xref: /linux/tools/testing/selftests/bpf/prog_tests/mptcp.c (revision e6bf1f7aea4df918e5ee848d9f6b7ce63135b4be)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2020, Tessares SA. */
3 /* Copyright (c) 2022, SUSE. */
4 
5 #include <linux/const.h>
6 #include <netinet/in.h>
7 #include <test_progs.h>
8 #include <unistd.h>
9 #include "cgroup_helpers.h"
10 #include "network_helpers.h"
11 #include "mptcp_sock.skel.h"
12 #include "mptcpify.skel.h"
13 #include "mptcp_subflow.skel.h"
14 
15 #define NS_TEST "mptcp_ns"
16 #define ADDR_1	"10.0.1.1"
17 #define ADDR_2	"10.0.1.2"
18 #define PORT_1	10001
19 
20 #ifndef IPPROTO_MPTCP
21 #define IPPROTO_MPTCP 262
22 #endif
23 
24 #ifndef SOL_MPTCP
25 #define SOL_MPTCP 284
26 #endif
27 #ifndef MPTCP_INFO
28 #define MPTCP_INFO		1
29 #endif
30 #ifndef MPTCP_INFO_FLAG_FALLBACK
31 #define MPTCP_INFO_FLAG_FALLBACK		_BITUL(0)
32 #endif
33 #ifndef MPTCP_INFO_FLAG_REMOTE_KEY_RECEIVED
34 #define MPTCP_INFO_FLAG_REMOTE_KEY_RECEIVED	_BITUL(1)
35 #endif
36 
37 #ifndef TCP_CA_NAME_MAX
38 #define TCP_CA_NAME_MAX	16
39 #endif
40 
41 struct __mptcp_info {
42 	__u8	mptcpi_subflows;
43 	__u8	mptcpi_add_addr_signal;
44 	__u8	mptcpi_add_addr_accepted;
45 	__u8	mptcpi_subflows_max;
46 	__u8	mptcpi_add_addr_signal_max;
47 	__u8	mptcpi_add_addr_accepted_max;
48 	__u32	mptcpi_flags;
49 	__u32	mptcpi_token;
50 	__u64	mptcpi_write_seq;
51 	__u64	mptcpi_snd_una;
52 	__u64	mptcpi_rcv_nxt;
53 	__u8	mptcpi_local_addr_used;
54 	__u8	mptcpi_local_addr_max;
55 	__u8	mptcpi_csum_enabled;
56 	__u32	mptcpi_retransmits;
57 	__u64	mptcpi_bytes_retrans;
58 	__u64	mptcpi_bytes_sent;
59 	__u64	mptcpi_bytes_received;
60 	__u64	mptcpi_bytes_acked;
61 };
62 
63 struct mptcp_storage {
64 	__u32 invoked;
65 	__u32 is_mptcp;
66 	struct sock *sk;
67 	__u32 token;
68 	struct sock *first;
69 	char ca_name[TCP_CA_NAME_MAX];
70 };
71 
72 static struct nstoken *create_netns(void)
73 {
74 	SYS(fail, "ip netns add %s", NS_TEST);
75 	SYS(fail, "ip -net %s link set dev lo up", NS_TEST);
76 
77 	return open_netns(NS_TEST);
78 fail:
79 	return NULL;
80 }
81 
82 static void cleanup_netns(struct nstoken *nstoken)
83 {
84 	if (nstoken)
85 		close_netns(nstoken);
86 
87 	SYS_NOFAIL("ip netns del %s", NS_TEST);
88 }
89 
90 static int start_mptcp_server(int family, const char *addr_str, __u16 port,
91 			      int timeout_ms)
92 {
93 	struct network_helper_opts opts = {
94 		.timeout_ms	= timeout_ms,
95 		.proto		= IPPROTO_MPTCP,
96 	};
97 
98 	return start_server_str(family, SOCK_STREAM, addr_str, port, &opts);
99 }
100 
101 static int verify_tsk(int map_fd, int client_fd)
102 {
103 	int err, cfd = client_fd;
104 	struct mptcp_storage val;
105 
106 	err = bpf_map_lookup_elem(map_fd, &cfd, &val);
107 	if (!ASSERT_OK(err, "bpf_map_lookup_elem"))
108 		return err;
109 
110 	if (!ASSERT_EQ(val.invoked, 1, "unexpected invoked count"))
111 		err++;
112 
113 	if (!ASSERT_EQ(val.is_mptcp, 0, "unexpected is_mptcp"))
114 		err++;
115 
116 	return err;
117 }
118 
119 static void get_msk_ca_name(char ca_name[])
120 {
121 	size_t len;
122 	int fd;
123 
124 	fd = open("/proc/sys/net/ipv4/tcp_congestion_control", O_RDONLY);
125 	if (!ASSERT_GE(fd, 0, "failed to open tcp_congestion_control"))
126 		return;
127 
128 	len = read(fd, ca_name, TCP_CA_NAME_MAX);
129 	if (!ASSERT_GT(len, 0, "failed to read ca_name"))
130 		goto err;
131 
132 	if (len > 0 && ca_name[len - 1] == '\n')
133 		ca_name[len - 1] = '\0';
134 
135 err:
136 	close(fd);
137 }
138 
139 static int verify_msk(int map_fd, int client_fd, __u32 token)
140 {
141 	char ca_name[TCP_CA_NAME_MAX];
142 	int err, cfd = client_fd;
143 	struct mptcp_storage val;
144 
145 	if (!ASSERT_GT(token, 0, "invalid token"))
146 		return -1;
147 
148 	get_msk_ca_name(ca_name);
149 
150 	err = bpf_map_lookup_elem(map_fd, &cfd, &val);
151 	if (!ASSERT_OK(err, "bpf_map_lookup_elem"))
152 		return err;
153 
154 	if (!ASSERT_EQ(val.invoked, 1, "unexpected invoked count"))
155 		err++;
156 
157 	if (!ASSERT_EQ(val.is_mptcp, 1, "unexpected is_mptcp"))
158 		err++;
159 
160 	if (!ASSERT_EQ(val.token, token, "unexpected token"))
161 		err++;
162 
163 	if (!ASSERT_EQ(val.first, val.sk, "unexpected first"))
164 		err++;
165 
166 	if (!ASSERT_STRNEQ(val.ca_name, ca_name, TCP_CA_NAME_MAX, "unexpected ca_name"))
167 		err++;
168 
169 	return err;
170 }
171 
172 static int run_test(int cgroup_fd, int server_fd, bool is_mptcp)
173 {
174 	int client_fd, prog_fd, map_fd, err;
175 	struct mptcp_sock *sock_skel;
176 
177 	sock_skel = mptcp_sock__open_and_load();
178 	if (!ASSERT_OK_PTR(sock_skel, "skel_open_load"))
179 		return libbpf_get_error(sock_skel);
180 
181 	err = mptcp_sock__attach(sock_skel);
182 	if (!ASSERT_OK(err, "skel_attach"))
183 		goto out;
184 
185 	prog_fd = bpf_program__fd(sock_skel->progs._sockops);
186 	map_fd = bpf_map__fd(sock_skel->maps.socket_storage_map);
187 	err = bpf_prog_attach(prog_fd, cgroup_fd, BPF_CGROUP_SOCK_OPS, 0);
188 	if (!ASSERT_OK(err, "bpf_prog_attach"))
189 		goto out;
190 
191 	client_fd = connect_to_fd(server_fd, 0);
192 	if (!ASSERT_GE(client_fd, 0, "connect to fd")) {
193 		err = -EIO;
194 		goto out;
195 	}
196 
197 	err += is_mptcp ? verify_msk(map_fd, client_fd, sock_skel->bss->token) :
198 			  verify_tsk(map_fd, client_fd);
199 
200 	close(client_fd);
201 
202 out:
203 	mptcp_sock__destroy(sock_skel);
204 	return err;
205 }
206 
207 static void test_base(void)
208 {
209 	struct nstoken *nstoken = NULL;
210 	int server_fd, cgroup_fd;
211 
212 	cgroup_fd = test__join_cgroup("/mptcp");
213 	if (!ASSERT_GE(cgroup_fd, 0, "test__join_cgroup"))
214 		return;
215 
216 	nstoken = create_netns();
217 	if (!ASSERT_OK_PTR(nstoken, "create_netns"))
218 		goto fail;
219 
220 	/* without MPTCP */
221 	server_fd = start_server(AF_INET, SOCK_STREAM, NULL, 0, 0);
222 	if (!ASSERT_GE(server_fd, 0, "start_server"))
223 		goto with_mptcp;
224 
225 	ASSERT_OK(run_test(cgroup_fd, server_fd, false), "run_test tcp");
226 
227 	close(server_fd);
228 
229 with_mptcp:
230 	/* with MPTCP */
231 	server_fd = start_mptcp_server(AF_INET, NULL, 0, 0);
232 	if (!ASSERT_GE(server_fd, 0, "start_mptcp_server"))
233 		goto fail;
234 
235 	ASSERT_OK(run_test(cgroup_fd, server_fd, true), "run_test mptcp");
236 
237 	close(server_fd);
238 
239 fail:
240 	cleanup_netns(nstoken);
241 	close(cgroup_fd);
242 }
243 
244 static void send_byte(int fd)
245 {
246 	char b = 0x55;
247 
248 	ASSERT_EQ(write(fd, &b, sizeof(b)), 1, "send single byte");
249 }
250 
251 static int verify_mptcpify(int server_fd, int client_fd)
252 {
253 	struct __mptcp_info info;
254 	socklen_t optlen;
255 	int protocol;
256 	int err = 0;
257 
258 	optlen = sizeof(protocol);
259 	if (!ASSERT_OK(getsockopt(server_fd, SOL_SOCKET, SO_PROTOCOL, &protocol, &optlen),
260 		       "getsockopt(SOL_PROTOCOL)"))
261 		return -1;
262 
263 	if (!ASSERT_EQ(protocol, IPPROTO_MPTCP, "protocol isn't MPTCP"))
264 		err++;
265 
266 	optlen = sizeof(info);
267 	if (!ASSERT_OK(getsockopt(client_fd, SOL_MPTCP, MPTCP_INFO, &info, &optlen),
268 		       "getsockopt(MPTCP_INFO)"))
269 		return -1;
270 
271 	if (!ASSERT_GE(info.mptcpi_flags, 0, "unexpected mptcpi_flags"))
272 		err++;
273 	if (!ASSERT_FALSE(info.mptcpi_flags & MPTCP_INFO_FLAG_FALLBACK,
274 			  "MPTCP fallback"))
275 		err++;
276 	if (!ASSERT_TRUE(info.mptcpi_flags & MPTCP_INFO_FLAG_REMOTE_KEY_RECEIVED,
277 			 "no remote key received"))
278 		err++;
279 
280 	return err;
281 }
282 
283 static int run_mptcpify(int cgroup_fd)
284 {
285 	int server_fd, client_fd, err = 0;
286 	struct mptcpify *mptcpify_skel;
287 
288 	mptcpify_skel = mptcpify__open_and_load();
289 	if (!ASSERT_OK_PTR(mptcpify_skel, "skel_open_load"))
290 		return libbpf_get_error(mptcpify_skel);
291 
292 	mptcpify_skel->bss->pid = getpid();
293 
294 	err = mptcpify__attach(mptcpify_skel);
295 	if (!ASSERT_OK(err, "skel_attach"))
296 		goto out;
297 
298 	/* without MPTCP */
299 	server_fd = start_server(AF_INET, SOCK_STREAM, NULL, 0, 0);
300 	if (!ASSERT_GE(server_fd, 0, "start_server")) {
301 		err = -EIO;
302 		goto out;
303 	}
304 
305 	client_fd = connect_to_fd(server_fd, 0);
306 	if (!ASSERT_GE(client_fd, 0, "connect to fd")) {
307 		err = -EIO;
308 		goto close_server;
309 	}
310 
311 	send_byte(client_fd);
312 
313 	err = verify_mptcpify(server_fd, client_fd);
314 
315 	close(client_fd);
316 close_server:
317 	close(server_fd);
318 out:
319 	mptcpify__destroy(mptcpify_skel);
320 	return err;
321 }
322 
323 static void test_mptcpify(void)
324 {
325 	struct nstoken *nstoken = NULL;
326 	int cgroup_fd;
327 
328 	cgroup_fd = test__join_cgroup("/mptcpify");
329 	if (!ASSERT_GE(cgroup_fd, 0, "test__join_cgroup"))
330 		return;
331 
332 	nstoken = create_netns();
333 	if (!ASSERT_OK_PTR(nstoken, "create_netns"))
334 		goto fail;
335 
336 	ASSERT_OK(run_mptcpify(cgroup_fd), "run_mptcpify");
337 
338 fail:
339 	cleanup_netns(nstoken);
340 	close(cgroup_fd);
341 }
342 
343 static int endpoint_init(char *flags)
344 {
345 	SYS(fail, "ip -net %s link add veth1 type veth peer name veth2", NS_TEST);
346 	SYS(fail, "ip -net %s addr add %s/24 dev veth1", NS_TEST, ADDR_1);
347 	SYS(fail, "ip -net %s link set dev veth1 up", NS_TEST);
348 	SYS(fail, "ip -net %s addr add %s/24 dev veth2", NS_TEST, ADDR_2);
349 	SYS(fail, "ip -net %s link set dev veth2 up", NS_TEST);
350 	if (SYS_NOFAIL("ip -net %s mptcp endpoint add %s %s", NS_TEST, ADDR_2, flags)) {
351 		printf("'ip mptcp' not supported, skip this test.\n");
352 		test__skip();
353 		goto fail;
354 	}
355 
356 	return 0;
357 fail:
358 	return -1;
359 }
360 
361 static void wait_for_new_subflows(int fd)
362 {
363 	socklen_t len;
364 	u8 subflows;
365 	int err, i;
366 
367 	len = sizeof(subflows);
368 	/* Wait max 5 sec for new subflows to be created */
369 	for (i = 0; i < 50; i++) {
370 		err = getsockopt(fd, SOL_MPTCP, MPTCP_INFO, &subflows, &len);
371 		if (!err && subflows > 0)
372 			break;
373 
374 		usleep(100000); /* 0.1s */
375 	}
376 }
377 
378 static void run_subflow(void)
379 {
380 	int server_fd, client_fd, err;
381 	char new[TCP_CA_NAME_MAX];
382 	char cc[TCP_CA_NAME_MAX];
383 	unsigned int mark;
384 	socklen_t len;
385 
386 	server_fd = start_mptcp_server(AF_INET, ADDR_1, PORT_1, 0);
387 	if (!ASSERT_OK_FD(server_fd, "start_mptcp_server"))
388 		return;
389 
390 	client_fd = connect_to_fd(server_fd, 0);
391 	if (!ASSERT_OK_FD(client_fd, "connect_to_fd"))
392 		goto close_server;
393 
394 	send_byte(client_fd);
395 	wait_for_new_subflows(client_fd);
396 
397 	len = sizeof(mark);
398 	err = getsockopt(client_fd, SOL_SOCKET, SO_MARK, &mark, &len);
399 	if (ASSERT_OK(err, "getsockopt(client_fd, SO_MARK)"))
400 		ASSERT_EQ(mark, 0, "mark");
401 
402 	len = sizeof(new);
403 	err = getsockopt(client_fd, SOL_TCP, TCP_CONGESTION, new, &len);
404 	if (ASSERT_OK(err, "getsockopt(client_fd, TCP_CONGESTION)")) {
405 		get_msk_ca_name(cc);
406 		ASSERT_STREQ(new, cc, "cc");
407 	}
408 
409 	close(client_fd);
410 close_server:
411 	close(server_fd);
412 }
413 
414 static void test_subflow(void)
415 {
416 	struct mptcp_subflow *skel;
417 	struct nstoken *nstoken;
418 	int cgroup_fd;
419 
420 	cgroup_fd = test__join_cgroup("/mptcp_subflow");
421 	if (!ASSERT_OK_FD(cgroup_fd, "join_cgroup: mptcp_subflow"))
422 		return;
423 
424 	skel = mptcp_subflow__open_and_load();
425 	if (!ASSERT_OK_PTR(skel, "skel_open_load: mptcp_subflow"))
426 		goto close_cgroup;
427 
428 	skel->bss->pid = getpid();
429 
430 	skel->links.mptcp_subflow =
431 		bpf_program__attach_cgroup(skel->progs.mptcp_subflow, cgroup_fd);
432 	if (!ASSERT_OK_PTR(skel->links.mptcp_subflow, "attach mptcp_subflow"))
433 		goto skel_destroy;
434 
435 	skel->links._getsockopt_subflow =
436 		bpf_program__attach_cgroup(skel->progs._getsockopt_subflow, cgroup_fd);
437 	if (!ASSERT_OK_PTR(skel->links._getsockopt_subflow, "attach _getsockopt_subflow"))
438 		goto skel_destroy;
439 
440 	nstoken = create_netns();
441 	if (!ASSERT_OK_PTR(nstoken, "create_netns: mptcp_subflow"))
442 		goto skel_destroy;
443 
444 	if (endpoint_init("subflow") < 0)
445 		goto close_netns;
446 
447 	run_subflow();
448 
449 close_netns:
450 	cleanup_netns(nstoken);
451 skel_destroy:
452 	mptcp_subflow__destroy(skel);
453 close_cgroup:
454 	close(cgroup_fd);
455 }
456 
457 void test_mptcp(void)
458 {
459 	if (test__start_subtest("base"))
460 		test_base();
461 	if (test__start_subtest("mptcpify"))
462 		test_mptcpify();
463 	if (test__start_subtest("subflow"))
464 		test_subflow();
465 }
466