Home | History | Annotate | Download | only in sys
      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 2009 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 /*
     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  * The matchtype specifies which entry will be accessed.
     90  * MT_EXACT: only find an exact match (non-normalized)
     91  * MT_FIRST: find the "first" normalized (case and Unicode
     92  *     form) match; the designated "first" match will not change as long
     93  *     as the set of entries with this normalization doesn't change
     94  * MT_BEST: if there is an exact match, find that, otherwise find the
     95  *     first normalized match
     96  */
     97 typedef enum matchtype
     98 {
     99 	MT_EXACT,
    100 	MT_BEST,
    101 	MT_FIRST
    102 } matchtype_t;
    103 
    104 typedef 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  * MT_EXACT will cause the zap object to only support MT_EXACT lookups,
    119  * otherwise any matchtype can be used for lookups.
    120  *
    121  * normflags specifies what normalization will be done.  values are:
    122  * 0: no normalization (legacy on-disk format, supports MT_EXACT matching
    123  *     only)
    124  * U8_TEXTPREP_TOLOWER: case normalization will be performed.
    125  *     MT_FIRST/MT_BEST matching will find entries that match without
    126  *     regard to case (eg. looking for "foo" can find an entry "Foo").
    127  * Eventually, other flags will permit unicode normalization as well.
    128  */
    129 uint64_t zap_create(objset_t *ds, dmu_object_type_t ot,
    130     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
    131 uint64_t zap_create_norm(objset_t *ds, int normflags, dmu_object_type_t ot,
    132     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
    133 uint64_t zap_create_flags(objset_t *os, int normflags, zap_flags_t flags,
    134     dmu_object_type_t ot, int leaf_blockshift, int indirect_blockshift,
    135     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
    136 
    137 /*
    138  * Create a new zapobj with no attributes from the given (unallocated)
    139  * object number.
    140  */
    141 int zap_create_claim(objset_t *ds, uint64_t obj, dmu_object_type_t ot,
    142     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
    143 int zap_create_claim_norm(objset_t *ds, uint64_t obj,
    144     int normflags, dmu_object_type_t ot,
    145     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
    146 
    147 /*
    148  * The zapobj passed in must be a valid ZAP object for all of the
    149  * following routines.
    150  */
    151 
    152 /*
    153  * Destroy this zapobj and all its attributes.
    154  *
    155  * Frees the object number using dmu_object_free.
    156  */
    157 int zap_destroy(objset_t *ds, uint64_t zapobj, dmu_tx_t *tx);
    158 
    159 /*
    160  * Manipulate attributes.
    161  *
    162  * 'integer_size' is in bytes, and must be 1, 2, 4, or 8.
    163  */
    164 
    165 /*
    166  * Retrieve the contents of the attribute with the given name.
    167  *
    168  * If the requested attribute does not exist, the call will fail and
    169  * return ENOENT.
    170  *
    171  * If 'integer_size' is smaller than the attribute's integer size, the
    172  * call will fail and return EINVAL.
    173  *
    174  * If 'integer_size' is equal to or larger than the attribute's integer
    175  * size, the call will succeed and return 0.  * When converting to a
    176  * larger integer size, the integers will be treated as unsigned (ie. no
    177  * sign-extension will be performed).
    178  *
    179  * 'num_integers' is the length (in integers) of 'buf'.
    180  *
    181  * If the attribute is longer than the buffer, as many integers as will
    182  * fit will be transferred to 'buf'.  If the entire attribute was not
    183  * transferred, the call will return EOVERFLOW.
    184  *
    185  * If rn_len is nonzero, realname will be set to the name of the found
    186  * entry (which may be different from the requested name if matchtype is
    187  * not MT_EXACT).
    188  *
    189  * If normalization_conflictp is not NULL, it will be set if there is
    190  * another name with the same case/unicode normalized form.
    191  */
    192 int zap_lookup(objset_t *ds, uint64_t zapobj, const char *name,
    193     uint64_t integer_size, uint64_t num_integers, void *buf);
    194 int zap_lookup_norm(objset_t *ds, uint64_t zapobj, const char *name,
    195     uint64_t integer_size, uint64_t num_integers, void *buf,
    196     matchtype_t mt, char *realname, int rn_len,
    197     boolean_t *normalization_conflictp);
    198 int zap_lookup_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
    199     int key_numints, uint64_t integer_size, uint64_t num_integers, void *buf);
    200 int zap_contains(objset_t *ds, uint64_t zapobj, const char *name);
    201 
    202 int zap_count_write(objset_t *os, uint64_t zapobj, const char *name,
    203     int add, uint64_t *towrite, uint64_t *tooverwrite);
    204 
    205 /*
    206  * Create an attribute with the given name and value.
    207  *
    208  * If an attribute with the given name already exists, the call will
    209  * fail and return EEXIST.
    210  */
    211 int zap_add(objset_t *ds, uint64_t zapobj, const char *key,
    212     int integer_size, uint64_t num_integers,
    213     const void *val, dmu_tx_t *tx);
    214 int zap_add_uint64(objset_t *ds, uint64_t zapobj, const uint64_t *key,
    215     int key_numints, int integer_size, uint64_t num_integers,
    216     const void *val, dmu_tx_t *tx);
    217 
    218 /*
    219  * Set the attribute with the given name to the given value.  If an
    220  * attribute with the given name does not exist, it will be created.  If
    221  * an attribute with the given name already exists, the previous value
    222  * will be overwritten.  The integer_size may be different from the
    223  * existing attribute's integer size, in which case the attribute's
    224  * integer size will be updated to the new value.
    225  */
    226 int zap_update(objset_t *ds, uint64_t zapobj, const char *name,
    227     int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
    228 int zap_update_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
    229     int key_numints,
    230     int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx);
    231 
    232 /*
    233  * Get the length (in integers) and the integer size of the specified
    234  * attribute.
    235  *
    236  * If the requested attribute does not exist, the call will fail and
    237  * return ENOENT.
    238  */
    239 int zap_length(objset_t *ds, uint64_t zapobj, const char *name,
    240     uint64_t *integer_size, uint64_t *num_integers);
    241 int zap_length_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
    242     int key_numints, uint64_t *integer_size, uint64_t *num_integers);
    243 
    244 /*
    245  * Remove the specified attribute.
    246  *
    247  * If the specified attribute does not exist, the call will fail and
    248  * return ENOENT.
    249  */
    250 int zap_remove(objset_t *ds, uint64_t zapobj, const char *name, dmu_tx_t *tx);
    251 int zap_remove_norm(objset_t *ds, uint64_t zapobj, const char *name,
    252     matchtype_t mt, dmu_tx_t *tx);
    253 int zap_remove_uint64(objset_t *os, uint64_t zapobj, const uint64_t *key,
    254     int key_numints, dmu_tx_t *tx);
    255 
    256 /*
    257  * Returns (in *count) the number of attributes in the specified zap
    258  * object.
    259  */
    260 int zap_count(objset_t *ds, uint64_t zapobj, uint64_t *count);
    261 
    262 
    263 /*
    264  * Returns (in name) the name of the entry whose (value & mask)
    265  * (za_first_integer) is value, or ENOENT if not found.  The string
    266  * pointed to by name must be at least 256 bytes long.  If mask==0, the
    267  * match must be exact (ie, same as mask=-1ULL).
    268  */
    269 int zap_value_search(objset_t *os, uint64_t zapobj,
    270     uint64_t value, uint64_t mask, char *name);
    271 
    272 /*
    273  * Transfer all the entries from fromobj into intoobj.  Only works on
    274  * int_size=8 num_integers=1 values.  Fails if there are any duplicated
    275  * entries.
    276  */
    277 int zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx);
    278 
    279 /*
    280  * Manipulate entries where the name + value are the "same" (the name is
    281  * a stringified version of the value).
    282  */
    283 int zap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
    284 int zap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx);
    285 int zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value);
    286 int zap_increment_int(objset_t *os, uint64_t obj, uint64_t key, int64_t delta,
    287     dmu_tx_t *tx);
    288 
    289 struct zap;
    290 struct zap_leaf;
    291 typedef struct zap_cursor {
    292 	/* This structure is opaque! */
    293 	objset_t *zc_objset;
    294 	struct zap *zc_zap;
    295 	struct zap_leaf *zc_leaf;
    296 	uint64_t zc_zapobj;
    297 	uint64_t zc_serialized;
    298 	uint64_t zc_hash;
    299 	uint32_t zc_cd;
    300 } zap_cursor_t;
    301 
    302 typedef struct {
    303 	int za_integer_length;
    304 	/*
    305 	 * za_normalization_conflict will be set if there are additional
    306 	 * entries with this normalized form (eg, "foo" and "Foo").
    307 	 */
    308 	boolean_t za_normalization_conflict;
    309 	uint64_t za_num_integers;
    310 	uint64_t za_first_integer;	/* no sign extension for <8byte ints */
    311 	char za_name[MAXNAMELEN];
    312 } zap_attribute_t;
    313 
    314 /*
    315  * The interface for listing all the attributes of a zapobj can be
    316  * thought of as cursor moving down a list of the attributes one by
    317  * one.  The cookie returned by the zap_cursor_serialize routine is
    318  * persistent across system calls (and across reboot, even).
    319  */
    320 
    321 /*
    322  * Initialize a zap cursor, pointing to the "first" attribute of the
    323  * zapobj.  You must _fini the cursor when you are done with it.
    324  */
    325 void zap_cursor_init(zap_cursor_t *zc, objset_t *ds, uint64_t zapobj);
    326 void zap_cursor_fini(zap_cursor_t *zc);
    327 
    328 /*
    329  * Get the attribute currently pointed to by the cursor.  Returns
    330  * ENOENT if at the end of the attributes.
    331  */
    332 int zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za);
    333 
    334 /*
    335  * Advance the cursor to the next attribute.
    336  */
    337 void zap_cursor_advance(zap_cursor_t *zc);
    338 
    339 /*
    340  * Get a persistent cookie pointing to the current position of the zap
    341  * cursor.  The low 4 bits in the cookie are always zero, and thus can
    342  * be used as to differentiate a serialized cookie from a different type
    343  * of value.  The cookie will be less than 2^32 as long as there are
    344  * fewer than 2^22 (4.2 million) entries in the zap object.
    345  */
    346 uint64_t zap_cursor_serialize(zap_cursor_t *zc);
    347 
    348 /*
    349  * Advance the cursor to the attribute having the given key.
    350  */
    351 int zap_cursor_move_to_key(zap_cursor_t *zc, const char *name, matchtype_t mt);
    352 
    353 /*
    354  * Initialize a zap cursor pointing to the position recorded by
    355  * zap_cursor_serialize (in the "serialized" argument).  You can also
    356  * use a "serialized" argument of 0 to start at the beginning of the
    357  * zapobj (ie.  zap_cursor_init_serialized(..., 0) is equivalent to
    358  * zap_cursor_init(...).)
    359  */
    360 void zap_cursor_init_serialized(zap_cursor_t *zc, objset_t *ds,
    361     uint64_t zapobj, uint64_t serialized);
    362 
    363 
    364 #define	ZAP_HISTOGRAM_SIZE 10
    365 
    366 typedef struct zap_stats {
    367 	/*
    368 	 * Size of the pointer table (in number of entries).
    369 	 * This is always a power of 2, or zero if it's a microzap.
    370 	 * In general, it should be considerably greater than zs_num_leafs.
    371 	 */
    372 	uint64_t zs_ptrtbl_len;
    373 
    374 	uint64_t zs_blocksize;		/* size of zap blocks */
    375 
    376 	/*
    377 	 * The number of blocks used.  Note that some blocks may be
    378 	 * wasted because old ptrtbl's and large name/value blocks are
    379 	 * not reused.  (Although their space is reclaimed, we don't
    380 	 * reuse those offsets in the object.)
    381 	 */
    382 	uint64_t zs_num_blocks;
    383 
    384 	/*
    385 	 * Pointer table values from zap_ptrtbl in the zap_phys_t
    386 	 */
    387 	uint64_t zs_ptrtbl_nextblk;	  /* next (larger) copy start block */
    388 	uint64_t zs_ptrtbl_blks_copied;   /* number source blocks copied */
    389 	uint64_t zs_ptrtbl_zt_blk;	  /* starting block number */
    390 	uint64_t zs_ptrtbl_zt_numblks;    /* number of blocks */
    391 	uint64_t zs_ptrtbl_zt_shift;	  /* bits to index it */
    392 
    393 	/*
    394 	 * Values of the other members of the zap_phys_t
    395 	 */
    396 	uint64_t zs_block_type;		/* ZBT_HEADER */
    397 	uint64_t zs_magic;		/* ZAP_MAGIC */
    398 	uint64_t zs_num_leafs;		/* The number of leaf blocks */
    399 	uint64_t zs_num_entries;	/* The number of zap entries */
    400 	uint64_t zs_salt;		/* salt to stir into hash function */
    401 
    402 	/*
    403 	 * Histograms.  For all histograms, the last index
    404 	 * (ZAP_HISTOGRAM_SIZE-1) includes any values which are greater
    405 	 * than what can be represented.  For example
    406 	 * zs_leafs_with_n5_entries[ZAP_HISTOGRAM_SIZE-1] is the number
    407 	 * of leafs with more than 45 entries.
    408 	 */
    409 
    410 	/*
    411 	 * zs_leafs_with_n_pointers[n] is the number of leafs with
    412 	 * 2^n pointers to it.
    413 	 */
    414 	uint64_t zs_leafs_with_2n_pointers[ZAP_HISTOGRAM_SIZE];
    415 
    416 	/*
    417 	 * zs_leafs_with_n_entries[n] is the number of leafs with
    418 	 * [n*5, (n+1)*5) entries.  In the current implementation, there
    419 	 * can be at most 55 entries in any block, but there may be
    420 	 * fewer if the name or value is large, or the block is not
    421 	 * completely full.
    422 	 */
    423 	uint64_t zs_blocks_with_n5_entries[ZAP_HISTOGRAM_SIZE];
    424 
    425 	/*
    426 	 * zs_leafs_n_tenths_full[n] is the number of leafs whose
    427 	 * fullness is in the range [n/10, (n+1)/10).
    428 	 */
    429 	uint64_t zs_blocks_n_tenths_full[ZAP_HISTOGRAM_SIZE];
    430 
    431 	/*
    432 	 * zs_entries_using_n_chunks[n] is the number of entries which
    433 	 * consume n 24-byte chunks.  (Note, large names/values only use
    434 	 * one chunk, but contribute to zs_num_blocks_large.)
    435 	 */
    436 	uint64_t zs_entries_using_n_chunks[ZAP_HISTOGRAM_SIZE];
    437 
    438 	/*
    439 	 * zs_buckets_with_n_entries[n] is the number of buckets (each
    440 	 * leaf has 64 buckets) with n entries.
    441 	 * zs_buckets_with_n_entries[1] should be very close to
    442 	 * zs_num_entries.
    443 	 */
    444 	uint64_t zs_buckets_with_n_entries[ZAP_HISTOGRAM_SIZE];
    445 } zap_stats_t;
    446 
    447 /*
    448  * Get statistics about a ZAP object.  Note: you need to be aware of the
    449  * internal implementation of the ZAP to correctly interpret some of the
    450  * statistics.  This interface shouldn't be relied on unless you really
    451  * know what you're doing.
    452  */
    453 int zap_get_stats(objset_t *ds, uint64_t zapobj, zap_stats_t *zs);
    454 
    455 #ifdef	__cplusplus
    456 }
    457 #endif
    458 
    459 #endif	/* _SYS_ZAP_H */
    460