| 1 | |
| 2 | #define _GNU_SOURCE
|
| 3 |
|
| 4 | #include <errno.h>
|
| 5 | #include <stdbool.h>
|
| 6 | #include <stdio.h>
|
| 7 | #include <string.h>
|
| 8 | #include <unistd.h>
|
| 9 | #include <sched.h>
|
| 10 |
|
| 11 | #include <arpa/inet.h>
|
| 12 | #include <sys/mount.h>
|
| 13 | #include <sys/stat.h>
|
| 14 | #include <sys/types.h>
|
| 15 | #include <sys/un.h>
|
| 16 | #include <sys/eventfd.h>
|
| 17 |
|
| 18 | #include <linux/err.h>
|
| 19 | #include <linux/in.h>
|
| 20 | #include <linux/in6.h>
|
| 21 | #include <linux/limits.h>
|
| 22 |
|
| 23 | #include <linux/ip.h>
|
| 24 | #include <netinet/udp.h>
|
| 25 | #include <netinet/tcp.h>
|
| 26 | #include <net/if.h>
|
| 27 |
|
| 28 | #include "bpf_util.h"
|
| 29 | #include "network_helpers.h"
|
| 30 | #include "test_progs.h"
|
| 31 |
|
| 32 | #ifdef TRAFFIC_MONITOR
|
| 33 |
|
| 34 | #define PCAP_DONT_INCLUDE_PCAP_BPF_H 1
|
| 35 | #include <pcap/pcap.h>
|
| 36 | #include <pcap/dlt.h>
|
| 37 | #endif
|
| 38 |
|
| 39 | #ifndef IPPROTO_MPTCP
|
| 40 | #define IPPROTO_MPTCP 262
|
| 41 | #endif
|
| 42 |
|
| 43 | #define clean_errno() (errno == 0 ? "None" : strerror(errno))
|
| 44 | #define log_err(MSG, ...) ({ \
|
| 45 | int __save = errno; \
|
| 46 | fprintf(stderr, "(%s:%d: errno: %s) " MSG "\n", \
|
| 47 | __FILE__, __LINE__, clean_errno(), \
|
| 48 | ##__VA_ARGS__); \
|
| 49 | errno = __save; \
|
| 50 | })
|
| 51 |
|
| 52 | struct ipv4_packet pkt_v4 = {
|
| 53 | .eth.h_proto = __bpf_constant_htons(ETH_P_IP),
|
| 54 | .iph.ihl = 5,
|
| 55 | .iph.protocol = IPPROTO_TCP,
|
| 56 | .iph.tot_len = __bpf_constant_htons(MAGIC_BYTES),
|
| 57 | .tcp.urg_ptr = 123,
|
| 58 | .tcp.doff = 5,
|
| 59 | };
|
| 60 |
|
| 61 | struct ipv6_packet pkt_v6 = {
|
| 62 | .eth.h_proto = __bpf_constant_htons(ETH_P_IPV6),
|
| 63 | .iph.nexthdr = IPPROTO_TCP,
|
| 64 | .iph.payload_len = __bpf_constant_htons(MAGIC_BYTES),
|
| 65 | .tcp.urg_ptr = 123,
|
| 66 | .tcp.doff = 5,
|
| 67 | };
|
| 68 |
|
| 69 | static const struct network_helper_opts default_opts;
|
| 70 |
|
| 71 | int settimeo(int fd, int timeout_ms)
|
| 72 | {
|
| 73 | struct timeval timeout = { .tv_sec = 3 };
|
| 74 |
|
| 75 | if (timeout_ms > 0) {
|
| 76 | timeout.tv_sec = timeout_ms / 1000;
|
| 77 | timeout.tv_usec = (timeout_ms % 1000) * 1000;
|
| 78 | }
|
| 79 |
|
| 80 | if (setsockopt(fd, SOL_SOCKET, SO_RCVTIMEO, &timeout,
|
| 81 | sizeof(timeout))) {
|
| 82 | log_err("Failed to set SO_RCVTIMEO");
|
| 83 | return -1;
|
| 84 | }
|
| 85 |
|
| 86 | if (setsockopt(fd, SOL_SOCKET, SO_SNDTIMEO, &timeout,
|
| 87 | sizeof(timeout))) {
|
| 88 | log_err("Failed to set SO_SNDTIMEO");
|
| 89 | return -1;
|
| 90 | }
|
| 91 |
|
| 92 | return 0;
|
| 93 | }
|
| 94 |
|
| 95 | #define save_errno_close(fd) ({ int __save = errno; close(fd); errno = __save; })
|
| 96 |
|
| 97 | int start_server_addr(int type, const struct sockaddr_storage *addr, socklen_t addrlen,
|
| 98 | const struct network_helper_opts *opts)
|
| 99 | {
|
| 100 | int on = 1, fd;
|
| 101 |
|
| 102 | if (!opts)
|
| 103 | opts = &default_opts;
|
| 104 |
|
| 105 | fd = socket(addr->ss_family, type, opts->proto);
|
| 106 | if (fd < 0) {
|
| 107 | log_err("Failed to create server socket");
|
| 108 | return -1;
|
| 109 | }
|
| 110 |
|
| 111 | if (settimeo(fd, opts->timeout_ms))
|
| 112 | goto error_close;
|
| 113 |
|
| 114 | if ((type & SOCK_TYPE_MASK) == SOCK_STREAM &&
|
| 115 | setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on))) {
|
| 116 | log_err("Failed to enable SO_REUSEADDR");
|
| 117 | goto error_close;
|
| 118 | }
|
| 119 |
|
| 120 | if (opts->post_socket_cb &&
|
| 121 | opts->post_socket_cb(fd, opts->cb_opts)) {
|
| 122 | log_err("Failed to call post_socket_cb");
|
| 123 | goto error_close;
|
| 124 | }
|
| 125 |
|
| 126 | if (bind(fd, (struct sockaddr *)addr, addrlen) < 0) {
|
| 127 | log_err("Failed to bind socket");
|
| 128 | goto error_close;
|
| 129 | }
|
| 130 |
|
| 131 | if ((type & SOCK_TYPE_MASK) == SOCK_STREAM) {
|
| 132 | if (listen(fd, opts->backlog ? MAX(opts->backlog, 0) : 1) < 0) {
|
| 133 | log_err("Failed to listed on socket");
|
| 134 | goto error_close;
|
| 135 | }
|
| 136 | }
|
| 137 |
|
| 138 | return fd;
|
| 139 |
|
| 140 | error_close:
|
| 141 | save_errno_close(fd);
|
| 142 | return -1;
|
| 143 | }
|
| 144 |
|
| 145 | int start_server_str(int family, int type, const char *addr_str, __u16 port,
|
| 146 | const struct network_helper_opts *opts)
|
| 147 | {
|
| 148 | struct sockaddr_storage addr;
|
| 149 | socklen_t addrlen;
|
| 150 |
|
| 151 | if (!opts)
|
| 152 | opts = &default_opts;
|
| 153 |
|
| 154 | if (make_sockaddr(family, addr_str, port, &addr, &addrlen))
|
| 155 | return -1;
|
| 156 |
|
| 157 | return start_server_addr(type, &addr, addrlen, opts);
|
| 158 | }
|
| 159 |
|
| 160 | int start_server(int family, int type, const char *addr_str, __u16 port,
|
| 161 | int timeout_ms)
|
| 162 | {
|
| 163 | struct network_helper_opts opts = {
|
| 164 | .timeout_ms = timeout_ms,
|
| 165 | };
|
| 166 |
|
| 167 | return start_server_str(family, type, addr_str, port, &opts);
|
| 168 | }
|
| 169 |
|
| 170 | static int reuseport_cb(int fd, void *opts)
|
| 171 | {
|
| 172 | int on = 1;
|
| 173 |
|
| 174 | return setsockopt(fd, SOL_SOCKET, SO_REUSEPORT, &on, sizeof(on));
|
| 175 | }
|
| 176 |
|
| 177 | int *start_reuseport_server(int family, int type, const char *addr_str,
|
| 178 | __u16 port, int timeout_ms, unsigned int nr_listens)
|
| 179 | {
|
| 180 | struct network_helper_opts opts = {
|
| 181 | .timeout_ms = timeout_ms,
|
| 182 | .post_socket_cb = reuseport_cb,
|
| 183 | };
|
| 184 | struct sockaddr_storage addr;
|
| 185 | unsigned int nr_fds = 0;
|
| 186 | socklen_t addrlen;
|
| 187 | int *fds;
|
| 188 |
|
| 189 | if (!nr_listens)
|
| 190 | return NULL;
|
| 191 |
|
| 192 | if (make_sockaddr(family, addr_str, port, &addr, &addrlen))
|
| 193 | return NULL;
|
| 194 |
|
| 195 | fds = malloc(sizeof(*fds) * nr_listens);
|
| 196 | if (!fds)
|
| 197 | return NULL;
|
| 198 |
|
| 199 | fds[0] = start_server_addr(type, &addr, addrlen, &opts);
|
| 200 | if (fds[0] == -1)
|
| 201 | goto close_fds;
|
| 202 | nr_fds = 1;
|
| 203 |
|
| 204 | if (getsockname(fds[0], (struct sockaddr *)&addr, &addrlen))
|
| 205 | goto close_fds;
|
| 206 |
|
| 207 | for (; nr_fds < nr_listens; nr_fds++) {
|
| 208 | fds[nr_fds] = start_server_addr(type, &addr, addrlen, &opts);
|
| 209 | if (fds[nr_fds] == -1)
|
| 210 | goto close_fds;
|
| 211 | }
|
| 212 |
|
| 213 | return fds;
|
| 214 |
|
| 215 | close_fds:
|
| 216 | free_fds(fds, nr_fds);
|
| 217 | return NULL;
|
| 218 | }
|
| 219 |
|
| 220 | void free_fds(int *fds, unsigned int nr_close_fds)
|
| 221 | {
|
| 222 | if (fds) {
|
| 223 | while (nr_close_fds)
|
| 224 | close(fds[--nr_close_fds]);
|
| 225 | free(fds);
|
| 226 | }
|
| 227 | }
|
| 228 |
|
| 229 | int fastopen_connect(int server_fd, const char *data, unsigned int data_len,
|
| 230 | int timeout_ms)
|
| 231 | {
|
| 232 | struct sockaddr_storage addr;
|
| 233 | socklen_t addrlen = sizeof(addr);
|
| 234 | struct sockaddr_in *addr_in;
|
| 235 | int fd, ret;
|
| 236 |
|
| 237 | if (getsockname(server_fd, (struct sockaddr *)&addr, &addrlen)) {
|
| 238 | log_err("Failed to get server addr");
|
| 239 | return -1;
|
| 240 | }
|
| 241 |
|
| 242 | addr_in = (struct sockaddr_in *)&addr;
|
| 243 | fd = socket(addr_in->sin_family, SOCK_STREAM, 0);
|
| 244 | if (fd < 0) {
|
| 245 | log_err("Failed to create client socket");
|
| 246 | return -1;
|
| 247 | }
|
| 248 |
|
| 249 | if (settimeo(fd, timeout_ms))
|
| 250 | goto error_close;
|
| 251 |
|
| 252 | ret = sendto(fd, data, data_len, MSG_FASTOPEN, (struct sockaddr *)&addr,
|
| 253 | addrlen);
|
| 254 | if (ret != data_len) {
|
| 255 | log_err("sendto(data, %u) != %d\n", data_len, ret);
|
| 256 | goto error_close;
|
| 257 | }
|
| 258 |
|
| 259 | return fd;
|
| 260 |
|
| 261 | error_close:
|
| 262 | save_errno_close(fd);
|
| 263 | return -1;
|
| 264 | }
|
| 265 |
|
| 266 | int client_socket(int family, int type,
|
| 267 | const struct network_helper_opts *opts)
|
| 268 | {
|
| 269 | int fd;
|
| 270 |
|
| 271 | if (!opts)
|
| 272 | opts = &default_opts;
|
| 273 |
|
| 274 | fd = socket(family, type, opts->proto);
|
| 275 | if (fd < 0) {
|
| 276 | log_err("Failed to create client socket");
|
| 277 | return -1;
|
| 278 | }
|
| 279 |
|
| 280 | if (settimeo(fd, opts->timeout_ms))
|
| 281 | goto error_close;
|
| 282 |
|
| 283 | if (opts->post_socket_cb &&
|
| 284 | opts->post_socket_cb(fd, opts->cb_opts))
|
| 285 | goto error_close;
|
| 286 |
|
| 287 | return fd;
|
| 288 |
|
| 289 | error_close:
|
| 290 | save_errno_close(fd);
|
| 291 | return -1;
|
| 292 | }
|
| 293 |
|
| 294 | int connect_to_addr(int type, const struct sockaddr_storage *addr, socklen_t addrlen,
|
| 295 | const struct network_helper_opts *opts)
|
| 296 | {
|
| 297 | int fd;
|
| 298 |
|
| 299 | if (!opts)
|
| 300 | opts = &default_opts;
|
| 301 |
|
| 302 | fd = client_socket(addr->ss_family, type, opts);
|
| 303 | if (fd < 0) {
|
| 304 | log_err("Failed to create client socket");
|
| 305 | return -1;
|
| 306 | }
|
| 307 |
|
| 308 | if (connect(fd, (const struct sockaddr *)addr, addrlen)) {
|
| 309 | log_err("Failed to connect to server");
|
| 310 | save_errno_close(fd);
|
| 311 | return -1;
|
| 312 | }
|
| 313 |
|
| 314 | return fd;
|
| 315 | }
|
| 316 |
|
| 317 | int connect_to_addr_str(int family, int type, const char *addr_str, __u16 port,
|
| 318 | const struct network_helper_opts *opts)
|
| 319 | {
|
| 320 | struct sockaddr_storage addr;
|
| 321 | socklen_t addrlen;
|
| 322 |
|
| 323 | if (!opts)
|
| 324 | opts = &default_opts;
|
| 325 |
|
| 326 | if (make_sockaddr(family, addr_str, port, &addr, &addrlen))
|
| 327 | return -1;
|
| 328 |
|
| 329 | return connect_to_addr(type, &addr, addrlen, opts);
|
| 330 | }
|
| 331 |
|
| 332 | int connect_to_fd_opts(int server_fd, const struct network_helper_opts *opts)
|
| 333 | {
|
| 334 | struct sockaddr_storage addr;
|
| 335 | socklen_t addrlen, optlen;
|
| 336 | int type;
|
| 337 |
|
| 338 | if (!opts)
|
| 339 | opts = &default_opts;
|
| 340 |
|
| 341 | optlen = sizeof(type);
|
| 342 | if (getsockopt(server_fd, SOL_SOCKET, SO_TYPE, &type, &optlen)) {
|
| 343 | log_err("getsockopt(SOL_TYPE)");
|
| 344 | return -1;
|
| 345 | }
|
| 346 |
|
| 347 | addrlen = sizeof(addr);
|
| 348 | if (getsockname(server_fd, (struct sockaddr *)&addr, &addrlen)) {
|
| 349 | log_err("Failed to get server addr");
|
| 350 | return -1;
|
| 351 | }
|
| 352 |
|
| 353 | return connect_to_addr(type, &addr, addrlen, opts);
|
| 354 | }
|
| 355 |
|
| 356 | int connect_to_fd(int server_fd, int timeout_ms)
|
| 357 | {
|
| 358 | struct network_helper_opts opts = {
|
| 359 | .timeout_ms = timeout_ms,
|
| 360 | };
|
| 361 | socklen_t optlen;
|
| 362 | int protocol;
|
| 363 |
|
| 364 | optlen = sizeof(protocol);
|
| 365 | if (getsockopt(server_fd, SOL_SOCKET, SO_PROTOCOL, &protocol, &optlen)) {
|
| 366 | log_err("getsockopt(SOL_PROTOCOL)");
|
| 367 | return -1;
|
| 368 | }
|
| 369 | opts.proto = protocol;
|
| 370 |
|
| 371 | return connect_to_fd_opts(server_fd, &opts);
|
| 372 | }
|
| 373 |
|
| 374 | int connect_fd_to_fd(int client_fd, int server_fd, int timeout_ms)
|
| 375 | {
|
| 376 | struct sockaddr_storage addr;
|
| 377 | socklen_t len = sizeof(addr);
|
| 378 |
|
| 379 | if (settimeo(client_fd, timeout_ms))
|
| 380 | return -1;
|
| 381 |
|
| 382 | if (getsockname(server_fd, (struct sockaddr *)&addr, &len)) {
|
| 383 | log_err("Failed to get server addr");
|
| 384 | return -1;
|
| 385 | }
|
| 386 |
|
| 387 | if (connect(client_fd, (const struct sockaddr *)&addr, len)) {
|
| 388 | log_err("Failed to connect to server");
|
| 389 | return -1;
|
| 390 | }
|
| 391 |
|
| 392 | return 0;
|
| 393 | }
|
| 394 |
|
| 395 | int make_sockaddr(int family, const char *addr_str, __u16 port,
|
| 396 | struct sockaddr_storage *addr, socklen_t *len)
|
| 397 | {
|
| 398 | if (family == AF_INET) {
|
| 399 | struct sockaddr_in *sin = (void *)addr;
|
| 400 |
|
| 401 | memset(addr, 0, sizeof(*sin));
|
| 402 | sin->sin_family = AF_INET;
|
| 403 | sin->sin_port = htons(port);
|
| 404 | if (addr_str &&
|
| 405 | inet_pton(AF_INET, addr_str, &sin->sin_addr) != 1) {
|
| 406 | log_err("inet_pton(AF_INET, %s)", addr_str);
|
| 407 | return -1;
|
| 408 | }
|
| 409 | if (len)
|
| 410 | *len = sizeof(*sin);
|
| 411 | return 0;
|
| 412 | } else if (family == AF_INET6) {
|
| 413 | struct sockaddr_in6 *sin6 = (void *)addr;
|
| 414 |
|
| 415 | memset(addr, 0, sizeof(*sin6));
|
| 416 | sin6->sin6_family = AF_INET6;
|
| 417 | sin6->sin6_port = htons(port);
|
| 418 | if (addr_str &&
|
| 419 | inet_pton(AF_INET6, addr_str, &sin6->sin6_addr) != 1) {
|
| 420 | log_err("inet_pton(AF_INET6, %s)", addr_str);
|
| 421 | return -1;
|
| 422 | }
|
| 423 | if (len)
|
| 424 | *len = sizeof(*sin6);
|
| 425 | return 0;
|
| 426 | } else if (family == AF_UNIX) {
|
| 427 | |
| 428 | * Note that we always use abstract unix sockets to avoid having
|
| 429 | * to clean up leftover files.
|
| 430 |
|
| 431 | struct sockaddr_un *sun = (void *)addr;
|
| 432 |
|
| 433 | memset(addr, 0, sizeof(*sun));
|
| 434 | sun->sun_family = family;
|
| 435 | sun->sun_path[0] = 0;
|
| 436 | strscpy(sun->sun_path + 1, addr_str, sizeof(sun->sun_path) - 1);
|
| 437 | if (len)
|
| 438 | *len = offsetof(struct sockaddr_un, sun_path) + 1 + strlen(addr_str);
|
| 439 | return 0;
|
| 440 | }
|
| 441 | return -1;
|
| 442 | }
|
| 443 |
|
| 444 | char *ping_command(int family)
|
| 445 | {
|
| 446 | if (family == AF_INET6) {
|
| 447 |
|
| 448 | if (!system("which ping6 >/dev/null 2>&1"))
|
| 449 | return "ping6";
|
| 450 | else
|
| 451 | return "ping -6";
|
| 452 | }
|
| 453 | return "ping";
|
| 454 | }
|
| 455 |
|
| 456 | int append_tid(char *str, size_t sz)
|
| 457 | {
|
| 458 | size_t end;
|
| 459 |
|
| 460 | if (!str)
|
| 461 | return -1;
|
| 462 |
|
| 463 | end = strlen(str);
|
| 464 | if (end + 8 > sz)
|
| 465 | return -1;
|
| 466 |
|
| 467 | sprintf(&str[end], "%07ld", sys_gettid());
|
| 468 | str[end + 7] = '\0';
|
| 469 |
|
| 470 | return 0;
|
| 471 | }
|
| 472 |
|
| 473 | int remove_netns(const char *name)
|
| 474 | {
|
| 475 | char *cmd;
|
| 476 | int r;
|
| 477 |
|
| 478 | r = asprintf(&cmd, "ip netns del %s >/dev/null 2>&1", name);
|
| 479 | if (r < 0) {
|
| 480 | log_err("Failed to malloc cmd");
|
| 481 | return -1;
|
| 482 | }
|
| 483 |
|
| 484 | r = system(cmd);
|
| 485 | free(cmd);
|
| 486 | return r;
|
| 487 | }
|
| 488 |
|
| 489 | int make_netns(const char *name)
|
| 490 | {
|
| 491 | char *cmd;
|
| 492 | int r;
|
| 493 |
|
| 494 | r = asprintf(&cmd, "ip netns add %s", name);
|
| 495 | if (r < 0) {
|
| 496 | log_err("Failed to malloc cmd");
|
| 497 | return -1;
|
| 498 | }
|
| 499 |
|
| 500 | r = system(cmd);
|
| 501 | free(cmd);
|
| 502 |
|
| 503 | if (r)
|
| 504 | return r;
|
| 505 |
|
| 506 | r = asprintf(&cmd, "ip -n %s link set lo up", name);
|
| 507 | if (r < 0) {
|
| 508 | log_err("Failed to malloc cmd for setting up lo");
|
| 509 | remove_netns(name);
|
| 510 | return -1;
|
| 511 | }
|
| 512 |
|
| 513 | r = system(cmd);
|
| 514 | free(cmd);
|
| 515 |
|
| 516 | return r;
|
| 517 | }
|
| 518 |
|
| 519 | struct nstoken {
|
| 520 | int orig_netns_fd;
|
| 521 | };
|
| 522 |
|
| 523 | struct nstoken *open_netns(const char *name)
|
| 524 | {
|
| 525 | int nsfd;
|
| 526 | char nspath[PATH_MAX];
|
| 527 | int err;
|
| 528 | struct nstoken *token;
|
| 529 |
|
| 530 | token = calloc(1, sizeof(struct nstoken));
|
| 531 | if (!token) {
|
| 532 | log_err("Failed to malloc token");
|
| 533 | return NULL;
|
| 534 | }
|
| 535 |
|
| 536 | token->orig_netns_fd = open("/proc/self/ns/net", O_RDONLY);
|
| 537 | if (token->orig_netns_fd == -1) {
|
| 538 | log_err("Failed to open(/proc/self/ns/net)");
|
| 539 | goto fail;
|
| 540 | }
|
| 541 |
|
| 542 | snprintf(nspath, sizeof(nspath), "%s/%s", "/var/run/netns", name);
|
| 543 | nsfd = open(nspath, O_RDONLY | O_CLOEXEC);
|
| 544 | if (nsfd == -1) {
|
| 545 | log_err("Failed to open(%s)", nspath);
|
| 546 | goto fail;
|
| 547 | }
|
| 548 |
|
| 549 | err = setns(nsfd, CLONE_NEWNET);
|
| 550 | close(nsfd);
|
| 551 | if (err) {
|
| 552 | log_err("Failed to setns(nsfd)");
|
| 553 | goto fail;
|
| 554 | }
|
| 555 |
|
| 556 | return token;
|
| 557 | fail:
|
| 558 | if (token->orig_netns_fd != -1)
|
| 559 | close(token->orig_netns_fd);
|
| 560 | free(token);
|
| 561 | return NULL;
|
| 562 | }
|
| 563 |
|
| 564 | void close_netns(struct nstoken *token)
|
| 565 | {
|
| 566 | if (!token)
|
| 567 | return;
|
| 568 |
|
| 569 | if (setns(token->orig_netns_fd, CLONE_NEWNET))
|
| 570 | log_err("Failed to setns(orig_netns_fd)");
|
| 571 | close(token->orig_netns_fd);
|
| 572 | free(token);
|
| 573 | }
|
| 574 |
|
| 575 | int open_tuntap(const char *dev_name, bool need_mac)
|
| 576 | {
|
| 577 | int err = 0;
|
| 578 | struct ifreq ifr;
|
| 579 | int fd = open("/dev/net/tun", O_RDWR);
|
| 580 |
|
| 581 | if (!ASSERT_GE(fd, 0, "open(/dev/net/tun)"))
|
| 582 | return -1;
|
| 583 |
|
| 584 | ifr.ifr_flags = IFF_NO_PI | (need_mac ? IFF_TAP : IFF_TUN);
|
| 585 | strscpy(ifr.ifr_name, dev_name);
|
| 586 |
|
| 587 | err = ioctl(fd, TUNSETIFF, &ifr);
|
| 588 | if (!ASSERT_OK(err, "ioctl(TUNSETIFF)")) {
|
| 589 | close(fd);
|
| 590 | return -1;
|
| 591 | }
|
| 592 |
|
| 593 | err = fcntl(fd, F_SETFL, O_NONBLOCK);
|
| 594 | if (!ASSERT_OK(err, "fcntl(O_NONBLOCK)")) {
|
| 595 | close(fd);
|
| 596 | return -1;
|
| 597 | }
|
| 598 |
|
| 599 | return fd;
|
| 600 | }
|
| 601 |
|
| 602 | int get_socket_local_port(int sock_fd)
|
| 603 | {
|
| 604 | struct sockaddr_storage addr;
|
| 605 | socklen_t addrlen = sizeof(addr);
|
| 606 | int err;
|
| 607 |
|
| 608 | err = getsockname(sock_fd, (struct sockaddr *)&addr, &addrlen);
|
| 609 | if (err < 0)
|
| 610 | return err;
|
| 611 |
|
| 612 | if (addr.ss_family == AF_INET) {
|
| 613 | struct sockaddr_in *sin = (struct sockaddr_in *)&addr;
|
| 614 |
|
| 615 | return sin->sin_port;
|
| 616 | } else if (addr.ss_family == AF_INET6) {
|
| 617 | struct sockaddr_in6 *sin = (struct sockaddr_in6 *)&addr;
|
| 618 |
|
| 619 | return sin->sin6_port;
|
| 620 | }
|
| 621 |
|
| 622 | return -1;
|
| 623 | }
|
| 624 |
|
| 625 | int get_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param)
|
| 626 | {
|
| 627 | struct ifreq ifr = {0};
|
| 628 | int sockfd, err;
|
| 629 |
|
| 630 | sockfd = socket(AF_INET, SOCK_DGRAM, 0);
|
| 631 | if (sockfd < 0)
|
| 632 | return -errno;
|
| 633 |
|
| 634 | memcpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
|
| 635 |
|
| 636 | ring_param->cmd = ETHTOOL_GRINGPARAM;
|
| 637 | ifr.ifr_data = (char *)ring_param;
|
| 638 |
|
| 639 | if (ioctl(sockfd, SIOCETHTOOL, &ifr) < 0) {
|
| 640 | err = errno;
|
| 641 | close(sockfd);
|
| 642 | return -err;
|
| 643 | }
|
| 644 |
|
| 645 | close(sockfd);
|
| 646 | return 0;
|
| 647 | }
|
| 648 |
|
| 649 | int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param)
|
| 650 | {
|
| 651 | struct ifreq ifr = {0};
|
| 652 | int sockfd, err;
|
| 653 |
|
| 654 | sockfd = socket(AF_INET, SOCK_DGRAM, 0);
|
| 655 | if (sockfd < 0)
|
| 656 | return -errno;
|
| 657 |
|
| 658 | memcpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
|
| 659 |
|
| 660 | ring_param->cmd = ETHTOOL_SRINGPARAM;
|
| 661 | ifr.ifr_data = (char *)ring_param;
|
| 662 |
|
| 663 | if (ioctl(sockfd, SIOCETHTOOL, &ifr) < 0) {
|
| 664 | err = errno;
|
| 665 | close(sockfd);
|
| 666 | return -err;
|
| 667 | }
|
| 668 |
|
| 669 | close(sockfd);
|
| 670 | return 0;
|
| 671 | }
|
| 672 |
|
| 673 | struct send_recv_arg {
|
| 674 | int fd;
|
| 675 | uint32_t bytes;
|
| 676 | int stop;
|
| 677 | };
|
| 678 |
|
| 679 | static void *send_recv_server(void *arg)
|
| 680 | {
|
| 681 | struct send_recv_arg *a = (struct send_recv_arg *)arg;
|
| 682 | ssize_t nr_sent = 0, bytes = 0;
|
| 683 | char batch[1500];
|
| 684 | int err = 0, fd;
|
| 685 |
|
| 686 | fd = accept(a->fd, NULL, NULL);
|
| 687 | while (fd == -1) {
|
| 688 | if (errno == EINTR)
|
| 689 | continue;
|
| 690 | err = -errno;
|
| 691 | goto done;
|
| 692 | }
|
| 693 |
|
| 694 | if (settimeo(fd, 0)) {
|
| 695 | err = -errno;
|
| 696 | goto done;
|
| 697 | }
|
| 698 |
|
| 699 | while (bytes < a->bytes && !READ_ONCE(a->stop)) {
|
| 700 | nr_sent = send(fd, &batch,
|
| 701 | MIN(a->bytes - bytes, sizeof(batch)), 0);
|
| 702 | if (nr_sent == -1 && errno == EINTR)
|
| 703 | continue;
|
| 704 | if (nr_sent == -1) {
|
| 705 | err = -errno;
|
| 706 | break;
|
| 707 | }
|
| 708 | bytes += nr_sent;
|
| 709 | }
|
| 710 |
|
| 711 | if (bytes != a->bytes) {
|
| 712 | log_err("send %zd expected %u", bytes, a->bytes);
|
| 713 | if (!err)
|
| 714 | err = bytes > a->bytes ? -E2BIG : -EINTR;
|
| 715 | }
|
| 716 |
|
| 717 | done:
|
| 718 | if (fd >= 0)
|
| 719 | close(fd);
|
| 720 | if (err) {
|
| 721 | WRITE_ONCE(a->stop, 1);
|
| 722 | return ERR_PTR(err);
|
| 723 | }
|
| 724 | return NULL;
|
| 725 | }
|
| 726 |
|
| 727 | int send_recv_data(int lfd, int fd, uint32_t total_bytes)
|
| 728 | {
|
| 729 | ssize_t nr_recv = 0, bytes = 0;
|
| 730 | struct send_recv_arg arg = {
|
| 731 | .fd = lfd,
|
| 732 | .bytes = total_bytes,
|
| 733 | .stop = 0,
|
| 734 | };
|
| 735 | pthread_t srv_thread;
|
| 736 | void *thread_ret;
|
| 737 | char batch[1500];
|
| 738 | int err = 0;
|
| 739 |
|
| 740 | err = pthread_create(&srv_thread, NULL, send_recv_server, (void *)&arg);
|
| 741 | if (err) {
|
| 742 | log_err("Failed to pthread_create");
|
| 743 | return err;
|
| 744 | }
|
| 745 |
|
| 746 |
|
| 747 | while (bytes < total_bytes && !READ_ONCE(arg.stop)) {
|
| 748 | nr_recv = recv(fd, &batch,
|
| 749 | MIN(total_bytes - bytes, sizeof(batch)), 0);
|
| 750 | if (nr_recv == -1 && errno == EINTR)
|
| 751 | continue;
|
| 752 | if (nr_recv == -1) {
|
| 753 | err = -errno;
|
| 754 | break;
|
| 755 | }
|
| 756 | bytes += nr_recv;
|
| 757 | }
|
| 758 |
|
| 759 | if (bytes != total_bytes) {
|
| 760 | log_err("recv %zd expected %u", bytes, total_bytes);
|
| 761 | if (!err)
|
| 762 | err = bytes > total_bytes ? -E2BIG : -EINTR;
|
| 763 | }
|
| 764 |
|
| 765 | WRITE_ONCE(arg.stop, 1);
|
| 766 | pthread_join(srv_thread, &thread_ret);
|
| 767 | if (IS_ERR(thread_ret)) {
|
| 768 | log_err("Failed in thread_ret %ld", PTR_ERR(thread_ret));
|
| 769 | err = err ? : PTR_ERR(thread_ret);
|
| 770 | }
|
| 771 |
|
| 772 | return err;
|
| 773 | }
|
| 774 |
|
| 775 | int tc_prog_attach(const char *dev, int ingress_fd, int egress_fd)
|
| 776 | {
|
| 777 | int ifindex, ret;
|
| 778 |
|
| 779 | if (!ASSERT_TRUE(ingress_fd >= 0 || egress_fd >= 0,
|
| 780 | "at least one program fd is valid"))
|
| 781 | return -1;
|
| 782 |
|
| 783 | ifindex = if_nametoindex(dev);
|
| 784 | if (!ASSERT_NEQ(ifindex, 0, "get ifindex"))
|
| 785 | return -1;
|
| 786 |
|
| 787 | DECLARE_LIBBPF_OPTS(bpf_tc_hook, hook, .ifindex = ifindex,
|
| 788 | .attach_point = BPF_TC_INGRESS | BPF_TC_EGRESS);
|
| 789 | DECLARE_LIBBPF_OPTS(bpf_tc_opts, opts1, .handle = 1,
|
| 790 | .priority = 1, .prog_fd = ingress_fd);
|
| 791 | DECLARE_LIBBPF_OPTS(bpf_tc_opts, opts2, .handle = 1,
|
| 792 | .priority = 1, .prog_fd = egress_fd);
|
| 793 |
|
| 794 | ret = bpf_tc_hook_create(&hook);
|
| 795 | if (!ASSERT_OK(ret, "create tc hook"))
|
| 796 | return ret;
|
| 797 |
|
| 798 | if (ingress_fd >= 0) {
|
| 799 | hook.attach_point = BPF_TC_INGRESS;
|
| 800 | ret = bpf_tc_attach(&hook, &opts1);
|
| 801 | if (!ASSERT_OK(ret, "bpf_tc_attach")) {
|
| 802 | bpf_tc_hook_destroy(&hook);
|
| 803 | return ret;
|
| 804 | }
|
| 805 | }
|
| 806 |
|
| 807 | if (egress_fd >= 0) {
|
| 808 | hook.attach_point = BPF_TC_EGRESS;
|
| 809 | ret = bpf_tc_attach(&hook, &opts2);
|
| 810 | if (!ASSERT_OK(ret, "bpf_tc_attach")) {
|
| 811 | bpf_tc_hook_destroy(&hook);
|
| 812 | return ret;
|
| 813 | }
|
| 814 | }
|
| 815 |
|
| 816 | return 0;
|
| 817 | }
|
| 818 |
|
| 819 | #ifdef TRAFFIC_MONITOR
|
| 820 | struct tmonitor_ctx {
|
| 821 | pcap_t *pcap;
|
| 822 | pcap_dumper_t *dumper;
|
| 823 | pthread_t thread;
|
| 824 | int wake_fd;
|
| 825 |
|
| 826 | volatile bool done;
|
| 827 | char pkt_fname[PATH_MAX];
|
| 828 | int pcap_fd;
|
| 829 | };
|
| 830 |
|
| 831 | static int __base_pr(const char *format, va_list args)
|
| 832 | {
|
| 833 | return vfprintf(stdout, format, args);
|
| 834 | }
|
| 835 |
|
| 836 | static tm_print_fn_t __tm_pr = __base_pr;
|
| 837 |
|
| 838 | tm_print_fn_t traffic_monitor_set_print(tm_print_fn_t fn)
|
| 839 | {
|
| 840 | tm_print_fn_t old_print_fn;
|
| 841 |
|
| 842 | old_print_fn = __atomic_exchange_n(&__tm_pr, fn, __ATOMIC_RELAXED);
|
| 843 |
|
| 844 | return old_print_fn;
|
| 845 | }
|
| 846 |
|
| 847 | void tm_print(const char *format, ...)
|
| 848 | {
|
| 849 | tm_print_fn_t print_fn;
|
| 850 | va_list args;
|
| 851 |
|
| 852 | print_fn = __atomic_load_n(&__tm_pr, __ATOMIC_RELAXED);
|
| 853 | if (!print_fn)
|
| 854 | return;
|
| 855 |
|
| 856 | va_start(args, format);
|
| 857 | print_fn(format, args);
|
| 858 | va_end(args);
|
| 859 | }
|
| 860 |
|
| 861 |
|
| 862 | static bool is_ethernet(const u_char *packet)
|
| 863 | {
|
| 864 | u16 arphdr_type;
|
| 865 |
|
| 866 | memcpy(&arphdr_type, packet + 8, 2);
|
| 867 | arphdr_type = ntohs(arphdr_type);
|
| 868 |
|
| 869 | |
| 870 | * Except the following cases, the protocol type contains the
|
| 871 | * Ethernet protocol type for the packet.
|
| 872 | *
|
| 873 | * https://www.tcpdump.org/linktypes/LINKTYPE_LINUX_SLL2.html
|
| 874 |
|
| 875 | switch (arphdr_type) {
|
| 876 | case 770:
|
| 877 | case 778:
|
| 878 | case 803:
|
| 879 | tm_print("Packet captured: arphdr_type=%d\n", arphdr_type);
|
| 880 | return false;
|
| 881 | }
|
| 882 | return true;
|
| 883 | }
|
| 884 |
|
| 885 | static const char * const pkt_types[] = {
|
| 886 | "In",
|
| 887 | "B",
|
| 888 | "M",
|
| 889 | "C",
|
| 890 | "Out",
|
| 891 | };
|
| 892 |
|
| 893 | static const char *pkt_type_str(u16 pkt_type)
|
| 894 | {
|
| 895 | if (pkt_type < ARRAY_SIZE(pkt_types))
|
| 896 | return pkt_types[pkt_type];
|
| 897 | return "Unknown";
|
| 898 | }
|
| 899 |
|
| 900 | #define MAX_FLAGS_STRLEN 21
|
| 901 |
|
| 902 | static void show_transport(const u_char *packet, u16 len, u32 ifindex,
|
| 903 | const char *src_addr, const char *dst_addr,
|
| 904 | u16 proto, bool ipv6, u8 pkt_type)
|
| 905 | {
|
| 906 | char *ifname, _ifname[IF_NAMESIZE], flags[MAX_FLAGS_STRLEN] = "";
|
| 907 | const char *transport_str;
|
| 908 | u16 src_port, dst_port;
|
| 909 | struct udphdr *udp;
|
| 910 | struct tcphdr *tcp;
|
| 911 |
|
| 912 | ifname = if_indextoname(ifindex, _ifname);
|
| 913 | if (!ifname) {
|
| 914 | snprintf(_ifname, sizeof(_ifname), "unknown(%d)", ifindex);
|
| 915 | ifname = _ifname;
|
| 916 | }
|
| 917 |
|
| 918 | if (proto == IPPROTO_UDP) {
|
| 919 | udp = (struct udphdr *)packet;
|
| 920 | src_port = ntohs(udp->source);
|
| 921 | dst_port = ntohs(udp->dest);
|
| 922 | transport_str = "UDP";
|
| 923 | } else if (proto == IPPROTO_TCP) {
|
| 924 | tcp = (struct tcphdr *)packet;
|
| 925 | src_port = ntohs(tcp->source);
|
| 926 | dst_port = ntohs(tcp->dest);
|
| 927 | transport_str = "TCP";
|
| 928 | } else if (proto == IPPROTO_ICMP) {
|
| 929 | tm_print("%-7s %-3s IPv4 %s > %s: ICMP, length %d, type %d, code %d\n",
|
| 930 | ifname, pkt_type_str(pkt_type), src_addr, dst_addr, len,
|
| 931 | packet[0], packet[1]);
|
| 932 | return;
|
| 933 | } else if (proto == IPPROTO_ICMPV6) {
|
| 934 | tm_print("%-7s %-3s IPv6 %s > %s: ICMPv6, length %d, type %d, code %d\n",
|
| 935 | ifname, pkt_type_str(pkt_type), src_addr, dst_addr, len,
|
| 936 | packet[0], packet[1]);
|
| 937 | return;
|
| 938 | } else {
|
| 939 | tm_print("%-7s %-3s %s %s > %s: protocol %d\n",
|
| 940 | ifname, pkt_type_str(pkt_type), ipv6 ? "IPv6" : "IPv4",
|
| 941 | src_addr, dst_addr, proto);
|
| 942 | return;
|
| 943 | }
|
| 944 |
|
| 945 |
|
| 946 |
|
| 947 | if (proto == IPPROTO_TCP)
|
| 948 | snprintf(flags, MAX_FLAGS_STRLEN, "%s%s%s%s",
|
| 949 | tcp->fin ? ", FIN" : "",
|
| 950 | tcp->syn ? ", SYN" : "",
|
| 951 | tcp->rst ? ", RST" : "",
|
| 952 | tcp->ack ? ", ACK" : "");
|
| 953 |
|
| 954 | if (ipv6)
|
| 955 | tm_print("%-7s %-3s IPv6 %s.%d > %s.%d: %s, length %d%s\n",
|
| 956 | ifname, pkt_type_str(pkt_type), src_addr, src_port,
|
| 957 | dst_addr, dst_port, transport_str, len, flags);
|
| 958 | else
|
| 959 | tm_print("%-7s %-3s IPv4 %s:%d > %s:%d: %s, length %d%s\n",
|
| 960 | ifname, pkt_type_str(pkt_type), src_addr, src_port,
|
| 961 | dst_addr, dst_port, transport_str, len, flags);
|
| 962 | }
|
| 963 |
|
| 964 | static void show_ipv6_packet(const u_char *packet, u32 ifindex, u8 pkt_type)
|
| 965 | {
|
| 966 | char src_buf[INET6_ADDRSTRLEN], dst_buf[INET6_ADDRSTRLEN];
|
| 967 | struct ipv6hdr *pkt = (struct ipv6hdr *)packet;
|
| 968 | const char *src, *dst;
|
| 969 | u_char proto;
|
| 970 |
|
| 971 | src = inet_ntop(AF_INET6, &pkt->saddr, src_buf, sizeof(src_buf));
|
| 972 | if (!src)
|
| 973 | src = "<invalid>";
|
| 974 | dst = inet_ntop(AF_INET6, &pkt->daddr, dst_buf, sizeof(dst_buf));
|
| 975 | if (!dst)
|
| 976 | dst = "<invalid>";
|
| 977 | proto = pkt->nexthdr;
|
| 978 | show_transport(packet + sizeof(struct ipv6hdr),
|
| 979 | ntohs(pkt->payload_len),
|
| 980 | ifindex, src, dst, proto, true, pkt_type);
|
| 981 | }
|
| 982 |
|
| 983 | static void show_ipv4_packet(const u_char *packet, u32 ifindex, u8 pkt_type)
|
| 984 | {
|
| 985 | char src_buf[INET_ADDRSTRLEN], dst_buf[INET_ADDRSTRLEN];
|
| 986 | struct iphdr *pkt = (struct iphdr *)packet;
|
| 987 | const char *src, *dst;
|
| 988 | u_char proto;
|
| 989 |
|
| 990 | src = inet_ntop(AF_INET, &pkt->saddr, src_buf, sizeof(src_buf));
|
| 991 | if (!src)
|
| 992 | src = "<invalid>";
|
| 993 | dst = inet_ntop(AF_INET, &pkt->daddr, dst_buf, sizeof(dst_buf));
|
| 994 | if (!dst)
|
| 995 | dst = "<invalid>";
|
| 996 | proto = pkt->protocol;
|
| 997 | show_transport(packet + sizeof(struct iphdr),
|
| 998 | ntohs(pkt->tot_len),
|
| 999 | ifindex, src, dst, proto, false, pkt_type);
|
| 1000 | }
|
| 1001 |
|
| 1002 | static void *traffic_monitor_thread(void *arg)
|
| 1003 | {
|
| 1004 | char *ifname, _ifname[IF_NAMESIZE];
|
| 1005 | const u_char *packet, *payload;
|
| 1006 | struct tmonitor_ctx *ctx = arg;
|
| 1007 | pcap_dumper_t *dumper = ctx->dumper;
|
| 1008 | int fd = ctx->pcap_fd, nfds, r;
|
| 1009 | int wake_fd = ctx->wake_fd;
|
| 1010 | struct pcap_pkthdr header;
|
| 1011 | pcap_t *pcap = ctx->pcap;
|
| 1012 | u32 ifindex;
|
| 1013 | fd_set fds;
|
| 1014 | u16 proto;
|
| 1015 | u8 ptype;
|
| 1016 |
|
| 1017 | nfds = (fd > wake_fd ? fd : wake_fd) + 1;
|
| 1018 | FD_ZERO(&fds);
|
| 1019 |
|
| 1020 | while (!ctx->done) {
|
| 1021 | FD_SET(fd, &fds);
|
| 1022 | FD_SET(wake_fd, &fds);
|
| 1023 | r = select(nfds, &fds, NULL, NULL, NULL);
|
| 1024 | if (!r)
|
| 1025 | continue;
|
| 1026 | if (r < 0) {
|
| 1027 | if (errno == EINTR)
|
| 1028 | continue;
|
| 1029 | log_err("Fail to select on pcap fd and wake fd");
|
| 1030 | break;
|
| 1031 | }
|
| 1032 |
|
| 1033 |
|
| 1034 | packet = pcap_next(pcap, &header);
|
| 1035 | if (!packet)
|
| 1036 | continue;
|
| 1037 |
|
| 1038 | |
| 1039 | * According to the man page of pcap_dump(), first argument
|
| 1040 | * is the pcap_dumper_t pointer even it's argument type is
|
| 1041 | * u_char *.
|
| 1042 |
|
| 1043 | pcap_dump((u_char *)dumper, &header, packet);
|
| 1044 |
|
| 1045 | |
| 1046 | * Not sure what other types of packets look like. Here, we
|
| 1047 | * parse only Ethernet and compatible packets.
|
| 1048 |
|
| 1049 | if (!is_ethernet(packet))
|
| 1050 | continue;
|
| 1051 |
|
| 1052 | |
| 1053 | * Skip SLL2 header
|
| 1054 | * https://www.tcpdump.org/linktypes/LINKTYPE_LINUX_SLL2.html
|
| 1055 | *
|
| 1056 | * Although the document doesn't mention that, the payload
|
| 1057 | * doesn't include the Ethernet header. The payload starts
|
| 1058 | * from the first byte of the network layer header.
|
| 1059 |
|
| 1060 | payload = packet + 20;
|
| 1061 |
|
| 1062 | memcpy(&proto, packet, 2);
|
| 1063 | proto = ntohs(proto);
|
| 1064 | memcpy(&ifindex, packet + 4, 4);
|
| 1065 | ifindex = ntohl(ifindex);
|
| 1066 | ptype = packet[10];
|
| 1067 |
|
| 1068 | if (proto == ETH_P_IPV6) {
|
| 1069 | show_ipv6_packet(payload, ifindex, ptype);
|
| 1070 | } else if (proto == ETH_P_IP) {
|
| 1071 | show_ipv4_packet(payload, ifindex, ptype);
|
| 1072 | } else {
|
| 1073 | ifname = if_indextoname(ifindex, _ifname);
|
| 1074 | if (!ifname) {
|
| 1075 | snprintf(_ifname, sizeof(_ifname), "unknown(%d)", ifindex);
|
| 1076 | ifname = _ifname;
|
| 1077 | }
|
| 1078 |
|
| 1079 | tm_print("%-7s %-3s Unknown network protocol type 0x%x\n",
|
| 1080 | ifname, pkt_type_str(ptype), proto);
|
| 1081 | }
|
| 1082 | }
|
| 1083 |
|
| 1084 | return NULL;
|
| 1085 | }
|
| 1086 |
|
| 1087 | |
| 1088 | * Prepare the pcap handle to capture packets.
|
| 1089 | *
|
| 1090 | * This pcap is non-blocking and immediate mode is enabled to receive
|
| 1091 | * captured packets as soon as possible. The snaplen is set to 1024 bytes
|
| 1092 | * to limit the size of captured content. The format of the link-layer
|
| 1093 | * header is set to DLT_LINUX_SLL2 to enable handling various link-layer
|
| 1094 | * technologies.
|
| 1095 |
|
| 1096 | static pcap_t *traffic_monitor_prepare_pcap(void)
|
| 1097 | {
|
| 1098 | char errbuf[PCAP_ERRBUF_SIZE];
|
| 1099 | pcap_t *pcap;
|
| 1100 | int r;
|
| 1101 |
|
| 1102 |
|
| 1103 | pcap = pcap_create("any", errbuf);
|
| 1104 | if (!pcap) {
|
| 1105 | log_err("Failed to open pcap: %s", errbuf);
|
| 1106 | return NULL;
|
| 1107 | }
|
| 1108 |
|
| 1109 | r = pcap_set_snaplen(pcap, 1024);
|
| 1110 | if (r) {
|
| 1111 | log_err("Failed to set snaplen: %s", pcap_geterr(pcap));
|
| 1112 | goto error;
|
| 1113 | }
|
| 1114 |
|
| 1115 | r = pcap_set_immediate_mode(pcap, 1);
|
| 1116 | if (r) {
|
| 1117 | log_err("Failed to set immediate mode: %s", pcap_geterr(pcap));
|
| 1118 | goto error;
|
| 1119 | }
|
| 1120 | r = pcap_setnonblock(pcap, 1, errbuf);
|
| 1121 | if (r) {
|
| 1122 | log_err("Failed to set nonblock: %s", errbuf);
|
| 1123 | goto error;
|
| 1124 | }
|
| 1125 | r = pcap_activate(pcap);
|
| 1126 | if (r) {
|
| 1127 | log_err("Failed to activate pcap: %s", pcap_geterr(pcap));
|
| 1128 | goto error;
|
| 1129 | }
|
| 1130 |
|
| 1131 | r = pcap_set_datalink(pcap, DLT_LINUX_SLL2);
|
| 1132 | if (r) {
|
| 1133 | log_err("Failed to set datalink: %s", pcap_geterr(pcap));
|
| 1134 | goto error;
|
| 1135 | }
|
| 1136 |
|
| 1137 | return pcap;
|
| 1138 | error:
|
| 1139 | pcap_close(pcap);
|
| 1140 | return NULL;
|
| 1141 | }
|
| 1142 |
|
| 1143 | static void encode_test_name(char *buf, size_t len, const char *test_name, const char *subtest_name)
|
| 1144 | {
|
| 1145 | char *p;
|
| 1146 |
|
| 1147 | if (subtest_name)
|
| 1148 | snprintf(buf, len, "%s__%s", test_name, subtest_name);
|
| 1149 | else
|
| 1150 | snprintf(buf, len, "%s", test_name);
|
| 1151 | while ((p = strchr(buf, '/')))
|
| 1152 | *p = '_';
|
| 1153 | while ((p = strchr(buf, ' ')))
|
| 1154 | *p = '_';
|
| 1155 | }
|
| 1156 |
|
| 1157 | #define PCAP_DIR "/tmp/tmon_pcap"
|
| 1158 |
|
| 1159 | |
| 1160 | * Start to monitor the network traffic in the given network namespace.
|
| 1161 | *
|
| 1162 | * netns: the name of the network namespace to monitor. If NULL, the
|
| 1163 | * current network namespace is monitored.
|
| 1164 | * test_name: the name of the running test.
|
| 1165 | * subtest_name: the name of the running subtest if there is. It should be
|
| 1166 | * NULL if it is not a subtest.
|
| 1167 | *
|
| 1168 | * This function will start a thread to capture packets going through NICs
|
| 1169 | * in the give network namespace.
|
| 1170 |
|
| 1171 | struct tmonitor_ctx *traffic_monitor_start(const char *netns, const char *test_name,
|
| 1172 | const char *subtest_name)
|
| 1173 | {
|
| 1174 | struct nstoken *nstoken = NULL;
|
| 1175 | struct tmonitor_ctx *ctx;
|
| 1176 | char test_name_buf[64];
|
| 1177 | static int tmon_seq;
|
| 1178 | int r;
|
| 1179 |
|
| 1180 | if (netns) {
|
| 1181 | nstoken = open_netns(netns);
|
| 1182 | if (!nstoken)
|
| 1183 | return NULL;
|
| 1184 | }
|
| 1185 | ctx = malloc(sizeof(*ctx));
|
| 1186 | if (!ctx) {
|
| 1187 | log_err("Failed to malloc ctx");
|
| 1188 | goto fail_ctx;
|
| 1189 | }
|
| 1190 | memset(ctx, 0, sizeof(*ctx));
|
| 1191 |
|
| 1192 | encode_test_name(test_name_buf, sizeof(test_name_buf), test_name, subtest_name);
|
| 1193 | snprintf(ctx->pkt_fname, sizeof(ctx->pkt_fname),
|
| 1194 | PCAP_DIR "/packets-%d-%d-%s-%s.log", getpid(), tmon_seq++,
|
| 1195 | test_name_buf, netns ? netns : "unknown");
|
| 1196 |
|
| 1197 | r = mkdir(PCAP_DIR, 0755);
|
| 1198 | if (r && errno != EEXIST) {
|
| 1199 | log_err("Failed to create " PCAP_DIR);
|
| 1200 | goto fail_pcap;
|
| 1201 | }
|
| 1202 |
|
| 1203 | ctx->pcap = traffic_monitor_prepare_pcap();
|
| 1204 | if (!ctx->pcap)
|
| 1205 | goto fail_pcap;
|
| 1206 | ctx->pcap_fd = pcap_get_selectable_fd(ctx->pcap);
|
| 1207 | if (ctx->pcap_fd < 0) {
|
| 1208 | log_err("Failed to get pcap fd");
|
| 1209 | goto fail_dumper;
|
| 1210 | }
|
| 1211 |
|
| 1212 |
|
| 1213 | ctx->dumper = pcap_dump_open(ctx->pcap, ctx->pkt_fname);
|
| 1214 | if (!ctx->dumper) {
|
| 1215 | log_err("Failed to open pcap dump: %s", ctx->pkt_fname);
|
| 1216 | goto fail_dumper;
|
| 1217 | }
|
| 1218 |
|
| 1219 |
|
| 1220 | ctx->wake_fd = eventfd(0, 0);
|
| 1221 | if (ctx->wake_fd < 0) {
|
| 1222 | log_err("Failed to create eventfd");
|
| 1223 | goto fail_eventfd;
|
| 1224 | }
|
| 1225 |
|
| 1226 | r = pthread_create(&ctx->thread, NULL, traffic_monitor_thread, ctx);
|
| 1227 | if (r) {
|
| 1228 | log_err("Failed to create thread");
|
| 1229 | goto fail;
|
| 1230 | }
|
| 1231 |
|
| 1232 | close_netns(nstoken);
|
| 1233 |
|
| 1234 | return ctx;
|
| 1235 |
|
| 1236 | fail:
|
| 1237 | close(ctx->wake_fd);
|
| 1238 |
|
| 1239 | fail_eventfd:
|
| 1240 | pcap_dump_close(ctx->dumper);
|
| 1241 | unlink(ctx->pkt_fname);
|
| 1242 |
|
| 1243 | fail_dumper:
|
| 1244 | pcap_close(ctx->pcap);
|
| 1245 |
|
| 1246 | fail_pcap:
|
| 1247 | free(ctx);
|
| 1248 |
|
| 1249 | fail_ctx:
|
| 1250 | close_netns(nstoken);
|
| 1251 |
|
| 1252 | return NULL;
|
| 1253 | }
|
| 1254 |
|
| 1255 | static void traffic_monitor_release(struct tmonitor_ctx *ctx)
|
| 1256 | {
|
| 1257 | pcap_close(ctx->pcap);
|
| 1258 | pcap_dump_close(ctx->dumper);
|
| 1259 |
|
| 1260 | close(ctx->wake_fd);
|
| 1261 |
|
| 1262 | free(ctx);
|
| 1263 | }
|
| 1264 |
|
| 1265 | |
| 1266 | * Stop the network traffic monitor.
|
| 1267 | *
|
| 1268 | * ctx: the context returned by traffic_monitor_start()
|
| 1269 |
|
| 1270 | void traffic_monitor_stop(struct tmonitor_ctx *ctx)
|
| 1271 | {
|
| 1272 | __u64 w = 1;
|
| 1273 |
|
| 1274 | if (!ctx)
|
| 1275 | return;
|
| 1276 |
|
| 1277 |
|
| 1278 | ctx->done = true;
|
| 1279 |
|
| 1280 | write(ctx->wake_fd, &w, sizeof(w));
|
| 1281 | pthread_join(ctx->thread, NULL);
|
| 1282 |
|
| 1283 | tm_print("Packet file: %s\n", strrchr(ctx->pkt_fname, '/') + 1);
|
| 1284 |
|
| 1285 | traffic_monitor_release(ctx);
|
| 1286 | }
|
| 1287 |
|
| 1288 | #endif
|