xref: /freebsd/sys/contrib/openzfs/include/sys/zap.h (revision 783d3ff6)
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 https://opensource.org/licenses/CDDL-1.0.
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, 2018 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 
86 #ifdef	__cplusplus
87 extern "C" {
88 #endif
89 
90 /*
91  * Specifies matching criteria for ZAP lookups.
92  * MT_NORMALIZE		Use ZAP normalization flags, which can include both
93  *			unicode normalization and case-insensitivity.
94  * MT_MATCH_CASE	Do case-sensitive lookups even if MT_NORMALIZE is
95  *			specified and ZAP normalization flags include
96  *			U8_TEXTPREP_TOUPPER.
97  */
98 typedef enum matchtype {
99 	MT_NORMALIZE = 1 << 0,
100 	MT_MATCH_CASE = 1 << 1,
101 } matchtype_t;
102 
103 typedef enum zap_flags {
104 	/* Use 64-bit hash value (serialized cursors will always use 64-bits) */
105 	ZAP_FLAG_HASH64 = 1 << 0,
106 	/* Key is binary, not string (zap_add_uint64() can be used) */
107 	ZAP_FLAG_UINT64_KEY = 1 << 1,
108 	/*
109 	 * First word of key (which must be an array of uint64) is
110 	 * already randomly distributed.
111 	 */
112 	ZAP_FLAG_PRE_HASHED_KEY = 1 << 2,
113 #if defined(__linux__) && defined(_KERNEL)
114 } zfs_zap_flags_t;
115 #define	zap_flags_t	zfs_zap_flags_t
116 #else
117 } zap_flags_t;
118 #endif
119 
120 /*
121  * Create a new zapobj with no attributes and return its object number.
122  */
123 uint64_t zap_create(objset_t *ds, dmu_object_type_t ot,
124     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
125 uint64_t zap_create_dnsize(objset_t *ds, dmu_object_type_t ot,
126     dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *tx);
127 uint64_t zap_create_norm(objset_t *ds, int normflags, dmu_object_type_t ot,
128     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
129 uint64_t zap_create_norm_dnsize(objset_t *ds, int normflags,
130     dmu_object_type_t ot, dmu_object_type_t bonustype, int bonuslen,
131     int dnodesize, dmu_tx_t *tx);
132 uint64_t zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
133     dmu_object_type_t ot, int leaf_blockshift, int indirect_blockshift,
134     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
135 uint64_t zap_create_flags_dnsize(objset_t *os, int normflags,
136     zap_flags_t flags, dmu_object_type_t ot, int leaf_blockshift,
137     int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen,
138     int dnodesize, dmu_tx_t *tx);
139 uint64_t zap_create_hold(objset_t *os, int normflags, zap_flags_t flags,
140     dmu_object_type_t ot, int leaf_blockshift, int indirect_blockshift,
141     dmu_object_type_t bonustype, int bonuslen, int dnodesize,
142     dnode_t **allocated_dnode, const void *tag, dmu_tx_t *tx);
143 
144 uint64_t zap_create_link(objset_t *os, dmu_object_type_t ot,
145     uint64_t parent_obj, const char *name, dmu_tx_t *tx);
146 uint64_t zap_create_link_dnsize(objset_t *os, dmu_object_type_t ot,
147     uint64_t parent_obj, const char *name, int dnodesize, dmu_tx_t *tx);
148 
149 /*
150  * Initialize an already-allocated object.
151  */
152 void mzap_create_impl(dnode_t *dn, int normflags, zap_flags_t flags,
153     dmu_tx_t *tx);
154 
155 /*
156  * Create a new zapobj with no attributes from the given (unallocated)
157  * object number.
158  */
159 int zap_create_claim(objset_t *ds, uint64_t obj, dmu_object_type_t ot,
160     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
161 int zap_create_claim_dnsize(objset_t *ds, uint64_t obj, dmu_object_type_t ot,
162     dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *tx);
163 int zap_create_claim_norm(objset_t *ds, uint64_t obj,
164     int normflags, dmu_object_type_t ot,
165     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
166 int zap_create_claim_norm_dnsize(objset_t *ds, uint64_t obj,
167     int normflags, dmu_object_type_t ot,
168     dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *tx);
169 
170 /*
171  * The zapobj passed in must be a valid ZAP object for all of the
172  * following routines.
173  */
174 
175 /*
176  * Destroy this zapobj and all its attributes.
177  *
178  * Frees the object number using dmu_object_free.
179  */
180 int zap_destroy(objset_t *ds, uint64_t zapobj, dmu_tx_t *tx);
181 
182 /*
183  * Manipulate attributes.
184  *
185  * 'integer_size' is in bytes, and must be 1, 2, 4, or 8.
186  */
187 
188 /*
189  * Retrieve the contents of the attribute with the given name.
190  *
191  * If the requested attribute does not exist, the call will fail and
192  * return ENOENT.
193  *
194  * If 'integer_size' is smaller than the attribute's integer size, the
195  * call will fail and return EINVAL.
196  *
197  * If 'integer_size' is equal to or larger than the attribute's integer
198  * size, the call will succeed and return 0.
199  *
200  * When converting to a larger integer size, the integers will be treated as
201  * unsigned (ie. no sign-extension will be performed).
202  *
203  * 'num_integers' is the length (in integers) of 'buf'.
204  *
205  * If the attribute is longer than the buffer, as many integers as will
206  * fit will be transferred to 'buf'.  If the entire attribute was not
207  * transferred, the call will return EOVERFLOW.
208  */
209 int zap_lookup(objset_t *ds, uint64_t zapobj, const char *name,
210     uint64_t integer_size, uint64_t num_integers, void *buf);
211 
212 /*
213  * If rn_len is nonzero, realname will be set to the name of the found
214  * entry (which may be different from the requested name if matchtype is
215  * not MT_EXACT).
216  *
217  * If normalization_conflictp is not NULL, it will be set if there is
218  * another name with the same case/unicode normalized form.
219  */
220 int zap_lookup_norm(objset_t *ds, uint64_t zapobj, const char *name,
221     uint64_t integer_size, uint64_t num_integers, void *buf,
222     matchtype_t mt, char *realname, int rn_len,
223     boolean_t *normalization_conflictp);
224 int zap_lookup_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
225     int key_numints, uint64_t integer_size, uint64_t num_integers, void *buf);
226 int zap_contains(objset_t *ds, uint64_t zapobj, const char *name);
227 int zap_prefetch(objset_t *os, uint64_t zapobj, const char *name);
228 int zap_prefetch_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
229     int key_numints);
230 
231 int zap_lookup_by_dnode(dnode_t *dn, const char *name,
232     uint64_t integer_size, uint64_t num_integers, void *buf);
233 int zap_lookup_norm_by_dnode(dnode_t *dn, const char *name,
234     uint64_t integer_size, uint64_t num_integers, void *buf,
235     matchtype_t mt, char *realname, int rn_len,
236     boolean_t *ncp);
237 
238 int zap_count_write_by_dnode(dnode_t *dn, const char *name,
239     int add, zfs_refcount_t *towrite, zfs_refcount_t *tooverwrite);
240 
241 /*
242  * Create an attribute with the given name and value.
243  *
244  * If an attribute with the given name already exists, the call will
245  * fail and return EEXIST.
246  */
247 int zap_add(objset_t *ds, uint64_t zapobj, const char *key,
248     int integer_size, uint64_t num_integers,
249     const void *val, dmu_tx_t *tx);
250 int zap_add_by_dnode(dnode_t *dn, const char *key,
251     int integer_size, uint64_t num_integers,
252     const void *val, dmu_tx_t *tx);
253 int zap_add_uint64(objset_t *ds, uint64_t zapobj, const uint64_t *key,
254     int key_numints, int integer_size, uint64_t num_integers,
255     const void *val, dmu_tx_t *tx);
256 int zap_add_uint64_by_dnode(dnode_t *dn, const uint64_t *key,
257     int key_numints, int integer_size, uint64_t num_integers,
258     const void *val, dmu_tx_t *tx);
259 
260 /*
261  * Set the attribute with the given name to the given value.  If an
262  * attribute with the given name does not exist, it will be created.  If
263  * an attribute with the given name already exists, the previous value
264  * will be overwritten.  The integer_size may be different from the
265  * existing attribute's integer size, in which case the attribute's
266  * integer size will be updated to the new value.
267  */
268 int zap_update(objset_t *ds, uint64_t zapobj, const char *name,
269     int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
270 int zap_update_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
271     int key_numints,
272     int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
273 int zap_update_uint64_by_dnode(dnode_t *dn, const uint64_t *key,
274     int key_numints,
275     int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
276 
277 /*
278  * Get the length (in integers) and the integer size of the specified
279  * attribute.
280  *
281  * If the requested attribute does not exist, the call will fail and
282  * return ENOENT.
283  */
284 int zap_length(objset_t *ds, uint64_t zapobj, const char *name,
285     uint64_t *integer_size, uint64_t *num_integers);
286 int zap_length_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
287     int key_numints, uint64_t *integer_size, uint64_t *num_integers);
288 
289 /*
290  * Remove the specified attribute.
291  *
292  * If the specified attribute does not exist, the call will fail and
293  * return ENOENT.
294  */
295 int zap_remove(objset_t *ds, uint64_t zapobj, const char *name, dmu_tx_t *tx);
296 int zap_remove_norm(objset_t *ds, uint64_t zapobj, const char *name,
297     matchtype_t mt, dmu_tx_t *tx);
298 int zap_remove_by_dnode(dnode_t *dn, const char *name, dmu_tx_t *tx);
299 int zap_remove_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
300     int key_numints, dmu_tx_t *tx);
301 int zap_remove_uint64_by_dnode(dnode_t *dn, const uint64_t *key,
302     int key_numints, dmu_tx_t *tx);
303 
304 /*
305  * Returns (in *count) the number of attributes in the specified zap
306  * object.
307  */
308 int zap_count(objset_t *ds, uint64_t zapobj, uint64_t *count);
309 
310 /*
311  * Returns (in name) the name of the entry whose (value & mask)
312  * (za_first_integer) is value, or ENOENT if not found.  The string
313  * pointed to by name must be at least 256 bytes long.  If mask==0, the
314  * match must be exact (ie, same as mask=-1ULL).
315  */
316 int zap_value_search(objset_t *os, uint64_t zapobj,
317     uint64_t value, uint64_t mask, char *name);
318 
319 /*
320  * Transfer all the entries from fromobj into intoobj.  Only works on
321  * int_size=8 num_integers=1 values.  Fails if there are any duplicated
322  * entries.
323  */
324 int zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx);
325 
326 /* Same as zap_join, but set the values to 'value'. */
327 int zap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj,
328     uint64_t value, dmu_tx_t *tx);
329 
330 /* Same as zap_join, but add together any duplicated entries. */
331 int zap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj,
332     dmu_tx_t *tx);
333 
334 /*
335  * Manipulate entries where the name + value are the "same" (the name is
336  * a stringified version of the value).
337  */
338 int zap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
339 int zap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
340 int zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value);
341 int zap_increment_int(objset_t *os, uint64_t obj, uint64_t key, int64_t delta,
342     dmu_tx_t *tx);
343 
344 /* Here the key is an int and the value is a different int. */
345 int zap_add_int_key(objset_t *os, uint64_t obj,
346     uint64_t key, uint64_t value, dmu_tx_t *tx);
347 int zap_update_int_key(objset_t *os, uint64_t obj,
348     uint64_t key, uint64_t value, dmu_tx_t *tx);
349 int zap_lookup_int_key(objset_t *os, uint64_t obj,
350     uint64_t key, uint64_t *valuep);
351 
352 int zap_increment(objset_t *os, uint64_t obj, const char *name, int64_t delta,
353     dmu_tx_t *tx);
354 
355 struct zap;
356 struct zap_leaf;
357 typedef struct zap_cursor {
358 	/* This structure is opaque! */
359 	objset_t *zc_objset;
360 	struct zap *zc_zap;
361 	struct zap_leaf *zc_leaf;
362 	uint64_t zc_zapobj;
363 	uint64_t zc_serialized;
364 	uint64_t zc_hash;
365 	uint32_t zc_cd;
366 	boolean_t zc_prefetch;
367 } zap_cursor_t;
368 
369 typedef struct {
370 	int za_integer_length;
371 	/*
372 	 * za_normalization_conflict will be set if there are additional
373 	 * entries with this normalized form (eg, "foo" and "Foo").
374 	 */
375 	boolean_t za_normalization_conflict;
376 	uint64_t za_num_integers;
377 	uint64_t za_first_integer;	/* no sign extension for <8byte ints */
378 	char za_name[ZAP_MAXNAMELEN];
379 } zap_attribute_t;
380 
381 /*
382  * The interface for listing all the attributes of a zapobj can be
383  * thought of as cursor moving down a list of the attributes one by
384  * one.  The cookie returned by the zap_cursor_serialize routine is
385  * persistent across system calls (and across reboot, even).
386  */
387 
388 /*
389  * Initialize a zap cursor, pointing to the "first" attribute of the
390  * zapobj.  You must _fini the cursor when you are done with it.
391  */
392 void zap_cursor_init(zap_cursor_t *zc, objset_t *os, uint64_t zapobj);
393 void zap_cursor_init_noprefetch(zap_cursor_t *zc, objset_t *os,
394     uint64_t zapobj);
395 void zap_cursor_fini(zap_cursor_t *zc);
396 
397 /*
398  * Get the attribute currently pointed to by the cursor.  Returns
399  * ENOENT if at the end of the attributes.
400  */
401 int zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za);
402 
403 /*
404  * Advance the cursor to the next attribute.
405  */
406 void zap_cursor_advance(zap_cursor_t *zc);
407 
408 /*
409  * Get a persistent cookie pointing to the current position of the zap
410  * cursor.  The low 4 bits in the cookie are always zero, and thus can
411  * be used as to differentiate a serialized cookie from a different type
412  * of value.  The cookie will be less than 2^32 as long as there are
413  * fewer than 2^22 (4.2 million) entries in the zap object.
414  */
415 uint64_t zap_cursor_serialize(zap_cursor_t *zc);
416 
417 /*
418  * Initialize a zap cursor pointing to the position recorded by
419  * zap_cursor_serialize (in the "serialized" argument).  You can also
420  * use a "serialized" argument of 0 to start at the beginning of the
421  * zapobj (ie.  zap_cursor_init_serialized(..., 0) is equivalent to
422  * zap_cursor_init(...).)
423  */
424 void zap_cursor_init_serialized(zap_cursor_t *zc, objset_t *ds,
425     uint64_t zapobj, uint64_t serialized);
426 
427 
428 #define	ZAP_HISTOGRAM_SIZE 10
429 
430 typedef struct zap_stats {
431 	/*
432 	 * Size of the pointer table (in number of entries).
433 	 * This is always a power of 2, or zero if it's a microzap.
434 	 * In general, it should be considerably greater than zs_num_leafs.
435 	 */
436 	uint64_t zs_ptrtbl_len;
437 
438 	uint64_t zs_blocksize;		/* size of zap blocks */
439 
440 	/*
441 	 * The number of blocks used.  Note that some blocks may be
442 	 * wasted because old ptrtbl's and large name/value blocks are
443 	 * not reused.  (Although their space is reclaimed, we don't
444 	 * reuse those offsets in the object.)
445 	 */
446 	uint64_t zs_num_blocks;
447 
448 	/*
449 	 * Pointer table values from zap_ptrtbl in the zap_phys_t
450 	 */
451 	uint64_t zs_ptrtbl_nextblk;	  /* next (larger) copy start block */
452 	uint64_t zs_ptrtbl_blks_copied;   /* number source blocks copied */
453 	uint64_t zs_ptrtbl_zt_blk;	  /* starting block number */
454 	uint64_t zs_ptrtbl_zt_numblks;    /* number of blocks */
455 	uint64_t zs_ptrtbl_zt_shift;	  /* bits to index it */
456 
457 	/*
458 	 * Values of the other members of the zap_phys_t
459 	 */
460 	uint64_t zs_block_type;		/* ZBT_HEADER */
461 	uint64_t zs_magic;		/* ZAP_MAGIC */
462 	uint64_t zs_num_leafs;		/* The number of leaf blocks */
463 	uint64_t zs_num_entries;	/* The number of zap entries */
464 	uint64_t zs_salt;		/* salt to stir into hash function */
465 
466 	/*
467 	 * Histograms.  For all histograms, the last index
468 	 * (ZAP_HISTOGRAM_SIZE-1) includes any values which are greater
469 	 * than what can be represented.  For example
470 	 * zs_leafs_with_n5_entries[ZAP_HISTOGRAM_SIZE-1] is the number
471 	 * of leafs with more than 45 entries.
472 	 */
473 
474 	/*
475 	 * zs_leafs_with_n_pointers[n] is the number of leafs with
476 	 * 2^n pointers to it.
477 	 */
478 	uint64_t zs_leafs_with_2n_pointers[ZAP_HISTOGRAM_SIZE];
479 
480 	/*
481 	 * zs_leafs_with_n_entries[n] is the number of leafs with
482 	 * [n*5, (n+1)*5) entries.  In the current implementation, there
483 	 * can be at most 55 entries in any block, but there may be
484 	 * fewer if the name or value is large, or the block is not
485 	 * completely full.
486 	 */
487 	uint64_t zs_blocks_with_n5_entries[ZAP_HISTOGRAM_SIZE];
488 
489 	/*
490 	 * zs_leafs_n_tenths_full[n] is the number of leafs whose
491 	 * fullness is in the range [n/10, (n+1)/10).
492 	 */
493 	uint64_t zs_blocks_n_tenths_full[ZAP_HISTOGRAM_SIZE];
494 
495 	/*
496 	 * zs_entries_using_n_chunks[n] is the number of entries which
497 	 * consume n 24-byte chunks.  (Note, large names/values only use
498 	 * one chunk, but contribute to zs_num_blocks_large.)
499 	 */
500 	uint64_t zs_entries_using_n_chunks[ZAP_HISTOGRAM_SIZE];
501 
502 	/*
503 	 * zs_buckets_with_n_entries[n] is the number of buckets (each
504 	 * leaf has 64 buckets) with n entries.
505 	 * zs_buckets_with_n_entries[1] should be very close to
506 	 * zs_num_entries.
507 	 */
508 	uint64_t zs_buckets_with_n_entries[ZAP_HISTOGRAM_SIZE];
509 } zap_stats_t;
510 
511 /*
512  * Get statistics about a ZAP object.  Note: you need to be aware of the
513  * internal implementation of the ZAP to correctly interpret some of the
514  * statistics.  This interface shouldn't be relied on unless you really
515  * know what you're doing.
516  */
517 int zap_get_stats(objset_t *ds, uint64_t zapobj, zap_stats_t *zs);
518 
519 #ifdef	__cplusplus
520 }
521 #endif
522 
523 #endif	/* _SYS_ZAP_H */
524