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