block/blk-mq-sched.c

   1 /*
   2  * blk-mq scheduling framework
   3  *
   4  * Copyright (C) 2016 Jens Axboe
   5  */
   6 #include <linux/kernel.h>
   7 #include <linux/module.h>
   8 #include <linux/blk-mq.h>
   9
  10 #include <trace/events/block.h>
  11
  12 #include "blk.h"
  13 #include "blk-mq.h"
  14 #include "blk-mq-sched.h"
  15 #include "blk-mq-tag.h"
  16 #include "blk-wbt.h"
  17
  18 void blk_mq_sched_free_hctx_data(struct request_queue *q,
  19                                  void (*exit)(struct blk_mq_hw_ctx *))
  20 {
  21         struct blk_mq_hw_ctx *hctx;
  22         int i;
  23
  24         queue_for_each_hw_ctx(q, hctx, i) {
  25                 if (exit && hctx->sched_data)
  26                         exit(hctx);
  27                 kfree(hctx->sched_data);
  28                 hctx->sched_data = NULL;
  29         }
  30 }
  31 EXPORT_SYMBOL_GPL(blk_mq_sched_free_hctx_data);
  32
  33 static void __blk_mq_sched_assign_ioc(struct request_queue *q,
  34                                       struct request *rq,
  35                                       struct bio *bio,
  36                                       struct io_context *ioc)
  37 {
  38         struct io_cq *icq;
  39
  40         spin_lock_irq(q->queue_lock);
  41         icq = ioc_lookup_icq(ioc, q);
  42         spin_unlock_irq(q->queue_lock);
  43
  44         if (!icq) {
  45                 icq = ioc_create_icq(ioc, q, GFP_ATOMIC);
  46                 if (!icq)
  47                         return;
  48         }
  49
  50         rq->elv.icq = icq;
  51         if (!blk_mq_sched_get_rq_priv(q, rq, bio)) {
  52                 rq->rq_flags |= RQF_ELVPRIV;
  53                 get_io_context(icq->ioc);
  54                 return;
  55         }
  56
  57         rq->elv.icq = NULL;
  58 }
  59
  60 static void blk_mq_sched_assign_ioc(struct request_queue *q,
  61                                     struct request *rq, struct bio *bio)
  62 {
  63         struct io_context *ioc;
  64
  65         ioc = rq_ioc(bio);
  66         if (ioc)
  67                 __blk_mq_sched_assign_ioc(q, rq, bio, ioc);
  68 }
  69
  70 struct request *blk_mq_sched_get_request(struct request_queue *q,
  71                                          struct bio *bio,
  72                                          unsigned int op,
  73                                          struct blk_mq_alloc_data *data)
  74 {
  75         struct elevator_queue *e = q->elevator;
  76         struct request *rq;
  77
  78         blk_queue_enter_live(q);
  79         data->q = q;
  80         if (likely(!data->ctx))
  81                 data->ctx = blk_mq_get_ctx(q);
  82         if (likely(!data->hctx))
  83                 data->hctx = blk_mq_map_queue(q, data->ctx->cpu);
  84
  85         /*
  86          * For a reserved tag, allocate a normal request since we might
  87          * have driver dependencies on the value of the internal tag.
  88          */
  89         if (e && !(data->flags & BLK_MQ_REQ_RESERVED)) {
  90                 data->flags |= BLK_MQ_REQ_INTERNAL;
  91
  92                 /*
  93                  * Flush requests are special and go directly to the
  94                  * dispatch list.
  95                  */
  96                 if (!op_is_flush(op) && e->type->ops.mq.get_request) {
  97                         rq = e->type->ops.mq.get_request(q, op, data);
  98                         if (rq)
  99                                 rq->rq_flags |= RQF_QUEUED;
 100                 } else
 101                         rq = __blk_mq_alloc_request(data, op);
 102         } else {
 103                 rq = __blk_mq_alloc_request(data, op);
 104         }
 105
 106         if (rq) {
 107                 if (!op_is_flush(op)) {
 108                         rq->elv.icq = NULL;
 109                         if (e && e->type->icq_cache)
 110                                 blk_mq_sched_assign_ioc(q, rq, bio);
 111                 }
 112                 data->hctx->queued++;
 113                 return rq;
 114         }
 115
 116         blk_queue_exit(q);
 117         return NULL;
 118 }
 119
 120 void blk_mq_sched_put_request(struct request *rq)
 121 {
 122         struct request_queue *q = rq->q;
 123         struct elevator_queue *e = q->elevator;
 124
 125         if (rq->rq_flags & RQF_ELVPRIV) {
 126                 blk_mq_sched_put_rq_priv(rq->q, rq);
 127                 if (rq->elv.icq) {
 128                         put_io_context(rq->elv.icq->ioc);
 129                         rq->elv.icq = NULL;
 130                 }
 131         }
 132
 133         if ((rq->rq_flags & RQF_QUEUED) && e && e->type->ops.mq.put_request)
 134                 e->type->ops.mq.put_request(rq);
 135         else
 136                 blk_mq_finish_request(rq);
 137 }
 138
 139 void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
 140 {
 141         struct request_queue *q = hctx->queue;
 142         struct elevator_queue *e = q->elevator;
 143         const bool has_sched_dispatch = e && e->type->ops.mq.dispatch_request;
 144         bool did_work = false;
 145         LIST_HEAD(rq_list);
 146
 147         if (unlikely(blk_mq_hctx_stopped(hctx)))
 148                 return;
 149
 150         hctx->run++;
 151
 152         /*
 153          * If we have previous entries on our dispatch list, grab them first for
 154          * more fair dispatch.
 155          */
 156         if (!list_empty_careful(&hctx->dispatch)) {
 157                 spin_lock(&hctx->lock);
 158                 if (!list_empty(&hctx->dispatch))
 159                         list_splice_init(&hctx->dispatch, &rq_list);
 160                 spin_unlock(&hctx->lock);
 161         }
 162
 163         /*
 164          * Only ask the scheduler for requests, if we didn't have residual
 165          * requests from the dispatch list. This is to avoid the case where
 166          * we only ever dispatch a fraction of the requests available because
 167          * of low device queue depth. Once we pull requests out of the IO
 168          * scheduler, we can no longer merge or sort them. So it's best to
 169          * leave them there for as long as we can. Mark the hw queue as
 170          * needing a restart in that case.
 171          */
 172         if (!list_empty(&rq_list)) {
 173                 blk_mq_sched_mark_restart_hctx(hctx);
 174                 did_work = blk_mq_dispatch_rq_list(q, &rq_list);
 175         } else if (!has_sched_dispatch) {
 176                 blk_mq_flush_busy_ctxs(hctx, &rq_list);
 177                 blk_mq_dispatch_rq_list(q, &rq_list);
 178         }
 179
 180         /*
 181          * We want to dispatch from the scheduler if we had no work left
 182          * on the dispatch list, OR if we did have work but weren't able
 183          * to make progress.
 184          */
 185         if (!did_work && has_sched_dispatch) {
 186                 do {
 187                         struct request *rq;
 188
 189                         rq = e->type->ops.mq.dispatch_request(hctx);
 190                         if (!rq)
 191                                 break;
 192                         list_add(&rq->queuelist, &rq_list);
 193                 } while (blk_mq_dispatch_rq_list(q, &rq_list));
 194         }
 195 }
 196
 197 bool blk_mq_sched_try_merge(struct request_queue *q, struct bio *bio,
 198                             struct request **merged_request)
 199 {
 200         struct request *rq;
 201
 202         switch (elv_merge(q, &rq, bio)) {
 203         case ELEVATOR_BACK_MERGE:
 204                 if (!blk_mq_sched_allow_merge(q, rq, bio))
 205                         return false;
 206                 if (!bio_attempt_back_merge(q, rq, bio))
 207                         return false;
 208                 *merged_request = attempt_back_merge(q, rq);
 209                 if (!*merged_request)
 210                         elv_merged_request(q, rq, ELEVATOR_BACK_MERGE);
 211                 return true;
 212         case ELEVATOR_FRONT_MERGE:
 213                 if (!blk_mq_sched_allow_merge(q, rq, bio))
 214                         return false;
 215                 if (!bio_attempt_front_merge(q, rq, bio))
 216                         return false;
 217                 *merged_request = attempt_front_merge(q, rq);
 218                 if (!*merged_request)
 219                         elv_merged_request(q, rq, ELEVATOR_FRONT_MERGE);
 220                 return true;
 221         default:
 222                 return false;
 223         }
 224 }
 225 EXPORT_SYMBOL_GPL(blk_mq_sched_try_merge);
 226
 227 bool __blk_mq_sched_bio_merge(struct request_queue *q, struct bio *bio)
 228 {
 229         struct elevator_queue *e = q->elevator;
 230
 231         if (e->type->ops.mq.bio_merge) {
 232                 struct blk_mq_ctx *ctx = blk_mq_get_ctx(q);
 233                 struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);
 234
 235                 blk_mq_put_ctx(ctx);
 236                 return e->type->ops.mq.bio_merge(hctx, bio);
 237         }
 238
 239         return false;
 240 }
 241
 242 bool blk_mq_sched_try_insert_merge(struct request_queue *q, struct request *rq)
 243 {
 244         return rq_mergeable(rq) && elv_attempt_insert_merge(q, rq);
 245 }
 246 EXPORT_SYMBOL_GPL(blk_mq_sched_try_insert_merge);
 247
 248 void blk_mq_sched_request_inserted(struct request *rq)
 249 {
 250         trace_block_rq_insert(rq->q, rq);
 251 }
 252 EXPORT_SYMBOL_GPL(blk_mq_sched_request_inserted);
 253
 254 static bool blk_mq_sched_bypass_insert(struct blk_mq_hw_ctx *hctx,
 255                                        struct request *rq)
 256 {
 257         if (rq->tag == -1) {
 258                 rq->rq_flags |= RQF_SORTED;
 259                 return false;
 260         }
 261
 262         /*
 263          * If we already have a real request tag, send directly to
 264          * the dispatch list.
 265          */
 266         spin_lock(&hctx->lock);
 267         list_add(&rq->queuelist, &hctx->dispatch);
 268         spin_unlock(&hctx->lock);
 269         return true;
 270 }
 271
 272 static bool blk_mq_sched_restart_hctx(struct blk_mq_hw_ctx *hctx)
 273 {
 274         if (test_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state)) {
 275                 clear_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state);
 276                 if (blk_mq_hctx_has_pending(hctx)) {
 277                         blk_mq_run_hw_queue(hctx, true);
 278                         return true;
 279                 }
 280         }
 281         return false;
 282 }
 283
 284 /**
 285  * list_for_each_entry_rcu_rr - iterate in a round-robin fashion over rcu list
 286  * @pos:    loop cursor.
 287  * @skip:   the list element that will not be examined. Iteration starts at
 288  *          @skip->next.
 289  * @head:   head of the list to examine. This list must have at least one
 290  *          element, namely @skip.
 291  * @member: name of the list_head structure within typeof(*pos).
 292  */
 293 #define list_for_each_entry_rcu_rr(pos, skip, head, member)             \
 294         for ((pos) = (skip);                                            \
 295              (pos = (pos)->member.next != (head) ? list_entry_rcu(      \
 296                         (pos)->member.next, typeof(*pos), member) :     \
 297               list_entry_rcu((pos)->member.next->next, typeof(*pos), member)), \
 298              (pos) != (skip); )
 299
 300 /*
 301  * Called after a driver tag has been freed to check whether a hctx needs to
 302  * be restarted. Restarts @hctx if its tag set is not shared. Restarts hardware
 303  * queues in a round-robin fashion if the tag set of @hctx is shared with other
 304  * hardware queues.
 305  */
 306 void blk_mq_sched_restart(struct blk_mq_hw_ctx *const hctx)
 307 {
 308         struct blk_mq_tags *const tags = hctx->tags;
 309         struct blk_mq_tag_set *const set = hctx->queue->tag_set;
 310         struct request_queue *const queue = hctx->queue, *q;
 311         struct blk_mq_hw_ctx *hctx2;
 312         unsigned int i, j;
 313
 314         if (set->flags & BLK_MQ_F_TAG_SHARED) {
 315                 rcu_read_lock();
 316                 list_for_each_entry_rcu_rr(q, queue, &set->tag_list,
 317                                            tag_set_list) {
 318                         queue_for_each_hw_ctx(q, hctx2, i)
 319                                 if (hctx2->tags == tags &&
 320                                     blk_mq_sched_restart_hctx(hctx2))
 321                                         goto done;
 322                 }
 323                 j = hctx->queue_num + 1;
 324                 for (i = 0; i < queue->nr_hw_queues; i++, j++) {
 325                         if (j == queue->nr_hw_queues)
 326                                 j = 0;
 327                         hctx2 = queue->queue_hw_ctx[j];
 328                         if (hctx2->tags == tags &&
 329                             blk_mq_sched_restart_hctx(hctx2))
 330                                 break;
 331                 }
 332 done:
 333                 rcu_read_unlock();
 334         } else {
 335                 blk_mq_sched_restart_hctx(hctx);
 336         }
 337 }
 338
 339 /*
 340  * Add flush/fua to the queue. If we fail getting a driver tag, then
 341  * punt to the requeue list. Requeue will re-invoke us from a context
 342  * that's safe to block from.
 343  */
 344 static void blk_mq_sched_insert_flush(struct blk_mq_hw_ctx *hctx,
 345                                       struct request *rq, bool can_block)
 346 {
 347         if (blk_mq_get_driver_tag(rq, &hctx, can_block)) {
 348                 blk_insert_flush(rq);
 349                 blk_mq_run_hw_queue(hctx, true);
 350         } else
 351                 blk_mq_add_to_requeue_list(rq, false, true);
 352 }
 353
 354 void blk_mq_sched_insert_request(struct request *rq, bool at_head,
 355                                  bool run_queue, bool async, bool can_block)
 356 {
 357         struct request_queue *q = rq->q;
 358         struct elevator_queue *e = q->elevator;
 359         struct blk_mq_ctx *ctx = rq->mq_ctx;
 360         struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);
 361
 362         if (rq->tag == -1 && op_is_flush(rq->cmd_flags)) {
 363                 blk_mq_sched_insert_flush(hctx, rq, can_block);
 364                 return;
 365         }
 366
 367         if (e && blk_mq_sched_bypass_insert(hctx, rq))
 368                 goto run;
 369
 370         if (e && e->type->ops.mq.insert_requests) {
 371                 LIST_HEAD(list);
 372
 373                 list_add(&rq->queuelist, &list);
 374                 e->type->ops.mq.insert_requests(hctx, &list, at_head);
 375         } else {
 376                 spin_lock(&ctx->lock);
 377                 __blk_mq_insert_request(hctx, rq, at_head);
 378                 spin_unlock(&ctx->lock);
 379         }
 380
 381 run:
 382         if (run_queue)
 383                 blk_mq_run_hw_queue(hctx, async);
 384 }
 385
 386 void blk_mq_sched_insert_requests(struct request_queue *q,
 387                                   struct blk_mq_ctx *ctx,
 388                                   struct list_head *list, bool run_queue_async)
 389 {
 390         struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);
 391         struct elevator_queue *e = hctx->queue->elevator;
 392
 393         if (e) {
 394                 struct request *rq, *next;
 395
 396                 /*
 397                  * We bypass requests that already have a driver tag assigned,
 398                  * which should only be flushes. Flushes are only ever inserted
 399                  * as single requests, so we shouldn't ever hit the
 400                  * WARN_ON_ONCE() below (but let's handle it just in case).
 401                  */
 402                 list_for_each_entry_safe(rq, next, list, queuelist) {
 403                         if (WARN_ON_ONCE(rq->tag != -1)) {
 404                                 list_del_init(&rq->queuelist);
 405                                 blk_mq_sched_bypass_insert(hctx, rq);
 406                         }
 407                 }
 408         }
 409
 410         if (e && e->type->ops.mq.insert_requests)
 411                 e->type->ops.mq.insert_requests(hctx, list, false);
 412         else
 413                 blk_mq_insert_requests(hctx, ctx, list);
 414
 415         blk_mq_run_hw_queue(hctx, run_queue_async);
 416 }
 417
 418 static void blk_mq_sched_free_tags(struct blk_mq_tag_set *set,
 419                                    struct blk_mq_hw_ctx *hctx,
 420                                    unsigned int hctx_idx)
 421 {
 422         if (hctx->sched_tags) {
 423                 blk_mq_free_rqs(set, hctx->sched_tags, hctx_idx);
 424                 blk_mq_free_rq_map(hctx->sched_tags);
 425                 hctx->sched_tags = NULL;
 426         }
 427 }
 428
 429 static int blk_mq_sched_alloc_tags(struct request_queue *q,
 430                                    struct blk_mq_hw_ctx *hctx,
 431                                    unsigned int hctx_idx)
 432 {
 433         struct blk_mq_tag_set *set = q->tag_set;
 434         int ret;
 435
 436         hctx->sched_tags = blk_mq_alloc_rq_map(set, hctx_idx, q->nr_requests,
 437                                                set->reserved_tags);
 438         if (!hctx->sched_tags)
 439                 return -ENOMEM;
 440
 441         ret = blk_mq_alloc_rqs(set, hctx->sched_tags, hctx_idx, q->nr_requests);
 442         if (ret)
 443                 blk_mq_sched_free_tags(set, hctx, hctx_idx);
 444
 445         return ret;
 446 }
 447
 448 static void blk_mq_sched_tags_teardown(struct request_queue *q)
 449 {
 450         struct blk_mq_tag_set *set = q->tag_set;
 451         struct blk_mq_hw_ctx *hctx;
 452         int i;
 453
 454         queue_for_each_hw_ctx(q, hctx, i)
 455                 blk_mq_sched_free_tags(set, hctx, i);
 456 }
 457
 458 int blk_mq_sched_init_hctx(struct request_queue *q, struct blk_mq_hw_ctx *hctx,
 459                            unsigned int hctx_idx)
 460 {
 461         struct elevator_queue *e = q->elevator;
 462         int ret;
 463
 464         if (!e)
 465                 return 0;
 466
 467         ret = blk_mq_sched_alloc_tags(q, hctx, hctx_idx);
 468         if (ret)
 469                 return ret;
 470
 471         if (e->type->ops.mq.init_hctx) {
 472                 ret = e->type->ops.mq.init_hctx(hctx, hctx_idx);
 473                 if (ret) {
 474                         blk_mq_sched_free_tags(q->tag_set, hctx, hctx_idx);
 475                         return ret;
 476                 }
 477         }
 478
 479         return 0;
 480 }
 481
 482 void blk_mq_sched_exit_hctx(struct request_queue *q, struct blk_mq_hw_ctx *hctx,
 483                             unsigned int hctx_idx)
 484 {
 485         struct elevator_queue *e = q->elevator;
 486
 487         if (!e)
 488                 return;
 489
 490         if (e->type->ops.mq.exit_hctx && hctx->sched_data) {
 491                 e->type->ops.mq.exit_hctx(hctx, hctx_idx);
 492                 hctx->sched_data = NULL;
 493         }
 494
 495         blk_mq_sched_free_tags(q->tag_set, hctx, hctx_idx);
 496 }
 497
 498 int blk_mq_init_sched(struct request_queue *q, struct elevator_type *e)
 499 {
 500         struct blk_mq_hw_ctx *hctx;
 501         struct elevator_queue *eq;
 502         unsigned int i;
 503         int ret;
 504
 505         if (!e) {
 506                 q->elevator = NULL;
 507                 return 0;
 508         }
 509
 510         /*
 511          * Default to 256, since we don't split into sync/async like the
 512          * old code did. Additionally, this is a per-hw queue depth.
 513          */
 514         q->nr_requests = 2 * BLKDEV_MAX_RQ;
 515
 516         queue_for_each_hw_ctx(q, hctx, i) {
 517                 ret = blk_mq_sched_alloc_tags(q, hctx, i);
 518                 if (ret)
 519                         goto err;
 520         }
 521
 522         ret = e->ops.mq.init_sched(q, e);
 523         if (ret)
 524                 goto err;
 525
 526         if (e->ops.mq.init_hctx) {
 527                 queue_for_each_hw_ctx(q, hctx, i) {
 528                         ret = e->ops.mq.init_hctx(hctx, i);
 529                         if (ret) {
 530                                 eq = q->elevator;
 531                                 blk_mq_exit_sched(q, eq);
 532                                 kobject_put(&eq->kobj);
 533                                 return ret;
 534                         }
 535                 }
 536         }
 537
 538         return 0;
 539
 540 err:
 541         blk_mq_sched_tags_teardown(q);
 542         q->elevator = NULL;
 543         return ret;
 544 }
 545
 546 void blk_mq_exit_sched(struct request_queue *q, struct elevator_queue *e)
 547 {
 548         struct blk_mq_hw_ctx *hctx;
 549         unsigned int i;
 550
 551         if (e->type->ops.mq.exit_hctx) {
 552                 queue_for_each_hw_ctx(q, hctx, i) {
 553                         if (hctx->sched_data) {
 554                                 e->type->ops.mq.exit_hctx(hctx, i);
 555                                 hctx->sched_data = NULL;
 556                         }
 557                 }
 558         }
 559         if (e->type->ops.mq.exit_sched)
 560                 e->type->ops.mq.exit_sched(e);
 561         blk_mq_sched_tags_teardown(q);
 562         q->elevator = NULL;
 563 }
 564
 565 int blk_mq_sched_init(struct request_queue *q)
 566 {
 567         int ret;
 568
 569         mutex_lock(&q->sysfs_lock);
 570         ret = elevator_init(q, NULL);
 571         mutex_unlock(&q->sysfs_lock);
 572
 573         return ret;
 574 }