1// SPDX-License-Identifier: GPL-2.0
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 <pthread.h>
18#include <stdio.h>
19#include <stdlib.h>
20#include <string.h>
21#include <time.h>
22#include <unistd.h>
23#include <arpa/inet.h>
24#include <sys/time.h>
25
26#include <bpf/bpf.h>
27
28#include "bpf_util.h"
29
30struct tlpm_node {
31	struct tlpm_node *next;
32	size_t n_bits;
33	uint8_t key[];
34};
35
36static struct tlpm_node *tlpm_match(struct tlpm_node *list,
37				    const uint8_t *key,
38				    size_t n_bits);
39
40static struct tlpm_node *tlpm_add(struct tlpm_node *list,
41				  const uint8_t *key,
42				  size_t n_bits)
43{
44	struct tlpm_node *node;
45	size_t n;
46
47	n = (n_bits + 7) / 8;
48
49	/* 'overwrite' an equivalent entry if one already exists */
50	node = tlpm_match(list, key, n_bits);
51	if (node && node->n_bits == n_bits) {
52		memcpy(node->key, key, n);
53		return list;
54	}
55
56	/* add new entry with @key/@n_bits to @list and return new head */
57
58	node = malloc(sizeof(*node) + n);
59	assert(node);
60
61	node->next = list;
62	node->n_bits = n_bits;
63	memcpy(node->key, key, n);
64
65	return node;
66}
67
68static void tlpm_clear(struct tlpm_node *list)
69{
70	struct tlpm_node *node;
71
72	/* free all entries in @list */
73
74	while ((node = list)) {
75		list = list->next;
76		free(node);
77	}
78}
79
80static struct tlpm_node *tlpm_match(struct tlpm_node *list,
81				    const uint8_t *key,
82				    size_t n_bits)
83{
84	struct tlpm_node *best = NULL;
85	size_t i;
86
87	/* Perform longest prefix-match on @key/@n_bits. That is, iterate all
88	 * entries and match each prefix against @key. Remember the "best"
89	 * entry we find (i.e., the longest prefix that matches) and return it
90	 * to the caller when done.
91	 */
92
93	for ( ; list; list = list->next) {
94		for (i = 0; i < n_bits && i < list->n_bits; ++i) {
95			if ((key[i / 8] & (1 << (7 - i % 8))) !=
96			    (list->key[i / 8] & (1 << (7 - i % 8))))
97				break;
98		}
99
100		if (i >= list->n_bits) {
101			if (!best || i > best->n_bits)
102				best = list;
103		}
104	}
105
106	return best;
107}
108
109static struct tlpm_node *tlpm_delete(struct tlpm_node *list,
110				     const uint8_t *key,
111				     size_t n_bits)
112{
113	struct tlpm_node *best = tlpm_match(list, key, n_bits);
114	struct tlpm_node *node;
115
116	if (!best || best->n_bits != n_bits)
117		return list;
118
119	if (best == list) {
120		node = best->next;
121		free(best);
122		return node;
123	}
124
125	for (node = list; node; node = node->next) {
126		if (node->next == best) {
127			node->next = best->next;
128			free(best);
129			return list;
130		}
131	}
132	/* should never get here */
133	assert(0);
134	return list;
135}
136
137static void test_lpm_basic(void)
138{
139	struct tlpm_node *list = NULL, *t1, *t2;
140
141	/* very basic, static tests to verify tlpm works as expected */
142
143	assert(!tlpm_match(list, (uint8_t[]){ 0xff }, 8));
144
145	t1 = list = tlpm_add(list, (uint8_t[]){ 0xff }, 8);
146	assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff }, 8));
147	assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff, 0xff }, 16));
148	assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff, 0x00 }, 16));
149	assert(!tlpm_match(list, (uint8_t[]){ 0x7f }, 8));
150	assert(!tlpm_match(list, (uint8_t[]){ 0xfe }, 8));
151	assert(!tlpm_match(list, (uint8_t[]){ 0xff }, 7));
152
153	t2 = list = tlpm_add(list, (uint8_t[]){ 0xff, 0xff }, 16);
154	assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff }, 8));
155	assert(t2 == tlpm_match(list, (uint8_t[]){ 0xff, 0xff }, 16));
156	assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff, 0xff }, 15));
157	assert(!tlpm_match(list, (uint8_t[]){ 0x7f, 0xff }, 16));
158
159	list = tlpm_delete(list, (uint8_t[]){ 0xff, 0xff }, 16);
160	assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff }, 8));
161	assert(t1 == tlpm_match(list, (uint8_t[]){ 0xff, 0xff }, 16));
162
163	list = tlpm_delete(list, (uint8_t[]){ 0xff }, 8);
164	assert(!tlpm_match(list, (uint8_t[]){ 0xff }, 8));
165
166	tlpm_clear(list);
167}
168
169static void test_lpm_order(void)
170{
171	struct tlpm_node *t1, *t2, *l1 = NULL, *l2 = NULL;
172	size_t i, j;
173
174	/* Verify the tlpm implementation works correctly regardless of the
175	 * order of entries. Insert a random set of entries into @l1, and copy
176	 * the same data in reverse order into @l2. Then verify a lookup of
177	 * random keys will yield the same result in both sets.
178	 */
179
180	for (i = 0; i < (1 << 12); ++i)
181		l1 = tlpm_add(l1, (uint8_t[]){
182					rand() % 0xff,
183					rand() % 0xff,
184				}, rand() % 16 + 1);
185
186	for (t1 = l1; t1; t1 = t1->next)
187		l2 = tlpm_add(l2, t1->key, t1->n_bits);
188
189	for (i = 0; i < (1 << 8); ++i) {
190		uint8_t key[] = { rand() % 0xff, rand() % 0xff };
191
192		t1 = tlpm_match(l1, key, 16);
193		t2 = tlpm_match(l2, key, 16);
194
195		assert(!t1 == !t2);
196		if (t1) {
197			assert(t1->n_bits == t2->n_bits);
198			for (j = 0; j < t1->n_bits; ++j)
199				assert((t1->key[j / 8] & (1 << (7 - j % 8))) ==
200				       (t2->key[j / 8] & (1 << (7 - j % 8))));
201		}
202	}
203
204	tlpm_clear(l1);
205	tlpm_clear(l2);
206}
207
208static void test_lpm_map(int keysize)
209{
210	LIBBPF_OPTS(bpf_map_create_opts, opts, .map_flags = BPF_F_NO_PREALLOC);
211	volatile size_t n_matches, n_matches_after_delete;
212	size_t i, j, n_nodes, n_lookups;
213	struct tlpm_node *t, *list = NULL;
214	struct bpf_lpm_trie_key_u8 *key;
215	uint8_t *data, *value;
216	int r, map;
217
218	/* Compare behavior of tlpm vs. bpf-lpm. Create a randomized set of
219	 * prefixes and insert it into both tlpm and bpf-lpm. Then run some
220	 * randomized lookups and verify both maps return the same result.
221	 */
222
223	n_matches = 0;
224	n_matches_after_delete = 0;
225	n_nodes = 1 << 8;
226	n_lookups = 1 << 16;
227
228	data = alloca(keysize);
229	memset(data, 0, keysize);
230
231	value = alloca(keysize + 1);
232	memset(value, 0, keysize + 1);
233
234	key = alloca(sizeof(*key) + keysize);
235	memset(key, 0, sizeof(*key) + keysize);
236
237	map = bpf_map_create(BPF_MAP_TYPE_LPM_TRIE, NULL,
238			     sizeof(*key) + keysize,
239			     keysize + 1,
240			     4096,
241			     &opts);
242	assert(map >= 0);
243
244	for (i = 0; i < n_nodes; ++i) {
245		for (j = 0; j < keysize; ++j)
246			value[j] = rand() & 0xff;
247		value[keysize] = rand() % (8 * keysize + 1);
248
249		list = tlpm_add(list, value, value[keysize]);
250
251		key->prefixlen = value[keysize];
252		memcpy(key->data, value, keysize);
253		r = bpf_map_update_elem(map, key, value, 0);
254		assert(!r);
255	}
256
257	for (i = 0; i < n_lookups; ++i) {
258		for (j = 0; j < keysize; ++j)
259			data[j] = rand() & 0xff;
260
261		t = tlpm_match(list, data, 8 * keysize);
262
263		key->prefixlen = 8 * keysize;
264		memcpy(key->data, data, keysize);
265		r = bpf_map_lookup_elem(map, key, value);
266		assert(!r || errno == ENOENT);
267		assert(!t == !!r);
268
269		if (t) {
270			++n_matches;
271			assert(t->n_bits == value[keysize]);
272			for (j = 0; j < t->n_bits; ++j)
273				assert((t->key[j / 8] & (1 << (7 - j % 8))) ==
274				       (value[j / 8] & (1 << (7 - j % 8))));
275		}
276	}
277
278	/* Remove the first half of the elements in the tlpm and the
279	 * corresponding nodes from the bpf-lpm.  Then run the same
280	 * large number of random lookups in both and make sure they match.
281	 * Note: we need to count the number of nodes actually inserted
282	 * since there may have been duplicates.
283	 */
284	for (i = 0, t = list; t; i++, t = t->next)
285		;
286	for (j = 0; j < i / 2; ++j) {
287		key->prefixlen = list->n_bits;
288		memcpy(key->data, list->key, keysize);
289		r = bpf_map_delete_elem(map, key);
290		assert(!r);
291		list = tlpm_delete(list, list->key, list->n_bits);
292		assert(list);
293	}
294	for (i = 0; i < n_lookups; ++i) {
295		for (j = 0; j < keysize; ++j)
296			data[j] = rand() & 0xff;
297
298		t = tlpm_match(list, data, 8 * keysize);
299
300		key->prefixlen = 8 * keysize;
301		memcpy(key->data, data, keysize);
302		r = bpf_map_lookup_elem(map, key, value);
303		assert(!r || errno == ENOENT);
304		assert(!t == !!r);
305
306		if (t) {
307			++n_matches_after_delete;
308			assert(t->n_bits == value[keysize]);
309			for (j = 0; j < t->n_bits; ++j)
310				assert((t->key[j / 8] & (1 << (7 - j % 8))) ==
311				       (value[j / 8] & (1 << (7 - j % 8))));
312		}
313	}
314
315	close(map);
316	tlpm_clear(list);
317
318	/* With 255 random nodes in the map, we are pretty likely to match
319	 * something on every lookup. For statistics, use this:
320	 *
321	 *     printf("          nodes: %zu\n"
322	 *            "        lookups: %zu\n"
323	 *            "        matches: %zu\n"
324	 *            "matches(delete): %zu\n",
325	 *            n_nodes, n_lookups, n_matches, n_matches_after_delete);
326	 */
327}
328
329/* Test the implementation with some 'real world' examples */
330
331static void test_lpm_ipaddr(void)
332{
333	LIBBPF_OPTS(bpf_map_create_opts, opts, .map_flags = BPF_F_NO_PREALLOC);
334	struct bpf_lpm_trie_key_u8 *key_ipv4;
335	struct bpf_lpm_trie_key_u8 *key_ipv6;
336	size_t key_size_ipv4;
337	size_t key_size_ipv6;
338	int map_fd_ipv4;
339	int map_fd_ipv6;
340	__u64 value;
341
342	key_size_ipv4 = sizeof(*key_ipv4) + sizeof(__u32);
343	key_size_ipv6 = sizeof(*key_ipv6) + sizeof(__u32) * 4;
344	key_ipv4 = alloca(key_size_ipv4);
345	key_ipv6 = alloca(key_size_ipv6);
346
347	map_fd_ipv4 = bpf_map_create(BPF_MAP_TYPE_LPM_TRIE, NULL,
348				     key_size_ipv4, sizeof(value),
349				     100, &opts);
350	assert(map_fd_ipv4 >= 0);
351
352	map_fd_ipv6 = bpf_map_create(BPF_MAP_TYPE_LPM_TRIE, NULL,
353				     key_size_ipv6, sizeof(value),
354				     100, &opts);
355	assert(map_fd_ipv6 >= 0);
356
357	/* Fill data some IPv4 and IPv6 address ranges */
358	value = 1;
359	key_ipv4->prefixlen = 16;
360	inet_pton(AF_INET, "192.168.0.0", key_ipv4->data);
361	assert(bpf_map_update_elem(map_fd_ipv4, key_ipv4, &value, 0) == 0);
362
363	value = 2;
364	key_ipv4->prefixlen = 24;
365	inet_pton(AF_INET, "192.168.0.0", key_ipv4->data);
366	assert(bpf_map_update_elem(map_fd_ipv4, key_ipv4, &value, 0) == 0);
367
368	value = 3;
369	key_ipv4->prefixlen = 24;
370	inet_pton(AF_INET, "192.168.128.0", key_ipv4->data);
371	assert(bpf_map_update_elem(map_fd_ipv4, key_ipv4, &value, 0) == 0);
372
373	value = 5;
374	key_ipv4->prefixlen = 24;
375	inet_pton(AF_INET, "192.168.1.0", key_ipv4->data);
376	assert(bpf_map_update_elem(map_fd_ipv4, key_ipv4, &value, 0) == 0);
377
378	value = 4;
379	key_ipv4->prefixlen = 23;
380	inet_pton(AF_INET, "192.168.0.0", key_ipv4->data);
381	assert(bpf_map_update_elem(map_fd_ipv4, key_ipv4, &value, 0) == 0);
382
383	value = 0xdeadbeef;
384	key_ipv6->prefixlen = 64;
385	inet_pton(AF_INET6, "2a00:1450:4001:814::200e", key_ipv6->data);
386	assert(bpf_map_update_elem(map_fd_ipv6, key_ipv6, &value, 0) == 0);
387
388	/* Set tprefixlen to maximum for lookups */
389	key_ipv4->prefixlen = 32;
390	key_ipv6->prefixlen = 128;
391
392	/* Test some lookups that should come back with a value */
393	inet_pton(AF_INET, "192.168.128.23", key_ipv4->data);
394	assert(bpf_map_lookup_elem(map_fd_ipv4, key_ipv4, &value) == 0);
395	assert(value == 3);
396
397	inet_pton(AF_INET, "192.168.0.1", key_ipv4->data);
398	assert(bpf_map_lookup_elem(map_fd_ipv4, key_ipv4, &value) == 0);
399	assert(value == 2);
400
401	inet_pton(AF_INET6, "2a00:1450:4001:814::", key_ipv6->data);
402	assert(bpf_map_lookup_elem(map_fd_ipv6, key_ipv6, &value) == 0);
403	assert(value == 0xdeadbeef);
404
405	inet_pton(AF_INET6, "2a00:1450:4001:814::1", key_ipv6->data);
406	assert(bpf_map_lookup_elem(map_fd_ipv6, key_ipv6, &value) == 0);
407	assert(value == 0xdeadbeef);
408
409	/* Test some lookups that should not match any entry */
410	inet_pton(AF_INET, "10.0.0.1", key_ipv4->data);
411	assert(bpf_map_lookup_elem(map_fd_ipv4, key_ipv4, &value) == -ENOENT);
412
413	inet_pton(AF_INET, "11.11.11.11", key_ipv4->data);
414	assert(bpf_map_lookup_elem(map_fd_ipv4, key_ipv4, &value) == -ENOENT);
415
416	inet_pton(AF_INET6, "2a00:ffff::", key_ipv6->data);
417	assert(bpf_map_lookup_elem(map_fd_ipv6, key_ipv6, &value) == -ENOENT);
418
419	close(map_fd_ipv4);
420	close(map_fd_ipv6);
421}
422
423static void test_lpm_delete(void)
424{
425	LIBBPF_OPTS(bpf_map_create_opts, opts, .map_flags = BPF_F_NO_PREALLOC);
426	struct bpf_lpm_trie_key_u8 *key;
427	size_t key_size;
428	int map_fd;
429	__u64 value;
430
431	key_size = sizeof(*key) + sizeof(__u32);
432	key = alloca(key_size);
433
434	map_fd = bpf_map_create(BPF_MAP_TYPE_LPM_TRIE, NULL,
435				key_size, sizeof(value),
436				100, &opts);
437	assert(map_fd >= 0);
438
439	/* Add nodes:
440	 * 192.168.0.0/16   (1)
441	 * 192.168.0.0/24   (2)
442	 * 192.168.128.0/24 (3)
443	 * 192.168.1.0/24   (4)
444	 *
445	 *         (1)
446	 *        /   \
447         *     (IM)    (3)
448	 *    /   \
449         *   (2)  (4)
450	 */
451	value = 1;
452	key->prefixlen = 16;
453	inet_pton(AF_INET, "192.168.0.0", key->data);
454	assert(bpf_map_update_elem(map_fd, key, &value, 0) == 0);
455
456	value = 2;
457	key->prefixlen = 24;
458	inet_pton(AF_INET, "192.168.0.0", key->data);
459	assert(bpf_map_update_elem(map_fd, key, &value, 0) == 0);
460
461	value = 3;
462	key->prefixlen = 24;
463	inet_pton(AF_INET, "192.168.128.0", key->data);
464	assert(bpf_map_update_elem(map_fd, key, &value, 0) == 0);
465
466	value = 4;
467	key->prefixlen = 24;
468	inet_pton(AF_INET, "192.168.1.0", key->data);
469	assert(bpf_map_update_elem(map_fd, key, &value, 0) == 0);
470
471	/* remove non-existent node */
472	key->prefixlen = 32;
473	inet_pton(AF_INET, "10.0.0.1", key->data);
474	assert(bpf_map_lookup_elem(map_fd, key, &value) == -ENOENT);
475
476	key->prefixlen = 30; // unused prefix so far
477	inet_pton(AF_INET, "192.255.0.0", key->data);
478	assert(bpf_map_delete_elem(map_fd, key) == -ENOENT);
479
480	key->prefixlen = 16; // same prefix as the root node
481	inet_pton(AF_INET, "192.255.0.0", key->data);
482	assert(bpf_map_delete_elem(map_fd, key) == -ENOENT);
483
484	/* assert initial lookup */
485	key->prefixlen = 32;
486	inet_pton(AF_INET, "192.168.0.1", key->data);
487	assert(bpf_map_lookup_elem(map_fd, key, &value) == 0);
488	assert(value == 2);
489
490	/* remove leaf node */
491	key->prefixlen = 24;
492	inet_pton(AF_INET, "192.168.0.0", key->data);
493	assert(bpf_map_delete_elem(map_fd, key) == 0);
494
495	key->prefixlen = 32;
496	inet_pton(AF_INET, "192.168.0.1", key->data);
497	assert(bpf_map_lookup_elem(map_fd, key, &value) == 0);
498	assert(value == 1);
499
500	/* remove leaf (and intermediary) node */
501	key->prefixlen = 24;
502	inet_pton(AF_INET, "192.168.1.0", key->data);
503	assert(bpf_map_delete_elem(map_fd, key) == 0);
504
505	key->prefixlen = 32;
506	inet_pton(AF_INET, "192.168.1.1", key->data);
507	assert(bpf_map_lookup_elem(map_fd, key, &value) == 0);
508	assert(value == 1);
509
510	/* remove root node */
511	key->prefixlen = 16;
512	inet_pton(AF_INET, "192.168.0.0", key->data);
513	assert(bpf_map_delete_elem(map_fd, key) == 0);
514
515	key->prefixlen = 32;
516	inet_pton(AF_INET, "192.168.128.1", key->data);
517	assert(bpf_map_lookup_elem(map_fd, key, &value) == 0);
518	assert(value == 3);
519
520	/* remove last node */
521	key->prefixlen = 24;
522	inet_pton(AF_INET, "192.168.128.0", key->data);
523	assert(bpf_map_delete_elem(map_fd, key) == 0);
524
525	key->prefixlen = 32;
526	inet_pton(AF_INET, "192.168.128.1", key->data);
527	assert(bpf_map_lookup_elem(map_fd, key, &value) == -ENOENT);
528
529	close(map_fd);
530}
531
532static void test_lpm_get_next_key(void)
533{
534	LIBBPF_OPTS(bpf_map_create_opts, opts, .map_flags = BPF_F_NO_PREALLOC);
535	struct bpf_lpm_trie_key_u8 *key_p, *next_key_p;
536	size_t key_size;
537	__u32 value = 0;
538	int map_fd;
539
540	key_size = sizeof(*key_p) + sizeof(__u32);
541	key_p = alloca(key_size);
542	next_key_p = alloca(key_size);
543
544	map_fd = bpf_map_create(BPF_MAP_TYPE_LPM_TRIE, NULL, key_size, sizeof(value), 100, &opts);
545	assert(map_fd >= 0);
546
547	/* empty tree. get_next_key should return ENOENT */
548	assert(bpf_map_get_next_key(map_fd, NULL, key_p) == -ENOENT);
549
550	/* get and verify the first key, get the second one should fail. */
551	key_p->prefixlen = 16;
552	inet_pton(AF_INET, "192.168.0.0", key_p->data);
553	assert(bpf_map_update_elem(map_fd, key_p, &value, 0) == 0);
554
555	memset(key_p, 0, key_size);
556	assert(bpf_map_get_next_key(map_fd, NULL, key_p) == 0);
557	assert(key_p->prefixlen == 16 && key_p->data[0] == 192 &&
558	       key_p->data[1] == 168);
559
560	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == -ENOENT);
561
562	/* no exact matching key should get the first one in post order. */
563	key_p->prefixlen = 8;
564	assert(bpf_map_get_next_key(map_fd, NULL, key_p) == 0);
565	assert(key_p->prefixlen == 16 && key_p->data[0] == 192 &&
566	       key_p->data[1] == 168);
567
568	/* add one more element (total two) */
569	key_p->prefixlen = 24;
570	inet_pton(AF_INET, "192.168.128.0", key_p->data);
571	assert(bpf_map_update_elem(map_fd, key_p, &value, 0) == 0);
572
573	memset(key_p, 0, key_size);
574	assert(bpf_map_get_next_key(map_fd, NULL, key_p) == 0);
575	assert(key_p->prefixlen == 24 && key_p->data[0] == 192 &&
576	       key_p->data[1] == 168 && key_p->data[2] == 128);
577
578	memset(next_key_p, 0, key_size);
579	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == 0);
580	assert(next_key_p->prefixlen == 16 && next_key_p->data[0] == 192 &&
581	       next_key_p->data[1] == 168);
582
583	memcpy(key_p, next_key_p, key_size);
584	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == -ENOENT);
585
586	/* Add one more element (total three) */
587	key_p->prefixlen = 24;
588	inet_pton(AF_INET, "192.168.0.0", key_p->data);
589	assert(bpf_map_update_elem(map_fd, key_p, &value, 0) == 0);
590
591	memset(key_p, 0, key_size);
592	assert(bpf_map_get_next_key(map_fd, NULL, key_p) == 0);
593	assert(key_p->prefixlen == 24 && key_p->data[0] == 192 &&
594	       key_p->data[1] == 168 && key_p->data[2] == 0);
595
596	memset(next_key_p, 0, key_size);
597	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == 0);
598	assert(next_key_p->prefixlen == 24 && next_key_p->data[0] == 192 &&
599	       next_key_p->data[1] == 168 && next_key_p->data[2] == 128);
600
601	memcpy(key_p, next_key_p, key_size);
602	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == 0);
603	assert(next_key_p->prefixlen == 16 && next_key_p->data[0] == 192 &&
604	       next_key_p->data[1] == 168);
605
606	memcpy(key_p, next_key_p, key_size);
607	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == -ENOENT);
608
609	/* Add one more element (total four) */
610	key_p->prefixlen = 24;
611	inet_pton(AF_INET, "192.168.1.0", key_p->data);
612	assert(bpf_map_update_elem(map_fd, key_p, &value, 0) == 0);
613
614	memset(key_p, 0, key_size);
615	assert(bpf_map_get_next_key(map_fd, NULL, key_p) == 0);
616	assert(key_p->prefixlen == 24 && key_p->data[0] == 192 &&
617	       key_p->data[1] == 168 && key_p->data[2] == 0);
618
619	memset(next_key_p, 0, key_size);
620	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == 0);
621	assert(next_key_p->prefixlen == 24 && next_key_p->data[0] == 192 &&
622	       next_key_p->data[1] == 168 && next_key_p->data[2] == 1);
623
624	memcpy(key_p, next_key_p, key_size);
625	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == 0);
626	assert(next_key_p->prefixlen == 24 && next_key_p->data[0] == 192 &&
627	       next_key_p->data[1] == 168 && next_key_p->data[2] == 128);
628
629	memcpy(key_p, next_key_p, key_size);
630	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == 0);
631	assert(next_key_p->prefixlen == 16 && next_key_p->data[0] == 192 &&
632	       next_key_p->data[1] == 168);
633
634	memcpy(key_p, next_key_p, key_size);
635	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == -ENOENT);
636
637	/* Add one more element (total five) */
638	key_p->prefixlen = 28;
639	inet_pton(AF_INET, "192.168.1.128", key_p->data);
640	assert(bpf_map_update_elem(map_fd, key_p, &value, 0) == 0);
641
642	memset(key_p, 0, key_size);
643	assert(bpf_map_get_next_key(map_fd, NULL, key_p) == 0);
644	assert(key_p->prefixlen == 24 && key_p->data[0] == 192 &&
645	       key_p->data[1] == 168 && key_p->data[2] == 0);
646
647	memset(next_key_p, 0, key_size);
648	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == 0);
649	assert(next_key_p->prefixlen == 28 && next_key_p->data[0] == 192 &&
650	       next_key_p->data[1] == 168 && next_key_p->data[2] == 1 &&
651	       next_key_p->data[3] == 128);
652
653	memcpy(key_p, next_key_p, key_size);
654	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == 0);
655	assert(next_key_p->prefixlen == 24 && next_key_p->data[0] == 192 &&
656	       next_key_p->data[1] == 168 && next_key_p->data[2] == 1);
657
658	memcpy(key_p, next_key_p, key_size);
659	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == 0);
660	assert(next_key_p->prefixlen == 24 && next_key_p->data[0] == 192 &&
661	       next_key_p->data[1] == 168 && next_key_p->data[2] == 128);
662
663	memcpy(key_p, next_key_p, key_size);
664	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == 0);
665	assert(next_key_p->prefixlen == 16 && next_key_p->data[0] == 192 &&
666	       next_key_p->data[1] == 168);
667
668	memcpy(key_p, next_key_p, key_size);
669	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == -ENOENT);
670
671	/* no exact matching key should return the first one in post order */
672	key_p->prefixlen = 22;
673	inet_pton(AF_INET, "192.168.1.0", key_p->data);
674	assert(bpf_map_get_next_key(map_fd, key_p, next_key_p) == 0);
675	assert(next_key_p->prefixlen == 24 && next_key_p->data[0] == 192 &&
676	       next_key_p->data[1] == 168 && next_key_p->data[2] == 0);
677
678	close(map_fd);
679}
680
681#define MAX_TEST_KEYS	4
682struct lpm_mt_test_info {
683	int cmd; /* 0: update, 1: delete, 2: lookup, 3: get_next_key */
684	int iter;
685	int map_fd;
686	struct {
687		__u32 prefixlen;
688		__u32 data;
689	} key[MAX_TEST_KEYS];
690};
691
692static void *lpm_test_command(void *arg)
693{
694	int i, j, ret, iter, key_size;
695	struct lpm_mt_test_info *info = arg;
696	struct bpf_lpm_trie_key_u8 *key_p;
697
698	key_size = sizeof(*key_p) + sizeof(__u32);
699	key_p = alloca(key_size);
700	for (iter = 0; iter < info->iter; iter++)
701		for (i = 0; i < MAX_TEST_KEYS; i++) {
702			/* first half of iterations in forward order,
703			 * and second half in backward order.
704			 */
705			j = (iter < (info->iter / 2)) ? i : MAX_TEST_KEYS - i - 1;
706			key_p->prefixlen = info->key[j].prefixlen;
707			memcpy(key_p->data, &info->key[j].data, sizeof(__u32));
708			if (info->cmd == 0) {
709				__u32 value = j;
710				/* update must succeed */
711				assert(bpf_map_update_elem(info->map_fd, key_p, &value, 0) == 0);
712			} else if (info->cmd == 1) {
713				ret = bpf_map_delete_elem(info->map_fd, key_p);
714				assert(ret == 0 || errno == ENOENT);
715			} else if (info->cmd == 2) {
716				__u32 value;
717				ret = bpf_map_lookup_elem(info->map_fd, key_p, &value);
718				assert(ret == 0 || errno == ENOENT);
719			} else {
720				struct bpf_lpm_trie_key_u8 *next_key_p = alloca(key_size);
721				ret = bpf_map_get_next_key(info->map_fd, key_p, next_key_p);
722				assert(ret == 0 || errno == ENOENT || errno == ENOMEM);
723			}
724		}
725
726	// Pass successful exit info back to the main thread
727	pthread_exit((void *)info);
728}
729
730static void setup_lpm_mt_test_info(struct lpm_mt_test_info *info, int map_fd)
731{
732	info->iter = 2000;
733	info->map_fd = map_fd;
734	info->key[0].prefixlen = 16;
735	inet_pton(AF_INET, "192.168.0.0", &info->key[0].data);
736	info->key[1].prefixlen = 24;
737	inet_pton(AF_INET, "192.168.0.0", &info->key[1].data);
738	info->key[2].prefixlen = 24;
739	inet_pton(AF_INET, "192.168.128.0", &info->key[2].data);
740	info->key[3].prefixlen = 24;
741	inet_pton(AF_INET, "192.168.1.0", &info->key[3].data);
742}
743
744static void test_lpm_multi_thread(void)
745{
746	LIBBPF_OPTS(bpf_map_create_opts, opts, .map_flags = BPF_F_NO_PREALLOC);
747	struct lpm_mt_test_info info[4];
748	size_t key_size, value_size;
749	pthread_t thread_id[4];
750	int i, map_fd;
751	void *ret;
752
753	/* create a trie */
754	value_size = sizeof(__u32);
755	key_size = sizeof(struct bpf_lpm_trie_key_hdr) + value_size;
756	map_fd = bpf_map_create(BPF_MAP_TYPE_LPM_TRIE, NULL, key_size, value_size, 100, &opts);
757
758	/* create 4 threads to test update, delete, lookup and get_next_key */
759	setup_lpm_mt_test_info(&info[0], map_fd);
760	for (i = 0; i < 4; i++) {
761		if (i != 0)
762			memcpy(&info[i], &info[0], sizeof(info[i]));
763		info[i].cmd = i;
764		assert(pthread_create(&thread_id[i], NULL, &lpm_test_command, &info[i]) == 0);
765	}
766
767	for (i = 0; i < 4; i++)
768		assert(pthread_join(thread_id[i], &ret) == 0 && ret == (void *)&info[i]);
769
770	close(map_fd);
771}
772
773int main(void)
774{
775	int i;
776
777	/* we want predictable, pseudo random tests */
778	srand(0xf00ba1);
779
780	/* Use libbpf 1.0 API mode */
781	libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
782
783	test_lpm_basic();
784	test_lpm_order();
785
786	/* Test with 8, 16, 24, 32, ... 128 bit prefix length */
787	for (i = 1; i <= 16; ++i)
788		test_lpm_map(i);
789
790	test_lpm_ipaddr();
791	test_lpm_delete();
792	test_lpm_get_next_key();
793	test_lpm_multi_thread();
794
795	printf("test_lpm: OK\n");
796	return 0;
797}
798