xref: /illumos-gate/usr/src/uts/common/fs/zfs/sa.c (revision 449d1bcc)
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 /*
23  * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
24  * Portions Copyright 2011 iXsystems, Inc
25  * Copyright (c) 2013, 2017 by Delphix. All rights reserved.
26  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
27  * Copyright (c) 2014 Integros [integros.com]
28  * Copyright 2019 Joyent, Inc.
29  * Copyright 2023 RackTop Systems, Inc.
30  */
31 
32 #include <sys/zfs_context.h>
33 #include <sys/types.h>
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/sysmacros.h>
37 #include <sys/dmu.h>
38 #include <sys/dmu_impl.h>
39 #include <sys/dmu_objset.h>
40 #include <sys/dmu_tx.h>
41 #include <sys/dbuf.h>
42 #include <sys/dnode.h>
43 #include <sys/zap.h>
44 #include <sys/sa.h>
45 #include <sys/sunddi.h>
46 #include <sys/sa_impl.h>
47 #include <sys/dnode.h>
48 #include <sys/errno.h>
49 #include <sys/zfs_context.h>
50 
51 #ifdef _KERNEL
52 #include <sys/zfs_znode.h>
53 #endif
54 
55 /*
56  * ZFS System attributes:
57  *
58  * A generic mechanism to allow for arbitrary attributes
59  * to be stored in a dnode.  The data will be stored in the bonus buffer of
60  * the dnode and if necessary a special "spill" block will be used to handle
61  * overflow situations.  The spill block will be sized to fit the data
62  * from 512 - 128K.  When a spill block is used the BP (blkptr_t) for the
63  * spill block is stored at the end of the current bonus buffer.  Any
64  * attributes that would be in the way of the blkptr_t will be relocated
65  * into the spill block.
66  *
67  * Attribute registration:
68  *
69  * Stored persistently on a per dataset basis
70  * a mapping between attribute "string" names and their actual attribute
71  * numeric values, length, and byteswap function.  The names are only used
72  * during registration.  All  attributes are known by their unique attribute
73  * id value.  If an attribute can have a variable size then the value
74  * 0 will be used to indicate this.
75  *
76  * Attribute Layout:
77  *
78  * Attribute layouts are a way to compactly store multiple attributes, but
79  * without taking the overhead associated with managing each attribute
80  * individually.  Since you will typically have the same set of attributes
81  * stored in the same order a single table will be used to represent that
82  * layout.  The ZPL for example will usually have only about 10 different
83  * layouts (regular files, device files, symlinks,
84  * regular files + scanstamp, files/dir with extended attributes, and then
85  * you have the possibility of all of those minus ACL, because it would
86  * be kicked out into the spill block)
87  *
88  * Layouts are simply an array of the attributes and their
89  * ordering i.e. [0, 1, 4, 5, 2]
90  *
91  * Each distinct layout is given a unique layout number and that is whats
92  * stored in the header at the beginning of the SA data buffer.
93  *
94  * A layout only covers a single dbuf (bonus or spill).  If a set of
95  * attributes is split up between the bonus buffer and a spill buffer then
96  * two different layouts will be used.  This allows us to byteswap the
97  * spill without looking at the bonus buffer and keeps the on disk format of
98  * the bonus and spill buffer the same.
99  *
100  * Adding a single attribute will cause the entire set of attributes to
101  * be rewritten and could result in a new layout number being constructed
102  * as part of the rewrite if no such layout exists for the new set of
103  * attribues.  The new attribute will be appended to the end of the already
104  * existing attributes.
105  *
106  * Both the attribute registration and attribute layout information are
107  * stored in normal ZAP attributes.  Their should be a small number of
108  * known layouts and the set of attributes is assumed to typically be quite
109  * small.
110  *
111  * The registered attributes and layout "table" information is maintained
112  * in core and a special "sa_os_t" is attached to the objset_t.
113  *
114  * A special interface is provided to allow for quickly applying
115  * a large set of attributes at once.  sa_replace_all_by_template() is
116  * used to set an array of attributes.  This is used by the ZPL when
117  * creating a brand new file.  The template that is passed into the function
118  * specifies the attribute, size for variable length attributes, location of
119  * data and special "data locator" function if the data isn't in a contiguous
120  * location.
121  *
122  * Byteswap implications:
123  *
124  * Since the SA attributes are not entirely self describing we can't do
125  * the normal byteswap processing.  The special ZAP layout attribute and
126  * attribute registration attributes define the byteswap function and the
127  * size of the attributes, unless it is variable sized.
128  * The normal ZFS byteswapping infrastructure assumes you don't need
129  * to read any objects in order to do the necessary byteswapping.  Whereas
130  * SA attributes can only be properly byteswapped if the dataset is opened
131  * and the layout/attribute ZAP attributes are available.  Because of this
132  * the SA attributes will be byteswapped when they are first accessed by
133  * the SA code that will read the SA data.
134  */
135 
136 typedef void (sa_iterfunc_t)(void *hdr, void *addr, sa_attr_type_t,
137     uint16_t length, int length_idx, boolean_t, void *userp);
138 
139 static int sa_build_index(sa_handle_t *hdl, sa_buf_type_t buftype);
140 static void sa_idx_tab_hold(objset_t *os, sa_idx_tab_t *idx_tab);
141 static sa_idx_tab_t *sa_find_idx_tab(objset_t *os, dmu_object_type_t bonustype,
142     sa_hdr_phys_t *hdr);
143 static void sa_idx_tab_rele(objset_t *os, void *arg);
144 static void sa_copy_data(sa_data_locator_t *func, void *start, void *target,
145     int buflen);
146 static int sa_modify_attrs(sa_handle_t *hdl, sa_attr_type_t newattr,
147     sa_data_op_t action, sa_data_locator_t *locator, void *datastart,
148     uint16_t buflen, dmu_tx_t *tx);
149 
150 arc_byteswap_func_t *sa_bswap_table[] = {
151 	byteswap_uint64_array,
152 	byteswap_uint32_array,
153 	byteswap_uint16_array,
154 	byteswap_uint8_array,
155 	zfs_acl_byteswap,
156 };
157 
158 #define	SA_COPY_DATA(f, s, t, l) \
159 	{ \
160 		if (f == NULL) { \
161 			if (l == 8) { \
162 				*(uint64_t *)t = *(uint64_t *)s; \
163 			} else if (l == 16) { \
164 				*(uint64_t *)t = *(uint64_t *)s; \
165 				*(uint64_t *)((uintptr_t)t + 8) = \
166 				    *(uint64_t *)((uintptr_t)s + 8); \
167 			} else { \
168 				bcopy(s, t, l); \
169 			} \
170 		} else \
171 			sa_copy_data(f, s, t, l); \
172 	}
173 
174 /*
175  * This table is fixed and cannot be changed.  Its purpose is to
176  * allow the SA code to work with both old/new ZPL file systems.
177  * It contains the list of legacy attributes.  These attributes aren't
178  * stored in the "attribute" registry zap objects, since older ZPL file systems
179  * won't have the registry.  Only objsets of type ZFS_TYPE_FILESYSTEM will
180  * use this static table.
181  */
182 sa_attr_reg_t sa_legacy_attrs[] = {
183 	{"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0},
184 	{"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1},
185 	{"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2},
186 	{"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3},
187 	{"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4},
188 	{"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5},
189 	{"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6},
190 	{"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7},
191 	{"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8},
192 	{"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9},
193 	{"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10},
194 	{"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11},
195 	{"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12},
196 	{"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13},
197 	{"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14},
198 	{"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15},
199 };
200 
201 /*
202  * This is only used for objects of type DMU_OT_ZNODE
203  */
204 sa_attr_type_t sa_legacy_zpl_layout[] = {
205     0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
206 };
207 
208 /*
209  * Special dummy layout used for buffers with no attributes.
210  */
211 sa_attr_type_t sa_dummy_zpl_layout[] = { 0 };
212 
213 static int sa_legacy_attr_count = 16;
214 static kmem_cache_t *sa_cache = NULL;
215 
216 /*ARGSUSED*/
217 static int
sa_cache_constructor(void * buf,void * unused,int kmflag)218 sa_cache_constructor(void *buf, void *unused, int kmflag)
219 {
220 	sa_handle_t *hdl = buf;
221 
222 	mutex_init(&hdl->sa_lock, NULL, MUTEX_DEFAULT, NULL);
223 	return (0);
224 }
225 
226 /*ARGSUSED*/
227 static void
sa_cache_destructor(void * buf,void * unused)228 sa_cache_destructor(void *buf, void *unused)
229 {
230 	sa_handle_t *hdl = buf;
231 	mutex_destroy(&hdl->sa_lock);
232 }
233 
234 void
sa_cache_init(void)235 sa_cache_init(void)
236 {
237 	sa_cache = kmem_cache_create("sa_cache",
238 	    sizeof (sa_handle_t), 0, sa_cache_constructor,
239 	    sa_cache_destructor, NULL, NULL, NULL, 0);
240 }
241 
242 void
sa_cache_fini(void)243 sa_cache_fini(void)
244 {
245 	if (sa_cache)
246 		kmem_cache_destroy(sa_cache);
247 }
248 
249 static int
layout_num_compare(const void * arg1,const void * arg2)250 layout_num_compare(const void *arg1, const void *arg2)
251 {
252 	const sa_lot_t *node1 = (const sa_lot_t *)arg1;
253 	const sa_lot_t *node2 = (const sa_lot_t *)arg2;
254 
255 	return (TREE_CMP(node1->lot_num, node2->lot_num));
256 }
257 
258 static int
layout_hash_compare(const void * arg1,const void * arg2)259 layout_hash_compare(const void *arg1, const void *arg2)
260 {
261 	const sa_lot_t *node1 = (const sa_lot_t *)arg1;
262 	const sa_lot_t *node2 = (const sa_lot_t *)arg2;
263 
264 	int cmp = TREE_CMP(node1->lot_hash, node2->lot_hash);
265 	if (likely(cmp))
266 		return (cmp);
267 
268 	return (TREE_CMP(node1->lot_instance, node2->lot_instance));
269 }
270 
271 boolean_t
sa_layout_equal(sa_lot_t * tbf,sa_attr_type_t * attrs,int count)272 sa_layout_equal(sa_lot_t *tbf, sa_attr_type_t *attrs, int count)
273 {
274 	int i;
275 
276 	if (count != tbf->lot_attr_count)
277 		return (1);
278 
279 	for (i = 0; i != count; i++) {
280 		if (attrs[i] != tbf->lot_attrs[i])
281 			return (1);
282 	}
283 	return (0);
284 }
285 
286 #define	SA_ATTR_HASH(attr) (zfs_crc64_table[(-1ULL ^ attr) & 0xFF])
287 
288 static uint64_t
sa_layout_info_hash(sa_attr_type_t * attrs,int attr_count)289 sa_layout_info_hash(sa_attr_type_t *attrs, int attr_count)
290 {
291 	int i;
292 	uint64_t crc = -1ULL;
293 
294 	for (i = 0; i != attr_count; i++)
295 		crc ^= SA_ATTR_HASH(attrs[i]);
296 
297 	return (crc);
298 }
299 
300 static int
sa_get_spill(sa_handle_t * hdl)301 sa_get_spill(sa_handle_t *hdl)
302 {
303 	int rc;
304 	if (hdl->sa_spill == NULL) {
305 		if ((rc = dmu_spill_hold_existing(hdl->sa_bonus, NULL,
306 		    &hdl->sa_spill)) == 0)
307 			VERIFY(0 == sa_build_index(hdl, SA_SPILL));
308 	} else {
309 		rc = 0;
310 	}
311 
312 	return (rc);
313 }
314 
315 /*
316  * Main attribute lookup/update function
317  * returns 0 for success or non zero for failures
318  *
319  * Operates on bulk array, first failure will abort further processing
320  */
321 int
sa_attr_op(sa_handle_t * hdl,sa_bulk_attr_t * bulk,int count,sa_data_op_t data_op,dmu_tx_t * tx)322 sa_attr_op(sa_handle_t *hdl, sa_bulk_attr_t *bulk, int count,
323     sa_data_op_t data_op, dmu_tx_t *tx)
324 {
325 	sa_os_t *sa = hdl->sa_os->os_sa;
326 	int i;
327 	int error = 0;
328 	sa_buf_type_t buftypes;
329 
330 	buftypes = 0;
331 
332 	ASSERT(count > 0);
333 	for (i = 0; i != count; i++) {
334 		ASSERT(bulk[i].sa_attr <= hdl->sa_os->os_sa->sa_num_attrs);
335 
336 		bulk[i].sa_addr = NULL;
337 		/* First check the bonus buffer */
338 
339 		if (hdl->sa_bonus_tab && TOC_ATTR_PRESENT(
340 		    hdl->sa_bonus_tab->sa_idx_tab[bulk[i].sa_attr])) {
341 			SA_ATTR_INFO(sa, hdl->sa_bonus_tab,
342 			    SA_GET_HDR(hdl, SA_BONUS),
343 			    bulk[i].sa_attr, bulk[i], SA_BONUS, hdl);
344 			if (tx && !(buftypes & SA_BONUS)) {
345 				dmu_buf_will_dirty(hdl->sa_bonus, tx);
346 				buftypes |= SA_BONUS;
347 			}
348 		}
349 		if (bulk[i].sa_addr == NULL &&
350 		    ((error = sa_get_spill(hdl)) == 0)) {
351 			if (TOC_ATTR_PRESENT(
352 			    hdl->sa_spill_tab->sa_idx_tab[bulk[i].sa_attr])) {
353 				SA_ATTR_INFO(sa, hdl->sa_spill_tab,
354 				    SA_GET_HDR(hdl, SA_SPILL),
355 				    bulk[i].sa_attr, bulk[i], SA_SPILL, hdl);
356 				if (tx && !(buftypes & SA_SPILL) &&
357 				    bulk[i].sa_size == bulk[i].sa_length) {
358 					dmu_buf_will_dirty(hdl->sa_spill, tx);
359 					buftypes |= SA_SPILL;
360 				}
361 			}
362 		}
363 		if (error && error != ENOENT) {
364 			return ((error == ECKSUM) ? EIO : error);
365 		}
366 
367 		switch (data_op) {
368 		case SA_LOOKUP:
369 			if (bulk[i].sa_addr == NULL)
370 				return (SET_ERROR(ENOENT));
371 			if (bulk[i].sa_data) {
372 				SA_COPY_DATA(bulk[i].sa_data_func,
373 				    bulk[i].sa_addr, bulk[i].sa_data,
374 				    MIN(bulk[i].sa_size, bulk[i].sa_length));
375 			}
376 			continue;
377 
378 		case SA_UPDATE:
379 			/* existing rewrite of attr */
380 			if (bulk[i].sa_addr &&
381 			    bulk[i].sa_size == bulk[i].sa_length) {
382 				SA_COPY_DATA(bulk[i].sa_data_func,
383 				    bulk[i].sa_data, bulk[i].sa_addr,
384 				    bulk[i].sa_length);
385 				continue;
386 			} else if (bulk[i].sa_addr) { /* attr size change */
387 				error = sa_modify_attrs(hdl, bulk[i].sa_attr,
388 				    SA_REPLACE, bulk[i].sa_data_func,
389 				    bulk[i].sa_data, bulk[i].sa_length, tx);
390 			} else { /* adding new attribute */
391 				error = sa_modify_attrs(hdl, bulk[i].sa_attr,
392 				    SA_ADD, bulk[i].sa_data_func,
393 				    bulk[i].sa_data, bulk[i].sa_length, tx);
394 			}
395 			if (error)
396 				return (error);
397 			break;
398 		}
399 	}
400 	return (error);
401 }
402 
403 static sa_lot_t *
sa_add_layout_entry(objset_t * os,sa_attr_type_t * attrs,int attr_count,uint64_t lot_num,uint64_t hash,boolean_t zapadd,dmu_tx_t * tx)404 sa_add_layout_entry(objset_t *os, sa_attr_type_t *attrs, int attr_count,
405     uint64_t lot_num, uint64_t hash, boolean_t zapadd, dmu_tx_t *tx)
406 {
407 	sa_os_t *sa = os->os_sa;
408 	sa_lot_t *tb, *findtb;
409 	int i, size;
410 	avl_index_t loc;
411 
412 	ASSERT(MUTEX_HELD(&sa->sa_lock));
413 	tb = kmem_zalloc(sizeof (sa_lot_t), KM_SLEEP);
414 	tb->lot_attr_count = attr_count;
415 
416 	if ((size = sizeof (sa_attr_type_t) * attr_count) != 0) {
417 		tb->lot_attrs = kmem_alloc(size, KM_SLEEP);
418 		bcopy(attrs, tb->lot_attrs, size);
419 	}
420 
421 	tb->lot_num = lot_num;
422 	tb->lot_hash = hash;
423 	tb->lot_instance = 0;
424 
425 	if (zapadd) {
426 		char attr_name[8];
427 
428 		if (sa->sa_layout_attr_obj == 0) {
429 			sa->sa_layout_attr_obj = zap_create_link(os,
430 			    DMU_OT_SA_ATTR_LAYOUTS,
431 			    sa->sa_master_obj, SA_LAYOUTS, tx);
432 		}
433 
434 		(void) snprintf(attr_name, sizeof (attr_name),
435 		    "%d", (int)lot_num);
436 		VERIFY(0 == zap_update(os, os->os_sa->sa_layout_attr_obj,
437 		    attr_name, 2, attr_count, attrs, tx));
438 	}
439 
440 	list_create(&tb->lot_idx_tab, sizeof (sa_idx_tab_t),
441 	    offsetof(sa_idx_tab_t, sa_next));
442 
443 	for (i = 0; i != attr_count; i++) {
444 		if (sa->sa_attr_table[tb->lot_attrs[i]].sa_length == 0)
445 			tb->lot_var_sizes++;
446 	}
447 
448 	avl_add(&sa->sa_layout_num_tree, tb);
449 
450 	/* verify we don't have a hash collision */
451 	if ((findtb = avl_find(&sa->sa_layout_hash_tree, tb, &loc)) != NULL) {
452 		for (; findtb && findtb->lot_hash == hash;
453 		    findtb = AVL_NEXT(&sa->sa_layout_hash_tree, findtb)) {
454 			if (findtb->lot_instance != tb->lot_instance)
455 				break;
456 			tb->lot_instance++;
457 		}
458 	}
459 	avl_add(&sa->sa_layout_hash_tree, tb);
460 	return (tb);
461 }
462 
463 static void
sa_find_layout(objset_t * os,uint64_t hash,sa_attr_type_t * attrs,int count,dmu_tx_t * tx,sa_lot_t ** lot)464 sa_find_layout(objset_t *os, uint64_t hash, sa_attr_type_t *attrs,
465     int count, dmu_tx_t *tx, sa_lot_t **lot)
466 {
467 	sa_lot_t *tb, tbsearch;
468 	avl_index_t loc;
469 	sa_os_t *sa = os->os_sa;
470 	boolean_t found = B_FALSE;
471 
472 	mutex_enter(&sa->sa_lock);
473 	tbsearch.lot_hash = hash;
474 	tbsearch.lot_instance = 0;
475 	tb = avl_find(&sa->sa_layout_hash_tree, &tbsearch, &loc);
476 	if (tb) {
477 		for (; tb && tb->lot_hash == hash;
478 		    tb = AVL_NEXT(&sa->sa_layout_hash_tree, tb)) {
479 			if (sa_layout_equal(tb, attrs, count) == 0) {
480 				found = B_TRUE;
481 				break;
482 			}
483 		}
484 	}
485 	if (!found) {
486 		tb = sa_add_layout_entry(os, attrs, count,
487 		    avl_numnodes(&sa->sa_layout_num_tree), hash, B_TRUE, tx);
488 	}
489 	mutex_exit(&sa->sa_lock);
490 	*lot = tb;
491 }
492 
493 static int
sa_resize_spill(sa_handle_t * hdl,uint32_t size,dmu_tx_t * tx)494 sa_resize_spill(sa_handle_t *hdl, uint32_t size, dmu_tx_t *tx)
495 {
496 	int error;
497 	uint32_t blocksize;
498 
499 	if (size == 0) {
500 		blocksize = SPA_MINBLOCKSIZE;
501 	} else if (size > SPA_OLD_MAXBLOCKSIZE) {
502 		ASSERT(0);
503 		return (SET_ERROR(EFBIG));
504 	} else {
505 		blocksize = P2ROUNDUP_TYPED(size, SPA_MINBLOCKSIZE, uint32_t);
506 	}
507 
508 	error = dbuf_spill_set_blksz(hdl->sa_spill, blocksize, tx);
509 	ASSERT(error == 0);
510 	return (error);
511 }
512 
513 static void
sa_copy_data(sa_data_locator_t * func,void * datastart,void * target,int buflen)514 sa_copy_data(sa_data_locator_t *func, void *datastart, void *target, int buflen)
515 {
516 	if (func == NULL) {
517 		bcopy(datastart, target, buflen);
518 	} else {
519 		boolean_t start;
520 		int bytes;
521 		void *dataptr;
522 		void *saptr = target;
523 		uint32_t length;
524 
525 		start = B_TRUE;
526 		bytes = 0;
527 		while (bytes < buflen) {
528 			func(&dataptr, &length, buflen, start, datastart);
529 			bcopy(dataptr, saptr, length);
530 			saptr = (void *)((caddr_t)saptr + length);
531 			bytes += length;
532 			start = B_FALSE;
533 		}
534 	}
535 }
536 
537 /*
538  * Determine several different sizes
539  * first the sa header size
540  * the number of bytes to be stored
541  * if spill would occur the index in the attribute array is returned
542  *
543  * the boolean will_spill will be set when spilling is necessary.  It
544  * is only set when the buftype is SA_BONUS
545  */
546 static int
sa_find_sizes(sa_os_t * sa,sa_bulk_attr_t * attr_desc,int attr_count,dmu_buf_t * db,sa_buf_type_t buftype,int full_space,int * index,int * total,boolean_t * will_spill)547 sa_find_sizes(sa_os_t *sa, sa_bulk_attr_t *attr_desc, int attr_count,
548     dmu_buf_t *db, sa_buf_type_t buftype, int full_space, int *index,
549     int *total, boolean_t *will_spill)
550 {
551 	int var_size = 0;
552 	int i;
553 	int hdrsize;
554 	int extra_hdrsize;
555 
556 	if (buftype == SA_BONUS && sa->sa_force_spill) {
557 		*total = 0;
558 		*index = 0;
559 		*will_spill = B_TRUE;
560 		return (0);
561 	}
562 
563 	*index = -1;
564 	*total = 0;
565 	*will_spill = B_FALSE;
566 
567 	extra_hdrsize = 0;
568 	hdrsize = (SA_BONUSTYPE_FROM_DB(db) == DMU_OT_ZNODE) ? 0 :
569 	    sizeof (sa_hdr_phys_t);
570 
571 	ASSERT(IS_P2ALIGNED(full_space, 8));
572 
573 	for (i = 0; i != attr_count; i++) {
574 		boolean_t is_var_sz;
575 
576 		*total = P2ROUNDUP(*total, 8);
577 		*total += attr_desc[i].sa_length;
578 		if (*will_spill)
579 			continue;
580 
581 		is_var_sz = (SA_REGISTERED_LEN(sa, attr_desc[i].sa_attr) == 0);
582 		if (is_var_sz) {
583 			var_size++;
584 		}
585 
586 		if (is_var_sz && var_size > 1) {
587 			/*
588 			 * Don't worry that the spill block might overflow.
589 			 * It will be resized if needed in sa_build_layouts().
590 			 */
591 			if (buftype == SA_SPILL ||
592 			    P2ROUNDUP(hdrsize + sizeof (uint16_t), 8) +
593 			    *total < full_space) {
594 				/*
595 				 * Account for header space used by array of
596 				 * optional sizes of variable-length attributes.
597 				 * Record the extra header size in case this
598 				 * increase needs to be reversed due to
599 				 * spill-over.
600 				 */
601 				hdrsize += sizeof (uint16_t);
602 				if (*index != -1)
603 					extra_hdrsize += sizeof (uint16_t);
604 			} else {
605 				ASSERT(buftype == SA_BONUS);
606 				if (*index == -1)
607 					*index = i;
608 				*will_spill = B_TRUE;
609 				continue;
610 			}
611 		}
612 
613 		/*
614 		 * find index of where spill *could* occur.
615 		 * Then continue to count of remainder attribute
616 		 * space.  The sum is used later for sizing bonus
617 		 * and spill buffer.
618 		 */
619 		if (buftype == SA_BONUS && *index == -1 &&
620 		    *total + P2ROUNDUP(hdrsize, 8) >
621 		    (full_space - sizeof (blkptr_t))) {
622 			*index = i;
623 		}
624 
625 		if (*total + P2ROUNDUP(hdrsize, 8) > full_space &&
626 		    buftype == SA_BONUS)
627 			*will_spill = B_TRUE;
628 	}
629 
630 	if (*will_spill)
631 		hdrsize -= extra_hdrsize;
632 
633 	hdrsize = P2ROUNDUP(hdrsize, 8);
634 	return (hdrsize);
635 }
636 
637 #define	BUF_SPACE_NEEDED(total, header) (total + header)
638 
639 /*
640  * Find layout that corresponds to ordering of attributes
641  * If not found a new layout number is created and added to
642  * persistent layout tables.
643  */
644 static int
sa_build_layouts(sa_handle_t * hdl,sa_bulk_attr_t * attr_desc,int attr_count,dmu_tx_t * tx)645 sa_build_layouts(sa_handle_t *hdl, sa_bulk_attr_t *attr_desc, int attr_count,
646     dmu_tx_t *tx)
647 {
648 	sa_os_t *sa = hdl->sa_os->os_sa;
649 	uint64_t hash;
650 	sa_buf_type_t buftype;
651 	sa_hdr_phys_t *sahdr;
652 	void *data_start;
653 	int buf_space;
654 	sa_attr_type_t *attrs, *attrs_start;
655 	int i, lot_count;
656 	int dnodesize;
657 	int hdrsize;
658 	int spillhdrsize = 0;
659 	int used;
660 	dmu_object_type_t bonustype;
661 	sa_lot_t *lot;
662 	int len_idx;
663 	int spill_used;
664 	int bonuslen;
665 	boolean_t spilling;
666 
667 	dmu_buf_will_dirty(hdl->sa_bonus, tx);
668 	bonustype = SA_BONUSTYPE_FROM_DB(hdl->sa_bonus);
669 
670 	dmu_object_dnsize_from_db(hdl->sa_bonus, &dnodesize);
671 	bonuslen = DN_BONUS_SIZE(dnodesize);
672 
673 	/* first determine bonus header size and sum of all attributes */
674 	hdrsize = sa_find_sizes(sa, attr_desc, attr_count, hdl->sa_bonus,
675 	    SA_BONUS, bonuslen, &i, &used, &spilling);
676 
677 	if (used > SPA_OLD_MAXBLOCKSIZE)
678 		return (SET_ERROR(EFBIG));
679 
680 	VERIFY(0 == dmu_set_bonus(hdl->sa_bonus, spilling ?
681 	    MIN(bonuslen - sizeof (blkptr_t), used + hdrsize) :
682 	    used + hdrsize, tx));
683 
684 	ASSERT((bonustype == DMU_OT_ZNODE && spilling == 0) ||
685 	    bonustype == DMU_OT_SA);
686 
687 	/* setup and size spill buffer when needed */
688 	if (spilling) {
689 		boolean_t dummy;
690 
691 		if (hdl->sa_spill == NULL) {
692 			VERIFY(dmu_spill_hold_by_bonus(hdl->sa_bonus, 0, NULL,
693 			    &hdl->sa_spill) == 0);
694 		}
695 		dmu_buf_will_dirty(hdl->sa_spill, tx);
696 
697 		spillhdrsize = sa_find_sizes(sa, &attr_desc[i],
698 		    attr_count - i, hdl->sa_spill, SA_SPILL,
699 		    hdl->sa_spill->db_size, &i, &spill_used, &dummy);
700 
701 		if (spill_used > SPA_OLD_MAXBLOCKSIZE)
702 			return (SET_ERROR(EFBIG));
703 
704 		buf_space = hdl->sa_spill->db_size - spillhdrsize;
705 		if (BUF_SPACE_NEEDED(spill_used, spillhdrsize) >
706 		    hdl->sa_spill->db_size)
707 			VERIFY(0 == sa_resize_spill(hdl,
708 			    BUF_SPACE_NEEDED(spill_used, spillhdrsize), tx));
709 	}
710 
711 	/* setup starting pointers to lay down data */
712 	data_start = (void *)((uintptr_t)hdl->sa_bonus->db_data + hdrsize);
713 	sahdr = (sa_hdr_phys_t *)hdl->sa_bonus->db_data;
714 	buftype = SA_BONUS;
715 
716 	if (spilling)
717 		buf_space = (sa->sa_force_spill) ?
718 		    0 : SA_BLKPTR_SPACE - hdrsize;
719 	else
720 		buf_space = hdl->sa_bonus->db_size - hdrsize;
721 
722 	attrs_start = attrs = kmem_alloc(sizeof (sa_attr_type_t) * attr_count,
723 	    KM_SLEEP);
724 	lot_count = 0;
725 
726 	for (i = 0, len_idx = 0, hash = -1ULL; i != attr_count; i++) {
727 		uint16_t length;
728 
729 		ASSERT(IS_P2ALIGNED(data_start, 8));
730 		ASSERT(IS_P2ALIGNED(buf_space, 8));
731 		attrs[i] = attr_desc[i].sa_attr;
732 		length = SA_REGISTERED_LEN(sa, attrs[i]);
733 		if (length == 0)
734 			length = attr_desc[i].sa_length;
735 
736 		if (buf_space < length) {  /* switch to spill buffer */
737 			VERIFY(spilling);
738 			VERIFY(bonustype == DMU_OT_SA);
739 			if (buftype == SA_BONUS && !sa->sa_force_spill) {
740 				sa_find_layout(hdl->sa_os, hash, attrs_start,
741 				    lot_count, tx, &lot);
742 				SA_SET_HDR(sahdr, lot->lot_num, hdrsize);
743 			}
744 
745 			buftype = SA_SPILL;
746 			hash = -1ULL;
747 			len_idx = 0;
748 
749 			sahdr = (sa_hdr_phys_t *)hdl->sa_spill->db_data;
750 			sahdr->sa_magic = SA_MAGIC;
751 			data_start = (void *)((uintptr_t)sahdr +
752 			    spillhdrsize);
753 			attrs_start = &attrs[i];
754 			buf_space = hdl->sa_spill->db_size - spillhdrsize;
755 			lot_count = 0;
756 		}
757 		hash ^= SA_ATTR_HASH(attrs[i]);
758 		attr_desc[i].sa_addr = data_start;
759 		attr_desc[i].sa_size = length;
760 		SA_COPY_DATA(attr_desc[i].sa_data_func, attr_desc[i].sa_data,
761 		    data_start, length);
762 		if (sa->sa_attr_table[attrs[i]].sa_length == 0) {
763 			sahdr->sa_lengths[len_idx++] = length;
764 		}
765 		data_start = (void *)P2ROUNDUP(((uintptr_t)data_start +
766 		    length), 8);
767 		buf_space -= P2ROUNDUP(length, 8);
768 		lot_count++;
769 	}
770 
771 	sa_find_layout(hdl->sa_os, hash, attrs_start, lot_count, tx, &lot);
772 
773 	/*
774 	 * Verify that old znodes always have layout number 0.
775 	 * Must be DMU_OT_SA for arbitrary layouts
776 	 */
777 	VERIFY((bonustype == DMU_OT_ZNODE && lot->lot_num == 0) ||
778 	    (bonustype == DMU_OT_SA && lot->lot_num > 1));
779 
780 	if (bonustype == DMU_OT_SA) {
781 		SA_SET_HDR(sahdr, lot->lot_num,
782 		    buftype == SA_BONUS ? hdrsize : spillhdrsize);
783 	}
784 
785 	kmem_free(attrs, sizeof (sa_attr_type_t) * attr_count);
786 	if (hdl->sa_bonus_tab) {
787 		sa_idx_tab_rele(hdl->sa_os, hdl->sa_bonus_tab);
788 		hdl->sa_bonus_tab = NULL;
789 	}
790 	if (!sa->sa_force_spill)
791 		VERIFY(0 == sa_build_index(hdl, SA_BONUS));
792 	if (hdl->sa_spill) {
793 		sa_idx_tab_rele(hdl->sa_os, hdl->sa_spill_tab);
794 		if (!spilling) {
795 			/*
796 			 * remove spill block that is no longer needed.
797 			 */
798 			dmu_buf_rele(hdl->sa_spill, NULL);
799 			hdl->sa_spill = NULL;
800 			hdl->sa_spill_tab = NULL;
801 			VERIFY(0 == dmu_rm_spill(hdl->sa_os,
802 			    sa_handle_object(hdl), tx));
803 		} else {
804 			VERIFY(0 == sa_build_index(hdl, SA_SPILL));
805 		}
806 	}
807 
808 	return (0);
809 }
810 
811 static void
sa_free_attr_table(sa_os_t * sa)812 sa_free_attr_table(sa_os_t *sa)
813 {
814 	int i;
815 
816 	if (sa->sa_attr_table == NULL)
817 		return;
818 
819 	for (i = 0; i != sa->sa_num_attrs; i++) {
820 		if (sa->sa_attr_table[i].sa_name)
821 			kmem_free(sa->sa_attr_table[i].sa_name,
822 			    strlen(sa->sa_attr_table[i].sa_name) + 1);
823 	}
824 
825 	kmem_free(sa->sa_attr_table,
826 	    sizeof (sa_attr_table_t) * sa->sa_num_attrs);
827 
828 	sa->sa_attr_table = NULL;
829 }
830 
831 static int
sa_attr_table_setup(objset_t * os,sa_attr_reg_t * reg_attrs,int count)832 sa_attr_table_setup(objset_t *os, sa_attr_reg_t *reg_attrs, int count)
833 {
834 	sa_os_t *sa = os->os_sa;
835 	uint64_t sa_attr_count = 0;
836 	uint64_t sa_reg_count = 0;
837 	int error = 0;
838 	uint64_t attr_value;
839 	sa_attr_table_t *tb;
840 	zap_cursor_t zc;
841 	zap_attribute_t za;
842 	int registered_count = 0;
843 	int i;
844 	dmu_objset_type_t ostype = dmu_objset_type(os);
845 
846 	sa->sa_user_table =
847 	    kmem_zalloc(count * sizeof (sa_attr_type_t), KM_SLEEP);
848 	sa->sa_user_table_sz = count * sizeof (sa_attr_type_t);
849 
850 	if (sa->sa_reg_attr_obj != 0) {
851 		error = zap_count(os, sa->sa_reg_attr_obj,
852 		    &sa_attr_count);
853 
854 		/*
855 		 * Make sure we retrieved a count and that it isn't zero
856 		 */
857 		if (error || (error == 0 && sa_attr_count == 0)) {
858 			if (error == 0)
859 				error = SET_ERROR(EINVAL);
860 			goto bail;
861 		}
862 		sa_reg_count = sa_attr_count;
863 	}
864 
865 	if (ostype == DMU_OST_ZFS && sa_attr_count == 0)
866 		sa_attr_count += sa_legacy_attr_count;
867 
868 	/* Allocate attribute numbers for attributes that aren't registered */
869 	for (i = 0; i != count; i++) {
870 		boolean_t found = B_FALSE;
871 		int j;
872 
873 		if (ostype == DMU_OST_ZFS) {
874 			for (j = 0; j != sa_legacy_attr_count; j++) {
875 				if (strcmp(reg_attrs[i].sa_name,
876 				    sa_legacy_attrs[j].sa_name) == 0) {
877 					sa->sa_user_table[i] =
878 					    sa_legacy_attrs[j].sa_attr;
879 					found = B_TRUE;
880 				}
881 			}
882 		}
883 		if (found)
884 			continue;
885 
886 		if (sa->sa_reg_attr_obj)
887 			error = zap_lookup(os, sa->sa_reg_attr_obj,
888 			    reg_attrs[i].sa_name, 8, 1, &attr_value);
889 		else
890 			error = SET_ERROR(ENOENT);
891 		switch (error) {
892 		case ENOENT:
893 			sa->sa_user_table[i] = (sa_attr_type_t)sa_attr_count;
894 			sa_attr_count++;
895 			break;
896 		case 0:
897 			sa->sa_user_table[i] = ATTR_NUM(attr_value);
898 			break;
899 		default:
900 			goto bail;
901 		}
902 	}
903 
904 	sa->sa_num_attrs = sa_attr_count;
905 	tb = sa->sa_attr_table =
906 	    kmem_zalloc(sizeof (sa_attr_table_t) * sa_attr_count, KM_SLEEP);
907 
908 	/*
909 	 * Attribute table is constructed from requested attribute list,
910 	 * previously foreign registered attributes, and also the legacy
911 	 * ZPL set of attributes.
912 	 */
913 
914 	if (sa->sa_reg_attr_obj) {
915 		for (zap_cursor_init(&zc, os, sa->sa_reg_attr_obj);
916 		    (error = zap_cursor_retrieve(&zc, &za)) == 0;
917 		    zap_cursor_advance(&zc)) {
918 			uint64_t value;
919 			value  = za.za_first_integer;
920 
921 			registered_count++;
922 			tb[ATTR_NUM(value)].sa_attr = ATTR_NUM(value);
923 			tb[ATTR_NUM(value)].sa_length = ATTR_LENGTH(value);
924 			tb[ATTR_NUM(value)].sa_byteswap = ATTR_BSWAP(value);
925 			tb[ATTR_NUM(value)].sa_registered = B_TRUE;
926 
927 			if (tb[ATTR_NUM(value)].sa_name) {
928 				continue;
929 			}
930 			tb[ATTR_NUM(value)].sa_name =
931 			    kmem_zalloc(strlen(za.za_name) +1, KM_SLEEP);
932 			(void) strlcpy(tb[ATTR_NUM(value)].sa_name, za.za_name,
933 			    strlen(za.za_name) +1);
934 		}
935 		zap_cursor_fini(&zc);
936 		/*
937 		 * Make sure we processed the correct number of registered
938 		 * attributes
939 		 */
940 		if (registered_count != sa_reg_count) {
941 			ASSERT(error != 0);
942 			goto bail;
943 		}
944 
945 	}
946 
947 	if (ostype == DMU_OST_ZFS) {
948 		for (i = 0; i != sa_legacy_attr_count; i++) {
949 			if (tb[i].sa_name)
950 				continue;
951 			tb[i].sa_attr = sa_legacy_attrs[i].sa_attr;
952 			tb[i].sa_length = sa_legacy_attrs[i].sa_length;
953 			tb[i].sa_byteswap = sa_legacy_attrs[i].sa_byteswap;
954 			tb[i].sa_registered = B_FALSE;
955 			tb[i].sa_name =
956 			    kmem_zalloc(strlen(sa_legacy_attrs[i].sa_name) +1,
957 			    KM_SLEEP);
958 			(void) strlcpy(tb[i].sa_name,
959 			    sa_legacy_attrs[i].sa_name,
960 			    strlen(sa_legacy_attrs[i].sa_name) + 1);
961 		}
962 	}
963 
964 	for (i = 0; i != count; i++) {
965 		sa_attr_type_t attr_id;
966 
967 		attr_id = sa->sa_user_table[i];
968 		if (tb[attr_id].sa_name)
969 			continue;
970 
971 		tb[attr_id].sa_length = reg_attrs[i].sa_length;
972 		tb[attr_id].sa_byteswap = reg_attrs[i].sa_byteswap;
973 		tb[attr_id].sa_attr = attr_id;
974 		tb[attr_id].sa_name =
975 		    kmem_zalloc(strlen(reg_attrs[i].sa_name) + 1, KM_SLEEP);
976 		(void) strlcpy(tb[attr_id].sa_name, reg_attrs[i].sa_name,
977 		    strlen(reg_attrs[i].sa_name) + 1);
978 	}
979 
980 	sa->sa_need_attr_registration =
981 	    (sa_attr_count != registered_count);
982 
983 	return (0);
984 bail:
985 	kmem_free(sa->sa_user_table, count * sizeof (sa_attr_type_t));
986 	sa->sa_user_table = NULL;
987 	sa_free_attr_table(sa);
988 	return ((error != 0) ? error : EINVAL);
989 }
990 
991 int
sa_setup(objset_t * os,uint64_t sa_obj,sa_attr_reg_t * reg_attrs,int count,sa_attr_type_t ** user_table)992 sa_setup(objset_t *os, uint64_t sa_obj, sa_attr_reg_t *reg_attrs, int count,
993     sa_attr_type_t **user_table)
994 {
995 	zap_cursor_t zc;
996 	zap_attribute_t za;
997 	sa_os_t *sa;
998 	dmu_objset_type_t ostype = dmu_objset_type(os);
999 	sa_attr_type_t *tb;
1000 	int error;
1001 
1002 	mutex_enter(&os->os_user_ptr_lock);
1003 	if (os->os_sa) {
1004 		mutex_enter(&os->os_sa->sa_lock);
1005 		mutex_exit(&os->os_user_ptr_lock);
1006 		tb = os->os_sa->sa_user_table;
1007 		mutex_exit(&os->os_sa->sa_lock);
1008 		*user_table = tb;
1009 		return (0);
1010 	}
1011 
1012 	sa = kmem_zalloc(sizeof (sa_os_t), KM_SLEEP);
1013 	mutex_init(&sa->sa_lock, NULL, MUTEX_DEFAULT, NULL);
1014 	sa->sa_master_obj = sa_obj;
1015 
1016 	os->os_sa = sa;
1017 	mutex_enter(&sa->sa_lock);
1018 	mutex_exit(&os->os_user_ptr_lock);
1019 	avl_create(&sa->sa_layout_num_tree, layout_num_compare,
1020 	    sizeof (sa_lot_t), offsetof(sa_lot_t, lot_num_node));
1021 	avl_create(&sa->sa_layout_hash_tree, layout_hash_compare,
1022 	    sizeof (sa_lot_t), offsetof(sa_lot_t, lot_hash_node));
1023 
1024 	if (sa_obj) {
1025 		error = zap_lookup(os, sa_obj, SA_LAYOUTS,
1026 		    8, 1, &sa->sa_layout_attr_obj);
1027 		if (error != 0 && error != ENOENT)
1028 			goto fail;
1029 		error = zap_lookup(os, sa_obj, SA_REGISTRY,
1030 		    8, 1, &sa->sa_reg_attr_obj);
1031 		if (error != 0 && error != ENOENT)
1032 			goto fail;
1033 	}
1034 
1035 	if ((error = sa_attr_table_setup(os, reg_attrs, count)) != 0)
1036 		goto fail;
1037 
1038 	if (sa->sa_layout_attr_obj != 0) {
1039 		uint64_t layout_count;
1040 
1041 		error = zap_count(os, sa->sa_layout_attr_obj,
1042 		    &layout_count);
1043 
1044 		/*
1045 		 * Layout number count should be > 0
1046 		 */
1047 		if (error || (error == 0 && layout_count == 0)) {
1048 			if (error == 0)
1049 				error = SET_ERROR(EINVAL);
1050 			goto fail;
1051 		}
1052 
1053 		for (zap_cursor_init(&zc, os, sa->sa_layout_attr_obj);
1054 		    (error = zap_cursor_retrieve(&zc, &za)) == 0;
1055 		    zap_cursor_advance(&zc)) {
1056 			sa_attr_type_t *lot_attrs;
1057 			uint64_t lot_num;
1058 
1059 			lot_attrs = kmem_zalloc(sizeof (sa_attr_type_t) *
1060 			    za.za_num_integers, KM_SLEEP);
1061 
1062 			if ((error = (zap_lookup(os, sa->sa_layout_attr_obj,
1063 			    za.za_name, 2, za.za_num_integers,
1064 			    lot_attrs))) != 0) {
1065 				kmem_free(lot_attrs, sizeof (sa_attr_type_t) *
1066 				    za.za_num_integers);
1067 				break;
1068 			}
1069 			VERIFY(ddi_strtoull(za.za_name, NULL, 10,
1070 			    (unsigned long long *)&lot_num) == 0);
1071 
1072 			(void) sa_add_layout_entry(os, lot_attrs,
1073 			    za.za_num_integers, lot_num,
1074 			    sa_layout_info_hash(lot_attrs,
1075 			    za.za_num_integers), B_FALSE, NULL);
1076 			kmem_free(lot_attrs, sizeof (sa_attr_type_t) *
1077 			    za.za_num_integers);
1078 		}
1079 		zap_cursor_fini(&zc);
1080 
1081 		/*
1082 		 * Make sure layout count matches number of entries added
1083 		 * to AVL tree
1084 		 */
1085 		if (avl_numnodes(&sa->sa_layout_num_tree) != layout_count) {
1086 			ASSERT(error != 0);
1087 			goto fail;
1088 		}
1089 	}
1090 
1091 	/* Add special layout number for old ZNODES */
1092 	if (ostype == DMU_OST_ZFS) {
1093 		(void) sa_add_layout_entry(os, sa_legacy_zpl_layout,
1094 		    sa_legacy_attr_count, 0,
1095 		    sa_layout_info_hash(sa_legacy_zpl_layout,
1096 		    sa_legacy_attr_count), B_FALSE, NULL);
1097 
1098 		(void) sa_add_layout_entry(os, sa_dummy_zpl_layout, 0, 1,
1099 		    0, B_FALSE, NULL);
1100 	}
1101 	*user_table = os->os_sa->sa_user_table;
1102 	mutex_exit(&sa->sa_lock);
1103 	return (0);
1104 fail:
1105 	os->os_sa = NULL;
1106 	sa_free_attr_table(sa);
1107 	if (sa->sa_user_table)
1108 		kmem_free(sa->sa_user_table, sa->sa_user_table_sz);
1109 	mutex_exit(&sa->sa_lock);
1110 	avl_destroy(&sa->sa_layout_hash_tree);
1111 	avl_destroy(&sa->sa_layout_num_tree);
1112 	mutex_destroy(&sa->sa_lock);
1113 	kmem_free(sa, sizeof (sa_os_t));
1114 	return ((error == ECKSUM) ? EIO : error);
1115 }
1116 
1117 void
sa_tear_down(objset_t * os)1118 sa_tear_down(objset_t *os)
1119 {
1120 	sa_os_t *sa = os->os_sa;
1121 	sa_lot_t *layout;
1122 	void *cookie;
1123 
1124 	kmem_free(sa->sa_user_table, sa->sa_user_table_sz);
1125 
1126 	/* Free up attr table */
1127 
1128 	sa_free_attr_table(sa);
1129 
1130 	cookie = NULL;
1131 	while (layout = avl_destroy_nodes(&sa->sa_layout_hash_tree, &cookie)) {
1132 		sa_idx_tab_t *tab;
1133 		while (tab = list_head(&layout->lot_idx_tab)) {
1134 			ASSERT(zfs_refcount_count(&tab->sa_refcount));
1135 			sa_idx_tab_rele(os, tab);
1136 		}
1137 	}
1138 
1139 	cookie = NULL;
1140 	while (layout = avl_destroy_nodes(&sa->sa_layout_num_tree, &cookie)) {
1141 		kmem_free(layout->lot_attrs,
1142 		    sizeof (sa_attr_type_t) * layout->lot_attr_count);
1143 		kmem_free(layout, sizeof (sa_lot_t));
1144 	}
1145 
1146 	avl_destroy(&sa->sa_layout_hash_tree);
1147 	avl_destroy(&sa->sa_layout_num_tree);
1148 	mutex_destroy(&sa->sa_lock);
1149 
1150 	kmem_free(sa, sizeof (sa_os_t));
1151 	os->os_sa = NULL;
1152 }
1153 
1154 void
sa_build_idx_tab(void * hdr,void * attr_addr,sa_attr_type_t attr,uint16_t length,int length_idx,boolean_t var_length,void * userp)1155 sa_build_idx_tab(void *hdr, void *attr_addr, sa_attr_type_t attr,
1156     uint16_t length, int length_idx, boolean_t var_length, void *userp)
1157 {
1158 	sa_idx_tab_t *idx_tab = userp;
1159 
1160 	if (var_length) {
1161 		ASSERT(idx_tab->sa_variable_lengths);
1162 		idx_tab->sa_variable_lengths[length_idx] = length;
1163 	}
1164 	TOC_ATTR_ENCODE(idx_tab->sa_idx_tab[attr], length_idx,
1165 	    (uint32_t)((uintptr_t)attr_addr - (uintptr_t)hdr));
1166 }
1167 
1168 static void
sa_attr_iter(objset_t * os,sa_hdr_phys_t * hdr,dmu_object_type_t type,sa_iterfunc_t func,sa_lot_t * tab,void * userp)1169 sa_attr_iter(objset_t *os, sa_hdr_phys_t *hdr, dmu_object_type_t type,
1170     sa_iterfunc_t func, sa_lot_t *tab, void *userp)
1171 {
1172 	void *data_start;
1173 	sa_lot_t *tb = tab;
1174 	sa_lot_t search;
1175 	avl_index_t loc;
1176 	sa_os_t *sa = os->os_sa;
1177 	int i;
1178 	uint16_t *length_start = NULL;
1179 	uint8_t length_idx = 0;
1180 
1181 	if (tab == NULL) {
1182 		search.lot_num = SA_LAYOUT_NUM(hdr, type);
1183 		tb = avl_find(&sa->sa_layout_num_tree, &search, &loc);
1184 		ASSERT(tb);
1185 	}
1186 
1187 	if (IS_SA_BONUSTYPE(type)) {
1188 		data_start = (void *)P2ROUNDUP(((uintptr_t)hdr +
1189 		    offsetof(sa_hdr_phys_t, sa_lengths) +
1190 		    (sizeof (uint16_t) * tb->lot_var_sizes)), 8);
1191 		length_start = hdr->sa_lengths;
1192 	} else {
1193 		data_start = hdr;
1194 	}
1195 
1196 	for (i = 0; i != tb->lot_attr_count; i++) {
1197 		int attr_length, reg_length;
1198 		uint8_t idx_len;
1199 
1200 		reg_length = sa->sa_attr_table[tb->lot_attrs[i]].sa_length;
1201 		if (reg_length) {
1202 			attr_length = reg_length;
1203 			idx_len = 0;
1204 		} else {
1205 			attr_length = length_start[length_idx];
1206 			idx_len = length_idx++;
1207 		}
1208 
1209 		func(hdr, data_start, tb->lot_attrs[i], attr_length,
1210 		    idx_len, reg_length == 0 ? B_TRUE : B_FALSE, userp);
1211 
1212 		data_start = (void *)P2ROUNDUP(((uintptr_t)data_start +
1213 		    attr_length), 8);
1214 	}
1215 }
1216 
1217 /*ARGSUSED*/
1218 void
sa_byteswap_cb(void * hdr,void * attr_addr,sa_attr_type_t attr,uint16_t length,int length_idx,boolean_t variable_length,void * userp)1219 sa_byteswap_cb(void *hdr, void *attr_addr, sa_attr_type_t attr,
1220     uint16_t length, int length_idx, boolean_t variable_length, void *userp)
1221 {
1222 	sa_handle_t *hdl = userp;
1223 	sa_os_t *sa = hdl->sa_os->os_sa;
1224 
1225 	sa_bswap_table[sa->sa_attr_table[attr].sa_byteswap](attr_addr, length);
1226 }
1227 
1228 void
sa_byteswap(sa_handle_t * hdl,sa_buf_type_t buftype)1229 sa_byteswap(sa_handle_t *hdl, sa_buf_type_t buftype)
1230 {
1231 	sa_hdr_phys_t *sa_hdr_phys = SA_GET_HDR(hdl, buftype);
1232 	dmu_buf_impl_t *db;
1233 	sa_os_t *sa = hdl->sa_os->os_sa;
1234 	int num_lengths = 1;
1235 	int i;
1236 
1237 	ASSERT(MUTEX_HELD(&sa->sa_lock));
1238 	if (sa_hdr_phys->sa_magic == SA_MAGIC)
1239 		return;
1240 
1241 	db = SA_GET_DB(hdl, buftype);
1242 
1243 	if (buftype == SA_SPILL) {
1244 		arc_release(db->db_buf, NULL);
1245 		arc_buf_thaw(db->db_buf);
1246 	}
1247 
1248 	sa_hdr_phys->sa_magic = BSWAP_32(sa_hdr_phys->sa_magic);
1249 	sa_hdr_phys->sa_layout_info = BSWAP_16(sa_hdr_phys->sa_layout_info);
1250 
1251 	/*
1252 	 * Determine number of variable lenghts in header
1253 	 * The standard 8 byte header has one for free and a
1254 	 * 16 byte header would have 4 + 1;
1255 	 */
1256 	if (SA_HDR_SIZE(sa_hdr_phys) > 8)
1257 		num_lengths += (SA_HDR_SIZE(sa_hdr_phys) - 8) >> 1;
1258 	for (i = 0; i != num_lengths; i++)
1259 		sa_hdr_phys->sa_lengths[i] =
1260 		    BSWAP_16(sa_hdr_phys->sa_lengths[i]);
1261 
1262 	sa_attr_iter(hdl->sa_os, sa_hdr_phys, DMU_OT_SA,
1263 	    sa_byteswap_cb, NULL, hdl);
1264 
1265 	if (buftype == SA_SPILL)
1266 		arc_buf_freeze(((dmu_buf_impl_t *)hdl->sa_spill)->db_buf);
1267 }
1268 
1269 static int
sa_build_index(sa_handle_t * hdl,sa_buf_type_t buftype)1270 sa_build_index(sa_handle_t *hdl, sa_buf_type_t buftype)
1271 {
1272 	sa_hdr_phys_t *sa_hdr_phys;
1273 	dmu_buf_impl_t *db = SA_GET_DB(hdl, buftype);
1274 	dmu_object_type_t bonustype = SA_BONUSTYPE_FROM_DB(db);
1275 	sa_os_t *sa = hdl->sa_os->os_sa;
1276 	sa_idx_tab_t *idx_tab;
1277 
1278 	sa_hdr_phys = SA_GET_HDR(hdl, buftype);
1279 
1280 	mutex_enter(&sa->sa_lock);
1281 
1282 	/* Do we need to byteswap? */
1283 
1284 	/* only check if not old znode */
1285 	if (IS_SA_BONUSTYPE(bonustype) && sa_hdr_phys->sa_magic != SA_MAGIC &&
1286 	    sa_hdr_phys->sa_magic != 0) {
1287 		VERIFY(BSWAP_32(sa_hdr_phys->sa_magic) == SA_MAGIC);
1288 		sa_byteswap(hdl, buftype);
1289 	}
1290 
1291 	idx_tab = sa_find_idx_tab(hdl->sa_os, bonustype, sa_hdr_phys);
1292 
1293 	if (buftype == SA_BONUS)
1294 		hdl->sa_bonus_tab = idx_tab;
1295 	else
1296 		hdl->sa_spill_tab = idx_tab;
1297 
1298 	mutex_exit(&sa->sa_lock);
1299 	return (0);
1300 }
1301 
1302 /*ARGSUSED*/
1303 static void
sa_evict_sync(void * dbu)1304 sa_evict_sync(void *dbu)
1305 {
1306 	panic("evicting sa dbuf\n");
1307 }
1308 
1309 static void
sa_idx_tab_rele(objset_t * os,void * arg)1310 sa_idx_tab_rele(objset_t *os, void *arg)
1311 {
1312 	sa_os_t *sa = os->os_sa;
1313 	sa_idx_tab_t *idx_tab = arg;
1314 
1315 	if (idx_tab == NULL)
1316 		return;
1317 
1318 	mutex_enter(&sa->sa_lock);
1319 	if (zfs_refcount_remove(&idx_tab->sa_refcount, NULL) == 0) {
1320 		list_remove(&idx_tab->sa_layout->lot_idx_tab, idx_tab);
1321 		if (idx_tab->sa_variable_lengths)
1322 			kmem_free(idx_tab->sa_variable_lengths,
1323 			    sizeof (uint16_t) *
1324 			    idx_tab->sa_layout->lot_var_sizes);
1325 		zfs_refcount_destroy(&idx_tab->sa_refcount);
1326 		kmem_free(idx_tab->sa_idx_tab,
1327 		    sizeof (uint32_t) * sa->sa_num_attrs);
1328 		kmem_free(idx_tab, sizeof (sa_idx_tab_t));
1329 	}
1330 	mutex_exit(&sa->sa_lock);
1331 }
1332 
1333 static void
sa_idx_tab_hold(objset_t * os,sa_idx_tab_t * idx_tab)1334 sa_idx_tab_hold(objset_t *os, sa_idx_tab_t *idx_tab)
1335 {
1336 	sa_os_t *sa = os->os_sa;
1337 
1338 	ASSERT(MUTEX_HELD(&sa->sa_lock));
1339 	(void) zfs_refcount_add(&idx_tab->sa_refcount, NULL);
1340 }
1341 
1342 void
sa_handle_destroy(sa_handle_t * hdl)1343 sa_handle_destroy(sa_handle_t *hdl)
1344 {
1345 	dmu_buf_t *db = hdl->sa_bonus;
1346 
1347 	mutex_enter(&hdl->sa_lock);
1348 	(void) dmu_buf_remove_user(db, &hdl->sa_dbu);
1349 
1350 	if (hdl->sa_bonus_tab)
1351 		sa_idx_tab_rele(hdl->sa_os, hdl->sa_bonus_tab);
1352 
1353 	if (hdl->sa_spill_tab)
1354 		sa_idx_tab_rele(hdl->sa_os, hdl->sa_spill_tab);
1355 
1356 	dmu_buf_rele(hdl->sa_bonus, NULL);
1357 
1358 	if (hdl->sa_spill)
1359 		dmu_buf_rele((dmu_buf_t *)hdl->sa_spill, NULL);
1360 	mutex_exit(&hdl->sa_lock);
1361 
1362 	kmem_cache_free(sa_cache, hdl);
1363 }
1364 
1365 int
sa_handle_get_from_db(objset_t * os,dmu_buf_t * db,void * userp,sa_handle_type_t hdl_type,sa_handle_t ** handlepp)1366 sa_handle_get_from_db(objset_t *os, dmu_buf_t *db, void *userp,
1367     sa_handle_type_t hdl_type, sa_handle_t **handlepp)
1368 {
1369 	int error = 0;
1370 	dmu_object_info_t doi;
1371 	sa_handle_t *handle = NULL;
1372 
1373 #ifdef ZFS_DEBUG
1374 	dmu_object_info_from_db(db, &doi);
1375 	ASSERT(doi.doi_bonus_type == DMU_OT_SA ||
1376 	    doi.doi_bonus_type == DMU_OT_ZNODE);
1377 #endif
1378 	/* find handle, if it exists */
1379 	/* if one doesn't exist then create a new one, and initialize it */
1380 
1381 	if (hdl_type == SA_HDL_SHARED)
1382 		handle = dmu_buf_get_user(db);
1383 
1384 	if (handle == NULL) {
1385 		sa_handle_t *winner = NULL;
1386 
1387 		handle = kmem_cache_alloc(sa_cache, KM_SLEEP);
1388 		handle->sa_dbu.dbu_evict_func_sync = NULL;
1389 		handle->sa_dbu.dbu_evict_func_async = NULL;
1390 		handle->sa_userp = userp;
1391 		handle->sa_bonus = db;
1392 		handle->sa_os = os;
1393 		handle->sa_spill = NULL;
1394 		handle->sa_bonus_tab = NULL;
1395 		handle->sa_spill_tab = NULL;
1396 
1397 		error = sa_build_index(handle, SA_BONUS);
1398 
1399 		if (hdl_type == SA_HDL_SHARED) {
1400 			dmu_buf_init_user(&handle->sa_dbu, sa_evict_sync, NULL,
1401 			    NULL);
1402 			winner = dmu_buf_set_user_ie(db, &handle->sa_dbu);
1403 		}
1404 
1405 		if (winner != NULL) {
1406 			kmem_cache_free(sa_cache, handle);
1407 			handle = winner;
1408 		}
1409 	}
1410 	*handlepp = handle;
1411 
1412 	return (error);
1413 }
1414 
1415 int
sa_handle_get(objset_t * objset,uint64_t objid,void * userp,sa_handle_type_t hdl_type,sa_handle_t ** handlepp)1416 sa_handle_get(objset_t *objset, uint64_t objid, void *userp,
1417     sa_handle_type_t hdl_type, sa_handle_t **handlepp)
1418 {
1419 	dmu_buf_t *db;
1420 	int error;
1421 
1422 	if (error = dmu_bonus_hold(objset, objid, NULL, &db))
1423 		return (error);
1424 
1425 	return (sa_handle_get_from_db(objset, db, userp, hdl_type,
1426 	    handlepp));
1427 }
1428 
1429 int
sa_buf_hold(objset_t * objset,uint64_t obj_num,void * tag,dmu_buf_t ** db)1430 sa_buf_hold(objset_t *objset, uint64_t obj_num, void *tag, dmu_buf_t **db)
1431 {
1432 	return (dmu_bonus_hold(objset, obj_num, tag, db));
1433 }
1434 
1435 void
sa_buf_rele(dmu_buf_t * db,void * tag)1436 sa_buf_rele(dmu_buf_t *db, void *tag)
1437 {
1438 	dmu_buf_rele(db, tag);
1439 }
1440 
1441 int
sa_lookup_impl(sa_handle_t * hdl,sa_bulk_attr_t * bulk,int count)1442 sa_lookup_impl(sa_handle_t *hdl, sa_bulk_attr_t *bulk, int count)
1443 {
1444 	ASSERT(hdl);
1445 	ASSERT(MUTEX_HELD(&hdl->sa_lock));
1446 	return (sa_attr_op(hdl, bulk, count, SA_LOOKUP, NULL));
1447 }
1448 
1449 static int
sa_lookup_locked(sa_handle_t * hdl,sa_attr_type_t attr,void * buf,uint32_t buflen)1450 sa_lookup_locked(sa_handle_t *hdl, sa_attr_type_t attr, void *buf,
1451     uint32_t buflen)
1452 {
1453 	int error;
1454 	sa_bulk_attr_t bulk;
1455 
1456 	bulk.sa_attr = attr;
1457 	bulk.sa_data = buf;
1458 	bulk.sa_length = buflen;
1459 	bulk.sa_data_func = NULL;
1460 
1461 	ASSERT(hdl);
1462 	error = sa_lookup_impl(hdl, &bulk, 1);
1463 	return (error);
1464 }
1465 
1466 int
sa_lookup(sa_handle_t * hdl,sa_attr_type_t attr,void * buf,uint32_t buflen)1467 sa_lookup(sa_handle_t *hdl, sa_attr_type_t attr, void *buf, uint32_t buflen)
1468 {
1469 	int error;
1470 
1471 	mutex_enter(&hdl->sa_lock);
1472 	error = sa_lookup_locked(hdl, attr, buf, buflen);
1473 	mutex_exit(&hdl->sa_lock);
1474 
1475 	return (error);
1476 }
1477 
1478 #ifdef _KERNEL
1479 int
sa_lookup_uio(sa_handle_t * hdl,sa_attr_type_t attr,uio_t * uio)1480 sa_lookup_uio(sa_handle_t *hdl, sa_attr_type_t attr, uio_t *uio)
1481 {
1482 	int error;
1483 	sa_bulk_attr_t bulk;
1484 
1485 	bulk.sa_data = NULL;
1486 	bulk.sa_attr = attr;
1487 	bulk.sa_data_func = NULL;
1488 
1489 	ASSERT(hdl);
1490 
1491 	mutex_enter(&hdl->sa_lock);
1492 	if ((error = sa_attr_op(hdl, &bulk, 1, SA_LOOKUP, NULL)) == 0) {
1493 		error = uiomove((void *)bulk.sa_addr, MIN(bulk.sa_size,
1494 		    uio->uio_resid), UIO_READ, uio);
1495 	}
1496 	mutex_exit(&hdl->sa_lock);
1497 	return (error);
1498 
1499 }
1500 
1501 /*
1502  * For the existing object that is upgraded from old system, its ondisk layout
1503  * has no slot for the project ID attribute. But quota accounting logic needs
1504  * to access related slots by offset directly. So we need to adjust these old
1505  * objects' layout to make the project ID to some unified and fixed offset.
1506  */
1507 int
sa_add_projid(sa_handle_t * hdl,dmu_tx_t * tx,uint64_t projid)1508 sa_add_projid(sa_handle_t *hdl, dmu_tx_t *tx, uint64_t projid)
1509 {
1510 	znode_t *zp = sa_get_userdata(hdl);
1511 	dmu_buf_t *db = sa_get_db(hdl);
1512 	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1513 	int count = 0, err = 0;
1514 	sa_bulk_attr_t *bulk, *attrs;
1515 	zfs_acl_locator_cb_t locate = { 0 };
1516 	uint64_t uid, gid, mode, rdev, xattr = 0, parent, gen, links;
1517 	uint64_t crtime[2], mtime[2], ctime[2], atime[2];
1518 	zfs_acl_phys_t znode_acl = { 0 };
1519 	char scanstamp[AV_SCANSTAMP_SZ];
1520 
1521 	if (zp->z_acl_cached == NULL) {
1522 		zfs_acl_t *aclp;
1523 
1524 		rw_enter(&zp->z_acl_lock, RW_WRITER);
1525 		err = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
1526 		rw_exit(&zp->z_acl_lock);
1527 		if (err != 0 && err != ENOENT)
1528 			return (err);
1529 	}
1530 
1531 	bulk = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
1532 	attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
1533 	mutex_enter(&hdl->sa_lock);
1534 	mutex_enter(&zp->z_lock);
1535 
1536 	err = sa_lookup_locked(hdl, SA_ZPL_PROJID(zfsvfs), &projid,
1537 	    sizeof (uint64_t));
1538 	if (unlikely(err == 0))
1539 		/* Someone has added project ID attr by race. */
1540 		err = EEXIST;
1541 	if (err != ENOENT)
1542 		goto out;
1543 
1544 	/* First do a bulk query of the attributes that aren't cached */
1545 	if (zp->z_is_sa) {
1546 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1547 		    &mode, 8);
1548 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1549 		    &gen, 8);
1550 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1551 		    &uid, 8);
1552 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1553 		    &gid, 8);
1554 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
1555 		    &parent, 8);
1556 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1557 		    &atime, 16);
1558 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
1559 		    &mtime, 16);
1560 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
1561 		    &ctime, 16);
1562 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL,
1563 		    &crtime, 16);
1564 		if (S_ISBLK(zp->z_mode) || S_ISCHR(zp->z_mode))
1565 			SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL,
1566 			    &rdev, 8);
1567 	} else {
1568 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1569 		    &atime, 16);
1570 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
1571 		    &mtime, 16);
1572 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
1573 		    &ctime, 16);
1574 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL,
1575 		    &crtime, 16);
1576 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1577 		    &gen, 8);
1578 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1579 		    &mode, 8);
1580 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
1581 		    &parent, 8);
1582 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL,
1583 		    &xattr, 8);
1584 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL,
1585 		    &rdev, 8);
1586 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1587 		    &uid, 8);
1588 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1589 		    &gid, 8);
1590 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
1591 		    &znode_acl, 88);
1592 	}
1593 	err = sa_bulk_lookup_locked(hdl, bulk, count);
1594 	if (err != 0)
1595 		goto out;
1596 
1597 	err = sa_lookup_locked(hdl, SA_ZPL_XATTR(zfsvfs), &xattr, 8);
1598 	if (err != 0 && err != ENOENT)
1599 		goto out;
1600 
1601 	zp->z_projid = projid;
1602 	zp->z_pflags |= ZFS_PROJID;
1603 	links = zp->z_links;
1604 	count = 0;
1605 	err = 0;
1606 
1607 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
1608 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_SIZE(zfsvfs), NULL,
1609 	    &zp->z_size, 8);
1610 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_GEN(zfsvfs), NULL, &gen, 8);
1611 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
1612 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
1613 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
1614 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1615 	    &zp->z_pflags, 8);
1616 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16);
1617 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
1618 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
1619 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL,
1620 	    &crtime, 16);
1621 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
1622 	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_PROJID(zfsvfs), NULL, &projid, 8);
1623 
1624 	if (S_ISBLK(zp->z_mode) || S_ISCHR(zp->z_mode))
1625 		SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_RDEV(zfsvfs), NULL,
1626 		    &rdev, 8);
1627 
1628 	if (zp->z_acl_cached != NULL) {
1629 		SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
1630 		    &zp->z_acl_cached->z_acl_count, 8);
1631 		if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
1632 			zfs_acl_xform(zp, zp->z_acl_cached, CRED());
1633 		locate.cb_aclp = zp->z_acl_cached;
1634 		SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_DACL_ACES(zfsvfs),
1635 		    zfs_acl_data_locator, &locate,
1636 		    zp->z_acl_cached->z_acl_bytes);
1637 	}
1638 
1639 	if (xattr)
1640 		SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_XATTR(zfsvfs), NULL,
1641 		    &xattr, 8);
1642 
1643 	if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
1644 		bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
1645 		    scanstamp, AV_SCANSTAMP_SZ);
1646 		SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_SCANSTAMP(zfsvfs), NULL,
1647 		    scanstamp, AV_SCANSTAMP_SZ);
1648 		zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
1649 	}
1650 
1651 	VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
1652 	VERIFY(sa_replace_all_by_template_locked(hdl, attrs, count, tx) == 0);
1653 	if (znode_acl.z_acl_extern_obj) {
1654 		VERIFY(0 == dmu_object_free(zfsvfs->z_os,
1655 		    znode_acl.z_acl_extern_obj, tx));
1656 	}
1657 
1658 	zp->z_is_sa = B_TRUE;
1659 
1660 out:
1661 	mutex_exit(&zp->z_lock);
1662 	mutex_exit(&hdl->sa_lock);
1663 	kmem_free(attrs, sizeof (sa_bulk_attr_t) * ZPL_END);
1664 	kmem_free(bulk, sizeof (sa_bulk_attr_t) * ZPL_END);
1665 	return (err);
1666 }
1667 #endif
1668 
1669 static sa_idx_tab_t *
sa_find_idx_tab(objset_t * os,dmu_object_type_t bonustype,sa_hdr_phys_t * hdr)1670 sa_find_idx_tab(objset_t *os, dmu_object_type_t bonustype, sa_hdr_phys_t *hdr)
1671 {
1672 	sa_idx_tab_t *idx_tab;
1673 	sa_os_t *sa = os->os_sa;
1674 	sa_lot_t *tb, search;
1675 	avl_index_t loc;
1676 
1677 	/*
1678 	 * Deterimine layout number.  If SA node and header == 0 then
1679 	 * force the index table to the dummy "1" empty layout.
1680 	 *
1681 	 * The layout number would only be zero for a newly created file
1682 	 * that has not added any attributes yet, or with crypto enabled which
1683 	 * doesn't write any attributes to the bonus buffer.
1684 	 */
1685 
1686 	search.lot_num = SA_LAYOUT_NUM(hdr, bonustype);
1687 
1688 	tb = avl_find(&sa->sa_layout_num_tree, &search, &loc);
1689 
1690 	/* Verify header size is consistent with layout information */
1691 	ASSERT(tb);
1692 	ASSERT(IS_SA_BONUSTYPE(bonustype) &&
1693 	    SA_HDR_SIZE_MATCH_LAYOUT(hdr, tb) || !IS_SA_BONUSTYPE(bonustype) ||
1694 	    (IS_SA_BONUSTYPE(bonustype) && hdr->sa_layout_info == 0));
1695 
1696 	/*
1697 	 * See if any of the already existing TOC entries can be reused?
1698 	 */
1699 
1700 	for (idx_tab = list_head(&tb->lot_idx_tab); idx_tab;
1701 	    idx_tab = list_next(&tb->lot_idx_tab, idx_tab)) {
1702 		boolean_t valid_idx = B_TRUE;
1703 		int i;
1704 
1705 		if (tb->lot_var_sizes != 0 &&
1706 		    idx_tab->sa_variable_lengths != NULL) {
1707 			for (i = 0; i != tb->lot_var_sizes; i++) {
1708 				if (hdr->sa_lengths[i] !=
1709 				    idx_tab->sa_variable_lengths[i]) {
1710 					valid_idx = B_FALSE;
1711 					break;
1712 				}
1713 			}
1714 		}
1715 		if (valid_idx) {
1716 			sa_idx_tab_hold(os, idx_tab);
1717 			return (idx_tab);
1718 		}
1719 	}
1720 
1721 	/* No such luck, create a new entry */
1722 	idx_tab = kmem_zalloc(sizeof (sa_idx_tab_t), KM_SLEEP);
1723 	idx_tab->sa_idx_tab =
1724 	    kmem_zalloc(sizeof (uint32_t) * sa->sa_num_attrs, KM_SLEEP);
1725 	idx_tab->sa_layout = tb;
1726 	zfs_refcount_create(&idx_tab->sa_refcount);
1727 	if (tb->lot_var_sizes)
1728 		idx_tab->sa_variable_lengths = kmem_alloc(sizeof (uint16_t) *
1729 		    tb->lot_var_sizes, KM_SLEEP);
1730 
1731 	sa_attr_iter(os, hdr, bonustype, sa_build_idx_tab,
1732 	    tb, idx_tab);
1733 	sa_idx_tab_hold(os, idx_tab);   /* one hold for consumer */
1734 	sa_idx_tab_hold(os, idx_tab);	/* one for layout */
1735 	list_insert_tail(&tb->lot_idx_tab, idx_tab);
1736 	return (idx_tab);
1737 }
1738 
1739 void
sa_default_locator(void ** dataptr,uint32_t * len,uint32_t total_len,boolean_t start,void * userdata)1740 sa_default_locator(void **dataptr, uint32_t *len, uint32_t total_len,
1741     boolean_t start, void *userdata)
1742 {
1743 	ASSERT(start);
1744 
1745 	*dataptr = userdata;
1746 	*len = total_len;
1747 }
1748 
1749 static void
sa_attr_register_sync(sa_handle_t * hdl,dmu_tx_t * tx)1750 sa_attr_register_sync(sa_handle_t *hdl, dmu_tx_t *tx)
1751 {
1752 	uint64_t attr_value = 0;
1753 	sa_os_t *sa = hdl->sa_os->os_sa;
1754 	sa_attr_table_t *tb = sa->sa_attr_table;
1755 	int i;
1756 
1757 	mutex_enter(&sa->sa_lock);
1758 
1759 	if (!sa->sa_need_attr_registration || sa->sa_master_obj == 0) {
1760 		mutex_exit(&sa->sa_lock);
1761 		return;
1762 	}
1763 
1764 	if (sa->sa_reg_attr_obj == 0) {
1765 		sa->sa_reg_attr_obj = zap_create_link(hdl->sa_os,
1766 		    DMU_OT_SA_ATTR_REGISTRATION,
1767 		    sa->sa_master_obj, SA_REGISTRY, tx);
1768 	}
1769 	for (i = 0; i != sa->sa_num_attrs; i++) {
1770 		if (sa->sa_attr_table[i].sa_registered)
1771 			continue;
1772 		ATTR_ENCODE(attr_value, tb[i].sa_attr, tb[i].sa_length,
1773 		    tb[i].sa_byteswap);
1774 		VERIFY(0 == zap_update(hdl->sa_os, sa->sa_reg_attr_obj,
1775 		    tb[i].sa_name, 8, 1, &attr_value, tx));
1776 		tb[i].sa_registered = B_TRUE;
1777 	}
1778 	sa->sa_need_attr_registration = B_FALSE;
1779 	mutex_exit(&sa->sa_lock);
1780 }
1781 
1782 /*
1783  * Replace all attributes with attributes specified in template.
1784  * If dnode had a spill buffer then those attributes will be
1785  * also be replaced, possibly with just an empty spill block
1786  *
1787  * This interface is intended to only be used for bulk adding of
1788  * attributes for a new file.  It will also be used by the ZPL
1789  * when converting and old formatted znode to native SA support.
1790  */
1791 int
sa_replace_all_by_template_locked(sa_handle_t * hdl,sa_bulk_attr_t * attr_desc,int attr_count,dmu_tx_t * tx)1792 sa_replace_all_by_template_locked(sa_handle_t *hdl, sa_bulk_attr_t *attr_desc,
1793     int attr_count, dmu_tx_t *tx)
1794 {
1795 	sa_os_t *sa = hdl->sa_os->os_sa;
1796 
1797 	if (sa->sa_need_attr_registration)
1798 		sa_attr_register_sync(hdl, tx);
1799 	return (sa_build_layouts(hdl, attr_desc, attr_count, tx));
1800 }
1801 
1802 int
sa_replace_all_by_template(sa_handle_t * hdl,sa_bulk_attr_t * attr_desc,int attr_count,dmu_tx_t * tx)1803 sa_replace_all_by_template(sa_handle_t *hdl, sa_bulk_attr_t *attr_desc,
1804     int attr_count, dmu_tx_t *tx)
1805 {
1806 	int error;
1807 
1808 	mutex_enter(&hdl->sa_lock);
1809 	error = sa_replace_all_by_template_locked(hdl, attr_desc,
1810 	    attr_count, tx);
1811 	mutex_exit(&hdl->sa_lock);
1812 	return (error);
1813 }
1814 
1815 /*
1816  * Add/remove a single attribute or replace a variable-sized attribute value
1817  * with a value of a different size, and then rewrite the entire set
1818  * of attributes.
1819  * Same-length attribute value replacement (including fixed-length attributes)
1820  * is handled more efficiently by the upper layers.
1821  */
1822 static int
sa_modify_attrs(sa_handle_t * hdl,sa_attr_type_t newattr,sa_data_op_t action,sa_data_locator_t * locator,void * datastart,uint16_t buflen,dmu_tx_t * tx)1823 sa_modify_attrs(sa_handle_t *hdl, sa_attr_type_t newattr,
1824     sa_data_op_t action, sa_data_locator_t *locator, void *datastart,
1825     uint16_t buflen, dmu_tx_t *tx)
1826 {
1827 	sa_os_t *sa = hdl->sa_os->os_sa;
1828 	dmu_buf_impl_t *db = (dmu_buf_impl_t *)hdl->sa_bonus;
1829 	dnode_t *dn;
1830 	sa_bulk_attr_t *attr_desc;
1831 	void *old_data[2];
1832 	int bonus_attr_count = 0;
1833 	int bonus_data_size = 0;
1834 	int spill_data_size = 0;
1835 	int spill_attr_count = 0;
1836 	int error;
1837 	uint16_t length, reg_length;
1838 	int i, j, k, length_idx;
1839 	sa_hdr_phys_t *hdr;
1840 	sa_idx_tab_t *idx_tab;
1841 	int attr_count;
1842 	int count;
1843 
1844 	ASSERT(MUTEX_HELD(&hdl->sa_lock));
1845 
1846 	/* First make of copy of the old data */
1847 
1848 	DB_DNODE_ENTER(db);
1849 	dn = DB_DNODE(db);
1850 	if (dn->dn_bonuslen != 0) {
1851 		bonus_data_size = hdl->sa_bonus->db_size;
1852 		old_data[0] = kmem_alloc(bonus_data_size, KM_SLEEP);
1853 		bcopy(hdl->sa_bonus->db_data, old_data[0],
1854 		    hdl->sa_bonus->db_size);
1855 		bonus_attr_count = hdl->sa_bonus_tab->sa_layout->lot_attr_count;
1856 	} else {
1857 		old_data[0] = NULL;
1858 	}
1859 	DB_DNODE_EXIT(db);
1860 
1861 	/* Bring spill buffer online if it isn't currently */
1862 
1863 	if ((error = sa_get_spill(hdl)) == 0) {
1864 		spill_data_size = hdl->sa_spill->db_size;
1865 		old_data[1] = kmem_alloc(spill_data_size, KM_SLEEP);
1866 		bcopy(hdl->sa_spill->db_data, old_data[1],
1867 		    hdl->sa_spill->db_size);
1868 		spill_attr_count =
1869 		    hdl->sa_spill_tab->sa_layout->lot_attr_count;
1870 	} else if (error && error != ENOENT) {
1871 		if (old_data[0])
1872 			kmem_free(old_data[0], bonus_data_size);
1873 		return (error);
1874 	} else {
1875 		old_data[1] = NULL;
1876 	}
1877 
1878 	/* build descriptor of all attributes */
1879 
1880 	attr_count = bonus_attr_count + spill_attr_count;
1881 	if (action == SA_ADD)
1882 		attr_count++;
1883 	else if (action == SA_REMOVE)
1884 		attr_count--;
1885 
1886 	attr_desc = kmem_zalloc(sizeof (sa_bulk_attr_t) * attr_count, KM_SLEEP);
1887 
1888 	/*
1889 	 * loop through bonus and spill buffer if it exists, and
1890 	 * build up new attr_descriptor to reset the attributes
1891 	 */
1892 	k = j = 0;
1893 	count = bonus_attr_count;
1894 	hdr = SA_GET_HDR(hdl, SA_BONUS);
1895 	idx_tab = SA_IDX_TAB_GET(hdl, SA_BONUS);
1896 	for (; k != 2; k++) {
1897 		/*
1898 		 * Iterate over each attribute in layout.  Fetch the
1899 		 * size of variable-length attributes needing rewrite
1900 		 * from sa_lengths[].
1901 		 */
1902 		for (i = 0, length_idx = 0; i != count; i++) {
1903 			sa_attr_type_t attr;
1904 
1905 			attr = idx_tab->sa_layout->lot_attrs[i];
1906 			reg_length = SA_REGISTERED_LEN(sa, attr);
1907 			if (reg_length == 0) {
1908 				length = hdr->sa_lengths[length_idx];
1909 				length_idx++;
1910 			} else {
1911 				length = reg_length;
1912 			}
1913 			if (attr == newattr) {
1914 				/*
1915 				 * There is nothing to do for SA_REMOVE,
1916 				 * so it is just skipped.
1917 				 */
1918 				if (action == SA_REMOVE)
1919 					continue;
1920 
1921 				/*
1922 				 * Duplicate attributes are not allowed, so the
1923 				 * action can not be SA_ADD here.
1924 				 */
1925 				ASSERT3S(action, ==, SA_REPLACE);
1926 
1927 				/*
1928 				 * Only a variable-sized attribute can be
1929 				 * replaced here, and its size must be changing.
1930 				 */
1931 				ASSERT3U(reg_length, ==, 0);
1932 				ASSERT3U(length, !=, buflen);
1933 				SA_ADD_BULK_ATTR(attr_desc, j, attr,
1934 				    locator, datastart, buflen);
1935 			} else {
1936 				SA_ADD_BULK_ATTR(attr_desc, j, attr,
1937 				    NULL, (void *)
1938 				    (TOC_OFF(idx_tab->sa_idx_tab[attr]) +
1939 				    (uintptr_t)old_data[k]), length);
1940 			}
1941 		}
1942 		if (k == 0 && hdl->sa_spill) {
1943 			hdr = SA_GET_HDR(hdl, SA_SPILL);
1944 			idx_tab = SA_IDX_TAB_GET(hdl, SA_SPILL);
1945 			count = spill_attr_count;
1946 		} else {
1947 			break;
1948 		}
1949 	}
1950 	if (action == SA_ADD) {
1951 		reg_length = SA_REGISTERED_LEN(sa, newattr);
1952 		IMPLY(reg_length != 0, reg_length == buflen);
1953 		SA_ADD_BULK_ATTR(attr_desc, j, newattr, locator,
1954 		    datastart, buflen);
1955 	}
1956 	ASSERT3U(j, ==, attr_count);
1957 
1958 	error = sa_build_layouts(hdl, attr_desc, attr_count, tx);
1959 
1960 	if (old_data[0])
1961 		kmem_free(old_data[0], bonus_data_size);
1962 	if (old_data[1])
1963 		kmem_free(old_data[1], spill_data_size);
1964 	kmem_free(attr_desc, sizeof (sa_bulk_attr_t) * attr_count);
1965 
1966 	return (error);
1967 }
1968 
1969 static int
sa_bulk_update_impl(sa_handle_t * hdl,sa_bulk_attr_t * bulk,int count,dmu_tx_t * tx)1970 sa_bulk_update_impl(sa_handle_t *hdl, sa_bulk_attr_t *bulk, int count,
1971     dmu_tx_t *tx)
1972 {
1973 	int error;
1974 	sa_os_t *sa = hdl->sa_os->os_sa;
1975 	dmu_object_type_t bonustype;
1976 
1977 	bonustype = SA_BONUSTYPE_FROM_DB(SA_GET_DB(hdl, SA_BONUS));
1978 
1979 	ASSERT(hdl);
1980 	ASSERT(MUTEX_HELD(&hdl->sa_lock));
1981 
1982 	/* sync out registration table if necessary */
1983 	if (sa->sa_need_attr_registration)
1984 		sa_attr_register_sync(hdl, tx);
1985 
1986 	error = sa_attr_op(hdl, bulk, count, SA_UPDATE, tx);
1987 	if (error == 0 && !IS_SA_BONUSTYPE(bonustype) && sa->sa_update_cb)
1988 		sa->sa_update_cb(hdl, tx);
1989 
1990 	return (error);
1991 }
1992 
1993 /*
1994  * update or add new attribute
1995  */
1996 int
sa_update(sa_handle_t * hdl,sa_attr_type_t type,void * buf,uint32_t buflen,dmu_tx_t * tx)1997 sa_update(sa_handle_t *hdl, sa_attr_type_t type,
1998     void *buf, uint32_t buflen, dmu_tx_t *tx)
1999 {
2000 	int error;
2001 	sa_bulk_attr_t bulk;
2002 
2003 	bulk.sa_attr = type;
2004 	bulk.sa_data_func = NULL;
2005 	bulk.sa_length = buflen;
2006 	bulk.sa_data = buf;
2007 
2008 	mutex_enter(&hdl->sa_lock);
2009 	error = sa_bulk_update_impl(hdl, &bulk, 1, tx);
2010 	mutex_exit(&hdl->sa_lock);
2011 	return (error);
2012 }
2013 
2014 int
sa_update_from_cb(sa_handle_t * hdl,sa_attr_type_t attr,uint32_t buflen,sa_data_locator_t * locator,void * userdata,dmu_tx_t * tx)2015 sa_update_from_cb(sa_handle_t *hdl, sa_attr_type_t attr,
2016     uint32_t buflen, sa_data_locator_t *locator, void *userdata, dmu_tx_t *tx)
2017 {
2018 	int error;
2019 	sa_bulk_attr_t bulk;
2020 
2021 	bulk.sa_attr = attr;
2022 	bulk.sa_data = userdata;
2023 	bulk.sa_data_func = locator;
2024 	bulk.sa_length = buflen;
2025 
2026 	mutex_enter(&hdl->sa_lock);
2027 	error = sa_bulk_update_impl(hdl, &bulk, 1, tx);
2028 	mutex_exit(&hdl->sa_lock);
2029 	return (error);
2030 }
2031 
2032 /*
2033  * Return size of an attribute
2034  */
2035 
2036 int
sa_size(sa_handle_t * hdl,sa_attr_type_t attr,int * size)2037 sa_size(sa_handle_t *hdl, sa_attr_type_t attr, int *size)
2038 {
2039 	sa_bulk_attr_t bulk;
2040 	int error;
2041 
2042 	bulk.sa_data = NULL;
2043 	bulk.sa_attr = attr;
2044 	bulk.sa_data_func = NULL;
2045 
2046 	ASSERT(hdl);
2047 	mutex_enter(&hdl->sa_lock);
2048 	if ((error = sa_attr_op(hdl, &bulk, 1, SA_LOOKUP, NULL)) != 0) {
2049 		mutex_exit(&hdl->sa_lock);
2050 		return (error);
2051 	}
2052 	*size = bulk.sa_size;
2053 
2054 	mutex_exit(&hdl->sa_lock);
2055 	return (0);
2056 }
2057 
2058 int
sa_bulk_lookup_locked(sa_handle_t * hdl,sa_bulk_attr_t * attrs,int count)2059 sa_bulk_lookup_locked(sa_handle_t *hdl, sa_bulk_attr_t *attrs, int count)
2060 {
2061 	ASSERT(hdl);
2062 	ASSERT(MUTEX_HELD(&hdl->sa_lock));
2063 	return (sa_lookup_impl(hdl, attrs, count));
2064 }
2065 
2066 int
sa_bulk_lookup(sa_handle_t * hdl,sa_bulk_attr_t * attrs,int count)2067 sa_bulk_lookup(sa_handle_t *hdl, sa_bulk_attr_t *attrs, int count)
2068 {
2069 	int error;
2070 
2071 	ASSERT(hdl);
2072 	mutex_enter(&hdl->sa_lock);
2073 	error = sa_bulk_lookup_locked(hdl, attrs, count);
2074 	mutex_exit(&hdl->sa_lock);
2075 	return (error);
2076 }
2077 
2078 int
sa_bulk_update(sa_handle_t * hdl,sa_bulk_attr_t * attrs,int count,dmu_tx_t * tx)2079 sa_bulk_update(sa_handle_t *hdl, sa_bulk_attr_t *attrs, int count, dmu_tx_t *tx)
2080 {
2081 	int error;
2082 
2083 	ASSERT(hdl);
2084 	mutex_enter(&hdl->sa_lock);
2085 	error = sa_bulk_update_impl(hdl, attrs, count, tx);
2086 	mutex_exit(&hdl->sa_lock);
2087 	return (error);
2088 }
2089 
2090 int
sa_remove(sa_handle_t * hdl,sa_attr_type_t attr,dmu_tx_t * tx)2091 sa_remove(sa_handle_t *hdl, sa_attr_type_t attr, dmu_tx_t *tx)
2092 {
2093 	int error;
2094 
2095 	mutex_enter(&hdl->sa_lock);
2096 	error = sa_modify_attrs(hdl, attr, SA_REMOVE, NULL,
2097 	    NULL, 0, tx);
2098 	mutex_exit(&hdl->sa_lock);
2099 	return (error);
2100 }
2101 
2102 void
sa_object_info(sa_handle_t * hdl,dmu_object_info_t * doi)2103 sa_object_info(sa_handle_t *hdl, dmu_object_info_t *doi)
2104 {
2105 	dmu_object_info_from_db((dmu_buf_t *)hdl->sa_bonus, doi);
2106 }
2107 
2108 void
sa_object_size(sa_handle_t * hdl,uint32_t * blksize,u_longlong_t * nblocks)2109 sa_object_size(sa_handle_t *hdl, uint32_t *blksize, u_longlong_t *nblocks)
2110 {
2111 	dmu_object_size_from_db((dmu_buf_t *)hdl->sa_bonus,
2112 	    blksize, nblocks);
2113 }
2114 
2115 void
sa_set_userp(sa_handle_t * hdl,void * ptr)2116 sa_set_userp(sa_handle_t *hdl, void *ptr)
2117 {
2118 	hdl->sa_userp = ptr;
2119 }
2120 
2121 dmu_buf_t *
sa_get_db(sa_handle_t * hdl)2122 sa_get_db(sa_handle_t *hdl)
2123 {
2124 	return ((dmu_buf_t *)hdl->sa_bonus);
2125 }
2126 
2127 void *
sa_get_userdata(sa_handle_t * hdl)2128 sa_get_userdata(sa_handle_t *hdl)
2129 {
2130 	return (hdl->sa_userp);
2131 }
2132 
2133 void
sa_register_update_callback_locked(objset_t * os,sa_update_cb_t * func)2134 sa_register_update_callback_locked(objset_t *os, sa_update_cb_t *func)
2135 {
2136 	ASSERT(MUTEX_HELD(&os->os_sa->sa_lock));
2137 	os->os_sa->sa_update_cb = func;
2138 }
2139 
2140 void
sa_register_update_callback(objset_t * os,sa_update_cb_t * func)2141 sa_register_update_callback(objset_t *os, sa_update_cb_t *func)
2142 {
2143 
2144 	mutex_enter(&os->os_sa->sa_lock);
2145 	sa_register_update_callback_locked(os, func);
2146 	mutex_exit(&os->os_sa->sa_lock);
2147 }
2148 
2149 uint64_t
sa_handle_object(sa_handle_t * hdl)2150 sa_handle_object(sa_handle_t *hdl)
2151 {
2152 	return (hdl->sa_bonus->db_object);
2153 }
2154 
2155 boolean_t
sa_enabled(objset_t * os)2156 sa_enabled(objset_t *os)
2157 {
2158 	return (os->os_sa == NULL);
2159 }
2160 
2161 int
sa_set_sa_object(objset_t * os,uint64_t sa_object)2162 sa_set_sa_object(objset_t *os, uint64_t sa_object)
2163 {
2164 	sa_os_t *sa = os->os_sa;
2165 
2166 	if (sa->sa_master_obj)
2167 		return (1);
2168 
2169 	sa->sa_master_obj = sa_object;
2170 
2171 	return (0);
2172 }
2173 
2174 int
sa_hdrsize(void * arg)2175 sa_hdrsize(void *arg)
2176 {
2177 	sa_hdr_phys_t *hdr = arg;
2178 
2179 	return (SA_HDR_SIZE(hdr));
2180 }
2181 
2182 void
sa_handle_lock(sa_handle_t * hdl)2183 sa_handle_lock(sa_handle_t *hdl)
2184 {
2185 	ASSERT(hdl);
2186 	mutex_enter(&hdl->sa_lock);
2187 }
2188 
2189 void
sa_handle_unlock(sa_handle_t * hdl)2190 sa_handle_unlock(sa_handle_t *hdl)
2191 {
2192 	ASSERT(hdl);
2193 	mutex_exit(&hdl->sa_lock);
2194 }
2195