]>
Commit | Line | Data |
---|---|---|
12471842 PL |
1 | /* |
2 | * This file is part of PowerDNS or dnsdist. | |
3 | * Copyright -- PowerDNS.COM B.V. and its contributors | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or modify | |
6 | * it under the terms of version 2 of the GNU General Public License as | |
7 | * published by the Free Software Foundation. | |
8 | * | |
9 | * In addition, for the avoidance of any doubt, permission is granted to | |
10 | * link this program with OpenSSL and to (re)distribute the binaries | |
11 | * produced as the result of such linking. | |
12 | * | |
13 | * This program is distributed in the hope that it will be useful, | |
14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | * GNU General Public License for more details. | |
17 | * | |
18 | * You should have received a copy of the GNU General Public License | |
19 | * along with this program; if not, write to the Free Software | |
20 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. | |
21 | */ | |
870a0fe4 AT |
22 | #ifdef HAVE_CONFIG_H |
23 | #include "config.h" | |
24 | #endif | |
002c970a | 25 | #include "iputils.hh" |
22cf1fda | 26 | #include <sys/socket.h> |
27 | ||
002c970a | 28 | /** these functions provide a very lightweight wrapper to the Berkeley sockets API. Errors -> exceptions! */ |
29 | ||
30 | static void RuntimeError(const boost::format& fmt) | |
31 | { | |
32 | throw runtime_error(fmt.str()); | |
33 | } | |
34 | ||
73ba5999 CHB |
35 | static void NetworkErr(const boost::format& fmt) |
36 | { | |
37 | throw NetworkError(fmt.str()); | |
38 | } | |
002c970a | 39 | |
40 | int SSocket(int family, int type, int flags) | |
41 | { | |
42 | int ret = socket(family, type, flags); | |
43 | if(ret < 0) | |
44 | RuntimeError(boost::format("creating socket of type %d: %s") % family % strerror(errno)); | |
45 | return ret; | |
46 | } | |
47 | ||
48 | int SConnect(int sockfd, const ComboAddress& remote) | |
49 | { | |
50 | int ret = connect(sockfd, (struct sockaddr*)&remote, remote.getSocklen()); | |
ccb4b5e2 PD |
51 | if(ret < 0) { |
52 | int savederrno = errno; | |
53 | RuntimeError(boost::format("connecting socket to %s: %s") % remote.toStringWithPort() % strerror(savederrno)); | |
54 | } | |
002c970a | 55 | return ret; |
56 | } | |
57 | ||
51959320 RG |
58 | int SConnectWithTimeout(int sockfd, const ComboAddress& remote, int timeout) |
59 | { | |
60 | int ret = connect(sockfd, (struct sockaddr*)&remote, remote.getSocklen()); | |
61 | if(ret < 0) { | |
62 | int savederrno = errno; | |
63 | if (savederrno == EINPROGRESS) { | |
399fd947 RG |
64 | if (timeout <= 0) { |
65 | return ret; | |
66 | } | |
67 | ||
51959320 RG |
68 | /* we wait until the connection has been established */ |
69 | bool error = false; | |
70 | bool disconnected = false; | |
71 | int res = waitForRWData(sockfd, false, timeout, 0, &error, &disconnected); | |
72 | if (res == 1) { | |
73 | if (error) { | |
74 | savederrno = 0; | |
75 | socklen_t errlen = sizeof(savederrno); | |
76 | if (getsockopt(sockfd, SOL_SOCKET, SO_ERROR, (void *)&savederrno, &errlen) == 0) { | |
73ba5999 | 77 | NetworkErr(boost::format("connecting to %s failed: %s") % remote.toStringWithPort() % string(strerror(savederrno))); |
51959320 RG |
78 | } |
79 | else { | |
73ba5999 | 80 | NetworkErr(boost::format("connecting to %s failed") % remote.toStringWithPort()); |
51959320 RG |
81 | } |
82 | } | |
83 | if (disconnected) { | |
73ba5999 | 84 | NetworkErr(boost::format("%s closed the connection") % remote.toStringWithPort()); |
51959320 RG |
85 | } |
86 | return 0; | |
87 | } | |
88 | else if (res == 0) { | |
73ba5999 | 89 | NetworkErr(boost::format("timeout while connecting to %s") % remote.toStringWithPort()); |
51959320 RG |
90 | } else if (res < 0) { |
91 | savederrno = errno; | |
73ba5999 | 92 | NetworkErr(boost::format("waiting to connect to %s: %s") % remote.toStringWithPort() % string(strerror(savederrno))); |
51959320 RG |
93 | } |
94 | } | |
95 | else { | |
73ba5999 | 96 | NetworkErr(boost::format("connecting to %s: %s") % remote.toStringWithPort() % string(strerror(savederrno))); |
51959320 RG |
97 | } |
98 | } | |
99 | ||
100 | return ret; | |
101 | } | |
102 | ||
002c970a | 103 | int SBind(int sockfd, const ComboAddress& local) |
104 | { | |
105 | int ret = bind(sockfd, (struct sockaddr*)&local, local.getSocklen()); | |
ccb4b5e2 PD |
106 | if(ret < 0) { |
107 | int savederrno = errno; | |
108 | RuntimeError(boost::format("binding socket to %s: %s") % local.toStringWithPort() % strerror(savederrno)); | |
109 | } | |
002c970a | 110 | return ret; |
111 | } | |
112 | ||
113 | int SAccept(int sockfd, ComboAddress& remote) | |
114 | { | |
115 | socklen_t remlen = remote.getSocklen(); | |
116 | ||
117 | int ret = accept(sockfd, (struct sockaddr*)&remote, &remlen); | |
118 | if(ret < 0) | |
119 | RuntimeError(boost::format("accepting new connection on socket: %s") % strerror(errno)); | |
120 | return ret; | |
121 | } | |
122 | ||
123 | int SListen(int sockfd, int limit) | |
124 | { | |
125 | int ret = listen(sockfd, limit); | |
126 | if(ret < 0) | |
127 | RuntimeError(boost::format("setting socket to listen: %s") % strerror(errno)); | |
128 | return ret; | |
129 | } | |
130 | ||
131 | int SSetsockopt(int sockfd, int level, int opname, int value) | |
132 | { | |
133 | int ret = setsockopt(sockfd, level, opname, &value, sizeof(value)); | |
134 | if(ret < 0) | |
135 | RuntimeError(boost::format("setsockopt for level %d and opname %d to %d failed: %s") % level % opname % value % strerror(errno)); | |
136 | return ret; | |
137 | } | |
138 | ||
139 | ||
3e3f0358 | 140 | bool HarvestTimestamp(struct msghdr* msgh, struct timeval* tv) |
141 | { | |
142 | #ifdef SO_TIMESTAMP | |
143 | struct cmsghdr *cmsg; | |
144 | for (cmsg = CMSG_FIRSTHDR(msgh); cmsg != NULL; cmsg = CMSG_NXTHDR(msgh,cmsg)) { | |
22cf1fda | 145 | if ((cmsg->cmsg_level == SOL_SOCKET) && (cmsg->cmsg_type == SO_TIMESTAMP || cmsg->cmsg_type == SCM_TIMESTAMP) && |
3e3f0358 | 146 | CMSG_LEN(sizeof(*tv)) == cmsg->cmsg_len) { |
147 | memcpy(tv, CMSG_DATA(cmsg), sizeof(*tv)); | |
148 | return true; | |
149 | } | |
150 | } | |
151 | #endif | |
152 | return false; | |
153 | } | |
2b3eefc3 | 154 | bool HarvestDestinationAddress(const struct msghdr* msgh, ComboAddress* destination) |
3e3f0358 | 155 | { |
156 | memset(destination, 0, sizeof(*destination)); | |
4d39d7f3 TIH |
157 | #ifdef __NetBSD__ |
158 | struct cmsghdr* cmsg; | |
159 | #else | |
2b3eefc3 | 160 | const struct cmsghdr* cmsg; |
4d39d7f3 | 161 | #endif |
2b3eefc3 | 162 | for (cmsg = CMSG_FIRSTHDR(msgh); cmsg != NULL; cmsg = CMSG_NXTHDR(const_cast<struct msghdr*>(msgh), const_cast<struct cmsghdr*>(cmsg))) { |
3e3f0358 | 163 | #if defined(IP_PKTINFO) |
164 | if ((cmsg->cmsg_level == IPPROTO_IP) && (cmsg->cmsg_type == IP_PKTINFO)) { | |
165 | struct in_pktinfo *i = (struct in_pktinfo *) CMSG_DATA(cmsg); | |
166 | destination->sin4.sin_addr = i->ipi_addr; | |
167 | destination->sin4.sin_family = AF_INET; | |
168 | return true; | |
169 | } | |
170 | #elif defined(IP_RECVDSTADDR) | |
171 | if ((cmsg->cmsg_level == IPPROTO_IP) && (cmsg->cmsg_type == IP_RECVDSTADDR)) { | |
172 | struct in_addr *i = (struct in_addr *) CMSG_DATA(cmsg); | |
173 | destination->sin4.sin_addr = *i; | |
174 | destination->sin4.sin_family = AF_INET; | |
175 | return true; | |
176 | } | |
177 | #endif | |
178 | ||
179 | if ((cmsg->cmsg_level == IPPROTO_IPV6) && (cmsg->cmsg_type == IPV6_PKTINFO)) { | |
180 | struct in6_pktinfo *i = (struct in6_pktinfo *) CMSG_DATA(cmsg); | |
181 | destination->sin6.sin6_addr = i->ipi6_addr; | |
182 | destination->sin4.sin_family = AF_INET6; | |
183 | return true; | |
184 | } | |
185 | } | |
186 | return false; | |
187 | } | |
188 | ||
189 | bool IsAnyAddress(const ComboAddress& addr) | |
190 | { | |
191 | if(addr.sin4.sin_family == AF_INET) | |
192 | return addr.sin4.sin_addr.s_addr == 0; | |
193 | else if(addr.sin4.sin_family == AF_INET6) | |
194 | return !memcmp(&addr.sin6.sin6_addr, &in6addr_any, sizeof(addr.sin6.sin6_addr)); | |
195 | ||
196 | return false; | |
197 | } | |
198 | ||
a683e8bd | 199 | ssize_t sendfromto(int sock, const char* data, size_t len, int flags, const ComboAddress& from, const ComboAddress& to) |
3e3f0358 | 200 | { |
201 | struct msghdr msgh; | |
202 | struct iovec iov; | |
203 | char cbuf[256]; | |
204 | ||
205 | /* Set up iov and msgh structures. */ | |
206 | memset(&msgh, 0, sizeof(struct msghdr)); | |
207 | iov.iov_base = (void*)data; | |
208 | iov.iov_len = len; | |
209 | msgh.msg_iov = &iov; | |
210 | msgh.msg_iovlen = 1; | |
211 | msgh.msg_name = (struct sockaddr*)&to; | |
212 | msgh.msg_namelen = to.getSocklen(); | |
213 | ||
214 | if(from.sin4.sin_family) { | |
fbe2a2e0 | 215 | addCMsgSrcAddr(&msgh, cbuf, &from, 0); |
3e3f0358 | 216 | } |
579cae19 PD |
217 | else { |
218 | msgh.msg_control=NULL; | |
219 | } | |
3e3f0358 | 220 | return sendmsg(sock, &msgh, flags); |
221 | } | |
b71b60ee | 222 | |
223 | // be careful: when using this for receive purposes, make sure addr->sin4.sin_family is set appropriately so getSocklen works! | |
224 | // be careful: when using this function for *send* purposes, be sure to set cbufsize to 0! | |
579cae19 | 225 | // be careful: if you don't call addCMsgSrcAddr after fillMSGHdr, make sure to set msg_control to NULL |
b71b60ee | 226 | void fillMSGHdr(struct msghdr* msgh, struct iovec* iov, char* cbuf, size_t cbufsize, char* data, size_t datalen, ComboAddress* addr) |
227 | { | |
228 | iov->iov_base = data; | |
229 | iov->iov_len = datalen; | |
230 | ||
231 | memset(msgh, 0, sizeof(struct msghdr)); | |
232 | ||
233 | msgh->msg_control = cbuf; | |
234 | msgh->msg_controllen = cbufsize; | |
235 | msgh->msg_name = addr; | |
236 | msgh->msg_namelen = addr->getSocklen(); | |
237 | msgh->msg_iov = iov; | |
238 | msgh->msg_iovlen = 1; | |
239 | msgh->msg_flags = 0; | |
240 | } | |
22779196 | 241 | |
5b6099b2 | 242 | // warning: various parts of PowerDNS assume 'truncate' will never throw |
243 | void ComboAddress::truncate(unsigned int bits) noexcept | |
22779196 | 244 | { |
245 | uint8_t* start; | |
246 | int len=4; | |
247 | if(sin4.sin_family==AF_INET) { | |
ecd43f08 | 248 | if(bits >= 32) |
22779196 | 249 | return; |
250 | start = (uint8_t*)&sin4.sin_addr.s_addr; | |
251 | len=4; | |
252 | } | |
253 | else { | |
ecd43f08 | 254 | if(bits >= 128) |
22779196 | 255 | return; |
256 | start = (uint8_t*)&sin6.sin6_addr.s6_addr; | |
257 | len=16; | |
258 | } | |
259 | ||
22779196 | 260 | auto tozero= len*8 - bits; // if set to 22, this will clear 1 byte, as it should |
261 | ||
262 | memset(start + len - tozero/8, 0, tozero/8); // blot out the whole bytes on the right | |
263 | ||
264 | auto bitsleft=tozero % 8; // 2 bits left to clear | |
265 | ||
266 | // a b c d, to truncate to 22 bits, we just zeroed 'd' and need to zero 2 bits from c | |
267 | // so and by '11111100', which is ~((1<<2)-1) = ~3 | |
268 | uint8_t* place = start + len - 1 - tozero/8; | |
22779196 | 269 | *place &= (~((1<<bitsleft)-1)); |
270 | } | |
f9f9592e | 271 | |
fbe2a2e0 RG |
272 | ssize_t sendMsgWithTimeout(int fd, const char* buffer, size_t len, int timeout, ComboAddress& dest, const ComboAddress& local, unsigned int localItf) |
273 | { | |
274 | struct msghdr msgh; | |
275 | struct iovec iov; | |
276 | char cbuf[256]; | |
277 | bool firstTry = true; | |
278 | fillMSGHdr(&msgh, &iov, cbuf, sizeof(cbuf), const_cast<char*>(buffer), len, &dest); | |
279 | addCMsgSrcAddr(&msgh, cbuf, &local, localItf); | |
280 | ||
281 | do { | |
282 | ssize_t written = sendmsg(fd, &msgh, 0); | |
283 | ||
284 | if (written > 0) | |
285 | return written; | |
286 | ||
287 | if (errno == EAGAIN) { | |
288 | if (firstTry) { | |
289 | int res = waitForRWData(fd, false, timeout, 0); | |
290 | if (res > 0) { | |
291 | /* there is room available */ | |
292 | firstTry = false; | |
293 | } | |
294 | else if (res == 0) { | |
295 | throw runtime_error("Timeout while waiting to write data"); | |
296 | } else { | |
297 | throw runtime_error("Error while waiting for room to write data"); | |
298 | } | |
299 | } | |
300 | else { | |
301 | throw runtime_error("Timeout while waiting to write data"); | |
302 | } | |
303 | } | |
304 | else { | |
305 | unixDie("failed in write2WithTimeout"); | |
306 | } | |
307 | } | |
308 | while (firstTry); | |
309 | ||
310 | return 0; | |
311 | } | |
312 | ||
f9f9592e | 313 | template class NetmaskTree<bool>; |
fbe2a2e0 | 314 | |
17bca36a RG |
315 | bool sendSizeAndMsgWithTimeout(int sock, uint16_t bufferLen, const char* buffer, int idleTimeout, const ComboAddress* dest, const ComboAddress* local, unsigned int localItf, int totalTimeout, int flags) |
316 | { | |
317 | uint16_t size = htons(bufferLen); | |
318 | char cbuf[256]; | |
319 | struct msghdr msgh; | |
320 | struct iovec iov[2]; | |
321 | int remainingTime = totalTimeout; | |
322 | time_t start = 0; | |
323 | if (totalTimeout) { | |
324 | start = time(NULL); | |
325 | } | |
326 | ||
327 | /* Set up iov and msgh structures. */ | |
328 | memset(&msgh, 0, sizeof(struct msghdr)); | |
329 | msgh.msg_control = nullptr; | |
330 | msgh.msg_controllen = 0; | |
331 | if (dest) { | |
332 | msgh.msg_name = reinterpret_cast<void*>(const_cast<ComboAddress*>(dest)); | |
333 | msgh.msg_namelen = dest->getSocklen(); | |
334 | } | |
335 | else { | |
336 | msgh.msg_name = nullptr; | |
337 | msgh.msg_namelen = 0; | |
338 | } | |
339 | ||
340 | msgh.msg_flags = 0; | |
341 | ||
342 | if (localItf != 0 && local) { | |
343 | addCMsgSrcAddr(&msgh, cbuf, local, localItf); | |
344 | } | |
345 | ||
346 | iov[0].iov_base = &size; | |
347 | iov[0].iov_len = sizeof(size); | |
348 | iov[1].iov_base = reinterpret_cast<void*>(const_cast<char*>(buffer)); | |
349 | iov[1].iov_len = bufferLen; | |
350 | ||
351 | size_t pos = 0; | |
352 | size_t sent = 0; | |
353 | size_t nbElements = sizeof(iov)/sizeof(*iov); | |
354 | while (true) { | |
355 | msgh.msg_iov = &iov[pos]; | |
356 | msgh.msg_iovlen = nbElements - pos; | |
357 | ||
358 | ssize_t res = sendmsg(sock, &msgh, flags); | |
359 | if (res > 0) { | |
360 | size_t written = static_cast<size_t>(res); | |
361 | sent += written; | |
362 | ||
363 | if (sent == (sizeof(size) + bufferLen)) { | |
364 | return true; | |
365 | } | |
366 | /* partial write, we need to keep only the (parts of) elements | |
367 | that have not been written. | |
368 | */ | |
369 | do { | |
370 | if (written < iov[pos].iov_len) { | |
371 | iov[pos].iov_len -= written; | |
d9442969 | 372 | iov[pos].iov_base = reinterpret_cast<void*>(reinterpret_cast<char*>(iov[pos].iov_base) + written); |
17bca36a RG |
373 | written = 0; |
374 | } | |
375 | else { | |
376 | written -= iov[pos].iov_len; | |
377 | iov[pos].iov_len = 0; | |
378 | pos++; | |
379 | } | |
380 | } | |
381 | while (written > 0 && pos < nbElements); | |
382 | } | |
383 | else if (res == -1) { | |
384 | if (errno == EINTR) { | |
385 | continue; | |
386 | } | |
387 | else if (errno == EAGAIN || errno == EWOULDBLOCK || errno == EINPROGRESS) { | |
388 | /* EINPROGRESS might happen with non blocking socket, | |
389 | especially with TCP Fast Open */ | |
390 | int ret = waitForRWData(sock, false, (totalTimeout == 0 || idleTimeout <= remainingTime) ? idleTimeout : remainingTime, 0); | |
391 | if (ret > 0) { | |
392 | /* there is room available */ | |
393 | } | |
394 | else if (ret == 0) { | |
395 | throw runtime_error("Timeout while waiting to send data"); | |
396 | } else { | |
397 | throw runtime_error("Error while waiting for room to send data"); | |
398 | } | |
399 | } | |
400 | else { | |
401 | unixDie("failed in sendSizeAndMsgWithTimeout"); | |
402 | } | |
403 | } | |
404 | if (totalTimeout) { | |
405 | time_t now = time(NULL); | |
406 | int elapsed = now - start; | |
407 | if (elapsed >= remainingTime) { | |
408 | throw runtime_error("Timeout while sending data"); | |
409 | } | |
410 | start = now; | |
411 | remainingTime -= elapsed; | |
412 | } | |
413 | } | |
414 | ||
415 | return false; | |
416 | } | |
840ed663 RG |
417 | |
418 | /* requires a non-blocking socket. | |
419 | On Linux, we could use MSG_DONTWAIT on a blocking socket | |
420 | but this is not portable. | |
421 | */ | |
422 | bool isTCPSocketUsable(int sock) | |
423 | { | |
424 | int err = 0; | |
425 | char buf = '\0'; | |
426 | size_t buf_size = sizeof(buf); | |
427 | ||
428 | do { | |
429 | ssize_t got = recv(sock, &buf, buf_size, MSG_PEEK); | |
430 | ||
431 | if (got > 0) { | |
432 | /* socket is usable, some data is even waiting to be read */ | |
433 | return true; | |
434 | } | |
435 | else if (got == 0) { | |
436 | /* other end has closed the socket */ | |
437 | return false; | |
438 | } | |
439 | else { | |
88479ac5 | 440 | err = errno; |
840ed663 RG |
441 | |
442 | if (err == EAGAIN || err == EWOULDBLOCK) { | |
443 | /* socket is usable, no data waiting */ | |
444 | return true; | |
445 | } | |
446 | else { | |
447 | if (err != EINTR) { | |
448 | /* something is wrong, could be ECONNRESET, | |
449 | ENOTCONN, EPIPE, but anyway this socket is | |
450 | not usable. */ | |
451 | return false; | |
452 | } | |
453 | } | |
454 | } | |
455 | } while (err == EINTR); | |
456 | ||
457 | return false; | |
458 | } |