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) 2010, Oracle and/or its affiliates. All rights reserved.
23  * Copyright (c) 2012, 2019 by Delphix. All rights reserved.
24  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
25  */
26 
27 #include <sys/dmu.h>
28 #include <sys/zap.h>
29 #include <sys/zfs_context.h>
30 #include <sys/dsl_pool.h>
31 #include <sys/dsl_dataset.h>
32 
33 /*
34  * Deadlist concurrency:
35  *
36  * Deadlists can only be modified from the syncing thread.
37  *
38  * Except for dsl_deadlist_insert(), it can only be modified with the
39  * dp_config_rwlock held with RW_WRITER.
40  *
41  * The accessors (dsl_deadlist_space() and dsl_deadlist_space_range()) can
42  * be called concurrently, from open context, with the dl_config_rwlock held
43  * with RW_READER.
44  *
45  * Therefore, we only need to provide locking between dsl_deadlist_insert() and
46  * the accessors, protecting:
47  *     dl_phys->dl_used,comp,uncomp
48  *     and protecting the dl_tree from being loaded.
49  * The locking is provided by dl_lock.  Note that locking on the bpobj_t
50  * provides its own locking, and dl_oldfmt is immutable.
51  */
52 
53 /*
54  * Livelist Overview
55  * ================
56  *
57  * Livelists use the same 'deadlist_t' struct as deadlists and are also used
58  * to track blkptrs over the lifetime of a dataset. Livelists however, belong
59  * to clones and track the blkptrs that are clone-specific (were born after
60  * the clone's creation). The exception is embedded block pointers which are
61  * not included in livelists because they do not need to be freed.
62  *
63  * When it comes time to delete the clone, the livelist provides a quick
64  * reference as to what needs to be freed. For this reason, livelists also track
65  * when clone-specific blkptrs are freed before deletion to prevent double
66  * frees. Each blkptr in a livelist is marked as a FREE or an ALLOC and the
67  * deletion algorithm iterates backwards over the livelist, matching
68  * FREE/ALLOC pairs and then freeing those ALLOCs which remain. livelists
69  * are also updated in the case when blkptrs are remapped: the old version
70  * of the blkptr is cancelled out with a FREE and the new version is tracked
71  * with an ALLOC.
72  *
73  * To bound the amount of memory required for deletion, livelists over a
74  * certain size are spread over multiple entries. Entries are grouped by
75  * birth txg so we can be sure the ALLOC/FREE pair for a given blkptr will
76  * be in the same entry. This allows us to delete livelists incrementally
77  * over multiple syncs, one entry at a time.
78  *
79  * During the lifetime of the clone, livelists can get extremely large.
80  * Their size is managed by periodic condensing (preemptively cancelling out
81  * FREE/ALLOC pairs). Livelists are disabled when a clone is promoted or when
82  * the shared space between the clone and its origin is so small that it
83  * doesn't make sense to use livelists anymore.
84  */
85 
86 /*
87  * The threshold sublist size at which we create a new sub-livelist for the
88  * next txg. However, since blkptrs of the same transaction group must be in
89  * the same sub-list, the actual sublist size may exceed this. When picking the
90  * size we had to balance the fact that larger sublists mean fewer sublists
91  * (decreasing the cost of insertion) against the consideration that sublists
92  * will be loaded into memory and shouldn't take up an inordinate amount of
93  * space. We settled on ~500000 entries, corresponding to roughly 128M.
94  */
95 unsigned long zfs_livelist_max_entries = 500000;
96 
97 /*
98  * We can approximate how much of a performance gain a livelist will give us
99  * based on the percentage of blocks shared between the clone and its origin.
100  * 0 percent shared means that the clone has completely diverged and that the
101  * old method is maximally effective: every read from the block tree will
102  * result in lots of frees. Livelists give us gains when they track blocks
103  * scattered across the tree, when one read in the old method might only
104  * result in a few frees. Once the clone has been overwritten enough,
105  * writes are no longer sparse and we'll no longer get much of a benefit from
106  * tracking them with a livelist. We chose a lower limit of 75 percent shared
107  * (25 percent overwritten). This means that 1/4 of all block pointers will be
108  * freed (e.g. each read frees 256, out of a max of 1024) so we expect livelists
109  * to make deletion 4x faster. Once the amount of shared space drops below this
110  * threshold, the clone will revert to the old deletion method.
111  */
112 int zfs_livelist_min_percent_shared = 75;
113 
114 static int
115 dsl_deadlist_compare(const void *arg1, const void *arg2)
116 {
117 	const dsl_deadlist_entry_t *dle1 = arg1;
118 	const dsl_deadlist_entry_t *dle2 = arg2;
119 
120 	return (TREE_CMP(dle1->dle_mintxg, dle2->dle_mintxg));
121 }
122 
123 static int
124 dsl_deadlist_cache_compare(const void *arg1, const void *arg2)
125 {
126 	const dsl_deadlist_cache_entry_t *dlce1 = arg1;
127 	const dsl_deadlist_cache_entry_t *dlce2 = arg2;
128 
129 	return (TREE_CMP(dlce1->dlce_mintxg, dlce2->dlce_mintxg));
130 }
131 
132 static void
133 dsl_deadlist_load_tree(dsl_deadlist_t *dl)
134 {
135 	zap_cursor_t zc;
136 	zap_attribute_t za;
137 	int error;
138 
139 	ASSERT(MUTEX_HELD(&dl->dl_lock));
140 
141 	ASSERT(!dl->dl_oldfmt);
142 	if (dl->dl_havecache) {
143 		/*
144 		 * After loading the tree, the caller may modify the tree,
145 		 * e.g. to add or remove nodes, or to make a node no longer
146 		 * refer to the empty_bpobj.  These changes would make the
147 		 * dl_cache incorrect.  Therefore we discard the cache here,
148 		 * so that it can't become incorrect.
149 		 */
150 		dsl_deadlist_cache_entry_t *dlce;
151 		void *cookie = NULL;
152 		while ((dlce = avl_destroy_nodes(&dl->dl_cache, &cookie))
153 		    != NULL) {
154 			kmem_free(dlce, sizeof (*dlce));
155 		}
156 		avl_destroy(&dl->dl_cache);
157 		dl->dl_havecache = B_FALSE;
158 	}
159 	if (dl->dl_havetree)
160 		return;
161 
162 	avl_create(&dl->dl_tree, dsl_deadlist_compare,
163 	    sizeof (dsl_deadlist_entry_t),
164 	    offsetof(dsl_deadlist_entry_t, dle_node));
165 	for (zap_cursor_init(&zc, dl->dl_os, dl->dl_object);
166 	    (error = zap_cursor_retrieve(&zc, &za)) == 0;
167 	    zap_cursor_advance(&zc)) {
168 		dsl_deadlist_entry_t *dle = kmem_alloc(sizeof (*dle), KM_SLEEP);
169 		dle->dle_mintxg = zfs_strtonum(za.za_name, NULL);
170 
171 		/*
172 		 * Prefetch all the bpobj's so that we do that i/o
173 		 * in parallel.  Then open them all in a second pass.
174 		 */
175 		dle->dle_bpobj.bpo_object = za.za_first_integer;
176 		dmu_prefetch(dl->dl_os, dle->dle_bpobj.bpo_object,
177 		    0, 0, 0, ZIO_PRIORITY_SYNC_READ);
178 
179 		avl_add(&dl->dl_tree, dle);
180 	}
181 	VERIFY3U(error, ==, ENOENT);
182 	zap_cursor_fini(&zc);
183 
184 	for (dsl_deadlist_entry_t *dle = avl_first(&dl->dl_tree);
185 	    dle != NULL; dle = AVL_NEXT(&dl->dl_tree, dle)) {
186 		VERIFY0(bpobj_open(&dle->dle_bpobj, dl->dl_os,
187 		    dle->dle_bpobj.bpo_object));
188 	}
189 	dl->dl_havetree = B_TRUE;
190 }
191 
192 /*
193  * Load only the non-empty bpobj's into the dl_cache.  The cache is an analog
194  * of the dl_tree, but contains only non-empty_bpobj nodes from the ZAP. It
195  * is used only for gathering space statistics.  The dl_cache has two
196  * advantages over the dl_tree:
197  *
198  * 1. Loading the dl_cache is ~5x faster than loading the dl_tree (if it's
199  * mostly empty_bpobj's), due to less CPU overhead to open the empty_bpobj
200  * many times and to inquire about its (zero) space stats many times.
201  *
202  * 2. The dl_cache uses less memory than the dl_tree.  We only need to load
203  * the dl_tree of snapshots when deleting a snapshot, after which we free the
204  * dl_tree with dsl_deadlist_discard_tree
205  */
206 static void
207 dsl_deadlist_load_cache(dsl_deadlist_t *dl)
208 {
209 	zap_cursor_t zc;
210 	zap_attribute_t za;
211 	int error;
212 
213 	ASSERT(MUTEX_HELD(&dl->dl_lock));
214 
215 	ASSERT(!dl->dl_oldfmt);
216 	if (dl->dl_havecache)
217 		return;
218 
219 	uint64_t empty_bpobj = dmu_objset_pool(dl->dl_os)->dp_empty_bpobj;
220 
221 	avl_create(&dl->dl_cache, dsl_deadlist_cache_compare,
222 	    sizeof (dsl_deadlist_cache_entry_t),
223 	    offsetof(dsl_deadlist_cache_entry_t, dlce_node));
224 	for (zap_cursor_init(&zc, dl->dl_os, dl->dl_object);
225 	    (error = zap_cursor_retrieve(&zc, &za)) == 0;
226 	    zap_cursor_advance(&zc)) {
227 		if (za.za_first_integer == empty_bpobj)
228 			continue;
229 		dsl_deadlist_cache_entry_t *dlce =
230 		    kmem_zalloc(sizeof (*dlce), KM_SLEEP);
231 		dlce->dlce_mintxg = zfs_strtonum(za.za_name, NULL);
232 
233 		/*
234 		 * Prefetch all the bpobj's so that we do that i/o
235 		 * in parallel.  Then open them all in a second pass.
236 		 */
237 		dlce->dlce_bpobj = za.za_first_integer;
238 		dmu_prefetch(dl->dl_os, dlce->dlce_bpobj,
239 		    0, 0, 0, ZIO_PRIORITY_SYNC_READ);
240 		avl_add(&dl->dl_cache, dlce);
241 	}
242 	VERIFY3U(error, ==, ENOENT);
243 	zap_cursor_fini(&zc);
244 
245 	for (dsl_deadlist_cache_entry_t *dlce = avl_first(&dl->dl_cache);
246 	    dlce != NULL; dlce = AVL_NEXT(&dl->dl_cache, dlce)) {
247 		bpobj_t bpo;
248 		VERIFY0(bpobj_open(&bpo, dl->dl_os, dlce->dlce_bpobj));
249 
250 		VERIFY0(bpobj_space(&bpo,
251 		    &dlce->dlce_bytes, &dlce->dlce_comp, &dlce->dlce_uncomp));
252 		bpobj_close(&bpo);
253 	}
254 	dl->dl_havecache = B_TRUE;
255 }
256 
257 /*
258  * Discard the tree to save memory.
259  */
260 void
261 dsl_deadlist_discard_tree(dsl_deadlist_t *dl)
262 {
263 	mutex_enter(&dl->dl_lock);
264 
265 	if (!dl->dl_havetree) {
266 		mutex_exit(&dl->dl_lock);
267 		return;
268 	}
269 	dsl_deadlist_entry_t *dle;
270 	void *cookie = NULL;
271 	while ((dle = avl_destroy_nodes(&dl->dl_tree, &cookie)) != NULL) {
272 		bpobj_close(&dle->dle_bpobj);
273 		kmem_free(dle, sizeof (*dle));
274 	}
275 	avl_destroy(&dl->dl_tree);
276 
277 	dl->dl_havetree = B_FALSE;
278 	mutex_exit(&dl->dl_lock);
279 }
280 
281 void
282 dsl_deadlist_iterate(dsl_deadlist_t *dl, deadlist_iter_t func, void *args)
283 {
284 	dsl_deadlist_entry_t *dle;
285 
286 	ASSERT(dsl_deadlist_is_open(dl));
287 
288 	mutex_enter(&dl->dl_lock);
289 	dsl_deadlist_load_tree(dl);
290 	mutex_exit(&dl->dl_lock);
291 	for (dle = avl_first(&dl->dl_tree); dle != NULL;
292 	    dle = AVL_NEXT(&dl->dl_tree, dle)) {
293 		if (func(args, dle) != 0)
294 			break;
295 	}
296 }
297 
298 void
299 dsl_deadlist_open(dsl_deadlist_t *dl, objset_t *os, uint64_t object)
300 {
301 	dmu_object_info_t doi;
302 
303 	ASSERT(!dsl_deadlist_is_open(dl));
304 
305 	mutex_init(&dl->dl_lock, NULL, MUTEX_DEFAULT, NULL);
306 	dl->dl_os = os;
307 	dl->dl_object = object;
308 	VERIFY0(dmu_bonus_hold(os, object, dl, &dl->dl_dbuf));
309 	dmu_object_info_from_db(dl->dl_dbuf, &doi);
310 	if (doi.doi_type == DMU_OT_BPOBJ) {
311 		dmu_buf_rele(dl->dl_dbuf, dl);
312 		dl->dl_dbuf = NULL;
313 		dl->dl_oldfmt = B_TRUE;
314 		VERIFY0(bpobj_open(&dl->dl_bpobj, os, object));
315 		return;
316 	}
317 
318 	dl->dl_oldfmt = B_FALSE;
319 	dl->dl_phys = dl->dl_dbuf->db_data;
320 	dl->dl_havetree = B_FALSE;
321 	dl->dl_havecache = B_FALSE;
322 }
323 
324 boolean_t
325 dsl_deadlist_is_open(dsl_deadlist_t *dl)
326 {
327 	return (dl->dl_os != NULL);
328 }
329 
330 void
331 dsl_deadlist_close(dsl_deadlist_t *dl)
332 {
333 	ASSERT(dsl_deadlist_is_open(dl));
334 	mutex_destroy(&dl->dl_lock);
335 
336 	if (dl->dl_oldfmt) {
337 		dl->dl_oldfmt = B_FALSE;
338 		bpobj_close(&dl->dl_bpobj);
339 		dl->dl_os = NULL;
340 		dl->dl_object = 0;
341 		return;
342 	}
343 
344 	if (dl->dl_havetree) {
345 		dsl_deadlist_entry_t *dle;
346 		void *cookie = NULL;
347 		while ((dle = avl_destroy_nodes(&dl->dl_tree, &cookie))
348 		    != NULL) {
349 			bpobj_close(&dle->dle_bpobj);
350 			kmem_free(dle, sizeof (*dle));
351 		}
352 		avl_destroy(&dl->dl_tree);
353 	}
354 	if (dl->dl_havecache) {
355 		dsl_deadlist_cache_entry_t *dlce;
356 		void *cookie = NULL;
357 		while ((dlce = avl_destroy_nodes(&dl->dl_cache, &cookie))
358 		    != NULL) {
359 			kmem_free(dlce, sizeof (*dlce));
360 		}
361 		avl_destroy(&dl->dl_cache);
362 	}
363 	dmu_buf_rele(dl->dl_dbuf, dl);
364 	dl->dl_dbuf = NULL;
365 	dl->dl_phys = NULL;
366 	dl->dl_os = NULL;
367 	dl->dl_object = 0;
368 }
369 
370 uint64_t
371 dsl_deadlist_alloc(objset_t *os, dmu_tx_t *tx)
372 {
373 	if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_DEADLISTS)
374 		return (bpobj_alloc(os, SPA_OLD_MAXBLOCKSIZE, tx));
375 	return (zap_create(os, DMU_OT_DEADLIST, DMU_OT_DEADLIST_HDR,
376 	    sizeof (dsl_deadlist_phys_t), tx));
377 }
378 
379 void
380 dsl_deadlist_free(objset_t *os, uint64_t dlobj, dmu_tx_t *tx)
381 {
382 	dmu_object_info_t doi;
383 	zap_cursor_t zc;
384 	zap_attribute_t za;
385 	int error;
386 
387 	VERIFY0(dmu_object_info(os, dlobj, &doi));
388 	if (doi.doi_type == DMU_OT_BPOBJ) {
389 		bpobj_free(os, dlobj, tx);
390 		return;
391 	}
392 
393 	for (zap_cursor_init(&zc, os, dlobj);
394 	    (error = zap_cursor_retrieve(&zc, &za)) == 0;
395 	    zap_cursor_advance(&zc)) {
396 		uint64_t obj = za.za_first_integer;
397 		if (obj == dmu_objset_pool(os)->dp_empty_bpobj)
398 			bpobj_decr_empty(os, tx);
399 		else
400 			bpobj_free(os, obj, tx);
401 	}
402 	VERIFY3U(error, ==, ENOENT);
403 	zap_cursor_fini(&zc);
404 	VERIFY0(dmu_object_free(os, dlobj, tx));
405 }
406 
407 static void
408 dle_enqueue(dsl_deadlist_t *dl, dsl_deadlist_entry_t *dle,
409     const blkptr_t *bp, boolean_t bp_freed, dmu_tx_t *tx)
410 {
411 	ASSERT(MUTEX_HELD(&dl->dl_lock));
412 	if (dle->dle_bpobj.bpo_object ==
413 	    dmu_objset_pool(dl->dl_os)->dp_empty_bpobj) {
414 		uint64_t obj = bpobj_alloc(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
415 		bpobj_close(&dle->dle_bpobj);
416 		bpobj_decr_empty(dl->dl_os, tx);
417 		VERIFY0(bpobj_open(&dle->dle_bpobj, dl->dl_os, obj));
418 		VERIFY0(zap_update_int_key(dl->dl_os, dl->dl_object,
419 		    dle->dle_mintxg, obj, tx));
420 	}
421 	bpobj_enqueue(&dle->dle_bpobj, bp, bp_freed, tx);
422 }
423 
424 static void
425 dle_enqueue_subobj(dsl_deadlist_t *dl, dsl_deadlist_entry_t *dle,
426     uint64_t obj, dmu_tx_t *tx)
427 {
428 	ASSERT(MUTEX_HELD(&dl->dl_lock));
429 	if (dle->dle_bpobj.bpo_object !=
430 	    dmu_objset_pool(dl->dl_os)->dp_empty_bpobj) {
431 		bpobj_enqueue_subobj(&dle->dle_bpobj, obj, tx);
432 	} else {
433 		bpobj_close(&dle->dle_bpobj);
434 		bpobj_decr_empty(dl->dl_os, tx);
435 		VERIFY0(bpobj_open(&dle->dle_bpobj, dl->dl_os, obj));
436 		VERIFY0(zap_update_int_key(dl->dl_os, dl->dl_object,
437 		    dle->dle_mintxg, obj, tx));
438 	}
439 }
440 
441 void
442 dsl_deadlist_insert(dsl_deadlist_t *dl, const blkptr_t *bp, boolean_t bp_freed,
443     dmu_tx_t *tx)
444 {
445 	dsl_deadlist_entry_t dle_tofind;
446 	dsl_deadlist_entry_t *dle;
447 	avl_index_t where;
448 
449 	if (dl->dl_oldfmt) {
450 		bpobj_enqueue(&dl->dl_bpobj, bp, bp_freed, tx);
451 		return;
452 	}
453 
454 	mutex_enter(&dl->dl_lock);
455 	dsl_deadlist_load_tree(dl);
456 
457 	dmu_buf_will_dirty(dl->dl_dbuf, tx);
458 
459 	int sign = bp_freed ? -1 : +1;
460 	dl->dl_phys->dl_used +=
461 	    sign * bp_get_dsize_sync(dmu_objset_spa(dl->dl_os), bp);
462 	dl->dl_phys->dl_comp += sign * BP_GET_PSIZE(bp);
463 	dl->dl_phys->dl_uncomp += sign * BP_GET_UCSIZE(bp);
464 
465 	dle_tofind.dle_mintxg = bp->blk_birth;
466 	dle = avl_find(&dl->dl_tree, &dle_tofind, &where);
467 	if (dle == NULL)
468 		dle = avl_nearest(&dl->dl_tree, where, AVL_BEFORE);
469 	else
470 		dle = AVL_PREV(&dl->dl_tree, dle);
471 
472 	if (dle == NULL) {
473 		zfs_panic_recover("blkptr at %p has invalid BLK_BIRTH %llu",
474 		    bp, (longlong_t)bp->blk_birth);
475 		dle = avl_first(&dl->dl_tree);
476 	}
477 
478 	ASSERT3P(dle, !=, NULL);
479 	dle_enqueue(dl, dle, bp, bp_freed, tx);
480 	mutex_exit(&dl->dl_lock);
481 }
482 
483 int
484 dsl_deadlist_insert_alloc_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
485 {
486 	dsl_deadlist_t *dl = arg;
487 	dsl_deadlist_insert(dl, bp, B_FALSE, tx);
488 	return (0);
489 }
490 
491 int
492 dsl_deadlist_insert_free_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
493 {
494 	dsl_deadlist_t *dl = arg;
495 	dsl_deadlist_insert(dl, bp, B_TRUE, tx);
496 	return (0);
497 }
498 
499 /*
500  * Insert new key in deadlist, which must be > all current entries.
501  * mintxg is not inclusive.
502  */
503 void
504 dsl_deadlist_add_key(dsl_deadlist_t *dl, uint64_t mintxg, dmu_tx_t *tx)
505 {
506 	uint64_t obj;
507 	dsl_deadlist_entry_t *dle;
508 
509 	if (dl->dl_oldfmt)
510 		return;
511 
512 	dle = kmem_alloc(sizeof (*dle), KM_SLEEP);
513 	dle->dle_mintxg = mintxg;
514 
515 	mutex_enter(&dl->dl_lock);
516 	dsl_deadlist_load_tree(dl);
517 
518 	obj = bpobj_alloc_empty(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
519 	VERIFY0(bpobj_open(&dle->dle_bpobj, dl->dl_os, obj));
520 	avl_add(&dl->dl_tree, dle);
521 
522 	VERIFY0(zap_add_int_key(dl->dl_os, dl->dl_object,
523 	    mintxg, obj, tx));
524 	mutex_exit(&dl->dl_lock);
525 }
526 
527 /*
528  * Remove this key, merging its entries into the previous key.
529  */
530 void
531 dsl_deadlist_remove_key(dsl_deadlist_t *dl, uint64_t mintxg, dmu_tx_t *tx)
532 {
533 	dsl_deadlist_entry_t dle_tofind;
534 	dsl_deadlist_entry_t *dle, *dle_prev;
535 
536 	if (dl->dl_oldfmt)
537 		return;
538 	mutex_enter(&dl->dl_lock);
539 	dsl_deadlist_load_tree(dl);
540 
541 	dle_tofind.dle_mintxg = mintxg;
542 	dle = avl_find(&dl->dl_tree, &dle_tofind, NULL);
543 	ASSERT3P(dle, !=, NULL);
544 	dle_prev = AVL_PREV(&dl->dl_tree, dle);
545 
546 	dle_enqueue_subobj(dl, dle_prev, dle->dle_bpobj.bpo_object, tx);
547 
548 	avl_remove(&dl->dl_tree, dle);
549 	bpobj_close(&dle->dle_bpobj);
550 	kmem_free(dle, sizeof (*dle));
551 
552 	VERIFY0(zap_remove_int(dl->dl_os, dl->dl_object, mintxg, tx));
553 	mutex_exit(&dl->dl_lock);
554 }
555 
556 /*
557  * Remove a deadlist entry and all of its contents by removing the entry from
558  * the deadlist's avl tree, freeing the entry's bpobj and adjusting the
559  * deadlist's space accounting accordingly.
560  */
561 void
562 dsl_deadlist_remove_entry(dsl_deadlist_t *dl, uint64_t mintxg, dmu_tx_t *tx)
563 {
564 	uint64_t used, comp, uncomp;
565 	dsl_deadlist_entry_t dle_tofind;
566 	dsl_deadlist_entry_t *dle;
567 	objset_t *os = dl->dl_os;
568 
569 	if (dl->dl_oldfmt)
570 		return;
571 
572 	mutex_enter(&dl->dl_lock);
573 	dsl_deadlist_load_tree(dl);
574 
575 	dle_tofind.dle_mintxg = mintxg;
576 	dle = avl_find(&dl->dl_tree, &dle_tofind, NULL);
577 	VERIFY3P(dle, !=, NULL);
578 
579 	avl_remove(&dl->dl_tree, dle);
580 	VERIFY0(zap_remove_int(os, dl->dl_object, mintxg, tx));
581 	VERIFY0(bpobj_space(&dle->dle_bpobj, &used, &comp, &uncomp));
582 	dmu_buf_will_dirty(dl->dl_dbuf, tx);
583 	dl->dl_phys->dl_used -= used;
584 	dl->dl_phys->dl_comp -= comp;
585 	dl->dl_phys->dl_uncomp -= uncomp;
586 	if (dle->dle_bpobj.bpo_object == dmu_objset_pool(os)->dp_empty_bpobj) {
587 		bpobj_decr_empty(os, tx);
588 	} else {
589 		bpobj_free(os, dle->dle_bpobj.bpo_object, tx);
590 	}
591 	bpobj_close(&dle->dle_bpobj);
592 	kmem_free(dle, sizeof (*dle));
593 	mutex_exit(&dl->dl_lock);
594 }
595 
596 /*
597  * Clear out the contents of a deadlist_entry by freeing its bpobj,
598  * replacing it with an empty bpobj and adjusting the deadlist's
599  * space accounting
600  */
601 void
602 dsl_deadlist_clear_entry(dsl_deadlist_entry_t *dle, dsl_deadlist_t *dl,
603     dmu_tx_t *tx)
604 {
605 	uint64_t new_obj, used, comp, uncomp;
606 	objset_t *os = dl->dl_os;
607 
608 	mutex_enter(&dl->dl_lock);
609 	VERIFY0(zap_remove_int(os, dl->dl_object, dle->dle_mintxg, tx));
610 	VERIFY0(bpobj_space(&dle->dle_bpobj, &used, &comp, &uncomp));
611 	dmu_buf_will_dirty(dl->dl_dbuf, tx);
612 	dl->dl_phys->dl_used -= used;
613 	dl->dl_phys->dl_comp -= comp;
614 	dl->dl_phys->dl_uncomp -= uncomp;
615 	if (dle->dle_bpobj.bpo_object == dmu_objset_pool(os)->dp_empty_bpobj)
616 		bpobj_decr_empty(os, tx);
617 	else
618 		bpobj_free(os, dle->dle_bpobj.bpo_object, tx);
619 	bpobj_close(&dle->dle_bpobj);
620 	new_obj = bpobj_alloc_empty(os, SPA_OLD_MAXBLOCKSIZE, tx);
621 	VERIFY0(bpobj_open(&dle->dle_bpobj, os, new_obj));
622 	VERIFY0(zap_add_int_key(os, dl->dl_object, dle->dle_mintxg,
623 	    new_obj, tx));
624 	ASSERT(bpobj_is_empty(&dle->dle_bpobj));
625 	mutex_exit(&dl->dl_lock);
626 }
627 
628 /*
629  * Return the first entry in deadlist's avl tree
630  */
631 dsl_deadlist_entry_t *
632 dsl_deadlist_first(dsl_deadlist_t *dl)
633 {
634 	dsl_deadlist_entry_t *dle;
635 
636 	mutex_enter(&dl->dl_lock);
637 	dsl_deadlist_load_tree(dl);
638 	dle = avl_first(&dl->dl_tree);
639 	mutex_exit(&dl->dl_lock);
640 
641 	return (dle);
642 }
643 
644 /*
645  * Return the last entry in deadlist's avl tree
646  */
647 dsl_deadlist_entry_t *
648 dsl_deadlist_last(dsl_deadlist_t *dl)
649 {
650 	dsl_deadlist_entry_t *dle;
651 
652 	mutex_enter(&dl->dl_lock);
653 	dsl_deadlist_load_tree(dl);
654 	dle = avl_last(&dl->dl_tree);
655 	mutex_exit(&dl->dl_lock);
656 
657 	return (dle);
658 }
659 
660 /*
661  * Walk ds's snapshots to regenerate generate ZAP & AVL.
662  */
663 static void
664 dsl_deadlist_regenerate(objset_t *os, uint64_t dlobj,
665     uint64_t mrs_obj, dmu_tx_t *tx)
666 {
667 	dsl_deadlist_t dl = { 0 };
668 	dsl_pool_t *dp = dmu_objset_pool(os);
669 
670 	dsl_deadlist_open(&dl, os, dlobj);
671 	if (dl.dl_oldfmt) {
672 		dsl_deadlist_close(&dl);
673 		return;
674 	}
675 
676 	while (mrs_obj != 0) {
677 		dsl_dataset_t *ds;
678 		VERIFY0(dsl_dataset_hold_obj(dp, mrs_obj, FTAG, &ds));
679 		dsl_deadlist_add_key(&dl,
680 		    dsl_dataset_phys(ds)->ds_prev_snap_txg, tx);
681 		mrs_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
682 		dsl_dataset_rele(ds, FTAG);
683 	}
684 	dsl_deadlist_close(&dl);
685 }
686 
687 uint64_t
688 dsl_deadlist_clone(dsl_deadlist_t *dl, uint64_t maxtxg,
689     uint64_t mrs_obj, dmu_tx_t *tx)
690 {
691 	dsl_deadlist_entry_t *dle;
692 	uint64_t newobj;
693 
694 	newobj = dsl_deadlist_alloc(dl->dl_os, tx);
695 
696 	if (dl->dl_oldfmt) {
697 		dsl_deadlist_regenerate(dl->dl_os, newobj, mrs_obj, tx);
698 		return (newobj);
699 	}
700 
701 	mutex_enter(&dl->dl_lock);
702 	dsl_deadlist_load_tree(dl);
703 
704 	for (dle = avl_first(&dl->dl_tree); dle;
705 	    dle = AVL_NEXT(&dl->dl_tree, dle)) {
706 		uint64_t obj;
707 
708 		if (dle->dle_mintxg >= maxtxg)
709 			break;
710 
711 		obj = bpobj_alloc_empty(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
712 		VERIFY0(zap_add_int_key(dl->dl_os, newobj,
713 		    dle->dle_mintxg, obj, tx));
714 	}
715 	mutex_exit(&dl->dl_lock);
716 	return (newobj);
717 }
718 
719 void
720 dsl_deadlist_space(dsl_deadlist_t *dl,
721     uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
722 {
723 	ASSERT(dsl_deadlist_is_open(dl));
724 	if (dl->dl_oldfmt) {
725 		VERIFY0(bpobj_space(&dl->dl_bpobj,
726 		    usedp, compp, uncompp));
727 		return;
728 	}
729 
730 	mutex_enter(&dl->dl_lock);
731 	*usedp = dl->dl_phys->dl_used;
732 	*compp = dl->dl_phys->dl_comp;
733 	*uncompp = dl->dl_phys->dl_uncomp;
734 	mutex_exit(&dl->dl_lock);
735 }
736 
737 /*
738  * return space used in the range (mintxg, maxtxg].
739  * Includes maxtxg, does not include mintxg.
740  * mintxg and maxtxg must both be keys in the deadlist (unless maxtxg is
741  * UINT64_MAX).
742  */
743 void
744 dsl_deadlist_space_range(dsl_deadlist_t *dl, uint64_t mintxg, uint64_t maxtxg,
745     uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
746 {
747 	dsl_deadlist_cache_entry_t *dlce;
748 	dsl_deadlist_cache_entry_t dlce_tofind;
749 	avl_index_t where;
750 
751 	if (dl->dl_oldfmt) {
752 		VERIFY0(bpobj_space_range(&dl->dl_bpobj,
753 		    mintxg, maxtxg, usedp, compp, uncompp));
754 		return;
755 	}
756 
757 	*usedp = *compp = *uncompp = 0;
758 
759 	mutex_enter(&dl->dl_lock);
760 	dsl_deadlist_load_cache(dl);
761 	dlce_tofind.dlce_mintxg = mintxg;
762 	dlce = avl_find(&dl->dl_cache, &dlce_tofind, &where);
763 
764 	/*
765 	 * If this mintxg doesn't exist, it may be an empty_bpobj which
766 	 * is omitted from the sparse tree.  Start at the next non-empty
767 	 * entry.
768 	 */
769 	if (dlce == NULL)
770 		dlce = avl_nearest(&dl->dl_cache, where, AVL_AFTER);
771 
772 	for (; dlce && dlce->dlce_mintxg < maxtxg;
773 	    dlce = AVL_NEXT(&dl->dl_tree, dlce)) {
774 		*usedp += dlce->dlce_bytes;
775 		*compp += dlce->dlce_comp;
776 		*uncompp += dlce->dlce_uncomp;
777 	}
778 
779 	mutex_exit(&dl->dl_lock);
780 }
781 
782 static void
783 dsl_deadlist_insert_bpobj(dsl_deadlist_t *dl, uint64_t obj, uint64_t birth,
784     dmu_tx_t *tx)
785 {
786 	dsl_deadlist_entry_t dle_tofind;
787 	dsl_deadlist_entry_t *dle;
788 	avl_index_t where;
789 	uint64_t used, comp, uncomp;
790 	bpobj_t bpo;
791 
792 	ASSERT(MUTEX_HELD(&dl->dl_lock));
793 
794 	VERIFY0(bpobj_open(&bpo, dl->dl_os, obj));
795 	VERIFY0(bpobj_space(&bpo, &used, &comp, &uncomp));
796 	bpobj_close(&bpo);
797 
798 	dsl_deadlist_load_tree(dl);
799 
800 	dmu_buf_will_dirty(dl->dl_dbuf, tx);
801 	dl->dl_phys->dl_used += used;
802 	dl->dl_phys->dl_comp += comp;
803 	dl->dl_phys->dl_uncomp += uncomp;
804 
805 	dle_tofind.dle_mintxg = birth;
806 	dle = avl_find(&dl->dl_tree, &dle_tofind, &where);
807 	if (dle == NULL)
808 		dle = avl_nearest(&dl->dl_tree, where, AVL_BEFORE);
809 	dle_enqueue_subobj(dl, dle, obj, tx);
810 }
811 
812 static int
813 dsl_deadlist_insert_cb(void *arg, const blkptr_t *bp, boolean_t bp_freed,
814     dmu_tx_t *tx)
815 {
816 	dsl_deadlist_t *dl = arg;
817 	dsl_deadlist_insert(dl, bp, bp_freed, tx);
818 	return (0);
819 }
820 
821 /*
822  * Merge the deadlist pointed to by 'obj' into dl.  obj will be left as
823  * an empty deadlist.
824  */
825 void
826 dsl_deadlist_merge(dsl_deadlist_t *dl, uint64_t obj, dmu_tx_t *tx)
827 {
828 	zap_cursor_t zc;
829 	zap_attribute_t za;
830 	dmu_buf_t *bonus;
831 	dsl_deadlist_phys_t *dlp;
832 	dmu_object_info_t doi;
833 	int error;
834 
835 	VERIFY0(dmu_object_info(dl->dl_os, obj, &doi));
836 	if (doi.doi_type == DMU_OT_BPOBJ) {
837 		bpobj_t bpo;
838 		VERIFY0(bpobj_open(&bpo, dl->dl_os, obj));
839 		VERIFY0(bpobj_iterate(&bpo, dsl_deadlist_insert_cb, dl, tx));
840 		bpobj_close(&bpo);
841 		return;
842 	}
843 
844 	mutex_enter(&dl->dl_lock);
845 	for (zap_cursor_init(&zc, dl->dl_os, obj);
846 	    (error = zap_cursor_retrieve(&zc, &za)) == 0;
847 	    zap_cursor_advance(&zc)) {
848 		uint64_t mintxg = zfs_strtonum(za.za_name, NULL);
849 		dsl_deadlist_insert_bpobj(dl, za.za_first_integer, mintxg, tx);
850 		VERIFY0(zap_remove_int(dl->dl_os, obj, mintxg, tx));
851 	}
852 	VERIFY3U(error, ==, ENOENT);
853 	zap_cursor_fini(&zc);
854 
855 	VERIFY0(dmu_bonus_hold(dl->dl_os, obj, FTAG, &bonus));
856 	dlp = bonus->db_data;
857 	dmu_buf_will_dirty(bonus, tx);
858 	memset(dlp, 0, sizeof (*dlp));
859 	dmu_buf_rele(bonus, FTAG);
860 	mutex_exit(&dl->dl_lock);
861 }
862 
863 /*
864  * Remove entries on dl that are born > mintxg, and put them on the bpobj.
865  */
866 void
867 dsl_deadlist_move_bpobj(dsl_deadlist_t *dl, bpobj_t *bpo, uint64_t mintxg,
868     dmu_tx_t *tx)
869 {
870 	dsl_deadlist_entry_t dle_tofind;
871 	dsl_deadlist_entry_t *dle;
872 	avl_index_t where;
873 
874 	ASSERT(!dl->dl_oldfmt);
875 
876 	mutex_enter(&dl->dl_lock);
877 	dmu_buf_will_dirty(dl->dl_dbuf, tx);
878 	dsl_deadlist_load_tree(dl);
879 
880 	dle_tofind.dle_mintxg = mintxg;
881 	dle = avl_find(&dl->dl_tree, &dle_tofind, &where);
882 	if (dle == NULL)
883 		dle = avl_nearest(&dl->dl_tree, where, AVL_AFTER);
884 	while (dle) {
885 		uint64_t used, comp, uncomp;
886 		dsl_deadlist_entry_t *dle_next;
887 
888 		bpobj_enqueue_subobj(bpo, dle->dle_bpobj.bpo_object, tx);
889 
890 		VERIFY0(bpobj_space(&dle->dle_bpobj,
891 		    &used, &comp, &uncomp));
892 		ASSERT3U(dl->dl_phys->dl_used, >=, used);
893 		ASSERT3U(dl->dl_phys->dl_comp, >=, comp);
894 		ASSERT3U(dl->dl_phys->dl_uncomp, >=, uncomp);
895 		dl->dl_phys->dl_used -= used;
896 		dl->dl_phys->dl_comp -= comp;
897 		dl->dl_phys->dl_uncomp -= uncomp;
898 
899 		VERIFY0(zap_remove_int(dl->dl_os, dl->dl_object,
900 		    dle->dle_mintxg, tx));
901 
902 		dle_next = AVL_NEXT(&dl->dl_tree, dle);
903 		avl_remove(&dl->dl_tree, dle);
904 		bpobj_close(&dle->dle_bpobj);
905 		kmem_free(dle, sizeof (*dle));
906 		dle = dle_next;
907 	}
908 	mutex_exit(&dl->dl_lock);
909 }
910 
911 typedef struct livelist_entry {
912 	blkptr_t le_bp;
913 	uint32_t le_refcnt;
914 	avl_node_t le_node;
915 } livelist_entry_t;
916 
917 static int
918 livelist_compare(const void *larg, const void *rarg)
919 {
920 	const blkptr_t *l = &((livelist_entry_t *)larg)->le_bp;
921 	const blkptr_t *r = &((livelist_entry_t *)rarg)->le_bp;
922 
923 	/* Sort them according to dva[0] */
924 	uint64_t l_dva0_vdev = DVA_GET_VDEV(&l->blk_dva[0]);
925 	uint64_t r_dva0_vdev = DVA_GET_VDEV(&r->blk_dva[0]);
926 
927 	if (l_dva0_vdev != r_dva0_vdev)
928 		return (TREE_CMP(l_dva0_vdev, r_dva0_vdev));
929 
930 	/* if vdevs are equal, sort by offsets. */
931 	uint64_t l_dva0_offset = DVA_GET_OFFSET(&l->blk_dva[0]);
932 	uint64_t r_dva0_offset = DVA_GET_OFFSET(&r->blk_dva[0]);
933 	if (l_dva0_offset == r_dva0_offset)
934 		ASSERT3U(l->blk_birth, ==, r->blk_birth);
935 	return (TREE_CMP(l_dva0_offset, r_dva0_offset));
936 }
937 
938 struct livelist_iter_arg {
939 	avl_tree_t *avl;
940 	bplist_t *to_free;
941 	zthr_t *t;
942 };
943 
944 /*
945  * Expects an AVL tree which is incrementally filled will FREE blkptrs
946  * and used to match up ALLOC/FREE pairs. ALLOC'd blkptrs without a
947  * corresponding FREE are stored in the supplied bplist.
948  *
949  * Note that multiple FREE and ALLOC entries for the same blkptr may
950  * be encountered when dedup is involved. For this reason we keep a
951  * refcount for all the FREE entries of each blkptr and ensure that
952  * each of those FREE entries has a corresponding ALLOC preceding it.
953  */
954 static int
955 dsl_livelist_iterate(void *arg, const blkptr_t *bp, boolean_t bp_freed,
956     dmu_tx_t *tx)
957 {
958 	struct livelist_iter_arg *lia = arg;
959 	avl_tree_t *avl = lia->avl;
960 	bplist_t *to_free = lia->to_free;
961 	zthr_t *t = lia->t;
962 	ASSERT(tx == NULL);
963 
964 	if ((t != NULL) && (zthr_has_waiters(t) || zthr_iscancelled(t)))
965 		return (SET_ERROR(EINTR));
966 
967 	livelist_entry_t node;
968 	node.le_bp = *bp;
969 	livelist_entry_t *found = avl_find(avl, &node, NULL);
970 	if (bp_freed) {
971 		if (found == NULL) {
972 			/* first free entry for this blkptr */
973 			livelist_entry_t *e =
974 			    kmem_alloc(sizeof (livelist_entry_t), KM_SLEEP);
975 			e->le_bp = *bp;
976 			e->le_refcnt = 1;
977 			avl_add(avl, e);
978 		} else {
979 			/* dedup block free */
980 			ASSERT(BP_GET_DEDUP(bp));
981 			ASSERT3U(BP_GET_CHECKSUM(bp), ==,
982 			    BP_GET_CHECKSUM(&found->le_bp));
983 			ASSERT3U(found->le_refcnt + 1, >, found->le_refcnt);
984 			found->le_refcnt++;
985 		}
986 	} else {
987 		if (found == NULL) {
988 			/* block is currently marked as allocated */
989 			bplist_append(to_free, bp);
990 		} else {
991 			/* alloc matches a free entry */
992 			ASSERT3U(found->le_refcnt, !=, 0);
993 			found->le_refcnt--;
994 			if (found->le_refcnt == 0) {
995 				/* all tracked free pairs have been matched */
996 				avl_remove(avl, found);
997 				kmem_free(found, sizeof (livelist_entry_t));
998 			} else {
999 				/*
1000 				 * This is definitely a deduped blkptr so
1001 				 * let's validate it.
1002 				 */
1003 				ASSERT(BP_GET_DEDUP(bp));
1004 				ASSERT3U(BP_GET_CHECKSUM(bp), ==,
1005 				    BP_GET_CHECKSUM(&found->le_bp));
1006 			}
1007 		}
1008 	}
1009 	return (0);
1010 }
1011 
1012 /*
1013  * Accepts a bpobj and a bplist. Will insert into the bplist the blkptrs
1014  * which have an ALLOC entry but no matching FREE
1015  */
1016 int
1017 dsl_process_sub_livelist(bpobj_t *bpobj, bplist_t *to_free, zthr_t *t,
1018     uint64_t *size)
1019 {
1020 	avl_tree_t avl;
1021 	avl_create(&avl, livelist_compare, sizeof (livelist_entry_t),
1022 	    offsetof(livelist_entry_t, le_node));
1023 
1024 	/* process the sublist */
1025 	struct livelist_iter_arg arg = {
1026 	    .avl = &avl,
1027 	    .to_free = to_free,
1028 	    .t = t
1029 	};
1030 	int err = bpobj_iterate_nofree(bpobj, dsl_livelist_iterate, &arg, size);
1031 	VERIFY(err != 0 || avl_numnodes(&avl) == 0);
1032 
1033 	void *cookie = NULL;
1034 	livelist_entry_t *le = NULL;
1035 	while ((le = avl_destroy_nodes(&avl, &cookie)) != NULL) {
1036 		kmem_free(le, sizeof (livelist_entry_t));
1037 	}
1038 	avl_destroy(&avl);
1039 	return (err);
1040 }
1041 
1042 ZFS_MODULE_PARAM(zfs_livelist, zfs_livelist_, max_entries, ULONG, ZMOD_RW,
1043 	"Size to start the next sub-livelist in a livelist");
1044 
1045 ZFS_MODULE_PARAM(zfs_livelist, zfs_livelist_, min_percent_shared, INT, ZMOD_RW,
1046 	"Threshold at which livelist is disabled");
1047