drivers/md/dm-rq.c

   1 /*
   2  * Copyright (C) 2016 Red Hat, Inc. All rights reserved.
   3  *
   4  * This file is released under the GPL.
   5  */
   6
   7 #include "dm-core.h"
   8 #include "dm-rq.h"
   9
  10 #include <linux/elevator.h> /* for rq_end_sector() */
  11 #include <linux/blk-mq.h>
  12
  13 #define DM_MSG_PREFIX "core-rq"
  14
  15 #define DM_MQ_NR_HW_QUEUES 1
  16 #define DM_MQ_QUEUE_DEPTH 2048
  17 static unsigned dm_mq_nr_hw_queues = DM_MQ_NR_HW_QUEUES;
  18 static unsigned dm_mq_queue_depth = DM_MQ_QUEUE_DEPTH;
  19
  20 /*
  21  * Request-based DM's mempools' reserved IOs set by the user.
  22  */
  23 #define RESERVED_REQUEST_BASED_IOS      256
  24 static unsigned reserved_rq_based_ios = RESERVED_REQUEST_BASED_IOS;
  25
  26 static bool use_blk_mq = IS_ENABLED(CONFIG_DM_MQ_DEFAULT);
  27
  28 bool dm_use_blk_mq_default(void)
  29 {
  30         return use_blk_mq;
  31 }
  32
  33 bool dm_use_blk_mq(struct mapped_device *md)
  34 {
  35         return md->use_blk_mq;
  36 }
  37 EXPORT_SYMBOL_GPL(dm_use_blk_mq);
  38
  39 unsigned dm_get_reserved_rq_based_ios(void)
  40 {
  41         return __dm_get_module_param(&reserved_rq_based_ios,
  42                                      RESERVED_REQUEST_BASED_IOS, DM_RESERVED_MAX_IOS);
  43 }
  44 EXPORT_SYMBOL_GPL(dm_get_reserved_rq_based_ios);
  45
  46 static unsigned dm_get_blk_mq_nr_hw_queues(void)
  47 {
  48         return __dm_get_module_param(&dm_mq_nr_hw_queues, 1, 32);
  49 }
  50
  51 static unsigned dm_get_blk_mq_queue_depth(void)
  52 {
  53         return __dm_get_module_param(&dm_mq_queue_depth,
  54                                      DM_MQ_QUEUE_DEPTH, BLK_MQ_MAX_DEPTH);
  55 }
  56
  57 int dm_request_based(struct mapped_device *md)
  58 {
  59         return blk_queue_stackable(md->queue);
  60 }
  61
  62 static void dm_old_start_queue(struct request_queue *q)
  63 {
  64         unsigned long flags;
  65
  66         spin_lock_irqsave(q->queue_lock, flags);
  67         if (blk_queue_stopped(q))
  68                 blk_start_queue(q);
  69         spin_unlock_irqrestore(q->queue_lock, flags);
  70 }
  71
  72 static void dm_mq_start_queue(struct request_queue *q)
  73 {
  74         blk_mq_start_stopped_hw_queues(q, true);
  75         blk_mq_kick_requeue_list(q);
  76 }
  77
  78 void dm_start_queue(struct request_queue *q)
  79 {
  80         if (!q->mq_ops)
  81                 dm_old_start_queue(q);
  82         else
  83                 dm_mq_start_queue(q);
  84 }
  85
  86 static void dm_old_stop_queue(struct request_queue *q)
  87 {
  88         unsigned long flags;
  89
  90         spin_lock_irqsave(q->queue_lock, flags);
  91         if (!blk_queue_stopped(q))
  92                 blk_stop_queue(q);
  93         spin_unlock_irqrestore(q->queue_lock, flags);
  94 }
  95
  96 static void dm_mq_stop_queue(struct request_queue *q)
  97 {
  98         if (blk_mq_queue_stopped(q))
  99                 return;
 100
 101         blk_mq_quiesce_queue(q);
 102 }
 103
 104 void dm_stop_queue(struct request_queue *q)
 105 {
 106         if (!q->mq_ops)
 107                 dm_old_stop_queue(q);
 108         else
 109                 dm_mq_stop_queue(q);
 110 }
 111
 112 /*
 113  * Partial completion handling for request-based dm
 114  */
 115 static void end_clone_bio(struct bio *clone)
 116 {
 117         struct dm_rq_clone_bio_info *info =
 118                 container_of(clone, struct dm_rq_clone_bio_info, clone);
 119         struct dm_rq_target_io *tio = info->tio;
 120         struct bio *bio = info->orig;
 121         unsigned int nr_bytes = info->orig->bi_iter.bi_size;
 122         int error = clone->bi_error;
 123
 124         bio_put(clone);
 125
 126         if (tio->error)
 127                 /*
 128                  * An error has already been detected on the request.
 129                  * Once error occurred, just let clone->end_io() handle
 130                  * the remainder.
 131                  */
 132                 return;
 133         else if (error) {
 134                 /*
 135                  * Don't notice the error to the upper layer yet.
 136                  * The error handling decision is made by the target driver,
 137                  * when the request is completed.
 138                  */
 139                 tio->error = error;
 140                 return;
 141         }
 142
 143         /*
 144          * I/O for the bio successfully completed.
 145          * Notice the data completion to the upper layer.
 146          */
 147
 148         /*
 149          * bios are processed from the head of the list.
 150          * So the completing bio should always be rq->bio.
 151          * If it's not, something wrong is happening.
 152          */
 153         if (tio->orig->bio != bio)
 154                 DMERR("bio completion is going in the middle of the request");
 155
 156         /*
 157          * Update the original request.
 158          * Do not use blk_end_request() here, because it may complete
 159          * the original request before the clone, and break the ordering.
 160          */
 161         blk_update_request(tio->orig, 0, nr_bytes);
 162 }
 163
 164 static struct dm_rq_target_io *tio_from_request(struct request *rq)
 165 {
 166         return blk_mq_rq_to_pdu(rq);
 167 }
 168
 169 static void rq_end_stats(struct mapped_device *md, struct request *orig)
 170 {
 171         if (unlikely(dm_stats_used(&md->stats))) {
 172                 struct dm_rq_target_io *tio = tio_from_request(orig);
 173                 tio->duration_jiffies = jiffies - tio->duration_jiffies;
 174                 dm_stats_account_io(&md->stats, rq_data_dir(orig),
 175                                     blk_rq_pos(orig), tio->n_sectors, true,
 176                                     tio->duration_jiffies, &tio->stats_aux);
 177         }
 178 }
 179
 180 /*
 181  * Don't touch any member of the md after calling this function because
 182  * the md may be freed in dm_put() at the end of this function.
 183  * Or do dm_get() before calling this function and dm_put() later.
 184  */
 185 static void rq_completed(struct mapped_device *md, int rw, bool run_queue)
 186 {
 187         struct request_queue *q = md->queue;
 188         unsigned long flags;
 189
 190         atomic_dec(&md->pending[rw]);
 191
 192         /* nudge anyone waiting on suspend queue */
 193         if (!md_in_flight(md))
 194                 wake_up(&md->wait);
 195
 196         /*
 197          * Run this off this callpath, as drivers could invoke end_io while
 198          * inside their request_fn (and holding the queue lock). Calling
 199          * back into ->request_fn() could deadlock attempting to grab the
 200          * queue lock again.
 201          */
 202         if (!q->mq_ops && run_queue) {
 203                 spin_lock_irqsave(q->queue_lock, flags);
 204                 blk_run_queue_async(q);
 205                 spin_unlock_irqrestore(q->queue_lock, flags);
 206         }
 207
 208         /*
 209          * dm_put() must be at the end of this function. See the comment above
 210          */
 211         dm_put(md);
 212 }
 213
 214 /*
 215  * Complete the clone and the original request.
 216  * Must be called without clone's queue lock held,
 217  * see end_clone_request() for more details.
 218  */
 219 static void dm_end_request(struct request *clone, int error)
 220 {
 221         int rw = rq_data_dir(clone);
 222         struct dm_rq_target_io *tio = clone->end_io_data;
 223         struct mapped_device *md = tio->md;
 224         struct request *rq = tio->orig;
 225
 226         blk_rq_unprep_clone(clone);
 227         tio->ti->type->release_clone_rq(clone);
 228
 229         rq_end_stats(md, rq);
 230         if (!rq->q->mq_ops)
 231                 blk_end_request_all(rq, error);
 232         else
 233                 blk_mq_end_request(rq, error);
 234         rq_completed(md, rw, true);
 235 }
 236
 237 /*
 238  * Requeue the original request of a clone.
 239  */
 240 static void dm_old_requeue_request(struct request *rq)
 241 {
 242         struct request_queue *q = rq->q;
 243         unsigned long flags;
 244
 245         spin_lock_irqsave(q->queue_lock, flags);
 246         blk_requeue_request(q, rq);
 247         blk_run_queue_async(q);
 248         spin_unlock_irqrestore(q->queue_lock, flags);
 249 }
 250
 251 static void __dm_mq_kick_requeue_list(struct request_queue *q, unsigned long msecs)
 252 {
 253         blk_mq_delay_kick_requeue_list(q, msecs);
 254 }
 255
 256 void dm_mq_kick_requeue_list(struct mapped_device *md)
 257 {
 258         __dm_mq_kick_requeue_list(dm_get_md_queue(md), 0);
 259 }
 260 EXPORT_SYMBOL(dm_mq_kick_requeue_list);
 261
 262 static void dm_mq_delay_requeue_request(struct request *rq, unsigned long msecs)
 263 {
 264         blk_mq_requeue_request(rq, false);
 265         __dm_mq_kick_requeue_list(rq->q, msecs);
 266 }
 267
 268 static void dm_requeue_original_request(struct dm_rq_target_io *tio, bool delay_requeue)
 269 {
 270         struct mapped_device *md = tio->md;
 271         struct request *rq = tio->orig;
 272         int rw = rq_data_dir(rq);
 273
 274         rq_end_stats(md, rq);
 275         if (tio->clone) {
 276                 blk_rq_unprep_clone(tio->clone);
 277                 tio->ti->type->release_clone_rq(tio->clone);
 278         }
 279
 280         if (!rq->q->mq_ops)
 281                 dm_old_requeue_request(rq);
 282         else
 283                 dm_mq_delay_requeue_request(rq, delay_requeue ? 5000 : 0);
 284
 285         rq_completed(md, rw, false);
 286 }
 287
 288 static void dm_done(struct request *clone, int error, bool mapped)
 289 {
 290         int r = error;
 291         struct dm_rq_target_io *tio = clone->end_io_data;
 292         dm_request_endio_fn rq_end_io = NULL;
 293
 294         if (tio->ti) {
 295                 rq_end_io = tio->ti->type->rq_end_io;
 296
 297                 if (mapped && rq_end_io)
 298                         r = rq_end_io(tio->ti, clone, error, &tio->info);
 299         }
 300
 301         if (unlikely(r == -EREMOTEIO && (req_op(clone) == REQ_OP_WRITE_SAME) &&
 302                      !clone->q->limits.max_write_same_sectors))
 303                 disable_write_same(tio->md);
 304
 305         if (r <= 0)
 306                 /* The target wants to complete the I/O */
 307                 dm_end_request(clone, r);
 308         else if (r == DM_ENDIO_INCOMPLETE)
 309                 /* The target will handle the I/O */
 310                 return;
 311         else if (r == DM_ENDIO_REQUEUE)
 312                 /* The target wants to requeue the I/O */
 313                 dm_requeue_original_request(tio, false);
 314         else {
 315                 DMWARN("unimplemented target endio return value: %d", r);
 316                 BUG();
 317         }
 318 }
 319
 320 /*
 321  * Request completion handler for request-based dm
 322  */
 323 static void dm_softirq_done(struct request *rq)
 324 {
 325         bool mapped = true;
 326         struct dm_rq_target_io *tio = tio_from_request(rq);
 327         struct request *clone = tio->clone;
 328         int rw;
 329
 330         if (!clone) {
 331                 struct mapped_device *md = tio->md;
 332
 333                 rq_end_stats(md, rq);
 334                 rw = rq_data_dir(rq);
 335                 if (!rq->q->mq_ops)
 336                         blk_end_request_all(rq, tio->error);
 337                 else
 338                         blk_mq_end_request(rq, tio->error);
 339                 rq_completed(md, rw, false);
 340                 return;
 341         }
 342
 343         if (rq->rq_flags & RQF_FAILED)
 344                 mapped = false;
 345
 346         dm_done(clone, tio->error, mapped);
 347 }
 348
 349 /*
 350  * Complete the clone and the original request with the error status
 351  * through softirq context.
 352  */
 353 static void dm_complete_request(struct request *rq, int error)
 354 {
 355         struct dm_rq_target_io *tio = tio_from_request(rq);
 356
 357         tio->error = error;
 358         if (!rq->q->mq_ops)
 359                 blk_complete_request(rq);
 360         else
 361                 blk_mq_complete_request(rq, error);
 362 }
 363
 364 /*
 365  * Complete the not-mapped clone and the original request with the error status
 366  * through softirq context.
 367  * Target's rq_end_io() function isn't called.
 368  * This may be used when the target's map_rq() or clone_and_map_rq() functions fail.
 369  */
 370 static void dm_kill_unmapped_request(struct request *rq, int error)
 371 {
 372         rq->rq_flags |= RQF_FAILED;
 373         dm_complete_request(rq, error);
 374 }
 375
 376 /*
 377  * Called with the clone's queue lock held (in the case of .request_fn)
 378  */
 379 static void end_clone_request(struct request *clone, int error)
 380 {
 381         struct dm_rq_target_io *tio = clone->end_io_data;
 382
 383         /*
 384          * Actual request completion is done in a softirq context which doesn't
 385          * hold the clone's queue lock.  Otherwise, deadlock could occur because:
 386          *     - another request may be submitted by the upper level driver
 387          *       of the stacking during the completion
 388          *     - the submission which requires queue lock may be done
 389          *       against this clone's queue
 390          */
 391         dm_complete_request(tio->orig, error);
 392 }
 393
 394 static void dm_dispatch_clone_request(struct request *clone, struct request *rq)
 395 {
 396         int r;
 397
 398         if (blk_queue_io_stat(clone->q))
 399                 clone->rq_flags |= RQF_IO_STAT;
 400
 401         clone->start_time = jiffies;
 402         r = blk_insert_cloned_request(clone->q, clone);
 403         if (r)
 404                 /* must complete clone in terms of original request */
 405                 dm_complete_request(rq, r);
 406 }
 407
 408 static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig,
 409                                  void *data)
 410 {
 411         struct dm_rq_target_io *tio = data;
 412         struct dm_rq_clone_bio_info *info =
 413                 container_of(bio, struct dm_rq_clone_bio_info, clone);
 414
 415         info->orig = bio_orig;
 416         info->tio = tio;
 417         bio->bi_end_io = end_clone_bio;
 418
 419         return 0;
 420 }
 421
 422 static int setup_clone(struct request *clone, struct request *rq,
 423                        struct dm_rq_target_io *tio, gfp_t gfp_mask)
 424 {
 425         int r;
 426
 427         r = blk_rq_prep_clone(clone, rq, tio->md->bs, gfp_mask,
 428                               dm_rq_bio_constructor, tio);
 429         if (r)
 430                 return r;
 431
 432         clone->end_io = end_clone_request;
 433         clone->end_io_data = tio;
 434
 435         tio->clone = clone;
 436
 437         return 0;
 438 }
 439
 440 static void map_tio_request(struct kthread_work *work);
 441
 442 static void init_tio(struct dm_rq_target_io *tio, struct request *rq,
 443                      struct mapped_device *md)
 444 {
 445         tio->md = md;
 446         tio->ti = NULL;
 447         tio->clone = NULL;
 448         tio->orig = rq;
 449         tio->error = 0;
 450         /*
 451          * Avoid initializing info for blk-mq; it passes
 452          * target-specific data through info.ptr
 453          * (see: dm_mq_init_request)
 454          */
 455         if (!md->init_tio_pdu)
 456                 memset(&tio->info, 0, sizeof(tio->info));
 457         if (md->kworker_task)
 458                 kthread_init_work(&tio->work, map_tio_request);
 459 }
 460
 461 /*
 462  * Returns:
 463  * DM_MAPIO_*       : the request has been processed as indicated
 464  * DM_MAPIO_REQUEUE : the original request needs to be immediately requeued
 465  * < 0              : the request was completed due to failure
 466  */
 467 static int map_request(struct dm_rq_target_io *tio)
 468 {
 469         int r;
 470         struct dm_target *ti = tio->ti;
 471         struct mapped_device *md = tio->md;
 472         struct request *rq = tio->orig;
 473         struct request *clone = NULL;
 474
 475         r = ti->type->clone_and_map_rq(ti, rq, &tio->info, &clone);
 476         switch (r) {
 477         case DM_MAPIO_SUBMITTED:
 478                 /* The target has taken the I/O to submit by itself later */
 479                 break;
 480         case DM_MAPIO_REMAPPED:
 481                 if (setup_clone(clone, rq, tio, GFP_ATOMIC)) {
 482                         /* -ENOMEM */
 483                         ti->type->release_clone_rq(clone);
 484                         return DM_MAPIO_REQUEUE;
 485                 }
 486
 487                 /* The target has remapped the I/O so dispatch it */
 488                 trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)),
 489                                      blk_rq_pos(rq));
 490                 dm_dispatch_clone_request(clone, rq);
 491                 break;
 492         case DM_MAPIO_REQUEUE:
 493                 /* The target wants to requeue the I/O */
 494                 break;
 495         case DM_MAPIO_DELAY_REQUEUE:
 496                 /* The target wants to requeue the I/O after a delay */
 497                 dm_requeue_original_request(tio, true);
 498                 break;
 499         default:
 500                 if (r > 0) {
 501                         DMWARN("unimplemented target map return value: %d", r);
 502                         BUG();
 503                 }
 504
 505                 /* The target wants to complete the I/O */
 506                 dm_kill_unmapped_request(rq, r);
 507         }
 508
 509         return r;
 510 }
 511
 512 static void dm_start_request(struct mapped_device *md, struct request *orig)
 513 {
 514         if (!orig->q->mq_ops)
 515                 blk_start_request(orig);
 516         else
 517                 blk_mq_start_request(orig);
 518         atomic_inc(&md->pending[rq_data_dir(orig)]);
 519
 520         if (md->seq_rq_merge_deadline_usecs) {
 521                 md->last_rq_pos = rq_end_sector(orig);
 522                 md->last_rq_rw = rq_data_dir(orig);
 523                 md->last_rq_start_time = ktime_get();
 524         }
 525
 526         if (unlikely(dm_stats_used(&md->stats))) {
 527                 struct dm_rq_target_io *tio = tio_from_request(orig);
 528                 tio->duration_jiffies = jiffies;
 529                 tio->n_sectors = blk_rq_sectors(orig);
 530                 dm_stats_account_io(&md->stats, rq_data_dir(orig),
 531                                     blk_rq_pos(orig), tio->n_sectors, false, 0,
 532                                     &tio->stats_aux);
 533         }
 534
 535         /*
 536          * Hold the md reference here for the in-flight I/O.
 537          * We can't rely on the reference count by device opener,
 538          * because the device may be closed during the request completion
 539          * when all bios are completed.
 540          * See the comment in rq_completed() too.
 541          */
 542         dm_get(md);
 543 }
 544
 545 static int __dm_rq_init_rq(struct mapped_device *md, struct request *rq)
 546 {
 547         struct dm_rq_target_io *tio = blk_mq_rq_to_pdu(rq);
 548
 549         /*
 550          * Must initialize md member of tio, otherwise it won't
 551          * be available in dm_mq_queue_rq.
 552          */
 553         tio->md = md;
 554
 555         if (md->init_tio_pdu) {
 556                 /* target-specific per-io data is immediately after the tio */
 557                 tio->info.ptr = tio + 1;
 558         }
 559
 560         return 0;
 561 }
 562
 563 static int dm_rq_init_rq(struct request_queue *q, struct request *rq, gfp_t gfp)
 564 {
 565         return __dm_rq_init_rq(q->rq_alloc_data, rq);
 566 }
 567
 568 static void map_tio_request(struct kthread_work *work)
 569 {
 570         struct dm_rq_target_io *tio = container_of(work, struct dm_rq_target_io, work);
 571
 572         if (map_request(tio) == DM_MAPIO_REQUEUE)
 573                 dm_requeue_original_request(tio, false);
 574 }
 575
 576 ssize_t dm_attr_rq_based_seq_io_merge_deadline_show(struct mapped_device *md, char *buf)
 577 {
 578         return sprintf(buf, "%u\n", md->seq_rq_merge_deadline_usecs);
 579 }
 580
 581 #define MAX_SEQ_RQ_MERGE_DEADLINE_USECS 100000
 582
 583 ssize_t dm_attr_rq_based_seq_io_merge_deadline_store(struct mapped_device *md,
 584                                                      const char *buf, size_t count)
 585 {
 586         unsigned deadline;
 587
 588         if (dm_get_md_type(md) != DM_TYPE_REQUEST_BASED)
 589                 return count;
 590
 591         if (kstrtouint(buf, 10, &deadline))
 592                 return -EINVAL;
 593
 594         if (deadline > MAX_SEQ_RQ_MERGE_DEADLINE_USECS)
 595                 deadline = MAX_SEQ_RQ_MERGE_DEADLINE_USECS;
 596
 597         md->seq_rq_merge_deadline_usecs = deadline;
 598
 599         return count;
 600 }
 601
 602 static bool dm_old_request_peeked_before_merge_deadline(struct mapped_device *md)
 603 {
 604         ktime_t kt_deadline;
 605
 606         if (!md->seq_rq_merge_deadline_usecs)
 607                 return false;
 608
 609         kt_deadline = ns_to_ktime((u64)md->seq_rq_merge_deadline_usecs * NSEC_PER_USEC);
 610         kt_deadline = ktime_add_safe(md->last_rq_start_time, kt_deadline);
 611
 612         return !ktime_after(ktime_get(), kt_deadline);
 613 }
 614
 615 /*
 616  * q->request_fn for old request-based dm.
 617  * Called with the queue lock held.
 618  */
 619 static void dm_old_request_fn(struct request_queue *q)
 620 {
 621         struct mapped_device *md = q->queuedata;
 622         struct dm_target *ti = md->immutable_target;
 623         struct request *rq;
 624         struct dm_rq_target_io *tio;
 625         sector_t pos = 0;
 626
 627         if (unlikely(!ti)) {
 628                 int srcu_idx;
 629                 struct dm_table *map = dm_get_live_table(md, &srcu_idx);
 630
 631                 if (unlikely(!map)) {
 632                         dm_put_live_table(md, srcu_idx);
 633                         return;
 634                 }
 635                 ti = dm_table_find_target(map, pos);
 636                 dm_put_live_table(md, srcu_idx);
 637         }
 638
 639         /*
 640          * For suspend, check blk_queue_stopped() and increment
 641          * ->pending within a single queue_lock not to increment the
 642          * number of in-flight I/Os after the queue is stopped in
 643          * dm_suspend().
 644          */
 645         while (!blk_queue_stopped(q)) {
 646                 rq = blk_peek_request(q);
 647                 if (!rq)
 648                         return;
 649
 650                 /* always use block 0 to find the target for flushes for now */
 651                 pos = 0;
 652                 if (req_op(rq) != REQ_OP_FLUSH)
 653                         pos = blk_rq_pos(rq);
 654
 655                 if ((dm_old_request_peeked_before_merge_deadline(md) &&
 656                      md_in_flight(md) && rq->bio && !bio_multiple_segments(rq->bio) &&
 657                      md->last_rq_pos == pos && md->last_rq_rw == rq_data_dir(rq)) ||
 658                     (ti->type->busy && ti->type->busy(ti))) {
 659                         blk_delay_queue(q, 10);
 660                         return;
 661                 }
 662
 663                 dm_start_request(md, rq);
 664
 665                 tio = tio_from_request(rq);
 666                 init_tio(tio, rq, md);
 667                 /* Establish tio->ti before queuing work (map_tio_request) */
 668                 tio->ti = ti;
 669                 kthread_queue_work(&md->kworker, &tio->work);
 670                 BUG_ON(!irqs_disabled());
 671         }
 672 }
 673
 674 /*
 675  * Fully initialize a .request_fn request-based queue.
 676  */
 677 int dm_old_init_request_queue(struct mapped_device *md, struct dm_table *t)
 678 {
 679         struct dm_target *immutable_tgt;
 680
 681         /* Fully initialize the queue */
 682         md->queue->cmd_size = sizeof(struct dm_rq_target_io);
 683         md->queue->rq_alloc_data = md;
 684         md->queue->request_fn = dm_old_request_fn;
 685         md->queue->init_rq_fn = dm_rq_init_rq;
 686
 687         immutable_tgt = dm_table_get_immutable_target(t);
 688         if (immutable_tgt && immutable_tgt->per_io_data_size) {
 689                 /* any target-specific per-io data is immediately after the tio */
 690                 md->queue->cmd_size += immutable_tgt->per_io_data_size;
 691                 md->init_tio_pdu = true;
 692         }
 693         if (blk_init_allocated_queue(md->queue) < 0)
 694                 return -EINVAL;
 695
 696         /* disable dm_old_request_fn's merge heuristic by default */
 697         md->seq_rq_merge_deadline_usecs = 0;
 698
 699         dm_init_normal_md_queue(md);
 700         blk_queue_softirq_done(md->queue, dm_softirq_done);
 701
 702         /* Initialize the request-based DM worker thread */
 703         kthread_init_worker(&md->kworker);
 704         md->kworker_task = kthread_run(kthread_worker_fn, &md->kworker,
 705                                        "kdmwork-%s", dm_device_name(md));
 706         if (IS_ERR(md->kworker_task)) {
 707                 int error = PTR_ERR(md->kworker_task);
 708                 md->kworker_task = NULL;
 709                 return error;
 710         }
 711
 712         elv_register_queue(md->queue);
 713
 714         return 0;
 715 }
 716
 717 static int dm_mq_init_request(void *data, struct request *rq,
 718                        unsigned int hctx_idx, unsigned int request_idx,
 719                        unsigned int numa_node)
 720 {
 721         return __dm_rq_init_rq(data, rq);
 722 }
 723
 724 static int dm_mq_queue_rq(struct blk_mq_hw_ctx *hctx,
 725                           const struct blk_mq_queue_data *bd)
 726 {
 727         struct request *rq = bd->rq;
 728         struct dm_rq_target_io *tio = blk_mq_rq_to_pdu(rq);
 729         struct mapped_device *md = tio->md;
 730         struct dm_target *ti = md->immutable_target;
 731
 732         if (unlikely(!ti)) {
 733                 int srcu_idx;
 734                 struct dm_table *map = dm_get_live_table(md, &srcu_idx);
 735
 736                 ti = dm_table_find_target(map, 0);
 737                 dm_put_live_table(md, srcu_idx);
 738         }
 739
 740         if (ti->type->busy && ti->type->busy(ti))
 741                 return BLK_MQ_RQ_QUEUE_BUSY;
 742
 743         dm_start_request(md, rq);
 744
 745         /* Init tio using md established in .init_request */
 746         init_tio(tio, rq, md);
 747
 748         /*
 749          * Establish tio->ti before calling map_request().
 750          */
 751         tio->ti = ti;
 752
 753         /* Direct call is fine since .queue_rq allows allocations */
 754         if (map_request(tio) == DM_MAPIO_REQUEUE) {
 755                 /* Undo dm_start_request() before requeuing */
 756                 rq_end_stats(md, rq);
 757                 rq_completed(md, rq_data_dir(rq), false);
 758                 return BLK_MQ_RQ_QUEUE_BUSY;
 759         }
 760
 761         return BLK_MQ_RQ_QUEUE_OK;
 762 }
 763
 764 static struct blk_mq_ops dm_mq_ops = {
 765         .queue_rq = dm_mq_queue_rq,
 766         .complete = dm_softirq_done,
 767         .init_request = dm_mq_init_request,
 768 };
 769
 770 int dm_mq_init_request_queue(struct mapped_device *md, struct dm_table *t)
 771 {
 772         struct request_queue *q;
 773         struct dm_target *immutable_tgt;
 774         int err;
 775
 776         if (!dm_table_all_blk_mq_devices(t)) {
 777                 DMERR("request-based dm-mq may only be stacked on blk-mq device(s)");
 778                 return -EINVAL;
 779         }
 780
 781         md->tag_set = kzalloc_node(sizeof(struct blk_mq_tag_set), GFP_KERNEL, md->numa_node_id);
 782         if (!md->tag_set)
 783                 return -ENOMEM;
 784
 785         md->tag_set->ops = &dm_mq_ops;
 786         md->tag_set->queue_depth = dm_get_blk_mq_queue_depth();
 787         md->tag_set->numa_node = md->numa_node_id;
 788         md->tag_set->flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_SG_MERGE;
 789         md->tag_set->nr_hw_queues = dm_get_blk_mq_nr_hw_queues();
 790         md->tag_set->driver_data = md;
 791
 792         md->tag_set->cmd_size = sizeof(struct dm_rq_target_io);
 793         immutable_tgt = dm_table_get_immutable_target(t);
 794         if (immutable_tgt && immutable_tgt->per_io_data_size) {
 795                 /* any target-specific per-io data is immediately after the tio */
 796                 md->tag_set->cmd_size += immutable_tgt->per_io_data_size;
 797                 md->init_tio_pdu = true;
 798         }
 799
 800         err = blk_mq_alloc_tag_set(md->tag_set);
 801         if (err)
 802                 goto out_kfree_tag_set;
 803
 804         q = blk_mq_init_allocated_queue(md->tag_set, md->queue);
 805         if (IS_ERR(q)) {
 806                 err = PTR_ERR(q);
 807                 goto out_tag_set;
 808         }
 809         dm_init_md_queue(md);
 810
 811         /* backfill 'mq' sysfs registration normally done in blk_register_queue */
 812         blk_mq_register_dev(disk_to_dev(md->disk), q);
 813
 814         return 0;
 815
 816 out_tag_set:
 817         blk_mq_free_tag_set(md->tag_set);
 818 out_kfree_tag_set:
 819         kfree(md->tag_set);
 820
 821         return err;
 822 }
 823
 824 void dm_mq_cleanup_mapped_device(struct mapped_device *md)
 825 {
 826         if (md->tag_set) {
 827                 blk_mq_free_tag_set(md->tag_set);
 828                 kfree(md->tag_set);
 829         }
 830 }
 831
 832 module_param(reserved_rq_based_ios, uint, S_IRUGO | S_IWUSR);
 833 MODULE_PARM_DESC(reserved_rq_based_ios, "Reserved IOs in request-based mempools");
 834
 835 module_param(use_blk_mq, bool, S_IRUGO | S_IWUSR);
 836 MODULE_PARM_DESC(use_blk_mq, "Use block multiqueue for request-based DM devices");
 837
 838 module_param(dm_mq_nr_hw_queues, uint, S_IRUGO | S_IWUSR);
 839 MODULE_PARM_DESC(dm_mq_nr_hw_queues, "Number of hardware queues for request-based dm-mq devices");
 840
 841 module_param(dm_mq_queue_depth, uint, S_IRUGO | S_IWUSR);
 842 MODULE_PARM_DESC(dm_mq_queue_depth, "Queue depth for request-based dm-mq devices");