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) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2012, 2016 by Delphix. All rights reserved.
25 * Copyright 2017 Nexenta Systems, Inc.
26 */
27
28 #ifndef _SYS_ZAP_H
29 #define _SYS_ZAP_H
30
31 /*
32 * ZAP - ZFS Attribute Processor
33 *
34 * The ZAP is a module which sits on top of the DMU (Data Management
35 * Unit) and implements a higher-level storage primitive using DMU
36 * objects. Its primary consumer is the ZPL (ZFS Posix Layer).
37 *
38 * A "zapobj" is a DMU object which the ZAP uses to stores attributes.
39 * Users should use only zap routines to access a zapobj - they should
40 * not access the DMU object directly using DMU routines.
41 *
42 * The attributes stored in a zapobj are name-value pairs. The name is
43 * a zero-terminated string of up to ZAP_MAXNAMELEN bytes (including
44 * terminating NULL). The value is an array of integers, which may be
45 * 1, 2, 4, or 8 bytes long. The total space used by the array (number
46 * of integers * integer length) can be up to ZAP_MAXVALUELEN bytes.
47 * Note that an 8-byte integer value can be used to store the location
48 * (object number) of another dmu object (which may be itself a zapobj).
49 * Note that you can use a zero-length attribute to store a single bit
50 * of information - the attribute is present or not.
51 *
52 * The ZAP routines are thread-safe. However, you must observe the
53 * DMU's restriction that a transaction may not be operated on
54 * concurrently.
55 *
56 * Any of the routines that return an int may return an I/O error (EIO
57 * or ECHECKSUM).
58 *
59 *
60 * Implementation / Performance Notes:
61 *
62 * The ZAP is intended to operate most efficiently on attributes with
63 * short (49 bytes or less) names and single 8-byte values, for which
64 * the microzap will be used. The ZAP should be efficient enough so
65 * that the user does not need to cache these attributes.
66 *
67 * The ZAP's locking scheme makes its routines thread-safe. Operations
68 * on different zapobjs will be processed concurrently. Operations on
69 * the same zapobj which only read data will be processed concurrently.
70 * Operations on the same zapobj which modify data will be processed
71 * concurrently when there are many attributes in the zapobj (because
72 * the ZAP uses per-block locking - more than 128 * (number of cpus)
73 * small attributes will suffice).
74 */
75
76 /*
77 * We're using zero-terminated byte strings (ie. ASCII or UTF-8 C
78 * strings) for the names of attributes, rather than a byte string
79 * bounded by an explicit length. If some day we want to support names
80 * in character sets which have embedded zeros (eg. UTF-16, UTF-32),
81 * we'll have to add routines for using length-bounded strings.
82 */
83
84 #include <sys/dmu.h>
85 #include <sys/refcount.h>
86
87 #ifdef __cplusplus
88 extern "C" {
89 #endif
90
91 /*
92 * Specifies matching criteria for ZAP lookups.
93 * MT_NORMALIZE Use ZAP normalization flags, which can include both
94 * unicode normalization and case-insensitivity.
95 * MT_MATCH_CASE Do case-sensitive lookups even if MT_NORMALIZE is
96 * specified and ZAP normalization flags include
97 * U8_TEXTPREP_TOUPPER.
98 */
99 typedef enum matchtype {
100 MT_NORMALIZE = 1 << 0,
101 MT_MATCH_CASE = 1 << 1,
102 } matchtype_t;
103
104 typedef enum zap_flags {
105 /* Use 64-bit hash value (serialized cursors will always use 64-bits) */
106 ZAP_FLAG_HASH64 = 1 << 0,
107 /* Key is binary, not string (zap_add_uint64() can be used) */
108 ZAP_FLAG_UINT64_KEY = 1 << 1,
109 /*
110 * First word of key (which must be an array of uint64) is
111 * already randomly distributed.
112 */
113 ZAP_FLAG_PRE_HASHED_KEY = 1 << 2,
114 } zap_flags_t;
115
116 /*
117 * Create a new zapobj with no attributes and return its object number.
118 */
119 uint64_t zap_create(objset_t *ds, dmu_object_type_t ot,
120 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
121 uint64_t zap_create_norm(objset_t *ds, int normflags, dmu_object_type_t ot,
122 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
123 uint64_t zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
124 dmu_object_type_t ot, int leaf_blockshift, int indirect_blockshift,
125 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
126 uint64_t zap_create_link(objset_t *os, dmu_object_type_t ot,
127 uint64_t parent_obj, const char *name, dmu_tx_t *tx);
128
129 /*
130 * Initialize an already-allocated object.
131 */
132 void mzap_create_impl(objset_t *os, uint64_t obj, int normflags,
133 zap_flags_t flags, dmu_tx_t *tx);
134
135 /*
136 * Create a new zapobj with no attributes from the given (unallocated)
137 * object number.
138 */
139 int zap_create_claim(objset_t *ds, uint64_t obj, dmu_object_type_t ot,
140 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
141 int zap_create_claim_norm(objset_t *ds, uint64_t obj,
142 int normflags, dmu_object_type_t ot,
143 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
144
145 /*
146 * The zapobj passed in must be a valid ZAP object for all of the
147 * following routines.
148 */
149
150 /*
151 * Destroy this zapobj and all its attributes.
152 *
153 * Frees the object number using dmu_object_free.
154 */
155 int zap_destroy(objset_t *ds, uint64_t zapobj, dmu_tx_t *tx);
156
157 /*
158 * Manipulate attributes.
159 *
160 * 'integer_size' is in bytes, and must be 1, 2, 4, or 8.
161 */
162
163 /*
164 * Retrieve the contents of the attribute with the given name.
165 *
166 * If the requested attribute does not exist, the call will fail and
167 * return ENOENT.
168 *
169 * If 'integer_size' is smaller than the attribute's integer size, the
170 * call will fail and return EINVAL.
171 *
172 * If 'integer_size' is equal to or larger than the attribute's integer
173 * size, the call will succeed and return 0.
174 *
175 * When converting to a larger integer size, the integers will be treated as
176 * unsigned (ie. no sign-extension will be performed).
177 *
178 * 'num_integers' is the length (in integers) of 'buf'.
179 *
180 * If the attribute is longer than the buffer, as many integers as will
181 * fit will be transferred to 'buf'. If the entire attribute was not
182 * transferred, the call will return EOVERFLOW.
183 */
184 int zap_lookup(objset_t *ds, uint64_t zapobj, const char *name,
185 uint64_t integer_size, uint64_t num_integers, void *buf);
186
187 /*
188 * If rn_len is nonzero, realname will be set to the name of the found
189 * entry (which may be different from the requested name if matchtype is
190 * not MT_EXACT).
191 *
192 * If normalization_conflictp is not NULL, it will be set if there is
193 * another name with the same case/unicode normalized form.
194 */
195 int zap_lookup_norm(objset_t *ds, uint64_t zapobj, const char *name,
196 uint64_t integer_size, uint64_t num_integers, void *buf,
197 matchtype_t mt, char *realname, int rn_len,
198 boolean_t *normalization_conflictp);
199 int zap_lookup_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
200 int key_numints, uint64_t integer_size, uint64_t num_integers, void *buf);
201 int zap_contains(objset_t *ds, uint64_t zapobj, const char *name);
202 int zap_prefetch_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
203 int key_numints);
204 int zap_lookup_by_dnode(dnode_t *dn, const char *name,
205 uint64_t integer_size, uint64_t num_integers, void *buf);
206 int zap_lookup_norm_by_dnode(dnode_t *dn, const char *name,
207 uint64_t integer_size, uint64_t num_integers, void *buf,
208 matchtype_t mt, char *realname, int rn_len,
209 boolean_t *ncp);
210
211 /*
212 * Create an attribute with the given name and value.
213 *
214 * If an attribute with the given name already exists, the call will
215 * fail and return EEXIST.
216 */
217 int zap_add(objset_t *ds, uint64_t zapobj, const char *key,
218 int integer_size, uint64_t num_integers,
219 const void *val, dmu_tx_t *tx);
220 int zap_add_by_dnode(dnode_t *dn, const char *key,
221 int integer_size, uint64_t num_integers,
222 const void *val, dmu_tx_t *tx);
223 int zap_add_uint64(objset_t *ds, uint64_t zapobj, const uint64_t *key,
224 int key_numints, int integer_size, uint64_t num_integers,
225 const void *val, dmu_tx_t *tx);
226
227 /*
228 * Set the attribute with the given name to the given value. If an
229 * attribute with the given name does not exist, it will be created. If
230 * an attribute with the given name already exists, the previous value
231 * will be overwritten. The integer_size may be different from the
232 * existing attribute's integer size, in which case the attribute's
233 * integer size will be updated to the new value.
234 */
235 int zap_update(objset_t *ds, uint64_t zapobj, const char *name,
236 int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
237 int zap_update_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
238 int key_numints,
239 int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
240
241 /*
242 * Get the length (in integers) and the integer size of the specified
243 * attribute.
244 *
245 * If the requested attribute does not exist, the call will fail and
246 * return ENOENT.
247 */
248 int zap_length(objset_t *ds, uint64_t zapobj, const char *name,
249 uint64_t *integer_size, uint64_t *num_integers);
250 int zap_length_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
251 int key_numints, uint64_t *integer_size, uint64_t *num_integers);
252
253 /*
254 * Remove the specified attribute.
255 *
256 * If the specified attribute does not exist, the call will fail and
257 * return ENOENT.
258 */
259 int zap_remove(objset_t *ds, uint64_t zapobj, const char *name, dmu_tx_t *tx);
260 int zap_remove_norm(objset_t *ds, uint64_t zapobj, const char *name,
261 matchtype_t mt, dmu_tx_t *tx);
262 int zap_remove_by_dnode(dnode_t *dn, const char *name, dmu_tx_t *tx);
263 int zap_remove_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
264 int key_numints, dmu_tx_t *tx);
265
266 /*
267 * Returns (in *count) the number of attributes in the specified zap
268 * object.
269 */
270 int zap_count(objset_t *ds, uint64_t zapobj, uint64_t *count);
271
272 /*
273 * Returns (in name) the name of the entry whose (value & mask)
274 * (za_first_integer) is value, or ENOENT if not found. The string
275 * pointed to by name must be at least 256 bytes long. If mask==0, the
276 * match must be exact (ie, same as mask=-1ULL).
277 */
278 int zap_value_search(objset_t *os, uint64_t zapobj,
279 uint64_t value, uint64_t mask, char *name);
280
281 /*
282 * Transfer all the entries from fromobj into intoobj. Only works on
283 * int_size=8 num_integers=1 values. Fails if there are any duplicated
284 * entries.
285 */
286 int zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx);
287
288 /* Same as zap_join, but set the values to 'value'. */
289 int zap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj,
290 uint64_t value, dmu_tx_t *tx, boolean_t exists_ok);
291
292 /* Same as zap_join, but add together any duplicated entries. */
293 int zap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj,
294 dmu_tx_t *tx);
295
296 /*
297 * Manipulate entries where the name + value are the "same" (the name is
298 * a stringified version of the value).
299 */
300 int zap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
301 int zap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
302 int zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value);
303 int zap_increment_int(objset_t *os, uint64_t obj, uint64_t key, int64_t delta,
304 dmu_tx_t *tx);
305
306 /* Here the key is an int and the value is a different int. */
307 int zap_add_int_key(objset_t *os, uint64_t obj,
308 uint64_t key, uint64_t value, dmu_tx_t *tx);
309 int zap_update_int_key(objset_t *os, uint64_t obj,
310 uint64_t key, uint64_t value, dmu_tx_t *tx);
311 int zap_lookup_int_key(objset_t *os, uint64_t obj,
312 uint64_t key, uint64_t *valuep);
313
314 int zap_increment(objset_t *os, uint64_t obj, const char *name, int64_t delta,
315 dmu_tx_t *tx);
316
317 struct zap;
318 struct zap_leaf;
319 typedef struct zap_cursor {
320 /* This structure is opaque! */
321 objset_t *zc_objset;
322 struct zap *zc_zap;
323 struct zap_leaf *zc_leaf;
324 uint64_t zc_zapobj;
325 uint64_t zc_serialized;
326 uint64_t zc_hash;
327 uint32_t zc_cd;
328 } zap_cursor_t;
329
330 typedef struct {
331 int za_integer_length;
332 /*
333 * za_normalization_conflict will be set if there are additional
334 * entries with this normalized form (eg, "foo" and "Foo").
335 */
336 boolean_t za_normalization_conflict;
337 uint64_t za_num_integers;
338 uint64_t za_first_integer; /* no sign extension for <8byte ints */
339 char za_name[ZAP_MAXNAMELEN];
340 } zap_attribute_t;
341
342 /*
343 * The interface for listing all the attributes of a zapobj can be
344 * thought of as cursor moving down a list of the attributes one by
345 * one. The cookie returned by the zap_cursor_serialize routine is
346 * persistent across system calls (and across reboot, even).
347 */
348
349 /*
350 * Initialize a zap cursor, pointing to the "first" attribute of the
351 * zapobj. You must _fini the cursor when you are done with it.
352 */
353 void zap_cursor_init(zap_cursor_t *zc, objset_t *ds, uint64_t zapobj);
354 void zap_cursor_fini(zap_cursor_t *zc);
355
356 /*
357 * Get the attribute currently pointed to by the cursor. Returns
358 * ENOENT if at the end of the attributes.
359 */
360 int zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za);
361
362 /*
363 * Advance the cursor to the next attribute.
364 */
365 void zap_cursor_advance(zap_cursor_t *zc);
366
367 /*
368 * Get a persistent cookie pointing to the current position of the zap
369 * cursor. The low 4 bits in the cookie are always zero, and thus can
370 * be used as to differentiate a serialized cookie from a different type
371 * of value. The cookie will be less than 2^32 as long as there are
372 * fewer than 2^22 (4.2 million) entries in the zap object.
373 */
374 uint64_t zap_cursor_serialize(zap_cursor_t *zc);
375
376 /*
377 * Initialize a zap cursor pointing to the position recorded by
378 * zap_cursor_serialize (in the "serialized" argument). You can also
379 * use a "serialized" argument of 0 to start at the beginning of the
380 * zapobj (ie. zap_cursor_init_serialized(..., 0) is equivalent to
381 * zap_cursor_init(...).)
382 */
383 void zap_cursor_init_serialized(zap_cursor_t *zc, objset_t *ds,
384 uint64_t zapobj, uint64_t serialized);
385
386
387 #define ZAP_HISTOGRAM_SIZE 10
388
389 typedef struct zap_stats {
390 /*
391 * Size of the pointer table (in number of entries).
392 * This is always a power of 2, or zero if it's a microzap.
393 * In general, it should be considerably greater than zs_num_leafs.
394 */
395 uint64_t zs_ptrtbl_len;
396
397 uint64_t zs_blocksize; /* size of zap blocks */
398
399 /*
400 * The number of blocks used. Note that some blocks may be
401 * wasted because old ptrtbl's and large name/value blocks are
402 * not reused. (Although their space is reclaimed, we don't
403 * reuse those offsets in the object.)
404 */
405 uint64_t zs_num_blocks;
406
407 /*
408 * Pointer table values from zap_ptrtbl in the zap_phys_t
409 */
410 uint64_t zs_ptrtbl_nextblk; /* next (larger) copy start block */
411 uint64_t zs_ptrtbl_blks_copied; /* number source blocks copied */
412 uint64_t zs_ptrtbl_zt_blk; /* starting block number */
413 uint64_t zs_ptrtbl_zt_numblks; /* number of blocks */
414 uint64_t zs_ptrtbl_zt_shift; /* bits to index it */
415
416 /*
417 * Values of the other members of the zap_phys_t
418 */
419 uint64_t zs_block_type; /* ZBT_HEADER */
420 uint64_t zs_magic; /* ZAP_MAGIC */
421 uint64_t zs_num_leafs; /* The number of leaf blocks */
422 uint64_t zs_num_entries; /* The number of zap entries */
423 uint64_t zs_salt; /* salt to stir into hash function */
424
425 /*
426 * Histograms. For all histograms, the last index
427 * (ZAP_HISTOGRAM_SIZE-1) includes any values which are greater
428 * than what can be represented. For example
429 * zs_leafs_with_n5_entries[ZAP_HISTOGRAM_SIZE-1] is the number
430 * of leafs with more than 45 entries.
431 */
432
433 /*
434 * zs_leafs_with_n_pointers[n] is the number of leafs with
435 * 2^n pointers to it.
436 */
437 uint64_t zs_leafs_with_2n_pointers[ZAP_HISTOGRAM_SIZE];
438
439 /*
440 * zs_leafs_with_n_entries[n] is the number of leafs with
441 * [n*5, (n+1)*5) entries. In the current implementation, there
442 * can be at most 55 entries in any block, but there may be
443 * fewer if the name or value is large, or the block is not
444 * completely full.
445 */
446 uint64_t zs_blocks_with_n5_entries[ZAP_HISTOGRAM_SIZE];
447
448 /*
449 * zs_leafs_n_tenths_full[n] is the number of leafs whose
450 * fullness is in the range [n/10, (n+1)/10).
451 */
452 uint64_t zs_blocks_n_tenths_full[ZAP_HISTOGRAM_SIZE];
453
454 /*
455 * zs_entries_using_n_chunks[n] is the number of entries which
456 * consume n 24-byte chunks. (Note, large names/values only use
457 * one chunk, but contribute to zs_num_blocks_large.)
458 */
459 uint64_t zs_entries_using_n_chunks[ZAP_HISTOGRAM_SIZE];
460
461 /*
462 * zs_buckets_with_n_entries[n] is the number of buckets (each
463 * leaf has 64 buckets) with n entries.
464 * zs_buckets_with_n_entries[1] should be very close to
465 * zs_num_entries.
466 */
467 uint64_t zs_buckets_with_n_entries[ZAP_HISTOGRAM_SIZE];
468 } zap_stats_t;
469
470 /*
471 * Get statistics about a ZAP object. Note: you need to be aware of the
472 * internal implementation of the ZAP to correctly interpret some of the
473 * statistics. This interface shouldn't be relied on unless you really
474 * know what you're doing.
475 */
476 int zap_get_stats(objset_t *ds, uint64_t zapobj, zap_stats_t *zs);
477
478 #ifdef __cplusplus
479 }
480 #endif
481
482 #endif /* _SYS_ZAP_H */