xref: /linux/tools/testing/selftests/net/mptcp/mptcp_connect.c (revision d53b8e36925256097a08d7cb749198d85cbf9b2b)
1 // SPDX-License-Identifier: GPL-2.0
2 
3 #define _GNU_SOURCE
4 
5 #include <errno.h>
6 #include <limits.h>
7 #include <fcntl.h>
8 #include <string.h>
9 #include <stdarg.h>
10 #include <stdbool.h>
11 #include <stdint.h>
12 #include <stdio.h>
13 #include <stdlib.h>
14 #include <strings.h>
15 #include <signal.h>
16 #include <unistd.h>
17 #include <time.h>
18 
19 #include <sys/ioctl.h>
20 #include <sys/poll.h>
21 #include <sys/random.h>
22 #include <sys/sendfile.h>
23 #include <sys/stat.h>
24 #include <sys/socket.h>
25 #include <sys/types.h>
26 #include <sys/mman.h>
27 
28 #include <netdb.h>
29 #include <netinet/in.h>
30 
31 #include <linux/tcp.h>
32 #include <linux/time_types.h>
33 #include <linux/sockios.h>
34 
35 extern int optind;
36 
37 #ifndef IPPROTO_MPTCP
38 #define IPPROTO_MPTCP 262
39 #endif
40 #ifndef TCP_ULP
41 #define TCP_ULP 31
42 #endif
43 
44 static int  poll_timeout = 10 * 1000;
45 static bool listen_mode;
46 static bool quit;
47 
48 enum cfg_mode {
49 	CFG_MODE_POLL,
50 	CFG_MODE_MMAP,
51 	CFG_MODE_SENDFILE,
52 };
53 
54 enum cfg_peek {
55 	CFG_NONE_PEEK,
56 	CFG_WITH_PEEK,
57 	CFG_AFTER_PEEK,
58 };
59 
60 static enum cfg_mode cfg_mode = CFG_MODE_POLL;
61 static enum cfg_peek cfg_peek = CFG_NONE_PEEK;
62 static const char *cfg_host;
63 static const char *cfg_port	= "12000";
64 static int cfg_sock_proto	= IPPROTO_MPTCP;
65 static int pf = AF_INET;
66 static int cfg_sndbuf;
67 static int cfg_rcvbuf;
68 static bool cfg_join;
69 static bool cfg_remove;
70 static unsigned int cfg_time;
71 static unsigned int cfg_do_w;
72 static int cfg_wait;
73 static uint32_t cfg_mark;
74 static char *cfg_input;
75 static int cfg_repeat = 1;
76 static int cfg_truncate;
77 static int cfg_rcv_trunc;
78 
79 struct cfg_cmsg_types {
80 	unsigned int cmsg_enabled:1;
81 	unsigned int timestampns:1;
82 	unsigned int tcp_inq:1;
83 };
84 
85 struct cfg_sockopt_types {
86 	unsigned int transparent:1;
87 	unsigned int mptfo:1;
88 };
89 
90 struct tcp_inq_state {
91 	unsigned int last;
92 	bool expect_eof;
93 };
94 
95 struct wstate {
96 	char buf[8192];
97 	unsigned int len;
98 	unsigned int off;
99 	unsigned int total_len;
100 };
101 
102 static struct tcp_inq_state tcp_inq;
103 
104 static struct cfg_cmsg_types cfg_cmsg_types;
105 static struct cfg_sockopt_types cfg_sockopt_types;
106 
107 static void die_usage(void)
108 {
109 	fprintf(stderr, "Usage: mptcp_connect [-6] [-c cmsg] [-f offset] [-i file] [-I num] [-j] [-l] "
110 		"[-m mode] [-M mark] [-o option] [-p port] [-P mode] [-r num] [-R num] "
111 		"[-s MPTCP|TCP] [-S num] [-t num] [-T num] [-w sec] connect_address\n");
112 	fprintf(stderr, "\t-6 use ipv6\n");
113 	fprintf(stderr, "\t-c cmsg -- test cmsg type <cmsg>\n");
114 	fprintf(stderr, "\t-f offset -- stop the I/O after receiving and sending the specified amount "
115 		"of bytes. If there are unread bytes in the receive queue, that will cause a MPTCP "
116 		"fastclose at close/shutdown. If offset is negative, expect the peer to close before "
117 		"all the local data as been sent, thus toleration errors on write and EPIPE signals\n");
118 	fprintf(stderr, "\t-i file -- read the data to send from the given file instead of stdin");
119 	fprintf(stderr, "\t-I num -- repeat the transfer 'num' times. In listen mode accepts num "
120 		"incoming connections, in client mode, disconnect and reconnect to the server\n");
121 	fprintf(stderr, "\t-j     -- add additional sleep at connection start and tear down "
122 		"-- for MPJ tests\n");
123 	fprintf(stderr, "\t-l     -- listens mode, accepts incoming connection\n");
124 	fprintf(stderr, "\t-m [poll|mmap|sendfile] -- use poll(default)/mmap+write/sendfile\n");
125 	fprintf(stderr, "\t-M mark -- set socket packet mark\n");
126 	fprintf(stderr, "\t-o option -- test sockopt <option>\n");
127 	fprintf(stderr, "\t-p num -- use port num\n");
128 	fprintf(stderr,
129 		"\t-P [saveWithPeek|saveAfterPeek] -- save data with/after MSG_PEEK form tcp socket\n");
130 	fprintf(stderr, "\t-r num -- enable slow mode, limiting each write to num bytes "
131 		"-- for remove addr tests\n");
132 	fprintf(stderr, "\t-R num -- set SO_RCVBUF to num\n");
133 	fprintf(stderr, "\t-s [MPTCP|TCP] -- use mptcp(default) or tcp sockets\n");
134 	fprintf(stderr, "\t-S num -- set SO_SNDBUF to num\n");
135 	fprintf(stderr, "\t-t num -- set poll timeout to num\n");
136 	fprintf(stderr, "\t-T num -- set expected runtime to num ms\n");
137 	fprintf(stderr, "\t-w num -- wait num sec before closing the socket\n");
138 	exit(1);
139 }
140 
141 static void xerror(const char *fmt, ...)
142 {
143 	va_list ap;
144 
145 	va_start(ap, fmt);
146 	vfprintf(stderr, fmt, ap);
147 	va_end(ap);
148 	exit(1);
149 }
150 
151 static void handle_signal(int nr)
152 {
153 	quit = true;
154 }
155 
156 static const char *getxinfo_strerr(int err)
157 {
158 	if (err == EAI_SYSTEM)
159 		return strerror(errno);
160 
161 	return gai_strerror(err);
162 }
163 
164 static void xgetnameinfo(const struct sockaddr *addr, socklen_t addrlen,
165 			 char *host, socklen_t hostlen,
166 			 char *serv, socklen_t servlen)
167 {
168 	int flags = NI_NUMERICHOST | NI_NUMERICSERV;
169 	int err = getnameinfo(addr, addrlen, host, hostlen, serv, servlen,
170 			      flags);
171 
172 	if (err) {
173 		const char *errstr = getxinfo_strerr(err);
174 
175 		fprintf(stderr, "Fatal: getnameinfo: %s\n", errstr);
176 		exit(1);
177 	}
178 }
179 
180 static void xgetaddrinfo(const char *node, const char *service,
181 			 const struct addrinfo *hints,
182 			 struct addrinfo **res)
183 {
184 	int err = getaddrinfo(node, service, hints, res);
185 
186 	if (err) {
187 		const char *errstr = getxinfo_strerr(err);
188 
189 		fprintf(stderr, "Fatal: getaddrinfo(%s:%s): %s\n",
190 			node ? node : "", service ? service : "", errstr);
191 		exit(1);
192 	}
193 }
194 
195 static void set_rcvbuf(int fd, unsigned int size)
196 {
197 	int err;
198 
199 	err = setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &size, sizeof(size));
200 	if (err) {
201 		perror("set SO_RCVBUF");
202 		exit(1);
203 	}
204 }
205 
206 static void set_sndbuf(int fd, unsigned int size)
207 {
208 	int err;
209 
210 	err = setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &size, sizeof(size));
211 	if (err) {
212 		perror("set SO_SNDBUF");
213 		exit(1);
214 	}
215 }
216 
217 static void set_mark(int fd, uint32_t mark)
218 {
219 	int err;
220 
221 	err = setsockopt(fd, SOL_SOCKET, SO_MARK, &mark, sizeof(mark));
222 	if (err) {
223 		perror("set SO_MARK");
224 		exit(1);
225 	}
226 }
227 
228 static void set_transparent(int fd, int pf)
229 {
230 	int one = 1;
231 
232 	switch (pf) {
233 	case AF_INET:
234 		if (-1 == setsockopt(fd, SOL_IP, IP_TRANSPARENT, &one, sizeof(one)))
235 			perror("IP_TRANSPARENT");
236 		break;
237 	case AF_INET6:
238 		if (-1 == setsockopt(fd, IPPROTO_IPV6, IPV6_TRANSPARENT, &one, sizeof(one)))
239 			perror("IPV6_TRANSPARENT");
240 		break;
241 	}
242 }
243 
244 static void set_mptfo(int fd, int pf)
245 {
246 	int qlen = 25;
247 
248 	if (setsockopt(fd, IPPROTO_TCP, TCP_FASTOPEN, &qlen, sizeof(qlen)) == -1)
249 		perror("TCP_FASTOPEN");
250 }
251 
252 static int do_ulp_so(int sock, const char *name)
253 {
254 	return setsockopt(sock, IPPROTO_TCP, TCP_ULP, name, strlen(name));
255 }
256 
257 #define X(m)	xerror("%s:%u: %s: failed for proto %d at line %u", __FILE__, __LINE__, (m), proto, line)
258 static void sock_test_tcpulp(int sock, int proto, unsigned int line)
259 {
260 	socklen_t buflen = 8;
261 	char buf[8] = "";
262 	int ret = getsockopt(sock, IPPROTO_TCP, TCP_ULP, buf, &buflen);
263 
264 	if (ret != 0)
265 		X("getsockopt");
266 
267 	if (buflen > 0) {
268 		if (strcmp(buf, "mptcp") != 0)
269 			xerror("unexpected ULP '%s' for proto %d at line %u", buf, proto, line);
270 		ret = do_ulp_so(sock, "tls");
271 		if (ret == 0)
272 			X("setsockopt");
273 	} else if (proto == IPPROTO_MPTCP) {
274 		ret = do_ulp_so(sock, "tls");
275 		if (ret != -1)
276 			X("setsockopt");
277 	}
278 
279 	ret = do_ulp_so(sock, "mptcp");
280 	if (ret != -1)
281 		X("setsockopt");
282 
283 #undef X
284 }
285 
286 #define SOCK_TEST_TCPULP(s, p) sock_test_tcpulp((s), (p), __LINE__)
287 
288 static int sock_listen_mptcp(const char * const listenaddr,
289 			     const char * const port)
290 {
291 	int sock = -1;
292 	struct addrinfo hints = {
293 		.ai_protocol = IPPROTO_TCP,
294 		.ai_socktype = SOCK_STREAM,
295 		.ai_flags = AI_PASSIVE | AI_NUMERICHOST
296 	};
297 
298 	hints.ai_family = pf;
299 
300 	struct addrinfo *a, *addr;
301 	int one = 1;
302 
303 	xgetaddrinfo(listenaddr, port, &hints, &addr);
304 	hints.ai_family = pf;
305 
306 	for (a = addr; a; a = a->ai_next) {
307 		sock = socket(a->ai_family, a->ai_socktype, cfg_sock_proto);
308 		if (sock < 0)
309 			continue;
310 
311 		SOCK_TEST_TCPULP(sock, cfg_sock_proto);
312 
313 		if (-1 == setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, &one,
314 				     sizeof(one)))
315 			perror("setsockopt");
316 
317 		if (cfg_sockopt_types.transparent)
318 			set_transparent(sock, pf);
319 
320 		if (cfg_sockopt_types.mptfo)
321 			set_mptfo(sock, pf);
322 
323 		if (bind(sock, a->ai_addr, a->ai_addrlen) == 0)
324 			break; /* success */
325 
326 		perror("bind");
327 		close(sock);
328 		sock = -1;
329 	}
330 
331 	freeaddrinfo(addr);
332 
333 	if (sock < 0) {
334 		fprintf(stderr, "Could not create listen socket\n");
335 		return sock;
336 	}
337 
338 	SOCK_TEST_TCPULP(sock, cfg_sock_proto);
339 
340 	if (listen(sock, 20)) {
341 		perror("listen");
342 		close(sock);
343 		return -1;
344 	}
345 
346 	SOCK_TEST_TCPULP(sock, cfg_sock_proto);
347 
348 	return sock;
349 }
350 
351 static int sock_connect_mptcp(const char * const remoteaddr,
352 			      const char * const port, int proto,
353 			      struct addrinfo **peer,
354 			      int infd, struct wstate *winfo)
355 {
356 	struct addrinfo hints = {
357 		.ai_protocol = IPPROTO_TCP,
358 		.ai_socktype = SOCK_STREAM,
359 	};
360 	struct addrinfo *a, *addr;
361 	int syn_copied = 0;
362 	int sock = -1;
363 
364 	hints.ai_family = pf;
365 
366 	xgetaddrinfo(remoteaddr, port, &hints, &addr);
367 	for (a = addr; a; a = a->ai_next) {
368 		sock = socket(a->ai_family, a->ai_socktype, proto);
369 		if (sock < 0) {
370 			perror("socket");
371 			continue;
372 		}
373 
374 		SOCK_TEST_TCPULP(sock, proto);
375 
376 		if (cfg_mark)
377 			set_mark(sock, cfg_mark);
378 
379 		if (cfg_sockopt_types.mptfo) {
380 			if (!winfo->total_len)
381 				winfo->total_len = winfo->len = read(infd, winfo->buf,
382 								     sizeof(winfo->buf));
383 
384 			syn_copied = sendto(sock, winfo->buf, winfo->len, MSG_FASTOPEN,
385 					    a->ai_addr, a->ai_addrlen);
386 			if (syn_copied >= 0) {
387 				winfo->off = syn_copied;
388 				winfo->len -= syn_copied;
389 				*peer = a;
390 				break; /* success */
391 			}
392 		} else {
393 			if (connect(sock, a->ai_addr, a->ai_addrlen) == 0) {
394 				*peer = a;
395 				break; /* success */
396 			}
397 		}
398 		if (cfg_sockopt_types.mptfo) {
399 			perror("sendto()");
400 			close(sock);
401 			sock = -1;
402 		} else {
403 			perror("connect()");
404 			close(sock);
405 			sock = -1;
406 		}
407 	}
408 
409 	freeaddrinfo(addr);
410 	if (sock != -1)
411 		SOCK_TEST_TCPULP(sock, proto);
412 	return sock;
413 }
414 
415 static size_t do_rnd_write(const int fd, char *buf, const size_t len)
416 {
417 	static bool first = true;
418 	unsigned int do_w;
419 	ssize_t bw;
420 
421 	do_w = rand() & 0xffff;
422 	if (do_w == 0 || do_w > len)
423 		do_w = len;
424 
425 	if (cfg_join && first && do_w > 100)
426 		do_w = 100;
427 
428 	if (cfg_remove && do_w > cfg_do_w)
429 		do_w = cfg_do_w;
430 
431 	bw = write(fd, buf, do_w);
432 	if (bw < 0)
433 		return bw;
434 
435 	/* let the join handshake complete, before going on */
436 	if (cfg_join && first) {
437 		usleep(200000);
438 		first = false;
439 	}
440 
441 	if (cfg_remove)
442 		usleep(200000);
443 
444 	return bw;
445 }
446 
447 static size_t do_write(const int fd, char *buf, const size_t len)
448 {
449 	size_t offset = 0;
450 
451 	while (offset < len) {
452 		size_t written;
453 		ssize_t bw;
454 
455 		bw = write(fd, buf + offset, len - offset);
456 		if (bw < 0) {
457 			perror("write");
458 			return 0;
459 		}
460 
461 		written = (size_t)bw;
462 		offset += written;
463 	}
464 
465 	return offset;
466 }
467 
468 static void process_cmsg(struct msghdr *msgh)
469 {
470 	struct __kernel_timespec ts;
471 	bool inq_found = false;
472 	bool ts_found = false;
473 	unsigned int inq = 0;
474 	struct cmsghdr *cmsg;
475 
476 	for (cmsg = CMSG_FIRSTHDR(msgh); cmsg ; cmsg = CMSG_NXTHDR(msgh, cmsg)) {
477 		if (cmsg->cmsg_level == SOL_SOCKET && cmsg->cmsg_type == SO_TIMESTAMPNS_NEW) {
478 			memcpy(&ts, CMSG_DATA(cmsg), sizeof(ts));
479 			ts_found = true;
480 			continue;
481 		}
482 		if (cmsg->cmsg_level == IPPROTO_TCP && cmsg->cmsg_type == TCP_CM_INQ) {
483 			memcpy(&inq, CMSG_DATA(cmsg), sizeof(inq));
484 			inq_found = true;
485 			continue;
486 		}
487 
488 	}
489 
490 	if (cfg_cmsg_types.timestampns) {
491 		if (!ts_found)
492 			xerror("TIMESTAMPNS not present\n");
493 	}
494 
495 	if (cfg_cmsg_types.tcp_inq) {
496 		if (!inq_found)
497 			xerror("TCP_INQ not present\n");
498 
499 		if (inq > 1024)
500 			xerror("tcp_inq %u is larger than one kbyte\n", inq);
501 		tcp_inq.last = inq;
502 	}
503 }
504 
505 static ssize_t do_recvmsg_cmsg(const int fd, char *buf, const size_t len)
506 {
507 	char msg_buf[8192];
508 	struct iovec iov = {
509 		.iov_base = buf,
510 		.iov_len = len,
511 	};
512 	struct msghdr msg = {
513 		.msg_iov = &iov,
514 		.msg_iovlen = 1,
515 		.msg_control = msg_buf,
516 		.msg_controllen = sizeof(msg_buf),
517 	};
518 	int flags = 0;
519 	unsigned int last_hint = tcp_inq.last;
520 	int ret = recvmsg(fd, &msg, flags);
521 
522 	if (ret <= 0) {
523 		if (ret == 0 && tcp_inq.expect_eof)
524 			return ret;
525 
526 		if (ret == 0 && cfg_cmsg_types.tcp_inq)
527 			if (last_hint != 1 && last_hint != 0)
528 				xerror("EOF but last tcp_inq hint was %u\n", last_hint);
529 
530 		return ret;
531 	}
532 
533 	if (tcp_inq.expect_eof)
534 		xerror("expected EOF, last_hint %u, now %u\n",
535 		       last_hint, tcp_inq.last);
536 
537 	if (msg.msg_controllen && !cfg_cmsg_types.cmsg_enabled)
538 		xerror("got %lu bytes of cmsg data, expected 0\n",
539 		       (unsigned long)msg.msg_controllen);
540 
541 	if (msg.msg_controllen == 0 && cfg_cmsg_types.cmsg_enabled)
542 		xerror("%s\n", "got no cmsg data");
543 
544 	if (msg.msg_controllen)
545 		process_cmsg(&msg);
546 
547 	if (cfg_cmsg_types.tcp_inq) {
548 		if ((size_t)ret < len && last_hint > (unsigned int)ret) {
549 			if (ret + 1 != (int)last_hint) {
550 				int next = read(fd, msg_buf, sizeof(msg_buf));
551 
552 				xerror("read %u of %u, last_hint was %u tcp_inq hint now %u next_read returned %d/%m\n",
553 				       ret, (unsigned int)len, last_hint, tcp_inq.last, next);
554 			} else {
555 				tcp_inq.expect_eof = true;
556 			}
557 		}
558 	}
559 
560 	return ret;
561 }
562 
563 static ssize_t do_rnd_read(const int fd, char *buf, const size_t len)
564 {
565 	int ret = 0;
566 	char tmp[16384];
567 	size_t cap = rand();
568 
569 	cap &= 0xffff;
570 
571 	if (cap == 0)
572 		cap = 1;
573 	else if (cap > len)
574 		cap = len;
575 
576 	if (cfg_peek == CFG_WITH_PEEK) {
577 		ret = recv(fd, buf, cap, MSG_PEEK);
578 		ret = (ret < 0) ? ret : read(fd, tmp, ret);
579 	} else if (cfg_peek == CFG_AFTER_PEEK) {
580 		ret = recv(fd, buf, cap, MSG_PEEK);
581 		ret = (ret < 0) ? ret : read(fd, buf, cap);
582 	} else if (cfg_cmsg_types.cmsg_enabled) {
583 		ret = do_recvmsg_cmsg(fd, buf, cap);
584 	} else {
585 		ret = read(fd, buf, cap);
586 	}
587 
588 	return ret;
589 }
590 
591 static void set_nonblock(int fd, bool nonblock)
592 {
593 	int flags = fcntl(fd, F_GETFL);
594 
595 	if (flags == -1)
596 		return;
597 
598 	if (nonblock)
599 		fcntl(fd, F_SETFL, flags | O_NONBLOCK);
600 	else
601 		fcntl(fd, F_SETFL, flags & ~O_NONBLOCK);
602 }
603 
604 static void shut_wr(int fd)
605 {
606 	/* Close our write side, ev. give some time
607 	 * for address notification and/or checking
608 	 * the current status
609 	 */
610 	if (cfg_wait)
611 		usleep(cfg_wait);
612 
613 	shutdown(fd, SHUT_WR);
614 }
615 
616 static int copyfd_io_poll(int infd, int peerfd, int outfd,
617 			  bool *in_closed_after_out, struct wstate *winfo)
618 {
619 	struct pollfd fds = {
620 		.fd = peerfd,
621 		.events = POLLIN | POLLOUT,
622 	};
623 	unsigned int total_wlen = 0, total_rlen = 0;
624 
625 	set_nonblock(peerfd, true);
626 
627 	for (;;) {
628 		char rbuf[8192];
629 		ssize_t len;
630 
631 		if (fds.events == 0 || quit)
632 			break;
633 
634 		switch (poll(&fds, 1, poll_timeout)) {
635 		case -1:
636 			if (errno == EINTR)
637 				continue;
638 			perror("poll");
639 			return 1;
640 		case 0:
641 			fprintf(stderr, "%s: poll timed out (events: "
642 				"POLLIN %u, POLLOUT %u)\n", __func__,
643 				fds.events & POLLIN, fds.events & POLLOUT);
644 			return 2;
645 		}
646 
647 		if (fds.revents & POLLIN) {
648 			ssize_t rb = sizeof(rbuf);
649 
650 			/* limit the total amount of read data to the trunc value*/
651 			if (cfg_truncate > 0) {
652 				if (rb + total_rlen > cfg_truncate)
653 					rb = cfg_truncate - total_rlen;
654 				len = read(peerfd, rbuf, rb);
655 			} else {
656 				len = do_rnd_read(peerfd, rbuf, sizeof(rbuf));
657 			}
658 			if (len == 0) {
659 				/* no more data to receive:
660 				 * peer has closed its write side
661 				 */
662 				fds.events &= ~POLLIN;
663 
664 				if ((fds.events & POLLOUT) == 0) {
665 					*in_closed_after_out = true;
666 					/* and nothing more to send */
667 					break;
668 				}
669 
670 			/* Else, still have data to transmit */
671 			} else if (len < 0) {
672 				if (cfg_rcv_trunc)
673 					return 0;
674 				perror("read");
675 				return 3;
676 			}
677 
678 			total_rlen += len;
679 			do_write(outfd, rbuf, len);
680 		}
681 
682 		if (fds.revents & POLLOUT) {
683 			if (winfo->len == 0) {
684 				winfo->off = 0;
685 				winfo->len = read(infd, winfo->buf, sizeof(winfo->buf));
686 			}
687 
688 			if (winfo->len > 0) {
689 				ssize_t bw;
690 
691 				/* limit the total amount of written data to the trunc value */
692 				if (cfg_truncate > 0 && winfo->len + total_wlen > cfg_truncate)
693 					winfo->len = cfg_truncate - total_wlen;
694 
695 				bw = do_rnd_write(peerfd, winfo->buf + winfo->off, winfo->len);
696 				if (bw < 0) {
697 					if (cfg_rcv_trunc)
698 						return 0;
699 					perror("write");
700 					return 111;
701 				}
702 
703 				winfo->off += bw;
704 				winfo->len -= bw;
705 				total_wlen += bw;
706 			} else if (winfo->len == 0) {
707 				/* We have no more data to send. */
708 				fds.events &= ~POLLOUT;
709 
710 				if ((fds.events & POLLIN) == 0)
711 					/* ... and peer also closed already */
712 					break;
713 
714 				shut_wr(peerfd);
715 			} else {
716 				if (errno == EINTR)
717 					continue;
718 				perror("read");
719 				return 4;
720 			}
721 		}
722 
723 		if (fds.revents & (POLLERR | POLLNVAL)) {
724 			if (cfg_rcv_trunc)
725 				return 0;
726 			fprintf(stderr, "Unexpected revents: "
727 				"POLLERR/POLLNVAL(%x)\n", fds.revents);
728 			return 5;
729 		}
730 
731 		if (cfg_truncate > 0 && total_wlen >= cfg_truncate &&
732 		    total_rlen >= cfg_truncate)
733 			break;
734 	}
735 
736 	/* leave some time for late join/announce */
737 	if (cfg_remove && !quit)
738 		usleep(cfg_wait);
739 
740 	return 0;
741 }
742 
743 static int do_recvfile(int infd, int outfd)
744 {
745 	ssize_t r;
746 
747 	do {
748 		char buf[16384];
749 
750 		r = do_rnd_read(infd, buf, sizeof(buf));
751 		if (r > 0) {
752 			if (write(outfd, buf, r) != r)
753 				break;
754 		} else if (r < 0) {
755 			perror("read");
756 		}
757 	} while (r > 0);
758 
759 	return (int)r;
760 }
761 
762 static int spool_buf(int fd, struct wstate *winfo)
763 {
764 	while (winfo->len) {
765 		int ret = write(fd, winfo->buf + winfo->off, winfo->len);
766 
767 		if (ret < 0) {
768 			perror("write");
769 			return 4;
770 		}
771 		winfo->off += ret;
772 		winfo->len -= ret;
773 	}
774 	return 0;
775 }
776 
777 static int do_mmap(int infd, int outfd, unsigned int size,
778 		   struct wstate *winfo)
779 {
780 	char *inbuf = mmap(NULL, size, PROT_READ, MAP_SHARED, infd, 0);
781 	ssize_t ret = 0, off = winfo->total_len;
782 	size_t rem;
783 
784 	if (inbuf == MAP_FAILED) {
785 		perror("mmap");
786 		return 1;
787 	}
788 
789 	ret = spool_buf(outfd, winfo);
790 	if (ret < 0)
791 		return ret;
792 
793 	rem = size - winfo->total_len;
794 
795 	while (rem > 0) {
796 		ret = write(outfd, inbuf + off, rem);
797 
798 		if (ret < 0) {
799 			perror("write");
800 			break;
801 		}
802 
803 		off += ret;
804 		rem -= ret;
805 	}
806 
807 	munmap(inbuf, size);
808 	return rem;
809 }
810 
811 static int get_infd_size(int fd)
812 {
813 	struct stat sb;
814 	ssize_t count;
815 	int err;
816 
817 	err = fstat(fd, &sb);
818 	if (err < 0) {
819 		perror("fstat");
820 		return -1;
821 	}
822 
823 	if ((sb.st_mode & S_IFMT) != S_IFREG) {
824 		fprintf(stderr, "%s: stdin is not a regular file\n", __func__);
825 		return -2;
826 	}
827 
828 	count = sb.st_size;
829 	if (count > INT_MAX) {
830 		fprintf(stderr, "File too large: %zu\n", count);
831 		return -3;
832 	}
833 
834 	return (int)count;
835 }
836 
837 static int do_sendfile(int infd, int outfd, unsigned int count,
838 		       struct wstate *winfo)
839 {
840 	int ret = spool_buf(outfd, winfo);
841 
842 	if (ret < 0)
843 		return ret;
844 
845 	count -= winfo->total_len;
846 
847 	while (count > 0) {
848 		ssize_t r;
849 
850 		r = sendfile(outfd, infd, NULL, count);
851 		if (r < 0) {
852 			perror("sendfile");
853 			return 3;
854 		}
855 
856 		count -= r;
857 	}
858 
859 	return 0;
860 }
861 
862 static int copyfd_io_mmap(int infd, int peerfd, int outfd,
863 			  unsigned int size, bool *in_closed_after_out,
864 			  struct wstate *winfo)
865 {
866 	int err;
867 
868 	if (listen_mode) {
869 		err = do_recvfile(peerfd, outfd);
870 		if (err)
871 			return err;
872 
873 		err = do_mmap(infd, peerfd, size, winfo);
874 	} else {
875 		err = do_mmap(infd, peerfd, size, winfo);
876 		if (err)
877 			return err;
878 
879 		shut_wr(peerfd);
880 
881 		err = do_recvfile(peerfd, outfd);
882 		*in_closed_after_out = true;
883 	}
884 
885 	return err;
886 }
887 
888 static int copyfd_io_sendfile(int infd, int peerfd, int outfd,
889 			      unsigned int size, bool *in_closed_after_out, struct wstate *winfo)
890 {
891 	int err;
892 
893 	if (listen_mode) {
894 		err = do_recvfile(peerfd, outfd);
895 		if (err)
896 			return err;
897 
898 		err = do_sendfile(infd, peerfd, size, winfo);
899 	} else {
900 		err = do_sendfile(infd, peerfd, size, winfo);
901 		if (err)
902 			return err;
903 
904 		shut_wr(peerfd);
905 
906 		err = do_recvfile(peerfd, outfd);
907 		*in_closed_after_out = true;
908 	}
909 
910 	return err;
911 }
912 
913 static int copyfd_io(int infd, int peerfd, int outfd, bool close_peerfd, struct wstate *winfo)
914 {
915 	bool in_closed_after_out = false;
916 	struct timespec start, end;
917 	int file_size;
918 	int ret;
919 
920 	if (cfg_time && (clock_gettime(CLOCK_MONOTONIC, &start) < 0))
921 		xerror("can not fetch start time %d", errno);
922 
923 	switch (cfg_mode) {
924 	case CFG_MODE_POLL:
925 		ret = copyfd_io_poll(infd, peerfd, outfd, &in_closed_after_out,
926 				     winfo);
927 		break;
928 
929 	case CFG_MODE_MMAP:
930 		file_size = get_infd_size(infd);
931 		if (file_size < 0)
932 			return file_size;
933 		ret = copyfd_io_mmap(infd, peerfd, outfd, file_size,
934 				     &in_closed_after_out, winfo);
935 		break;
936 
937 	case CFG_MODE_SENDFILE:
938 		file_size = get_infd_size(infd);
939 		if (file_size < 0)
940 			return file_size;
941 		ret = copyfd_io_sendfile(infd, peerfd, outfd, file_size,
942 					 &in_closed_after_out, winfo);
943 		break;
944 
945 	default:
946 		fprintf(stderr, "Invalid mode %d\n", cfg_mode);
947 
948 		die_usage();
949 		return 1;
950 	}
951 
952 	if (ret)
953 		return ret;
954 
955 	if (close_peerfd)
956 		close(peerfd);
957 
958 	if (cfg_time) {
959 		unsigned int delta_ms;
960 
961 		if (clock_gettime(CLOCK_MONOTONIC, &end) < 0)
962 			xerror("can not fetch end time %d", errno);
963 		delta_ms = (end.tv_sec - start.tv_sec) * 1000 + (end.tv_nsec - start.tv_nsec) / 1000000;
964 		if (delta_ms > cfg_time) {
965 			xerror("transfer slower than expected! runtime %d ms, expected %d ms",
966 			       delta_ms, cfg_time);
967 		}
968 
969 		/* show the runtime only if this end shutdown(wr) before receiving the EOF,
970 		 * (that is, if this end got the longer runtime)
971 		 */
972 		if (in_closed_after_out)
973 			fprintf(stderr, "%d", delta_ms);
974 	}
975 
976 	return 0;
977 }
978 
979 static void check_sockaddr(int pf, struct sockaddr_storage *ss,
980 			   socklen_t salen)
981 {
982 	struct sockaddr_in6 *sin6;
983 	struct sockaddr_in *sin;
984 	socklen_t wanted_size = 0;
985 
986 	switch (pf) {
987 	case AF_INET:
988 		wanted_size = sizeof(*sin);
989 		sin = (void *)ss;
990 		if (!sin->sin_port)
991 			fprintf(stderr, "accept: something wrong: ip connection from port 0");
992 		break;
993 	case AF_INET6:
994 		wanted_size = sizeof(*sin6);
995 		sin6 = (void *)ss;
996 		if (!sin6->sin6_port)
997 			fprintf(stderr, "accept: something wrong: ipv6 connection from port 0");
998 		break;
999 	default:
1000 		fprintf(stderr, "accept: Unknown pf %d, salen %u\n", pf, salen);
1001 		return;
1002 	}
1003 
1004 	if (salen != wanted_size)
1005 		fprintf(stderr, "accept: size mismatch, got %d expected %d\n",
1006 			(int)salen, wanted_size);
1007 
1008 	if (ss->ss_family != pf)
1009 		fprintf(stderr, "accept: pf mismatch, expect %d, ss_family is %d\n",
1010 			(int)ss->ss_family, pf);
1011 }
1012 
1013 static void check_getpeername(int fd, struct sockaddr_storage *ss, socklen_t salen)
1014 {
1015 	struct sockaddr_storage peerss;
1016 	socklen_t peersalen = sizeof(peerss);
1017 
1018 	if (getpeername(fd, (struct sockaddr *)&peerss, &peersalen) < 0) {
1019 		perror("getpeername");
1020 		return;
1021 	}
1022 
1023 	if (peersalen != salen) {
1024 		fprintf(stderr, "%s: %d vs %d\n", __func__, peersalen, salen);
1025 		return;
1026 	}
1027 
1028 	if (memcmp(ss, &peerss, peersalen)) {
1029 		char a[INET6_ADDRSTRLEN];
1030 		char b[INET6_ADDRSTRLEN];
1031 		char c[INET6_ADDRSTRLEN];
1032 		char d[INET6_ADDRSTRLEN];
1033 
1034 		xgetnameinfo((struct sockaddr *)ss, salen,
1035 			     a, sizeof(a), b, sizeof(b));
1036 
1037 		xgetnameinfo((struct sockaddr *)&peerss, peersalen,
1038 			     c, sizeof(c), d, sizeof(d));
1039 
1040 		fprintf(stderr, "%s: memcmp failure: accept %s vs peername %s, %s vs %s salen %d vs %d\n",
1041 			__func__, a, c, b, d, peersalen, salen);
1042 	}
1043 }
1044 
1045 static void check_getpeername_connect(int fd)
1046 {
1047 	struct sockaddr_storage ss;
1048 	socklen_t salen = sizeof(ss);
1049 	char a[INET6_ADDRSTRLEN];
1050 	char b[INET6_ADDRSTRLEN];
1051 
1052 	if (getpeername(fd, (struct sockaddr *)&ss, &salen) < 0) {
1053 		perror("getpeername");
1054 		return;
1055 	}
1056 
1057 	xgetnameinfo((struct sockaddr *)&ss, salen,
1058 		     a, sizeof(a), b, sizeof(b));
1059 
1060 	if (strcmp(cfg_host, a) || strcmp(cfg_port, b))
1061 		fprintf(stderr, "%s: %s vs %s, %s vs %s\n", __func__,
1062 			cfg_host, a, cfg_port, b);
1063 }
1064 
1065 static void maybe_close(int fd)
1066 {
1067 	unsigned int r = rand();
1068 
1069 	if (!(cfg_join || cfg_remove || cfg_repeat > 1) && (r & 1))
1070 		close(fd);
1071 }
1072 
1073 int main_loop_s(int listensock)
1074 {
1075 	struct sockaddr_storage ss;
1076 	struct wstate winfo;
1077 	struct pollfd polls;
1078 	socklen_t salen;
1079 	int remotesock;
1080 	int fd = 0;
1081 
1082 again:
1083 	polls.fd = listensock;
1084 	polls.events = POLLIN;
1085 
1086 	switch (poll(&polls, 1, poll_timeout)) {
1087 	case -1:
1088 		perror("poll");
1089 		return 1;
1090 	case 0:
1091 		fprintf(stderr, "%s: timed out\n", __func__);
1092 		close(listensock);
1093 		return 2;
1094 	}
1095 
1096 	salen = sizeof(ss);
1097 	remotesock = accept(listensock, (struct sockaddr *)&ss, &salen);
1098 	if (remotesock >= 0) {
1099 		maybe_close(listensock);
1100 		check_sockaddr(pf, &ss, salen);
1101 		check_getpeername(remotesock, &ss, salen);
1102 
1103 		if (cfg_input) {
1104 			fd = open(cfg_input, O_RDONLY);
1105 			if (fd < 0)
1106 				xerror("can't open %s: %d", cfg_input, errno);
1107 		}
1108 
1109 		SOCK_TEST_TCPULP(remotesock, 0);
1110 
1111 		memset(&winfo, 0, sizeof(winfo));
1112 		copyfd_io(fd, remotesock, 1, true, &winfo);
1113 	} else {
1114 		perror("accept");
1115 		return 1;
1116 	}
1117 
1118 	if (cfg_input)
1119 		close(fd);
1120 
1121 	if (--cfg_repeat > 0)
1122 		goto again;
1123 
1124 	return 0;
1125 }
1126 
1127 static void init_rng(void)
1128 {
1129 	unsigned int foo;
1130 
1131 	if (getrandom(&foo, sizeof(foo), 0) == -1) {
1132 		perror("getrandom");
1133 		exit(1);
1134 	}
1135 
1136 	srand(foo);
1137 }
1138 
1139 static void xsetsockopt(int fd, int level, int optname, const void *optval, socklen_t optlen)
1140 {
1141 	int err;
1142 
1143 	err = setsockopt(fd, level, optname, optval, optlen);
1144 	if (err) {
1145 		perror("setsockopt");
1146 		exit(1);
1147 	}
1148 }
1149 
1150 static void apply_cmsg_types(int fd, const struct cfg_cmsg_types *cmsg)
1151 {
1152 	static const unsigned int on = 1;
1153 
1154 	if (cmsg->timestampns)
1155 		xsetsockopt(fd, SOL_SOCKET, SO_TIMESTAMPNS_NEW, &on, sizeof(on));
1156 	if (cmsg->tcp_inq)
1157 		xsetsockopt(fd, IPPROTO_TCP, TCP_INQ, &on, sizeof(on));
1158 }
1159 
1160 static void parse_cmsg_types(const char *type)
1161 {
1162 	char *next = strchr(type, ',');
1163 	unsigned int len = 0;
1164 
1165 	cfg_cmsg_types.cmsg_enabled = 1;
1166 
1167 	if (next) {
1168 		parse_cmsg_types(next + 1);
1169 		len = next - type;
1170 	} else {
1171 		len = strlen(type);
1172 	}
1173 
1174 	if (strncmp(type, "TIMESTAMPNS", len) == 0) {
1175 		cfg_cmsg_types.timestampns = 1;
1176 		return;
1177 	}
1178 
1179 	if (strncmp(type, "TCPINQ", len) == 0) {
1180 		cfg_cmsg_types.tcp_inq = 1;
1181 		return;
1182 	}
1183 
1184 	fprintf(stderr, "Unrecognized cmsg option %s\n", type);
1185 	exit(1);
1186 }
1187 
1188 static void parse_setsock_options(const char *name)
1189 {
1190 	char *next = strchr(name, ',');
1191 	unsigned int len = 0;
1192 
1193 	if (next) {
1194 		parse_setsock_options(next + 1);
1195 		len = next - name;
1196 	} else {
1197 		len = strlen(name);
1198 	}
1199 
1200 	if (strncmp(name, "TRANSPARENT", len) == 0) {
1201 		cfg_sockopt_types.transparent = 1;
1202 		return;
1203 	}
1204 
1205 	if (strncmp(name, "MPTFO", len) == 0) {
1206 		cfg_sockopt_types.mptfo = 1;
1207 		return;
1208 	}
1209 
1210 	fprintf(stderr, "Unrecognized setsockopt option %s\n", name);
1211 	exit(1);
1212 }
1213 
1214 void xdisconnect(int fd, int addrlen)
1215 {
1216 	struct sockaddr_storage empty;
1217 	int msec_sleep = 10;
1218 	int queued = 1;
1219 	int i;
1220 
1221 	shutdown(fd, SHUT_WR);
1222 
1223 	/* while until the pending data is completely flushed, the later
1224 	 * disconnect will bypass/ignore/drop any pending data.
1225 	 */
1226 	for (i = 0; ; i += msec_sleep) {
1227 		if (ioctl(fd, SIOCOUTQ, &queued) < 0)
1228 			xerror("can't query out socket queue: %d", errno);
1229 
1230 		if (!queued)
1231 			break;
1232 
1233 		if (i > poll_timeout)
1234 			xerror("timeout while waiting for spool to complete");
1235 		usleep(msec_sleep * 1000);
1236 	}
1237 
1238 	memset(&empty, 0, sizeof(empty));
1239 	empty.ss_family = AF_UNSPEC;
1240 	if (connect(fd, (struct sockaddr *)&empty, addrlen) < 0)
1241 		xerror("can't disconnect: %d", errno);
1242 }
1243 
1244 int main_loop(void)
1245 {
1246 	int fd = 0, ret, fd_in = 0;
1247 	struct addrinfo *peer;
1248 	struct wstate winfo;
1249 
1250 	if (cfg_input && cfg_sockopt_types.mptfo) {
1251 		fd_in = open(cfg_input, O_RDONLY);
1252 		if (fd < 0)
1253 			xerror("can't open %s:%d", cfg_input, errno);
1254 	}
1255 
1256 	memset(&winfo, 0, sizeof(winfo));
1257 	fd = sock_connect_mptcp(cfg_host, cfg_port, cfg_sock_proto, &peer, fd_in, &winfo);
1258 	if (fd < 0)
1259 		return 2;
1260 
1261 again:
1262 	check_getpeername_connect(fd);
1263 
1264 	SOCK_TEST_TCPULP(fd, cfg_sock_proto);
1265 
1266 	if (cfg_rcvbuf)
1267 		set_rcvbuf(fd, cfg_rcvbuf);
1268 	if (cfg_sndbuf)
1269 		set_sndbuf(fd, cfg_sndbuf);
1270 	if (cfg_cmsg_types.cmsg_enabled)
1271 		apply_cmsg_types(fd, &cfg_cmsg_types);
1272 
1273 	if (cfg_input && !cfg_sockopt_types.mptfo) {
1274 		fd_in = open(cfg_input, O_RDONLY);
1275 		if (fd < 0)
1276 			xerror("can't open %s:%d", cfg_input, errno);
1277 	}
1278 
1279 	ret = copyfd_io(fd_in, fd, 1, 0, &winfo);
1280 	if (ret)
1281 		return ret;
1282 
1283 	if (cfg_truncate > 0) {
1284 		xdisconnect(fd, peer->ai_addrlen);
1285 	} else if (--cfg_repeat > 0) {
1286 		xdisconnect(fd, peer->ai_addrlen);
1287 
1288 		/* the socket could be unblocking at this point, we need the
1289 		 * connect to be blocking
1290 		 */
1291 		set_nonblock(fd, false);
1292 		if (connect(fd, peer->ai_addr, peer->ai_addrlen))
1293 			xerror("can't reconnect: %d", errno);
1294 		if (cfg_input)
1295 			close(fd_in);
1296 		memset(&winfo, 0, sizeof(winfo));
1297 		goto again;
1298 	} else {
1299 		close(fd);
1300 	}
1301 
1302 	return 0;
1303 }
1304 
1305 int parse_proto(const char *proto)
1306 {
1307 	if (!strcasecmp(proto, "MPTCP"))
1308 		return IPPROTO_MPTCP;
1309 	if (!strcasecmp(proto, "TCP"))
1310 		return IPPROTO_TCP;
1311 
1312 	fprintf(stderr, "Unknown protocol: %s\n.", proto);
1313 	die_usage();
1314 
1315 	/* silence compiler warning */
1316 	return 0;
1317 }
1318 
1319 int parse_mode(const char *mode)
1320 {
1321 	if (!strcasecmp(mode, "poll"))
1322 		return CFG_MODE_POLL;
1323 	if (!strcasecmp(mode, "mmap"))
1324 		return CFG_MODE_MMAP;
1325 	if (!strcasecmp(mode, "sendfile"))
1326 		return CFG_MODE_SENDFILE;
1327 
1328 	fprintf(stderr, "Unknown test mode: %s\n", mode);
1329 	fprintf(stderr, "Supported modes are:\n");
1330 	fprintf(stderr, "\t\t\"poll\" - interleaved read/write using poll()\n");
1331 	fprintf(stderr, "\t\t\"mmap\" - send entire input file (mmap+write), then read response (-l will read input first)\n");
1332 	fprintf(stderr, "\t\t\"sendfile\" - send entire input file (sendfile), then read response (-l will read input first)\n");
1333 
1334 	die_usage();
1335 
1336 	/* silence compiler warning */
1337 	return 0;
1338 }
1339 
1340 int parse_peek(const char *mode)
1341 {
1342 	if (!strcasecmp(mode, "saveWithPeek"))
1343 		return CFG_WITH_PEEK;
1344 	if (!strcasecmp(mode, "saveAfterPeek"))
1345 		return CFG_AFTER_PEEK;
1346 
1347 	fprintf(stderr, "Unknown: %s\n", mode);
1348 	fprintf(stderr, "Supported MSG_PEEK mode are:\n");
1349 	fprintf(stderr,
1350 		"\t\t\"saveWithPeek\" - recv data with flags 'MSG_PEEK' and save the peek data into file\n");
1351 	fprintf(stderr,
1352 		"\t\t\"saveAfterPeek\" - read and save data into file after recv with flags 'MSG_PEEK'\n");
1353 
1354 	die_usage();
1355 
1356 	/* silence compiler warning */
1357 	return 0;
1358 }
1359 
1360 static int parse_int(const char *size)
1361 {
1362 	unsigned long s;
1363 
1364 	errno = 0;
1365 
1366 	s = strtoul(size, NULL, 0);
1367 
1368 	if (errno) {
1369 		fprintf(stderr, "Invalid sndbuf size %s (%s)\n",
1370 			size, strerror(errno));
1371 		die_usage();
1372 	}
1373 
1374 	if (s > INT_MAX) {
1375 		fprintf(stderr, "Invalid sndbuf size %s (%s)\n",
1376 			size, strerror(ERANGE));
1377 		die_usage();
1378 	}
1379 
1380 	return (int)s;
1381 }
1382 
1383 static void parse_opts(int argc, char **argv)
1384 {
1385 	int c;
1386 
1387 	while ((c = getopt(argc, argv, "6c:f:hi:I:jlm:M:o:p:P:r:R:s:S:t:T:w:")) != -1) {
1388 		switch (c) {
1389 		case 'f':
1390 			cfg_truncate = atoi(optarg);
1391 
1392 			/* when receiving a fastclose, ignore PIPE signals and
1393 			 * all the I/O errors later in the code
1394 			 */
1395 			if (cfg_truncate < 0) {
1396 				cfg_rcv_trunc = true;
1397 				signal(SIGPIPE, handle_signal);
1398 			}
1399 			break;
1400 		case 'j':
1401 			cfg_join = true;
1402 			cfg_mode = CFG_MODE_POLL;
1403 			break;
1404 		case 'r':
1405 			cfg_remove = true;
1406 			cfg_mode = CFG_MODE_POLL;
1407 			cfg_wait = 400000;
1408 			cfg_do_w = atoi(optarg);
1409 			if (cfg_do_w <= 0)
1410 				cfg_do_w = 50;
1411 			break;
1412 		case 'i':
1413 			cfg_input = optarg;
1414 			break;
1415 		case 'I':
1416 			cfg_repeat = atoi(optarg);
1417 			break;
1418 		case 'l':
1419 			listen_mode = true;
1420 			break;
1421 		case 'p':
1422 			cfg_port = optarg;
1423 			break;
1424 		case 's':
1425 			cfg_sock_proto = parse_proto(optarg);
1426 			break;
1427 		case 'h':
1428 			die_usage();
1429 			break;
1430 		case '6':
1431 			pf = AF_INET6;
1432 			break;
1433 		case 't':
1434 			poll_timeout = atoi(optarg) * 1000;
1435 			if (poll_timeout <= 0)
1436 				poll_timeout = -1;
1437 			break;
1438 		case 'T':
1439 			cfg_time = atoi(optarg);
1440 			break;
1441 		case 'm':
1442 			cfg_mode = parse_mode(optarg);
1443 			break;
1444 		case 'S':
1445 			cfg_sndbuf = parse_int(optarg);
1446 			break;
1447 		case 'R':
1448 			cfg_rcvbuf = parse_int(optarg);
1449 			break;
1450 		case 'w':
1451 			cfg_wait = atoi(optarg)*1000000;
1452 			break;
1453 		case 'M':
1454 			cfg_mark = strtol(optarg, NULL, 0);
1455 			break;
1456 		case 'P':
1457 			cfg_peek = parse_peek(optarg);
1458 			break;
1459 		case 'c':
1460 			parse_cmsg_types(optarg);
1461 			break;
1462 		case 'o':
1463 			parse_setsock_options(optarg);
1464 			break;
1465 		}
1466 	}
1467 
1468 	if (optind + 1 != argc)
1469 		die_usage();
1470 	cfg_host = argv[optind];
1471 
1472 	if (strchr(cfg_host, ':'))
1473 		pf = AF_INET6;
1474 }
1475 
1476 int main(int argc, char *argv[])
1477 {
1478 	init_rng();
1479 
1480 	signal(SIGUSR1, handle_signal);
1481 	parse_opts(argc, argv);
1482 
1483 	if (listen_mode) {
1484 		int fd = sock_listen_mptcp(cfg_host, cfg_port);
1485 
1486 		if (fd < 0)
1487 			return 1;
1488 
1489 		if (cfg_rcvbuf)
1490 			set_rcvbuf(fd, cfg_rcvbuf);
1491 		if (cfg_sndbuf)
1492 			set_sndbuf(fd, cfg_sndbuf);
1493 		if (cfg_mark)
1494 			set_mark(fd, cfg_mark);
1495 		if (cfg_cmsg_types.cmsg_enabled)
1496 			apply_cmsg_types(fd, &cfg_cmsg_types);
1497 
1498 		return main_loop_s(fd);
1499 	}
1500 
1501 	return main_loop();
1502 }
1503