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 http://www.opensolaris.org/os/licensing.
  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, 2016 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/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
  87                 size = sizeof (bpobj_phys_t);
  88
  89         return (dmu_object_alloc(os, DMU_OT_BPOBJ, blocksize,
  90             DMU_OT_BPOBJ_HDR, size, tx));
  91 }
  92
  93 void
  94 bpobj_free(objset_t *os, uint64_t obj, dmu_tx_t *tx)
  95 {
  96         int64_t i;
  97         bpobj_t bpo;
  98         dmu_object_info_t doi;
  99         int epb;
 100         dmu_buf_t *dbuf = NULL;
 101
 102         ASSERT(obj != dmu_objset_pool(os)->dp_empty_bpobj);
 103         VERIFY3U(0, ==, bpobj_open(&bpo, os, obj));
 104
 105         mutex_enter(&bpo.bpo_lock);
 106
 107         if (!bpo.bpo_havesubobj || bpo.bpo_phys->bpo_subobjs == 0)
 108                 goto out;
 109
 110         VERIFY3U(0, ==, dmu_object_info(os, bpo.bpo_phys->bpo_subobjs, &doi));
 111         epb = doi.doi_data_block_size / sizeof (uint64_t);
 112
 113         for (i = bpo.bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
 114                 uint64_t *objarray;
 115                 uint64_t offset, blkoff;
 116
 117                 offset = i * sizeof (uint64_t);
 118                 blkoff = P2PHASE(i, epb);
 119
 120                 if (dbuf == NULL || dbuf->db_offset > offset) {
 121                         if (dbuf)
 122                                 dmu_buf_rele(dbuf, FTAG);
 123                         VERIFY3U(0, ==, dmu_buf_hold(os,
 124                             bpo.bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0));
 125                 }
 126
 127                 ASSERT3U(offset, >=, dbuf->db_offset);
 128                 ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
 129
 130                 objarray = dbuf->db_data;
 131                 bpobj_free(os, objarray[blkoff], tx);
 132         }
 133         if (dbuf) {
 134                 dmu_buf_rele(dbuf, FTAG);
 135                 dbuf = NULL;
 136         }
 137         VERIFY3U(0, ==, dmu_object_free(os, bpo.bpo_phys->bpo_subobjs, tx));
 138
 139 out:
 140         mutex_exit(&bpo.bpo_lock);
 141         bpobj_close(&bpo);
 142
 143         VERIFY3U(0, ==, dmu_object_free(os, obj, tx));
 144 }
 145
 146 int
 147 bpobj_open(bpobj_t *bpo, objset_t *os, uint64_t object)
 148 {
 149         dmu_object_info_t doi;
 150         int err;
 151
 152         err = dmu_object_info(os, object, &doi);
 153         if (err)
 154                 return (err);
 155
 156         bzero(bpo, sizeof (*bpo));
 157         mutex_init(&bpo->bpo_lock, NULL, MUTEX_DEFAULT, NULL);
 158
 159         ASSERT(bpo->bpo_dbuf == NULL);
 160         ASSERT(bpo->bpo_phys == NULL);
 161         ASSERT(object != 0);
 162         ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ);
 163         ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_BPOBJ_HDR);
 164
 165         err = dmu_bonus_hold(os, object, bpo, &bpo->bpo_dbuf);
 166         if (err)
 167                 return (err);
 168
 169         bpo->bpo_os = os;
 170         bpo->bpo_object = object;
 171         bpo->bpo_epb = doi.doi_data_block_size >> SPA_BLKPTRSHIFT;
 172         bpo->bpo_havecomp = (doi.doi_bonus_size > BPOBJ_SIZE_V0);
 173         bpo->bpo_havesubobj = (doi.doi_bonus_size > BPOBJ_SIZE_V1);
 174         bpo->bpo_phys = bpo->bpo_dbuf->db_data;
 175         return (0);
 176 }
 177
 178 void
 179 bpobj_close(bpobj_t *bpo)
 180 {
 181         /* Lame workaround for closing a bpobj that was never opened. */
 182         if (bpo->bpo_object == 0)
 183                 return;
 184
 185         dmu_buf_rele(bpo->bpo_dbuf, bpo);
 186         if (bpo->bpo_cached_dbuf != NULL)
 187                 dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
 188         bpo->bpo_dbuf = NULL;
 189         bpo->bpo_phys = NULL;
 190         bpo->bpo_cached_dbuf = NULL;
 191         bpo->bpo_object = 0;
 192
 193         mutex_destroy(&bpo->bpo_lock);
 194 }
 195
 196 static boolean_t
 197 bpobj_hasentries(bpobj_t *bpo)
 198 {
 199         return (bpo->bpo_phys->bpo_num_blkptrs != 0 ||
 200             (bpo->bpo_havesubobj && bpo->bpo_phys->bpo_num_subobjs != 0));
 201 }
 202
 203 static int
 204 bpobj_iterate_impl(bpobj_t *bpo, bpobj_itor_t func, void *arg, dmu_tx_t *tx,
 205     boolean_t free)
 206 {
 207         dmu_object_info_t doi;
 208         int epb;
 209         int64_t i;
 210         int err = 0;
 211         dmu_buf_t *dbuf = NULL;
 212
 213         mutex_enter(&bpo->bpo_lock);
 214
 215         if (!bpobj_hasentries(bpo))
 216                 goto out;
 217
 218         if (free)
 219                 dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
 220
 221         for (i = bpo->bpo_phys->bpo_num_blkptrs - 1; i >= 0; i--) {
 222                 blkptr_t *bparray;
 223                 blkptr_t *bp;
 224                 uint64_t offset, blkoff;
 225
 226                 offset = i * sizeof (blkptr_t);
 227                 blkoff = P2PHASE(i, bpo->bpo_epb);
 228
 229                 if (dbuf == NULL || dbuf->db_offset > offset) {
 230                         if (dbuf)
 231                                 dmu_buf_rele(dbuf, FTAG);
 232                         err = dmu_buf_hold(bpo->bpo_os, bpo->bpo_object, offset,
 233                             FTAG, &dbuf, 0);
 234                         if (err)
 235                                 break;
 236                 }
 237
 238                 ASSERT3U(offset, >=, dbuf->db_offset);
 239                 ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
 240
 241                 bparray = dbuf->db_data;
 242                 bp = &bparray[blkoff];
 243                 err = func(arg, bp, tx);
 244                 if (err)
 245                         break;
 246                 if (free) {
 247                         bpo->bpo_phys->bpo_bytes -=
 248                             bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
 249                         ASSERT3S(bpo->bpo_phys->bpo_bytes, >=, 0);
 250                         if (bpo->bpo_havecomp) {
 251                                 bpo->bpo_phys->bpo_comp -= BP_GET_PSIZE(bp);
 252                                 bpo->bpo_phys->bpo_uncomp -= BP_GET_UCSIZE(bp);
 253                         }
 254                         bpo->bpo_phys->bpo_num_blkptrs--;
 255                         ASSERT3S(bpo->bpo_phys->bpo_num_blkptrs, >=, 0);
 256                 }
 257         }
 258         if (dbuf) {
 259                 dmu_buf_rele(dbuf, FTAG);
 260                 dbuf = NULL;
 261         }
 262         if (free) {
 263                 VERIFY3U(0, ==, dmu_free_range(bpo->bpo_os, bpo->bpo_object,
 264                     (i + 1) * sizeof (blkptr_t), -1ULL, tx));
 265         }
 266         if (err || !bpo->bpo_havesubobj || bpo->bpo_phys->bpo_subobjs == 0)
 267                 goto out;
 268
 269         ASSERT(bpo->bpo_havecomp);
 270         err = dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, &doi);
 271         if (err) {
 272                 mutex_exit(&bpo->bpo_lock);
 273                 return (err);
 274         }
 275         ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);
 276         epb = doi.doi_data_block_size / sizeof (uint64_t);
 277
 278         for (i = bpo->bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
 279                 uint64_t *objarray;
 280                 uint64_t offset, blkoff;
 281                 bpobj_t sublist;
 282                 uint64_t used_before, comp_before, uncomp_before;
 283                 uint64_t used_after, comp_after, uncomp_after;
 284
 285                 offset = i * sizeof (uint64_t);
 286                 blkoff = P2PHASE(i, epb);
 287
 288                 if (dbuf == NULL || dbuf->db_offset > offset) {
 289                         if (dbuf)
 290                                 dmu_buf_rele(dbuf, FTAG);
 291                         err = dmu_buf_hold(bpo->bpo_os,
 292                             bpo->bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0);
 293                         if (err)
 294                                 break;
 295                 }
 296
 297                 ASSERT3U(offset, >=, dbuf->db_offset);
 298                 ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
 299
 300                 objarray = dbuf->db_data;
 301                 err = bpobj_open(&sublist, bpo->bpo_os, objarray[blkoff]);
 302                 if (err)
 303                         break;
 304                 if (free) {
 305                         err = bpobj_space(&sublist,
 306                             &used_before, &comp_before, &uncomp_before);
 307                         if (err != 0) {
 308                                 bpobj_close(&sublist);
 309                                 break;
 310                         }
 311                 }
 312                 err = bpobj_iterate_impl(&sublist, func, arg, tx, free);
 313                 if (free) {
 314                         VERIFY3U(0, ==, bpobj_space(&sublist,
 315                             &used_after, &comp_after, &uncomp_after));
 316                         bpo->bpo_phys->bpo_bytes -= used_before - used_after;
 317                         ASSERT3S(bpo->bpo_phys->bpo_bytes, >=, 0);
 318                         bpo->bpo_phys->bpo_comp -= comp_before - comp_after;
 319                         bpo->bpo_phys->bpo_uncomp -=
 320                             uncomp_before - uncomp_after;
 321                 }
 322
 323                 bpobj_close(&sublist);
 324                 if (err)
 325                         break;
 326                 if (free) {
 327                         err = dmu_object_free(bpo->bpo_os,
 328                             objarray[blkoff], tx);
 329                         if (err)
 330                                 break;
 331                         bpo->bpo_phys->bpo_num_subobjs--;
 332                         ASSERT3S(bpo->bpo_phys->bpo_num_subobjs, >=, 0);
 333                 }
 334         }
 335         if (dbuf) {
 336                 dmu_buf_rele(dbuf, FTAG);
 337                 dbuf = NULL;
 338         }
 339         if (free) {
 340                 VERIFY3U(0, ==, dmu_free_range(bpo->bpo_os,
 341                     bpo->bpo_phys->bpo_subobjs,
 342                     (i + 1) * sizeof (uint64_t), -1ULL, tx));
 343         }
 344
 345 out:
 346         /* If there are no entries, there should be no bytes. */
 347         if (!bpobj_hasentries(bpo)) {
 348                 ASSERT0(bpo->bpo_phys->bpo_bytes);
 349                 ASSERT0(bpo->bpo_phys->bpo_comp);
 350                 ASSERT0(bpo->bpo_phys->bpo_uncomp);
 351         }
 352
 353         mutex_exit(&bpo->bpo_lock);
 354         return (err);
 355 }
 356
 357 /*
 358  * Iterate and remove the entries.  If func returns nonzero, iteration
 359  * will stop and that entry will not be removed.
 360  */
 361 int
 362 bpobj_iterate(bpobj_t *bpo, bpobj_itor_t func, void *arg, dmu_tx_t *tx)
 363 {
 364         return (bpobj_iterate_impl(bpo, func, arg, tx, B_TRUE));
 365 }
 366
 367 /*
 368  * Iterate the entries.  If func returns nonzero, iteration will stop.
 369  */
 370 int
 371 bpobj_iterate_nofree(bpobj_t *bpo, bpobj_itor_t func, void *arg, dmu_tx_t *tx)
 372 {
 373         return (bpobj_iterate_impl(bpo, func, arg, tx, B_FALSE));
 374 }
 375
 376 void
 377 bpobj_enqueue_subobj(bpobj_t *bpo, uint64_t subobj, dmu_tx_t *tx)
 378 {
 379         bpobj_t subbpo;
 380         uint64_t used, comp, uncomp, subsubobjs;
 381         ASSERTV(dmu_object_info_t doi);
 382
 383         ASSERT(bpo->bpo_havesubobj);
 384         ASSERT(bpo->bpo_havecomp);
 385         ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
 386
 387         if (subobj == dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj) {
 388                 bpobj_decr_empty(bpo->bpo_os, tx);
 389                 return;
 390         }
 391
 392         VERIFY3U(0, ==, bpobj_open(&subbpo, bpo->bpo_os, subobj));
 393         VERIFY3U(0, ==, bpobj_space(&subbpo, &used, &comp, &uncomp));
 394
 395         if (!bpobj_hasentries(&subbpo)) {
 396                 /* No point in having an empty subobj. */
 397                 bpobj_close(&subbpo);
 398                 bpobj_free(bpo->bpo_os, subobj, tx);
 399                 return;
 400         }
 401
 402         mutex_enter(&bpo->bpo_lock);
 403         dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
 404         if (bpo->bpo_phys->bpo_subobjs == 0) {
 405                 bpo->bpo_phys->bpo_subobjs = dmu_object_alloc(bpo->bpo_os,
 406                     DMU_OT_BPOBJ_SUBOBJ, SPA_OLD_MAXBLOCKSIZE,
 407                     DMU_OT_NONE, 0, tx);
 408         }
 409
 410         ASSERT0(dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, &doi));
 411         ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);
 412
 413         dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
 414             bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
 415             sizeof (subobj), &subobj, tx);
 416         bpo->bpo_phys->bpo_num_subobjs++;
 417
 418         /*
 419          * If subobj has only one block of subobjs, then move subobj's
 420          * subobjs to bpo's subobj list directly.  This reduces
 421          * recursion in bpobj_iterate due to nested subobjs.
 422          */
 423         subsubobjs = subbpo.bpo_phys->bpo_subobjs;
 424         if (subsubobjs != 0) {
 425                 dmu_object_info_t doi;
 426
 427                 VERIFY3U(0, ==, dmu_object_info(bpo->bpo_os, subsubobjs, &doi));
 428                 if (doi.doi_max_offset == doi.doi_data_block_size) {
 429                         dmu_buf_t *subdb;
 430                         uint64_t numsubsub = subbpo.bpo_phys->bpo_num_subobjs;
 431
 432                         VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, subsubobjs,
 433                             0, FTAG, &subdb, 0));
 434                         /*
 435                          * Make sure that we are not asking dmu_write()
 436                          * to write more data than we have in our buffer.
 437                          */
 438                         VERIFY3U(subdb->db_size, >=,
 439                             numsubsub * sizeof (subobj));
 440                         dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
 441                             bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
 442                             numsubsub * sizeof (subobj), subdb->db_data, tx);
 443                         dmu_buf_rele(subdb, FTAG);
 444                         bpo->bpo_phys->bpo_num_subobjs += numsubsub;
 445
 446                         dmu_buf_will_dirty(subbpo.bpo_dbuf, tx);
 447                         subbpo.bpo_phys->bpo_subobjs = 0;
 448                         VERIFY3U(0, ==, dmu_object_free(bpo->bpo_os,
 449                             subsubobjs, tx));
 450                 }
 451         }
 452         bpo->bpo_phys->bpo_bytes += used;
 453         bpo->bpo_phys->bpo_comp += comp;
 454         bpo->bpo_phys->bpo_uncomp += uncomp;
 455         mutex_exit(&bpo->bpo_lock);
 456
 457         bpobj_close(&subbpo);
 458 }
 459
 460 void
 461 bpobj_enqueue(bpobj_t *bpo, const blkptr_t *bp, dmu_tx_t *tx)
 462 {
 463         blkptr_t stored_bp = *bp;
 464         uint64_t offset;
 465         int blkoff;
 466         blkptr_t *bparray;
 467
 468         ASSERT(!BP_IS_HOLE(bp));
 469         ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
 470
 471         if (BP_IS_EMBEDDED(bp)) {
 472                 /*
 473                  * The bpobj will compress better without the payload.
 474                  *
 475                  * Note that we store EMBEDDED bp's because they have an
 476                  * uncompressed size, which must be accounted for.  An
 477                  * alternative would be to add their size to bpo_uncomp
 478                  * without storing the bp, but that would create additional
 479                  * complications: bpo_uncomp would be inconsistent with the
 480                  * set of BP's stored, and bpobj_iterate() wouldn't visit
 481                  * all the space accounted for in the bpobj.
 482                  */
 483                 bzero(&stored_bp, sizeof (stored_bp));
 484                 stored_bp.blk_prop = bp->blk_prop;
 485                 stored_bp.blk_birth = bp->blk_birth;
 486         } else if (!BP_GET_DEDUP(bp)) {
 487                 /* The bpobj will compress better without the checksum */
 488                 bzero(&stored_bp.blk_cksum, sizeof (stored_bp.blk_cksum));
 489         }
 490
 491         /* We never need the fill count. */
 492         stored_bp.blk_fill = 0;
 493
 494         mutex_enter(&bpo->bpo_lock);
 495
 496         offset = bpo->bpo_phys->bpo_num_blkptrs * sizeof (stored_bp);
 497         blkoff = P2PHASE(bpo->bpo_phys->bpo_num_blkptrs, bpo->bpo_epb);
 498
 499         if (bpo->bpo_cached_dbuf == NULL ||
 500             offset < bpo->bpo_cached_dbuf->db_offset ||
 501             offset >= bpo->bpo_cached_dbuf->db_offset +
 502             bpo->bpo_cached_dbuf->db_size) {
 503                 if (bpo->bpo_cached_dbuf)
 504                         dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
 505                 VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, bpo->bpo_object,
 506                     offset, bpo, &bpo->bpo_cached_dbuf, 0));
 507         }
 508
 509         dmu_buf_will_dirty(bpo->bpo_cached_dbuf, tx);
 510         bparray = bpo->bpo_cached_dbuf->db_data;
 511         bparray[blkoff] = stored_bp;
 512
 513         dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
 514         bpo->bpo_phys->bpo_num_blkptrs++;
 515         bpo->bpo_phys->bpo_bytes +=
 516             bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
 517         if (bpo->bpo_havecomp) {
 518                 bpo->bpo_phys->bpo_comp += BP_GET_PSIZE(bp);
 519                 bpo->bpo_phys->bpo_uncomp += BP_GET_UCSIZE(bp);
 520         }
 521         mutex_exit(&bpo->bpo_lock);
 522 }
 523
 524 struct space_range_arg {
 525         spa_t *spa;
 526         uint64_t mintxg;
 527         uint64_t maxtxg;
 528         uint64_t used;
 529         uint64_t comp;
 530         uint64_t uncomp;
 531 };
 532
 533 /* ARGSUSED */
 534 static int
 535 space_range_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
 536 {
 537         struct space_range_arg *sra = arg;
 538
 539         if (bp->blk_birth > sra->mintxg && bp->blk_birth <= sra->maxtxg) {
 540                 if (dsl_pool_sync_context(spa_get_dsl(sra->spa)))
 541                         sra->used += bp_get_dsize_sync(sra->spa, bp);
 542                 else
 543                         sra->used += bp_get_dsize(sra->spa, bp);
 544                 sra->comp += BP_GET_PSIZE(bp);
 545                 sra->uncomp += BP_GET_UCSIZE(bp);
 546         }
 547         return (0);
 548 }
 549
 550 int
 551 bpobj_space(bpobj_t *bpo, uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
 552 {
 553         mutex_enter(&bpo->bpo_lock);
 554
 555         *usedp = bpo->bpo_phys->bpo_bytes;
 556         if (bpo->bpo_havecomp) {
 557                 *compp = bpo->bpo_phys->bpo_comp;
 558                 *uncompp = bpo->bpo_phys->bpo_uncomp;
 559                 mutex_exit(&bpo->bpo_lock);
 560                 return (0);
 561         } else {
 562                 mutex_exit(&bpo->bpo_lock);
 563                 return (bpobj_space_range(bpo, 0, UINT64_MAX,
 564                     usedp, compp, uncompp));
 565         }
 566 }
 567
 568 /*
 569  * Return the amount of space in the bpobj which is:
 570  * mintxg < blk_birth <= maxtxg
 571  */
 572 int
 573 bpobj_space_range(bpobj_t *bpo, uint64_t mintxg, uint64_t maxtxg,
 574     uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
 575 {
 576         struct space_range_arg sra = { 0 };
 577         int err;
 578
 579         /*
 580          * As an optimization, if they want the whole txg range, just
 581          * get bpo_bytes rather than iterating over the bps.
 582          */
 583         if (mintxg < TXG_INITIAL && maxtxg == UINT64_MAX && bpo->bpo_havecomp)
 584                 return (bpobj_space(bpo, usedp, compp, uncompp));
 585
 586         sra.spa = dmu_objset_spa(bpo->bpo_os);
 587         sra.mintxg = mintxg;
 588         sra.maxtxg = maxtxg;
 589
 590         err = bpobj_iterate_nofree(bpo, space_range_cb, &sra, NULL);
 591         *usedp = sra.used;
 592         *compp = sra.comp;
 593         *uncompp = sra.uncomp;
 594         return (err);
 595 }