1 // SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
2 
3 /*
4  * Topology:
5  * ---------
6  *   NS0 namespace         |   NS1 namespace        | NS2 namespace
7  *                         |                        |
8  *   +---------------+     |   +---------------+    |
9  *   |    ipsec0     |---------|    ipsec0     |    |
10  *   | 192.168.1.100 |     |   | 192.168.1.200 |    |
11  *   | if_id: bpf    |     |   +---------------+    |
12  *   +---------------+     |                        |
13  *           |             |                        |   +---------------+
14  *           |             |                        |   |    ipsec0     |
15  *           \------------------------------------------| 192.168.1.200 |
16  *                         |                        |   +---------------+
17  *                         |                        |
18  *                         |                        | (overlay network)
19  *      ------------------------------------------------------
20  *                         |                        | (underlay network)
21  *   +--------------+      |   +--------------+     |
22  *   |    veth01    |----------|    veth10    |     |
23  *   | 172.16.1.100 |      |   | 172.16.1.200 |     |
24  *   ---------------+      |   +--------------+     |
25  *                         |                        |
26  *   +--------------+      |                        |   +--------------+
27  *   |    veth02    |-----------------------------------|    veth20    |
28  *   | 172.16.2.100 |      |                        |   | 172.16.2.200 |
29  *   +--------------+      |                        |   +--------------+
30  *
31  *
32  * Test Packet flow
33  * -----------
34  *  The tests perform 'ping 192.168.1.200' from the NS0 namespace:
35  *  1) request is routed to NS0 ipsec0
36  *  2) NS0 ipsec0 tc egress BPF program is triggered and sets the if_id based
37  *     on the requested value. This makes the ipsec0 device in external mode
38  *     select the destination tunnel
39  *  3) ping reaches the other namespace (NS1 or NS2 based on which if_id was
40  *     used) and response is sent
41  *  4) response is received on NS0 ipsec0, tc ingress program is triggered and
42  *     records the response if_id
43  *  5) requested if_id is compared with received if_id
44  */
45 
46 #include <net/if.h>
47 #include <linux/rtnetlink.h>
48 #include <linux/if_link.h>
49 
50 #include "test_progs.h"
51 #include "network_helpers.h"
52 #include "xfrm_info.skel.h"
53 
54 #define NS0 "xfrm_test_ns0"
55 #define NS1 "xfrm_test_ns1"
56 #define NS2 "xfrm_test_ns2"
57 
58 #define IF_ID_0_TO_1 1
59 #define IF_ID_0_TO_2 2
60 #define IF_ID_1 3
61 #define IF_ID_2 4
62 
63 #define IP4_ADDR_VETH01 "172.16.1.100"
64 #define IP4_ADDR_VETH10 "172.16.1.200"
65 #define IP4_ADDR_VETH02 "172.16.2.100"
66 #define IP4_ADDR_VETH20 "172.16.2.200"
67 
68 #define ESP_DUMMY_PARAMS \
69     "proto esp aead 'rfc4106(gcm(aes))' " \
70     "0xe4d8f4b4da1df18a3510b3781496daa82488b713 128 mode tunnel "
71 
attach_tc_prog(struct bpf_tc_hook * hook,int igr_fd,int egr_fd)72 static int attach_tc_prog(struct bpf_tc_hook *hook, int igr_fd, int egr_fd)
73 {
74 	LIBBPF_OPTS(bpf_tc_opts, opts1, .handle = 1, .priority = 1,
75 		    .prog_fd = igr_fd);
76 	LIBBPF_OPTS(bpf_tc_opts, opts2, .handle = 1, .priority = 1,
77 		    .prog_fd = egr_fd);
78 	int ret;
79 
80 	ret = bpf_tc_hook_create(hook);
81 	if (!ASSERT_OK(ret, "create tc hook"))
82 		return ret;
83 
84 	if (igr_fd >= 0) {
85 		hook->attach_point = BPF_TC_INGRESS;
86 		ret = bpf_tc_attach(hook, &opts1);
87 		if (!ASSERT_OK(ret, "bpf_tc_attach")) {
88 			bpf_tc_hook_destroy(hook);
89 			return ret;
90 		}
91 	}
92 
93 	if (egr_fd >= 0) {
94 		hook->attach_point = BPF_TC_EGRESS;
95 		ret = bpf_tc_attach(hook, &opts2);
96 		if (!ASSERT_OK(ret, "bpf_tc_attach")) {
97 			bpf_tc_hook_destroy(hook);
98 			return ret;
99 		}
100 	}
101 
102 	return 0;
103 }
104 
cleanup(void)105 static void cleanup(void)
106 {
107 	SYS_NOFAIL("test -f /var/run/netns/" NS0 " && ip netns delete " NS0);
108 	SYS_NOFAIL("test -f /var/run/netns/" NS1 " && ip netns delete " NS1);
109 	SYS_NOFAIL("test -f /var/run/netns/" NS2 " && ip netns delete " NS2);
110 }
111 
config_underlay(void)112 static int config_underlay(void)
113 {
114 	SYS(fail, "ip netns add " NS0);
115 	SYS(fail, "ip netns add " NS1);
116 	SYS(fail, "ip netns add " NS2);
117 
118 	/* NS0 <-> NS1 [veth01 <-> veth10] */
119 	SYS(fail, "ip link add veth01 netns " NS0 " type veth peer name veth10 netns " NS1);
120 	SYS(fail, "ip -net " NS0 " addr add " IP4_ADDR_VETH01 "/24 dev veth01");
121 	SYS(fail, "ip -net " NS0 " link set dev veth01 up");
122 	SYS(fail, "ip -net " NS1 " addr add " IP4_ADDR_VETH10 "/24 dev veth10");
123 	SYS(fail, "ip -net " NS1 " link set dev veth10 up");
124 
125 	/* NS0 <-> NS2 [veth02 <-> veth20] */
126 	SYS(fail, "ip link add veth02 netns " NS0 " type veth peer name veth20 netns " NS2);
127 	SYS(fail, "ip -net " NS0 " addr add " IP4_ADDR_VETH02 "/24 dev veth02");
128 	SYS(fail, "ip -net " NS0 " link set dev veth02 up");
129 	SYS(fail, "ip -net " NS2 " addr add " IP4_ADDR_VETH20 "/24 dev veth20");
130 	SYS(fail, "ip -net " NS2 " link set dev veth20 up");
131 
132 	return 0;
133 fail:
134 	return -1;
135 }
136 
setup_xfrm_tunnel_ns(const char * ns,const char * ipv4_local,const char * ipv4_remote,int if_id)137 static int setup_xfrm_tunnel_ns(const char *ns, const char *ipv4_local,
138 				const char *ipv4_remote, int if_id)
139 {
140 	/* State: local -> remote */
141 	SYS(fail, "ip -net %s xfrm state add src %s dst %s spi 1 "
142 	    ESP_DUMMY_PARAMS "if_id %d", ns, ipv4_local, ipv4_remote, if_id);
143 
144 	/* State: local <- remote */
145 	SYS(fail, "ip -net %s xfrm state add src %s dst %s spi 1 "
146 	    ESP_DUMMY_PARAMS "if_id %d", ns, ipv4_remote, ipv4_local, if_id);
147 
148 	/* Policy: local -> remote */
149 	SYS(fail, "ip -net %s xfrm policy add dir out src 0.0.0.0/0 dst 0.0.0.0/0 "
150 	    "if_id %d tmpl src %s dst %s proto esp mode tunnel if_id %d", ns,
151 	    if_id, ipv4_local, ipv4_remote, if_id);
152 
153 	/* Policy: local <- remote */
154 	SYS(fail, "ip -net %s xfrm policy add dir in src 0.0.0.0/0 dst 0.0.0.0/0 "
155 	    "if_id %d tmpl src %s dst %s proto esp mode tunnel if_id %d", ns,
156 	    if_id, ipv4_remote, ipv4_local, if_id);
157 
158 	return 0;
159 fail:
160 	return -1;
161 }
162 
setup_xfrm_tunnel(const char * ns_a,const char * ns_b,const char * ipv4_a,const char * ipv4_b,int if_id_a,int if_id_b)163 static int setup_xfrm_tunnel(const char *ns_a, const char *ns_b,
164 			     const char *ipv4_a, const char *ipv4_b,
165 			     int if_id_a, int if_id_b)
166 {
167 	return setup_xfrm_tunnel_ns(ns_a, ipv4_a, ipv4_b, if_id_a) ||
168 		setup_xfrm_tunnel_ns(ns_b, ipv4_b, ipv4_a, if_id_b);
169 }
170 
rtattr_add(struct nlmsghdr * nh,unsigned short type,unsigned short len)171 static struct rtattr *rtattr_add(struct nlmsghdr *nh, unsigned short type,
172 				 unsigned short len)
173 {
174 	struct rtattr *rta =
175 		(struct rtattr *)((uint8_t *)nh + RTA_ALIGN(nh->nlmsg_len));
176 	rta->rta_type = type;
177 	rta->rta_len = RTA_LENGTH(len);
178 	nh->nlmsg_len = RTA_ALIGN(nh->nlmsg_len) + RTA_ALIGN(rta->rta_len);
179 	return rta;
180 }
181 
rtattr_add_str(struct nlmsghdr * nh,unsigned short type,const char * s)182 static struct rtattr *rtattr_add_str(struct nlmsghdr *nh, unsigned short type,
183 				     const char *s)
184 {
185 	struct rtattr *rta = rtattr_add(nh, type, strlen(s));
186 
187 	memcpy(RTA_DATA(rta), s, strlen(s));
188 	return rta;
189 }
190 
rtattr_begin(struct nlmsghdr * nh,unsigned short type)191 static struct rtattr *rtattr_begin(struct nlmsghdr *nh, unsigned short type)
192 {
193 	return rtattr_add(nh, type, 0);
194 }
195 
rtattr_end(struct nlmsghdr * nh,struct rtattr * attr)196 static void rtattr_end(struct nlmsghdr *nh, struct rtattr *attr)
197 {
198 	uint8_t *end = (uint8_t *)nh + nh->nlmsg_len;
199 
200 	attr->rta_len = end - (uint8_t *)attr;
201 }
202 
setup_xfrmi_external_dev(const char * ns)203 static int setup_xfrmi_external_dev(const char *ns)
204 {
205 	struct {
206 		struct nlmsghdr nh;
207 		struct ifinfomsg info;
208 		unsigned char data[128];
209 	} req;
210 	struct rtattr *link_info, *info_data;
211 	struct nstoken *nstoken;
212 	int ret = -1, sock = -1;
213 	struct nlmsghdr *nh;
214 
215 	memset(&req, 0, sizeof(req));
216 	nh = &req.nh;
217 	nh->nlmsg_len = NLMSG_LENGTH(sizeof(req.info));
218 	nh->nlmsg_type = RTM_NEWLINK;
219 	nh->nlmsg_flags |= NLM_F_CREATE | NLM_F_REQUEST;
220 
221 	rtattr_add_str(nh, IFLA_IFNAME, "ipsec0");
222 	link_info = rtattr_begin(nh, IFLA_LINKINFO);
223 	rtattr_add_str(nh, IFLA_INFO_KIND, "xfrm");
224 	info_data = rtattr_begin(nh, IFLA_INFO_DATA);
225 	rtattr_add(nh, IFLA_XFRM_COLLECT_METADATA, 0);
226 	rtattr_end(nh, info_data);
227 	rtattr_end(nh, link_info);
228 
229 	nstoken = open_netns(ns);
230 	if (!ASSERT_OK_PTR(nstoken, "setns"))
231 		goto done;
232 
233 	sock = socket(AF_NETLINK, SOCK_RAW | SOCK_CLOEXEC, NETLINK_ROUTE);
234 	if (!ASSERT_GE(sock, 0, "netlink socket"))
235 		goto done;
236 	ret = send(sock, nh, nh->nlmsg_len, 0);
237 	if (!ASSERT_EQ(ret, nh->nlmsg_len, "netlink send length"))
238 		goto done;
239 
240 	ret = 0;
241 done:
242 	if (sock != -1)
243 		close(sock);
244 	if (nstoken)
245 		close_netns(nstoken);
246 	return ret;
247 }
248 
config_overlay(void)249 static int config_overlay(void)
250 {
251 	if (setup_xfrm_tunnel(NS0, NS1, IP4_ADDR_VETH01, IP4_ADDR_VETH10,
252 			      IF_ID_0_TO_1, IF_ID_1))
253 		goto fail;
254 	if (setup_xfrm_tunnel(NS0, NS2, IP4_ADDR_VETH02, IP4_ADDR_VETH20,
255 			      IF_ID_0_TO_2, IF_ID_2))
256 		goto fail;
257 
258 	/* Older iproute2 doesn't support this option */
259 	if (!ASSERT_OK(setup_xfrmi_external_dev(NS0), "xfrmi"))
260 		goto fail;
261 
262 	SYS(fail, "ip -net " NS0 " addr add 192.168.1.100/24 dev ipsec0");
263 	SYS(fail, "ip -net " NS0 " link set dev ipsec0 up");
264 
265 	SYS(fail, "ip -net " NS1 " link add ipsec0 type xfrm if_id %d", IF_ID_1);
266 	SYS(fail, "ip -net " NS1 " addr add 192.168.1.200/24 dev ipsec0");
267 	SYS(fail, "ip -net " NS1 " link set dev ipsec0 up");
268 
269 	SYS(fail, "ip -net " NS2 " link add ipsec0 type xfrm if_id %d", IF_ID_2);
270 	SYS(fail, "ip -net " NS2 " addr add 192.168.1.200/24 dev ipsec0");
271 	SYS(fail, "ip -net " NS2 " link set dev ipsec0 up");
272 
273 	return 0;
274 fail:
275 	return -1;
276 }
277 
test_xfrm_ping(struct xfrm_info * skel,u32 if_id)278 static int test_xfrm_ping(struct xfrm_info *skel, u32 if_id)
279 {
280 	skel->bss->req_if_id = if_id;
281 
282 	SYS(fail, "ping -i 0.01 -c 3 -w 10 -q 192.168.1.200 > /dev/null");
283 
284 	if (!ASSERT_EQ(skel->bss->resp_if_id, if_id, "if_id"))
285 		goto fail;
286 
287 	return 0;
288 fail:
289 	return -1;
290 }
291 
_test_xfrm_info(void)292 static void _test_xfrm_info(void)
293 {
294 	LIBBPF_OPTS(bpf_tc_hook, tc_hook, .attach_point = BPF_TC_INGRESS);
295 	int get_xfrm_info_prog_fd, set_xfrm_info_prog_fd;
296 	struct nstoken *nstoken = NULL;
297 	struct xfrm_info *skel;
298 	int ifindex;
299 
300 	/* load and attach bpf progs to ipsec dev tc hook point */
301 	skel = xfrm_info__open_and_load();
302 	if (!ASSERT_OK_PTR(skel, "xfrm_info__open_and_load"))
303 		goto done;
304 	nstoken = open_netns(NS0);
305 	if (!ASSERT_OK_PTR(nstoken, "setns " NS0))
306 		goto done;
307 	ifindex = if_nametoindex("ipsec0");
308 	if (!ASSERT_NEQ(ifindex, 0, "ipsec0 ifindex"))
309 		goto done;
310 	tc_hook.ifindex = ifindex;
311 	set_xfrm_info_prog_fd = bpf_program__fd(skel->progs.set_xfrm_info);
312 	get_xfrm_info_prog_fd = bpf_program__fd(skel->progs.get_xfrm_info);
313 	if (!ASSERT_GE(set_xfrm_info_prog_fd, 0, "bpf_program__fd"))
314 		goto done;
315 	if (!ASSERT_GE(get_xfrm_info_prog_fd, 0, "bpf_program__fd"))
316 		goto done;
317 	if (attach_tc_prog(&tc_hook, get_xfrm_info_prog_fd,
318 			   set_xfrm_info_prog_fd))
319 		goto done;
320 
321 	/* perform test */
322 	if (!ASSERT_EQ(test_xfrm_ping(skel, IF_ID_0_TO_1), 0, "ping " NS1))
323 		goto done;
324 	if (!ASSERT_EQ(test_xfrm_ping(skel, IF_ID_0_TO_2), 0, "ping " NS2))
325 		goto done;
326 
327 done:
328 	if (nstoken)
329 		close_netns(nstoken);
330 	xfrm_info__destroy(skel);
331 }
332 
test_xfrm_info(void)333 void test_xfrm_info(void)
334 {
335 	cleanup();
336 
337 	if (!ASSERT_OK(config_underlay(), "config_underlay"))
338 		goto done;
339 	if (!ASSERT_OK(config_overlay(), "config_overlay"))
340 		goto done;
341 
342 	if (test__start_subtest("xfrm_info"))
343 		_test_xfrm_info();
344 
345 done:
346 	cleanup();
347 }
348