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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2013 by Delphix. All rights reserved.
24 */
25
26 #ifndef _SYS_ZAP_H
27 #define _SYS_ZAP_H
28
29 /*
30 * ZAP - ZFS Attribute Processor
31 *
32 * The ZAP is a module which sits on top of the DMU (Data Management
33 * Unit) and implements a higher-level storage primitive using DMU
34 * objects. Its primary consumer is the ZPL (ZFS Posix Layer).
35 *
36 * A "zapobj" is a DMU object which the ZAP uses to stores attributes.
37 * Users should use only zap routines to access a zapobj - they should
38 * not access the DMU object directly using DMU routines.
39 *
40 * The attributes stored in a zapobj are name-value pairs. The name is
41 * a zero-terminated string of up to ZAP_MAXNAMELEN bytes (including
42 * terminating NULL). The value is an array of integers, which may be
43 * 1, 2, 4, or 8 bytes long. The total space used by the array (number
44 * of integers * integer length) can be up to ZAP_MAXVALUELEN bytes.
45 * Note that an 8-byte integer value can be used to store the location
46 * (object number) of another dmu object (which may be itself a zapobj).
47 * Note that you can use a zero-length attribute to store a single bit
48 * of information - the attribute is present or not.
49 *
50 * The ZAP routines are thread-safe. However, you must observe the
51 * DMU's restriction that a transaction may not be operated on
52 * concurrently.
53 *
54 * Any of the routines that return an int may return an I/O error (EIO
55 * or ECHECKSUM).
56 *
57 *
58 * Implementation / Performance Notes:
59 *
60 * The ZAP is intended to operate most efficiently on attributes with
61 * short (49 bytes or less) names and single 8-byte values, for which
62 * the microzap will be used. The ZAP should be efficient enough so
63 * that the user does not need to cache these attributes.
64 *
65 * The ZAP's locking scheme makes its routines thread-safe. Operations
66 * on different zapobjs will be processed concurrently. Operations on
67 * the same zapobj which only read data will be processed concurrently.
68 * Operations on the same zapobj which modify data will be processed
69 * concurrently when there are many attributes in the zapobj (because
70 * the ZAP uses per-block locking - more than 128 * (number of cpus)
71 * small attributes will suffice).
72 */
73
74 /*
75 * We're using zero-terminated byte strings (ie. ASCII or UTF-8 C
76 * strings) for the names of attributes, rather than a byte string
77 * bounded by an explicit length. If some day we want to support names
78 * in character sets which have embedded zeros (eg. UTF-16, UTF-32),
79 * we'll have to add routines for using length-bounded strings.
80 */
81
82 #include <sys/dmu.h>
83
84 #ifdef __cplusplus
85 extern "C" {
86 #endif
87
88 /*
89 * Specifies matching criteria for ZAP lookups.
90 */
91 typedef enum matchtype
92 {
93 /* Only find an exact match (non-normalized) */
94 MT_EXACT,
95 /*
96 * If there is an exact match, find that, otherwise find the
97 * first normalized match.
98 */
99 MT_BEST,
100 /*
101 * Find the "first" normalized (case and Unicode form) match;
102 * the designated "first" match will not change as long as the
103 * set of entries with this normalization doesn't change.
104 */
105 MT_FIRST
106 } matchtype_t;
107
108 typedef enum zap_flags {
109 /* Use 64-bit hash value (serialized cursors will always use 64-bits) */
110 ZAP_FLAG_HASH64 = 1 << 0,
111 /* Key is binary, not string (zap_add_uint64() can be used) */
112 ZAP_FLAG_UINT64_KEY = 1 << 1,
113 /*
114 * First word of key (which must be an array of uint64) is
115 * already randomly distributed.
116 */
117 ZAP_FLAG_PRE_HASHED_KEY = 1 << 2,
118 } zap_flags_t;
119
120 /*
121 * Create a new zapobj with no attributes and return its object number.
122 * MT_EXACT will cause the zap object to only support MT_EXACT lookups,
123 * otherwise any matchtype can be used for lookups.
124 *
125 * normflags specifies what normalization will be done. values are:
126 * 0: no normalization (legacy on-disk format, supports MT_EXACT matching
127 * only)
128 * U8_TEXTPREP_TOLOWER: case normalization will be performed.
129 * MT_FIRST/MT_BEST matching will find entries that match without
130 * regard to case (eg. looking for "foo" can find an entry "Foo").
131 * Eventually, other flags will permit unicode normalization as well.
132 */
133 uint64_t zap_create(objset_t *ds, dmu_object_type_t ot,
134 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
135 uint64_t zap_create_norm(objset_t *ds, int normflags, dmu_object_type_t ot,
136 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
137 uint64_t zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
138 dmu_object_type_t ot, int leaf_blockshift, int indirect_blockshift,
139 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
140 uint64_t zap_create_link(objset_t *os, dmu_object_type_t ot,
141 uint64_t parent_obj, const char *name, dmu_tx_t *tx);
142
143 /*
144 * Create a new zapobj with no attributes from the given (unallocated)
145 * object number.
146 */
147 int zap_create_claim(objset_t *ds, uint64_t obj, dmu_object_type_t ot,
148 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
149 int zap_create_claim_norm(objset_t *ds, uint64_t obj,
150 int normflags, dmu_object_type_t ot,
151 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
152
153 /*
154 * The zapobj passed in must be a valid ZAP object for all of the
155 * following routines.
156 */
157
158 /*
159 * Destroy this zapobj and all its attributes.
160 *
161 * Frees the object number using dmu_object_free.
162 */
163 int zap_destroy(objset_t *ds, uint64_t zapobj, dmu_tx_t *tx);
164
165 /*
166 * Manipulate attributes.
167 *
168 * 'integer_size' is in bytes, and must be 1, 2, 4, or 8.
169 */
170
171 /*
172 * Retrieve the contents of the attribute with the given name.
173 *
174 * If the requested attribute does not exist, the call will fail and
175 * return ENOENT.
176 *
177 * If 'integer_size' is smaller than the attribute's integer size, the
178 * call will fail and return EINVAL.
179 *
180 * If 'integer_size' is equal to or larger than the attribute's integer
181 * size, the call will succeed and return 0.
182 *
183 * When converting to a larger integer size, the integers will be treated as
184 * unsigned (ie. no sign-extension will be performed).
185 *
186 * 'num_integers' is the length (in integers) of 'buf'.
187 *
188 * If the attribute is longer than the buffer, as many integers as will
189 * fit will be transferred to 'buf'. If the entire attribute was not
190 * transferred, the call will return EOVERFLOW.
191 */
192 int zap_lookup(objset_t *ds, uint64_t zapobj, const char *name,
193 uint64_t integer_size, uint64_t num_integers, void *buf);
194
195 /*
196 * If rn_len is nonzero, realname will be set to the name of the found
197 * entry (which may be different from the requested name if matchtype is
198 * not MT_EXACT).
199 *
200 * If normalization_conflictp is not NULL, it will be set if there is
201 * another name with the same case/unicode normalized form.
202 */
203 int zap_lookup_norm(objset_t *ds, uint64_t zapobj, const char *name,
204 uint64_t integer_size, uint64_t num_integers, void *buf,
205 matchtype_t mt, char *realname, int rn_len,
206 boolean_t *normalization_conflictp);
207 int zap_lookup_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
208 int key_numints, uint64_t integer_size, uint64_t num_integers, void *buf);
209 int zap_contains(objset_t *ds, uint64_t zapobj, const char *name);
210 int zap_prefetch_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
211 int key_numints);
212
213 int zap_count_write(objset_t *os, uint64_t zapobj, const char *name,
214 int add, uint64_t *towrite, uint64_t *tooverwrite);
215
216 /*
217 * Create an attribute with the given name and value.
218 *
219 * If an attribute with the given name already exists, the call will
220 * fail and return EEXIST.
221 */
222 int zap_add(objset_t *ds, uint64_t zapobj, const char *key,
223 int integer_size, uint64_t num_integers,
224 const void *val, dmu_tx_t *tx);
225 int zap_add_uint64(objset_t *ds, uint64_t zapobj, const uint64_t *key,
226 int key_numints, int integer_size, uint64_t num_integers,
227 const void *val, dmu_tx_t *tx);
228
229 /*
230 * Set the attribute with the given name to the given value. If an
231 * attribute with the given name does not exist, it will be created. If
232 * an attribute with the given name already exists, the previous value
233 * will be overwritten. The integer_size may be different from the
234 * existing attribute's integer size, in which case the attribute's
235 * integer size will be updated to the new value.
236 */
237 int zap_update(objset_t *ds, uint64_t zapobj, const char *name,
238 int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
239 int zap_update_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
240 int key_numints,
241 int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
242
243 /*
244 * Get the length (in integers) and the integer size of the specified
245 * attribute.
246 *
247 * If the requested attribute does not exist, the call will fail and
248 * return ENOENT.
249 */
250 int zap_length(objset_t *ds, uint64_t zapobj, const char *name,
251 uint64_t *integer_size, uint64_t *num_integers);
252 int zap_length_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
253 int key_numints, uint64_t *integer_size, uint64_t *num_integers);
254
255 /*
256 * Remove the specified attribute.
257 *
258 * If the specified attribute does not exist, the call will fail and
259 * return ENOENT.
260 */
261 int zap_remove(objset_t *ds, uint64_t zapobj, const char *name, dmu_tx_t *tx);
262 int zap_remove_norm(objset_t *ds, uint64_t zapobj, const char *name,
263 matchtype_t mt, dmu_tx_t *tx);
264 int zap_remove_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
265 int key_numints, dmu_tx_t *tx);
266
267 /*
268 * Returns (in *count) the number of attributes in the specified zap
269 * object.
270 */
271 int zap_count(objset_t *ds, uint64_t zapobj, uint64_t *count);
272
273 /*
274 * Returns (in name) the name of the entry whose (value & mask)
275 * (za_first_integer) is value, or ENOENT if not found. The string
276 * pointed to by name must be at least 256 bytes long. If mask==0, the
277 * match must be exact (ie, same as mask=-1ULL).
278 */
279 int zap_value_search(objset_t *os, uint64_t zapobj,
280 uint64_t value, uint64_t mask, char *name);
281
282 /*
283 * Transfer all the entries from fromobj into intoobj. Only works on
284 * int_size=8 num_integers=1 values. Fails if there are any duplicated
285 * entries.
286 */
287 int zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx);
288
289 /* Same as zap_join, but set the values to 'value'. */
290 int zap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj,
291 uint64_t value, dmu_tx_t *tx);
292
293 /* Same as zap_join, but add together any duplicated entries. */
294 int zap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj,
295 dmu_tx_t *tx);
296
297 /*
298 * Manipulate entries where the name + value are the "same" (the name is
299 * a stringified version of the value).
300 */
301 int zap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
302 int zap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
303 int zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value);
304 int zap_increment_int(objset_t *os, uint64_t obj, uint64_t key, int64_t delta,
305 dmu_tx_t *tx);
306
307 /* Here the key is an int and the value is a different int. */
308 int zap_add_int_key(objset_t *os, uint64_t obj,
309 uint64_t key, uint64_t value, dmu_tx_t *tx);
310 int zap_update_int_key(objset_t *os, uint64_t obj,
311 uint64_t key, uint64_t value, dmu_tx_t *tx);
312 int zap_lookup_int_key(objset_t *os, uint64_t obj,
313 uint64_t key, uint64_t *valuep);
314
315 int zap_increment(objset_t *os, uint64_t obj, const char *name, int64_t delta,
316 dmu_tx_t *tx);
317
318 struct zap;
319 struct zap_leaf;
320 typedef struct zap_cursor {
321 /* This structure is opaque! */
322 objset_t *zc_objset;
323 struct zap *zc_zap;
324 struct zap_leaf *zc_leaf;
325 uint64_t zc_zapobj;
326 uint64_t zc_serialized;
327 uint64_t zc_hash;
328 uint32_t zc_cd;
329 } zap_cursor_t;
330
331 typedef struct {
332 int za_integer_length;
333 /*
334 * za_normalization_conflict will be set if there are additional
335 * entries with this normalized form (eg, "foo" and "Foo").
336 */
337 boolean_t za_normalization_conflict;
338 uint64_t za_num_integers;
339 uint64_t za_first_integer; /* no sign extension for <8byte ints */
340 char za_name[MAXNAMELEN];
341 } zap_attribute_t;
342
343 /*
344 * The interface for listing all the attributes of a zapobj can be
345 * thought of as cursor moving down a list of the attributes one by
346 * one. The cookie returned by the zap_cursor_serialize routine is
347 * persistent across system calls (and across reboot, even).
348 */
349
350 /*
351 * Initialize a zap cursor, pointing to the "first" attribute of the
352 * zapobj. You must _fini the cursor when you are done with it.
353 */
354 void zap_cursor_init(zap_cursor_t *zc, objset_t *ds, uint64_t zapobj);
355 void zap_cursor_fini(zap_cursor_t *zc);
356
357 /*
358 * Get the attribute currently pointed to by the cursor. Returns
359 * ENOENT if at the end of the attributes.
360 */
361 int zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za);
362
363 /*
364 * Advance the cursor to the next attribute.
365 */
366 void zap_cursor_advance(zap_cursor_t *zc);
367
368 /*
369 * Get a persistent cookie pointing to the current position of the zap
370 * cursor. The low 4 bits in the cookie are always zero, and thus can
371 * be used as to differentiate a serialized cookie from a different type
372 * of value. The cookie will be less than 2^32 as long as there are
373 * fewer than 2^22 (4.2 million) entries in the zap object.
374 */
375 uint64_t zap_cursor_serialize(zap_cursor_t *zc);
376
377 /*
378 * Initialize a zap cursor pointing to the position recorded by
379 * zap_cursor_serialize (in the "serialized" argument). You can also
380 * use a "serialized" argument of 0 to start at the beginning of the
381 * zapobj (ie. zap_cursor_init_serialized(..., 0) is equivalent to
382 * zap_cursor_init(...).)
383 */
384 void zap_cursor_init_serialized(zap_cursor_t *zc, objset_t *ds,
385 uint64_t zapobj, uint64_t serialized);
386
387
388 #define ZAP_HISTOGRAM_SIZE 10
389
390 typedef struct zap_stats {
391 /*
392 * Size of the pointer table (in number of entries).
393 * This is always a power of 2, or zero if it's a microzap.
394 * In general, it should be considerably greater than zs_num_leafs.
395 */
396 uint64_t zs_ptrtbl_len;
397
398 uint64_t zs_blocksize; /* size of zap blocks */
399
400 /*
401 * The number of blocks used. Note that some blocks may be
402 * wasted because old ptrtbl's and large name/value blocks are
403 * not reused. (Although their space is reclaimed, we don't
404 * reuse those offsets in the object.)
405 */
406 uint64_t zs_num_blocks;
407
408 /*
409 * Pointer table values from zap_ptrtbl in the zap_phys_t
410 */
411 uint64_t zs_ptrtbl_nextblk; /* next (larger) copy start block */
412 uint64_t zs_ptrtbl_blks_copied; /* number source blocks copied */
413 uint64_t zs_ptrtbl_zt_blk; /* starting block number */
414 uint64_t zs_ptrtbl_zt_numblks; /* number of blocks */
415 uint64_t zs_ptrtbl_zt_shift; /* bits to index it */
416
417 /*
418 * Values of the other members of the zap_phys_t
419 */
420 uint64_t zs_block_type; /* ZBT_HEADER */
421 uint64_t zs_magic; /* ZAP_MAGIC */
422 uint64_t zs_num_leafs; /* The number of leaf blocks */
423 uint64_t zs_num_entries; /* The number of zap entries */
424 uint64_t zs_salt; /* salt to stir into hash function */
425
426 /*
427 * Histograms. For all histograms, the last index
428 * (ZAP_HISTOGRAM_SIZE-1) includes any values which are greater
429 * than what can be represented. For example
430 * zs_leafs_with_n5_entries[ZAP_HISTOGRAM_SIZE-1] is the number
431 * of leafs with more than 45 entries.
432 */
433
434 /*
435 * zs_leafs_with_n_pointers[n] is the number of leafs with
436 * 2^n pointers to it.
437 */
438 uint64_t zs_leafs_with_2n_pointers[ZAP_HISTOGRAM_SIZE];
439
440 /*
441 * zs_leafs_with_n_entries[n] is the number of leafs with
442 * [n*5, (n+1)*5) entries. In the current implementation, there
443 * can be at most 55 entries in any block, but there may be
444 * fewer if the name or value is large, or the block is not
445 * completely full.
446 */
447 uint64_t zs_blocks_with_n5_entries[ZAP_HISTOGRAM_SIZE];
448
449 /*
450 * zs_leafs_n_tenths_full[n] is the number of leafs whose
451 * fullness is in the range [n/10, (n+1)/10).
452 */
453 uint64_t zs_blocks_n_tenths_full[ZAP_HISTOGRAM_SIZE];
454
455 /*
456 * zs_entries_using_n_chunks[n] is the number of entries which
457 * consume n 24-byte chunks. (Note, large names/values only use
458 * one chunk, but contribute to zs_num_blocks_large.)
459 */
460 uint64_t zs_entries_using_n_chunks[ZAP_HISTOGRAM_SIZE];
461
462 /*
463 * zs_buckets_with_n_entries[n] is the number of buckets (each
464 * leaf has 64 buckets) with n entries.
465 * zs_buckets_with_n_entries[1] should be very close to
466 * zs_num_entries.
467 */
468 uint64_t zs_buckets_with_n_entries[ZAP_HISTOGRAM_SIZE];
469 } zap_stats_t;
470
471 /*
472 * Get statistics about a ZAP object. Note: you need to be aware of the
473 * internal implementation of the ZAP to correctly interpret some of the
474 * statistics. This interface shouldn't be relied on unless you really
475 * know what you're doing.
476 */
477 int zap_get_stats(objset_t *ds, uint64_t zapobj, zap_stats_t *zs);
478
479 #ifdef __cplusplus
480 }
481 #endif
482
483 #endif /* _SYS_ZAP_H */