xref: /titanic_44/usr/src/uts/common/fs/zfs/dmu.c (revision 5679c89fcd2facbb4334df8870d3d7a4d2b11673)
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 2008 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #include <sys/dmu.h>
27 #include <sys/dmu_impl.h>
28 #include <sys/dmu_tx.h>
29 #include <sys/dbuf.h>
30 #include <sys/dnode.h>
31 #include <sys/zfs_context.h>
32 #include <sys/dmu_objset.h>
33 #include <sys/dmu_traverse.h>
34 #include <sys/dsl_dataset.h>
35 #include <sys/dsl_dir.h>
36 #include <sys/dsl_pool.h>
37 #include <sys/dsl_synctask.h>
38 #include <sys/dsl_prop.h>
39 #include <sys/dmu_zfetch.h>
40 #include <sys/zfs_ioctl.h>
41 #include <sys/zap.h>
42 #include <sys/zio_checksum.h>
43 #ifdef _KERNEL
44 #include <sys/vmsystm.h>
45 #include <sys/zfs_znode.h>
46 #endif
47 
48 const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = {
49 	{	byteswap_uint8_array,	TRUE,	"unallocated"		},
50 	{	zap_byteswap,		TRUE,	"object directory"	},
51 	{	byteswap_uint64_array,	TRUE,	"object array"		},
52 	{	byteswap_uint8_array,	TRUE,	"packed nvlist"		},
53 	{	byteswap_uint64_array,	TRUE,	"packed nvlist size"	},
54 	{	byteswap_uint64_array,	TRUE,	"bplist"		},
55 	{	byteswap_uint64_array,	TRUE,	"bplist header"		},
56 	{	byteswap_uint64_array,	TRUE,	"SPA space map header"	},
57 	{	byteswap_uint64_array,	TRUE,	"SPA space map"		},
58 	{	byteswap_uint64_array,	TRUE,	"ZIL intent log"	},
59 	{	dnode_buf_byteswap,	TRUE,	"DMU dnode"		},
60 	{	dmu_objset_byteswap,	TRUE,	"DMU objset"		},
61 	{	byteswap_uint64_array,	TRUE,	"DSL directory"		},
62 	{	zap_byteswap,		TRUE,	"DSL directory child map"},
63 	{	zap_byteswap,		TRUE,	"DSL dataset snap map"	},
64 	{	zap_byteswap,		TRUE,	"DSL props"		},
65 	{	byteswap_uint64_array,	TRUE,	"DSL dataset"		},
66 	{	zfs_znode_byteswap,	TRUE,	"ZFS znode"		},
67 	{	zfs_oldacl_byteswap,	TRUE,	"ZFS V0 ACL"		},
68 	{	byteswap_uint8_array,	FALSE,	"ZFS plain file"	},
69 	{	zap_byteswap,		TRUE,	"ZFS directory"		},
70 	{	zap_byteswap,		TRUE,	"ZFS master node"	},
71 	{	zap_byteswap,		TRUE,	"ZFS delete queue"	},
72 	{	byteswap_uint8_array,	FALSE,	"zvol object"		},
73 	{	zap_byteswap,		TRUE,	"zvol prop"		},
74 	{	byteswap_uint8_array,	FALSE,	"other uint8[]"		},
75 	{	byteswap_uint64_array,	FALSE,	"other uint64[]"	},
76 	{	zap_byteswap,		TRUE,	"other ZAP"		},
77 	{	zap_byteswap,		TRUE,	"persistent error log"	},
78 	{	byteswap_uint8_array,	TRUE,	"SPA history"		},
79 	{	byteswap_uint64_array,	TRUE,	"SPA history offsets"	},
80 	{	zap_byteswap,		TRUE,	"Pool properties"	},
81 	{	zap_byteswap,		TRUE,	"DSL permissions"	},
82 	{	zfs_acl_byteswap,	TRUE,	"ZFS ACL"		},
83 	{	byteswap_uint8_array,	TRUE,	"ZFS SYSACL"		},
84 	{	byteswap_uint8_array,	TRUE,	"FUID table"		},
85 	{	byteswap_uint64_array,	TRUE,	"FUID table size"	},
86 	{	zap_byteswap,		TRUE,	"DSL dataset next clones"},
87 	{	zap_byteswap,		TRUE,	"scrub work queue"	},
88 };
89 
90 int
91 dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
92     void *tag, dmu_buf_t **dbp)
93 {
94 	dnode_t *dn;
95 	uint64_t blkid;
96 	dmu_buf_impl_t *db;
97 	int err;
98 
99 	err = dnode_hold(os->os, object, FTAG, &dn);
100 	if (err)
101 		return (err);
102 	blkid = dbuf_whichblock(dn, offset);
103 	rw_enter(&dn->dn_struct_rwlock, RW_READER);
104 	db = dbuf_hold(dn, blkid, tag);
105 	rw_exit(&dn->dn_struct_rwlock);
106 	if (db == NULL) {
107 		err = EIO;
108 	} else {
109 		err = dbuf_read(db, NULL, DB_RF_CANFAIL);
110 		if (err) {
111 			dbuf_rele(db, tag);
112 			db = NULL;
113 		}
114 	}
115 
116 	dnode_rele(dn, FTAG);
117 	*dbp = &db->db;
118 	return (err);
119 }
120 
121 int
122 dmu_bonus_max(void)
123 {
124 	return (DN_MAX_BONUSLEN);
125 }
126 
127 int
128 dmu_set_bonus(dmu_buf_t *db, int newsize, dmu_tx_t *tx)
129 {
130 	dnode_t *dn = ((dmu_buf_impl_t *)db)->db_dnode;
131 
132 	if (dn->dn_bonus != (dmu_buf_impl_t *)db)
133 		return (EINVAL);
134 	if (newsize < 0 || newsize > db->db_size)
135 		return (EINVAL);
136 	dnode_setbonuslen(dn, newsize, tx);
137 	return (0);
138 }
139 
140 /*
141  * returns ENOENT, EIO, or 0.
142  */
143 int
144 dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **dbp)
145 {
146 	dnode_t *dn;
147 	dmu_buf_impl_t *db;
148 	int error;
149 
150 	error = dnode_hold(os->os, object, FTAG, &dn);
151 	if (error)
152 		return (error);
153 
154 	rw_enter(&dn->dn_struct_rwlock, RW_READER);
155 	if (dn->dn_bonus == NULL) {
156 		rw_exit(&dn->dn_struct_rwlock);
157 		rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
158 		if (dn->dn_bonus == NULL)
159 			dbuf_create_bonus(dn);
160 	}
161 	db = dn->dn_bonus;
162 	rw_exit(&dn->dn_struct_rwlock);
163 
164 	/* as long as the bonus buf is held, the dnode will be held */
165 	if (refcount_add(&db->db_holds, tag) == 1)
166 		VERIFY(dnode_add_ref(dn, db));
167 
168 	dnode_rele(dn, FTAG);
169 
170 	VERIFY(0 == dbuf_read(db, NULL, DB_RF_MUST_SUCCEED));
171 
172 	*dbp = &db->db;
173 	return (0);
174 }
175 
176 /*
177  * Note: longer-term, we should modify all of the dmu_buf_*() interfaces
178  * to take a held dnode rather than <os, object> -- the lookup is wasteful,
179  * and can induce severe lock contention when writing to several files
180  * whose dnodes are in the same block.
181  */
182 static int
183 dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset,
184     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp)
185 {
186 	dsl_pool_t *dp = NULL;
187 	dmu_buf_t **dbp;
188 	uint64_t blkid, nblks, i;
189 	uint32_t flags;
190 	int err;
191 	zio_t *zio;
192 	hrtime_t start;
193 
194 	ASSERT(length <= DMU_MAX_ACCESS);
195 
196 	flags = DB_RF_CANFAIL | DB_RF_NEVERWAIT;
197 	if (length > zfetch_array_rd_sz)
198 		flags |= DB_RF_NOPREFETCH;
199 
200 	rw_enter(&dn->dn_struct_rwlock, RW_READER);
201 	if (dn->dn_datablkshift) {
202 		int blkshift = dn->dn_datablkshift;
203 		nblks = (P2ROUNDUP(offset+length, 1ULL<<blkshift) -
204 		    P2ALIGN(offset, 1ULL<<blkshift)) >> blkshift;
205 	} else {
206 		if (offset + length > dn->dn_datablksz) {
207 			zfs_panic_recover("zfs: accessing past end of object "
208 			    "%llx/%llx (size=%u access=%llu+%llu)",
209 			    (longlong_t)dn->dn_objset->
210 			    os_dsl_dataset->ds_object,
211 			    (longlong_t)dn->dn_object, dn->dn_datablksz,
212 			    (longlong_t)offset, (longlong_t)length);
213 			return (EIO);
214 		}
215 		nblks = 1;
216 	}
217 	dbp = kmem_zalloc(sizeof (dmu_buf_t *) * nblks, KM_SLEEP);
218 
219 	if (dn->dn_objset->os_dsl_dataset)
220 		dp = dn->dn_objset->os_dsl_dataset->ds_dir->dd_pool;
221 	if (dp && dsl_pool_sync_context(dp))
222 		start = gethrtime();
223 	zio = zio_root(dn->dn_objset->os_spa, NULL, NULL, ZIO_FLAG_CANFAIL);
224 	blkid = dbuf_whichblock(dn, offset);
225 	for (i = 0; i < nblks; i++) {
226 		dmu_buf_impl_t *db = dbuf_hold(dn, blkid+i, tag);
227 		if (db == NULL) {
228 			rw_exit(&dn->dn_struct_rwlock);
229 			dmu_buf_rele_array(dbp, nblks, tag);
230 			zio_nowait(zio);
231 			return (EIO);
232 		}
233 		/* initiate async i/o */
234 		if (read) {
235 			rw_exit(&dn->dn_struct_rwlock);
236 			(void) dbuf_read(db, zio, flags);
237 			rw_enter(&dn->dn_struct_rwlock, RW_READER);
238 		}
239 		dbp[i] = &db->db;
240 	}
241 	rw_exit(&dn->dn_struct_rwlock);
242 
243 	/* wait for async i/o */
244 	err = zio_wait(zio);
245 	/* track read overhead when we are in sync context */
246 	if (dp && dsl_pool_sync_context(dp))
247 		dp->dp_read_overhead += gethrtime() - start;
248 	if (err) {
249 		dmu_buf_rele_array(dbp, nblks, tag);
250 		return (err);
251 	}
252 
253 	/* wait for other io to complete */
254 	if (read) {
255 		for (i = 0; i < nblks; i++) {
256 			dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbp[i];
257 			mutex_enter(&db->db_mtx);
258 			while (db->db_state == DB_READ ||
259 			    db->db_state == DB_FILL)
260 				cv_wait(&db->db_changed, &db->db_mtx);
261 			if (db->db_state == DB_UNCACHED)
262 				err = EIO;
263 			mutex_exit(&db->db_mtx);
264 			if (err) {
265 				dmu_buf_rele_array(dbp, nblks, tag);
266 				return (err);
267 			}
268 		}
269 	}
270 
271 	*numbufsp = nblks;
272 	*dbpp = dbp;
273 	return (0);
274 }
275 
276 static int
277 dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
278     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp)
279 {
280 	dnode_t *dn;
281 	int err;
282 
283 	err = dnode_hold(os->os, object, FTAG, &dn);
284 	if (err)
285 		return (err);
286 
287 	err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
288 	    numbufsp, dbpp);
289 
290 	dnode_rele(dn, FTAG);
291 
292 	return (err);
293 }
294 
295 int
296 dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
297     uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp)
298 {
299 	dnode_t *dn = ((dmu_buf_impl_t *)db)->db_dnode;
300 	int err;
301 
302 	err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
303 	    numbufsp, dbpp);
304 
305 	return (err);
306 }
307 
308 void
309 dmu_buf_rele_array(dmu_buf_t **dbp_fake, int numbufs, void *tag)
310 {
311 	int i;
312 	dmu_buf_impl_t **dbp = (dmu_buf_impl_t **)dbp_fake;
313 
314 	if (numbufs == 0)
315 		return;
316 
317 	for (i = 0; i < numbufs; i++) {
318 		if (dbp[i])
319 			dbuf_rele(dbp[i], tag);
320 	}
321 
322 	kmem_free(dbp, sizeof (dmu_buf_t *) * numbufs);
323 }
324 
325 void
326 dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset, uint64_t len)
327 {
328 	dnode_t *dn;
329 	uint64_t blkid;
330 	int nblks, i, err;
331 
332 	if (zfs_prefetch_disable)
333 		return;
334 
335 	if (len == 0) {  /* they're interested in the bonus buffer */
336 		dn = os->os->os_meta_dnode;
337 
338 		if (object == 0 || object >= DN_MAX_OBJECT)
339 			return;
340 
341 		rw_enter(&dn->dn_struct_rwlock, RW_READER);
342 		blkid = dbuf_whichblock(dn, object * sizeof (dnode_phys_t));
343 		dbuf_prefetch(dn, blkid);
344 		rw_exit(&dn->dn_struct_rwlock);
345 		return;
346 	}
347 
348 	/*
349 	 * XXX - Note, if the dnode for the requested object is not
350 	 * already cached, we will do a *synchronous* read in the
351 	 * dnode_hold() call.  The same is true for any indirects.
352 	 */
353 	err = dnode_hold(os->os, object, FTAG, &dn);
354 	if (err != 0)
355 		return;
356 
357 	rw_enter(&dn->dn_struct_rwlock, RW_READER);
358 	if (dn->dn_datablkshift) {
359 		int blkshift = dn->dn_datablkshift;
360 		nblks = (P2ROUNDUP(offset+len, 1<<blkshift) -
361 		    P2ALIGN(offset, 1<<blkshift)) >> blkshift;
362 	} else {
363 		nblks = (offset < dn->dn_datablksz);
364 	}
365 
366 	if (nblks != 0) {
367 		blkid = dbuf_whichblock(dn, offset);
368 		for (i = 0; i < nblks; i++)
369 			dbuf_prefetch(dn, blkid+i);
370 	}
371 
372 	rw_exit(&dn->dn_struct_rwlock);
373 
374 	dnode_rele(dn, FTAG);
375 }
376 
377 static int
378 get_next_chunk(dnode_t *dn, uint64_t *offset, uint64_t limit)
379 {
380 	uint64_t len = *offset - limit;
381 	uint64_t chunk_len = dn->dn_datablksz * DMU_MAX_DELETEBLKCNT;
382 	uint64_t subchunk =
383 	    dn->dn_datablksz * EPB(dn->dn_indblkshift, SPA_BLKPTRSHIFT);
384 
385 	ASSERT(limit <= *offset);
386 
387 	if (len <= chunk_len) {
388 		*offset = limit;
389 		return (0);
390 	}
391 
392 	ASSERT(ISP2(subchunk));
393 
394 	while (*offset > limit) {
395 		uint64_t initial_offset = P2ROUNDUP(*offset, subchunk);
396 		uint64_t delta;
397 		int err;
398 
399 		/* skip over allocated data */
400 		err = dnode_next_offset(dn,
401 		    DNODE_FIND_HOLE|DNODE_FIND_BACKWARDS, offset, 1, 1, 0);
402 		if (err == ESRCH)
403 			*offset = limit;
404 		else if (err)
405 			return (err);
406 
407 		ASSERT3U(*offset, <=, initial_offset);
408 		*offset = P2ALIGN(*offset, subchunk);
409 		delta = initial_offset - *offset;
410 		if (delta >= chunk_len) {
411 			*offset += delta - chunk_len;
412 			return (0);
413 		}
414 		chunk_len -= delta;
415 
416 		/* skip over unallocated data */
417 		err = dnode_next_offset(dn,
418 		    DNODE_FIND_BACKWARDS, offset, 1, 1, 0);
419 		if (err == ESRCH)
420 			*offset = limit;
421 		else if (err)
422 			return (err);
423 
424 		if (*offset < limit)
425 			*offset = limit;
426 		ASSERT3U(*offset, <, initial_offset);
427 	}
428 	return (0);
429 }
430 
431 static int
432 dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset,
433     uint64_t length, boolean_t free_dnode)
434 {
435 	dmu_tx_t *tx;
436 	uint64_t object_size, start, end, len;
437 	boolean_t trunc = (length == DMU_OBJECT_END);
438 	int align, err;
439 
440 	align = 1 << dn->dn_datablkshift;
441 	ASSERT(align > 0);
442 	object_size = align == 1 ? dn->dn_datablksz :
443 	    (dn->dn_maxblkid + 1) << dn->dn_datablkshift;
444 
445 	if (trunc || (end = offset + length) > object_size)
446 		end = object_size;
447 	if (end <= offset)
448 		return (0);
449 	length = end - offset;
450 
451 	while (length) {
452 		start = end;
453 		err = get_next_chunk(dn, &start, offset);
454 		if (err)
455 			return (err);
456 		len = trunc ? DMU_OBJECT_END : end - start;
457 
458 		tx = dmu_tx_create(os);
459 		dmu_tx_hold_free(tx, dn->dn_object, start, len);
460 		err = dmu_tx_assign(tx, TXG_WAIT);
461 		if (err) {
462 			dmu_tx_abort(tx);
463 			return (err);
464 		}
465 
466 		dnode_free_range(dn, start, trunc ? -1 : len, tx);
467 
468 		if (start == 0 && free_dnode) {
469 			ASSERT(trunc);
470 			dnode_free(dn, tx);
471 		}
472 
473 		length -= end - start;
474 
475 		dmu_tx_commit(tx);
476 		end = start;
477 	}
478 	return (0);
479 }
480 
481 int
482 dmu_free_long_range(objset_t *os, uint64_t object,
483     uint64_t offset, uint64_t length)
484 {
485 	dnode_t *dn;
486 	int err;
487 
488 	err = dnode_hold(os->os, object, FTAG, &dn);
489 	if (err != 0)
490 		return (err);
491 	err = dmu_free_long_range_impl(os, dn, offset, length, FALSE);
492 	dnode_rele(dn, FTAG);
493 	return (err);
494 }
495 
496 int
497 dmu_free_object(objset_t *os, uint64_t object)
498 {
499 	dnode_t *dn;
500 	dmu_tx_t *tx;
501 	int err;
502 
503 	err = dnode_hold_impl(os->os, object, DNODE_MUST_BE_ALLOCATED,
504 	    FTAG, &dn);
505 	if (err != 0)
506 		return (err);
507 	if (dn->dn_nlevels == 1) {
508 		tx = dmu_tx_create(os);
509 		dmu_tx_hold_bonus(tx, object);
510 		dmu_tx_hold_free(tx, dn->dn_object, 0, DMU_OBJECT_END);
511 		err = dmu_tx_assign(tx, TXG_WAIT);
512 		if (err == 0) {
513 			dnode_free_range(dn, 0, DMU_OBJECT_END, tx);
514 			dnode_free(dn, tx);
515 			dmu_tx_commit(tx);
516 		} else {
517 			dmu_tx_abort(tx);
518 		}
519 	} else {
520 		err = dmu_free_long_range_impl(os, dn, 0, DMU_OBJECT_END, TRUE);
521 	}
522 	dnode_rele(dn, FTAG);
523 	return (err);
524 }
525 
526 int
527 dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
528     uint64_t size, dmu_tx_t *tx)
529 {
530 	dnode_t *dn;
531 	int err = dnode_hold(os->os, object, FTAG, &dn);
532 	if (err)
533 		return (err);
534 	ASSERT(offset < UINT64_MAX);
535 	ASSERT(size == -1ULL || size <= UINT64_MAX - offset);
536 	dnode_free_range(dn, offset, size, tx);
537 	dnode_rele(dn, FTAG);
538 	return (0);
539 }
540 
541 int
542 dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
543     void *buf)
544 {
545 	dnode_t *dn;
546 	dmu_buf_t **dbp;
547 	int numbufs, i, err;
548 
549 	err = dnode_hold(os->os, object, FTAG, &dn);
550 	if (err)
551 		return (err);
552 
553 	/*
554 	 * Deal with odd block sizes, where there can't be data past the first
555 	 * block.  If we ever do the tail block optimization, we will need to
556 	 * handle that here as well.
557 	 */
558 	if (dn->dn_datablkshift == 0) {
559 		int newsz = offset > dn->dn_datablksz ? 0 :
560 		    MIN(size, dn->dn_datablksz - offset);
561 		bzero((char *)buf + newsz, size - newsz);
562 		size = newsz;
563 	}
564 
565 	while (size > 0) {
566 		uint64_t mylen = MIN(size, DMU_MAX_ACCESS / 2);
567 
568 		/*
569 		 * NB: we could do this block-at-a-time, but it's nice
570 		 * to be reading in parallel.
571 		 */
572 		err = dmu_buf_hold_array_by_dnode(dn, offset, mylen,
573 		    TRUE, FTAG, &numbufs, &dbp);
574 		if (err)
575 			break;
576 
577 		for (i = 0; i < numbufs; i++) {
578 			int tocpy;
579 			int bufoff;
580 			dmu_buf_t *db = dbp[i];
581 
582 			ASSERT(size > 0);
583 
584 			bufoff = offset - db->db_offset;
585 			tocpy = (int)MIN(db->db_size - bufoff, size);
586 
587 			bcopy((char *)db->db_data + bufoff, buf, tocpy);
588 
589 			offset += tocpy;
590 			size -= tocpy;
591 			buf = (char *)buf + tocpy;
592 		}
593 		dmu_buf_rele_array(dbp, numbufs, FTAG);
594 	}
595 	dnode_rele(dn, FTAG);
596 	return (err);
597 }
598 
599 void
600 dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
601     const void *buf, dmu_tx_t *tx)
602 {
603 	dmu_buf_t **dbp;
604 	int numbufs, i;
605 
606 	if (size == 0)
607 		return;
608 
609 	VERIFY(0 == dmu_buf_hold_array(os, object, offset, size,
610 	    FALSE, FTAG, &numbufs, &dbp));
611 
612 	for (i = 0; i < numbufs; i++) {
613 		int tocpy;
614 		int bufoff;
615 		dmu_buf_t *db = dbp[i];
616 
617 		ASSERT(size > 0);
618 
619 		bufoff = offset - db->db_offset;
620 		tocpy = (int)MIN(db->db_size - bufoff, size);
621 
622 		ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
623 
624 		if (tocpy == db->db_size)
625 			dmu_buf_will_fill(db, tx);
626 		else
627 			dmu_buf_will_dirty(db, tx);
628 
629 		bcopy(buf, (char *)db->db_data + bufoff, tocpy);
630 
631 		if (tocpy == db->db_size)
632 			dmu_buf_fill_done(db, tx);
633 
634 		offset += tocpy;
635 		size -= tocpy;
636 		buf = (char *)buf + tocpy;
637 	}
638 	dmu_buf_rele_array(dbp, numbufs, FTAG);
639 }
640 
641 void
642 dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
643     dmu_tx_t *tx)
644 {
645 	dmu_buf_t **dbp;
646 	int numbufs, i;
647 
648 	if (size == 0)
649 		return;
650 
651 	VERIFY(0 == dmu_buf_hold_array(os, object, offset, size,
652 	    FALSE, FTAG, &numbufs, &dbp));
653 
654 	for (i = 0; i < numbufs; i++) {
655 		dmu_buf_t *db = dbp[i];
656 
657 		dmu_buf_will_not_fill(db, tx);
658 	}
659 	dmu_buf_rele_array(dbp, numbufs, FTAG);
660 }
661 
662 #ifdef _KERNEL
663 int
664 dmu_read_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size)
665 {
666 	dmu_buf_t **dbp;
667 	int numbufs, i, err;
668 
669 	/*
670 	 * NB: we could do this block-at-a-time, but it's nice
671 	 * to be reading in parallel.
672 	 */
673 	err = dmu_buf_hold_array(os, object, uio->uio_loffset, size, TRUE, FTAG,
674 	    &numbufs, &dbp);
675 	if (err)
676 		return (err);
677 
678 	for (i = 0; i < numbufs; i++) {
679 		int tocpy;
680 		int bufoff;
681 		dmu_buf_t *db = dbp[i];
682 
683 		ASSERT(size > 0);
684 
685 		bufoff = uio->uio_loffset - db->db_offset;
686 		tocpy = (int)MIN(db->db_size - bufoff, size);
687 
688 		err = uiomove((char *)db->db_data + bufoff, tocpy,
689 		    UIO_READ, uio);
690 		if (err)
691 			break;
692 
693 		size -= tocpy;
694 	}
695 	dmu_buf_rele_array(dbp, numbufs, FTAG);
696 
697 	return (err);
698 }
699 
700 int
701 dmu_write_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size,
702     dmu_tx_t *tx)
703 {
704 	dmu_buf_t **dbp;
705 	int numbufs, i;
706 	int err = 0;
707 
708 	if (size == 0)
709 		return (0);
710 
711 	err = dmu_buf_hold_array(os, object, uio->uio_loffset, size,
712 	    FALSE, FTAG, &numbufs, &dbp);
713 	if (err)
714 		return (err);
715 
716 	for (i = 0; i < numbufs; i++) {
717 		int tocpy;
718 		int bufoff;
719 		dmu_buf_t *db = dbp[i];
720 
721 		ASSERT(size > 0);
722 
723 		bufoff = uio->uio_loffset - db->db_offset;
724 		tocpy = (int)MIN(db->db_size - bufoff, size);
725 
726 		ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
727 
728 		if (tocpy == db->db_size)
729 			dmu_buf_will_fill(db, tx);
730 		else
731 			dmu_buf_will_dirty(db, tx);
732 
733 		/*
734 		 * XXX uiomove could block forever (eg. nfs-backed
735 		 * pages).  There needs to be a uiolockdown() function
736 		 * to lock the pages in memory, so that uiomove won't
737 		 * block.
738 		 */
739 		err = uiomove((char *)db->db_data + bufoff, tocpy,
740 		    UIO_WRITE, uio);
741 
742 		if (tocpy == db->db_size)
743 			dmu_buf_fill_done(db, tx);
744 
745 		if (err)
746 			break;
747 
748 		size -= tocpy;
749 	}
750 	dmu_buf_rele_array(dbp, numbufs, FTAG);
751 	return (err);
752 }
753 
754 int
755 dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
756     page_t *pp, dmu_tx_t *tx)
757 {
758 	dmu_buf_t **dbp;
759 	int numbufs, i;
760 	int err;
761 
762 	if (size == 0)
763 		return (0);
764 
765 	err = dmu_buf_hold_array(os, object, offset, size,
766 	    FALSE, FTAG, &numbufs, &dbp);
767 	if (err)
768 		return (err);
769 
770 	for (i = 0; i < numbufs; i++) {
771 		int tocpy, copied, thiscpy;
772 		int bufoff;
773 		dmu_buf_t *db = dbp[i];
774 		caddr_t va;
775 
776 		ASSERT(size > 0);
777 		ASSERT3U(db->db_size, >=, PAGESIZE);
778 
779 		bufoff = offset - db->db_offset;
780 		tocpy = (int)MIN(db->db_size - bufoff, size);
781 
782 		ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
783 
784 		if (tocpy == db->db_size)
785 			dmu_buf_will_fill(db, tx);
786 		else
787 			dmu_buf_will_dirty(db, tx);
788 
789 		for (copied = 0; copied < tocpy; copied += PAGESIZE) {
790 			ASSERT3U(pp->p_offset, ==, db->db_offset + bufoff);
791 			thiscpy = MIN(PAGESIZE, tocpy - copied);
792 			va = zfs_map_page(pp, S_READ);
793 			bcopy(va, (char *)db->db_data + bufoff, thiscpy);
794 			zfs_unmap_page(pp, va);
795 			pp = pp->p_next;
796 			bufoff += PAGESIZE;
797 		}
798 
799 		if (tocpy == db->db_size)
800 			dmu_buf_fill_done(db, tx);
801 
802 		if (err)
803 			break;
804 
805 		offset += tocpy;
806 		size -= tocpy;
807 	}
808 	dmu_buf_rele_array(dbp, numbufs, FTAG);
809 	return (err);
810 }
811 #endif
812 
813 typedef struct {
814 	dbuf_dirty_record_t	*dr;
815 	dmu_sync_cb_t		*done;
816 	void			*arg;
817 } dmu_sync_arg_t;
818 
819 /* ARGSUSED */
820 static void
821 dmu_sync_ready(zio_t *zio, arc_buf_t *buf, void *varg)
822 {
823 	blkptr_t *bp = zio->io_bp;
824 
825 	if (!BP_IS_HOLE(bp)) {
826 		dmu_sync_arg_t *in = varg;
827 		dbuf_dirty_record_t *dr = in->dr;
828 		dmu_buf_impl_t *db = dr->dr_dbuf;
829 		ASSERT(BP_GET_TYPE(bp) == db->db_dnode->dn_type);
830 		ASSERT(BP_GET_LEVEL(bp) == 0);
831 		bp->blk_fill = 1;
832 	}
833 }
834 
835 /* ARGSUSED */
836 static void
837 dmu_sync_done(zio_t *zio, arc_buf_t *buf, void *varg)
838 {
839 	dmu_sync_arg_t *in = varg;
840 	dbuf_dirty_record_t *dr = in->dr;
841 	dmu_buf_impl_t *db = dr->dr_dbuf;
842 	dmu_sync_cb_t *done = in->done;
843 
844 	mutex_enter(&db->db_mtx);
845 	ASSERT(dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC);
846 	dr->dt.dl.dr_overridden_by = *zio->io_bp; /* structure assignment */
847 	dr->dt.dl.dr_override_state = DR_OVERRIDDEN;
848 	cv_broadcast(&db->db_changed);
849 	mutex_exit(&db->db_mtx);
850 
851 	if (done)
852 		done(&(db->db), in->arg);
853 
854 	kmem_free(in, sizeof (dmu_sync_arg_t));
855 }
856 
857 /*
858  * Intent log support: sync the block associated with db to disk.
859  * N.B. and XXX: the caller is responsible for making sure that the
860  * data isn't changing while dmu_sync() is writing it.
861  *
862  * Return values:
863  *
864  *	EEXIST: this txg has already been synced, so there's nothing to to.
865  *		The caller should not log the write.
866  *
867  *	ENOENT: the block was dbuf_free_range()'d, so there's nothing to do.
868  *		The caller should not log the write.
869  *
870  *	EALREADY: this block is already in the process of being synced.
871  *		The caller should track its progress (somehow).
872  *
873  *	EINPROGRESS: the IO has been initiated.
874  *		The caller should log this blkptr in the callback.
875  *
876  *	0: completed.  Sets *bp to the blkptr just written.
877  *		The caller should log this blkptr immediately.
878  */
879 int
880 dmu_sync(zio_t *pio, dmu_buf_t *db_fake,
881     blkptr_t *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg)
882 {
883 	dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
884 	objset_impl_t *os = db->db_objset;
885 	dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
886 	tx_state_t *tx = &dp->dp_tx;
887 	dbuf_dirty_record_t *dr;
888 	dmu_sync_arg_t *in;
889 	zbookmark_t zb;
890 	writeprops_t wp = { 0 };
891 	zio_t *zio;
892 	int err;
893 
894 	ASSERT(BP_IS_HOLE(bp));
895 	ASSERT(txg != 0);
896 
897 	dprintf("dmu_sync txg=%llu, s,o,q %llu %llu %llu\n",
898 	    txg, tx->tx_synced_txg, tx->tx_open_txg, tx->tx_quiesced_txg);
899 
900 	/*
901 	 * XXX - would be nice if we could do this without suspending...
902 	 */
903 	txg_suspend(dp);
904 
905 	/*
906 	 * If this txg already synced, there's nothing to do.
907 	 */
908 	if (txg <= tx->tx_synced_txg) {
909 		txg_resume(dp);
910 		/*
911 		 * If we're running ziltest, we need the blkptr regardless.
912 		 */
913 		if (txg > spa_freeze_txg(dp->dp_spa)) {
914 			/* if db_blkptr == NULL, this was an empty write */
915 			if (db->db_blkptr)
916 				*bp = *db->db_blkptr; /* structure assignment */
917 			return (0);
918 		}
919 		return (EEXIST);
920 	}
921 
922 	mutex_enter(&db->db_mtx);
923 
924 	if (txg == tx->tx_syncing_txg) {
925 		while (db->db_data_pending) {
926 			/*
927 			 * IO is in-progress.  Wait for it to finish.
928 			 * XXX - would be nice to be able to somehow "attach"
929 			 * this zio to the parent zio passed in.
930 			 */
931 			cv_wait(&db->db_changed, &db->db_mtx);
932 			if (!db->db_data_pending &&
933 			    db->db_blkptr && BP_IS_HOLE(db->db_blkptr)) {
934 				/*
935 				 * IO was compressed away
936 				 */
937 				*bp = *db->db_blkptr; /* structure assignment */
938 				mutex_exit(&db->db_mtx);
939 				txg_resume(dp);
940 				return (0);
941 			}
942 			ASSERT(db->db_data_pending ||
943 			    (db->db_blkptr && db->db_blkptr->blk_birth == txg));
944 		}
945 
946 		if (db->db_blkptr && db->db_blkptr->blk_birth == txg) {
947 			/*
948 			 * IO is already completed.
949 			 */
950 			*bp = *db->db_blkptr; /* structure assignment */
951 			mutex_exit(&db->db_mtx);
952 			txg_resume(dp);
953 			return (0);
954 		}
955 	}
956 
957 	dr = db->db_last_dirty;
958 	while (dr && dr->dr_txg > txg)
959 		dr = dr->dr_next;
960 	if (dr == NULL || dr->dr_txg < txg) {
961 		/*
962 		 * This dbuf isn't dirty, must have been free_range'd.
963 		 * There's no need to log writes to freed blocks, so we're done.
964 		 */
965 		mutex_exit(&db->db_mtx);
966 		txg_resume(dp);
967 		return (ENOENT);
968 	}
969 
970 	ASSERT(dr->dr_txg == txg);
971 	if (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC) {
972 		/*
973 		 * We have already issued a sync write for this buffer.
974 		 */
975 		mutex_exit(&db->db_mtx);
976 		txg_resume(dp);
977 		return (EALREADY);
978 	} else if (dr->dt.dl.dr_override_state == DR_OVERRIDDEN) {
979 		/*
980 		 * This buffer has already been synced.  It could not
981 		 * have been dirtied since, or we would have cleared the state.
982 		 */
983 		*bp = dr->dt.dl.dr_overridden_by; /* structure assignment */
984 		mutex_exit(&db->db_mtx);
985 		txg_resume(dp);
986 		return (0);
987 	}
988 
989 	dr->dt.dl.dr_override_state = DR_IN_DMU_SYNC;
990 	in = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP);
991 	in->dr = dr;
992 	in->done = done;
993 	in->arg = arg;
994 	mutex_exit(&db->db_mtx);
995 	txg_resume(dp);
996 
997 	zb.zb_objset = os->os_dsl_dataset->ds_object;
998 	zb.zb_object = db->db.db_object;
999 	zb.zb_level = db->db_level;
1000 	zb.zb_blkid = db->db_blkid;
1001 
1002 	wp.wp_type = db->db_dnode->dn_type;
1003 	wp.wp_level = db->db_level;
1004 	wp.wp_copies = os->os_copies;
1005 	wp.wp_dnchecksum = db->db_dnode->dn_checksum;
1006 	wp.wp_oschecksum = os->os_checksum;
1007 	wp.wp_dncompress = db->db_dnode->dn_compress;
1008 	wp.wp_oscompress = os->os_compress;
1009 
1010 	ASSERT(BP_IS_HOLE(bp));
1011 
1012 	zio = arc_write(pio, os->os_spa, &wp, DBUF_IS_L2CACHEABLE(db),
1013 	    txg, bp, dr->dt.dl.dr_data, dmu_sync_ready, dmu_sync_done, in,
1014 	    ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
1015 	if (pio) {
1016 		zio_nowait(zio);
1017 		err = EINPROGRESS;
1018 	} else {
1019 		err = zio_wait(zio);
1020 		ASSERT(err == 0);
1021 	}
1022 	return (err);
1023 }
1024 
1025 int
1026 dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs,
1027 	dmu_tx_t *tx)
1028 {
1029 	dnode_t *dn;
1030 	int err;
1031 
1032 	err = dnode_hold(os->os, object, FTAG, &dn);
1033 	if (err)
1034 		return (err);
1035 	err = dnode_set_blksz(dn, size, ibs, tx);
1036 	dnode_rele(dn, FTAG);
1037 	return (err);
1038 }
1039 
1040 void
1041 dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
1042 	dmu_tx_t *tx)
1043 {
1044 	dnode_t *dn;
1045 
1046 	/* XXX assumes dnode_hold will not get an i/o error */
1047 	(void) dnode_hold(os->os, object, FTAG, &dn);
1048 	ASSERT(checksum < ZIO_CHECKSUM_FUNCTIONS);
1049 	dn->dn_checksum = checksum;
1050 	dnode_setdirty(dn, tx);
1051 	dnode_rele(dn, FTAG);
1052 }
1053 
1054 void
1055 dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
1056 	dmu_tx_t *tx)
1057 {
1058 	dnode_t *dn;
1059 
1060 	/* XXX assumes dnode_hold will not get an i/o error */
1061 	(void) dnode_hold(os->os, object, FTAG, &dn);
1062 	ASSERT(compress < ZIO_COMPRESS_FUNCTIONS);
1063 	dn->dn_compress = compress;
1064 	dnode_setdirty(dn, tx);
1065 	dnode_rele(dn, FTAG);
1066 }
1067 
1068 int
1069 dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off)
1070 {
1071 	dnode_t *dn;
1072 	int i, err;
1073 
1074 	err = dnode_hold(os->os, object, FTAG, &dn);
1075 	if (err)
1076 		return (err);
1077 	/*
1078 	 * Sync any current changes before
1079 	 * we go trundling through the block pointers.
1080 	 */
1081 	for (i = 0; i < TXG_SIZE; i++) {
1082 		if (list_link_active(&dn->dn_dirty_link[i]))
1083 			break;
1084 	}
1085 	if (i != TXG_SIZE) {
1086 		dnode_rele(dn, FTAG);
1087 		txg_wait_synced(dmu_objset_pool(os), 0);
1088 		err = dnode_hold(os->os, object, FTAG, &dn);
1089 		if (err)
1090 			return (err);
1091 	}
1092 
1093 	err = dnode_next_offset(dn, (hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0);
1094 	dnode_rele(dn, FTAG);
1095 
1096 	return (err);
1097 }
1098 
1099 void
1100 dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi)
1101 {
1102 	rw_enter(&dn->dn_struct_rwlock, RW_READER);
1103 	mutex_enter(&dn->dn_mtx);
1104 
1105 	doi->doi_data_block_size = dn->dn_datablksz;
1106 	doi->doi_metadata_block_size = dn->dn_indblkshift ?
1107 	    1ULL << dn->dn_indblkshift : 0;
1108 	doi->doi_indirection = dn->dn_nlevels;
1109 	doi->doi_checksum = dn->dn_checksum;
1110 	doi->doi_compress = dn->dn_compress;
1111 	doi->doi_physical_blks = (DN_USED_BYTES(dn->dn_phys) +
1112 	    SPA_MINBLOCKSIZE/2) >> SPA_MINBLOCKSHIFT;
1113 	doi->doi_max_block_offset = dn->dn_phys->dn_maxblkid;
1114 	doi->doi_type = dn->dn_type;
1115 	doi->doi_bonus_size = dn->dn_bonuslen;
1116 	doi->doi_bonus_type = dn->dn_bonustype;
1117 
1118 	mutex_exit(&dn->dn_mtx);
1119 	rw_exit(&dn->dn_struct_rwlock);
1120 }
1121 
1122 /*
1123  * Get information on a DMU object.
1124  * If doi is NULL, just indicates whether the object exists.
1125  */
1126 int
1127 dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi)
1128 {
1129 	dnode_t *dn;
1130 	int err = dnode_hold(os->os, object, FTAG, &dn);
1131 
1132 	if (err)
1133 		return (err);
1134 
1135 	if (doi != NULL)
1136 		dmu_object_info_from_dnode(dn, doi);
1137 
1138 	dnode_rele(dn, FTAG);
1139 	return (0);
1140 }
1141 
1142 /*
1143  * As above, but faster; can be used when you have a held dbuf in hand.
1144  */
1145 void
1146 dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi)
1147 {
1148 	dmu_object_info_from_dnode(((dmu_buf_impl_t *)db)->db_dnode, doi);
1149 }
1150 
1151 /*
1152  * Faster still when you only care about the size.
1153  * This is specifically optimized for zfs_getattr().
1154  */
1155 void
1156 dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, u_longlong_t *nblk512)
1157 {
1158 	dnode_t *dn = ((dmu_buf_impl_t *)db)->db_dnode;
1159 
1160 	*blksize = dn->dn_datablksz;
1161 	/* add 1 for dnode space */
1162 	*nblk512 = ((DN_USED_BYTES(dn->dn_phys) + SPA_MINBLOCKSIZE/2) >>
1163 	    SPA_MINBLOCKSHIFT) + 1;
1164 }
1165 
1166 void
1167 byteswap_uint64_array(void *vbuf, size_t size)
1168 {
1169 	uint64_t *buf = vbuf;
1170 	size_t count = size >> 3;
1171 	int i;
1172 
1173 	ASSERT((size & 7) == 0);
1174 
1175 	for (i = 0; i < count; i++)
1176 		buf[i] = BSWAP_64(buf[i]);
1177 }
1178 
1179 void
1180 byteswap_uint32_array(void *vbuf, size_t size)
1181 {
1182 	uint32_t *buf = vbuf;
1183 	size_t count = size >> 2;
1184 	int i;
1185 
1186 	ASSERT((size & 3) == 0);
1187 
1188 	for (i = 0; i < count; i++)
1189 		buf[i] = BSWAP_32(buf[i]);
1190 }
1191 
1192 void
1193 byteswap_uint16_array(void *vbuf, size_t size)
1194 {
1195 	uint16_t *buf = vbuf;
1196 	size_t count = size >> 1;
1197 	int i;
1198 
1199 	ASSERT((size & 1) == 0);
1200 
1201 	for (i = 0; i < count; i++)
1202 		buf[i] = BSWAP_16(buf[i]);
1203 }
1204 
1205 /* ARGSUSED */
1206 void
1207 byteswap_uint8_array(void *vbuf, size_t size)
1208 {
1209 }
1210 
1211 void
1212 dmu_init(void)
1213 {
1214 	dbuf_init();
1215 	dnode_init();
1216 	arc_init();
1217 	l2arc_init();
1218 }
1219 
1220 void
1221 dmu_fini(void)
1222 {
1223 	arc_fini();
1224 	dnode_fini();
1225 	dbuf_fini();
1226 	l2arc_fini();
1227 }
1228