xref: /linux/tools/testing/selftests/net/mptcp/mptcp_connect.c (revision 730444f094b12052916ebd7e14fe57bc3d47bf38)
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 <arpa/inet.h>
29 
30 #include <netdb.h>
31 #include <netinet/in.h>
32 
33 #include <linux/tcp.h>
34 #include <linux/time_types.h>
35 #include <linux/sockios.h>
36 #include <linux/compiler.h>
37 
38 extern int optind;
39 
40 #ifndef IPPROTO_MPTCP
41 #define IPPROTO_MPTCP 262
42 #endif
43 #ifndef TCP_ULP
44 #define TCP_ULP 31
45 #endif
46 
47 static int  poll_timeout = 10 * 1000;
48 static bool listen_mode;
49 static bool quit;
50 
51 enum cfg_mode {
52 	CFG_MODE_POLL,
53 	CFG_MODE_MMAP,
54 	CFG_MODE_SENDFILE,
55 	CFG_MODE_SPLICE,
56 };
57 
58 enum cfg_peek {
59 	CFG_NONE_PEEK,
60 	CFG_WITH_PEEK,
61 	CFG_AFTER_PEEK,
62 };
63 
64 static enum cfg_mode cfg_mode = CFG_MODE_POLL;
65 static enum cfg_peek cfg_peek = CFG_NONE_PEEK;
66 static const char *cfg_host;
67 static const char *cfg_port	= "12000";
68 static int cfg_sock_proto	= IPPROTO_MPTCP;
69 static int pf = AF_INET;
70 static int cfg_sndbuf;
71 static int cfg_rcvbuf;
72 static bool cfg_join;
73 static bool cfg_remove;
74 static unsigned int cfg_time;
75 static unsigned int cfg_do_w;
76 static int cfg_wait;
77 static uint32_t cfg_mark;
78 static char *cfg_input;
79 static int cfg_repeat = 1;
80 static int cfg_truncate;
81 static int cfg_rcv_trunc;
82 
83 struct cfg_cmsg_types {
84 	unsigned int cmsg_enabled:1;
85 	unsigned int timestampns:1;
86 	unsigned int tcp_inq:1;
87 };
88 
89 struct cfg_sockopt_types {
90 	unsigned int transparent:1;
91 	unsigned int mptfo:1;
92 };
93 
94 struct tcp_inq_state {
95 	unsigned int last;
96 	bool expect_eof;
97 };
98 
99 struct wstate {
100 	char buf[8192];
101 	unsigned int len;
102 	unsigned int off;
103 	unsigned int total_len;
104 };
105 
106 static struct tcp_inq_state tcp_inq;
107 
108 static struct cfg_cmsg_types cfg_cmsg_types;
109 static struct cfg_sockopt_types cfg_sockopt_types;
110 
111 static void die_usage(void)
112 {
113 	fprintf(stderr, "Usage: mptcp_connect [-6] [-c cmsg] [-f offset] [-i file] [-I num] [-j] [-l] "
114 		"[-m mode] [-M mark] [-o option] [-p port] [-P mode] [-r num] [-R num] "
115 		"[-s MPTCP|TCP] [-S num] [-t num] [-T num] [-w sec] connect_address\n");
116 	fprintf(stderr, "\t-6 use ipv6\n");
117 	fprintf(stderr, "\t-c cmsg -- test cmsg type <cmsg>\n");
118 	fprintf(stderr, "\t-f offset -- stop the I/O after receiving and sending the specified amount "
119 		"of bytes. If there are unread bytes in the receive queue, that will cause a MPTCP "
120 		"fastclose at close/shutdown. If offset is negative, expect the peer to close before "
121 		"all the local data as been sent, thus toleration errors on write and EPIPE signals\n");
122 	fprintf(stderr, "\t-i file -- read the data to send from the given file instead of stdin");
123 	fprintf(stderr, "\t-I num -- repeat the transfer 'num' times. In listen mode accepts num "
124 		"incoming connections, in client mode, disconnect and reconnect to the server\n");
125 	fprintf(stderr, "\t-j     -- add additional sleep at connection start and tear down "
126 		"-- for MPJ tests\n");
127 	fprintf(stderr, "\t-l     -- listens mode, accepts incoming connection\n");
128 	fprintf(stderr, "\t-m [poll|mmap|sendfile|splice] -- use poll(default)/mmap+write/sendfile/splice\n");
129 	fprintf(stderr, "\t-M mark -- set socket packet mark\n");
130 	fprintf(stderr, "\t-o option -- test sockopt <option>\n");
131 	fprintf(stderr, "\t-p num -- use port num\n");
132 	fprintf(stderr,
133 		"\t-P [saveWithPeek|saveAfterPeek] -- save data with/after MSG_PEEK form tcp socket\n");
134 	fprintf(stderr, "\t-r num -- enable slow mode, limiting each write to num bytes "
135 		"-- for remove addr tests\n");
136 	fprintf(stderr, "\t-R num -- set SO_RCVBUF to num\n");
137 	fprintf(stderr, "\t-s [MPTCP|TCP] -- use mptcp(default) or tcp sockets\n");
138 	fprintf(stderr, "\t-S num -- set SO_SNDBUF to num\n");
139 	fprintf(stderr, "\t-t num -- set poll timeout to num\n");
140 	fprintf(stderr, "\t-T num -- set expected runtime to num ms\n");
141 	fprintf(stderr, "\t-w num -- wait num sec before closing the socket\n");
142 	exit(1);
143 }
144 
145 static void __noreturn xerror(const char *fmt, ...)
146 {
147 	va_list ap;
148 
149 	va_start(ap, fmt);
150 	vfprintf(stderr, fmt, ap);
151 	va_end(ap);
152 	exit(1);
153 }
154 
155 static void handle_signal(int nr)
156 {
157 	quit = true;
158 }
159 
160 static const char *getxinfo_strerr(int err)
161 {
162 	if (err == EAI_SYSTEM)
163 		return strerror(errno);
164 
165 	return gai_strerror(err);
166 }
167 
168 static void xgetnameinfo(const struct sockaddr *addr, socklen_t addrlen,
169 			 char *host, socklen_t hostlen,
170 			 char *serv, socklen_t servlen)
171 {
172 	int flags = NI_NUMERICHOST | NI_NUMERICSERV;
173 	int err = getnameinfo(addr, addrlen, host, hostlen, serv, servlen,
174 			      flags);
175 
176 	if (err) {
177 		const char *errstr = getxinfo_strerr(err);
178 
179 		fprintf(stderr, "Fatal: getnameinfo: %s\n", errstr);
180 		exit(1);
181 	}
182 }
183 
184 static void xgetaddrinfo(const char *node, const char *service,
185 			 struct addrinfo *hints,
186 			 struct addrinfo **res)
187 {
188 	int err;
189 
190 again:
191 	err = getaddrinfo(node, service, hints, res);
192 	if (err) {
193 		const char *errstr;
194 
195 		/* glibc starts to support MPTCP since v2.42.
196 		 * For older versions, use IPPROTO_TCP to resolve,
197 		 * and use TCP/MPTCP to create socket.
198 		 * Link: https://sourceware.org/git/?p=glibc.git;a=commit;h=a8e9022e0f82
199 		 */
200 		if (err == EAI_SOCKTYPE) {
201 			hints->ai_protocol = IPPROTO_TCP;
202 			goto again;
203 		}
204 
205 		errstr = getxinfo_strerr(err);
206 
207 		fprintf(stderr, "Fatal: getaddrinfo(%s:%s): %s\n",
208 			node ? node : "", service ? service : "", errstr);
209 		exit(1);
210 	}
211 }
212 
213 static void set_rcvbuf(int fd, unsigned int size)
214 {
215 	int err;
216 
217 	err = setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &size, sizeof(size));
218 	if (err) {
219 		perror("set SO_RCVBUF");
220 		exit(1);
221 	}
222 }
223 
224 static void set_sndbuf(int fd, unsigned int size)
225 {
226 	int err;
227 
228 	err = setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &size, sizeof(size));
229 	if (err) {
230 		perror("set SO_SNDBUF");
231 		exit(1);
232 	}
233 }
234 
235 static void set_mark(int fd, uint32_t mark)
236 {
237 	int err;
238 
239 	err = setsockopt(fd, SOL_SOCKET, SO_MARK, &mark, sizeof(mark));
240 	if (err) {
241 		perror("set SO_MARK");
242 		exit(1);
243 	}
244 }
245 
246 static void set_transparent(int fd, int pf)
247 {
248 	int one = 1;
249 
250 	switch (pf) {
251 	case AF_INET:
252 		if (-1 == setsockopt(fd, SOL_IP, IP_TRANSPARENT, &one, sizeof(one)))
253 			perror("IP_TRANSPARENT");
254 		break;
255 	case AF_INET6:
256 		if (-1 == setsockopt(fd, IPPROTO_IPV6, IPV6_TRANSPARENT, &one, sizeof(one)))
257 			perror("IPV6_TRANSPARENT");
258 		break;
259 	}
260 }
261 
262 static void set_mptfo(int fd)
263 {
264 	int qlen = 25;
265 
266 	if (setsockopt(fd, IPPROTO_TCP, TCP_FASTOPEN, &qlen, sizeof(qlen)) == -1)
267 		perror("TCP_FASTOPEN");
268 }
269 
270 static int do_ulp_so(int sock, const char *name)
271 {
272 	return setsockopt(sock, IPPROTO_TCP, TCP_ULP, name, strlen(name));
273 }
274 
275 #define X(m)	xerror("%s:%u: %s: failed for proto %d at line %u", __FILE__, __LINE__, (m), proto, line)
276 static void sock_test_tcpulp(int sock, int proto, unsigned int line)
277 {
278 	socklen_t buflen = 8;
279 	char buf[8] = "";
280 	int ret = getsockopt(sock, IPPROTO_TCP, TCP_ULP, buf, &buflen);
281 
282 	if (ret != 0)
283 		X("getsockopt");
284 
285 	if (buflen > 0) {
286 		if (strcmp(buf, "mptcp") != 0)
287 			xerror("unexpected ULP '%s' for proto %d at line %u", buf, proto, line);
288 		ret = do_ulp_so(sock, "tls");
289 		if (ret == 0)
290 			X("setsockopt");
291 	} else if (proto == IPPROTO_MPTCP) {
292 		ret = do_ulp_so(sock, "tls");
293 		if (ret != -1)
294 			X("setsockopt");
295 	}
296 
297 	ret = do_ulp_so(sock, "mptcp");
298 	if (ret != -1)
299 		X("setsockopt");
300 
301 #undef X
302 }
303 
304 #define SOCK_TEST_TCPULP(s, p) sock_test_tcpulp((s), (p), __LINE__)
305 
306 static int sock_listen_mptcp(const char * const listenaddr,
307 			     const char * const port)
308 {
309 	int sock = -1;
310 	struct addrinfo hints = {
311 		.ai_protocol = IPPROTO_MPTCP,
312 		.ai_socktype = SOCK_STREAM,
313 		.ai_flags = AI_PASSIVE | AI_NUMERICHOST
314 	};
315 
316 	hints.ai_family = pf;
317 
318 	struct addrinfo *a, *addr;
319 	int one = 1;
320 
321 	xgetaddrinfo(listenaddr, port, &hints, &addr);
322 	hints.ai_family = pf;
323 
324 	for (a = addr; a; a = a->ai_next) {
325 		sock = socket(a->ai_family, a->ai_socktype, cfg_sock_proto);
326 		if (sock < 0)
327 			continue;
328 
329 		SOCK_TEST_TCPULP(sock, cfg_sock_proto);
330 
331 		if (-1 == setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, &one,
332 				     sizeof(one)))
333 			perror("setsockopt");
334 
335 		if (cfg_sockopt_types.transparent)
336 			set_transparent(sock, pf);
337 
338 		if (cfg_sockopt_types.mptfo)
339 			set_mptfo(sock);
340 
341 		if (bind(sock, a->ai_addr, a->ai_addrlen) == 0)
342 			break; /* success */
343 
344 		perror("bind");
345 		close(sock);
346 		sock = -1;
347 	}
348 
349 	freeaddrinfo(addr);
350 
351 	if (sock < 0) {
352 		fprintf(stderr, "Could not create listen socket\n");
353 		return sock;
354 	}
355 
356 	SOCK_TEST_TCPULP(sock, cfg_sock_proto);
357 
358 	if (listen(sock, 20)) {
359 		perror("listen");
360 		close(sock);
361 		return -1;
362 	}
363 
364 	SOCK_TEST_TCPULP(sock, cfg_sock_proto);
365 
366 	return sock;
367 }
368 
369 static int sock_connect_mptcp(const char * const remoteaddr,
370 			      const char * const port, int proto,
371 			      struct addrinfo **peer,
372 			      int infd, struct wstate *winfo)
373 {
374 	struct addrinfo hints = {
375 		.ai_protocol = IPPROTO_MPTCP,
376 		.ai_socktype = SOCK_STREAM,
377 	};
378 	struct addrinfo *a, *addr;
379 	int syn_copied = 0;
380 	int sock = -1;
381 
382 	hints.ai_family = pf;
383 
384 	/* Keep the resolved address alive for the whole execution: it is
385 	 * used again when reconnecting, and will be released at exit time.
386 	 */
387 	xgetaddrinfo(remoteaddr, port, &hints, &addr);
388 	for (a = addr; a; a = a->ai_next) {
389 		sock = socket(a->ai_family, a->ai_socktype, proto);
390 		if (sock < 0) {
391 			perror("socket");
392 			continue;
393 		}
394 
395 		SOCK_TEST_TCPULP(sock, proto);
396 
397 		if (cfg_mark)
398 			set_mark(sock, cfg_mark);
399 
400 		if (cfg_sockopt_types.mptfo) {
401 			if (!winfo->total_len)
402 				winfo->total_len = winfo->len = read(infd, winfo->buf,
403 								     sizeof(winfo->buf));
404 
405 			syn_copied = sendto(sock, winfo->buf, winfo->len, MSG_FASTOPEN,
406 					    a->ai_addr, a->ai_addrlen);
407 			if (syn_copied >= 0) {
408 				winfo->off = syn_copied;
409 				winfo->len -= syn_copied;
410 				*peer = a;
411 				break; /* success */
412 			}
413 			perror("sendto()");
414 		} else {
415 			if (connect(sock, a->ai_addr, a->ai_addrlen) == 0) {
416 				*peer = a;
417 				break; /* success */
418 			}
419 			perror("connect()");
420 		}
421 
422 		/* error */
423 		close(sock);
424 		sock = -1;
425 	}
426 
427 	if (sock != -1)
428 		SOCK_TEST_TCPULP(sock, proto);
429 	return sock;
430 }
431 
432 static size_t do_rnd_write(const int fd, char *buf, const size_t len)
433 {
434 	static bool first = true;
435 	unsigned int do_w;
436 	ssize_t bw;
437 
438 	do_w = rand() & 0xffff;
439 	if (do_w == 0 || do_w > len)
440 		do_w = len;
441 
442 	if (cfg_join && first && do_w > 100)
443 		do_w = 100;
444 
445 	if (cfg_remove && do_w > cfg_do_w)
446 		do_w = cfg_do_w;
447 
448 	bw = write(fd, buf, do_w);
449 	if (bw < 0)
450 		return bw;
451 
452 	/* let the join handshake complete, before going on */
453 	if (cfg_join && first) {
454 		usleep(200000);
455 		first = false;
456 	}
457 
458 	if (cfg_remove)
459 		usleep(200000);
460 
461 	return bw;
462 }
463 
464 static size_t do_write(const int fd, char *buf, const size_t len)
465 {
466 	size_t offset = 0;
467 
468 	while (offset < len) {
469 		size_t written;
470 		ssize_t bw;
471 
472 		bw = write(fd, buf + offset, len - offset);
473 		if (bw < 0) {
474 			perror("write");
475 			return 0;
476 		}
477 
478 		written = (size_t)bw;
479 		offset += written;
480 	}
481 
482 	return offset;
483 }
484 
485 static void process_cmsg(struct msghdr *msgh)
486 {
487 	struct __kernel_timespec ts;
488 	bool inq_found = false;
489 	bool ts_found = false;
490 	unsigned int inq = 0;
491 	struct cmsghdr *cmsg;
492 
493 	for (cmsg = CMSG_FIRSTHDR(msgh); cmsg ; cmsg = CMSG_NXTHDR(msgh, cmsg)) {
494 		if (cmsg->cmsg_level == SOL_SOCKET && cmsg->cmsg_type == SO_TIMESTAMPNS_NEW) {
495 			memcpy(&ts, CMSG_DATA(cmsg), sizeof(ts));
496 			ts_found = true;
497 			continue;
498 		}
499 		if (cmsg->cmsg_level == IPPROTO_TCP && cmsg->cmsg_type == TCP_CM_INQ) {
500 			memcpy(&inq, CMSG_DATA(cmsg), sizeof(inq));
501 			inq_found = true;
502 			continue;
503 		}
504 
505 	}
506 
507 	if (cfg_cmsg_types.timestampns) {
508 		if (!ts_found)
509 			xerror("TIMESTAMPNS not present\n");
510 	}
511 
512 	if (cfg_cmsg_types.tcp_inq) {
513 		if (!inq_found)
514 			xerror("TCP_INQ not present\n");
515 
516 		if (inq > 1024)
517 			xerror("tcp_inq %u is larger than one kbyte\n", inq);
518 		tcp_inq.last = inq;
519 	}
520 }
521 
522 static ssize_t do_recvmsg_cmsg(const int fd, char *buf, const size_t len)
523 {
524 	char msg_buf[8192];
525 	struct iovec iov = {
526 		.iov_base = buf,
527 		.iov_len = len,
528 	};
529 	struct msghdr msg = {
530 		.msg_iov = &iov,
531 		.msg_iovlen = 1,
532 		.msg_control = msg_buf,
533 		.msg_controllen = sizeof(msg_buf),
534 	};
535 	int flags = 0;
536 	unsigned int last_hint = tcp_inq.last;
537 	int ret = recvmsg(fd, &msg, flags);
538 
539 	if (ret <= 0) {
540 		if (ret == 0 && tcp_inq.expect_eof)
541 			return ret;
542 
543 		if (ret == 0 && cfg_cmsg_types.tcp_inq)
544 			if (last_hint != 1 && last_hint != 0)
545 				xerror("EOF but last tcp_inq hint was %u\n", last_hint);
546 
547 		return ret;
548 	}
549 
550 	if (tcp_inq.expect_eof)
551 		xerror("expected EOF, last_hint %u, now %u\n",
552 		       last_hint, tcp_inq.last);
553 
554 	if (msg.msg_controllen && !cfg_cmsg_types.cmsg_enabled)
555 		xerror("got %lu bytes of cmsg data, expected 0\n",
556 		       (unsigned long)msg.msg_controllen);
557 
558 	if (msg.msg_controllen == 0 && cfg_cmsg_types.cmsg_enabled)
559 		xerror("%s\n", "got no cmsg data");
560 
561 	if (msg.msg_controllen)
562 		process_cmsg(&msg);
563 
564 	if (cfg_cmsg_types.tcp_inq) {
565 		if ((size_t)ret < len && last_hint > (unsigned int)ret) {
566 			if (ret + 1 != (int)last_hint) {
567 				int next = read(fd, msg_buf, sizeof(msg_buf));
568 
569 				xerror("read %u of %u, last_hint was %u tcp_inq hint now %u next_read returned %d/%m\n",
570 				       ret, (unsigned int)len, last_hint, tcp_inq.last, next);
571 			} else {
572 				tcp_inq.expect_eof = true;
573 			}
574 		}
575 	}
576 
577 	return ret;
578 }
579 
580 static ssize_t do_rnd_read(const int fd, char *buf, const size_t len)
581 {
582 	int ret = 0;
583 	char tmp[16384];
584 	size_t cap = rand();
585 
586 	cap &= 0xffff;
587 
588 	if (cap == 0)
589 		cap = 1;
590 	else if (cap > len)
591 		cap = len;
592 
593 	if (cfg_peek == CFG_WITH_PEEK) {
594 		ret = recv(fd, buf, cap, MSG_PEEK);
595 		ret = (ret < 0) ? ret : read(fd, tmp, ret);
596 	} else if (cfg_peek == CFG_AFTER_PEEK) {
597 		ret = recv(fd, buf, cap, MSG_PEEK);
598 		ret = (ret < 0) ? ret : read(fd, buf, cap);
599 	} else if (cfg_cmsg_types.cmsg_enabled) {
600 		ret = do_recvmsg_cmsg(fd, buf, cap);
601 	} else {
602 		ret = read(fd, buf, cap);
603 	}
604 
605 	return ret;
606 }
607 
608 static void set_nonblock(int fd, bool nonblock)
609 {
610 	int flags = fcntl(fd, F_GETFL);
611 
612 	if (flags == -1)
613 		return;
614 
615 	if (nonblock)
616 		fcntl(fd, F_SETFL, flags | O_NONBLOCK);
617 	else
618 		fcntl(fd, F_SETFL, flags & ~O_NONBLOCK);
619 }
620 
621 static void shut_wr(int fd)
622 {
623 	/* Close our write side, ev. give some time
624 	 * for address notification and/or checking
625 	 * the current status
626 	 */
627 	if (cfg_wait)
628 		usleep(cfg_wait);
629 
630 	shutdown(fd, SHUT_WR);
631 }
632 
633 static int copyfd_io_poll(int infd, int peerfd, int outfd,
634 			  bool *in_closed_after_out, struct wstate *winfo)
635 {
636 	struct pollfd fds = {
637 		.fd = peerfd,
638 		.events = POLLIN | POLLOUT,
639 	};
640 	unsigned int total_wlen = 0, total_rlen = 0;
641 
642 	set_nonblock(peerfd, true);
643 
644 	for (;;) {
645 		char rbuf[8192];
646 		ssize_t len;
647 
648 		if (fds.events == 0 || quit)
649 			break;
650 
651 		switch (poll(&fds, 1, poll_timeout)) {
652 		case -1:
653 			if (errno == EINTR)
654 				continue;
655 			perror("poll");
656 			return 1;
657 		case 0:
658 			fprintf(stderr, "%s: poll timed out (events: "
659 				"POLLIN %u, POLLOUT %u)\n", __func__,
660 				fds.events & POLLIN, fds.events & POLLOUT);
661 			return 2;
662 		}
663 
664 		if (fds.revents & POLLIN) {
665 			ssize_t rb = sizeof(rbuf);
666 
667 			/* limit the total amount of read data to the trunc value*/
668 			if (cfg_truncate > 0) {
669 				if (rb + total_rlen > cfg_truncate)
670 					rb = cfg_truncate - total_rlen;
671 				len = read(peerfd, rbuf, rb);
672 			} else {
673 				len = do_rnd_read(peerfd, rbuf, sizeof(rbuf));
674 			}
675 			if (len == 0) {
676 				/* no more data to receive:
677 				 * peer has closed its write side
678 				 */
679 				fds.events &= ~POLLIN;
680 
681 				if ((fds.events & POLLOUT) == 0) {
682 					*in_closed_after_out = true;
683 					/* and nothing more to send */
684 					break;
685 				}
686 
687 			/* Else, still have data to transmit */
688 			} else if (len < 0) {
689 				if (cfg_rcv_trunc)
690 					return 0;
691 				perror("read");
692 				return 3;
693 			}
694 
695 			total_rlen += len;
696 			do_write(outfd, rbuf, len);
697 		}
698 
699 		if (fds.revents & POLLOUT) {
700 			if (winfo->len == 0) {
701 				winfo->off = 0;
702 				winfo->len = read(infd, winfo->buf, sizeof(winfo->buf));
703 			}
704 
705 			if (winfo->len > 0) {
706 				ssize_t bw;
707 
708 				/* limit the total amount of written data to the trunc value */
709 				if (cfg_truncate > 0 && winfo->len + total_wlen > cfg_truncate)
710 					winfo->len = cfg_truncate - total_wlen;
711 
712 				bw = do_rnd_write(peerfd, winfo->buf + winfo->off, winfo->len);
713 				if (bw < 0) {
714 					/* expected reset, continue to read */
715 					if (cfg_rcv_trunc &&
716 					    (errno == ECONNRESET ||
717 					     errno == EPIPE)) {
718 						fds.events &= ~POLLOUT;
719 						continue;
720 					}
721 
722 					perror("write");
723 					return 111;
724 				}
725 
726 				winfo->off += bw;
727 				winfo->len -= bw;
728 				total_wlen += bw;
729 			} else if (winfo->len == 0) {
730 				/* We have no more data to send. */
731 				fds.events &= ~POLLOUT;
732 
733 				if ((fds.events & POLLIN) == 0)
734 					/* ... and peer also closed already */
735 					break;
736 
737 				shut_wr(peerfd);
738 			} else {
739 				if (errno == EINTR)
740 					continue;
741 				perror("read");
742 				return 4;
743 			}
744 		}
745 
746 		if (fds.revents & (POLLERR | POLLNVAL)) {
747 			if (cfg_rcv_trunc) {
748 				fds.events &= ~(POLLERR | POLLNVAL);
749 				continue;
750 			}
751 			fprintf(stderr, "Unexpected revents: "
752 				"POLLERR/POLLNVAL(%x)\n", fds.revents);
753 			return 5;
754 		}
755 
756 		if (cfg_truncate > 0 && total_wlen >= cfg_truncate &&
757 		    total_rlen >= cfg_truncate)
758 			break;
759 	}
760 
761 	/* leave some time for late join/announce */
762 	if (cfg_remove && !quit)
763 		usleep(cfg_wait);
764 
765 	return 0;
766 }
767 
768 static int do_recvfile(int infd, int outfd)
769 {
770 	ssize_t r;
771 
772 	do {
773 		char buf[16384];
774 
775 		r = do_rnd_read(infd, buf, sizeof(buf));
776 		if (r > 0) {
777 			if (write(outfd, buf, r) != r)
778 				break;
779 		} else if (r < 0) {
780 			perror("read");
781 		}
782 	} while (r > 0);
783 
784 	return (int)r;
785 }
786 
787 static int spool_buf(int fd, struct wstate *winfo)
788 {
789 	while (winfo->len) {
790 		int ret = write(fd, winfo->buf + winfo->off, winfo->len);
791 
792 		if (ret < 0) {
793 			perror("write");
794 			return 4;
795 		}
796 		winfo->off += ret;
797 		winfo->len -= ret;
798 	}
799 	return 0;
800 }
801 
802 static int do_mmap(int infd, int outfd, unsigned int size,
803 		   struct wstate *winfo)
804 {
805 	char *inbuf = mmap(NULL, size, PROT_READ, MAP_SHARED, infd, 0);
806 	ssize_t ret = 0, off = winfo->total_len;
807 	size_t rem;
808 
809 	if (inbuf == MAP_FAILED) {
810 		perror("mmap");
811 		return 1;
812 	}
813 
814 	ret = spool_buf(outfd, winfo);
815 	if (ret < 0)
816 		return ret;
817 
818 	rem = size - winfo->total_len;
819 
820 	while (rem > 0) {
821 		ret = write(outfd, inbuf + off, rem);
822 
823 		if (ret < 0) {
824 			perror("write");
825 			break;
826 		}
827 
828 		off += ret;
829 		rem -= ret;
830 	}
831 
832 	munmap(inbuf, size);
833 	return rem;
834 }
835 
836 static int get_infd_size(int fd)
837 {
838 	struct stat sb;
839 	ssize_t count;
840 	int err;
841 
842 	err = fstat(fd, &sb);
843 	if (err < 0) {
844 		perror("fstat");
845 		return -1;
846 	}
847 
848 	if ((sb.st_mode & S_IFMT) != S_IFREG) {
849 		fprintf(stderr, "%s: stdin is not a regular file\n", __func__);
850 		return -2;
851 	}
852 
853 	count = sb.st_size;
854 	if (count > INT_MAX) {
855 		fprintf(stderr, "File too large: %zu\n", count);
856 		return -3;
857 	}
858 
859 	return (int)count;
860 }
861 
862 static int do_sendfile(int infd, int outfd, unsigned int count,
863 		       struct wstate *winfo)
864 {
865 	int ret = spool_buf(outfd, winfo);
866 
867 	if (ret < 0)
868 		return ret;
869 
870 	count -= winfo->total_len;
871 
872 	while (count > 0) {
873 		ssize_t r;
874 
875 		r = sendfile(outfd, infd, NULL, count);
876 		if (r < 0) {
877 			perror("sendfile");
878 			return 3;
879 		}
880 
881 		count -= r;
882 	}
883 
884 	return 0;
885 }
886 
887 static int copyfd_io_mmap(int infd, int peerfd, int outfd,
888 			  unsigned int size, bool *in_closed_after_out,
889 			  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_mmap(infd, peerfd, size, winfo);
899 	} else {
900 		err = do_mmap(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_sendfile(int infd, int peerfd, int outfd,
914 			      unsigned int size, bool *in_closed_after_out, struct wstate *winfo)
915 {
916 	int err;
917 
918 	if (listen_mode) {
919 		err = do_recvfile(peerfd, outfd);
920 		if (err)
921 			return err;
922 
923 		err = do_sendfile(infd, peerfd, size, winfo);
924 	} else {
925 		err = do_sendfile(infd, peerfd, size, winfo);
926 		if (err)
927 			return err;
928 
929 		shut_wr(peerfd);
930 
931 		err = do_recvfile(peerfd, outfd);
932 		*in_closed_after_out = true;
933 	}
934 
935 	return err;
936 }
937 
938 static int do_splice(const int infd, const int outfd, const size_t len,
939 		     struct wstate *winfo)
940 {
941 	ssize_t in_bytes, out_bytes;
942 	int pipefd[2];
943 	int err;
944 
945 	err = pipe(pipefd);
946 	if (err) {
947 		perror("pipe");
948 		return 2;
949 	}
950 
951 again:
952 	in_bytes = splice(infd, NULL, pipefd[1], NULL, len - winfo->total_len,
953 			  SPLICE_F_MOVE | SPLICE_F_MORE);
954 	if (in_bytes < 0) {
955 		perror("splice in");
956 		err = 3;
957 	} else if (in_bytes > 0) {
958 		out_bytes = splice(pipefd[0], NULL, outfd, NULL, in_bytes,
959 				   SPLICE_F_MOVE | SPLICE_F_MORE);
960 		if (out_bytes < 0) {
961 			perror("splice out");
962 			err = 4;
963 		} else if (in_bytes != out_bytes) {
964 			fprintf(stderr, "Unexpected transfer: %zu vs %zu\n",
965 				in_bytes, out_bytes);
966 			err = 5;
967 		} else {
968 			goto again;
969 		}
970 	}
971 
972 	close(pipefd[0]);
973 	close(pipefd[1]);
974 
975 	return err;
976 }
977 
978 static int copyfd_io_splice(int infd, int peerfd, int outfd, unsigned int size,
979 			    bool *in_closed_after_out, struct wstate *winfo)
980 {
981 	int err;
982 
983 	if (listen_mode) {
984 		err = do_splice(peerfd, outfd, size, winfo);
985 		if (err)
986 			return err;
987 
988 		err = do_splice(infd, peerfd, size, winfo);
989 	} else {
990 		err = do_splice(infd, peerfd, size, winfo);
991 		if (err)
992 			return err;
993 
994 		shut_wr(peerfd);
995 
996 		err = do_splice(peerfd, outfd, size, winfo);
997 		*in_closed_after_out = true;
998 	}
999 
1000 	return err;
1001 }
1002 
1003 static int copyfd_io(int infd, int peerfd, int outfd, bool close_peerfd, struct wstate *winfo)
1004 {
1005 	bool in_closed_after_out = false;
1006 	struct timespec start, end;
1007 	int file_size;
1008 	int ret;
1009 
1010 	if (cfg_time && (clock_gettime(CLOCK_MONOTONIC, &start) < 0))
1011 		xerror("can not fetch start time %d", errno);
1012 
1013 	switch (cfg_mode) {
1014 	case CFG_MODE_POLL:
1015 		ret = copyfd_io_poll(infd, peerfd, outfd, &in_closed_after_out,
1016 				     winfo);
1017 		break;
1018 
1019 	case CFG_MODE_MMAP:
1020 		file_size = get_infd_size(infd);
1021 		if (file_size < 0)
1022 			return file_size;
1023 		ret = copyfd_io_mmap(infd, peerfd, outfd, file_size,
1024 				     &in_closed_after_out, winfo);
1025 		break;
1026 
1027 	case CFG_MODE_SENDFILE:
1028 		file_size = get_infd_size(infd);
1029 		if (file_size < 0)
1030 			return file_size;
1031 		ret = copyfd_io_sendfile(infd, peerfd, outfd, file_size,
1032 					 &in_closed_after_out, winfo);
1033 		break;
1034 
1035 	case CFG_MODE_SPLICE:
1036 		file_size = get_infd_size(infd);
1037 		if (file_size < 0)
1038 			return file_size;
1039 		ret = copyfd_io_splice(infd, peerfd, outfd, file_size,
1040 				       &in_closed_after_out, winfo);
1041 		break;
1042 
1043 	default:
1044 		fprintf(stderr, "Invalid mode %d\n", cfg_mode);
1045 
1046 		die_usage();
1047 		return 1;
1048 	}
1049 
1050 	if (ret)
1051 		return ret;
1052 
1053 	if (close_peerfd)
1054 		close(peerfd);
1055 
1056 	if (cfg_time) {
1057 		unsigned int delta_ms;
1058 
1059 		if (clock_gettime(CLOCK_MONOTONIC, &end) < 0)
1060 			xerror("can not fetch end time %d", errno);
1061 		delta_ms = (end.tv_sec - start.tv_sec) * 1000 + (end.tv_nsec - start.tv_nsec) / 1000000;
1062 		if (delta_ms > cfg_time) {
1063 			xerror("transfer slower than expected! runtime %d ms, expected %d ms",
1064 			       delta_ms, cfg_time);
1065 		}
1066 
1067 		/* show the runtime only if this end shutdown(wr) before receiving the EOF,
1068 		 * (that is, if this end got the longer runtime)
1069 		 */
1070 		if (in_closed_after_out)
1071 			fprintf(stderr, "%d", delta_ms);
1072 	}
1073 
1074 	return 0;
1075 }
1076 
1077 static void check_sockaddr(int pf, struct sockaddr_storage *ss,
1078 			   socklen_t salen)
1079 {
1080 	struct sockaddr_in6 *sin6;
1081 	struct sockaddr_in *sin;
1082 	socklen_t wanted_size = 0;
1083 
1084 	switch (pf) {
1085 	case AF_INET:
1086 		wanted_size = sizeof(*sin);
1087 		sin = (void *)ss;
1088 		if (!sin->sin_port)
1089 			fprintf(stderr, "accept: something wrong: ip connection from port 0");
1090 		break;
1091 	case AF_INET6:
1092 		wanted_size = sizeof(*sin6);
1093 		sin6 = (void *)ss;
1094 		if (!sin6->sin6_port)
1095 			fprintf(stderr, "accept: something wrong: ipv6 connection from port 0");
1096 		break;
1097 	default:
1098 		fprintf(stderr, "accept: Unknown pf %d, salen %u\n", pf, salen);
1099 		return;
1100 	}
1101 
1102 	if (salen != wanted_size)
1103 		fprintf(stderr, "accept: size mismatch, got %d expected %d\n",
1104 			(int)salen, wanted_size);
1105 
1106 	if (ss->ss_family != pf)
1107 		fprintf(stderr, "accept: pf mismatch, expect %d, ss_family is %d\n",
1108 			(int)ss->ss_family, pf);
1109 }
1110 
1111 static void check_getpeername(int fd, struct sockaddr_storage *ss, socklen_t salen)
1112 {
1113 	struct sockaddr_storage peerss;
1114 	socklen_t peersalen = sizeof(peerss);
1115 
1116 	if (getpeername(fd, (struct sockaddr *)&peerss, &peersalen) < 0) {
1117 		perror("getpeername");
1118 		return;
1119 	}
1120 
1121 	if (peersalen != salen) {
1122 		fprintf(stderr, "%s: %d vs %d\n", __func__, peersalen, salen);
1123 		return;
1124 	}
1125 
1126 	if (memcmp(ss, &peerss, peersalen)) {
1127 		char a[INET6_ADDRSTRLEN];
1128 		char b[INET6_ADDRSTRLEN];
1129 		char c[INET6_ADDRSTRLEN];
1130 		char d[INET6_ADDRSTRLEN];
1131 
1132 		xgetnameinfo((struct sockaddr *)ss, salen,
1133 			     a, sizeof(a), b, sizeof(b));
1134 
1135 		xgetnameinfo((struct sockaddr *)&peerss, peersalen,
1136 			     c, sizeof(c), d, sizeof(d));
1137 
1138 		fprintf(stderr, "%s: memcmp failure: accept %s vs peername %s, %s vs %s salen %d vs %d\n",
1139 			__func__, a, c, b, d, peersalen, salen);
1140 	}
1141 }
1142 
1143 static void check_getpeername_connect(int fd)
1144 {
1145 	struct sockaddr_storage ss;
1146 	socklen_t salen = sizeof(ss);
1147 	char a[INET6_ADDRSTRLEN];
1148 	char b[INET6_ADDRSTRLEN];
1149 	const char *iface;
1150 	size_t len;
1151 
1152 	if (getpeername(fd, (struct sockaddr *)&ss, &salen) < 0) {
1153 		perror("getpeername");
1154 		return;
1155 	}
1156 
1157 	xgetnameinfo((struct sockaddr *)&ss, salen,
1158 		     a, sizeof(a), b, sizeof(b));
1159 
1160 	iface = strchr(cfg_host, '%');
1161 	if (iface)
1162 		len = iface - cfg_host;
1163 	else
1164 		len = strlen(cfg_host) + 1;
1165 
1166 	if (strncmp(cfg_host, a, len) || strcmp(cfg_port, b))
1167 		fprintf(stderr, "%s: %s vs %s, %s vs %s\n", __func__,
1168 			cfg_host, a, cfg_port, b);
1169 }
1170 
1171 static void maybe_close(int fd)
1172 {
1173 	unsigned int r = rand();
1174 
1175 	if (!(cfg_join || cfg_remove || cfg_repeat > 1) && (r & 1))
1176 		close(fd);
1177 }
1178 
1179 int main_loop_s(int listensock)
1180 {
1181 	struct sockaddr_storage ss;
1182 	struct wstate winfo;
1183 	struct pollfd polls;
1184 	socklen_t salen;
1185 	int remotesock;
1186 	int err = 0;
1187 	int fd = 0;
1188 
1189 again:
1190 	polls.fd = listensock;
1191 	polls.events = POLLIN;
1192 
1193 	switch (poll(&polls, 1, poll_timeout)) {
1194 	case -1:
1195 		perror("poll");
1196 		return 1;
1197 	case 0:
1198 		fprintf(stderr, "%s: timed out\n", __func__);
1199 		close(listensock);
1200 		return 2;
1201 	}
1202 
1203 	salen = sizeof(ss);
1204 	remotesock = accept(listensock, (struct sockaddr *)&ss, &salen);
1205 	if (remotesock >= 0) {
1206 		maybe_close(listensock);
1207 		check_sockaddr(pf, &ss, salen);
1208 		check_getpeername(remotesock, &ss, salen);
1209 
1210 		if (cfg_input) {
1211 			fd = open(cfg_input, O_RDONLY);
1212 			if (fd < 0)
1213 				xerror("can't open %s: %d", cfg_input, errno);
1214 		}
1215 
1216 		SOCK_TEST_TCPULP(remotesock, 0);
1217 
1218 		memset(&winfo, 0, sizeof(winfo));
1219 		err = copyfd_io(fd, remotesock, 1, true, &winfo);
1220 	} else {
1221 		perror("accept");
1222 		return 1;
1223 	}
1224 
1225 	if (cfg_input)
1226 		close(fd);
1227 
1228 	if (!err && --cfg_repeat > 0)
1229 		goto again;
1230 
1231 	return err;
1232 }
1233 
1234 static void init_rng(void)
1235 {
1236 	unsigned int foo;
1237 
1238 	if (getrandom(&foo, sizeof(foo), 0) == -1) {
1239 		perror("getrandom");
1240 		exit(1);
1241 	}
1242 
1243 	srand(foo);
1244 }
1245 
1246 static void xsetsockopt(int fd, int level, int optname, const void *optval, socklen_t optlen)
1247 {
1248 	int err;
1249 
1250 	err = setsockopt(fd, level, optname, optval, optlen);
1251 	if (err) {
1252 		perror("setsockopt");
1253 		exit(1);
1254 	}
1255 }
1256 
1257 static void apply_cmsg_types(int fd, const struct cfg_cmsg_types *cmsg)
1258 {
1259 	static const unsigned int on = 1;
1260 
1261 	if (cmsg->timestampns)
1262 		xsetsockopt(fd, SOL_SOCKET, SO_TIMESTAMPNS_NEW, &on, sizeof(on));
1263 	if (cmsg->tcp_inq)
1264 		xsetsockopt(fd, IPPROTO_TCP, TCP_INQ, &on, sizeof(on));
1265 }
1266 
1267 static void parse_cmsg_types(const char *type)
1268 {
1269 	const char *next = strchr(type, ',');
1270 	unsigned int len = 0;
1271 
1272 	cfg_cmsg_types.cmsg_enabled = 1;
1273 
1274 	if (next) {
1275 		parse_cmsg_types(next + 1);
1276 		len = next - type;
1277 	} else {
1278 		len = strlen(type);
1279 	}
1280 
1281 	if (strncmp(type, "TIMESTAMPNS", len) == 0) {
1282 		cfg_cmsg_types.timestampns = 1;
1283 		return;
1284 	}
1285 
1286 	if (strncmp(type, "TCPINQ", len) == 0) {
1287 		cfg_cmsg_types.tcp_inq = 1;
1288 		return;
1289 	}
1290 
1291 	fprintf(stderr, "Unrecognized cmsg option %s\n", type);
1292 	exit(1);
1293 }
1294 
1295 static void parse_setsock_options(const char *name)
1296 {
1297 	const char *next = strchr(name, ',');
1298 	unsigned int len = 0;
1299 
1300 	if (next) {
1301 		parse_setsock_options(next + 1);
1302 		len = next - name;
1303 	} else {
1304 		len = strlen(name);
1305 	}
1306 
1307 	if (strncmp(name, "TRANSPARENT", len) == 0) {
1308 		cfg_sockopt_types.transparent = 1;
1309 		return;
1310 	}
1311 
1312 	if (strncmp(name, "MPTFO", len) == 0) {
1313 		cfg_sockopt_types.mptfo = 1;
1314 		return;
1315 	}
1316 
1317 	fprintf(stderr, "Unrecognized setsockopt option %s\n", name);
1318 	exit(1);
1319 }
1320 
1321 void xdisconnect(int fd)
1322 {
1323 	socklen_t addrlen = sizeof(struct sockaddr_storage);
1324 	struct sockaddr_storage addr, empty;
1325 	int msec_sleep = 10;
1326 	void *raw_addr;
1327 	int i, cmdlen;
1328 	char cmd[128];
1329 
1330 	/* get the local address and convert it to string */
1331 	if (getsockname(fd, (struct sockaddr *)&addr, &addrlen) < 0)
1332 		xerror("getsockname");
1333 
1334 	if (addr.ss_family == AF_INET)
1335 		raw_addr = &(((struct sockaddr_in *)&addr)->sin_addr);
1336 	else if (addr.ss_family == AF_INET6)
1337 		raw_addr = &(((struct sockaddr_in6 *)&addr)->sin6_addr);
1338 	else
1339 		xerror("bad family");
1340 
1341 	strcpy(cmd, "ss -Mnt | grep -q ");
1342 	cmdlen = strlen(cmd);
1343 	if (!inet_ntop(addr.ss_family, raw_addr, &cmd[cmdlen],
1344 		       sizeof(cmd) - cmdlen))
1345 		xerror("inet_ntop");
1346 
1347 	shutdown(fd, SHUT_WR);
1348 
1349 	/*
1350 	 * wait until the pending data is completely flushed and all
1351 	 * the sockets reached the closed status.
1352 	 * disconnect will bypass/ignore/drop any pending data.
1353 	 */
1354 	for (i = 0; ; i += msec_sleep) {
1355 		/* closed socket are not listed by 'ss' */
1356 		if (system(cmd) != 0)
1357 			break;
1358 
1359 		if (i > poll_timeout)
1360 			xerror("timeout while waiting for spool to complete");
1361 		usleep(msec_sleep * 1000);
1362 	}
1363 
1364 	memset(&empty, 0, sizeof(empty));
1365 	empty.ss_family = AF_UNSPEC;
1366 	if (connect(fd, (struct sockaddr *)&empty, addrlen) < 0)
1367 		xerror("can't disconnect: %d", errno);
1368 }
1369 
1370 int main_loop(void)
1371 {
1372 	struct addrinfo *peer = NULL;
1373 	int fd = 0, ret, fd_in = 0;
1374 	struct wstate winfo;
1375 
1376 	if (cfg_input && cfg_sockopt_types.mptfo) {
1377 		fd_in = open(cfg_input, O_RDONLY);
1378 		if (fd_in < 0)
1379 			xerror("can't open %s:%d", cfg_input, errno);
1380 	}
1381 
1382 	memset(&winfo, 0, sizeof(winfo));
1383 	fd = sock_connect_mptcp(cfg_host, cfg_port, cfg_sock_proto, &peer, fd_in, &winfo);
1384 	if (fd < 0)
1385 		return 2;
1386 
1387 again:
1388 	check_getpeername_connect(fd);
1389 
1390 	SOCK_TEST_TCPULP(fd, cfg_sock_proto);
1391 
1392 	if (cfg_rcvbuf)
1393 		set_rcvbuf(fd, cfg_rcvbuf);
1394 	if (cfg_sndbuf)
1395 		set_sndbuf(fd, cfg_sndbuf);
1396 	if (cfg_cmsg_types.cmsg_enabled)
1397 		apply_cmsg_types(fd, &cfg_cmsg_types);
1398 
1399 	if (cfg_input && !cfg_sockopt_types.mptfo) {
1400 		fd_in = open(cfg_input, O_RDONLY);
1401 		if (fd_in < 0)
1402 			xerror("can't open %s:%d", cfg_input, errno);
1403 	}
1404 
1405 	ret = copyfd_io(fd_in, fd, 1, 0, &winfo);
1406 	if (ret)
1407 		goto out;
1408 
1409 	if (cfg_truncate > 0) {
1410 		shutdown(fd, SHUT_WR);
1411 	} else if (--cfg_repeat > 0) {
1412 		xdisconnect(fd);
1413 
1414 		/* the socket could be unblocking at this point, we need the
1415 		 * connect to be blocking
1416 		 */
1417 		set_nonblock(fd, false);
1418 		if (connect(fd, peer->ai_addr, peer->ai_addrlen))
1419 			xerror("can't reconnect: %d", errno);
1420 		if (cfg_input)
1421 			close(fd_in);
1422 		memset(&winfo, 0, sizeof(winfo));
1423 		goto again;
1424 	} else {
1425 		close(fd);
1426 	}
1427 
1428 out:
1429 	if (cfg_input)
1430 		close(fd_in);
1431 	return ret;
1432 }
1433 
1434 int parse_proto(const char *proto)
1435 {
1436 	if (!strcasecmp(proto, "MPTCP"))
1437 		return IPPROTO_MPTCP;
1438 	if (!strcasecmp(proto, "TCP"))
1439 		return IPPROTO_TCP;
1440 
1441 	fprintf(stderr, "Unknown protocol: %s\n.", proto);
1442 	die_usage();
1443 
1444 	/* silence compiler warning */
1445 	return 0;
1446 }
1447 
1448 int parse_mode(const char *mode)
1449 {
1450 	if (!strcasecmp(mode, "poll"))
1451 		return CFG_MODE_POLL;
1452 	if (!strcasecmp(mode, "mmap"))
1453 		return CFG_MODE_MMAP;
1454 	if (!strcasecmp(mode, "sendfile"))
1455 		return CFG_MODE_SENDFILE;
1456 	if (!strcasecmp(mode, "splice"))
1457 		return CFG_MODE_SPLICE;
1458 
1459 	fprintf(stderr, "Unknown test mode: %s\n", mode);
1460 	fprintf(stderr, "Supported modes are:\n");
1461 	fprintf(stderr, "\t\t\"poll\" - interleaved read/write using poll()\n");
1462 	fprintf(stderr, "\t\t\"mmap\" - send entire input file (mmap+write), then read response (-l will read input first)\n");
1463 	fprintf(stderr, "\t\t\"sendfile\" - send entire input file (sendfile), then read response (-l will read input first)\n");
1464 	fprintf(stderr, "\t\t\"splice\" - send entire input file (splice), then read response (-l will read input first)\n");
1465 
1466 	die_usage();
1467 
1468 	/* silence compiler warning */
1469 	return 0;
1470 }
1471 
1472 int parse_peek(const char *mode)
1473 {
1474 	if (!strcasecmp(mode, "saveWithPeek"))
1475 		return CFG_WITH_PEEK;
1476 	if (!strcasecmp(mode, "saveAfterPeek"))
1477 		return CFG_AFTER_PEEK;
1478 
1479 	fprintf(stderr, "Unknown: %s\n", mode);
1480 	fprintf(stderr, "Supported MSG_PEEK mode are:\n");
1481 	fprintf(stderr,
1482 		"\t\t\"saveWithPeek\" - recv data with flags 'MSG_PEEK' and save the peek data into file\n");
1483 	fprintf(stderr,
1484 		"\t\t\"saveAfterPeek\" - read and save data into file after recv with flags 'MSG_PEEK'\n");
1485 
1486 	die_usage();
1487 
1488 	/* silence compiler warning */
1489 	return 0;
1490 }
1491 
1492 static int parse_int(const char *size)
1493 {
1494 	unsigned long s;
1495 
1496 	errno = 0;
1497 
1498 	s = strtoul(size, NULL, 0);
1499 
1500 	if (errno) {
1501 		fprintf(stderr, "Invalid sndbuf size %s (%s)\n",
1502 			size, strerror(errno));
1503 		die_usage();
1504 	}
1505 
1506 	if (s > INT_MAX) {
1507 		fprintf(stderr, "Invalid sndbuf size %s (%s)\n",
1508 			size, strerror(ERANGE));
1509 		die_usage();
1510 	}
1511 
1512 	return (int)s;
1513 }
1514 
1515 static void parse_opts(int argc, char **argv)
1516 {
1517 	int c;
1518 
1519 	while ((c = getopt(argc, argv, "6c:f:hi:I:jlm:M:o:p:P:r:R:s:S:t:T:w:")) != -1) {
1520 		switch (c) {
1521 		case 'f':
1522 			cfg_truncate = atoi(optarg);
1523 
1524 			/* when receiving a fastclose, ignore PIPE signals and
1525 			 * all the I/O errors later in the code
1526 			 */
1527 			if (cfg_truncate < 0) {
1528 				cfg_rcv_trunc = true;
1529 				signal(SIGPIPE, SIG_IGN);
1530 			}
1531 			break;
1532 		case 'j':
1533 			cfg_join = true;
1534 			cfg_mode = CFG_MODE_POLL;
1535 			break;
1536 		case 'r':
1537 			cfg_remove = true;
1538 			cfg_mode = CFG_MODE_POLL;
1539 			cfg_wait = 400000;
1540 			cfg_do_w = atoi(optarg);
1541 			if (cfg_do_w <= 0)
1542 				cfg_do_w = 50;
1543 			break;
1544 		case 'i':
1545 			cfg_input = optarg;
1546 			break;
1547 		case 'I':
1548 			cfg_repeat = atoi(optarg);
1549 			break;
1550 		case 'l':
1551 			listen_mode = true;
1552 			break;
1553 		case 'p':
1554 			cfg_port = optarg;
1555 			break;
1556 		case 's':
1557 			cfg_sock_proto = parse_proto(optarg);
1558 			break;
1559 		case 'h':
1560 			die_usage();
1561 			break;
1562 		case '6':
1563 			pf = AF_INET6;
1564 			break;
1565 		case 't':
1566 			poll_timeout = atoi(optarg) * 1000;
1567 			if (poll_timeout <= 0)
1568 				poll_timeout = -1;
1569 			break;
1570 		case 'T':
1571 			cfg_time = atoi(optarg);
1572 			break;
1573 		case 'm':
1574 			cfg_mode = parse_mode(optarg);
1575 			break;
1576 		case 'S':
1577 			cfg_sndbuf = parse_int(optarg);
1578 			break;
1579 		case 'R':
1580 			cfg_rcvbuf = parse_int(optarg);
1581 			break;
1582 		case 'w':
1583 			cfg_wait = atoi(optarg)*1000000;
1584 			break;
1585 		case 'M':
1586 			cfg_mark = strtol(optarg, NULL, 0);
1587 			break;
1588 		case 'P':
1589 			cfg_peek = parse_peek(optarg);
1590 			break;
1591 		case 'c':
1592 			parse_cmsg_types(optarg);
1593 			break;
1594 		case 'o':
1595 			parse_setsock_options(optarg);
1596 			break;
1597 		}
1598 	}
1599 
1600 	if (optind + 1 != argc)
1601 		die_usage();
1602 	cfg_host = argv[optind];
1603 
1604 	if (strchr(cfg_host, ':'))
1605 		pf = AF_INET6;
1606 }
1607 
1608 int main(int argc, char *argv[])
1609 {
1610 	init_rng();
1611 
1612 	signal(SIGUSR1, handle_signal);
1613 	parse_opts(argc, argv);
1614 
1615 	if (listen_mode) {
1616 		int fd = sock_listen_mptcp(cfg_host, cfg_port);
1617 
1618 		if (fd < 0)
1619 			return 1;
1620 
1621 		if (cfg_rcvbuf)
1622 			set_rcvbuf(fd, cfg_rcvbuf);
1623 		if (cfg_sndbuf)
1624 			set_sndbuf(fd, cfg_sndbuf);
1625 		if (cfg_mark)
1626 			set_mark(fd, cfg_mark);
1627 		if (cfg_cmsg_types.cmsg_enabled)
1628 			apply_cmsg_types(fd, &cfg_cmsg_types);
1629 
1630 		return main_loop_s(fd);
1631 	}
1632 
1633 	return main_loop();
1634 }
1635