linux/drivers/md/dm-table.c
<<
>>
Prefs
   1/*
   2 * Copyright (C) 2001 Sistina Software (UK) Limited.
   3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
   4 *
   5 * This file is released under the GPL.
   6 */
   7
   8#include "dm.h"
   9
  10#include <linux/module.h>
  11#include <linux/vmalloc.h>
  12#include <linux/blkdev.h>
  13#include <linux/namei.h>
  14#include <linux/ctype.h>
  15#include <linux/string.h>
  16#include <linux/slab.h>
  17#include <linux/interrupt.h>
  18#include <linux/mutex.h>
  19#include <linux/delay.h>
  20#include <linux/atomic.h>
  21
  22#define DM_MSG_PREFIX "table"
  23
  24#define MAX_DEPTH 16
  25#define NODE_SIZE L1_CACHE_BYTES
  26#define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t))
  27#define CHILDREN_PER_NODE (KEYS_PER_NODE + 1)
  28
  29/*
  30 * The table has always exactly one reference from either mapped_device->map
  31 * or hash_cell->new_map. This reference is not counted in table->holders.
  32 * A pair of dm_create_table/dm_destroy_table functions is used for table
  33 * creation/destruction.
  34 *
  35 * Temporary references from the other code increase table->holders. A pair
  36 * of dm_table_get/dm_table_put functions is used to manipulate it.
  37 *
  38 * When the table is about to be destroyed, we wait for table->holders to
  39 * drop to zero.
  40 */
  41
  42struct dm_table {
  43        struct mapped_device *md;
  44        atomic_t holders;
  45        unsigned type;
  46
  47        /* btree table */
  48        unsigned int depth;
  49        unsigned int counts[MAX_DEPTH]; /* in nodes */
  50        sector_t *index[MAX_DEPTH];
  51
  52        unsigned int num_targets;
  53        unsigned int num_allocated;
  54        sector_t *highs;
  55        struct dm_target *targets;
  56
  57        struct target_type *immutable_target_type;
  58        unsigned integrity_supported:1;
  59        unsigned singleton:1;
  60
  61        /*
  62         * Indicates the rw permissions for the new logical
  63         * device.  This should be a combination of FMODE_READ
  64         * and FMODE_WRITE.
  65         */
  66        fmode_t mode;
  67
  68        /* a list of devices used by this table */
  69        struct list_head devices;
  70
  71        /* events get handed up using this callback */
  72        void (*event_fn)(void *);
  73        void *event_context;
  74
  75        struct dm_md_mempools *mempools;
  76
  77        struct list_head target_callbacks;
  78};
  79
  80/*
  81 * Similar to ceiling(log_size(n))
  82 */
  83static unsigned int int_log(unsigned int n, unsigned int base)
  84{
  85        int result = 0;
  86
  87        while (n > 1) {
  88                n = dm_div_up(n, base);
  89                result++;
  90        }
  91
  92        return result;
  93}
  94
  95/*
  96 * Calculate the index of the child node of the n'th node k'th key.
  97 */
  98static inline unsigned int get_child(unsigned int n, unsigned int k)
  99{
 100        return (n * CHILDREN_PER_NODE) + k;
 101}
 102
 103/*
 104 * Return the n'th node of level l from table t.
 105 */
 106static inline sector_t *get_node(struct dm_table *t,
 107                                 unsigned int l, unsigned int n)
 108{
 109        return t->index[l] + (n * KEYS_PER_NODE);
 110}
 111
 112/*
 113 * Return the highest key that you could lookup from the n'th
 114 * node on level l of the btree.
 115 */
 116static sector_t high(struct dm_table *t, unsigned int l, unsigned int n)
 117{
 118        for (; l < t->depth - 1; l++)
 119                n = get_child(n, CHILDREN_PER_NODE - 1);
 120
 121        if (n >= t->counts[l])
 122                return (sector_t) - 1;
 123
 124        return get_node(t, l, n)[KEYS_PER_NODE - 1];
 125}
 126
 127/*
 128 * Fills in a level of the btree based on the highs of the level
 129 * below it.
 130 */
 131static int setup_btree_index(unsigned int l, struct dm_table *t)
 132{
 133        unsigned int n, k;
 134        sector_t *node;
 135
 136        for (n = 0U; n < t->counts[l]; n++) {
 137                node = get_node(t, l, n);
 138
 139                for (k = 0U; k < KEYS_PER_NODE; k++)
 140                        node[k] = high(t, l + 1, get_child(n, k));
 141        }
 142
 143        return 0;
 144}
 145
 146void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size)
 147{
 148        unsigned long size;
 149        void *addr;
 150
 151        /*
 152         * Check that we're not going to overflow.
 153         */
 154        if (nmemb > (ULONG_MAX / elem_size))
 155                return NULL;
 156
 157        size = nmemb * elem_size;
 158        addr = vzalloc(size);
 159
 160        return addr;
 161}
 162EXPORT_SYMBOL(dm_vcalloc);
 163
 164/*
 165 * highs, and targets are managed as dynamic arrays during a
 166 * table load.
 167 */
 168static int alloc_targets(struct dm_table *t, unsigned int num)
 169{
 170        sector_t *n_highs;
 171        struct dm_target *n_targets;
 172        int n = t->num_targets;
 173
 174        /*
 175         * Allocate both the target array and offset array at once.
 176         * Append an empty entry to catch sectors beyond the end of
 177         * the device.
 178         */
 179        n_highs = (sector_t *) dm_vcalloc(num + 1, sizeof(struct dm_target) +
 180                                          sizeof(sector_t));
 181        if (!n_highs)
 182                return -ENOMEM;
 183
 184        n_targets = (struct dm_target *) (n_highs + num);
 185
 186        if (n) {
 187                memcpy(n_highs, t->highs, sizeof(*n_highs) * n);
 188                memcpy(n_targets, t->targets, sizeof(*n_targets) * n);
 189        }
 190
 191        memset(n_highs + n, -1, sizeof(*n_highs) * (num - n));
 192        vfree(t->highs);
 193
 194        t->num_allocated = num;
 195        t->highs = n_highs;
 196        t->targets = n_targets;
 197
 198        return 0;
 199}
 200
 201int dm_table_create(struct dm_table **result, fmode_t mode,
 202                    unsigned num_targets, struct mapped_device *md)
 203{
 204        struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL);
 205
 206        if (!t)
 207                return -ENOMEM;
 208
 209        INIT_LIST_HEAD(&t->devices);
 210        INIT_LIST_HEAD(&t->target_callbacks);
 211        atomic_set(&t->holders, 0);
 212
 213        if (!num_targets)
 214                num_targets = KEYS_PER_NODE;
 215
 216        num_targets = dm_round_up(num_targets, KEYS_PER_NODE);
 217
 218        if (alloc_targets(t, num_targets)) {
 219                kfree(t);
 220                return -ENOMEM;
 221        }
 222
 223        t->mode = mode;
 224        t->md = md;
 225        *result = t;
 226        return 0;
 227}
 228
 229static void free_devices(struct list_head *devices)
 230{
 231        struct list_head *tmp, *next;
 232
 233        list_for_each_safe(tmp, next, devices) {
 234                struct dm_dev_internal *dd =
 235                    list_entry(tmp, struct dm_dev_internal, list);
 236                DMWARN("dm_table_destroy: dm_put_device call missing for %s",
 237                       dd->dm_dev.name);
 238                kfree(dd);
 239        }
 240}
 241
 242void dm_table_destroy(struct dm_table *t)
 243{
 244        unsigned int i;
 245
 246        if (!t)
 247                return;
 248
 249        while (atomic_read(&t->holders))
 250                msleep(1);
 251        smp_mb();
 252
 253        /* free the indexes */
 254        if (t->depth >= 2)
 255                vfree(t->index[t->depth - 2]);
 256
 257        /* free the targets */
 258        for (i = 0; i < t->num_targets; i++) {
 259                struct dm_target *tgt = t->targets + i;
 260
 261                if (tgt->type->dtr)
 262                        tgt->type->dtr(tgt);
 263
 264                dm_put_target_type(tgt->type);
 265        }
 266
 267        vfree(t->highs);
 268
 269        /* free the device list */
 270        free_devices(&t->devices);
 271
 272        dm_free_md_mempools(t->mempools);
 273
 274        kfree(t);
 275}
 276
 277void dm_table_get(struct dm_table *t)
 278{
 279        atomic_inc(&t->holders);
 280}
 281EXPORT_SYMBOL(dm_table_get);
 282
 283void dm_table_put(struct dm_table *t)
 284{
 285        if (!t)
 286                return;
 287
 288        smp_mb__before_atomic_dec();
 289        atomic_dec(&t->holders);
 290}
 291EXPORT_SYMBOL(dm_table_put);
 292
 293/*
 294 * Checks to see if we need to extend highs or targets.
 295 */
 296static inline int check_space(struct dm_table *t)
 297{
 298        if (t->num_targets >= t->num_allocated)
 299                return alloc_targets(t, t->num_allocated * 2);
 300
 301        return 0;
 302}
 303
 304/*
 305 * See if we've already got a device in the list.
 306 */
 307static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev)
 308{
 309        struct dm_dev_internal *dd;
 310
 311        list_for_each_entry (dd, l, list)
 312                if (dd->dm_dev.bdev->bd_dev == dev)
 313                        return dd;
 314
 315        return NULL;
 316}
 317
 318/*
 319 * Open a device so we can use it as a map destination.
 320 */
 321static int open_dev(struct dm_dev_internal *d, dev_t dev,
 322                    struct mapped_device *md)
 323{
 324        static char *_claim_ptr = "I belong to device-mapper";
 325        struct block_device *bdev;
 326
 327        int r;
 328
 329        BUG_ON(d->dm_dev.bdev);
 330
 331        bdev = blkdev_get_by_dev(dev, d->dm_dev.mode | FMODE_EXCL, _claim_ptr);
 332        if (IS_ERR(bdev))
 333                return PTR_ERR(bdev);
 334
 335        r = bd_link_disk_holder(bdev, dm_disk(md));
 336        if (r) {
 337                blkdev_put(bdev, d->dm_dev.mode | FMODE_EXCL);
 338                return r;
 339        }
 340
 341        d->dm_dev.bdev = bdev;
 342        return 0;
 343}
 344
 345/*
 346 * Close a device that we've been using.
 347 */
 348static void close_dev(struct dm_dev_internal *d, struct mapped_device *md)
 349{
 350        if (!d->dm_dev.bdev)
 351                return;
 352
 353        bd_unlink_disk_holder(d->dm_dev.bdev, dm_disk(md));
 354        blkdev_put(d->dm_dev.bdev, d->dm_dev.mode | FMODE_EXCL);
 355        d->dm_dev.bdev = NULL;
 356}
 357
 358/*
 359 * If possible, this checks an area of a destination device is invalid.
 360 */
 361static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev,
 362                                  sector_t start, sector_t len, void *data)
 363{
 364        struct request_queue *q;
 365        struct queue_limits *limits = data;
 366        struct block_device *bdev = dev->bdev;
 367        sector_t dev_size =
 368                i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
 369        unsigned short logical_block_size_sectors =
 370                limits->logical_block_size >> SECTOR_SHIFT;
 371        char b[BDEVNAME_SIZE];
 372
 373        /*
 374         * Some devices exist without request functions,
 375         * such as loop devices not yet bound to backing files.
 376         * Forbid the use of such devices.
 377         */
 378        q = bdev_get_queue(bdev);
 379        if (!q || !q->make_request_fn) {
 380                DMWARN("%s: %s is not yet initialised: "
 381                       "start=%llu, len=%llu, dev_size=%llu",
 382                       dm_device_name(ti->table->md), bdevname(bdev, b),
 383                       (unsigned long long)start,
 384                       (unsigned long long)len,
 385                       (unsigned long long)dev_size);
 386                return 1;
 387        }
 388
 389        if (!dev_size)
 390                return 0;
 391
 392        if ((start >= dev_size) || (start + len > dev_size)) {
 393                DMWARN("%s: %s too small for target: "
 394                       "start=%llu, len=%llu, dev_size=%llu",
 395                       dm_device_name(ti->table->md), bdevname(bdev, b),
 396                       (unsigned long long)start,
 397                       (unsigned long long)len,
 398                       (unsigned long long)dev_size);
 399                return 1;
 400        }
 401
 402        if (logical_block_size_sectors <= 1)
 403                return 0;
 404
 405        if (start & (logical_block_size_sectors - 1)) {
 406                DMWARN("%s: start=%llu not aligned to h/w "
 407                       "logical block size %u of %s",
 408                       dm_device_name(ti->table->md),
 409                       (unsigned long long)start,
 410                       limits->logical_block_size, bdevname(bdev, b));
 411                return 1;
 412        }
 413
 414        if (len & (logical_block_size_sectors - 1)) {
 415                DMWARN("%s: len=%llu not aligned to h/w "
 416                       "logical block size %u of %s",
 417                       dm_device_name(ti->table->md),
 418                       (unsigned long long)len,
 419                       limits->logical_block_size, bdevname(bdev, b));
 420                return 1;
 421        }
 422
 423        return 0;
 424}
 425
 426/*
 427 * This upgrades the mode on an already open dm_dev, being
 428 * careful to leave things as they were if we fail to reopen the
 429 * device and not to touch the existing bdev field in case
 430 * it is accessed concurrently inside dm_table_any_congested().
 431 */
 432static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
 433                        struct mapped_device *md)
 434{
 435        int r;
 436        struct dm_dev_internal dd_new, dd_old;
 437
 438        dd_new = dd_old = *dd;
 439
 440        dd_new.dm_dev.mode |= new_mode;
 441        dd_new.dm_dev.bdev = NULL;
 442
 443        r = open_dev(&dd_new, dd->dm_dev.bdev->bd_dev, md);
 444        if (r)
 445                return r;
 446
 447        dd->dm_dev.mode |= new_mode;
 448        close_dev(&dd_old, md);
 449
 450        return 0;
 451}
 452
 453/*
 454 * Add a device to the list, or just increment the usage count if
 455 * it's already present.
 456 */
 457int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
 458                  struct dm_dev **result)
 459{
 460        int r;
 461        dev_t uninitialized_var(dev);
 462        struct dm_dev_internal *dd;
 463        unsigned int major, minor;
 464        struct dm_table *t = ti->table;
 465        char dummy;
 466
 467        BUG_ON(!t);
 468
 469        if (sscanf(path, "%u:%u%c", &major, &minor, &dummy) == 2) {
 470                /* Extract the major/minor numbers */
 471                dev = MKDEV(major, minor);
 472                if (MAJOR(dev) != major || MINOR(dev) != minor)
 473                        return -EOVERFLOW;
 474        } else {
 475                /* convert the path to a device */
 476                struct block_device *bdev = lookup_bdev(path);
 477
 478                if (IS_ERR(bdev))
 479                        return PTR_ERR(bdev);
 480                dev = bdev->bd_dev;
 481                bdput(bdev);
 482        }
 483
 484        dd = find_device(&t->devices, dev);
 485        if (!dd) {
 486                dd = kmalloc(sizeof(*dd), GFP_KERNEL);
 487                if (!dd)
 488                        return -ENOMEM;
 489
 490                dd->dm_dev.mode = mode;
 491                dd->dm_dev.bdev = NULL;
 492
 493                if ((r = open_dev(dd, dev, t->md))) {
 494                        kfree(dd);
 495                        return r;
 496                }
 497
 498                format_dev_t(dd->dm_dev.name, dev);
 499
 500                atomic_set(&dd->count, 0);
 501                list_add(&dd->list, &t->devices);
 502
 503        } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) {
 504                r = upgrade_mode(dd, mode, t->md);
 505                if (r)
 506                        return r;
 507        }
 508        atomic_inc(&dd->count);
 509
 510        *result = &dd->dm_dev;
 511        return 0;
 512}
 513EXPORT_SYMBOL(dm_get_device);
 514
 515int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev,
 516                         sector_t start, sector_t len, void *data)
 517{
 518        struct queue_limits *limits = data;
 519        struct block_device *bdev = dev->bdev;
 520        struct request_queue *q = bdev_get_queue(bdev);
 521        char b[BDEVNAME_SIZE];
 522
 523        if (unlikely(!q)) {
 524                DMWARN("%s: Cannot set limits for nonexistent device %s",
 525                       dm_device_name(ti->table->md), bdevname(bdev, b));
 526                return 0;
 527        }
 528
 529        if (bdev_stack_limits(limits, bdev, start) < 0)
 530                DMWARN("%s: adding target device %s caused an alignment inconsistency: "
 531                       "physical_block_size=%u, logical_block_size=%u, "
 532                       "alignment_offset=%u, start=%llu",
 533                       dm_device_name(ti->table->md), bdevname(bdev, b),
 534                       q->limits.physical_block_size,
 535                       q->limits.logical_block_size,
 536                       q->limits.alignment_offset,
 537                       (unsigned long long) start << SECTOR_SHIFT);
 538
 539        /*
 540         * Check if merge fn is supported.
 541         * If not we'll force DM to use PAGE_SIZE or
 542         * smaller I/O, just to be safe.
 543         */
 544        if (dm_queue_merge_is_compulsory(q) && !ti->type->merge)
 545                blk_limits_max_hw_sectors(limits,
 546                                          (unsigned int) (PAGE_SIZE >> 9));
 547        return 0;
 548}
 549EXPORT_SYMBOL_GPL(dm_set_device_limits);
 550
 551/*
 552 * Decrement a device's use count and remove it if necessary.
 553 */
 554void dm_put_device(struct dm_target *ti, struct dm_dev *d)
 555{
 556        struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal,
 557                                                  dm_dev);
 558
 559        if (atomic_dec_and_test(&dd->count)) {
 560                close_dev(dd, ti->table->md);
 561                list_del(&dd->list);
 562                kfree(dd);
 563        }
 564}
 565EXPORT_SYMBOL(dm_put_device);
 566
 567/*
 568 * Checks to see if the target joins onto the end of the table.
 569 */
 570static int adjoin(struct dm_table *table, struct dm_target *ti)
 571{
 572        struct dm_target *prev;
 573
 574        if (!table->num_targets)
 575                return !ti->begin;
 576
 577        prev = &table->targets[table->num_targets - 1];
 578        return (ti->begin == (prev->begin + prev->len));
 579}
 580
 581/*
 582 * Used to dynamically allocate the arg array.
 583 */
 584static char **realloc_argv(unsigned *array_size, char **old_argv)
 585{
 586        char **argv;
 587        unsigned new_size;
 588
 589        new_size = *array_size ? *array_size * 2 : 64;
 590        argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL);
 591        if (argv) {
 592                memcpy(argv, old_argv, *array_size * sizeof(*argv));
 593                *array_size = new_size;
 594        }
 595
 596        kfree(old_argv);
 597        return argv;
 598}
 599
 600/*
 601 * Destructively splits up the argument list to pass to ctr.
 602 */
 603int dm_split_args(int *argc, char ***argvp, char *input)
 604{
 605        char *start, *end = input, *out, **argv = NULL;
 606        unsigned array_size = 0;
 607
 608        *argc = 0;
 609
 610        if (!input) {
 611                *argvp = NULL;
 612                return 0;
 613        }
 614
 615        argv = realloc_argv(&array_size, argv);
 616        if (!argv)
 617                return -ENOMEM;
 618
 619        while (1) {
 620                /* Skip whitespace */
 621                start = skip_spaces(end);
 622
 623                if (!*start)
 624                        break;  /* success, we hit the end */
 625
 626                /* 'out' is used to remove any back-quotes */
 627                end = out = start;
 628                while (*end) {
 629                        /* Everything apart from '\0' can be quoted */
 630                        if (*end == '\\' && *(end + 1)) {
 631                                *out++ = *(end + 1);
 632                                end += 2;
 633                                continue;
 634                        }
 635
 636                        if (isspace(*end))
 637                                break;  /* end of token */
 638
 639                        *out++ = *end++;
 640                }
 641
 642                /* have we already filled the array ? */
 643                if ((*argc + 1) > array_size) {
 644                        argv = realloc_argv(&array_size, argv);
 645                        if (!argv)
 646                                return -ENOMEM;
 647                }
 648
 649                /* we know this is whitespace */
 650                if (*end)
 651                        end++;
 652
 653                /* terminate the string and put it in the array */
 654                *out = '\0';
 655                argv[*argc] = start;
 656                (*argc)++;
 657        }
 658
 659        *argvp = argv;
 660        return 0;
 661}
 662
 663/*
 664 * Impose necessary and sufficient conditions on a devices's table such
 665 * that any incoming bio which respects its logical_block_size can be
 666 * processed successfully.  If it falls across the boundary between
 667 * two or more targets, the size of each piece it gets split into must
 668 * be compatible with the logical_block_size of the target processing it.
 669 */
 670static int validate_hardware_logical_block_alignment(struct dm_table *table,
 671                                                 struct queue_limits *limits)
 672{
 673        /*
 674         * This function uses arithmetic modulo the logical_block_size
 675         * (in units of 512-byte sectors).
 676         */
 677        unsigned short device_logical_block_size_sects =
 678                limits->logical_block_size >> SECTOR_SHIFT;
 679
 680        /*
 681         * Offset of the start of the next table entry, mod logical_block_size.
 682         */
 683        unsigned short next_target_start = 0;
 684
 685        /*
 686         * Given an aligned bio that extends beyond the end of a
 687         * target, how many sectors must the next target handle?
 688         */
 689        unsigned short remaining = 0;
 690
 691        struct dm_target *uninitialized_var(ti);
 692        struct queue_limits ti_limits;
 693        unsigned i = 0;
 694
 695        /*
 696         * Check each entry in the table in turn.
 697         */
 698        while (i < dm_table_get_num_targets(table)) {
 699                ti = dm_table_get_target(table, i++);
 700
 701                blk_set_stacking_limits(&ti_limits);
 702
 703                /* combine all target devices' limits */
 704                if (ti->type->iterate_devices)
 705                        ti->type->iterate_devices(ti, dm_set_device_limits,
 706                                                  &ti_limits);
 707
 708                /*
 709                 * If the remaining sectors fall entirely within this
 710                 * table entry are they compatible with its logical_block_size?
 711                 */
 712                if (remaining < ti->len &&
 713                    remaining & ((ti_limits.logical_block_size >>
 714                                  SECTOR_SHIFT) - 1))
 715                        break;  /* Error */
 716
 717                next_target_start =
 718                    (unsigned short) ((next_target_start + ti->len) &
 719                                      (device_logical_block_size_sects - 1));
 720                remaining = next_target_start ?
 721                    device_logical_block_size_sects - next_target_start : 0;
 722        }
 723
 724        if (remaining) {
 725                DMWARN("%s: table line %u (start sect %llu len %llu) "
 726                       "not aligned to h/w logical block size %u",
 727                       dm_device_name(table->md), i,
 728                       (unsigned long long) ti->begin,
 729                       (unsigned long long) ti->len,
 730                       limits->logical_block_size);
 731                return -EINVAL;
 732        }
 733
 734        return 0;
 735}
 736
 737int dm_table_add_target(struct dm_table *t, const char *type,
 738                        sector_t start, sector_t len, char *params)
 739{
 740        int r = -EINVAL, argc;
 741        char **argv;
 742        struct dm_target *tgt;
 743
 744        if (t->singleton) {
 745                DMERR("%s: target type %s must appear alone in table",
 746                      dm_device_name(t->md), t->targets->type->name);
 747                return -EINVAL;
 748        }
 749
 750        if ((r = check_space(t)))
 751                return r;
 752
 753        tgt = t->targets + t->num_targets;
 754        memset(tgt, 0, sizeof(*tgt));
 755
 756        if (!len) {
 757                DMERR("%s: zero-length target", dm_device_name(t->md));
 758                return -EINVAL;
 759        }
 760
 761        tgt->type = dm_get_target_type(type);
 762        if (!tgt->type) {
 763                DMERR("%s: %s: unknown target type", dm_device_name(t->md),
 764                      type);
 765                return -EINVAL;
 766        }
 767
 768        if (dm_target_needs_singleton(tgt->type)) {
 769                if (t->num_targets) {
 770                        DMERR("%s: target type %s must appear alone in table",
 771                              dm_device_name(t->md), type);
 772                        return -EINVAL;
 773                }
 774                t->singleton = 1;
 775        }
 776
 777        if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) {
 778                DMERR("%s: target type %s may not be included in read-only tables",
 779                      dm_device_name(t->md), type);
 780                return -EINVAL;
 781        }
 782
 783        if (t->immutable_target_type) {
 784                if (t->immutable_target_type != tgt->type) {
 785                        DMERR("%s: immutable target type %s cannot be mixed with other target types",
 786                              dm_device_name(t->md), t->immutable_target_type->name);
 787                        return -EINVAL;
 788                }
 789        } else if (dm_target_is_immutable(tgt->type)) {
 790                if (t->num_targets) {
 791                        DMERR("%s: immutable target type %s cannot be mixed with other target types",
 792                              dm_device_name(t->md), tgt->type->name);
 793                        return -EINVAL;
 794                }
 795                t->immutable_target_type = tgt->type;
 796        }
 797
 798        tgt->table = t;
 799        tgt->begin = start;
 800        tgt->len = len;
 801        tgt->error = "Unknown error";
 802
 803        /*
 804         * Does this target adjoin the previous one ?
 805         */
 806        if (!adjoin(t, tgt)) {
 807                tgt->error = "Gap in table";
 808                r = -EINVAL;
 809                goto bad;
 810        }
 811
 812        r = dm_split_args(&argc, &argv, params);
 813        if (r) {
 814                tgt->error = "couldn't split parameters (insufficient memory)";
 815                goto bad;
 816        }
 817
 818        r = tgt->type->ctr(tgt, argc, argv);
 819        kfree(argv);
 820        if (r)
 821                goto bad;
 822
 823        t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
 824
 825        if (!tgt->num_discard_bios && tgt->discards_supported)
 826                DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.",
 827                       dm_device_name(t->md), type);
 828
 829        return 0;
 830
 831 bad:
 832        DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
 833        dm_put_target_type(tgt->type);
 834        return r;
 835}
 836
 837/*
 838 * Target argument parsing helpers.
 839 */
 840static int validate_next_arg(struct dm_arg *arg, struct dm_arg_set *arg_set,
 841                             unsigned *value, char **error, unsigned grouped)
 842{
 843        const char *arg_str = dm_shift_arg(arg_set);
 844        char dummy;
 845
 846        if (!arg_str ||
 847            (sscanf(arg_str, "%u%c", value, &dummy) != 1) ||
 848            (*value < arg->min) ||
 849            (*value > arg->max) ||
 850            (grouped && arg_set->argc < *value)) {
 851                *error = arg->error;
 852                return -EINVAL;
 853        }
 854
 855        return 0;
 856}
 857
 858int dm_read_arg(struct dm_arg *arg, struct dm_arg_set *arg_set,
 859                unsigned *value, char **error)
 860{
 861        return validate_next_arg(arg, arg_set, value, error, 0);
 862}
 863EXPORT_SYMBOL(dm_read_arg);
 864
 865int dm_read_arg_group(struct dm_arg *arg, struct dm_arg_set *arg_set,
 866                      unsigned *value, char **error)
 867{
 868        return validate_next_arg(arg, arg_set, value, error, 1);
 869}
 870EXPORT_SYMBOL(dm_read_arg_group);
 871
 872const char *dm_shift_arg(struct dm_arg_set *as)
 873{
 874        char *r;
 875
 876        if (as->argc) {
 877                as->argc--;
 878                r = *as->argv;
 879                as->argv++;
 880                return r;
 881        }
 882
 883        return NULL;
 884}
 885EXPORT_SYMBOL(dm_shift_arg);
 886
 887void dm_consume_args(struct dm_arg_set *as, unsigned num_args)
 888{
 889        BUG_ON(as->argc < num_args);
 890        as->argc -= num_args;
 891        as->argv += num_args;
 892}
 893EXPORT_SYMBOL(dm_consume_args);
 894
 895static int dm_table_set_type(struct dm_table *t)
 896{
 897        unsigned i;
 898        unsigned bio_based = 0, request_based = 0;
 899        struct dm_target *tgt;
 900        struct dm_dev_internal *dd;
 901        struct list_head *devices;
 902
 903        for (i = 0; i < t->num_targets; i++) {
 904                tgt = t->targets + i;
 905                if (dm_target_request_based(tgt))
 906                        request_based = 1;
 907                else
 908                        bio_based = 1;
 909
 910                if (bio_based && request_based) {
 911                        DMWARN("Inconsistent table: different target types"
 912                               " can't be mixed up");
 913                        return -EINVAL;
 914                }
 915        }
 916
 917        if (bio_based) {
 918                /* We must use this table as bio-based */
 919                t->type = DM_TYPE_BIO_BASED;
 920                return 0;
 921        }
 922
 923        BUG_ON(!request_based); /* No targets in this table */
 924
 925        /* Non-request-stackable devices can't be used for request-based dm */
 926        devices = dm_table_get_devices(t);
 927        list_for_each_entry(dd, devices, list) {
 928                if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev.bdev))) {
 929                        DMWARN("table load rejected: including"
 930                               " non-request-stackable devices");
 931                        return -EINVAL;
 932                }
 933        }
 934
 935        /*
 936         * Request-based dm supports only tables that have a single target now.
 937         * To support multiple targets, request splitting support is needed,
 938         * and that needs lots of changes in the block-layer.
 939         * (e.g. request completion process for partial completion.)
 940         */
 941        if (t->num_targets > 1) {
 942                DMWARN("Request-based dm doesn't support multiple targets yet");
 943                return -EINVAL;
 944        }
 945
 946        t->type = DM_TYPE_REQUEST_BASED;
 947
 948        return 0;
 949}
 950
 951unsigned dm_table_get_type(struct dm_table *t)
 952{
 953        return t->type;
 954}
 955
 956struct target_type *dm_table_get_immutable_target_type(struct dm_table *t)
 957{
 958        return t->immutable_target_type;
 959}
 960
 961bool dm_table_request_based(struct dm_table *t)
 962{
 963        return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED;
 964}
 965
 966int dm_table_alloc_md_mempools(struct dm_table *t)
 967{
 968        unsigned type = dm_table_get_type(t);
 969        unsigned per_bio_data_size = 0;
 970        struct dm_target *tgt;
 971        unsigned i;
 972
 973        if (unlikely(type == DM_TYPE_NONE)) {
 974                DMWARN("no table type is set, can't allocate mempools");
 975                return -EINVAL;
 976        }
 977
 978        if (type == DM_TYPE_BIO_BASED)
 979                for (i = 0; i < t->num_targets; i++) {
 980                        tgt = t->targets + i;
 981                        per_bio_data_size = max(per_bio_data_size, tgt->per_bio_data_size);
 982                }
 983
 984        t->mempools = dm_alloc_md_mempools(type, t->integrity_supported, per_bio_data_size);
 985        if (!t->mempools)
 986                return -ENOMEM;
 987
 988        return 0;
 989}
 990
 991void dm_table_free_md_mempools(struct dm_table *t)
 992{
 993        dm_free_md_mempools(t->mempools);
 994        t->mempools = NULL;
 995}
 996
 997struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t)
 998{
 999        return t->mempools;
1000}
1001
1002static int setup_indexes(struct dm_table *t)
1003{
1004        int i;
1005        unsigned int total = 0;
1006        sector_t *indexes;
1007
1008        /* allocate the space for *all* the indexes */
1009        for (i = t->depth - 2; i >= 0; i--) {
1010                t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
1011                total += t->counts[i];
1012        }
1013
1014        indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
1015        if (!indexes)
1016                return -ENOMEM;
1017
1018        /* set up internal nodes, bottom-up */
1019        for (i = t->depth - 2; i >= 0; i--) {
1020                t->index[i] = indexes;
1021                indexes += (KEYS_PER_NODE * t->counts[i]);
1022                setup_btree_index(i, t);
1023        }
1024
1025        return 0;
1026}
1027
1028/*
1029 * Builds the btree to index the map.
1030 */
1031static int dm_table_build_index(struct dm_table *t)
1032{
1033        int r = 0;
1034        unsigned int leaf_nodes;
1035
1036        /* how many indexes will the btree have ? */
1037        leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
1038        t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
1039
1040        /* leaf layer has already been set up */
1041        t->counts[t->depth - 1] = leaf_nodes;
1042        t->index[t->depth - 1] = t->highs;
1043
1044        if (t->depth >= 2)
1045                r = setup_indexes(t);
1046
1047        return r;
1048}
1049
1050/*
1051 * Get a disk whose integrity profile reflects the table's profile.
1052 * If %match_all is true, all devices' profiles must match.
1053 * If %match_all is false, all devices must at least have an
1054 * allocated integrity profile; but uninitialized is ok.
1055 * Returns NULL if integrity support was inconsistent or unavailable.
1056 */
1057static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t,
1058                                                    bool match_all)
1059{
1060        struct list_head *devices = dm_table_get_devices(t);
1061        struct dm_dev_internal *dd = NULL;
1062        struct gendisk *prev_disk = NULL, *template_disk = NULL;
1063
1064        list_for_each_entry(dd, devices, list) {
1065                template_disk = dd->dm_dev.bdev->bd_disk;
1066                if (!blk_get_integrity(template_disk))
1067                        goto no_integrity;
1068                if (!match_all && !blk_integrity_is_initialized(template_disk))
1069                        continue; /* skip uninitialized profiles */
1070                else if (prev_disk &&
1071                         blk_integrity_compare(prev_disk, template_disk) < 0)
1072                        goto no_integrity;
1073                prev_disk = template_disk;
1074        }
1075
1076        return template_disk;
1077
1078no_integrity:
1079        if (prev_disk)
1080                DMWARN("%s: integrity not set: %s and %s profile mismatch",
1081                       dm_device_name(t->md),
1082                       prev_disk->disk_name,
1083                       template_disk->disk_name);
1084        return NULL;
1085}
1086
1087/*
1088 * Register the mapped device for blk_integrity support if
1089 * the underlying devices have an integrity profile.  But all devices
1090 * may not have matching profiles (checking all devices isn't reliable
1091 * during table load because this table may use other DM device(s) which
1092 * must be resumed before they will have an initialized integity profile).
1093 * Stacked DM devices force a 2 stage integrity profile validation:
1094 * 1 - during load, validate all initialized integrity profiles match
1095 * 2 - during resume, validate all integrity profiles match
1096 */
1097static int dm_table_prealloc_integrity(struct dm_table *t, struct mapped_device *md)
1098{
1099        struct gendisk *template_disk = NULL;
1100
1101        template_disk = dm_table_get_integrity_disk(t, false);
1102        if (!template_disk)
1103                return 0;
1104
1105        if (!blk_integrity_is_initialized(dm_disk(md))) {
1106                t->integrity_supported = 1;
1107                return blk_integrity_register(dm_disk(md), NULL);
1108        }
1109
1110        /*
1111         * If DM device already has an initalized integrity
1112         * profile the new profile should not conflict.
1113         */
1114        if (blk_integrity_is_initialized(template_disk) &&
1115            blk_integrity_compare(dm_disk(md), template_disk) < 0) {
1116                DMWARN("%s: conflict with existing integrity profile: "
1117                       "%s profile mismatch",
1118                       dm_device_name(t->md),
1119                       template_disk->disk_name);
1120                return 1;
1121        }
1122
1123        /* Preserve existing initialized integrity profile */
1124        t->integrity_supported = 1;
1125        return 0;
1126}
1127
1128/*
1129 * Prepares the table for use by building the indices,
1130 * setting the type, and allocating mempools.
1131 */
1132int dm_table_complete(struct dm_table *t)
1133{
1134        int r;
1135
1136        r = dm_table_set_type(t);
1137        if (r) {
1138                DMERR("unable to set table type");
1139                return r;
1140        }
1141
1142        r = dm_table_build_index(t);
1143        if (r) {
1144                DMERR("unable to build btrees");
1145                return r;
1146        }
1147
1148        r = dm_table_prealloc_integrity(t, t->md);
1149        if (r) {
1150                DMERR("could not register integrity profile.");
1151                return r;
1152        }
1153
1154        r = dm_table_alloc_md_mempools(t);
1155        if (r)
1156                DMERR("unable to allocate mempools");
1157
1158        return r;
1159}
1160
1161static DEFINE_MUTEX(_event_lock);
1162void dm_table_event_callback(struct dm_table *t,
1163                             void (*fn)(void *), void *context)
1164{
1165        mutex_lock(&_event_lock);
1166        t->event_fn = fn;
1167        t->event_context = context;
1168        mutex_unlock(&_event_lock);
1169}
1170
1171void dm_table_event(struct dm_table *t)
1172{
1173        /*
1174         * You can no longer call dm_table_event() from interrupt
1175         * context, use a bottom half instead.
1176         */
1177        BUG_ON(in_interrupt());
1178
1179        mutex_lock(&_event_lock);
1180        if (t->event_fn)
1181                t->event_fn(t->event_context);
1182        mutex_unlock(&_event_lock);
1183}
1184EXPORT_SYMBOL(dm_table_event);
1185
1186sector_t dm_table_get_size(struct dm_table *t)
1187{
1188        return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
1189}
1190EXPORT_SYMBOL(dm_table_get_size);
1191
1192struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
1193{
1194        if (index >= t->num_targets)
1195                return NULL;
1196
1197        return t->targets + index;
1198}
1199
1200/*
1201 * Search the btree for the correct target.
1202 *
1203 * Caller should check returned pointer with dm_target_is_valid()
1204 * to trap I/O beyond end of device.
1205 */
1206struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
1207{
1208        unsigned int l, n = 0, k = 0;
1209        sector_t *node;
1210
1211        for (l = 0; l < t->depth; l++) {
1212                n = get_child(n, k);
1213                node = get_node(t, l, n);
1214
1215                for (k = 0; k < KEYS_PER_NODE; k++)
1216                        if (node[k] >= sector)
1217                                break;
1218        }
1219
1220        return &t->targets[(KEYS_PER_NODE * n) + k];
1221}
1222
1223static int count_device(struct dm_target *ti, struct dm_dev *dev,
1224                        sector_t start, sector_t len, void *data)
1225{
1226        unsigned *num_devices = data;
1227
1228        (*num_devices)++;
1229
1230        return 0;
1231}
1232
1233/*
1234 * Check whether a table has no data devices attached using each
1235 * target's iterate_devices method.
1236 * Returns false if the result is unknown because a target doesn't
1237 * support iterate_devices.
1238 */
1239bool dm_table_has_no_data_devices(struct dm_table *table)
1240{
1241        struct dm_target *uninitialized_var(ti);
1242        unsigned i = 0, num_devices = 0;
1243
1244        while (i < dm_table_get_num_targets(table)) {
1245                ti = dm_table_get_target(table, i++);
1246
1247                if (!ti->type->iterate_devices)
1248                        return false;
1249
1250                ti->type->iterate_devices(ti, count_device, &num_devices);
1251                if (num_devices)
1252                        return false;
1253        }
1254
1255        return true;
1256}
1257
1258/*
1259 * Establish the new table's queue_limits and validate them.
1260 */
1261int dm_calculate_queue_limits(struct dm_table *table,
1262                              struct queue_limits *limits)
1263{
1264        struct dm_target *uninitialized_var(ti);
1265        struct queue_limits ti_limits;
1266        unsigned i = 0;
1267
1268        blk_set_stacking_limits(limits);
1269
1270        while (i < dm_table_get_num_targets(table)) {
1271                blk_set_stacking_limits(&ti_limits);
1272
1273                ti = dm_table_get_target(table, i++);
1274
1275                if (!ti->type->iterate_devices)
1276                        goto combine_limits;
1277
1278                /*
1279                 * Combine queue limits of all the devices this target uses.
1280                 */
1281                ti->type->iterate_devices(ti, dm_set_device_limits,
1282                                          &ti_limits);
1283
1284                /* Set I/O hints portion of queue limits */
1285                if (ti->type->io_hints)
1286                        ti->type->io_hints(ti, &ti_limits);
1287
1288                /*
1289                 * Check each device area is consistent with the target's
1290                 * overall queue limits.
1291                 */
1292                if (ti->type->iterate_devices(ti, device_area_is_invalid,
1293                                              &ti_limits))
1294                        return -EINVAL;
1295
1296combine_limits:
1297                /*
1298                 * Merge this target's queue limits into the overall limits
1299                 * for the table.
1300                 */
1301                if (blk_stack_limits(limits, &ti_limits, 0) < 0)
1302                        DMWARN("%s: adding target device "
1303                               "(start sect %llu len %llu) "
1304                               "caused an alignment inconsistency",
1305                               dm_device_name(table->md),
1306                               (unsigned long long) ti->begin,
1307                               (unsigned long long) ti->len);
1308        }
1309
1310        return validate_hardware_logical_block_alignment(table, limits);
1311}
1312
1313/*
1314 * Set the integrity profile for this device if all devices used have
1315 * matching profiles.  We're quite deep in the resume path but still
1316 * don't know if all devices (particularly DM devices this device
1317 * may be stacked on) have matching profiles.  Even if the profiles
1318 * don't match we have no way to fail (to resume) at this point.
1319 */
1320static void dm_table_set_integrity(struct dm_table *t)
1321{
1322        struct gendisk *template_disk = NULL;
1323
1324        if (!blk_get_integrity(dm_disk(t->md)))
1325                return;
1326
1327        template_disk = dm_table_get_integrity_disk(t, true);
1328        if (template_disk)
1329                blk_integrity_register(dm_disk(t->md),
1330                                       blk_get_integrity(template_disk));
1331        else if (blk_integrity_is_initialized(dm_disk(t->md)))
1332                DMWARN("%s: device no longer has a valid integrity profile",
1333                       dm_device_name(t->md));
1334        else
1335                DMWARN("%s: unable to establish an integrity profile",
1336                       dm_device_name(t->md));
1337}
1338
1339static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev,
1340                                sector_t start, sector_t len, void *data)
1341{
1342        unsigned flush = (*(unsigned *)data);
1343        struct request_queue *q = bdev_get_queue(dev->bdev);
1344
1345        return q && (q->flush_flags & flush);
1346}
1347
1348static bool dm_table_supports_flush(struct dm_table *t, unsigned flush)
1349{
1350        struct dm_target *ti;
1351        unsigned i = 0;
1352
1353        /*
1354         * Require at least one underlying device to support flushes.
1355         * t->devices includes internal dm devices such as mirror logs
1356         * so we need to use iterate_devices here, which targets
1357         * supporting flushes must provide.
1358         */
1359        while (i < dm_table_get_num_targets(t)) {
1360                ti = dm_table_get_target(t, i++);
1361
1362                if (!ti->num_flush_bios)
1363                        continue;
1364
1365                if (ti->flush_supported)
1366                        return 1;
1367
1368                if (ti->type->iterate_devices &&
1369                    ti->type->iterate_devices(ti, device_flush_capable, &flush))
1370                        return 1;
1371        }
1372
1373        return 0;
1374}
1375
1376static bool dm_table_discard_zeroes_data(struct dm_table *t)
1377{
1378        struct dm_target *ti;
1379        unsigned i = 0;
1380
1381        /* Ensure that all targets supports discard_zeroes_data. */
1382        while (i < dm_table_get_num_targets(t)) {
1383                ti = dm_table_get_target(t, i++);
1384
1385                if (ti->discard_zeroes_data_unsupported)
1386                        return 0;
1387        }
1388
1389        return 1;
1390}
1391
1392static int device_is_nonrot(struct dm_target *ti, struct dm_dev *dev,
1393                            sector_t start, sector_t len, void *data)
1394{
1395        struct request_queue *q = bdev_get_queue(dev->bdev);
1396
1397        return q && blk_queue_nonrot(q);
1398}
1399
1400static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev,
1401                             sector_t start, sector_t len, void *data)
1402{
1403        struct request_queue *q = bdev_get_queue(dev->bdev);
1404
1405        return q && !blk_queue_add_random(q);
1406}
1407
1408static bool dm_table_all_devices_attribute(struct dm_table *t,
1409                                           iterate_devices_callout_fn func)
1410{
1411        struct dm_target *ti;
1412        unsigned i = 0;
1413
1414        while (i < dm_table_get_num_targets(t)) {
1415                ti = dm_table_get_target(t, i++);
1416
1417                if (!ti->type->iterate_devices ||
1418                    !ti->type->iterate_devices(ti, func, NULL))
1419                        return 0;
1420        }
1421
1422        return 1;
1423}
1424
1425static int device_not_write_same_capable(struct dm_target *ti, struct dm_dev *dev,
1426                                         sector_t start, sector_t len, void *data)
1427{
1428        struct request_queue *q = bdev_get_queue(dev->bdev);
1429
1430        return q && !q->limits.max_write_same_sectors;
1431}
1432
1433static bool dm_table_supports_write_same(struct dm_table *t)
1434{
1435        struct dm_target *ti;
1436        unsigned i = 0;
1437
1438        while (i < dm_table_get_num_targets(t)) {
1439                ti = dm_table_get_target(t, i++);
1440
1441                if (!ti->num_write_same_bios)
1442                        return false;
1443
1444                if (!ti->type->iterate_devices ||
1445                    ti->type->iterate_devices(ti, device_not_write_same_capable, NULL))
1446                        return false;
1447        }
1448
1449        return true;
1450}
1451
1452void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
1453                               struct queue_limits *limits)
1454{
1455        unsigned flush = 0;
1456
1457        /*
1458         * Copy table's limits to the DM device's request_queue
1459         */
1460        q->limits = *limits;
1461
1462        if (!dm_table_supports_discards(t))
1463                queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q);
1464        else
1465                queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q);
1466
1467        if (dm_table_supports_flush(t, REQ_FLUSH)) {
1468                flush |= REQ_FLUSH;
1469                if (dm_table_supports_flush(t, REQ_FUA))
1470                        flush |= REQ_FUA;
1471        }
1472        blk_queue_flush(q, flush);
1473
1474        if (!dm_table_discard_zeroes_data(t))
1475                q->limits.discard_zeroes_data = 0;
1476
1477        /* Ensure that all underlying devices are non-rotational. */
1478        if (dm_table_all_devices_attribute(t, device_is_nonrot))
1479                queue_flag_set_unlocked(QUEUE_FLAG_NONROT, q);
1480        else
1481                queue_flag_clear_unlocked(QUEUE_FLAG_NONROT, q);
1482
1483        if (!dm_table_supports_write_same(t))
1484                q->limits.max_write_same_sectors = 0;
1485
1486        dm_table_set_integrity(t);
1487
1488        /*
1489         * Determine whether or not this queue's I/O timings contribute
1490         * to the entropy pool, Only request-based targets use this.
1491         * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not
1492         * have it set.
1493         */
1494        if (blk_queue_add_random(q) && dm_table_all_devices_attribute(t, device_is_not_random))
1495                queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, q);
1496
1497        /*
1498         * QUEUE_FLAG_STACKABLE must be set after all queue settings are
1499         * visible to other CPUs because, once the flag is set, incoming bios
1500         * are processed by request-based dm, which refers to the queue
1501         * settings.
1502         * Until the flag set, bios are passed to bio-based dm and queued to
1503         * md->deferred where queue settings are not needed yet.
1504         * Those bios are passed to request-based dm at the resume time.
1505         */
1506        smp_mb();
1507        if (dm_table_request_based(t))
1508                queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q);
1509}
1510
1511unsigned int dm_table_get_num_targets(struct dm_table *t)
1512{
1513        return t->num_targets;
1514}
1515
1516struct list_head *dm_table_get_devices(struct dm_table *t)
1517{
1518        return &t->devices;
1519}
1520
1521fmode_t dm_table_get_mode(struct dm_table *t)
1522{
1523        return t->mode;
1524}
1525EXPORT_SYMBOL(dm_table_get_mode);
1526
1527static void suspend_targets(struct dm_table *t, unsigned postsuspend)
1528{
1529        int i = t->num_targets;
1530        struct dm_target *ti = t->targets;
1531
1532        while (i--) {
1533                if (postsuspend) {
1534                        if (ti->type->postsuspend)
1535                                ti->type->postsuspend(ti);
1536                } else if (ti->type->presuspend)
1537                        ti->type->presuspend(ti);
1538
1539                ti++;
1540        }
1541}
1542
1543void dm_table_presuspend_targets(struct dm_table *t)
1544{
1545        if (!t)
1546                return;
1547
1548        suspend_targets(t, 0);
1549}
1550
1551void dm_table_postsuspend_targets(struct dm_table *t)
1552{
1553        if (!t)
1554                return;
1555
1556        suspend_targets(t, 1);
1557}
1558
1559int dm_table_resume_targets(struct dm_table *t)
1560{
1561        int i, r = 0;
1562
1563        for (i = 0; i < t->num_targets; i++) {
1564                struct dm_target *ti = t->targets + i;
1565
1566                if (!ti->type->preresume)
1567                        continue;
1568
1569                r = ti->type->preresume(ti);
1570                if (r)
1571                        return r;
1572        }
1573
1574        for (i = 0; i < t->num_targets; i++) {
1575                struct dm_target *ti = t->targets + i;
1576
1577                if (ti->type->resume)
1578                        ti->type->resume(ti);
1579        }
1580
1581        return 0;
1582}
1583
1584void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb)
1585{
1586        list_add(&cb->list, &t->target_callbacks);
1587}
1588EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks);
1589
1590int dm_table_any_congested(struct dm_table *t, int bdi_bits)
1591{
1592        struct dm_dev_internal *dd;
1593        struct list_head *devices = dm_table_get_devices(t);
1594        struct dm_target_callbacks *cb;
1595        int r = 0;
1596
1597        list_for_each_entry(dd, devices, list) {
1598                struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev);
1599                char b[BDEVNAME_SIZE];
1600
1601                if (likely(q))
1602                        r |= bdi_congested(&q->backing_dev_info, bdi_bits);
1603                else
1604                        DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
1605                                     dm_device_name(t->md),
1606                                     bdevname(dd->dm_dev.bdev, b));
1607        }
1608
1609        list_for_each_entry(cb, &t->target_callbacks, list)
1610                if (cb->congested_fn)
1611                        r |= cb->congested_fn(cb, bdi_bits);
1612
1613        return r;
1614}
1615
1616int dm_table_any_busy_target(struct dm_table *t)
1617{
1618        unsigned i;
1619        struct dm_target *ti;
1620
1621        for (i = 0; i < t->num_targets; i++) {
1622                ti = t->targets + i;
1623                if (ti->type->busy && ti->type->busy(ti))
1624                        return 1;
1625        }
1626
1627        return 0;
1628}
1629
1630struct mapped_device *dm_table_get_md(struct dm_table *t)
1631{
1632        return t->md;
1633}
1634EXPORT_SYMBOL(dm_table_get_md);
1635
1636static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev,
1637                                  sector_t start, sector_t len, void *data)
1638{
1639        struct request_queue *q = bdev_get_queue(dev->bdev);
1640
1641        return q && blk_queue_discard(q);
1642}
1643
1644bool dm_table_supports_discards(struct dm_table *t)
1645{
1646        struct dm_target *ti;
1647        unsigned i = 0;
1648
1649        /*
1650         * Unless any target used by the table set discards_supported,
1651         * require at least one underlying device to support discards.
1652         * t->devices includes internal dm devices such as mirror logs
1653         * so we need to use iterate_devices here, which targets
1654         * supporting discard selectively must provide.
1655         */
1656        while (i < dm_table_get_num_targets(t)) {
1657                ti = dm_table_get_target(t, i++);
1658
1659                if (!ti->num_discard_bios)
1660                        continue;
1661
1662                if (ti->discards_supported)
1663                        return 1;
1664
1665                if (ti->type->iterate_devices &&
1666                    ti->type->iterate_devices(ti, device_discard_capable, NULL))
1667                        return 1;
1668        }
1669
1670        return 0;
1671}
1672
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.