xref: /titanic_50/usr/src/uts/common/fs/zfs/sys/zap.h (revision a4e4e13f4001644f2f960e3be0056c22b3a40fd1)
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 2007 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #ifndef	_SYS_ZAP_H
27 #define	_SYS_ZAP_H
28 
29 #pragma ident	"%Z%%M%	%I%	%E% SMI"
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 #define	ZAP_MAXNAMELEN 256
91 #define	ZAP_MAXVALUELEN 1024
92 
93 /*
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
101  */
102 typedef enum matchtype
103 {
104 	MT_EXACT,
105 	MT_BEST,
106 	MT_FIRST
107 } matchtype_t;
108 
109 /*
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.
113  *
114  * normflags specifies what normalization will be done.  values are:
115  * 0: no normalization (legacy on-disk format, supports MT_EXACT matching
116  *     only)
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.
121  */
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);
126 
127 /*
128  * Create a new zapobj with no attributes from the given (unallocated)
129  * object number.
130  */
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);
136 
137 /*
138  * The zapobj passed in must be a valid ZAP object for all of the
139  * following routines.
140  */
141 
142 /*
143  * Destroy this zapobj and all its attributes.
144  *
145  * Frees the object number using dmu_object_free.
146  */
147 int zap_destroy(objset_t *ds, uint64_t zapobj, dmu_tx_t *tx);
148 
149 /*
150  * Manipulate attributes.
151  *
152  * 'integer_size' is in bytes, and must be 1, 2, 4, or 8.
153  */
154 
155 /*
156  * Retrieve the contents of the attribute with the given name.
157  *
158  * If the requested attribute does not exist, the call will fail and
159  * return ENOENT.
160  *
161  * If 'integer_size' is smaller than the attribute's integer size, the
162  * call will fail and return EINVAL.
163  *
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).
168  *
169  * 'num_integers' is the length (in integers) of 'buf'.
170  *
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.
174  *
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
177  * not MT_EXACT).
178  *
179  * If normalization_conflictp is not NULL, it will be set if there is
180  * another name with the same case/unicode normalized form.
181  */
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);
188 
189 /*
190  * Create an attribute with the given name and value.
191  *
192  * If an attribute with the given name already exists, the call will
193  * fail and return EEXIST.
194  */
195 int zap_add(objset_t *ds, uint64_t zapobj, const char *name,
196     int integer_size, uint64_t num_integers,
197     const void *val, dmu_tx_t *tx);
198 
199 /*
200  * Set the attribute with the given name to the given value.  If an
201  * attribute with the given name does not exist, it will be created.  If
202  * an attribute with the given name already exists, the previous value
203  * will be overwritten.  The integer_size may be different from the
204  * existing attribute's integer size, in which case the attribute's
205  * integer size will be updated to the new value.
206  */
207 int zap_update(objset_t *ds, uint64_t zapobj, const char *name,
208     int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
209 
210 /*
211  * Get the length (in integers) and the integer size of the specified
212  * attribute.
213  *
214  * If the requested attribute does not exist, the call will fail and
215  * return ENOENT.
216  */
217 int zap_length(objset_t *ds, uint64_t zapobj, const char *name,
218     uint64_t *integer_size, uint64_t *num_integers);
219 
220 /*
221  * Remove the specified attribute.
222  *
223  * If the specified attribute does not exist, the call will fail and
224  * return ENOENT.
225  */
226 int zap_remove(objset_t *ds, uint64_t zapobj, const char *name, dmu_tx_t *tx);
227 int zap_remove_norm(objset_t *ds, uint64_t zapobj, const char *name,
228     matchtype_t mt, dmu_tx_t *tx);
229 
230 /*
231  * Returns (in *count) the number of attributes in the specified zap
232  * object.
233  */
234 int zap_count(objset_t *ds, uint64_t zapobj, uint64_t *count);
235 
236 
237 /*
238  * Returns (in name) the name of the entry whose (value & mask)
239  * (za_first_integer) is value, or ENOENT if not found.  The string
240  * pointed to by name must be at least 256 bytes long.  If mask==0, the
241  * match must be exact (ie, same as mask=-1ULL).
242  */
243 int zap_value_search(objset_t *os, uint64_t zapobj,
244     uint64_t value, uint64_t mask, char *name);
245 
246 struct zap;
247 struct zap_leaf;
248 typedef struct zap_cursor {
249 	/* This structure is opaque! */
250 	objset_t *zc_objset;
251 	struct zap *zc_zap;
252 	struct zap_leaf *zc_leaf;
253 	uint64_t zc_zapobj;
254 	uint64_t zc_hash;
255 	uint32_t zc_cd;
256 } zap_cursor_t;
257 
258 typedef struct {
259 	int za_integer_length;
260 	/*
261 	 * za_normalization_conflict will be set if there are additional
262 	 * entries with this normalized form (eg, "foo" and "Foo").
263 	 */
264 	boolean_t za_normalization_conflict;
265 	uint64_t za_num_integers;
266 	uint64_t za_first_integer;	/* no sign extension for <8byte ints */
267 	char za_name[MAXNAMELEN];
268 } zap_attribute_t;
269 
270 /*
271  * The interface for listing all the attributes of a zapobj can be
272  * thought of as cursor moving down a list of the attributes one by
273  * one.  The cookie returned by the zap_cursor_serialize routine is
274  * persistent across system calls (and across reboot, even).
275  */
276 
277 /*
278  * Initialize a zap cursor, pointing to the "first" attribute of the
279  * zapobj.  You must _fini the cursor when you are done with it.
280  */
281 void zap_cursor_init(zap_cursor_t *zc, objset_t *ds, uint64_t zapobj);
282 void zap_cursor_fini(zap_cursor_t *zc);
283 
284 /*
285  * Get the attribute currently pointed to by the cursor.  Returns
286  * ENOENT if at the end of the attributes.
287  */
288 int zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za);
289 
290 /*
291  * Advance the cursor to the next attribute.
292  */
293 void zap_cursor_advance(zap_cursor_t *zc);
294 
295 /*
296  * Get a persistent cookie pointing to the current position of the zap
297  * cursor.  The low 4 bits in the cookie are always zero, and thus can
298  * be used as to differentiate a serialized cookie from a different type
299  * of value.  The cookie will be less than 2^32 as long as there are
300  * fewer than 2^22 (4.2 million) entries in the zap object.
301  */
302 uint64_t zap_cursor_serialize(zap_cursor_t *zc);
303 
304 /*
305  * Initialize a zap cursor pointing to the position recorded by
306  * zap_cursor_serialize (in the "serialized" argument).  You can also
307  * use a "serialized" argument of 0 to start at the beginning of the
308  * zapobj (ie.  zap_cursor_init_serialized(..., 0) is equivalent to
309  * zap_cursor_init(...).)
310  */
311 void zap_cursor_init_serialized(zap_cursor_t *zc, objset_t *ds,
312     uint64_t zapobj, uint64_t serialized);
313 
314 
315 #define	ZAP_HISTOGRAM_SIZE 10
316 
317 typedef struct zap_stats {
318 	/*
319 	 * Size of the pointer table (in number of entries).
320 	 * This is always a power of 2, or zero if it's a microzap.
321 	 * In general, it should be considerably greater than zs_num_leafs.
322 	 */
323 	uint64_t zs_ptrtbl_len;
324 
325 	uint64_t zs_blocksize;		/* size of zap blocks */
326 
327 	/*
328 	 * The number of blocks used.  Note that some blocks may be
329 	 * wasted because old ptrtbl's and large name/value blocks are
330 	 * not reused.  (Although their space is reclaimed, we don't
331 	 * reuse those offsets in the object.)
332 	 */
333 	uint64_t zs_num_blocks;
334 
335 	/*
336 	 * Pointer table values from zap_ptrtbl in the zap_phys_t
337 	 */
338 	uint64_t zs_ptrtbl_nextblk;	  /* next (larger) copy start block */
339 	uint64_t zs_ptrtbl_blks_copied;   /* number source blocks copied */
340 	uint64_t zs_ptrtbl_zt_blk;	  /* starting block number */
341 	uint64_t zs_ptrtbl_zt_numblks;    /* number of blocks */
342 	uint64_t zs_ptrtbl_zt_shift;	  /* bits to index it */
343 
344 	/*
345 	 * Values of the other members of the zap_phys_t
346 	 */
347 	uint64_t zs_block_type;		/* ZBT_HEADER */
348 	uint64_t zs_magic;		/* ZAP_MAGIC */
349 	uint64_t zs_num_leafs;		/* The number of leaf blocks */
350 	uint64_t zs_num_entries;	/* The number of zap entries */
351 	uint64_t zs_salt;		/* salt to stir into hash function */
352 
353 	/*
354 	 * Histograms.  For all histograms, the last index
355 	 * (ZAP_HISTOGRAM_SIZE-1) includes any values which are greater
356 	 * than what can be represented.  For example
357 	 * zs_leafs_with_n5_entries[ZAP_HISTOGRAM_SIZE-1] is the number
358 	 * of leafs with more than 45 entries.
359 	 */
360 
361 	/*
362 	 * zs_leafs_with_n_pointers[n] is the number of leafs with
363 	 * 2^n pointers to it.
364 	 */
365 	uint64_t zs_leafs_with_2n_pointers[ZAP_HISTOGRAM_SIZE];
366 
367 	/*
368 	 * zs_leafs_with_n_entries[n] is the number of leafs with
369 	 * [n*5, (n+1)*5) entries.  In the current implementation, there
370 	 * can be at most 55 entries in any block, but there may be
371 	 * fewer if the name or value is large, or the block is not
372 	 * completely full.
373 	 */
374 	uint64_t zs_blocks_with_n5_entries[ZAP_HISTOGRAM_SIZE];
375 
376 	/*
377 	 * zs_leafs_n_tenths_full[n] is the number of leafs whose
378 	 * fullness is in the range [n/10, (n+1)/10).
379 	 */
380 	uint64_t zs_blocks_n_tenths_full[ZAP_HISTOGRAM_SIZE];
381 
382 	/*
383 	 * zs_entries_using_n_chunks[n] is the number of entries which
384 	 * consume n 24-byte chunks.  (Note, large names/values only use
385 	 * one chunk, but contribute to zs_num_blocks_large.)
386 	 */
387 	uint64_t zs_entries_using_n_chunks[ZAP_HISTOGRAM_SIZE];
388 
389 	/*
390 	 * zs_buckets_with_n_entries[n] is the number of buckets (each
391 	 * leaf has 64 buckets) with n entries.
392 	 * zs_buckets_with_n_entries[1] should be very close to
393 	 * zs_num_entries.
394 	 */
395 	uint64_t zs_buckets_with_n_entries[ZAP_HISTOGRAM_SIZE];
396 } zap_stats_t;
397 
398 /*
399  * Get statistics about a ZAP object.  Note: you need to be aware of the
400  * internal implementation of the ZAP to correctly interpret some of the
401  * statistics.  This interface shouldn't be relied on unless you really
402  * know what you're doing.
403  */
404 int zap_get_stats(objset_t *ds, uint64_t zapobj, zap_stats_t *zs);
405 
406 #ifdef	__cplusplus
407 }
408 #endif
409 
410 #endif	/* _SYS_ZAP_H */
411