xref: /linux/tools/testing/selftests/net/mptcp/mptcp_connect.c (revision d3b402c5a2d47f51eb0581da1a7b142f82cb10d1)
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 
die_usage(void)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 
xerror(const char * fmt,...)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 
handle_signal(int nr)155 static void handle_signal(int nr)
156 {
157 	quit = true;
158 }
159 
getxinfo_strerr(int err)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 
xgetnameinfo(const struct sockaddr * addr,socklen_t addrlen,char * host,socklen_t hostlen,char * serv,socklen_t servlen)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 
xgetaddrinfo(const char * node,const char * service,struct addrinfo * hints,struct addrinfo ** res)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 
set_rcvbuf(int fd,unsigned int size)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 
set_sndbuf(int fd,unsigned int size)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 
set_mark(int fd,uint32_t mark)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 
set_transparent(int fd,int pf)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 
set_mptfo(int fd)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 
do_ulp_so(int sock,const char * name)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)
sock_test_tcpulp(int sock,int proto,unsigned int 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 
sock_listen_mptcp(const char * const listenaddr,const char * const port)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 
sock_connect_mptcp(const char * const remoteaddr,const char * const port,int proto,struct addrinfo ** peer,int infd,struct wstate * winfo)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 	xgetaddrinfo(remoteaddr, port, &hints, &addr);
385 	for (a = addr; a; a = a->ai_next) {
386 		sock = socket(a->ai_family, a->ai_socktype, proto);
387 		if (sock < 0) {
388 			perror("socket");
389 			continue;
390 		}
391 
392 		SOCK_TEST_TCPULP(sock, proto);
393 
394 		if (cfg_mark)
395 			set_mark(sock, cfg_mark);
396 
397 		if (cfg_sockopt_types.mptfo) {
398 			if (!winfo->total_len)
399 				winfo->total_len = winfo->len = read(infd, winfo->buf,
400 								     sizeof(winfo->buf));
401 
402 			syn_copied = sendto(sock, winfo->buf, winfo->len, MSG_FASTOPEN,
403 					    a->ai_addr, a->ai_addrlen);
404 			if (syn_copied >= 0) {
405 				winfo->off = syn_copied;
406 				winfo->len -= syn_copied;
407 				*peer = a;
408 				break; /* success */
409 			}
410 			perror("sendto()");
411 		} else {
412 			if (connect(sock, a->ai_addr, a->ai_addrlen) == 0) {
413 				*peer = a;
414 				break; /* success */
415 			}
416 			perror("connect()");
417 		}
418 
419 		/* error */
420 		close(sock);
421 		sock = -1;
422 	}
423 
424 	freeaddrinfo(addr);
425 	if (sock != -1)
426 		SOCK_TEST_TCPULP(sock, proto);
427 	return sock;
428 }
429 
do_rnd_write(const int fd,char * buf,const size_t len)430 static size_t do_rnd_write(const int fd, char *buf, const size_t len)
431 {
432 	static bool first = true;
433 	unsigned int do_w;
434 	ssize_t bw;
435 
436 	do_w = rand() & 0xffff;
437 	if (do_w == 0 || do_w > len)
438 		do_w = len;
439 
440 	if (cfg_join && first && do_w > 100)
441 		do_w = 100;
442 
443 	if (cfg_remove && do_w > cfg_do_w)
444 		do_w = cfg_do_w;
445 
446 	bw = write(fd, buf, do_w);
447 	if (bw < 0)
448 		return bw;
449 
450 	/* let the join handshake complete, before going on */
451 	if (cfg_join && first) {
452 		usleep(200000);
453 		first = false;
454 	}
455 
456 	if (cfg_remove)
457 		usleep(200000);
458 
459 	return bw;
460 }
461 
do_write(const int fd,char * buf,const size_t len)462 static size_t do_write(const int fd, char *buf, const size_t len)
463 {
464 	size_t offset = 0;
465 
466 	while (offset < len) {
467 		size_t written;
468 		ssize_t bw;
469 
470 		bw = write(fd, buf + offset, len - offset);
471 		if (bw < 0) {
472 			perror("write");
473 			return 0;
474 		}
475 
476 		written = (size_t)bw;
477 		offset += written;
478 	}
479 
480 	return offset;
481 }
482 
process_cmsg(struct msghdr * msgh)483 static void process_cmsg(struct msghdr *msgh)
484 {
485 	struct __kernel_timespec ts;
486 	bool inq_found = false;
487 	bool ts_found = false;
488 	unsigned int inq = 0;
489 	struct cmsghdr *cmsg;
490 
491 	for (cmsg = CMSG_FIRSTHDR(msgh); cmsg ; cmsg = CMSG_NXTHDR(msgh, cmsg)) {
492 		if (cmsg->cmsg_level == SOL_SOCKET && cmsg->cmsg_type == SO_TIMESTAMPNS_NEW) {
493 			memcpy(&ts, CMSG_DATA(cmsg), sizeof(ts));
494 			ts_found = true;
495 			continue;
496 		}
497 		if (cmsg->cmsg_level == IPPROTO_TCP && cmsg->cmsg_type == TCP_CM_INQ) {
498 			memcpy(&inq, CMSG_DATA(cmsg), sizeof(inq));
499 			inq_found = true;
500 			continue;
501 		}
502 
503 	}
504 
505 	if (cfg_cmsg_types.timestampns) {
506 		if (!ts_found)
507 			xerror("TIMESTAMPNS not present\n");
508 	}
509 
510 	if (cfg_cmsg_types.tcp_inq) {
511 		if (!inq_found)
512 			xerror("TCP_INQ not present\n");
513 
514 		if (inq > 1024)
515 			xerror("tcp_inq %u is larger than one kbyte\n", inq);
516 		tcp_inq.last = inq;
517 	}
518 }
519 
do_recvmsg_cmsg(const int fd,char * buf,const size_t len)520 static ssize_t do_recvmsg_cmsg(const int fd, char *buf, const size_t len)
521 {
522 	char msg_buf[8192];
523 	struct iovec iov = {
524 		.iov_base = buf,
525 		.iov_len = len,
526 	};
527 	struct msghdr msg = {
528 		.msg_iov = &iov,
529 		.msg_iovlen = 1,
530 		.msg_control = msg_buf,
531 		.msg_controllen = sizeof(msg_buf),
532 	};
533 	int flags = 0;
534 	unsigned int last_hint = tcp_inq.last;
535 	int ret = recvmsg(fd, &msg, flags);
536 
537 	if (ret <= 0) {
538 		if (ret == 0 && tcp_inq.expect_eof)
539 			return ret;
540 
541 		if (ret == 0 && cfg_cmsg_types.tcp_inq)
542 			if (last_hint != 1 && last_hint != 0)
543 				xerror("EOF but last tcp_inq hint was %u\n", last_hint);
544 
545 		return ret;
546 	}
547 
548 	if (tcp_inq.expect_eof)
549 		xerror("expected EOF, last_hint %u, now %u\n",
550 		       last_hint, tcp_inq.last);
551 
552 	if (msg.msg_controllen && !cfg_cmsg_types.cmsg_enabled)
553 		xerror("got %lu bytes of cmsg data, expected 0\n",
554 		       (unsigned long)msg.msg_controllen);
555 
556 	if (msg.msg_controllen == 0 && cfg_cmsg_types.cmsg_enabled)
557 		xerror("%s\n", "got no cmsg data");
558 
559 	if (msg.msg_controllen)
560 		process_cmsg(&msg);
561 
562 	if (cfg_cmsg_types.tcp_inq) {
563 		if ((size_t)ret < len && last_hint > (unsigned int)ret) {
564 			if (ret + 1 != (int)last_hint) {
565 				int next = read(fd, msg_buf, sizeof(msg_buf));
566 
567 				xerror("read %u of %u, last_hint was %u tcp_inq hint now %u next_read returned %d/%m\n",
568 				       ret, (unsigned int)len, last_hint, tcp_inq.last, next);
569 			} else {
570 				tcp_inq.expect_eof = true;
571 			}
572 		}
573 	}
574 
575 	return ret;
576 }
577 
do_rnd_read(const int fd,char * buf,const size_t len)578 static ssize_t do_rnd_read(const int fd, char *buf, const size_t len)
579 {
580 	int ret = 0;
581 	char tmp[16384];
582 	size_t cap = rand();
583 
584 	cap &= 0xffff;
585 
586 	if (cap == 0)
587 		cap = 1;
588 	else if (cap > len)
589 		cap = len;
590 
591 	if (cfg_peek == CFG_WITH_PEEK) {
592 		ret = recv(fd, buf, cap, MSG_PEEK);
593 		ret = (ret < 0) ? ret : read(fd, tmp, ret);
594 	} else if (cfg_peek == CFG_AFTER_PEEK) {
595 		ret = recv(fd, buf, cap, MSG_PEEK);
596 		ret = (ret < 0) ? ret : read(fd, buf, cap);
597 	} else if (cfg_cmsg_types.cmsg_enabled) {
598 		ret = do_recvmsg_cmsg(fd, buf, cap);
599 	} else {
600 		ret = read(fd, buf, cap);
601 	}
602 
603 	return ret;
604 }
605 
set_nonblock(int fd,bool nonblock)606 static void set_nonblock(int fd, bool nonblock)
607 {
608 	int flags = fcntl(fd, F_GETFL);
609 
610 	if (flags == -1)
611 		return;
612 
613 	if (nonblock)
614 		fcntl(fd, F_SETFL, flags | O_NONBLOCK);
615 	else
616 		fcntl(fd, F_SETFL, flags & ~O_NONBLOCK);
617 }
618 
shut_wr(int fd)619 static void shut_wr(int fd)
620 {
621 	/* Close our write side, ev. give some time
622 	 * for address notification and/or checking
623 	 * the current status
624 	 */
625 	if (cfg_wait)
626 		usleep(cfg_wait);
627 
628 	shutdown(fd, SHUT_WR);
629 }
630 
copyfd_io_poll(int infd,int peerfd,int outfd,bool * in_closed_after_out,struct wstate * winfo)631 static int copyfd_io_poll(int infd, int peerfd, int outfd,
632 			  bool *in_closed_after_out, struct wstate *winfo)
633 {
634 	struct pollfd fds = {
635 		.fd = peerfd,
636 		.events = POLLIN | POLLOUT,
637 	};
638 	unsigned int total_wlen = 0, total_rlen = 0;
639 
640 	set_nonblock(peerfd, true);
641 
642 	for (;;) {
643 		char rbuf[8192];
644 		ssize_t len;
645 
646 		if (fds.events == 0 || quit)
647 			break;
648 
649 		switch (poll(&fds, 1, poll_timeout)) {
650 		case -1:
651 			if (errno == EINTR)
652 				continue;
653 			perror("poll");
654 			return 1;
655 		case 0:
656 			fprintf(stderr, "%s: poll timed out (events: "
657 				"POLLIN %u, POLLOUT %u)\n", __func__,
658 				fds.events & POLLIN, fds.events & POLLOUT);
659 			return 2;
660 		}
661 
662 		if (fds.revents & POLLIN) {
663 			ssize_t rb = sizeof(rbuf);
664 
665 			/* limit the total amount of read data to the trunc value*/
666 			if (cfg_truncate > 0) {
667 				if (rb + total_rlen > cfg_truncate)
668 					rb = cfg_truncate - total_rlen;
669 				len = read(peerfd, rbuf, rb);
670 			} else {
671 				len = do_rnd_read(peerfd, rbuf, sizeof(rbuf));
672 			}
673 			if (len == 0) {
674 				/* no more data to receive:
675 				 * peer has closed its write side
676 				 */
677 				fds.events &= ~POLLIN;
678 
679 				if ((fds.events & POLLOUT) == 0) {
680 					*in_closed_after_out = true;
681 					/* and nothing more to send */
682 					break;
683 				}
684 
685 			/* Else, still have data to transmit */
686 			} else if (len < 0) {
687 				if (cfg_rcv_trunc)
688 					return 0;
689 				perror("read");
690 				return 3;
691 			}
692 
693 			total_rlen += len;
694 			do_write(outfd, rbuf, len);
695 		}
696 
697 		if (fds.revents & POLLOUT) {
698 			if (winfo->len == 0) {
699 				winfo->off = 0;
700 				winfo->len = read(infd, winfo->buf, sizeof(winfo->buf));
701 			}
702 
703 			if (winfo->len > 0) {
704 				ssize_t bw;
705 
706 				/* limit the total amount of written data to the trunc value */
707 				if (cfg_truncate > 0 && winfo->len + total_wlen > cfg_truncate)
708 					winfo->len = cfg_truncate - total_wlen;
709 
710 				bw = do_rnd_write(peerfd, winfo->buf + winfo->off, winfo->len);
711 				if (bw < 0) {
712 					/* expected reset, continue to read */
713 					if (cfg_rcv_trunc &&
714 					    (errno == ECONNRESET ||
715 					     errno == EPIPE)) {
716 						fds.events &= ~POLLOUT;
717 						continue;
718 					}
719 
720 					perror("write");
721 					return 111;
722 				}
723 
724 				winfo->off += bw;
725 				winfo->len -= bw;
726 				total_wlen += bw;
727 			} else if (winfo->len == 0) {
728 				/* We have no more data to send. */
729 				fds.events &= ~POLLOUT;
730 
731 				if ((fds.events & POLLIN) == 0)
732 					/* ... and peer also closed already */
733 					break;
734 
735 				shut_wr(peerfd);
736 			} else {
737 				if (errno == EINTR)
738 					continue;
739 				perror("read");
740 				return 4;
741 			}
742 		}
743 
744 		if (fds.revents & (POLLERR | POLLNVAL)) {
745 			if (cfg_rcv_trunc) {
746 				fds.events &= ~(POLLERR | POLLNVAL);
747 				continue;
748 			}
749 			fprintf(stderr, "Unexpected revents: "
750 				"POLLERR/POLLNVAL(%x)\n", fds.revents);
751 			return 5;
752 		}
753 
754 		if (cfg_truncate > 0 && total_wlen >= cfg_truncate &&
755 		    total_rlen >= cfg_truncate)
756 			break;
757 	}
758 
759 	/* leave some time for late join/announce */
760 	if (cfg_remove && !quit)
761 		usleep(cfg_wait);
762 
763 	return 0;
764 }
765 
do_recvfile(int infd,int outfd)766 static int do_recvfile(int infd, int outfd)
767 {
768 	ssize_t r;
769 
770 	do {
771 		char buf[16384];
772 
773 		r = do_rnd_read(infd, buf, sizeof(buf));
774 		if (r > 0) {
775 			if (write(outfd, buf, r) != r)
776 				break;
777 		} else if (r < 0) {
778 			perror("read");
779 		}
780 	} while (r > 0);
781 
782 	return (int)r;
783 }
784 
spool_buf(int fd,struct wstate * winfo)785 static int spool_buf(int fd, struct wstate *winfo)
786 {
787 	while (winfo->len) {
788 		int ret = write(fd, winfo->buf + winfo->off, winfo->len);
789 
790 		if (ret < 0) {
791 			perror("write");
792 			return 4;
793 		}
794 		winfo->off += ret;
795 		winfo->len -= ret;
796 	}
797 	return 0;
798 }
799 
do_mmap(int infd,int outfd,unsigned int size,struct wstate * winfo)800 static int do_mmap(int infd, int outfd, unsigned int size,
801 		   struct wstate *winfo)
802 {
803 	char *inbuf = mmap(NULL, size, PROT_READ, MAP_SHARED, infd, 0);
804 	ssize_t ret = 0, off = winfo->total_len;
805 	size_t rem;
806 
807 	if (inbuf == MAP_FAILED) {
808 		perror("mmap");
809 		return 1;
810 	}
811 
812 	ret = spool_buf(outfd, winfo);
813 	if (ret < 0)
814 		return ret;
815 
816 	rem = size - winfo->total_len;
817 
818 	while (rem > 0) {
819 		ret = write(outfd, inbuf + off, rem);
820 
821 		if (ret < 0) {
822 			perror("write");
823 			break;
824 		}
825 
826 		off += ret;
827 		rem -= ret;
828 	}
829 
830 	munmap(inbuf, size);
831 	return rem;
832 }
833 
get_infd_size(int fd)834 static int get_infd_size(int fd)
835 {
836 	struct stat sb;
837 	ssize_t count;
838 	int err;
839 
840 	err = fstat(fd, &sb);
841 	if (err < 0) {
842 		perror("fstat");
843 		return -1;
844 	}
845 
846 	if ((sb.st_mode & S_IFMT) != S_IFREG) {
847 		fprintf(stderr, "%s: stdin is not a regular file\n", __func__);
848 		return -2;
849 	}
850 
851 	count = sb.st_size;
852 	if (count > INT_MAX) {
853 		fprintf(stderr, "File too large: %zu\n", count);
854 		return -3;
855 	}
856 
857 	return (int)count;
858 }
859 
do_sendfile(int infd,int outfd,unsigned int count,struct wstate * winfo)860 static int do_sendfile(int infd, int outfd, unsigned int count,
861 		       struct wstate *winfo)
862 {
863 	int ret = spool_buf(outfd, winfo);
864 
865 	if (ret < 0)
866 		return ret;
867 
868 	count -= winfo->total_len;
869 
870 	while (count > 0) {
871 		ssize_t r;
872 
873 		r = sendfile(outfd, infd, NULL, count);
874 		if (r < 0) {
875 			perror("sendfile");
876 			return 3;
877 		}
878 
879 		count -= r;
880 	}
881 
882 	return 0;
883 }
884 
copyfd_io_mmap(int infd,int peerfd,int outfd,unsigned int size,bool * in_closed_after_out,struct wstate * winfo)885 static int copyfd_io_mmap(int infd, int peerfd, int outfd,
886 			  unsigned int size, bool *in_closed_after_out,
887 			  struct wstate *winfo)
888 {
889 	int err;
890 
891 	if (listen_mode) {
892 		err = do_recvfile(peerfd, outfd);
893 		if (err)
894 			return err;
895 
896 		err = do_mmap(infd, peerfd, size, winfo);
897 	} else {
898 		err = do_mmap(infd, peerfd, size, winfo);
899 		if (err)
900 			return err;
901 
902 		shut_wr(peerfd);
903 
904 		err = do_recvfile(peerfd, outfd);
905 		*in_closed_after_out = true;
906 	}
907 
908 	return err;
909 }
910 
copyfd_io_sendfile(int infd,int peerfd,int outfd,unsigned int size,bool * in_closed_after_out,struct wstate * winfo)911 static int copyfd_io_sendfile(int infd, int peerfd, int outfd,
912 			      unsigned int size, bool *in_closed_after_out, struct wstate *winfo)
913 {
914 	int err;
915 
916 	if (listen_mode) {
917 		err = do_recvfile(peerfd, outfd);
918 		if (err)
919 			return err;
920 
921 		err = do_sendfile(infd, peerfd, size, winfo);
922 	} else {
923 		err = do_sendfile(infd, peerfd, size, winfo);
924 		if (err)
925 			return err;
926 
927 		shut_wr(peerfd);
928 
929 		err = do_recvfile(peerfd, outfd);
930 		*in_closed_after_out = true;
931 	}
932 
933 	return err;
934 }
935 
do_splice(const int infd,const int outfd,const size_t len,struct wstate * winfo)936 static int do_splice(const int infd, const int outfd, const size_t len,
937 		     struct wstate *winfo)
938 {
939 	ssize_t in_bytes, out_bytes;
940 	int pipefd[2];
941 	int err;
942 
943 	err = pipe(pipefd);
944 	if (err) {
945 		perror("pipe");
946 		return 2;
947 	}
948 
949 again:
950 	in_bytes = splice(infd, NULL, pipefd[1], NULL, len - winfo->total_len,
951 			  SPLICE_F_MOVE | SPLICE_F_MORE);
952 	if (in_bytes < 0) {
953 		perror("splice in");
954 		err = 3;
955 	} else if (in_bytes > 0) {
956 		out_bytes = splice(pipefd[0], NULL, outfd, NULL, in_bytes,
957 				   SPLICE_F_MOVE | SPLICE_F_MORE);
958 		if (out_bytes < 0) {
959 			perror("splice out");
960 			err = 4;
961 		} else if (in_bytes != out_bytes) {
962 			fprintf(stderr, "Unexpected transfer: %zu vs %zu\n",
963 				in_bytes, out_bytes);
964 			err = 5;
965 		} else {
966 			goto again;
967 		}
968 	}
969 
970 	close(pipefd[0]);
971 	close(pipefd[1]);
972 
973 	return err;
974 }
975 
copyfd_io_splice(int infd,int peerfd,int outfd,unsigned int size,bool * in_closed_after_out,struct wstate * winfo)976 static int copyfd_io_splice(int infd, int peerfd, int outfd, unsigned int size,
977 			    bool *in_closed_after_out, struct wstate *winfo)
978 {
979 	int err;
980 
981 	if (listen_mode) {
982 		err = do_splice(peerfd, outfd, size, winfo);
983 		if (err)
984 			return err;
985 
986 		err = do_splice(infd, peerfd, size, winfo);
987 	} else {
988 		err = do_splice(infd, peerfd, size, winfo);
989 		if (err)
990 			return err;
991 
992 		shut_wr(peerfd);
993 
994 		err = do_splice(peerfd, outfd, size, winfo);
995 		*in_closed_after_out = true;
996 	}
997 
998 	return err;
999 }
1000 
copyfd_io(int infd,int peerfd,int outfd,bool close_peerfd,struct wstate * winfo)1001 static int copyfd_io(int infd, int peerfd, int outfd, bool close_peerfd, struct wstate *winfo)
1002 {
1003 	bool in_closed_after_out = false;
1004 	struct timespec start, end;
1005 	int file_size;
1006 	int ret;
1007 
1008 	if (cfg_time && (clock_gettime(CLOCK_MONOTONIC, &start) < 0))
1009 		xerror("can not fetch start time %d", errno);
1010 
1011 	switch (cfg_mode) {
1012 	case CFG_MODE_POLL:
1013 		ret = copyfd_io_poll(infd, peerfd, outfd, &in_closed_after_out,
1014 				     winfo);
1015 		break;
1016 
1017 	case CFG_MODE_MMAP:
1018 		file_size = get_infd_size(infd);
1019 		if (file_size < 0)
1020 			return file_size;
1021 		ret = copyfd_io_mmap(infd, peerfd, outfd, file_size,
1022 				     &in_closed_after_out, winfo);
1023 		break;
1024 
1025 	case CFG_MODE_SENDFILE:
1026 		file_size = get_infd_size(infd);
1027 		if (file_size < 0)
1028 			return file_size;
1029 		ret = copyfd_io_sendfile(infd, peerfd, outfd, file_size,
1030 					 &in_closed_after_out, winfo);
1031 		break;
1032 
1033 	case CFG_MODE_SPLICE:
1034 		file_size = get_infd_size(infd);
1035 		if (file_size < 0)
1036 			return file_size;
1037 		ret = copyfd_io_splice(infd, peerfd, outfd, file_size,
1038 				       &in_closed_after_out, winfo);
1039 		break;
1040 
1041 	default:
1042 		fprintf(stderr, "Invalid mode %d\n", cfg_mode);
1043 
1044 		die_usage();
1045 		return 1;
1046 	}
1047 
1048 	if (ret)
1049 		return ret;
1050 
1051 	if (close_peerfd)
1052 		close(peerfd);
1053 
1054 	if (cfg_time) {
1055 		unsigned int delta_ms;
1056 
1057 		if (clock_gettime(CLOCK_MONOTONIC, &end) < 0)
1058 			xerror("can not fetch end time %d", errno);
1059 		delta_ms = (end.tv_sec - start.tv_sec) * 1000 + (end.tv_nsec - start.tv_nsec) / 1000000;
1060 		if (delta_ms > cfg_time) {
1061 			xerror("transfer slower than expected! runtime %d ms, expected %d ms",
1062 			       delta_ms, cfg_time);
1063 		}
1064 
1065 		/* show the runtime only if this end shutdown(wr) before receiving the EOF,
1066 		 * (that is, if this end got the longer runtime)
1067 		 */
1068 		if (in_closed_after_out)
1069 			fprintf(stderr, "%d", delta_ms);
1070 	}
1071 
1072 	return 0;
1073 }
1074 
check_sockaddr(int pf,struct sockaddr_storage * ss,socklen_t salen)1075 static void check_sockaddr(int pf, struct sockaddr_storage *ss,
1076 			   socklen_t salen)
1077 {
1078 	struct sockaddr_in6 *sin6;
1079 	struct sockaddr_in *sin;
1080 	socklen_t wanted_size = 0;
1081 
1082 	switch (pf) {
1083 	case AF_INET:
1084 		wanted_size = sizeof(*sin);
1085 		sin = (void *)ss;
1086 		if (!sin->sin_port)
1087 			fprintf(stderr, "accept: something wrong: ip connection from port 0");
1088 		break;
1089 	case AF_INET6:
1090 		wanted_size = sizeof(*sin6);
1091 		sin6 = (void *)ss;
1092 		if (!sin6->sin6_port)
1093 			fprintf(stderr, "accept: something wrong: ipv6 connection from port 0");
1094 		break;
1095 	default:
1096 		fprintf(stderr, "accept: Unknown pf %d, salen %u\n", pf, salen);
1097 		return;
1098 	}
1099 
1100 	if (salen != wanted_size)
1101 		fprintf(stderr, "accept: size mismatch, got %d expected %d\n",
1102 			(int)salen, wanted_size);
1103 
1104 	if (ss->ss_family != pf)
1105 		fprintf(stderr, "accept: pf mismatch, expect %d, ss_family is %d\n",
1106 			(int)ss->ss_family, pf);
1107 }
1108 
check_getpeername(int fd,struct sockaddr_storage * ss,socklen_t salen)1109 static void check_getpeername(int fd, struct sockaddr_storage *ss, socklen_t salen)
1110 {
1111 	struct sockaddr_storage peerss;
1112 	socklen_t peersalen = sizeof(peerss);
1113 
1114 	if (getpeername(fd, (struct sockaddr *)&peerss, &peersalen) < 0) {
1115 		perror("getpeername");
1116 		return;
1117 	}
1118 
1119 	if (peersalen != salen) {
1120 		fprintf(stderr, "%s: %d vs %d\n", __func__, peersalen, salen);
1121 		return;
1122 	}
1123 
1124 	if (memcmp(ss, &peerss, peersalen)) {
1125 		char a[INET6_ADDRSTRLEN];
1126 		char b[INET6_ADDRSTRLEN];
1127 		char c[INET6_ADDRSTRLEN];
1128 		char d[INET6_ADDRSTRLEN];
1129 
1130 		xgetnameinfo((struct sockaddr *)ss, salen,
1131 			     a, sizeof(a), b, sizeof(b));
1132 
1133 		xgetnameinfo((struct sockaddr *)&peerss, peersalen,
1134 			     c, sizeof(c), d, sizeof(d));
1135 
1136 		fprintf(stderr, "%s: memcmp failure: accept %s vs peername %s, %s vs %s salen %d vs %d\n",
1137 			__func__, a, c, b, d, peersalen, salen);
1138 	}
1139 }
1140 
check_getpeername_connect(int fd)1141 static void check_getpeername_connect(int fd)
1142 {
1143 	struct sockaddr_storage ss;
1144 	socklen_t salen = sizeof(ss);
1145 	char a[INET6_ADDRSTRLEN];
1146 	char b[INET6_ADDRSTRLEN];
1147 	const char *iface;
1148 	size_t len;
1149 
1150 	if (getpeername(fd, (struct sockaddr *)&ss, &salen) < 0) {
1151 		perror("getpeername");
1152 		return;
1153 	}
1154 
1155 	xgetnameinfo((struct sockaddr *)&ss, salen,
1156 		     a, sizeof(a), b, sizeof(b));
1157 
1158 	iface = strchr(cfg_host, '%');
1159 	if (iface)
1160 		len = iface - cfg_host;
1161 	else
1162 		len = strlen(cfg_host) + 1;
1163 
1164 	if (strncmp(cfg_host, a, len) || strcmp(cfg_port, b))
1165 		fprintf(stderr, "%s: %s vs %s, %s vs %s\n", __func__,
1166 			cfg_host, a, cfg_port, b);
1167 }
1168 
maybe_close(int fd)1169 static void maybe_close(int fd)
1170 {
1171 	unsigned int r = rand();
1172 
1173 	if (!(cfg_join || cfg_remove || cfg_repeat > 1) && (r & 1))
1174 		close(fd);
1175 }
1176 
main_loop_s(int listensock)1177 int main_loop_s(int listensock)
1178 {
1179 	struct sockaddr_storage ss;
1180 	struct wstate winfo;
1181 	struct pollfd polls;
1182 	socklen_t salen;
1183 	int remotesock;
1184 	int err = 0;
1185 	int fd = 0;
1186 
1187 again:
1188 	polls.fd = listensock;
1189 	polls.events = POLLIN;
1190 
1191 	switch (poll(&polls, 1, poll_timeout)) {
1192 	case -1:
1193 		perror("poll");
1194 		return 1;
1195 	case 0:
1196 		fprintf(stderr, "%s: timed out\n", __func__);
1197 		close(listensock);
1198 		return 2;
1199 	}
1200 
1201 	salen = sizeof(ss);
1202 	remotesock = accept(listensock, (struct sockaddr *)&ss, &salen);
1203 	if (remotesock >= 0) {
1204 		maybe_close(listensock);
1205 		check_sockaddr(pf, &ss, salen);
1206 		check_getpeername(remotesock, &ss, salen);
1207 
1208 		if (cfg_input) {
1209 			fd = open(cfg_input, O_RDONLY);
1210 			if (fd < 0)
1211 				xerror("can't open %s: %d", cfg_input, errno);
1212 		}
1213 
1214 		SOCK_TEST_TCPULP(remotesock, 0);
1215 
1216 		memset(&winfo, 0, sizeof(winfo));
1217 		err = copyfd_io(fd, remotesock, 1, true, &winfo);
1218 	} else {
1219 		perror("accept");
1220 		return 1;
1221 	}
1222 
1223 	if (cfg_input)
1224 		close(fd);
1225 
1226 	if (!err && --cfg_repeat > 0)
1227 		goto again;
1228 
1229 	return err;
1230 }
1231 
init_rng(void)1232 static void init_rng(void)
1233 {
1234 	unsigned int foo;
1235 
1236 	if (getrandom(&foo, sizeof(foo), 0) == -1) {
1237 		perror("getrandom");
1238 		exit(1);
1239 	}
1240 
1241 	srand(foo);
1242 }
1243 
xsetsockopt(int fd,int level,int optname,const void * optval,socklen_t optlen)1244 static void xsetsockopt(int fd, int level, int optname, const void *optval, socklen_t optlen)
1245 {
1246 	int err;
1247 
1248 	err = setsockopt(fd, level, optname, optval, optlen);
1249 	if (err) {
1250 		perror("setsockopt");
1251 		exit(1);
1252 	}
1253 }
1254 
apply_cmsg_types(int fd,const struct cfg_cmsg_types * cmsg)1255 static void apply_cmsg_types(int fd, const struct cfg_cmsg_types *cmsg)
1256 {
1257 	static const unsigned int on = 1;
1258 
1259 	if (cmsg->timestampns)
1260 		xsetsockopt(fd, SOL_SOCKET, SO_TIMESTAMPNS_NEW, &on, sizeof(on));
1261 	if (cmsg->tcp_inq)
1262 		xsetsockopt(fd, IPPROTO_TCP, TCP_INQ, &on, sizeof(on));
1263 }
1264 
parse_cmsg_types(const char * type)1265 static void parse_cmsg_types(const char *type)
1266 {
1267 	char *next = strchr(type, ',');
1268 	unsigned int len = 0;
1269 
1270 	cfg_cmsg_types.cmsg_enabled = 1;
1271 
1272 	if (next) {
1273 		parse_cmsg_types(next + 1);
1274 		len = next - type;
1275 	} else {
1276 		len = strlen(type);
1277 	}
1278 
1279 	if (strncmp(type, "TIMESTAMPNS", len) == 0) {
1280 		cfg_cmsg_types.timestampns = 1;
1281 		return;
1282 	}
1283 
1284 	if (strncmp(type, "TCPINQ", len) == 0) {
1285 		cfg_cmsg_types.tcp_inq = 1;
1286 		return;
1287 	}
1288 
1289 	fprintf(stderr, "Unrecognized cmsg option %s\n", type);
1290 	exit(1);
1291 }
1292 
parse_setsock_options(const char * name)1293 static void parse_setsock_options(const char *name)
1294 {
1295 	char *next = strchr(name, ',');
1296 	unsigned int len = 0;
1297 
1298 	if (next) {
1299 		parse_setsock_options(next + 1);
1300 		len = next - name;
1301 	} else {
1302 		len = strlen(name);
1303 	}
1304 
1305 	if (strncmp(name, "TRANSPARENT", len) == 0) {
1306 		cfg_sockopt_types.transparent = 1;
1307 		return;
1308 	}
1309 
1310 	if (strncmp(name, "MPTFO", len) == 0) {
1311 		cfg_sockopt_types.mptfo = 1;
1312 		return;
1313 	}
1314 
1315 	fprintf(stderr, "Unrecognized setsockopt option %s\n", name);
1316 	exit(1);
1317 }
1318 
xdisconnect(int fd)1319 void xdisconnect(int fd)
1320 {
1321 	socklen_t addrlen = sizeof(struct sockaddr_storage);
1322 	struct sockaddr_storage addr, empty;
1323 	int msec_sleep = 10;
1324 	void *raw_addr;
1325 	int i, cmdlen;
1326 	char cmd[128];
1327 
1328 	/* get the local address and convert it to string */
1329 	if (getsockname(fd, (struct sockaddr *)&addr, &addrlen) < 0)
1330 		xerror("getsockname");
1331 
1332 	if (addr.ss_family == AF_INET)
1333 		raw_addr = &(((struct sockaddr_in *)&addr)->sin_addr);
1334 	else if (addr.ss_family == AF_INET6)
1335 		raw_addr = &(((struct sockaddr_in6 *)&addr)->sin6_addr);
1336 	else
1337 		xerror("bad family");
1338 
1339 	strcpy(cmd, "ss -Mnt | grep -q ");
1340 	cmdlen = strlen(cmd);
1341 	if (!inet_ntop(addr.ss_family, raw_addr, &cmd[cmdlen],
1342 		       sizeof(cmd) - cmdlen))
1343 		xerror("inet_ntop");
1344 
1345 	shutdown(fd, SHUT_WR);
1346 
1347 	/*
1348 	 * wait until the pending data is completely flushed and all
1349 	 * the sockets reached the closed status.
1350 	 * disconnect will bypass/ignore/drop any pending data.
1351 	 */
1352 	for (i = 0; ; i += msec_sleep) {
1353 		/* closed socket are not listed by 'ss' */
1354 		if (system(cmd) != 0)
1355 			break;
1356 
1357 		if (i > poll_timeout)
1358 			xerror("timeout while waiting for spool to complete");
1359 		usleep(msec_sleep * 1000);
1360 	}
1361 
1362 	memset(&empty, 0, sizeof(empty));
1363 	empty.ss_family = AF_UNSPEC;
1364 	if (connect(fd, (struct sockaddr *)&empty, addrlen) < 0)
1365 		xerror("can't disconnect: %d", errno);
1366 }
1367 
main_loop(void)1368 int main_loop(void)
1369 {
1370 	struct addrinfo *peer = NULL;
1371 	int fd = 0, ret, fd_in = 0;
1372 	struct wstate winfo;
1373 
1374 	if (cfg_input && cfg_sockopt_types.mptfo) {
1375 		fd_in = open(cfg_input, O_RDONLY);
1376 		if (fd_in < 0)
1377 			xerror("can't open %s:%d", cfg_input, errno);
1378 	}
1379 
1380 	memset(&winfo, 0, sizeof(winfo));
1381 	fd = sock_connect_mptcp(cfg_host, cfg_port, cfg_sock_proto, &peer, fd_in, &winfo);
1382 	if (fd < 0)
1383 		return 2;
1384 
1385 again:
1386 	check_getpeername_connect(fd);
1387 
1388 	SOCK_TEST_TCPULP(fd, cfg_sock_proto);
1389 
1390 	if (cfg_rcvbuf)
1391 		set_rcvbuf(fd, cfg_rcvbuf);
1392 	if (cfg_sndbuf)
1393 		set_sndbuf(fd, cfg_sndbuf);
1394 	if (cfg_cmsg_types.cmsg_enabled)
1395 		apply_cmsg_types(fd, &cfg_cmsg_types);
1396 
1397 	if (cfg_input && !cfg_sockopt_types.mptfo) {
1398 		fd_in = open(cfg_input, O_RDONLY);
1399 		if (fd_in < 0)
1400 			xerror("can't open %s:%d", cfg_input, errno);
1401 	}
1402 
1403 	ret = copyfd_io(fd_in, fd, 1, 0, &winfo);
1404 	if (ret)
1405 		goto out;
1406 
1407 	if (cfg_truncate > 0) {
1408 		shutdown(fd, SHUT_WR);
1409 	} else if (--cfg_repeat > 0) {
1410 		xdisconnect(fd);
1411 
1412 		/* the socket could be unblocking at this point, we need the
1413 		 * connect to be blocking
1414 		 */
1415 		set_nonblock(fd, false);
1416 		if (connect(fd, peer->ai_addr, peer->ai_addrlen))
1417 			xerror("can't reconnect: %d", errno);
1418 		if (cfg_input)
1419 			close(fd_in);
1420 		memset(&winfo, 0, sizeof(winfo));
1421 		goto again;
1422 	} else {
1423 		close(fd);
1424 	}
1425 
1426 out:
1427 	if (cfg_input)
1428 		close(fd_in);
1429 	return ret;
1430 }
1431 
parse_proto(const char * proto)1432 int parse_proto(const char *proto)
1433 {
1434 	if (!strcasecmp(proto, "MPTCP"))
1435 		return IPPROTO_MPTCP;
1436 	if (!strcasecmp(proto, "TCP"))
1437 		return IPPROTO_TCP;
1438 
1439 	fprintf(stderr, "Unknown protocol: %s\n.", proto);
1440 	die_usage();
1441 
1442 	/* silence compiler warning */
1443 	return 0;
1444 }
1445 
parse_mode(const char * mode)1446 int parse_mode(const char *mode)
1447 {
1448 	if (!strcasecmp(mode, "poll"))
1449 		return CFG_MODE_POLL;
1450 	if (!strcasecmp(mode, "mmap"))
1451 		return CFG_MODE_MMAP;
1452 	if (!strcasecmp(mode, "sendfile"))
1453 		return CFG_MODE_SENDFILE;
1454 	if (!strcasecmp(mode, "splice"))
1455 		return CFG_MODE_SPLICE;
1456 
1457 	fprintf(stderr, "Unknown test mode: %s\n", mode);
1458 	fprintf(stderr, "Supported modes are:\n");
1459 	fprintf(stderr, "\t\t\"poll\" - interleaved read/write using poll()\n");
1460 	fprintf(stderr, "\t\t\"mmap\" - send entire input file (mmap+write), then read response (-l will read input first)\n");
1461 	fprintf(stderr, "\t\t\"sendfile\" - send entire input file (sendfile), then read response (-l will read input first)\n");
1462 	fprintf(stderr, "\t\t\"splice\" - send entire input file (splice), then read response (-l will read input first)\n");
1463 
1464 	die_usage();
1465 
1466 	/* silence compiler warning */
1467 	return 0;
1468 }
1469 
parse_peek(const char * mode)1470 int parse_peek(const char *mode)
1471 {
1472 	if (!strcasecmp(mode, "saveWithPeek"))
1473 		return CFG_WITH_PEEK;
1474 	if (!strcasecmp(mode, "saveAfterPeek"))
1475 		return CFG_AFTER_PEEK;
1476 
1477 	fprintf(stderr, "Unknown: %s\n", mode);
1478 	fprintf(stderr, "Supported MSG_PEEK mode are:\n");
1479 	fprintf(stderr,
1480 		"\t\t\"saveWithPeek\" - recv data with flags 'MSG_PEEK' and save the peek data into file\n");
1481 	fprintf(stderr,
1482 		"\t\t\"saveAfterPeek\" - read and save data into file after recv with flags 'MSG_PEEK'\n");
1483 
1484 	die_usage();
1485 
1486 	/* silence compiler warning */
1487 	return 0;
1488 }
1489 
parse_int(const char * size)1490 static int parse_int(const char *size)
1491 {
1492 	unsigned long s;
1493 
1494 	errno = 0;
1495 
1496 	s = strtoul(size, NULL, 0);
1497 
1498 	if (errno) {
1499 		fprintf(stderr, "Invalid sndbuf size %s (%s)\n",
1500 			size, strerror(errno));
1501 		die_usage();
1502 	}
1503 
1504 	if (s > INT_MAX) {
1505 		fprintf(stderr, "Invalid sndbuf size %s (%s)\n",
1506 			size, strerror(ERANGE));
1507 		die_usage();
1508 	}
1509 
1510 	return (int)s;
1511 }
1512 
parse_opts(int argc,char ** argv)1513 static void parse_opts(int argc, char **argv)
1514 {
1515 	int c;
1516 
1517 	while ((c = getopt(argc, argv, "6c:f:hi:I:jlm:M:o:p:P:r:R:s:S:t:T:w:")) != -1) {
1518 		switch (c) {
1519 		case 'f':
1520 			cfg_truncate = atoi(optarg);
1521 
1522 			/* when receiving a fastclose, ignore PIPE signals and
1523 			 * all the I/O errors later in the code
1524 			 */
1525 			if (cfg_truncate < 0) {
1526 				cfg_rcv_trunc = true;
1527 				signal(SIGPIPE, SIG_IGN);
1528 			}
1529 			break;
1530 		case 'j':
1531 			cfg_join = true;
1532 			cfg_mode = CFG_MODE_POLL;
1533 			break;
1534 		case 'r':
1535 			cfg_remove = true;
1536 			cfg_mode = CFG_MODE_POLL;
1537 			cfg_wait = 400000;
1538 			cfg_do_w = atoi(optarg);
1539 			if (cfg_do_w <= 0)
1540 				cfg_do_w = 50;
1541 			break;
1542 		case 'i':
1543 			cfg_input = optarg;
1544 			break;
1545 		case 'I':
1546 			cfg_repeat = atoi(optarg);
1547 			break;
1548 		case 'l':
1549 			listen_mode = true;
1550 			break;
1551 		case 'p':
1552 			cfg_port = optarg;
1553 			break;
1554 		case 's':
1555 			cfg_sock_proto = parse_proto(optarg);
1556 			break;
1557 		case 'h':
1558 			die_usage();
1559 			break;
1560 		case '6':
1561 			pf = AF_INET6;
1562 			break;
1563 		case 't':
1564 			poll_timeout = atoi(optarg) * 1000;
1565 			if (poll_timeout <= 0)
1566 				poll_timeout = -1;
1567 			break;
1568 		case 'T':
1569 			cfg_time = atoi(optarg);
1570 			break;
1571 		case 'm':
1572 			cfg_mode = parse_mode(optarg);
1573 			break;
1574 		case 'S':
1575 			cfg_sndbuf = parse_int(optarg);
1576 			break;
1577 		case 'R':
1578 			cfg_rcvbuf = parse_int(optarg);
1579 			break;
1580 		case 'w':
1581 			cfg_wait = atoi(optarg)*1000000;
1582 			break;
1583 		case 'M':
1584 			cfg_mark = strtol(optarg, NULL, 0);
1585 			break;
1586 		case 'P':
1587 			cfg_peek = parse_peek(optarg);
1588 			break;
1589 		case 'c':
1590 			parse_cmsg_types(optarg);
1591 			break;
1592 		case 'o':
1593 			parse_setsock_options(optarg);
1594 			break;
1595 		}
1596 	}
1597 
1598 	if (optind + 1 != argc)
1599 		die_usage();
1600 	cfg_host = argv[optind];
1601 
1602 	if (strchr(cfg_host, ':'))
1603 		pf = AF_INET6;
1604 }
1605 
main(int argc,char * argv[])1606 int main(int argc, char *argv[])
1607 {
1608 	init_rng();
1609 
1610 	signal(SIGUSR1, handle_signal);
1611 	parse_opts(argc, argv);
1612 
1613 	if (listen_mode) {
1614 		int fd = sock_listen_mptcp(cfg_host, cfg_port);
1615 
1616 		if (fd < 0)
1617 			return 1;
1618 
1619 		if (cfg_rcvbuf)
1620 			set_rcvbuf(fd, cfg_rcvbuf);
1621 		if (cfg_sndbuf)
1622 			set_sndbuf(fd, cfg_sndbuf);
1623 		if (cfg_mark)
1624 			set_mark(fd, cfg_mark);
1625 		if (cfg_cmsg_types.cmsg_enabled)
1626 			apply_cmsg_types(fd, &cfg_cmsg_types);
1627 
1628 		return main_loop_s(fd);
1629 	}
1630 
1631 	return main_loop();
1632 }
1633