]>
Commit | Line | Data |
---|---|---|
12c86877 BH |
1 | /* |
2 | PowerDNS Versatile Database Driven Nameserver | |
2e7834cb | 3 | Copyright (C) 2002-2012 PowerDNS.COM BV |
12c86877 BH |
4 | |
5 | This program is free software; you can redistribute it and/or modify | |
22dc646a BH |
6 | it under the terms of the GNU General Public License version 2 |
7 | as published by the Free Software Foundation | |
f782fe38 MH |
8 | |
9 | Additionally, the license of this program contains a special | |
10 | exception which allows to distribute the program in binary form when | |
11 | it is linked against OpenSSL. | |
12 | ||
12c86877 BH |
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 | |
06bd9ccf | 20 | Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA |
12c86877 | 21 | */ |
870a0fe4 AT |
22 | #ifdef HAVE_CONFIG_H |
23 | #include "config.h" | |
24 | #endif | |
b6f3b03a | 25 | #include <boost/algorithm/string.hpp> |
bf269e28 | 26 | #include "auth-packetcache.hh" |
1258abe0 | 27 | #include "utility.hh" |
519f5484 | 28 | #include "threadname.hh" |
add640c0 | 29 | #include "dnssecinfra.hh" |
4c1474f3 | 30 | #include "dnsseckeeper.hh" |
12c86877 | 31 | #include <cstdio> |
4888e4b2 | 32 | #include "base32.hh" |
12c86877 BH |
33 | #include <cstring> |
34 | #include <cstdlib> | |
35 | #include <sys/types.h> | |
940d7811 | 36 | #include <netinet/tcp.h> |
12c86877 BH |
37 | #include <iostream> |
38 | #include <string> | |
39 | #include "tcpreceiver.hh" | |
67d74e49 | 40 | #include "sstuff.hh" |
fa8fd4d2 | 41 | |
12c86877 BH |
42 | #include <errno.h> |
43 | #include <signal.h> | |
78bcb858 | 44 | #include "base64.hh" |
12c86877 BH |
45 | #include "ueberbackend.hh" |
46 | #include "dnspacket.hh" | |
47 | #include "nameserver.hh" | |
48 | #include "distributor.hh" | |
49 | #include "lock.hh" | |
50 | #include "logger.hh" | |
51 | #include "arguments.hh" | |
379ab445 | 52 | |
3e8216c8 | 53 | #include "common_startup.hh" |
12c86877 BH |
54 | #include "packethandler.hh" |
55 | #include "statbag.hh" | |
56 | #include "resolver.hh" | |
57 | #include "communicator.hh" | |
61b26744 | 58 | #include "namespaces.hh" |
8e9b7d99 | 59 | #include "signingpipe.hh" |
273d88b2 | 60 | #include "stubresolver.hh" |
bf269e28 | 61 | extern AuthPacketCache PC; |
12c86877 BH |
62 | extern StatBag S; |
63 | ||
64 | /** | |
65 | \file tcpreceiver.cc | |
66 | \brief This file implements the tcpreceiver that receives and answers questions over TCP/IP | |
67 | */ | |
68 | ||
ac2bb9e7 | 69 | pthread_mutex_t TCPNameserver::s_plock = PTHREAD_MUTEX_INITIALIZER; |
12c86877 BH |
70 | Semaphore *TCPNameserver::d_connectionroom_sem; |
71 | PacketHandler *TCPNameserver::s_P; | |
9f1d5826 | 72 | NetmaskGroup TCPNameserver::d_ng; |
cb0af1a1 RG |
73 | size_t TCPNameserver::d_maxTransactionsPerConn; |
74 | size_t TCPNameserver::d_maxConnectionsPerClient; | |
75 | unsigned int TCPNameserver::d_idleTimeout; | |
76 | unsigned int TCPNameserver::d_maxConnectionDuration; | |
77 | std::mutex TCPNameserver::s_clientsCountMutex; | |
78 | std::map<ComboAddress,size_t,ComboAddress::addressOnlyLessThan> TCPNameserver::s_clientsCount; | |
12c86877 | 79 | |
12c86877 BH |
80 | void TCPNameserver::go() |
81 | { | |
e6a9dde5 | 82 | g_log<<Logger::Error<<"Creating backend connection for TCP"<<endl; |
12c86877 BH |
83 | s_P=0; |
84 | try { | |
85 | s_P=new PacketHandler; | |
86 | } | |
3f81d239 | 87 | catch(PDNSException &ae) { |
e6a9dde5 | 88 | g_log<<Logger::Error<<"TCP server is unable to launch backends - will try again when questions come in: "<<ae.reason<<endl; |
12c86877 BH |
89 | } |
90 | pthread_create(&d_tid, 0, launcher, static_cast<void *>(this)); | |
91 | } | |
92 | ||
93 | void *TCPNameserver::launcher(void *data) | |
94 | { | |
95 | static_cast<TCPNameserver *>(data)->thread(); | |
96 | return 0; | |
97 | } | |
98 | ||
3f81d239 | 99 | // throws PDNSException if things didn't go according to plan, returns 0 if really 0 bytes were read |
cb0af1a1 | 100 | static int readnWithTimeout(int fd, void* buffer, unsigned int n, unsigned int idleTimeout, bool throwOnEOF=true, unsigned int totalTimeout=0) |
12c86877 | 101 | { |
6a3e5d1a BH |
102 | unsigned int bytes=n; |
103 | char *ptr = (char*)buffer; | |
104 | int ret; | |
cb0af1a1 RG |
105 | time_t start = 0; |
106 | unsigned int remainingTotal = totalTimeout; | |
107 | if (totalTimeout) { | |
108 | start = time(NULL); | |
109 | } | |
6a3e5d1a BH |
110 | while(bytes) { |
111 | ret=read(fd, ptr, bytes); | |
112 | if(ret < 0) { | |
113 | if(errno==EAGAIN) { | |
cb0af1a1 | 114 | ret=waitForData(fd, (totalTimeout == 0 || idleTimeout <= remainingTotal) ? idleTimeout : remainingTotal); |
4957a608 BH |
115 | if(ret < 0) |
116 | throw NetworkError("Waiting for data read"); | |
117 | if(!ret) | |
118 | throw NetworkError("Timeout reading data"); | |
119 | continue; | |
6a3e5d1a BH |
120 | } |
121 | else | |
4957a608 | 122 | throw NetworkError("Reading data: "+stringerror()); |
6a3e5d1a BH |
123 | } |
124 | if(!ret) { | |
125 | if(!throwOnEOF && n == bytes) | |
4957a608 | 126 | return 0; |
6a3e5d1a | 127 | else |
4957a608 | 128 | throw NetworkError("Did not fulfill read from TCP due to EOF"); |
6a3e5d1a BH |
129 | } |
130 | ||
131 | ptr += ret; | |
132 | bytes -= ret; | |
cb0af1a1 RG |
133 | if (totalTimeout) { |
134 | time_t now = time(NULL); | |
135 | unsigned int elapsed = now - start; | |
136 | if (elapsed >= remainingTotal) { | |
137 | throw NetworkError("Timeout while reading data"); | |
138 | } | |
139 | start = now; | |
140 | remainingTotal -= elapsed; | |
141 | } | |
6a3e5d1a BH |
142 | } |
143 | return n; | |
144 | } | |
12c86877 | 145 | |
6a3e5d1a | 146 | // ditto |
cb0af1a1 | 147 | static void writenWithTimeout(int fd, const void *buffer, unsigned int n, unsigned int idleTimeout) |
6a3e5d1a BH |
148 | { |
149 | unsigned int bytes=n; | |
150 | const char *ptr = (char*)buffer; | |
151 | int ret; | |
152 | while(bytes) { | |
153 | ret=write(fd, ptr, bytes); | |
154 | if(ret < 0) { | |
155 | if(errno==EAGAIN) { | |
cb0af1a1 | 156 | ret=waitForRWData(fd, false, idleTimeout, 0); |
4957a608 BH |
157 | if(ret < 0) |
158 | throw NetworkError("Waiting for data write"); | |
159 | if(!ret) | |
160 | throw NetworkError("Timeout writing data"); | |
161 | continue; | |
6a3e5d1a BH |
162 | } |
163 | else | |
4957a608 | 164 | throw NetworkError("Writing data: "+stringerror()); |
6a3e5d1a | 165 | } |
12c86877 | 166 | if(!ret) { |
67d74e49 | 167 | throw NetworkError("Did not fulfill TCP write due to EOF"); |
12c86877 | 168 | } |
6a3e5d1a BH |
169 | |
170 | ptr += ret; | |
171 | bytes -= ret; | |
12c86877 | 172 | } |
12c86877 BH |
173 | } |
174 | ||
6a3e5d1a | 175 | void connectWithTimeout(int fd, struct sockaddr* remote, size_t socklen) |
12c86877 | 176 | { |
6a3e5d1a BH |
177 | int err; |
178 | Utility::socklen_t len=sizeof(err); | |
179 | ||
76473b92 | 180 | if((err=connect(fd, remote, socklen))<0 && errno!=EINPROGRESS) |
67d74e49 | 181 | throw NetworkError("connect: "+stringerror()); |
6a3e5d1a BH |
182 | |
183 | if(!err) | |
184 | goto done; | |
185 | ||
186 | err=waitForRWData(fd, false, 5, 0); | |
187 | if(err == 0) | |
67d74e49 | 188 | throw NetworkError("Timeout connecting to remote"); |
6a3e5d1a | 189 | if(err < 0) |
67d74e49 | 190 | throw NetworkError("Error connecting to remote"); |
12c86877 | 191 | |
6a3e5d1a | 192 | if(getsockopt(fd, SOL_SOCKET,SO_ERROR,(char *)&err,&len)<0) |
67d74e49 | 193 | throw NetworkError("Error connecting to remote: "+stringerror()); // Solaris |
6a3e5d1a BH |
194 | |
195 | if(err) | |
67d74e49 | 196 | throw NetworkError("Error connecting to remote: "+string(strerror(err))); |
6a3e5d1a BH |
197 | |
198 | done: | |
199 | ; | |
200 | } | |
12c86877 | 201 | |
6a3e5d1a BH |
202 | void TCPNameserver::sendPacket(shared_ptr<DNSPacket> p, int outsock) |
203 | { | |
b552d7b1 | 204 | g_rs.submitResponse(*p, false); |
9951e2d0 | 205 | |
fbaa5e09 BH |
206 | uint16_t len=htons(p->getString().length()); |
207 | string buffer((const char*)&len, 2); | |
208 | buffer.append(p->getString()); | |
cb0af1a1 | 209 | writenWithTimeout(outsock, buffer.c_str(), buffer.length(), d_idleTimeout); |
6a3e5d1a BH |
210 | } |
211 | ||
212 | ||
cb0af1a1 | 213 | void TCPNameserver::getQuestion(int fd, char *mesg, int pktlen, const ComboAddress &remote, unsigned int totalTime) |
6a3e5d1a BH |
214 | try |
215 | { | |
cb0af1a1 | 216 | readnWithTimeout(fd, mesg, pktlen, d_idleTimeout, true, totalTime); |
6a3e5d1a | 217 | } |
67d74e49 BH |
218 | catch(NetworkError& ae) { |
219 | throw NetworkError("Error reading DNS data from TCP client "+remote.toString()+": "+ae.what()); | |
12c86877 BH |
220 | } |
221 | ||
5fd567ec | 222 | static void incTCPAnswerCount(const ComboAddress& remote) |
223 | { | |
224 | S.inc("tcp-answers"); | |
225 | if(remote.sin4.sin_family == AF_INET6) | |
226 | S.inc("tcp6-answers"); | |
227 | else | |
228 | S.inc("tcp4-answers"); | |
229 | } | |
cb0af1a1 RG |
230 | |
231 | static bool maxConnectionDurationReached(unsigned int maxConnectionDuration, time_t start, unsigned int& remainingTime) | |
232 | { | |
233 | if (maxConnectionDuration) { | |
234 | time_t elapsed = time(NULL) - start; | |
235 | if (elapsed >= maxConnectionDuration) { | |
236 | return true; | |
237 | } | |
238 | remainingTime = maxConnectionDuration - elapsed; | |
239 | } | |
240 | return false; | |
241 | } | |
242 | ||
243 | void TCPNameserver::decrementClientCount(const ComboAddress& remote) | |
244 | { | |
245 | if (d_maxConnectionsPerClient) { | |
246 | std::lock_guard<std::mutex> lock(s_clientsCountMutex); | |
247 | s_clientsCount[remote]--; | |
248 | if (s_clientsCount[remote] == 0) { | |
249 | s_clientsCount.erase(remote); | |
250 | } | |
251 | } | |
252 | } | |
253 | ||
12c86877 BH |
254 | void *TCPNameserver::doConnection(void *data) |
255 | { | |
519f5484 | 256 | setThreadName("pdns/tcpConnect"); |
ff76e8b4 | 257 | shared_ptr<DNSPacket> packet; |
b014ad98 BH |
258 | // Fix gcc-4.0 error (on AMD64) |
259 | int fd=(int)(long)data; // gotta love C (generates a harmless warning on opteron) | |
4f5e7925 | 260 | ComboAddress remote; |
261 | socklen_t remotelen=sizeof(remote); | |
cb0af1a1 RG |
262 | size_t transactions = 0; |
263 | time_t start = 0; | |
264 | if (d_maxConnectionDuration) { | |
265 | start = time(NULL); | |
266 | } | |
4f5e7925 | 267 | |
12c86877 | 268 | pthread_detach(pthread_self()); |
4f5e7925 | 269 | if(getpeername(fd, (struct sockaddr *)&remote, &remotelen) < 0) { |
e6a9dde5 | 270 | g_log<<Logger::Warning<<"Received question from socket which had no remote address, dropping ("<<stringerror()<<")"<<endl; |
4f5e7925 | 271 | d_connectionroom_sem->post(); |
a7b68ae7 RG |
272 | try { |
273 | closesocket(fd); | |
274 | } | |
275 | catch(const PDNSException& e) { | |
e6a9dde5 | 276 | g_log<<Logger::Error<<"Error closing TCP socket: "<<e.reason<<endl; |
a7b68ae7 | 277 | } |
4f5e7925 | 278 | return 0; |
279 | } | |
280 | ||
3897b9e1 | 281 | setNonBlocking(fd); |
12c86877 | 282 | try { |
c2b4ccc0 | 283 | int mesgsize=65535; |
284 | scoped_array<char> mesg(new char[mesgsize]); | |
12c86877 | 285 | |
e6a9dde5 | 286 | DLOG(g_log<<"TCP Connection accepted on fd "<<fd<<endl); |
21a303f3 | 287 | bool logDNSQueries= ::arg().mustDo("log-dns-queries"); |
12c86877 | 288 | for(;;) { |
cb0af1a1 RG |
289 | unsigned int remainingTime = 0; |
290 | transactions++; | |
291 | if (d_maxTransactionsPerConn && transactions > d_maxTransactionsPerConn) { | |
e6a9dde5 | 292 | g_log << Logger::Notice<<"TCP Remote "<< remote <<" exceeded the number of transactions per connection, dropping."; |
cb0af1a1 RG |
293 | break; |
294 | } | |
295 | if (maxConnectionDurationReached(d_maxConnectionDuration, start, remainingTime)) { | |
e6a9dde5 | 296 | g_log << Logger::Notice<<"TCP Remote "<< remote <<" exceeded the maximum TCP connection duration, dropping."; |
cb0af1a1 RG |
297 | break; |
298 | } | |
6a3e5d1a BH |
299 | |
300 | uint16_t pktlen; | |
cb0af1a1 | 301 | if(!readnWithTimeout(fd, &pktlen, 2, d_idleTimeout, false, remainingTime)) |
4957a608 | 302 | break; |
6a3e5d1a | 303 | else |
4957a608 | 304 | pktlen=ntohs(pktlen); |
12c86877 | 305 | |
366e1e5e AT |
306 | // this check will always be false *if* no one touches |
307 | // the mesg array. pktlen can be maximum of 65535 as | |
308 | // it is 2 byte unsigned variable. In getQuestion, we | |
309 | // write to 0 up to pktlen-1 so 65535 is just right. | |
310 | ||
311 | // do not remove this check as it will catch if someone | |
312 | // decreases the mesg buffer size for some reason. | |
c2b4ccc0 | 313 | if(pktlen > mesgsize) { |
e6a9dde5 | 314 | g_log<<Logger::Warning<<"Received an overly large question from "<<remote.toString()<<", dropping"<<endl; |
4957a608 | 315 | break; |
12c86877 BH |
316 | } |
317 | ||
cb0af1a1 | 318 | if (maxConnectionDurationReached(d_maxConnectionDuration, start, remainingTime)) { |
e6a9dde5 | 319 | g_log << Logger::Notice<<"TCP Remote "<< remote <<" exceeded the maximum TCP connection duration, dropping."; |
cb0af1a1 RG |
320 | break; |
321 | } | |
322 | ||
323 | getQuestion(fd, mesg.get(), pktlen, remote, remainingTime); | |
12c86877 | 324 | S.inc("tcp-queries"); |
5fd567ec | 325 | if(remote.sin4.sin_family == AF_INET6) |
326 | S.inc("tcp6-queries"); | |
327 | else | |
328 | S.inc("tcp4-queries"); | |
3e579e91 | 329 | |
27c0050c | 330 | packet=shared_ptr<DNSPacket>(new DNSPacket(true)); |
809fe23f | 331 | packet->setRemote(&remote); |
e9dd48f9 | 332 | packet->d_tcp=true; |
ff76e8b4 | 333 | packet->setSocket(fd); |
c2b4ccc0 | 334 | if(packet->parse(mesg.get(), pktlen)<0) |
4957a608 | 335 | break; |
c1663439 | 336 | |
6e59a580 KM |
337 | if(packet->qtype.getCode()==QType::AXFR) { |
338 | if(doAXFR(packet->qdomain, packet, fd)) | |
5fd567ec | 339 | incTCPAnswerCount(remote); |
6e59a580 KM |
340 | continue; |
341 | } | |
342 | ||
343 | if(packet->qtype.getCode()==QType::IXFR) { | |
344 | if(doIXFR(packet, fd)) | |
5fd567ec | 345 | incTCPAnswerCount(remote); |
4957a608 | 346 | continue; |
12c86877 BH |
347 | } |
348 | ||
ff76e8b4 | 349 | shared_ptr<DNSPacket> reply; |
27c0050c | 350 | shared_ptr<DNSPacket> cached= shared_ptr<DNSPacket>(new DNSPacket(false)); |
fe498ace | 351 | if(logDNSQueries) { |
cece60fc CH |
352 | string remote_text; |
353 | if(packet->hasEDNSSubnet()) | |
354 | remote_text = packet->getRemote().toString() + "<-" + packet->getRealRemote().toString(); | |
fe498ace | 355 | else |
cece60fc | 356 | remote_text = packet->getRemote().toString(); |
e6a9dde5 | 357 | g_log << Logger::Notice<<"TCP Remote "<< remote_text <<" wants '" << packet->qdomain<<"|"<<packet->qtype.getName() << |
bb5903e2 | 358 | "', do = " <<packet->d_dnssecOk <<", bufsize = "<< packet->getMaxReplyLen()<<": "; |
fe498ace | 359 | } |
bb5903e2 | 360 | |
9a037bfa KM |
361 | if(PC.enabled()) { |
362 | if(packet->couldBeCached() && PC.get(packet.get(), cached.get())) { // short circuit - does the PacketCache recognize this question? | |
363 | if(logDNSQueries) | |
364 | g_log<<"packetcache HIT"<<endl; | |
365 | cached->setRemote(&packet->d_remote); | |
366 | cached->d.id=packet->d.id; | |
367 | cached->d.rd=packet->d.rd; // copy in recursion desired bit | |
368 | cached->commitD(); // commit d to the packet inlined | |
369 | ||
370 | sendPacket(cached, fd); // presigned, don't do it again | |
371 | continue; | |
372 | } | |
21a303f3 | 373 | if(logDNSQueries) |
9a037bfa | 374 | g_log<<"packetcache MISS"<<endl; |
12c86877 | 375 | } |
12c86877 | 376 | { |
4957a608 BH |
377 | Lock l(&s_plock); |
378 | if(!s_P) { | |
e6a9dde5 | 379 | g_log<<Logger::Error<<"TCP server is without backend connections, launching"<<endl; |
4957a608 BH |
380 | s_P=new PacketHandler; |
381 | } | |
4957a608 | 382 | |
e89efca5 | 383 | reply=shared_ptr<DNSPacket>(s_P->doQuestion(packet.get())); // we really need to ask the backend :-) |
12c86877 BH |
384 | } |
385 | ||
12c86877 | 386 | if(!reply) // unable to write an answer? |
4957a608 | 387 | break; |
b552d7b1 | 388 | |
ff76e8b4 | 389 | sendPacket(reply, fd); |
12c86877 | 390 | } |
12c86877 | 391 | } |
3f81d239 | 392 | catch(PDNSException &ae) { |
556252ea BH |
393 | Lock l(&s_plock); |
394 | delete s_P; | |
395 | s_P = 0; // on next call, backend will be recycled | |
e6a9dde5 | 396 | g_log<<Logger::Error<<"TCP nameserver had error, cycling backend: "<<ae.reason<<endl; |
ef1d2f44 | 397 | } |
0afa9049 | 398 | catch(NetworkError &e) { |
e6a9dde5 | 399 | g_log<<Logger::Info<<"TCP Connection Thread died because of network error: "<<e.what()<<endl; |
0afa9049 BH |
400 | } |
401 | ||
adc10f99 | 402 | catch(std::exception &e) { |
e6a9dde5 | 403 | g_log<<Logger::Error<<"TCP Connection Thread died because of STL error: "<<e.what()<<endl; |
12c86877 BH |
404 | } |
405 | catch( ... ) | |
406 | { | |
e6a9dde5 | 407 | g_log << Logger::Error << "TCP Connection Thread caught unknown exception." << endl; |
12c86877 | 408 | } |
12c86877 | 409 | d_connectionroom_sem->post(); |
a7b68ae7 RG |
410 | |
411 | try { | |
412 | closesocket(fd); | |
413 | } | |
414 | catch(const PDNSException& e) { | |
e6a9dde5 | 415 | g_log<<Logger::Error<<"Error closing TCP socket: "<<e.reason<<endl; |
a7b68ae7 | 416 | } |
cb0af1a1 | 417 | decrementClientCount(remote); |
12c86877 BH |
418 | |
419 | return 0; | |
420 | } | |
421 | ||
78bcb858 | 422 | |
e082fb4c | 423 | // call this method with s_plock held! |
ff76e8b4 | 424 | bool TCPNameserver::canDoAXFR(shared_ptr<DNSPacket> q) |
12c86877 | 425 | { |
379ab445 | 426 | if(::arg().mustDo("disable-axfr")) |
318c3ec6 BH |
427 | return false; |
428 | ||
78bcb858 BH |
429 | if(q->d_havetsig) { // if you have one, it must be good |
430 | TSIGRecordContent trc; | |
7abbc40f PD |
431 | DNSName keyname; |
432 | string secret; | |
ea3816cf | 433 | if(!q->checkForCorrectTSIG(s_P->getBackend(), &keyname, &secret, &trc)) { |
78bcb858 | 434 | return false; |
7f9ac49b AT |
435 | } else { |
436 | getTSIGHashEnum(trc.d_algoName, q->d_tsig_algo); | |
437 | if (q->d_tsig_algo == TSIG_GSS) { | |
1635f12b | 438 | GssContext gssctx(keyname); |
7f9ac49b | 439 | if (!gssctx.getPeerPrincipal(q->d_peer_principal)) { |
e6a9dde5 | 440 | g_log<<Logger::Warning<<"Failed to extract peer principal from GSS context with keyname '"<<keyname<<"'"<<endl; |
7f9ac49b AT |
441 | } |
442 | } | |
443 | } | |
444 | ||
ea99d474 | 445 | DNSSECKeeper dk(s_P->getBackend()); |
5e29f2f9 | 446 | |
84fc3f8b AT |
447 | if (q->d_tsig_algo == TSIG_GSS) { |
448 | vector<string> princs; | |
449 | s_P->getBackend()->getDomainMetadata(q->qdomain, "GSS-ALLOW-AXFR-PRINCIPAL", princs); | |
ff05fd12 | 450 | for(const std::string& princ : princs) { |
84fc3f8b | 451 | if (q->d_peer_principal == princ) { |
e6a9dde5 | 452 | g_log<<Logger::Warning<<"AXFR of domain '"<<q->qdomain<<"' allowed: TSIG signed request with authorized principal '"<<q->d_peer_principal<<"' and algorithm 'gss-tsig'"<<endl; |
84fc3f8b AT |
453 | return true; |
454 | } | |
455 | } | |
e6a9dde5 | 456 | g_log<<Logger::Warning<<"AXFR of domain '"<<q->qdomain<<"' denied: TSIG signed request with principal '"<<q->d_peer_principal<<"' and algorithm 'gss-tsig' is not permitted"<<endl; |
84fc3f8b AT |
457 | return false; |
458 | } | |
459 | ||
3d03fee8 | 460 | if(!dk.TSIGGrantsAccess(q->qdomain, keyname)) { |
e6a9dde5 | 461 | g_log<<Logger::Error<<"AXFR '"<<q->qdomain<<"' denied: key with name '"<<keyname<<"' and algorithm '"<<getTSIGAlgoName(q->d_tsig_algo)<<"' does not grant access to zone"<<endl; |
78bcb858 BH |
462 | return false; |
463 | } | |
464 | else { | |
e6a9dde5 | 465 | g_log<<Logger::Warning<<"AXFR of domain '"<<q->qdomain<<"' allowed: TSIG signed request with authorized key '"<<keyname<<"' and algorithm '"<<getTSIGAlgoName(q->d_tsig_algo)<<"'"<<endl; |
78bcb858 BH |
466 | return true; |
467 | } | |
468 | } | |
93afc0a3 PD |
469 | |
470 | // cerr<<"checking allow-axfr-ips"<<endl; | |
471 | if(!(::arg()["allow-axfr-ips"].empty()) && d_ng.match( (ComboAddress *) &q->d_remote )) { | |
e6a9dde5 | 472 | g_log<<Logger::Warning<<"AXFR of domain '"<<q->qdomain<<"' allowed: client IP "<<q->getRemote()<<" is in allow-axfr-ips"<<endl; |
12c86877 | 473 | return true; |
ab5edd12 | 474 | } |
93afc0a3 PD |
475 | |
476 | FindNS fns; | |
477 | ||
478 | // cerr<<"doing per-zone-axfr-acls"<<endl; | |
479 | SOAData sd; | |
79ba7763 | 480 | if(s_P->getBackend()->getSOAUncached(q->qdomain,sd)) { |
93afc0a3 PD |
481 | // cerr<<"got backend and SOA"<<endl; |
482 | DNSBackend *B=sd.db; | |
483 | vector<string> acl; | |
894bcf36 | 484 | s_P->getBackend()->getDomainMetadata(q->qdomain, "ALLOW-AXFR-FROM", acl); |
93afc0a3 PD |
485 | for (vector<string>::const_iterator i = acl.begin(); i != acl.end(); ++i) { |
486 | // cerr<<"matching against "<<*i<<endl; | |
487 | if(pdns_iequals(*i, "AUTO-NS")) { | |
488 | // cerr<<"AUTO-NS magic please!"<<endl; | |
489 | ||
490 | DNSResourceRecord rr; | |
7abbc40f | 491 | set<DNSName> nsset; |
93afc0a3 PD |
492 | |
493 | B->lookup(QType(QType::NS),q->qdomain); | |
494 | while(B->get(rr)) | |
290a083d | 495 | nsset.insert(DNSName(rr.content)); |
7abbc40f | 496 | for(const auto & j: nsset) { |
35b942fe | 497 | vector<string> nsips=fns.lookup(j, s_P->getBackend()); |
93afc0a3 PD |
498 | for(vector<string>::const_iterator k=nsips.begin();k!=nsips.end();++k) { |
499 | // cerr<<"got "<<*k<<" from AUTO-NS"<<endl; | |
ded6b08d | 500 | if(*k == q->getRemote().toString()) |
93afc0a3 PD |
501 | { |
502 | // cerr<<"got AUTO-NS hit"<<endl; | |
e6a9dde5 | 503 | g_log<<Logger::Warning<<"AXFR of domain '"<<q->qdomain<<"' allowed: client IP "<<q->getRemote()<<" is in NSset"<<endl; |
93afc0a3 PD |
504 | return true; |
505 | } | |
506 | } | |
507 | } | |
508 | } | |
509 | else | |
510 | { | |
511 | Netmask nm = Netmask(*i); | |
512 | if(nm.match( (ComboAddress *) &q->d_remote )) | |
513 | { | |
e6a9dde5 | 514 | g_log<<Logger::Warning<<"AXFR of domain '"<<q->qdomain<<"' allowed: client IP "<<q->getRemote()<<" is in per-domain ACL"<<endl; |
93afc0a3 PD |
515 | // cerr<<"hit!"<<endl; |
516 | return true; | |
517 | } | |
518 | } | |
519 | } | |
520 | } | |
521 | ||
12c86877 BH |
522 | extern CommunicatorClass Communicator; |
523 | ||
ded6b08d | 524 | if(Communicator.justNotified(q->qdomain, q->getRemote().toString())) { // we just notified this ip |
e6a9dde5 | 525 | g_log<<Logger::Warning<<"Approved AXFR of '"<<q->qdomain<<"' from recently notified slave "<<q->getRemote()<<endl; |
12c86877 BH |
526 | return true; |
527 | } | |
528 | ||
e6a9dde5 | 529 | g_log<<Logger::Error<<"AXFR of domain '"<<q->qdomain<<"' denied: client IP "<<q->getRemote()<<" has no permission"<<endl; |
12c86877 BH |
530 | return false; |
531 | } | |
532 | ||
b317b510 | 533 | namespace { |
54d84273 PD |
534 | struct NSECXEntry |
535 | { | |
22a0ef16 | 536 | NSECBitmap d_set; |
54d84273 | 537 | unsigned int d_ttl; |
feef1ece | 538 | bool d_auth; |
54d84273 | 539 | }; |
8e9b7d99 | 540 | |
54d84273 PD |
541 | shared_ptr<DNSPacket> getFreshAXFRPacket(shared_ptr<DNSPacket> q) |
542 | { | |
543 | shared_ptr<DNSPacket> ret = shared_ptr<DNSPacket>(q->replyPacket()); | |
544 | ret->setCompress(false); | |
545 | ret->d_dnssecOk=false; // RFC 5936, 2.2.5 | |
546 | ret->d_tcp = true; | |
547 | return ret; | |
548 | } | |
8e9b7d99 BH |
549 | } |
550 | ||
54d84273 | 551 | |
12c86877 | 552 | /** do the actual zone transfer. Return 0 in case of error, 1 in case of success */ |
7abbc40f | 553 | int TCPNameserver::doAXFR(const DNSName &target, shared_ptr<DNSPacket> q, int outsock) |
12c86877 | 554 | { |
8e9b7d99 | 555 | shared_ptr<DNSPacket> outpacket= getFreshAXFRPacket(q); |
c67e46a1 | 556 | if(q->d_dnssecOk) |
05e24311 | 557 | outpacket->d_dnssecOk=true; // RFC 5936, 2.2.5 'SHOULD' |
22893145 | 558 | |
e6a9dde5 | 559 | g_log<<Logger::Error<<"AXFR of domain '"<<target<<"' initiated by "<<q->getRemote()<<endl; |
12c86877 | 560 | |
22893145 | 561 | // determine if zone exists and AXFR is allowed using existing backend before spawning a new backend. |
12c86877 BH |
562 | SOAData sd; |
563 | { | |
564 | Lock l(&s_plock); | |
e6a9dde5 | 565 | DLOG(g_log<<"Looking for SOA"<<endl); // find domain_id via SOA and list complete domain. No SOA, no AXFR |
12a965c5 | 566 | if(!s_P) { |
e6a9dde5 | 567 | g_log<<Logger::Error<<"TCP server is without backend connections in doAXFR, launching"<<endl; |
12a965c5 BH |
568 | s_P=new PacketHandler; |
569 | } | |
12c86877 | 570 | |
ea99d474 | 571 | // canDoAXFR does all the ACL checks, and has the if(disable-axfr) shortcut, call it first. |
8090f5a2 | 572 | if (!canDoAXFR(q)) { |
e6a9dde5 | 573 | g_log<<Logger::Error<<"AXFR of domain '"<<target<<"' failed: "<<q->getRemote()<<" may not request AXFR"<<endl; |
9c556f63 | 574 | outpacket->setRcode(RCode::NotAuth); |
8090f5a2 AT |
575 | sendPacket(outpacket,outsock); |
576 | return 0; | |
577 | } | |
578 | ||
8090f5a2 | 579 | if(!s_P->getBackend()->getSOAUncached(target, sd)) { |
e6a9dde5 | 580 | g_log<<Logger::Error<<"AXFR of domain '"<<target<<"' failed: not authoritative"<<endl; |
9c556f63 | 581 | outpacket->setRcode(RCode::NotAuth); |
ff76e8b4 | 582 | sendPacket(outpacket,outsock); |
12c86877 BH |
583 | return 0; |
584 | } | |
3de83124 | 585 | } |
22893145 | 586 | |
8e9b7d99 | 587 | UeberBackend db; |
79ba7763 | 588 | if(!db.getSOAUncached(target, sd)) { |
e6a9dde5 | 589 | g_log<<Logger::Error<<"AXFR of domain '"<<target<<"' failed: not authoritative in second instance"<<endl; |
79ba7763 | 590 | outpacket->setRcode(RCode::NotAuth); |
ff76e8b4 | 591 | sendPacket(outpacket,outsock); |
3de83124 | 592 | return 0; |
12c86877 | 593 | } |
3de83124 | 594 | |
ea99d474 | 595 | DNSSECKeeper dk(&db); |
22893145 CH |
596 | dk.clearCaches(target); |
597 | bool securedZone = dk.isSecuredZone(target); | |
598 | bool presignedZone = dk.isPresigned(target); | |
599 | ||
600 | bool noAXFRBecauseOfNSEC3Narrow=false; | |
601 | NSEC3PARAMRecordContent ns3pr; | |
602 | bool narrow; | |
603 | bool NSEC3Zone=false; | |
dacacb23 | 604 | if(securedZone && dk.getNSEC3PARAM(target, &ns3pr, &narrow)) { |
22893145 CH |
605 | NSEC3Zone=true; |
606 | if(narrow) { | |
e6a9dde5 | 607 | g_log<<Logger::Error<<"Not doing AXFR of an NSEC3 narrow zone '"<<target<<"' for "<<q->getRemote()<<endl; |
22893145 CH |
608 | noAXFRBecauseOfNSEC3Narrow=true; |
609 | } | |
610 | } | |
611 | ||
612 | if(noAXFRBecauseOfNSEC3Narrow) { | |
e6a9dde5 | 613 | g_log<<Logger::Error<<"AXFR of domain '"<<target<<"' denied to "<<q->getRemote()<<endl; |
22893145 CH |
614 | outpacket->setRcode(RCode::Refused); |
615 | // FIXME: should actually figure out if we are auth over a zone, and send out 9 if we aren't | |
616 | sendPacket(outpacket,outsock); | |
617 | return 0; | |
618 | } | |
619 | ||
78bcb858 | 620 | TSIGRecordContent trc; |
7abbc40f PD |
621 | DNSName tsigkeyname; |
622 | string tsigsecret; | |
78bcb858 | 623 | |
ea3816cf | 624 | bool haveTSIGDetails = q->getTSIGDetails(&trc, &tsigkeyname); |
78bcb858 | 625 | |
60a1c204 | 626 | if(haveTSIGDetails && !tsigkeyname.empty()) { |
2c26f25a | 627 | string tsig64; |
3343ad1f | 628 | DNSName algorithm=trc.d_algoName; // FIXME400: check |
290a083d | 629 | if (algorithm == DNSName("hmac-md5.sig-alg.reg.int")) |
630 | algorithm = DNSName("hmac-md5"); | |
631 | if (algorithm != DNSName("gss-tsig")) { | |
ea99d474 | 632 | if(!db.getTSIGKey(tsigkeyname, &algorithm, &tsig64)) { |
e6a9dde5 | 633 | g_log<<Logger::Error<<"TSIG key '"<<tsigkeyname<<"' for domain '"<<target<<"' not found"<<endl; |
53ace5d5 PL |
634 | return 0; |
635 | } | |
636 | if (B64Decode(tsig64, tsigsecret) == -1) { | |
e6a9dde5 | 637 | g_log<<Logger::Error<<"Unable to Base-64 decode TSIG key '"<<tsigkeyname<<"' for domain '"<<target<<"'"<<endl; |
53ace5d5 PL |
638 | return 0; |
639 | } | |
84fc3f8b | 640 | } |
78bcb858 | 641 | } |
8e9b7d99 | 642 | |
8e9b7d99 | 643 | |
8267bd2c | 644 | // SOA *must* go out first, our signing pipe might reorder |
e6a9dde5 | 645 | DLOG(g_log<<"Sending out SOA"<<endl); |
13f9e280 CH |
646 | DNSZoneRecord soa = makeEditedDNSZRFromSOAData(dk, sd); |
647 | outpacket->addRecord(soa); | |
3c68fb14 | 648 | if(securedZone && !presignedZone) { |
7abbc40f | 649 | set<DNSName> authSet; |
8d3cbffa | 650 | authSet.insert(target); |
ea99d474 | 651 | addRRSigs(dk, db, authSet, outpacket->getRRS()); |
8d3cbffa | 652 | } |
8e9b7d99 | 653 | |
60a1c204 | 654 | if(haveTSIGDetails && !tsigkeyname.empty()) |
78bcb858 BH |
655 | outpacket->setTSIGDetails(trc, tsigkeyname, tsigsecret, trc.d_mac); // first answer is 'normal' |
656 | ||
8267bd2c | 657 | sendPacket(outpacket, outsock); |
78bcb858 BH |
658 | |
659 | trc.d_mac = outpacket->d_trc.d_mac; | |
8267bd2c BH |
660 | outpacket = getFreshAXFRPacket(q); |
661 | ||
3c68fb14 | 662 | ChunkedSigningPipe csp(target, (securedZone && !presignedZone), ::arg().asNum("signing-threads", 1)); |
8e9b7d99 | 663 | |
6e8694df | 664 | typedef map<DNSName, NSECXEntry, CanonDNSNameCompare> nsecxrepo_t; |
9d3151d9 | 665 | nsecxrepo_t nsecxrepo; |
4888e4b2 BH |
666 | |
667 | // this is where the DNSKEYs go in | |
0c350cb5 | 668 | |
4c1474f3 | 669 | DNSSECKeeper::keyset_t keys = dk.getKeys(target); |
0c350cb5 | 670 | |
90ba52e0 | 671 | DNSZoneRecord zrr; |
0c350cb5 | 672 | |
90ba52e0 | 673 | zrr.dr.d_name = target; |
674 | zrr.dr.d_ttl = sd.default_ttl; | |
675 | zrr.auth = 1; // please sign! | |
794c2f92 | 676 | |
991a0977 | 677 | string publishCDNSKEY, publishCDS; |
0900d2d3 CH |
678 | dk.getFromMeta(q->qdomain, "PUBLISH-CDNSKEY", publishCDNSKEY); |
679 | dk.getFromMeta(q->qdomain, "PUBLISH-CDS", publishCDS); | |
90ba52e0 | 680 | vector<DNSZoneRecord> cds, cdnskey; |
f889ab99 PL |
681 | DNSSECKeeper::keyset_t entryPoints = dk.getEntryPoints(q->qdomain); |
682 | set<uint32_t> entryPointIds; | |
683 | for (auto const& value : entryPoints) | |
684 | entryPointIds.insert(value.second.id); | |
991a0977 | 685 | |
ff05fd12 | 686 | for(const DNSSECKeeper::keyset_t::value_type& value : keys) { |
90ba52e0 | 687 | zrr.dr.d_type = QType::DNSKEY; |
688 | zrr.dr.d_content = std::make_shared<DNSKEYRecordContent>(value.first.getDNSKEY()); | |
6e8694df | 689 | DNSName keyname = NSEC3Zone ? DNSName(toBase32Hex(hashQNameWithSalt(ns3pr, zrr.dr.d_name))) : zrr.dr.d_name; |
9d3151d9 | 690 | NSECXEntry& ne = nsecxrepo[keyname]; |
b317b510 | 691 | |
22a0ef16 | 692 | ne.d_set.set(zrr.dr.d_type); |
794c2f92 | 693 | ne.d_ttl = sd.default_ttl; |
90ba52e0 | 694 | csp.submit(zrr); |
991a0977 PL |
695 | |
696 | // generate CDS and CDNSKEY records | |
f889ab99 | 697 | if(entryPointIds.count(value.second.id) > 0){ |
991a0977 | 698 | if(publishCDNSKEY == "1") { |
90ba52e0 | 699 | zrr.dr.d_type=QType::CDNSKEY; |
700 | zrr.dr.d_content = std::make_shared<DNSKEYRecordContent>(value.first.getDNSKEY()); | |
701 | cdnskey.push_back(zrr); | |
991a0977 PL |
702 | } |
703 | ||
704 | if(!publishCDS.empty()){ | |
90ba52e0 | 705 | zrr.dr.d_type=QType::CDS; |
991a0977 PL |
706 | vector<string> digestAlgos; |
707 | stringtok(digestAlgos, publishCDS, ", "); | |
56225bd3 | 708 | for(auto const &digestAlgo : digestAlgos) { |
90ba52e0 | 709 | zrr.dr.d_content=std::make_shared<DSRecordContent>(makeDSFromDNSKey(target, value.first.getDNSKEY(), pdns_stou(digestAlgo))); |
710 | cds.push_back(zrr); | |
991a0977 PL |
711 | } |
712 | } | |
713 | } | |
4c1474f3 | 714 | } |
0c350cb5 | 715 | |
cc8df07f | 716 | if(::arg().mustDo("direct-dnskey")) { |
6dae726d | 717 | sd.db->lookup(QType(QType::DNSKEY), target, NULL, sd.domain_id); |
90ba52e0 | 718 | while(sd.db->get(zrr)) { |
719 | zrr.dr.d_ttl = sd.default_ttl; | |
720 | csp.submit(zrr); | |
6dae726d PD |
721 | } |
722 | } | |
723 | ||
b8adb30d KM |
724 | uint8_t flags; |
725 | ||
95c5bc40 | 726 | if(NSEC3Zone) { // now stuff in the NSEC3PARAM |
b8adb30d | 727 | flags = ns3pr.d_flags; |
90ba52e0 | 728 | zrr.dr.d_type = QType::NSEC3PARAM; |
95c5bc40 | 729 | ns3pr.d_flags = 0; |
90ba52e0 | 730 | zrr.dr.d_content = std::make_shared<NSEC3PARAMRecordContent>(ns3pr); |
b8adb30d | 731 | ns3pr.d_flags = flags; |
6e8694df | 732 | DNSName keyname = DNSName(toBase32Hex(hashQNameWithSalt(ns3pr, zrr.dr.d_name))); |
ce464268 BH |
733 | NSECXEntry& ne = nsecxrepo[keyname]; |
734 | ||
22a0ef16 | 735 | ne.d_set.set(zrr.dr.d_type); |
90ba52e0 | 736 | csp.submit(zrr); |
ce464268 | 737 | } |
8e9b7d99 | 738 | |
0c350cb5 BH |
739 | // now start list zone |
740 | if(!(sd.db->list(target, sd.domain_id))) { | |
e6a9dde5 | 741 | g_log<<Logger::Error<<"Backend signals error condition"<<endl; |
9c556f63 | 742 | outpacket->setRcode(RCode::ServFail); |
0c350cb5 BH |
743 | sendPacket(outpacket,outsock); |
744 | return 0; | |
745 | } | |
746 | ||
b772ffea | 747 | |
5633a4af | 748 | const bool rectify = !(presignedZone || ::arg().mustDo("disable-axfr-rectify")); |
7abbc40f | 749 | set<DNSName> qnames, nsset, terms; |
90ba52e0 | 750 | vector<DNSZoneRecord> zrrs; |
b772ffea | 751 | |
991a0977 | 752 | // Add the CDNSKEY and CDS records we created earlier |
cece60fc CH |
753 | for (auto const &synth_zrr : cds) |
754 | zrrs.push_back(synth_zrr); | |
90ba52e0 | 755 | |
cece60fc CH |
756 | for (auto const &synth_zrr : cdnskey) |
757 | zrrs.push_back(synth_zrr); | |
90ba52e0 | 758 | |
759 | while(sd.db->get(zrr)) { | |
8bf260dd | 760 | zrr.dr.d_name.makeUsLowerCase(); |
90ba52e0 | 761 | if(zrr.dr.d_name.isPartOf(target)) { |
762 | if (zrr.dr.d_type == QType::ALIAS && ::arg().mustDo("outgoing-axfr-expand-alias")) { | |
763 | vector<DNSZoneRecord> ips; | |
764 | int ret1 = stubDoResolve(getRR<ALIASRecordContent>(zrr.dr)->d_content, QType::A, ips); | |
765 | int ret2 = stubDoResolve(getRR<ALIASRecordContent>(zrr.dr)->d_content, QType::AAAA, ips); | |
273d88b2 | 766 | if(ret1 != RCode::NoError || ret2 != RCode::NoError) { |
e6a9dde5 | 767 | g_log<<Logger::Error<<"Error resolving for ALIAS "<<zrr.dr.d_content->getZoneRepresentation()<<", aborting AXFR"<<endl; |
9c556f63 | 768 | outpacket->setRcode(RCode::ServFail); |
273d88b2 PD |
769 | sendPacket(outpacket,outsock); |
770 | return 0; | |
771 | } | |
d86e1bf7 | 772 | for(const auto& ip: ips) { |
90ba52e0 | 773 | zrr.dr.d_type = ip.dr.d_type; |
0438fd89 | 774 | zrr.dr.d_content = ip.dr.d_content; |
90ba52e0 | 775 | zrrs.push_back(zrr); |
d86e1bf7 | 776 | } |
a68df29d | 777 | continue; |
d86e1bf7 PD |
778 | } |
779 | ||
b772ffea | 780 | if (rectify) { |
90ba52e0 | 781 | if (zrr.dr.d_type) { |
782 | qnames.insert(zrr.dr.d_name); | |
783 | if(zrr.dr.d_type == QType::NS && zrr.dr.d_name!=target) | |
784 | nsset.insert(zrr.dr.d_name); | |
b772ffea KM |
785 | } else { |
786 | // remove existing ents | |
787 | continue; | |
788 | } | |
789 | } | |
a68df29d | 790 | zrrs.push_back(zrr); |
b772ffea | 791 | } else { |
90ba52e0 | 792 | if (zrr.dr.d_type) |
e6a9dde5 | 793 | g_log<<Logger::Warning<<"Zone '"<<target<<"' contains out-of-zone data '"<<zrr.dr.d_name<<"|"<<DNSRecordContent::NumberToType(zrr.dr.d_type)<<"', ignoring"<<endl; |
b772ffea KM |
794 | } |
795 | } | |
796 | ||
75f2589f | 797 | // Group records by name and type, signpipe stumbles over interrupted rrsets |
22a676e0 | 798 | if(securedZone && !presignedZone) { |
8daafcc1 KM |
799 | sort(zrrs.begin(), zrrs.end(), [](const DNSZoneRecord& a, const DNSZoneRecord& b) { |
800 | return tie(a.dr.d_name, a.dr.d_type) < tie(b.dr.d_name, b.dr.d_type); | |
801 | }); | |
802 | } | |
75f2589f | 803 | |
b772ffea KM |
804 | if(rectify) { |
805 | // set auth | |
2010ac95 RG |
806 | for(DNSZoneRecord &loopZRR : zrrs) { |
807 | loopZRR.auth=true; | |
808 | if (loopZRR.dr.d_type != QType::NS || loopZRR.dr.d_name!=target) { | |
809 | DNSName shorter(loopZRR.dr.d_name); | |
b772ffea | 810 | do { |
e325f20c | 811 | if (shorter==target) // apex is always auth |
cb045f61 | 812 | break; |
2010ac95 RG |
813 | if(nsset.count(shorter) && !(loopZRR.dr.d_name==shorter && loopZRR.dr.d_type == QType::DS)) { |
814 | loopZRR.auth=false; | |
cb045f61 | 815 | break; |
9f70b77a | 816 | } |
7abbc40f | 817 | } while(shorter.chopOff()); |
9f70b77a | 818 | } |
b772ffea KM |
819 | } |
820 | ||
821 | if(NSEC3Zone) { | |
822 | // ents are only required for NSEC3 zones | |
823 | uint32_t maxent = ::arg().asNum("max-ent-entries"); | |
6ded341a | 824 | set<DNSName> nsec3set, nonterm; |
2010ac95 | 825 | for (auto &loopZRR: zrrs) { |
6ded341a | 826 | bool skip=false; |
2010ac95 | 827 | DNSName shorter = loopZRR.dr.d_name; |
6ded341a KM |
828 | if (shorter != target && shorter.chopOff() && shorter != target) { |
829 | do { | |
830 | if(nsset.count(shorter)) { | |
831 | skip=true; | |
832 | break; | |
833 | } | |
834 | } while(shorter.chopOff() && shorter != target); | |
835 | } | |
2010ac95 RG |
836 | shorter = loopZRR.dr.d_name; |
837 | if(!skip && (loopZRR.dr.d_type != QType::NS || !ns3pr.d_flags)) { | |
6ded341a KM |
838 | do { |
839 | if(!nsec3set.count(shorter)) { | |
840 | nsec3set.insert(shorter); | |
841 | } | |
842 | } while(shorter != target && shorter.chopOff()); | |
843 | } | |
844 | } | |
845 | ||
2010ac95 RG |
846 | for(DNSZoneRecord &loopZRR : zrrs) { |
847 | DNSName shorter(loopZRR.dr.d_name); | |
e325f20c | 848 | while(shorter != target && shorter.chopOff()) { |
6ded341a | 849 | if(!qnames.count(shorter) && !nonterm.count(shorter) && nsec3set.count(shorter)) { |
b772ffea | 850 | if(!(maxent)) { |
e6a9dde5 | 851 | g_log<<Logger::Warning<<"Zone '"<<target<<"' has too many empty non terminals."<<endl; |
b772ffea KM |
852 | return 0; |
853 | } | |
6ded341a KM |
854 | nonterm.insert(shorter); |
855 | --maxent; | |
b772ffea KM |
856 | } |
857 | } | |
858 | } | |
859 | ||
9e23e712 | 860 | for(const auto& nt : nonterm) { |
2010ac95 RG |
861 | DNSZoneRecord tempRR; |
862 | tempRR.dr.d_name=nt; | |
863 | tempRR.dr.d_type=QType::ENT; | |
864 | tempRR.auth=true; | |
865 | zrrs.push_back(tempRR); | |
b772ffea KM |
866 | } |
867 | } | |
868 | } | |
869 | ||
870 | ||
12c86877 | 871 | /* now write all other records */ |
8e9b7d99 | 872 | |
6e8694df | 873 | DNSName keyname; |
3370c993 | 874 | unsigned int udiff; |
1c6d9830 BH |
875 | DTime dt; |
876 | dt.set(); | |
bec14a20 | 877 | int records=0; |
2010ac95 | 878 | for(DNSZoneRecord &loopZRR : zrrs) { |
3c68fb14 | 879 | if (!presignedZone && loopZRR.dr.d_type == QType::RRSIG) |
794c2f92 | 880 | continue; |
6dae726d | 881 | |
991a0977 | 882 | // only skip the DNSKEY, CDNSKEY and CDS if direct-dnskey is enabled, to avoid changing behaviour |
6dae726d | 883 | // when it is not enabled. |
2010ac95 | 884 | if(::arg().mustDo("direct-dnskey") && (loopZRR.dr.d_type == QType::DNSKEY || loopZRR.dr.d_type == QType::CDNSKEY || loopZRR.dr.d_type == QType::CDS)) |
6dae726d PD |
885 | continue; |
886 | ||
bec14a20 | 887 | records++; |
2010ac95 RG |
888 | if(securedZone && (loopZRR.auth || loopZRR.dr.d_type == QType::NS)) { |
889 | if (NSEC3Zone || loopZRR.dr.d_type) { | |
3c68fb14 KM |
890 | if (presignedZone && NSEC3Zone && loopZRR.dr.d_type == QType::RRSIG && getRR<RRSIGRecordContent>(loopZRR.dr)->d_type == QType::NSEC3) { |
891 | keyname = loopZRR.dr.d_name.makeRelative(sd.qname); | |
892 | } else { | |
893 | keyname = NSEC3Zone ? DNSName(toBase32Hex(hashQNameWithSalt(ns3pr, loopZRR.dr.d_name))) : loopZRR.dr.d_name; | |
894 | } | |
b5baefaf PD |
895 | NSECXEntry& ne = nsecxrepo[keyname]; |
896 | ne.d_ttl = sd.default_ttl; | |
3c68fb14 KM |
897 | ne.d_auth = (ne.d_auth || loopZRR.auth || (NSEC3Zone && (!ns3pr.d_flags))); |
898 | if (loopZRR.dr.d_type && loopZRR.dr.d_type != QType::RRSIG) { | |
22a0ef16 | 899 | ne.d_set.set(loopZRR.dr.d_type); |
b5baefaf PD |
900 | } |
901 | } | |
b317b510 | 902 | } |
b5baefaf | 903 | |
2010ac95 | 904 | if (!loopZRR.dr.d_type) |
b5baefaf PD |
905 | continue; // skip empty non-terminals |
906 | ||
2010ac95 | 907 | if(loopZRR.dr.d_type == QType::SOA) |
12c86877 | 908 | continue; // skip SOA - would indicate end of AXFR |
add640c0 | 909 | |
2010ac95 | 910 | if(csp.submit(loopZRR)) { |
1c6d9830 BH |
911 | for(;;) { |
912 | outpacket->getRRS() = csp.getChunk(); | |
913 | if(!outpacket->getRRS().empty()) { | |
60a1c204 | 914 | if(haveTSIGDetails && !tsigkeyname.empty()) |
54d84273 | 915 | outpacket->setTSIGDetails(trc, tsigkeyname, tsigsecret, trc.d_mac, true); |
1c6d9830 | 916 | sendPacket(outpacket, outsock); |
78bcb858 | 917 | trc.d_mac=outpacket->d_trc.d_mac; |
1c6d9830 BH |
918 | outpacket=getFreshAXFRPacket(q); |
919 | } | |
920 | else | |
921 | break; | |
922 | } | |
12c86877 BH |
923 | } |
924 | } | |
78bcb858 | 925 | /* |
3370c993 | 926 | udiff=dt.udiffNoReset(); |
1c6d9830 BH |
927 | cerr<<"Starting NSEC: "<<csp.d_signed/(udiff/1000000.0)<<" sigs/s, "<<csp.d_signed<<" / "<<udiff/1000000.0<<endl; |
928 | cerr<<"Outstanding: "<<csp.d_outstanding<<", "<<csp.d_queued - csp.d_signed << endl; | |
929 | cerr<<"Ready for consumption: "<<csp.getReady()<<endl; | |
78bcb858 | 930 | */ |
feef1ece | 931 | if(securedZone) { |
4888e4b2 | 932 | if(NSEC3Zone) { |
9d3151d9 | 933 | for(nsecxrepo_t::const_iterator iter = nsecxrepo.begin(); iter != nsecxrepo.end(); ++iter) { |
3c68fb14 | 934 | if(iter->second.d_auth) { |
feef1ece | 935 | NSEC3RecordContent n3rc; |
22a0ef16 | 936 | n3rc.set(iter->second.d_set); |
27d4a65b RG |
937 | const auto numberOfTypesSet = n3rc.numberOfTypesSet(); |
938 | if (numberOfTypesSet != 0 && (numberOfTypesSet != 1 || !n3rc.isSet(QType::NS))) { | |
939 | n3rc.set(QType::RRSIG); | |
940 | } | |
941 | n3rc.d_salt = ns3pr.d_salt; | |
feef1ece PD |
942 | n3rc.d_flags = ns3pr.d_flags; |
943 | n3rc.d_iterations = ns3pr.d_iterations; | |
27d4a65b | 944 | n3rc.d_algorithm = DNSSECKeeper::SHA1; // SHA1, fixed in PowerDNS for now |
feef1ece | 945 | nsecxrepo_t::const_iterator inext = iter; |
cb167afd | 946 | ++inext; |
feef1ece PD |
947 | if(inext == nsecxrepo.end()) |
948 | inext = nsecxrepo.begin(); | |
3c68fb14 | 949 | while(!inext->second.d_auth && inext != iter) |
feef1ece | 950 | { |
cb167afd | 951 | ++inext; |
feef1ece PD |
952 | if(inext == nsecxrepo.end()) |
953 | inext = nsecxrepo.begin(); | |
954 | } | |
6e8694df KM |
955 | n3rc.d_nexthash = fromBase32Hex(inext->first.toStringNoDot()); |
956 | zrr.dr.d_name = iter->first+sd.qname; | |
90ba52e0 | 957 | |
958 | zrr.dr.d_ttl = sd.default_ttl; | |
27d4a65b | 959 | zrr.dr.d_content = std::make_shared<NSEC3RecordContent>(std::move(n3rc)); |
90ba52e0 | 960 | zrr.dr.d_type = QType::NSEC3; |
961 | zrr.dr.d_place = DNSResourceRecord::ANSWER; | |
962 | zrr.auth=true; | |
963 | if(csp.submit(zrr)) { | |
feef1ece PD |
964 | for(;;) { |
965 | outpacket->getRRS() = csp.getChunk(); | |
966 | if(!outpacket->getRRS().empty()) { | |
60a1c204 | 967 | if(haveTSIGDetails && !tsigkeyname.empty()) |
feef1ece PD |
968 | outpacket->setTSIGDetails(trc, tsigkeyname, tsigsecret, trc.d_mac, true); |
969 | sendPacket(outpacket, outsock); | |
970 | trc.d_mac=outpacket->d_trc.d_mac; | |
971 | outpacket=getFreshAXFRPacket(q); | |
972 | } | |
973 | else | |
974 | break; | |
1c6d9830 | 975 | } |
1c6d9830 | 976 | } |
8e9b7d99 | 977 | } |
4888e4b2 BH |
978 | } |
979 | } | |
9d3151d9 | 980 | else for(nsecxrepo_t::const_iterator iter = nsecxrepo.begin(); iter != nsecxrepo.end(); ++iter) { |
ed9c3a50 | 981 | NSECRecordContent nrc; |
22a0ef16 | 982 | nrc.set(iter->second.d_set); |
27d4a65b RG |
983 | nrc.set(QType::RRSIG); |
984 | nrc.set(QType::NSEC); | |
6e8694df KM |
985 | |
986 | if(boost::next(iter) != nsecxrepo.end()) | |
987 | nrc.d_next = boost::next(iter)->first; | |
ed9c3a50 | 988 | else |
6e8694df KM |
989 | nrc.d_next=nsecxrepo.begin()->first; |
990 | zrr.dr.d_name = iter->first; | |
991 | ||
90ba52e0 | 992 | zrr.dr.d_ttl = sd.default_ttl; |
27d4a65b | 993 | zrr.dr.d_content = std::make_shared<NSECRecordContent>(std::move(nrc)); |
90ba52e0 | 994 | zrr.dr.d_type = QType::NSEC; |
995 | zrr.dr.d_place = DNSResourceRecord::ANSWER; | |
996 | zrr.auth=true; | |
997 | if(csp.submit(zrr)) { | |
1c6d9830 BH |
998 | for(;;) { |
999 | outpacket->getRRS() = csp.getChunk(); | |
1000 | if(!outpacket->getRRS().empty()) { | |
60a1c204 | 1001 | if(haveTSIGDetails && !tsigkeyname.empty()) |
78bcb858 | 1002 | outpacket->setTSIGDetails(trc, tsigkeyname, tsigsecret, trc.d_mac, true); |
1c6d9830 | 1003 | sendPacket(outpacket, outsock); |
78bcb858 | 1004 | trc.d_mac=outpacket->d_trc.d_mac; |
1c6d9830 BH |
1005 | outpacket=getFreshAXFRPacket(q); |
1006 | } | |
1007 | else | |
1008 | break; | |
1009 | } | |
8e9b7d99 | 1010 | } |
add640c0 | 1011 | } |
add640c0 | 1012 | } |
78bcb858 | 1013 | /* |
3370c993 | 1014 | udiff=dt.udiffNoReset(); |
1c6d9830 BH |
1015 | cerr<<"Flushing pipe: "<<csp.d_signed/(udiff/1000000.0)<<" sigs/s, "<<csp.d_signed<<" / "<<udiff/1000000.0<<endl; |
1016 | cerr<<"Outstanding: "<<csp.d_outstanding<<", "<<csp.d_queued - csp.d_signed << endl; | |
1017 | cerr<<"Ready for consumption: "<<csp.getReady()<<endl; | |
78bcb858 | 1018 | * */ |
bec14a20 BH |
1019 | for(;;) { |
1020 | outpacket->getRRS() = csp.getChunk(true); // flush the pipe | |
1021 | if(!outpacket->getRRS().empty()) { | |
60a1c204 | 1022 | if(haveTSIGDetails && !tsigkeyname.empty()) |
78bcb858 | 1023 | outpacket->setTSIGDetails(trc, tsigkeyname, tsigsecret, trc.d_mac, true); // first answer is 'normal' |
bec14a20 | 1024 | sendPacket(outpacket, outsock); |
78bcb858 | 1025 | trc.d_mac=outpacket->d_trc.d_mac; |
bec14a20 BH |
1026 | outpacket=getFreshAXFRPacket(q); |
1027 | } | |
1028 | else | |
1029 | break; | |
12c86877 | 1030 | } |
8e9b7d99 | 1031 | |
1c6d9830 | 1032 | udiff=dt.udiffNoReset(); |
f1f85f12 | 1033 | if(securedZone) |
e6a9dde5 | 1034 | g_log<<Logger::Info<<"Done signing: "<<csp.d_signed/(udiff/1000000.0)<<" sigs/s, "<<endl; |
1c6d9830 | 1035 | |
e6a9dde5 | 1036 | DLOG(g_log<<"Done writing out records"<<endl); |
12c86877 | 1037 | /* and terminate with yet again the SOA record */ |
8e9b7d99 | 1038 | outpacket=getFreshAXFRPacket(q); |
13f9e280 | 1039 | outpacket->addRecord(soa); |
60a1c204 | 1040 | if(haveTSIGDetails && !tsigkeyname.empty()) |
78bcb858 | 1041 | outpacket->setTSIGDetails(trc, tsigkeyname, tsigsecret, trc.d_mac, true); |
92c90b44 | 1042 | |
ff76e8b4 | 1043 | sendPacket(outpacket, outsock); |
78bcb858 | 1044 | |
e6a9dde5 PL |
1045 | DLOG(g_log<<"last packet - close"<<endl); |
1046 | g_log<<Logger::Error<<"AXFR of domain '"<<target<<"' to "<<q->getRemote()<<" finished"<<endl; | |
12c86877 BH |
1047 | |
1048 | return 1; | |
1049 | } | |
1050 | ||
6e59a580 KM |
1051 | int TCPNameserver::doIXFR(shared_ptr<DNSPacket> q, int outsock) |
1052 | { | |
1053 | shared_ptr<DNSPacket> outpacket=getFreshAXFRPacket(q); | |
1054 | if(q->d_dnssecOk) | |
1055 | outpacket->d_dnssecOk=true; // RFC 5936, 2.2.5 'SHOULD' | |
1056 | ||
6e59a580 | 1057 | uint32_t serial = 0; |
27c0050c | 1058 | MOADNSParser mdp(false, q->getString()); |
6e59a580 KM |
1059 | for(MOADNSParser::answers_t::const_iterator i=mdp.d_answers.begin(); i != mdp.d_answers.end(); ++i) { |
1060 | const DNSRecord *rr = &i->first; | |
e693ff5a | 1061 | if (rr->d_type == QType::SOA && rr->d_place == DNSResourceRecord::AUTHORITY) { |
6e59a580 KM |
1062 | vector<string>parts; |
1063 | stringtok(parts, rr->d_content->getZoneRepresentation()); | |
1064 | if (parts.size() >= 3) { | |
95dd3b90 RG |
1065 | try { |
1066 | serial=pdns_stou(parts[2]); | |
1067 | } | |
1068 | catch(const std::out_of_range& oor) { | |
e6a9dde5 | 1069 | g_log<<Logger::Error<<"Invalid serial in IXFR query"<<endl; |
95dd3b90 RG |
1070 | outpacket->setRcode(RCode::FormErr); |
1071 | sendPacket(outpacket,outsock); | |
1072 | return 0; | |
1073 | } | |
6e59a580 | 1074 | } else { |
e6a9dde5 | 1075 | g_log<<Logger::Error<<"No serial in IXFR query"<<endl; |
6e59a580 KM |
1076 | outpacket->setRcode(RCode::FormErr); |
1077 | sendPacket(outpacket,outsock); | |
1078 | return 0; | |
1079 | } | |
3e67ea8b | 1080 | } else if (rr->d_type != QType::TSIG && rr->d_type != QType::OPT) { |
e6a9dde5 | 1081 | g_log<<Logger::Error<<"Additional records in IXFR query, type: "<<QType(rr->d_type).getName()<<endl; |
6e59a580 KM |
1082 | outpacket->setRcode(RCode::FormErr); |
1083 | sendPacket(outpacket,outsock); | |
1084 | return 0; | |
1085 | } | |
1086 | } | |
1087 | ||
e6a9dde5 | 1088 | g_log<<Logger::Error<<"IXFR of domain '"<<q->qdomain<<"' initiated by "<<q->getRemote()<<" with serial "<<serial<<endl; |
6e59a580 | 1089 | |
22893145 | 1090 | // determine if zone exists and AXFR is allowed using existing backend before spawning a new backend. |
6e59a580 | 1091 | SOAData sd; |
6e59a580 KM |
1092 | { |
1093 | Lock l(&s_plock); | |
e6a9dde5 | 1094 | DLOG(g_log<<"Looking for SOA"<<endl); // find domain_id via SOA and list complete domain. No SOA, no IXFR |
6e59a580 | 1095 | if(!s_P) { |
e6a9dde5 | 1096 | g_log<<Logger::Error<<"TCP server is without backend connections in doIXFR, launching"<<endl; |
6e59a580 KM |
1097 | s_P=new PacketHandler; |
1098 | } | |
1099 | ||
22893145 CH |
1100 | // canDoAXFR does all the ACL checks, and has the if(disable-axfr) shortcut, call it first. |
1101 | if(!canDoAXFR(q) || !s_P->getBackend()->getSOAUncached(q->qdomain, sd)) { | |
e6a9dde5 | 1102 | g_log<<Logger::Error<<"IXFR of domain '"<<q->qdomain<<"' failed: not authoritative"<<endl; |
9c556f63 | 1103 | outpacket->setRcode(RCode::NotAuth); |
6e59a580 KM |
1104 | sendPacket(outpacket,outsock); |
1105 | return 0; | |
1106 | } | |
1107 | } | |
1108 | ||
22893145 CH |
1109 | DNSSECKeeper dk; |
1110 | NSEC3PARAMRecordContent ns3pr; | |
1111 | bool narrow; | |
1112 | ||
1113 | dk.clearCaches(q->qdomain); | |
1114 | bool securedZone = dk.isSecuredZone(q->qdomain); | |
1115 | if(dk.getNSEC3PARAM(q->qdomain, &ns3pr, &narrow)) { | |
1116 | if(narrow) { | |
e6a9dde5 PL |
1117 | g_log<<Logger::Error<<"Not doing IXFR of an NSEC3 narrow zone."<<endl; |
1118 | g_log<<Logger::Error<<"IXFR of domain '"<<q->qdomain<<"' denied to "<<q->getRemote()<<endl; | |
22893145 CH |
1119 | outpacket->setRcode(RCode::Refused); |
1120 | sendPacket(outpacket,outsock); | |
1121 | return 0; | |
1122 | } | |
1123 | } | |
1124 | ||
7abbc40f | 1125 | DNSName target = q->qdomain; |
6e59a580 KM |
1126 | |
1127 | UeberBackend db; | |
79ba7763 | 1128 | if(!db.getSOAUncached(target, sd)) { |
e6a9dde5 | 1129 | g_log<<Logger::Error<<"IXFR of domain '"<<target<<"' failed: not authoritative in second instance"<<endl; |
79ba7763 | 1130 | outpacket->setRcode(RCode::NotAuth); |
6e59a580 KM |
1131 | sendPacket(outpacket,outsock); |
1132 | return 0; | |
1133 | } | |
24d9e514 | 1134 | |
13f9e280 | 1135 | if (!rfc1982LessThan(serial, calculateEditSOA(sd.serial, dk, sd.qname))) { |
6e59a580 | 1136 | TSIGRecordContent trc; |
7abbc40f PD |
1137 | DNSName tsigkeyname; |
1138 | string tsigsecret; | |
6e59a580 | 1139 | |
ea3816cf | 1140 | bool haveTSIGDetails = q->getTSIGDetails(&trc, &tsigkeyname); |
6e59a580 | 1141 | |
60a1c204 | 1142 | if(haveTSIGDetails && !tsigkeyname.empty()) { |
bb7fb11c | 1143 | string tsig64; |
3343ad1f | 1144 | DNSName algorithm=trc.d_algoName; // FIXME400: was toLowerCanonic, compare output |
290a083d | 1145 | if (algorithm == DNSName("hmac-md5.sig-alg.reg.int")) |
1146 | algorithm = DNSName("hmac-md5"); | |
6e59a580 | 1147 | Lock l(&s_plock); |
53ace5d5 | 1148 | if(!s_P->getBackend()->getTSIGKey(tsigkeyname, &algorithm, &tsig64)) { |
e6a9dde5 | 1149 | g_log<<Logger::Error<<"TSIG key '"<<tsigkeyname<<"' for domain '"<<target<<"' not found"<<endl; |
53ace5d5 PL |
1150 | return 0; |
1151 | } | |
1152 | if (B64Decode(tsig64, tsigsecret) == -1) { | |
e6a9dde5 | 1153 | g_log<<Logger::Error<<"Unable to Base-64 decode TSIG key '"<<tsigkeyname<<"' for domain '"<<target<<"'"<<endl; |
53ace5d5 PL |
1154 | return 0; |
1155 | } | |
6e59a580 KM |
1156 | } |
1157 | ||
1158 | UeberBackend signatureDB; | |
1159 | ||
1160 | // SOA *must* go out first, our signing pipe might reorder | |
e6a9dde5 | 1161 | DLOG(g_log<<"Sending out SOA"<<endl); |
13f9e280 CH |
1162 | DNSZoneRecord soa = makeEditedDNSZRFromSOAData(dk, sd); |
1163 | outpacket->addRecord(soa); | |
6e59a580 | 1164 | if(securedZone) { |
7abbc40f | 1165 | set<DNSName> authSet; |
6e59a580 KM |
1166 | authSet.insert(target); |
1167 | addRRSigs(dk, signatureDB, authSet, outpacket->getRRS()); | |
1168 | } | |
1169 | ||
60a1c204 | 1170 | if(haveTSIGDetails && !tsigkeyname.empty()) |
6e59a580 KM |
1171 | outpacket->setTSIGDetails(trc, tsigkeyname, tsigsecret, trc.d_mac); // first answer is 'normal' |
1172 | ||
1173 | sendPacket(outpacket, outsock); | |
1174 | ||
e6a9dde5 | 1175 | g_log<<Logger::Error<<"IXFR of domain '"<<target<<"' to "<<q->getRemote()<<" finished"<<endl; |
6e59a580 KM |
1176 | |
1177 | return 1; | |
1178 | } | |
1179 | ||
e6a9dde5 | 1180 | g_log<<Logger::Error<<"IXFR fallback to AXFR for domain '"<<target<<"' our serial "<<sd.serial<<endl; |
6e59a580 KM |
1181 | return doAXFR(q->qdomain, q, outsock); |
1182 | } | |
1183 | ||
12c86877 BH |
1184 | TCPNameserver::~TCPNameserver() |
1185 | { | |
1186 | delete d_connectionroom_sem; | |
1187 | } | |
1188 | ||
1189 | TCPNameserver::TCPNameserver() | |
1190 | { | |
cb0af1a1 RG |
1191 | d_maxTransactionsPerConn = ::arg().asNum("max-tcp-transactions-per-conn"); |
1192 | d_idleTimeout = ::arg().asNum("tcp-idle-timeout"); | |
1193 | d_maxConnectionDuration = ::arg().asNum("max-tcp-connection-duration"); | |
1194 | d_maxConnectionsPerClient = ::arg().asNum("max-tcp-connections-per-client"); | |
1195 | ||
379ab445 BH |
1196 | // sem_init(&d_connectionroom_sem,0,::arg().asNum("max-tcp-connections")); |
1197 | d_connectionroom_sem = new Semaphore( ::arg().asNum( "max-tcp-connections" )); | |
117e1bf2 | 1198 | d_tid=0; |
12c86877 | 1199 | vector<string>locals; |
379ab445 | 1200 | stringtok(locals,::arg()["local-address"]," ,"); |
12c86877 BH |
1201 | |
1202 | vector<string>locals6; | |
379ab445 | 1203 | stringtok(locals6,::arg()["local-ipv6"]," ,"); |
12c86877 | 1204 | |
12c86877 | 1205 | if(locals.empty() && locals6.empty()) |
3f81d239 | 1206 | throw PDNSException("No local address specified"); |
12c86877 | 1207 | |
68b011bd | 1208 | d_ng.toMasks(::arg()["allow-axfr-ips"] ); |
9f1d5826 | 1209 | |
12c86877 | 1210 | signal(SIGPIPE,SIG_IGN); |
12c86877 BH |
1211 | |
1212 | for(vector<string>::const_iterator laddr=locals.begin();laddr!=locals.end();++laddr) { | |
12c86877 | 1213 | int s=socket(AF_INET,SOCK_STREAM,0); |
326484be | 1214 | |
12c86877 | 1215 | if(s<0) |
3f81d239 | 1216 | throw PDNSException("Unable to acquire TCP socket: "+stringerror()); |
12c86877 | 1217 | |
3897b9e1 | 1218 | setCloseOnExec(s); |
fb316318 | 1219 | |
379ab445 | 1220 | ComboAddress local(*laddr, ::arg().asNum("local-port")); |
12c86877 BH |
1221 | |
1222 | int tmp=1; | |
1223 | if(setsockopt(s,SOL_SOCKET,SO_REUSEADDR,(char*)&tmp,sizeof tmp)<0) { | |
e6a9dde5 | 1224 | g_log<<Logger::Error<<"Setsockopt failed"<<endl; |
5bd2ea7b | 1225 | _exit(1); |
12c86877 | 1226 | } |
940d7811 RG |
1227 | |
1228 | if (::arg().asNum("tcp-fast-open") > 0) { | |
1229 | #ifdef TCP_FASTOPEN | |
1230 | int fastOpenQueueSize = ::arg().asNum("tcp-fast-open"); | |
1231 | if (setsockopt(s, IPPROTO_TCP, TCP_FASTOPEN, &fastOpenQueueSize, sizeof fastOpenQueueSize) < 0) { | |
e6a9dde5 | 1232 | g_log<<Logger::Error<<"Failed to enable TCP Fast Open for listening socket: "<<strerror(errno)<<endl; |
940d7811 RG |
1233 | } |
1234 | #else | |
e6a9dde5 | 1235 | g_log<<Logger::Warning<<"TCP Fast Open configured but not supported for listening socket"<<endl; |
940d7811 RG |
1236 | #endif |
1237 | } | |
1238 | ||
fec7dd5a SS |
1239 | if( ::arg().mustDo("non-local-bind") ) |
1240 | Utility::setBindAny(AF_INET, s); | |
1241 | ||
379ab445 | 1242 | if(::bind(s, (sockaddr*)&local, local.getSocklen())<0) { |
2c896042 | 1243 | close(s); |
5ecb2885 | 1244 | if( errno == EADDRNOTAVAIL && ! ::arg().mustDo("local-address-nonexist-fail") ) { |
e6a9dde5 | 1245 | g_log<<Logger::Error<<"IPv4 Address " << *laddr << " does not exist on this server - skipping TCP bind" << endl; |
5ecb2885 MZ |
1246 | continue; |
1247 | } else { | |
e6a9dde5 | 1248 | g_log<<Logger::Error<<"Unable to bind to TCP socket " << *laddr << ": "<<strerror(errno)<<endl; |
2ab7e9ac | 1249 | throw PDNSException("Unable to bind to TCP socket"); |
5ecb2885 | 1250 | } |
12c86877 BH |
1251 | } |
1252 | ||
1253 | listen(s,128); | |
e6a9dde5 | 1254 | g_log<<Logger::Error<<"TCP server bound to "<<local.toStringWithPort()<<endl; |
12c86877 | 1255 | d_sockets.push_back(s); |
8edfedf1 BH |
1256 | struct pollfd pfd; |
1257 | memset(&pfd, 0, sizeof(pfd)); | |
1258 | pfd.fd = s; | |
1259 | pfd.events = POLLIN; | |
1260 | ||
1261 | d_prfds.push_back(pfd); | |
12c86877 BH |
1262 | } |
1263 | ||
12c86877 | 1264 | for(vector<string>::const_iterator laddr=locals6.begin();laddr!=locals6.end();++laddr) { |
12c86877 BH |
1265 | int s=socket(AF_INET6,SOCK_STREAM,0); |
1266 | ||
1267 | if(s<0) | |
3f81d239 | 1268 | throw PDNSException("Unable to acquire TCPv6 socket: "+stringerror()); |
178d5134 | 1269 | |
3897b9e1 | 1270 | setCloseOnExec(s); |
fb316318 | 1271 | |
379ab445 | 1272 | ComboAddress local(*laddr, ::arg().asNum("local-port")); |
12c86877 BH |
1273 | |
1274 | int tmp=1; | |
1275 | if(setsockopt(s,SOL_SOCKET,SO_REUSEADDR,(char*)&tmp,sizeof tmp)<0) { | |
e6a9dde5 | 1276 | g_log<<Logger::Error<<"Setsockopt failed"<<endl; |
5bd2ea7b | 1277 | _exit(1); |
12c86877 | 1278 | } |
940d7811 RG |
1279 | |
1280 | if (::arg().asNum("tcp-fast-open") > 0) { | |
1281 | #ifdef TCP_FASTOPEN | |
1282 | int fastOpenQueueSize = ::arg().asNum("tcp-fast-open"); | |
1283 | if (setsockopt(s, IPPROTO_TCP, TCP_FASTOPEN, &fastOpenQueueSize, sizeof fastOpenQueueSize) < 0) { | |
e6a9dde5 | 1284 | g_log<<Logger::Error<<"Failed to enable TCP Fast Open for listening socket: "<<strerror(errno)<<endl; |
940d7811 RG |
1285 | } |
1286 | #else | |
e6a9dde5 | 1287 | g_log<<Logger::Warning<<"TCP Fast Open configured but not supported for listening socket"<<endl; |
940d7811 RG |
1288 | #endif |
1289 | } | |
1290 | ||
fec7dd5a SS |
1291 | if( ::arg().mustDo("non-local-bind") ) |
1292 | Utility::setBindAny(AF_INET6, s); | |
326484be | 1293 | if(setsockopt(s, IPPROTO_IPV6, IPV6_V6ONLY, &tmp, sizeof(tmp)) < 0) { |
e6a9dde5 | 1294 | g_log<<Logger::Error<<"Failed to set IPv6 socket to IPv6 only, continuing anyhow: "<<strerror(errno)<<endl; |
326484be | 1295 | } |
ff76e8b4 | 1296 | if(bind(s, (const sockaddr*)&local, local.getSocklen())<0) { |
2c896042 | 1297 | close(s); |
5ecb2885 | 1298 | if( errno == EADDRNOTAVAIL && ! ::arg().mustDo("local-ipv6-nonexist-fail") ) { |
e6a9dde5 | 1299 | g_log<<Logger::Error<<"IPv6 Address " << *laddr << " does not exist on this server - skipping TCP bind" << endl; |
5ecb2885 MZ |
1300 | continue; |
1301 | } else { | |
e6a9dde5 | 1302 | g_log<<Logger::Error<<"Unable to bind to TCPv6 socket" << *laddr << ": "<<strerror(errno)<<endl; |
2ab7e9ac | 1303 | throw PDNSException("Unable to bind to TCPv6 socket"); |
5ecb2885 | 1304 | } |
12c86877 BH |
1305 | } |
1306 | ||
1307 | listen(s,128); | |
e6a9dde5 | 1308 | g_log<<Logger::Error<<"TCPv6 server bound to "<<local.toStringWithPort()<<endl; // this gets %eth0 right |
12c86877 | 1309 | d_sockets.push_back(s); |
8edfedf1 BH |
1310 | |
1311 | struct pollfd pfd; | |
1312 | memset(&pfd, 0, sizeof(pfd)); | |
1313 | pfd.fd = s; | |
1314 | pfd.events = POLLIN; | |
1315 | ||
1316 | d_prfds.push_back(pfd); | |
12c86877 | 1317 | } |
12c86877 BH |
1318 | } |
1319 | ||
1320 | ||
ff76e8b4 | 1321 | //! Start of TCP operations thread, we launch a new thread for each incoming TCP question |
12c86877 BH |
1322 | void TCPNameserver::thread() |
1323 | { | |
519f5484 | 1324 | setThreadName("pdns/tcpnameser"); |
12c86877 BH |
1325 | try { |
1326 | for(;;) { | |
1327 | int fd; | |
cb0af1a1 RG |
1328 | ComboAddress remote; |
1329 | Utility::socklen_t addrlen=remote.getSocklen(); | |
12c86877 | 1330 | |
8edfedf1 | 1331 | int ret=poll(&d_prfds[0], d_prfds.size(), -1); // blocks, forever if need be |
8a63d3ce | 1332 | if(ret <= 0) |
4957a608 | 1333 | continue; |
8a63d3ce | 1334 | |
12c86877 | 1335 | int sock=-1; |
8ce9e4e6 | 1336 | for(const pollfd& pfd : d_prfds) { |
c1ee10a6 | 1337 | if(pfd.revents & POLLIN) { |
4957a608 | 1338 | sock = pfd.fd; |
cb0af1a1 RG |
1339 | remote.sin4.sin_family = AF_INET6; |
1340 | addrlen=remote.getSocklen(); | |
4957a608 BH |
1341 | |
1342 | if((fd=accept(sock, (sockaddr*)&remote, &addrlen))<0) { | |
e6a9dde5 | 1343 | g_log<<Logger::Error<<"TCP question accept error: "<<strerror(errno)<<endl; |
4957a608 BH |
1344 | |
1345 | if(errno==EMFILE) { | |
e6a9dde5 | 1346 | g_log<<Logger::Error<<"TCP handler out of filedescriptors, exiting, won't recover from this"<<endl; |
5bd2ea7b | 1347 | _exit(1); |
4957a608 BH |
1348 | } |
1349 | } | |
1350 | else { | |
cb0af1a1 RG |
1351 | if (d_maxConnectionsPerClient) { |
1352 | std::lock_guard<std::mutex> lock(s_clientsCountMutex); | |
1353 | if (s_clientsCount[remote] >= d_maxConnectionsPerClient) { | |
e6a9dde5 | 1354 | g_log<<Logger::Notice<<"Limit of simultaneous TCP connections per client reached for "<< remote<<", dropping"<<endl; |
cb0af1a1 RG |
1355 | close(fd); |
1356 | continue; | |
1357 | } | |
1358 | s_clientsCount[remote]++; | |
1359 | } | |
1360 | ||
4957a608 BH |
1361 | pthread_t tid; |
1362 | d_connectionroom_sem->wait(); // blocks if no connections are available | |
1363 | ||
1364 | int room; | |
1365 | d_connectionroom_sem->getValue( &room); | |
1366 | if(room<1) | |
e6a9dde5 | 1367 | g_log<<Logger::Warning<<"Limit of simultaneous TCP connections reached - raise max-tcp-connections"<<endl; |
4957a608 | 1368 | |
222efdc0 | 1369 | if(pthread_create(&tid, 0, &doConnection, reinterpret_cast<void*>(fd))) { |
e6a9dde5 | 1370 | g_log<<Logger::Error<<"Error creating thread: "<<stringerror()<<endl; |
4957a608 | 1371 | d_connectionroom_sem->post(); |
48e8d70b | 1372 | close(fd); |
cb0af1a1 | 1373 | decrementClientCount(remote); |
4957a608 BH |
1374 | } |
1375 | } | |
1376 | } | |
12c86877 BH |
1377 | } |
1378 | } | |
1379 | } | |
3f81d239 | 1380 | catch(PDNSException &AE) { |
e6a9dde5 | 1381 | g_log<<Logger::Error<<"TCP Nameserver thread dying because of fatal error: "<<AE.reason<<endl; |
12c86877 BH |
1382 | } |
1383 | catch(...) { | |
e6a9dde5 | 1384 | g_log<<Logger::Error<<"TCPNameserver dying because of an unexpected fatal error"<<endl; |
12c86877 | 1385 | } |
5bd2ea7b | 1386 | _exit(1); // take rest of server with us |
12c86877 BH |
1387 | } |
1388 | ||
1389 |