summaryrefslogtreecommitdiffstats
path: root/module/zfs/dmu_object.c
blob: be4eba91c3598c513d4a965e38d851485ca6994a (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
 * Copyright (c) 2013, 2017 by Delphix. All rights reserved.
 * Copyright 2014 HybridCluster. All rights reserved.
 */

#include <sys/dbuf.h>
#include <sys/dmu.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_tx.h>
#include <sys/dnode.h>
#include <sys/zap.h>
#include <sys/zfeature.h>
#include <sys/dsl_dataset.h>

/*
 * Each of the concurrent object allocators will grab
 * 2^dmu_object_alloc_chunk_shift dnode slots at a time.  The default is to
 * grab 128 slots, which is 4 blocks worth.  This was experimentally
 * determined to be the lowest value that eliminates the measurable effect
 * of lock contention from this code path.
 */
int dmu_object_alloc_chunk_shift = 7;

static uint64_t
dmu_object_alloc_impl(objset_t *os, dmu_object_type_t ot, int blocksize,
    int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen,
    int dnodesize, dnode_t **allocated_dnode, void *tag, dmu_tx_t *tx)
{
	uint64_t object;
	uint64_t L1_dnode_count = DNODES_PER_BLOCK <<
	    (DMU_META_DNODE(os)->dn_indblkshift - SPA_BLKPTRSHIFT);
	dnode_t *dn = NULL;
	int dn_slots = dnodesize >> DNODE_SHIFT;
	boolean_t restarted = B_FALSE;
	uint64_t *cpuobj = NULL;
	int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift;
	int error;

	kpreempt_disable();
	cpuobj = &os->os_obj_next_percpu[CPU_SEQID %
	    os->os_obj_next_percpu_len];
	kpreempt_enable();

	if (dn_slots == 0) {
		dn_slots = DNODE_MIN_SLOTS;
	} else {
		ASSERT3S(dn_slots, >=, DNODE_MIN_SLOTS);
		ASSERT3S(dn_slots, <=, DNODE_MAX_SLOTS);
	}

	/*
	 * The "chunk" of dnodes that is assigned to a CPU-specific
	 * allocator needs to be at least one block's worth, to avoid
	 * lock contention on the dbuf.  It can be at most one L1 block's
	 * worth, so that the "rescan after polishing off a L1's worth"
	 * logic below will be sure to kick in.
	 */
	if (dnodes_per_chunk < DNODES_PER_BLOCK)
		dnodes_per_chunk = DNODES_PER_BLOCK;
	if (dnodes_per_chunk > L1_dnode_count)
		dnodes_per_chunk = L1_dnode_count;

	/*
	 * The caller requested the dnode be returned as a performance
	 * optimization in order to avoid releasing the hold only to
	 * immediately reacquire it.  Since they caller is responsible
	 * for releasing the hold they must provide the tag.
	 */
	if (allocated_dnode != NULL) {
		ASSERT3P(tag, !=, NULL);
	} else {
		ASSERT3P(tag, ==, NULL);
		tag = FTAG;
	}

	object = *cpuobj;
	for (;;) {
		/*
		 * If we finished a chunk of dnodes, get a new one from
		 * the global allocator.
		 */
		if ((P2PHASE(object, dnodes_per_chunk) == 0) ||
		    (P2PHASE(object + dn_slots - 1, dnodes_per_chunk) <
		    dn_slots)) {
			DNODE_STAT_BUMP(dnode_alloc_next_chunk);
			mutex_enter(&os->os_obj_lock);
			ASSERT0(P2PHASE(os->os_obj_next_chunk,
			    dnodes_per_chunk));
			object = os->os_obj_next_chunk;

			/*
			 * Each time we polish off a L1 bp worth of dnodes
			 * (2^12 objects), move to another L1 bp that's
			 * still reasonably sparse (at most 1/4 full). Look
			 * from the beginning at most once per txg. If we
			 * still can't allocate from that L1 block, search
			 * for an empty L0 block, which will quickly skip
			 * to the end of the metadnode if no nearby L0
			 * blocks are empty. This fallback avoids a
			 * pathology where full dnode blocks containing
			 * large dnodes appear sparse because they have a
			 * low blk_fill, leading to many failed allocation
			 * attempts. In the long term a better mechanism to
			 * search for sparse metadnode regions, such as
			 * spacemaps, could be implemented.
			 *
			 * os_scan_dnodes is set during txg sync if enough
			 * objects have been freed since the previous
			 * rescan to justify backfilling again.
			 *
			 * Note that dmu_traverse depends on the behavior
			 * that we use multiple blocks of the dnode object
			 * before going back to reuse objects.  Any change
			 * to this algorithm should preserve that property
			 * or find another solution to the issues described
			 * in traverse_visitbp.
			 */
			if (P2PHASE(object, L1_dnode_count) == 0) {
				uint64_t offset;
				uint64_t blkfill;
				int minlvl;
				if (os->os_rescan_dnodes) {
					offset = 0;
					os->os_rescan_dnodes = B_FALSE;
				} else {
					offset = object << DNODE_SHIFT;
				}
				blkfill = restarted ? 1 : DNODES_PER_BLOCK >> 2;
				minlvl = restarted ? 1 : 2;
				restarted = B_TRUE;
				error = dnode_next_offset(DMU_META_DNODE(os),
				    DNODE_FIND_HOLE, &offset, minlvl,
				    blkfill, 0);
				if (error == 0) {
					object = offset >> DNODE_SHIFT;
				}
			}
			/*
			 * Note: if "restarted", we may find a L0 that
			 * is not suitably aligned.
			 */
			os->os_obj_next_chunk =
			    P2ALIGN(object, dnodes_per_chunk) +
			    dnodes_per_chunk;
			(void) atomic_swap_64(cpuobj, object);
			mutex_exit(&os->os_obj_lock);
		}

		/*
		 * The value of (*cpuobj) before adding dn_slots is the object
		 * ID assigned to us.  The value afterwards is the object ID
		 * assigned to whoever wants to do an allocation next.
		 */
		object = atomic_add_64_nv(cpuobj, dn_slots) - dn_slots;

		/*
		 * XXX We should check for an i/o error here and return
		 * up to our caller.  Actually we should pre-read it in
		 * dmu_tx_assign(), but there is currently no mechanism
		 * to do so.
		 */
		error = dnode_hold_impl(os, object, DNODE_MUST_BE_FREE,
		    dn_slots, tag, &dn);
		if (error == 0) {
			rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
			/*
			 * Another thread could have allocated it; check
			 * again now that we have the struct lock.
			 */
			if (dn->dn_type == DMU_OT_NONE) {
				dnode_allocate(dn, ot, blocksize,
				    indirect_blockshift, bonustype,
				    bonuslen, dn_slots, tx);
				rw_exit(&dn->dn_struct_rwlock);
				dmu_tx_add_new_object(tx, dn);

				/*
				 * Caller requested the allocated dnode be
				 * returned and is responsible for the hold.
				 */
				if (allocated_dnode != NULL)
					*allocated_dnode = dn;
				else
					dnode_rele(dn, tag);

				return (object);
			}
			rw_exit(&dn->dn_struct_rwlock);
			dnode_rele(dn, tag);
			DNODE_STAT_BUMP(dnode_alloc_race);
		}

		/*
		 * Skip to next known valid starting point on error.  This
		 * is the start of the next block of dnodes.
		 */
		if (dmu_object_next(os, &object, B_TRUE, 0) != 0) {
			object = P2ROUNDUP(object + 1, DNODES_PER_BLOCK);
			DNODE_STAT_BUMP(dnode_alloc_next_block);
		}
		(void) atomic_swap_64(cpuobj, object);
	}
}

uint64_t
dmu_object_alloc(objset_t *os, dmu_object_type_t ot, int blocksize,
    dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
{
	return dmu_object_alloc_impl(os, ot, blocksize, 0, bonustype,
	    bonuslen, 0, NULL, NULL, tx);
}

uint64_t
dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize,
    int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen,
    dmu_tx_t *tx)
{
	return dmu_object_alloc_impl(os, ot, blocksize, indirect_blockshift,
	    bonustype, bonuslen, 0, NULL, NULL, tx);
}

uint64_t
dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot, int blocksize,
    dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *tx)
{
	return (dmu_object_alloc_impl(os, ot, blocksize, 0, bonustype,
	    bonuslen, dnodesize, NULL, NULL, tx));
}

/*
 * Allocate a new object and return a pointer to the newly allocated dnode
 * via the allocated_dnode argument.  The returned dnode will be held and
 * the caller is responsible for releasing the hold by calling dnode_rele().
 */
uint64_t
dmu_object_alloc_hold(objset_t *os, dmu_object_type_t ot, int blocksize,
    int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen,
    int dnodesize, dnode_t **allocated_dnode, void *tag, dmu_tx_t *tx)
{
	return (dmu_object_alloc_impl(os, ot, blocksize, indirect_blockshift,
	    bonustype, bonuslen, dnodesize, allocated_dnode, tag, tx));
}

int
dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
    int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
{
	return (dmu_object_claim_dnsize(os, object, ot, blocksize, bonustype,
	    bonuslen, 0, tx));
}

int
dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot,
    int blocksize, dmu_object_type_t bonustype, int bonuslen,
    int dnodesize, dmu_tx_t *tx)
{
	dnode_t *dn;
	int dn_slots = dnodesize >> DNODE_SHIFT;
	int err;

	if (dn_slots == 0)
		dn_slots = DNODE_MIN_SLOTS;
	ASSERT3S(dn_slots, >=, DNODE_MIN_SLOTS);
	ASSERT3S(dn_slots, <=, DNODE_MAX_SLOTS);

	if (object == DMU_META_DNODE_OBJECT && !dmu_tx_private_ok(tx))
		return (SET_ERROR(EBADF));

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_FREE, dn_slots,
	    FTAG, &dn);
	if (err)
		return (err);

	dnode_allocate(dn, ot, blocksize, 0, bonustype, bonuslen, dn_slots, tx);
	dmu_tx_add_new_object(tx, dn);

	dnode_rele(dn, FTAG);

	return (0);
}

int
dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
    int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
{
	return (dmu_object_reclaim_dnsize(os, object, ot, blocksize, bonustype,
	    bonuslen, DNODE_MIN_SIZE, B_FALSE, tx));
}

int
dmu_object_reclaim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot,
    int blocksize, dmu_object_type_t bonustype, int bonuslen, int dnodesize,
    boolean_t keep_spill, dmu_tx_t *tx)
{
	dnode_t *dn;
	int dn_slots = dnodesize >> DNODE_SHIFT;
	int err;

	if (dn_slots == 0)
		dn_slots = DNODE_MIN_SLOTS;

	if (object == DMU_META_DNODE_OBJECT)
		return (SET_ERROR(EBADF));

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0,
	    FTAG, &dn);
	if (err)
		return (err);

	dnode_reallocate(dn, ot, blocksize, bonustype, bonuslen, dn_slots,
	    keep_spill, tx);

	dnode_rele(dn, FTAG);
	return (err);
}

int
dmu_object_rm_spill(objset_t *os, uint64_t object, dmu_tx_t *tx)
{
	dnode_t *dn;
	int err;

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0,
	    FTAG, &dn);
	if (err)
		return (err);

	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
	if (dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
		dbuf_rm_spill(dn, tx);
		dnode_rm_spill(dn, tx);
	}
	rw_exit(&dn->dn_struct_rwlock);

	dnode_rele(dn, FTAG);
	return (err);
}

int
dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx)
{
	dnode_t *dn;
	int err;

	ASSERT(object != DMU_META_DNODE_OBJECT || dmu_tx_private_ok(tx));

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0,
	    FTAG, &dn);
	if (err)
		return (err);

	ASSERT(dn->dn_type != DMU_OT_NONE);
	/*
	 * If we don't create this free range, we'll leak indirect blocks when
	 * we get to freeing the dnode in syncing context.
	 */
	dnode_free_range(dn, 0, DMU_OBJECT_END, tx);
	dnode_free(dn, tx);
	dnode_rele(dn, FTAG);

	return (0);
}

/*
 * Return (in *objectp) the next object which is allocated (or a hole)
 * after *object, taking into account only objects that may have been modified
 * after the specified txg.
 */
int
dmu_object_next(objset_t *os, uint64_t *objectp, boolean_t hole, uint64_t txg)
{
	uint64_t offset;
	uint64_t start_obj;
	struct dsl_dataset *ds = os->os_dsl_dataset;
	int error;

	if (*objectp == 0) {
		start_obj = 1;
	} else if (ds && dsl_dataset_feature_is_active(ds,
	    SPA_FEATURE_LARGE_DNODE)) {
		uint64_t i = *objectp + 1;
		uint64_t last_obj = *objectp | (DNODES_PER_BLOCK - 1);
		dmu_object_info_t doi;

		/*
		 * Scan through the remaining meta dnode block.  The contents
		 * of each slot in the block are known so it can be quickly
		 * checked.  If the block is exhausted without a match then
		 * hand off to dnode_next_offset() for further scanning.
		 */
		while (i <= last_obj) {
			error = dmu_object_info(os, i, &doi);
			if (error == ENOENT) {
				if (hole) {
					*objectp = i;
					return (0);
				} else {
					i++;
				}
			} else if (error == EEXIST) {
				i++;
			} else if (error == 0) {
				if (hole) {
					i += doi.doi_dnodesize >> DNODE_SHIFT;
				} else {
					*objectp = i;
					return (0);
				}
			} else {
				return (error);
			}
		}

		start_obj = i;
	} else {
		start_obj = *objectp + 1;
	}

	offset = start_obj << DNODE_SHIFT;

	error = dnode_next_offset(DMU_META_DNODE(os),
	    (hole ? DNODE_FIND_HOLE : 0), &offset, 0, DNODES_PER_BLOCK, txg);

	*objectp = offset >> DNODE_SHIFT;

	return (error);
}

/*
 * Turn this object from old_type into DMU_OTN_ZAP_METADATA, and bump the
 * refcount on SPA_FEATURE_EXTENSIBLE_DATASET.
 *
 * Only for use from syncing context, on MOS objects.
 */
void
dmu_object_zapify(objset_t *mos, uint64_t object, dmu_object_type_t old_type,
    dmu_tx_t *tx)
{
	dnode_t *dn;

	ASSERT(dmu_tx_is_syncing(tx));

	VERIFY0(dnode_hold(mos, object, FTAG, &dn));
	if (dn->dn_type == DMU_OTN_ZAP_METADATA) {
		dnode_rele(dn, FTAG);
		return;
	}
	ASSERT3U(dn->dn_type, ==, old_type);
	ASSERT0(dn->dn_maxblkid);

	/*
	 * We must initialize the ZAP data before changing the type,
	 * so that concurrent calls to *_is_zapified() can determine if
	 * the object has been completely zapified by checking the type.
	 */
	mzap_create_impl(dn, 0, 0, tx);

	dn->dn_next_type[tx->tx_txg & TXG_MASK] = dn->dn_type =
	    DMU_OTN_ZAP_METADATA;
	dnode_setdirty(dn, tx);
	dnode_rele(dn, FTAG);

	spa_feature_incr(dmu_objset_spa(mos),
	    SPA_FEATURE_EXTENSIBLE_DATASET, tx);
}

void
dmu_object_free_zapified(objset_t *mos, uint64_t object, dmu_tx_t *tx)
{
	dnode_t *dn;
	dmu_object_type_t t;

	ASSERT(dmu_tx_is_syncing(tx));

	VERIFY0(dnode_hold(mos, object, FTAG, &dn));
	t = dn->dn_type;
	dnode_rele(dn, FTAG);

	if (t == DMU_OTN_ZAP_METADATA) {
		spa_feature_decr(dmu_objset_spa(mos),
		    SPA_FEATURE_EXTENSIBLE_DATASET, tx);
	}
	VERIFY0(dmu_object_free(mos, object, tx));
}

EXPORT_SYMBOL(dmu_object_alloc);
EXPORT_SYMBOL(dmu_object_alloc_ibs);
EXPORT_SYMBOL(dmu_object_alloc_dnsize);
EXPORT_SYMBOL(dmu_object_alloc_hold);
EXPORT_SYMBOL(dmu_object_claim);
EXPORT_SYMBOL(dmu_object_claim_dnsize);
EXPORT_SYMBOL(dmu_object_reclaim);
EXPORT_SYMBOL(dmu_object_reclaim_dnsize);
EXPORT_SYMBOL(dmu_object_rm_spill);
EXPORT_SYMBOL(dmu_object_free);
EXPORT_SYMBOL(dmu_object_next);
EXPORT_SYMBOL(dmu_object_zapify);
EXPORT_SYMBOL(dmu_object_free_zapified);

/* BEGIN CSTYLED */
ZFS_MODULE_PARAM(zfs, , dmu_object_alloc_chunk_shift, INT, ZMOD_RW,
	"CPU-specific allocator grabs 2^N objects at once");
/* END CSTYLED */