lib/classifier-private.h

   1 /*
   2  * Copyright (c) 2014, 2015, 2016 Nicira, Inc.
   3  *
   4  * Licensed under the Apache License, Version 2.0 (the "License");
   5  * you may not use this file except in compliance with the License.
   6  * You may obtain a copy of the License at:
   7  *
   8  *     http://www.apache.org/licenses/LICENSE-2.0
   9  *
  10  * Unless required by applicable law or agreed to in writing, software
  11  * distributed under the License is distributed on an "AS IS" BASIS,
  12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13  * See the License for the specific language governing permissions and
  14  * limitations under the License.
  15  */
  16
  17 #ifndef CLASSIFIER_PRIVATE_H
  18 #define CLASSIFIER_PRIVATE_H 1
  19
  20 #include "ccmap.h"
  21 #include "cmap.h"
  22 #include "flow.h"
  23 #include "hash.h"
  24 #include "rculist.h"
  25
  26 /* Classifier internal definitions, subject to change at any time. */
  27
  28 /* A set of rules that all have the same fields wildcarded. */
  29 struct cls_subtable {
  30     struct cmap_node cmap_node;    /* Within classifier's 'subtables_map'. */
  31
  32     /* These fields are only used by writers. */
  33     int max_priority;              /* Max priority of any rule in subtable. */
  34     unsigned int max_count;        /* Count of max_priority rules. */
  35
  36     /* Accessed by iterators. */
  37     struct rculist rules_list;              /* Unordered. */
  38
  39     /* Identical, but lower priority rules are not inserted to any of the
  40      * following data structures. */
  41
  42     /* These fields are accessed by readers who care about wildcarding. */
  43     const uint8_t n_indices;                   /* How many indices to use. */
  44     const struct flowmap index_maps[CLS_MAX_INDICES + 1]; /* Stage maps. */
  45     unsigned int trie_plen[CLS_MAX_TRIES];  /* Trie prefix length in 'mask'
  46                                              * (runtime configurable). */
  47     const int ports_mask_len;
  48     struct ccmap indices[CLS_MAX_INDICES];  /* Staged lookup indices. */
  49     rcu_trie_ptr ports_trie;                /* NULL if none. */
  50
  51     /* These fields are accessed by all readers. */
  52     struct cmap rules;                      /* Contains 'cls_match'es. */
  53     const struct minimask mask;             /* Wildcards for fields. */
  54     /* 'mask' must be the last field. */
  55 };
  56
  57 /* Internal representation of a rule in a "struct cls_subtable".
  58  *
  59  * The 'next' member is an element in a singly linked, null-terminated list.
  60  * This list links together identical "cls_match"es in order of decreasing
  61  * priority.  The classifier code maintains the invariant that at most one rule
  62  * of a given priority is visible for any given lookup version.
  63  */
  64 struct cls_match {
  65     /* Accessed by everybody. */
  66     OVSRCU_TYPE(struct cls_match *) next; /* Equal, lower-priority matches. */
  67     OVSRCU_TYPE(struct cls_conjunction_set *) conj_set;
  68
  69     /* Accessed by readers interested in wildcarding. */
  70     const int priority;         /* Larger numbers are higher priorities. */
  71
  72     /* Accessed by all readers. */
  73     struct cmap_node cmap_node; /* Within struct cls_subtable 'rules'. */
  74
  75     /* Rule versioning. */
  76     struct versions versions;
  77
  78     const struct cls_rule *cls_rule;
  79     const struct miniflow flow; /* Matching rule. Mask is in the subtable. */
  80     /* 'flow' must be the last field. */
  81 };
  82
  83 /* Utilities for accessing the 'cls_match' member of struct cls_rule. */
  84 static inline struct cls_match *
  85 get_cls_match_protected(const struct cls_rule *rule)
  86 {
  87     return ovsrcu_get_protected(struct cls_match *, &rule->cls_match);
  88 }
  89
  90 static inline struct cls_match *
  91 get_cls_match(const struct cls_rule *rule)
  92 {
  93     return ovsrcu_get(struct cls_match *, &rule->cls_match);
  94 }
  95
  96 /* Must be RCU postponed. */
  97 void cls_match_free_cb(struct cls_match *);
  98
  99 static inline void
 100 cls_match_set_remove_version(struct cls_match *rule, ovs_version_t version)
 101 {
 102     versions_set_remove_version(&rule->versions, version);
 103 }
 104
 105 static inline bool
 106 cls_match_visible_in_version(const struct cls_match *rule,
 107                              ovs_version_t version)
 108 {
 109     return versions_visible_in_version(&rule->versions, version);
 110 }
 111
 112 static inline bool
 113 cls_match_is_eventually_invisible(const struct cls_match *rule)
 114 {
 115     return versions_is_eventually_invisible(&rule->versions);
 116 }
 117
 118 \f
 119 /* cls_match 'next' */
 120
 121 static inline const struct cls_match *
 122 cls_match_next(const struct cls_match *rule)
 123 {
 124     return ovsrcu_get(struct cls_match *, &rule->next);
 125 }
 126
 127 static inline struct cls_match *
 128 cls_match_next_protected(const struct cls_match *rule)
 129 {
 130     return ovsrcu_get_protected(struct cls_match *, &rule->next);
 131 }
 132
 133 /* Puts 'rule' in the position between 'prev' and 'next'.  If 'prev' == NULL,
 134  * then the 'rule' is the new list head, and if 'next' == NULL, the rule is the
 135  * new list tail.
 136  * If there are any nodes between 'prev' and 'next', they are dropped from the
 137  * list. */
 138 static inline void
 139 cls_match_insert(struct cls_match *prev, struct cls_match *next,
 140                  struct cls_match *rule)
 141 {
 142     ovsrcu_set_hidden(&rule->next, next);
 143
 144     if (prev) {
 145         ovsrcu_set(&prev->next, rule);
 146     }
 147 }
 148
 149 /* Puts 'new_rule' in the position of 'old_rule', which is the next node after
 150  * 'prev'. If 'prev' == NULL, then the 'new_rule' is the new list head.
 151  *
 152  * The replaced cls_match still links to the later rules, and may still be
 153  * referenced by other threads until all other threads quiesce.  The replaced
 154  * rule may not be re-inserted, re-initialized, or deleted until after all
 155  * other threads have quiesced (use ovsrcu_postpone). */
 156 static inline void
 157 cls_match_replace(struct cls_match *prev,
 158                   struct cls_match *old_rule, struct cls_match *new_rule)
 159 {
 160     cls_match_insert(prev, cls_match_next_protected(old_rule), new_rule);
 161 }
 162
 163 /* Removes 'rule' following 'prev' from the list. If 'prev' is NULL, then the
 164  * 'rule' is a list head, and the caller is responsible for maintaining its
 165  * list head pointer (if any).
 166  *
 167  * Afterward, the removed rule is not linked to any more, but still links to
 168  * the following rules, and may still be referenced by other threads until all
 169  * other threads quiesce.  The removed rule may not be re-inserted,
 170  * re-initialized, or deleted until after all other threads have quiesced (use
 171  * ovsrcu_postpone).
 172  */
 173 static inline void
 174 cls_match_remove(struct cls_match *prev, struct cls_match *rule)
 175 {
 176     if (prev) {
 177         ovsrcu_set(&prev->next, cls_match_next_protected(rule));
 178     }
 179 }
 180
 181 #define CLS_MATCH_FOR_EACH(ITER, HEAD)                              \
 182     for ((ITER) = (HEAD); (ITER); (ITER) = cls_match_next(ITER))
 183
 184 #define CLS_MATCH_FOR_EACH_AFTER_HEAD(ITER, HEAD)   \
 185     CLS_MATCH_FOR_EACH(ITER, cls_match_next(HEAD))
 186
 187 /* Iterate cls_matches keeping the previous pointer for modifications. */
 188 #define FOR_EACH_RULE_IN_LIST_PROTECTED(ITER, PREV, HEAD)           \
 189     for ((PREV) = NULL, (ITER) = (HEAD);                            \
 190          (ITER);                                                    \
 191          (PREV) = (ITER), (ITER) = cls_match_next_protected(ITER))
 192
 193 \f
 194 /* A longest-prefix match tree. */
 195 struct trie_node {
 196     uint32_t prefix;           /* Prefix bits for this node, MSB first. */
 197     uint8_t  n_bits;           /* Never zero, except for the root node. */
 198     unsigned int n_rules;      /* Number of rules that have this prefix. */
 199     rcu_trie_ptr edges[2];     /* Both NULL if leaf. */
 200 };
 201
 202 /* Max bits per node.  Must fit in struct trie_node's 'prefix'.
 203  * Also tested with 16, 8, and 5 to stress the implementation. */
 204 #define TRIE_PREFIX_BITS 32
 205 \f
 206 /* flow/miniflow/minimask/minimatch utilities.
 207  * These are only used by the classifier, so place them here to allow
 208  * for better optimization. */
 209
 210 /* Returns a hash value for the bits of 'flow' where there are 1-bits in
 211  * 'mask', given 'basis'.
 212  *
 213  * The hash values returned by this function are the same as those returned by
 214  * miniflow_hash_in_minimask(), only the form of the arguments differ. */
 215 static inline uint32_t
 216 flow_hash_in_minimask(const struct flow *flow, const struct minimask *mask,
 217                       uint32_t basis)
 218 {
 219     const uint64_t *mask_values = miniflow_get_values(&mask->masks);
 220     const uint64_t *flow_u64 = (const uint64_t *)flow;
 221     const uint64_t *p = mask_values;
 222     uint32_t hash = basis;
 223     map_t map;
 224
 225     FLOWMAP_FOR_EACH_MAP (map, mask->masks.map) {
 226         size_t idx;
 227
 228         MAP_FOR_EACH_INDEX (idx, map) {
 229             hash = hash_add64(hash, flow_u64[idx] & *p++);
 230         }
 231         flow_u64 += MAP_T_BITS;
 232     }
 233
 234     return hash_finish(hash, (p - mask_values) * 8);
 235 }
 236
 237 /* Returns a hash value for the bits of 'flow' where there are 1-bits in
 238  * 'mask', given 'basis'.
 239  *
 240  * The hash values returned by this function are the same as those returned by
 241  * flow_hash_in_minimask(), only the form of the arguments differ. */
 242 static inline uint32_t
 243 miniflow_hash_in_minimask(const struct miniflow *flow,
 244                           const struct minimask *mask, uint32_t basis)
 245 {
 246     const uint64_t *mask_values = miniflow_get_values(&mask->masks);
 247     const uint64_t *p = mask_values;
 248     uint32_t hash = basis;
 249     uint64_t value;
 250
 251     MINIFLOW_FOR_EACH_IN_FLOWMAP(value, flow, mask->masks.map) {
 252         hash = hash_add64(hash, value & *p++);
 253     }
 254
 255     return hash_finish(hash, (p - mask_values) * 8);
 256 }
 257
 258 /* Returns a hash value for the values of 'flow', indicated by 'range', where
 259  * there are 1-bits in 'mask', given 'basis'.  'range' must be a continuous
 260  * subset of the bits in 'mask''s map, representing a continuous range of the
 261  * minimask's mask data.  '*offset' must be the number of 64-bit units of the
 262  * minimask's data to skip to get to the first unit covered by 'range'. On
 263  * return '*offset' is updated with the number of 64-bit units of the minimask
 264  * consumed.
 265  *
 266  * Typically this function is called for successive ranges of minimask's masks,
 267  * and the first invocation passes '*offset' as zero.
 268  *
 269  * The hash values returned by this function are the same as those returned by
 270  * minimatch_hash_range(), only the form of the arguments differ. */
 271 static inline uint32_t
 272 flow_hash_in_minimask_range(const struct flow *flow,
 273                             const struct minimask *mask,
 274                             const struct flowmap range,
 275                             unsigned int *offset,
 276                             uint32_t *basis)
 277 {
 278     const uint64_t *mask_values = miniflow_get_values(&mask->masks);
 279     const uint64_t *flow_u64 = (const uint64_t *)flow;
 280     const uint64_t *p = mask_values + *offset;
 281     uint32_t hash = *basis;
 282     map_t map;
 283
 284     FLOWMAP_FOR_EACH_MAP (map, range) {
 285         size_t idx;
 286
 287         MAP_FOR_EACH_INDEX (idx, map) {
 288             hash = hash_add64(hash, flow_u64[idx] & *p++);
 289         }
 290         flow_u64 += MAP_T_BITS;
 291     }
 292
 293     *basis = hash; /* Allow continuation from the unfinished value. */
 294     *offset = p - mask_values;
 295     return hash_finish(hash, *offset * 8);
 296 }
 297
 298 /* Fold minimask 'mask''s wildcard mask into 'wc's wildcard mask. */
 299 static inline void
 300 flow_wildcards_fold_minimask(struct flow_wildcards *wc,
 301                              const struct minimask *mask)
 302 {
 303     flow_union_with_miniflow(&wc->masks, &mask->masks);
 304 }
 305
 306 /* Fold minimask 'mask''s wildcard mask into 'wc's wildcard mask for bits in
 307  * 'fmap'.  1-bits in 'fmap' are a subset of 1-bits in 'mask''s map. */
 308 static inline void
 309 flow_wildcards_fold_minimask_in_map(struct flow_wildcards *wc,
 310                                     const struct minimask *mask,
 311                                     const struct flowmap fmap)
 312 {
 313     flow_union_with_miniflow_subset(&wc->masks, &mask->masks, fmap);
 314 }
 315
 316 /* Returns a hash value for 'mask', given 'basis'. */
 317 static inline uint32_t
 318 minimask_hash(const struct minimask *mask, uint32_t basis)
 319 {
 320     const uint64_t *p = miniflow_get_values(&mask->masks);
 321     size_t n_values = miniflow_n_values(&mask->masks);
 322     uint32_t hash = basis;
 323
 324     for (size_t i = 0; i < n_values; i++) {
 325         hash = hash_add64(hash, *p++);
 326     }
 327
 328     map_t map;
 329     FLOWMAP_FOR_EACH_MAP (map, mask->masks.map) {
 330         hash = hash_add64(hash, map);
 331     }
 332
 333     return hash_finish(hash, n_values);
 334 }
 335
 336 /* Returns a hash value for the values of 'match->flow', indicated by 'range',
 337  * where there are 1-bits in 'match->mask', given 'basis'.  'range' must be a
 338  * continuous subset of the bits in the map of 'match', representing a
 339  * continuous range of the mask data of 'match'.  '*offset' must be the number
 340  * of 64-bit units of the match data to skip to get to the first unit covered
 341  * by 'range'.  On return '*offset' is updated with the number of 64-bit units
 342  * of the match consumed.
 343  *
 344  * Typically this function is called for successive ranges of minimask's masks,
 345  * and the first invocation passes '*offset' as zero.
 346  *
 347  * The hash values returned by this function are the same as those returned by
 348  * flow_hash_in_minimask_range(), only the form of the arguments differ. */
 349 static inline uint32_t
 350 minimatch_hash_range(const struct minimatch *match,
 351                      const struct flowmap range, unsigned int *offset,
 352                      uint32_t *basis)
 353 {
 354     const uint64_t *p = miniflow_get_values(match->flow) + *offset;
 355     const uint64_t *q = miniflow_get_values(&match->mask->masks) + *offset;
 356     unsigned int n = flowmap_n_1bits(range);
 357     uint32_t hash = *basis;
 358
 359     for (unsigned int i = 0; i < n; i++) {
 360         hash = hash_add64(hash, p[i] & q[i]);
 361     }
 362     *basis = hash; /* Allow continuation from the unfinished value. */
 363     *offset += n;
 364     return hash_finish(hash, *offset * 8);
 365 }
 366
 367 #endif