diff options
author | Andrii Nakryiko <andriin@fb.com> | 2019-05-24 14:59:00 -0400 |
---|---|---|
committer | Alexei Starovoitov <ast@kernel.org> | 2019-05-24 17:05:57 -0400 |
commit | e3b924224028c6fc31545e3812eecbe2ddbf35f6 (patch) | |
tree | f5fdfaf2fac6e21b7c7b9d044ab73cf068fa6423 | |
parent | 9db324314d29442c8bb8212dd40a3bb26f86c1c9 (diff) |
libbpf: add resizable non-thread safe internal hashmap
There is a need for fast point lookups inside libbpf for multiple use
cases (e.g., name resolution for BTF-to-C conversion, by-name lookups in
BTF for upcoming BPF CO-RE relocation support, etc). This patch
implements simple resizable non-thread safe hashmap using single linked
list chains.
Four different insert strategies are supported:
- HASHMAP_ADD - only add key/value if key doesn't exist yet;
- HASHMAP_SET - add key/value pair if key doesn't exist yet; otherwise,
update value;
- HASHMAP_UPDATE - update value, if key already exists; otherwise, do
nothing and return -ENOENT;
- HASHMAP_APPEND - always add key/value pair, even if key already exists.
This turns hashmap into a multimap by allowing multiple values to be
associated with the same key. Most useful read API for such hashmap is
hashmap__for_each_key_entry() iteration. If hashmap__find() is still
used, it will return last inserted key/value entry (first in a bucket
chain).
For HASHMAP_SET and HASHMAP_UPDATE, old key/value pair is returned, so
that calling code can handle proper memory management, if necessary.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
-rw-r--r-- | tools/lib/bpf/Build | 2 | ||||
-rw-r--r-- | tools/lib/bpf/hashmap.c | 229 | ||||
-rw-r--r-- | tools/lib/bpf/hashmap.h | 173 |
3 files changed, 403 insertions, 1 deletions
diff --git a/tools/lib/bpf/Build b/tools/lib/bpf/Build index ee9d5362f35b..dcf0d36598e0 100644 --- a/tools/lib/bpf/Build +++ b/tools/lib/bpf/Build | |||
@@ -1 +1 @@ | |||
libbpf-y := libbpf.o bpf.o nlattr.o btf.o libbpf_errno.o str_error.o netlink.o bpf_prog_linfo.o libbpf_probes.o xsk.o | libbpf-y := libbpf.o bpf.o nlattr.o btf.o libbpf_errno.o str_error.o netlink.o bpf_prog_linfo.o libbpf_probes.o xsk.o hashmap.o | ||
diff --git a/tools/lib/bpf/hashmap.c b/tools/lib/bpf/hashmap.c new file mode 100644 index 000000000000..6122272943e6 --- /dev/null +++ b/tools/lib/bpf/hashmap.c | |||
@@ -0,0 +1,229 @@ | |||
1 | // SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) | ||
2 | |||
3 | /* | ||
4 | * Generic non-thread safe hash map implementation. | ||
5 | * | ||
6 | * Copyright (c) 2019 Facebook | ||
7 | */ | ||
8 | #include <stdint.h> | ||
9 | #include <stdlib.h> | ||
10 | #include <stdio.h> | ||
11 | #include <errno.h> | ||
12 | #include <linux/err.h> | ||
13 | #include "hashmap.h" | ||
14 | |||
15 | /* start with 4 buckets */ | ||
16 | #define HASHMAP_MIN_CAP_BITS 2 | ||
17 | |||
18 | static void hashmap_add_entry(struct hashmap_entry **pprev, | ||
19 | struct hashmap_entry *entry) | ||
20 | { | ||
21 | entry->next = *pprev; | ||
22 | *pprev = entry; | ||
23 | } | ||
24 | |||
25 | static void hashmap_del_entry(struct hashmap_entry **pprev, | ||
26 | struct hashmap_entry *entry) | ||
27 | { | ||
28 | *pprev = entry->next; | ||
29 | entry->next = NULL; | ||
30 | } | ||
31 | |||
32 | void hashmap__init(struct hashmap *map, hashmap_hash_fn hash_fn, | ||
33 | hashmap_equal_fn equal_fn, void *ctx) | ||
34 | { | ||
35 | map->hash_fn = hash_fn; | ||
36 | map->equal_fn = equal_fn; | ||
37 | map->ctx = ctx; | ||
38 | |||
39 | map->buckets = NULL; | ||
40 | map->cap = 0; | ||
41 | map->cap_bits = 0; | ||
42 | map->sz = 0; | ||
43 | } | ||
44 | |||
45 | struct hashmap *hashmap__new(hashmap_hash_fn hash_fn, | ||
46 | hashmap_equal_fn equal_fn, | ||
47 | void *ctx) | ||
48 | { | ||
49 | struct hashmap *map = malloc(sizeof(struct hashmap)); | ||
50 | |||
51 | if (!map) | ||
52 | return ERR_PTR(-ENOMEM); | ||
53 | hashmap__init(map, hash_fn, equal_fn, ctx); | ||
54 | return map; | ||
55 | } | ||
56 | |||
57 | void hashmap__clear(struct hashmap *map) | ||
58 | { | ||
59 | free(map->buckets); | ||
60 | map->cap = map->cap_bits = map->sz = 0; | ||
61 | } | ||
62 | |||
63 | void hashmap__free(struct hashmap *map) | ||
64 | { | ||
65 | if (!map) | ||
66 | return; | ||
67 | |||
68 | hashmap__clear(map); | ||
69 | free(map); | ||
70 | } | ||
71 | |||
72 | size_t hashmap__size(const struct hashmap *map) | ||
73 | { | ||
74 | return map->sz; | ||
75 | } | ||
76 | |||
77 | size_t hashmap__capacity(const struct hashmap *map) | ||
78 | { | ||
79 | return map->cap; | ||
80 | } | ||
81 | |||
82 | static bool hashmap_needs_to_grow(struct hashmap *map) | ||
83 | { | ||
84 | /* grow if empty or more than 75% filled */ | ||
85 | return (map->cap == 0) || ((map->sz + 1) * 4 / 3 > map->cap); | ||
86 | } | ||
87 | |||
88 | static int hashmap_grow(struct hashmap *map) | ||
89 | { | ||
90 | struct hashmap_entry **new_buckets; | ||
91 | struct hashmap_entry *cur, *tmp; | ||
92 | size_t new_cap_bits, new_cap; | ||
93 | size_t h; | ||
94 | int bkt; | ||
95 | |||
96 | new_cap_bits = map->cap_bits + 1; | ||
97 | if (new_cap_bits < HASHMAP_MIN_CAP_BITS) | ||
98 | new_cap_bits = HASHMAP_MIN_CAP_BITS; | ||
99 | |||
100 | new_cap = 1UL << new_cap_bits; | ||
101 | new_buckets = calloc(new_cap, sizeof(new_buckets[0])); | ||
102 | if (!new_buckets) | ||
103 | return -ENOMEM; | ||
104 | |||
105 | hashmap__for_each_entry_safe(map, cur, tmp, bkt) { | ||
106 | h = hash_bits(map->hash_fn(cur->key, map->ctx), new_cap_bits); | ||
107 | hashmap_add_entry(&new_buckets[h], cur); | ||
108 | } | ||
109 | |||
110 | map->cap = new_cap; | ||
111 | map->cap_bits = new_cap_bits; | ||
112 | free(map->buckets); | ||
113 | map->buckets = new_buckets; | ||
114 | |||
115 | return 0; | ||
116 | } | ||
117 | |||
118 | static bool hashmap_find_entry(const struct hashmap *map, | ||
119 | const void *key, size_t hash, | ||
120 | struct hashmap_entry ***pprev, | ||
121 | struct hashmap_entry **entry) | ||
122 | { | ||
123 | struct hashmap_entry *cur, **prev_ptr; | ||
124 | |||
125 | if (!map->buckets) | ||
126 | return false; | ||
127 | |||
128 | for (prev_ptr = &map->buckets[hash], cur = *prev_ptr; | ||
129 | cur; | ||
130 | prev_ptr = &cur->next, cur = cur->next) { | ||
131 | if (map->equal_fn(cur->key, key, map->ctx)) { | ||
132 | if (pprev) | ||
133 | *pprev = prev_ptr; | ||
134 | *entry = cur; | ||
135 | return true; | ||
136 | } | ||
137 | } | ||
138 | |||
139 | return false; | ||
140 | } | ||
141 | |||
142 | int hashmap__insert(struct hashmap *map, const void *key, void *value, | ||
143 | enum hashmap_insert_strategy strategy, | ||
144 | const void **old_key, void **old_value) | ||
145 | { | ||
146 | struct hashmap_entry *entry; | ||
147 | size_t h; | ||
148 | int err; | ||
149 | |||
150 | if (old_key) | ||
151 | *old_key = NULL; | ||
152 | if (old_value) | ||
153 | *old_value = NULL; | ||
154 | |||
155 | h = hash_bits(map->hash_fn(key, map->ctx), map->cap_bits); | ||
156 | if (strategy != HASHMAP_APPEND && | ||
157 | hashmap_find_entry(map, key, h, NULL, &entry)) { | ||
158 | if (old_key) | ||
159 | *old_key = entry->key; | ||
160 | if (old_value) | ||
161 | *old_value = entry->value; | ||
162 | |||
163 | if (strategy == HASHMAP_SET || strategy == HASHMAP_UPDATE) { | ||
164 | entry->key = key; | ||
165 | entry->value = value; | ||
166 | return 0; | ||
167 | } else if (strategy == HASHMAP_ADD) { | ||
168 | return -EEXIST; | ||
169 | } | ||
170 | } | ||
171 | |||
172 | if (strategy == HASHMAP_UPDATE) | ||
173 | return -ENOENT; | ||
174 | |||
175 | if (hashmap_needs_to_grow(map)) { | ||
176 | err = hashmap_grow(map); | ||
177 | if (err) | ||
178 | return err; | ||
179 | h = hash_bits(map->hash_fn(key, map->ctx), map->cap_bits); | ||
180 | } | ||
181 | |||
182 | entry = malloc(sizeof(struct hashmap_entry)); | ||
183 | if (!entry) | ||
184 | return -ENOMEM; | ||
185 | |||
186 | entry->key = key; | ||
187 | entry->value = value; | ||
188 | hashmap_add_entry(&map->buckets[h], entry); | ||
189 | map->sz++; | ||
190 | |||
191 | return 0; | ||
192 | } | ||
193 | |||
194 | bool hashmap__find(const struct hashmap *map, const void *key, void **value) | ||
195 | { | ||
196 | struct hashmap_entry *entry; | ||
197 | size_t h; | ||
198 | |||
199 | h = hash_bits(map->hash_fn(key, map->ctx), map->cap_bits); | ||
200 | if (!hashmap_find_entry(map, key, h, NULL, &entry)) | ||
201 | return false; | ||
202 | |||
203 | if (value) | ||
204 | *value = entry->value; | ||
205 | return true; | ||
206 | } | ||
207 | |||
208 | bool hashmap__delete(struct hashmap *map, const void *key, | ||
209 | const void **old_key, void **old_value) | ||
210 | { | ||
211 | struct hashmap_entry **pprev, *entry; | ||
212 | size_t h; | ||
213 | |||
214 | h = hash_bits(map->hash_fn(key, map->ctx), map->cap_bits); | ||
215 | if (!hashmap_find_entry(map, key, h, &pprev, &entry)) | ||
216 | return false; | ||
217 | |||
218 | if (old_key) | ||
219 | *old_key = entry->key; | ||
220 | if (old_value) | ||
221 | *old_value = entry->value; | ||
222 | |||
223 | hashmap_del_entry(pprev, entry); | ||
224 | free(entry); | ||
225 | map->sz--; | ||
226 | |||
227 | return true; | ||
228 | } | ||
229 | |||
diff --git a/tools/lib/bpf/hashmap.h b/tools/lib/bpf/hashmap.h new file mode 100644 index 000000000000..03748a742146 --- /dev/null +++ b/tools/lib/bpf/hashmap.h | |||
@@ -0,0 +1,173 @@ | |||
1 | /* SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) */ | ||
2 | |||
3 | /* | ||
4 | * Generic non-thread safe hash map implementation. | ||
5 | * | ||
6 | * Copyright (c) 2019 Facebook | ||
7 | */ | ||
8 | #ifndef __LIBBPF_HASHMAP_H | ||
9 | #define __LIBBPF_HASHMAP_H | ||
10 | |||
11 | #include <stdbool.h> | ||
12 | #include <stddef.h> | ||
13 | #include "libbpf_internal.h" | ||
14 | |||
15 | static inline size_t hash_bits(size_t h, int bits) | ||
16 | { | ||
17 | /* shuffle bits and return requested number of upper bits */ | ||
18 | return (h * 11400714819323198485llu) >> (__WORDSIZE - bits); | ||
19 | } | ||
20 | |||
21 | typedef size_t (*hashmap_hash_fn)(const void *key, void *ctx); | ||
22 | typedef bool (*hashmap_equal_fn)(const void *key1, const void *key2, void *ctx); | ||
23 | |||
24 | struct hashmap_entry { | ||
25 | const void *key; | ||
26 | void *value; | ||
27 | struct hashmap_entry *next; | ||
28 | }; | ||
29 | |||
30 | struct hashmap { | ||
31 | hashmap_hash_fn hash_fn; | ||
32 | hashmap_equal_fn equal_fn; | ||
33 | void *ctx; | ||
34 | |||
35 | struct hashmap_entry **buckets; | ||
36 | size_t cap; | ||
37 | size_t cap_bits; | ||
38 | size_t sz; | ||
39 | }; | ||
40 | |||
41 | #define HASHMAP_INIT(hash_fn, equal_fn, ctx) { \ | ||
42 | .hash_fn = (hash_fn), \ | ||
43 | .equal_fn = (equal_fn), \ | ||
44 | .ctx = (ctx), \ | ||
45 | .buckets = NULL, \ | ||
46 | .cap = 0, \ | ||
47 | .cap_bits = 0, \ | ||
48 | .sz = 0, \ | ||
49 | } | ||
50 | |||
51 | void hashmap__init(struct hashmap *map, hashmap_hash_fn hash_fn, | ||
52 | hashmap_equal_fn equal_fn, void *ctx); | ||
53 | struct hashmap *hashmap__new(hashmap_hash_fn hash_fn, | ||
54 | hashmap_equal_fn equal_fn, | ||
55 | void *ctx); | ||
56 | void hashmap__clear(struct hashmap *map); | ||
57 | void hashmap__free(struct hashmap *map); | ||
58 | |||
59 | size_t hashmap__size(const struct hashmap *map); | ||
60 | size_t hashmap__capacity(const struct hashmap *map); | ||
61 | |||
62 | /* | ||
63 | * Hashmap insertion strategy: | ||
64 | * - HASHMAP_ADD - only add key/value if key doesn't exist yet; | ||
65 | * - HASHMAP_SET - add key/value pair if key doesn't exist yet; otherwise, | ||
66 | * update value; | ||
67 | * - HASHMAP_UPDATE - update value, if key already exists; otherwise, do | ||
68 | * nothing and return -ENOENT; | ||
69 | * - HASHMAP_APPEND - always add key/value pair, even if key already exists. | ||
70 | * This turns hashmap into a multimap by allowing multiple values to be | ||
71 | * associated with the same key. Most useful read API for such hashmap is | ||
72 | * hashmap__for_each_key_entry() iteration. If hashmap__find() is still | ||
73 | * used, it will return last inserted key/value entry (first in a bucket | ||
74 | * chain). | ||
75 | */ | ||
76 | enum hashmap_insert_strategy { | ||
77 | HASHMAP_ADD, | ||
78 | HASHMAP_SET, | ||
79 | HASHMAP_UPDATE, | ||
80 | HASHMAP_APPEND, | ||
81 | }; | ||
82 | |||
83 | /* | ||
84 | * hashmap__insert() adds key/value entry w/ various semantics, depending on | ||
85 | * provided strategy value. If a given key/value pair replaced already | ||
86 | * existing key/value pair, both old key and old value will be returned | ||
87 | * through old_key and old_value to allow calling code do proper memory | ||
88 | * management. | ||
89 | */ | ||
90 | int hashmap__insert(struct hashmap *map, const void *key, void *value, | ||
91 | enum hashmap_insert_strategy strategy, | ||
92 | const void **old_key, void **old_value); | ||
93 | |||
94 | static inline int hashmap__add(struct hashmap *map, | ||
95 | const void *key, void *value) | ||
96 | { | ||
97 | return hashmap__insert(map, key, value, HASHMAP_ADD, NULL, NULL); | ||
98 | } | ||
99 | |||
100 | static inline int hashmap__set(struct hashmap *map, | ||
101 | const void *key, void *value, | ||
102 | const void **old_key, void **old_value) | ||
103 | { | ||
104 | return hashmap__insert(map, key, value, HASHMAP_SET, | ||
105 | old_key, old_value); | ||
106 | } | ||
107 | |||
108 | static inline int hashmap__update(struct hashmap *map, | ||
109 | const void *key, void *value, | ||
110 | const void **old_key, void **old_value) | ||
111 | { | ||
112 | return hashmap__insert(map, key, value, HASHMAP_UPDATE, | ||
113 | old_key, old_value); | ||
114 | } | ||
115 | |||
116 | static inline int hashmap__append(struct hashmap *map, | ||
117 | const void *key, void *value) | ||
118 | { | ||
119 | return hashmap__insert(map, key, value, HASHMAP_APPEND, NULL, NULL); | ||
120 | } | ||
121 | |||
122 | bool hashmap__delete(struct hashmap *map, const void *key, | ||
123 | const void **old_key, void **old_value); | ||
124 | |||
125 | bool hashmap__find(const struct hashmap *map, const void *key, void **value); | ||
126 | |||
127 | /* | ||
128 | * hashmap__for_each_entry - iterate over all entries in hashmap | ||
129 | * @map: hashmap to iterate | ||
130 | * @cur: struct hashmap_entry * used as a loop cursor | ||
131 | * @bkt: integer used as a bucket loop cursor | ||
132 | */ | ||
133 | #define hashmap__for_each_entry(map, cur, bkt) \ | ||
134 | for (bkt = 0; bkt < map->cap; bkt++) \ | ||
135 | for (cur = map->buckets[bkt]; cur; cur = cur->next) | ||
136 | |||
137 | /* | ||
138 | * hashmap__for_each_entry_safe - iterate over all entries in hashmap, safe | ||
139 | * against removals | ||
140 | * @map: hashmap to iterate | ||
141 | * @cur: struct hashmap_entry * used as a loop cursor | ||
142 | * @tmp: struct hashmap_entry * used as a temporary next cursor storage | ||
143 | * @bkt: integer used as a bucket loop cursor | ||
144 | */ | ||
145 | #define hashmap__for_each_entry_safe(map, cur, tmp, bkt) \ | ||
146 | for (bkt = 0; bkt < map->cap; bkt++) \ | ||
147 | for (cur = map->buckets[bkt]; \ | ||
148 | cur && ({tmp = cur->next; true; }); \ | ||
149 | cur = tmp) | ||
150 | |||
151 | /* | ||
152 | * hashmap__for_each_key_entry - iterate over entries associated with given key | ||
153 | * @map: hashmap to iterate | ||
154 | * @cur: struct hashmap_entry * used as a loop cursor | ||
155 | * @key: key to iterate entries for | ||
156 | */ | ||
157 | #define hashmap__for_each_key_entry(map, cur, _key) \ | ||
158 | for (cur = ({ size_t bkt = hash_bits(map->hash_fn((_key), map->ctx),\ | ||
159 | map->cap_bits); \ | ||
160 | map->buckets ? map->buckets[bkt] : NULL; }); \ | ||
161 | cur; \ | ||
162 | cur = cur->next) \ | ||
163 | if (map->equal_fn(cur->key, (_key), map->ctx)) | ||
164 | |||
165 | #define hashmap__for_each_key_entry_safe(map, cur, tmp, _key) \ | ||
166 | for (cur = ({ size_t bkt = hash_bits(map->hash_fn((_key), map->ctx),\ | ||
167 | map->cap_bits); \ | ||
168 | cur = map->buckets ? map->buckets[bkt] : NULL; }); \ | ||
169 | cur && ({ tmp = cur->next; true; }); \ | ||
170 | cur = tmp) \ | ||
171 | if (map->equal_fn(cur->key, (_key), map->ctx)) | ||
172 | |||
173 | #endif /* __LIBBPF_HASHMAP_H */ | ||