Back to home page

LXR

 
 

    


0001 /*
0002  *  CFQ, or complete fairness queueing, disk scheduler.
0003  *
0004  *  Based on ideas from a previously unfinished io
0005  *  scheduler (round robin per-process disk scheduling) and Andrea Arcangeli.
0006  *
0007  *  Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
0008  */
0009 #include <linux/module.h>
0010 #include <linux/slab.h>
0011 #include <linux/blkdev.h>
0012 #include <linux/elevator.h>
0013 #include <linux/ktime.h>
0014 #include <linux/rbtree.h>
0015 #include <linux/ioprio.h>
0016 #include <linux/blktrace_api.h>
0017 #include <linux/blk-cgroup.h>
0018 #include "blk.h"
0019 #include "blk-wbt.h"
0020 
0021 /*
0022  * tunables
0023  */
0024 /* max queue in one round of service */
0025 static const int cfq_quantum = 8;
0026 static const u64 cfq_fifo_expire[2] = { NSEC_PER_SEC / 4, NSEC_PER_SEC / 8 };
0027 /* maximum backwards seek, in KiB */
0028 static const int cfq_back_max = 16 * 1024;
0029 /* penalty of a backwards seek */
0030 static const int cfq_back_penalty = 2;
0031 static const u64 cfq_slice_sync = NSEC_PER_SEC / 10;
0032 static u64 cfq_slice_async = NSEC_PER_SEC / 25;
0033 static const int cfq_slice_async_rq = 2;
0034 static u64 cfq_slice_idle = NSEC_PER_SEC / 125;
0035 static u64 cfq_group_idle = NSEC_PER_SEC / 125;
0036 static const u64 cfq_target_latency = (u64)NSEC_PER_SEC * 3/10; /* 300 ms */
0037 static const int cfq_hist_divisor = 4;
0038 
0039 /*
0040  * offset from end of service tree
0041  */
0042 #define CFQ_IDLE_DELAY      (NSEC_PER_SEC / 5)
0043 
0044 /*
0045  * below this threshold, we consider thinktime immediate
0046  */
0047 #define CFQ_MIN_TT      (2 * NSEC_PER_SEC / HZ)
0048 
0049 #define CFQ_SLICE_SCALE     (5)
0050 #define CFQ_HW_QUEUE_MIN    (5)
0051 #define CFQ_SERVICE_SHIFT       12
0052 
0053 #define CFQQ_SEEK_THR       (sector_t)(8 * 100)
0054 #define CFQQ_CLOSE_THR      (sector_t)(8 * 1024)
0055 #define CFQQ_SECT_THR_NONROT    (sector_t)(2 * 32)
0056 #define CFQQ_SEEKY(cfqq)    (hweight32(cfqq->seek_history) > 32/8)
0057 
0058 #define RQ_CIC(rq)      icq_to_cic((rq)->elv.icq)
0059 #define RQ_CFQQ(rq)     (struct cfq_queue *) ((rq)->elv.priv[0])
0060 #define RQ_CFQG(rq)     (struct cfq_group *) ((rq)->elv.priv[1])
0061 
0062 static struct kmem_cache *cfq_pool;
0063 
0064 #define CFQ_PRIO_LISTS      IOPRIO_BE_NR
0065 #define cfq_class_idle(cfqq)    ((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
0066 #define cfq_class_rt(cfqq)  ((cfqq)->ioprio_class == IOPRIO_CLASS_RT)
0067 
0068 #define sample_valid(samples)   ((samples) > 80)
0069 #define rb_entry_cfqg(node) rb_entry((node), struct cfq_group, rb_node)
0070 
0071 /* blkio-related constants */
0072 #define CFQ_WEIGHT_LEGACY_MIN   10
0073 #define CFQ_WEIGHT_LEGACY_DFL   500
0074 #define CFQ_WEIGHT_LEGACY_MAX   1000
0075 
0076 struct cfq_ttime {
0077     u64 last_end_request;
0078 
0079     u64 ttime_total;
0080     u64 ttime_mean;
0081     unsigned long ttime_samples;
0082 };
0083 
0084 /*
0085  * Most of our rbtree usage is for sorting with min extraction, so
0086  * if we cache the leftmost node we don't have to walk down the tree
0087  * to find it. Idea borrowed from Ingo Molnars CFS scheduler. We should
0088  * move this into the elevator for the rq sorting as well.
0089  */
0090 struct cfq_rb_root {
0091     struct rb_root rb;
0092     struct rb_node *left;
0093     unsigned count;
0094     u64 min_vdisktime;
0095     struct cfq_ttime ttime;
0096 };
0097 #define CFQ_RB_ROOT (struct cfq_rb_root) { .rb = RB_ROOT, \
0098             .ttime = {.last_end_request = ktime_get_ns(),},}
0099 
0100 /*
0101  * Per process-grouping structure
0102  */
0103 struct cfq_queue {
0104     /* reference count */
0105     int ref;
0106     /* various state flags, see below */
0107     unsigned int flags;
0108     /* parent cfq_data */
0109     struct cfq_data *cfqd;
0110     /* service_tree member */
0111     struct rb_node rb_node;
0112     /* service_tree key */
0113     u64 rb_key;
0114     /* prio tree member */
0115     struct rb_node p_node;
0116     /* prio tree root we belong to, if any */
0117     struct rb_root *p_root;
0118     /* sorted list of pending requests */
0119     struct rb_root sort_list;
0120     /* if fifo isn't expired, next request to serve */
0121     struct request *next_rq;
0122     /* requests queued in sort_list */
0123     int queued[2];
0124     /* currently allocated requests */
0125     int allocated[2];
0126     /* fifo list of requests in sort_list */
0127     struct list_head fifo;
0128 
0129     /* time when queue got scheduled in to dispatch first request. */
0130     u64 dispatch_start;
0131     u64 allocated_slice;
0132     u64 slice_dispatch;
0133     /* time when first request from queue completed and slice started. */
0134     u64 slice_start;
0135     u64 slice_end;
0136     s64 slice_resid;
0137 
0138     /* pending priority requests */
0139     int prio_pending;
0140     /* number of requests that are on the dispatch list or inside driver */
0141     int dispatched;
0142 
0143     /* io prio of this group */
0144     unsigned short ioprio, org_ioprio;
0145     unsigned short ioprio_class, org_ioprio_class;
0146 
0147     pid_t pid;
0148 
0149     u32 seek_history;
0150     sector_t last_request_pos;
0151 
0152     struct cfq_rb_root *service_tree;
0153     struct cfq_queue *new_cfqq;
0154     struct cfq_group *cfqg;
0155     /* Number of sectors dispatched from queue in single dispatch round */
0156     unsigned long nr_sectors;
0157 };
0158 
0159 /*
0160  * First index in the service_trees.
0161  * IDLE is handled separately, so it has negative index
0162  */
0163 enum wl_class_t {
0164     BE_WORKLOAD = 0,
0165     RT_WORKLOAD = 1,
0166     IDLE_WORKLOAD = 2,
0167     CFQ_PRIO_NR,
0168 };
0169 
0170 /*
0171  * Second index in the service_trees.
0172  */
0173 enum wl_type_t {
0174     ASYNC_WORKLOAD = 0,
0175     SYNC_NOIDLE_WORKLOAD = 1,
0176     SYNC_WORKLOAD = 2
0177 };
0178 
0179 struct cfqg_stats {
0180 #ifdef CONFIG_CFQ_GROUP_IOSCHED
0181     /* number of ios merged */
0182     struct blkg_rwstat      merged;
0183     /* total time spent on device in ns, may not be accurate w/ queueing */
0184     struct blkg_rwstat      service_time;
0185     /* total time spent waiting in scheduler queue in ns */
0186     struct blkg_rwstat      wait_time;
0187     /* number of IOs queued up */
0188     struct blkg_rwstat      queued;
0189     /* total disk time and nr sectors dispatched by this group */
0190     struct blkg_stat        time;
0191 #ifdef CONFIG_DEBUG_BLK_CGROUP
0192     /* time not charged to this cgroup */
0193     struct blkg_stat        unaccounted_time;
0194     /* sum of number of ios queued across all samples */
0195     struct blkg_stat        avg_queue_size_sum;
0196     /* count of samples taken for average */
0197     struct blkg_stat        avg_queue_size_samples;
0198     /* how many times this group has been removed from service tree */
0199     struct blkg_stat        dequeue;
0200     /* total time spent waiting for it to be assigned a timeslice. */
0201     struct blkg_stat        group_wait_time;
0202     /* time spent idling for this blkcg_gq */
0203     struct blkg_stat        idle_time;
0204     /* total time with empty current active q with other requests queued */
0205     struct blkg_stat        empty_time;
0206     /* fields after this shouldn't be cleared on stat reset */
0207     uint64_t            start_group_wait_time;
0208     uint64_t            start_idle_time;
0209     uint64_t            start_empty_time;
0210     uint16_t            flags;
0211 #endif  /* CONFIG_DEBUG_BLK_CGROUP */
0212 #endif  /* CONFIG_CFQ_GROUP_IOSCHED */
0213 };
0214 
0215 /* Per-cgroup data */
0216 struct cfq_group_data {
0217     /* must be the first member */
0218     struct blkcg_policy_data cpd;
0219 
0220     unsigned int weight;
0221     unsigned int leaf_weight;
0222 };
0223 
0224 /* This is per cgroup per device grouping structure */
0225 struct cfq_group {
0226     /* must be the first member */
0227     struct blkg_policy_data pd;
0228 
0229     /* group service_tree member */
0230     struct rb_node rb_node;
0231 
0232     /* group service_tree key */
0233     u64 vdisktime;
0234 
0235     /*
0236      * The number of active cfqgs and sum of their weights under this
0237      * cfqg.  This covers this cfqg's leaf_weight and all children's
0238      * weights, but does not cover weights of further descendants.
0239      *
0240      * If a cfqg is on the service tree, it's active.  An active cfqg
0241      * also activates its parent and contributes to the children_weight
0242      * of the parent.
0243      */
0244     int nr_active;
0245     unsigned int children_weight;
0246 
0247     /*
0248      * vfraction is the fraction of vdisktime that the tasks in this
0249      * cfqg are entitled to.  This is determined by compounding the
0250      * ratios walking up from this cfqg to the root.
0251      *
0252      * It is in fixed point w/ CFQ_SERVICE_SHIFT and the sum of all
0253      * vfractions on a service tree is approximately 1.  The sum may
0254      * deviate a bit due to rounding errors and fluctuations caused by
0255      * cfqgs entering and leaving the service tree.
0256      */
0257     unsigned int vfraction;
0258 
0259     /*
0260      * There are two weights - (internal) weight is the weight of this
0261      * cfqg against the sibling cfqgs.  leaf_weight is the wight of
0262      * this cfqg against the child cfqgs.  For the root cfqg, both
0263      * weights are kept in sync for backward compatibility.
0264      */
0265     unsigned int weight;
0266     unsigned int new_weight;
0267     unsigned int dev_weight;
0268 
0269     unsigned int leaf_weight;
0270     unsigned int new_leaf_weight;
0271     unsigned int dev_leaf_weight;
0272 
0273     /* number of cfqq currently on this group */
0274     int nr_cfqq;
0275 
0276     /*
0277      * Per group busy queues average. Useful for workload slice calc. We
0278      * create the array for each prio class but at run time it is used
0279      * only for RT and BE class and slot for IDLE class remains unused.
0280      * This is primarily done to avoid confusion and a gcc warning.
0281      */
0282     unsigned int busy_queues_avg[CFQ_PRIO_NR];
0283     /*
0284      * rr lists of queues with requests. We maintain service trees for
0285      * RT and BE classes. These trees are subdivided in subclasses
0286      * of SYNC, SYNC_NOIDLE and ASYNC based on workload type. For IDLE
0287      * class there is no subclassification and all the cfq queues go on
0288      * a single tree service_tree_idle.
0289      * Counts are embedded in the cfq_rb_root
0290      */
0291     struct cfq_rb_root service_trees[2][3];
0292     struct cfq_rb_root service_tree_idle;
0293 
0294     u64 saved_wl_slice;
0295     enum wl_type_t saved_wl_type;
0296     enum wl_class_t saved_wl_class;
0297 
0298     /* number of requests that are on the dispatch list or inside driver */
0299     int dispatched;
0300     struct cfq_ttime ttime;
0301     struct cfqg_stats stats;    /* stats for this cfqg */
0302 
0303     /* async queue for each priority case */
0304     struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
0305     struct cfq_queue *async_idle_cfqq;
0306 
0307 };
0308 
0309 struct cfq_io_cq {
0310     struct io_cq        icq;        /* must be the first member */
0311     struct cfq_queue    *cfqq[2];
0312     struct cfq_ttime    ttime;
0313     int         ioprio;     /* the current ioprio */
0314 #ifdef CONFIG_CFQ_GROUP_IOSCHED
0315     uint64_t        blkcg_serial_nr; /* the current blkcg serial */
0316 #endif
0317 };
0318 
0319 /*
0320  * Per block device queue structure
0321  */
0322 struct cfq_data {
0323     struct request_queue *queue;
0324     /* Root service tree for cfq_groups */
0325     struct cfq_rb_root grp_service_tree;
0326     struct cfq_group *root_group;
0327 
0328     /*
0329      * The priority currently being served
0330      */
0331     enum wl_class_t serving_wl_class;
0332     enum wl_type_t serving_wl_type;
0333     u64 workload_expires;
0334     struct cfq_group *serving_group;
0335 
0336     /*
0337      * Each priority tree is sorted by next_request position.  These
0338      * trees are used when determining if two or more queues are
0339      * interleaving requests (see cfq_close_cooperator).
0340      */
0341     struct rb_root prio_trees[CFQ_PRIO_LISTS];
0342 
0343     unsigned int busy_queues;
0344     unsigned int busy_sync_queues;
0345 
0346     int rq_in_driver;
0347     int rq_in_flight[2];
0348 
0349     /*
0350      * queue-depth detection
0351      */
0352     int rq_queued;
0353     int hw_tag;
0354     /*
0355      * hw_tag can be
0356      * -1 => indeterminate, (cfq will behave as if NCQ is present, to allow better detection)
0357      *  1 => NCQ is present (hw_tag_est_depth is the estimated max depth)
0358      *  0 => no NCQ
0359      */
0360     int hw_tag_est_depth;
0361     unsigned int hw_tag_samples;
0362 
0363     /*
0364      * idle window management
0365      */
0366     struct hrtimer idle_slice_timer;
0367     struct work_struct unplug_work;
0368 
0369     struct cfq_queue *active_queue;
0370     struct cfq_io_cq *active_cic;
0371 
0372     sector_t last_position;
0373 
0374     /*
0375      * tunables, see top of file
0376      */
0377     unsigned int cfq_quantum;
0378     unsigned int cfq_back_penalty;
0379     unsigned int cfq_back_max;
0380     unsigned int cfq_slice_async_rq;
0381     unsigned int cfq_latency;
0382     u64 cfq_fifo_expire[2];
0383     u64 cfq_slice[2];
0384     u64 cfq_slice_idle;
0385     u64 cfq_group_idle;
0386     u64 cfq_target_latency;
0387 
0388     /*
0389      * Fallback dummy cfqq for extreme OOM conditions
0390      */
0391     struct cfq_queue oom_cfqq;
0392 
0393     u64 last_delayed_sync;
0394 };
0395 
0396 static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd);
0397 static void cfq_put_queue(struct cfq_queue *cfqq);
0398 
0399 static struct cfq_rb_root *st_for(struct cfq_group *cfqg,
0400                         enum wl_class_t class,
0401                         enum wl_type_t type)
0402 {
0403     if (!cfqg)
0404         return NULL;
0405 
0406     if (class == IDLE_WORKLOAD)
0407         return &cfqg->service_tree_idle;
0408 
0409     return &cfqg->service_trees[class][type];
0410 }
0411 
0412 enum cfqq_state_flags {
0413     CFQ_CFQQ_FLAG_on_rr = 0,    /* on round-robin busy list */
0414     CFQ_CFQQ_FLAG_wait_request, /* waiting for a request */
0415     CFQ_CFQQ_FLAG_must_dispatch,    /* must be allowed a dispatch */
0416     CFQ_CFQQ_FLAG_must_alloc_slice, /* per-slice must_alloc flag */
0417     CFQ_CFQQ_FLAG_fifo_expire,  /* FIFO checked in this slice */
0418     CFQ_CFQQ_FLAG_idle_window,  /* slice idling enabled */
0419     CFQ_CFQQ_FLAG_prio_changed, /* task priority has changed */
0420     CFQ_CFQQ_FLAG_slice_new,    /* no requests dispatched in slice */
0421     CFQ_CFQQ_FLAG_sync,     /* synchronous queue */
0422     CFQ_CFQQ_FLAG_coop,     /* cfqq is shared */
0423     CFQ_CFQQ_FLAG_split_coop,   /* shared cfqq will be splitted */
0424     CFQ_CFQQ_FLAG_deep,     /* sync cfqq experienced large depth */
0425     CFQ_CFQQ_FLAG_wait_busy,    /* Waiting for next request */
0426 };
0427 
0428 #define CFQ_CFQQ_FNS(name)                      \
0429 static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)     \
0430 {                                   \
0431     (cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name);           \
0432 }                                   \
0433 static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)    \
0434 {                                   \
0435     (cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);          \
0436 }                                   \
0437 static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)     \
0438 {                                   \
0439     return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;  \
0440 }
0441 
0442 CFQ_CFQQ_FNS(on_rr);
0443 CFQ_CFQQ_FNS(wait_request);
0444 CFQ_CFQQ_FNS(must_dispatch);
0445 CFQ_CFQQ_FNS(must_alloc_slice);
0446 CFQ_CFQQ_FNS(fifo_expire);
0447 CFQ_CFQQ_FNS(idle_window);
0448 CFQ_CFQQ_FNS(prio_changed);
0449 CFQ_CFQQ_FNS(slice_new);
0450 CFQ_CFQQ_FNS(sync);
0451 CFQ_CFQQ_FNS(coop);
0452 CFQ_CFQQ_FNS(split_coop);
0453 CFQ_CFQQ_FNS(deep);
0454 CFQ_CFQQ_FNS(wait_busy);
0455 #undef CFQ_CFQQ_FNS
0456 
0457 #if defined(CONFIG_CFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP)
0458 
0459 /* cfqg stats flags */
0460 enum cfqg_stats_flags {
0461     CFQG_stats_waiting = 0,
0462     CFQG_stats_idling,
0463     CFQG_stats_empty,
0464 };
0465 
0466 #define CFQG_FLAG_FNS(name)                     \
0467 static inline void cfqg_stats_mark_##name(struct cfqg_stats *stats) \
0468 {                                   \
0469     stats->flags |= (1 << CFQG_stats_##name);           \
0470 }                                   \
0471 static inline void cfqg_stats_clear_##name(struct cfqg_stats *stats)    \
0472 {                                   \
0473     stats->flags &= ~(1 << CFQG_stats_##name);          \
0474 }                                   \
0475 static inline int cfqg_stats_##name(struct cfqg_stats *stats)       \
0476 {                                   \
0477     return (stats->flags & (1 << CFQG_stats_##name)) != 0;      \
0478 }                                   \
0479 
0480 CFQG_FLAG_FNS(waiting)
0481 CFQG_FLAG_FNS(idling)
0482 CFQG_FLAG_FNS(empty)
0483 #undef CFQG_FLAG_FNS
0484 
0485 /* This should be called with the queue_lock held. */
0486 static void cfqg_stats_update_group_wait_time(struct cfqg_stats *stats)
0487 {
0488     unsigned long long now;
0489 
0490     if (!cfqg_stats_waiting(stats))
0491         return;
0492 
0493     now = sched_clock();
0494     if (time_after64(now, stats->start_group_wait_time))
0495         blkg_stat_add(&stats->group_wait_time,
0496                   now - stats->start_group_wait_time);
0497     cfqg_stats_clear_waiting(stats);
0498 }
0499 
0500 /* This should be called with the queue_lock held. */
0501 static void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg,
0502                          struct cfq_group *curr_cfqg)
0503 {
0504     struct cfqg_stats *stats = &cfqg->stats;
0505 
0506     if (cfqg_stats_waiting(stats))
0507         return;
0508     if (cfqg == curr_cfqg)
0509         return;
0510     stats->start_group_wait_time = sched_clock();
0511     cfqg_stats_mark_waiting(stats);
0512 }
0513 
0514 /* This should be called with the queue_lock held. */
0515 static void cfqg_stats_end_empty_time(struct cfqg_stats *stats)
0516 {
0517     unsigned long long now;
0518 
0519     if (!cfqg_stats_empty(stats))
0520         return;
0521 
0522     now = sched_clock();
0523     if (time_after64(now, stats->start_empty_time))
0524         blkg_stat_add(&stats->empty_time,
0525                   now - stats->start_empty_time);
0526     cfqg_stats_clear_empty(stats);
0527 }
0528 
0529 static void cfqg_stats_update_dequeue(struct cfq_group *cfqg)
0530 {
0531     blkg_stat_add(&cfqg->stats.dequeue, 1);
0532 }
0533 
0534 static void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg)
0535 {
0536     struct cfqg_stats *stats = &cfqg->stats;
0537 
0538     if (blkg_rwstat_total(&stats->queued))
0539         return;
0540 
0541     /*
0542      * group is already marked empty. This can happen if cfqq got new
0543      * request in parent group and moved to this group while being added
0544      * to service tree. Just ignore the event and move on.
0545      */
0546     if (cfqg_stats_empty(stats))
0547         return;
0548 
0549     stats->start_empty_time = sched_clock();
0550     cfqg_stats_mark_empty(stats);
0551 }
0552 
0553 static void cfqg_stats_update_idle_time(struct cfq_group *cfqg)
0554 {
0555     struct cfqg_stats *stats = &cfqg->stats;
0556 
0557     if (cfqg_stats_idling(stats)) {
0558         unsigned long long now = sched_clock();
0559 
0560         if (time_after64(now, stats->start_idle_time))
0561             blkg_stat_add(&stats->idle_time,
0562                       now - stats->start_idle_time);
0563         cfqg_stats_clear_idling(stats);
0564     }
0565 }
0566 
0567 static void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg)
0568 {
0569     struct cfqg_stats *stats = &cfqg->stats;
0570 
0571     BUG_ON(cfqg_stats_idling(stats));
0572 
0573     stats->start_idle_time = sched_clock();
0574     cfqg_stats_mark_idling(stats);
0575 }
0576 
0577 static void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg)
0578 {
0579     struct cfqg_stats *stats = &cfqg->stats;
0580 
0581     blkg_stat_add(&stats->avg_queue_size_sum,
0582               blkg_rwstat_total(&stats->queued));
0583     blkg_stat_add(&stats->avg_queue_size_samples, 1);
0584     cfqg_stats_update_group_wait_time(stats);
0585 }
0586 
0587 #else   /* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */
0588 
0589 static inline void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg, struct cfq_group *curr_cfqg) { }
0590 static inline void cfqg_stats_end_empty_time(struct cfqg_stats *stats) { }
0591 static inline void cfqg_stats_update_dequeue(struct cfq_group *cfqg) { }
0592 static inline void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg) { }
0593 static inline void cfqg_stats_update_idle_time(struct cfq_group *cfqg) { }
0594 static inline void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg) { }
0595 static inline void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg) { }
0596 
0597 #endif  /* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */
0598 
0599 #ifdef CONFIG_CFQ_GROUP_IOSCHED
0600 
0601 static inline struct cfq_group *pd_to_cfqg(struct blkg_policy_data *pd)
0602 {
0603     return pd ? container_of(pd, struct cfq_group, pd) : NULL;
0604 }
0605 
0606 static struct cfq_group_data
0607 *cpd_to_cfqgd(struct blkcg_policy_data *cpd)
0608 {
0609     return cpd ? container_of(cpd, struct cfq_group_data, cpd) : NULL;
0610 }
0611 
0612 static inline struct blkcg_gq *cfqg_to_blkg(struct cfq_group *cfqg)
0613 {
0614     return pd_to_blkg(&cfqg->pd);
0615 }
0616 
0617 static struct blkcg_policy blkcg_policy_cfq;
0618 
0619 static inline struct cfq_group *blkg_to_cfqg(struct blkcg_gq *blkg)
0620 {
0621     return pd_to_cfqg(blkg_to_pd(blkg, &blkcg_policy_cfq));
0622 }
0623 
0624 static struct cfq_group_data *blkcg_to_cfqgd(struct blkcg *blkcg)
0625 {
0626     return cpd_to_cfqgd(blkcg_to_cpd(blkcg, &blkcg_policy_cfq));
0627 }
0628 
0629 static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg)
0630 {
0631     struct blkcg_gq *pblkg = cfqg_to_blkg(cfqg)->parent;
0632 
0633     return pblkg ? blkg_to_cfqg(pblkg) : NULL;
0634 }
0635 
0636 static inline bool cfqg_is_descendant(struct cfq_group *cfqg,
0637                       struct cfq_group *ancestor)
0638 {
0639     return cgroup_is_descendant(cfqg_to_blkg(cfqg)->blkcg->css.cgroup,
0640                     cfqg_to_blkg(ancestor)->blkcg->css.cgroup);
0641 }
0642 
0643 static inline void cfqg_get(struct cfq_group *cfqg)
0644 {
0645     return blkg_get(cfqg_to_blkg(cfqg));
0646 }
0647 
0648 static inline void cfqg_put(struct cfq_group *cfqg)
0649 {
0650     return blkg_put(cfqg_to_blkg(cfqg));
0651 }
0652 
0653 #define cfq_log_cfqq(cfqd, cfqq, fmt, args...)  do {            \
0654     char __pbuf[128];                       \
0655                                     \
0656     blkg_path(cfqg_to_blkg((cfqq)->cfqg), __pbuf, sizeof(__pbuf));  \
0657     blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c %s " fmt, (cfqq)->pid, \
0658             cfq_cfqq_sync((cfqq)) ? 'S' : 'A',      \
0659             cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
0660               __pbuf, ##args);              \
0661 } while (0)
0662 
0663 #define cfq_log_cfqg(cfqd, cfqg, fmt, args...)  do {            \
0664     char __pbuf[128];                       \
0665                                     \
0666     blkg_path(cfqg_to_blkg(cfqg), __pbuf, sizeof(__pbuf));      \
0667     blk_add_trace_msg((cfqd)->queue, "%s " fmt, __pbuf, ##args);    \
0668 } while (0)
0669 
0670 static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
0671                         struct cfq_group *curr_cfqg,
0672                         unsigned int op)
0673 {
0674     blkg_rwstat_add(&cfqg->stats.queued, op, 1);
0675     cfqg_stats_end_empty_time(&cfqg->stats);
0676     cfqg_stats_set_start_group_wait_time(cfqg, curr_cfqg);
0677 }
0678 
0679 static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
0680             uint64_t time, unsigned long unaccounted_time)
0681 {
0682     blkg_stat_add(&cfqg->stats.time, time);
0683 #ifdef CONFIG_DEBUG_BLK_CGROUP
0684     blkg_stat_add(&cfqg->stats.unaccounted_time, unaccounted_time);
0685 #endif
0686 }
0687 
0688 static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg,
0689                            unsigned int op)
0690 {
0691     blkg_rwstat_add(&cfqg->stats.queued, op, -1);
0692 }
0693 
0694 static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg,
0695                            unsigned int op)
0696 {
0697     blkg_rwstat_add(&cfqg->stats.merged, op, 1);
0698 }
0699 
0700 static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
0701             uint64_t start_time, uint64_t io_start_time,
0702             unsigned int op)
0703 {
0704     struct cfqg_stats *stats = &cfqg->stats;
0705     unsigned long long now = sched_clock();
0706 
0707     if (time_after64(now, io_start_time))
0708         blkg_rwstat_add(&stats->service_time, op, now - io_start_time);
0709     if (time_after64(io_start_time, start_time))
0710         blkg_rwstat_add(&stats->wait_time, op,
0711                 io_start_time - start_time);
0712 }
0713 
0714 /* @stats = 0 */
0715 static void cfqg_stats_reset(struct cfqg_stats *stats)
0716 {
0717     /* queued stats shouldn't be cleared */
0718     blkg_rwstat_reset(&stats->merged);
0719     blkg_rwstat_reset(&stats->service_time);
0720     blkg_rwstat_reset(&stats->wait_time);
0721     blkg_stat_reset(&stats->time);
0722 #ifdef CONFIG_DEBUG_BLK_CGROUP
0723     blkg_stat_reset(&stats->unaccounted_time);
0724     blkg_stat_reset(&stats->avg_queue_size_sum);
0725     blkg_stat_reset(&stats->avg_queue_size_samples);
0726     blkg_stat_reset(&stats->dequeue);
0727     blkg_stat_reset(&stats->group_wait_time);
0728     blkg_stat_reset(&stats->idle_time);
0729     blkg_stat_reset(&stats->empty_time);
0730 #endif
0731 }
0732 
0733 /* @to += @from */
0734 static void cfqg_stats_add_aux(struct cfqg_stats *to, struct cfqg_stats *from)
0735 {
0736     /* queued stats shouldn't be cleared */
0737     blkg_rwstat_add_aux(&to->merged, &from->merged);
0738     blkg_rwstat_add_aux(&to->service_time, &from->service_time);
0739     blkg_rwstat_add_aux(&to->wait_time, &from->wait_time);
0740     blkg_stat_add_aux(&from->time, &from->time);
0741 #ifdef CONFIG_DEBUG_BLK_CGROUP
0742     blkg_stat_add_aux(&to->unaccounted_time, &from->unaccounted_time);
0743     blkg_stat_add_aux(&to->avg_queue_size_sum, &from->avg_queue_size_sum);
0744     blkg_stat_add_aux(&to->avg_queue_size_samples, &from->avg_queue_size_samples);
0745     blkg_stat_add_aux(&to->dequeue, &from->dequeue);
0746     blkg_stat_add_aux(&to->group_wait_time, &from->group_wait_time);
0747     blkg_stat_add_aux(&to->idle_time, &from->idle_time);
0748     blkg_stat_add_aux(&to->empty_time, &from->empty_time);
0749 #endif
0750 }
0751 
0752 /*
0753  * Transfer @cfqg's stats to its parent's aux counts so that the ancestors'
0754  * recursive stats can still account for the amount used by this cfqg after
0755  * it's gone.
0756  */
0757 static void cfqg_stats_xfer_dead(struct cfq_group *cfqg)
0758 {
0759     struct cfq_group *parent = cfqg_parent(cfqg);
0760 
0761     lockdep_assert_held(cfqg_to_blkg(cfqg)->q->queue_lock);
0762 
0763     if (unlikely(!parent))
0764         return;
0765 
0766     cfqg_stats_add_aux(&parent->stats, &cfqg->stats);
0767     cfqg_stats_reset(&cfqg->stats);
0768 }
0769 
0770 #else   /* CONFIG_CFQ_GROUP_IOSCHED */
0771 
0772 static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg) { return NULL; }
0773 static inline bool cfqg_is_descendant(struct cfq_group *cfqg,
0774                       struct cfq_group *ancestor)
0775 {
0776     return true;
0777 }
0778 static inline void cfqg_get(struct cfq_group *cfqg) { }
0779 static inline void cfqg_put(struct cfq_group *cfqg) { }
0780 
0781 #define cfq_log_cfqq(cfqd, cfqq, fmt, args...)  \
0782     blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c " fmt, (cfqq)->pid, \
0783             cfq_cfqq_sync((cfqq)) ? 'S' : 'A',      \
0784             cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
0785                 ##args)
0786 #define cfq_log_cfqg(cfqd, cfqg, fmt, args...)      do {} while (0)
0787 
0788 static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
0789             struct cfq_group *curr_cfqg, unsigned int op) { }
0790 static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
0791             uint64_t time, unsigned long unaccounted_time) { }
0792 static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg,
0793             unsigned int op) { }
0794 static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg,
0795             unsigned int op) { }
0796 static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
0797             uint64_t start_time, uint64_t io_start_time,
0798             unsigned int op) { }
0799 
0800 #endif  /* CONFIG_CFQ_GROUP_IOSCHED */
0801 
0802 #define cfq_log(cfqd, fmt, args...) \
0803     blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)
0804 
0805 /* Traverses through cfq group service trees */
0806 #define for_each_cfqg_st(cfqg, i, j, st) \
0807     for (i = 0; i <= IDLE_WORKLOAD; i++) \
0808         for (j = 0, st = i < IDLE_WORKLOAD ? &cfqg->service_trees[i][j]\
0809             : &cfqg->service_tree_idle; \
0810             (i < IDLE_WORKLOAD && j <= SYNC_WORKLOAD) || \
0811             (i == IDLE_WORKLOAD && j == 0); \
0812             j++, st = i < IDLE_WORKLOAD ? \
0813             &cfqg->service_trees[i][j]: NULL) \
0814 
0815 static inline bool cfq_io_thinktime_big(struct cfq_data *cfqd,
0816     struct cfq_ttime *ttime, bool group_idle)
0817 {
0818     u64 slice;
0819     if (!sample_valid(ttime->ttime_samples))
0820         return false;
0821     if (group_idle)
0822         slice = cfqd->cfq_group_idle;
0823     else
0824         slice = cfqd->cfq_slice_idle;
0825     return ttime->ttime_mean > slice;
0826 }
0827 
0828 static inline bool iops_mode(struct cfq_data *cfqd)
0829 {
0830     /*
0831      * If we are not idling on queues and it is a NCQ drive, parallel
0832      * execution of requests is on and measuring time is not possible
0833      * in most of the cases until and unless we drive shallower queue
0834      * depths and that becomes a performance bottleneck. In such cases
0835      * switch to start providing fairness in terms of number of IOs.
0836      */
0837     if (!cfqd->cfq_slice_idle && cfqd->hw_tag)
0838         return true;
0839     else
0840         return false;
0841 }
0842 
0843 static inline enum wl_class_t cfqq_class(struct cfq_queue *cfqq)
0844 {
0845     if (cfq_class_idle(cfqq))
0846         return IDLE_WORKLOAD;
0847     if (cfq_class_rt(cfqq))
0848         return RT_WORKLOAD;
0849     return BE_WORKLOAD;
0850 }
0851 
0852 
0853 static enum wl_type_t cfqq_type(struct cfq_queue *cfqq)
0854 {
0855     if (!cfq_cfqq_sync(cfqq))
0856         return ASYNC_WORKLOAD;
0857     if (!cfq_cfqq_idle_window(cfqq))
0858         return SYNC_NOIDLE_WORKLOAD;
0859     return SYNC_WORKLOAD;
0860 }
0861 
0862 static inline int cfq_group_busy_queues_wl(enum wl_class_t wl_class,
0863                     struct cfq_data *cfqd,
0864                     struct cfq_group *cfqg)
0865 {
0866     if (wl_class == IDLE_WORKLOAD)
0867         return cfqg->service_tree_idle.count;
0868 
0869     return cfqg->service_trees[wl_class][ASYNC_WORKLOAD].count +
0870         cfqg->service_trees[wl_class][SYNC_NOIDLE_WORKLOAD].count +
0871         cfqg->service_trees[wl_class][SYNC_WORKLOAD].count;
0872 }
0873 
0874 static inline int cfqg_busy_async_queues(struct cfq_data *cfqd,
0875                     struct cfq_group *cfqg)
0876 {
0877     return cfqg->service_trees[RT_WORKLOAD][ASYNC_WORKLOAD].count +
0878         cfqg->service_trees[BE_WORKLOAD][ASYNC_WORKLOAD].count;
0879 }
0880 
0881 static void cfq_dispatch_insert(struct request_queue *, struct request *);
0882 static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, bool is_sync,
0883                        struct cfq_io_cq *cic, struct bio *bio);
0884 
0885 static inline struct cfq_io_cq *icq_to_cic(struct io_cq *icq)
0886 {
0887     /* cic->icq is the first member, %NULL will convert to %NULL */
0888     return container_of(icq, struct cfq_io_cq, icq);
0889 }
0890 
0891 static inline struct cfq_io_cq *cfq_cic_lookup(struct cfq_data *cfqd,
0892                            struct io_context *ioc)
0893 {
0894     if (ioc)
0895         return icq_to_cic(ioc_lookup_icq(ioc, cfqd->queue));
0896     return NULL;
0897 }
0898 
0899 static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_cq *cic, bool is_sync)
0900 {
0901     return cic->cfqq[is_sync];
0902 }
0903 
0904 static inline void cic_set_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq,
0905                 bool is_sync)
0906 {
0907     cic->cfqq[is_sync] = cfqq;
0908 }
0909 
0910 static inline struct cfq_data *cic_to_cfqd(struct cfq_io_cq *cic)
0911 {
0912     return cic->icq.q->elevator->elevator_data;
0913 }
0914 
0915 /*
0916  * scheduler run of queue, if there are requests pending and no one in the
0917  * driver that will restart queueing
0918  */
0919 static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
0920 {
0921     if (cfqd->busy_queues) {
0922         cfq_log(cfqd, "schedule dispatch");
0923         kblockd_schedule_work(&cfqd->unplug_work);
0924     }
0925 }
0926 
0927 /*
0928  * Scale schedule slice based on io priority. Use the sync time slice only
0929  * if a queue is marked sync and has sync io queued. A sync queue with async
0930  * io only, should not get full sync slice length.
0931  */
0932 static inline u64 cfq_prio_slice(struct cfq_data *cfqd, bool sync,
0933                  unsigned short prio)
0934 {
0935     u64 base_slice = cfqd->cfq_slice[sync];
0936     u64 slice = div_u64(base_slice, CFQ_SLICE_SCALE);
0937 
0938     WARN_ON(prio >= IOPRIO_BE_NR);
0939 
0940     return base_slice + (slice * (4 - prio));
0941 }
0942 
0943 static inline u64
0944 cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
0945 {
0946     return cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio);
0947 }
0948 
0949 /**
0950  * cfqg_scale_charge - scale disk time charge according to cfqg weight
0951  * @charge: disk time being charged
0952  * @vfraction: vfraction of the cfqg, fixed point w/ CFQ_SERVICE_SHIFT
0953  *
0954  * Scale @charge according to @vfraction, which is in range (0, 1].  The
0955  * scaling is inversely proportional.
0956  *
0957  * scaled = charge / vfraction
0958  *
0959  * The result is also in fixed point w/ CFQ_SERVICE_SHIFT.
0960  */
0961 static inline u64 cfqg_scale_charge(u64 charge,
0962                     unsigned int vfraction)
0963 {
0964     u64 c = charge << CFQ_SERVICE_SHIFT;    /* make it fixed point */
0965 
0966     /* charge / vfraction */
0967     c <<= CFQ_SERVICE_SHIFT;
0968     return div_u64(c, vfraction);
0969 }
0970 
0971 static inline u64 max_vdisktime(u64 min_vdisktime, u64 vdisktime)
0972 {
0973     s64 delta = (s64)(vdisktime - min_vdisktime);
0974     if (delta > 0)
0975         min_vdisktime = vdisktime;
0976 
0977     return min_vdisktime;
0978 }
0979 
0980 static inline u64 min_vdisktime(u64 min_vdisktime, u64 vdisktime)
0981 {
0982     s64 delta = (s64)(vdisktime - min_vdisktime);
0983     if (delta < 0)
0984         min_vdisktime = vdisktime;
0985 
0986     return min_vdisktime;
0987 }
0988 
0989 static void update_min_vdisktime(struct cfq_rb_root *st)
0990 {
0991     struct cfq_group *cfqg;
0992 
0993     if (st->left) {
0994         cfqg = rb_entry_cfqg(st->left);
0995         st->min_vdisktime = max_vdisktime(st->min_vdisktime,
0996                           cfqg->vdisktime);
0997     }
0998 }
0999 
1000 /*
1001  * get averaged number of queues of RT/BE priority.
1002  * average is updated, with a formula that gives more weight to higher numbers,
1003  * to quickly follows sudden increases and decrease slowly
1004  */
1005 
1006 static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd,
1007                     struct cfq_group *cfqg, bool rt)
1008 {
1009     unsigned min_q, max_q;
1010     unsigned mult  = cfq_hist_divisor - 1;
1011     unsigned round = cfq_hist_divisor / 2;
1012     unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg);
1013 
1014     min_q = min(cfqg->busy_queues_avg[rt], busy);
1015     max_q = max(cfqg->busy_queues_avg[rt], busy);
1016     cfqg->busy_queues_avg[rt] = (mult * max_q + min_q + round) /
1017         cfq_hist_divisor;
1018     return cfqg->busy_queues_avg[rt];
1019 }
1020 
1021 static inline u64
1022 cfq_group_slice(struct cfq_data *cfqd, struct cfq_group *cfqg)
1023 {
1024     return cfqd->cfq_target_latency * cfqg->vfraction >> CFQ_SERVICE_SHIFT;
1025 }
1026 
1027 static inline u64
1028 cfq_scaled_cfqq_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1029 {
1030     u64 slice = cfq_prio_to_slice(cfqd, cfqq);
1031     if (cfqd->cfq_latency) {
1032         /*
1033          * interested queues (we consider only the ones with the same
1034          * priority class in the cfq group)
1035          */
1036         unsigned iq = cfq_group_get_avg_queues(cfqd, cfqq->cfqg,
1037                         cfq_class_rt(cfqq));
1038         u64 sync_slice = cfqd->cfq_slice[1];
1039         u64 expect_latency = sync_slice * iq;
1040         u64 group_slice = cfq_group_slice(cfqd, cfqq->cfqg);
1041 
1042         if (expect_latency > group_slice) {
1043             u64 base_low_slice = 2 * cfqd->cfq_slice_idle;
1044             u64 low_slice;
1045 
1046             /* scale low_slice according to IO priority
1047              * and sync vs async */
1048             low_slice = div64_u64(base_low_slice*slice, sync_slice);
1049             low_slice = min(slice, low_slice);
1050             /* the adapted slice value is scaled to fit all iqs
1051              * into the target latency */
1052             slice = div64_u64(slice*group_slice, expect_latency);
1053             slice = max(slice, low_slice);
1054         }
1055     }
1056     return slice;
1057 }
1058 
1059 static inline void
1060 cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1061 {
1062     u64 slice = cfq_scaled_cfqq_slice(cfqd, cfqq);
1063     u64 now = ktime_get_ns();
1064 
1065     cfqq->slice_start = now;
1066     cfqq->slice_end = now + slice;
1067     cfqq->allocated_slice = slice;
1068     cfq_log_cfqq(cfqd, cfqq, "set_slice=%llu", cfqq->slice_end - now);
1069 }
1070 
1071 /*
1072  * We need to wrap this check in cfq_cfqq_slice_new(), since ->slice_end
1073  * isn't valid until the first request from the dispatch is activated
1074  * and the slice time set.
1075  */
1076 static inline bool cfq_slice_used(struct cfq_queue *cfqq)
1077 {
1078     if (cfq_cfqq_slice_new(cfqq))
1079         return false;
1080     if (ktime_get_ns() < cfqq->slice_end)
1081         return false;
1082 
1083     return true;
1084 }
1085 
1086 /*
1087  * Lifted from AS - choose which of rq1 and rq2 that is best served now.
1088  * We choose the request that is closest to the head right now. Distance
1089  * behind the head is penalized and only allowed to a certain extent.
1090  */
1091 static struct request *
1092 cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2, sector_t last)
1093 {
1094     sector_t s1, s2, d1 = 0, d2 = 0;
1095     unsigned long back_max;
1096 #define CFQ_RQ1_WRAP    0x01 /* request 1 wraps */
1097 #define CFQ_RQ2_WRAP    0x02 /* request 2 wraps */
1098     unsigned wrap = 0; /* bit mask: requests behind the disk head? */
1099 
1100     if (rq1 == NULL || rq1 == rq2)
1101         return rq2;
1102     if (rq2 == NULL)
1103         return rq1;
1104 
1105     if (rq_is_sync(rq1) != rq_is_sync(rq2))
1106         return rq_is_sync(rq1) ? rq1 : rq2;
1107 
1108     if ((rq1->cmd_flags ^ rq2->cmd_flags) & REQ_PRIO)
1109         return rq1->cmd_flags & REQ_PRIO ? rq1 : rq2;
1110 
1111     s1 = blk_rq_pos(rq1);
1112     s2 = blk_rq_pos(rq2);
1113 
1114     /*
1115      * by definition, 1KiB is 2 sectors
1116      */
1117     back_max = cfqd->cfq_back_max * 2;
1118 
1119     /*
1120      * Strict one way elevator _except_ in the case where we allow
1121      * short backward seeks which are biased as twice the cost of a
1122      * similar forward seek.
1123      */
1124     if (s1 >= last)
1125         d1 = s1 - last;
1126     else if (s1 + back_max >= last)
1127         d1 = (last - s1) * cfqd->cfq_back_penalty;
1128     else
1129         wrap |= CFQ_RQ1_WRAP;
1130 
1131     if (s2 >= last)
1132         d2 = s2 - last;
1133     else if (s2 + back_max >= last)
1134         d2 = (last - s2) * cfqd->cfq_back_penalty;
1135     else
1136         wrap |= CFQ_RQ2_WRAP;
1137 
1138     /* Found required data */
1139 
1140     /*
1141      * By doing switch() on the bit mask "wrap" we avoid having to
1142      * check two variables for all permutations: --> faster!
1143      */
1144     switch (wrap) {
1145     case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
1146         if (d1 < d2)
1147             return rq1;
1148         else if (d2 < d1)
1149             return rq2;
1150         else {
1151             if (s1 >= s2)
1152                 return rq1;
1153             else
1154                 return rq2;
1155         }
1156 
1157     case CFQ_RQ2_WRAP:
1158         return rq1;
1159     case CFQ_RQ1_WRAP:
1160         return rq2;
1161     case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */
1162     default:
1163         /*
1164          * Since both rqs are wrapped,
1165          * start with the one that's further behind head
1166          * (--> only *one* back seek required),
1167          * since back seek takes more time than forward.
1168          */
1169         if (s1 <= s2)
1170             return rq1;
1171         else
1172             return rq2;
1173     }
1174 }
1175 
1176 /*
1177  * The below is leftmost cache rbtree addon
1178  */
1179 static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root)
1180 {
1181     /* Service tree is empty */
1182     if (!root->count)
1183         return NULL;
1184 
1185     if (!root->left)
1186         root->left = rb_first(&root->rb);
1187 
1188     if (root->left)
1189         return rb_entry(root->left, struct cfq_queue, rb_node);
1190 
1191     return NULL;
1192 }
1193 
1194 static struct cfq_group *cfq_rb_first_group(struct cfq_rb_root *root)
1195 {
1196     if (!root->left)
1197         root->left = rb_first(&root->rb);
1198 
1199     if (root->left)
1200         return rb_entry_cfqg(root->left);
1201 
1202     return NULL;
1203 }
1204 
1205 static void rb_erase_init(struct rb_node *n, struct rb_root *root)
1206 {
1207     rb_erase(n, root);
1208     RB_CLEAR_NODE(n);
1209 }
1210 
1211 static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root)
1212 {
1213     if (root->left == n)
1214         root->left = NULL;
1215     rb_erase_init(n, &root->rb);
1216     --root->count;
1217 }
1218 
1219 /*
1220  * would be nice to take fifo expire time into account as well
1221  */
1222 static struct request *
1223 cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1224           struct request *last)
1225 {
1226     struct rb_node *rbnext = rb_next(&last->rb_node);
1227     struct rb_node *rbprev = rb_prev(&last->rb_node);
1228     struct request *next = NULL, *prev = NULL;
1229 
1230     BUG_ON(RB_EMPTY_NODE(&last->rb_node));
1231 
1232     if (rbprev)
1233         prev = rb_entry_rq(rbprev);
1234 
1235     if (rbnext)
1236         next = rb_entry_rq(rbnext);
1237     else {
1238         rbnext = rb_first(&cfqq->sort_list);
1239         if (rbnext && rbnext != &last->rb_node)
1240             next = rb_entry_rq(rbnext);
1241     }
1242 
1243     return cfq_choose_req(cfqd, next, prev, blk_rq_pos(last));
1244 }
1245 
1246 static u64 cfq_slice_offset(struct cfq_data *cfqd,
1247                 struct cfq_queue *cfqq)
1248 {
1249     /*
1250      * just an approximation, should be ok.
1251      */
1252     return (cfqq->cfqg->nr_cfqq - 1) * (cfq_prio_slice(cfqd, 1, 0) -
1253                cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio));
1254 }
1255 
1256 static inline s64
1257 cfqg_key(struct cfq_rb_root *st, struct cfq_group *cfqg)
1258 {
1259     return cfqg->vdisktime - st->min_vdisktime;
1260 }
1261 
1262 static void
1263 __cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
1264 {
1265     struct rb_node **node = &st->rb.rb_node;
1266     struct rb_node *parent = NULL;
1267     struct cfq_group *__cfqg;
1268     s64 key = cfqg_key(st, cfqg);
1269     int left = 1;
1270 
1271     while (*node != NULL) {
1272         parent = *node;
1273         __cfqg = rb_entry_cfqg(parent);
1274 
1275         if (key < cfqg_key(st, __cfqg))
1276             node = &parent->rb_left;
1277         else {
1278             node = &parent->rb_right;
1279             left = 0;
1280         }
1281     }
1282 
1283     if (left)
1284         st->left = &cfqg->rb_node;
1285 
1286     rb_link_node(&cfqg->rb_node, parent, node);
1287     rb_insert_color(&cfqg->rb_node, &st->rb);
1288 }
1289 
1290 /*
1291  * This has to be called only on activation of cfqg
1292  */
1293 static void
1294 cfq_update_group_weight(struct cfq_group *cfqg)
1295 {
1296     if (cfqg->new_weight) {
1297         cfqg->weight = cfqg->new_weight;
1298         cfqg->new_weight = 0;
1299     }
1300 }
1301 
1302 static void
1303 cfq_update_group_leaf_weight(struct cfq_group *cfqg)
1304 {
1305     BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
1306 
1307     if (cfqg->new_leaf_weight) {
1308         cfqg->leaf_weight = cfqg->new_leaf_weight;
1309         cfqg->new_leaf_weight = 0;
1310     }
1311 }
1312 
1313 static void
1314 cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
1315 {
1316     unsigned int vfr = 1 << CFQ_SERVICE_SHIFT;  /* start with 1 */
1317     struct cfq_group *pos = cfqg;
1318     struct cfq_group *parent;
1319     bool propagate;
1320 
1321     /* add to the service tree */
1322     BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
1323 
1324     /*
1325      * Update leaf_weight.  We cannot update weight at this point
1326      * because cfqg might already have been activated and is
1327      * contributing its current weight to the parent's child_weight.
1328      */
1329     cfq_update_group_leaf_weight(cfqg);
1330     __cfq_group_service_tree_add(st, cfqg);
1331 
1332     /*
1333      * Activate @cfqg and calculate the portion of vfraction @cfqg is
1334      * entitled to.  vfraction is calculated by walking the tree
1335      * towards the root calculating the fraction it has at each level.
1336      * The compounded ratio is how much vfraction @cfqg owns.
1337      *
1338      * Start with the proportion tasks in this cfqg has against active
1339      * children cfqgs - its leaf_weight against children_weight.
1340      */
1341     propagate = !pos->nr_active++;
1342     pos->children_weight += pos->leaf_weight;
1343     vfr = vfr * pos->leaf_weight / pos->children_weight;
1344 
1345     /*
1346      * Compound ->weight walking up the tree.  Both activation and
1347      * vfraction calculation are done in the same loop.  Propagation
1348      * stops once an already activated node is met.  vfraction
1349      * calculation should always continue to the root.
1350      */
1351     while ((parent = cfqg_parent(pos))) {
1352         if (propagate) {
1353             cfq_update_group_weight(pos);
1354             propagate = !parent->nr_active++;
1355             parent->children_weight += pos->weight;
1356         }
1357         vfr = vfr * pos->weight / parent->children_weight;
1358         pos = parent;
1359     }
1360 
1361     cfqg->vfraction = max_t(unsigned, vfr, 1);
1362 }
1363 
1364 static void
1365 cfq_group_notify_queue_add(struct cfq_data *cfqd, struct cfq_group *cfqg)
1366 {
1367     struct cfq_rb_root *st = &cfqd->grp_service_tree;
1368     struct cfq_group *__cfqg;
1369     struct rb_node *n;
1370 
1371     cfqg->nr_cfqq++;
1372     if (!RB_EMPTY_NODE(&cfqg->rb_node))
1373         return;
1374 
1375     /*
1376      * Currently put the group at the end. Later implement something
1377      * so that groups get lesser vtime based on their weights, so that
1378      * if group does not loose all if it was not continuously backlogged.
1379      */
1380     n = rb_last(&st->rb);
1381     if (n) {
1382         __cfqg = rb_entry_cfqg(n);
1383         cfqg->vdisktime = __cfqg->vdisktime + CFQ_IDLE_DELAY;
1384     } else
1385         cfqg->vdisktime = st->min_vdisktime;
1386     cfq_group_service_tree_add(st, cfqg);
1387 }
1388 
1389 static void
1390 cfq_group_service_tree_del(struct cfq_rb_root *st, struct cfq_group *cfqg)
1391 {
1392     struct cfq_group *pos = cfqg;
1393     bool propagate;
1394 
1395     /*
1396      * Undo activation from cfq_group_service_tree_add().  Deactivate
1397      * @cfqg and propagate deactivation upwards.
1398      */
1399     propagate = !--pos->nr_active;
1400     pos->children_weight -= pos->leaf_weight;
1401 
1402     while (propagate) {
1403         struct cfq_group *parent = cfqg_parent(pos);
1404 
1405         /* @pos has 0 nr_active at this point */
1406         WARN_ON_ONCE(pos->children_weight);
1407         pos->vfraction = 0;
1408 
1409         if (!parent)
1410             break;
1411 
1412         propagate = !--parent->nr_active;
1413         parent->children_weight -= pos->weight;
1414         pos = parent;
1415     }
1416 
1417     /* remove from the service tree */
1418     if (!RB_EMPTY_NODE(&cfqg->rb_node))
1419         cfq_rb_erase(&cfqg->rb_node, st);
1420 }
1421 
1422 static void
1423 cfq_group_notify_queue_del(struct cfq_data *cfqd, struct cfq_group *cfqg)
1424 {
1425     struct cfq_rb_root *st = &cfqd->grp_service_tree;
1426 
1427     BUG_ON(cfqg->nr_cfqq < 1);
1428     cfqg->nr_cfqq--;
1429 
1430     /* If there are other cfq queues under this group, don't delete it */
1431     if (cfqg->nr_cfqq)
1432         return;
1433 
1434     cfq_log_cfqg(cfqd, cfqg, "del_from_rr group");
1435     cfq_group_service_tree_del(st, cfqg);
1436     cfqg->saved_wl_slice = 0;
1437     cfqg_stats_update_dequeue(cfqg);
1438 }
1439 
1440 static inline u64 cfq_cfqq_slice_usage(struct cfq_queue *cfqq,
1441                        u64 *unaccounted_time)
1442 {
1443     u64 slice_used;
1444     u64 now = ktime_get_ns();
1445 
1446     /*
1447      * Queue got expired before even a single request completed or
1448      * got expired immediately after first request completion.
1449      */
1450     if (!cfqq->slice_start || cfqq->slice_start == now) {
1451         /*
1452          * Also charge the seek time incurred to the group, otherwise
1453          * if there are mutiple queues in the group, each can dispatch
1454          * a single request on seeky media and cause lots of seek time
1455          * and group will never know it.
1456          */
1457         slice_used = max_t(u64, (now - cfqq->dispatch_start),
1458                     jiffies_to_nsecs(1));
1459     } else {
1460         slice_used = now - cfqq->slice_start;
1461         if (slice_used > cfqq->allocated_slice) {
1462             *unaccounted_time = slice_used - cfqq->allocated_slice;
1463             slice_used = cfqq->allocated_slice;
1464         }
1465         if (cfqq->slice_start > cfqq->dispatch_start)
1466             *unaccounted_time += cfqq->slice_start -
1467                     cfqq->dispatch_start;
1468     }
1469 
1470     return slice_used;
1471 }
1472 
1473 static void cfq_group_served(struct cfq_data *cfqd, struct cfq_group *cfqg,
1474                 struct cfq_queue *cfqq)
1475 {
1476     struct cfq_rb_root *st = &cfqd->grp_service_tree;
1477     u64 used_sl, charge, unaccounted_sl = 0;
1478     int nr_sync = cfqg->nr_cfqq - cfqg_busy_async_queues(cfqd, cfqg)
1479             - cfqg->service_tree_idle.count;
1480     unsigned int vfr;
1481     u64 now = ktime_get_ns();
1482 
1483     BUG_ON(nr_sync < 0);
1484     used_sl = charge = cfq_cfqq_slice_usage(cfqq, &unaccounted_sl);
1485 
1486     if (iops_mode(cfqd))
1487         charge = cfqq->slice_dispatch;
1488     else if (!cfq_cfqq_sync(cfqq) && !nr_sync)
1489         charge = cfqq->allocated_slice;
1490 
1491     /*
1492      * Can't update vdisktime while on service tree and cfqg->vfraction
1493      * is valid only while on it.  Cache vfr, leave the service tree,
1494      * update vdisktime and go back on.  The re-addition to the tree
1495      * will also update the weights as necessary.
1496      */
1497     vfr = cfqg->vfraction;
1498     cfq_group_service_tree_del(st, cfqg);
1499     cfqg->vdisktime += cfqg_scale_charge(charge, vfr);
1500     cfq_group_service_tree_add(st, cfqg);
1501 
1502     /* This group is being expired. Save the context */
1503     if (cfqd->workload_expires > now) {
1504         cfqg->saved_wl_slice = cfqd->workload_expires - now;
1505         cfqg->saved_wl_type = cfqd->serving_wl_type;
1506         cfqg->saved_wl_class = cfqd->serving_wl_class;
1507     } else
1508         cfqg->saved_wl_slice = 0;
1509 
1510     cfq_log_cfqg(cfqd, cfqg, "served: vt=%llu min_vt=%llu", cfqg->vdisktime,
1511                     st->min_vdisktime);
1512     cfq_log_cfqq(cfqq->cfqd, cfqq,
1513              "sl_used=%llu disp=%llu charge=%llu iops=%u sect=%lu",
1514              used_sl, cfqq->slice_dispatch, charge,
1515              iops_mode(cfqd), cfqq->nr_sectors);
1516     cfqg_stats_update_timeslice_used(cfqg, used_sl, unaccounted_sl);
1517     cfqg_stats_set_start_empty_time(cfqg);
1518 }
1519 
1520 /**
1521  * cfq_init_cfqg_base - initialize base part of a cfq_group
1522  * @cfqg: cfq_group to initialize
1523  *
1524  * Initialize the base part which is used whether %CONFIG_CFQ_GROUP_IOSCHED
1525  * is enabled or not.
1526  */
1527 static void cfq_init_cfqg_base(struct cfq_group *cfqg)
1528 {
1529     struct cfq_rb_root *st;
1530     int i, j;
1531 
1532     for_each_cfqg_st(cfqg, i, j, st)
1533         *st = CFQ_RB_ROOT;
1534     RB_CLEAR_NODE(&cfqg->rb_node);
1535 
1536     cfqg->ttime.last_end_request = ktime_get_ns();
1537 }
1538 
1539 #ifdef CONFIG_CFQ_GROUP_IOSCHED
1540 static int __cfq_set_weight(struct cgroup_subsys_state *css, u64 val,
1541                 bool on_dfl, bool reset_dev, bool is_leaf_weight);
1542 
1543 static void cfqg_stats_exit(struct cfqg_stats *stats)
1544 {
1545     blkg_rwstat_exit(&stats->merged);
1546     blkg_rwstat_exit(&stats->service_time);
1547     blkg_rwstat_exit(&stats->wait_time);
1548     blkg_rwstat_exit(&stats->queued);
1549     blkg_stat_exit(&stats->time);
1550 #ifdef CONFIG_DEBUG_BLK_CGROUP
1551     blkg_stat_exit(&stats->unaccounted_time);
1552     blkg_stat_exit(&stats->avg_queue_size_sum);
1553     blkg_stat_exit(&stats->avg_queue_size_samples);
1554     blkg_stat_exit(&stats->dequeue);
1555     blkg_stat_exit(&stats->group_wait_time);
1556     blkg_stat_exit(&stats->idle_time);
1557     blkg_stat_exit(&stats->empty_time);
1558 #endif
1559 }
1560 
1561 static int cfqg_stats_init(struct cfqg_stats *stats, gfp_t gfp)
1562 {
1563     if (blkg_rwstat_init(&stats->merged, gfp) ||
1564         blkg_rwstat_init(&stats->service_time, gfp) ||
1565         blkg_rwstat_init(&stats->wait_time, gfp) ||
1566         blkg_rwstat_init(&stats->queued, gfp) ||
1567         blkg_stat_init(&stats->time, gfp))
1568         goto err;
1569 
1570 #ifdef CONFIG_DEBUG_BLK_CGROUP
1571     if (blkg_stat_init(&stats->unaccounted_time, gfp) ||
1572         blkg_stat_init(&stats->avg_queue_size_sum, gfp) ||
1573         blkg_stat_init(&stats->avg_queue_size_samples, gfp) ||
1574         blkg_stat_init(&stats->dequeue, gfp) ||
1575         blkg_stat_init(&stats->group_wait_time, gfp) ||
1576         blkg_stat_init(&stats->idle_time, gfp) ||
1577         blkg_stat_init(&stats->empty_time, gfp))
1578         goto err;
1579 #endif
1580     return 0;
1581 err:
1582     cfqg_stats_exit(stats);
1583     return -ENOMEM;
1584 }
1585 
1586 static struct blkcg_policy_data *cfq_cpd_alloc(gfp_t gfp)
1587 {
1588     struct cfq_group_data *cgd;
1589 
1590     cgd = kzalloc(sizeof(*cgd), gfp);
1591     if (!cgd)
1592         return NULL;
1593     return &cgd->cpd;
1594 }
1595 
1596 static void cfq_cpd_init(struct blkcg_policy_data *cpd)
1597 {
1598     struct cfq_group_data *cgd = cpd_to_cfqgd(cpd);
1599     unsigned int weight = cgroup_subsys_on_dfl(io_cgrp_subsys) ?
1600                   CGROUP_WEIGHT_DFL : CFQ_WEIGHT_LEGACY_DFL;
1601 
1602     if (cpd_to_blkcg(cpd) == &blkcg_root)
1603         weight *= 2;
1604 
1605     cgd->weight = weight;
1606     cgd->leaf_weight = weight;
1607 }
1608 
1609 static void cfq_cpd_free(struct blkcg_policy_data *cpd)
1610 {
1611     kfree(cpd_to_cfqgd(cpd));
1612 }
1613 
1614 static void cfq_cpd_bind(struct blkcg_policy_data *cpd)
1615 {
1616     struct blkcg *blkcg = cpd_to_blkcg(cpd);
1617     bool on_dfl = cgroup_subsys_on_dfl(io_cgrp_subsys);
1618     unsigned int weight = on_dfl ? CGROUP_WEIGHT_DFL : CFQ_WEIGHT_LEGACY_DFL;
1619 
1620     if (blkcg == &blkcg_root)
1621         weight *= 2;
1622 
1623     WARN_ON_ONCE(__cfq_set_weight(&blkcg->css, weight, on_dfl, true, false));
1624     WARN_ON_ONCE(__cfq_set_weight(&blkcg->css, weight, on_dfl, true, true));
1625 }
1626 
1627 static struct blkg_policy_data *cfq_pd_alloc(gfp_t gfp, int node)
1628 {
1629     struct cfq_group *cfqg;
1630 
1631     cfqg = kzalloc_node(sizeof(*cfqg), gfp, node);
1632     if (!cfqg)
1633         return NULL;
1634 
1635     cfq_init_cfqg_base(cfqg);
1636     if (cfqg_stats_init(&cfqg->stats, gfp)) {
1637         kfree(cfqg);
1638         return NULL;
1639     }
1640 
1641     return &cfqg->pd;
1642 }
1643 
1644 static void cfq_pd_init(struct blkg_policy_data *pd)
1645 {
1646     struct cfq_group *cfqg = pd_to_cfqg(pd);
1647     struct cfq_group_data *cgd = blkcg_to_cfqgd(pd->blkg->blkcg);
1648 
1649     cfqg->weight = cgd->weight;
1650     cfqg->leaf_weight = cgd->leaf_weight;
1651 }
1652 
1653 static void cfq_pd_offline(struct blkg_policy_data *pd)
1654 {
1655     struct cfq_group *cfqg = pd_to_cfqg(pd);
1656     int i;
1657 
1658     for (i = 0; i < IOPRIO_BE_NR; i++) {
1659         if (cfqg->async_cfqq[0][i])
1660             cfq_put_queue(cfqg->async_cfqq[0][i]);
1661         if (cfqg->async_cfqq[1][i])
1662             cfq_put_queue(cfqg->async_cfqq[1][i]);
1663     }
1664 
1665     if (cfqg->async_idle_cfqq)
1666         cfq_put_queue(cfqg->async_idle_cfqq);
1667 
1668     /*
1669      * @blkg is going offline and will be ignored by
1670      * blkg_[rw]stat_recursive_sum().  Transfer stats to the parent so
1671      * that they don't get lost.  If IOs complete after this point, the
1672      * stats for them will be lost.  Oh well...
1673      */
1674     cfqg_stats_xfer_dead(cfqg);
1675 }
1676 
1677 static void cfq_pd_free(struct blkg_policy_data *pd)
1678 {
1679     struct cfq_group *cfqg = pd_to_cfqg(pd);
1680 
1681     cfqg_stats_exit(&cfqg->stats);
1682     return kfree(cfqg);
1683 }
1684 
1685 static void cfq_pd_reset_stats(struct blkg_policy_data *pd)
1686 {
1687     struct cfq_group *cfqg = pd_to_cfqg(pd);
1688 
1689     cfqg_stats_reset(&cfqg->stats);
1690 }
1691 
1692 static struct cfq_group *cfq_lookup_cfqg(struct cfq_data *cfqd,
1693                      struct blkcg *blkcg)
1694 {
1695     struct blkcg_gq *blkg;
1696 
1697     blkg = blkg_lookup(blkcg, cfqd->queue);
1698     if (likely(blkg))
1699         return blkg_to_cfqg(blkg);
1700     return NULL;
1701 }
1702 
1703 static void cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg)
1704 {
1705     cfqq->cfqg = cfqg;
1706     /* cfqq reference on cfqg */
1707     cfqg_get(cfqg);
1708 }
1709 
1710 static u64 cfqg_prfill_weight_device(struct seq_file *sf,
1711                      struct blkg_policy_data *pd, int off)
1712 {
1713     struct cfq_group *cfqg = pd_to_cfqg(pd);
1714 
1715     if (!cfqg->dev_weight)
1716         return 0;
1717     return __blkg_prfill_u64(sf, pd, cfqg->dev_weight);
1718 }
1719 
1720 static int cfqg_print_weight_device(struct seq_file *sf, void *v)
1721 {
1722     blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1723               cfqg_prfill_weight_device, &blkcg_policy_cfq,
1724               0, false);
1725     return 0;
1726 }
1727 
1728 static u64 cfqg_prfill_leaf_weight_device(struct seq_file *sf,
1729                       struct blkg_policy_data *pd, int off)
1730 {
1731     struct cfq_group *cfqg = pd_to_cfqg(pd);
1732 
1733     if (!cfqg->dev_leaf_weight)
1734         return 0;
1735     return __blkg_prfill_u64(sf, pd, cfqg->dev_leaf_weight);
1736 }
1737 
1738 static int cfqg_print_leaf_weight_device(struct seq_file *sf, void *v)
1739 {
1740     blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1741               cfqg_prfill_leaf_weight_device, &blkcg_policy_cfq,
1742               0, false);
1743     return 0;
1744 }
1745 
1746 static int cfq_print_weight(struct seq_file *sf, void *v)
1747 {
1748     struct blkcg *blkcg = css_to_blkcg(seq_css(sf));
1749     struct cfq_group_data *cgd = blkcg_to_cfqgd(blkcg);
1750     unsigned int val = 0;
1751 
1752     if (cgd)
1753         val = cgd->weight;
1754 
1755     seq_printf(sf, "%u\n", val);
1756     return 0;
1757 }
1758 
1759 static int cfq_print_leaf_weight(struct seq_file *sf, void *v)
1760 {
1761     struct blkcg *blkcg = css_to_blkcg(seq_css(sf));
1762     struct cfq_group_data *cgd = blkcg_to_cfqgd(blkcg);
1763     unsigned int val = 0;
1764 
1765     if (cgd)
1766         val = cgd->leaf_weight;
1767 
1768     seq_printf(sf, "%u\n", val);
1769     return 0;
1770 }
1771 
1772 static ssize_t __cfqg_set_weight_device(struct kernfs_open_file *of,
1773                     char *buf, size_t nbytes, loff_t off,
1774                     bool on_dfl, bool is_leaf_weight)
1775 {
1776     unsigned int min = on_dfl ? CGROUP_WEIGHT_MIN : CFQ_WEIGHT_LEGACY_MIN;
1777     unsigned int max = on_dfl ? CGROUP_WEIGHT_MAX : CFQ_WEIGHT_LEGACY_MAX;
1778     struct blkcg *blkcg = css_to_blkcg(of_css(of));
1779     struct blkg_conf_ctx ctx;
1780     struct cfq_group *cfqg;
1781     struct cfq_group_data *cfqgd;
1782     int ret;
1783     u64 v;
1784 
1785     ret = blkg_conf_prep(blkcg, &blkcg_policy_cfq, buf, &ctx);
1786     if (ret)
1787         return ret;
1788 
1789     if (sscanf(ctx.body, "%llu", &v) == 1) {
1790         /* require "default" on dfl */
1791         ret = -ERANGE;
1792         if (!v && on_dfl)
1793             goto out_finish;
1794     } else if (!strcmp(strim(ctx.body), "default")) {
1795         v = 0;
1796     } else {
1797         ret = -EINVAL;
1798         goto out_finish;
1799     }
1800 
1801     cfqg = blkg_to_cfqg(ctx.blkg);
1802     cfqgd = blkcg_to_cfqgd(blkcg);
1803 
1804     ret = -ERANGE;
1805     if (!v || (v >= min && v <= max)) {
1806         if (!is_leaf_weight) {
1807             cfqg->dev_weight = v;
1808             cfqg->new_weight = v ?: cfqgd->weight;
1809         } else {
1810             cfqg->dev_leaf_weight = v;
1811             cfqg->new_leaf_weight = v ?: cfqgd->leaf_weight;
1812         }
1813         ret = 0;
1814     }
1815 out_finish:
1816     blkg_conf_finish(&ctx);
1817     return ret ?: nbytes;
1818 }
1819 
1820 static ssize_t cfqg_set_weight_device(struct kernfs_open_file *of,
1821                       char *buf, size_t nbytes, loff_t off)
1822 {
1823     return __cfqg_set_weight_device(of, buf, nbytes, off, false, false);
1824 }
1825 
1826 static ssize_t cfqg_set_leaf_weight_device(struct kernfs_open_file *of,
1827                        char *buf, size_t nbytes, loff_t off)
1828 {
1829     return __cfqg_set_weight_device(of, buf, nbytes, off, false, true);
1830 }
1831 
1832 static int __cfq_set_weight(struct cgroup_subsys_state *css, u64 val,
1833                 bool on_dfl, bool reset_dev, bool is_leaf_weight)
1834 {
1835     unsigned int min = on_dfl ? CGROUP_WEIGHT_MIN : CFQ_WEIGHT_LEGACY_MIN;
1836     unsigned int max = on_dfl ? CGROUP_WEIGHT_MAX : CFQ_WEIGHT_LEGACY_MAX;
1837     struct blkcg *blkcg = css_to_blkcg(css);
1838     struct blkcg_gq *blkg;
1839     struct cfq_group_data *cfqgd;
1840     int ret = 0;
1841 
1842     if (val < min || val > max)
1843         return -ERANGE;
1844 
1845     spin_lock_irq(&blkcg->lock);
1846     cfqgd = blkcg_to_cfqgd(blkcg);
1847     if (!cfqgd) {
1848         ret = -EINVAL;
1849         goto out;
1850     }
1851 
1852     if (!is_leaf_weight)
1853         cfqgd->weight = val;
1854     else
1855         cfqgd->leaf_weight = val;
1856 
1857     hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
1858         struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1859 
1860         if (!cfqg)
1861             continue;
1862 
1863         if (!is_leaf_weight) {
1864             if (reset_dev)
1865                 cfqg->dev_weight = 0;
1866             if (!cfqg->dev_weight)
1867                 cfqg->new_weight = cfqgd->weight;
1868         } else {
1869             if (reset_dev)
1870                 cfqg->dev_leaf_weight = 0;
1871             if (!cfqg->dev_leaf_weight)
1872                 cfqg->new_leaf_weight = cfqgd->leaf_weight;
1873         }
1874     }
1875 
1876 out:
1877     spin_unlock_irq(&blkcg->lock);
1878     return ret;
1879 }
1880 
1881 static int cfq_set_weight(struct cgroup_subsys_state *css, struct cftype *cft,
1882               u64 val)
1883 {
1884     return __cfq_set_weight(css, val, false, false, false);
1885 }
1886 
1887 static int cfq_set_leaf_weight(struct cgroup_subsys_state *css,
1888                    struct cftype *cft, u64 val)
1889 {
1890     return __cfq_set_weight(css, val, false, false, true);
1891 }
1892 
1893 static int cfqg_print_stat(struct seq_file *sf, void *v)
1894 {
1895     blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), blkg_prfill_stat,
1896               &blkcg_policy_cfq, seq_cft(sf)->private, false);
1897     return 0;
1898 }
1899 
1900 static int cfqg_print_rwstat(struct seq_file *sf, void *v)
1901 {
1902     blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), blkg_prfill_rwstat,
1903               &blkcg_policy_cfq, seq_cft(sf)->private, true);
1904     return 0;
1905 }
1906 
1907 static u64 cfqg_prfill_stat_recursive(struct seq_file *sf,
1908                       struct blkg_policy_data *pd, int off)
1909 {
1910     u64 sum = blkg_stat_recursive_sum(pd_to_blkg(pd),
1911                       &blkcg_policy_cfq, off);
1912     return __blkg_prfill_u64(sf, pd, sum);
1913 }
1914 
1915 static u64 cfqg_prfill_rwstat_recursive(struct seq_file *sf,
1916                     struct blkg_policy_data *pd, int off)
1917 {
1918     struct blkg_rwstat sum = blkg_rwstat_recursive_sum(pd_to_blkg(pd),
1919                             &blkcg_policy_cfq, off);
1920     return __blkg_prfill_rwstat(sf, pd, &sum);
1921 }
1922 
1923 static int cfqg_print_stat_recursive(struct seq_file *sf, void *v)
1924 {
1925     blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1926               cfqg_prfill_stat_recursive, &blkcg_policy_cfq,
1927               seq_cft(sf)->private, false);
1928     return 0;
1929 }
1930 
1931 static int cfqg_print_rwstat_recursive(struct seq_file *sf, void *v)
1932 {
1933     blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1934               cfqg_prfill_rwstat_recursive, &blkcg_policy_cfq,
1935               seq_cft(sf)->private, true);
1936     return 0;
1937 }
1938 
1939 static u64 cfqg_prfill_sectors(struct seq_file *sf, struct blkg_policy_data *pd,
1940                    int off)
1941 {
1942     u64 sum = blkg_rwstat_total(&pd->blkg->stat_bytes);
1943 
1944     return __blkg_prfill_u64(sf, pd, sum >> 9);
1945 }
1946 
1947 static int cfqg_print_stat_sectors(struct seq_file *sf, void *v)
1948 {
1949     blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1950               cfqg_prfill_sectors, &blkcg_policy_cfq, 0, false);
1951     return 0;
1952 }
1953 
1954 static u64 cfqg_prfill_sectors_recursive(struct seq_file *sf,
1955                      struct blkg_policy_data *pd, int off)
1956 {
1957     struct blkg_rwstat tmp = blkg_rwstat_recursive_sum(pd->blkg, NULL,
1958                     offsetof(struct blkcg_gq, stat_bytes));
1959     u64 sum = atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_READ]) +
1960         atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_WRITE]);
1961 
1962     return __blkg_prfill_u64(sf, pd, sum >> 9);
1963 }
1964 
1965 static int cfqg_print_stat_sectors_recursive(struct seq_file *sf, void *v)
1966 {
1967     blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1968               cfqg_prfill_sectors_recursive, &blkcg_policy_cfq, 0,
1969               false);
1970     return 0;
1971 }
1972 
1973 #ifdef CONFIG_DEBUG_BLK_CGROUP
1974 static u64 cfqg_prfill_avg_queue_size(struct seq_file *sf,
1975                       struct blkg_policy_data *pd, int off)
1976 {
1977     struct cfq_group *cfqg = pd_to_cfqg(pd);
1978     u64 samples = blkg_stat_read(&cfqg->stats.avg_queue_size_samples);
1979     u64 v = 0;
1980 
1981     if (samples) {
1982         v = blkg_stat_read(&cfqg->stats.avg_queue_size_sum);
1983         v = div64_u64(v, samples);
1984     }
1985     __blkg_prfill_u64(sf, pd, v);
1986     return 0;
1987 }
1988 
1989 /* print avg_queue_size */
1990 static int cfqg_print_avg_queue_size(struct seq_file *sf, void *v)
1991 {
1992     blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1993               cfqg_prfill_avg_queue_size, &blkcg_policy_cfq,
1994               0, false);
1995     return 0;
1996 }
1997 #endif  /* CONFIG_DEBUG_BLK_CGROUP */
1998 
1999 static struct cftype cfq_blkcg_legacy_files[] = {
2000     /* on root, weight is mapped to leaf_weight */
2001     {
2002         .name = "weight_device",
2003         .flags = CFTYPE_ONLY_ON_ROOT,
2004         .seq_show = cfqg_print_leaf_weight_device,
2005         .write = cfqg_set_leaf_weight_device,
2006     },
2007     {
2008         .name = "weight",
2009         .flags = CFTYPE_ONLY_ON_ROOT,
2010         .seq_show = cfq_print_leaf_weight,
2011         .write_u64 = cfq_set_leaf_weight,
2012     },
2013 
2014     /* no such mapping necessary for !roots */
2015     {
2016         .name = "weight_device",
2017         .flags = CFTYPE_NOT_ON_ROOT,
2018         .seq_show = cfqg_print_weight_device,
2019         .write = cfqg_set_weight_device,
2020     },
2021     {
2022         .name = "weight",
2023         .flags = CFTYPE_NOT_ON_ROOT,
2024         .seq_show = cfq_print_weight,
2025         .write_u64 = cfq_set_weight,
2026     },
2027 
2028     {
2029         .name = "leaf_weight_device",
2030         .seq_show = cfqg_print_leaf_weight_device,
2031         .write = cfqg_set_leaf_weight_device,
2032     },
2033     {
2034         .name = "leaf_weight",
2035         .seq_show = cfq_print_leaf_weight,
2036         .write_u64 = cfq_set_leaf_weight,
2037     },
2038 
2039     /* statistics, covers only the tasks in the cfqg */
2040     {
2041         .name = "time",
2042         .private = offsetof(struct cfq_group, stats.time),
2043         .seq_show = cfqg_print_stat,
2044     },
2045     {
2046         .name = "sectors",
2047         .seq_show = cfqg_print_stat_sectors,
2048     },
2049     {
2050         .name = "io_service_bytes",
2051         .private = (unsigned long)&blkcg_policy_cfq,
2052         .seq_show = blkg_print_stat_bytes,
2053     },
2054     {
2055         .name = "io_serviced",
2056         .private = (unsigned long)&blkcg_policy_cfq,
2057         .seq_show = blkg_print_stat_ios,
2058     },
2059     {
2060         .name = "io_service_time",
2061         .private = offsetof(struct cfq_group, stats.service_time),
2062         .seq_show = cfqg_print_rwstat,
2063     },
2064     {
2065         .name = "io_wait_time",
2066         .private = offsetof(struct cfq_group, stats.wait_time),
2067         .seq_show = cfqg_print_rwstat,
2068     },
2069     {
2070         .name = "io_merged",
2071         .private = offsetof(struct cfq_group, stats.merged),
2072         .seq_show = cfqg_print_rwstat,
2073     },
2074     {
2075         .name = "io_queued",
2076         .private = offsetof(struct cfq_group, stats.queued),
2077         .seq_show = cfqg_print_rwstat,
2078     },
2079 
2080     /* the same statictics which cover the cfqg and its descendants */
2081     {
2082         .name = "time_recursive",
2083         .private = offsetof(struct cfq_group, stats.time),
2084         .seq_show = cfqg_print_stat_recursive,
2085     },
2086     {
2087         .name = "sectors_recursive",
2088         .seq_show = cfqg_print_stat_sectors_recursive,
2089     },
2090     {
2091         .name = "io_service_bytes_recursive",
2092         .private = (unsigned long)&blkcg_policy_cfq,
2093         .seq_show = blkg_print_stat_bytes_recursive,
2094     },
2095     {
2096         .name = "io_serviced_recursive",
2097         .private = (unsigned long)&blkcg_policy_cfq,
2098         .seq_show = blkg_print_stat_ios_recursive,
2099     },
2100     {
2101         .name = "io_service_time_recursive",
2102         .private = offsetof(struct cfq_group, stats.service_time),
2103         .seq_show = cfqg_print_rwstat_recursive,
2104     },
2105     {
2106         .name = "io_wait_time_recursive",
2107         .private = offsetof(struct cfq_group, stats.wait_time),
2108         .seq_show = cfqg_print_rwstat_recursive,
2109     },
2110     {
2111         .name = "io_merged_recursive",
2112         .private = offsetof(struct cfq_group, stats.merged),
2113         .seq_show = cfqg_print_rwstat_recursive,
2114     },
2115     {
2116         .name = "io_queued_recursive",
2117         .private = offsetof(struct cfq_group, stats.queued),
2118         .seq_show = cfqg_print_rwstat_recursive,
2119     },
2120 #ifdef CONFIG_DEBUG_BLK_CGROUP
2121     {
2122         .name = "avg_queue_size",
2123         .seq_show = cfqg_print_avg_queue_size,
2124     },
2125     {
2126         .name = "group_wait_time",
2127         .private = offsetof(struct cfq_group, stats.group_wait_time),
2128         .seq_show = cfqg_print_stat,
2129     },
2130     {
2131         .name = "idle_time",
2132         .private = offsetof(struct cfq_group, stats.idle_time),
2133         .seq_show = cfqg_print_stat,
2134     },
2135     {
2136         .name = "empty_time",
2137         .private = offsetof(struct cfq_group, stats.empty_time),
2138         .seq_show = cfqg_print_stat,
2139     },
2140     {
2141         .name = "dequeue",
2142         .private = offsetof(struct cfq_group, stats.dequeue),
2143         .seq_show = cfqg_print_stat,
2144     },
2145     {
2146         .name = "unaccounted_time",
2147         .private = offsetof(struct cfq_group, stats.unaccounted_time),
2148         .seq_show = cfqg_print_stat,
2149     },
2150 #endif  /* CONFIG_DEBUG_BLK_CGROUP */
2151     { } /* terminate */
2152 };
2153 
2154 static int cfq_print_weight_on_dfl(struct seq_file *sf, void *v)
2155 {
2156     struct blkcg *blkcg = css_to_blkcg(seq_css(sf));
2157     struct cfq_group_data *cgd = blkcg_to_cfqgd(blkcg);
2158 
2159     seq_printf(sf, "default %u\n", cgd->weight);
2160     blkcg_print_blkgs(sf, blkcg, cfqg_prfill_weight_device,
2161               &blkcg_policy_cfq, 0, false);
2162     return 0;
2163 }
2164 
2165 static ssize_t cfq_set_weight_on_dfl(struct kernfs_open_file *of,
2166                      char *buf, size_t nbytes, loff_t off)
2167 {
2168     char *endp;
2169     int ret;
2170     u64 v;
2171 
2172     buf = strim(buf);
2173 
2174     /* "WEIGHT" or "default WEIGHT" sets the default weight */
2175     v = simple_strtoull(buf, &endp, 0);
2176     if (*endp == '\0' || sscanf(buf, "default %llu", &v) == 1) {
2177         ret = __cfq_set_weight(of_css(of), v, true, false, false);
2178         return ret ?: nbytes;
2179     }
2180 
2181     /* "MAJ:MIN WEIGHT" */
2182     return __cfqg_set_weight_device(of, buf, nbytes, off, true, false);
2183 }
2184 
2185 static struct cftype cfq_blkcg_files[] = {
2186     {
2187         .name = "weight",
2188         .flags = CFTYPE_NOT_ON_ROOT,
2189         .seq_show = cfq_print_weight_on_dfl,
2190         .write = cfq_set_weight_on_dfl,
2191     },
2192     { } /* terminate */
2193 };
2194 
2195 #else /* GROUP_IOSCHED */
2196 static struct cfq_group *cfq_lookup_cfqg(struct cfq_data *cfqd,
2197                      struct blkcg *blkcg)
2198 {
2199     return cfqd->root_group;
2200 }
2201 
2202 static inline void
2203 cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg) {
2204     cfqq->cfqg = cfqg;
2205 }
2206 
2207 #endif /* GROUP_IOSCHED */
2208 
2209 /*
2210  * The cfqd->service_trees holds all pending cfq_queue's that have
2211  * requests waiting to be processed. It is sorted in the order that
2212  * we will service the queues.
2213  */
2214 static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2215                  bool add_front)
2216 {
2217     struct rb_node **p, *parent;
2218     struct cfq_queue *__cfqq;
2219     u64 rb_key;
2220     struct cfq_rb_root *st;
2221     int left;
2222     int new_cfqq = 1;
2223     u64 now = ktime_get_ns();
2224 
2225     st = st_for(cfqq->cfqg, cfqq_class(cfqq), cfqq_type(cfqq));
2226     if (cfq_class_idle(cfqq)) {
2227         rb_key = CFQ_IDLE_DELAY;
2228         parent = rb_last(&st->rb);
2229         if (parent && parent != &cfqq->rb_node) {
2230             __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2231             rb_key += __cfqq->rb_key;
2232         } else
2233             rb_key += now;
2234     } else if (!add_front) {
2235         /*
2236          * Get our rb key offset. Subtract any residual slice
2237          * value carried from last service. A negative resid
2238          * count indicates slice overrun, and this should position
2239          * the next service time further away in the tree.
2240          */
2241         rb_key = cfq_slice_offset(cfqd, cfqq) + now;
2242         rb_key -= cfqq->slice_resid;
2243         cfqq->slice_resid = 0;
2244     } else {
2245         rb_key = -NSEC_PER_SEC;
2246         __cfqq = cfq_rb_first(st);
2247         rb_key += __cfqq ? __cfqq->rb_key : now;
2248     }
2249 
2250     if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2251         new_cfqq = 0;
2252         /*
2253          * same position, nothing more to do
2254          */
2255         if (rb_key == cfqq->rb_key && cfqq->service_tree == st)
2256             return;
2257 
2258         cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2259         cfqq->service_tree = NULL;
2260     }
2261 
2262     left = 1;
2263     parent = NULL;
2264     cfqq->service_tree = st;
2265     p = &st->rb.rb_node;
2266     while (*p) {
2267         parent = *p;
2268         __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2269 
2270         /*
2271          * sort by key, that represents service time.
2272          */
2273         if (rb_key < __cfqq->rb_key)
2274             p = &parent->rb_left;
2275         else {
2276             p = &parent->rb_right;
2277             left = 0;
2278         }
2279     }
2280 
2281     if (left)
2282         st->left = &cfqq->rb_node;
2283 
2284     cfqq->rb_key = rb_key;
2285     rb_link_node(&cfqq->rb_node, parent, p);
2286     rb_insert_color(&cfqq->rb_node, &st->rb);
2287     st->count++;
2288     if (add_front || !new_cfqq)
2289         return;
2290     cfq_group_notify_queue_add(cfqd, cfqq->cfqg);
2291 }
2292 
2293 static struct cfq_queue *
2294 cfq_prio_tree_lookup(struct cfq_data *cfqd, struct rb_root *root,
2295              sector_t sector, struct rb_node **ret_parent,
2296              struct rb_node ***rb_link)
2297 {
2298     struct rb_node **p, *parent;
2299     struct cfq_queue *cfqq = NULL;
2300 
2301     parent = NULL;
2302     p = &root->rb_node;
2303     while (*p) {
2304         struct rb_node **n;
2305 
2306         parent = *p;
2307         cfqq = rb_entry(parent, struct cfq_queue, p_node);
2308 
2309         /*
2310          * Sort strictly based on sector.  Smallest to the left,
2311          * largest to the right.
2312          */
2313         if (sector > blk_rq_pos(cfqq->next_rq))
2314             n = &(*p)->rb_right;
2315         else if (sector < blk_rq_pos(cfqq->next_rq))
2316             n = &(*p)->rb_left;
2317         else
2318             break;
2319         p = n;
2320         cfqq = NULL;
2321     }
2322 
2323     *ret_parent = parent;
2324     if (rb_link)
2325         *rb_link = p;
2326     return cfqq;
2327 }
2328 
2329 static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2330 {
2331     struct rb_node **p, *parent;
2332     struct cfq_queue *__cfqq;
2333 
2334     if (cfqq->p_root) {
2335         rb_erase(&cfqq->p_node, cfqq->p_root);
2336         cfqq->p_root = NULL;
2337     }
2338 
2339     if (cfq_class_idle(cfqq))
2340         return;
2341     if (!cfqq->next_rq)
2342         return;
2343 
2344     cfqq->p_root = &cfqd->prio_trees[cfqq->org_ioprio];
2345     __cfqq = cfq_prio_tree_lookup(cfqd, cfqq->p_root,
2346                       blk_rq_pos(cfqq->next_rq), &parent, &p);
2347     if (!__cfqq) {
2348         rb_link_node(&cfqq->p_node, parent, p);
2349         rb_insert_color(&cfqq->p_node, cfqq->p_root);
2350     } else
2351         cfqq->p_root = NULL;
2352 }
2353 
2354 /*
2355  * Update cfqq's position in the service tree.
2356  */
2357 static void cfq_resort_rr_list(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2358 {
2359     /*
2360      * Resorting requires the cfqq to be on the RR list already.
2361      */
2362     if (cfq_cfqq_on_rr(cfqq)) {
2363         cfq_service_tree_add(cfqd, cfqq, 0);
2364         cfq_prio_tree_add(cfqd, cfqq);
2365     }
2366 }
2367 
2368 /*
2369  * add to busy list of queues for service, trying to be fair in ordering
2370  * the pending list according to last request service
2371  */
2372 static void cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2373 {
2374     cfq_log_cfqq(cfqd, cfqq, "add_to_rr");
2375     BUG_ON(cfq_cfqq_on_rr(cfqq));
2376     cfq_mark_cfqq_on_rr(cfqq);
2377     cfqd->busy_queues++;
2378     if (cfq_cfqq_sync(cfqq))
2379         cfqd->busy_sync_queues++;
2380 
2381     cfq_resort_rr_list(cfqd, cfqq);
2382 }
2383 
2384 /*
2385  * Called when the cfqq no longer has requests pending, remove it from
2386  * the service tree.
2387  */
2388 static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2389 {
2390     cfq_log_cfqq(cfqd, cfqq, "del_from_rr");
2391     BUG_ON(!cfq_cfqq_on_rr(cfqq));
2392     cfq_clear_cfqq_on_rr(cfqq);
2393 
2394     if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2395         cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2396         cfqq->service_tree = NULL;
2397     }
2398     if (cfqq->p_root) {
2399         rb_erase(&cfqq->p_node, cfqq->p_root);
2400         cfqq->p_root = NULL;
2401     }
2402 
2403     cfq_group_notify_queue_del(cfqd, cfqq->cfqg);
2404     BUG_ON(!cfqd->busy_queues);
2405     cfqd->busy_queues--;
2406     if (cfq_cfqq_sync(cfqq))
2407         cfqd->busy_sync_queues--;
2408 }
2409 
2410 /*
2411  * rb tree support functions
2412  */
2413 static void cfq_del_rq_rb(struct request *rq)
2414 {
2415     struct cfq_queue *cfqq = RQ_CFQQ(rq);
2416     const int sync = rq_is_sync(rq);
2417 
2418     BUG_ON(!cfqq->queued[sync]);
2419     cfqq->queued[sync]--;
2420 
2421     elv_rb_del(&cfqq->sort_list, rq);
2422 
2423     if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list)) {
2424         /*
2425          * Queue will be deleted from service tree when we actually
2426          * expire it later. Right now just remove it from prio tree
2427          * as it is empty.
2428          */
2429         if (cfqq->p_root) {
2430             rb_erase(&cfqq->p_node, cfqq->p_root);
2431             cfqq->p_root = NULL;
2432         }
2433     }
2434 }
2435 
2436 static void cfq_add_rq_rb(struct request *rq)
2437 {
2438     struct cfq_queue *cfqq = RQ_CFQQ(rq);
2439     struct cfq_data *cfqd = cfqq->cfqd;
2440     struct request *prev;
2441 
2442     cfqq->queued[rq_is_sync(rq)]++;
2443 
2444     elv_rb_add(&cfqq->sort_list, rq);
2445 
2446     if (!cfq_cfqq_on_rr(cfqq))
2447         cfq_add_cfqq_rr(cfqd, cfqq);
2448 
2449     /*
2450      * check if this request is a better next-serve candidate
2451      */
2452     prev = cfqq->next_rq;
2453     cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq, cfqd->last_position);
2454 
2455     /*
2456      * adjust priority tree position, if ->next_rq changes
2457      */
2458     if (prev != cfqq->next_rq)
2459         cfq_prio_tree_add(cfqd, cfqq);
2460 
2461     BUG_ON(!cfqq->next_rq);
2462 }
2463 
2464 static void cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
2465 {
2466     elv_rb_del(&cfqq->sort_list, rq);
2467     cfqq->queued[rq_is_sync(rq)]--;
2468     cfqg_stats_update_io_remove(RQ_CFQG(rq), rq->cmd_flags);
2469     cfq_add_rq_rb(rq);
2470     cfqg_stats_update_io_add(RQ_CFQG(rq), cfqq->cfqd->serving_group,
2471                  rq->cmd_flags);
2472 }
2473 
2474 static struct request *
2475 cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio)
2476 {
2477     struct task_struct *tsk = current;
2478     struct cfq_io_cq *cic;
2479     struct cfq_queue *cfqq;
2480 
2481     cic = cfq_cic_lookup(cfqd, tsk->io_context);
2482     if (!cic)
2483         return NULL;
2484 
2485     cfqq = cic_to_cfqq(cic, op_is_sync(bio->bi_opf));
2486     if (cfqq)
2487         return elv_rb_find(&cfqq->sort_list, bio_end_sector(bio));
2488 
2489     return NULL;
2490 }
2491 
2492 static void cfq_activate_request(struct request_queue *q, struct request *rq)
2493 {
2494     struct cfq_data *cfqd = q->elevator->elevator_data;
2495 
2496     cfqd->rq_in_driver++;
2497     cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "activate rq, drv=%d",
2498                         cfqd->rq_in_driver);
2499 
2500     cfqd->last_position = blk_rq_pos(rq) + blk_rq_sectors(rq);
2501 }
2502 
2503 static void cfq_deactivate_request(struct request_queue *q, struct request *rq)
2504 {
2505     struct cfq_data *cfqd = q->elevator->elevator_data;
2506 
2507     WARN_ON(!cfqd->rq_in_driver);
2508     cfqd->rq_in_driver--;
2509     cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "deactivate rq, drv=%d",
2510                         cfqd->rq_in_driver);
2511 }
2512 
2513 static void cfq_remove_request(struct request *rq)
2514 {
2515     struct cfq_queue *cfqq = RQ_CFQQ(rq);
2516 
2517     if (cfqq->next_rq == rq)
2518         cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq);
2519 
2520     list_del_init(&rq->queuelist);
2521     cfq_del_rq_rb(rq);
2522 
2523     cfqq->cfqd->rq_queued--;
2524     cfqg_stats_update_io_remove(RQ_CFQG(rq), rq->cmd_flags);
2525     if (rq->cmd_flags & REQ_PRIO) {
2526         WARN_ON(!cfqq->prio_pending);
2527         cfqq->prio_pending--;
2528     }
2529 }
2530 
2531 static int cfq_merge(struct request_queue *q, struct request **req,
2532              struct bio *bio)
2533 {
2534     struct cfq_data *cfqd = q->elevator->elevator_data;
2535     struct request *__rq;
2536 
2537     __rq = cfq_find_rq_fmerge(cfqd, bio);
2538     if (__rq && elv_bio_merge_ok(__rq, bio)) {
2539         *req = __rq;
2540         return ELEVATOR_FRONT_MERGE;
2541     }
2542 
2543     return ELEVATOR_NO_MERGE;
2544 }
2545 
2546 static void cfq_merged_request(struct request_queue *q, struct request *req,
2547                    int type)
2548 {
2549     if (type == ELEVATOR_FRONT_MERGE) {
2550         struct cfq_queue *cfqq = RQ_CFQQ(req);
2551 
2552         cfq_reposition_rq_rb(cfqq, req);
2553     }
2554 }
2555 
2556 static void cfq_bio_merged(struct request_queue *q, struct request *req,
2557                 struct bio *bio)
2558 {
2559     cfqg_stats_update_io_merged(RQ_CFQG(req), bio->bi_opf);
2560 }
2561 
2562 static void
2563 cfq_merged_requests(struct request_queue *q, struct request *rq,
2564             struct request *next)
2565 {
2566     struct cfq_queue *cfqq = RQ_CFQQ(rq);
2567     struct cfq_data *cfqd = q->elevator->elevator_data;
2568 
2569     /*
2570      * reposition in fifo if next is older than rq
2571      */
2572     if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
2573         next->fifo_time < rq->fifo_time &&
2574         cfqq == RQ_CFQQ(next)) {
2575         list_move(&rq->queuelist, &next->queuelist);
2576         rq->fifo_time = next->fifo_time;
2577     }
2578 
2579     if (cfqq->next_rq == next)
2580         cfqq->next_rq = rq;
2581     cfq_remove_request(next);
2582     cfqg_stats_update_io_merged(RQ_CFQG(rq), next->cmd_flags);
2583 
2584     cfqq = RQ_CFQQ(next);
2585     /*
2586      * all requests of this queue are merged to other queues, delete it
2587      * from the service tree. If it's the active_queue,
2588      * cfq_dispatch_requests() will choose to expire it or do idle
2589      */
2590     if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list) &&
2591         cfqq != cfqd->active_queue)
2592         cfq_del_cfqq_rr(cfqd, cfqq);
2593 }
2594 
2595 static int cfq_allow_bio_merge(struct request_queue *q, struct request *rq,
2596                    struct bio *bio)
2597 {
2598     struct cfq_data *cfqd = q->elevator->elevator_data;
2599     bool is_sync = op_is_sync(bio->bi_opf);
2600     struct cfq_io_cq *cic;
2601     struct cfq_queue *cfqq;
2602 
2603     /*
2604      * Disallow merge of a sync bio into an async request.
2605      */
2606     if (is_sync && !rq_is_sync(rq))
2607         return false;
2608 
2609     /*
2610      * Lookup the cfqq that this bio will be queued with and allow
2611      * merge only if rq is queued there.
2612      */
2613     cic = cfq_cic_lookup(cfqd, current->io_context);
2614     if (!cic)
2615         return false;
2616 
2617     cfqq = cic_to_cfqq(cic, is_sync);
2618     return cfqq == RQ_CFQQ(rq);
2619 }
2620 
2621 static int cfq_allow_rq_merge(struct request_queue *q, struct request *rq,
2622                   struct request *next)
2623 {
2624     return RQ_CFQQ(rq) == RQ_CFQQ(next);
2625 }
2626 
2627 static inline void cfq_del_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2628 {
2629     hrtimer_try_to_cancel(&cfqd->idle_slice_timer);
2630     cfqg_stats_update_idle_time(cfqq->cfqg);
2631 }
2632 
2633 static void __cfq_set_active_queue(struct cfq_data *cfqd,
2634                    struct cfq_queue *cfqq)
2635 {
2636     if (cfqq) {
2637         cfq_log_cfqq(cfqd, cfqq, "set_active wl_class:%d wl_type:%d",
2638                 cfqd->serving_wl_class, cfqd->serving_wl_type);
2639         cfqg_stats_update_avg_queue_size(cfqq->cfqg);
2640         cfqq->slice_start = 0;
2641         cfqq->dispatch_start = ktime_get_ns();
2642         cfqq->allocated_slice = 0;
2643         cfqq->slice_end = 0;
2644         cfqq->slice_dispatch = 0;
2645         cfqq->nr_sectors = 0;
2646 
2647         cfq_clear_cfqq_wait_request(cfqq);
2648         cfq_clear_cfqq_must_dispatch(cfqq);
2649         cfq_clear_cfqq_must_alloc_slice(cfqq);
2650         cfq_clear_cfqq_fifo_expire(cfqq);
2651         cfq_mark_cfqq_slice_new(cfqq);
2652 
2653         cfq_del_timer(cfqd, cfqq);
2654     }
2655 
2656     cfqd->active_queue = cfqq;
2657 }
2658 
2659 /*
2660  * current cfqq expired its slice (or was too idle), select new one
2661  */
2662 static void
2663 __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2664             bool timed_out)
2665 {
2666     cfq_log_cfqq(cfqd, cfqq, "slice expired t=%d", timed_out);
2667 
2668     if (cfq_cfqq_wait_request(cfqq))
2669         cfq_del_timer(cfqd, cfqq);
2670 
2671     cfq_clear_cfqq_wait_request(cfqq);
2672     cfq_clear_cfqq_wait_busy(cfqq);
2673 
2674     /*
2675      * If this cfqq is shared between multiple processes, check to
2676      * make sure that those processes are still issuing I/Os within
2677      * the mean seek distance.  If not, it may be time to break the
2678      * queues apart again.
2679      */
2680     if (cfq_cfqq_coop(cfqq) && CFQQ_SEEKY(cfqq))
2681         cfq_mark_cfqq_split_coop(cfqq);
2682 
2683     /*
2684      * store what was left of this slice, if the queue idled/timed out
2685      */
2686     if (timed_out) {
2687         if (cfq_cfqq_slice_new(cfqq))
2688             cfqq->slice_resid = cfq_scaled_cfqq_slice(cfqd, cfqq);
2689         else
2690             cfqq->slice_resid = cfqq->slice_end - ktime_get_ns();
2691         cfq_log_cfqq(cfqd, cfqq, "resid=%lld", cfqq->slice_resid);
2692     }
2693 
2694     cfq_group_served(cfqd, cfqq->cfqg, cfqq);
2695 
2696     if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list))
2697         cfq_del_cfqq_rr(cfqd, cfqq);
2698 
2699     cfq_resort_rr_list(cfqd, cfqq);
2700 
2701     if (cfqq == cfqd->active_queue)
2702         cfqd->active_queue = NULL;
2703 
2704     if (cfqd->active_cic) {
2705         put_io_context(cfqd->active_cic->icq.ioc);
2706         cfqd->active_cic = NULL;
2707     }
2708 }
2709 
2710 static inline void cfq_slice_expired(struct cfq_data *cfqd, bool timed_out)
2711 {
2712     struct cfq_queue *cfqq = cfqd->active_queue;
2713 
2714     if (cfqq)
2715         __cfq_slice_expired(cfqd, cfqq, timed_out);
2716 }
2717 
2718 /*
2719  * Get next queue for service. Unless we have a queue preemption,
2720  * we'll simply select the first cfqq in the service tree.
2721  */
2722 static struct cfq_queue *cfq_get_next_queue(struct cfq_data *cfqd)
2723 {
2724     struct cfq_rb_root *st = st_for(cfqd->serving_group,
2725             cfqd->serving_wl_class, cfqd->serving_wl_type);
2726 
2727     if (!cfqd->rq_queued)
2728         return NULL;
2729 
2730     /* There is nothing to dispatch */
2731     if (!st)
2732         return NULL;
2733     if (RB_EMPTY_ROOT(&st->rb))
2734         return NULL;
2735     return cfq_rb_first(st);
2736 }
2737 
2738 static struct cfq_queue *cfq_get_next_queue_forced(struct cfq_data *cfqd)
2739 {
2740     struct cfq_group *cfqg;
2741     struct cfq_queue *cfqq;
2742     int i, j;
2743     struct cfq_rb_root *st;
2744 
2745     if (!cfqd->rq_queued)
2746         return NULL;
2747 
2748     cfqg = cfq_get_next_cfqg(cfqd);
2749     if (!cfqg)
2750         return NULL;
2751 
2752     for_each_cfqg_st(cfqg, i, j, st)
2753         if ((cfqq = cfq_rb_first(st)) != NULL)
2754             return cfqq;
2755     return NULL;
2756 }
2757 
2758 /*
2759  * Get and set a new active queue for service.
2760  */
2761 static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd,
2762                           struct cfq_queue *cfqq)
2763 {
2764     if (!cfqq)
2765         cfqq = cfq_get_next_queue(cfqd);
2766 
2767     __cfq_set_active_queue(cfqd, cfqq);
2768     return cfqq;
2769 }
2770 
2771 static inline sector_t cfq_dist_from_last(struct cfq_data *cfqd,
2772                       struct request *rq)
2773 {
2774     if (blk_rq_pos(rq) >= cfqd->last_position)
2775         return blk_rq_pos(rq) - cfqd->last_position;
2776     else
2777         return cfqd->last_position - blk_rq_pos(rq);
2778 }
2779 
2780 static inline int cfq_rq_close(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2781                    struct request *rq)
2782 {
2783     return cfq_dist_from_last(cfqd, rq) <= CFQQ_CLOSE_THR;
2784 }
2785 
2786 static struct cfq_queue *cfqq_close(struct cfq_data *cfqd,
2787                     struct cfq_queue *cur_cfqq)
2788 {
2789     struct rb_root *root = &cfqd->prio_trees[cur_cfqq->org_ioprio];
2790     struct rb_node *parent, *node;
2791     struct cfq_queue *__cfqq;
2792     sector_t sector = cfqd->last_position;
2793 
2794     if (RB_EMPTY_ROOT(root))
2795         return NULL;
2796 
2797     /*
2798      * First, if we find a request starting at the end of the last
2799      * request, choose it.
2800      */
2801     __cfqq = cfq_prio_tree_lookup(cfqd, root, sector, &parent, NULL);
2802     if (__cfqq)
2803         return __cfqq;
2804 
2805     /*
2806      * If the exact sector wasn't found, the parent of the NULL leaf
2807      * will contain the closest sector.
2808      */
2809     __cfqq = rb_entry(parent, struct cfq_queue, p_node);
2810     if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2811         return __cfqq;
2812 
2813     if (blk_rq_pos(__cfqq->next_rq) < sector)
2814         node = rb_next(&__cfqq->p_node);
2815     else
2816         node = rb_prev(&__cfqq->p_node);
2817     if (!node)
2818         return NULL;
2819 
2820     __cfqq = rb_entry(node, struct cfq_queue, p_node);
2821     if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2822         return __cfqq;
2823 
2824     return NULL;
2825 }
2826 
2827 /*
2828  * cfqd - obvious
2829  * cur_cfqq - passed in so that we don't decide that the current queue is
2830  *        closely cooperating with itself.
2831  *
2832  * So, basically we're assuming that that cur_cfqq has dispatched at least
2833  * one request, and that cfqd->last_position reflects a position on the disk
2834  * associated with the I/O issued by cur_cfqq.  I'm not sure this is a valid
2835  * assumption.
2836  */
2837 static struct cfq_queue *cfq_close_cooperator(struct cfq_data *cfqd,
2838                           struct cfq_queue *cur_cfqq)
2839 {
2840     struct cfq_queue *cfqq;
2841 
2842     if (cfq_class_idle(cur_cfqq))
2843         return NULL;
2844     if (!cfq_cfqq_sync(cur_cfqq))
2845         return NULL;
2846     if (CFQQ_SEEKY(cur_cfqq))
2847         return NULL;
2848 
2849     /*
2850      * Don't search priority tree if it's the only queue in the group.
2851      */
2852     if (cur_cfqq->cfqg->nr_cfqq == 1)
2853         return NULL;
2854 
2855     /*
2856      * We should notice if some of the queues are cooperating, eg
2857      * working closely on the same area of the disk. In that case,
2858      * we can group them together and don't waste time idling.
2859      */
2860     cfqq = cfqq_close(cfqd, cur_cfqq);
2861     if (!cfqq)
2862         return NULL;
2863 
2864     /* If new queue belongs to different cfq_group, don't choose it */
2865     if (cur_cfqq->cfqg != cfqq->cfqg)
2866         return NULL;
2867 
2868     /*
2869      * It only makes sense to merge sync queues.
2870      */
2871     if (!cfq_cfqq_sync(cfqq))
2872         return NULL;
2873     if (CFQQ_SEEKY(cfqq))
2874         return NULL;
2875 
2876     /*
2877      * Do not merge queues of different priority classes
2878      */
2879     if (cfq_class_rt(cfqq) != cfq_class_rt(cur_cfqq))
2880         return NULL;
2881 
2882     return cfqq;
2883 }
2884 
2885 /*
2886  * Determine whether we should enforce idle window for this queue.
2887  */
2888 
2889 static bool cfq_should_idle(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2890 {
2891     enum wl_class_t wl_class = cfqq_class(cfqq);
2892     struct cfq_rb_root *st = cfqq->service_tree;
2893 
2894     BUG_ON(!st);
2895     BUG_ON(!st->count);
2896 
2897     if (!cfqd->cfq_slice_idle)
2898         return false;
2899 
2900     /* We never do for idle class queues. */
2901     if (wl_class == IDLE_WORKLOAD)
2902         return false;
2903 
2904     /* We do for queues that were marked with idle window flag. */
2905     if (cfq_cfqq_idle_window(cfqq) &&
2906        !(blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag))
2907         return true;
2908 
2909     /*
2910      * Otherwise, we do only if they are the last ones
2911      * in their service tree.
2912      */
2913     if (st->count == 1 && cfq_cfqq_sync(cfqq) &&
2914        !cfq_io_thinktime_big(cfqd, &st->ttime, false))
2915         return true;
2916     cfq_log_cfqq(cfqd, cfqq, "Not idling. st->count:%d", st->count);
2917     return false;
2918 }
2919 
2920 static void cfq_arm_slice_timer(struct cfq_data *cfqd)
2921 {
2922     struct cfq_queue *cfqq = cfqd->active_queue;
2923     struct cfq_rb_root *st = cfqq->service_tree;
2924     struct cfq_io_cq *cic;
2925     u64 sl, group_idle = 0;
2926     u64 now = ktime_get_ns();
2927 
2928     /*
2929      * SSD device without seek penalty, disable idling. But only do so
2930      * for devices that support queuing, otherwise we still have a problem
2931      * with sync vs async workloads.
2932      */
2933     if (blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag)
2934         return;
2935 
2936     WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list));
2937     WARN_ON(cfq_cfqq_slice_new(cfqq));
2938 
2939     /*
2940      * idle is disabled, either manually or by past process history
2941      */
2942     if (!cfq_should_idle(cfqd, cfqq)) {
2943         /* no queue idling. Check for group idling */
2944         if (cfqd->cfq_group_idle)
2945             group_idle = cfqd->cfq_group_idle;
2946         else
2947             return;
2948     }
2949 
2950     /*
2951      * still active requests from this queue, don't idle
2952      */
2953     if (cfqq->dispatched)
2954         return;
2955 
2956     /*
2957      * task has exited, don't wait
2958      */
2959     cic = cfqd->active_cic;
2960     if (!cic || !atomic_read(&cic->icq.ioc->active_ref))
2961         return;
2962 
2963     /*
2964      * If our average think time is larger than the remaining time
2965      * slice, then don't idle. This avoids overrunning the allotted
2966      * time slice.
2967      */
2968     if (sample_valid(cic->ttime.ttime_samples) &&
2969         (cfqq->slice_end - now < cic->ttime.ttime_mean)) {
2970         cfq_log_cfqq(cfqd, cfqq, "Not idling. think_time:%llu",
2971                  cic->ttime.ttime_mean);
2972         return;
2973     }
2974 
2975     /*
2976      * There are other queues in the group or this is the only group and
2977      * it has too big thinktime, don't do group idle.
2978      */
2979     if (group_idle &&
2980         (cfqq->cfqg->nr_cfqq > 1 ||
2981          cfq_io_thinktime_big(cfqd, &st->ttime, true)))
2982         return;
2983 
2984     cfq_mark_cfqq_wait_request(cfqq);
2985 
2986     if (group_idle)
2987         sl = cfqd->cfq_group_idle;
2988     else
2989         sl = cfqd->cfq_slice_idle;
2990 
2991     hrtimer_start(&cfqd->idle_slice_timer, ns_to_ktime(sl),
2992               HRTIMER_MODE_REL);
2993     cfqg_stats_set_start_idle_time(cfqq->cfqg);
2994     cfq_log_cfqq(cfqd, cfqq, "arm_idle: %llu group_idle: %d", sl,
2995             group_idle ? 1 : 0);
2996 }
2997 
2998 /*
2999  * Move request from internal lists to the request queue dispatch list.
3000  */
3001 static void cfq_dispatch_insert(struct request_queue *q, struct request *rq)
3002 {
3003     struct cfq_data *cfqd = q->elevator->elevator_data;
3004     struct cfq_queue *cfqq = RQ_CFQQ(rq);
3005 
3006     cfq_log_cfqq(cfqd, cfqq, "dispatch_insert");
3007 
3008     cfqq->next_rq = cfq_find_next_rq(cfqd, cfqq, rq);
3009     cfq_remove_request(rq);
3010     cfqq->dispatched++;
3011     (RQ_CFQG(rq))->dispatched++;
3012     elv_dispatch_sort(q, rq);
3013 
3014     cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]++;
3015     cfqq->nr_sectors += blk_rq_sectors(rq);
3016 }
3017 
3018 /*
3019  * return expired entry, or NULL to just start from scratch in rbtree
3020  */
3021 static struct request *cfq_check_fifo(struct cfq_queue *cfqq)
3022 {
3023     struct request *rq = NULL;
3024 
3025     if (cfq_cfqq_fifo_expire(cfqq))
3026         return NULL;
3027 
3028     cfq_mark_cfqq_fifo_expire(cfqq);
3029 
3030     if (list_empty(&cfqq->fifo))
3031         return NULL;
3032 
3033     rq = rq_entry_fifo(cfqq->fifo.next);
3034     if (ktime_get_ns() < rq->fifo_time)
3035         rq = NULL;
3036 
3037     return rq;
3038 }
3039 
3040 static inline int
3041 cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3042 {
3043     const int base_rq = cfqd->cfq_slice_async_rq;
3044 
3045     WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
3046 
3047     return 2 * base_rq * (IOPRIO_BE_NR - cfqq->ioprio);
3048 }
3049 
3050 /*
3051  * Must be called with the queue_lock held.
3052  */
3053 static int cfqq_process_refs(struct cfq_queue *cfqq)
3054 {
3055     int process_refs, io_refs;
3056 
3057     io_refs = cfqq->allocated[READ] + cfqq->allocated[WRITE];
3058     process_refs = cfqq->ref - io_refs;
3059     BUG_ON(process_refs < 0);
3060     return process_refs;
3061 }
3062 
3063 static void cfq_setup_merge(struct cfq_queue *cfqq, struct cfq_queue *new_cfqq)
3064 {
3065     int process_refs, new_process_refs;
3066     struct cfq_queue *__cfqq;
3067 
3068     /*
3069      * If there are no process references on the new_cfqq, then it is
3070      * unsafe to follow the ->new_cfqq chain as other cfqq's in the
3071      * chain may have dropped their last reference (not just their
3072      * last process reference).
3073      */
3074     if (!cfqq_process_refs(new_cfqq))
3075         return;
3076 
3077     /* Avoid a circular list and skip interim queue merges */
3078     while ((__cfqq = new_cfqq->new_cfqq)) {
3079         if (__cfqq == cfqq)
3080             return;
3081         new_cfqq = __cfqq;
3082     }
3083 
3084     process_refs = cfqq_process_refs(cfqq);
3085     new_process_refs = cfqq_process_refs(new_cfqq);
3086     /*
3087      * If the process for the cfqq has gone away, there is no
3088      * sense in merging the queues.
3089      */
3090     if (process_refs == 0 || new_process_refs == 0)
3091         return;
3092 
3093     /*
3094      * Merge in the direction of the lesser amount of work.
3095      */
3096     if (new_process_refs >= process_refs) {
3097         cfqq->new_cfqq = new_cfqq;
3098         new_cfqq->ref += process_refs;
3099     } else {
3100         new_cfqq->new_cfqq = cfqq;
3101         cfqq->ref += new_process_refs;
3102     }
3103 }
3104 
3105 static enum wl_type_t cfq_choose_wl_type(struct cfq_data *cfqd,
3106             struct cfq_group *cfqg, enum wl_class_t wl_class)
3107 {
3108     struct cfq_queue *queue;
3109     int i;
3110     bool key_valid = false;
3111     u64 lowest_key = 0;
3112     enum wl_type_t cur_best = SYNC_NOIDLE_WORKLOAD;
3113 
3114     for (i = 0; i <= SYNC_WORKLOAD; ++i) {
3115         /* select the one with lowest rb_key */
3116         queue = cfq_rb_first(st_for(cfqg, wl_class, i));
3117         if (queue &&
3118             (!key_valid || queue->rb_key < lowest_key)) {
3119             lowest_key = queue->rb_key;
3120             cur_best = i;
3121             key_valid = true;
3122         }
3123     }
3124 
3125     return cur_best;
3126 }
3127 
3128 static void
3129 choose_wl_class_and_type(struct cfq_data *cfqd, struct cfq_group *cfqg)
3130 {
3131     u64 slice;
3132     unsigned count;
3133     struct cfq_rb_root *st;
3134     u64 group_slice;
3135     enum wl_class_t original_class = cfqd->serving_wl_class;
3136     u64 now = ktime_get_ns();
3137 
3138     /* Choose next priority. RT > BE > IDLE */
3139     if (cfq_group_busy_queues_wl(RT_WORKLOAD, cfqd, cfqg))
3140         cfqd->serving_wl_class = RT_WORKLOAD;
3141     else if (cfq_group_busy_queues_wl(BE_WORKLOAD, cfqd, cfqg))
3142         cfqd->serving_wl_class = BE_WORKLOAD;
3143     else {
3144         cfqd->serving_wl_class = IDLE_WORKLOAD;
3145         cfqd->workload_expires = now + jiffies_to_nsecs(1);
3146         return;
3147     }
3148 
3149     if (original_class != cfqd->serving_wl_class)
3150         goto new_workload;
3151 
3152     /*
3153      * For RT and BE, we have to choose also the type
3154      * (SYNC, SYNC_NOIDLE, ASYNC), and to compute a workload
3155      * expiration time
3156      */
3157     st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
3158     count = st->count;
3159 
3160     /*
3161      * check workload expiration, and that we still have other queues ready
3162      */
3163     if (count && !(now > cfqd->workload_expires))
3164         return;
3165 
3166 new_workload:
3167     /* otherwise select new workload type */
3168     cfqd->serving_wl_type = cfq_choose_wl_type(cfqd, cfqg,
3169                     cfqd->serving_wl_class);
3170     st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
3171     count = st->count;
3172 
3173     /*
3174      * the workload slice is computed as a fraction of target latency
3175      * proportional to the number of queues in that workload, over
3176      * all the queues in the same priority class
3177      */
3178     group_slice = cfq_group_slice(cfqd, cfqg);
3179 
3180     slice = div_u64(group_slice * count,
3181         max_t(unsigned, cfqg->busy_queues_avg[cfqd->serving_wl_class],
3182               cfq_group_busy_queues_wl(cfqd->serving_wl_class, cfqd,
3183                     cfqg)));
3184 
3185     if (cfqd->serving_wl_type == ASYNC_WORKLOAD) {
3186         u64 tmp;
3187 
3188         /*
3189          * Async queues are currently system wide. Just taking
3190          * proportion of queues with-in same group will lead to higher
3191          * async ratio system wide as generally root group is going
3192          * to have higher weight. A more accurate thing would be to
3193          * calculate system wide asnc/sync ratio.
3194          */
3195         tmp = cfqd->cfq_target_latency *
3196             cfqg_busy_async_queues(cfqd, cfqg);
3197         tmp = div_u64(tmp, cfqd->busy_queues);
3198         slice = min_t(u64, slice, tmp);
3199 
3200         /* async workload slice is scaled down according to
3201          * the sync/async slice ratio. */
3202         slice = div64_u64(slice*cfqd->cfq_slice[0], cfqd->cfq_slice[1]);
3203     } else
3204         /* sync workload slice is at least 2 * cfq_slice_idle */
3205         slice = max(slice, 2 * cfqd->cfq_slice_idle);
3206 
3207     slice = max_t(u64, slice, CFQ_MIN_TT);
3208     cfq_log(cfqd, "workload slice:%llu", slice);
3209     cfqd->workload_expires = now + slice;
3210 }
3211 
3212 static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd)
3213 {
3214     struct cfq_rb_root *st = &cfqd->grp_service_tree;
3215     struct cfq_group *cfqg;
3216 
3217     if (RB_EMPTY_ROOT(&st->rb))
3218         return NULL;
3219     cfqg = cfq_rb_first_group(st);
3220     update_min_vdisktime(st);
3221     return cfqg;
3222 }
3223 
3224 static void cfq_choose_cfqg(struct cfq_data *cfqd)
3225 {
3226     struct cfq_group *cfqg = cfq_get_next_cfqg(cfqd);
3227     u64 now = ktime_get_ns();
3228 
3229     cfqd->serving_group = cfqg;
3230 
3231     /* Restore the workload type data */
3232     if (cfqg->saved_wl_slice) {
3233         cfqd->workload_expires = now + cfqg->saved_wl_slice;
3234         cfqd->serving_wl_type = cfqg->saved_wl_type;
3235         cfqd->serving_wl_class = cfqg->saved_wl_class;
3236     } else
3237         cfqd->workload_expires = now - 1;
3238 
3239     choose_wl_class_and_type(cfqd, cfqg);
3240 }
3241 
3242 /*
3243  * Select a queue for service. If we have a current active queue,
3244  * check whether to continue servicing it, or retrieve and set a new one.
3245  */
3246 static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
3247 {
3248     struct cfq_queue *cfqq, *new_cfqq = NULL;
3249     u64 now = ktime_get_ns();
3250 
3251     cfqq = cfqd->active_queue;
3252     if (!cfqq)
3253         goto new_queue;
3254 
3255     if (!cfqd->rq_queued)
3256         return NULL;
3257 
3258     /*
3259      * We were waiting for group to get backlogged. Expire the queue
3260      */
3261     if (cfq_cfqq_wait_busy(cfqq) && !RB_EMPTY_ROOT(&cfqq->sort_list))
3262         goto expire;
3263 
3264     /*
3265      * The active queue has run out of time, expire it and select new.
3266      */
3267     if (cfq_slice_used(cfqq) && !cfq_cfqq_must_dispatch(cfqq)) {
3268         /*
3269          * If slice had not expired at the completion of last request
3270          * we might not have turned on wait_busy flag. Don't expire
3271          * the queue yet. Allow the group to get backlogged.
3272          *
3273          * The very fact that we have used the slice, that means we
3274          * have been idling all along on this queue and it should be
3275          * ok to wait for this request to complete.
3276          */
3277         if (cfqq->cfqg->nr_cfqq == 1 && RB_EMPTY_ROOT(&cfqq->sort_list)
3278             && cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3279             cfqq = NULL;
3280             goto keep_queue;
3281         } else
3282             goto check_group_idle;
3283     }
3284 
3285     /*
3286      * The active queue has requests and isn't expired, allow it to
3287      * dispatch.
3288      */
3289     if (!RB_EMPTY_ROOT(&cfqq->sort_list))
3290         goto keep_queue;
3291 
3292     /*
3293      * If another queue has a request waiting within our mean seek
3294      * distance, let it run.  The expire code will check for close
3295      * cooperators and put the close queue at the front of the service
3296      * tree.  If possible, merge the expiring queue with the new cfqq.
3297      */
3298     new_cfqq = cfq_close_cooperator(cfqd, cfqq);
3299     if (new_cfqq) {
3300         if (!cfqq->new_cfqq)
3301             cfq_setup_merge(cfqq, new_cfqq);
3302         goto expire;
3303     }
3304 
3305     /*
3306      * No requests pending. If the active queue still has requests in
3307      * flight or is idling for a new request, allow either of these
3308      * conditions to happen (or time out) before selecting a new queue.
3309      */
3310     if (hrtimer_active(&cfqd->idle_slice_timer)) {
3311         cfqq = NULL;
3312         goto keep_queue;
3313     }
3314 
3315     /*
3316      * This is a deep seek queue, but the device is much faster than
3317      * the queue can deliver, don't idle
3318      **/
3319     if (CFQQ_SEEKY(cfqq) && cfq_cfqq_idle_window(cfqq) &&
3320         (cfq_cfqq_slice_new(cfqq) ||
3321         (cfqq->slice_end - now > now - cfqq->slice_start))) {
3322         cfq_clear_cfqq_deep(cfqq);
3323         cfq_clear_cfqq_idle_window(cfqq);
3324     }
3325 
3326     if (cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3327         cfqq = NULL;
3328         goto keep_queue;
3329     }
3330 
3331     /*
3332      * If group idle is enabled and there are requests dispatched from
3333      * this group, wait for requests to complete.
3334      */
3335 check_group_idle:
3336     if (cfqd->cfq_group_idle && cfqq->cfqg->nr_cfqq == 1 &&
3337         cfqq->cfqg->dispatched &&
3338         !cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true)) {
3339         cfqq = NULL;
3340         goto keep_queue;
3341     }
3342 
3343 expire:
3344     cfq_slice_expired(cfqd, 0);
3345 new_queue:
3346     /*
3347      * Current queue expired. Check if we have to switch to a new
3348      * service tree
3349      */
3350     if (!new_cfqq)
3351         cfq_choose_cfqg(cfqd);
3352 
3353     cfqq = cfq_set_active_queue(cfqd, new_cfqq);
3354 keep_queue:
3355     return cfqq;
3356 }
3357 
3358 static int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq)
3359 {
3360     int dispatched = 0;
3361 
3362     while (cfqq->next_rq) {
3363         cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq);
3364         dispatched++;
3365     }
3366 
3367     BUG_ON(!list_empty(&cfqq->fifo));
3368 
3369     /* By default cfqq is not expired if it is empty. Do it explicitly */
3370     __cfq_slice_expired(cfqq->cfqd, cfqq, 0);
3371     return dispatched;
3372 }
3373 
3374 /*
3375  * Drain our current requests. Used for barriers and when switching
3376  * io schedulers on-the-fly.
3377  */
3378 static int cfq_forced_dispatch(struct cfq_data *cfqd)
3379 {
3380     struct cfq_queue *cfqq;
3381     int dispatched = 0;
3382 
3383     /* Expire the timeslice of the current active queue first */
3384     cfq_slice_expired(cfqd, 0);
3385     while ((cfqq = cfq_get_next_queue_forced(cfqd)) != NULL) {
3386         __cfq_set_active_queue(cfqd, cfqq);
3387         dispatched += __cfq_forced_dispatch_cfqq(cfqq);
3388     }
3389 
3390     BUG_ON(cfqd->busy_queues);
3391 
3392     cfq_log(cfqd, "forced_dispatch=%d", dispatched);
3393     return dispatched;
3394 }
3395 
3396 static inline bool cfq_slice_used_soon(struct cfq_data *cfqd,
3397     struct cfq_queue *cfqq)
3398 {
3399     u64 now = ktime_get_ns();
3400 
3401     /* the queue hasn't finished any request, can't estimate */
3402     if (cfq_cfqq_slice_new(cfqq))
3403         return true;
3404     if (now + cfqd->cfq_slice_idle * cfqq->dispatched > cfqq->slice_end)
3405         return true;
3406 
3407     return false;
3408 }
3409 
3410 static bool cfq_may_dispatch(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3411 {
3412     unsigned int max_dispatch;
3413 
3414     if (cfq_cfqq_must_dispatch(cfqq))
3415         return true;
3416 
3417     /*
3418      * Drain async requests before we start sync IO
3419      */
3420     if (cfq_should_idle(cfqd, cfqq) && cfqd->rq_in_flight[BLK_RW_ASYNC])
3421         return false;
3422 
3423     /*
3424      * If this is an async queue and we have sync IO in flight, let it wait
3425      */
3426     if (cfqd->rq_in_flight[BLK_RW_SYNC] && !cfq_cfqq_sync(cfqq))
3427         return false;
3428 
3429     max_dispatch = max_t(unsigned int, cfqd->cfq_quantum / 2, 1);
3430     if (cfq_class_idle(cfqq))
3431         max_dispatch = 1;
3432 
3433     /*
3434      * Does this cfqq already have too much IO in flight?
3435      */
3436     if (cfqq->dispatched >= max_dispatch) {
3437         bool promote_sync = false;
3438         /*
3439          * idle queue must always only have a single IO in flight
3440          */
3441         if (cfq_class_idle(cfqq))
3442             return false;
3443 
3444         /*
3445          * If there is only one sync queue
3446          * we can ignore async queue here and give the sync
3447          * queue no dispatch limit. The reason is a sync queue can
3448          * preempt async queue, limiting the sync queue doesn't make
3449          * sense. This is useful for aiostress test.
3450          */
3451         if (cfq_cfqq_sync(cfqq) && cfqd->busy_sync_queues == 1)
3452             promote_sync = true;
3453 
3454         /*
3455          * We have other queues, don't allow more IO from this one
3456          */
3457         if (cfqd->busy_queues > 1 && cfq_slice_used_soon(cfqd, cfqq) &&
3458                 !promote_sync)
3459             return false;
3460 
3461         /*
3462          * Sole queue user, no limit
3463          */
3464         if (cfqd->busy_queues == 1 || promote_sync)
3465             max_dispatch = -1;
3466         else
3467             /*
3468              * Normally we start throttling cfqq when cfq_quantum/2
3469              * requests have been dispatched. But we can drive
3470              * deeper queue depths at the beginning of slice
3471              * subjected to upper limit of cfq_quantum.
3472              * */
3473             max_dispatch = cfqd->cfq_quantum;
3474     }
3475 
3476     /*
3477      * Async queues must wait a bit before being allowed dispatch.
3478      * We also ramp up the dispatch depth gradually for async IO,
3479      * based on the last sync IO we serviced
3480      */
3481     if (!cfq_cfqq_sync(cfqq) && cfqd->cfq_latency) {
3482         u64 last_sync = ktime_get_ns() - cfqd->last_delayed_sync;
3483         unsigned int depth;
3484 
3485         depth = div64_u64(last_sync, cfqd->cfq_slice[1]);
3486         if (!depth && !cfqq->dispatched)
3487             depth = 1;
3488         if (depth < max_dispatch)
3489             max_dispatch = depth;
3490     }
3491 
3492     /*
3493      * If we're below the current max, allow a dispatch
3494      */
3495     return cfqq->dispatched < max_dispatch;
3496 }
3497 
3498 /*
3499  * Dispatch a request from cfqq, moving them to the request queue
3500  * dispatch list.
3501  */
3502 static bool cfq_dispatch_request(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3503 {
3504     struct request *rq;
3505 
3506     BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list));
3507 
3508     rq = cfq_check_fifo(cfqq);
3509     if (rq)
3510         cfq_mark_cfqq_must_dispatch(cfqq);
3511 
3512     if (!cfq_may_dispatch(cfqd, cfqq))
3513         return false;
3514 
3515     /*
3516      * follow expired path, else get first next available
3517      */
3518     if (!rq)
3519         rq = cfqq->next_rq;
3520     else
3521         cfq_log_cfqq(cfqq->cfqd, cfqq, "fifo=%p", rq);
3522 
3523     /*
3524      * insert request into driver dispatch list
3525      */
3526     cfq_dispatch_insert(cfqd->queue, rq);
3527 
3528     if (!cfqd->active_cic) {
3529         struct cfq_io_cq *cic = RQ_CIC(rq);
3530 
3531         atomic_long_inc(&cic->icq.ioc->refcount);
3532         cfqd->active_cic = cic;
3533     }
3534 
3535     return true;
3536 }
3537 
3538 /*
3539  * Find the cfqq that we need to service and move a request from that to the
3540  * dispatch list
3541  */
3542 static int cfq_dispatch_requests(struct request_queue *q, int force)
3543 {
3544     struct cfq_data *cfqd = q->elevator->elevator_data;
3545     struct cfq_queue *cfqq;
3546 
3547     if (!cfqd->busy_queues)
3548         return 0;
3549 
3550     if (unlikely(force))
3551         return cfq_forced_dispatch(cfqd);
3552 
3553     cfqq = cfq_select_queue(cfqd);
3554     if (!cfqq)
3555         return 0;
3556 
3557     /*
3558      * Dispatch a request from this cfqq, if it is allowed
3559      */
3560     if (!cfq_dispatch_request(cfqd, cfqq))
3561         return 0;
3562 
3563     cfqq->slice_dispatch++;
3564     cfq_clear_cfqq_must_dispatch(cfqq);
3565 
3566     /*
3567      * expire an async queue immediately if it has used up its slice. idle
3568      * queue always expire after 1 dispatch round.
3569      */
3570     if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) &&
3571         cfqq->slice_dispatch >= cfq_prio_to_maxrq(cfqd, cfqq)) ||
3572         cfq_class_idle(cfqq))) {
3573         cfqq->slice_end = ktime_get_ns() + 1;
3574         cfq_slice_expired(cfqd, 0);
3575     }
3576 
3577     cfq_log_cfqq(cfqd, cfqq, "dispatched a request");
3578     return 1;
3579 }
3580 
3581 /*
3582  * task holds one reference to the queue, dropped when task exits. each rq
3583  * in-flight on this queue also holds a reference, dropped when rq is freed.
3584  *
3585  * Each cfq queue took a reference on the parent group. Drop it now.
3586  * queue lock must be held here.
3587  */
3588 static void cfq_put_queue(struct cfq_queue *cfqq)
3589 {
3590     struct cfq_data *cfqd = cfqq->cfqd;
3591     struct cfq_group *cfqg;
3592 
3593     BUG_ON(cfqq->ref <= 0);
3594 
3595     cfqq->ref--;
3596     if (cfqq->ref)
3597         return;
3598 
3599     cfq_log_cfqq(cfqd, cfqq, "put_queue");
3600     BUG_ON(rb_first(&cfqq->sort_list));
3601     BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
3602     cfqg = cfqq->cfqg;
3603 
3604     if (unlikely(cfqd->active_queue == cfqq)) {
3605         __cfq_slice_expired(cfqd, cfqq, 0);
3606         cfq_schedule_dispatch(cfqd);
3607     }
3608 
3609     BUG_ON(cfq_cfqq_on_rr(cfqq));
3610     kmem_cache_free(cfq_pool, cfqq);
3611     cfqg_put(cfqg);
3612 }
3613 
3614 static void cfq_put_cooperator(struct cfq_queue *cfqq)
3615 {
3616     struct cfq_queue *__cfqq, *next;
3617 
3618     /*
3619      * If this queue was scheduled to merge with another queue, be
3620      * sure to drop the reference taken on that queue (and others in
3621      * the merge chain).  See cfq_setup_merge and cfq_merge_cfqqs.
3622      */
3623     __cfqq = cfqq->new_cfqq;
3624     while (__cfqq) {
3625         if (__cfqq == cfqq) {
3626             WARN(1, "cfqq->new_cfqq loop detected\n");
3627             break;
3628         }
3629         next = __cfqq->new_cfqq;
3630         cfq_put_queue(__cfqq);
3631         __cfqq = next;
3632     }
3633 }
3634 
3635 static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3636 {
3637     if (unlikely(cfqq == cfqd->active_queue)) {
3638         __cfq_slice_expired(cfqd, cfqq, 0);
3639         cfq_schedule_dispatch(cfqd);
3640     }
3641 
3642     cfq_put_cooperator(cfqq);
3643 
3644     cfq_put_queue(cfqq);
3645 }
3646 
3647 static void cfq_init_icq(struct io_cq *icq)
3648 {
3649     struct cfq_io_cq *cic = icq_to_cic(icq);
3650 
3651     cic->ttime.last_end_request = ktime_get_ns();
3652 }
3653 
3654 static void cfq_exit_icq(struct io_cq *icq)
3655 {
3656     struct cfq_io_cq *cic = icq_to_cic(icq);
3657     struct cfq_data *cfqd = cic_to_cfqd(cic);
3658 
3659     if (cic_to_cfqq(cic, false)) {
3660         cfq_exit_cfqq(cfqd, cic_to_cfqq(cic, false));
3661         cic_set_cfqq(cic, NULL, false);
3662     }
3663 
3664     if (cic_to_cfqq(cic, true)) {
3665         cfq_exit_cfqq(cfqd, cic_to_cfqq(cic, true));
3666         cic_set_cfqq(cic, NULL, true);
3667     }
3668 }
3669 
3670 static void cfq_init_prio_data(struct cfq_queue *cfqq, struct cfq_io_cq *cic)
3671 {
3672     struct task_struct *tsk = current;
3673     int ioprio_class;
3674 
3675     if (!cfq_cfqq_prio_changed(cfqq))
3676         return;
3677 
3678     ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3679     switch (ioprio_class) {
3680     default:
3681         printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
3682     case IOPRIO_CLASS_NONE:
3683         /*
3684          * no prio set, inherit CPU scheduling settings
3685          */
3686         cfqq->ioprio = task_nice_ioprio(tsk);
3687         cfqq->ioprio_class = task_nice_ioclass(tsk);
3688         break;
3689     case IOPRIO_CLASS_RT:
3690         cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3691         cfqq->ioprio_class = IOPRIO_CLASS_RT;
3692         break;
3693     case IOPRIO_CLASS_BE:
3694         cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3695         cfqq->ioprio_class = IOPRIO_CLASS_BE;
3696         break;
3697     case IOPRIO_CLASS_IDLE:
3698         cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
3699         cfqq->ioprio = 7;
3700         cfq_clear_cfqq_idle_window(cfqq);
3701         break;
3702     }
3703 
3704     /*
3705      * keep track of original prio settings in case we have to temporarily
3706      * elevate the priority of this queue
3707      */
3708     cfqq->org_ioprio = cfqq->ioprio;
3709     cfqq->org_ioprio_class = cfqq->ioprio_class;
3710     cfq_clear_cfqq_prio_changed(cfqq);
3711 }
3712 
3713 static void check_ioprio_changed(struct cfq_io_cq *cic, struct bio *bio)
3714 {
3715     int ioprio = cic->icq.ioc->ioprio;
3716     struct cfq_data *cfqd = cic_to_cfqd(cic);
3717     struct cfq_queue *cfqq;
3718 
3719     /*
3720      * Check whether ioprio has changed.  The condition may trigger
3721      * spuriously on a newly created cic but there's no harm.
3722      */
3723     if (unlikely(!cfqd) || likely(cic->ioprio == ioprio))
3724         return;
3725 
3726     cfqq = cic_to_cfqq(cic, false);
3727     if (cfqq) {
3728         cfq_put_queue(cfqq);
3729         cfqq = cfq_get_queue(cfqd, BLK_RW_ASYNC, cic, bio);
3730         cic_set_cfqq(cic, cfqq, false);
3731     }
3732 
3733     cfqq = cic_to_cfqq(cic, true);
3734     if (cfqq)
3735         cfq_mark_cfqq_prio_changed(cfqq);
3736 
3737     cic->ioprio = ioprio;
3738 }
3739 
3740 static void cfq_init_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3741               pid_t pid, bool is_sync)
3742 {
3743     RB_CLEAR_NODE(&cfqq->rb_node);
3744     RB_CLEAR_NODE(&cfqq->p_node);
3745     INIT_LIST_HEAD(&cfqq->fifo);
3746 
3747     cfqq->ref = 0;
3748     cfqq->cfqd = cfqd;
3749 
3750     cfq_mark_cfqq_prio_changed(cfqq);
3751 
3752     if (is_sync) {
3753         if (!cfq_class_idle(cfqq))
3754             cfq_mark_cfqq_idle_window(cfqq);
3755         cfq_mark_cfqq_sync(cfqq);
3756     }
3757     cfqq->pid = pid;
3758 }
3759 
3760 #ifdef CONFIG_CFQ_GROUP_IOSCHED
3761 static bool check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio)
3762 {
3763     struct cfq_data *cfqd = cic_to_cfqd(cic);
3764     struct cfq_queue *cfqq;
3765     uint64_t serial_nr;
3766     bool nonroot_cg;
3767 
3768     rcu_read_lock();
3769     serial_nr = bio_blkcg(bio)->css.serial_nr;
3770     nonroot_cg = bio_blkcg(bio) != &blkcg_root;
3771     rcu_read_unlock();
3772 
3773     /*
3774      * Check whether blkcg has changed.  The condition may trigger
3775      * spuriously on a newly created cic but there's no harm.
3776      */
3777     if (unlikely(!cfqd) || likely(cic->blkcg_serial_nr == serial_nr))
3778         return nonroot_cg;
3779 
3780     /*
3781      * Drop reference to queues.  New queues will be assigned in new
3782      * group upon arrival of fresh requests.
3783      */
3784     cfqq = cic_to_cfqq(cic, false);
3785     if (cfqq) {
3786         cfq_log_cfqq(cfqd, cfqq, "changed cgroup");
3787         cic_set_cfqq(cic, NULL, false);
3788         cfq_put_queue(cfqq);
3789     }
3790 
3791     cfqq = cic_to_cfqq(cic, true);
3792     if (cfqq) {
3793         cfq_log_cfqq(cfqd, cfqq, "changed cgroup");
3794         cic_set_cfqq(cic, NULL, true);
3795         cfq_put_queue(cfqq);
3796     }
3797 
3798     cic->blkcg_serial_nr = serial_nr;
3799     return nonroot_cg;
3800 }
3801 #else
3802 static inline bool check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio)
3803 {
3804     return false;
3805 }
3806 #endif  /* CONFIG_CFQ_GROUP_IOSCHED */
3807 
3808 static struct cfq_queue **
3809 cfq_async_queue_prio(struct cfq_group *cfqg, int ioprio_class, int ioprio)
3810 {
3811     switch (ioprio_class) {
3812     case IOPRIO_CLASS_RT:
3813         return &cfqg->async_cfqq[0][ioprio];
3814     case IOPRIO_CLASS_NONE:
3815         ioprio = IOPRIO_NORM;
3816         /* fall through */
3817     case IOPRIO_CLASS_BE:
3818         return &cfqg->async_cfqq[1][ioprio];
3819     case IOPRIO_CLASS_IDLE:
3820         return &cfqg->async_idle_cfqq;
3821     default:
3822         BUG();
3823     }
3824 }
3825 
3826 static struct cfq_queue *
3827 cfq_get_queue(struct cfq_data *cfqd, bool is_sync, struct cfq_io_cq *cic,
3828           struct bio *bio)
3829 {
3830     int ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3831     int ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3832     struct cfq_queue **async_cfqq = NULL;
3833     struct cfq_queue *cfqq;
3834     struct cfq_group *cfqg;
3835 
3836     rcu_read_lock();
3837     cfqg = cfq_lookup_cfqg(cfqd, bio_blkcg(bio));
3838     if (!cfqg) {
3839         cfqq = &cfqd->oom_cfqq;
3840         goto out;
3841     }
3842 
3843     if (!is_sync) {
3844         if (!ioprio_valid(cic->ioprio)) {
3845             struct task_struct *tsk = current;
3846             ioprio = task_nice_ioprio(tsk);
3847             ioprio_class = task_nice_ioclass(tsk);
3848         }
3849         async_cfqq = cfq_async_queue_prio(cfqg, ioprio_class, ioprio);
3850         cfqq = *async_cfqq;
3851         if (cfqq)
3852             goto out;
3853     }
3854 
3855     cfqq = kmem_cache_alloc_node(cfq_pool,
3856                      GFP_NOWAIT | __GFP_ZERO | __GFP_NOWARN,
3857                      cfqd->queue->node);
3858     if (!cfqq) {
3859         cfqq = &cfqd->oom_cfqq;
3860         goto out;
3861     }
3862 
3863     cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync);
3864     cfq_init_prio_data(cfqq, cic);
3865     cfq_link_cfqq_cfqg(cfqq, cfqg);
3866     cfq_log_cfqq(cfqd, cfqq, "alloced");
3867 
3868     if (async_cfqq) {
3869         /* a new async queue is created, pin and remember */
3870         cfqq->ref++;
3871         *async_cfqq = cfqq;
3872     }
3873 out:
3874     cfqq->ref++;
3875     rcu_read_unlock();
3876     return cfqq;
3877 }
3878 
3879 static void
3880 __cfq_update_io_thinktime(struct cfq_ttime *ttime, u64 slice_idle)
3881 {
3882     u64 elapsed = ktime_get_ns() - ttime->last_end_request;
3883     elapsed = min(elapsed, 2UL * slice_idle);
3884 
3885     ttime->ttime_samples = (7*ttime->ttime_samples + 256) / 8;
3886     ttime->ttime_total = div_u64(7*ttime->ttime_total + 256*elapsed,  8);
3887     ttime->ttime_mean = div64_ul(ttime->ttime_total + 128,
3888                      ttime->ttime_samples);
3889 }
3890 
3891 static void
3892 cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3893             struct cfq_io_cq *cic)
3894 {
3895     if (cfq_cfqq_sync(cfqq)) {
3896         __cfq_update_io_thinktime(&cic->ttime, cfqd->cfq_slice_idle);
3897         __cfq_update_io_thinktime(&cfqq->service_tree->ttime,
3898             cfqd->cfq_slice_idle);
3899     }
3900 #ifdef CONFIG_CFQ_GROUP_IOSCHED
3901     __cfq_update_io_thinktime(&cfqq->cfqg->ttime, cfqd->cfq_group_idle);
3902 #endif
3903 }
3904 
3905 static void
3906 cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3907                struct request *rq)
3908 {
3909     sector_t sdist = 0;
3910     sector_t n_sec = blk_rq_sectors(rq);
3911     if (cfqq->last_request_pos) {
3912         if (cfqq->last_request_pos < blk_rq_pos(rq))
3913             sdist = blk_rq_pos(rq) - cfqq->last_request_pos;
3914         else
3915             sdist = cfqq->last_request_pos - blk_rq_pos(rq);
3916     }
3917 
3918     cfqq->seek_history <<= 1;
3919     if (blk_queue_nonrot(cfqd->queue))
3920         cfqq->seek_history |= (n_sec < CFQQ_SECT_THR_NONROT);
3921     else
3922         cfqq->seek_history |= (sdist > CFQQ_SEEK_THR);
3923 }
3924 
3925 static inline bool req_noidle(struct request *req)
3926 {
3927     return req_op(req) == REQ_OP_WRITE &&
3928         (req->cmd_flags & (REQ_SYNC | REQ_IDLE)) == REQ_SYNC;
3929 }
3930 
3931 /*
3932  * Disable idle window if the process thinks too long or seeks so much that
3933  * it doesn't matter
3934  */
3935 static void
3936 cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3937                struct cfq_io_cq *cic)
3938 {
3939     int old_idle, enable_idle;
3940 
3941     /*
3942      * Don't idle for async or idle io prio class
3943      */
3944     if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq))
3945         return;
3946 
3947     enable_idle = old_idle = cfq_cfqq_idle_window(cfqq);
3948 
3949     if (cfqq->queued[0] + cfqq->queued[1] >= 4)
3950         cfq_mark_cfqq_deep(cfqq);
3951 
3952     if (cfqq->next_rq && req_noidle(cfqq->next_rq))
3953         enable_idle = 0;
3954     else if (!atomic_read(&cic->icq.ioc->active_ref) ||
3955          !cfqd->cfq_slice_idle ||
3956          (!cfq_cfqq_deep(cfqq) && CFQQ_SEEKY(cfqq)))
3957         enable_idle = 0;
3958     else if (sample_valid(cic->ttime.ttime_samples)) {
3959         if (cic->ttime.ttime_mean > cfqd->cfq_slice_idle)
3960             enable_idle = 0;
3961         else
3962             enable_idle = 1;
3963     }
3964 
3965     if (old_idle != enable_idle) {
3966         cfq_log_cfqq(cfqd, cfqq, "idle=%d", enable_idle);
3967         if (enable_idle)
3968             cfq_mark_cfqq_idle_window(cfqq);
3969         else
3970             cfq_clear_cfqq_idle_window(cfqq);
3971     }
3972 }
3973 
3974 /*
3975  * Check if new_cfqq should preempt the currently active queue. Return 0 for
3976  * no or if we aren't sure, a 1 will cause a preempt.
3977  */
3978 static bool
3979 cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
3980            struct request *rq)
3981 {
3982     struct cfq_queue *cfqq;
3983 
3984     cfqq = cfqd->active_queue;
3985     if (!cfqq)
3986         return false;
3987 
3988     if (cfq_class_idle(new_cfqq))
3989         return false;
3990 
3991     if (cfq_class_idle(cfqq))
3992         return true;
3993 
3994     /*
3995      * Don't allow a non-RT request to preempt an ongoing RT cfqq timeslice.
3996      */
3997     if (cfq_class_rt(cfqq) && !cfq_class_rt(new_cfqq))
3998         return false;
3999 
4000     /*
4001      * if the new request is sync, but the currently running queue is
4002      * not, let the sync request have priority.
4003      */
4004     if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq) && !cfq_cfqq_must_dispatch(cfqq))
4005         return true;
4006 
4007     /*
4008      * Treat ancestors of current cgroup the same way as current cgroup.
4009      * For anybody else we disallow preemption to guarantee service
4010      * fairness among cgroups.
4011      */
4012     if (!cfqg_is_descendant(cfqq->cfqg, new_cfqq->cfqg))
4013         return false;
4014 
4015     if (cfq_slice_used(cfqq))
4016         return true;
4017 
4018     /*
4019      * Allow an RT request to pre-empt an ongoing non-RT cfqq timeslice.
4020      */
4021     if (cfq_class_rt(new_cfqq) && !cfq_class_rt(cfqq))
4022         return true;
4023 
4024     WARN_ON_ONCE(cfqq->ioprio_class != new_cfqq->ioprio_class);
4025     /* Allow preemption only if we are idling on sync-noidle tree */
4026     if (cfqd->serving_wl_type == SYNC_NOIDLE_WORKLOAD &&
4027         cfqq_type(new_cfqq) == SYNC_NOIDLE_WORKLOAD &&
4028         RB_EMPTY_ROOT(&cfqq->sort_list))
4029         return true;
4030 
4031     /*
4032      * So both queues are sync. Let the new request get disk time if
4033      * it's a metadata request and the current queue is doing regular IO.
4034      */
4035     if ((rq->cmd_flags & REQ_PRIO) && !cfqq->prio_pending)
4036         return true;
4037 
4038     /* An idle queue should not be idle now for some reason */
4039     if (RB_EMPTY_ROOT(&cfqq->sort_list) && !cfq_should_idle(cfqd, cfqq))
4040         return true;
4041 
4042     if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq))
4043         return false;
4044 
4045     /*
4046      * if this request is as-good as one we would expect from the
4047      * current cfqq, let it preempt
4048      */
4049     if (cfq_rq_close(cfqd, cfqq, rq))
4050         return true;
4051 
4052     return false;
4053 }
4054 
4055 /*
4056  * cfqq preempts the active queue. if we allowed preempt with no slice left,
4057  * let it have half of its nominal slice.
4058  */
4059 static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
4060 {
4061     enum wl_type_t old_type = cfqq_type(cfqd->active_queue);
4062 
4063     cfq_log_cfqq(cfqd, cfqq, "preempt");
4064     cfq_slice_expired(cfqd, 1);
4065 
4066     /*
4067      * workload type is changed, don't save slice, otherwise preempt
4068      * doesn't happen
4069      */
4070     if (old_type != cfqq_type(cfqq))
4071         cfqq->cfqg->saved_wl_slice = 0;
4072 
4073     /*
4074      * Put the new queue at the front of the of the current list,
4075      * so we know that it will be selected next.
4076      */
4077     BUG_ON(!cfq_cfqq_on_rr(cfqq));
4078 
4079     cfq_service_tree_add(cfqd, cfqq, 1);
4080 
4081     cfqq->slice_end = 0;
4082     cfq_mark_cfqq_slice_new(cfqq);
4083 }
4084 
4085 /*
4086  * Called when a new fs request (rq) is added (to cfqq). Check if there's
4087  * something we should do about it
4088  */
4089 static void
4090 cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
4091         struct request *rq)
4092 {
4093     struct cfq_io_cq *cic = RQ_CIC(rq);
4094 
4095     cfqd->rq_queued++;
4096     if (rq->cmd_flags & REQ_PRIO)
4097         cfqq->prio_pending++;
4098 
4099     cfq_update_io_thinktime(cfqd, cfqq, cic);
4100     cfq_update_io_seektime(cfqd, cfqq, rq);
4101     cfq_update_idle_window(cfqd, cfqq, cic);
4102 
4103     cfqq->last_request_pos = blk_rq_pos(rq) + blk_rq_sectors(rq);
4104 
4105     if (cfqq == cfqd->active_queue) {
4106         /*
4107          * Remember that we saw a request from this process, but
4108          * don't start queuing just yet. Otherwise we risk seeing lots
4109          * of tiny requests, because we disrupt the normal plugging
4110          * and merging. If the request is already larger than a single
4111          * page, let it rip immediately. For that case we assume that
4112          * merging is already done. Ditto for a busy system that
4113          * has other work pending, don't risk delaying until the
4114          * idle timer unplug to continue working.
4115          */
4116         if (cfq_cfqq_wait_request(cfqq)) {
4117             if (blk_rq_bytes(rq) > PAGE_SIZE ||
4118                 cfqd->busy_queues > 1) {
4119                 cfq_del_timer(cfqd, cfqq);
4120                 cfq_clear_cfqq_wait_request(cfqq);
4121                 __blk_run_queue(cfqd->queue);
4122             } else {
4123                 cfqg_stats_update_idle_time(cfqq->cfqg);
4124                 cfq_mark_cfqq_must_dispatch(cfqq);
4125             }
4126         }
4127     } else if (cfq_should_preempt(cfqd, cfqq, rq)) {
4128         /*
4129          * not the active queue - expire current slice if it is
4130          * idle and has expired it's mean thinktime or this new queue
4131          * has some old slice time left and is of higher priority or
4132          * this new queue is RT and the current one is BE
4133          */
4134         cfq_preempt_queue(cfqd, cfqq);
4135         __blk_run_queue(cfqd->queue);
4136     }
4137 }
4138 
4139 static void cfq_insert_request(struct request_queue *q, struct request *rq)
4140 {
4141     struct cfq_data *cfqd = q->elevator->elevator_data;
4142     struct cfq_queue *cfqq = RQ_CFQQ(rq);
4143 
4144     cfq_log_cfqq(cfqd, cfqq, "insert_request");
4145     cfq_init_prio_data(cfqq, RQ_CIC(rq));
4146 
4147     rq->fifo_time = ktime_get_ns() + cfqd->cfq_fifo_expire[rq_is_sync(rq)];
4148     list_add_tail(&rq->queuelist, &cfqq->fifo);
4149     cfq_add_rq_rb(rq);
4150     cfqg_stats_update_io_add(RQ_CFQG(rq), cfqd->serving_group,
4151                  rq->cmd_flags);
4152     cfq_rq_enqueued(cfqd, cfqq, rq);
4153 }
4154 
4155 /*
4156  * Update hw_tag based on peak queue depth over 50 samples under
4157  * sufficient load.
4158  */
4159 static void cfq_update_hw_tag(struct cfq_data *cfqd)
4160 {
4161     struct cfq_queue *cfqq = cfqd->active_queue;
4162 
4163     if (cfqd->rq_in_driver > cfqd->hw_tag_est_depth)
4164         cfqd->hw_tag_est_depth = cfqd->rq_in_driver;
4165 
4166     if (cfqd->hw_tag == 1)
4167         return;
4168 
4169     if (cfqd->rq_queued <= CFQ_HW_QUEUE_MIN &&
4170         cfqd->rq_in_driver <= CFQ_HW_QUEUE_MIN)
4171         return;
4172 
4173     /*
4174      * If active queue hasn't enough requests and can idle, cfq might not
4175      * dispatch sufficient requests to hardware. Don't zero hw_tag in this
4176      * case
4177      */
4178     if (cfqq && cfq_cfqq_idle_window(cfqq) &&
4179         cfqq->dispatched + cfqq->queued[0] + cfqq->queued[1] <
4180         CFQ_HW_QUEUE_MIN && cfqd->rq_in_driver < CFQ_HW_QUEUE_MIN)
4181         return;
4182 
4183     if (cfqd->hw_tag_samples++ < 50)
4184         return;
4185 
4186     if (cfqd->hw_tag_est_depth >= CFQ_HW_QUEUE_MIN)
4187         cfqd->hw_tag = 1;
4188     else
4189         cfqd->hw_tag = 0;
4190 }
4191 
4192 static bool cfq_should_wait_busy(struct cfq_data *cfqd, struct cfq_queue *cfqq)
4193 {
4194     struct cfq_io_cq *cic = cfqd->active_cic;
4195     u64 now = ktime_get_ns();
4196 
4197     /* If the queue already has requests, don't wait */
4198     if (!RB_EMPTY_ROOT(&cfqq->sort_list))
4199         return false;
4200 
4201     /* If there are other queues in the group, don't wait */
4202     if (cfqq->cfqg->nr_cfqq > 1)
4203         return false;
4204 
4205     /* the only queue in the group, but think time is big */
4206     if (cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true))
4207         return false;
4208 
4209     if (cfq_slice_used(cfqq))
4210         return true;
4211 
4212     /* if slice left is less than think time, wait busy */
4213     if (cic && sample_valid(cic->ttime.ttime_samples)
4214         && (cfqq->slice_end - now < cic->ttime.ttime_mean))
4215         return true;
4216 
4217     /*
4218      * If think times is less than a jiffy than ttime_mean=0 and above
4219      * will not be true. It might happen that slice has not expired yet
4220      * but will expire soon (4-5 ns) during select_queue(). To cover the
4221      * case where think time is less than a jiffy, mark the queue wait
4222      * busy if only 1 jiffy is left in the slice.
4223      */
4224     if (cfqq->slice_end - now <= jiffies_to_nsecs(1))
4225         return true;
4226 
4227     return false;
4228 }
4229 
4230 static void cfq_completed_request(struct request_queue *q, struct request *rq)
4231 {
4232     struct cfq_queue *cfqq = RQ_CFQQ(rq);
4233     struct cfq_data *cfqd = cfqq->cfqd;
4234     const int sync = rq_is_sync(rq);
4235     u64 now = ktime_get_ns();
4236 
4237     cfq_log_cfqq(cfqd, cfqq, "complete rqnoidle %d", req_noidle(rq));
4238 
4239     cfq_update_hw_tag(cfqd);
4240 
4241     WARN_ON(!cfqd->rq_in_driver);
4242     WARN_ON(!cfqq->dispatched);
4243     cfqd->rq_in_driver--;
4244     cfqq->dispatched--;
4245     (RQ_CFQG(rq))->dispatched--;
4246     cfqg_stats_update_completion(cfqq->cfqg, rq_start_time_ns(rq),
4247                      rq_io_start_time_ns(rq), rq->cmd_flags);
4248 
4249     cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]--;
4250 
4251     if (sync) {
4252         struct cfq_rb_root *st;
4253 
4254         RQ_CIC(rq)->ttime.last_end_request = now;
4255 
4256         if (cfq_cfqq_on_rr(cfqq))
4257             st = cfqq->service_tree;
4258         else
4259             st = st_for(cfqq->cfqg, cfqq_class(cfqq),
4260                     cfqq_type(cfqq));
4261 
4262         st->ttime.last_end_request = now;
4263         /*
4264          * We have to do this check in jiffies since start_time is in
4265          * jiffies and it is not trivial to convert to ns. If
4266          * cfq_fifo_expire[1] ever comes close to 1 jiffie, this test
4267          * will become problematic but so far we are fine (the default
4268          * is 128 ms).
4269          */
4270         if (!time_after(rq->start_time +
4271                   nsecs_to_jiffies(cfqd->cfq_fifo_expire[1]),
4272                 jiffies))
4273             cfqd->last_delayed_sync = now;
4274     }
4275 
4276 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4277     cfqq->cfqg->ttime.last_end_request = now;
4278 #endif
4279 
4280     /*
4281      * If this is the active queue, check if it needs to be expired,
4282      * or if we want to idle in case it has no pending requests.
4283      */
4284     if (cfqd->active_queue == cfqq) {
4285         const bool cfqq_empty = RB_EMPTY_ROOT(&cfqq->sort_list);
4286 
4287         if (cfq_cfqq_slice_new(cfqq)) {
4288             cfq_set_prio_slice(cfqd, cfqq);
4289             cfq_clear_cfqq_slice_new(cfqq);
4290         }
4291 
4292         /*
4293          * Should we wait for next request to come in before we expire
4294          * the queue.
4295          */
4296         if (cfq_should_wait_busy(cfqd, cfqq)) {
4297             u64 extend_sl = cfqd->cfq_slice_idle;
4298             if (!cfqd->cfq_slice_idle)
4299                 extend_sl = cfqd->cfq_group_idle;
4300             cfqq->slice_end = now + extend_sl;
4301             cfq_mark_cfqq_wait_busy(cfqq);
4302             cfq_log_cfqq(cfqd, cfqq, "will busy wait");
4303         }
4304 
4305         /*
4306          * Idling is not enabled on:
4307          * - expired queues
4308          * - idle-priority queues
4309          * - async queues
4310          * - queues with still some requests queued
4311          * - when there is a close cooperator
4312          */
4313         if (cfq_slice_used(cfqq) || cfq_class_idle(cfqq))
4314             cfq_slice_expired(cfqd, 1);
4315         else if (sync && cfqq_empty &&
4316              !cfq_close_cooperator(cfqd, cfqq)) {
4317             cfq_arm_slice_timer(cfqd);
4318         }
4319     }
4320 
4321     if (!cfqd->rq_in_driver)
4322         cfq_schedule_dispatch(cfqd);
4323 }
4324 
4325 static void cfqq_boost_on_prio(struct cfq_queue *cfqq, unsigned int op)
4326 {
4327     /*
4328      * If REQ_PRIO is set, boost class and prio level, if it's below
4329      * BE/NORM. If prio is not set, restore the potentially boosted
4330      * class/prio level.
4331      */
4332     if (!(op & REQ_PRIO)) {
4333         cfqq->ioprio_class = cfqq->org_ioprio_class;
4334         cfqq->ioprio = cfqq->org_ioprio;
4335     } else {
4336         if (cfq_class_idle(cfqq))
4337             cfqq->ioprio_class = IOPRIO_CLASS_BE;
4338         if (cfqq->ioprio > IOPRIO_NORM)
4339             cfqq->ioprio = IOPRIO_NORM;
4340     }
4341 }
4342 
4343 static inline int __cfq_may_queue(struct cfq_queue *cfqq)
4344 {
4345     if (cfq_cfqq_wait_request(cfqq) && !cfq_cfqq_must_alloc_slice(cfqq)) {
4346         cfq_mark_cfqq_must_alloc_slice(cfqq);
4347         return ELV_MQUEUE_MUST;
4348     }
4349 
4350     return ELV_MQUEUE_MAY;
4351 }
4352 
4353 static int cfq_may_queue(struct request_queue *q, unsigned int op)
4354 {
4355     struct cfq_data *cfqd = q->elevator->elevator_data;
4356     struct task_struct *tsk = current;
4357     struct cfq_io_cq *cic;
4358     struct cfq_queue *cfqq;
4359 
4360     /*
4361      * don't force setup of a queue from here, as a call to may_queue
4362      * does not necessarily imply that a request actually will be queued.
4363      * so just lookup a possibly existing queue, or return 'may queue'
4364      * if that fails
4365      */
4366     cic = cfq_cic_lookup(cfqd, tsk->io_context);
4367     if (!cic)
4368         return ELV_MQUEUE_MAY;
4369 
4370     cfqq = cic_to_cfqq(cic, op_is_sync(op));
4371     if (cfqq) {
4372         cfq_init_prio_data(cfqq, cic);
4373         cfqq_boost_on_prio(cfqq, op);
4374 
4375         return __cfq_may_queue(cfqq);
4376     }
4377 
4378     return ELV_MQUEUE_MAY;
4379 }
4380 
4381 /*
4382  * queue lock held here
4383  */
4384 static void cfq_put_request(struct request *rq)
4385 {
4386     struct cfq_queue *cfqq = RQ_CFQQ(rq);
4387 
4388     if (cfqq) {
4389         const int rw = rq_data_dir(rq);
4390 
4391         BUG_ON(!cfqq->allocated[rw]);
4392         cfqq->allocated[rw]--;
4393 
4394         /* Put down rq reference on cfqg */
4395         cfqg_put(RQ_CFQG(rq));
4396         rq->elv.priv[0] = NULL;
4397         rq->elv.priv[1] = NULL;
4398 
4399         cfq_put_queue(cfqq);
4400     }
4401 }
4402 
4403 static struct cfq_queue *
4404 cfq_merge_cfqqs(struct cfq_data *cfqd, struct cfq_io_cq *cic,
4405         struct cfq_queue *cfqq)
4406 {
4407     cfq_log_cfqq(cfqd, cfqq, "merging with queue %p", cfqq->new_cfqq);
4408     cic_set_cfqq(cic, cfqq->new_cfqq, 1);
4409     cfq_mark_cfqq_coop(cfqq->new_cfqq);
4410     cfq_put_queue(cfqq);
4411     return cic_to_cfqq(cic, 1);
4412 }
4413 
4414 /*
4415  * Returns NULL if a new cfqq should be allocated, or the old cfqq if this
4416  * was the last process referring to said cfqq.
4417  */
4418 static struct cfq_queue *
4419 split_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq)
4420 {
4421     if (cfqq_process_refs(cfqq) == 1) {
4422         cfqq->pid = current->pid;
4423         cfq_clear_cfqq_coop(cfqq);
4424         cfq_clear_cfqq_split_coop(cfqq);
4425         return cfqq;
4426     }
4427 
4428     cic_set_cfqq(cic, NULL, 1);
4429 
4430     cfq_put_cooperator(cfqq);
4431 
4432     cfq_put_queue(cfqq);
4433     return NULL;
4434 }
4435 /*
4436  * Allocate cfq data structures associated with this request.
4437  */
4438 static int
4439 cfq_set_request(struct request_queue *q, struct request *rq, struct bio *bio,
4440         gfp_t gfp_mask)
4441 {
4442     struct cfq_data *cfqd = q->elevator->elevator_data;
4443     struct cfq_io_cq *cic = icq_to_cic(rq->elv.icq);
4444     const int rw = rq_data_dir(rq);
4445     const bool is_sync = rq_is_sync(rq);
4446     struct cfq_queue *cfqq;
4447     bool disable_wbt;
4448 
4449     spin_lock_irq(q->queue_lock);
4450 
4451     check_ioprio_changed(cic, bio);
4452     disable_wbt = check_blkcg_changed(cic, bio);
4453 new_queue:
4454     cfqq = cic_to_cfqq(cic, is_sync);
4455     if (!cfqq || cfqq == &cfqd->oom_cfqq) {
4456         if (cfqq)
4457             cfq_put_queue(cfqq);
4458         cfqq = cfq_get_queue(cfqd, is_sync, cic, bio);
4459         cic_set_cfqq(cic, cfqq, is_sync);
4460     } else {
4461         /*
4462          * If the queue was seeky for too long, break it apart.
4463          */
4464         if (cfq_cfqq_coop(cfqq) && cfq_cfqq_split_coop(cfqq)) {
4465             cfq_log_cfqq(cfqd, cfqq, "breaking apart cfqq");
4466             cfqq = split_cfqq(cic, cfqq);
4467             if (!cfqq)
4468                 goto new_queue;
4469         }
4470 
4471         /*
4472          * Check to see if this queue is scheduled to merge with
4473          * another, closely cooperating queue.  The merging of
4474          * queues happens here as it must be done in process context.
4475          * The reference on new_cfqq was taken in merge_cfqqs.
4476          */
4477         if (cfqq->new_cfqq)
4478             cfqq = cfq_merge_cfqqs(cfqd, cic, cfqq);
4479     }
4480 
4481     cfqq->allocated[rw]++;
4482 
4483     cfqq->ref++;
4484     cfqg_get(cfqq->cfqg);
4485     rq->elv.priv[0] = cfqq;
4486     rq->elv.priv[1] = cfqq->cfqg;
4487     spin_unlock_irq(q->queue_lock);
4488 
4489     if (disable_wbt)
4490         wbt_disable_default(q);
4491 
4492     return 0;
4493 }
4494 
4495 static void cfq_kick_queue(struct work_struct *work)
4496 {
4497     struct cfq_data *cfqd =
4498         container_of(work, struct cfq_data, unplug_work);
4499     struct request_queue *q = cfqd->queue;
4500 
4501     spin_lock_irq(q->queue_lock);
4502     __blk_run_queue(cfqd->queue);
4503     spin_unlock_irq(q->queue_lock);
4504 }
4505 
4506 /*
4507  * Timer running if the active_queue is currently idling inside its time slice
4508  */
4509 static enum hrtimer_restart cfq_idle_slice_timer(struct hrtimer *timer)
4510 {
4511     struct cfq_data *cfqd = container_of(timer, struct cfq_data,
4512                          idle_slice_timer);
4513     struct cfq_queue *cfqq;
4514     unsigned long flags;
4515     int timed_out = 1;
4516 
4517     cfq_log(cfqd, "idle timer fired");
4518 
4519     spin_lock_irqsave(cfqd->queue->queue_lock, flags);
4520 
4521     cfqq = cfqd->active_queue;
4522     if (cfqq) {
4523         timed_out = 0;
4524 
4525         /*
4526          * We saw a request before the queue expired, let it through
4527          */
4528         if (cfq_cfqq_must_dispatch(cfqq))
4529             goto out_kick;
4530 
4531         /*
4532          * expired
4533          */
4534         if (cfq_slice_used(cfqq))
4535             goto expire;
4536 
4537         /*
4538          * only expire and reinvoke request handler, if there are
4539          * other queues with pending requests
4540          */
4541         if (!cfqd->busy_queues)
4542             goto out_cont;
4543 
4544         /*
4545          * not expired and it has a request pending, let it dispatch
4546          */
4547         if (!RB_EMPTY_ROOT(&cfqq->sort_list))
4548             goto out_kick;
4549 
4550         /*
4551          * Queue depth flag is reset only when the idle didn't succeed
4552          */
4553         cfq_clear_cfqq_deep(cfqq);
4554     }
4555 expire:
4556     cfq_slice_expired(cfqd, timed_out);
4557 out_kick:
4558     cfq_schedule_dispatch(cfqd);
4559 out_cont:
4560     spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
4561     return HRTIMER_NORESTART;
4562 }
4563 
4564 static void cfq_shutdown_timer_wq(struct cfq_data *cfqd)
4565 {
4566     hrtimer_cancel(&cfqd->idle_slice_timer);
4567     cancel_work_sync(&cfqd->unplug_work);
4568 }
4569 
4570 static void cfq_exit_queue(struct elevator_queue *e)
4571 {
4572     struct cfq_data *cfqd = e->elevator_data;
4573     struct request_queue *q = cfqd->queue;
4574 
4575     cfq_shutdown_timer_wq(cfqd);
4576 
4577     spin_lock_irq(q->queue_lock);
4578 
4579     if (cfqd->active_queue)
4580         __cfq_slice_expired(cfqd, cfqd->active_queue, 0);
4581 
4582     spin_unlock_irq(q->queue_lock);
4583 
4584     cfq_shutdown_timer_wq(cfqd);
4585 
4586 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4587     blkcg_deactivate_policy(q, &blkcg_policy_cfq);
4588 #else
4589     kfree(cfqd->root_group);
4590 #endif
4591     kfree(cfqd);
4592 }
4593 
4594 static int cfq_init_queue(struct request_queue *q, struct elevator_type *e)
4595 {
4596     struct cfq_data *cfqd;
4597     struct blkcg_gq *blkg __maybe_unused;
4598     int i, ret;
4599     struct elevator_queue *eq;
4600 
4601     eq = elevator_alloc(q, e);
4602     if (!eq)
4603         return -ENOMEM;
4604 
4605     cfqd = kzalloc_node(sizeof(*cfqd), GFP_KERNEL, q->node);
4606     if (!cfqd) {
4607         kobject_put(&eq->kobj);
4608         return -ENOMEM;
4609     }
4610     eq->elevator_data = cfqd;
4611 
4612     cfqd->queue = q;
4613     spin_lock_irq(q->queue_lock);
4614     q->elevator = eq;
4615     spin_unlock_irq(q->queue_lock);
4616 
4617     /* Init root service tree */
4618     cfqd->grp_service_tree = CFQ_RB_ROOT;
4619 
4620     /* Init root group and prefer root group over other groups by default */
4621 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4622     ret = blkcg_activate_policy(q, &blkcg_policy_cfq);
4623     if (ret)
4624         goto out_free;
4625 
4626     cfqd->root_group = blkg_to_cfqg(q->root_blkg);
4627 #else
4628     ret = -ENOMEM;
4629     cfqd->root_group = kzalloc_node(sizeof(*cfqd->root_group),
4630                     GFP_KERNEL, cfqd->queue->node);
4631     if (!cfqd->root_group)
4632         goto out_free;
4633 
4634     cfq_init_cfqg_base(cfqd->root_group);
4635     cfqd->root_group->weight = 2 * CFQ_WEIGHT_LEGACY_DFL;
4636     cfqd->root_group->leaf_weight = 2 * CFQ_WEIGHT_LEGACY_DFL;
4637 #endif
4638 
4639     /*
4640      * Not strictly needed (since RB_ROOT just clears the node and we
4641      * zeroed cfqd on alloc), but better be safe in case someone decides
4642      * to add magic to the rb code
4643      */
4644     for (i = 0; i < CFQ_PRIO_LISTS; i++)
4645         cfqd->prio_trees[i] = RB_ROOT;
4646 
4647     /*
4648      * Our fallback cfqq if cfq_get_queue() runs into OOM issues.
4649      * Grab a permanent reference to it, so that the normal code flow
4650      * will not attempt to free it.  oom_cfqq is linked to root_group
4651      * but shouldn't hold a reference as it'll never be unlinked.  Lose
4652      * the reference from linking right away.
4653      */
4654     cfq_init_cfqq(cfqd, &cfqd->oom_cfqq, 1, 0);
4655     cfqd->oom_cfqq.ref++;
4656 
4657     spin_lock_irq(q->queue_lock);
4658     cfq_link_cfqq_cfqg(&cfqd->oom_cfqq, cfqd->root_group);
4659     cfqg_put(cfqd->root_group);
4660     spin_unlock_irq(q->queue_lock);
4661 
4662     hrtimer_init(&cfqd->idle_slice_timer, CLOCK_MONOTONIC,
4663              HRTIMER_MODE_REL);
4664     cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
4665 
4666     INIT_WORK(&cfqd->unplug_work, cfq_kick_queue);
4667 
4668     cfqd->cfq_quantum = cfq_quantum;
4669     cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0];
4670     cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1];
4671     cfqd->cfq_back_max = cfq_back_max;
4672     cfqd->cfq_back_penalty = cfq_back_penalty;
4673     cfqd->cfq_slice[0] = cfq_slice_async;
4674     cfqd->cfq_slice[1] = cfq_slice_sync;
4675     cfqd->cfq_target_latency = cfq_target_latency;
4676     cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
4677     cfqd->cfq_slice_idle = cfq_slice_idle;
4678     cfqd->cfq_group_idle = cfq_group_idle;
4679     cfqd->cfq_latency = 1;
4680     cfqd->hw_tag = -1;
4681     /*
4682      * we optimistically start assuming sync ops weren't delayed in last
4683      * second, in order to have larger depth for async operations.
4684      */
4685     cfqd->last_delayed_sync = ktime_get_ns() - NSEC_PER_SEC;
4686     return 0;
4687 
4688 out_free:
4689     kfree(cfqd);
4690     kobject_put(&eq->kobj);
4691     return ret;
4692 }
4693 
4694 static void cfq_registered_queue(struct request_queue *q)
4695 {
4696     struct elevator_queue *e = q->elevator;
4697     struct cfq_data *cfqd = e->elevator_data;
4698 
4699     /*
4700      * Default to IOPS mode with no idling for SSDs
4701      */
4702     if (blk_queue_nonrot(q))
4703         cfqd->cfq_slice_idle = 0;
4704 }
4705 
4706 /*
4707  * sysfs parts below -->
4708  */
4709 static ssize_t
4710 cfq_var_show(unsigned int var, char *page)
4711 {
4712     return sprintf(page, "%u\n", var);
4713 }
4714 
4715 static ssize_t
4716 cfq_var_store(unsigned int *var, const char *page, size_t count)
4717 {
4718     char *p = (char *) page;
4719 
4720     *var = simple_strtoul(p, &p, 10);
4721     return count;
4722 }
4723 
4724 #define SHOW_FUNCTION(__FUNC, __VAR, __CONV)                \
4725 static ssize_t __FUNC(struct elevator_queue *e, char *page)     \
4726 {                                   \
4727     struct cfq_data *cfqd = e->elevator_data;           \
4728     u64 __data = __VAR;                     \
4729     if (__CONV)                         \
4730         __data = div_u64(__data, NSEC_PER_MSEC);            \
4731     return cfq_var_show(__data, (page));                \
4732 }
4733 SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0);
4734 SHOW_FUNCTION(cfq_fifo_expire_sync_show, cfqd->cfq_fifo_expire[1], 1);
4735 SHOW_FUNCTION(cfq_fifo_expire_async_show, cfqd->cfq_fifo_expire[0], 1);
4736 SHOW_FUNCTION(cfq_back_seek_max_show, cfqd->cfq_back_max, 0);
4737 SHOW_FUNCTION(cfq_back_seek_penalty_show, cfqd->cfq_back_penalty, 0);
4738 SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1);
4739 SHOW_FUNCTION(cfq_group_idle_show, cfqd->cfq_group_idle, 1);
4740 SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1);
4741 SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
4742 SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
4743 SHOW_FUNCTION(cfq_low_latency_show, cfqd->cfq_latency, 0);
4744 SHOW_FUNCTION(cfq_target_latency_show, cfqd->cfq_target_latency, 1);
4745 #undef SHOW_FUNCTION
4746 
4747 #define USEC_SHOW_FUNCTION(__FUNC, __VAR)               \
4748 static ssize_t __FUNC(struct elevator_queue *e, char *page)     \
4749 {                                   \
4750     struct cfq_data *cfqd = e->elevator_data;           \
4751     u64 __data = __VAR;                     \
4752     __data = div_u64(__data, NSEC_PER_USEC);            \
4753     return cfq_var_show(__data, (page));                \
4754 }
4755 USEC_SHOW_FUNCTION(cfq_slice_idle_us_show, cfqd->cfq_slice_idle);
4756 USEC_SHOW_FUNCTION(cfq_group_idle_us_show, cfqd->cfq_group_idle);
4757 USEC_SHOW_FUNCTION(cfq_slice_sync_us_show, cfqd->cfq_slice[1]);
4758 USEC_SHOW_FUNCTION(cfq_slice_async_us_show, cfqd->cfq_slice[0]);
4759 USEC_SHOW_FUNCTION(cfq_target_latency_us_show, cfqd->cfq_target_latency);
4760 #undef USEC_SHOW_FUNCTION
4761 
4762 #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV)         \
4763 static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count) \
4764 {                                   \
4765     struct cfq_data *cfqd = e->elevator_data;           \
4766     unsigned int __data;                        \
4767     int ret = cfq_var_store(&__data, (page), count);        \
4768     if (__data < (MIN))                     \
4769         __data = (MIN);                     \
4770     else if (__data > (MAX))                    \
4771         __data = (MAX);                     \
4772     if (__CONV)                         \
4773         *(__PTR) = (u64)__data * NSEC_PER_MSEC;         \
4774     else                                \
4775         *(__PTR) = __data;                  \
4776     return ret;                         \
4777 }
4778 STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
4779 STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1,
4780         UINT_MAX, 1);
4781 STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1,
4782         UINT_MAX, 1);
4783 STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
4784 STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1,
4785         UINT_MAX, 0);
4786 STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1);
4787 STORE_FUNCTION(cfq_group_idle_store, &cfqd->cfq_group_idle, 0, UINT_MAX, 1);
4788 STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1);
4789 STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1);
4790 STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1,
4791         UINT_MAX, 0);
4792 STORE_FUNCTION(cfq_low_latency_store, &cfqd->cfq_latency, 0, 1, 0);
4793 STORE_FUNCTION(cfq_target_latency_store, &cfqd->cfq_target_latency, 1, UINT_MAX, 1);
4794 #undef STORE_FUNCTION
4795 
4796 #define USEC_STORE_FUNCTION(__FUNC, __PTR, MIN, MAX)            \
4797 static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count) \
4798 {                                   \
4799     struct cfq_data *cfqd = e->elevator_data;           \
4800     unsigned int __data;                        \
4801     int ret = cfq_var_store(&__data, (page), count);        \
4802     if (__data < (MIN))                     \
4803         __data = (MIN);                     \
4804     else if (__data > (MAX))                    \
4805         __data = (MAX);                     \
4806     *(__PTR) = (u64)__data * NSEC_PER_USEC;             \
4807     return ret;                         \
4808 }
4809 USEC_STORE_FUNCTION(cfq_slice_idle_us_store, &cfqd->cfq_slice_idle, 0, UINT_MAX);
4810 USEC_STORE_FUNCTION(cfq_group_idle_us_store, &cfqd->cfq_group_idle, 0, UINT_MAX);
4811 USEC_STORE_FUNCTION(cfq_slice_sync_us_store, &cfqd->cfq_slice[1], 1, UINT_MAX);
4812 USEC_STORE_FUNCTION(cfq_slice_async_us_store, &cfqd->cfq_slice[0], 1, UINT_MAX);
4813 USEC_STORE_FUNCTION(cfq_target_latency_us_store, &cfqd->cfq_target_latency, 1, UINT_MAX);
4814 #undef USEC_STORE_FUNCTION
4815 
4816 #define CFQ_ATTR(name) \
4817     __ATTR(name, S_IRUGO|S_IWUSR, cfq_##name##_show, cfq_##name##_store)
4818 
4819 static struct elv_fs_entry cfq_attrs[] = {
4820     CFQ_ATTR(quantum),
4821     CFQ_ATTR(fifo_expire_sync),
4822     CFQ_ATTR(fifo_expire_async),
4823     CFQ_ATTR(back_seek_max),
4824     CFQ_ATTR(back_seek_penalty),
4825     CFQ_ATTR(slice_sync),
4826     CFQ_ATTR(slice_sync_us),
4827     CFQ_ATTR(slice_async),
4828     CFQ_ATTR(slice_async_us),
4829     CFQ_ATTR(slice_async_rq),
4830     CFQ_ATTR(slice_idle),
4831     CFQ_ATTR(slice_idle_us),
4832     CFQ_ATTR(group_idle),
4833     CFQ_ATTR(group_idle_us),
4834     CFQ_ATTR(low_latency),
4835     CFQ_ATTR(target_latency),
4836     CFQ_ATTR(target_latency_us),
4837     __ATTR_NULL
4838 };
4839 
4840 static struct elevator_type iosched_cfq = {
4841     .ops = {
4842         .elevator_merge_fn =        cfq_merge,
4843         .elevator_merged_fn =       cfq_merged_request,
4844         .elevator_merge_req_fn =    cfq_merged_requests,
4845         .elevator_allow_bio_merge_fn =  cfq_allow_bio_merge,
4846         .elevator_allow_rq_merge_fn =   cfq_allow_rq_merge,
4847         .elevator_bio_merged_fn =   cfq_bio_merged,
4848         .elevator_dispatch_fn =     cfq_dispatch_requests,
4849         .elevator_add_req_fn =      cfq_insert_request,
4850         .elevator_activate_req_fn = cfq_activate_request,
4851         .elevator_deactivate_req_fn =   cfq_deactivate_request,
4852         .elevator_completed_req_fn =    cfq_completed_request,
4853         .elevator_former_req_fn =   elv_rb_former_request,
4854         .elevator_latter_req_fn =   elv_rb_latter_request,
4855         .elevator_init_icq_fn =     cfq_init_icq,
4856         .elevator_exit_icq_fn =     cfq_exit_icq,
4857         .elevator_set_req_fn =      cfq_set_request,
4858         .elevator_put_req_fn =      cfq_put_request,
4859         .elevator_may_queue_fn =    cfq_may_queue,
4860         .elevator_init_fn =     cfq_init_queue,
4861         .elevator_exit_fn =     cfq_exit_queue,
4862         .elevator_registered_fn =   cfq_registered_queue,
4863     },
4864     .icq_size   =   sizeof(struct cfq_io_cq),
4865     .icq_align  =   __alignof__(struct cfq_io_cq),
4866     .elevator_attrs =   cfq_attrs,
4867     .elevator_name  =   "cfq",
4868     .elevator_owner =   THIS_MODULE,
4869 };
4870 
4871 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4872 static struct blkcg_policy blkcg_policy_cfq = {
4873     .dfl_cftypes        = cfq_blkcg_files,
4874     .legacy_cftypes     = cfq_blkcg_legacy_files,
4875 
4876     .cpd_alloc_fn       = cfq_cpd_alloc,
4877     .cpd_init_fn        = cfq_cpd_init,
4878     .cpd_free_fn        = cfq_cpd_free,
4879     .cpd_bind_fn        = cfq_cpd_bind,
4880 
4881     .pd_alloc_fn        = cfq_pd_alloc,
4882     .pd_init_fn     = cfq_pd_init,
4883     .pd_offline_fn      = cfq_pd_offline,
4884     .pd_free_fn     = cfq_pd_free,
4885     .pd_reset_stats_fn  = cfq_pd_reset_stats,
4886 };
4887 #endif
4888 
4889 static int __init cfq_init(void)
4890 {
4891     int ret;
4892 
4893 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4894     ret = blkcg_policy_register(&blkcg_policy_cfq);
4895     if (ret)
4896         return ret;
4897 #else
4898     cfq_group_idle = 0;
4899 #endif
4900 
4901     ret = -ENOMEM;
4902     cfq_pool = KMEM_CACHE(cfq_queue, 0);
4903     if (!cfq_pool)
4904         goto err_pol_unreg;
4905 
4906     ret = elv_register(&iosched_cfq);
4907     if (ret)
4908         goto err_free_pool;
4909 
4910     return 0;
4911 
4912 err_free_pool:
4913     kmem_cache_destroy(cfq_pool);
4914 err_pol_unreg:
4915 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4916     blkcg_policy_unregister(&blkcg_policy_cfq);
4917 #endif
4918     return ret;
4919 }
4920 
4921 static void __exit cfq_exit(void)
4922 {
4923 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4924     blkcg_policy_unregister(&blkcg_policy_cfq);
4925 #endif
4926     elv_unregister(&iosched_cfq);
4927     kmem_cache_destroy(cfq_pool);
4928 }
4929 
4930 module_init(cfq_init);
4931 module_exit(cfq_exit);
4932 
4933 MODULE_AUTHOR("Jens Axboe");
4934 MODULE_LICENSE("GPL");
4935 MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler");