xref: /openbsd-src/usr.sbin/unbound/testcode/delayer.c (revision 0e9b6f9f434b93ba2f11e3b38a3da379b202585e)
1 /*
2  * testcode/delayer.c - debug program that delays queries to a server.
3  *
4  * Copyright (c) 2008, NLnet Labs. All rights reserved.
5  *
6  * This software is open source.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  *
12  * Redistributions of source code must retain the above copyright notice,
13  * this list of conditions and the following disclaimer.
14  *
15  * Redistributions in binary form must reproduce the above copyright notice,
16  * this list of conditions and the following disclaimer in the documentation
17  * and/or other materials provided with the distribution.
18  *
19  * Neither the name of the NLNET LABS nor the names of its contributors may
20  * be used to endorse or promote products derived from this software without
21  * specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
24  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
25  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
26  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
27  * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
28  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
29  * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
30  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34  */
35 
36 /**
37  * \file
38  *
39  * This program delays queries made. It performs as a proxy to another
40  * server and delays queries to it.
41  */
42 
43 #include "config.h"
44 #ifdef HAVE_GETOPT_H
45 #include <getopt.h>
46 #endif
47 #ifdef HAVE_TIME_H
48 #include <time.h>
49 #endif
50 #include <sys/time.h>
51 #include "util/net_help.h"
52 #include "util/config_file.h"
53 #include "sldns/sbuffer.h"
54 #include <signal.h>
55 
56 /** number of reads per select for delayer */
57 #define TRIES_PER_SELECT 100
58 
59 /**
60  * The ring buffer
61  */
62 struct ringbuf {
63 	/** base of buffer */
64 	uint8_t* buf;
65 	/** size of buffer */
66 	size_t size;
67 	/** low mark, items start here */
68 	size_t low;
69 	/** high mark, items end here */
70 	size_t high;
71 };
72 
73 /**
74  * List of proxy fds that return replies from the server to our clients.
75  */
76 struct proxy {
77 	/** the fd to listen for replies from server */
78 	int s;
79 	/** last time this was used */
80 	struct timeval lastuse;
81 	/** remote address */
82 	struct sockaddr_storage addr;
83 	/** length of addr */
84 	socklen_t addr_len;
85 	/** number of queries waiting (in total) */
86 	size_t numwait;
87 	/** number of queries sent to server (in total) */
88 	size_t numsent;
89 	/** numberof answers returned to client (in total) */
90 	size_t numreturn;
91 	/** how many times repurposed */
92 	size_t numreuse;
93 	/** next in proxylist */
94 	struct proxy* next;
95 };
96 
97 /**
98  * An item that has to be TCP relayed
99  */
100 struct tcp_send_list {
101 	/** the data item */
102 	uint8_t* item;
103 	/** size of item */
104 	size_t len;
105 	/** time when the item can be transmitted on */
106 	struct timeval wait;
107 	/** how much of the item has already been transmitted */
108 	size_t done;
109 	/** next in list */
110 	struct tcp_send_list* next;
111 };
112 
113 /**
114  * List of TCP proxy fd pairs to TCP connect client to server
115  */
116 struct tcp_proxy {
117 	/** the fd to listen for client query */
118 	int client_s;
119 	/** the fd to listen for server answer */
120 	int server_s;
121 
122 	/** remote client address */
123 	struct sockaddr_storage addr;
124 	/** length of address */
125 	socklen_t addr_len;
126 	/** timeout on this entry */
127 	struct timeval timeout;
128 
129 	/** list of query items to send to server */
130 	struct tcp_send_list* querylist;
131 	/** last in query list */
132 	struct tcp_send_list* querylast;
133 	/** list of answer items to send to client */
134 	struct tcp_send_list* answerlist;
135 	/** last in answerlist */
136 	struct tcp_send_list* answerlast;
137 
138 	/** next in list */
139 	struct tcp_proxy* next;
140 };
141 
142 /** usage information for delayer */
usage(char * argv[])143 static void usage(char* argv[])
144 {
145 	printf("usage: %s [options]\n", argv[0]);
146 	printf("	-f addr : use addr, forward to that server, @port.\n");
147 	printf("	-b addr : bind to this address to listen.\n");
148 	printf("	-p port : bind to this port (use 0 for random).\n");
149 	printf("	-m mem	: use this much memory for waiting queries.\n");
150 	printf("	-d delay: UDP queries are delayed n milliseconds.\n");
151 	printf("		  TCP is delayed twice (on send, on recv).\n");
152 	printf("	-h 	: this help message\n");
153 	exit(1);
154 }
155 
156 /** timeval compare, t1 < t2 */
157 static int
dl_tv_smaller(struct timeval * t1,const struct timeval * t2)158 dl_tv_smaller(struct timeval* t1, const struct timeval* t2)
159 {
160 #ifndef S_SPLINT_S
161 	if(t1->tv_sec < t2->tv_sec)
162 		return 1;
163 	if(t1->tv_sec == t2->tv_sec &&
164 		t1->tv_usec < t2->tv_usec)
165 		return 1;
166 #endif
167 	return 0;
168 }
169 
170 /** timeval add, t1 += t2 */
171 static void
dl_tv_add(struct timeval * t1,const struct timeval * t2)172 dl_tv_add(struct timeval* t1, const struct timeval* t2)
173 {
174 #ifndef S_SPLINT_S
175 	t1->tv_sec += t2->tv_sec;
176 	t1->tv_usec += t2->tv_usec;
177 	while(t1->tv_usec >= 1000000) {
178 		t1->tv_usec -= 1000000;
179 		t1->tv_sec++;
180 	}
181 #endif
182 }
183 
184 /** timeval subtract, t1 -= t2 */
185 static void
dl_tv_subtract(struct timeval * t1,const struct timeval * t2)186 dl_tv_subtract(struct timeval* t1, const struct timeval* t2)
187 {
188 #ifndef S_SPLINT_S
189 	t1->tv_sec -= t2->tv_sec;
190 	if(t1->tv_usec >= t2->tv_usec) {
191 		t1->tv_usec -= t2->tv_usec;
192 	} else {
193 		t1->tv_sec--;
194 		t1->tv_usec = 1000000-(t2->tv_usec-t1->tv_usec);
195 	}
196 #endif
197 }
198 
199 
200 /** create new ring buffer */
201 static struct ringbuf*
ring_create(size_t sz)202 ring_create(size_t sz)
203 {
204 	struct ringbuf* r = (struct ringbuf*)calloc(1, sizeof(*r));
205 	if(!r) fatal_exit("out of memory");
206 	r->buf = (uint8_t*)malloc(sz);
207 	if(!r->buf) fatal_exit("out of memory");
208 	r->size = sz;
209 	r->low = 0;
210 	r->high = 0;
211 	return r;
212 }
213 
214 /** delete ring buffer */
215 static void
ring_delete(struct ringbuf * r)216 ring_delete(struct ringbuf* r)
217 {
218 	if(!r) return;
219 	free(r->buf);
220 	free(r);
221 }
222 
223 /** add entry to ringbuffer */
224 static void
ring_add(struct ringbuf * r,sldns_buffer * pkt,struct timeval * now,struct timeval * delay,struct proxy * p)225 ring_add(struct ringbuf* r, sldns_buffer* pkt, struct timeval* now,
226 	struct timeval* delay, struct proxy* p)
227 {
228 	/* time -- proxy* -- 16bitlen -- message */
229 	uint16_t len = (uint16_t)sldns_buffer_limit(pkt);
230 	struct timeval when;
231 	size_t needed;
232 	uint8_t* where = NULL;
233 	log_assert(sldns_buffer_limit(pkt) <= 65535);
234 	needed = sizeof(when) + sizeof(p) + sizeof(len) + len;
235 	/* put item into ringbuffer */
236 	if(r->low < r->high) {
237 		/* used part is in the middle */
238 		if(r->size - r->high >= needed) {
239 			where = r->buf + r->high;
240 			r->high += needed;
241 		} else if(r->low > needed) {
242 			/* wrap around ringbuffer */
243 			/* make sure r->low == r->high means empty */
244 			/* so r->low == r->high cannot be used to signify
245 			 * a completely full ringbuf */
246 			if(r->size - r->high > sizeof(when)+sizeof(p)) {
247 				/* zero entry at end of buffer */
248 				memset(r->buf+r->high, 0,
249 					sizeof(when)+sizeof(p));
250 			}
251 			where = r->buf;
252 			r->high = needed;
253 		} else {
254 			/* drop message */
255 			log_warn("warning: mem full, dropped message");
256 			return;
257 		}
258 	} else {
259 		/* empty */
260 		if(r->high == r->low) {
261 			where = r->buf;
262 			r->low = 0;
263 			r->high = needed;
264 		/* unused part is in the middle */
265 		/* so ringbuffer has wrapped around */
266 		} else if(r->low - r->high > needed) {
267 			where = r->buf + r->high;
268 			r->high += needed;
269 		} else {
270 			log_warn("warning: mem full, dropped message");
271 			return;
272 		}
273 	}
274 	when = *now;
275 	dl_tv_add(&when, delay);
276 	/* copy it at where part */
277 	log_assert(where != NULL);
278 	memmove(where, &when, sizeof(when));
279 	memmove(where+sizeof(when), &p, sizeof(p));
280 	memmove(where+sizeof(when)+sizeof(p), &len, sizeof(len));
281 	memmove(where+sizeof(when)+sizeof(p)+sizeof(len),
282 		sldns_buffer_begin(pkt), len);
283 }
284 
285 /** see if the ringbuffer is empty */
286 static int
ring_empty(struct ringbuf * r)287 ring_empty(struct ringbuf* r)
288 {
289 	return (r->low == r->high);
290 }
291 
292 /** peek at timevalue for next item in ring */
293 static struct timeval*
ring_peek_time(struct ringbuf * r)294 ring_peek_time(struct ringbuf* r)
295 {
296 	if(ring_empty(r))
297 		return NULL;
298 	return (struct timeval*)&r->buf[r->low];
299 }
300 
301 /** get entry from ringbuffer */
302 static int
ring_pop(struct ringbuf * r,sldns_buffer * pkt,struct timeval * tv,struct proxy ** p)303 ring_pop(struct ringbuf* r, sldns_buffer* pkt, struct timeval* tv,
304 	struct proxy** p)
305 {
306 	/* time -- proxy* -- 16bitlen -- message */
307 	uint16_t len;
308 	uint8_t* where = NULL;
309 	size_t done;
310 	if(r->low == r->high)
311 		return 0;
312 	where = r->buf + r->low;
313 	memmove(tv, where, sizeof(*tv));
314 	memmove(p, where+sizeof(*tv), sizeof(*p));
315 	memmove(&len, where+sizeof(*tv)+sizeof(*p), sizeof(len));
316 	memmove(sldns_buffer_begin(pkt),
317 		where+sizeof(*tv)+sizeof(*p)+sizeof(len), len);
318 	sldns_buffer_set_limit(pkt, (size_t)len);
319 	done = sizeof(*tv)+sizeof(*p)+sizeof(len)+len;
320 	/* move lowmark */
321 	if(r->low < r->high) {
322 		/* used part in middle */
323 		log_assert(r->high - r->low >= done);
324 		r->low += done;
325 	} else {
326 		/* unused part in middle */
327 		log_assert(r->size - r->low >= done);
328 		r->low += done;
329 		if(r->size - r->low > sizeof(*tv)+sizeof(*p)) {
330 			/* see if it is zeroed; means end of buffer */
331 			struct proxy* pz;
332 			memmove(&pz, r->buf+r->low+sizeof(*tv), sizeof(pz));
333 			if(pz == NULL)
334 				r->low = 0;
335 		} else r->low = 0;
336 	}
337 	if(r->low == r->high) {
338 		r->low = 0; /* reset if empty */
339 		r->high = 0;
340 	}
341 	return 1;
342 }
343 
344 /** signal handler global info */
345 static volatile int do_quit = 0;
346 
347 /** signal handler for user quit */
delayer_sigh(int sig)348 static RETSIGTYPE delayer_sigh(int sig)
349 {
350 	char str[] = "exit on signal   \n";
351 	str[15] = '0' + (sig/10)%10;
352 	str[16] = '0' + sig%10;
353 	/* simple cast to void will not silence Wunused-result */
354 	(void)!write(STDOUT_FILENO, str, strlen(str));
355 	do_quit = 1;
356 }
357 
358 /** send out waiting packets */
359 static void
service_send(struct ringbuf * ring,struct timeval * now,sldns_buffer * pkt,struct sockaddr_storage * srv_addr,socklen_t srv_len)360 service_send(struct ringbuf* ring, struct timeval* now, sldns_buffer* pkt,
361 	struct sockaddr_storage* srv_addr, socklen_t srv_len)
362 {
363 	struct proxy* p;
364 	struct timeval tv;
365 	ssize_t sent;
366 	while(!ring_empty(ring) &&
367 		dl_tv_smaller(ring_peek_time(ring), now)) {
368 		/* this items needs to be sent out */
369 		if(!ring_pop(ring, pkt, &tv, &p))
370 			fatal_exit("ringbuf error: pop failed");
371 		verbose(1, "send out query %d.%6.6d",
372 			(unsigned)tv.tv_sec, (unsigned)tv.tv_usec);
373 		log_addr(1, "from client", &p->addr, p->addr_len);
374 		/* send it */
375 		sent = sendto(p->s, (void*)sldns_buffer_begin(pkt),
376 			sldns_buffer_limit(pkt), 0,
377 			(struct sockaddr*)srv_addr, srv_len);
378 		if(sent == -1) {
379 			log_err("sendto: %s", sock_strerror(errno));
380 		} else if(sent != (ssize_t)sldns_buffer_limit(pkt)) {
381 			log_err("sendto: partial send");
382 		}
383 		p->lastuse = *now;
384 		p->numsent++;
385 	}
386 }
387 
388 /** do proxy for one readable client */
389 static void
do_proxy(struct proxy * p,int retsock,sldns_buffer * pkt)390 do_proxy(struct proxy* p, int retsock, sldns_buffer* pkt)
391 {
392 	int i;
393 	ssize_t r;
394 	for(i=0; i<TRIES_PER_SELECT; i++) {
395 		r = recv(p->s, (void*)sldns_buffer_begin(pkt),
396 			sldns_buffer_capacity(pkt), 0);
397 		if(r == -1) {
398 #ifndef USE_WINSOCK
399 			if(errno == EAGAIN || errno == EINTR)
400 				return;
401 #else
402 			if(WSAGetLastError() == WSAEINPROGRESS ||
403 				WSAGetLastError() == WSAEWOULDBLOCK)
404 				return;
405 #endif
406 			log_err("recv: %s", sock_strerror(errno));
407 			return;
408 		}
409 		sldns_buffer_set_limit(pkt, (size_t)r);
410 		log_addr(1, "return reply to client", &p->addr, p->addr_len);
411 		/* send reply back to the real client */
412 		p->numreturn++;
413 		r = sendto(retsock, (void*)sldns_buffer_begin(pkt), (size_t)r,
414 			0, (struct sockaddr*)&p->addr, p->addr_len);
415 		if(r == -1) {
416 			log_err("sendto: %s", sock_strerror(errno));
417 		}
418 	}
419 }
420 
421 /** proxy return replies to clients */
422 static void
service_proxy(fd_set * rset,int retsock,struct proxy * proxies,sldns_buffer * pkt,struct timeval * now)423 service_proxy(fd_set* rset, int retsock, struct proxy* proxies,
424 	sldns_buffer* pkt, struct timeval* now)
425 {
426 	struct proxy* p;
427 	for(p = proxies; p; p = p->next) {
428 		if(FD_ISSET(p->s, rset)) {
429 			p->lastuse = *now;
430 			do_proxy(p, retsock, pkt);
431 		}
432 	}
433 }
434 
435 /** find or else create proxy for this remote client */
436 static struct proxy*
find_create_proxy(struct sockaddr_storage * from,socklen_t from_len,fd_set * rorig,int * max,struct proxy ** proxies,int serv_ip6,struct timeval * now,struct timeval * reuse_timeout)437 find_create_proxy(struct sockaddr_storage* from, socklen_t from_len,
438 	fd_set* rorig, int* max, struct proxy** proxies, int serv_ip6,
439 	struct timeval* now, struct timeval* reuse_timeout)
440 {
441 	struct proxy* p;
442 	struct timeval t;
443 	for(p = *proxies; p; p = p->next) {
444 		if(sockaddr_cmp(from, from_len, &p->addr, p->addr_len)==0)
445 			return p;
446 	}
447 	/* possibly: reuse lapsed entries */
448 	for(p = *proxies; p; p = p->next) {
449 		if(p->numwait > p->numsent || p->numsent > p->numreturn)
450 			continue;
451 		t = *now;
452 		dl_tv_subtract(&t, &p->lastuse);
453 		if(dl_tv_smaller(&t, reuse_timeout))
454 			continue;
455 		/* yes! */
456 		verbose(1, "reuse existing entry");
457 		memmove(&p->addr, from, from_len);
458 		p->addr_len = from_len;
459 		p->numreuse++;
460 		return p;
461 	}
462 	/* create new */
463 	p = (struct proxy*)calloc(1, sizeof(*p));
464 	if(!p) fatal_exit("out of memory");
465 	p->s = socket(serv_ip6?AF_INET6:AF_INET, SOCK_DGRAM, 0);
466 	if(p->s == -1) {
467 		fatal_exit("socket: %s", sock_strerror(errno));
468 	}
469 	fd_set_nonblock(p->s);
470 	memmove(&p->addr, from, from_len);
471 	p->addr_len = from_len;
472 	p->next = *proxies;
473 	*proxies = p;
474 	FD_SET(FD_SET_T p->s, rorig);
475 	if(p->s+1 > *max)
476 		*max = p->s+1;
477 	return p;
478 }
479 
480 /** recv new waiting packets */
481 static void
service_recv(int s,struct ringbuf * ring,sldns_buffer * pkt,fd_set * rorig,int * max,struct proxy ** proxies,struct sockaddr_storage * srv_addr,socklen_t srv_len,struct timeval * now,struct timeval * delay,struct timeval * reuse)482 service_recv(int s, struct ringbuf* ring, sldns_buffer* pkt,
483 	fd_set* rorig, int* max, struct proxy** proxies,
484 	struct sockaddr_storage* srv_addr, socklen_t srv_len,
485 	struct timeval* now, struct timeval* delay, struct timeval* reuse)
486 {
487 	int i;
488 	struct sockaddr_storage from;
489 	socklen_t from_len;
490 	ssize_t len;
491 	struct proxy* p;
492 	for(i=0; i<TRIES_PER_SELECT; i++) {
493 		from_len = (socklen_t)sizeof(from);
494 		len = recvfrom(s, (void*)sldns_buffer_begin(pkt),
495 			sldns_buffer_capacity(pkt), 0,
496 			(struct sockaddr*)&from, &from_len);
497 		if(len < 0) {
498 #ifndef USE_WINSOCK
499 			if(errno == EAGAIN || errno == EINTR)
500 				return;
501 #else
502 			if(WSAGetLastError() == WSAEWOULDBLOCK ||
503 				WSAGetLastError() == WSAEINPROGRESS)
504 				return;
505 #endif
506 			fatal_exit("recvfrom: %s", sock_strerror(errno));
507 		}
508 		sldns_buffer_set_limit(pkt, (size_t)len);
509 		/* find its proxy element */
510 		p = find_create_proxy(&from, from_len, rorig, max, proxies,
511 			addr_is_ip6(srv_addr, srv_len), now, reuse);
512 		if(!p) fatal_exit("error: cannot find or create proxy");
513 		p->lastuse = *now;
514 		ring_add(ring, pkt, now, delay, p);
515 		p->numwait++;
516 		log_addr(1, "recv from client", &p->addr, p->addr_len);
517 	}
518 }
519 
520 /** delete tcp proxy */
521 static void
tcp_proxy_delete(struct tcp_proxy * p)522 tcp_proxy_delete(struct tcp_proxy* p)
523 {
524 	struct tcp_send_list* s, *sn;
525 	if(!p)
526 		return;
527 	log_addr(1, "delete tcp proxy", &p->addr, p->addr_len);
528 	s = p->querylist;
529 	while(s) {
530 		sn = s->next;
531 		free(s->item);
532 		free(s);
533 		s = sn;
534 	}
535 	s = p->answerlist;
536 	while(s) {
537 		sn = s->next;
538 		free(s->item);
539 		free(s);
540 		s = sn;
541 	}
542 	sock_close(p->client_s);
543 	if(p->server_s != -1)
544 		sock_close(p->server_s);
545 	free(p);
546 }
547 
548 /** accept new TCP connections, and set them up */
549 static void
service_tcp_listen(int s,fd_set * rorig,int * max,struct tcp_proxy ** proxies,struct sockaddr_storage * srv_addr,socklen_t srv_len,struct timeval * now,struct timeval * tcp_timeout)550 service_tcp_listen(int s, fd_set* rorig, int* max, struct tcp_proxy** proxies,
551 	struct sockaddr_storage* srv_addr, socklen_t srv_len,
552 	struct timeval* now, struct timeval* tcp_timeout)
553 {
554 	int newfd;
555 	struct sockaddr_storage addr;
556 	struct tcp_proxy* p;
557 	socklen_t addr_len;
558 	newfd = accept(s, (struct sockaddr*)&addr, &addr_len);
559 	if(newfd == -1) {
560 #ifndef USE_WINSOCK
561 		if(errno == EAGAIN || errno == EINTR)
562 			return;
563 #else
564 		if(WSAGetLastError() == WSAEWOULDBLOCK ||
565 			WSAGetLastError() == WSAEINPROGRESS ||
566 			WSAGetLastError() == WSAECONNRESET)
567 			return;
568 #endif
569 		fatal_exit("accept: %s", sock_strerror(errno));
570 	}
571 	p = (struct tcp_proxy*)calloc(1, sizeof(*p));
572 	if(!p) fatal_exit("out of memory");
573 	memmove(&p->addr, &addr, addr_len);
574 	p->addr_len = addr_len;
575 	log_addr(1, "new tcp proxy", &p->addr, p->addr_len);
576 	p->client_s = newfd;
577 	p->server_s = socket(addr_is_ip6(srv_addr, srv_len)?AF_INET6:AF_INET,
578 		SOCK_STREAM, 0);
579 	if(p->server_s == -1) {
580 		fatal_exit("tcp socket: %s", sock_strerror(errno));
581 	}
582 	fd_set_nonblock(p->client_s);
583 	fd_set_nonblock(p->server_s);
584 	if(connect(p->server_s, (struct sockaddr*)srv_addr, srv_len) == -1) {
585 #ifndef USE_WINSOCK
586 		if(errno != EINPROGRESS) {
587 			log_err("tcp connect: %s", strerror(errno));
588 #else
589 		if(WSAGetLastError() != WSAEWOULDBLOCK &&
590 			WSAGetLastError() != WSAEINPROGRESS) {
591 			log_err("tcp connect: %s",
592 				wsa_strerror(WSAGetLastError()));
593 #endif
594 			sock_close(p->server_s);
595 			sock_close(p->client_s);
596 			free(p);
597 			return;
598 		}
599 	}
600 	p->timeout = *now;
601 	dl_tv_add(&p->timeout, tcp_timeout);
602 
603 	/* listen to client and server */
604 	FD_SET(FD_SET_T p->client_s, rorig);
605 	FD_SET(FD_SET_T p->server_s, rorig);
606 	if(p->client_s+1 > *max)
607 		*max = p->client_s+1;
608 	if(p->server_s+1 > *max)
609 		*max = p->server_s+1;
610 
611 	/* add into proxy list */
612 	p->next = *proxies;
613 	*proxies = p;
614 }
615 
616 /** relay TCP, read a part */
617 static int
618 tcp_relay_read(int s, struct tcp_send_list** first,
619 	struct tcp_send_list** last, struct timeval* now,
620 	struct timeval* delay, sldns_buffer* pkt)
621 {
622 	struct tcp_send_list* item;
623 	ssize_t r = recv(s, (void*)sldns_buffer_begin(pkt),
624 		sldns_buffer_capacity(pkt), 0);
625 	if(r == -1) {
626 #ifndef USE_WINSOCK
627 		if(errno == EINTR || errno == EAGAIN)
628 			return 1;
629 #else
630 		if(WSAGetLastError() == WSAEINPROGRESS ||
631 			WSAGetLastError() == WSAEWOULDBLOCK)
632 			return 1;
633 #endif
634 		log_err("tcp read: %s", sock_strerror(errno));
635 		return 0;
636 	} else if(r == 0) {
637 		/* connection closed */
638 		return 0;
639 	}
640 	item = (struct tcp_send_list*)malloc(sizeof(*item));
641 	if(!item) {
642 		log_err("out of memory");
643 		return 0;
644 	}
645 	verbose(1, "read item len %d", (int)r);
646 	item->len = (size_t)r;
647 	item->item = memdup(sldns_buffer_begin(pkt), item->len);
648 	if(!item->item) {
649 		free(item);
650 		log_err("out of memory");
651 		return 0;
652 	}
653 	item->done = 0;
654 	item->wait = *now;
655 	dl_tv_add(&item->wait, delay);
656 	item->next = NULL;
657 
658 	/* link in */
659 	if(*first) {
660 		(*last)->next = item;
661 	} else {
662 		*first = item;
663 	}
664 	*last = item;
665 	return 1;
666 }
667 
668 /** relay TCP, write a part */
669 static int
670 tcp_relay_write(int s, struct tcp_send_list** first,
671 	struct tcp_send_list** last, struct timeval* now)
672 {
673 	ssize_t r;
674 	struct tcp_send_list* p;
675 	while(*first) {
676 		p = *first;
677 		/* is the item ready? */
678 		if(!dl_tv_smaller(&p->wait, now))
679 			return 1;
680 		/* write it */
681 		r = send(s, (void*)(p->item + p->done), p->len - p->done, 0);
682 		if(r == -1) {
683 #ifndef USE_WINSOCK
684 			if(errno == EAGAIN || errno == EINTR)
685 				return 1;
686 #else
687 			if(WSAGetLastError() == WSAEWOULDBLOCK ||
688 				WSAGetLastError() == WSAEINPROGRESS)
689 				return 1;
690 #endif
691 			log_err("tcp write: %s", sock_strerror(errno));
692 			return 0;
693 		} else if(r == 0) {
694 			/* closed */
695 			return 0;
696 		}
697 		/* account it */
698 		p->done += (size_t)r;
699 		verbose(1, "write item %d of %d", (int)p->done, (int)p->len);
700 		if(p->done >= p->len) {
701 			free(p->item);
702 			*first = p->next;
703 			if(!*first)
704 				*last = NULL;
705 			free(p);
706 		} else {
707 			/* partial write */
708 			return 1;
709 		}
710 	}
711 	return 1;
712 }
713 
714 /** perform TCP relaying */
715 static void
716 service_tcp_relay(struct tcp_proxy** tcp_proxies, struct timeval* now,
717 	struct timeval* delay, struct timeval* tcp_timeout, sldns_buffer* pkt,
718 	fd_set* rset, fd_set* rorig, fd_set* worig)
719 {
720 	struct tcp_proxy* p, **prev;
721 	struct timeval tout;
722 	int delete_it;
723 	p = *tcp_proxies;
724 	prev = tcp_proxies;
725 	tout = *now;
726 	dl_tv_add(&tout, tcp_timeout);
727 
728 	while(p) {
729 		delete_it = 0;
730 		/* can we receive further queries? */
731 		if(!delete_it && FD_ISSET(p->client_s, rset)) {
732 			p->timeout = tout;
733 			log_addr(1, "read tcp query", &p->addr, p->addr_len);
734 			if(!tcp_relay_read(p->client_s, &p->querylist,
735 				&p->querylast, now, delay, pkt))
736 				delete_it = 1;
737 		}
738 		/* can we receive further answers? */
739 		if(!delete_it && p->server_s != -1 &&
740 			FD_ISSET(p->server_s, rset)) {
741 			p->timeout = tout;
742 			log_addr(1, "read tcp answer", &p->addr, p->addr_len);
743 			if(!tcp_relay_read(p->server_s, &p->answerlist,
744 				&p->answerlast, now, delay, pkt)) {
745 				sock_close(p->server_s);
746 				FD_CLR(FD_SET_T p->server_s, worig);
747 				FD_CLR(FD_SET_T p->server_s, rorig);
748 				p->server_s = -1;
749 			}
750 		}
751 		/* can we send on further queries */
752 		if(!delete_it && p->querylist && p->server_s != -1) {
753 			p->timeout = tout;
754 			if(dl_tv_smaller(&p->querylist->wait, now))
755 				log_addr(1, "write tcp query",
756 					&p->addr, p->addr_len);
757 			if(!tcp_relay_write(p->server_s, &p->querylist,
758 				&p->querylast, now))
759 				delete_it = 1;
760 			if(p->querylist &&
761 				dl_tv_smaller(&p->querylist->wait, now))
762 				FD_SET(FD_SET_T p->server_s, worig);
763 			else 	FD_CLR(FD_SET_T p->server_s, worig);
764 		}
765 
766 		/* can we send on further answers */
767 		if(!delete_it && p->answerlist) {
768 			p->timeout = tout;
769 			if(dl_tv_smaller(&p->answerlist->wait, now))
770 				log_addr(1, "write tcp answer",
771 					&p->addr, p->addr_len);
772 			if(!tcp_relay_write(p->client_s, &p->answerlist,
773 				&p->answerlast, now))
774 				delete_it = 1;
775 			if(p->answerlist && dl_tv_smaller(&p->answerlist->wait,
776 				now))
777 				FD_SET(FD_SET_T p->client_s, worig);
778 			else 	FD_CLR(FD_SET_T p->client_s, worig);
779 			if(!p->answerlist && p->server_s == -1)
780 				delete_it = 1;
781 		}
782 
783 		/* does this entry timeout? (unused too long) */
784 		if(dl_tv_smaller(&p->timeout, now)) {
785 			delete_it = 1;
786 		}
787 		if(delete_it) {
788 			struct tcp_proxy* np = p->next;
789 			*prev = np;
790 			FD_CLR(FD_SET_T p->client_s, rorig);
791 			FD_CLR(FD_SET_T p->client_s, worig);
792 			if(p->server_s != -1) {
793 				FD_CLR(FD_SET_T p->server_s, rorig);
794 				FD_CLR(FD_SET_T p->server_s, worig);
795 			}
796 			tcp_proxy_delete(p);
797 			p = np;
798 			continue;
799 		}
800 
801 		prev = &p->next;
802 		p = p->next;
803 	}
804 }
805 
806 /** find waiting time */
807 static int
808 service_findwait(struct timeval* now, struct timeval* wait,
809 	struct ringbuf* ring, struct tcp_proxy* tcplist)
810 {
811 	/* first item is the time to wait */
812 	struct timeval* peek = ring_peek_time(ring);
813 	struct timeval tcv;
814 	int have_tcpval = 0;
815 	struct tcp_proxy* p;
816 
817 	/* also for TCP list the first in sendlists is the time to wait */
818 	for(p=tcplist; p; p=p->next) {
819 		if(!have_tcpval)
820 			tcv = p->timeout;
821 		have_tcpval = 1;
822 		if(dl_tv_smaller(&p->timeout, &tcv))
823 			tcv = p->timeout;
824 		if(p->querylist && dl_tv_smaller(&p->querylist->wait, &tcv))
825 			tcv = p->querylist->wait;
826 		if(p->answerlist && dl_tv_smaller(&p->answerlist->wait, &tcv))
827 			tcv = p->answerlist->wait;
828 	}
829 	if(peek) {
830 		/* peek can be unaligned */
831 		/* use wait as a temp variable */
832 		memmove(wait, peek, sizeof(*wait));
833 		if(!have_tcpval)
834 			tcv = *wait;
835 		else if(dl_tv_smaller(wait, &tcv))
836 			tcv = *wait;
837 		have_tcpval = 1;
838 	}
839 	if(have_tcpval) {
840 		*wait = tcv;
841 		dl_tv_subtract(wait, now);
842 		return 1;
843 	}
844 	/* nothing, block */
845 	return 0;
846 }
847 
848 /** clear proxy list */
849 static void
850 proxy_list_clear(struct proxy* p)
851 {
852 	char from[109];
853 	struct proxy* np;
854 	int i=0, port;
855 	while(p) {
856 		np = p->next;
857 		port = (int)ntohs(((struct sockaddr_in*)&p->addr)->sin_port);
858 		if(addr_is_ip6(&p->addr, p->addr_len)) {
859 			if(inet_ntop(AF_INET6,
860 				&((struct sockaddr_in6*)&p->addr)->sin6_addr,
861 				from, (socklen_t)sizeof(from)) == 0)
862 				(void)strlcpy(from, "err", sizeof(from));
863 		} else {
864 			if(inet_ntop(AF_INET,
865 				&((struct sockaddr_in*)&p->addr)->sin_addr,
866 				from, (socklen_t)sizeof(from)) == 0)
867 				(void)strlcpy(from, "err", sizeof(from));
868 		}
869 		printf("client[%d]: last %s@%d of %d : %u in, %u out, "
870 			"%u returned\n", i++, from, port, (int)p->numreuse+1,
871 			(unsigned)p->numwait, (unsigned)p->numsent,
872 			(unsigned)p->numreturn);
873 		sock_close(p->s);
874 		free(p);
875 		p = np;
876 	}
877 }
878 
879 /** clear TCP proxy list */
880 static void
881 tcp_proxy_list_clear(struct tcp_proxy* p)
882 {
883 	struct tcp_proxy* np;
884 	while(p) {
885 		np = p->next;
886 		tcp_proxy_delete(p);
887 		p = np;
888 	}
889 }
890 
891 /** delayer service loop */
892 static void
893 service_loop(int udp_s, int listen_s, struct ringbuf* ring,
894 	struct timeval* delay, struct timeval* reuse,
895 	struct sockaddr_storage* srv_addr, socklen_t srv_len,
896 	sldns_buffer* pkt)
897 {
898 	fd_set rset, rorig;
899 	fd_set wset, worig;
900 	struct timeval now, wait;
901 	int max, have_wait = 0;
902 	struct proxy* proxies = NULL;
903 	struct tcp_proxy* tcp_proxies = NULL;
904 	struct timeval tcp_timeout;
905 	tcp_timeout.tv_sec = 120;
906 	tcp_timeout.tv_usec = 0;
907 #ifndef S_SPLINT_S
908 	FD_ZERO(&rorig);
909 	FD_ZERO(&worig);
910 	FD_SET(FD_SET_T udp_s, &rorig);
911 	FD_SET(FD_SET_T listen_s, &rorig);
912 #endif
913 	max = udp_s + 1;
914 	if(listen_s + 1 > max) max = listen_s + 1;
915 	while(!do_quit) {
916 		/* wait for events */
917 		rset = rorig;
918 		wset = worig;
919 		if(have_wait)
920 			verbose(1, "wait for %d.%6.6d",
921 			(unsigned)wait.tv_sec, (unsigned)wait.tv_usec);
922 		else	verbose(1, "wait");
923 		if(select(max, &rset, &wset, NULL, have_wait?&wait:NULL) < 0) {
924 			if(errno == EAGAIN || errno == EINTR)
925 				continue;
926 			fatal_exit("select: %s", strerror(errno));
927 		}
928 		/* get current time */
929 		if(gettimeofday(&now, NULL) < 0) {
930 			if(errno == EAGAIN || errno == EINTR)
931 				continue;
932 			fatal_exit("gettimeofday: %s", strerror(errno));
933 		}
934 		verbose(1, "process at %u.%6.6u\n",
935 			(unsigned)now.tv_sec, (unsigned)now.tv_usec);
936 		/* sendout delayed queries to master server (frees up buffer)*/
937 		service_send(ring, &now, pkt, srv_addr, srv_len);
938 		/* proxy return replies */
939 		service_proxy(&rset, udp_s, proxies, pkt, &now);
940 		/* see what can be received to start waiting */
941 		service_recv(udp_s, ring, pkt, &rorig, &max, &proxies,
942 			srv_addr, srv_len, &now, delay, reuse);
943 		/* see if there are new tcp connections */
944 		service_tcp_listen(listen_s, &rorig, &max, &tcp_proxies,
945 			srv_addr, srv_len, &now, &tcp_timeout);
946 		/* service tcp connections */
947 		service_tcp_relay(&tcp_proxies, &now, delay, &tcp_timeout,
948 			pkt, &rset, &rorig, &worig);
949 		/* see what next timeout is (if any) */
950 		have_wait = service_findwait(&now, &wait, ring, tcp_proxies);
951 	}
952 	proxy_list_clear(proxies);
953 	tcp_proxy_list_clear(tcp_proxies);
954 }
955 
956 /** delayer main service routine */
957 static void
958 service(const char* bind_str, int bindport, const char* serv_str,
959 	size_t memsize, int delay_msec)
960 {
961 	struct sockaddr_storage bind_addr, srv_addr;
962 	socklen_t bind_len, srv_len;
963 	struct ringbuf* ring = ring_create(memsize);
964 	struct timeval delay, reuse;
965 	sldns_buffer* pkt;
966 	int i, s, listen_s;
967 #ifndef S_SPLINT_S
968 	delay.tv_sec = delay_msec / 1000;
969 	delay.tv_usec = (delay_msec % 1000)*1000;
970 #endif
971 	reuse = delay; /* reuse is max(4*delay, 1 second) */
972 	dl_tv_add(&reuse, &delay);
973 	dl_tv_add(&reuse, &delay);
974 	dl_tv_add(&reuse, &delay);
975 	if(reuse.tv_sec == 0)
976 		reuse.tv_sec = 1;
977 	if(!extstrtoaddr(serv_str, &srv_addr, &srv_len, UNBOUND_DNS_PORT)) {
978 		printf("cannot parse forward address: %s\n", serv_str);
979 		exit(1);
980 	}
981 	pkt = sldns_buffer_new(65535);
982 	if(!pkt)
983 		fatal_exit("out of memory");
984 	if( signal(SIGINT, delayer_sigh) == SIG_ERR ||
985 #ifdef SIGHUP
986 		signal(SIGHUP, delayer_sigh) == SIG_ERR ||
987 #endif
988 #ifdef SIGQUIT
989 		signal(SIGQUIT, delayer_sigh) == SIG_ERR ||
990 #endif
991 #ifdef SIGBREAK
992 		signal(SIGBREAK, delayer_sigh) == SIG_ERR ||
993 #endif
994 #ifdef SIGALRM
995 		signal(SIGALRM, delayer_sigh) == SIG_ERR ||
996 #endif
997 		signal(SIGTERM, delayer_sigh) == SIG_ERR)
998 		fatal_exit("could not bind to signal");
999 	/* bind UDP port */
1000 	if((s = socket(str_is_ip6(bind_str)?AF_INET6:AF_INET,
1001 		SOCK_DGRAM, 0)) == -1) {
1002 		fatal_exit("socket: %s", sock_strerror(errno));
1003 	}
1004 	i=0;
1005 	if(bindport == 0) {
1006 		bindport = 1024 + ((int)arc4random())%64000;
1007 		i = 100;
1008 	}
1009 	while(1) {
1010 		if(!ipstrtoaddr(bind_str, bindport, &bind_addr, &bind_len)) {
1011 			printf("cannot parse listen address: %s\n", bind_str);
1012 			exit(1);
1013 		}
1014 		if(bind(s, (struct sockaddr*)&bind_addr, bind_len) == -1) {
1015 			log_err("bind: %s", sock_strerror(errno));
1016 			if(i--==0)
1017 				fatal_exit("cannot bind any port");
1018 			bindport = 1024 + ((int)arc4random())%64000;
1019 		} else break;
1020 	}
1021 	fd_set_nonblock(s);
1022 	/* and TCP port */
1023 	if((listen_s = socket(str_is_ip6(bind_str)?AF_INET6:AF_INET,
1024 		SOCK_STREAM, 0)) == -1) {
1025 		fatal_exit("tcp socket: %s", sock_strerror(errno));
1026 	}
1027 #ifdef SO_REUSEADDR
1028 	if(1) {
1029 		int on = 1;
1030 		if(setsockopt(listen_s, SOL_SOCKET, SO_REUSEADDR, (void*)&on,
1031 			(socklen_t)sizeof(on)) < 0)
1032 			fatal_exit("setsockopt(.. SO_REUSEADDR ..) failed: %s",
1033 				sock_strerror(errno));
1034 	}
1035 #endif
1036 	if(bind(listen_s, (struct sockaddr*)&bind_addr, bind_len) == -1) {
1037 		fatal_exit("tcp bind: %s", sock_strerror(errno));
1038 	}
1039 	if(listen(listen_s, 5) == -1) {
1040 		fatal_exit("tcp listen: %s", sock_strerror(errno));
1041 	}
1042 	fd_set_nonblock(listen_s);
1043 	printf("listening on port: %d\n", bindport);
1044 
1045 	/* process loop */
1046 	do_quit = 0;
1047 	service_loop(s, listen_s, ring, &delay, &reuse, &srv_addr, srv_len,
1048 		pkt);
1049 
1050 	/* cleanup */
1051 	verbose(1, "cleanup");
1052 	sock_close(s);
1053 	sock_close(listen_s);
1054 	sldns_buffer_free(pkt);
1055 	ring_delete(ring);
1056 }
1057 
1058 /** getopt global, in case header files fail to declare it. */
1059 extern int optind;
1060 /** getopt global, in case header files fail to declare it. */
1061 extern char* optarg;
1062 
1063 /** main program for delayer */
1064 int main(int argc, char** argv)
1065 {
1066 	int c;		/* defaults */
1067 	const char* server = "127.0.0.1@53";
1068 	const char* bindto = "0.0.0.0";
1069 	int bindport = 0;
1070 	size_t memsize = 10*1024*1024;
1071 	int delay = 100;
1072 
1073 	verbosity = 0;
1074 	log_init(0, 0, 0);
1075 	log_ident_set("delayer");
1076 	if(argc == 1) usage(argv);
1077 	while( (c=getopt(argc, argv, "b:d:f:hm:p:")) != -1) {
1078 		switch(c) {
1079 			case 'b':
1080 				bindto = optarg;
1081 				break;
1082 			case 'd':
1083 				if(atoi(optarg)==0 && strcmp(optarg,"0")!=0) {
1084 					printf("bad delay: %s\n", optarg);
1085 					return 1;
1086 				}
1087 				delay = atoi(optarg);
1088 				break;
1089 			case 'f':
1090 				server = optarg;
1091 				break;
1092 			case 'm':
1093 				if(!cfg_parse_memsize(optarg, &memsize)) {
1094 					printf("bad memsize: %s\n", optarg);
1095 					return 1;
1096 				}
1097 				break;
1098 			case 'p':
1099 				if(atoi(optarg)==0 && strcmp(optarg,"0")!=0) {
1100 					printf("bad port nr: %s\n", optarg);
1101 					return 1;
1102 				}
1103 				bindport = atoi(optarg);
1104 				break;
1105 			case 'h':
1106 			case '?':
1107 			default:
1108 				usage(argv);
1109 		}
1110 	}
1111 	argc -= optind;
1112 	argv += optind;
1113 	if(argc != 0)
1114 		usage(argv);
1115 
1116 	printf("bind to %s @ %d and forward to %s after %d msec\n",
1117 		bindto, bindport, server, delay);
1118 	service(bindto, bindport, server, memsize, delay);
1119 	return 0;
1120 }
1121