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) 2014 Integros [integros.com]
25 * Copyright (c) 2017 Datto Inc.
26 */
27
28 #include <sys/bpobj.h>
29 #include <sys/zfs_context.h>
30 #include <sys/refcount.h>
31 #include <sys/dsl_pool.h>
32 #include <sys/zfeature.h>
33 #include <sys/zap.h>
34
35 /*
36 * Return an empty bpobj, preferably the empty dummy one (dp_empty_bpobj).
37 */
38 uint64_t
bpobj_alloc_empty(objset_t * os,int blocksize,dmu_tx_t * tx)39 bpobj_alloc_empty(objset_t *os, int blocksize, dmu_tx_t *tx)
40 {
41 spa_t *spa = dmu_objset_spa(os);
42 dsl_pool_t *dp = dmu_objset_pool(os);
43
44 if (spa_feature_is_enabled(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
45 if (!spa_feature_is_active(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
46 ASSERT0(dp->dp_empty_bpobj);
47 dp->dp_empty_bpobj =
48 bpobj_alloc(os, SPA_OLD_MAXBLOCKSIZE, tx);
49 VERIFY(zap_add(os,
50 DMU_POOL_DIRECTORY_OBJECT,
51 DMU_POOL_EMPTY_BPOBJ, sizeof (uint64_t), 1,
52 &dp->dp_empty_bpobj, tx) == 0);
53 }
54 spa_feature_incr(spa, SPA_FEATURE_EMPTY_BPOBJ, tx);
55 ASSERT(dp->dp_empty_bpobj != 0);
56 return (dp->dp_empty_bpobj);
57 } else {
58 return (bpobj_alloc(os, blocksize, tx));
59 }
60 }
61
62 void
bpobj_decr_empty(objset_t * os,dmu_tx_t * tx)63 bpobj_decr_empty(objset_t *os, dmu_tx_t *tx)
64 {
65 dsl_pool_t *dp = dmu_objset_pool(os);
66
67 spa_feature_decr(dmu_objset_spa(os), SPA_FEATURE_EMPTY_BPOBJ, tx);
68 if (!spa_feature_is_active(dmu_objset_spa(os),
69 SPA_FEATURE_EMPTY_BPOBJ)) {
70 VERIFY3U(0, ==, zap_remove(dp->dp_meta_objset,
71 DMU_POOL_DIRECTORY_OBJECT,
72 DMU_POOL_EMPTY_BPOBJ, tx));
73 VERIFY3U(0, ==, dmu_object_free(os, dp->dp_empty_bpobj, tx));
74 dp->dp_empty_bpobj = 0;
75 }
76 }
77
78 uint64_t
bpobj_alloc(objset_t * os,int blocksize,dmu_tx_t * tx)79 bpobj_alloc(objset_t *os, int blocksize, dmu_tx_t *tx)
80 {
81 int size;
82
83 if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_BPOBJ_ACCOUNT)
84 size = BPOBJ_SIZE_V0;
85 else if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_DEADLISTS)
86 size = BPOBJ_SIZE_V1;
87 else
88 size = sizeof (bpobj_phys_t);
89
90 return (dmu_object_alloc(os, DMU_OT_BPOBJ, blocksize,
91 DMU_OT_BPOBJ_HDR, size, tx));
92 }
93
94 void
bpobj_free(objset_t * os,uint64_t obj,dmu_tx_t * tx)95 bpobj_free(objset_t *os, uint64_t obj, dmu_tx_t *tx)
96 {
97 int64_t i;
98 bpobj_t bpo;
99 dmu_object_info_t doi;
100 int epb;
101 dmu_buf_t *dbuf = NULL;
102
103 ASSERT(obj != dmu_objset_pool(os)->dp_empty_bpobj);
104 VERIFY3U(0, ==, bpobj_open(&bpo, os, obj));
105
106 mutex_enter(&bpo.bpo_lock);
107
108 if (!bpo.bpo_havesubobj || bpo.bpo_phys->bpo_subobjs == 0)
109 goto out;
110
111 VERIFY3U(0, ==, dmu_object_info(os, bpo.bpo_phys->bpo_subobjs, &doi));
112 epb = doi.doi_data_block_size / sizeof (uint64_t);
113
114 for (i = bpo.bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
115 uint64_t *objarray;
116 uint64_t offset, blkoff;
117
118 offset = i * sizeof (uint64_t);
119 blkoff = P2PHASE(i, epb);
120
121 if (dbuf == NULL || dbuf->db_offset > offset) {
122 if (dbuf)
123 dmu_buf_rele(dbuf, FTAG);
124 VERIFY3U(0, ==, dmu_buf_hold(os,
125 bpo.bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0));
126 }
127
128 ASSERT3U(offset, >=, dbuf->db_offset);
129 ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
130
131 objarray = dbuf->db_data;
132 bpobj_free(os, objarray[blkoff], tx);
133 }
134 if (dbuf) {
135 dmu_buf_rele(dbuf, FTAG);
136 dbuf = NULL;
137 }
138 VERIFY3U(0, ==, dmu_object_free(os, bpo.bpo_phys->bpo_subobjs, tx));
139
140 out:
141 mutex_exit(&bpo.bpo_lock);
142 bpobj_close(&bpo);
143
144 VERIFY3U(0, ==, dmu_object_free(os, obj, tx));
145 }
146
147 int
bpobj_open(bpobj_t * bpo,objset_t * os,uint64_t object)148 bpobj_open(bpobj_t *bpo, objset_t *os, uint64_t object)
149 {
150 dmu_object_info_t doi;
151 int err;
152
153 err = dmu_object_info(os, object, &doi);
154 if (err)
155 return (err);
156
157 bzero(bpo, sizeof (*bpo));
158 mutex_init(&bpo->bpo_lock, NULL, MUTEX_DEFAULT, NULL);
159
160 ASSERT(bpo->bpo_dbuf == NULL);
161 ASSERT(bpo->bpo_phys == NULL);
162 ASSERT(object != 0);
163 ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ);
164 ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_BPOBJ_HDR);
165
166 err = dmu_bonus_hold(os, object, bpo, &bpo->bpo_dbuf);
167 if (err)
168 return (err);
169
170 bpo->bpo_os = os;
171 bpo->bpo_object = object;
172 bpo->bpo_epb = doi.doi_data_block_size >> SPA_BLKPTRSHIFT;
173 bpo->bpo_havecomp = (doi.doi_bonus_size > BPOBJ_SIZE_V0);
174 bpo->bpo_havesubobj = (doi.doi_bonus_size > BPOBJ_SIZE_V1);
175 bpo->bpo_phys = bpo->bpo_dbuf->db_data;
176 return (0);
177 }
178
179 boolean_t
bpobj_is_open(const bpobj_t * bpo)180 bpobj_is_open(const bpobj_t *bpo)
181 {
182 return (bpo->bpo_object != 0);
183 }
184
185 void
bpobj_close(bpobj_t * bpo)186 bpobj_close(bpobj_t *bpo)
187 {
188 /* Lame workaround for closing a bpobj that was never opened. */
189 if (bpo->bpo_object == 0)
190 return;
191
192 dmu_buf_rele(bpo->bpo_dbuf, bpo);
193 if (bpo->bpo_cached_dbuf != NULL)
194 dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
195 bpo->bpo_dbuf = NULL;
196 bpo->bpo_phys = NULL;
197 bpo->bpo_cached_dbuf = NULL;
198 bpo->bpo_object = 0;
199
200 mutex_destroy(&bpo->bpo_lock);
201 }
202
203 static boolean_t
bpobj_is_empty_impl(bpobj_t * bpo)204 bpobj_is_empty_impl(bpobj_t *bpo)
205 {
206 ASSERT(MUTEX_HELD(&bpo->bpo_lock));
207 return (bpo->bpo_phys->bpo_num_blkptrs == 0 &&
208 (!bpo->bpo_havesubobj || bpo->bpo_phys->bpo_num_subobjs == 0));
209 }
210
211
212 boolean_t
bpobj_is_empty(bpobj_t * bpo)213 bpobj_is_empty(bpobj_t *bpo)
214 {
215 mutex_enter(&bpo->bpo_lock);
216 boolean_t is_empty = bpobj_is_empty_impl(bpo);
217 mutex_exit(&bpo->bpo_lock);
218 return (is_empty);
219 }
220
221 static int
bpobj_iterate_impl(bpobj_t * bpo,bpobj_itor_t func,void * arg,dmu_tx_t * tx,boolean_t free)222 bpobj_iterate_impl(bpobj_t *bpo, bpobj_itor_t func, void *arg, dmu_tx_t *tx,
223 boolean_t free)
224 {
225 dmu_object_info_t doi;
226 int epb;
227 int64_t i;
228 int err = 0;
229 dmu_buf_t *dbuf = NULL;
230
231 ASSERT(bpobj_is_open(bpo));
232 mutex_enter(&bpo->bpo_lock);
233
234 if (free)
235 dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
236
237 for (i = bpo->bpo_phys->bpo_num_blkptrs - 1; i >= 0; i--) {
238 blkptr_t *bparray;
239 blkptr_t *bp;
240 uint64_t offset, blkoff;
241
242 offset = i * sizeof (blkptr_t);
243 blkoff = P2PHASE(i, bpo->bpo_epb);
244
245 if (dbuf == NULL || dbuf->db_offset > offset) {
246 if (dbuf)
247 dmu_buf_rele(dbuf, FTAG);
248 err = dmu_buf_hold(bpo->bpo_os, bpo->bpo_object, offset,
249 FTAG, &dbuf, 0);
250 if (err)
251 break;
252 }
253
254 ASSERT3U(offset, >=, dbuf->db_offset);
255 ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
256
257 bparray = dbuf->db_data;
258 bp = &bparray[blkoff];
259 err = func(arg, bp, tx);
260 if (err)
261 break;
262 if (free) {
263 bpo->bpo_phys->bpo_bytes -=
264 bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
265 ASSERT3S(bpo->bpo_phys->bpo_bytes, >=, 0);
266 if (bpo->bpo_havecomp) {
267 bpo->bpo_phys->bpo_comp -= BP_GET_PSIZE(bp);
268 bpo->bpo_phys->bpo_uncomp -= BP_GET_UCSIZE(bp);
269 }
270 bpo->bpo_phys->bpo_num_blkptrs--;
271 ASSERT3S(bpo->bpo_phys->bpo_num_blkptrs, >=, 0);
272 }
273 }
274 if (dbuf) {
275 dmu_buf_rele(dbuf, FTAG);
276 dbuf = NULL;
277 }
278 if (free) {
279 VERIFY3U(0, ==, dmu_free_range(bpo->bpo_os, bpo->bpo_object,
280 (i + 1) * sizeof (blkptr_t), DMU_OBJECT_END, tx));
281 }
282 if (err || !bpo->bpo_havesubobj || bpo->bpo_phys->bpo_subobjs == 0)
283 goto out;
284
285 ASSERT(bpo->bpo_havecomp);
286 err = dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, &doi);
287 if (err) {
288 mutex_exit(&bpo->bpo_lock);
289 return (err);
290 }
291 ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);
292 epb = doi.doi_data_block_size / sizeof (uint64_t);
293
294 for (i = bpo->bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
295 uint64_t *objarray;
296 uint64_t offset, blkoff;
297 bpobj_t sublist;
298 uint64_t used_before, comp_before, uncomp_before;
299 uint64_t used_after, comp_after, uncomp_after;
300
301 offset = i * sizeof (uint64_t);
302 blkoff = P2PHASE(i, epb);
303
304 if (dbuf == NULL || dbuf->db_offset > offset) {
305 if (dbuf)
306 dmu_buf_rele(dbuf, FTAG);
307 err = dmu_buf_hold(bpo->bpo_os,
308 bpo->bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0);
309 if (err)
310 break;
311 }
312
313 ASSERT3U(offset, >=, dbuf->db_offset);
314 ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
315
316 objarray = dbuf->db_data;
317 err = bpobj_open(&sublist, bpo->bpo_os, objarray[blkoff]);
318 if (err)
319 break;
320 if (free) {
321 err = bpobj_space(&sublist,
322 &used_before, &comp_before, &uncomp_before);
323 if (err != 0) {
324 bpobj_close(&sublist);
325 break;
326 }
327 }
328 err = bpobj_iterate_impl(&sublist, func, arg, tx, free);
329 if (free) {
330 VERIFY3U(0, ==, bpobj_space(&sublist,
331 &used_after, &comp_after, &uncomp_after));
332 bpo->bpo_phys->bpo_bytes -= used_before - used_after;
333 ASSERT3S(bpo->bpo_phys->bpo_bytes, >=, 0);
334 bpo->bpo_phys->bpo_comp -= comp_before - comp_after;
335 bpo->bpo_phys->bpo_uncomp -=
336 uncomp_before - uncomp_after;
337 }
338
339 bpobj_close(&sublist);
340 if (err)
341 break;
342 if (free) {
343 err = dmu_object_free(bpo->bpo_os,
344 objarray[blkoff], tx);
345 if (err)
346 break;
347 bpo->bpo_phys->bpo_num_subobjs--;
348 ASSERT3S(bpo->bpo_phys->bpo_num_subobjs, >=, 0);
349 }
350 }
351 if (dbuf) {
352 dmu_buf_rele(dbuf, FTAG);
353 dbuf = NULL;
354 }
355 if (free) {
356 VERIFY3U(0, ==, dmu_free_range(bpo->bpo_os,
357 bpo->bpo_phys->bpo_subobjs,
358 (i + 1) * sizeof (uint64_t), DMU_OBJECT_END, tx));
359 }
360
361 out:
362 /* If there are no entries, there should be no bytes. */
363 if (bpobj_is_empty_impl(bpo)) {
364 ASSERT0(bpo->bpo_phys->bpo_bytes);
365 ASSERT0(bpo->bpo_phys->bpo_comp);
366 ASSERT0(bpo->bpo_phys->bpo_uncomp);
367 }
368
369 mutex_exit(&bpo->bpo_lock);
370 return (err);
371 }
372
373 /*
374 * Iterate and remove the entries. If func returns nonzero, iteration
375 * will stop and that entry will not be removed.
376 */
377 int
bpobj_iterate(bpobj_t * bpo,bpobj_itor_t func,void * arg,dmu_tx_t * tx)378 bpobj_iterate(bpobj_t *bpo, bpobj_itor_t func, void *arg, dmu_tx_t *tx)
379 {
380 return (bpobj_iterate_impl(bpo, func, arg, tx, B_TRUE));
381 }
382
383 /*
384 * Iterate the entries. If func returns nonzero, iteration will stop.
385 */
386 int
bpobj_iterate_nofree(bpobj_t * bpo,bpobj_itor_t func,void * arg,dmu_tx_t * tx)387 bpobj_iterate_nofree(bpobj_t *bpo, bpobj_itor_t func, void *arg, dmu_tx_t *tx)
388 {
389 return (bpobj_iterate_impl(bpo, func, arg, tx, B_FALSE));
390 }
391
392 void
bpobj_enqueue_subobj(bpobj_t * bpo,uint64_t subobj,dmu_tx_t * tx)393 bpobj_enqueue_subobj(bpobj_t *bpo, uint64_t subobj, dmu_tx_t *tx)
394 {
395 bpobj_t subbpo;
396 uint64_t used, comp, uncomp, subsubobjs;
397
398 ASSERT(bpobj_is_open(bpo));
399 ASSERT(subobj != 0);
400 ASSERT(bpo->bpo_havesubobj);
401 ASSERT(bpo->bpo_havecomp);
402 ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
403
404 if (subobj == dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj) {
405 bpobj_decr_empty(bpo->bpo_os, tx);
406 return;
407 }
408
409 VERIFY3U(0, ==, bpobj_open(&subbpo, bpo->bpo_os, subobj));
410 VERIFY3U(0, ==, bpobj_space(&subbpo, &used, &comp, &uncomp));
411
412 if (bpobj_is_empty(&subbpo)) {
413 /* No point in having an empty subobj. */
414 bpobj_close(&subbpo);
415 bpobj_free(bpo->bpo_os, subobj, tx);
416 return;
417 }
418
419 mutex_enter(&bpo->bpo_lock);
420 dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
421 if (bpo->bpo_phys->bpo_subobjs == 0) {
422 bpo->bpo_phys->bpo_subobjs = dmu_object_alloc(bpo->bpo_os,
423 DMU_OT_BPOBJ_SUBOBJ, SPA_OLD_MAXBLOCKSIZE,
424 DMU_OT_NONE, 0, tx);
425 }
426
427 dmu_object_info_t doi;
428 ASSERT0(dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, &doi));
429 ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);
430
431 dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
432 bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
433 sizeof (subobj), &subobj, tx);
434 bpo->bpo_phys->bpo_num_subobjs++;
435
436 /*
437 * If subobj has only one block of subobjs, then move subobj's
438 * subobjs to bpo's subobj list directly. This reduces
439 * recursion in bpobj_iterate due to nested subobjs.
440 */
441 subsubobjs = subbpo.bpo_phys->bpo_subobjs;
442 if (subsubobjs != 0) {
443 dmu_object_info_t doi;
444
445 VERIFY3U(0, ==, dmu_object_info(bpo->bpo_os, subsubobjs, &doi));
446 if (doi.doi_max_offset == doi.doi_data_block_size) {
447 dmu_buf_t *subdb;
448 uint64_t numsubsub = subbpo.bpo_phys->bpo_num_subobjs;
449
450 VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, subsubobjs,
451 0, FTAG, &subdb, 0));
452 /*
453 * Make sure that we are not asking dmu_write()
454 * to write more data than we have in our buffer.
455 */
456 VERIFY3U(subdb->db_size, >=,
457 numsubsub * sizeof (subobj));
458 dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
459 bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
460 numsubsub * sizeof (subobj), subdb->db_data, tx);
461 dmu_buf_rele(subdb, FTAG);
462 bpo->bpo_phys->bpo_num_subobjs += numsubsub;
463
464 dmu_buf_will_dirty(subbpo.bpo_dbuf, tx);
465 subbpo.bpo_phys->bpo_subobjs = 0;
466 VERIFY3U(0, ==, dmu_object_free(bpo->bpo_os,
467 subsubobjs, tx));
468 }
469 }
470 bpo->bpo_phys->bpo_bytes += used;
471 bpo->bpo_phys->bpo_comp += comp;
472 bpo->bpo_phys->bpo_uncomp += uncomp;
473 mutex_exit(&bpo->bpo_lock);
474
475 bpobj_close(&subbpo);
476 }
477
478 void
bpobj_enqueue(bpobj_t * bpo,const blkptr_t * bp,dmu_tx_t * tx)479 bpobj_enqueue(bpobj_t *bpo, const blkptr_t *bp, dmu_tx_t *tx)
480 {
481 blkptr_t stored_bp = *bp;
482 uint64_t offset;
483 int blkoff;
484 blkptr_t *bparray;
485
486 ASSERT(bpobj_is_open(bpo));
487 ASSERT(!BP_IS_HOLE(bp));
488 ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
489
490 if (BP_IS_EMBEDDED(bp)) {
491 /*
492 * The bpobj will compress better without the payload.
493 *
494 * Note that we store EMBEDDED bp's because they have an
495 * uncompressed size, which must be accounted for. An
496 * alternative would be to add their size to bpo_uncomp
497 * without storing the bp, but that would create additional
498 * complications: bpo_uncomp would be inconsistent with the
499 * set of BP's stored, and bpobj_iterate() wouldn't visit
500 * all the space accounted for in the bpobj.
501 */
502 bzero(&stored_bp, sizeof (stored_bp));
503 stored_bp.blk_prop = bp->blk_prop;
504 stored_bp.blk_birth = bp->blk_birth;
505 } else if (!BP_GET_DEDUP(bp)) {
506 /* The bpobj will compress better without the checksum */
507 bzero(&stored_bp.blk_cksum, sizeof (stored_bp.blk_cksum));
508 }
509
510 /* We never need the fill count. */
511 stored_bp.blk_fill = 0;
512
513 mutex_enter(&bpo->bpo_lock);
514
515 offset = bpo->bpo_phys->bpo_num_blkptrs * sizeof (stored_bp);
516 blkoff = P2PHASE(bpo->bpo_phys->bpo_num_blkptrs, bpo->bpo_epb);
517
518 if (bpo->bpo_cached_dbuf == NULL ||
519 offset < bpo->bpo_cached_dbuf->db_offset ||
520 offset >= bpo->bpo_cached_dbuf->db_offset +
521 bpo->bpo_cached_dbuf->db_size) {
522 if (bpo->bpo_cached_dbuf)
523 dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
524 VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, bpo->bpo_object,
525 offset, bpo, &bpo->bpo_cached_dbuf, 0));
526 }
527
528 dmu_buf_will_dirty(bpo->bpo_cached_dbuf, tx);
529 bparray = bpo->bpo_cached_dbuf->db_data;
530 bparray[blkoff] = stored_bp;
531
532 dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
533 bpo->bpo_phys->bpo_num_blkptrs++;
534 bpo->bpo_phys->bpo_bytes +=
535 bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
536 if (bpo->bpo_havecomp) {
537 bpo->bpo_phys->bpo_comp += BP_GET_PSIZE(bp);
538 bpo->bpo_phys->bpo_uncomp += BP_GET_UCSIZE(bp);
539 }
540 mutex_exit(&bpo->bpo_lock);
541 }
542
543 struct space_range_arg {
544 spa_t *spa;
545 uint64_t mintxg;
546 uint64_t maxtxg;
547 uint64_t used;
548 uint64_t comp;
549 uint64_t uncomp;
550 };
551
552 /* ARGSUSED */
553 static int
space_range_cb(void * arg,const blkptr_t * bp,dmu_tx_t * tx)554 space_range_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
555 {
556 struct space_range_arg *sra = arg;
557
558 if (bp->blk_birth > sra->mintxg && bp->blk_birth <= sra->maxtxg) {
559 if (dsl_pool_sync_context(spa_get_dsl(sra->spa)))
560 sra->used += bp_get_dsize_sync(sra->spa, bp);
561 else
562 sra->used += bp_get_dsize(sra->spa, bp);
563 sra->comp += BP_GET_PSIZE(bp);
564 sra->uncomp += BP_GET_UCSIZE(bp);
565 }
566 return (0);
567 }
568
569 int
bpobj_space(bpobj_t * bpo,uint64_t * usedp,uint64_t * compp,uint64_t * uncompp)570 bpobj_space(bpobj_t *bpo, uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
571 {
572 ASSERT(bpobj_is_open(bpo));
573 mutex_enter(&bpo->bpo_lock);
574
575 *usedp = bpo->bpo_phys->bpo_bytes;
576 if (bpo->bpo_havecomp) {
577 *compp = bpo->bpo_phys->bpo_comp;
578 *uncompp = bpo->bpo_phys->bpo_uncomp;
579 mutex_exit(&bpo->bpo_lock);
580 return (0);
581 } else {
582 mutex_exit(&bpo->bpo_lock);
583 return (bpobj_space_range(bpo, 0, UINT64_MAX,
584 usedp, compp, uncompp));
585 }
586 }
587
588 /*
589 * Return the amount of space in the bpobj which is:
590 * mintxg < blk_birth <= maxtxg
591 */
592 int
bpobj_space_range(bpobj_t * bpo,uint64_t mintxg,uint64_t maxtxg,uint64_t * usedp,uint64_t * compp,uint64_t * uncompp)593 bpobj_space_range(bpobj_t *bpo, uint64_t mintxg, uint64_t maxtxg,
594 uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
595 {
596 struct space_range_arg sra = { 0 };
597 int err;
598
599 ASSERT(bpobj_is_open(bpo));
600
601 /*
602 * As an optimization, if they want the whole txg range, just
603 * get bpo_bytes rather than iterating over the bps.
604 */
605 if (mintxg < TXG_INITIAL && maxtxg == UINT64_MAX && bpo->bpo_havecomp)
606 return (bpobj_space(bpo, usedp, compp, uncompp));
607
608 sra.spa = dmu_objset_spa(bpo->bpo_os);
609 sra.mintxg = mintxg;
610 sra.maxtxg = maxtxg;
611
612 err = bpobj_iterate_nofree(bpo, space_range_cb, &sra, NULL);
613 *usedp = sra.used;
614 *compp = sra.comp;
615 *uncompp = sra.uncomp;
616 return (err);
617 }
618