block/blk-throttle.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Interface for controlling IO bandwidth on a request queue
   4  *
   5  * Copyright (C) 2010 Vivek Goyal <vgoyal@redhat.com>
   6  */
   7
   8 #include <linux/module.h>
   9 #include <linux/slab.h>
  10 #include <linux/blkdev.h>
  11 #include <linux/bio.h>
  12 #include <linux/blktrace_api.h>
  13 #include <linux/blk-cgroup.h>
  14 #include "blk.h"
  15 #include "blk-cgroup-rwstat.h"
  16
  17 /* Max dispatch from a group in 1 round */
  18 #define THROTL_GRP_QUANTUM 8
  19
  20 /* Total max dispatch from all groups in one round */
  21 #define THROTL_QUANTUM 32
  22
  23 /* Throttling is performed over a slice and after that slice is renewed */
  24 #define DFL_THROTL_SLICE_HD (HZ / 10)
  25 #define DFL_THROTL_SLICE_SSD (HZ / 50)
  26 #define MAX_THROTL_SLICE (HZ)
  27 #define MAX_IDLE_TIME (5L * 1000 * 1000) /* 5 s */
  28 #define MIN_THROTL_BPS (320 * 1024)
  29 #define MIN_THROTL_IOPS (10)
  30 #define DFL_LATENCY_TARGET (-1L)
  31 #define DFL_IDLE_THRESHOLD (0)
  32 #define DFL_HD_BASELINE_LATENCY (4000L) /* 4ms */
  33 #define LATENCY_FILTERED_SSD (0)
  34 /*
  35  * For HD, very small latency comes from sequential IO. Such IO is helpless to
  36  * help determine if its IO is impacted by others, hence we ignore the IO
  37  */
  38 #define LATENCY_FILTERED_HD (1000L) /* 1ms */
  39
  40 static struct blkcg_policy blkcg_policy_throtl;
  41
  42 /* A workqueue to queue throttle related work */
  43 static struct workqueue_struct *kthrotld_workqueue;
  44
  45 /*
  46  * To implement hierarchical throttling, throtl_grps form a tree and bios
  47  * are dispatched upwards level by level until they reach the top and get
  48  * issued.  When dispatching bios from the children and local group at each
  49  * level, if the bios are dispatched into a single bio_list, there's a risk
  50  * of a local or child group which can queue many bios at once filling up
  51  * the list starving others.
  52  *
  53  * To avoid such starvation, dispatched bios are queued separately
  54  * according to where they came from.  When they are again dispatched to
  55  * the parent, they're popped in round-robin order so that no single source
  56  * hogs the dispatch window.
  57  *
  58  * throtl_qnode is used to keep the queued bios separated by their sources.
  59  * Bios are queued to throtl_qnode which in turn is queued to
  60  * throtl_service_queue and then dispatched in round-robin order.
  61  *
  62  * It's also used to track the reference counts on blkg's.  A qnode always
  63  * belongs to a throtl_grp and gets queued on itself or the parent, so
  64  * incrementing the reference of the associated throtl_grp when a qnode is
  65  * queued and decrementing when dequeued is enough to keep the whole blkg
  66  * tree pinned while bios are in flight.
  67  */
  68 struct throtl_qnode {
  69         struct list_head        node;           /* service_queue->queued[] */
  70         struct bio_list         bios;           /* queued bios */
  71         struct throtl_grp       *tg;            /* tg this qnode belongs to */
  72 };
  73
  74 struct throtl_service_queue {
  75         struct throtl_service_queue *parent_sq; /* the parent service_queue */
  76
  77         /*
  78          * Bios queued directly to this service_queue or dispatched from
  79          * children throtl_grp's.
  80          */
  81         struct list_head        queued[2];      /* throtl_qnode [READ/WRITE] */
  82         unsigned int            nr_queued[2];   /* number of queued bios */
  83
  84         /*
  85          * RB tree of active children throtl_grp's, which are sorted by
  86          * their ->disptime.
  87          */
  88         struct rb_root_cached   pending_tree;   /* RB tree of active tgs */
  89         unsigned int            nr_pending;     /* # queued in the tree */
  90         unsigned long           first_pending_disptime; /* disptime of the first tg */
  91         struct timer_list       pending_timer;  /* fires on first_pending_disptime */
  92 };
  93
  94 enum tg_state_flags {
  95         THROTL_TG_PENDING       = 1 << 0,       /* on parent's pending tree */
  96         THROTL_TG_WAS_EMPTY     = 1 << 1,       /* bio_lists[] became non-empty */
  97 };
  98
  99 #define rb_entry_tg(node)       rb_entry((node), struct throtl_grp, rb_node)
 100
 101 enum {
 102         LIMIT_LOW,
 103         LIMIT_MAX,
 104         LIMIT_CNT,
 105 };
 106
 107 struct throtl_grp {
 108         /* must be the first member */
 109         struct blkg_policy_data pd;
 110
 111         /* active throtl group service_queue member */
 112         struct rb_node rb_node;
 113
 114         /* throtl_data this group belongs to */
 115         struct throtl_data *td;
 116
 117         /* this group's service queue */
 118         struct throtl_service_queue service_queue;
 119
 120         /*
 121          * qnode_on_self is used when bios are directly queued to this
 122          * throtl_grp so that local bios compete fairly with bios
 123          * dispatched from children.  qnode_on_parent is used when bios are
 124          * dispatched from this throtl_grp into its parent and will compete
 125          * with the sibling qnode_on_parents and the parent's
 126          * qnode_on_self.
 127          */
 128         struct throtl_qnode qnode_on_self[2];
 129         struct throtl_qnode qnode_on_parent[2];
 130
 131         /*
 132          * Dispatch time in jiffies. This is the estimated time when group
 133          * will unthrottle and is ready to dispatch more bio. It is used as
 134          * key to sort active groups in service tree.
 135          */
 136         unsigned long disptime;
 137
 138         unsigned int flags;
 139
 140         /* are there any throtl rules between this group and td? */
 141         bool has_rules[2];
 142
 143         /* internally used bytes per second rate limits */
 144         uint64_t bps[2][LIMIT_CNT];
 145         /* user configured bps limits */
 146         uint64_t bps_conf[2][LIMIT_CNT];
 147
 148         /* internally used IOPS limits */
 149         unsigned int iops[2][LIMIT_CNT];
 150         /* user configured IOPS limits */
 151         unsigned int iops_conf[2][LIMIT_CNT];
 152
 153         /* Number of bytes dispatched in current slice */
 154         uint64_t bytes_disp[2];
 155         /* Number of bio's dispatched in current slice */
 156         unsigned int io_disp[2];
 157
 158         unsigned long last_low_overflow_time[2];
 159
 160         uint64_t last_bytes_disp[2];
 161         unsigned int last_io_disp[2];
 162
 163         unsigned long last_check_time;
 164
 165         unsigned long latency_target; /* us */
 166         unsigned long latency_target_conf; /* us */
 167         /* When did we start a new slice */
 168         unsigned long slice_start[2];
 169         unsigned long slice_end[2];
 170
 171         unsigned long last_finish_time; /* ns / 1024 */
 172         unsigned long checked_last_finish_time; /* ns / 1024 */
 173         unsigned long avg_idletime; /* ns / 1024 */
 174         unsigned long idletime_threshold; /* us */
 175         unsigned long idletime_threshold_conf; /* us */
 176
 177         unsigned int bio_cnt; /* total bios */
 178         unsigned int bad_bio_cnt; /* bios exceeding latency threshold */
 179         unsigned long bio_cnt_reset_time;
 180
 181         atomic_t io_split_cnt[2];
 182         atomic_t last_io_split_cnt[2];
 183
 184         struct blkg_rwstat stat_bytes;
 185         struct blkg_rwstat stat_ios;
 186 };
 187
 188 /* We measure latency for request size from <= 4k to >= 1M */
 189 #define LATENCY_BUCKET_SIZE 9
 190
 191 struct latency_bucket {
 192         unsigned long total_latency; /* ns / 1024 */
 193         int samples;
 194 };
 195
 196 struct avg_latency_bucket {
 197         unsigned long latency; /* ns / 1024 */
 198         bool valid;
 199 };
 200
 201 struct throtl_data
 202 {
 203         /* service tree for active throtl groups */
 204         struct throtl_service_queue service_queue;
 205
 206         struct request_queue *queue;
 207
 208         /* Total Number of queued bios on READ and WRITE lists */
 209         unsigned int nr_queued[2];
 210
 211         unsigned int throtl_slice;
 212
 213         /* Work for dispatching throttled bios */
 214         struct work_struct dispatch_work;
 215         unsigned int limit_index;
 216         bool limit_valid[LIMIT_CNT];
 217
 218         unsigned long low_upgrade_time;
 219         unsigned long low_downgrade_time;
 220
 221         unsigned int scale;
 222
 223         struct latency_bucket tmp_buckets[2][LATENCY_BUCKET_SIZE];
 224         struct avg_latency_bucket avg_buckets[2][LATENCY_BUCKET_SIZE];
 225         struct latency_bucket __percpu *latency_buckets[2];
 226         unsigned long last_calculate_time;
 227         unsigned long filtered_latency;
 228
 229         bool track_bio_latency;
 230 };
 231
 232 static void throtl_pending_timer_fn(struct timer_list *t);
 233
 234 static inline struct throtl_grp *pd_to_tg(struct blkg_policy_data *pd)
 235 {
 236         return pd ? container_of(pd, struct throtl_grp, pd) : NULL;
 237 }
 238
 239 static inline struct throtl_grp *blkg_to_tg(struct blkcg_gq *blkg)
 240 {
 241         return pd_to_tg(blkg_to_pd(blkg, &blkcg_policy_throtl));
 242 }
 243
 244 static inline struct blkcg_gq *tg_to_blkg(struct throtl_grp *tg)
 245 {
 246         return pd_to_blkg(&tg->pd);
 247 }
 248
 249 /**
 250  * sq_to_tg - return the throl_grp the specified service queue belongs to
 251  * @sq: the throtl_service_queue of interest
 252  *
 253  * Return the throtl_grp @sq belongs to.  If @sq is the top-level one
 254  * embedded in throtl_data, %NULL is returned.
 255  */
 256 static struct throtl_grp *sq_to_tg(struct throtl_service_queue *sq)
 257 {
 258         if (sq && sq->parent_sq)
 259                 return container_of(sq, struct throtl_grp, service_queue);
 260         else
 261                 return NULL;
 262 }
 263
 264 /**
 265  * sq_to_td - return throtl_data the specified service queue belongs to
 266  * @sq: the throtl_service_queue of interest
 267  *
 268  * A service_queue can be embedded in either a throtl_grp or throtl_data.
 269  * Determine the associated throtl_data accordingly and return it.
 270  */
 271 static struct throtl_data *sq_to_td(struct throtl_service_queue *sq)
 272 {
 273         struct throtl_grp *tg = sq_to_tg(sq);
 274
 275         if (tg)
 276                 return tg->td;
 277         else
 278                 return container_of(sq, struct throtl_data, service_queue);
 279 }
 280
 281 /*
 282  * cgroup's limit in LIMIT_MAX is scaled if low limit is set. This scale is to
 283  * make the IO dispatch more smooth.
 284  * Scale up: linearly scale up according to lapsed time since upgrade. For
 285  *           every throtl_slice, the limit scales up 1/2 .low limit till the
 286  *           limit hits .max limit
 287  * Scale down: exponentially scale down if a cgroup doesn't hit its .low limit
 288  */
 289 static uint64_t throtl_adjusted_limit(uint64_t low, struct throtl_data *td)
 290 {
 291         /* arbitrary value to avoid too big scale */
 292         if (td->scale < 4096 && time_after_eq(jiffies,
 293             td->low_upgrade_time + td->scale * td->throtl_slice))
 294                 td->scale = (jiffies - td->low_upgrade_time) / td->throtl_slice;
 295
 296         return low + (low >> 1) * td->scale;
 297 }
 298
 299 static uint64_t tg_bps_limit(struct throtl_grp *tg, int rw)
 300 {
 301         struct blkcg_gq *blkg = tg_to_blkg(tg);
 302         struct throtl_data *td;
 303         uint64_t ret;
 304
 305         if (cgroup_subsys_on_dfl(io_cgrp_subsys) && !blkg->parent)
 306                 return U64_MAX;
 307
 308         td = tg->td;
 309         ret = tg->bps[rw][td->limit_index];
 310         if (ret == 0 && td->limit_index == LIMIT_LOW) {
 311                 /* intermediate node or iops isn't 0 */
 312                 if (!list_empty(&blkg->blkcg->css.children) ||
 313                     tg->iops[rw][td->limit_index])
 314                         return U64_MAX;
 315                 else
 316                         return MIN_THROTL_BPS;
 317         }
 318
 319         if (td->limit_index == LIMIT_MAX && tg->bps[rw][LIMIT_LOW] &&
 320             tg->bps[rw][LIMIT_LOW] != tg->bps[rw][LIMIT_MAX]) {
 321                 uint64_t adjusted;
 322
 323                 adjusted = throtl_adjusted_limit(tg->bps[rw][LIMIT_LOW], td);
 324                 ret = min(tg->bps[rw][LIMIT_MAX], adjusted);
 325         }
 326         return ret;
 327 }
 328
 329 static unsigned int tg_iops_limit(struct throtl_grp *tg, int rw)
 330 {
 331         struct blkcg_gq *blkg = tg_to_blkg(tg);
 332         struct throtl_data *td;
 333         unsigned int ret;
 334
 335         if (cgroup_subsys_on_dfl(io_cgrp_subsys) && !blkg->parent)
 336                 return UINT_MAX;
 337
 338         td = tg->td;
 339         ret = tg->iops[rw][td->limit_index];
 340         if (ret == 0 && tg->td->limit_index == LIMIT_LOW) {
 341                 /* intermediate node or bps isn't 0 */
 342                 if (!list_empty(&blkg->blkcg->css.children) ||
 343                     tg->bps[rw][td->limit_index])
 344                         return UINT_MAX;
 345                 else
 346                         return MIN_THROTL_IOPS;
 347         }
 348
 349         if (td->limit_index == LIMIT_MAX && tg->iops[rw][LIMIT_LOW] &&
 350             tg->iops[rw][LIMIT_LOW] != tg->iops[rw][LIMIT_MAX]) {
 351                 uint64_t adjusted;
 352
 353                 adjusted = throtl_adjusted_limit(tg->iops[rw][LIMIT_LOW], td);
 354                 if (adjusted > UINT_MAX)
 355                         adjusted = UINT_MAX;
 356                 ret = min_t(unsigned int, tg->iops[rw][LIMIT_MAX], adjusted);
 357         }
 358         return ret;
 359 }
 360
 361 #define request_bucket_index(sectors) \
 362         clamp_t(int, order_base_2(sectors) - 3, 0, LATENCY_BUCKET_SIZE - 1)
 363
 364 /**
 365  * throtl_log - log debug message via blktrace
 366  * @sq: the service_queue being reported
 367  * @fmt: printf format string
 368  * @args: printf args
 369  *
 370  * The messages are prefixed with "throtl BLKG_NAME" if @sq belongs to a
 371  * throtl_grp; otherwise, just "throtl".
 372  */
 373 #define throtl_log(sq, fmt, args...)    do {                            \
 374         struct throtl_grp *__tg = sq_to_tg((sq));                       \
 375         struct throtl_data *__td = sq_to_td((sq));                      \
 376                                                                         \
 377         (void)__td;                                                     \
 378         if (likely(!blk_trace_note_message_enabled(__td->queue)))       \
 379                 break;                                                  \
 380         if ((__tg)) {                                                   \
 381                 blk_add_cgroup_trace_msg(__td->queue,                   \
 382                         tg_to_blkg(__tg)->blkcg, "throtl " fmt, ##args);\
 383         } else {                                                        \
 384                 blk_add_trace_msg(__td->queue, "throtl " fmt, ##args);  \
 385         }                                                               \
 386 } while (0)
 387
 388 static inline unsigned int throtl_bio_data_size(struct bio *bio)
 389 {
 390         /* assume it's one sector */
 391         if (unlikely(bio_op(bio) == REQ_OP_DISCARD))
 392                 return 512;
 393         return bio->bi_iter.bi_size;
 394 }
 395
 396 static void throtl_qnode_init(struct throtl_qnode *qn, struct throtl_grp *tg)
 397 {
 398         INIT_LIST_HEAD(&qn->node);
 399         bio_list_init(&qn->bios);
 400         qn->tg = tg;
 401 }
 402
 403 /**
 404  * throtl_qnode_add_bio - add a bio to a throtl_qnode and activate it
 405  * @bio: bio being added
 406  * @qn: qnode to add bio to
 407  * @queued: the service_queue->queued[] list @qn belongs to
 408  *
 409  * Add @bio to @qn and put @qn on @queued if it's not already on.
 410  * @qn->tg's reference count is bumped when @qn is activated.  See the
 411  * comment on top of throtl_qnode definition for details.
 412  */
 413 static void throtl_qnode_add_bio(struct bio *bio, struct throtl_qnode *qn,
 414                                  struct list_head *queued)
 415 {
 416         bio_list_add(&qn->bios, bio);
 417         if (list_empty(&qn->node)) {
 418                 list_add_tail(&qn->node, queued);
 419                 blkg_get(tg_to_blkg(qn->tg));
 420         }
 421 }
 422
 423 /**
 424  * throtl_peek_queued - peek the first bio on a qnode list
 425  * @queued: the qnode list to peek
 426  */
 427 static struct bio *throtl_peek_queued(struct list_head *queued)
 428 {
 429         struct throtl_qnode *qn;
 430         struct bio *bio;
 431
 432         if (list_empty(queued))
 433                 return NULL;
 434
 435         qn = list_first_entry(queued, struct throtl_qnode, node);
 436         bio = bio_list_peek(&qn->bios);
 437         WARN_ON_ONCE(!bio);
 438         return bio;
 439 }
 440
 441 /**
 442  * throtl_pop_queued - pop the first bio form a qnode list
 443  * @queued: the qnode list to pop a bio from
 444  * @tg_to_put: optional out argument for throtl_grp to put
 445  *
 446  * Pop the first bio from the qnode list @queued.  After popping, the first
 447  * qnode is removed from @queued if empty or moved to the end of @queued so
 448  * that the popping order is round-robin.
 449  *
 450  * When the first qnode is removed, its associated throtl_grp should be put
 451  * too.  If @tg_to_put is NULL, this function automatically puts it;
 452  * otherwise, *@tg_to_put is set to the throtl_grp to put and the caller is
 453  * responsible for putting it.
 454  */
 455 static struct bio *throtl_pop_queued(struct list_head *queued,
 456                                      struct throtl_grp **tg_to_put)
 457 {
 458         struct throtl_qnode *qn;
 459         struct bio *bio;
 460
 461         if (list_empty(queued))
 462                 return NULL;
 463
 464         qn = list_first_entry(queued, struct throtl_qnode, node);
 465         bio = bio_list_pop(&qn->bios);
 466         WARN_ON_ONCE(!bio);
 467
 468         if (bio_list_empty(&qn->bios)) {
 469                 list_del_init(&qn->node);
 470                 if (tg_to_put)
 471                         *tg_to_put = qn->tg;
 472                 else
 473                         blkg_put(tg_to_blkg(qn->tg));
 474         } else {
 475                 list_move_tail(&qn->node, queued);
 476         }
 477
 478         return bio;
 479 }
 480
 481 /* init a service_queue, assumes the caller zeroed it */
 482 static void throtl_service_queue_init(struct throtl_service_queue *sq)
 483 {
 484         INIT_LIST_HEAD(&sq->queued[0]);
 485         INIT_LIST_HEAD(&sq->queued[1]);
 486         sq->pending_tree = RB_ROOT_CACHED;
 487         timer_setup(&sq->pending_timer, throtl_pending_timer_fn, 0);
 488 }
 489
 490 static struct blkg_policy_data *throtl_pd_alloc(gfp_t gfp,
 491                                                 struct request_queue *q,
 492                                                 struct blkcg *blkcg)
 493 {
 494         struct throtl_grp *tg;
 495         int rw;
 496
 497         tg = kzalloc_node(sizeof(*tg), gfp, q->node);
 498         if (!tg)
 499                 return NULL;
 500
 501         if (blkg_rwstat_init(&tg->stat_bytes, gfp))
 502                 goto err_free_tg;
 503
 504         if (blkg_rwstat_init(&tg->stat_ios, gfp))
 505                 goto err_exit_stat_bytes;
 506
 507         throtl_service_queue_init(&tg->service_queue);
 508
 509         for (rw = READ; rw <= WRITE; rw++) {
 510                 throtl_qnode_init(&tg->qnode_on_self[rw], tg);
 511                 throtl_qnode_init(&tg->qnode_on_parent[rw], tg);
 512         }
 513
 514         RB_CLEAR_NODE(&tg->rb_node);
 515         tg->bps[READ][LIMIT_MAX] = U64_MAX;
 516         tg->bps[WRITE][LIMIT_MAX] = U64_MAX;
 517         tg->iops[READ][LIMIT_MAX] = UINT_MAX;
 518         tg->iops[WRITE][LIMIT_MAX] = UINT_MAX;
 519         tg->bps_conf[READ][LIMIT_MAX] = U64_MAX;
 520         tg->bps_conf[WRITE][LIMIT_MAX] = U64_MAX;
 521         tg->iops_conf[READ][LIMIT_MAX] = UINT_MAX;
 522         tg->iops_conf[WRITE][LIMIT_MAX] = UINT_MAX;
 523         /* LIMIT_LOW will have default value 0 */
 524
 525         tg->latency_target = DFL_LATENCY_TARGET;
 526         tg->latency_target_conf = DFL_LATENCY_TARGET;
 527         tg->idletime_threshold = DFL_IDLE_THRESHOLD;
 528         tg->idletime_threshold_conf = DFL_IDLE_THRESHOLD;
 529
 530         return &tg->pd;
 531
 532 err_exit_stat_bytes:
 533         blkg_rwstat_exit(&tg->stat_bytes);
 534 err_free_tg:
 535         kfree(tg);
 536         return NULL;
 537 }
 538
 539 static void throtl_pd_init(struct blkg_policy_data *pd)
 540 {
 541         struct throtl_grp *tg = pd_to_tg(pd);
 542         struct blkcg_gq *blkg = tg_to_blkg(tg);
 543         struct throtl_data *td = blkg->q->td;
 544         struct throtl_service_queue *sq = &tg->service_queue;
 545
 546         /*
 547          * If on the default hierarchy, we switch to properly hierarchical
 548          * behavior where limits on a given throtl_grp are applied to the
 549          * whole subtree rather than just the group itself.  e.g. If 16M
 550          * read_bps limit is set on the root group, the whole system can't
 551          * exceed 16M for the device.
 552          *
 553          * If not on the default hierarchy, the broken flat hierarchy
 554          * behavior is retained where all throtl_grps are treated as if
 555          * they're all separate root groups right below throtl_data.
 556          * Limits of a group don't interact with limits of other groups
 557          * regardless of the position of the group in the hierarchy.
 558          */
 559         sq->parent_sq = &td->service_queue;
 560         if (cgroup_subsys_on_dfl(io_cgrp_subsys) && blkg->parent)
 561                 sq->parent_sq = &blkg_to_tg(blkg->parent)->service_queue;
 562         tg->td = td;
 563 }
 564
 565 /*
 566  * Set has_rules[] if @tg or any of its parents have limits configured.
 567  * This doesn't require walking up to the top of the hierarchy as the
 568  * parent's has_rules[] is guaranteed to be correct.
 569  */
 570 static void tg_update_has_rules(struct throtl_grp *tg)
 571 {
 572         struct throtl_grp *parent_tg = sq_to_tg(tg->service_queue.parent_sq);
 573         struct throtl_data *td = tg->td;
 574         int rw;
 575
 576         for (rw = READ; rw <= WRITE; rw++)
 577                 tg->has_rules[rw] = (parent_tg && parent_tg->has_rules[rw]) ||
 578                         (td->limit_valid[td->limit_index] &&
 579                          (tg_bps_limit(tg, rw) != U64_MAX ||
 580                           tg_iops_limit(tg, rw) != UINT_MAX));
 581 }
 582
 583 static void throtl_pd_online(struct blkg_policy_data *pd)
 584 {
 585         struct throtl_grp *tg = pd_to_tg(pd);
 586         /*
 587          * We don't want new groups to escape the limits of its ancestors.
 588          * Update has_rules[] after a new group is brought online.
 589          */
 590         tg_update_has_rules(tg);
 591 }
 592
 593 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
 594 static void blk_throtl_update_limit_valid(struct throtl_data *td)
 595 {
 596         struct cgroup_subsys_state *pos_css;
 597         struct blkcg_gq *blkg;
 598         bool low_valid = false;
 599
 600         rcu_read_lock();
 601         blkg_for_each_descendant_post(blkg, pos_css, td->queue->root_blkg) {
 602                 struct throtl_grp *tg = blkg_to_tg(blkg);
 603
 604                 if (tg->bps[READ][LIMIT_LOW] || tg->bps[WRITE][LIMIT_LOW] ||
 605                     tg->iops[READ][LIMIT_LOW] || tg->iops[WRITE][LIMIT_LOW]) {
 606                         low_valid = true;
 607                         break;
 608                 }
 609         }
 610         rcu_read_unlock();
 611
 612         td->limit_valid[LIMIT_LOW] = low_valid;
 613 }
 614 #else
 615 static inline void blk_throtl_update_limit_valid(struct throtl_data *td)
 616 {
 617 }
 618 #endif
 619
 620 static void throtl_upgrade_state(struct throtl_data *td);
 621 static void throtl_pd_offline(struct blkg_policy_data *pd)
 622 {
 623         struct throtl_grp *tg = pd_to_tg(pd);
 624
 625         tg->bps[READ][LIMIT_LOW] = 0;
 626         tg->bps[WRITE][LIMIT_LOW] = 0;
 627         tg->iops[READ][LIMIT_LOW] = 0;
 628         tg->iops[WRITE][LIMIT_LOW] = 0;
 629
 630         blk_throtl_update_limit_valid(tg->td);
 631
 632         if (!tg->td->limit_valid[tg->td->limit_index])
 633                 throtl_upgrade_state(tg->td);
 634 }
 635
 636 static void throtl_pd_free(struct blkg_policy_data *pd)
 637 {
 638         struct throtl_grp *tg = pd_to_tg(pd);
 639
 640         del_timer_sync(&tg->service_queue.pending_timer);
 641         blkg_rwstat_exit(&tg->stat_bytes);
 642         blkg_rwstat_exit(&tg->stat_ios);
 643         kfree(tg);
 644 }
 645
 646 static struct throtl_grp *
 647 throtl_rb_first(struct throtl_service_queue *parent_sq)
 648 {
 649         struct rb_node *n;
 650
 651         n = rb_first_cached(&parent_sq->pending_tree);
 652         WARN_ON_ONCE(!n);
 653         if (!n)
 654                 return NULL;
 655         return rb_entry_tg(n);
 656 }
 657
 658 static void throtl_rb_erase(struct rb_node *n,
 659                             struct throtl_service_queue *parent_sq)
 660 {
 661         rb_erase_cached(n, &parent_sq->pending_tree);
 662         RB_CLEAR_NODE(n);
 663         --parent_sq->nr_pending;
 664 }
 665
 666 static void update_min_dispatch_time(struct throtl_service_queue *parent_sq)
 667 {
 668         struct throtl_grp *tg;
 669
 670         tg = throtl_rb_first(parent_sq);
 671         if (!tg)
 672                 return;
 673
 674         parent_sq->first_pending_disptime = tg->disptime;
 675 }
 676
 677 static void tg_service_queue_add(struct throtl_grp *tg)
 678 {
 679         struct throtl_service_queue *parent_sq = tg->service_queue.parent_sq;
 680         struct rb_node **node = &parent_sq->pending_tree.rb_root.rb_node;
 681         struct rb_node *parent = NULL;
 682         struct throtl_grp *__tg;
 683         unsigned long key = tg->disptime;
 684         bool leftmost = true;
 685
 686         while (*node != NULL) {
 687                 parent = *node;
 688                 __tg = rb_entry_tg(parent);
 689
 690                 if (time_before(key, __tg->disptime))
 691                         node = &parent->rb_left;
 692                 else {
 693                         node = &parent->rb_right;
 694                         leftmost = false;
 695                 }
 696         }
 697
 698         rb_link_node(&tg->rb_node, parent, node);
 699         rb_insert_color_cached(&tg->rb_node, &parent_sq->pending_tree,
 700                                leftmost);
 701 }
 702
 703 static void throtl_enqueue_tg(struct throtl_grp *tg)
 704 {
 705         if (!(tg->flags & THROTL_TG_PENDING)) {
 706                 tg_service_queue_add(tg);
 707                 tg->flags |= THROTL_TG_PENDING;
 708                 tg->service_queue.parent_sq->nr_pending++;
 709         }
 710 }
 711
 712 static void throtl_dequeue_tg(struct throtl_grp *tg)
 713 {
 714         if (tg->flags & THROTL_TG_PENDING) {
 715                 throtl_rb_erase(&tg->rb_node, tg->service_queue.parent_sq);
 716                 tg->flags &= ~THROTL_TG_PENDING;
 717         }
 718 }
 719
 720 /* Call with queue lock held */
 721 static void throtl_schedule_pending_timer(struct throtl_service_queue *sq,
 722                                           unsigned long expires)
 723 {
 724         unsigned long max_expire = jiffies + 8 * sq_to_td(sq)->throtl_slice;
 725
 726         /*
 727          * Since we are adjusting the throttle limit dynamically, the sleep
 728          * time calculated according to previous limit might be invalid. It's
 729          * possible the cgroup sleep time is very long and no other cgroups
 730          * have IO running so notify the limit changes. Make sure the cgroup
 731          * doesn't sleep too long to avoid the missed notification.
 732          */
 733         if (time_after(expires, max_expire))
 734                 expires = max_expire;
 735         mod_timer(&sq->pending_timer, expires);
 736         throtl_log(sq, "schedule timer. delay=%lu jiffies=%lu",
 737                    expires - jiffies, jiffies);
 738 }
 739
 740 /**
 741  * throtl_schedule_next_dispatch - schedule the next dispatch cycle
 742  * @sq: the service_queue to schedule dispatch for
 743  * @force: force scheduling
 744  *
 745  * Arm @sq->pending_timer so that the next dispatch cycle starts on the
 746  * dispatch time of the first pending child.  Returns %true if either timer
 747  * is armed or there's no pending child left.  %false if the current
 748  * dispatch window is still open and the caller should continue
 749  * dispatching.
 750  *
 751  * If @force is %true, the dispatch timer is always scheduled and this
 752  * function is guaranteed to return %true.  This is to be used when the
 753  * caller can't dispatch itself and needs to invoke pending_timer
 754  * unconditionally.  Note that forced scheduling is likely to induce short
 755  * delay before dispatch starts even if @sq->first_pending_disptime is not
 756  * in the future and thus shouldn't be used in hot paths.
 757  */
 758 static bool throtl_schedule_next_dispatch(struct throtl_service_queue *sq,
 759                                           bool force)
 760 {
 761         /* any pending children left? */
 762         if (!sq->nr_pending)
 763                 return true;
 764
 765         update_min_dispatch_time(sq);
 766
 767         /* is the next dispatch time in the future? */
 768         if (force || time_after(sq->first_pending_disptime, jiffies)) {
 769                 throtl_schedule_pending_timer(sq, sq->first_pending_disptime);
 770                 return true;
 771         }
 772
 773         /* tell the caller to continue dispatching */
 774         return false;
 775 }
 776
 777 static inline void throtl_start_new_slice_with_credit(struct throtl_grp *tg,
 778                 bool rw, unsigned long start)
 779 {
 780         tg->bytes_disp[rw] = 0;
 781         tg->io_disp[rw] = 0;
 782
 783         atomic_set(&tg->io_split_cnt[rw], 0);
 784
 785         /*
 786          * Previous slice has expired. We must have trimmed it after last
 787          * bio dispatch. That means since start of last slice, we never used
 788          * that bandwidth. Do try to make use of that bandwidth while giving
 789          * credit.
 790          */
 791         if (time_after_eq(start, tg->slice_start[rw]))
 792                 tg->slice_start[rw] = start;
 793
 794         tg->slice_end[rw] = jiffies + tg->td->throtl_slice;
 795         throtl_log(&tg->service_queue,
 796                    "[%c] new slice with credit start=%lu end=%lu jiffies=%lu",
 797                    rw == READ ? 'R' : 'W', tg->slice_start[rw],
 798                    tg->slice_end[rw], jiffies);
 799 }
 800
 801 static inline void throtl_start_new_slice(struct throtl_grp *tg, bool rw)
 802 {
 803         tg->bytes_disp[rw] = 0;
 804         tg->io_disp[rw] = 0;
 805         tg->slice_start[rw] = jiffies;
 806         tg->slice_end[rw] = jiffies + tg->td->throtl_slice;
 807
 808         atomic_set(&tg->io_split_cnt[rw], 0);
 809
 810         throtl_log(&tg->service_queue,
 811                    "[%c] new slice start=%lu end=%lu jiffies=%lu",
 812                    rw == READ ? 'R' : 'W', tg->slice_start[rw],
 813                    tg->slice_end[rw], jiffies);
 814 }
 815
 816 static inline void throtl_set_slice_end(struct throtl_grp *tg, bool rw,
 817                                         unsigned long jiffy_end)
 818 {
 819         tg->slice_end[rw] = roundup(jiffy_end, tg->td->throtl_slice);
 820 }
 821
 822 static inline void throtl_extend_slice(struct throtl_grp *tg, bool rw,
 823                                        unsigned long jiffy_end)
 824 {
 825         throtl_set_slice_end(tg, rw, jiffy_end);
 826         throtl_log(&tg->service_queue,
 827                    "[%c] extend slice start=%lu end=%lu jiffies=%lu",
 828                    rw == READ ? 'R' : 'W', tg->slice_start[rw],
 829                    tg->slice_end[rw], jiffies);
 830 }
 831
 832 /* Determine if previously allocated or extended slice is complete or not */
 833 static bool throtl_slice_used(struct throtl_grp *tg, bool rw)
 834 {
 835         if (time_in_range(jiffies, tg->slice_start[rw], tg->slice_end[rw]))
 836                 return false;
 837
 838         return true;
 839 }
 840
 841 /* Trim the used slices and adjust slice start accordingly */
 842 static inline void throtl_trim_slice(struct throtl_grp *tg, bool rw)
 843 {
 844         unsigned long nr_slices, time_elapsed, io_trim;
 845         u64 bytes_trim, tmp;
 846
 847         BUG_ON(time_before(tg->slice_end[rw], tg->slice_start[rw]));
 848
 849         /*
 850          * If bps are unlimited (-1), then time slice don't get
 851          * renewed. Don't try to trim the slice if slice is used. A new
 852          * slice will start when appropriate.
 853          */
 854         if (throtl_slice_used(tg, rw))
 855                 return;
 856
 857         /*
 858          * A bio has been dispatched. Also adjust slice_end. It might happen
 859          * that initially cgroup limit was very low resulting in high
 860          * slice_end, but later limit was bumped up and bio was dispatched
 861          * sooner, then we need to reduce slice_end. A high bogus slice_end
 862          * is bad because it does not allow new slice to start.
 863          */
 864
 865         throtl_set_slice_end(tg, rw, jiffies + tg->td->throtl_slice);
 866
 867         time_elapsed = jiffies - tg->slice_start[rw];
 868
 869         nr_slices = time_elapsed / tg->td->throtl_slice;
 870
 871         if (!nr_slices)
 872                 return;
 873         tmp = tg_bps_limit(tg, rw) * tg->td->throtl_slice * nr_slices;
 874         do_div(tmp, HZ);
 875         bytes_trim = tmp;
 876
 877         io_trim = (tg_iops_limit(tg, rw) * tg->td->throtl_slice * nr_slices) /
 878                 HZ;
 879
 880         if (!bytes_trim && !io_trim)
 881                 return;
 882
 883         if (tg->bytes_disp[rw] >= bytes_trim)
 884                 tg->bytes_disp[rw] -= bytes_trim;
 885         else
 886                 tg->bytes_disp[rw] = 0;
 887
 888         if (tg->io_disp[rw] >= io_trim)
 889                 tg->io_disp[rw] -= io_trim;
 890         else
 891                 tg->io_disp[rw] = 0;
 892
 893         tg->slice_start[rw] += nr_slices * tg->td->throtl_slice;
 894
 895         throtl_log(&tg->service_queue,
 896                    "[%c] trim slice nr=%lu bytes=%llu io=%lu start=%lu end=%lu jiffies=%lu",
 897                    rw == READ ? 'R' : 'W', nr_slices, bytes_trim, io_trim,
 898                    tg->slice_start[rw], tg->slice_end[rw], jiffies);
 899 }
 900
 901 static bool tg_with_in_iops_limit(struct throtl_grp *tg, struct bio *bio,
 902                                   u32 iops_limit, unsigned long *wait)
 903 {
 904         bool rw = bio_data_dir(bio);
 905         unsigned int io_allowed;
 906         unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
 907         u64 tmp;
 908
 909         if (iops_limit == UINT_MAX) {
 910                 if (wait)
 911                         *wait = 0;
 912                 return true;
 913         }
 914
 915         jiffy_elapsed = jiffies - tg->slice_start[rw];
 916
 917         /* Round up to the next throttle slice, wait time must be nonzero */
 918         jiffy_elapsed_rnd = roundup(jiffy_elapsed + 1, tg->td->throtl_slice);
 919
 920         /*
 921          * jiffy_elapsed_rnd should not be a big value as minimum iops can be
 922          * 1 then at max jiffy elapsed should be equivalent of 1 second as we
 923          * will allow dispatch after 1 second and after that slice should
 924          * have been trimmed.
 925          */
 926
 927         tmp = (u64)iops_limit * jiffy_elapsed_rnd;
 928         do_div(tmp, HZ);
 929
 930         if (tmp > UINT_MAX)
 931                 io_allowed = UINT_MAX;
 932         else
 933                 io_allowed = tmp;
 934
 935         if (tg->io_disp[rw] + 1 <= io_allowed) {
 936                 if (wait)
 937                         *wait = 0;
 938                 return true;
 939         }
 940
 941         /* Calc approx time to dispatch */
 942         jiffy_wait = jiffy_elapsed_rnd - jiffy_elapsed;
 943
 944         if (wait)
 945                 *wait = jiffy_wait;
 946         return false;
 947 }
 948
 949 static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
 950                                  u64 bps_limit, unsigned long *wait)
 951 {
 952         bool rw = bio_data_dir(bio);
 953         u64 bytes_allowed, extra_bytes, tmp;
 954         unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
 955         unsigned int bio_size = throtl_bio_data_size(bio);
 956
 957         if (bps_limit == U64_MAX) {
 958                 if (wait)
 959                         *wait = 0;
 960                 return true;
 961         }
 962
 963         jiffy_elapsed = jiffy_elapsed_rnd = jiffies - tg->slice_start[rw];
 964
 965         /* Slice has just started. Consider one slice interval */
 966         if (!jiffy_elapsed)
 967                 jiffy_elapsed_rnd = tg->td->throtl_slice;
 968
 969         jiffy_elapsed_rnd = roundup(jiffy_elapsed_rnd, tg->td->throtl_slice);
 970
 971         tmp = bps_limit * jiffy_elapsed_rnd;
 972         do_div(tmp, HZ);
 973         bytes_allowed = tmp;
 974
 975         if (tg->bytes_disp[rw] + bio_size <= bytes_allowed) {
 976                 if (wait)
 977                         *wait = 0;
 978                 return true;
 979         }
 980
 981         /* Calc approx time to dispatch */
 982         extra_bytes = tg->bytes_disp[rw] + bio_size - bytes_allowed;
 983         jiffy_wait = div64_u64(extra_bytes * HZ, bps_limit);
 984
 985         if (!jiffy_wait)
 986                 jiffy_wait = 1;
 987
 988         /*
 989          * This wait time is without taking into consideration the rounding
 990          * up we did. Add that time also.
 991          */
 992         jiffy_wait = jiffy_wait + (jiffy_elapsed_rnd - jiffy_elapsed);
 993         if (wait)
 994                 *wait = jiffy_wait;
 995         return false;
 996 }
 997
 998 /*
 999  * Returns whether one can dispatch a bio or not. Also returns approx number
1000  * of jiffies to wait before this bio is with-in IO rate and can be dispatched
1001  */
1002 static bool tg_may_dispatch(struct throtl_grp *tg, struct bio *bio,
1003                             unsigned long *wait)
1004 {
1005         bool rw = bio_data_dir(bio);
1006         unsigned long bps_wait = 0, iops_wait = 0, max_wait = 0;
1007         u64 bps_limit = tg_bps_limit(tg, rw);
1008         u32 iops_limit = tg_iops_limit(tg, rw);
1009
1010         /*
1011          * Currently whole state machine of group depends on first bio
1012          * queued in the group bio list. So one should not be calling
1013          * this function with a different bio if there are other bios
1014          * queued.
1015          */
1016         BUG_ON(tg->service_queue.nr_queued[rw] &&
1017                bio != throtl_peek_queued(&tg->service_queue.queued[rw]));
1018
1019         /* If tg->bps = -1, then BW is unlimited */
1020         if (bps_limit == U64_MAX && iops_limit == UINT_MAX) {
1021                 if (wait)
1022                         *wait = 0;
1023                 return true;
1024         }
1025
1026         /*
1027          * If previous slice expired, start a new one otherwise renew/extend
1028          * existing slice to make sure it is at least throtl_slice interval
1029          * long since now. New slice is started only for empty throttle group.
1030          * If there is queued bio, that means there should be an active
1031          * slice and it should be extended instead.
1032          */
1033         if (throtl_slice_used(tg, rw) && !(tg->service_queue.nr_queued[rw]))
1034                 throtl_start_new_slice(tg, rw);
1035         else {
1036                 if (time_before(tg->slice_end[rw],
1037                     jiffies + tg->td->throtl_slice))
1038                         throtl_extend_slice(tg, rw,
1039                                 jiffies + tg->td->throtl_slice);
1040         }
1041
1042         if (iops_limit != UINT_MAX)
1043                 tg->io_disp[rw] += atomic_xchg(&tg->io_split_cnt[rw], 0);
1044
1045         if (tg_with_in_bps_limit(tg, bio, bps_limit, &bps_wait) &&
1046             tg_with_in_iops_limit(tg, bio, iops_limit, &iops_wait)) {
1047                 if (wait)
1048                         *wait = 0;
1049                 return true;
1050         }
1051
1052         max_wait = max(bps_wait, iops_wait);
1053
1054         if (wait)
1055                 *wait = max_wait;
1056
1057         if (time_before(tg->slice_end[rw], jiffies + max_wait))
1058                 throtl_extend_slice(tg, rw, jiffies + max_wait);
1059
1060         return false;
1061 }
1062
1063 static void throtl_charge_bio(struct throtl_grp *tg, struct bio *bio)
1064 {
1065         bool rw = bio_data_dir(bio);
1066         unsigned int bio_size = throtl_bio_data_size(bio);
1067
1068         /* Charge the bio to the group */
1069         tg->bytes_disp[rw] += bio_size;
1070         tg->io_disp[rw]++;
1071         tg->last_bytes_disp[rw] += bio_size;
1072         tg->last_io_disp[rw]++;
1073
1074         /*
1075          * BIO_THROTTLED is used to prevent the same bio to be throttled
1076          * more than once as a throttled bio will go through blk-throtl the
1077          * second time when it eventually gets issued.  Set it when a bio
1078          * is being charged to a tg.
1079          */
1080         if (!bio_flagged(bio, BIO_THROTTLED))
1081                 bio_set_flag(bio, BIO_THROTTLED);
1082 }
1083
1084 /**
1085  * throtl_add_bio_tg - add a bio to the specified throtl_grp
1086  * @bio: bio to add
1087  * @qn: qnode to use
1088  * @tg: the target throtl_grp
1089  *
1090  * Add @bio to @tg's service_queue using @qn.  If @qn is not specified,
1091  * tg->qnode_on_self[] is used.
1092  */
1093 static void throtl_add_bio_tg(struct bio *bio, struct throtl_qnode *qn,
1094                               struct throtl_grp *tg)
1095 {
1096         struct throtl_service_queue *sq = &tg->service_queue;
1097         bool rw = bio_data_dir(bio);
1098
1099         if (!qn)
1100                 qn = &tg->qnode_on_self[rw];
1101
1102         /*
1103          * If @tg doesn't currently have any bios queued in the same
1104          * direction, queueing @bio can change when @tg should be
1105          * dispatched.  Mark that @tg was empty.  This is automatically
1106          * cleared on the next tg_update_disptime().
1107          */
1108         if (!sq->nr_queued[rw])
1109                 tg->flags |= THROTL_TG_WAS_EMPTY;
1110
1111         throtl_qnode_add_bio(bio, qn, &sq->queued[rw]);
1112
1113         sq->nr_queued[rw]++;
1114         throtl_enqueue_tg(tg);
1115 }
1116
1117 static void tg_update_disptime(struct throtl_grp *tg)
1118 {
1119         struct throtl_service_queue *sq = &tg->service_queue;
1120         unsigned long read_wait = -1, write_wait = -1, min_wait = -1, disptime;
1121         struct bio *bio;
1122
1123         bio = throtl_peek_queued(&sq->queued[READ]);
1124         if (bio)
1125                 tg_may_dispatch(tg, bio, &read_wait);
1126
1127         bio = throtl_peek_queued(&sq->queued[WRITE]);
1128         if (bio)
1129                 tg_may_dispatch(tg, bio, &write_wait);
1130
1131         min_wait = min(read_wait, write_wait);
1132         disptime = jiffies + min_wait;
1133
1134         /* Update dispatch time */
1135         throtl_dequeue_tg(tg);
1136         tg->disptime = disptime;
1137         throtl_enqueue_tg(tg);
1138
1139         /* see throtl_add_bio_tg() */
1140         tg->flags &= ~THROTL_TG_WAS_EMPTY;
1141 }
1142
1143 static void start_parent_slice_with_credit(struct throtl_grp *child_tg,
1144                                         struct throtl_grp *parent_tg, bool rw)
1145 {
1146         if (throtl_slice_used(parent_tg, rw)) {
1147                 throtl_start_new_slice_with_credit(parent_tg, rw,
1148                                 child_tg->slice_start[rw]);
1149         }
1150
1151 }
1152
1153 static void tg_dispatch_one_bio(struct throtl_grp *tg, bool rw)
1154 {
1155         struct throtl_service_queue *sq = &tg->service_queue;
1156         struct throtl_service_queue *parent_sq = sq->parent_sq;
1157         struct throtl_grp *parent_tg = sq_to_tg(parent_sq);
1158         struct throtl_grp *tg_to_put = NULL;
1159         struct bio *bio;
1160
1161         /*
1162          * @bio is being transferred from @tg to @parent_sq.  Popping a bio
1163          * from @tg may put its reference and @parent_sq might end up
1164          * getting released prematurely.  Remember the tg to put and put it
1165          * after @bio is transferred to @parent_sq.
1166          */
1167         bio = throtl_pop_queued(&sq->queued[rw], &tg_to_put);
1168         sq->nr_queued[rw]--;
1169
1170         throtl_charge_bio(tg, bio);
1171
1172         /*
1173          * If our parent is another tg, we just need to transfer @bio to
1174          * the parent using throtl_add_bio_tg().  If our parent is
1175          * @td->service_queue, @bio is ready to be issued.  Put it on its
1176          * bio_lists[] and decrease total number queued.  The caller is
1177          * responsible for issuing these bios.
1178          */
1179         if (parent_tg) {
1180                 throtl_add_bio_tg(bio, &tg->qnode_on_parent[rw], parent_tg);
1181                 start_parent_slice_with_credit(tg, parent_tg, rw);
1182         } else {
1183                 throtl_qnode_add_bio(bio, &tg->qnode_on_parent[rw],
1184                                      &parent_sq->queued[rw]);
1185                 BUG_ON(tg->td->nr_queued[rw] <= 0);
1186                 tg->td->nr_queued[rw]--;
1187         }
1188
1189         throtl_trim_slice(tg, rw);
1190
1191         if (tg_to_put)
1192                 blkg_put(tg_to_blkg(tg_to_put));
1193 }
1194
1195 static int throtl_dispatch_tg(struct throtl_grp *tg)
1196 {
1197         struct throtl_service_queue *sq = &tg->service_queue;
1198         unsigned int nr_reads = 0, nr_writes = 0;
1199         unsigned int max_nr_reads = THROTL_GRP_QUANTUM * 3 / 4;
1200         unsigned int max_nr_writes = THROTL_GRP_QUANTUM - max_nr_reads;
1201         struct bio *bio;
1202
1203         /* Try to dispatch 75% READS and 25% WRITES */
1204
1205         while ((bio = throtl_peek_queued(&sq->queued[READ])) &&
1206                tg_may_dispatch(tg, bio, NULL)) {
1207
1208                 tg_dispatch_one_bio(tg, bio_data_dir(bio));
1209                 nr_reads++;
1210
1211                 if (nr_reads >= max_nr_reads)
1212                         break;
1213         }
1214
1215         while ((bio = throtl_peek_queued(&sq->queued[WRITE])) &&
1216                tg_may_dispatch(tg, bio, NULL)) {
1217
1218                 tg_dispatch_one_bio(tg, bio_data_dir(bio));
1219                 nr_writes++;
1220
1221                 if (nr_writes >= max_nr_writes)
1222                         break;
1223         }
1224
1225         return nr_reads + nr_writes;
1226 }
1227
1228 static int throtl_select_dispatch(struct throtl_service_queue *parent_sq)
1229 {
1230         unsigned int nr_disp = 0;
1231
1232         while (1) {
1233                 struct throtl_grp *tg;
1234                 struct throtl_service_queue *sq;
1235
1236                 if (!parent_sq->nr_pending)
1237                         break;
1238
1239                 tg = throtl_rb_first(parent_sq);
1240                 if (!tg)
1241                         break;
1242
1243                 if (time_before(jiffies, tg->disptime))
1244                         break;
1245
1246                 throtl_dequeue_tg(tg);
1247
1248                 nr_disp += throtl_dispatch_tg(tg);
1249
1250                 sq = &tg->service_queue;
1251                 if (sq->nr_queued[0] || sq->nr_queued[1])
1252                         tg_update_disptime(tg);
1253
1254                 if (nr_disp >= THROTL_QUANTUM)
1255                         break;
1256         }
1257
1258         return nr_disp;
1259 }
1260
1261 static bool throtl_can_upgrade(struct throtl_data *td,
1262         struct throtl_grp *this_tg);
1263 /**
1264  * throtl_pending_timer_fn - timer function for service_queue->pending_timer
1265  * @t: the pending_timer member of the throtl_service_queue being serviced
1266  *
1267  * This timer is armed when a child throtl_grp with active bio's become
1268  * pending and queued on the service_queue's pending_tree and expires when
1269  * the first child throtl_grp should be dispatched.  This function
1270  * dispatches bio's from the children throtl_grps to the parent
1271  * service_queue.
1272  *
1273  * If the parent's parent is another throtl_grp, dispatching is propagated
1274  * by either arming its pending_timer or repeating dispatch directly.  If
1275  * the top-level service_tree is reached, throtl_data->dispatch_work is
1276  * kicked so that the ready bio's are issued.
1277  */
1278 static void throtl_pending_timer_fn(struct timer_list *t)
1279 {
1280         struct throtl_service_queue *sq = from_timer(sq, t, pending_timer);
1281         struct throtl_grp *tg = sq_to_tg(sq);
1282         struct throtl_data *td = sq_to_td(sq);
1283         struct request_queue *q = td->queue;
1284         struct throtl_service_queue *parent_sq;
1285         bool dispatched;
1286         int ret;
1287
1288         spin_lock_irq(&q->queue_lock);
1289         if (throtl_can_upgrade(td, NULL))
1290                 throtl_upgrade_state(td);
1291
1292 again:
1293         parent_sq = sq->parent_sq;
1294         dispatched = false;
1295
1296         while (true) {
1297                 throtl_log(sq, "dispatch nr_queued=%u read=%u write=%u",
1298                            sq->nr_queued[READ] + sq->nr_queued[WRITE],
1299                            sq->nr_queued[READ], sq->nr_queued[WRITE]);
1300
1301                 ret = throtl_select_dispatch(sq);
1302                 if (ret) {
1303                         throtl_log(sq, "bios disp=%u", ret);
1304                         dispatched = true;
1305                 }
1306
1307                 if (throtl_schedule_next_dispatch(sq, false))
1308                         break;
1309
1310                 /* this dispatch windows is still open, relax and repeat */
1311                 spin_unlock_irq(&q->queue_lock);
1312                 cpu_relax();
1313                 spin_lock_irq(&q->queue_lock);
1314         }
1315
1316         if (!dispatched)
1317                 goto out_unlock;
1318
1319         if (parent_sq) {
1320                 /* @parent_sq is another throl_grp, propagate dispatch */
1321                 if (tg->flags & THROTL_TG_WAS_EMPTY) {
1322                         tg_update_disptime(tg);
1323                         if (!throtl_schedule_next_dispatch(parent_sq, false)) {
1324                                 /* window is already open, repeat dispatching */
1325                                 sq = parent_sq;
1326                                 tg = sq_to_tg(sq);
1327                                 goto again;
1328                         }
1329                 }
1330         } else {
1331                 /* reached the top-level, queue issuing */
1332                 queue_work(kthrotld_workqueue, &td->dispatch_work);
1333         }
1334 out_unlock:
1335         spin_unlock_irq(&q->queue_lock);
1336 }
1337
1338 /**
1339  * blk_throtl_dispatch_work_fn - work function for throtl_data->dispatch_work
1340  * @work: work item being executed
1341  *
1342  * This function is queued for execution when bios reach the bio_lists[]
1343  * of throtl_data->service_queue.  Those bios are ready and issued by this
1344  * function.
1345  */
1346 static void blk_throtl_dispatch_work_fn(struct work_struct *work)
1347 {
1348         struct throtl_data *td = container_of(work, struct throtl_data,
1349                                               dispatch_work);
1350         struct throtl_service_queue *td_sq = &td->service_queue;
1351         struct request_queue *q = td->queue;
1352         struct bio_list bio_list_on_stack;
1353         struct bio *bio;
1354         struct blk_plug plug;
1355         int rw;
1356
1357         bio_list_init(&bio_list_on_stack);
1358
1359         spin_lock_irq(&q->queue_lock);
1360         for (rw = READ; rw <= WRITE; rw++)
1361                 while ((bio = throtl_pop_queued(&td_sq->queued[rw], NULL)))
1362                         bio_list_add(&bio_list_on_stack, bio);
1363         spin_unlock_irq(&q->queue_lock);
1364
1365         if (!bio_list_empty(&bio_list_on_stack)) {
1366                 blk_start_plug(&plug);
1367                 while ((bio = bio_list_pop(&bio_list_on_stack)))
1368                         submit_bio_noacct(bio);
1369                 blk_finish_plug(&plug);
1370         }
1371 }
1372
1373 static u64 tg_prfill_conf_u64(struct seq_file *sf, struct blkg_policy_data *pd,
1374                               int off)
1375 {
1376         struct throtl_grp *tg = pd_to_tg(pd);
1377         u64 v = *(u64 *)((void *)tg + off);
1378
1379         if (v == U64_MAX)
1380                 return 0;
1381         return __blkg_prfill_u64(sf, pd, v);
1382 }
1383
1384 static u64 tg_prfill_conf_uint(struct seq_file *sf, struct blkg_policy_data *pd,
1385                                int off)
1386 {
1387         struct throtl_grp *tg = pd_to_tg(pd);
1388         unsigned int v = *(unsigned int *)((void *)tg + off);
1389
1390         if (v == UINT_MAX)
1391                 return 0;
1392         return __blkg_prfill_u64(sf, pd, v);
1393 }
1394
1395 static int tg_print_conf_u64(struct seq_file *sf, void *v)
1396 {
1397         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), tg_prfill_conf_u64,
1398                           &blkcg_policy_throtl, seq_cft(sf)->private, false);
1399         return 0;
1400 }
1401
1402 static int tg_print_conf_uint(struct seq_file *sf, void *v)
1403 {
1404         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), tg_prfill_conf_uint,
1405                           &blkcg_policy_throtl, seq_cft(sf)->private, false);
1406         return 0;
1407 }
1408
1409 static void tg_conf_updated(struct throtl_grp *tg, bool global)
1410 {
1411         struct throtl_service_queue *sq = &tg->service_queue;
1412         struct cgroup_subsys_state *pos_css;
1413         struct blkcg_gq *blkg;
1414
1415         throtl_log(&tg->service_queue,
1416                    "limit change rbps=%llu wbps=%llu riops=%u wiops=%u",
1417                    tg_bps_limit(tg, READ), tg_bps_limit(tg, WRITE),
1418                    tg_iops_limit(tg, READ), tg_iops_limit(tg, WRITE));
1419
1420         /*
1421          * Update has_rules[] flags for the updated tg's subtree.  A tg is
1422          * considered to have rules if either the tg itself or any of its
1423          * ancestors has rules.  This identifies groups without any
1424          * restrictions in the whole hierarchy and allows them to bypass
1425          * blk-throttle.
1426          */
1427         blkg_for_each_descendant_pre(blkg, pos_css,
1428                         global ? tg->td->queue->root_blkg : tg_to_blkg(tg)) {
1429                 struct throtl_grp *this_tg = blkg_to_tg(blkg);
1430                 struct throtl_grp *parent_tg;
1431
1432                 tg_update_has_rules(this_tg);
1433                 /* ignore root/second level */
1434                 if (!cgroup_subsys_on_dfl(io_cgrp_subsys) || !blkg->parent ||
1435                     !blkg->parent->parent)
1436                         continue;
1437                 parent_tg = blkg_to_tg(blkg->parent);
1438                 /*
1439                  * make sure all children has lower idle time threshold and
1440                  * higher latency target
1441                  */
1442                 this_tg->idletime_threshold = min(this_tg->idletime_threshold,
1443                                 parent_tg->idletime_threshold);
1444                 this_tg->latency_target = max(this_tg->latency_target,
1445                                 parent_tg->latency_target);
1446         }
1447
1448         /*
1449          * We're already holding queue_lock and know @tg is valid.  Let's
1450          * apply the new config directly.
1451          *
1452          * Restart the slices for both READ and WRITES. It might happen
1453          * that a group's limit are dropped suddenly and we don't want to
1454          * account recently dispatched IO with new low rate.
1455          */
1456         throtl_start_new_slice(tg, READ);
1457         throtl_start_new_slice(tg, WRITE);
1458
1459         if (tg->flags & THROTL_TG_PENDING) {
1460                 tg_update_disptime(tg);
1461                 throtl_schedule_next_dispatch(sq->parent_sq, true);
1462         }
1463 }
1464
1465 static ssize_t tg_set_conf(struct kernfs_open_file *of,
1466                            char *buf, size_t nbytes, loff_t off, bool is_u64)
1467 {
1468         struct blkcg *blkcg = css_to_blkcg(of_css(of));
1469         struct blkg_conf_ctx ctx;
1470         struct throtl_grp *tg;
1471         int ret;
1472         u64 v;
1473
1474         ret = blkg_conf_prep(blkcg, &blkcg_policy_throtl, buf, &ctx);
1475         if (ret)
1476                 return ret;
1477
1478         ret = -EINVAL;
1479         if (sscanf(ctx.body, "%llu", &v) != 1)
1480                 goto out_finish;
1481         if (!v)
1482                 v = U64_MAX;
1483
1484         tg = blkg_to_tg(ctx.blkg);
1485
1486         if (is_u64)
1487                 *(u64 *)((void *)tg + of_cft(of)->private) = v;
1488         else
1489                 *(unsigned int *)((void *)tg + of_cft(of)->private) = v;
1490
1491         tg_conf_updated(tg, false);
1492         ret = 0;
1493 out_finish:
1494         blkg_conf_finish(&ctx);
1495         return ret ?: nbytes;
1496 }
1497
1498 static ssize_t tg_set_conf_u64(struct kernfs_open_file *of,
1499                                char *buf, size_t nbytes, loff_t off)
1500 {
1501         return tg_set_conf(of, buf, nbytes, off, true);
1502 }
1503
1504 static ssize_t tg_set_conf_uint(struct kernfs_open_file *of,
1505                                 char *buf, size_t nbytes, loff_t off)
1506 {
1507         return tg_set_conf(of, buf, nbytes, off, false);
1508 }
1509
1510 static int tg_print_rwstat(struct seq_file *sf, void *v)
1511 {
1512         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1513                           blkg_prfill_rwstat, &blkcg_policy_throtl,
1514                           seq_cft(sf)->private, true);
1515         return 0;
1516 }
1517
1518 static u64 tg_prfill_rwstat_recursive(struct seq_file *sf,
1519                                       struct blkg_policy_data *pd, int off)
1520 {
1521         struct blkg_rwstat_sample sum;
1522
1523         blkg_rwstat_recursive_sum(pd_to_blkg(pd), &blkcg_policy_throtl, off,
1524                                   &sum);
1525         return __blkg_prfill_rwstat(sf, pd, &sum);
1526 }
1527
1528 static int tg_print_rwstat_recursive(struct seq_file *sf, void *v)
1529 {
1530         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1531                           tg_prfill_rwstat_recursive, &blkcg_policy_throtl,
1532                           seq_cft(sf)->private, true);
1533         return 0;
1534 }
1535
1536 static struct cftype throtl_legacy_files[] = {
1537         {
1538                 .name = "throttle.read_bps_device",
1539                 .private = offsetof(struct throtl_grp, bps[READ][LIMIT_MAX]),
1540                 .seq_show = tg_print_conf_u64,
1541                 .write = tg_set_conf_u64,
1542         },
1543         {
1544                 .name = "throttle.write_bps_device",
1545                 .private = offsetof(struct throtl_grp, bps[WRITE][LIMIT_MAX]),
1546                 .seq_show = tg_print_conf_u64,
1547                 .write = tg_set_conf_u64,
1548         },
1549         {
1550                 .name = "throttle.read_iops_device",
1551                 .private = offsetof(struct throtl_grp, iops[READ][LIMIT_MAX]),
1552                 .seq_show = tg_print_conf_uint,
1553                 .write = tg_set_conf_uint,
1554         },
1555         {
1556                 .name = "throttle.write_iops_device",
1557                 .private = offsetof(struct throtl_grp, iops[WRITE][LIMIT_MAX]),
1558                 .seq_show = tg_print_conf_uint,
1559                 .write = tg_set_conf_uint,
1560         },
1561         {
1562                 .name = "throttle.io_service_bytes",
1563                 .private = offsetof(struct throtl_grp, stat_bytes),
1564                 .seq_show = tg_print_rwstat,
1565         },
1566         {
1567                 .name = "throttle.io_service_bytes_recursive",
1568                 .private = offsetof(struct throtl_grp, stat_bytes),
1569                 .seq_show = tg_print_rwstat_recursive,
1570         },
1571         {
1572                 .name = "throttle.io_serviced",
1573                 .private = offsetof(struct throtl_grp, stat_ios),
1574                 .seq_show = tg_print_rwstat,
1575         },
1576         {
1577                 .name = "throttle.io_serviced_recursive",
1578                 .private = offsetof(struct throtl_grp, stat_ios),
1579                 .seq_show = tg_print_rwstat_recursive,
1580         },
1581         { }     /* terminate */
1582 };
1583
1584 static u64 tg_prfill_limit(struct seq_file *sf, struct blkg_policy_data *pd,
1585                          int off)
1586 {
1587         struct throtl_grp *tg = pd_to_tg(pd);
1588         const char *dname = blkg_dev_name(pd->blkg);
1589         char bufs[4][21] = { "max", "max", "max", "max" };
1590         u64 bps_dft;
1591         unsigned int iops_dft;
1592         char idle_time[26] = "";
1593         char latency_time[26] = "";
1594
1595         if (!dname)
1596                 return 0;
1597
1598         if (off == LIMIT_LOW) {
1599                 bps_dft = 0;
1600                 iops_dft = 0;
1601         } else {
1602                 bps_dft = U64_MAX;
1603                 iops_dft = UINT_MAX;
1604         }
1605
1606         if (tg->bps_conf[READ][off] == bps_dft &&
1607             tg->bps_conf[WRITE][off] == bps_dft &&
1608             tg->iops_conf[READ][off] == iops_dft &&
1609             tg->iops_conf[WRITE][off] == iops_dft &&
1610             (off != LIMIT_LOW ||
1611              (tg->idletime_threshold_conf == DFL_IDLE_THRESHOLD &&
1612               tg->latency_target_conf == DFL_LATENCY_TARGET)))
1613                 return 0;
1614
1615         if (tg->bps_conf[READ][off] != U64_MAX)
1616                 snprintf(bufs[0], sizeof(bufs[0]), "%llu",
1617                         tg->bps_conf[READ][off]);
1618         if (tg->bps_conf[WRITE][off] != U64_MAX)
1619                 snprintf(bufs[1], sizeof(bufs[1]), "%llu",
1620                         tg->bps_conf[WRITE][off]);
1621         if (tg->iops_conf[READ][off] != UINT_MAX)
1622                 snprintf(bufs[2], sizeof(bufs[2]), "%u",
1623                         tg->iops_conf[READ][off]);
1624         if (tg->iops_conf[WRITE][off] != UINT_MAX)
1625                 snprintf(bufs[3], sizeof(bufs[3]), "%u",
1626                         tg->iops_conf[WRITE][off]);
1627         if (off == LIMIT_LOW) {
1628                 if (tg->idletime_threshold_conf == ULONG_MAX)
1629                         strcpy(idle_time, " idle=max");
1630                 else
1631                         snprintf(idle_time, sizeof(idle_time), " idle=%lu",
1632                                 tg->idletime_threshold_conf);
1633
1634                 if (tg->latency_target_conf == ULONG_MAX)
1635                         strcpy(latency_time, " latency=max");
1636                 else
1637                         snprintf(latency_time, sizeof(latency_time),
1638                                 " latency=%lu", tg->latency_target_conf);
1639         }
1640
1641         seq_printf(sf, "%s rbps=%s wbps=%s riops=%s wiops=%s%s%s\n",
1642                    dname, bufs[0], bufs[1], bufs[2], bufs[3], idle_time,
1643                    latency_time);
1644         return 0;
1645 }
1646
1647 static int tg_print_limit(struct seq_file *sf, void *v)
1648 {
1649         blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), tg_prfill_limit,
1650                           &blkcg_policy_throtl, seq_cft(sf)->private, false);
1651         return 0;
1652 }
1653
1654 static ssize_t tg_set_limit(struct kernfs_open_file *of,
1655                           char *buf, size_t nbytes, loff_t off)
1656 {
1657         struct blkcg *blkcg = css_to_blkcg(of_css(of));
1658         struct blkg_conf_ctx ctx;
1659         struct throtl_grp *tg;
1660         u64 v[4];
1661         unsigned long idle_time;
1662         unsigned long latency_time;
1663         int ret;
1664         int index = of_cft(of)->private;
1665
1666         ret = blkg_conf_prep(blkcg, &blkcg_policy_throtl, buf, &ctx);
1667         if (ret)
1668                 return ret;
1669
1670         tg = blkg_to_tg(ctx.blkg);
1671
1672         v[0] = tg->bps_conf[READ][index];
1673         v[1] = tg->bps_conf[WRITE][index];
1674         v[2] = tg->iops_conf[READ][index];
1675         v[3] = tg->iops_conf[WRITE][index];
1676
1677         idle_time = tg->idletime_threshold_conf;
1678         latency_time = tg->latency_target_conf;
1679         while (true) {
1680                 char tok[27];   /* wiops=18446744073709551616 */
1681                 char *p;
1682                 u64 val = U64_MAX;
1683                 int len;
1684
1685                 if (sscanf(ctx.body, "%26s%n", tok, &len) != 1)
1686                         break;
1687                 if (tok[0] == '\0')
1688                         break;
1689                 ctx.body += len;
1690
1691                 ret = -EINVAL;
1692                 p = tok;
1693                 strsep(&p, "=");
1694                 if (!p || (sscanf(p, "%llu", &val) != 1 && strcmp(p, "max")))
1695                         goto out_finish;
1696
1697                 ret = -ERANGE;
1698                 if (!val)
1699                         goto out_finish;
1700
1701                 ret = -EINVAL;
1702                 if (!strcmp(tok, "rbps") && val > 1)
1703                         v[0] = val;
1704                 else if (!strcmp(tok, "wbps") && val > 1)
1705                         v[1] = val;
1706                 else if (!strcmp(tok, "riops") && val > 1)
1707                         v[2] = min_t(u64, val, UINT_MAX);
1708                 else if (!strcmp(tok, "wiops") && val > 1)
1709                         v[3] = min_t(u64, val, UINT_MAX);
1710                 else if (off == LIMIT_LOW && !strcmp(tok, "idle"))
1711                         idle_time = val;
1712                 else if (off == LIMIT_LOW && !strcmp(tok, "latency"))
1713                         latency_time = val;
1714                 else
1715                         goto out_finish;
1716         }
1717
1718         tg->bps_conf[READ][index] = v[0];
1719         tg->bps_conf[WRITE][index] = v[1];
1720         tg->iops_conf[READ][index] = v[2];
1721         tg->iops_conf[WRITE][index] = v[3];
1722
1723         if (index == LIMIT_MAX) {
1724                 tg->bps[READ][index] = v[0];
1725                 tg->bps[WRITE][index] = v[1];
1726                 tg->iops[READ][index] = v[2];
1727                 tg->iops[WRITE][index] = v[3];
1728         }
1729         tg->bps[READ][LIMIT_LOW] = min(tg->bps_conf[READ][LIMIT_LOW],
1730                 tg->bps_conf[READ][LIMIT_MAX]);
1731         tg->bps[WRITE][LIMIT_LOW] = min(tg->bps_conf[WRITE][LIMIT_LOW],
1732                 tg->bps_conf[WRITE][LIMIT_MAX]);
1733         tg->iops[READ][LIMIT_LOW] = min(tg->iops_conf[READ][LIMIT_LOW],
1734                 tg->iops_conf[READ][LIMIT_MAX]);
1735         tg->iops[WRITE][LIMIT_LOW] = min(tg->iops_conf[WRITE][LIMIT_LOW],
1736                 tg->iops_conf[WRITE][LIMIT_MAX]);
1737         tg->idletime_threshold_conf = idle_time;
1738         tg->latency_target_conf = latency_time;
1739
1740         /* force user to configure all settings for low limit  */
1741         if (!(tg->bps[READ][LIMIT_LOW] || tg->iops[READ][LIMIT_LOW] ||
1742               tg->bps[WRITE][LIMIT_LOW] || tg->iops[WRITE][LIMIT_LOW]) ||
1743             tg->idletime_threshold_conf == DFL_IDLE_THRESHOLD ||
1744             tg->latency_target_conf == DFL_LATENCY_TARGET) {
1745                 tg->bps[READ][LIMIT_LOW] = 0;
1746                 tg->bps[WRITE][LIMIT_LOW] = 0;
1747                 tg->iops[READ][LIMIT_LOW] = 0;
1748                 tg->iops[WRITE][LIMIT_LOW] = 0;
1749                 tg->idletime_threshold = DFL_IDLE_THRESHOLD;
1750                 tg->latency_target = DFL_LATENCY_TARGET;
1751         } else if (index == LIMIT_LOW) {
1752                 tg->idletime_threshold = tg->idletime_threshold_conf;
1753                 tg->latency_target = tg->latency_target_conf;
1754         }
1755
1756         blk_throtl_update_limit_valid(tg->td);
1757         if (tg->td->limit_valid[LIMIT_LOW]) {
1758                 if (index == LIMIT_LOW)
1759                         tg->td->limit_index = LIMIT_LOW;
1760         } else
1761                 tg->td->limit_index = LIMIT_MAX;
1762         tg_conf_updated(tg, index == LIMIT_LOW &&
1763                 tg->td->limit_valid[LIMIT_LOW]);
1764         ret = 0;
1765 out_finish:
1766         blkg_conf_finish(&ctx);
1767         return ret ?: nbytes;
1768 }
1769
1770 static struct cftype throtl_files[] = {
1771 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
1772         {
1773                 .name = "low",
1774                 .flags = CFTYPE_NOT_ON_ROOT,
1775                 .seq_show = tg_print_limit,
1776                 .write = tg_set_limit,
1777                 .private = LIMIT_LOW,
1778         },
1779 #endif
1780         {
1781                 .name = "max",
1782                 .flags = CFTYPE_NOT_ON_ROOT,
1783                 .seq_show = tg_print_limit,
1784                 .write = tg_set_limit,
1785                 .private = LIMIT_MAX,
1786         },
1787         { }     /* terminate */
1788 };
1789
1790 static void throtl_shutdown_wq(struct request_queue *q)
1791 {
1792         struct throtl_data *td = q->td;
1793
1794         cancel_work_sync(&td->dispatch_work);
1795 }
1796
1797 static struct blkcg_policy blkcg_policy_throtl = {
1798         .dfl_cftypes            = throtl_files,
1799         .legacy_cftypes         = throtl_legacy_files,
1800
1801         .pd_alloc_fn            = throtl_pd_alloc,
1802         .pd_init_fn             = throtl_pd_init,
1803         .pd_online_fn           = throtl_pd_online,
1804         .pd_offline_fn          = throtl_pd_offline,
1805         .pd_free_fn             = throtl_pd_free,
1806 };
1807
1808 static unsigned long __tg_last_low_overflow_time(struct throtl_grp *tg)
1809 {
1810         unsigned long rtime = jiffies, wtime = jiffies;
1811
1812         if (tg->bps[READ][LIMIT_LOW] || tg->iops[READ][LIMIT_LOW])
1813                 rtime = tg->last_low_overflow_time[READ];
1814         if (tg->bps[WRITE][LIMIT_LOW] || tg->iops[WRITE][LIMIT_LOW])
1815                 wtime = tg->last_low_overflow_time[WRITE];
1816         return min(rtime, wtime);
1817 }
1818
1819 /* tg should not be an intermediate node */
1820 static unsigned long tg_last_low_overflow_time(struct throtl_grp *tg)
1821 {
1822         struct throtl_service_queue *parent_sq;
1823         struct throtl_grp *parent = tg;
1824         unsigned long ret = __tg_last_low_overflow_time(tg);
1825
1826         while (true) {
1827                 parent_sq = parent->service_queue.parent_sq;
1828                 parent = sq_to_tg(parent_sq);
1829                 if (!parent)
1830                         break;
1831
1832                 /*
1833                  * The parent doesn't have low limit, it always reaches low
1834                  * limit. Its overflow time is useless for children
1835                  */
1836                 if (!parent->bps[READ][LIMIT_LOW] &&
1837                     !parent->iops[READ][LIMIT_LOW] &&
1838                     !parent->bps[WRITE][LIMIT_LOW] &&
1839                     !parent->iops[WRITE][LIMIT_LOW])
1840                         continue;
1841                 if (time_after(__tg_last_low_overflow_time(parent), ret))
1842                         ret = __tg_last_low_overflow_time(parent);
1843         }
1844         return ret;
1845 }
1846
1847 static bool throtl_tg_is_idle(struct throtl_grp *tg)
1848 {
1849         /*
1850          * cgroup is idle if:
1851          * - single idle is too long, longer than a fixed value (in case user
1852          *   configure a too big threshold) or 4 times of idletime threshold
1853          * - average think time is more than threshold
1854          * - IO latency is largely below threshold
1855          */
1856         unsigned long time;
1857         bool ret;
1858
1859         time = min_t(unsigned long, MAX_IDLE_TIME, 4 * tg->idletime_threshold);
1860         ret = tg->latency_target == DFL_LATENCY_TARGET ||
1861               tg->idletime_threshold == DFL_IDLE_THRESHOLD ||
1862               (ktime_get_ns() >> 10) - tg->last_finish_time > time ||
1863               tg->avg_idletime > tg->idletime_threshold ||
1864               (tg->latency_target && tg->bio_cnt &&
1865                 tg->bad_bio_cnt * 5 < tg->bio_cnt);
1866         throtl_log(&tg->service_queue,
1867                 "avg_idle=%ld, idle_threshold=%ld, bad_bio=%d, total_bio=%d, is_idle=%d, scale=%d",
1868                 tg->avg_idletime, tg->idletime_threshold, tg->bad_bio_cnt,
1869                 tg->bio_cnt, ret, tg->td->scale);
1870         return ret;
1871 }
1872
1873 static bool throtl_tg_can_upgrade(struct throtl_grp *tg)
1874 {
1875         struct throtl_service_queue *sq = &tg->service_queue;
1876         bool read_limit, write_limit;
1877
1878         /*
1879          * if cgroup reaches low limit (if low limit is 0, the cgroup always
1880          * reaches), it's ok to upgrade to next limit
1881          */
1882         read_limit = tg->bps[READ][LIMIT_LOW] || tg->iops[READ][LIMIT_LOW];
1883         write_limit = tg->bps[WRITE][LIMIT_LOW] || tg->iops[WRITE][LIMIT_LOW];
1884         if (!read_limit && !write_limit)
1885                 return true;
1886         if (read_limit && sq->nr_queued[READ] &&
1887             (!write_limit || sq->nr_queued[WRITE]))
1888                 return true;
1889         if (write_limit && sq->nr_queued[WRITE] &&
1890             (!read_limit || sq->nr_queued[READ]))
1891                 return true;
1892
1893         if (time_after_eq(jiffies,
1894                 tg_last_low_overflow_time(tg) + tg->td->throtl_slice) &&
1895             throtl_tg_is_idle(tg))
1896                 return true;
1897         return false;
1898 }
1899
1900 static bool throtl_hierarchy_can_upgrade(struct throtl_grp *tg)
1901 {
1902         while (true) {
1903                 if (throtl_tg_can_upgrade(tg))
1904                         return true;
1905                 tg = sq_to_tg(tg->service_queue.parent_sq);
1906                 if (!tg || !tg_to_blkg(tg)->parent)
1907                         return false;
1908         }
1909         return false;
1910 }
1911
1912 static bool throtl_can_upgrade(struct throtl_data *td,
1913         struct throtl_grp *this_tg)
1914 {
1915         struct cgroup_subsys_state *pos_css;
1916         struct blkcg_gq *blkg;
1917
1918         if (td->limit_index != LIMIT_LOW)
1919                 return false;
1920
1921         if (time_before(jiffies, td->low_downgrade_time + td->throtl_slice))
1922                 return false;
1923
1924         rcu_read_lock();
1925         blkg_for_each_descendant_post(blkg, pos_css, td->queue->root_blkg) {
1926                 struct throtl_grp *tg = blkg_to_tg(blkg);
1927
1928                 if (tg == this_tg)
1929                         continue;
1930                 if (!list_empty(&tg_to_blkg(tg)->blkcg->css.children))
1931                         continue;
1932                 if (!throtl_hierarchy_can_upgrade(tg)) {
1933                         rcu_read_unlock();
1934                         return false;
1935                 }
1936         }
1937         rcu_read_unlock();
1938         return true;
1939 }
1940
1941 static void throtl_upgrade_check(struct throtl_grp *tg)
1942 {
1943         unsigned long now = jiffies;
1944
1945         if (tg->td->limit_index != LIMIT_LOW)
1946                 return;
1947
1948         if (time_after(tg->last_check_time + tg->td->throtl_slice, now))
1949                 return;
1950
1951         tg->last_check_time = now;
1952
1953         if (!time_after_eq(now,
1954              __tg_last_low_overflow_time(tg) + tg->td->throtl_slice))
1955                 return;
1956
1957         if (throtl_can_upgrade(tg->td, NULL))
1958                 throtl_upgrade_state(tg->td);
1959 }
1960
1961 static void throtl_upgrade_state(struct throtl_data *td)
1962 {
1963         struct cgroup_subsys_state *pos_css;
1964         struct blkcg_gq *blkg;
1965
1966         throtl_log(&td->service_queue, "upgrade to max");
1967         td->limit_index = LIMIT_MAX;
1968         td->low_upgrade_time = jiffies;
1969         td->scale = 0;
1970         rcu_read_lock();
1971         blkg_for_each_descendant_post(blkg, pos_css, td->queue->root_blkg) {
1972                 struct throtl_grp *tg = blkg_to_tg(blkg);
1973                 struct throtl_service_queue *sq = &tg->service_queue;
1974
1975                 tg->disptime = jiffies - 1;
1976                 throtl_select_dispatch(sq);
1977                 throtl_schedule_next_dispatch(sq, true);
1978         }
1979         rcu_read_unlock();
1980         throtl_select_dispatch(&td->service_queue);
1981         throtl_schedule_next_dispatch(&td->service_queue, true);
1982         queue_work(kthrotld_workqueue, &td->dispatch_work);
1983 }
1984
1985 static void throtl_downgrade_state(struct throtl_data *td)
1986 {
1987         td->scale /= 2;
1988
1989         throtl_log(&td->service_queue, "downgrade, scale %d", td->scale);
1990         if (td->scale) {
1991                 td->low_upgrade_time = jiffies - td->scale * td->throtl_slice;
1992                 return;
1993         }
1994
1995         td->limit_index = LIMIT_LOW;
1996         td->low_downgrade_time = jiffies;
1997 }
1998
1999 static bool throtl_tg_can_downgrade(struct throtl_grp *tg)
2000 {
2001         struct throtl_data *td = tg->td;
2002         unsigned long now = jiffies;
2003
2004         /*
2005          * If cgroup is below low limit, consider downgrade and throttle other
2006          * cgroups
2007          */
2008         if (time_after_eq(now, td->low_upgrade_time + td->throtl_slice) &&
2009             time_after_eq(now, tg_last_low_overflow_time(tg) +
2010                                         td->throtl_slice) &&
2011             (!throtl_tg_is_idle(tg) ||
2012              !list_empty(&tg_to_blkg(tg)->blkcg->css.children)))
2013                 return true;
2014         return false;
2015 }
2016
2017 static bool throtl_hierarchy_can_downgrade(struct throtl_grp *tg)
2018 {
2019         while (true) {
2020                 if (!throtl_tg_can_downgrade(tg))
2021                         return false;
2022                 tg = sq_to_tg(tg->service_queue.parent_sq);
2023                 if (!tg || !tg_to_blkg(tg)->parent)
2024                         break;
2025         }
2026         return true;
2027 }
2028
2029 static void throtl_downgrade_check(struct throtl_grp *tg)
2030 {
2031         uint64_t bps;
2032         unsigned int iops;
2033         unsigned long elapsed_time;
2034         unsigned long now = jiffies;
2035
2036         if (tg->td->limit_index != LIMIT_MAX ||
2037             !tg->td->limit_valid[LIMIT_LOW])
2038                 return;
2039         if (!list_empty(&tg_to_blkg(tg)->blkcg->css.children))
2040                 return;
2041         if (time_after(tg->last_check_time + tg->td->throtl_slice, now))
2042                 return;
2043
2044         elapsed_time = now - tg->last_check_time;
2045         tg->last_check_time = now;
2046
2047         if (time_before(now, tg_last_low_overflow_time(tg) +
2048                         tg->td->throtl_slice))
2049                 return;
2050
2051         if (tg->bps[READ][LIMIT_LOW]) {
2052                 bps = tg->last_bytes_disp[READ] * HZ;
2053                 do_div(bps, elapsed_time);
2054                 if (bps >= tg->bps[READ][LIMIT_LOW])
2055                         tg->last_low_overflow_time[READ] = now;
2056         }
2057
2058         if (tg->bps[WRITE][LIMIT_LOW]) {
2059                 bps = tg->last_bytes_disp[WRITE] * HZ;
2060                 do_div(bps, elapsed_time);
2061                 if (bps >= tg->bps[WRITE][LIMIT_LOW])
2062                         tg->last_low_overflow_time[WRITE] = now;
2063         }
2064
2065         if (tg->iops[READ][LIMIT_LOW]) {
2066                 tg->last_io_disp[READ] += atomic_xchg(&tg->last_io_split_cnt[READ], 0);
2067                 iops = tg->last_io_disp[READ] * HZ / elapsed_time;
2068                 if (iops >= tg->iops[READ][LIMIT_LOW])
2069                         tg->last_low_overflow_time[READ] = now;
2070         }
2071
2072         if (tg->iops[WRITE][LIMIT_LOW]) {
2073                 tg->last_io_disp[WRITE] += atomic_xchg(&tg->last_io_split_cnt[WRITE], 0);
2074                 iops = tg->last_io_disp[WRITE] * HZ / elapsed_time;
2075                 if (iops >= tg->iops[WRITE][LIMIT_LOW])
2076                         tg->last_low_overflow_time[WRITE] = now;
2077         }
2078
2079         /*
2080          * If cgroup is below low limit, consider downgrade and throttle other
2081          * cgroups
2082          */
2083         if (throtl_hierarchy_can_downgrade(tg))
2084                 throtl_downgrade_state(tg->td);
2085
2086         tg->last_bytes_disp[READ] = 0;
2087         tg->last_bytes_disp[WRITE] = 0;
2088         tg->last_io_disp[READ] = 0;
2089         tg->last_io_disp[WRITE] = 0;
2090 }
2091
2092 static void blk_throtl_update_idletime(struct throtl_grp *tg)
2093 {
2094         unsigned long now;
2095         unsigned long last_finish_time = tg->last_finish_time;
2096
2097         if (last_finish_time == 0)
2098                 return;
2099
2100         now = ktime_get_ns() >> 10;
2101         if (now <= last_finish_time ||
2102             last_finish_time == tg->checked_last_finish_time)
2103                 return;
2104
2105         tg->avg_idletime = (tg->avg_idletime * 7 + now - last_finish_time) >> 3;
2106         tg->checked_last_finish_time = last_finish_time;
2107 }
2108
2109 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
2110 static void throtl_update_latency_buckets(struct throtl_data *td)
2111 {
2112         struct avg_latency_bucket avg_latency[2][LATENCY_BUCKET_SIZE];
2113         int i, cpu, rw;
2114         unsigned long last_latency[2] = { 0 };
2115         unsigned long latency[2];
2116
2117         if (!blk_queue_nonrot(td->queue) || !td->limit_valid[LIMIT_LOW])
2118                 return;
2119         if (time_before(jiffies, td->last_calculate_time + HZ))
2120                 return;
2121         td->last_calculate_time = jiffies;
2122
2123         memset(avg_latency, 0, sizeof(avg_latency));
2124         for (rw = READ; rw <= WRITE; rw++) {
2125                 for (i = 0; i < LATENCY_BUCKET_SIZE; i++) {
2126                         struct latency_bucket *tmp = &td->tmp_buckets[rw][i];
2127
2128                         for_each_possible_cpu(cpu) {
2129                                 struct latency_bucket *bucket;
2130
2131                                 /* this isn't race free, but ok in practice */
2132                                 bucket = per_cpu_ptr(td->latency_buckets[rw],
2133                                         cpu);
2134                                 tmp->total_latency += bucket[i].total_latency;
2135                                 tmp->samples += bucket[i].samples;
2136                                 bucket[i].total_latency = 0;
2137                                 bucket[i].samples = 0;
2138                         }
2139
2140                         if (tmp->samples >= 32) {
2141                                 int samples = tmp->samples;
2142
2143                                 latency[rw] = tmp->total_latency;
2144
2145                                 tmp->total_latency = 0;
2146                                 tmp->samples = 0;
2147                                 latency[rw] /= samples;
2148                                 if (latency[rw] == 0)
2149                                         continue;
2150                                 avg_latency[rw][i].latency = latency[rw];
2151                         }
2152                 }
2153         }
2154
2155         for (rw = READ; rw <= WRITE; rw++) {
2156                 for (i = 0; i < LATENCY_BUCKET_SIZE; i++) {
2157                         if (!avg_latency[rw][i].latency) {
2158                                 if (td->avg_buckets[rw][i].latency < last_latency[rw])
2159                                         td->avg_buckets[rw][i].latency =
2160                                                 last_latency[rw];
2161                                 continue;
2162                         }
2163
2164                         if (!td->avg_buckets[rw][i].valid)
2165                                 latency[rw] = avg_latency[rw][i].latency;
2166                         else
2167                                 latency[rw] = (td->avg_buckets[rw][i].latency * 7 +
2168                                         avg_latency[rw][i].latency) >> 3;
2169
2170                         td->avg_buckets[rw][i].latency = max(latency[rw],
2171                                 last_latency[rw]);
2172                         td->avg_buckets[rw][i].valid = true;
2173                         last_latency[rw] = td->avg_buckets[rw][i].latency;
2174                 }
2175         }
2176
2177         for (i = 0; i < LATENCY_BUCKET_SIZE; i++)
2178                 throtl_log(&td->service_queue,
2179                         "Latency bucket %d: read latency=%ld, read valid=%d, "
2180                         "write latency=%ld, write valid=%d", i,
2181                         td->avg_buckets[READ][i].latency,
2182                         td->avg_buckets[READ][i].valid,
2183                         td->avg_buckets[WRITE][i].latency,
2184                         td->avg_buckets[WRITE][i].valid);
2185 }
2186 #else
2187 static inline void throtl_update_latency_buckets(struct throtl_data *td)
2188 {
2189 }
2190 #endif
2191
2192 void blk_throtl_charge_bio_split(struct bio *bio)
2193 {
2194         struct blkcg_gq *blkg = bio->bi_blkg;
2195         struct throtl_grp *parent = blkg_to_tg(blkg);
2196         struct throtl_service_queue *parent_sq;
2197         bool rw = bio_data_dir(bio);
2198
2199         do {
2200                 if (!parent->has_rules[rw])
2201                         break;
2202
2203                 atomic_inc(&parent->io_split_cnt[rw]);
2204                 atomic_inc(&parent->last_io_split_cnt[rw]);
2205
2206                 parent_sq = parent->service_queue.parent_sq;
2207                 parent = sq_to_tg(parent_sq);
2208         } while (parent);
2209 }
2210
2211 bool blk_throtl_bio(struct bio *bio)
2212 {
2213         struct request_queue *q = bio->bi_bdev->bd_disk->queue;
2214         struct blkcg_gq *blkg = bio->bi_blkg;
2215         struct throtl_qnode *qn = NULL;
2216         struct throtl_grp *tg = blkg_to_tg(blkg);
2217         struct throtl_service_queue *sq;
2218         bool rw = bio_data_dir(bio);
2219         bool throttled = false;
2220         struct throtl_data *td = tg->td;
2221
2222         rcu_read_lock();
2223
2224         /* see throtl_charge_bio() */
2225         if (bio_flagged(bio, BIO_THROTTLED))
2226                 goto out;
2227
2228         if (!cgroup_subsys_on_dfl(io_cgrp_subsys)) {
2229                 blkg_rwstat_add(&tg->stat_bytes, bio->bi_opf,
2230                                 bio->bi_iter.bi_size);
2231                 blkg_rwstat_add(&tg->stat_ios, bio->bi_opf, 1);
2232         }
2233
2234         if (!tg->has_rules[rw])
2235                 goto out;
2236
2237         spin_lock_irq(&q->queue_lock);
2238
2239         throtl_update_latency_buckets(td);
2240
2241         blk_throtl_update_idletime(tg);
2242
2243         sq = &tg->service_queue;
2244
2245 again:
2246         while (true) {
2247                 if (tg->last_low_overflow_time[rw] == 0)
2248                         tg->last_low_overflow_time[rw] = jiffies;
2249                 throtl_downgrade_check(tg);
2250                 throtl_upgrade_check(tg);
2251                 /* throtl is FIFO - if bios are already queued, should queue */
2252                 if (sq->nr_queued[rw])
2253                         break;
2254
2255                 /* if above limits, break to queue */
2256                 if (!tg_may_dispatch(tg, bio, NULL)) {
2257                         tg->last_low_overflow_time[rw] = jiffies;
2258                         if (throtl_can_upgrade(td, tg)) {
2259                                 throtl_upgrade_state(td);
2260                                 goto again;
2261                         }
2262                         break;
2263                 }
2264
2265                 /* within limits, let's charge and dispatch directly */
2266                 throtl_charge_bio(tg, bio);
2267
2268                 /*
2269                  * We need to trim slice even when bios are not being queued
2270                  * otherwise it might happen that a bio is not queued for
2271                  * a long time and slice keeps on extending and trim is not
2272                  * called for a long time. Now if limits are reduced suddenly
2273                  * we take into account all the IO dispatched so far at new
2274                  * low rate and * newly queued IO gets a really long dispatch
2275                  * time.
2276                  *
2277                  * So keep on trimming slice even if bio is not queued.
2278                  */
2279                 throtl_trim_slice(tg, rw);
2280
2281                 /*
2282                  * @bio passed through this layer without being throttled.
2283                  * Climb up the ladder.  If we're already at the top, it
2284                  * can be executed directly.
2285                  */
2286                 qn = &tg->qnode_on_parent[rw];
2287                 sq = sq->parent_sq;
2288                 tg = sq_to_tg(sq);
2289                 if (!tg)
2290                         goto out_unlock;
2291         }
2292
2293         /* out-of-limit, queue to @tg */
2294         throtl_log(sq, "[%c] bio. bdisp=%llu sz=%u bps=%llu iodisp=%u iops=%u queued=%d/%d",
2295                    rw == READ ? 'R' : 'W',
2296                    tg->bytes_disp[rw], bio->bi_iter.bi_size,
2297                    tg_bps_limit(tg, rw),
2298                    tg->io_disp[rw], tg_iops_limit(tg, rw),
2299                    sq->nr_queued[READ], sq->nr_queued[WRITE]);
2300
2301         tg->last_low_overflow_time[rw] = jiffies;
2302
2303         td->nr_queued[rw]++;
2304         throtl_add_bio_tg(bio, qn, tg);
2305         throttled = true;
2306
2307         /*
2308          * Update @tg's dispatch time and force schedule dispatch if @tg
2309          * was empty before @bio.  The forced scheduling isn't likely to
2310          * cause undue delay as @bio is likely to be dispatched directly if
2311          * its @tg's disptime is not in the future.
2312          */
2313         if (tg->flags & THROTL_TG_WAS_EMPTY) {
2314                 tg_update_disptime(tg);
2315                 throtl_schedule_next_dispatch(tg->service_queue.parent_sq, true);
2316         }
2317
2318 out_unlock:
2319         spin_unlock_irq(&q->queue_lock);
2320 out:
2321         bio_set_flag(bio, BIO_THROTTLED);
2322
2323 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
2324         if (throttled || !td->track_bio_latency)
2325                 bio->bi_issue.value |= BIO_ISSUE_THROTL_SKIP_LATENCY;
2326 #endif
2327         rcu_read_unlock();
2328         return throttled;
2329 }
2330
2331 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
2332 static void throtl_track_latency(struct throtl_data *td, sector_t size,
2333         int op, unsigned long time)
2334 {
2335         struct latency_bucket *latency;
2336         int index;
2337
2338         if (!td || td->limit_index != LIMIT_LOW ||
2339             !(op == REQ_OP_READ || op == REQ_OP_WRITE) ||
2340             !blk_queue_nonrot(td->queue))
2341                 return;
2342
2343         index = request_bucket_index(size);
2344
2345         latency = get_cpu_ptr(td->latency_buckets[op]);
2346         latency[index].total_latency += time;
2347         latency[index].samples++;
2348         put_cpu_ptr(td->latency_buckets[op]);
2349 }
2350
2351 void blk_throtl_stat_add(struct request *rq, u64 time_ns)
2352 {
2353         struct request_queue *q = rq->q;
2354         struct throtl_data *td = q->td;
2355
2356         throtl_track_latency(td, blk_rq_stats_sectors(rq), req_op(rq),
2357                              time_ns >> 10);
2358 }
2359
2360 void blk_throtl_bio_endio(struct bio *bio)
2361 {
2362         struct blkcg_gq *blkg;
2363         struct throtl_grp *tg;
2364         u64 finish_time_ns;
2365         unsigned long finish_time;
2366         unsigned long start_time;
2367         unsigned long lat;
2368         int rw = bio_data_dir(bio);
2369
2370         blkg = bio->bi_blkg;
2371         if (!blkg)
2372                 return;
2373         tg = blkg_to_tg(blkg);
2374         if (!tg->td->limit_valid[LIMIT_LOW])
2375                 return;
2376
2377         finish_time_ns = ktime_get_ns();
2378         tg->last_finish_time = finish_time_ns >> 10;
2379
2380         start_time = bio_issue_time(&bio->bi_issue) >> 10;
2381         finish_time = __bio_issue_time(finish_time_ns) >> 10;
2382         if (!start_time || finish_time <= start_time)
2383                 return;
2384
2385         lat = finish_time - start_time;
2386         /* this is only for bio based driver */
2387         if (!(bio->bi_issue.value & BIO_ISSUE_THROTL_SKIP_LATENCY))
2388                 throtl_track_latency(tg->td, bio_issue_size(&bio->bi_issue),
2389                                      bio_op(bio), lat);
2390
2391         if (tg->latency_target && lat >= tg->td->filtered_latency) {
2392                 int bucket;
2393                 unsigned int threshold;
2394
2395                 bucket = request_bucket_index(bio_issue_size(&bio->bi_issue));
2396                 threshold = tg->td->avg_buckets[rw][bucket].latency +
2397                         tg->latency_target;
2398                 if (lat > threshold)
2399                         tg->bad_bio_cnt++;
2400                 /*
2401                  * Not race free, could get wrong count, which means cgroups
2402                  * will be throttled
2403                  */
2404                 tg->bio_cnt++;
2405         }
2406
2407         if (time_after(jiffies, tg->bio_cnt_reset_time) || tg->bio_cnt > 1024) {
2408                 tg->bio_cnt_reset_time = tg->td->throtl_slice + jiffies;
2409                 tg->bio_cnt /= 2;
2410                 tg->bad_bio_cnt /= 2;
2411         }
2412 }
2413 #endif
2414
2415 int blk_throtl_init(struct request_queue *q)
2416 {
2417         struct throtl_data *td;
2418         int ret;
2419
2420         td = kzalloc_node(sizeof(*td), GFP_KERNEL, q->node);
2421         if (!td)
2422                 return -ENOMEM;
2423         td->latency_buckets[READ] = __alloc_percpu(sizeof(struct latency_bucket) *
2424                 LATENCY_BUCKET_SIZE, __alignof__(u64));
2425         if (!td->latency_buckets[READ]) {
2426                 kfree(td);
2427                 return -ENOMEM;
2428         }
2429         td->latency_buckets[WRITE] = __alloc_percpu(sizeof(struct latency_bucket) *
2430                 LATENCY_BUCKET_SIZE, __alignof__(u64));
2431         if (!td->latency_buckets[WRITE]) {
2432                 free_percpu(td->latency_buckets[READ]);
2433                 kfree(td);
2434                 return -ENOMEM;
2435         }
2436
2437         INIT_WORK(&td->dispatch_work, blk_throtl_dispatch_work_fn);
2438         throtl_service_queue_init(&td->service_queue);
2439
2440         q->td = td;
2441         td->queue = q;
2442
2443         td->limit_valid[LIMIT_MAX] = true;
2444         td->limit_index = LIMIT_MAX;
2445         td->low_upgrade_time = jiffies;
2446         td->low_downgrade_time = jiffies;
2447
2448         /* activate policy */
2449         ret = blkcg_activate_policy(q, &blkcg_policy_throtl);
2450         if (ret) {
2451                 free_percpu(td->latency_buckets[READ]);
2452                 free_percpu(td->latency_buckets[WRITE]);
2453                 kfree(td);
2454         }
2455         return ret;
2456 }
2457
2458 void blk_throtl_exit(struct request_queue *q)
2459 {
2460         BUG_ON(!q->td);
2461         del_timer_sync(&q->td->service_queue.pending_timer);
2462         throtl_shutdown_wq(q);
2463         blkcg_deactivate_policy(q, &blkcg_policy_throtl);
2464         free_percpu(q->td->latency_buckets[READ]);
2465         free_percpu(q->td->latency_buckets[WRITE]);
2466         kfree(q->td);
2467 }
2468
2469 void blk_throtl_register_queue(struct request_queue *q)
2470 {
2471         struct throtl_data *td;
2472         int i;
2473
2474         td = q->td;
2475         BUG_ON(!td);
2476
2477         if (blk_queue_nonrot(q)) {
2478                 td->throtl_slice = DFL_THROTL_SLICE_SSD;
2479                 td->filtered_latency = LATENCY_FILTERED_SSD;
2480         } else {
2481                 td->throtl_slice = DFL_THROTL_SLICE_HD;
2482                 td->filtered_latency = LATENCY_FILTERED_HD;
2483                 for (i = 0; i < LATENCY_BUCKET_SIZE; i++) {
2484                         td->avg_buckets[READ][i].latency = DFL_HD_BASELINE_LATENCY;
2485                         td->avg_buckets[WRITE][i].latency = DFL_HD_BASELINE_LATENCY;
2486                 }
2487         }
2488 #ifndef CONFIG_BLK_DEV_THROTTLING_LOW
2489         /* if no low limit, use previous default */
2490         td->throtl_slice = DFL_THROTL_SLICE_HD;
2491 #endif
2492
2493         td->track_bio_latency = !queue_is_mq(q);
2494         if (!td->track_bio_latency)
2495                 blk_stat_enable_accounting(q);
2496 }
2497
2498 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
2499 ssize_t blk_throtl_sample_time_show(struct request_queue *q, char *page)
2500 {
2501         if (!q->td)
2502                 return -EINVAL;
2503         return sprintf(page, "%u\n", jiffies_to_msecs(q->td->throtl_slice));
2504 }
2505
2506 ssize_t blk_throtl_sample_time_store(struct request_queue *q,
2507         const char *page, size_t count)
2508 {
2509         unsigned long v;
2510         unsigned long t;
2511
2512         if (!q->td)
2513                 return -EINVAL;
2514         if (kstrtoul(page, 10, &v))
2515                 return -EINVAL;
2516         t = msecs_to_jiffies(v);
2517         if (t == 0 || t > MAX_THROTL_SLICE)
2518                 return -EINVAL;
2519         q->td->throtl_slice = t;
2520         return count;
2521 }
2522 #endif
2523
2524 static int __init throtl_init(void)
2525 {
2526         kthrotld_workqueue = alloc_workqueue("kthrotld", WQ_MEM_RECLAIM, 0);
2527         if (!kthrotld_workqueue)
2528                 panic("Failed to create kthrotld\n");
2529
2530         return blkcg_policy_register(&blkcg_policy_throtl);
2531 }
2532
2533 module_init(throtl_init);