sys/contrib/openzfs/module/zfs/bpobj.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or https://opensource.org/licenses/CDDL-1.0.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*
  22  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  23  * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
  24  * Copyright (c) 2017 Datto Inc.
  25  */
  26
  27 #include <sys/bpobj.h>
  28 #include <sys/zfs_context.h>
  29 #include <sys/zfs_refcount.h>
  30 #include <sys/dsl_pool.h>
  31 #include <sys/zfeature.h>
  32 #include <sys/zap.h>
  33
  34 /*
  35  * Return an empty bpobj, preferably the empty dummy one (dp_empty_bpobj).
  36  */
  37 uint64_t
  38 bpobj_alloc_empty(objset_t *os, int blocksize, dmu_tx_t *tx)
  39 {
  40         spa_t *spa = dmu_objset_spa(os);
  41         dsl_pool_t *dp = dmu_objset_pool(os);
  42
  43         if (spa_feature_is_enabled(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
  44                 if (!spa_feature_is_active(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
  45                         ASSERT0(dp->dp_empty_bpobj);
  46                         dp->dp_empty_bpobj =
  47                             bpobj_alloc(os, SPA_OLD_MAXBLOCKSIZE, tx);
  48                         VERIFY(zap_add(os,
  49                             DMU_POOL_DIRECTORY_OBJECT,
  50                             DMU_POOL_EMPTY_BPOBJ, sizeof (uint64_t), 1,
  51                             &dp->dp_empty_bpobj, tx) == 0);
  52                 }
  53                 spa_feature_incr(spa, SPA_FEATURE_EMPTY_BPOBJ, tx);
  54                 ASSERT(dp->dp_empty_bpobj != 0);
  55                 return (dp->dp_empty_bpobj);
  56         } else {
  57                 return (bpobj_alloc(os, blocksize, tx));
  58         }
  59 }
  60
  61 void
  62 bpobj_decr_empty(objset_t *os, dmu_tx_t *tx)
  63 {
  64         dsl_pool_t *dp = dmu_objset_pool(os);
  65
  66         spa_feature_decr(dmu_objset_spa(os), SPA_FEATURE_EMPTY_BPOBJ, tx);
  67         if (!spa_feature_is_active(dmu_objset_spa(os),
  68             SPA_FEATURE_EMPTY_BPOBJ)) {
  69                 VERIFY3U(0, ==, zap_remove(dp->dp_meta_objset,
  70                     DMU_POOL_DIRECTORY_OBJECT,
  71                     DMU_POOL_EMPTY_BPOBJ, tx));
  72                 VERIFY3U(0, ==, dmu_object_free(os, dp->dp_empty_bpobj, tx));
  73                 dp->dp_empty_bpobj = 0;
  74         }
  75 }
  76
  77 uint64_t
  78 bpobj_alloc(objset_t *os, int blocksize, dmu_tx_t *tx)
  79 {
  80         int size;
  81
  82         if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_BPOBJ_ACCOUNT)
  83                 size = BPOBJ_SIZE_V0;
  84         else if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_DEADLISTS)
  85                 size = BPOBJ_SIZE_V1;
  86         else if (!spa_feature_is_active(dmu_objset_spa(os),
  87             SPA_FEATURE_LIVELIST))
  88                 size = BPOBJ_SIZE_V2;
  89         else
  90                 size = sizeof (bpobj_phys_t);
  91
  92         return (dmu_object_alloc(os, DMU_OT_BPOBJ, blocksize,
  93             DMU_OT_BPOBJ_HDR, size, tx));
  94 }
  95
  96 void
  97 bpobj_free(objset_t *os, uint64_t obj, dmu_tx_t *tx)
  98 {
  99         int64_t i;
 100         bpobj_t bpo;
 101         dmu_object_info_t doi;
 102         int epb;
 103         dmu_buf_t *dbuf = NULL;
 104
 105         ASSERT(obj != dmu_objset_pool(os)->dp_empty_bpobj);
 106         VERIFY3U(0, ==, bpobj_open(&bpo, os, obj));
 107
 108         mutex_enter(&bpo.bpo_lock);
 109
 110         if (!bpo.bpo_havesubobj || bpo.bpo_phys->bpo_subobjs == 0)
 111                 goto out;
 112
 113         VERIFY3U(0, ==, dmu_object_info(os, bpo.bpo_phys->bpo_subobjs, &doi));
 114         epb = doi.doi_data_block_size / sizeof (uint64_t);
 115
 116         for (i = bpo.bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
 117                 uint64_t *objarray;
 118                 uint64_t offset, blkoff;
 119
 120                 offset = i * sizeof (uint64_t);
 121                 blkoff = P2PHASE(i, epb);
 122
 123                 if (dbuf == NULL || dbuf->db_offset > offset) {
 124                         if (dbuf)
 125                                 dmu_buf_rele(dbuf, FTAG);
 126                         VERIFY3U(0, ==, dmu_buf_hold(os,
 127                             bpo.bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0));
 128                 }
 129
 130                 ASSERT3U(offset, >=, dbuf->db_offset);
 131                 ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
 132
 133                 objarray = dbuf->db_data;
 134                 bpobj_free(os, objarray[blkoff], tx);
 135         }
 136         if (dbuf) {
 137                 dmu_buf_rele(dbuf, FTAG);
 138                 dbuf = NULL;
 139         }
 140         VERIFY3U(0, ==, dmu_object_free(os, bpo.bpo_phys->bpo_subobjs, tx));
 141
 142 out:
 143         mutex_exit(&bpo.bpo_lock);
 144         bpobj_close(&bpo);
 145
 146         VERIFY3U(0, ==, dmu_object_free(os, obj, tx));
 147 }
 148
 149 int
 150 bpobj_open(bpobj_t *bpo, objset_t *os, uint64_t object)
 151 {
 152         dmu_object_info_t doi;
 153         int err;
 154
 155         err = dmu_object_info(os, object, &doi);
 156         if (err)
 157                 return (err);
 158
 159         memset(bpo, 0, sizeof (*bpo));
 160         mutex_init(&bpo->bpo_lock, NULL, MUTEX_DEFAULT, NULL);
 161
 162         ASSERT(bpo->bpo_dbuf == NULL);
 163         ASSERT(bpo->bpo_phys == NULL);
 164         ASSERT(object != 0);
 165         ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ);
 166         ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_BPOBJ_HDR);
 167
 168         err = dmu_bonus_hold(os, object, bpo, &bpo->bpo_dbuf);
 169         if (err)
 170                 return (err);
 171
 172         bpo->bpo_os = os;
 173         bpo->bpo_object = object;
 174         bpo->bpo_epb = doi.doi_data_block_size >> SPA_BLKPTRSHIFT;
 175         bpo->bpo_havecomp = (doi.doi_bonus_size > BPOBJ_SIZE_V0);
 176         bpo->bpo_havesubobj = (doi.doi_bonus_size > BPOBJ_SIZE_V1);
 177         bpo->bpo_havefreed = (doi.doi_bonus_size > BPOBJ_SIZE_V2);
 178         bpo->bpo_phys = bpo->bpo_dbuf->db_data;
 179         return (0);
 180 }
 181
 182 boolean_t
 183 bpobj_is_open(const bpobj_t *bpo)
 184 {
 185         return (bpo->bpo_object != 0);
 186 }
 187
 188 void
 189 bpobj_close(bpobj_t *bpo)
 190 {
 191         /* Lame workaround for closing a bpobj that was never opened. */
 192         if (bpo->bpo_object == 0)
 193                 return;
 194
 195         dmu_buf_rele(bpo->bpo_dbuf, bpo);
 196         if (bpo->bpo_cached_dbuf != NULL)
 197                 dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
 198         bpo->bpo_dbuf = NULL;
 199         bpo->bpo_phys = NULL;
 200         bpo->bpo_cached_dbuf = NULL;
 201         bpo->bpo_object = 0;
 202
 203         mutex_destroy(&bpo->bpo_lock);
 204 }
 205
 206 static boolean_t
 207 bpobj_is_empty_impl(bpobj_t *bpo)
 208 {
 209         ASSERT(MUTEX_HELD(&bpo->bpo_lock));
 210         return (bpo->bpo_phys->bpo_num_blkptrs == 0 &&
 211             (!bpo->bpo_havesubobj || bpo->bpo_phys->bpo_num_subobjs == 0));
 212 }
 213
 214 boolean_t
 215 bpobj_is_empty(bpobj_t *bpo)
 216 {
 217         mutex_enter(&bpo->bpo_lock);
 218         boolean_t is_empty = bpobj_is_empty_impl(bpo);
 219         mutex_exit(&bpo->bpo_lock);
 220         return (is_empty);
 221 }
 222
 223 /*
 224  * A recursive iteration of the bpobjs would be nice here but we run the risk
 225  * of overflowing function stack space.  Instead, find each subobj and add it
 226  * to the head of our list so it can be scanned for subjobjs.  Like a
 227  * recursive implementation, the "deepest" subobjs will be freed first.
 228  * When a subobj is found to have no additional subojs, free it.
 229  */
 230 typedef struct bpobj_info {
 231         bpobj_t *bpi_bpo;
 232         /*
 233          * This object is a subobj of bpi_parent,
 234          * at bpi_index in its subobj array.
 235          */
 236         struct bpobj_info *bpi_parent;
 237         uint64_t bpi_index;
 238         /* How many of our subobj's are left to process. */
 239         uint64_t bpi_unprocessed_subobjs;
 240         /* True after having visited this bpo's directly referenced BPs. */
 241         boolean_t bpi_visited;
 242         list_node_t bpi_node;
 243 } bpobj_info_t;
 244
 245 static bpobj_info_t *
 246 bpi_alloc(bpobj_t *bpo, bpobj_info_t *parent, uint64_t index)
 247 {
 248         bpobj_info_t *bpi = kmem_zalloc(sizeof (bpobj_info_t), KM_SLEEP);
 249         bpi->bpi_bpo = bpo;
 250         bpi->bpi_parent = parent;
 251         bpi->bpi_index = index;
 252         if (bpo->bpo_havesubobj && bpo->bpo_phys->bpo_subobjs != 0) {
 253                 bpi->bpi_unprocessed_subobjs = bpo->bpo_phys->bpo_num_subobjs;
 254         }
 255         return (bpi);
 256 }
 257
 258 /*
 259  * Update bpobj and all of its parents with new space accounting.
 260  */
 261 static void
 262 propagate_space_reduction(bpobj_info_t *bpi, int64_t freed,
 263     int64_t comp_freed, int64_t uncomp_freed, dmu_tx_t *tx)
 264 {
 265
 266         for (; bpi != NULL; bpi = bpi->bpi_parent) {
 267                 bpobj_t *p = bpi->bpi_bpo;
 268                 ASSERT(dmu_buf_is_dirty(p->bpo_dbuf, tx));
 269                 p->bpo_phys->bpo_bytes -= freed;
 270                 ASSERT3S(p->bpo_phys->bpo_bytes, >=, 0);
 271                 if (p->bpo_havecomp) {
 272                         p->bpo_phys->bpo_comp -= comp_freed;
 273                         p->bpo_phys->bpo_uncomp -= uncomp_freed;
 274                 }
 275         }
 276 }
 277
 278 static int
 279 bpobj_iterate_blkptrs(bpobj_info_t *bpi, bpobj_itor_t func, void *arg,
 280     int64_t start, dmu_tx_t *tx, boolean_t free)
 281 {
 282         int err = 0;
 283         int64_t freed = 0, comp_freed = 0, uncomp_freed = 0;
 284         dmu_buf_t *dbuf = NULL;
 285         bpobj_t *bpo = bpi->bpi_bpo;
 286
 287         int64_t i = bpo->bpo_phys->bpo_num_blkptrs - 1;
 288         uint64_t pe = P2ALIGN_TYPED(i, bpo->bpo_epb, uint64_t) *
 289             sizeof (blkptr_t);
 290         uint64_t ps = start * sizeof (blkptr_t);
 291         uint64_t pb = MAX((pe > dmu_prefetch_max) ? pe - dmu_prefetch_max : 0,
 292             ps);
 293         if (pe > pb) {
 294                 dmu_prefetch(bpo->bpo_os, bpo->bpo_object, 0, pb, pe - pb,
 295                     ZIO_PRIORITY_ASYNC_READ);
 296         }
 297         for (; i >= start; i--) {
 298                 uint64_t offset = i * sizeof (blkptr_t);
 299                 uint64_t blkoff = P2PHASE(i, bpo->bpo_epb);
 300
 301                 if (dbuf == NULL || dbuf->db_offset > offset) {
 302                         if (dbuf)
 303                                 dmu_buf_rele(dbuf, FTAG);
 304                         err = dmu_buf_hold(bpo->bpo_os, bpo->bpo_object,
 305                             offset, FTAG, &dbuf, DMU_READ_NO_PREFETCH);
 306                         if (err)
 307                                 break;
 308                         pe = pb;
 309                         pb = MAX((dbuf->db_offset > dmu_prefetch_max) ?
 310                             dbuf->db_offset - dmu_prefetch_max : 0, ps);
 311                         if (pe > pb) {
 312                                 dmu_prefetch(bpo->bpo_os, bpo->bpo_object, 0,
 313                                     pb, pe - pb, ZIO_PRIORITY_ASYNC_READ);
 314                         }
 315                 }
 316
 317                 ASSERT3U(offset, >=, dbuf->db_offset);
 318                 ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
 319
 320                 blkptr_t *bparray = dbuf->db_data;
 321                 blkptr_t *bp = &bparray[blkoff];
 322
 323                 boolean_t bp_freed = BP_GET_FREE(bp);
 324                 err = func(arg, bp, bp_freed, tx);
 325                 if (err)
 326                         break;
 327
 328                 if (free) {
 329                         int sign = bp_freed ? -1 : +1;
 330                         spa_t *spa = dmu_objset_spa(bpo->bpo_os);
 331                         freed += sign * bp_get_dsize_sync(spa, bp);
 332                         comp_freed += sign * BP_GET_PSIZE(bp);
 333                         uncomp_freed += sign * BP_GET_UCSIZE(bp);
 334                         ASSERT(dmu_buf_is_dirty(bpo->bpo_dbuf, tx));
 335                         bpo->bpo_phys->bpo_num_blkptrs--;
 336                         ASSERT3S(bpo->bpo_phys->bpo_num_blkptrs, >=, 0);
 337                         if (bp_freed) {
 338                                 ASSERT(bpo->bpo_havefreed);
 339                                 bpo->bpo_phys->bpo_num_freed--;
 340                                 ASSERT3S(bpo->bpo_phys->bpo_num_freed, >=, 0);
 341                         }
 342                 }
 343         }
 344         if (free) {
 345                 propagate_space_reduction(bpi, freed, comp_freed,
 346                     uncomp_freed, tx);
 347                 VERIFY0(dmu_free_range(bpo->bpo_os,
 348                     bpo->bpo_object,
 349                     bpo->bpo_phys->bpo_num_blkptrs * sizeof (blkptr_t),
 350                     DMU_OBJECT_END, tx));
 351         }
 352         if (dbuf) {
 353                 dmu_buf_rele(dbuf, FTAG);
 354                 dbuf = NULL;
 355         }
 356         return (err);
 357 }
 358
 359 /*
 360  * Given an initial bpo, start by freeing the BPs that are directly referenced
 361  * by that bpo. If the bpo has subobjs, read in its last subobj and push the
 362  * subobj to our stack. By popping items off our stack, eventually we will
 363  * encounter a bpo that has no subobjs.  We can free its bpobj_info_t, and if
 364  * requested also free the now-empty bpo from disk and decrement
 365  * its parent's subobj count. We continue popping each subobj from our stack,
 366  * visiting its last subobj until they too have no more subobjs, and so on.
 367  */
 368 static int
 369 bpobj_iterate_impl(bpobj_t *initial_bpo, bpobj_itor_t func, void *arg,
 370     dmu_tx_t *tx, boolean_t free, uint64_t *bpobj_size)
 371 {
 372         list_t stack;
 373         bpobj_info_t *bpi;
 374         int err = 0;
 375
 376         /*
 377          * Create a "stack" for us to work with without worrying about
 378          * stack overflows. Initialize it with the initial_bpo.
 379          */
 380         list_create(&stack, sizeof (bpobj_info_t),
 381             offsetof(bpobj_info_t, bpi_node));
 382         mutex_enter(&initial_bpo->bpo_lock);
 383
 384         if (bpobj_size != NULL)
 385                 *bpobj_size = initial_bpo->bpo_phys->bpo_num_blkptrs;
 386
 387         list_insert_head(&stack, bpi_alloc(initial_bpo, NULL, 0));
 388
 389         while ((bpi = list_head(&stack)) != NULL) {
 390                 bpobj_t *bpo = bpi->bpi_bpo;
 391
 392                 ASSERT3P(bpo, !=, NULL);
 393                 ASSERT(MUTEX_HELD(&bpo->bpo_lock));
 394                 ASSERT(bpobj_is_open(bpo));
 395
 396                 if (free)
 397                         dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
 398
 399                 if (bpi->bpi_visited == B_FALSE) {
 400                         err = bpobj_iterate_blkptrs(bpi, func, arg, 0, tx,
 401                             free);
 402                         bpi->bpi_visited = B_TRUE;
 403                         if (err != 0)
 404                                 break;
 405                 }
 406                 /*
 407                  * We've finished with this bpo's directly-referenced BP's and
 408                  * it has no more unprocessed subobjs. We can free its
 409                  * bpobj_info_t (unless it is the topmost, initial_bpo).
 410                  * If we are freeing from disk, we can also do that.
 411                  */
 412                 if (bpi->bpi_unprocessed_subobjs == 0) {
 413                         /*
 414                          * If there are no entries, there should
 415                          * be no bytes.
 416                          */
 417                         if (bpobj_is_empty_impl(bpo)) {
 418                                 ASSERT0(bpo->bpo_phys->bpo_bytes);
 419                                 ASSERT0(bpo->bpo_phys->bpo_comp);
 420                                 ASSERT0(bpo->bpo_phys->bpo_uncomp);
 421                         }
 422
 423                         /* The initial_bpo has no parent and is not closed. */
 424                         if (bpi->bpi_parent != NULL) {
 425                                 if (free) {
 426                                         bpobj_t *p = bpi->bpi_parent->bpi_bpo;
 427
 428                                         ASSERT0(bpo->bpo_phys->bpo_num_blkptrs);
 429                                         ASSERT3U(p->bpo_phys->bpo_num_subobjs,
 430                                             >, 0);
 431                                         ASSERT3U(bpi->bpi_index, ==,
 432                                             p->bpo_phys->bpo_num_subobjs - 1);
 433                                         ASSERT(dmu_buf_is_dirty(bpo->bpo_dbuf,
 434                                             tx));
 435
 436                                         p->bpo_phys->bpo_num_subobjs--;
 437
 438                                         VERIFY0(dmu_free_range(p->bpo_os,
 439                                             p->bpo_phys->bpo_subobjs,
 440                                             bpi->bpi_index * sizeof (uint64_t),
 441                                             sizeof (uint64_t), tx));
 442
 443                                         /* eliminate the empty subobj list */
 444                                         if (bpo->bpo_havesubobj &&
 445                                             bpo->bpo_phys->bpo_subobjs != 0) {
 446                                                 ASSERT0(bpo->bpo_phys->
 447                                                     bpo_num_subobjs);
 448                                                 err = dmu_object_free(
 449                                                     bpo->bpo_os,
 450                                                     bpo->bpo_phys->bpo_subobjs,
 451                                                     tx);
 452                                                 if (err)
 453                                                         break;
 454                                                 bpo->bpo_phys->bpo_subobjs = 0;
 455                                         }
 456                                         err = dmu_object_free(p->bpo_os,
 457                                             bpo->bpo_object, tx);
 458                                         if (err)
 459                                                 break;
 460                                 }
 461
 462                                 mutex_exit(&bpo->bpo_lock);
 463                                 bpobj_close(bpo);
 464                                 kmem_free(bpo, sizeof (bpobj_t));
 465                         } else {
 466                                 mutex_exit(&bpo->bpo_lock);
 467                         }
 468
 469                         /*
 470                          * Finished processing this bpo. Unlock, and free
 471                          * our "stack" info.
 472                          */
 473                         list_remove_head(&stack);
 474                         kmem_free(bpi, sizeof (bpobj_info_t));
 475                 } else {
 476                         /*
 477                          * We have unprocessed subobjs. Process the next one.
 478                          */
 479                         ASSERT(bpo->bpo_havecomp);
 480                         ASSERT3P(bpobj_size, ==, NULL);
 481
 482                         /* Add the last subobj to stack. */
 483                         int64_t i = bpi->bpi_unprocessed_subobjs - 1;
 484                         uint64_t offset = i * sizeof (uint64_t);
 485
 486                         uint64_t subobj;
 487                         err = dmu_read(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
 488                             offset, sizeof (uint64_t), &subobj,
 489                             DMU_READ_NO_PREFETCH);
 490                         if (err)
 491                                 break;
 492
 493                         bpobj_t *subbpo = kmem_alloc(sizeof (bpobj_t),
 494                             KM_SLEEP);
 495                         err = bpobj_open(subbpo, bpo->bpo_os, subobj);
 496                         if (err) {
 497                                 kmem_free(subbpo, sizeof (bpobj_t));
 498                                 break;
 499                         }
 500
 501                         if (subbpo->bpo_havesubobj &&
 502                             subbpo->bpo_phys->bpo_subobjs != 0) {
 503                                 dmu_prefetch(subbpo->bpo_os,
 504                                     subbpo->bpo_phys->bpo_subobjs, 0, 0, 0,
 505                                     ZIO_PRIORITY_ASYNC_READ);
 506                         }
 507
 508                         list_insert_head(&stack, bpi_alloc(subbpo, bpi, i));
 509                         mutex_enter(&subbpo->bpo_lock);
 510                         bpi->bpi_unprocessed_subobjs--;
 511                 }
 512         }
 513         /*
 514          * Cleanup anything left on the "stack" after we left the loop.
 515          * Every bpo on the stack is locked so we must remember to undo
 516          * that now (in LIFO order).
 517          */
 518         while ((bpi = list_remove_head(&stack)) != NULL) {
 519                 bpobj_t *bpo = bpi->bpi_bpo;
 520                 ASSERT(err != 0);
 521                 ASSERT3P(bpo, !=, NULL);
 522
 523                 mutex_exit(&bpo->bpo_lock);
 524
 525                 /* do not free the initial_bpo */
 526                 if (bpi->bpi_parent != NULL) {
 527                         bpobj_close(bpi->bpi_bpo);
 528                         kmem_free(bpi->bpi_bpo, sizeof (bpobj_t));
 529                 }
 530                 kmem_free(bpi, sizeof (bpobj_info_t));
 531         }
 532
 533         list_destroy(&stack);
 534
 535         return (err);
 536 }
 537
 538 /*
 539  * Iterate and remove the entries.  If func returns nonzero, iteration
 540  * will stop and that entry will not be removed.
 541  */
 542 int
 543 bpobj_iterate(bpobj_t *bpo, bpobj_itor_t func, void *arg, dmu_tx_t *tx)
 544 {
 545         return (bpobj_iterate_impl(bpo, func, arg, tx, B_TRUE, NULL));
 546 }
 547
 548 /*
 549  * Iterate the entries.  If func returns nonzero, iteration will stop.
 550  *
 551  * If there are no subobjs:
 552  *
 553  * *bpobj_size can be used to return the number of block pointers in the
 554  * bpobj.  Note that this may be different from the number of block pointers
 555  * that are iterated over, if iteration is terminated early (e.g. by the func
 556  * returning nonzero).
 557  *
 558  * If there are concurrent (or subsequent) modifications to the bpobj then the
 559  * returned *bpobj_size can be passed as "start" to
 560  * livelist_bpobj_iterate_from_nofree() to iterate the newly added entries.
 561  */
 562 int
 563 bpobj_iterate_nofree(bpobj_t *bpo, bpobj_itor_t func, void *arg,
 564     uint64_t *bpobj_size)
 565 {
 566         return (bpobj_iterate_impl(bpo, func, arg, NULL, B_FALSE, bpobj_size));
 567 }
 568
 569 /*
 570  * Iterate over the blkptrs in the bpobj beginning at index start. If func
 571  * returns nonzero, iteration will stop. This is a livelist specific function
 572  * since it assumes that there are no subobjs present.
 573  */
 574 int
 575 livelist_bpobj_iterate_from_nofree(bpobj_t *bpo, bpobj_itor_t func, void *arg,
 576     int64_t start)
 577 {
 578         if (bpo->bpo_havesubobj)
 579                 VERIFY0(bpo->bpo_phys->bpo_subobjs);
 580         bpobj_info_t *bpi = bpi_alloc(bpo, NULL, 0);
 581         int err = bpobj_iterate_blkptrs(bpi, func, arg, start, NULL, B_FALSE);
 582         kmem_free(bpi, sizeof (bpobj_info_t));
 583         return (err);
 584 }
 585
 586 /*
 587  * Logically add subobj's contents to the parent bpobj.
 588  *
 589  * In the most general case, this is accomplished in constant time by adding
 590  * a reference to subobj.  This case is used when enqueuing a large subobj:
 591  * +--------------+                        +--------------+
 592  * | bpobj        |----------------------->| subobj list  |
 593  * +----+----+----+----+----+              +-----+-----+--+--+
 594  * | bp | bp | bp | bp | bp |              | obj | obj | obj |
 595  * +----+----+----+----+----+              +-----+-----+-----+
 596  *
 597  * +--------------+                        +--------------+
 598  * | sub-bpobj    |----------------------> | subsubobj    |
 599  * +----+----+----+----+---------+----+    +-----+-----+--+--------+-----+
 600  * | bp | bp | bp | bp |   ...   | bp |    | obj | obj |    ...    | obj |
 601  * +----+----+----+----+---------+----+    +-----+-----+-----------+-----+
 602  *
 603  * Result: sub-bpobj added to parent's subobj list.
 604  * +--------------+                        +--------------+
 605  * | bpobj        |----------------------->| subobj list  |
 606  * +----+----+----+----+----+              +-----+-----+--+--+-----+
 607  * | bp | bp | bp | bp | bp |              | obj | obj | obj | OBJ |
 608  * +----+----+----+----+----+              +-----+-----+-----+--|--+
 609  *                                                              |
 610  *       /-----------------------------------------------------/
 611  *       v
 612  * +--------------+                        +--------------+
 613  * | sub-bpobj    |----------------------> | subsubobj    |
 614  * +----+----+----+----+---------+----+    +-----+-----+--+--------+-----+
 615  * | bp | bp | bp | bp |   ...   | bp |    | obj | obj |    ...    | obj |
 616  * +----+----+----+----+---------+----+    +-----+-----+-----------+-----+
 617  *
 618  *
 619  * In a common case, the subobj is small: its bp's and its list of subobj's
 620  * are each stored in a single block.  In this case we copy the subobj's
 621  * contents to the parent:
 622  * +--------------+                        +--------------+
 623  * | bpobj        |----------------------->| subobj list  |
 624  * +----+----+----+----+----+              +-----+-----+--+--+
 625  * | bp | bp | bp | bp | bp |              | obj | obj | obj |
 626  * +----+----+----+----+----+              +-----+-----+-----+
 627  *                          ^                                ^
 628  * +--------------+         |              +--------------+  |
 629  * | sub-bpobj    |---------^------------> | subsubobj    |  ^
 630  * +----+----+----+         |              +-----+-----+--+  |
 631  * | BP | BP |-->-->-->-->-/               | OBJ | OBJ |-->-/
 632  * +----+----+                             +-----+-----+
 633  *
 634  * Result: subobj destroyed, contents copied to parent:
 635  * +--------------+                        +--------------+
 636  * | bpobj        |----------------------->| subobj list  |
 637  * +----+----+----+----+----+----+----+    +-----+-----+--+--+-----+-----+
 638  * | bp | bp | bp | bp | bp | BP | BP |    | obj | obj | obj | OBJ | OBJ |
 639  * +----+----+----+----+----+----+----+    +-----+-----+-----+-----+-----+
 640  *
 641  *
 642  * If the subobj has many BP's but few subobj's, we can copy the sub-subobj's
 643  * but retain the sub-bpobj:
 644  * +--------------+                        +--------------+
 645  * | bpobj        |----------------------->| subobj list  |
 646  * +----+----+----+----+----+              +-----+-----+--+--+
 647  * | bp | bp | bp | bp | bp |              | obj | obj | obj |
 648  * +----+----+----+----+----+              +-----+-----+-----+
 649  *                                                           ^
 650  * +--------------+                        +--------------+  |
 651  * | sub-bpobj    |----------------------> | subsubobj    |  ^
 652  * +----+----+----+----+---------+----+    +-----+-----+--+  |
 653  * | bp | bp | bp | bp |   ...   | bp |    | OBJ | OBJ |-->-/
 654  * +----+----+----+----+---------+----+    +-----+-----+
 655  *
 656  * Result: sub-sub-bpobjs and subobj added to parent's subobj list.
 657  * +--------------+                     +--------------+
 658  * | bpobj        |-------------------->| subobj list  |
 659  * +----+----+----+----+----+           +-----+-----+--+--+-----+-----+------+
 660  * | bp | bp | bp | bp | bp |           | obj | obj | obj | OBJ | OBJ | OBJ* |
 661  * +----+----+----+----+----+           +-----+-----+-----+-----+-----+--|---+
 662  *                                                                       |
 663  *       /--------------------------------------------------------------/
 664  *       v
 665  * +--------------+
 666  * | sub-bpobj    |
 667  * +----+----+----+----+---------+----+
 668  * | bp | bp | bp | bp |   ...   | bp |
 669  * +----+----+----+----+---------+----+
 670  */
 671 void
 672 bpobj_enqueue_subobj(bpobj_t *bpo, uint64_t subobj, dmu_tx_t *tx)
 673 {
 674         bpobj_t subbpo;
 675         uint64_t used, comp, uncomp, subsubobjs;
 676         boolean_t copy_subsub = B_TRUE;
 677         boolean_t copy_bps = B_TRUE;
 678
 679         ASSERT(bpobj_is_open(bpo));
 680         ASSERT(subobj != 0);
 681         ASSERT(bpo->bpo_havesubobj);
 682         ASSERT(bpo->bpo_havecomp);
 683         ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
 684
 685         if (subobj == dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj) {
 686                 bpobj_decr_empty(bpo->bpo_os, tx);
 687                 return;
 688         }
 689
 690         VERIFY3U(0, ==, bpobj_open(&subbpo, bpo->bpo_os, subobj));
 691         if (bpobj_is_empty(&subbpo)) {
 692                 /* No point in having an empty subobj. */
 693                 bpobj_close(&subbpo);
 694                 bpobj_free(bpo->bpo_os, subobj, tx);
 695                 return;
 696         }
 697         VERIFY3U(0, ==, bpobj_space(&subbpo, &used, &comp, &uncomp));
 698
 699         mutex_enter(&bpo->bpo_lock);
 700         dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
 701
 702         dmu_object_info_t doi;
 703
 704         if (bpo->bpo_phys->bpo_subobjs != 0) {
 705                 ASSERT0(dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
 706                     &doi));
 707                 ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);
 708         }
 709
 710         /*
 711          * If subobj has only one block of subobjs, then move subobj's
 712          * subobjs to bpo's subobj list directly.  This reduces recursion in
 713          * bpobj_iterate due to nested subobjs.
 714          */
 715         subsubobjs = subbpo.bpo_phys->bpo_subobjs;
 716         if (subsubobjs != 0) {
 717                 VERIFY0(dmu_object_info(bpo->bpo_os, subsubobjs, &doi));
 718                 if (doi.doi_max_offset > doi.doi_data_block_size) {
 719                         copy_subsub = B_FALSE;
 720                 }
 721         }
 722
 723         /*
 724          * If, in addition to having only one block of subobj's, subobj has
 725          * only one block of bp's, then move subobj's bp's to bpo's bp list
 726          * directly. This reduces recursion in bpobj_iterate due to nested
 727          * subobjs.
 728          */
 729         VERIFY3U(0, ==, dmu_object_info(bpo->bpo_os, subobj, &doi));
 730         if (doi.doi_max_offset > doi.doi_data_block_size || !copy_subsub) {
 731                 copy_bps = B_FALSE;
 732         }
 733
 734         if (copy_subsub && subsubobjs != 0) {
 735                 dmu_buf_t *subdb;
 736                 uint64_t numsubsub = subbpo.bpo_phys->bpo_num_subobjs;
 737
 738                 VERIFY0(dmu_buf_hold(bpo->bpo_os, subsubobjs,
 739                     0, FTAG, &subdb, 0));
 740                 /*
 741                  * Make sure that we are not asking dmu_write()
 742                  * to write more data than we have in our buffer.
 743                  */
 744                 VERIFY3U(subdb->db_size, >=,
 745                     numsubsub * sizeof (subobj));
 746                 if (bpo->bpo_phys->bpo_subobjs == 0) {
 747                         bpo->bpo_phys->bpo_subobjs =
 748                             dmu_object_alloc(bpo->bpo_os,
 749                             DMU_OT_BPOBJ_SUBOBJ, SPA_OLD_MAXBLOCKSIZE,
 750                             DMU_OT_NONE, 0, tx);
 751                 }
 752                 dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
 753                     bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
 754                     numsubsub * sizeof (subobj), subdb->db_data, tx);
 755                 dmu_buf_rele(subdb, FTAG);
 756                 bpo->bpo_phys->bpo_num_subobjs += numsubsub;
 757
 758                 dmu_buf_will_dirty(subbpo.bpo_dbuf, tx);
 759                 subbpo.bpo_phys->bpo_subobjs = 0;
 760                 VERIFY0(dmu_object_free(bpo->bpo_os, subsubobjs, tx));
 761         }
 762
 763         if (copy_bps) {
 764                 dmu_buf_t *bps;
 765                 uint64_t numbps = subbpo.bpo_phys->bpo_num_blkptrs;
 766
 767                 ASSERT(copy_subsub);
 768                 VERIFY0(dmu_buf_hold(bpo->bpo_os, subobj,
 769                     0, FTAG, &bps, 0));
 770
 771                 /*
 772                  * Make sure that we are not asking dmu_write()
 773                  * to write more data than we have in our buffer.
 774                  */
 775                 VERIFY3U(bps->db_size, >=, numbps * sizeof (blkptr_t));
 776                 dmu_write(bpo->bpo_os, bpo->bpo_object,
 777                     bpo->bpo_phys->bpo_num_blkptrs * sizeof (blkptr_t),
 778                     numbps * sizeof (blkptr_t),
 779                     bps->db_data, tx);
 780                 dmu_buf_rele(bps, FTAG);
 781                 bpo->bpo_phys->bpo_num_blkptrs += numbps;
 782
 783                 bpobj_close(&subbpo);
 784                 VERIFY0(dmu_object_free(bpo->bpo_os, subobj, tx));
 785         } else {
 786                 bpobj_close(&subbpo);
 787                 if (bpo->bpo_phys->bpo_subobjs == 0) {
 788                         bpo->bpo_phys->bpo_subobjs =
 789                             dmu_object_alloc(bpo->bpo_os,
 790                             DMU_OT_BPOBJ_SUBOBJ, SPA_OLD_MAXBLOCKSIZE,
 791                             DMU_OT_NONE, 0, tx);
 792                 }
 793
 794                 dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
 795                     bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
 796                     sizeof (subobj), &subobj, tx);
 797                 bpo->bpo_phys->bpo_num_subobjs++;
 798         }
 799
 800         bpo->bpo_phys->bpo_bytes += used;
 801         bpo->bpo_phys->bpo_comp += comp;
 802         bpo->bpo_phys->bpo_uncomp += uncomp;
 803         mutex_exit(&bpo->bpo_lock);
 804
 805 }
 806
 807 /*
 808  * Prefetch metadata required for bpobj_enqueue_subobj().
 809  */
 810 void
 811 bpobj_prefetch_subobj(bpobj_t *bpo, uint64_t subobj)
 812 {
 813         dmu_object_info_t doi;
 814         bpobj_t subbpo;
 815         uint64_t subsubobjs;
 816         boolean_t copy_subsub = B_TRUE;
 817         boolean_t copy_bps = B_TRUE;
 818
 819         ASSERT(bpobj_is_open(bpo));
 820         ASSERT(subobj != 0);
 821
 822         if (subobj == dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj)
 823                 return;
 824
 825         if (bpobj_open(&subbpo, bpo->bpo_os, subobj) != 0)
 826                 return;
 827         if (bpobj_is_empty(&subbpo)) {
 828                 bpobj_close(&subbpo);
 829                 return;
 830         }
 831         subsubobjs = subbpo.bpo_phys->bpo_subobjs;
 832         bpobj_close(&subbpo);
 833
 834         if (subsubobjs != 0) {
 835                 if (dmu_object_info(bpo->bpo_os, subsubobjs, &doi) != 0)
 836                         return;
 837                 if (doi.doi_max_offset > doi.doi_data_block_size)
 838                         copy_subsub = B_FALSE;
 839         }
 840
 841         if (dmu_object_info(bpo->bpo_os, subobj, &doi) != 0)
 842                 return;
 843         if (doi.doi_max_offset > doi.doi_data_block_size || !copy_subsub)
 844                 copy_bps = B_FALSE;
 845
 846         if (copy_subsub && subsubobjs != 0) {
 847                 if (bpo->bpo_phys->bpo_subobjs) {
 848                         dmu_prefetch(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, 0,
 849                             bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj), 1,
 850                             ZIO_PRIORITY_ASYNC_READ);
 851                 }
 852                 dmu_prefetch(bpo->bpo_os, subsubobjs, 0, 0, 1,
 853                     ZIO_PRIORITY_ASYNC_READ);
 854         }
 855
 856         if (copy_bps) {
 857                 dmu_prefetch(bpo->bpo_os, bpo->bpo_object, 0,
 858                     bpo->bpo_phys->bpo_num_blkptrs * sizeof (blkptr_t), 1,
 859                     ZIO_PRIORITY_ASYNC_READ);
 860                 dmu_prefetch(bpo->bpo_os, subobj, 0, 0, 1,
 861                     ZIO_PRIORITY_ASYNC_READ);
 862         } else if (bpo->bpo_phys->bpo_subobjs) {
 863                 dmu_prefetch(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, 0,
 864                     bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj), 1,
 865                     ZIO_PRIORITY_ASYNC_READ);
 866         }
 867 }
 868
 869 void
 870 bpobj_enqueue(bpobj_t *bpo, const blkptr_t *bp, boolean_t bp_freed,
 871     dmu_tx_t *tx)
 872 {
 873         blkptr_t stored_bp = *bp;
 874         uint64_t offset;
 875         int blkoff;
 876         blkptr_t *bparray;
 877
 878         ASSERT(bpobj_is_open(bpo));
 879         ASSERT(!BP_IS_HOLE(bp));
 880         ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
 881
 882         if (BP_IS_EMBEDDED(bp)) {
 883                 /*
 884                  * The bpobj will compress better without the payload.
 885                  *
 886                  * Note that we store EMBEDDED bp's because they have an
 887                  * uncompressed size, which must be accounted for.  An
 888                  * alternative would be to add their size to bpo_uncomp
 889                  * without storing the bp, but that would create additional
 890                  * complications: bpo_uncomp would be inconsistent with the
 891                  * set of BP's stored, and bpobj_iterate() wouldn't visit
 892                  * all the space accounted for in the bpobj.
 893                  */
 894                 memset(&stored_bp, 0, sizeof (stored_bp));
 895                 stored_bp.blk_prop = bp->blk_prop;
 896                 stored_bp.blk_birth = bp->blk_birth;
 897         } else if (!BP_GET_DEDUP(bp)) {
 898                 /* The bpobj will compress better without the checksum */
 899                 memset(&stored_bp.blk_cksum, 0, sizeof (stored_bp.blk_cksum));
 900         }
 901
 902         stored_bp.blk_fill = 0;
 903         BP_SET_FREE(&stored_bp, bp_freed);
 904
 905         mutex_enter(&bpo->bpo_lock);
 906
 907         offset = bpo->bpo_phys->bpo_num_blkptrs * sizeof (stored_bp);
 908         blkoff = P2PHASE(bpo->bpo_phys->bpo_num_blkptrs, bpo->bpo_epb);
 909
 910         if (bpo->bpo_cached_dbuf == NULL ||
 911             offset < bpo->bpo_cached_dbuf->db_offset ||
 912             offset >= bpo->bpo_cached_dbuf->db_offset +
 913             bpo->bpo_cached_dbuf->db_size) {
 914                 if (bpo->bpo_cached_dbuf)
 915                         dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
 916                 VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, bpo->bpo_object,
 917                     offset, bpo, &bpo->bpo_cached_dbuf, 0));
 918                 ASSERT3P(bpo->bpo_cached_dbuf, !=, NULL);
 919         }
 920
 921         dmu_buf_will_dirty(bpo->bpo_cached_dbuf, tx);
 922         bparray = bpo->bpo_cached_dbuf->db_data;
 923         bparray[blkoff] = stored_bp;
 924
 925         dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
 926         bpo->bpo_phys->bpo_num_blkptrs++;
 927         int sign = bp_freed ? -1 : +1;
 928         bpo->bpo_phys->bpo_bytes += sign *
 929             bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
 930         if (bpo->bpo_havecomp) {
 931                 bpo->bpo_phys->bpo_comp += sign * BP_GET_PSIZE(bp);
 932                 bpo->bpo_phys->bpo_uncomp += sign * BP_GET_UCSIZE(bp);
 933         }
 934         if (bp_freed) {
 935                 ASSERT(bpo->bpo_havefreed);
 936                 bpo->bpo_phys->bpo_num_freed++;
 937         }
 938         mutex_exit(&bpo->bpo_lock);
 939 }
 940
 941 struct space_range_arg {
 942         spa_t *spa;
 943         uint64_t mintxg;
 944         uint64_t maxtxg;
 945         uint64_t used;
 946         uint64_t comp;
 947         uint64_t uncomp;
 948 };
 949
 950 static int
 951 space_range_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed, dmu_tx_t *tx)
 952 {
 953         (void) bp_freed, (void) tx;
 954         struct space_range_arg *sra = arg;
 955
 956         if (bp->blk_birth > sra->mintxg && bp->blk_birth <= sra->maxtxg) {
 957                 if (dsl_pool_sync_context(spa_get_dsl(sra->spa)))
 958                         sra->used += bp_get_dsize_sync(sra->spa, bp);
 959                 else
 960                         sra->used += bp_get_dsize(sra->spa, bp);
 961                 sra->comp += BP_GET_PSIZE(bp);
 962                 sra->uncomp += BP_GET_UCSIZE(bp);
 963         }
 964         return (0);
 965 }
 966
 967 int
 968 bpobj_space(bpobj_t *bpo, uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
 969 {
 970         ASSERT(bpobj_is_open(bpo));
 971         mutex_enter(&bpo->bpo_lock);
 972
 973         *usedp = bpo->bpo_phys->bpo_bytes;
 974         if (bpo->bpo_havecomp) {
 975                 *compp = bpo->bpo_phys->bpo_comp;
 976                 *uncompp = bpo->bpo_phys->bpo_uncomp;
 977                 mutex_exit(&bpo->bpo_lock);
 978                 return (0);
 979         } else {
 980                 mutex_exit(&bpo->bpo_lock);
 981                 return (bpobj_space_range(bpo, 0, UINT64_MAX,
 982                     usedp, compp, uncompp));
 983         }
 984 }
 985
 986 /*
 987  * Return the amount of space in the bpobj which is:
 988  * mintxg < blk_birth <= maxtxg
 989  */
 990 int
 991 bpobj_space_range(bpobj_t *bpo, uint64_t mintxg, uint64_t maxtxg,
 992     uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
 993 {
 994         struct space_range_arg sra = { 0 };
 995         int err;
 996
 997         ASSERT(bpobj_is_open(bpo));
 998
 999         /*
1000          * As an optimization, if they want the whole txg range, just
1001          * get bpo_bytes rather than iterating over the bps.
1002          */
1003         if (mintxg < TXG_INITIAL && maxtxg == UINT64_MAX && bpo->bpo_havecomp)
1004                 return (bpobj_space(bpo, usedp, compp, uncompp));
1005
1006         sra.spa = dmu_objset_spa(bpo->bpo_os);
1007         sra.mintxg = mintxg;
1008         sra.maxtxg = maxtxg;
1009
1010         err = bpobj_iterate_nofree(bpo, space_range_cb, &sra, NULL);
1011         *usedp = sra.used;
1012         *compp = sra.comp;
1013         *uncompp = sra.uncomp;
1014         return (err);
1015 }
1016
1017 /*
1018  * A bpobj_itor_t to append blkptrs to a bplist. Note that while blkptrs in a
1019  * bpobj are designated as free or allocated that information is not preserved
1020  * in bplists.
1021  */
1022 int
1023 bplist_append_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed,
1024     dmu_tx_t *tx)
1025 {
1026         (void) bp_freed, (void) tx;
1027         bplist_t *bpl = arg;
1028         bplist_append(bpl, bp);
1029         return (0);
1030 }