1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2019 Facebook */
3
4 #include <sys/socket.h>
5 #include <sys/epoll.h>
6 #include <netinet/in.h>
7 #include <arpa/inet.h>
8 #include <unistd.h>
9 #include <stdlib.h>
10 #include <string.h>
11 #include <errno.h>
12
13 #include <bpf/bpf.h>
14 #include <bpf/libbpf.h>
15
16 #include "cgroup_helpers.h"
17 #include "bpf_rlimit.h"
18
19 enum bpf_addr_array_idx {
20 ADDR_SRV_IDX,
21 ADDR_CLI_IDX,
22 __NR_BPF_ADDR_ARRAY_IDX,
23 };
24
25 enum bpf_result_array_idx {
26 EGRESS_SRV_IDX,
27 EGRESS_CLI_IDX,
28 INGRESS_LISTEN_IDX,
29 __NR_BPF_RESULT_ARRAY_IDX,
30 };
31
32 enum bpf_linum_array_idx {
33 EGRESS_LINUM_IDX,
34 INGRESS_LINUM_IDX,
35 __NR_BPF_LINUM_ARRAY_IDX,
36 };
37
38 struct bpf_spinlock_cnt {
39 struct bpf_spin_lock lock;
40 __u32 cnt;
41 };
42
43 #define CHECK(condition, tag, format...) ({ \
44 int __ret = !!(condition); \
45 if (__ret) { \
46 printf("%s(%d):FAIL:%s ", __func__, __LINE__, tag); \
47 printf(format); \
48 printf("\n"); \
49 exit(-1); \
50 } \
51 })
52
53 #define TEST_CGROUP "/test-bpf-sock-fields"
54 #define DATA "Hello BPF!"
55 #define DATA_LEN sizeof(DATA)
56
57 static struct sockaddr_in6 srv_sa6, cli_sa6;
58 static int sk_pkt_out_cnt10_fd;
59 static int sk_pkt_out_cnt_fd;
60 static int linum_map_fd;
61 static int addr_map_fd;
62 static int tp_map_fd;
63 static int sk_map_fd;
64
65 static __u32 addr_srv_idx = ADDR_SRV_IDX;
66 static __u32 addr_cli_idx = ADDR_CLI_IDX;
67
68 static __u32 egress_srv_idx = EGRESS_SRV_IDX;
69 static __u32 egress_cli_idx = EGRESS_CLI_IDX;
70 static __u32 ingress_listen_idx = INGRESS_LISTEN_IDX;
71
72 static __u32 egress_linum_idx = EGRESS_LINUM_IDX;
73 static __u32 ingress_linum_idx = INGRESS_LINUM_IDX;
74
init_loopback6(struct sockaddr_in6 * sa6)75 static void init_loopback6(struct sockaddr_in6 *sa6)
76 {
77 memset(sa6, 0, sizeof(*sa6));
78 sa6->sin6_family = AF_INET6;
79 sa6->sin6_addr = in6addr_loopback;
80 }
81
print_sk(const struct bpf_sock * sk)82 static void print_sk(const struct bpf_sock *sk)
83 {
84 char src_ip4[24], dst_ip4[24];
85 char src_ip6[64], dst_ip6[64];
86
87 inet_ntop(AF_INET, &sk->src_ip4, src_ip4, sizeof(src_ip4));
88 inet_ntop(AF_INET6, &sk->src_ip6, src_ip6, sizeof(src_ip6));
89 inet_ntop(AF_INET, &sk->dst_ip4, dst_ip4, sizeof(dst_ip4));
90 inet_ntop(AF_INET6, &sk->dst_ip6, dst_ip6, sizeof(dst_ip6));
91
92 printf("state:%u bound_dev_if:%u family:%u type:%u protocol:%u mark:%u priority:%u "
93 "src_ip4:%x(%s) src_ip6:%x:%x:%x:%x(%s) src_port:%u "
94 "dst_ip4:%x(%s) dst_ip6:%x:%x:%x:%x(%s) dst_port:%u\n",
95 sk->state, sk->bound_dev_if, sk->family, sk->type, sk->protocol,
96 sk->mark, sk->priority,
97 sk->src_ip4, src_ip4,
98 sk->src_ip6[0], sk->src_ip6[1], sk->src_ip6[2], sk->src_ip6[3],
99 src_ip6, sk->src_port,
100 sk->dst_ip4, dst_ip4,
101 sk->dst_ip6[0], sk->dst_ip6[1], sk->dst_ip6[2], sk->dst_ip6[3],
102 dst_ip6, ntohs(sk->dst_port));
103 }
104
print_tp(const struct bpf_tcp_sock * tp)105 static void print_tp(const struct bpf_tcp_sock *tp)
106 {
107 printf("snd_cwnd:%u srtt_us:%u rtt_min:%u snd_ssthresh:%u rcv_nxt:%u "
108 "snd_nxt:%u snd:una:%u mss_cache:%u ecn_flags:%u "
109 "rate_delivered:%u rate_interval_us:%u packets_out:%u "
110 "retrans_out:%u total_retrans:%u segs_in:%u data_segs_in:%u "
111 "segs_out:%u data_segs_out:%u lost_out:%u sacked_out:%u "
112 "bytes_received:%llu bytes_acked:%llu\n",
113 tp->snd_cwnd, tp->srtt_us, tp->rtt_min, tp->snd_ssthresh,
114 tp->rcv_nxt, tp->snd_nxt, tp->snd_una, tp->mss_cache,
115 tp->ecn_flags, tp->rate_delivered, tp->rate_interval_us,
116 tp->packets_out, tp->retrans_out, tp->total_retrans,
117 tp->segs_in, tp->data_segs_in, tp->segs_out,
118 tp->data_segs_out, tp->lost_out, tp->sacked_out,
119 tp->bytes_received, tp->bytes_acked);
120 }
121
check_result(void)122 static void check_result(void)
123 {
124 struct bpf_tcp_sock srv_tp, cli_tp, listen_tp;
125 struct bpf_sock srv_sk, cli_sk, listen_sk;
126 __u32 ingress_linum, egress_linum;
127 int err;
128
129 err = bpf_map_lookup_elem(linum_map_fd, &egress_linum_idx,
130 &egress_linum);
131 CHECK(err == -1, "bpf_map_lookup_elem(linum_map_fd)",
132 "err:%d errno:%d", err, errno);
133
134 err = bpf_map_lookup_elem(linum_map_fd, &ingress_linum_idx,
135 &ingress_linum);
136 CHECK(err == -1, "bpf_map_lookup_elem(linum_map_fd)",
137 "err:%d errno:%d", err, errno);
138
139 err = bpf_map_lookup_elem(sk_map_fd, &egress_srv_idx, &srv_sk);
140 CHECK(err == -1, "bpf_map_lookup_elem(sk_map_fd, &egress_srv_idx)",
141 "err:%d errno:%d", err, errno);
142 err = bpf_map_lookup_elem(tp_map_fd, &egress_srv_idx, &srv_tp);
143 CHECK(err == -1, "bpf_map_lookup_elem(tp_map_fd, &egress_srv_idx)",
144 "err:%d errno:%d", err, errno);
145
146 err = bpf_map_lookup_elem(sk_map_fd, &egress_cli_idx, &cli_sk);
147 CHECK(err == -1, "bpf_map_lookup_elem(sk_map_fd, &egress_cli_idx)",
148 "err:%d errno:%d", err, errno);
149 err = bpf_map_lookup_elem(tp_map_fd, &egress_cli_idx, &cli_tp);
150 CHECK(err == -1, "bpf_map_lookup_elem(tp_map_fd, &egress_cli_idx)",
151 "err:%d errno:%d", err, errno);
152
153 err = bpf_map_lookup_elem(sk_map_fd, &ingress_listen_idx, &listen_sk);
154 CHECK(err == -1, "bpf_map_lookup_elem(sk_map_fd, &ingress_listen_idx)",
155 "err:%d errno:%d", err, errno);
156 err = bpf_map_lookup_elem(tp_map_fd, &ingress_listen_idx, &listen_tp);
157 CHECK(err == -1, "bpf_map_lookup_elem(tp_map_fd, &ingress_listen_idx)",
158 "err:%d errno:%d", err, errno);
159
160 printf("listen_sk: ");
161 print_sk(&listen_sk);
162 printf("\n");
163
164 printf("srv_sk: ");
165 print_sk(&srv_sk);
166 printf("\n");
167
168 printf("cli_sk: ");
169 print_sk(&cli_sk);
170 printf("\n");
171
172 printf("listen_tp: ");
173 print_tp(&listen_tp);
174 printf("\n");
175
176 printf("srv_tp: ");
177 print_tp(&srv_tp);
178 printf("\n");
179
180 printf("cli_tp: ");
181 print_tp(&cli_tp);
182 printf("\n");
183
184 CHECK(listen_sk.state != 10 ||
185 listen_sk.family != AF_INET6 ||
186 listen_sk.protocol != IPPROTO_TCP ||
187 memcmp(listen_sk.src_ip6, &in6addr_loopback,
188 sizeof(listen_sk.src_ip6)) ||
189 listen_sk.dst_ip6[0] || listen_sk.dst_ip6[1] ||
190 listen_sk.dst_ip6[2] || listen_sk.dst_ip6[3] ||
191 listen_sk.src_port != ntohs(srv_sa6.sin6_port) ||
192 listen_sk.dst_port,
193 "Unexpected listen_sk",
194 "Check listen_sk output. ingress_linum:%u",
195 ingress_linum);
196
197 CHECK(srv_sk.state == 10 ||
198 !srv_sk.state ||
199 srv_sk.family != AF_INET6 ||
200 srv_sk.protocol != IPPROTO_TCP ||
201 memcmp(srv_sk.src_ip6, &in6addr_loopback,
202 sizeof(srv_sk.src_ip6)) ||
203 memcmp(srv_sk.dst_ip6, &in6addr_loopback,
204 sizeof(srv_sk.dst_ip6)) ||
205 srv_sk.src_port != ntohs(srv_sa6.sin6_port) ||
206 srv_sk.dst_port != cli_sa6.sin6_port,
207 "Unexpected srv_sk", "Check srv_sk output. egress_linum:%u",
208 egress_linum);
209
210 CHECK(cli_sk.state == 10 ||
211 !cli_sk.state ||
212 cli_sk.family != AF_INET6 ||
213 cli_sk.protocol != IPPROTO_TCP ||
214 memcmp(cli_sk.src_ip6, &in6addr_loopback,
215 sizeof(cli_sk.src_ip6)) ||
216 memcmp(cli_sk.dst_ip6, &in6addr_loopback,
217 sizeof(cli_sk.dst_ip6)) ||
218 cli_sk.src_port != ntohs(cli_sa6.sin6_port) ||
219 cli_sk.dst_port != srv_sa6.sin6_port,
220 "Unexpected cli_sk", "Check cli_sk output. egress_linum:%u",
221 egress_linum);
222
223 CHECK(listen_tp.data_segs_out ||
224 listen_tp.data_segs_in ||
225 listen_tp.total_retrans ||
226 listen_tp.bytes_acked,
227 "Unexpected listen_tp", "Check listen_tp output. ingress_linum:%u",
228 ingress_linum);
229
230 CHECK(srv_tp.data_segs_out != 2 ||
231 srv_tp.data_segs_in ||
232 srv_tp.snd_cwnd != 10 ||
233 srv_tp.total_retrans ||
234 srv_tp.bytes_acked != 2 * DATA_LEN,
235 "Unexpected srv_tp", "Check srv_tp output. egress_linum:%u",
236 egress_linum);
237
238 CHECK(cli_tp.data_segs_out ||
239 cli_tp.data_segs_in != 2 ||
240 cli_tp.snd_cwnd != 10 ||
241 cli_tp.total_retrans ||
242 cli_tp.bytes_received != 2 * DATA_LEN,
243 "Unexpected cli_tp", "Check cli_tp output. egress_linum:%u",
244 egress_linum);
245 }
246
check_sk_pkt_out_cnt(int accept_fd,int cli_fd)247 static void check_sk_pkt_out_cnt(int accept_fd, int cli_fd)
248 {
249 struct bpf_spinlock_cnt pkt_out_cnt = {}, pkt_out_cnt10 = {};
250 int err;
251
252 pkt_out_cnt.cnt = ~0;
253 pkt_out_cnt10.cnt = ~0;
254 err = bpf_map_lookup_elem(sk_pkt_out_cnt_fd, &accept_fd, &pkt_out_cnt);
255 if (!err)
256 err = bpf_map_lookup_elem(sk_pkt_out_cnt10_fd, &accept_fd,
257 &pkt_out_cnt10);
258
259 /* The bpf prog only counts for fullsock and
260 * passive conneciton did not become fullsock until 3WHS
261 * had been finished.
262 * The bpf prog only counted two data packet out but we
263 * specially init accept_fd's pkt_out_cnt by 2 in
264 * init_sk_storage(). Hence, 4 here.
265 */
266 CHECK(err || pkt_out_cnt.cnt != 4 || pkt_out_cnt10.cnt != 40,
267 "bpf_map_lookup_elem(sk_pkt_out_cnt, &accept_fd)",
268 "err:%d errno:%d pkt_out_cnt:%u pkt_out_cnt10:%u",
269 err, errno, pkt_out_cnt.cnt, pkt_out_cnt10.cnt);
270
271 pkt_out_cnt.cnt = ~0;
272 pkt_out_cnt10.cnt = ~0;
273 err = bpf_map_lookup_elem(sk_pkt_out_cnt_fd, &cli_fd, &pkt_out_cnt);
274 if (!err)
275 err = bpf_map_lookup_elem(sk_pkt_out_cnt10_fd, &cli_fd,
276 &pkt_out_cnt10);
277 /* Active connection is fullsock from the beginning.
278 * 1 SYN and 1 ACK during 3WHS
279 * 2 Acks on data packet.
280 *
281 * The bpf_prog initialized it to 0xeB9F.
282 */
283 CHECK(err || pkt_out_cnt.cnt != 0xeB9F + 4 ||
284 pkt_out_cnt10.cnt != 0xeB9F + 40,
285 "bpf_map_lookup_elem(sk_pkt_out_cnt, &cli_fd)",
286 "err:%d errno:%d pkt_out_cnt:%u pkt_out_cnt10:%u",
287 err, errno, pkt_out_cnt.cnt, pkt_out_cnt10.cnt);
288 }
289
init_sk_storage(int sk_fd,__u32 pkt_out_cnt)290 static void init_sk_storage(int sk_fd, __u32 pkt_out_cnt)
291 {
292 struct bpf_spinlock_cnt scnt = {};
293 int err;
294
295 scnt.cnt = pkt_out_cnt;
296 err = bpf_map_update_elem(sk_pkt_out_cnt_fd, &sk_fd, &scnt,
297 BPF_NOEXIST);
298 CHECK(err, "bpf_map_update_elem(sk_pkt_out_cnt_fd)",
299 "err:%d errno:%d", err, errno);
300
301 scnt.cnt *= 10;
302 err = bpf_map_update_elem(sk_pkt_out_cnt10_fd, &sk_fd, &scnt,
303 BPF_NOEXIST);
304 CHECK(err, "bpf_map_update_elem(sk_pkt_out_cnt10_fd)",
305 "err:%d errno:%d", err, errno);
306 }
307
test(void)308 static void test(void)
309 {
310 int listen_fd, cli_fd, accept_fd, epfd, err;
311 struct epoll_event ev;
312 socklen_t addrlen;
313 int i;
314
315 addrlen = sizeof(struct sockaddr_in6);
316 ev.events = EPOLLIN;
317
318 epfd = epoll_create(1);
319 CHECK(epfd == -1, "epoll_create()", "epfd:%d errno:%d", epfd, errno);
320
321 /* Prepare listen_fd */
322 listen_fd = socket(AF_INET6, SOCK_STREAM | SOCK_NONBLOCK, 0);
323 CHECK(listen_fd == -1, "socket()", "listen_fd:%d errno:%d",
324 listen_fd, errno);
325
326 init_loopback6(&srv_sa6);
327 err = bind(listen_fd, (struct sockaddr *)&srv_sa6, sizeof(srv_sa6));
328 CHECK(err, "bind(listen_fd)", "err:%d errno:%d", err, errno);
329
330 err = getsockname(listen_fd, (struct sockaddr *)&srv_sa6, &addrlen);
331 CHECK(err, "getsockname(listen_fd)", "err:%d errno:%d", err, errno);
332
333 err = listen(listen_fd, 1);
334 CHECK(err, "listen(listen_fd)", "err:%d errno:%d", err, errno);
335
336 /* Prepare cli_fd */
337 cli_fd = socket(AF_INET6, SOCK_STREAM | SOCK_NONBLOCK, 0);
338 CHECK(cli_fd == -1, "socket()", "cli_fd:%d errno:%d", cli_fd, errno);
339
340 init_loopback6(&cli_sa6);
341 err = bind(cli_fd, (struct sockaddr *)&cli_sa6, sizeof(cli_sa6));
342 CHECK(err, "bind(cli_fd)", "err:%d errno:%d", err, errno);
343
344 err = getsockname(cli_fd, (struct sockaddr *)&cli_sa6, &addrlen);
345 CHECK(err, "getsockname(cli_fd)", "err:%d errno:%d",
346 err, errno);
347
348 /* Update addr_map with srv_sa6 and cli_sa6 */
349 err = bpf_map_update_elem(addr_map_fd, &addr_srv_idx, &srv_sa6, 0);
350 CHECK(err, "map_update", "err:%d errno:%d", err, errno);
351
352 err = bpf_map_update_elem(addr_map_fd, &addr_cli_idx, &cli_sa6, 0);
353 CHECK(err, "map_update", "err:%d errno:%d", err, errno);
354
355 /* Connect from cli_sa6 to srv_sa6 */
356 err = connect(cli_fd, (struct sockaddr *)&srv_sa6, addrlen);
357 printf("srv_sa6.sin6_port:%u cli_sa6.sin6_port:%u\n\n",
358 ntohs(srv_sa6.sin6_port), ntohs(cli_sa6.sin6_port));
359 CHECK(err && errno != EINPROGRESS,
360 "connect(cli_fd)", "err:%d errno:%d", err, errno);
361
362 ev.data.fd = listen_fd;
363 err = epoll_ctl(epfd, EPOLL_CTL_ADD, listen_fd, &ev);
364 CHECK(err, "epoll_ctl(EPOLL_CTL_ADD, listen_fd)", "err:%d errno:%d",
365 err, errno);
366
367 /* Accept the connection */
368 /* Have some timeout in accept(listen_fd). Just in case. */
369 err = epoll_wait(epfd, &ev, 1, 1000);
370 CHECK(err != 1 || ev.data.fd != listen_fd,
371 "epoll_wait(listen_fd)",
372 "err:%d errno:%d ev.data.fd:%d listen_fd:%d",
373 err, errno, ev.data.fd, listen_fd);
374
375 accept_fd = accept(listen_fd, NULL, NULL);
376 CHECK(accept_fd == -1, "accept(listen_fd)", "accept_fd:%d errno:%d",
377 accept_fd, errno);
378 close(listen_fd);
379
380 ev.data.fd = cli_fd;
381 err = epoll_ctl(epfd, EPOLL_CTL_ADD, cli_fd, &ev);
382 CHECK(err, "epoll_ctl(EPOLL_CTL_ADD, cli_fd)", "err:%d errno:%d",
383 err, errno);
384
385 init_sk_storage(accept_fd, 2);
386
387 for (i = 0; i < 2; i++) {
388 /* Send some data from accept_fd to cli_fd */
389 err = send(accept_fd, DATA, DATA_LEN, 0);
390 CHECK(err != DATA_LEN, "send(accept_fd)", "err:%d errno:%d",
391 err, errno);
392
393 /* Have some timeout in recv(cli_fd). Just in case. */
394 err = epoll_wait(epfd, &ev, 1, 1000);
395 CHECK(err != 1 || ev.data.fd != cli_fd,
396 "epoll_wait(cli_fd)", "err:%d errno:%d ev.data.fd:%d cli_fd:%d",
397 err, errno, ev.data.fd, cli_fd);
398
399 err = recv(cli_fd, NULL, 0, MSG_TRUNC);
400 CHECK(err, "recv(cli_fd)", "err:%d errno:%d", err, errno);
401 }
402
403 check_sk_pkt_out_cnt(accept_fd, cli_fd);
404
405 close(epfd);
406 close(accept_fd);
407 close(cli_fd);
408
409 check_result();
410 }
411
main(int argc,char ** argv)412 int main(int argc, char **argv)
413 {
414 struct bpf_prog_load_attr attr = {
415 .file = "test_sock_fields_kern.o",
416 .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
417 .prog_flags = BPF_F_TEST_RND_HI32,
418 };
419 int cgroup_fd, egress_fd, ingress_fd, err;
420 struct bpf_program *ingress_prog;
421 struct bpf_object *obj;
422 struct bpf_map *map;
423
424 err = setup_cgroup_environment();
425 CHECK(err, "setup_cgroup_environment()", "err:%d errno:%d",
426 err, errno);
427
428 atexit(cleanup_cgroup_environment);
429
430 /* Create a cgroup, get fd, and join it */
431 cgroup_fd = create_and_get_cgroup(TEST_CGROUP);
432 CHECK(cgroup_fd == -1, "create_and_get_cgroup()",
433 "cgroup_fd:%d errno:%d", cgroup_fd, errno);
434
435 err = join_cgroup(TEST_CGROUP);
436 CHECK(err, "join_cgroup", "err:%d errno:%d", err, errno);
437
438 err = bpf_prog_load_xattr(&attr, &obj, &egress_fd);
439 CHECK(err, "bpf_prog_load_xattr()", "err:%d", err);
440
441 ingress_prog = bpf_object__find_program_by_title(obj,
442 "cgroup_skb/ingress");
443 CHECK(!ingress_prog,
444 "bpf_object__find_program_by_title(cgroup_skb/ingress)",
445 "not found");
446 ingress_fd = bpf_program__fd(ingress_prog);
447
448 err = bpf_prog_attach(egress_fd, cgroup_fd, BPF_CGROUP_INET_EGRESS, 0);
449 CHECK(err == -1, "bpf_prog_attach(CPF_CGROUP_INET_EGRESS)",
450 "err:%d errno%d", err, errno);
451
452 err = bpf_prog_attach(ingress_fd, cgroup_fd,
453 BPF_CGROUP_INET_INGRESS, 0);
454 CHECK(err == -1, "bpf_prog_attach(CPF_CGROUP_INET_INGRESS)",
455 "err:%d errno%d", err, errno);
456 close(cgroup_fd);
457
458 map = bpf_object__find_map_by_name(obj, "addr_map");
459 CHECK(!map, "cannot find addr_map", "(null)");
460 addr_map_fd = bpf_map__fd(map);
461
462 map = bpf_object__find_map_by_name(obj, "sock_result_map");
463 CHECK(!map, "cannot find sock_result_map", "(null)");
464 sk_map_fd = bpf_map__fd(map);
465
466 map = bpf_object__find_map_by_name(obj, "tcp_sock_result_map");
467 CHECK(!map, "cannot find tcp_sock_result_map", "(null)");
468 tp_map_fd = bpf_map__fd(map);
469
470 map = bpf_object__find_map_by_name(obj, "linum_map");
471 CHECK(!map, "cannot find linum_map", "(null)");
472 linum_map_fd = bpf_map__fd(map);
473
474 map = bpf_object__find_map_by_name(obj, "sk_pkt_out_cnt");
475 CHECK(!map, "cannot find sk_pkt_out_cnt", "(null)");
476 sk_pkt_out_cnt_fd = bpf_map__fd(map);
477
478 map = bpf_object__find_map_by_name(obj, "sk_pkt_out_cnt10");
479 CHECK(!map, "cannot find sk_pkt_out_cnt10", "(null)");
480 sk_pkt_out_cnt10_fd = bpf_map__fd(map);
481
482 test();
483
484 bpf_object__close(obj);
485 cleanup_cgroup_environment();
486
487 printf("PASS\n");
488
489 return 0;
490 }
491