License cleanup: add SPDX GPL-2.0 license identifier to files with no license
[linux-block.git] / tools / testing / selftests / bpf / test_lpm_map.c
CommitLineData
b2441318 1// SPDX-License-Identifier: GPL-2.0
4d3381f5
DH
2/*
3 * Randomized tests for eBPF longest-prefix-match maps
4 *
5 * This program runs randomized tests against the lpm-bpf-map. It implements a
6 * "Trivial Longest Prefix Match" (tlpm) based on simple, linear, singly linked
7 * lists. The implementation should be pretty straightforward.
8 *
9 * Based on tlpm, this inserts randomized data into bpf-lpm-maps and verifies
10 * the trie-based bpf-map implementation behaves the same way as tlpm.
11 */
12
13#include <assert.h>
14#include <errno.h>
15#include <inttypes.h>
16#include <linux/bpf.h>
17#include <stdio.h>
18#include <stdlib.h>
19#include <string.h>
20#include <time.h>
21#include <unistd.h>
22#include <arpa/inet.h>
23#include <sys/time.h>
24#include <sys/resource.h>
25
10ecc728 26#include <bpf/bpf.h>
4d3381f5
DH
27#include "bpf_util.h"
28
29struct tlpm_node {
30 struct tlpm_node *next;
31 size_t n_bits;
32 uint8_t key[];
33};
34
35static struct tlpm_node *tlpm_add(struct tlpm_node *list,
36 const uint8_t *key,
37 size_t n_bits)
38{
39 struct tlpm_node *node;
40 size_t n;
41
42 /* add new entry with @key/@n_bits to @list and return new head */
43
44 n = (n_bits + 7) / 8;
45 node = malloc(sizeof(*node) + n);
46 assert(node);
47
48 node->next = list;
49 node->n_bits = n_bits;
50 memcpy(node->key, key, n);
51
52 return node;
53}
54
55static void tlpm_clear(struct tlpm_node *list)
56{
57 struct tlpm_node *node;
58
59 /* free all entries in @list */
60
61 while ((node = list)) {
62 list = list->next;
63 free(node);
64 }
65}
66
67static struct tlpm_node *tlpm_match(struct tlpm_node *list,
68 const uint8_t *key,
69 size_t n_bits)
70{
71 struct tlpm_node *best = NULL;
72 size_t i;
73
74 /* Perform longest prefix-match on @key/@n_bits. That is, iterate all
75 * entries and match each prefix against @key. Remember the "best"
76 * entry we find (i.e., the longest prefix that matches) and return it
77 * to the caller when done.
78 */
79
80 for ( ; list; list = list->next) {
81 for (i = 0; i < n_bits && i < list->n_bits; ++i) {
82 if ((key[i / 8] & (1 << (7 - i % 8))) !=
83 (list->key[i / 8] & (1 << (7 - i % 8))))
84 break;
85 }
86
87 if (i >= list->n_bits) {
88 if (!best || i > best->n_bits)
89 best = list;
90 }
91 }
92
93 return best;
94}
95
96static void test_lpm_basic(void)
97{
98 struct tlpm_node *list = NULL, *t1, *t2;
99
100 /* very basic, static tests to verify tlpm works as expected */
101
102 assert(!tlpm_match(list, (uint8_t[]){ 0xff }, 8));
103
104 t1 = list = tlpm_add(list, (uint8_t[]){ 0xff }, 8);
105 assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff }, 8));
106 assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff, 0xff }, 16));
107 assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff, 0x00 }, 16));
108 assert(!tlpm_match(list, (uint8_t[]){ 0x7f }, 8));
109 assert(!tlpm_match(list, (uint8_t[]){ 0xfe }, 8));
110 assert(!tlpm_match(list, (uint8_t[]){ 0xff }, 7));
111
112 t2 = list = tlpm_add(list, (uint8_t[]){ 0xff, 0xff }, 16);
113 assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff }, 8));
114 assert(t2 == tlpm_match(list, (uint8_t[]){ 0xff, 0xff }, 16));
115 assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff, 0xff }, 15));
116 assert(!tlpm_match(list, (uint8_t[]){ 0x7f, 0xff }, 16));
117
118 tlpm_clear(list);
119}
120
121static void test_lpm_order(void)
122{
123 struct tlpm_node *t1, *t2, *l1 = NULL, *l2 = NULL;
124 size_t i, j;
125
126 /* Verify the tlpm implementation works correctly regardless of the
127 * order of entries. Insert a random set of entries into @l1, and copy
128 * the same data in reverse order into @l2. Then verify a lookup of
129 * random keys will yield the same result in both sets.
130 */
131
132 for (i = 0; i < (1 << 12); ++i)
133 l1 = tlpm_add(l1, (uint8_t[]){
134 rand() % 0xff,
135 rand() % 0xff,
136 }, rand() % 16 + 1);
137
138 for (t1 = l1; t1; t1 = t1->next)
139 l2 = tlpm_add(l2, t1->key, t1->n_bits);
140
141 for (i = 0; i < (1 << 8); ++i) {
142 uint8_t key[] = { rand() % 0xff, rand() % 0xff };
143
144 t1 = tlpm_match(l1, key, 16);
145 t2 = tlpm_match(l2, key, 16);
146
147 assert(!t1 == !t2);
148 if (t1) {
149 assert(t1->n_bits == t2->n_bits);
150 for (j = 0; j < t1->n_bits; ++j)
151 assert((t1->key[j / 8] & (1 << (7 - j % 8))) ==
152 (t2->key[j / 8] & (1 << (7 - j % 8))));
153 }
154 }
155
156 tlpm_clear(l1);
157 tlpm_clear(l2);
158}
159
160static void test_lpm_map(int keysize)
161{
162 size_t i, j, n_matches, n_nodes, n_lookups;
163 struct tlpm_node *t, *list = NULL;
164 struct bpf_lpm_trie_key *key;
165 uint8_t *data, *value;
166 int r, map;
167
168 /* Compare behavior of tlpm vs. bpf-lpm. Create a randomized set of
169 * prefixes and insert it into both tlpm and bpf-lpm. Then run some
170 * randomized lookups and verify both maps return the same result.
171 */
172
173 n_matches = 0;
174 n_nodes = 1 << 8;
175 n_lookups = 1 << 16;
176
177 data = alloca(keysize);
178 memset(data, 0, keysize);
179
180 value = alloca(keysize + 1);
181 memset(value, 0, keysize + 1);
182
183 key = alloca(sizeof(*key) + keysize);
184 memset(key, 0, sizeof(*key) + keysize);
185
f4874d01 186 map = bpf_create_map(BPF_MAP_TYPE_LPM_TRIE,
4d3381f5
DH
187 sizeof(*key) + keysize,
188 keysize + 1,
189 4096,
190 BPF_F_NO_PREALLOC);
191 assert(map >= 0);
192
193 for (i = 0; i < n_nodes; ++i) {
194 for (j = 0; j < keysize; ++j)
195 value[j] = rand() & 0xff;
196 value[keysize] = rand() % (8 * keysize + 1);
197
198 list = tlpm_add(list, value, value[keysize]);
199
200 key->prefixlen = value[keysize];
201 memcpy(key->data, value, keysize);
10ecc728 202 r = bpf_map_update_elem(map, key, value, 0);
4d3381f5
DH
203 assert(!r);
204 }
205
206 for (i = 0; i < n_lookups; ++i) {
207 for (j = 0; j < keysize; ++j)
208 data[j] = rand() & 0xff;
209
210 t = tlpm_match(list, data, 8 * keysize);
211
212 key->prefixlen = 8 * keysize;
213 memcpy(key->data, data, keysize);
e5ff7c40 214 r = bpf_map_lookup_elem(map, key, value);
4d3381f5
DH
215 assert(!r || errno == ENOENT);
216 assert(!t == !!r);
217
218 if (t) {
219 ++n_matches;
220 assert(t->n_bits == value[keysize]);
221 for (j = 0; j < t->n_bits; ++j)
222 assert((t->key[j / 8] & (1 << (7 - j % 8))) ==
223 (value[j / 8] & (1 << (7 - j % 8))));
224 }
225 }
226
227 close(map);
228 tlpm_clear(list);
229
230 /* With 255 random nodes in the map, we are pretty likely to match
231 * something on every lookup. For statistics, use this:
232 *
233 * printf(" nodes: %zu\n"
234 * "lookups: %zu\n"
235 * "matches: %zu\n", n_nodes, n_lookups, n_matches);
236 */
237}
238
239/* Test the implementation with some 'real world' examples */
240
241static void test_lpm_ipaddr(void)
242{
243 struct bpf_lpm_trie_key *key_ipv4;
244 struct bpf_lpm_trie_key *key_ipv6;
245 size_t key_size_ipv4;
246 size_t key_size_ipv6;
247 int map_fd_ipv4;
248 int map_fd_ipv6;
249 __u64 value;
250
251 key_size_ipv4 = sizeof(*key_ipv4) + sizeof(__u32);
252 key_size_ipv6 = sizeof(*key_ipv6) + sizeof(__u32) * 4;
253 key_ipv4 = alloca(key_size_ipv4);
254 key_ipv6 = alloca(key_size_ipv6);
255
f4874d01 256 map_fd_ipv4 = bpf_create_map(BPF_MAP_TYPE_LPM_TRIE,
4d3381f5
DH
257 key_size_ipv4, sizeof(value),
258 100, BPF_F_NO_PREALLOC);
259 assert(map_fd_ipv4 >= 0);
260
f4874d01 261 map_fd_ipv6 = bpf_create_map(BPF_MAP_TYPE_LPM_TRIE,
4d3381f5
DH
262 key_size_ipv6, sizeof(value),
263 100, BPF_F_NO_PREALLOC);
264 assert(map_fd_ipv6 >= 0);
265
266 /* Fill data some IPv4 and IPv6 address ranges */
267 value = 1;
268 key_ipv4->prefixlen = 16;
269 inet_pton(AF_INET, "192.168.0.0", key_ipv4->data);
10ecc728 270 assert(bpf_map_update_elem(map_fd_ipv4, key_ipv4, &value, 0) == 0);
4d3381f5
DH
271
272 value = 2;
273 key_ipv4->prefixlen = 24;
274 inet_pton(AF_INET, "192.168.0.0", key_ipv4->data);
10ecc728 275 assert(bpf_map_update_elem(map_fd_ipv4, key_ipv4, &value, 0) == 0);
4d3381f5
DH
276
277 value = 3;
278 key_ipv4->prefixlen = 24;
279 inet_pton(AF_INET, "192.168.128.0", key_ipv4->data);
10ecc728 280 assert(bpf_map_update_elem(map_fd_ipv4, key_ipv4, &value, 0) == 0);
4d3381f5
DH
281
282 value = 5;
283 key_ipv4->prefixlen = 24;
284 inet_pton(AF_INET, "192.168.1.0", key_ipv4->data);
10ecc728 285 assert(bpf_map_update_elem(map_fd_ipv4, key_ipv4, &value, 0) == 0);
4d3381f5
DH
286
287 value = 4;
288 key_ipv4->prefixlen = 23;
289 inet_pton(AF_INET, "192.168.0.0", key_ipv4->data);
10ecc728 290 assert(bpf_map_update_elem(map_fd_ipv4, key_ipv4, &value, 0) == 0);
4d3381f5
DH
291
292 value = 0xdeadbeef;
293 key_ipv6->prefixlen = 64;
294 inet_pton(AF_INET6, "2a00:1450:4001:814::200e", key_ipv6->data);
10ecc728 295 assert(bpf_map_update_elem(map_fd_ipv6, key_ipv6, &value, 0) == 0);
4d3381f5
DH
296
297 /* Set tprefixlen to maximum for lookups */
298 key_ipv4->prefixlen = 32;
299 key_ipv6->prefixlen = 128;
300
301 /* Test some lookups that should come back with a value */
302 inet_pton(AF_INET, "192.168.128.23", key_ipv4->data);
e5ff7c40 303 assert(bpf_map_lookup_elem(map_fd_ipv4, key_ipv4, &value) == 0);
4d3381f5
DH
304 assert(value == 3);
305
306 inet_pton(AF_INET, "192.168.0.1", key_ipv4->data);
e5ff7c40 307 assert(bpf_map_lookup_elem(map_fd_ipv4, key_ipv4, &value) == 0);
4d3381f5
DH
308 assert(value == 2);
309
310 inet_pton(AF_INET6, "2a00:1450:4001:814::", key_ipv6->data);
e5ff7c40 311 assert(bpf_map_lookup_elem(map_fd_ipv6, key_ipv6, &value) == 0);
4d3381f5
DH
312 assert(value == 0xdeadbeef);
313
314 inet_pton(AF_INET6, "2a00:1450:4001:814::1", key_ipv6->data);
e5ff7c40 315 assert(bpf_map_lookup_elem(map_fd_ipv6, key_ipv6, &value) == 0);
4d3381f5
DH
316 assert(value == 0xdeadbeef);
317
318 /* Test some lookups that should not match any entry */
319 inet_pton(AF_INET, "10.0.0.1", key_ipv4->data);
e5ff7c40 320 assert(bpf_map_lookup_elem(map_fd_ipv4, key_ipv4, &value) == -1 &&
4d3381f5
DH
321 errno == ENOENT);
322
323 inet_pton(AF_INET, "11.11.11.11", key_ipv4->data);
e5ff7c40 324 assert(bpf_map_lookup_elem(map_fd_ipv4, key_ipv4, &value) == -1 &&
4d3381f5
DH
325 errno == ENOENT);
326
327 inet_pton(AF_INET6, "2a00:ffff::", key_ipv6->data);
e5ff7c40 328 assert(bpf_map_lookup_elem(map_fd_ipv6, key_ipv6, &value) == -1 &&
4d3381f5
DH
329 errno == ENOENT);
330
331 close(map_fd_ipv4);
332 close(map_fd_ipv6);
333}
334
335int main(void)
336{
337 struct rlimit limit = { RLIM_INFINITY, RLIM_INFINITY };
338 int i, ret;
339
340 /* we want predictable, pseudo random tests */
341 srand(0xf00ba1);
342
343 /* allow unlimited locked memory */
344 ret = setrlimit(RLIMIT_MEMLOCK, &limit);
345 if (ret < 0)
346 perror("Unable to lift memlock rlimit");
347
348 test_lpm_basic();
349 test_lpm_order();
350
351 /* Test with 8, 16, 24, 32, ... 128 bit prefix length */
352 for (i = 1; i <= 16; ++i)
353 test_lpm_map(i);
354
355 test_lpm_ipaddr();
356
357 printf("test_lpm: OK\n");
358 return 0;
359}