drivers/md/dm-kcopyd.c

   1 /*
   2  * Copyright (C) 2002 Sistina Software (UK) Limited.
   3  * Copyright (C) 2006 Red Hat GmbH
   4  *
   5  * This file is released under the GPL.
   6  *
   7  * Kcopyd provides a simple interface for copying an area of one
   8  * block-device to one or more other block-devices, with an asynchronous
   9  * completion notification.
  10  */
  11
  12 #include <linux/types.h>
  13 #include <asm/atomic.h>
  14 #include <linux/blkdev.h>
  15 #include <linux/fs.h>
  16 #include <linux/init.h>
  17 #include <linux/list.h>
  18 #include <linux/mempool.h>
  19 #include <linux/module.h>
  20 #include <linux/pagemap.h>
  21 #include <linux/slab.h>
  22 #include <linux/vmalloc.h>
  23 #include <linux/workqueue.h>
  24 #include <linux/mutex.h>
  25
  26 #include "kcopyd.h"
  27 #include "dm.h"
  28
  29 /*-----------------------------------------------------------------
  30  * Each kcopyd client has its own little pool of preallocated
  31  * pages for kcopyd io.
  32  *---------------------------------------------------------------*/
  33 struct dm_kcopyd_client {
  34         spinlock_t lock;
  35         struct page_list *pages;
  36         unsigned int nr_pages;
  37         unsigned int nr_free_pages;
  38
  39         struct dm_io_client *io_client;
  40
  41         wait_queue_head_t destroyq;
  42         atomic_t nr_jobs;
  43
  44         mempool_t *job_pool;
  45
  46         struct workqueue_struct *kcopyd_wq;
  47         struct work_struct kcopyd_work;
  48
  49 /*
  50  * We maintain three lists of jobs:
  51  *
  52  * i)   jobs waiting for pages
  53  * ii)  jobs that have pages, and are waiting for the io to be issued.
  54  * iii) jobs that have completed.
  55  *
  56  * All three of these are protected by job_lock.
  57  */
  58         spinlock_t job_lock;
  59         struct list_head complete_jobs;
  60         struct list_head io_jobs;
  61         struct list_head pages_jobs;
  62 };
  63
  64 static void wake(struct dm_kcopyd_client *kc)
  65 {
  66         queue_work(kc->kcopyd_wq, &kc->kcopyd_work);
  67 }
  68
  69 static struct page_list *alloc_pl(void)
  70 {
  71         struct page_list *pl;
  72
  73         pl = kmalloc(sizeof(*pl), GFP_KERNEL);
  74         if (!pl)
  75                 return NULL;
  76
  77         pl->page = alloc_page(GFP_KERNEL);
  78         if (!pl->page) {
  79                 kfree(pl);
  80                 return NULL;
  81         }
  82
  83         return pl;
  84 }
  85
  86 static void free_pl(struct page_list *pl)
  87 {
  88         __free_page(pl->page);
  89         kfree(pl);
  90 }
  91
  92 static int kcopyd_get_pages(struct dm_kcopyd_client *kc,
  93                             unsigned int nr, struct page_list **pages)
  94 {
  95         struct page_list *pl;
  96
  97         spin_lock(&kc->lock);
  98         if (kc->nr_free_pages < nr) {
  99                 spin_unlock(&kc->lock);
 100                 return -ENOMEM;
 101         }
 102
 103         kc->nr_free_pages -= nr;
 104         for (*pages = pl = kc->pages; --nr; pl = pl->next)
 105                 ;
 106
 107         kc->pages = pl->next;
 108         pl->next = NULL;
 109
 110         spin_unlock(&kc->lock);
 111
 112         return 0;
 113 }
 114
 115 static void kcopyd_put_pages(struct dm_kcopyd_client *kc, struct page_list *pl)
 116 {
 117         struct page_list *cursor;
 118
 119         spin_lock(&kc->lock);
 120         for (cursor = pl; cursor->next; cursor = cursor->next)
 121                 kc->nr_free_pages++;
 122
 123         kc->nr_free_pages++;
 124         cursor->next = kc->pages;
 125         kc->pages = pl;
 126         spin_unlock(&kc->lock);
 127 }
 128
 129 /*
 130  * These three functions resize the page pool.
 131  */
 132 static void drop_pages(struct page_list *pl)
 133 {
 134         struct page_list *next;
 135
 136         while (pl) {
 137                 next = pl->next;
 138                 free_pl(pl);
 139                 pl = next;
 140         }
 141 }
 142
 143 static int client_alloc_pages(struct dm_kcopyd_client *kc, unsigned int nr)
 144 {
 145         unsigned int i;
 146         struct page_list *pl = NULL, *next;
 147
 148         for (i = 0; i < nr; i++) {
 149                 next = alloc_pl();
 150                 if (!next) {
 151                         if (pl)
 152                                 drop_pages(pl);
 153                         return -ENOMEM;
 154                 }
 155                 next->next = pl;
 156                 pl = next;
 157         }
 158
 159         kcopyd_put_pages(kc, pl);
 160         kc->nr_pages += nr;
 161         return 0;
 162 }
 163
 164 static void client_free_pages(struct dm_kcopyd_client *kc)
 165 {
 166         BUG_ON(kc->nr_free_pages != kc->nr_pages);
 167         drop_pages(kc->pages);
 168         kc->pages = NULL;
 169         kc->nr_free_pages = kc->nr_pages = 0;
 170 }
 171
 172 /*-----------------------------------------------------------------
 173  * kcopyd_jobs need to be allocated by the *clients* of kcopyd,
 174  * for this reason we use a mempool to prevent the client from
 175  * ever having to do io (which could cause a deadlock).
 176  *---------------------------------------------------------------*/
 177 struct kcopyd_job {
 178         struct dm_kcopyd_client *kc;
 179         struct list_head list;
 180         unsigned long flags;
 181
 182         /*
 183          * Error state of the job.
 184          */
 185         int read_err;
 186         unsigned long write_err;
 187
 188         /*
 189          * Either READ or WRITE
 190          */
 191         int rw;
 192         struct dm_io_region source;
 193
 194         /*
 195          * The destinations for the transfer.
 196          */
 197         unsigned int num_dests;
 198         struct dm_io_region dests[DM_KCOPYD_MAX_REGIONS];
 199
 200         sector_t offset;
 201         unsigned int nr_pages;
 202         struct page_list *pages;
 203
 204         /*
 205          * Set this to ensure you are notified when the job has
 206          * completed.  'context' is for callback to use.
 207          */
 208         dm_kcopyd_notify_fn fn;
 209         void *context;
 210
 211         /*
 212          * These fields are only used if the job has been split
 213          * into more manageable parts.
 214          */
 215         struct mutex lock;
 216         atomic_t sub_jobs;
 217         sector_t progress;
 218 };
 219
 220 /* FIXME: this should scale with the number of pages */
 221 #define MIN_JOBS 512
 222
 223 static struct kmem_cache *_job_cache;
 224
 225 int __init dm_kcopyd_init(void)
 226 {
 227         _job_cache = KMEM_CACHE(kcopyd_job, 0);
 228         if (!_job_cache)
 229                 return -ENOMEM;
 230
 231         return 0;
 232 }
 233
 234 void dm_kcopyd_exit(void)
 235 {
 236         kmem_cache_destroy(_job_cache);
 237         _job_cache = NULL;
 238 }
 239
 240 /*
 241  * Functions to push and pop a job onto the head of a given job
 242  * list.
 243  */
 244 static struct kcopyd_job *pop(struct list_head *jobs,
 245                               struct dm_kcopyd_client *kc)
 246 {
 247         struct kcopyd_job *job = NULL;
 248         unsigned long flags;
 249
 250         spin_lock_irqsave(&kc->job_lock, flags);
 251
 252         if (!list_empty(jobs)) {
 253                 job = list_entry(jobs->next, struct kcopyd_job, list);
 254                 list_del(&job->list);
 255         }
 256         spin_unlock_irqrestore(&kc->job_lock, flags);
 257
 258         return job;
 259 }
 260
 261 static void push(struct list_head *jobs, struct kcopyd_job *job)
 262 {
 263         unsigned long flags;
 264         struct dm_kcopyd_client *kc = job->kc;
 265
 266         spin_lock_irqsave(&kc->job_lock, flags);
 267         list_add_tail(&job->list, jobs);
 268         spin_unlock_irqrestore(&kc->job_lock, flags);
 269 }
 270
 271 /*
 272  * These three functions process 1 item from the corresponding
 273  * job list.
 274  *
 275  * They return:
 276  * < 0: error
 277  *   0: success
 278  * > 0: can't process yet.
 279  */
 280 static int run_complete_job(struct kcopyd_job *job)
 281 {
 282         void *context = job->context;
 283         int read_err = job->read_err;
 284         unsigned long write_err = job->write_err;
 285         dm_kcopyd_notify_fn fn = job->fn;
 286         struct dm_kcopyd_client *kc = job->kc;
 287
 288         kcopyd_put_pages(kc, job->pages);
 289         mempool_free(job, kc->job_pool);
 290         fn(read_err, write_err, context);
 291
 292         if (atomic_dec_and_test(&kc->nr_jobs))
 293                 wake_up(&kc->destroyq);
 294
 295         return 0;
 296 }
 297
 298 static void complete_io(unsigned long error, void *context)
 299 {
 300         struct kcopyd_job *job = (struct kcopyd_job *) context;
 301         struct dm_kcopyd_client *kc = job->kc;
 302
 303         if (error) {
 304                 if (job->rw == WRITE)
 305                         job->write_err |= error;
 306                 else
 307                         job->read_err = 1;
 308
 309                 if (!test_bit(DM_KCOPYD_IGNORE_ERROR, &job->flags)) {
 310                         push(&kc->complete_jobs, job);
 311                         wake(kc);
 312                         return;
 313                 }
 314         }
 315
 316         if (job->rw == WRITE)
 317                 push(&kc->complete_jobs, job);
 318
 319         else {
 320                 job->rw = WRITE;
 321                 push(&kc->io_jobs, job);
 322         }
 323
 324         wake(kc);
 325 }
 326
 327 /*
 328  * Request io on as many buffer heads as we can currently get for
 329  * a particular job.
 330  */
 331 static int run_io_job(struct kcopyd_job *job)
 332 {
 333         int r;
 334         struct dm_io_request io_req = {
 335                 .bi_rw = job->rw,
 336                 .mem.type = DM_IO_PAGE_LIST,
 337                 .mem.ptr.pl = job->pages,
 338                 .mem.offset = job->offset,
 339                 .notify.fn = complete_io,
 340                 .notify.context = job,
 341                 .client = job->kc->io_client,
 342         };
 343
 344         if (job->rw == READ)
 345                 r = dm_io(&io_req, 1, &job->source, NULL);
 346         else
 347                 r = dm_io(&io_req, job->num_dests, job->dests, NULL);
 348
 349         return r;
 350 }
 351
 352 static int run_pages_job(struct kcopyd_job *job)
 353 {
 354         int r;
 355
 356         job->nr_pages = dm_div_up(job->dests[0].count + job->offset,
 357                                   PAGE_SIZE >> 9);
 358         r = kcopyd_get_pages(job->kc, job->nr_pages, &job->pages);
 359         if (!r) {
 360                 /* this job is ready for io */
 361                 push(&job->kc->io_jobs, job);
 362                 return 0;
 363         }
 364
 365         if (r == -ENOMEM)
 366                 /* can't complete now */
 367                 return 1;
 368
 369         return r;
 370 }
 371
 372 /*
 373  * Run through a list for as long as possible.  Returns the count
 374  * of successful jobs.
 375  */
 376 static int process_jobs(struct list_head *jobs, struct dm_kcopyd_client *kc,
 377                         int (*fn) (struct kcopyd_job *))
 378 {
 379         struct kcopyd_job *job;
 380         int r, count = 0;
 381
 382         while ((job = pop(jobs, kc))) {
 383
 384                 r = fn(job);
 385
 386                 if (r < 0) {
 387                         /* error this rogue job */
 388                         if (job->rw == WRITE)
 389                                 job->write_err = (unsigned long) -1L;
 390                         else
 391                                 job->read_err = 1;
 392                         push(&kc->complete_jobs, job);
 393                         break;
 394                 }
 395
 396                 if (r > 0) {
 397                         /*
 398                          * We couldn't service this job ATM, so
 399                          * push this job back onto the list.
 400                          */
 401                         push(jobs, job);
 402                         break;
 403                 }
 404
 405                 count++;
 406         }
 407
 408         return count;
 409 }
 410
 411 /*
 412  * kcopyd does this every time it's woken up.
 413  */
 414 static void do_work(struct work_struct *work)
 415 {
 416         struct dm_kcopyd_client *kc = container_of(work,
 417                                         struct dm_kcopyd_client, kcopyd_work);
 418
 419         /*
 420          * The order that these are called is *very* important.
 421          * complete jobs can free some pages for pages jobs.
 422          * Pages jobs when successful will jump onto the io jobs
 423          * list.  io jobs call wake when they complete and it all
 424          * starts again.
 425          */
 426         process_jobs(&kc->complete_jobs, kc, run_complete_job);
 427         process_jobs(&kc->pages_jobs, kc, run_pages_job);
 428         process_jobs(&kc->io_jobs, kc, run_io_job);
 429 }
 430
 431 /*
 432  * If we are copying a small region we just dispatch a single job
 433  * to do the copy, otherwise the io has to be split up into many
 434  * jobs.
 435  */
 436 static void dispatch_job(struct kcopyd_job *job)
 437 {
 438         struct dm_kcopyd_client *kc = job->kc;
 439         atomic_inc(&kc->nr_jobs);
 440         push(&kc->pages_jobs, job);
 441         wake(kc);
 442 }
 443
 444 #define SUB_JOB_SIZE 128
 445 static void segment_complete(int read_err, unsigned long write_err,
 446                              void *context)
 447 {
 448         /* FIXME: tidy this function */
 449         sector_t progress = 0;
 450         sector_t count = 0;
 451         struct kcopyd_job *job = (struct kcopyd_job *) context;
 452
 453         mutex_lock(&job->lock);
 454
 455         /* update the error */
 456         if (read_err)
 457                 job->read_err = 1;
 458
 459         if (write_err)
 460                 job->write_err |= write_err;
 461
 462         /*
 463          * Only dispatch more work if there hasn't been an error.
 464          */
 465         if ((!job->read_err && !job->write_err) ||
 466             test_bit(DM_KCOPYD_IGNORE_ERROR, &job->flags)) {
 467                 /* get the next chunk of work */
 468                 progress = job->progress;
 469                 count = job->source.count - progress;
 470                 if (count) {
 471                         if (count > SUB_JOB_SIZE)
 472                                 count = SUB_JOB_SIZE;
 473
 474                         job->progress += count;
 475                 }
 476         }
 477         mutex_unlock(&job->lock);
 478
 479         if (count) {
 480                 int i;
 481                 struct kcopyd_job *sub_job = mempool_alloc(job->kc->job_pool,
 482                                                            GFP_NOIO);
 483
 484                 *sub_job = *job;
 485                 sub_job->source.sector += progress;
 486                 sub_job->source.count = count;
 487
 488                 for (i = 0; i < job->num_dests; i++) {
 489                         sub_job->dests[i].sector += progress;
 490                         sub_job->dests[i].count = count;
 491                 }
 492
 493                 sub_job->fn = segment_complete;
 494                 sub_job->context = job;
 495                 dispatch_job(sub_job);
 496
 497         } else if (atomic_dec_and_test(&job->sub_jobs)) {
 498
 499                 /*
 500                  * To avoid a race we must keep the job around
 501                  * until after the notify function has completed.
 502                  * Otherwise the client may try and stop the job
 503                  * after we've completed.
 504                  */
 505                 job->fn(read_err, write_err, job->context);
 506                 mempool_free(job, job->kc->job_pool);
 507         }
 508 }
 509
 510 /*
 511  * Create some little jobs that will do the move between
 512  * them.
 513  */
 514 #define SPLIT_COUNT 8
 515 static void split_job(struct kcopyd_job *job)
 516 {
 517         int i;
 518
 519         atomic_set(&job->sub_jobs, SPLIT_COUNT);
 520         for (i = 0; i < SPLIT_COUNT; i++)
 521                 segment_complete(0, 0u, job);
 522 }
 523
 524 int dm_kcopyd_copy(struct dm_kcopyd_client *kc, struct dm_io_region *from,
 525                    unsigned int num_dests, struct dm_io_region *dests,
 526                    unsigned int flags, dm_kcopyd_notify_fn fn, void *context)
 527 {
 528         struct kcopyd_job *job;
 529
 530         /*
 531          * Allocate a new job.
 532          */
 533         job = mempool_alloc(kc->job_pool, GFP_NOIO);
 534
 535         /*
 536          * set up for the read.
 537          */
 538         job->kc = kc;
 539         job->flags = flags;
 540         job->read_err = 0;
 541         job->write_err = 0;
 542         job->rw = READ;
 543
 544         job->source = *from;
 545
 546         job->num_dests = num_dests;
 547         memcpy(&job->dests, dests, sizeof(*dests) * num_dests);
 548
 549         job->offset = 0;
 550         job->nr_pages = 0;
 551         job->pages = NULL;
 552
 553         job->fn = fn;
 554         job->context = context;
 555
 556         if (job->source.count < SUB_JOB_SIZE)
 557                 dispatch_job(job);
 558
 559         else {
 560                 mutex_init(&job->lock);
 561                 job->progress = 0;
 562                 split_job(job);
 563         }
 564
 565         return 0;
 566 }
 567 EXPORT_SYMBOL(dm_kcopyd_copy);
 568
 569 /*
 570  * Cancels a kcopyd job, eg. someone might be deactivating a
 571  * mirror.
 572  */
 573 #if 0
 574 int kcopyd_cancel(struct kcopyd_job *job, int block)
 575 {
 576         /* FIXME: finish */
 577         return -1;
 578 }
 579 #endif  /*  0  */
 580
 581 /*-----------------------------------------------------------------
 582  * Client setup
 583  *---------------------------------------------------------------*/
 584 int dm_kcopyd_client_create(unsigned int nr_pages,
 585                             struct dm_kcopyd_client **result)
 586 {
 587         int r = -ENOMEM;
 588         struct dm_kcopyd_client *kc;
 589
 590         kc = kmalloc(sizeof(*kc), GFP_KERNEL);
 591         if (!kc)
 592                 return -ENOMEM;
 593
 594         spin_lock_init(&kc->lock);
 595         spin_lock_init(&kc->job_lock);
 596         INIT_LIST_HEAD(&kc->complete_jobs);
 597         INIT_LIST_HEAD(&kc->io_jobs);
 598         INIT_LIST_HEAD(&kc->pages_jobs);
 599
 600         kc->job_pool = mempool_create_slab_pool(MIN_JOBS, _job_cache);
 601         if (!kc->job_pool)
 602                 goto bad_slab;
 603
 604         INIT_WORK(&kc->kcopyd_work, do_work);
 605         kc->kcopyd_wq = create_singlethread_workqueue("kcopyd");
 606         if (!kc->kcopyd_wq)
 607                 goto bad_workqueue;
 608
 609         kc->pages = NULL;
 610         kc->nr_pages = kc->nr_free_pages = 0;
 611         r = client_alloc_pages(kc, nr_pages);
 612         if (r)
 613                 goto bad_client_pages;
 614
 615         kc->io_client = dm_io_client_create(nr_pages);
 616         if (IS_ERR(kc->io_client)) {
 617                 r = PTR_ERR(kc->io_client);
 618                 goto bad_io_client;
 619         }
 620
 621         init_waitqueue_head(&kc->destroyq);
 622         atomic_set(&kc->nr_jobs, 0);
 623
 624         *result = kc;
 625         return 0;
 626
 627 bad_io_client:
 628         client_free_pages(kc);
 629 bad_client_pages:
 630         destroy_workqueue(kc->kcopyd_wq);
 631 bad_workqueue:
 632         mempool_destroy(kc->job_pool);
 633 bad_slab:
 634         kfree(kc);
 635
 636         return r;
 637 }
 638 EXPORT_SYMBOL(dm_kcopyd_client_create);
 639
 640 void dm_kcopyd_client_destroy(struct dm_kcopyd_client *kc)
 641 {
 642         /* Wait for completion of all jobs submitted by this client. */
 643         wait_event(kc->destroyq, !atomic_read(&kc->nr_jobs));
 644
 645         BUG_ON(!list_empty(&kc->complete_jobs));
 646         BUG_ON(!list_empty(&kc->io_jobs));
 647         BUG_ON(!list_empty(&kc->pages_jobs));
 648         destroy_workqueue(kc->kcopyd_wq);
 649         dm_io_client_destroy(kc->io_client);
 650         client_free_pages(kc);
 651         mempool_destroy(kc->job_pool);
 652         kfree(kc);
 653 }
 654 EXPORT_SYMBOL(dm_kcopyd_client_destroy);