xref: /openbmc/linux/tools/testing/selftests/bpf/prog_tests/test_tunnel.c (revision 1bffcea42926b26e092045ac398850e80d950bb2)
1 // SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
2 
3 /*
4  * End-to-end eBPF tunnel test suite
5  *   The file tests BPF network tunnel implementation.
6  *
7  * Topology:
8  * ---------
9  *     root namespace   |     at_ns0 namespace
10  *                       |
11  *       -----------     |     -----------
12  *       | tnl dev |     |     | tnl dev |  (overlay network)
13  *       -----------     |     -----------
14  *       metadata-mode   |     metadata-mode
15  *        with bpf       |       with bpf
16  *                       |
17  *       ----------      |     ----------
18  *       |  veth1  | --------- |  veth0  |  (underlay network)
19  *       ----------    peer    ----------
20  *
21  *
22  *  Device Configuration
23  *  --------------------
24  *  root namespace with metadata-mode tunnel + BPF
25  *  Device names and addresses:
26  *	veth1 IP 1: 172.16.1.200, IPv6: 00::22 (underlay)
27  *		IP 2: 172.16.1.20, IPv6: 00::bb (underlay)
28  *	tunnel dev <type>11, ex: gre11, IPv4: 10.1.1.200, IPv6: 1::22 (overlay)
29  *
30  *  Namespace at_ns0 with native tunnel
31  *  Device names and addresses:
32  *	veth0 IPv4: 172.16.1.100, IPv6: 00::11 (underlay)
33  *	tunnel dev <type>00, ex: gre00, IPv4: 10.1.1.100, IPv6: 1::11 (overlay)
34  *
35  *
36  * End-to-end ping packet flow
37  *  ---------------------------
38  *  Most of the tests start by namespace creation, device configuration,
39  *  then ping the underlay and overlay network.  When doing 'ping 10.1.1.100'
40  *  from root namespace, the following operations happen:
41  *  1) Route lookup shows 10.1.1.100/24 belongs to tnl dev, fwd to tnl dev.
42  *  2) Tnl device's egress BPF program is triggered and set the tunnel metadata,
43  *     with local_ip=172.16.1.200, remote_ip=172.16.1.100. BPF program choose
44  *     the primary or secondary ip of veth1 as the local ip of tunnel. The
45  *     choice is made based on the value of bpf map local_ip_map.
46  *  3) Outer tunnel header is prepended and route the packet to veth1's egress.
47  *  4) veth0's ingress queue receive the tunneled packet at namespace at_ns0.
48  *  5) Tunnel protocol handler, ex: vxlan_rcv, decap the packet.
49  *  6) Forward the packet to the overlay tnl dev.
50  */
51 
52 #include <arpa/inet.h>
53 #include <linux/if_tun.h>
54 #include <linux/limits.h>
55 #include <linux/sysctl.h>
56 #include <linux/time_types.h>
57 #include <linux/net_tstamp.h>
58 #include <net/if.h>
59 #include <stdbool.h>
60 #include <stdio.h>
61 #include <sys/stat.h>
62 #include <unistd.h>
63 
64 #include "test_progs.h"
65 #include "network_helpers.h"
66 #include "test_tunnel_kern.skel.h"
67 
68 #define IP4_ADDR_VETH0 "172.16.1.100"
69 #define IP4_ADDR1_VETH1 "172.16.1.200"
70 #define IP4_ADDR2_VETH1 "172.16.1.20"
71 #define IP4_ADDR_TUNL_DEV0 "10.1.1.100"
72 #define IP4_ADDR_TUNL_DEV1 "10.1.1.200"
73 
74 #define IP6_ADDR_VETH0 "::11"
75 #define IP6_ADDR1_VETH1 "::22"
76 #define IP6_ADDR2_VETH1 "::bb"
77 
78 #define IP4_ADDR1_HEX_VETH1 0xac1001c8
79 #define IP4_ADDR2_HEX_VETH1 0xac100114
80 #define IP6_ADDR1_HEX_VETH1 0x22
81 #define IP6_ADDR2_HEX_VETH1 0xbb
82 
83 #define MAC_TUNL_DEV0 "52:54:00:d9:01:00"
84 #define MAC_TUNL_DEV1 "52:54:00:d9:02:00"
85 #define MAC_VETH1 "52:54:00:d9:03:00"
86 
87 #define VXLAN_TUNL_DEV0 "vxlan00"
88 #define VXLAN_TUNL_DEV1 "vxlan11"
89 #define IP6VXLAN_TUNL_DEV0 "ip6vxlan00"
90 #define IP6VXLAN_TUNL_DEV1 "ip6vxlan11"
91 
92 #define PING_ARGS "-i 0.01 -c 3 -w 10 -q"
93 
94 static int config_device(void)
95 {
96 	SYS(fail, "ip netns add at_ns0");
97 	SYS(fail, "ip link add veth0 address " MAC_VETH1 " type veth peer name veth1");
98 	SYS(fail, "ip link set veth0 netns at_ns0");
99 	SYS(fail, "ip addr add " IP4_ADDR1_VETH1 "/24 dev veth1");
100 	SYS(fail, "ip link set dev veth1 up mtu 1500");
101 	SYS(fail, "ip netns exec at_ns0 ip addr add " IP4_ADDR_VETH0 "/24 dev veth0");
102 	SYS(fail, "ip netns exec at_ns0 ip link set dev veth0 up mtu 1500");
103 
104 	return 0;
105 fail:
106 	return -1;
107 }
108 
109 static void cleanup(void)
110 {
111 	SYS_NOFAIL("test -f /var/run/netns/at_ns0 && ip netns delete at_ns0");
112 	SYS_NOFAIL("ip link del veth1 2> /dev/null");
113 	SYS_NOFAIL("ip link del %s 2> /dev/null", VXLAN_TUNL_DEV1);
114 	SYS_NOFAIL("ip link del %s 2> /dev/null", IP6VXLAN_TUNL_DEV1);
115 }
116 
117 static int add_vxlan_tunnel(void)
118 {
119 	/* at_ns0 namespace */
120 	SYS(fail, "ip netns exec at_ns0 ip link add dev %s type vxlan external gbp dstport 4789",
121 	    VXLAN_TUNL_DEV0);
122 	SYS(fail, "ip netns exec at_ns0 ip link set dev %s address %s up",
123 	    VXLAN_TUNL_DEV0, MAC_TUNL_DEV0);
124 	SYS(fail, "ip netns exec at_ns0 ip addr add dev %s %s/24",
125 	    VXLAN_TUNL_DEV0, IP4_ADDR_TUNL_DEV0);
126 	SYS(fail, "ip netns exec at_ns0 ip neigh add %s lladdr %s dev %s",
127 	    IP4_ADDR_TUNL_DEV1, MAC_TUNL_DEV1, VXLAN_TUNL_DEV0);
128 	SYS(fail, "ip netns exec at_ns0 ip neigh add %s lladdr %s dev veth0",
129 	    IP4_ADDR2_VETH1, MAC_VETH1);
130 
131 	/* root namespace */
132 	SYS(fail, "ip link add dev %s type vxlan external gbp dstport 4789",
133 	    VXLAN_TUNL_DEV1);
134 	SYS(fail, "ip link set dev %s address %s up", VXLAN_TUNL_DEV1, MAC_TUNL_DEV1);
135 	SYS(fail, "ip addr add dev %s %s/24", VXLAN_TUNL_DEV1, IP4_ADDR_TUNL_DEV1);
136 	SYS(fail, "ip neigh add %s lladdr %s dev %s",
137 	    IP4_ADDR_TUNL_DEV0, MAC_TUNL_DEV0, VXLAN_TUNL_DEV1);
138 
139 	return 0;
140 fail:
141 	return -1;
142 }
143 
144 static void delete_vxlan_tunnel(void)
145 {
146 	SYS_NOFAIL("ip netns exec at_ns0 ip link delete dev %s",
147 		   VXLAN_TUNL_DEV0);
148 	SYS_NOFAIL("ip link delete dev %s", VXLAN_TUNL_DEV1);
149 }
150 
151 static int add_ip6vxlan_tunnel(void)
152 {
153 	SYS(fail, "ip netns exec at_ns0 ip -6 addr add %s/96 dev veth0",
154 	    IP6_ADDR_VETH0);
155 	SYS(fail, "ip netns exec at_ns0 ip link set dev veth0 up");
156 	SYS(fail, "ip -6 addr add %s/96 dev veth1", IP6_ADDR1_VETH1);
157 	SYS(fail, "ip -6 addr add %s/96 dev veth1", IP6_ADDR2_VETH1);
158 	SYS(fail, "ip link set dev veth1 up");
159 
160 	/* at_ns0 namespace */
161 	SYS(fail, "ip netns exec at_ns0 ip link add dev %s type vxlan external dstport 4789",
162 	    IP6VXLAN_TUNL_DEV0);
163 	SYS(fail, "ip netns exec at_ns0 ip addr add dev %s %s/24",
164 	    IP6VXLAN_TUNL_DEV0, IP4_ADDR_TUNL_DEV0);
165 	SYS(fail, "ip netns exec at_ns0 ip link set dev %s address %s up",
166 	    IP6VXLAN_TUNL_DEV0, MAC_TUNL_DEV0);
167 
168 	/* root namespace */
169 	SYS(fail, "ip link add dev %s type vxlan external dstport 4789",
170 	    IP6VXLAN_TUNL_DEV1);
171 	SYS(fail, "ip addr add dev %s %s/24", IP6VXLAN_TUNL_DEV1, IP4_ADDR_TUNL_DEV1);
172 	SYS(fail, "ip link set dev %s address %s up",
173 	    IP6VXLAN_TUNL_DEV1, MAC_TUNL_DEV1);
174 
175 	return 0;
176 fail:
177 	return -1;
178 }
179 
180 static void delete_ip6vxlan_tunnel(void)
181 {
182 	SYS_NOFAIL("ip netns exec at_ns0 ip -6 addr delete %s/96 dev veth0",
183 		   IP6_ADDR_VETH0);
184 	SYS_NOFAIL("ip -6 addr delete %s/96 dev veth1", IP6_ADDR1_VETH1);
185 	SYS_NOFAIL("ip -6 addr delete %s/96 dev veth1", IP6_ADDR2_VETH1);
186 	SYS_NOFAIL("ip netns exec at_ns0 ip link delete dev %s",
187 		   IP6VXLAN_TUNL_DEV0);
188 	SYS_NOFAIL("ip link delete dev %s", IP6VXLAN_TUNL_DEV1);
189 }
190 
191 static int test_ping(int family, const char *addr)
192 {
193 	SYS(fail, "%s %s %s > /dev/null", ping_command(family), PING_ARGS, addr);
194 	return 0;
195 fail:
196 	return -1;
197 }
198 
199 static int attach_tc_prog(struct bpf_tc_hook *hook, int igr_fd, int egr_fd)
200 {
201 	DECLARE_LIBBPF_OPTS(bpf_tc_opts, opts1, .handle = 1,
202 			    .priority = 1, .prog_fd = igr_fd);
203 	DECLARE_LIBBPF_OPTS(bpf_tc_opts, opts2, .handle = 1,
204 			    .priority = 1, .prog_fd = egr_fd);
205 	int ret;
206 
207 	ret = bpf_tc_hook_create(hook);
208 	if (!ASSERT_OK(ret, "create tc hook"))
209 		return ret;
210 
211 	if (igr_fd >= 0) {
212 		hook->attach_point = BPF_TC_INGRESS;
213 		ret = bpf_tc_attach(hook, &opts1);
214 		if (!ASSERT_OK(ret, "bpf_tc_attach")) {
215 			bpf_tc_hook_destroy(hook);
216 			return ret;
217 		}
218 	}
219 
220 	if (egr_fd >= 0) {
221 		hook->attach_point = BPF_TC_EGRESS;
222 		ret = bpf_tc_attach(hook, &opts2);
223 		if (!ASSERT_OK(ret, "bpf_tc_attach")) {
224 			bpf_tc_hook_destroy(hook);
225 			return ret;
226 		}
227 	}
228 
229 	return 0;
230 }
231 
232 static void test_vxlan_tunnel(void)
233 {
234 	struct test_tunnel_kern *skel = NULL;
235 	struct nstoken *nstoken;
236 	int local_ip_map_fd = -1;
237 	int set_src_prog_fd, get_src_prog_fd;
238 	int set_dst_prog_fd;
239 	int key = 0, ifindex = -1;
240 	uint local_ip;
241 	int err;
242 	DECLARE_LIBBPF_OPTS(bpf_tc_hook, tc_hook,
243 			    .attach_point = BPF_TC_INGRESS);
244 
245 	/* add vxlan tunnel */
246 	err = add_vxlan_tunnel();
247 	if (!ASSERT_OK(err, "add vxlan tunnel"))
248 		goto done;
249 
250 	/* load and attach bpf prog to tunnel dev tc hook point */
251 	skel = test_tunnel_kern__open_and_load();
252 	if (!ASSERT_OK_PTR(skel, "test_tunnel_kern__open_and_load"))
253 		goto done;
254 	ifindex = if_nametoindex(VXLAN_TUNL_DEV1);
255 	if (!ASSERT_NEQ(ifindex, 0, "vxlan11 ifindex"))
256 		goto done;
257 	tc_hook.ifindex = ifindex;
258 	get_src_prog_fd = bpf_program__fd(skel->progs.vxlan_get_tunnel_src);
259 	set_src_prog_fd = bpf_program__fd(skel->progs.vxlan_set_tunnel_src);
260 	if (!ASSERT_GE(get_src_prog_fd, 0, "bpf_program__fd"))
261 		goto done;
262 	if (!ASSERT_GE(set_src_prog_fd, 0, "bpf_program__fd"))
263 		goto done;
264 	if (attach_tc_prog(&tc_hook, get_src_prog_fd, set_src_prog_fd))
265 		goto done;
266 
267 	/* load and attach bpf prog to veth dev tc hook point */
268 	ifindex = if_nametoindex("veth1");
269 	if (!ASSERT_NEQ(ifindex, 0, "veth1 ifindex"))
270 		goto done;
271 	tc_hook.ifindex = ifindex;
272 	set_dst_prog_fd = bpf_program__fd(skel->progs.veth_set_outer_dst);
273 	if (!ASSERT_GE(set_dst_prog_fd, 0, "bpf_program__fd"))
274 		goto done;
275 	if (attach_tc_prog(&tc_hook, set_dst_prog_fd, -1))
276 		goto done;
277 
278 	/* load and attach prog set_md to tunnel dev tc hook point at_ns0 */
279 	nstoken = open_netns("at_ns0");
280 	if (!ASSERT_OK_PTR(nstoken, "setns src"))
281 		goto done;
282 	ifindex = if_nametoindex(VXLAN_TUNL_DEV0);
283 	if (!ASSERT_NEQ(ifindex, 0, "vxlan00 ifindex"))
284 		goto done;
285 	tc_hook.ifindex = ifindex;
286 	set_dst_prog_fd = bpf_program__fd(skel->progs.vxlan_set_tunnel_dst);
287 	if (!ASSERT_GE(set_dst_prog_fd, 0, "bpf_program__fd"))
288 		goto done;
289 	if (attach_tc_prog(&tc_hook, -1, set_dst_prog_fd))
290 		goto done;
291 	close_netns(nstoken);
292 
293 	/* use veth1 ip 2 as tunnel source ip */
294 	local_ip_map_fd = bpf_map__fd(skel->maps.local_ip_map);
295 	if (!ASSERT_GE(local_ip_map_fd, 0, "bpf_map__fd"))
296 		goto done;
297 	local_ip = IP4_ADDR2_HEX_VETH1;
298 	err = bpf_map_update_elem(local_ip_map_fd, &key, &local_ip, BPF_ANY);
299 	if (!ASSERT_OK(err, "update bpf local_ip_map"))
300 		goto done;
301 
302 	/* ping test */
303 	err = test_ping(AF_INET, IP4_ADDR_TUNL_DEV0);
304 	if (!ASSERT_OK(err, "test_ping"))
305 		goto done;
306 
307 done:
308 	/* delete vxlan tunnel */
309 	delete_vxlan_tunnel();
310 	if (local_ip_map_fd >= 0)
311 		close(local_ip_map_fd);
312 	if (skel)
313 		test_tunnel_kern__destroy(skel);
314 }
315 
316 static void test_ip6vxlan_tunnel(void)
317 {
318 	struct test_tunnel_kern *skel = NULL;
319 	struct nstoken *nstoken;
320 	int local_ip_map_fd = -1;
321 	int set_src_prog_fd, get_src_prog_fd;
322 	int set_dst_prog_fd;
323 	int key = 0, ifindex = -1;
324 	uint local_ip;
325 	int err;
326 	DECLARE_LIBBPF_OPTS(bpf_tc_hook, tc_hook,
327 			    .attach_point = BPF_TC_INGRESS);
328 
329 	/* add vxlan tunnel */
330 	err = add_ip6vxlan_tunnel();
331 	if (!ASSERT_OK(err, "add_ip6vxlan_tunnel"))
332 		goto done;
333 
334 	/* load and attach bpf prog to tunnel dev tc hook point */
335 	skel = test_tunnel_kern__open_and_load();
336 	if (!ASSERT_OK_PTR(skel, "test_tunnel_kern__open_and_load"))
337 		goto done;
338 	ifindex = if_nametoindex(IP6VXLAN_TUNL_DEV1);
339 	if (!ASSERT_NEQ(ifindex, 0, "ip6vxlan11 ifindex"))
340 		goto done;
341 	tc_hook.ifindex = ifindex;
342 	get_src_prog_fd = bpf_program__fd(skel->progs.ip6vxlan_get_tunnel_src);
343 	set_src_prog_fd = bpf_program__fd(skel->progs.ip6vxlan_set_tunnel_src);
344 	if (!ASSERT_GE(set_src_prog_fd, 0, "bpf_program__fd"))
345 		goto done;
346 	if (!ASSERT_GE(get_src_prog_fd, 0, "bpf_program__fd"))
347 		goto done;
348 	if (attach_tc_prog(&tc_hook, get_src_prog_fd, set_src_prog_fd))
349 		goto done;
350 
351 	/* load and attach prog set_md to tunnel dev tc hook point at_ns0 */
352 	nstoken = open_netns("at_ns0");
353 	if (!ASSERT_OK_PTR(nstoken, "setns src"))
354 		goto done;
355 	ifindex = if_nametoindex(IP6VXLAN_TUNL_DEV0);
356 	if (!ASSERT_NEQ(ifindex, 0, "ip6vxlan00 ifindex"))
357 		goto done;
358 	tc_hook.ifindex = ifindex;
359 	set_dst_prog_fd = bpf_program__fd(skel->progs.ip6vxlan_set_tunnel_dst);
360 	if (!ASSERT_GE(set_dst_prog_fd, 0, "bpf_program__fd"))
361 		goto done;
362 	if (attach_tc_prog(&tc_hook, -1, set_dst_prog_fd))
363 		goto done;
364 	close_netns(nstoken);
365 
366 	/* use veth1 ip 2 as tunnel source ip */
367 	local_ip_map_fd = bpf_map__fd(skel->maps.local_ip_map);
368 	if (!ASSERT_GE(local_ip_map_fd, 0, "get local_ip_map fd"))
369 		goto done;
370 	local_ip = IP6_ADDR2_HEX_VETH1;
371 	err = bpf_map_update_elem(local_ip_map_fd, &key, &local_ip, BPF_ANY);
372 	if (!ASSERT_OK(err, "update bpf local_ip_map"))
373 		goto done;
374 
375 	/* ping test */
376 	err = test_ping(AF_INET, IP4_ADDR_TUNL_DEV0);
377 	if (!ASSERT_OK(err, "test_ping"))
378 		goto done;
379 
380 done:
381 	/* delete ipv6 vxlan tunnel */
382 	delete_ip6vxlan_tunnel();
383 	if (local_ip_map_fd >= 0)
384 		close(local_ip_map_fd);
385 	if (skel)
386 		test_tunnel_kern__destroy(skel);
387 }
388 
389 #define RUN_TEST(name)							\
390 	({								\
391 		if (test__start_subtest(#name)) {			\
392 			test_ ## name();				\
393 		}							\
394 	})
395 
396 static void *test_tunnel_run_tests(void *arg)
397 {
398 	cleanup();
399 	config_device();
400 
401 	RUN_TEST(vxlan_tunnel);
402 	RUN_TEST(ip6vxlan_tunnel);
403 
404 	cleanup();
405 
406 	return NULL;
407 }
408 
409 void test_tunnel(void)
410 {
411 	pthread_t test_thread;
412 	int err;
413 
414 	/* Run the tests in their own thread to isolate the namespace changes
415 	 * so they do not affect the environment of other tests.
416 	 * (specifically needed because of unshare(CLONE_NEWNS) in open_netns())
417 	 */
418 	err = pthread_create(&test_thread, NULL, &test_tunnel_run_tests, NULL);
419 	if (ASSERT_OK(err, "pthread_create"))
420 		ASSERT_OK(pthread_join(test_thread, NULL), "pthread_join");
421 }
422