[ovs-dev] [nxm 17/42] classifier: Rewrite.

Ben Pfaff blp at nicira.com
Thu Oct 28 10:27:48 PDT 2010


The old classifier was not adaptive: it required knowing the structure of
the flows that were likely to be in use to get good performance.  It is
likely that it degenerated to linear search in any real-world case.

This new classifier is adaptive and should perform better in the real
world.
---
 lib/classifier.c        |  992 ++++++++++++++++++++---------------------------
 lib/classifier.h        |  130 ++----
 lib/flow.c              |   77 ++++-
 lib/flow.h              |   28 ++-
 tests/classifier.at     |    8 +-
 tests/test-classifier.c |  584 +++++++++++++---------------
 6 files changed, 835 insertions(+), 984 deletions(-)

diff --git a/lib/classifier.c b/lib/classifier.c
index ae2019f..90c7166 100644
--- a/lib/classifier.c
+++ b/lib/classifier.c
@@ -22,36 +22,96 @@
 #include "dynamic-string.h"
 #include "flow.h"
 #include "hash.h"
+#include "packets.h"
 
-const struct cls_field cls_fields[CLS_N_FIELDS + 1] = {
-#define CLS_FIELD(WILDCARDS, MEMBER, NAME)      \
-    { offsetof(struct flow, MEMBER),            \
-      sizeof ((struct flow *)0)->MEMBER,        \
-      WILDCARDS,                                \
-      #NAME },
-    CLS_FIELDS
-#undef CLS_FIELD
-    { sizeof(struct flow), 0, 0, "exact" },
-};
-
-static uint32_t hash_fields(const struct flow *, int table_idx);
-static bool equal_fields(const struct flow *, const struct flow *,
-                         int table_idx);
-
-static int table_idx_from_wildcards(uint32_t wildcards);
-static struct cls_rule *table_insert(struct hmap *, struct cls_rule *);
-static struct cls_rule *insert_exact_rule(struct classifier *,
-                                          struct cls_rule *);
-static struct cls_bucket *find_bucket(struct hmap *, size_t hash,
-                                      const struct cls_rule *);
-static struct cls_rule *search_table(const struct hmap *table, int field_idx,
-                                     const struct cls_rule *);
-static struct cls_rule *search_exact_table(const struct classifier *,
-                                           size_t hash, const struct flow *);
-static bool rules_match_1wild(const struct cls_rule *fixed,
-                              const struct cls_rule *wild, int field_idx);
-static bool rules_match_2wild(const struct cls_rule *wild1,
-                              const struct cls_rule *wild2, int field_idx);
+static struct cls_table *find_table(const struct classifier *,
+                                    const struct flow_wildcards *);
+static struct cls_table *insert_table(struct classifier *,
+                                      const struct flow_wildcards *);
+
+static struct cls_table *classifier_first_table(const struct classifier *);
+static struct cls_table *classifier_next_table(const struct classifier *,
+                                               const struct cls_table *);
+void classifier_destroy_table(struct classifier *, struct cls_table *);
+
+static bool should_include(const struct cls_table *, int include);
+
+static struct cls_rule *find_match(const struct cls_table *,
+                                   const struct flow *);
+static struct cls_rule *find_equal(struct cls_table *, const struct flow *,
+                                   uint32_t hash);
+static struct cls_rule *insert_rule(struct cls_table *, struct cls_rule *);
+
+static bool flow_equal_except(const struct flow *, const struct flow *,
+                                const struct flow_wildcards *);
+static void zero_wildcards(struct flow *, const struct flow_wildcards *);
+
+/* Iterates RULE over HEAD and all of the cls_rules on HEAD->list. */
+#define FOR_EACH_RULE_IN_LIST(RULE, HEAD)                               \
+    for ((RULE) = (HEAD); (RULE) != NULL; (RULE) = next_rule_in_list(RULE))
+#define FOR_EACH_RULE_IN_LIST_SAFE(RULE, NEXT, HEAD)                    \
+    for ((RULE) = (HEAD);                                               \
+         (RULE) != NULL && ((NEXT) = next_rule_in_list(RULE), true);    \
+         (RULE) = (NEXT))
+
+static struct cls_rule *next_rule_in_list(struct cls_rule *);
+
+static struct cls_table *
+cls_table_from_hmap_node(const struct hmap_node *node)
+{
+    return node ? CONTAINER_OF(node, struct cls_table, hmap_node) : NULL;
+}
+
+static struct cls_rule *
+cls_rule_from_hmap_node(const struct hmap_node *node)
+{
+    return node ? CONTAINER_OF(node, struct cls_rule, hmap_node) : NULL;
+}
+
+/* Returns the cls_table within 'cls' that has no wildcards, or NULL if there
+ * is none.  */
+struct cls_table *
+classifier_exact_table(const struct classifier *cls)
+{
+    struct flow_wildcards exact_wc;
+    flow_wildcards_init_exact(&exact_wc);
+    return find_table(cls, &exact_wc);
+}
+
+/* Returns the first rule in 'table', or a null pointer if 'table' is NULL.
+ *
+ * (The classifier never creates empty tables and destroys tables as soon as
+ * they become empty, so if 'table' is null then the return value should always
+ * be nonnull.  However, the implementation of this function is robust
+ * anyhow.) */
+struct cls_rule *
+cls_table_first_rule(const struct cls_table *table)
+{
+    return table ? cls_rule_from_hmap_node(hmap_first(&table->rules)) : NULL;
+}
+
+/* Returns the next rule in 'table' following 'rule', or a null pointer if
+ * 'rule' is the last rule in 'table'. */
+struct cls_rule *
+cls_table_next_rule(const struct cls_table *table, const struct cls_rule *rule)
+{
+    struct cls_rule *next
+        = CONTAINER_OF(rule->list.next, struct cls_rule, hmap_node);
+
+    return (next->priority < rule->priority
+            ? next
+            : cls_rule_from_hmap_node(hmap_next(&table->rules,
+                                                &next->hmap_node)));
+}
+
+static void
+cls_rule_init__(struct cls_rule *rule,
+                const struct flow *flow, uint32_t wildcards)
+{
+    rule->flow = *flow;
+    flow_wildcards_init(&rule->wc, wildcards);
+    cls_rule_zero_wildcards(rule);
+}
 
 /* Converts the flow in 'flow' into a cls_rule in 'rule', with the given
  * 'wildcards' and 'priority'.*/
@@ -59,10 +119,8 @@ void
 cls_rule_from_flow(const struct flow *flow, uint32_t wildcards,
                    unsigned int priority, struct cls_rule *rule)
 {
-    rule->flow = *flow;
-    flow_wildcards_init(&rule->wc, wildcards);
+    cls_rule_init__(rule, flow, wildcards);
     rule->priority = priority;
-    rule->table_idx = table_idx_from_wildcards(rule->wc.wildcards);
 }
 
 /* Converts the ofp_match in 'match' into a cls_rule in 'rule', with the given
@@ -74,10 +132,25 @@ cls_rule_from_match(const struct ofp_match *match, unsigned int priority,
                     struct cls_rule *rule)
 {
     uint32_t wildcards;
-    flow_from_match(match, tun_id_from_cookie, cookie, &rule->flow, &wildcards);
-    flow_wildcards_init(&rule->wc, wildcards);
+    struct flow flow;
+
+    flow_from_match(match, tun_id_from_cookie, cookie, &flow, &wildcards);
+    cls_rule_init__(rule, &flow, wildcards);
     rule->priority = rule->wc.wildcards ? priority : UINT16_MAX;
-    rule->table_idx = table_idx_from_wildcards(rule->wc.wildcards);
+}
+
+/* For each bit or field wildcarded in 'rule', sets the corresponding bit or
+ * field in 'flow' to all-0-bits.  It is important to maintain this invariant
+ * in a clr_rule that might be inserted into a classifier.
+ *
+ * It is never necessary to call this function directly for a cls_rule that is
+ * initialized or modified only by cls_rule_*() functions.  It is useful to
+ * restore the invariant in a cls_rule whose 'wc' member is modified by hand.
+ */
+void
+cls_rule_zero_wildcards(struct cls_rule *rule)
+{
+    zero_wildcards(&rule->flow, &rule->wc);
 }
 
 /* Converts 'rule' to a string and returns the string.  The caller must free
@@ -109,13 +182,8 @@ cls_rule_print(const struct cls_rule *rule)
 void
 classifier_init(struct classifier *cls)
 {
-    int i;
-
     cls->n_rules = 0;
-    for (i = 0; i < ARRAY_SIZE(cls->tables); i++) {
-        hmap_init(&cls->tables[i]);
-    }
-    hmap_init(&cls->exact_table);
+    hmap_init(&cls->tables);
 }
 
 /* Destroys 'cls'.  Rules within 'cls', if any, are not freed; this is the
@@ -124,21 +192,18 @@ void
 classifier_destroy(struct classifier *cls)
 {
     if (cls) {
-        struct cls_bucket *bucket, *next_bucket;
-        struct hmap *tbl;
+        struct cls_table *table, *next_table;
 
-        for (tbl = &cls->tables[0]; tbl < &cls->tables[CLS_N_FIELDS]; tbl++) {
-            HMAP_FOR_EACH_SAFE (bucket, next_bucket, hmap_node, tbl) {
-                free(bucket);
-            }
-            hmap_destroy(tbl);
+        HMAP_FOR_EACH_SAFE (table, next_table, hmap_node, &cls->tables) {
+            hmap_destroy(&table->rules);
+            hmap_remove(&cls->tables, &table->hmap_node);
+            free(table);
         }
-        hmap_destroy(&cls->exact_table);
+        hmap_destroy(&cls->tables);
     }
 }
 
-/* Returns true if 'cls' does not contain any classification rules, false
- * otherwise. */
+/* Returns true if 'cls' contains no classification rules, false otherwise. */
 bool
 classifier_is_empty(const struct classifier *cls)
 {
@@ -156,10 +221,12 @@ classifier_count(const struct classifier *cls)
 int
 classifier_count_exact(const struct classifier *cls)
 {
-    return hmap_count(&cls->exact_table);
+    struct cls_table *exact_table = classifier_exact_table(cls);
+    return exact_table ? exact_table->n_table_rules : 0;
 }
 
-/* Inserts 'rule' into 'cls'.  Transfers ownership of 'rule' to 'cls'.
+/* Inserts 'rule' into 'cls'.  Until 'rule' is removed from 'cls', the caller
+ * must not modify or free it.
  *
  * If 'cls' already contains an identical rule (including wildcards, values of
  * fixed fields, and priority), replaces the old rule by 'rule' and returns the
@@ -173,127 +240,100 @@ classifier_count_exact(const struct classifier *cls)
 struct cls_rule *
 classifier_insert(struct classifier *cls, struct cls_rule *rule)
 {
-    struct cls_rule *old;
-    assert((rule->wc.wildcards == 0) == (rule->table_idx == CLS_F_IDX_EXACT));
-    old = (rule->wc.wildcards
-           ? table_insert(&cls->tables[rule->table_idx], rule)
-           : insert_exact_rule(cls, rule));
-    if (!old) {
+    struct cls_rule *old_rule;
+    struct cls_table *table;
+
+    table = find_table(cls, &rule->wc);
+    if (!table) {
+        table = insert_table(cls, &rule->wc);
+    }
+
+    old_rule = insert_rule(table, rule);
+    if (!old_rule) {
+        table->n_table_rules++;
         cls->n_rules++;
     }
-    return old;
+    return old_rule;
 }
 
-/* Removes 'rule' from 'cls'.  It is caller's responsibility to free 'rule', if
- * this is desirable. */
+/* Removes 'rule' from 'cls'.  It is the caller's responsibility to free
+ * 'rule', if this is desirable. */
 void
 classifier_remove(struct classifier *cls, struct cls_rule *rule)
 {
-    if (rule->wc.wildcards) {
-        /* Remove 'rule' from bucket.  If that empties the bucket, remove the
-         * bucket from its table. */
-        struct hmap *table = &cls->tables[rule->table_idx];
-        struct list *rules = list_remove(&rule->node.list);
-        if (list_is_empty(rules)) {
-            /* This code is a little tricky.  list_remove() returns the list
-             * element just after the one removed.  Since the list is now
-             * empty, this will be the address of the 'rules' member of the
-             * bucket that was just emptied, so pointer arithmetic (via
-             * CONTAINER_OF) can find that bucket. */
-            struct cls_bucket *bucket;
-            bucket = CONTAINER_OF(rules, struct cls_bucket, rules);
-            hmap_remove(table, &bucket->hmap_node);
-            free(bucket);
-        }
-    } else {
-        /* Remove 'rule' from cls->exact_table. */
-        hmap_remove(&cls->exact_table, &rule->node.hmap);
-    }
-    cls->n_rules--;
-}
+    struct cls_rule *head;
+    struct cls_table *table;
 
-static struct cls_rule *
-classifier_lookup_exact(const struct classifier *cls, const struct flow *flow)
-{
-    return (!hmap_is_empty(&cls->exact_table)
-            ? search_exact_table(cls, flow_hash(flow, 0), flow)
-            : NULL);
-}
+    table = find_table(cls, &rule->wc);
+    head = find_equal(table, &rule->flow, rule->hmap_node.hash);
+    if (head != rule) {
+        list_remove(&rule->list);
+    } else if (list_is_empty(&rule->list)) {
+        hmap_remove(&table->rules, &rule->hmap_node);
+    } else {
+        struct cls_rule *next = CONTAINER_OF(rule->list.next,
+                                             struct cls_rule, list);
 
-static struct cls_rule *
-classifier_lookup_wild(const struct classifier *cls, const struct flow *flow)
-{
-    struct cls_rule *best = NULL;
-    if (cls->n_rules > hmap_count(&cls->exact_table)) {
-        struct cls_rule target;
-        int i;
+        list_remove(&rule->list);
+        hmap_replace(&table->rules, &rule->hmap_node, &next->hmap_node);
+    }
 
-        cls_rule_from_flow(flow, 0, 0, &target);
-        for (i = 0; i < CLS_N_FIELDS; i++) {
-            struct cls_rule *rule = search_table(&cls->tables[i], i, &target);
-            if (rule && (!best || rule->priority > best->priority)) {
-                best = rule;
-            }
-        }
+    if (--table->n_table_rules == 0 && !table->n_refs) {
+        classifier_destroy_table(cls, table);
     }
-    return best;
+
+    cls->n_rules--;
 }
 
 /* Finds and returns the highest-priority rule in 'cls' that matches 'flow'.
  * Returns a null pointer if no rules in 'cls' match 'flow'.  If multiple rules
  * of equal priority match 'flow', returns one arbitrarily.
  *
- * (When multiple rules of equal priority happen to fall into the same bucket,
- * rules added more recently take priority over rules added less recently, but
- * this is subject to change and should not be depended upon.) */
+ * 'include' is a combination of CLS_INC_* values that specify tables to
+ * include in the search. */
 struct cls_rule *
 classifier_lookup(const struct classifier *cls, const struct flow *flow,
                   int include)
 {
-    if (include & CLS_INC_EXACT) {
-        struct cls_rule *rule = classifier_lookup_exact(cls, flow);
-        if (rule) {
-            return rule;
-        }
-    }
+    struct cls_table *table;
+    struct cls_rule *best;
 
-    if (include & CLS_INC_WILD) {
-        return classifier_lookup_wild(cls, flow);
+    best = NULL;
+    HMAP_FOR_EACH (table, hmap_node, &cls->tables) {
+        if (should_include(table, include)) {
+            struct cls_rule *rule = find_match(table, flow);
+            if (rule && (!best || rule->priority > best->priority)) {
+                best = rule;
+            }
+        }
     }
-
-    return NULL;
+    return best;
 }
 
+/* Finds and returns a rule in 'cls' with exactly the same priority and
+ * matching criteria as 'target'.  Returns a null pointer if 'cls' doesn't
+ * contain an exact match.
+ *
+ * Priority is ignored for exact-match rules. */
 struct cls_rule *
 classifier_find_rule_exactly(const struct classifier *cls,
                              const struct cls_rule *target)
 {
-    struct cls_bucket *bucket;
-    int table_idx;
-    uint32_t hash;
+    struct cls_rule *head, *rule;
+    struct cls_table *table;
 
-    if (!target->wc.wildcards) {
-        /* Ignores 'target->priority'. */
-        return search_exact_table(cls, flow_hash(&target->flow, 0),
-                                  &target->flow);
+    table = find_table(cls, &target->wc);
+    if (!table) {
+        return NULL;
     }
 
-    assert(target->wc.wildcards == (target->wc.wildcards & OVSFW_ALL));
-    table_idx = table_idx_from_wildcards(target->wc.wildcards);
-    hash = hash_fields(&target->flow, table_idx);
-    HMAP_FOR_EACH_WITH_HASH (bucket, hmap_node, hash,
-                             &cls->tables[table_idx]) {
-        if (equal_fields(&bucket->fixed, &target->flow, table_idx)) {
-            struct cls_rule *pos;
-            LIST_FOR_EACH (pos, node.list, &bucket->rules) {
-                if (pos->priority < target->priority) {
-                    return NULL;
-                } else if (pos->priority == target->priority &&
-                           pos->wc.wildcards == target->wc.wildcards &&
-                           flow_equal(&target->flow, &pos->flow)) {
-                    return pos;
-                }
-            }
+    head = find_equal(table, &target->flow, flow_hash(&target->flow, 0));
+    if (!target->wc.wildcards) {
+        return head;
+    }
+    FOR_EACH_RULE_IN_LIST (rule, head) {
+        if (target->priority >= rule->priority) {
+            return target->priority == rule->priority ? rule : NULL;
         }
     }
     return NULL;
@@ -306,22 +346,19 @@ bool
 classifier_rule_overlaps(const struct classifier *cls,
                          const struct cls_rule *target)
 {
-    const struct hmap *tbl;
-
-    if (!target->wc.wildcards) {
-        return (search_exact_table(cls, flow_hash(&target->flow, 0),
-                                   &target->flow) != NULL);
-    }
+    struct cls_table *table;
 
-    for (tbl = &cls->tables[0]; tbl < &cls->tables[CLS_N_FIELDS]; tbl++) {
-        struct cls_bucket *bucket;
+    HMAP_FOR_EACH (table, hmap_node, &cls->tables) {
+        struct flow_wildcards wc;
+        struct cls_rule *head;
 
-        HMAP_FOR_EACH (bucket, hmap_node, tbl) {
+        flow_wildcards_combine(&wc, &target->wc, &table->wc);
+        HMAP_FOR_EACH (head, hmap_node, &table->rules) {
             struct cls_rule *rule;
 
-            LIST_FOR_EACH (rule, node.list, &bucket->rules) {
+            FOR_EACH_RULE_IN_LIST (rule, head) {
                 if (rule->priority == target->priority
-                    && rules_match_2wild(rule, target, 0)) {
+                    && flow_equal_except(&target->flow, &rule->flow, &wc)) {
                     return true;
                 }
             }
@@ -331,511 +368,314 @@ classifier_rule_overlaps(const struct classifier *cls,
     return false;
 }
 
-/* Ignores target->priority.
+/* Searches 'cls' for rules that exactly match 'target' or are more specific
+ * than 'target'.  That is, a given 'rule' matches 'target' if, for every
+ * field:
+ *
+ *   - 'target' and 'rule' specify the same (non-wildcarded) value for the
+ *     field, or
+ *
+ *   - 'target' wildcards the field,
+ *
+ * but not if:
+ *
+ *   - 'target' and 'rule' specify different values for the field, or
+ *
+ *   - 'target' specifies a value for the field but 'rule' wildcards it.
+ *
+ * Equivalently, the truth table for whether a field matches is:
+ *
+ *                                     rule
+ *
+ *                             wildcard    exact
+ *                            +---------+---------+
+ *                   t   wild |   yes   |   yes   |
+ *                   a   card |         |         |
+ *                   r        +---------+---------+
+ *                   g  exact |    no   |if values|
+ *                   e        |         |are equal|
+ *                   t        +---------+---------+
+ *
+ * This is the matching rule used by OpenFlow 1.0 non-strict OFPT_FLOW_MOD
+ * commands and by OpenFlow 1.0 aggregate and flow stats.
+ *
+ * Ignores target->priority.
  *
  * 'callback' is allowed to delete the rule that is passed as its argument, but
- * it must not delete (or move) any other rules in 'cls' that are in the same
- * table as the argument rule.  Two rules are in the same table if their
- * cls_rule structs have the same table_idx; as a special case, a rule with
- * wildcards and an exact-match rule will never be in the same table. */
+ * it must not delete (or move) any other rules in 'cls' that have the same
+ * wildcards as the argument rule. */
 void
-classifier_for_each_match(const struct classifier *cls,
+classifier_for_each_match(const struct classifier *cls_,
                           const struct cls_rule *target,
                           int include, cls_cb_func *callback, void *aux)
 {
-    if (include & CLS_INC_WILD) {
-        const struct hmap *table;
-
-        for (table = &cls->tables[0]; table < &cls->tables[CLS_N_FIELDS];
-             table++) {
-            struct cls_bucket *bucket, *next_bucket;
-
-            HMAP_FOR_EACH_SAFE (bucket, next_bucket, hmap_node, table) {
-                /* XXX there is a bit of room for optimization here based on
-                 * rejecting entire buckets on their fixed fields, but it will
-                 * only be worthwhile for big buckets (which we hope we won't
-                 * get anyway, but...) */
-                struct cls_rule *prev_rule, *rule;
-
-                /* We can't just use LIST_FOR_EACH_SAFE here because, if the
-                 * callback deletes the last rule in the bucket, then the
-                 * bucket itself will be destroyed.  The bucket contains the
-                 * list head so that's a use-after-free error. */
-                prev_rule = NULL;
-                LIST_FOR_EACH (rule, node.list, &bucket->rules) {
-                    if (rules_match_1wild(rule, target, 0)) {
-                        if (prev_rule) {
-                            callback(prev_rule, aux);
-                        }
-                        prev_rule = rule;
+    struct classifier *cls = (struct classifier *) cls_;
+    struct cls_table *table, *next_table;
+
+    for (table = classifier_first_table(cls); table; table = next_table) {
+        if (should_include(table, include)
+            && !flow_wildcards_has_extra(&table->wc, &target->wc)) {
+            /* We have eliminated the "no" case in the truth table above.  Two
+             * of the three remaining cases are trivial.  We only need to check
+             * the fourth case, where both 'rule' and 'target' require an exact
+             * match. */
+            struct cls_rule *head, *next_head;
+
+            table->n_refs++;
+            HMAP_FOR_EACH_SAFE (head, next_head, hmap_node, &table->rules) {
+                if (flow_equal_except(&head->flow, &target->flow,
+                                      &target->wc)) {
+                    struct cls_rule *rule, *next_rule;
+
+                    FOR_EACH_RULE_IN_LIST_SAFE (rule, next_rule, head) {
+                        callback(rule, aux);
                     }
                 }
-                if (prev_rule) {
-                    callback(prev_rule, aux);
-                }
             }
-        }
-    }
-
-    if (include & CLS_INC_EXACT) {
-        if (target->wc.wildcards) {
-            struct cls_rule *rule, *next_rule;
-
-            HMAP_FOR_EACH_SAFE (rule, next_rule, node.hmap,
-                                &cls->exact_table) {
-                if (rules_match_1wild(rule, target, 0)) {
-                    callback(rule, aux);
-                }
+            next_table = classifier_next_table(cls, table);
+            if (!--table->n_refs && !table->n_table_rules) {
+                classifier_destroy_table(cls, table);
             }
         } else {
-            /* Optimization: there can be at most one match in the exact
-             * table. */
-            size_t hash = flow_hash(&target->flow, 0);
-            struct cls_rule *rule = search_exact_table(cls, hash,
-                                                       &target->flow);
-            if (rule) {
-                callback(rule, aux);
-            }
+            next_table = classifier_next_table(cls, table);
         }
     }
 }
 
 /* 'callback' is allowed to delete the rule that is passed as its argument, but
- * it must not delete (or move) any other rules in 'cls' that are in the same
- * table as the argument rule.  Two rules are in the same table if their
- * cls_rule structs have the same table_idx; as a special case, a rule with
- * wildcards and an exact-match rule will never be in the same table.
+ * it must not delete (or move) any other rules in 'cls' that have the same
+ * wildcards as the argument rule.
  *
  * If 'include' is CLS_INC_EXACT then CLASSIFIER_FOR_EACH_EXACT_RULE(_SAFE) is
  * probably easier to use. */
 void
-classifier_for_each(const struct classifier *cls, int include,
+classifier_for_each(const struct classifier *cls_, int include,
                     void (*callback)(struct cls_rule *, void *aux),
                     void *aux)
 {
-    if (include & CLS_INC_WILD) {
-        const struct hmap *tbl;
+    struct classifier *cls = (struct classifier *) cls_;
+    struct cls_table *table, *next_table;
 
-        for (tbl = &cls->tables[0]; tbl < &cls->tables[CLS_N_FIELDS]; tbl++) {
-            struct cls_bucket *bucket, *next_bucket;
+    for (table = classifier_first_table(cls); table; table = next_table) {
+        if (should_include(table, include)) {
+            struct cls_rule *head, *next_head;
 
-            HMAP_FOR_EACH_SAFE (bucket, next_bucket, hmap_node, tbl) {
-                struct cls_rule *prev_rule, *rule;
+            table->n_refs++;
+            HMAP_FOR_EACH_SAFE (head, next_head, hmap_node, &table->rules) {
+                struct cls_rule *rule, *next_rule;
 
-                /* We can't just use LIST_FOR_EACH_SAFE here because, if the
-                 * callback deletes the last rule in the bucket, then the
-                 * bucket itself will be destroyed.  The bucket contains the
-                 * list head so that's a use-after-free error. */
-                prev_rule = NULL;
-                LIST_FOR_EACH (rule, node.list, &bucket->rules) {
-                    if (prev_rule) {
-                        callback(prev_rule, aux);
-                    }
-                    prev_rule = rule;
-                }
-                if (prev_rule) {
-                    callback(prev_rule, aux);
+                FOR_EACH_RULE_IN_LIST_SAFE (rule, next_rule, head) {
+                    callback(rule, aux);
                 }
             }
+            next_table = classifier_next_table(cls, table);
+            if (!--table->n_refs && !table->n_table_rules) {
+                hmap_remove(&cls->tables, &table->hmap_node);
+                free(table);
+            }
+        } else {
+            next_table = classifier_next_table(cls, table);
         }
     }
+}
+
+static struct cls_table *
+find_table(const struct classifier *cls, const struct flow_wildcards *wc)
+{
+    struct cls_table *table;
 
-    if (include & CLS_INC_EXACT) {
-        struct cls_rule *rule, *next_rule;
-
-        HMAP_FOR_EACH_SAFE (rule, next_rule, node.hmap, &cls->exact_table) {
-            callback(rule, aux);
+    HMAP_FOR_EACH_IN_BUCKET (table, hmap_node, flow_wildcards_hash(wc),
+                             &cls->tables) {
+        if (flow_wildcards_equal(wc, &table->wc)) {
+            return table;
         }
     }
+    return table;
 }
-
-static struct cls_bucket *create_bucket(struct hmap *, size_t hash,
-                                        const struct flow *fixed);
-static struct cls_rule *bucket_insert(struct cls_bucket *, struct cls_rule *);
-
-static inline bool equal_bytes(const void *, const void *, size_t n);
-
-/* Returns a hash computed across the fields in 'flow' whose field indexes
- * (CLS_F_IDX_*) are less than 'table_idx'.  (If 'table_idx' is
- * CLS_F_IDX_EXACT, hashes all the fields in 'flow'). */
-static uint32_t
-hash_fields(const struct flow *flow, int table_idx)
-{
-    /* I just know I'm going to hell for writing code this way.
-     *
-     * GCC generates pretty good code here, with only a single taken
-     * conditional jump per execution.  Now the question is, would we be better
-     * off marking this function ALWAYS_INLINE and writing a wrapper that
-     * switches on the value of 'table_idx' to get rid of all the conditional
-     * jumps entirely (except for one in the wrapper)?  Honestly I really,
-     * really hope that it doesn't matter in practice.
-     *
-     * We could do better by calculating hashes incrementally, instead of
-     * starting over from the top each time.  But that would be even uglier. */
-    uint32_t a, b, c;
-    uint32_t tmp[3];
-    size_t n;
-
-    a = b = c = 0xdeadbeef + table_idx;
-    n = 0;
-
-#define CLS_FIELD(WILDCARDS, MEMBER, NAME)                      \
-    if (table_idx == CLS_F_IDX_##NAME) {                        \
-        /* Done. */                                             \
-        memset((uint8_t *) tmp + n, 0, sizeof tmp - n);         \
-        goto finish;                                            \
-    } else {                                                    \
-        const size_t size = sizeof flow->MEMBER;                \
-        const uint8_t *p1 = (const uint8_t *) &flow->MEMBER;    \
-        const size_t p1_size = MIN(sizeof tmp - n, size);       \
-        const uint8_t *p2 = p1 + p1_size;                       \
-        const size_t p2_size = size - p1_size;                  \
-                                                                \
-        /* Append to 'tmp' as much data as will fit. */         \
-        memcpy((uint8_t *) tmp + n, p1, p1_size);               \
-        n += p1_size;                                           \
-                                                                \
-        /* If 'tmp' is full, mix. */                            \
-        if (n == sizeof tmp) {                                  \
-            a += tmp[0];                                        \
-            b += tmp[1];                                        \
-            c += tmp[2];                                        \
-            HASH_MIX(a, b, c);                                  \
-            n = 0;                                              \
-        }                                                       \
-                                                                \
-        /* Append to 'tmp' any data that didn't fit. */         \
-        memcpy(tmp, p2, p2_size);                               \
-        n += p2_size;                                           \
-    }
-    CLS_FIELDS
-#undef CLS_FIELD
 
-finish:
-    a += tmp[0];
-    b += tmp[1];
-    c += tmp[2];
-    HASH_FINAL(a, b, c);
-    return c;
-}
+static struct cls_table *
+insert_table(struct classifier *cls, const struct flow_wildcards *wc)
+{
+    struct cls_table *table;
 
-/* Compares the fields in 'a' and 'b' whose field indexes (CLS_F_IDX_*) are
- * less than 'table_idx'.  (If 'table_idx' is CLS_F_IDX_EXACT, compares all the
- * fields in 'a' and 'b').
- *
- * Returns true if all the compared fields are equal, false otherwise. */
-static bool
-equal_fields(const struct flow *a, const struct flow *b, int table_idx)
-{
-    /* XXX The generated code could be better here. */
-#define CLS_FIELD(WILDCARDS, MEMBER, NAME)                              \
-    if (table_idx == CLS_F_IDX_##NAME) {                                \
-        return true;                                                    \
-    } else if (!equal_bytes(&a->MEMBER, &b->MEMBER, sizeof a->MEMBER)) { \
-        return false;                                                   \
-    }
-    CLS_FIELDS
-#undef CLS_FIELD
+    table = xzalloc(sizeof *table);
+    hmap_init(&table->rules);
+    table->wc = *wc;
+    hmap_insert(&cls->tables, &table->hmap_node, flow_wildcards_hash(wc));
 
-    return true;
+    return table;
 }
 
-static int
-table_idx_from_wildcards(uint32_t wildcards)
+static struct cls_table *
+classifier_first_table(const struct classifier *cls)
 {
-    if (!wildcards) {
-        return CLS_F_IDX_EXACT;
-    }
-#define CLS_FIELD(WILDCARDS, MEMBER, NAME) \
-    if (wildcards & WILDCARDS) {           \
-        return CLS_F_IDX_##NAME;           \
-    }
-    CLS_FIELDS
-#undef CLS_FIELD
-    NOT_REACHED();
+    return cls_table_from_hmap_node(hmap_first(&cls->tables));
 }
 
-/* Inserts 'rule' into 'table'.  Returns the rule, if any, that was displaced
- * in favor of 'rule'. */
-static struct cls_rule *
-table_insert(struct hmap *table, struct cls_rule *rule)
+static struct cls_table *
+classifier_next_table(const struct classifier *cls,
+                      const struct cls_table *table)
 {
-    struct cls_bucket *bucket;
-    size_t hash;
+    return cls_table_from_hmap_node(hmap_next(&cls->tables,
+                                              &table->hmap_node));
+}
 
-    hash = hash_fields(&rule->flow, rule->table_idx);
-    bucket = find_bucket(table, hash, rule);
-    if (!bucket) {
-        bucket = create_bucket(table, hash, &rule->flow);
-    }
+void
+classifier_destroy_table(struct classifier *cls, struct cls_table *table)
+{
+    hmap_remove(&cls->tables, &table->hmap_node);
+    hmap_destroy(&table->rules);
+    free(table);
+}
 
-    return bucket_insert(bucket, rule);
+/* Returns true if 'table' should be included by an operation with the
+ * specified 'include' (a combination of CLS_INC_*). */
+static bool
+should_include(const struct cls_table *table, int include)
+{
+    return include & (table->wc.wildcards ? CLS_INC_WILD : CLS_INC_EXACT);
 }
 
-/* Inserts 'rule' into 'bucket', given that 'field' is the first wildcarded
- * field in 'rule'.
- *
- * Returns the rule, if any, that was displaced in favor of 'rule'. */
 static struct cls_rule *
-bucket_insert(struct cls_bucket *bucket, struct cls_rule *rule)
-{
-    struct cls_rule *pos;
-    LIST_FOR_EACH (pos, node.list, &bucket->rules) {
-        if (pos->priority == rule->priority) {
-            if (pos->wc.wildcards == rule->wc.wildcards
-                && rules_match_1wild(pos, rule, rule->table_idx))
-            {
-                list_replace(&rule->node.list, &pos->node.list);
-                return pos;
-            }
-        } else if (pos->priority < rule->priority) {
-            break;
+find_match(const struct cls_table *table, const struct flow *flow)
+{
+    struct cls_rule *rule;
+    struct flow f;
+
+    f = *flow;
+    zero_wildcards(&f, &table->wc);
+    HMAP_FOR_EACH_WITH_HASH (rule, hmap_node, flow_hash(&f, 0),
+                             &table->rules) {
+        if (flow_equal(&f, &rule->flow)) {
+            return rule;
         }
     }
-    list_insert(&pos->node.list, &rule->node.list);
     return NULL;
 }
 
 static struct cls_rule *
-insert_exact_rule(struct classifier *cls, struct cls_rule *rule)
+find_equal(struct cls_table *table, const struct flow *flow, uint32_t hash)
 {
-    struct cls_rule *old_rule;
-    size_t hash;
-
-    hash = flow_hash(&rule->flow, 0);
-    old_rule = search_exact_table(cls, hash, &rule->flow);
-    if (old_rule) {
-        hmap_remove(&cls->exact_table, &old_rule->node.hmap);
-    }
-    hmap_insert(&cls->exact_table, &rule->node.hmap, hash);
-    return old_rule;
-}
+    struct cls_rule *head;
 
-/* Returns the bucket in 'table' that has the given 'hash' and the same fields
- * as 'rule->flow' (up to 'rule->table_idx'), or a null pointer if no bucket
- * matches. */
-static struct cls_bucket *
-find_bucket(struct hmap *table, size_t hash, const struct cls_rule *rule)
-{
-    struct cls_bucket *bucket;
-    HMAP_FOR_EACH_WITH_HASH (bucket, hmap_node, hash, table) {
-        if (equal_fields(&bucket->fixed, &rule->flow, rule->table_idx)) {
-            return bucket;
+    HMAP_FOR_EACH_WITH_HASH (head, hmap_node, hash, &table->rules) {
+        if (flow_equal(&head->flow, flow)) {
+            return head;
         }
     }
     return NULL;
 }
 
-/* Creates a bucket and inserts it in 'table' with the given 'hash' and 'fixed'
- * values.  Returns the new bucket. */
-static struct cls_bucket *
-create_bucket(struct hmap *table, size_t hash, const struct flow *fixed)
-{
-    struct cls_bucket *bucket = xmalloc(sizeof *bucket);
-    list_init(&bucket->rules);
-    bucket->fixed = *fixed;
-    hmap_insert(table, &bucket->hmap_node, hash);
-    return bucket;
-}
-
-/* Returns true if the 'n' bytes in 'a' and 'b' are equal, false otherwise. */
-static inline bool ALWAYS_INLINE
-equal_bytes(const void *a, const void *b, size_t n)
+static struct cls_rule *
+insert_rule(struct cls_table *table, struct cls_rule *new)
 {
-#ifdef __i386__
-    /* For some reason GCC generates stupid code for memcmp() of small
-     * constant integer lengths.  Help it out.
-     *
-     * This function is always inlined, and it is always called with 'n' as a
-     * compile-time constant, so the switch statement gets optimized out and
-     * this whole function just expands to an instruction or two. */
-    switch (n) {
-    case 1:
-        return *(uint8_t *) a == *(uint8_t *) b;
+    struct cls_rule *head;
 
-    case 2:
-        return *(uint16_t *) a == *(uint16_t *) b;
+    new->hmap_node.hash = flow_hash(&new->flow, 0);
 
-    case 4:
-        return *(uint32_t *) a == *(uint32_t *) b;
+    head = find_equal(table, &new->flow, new->hmap_node.hash);
+    if (!head) {
+        hmap_insert(&table->rules, &new->hmap_node, new->hmap_node.hash);
+        list_init(&new->list);
+        return NULL;
+    } else {
+        /* Scan the list for the insertion point that will keep the list in
+         * order of decreasing priority. */
+        struct cls_rule *rule;
+        FOR_EACH_RULE_IN_LIST (rule, head) {
+            if (new->priority >= rule->priority) {
+                if (rule == head) {
+                    /* 'new' is the new highest-priority flow in the list. */
+                    hmap_replace(&table->rules,
+                                 &rule->hmap_node, &new->hmap_node);
+                }
 
-    case 6:
-        return (*(uint32_t *) a == *(uint32_t *) b
-                && ((uint16_t *) a)[2] == ((uint16_t *) b)[2]);
+                if (new->priority == rule->priority) {
+                    list_replace(&new->list, &rule->list);
+                    return rule;
+                } else {
+                    list_insert(&rule->list, &new->list);
+                    return NULL;
+                }
+            }
+        }
 
-    default:
-        abort();
+        /* Insert 'new' at the end of the list. */
+        list_push_back(&head->list, &new->list);
+        return NULL;
     }
-#else
-    /* I hope GCC is smarter on your platform. */
-    return !memcmp(a, b, n);
-#endif
 }
 
-/* Returns the 32-bit unsigned integer at 'p'. */
-static inline uint32_t
-read_uint32(const void *p)
+static struct cls_rule *
+next_rule_in_list(struct cls_rule *rule)
 {
-    /* GCC optimizes this into a single machine instruction on x86. */
-    uint32_t x;
-    memcpy(&x, p, sizeof x);
-    return x;
+    struct cls_rule *next = OBJECT_CONTAINING(rule->list.next, next, list);
+    return next->priority < rule->priority ? next : NULL;
 }
 
-/* Compares the specified field in 'a' and 'b'.  Returns true if the fields are
- * equal, or if the ofp_match wildcard bits in 'wildcards' are set such that
- * non-equal values may be ignored.  'nw_src_mask' and 'nw_dst_mask' must be
- * those that would be set for 'wildcards' by cls_rule_set_masks().
- *
- * The compared field is the one with wildcard bit or bits 'field_wc', offset
- * 'rule_ofs' within cls_rule's "fields" member, and length 'len', in bytes. */
-static inline bool ALWAYS_INLINE
-field_matches(const struct flow *a_, const struct flow *b_,
-              uint32_t wildcards, uint32_t nw_src_mask, uint32_t nw_dst_mask,
-              uint32_t field_wc, int ofs, int len)
-{
-    /* This function is always inlined, and it is always called with 'field_wc'
-     * as a compile-time constant, so the "if" conditionals here generate no
-     * code. */
-    const void *a = (const uint8_t *) a_ + ofs;
-    const void *b = (const uint8_t *) b_ + ofs;
-    if (!(field_wc & (field_wc - 1))) {
-        /* Handle all the single-bit wildcard cases. */
-        return wildcards & field_wc || equal_bytes(a, b, len);
-    } else if (field_wc == OFPFW_NW_SRC_MASK ||
-               field_wc == OFPFW_NW_DST_MASK) {
-        uint32_t a_ip = read_uint32(a);
-        uint32_t b_ip = read_uint32(b);
-        uint32_t mask = (field_wc == OFPFW_NW_SRC_MASK
-                         ? nw_src_mask : nw_dst_mask);
-        return ((a_ip ^ b_ip) & mask) == 0;
-    } else {
-        abort();
-    }
-}
-
-/* Returns true if 'a' and 'b' match, ignoring fields for which the wildcards
- * in 'wildcards' are set.  'nw_src_mask' and 'nw_dst_mask' must be those that
- * would be set for 'wildcards' by cls_rule_set_masks().  'field_idx' is the
- * index of the first field to be compared; fields before 'field_idx' are
- * assumed to match.  (Always returns true if 'field_idx' is CLS_N_FIELDS.) */
-static bool
-rules_match(const struct cls_rule *a, const struct cls_rule *b,
-            uint32_t wildcards, uint32_t nw_src_mask, uint32_t nw_dst_mask,
-            int field_idx)
-{
-    /* This is related to Duff's device (see
-     * http://en.wikipedia.org/wiki/Duff's_device).  */
-    switch (field_idx) {
-#define CLS_FIELD(WILDCARDS, MEMBER, NAME)                          \
-        case CLS_F_IDX_##NAME:                                      \
-            if (!field_matches(&a->flow, &b->flow,                  \
-                               wildcards, nw_src_mask, nw_dst_mask, \
-                               WILDCARDS, offsetof(struct flow, MEMBER), \
-                               sizeof a->flow.MEMBER)) {            \
-                return false;                                       \
-            }                                                       \
-        /* Fall though */
-        CLS_FIELDS
-#undef CLS_FIELD
-    }
-    return true;
-}
-
-/* Returns true if 'fixed' and 'wild' match.  All fields in 'fixed' must have
- * fixed values; 'wild' may contain wildcards.
- *
- * 'field_idx' is the index of the first field to be compared; fields before
- * 'field_idx' are assumed to match.  Always returns true if 'field_idx' is
- * CLS_N_FIELDS. */
 static bool
-rules_match_1wild(const struct cls_rule *fixed, const struct cls_rule *wild,
-                  int field_idx)
+flow_equal_except(const struct flow *a, const struct flow *b,
+                  const struct flow_wildcards *wildcards)
 {
-    return rules_match(fixed, wild, wild->wc.wildcards, wild->wc.nw_src_mask,
-                       wild->wc.nw_dst_mask, field_idx);
-}
+    const uint32_t wc = wildcards->wildcards;
 
-/* Returns true if 'wild1' and 'wild2' match, that is, if their fields
- * are equal modulo wildcards in 'wild1' or 'wild2'.
- *
- * 'field_idx' is the index of the first field to be compared; fields before
- * 'field_idx' are assumed to match.  Always returns true if 'field_idx' is
- * CLS_N_FIELDS. */
-static bool
-rules_match_2wild(const struct cls_rule *wild1, const struct cls_rule *wild2,
-                  int field_idx)
-{
-    return rules_match(wild1, wild2,
-                       wild1->wc.wildcards | wild2->wc.wildcards,
-                       wild1->wc.nw_src_mask & wild2->wc.nw_src_mask,
-                       wild1->wc.nw_dst_mask & wild2->wc.nw_dst_mask,
-                       field_idx);
+    BUILD_ASSERT_DECL(FLOW_SIG_SIZE == 37);
+
+    return ((wc & NXFW_TUN_ID || a->tun_id == b->tun_id)
+            && !((a->nw_src ^ b->nw_src) & wildcards->nw_src_mask)
+            && !((a->nw_dst ^ b->nw_dst) & wildcards->nw_dst_mask)
+            && (wc & OFPFW_IN_PORT || a->in_port == b->in_port)
+            && (wc & OFPFW_DL_VLAN || a->dl_vlan == b->dl_vlan)
+            && (wc & OFPFW_DL_TYPE || a->dl_type == b->dl_type)
+            && (wc & OFPFW_TP_SRC || a->tp_src == b->tp_src)
+            && (wc & OFPFW_TP_DST || a->tp_dst == b->tp_dst)
+            && (wc & OFPFW_DL_SRC || eth_addr_equals(a->dl_src, b->dl_src))
+            && (wc & OFPFW_DL_DST || eth_addr_equals(a->dl_dst, b->dl_dst))
+            && (wc & OFPFW_NW_PROTO || a->nw_proto == b->nw_proto)
+            && (wc & OFPFW_DL_VLAN_PCP || a->dl_vlan_pcp == b->dl_vlan_pcp)
+            && (wc & OFPFW_NW_TOS || a->nw_tos == b->nw_tos));
 }
 
-/* Searches 'bucket' for a rule that matches 'target'.  Returns the
- * highest-priority match, if one is found, or a null pointer if there is no
- * match.
- *
- * 'field_idx' must be the index of the first wildcarded field in 'bucket'. */
-static struct cls_rule *
-search_bucket(struct cls_bucket *bucket, int field_idx,
-              const struct cls_rule *target)
+static void
+zero_wildcards(struct flow *flow, const struct flow_wildcards *wildcards)
 {
-    struct cls_rule *pos;
+    const uint32_t wc = wildcards->wildcards;
 
-    if (!equal_fields(&bucket->fixed, &target->flow, field_idx)) {
-        return NULL;
-    }
+    BUILD_ASSERT_DECL(FLOW_SIG_SIZE == 37);
 
-    LIST_FOR_EACH (pos, node.list, &bucket->rules) {
-        if (rules_match_1wild(target, pos, field_idx)) {
-            return pos;
-        }
+    if (wc & NXFW_TUN_ID) {
+        flow->tun_id = 0;
     }
-    return NULL;
-}
-
-/* Searches 'table' for a rule that matches 'target'.  Returns the
- * highest-priority match, if one is found, or a null pointer if there is no
- * match.
- *
- * 'field_idx' must be the index of the first wildcarded field in 'table'. */
-static struct cls_rule *
-search_table(const struct hmap *table, int field_idx,
-             const struct cls_rule *target)
-{
-    struct cls_bucket *bucket;
-
-    switch (hmap_count(table)) {
-        /* In these special cases there's no need to hash.  */
-    case 0:
-        return NULL;
-    case 1:
-        bucket = CONTAINER_OF(hmap_first(table), struct cls_bucket, hmap_node);
-        return search_bucket(bucket, field_idx, target);
+    flow->nw_src &= wildcards->nw_src_mask;
+    flow->nw_dst &= wildcards->nw_dst_mask;
+    if (wc & OFPFW_IN_PORT) {
+        flow->in_port = 0;
     }
-
-    HMAP_FOR_EACH_WITH_HASH (bucket, hmap_node,
-                             hash_fields(&target->flow, field_idx), table) {
-        struct cls_rule *rule = search_bucket(bucket, field_idx, target);
-        if (rule) {
-            return rule;
-        }
+    if (wc & OFPFW_DL_VLAN) {
+        flow->dl_vlan = 0;
     }
-    return NULL;
-}
-
-static struct cls_rule *
-search_exact_table(const struct classifier *cls, size_t hash,
-                   const struct flow *target)
-{
-    struct cls_rule *rule;
-
-    HMAP_FOR_EACH_WITH_HASH (rule, node.hmap, hash, &cls->exact_table) {
-        if (flow_equal(&rule->flow, target)) {
-            return rule;
-        }
+    if (wc & OFPFW_DL_TYPE) {
+        flow->dl_type = 0;
+    }
+    if (wc & OFPFW_TP_SRC) {
+        flow->tp_src = 0;
+    }
+    if (wc & OFPFW_TP_DST) {
+        flow->tp_dst = 0;
+    }
+    if (wc & OFPFW_DL_SRC) {
+        memset(flow->dl_src, 0, sizeof flow->dl_src);
+    }
+    if (wc & OFPFW_DL_DST) {
+        memset(flow->dl_dst, 0, sizeof flow->dl_dst);
+    }
+    if (wc & OFPFW_NW_PROTO) {
+        flow->nw_proto = 0;
+    }
+    if (wc & OFPFW_DL_VLAN_PCP) {
+        flow->dl_vlan_pcp = 0;
+    }
+    if (wc & OFPFW_NW_TOS) {
+        flow->nw_tos = 0;
     }
-    return NULL;
 }
diff --git a/lib/classifier.h b/lib/classifier.h
index d2e2b8b..91e7494 100644
--- a/lib/classifier.h
+++ b/lib/classifier.h
@@ -19,26 +19,11 @@
 
 /* Flow classifier.
  *
- * This flow classifier assumes that we can arrange the fields in a flow in an
- * order such that the set of wildcarded fields in a rule tend to fall toward
- * the end of the ordering.  That is, if field F is wildcarded, then all the
- * fields after F tend to be wildcarded as well.  If this assumption is
- * violated, then the classifier will still classify flows correctly, but its
- * performance will suffer.
- *
- * The classifier uses a collection of CLS_N_FIELDS hash tables for wildcarded
- * flows.  Each of these tables contains the flows that wildcard a given field
- * and do not wildcard any of the fields that precede F in the ordering.  The
- * key for each hash table is the value of the fields preceding F that are not
- * wildcarded.  All the flows that fall within a table and have the same key
- * are kept as a linked list ordered from highest to lowest priority.
- *
- * The classifier also maintains a separate hash table of exact-match flows.
- *
- * To search the classifier we first search the table of exact-match flows,
- * since exact-match flows always have highest priority.  If there is a match,
- * we're done.  Otherwise, we search each of the CLS_N_FIELDS hash tables in
- * turn, looking for the highest-priority match, and return it (if any).
+ * A classifier is a "struct classifier",
+ *      a hash map from a set of wildcards to a "struct cls_table",
+ *              a hash map from fixed field values to "struct cls_rule",
+ *                      which can contain a list of otherwise identical rules
+ *                      with lower priorities.
  */
 
 #include "flow.h"
@@ -47,80 +32,38 @@
 #include "openflow/nicira-ext.h"
 #include "openflow/openflow.h"
 
-/* Number of bytes of fields in a rule. */
-#define CLS_N_BYTES 37
-
-/* Fields in a rule.
- *
- * This definition sets the ordering of fields, which is important for
- * performance (see above).  To adjust the ordering, change the order of the
- * lines. */
-#define CLS_FIELDS                                          \
-    /*                           struct flow  all-caps */   \
-    /*        wildcard bit(s)    member name  name     */   \
-    /*        -----------------  -----------  -------- */   \
-    CLS_FIELD(OFPFW_IN_PORT,     in_port,     IN_PORT)      \
-    CLS_FIELD(NXFW_TUN_ID,       tun_id,      TUN_ID)       \
-    CLS_FIELD(OFPFW_DL_VLAN,     dl_vlan,     DL_VLAN)      \
-    CLS_FIELD(OFPFW_DL_VLAN_PCP, dl_vlan_pcp, DL_VLAN_PCP)  \
-    CLS_FIELD(OFPFW_DL_SRC,      dl_src,      DL_SRC)       \
-    CLS_FIELD(OFPFW_DL_DST,      dl_dst,      DL_DST)       \
-    CLS_FIELD(OFPFW_DL_TYPE,     dl_type,     DL_TYPE)      \
-    CLS_FIELD(OFPFW_NW_SRC_MASK, nw_src,      NW_SRC)       \
-    CLS_FIELD(OFPFW_NW_DST_MASK, nw_dst,      NW_DST)       \
-    CLS_FIELD(OFPFW_NW_PROTO,    nw_proto,    NW_PROTO)     \
-    CLS_FIELD(OFPFW_NW_TOS,      nw_tos,      NW_TOS)       \
-    CLS_FIELD(OFPFW_TP_SRC,      tp_src,      TP_SRC)       \
-    CLS_FIELD(OFPFW_TP_DST,      tp_dst,      TP_DST)
-
-/* Field indexes.
- *
- * (These are also indexed into struct classifier's 'tables' array.) */
-enum {
-#define CLS_FIELD(WILDCARDS, MEMBER, NAME) CLS_F_IDX_##NAME,
-    CLS_FIELDS
-#undef CLS_FIELD
-    CLS_F_IDX_EXACT,            /* Exact-match table. */
-    CLS_N_FIELDS = CLS_F_IDX_EXACT
-};
-
-/* Field information. */
-struct cls_field {
-    int ofs;                    /* Offset in struct flow. */
-    int len;                    /* Length in bytes. */
-    uint32_t wildcards;         /* OFPFW_* bit or bits for this field. */
-    const char *name;           /* Name (for debugging). */
-};
-extern const struct cls_field cls_fields[CLS_N_FIELDS + 1];
-
 /* A flow classifier. */
 struct classifier {
-    int n_rules;                /* Sum of hmap_count() over tables[]. */
-    struct hmap tables[CLS_N_FIELDS]; /* Contain cls_bucket elements. */
-    struct hmap exact_table;          /* Contain cls_rule elements. */
+    int n_rules;                /* Total number of rules. */
+    struct hmap tables;         /* Contains "struct cls_table"s.  */
 };
 
-/* A group of rules with the same fixed values for initial fields. */
-struct cls_bucket {
-    struct hmap_node hmap_node; /* Within struct classifier 'tables'. */
-    struct list rules;          /* In order from highest to lowest priority. */
-    struct flow fixed;          /* Values for fixed fields. */
+/* A set of rules that all have the same wildcarded fields. */
+struct cls_table {
+    struct hmap_node hmap_node; /* Within struct classifier 'wctables'. */
+    struct hmap rules;          /* Contains "struct cls_rule"s. */
+    struct flow_wildcards wc;   /* Wildcards for fields. */
+    int n_table_rules;          /* Number of rules, including duplicates. */
+    int n_refs;                 /* Reference count used during iteration. */
 };
 
 /* A flow classification rule.
  *
- * Use cls_rule_from_flow() or cls_rule_from_match() to initialize a cls_rule
- * or you will almost certainly not initialize 'table_idx' correctly, with
- * disastrous results! */
+ * Use one of the cls_rule_*() functions to initialize a cls_rule.
+ *
+ * The cls_rule_*() functions below maintain the following important
+ * invariant that the classifier depends on:
+ *
+ *   - If a bit or a field is wildcarded in 'wc', then the corresponding bit or
+ *     field in 'flow' is set to all-0-bits.  (The cls_rule_zero_wildcards()
+ *     function can be used to restore this invariant after adding wildcards.)
+ */
 struct cls_rule {
-    union {
-        struct list list;       /* Within struct cls_bucket 'rules'. */
-        struct hmap_node hmap;  /* Within struct classifier 'exact_table'. */
-    } node;
+    struct hmap_node hmap_node; /* Within struct cls_table 'rules'. */
+    struct list list;           /* List of identical, lower-priority rules. */
     struct flow flow;           /* All field values. */
     struct flow_wildcards wc;   /* Wildcards for fields. */
     unsigned int priority;      /* Larger numbers are higher priorities. */
-    unsigned int table_idx;     /* Index into struct classifier 'tables'. */
 };
 
 enum {
@@ -134,6 +77,9 @@ void cls_rule_from_flow(const struct flow *, uint32_t wildcards,
 void cls_rule_from_match(const struct ofp_match *, unsigned int priority,
                          bool tun_id_from_cookie, uint64_t cookie,
                          struct cls_rule *);
+
+void cls_rule_zero_wildcards(struct cls_rule *);
+
 char *cls_rule_to_string(const struct cls_rule *);
 void cls_rule_print(const struct cls_rule *);
 
@@ -160,10 +106,22 @@ void classifier_for_each_match(const struct classifier *,
 struct cls_rule *classifier_find_rule_exactly(const struct classifier *,
                                               const struct cls_rule *);
 
-#define CLASSIFIER_FOR_EACH_EXACT_RULE(RULE, MEMBER, CLS) \
-        HMAP_FOR_EACH (RULE, MEMBER.node.hmap, &(CLS)->exact_table)
+/* Iteration shorthands. */
+
+struct cls_table *classifier_exact_table(const struct classifier *);
+struct cls_rule *cls_table_first_rule(const struct cls_table *);
+struct cls_rule *cls_table_next_rule(const struct cls_table *,
+                                     const struct cls_rule *);
+
+#define CLS_TABLE_FOR_EACH_RULE(RULE, MEMBER, TABLE)                    \
+    for ((RULE) = OBJECT_CONTAINING(cls_table_first_rule(TABLE),        \
+                                    RULE, MEMBER);                      \
+         &(RULE)->MEMBER != NULL;                                       \
+         (RULE) = OBJECT_CONTAINING(cls_table_next_rule(TABLE,          \
+                                                        &(RULE)->MEMBER), \
+                                    RULE, MEMBER))
 
-#define CLASSIFIER_FOR_EACH_EXACT_RULE_SAFE(RULE, NEXT, MEMBER, CLS) \
-        HMAP_FOR_EACH_SAFE (RULE, NEXT, MEMBER.node.hmap, &(CLS)->exact_table)
+#define CLASSIFIER_FOR_EACH_EXACT_RULE(RULE, MEMBER, CLS)               \
+    CLS_TABLE_FOR_EACH_RULE (RULE, MEMBER, classifier_exact_table(CLS))
 
 #endif /* classifier.h */
diff --git a/lib/flow.c b/lib/flow.c
index e932193..f20ff16 100644
--- a/lib/flow.c
+++ b/lib/flow.c
@@ -367,12 +367,31 @@ flow_nw_bits_to_mask(uint32_t wildcards, int shift)
     return wildcards < 32 ? htonl(~((1u << wildcards) - 1)) : 0;
 }
 
+/* Return 'wildcards' in "normal form":
+ *
+ *   - Forces unknown bits to 0.
+ *
+ *   - Forces nw_src and nw_dst masks greater than 32 to exactly 32.
+ */
+static inline uint32_t
+flow_wildcards_normalize(uint32_t wildcards)
+{
+    wildcards &= wildcards & OVSFW_ALL;
+    if (wildcards & (0x20 << OFPFW_NW_SRC_SHIFT)) {
+        wildcards &= ~(0x1f << OFPFW_NW_SRC_SHIFT);
+    }
+    if (wildcards & (0x20 << OFPFW_NW_DST_SHIFT)) {
+        wildcards &= ~(0x1f << OFPFW_NW_DST_SHIFT);
+    }
+    return wildcards;
+}
+
 /* Initializes 'wc' from 'wildcards', which may be any combination of the
  * OFPFW_* and OVSFW_* wildcard bits. */
 void
 flow_wildcards_init(struct flow_wildcards *wc, uint32_t wildcards)
 {
-    wc->wildcards = wildcards & OVSFW_ALL;
+    wc->wildcards = flow_wildcards_normalize(wildcards);
     wc->nw_src_mask = flow_nw_bits_to_mask(wc->wildcards, OFPFW_NW_SRC_SHIFT);
     wc->nw_dst_mask = flow_nw_bits_to_mask(wc->wildcards, OFPFW_NW_DST_SHIFT);
 }
@@ -385,6 +404,62 @@ flow_wildcards_init_exact(struct flow_wildcards *wc)
     flow_wildcards_init(wc, 0);
 }
 
+static inline uint32_t
+combine_nw_bits(uint32_t wb1, uint32_t wb2, int shift)
+{
+    uint32_t sb1 = (wb1 >> shift) & 0x3f;
+    uint32_t sb2 = (wb2 >> shift) & 0x3f;
+    return MAX(sb1, sb2) << shift;
+}
+
+/* Initializes 'dst' as the combination of wildcards in 'src1' and 'src2'.
+ * That is, a bit or a field is wildcarded in 'dst' if it is wildcarded in
+ * 'src1' or 'src2' or both.  */
+void
+flow_wildcards_combine(struct flow_wildcards *dst,
+                       const struct flow_wildcards *src1,
+                       const struct flow_wildcards *src2)
+{
+    uint32_t wb1 = src1->wildcards;
+    uint32_t wb2 = src2->wildcards;
+
+    dst->wildcards = (wb1 | wb2) & ~(OFPFW_NW_SRC_MASK | OFPFW_NW_DST_MASK);
+    dst->wildcards |= combine_nw_bits(wb1, wb2, OFPFW_NW_SRC_SHIFT);
+    dst->wildcards |= combine_nw_bits(wb1, wb2, OFPFW_NW_DST_SHIFT);
+    dst->nw_src_mask = src1->nw_src_mask & src2->nw_src_mask;
+    dst->nw_dst_mask = src1->nw_dst_mask & src2->nw_dst_mask;
+}
+
+/* Returns a hash of the wildcards in 'wc'. */
+uint32_t
+flow_wildcards_hash(const struct flow_wildcards *wc)
+{
+    /* There is no need to include nw_src_mask or nw_dst_mask because they do
+     * not add any information (they can be computed from wc->wildcards).  */
+    return hash_int(wc->wildcards, 0);
+}
+
+/* Returns true if 'a' and 'b' represent the same wildcards, false if they are
+ * different. */
+bool
+flow_wildcards_equal(const struct flow_wildcards *a,
+                     const struct flow_wildcards *b)
+{
+    return a->wildcards == b->wildcards;
+}
+
+/* Returns true if at least one bit or field is wildcarded in 'a' but not in
+ * 'b', false otherwise. */
+bool
+flow_wildcards_has_extra(const struct flow_wildcards *a,
+                         const struct flow_wildcards *b)
+{
+#define OFPFW_NW_MASK (OFPFW_NW_SRC_MASK | OFPFW_NW_DST_MASK)
+    return ((a->wildcards & ~(b->wildcards | OFPFW_NW_MASK))
+            || (a->nw_src_mask & b->nw_src_mask) != b->nw_src_mask
+            || (a->nw_dst_mask & b->nw_dst_mask) != b->nw_dst_mask);
+}
+
 static int
 count_ones(ovs_be32 mask)
 {
diff --git a/lib/flow.h b/lib/flow.h
index 7667cb0..67f4827 100644
--- a/lib/flow.h
+++ b/lib/flow.h
@@ -88,7 +88,23 @@ flow_hash(const struct flow *flow, uint32_t basis)
     return hash_bytes(flow, FLOW_SIG_SIZE, basis);
 }
 
-/* Information on wildcards for a flow, as a supplement to struct flow. */
+/* Information on wildcards for a flow, as a supplement to struct flow.
+ *
+ * The flow_wildcards_*() functions below both depend on and maintain the
+ * following important invariants:
+ *
+ * 1. 'wildcards' is nonzero if and only if at least one bit or field is
+ *    wildcarded.
+ *
+ * 2. Bits in 'wildcards' not included in OVSFW_ALL are set to 0.  (This is a
+ *    corollary to invariant #1.)
+ *
+ * 3. The fields in 'wildcards' masked by OFPFW_NW_SRC_MASK and
+ *    OFPFW_NW_DST_MASK have values between 0 and 32, inclusive.
+ *
+ * 4. The fields masked by OFPFW_NW_SRC_MASK and OFPFW_NW_DST_MASK correspond
+ *    correctly to the masks in 'nw_src_mask' and 'nw_dst_mask', respectively.
+ */
 struct flow_wildcards {
     uint32_t wildcards;         /* enum ofp_flow_wildcards. */
     ovs_be32 nw_src_mask;       /* 1-bit in each significant nw_src bit. */
@@ -102,4 +118,14 @@ void flow_wildcards_init_exact(struct flow_wildcards *);
 bool flow_wildcards_set_nw_src_mask(struct flow_wildcards *, ovs_be32);
 bool flow_wildcards_set_nw_dst_mask(struct flow_wildcards *, ovs_be32);
 
+void flow_wildcards_combine(struct flow_wildcards *dst,
+                            const struct flow_wildcards *src1,
+                            const struct flow_wildcards *src2);
+bool flow_wildcards_has_extra(const struct flow_wildcards *,
+                              const struct flow_wildcards *);
+
+uint32_t flow_wildcards_hash(const struct flow_wildcards *);
+bool flow_wildcards_equal(const struct flow_wildcards *,
+                          const struct flow_wildcards *);
+
 #endif /* flow.h */
diff --git a/tests/classifier.at b/tests/classifier.at
index f9e6953..26ca079 100644
--- a/tests/classifier.at
+++ b/tests/classifier.at
@@ -5,12 +5,10 @@ m4_foreach(
    [destroy-null],
    [single-rule],
    [rule-replacement],
-   [two-rules-in-one-bucket],
-   [two-rules-in-one-table],
-   [two-rules-in-different-tables],
-   [many-rules-in-one-bucket],
+   [many-rules-in-one-list],
    [many-rules-in-one-table],
-   [many-rules-in-different-tables]],
+   [many-rules-in-two-tables],
+   [many-rules-in-five-tables]],
   [AT_SETUP([flow classifier - m4_bpatsubst(testname, [-], [ ])])
    AT_CHECK([test-classifier testname], [0], [], [])
    AT_CLEANUP])])
diff --git a/tests/test-classifier.c b/tests/test-classifier.c
index 09853fc..4eccafb 100644
--- a/tests/test-classifier.c
+++ b/tests/test-classifier.c
@@ -37,6 +37,53 @@
 #undef NDEBUG
 #include <assert.h>
 
+/* Fields in a rule. */
+#define CLS_FIELDS                                          \
+    /*                           struct flow  all-caps */   \
+    /*        wildcard bit(s)    member name  name     */   \
+    /*        -----------------  -----------  -------- */   \
+    CLS_FIELD(NXFW_TUN_ID,       tun_id,      TUN_ID)       \
+    CLS_FIELD(OFPFW_NW_SRC_MASK, nw_src,      NW_SRC)       \
+    CLS_FIELD(OFPFW_NW_DST_MASK, nw_dst,      NW_DST)       \
+    CLS_FIELD(OFPFW_IN_PORT,     in_port,     IN_PORT)      \
+    CLS_FIELD(OFPFW_DL_VLAN,     dl_vlan,     DL_VLAN)      \
+    CLS_FIELD(OFPFW_DL_TYPE,     dl_type,     DL_TYPE)      \
+    CLS_FIELD(OFPFW_TP_SRC,      tp_src,      TP_SRC)       \
+    CLS_FIELD(OFPFW_TP_DST,      tp_dst,      TP_DST)       \
+    CLS_FIELD(OFPFW_DL_SRC,      dl_src,      DL_SRC)       \
+    CLS_FIELD(OFPFW_DL_DST,      dl_dst,      DL_DST)       \
+    CLS_FIELD(OFPFW_NW_PROTO,    nw_proto,    NW_PROTO)     \
+    CLS_FIELD(OFPFW_DL_VLAN_PCP, dl_vlan_pcp, DL_VLAN_PCP)  \
+    CLS_FIELD(OFPFW_NW_TOS,      nw_tos,      NW_TOS)
+
+/* Field indexes.
+ *
+ * (These are also indexed into struct classifier's 'tables' array.) */
+enum {
+#define CLS_FIELD(WILDCARDS, MEMBER, NAME) CLS_F_IDX_##NAME,
+    CLS_FIELDS
+#undef CLS_FIELD
+    CLS_N_FIELDS
+};
+
+/* Field information. */
+struct cls_field {
+    int ofs;                    /* Offset in struct flow. */
+    int len;                    /* Length in bytes. */
+    uint32_t wildcards;         /* OFPFW_* bit or bits for this field. */
+    const char *name;           /* Name (for debugging). */
+};
+
+static const struct cls_field cls_fields[CLS_N_FIELDS] = {
+#define CLS_FIELD(WILDCARDS, MEMBER, NAME)      \
+    { offsetof(struct flow, MEMBER),            \
+      sizeof ((struct flow *)0)->MEMBER,        \
+      WILDCARDS,                                \
+      #NAME },
+    CLS_FIELDS
+#undef CLS_FIELD
+};
+
 struct test_rule {
     int aux;                    /* Auxiliary data. */
     struct cls_rule cls_rule;   /* Classifier rule data. */
@@ -214,6 +261,7 @@ tcls_delete_matches(struct tcls *cls,
         struct test_rule *pos = cls->rules[i];
         uint32_t wildcards = pos->cls_rule.wc.wildcards;
         if (include & (wildcards ? CLS_INC_WILD : CLS_INC_EXACT)
+            && !flow_wildcards_has_extra(&pos->cls_rule.wc, &target->wc)
             && match(target, &pos->cls_rule.flow)) {
             tcls_remove(cls, pos);
         } else {
@@ -391,35 +439,38 @@ destroy_classifier(struct classifier *cls)
 
 static void
 check_tables(const struct classifier *cls,
-             int n_tables, int n_buckets, int n_rules)
+             int n_tables, int n_rules, int n_dups)
 {
+    const struct cls_table *table;
     int found_tables = 0;
-    int found_buckets = 0;
     int found_rules = 0;
-    int i;
+    int found_dups = 0;
 
-    BUILD_ASSERT(CLS_N_FIELDS == ARRAY_SIZE(cls->tables));
-    for (i = 0; i < CLS_N_FIELDS; i++) {
-        const struct cls_bucket *bucket;
-        if (!hmap_is_empty(&cls->tables[i])) {
-            found_tables++;
-        }
-        HMAP_FOR_EACH (bucket, hmap_node, &cls->tables[i]) {
-            found_buckets++;
-            assert(!list_is_empty(&bucket->rules));
-            found_rules += list_size(&bucket->rules);
-        }
-    }
+    HMAP_FOR_EACH (table, hmap_node, &cls->tables) {
+        const struct cls_rule *head;
+
+        assert(!hmap_is_empty(&table->rules));
 
-    if (!hmap_is_empty(&cls->exact_table)) {
         found_tables++;
-        found_buckets++;
-        found_rules += hmap_count(&cls->exact_table);
+        HMAP_FOR_EACH (head, hmap_node, &table->rules) {
+            unsigned int prev_priority = UINT_MAX;
+            const struct cls_rule *rule;
+
+            found_rules++;
+            LIST_FOR_EACH (rule, list, &head->list) {
+                assert(rule->priority < prev_priority);
+                prev_priority = rule->priority;
+                found_rules++;
+                found_dups++;
+                assert(classifier_find_rule_exactly(cls, rule) == rule);
+            }
+        }
     }
 
-    assert(n_tables == -1 || found_tables == n_tables);
+    assert(found_tables == hmap_count(&cls->tables));
+    assert(n_tables == -1 || n_tables == hmap_count(&cls->tables));
     assert(n_rules == -1 || found_rules == n_rules);
-    assert(n_buckets == -1 || found_buckets == n_buckets);
+    assert(n_dups == -1 || found_dups == n_dups);
 }
 
 static struct test_rule *
@@ -501,7 +552,7 @@ test_single_rule(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
 
         tcls_rule = tcls_insert(&tcls, rule);
         assert(!classifier_insert(&cls, &rule->cls_rule));
-        check_tables(&cls, 1, 1, 1);
+        check_tables(&cls, 1, 1, 0);
         compare_classifiers(&cls, &tcls);
 
         classifier_remove(&cls, &rule->cls_rule);
@@ -537,7 +588,7 @@ test_rule_replacement(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
         tcls_init(&tcls);
         tcls_insert(&tcls, rule1);
         assert(!classifier_insert(&cls, &rule1->cls_rule));
-        check_tables(&cls, 1, 1, 1);
+        check_tables(&cls, 1, 1, 0);
         compare_classifiers(&cls, &tcls);
         tcls_destroy(&tcls);
 
@@ -546,7 +597,7 @@ test_rule_replacement(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
         assert(test_rule_from_cls_rule(
                    classifier_insert(&cls, &rule2->cls_rule)) == rule1);
         free(rule1);
-        check_tables(&cls, 1, 1, 1);
+        check_tables(&cls, 1, 1, 0);
         compare_classifiers(&cls, &tcls);
         tcls_destroy(&tcls);
         destroy_classifier(&cls);
@@ -554,354 +605,248 @@ test_rule_replacement(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
 }
 
 static int
-table_mask(int table)
+factorial(int n_items)
 {
-    return ((1u << CLS_N_FIELDS) - 1) & ~((1u << table) - 1);
+    int n, i;
+
+    n = 1;
+    for (i = 2; i <= n_items; i++) {
+        n *= i;
+    }
+    return n;
 }
 
-static int
-random_wcf_in_table(int table, int seed)
+static void
+swap(int *a, int *b)
 {
-    int wc_fields = (1u << table) | hash_int(seed, 0);
-    return wc_fields & table_mask(table);
+    int tmp = *a;
+    *a = *b;
+    *b = tmp;
 }
 
-/* Tests classification with two rules at a time that fall into the same
- * bucket. */
 static void
-test_two_rules_in_one_bucket(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
+reverse(int *a, int n)
 {
-    int table, rel_pri, wcf_pat, value_pat;
-
-    for (table = 0; table <= CLS_N_FIELDS; table++) {
-        for (rel_pri = -1; rel_pri <= +1; rel_pri++) {
-            for (wcf_pat = 0; wcf_pat < 4; wcf_pat++) {
-                int n_value_pats = table == CLS_N_FIELDS - 1 ? 1 : 2;
-                for (value_pat = 0; value_pat < n_value_pats; value_pat++) {
-                    struct test_rule *rule1, *tcls_rule1;
-                    struct test_rule *rule2, *tcls_rule2;
-                    struct test_rule *displaced_rule;
-                    struct classifier cls;
-                    struct tcls tcls;
-                    unsigned int pri1, pri2;
-                    int wcf1, wcf2;
-
-                    if (table != CLS_F_IDX_EXACT) {
-                        /* We can use identical priorities in this test because
-                         * the classifier always chooses the rule added later
-                         * for equal-priority rules that fall into the same
-                         * bucket.  */
-                        pri1 = table * 257 + 50;
-                        pri2 = pri1 + rel_pri;
-
-                        wcf1 = (wcf_pat & 1
-                                ? random_wcf_in_table(table, pri1)
-                                : 1u << table);
-                        wcf2 = (wcf_pat & 2
-                                ? random_wcf_in_table(table, pri2)
-                                : 1u << table);
-                        if (value_pat) {
-                            wcf1 &= ~(1u << (CLS_N_FIELDS - 1));
-                            wcf2 &= ~(1u << (CLS_N_FIELDS - 1));
-                        }
-                    } else {
-                        /* This classifier always puts exact-match rules at
-                         * maximum priority.  */
-                        pri1 = pri2 = UINT_MAX;
-
-                        /* No wildcard fields. */
-                        wcf1 = wcf2 = 0;
-                    }
-
-                    rule1 = make_rule(wcf1, pri1, 0);
-                    rule2 = make_rule(wcf2, pri2,
-                                      value_pat << (CLS_N_FIELDS - 1));
-
-                    classifier_init(&cls);
-                    tcls_init(&tcls);
-
-                    tcls_rule1 = tcls_insert(&tcls, rule1);
-                    tcls_rule2 = tcls_insert(&tcls, rule2);
-                    assert(!classifier_insert(&cls, &rule1->cls_rule));
-                    displaced_rule = test_rule_from_cls_rule(
-                        classifier_insert(&cls, &rule2->cls_rule));
-                    if (wcf1 != wcf2 || pri1 != pri2 || value_pat) {
-                        assert(!displaced_rule);
+    int i;
 
-                        check_tables(&cls, 1, 1, 2);
-                        compare_classifiers(&cls, &tcls);
+    for (i = 0; i < n / 2; i++) {
+        int j = n - (i + 1);
+        swap(&a[i], &a[j]);
+    }
+}
 
-                        classifier_remove(&cls, &rule1->cls_rule);
-                        tcls_remove(&tcls, tcls_rule1);
-                        check_tables(&cls, 1, 1, 1);
-                        compare_classifiers(&cls, &tcls);
-                    } else {
-                        assert(displaced_rule == rule1);
-                        check_tables(&cls, 1, 1, 1);
-                        compare_classifiers(&cls, &tcls);
-                    }
-                    free(rule1);
+static bool
+next_permutation(int *a, int n)
+{
+    int k;
 
-                    classifier_remove(&cls, &rule2->cls_rule);
-                    tcls_remove(&tcls, tcls_rule2);
-                    compare_classifiers(&cls, &tcls);
-                    free(rule2);
+    for (k = n - 2; k >= 0; k--) {
+        if (a[k] < a[k + 1]) {
+            int l;
 
-                    destroy_classifier(&cls);
-                    tcls_destroy(&tcls);
+            for (l = n - 1; ; l--) {
+                if (a[l] > a[k]) {
+                    swap(&a[k], &a[l]);
+                    reverse(a + (k + 1), n - (k + 1));
+                    return true;
                 }
             }
         }
     }
+    return false;
 }
 
-/* Tests classification with two rules at a time that fall into the same
- * table but different buckets. */
+/* Tests classification with rules that have the same matching criteria. */
 static void
-test_two_rules_in_one_table(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
-{
-    int table, rel_pri, wcf_pat;
-
-    /* Skip tables 0 and CLS_F_IDX_EXACT because they have one bucket. */
-    for (table = 1; table < CLS_N_FIELDS; table++) {
-        for (rel_pri = -1; rel_pri <= +1; rel_pri++) {
-            for (wcf_pat = 0; wcf_pat < 5; wcf_pat++) {
-                struct test_rule *rule1, *tcls_rule1;
-                struct test_rule *rule2, *tcls_rule2;
-                struct classifier cls;
-                struct tcls tcls;
-                unsigned int pri1, pri2;
-                int wcf1, wcf2;
-                int value_mask, value_pat1, value_pat2;
-                int i;
-
-                /* We can use identical priorities in this test because the
-                 * classifier always chooses the rule added later for
-                 * equal-priority rules that fall into the same table.  */
-                pri1 = table * 257 + 50;
-                pri2 = pri1 + rel_pri;
-
-                if (wcf_pat & 4) {
-                    wcf1 = wcf2 = random_wcf_in_table(table, pri1);
-                } else {
-                    wcf1 = (wcf_pat & 1
-                            ? random_wcf_in_table(table, pri1)
-                            : 1u << table);
-                    wcf2 = (wcf_pat & 2
-                            ? random_wcf_in_table(table, pri2)
-                            : 1u << table);
-                }
-
-                /* Generate value patterns that will put the two rules into
-                 * different buckets. */
-                value_mask = ((1u << table) - 1);
-                value_pat1 = hash_int(pri1, 1) & value_mask;
-                i = 0;
-                do {
-                    value_pat2 = (hash_int(pri2, i++) & value_mask);
-                } while (value_pat1 == value_pat2);
-                rule1 = make_rule(wcf1, pri1, value_pat1);
-                rule2 = make_rule(wcf2, pri2, value_pat2);
-
-                classifier_init(&cls);
-                tcls_init(&tcls);
-
-                tcls_rule1 = tcls_insert(&tcls, rule1);
-                tcls_rule2 = tcls_insert(&tcls, rule2);
-                assert(!classifier_insert(&cls, &rule1->cls_rule));
-                assert(!classifier_insert(&cls, &rule2->cls_rule));
-                check_tables(&cls, 1, 2, 2);
-                compare_classifiers(&cls, &tcls);
+test_many_rules_in_one_list (int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
+{
+    enum { N_RULES = 3 };
+    int n_pris;
 
-                classifier_remove(&cls, &rule1->cls_rule);
-                tcls_remove(&tcls, tcls_rule1);
-                check_tables(&cls, 1, 1, 1);
-                compare_classifiers(&cls, &tcls);
-                free(rule1);
+    for (n_pris = N_RULES; n_pris >= 1; n_pris--) {
+        int ops[N_RULES * 2];
+        int pris[N_RULES];
+        int n_permutations;
+        int i;
 
-                classifier_remove(&cls, &rule2->cls_rule);
-                tcls_remove(&tcls, tcls_rule2);
-                compare_classifiers(&cls, &tcls);
-                free(rule2);
+        pris[0] = 0;
+        for (i = 1; i < N_RULES; i++) {
+            pris[i] = pris[i - 1] + (n_pris > i);
+        }
 
-                classifier_destroy(&cls);
-                tcls_destroy(&tcls);
-            }
+        for (i = 0; i < N_RULES * 2; i++) {
+            ops[i] = i / 2;
         }
-    }
-}
 
-/* Tests classification with two rules at a time that fall into different
- * tables. */
-static void
-test_two_rules_in_different_tables(int argc OVS_UNUSED,
-                                   char *argv[] OVS_UNUSED)
-{
-    int table1, table2, rel_pri, wcf_pat;
-
-    for (table1 = 0; table1 < CLS_N_FIELDS; table1++) {
-        for (table2 = table1 + 1; table2 <= CLS_N_FIELDS; table2++) {
-            for (rel_pri = 0; rel_pri < 2; rel_pri++) {
-                for (wcf_pat = 0; wcf_pat < 4; wcf_pat++) {
-                    struct test_rule *rule1, *tcls_rule1;
-                    struct test_rule *rule2, *tcls_rule2;
-                    struct classifier cls;
-                    struct tcls tcls;
-                    unsigned int pri1, pri2;
-                    int wcf1, wcf2;
-
-                    /* We must use unique priorities in this test because the
-                     * classifier makes the rule choice undefined for rules of
-                     * equal priority that fall into different tables.  (In
-                     * practice, lower-numbered tables win.)  */
-                    pri1 = table1 * 257 + 50;
-                    pri2 = rel_pri ? pri1 - 1 : pri1 + 1;
-
-                    wcf1 = (wcf_pat & 1
-                            ? random_wcf_in_table(table1, pri1)
-                            : 1u << table1);
-                    wcf2 = (wcf_pat & 2
-                            ? random_wcf_in_table(table2, pri2)
-                            : 1u << table2);
-
-                    if (table2 == CLS_F_IDX_EXACT) {
-                        pri2 = UINT16_MAX;
-                        wcf2 = 0;
-                    }
+        n_permutations = 0;
+        do {
+            struct test_rule *rules[N_RULES];
+            struct test_rule *tcls_rules[N_RULES];
+            int pri_rules[N_RULES];
+            struct classifier cls;
+            struct tcls tcls;
 
-                    rule1 = make_rule(wcf1, pri1, 0);
-                    rule2 = make_rule(wcf2, pri2, 0);
+            n_permutations++;
 
-                    classifier_init(&cls);
-                    tcls_init(&tcls);
+            for (i = 0; i < N_RULES; i++) {
+                rules[i] = make_rule(456, pris[i], 0);
+                tcls_rules[i] = NULL;
+                pri_rules[i] = -1;
+            }
 
-                    tcls_rule1 = tcls_insert(&tcls, rule1);
-                    tcls_rule2 = tcls_insert(&tcls, rule2);
-                    assert(!classifier_insert(&cls, &rule1->cls_rule));
-                    assert(!classifier_insert(&cls, &rule2->cls_rule));
-                    check_tables(&cls, 2, 2, 2);
-                    compare_classifiers(&cls, &tcls);
+            classifier_init(&cls);
+            tcls_init(&tcls);
 
-                    classifier_remove(&cls, &rule1->cls_rule);
-                    tcls_remove(&tcls, tcls_rule1);
-                    check_tables(&cls, 1, 1, 1);
-                    compare_classifiers(&cls, &tcls);
-                    free(rule1);
+            for (i = 0; i < ARRAY_SIZE(ops); i++) {
+                int j = ops[i];
+                int m, n;
 
-                    classifier_remove(&cls, &rule2->cls_rule);
-                    tcls_remove(&tcls, tcls_rule2);
-                    compare_classifiers(&cls, &tcls);
-                    free(rule2);
+                if (!tcls_rules[j]) {
+                    struct test_rule *displaced_rule;
+
+                    tcls_rules[j] = tcls_insert(&tcls, rules[j]);
+                    displaced_rule = test_rule_from_cls_rule(
+                        classifier_insert(&cls, &rules[j]->cls_rule));
+                    if (pri_rules[pris[j]] >= 0) {
+                        int k = pri_rules[pris[j]];
+                        assert(displaced_rule != NULL);
+                        assert(displaced_rule != rules[j]);
+                        assert(pris[j] == displaced_rule->cls_rule.priority);
+                        tcls_rules[k] = NULL;
+                    } else {
+                        assert(displaced_rule == NULL);
+                    }
+                    pri_rules[pris[j]] = j;
+                } else {
+                    classifier_remove(&cls, &rules[j]->cls_rule);
+                    tcls_remove(&tcls, tcls_rules[j]);
+                    tcls_rules[j] = NULL;
+                    pri_rules[pris[j]] = -1;
+                }
 
-                    classifier_destroy(&cls);
-                    tcls_destroy(&tcls);
+                n = 0;
+                for (m = 0; m < N_RULES; m++) {
+                    n += tcls_rules[m] != NULL;
                 }
+                check_tables(&cls, n > 0, n, n - 1);
+
+                compare_classifiers(&cls, &tcls);
             }
-        }
+
+            classifier_destroy(&cls);
+            tcls_destroy(&tcls);
+
+            for (i = 0; i < N_RULES; i++) {
+                free(rules[i]);
+            }
+        } while (next_permutation(ops, ARRAY_SIZE(ops)));
+        assert(n_permutations == (factorial(N_RULES * 2) >> N_RULES));
     }
 }
 
-/* Tests classification with many rules at a time that fall into the same
- * bucket but have unique priorities (and various wildcards). */
-static void
-test_many_rules_in_one_bucket(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
+static int
+count_ones(unsigned long int x)
 {
-    enum { MAX_RULES = 50 };
-    int iteration, table;
-
-    for (iteration = 0; iteration < 3; iteration++) {
-        for (table = 0; table <= CLS_N_FIELDS; table++) {
-            unsigned int priorities[MAX_RULES];
-            struct classifier cls;
-            struct tcls tcls;
-            int i;
+    int n = 0;
 
-            srand(hash_int(table, iteration));
-            for (i = 0; i < MAX_RULES; i++) {
-                priorities[i] = i * 129;
-            }
-            shuffle(priorities, ARRAY_SIZE(priorities));
+    while (x) {
+        x &= x - 1;
+        n++;
+    }
 
-            classifier_init(&cls);
-            tcls_init(&tcls);
+    return n;
+}
 
-            for (i = 0; i < MAX_RULES; i++) {
-                struct test_rule *rule;
-                unsigned int priority = priorities[i];
-                int wcf;
-
-                wcf = random_wcf_in_table(table, priority);
-                rule = make_rule(wcf, priority,
-                                 table == CLS_F_IDX_EXACT ? i : 1234);
-                tcls_insert(&tcls, rule);
-                assert(!classifier_insert(&cls, &rule->cls_rule));
-                check_tables(&cls, 1, 1, i + 1);
-                compare_classifiers(&cls, &tcls);
-            }
+static bool
+array_contains(int *array, int n, int value)
+{
+    int i;
 
-            destroy_classifier(&cls);
-            tcls_destroy(&tcls);
+    for (i = 0; i < n; i++) {
+        if (array[i] == value) {
+            return true;
         }
     }
+
+    return false;
 }
 
-/* Tests classification with many rules at a time that fall into the same
- * table but random buckets. */
+/* Tests classification with two rules at a time that fall into the same
+ * table but different lists. */
 static void
 test_many_rules_in_one_table(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
 {
-    enum { MAX_RULES = 50 };
-    int iteration, table;
+    int iteration;
 
-    for (iteration = 0; iteration < 3; iteration++) {
-        for (table = 0; table < CLS_N_FIELDS; table++) {
-            unsigned int priorities[MAX_RULES];
-            struct classifier cls;
-            struct tcls tcls;
-            int i;
+    for (iteration = 0; iteration < 50; iteration++) {
+        enum { N_RULES = 20 };
+        struct test_rule *rules[N_RULES];
+        struct test_rule *tcls_rules[N_RULES];
+        struct classifier cls;
+        struct tcls tcls;
+        int value_pats[N_RULES];
+        int value_mask;
+        int wcf;
+        int i;
 
-            srand(hash_int(table, iteration));
-            for (i = 0; i < MAX_RULES; i++) {
-                priorities[i] = i * 129;
-            }
-            shuffle(priorities, ARRAY_SIZE(priorities));
+        do {
+            wcf = rand() & ((1u << CLS_N_FIELDS) - 1);
+            value_mask = ~wcf & ((1u << CLS_N_FIELDS) - 1);
+        } while ((1 << count_ones(value_mask)) < N_RULES);
 
-            classifier_init(&cls);
-            tcls_init(&tcls);
+        classifier_init(&cls);
+        tcls_init(&tcls);
 
-            for (i = 0; i < MAX_RULES; i++) {
-                struct test_rule *rule;
-                unsigned int priority = priorities[i];
-                int wcf;
+        for (i = 0; i < N_RULES; i++) {
+            unsigned int priority = rand();
 
-                wcf = random_wcf_in_table(table, priority);
-                rule = make_rule(wcf, priority, hash_int(priority, 1));
-                tcls_insert(&tcls, rule);
-                assert(!classifier_insert(&cls, &rule->cls_rule));
-                check_tables(&cls, 1, -1, i + 1);
-                compare_classifiers(&cls, &tcls);
-            }
+            do {
+                value_pats[i] = rand() & value_mask;
+            } while (array_contains(value_pats, i, value_pats[i]));
 
-            destroy_classifier(&cls);
-            tcls_destroy(&tcls);
+            rules[i] = make_rule(wcf, priority, value_pats[i]);
+            tcls_rules[i] = tcls_insert(&tcls, rules[i]);
+            assert(!classifier_insert(&cls, &rules[i]->cls_rule));
+
+            check_tables(&cls, 1, i + 1, 0);
+            compare_classifiers(&cls, &tcls);
+        }
+
+        for (i = 0; i < N_RULES; i++) {
+            tcls_remove(&tcls, tcls_rules[i]);
+            classifier_remove(&cls, &rules[i]->cls_rule);
+            free(rules[i]);
+
+            check_tables(&cls, i < N_RULES - 1, N_RULES - (i + 1), 0);
+            compare_classifiers(&cls, &tcls);
         }
+
+        classifier_destroy(&cls);
+        tcls_destroy(&tcls);
     }
 }
 
-/* Tests classification with many rules at a time that fall into random buckets
- * in random tables. */
+/* Tests classification with many rules at a time that fall into random lists
+ * in 'n' tables. */
 static void
-test_many_rules_in_different_tables(int argc OVS_UNUSED,
-                                    char *argv[] OVS_UNUSED)
+test_many_rules_in_n_tables(int n_tables)
 {
     enum { MAX_RULES = 50 };
+    int wcfs[10];
     int iteration;
+    int i;
+
+    assert(n_tables < 10);
+    for (i = 0; i < n_tables; i++) {
+        do {
+            wcfs[i] = rand() & ((1u << CLS_N_FIELDS) - 1);
+        } while (array_contains(wcfs, i, wcfs[i]));
+    }
 
     for (iteration = 0; iteration < 30; iteration++) {
         unsigned int priorities[MAX_RULES];
         struct classifier cls;
         struct tcls tcls;
-        int i;
 
         srand(iteration);
         for (i = 0; i < MAX_RULES; i++) {
@@ -915,13 +860,12 @@ test_many_rules_in_different_tables(int argc OVS_UNUSED,
         for (i = 0; i < MAX_RULES; i++) {
             struct test_rule *rule;
             unsigned int priority = priorities[i];
-            int table = rand() % (CLS_N_FIELDS + 1);
-            int wcf = random_wcf_in_table(table, rand());
+            int wcf = wcfs[rand() % n_tables];
             int value_pat = rand() & ((1u << CLS_N_FIELDS) - 1);
             rule = make_rule(wcf, priority, value_pat);
             tcls_insert(&tcls, rule);
             assert(!classifier_insert(&cls, &rule->cls_rule));
-            check_tables(&cls, -1, -1, i + 1);
+            check_tables(&cls, -1, i + 1, -1);
             compare_classifiers(&cls, &tcls);
         }
 
@@ -935,6 +879,7 @@ test_many_rules_in_different_tables(int argc OVS_UNUSED,
                                       free_rule, &cls);
             tcls_delete_matches(&tcls, &rule->cls_rule, include);
             compare_classifiers(&cls, &tcls);
+            check_tables(&cls, -1, -1, -1);
             free(rule);
         }
 
@@ -942,26 +887,35 @@ test_many_rules_in_different_tables(int argc OVS_UNUSED,
         tcls_destroy(&tcls);
     }
 }
+
+static void
+test_many_rules_in_two_tables(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
+{
+    test_many_rules_in_n_tables(2);
+}
+
+static void
+test_many_rules_in_five_tables(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
+{
+    test_many_rules_in_n_tables(5);
+}
 
 static const struct command commands[] = {
     {"empty", 0, 0, test_empty},
     {"destroy-null", 0, 0, test_destroy_null},
     {"single-rule", 0, 0, test_single_rule},
     {"rule-replacement", 0, 0, test_rule_replacement},
-    {"two-rules-in-one-bucket", 0, 0, test_two_rules_in_one_bucket},
-    {"two-rules-in-one-table", 0, 0, test_two_rules_in_one_table},
-    {"two-rules-in-different-tables", 0, 0,
-     test_two_rules_in_different_tables},
-    {"many-rules-in-one-bucket", 0, 0, test_many_rules_in_one_bucket},
+    {"many-rules-in-one-list", 0, 0, test_many_rules_in_one_list},
     {"many-rules-in-one-table", 0, 0, test_many_rules_in_one_table},
-    {"many-rules-in-different-tables", 0, 0,
-     test_many_rules_in_different_tables},
+    {"many-rules-in-two-tables", 0, 0, test_many_rules_in_two_tables},
+    {"many-rules-in-five-tables", 0, 0, test_many_rules_in_five_tables},
     {NULL, 0, 0, NULL},
 };
 
 int
 main(int argc, char *argv[])
 {
+    set_program_name(argv[0]);
     init_values();
     run_command(argc - 1, argv + 1, commands);
     return 0;
-- 
1.7.1





More information about the dev mailing list