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.
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.
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]
22 * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
29 #pragma ident "%Z%%M% %I% %E% SMI"
32 * ZAP - ZFS Attribute Processor
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).
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.
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.
52 * The ZAP routines are thread-safe. However, you must observe the
53 * DMU's restriction that a transaction may not be operated on
56 * Any of the routines that return an int may return an I/O error (EIO
60 * Implementation / Performance Notes:
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.
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).
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.
90 #define ZAP_MAXNAMELEN 256
91 #define ZAP_MAXVALUELEN 1024
94 * The matchtype specifies which entry will be accessed.
95 * MT_EXACT: only find an exact match (non-normalized)
96 * MT_FIRST: find the "first" normalized (case and Unicode
97 * form) match; the designated "first" match will not change as long
98 * as the set of entries with this normalization doesn't change
99 * MT_BEST: if there is an exact match, find that, otherwise find the
100 * first normalized match
102 typedef enum matchtype
110 * Create a new zapobj with no attributes and return its object number.
111 * MT_EXACT will cause the zap object to only support MT_EXACT lookups,
112 * otherwise any matchtype can be used for lookups.
114 * normflags specifies what normalization will be done. values are:
115 * 0: no normalization (legacy on-disk format, supports MT_EXACT matching
117 * U8_TEXTPREP_TOLOWER: case normalization will be performed.
118 * MT_FIRST/MT_BEST matching will find entries that match without
119 * regard to case (eg. looking for "foo" can find an entry "Foo").
120 * Eventually, other flags will permit unicode normalization as well.
122 uint64_t zap_create(objset_t *ds, dmu_object_type_t ot,
123 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
124 uint64_t zap_create_norm(objset_t *ds, int normflags, dmu_object_type_t ot,
125 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
128 * Create a new zapobj with no attributes from the given (unallocated)
131 int zap_create_claim(objset_t *ds, uint64_t obj, dmu_object_type_t ot,
132 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
133 int zap_create_claim_norm(objset_t *ds, uint64_t obj,
134 int normflags, dmu_object_type_t ot,
135 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
138 * The zapobj passed in must be a valid ZAP object for all of the
139 * following routines.
143 * Destroy this zapobj and all its attributes.
145 * Frees the object number using dmu_object_free.
147 int zap_destroy(objset_t *ds, uint64_t zapobj, dmu_tx_t *tx);
150 * Manipulate attributes.
152 * 'integer_size' is in bytes, and must be 1, 2, 4, or 8.
156 * Retrieve the contents of the attribute with the given name.
158 * If the requested attribute does not exist, the call will fail and
161 * If 'integer_size' is smaller than the attribute's integer size, the
162 * call will fail and return EINVAL.
164 * If 'integer_size' is equal to or larger than the attribute's integer
165 * size, the call will succeed and return 0. * When converting to a
166 * larger integer size, the integers will be treated as unsigned (ie. no
167 * sign-extension will be performed).
169 * 'num_integers' is the length (in integers) of 'buf'.
171 * If the attribute is longer than the buffer, as many integers as will
172 * fit will be transferred to 'buf'. If the entire attribute was not
173 * transferred, the call will return EOVERFLOW.
175 * If rn_len is nonzero, realname will be set to the name of the found
176 * entry (which may be different from the requested name if matchtype is
179 * If normalization_conflictp is not NULL, it will be set if there is
180 * another name with the same case/unicode normalized form.
182 int zap_lookup(objset_t *ds, uint64_t zapobj, const char *name,
183 uint64_t integer_size, uint64_t num_integers, void *buf);
184 int zap_lookup_norm(objset_t *ds, uint64_t zapobj, const char *name,
185 uint64_t integer_size, uint64_t num_integers, void *buf,
186 matchtype_t mt, char *realname, int rn_len,
187 boolean_t *normalization_conflictp);
189 int zap_count_write(objset_t *os, uint64_t zapobj, const char *name,
190 int add, uint64_t *towrite, uint64_t *tooverwrite);
193 * Create an attribute with the given name and value.
195 * If an attribute with the given name already exists, the call will
196 * fail and return EEXIST.
198 int zap_add(objset_t *ds, uint64_t zapobj, const char *name,
199 int integer_size, uint64_t num_integers,
200 const void *val, dmu_tx_t *tx);
203 * Set the attribute with the given name to the given value. If an
204 * attribute with the given name does not exist, it will be created. If
205 * an attribute with the given name already exists, the previous value
206 * will be overwritten. The integer_size may be different from the
207 * existing attribute's integer size, in which case the attribute's
208 * integer size will be updated to the new value.
210 int zap_update(objset_t *ds, uint64_t zapobj, const char *name,
211 int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
214 * Get the length (in integers) and the integer size of the specified
217 * If the requested attribute does not exist, the call will fail and
220 int zap_length(objset_t *ds, uint64_t zapobj, const char *name,
221 uint64_t *integer_size, uint64_t *num_integers);
224 * Remove the specified attribute.
226 * If the specified attribute does not exist, the call will fail and
229 int zap_remove(objset_t *ds, uint64_t zapobj, const char *name, dmu_tx_t *tx);
230 int zap_remove_norm(objset_t *ds, uint64_t zapobj, const char *name,
231 matchtype_t mt, dmu_tx_t *tx);
234 * Returns (in *count) the number of attributes in the specified zap
237 int zap_count(objset_t *ds, uint64_t zapobj, uint64_t *count);
241 * Returns (in name) the name of the entry whose (value & mask)
242 * (za_first_integer) is value, or ENOENT if not found. The string
243 * pointed to by name must be at least 256 bytes long. If mask==0, the
244 * match must be exact (ie, same as mask=-1ULL).
246 int zap_value_search(objset_t *os, uint64_t zapobj,
247 uint64_t value, uint64_t mask, char *name);
250 * Transfer all the entries from fromobj into intoobj. Only works on
251 * int_size=8 num_integers=1 values. Fails if there are any duplicated
254 int zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx);
257 * Manipulate entries where the name + value are the "same" (the name is
258 * a stringified version of the value).
260 int zap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
261 int zap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
262 int zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value);
266 typedef struct zap_cursor {
267 /* This structure is opaque! */
270 struct zap_leaf *zc_leaf;
277 int za_integer_length;
279 * za_normalization_conflict will be set if there are additional
280 * entries with this normalized form (eg, "foo" and "Foo").
282 boolean_t za_normalization_conflict;
283 uint64_t za_num_integers;
284 uint64_t za_first_integer; /* no sign extension for <8byte ints */
285 char za_name[MAXNAMELEN];
289 * The interface for listing all the attributes of a zapobj can be
290 * thought of as cursor moving down a list of the attributes one by
291 * one. The cookie returned by the zap_cursor_serialize routine is
292 * persistent across system calls (and across reboot, even).
296 * Initialize a zap cursor, pointing to the "first" attribute of the
297 * zapobj. You must _fini the cursor when you are done with it.
299 void zap_cursor_init(zap_cursor_t *zc, objset_t *ds, uint64_t zapobj);
300 void zap_cursor_fini(zap_cursor_t *zc);
303 * Get the attribute currently pointed to by the cursor. Returns
304 * ENOENT if at the end of the attributes.
306 int zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za);
309 * Advance the cursor to the next attribute.
311 void zap_cursor_advance(zap_cursor_t *zc);
314 * Get a persistent cookie pointing to the current position of the zap
315 * cursor. The low 4 bits in the cookie are always zero, and thus can
316 * be used as to differentiate a serialized cookie from a different type
317 * of value. The cookie will be less than 2^32 as long as there are
318 * fewer than 2^22 (4.2 million) entries in the zap object.
320 uint64_t zap_cursor_serialize(zap_cursor_t *zc);
323 * Initialize a zap cursor pointing to the position recorded by
324 * zap_cursor_serialize (in the "serialized" argument). You can also
325 * use a "serialized" argument of 0 to start at the beginning of the
326 * zapobj (ie. zap_cursor_init_serialized(..., 0) is equivalent to
327 * zap_cursor_init(...).)
329 void zap_cursor_init_serialized(zap_cursor_t *zc, objset_t *ds,
330 uint64_t zapobj, uint64_t serialized);
333 #define ZAP_HISTOGRAM_SIZE 10
335 typedef struct zap_stats {
337 * Size of the pointer table (in number of entries).
338 * This is always a power of 2, or zero if it's a microzap.
339 * In general, it should be considerably greater than zs_num_leafs.
341 uint64_t zs_ptrtbl_len;
343 uint64_t zs_blocksize; /* size of zap blocks */
346 * The number of blocks used. Note that some blocks may be
347 * wasted because old ptrtbl's and large name/value blocks are
348 * not reused. (Although their space is reclaimed, we don't
349 * reuse those offsets in the object.)
351 uint64_t zs_num_blocks;
354 * Pointer table values from zap_ptrtbl in the zap_phys_t
356 uint64_t zs_ptrtbl_nextblk; /* next (larger) copy start block */
357 uint64_t zs_ptrtbl_blks_copied; /* number source blocks copied */
358 uint64_t zs_ptrtbl_zt_blk; /* starting block number */
359 uint64_t zs_ptrtbl_zt_numblks; /* number of blocks */
360 uint64_t zs_ptrtbl_zt_shift; /* bits to index it */
363 * Values of the other members of the zap_phys_t
365 uint64_t zs_block_type; /* ZBT_HEADER */
366 uint64_t zs_magic; /* ZAP_MAGIC */
367 uint64_t zs_num_leafs; /* The number of leaf blocks */
368 uint64_t zs_num_entries; /* The number of zap entries */
369 uint64_t zs_salt; /* salt to stir into hash function */
372 * Histograms. For all histograms, the last index
373 * (ZAP_HISTOGRAM_SIZE-1) includes any values which are greater
374 * than what can be represented. For example
375 * zs_leafs_with_n5_entries[ZAP_HISTOGRAM_SIZE-1] is the number
376 * of leafs with more than 45 entries.
380 * zs_leafs_with_n_pointers[n] is the number of leafs with
381 * 2^n pointers to it.
383 uint64_t zs_leafs_with_2n_pointers[ZAP_HISTOGRAM_SIZE];
386 * zs_leafs_with_n_entries[n] is the number of leafs with
387 * [n*5, (n+1)*5) entries. In the current implementation, there
388 * can be at most 55 entries in any block, but there may be
389 * fewer if the name or value is large, or the block is not
392 uint64_t zs_blocks_with_n5_entries[ZAP_HISTOGRAM_SIZE];
395 * zs_leafs_n_tenths_full[n] is the number of leafs whose
396 * fullness is in the range [n/10, (n+1)/10).
398 uint64_t zs_blocks_n_tenths_full[ZAP_HISTOGRAM_SIZE];
401 * zs_entries_using_n_chunks[n] is the number of entries which
402 * consume n 24-byte chunks. (Note, large names/values only use
403 * one chunk, but contribute to zs_num_blocks_large.)
405 uint64_t zs_entries_using_n_chunks[ZAP_HISTOGRAM_SIZE];
408 * zs_buckets_with_n_entries[n] is the number of buckets (each
409 * leaf has 64 buckets) with n entries.
410 * zs_buckets_with_n_entries[1] should be very close to
413 uint64_t zs_buckets_with_n_entries[ZAP_HISTOGRAM_SIZE];
417 * Get statistics about a ZAP object. Note: you need to be aware of the
418 * internal implementation of the ZAP to correctly interpret some of the
419 * statistics. This interface shouldn't be relied on unless you really
420 * know what you're doing.
422 int zap_get_stats(objset_t *ds, uint64_t zapobj, zap_stats_t *zs);
428 #endif /* _SYS_ZAP_H */