1 // SPDX-License-Identifier: GPL-2.0 2 /* Copyright (c) 2019 Facebook */ 3 4 #include <netinet/in.h> 5 #include <arpa/inet.h> 6 #include <unistd.h> 7 #include <stdlib.h> 8 #include <string.h> 9 #include <errno.h> 10 11 #include <bpf/bpf.h> 12 #include <bpf/libbpf.h> 13 #include <linux/compiler.h> 14 15 #include "network_helpers.h" 16 #include "cgroup_helpers.h" 17 #include "test_progs.h" 18 #include "bpf_rlimit.h" 19 #include "test_sock_fields.skel.h" 20 21 enum bpf_linum_array_idx { 22 EGRESS_LINUM_IDX, 23 INGRESS_LINUM_IDX, 24 __NR_BPF_LINUM_ARRAY_IDX, 25 }; 26 27 struct bpf_spinlock_cnt { 28 struct bpf_spin_lock lock; 29 __u32 cnt; 30 }; 31 32 #define PARENT_CGROUP "/test-bpf-sock-fields" 33 #define CHILD_CGROUP "/test-bpf-sock-fields/child" 34 #define DATA "Hello BPF!" 35 #define DATA_LEN sizeof(DATA) 36 37 static struct sockaddr_in6 srv_sa6, cli_sa6; 38 static int sk_pkt_out_cnt10_fd; 39 static struct test_sock_fields *skel; 40 static int sk_pkt_out_cnt_fd; 41 static __u64 parent_cg_id; 42 static __u64 child_cg_id; 43 static int linum_map_fd; 44 static __u32 duration; 45 46 static __u32 egress_linum_idx = EGRESS_LINUM_IDX; 47 static __u32 ingress_linum_idx = INGRESS_LINUM_IDX; 48 49 static void print_sk(const struct bpf_sock *sk, const char *prefix) 50 { 51 char src_ip4[24], dst_ip4[24]; 52 char src_ip6[64], dst_ip6[64]; 53 54 inet_ntop(AF_INET, &sk->src_ip4, src_ip4, sizeof(src_ip4)); 55 inet_ntop(AF_INET6, &sk->src_ip6, src_ip6, sizeof(src_ip6)); 56 inet_ntop(AF_INET, &sk->dst_ip4, dst_ip4, sizeof(dst_ip4)); 57 inet_ntop(AF_INET6, &sk->dst_ip6, dst_ip6, sizeof(dst_ip6)); 58 59 printf("%s: state:%u bound_dev_if:%u family:%u type:%u protocol:%u mark:%u priority:%u " 60 "src_ip4:%x(%s) src_ip6:%x:%x:%x:%x(%s) src_port:%u " 61 "dst_ip4:%x(%s) dst_ip6:%x:%x:%x:%x(%s) dst_port:%u\n", 62 prefix, 63 sk->state, sk->bound_dev_if, sk->family, sk->type, sk->protocol, 64 sk->mark, sk->priority, 65 sk->src_ip4, src_ip4, 66 sk->src_ip6[0], sk->src_ip6[1], sk->src_ip6[2], sk->src_ip6[3], 67 src_ip6, sk->src_port, 68 sk->dst_ip4, dst_ip4, 69 sk->dst_ip6[0], sk->dst_ip6[1], sk->dst_ip6[2], sk->dst_ip6[3], 70 dst_ip6, ntohs(sk->dst_port)); 71 } 72 73 static void print_tp(const struct bpf_tcp_sock *tp, const char *prefix) 74 { 75 printf("%s: snd_cwnd:%u srtt_us:%u rtt_min:%u snd_ssthresh:%u rcv_nxt:%u " 76 "snd_nxt:%u snd:una:%u mss_cache:%u ecn_flags:%u " 77 "rate_delivered:%u rate_interval_us:%u packets_out:%u " 78 "retrans_out:%u total_retrans:%u segs_in:%u data_segs_in:%u " 79 "segs_out:%u data_segs_out:%u lost_out:%u sacked_out:%u " 80 "bytes_received:%llu bytes_acked:%llu\n", 81 prefix, 82 tp->snd_cwnd, tp->srtt_us, tp->rtt_min, tp->snd_ssthresh, 83 tp->rcv_nxt, tp->snd_nxt, tp->snd_una, tp->mss_cache, 84 tp->ecn_flags, tp->rate_delivered, tp->rate_interval_us, 85 tp->packets_out, tp->retrans_out, tp->total_retrans, 86 tp->segs_in, tp->data_segs_in, tp->segs_out, 87 tp->data_segs_out, tp->lost_out, tp->sacked_out, 88 tp->bytes_received, tp->bytes_acked); 89 } 90 91 static void check_result(void) 92 { 93 struct bpf_tcp_sock srv_tp, cli_tp, listen_tp; 94 struct bpf_sock srv_sk, cli_sk, listen_sk; 95 __u32 ingress_linum, egress_linum; 96 int err; 97 98 err = bpf_map_lookup_elem(linum_map_fd, &egress_linum_idx, 99 &egress_linum); 100 CHECK(err == -1, "bpf_map_lookup_elem(linum_map_fd)", 101 "err:%d errno:%d\n", err, errno); 102 103 err = bpf_map_lookup_elem(linum_map_fd, &ingress_linum_idx, 104 &ingress_linum); 105 CHECK(err == -1, "bpf_map_lookup_elem(linum_map_fd)", 106 "err:%d errno:%d\n", err, errno); 107 108 memcpy(&srv_sk, &skel->bss->srv_sk, sizeof(srv_sk)); 109 memcpy(&srv_tp, &skel->bss->srv_tp, sizeof(srv_tp)); 110 memcpy(&cli_sk, &skel->bss->cli_sk, sizeof(cli_sk)); 111 memcpy(&cli_tp, &skel->bss->cli_tp, sizeof(cli_tp)); 112 memcpy(&listen_sk, &skel->bss->listen_sk, sizeof(listen_sk)); 113 memcpy(&listen_tp, &skel->bss->listen_tp, sizeof(listen_tp)); 114 115 print_sk(&listen_sk, "listen_sk"); 116 print_sk(&srv_sk, "srv_sk"); 117 print_sk(&cli_sk, "cli_sk"); 118 print_tp(&listen_tp, "listen_tp"); 119 print_tp(&srv_tp, "srv_tp"); 120 print_tp(&cli_tp, "cli_tp"); 121 122 CHECK(listen_sk.state != 10 || 123 listen_sk.family != AF_INET6 || 124 listen_sk.protocol != IPPROTO_TCP || 125 memcmp(listen_sk.src_ip6, &in6addr_loopback, 126 sizeof(listen_sk.src_ip6)) || 127 listen_sk.dst_ip6[0] || listen_sk.dst_ip6[1] || 128 listen_sk.dst_ip6[2] || listen_sk.dst_ip6[3] || 129 listen_sk.src_port != ntohs(srv_sa6.sin6_port) || 130 listen_sk.dst_port, 131 "listen_sk", 132 "Unexpected. Check listen_sk output. ingress_linum:%u\n", 133 ingress_linum); 134 135 CHECK(srv_sk.state == 10 || 136 !srv_sk.state || 137 srv_sk.family != AF_INET6 || 138 srv_sk.protocol != IPPROTO_TCP || 139 memcmp(srv_sk.src_ip6, &in6addr_loopback, 140 sizeof(srv_sk.src_ip6)) || 141 memcmp(srv_sk.dst_ip6, &in6addr_loopback, 142 sizeof(srv_sk.dst_ip6)) || 143 srv_sk.src_port != ntohs(srv_sa6.sin6_port) || 144 srv_sk.dst_port != cli_sa6.sin6_port, 145 "srv_sk", "Unexpected. Check srv_sk output. egress_linum:%u\n", 146 egress_linum); 147 148 CHECK(!skel->bss->lsndtime, "srv_tp", "Unexpected lsndtime:0\n"); 149 150 CHECK(cli_sk.state == 10 || 151 !cli_sk.state || 152 cli_sk.family != AF_INET6 || 153 cli_sk.protocol != IPPROTO_TCP || 154 memcmp(cli_sk.src_ip6, &in6addr_loopback, 155 sizeof(cli_sk.src_ip6)) || 156 memcmp(cli_sk.dst_ip6, &in6addr_loopback, 157 sizeof(cli_sk.dst_ip6)) || 158 cli_sk.src_port != ntohs(cli_sa6.sin6_port) || 159 cli_sk.dst_port != srv_sa6.sin6_port, 160 "cli_sk", "Unexpected. Check cli_sk output. egress_linum:%u\n", 161 egress_linum); 162 163 CHECK(listen_tp.data_segs_out || 164 listen_tp.data_segs_in || 165 listen_tp.total_retrans || 166 listen_tp.bytes_acked, 167 "listen_tp", 168 "Unexpected. Check listen_tp output. ingress_linum:%u\n", 169 ingress_linum); 170 171 CHECK(srv_tp.data_segs_out != 2 || 172 srv_tp.data_segs_in || 173 srv_tp.snd_cwnd != 10 || 174 srv_tp.total_retrans || 175 srv_tp.bytes_acked < 2 * DATA_LEN, 176 "srv_tp", "Unexpected. Check srv_tp output. egress_linum:%u\n", 177 egress_linum); 178 179 CHECK(cli_tp.data_segs_out || 180 cli_tp.data_segs_in != 2 || 181 cli_tp.snd_cwnd != 10 || 182 cli_tp.total_retrans || 183 cli_tp.bytes_received < 2 * DATA_LEN, 184 "cli_tp", "Unexpected. Check cli_tp output. egress_linum:%u\n", 185 egress_linum); 186 187 CHECK(skel->bss->parent_cg_id != parent_cg_id, 188 "parent_cg_id", "%zu != %zu\n", 189 (size_t)skel->bss->parent_cg_id, (size_t)parent_cg_id); 190 191 CHECK(skel->bss->child_cg_id != child_cg_id, 192 "child_cg_id", "%zu != %zu\n", 193 (size_t)skel->bss->child_cg_id, (size_t)child_cg_id); 194 } 195 196 static void check_sk_pkt_out_cnt(int accept_fd, int cli_fd) 197 { 198 struct bpf_spinlock_cnt pkt_out_cnt = {}, pkt_out_cnt10 = {}; 199 int err; 200 201 pkt_out_cnt.cnt = ~0; 202 pkt_out_cnt10.cnt = ~0; 203 err = bpf_map_lookup_elem(sk_pkt_out_cnt_fd, &accept_fd, &pkt_out_cnt); 204 if (!err) 205 err = bpf_map_lookup_elem(sk_pkt_out_cnt10_fd, &accept_fd, 206 &pkt_out_cnt10); 207 208 /* The bpf prog only counts for fullsock and 209 * passive connection did not become fullsock until 3WHS 210 * had been finished, so the bpf prog only counted two data 211 * packet out. 212 */ 213 CHECK(err || pkt_out_cnt.cnt < 0xeB9F + 2 || 214 pkt_out_cnt10.cnt < 0xeB9F + 20, 215 "bpf_map_lookup_elem(sk_pkt_out_cnt, &accept_fd)", 216 "err:%d errno:%d pkt_out_cnt:%u pkt_out_cnt10:%u\n", 217 err, errno, pkt_out_cnt.cnt, pkt_out_cnt10.cnt); 218 219 pkt_out_cnt.cnt = ~0; 220 pkt_out_cnt10.cnt = ~0; 221 err = bpf_map_lookup_elem(sk_pkt_out_cnt_fd, &cli_fd, &pkt_out_cnt); 222 if (!err) 223 err = bpf_map_lookup_elem(sk_pkt_out_cnt10_fd, &cli_fd, 224 &pkt_out_cnt10); 225 /* Active connection is fullsock from the beginning. 226 * 1 SYN and 1 ACK during 3WHS 227 * 2 Acks on data packet. 228 * 229 * The bpf_prog initialized it to 0xeB9F. 230 */ 231 CHECK(err || pkt_out_cnt.cnt < 0xeB9F + 4 || 232 pkt_out_cnt10.cnt < 0xeB9F + 40, 233 "bpf_map_lookup_elem(sk_pkt_out_cnt, &cli_fd)", 234 "err:%d errno:%d pkt_out_cnt:%u pkt_out_cnt10:%u\n", 235 err, errno, pkt_out_cnt.cnt, pkt_out_cnt10.cnt); 236 } 237 238 static int init_sk_storage(int sk_fd, __u32 pkt_out_cnt) 239 { 240 struct bpf_spinlock_cnt scnt = {}; 241 int err; 242 243 scnt.cnt = pkt_out_cnt; 244 err = bpf_map_update_elem(sk_pkt_out_cnt_fd, &sk_fd, &scnt, 245 BPF_NOEXIST); 246 if (CHECK(err, "bpf_map_update_elem(sk_pkt_out_cnt_fd)", 247 "err:%d errno:%d\n", err, errno)) 248 return err; 249 250 err = bpf_map_update_elem(sk_pkt_out_cnt10_fd, &sk_fd, &scnt, 251 BPF_NOEXIST); 252 if (CHECK(err, "bpf_map_update_elem(sk_pkt_out_cnt10_fd)", 253 "err:%d errno:%d\n", err, errno)) 254 return err; 255 256 return 0; 257 } 258 259 static void test(void) 260 { 261 int listen_fd = -1, cli_fd = -1, accept_fd = -1, err, i; 262 socklen_t addrlen = sizeof(struct sockaddr_in6); 263 char buf[DATA_LEN]; 264 265 /* Prepare listen_fd */ 266 listen_fd = start_server(AF_INET6, SOCK_STREAM, "::1", 0, 0); 267 /* start_server() has logged the error details */ 268 if (CHECK_FAIL(listen_fd == -1)) 269 goto done; 270 271 err = getsockname(listen_fd, (struct sockaddr *)&srv_sa6, &addrlen); 272 if (CHECK(err, "getsockname(listen_fd)", "err:%d errno:%d\n", err, 273 errno)) 274 goto done; 275 memcpy(&skel->bss->srv_sa6, &srv_sa6, sizeof(srv_sa6)); 276 277 cli_fd = connect_to_fd(listen_fd, 0); 278 if (CHECK_FAIL(cli_fd == -1)) 279 goto done; 280 281 err = getsockname(cli_fd, (struct sockaddr *)&cli_sa6, &addrlen); 282 if (CHECK(err, "getsockname(cli_fd)", "err:%d errno:%d\n", 283 err, errno)) 284 goto done; 285 286 accept_fd = accept(listen_fd, NULL, NULL); 287 if (CHECK(accept_fd == -1, "accept(listen_fd)", 288 "accept_fd:%d errno:%d\n", 289 accept_fd, errno)) 290 goto done; 291 292 if (init_sk_storage(accept_fd, 0xeB9F)) 293 goto done; 294 295 for (i = 0; i < 2; i++) { 296 /* Send some data from accept_fd to cli_fd. 297 * MSG_EOR to stop kernel from coalescing two pkts. 298 */ 299 err = send(accept_fd, DATA, DATA_LEN, MSG_EOR); 300 if (CHECK(err != DATA_LEN, "send(accept_fd)", 301 "err:%d errno:%d\n", err, errno)) 302 goto done; 303 304 err = recv(cli_fd, buf, DATA_LEN, 0); 305 if (CHECK(err != DATA_LEN, "recv(cli_fd)", "err:%d errno:%d\n", 306 err, errno)) 307 goto done; 308 } 309 310 shutdown(cli_fd, SHUT_WR); 311 err = recv(accept_fd, buf, 1, 0); 312 if (CHECK(err, "recv(accept_fd) for fin", "err:%d errno:%d\n", 313 err, errno)) 314 goto done; 315 shutdown(accept_fd, SHUT_WR); 316 err = recv(cli_fd, buf, 1, 0); 317 if (CHECK(err, "recv(cli_fd) for fin", "err:%d errno:%d\n", 318 err, errno)) 319 goto done; 320 check_sk_pkt_out_cnt(accept_fd, cli_fd); 321 check_result(); 322 323 done: 324 if (accept_fd != -1) 325 close(accept_fd); 326 if (cli_fd != -1) 327 close(cli_fd); 328 if (listen_fd != -1) 329 close(listen_fd); 330 } 331 332 void test_sock_fields(void) 333 { 334 struct bpf_link *egress_link = NULL, *ingress_link = NULL; 335 int parent_cg_fd = -1, child_cg_fd = -1; 336 337 /* Create a cgroup, get fd, and join it */ 338 parent_cg_fd = test__join_cgroup(PARENT_CGROUP); 339 if (CHECK_FAIL(parent_cg_fd < 0)) 340 return; 341 parent_cg_id = get_cgroup_id(PARENT_CGROUP); 342 if (CHECK_FAIL(!parent_cg_id)) 343 goto done; 344 345 child_cg_fd = test__join_cgroup(CHILD_CGROUP); 346 if (CHECK_FAIL(child_cg_fd < 0)) 347 goto done; 348 child_cg_id = get_cgroup_id(CHILD_CGROUP); 349 if (CHECK_FAIL(!child_cg_id)) 350 goto done; 351 352 skel = test_sock_fields__open_and_load(); 353 if (CHECK(!skel, "test_sock_fields__open_and_load", "failed\n")) 354 goto done; 355 356 egress_link = bpf_program__attach_cgroup(skel->progs.egress_read_sock_fields, 357 child_cg_fd); 358 if (CHECK(IS_ERR(egress_link), "attach_cgroup(egress)", "err:%ld\n", 359 PTR_ERR(egress_link))) 360 goto done; 361 362 ingress_link = bpf_program__attach_cgroup(skel->progs.ingress_read_sock_fields, 363 child_cg_fd); 364 if (CHECK(IS_ERR(ingress_link), "attach_cgroup(ingress)", "err:%ld\n", 365 PTR_ERR(ingress_link))) 366 goto done; 367 368 linum_map_fd = bpf_map__fd(skel->maps.linum_map); 369 sk_pkt_out_cnt_fd = bpf_map__fd(skel->maps.sk_pkt_out_cnt); 370 sk_pkt_out_cnt10_fd = bpf_map__fd(skel->maps.sk_pkt_out_cnt10); 371 372 test(); 373 374 done: 375 bpf_link__destroy(egress_link); 376 bpf_link__destroy(ingress_link); 377 test_sock_fields__destroy(skel); 378 if (child_cg_fd != -1) 379 close(child_cg_fd); 380 if (parent_cg_fd != -1) 381 close(parent_cg_fd); 382 } 383