfqq->cfqg);
2499 }
2500
2501 static void __cfq_set_active_queue(struct cfq_data *cfqd,
2502                                    struct cfq_queue *cfqq)
2503 {
2504         if (cfqq) {
2505                 cfq_log_cfqq(cfqd, cfqq, "set_active wl_class:%d wl_type:%d",
2506                                 cfqd->serving_wl_class, cfqd->serving_wl_type);
2507                 cfqg_stats_update_avg_queue_size(cfqq->cfqg);
2508                 cfqq->slice_start = 0;
2509                 cfqq->dispatch_start = jiffies;
2510                 cfqq->allocated_slice = 0;
2511                 cfqq->slice_end = 0;
2512                 cfqq->slice_dispatch = 0;
2513                 cfqq->nr_sectors = 0;
2514
2515                 cfq_clear_cfqq_wait_request(cfqq);
2516                 cfq_clear_cfqq_must_dispatch(cfqq);
2517                 cfq_clear_cfqq_must_alloc_slice(cfqq);
2518                 cfq_clear_cfqq_fifo_expire(cfqq);
2519                 cfq_mark_cfqq_slice_new(cfqq);
2520
2521                 cfq_del_timer(cfqd, cfqq);
2522         }
2523
2524         cfqd->active_queue = cfqq;
2525 }
2526
2527 /*
2528  * current cfqq expired its slice (or was too idle), select new one
2529  */
2530 static void
2531 __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2532                     bool timed_out)
2533 {
2534         cfq_log_cfqq(cfqd, cfqq, "slice expired t=%d", timed_out);
2535
2536         if (cfq_cfqq_wait_request(cfqq))
2537                 cfq_del_timer(cfqd, cfqq);
2538
2539         cfq_clear_cfqq_wait_request(cfqq);
2540         cfq_clear_cfqq_wait_busy(cfqq);
2541
2542         /*
2543          * If this cfqq is shared between multiple processes, check to
2544          * make sure that those processes are still issuing I/Os within
2545          * the mean seek distance.  If not, it may be time to break the
2546          * queues apart again.
2547          */
2548         if (cfq_cfqq_coop(cfqq) && CFQQ_SEEKY(cfqq))
2549                 cfq_mark_cfqq_split_coop(cfqq);
2550
2551         /*
2552          * store what was left of this slice, if the queue idled/timed out
2553          */
2554         if (timed_out) {
2555                 if (cfq_cfqq_slice_new(cfqq))
2556                         cfqq->slice_resid = cfq_scaled_cfqq_slice(cfqd, cfqq);
2557                 else
2558                         cfqq->slice_resid = cfqq->slice_end - jiffies;
2559                 cfq_log_cfqq(cfqd, cfqq, "resid=%ld", cfqq->slice_resid);
2560         }
2561
2562         cfq_group_served(cfqd, cfqq->cfqg, cfqq);
2563
2564         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list))
2565                 cfq_del_cfqq_rr(cfqd, cfqq);
2566
2567         cfq_resort_rr_list(cfqd, cfqq);
2568
2569         if (cfqq == cfqd->active_queue)
2570                 cfqd->active_queue = NULL;
2571
2572         if (cfqd->active_cic) {
2573                 put_io_context(cfqd->active_cic->icq.ioc);
2574                 cfqd->active_cic = NULL;
2575         }
2576 }
2577
2578 static inline void cfq_slice_expired(struct cfq_data *cfqd, bool timed_out)
2579 {
2580         struct cfq_queue *cfqq = cfqd->active_queue;
2581
2582         if (cfqq)
2583                 __cfq_slice_expired(cfqd, cfqq, timed_out);
2584 }
2585
2586 /*
2587  * Get next queue for service. Unless we have a queue preemption,
2588  * we'll simply select the first cfqq in the service tree.
2589  */
2590 static struct cfq_queue *cfq_get_next_queue(struct cfq_data *cfqd)
2591 {
2592         struct cfq_rb_root *st = st_for(cfqd->serving_group,
2593                         cfqd->serving_wl_class, cfqd->serving_wl_type);
2594
2595         if (!cfqd->rq_queued)
2596                 return NULL;
2597
2598         /* There is nothing to dispatch */
2599         if (!st)
2600                 return NULL;
2601         if (RB_EMPTY_ROOT(&st->rb))
2602                 return NULL;
2603         return cfq_rb_first(st);
2604 }
2605
2606 static struct cfq_queue *cfq_get_next_queue_forced(struct cfq_data *cfqd)
2607 {
2608         struct cfq_group *cfqg;
2609         struct cfq_queue *cfqq;
2610         int i, j;
2611         struct cfq_rb_root *st;
2612
2613         if (!cfqd->rq_queued)
2614                 return NULL;
2615
2616         cfqg = cfq_get_next_cfqg(cfqd);
2617         if (!cfqg)
2618                 return NULL;
2619
2620         for_each_cfqg_st(cfqg, i, j, st)
2621                 if ((cfqq = cfq_rb_first(st)) != NULL)
2622                         return cfqq;
2623         return NULL;
2624 }
2625
2626 /*
2627  * Get and set a new active queue for service.
2628  */
2629 static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd,
2630                                               struct cfq_queue *cfqq)
2631 {
2632         if (!cfqq)
2633                 cfqq = cfq_get_next_queue(cfqd);
2634
2635         __cfq_set_active_queue(cfqd, cfqq);
2636         return cfqq;
2637 }
2638
2639 static inline sector_t cfq_dist_from_last(struct cfq_data *cfqd,
2640                                           struct request *rq)
2641 {
2642         if (blk_rq_pos(rq) >= cfqd->last_position)
2643                 return blk_rq_pos(rq) - cfqd->last_position;
2644         else
2645                 return cfqd->last_position - blk_rq_pos(rq);
2646 }
2647
2648 static inline int cfq_rq_close(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2649                                struct request *rq)
2650 {
2651         return cfq_dist_from_last(cfqd, rq) <= CFQQ_CLOSE_THR;
2652 }
2653
2654 static struct cfq_queue *cfqq_close(struct cfq_data *cfqd,
2655                                     struct cfq_queue *cur_cfqq)
2656 {
2657         struct rb_root *root = &cfqd->prio_trees[cur_cfqq->org_ioprio];
2658         struct rb_node *parent, *node;
2659         struct cfq_queue *__cfqq;
2660         sector_t sector = cfqd->last_position;
2661
2662         if (RB_EMPTY_ROOT(root))
2663                 return NULL;
2664
2665         /*
2666          * First, if we find a request starting at the end of the last
2667          * request, choose it.
2668          */
2669         __cfqq = cfq_prio_tree_lookup(cfqd, root, sector, &parent, NULL);
2670         if (__cfqq)
2671                 return __cfqq;
2672
2673         /*
2674          * If the exact sector wasn't found, the parent of the NULL leaf
2675          * will contain the closest sector.
2676          */
2677         __cfqq = rb_entry(parent, struct cfq_queue, p_node);
2678         if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2679                 return __cfqq;
2680
2681         if (blk_rq_pos(__cfqq->next_rq) < sector)
2682                 node = rb_next(&__cfqq->p_node);
2683         else
2684                 node = rb_prev(&__cfqq->p_node);
2685         if (!node)
2686                 return NULL;
2687
2688         __cfqq = rb_entry(node, struct cfq_queue, p_node);
2689         if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2690                 return __cfqq;
2691
2692         return NULL;
2693 }
2694
2695 /*
2696  * cfqd - obvious
2697  * cur_cfqq - passed in so that we don't decide that the current queue is
2698  *            closely cooperating with itself.
2699  *
2700  * So, basically we're assuming that that cur_cfqq has dispatched at least
2701  * one request, and that cfqd->last_position reflects a position on the disk
2702  * associated with the I/O issued by cur_cfqq.  I'm not sure this is a valid
2703  * assumption.
2704  */
2705 static struct cfq_queue *cfq_close_cooperator(struct cfq_data *cfqd,
2706                                               struct cfq_queue *cur_cfqq)
2707 {
2708         struct cfq_queue *cfqq;
2709
2710         if (cfq_class_idle(cur_cfqq))
2711                 return NULL;
2712         if (!cfq_cfqq_sync(cur_cfqq))
2713                 return NULL;
2714         if (CFQQ_SEEKY(cur_cfqq))
2715                 return NULL;
2716
2717         /*
2718          * Don't search priority tree if it's the only queue in the group.
2719          */
2720         if (cur_cfqq->cfqg->nr_cfqq == 1)
2721                 return NULL;
2722
2723         /*
2724          * We should notice if some of the queues are cooperating, eg
2725          * working closely on the same area of the disk. In that case,
2726          * we can group them together and don't waste time idling.
2727          */
2728         cfqq = cfqq_close(cfqd, cur_cfqq);
2729         if (!cfqq)
2730                 return NULL;
2731
2732         /* If new queue belongs to different cfq_group, don't choose it */
2733         if (cur_cfqq->cfqg != cfqq->cfqg)
2734                 return NULL;
2735
2736         /*
2737          * It only makes sense to merge sync queues.
2738          */
2739         if (!cfq_cfqq_sync(cfqq))
2740                 return NULL;
2741         if (CFQQ_SEEKY(cfqq))
2742                 return NULL;
2743
2744         /*
2745          * Do not merge queues of different priority classes
2746          */
2747         if (cfq_class_rt(cfqq) != cfq_class_rt(cur_cfqq))
2748                 return NULL;
2749
2750         return cfqq;
2751 }
2752
2753 /*
2754  * Determine whether we should enforce idle window for this queue.
2755  */
2756
2757 static bool cfq_should_idle(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2758 {
2759         enum wl_class_t wl_class = cfqq_class(cfqq);
2760         struct cfq_rb_root *st = cfqq->service_tree;
2761
2762         BUG_ON(!st);
2763         BUG_ON(!st->count);
2764
2765         if (!cfqd->cfq_slice_idle)
2766                 return false;
2767
2768         /* We never do for idle class queues. */
2769         if (wl_class == IDLE_WORKLOAD)
2770                 return false;
2771
2772         /* We do for queues that were marked with idle window flag. */
2773         if (cfq_cfqq_idle_window(cfqq) &&
2774            !(blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag))
2775                 return true;
2776
2777         /*
2778          * Otherwise, we do only if they are the last ones
2779          * in their service tree.
2780          */
2781         if (st->count == 1 && cfq_cfqq_sync(cfqq) &&
2782            !cfq_io_thinktime_big(cfqd, &st->ttime, false))
2783                 return true;
2784         cfq_log_cfqq(cfqd, cfqq, "Not idling. st->count:%d", st->count);
2785         return false;
2786 }
2787
2788 static void cfq_arm_slice_timer(struct cfq_data *cfqd)
2789 {
2790         struct cfq_queue *cfqq = cfqd->active_queue;
2791         struct cfq_io_cq *cic;
2792         unsigned long sl, group_idle = 0;
2793
2794         /*
2795          * SSD device without seek penalty, disable idling. But only do so
2796          * for devices that support queuing, otherwise we still have a problem
2797          * with sync vs async workloads.
2798          */
2799         if (blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag)
2800                 return;
2801
2802         WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list));
2803         WARN_ON(cfq_cfqq_slice_new(cfqq));
2804
2805         /*
2806          * idle is disabled, either manually or by past process history
2807          */
2808         if (!cfq_should_idle(cfqd, cfqq)) {
2809                 /* no queue idling. Check for group idling */
2810                 if (cfqd->cfq_group_idle)
2811                         group_idle = cfqd->cfq_group_idle;
2812                 else
2813                         return;
2814         }
2815
2816         /*
2817          * still active requests from this queue, don't idle
2818          */
2819         if (cfqq->dispatched)
2820                 return;
2821
2822         /*
2823          * task has exited, don't wait
2824          */
2825         cic = cfqd->active_cic;
2826         if (!cic || !atomic_read(&cic->icq.ioc->active_ref))
2827                 return;
2828
2829         /*
2830          * If our average think time is larger than the remaining time
2831          * slice, then don't idle. This avoids overrunning the allotted
2832          * time slice.
2833          */
2834         if (sample_valid(cic->ttime.ttime_samples) &&
2835             (cfqq->slice_end - jiffies < cic->ttime.ttime_mean)) {
2836                 cfq_log_cfqq(cfqd, cfqq, "Not idling. think_time:%lu",
2837                              cic->ttime.ttime_mean);
2838                 return;
2839         }
2840
2841         /* There are other queues in the group, don't do group idle */
2842         if (group_idle && cfqq->cfqg->nr_cfqq > 1)
2843                 return;
2844
2845         cfq_mark_cfqq_wait_request(cfqq);
2846
2847         if (group_idle)
2848                 sl = cfqd->cfq_group_idle;
2849         else
2850                 sl = cfqd->cfq_slice_idle;
2851
2852         mod_timer(&cfqd->idle_slice_timer, jiffies + sl);
2853         cfqg_stats_set_start_idle_time(cfqq->cfqg);
2854         cfq_log_cfqq(cfqd, cfqq, "arm_idle: %lu group_idle: %d", sl,
2855                         group_idle ? 1 : 0);
2856 }
2857
2858 /*
2859  * Move request from internal lists to the request queue dispatch list.
2860  */
2861 static void cfq_dispatch_insert(struct request_queue *q, struct request *rq)
2862 {
2863         struct cfq_data *cfqd = q->elevator->elevator_data;
2864         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2865
2866         cfq_log_cfqq(cfqd, cfqq, "dispatch_insert");
2867
2868         cfqq->next_rq = cfq_find_next_rq(cfqd, cfqq, rq);
2869         cfq_remove_request(rq);
2870         cfqq->dispatched++;
2871         (RQ_CFQG(rq))->dispatched++;
2872         elv_dispatch_sort(q, rq);
2873
2874         cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]++;
2875         cfqq->nr_sectors += blk_rq_sectors(rq);
2876         cfqg_stats_update_dispatch(cfqq->cfqg, blk_rq_bytes(rq), rq->cmd_flags);
2877 }
2878
2879 /*
2880  * return expired entry, or NULL to just start from scratch in rbtree
2881  */
2882 static struct request *cfq_check_fifo(struct cfq_queue *cfqq)
2883 {
2884         struct request *rq = NULL;
2885
2886         if (cfq_cfqq_fifo_expire(cfqq))
2887                 return NULL;
2888
2889         cfq_mark_cfqq_fifo_expire(cfqq);
2890
2891         if (list_empty(&cfqq->fifo))
2892                 return NULL;
2893
2894         rq = rq_entry_fifo(cfqq->fifo.next);
2895         if (time_before(jiffies, rq->fifo_time))
2896                 rq = NULL;
2897
2898         cfq_log_cfqq(cfqq->cfqd, cfqq, "fifo=%p", rq);
2899         return rq;
2900 }
2901
2902 static inline int
2903 cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2904 {
2905         const int base_rq = cfqd->cfq_slice_async_rq;
2906
2907         WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
2908
2909         return 2 * base_rq * (IOPRIO_BE_NR - cfqq->ioprio);
2910 }
2911
2912 /*
2913  * Must be called with the queue_lock held.
2914  */
2915 static int cfqq_process_refs(struct cfq_queue *cfqq)
2916 {
2917         int process_refs, io_refs;
2918
2919         io_refs = cfqq->allocated[READ] + cfqq->allocated[WRITE];
2920         process_refs = cfqq->ref - io_refs;
2921         BUG_ON(process_refs < 0);
2922         return process_refs;
2923 }
2924
2925 static void cfq_setup_merge(struct cfq_queue *cfqq, struct cfq_queue *new_cfqq)
2926 {
2927         int process_refs, new_process_refs;
2928         struct cfq_queue *__cfqq;
2929
2930         /*
2931          * If there are no process references on the new_cfqq, then it is
2932          * unsafe to follow the ->new_cfqq chain as other cfqq's in the
2933          * chain may have dropped their last reference (not just their
2934          * last process reference).
2935          */
2936         if (!cfqq_process_refs(new_cfqq))
2937                 return;
2938
2939         /* Avoid a circular list and skip interim queue merges */
2940         while ((__cfqq = new_cfqq->new_cfqq)) {
2941                 if (__cfqq == cfqq)
2942                         return;
2943                 new_cfqq = __cfqq;
2944         }
2945
2946         process_refs = cfqq_process_refs(cfqq);
2947         new_process_refs = cfqq_process_refs(new_cfqq);
2948         /*
2949          * If the process for the cfqq has gone away, there is no
2950          * sense in merging the queues.
2951          */
2952         if (process_refs == 0 || new_process_refs == 0)
2953                 return;
2954
2955         /*
2956          * Merge in the direction of the lesser amount of work.
2957          */
2958         if (new_process_refs >= process_refs) {
2959                 cfqq->new_cfqq = new_cfqq;
2960                 new_cfqq->ref += process_refs;
2961         } else {
2962                 new_cfqq->new_cfqq = cfqq;
2963                 cfqq->ref += new_process_refs;
2964         }
2965 }
2966
2967 static enum wl_type_t cfq_choose_wl_type(struct cfq_data *cfqd,
2968                         struct cfq_group *cfqg, enum wl_class_t wl_class)
2969 {
2970         struct cfq_queue *queue;
2971         int i;
2972         bool key_valid = false;
2973         unsigned long lowest_key = 0;
2974         enum wl_type_t cur_best = SYNC_NOIDLE_WORKLOAD;
2975
2976         for (i = 0; i <= SYNC_WORKLOAD; ++i) {
2977                 /* select the one with lowest rb_key */
2978                 queue = cfq_rb_first(st_for(cfqg, wl_class, i));
2979                 if (queue &&
2980                     (!key_valid || time_before(queue->rb_key, lowest_key))) {
2981                         lowest_key = queue->rb_key;
2982                         cur_best = i;
2983                         key_valid = true;
2984                 }
2985         }
2986
2987         return cur_best;
2988 }
2989
2990 static void
2991 choose_wl_class_and_type(struct cfq_data *cfqd, struct cfq_group *cfqg)
2992 {
2993         unsigned slice;
2994         unsigned count;
2995         struct cfq_rb_root *st;
2996         unsigned group_slice;
2997         enum wl_class_t original_class = cfqd->serving_wl_class;
2998
2999         /* Choose next priority. RT > BE > IDLE */
3000         if (cfq_group_busy_queues_wl(RT_WORKLOAD, cfqd, cfqg))
3001                 cfqd->serving_wl_class = RT_WORKLOAD;
3002         else if (cfq_group_busy_queues_wl(BE_WORKLOAD, cfqd, cfqg))
3003                 cfqd->serving_wl_class = BE_WORKLOAD;
3004         else {
3005                 cfqd->serving_wl_class = IDLE_WORKLOAD;
3006                 cfqd->workload_expires = jiffies + 1;
3007                 return;
3008         }
3009
3010         if (original_class != cfqd->serving_wl_class)
3011                 goto new_workload;
3012
3013         /*
3014          * For RT and BE, we have to choose also the type
3015          * (SYNC, SYNC_NOIDLE, ASYNC), and to compute a workload
3016          * expiration time
3017          */
3018         st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
3019         count = st->count;
3020
3021         /*
3022          * check workload expiration, and that we still have other queues ready
3023          */
3024         if (count && !time_after(jiffies, cfqd->workload_expires))
3025                 return;
3026
3027 new_workload:
3028         /* otherwise select new workload type */
3029         cfqd->serving_wl_type = cfq_choose_wl_type(cfqd, cfqg,
3030                                         cfqd->serving_wl_class);
3031         st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
3032         count = st->count;
3033
3034         /*
3035          * the workload slice is computed as a fraction of target latency
3036          * proportional to the number of queues in that workload, over
3037          * all the queues in the same priority class
3038          */
3039         group_slice = cfq_group_slice(cfqd, cfqg);
3040
3041         slice = group_slice * count /
3042                 max_t(unsigned, cfqg->busy_queues_avg[cfqd->serving_wl_class],
3043                       cfq_group_busy_queues_wl(cfqd->serving_wl_class, cfqd,
3044                                         cfqg));
3045
3046         if (cfqd->serving_wl_type == ASYNC_WORKLOAD) {
3047                 unsigned int tmp;
3048
3049                 /*
3050                  * Async queues are currently system wide. Just taking
3051                  * proportion of queues with-in same group will lead to higher
3052                  * async ratio system wide as generally root group is going
3053                  * to have higher weight. A more accurate thing would be to
3054                  * calculate system wide asnc/sync ratio.
3055                  */
3056                 tmp = cfqd->cfq_target_latency *
3057                         cfqg_busy_async_queues(cfqd, cfqg);
3058                 tmp = tmp/cfqd->busy_queues;
3059                 slice = min_t(unsigned, slice, tmp);
3060
3061                 /* async workload slice is scaled down according to
3062                  * the sync/async slice ratio. */
3063                 slice = slice * cfqd->cfq_slice[0] / cfqd->cfq_slice[1];
3064         } else
3065                 /* sync workload slice is at least 2 * cfq_slice_idle */
3066                 slice = max(slice, 2 * cfqd->cfq_slice_idle);
3067
3068         slice = max_t(unsigned, slice, CFQ_MIN_TT);
3069         cfq_log(cfqd, "workload slice:%d", slice);
3070         cfqd->workload_expires = jiffies + slice;
3071 }
3072
3073 static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd)
3074 {
3075         struct cfq_rb_root *st = &cfqd->grp_service_tree;
3076         struct cfq_group *cfqg;
3077
3078         if (RB_EMPTY_ROOT(&st->rb))
3079                 return NULL;
3080         cfqg = cfq_rb_first_group(st);
3081         update_min_vdisktime(st);
3082         return cfqg;
3083 }
3084
3085 static void cfq_choose_cfqg(struct cfq_data *cfqd)
3086 {
3087         struct cfq_group *cfqg = cfq_get_next_cfqg(cfqd);
3088
3089         cfqd->serving_group = cfqg;
3090
3091         /* Restore the workload type data */
3092         if (cfqg->saved_wl_slice) {
3093                 cfqd->workload_expires = jiffies + cfqg->saved_wl_slice;
3094                 cfqd->serving_wl_type = cfqg->saved_wl_type;
3095                 cfqd->serving_wl_class = cfqg->saved_wl_class;
3096         } else
3097                 cfqd->workload_expires = jiffies - 1;
3098
3099         choose_wl_class_and_type(cfqd, cfqg);
3100 }
3101
3102 /*
3103  * Select a queue for service. If we have a current active queue,
3104  * check whether to continue servicing it, or retrieve and set a new one.
3105  */
3106 static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
3107 {
3108         struct cfq_queue *cfqq, *new_cfqq = NULL;
3109
3110         cfqq = cfqd->active_queue;
3111         if (!cfqq)
3112                 goto new_queue;
3113
3114         if (!cfqd->rq_queued)
3115                 return NULL;
3116
3117         /*
3118          * We were waiting for group to get backlogged. Expire the queue
3119          */
3120         if (cfq_cfqq_wait_busy(cfqq) && !RB_EMPTY_ROOT(&cfqq->sort_list))
3121                 goto expire;
3122
3123         /*
3124          * The active queue has run out of time, expire it and select new.
3125          */
3126         if (cfq_slice_used(cfqq) && !cfq_cfqq_must_dispatch(cfqq)) {
3127                 /*
3128                  * If slice had not expired at the completion of last request
3129                  * we might not have turned on wait_busy flag. Don't expire
3130                  * the queue yet. Allow the group to get backlogged.
3131                  *
3132                  * The very fact that we have used the slice, that means we
3133                  * have been idling all along on this queue and it should be
3134                  * ok to wait for this request to complete.
3135                  */
3136                 if (cfqq->cfqg->nr_cfqq == 1 && RB_EMPTY_ROOT(&cfqq->sort_list)
3137                     && cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3138                         cfqq = NULL;
3139                         goto keep_queue;
3140                 } else
3141                         goto check_group_idle;
3142         }
3143
3144         /*
3145          * The active queue has requests and isn't expired, allow it to
3146          * dispatch.
3147          */
3148         if (!RB_EMPTY_ROOT(&cfqq->sort_list))
3149                 goto keep_queue;
3150
3151         /*
3152          * If another queue has a request waiting within our mean seek
3153          * distance, let it run.  The expire code will check for close
3154          * cooperators and put the close queue at the front of the service
3155          * tree.  If possible, merge the expiring queue with the new cfqq.
3156          */
3157         new_cfqq = cfq_close_cooperator(cfqd, cfqq);
3158         if (new_cfqq) {
3159                 if (!cfqq->new_cfqq)
3160                         cfq_setup_merge(cfqq, new_cfqq);
3161                 goto expire;
3162         }
3163
3164         /*
3165          * No requests pending. If the active queue still has requests in
3166          * flight or is idling for a new request, allow either of these
3167          * conditions to happen (or time out) before selecting a new queue.
3168          */
3169         if (timer_pending(&cfqd->idle_slice_timer)) {
3170                 cfqq = NULL;
3171                 goto keep_queue;
3172         }
3173
3174         /*
3175          * This is a deep seek queue, but the device is much faster than
3176          * the queue can deliver, don't idle
3177          **/
3178         if (CFQQ_SEEKY(cfqq) && cfq_cfqq_idle_window(cfqq) &&
3179             (cfq_cfqq_slice_new(cfqq) ||
3180             (cfqq->slice_end - jiffies > jiffies - cfqq->slice_start))) {
3181                 cfq_clear_cfqq_deep(cfqq);
3182                 cfq_clear_cfqq_idle_window(cfqq);
3183         }
3184
3185         if (cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3186                 cfqq = NULL;
3187                 goto keep_queue;
3188         }
3189
3190         /*
3191          * If group idle is enabled and there are requests dispatched from
3192          * this group, wait for requests to complete.
3193          */
3194 check_group_idle:
3195         if (cfqd->cfq_group_idle && cfqq->cfqg->nr_cfqq == 1 &&
3196             cfqq->cfqg->dispatched &&
3197             !cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true)) {
3198                 cfqq = NULL;
3199                 goto keep_queue;
3200         }
3201
3202 expire:
3203         cfq_slice_expired(cfqd, 0);
3204 new_queue:
3205         /*
3206          * Current queue expired. Check if we have to switch to a new
3207          * service tree
3208          */
3209         if (!new_cfqq)
3210                 cfq_choose_cfqg(cfqd);
3211
3212         cfqq = cfq_set_active_queue(cfqd, new_cfqq);
3213 keep_queue:
3214         return cfqq;
3215 }
3216
3217 static int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq)
3218 {
3219         int dispatched = 0;
3220
3221         while (cfqq->next_rq) {
3222                 cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq);
3223                 dispatched++;
3224         }
3225
3226         BUG_ON(!list_empty(&cfqq->fifo));
3227
3228         /* By default cfqq is not expired if it is empty. Do it explicitly */
3229         __cfq_slice_expired(cfqq->cfqd, cfqq, 0);
3230         return dispatched;
3231 }
3232
3233 /*
3234  * Drain our current requests. Used for barriers and when switching
3235  * io schedulers on-the-fly.
3236  */
3237 static int cfq_forced_dispatch(struct cfq_data *cfqd)
3238 {
3239         struct cfq_queue *cfqq;
3240         int dispatched = 0;
3241
3242         /* Expire the timeslice of the current active queue first */
3243         cfq_slice_expired(cfqd, 0);
3244         while ((cfqq = cfq_get_next_queue_forced(cfqd)) != NULL) {
3245                 __cfq_set_active_queue(cfqd, cfqq);
3246                 dispatched += __cfq_forced_dispatch_cfqq(cfqq);
3247         }
3248
3249         BUG_ON(cfqd->busy_queues);
3250
3251         cfq_log(cfqd, "forced_dispatch=%d", dispatched);
3252         return dispatched;
3253 }
3254
3255 static inline bool cfq_slice_used_soon(struct cfq_data *cfqd,
3256         struct cfq_queue *cfqq)
3257 {
3258         /* the queue hasn't finished any request, can't estimate */
3259         if (cfq_cfqq_slice_new(cfqq))
3260                 return true;
3261         if (time_after(jiffies + cfqd->cfq_slice_idle * cfqq->dispatched,
3262                 cfqq->slice_end))
3263                 return true;
3264
3265         return false;
3266 }
3267
3268 static bool cfq_may_dispatch(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3269 {
3270         unsigned int max_dispatch;
3271
3272         /*
3273          * Drain async requests before we start sync IO
3274          */
3275         if (cfq_should_idle(cfqd, cfqq) && cfqd->rq_in_flight[BLK_RW_ASYNC])
3276                 return false;
3277
3278         /*
3279          * If this is an async queue and we have sync IO in flight, let it wait
3280          */
3281         if (cfqd->rq_in_flight[BLK_RW_SYNC] && !cfq_cfqq_sync(cfqq))
3282                 return false;
3283
3284         max_dispatch = max_t(unsigned int, cfqd->cfq_quantum / 2, 1);
3285         if (cfq_class_idle(cfqq))
3286                 max_dispatch = 1;
3287
3288         /*
3289          * Does this cfqq already have too much IO in flight?
3290          */
3291         if (cfqq->dispatched >= max_dispatch) {
3292                 bool promote_sync = false;
3293                 /*
3294                  * idle queue must always only have a single IO in flight
3295                  */
3296                 if (cfq_class_idle(cfqq))
3297                         return false;
3298
3299                 /*
3300                  * If there is only one sync queue
3301                  * we can ignore async queue here and give the sync
3302                  * queue no dispatch limit. The reason is a sync queue can
3303                  * preempt async queue, limiting the sync queue doesn't make
3304                  * sense. This is useful for aiostress test.
3305                  */
3306                 if (cfq_cfqq_sync(cfqq) && cfqd->busy_sync_queues == 1)
3307                         promote_sync = true;
3308
3309                 /*
3310                  * We have other queues, don't allow more IO from this one
3311                  */
3312                 if (cfqd->busy_queues > 1 && cfq_slice_used_soon(cfqd, cfqq) &&
3313                                 !promote_sync)
3314                         return false;
3315
3316                 /*
3317                  * Sole queue user, no limit
3318                  */
3319                 if (cfqd->busy_queues == 1 || promote_sync)
3320                         max_dispatch = -1;
3321                 else
3322                         /*
3323                          * Normally we start throttling cfqq when cfq_quantum/2
3324                          * requests have been dispatched. But we can drive
3325                          * deeper queue depths at the beginning of slice
3326                          * subjected to upper limit of cfq_quantum.
3327                          * */
3328                         max_dispatch = cfqd->cfq_quantum;
3329         }
3330
3331         /*
3332          * Async queues must wait a bit before being allowed dispatch.
3333          * We also ramp up the dispatch depth gradually for async IO,
3334          * based on the last sync IO we serviced
3335          */
3336         if (!cfq_cfqq_sync(cfqq) && cfqd->cfq_latency) {
3337                 unsigned long last_sync = jiffies - cfqd->last_delayed_sync;
3338                 unsigned int depth;
3339
3340                 depth = last_sync / cfqd->cfq_slice[1];
3341                 if (!depth && !cfqq->dispatched)
3342                         depth = 1;
3343                 if (depth < max_dispatch)
3344                         max_dispatch = depth;
3345         }
3346
3347         /*
3348          * If we're below the current max, allow a dispatch
3349          */
3350         return cfqq->dispatched < max_dispatch;
3351 }
3352
3353 /*
3354  * Dispatch a request from cfqq, moving them to the request queue
3355  * dispatch list.
3356  */
3357 static bool cfq_dispatch_request(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3358 {
3359         struct request *rq;
3360
3361         BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list));
3362
3363         if (!cfq_may_dispatch(cfqd, cfqq))
3364                 return false;
3365
3366         /*
3367          * follow expired path, else get first next available
3368          */
3369         rq = cfq_check_fifo(cfqq);
3370         if (!rq)
3371                 rq = cfqq->next_rq;
3372
3373         /*
3374          * insert request into driver dispatch list
3375          */
3376         cfq_dispatch_insert(cfqd->queue, rq);
3377
3378         if (!cfqd->active_cic) {
3379                 struct cfq_io_cq *cic = RQ_CIC(rq);
3380
3381                 atomic_long_inc(&cic->icq.ioc->refcount);
3382                 cfqd->active_cic = cic;
3383         }
3384
3385         return true;
3386 }
3387
3388 /*
3389  * Find the cfqq that we need to service and move a request from that to the
3390  * dispatch list
3391  */
3392 static int cfq_dispatch_requests(struct request_queue *q, int force)
3393 {
3394         struct cfq_data *cfqd = q->elevator->elevator_data;
3395         struct cfq_queue *cfqq;
3396
3397         if (!cfqd->busy_queues)
3398                 return 0;
3399
3400         if (unlikely(force))
3401                 return cfq_forced_dispatch(cfqd);
3402
3403         cfqq = cfq_select_queue(cfqd);
3404         if (!cfqq)
3405                 return 0;
3406
3407         /*
3408          * Dispatch a request from this cfqq, if it is allowed
3409          */
3410         if (!cfq_dispatch_request(cfqd, cfqq))
3411                 return 0;
3412
3413         cfqq->slice_dispatch++;
3414         cfq_clear_cfqq_must_dispatch(cfqq);
3415
3416         /*
3417          * expire an async queue immediately if it has used up its slice. idle
3418          * queue always expire after 1 dispatch round.
3419          */
3420         if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) &&
3421             cfqq->slice_dispatch >= cfq_prio_to_maxrq(cfqd, cfqq)) ||
3422             cfq_class_idle(cfqq))) {
3423                 cfqq->slice_end = jiffies + 1;
3424                 cfq_slice_expired(cfqd, 0);
3425         }
3426
3427         cfq_log_cfqq(cfqd, cfqq, "dispatched a request");
3428         return 1;
3429 }
3430
3431 /*
3432  * task holds one reference to the queue, dropped when task exits. each rq
3433  * in-flight on this queue also holds a reference, dropped when rq is freed.
3434  *
3435  * Each cfq queue took a reference on the parent group. Drop it now.
3436  * queue lock must be held here.
3437  */
3438 static void cfq_put_queue(struct cfq_queue *cfqq)
3439 {
3440         struct cfq_data *cfqd = cfqq->cfqd;
3441         struct cfq_group *cfqg;
3442
3443         BUG_ON(cfqq->ref <= 0);
3444
3445         cfqq->ref--;
3446         if (cfqq->ref)
3447                 return;
3448
3449         cfq_log_cfqq(cfqd, cfqq, "put_queue");
3450         BUG_ON(rb_first(&cfqq->sort_list));
3451         BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
3452         cfqg = cfqq->cfqg;
3453
3454         if (unlikely(cfqd->active_queue == cfqq)) {
3455                 __cfq_slice_expired(cfqd, cfqq, 0);
3456                 cfq_schedule_dispatch(cfqd);
3457         }
3458
3459         BUG_ON(cfq_cfqq_on_rr(cfqq));
3460         kmem_cache_free(cfq_pool, cfqq);
3461         cfqg_put(cfqg);
3462 }
3463
3464 static void cfq_put_cooperator(struct cfq_queue *cfqq)
3465 {
3466         struct cfq_queue *__cfqq, *next;
3467
3468         /*
3469          * If this queue was scheduled to merge with another queue, be
3470          * sure to drop the reference taken on that queue (and others in
3471          * the merge chain).  See cfq_setup_merge and cfq_merge_cfqqs.
3472          */
3473         __cfqq = cfqq->new_cfqq;
3474         while (__cfqq) {
3475                 if (__cfqq == cfqq) {
3476                         WARN(1, "cfqq->new_cfqq loop detected\n");
3477                         break;
3478                 }
3479                 next = __cfqq->new_cfqq;
3480                 cfq_put_queue(__cfqq);
3481                 __cfqq = next;
3482         }
3483 }
3484
3485 static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3486 {
3487         if (unlikely(cfqq == cfqd->active_queue)) {
3488                 __cfq_slice_expired(cfqd, cfqq, 0);
3489                 cfq_schedule_dispatch(cfqd);
3490         }
3491
3492         cfq_put_cooperator(cfqq);
3493
3494         cfq_put_queue(cfqq);
3495 }
3496
3497 static void cfq_init_icq(struct io_cq *icq)
3498 {
3499         struct cfq_io_cq *cic = icq_to_cic(icq);
3500
3501         cic->ttime.last_end_request = jiffies;
3502 }
3503
3504 static void cfq_exit_icq(struct io_cq *icq)
3505 {
3506         struct cfq_io_cq *cic = icq_to_cic(icq);
3507         struct cfq_data *cfqd = cic_to_cfqd(cic);
3508
3509         if (cic->cfqq[BLK_RW_ASYNC]) {
3510                 cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_ASYNC]);
3511                 cic->cfqq[BLK_RW_ASYNC] = NULL;
3512         }
3513
3514         if (cic->cfqq[BLK_RW_SYNC]) {
3515                 cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_SYNC]);
3516                 cic->cfqq[BLK_RW_SYNC] = NULL;
3517         }
3518 }
3519
3520 static void cfq_init_prio_data(struct cfq_queue *cfqq, struct cfq_io_cq *cic)
3521 {
3522         struct task_struct *tsk = current;
3523         int ioprio_class;
3524
3525         if (!cfq_cfqq_prio_changed(cfqq))
3526                 return;
3527
3528         ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3529         switch (ioprio_class) {
3530         default:
3531                 printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
3532         case IOPRIO_CLASS_NONE:
3533                 /*
3534                  * no prio set, inherit CPU scheduling settings
3535                  */
3536                 cfqq->ioprio = task_nice_ioprio(tsk);
3537                 cfqq->ioprio_class = task_nice_ioclass(tsk);
3538                 break;
3539         case IOPRIO_CLASS_RT:
3540                 cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3541                 cfqq->ioprio_class = IOPRIO_CLASS_RT;
3542                 break;
3543         case IOPRIO_CLASS_BE:
3544                 cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3545                 cfqq->ioprio_class = IOPRIO_CLASS_BE;
3546                 break;
3547         case IOPRIO_CLASS_IDLE:
3548                 cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
3549                 cfqq->ioprio = 7;
3550                 cfq_clear_cfqq_idle_window(cfqq);
3551                 break;
3552         }
3553
3554         /*
3555          * keep track of original prio settings in case we have to temporarily
3556          * elevate the priority of this queue
3557          */
3558         cfqq->org_ioprio = cfqq->ioprio;
3559         cfq_clear_cfqq_prio_changed(cfqq);
3560 }
3561
3562 static void check_ioprio_changed(struct cfq_io_cq *cic, struct bio *bio)
3563 {
3564         int ioprio = cic->icq.ioc->ioprio;
3565         struct cfq_data *cfqd = cic_to_cfqd(cic);
3566         struct cfq_queue *cfqq;
3567
3568         /*
3569          * Check whether ioprio has changed.  The condition may trigger
3570          * spuriously on a newly created cic but there's no harm.
3571          */
3572         if (unlikely(!cfqd) || likely(cic->ioprio == ioprio))
3573                 return;
3574
3575         cfqq = cic->cfqq[BLK_RW_ASYNC];
3576         if (cfqq) {
3577                 struct cfq_queue *new_cfqq;
3578                 new_cfqq = cfq_get_queue(cfqd, BLK_RW_ASYNC, cic, bio,
3579                                          GFP_ATOMIC);
3580                 if (new_cfqq) {
3581                         cic->cfqq[BLK_RW_ASYNC] = new_cfqq;
3582                         cfq_put_queue(cfqq);
3583                 }
3584         }
3585
3586         cfqq = cic->cfqq[BLK_RW_SYNC];
3587         if (cfqq)
3588                 cfq_mark_cfqq_prio_changed(cfqq);
3589
3590         cic->ioprio = ioprio;
3591 }
3592
3593 static void cfq_init_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3594                           pid_t pid, bool is_sync)
3595 {
3596         RB_CLEAR_NODE(&cfqq->rb_node);
3597         RB_CLEAR_NODE(&cfqq->p_node);
3598         INIT_LIST_HEAD(&cfqq->fifo);
3599
3600         cfqq->ref = 0;
3601         cfqq->cfqd = cfqd;
3602
3603         cfq_mark_cfqq_prio_changed(cfqq);
3604
3605         if (is_sync) {
3606                 if (!cfq_class_idle(cfqq))
3607                         cfq_mark_cfqq_idle_window(cfqq);
3608                 cfq_mark_cfqq_sync(cfqq);
3609         }
3610         cfqq->pid = pid;
3611 }
3612
3613 #ifdef CONFIG_CFQ_GROUP_IOSCHED
3614 static void check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio)
3615 {
3616         struct cfq_data *cfqd = cic_to_cfqd(cic);
3617         struct cfq_queue *sync_cfqq;
3618         uint64_t serial_nr;
3619
3620         rcu_read_lock();
3621         serial_nr = bio_blkcg(bio)->css.serial_nr;
3622         rcu_read_unlock();
3623
3624         /*
3625          * Check whether blkcg has changed.  The condition may trigger
3626          * spuriously on a newly created cic but there's no harm.
3627          */
3628         if (unlikely(!cfqd) || likely(cic->blkcg_serial_nr == serial_nr))
3629                 return;
3630
3631         sync_cfqq = cic_to_cfqq(cic, 1);
3632         if (sync_cfqq) {
3633                 /*
3634                  * Drop reference to sync queue. A new sync queue will be
3635                  * assigned in new group upon arrival of a fresh request.
3636                  */
3637                 cfq_log_cfqq(cfqd, sync_cfqq, "changed cgroup");
3638                 cic_set_cfqq(cic, NULL, 1);
3639                 cfq_put_queue(sync_cfqq);
3640         }
3641
3642         cic->blkcg_serial_nr = serial_nr;
3643 }
3644 #else
3645 static inline void check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio) { }
3646 #endif  /* CONFIG_CFQ_GROUP_IOSCHED */
3647
3648 static struct cfq_queue *
3649 cfq_find_alloc_queue(struct cfq_data *cfqd, bool is_sync, struct cfq_io_cq *cic,
3650                      struct bio *bio, gfp_t gfp_mask)
3651 {
3652         struct blkcg *blkcg;
3653         struct cfq_queue *cfqq, *new_cfqq = NULL;
3654         struct cfq_group *cfqg;
3655
3656 retry:
3657         rcu_read_lock();
3658
3659         blkcg = bio_blkcg(bio);
3660         cfqg = cfq_lookup_create_cfqg(cfqd, blkcg);
3661         if (!cfqg) {
3662                 cfqq = &cfqd->oom_cfqq;
3663                 goto out;
3664         }
3665
3666         cfqq = cic_to_cfqq(cic, is_sync);
3667
3668         /*
3669          * Always try a new alloc if we fell back to the OOM cfqq
3670          * originally, since it should just be a temporary situation.
3671          */
3672         if (!cfqq || cfqq == &cfqd->oom_cfqq) {
3673                 cfqq = NULL;
3674                 if (new_cfqq) {
3675                         cfqq = new_cfqq;
3676                         new_cfqq = NULL;
3677                 } else if (gfp_mask & __GFP_WAIT) {
3678                         rcu_read_unlock();
3679                         spin_unlock_irq(cfqd->queue->queue_lock);
3680                         new_cfqq = kmem_cache_alloc_node(cfq_pool,
3681                                         gfp_mask | __GFP_ZERO,
3682                                         cfqd->queue->node);
3683                         spin_lock_irq(cfqd->queue->queue_lock);
3684                         if (new_cfqq)
3685                                 goto retry;
3686                         else
3687                                 return &cfqd->oom_cfqq;
3688                 } else {
3689                         cfqq = kmem_cache_alloc_node(cfq_pool,
3690                                         gfp_mask | __GFP_ZERO,
3691                                         cfqd->queue->node);
3692                 }
3693
3694                 if (cfqq) {
3695                         cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync);
3696                         cfq_init_prio_data(cfqq, cic);
3697                         cfq_link_cfqq_cfqg(cfqq, cfqg);
3698                         cfq_log_cfqq(cfqd, cfqq, "alloced");
3699                 } else
3700                         cfqq = &cfqd->oom_cfqq;
3701         }
3702 out:
3703         if (new_cfqq)
3704                 kmem_cache_free(cfq_pool, new_cfqq);
3705
3706         rcu_read_unlock();
3707         return cfqq;
3708 }
3709
3710 static struct cfq_queue **
3711 cfq_async_queue_prio(struct cfq_data *cfqd, int ioprio_class, int ioprio)
3712 {
3713         switch (ioprio_class) {
3714         case IOPRIO_CLASS_RT:
3715                 return &cfqd->async_cfqq[0][ioprio];
3716         case IOPRIO_CLASS_NONE:
3717                 ioprio = IOPRIO_NORM;
3718                 /* fall through */
3719         case IOPRIO_CLASS_BE:
3720                 return &cfqd->async_cfqq[1][ioprio];
3721         case IOPRIO_CLASS_IDLE:
3722                 return &cfqd->async_idle_cfqq;
3723         default:
3724                 BUG();
3725         }
3726 }
3727
3728 static struct cfq_queue *
3729 cfq_get_queue(struct cfq_data *cfqd, bool is_sync, struct cfq_io_cq *cic,
3730               struct bio *bio, gfp_t gfp_mask)
3731 {
3732         int ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3733         int ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3734         struct cfq_queue **async_cfqq = NULL;
3735         struct cfq_queue *cfqq = NULL;
3736
3737         if (!is_sync) {
3738                 if (!ioprio_valid(cic->ioprio)) {
3739                         struct task_struct *tsk = current;
3740                         ioprio = task_nice_ioprio(tsk);
3741                         ioprio_class = task_nice_ioclass(tsk);
3742                 }
3743                 async_cfqq = cfq_async_queue_prio(cfqd, ioprio_class, ioprio);
3744                 cfqq = *async_cfqq;
3745         }
3746
3747         if (!cfqq)
3748                 cfqq = cfq_find_alloc_queue(cfqd, is_sync, cic, bio, gfp_mask);
3749
3750         /*
3751          * pin the queue now that it's allocated, scheduler exit will prune it
3752          */
3753         if (!is_sync && !(*async_cfqq)) {
3754                 cfqq->ref++;
3755                 *async_cfqq = cfqq;
3756         }
3757
3758         cfqq->ref++;
3759         return cfqq;
3760 }
3761
3762 static void
3763 __cfq_update_io_thinktime(struct cfq_ttime *ttime, unsigned long slice_idle)
3764 {
3765         unsigned long elapsed = jiffies - ttime->last_end_request;
3766         elapsed = min(elapsed, 2UL * slice_idle);
3767
3768         ttime->ttime_samples = (7*ttime->ttime_samples + 256) / 8;
3769         ttime->ttime_total = (7*ttime->ttime_total + 256*elapsed) / 8;
3770         ttime->ttime_mean = (ttime->ttime_total + 128) / ttime->ttime_samples;
3771 }
3772
3773 static void
3774 cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3775                         struct cfq_io_cq *cic)
3776 {
3777         if (cfq_cfqq_sync(cfqq)) {
3778                 __cfq_update_io_thinktime(&cic->ttime, cfqd->cfq_slice_idle);
3779                 __cfq_update_io_thinktime(&cfqq->service_tree->ttime,
3780                         cfqd->cfq_slice_idle);
3781         }
3782 #ifdef CONFIG_CFQ_GROUP_IOSCHED
3783         __cfq_update_io_thinktime(&cfqq->cfqg->ttime, cfqd->cfq_group_idle);
3784 #endif
3785 }
3786
3787 static void
3788 cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3789                        struct request *rq)
3790 {
3791         sector_t sdist = 0;
3792         sector_t n_sec = blk_rq_sectors(rq);
3793         if (cfqq->last_request_pos) {
3794                 if (cfqq->last_request_pos < blk_rq_pos(rq))
3795                         sdist = blk_rq_pos(rq) - cfqq->last_request_pos;
3796                 else
3797                         sdist = cfqq->last_request_pos - blk_rq_pos(rq);
3798         }
3799
3800         cfqq->seek_history <<= 1;
3801         if (blk_queue_nonrot(cfqd->queue))
3802                 cfqq->seek_history |= (n_sec < CFQQ_SECT_THR_NONROT);
3803         else
3804                 cfqq->seek_history |= (sdist > CFQQ_SEEK_THR);
3805 }
3806
3807 /*
3808  * Disable idle window if the process thinks too long or seeks so much that
3809  * it doesn't matter
3810  */
3811 static void
3812 cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3813                        struct cfq_io_cq *cic)
3814 {
3815         int old_idle, enable_idle;
3816
3817         /*
3818          * Don't idle for async or idle io prio class
3819          */
3820         if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq))
3821                 return;
3822
3823         enable_idle = old_idle = cfq_cfqq_idle_window(cfqq);
3824
3825         if (cfqq->queued[0] + cfqq->queued[1] >= 4)
3826                 cfq_mark_cfqq_deep(cfqq);
3827
3828         if (cfqq->next_rq && (cfqq->next_rq->cmd_flags & REQ_NOIDLE))
3829                 enable_idle = 0;
3830         else if (!atomic_read(&cic->icq.ioc->active_ref) ||
3831                  !cfqd->cfq_slice_idle ||
3832                  (!cfq_cfqq_deep(cfqq) && CFQQ_SEEKY(cfqq)))
3833                 enable_idle = 0;
3834         else if (sample_valid(cic->ttime.ttime_samples)) {
3835                 if (cic->ttime.ttime_mean > cfqd->cfq_slice_idle)
3836                         enable_idle = 0;
3837                 else
3838                         enable_idle = 1;
3839         }
3840
3841         if (old_idle != enable_idle) {
3842                 cfq_log_cfqq(cfqd, cfqq, "idle=%d", enable_idle);
3843                 if (enable_idle)
3844                         cfq_mark_cfqq_idle_window(cfqq);
3845                 else
3846                         cfq_clear_cfqq_idle_window(cfqq);
3847         }
3848 }
3849
3850 /*
3851  * Check if new_cfqq should preempt the currently active queue. Return 0 for
3852  * no or if we aren't sure, a 1 will cause a preempt.
3853  */
3854 static bool
3855 cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
3856                    struct request *rq)
3857 {
3858         struct cfq_queue *cfqq;
3859
3860         cfqq = cfqd->active_queue;
3861         if (!cfqq)
3862                 return false;
3863
3864         if (cfq_class_idle(new_cfqq))
3865                 return false;
3866
3867         if (cfq_class_idle(cfqq))
3868                 return true;
3869
3870         /*
3871          * Don't allow a non-RT request to preempt an ongoing RT cfqq timeslice.
3872          */
3873         if (cfq_class_rt(cfqq) && !cfq_class_rt(new_cfqq))
3874                 return false;
3875
3876         /*
3877          * if the new request is sync, but the currently running queue is
3878          * not, let the sync request have priority.
3879          */
3880         if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq))
3881                 return true;
3882
3883         if (new_cfqq->cfqg != cfqq->cfqg)
3884                 return false;
3885
3886         if (cfq_slice_used(cfqq))
3887                 return true;
3888
3889         /* Allow preemption only if we are idling on sync-noidle tree */
3890         if (cfqd->serving_wl_type == SYNC_NOIDLE_WORKLOAD &&
3891             cfqq_type(new_cfqq) == SYNC_NOIDLE_WORKLOAD &&
3892             new_cfqq->service_tree->count == 2 &&
3893             RB_EMPTY_ROOT(&cfqq->sort_list))
3894                 return true;
3895
3896         /*
3897          * So both queues are sync. Let the new request get disk time if
3898          * it's a metadata request and the current queue is doing regular IO.
3899          */
3900         if ((rq->cmd_flags & REQ_PRIO) && !cfqq->prio_pending)
3901                 return true;
3902
3903         /*
3904          * Allow an RT request to pre-empt an ongoing non-RT cfqq timeslice.
3905          */
3906         if (cfq_class_rt(new_cfqq) && !cfq_class_rt(cfqq))
3907                 return true;
3908
3909         /* An idle queue should not be idle now for some reason */
3910         if (RB_EMPTY_ROOT(&cfqq->sort_list) && !cfq_should_idle(cfqd, cfqq))
3911                 return true;
3912
3913         if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq))
3914                 return false;
3915
3916         /*
3917          * if this request is as-good as one we would expect from the
3918          * current cfqq, let it preempt
3919          */
3920         if (cfq_rq_close(cfqd, cfqq, rq))
3921                 return true;
3922
3923         return false;
3924 }
3925
3926 /*
3927  * cfqq preempts the active queue. if we allowed preempt with no slice left,
3928  * let it have half of its nominal slice.
3929  */
3930 static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3931 {
3932         enum wl_type_t old_type = cfqq_type(cfqd->active_queue);
3933
3934         cfq_log_cfqq(cfqd, cfqq, "preempt");
3935         cfq_slice_expired(cfqd, 1);
3936
3937         /*
3938          * workload type is changed, don't save slice, otherwise preempt
3939          * doesn't happen
3940          */
3941         if (old_type != cfqq_type(cfqq))
3942                 cfqq->cfqg->saved_wl_slice = 0;
3943
3944         /*
3945          * Put the new queue at the front of the of the current list,
3946          * so we know that it will be selected next.
3947          */
3948         BUG_ON(!cfq_cfqq_on_rr(cfqq));
3949
3950         cfq_service_tree_add(cfqd, cfqq, 1);
3951
3952         cfqq->slice_end = 0;
3953         cfq_mark_cfqq_slice_new(cfqq);
3954 }
3955
3956 /*
3957  * Called when a new fs request (rq) is added (to cfqq). Check if there's
3958  * something we should do about it
3959  */
3960 static void
3961 cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3962                 struct request *rq)
3963 {
3964         struct cfq_io_cq *cic = RQ_CIC(rq);
3965
3966         cfqd->rq_queued++;
3967         if (rq->cmd_flags & REQ_PRIO)
3968                 cfqq->prio_pending++;
3969
3970         cfq_update_io_thinktime(cfqd, cfqq, cic);
3971         cfq_update_io_seektime(cfqd, cfqq, rq);
3972         cfq_update_idle_window(cfqd, cfqq, cic);
3973
3974         cfqq->last_request_pos = blk_rq_pos(rq) + blk_rq_sectors(rq);
3975
3976         if (cfqq == cfqd->active_queue) {
3977                 /*
3978                  * Remember that we saw a request from this process, but
3979                  * don't start queuing just yet. Otherwise we risk seeing lots
3980                  * of tiny requests, because we disrupt the normal plugging
3981                  * and merging. If the request is already larger than a single
3982                  * page, let it rip immediately. For that case we assume that
3983                  * merging is already done. Ditto for a busy system that
3984                  * has other work pending, don't risk delaying until the
3985                  * idle timer unplug to continue working.
3986                  */
3987                 if (cfq_cfqq_wait_request(cfqq)) {
3988                         if (blk_rq_bytes(rq) > PAGE_CACHE_SIZE ||
3989                             cfqd->busy_queues > 1) {
3990                                 cfq_del_timer(cfqd, cfqq);
3991                                 cfq_clear_cfqq_wait_request(cfqq);
3992                                 __blk_run_queue(cfqd->queue);
3993                         } else {
3994                                 cfqg_stats_update_idle_time(cfqq->cfqg);
3995                                 cfq_mark_cfqq_must_dispatch(cfqq);
3996                         }
3997                 }
3998         } else if (cfq_should_preempt(cfqd, cfqq, rq)) {
3999                 /*
4000                  * not the active queue - expire current slice if it is
4001                  * idle and has expired it's mean thinktime or this new queue
4002                  * has some old slice time left and is of higher priority or
4003                  * this new queue is RT and the current one is BE
4004                  */
4005                 cfq_preempt_queue(cfqd, cfqq);
4006                 __blk_run_queue(cfqd->queue);
4007         }
4008 }
4009
4010 static void cfq_insert_request(struct request_queue *q, struct request *rq)
4011 {
4012         struct cfq_data *cfqd = q->elevator->elevator_data;
4013         struct cfq_queue *cfqq = RQ_CFQQ(rq);
4014
4015         cfq_log_cfqq(cfqd, cfqq, "insert_request");
4016         cfq_init_prio_data(cfqq, RQ_CIC(rq));
4017
4018         rq->fifo_time = jiffies + cfqd->cfq_fifo_expire[rq_is_sync(rq)];
4019         list_add_tail(&rq->queuelist, &cfqq->fifo);
4020         cfq_add_rq_rb(rq);
4021         cfqg_stats_update_io_add(RQ_CFQG(rq), cfqd->serving_group,
4022                                  rq->cmd_flags);
4023         cfq_rq_enqueued(cfqd, cfqq, rq);
4024 }
4025
4026 /*
4027  * Update hw_tag based on peak queue depth over 50 samples under
4028  * sufficient load.
4029  */
4030 static void cfq_update_hw_tag(struct cfq_data *cfqd)
4031 {
4032         struct cfq_queue *cfqq = cfqd->active_queue;
4033
4034         if (cfqd->rq_in_driver > cfqd->hw_tag_est_depth)
4035                 cfqd->hw_tag_est_depth = cfqd->rq_in_driver;
4036
4037         if (cfqd->hw_tag == 1)
4038                 return;
4039
4040         if (cfqd->rq_queued <= CFQ_HW_QUEUE_MIN &&
4041             cfqd->rq_in_driver <= CFQ_HW_QUEUE_MIN)
4042                 return;
4043
4044         /*
4045          * If active queue hasn't enough requests and can idle, cfq might not
4046          * dispatch sufficient requests to hardware. Don't zero hw_tag in this
4047          * case
4048          */
4049         if (cfqq && cfq_cfqq_idle_window(cfqq) &&
4050             cfqq->dispatched + cfqq->queued[0] + cfqq->queued[1] <
4051             CFQ_HW_QUEUE_MIN && cfqd->rq_in_driver < CFQ_HW_QUEUE_MIN)
4052                 return;
4053
4054         if (cfqd->hw_tag_samples++ < 50)
4055                 return;
4056
4057         if (cfqd->hw_tag_est_depth >= CFQ_HW_QUEUE_MIN)
4058                 cfqd->hw_tag = 1;
4059         else
4060                 cfqd->hw_tag = 0;
4061 }
4062
4063 static bool cfq_should_wait_busy(struct cfq_data *cfqd, struct cfq_queue *cfqq)
4064 {
4065         struct cfq_io_cq *cic = cfqd->active_cic;
4066
4067         /* If the queue already has requests, don't wait */
4068         if (!RB_EMPTY_ROOT(&cfqq->sort_list))
4069                 return false;
4070
4071         /* If there are other queues in the group, don't wait */
4072         if (cfqq->cfqg->nr_cfqq > 1)
4073                 return false;
4074
4075         /* the only queue in the group, but think time is big */
4076         if (cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true))
4077                 return false;
4078
4079         if (cfq_slice_used(cfqq))
4080                 return true;
4081
4082         /* if slice left is less than think time, wait busy */
4083         if (cic && sample_valid(cic->ttime.ttime_samples)
4084             && (cfqq->slice_end - jiffies < cic->ttime.ttime_mean))
4085                 return true;
4086
4087         /*
4088          * If think times is less than a jiffy than ttime_mean=0 and above
4089          * will not be true. It might happen that slice has not expired yet
4090          * but will expire soon (4-5 ns) during select_queue(). To cover the
4091          * case where think time is less than a jiffy, mark the queue wait
4092          * busy if only 1 jiffy is left in the slice.
4093          */
4094         if (cfqq->slice_end - jiffies == 1)
4095                 return true;
4096
4097         return false;
4098 }
4099
4100 static void cfq_completed_request(struct request_queue *q, struct request *rq)
4101 {
4102         struct cfq_queue *cfqq = RQ_CFQQ(rq);
4103         struct cfq_data *cfqd = cfqq->cfqd;
4104         const int sync = rq_is_sync(rq);
4105         unsigned long now;
4106
4107         now = jiffies;
4108         cfq_log_cfqq(cfqd, cfqq, "complete rqnoidle %d",
4109                      !!(rq->cmd_flags & REQ_NOIDLE));
4110
4111         cfq_update_hw_tag(cfqd);
4112
4113         WARN_ON(!cfqd->rq_in_driver);
4114         WARN_ON(!cfqq->dispatched);
4115         cfqd->rq_in_driver--;
4116         cfqq->dispatched--;
4117         (RQ_CFQG(rq))->dispatched--;
4118         cfqg_stats_update_completion(cfqq->cfqg, rq_start_time_ns(rq),
4119                                      rq_io_start_time_ns(rq), rq->cmd_flags);
4120
4121         cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]--;
4122
4123         if (sync) {
4124                 struct cfq_rb_root *st;
4125
4126                 RQ_CIC(rq)->ttime.last_end_request = now;
4127
4128                 if (cfq_cfqq_on_rr(cfqq))
4129                         st = cfqq->service_tree;
4130                 else
4131                         st = st_for(cfqq->cfqg, cfqq_class(cfqq),
4132                                         cfqq_type(cfqq));
4133
4134                 st->ttime.last_end_request = now;
4135                 if (!time_after(rq->start_time + cfqd->cfq_fifo_expire[1], now))
4136                         cfqd->last_delayed_sync = now;
4137         }
4138
4139 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4140         cfqq->cfqg->ttime.last_end_request = now;
4141 #endif
4142
4143         /*
4144          * If this is the active queue, check if it needs to be expired,
4145          * or if we want to idle in case it has no pending requests.
4146          */
4147         if (cfqd->active_queue == cfqq) {
4148                 const bool cfqq_empty = RB_EMPTY_ROOT(&cfqq->sort_list);
4149
4150                 if (cfq_cfqq_slice_new(cfqq)) {
4151                         cfq_set_prio_slice(cfqd, cfqq);
4152                         cfq_clear_cfqq_slice_new(cfqq);
4153                 }
4154
4155                 /*
4156                  * Should we wait for next request to come in before we expire
4157                  * the queue.
4158                  */
4159                 if (cfq_should_wait_busy(cfqd, cfqq)) {
4160                         unsigned long extend_sl = cfqd->cfq_slice_idle;
4161                         if (!cfqd->cfq_slice_idle)
4162                                 extend_sl = cfqd->cfq_group_idle;
4163                         cfqq->slice_end = jiffies + extend_sl;
4164                         cfq_mark_cfqq_wait_busy(cfqq);
4165                         cfq_log_cfqq(cfqd, cfqq, "will busy wait");
4166                 }
4167
4168                 /*
4169                  * Idling is not enabled on:
4170                  * - expired queues
4171                  * - idle-priority queues
4172                  * - async queues
4173                  * - queues with still some requests queued
4174                  * - when there is a close cooperator
4175                  */
4176                 if (cfq_slice_used(cfqq) || cfq_class_idle(cfqq))
4177                         cfq_slice_expired(cfqd, 1);
4178                 else if (sync && cfqq_empty &&
4179                          !cfq_close_cooperator(cfqd, cfqq)) {
4180                         cfq_arm_slice_timer(cfqd);
4181                 }
4182         }
4183
4184         if (!cfqd->rq_in_driver)
4185                 cfq_schedule_dispatch(cfqd);
4186 }
4187
4188 static inline int __cfq_may_queue(struct cfq_queue *cfqq)
4189 {
4190         if (cfq_cfqq_wait_request(cfqq) && !cfq_cfqq_must_alloc_slice(cfqq)) {
4191                 cfq_mark_cfqq_must_alloc_slice(cfqq);
4192                 return ELV_MQUEUE_MUST;
4193         }
4194
4195         return ELV_MQUEUE_MAY;
4196 }
4197
4198 static int cfq_may_queue(struct request_queue *q, int rw)
4199 {
4200         struct cfq_data *cfqd = q->elevator->elevator_data;
4201         struct task_struct *tsk = current;
4202         struct cfq_io_cq *cic;
4203         struct cfq_queue *cfqq;
4204
4205         /*
4206          * don't force setup of a queue from here, as a call to may_queue
4207          * does not necessarily imply that a request actually will be queued.
4208          * so just lookup a possibly existing queue, or return 'may queue'
4209          * if that fails
4210          */
4211         cic = cfq_cic_lookup(cfqd, tsk->io_context);
4212         if (!cic)
4213                 return ELV_MQUEUE_MAY;
4214
4215         cfqq = cic_to_cfqq(cic, rw_is_sync(rw));
4216         if (cfqq) {
4217                 cfq_init_prio_data(cfqq, cic);
4218
4219                 return __cfq_may_queue(cfqq);
4220         }
4221
4222         return ELV_MQUEUE_MAY;
4223 }
4224
4225 /*
4226  * queue lock held here
4227  */
4228 static void cfq_put_request(struct request *rq)
4229 {
4230         struct cfq_queue *cfqq = RQ_CFQQ(rq);
4231
4232         if (cfqq) {
4233                 const int rw = rq_data_dir(rq);
4234
4235                 BUG_ON(!cfqq->allocated[rw]);
4236                 cfqq->allocated[rw]--;
4237
4238                 /* Put down rq reference on cfqg */
4239                 cfqg_put(RQ_CFQG(rq));
4240                 rq->elv.priv[0] = NULL;
4241                 rq->elv.priv[1] = NULL;
4242
4243                 cfq_put_queue(cfqq);
4244         }
4245 }
4246
4247 static struct cfq_queue *
4248 cfq_merge_cfqqs(struct cfq_data *cfqd, struct cfq_io_cq *cic,
4249                 struct cfq_queue *cfqq)
4250 {
4251         cfq_log_cfqq(cfqd, cfqq, "merging with queue %p", cfqq->new_cfqq);
4252         cic_set_cfqq(cic, cfqq->new_cfqq, 1);
4253         cfq_mark_cfqq_coop(cfqq->new_cfqq);
4254         cfq_put_queue(cfqq);
4255         return cic_to_cfqq(cic, 1);
4256 }
4257
4258 /*
4259  * Returns NULL if a new cfqq should be allocated, or the old cfqq if this
4260  * was the last process referring to said cfqq.
4261  */
4262 static struct cfq_queue *
4263 split_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq)
4264 {
4265         if (cfqq_process_refs(cfqq) == 1) {
4266                 cfqq->pid = current->pid;
4267                 cfq_clear_cfqq_coop(cfqq);
4268                 cfq_clear_cfqq_split_coop(cfqq);
4269                 return cfqq;
4270         }
4271
4272         cic_set_cfqq(cic, NULL, 1);
4273
4274         cfq_put_cooperator(cfqq);
4275
4276         cfq_put_queue(cfqq);
4277         return NULL;
4278 }
4279 /*
4280  * Allocate cfq data structures associated with this request.
4281  */
4282 static int
4283 cfq_set_request(struct request_queue *q, struct request *rq, struct bio *bio,
4284                 gfp_t gfp_mask)
4285 {
4286         struct cfq_data *cfqd = q->elevator->elevator_data;
4287         struct cfq_io_cq *cic = icq_to_cic(rq->elv.icq);
4288         const int rw = rq_data_dir(rq);
4289         const bool is_sync = rq_is_sync(rq);
4290         struct cfq_queue *cfqq;
4291
4292         might_sleep_if(gfp_mask & __GFP_WAIT);
4293
4294         spin_lock_irq(q->queue_lock);
4295
4296         check_ioprio_changed(cic, bio);
4297         check_blkcg_changed(cic, bio);
4298 new_queue:
4299         cfqq = cic_to_cfqq(cic, is_sync);
4300         if (!cfqq || cfqq == &cfqd->oom_cfqq) {
4301                 cfqq = cfq_get_queue(cfqd, is_sync, cic, bio, gfp_mask);
4302                 cic_set_cfqq(cic, cfqq, is_sync);
4303         } else {
4304                 /*
4305                  * If the queue was seeky for too long, break it apart.
4306                  */
4307                 if (cfq_cfqq_coop(cfqq) && cfq_cfqq_split_coop(cfqq)) {
4308                         cfq_log_cfqq(cfqd, cfqq, "breaking apart cfqq");
4309                         cfqq = split_cfqq(cic, cfqq);
4310                         if (!cfqq)
4311                                 goto new_queue;
4312                 }
4313
4314                 /*
4315                  * Check to see if this queue is scheduled to merge with
4316                  * another, closely cooperating queue.  The merging of
4317                  * queues happens here as it must be done in process context.
4318                  * The reference on new_cfqq was taken in merge_cfqqs.
4319                  */
4320                 if (cfqq->new_cfqq)
4321                         cfqq = cfq_merge_cfqqs(cfqd, cic, cfqq);
4322         }
4323
4324         cfqq->allocated[rw]++;
4325
4326         cfqq->ref++;
4327         cfqg_get(cfqq->cfqg);
4328         rq->elv.priv[0] = cfqq;
4329         rq->elv.priv[1] = cfqq->cfqg;
4330         spin_unlock_irq(q->queue_lock);
4331         return 0;
4332 }
4333
4334 static void cfq_kick_queue(struct work_struct *work)
4335 {
4336         struct cfq_data *cfqd =
4337                 container_of(work, struct cfq_data, unplug_work);
4338         struct request_queue *q = cfqd->queue;
4339
4340         spin_lock_irq(q->queue_lock);
4341         __blk_run_queue(cfqd->queue);
4342         spin_unlock_irq(q->queue_lock);
4343 }
4344
4345 /*
4346  * Timer running if the active_queue is currently idling inside its time slice
4347  */
4348 static void cfq_idle_slice_timer(unsigned long data)
4349 {
4350         struct cfq_data *cfqd = (struct cfq_data *) data;
4351         struct cfq_queue *cfqq;
4352         unsigned long flags;
4353         int timed_out = 1;
4354
4355         cfq_log(cfqd, "idle timer fired");
4356
4357         spin_lock_irqsave(cfqd->queue->queue_lock, flags);
4358
4359         cfqq = cfqd->active_queue;
4360         if (cfqq) {
4361                 timed_out = 0;
4362
4363                 /*
4364                  * We saw a request before the queue expired, let it through
4365                  */
4366                 if (cfq_cfqq_must_dispatch(cfqq))
4367                         goto out_kick;
4368
4369                 /*
4370                  * expired
4371                  */
4372                 if (cfq_slice_used(cfqq))
4373                         goto expire;
4374
4375                 /*
4376                  * only expire and reinvoke request handler, if there are
4377                  * other queues with pending requests
4378                  */
4379                 if (!cfqd->busy_queues)
4380                         goto out_cont;
4381
4382                 /*
4383                  * not expired and it has a request pending, let it dispatch
4384                  */
4385                 if (!RB_EMPTY_ROOT(&cfqq->sort_list))
4386                         goto out_kick;
4387
4388                 /*
4389                  * Queue depth flag is reset only when the idle didn't succeed
4390                  */
4391                 cfq_clear_cfqq_deep(cfqq);
4392         }
4393 expire:
4394         cfq_slice_expired(cfqd, timed_out);
4395 out_kick:
4396         cfq_schedule_dispatch(cfqd);
4397 out_cont:
4398         spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
4399 }
4400
4401 static void cfq_shutdown_timer_wq(struct cfq_data *cfqd)
4402 {
4403         del_timer_sync(&cfqd->idle_slice_timer);
4404         cancel_work_sync(&cfqd->unplug_work);
4405 }
4406
4407 static void cfq_put_async_queues(struct cfq_data *cfqd)
4408 {
4409         int i;
4410
4411         for (i = 0; i < IOPRIO_BE_NR; i++) {
4412                 if (cfqd->async_cfqq[0][i])
4413                         cfq_put_queue(cfqd->async_cfqq[0][i]);
4414                 if (cfqd->async_cfqq[1][i])
4415                         cfq_put_queue(cfqd->async_cfqq[1][i]);
4416         }
4417
4418         if (cfqd->async_idle_cfqq)
4419                 cfq_put_queue(cfqd->async_idle_cfqq);
4420 }
4421
4422 static void cfq_exit_queue(struct elevator_queue *e)
4423 {
4424         struct cfq_data *cfqd = e->elevator_data;
4425         struct request_queue *q = cfqd->queue;
4426
4427         cfq_shutdown_timer_wq(cfqd);
4428
4429         spin_lock_irq(q->queue_lock);
4430
4431         if (cfqd->active_queue)
4432                 __cfq_slice_expired(cfqd, cfqd->active_queue, 0);
4433
4434         cfq_put_async_queues(cfqd);
4435
4436         spin_unlock_irq(q->queue_lock);
4437
4438         cfq_shutdown_timer_wq(cfqd);
4439
4440 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4441         blkcg_deactivate_policy(q, &blkcg_policy_cfq);
4442 #else
4443         kfree(cfqd->root_group);
4444 #endif
4445         kfree(cfqd);
4446 }
4447
4448 static int cfq_init_queue(struct request_queue *q, struct elevator_type *e)
4449 {
4450         struct cfq_data *cfqd;
4451         struct blkcg_gq *blkg __maybe_unused;
4452         int i, ret;
4453         struct elevator_queue *eq;
4454
4455         eq = elevator_alloc(q, e);
4456         if (!eq)
4457                 return -ENOMEM;
4458
4459         cfqd = kzalloc_node(sizeof(*cfqd), GFP_KERNEL, q->node);
4460         if (!cfqd) {
4461                 kobject_put(&eq->kobj);
4462                 return -ENOMEM;
4463         }
4464         eq->elevator_data = cfqd;
4465
4466         cfqd->queue = q;
4467         spin_lock_irq(q->queue_lock);
4468         q->elevator = eq;
4469         spin_unlock_irq(q->queue_lock);
4470
4471         /* Init root service tree */
4472         cfqd->grp_service_tree = CFQ_RB_ROOT;
4473
4474         /* Init root group and prefer root group over other groups by default */
4475 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4476         ret = blkcg_activate_policy(q, &blkcg_policy_cfq);
4477         if (ret)
4478                 goto out_free;
4479
4480         cfqd->root_group = blkg_to_cfqg(q->root_blkg);
4481 #else
4482         ret = -ENOMEM;
4483         cfqd->root_group = kzalloc_node(sizeof(*cfqd->root_group),
4484                                         GFP_KERNEL, cfqd->queue->node);
4485         if (!cfqd->root_group)
4486                 goto out_free;
4487
4488         cfq_init_cfqg_base(cfqd->root_group);
4489 #endif
4490         cfqd->root_group->weight = 2 * CFQ_WEIGHT_DEFAULT;
4491         cfqd->root_group->leaf_weight = 2 * CFQ_WEIGHT_DEFAULT;
4492
4493         /*
4494          * Not strictly needed (since RB_ROOT just clears the node and we
4495          * zeroed cfqd on alloc), but better be safe in case someone decides
4496          * to add magic to the rb code
4497          */
4498         for (i = 0; i < CFQ_PRIO_LISTS; i++)
4499                 cfqd->prio_trees[i] = RB_ROOT;
4500
4501         /*
4502          * Our fallback cfqq if cfq_find_alloc_queue() runs into OOM issues.
4503          * Grab a permanent reference to it, so that the normal code flow
4504          * will not attempt to free it.  oom_cfqq is linked to root_group
4505          * but shouldn't hold a reference as it'll never be unlinked.  Lose
4506          * the reference from linking right away.
4507          */
4508         cfq_init_cfqq(cfqd, &cfqd->oom_cfqq, 1, 0);
4509         cfqd->oom_cfqq.ref++;
4510
4511         spin_lock_irq(q->queue_lock);
4512         cfq_link_cfqq_cfqg(&cfqd->oom_cfqq, cfqd->root_group);
4513         cfqg_put(cfqd->root_group);
4514         spin_unlock_irq(q->queue_lock);
4515
4516         init_timer(&cfqd->idle_slice_timer);
4517         cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
4518         cfqd->idle_slice_timer.data = (unsigned long) cfqd;
4519
4520         INIT_WORK(&cfqd->unplug_work, cfq_kick_queue);
4521
4522         cfqd->cfq_quantum = cfq_quantum;
4523         cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0];
4524         cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1];
4525         cfqd->cfq_back_max = cfq_back_max;
4526         cfqd->cfq_back_penalty = cfq_back_penalty;
4527         cfqd->cfq_slice[0] = cfq_slice_async;
4528         cfqd->cfq_slice[1] = cfq_slice_sync;
4529         cfqd->cfq_target_latency = cfq_target_latency;
4530         cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
4531         cfqd->cfq_slice_idle = cfq_slice_idle;
4532         cfqd->cfq_group_idle = cfq_group_idle;
4533         cfqd->cfq_latency = 1;
4534         cfqd->hw_tag = -1;
4535         /*
4536          * we optimistically start assuming sync ops weren't delayed in last
4537          * second, in order to have larger depth for async operations.
4538          */
4539         cfqd->last_delayed_sync = jiffies - HZ;
4540         return 0;
4541
4542 out_free:
4543         kfree(cfqd);
4544         kobject_put(&eq->kobj);
4545         return ret;
4546 }
4547
4548 static void cfq_registered_queue(struct request_queue *q)
4549 {
4550         struct elevator_queue *e = q->elevator;
4551         struct cfq_data *cfqd = e->elevator_data;
4552
4553         /*
4554          * Default to IOPS mode with no idling for SSDs
4555          */
4556         if (blk_queue_nonrot(q))
4557                 cfqd->cfq_slice_idle = 0;
4558 }
4559
4560 /*
4561  * sysfs parts below -->
4562  */
4563 static ssize_t
4564 cfq_var_show(unsigned int var, char *page)
4565 {
4566         return sprintf(page, "%u\n", var);
4567 }
4568
4569 static ssize_t
4570 cfq_var_store(unsigned int *var, const char *page, size_t count)
4571 {
4572         char *p = (char *) page;
4573
4574         *var = simple_strtoul(p, &p, 10);
4575         return count;
4576 }
4577
4578 #define SHOW_FUNCTION(__FUNC, __VAR, __CONV)                            \
4579 static ssize_t __FUNC(struct elevator_queue *e, char *page)             \
4580 {                                                                       \
4581         struct cfq_data *cfqd = e->elevator_data;                       \
4582         unsigned int __data = __VAR;                                    \
4583         if (__CONV)                                                     \
4584                 __data = jiffies_to_msecs(__data);                      \
4585         return cfq_var_show(__data, (page));                            \
4586 }
4587 SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0);
4588 SHOW_FUNCTION(cfq_fifo_expire_sync_show, cfqd->cfq_fifo_expire[1], 1);
4589 SHOW_FUNCTION(cfq_fifo_expire_async_show, cfqd->cfq_fifo_expire[0], 1);
4590 SHOW_FUNCTION(cfq_back_seek_max_show, cfqd->cfq_back_max, 0);
4591 SHOW_FUNCTION(cfq_back_seek_penalty_show, cfqd->cfq_back_penalty, 0);
4592 SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1);
4593 SHOW_FUNCTION(cfq_group_idle_show, cfqd->cfq_group_idle, 1);
4594 SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1);
4595 SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
4596 SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
4597 SHOW_FUNCTION(cfq_low_latency_show, cfqd->cfq_latency, 0);
4598 SHOW_FUNCTION(cfq_target_latency_show, cfqd->cfq_target_latency, 1);
4599 #undef SHOW_FUNCTION
4600
4601 #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV)                 \
4602 static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count) \
4603 {                                                                       \
4604         struct cfq_data *cfqd = e->elevator_data;                       \
4605         unsigned int __data;                                            \
4606         int ret = cfq_var_store(&__data, (page), count);                \
4607         if (__data < (MIN))                                             \
4608                 __data = (MIN);                                         \
4609         else if (__data > (MAX))                                        \
4610                 __data = (MAX);                                         \
4611         if (__CONV)                                                     \
4612                 *(__PTR) = msecs_to_jiffies(__data);                    \
4613         else                                                            \
4614                 *(__PTR) = __data;                                      \
4615         return ret;                                                     \
4616 }
4617 STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
4618 STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1,
4619                 UINT_MAX, 1);
4620 STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1,
4621                 UINT_MAX, 1);
4622 STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
4623 STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1,
4624                 UINT_MAX, 0);
4625 STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1);
4626 STORE_FUNCTION(cfq_group_idle_store, &cfqd->cfq_group_idle, 0, UINT_MAX, 1);
4627 STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1);
4628 STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1);
4629 STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1,
4630                 UINT_MAX, 0);
4631 STORE_FUNCTION(cfq_low_latency_store, &cfqd->cfq_latency, 0, 1, 0);
4632 STORE_FUNCTION(cfq_target_latency_store, &cfqd->cfq_target_latency, 1, UINT_MAX, 1);
4633 #undef STORE_FUNCTION
4634
4635 #define CFQ_ATTR(name) \
4636         __ATTR(name, S_IRUGO|S_IWUSR, cfq_##name##_show, cfq_##name##_store)
4637
4638 static struct elv_fs_entry cfq_attrs[] = {
4639         CFQ_ATTR(quantum),
4640         CFQ_ATTR(fifo_expire_sync),
4641         CFQ_ATTR(fifo_expire_async),
4642         CFQ_ATTR(back_seek_max),
4643         CFQ_ATTR(back_seek_penalty),
4644         CFQ_ATTR(slice_sync),
4645         CFQ_ATTR(slice_async),
4646         CFQ_ATTR(slice_async_rq),
4647         CFQ_ATTR(slice_idle),
4648         CFQ_ATTR(group_idle),
4649         CFQ_ATTR(low_latency),
4650         CFQ_ATTR(target_latency),
4651         __ATTR_NULL
4652 };
4653
4654 static struct elevator_type iosched_cfq = {
4655         .ops = {
4656                 .elevator_merge_fn =            cfq_merge,
4657                 .elevator_merged_fn =           cfq_merged_request,
4658                 .elevator_merge_req_fn =        cfq_merged_requests,
4659                 .elevator_allow_merge_fn =      cfq_allow_merge,
4660                 .elevator_bio_merged_fn =       cfq_bio_merged,
4661                 .elevator_dispatch_fn =         cfq_dispatch_requests,
4662                 .elevator_add_req_fn =          cfq_insert_request,
4663                 .elevator_activate_req_fn =     cfq_activate_request,
4664                 .elevator_deactivate_req_fn =   cfq_deactivate_request,
4665                 .elevator_completed_req_fn =    cfq_completed_request,
4666                 .elevator_former_req_fn =       elv_rb_former_request,
4667                 .elevator_latter_req_fn =       elv_rb_latter_request,
4668                 .elevator_init_icq_fn =         cfq_init_icq,
4669                 .elevator_exit_icq_fn =         cfq_exit_icq,
4670                 .elevator_set_req_fn =          cfq_set_request,
4671                 .elevator_put_req_fn =          cfq_put_request,
4672                 .elevator_may_queue_fn =        cfq_may_queue,
4673                 .elevator_init_fn =             cfq_init_queue,
4674                 .elevator_exit_fn =             cfq_exit_queue,
4675                 .elevator_registered_fn =       cfq_registered_queue,
4676         },
4677         .icq_size       =       sizeof(struct cfq_io_cq),
4678         .icq_align      =       __alignof__(struct cfq_io_cq),
4679         .elevator_attrs =       cfq_attrs,
4680         .elevator_name  =       "cfq",
4681         .elevator_owner =       THIS_MODULE,
4682 };
4683
4684 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4685 static struct blkcg_policy blkcg_policy_cfq = {
4686         .pd_size                = sizeof(struct cfq_group),
4687         .cpd_size               = sizeof(struct cfq_group_data),
4688         .cftypes                = cfq_blkcg_files,
4689
4690         .cpd_init_fn            = cfq_cpd_init,
4691         .pd_init_fn             = cfq_pd_init,
4692         .pd_offline_fn          = cfq_pd_offline,
4693         .pd_reset_stats_fn      = cfq_pd_reset_stats,
4694 };
4695 #endif
4696
4697 static int __init cfq_init(void)
4698 {
4699         int ret;
4700
4701         /*
4702          * could be 0 on HZ < 1000 setups
4703          */
4704         if (!cfq_slice_async)
4705                 cfq_slice_async = 1;
4706         if (!cfq_slice_idle)
4707                 cfq_slice_idle = 1;
4708
4709 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4710         if (!cfq_group_idle)
4711                 cfq_group_idle = 1;
4712
4713         ret = blkcg_policy_register(&blkcg_policy_cfq);
4714         if (ret)
4715                 return ret;
4716 #else
4717         cfq_group_idle = 0;
4718 #endif
4719
4720         ret = -ENOMEM;
4721         cfq_pool = KMEM_CACHE(cfq_queue, 0);
4722         if (!cfq_pool)
4723                 goto err_pol_unreg;
4724
4725         ret = elv_register(&iosched_cfq);
4726         if (ret)
4727                 goto err_free_pool;
4728
4729         return 0;
4730
4731 err_free_pool:
4732         kmem_cache_destroy(cfq_pool);
4733 err_pol_unreg:
4734 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4735         blkcg_policy_unregister(&blkcg_policy_cfq);
4736 #endif
4737         return ret;
4738 }
4739
4740 static void __exit cfq_exit(void)
4741 {
4742 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4743         blkcg_policy_unregister(&blkcg_policy_cfq);
4744 #endif
4745         elv_unregister(&iosched_cfq);
4746         kmem_cache_destroy(cfq_pool);
4747 }
4748
4749 module_init(cfq_init);
4750 module_exit(cfq_exit);
4751
4752 MODULE_AUTHOR("Jens Axboe");
4753 MODULE_LICENSE("GPL");
4754 MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler");