xref: /titanic_52/usr/src/uts/common/fs/zfs/sys/zap.h (revision 40a5c998e5462ed47916e2edfcc5016073cf8851)
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) 2012, 2015 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  * Initialize an already-allocated object.
145  */
146 void 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  */
153 int 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);
155 int zap_create_claim_norm(objset_t *ds, uint64_t obj,
156     int normflags, dmu_object_type_t ot,
157     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
158 
159 /*
160  * The zapobj passed in must be a valid ZAP object for all of the
161  * following routines.
162  */
163 
164 /*
165  * Destroy this zapobj and all its attributes.
166  *
167  * Frees the object number using dmu_object_free.
168  */
169 int zap_destroy(objset_t *ds, uint64_t zapobj, dmu_tx_t *tx);
170 
171 /*
172  * Manipulate attributes.
173  *
174  * 'integer_size' is in bytes, and must be 1, 2, 4, or 8.
175  */
176 
177 /*
178  * Retrieve the contents of the attribute with the given name.
179  *
180  * If the requested attribute does not exist, the call will fail and
181  * return ENOENT.
182  *
183  * If 'integer_size' is smaller than the attribute's integer size, the
184  * call will fail and return EINVAL.
185  *
186  * If 'integer_size' is equal to or larger than the attribute's integer
187  * size, the call will succeed and return 0.
188  *
189  * When converting to a larger integer size, the integers will be treated as
190  * unsigned (ie. no sign-extension will be performed).
191  *
192  * 'num_integers' is the length (in integers) of 'buf'.
193  *
194  * If the attribute is longer than the buffer, as many integers as will
195  * fit will be transferred to 'buf'.  If the entire attribute was not
196  * transferred, the call will return EOVERFLOW.
197  */
198 int zap_lookup(objset_t *ds, uint64_t zapobj, const char *name,
199     uint64_t integer_size, uint64_t num_integers, void *buf);
200 
201 /*
202  * If rn_len is nonzero, realname will be set to the name of the found
203  * entry (which may be different from the requested name if matchtype is
204  * not MT_EXACT).
205  *
206  * If normalization_conflictp is not NULL, it will be set if there is
207  * another name with the same case/unicode normalized form.
208  */
209 int zap_lookup_norm(objset_t *ds, uint64_t zapobj, const char *name,
210     uint64_t integer_size, uint64_t num_integers, void *buf,
211     matchtype_t mt, char *realname, int rn_len,
212     boolean_t *normalization_conflictp);
213 int zap_lookup_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
214     int key_numints, uint64_t integer_size, uint64_t num_integers, void *buf);
215 int zap_contains(objset_t *ds, uint64_t zapobj, const char *name);
216 int zap_prefetch_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
217     int key_numints);
218 
219 int zap_count_write(objset_t *os, uint64_t zapobj, const char *name,
220     int add, uint64_t *towrite, uint64_t *tooverwrite);
221 
222 /*
223  * Create an attribute with the given name and value.
224  *
225  * If an attribute with the given name already exists, the call will
226  * fail and return EEXIST.
227  */
228 int zap_add(objset_t *ds, uint64_t zapobj, const char *key,
229     int integer_size, uint64_t num_integers,
230     const void *val, dmu_tx_t *tx);
231 int zap_add_uint64(objset_t *ds, uint64_t zapobj, const uint64_t *key,
232     int key_numints, int integer_size, uint64_t num_integers,
233     const void *val, dmu_tx_t *tx);
234 
235 /*
236  * Set the attribute with the given name to the given value.  If an
237  * attribute with the given name does not exist, it will be created.  If
238  * an attribute with the given name already exists, the previous value
239  * will be overwritten.  The integer_size may be different from the
240  * existing attribute's integer size, in which case the attribute's
241  * integer size will be updated to the new value.
242  */
243 int zap_update(objset_t *ds, uint64_t zapobj, const char *name,
244     int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
245 int zap_update_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
246     int key_numints,
247     int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
248 
249 /*
250  * Get the length (in integers) and the integer size of the specified
251  * attribute.
252  *
253  * If the requested attribute does not exist, the call will fail and
254  * return ENOENT.
255  */
256 int zap_length(objset_t *ds, uint64_t zapobj, const char *name,
257     uint64_t *integer_size, uint64_t *num_integers);
258 int zap_length_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
259     int key_numints, uint64_t *integer_size, uint64_t *num_integers);
260 
261 /*
262  * Remove the specified attribute.
263  *
264  * If the specified attribute does not exist, the call will fail and
265  * return ENOENT.
266  */
267 int zap_remove(objset_t *ds, uint64_t zapobj, const char *name, dmu_tx_t *tx);
268 int zap_remove_norm(objset_t *ds, uint64_t zapobj, const char *name,
269     matchtype_t mt, dmu_tx_t *tx);
270 int zap_remove_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
271     int key_numints, dmu_tx_t *tx);
272 
273 /*
274  * Returns (in *count) the number of attributes in the specified zap
275  * object.
276  */
277 int zap_count(objset_t *ds, uint64_t zapobj, uint64_t *count);
278 
279 /*
280  * Returns (in name) the name of the entry whose (value & mask)
281  * (za_first_integer) is value, or ENOENT if not found.  The string
282  * pointed to by name must be at least 256 bytes long.  If mask==0, the
283  * match must be exact (ie, same as mask=-1ULL).
284  */
285 int zap_value_search(objset_t *os, uint64_t zapobj,
286     uint64_t value, uint64_t mask, char *name);
287 
288 /*
289  * Transfer all the entries from fromobj into intoobj.  Only works on
290  * int_size=8 num_integers=1 values.  Fails if there are any duplicated
291  * entries.
292  */
293 int zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx);
294 
295 /* Same as zap_join, but set the values to 'value'. */
296 int zap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj,
297     uint64_t value, dmu_tx_t *tx);
298 
299 /* Same as zap_join, but add together any duplicated entries. */
300 int zap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj,
301     dmu_tx_t *tx);
302 
303 /*
304  * Manipulate entries where the name + value are the "same" (the name is
305  * a stringified version of the value).
306  */
307 int zap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
308 int zap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
309 int zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value);
310 int zap_increment_int(objset_t *os, uint64_t obj, uint64_t key, int64_t delta,
311     dmu_tx_t *tx);
312 
313 /* Here the key is an int and the value is a different int. */
314 int zap_add_int_key(objset_t *os, uint64_t obj,
315     uint64_t key, uint64_t value, dmu_tx_t *tx);
316 int zap_update_int_key(objset_t *os, uint64_t obj,
317     uint64_t key, uint64_t value, dmu_tx_t *tx);
318 int zap_lookup_int_key(objset_t *os, uint64_t obj,
319     uint64_t key, uint64_t *valuep);
320 
321 int zap_increment(objset_t *os, uint64_t obj, const char *name, int64_t delta,
322     dmu_tx_t *tx);
323 
324 struct zap;
325 struct zap_leaf;
326 typedef struct zap_cursor {
327 	/* This structure is opaque! */
328 	objset_t *zc_objset;
329 	struct zap *zc_zap;
330 	struct zap_leaf *zc_leaf;
331 	uint64_t zc_zapobj;
332 	uint64_t zc_serialized;
333 	uint64_t zc_hash;
334 	uint32_t zc_cd;
335 } zap_cursor_t;
336 
337 typedef struct {
338 	int za_integer_length;
339 	/*
340 	 * za_normalization_conflict will be set if there are additional
341 	 * entries with this normalized form (eg, "foo" and "Foo").
342 	 */
343 	boolean_t za_normalization_conflict;
344 	uint64_t za_num_integers;
345 	uint64_t za_first_integer;	/* no sign extension for <8byte ints */
346 	char za_name[ZAP_MAXNAMELEN];
347 } zap_attribute_t;
348 
349 /*
350  * The interface for listing all the attributes of a zapobj can be
351  * thought of as cursor moving down a list of the attributes one by
352  * one.  The cookie returned by the zap_cursor_serialize routine is
353  * persistent across system calls (and across reboot, even).
354  */
355 
356 /*
357  * Initialize a zap cursor, pointing to the "first" attribute of the
358  * zapobj.  You must _fini the cursor when you are done with it.
359  */
360 void zap_cursor_init(zap_cursor_t *zc, objset_t *ds, uint64_t zapobj);
361 void zap_cursor_fini(zap_cursor_t *zc);
362 
363 /*
364  * Get the attribute currently pointed to by the cursor.  Returns
365  * ENOENT if at the end of the attributes.
366  */
367 int zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za);
368 
369 /*
370  * Advance the cursor to the next attribute.
371  */
372 void zap_cursor_advance(zap_cursor_t *zc);
373 
374 /*
375  * Get a persistent cookie pointing to the current position of the zap
376  * cursor.  The low 4 bits in the cookie are always zero, and thus can
377  * be used as to differentiate a serialized cookie from a different type
378  * of value.  The cookie will be less than 2^32 as long as there are
379  * fewer than 2^22 (4.2 million) entries in the zap object.
380  */
381 uint64_t zap_cursor_serialize(zap_cursor_t *zc);
382 
383 /*
384  * Initialize a zap cursor pointing to the position recorded by
385  * zap_cursor_serialize (in the "serialized" argument).  You can also
386  * use a "serialized" argument of 0 to start at the beginning of the
387  * zapobj (ie.  zap_cursor_init_serialized(..., 0) is equivalent to
388  * zap_cursor_init(...).)
389  */
390 void zap_cursor_init_serialized(zap_cursor_t *zc, objset_t *ds,
391     uint64_t zapobj, uint64_t serialized);
392 
393 
394 #define	ZAP_HISTOGRAM_SIZE 10
395 
396 typedef struct zap_stats {
397 	/*
398 	 * Size of the pointer table (in number of entries).
399 	 * This is always a power of 2, or zero if it's a microzap.
400 	 * In general, it should be considerably greater than zs_num_leafs.
401 	 */
402 	uint64_t zs_ptrtbl_len;
403 
404 	uint64_t zs_blocksize;		/* size of zap blocks */
405 
406 	/*
407 	 * The number of blocks used.  Note that some blocks may be
408 	 * wasted because old ptrtbl's and large name/value blocks are
409 	 * not reused.  (Although their space is reclaimed, we don't
410 	 * reuse those offsets in the object.)
411 	 */
412 	uint64_t zs_num_blocks;
413 
414 	/*
415 	 * Pointer table values from zap_ptrtbl in the zap_phys_t
416 	 */
417 	uint64_t zs_ptrtbl_nextblk;	  /* next (larger) copy start block */
418 	uint64_t zs_ptrtbl_blks_copied;   /* number source blocks copied */
419 	uint64_t zs_ptrtbl_zt_blk;	  /* starting block number */
420 	uint64_t zs_ptrtbl_zt_numblks;    /* number of blocks */
421 	uint64_t zs_ptrtbl_zt_shift;	  /* bits to index it */
422 
423 	/*
424 	 * Values of the other members of the zap_phys_t
425 	 */
426 	uint64_t zs_block_type;		/* ZBT_HEADER */
427 	uint64_t zs_magic;		/* ZAP_MAGIC */
428 	uint64_t zs_num_leafs;		/* The number of leaf blocks */
429 	uint64_t zs_num_entries;	/* The number of zap entries */
430 	uint64_t zs_salt;		/* salt to stir into hash function */
431 
432 	/*
433 	 * Histograms.  For all histograms, the last index
434 	 * (ZAP_HISTOGRAM_SIZE-1) includes any values which are greater
435 	 * than what can be represented.  For example
436 	 * zs_leafs_with_n5_entries[ZAP_HISTOGRAM_SIZE-1] is the number
437 	 * of leafs with more than 45 entries.
438 	 */
439 
440 	/*
441 	 * zs_leafs_with_n_pointers[n] is the number of leafs with
442 	 * 2^n pointers to it.
443 	 */
444 	uint64_t zs_leafs_with_2n_pointers[ZAP_HISTOGRAM_SIZE];
445 
446 	/*
447 	 * zs_leafs_with_n_entries[n] is the number of leafs with
448 	 * [n*5, (n+1)*5) entries.  In the current implementation, there
449 	 * can be at most 55 entries in any block, but there may be
450 	 * fewer if the name or value is large, or the block is not
451 	 * completely full.
452 	 */
453 	uint64_t zs_blocks_with_n5_entries[ZAP_HISTOGRAM_SIZE];
454 
455 	/*
456 	 * zs_leafs_n_tenths_full[n] is the number of leafs whose
457 	 * fullness is in the range [n/10, (n+1)/10).
458 	 */
459 	uint64_t zs_blocks_n_tenths_full[ZAP_HISTOGRAM_SIZE];
460 
461 	/*
462 	 * zs_entries_using_n_chunks[n] is the number of entries which
463 	 * consume n 24-byte chunks.  (Note, large names/values only use
464 	 * one chunk, but contribute to zs_num_blocks_large.)
465 	 */
466 	uint64_t zs_entries_using_n_chunks[ZAP_HISTOGRAM_SIZE];
467 
468 	/*
469 	 * zs_buckets_with_n_entries[n] is the number of buckets (each
470 	 * leaf has 64 buckets) with n entries.
471 	 * zs_buckets_with_n_entries[1] should be very close to
472 	 * zs_num_entries.
473 	 */
474 	uint64_t zs_buckets_with_n_entries[ZAP_HISTOGRAM_SIZE];
475 } zap_stats_t;
476 
477 /*
478  * Get statistics about a ZAP object.  Note: you need to be aware of the
479  * internal implementation of the ZAP to correctly interpret some of the
480  * statistics.  This interface shouldn't be relied on unless you really
481  * know what you're doing.
482  */
483 int zap_get_stats(objset_t *ds, uint64_t zapobj, zap_stats_t *zs);
484 
485 #ifdef	__cplusplus
486 }
487 #endif
488 
489 #endif	/* _SYS_ZAP_H */
490