1 /*

2

3 Copyright (c) 2009 Carl Byington - 510 Software Group, released under

4 the GPL version 3 or any later version at your choice available at

5 http://www.gnu.org/licenses/gpl-3.0.txt

6

7 Based on a sample milter Copyright (c) 2000-2003 Sendmail, Inc. and its

8 suppliers. Inspired by the DCC by Rhyolite Software

9

10 -b port The port used to talk to the dcc interface daemon

11 -r port The port used to talk to our internal dns resolver processes

12 -p port The port through which the MTA will connect to this milter.

13 -t sec The timeout value.

14 -c Check the config, and print a copy to stdout. Don't start the

15 milter or do anything with the socket.

16 -s Stress test by loading and deleting the current config in a loop.

17 -d level set the debug level

18 -e f|t Print the results of looking up from address f and to address

19 t in the current config

20

21 */

22

23

24 // from sendmail sample

25 #include <sys/types.h>

26 #include <sys/stat.h>

27 #include <errno.h>

28 #include <sysexits.h>

29 #include <unistd.h>

30

31 // needed for socket io

32 #include <sys/ioctl.h>

33 #include <net/if.h>

34 #include <arpa/inet.h>

35 #include <netinet/in.h>

36 #include <netinet/tcp.h>

37 #include <netdb.h>

38 #include <sys/socket.h>

39 #include <sys/un.h>

40

41 // needed for thread

42 #include <pthread.h>

43

44 // needed for std c++ collections

45 #include <set>

46 #include <map>

47 #include <list>

48

49 // for the dns resolver

50 #include <netinet/in.h>

51 #include <arpa/nameser.h>

52 #include <resolv.h>

53

54 // misc stuff needed here

55 #include <ctype.h>

56 #include <syslog.h>

57 #include <pwd.h>

58 #include <sys/wait.h> /* header for waitpid() and various macros */

59 #include <signal.h> /* header for signal functions */

60

61 #include "includes.h"

62

63 #ifndef HAVE_DAEMON

64 #include "daemon.h"

65 #include "daemon.c"

66 #endif

67

68 #ifndef HAVE_MEMRCHR

69 void *memrchr(const void *a, int c, size_t len);

70 void *memrchr(const void *a, int c, size_t len) {

71 const unsigned char *p = (const unsigned char *)a;

72 for (p += len-1; (const void *)p >= a; p--)

73 if (*p == c)

74 return (void *)p;

75 return (void *)0;

76 }

77 #endif

78

79 extern "C" {

80 sfsistat mlfi_connect(SMFICTX *ctx, char *hostname, _SOCK_ADDR *hostaddr);

81 sfsistat mlfi_helo(SMFICTX * ctx, char *helohost);

82 sfsistat mlfi_envfrom(SMFICTX *ctx, char **argv);

83 sfsistat mlfi_envrcpt(SMFICTX *ctx, char **argv);

84 sfsistat mlfi_header(SMFICTX* ctx, char* headerf, char* headerv);

85 sfsistat mlfi_body(SMFICTX *ctx, u_char *data, size_t len);

86 sfsistat mlfi_eom(SMFICTX *ctx);

87 sfsistat mlfi_abort(SMFICTX *ctx);

88 sfsistat mlfi_close(SMFICTX *ctx);

89 void sig_chld(int signo);

90 }

91

92 int debug_syslog = 0;

93 bool syslog_opened = false;

94 bool use_syslog = true; // false to printf

95 bool loader_run = true; // used to stop the config loader thread

96 CONFIG *config = NULL; // protected by the config_mutex

97 int generation = 0; // protected by the config_mutex

98 const int maxlen = 1000; // used for snprintf buffers

99 regex_t srs_pattern; // used to detect srs coding in mail addresses

100

101 pthread_mutex_t config_mutex;

102 pthread_mutex_t syslog_mutex;

103 pthread_mutex_t resolve_mutex;

104 pthread_mutex_t fd_pool_mutex;

105 pthread_mutex_t rate_mutex;

106

107 std::set<int> fd_pool;

108 int NULL_SOCKET = -1;

109 const time_t ERROR_SOCKET_TIME = 60; // number of seconds between attempts to open a socket to the dns resolver process

110 const char *resolver_port = NULL; // unix domain socket to talk to the dns resolver process

111 int resolver_socket = NULL_SOCKET; // socket used to listen for resolver requests

112 const char *dccifd_port = NULL; // unix domain socket to talk to the dcc interface daemon

113 time_t last_error_time;

114 int resolver_sock_count = 0; // protected with fd_pool_mutex

115 int resolver_pool_size = 0; // protected with fd_pool_mutex

116 rcpt_rates rcpt_counts; // protected with rate_mutex

117

118

119 struct ns_map {

120 // all the strings are owned by the keys/values in the ns_host string map

121 string_map ns_host; // nameserver name -> host name that uses this name server

122 ns_mapper ns_ip; // nameserver name -> ip address of the name server

123 ~ns_map();

124 void add(const char *name, const char *refer);

125 };

126

127

128 ns_map::~ns_map() {

129 for (string_map::iterator i=ns_host.begin(); i!=ns_host.end(); i++) {

130 const char *x = (*i).first;

131 const char *y = (*i).second;

132 free((void*)x);

133 free((void*)y);

134 }

135 ns_ip.clear();

136 ns_host.clear();

137 }

138

139

140 void ns_map::add(const char *name, const char *refer) {

141 string_map::iterator i = ns_host.find(name);

142 if (i != ns_host.end()) return;

143 char *x = strdup(name);

144 char *y = strdup(refer);

145 ns_ip[x] = 0;

146 ns_host[x] = y;

147

148 }

149

150 // packed structure to allow a single socket write to dump the

151 // length and the following answer. The packing attribute is gcc specific.

152 struct glommer {

153 size_t length;

154 #ifdef NS_PACKETSZ

155 u_char answer[NS_PACKETSZ*4]; // with a resolver, we return resolver answers

156 #else

157 int answer; // without a resolver, we return a single ip4 address, 0 == no answer

158 #endif

159 } __attribute__ ((packed));

160

161

162 ////////////////////////////////////////////////

163 // helper to manipulate recipient counts

164 //

165 int incr_rcpt_count(const char *user);

166 int incr_rcpt_count(const char *user) {

167 pthread_mutex_lock(&rate_mutex);

168 rcpt_rates::iterator i = rcpt_counts.find(user);

169 int c = 1;

170 if (i == rcpt_counts.end()) {

171 user = strdup(user);

172 rcpt_counts[user] = c;

173 }

174 else {

175 c = ++((*i).second);

176 }

177 pthread_mutex_unlock(&rate_mutex);

178 return c;

179 }

180

181 ////////////////////////////////////////////////

182 // helper to discard the strings held by a context_map

183 //

184 void discard(context_map &cm);

185 void discard(context_map &cm) {

186 for (context_map::iterator i=cm.begin(); i!=cm.end(); i++) {

187 const char *x = (*i).first;

188 free((void*)x);

189 }

190 cm.clear();

191 }

192

193

194 ////////////////////////////////////////////////

195 // helper to register a string in a context_map

196 //

197 void register_string(context_map &cm, const char *name, CONTEXT *con);

198 void register_string(context_map &cm, const char *name, CONTEXT *con) {

199 context_map::iterator i = cm.find(name);

200 if (i != cm.end()) return;

201 char *x = strdup(name);

202 cm[x] = con;

203 }

204

205

206 ////////////////////////////////////////////////

207 // disconnect the fd from the dns resolver process

208 //

209 void my_disconnect(int sock, bool decrement = true);

210 void my_disconnect(int sock, bool decrement) {

211 if (sock != NULL_SOCKET) {

212 if (decrement) {

213 pthread_mutex_lock(&fd_pool_mutex);

214 resolver_sock_count--;

215 pthread_mutex_unlock(&fd_pool_mutex);

216 }

217 shutdown(sock, SHUT_RDWR);

218 close(sock);

219 }

220 }

221

222

223 ////////////////////////////////////////////////

224 // return fd connected to the dns resolver process

225 //

226 int my_connect();

227 int my_connect() {

228 // if we have had recent errors, don't even try to open the socket

229 if ((time(NULL) - last_error_time) < ERROR_SOCKET_TIME) return NULL_SOCKET;

230

231 // nothing recent, maybe this time it will work

232 int sock = NULL_SOCKET;

233 sockaddr_un server;

234 memset(&server, '\0', sizeof(server));

235 server.sun_family = AF_UNIX;

236 strncpy(server.sun_path, resolver_port, sizeof(server.sun_path)-1);

237 sock = socket(AF_UNIX, SOCK_STREAM, 0);

238 if (sock != NULL_SOCKET) {

239 bool rc = (connect(sock, (sockaddr *)&server, sizeof(server)) == 0);

240 if (!rc) {

241 my_disconnect(sock, false);

242 sock = NULL_SOCKET;

243 last_error_time = time(NULL);

244 }

245 }

246 else last_error_time = time(NULL);

247 if (sock != NULL_SOCKET) {

248 pthread_mutex_lock(&fd_pool_mutex);

249 resolver_sock_count++;

250 pthread_mutex_unlock(&fd_pool_mutex);

251 }

252 return sock;

253 }

254

255

256 ////////////////////////////////////////////////

257 // ask a dns question and get an A record answer - we don't try

258 // very hard, just using the default resolver retry settings.

259 // If we cannot get an answer, we just accept the mail.

260 //

261 //

262 int dns_interface(mlfiPriv &priv, const char *question, bool maybe_ip, ns_map *nameservers);

263 int dns_interface(mlfiPriv &priv, const char *question, bool maybe_ip, ns_map *nameservers) {

264 // tell sendmail we are still working

265 #if _FFR_SMFI_PROGRESS

266 if (priv.eom) smfi_progress(priv.ctx);

267 #endif

268

269 // this part can be done without locking the resolver mutex. Each

270 // milter thread is talking over its own socket to a separate resolver

271 // process, which does the actual dns resolution.

272 if (priv.err) return 0; // cannot ask more questions on this socket.

273 if (maybe_ip) {

274 // might be a bare ip address, try this first to avoid dns lookups that may not be needed

275 in_addr ip;

276 if (inet_aton(question, &ip)) {

277 return (int)ip.s_addr;

278 }

279 }

280 int n = strlen(question);

281 if (question[n-1] == '.') {

282 priv.my_write(question, n+1); // write the question including the null terminator

283 }

284 else {

285 priv.my_write(question, n); // write the question

286 priv.my_write(".", 2); // and the fully qualified . terminator and null string terminator

287 }

288 glommer glom;

289 char *buf = (char *)&glom;

290 priv.my_read(buf, sizeof(glom.length));

291 buf += sizeof(glom.length);

292 #ifdef RESOLVER_DEBUG

293 char text[1000];

294 snprintf(text, sizeof(text), "dns_interface() wrote question %s and has answer length %d", question, glom.length);

295 my_syslog(text);

296 #endif

297 if (glom.length == 0) return 0;

298 if (glom.length > sizeof(glom.answer)) {

299 priv.err = true;

300 return 0; // cannot process overlarge answers

301 }

302 priv.my_read(buf, glom.length);

303

304 #ifdef NS_PACKETSZ

305 // now we need to lock the resolver mutex to keep the milter threads from

306 // stepping on each other while parsing the dns answer.

307 int ret_address = 0;

308 pthread_mutex_lock(&resolve_mutex);

309 // parse the answer

310 ns_msg handle;

311 ns_rr rr;

312 if (ns_initparse(glom.answer, glom.length, &handle) == 0) {

313 // look for ns names

314 if (nameservers) {

315 ns_map &ns = *nameservers;

316 int rrnum = 0;

317 while (ns_parserr(&handle, ns_s_ns, rrnum++, &rr) == 0) {

318 if (ns_rr_type(rr) == ns_t_ns) {

319 char nam[NS_MAXDNAME+1];

320 char *n = nam;

321 const u_char *p = ns_rr_rdata(rr);

322 while (((n-nam) < NS_MAXDNAME) && ((size_t)(p-glom.answer) < glom.length) && *p) {

323 size_t s = *(p++);

324 if (s > 191) {

325 // compression pointer

326 s = (s-192)*256 + *(p++);

327 if (s >= glom.length) break; // pointer outside bounds of answer

328 p = glom.answer + s;

329 s = *(p++);

330 }

331 if (s > 0) {

332 if ((size_t)(n-nam) >= (NS_MAXDNAME-s)) break; // destination would overflow name buffer

333 if ((size_t)(p-glom.answer) >= (glom.length-s)) break; // source outside bounds of answer

334 memcpy(n, p, s);

335 n += s;

336 p += s;

337 *(n++) = '.';

338 }

339 }

340 if (n-nam) n--; // remove trailing .

341 *n = '\0'; // null terminate it

342 ns.add(nam, question); // ns host to lookup later

343 }

344 }

345 rrnum = 0;

346 while (ns_parserr(&handle, ns_s_ar, rrnum++, &rr) == 0) {

347 if (ns_rr_type(rr) == ns_t_a) {

348 char* nam = (char*)ns_rr_name(rr);

349 ns_mapper::iterator i = ns.ns_ip.find(nam);

350 if (i != ns.ns_ip.end()) {

351 // we want this ip address

352 int address;

353 memcpy(&address, ns_rr_rdata(rr), sizeof(address));

354 ns.ns_ip[nam] = address;

355 }

356 }

357 }

358 }

359 int rrnum = 0;

360 while (ns_parserr(&handle, ns_s_an, rrnum++, &rr) == 0) {

361 if (ns_rr_type(rr) == ns_t_a) {

362 int address;

363 memcpy(&address, ns_rr_rdata(rr), sizeof(address));

364 ret_address = address;

365 }

366 }

367 }

368 pthread_mutex_unlock(&resolve_mutex);

369 #ifdef RESOLVER_DEBUG

370 snprintf(text, sizeof(text), "dns_interface() found ip %d", ret_address);

371 my_syslog(text);

372 #endif

373 return ret_address;

374 #else

375 return glom.answer;

376 #endif

377 }

378

379

380 ////////////////////////////////////////////////

381 // lookup the domain name part of a hostname on the uribl

382 //

383 // if we find part of the hostname on the uribl, return

384 // true and point found to the part of the hostname that we found

385 // as a string registered in hosts.

386 // otherwise, return false and preserve the value of found.

387 //

388 bool uriblookup(mlfiPriv &priv, string_set &hosts, const char *hostname, const char *top, const char *&found) ;

389 bool uriblookup(mlfiPriv &priv, string_set &hosts, const char *hostname, const char *top, const char *&found) {

390 // top is pointer to '.' char at end of base domain, or null for ip address form

391 // so for hostname of www.fred.mydomain.co.uk

392 // top points to-----------------------^

393 // and we end up looking at only mydomain.co.uk, ignoring the www.fred stuff

394 char buf[maxlen];

395 if (top) {

396 // add one more component

397 const char *x = (const char *)memrchr(hostname, '.', top-hostname);

398 if (x) hostname = x+1;

399 }

400 snprintf(buf, sizeof(buf), "%s.%s.", hostname, priv.uribl_suffix);

401 if (dns_interface(priv, buf, false, NULL)) {

402 if (debug_syslog > 2) {

403 char tmp[maxlen];

404 snprintf(tmp, sizeof(tmp), "found %s on %s", hostname, priv.uribl_suffix);

405 my_syslog(tmp);

406 }

407 found = register_string(hosts, hostname);

408 return true;

409 }

410 return false;

411 }

412

413

414 ////////////////////////////////////////////////

415 // uribl checker

416 // -------------

417 // hostname MUST not have a trailing dot

418 // If tld, two level lookup.

419 // Else, look up three level domain.

420 //

421 // if we find part of the hostname on the uribl, return

422 // true and point found to the part of the hostname that we found

423 // as a string registered in hosts.

424 // otherwise, return false and preserve the value of found.

425 //

426 bool check_uribl(mlfiPriv &priv, string_set &hosts, const char *hostname, const char *&found) ;

427 bool check_uribl(mlfiPriv &priv, string_set &hosts, const char *hostname, const char *&found) {

428 in_addr ip;

429 if (inet_aton(hostname, &ip)) {

430 const u_char *src = (const u_char *)&ip.s_addr;

431 if (src[0] == 127) return false; // don't do dns lookups on localhost

432 if (src[0] == 10) return false; // don't do dns lookups on rfc1918 space

433 if ((src[0] == 192) && (src[1] == 168)) return false;

434 if ((src[0] == 172) && (16 <= src[1]) && (src[1] <= 31)) return false;

435 char adr[sizeof "255.255.255.255 "];

436 snprintf(adr, sizeof(adr), "%u.%u.%u.%u", src[3], src[2], src[1], src[0]);

437 // cannot use inet_ntop here since we want the octets reversed.

438 return (uriblookup(priv, hosts, adr, NULL, found));

439 }

440

441 const char *top, *top2, *top3;

442 top = strrchr(hostname, '.');

443 if (top) {

444 top2 = (const char *)memrchr(hostname, '.', top-hostname);

445

446 if (top2) {

447 string_set::iterator i = priv.memory->get_cctlds()->find(top2+1);

448 string_set::iterator x = priv.memory->get_cctlds()->end();

449 // if we have a 2-level-cctld, just look at top three levels of the name

450 if (i != x) return uriblookup(priv, hosts, hostname, top2, found);

451

452 // if we have more than 3 levels in the name, look at the top three levels of the name

453 top3 = (const char *)memrchr(hostname, '.', top2-hostname);

454 if (top3 && uriblookup(priv, hosts, hostname, top2, found)) return true;

455

456 // if that was not found, fall thru to looking at the top two levels

457 }

458 // look at the top two levels of the name

459 return uriblookup(priv, hosts, hostname, top, found);

460 }

461 return false;

462 }

463

464

465 mlfiPriv::mlfiPriv() {

466 pthread_mutex_lock(&config_mutex);

467 pc = config;

468 pc->reference_count++;

469 pthread_mutex_unlock(&config_mutex);

470 get_fd();

471 ctx = NULL;

472 eom = false;

473 ip = 0;

474 helo = NULL;

475 mailaddr = NULL;

476 queueid = NULL;

477 authenticated = NULL;

478 client_name = NULL;

479 host_uribl = NULL;

480 helo_uribl = false;

481 client_uribl = false;

482 from_uribl = false;

483 have_whites = false;

484 only_whites = true;

485 want_spamassassin = false;

486 want_dccgrey = false;

487 want_dccbulk = false;

488 allow_autowhitelisting = true;

489 content_context = NULL;

490 memory = NULL;

491 scanner = NULL;

492 content_suffix = NULL;

493 content_message = NULL;

494 uribl_suffix = NULL;

495 uribl_message = NULL;

496 content_host_ignore = NULL;

497 assassin = NULL;

498 dccifd = NULL;

499 }

500

501 mlfiPriv::~mlfiPriv() {

502 return_fd();

503 pthread_mutex_lock(&config_mutex);

504 pc->reference_count--;

505 bool last = (!pc->reference_count) && (pc != config);

506 pthread_mutex_unlock(&config_mutex);

507 if (last) delete pc; // free this config, since we were the last reference to it

508 if (helo) free((void*)helo);

509 reset(true);

510 }

511

512 void mlfiPriv::reset(bool final) {

513 while (!delayer.empty()) {

514 DELAYWHITEP dwp = delayer.front();

515 const char *loto = dwp->get_loto();

516 if (loto) free((void*)loto);

517 delete dwp;

518 delayer.pop_front();

519 }

520 if (mailaddr) free((void*)mailaddr);

521 if (queueid) free((void*)queueid);

522 if (authenticated) free((void*)authenticated);

523 if (client_name) free((void*)client_name);

524 discard(hosts_uribl);

525 delayer.clear();

526 discard(env_to);

527 if (memory) delete memory;

528 if (scanner) delete scanner;

529 if (assassin) delete assassin;

530 if (dccifd) delete dccifd;

531 if (!final) {

532 ctx = NULL;

533 eom = false;

534 mailaddr = NULL;

535 queueid = NULL;

536 authenticated = NULL;

537 client_name = NULL;

538 host_uribl = NULL;

539 helo_uribl = false;

540 client_uribl = false;

541 from_uribl = false;

542 have_whites = false;

543 only_whites = true;

544 want_spamassassin = false;

545 want_dccgrey = false;

546 want_dccbulk = false;

547 allow_autowhitelisting = true;

548 content_context = NULL;

549 memory = NULL;

550 scanner = NULL;

551 content_suffix = NULL;

552 content_message = NULL;

553 uribl_suffix = NULL;

554 uribl_message = NULL;

555 content_host_ignore = NULL;

556 assassin = NULL;

557 dccifd = NULL;

558 }

559 }

560

561 void mlfiPriv::get_fd() {

562 err = true;

563 fd = NULL_SOCKET;

564 int result = pthread_mutex_lock(&fd_pool_mutex);

565 if (!result) {

566 std::set<int>::iterator i;

567 i = fd_pool.begin();

568 if (i != fd_pool.end()) {

569 // have at least one fd in the pool

570 err = false;

571 fd = *i;

572 fd_pool.erase(fd);

573 resolver_pool_size--;

574 pthread_mutex_unlock(&fd_pool_mutex);

575 }

576 else {

577 // pool is empty, get a new fd

578 pthread_mutex_unlock(&fd_pool_mutex);

579 fd = my_connect();

580 err = (fd == NULL_SOCKET);

581 }

582 }

583 else {

584 // cannot lock the pool, just get a new fd

585 fd = my_connect();

586 err = (fd == NULL_SOCKET);

587 }

588 }

589

590 void mlfiPriv::return_fd() {

591 if (err) {

592 // this fd got a socket error, so close it, rather than returning it to the pool

593 my_disconnect(fd);

594 }

595 else {

596 int result = pthread_mutex_lock(&fd_pool_mutex);

597 if (!result) {

598 if ((resolver_sock_count > resolver_pool_size*5) || (resolver_pool_size < 5)) {

599 // return the fd to the pool

600 fd_pool.insert(fd);

601 resolver_pool_size++;

602 pthread_mutex_unlock(&fd_pool_mutex);

603 }

604 else {

605 // more than 20% of the open resolver sockets are in the pool, and the

606 // pool as at least 5 sockets. that is enough, so just close this one.

607 pthread_mutex_unlock(&fd_pool_mutex);

608 my_disconnect(fd);

609 }

610 }

611 else {

612 // could not lock the pool, so just close the fd

613 my_disconnect(fd);

614 }

615 }

616 }

617

618 size_t mlfiPriv::my_write(const char *buf, size_t len) {

619 if (err) return 0;

620 size_t rs = 0;

621 while (len) {

622 size_t ws = write(fd, buf, len);

623 if (ws > 0) {

624 rs += ws;

625 len -= ws;

626 buf += ws;

627 }

628 else {

629 // peer closed the socket!

630 rs = 0;

631 err = true;

632 break;

633 }

634 }

635 return rs;

636 }

637

638 size_t mlfiPriv::my_read(char *buf, size_t len) {

639 if (err) return 0;

640 size_t rs = 0;

641 while (len) {

642 size_t ws = read(fd, buf, len);

643 if (ws > 0) {

644 rs += ws;

645 len -= ws;

646 buf += ws;

647 }

648 else {

649 // peer closed the socket!

650 rs = 0;

651 err = true;

652 break;

653 }

654 }

655 return rs;

656 }

657

658 void mlfiPriv::need_content_filter(const char *rcpt, CONTEXT &con) {

659 if (!memory) {

660 // first recipient that needs content filtering sets

661 // some of the content filtering parameters

662 memory = new recorder(this, con.get_html_tags(), con.get_content_tlds(), con.get_content_cctlds());

663 scanner = new url_scanner(memory);

664 content_suffix = con.get_content_suffix();

665 content_message = con.get_content_message();

666 uribl_suffix = con.get_uribl_suffix();

667 uribl_message = con.get_uribl_message();

668 content_host_ignore = &con.get_content_host_ignore();

669 // if we are using uribl, test helo and client names here

670 if (uribl_suffix) {

671 if (helo) {

672 helo_uribl = check_uribl(*this, hosts_uribl, helo, host_uribl);

673 if (client_name && !helo_uribl) {

674 client_uribl = check_uribl(*this, hosts_uribl, client_name, host_uribl);

675 if (mailaddr && !client_uribl) {

676 char *f = strchr(mailaddr, '@');

677 if (f) from_uribl = check_uribl(*this, hosts_uribl, f+1, host_uribl);

678 }

679 }

680 }

681 }

682 }

683 }

684

685

686 mlfiPriv* fetch_priv_from_ctx(SMFICTX *ctx);

687 mlfiPriv* fetch_priv_from_ctx(SMFICTX *ctx)

688 {

689 mlfiPriv *priv = (struct mlfiPriv *)smfi_getpriv(ctx);

690 priv->ctx = ctx;

691 return priv;

692 }

693 #define MLFIPRIV fetch_priv_from_ctx(ctx)

694

695

696

697 ////////////////////////////////////////////////

698 // syslog a message

699 //

700 void my_syslog(mlfiPriv *priv, const char *text) {

701 char buf[maxlen];

702 if (priv) {

703 snprintf(buf, sizeof(buf), "%s: %s", priv->queueid, text);

704 text = buf;

705 }

706 if (use_syslog) {

707 pthread_mutex_lock(&syslog_mutex);

708 if (!syslog_opened) {

709 openlog("dnsbl", LOG_PID, LOG_MAIL);

710 syslog_opened = true;

711 }

712 syslog(LOG_NOTICE, "%s", text);

713 pthread_mutex_unlock(&syslog_mutex);

714 }

715 else {

716 printf("%s \n", text);

717 }

718 }

719

720 void my_syslog(mlfiPriv *priv, const string text) {

721 if (debug_syslog > 3) {

722 char buf[maxlen];

723 strncpy(buf, text.c_str(), sizeof(buf));

724 buf[maxlen-1] = '\0'; // ensure null termination

725 my_syslog(priv, buf);

726 }

727 }

728

729 void my_syslog(const char *text) {

730 my_syslog(NULL, text);

731 }

732

733

734 ////////////////////////////////////////////////

735 // read a resolver request from the socket, process it, and

736 // write the result back to the socket.

737

738 void process_resolver_requests(int socket);

739 void process_resolver_requests(int socket) {

740 #ifdef NS_MAXDNAME

741 char question[NS_MAXDNAME];

742 #else

743 char question[1000];

744 #endif

745 glommer glom;

746

747 int maxq = sizeof(question);

748 while (true) {

749 // read a question

750 int rs = 0;

751 while (rs < maxq) {

752 int ns = read(socket, question+rs, maxq-rs);

753 if (ns > 0) {

754 rs += ns;

755 if (question[rs-1] == '\0') {

756 // last byte read was the null terminator, we are done

757 break;

758 }

759 }

760 else {

761 // peer closed the socket

762 #ifdef RESOLVER_DEBUG

763 my_syslog("process_resolver_requests() peer closed socket while reading question");

764 #endif

765 shutdown(socket, SHUT_RDWR);

766 close(socket);

767 return;

768 }

769 }

770 question[rs-1] = '\0'; // ensure null termination

771

772 // find the answer

773 #ifdef NS_PACKETSZ

774 #ifdef RESOLVER_DEBUG

775 char text[1000];

776 snprintf(text, sizeof(text), "process_resolver_requests() has a question %s", question);

777 my_syslog(text);

778 #endif

779 int res_result = res_search(question, ns_c_in, ns_t_a, glom.answer, sizeof(glom.answer));

780 if (res_result < 0) glom.length = 0; // represent all errors as zero length answers

781 else glom.length = (size_t)res_result;

782 #else

783 glom.length = sizeof(glom.answer);

784 glom.answer = 0;

785 struct hostent *host = gethostbyname(question);

786 if (host && (host->h_addrtype == AF_INET)) {

787 memcpy(&glom.answer, host->h_addr, sizeof(glom.answer));

788 }

789 #endif

790

791 // write the answer

792 char *buf = (char *)&glom;

793 int len = glom.length + sizeof(glom.length);

794 #ifdef RESOLVER_DEBUG

795 snprintf(text, sizeof(text), "process_resolver_requests() writing answer length %d for total %d", glom.length, len);

796 my_syslog(text);

797 #endif

798 int ws = 0;

799 while (len > ws) {

800 int ns = write(socket, buf+ws, len-ws);

801 if (ns > 0) {

802 ws += ns;

803 }

804 else {

805 // peer closed the socket!

806 #ifdef RESOLVER_DEBUG

807 my_syslog("process_resolver_requests() peer closed socket while writing answer");

808 #endif

809 shutdown(socket, SHUT_RDWR);

810 close(socket);

811 return;

812 }

813 }

814 }

815 }

816

817

818 ////////////////////////////////////////////////

819 // check a single dnsbl

820 //

821 bool check_single(mlfiPriv &priv, int ip, const char *suffix);

822 bool check_single(mlfiPriv &priv, int ip, const char *suffix) {

823 // make a dns question

824 const u_char *src = (const u_char *)&ip;

825 if (src[0] == 127) return false; // don't do dns lookups on localhost

826 if (src[0] == 10) return false; // don't do dns lookups on rfc1918 space

827 if ((src[0] == 192) && (src[1] == 168)) return false;

828 if ((src[0] == 172) && (16 <= src[1]) && (src[1] <= 31)) return false;

829 #ifdef NS_MAXDNAME

830 char question[NS_MAXDNAME];

831 #else

832 char question[1000];

833 #endif

834 snprintf(question, sizeof(question), "%u.%u.%u.%u.%s.", src[3], src[2], src[1], src[0], suffix);

835 // ask the question, if we get an A record it implies a blacklisted ip address

836 return dns_interface(priv, question, false, NULL);

837 }

838

839

840 ////////////////////////////////////////////////

841 // check a single dnsbl

842 //

843 bool check_single(mlfiPriv &priv, int ip, DNSBL &bl);

844 bool check_single(mlfiPriv &priv, int ip, DNSBL &bl) {

845 return check_single(priv, ip, bl.suffix);

846 }

847

848

849 ////////////////////////////////////////////////

850 // check the dnsbls specified for this recipient

851 //

852 bool check_dnsbl(mlfiPriv &priv, dnsblp_list &dnsbll, DNSBLP &rejectlist);

853 bool check_dnsbl(mlfiPriv &priv, dnsblp_list &dnsbll, DNSBLP &rejectlist) {

854 for (dnsblp_list::iterator i=dnsbll.begin(); i!=dnsbll.end(); i++) {

855 DNSBLP dp = *i; // non null by construction

856 bool st;

857 map<DNSBLP, bool>::iterator f = priv.checked.find(dp);

858 if (f == priv.checked.end()) {

859 // have not checked this list yet

860 st = check_single(priv, priv.ip, *dp);

861 rejectlist = dp;

862 priv.checked[dp] = st;

863 }

864 else {

865 st = (*f).second;

866 rejectlist = (*f).first;

867 }

868 if (st) return st;

869 }

870 return false;

871 }

872

873

874 ////////////////////////////////////////////////

875 // check the hosts from the body against the content filter and uribl dnsbls

876 //

877 //

878 bool check_hosts(mlfiPriv &priv, bool random, int limit, const char *&msg, const char *&host, int &ip, const char *&found);

879 bool check_hosts(mlfiPriv &priv, bool random, int limit, const char *&msg, const char *&host, int &ip, const char *&found) {

880 found = NULL; // normally ip address style

881 if (!priv.content_suffix && !priv.uribl_suffix) return false; // nothing to check

882 string_set &hosts = priv.memory->get_hosts();

883 string_set &ignore = *priv.content_host_ignore;

884

885 int count = 0;

886 int cnt = hosts.size(); // number of hosts we could look at

887 int_set ips;

888 ns_map nameservers;

889 for (string_set::iterator i=hosts.begin(); i!=hosts.end(); i++) {

890 host = *i; // a reference into hosts, which will live until this smtp transaction is closed

891

892 // don't bother looking up hosts on the ignore list

893 string_set::iterator j = ignore.find(host);

894 if (j != ignore.end()) continue;

895

896 // try to only look at limit/cnt fraction of the available cnt host names in random mode

897 if ((cnt > limit) && (limit > 0) && random) {

898 int r = rand() % cnt;

899 if (r >= limit) {

900 if (debug_syslog > 2) {

901 char buf[maxlen];

902 snprintf(buf, sizeof(buf), "host %s skipped", host);

903 my_syslog(&priv, buf);

904 }

905 continue;

906 }

907 }

908 count++;

909 ip = dns_interface(priv, host, true, &nameservers);

910 if (debug_syslog > 2) {

911 char buf[maxlen];

912 if (ip) {

913 char adr[sizeof "255.255.255.255 "];

914 adr[0] = '\0';

915 inet_ntop(AF_INET, (const u_char *)&ip, adr, sizeof(adr));

916 snprintf(buf, sizeof(buf), "host %s found at %s", host, adr);

917 }

918 else {

919 snprintf(buf, sizeof(buf), "host %s not found", host);

920 }

921 my_syslog(&priv, buf);

922 }

923 if (ip) {

924 int_set::iterator i = ips.find(ip);

925 if (i == ips.end()) {

926 // we haven't looked this up yet

927 ips.insert(ip);

928 // check dnsbl style list

929 if (priv.content_suffix && check_single(priv, ip, priv.content_suffix)) {

930 msg = priv.content_message;

931 return true;

932 }

933 // Check uribl & surbl style list

934 if (priv.uribl_suffix && check_uribl(priv, hosts, host, found)) {

935 msg = priv.uribl_message;

936 return true;

937 }

938 }

939 }

940 }

941 limit *= 4; // allow average of 3 ns per host name

942 for (ns_mapper::iterator i=nameservers.ns_ip.begin(); i!=nameservers.ns_ip.end(); i++) {

943 count++;

944 if ((count > limit) && (limit > 0)) return false; // too many name servers to check them all

945 host = (*i).first; // a transient reference that needs to be replaced before we return it

946 ip = (*i).second;

947 if (!ip) ip = dns_interface(priv, host, false, NULL);

948 if (debug_syslog > 2) {

949 char buf[maxlen];

950 if (ip) {

951 char adr[sizeof "255.255.255.255 "];

952 adr[0] = '\0';

953 inet_ntop(AF_INET, (const u_char *)&ip, adr, sizeof(adr));

954 snprintf(buf, sizeof(buf), "ns %s found at %s", host, adr);

955 }

956 else {

957 snprintf(buf, sizeof(buf), "ns %s not found", host);

958 }

959 my_syslog(&priv, buf);

960 }

961 if (ip) {

962 int_set::iterator i = ips.find(ip);

963 if (i == ips.end()) {

964 ips.insert(ip);

965 if (check_single(priv, ip, priv.content_suffix)) {

966 msg = priv.content_message;

967 string_map::iterator j = nameservers.ns_host.find(host);

968 if (j != nameservers.ns_host.end()) {

969 const char *refer = (*j).second;

970 char buf[maxlen];

971 snprintf(buf, sizeof(buf), "%s with nameserver %s", refer, host);

972 host = register_string(hosts, buf); // put a copy into hosts, and return that reference

973 }

974 else {

975 host = register_string(hosts, host); // put a copy into hosts, and return that reference

976 }

977 return true;

978 }

979 }

980 }

981 }

982 return false;

983 }

984

985

986 ////////////////////////////////////////////////

987 //

988 // this email address is passed in from sendmail, and will normally be

989 // enclosed in <>. I think older versions of sendmail supplied the <>

990 // wrapper if the mail client did not, but the current version does not do

991 // that. So the <> wrapper is now optional. It may have mixed case, just

992 // as the mail client sent it. We dup the string and convert the duplicate

993 // to lower case. Some clients enclose the entire address in single quotes,

994 // so we strip those as well. We also remove the SRS coding.

995 //

996 const char *to_lower_string(const char *email);

997 const char *to_lower_string(const char *email) {

998 int n = strlen(email);

999 if (email[0] == '<') {

1000 // assume it also ends with >

1001 n -= 2;

1002 if (n < 1) return strdup(email); // return "<>"

1003 email++;

1004 }

1005 if ((email[0] == '\'') && (email[n-1] == '\'') && (n > 2)) {

1006 n -= 2;

1007 email++;

1008 }

1009 char *key = strdup(email);

1010 key[n] = '\0';

1011 for (int i=0; i<n; i++) key[i] = tolower(key[i]);

1012 if ((n > 12) && (strncmp(key, "srs", 3) == 0)) {

1013 // might have srs coding to be removed

1014 const int nmatch = 6;

1015 regmatch_t match[nmatch];

1016 if (0 == regexec(&srs_pattern, key, nmatch, match, 0)) {

1017 int s4 = match[4].rm_so; // domain

1018 int e4 = match[4].rm_eo;

1019 int s5 = match[5].rm_so; // user

1020 int e5 = match[5].rm_eo;

1021 if ((s4 != -1) && (s5 != -1)) {

1022 char *newkey = strdup(key); // large enough

1023 key[e4] = '\0';

1024 key[e5] = '\0';

1025 strcpy(newkey, key+s5); // user

1026 strcat(newkey, "@"); // @

1027 strcat(newkey, key+s4); // domain

1028 free(key);

1029 key = newkey;

1030 }

1031 }

1032 }

1033 return key;

1034 }

1035

1036

1037 ////////////////////////////////////////////////

1038 // start of sendmail milter interfaces

1039 //

1040 sfsistat mlfi_connect(SMFICTX *ctx, char *hostname, _SOCK_ADDR *hostaddr)

1041 {

1042 // allocate some private memory

1043 mlfiPriv *priv = new mlfiPriv;

1044 if (hostaddr->sa_family == AF_INET) {

1045 priv->ip = ((struct sockaddr_in *)hostaddr)->sin_addr.s_addr;

1046 }

1047 // save the private data

1048 smfi_setpriv(ctx, (void*)priv);

1049

1050 // continue processing

1051 return SMFIS_CONTINUE;

1052 }

1053

1054 sfsistat mlfi_helo(SMFICTX * ctx, char *helohost)

1055 {

1056 mlfiPriv &priv = *MLFIPRIV;

1057 priv.helo = strdup(helohost);

1058 return SMFIS_CONTINUE;

1059 }

1060

1061 sfsistat mlfi_envfrom(SMFICTX *ctx, char **from)

1062 {

1063 mlfiPriv &priv = *MLFIPRIV;

1064 priv.mailaddr = to_lower_string(from[0]);

1065 priv.queueid = strdup(smfi_getsymval(ctx, (char*)"i"));

1066 priv.authenticated = smfi_getsymval(ctx, (char*)"{auth_authen}");

1067 priv.client_name = smfi_getsymval(ctx, (char*)"_");

1068 if (!priv.helo) priv.helo = strdup("unknown");

1069 if (priv.authenticated) priv.authenticated = strdup(priv.authenticated);

1070 if (priv.client_name) priv.client_name = strdup(priv.client_name);

1071 if (spamc != spamc_empty) {

1072 priv.assassin = new SpamAssassin(&priv, priv.ip, priv.helo, priv.mailaddr, priv.queueid);

1073 }

1074 if (dccifd_port) {

1075 priv.dccifd = new DccInterface(dccifd_port, &priv, priv.ip, priv.helo, priv.mailaddr);

1076 }

1077 return SMFIS_CONTINUE;

1078 }

1079

1080 sfsistat mlfi_envrcpt(SMFICTX *ctx, char **rcpt)

1081 {

1082 DNSBLP rejectlist = NULL; // list that caused the reject

1083 mlfiPriv &priv = *MLFIPRIV;

1084 CONFIG &dc = *priv.pc;

1085 const char *rcptaddr = rcpt[0];

1086 const char *loto = to_lower_string(rcptaddr);

1087 bool self = (strcmp(loto, priv.mailaddr) == 0);

1088

1089 // some version of sendmail allowed rcpt to:<> and passed it thru to the milters

1090 if (strcmp(loto, "<>") == 0) {

1091 smfi_setreply(ctx, (char*)"550", (char*)"5.7.1", (char*)"bogus recipient");

1092 return SMFIS_REJECT;

1093 }

1094 // priv.mailaddr sending original message to loto

1095 CONTEXT &con = *(dc.find_context(loto)->find_context(priv.mailaddr));

1096 VERIFYP ver = con.find_verify(loto);

1097 const char *fromvalue = con.find_from(priv.mailaddr, true, priv.queueid);

1098 // tell spam assassin and dccifd about this recipient

1099 if (priv.assassin) priv.assassin->mlfi_envrcpt(ctx, loto);

1100 if (priv.dccifd) priv.dccifd->mlfi_envrcpt(ctx, loto, con.get_grey() && !priv.authenticated);

1101 // loto sending a reply back to priv.mailaddr

1102 CONTEXT &con2 = *(dc.find_context(priv.mailaddr)->find_context(loto));

1103 const char *replyvalue = con2.find_from(loto);

1104 if (debug_syslog > 1) {

1105 char buf[maxlen];

1106 char msg[maxlen];

1107 snprintf(msg, sizeof(msg), "from <%s> to <%s> using context %s state %s reply state %s", priv.mailaddr, loto, con.get_full_name(buf,maxlen), fromvalue, replyvalue);

1108 my_syslog(&priv, msg);

1109 }

1110 free((void*)loto);

1111 status st;

1112 if (replyvalue == token_black) {

1113 smfi_setreply(ctx, (char*)"550", (char*)"5.7.1", (char*)"recipient can not reply due to blacklisting");

1114 return SMFIS_REJECT;

1115 }

1116 if (priv.authenticated) {

1117 int c = incr_rcpt_count(priv.authenticated);

1118 int l = dc.default_context->find_rate(priv.authenticated);

1119 if (debug_syslog > 1) {

1120 char msg[maxlen];

1121 snprintf(msg, sizeof(msg), "authenticated id %s (%d recipients, %d limit)", priv.authenticated, c, l);

1122 my_syslog(&priv, msg);

1123 }

1124 if (c > l) {

1125 smfi_setreply(ctx, (char*)"550", (char*)"5.7.1", (char*)"recipient rate limit exceeded");

1126 return SMFIS_REJECT;

1127 }

1128 st = white;

1129 }

1130 else if (fromvalue == token_black) {

1131 st = black;

1132 }

1133 else if ((fromvalue == token_white) && !self) {

1134 st = white;

1135 }

1136 else {

1137 // check the dns based lists

1138 st = (check_dnsbl(priv, con.get_dnsbl_list(), rejectlist)) ? reject : oksofar;

1139 }

1140 if (st == reject) {

1141 // reject the recipient based on some dnsbl

1142 char adr[sizeof "255.255.255.255 "];

1143 adr[0] = '\0';

1144 inet_ntop(AF_INET, (const u_char *)&priv.ip, adr, sizeof(adr));

1145 char buf[maxlen];

1146 snprintf(buf, sizeof(buf), rejectlist->message, adr, adr);

1147 smfi_setreply(ctx, (char*)"550", (char*)"5.7.1", buf);

1148 return SMFIS_REJECT;

1149 }

1150 if (st == oksofar) {

1151 const char *msg = con.generic_match(priv.client_name);

1152 if (msg) {

1153 // reject the recipient based on generic reverse dns

1154 char buf[maxlen];

1155 snprintf(buf, sizeof(buf), msg, priv.client_name);

1156 smfi_setreply(ctx, (char*)"550", (char*)"5.7.1", buf);

1157 return SMFIS_REJECT;

1158 }

1159 }

1160 if (st == black) {

1161 // reject the recipient based on blacklisting either from or to

1162 smfi_setreply(ctx, (char*)"550", (char*)"5.7.1", (char*)"no such user");

1163 return SMFIS_REJECT;

1164 }

1165 if (ver) {

1166 // try to verify this from/to pair of addresses even if it might be explicitly whitelisted

1167 const char *loto = to_lower_string(rcptaddr);

1168 bool rc = ver->ok(priv.mailaddr, loto);

1169 free((void*)loto);

1170 if (!rc) {

1171 smfi_setreply(ctx, (char*)"550", (char*)"5.7.1", (char*)"no such user");

1172 return SMFIS_REJECT;

1173 }

1174 }

1175 // we will accept the recipient, but add an auto-whitelist entry

1176 // if needed to ensure we can accept replies

1177 loto = to_lower_string(rcptaddr);

1178 WHITELISTERP w = con2.find_autowhite(loto, priv.mailaddr);

1179 // check if local part is too big

1180 const int max_local_size = 30;

1181 const char *p = strchr(loto, '@');

1182 int len = (p) ? p-loto : max_local_size;

1183 if (len >= max_local_size) w = NULL; // too big, pretend we don't have a whitelister

1184 // record it if we have a whitelister

1185 if (w) {

1186 DELAYWHITEP dwp = new DELAYWHITE(loto, w, &con2); // dwp takes ownership of the string

1187 priv.delayer.push_back(dwp);

1188 }

1189 else {

1190 free((void*)loto); // or we free it here

1191 }

1192

1193 // accept the recipient

1194 if (!con.get_content_filtering()) st = white;

1195

1196 if (st == oksofar) {

1197 // remember first content filtering context

1198 if (con.get_content_filtering()) {

1199 if (!priv.content_context) priv.content_context = &con;

1200 else if (con.get_require() && (priv.content_context != &con)) {

1201 smfi_setreply(ctx, (char*)"452", (char*)"4.2.1", (char*)"incompatible filtering contexts");

1202 return SMFIS_TEMPFAIL;

1203 }

1204 priv.need_content_filter(rcptaddr, con);

1205 char bu[maxlen];

1206 bool uri = false;

1207 // content filtering implies also checking helo name on uribl (if enabled)

1208 if (priv.helo_uribl) {

1209 snprintf(bu, sizeof(bu), "(helo %s)", priv.host_uribl);

1210 uri = true;

1211 }

1212 // content filterint implies also checking client reverse dns name on uribl (if enabled)

1213 if (priv.client_uribl) {

1214 snprintf(bu, sizeof(bu), "(rdns %s)", priv.host_uribl);

1215 uri = true;

1216 }

1217 // content filterint implies also checking mail from domain name on uribl (if enabled)

1218 if (priv.from_uribl) {

1219 snprintf(bu, sizeof(bu), "(from %s)", priv.host_uribl);

1220 uri = true;

1221 }

1222 if (uri) {

1223 char buf[maxlen];

1224 snprintf(buf, sizeof(buf), priv.uribl_message, bu, priv.host_uribl);

1225 smfi_setreply(ctx, (char*)"550", (char*)"5.7.1", buf);

1226 return SMFIS_REJECT;

1227 }

1228 }

1229 // remember the non-whites

1230 register_string(priv.env_to, rcptaddr, &con);

1231 priv.only_whites = false;

1232 priv.want_spamassassin |= (priv.assassin) && // have spam assassin available and

1233 (con.get_spamassassin_limit() != 0); // want to use it with a non-zero score

1234 priv.want_dccgrey |= (priv.dccifd) && // have dcc interface and

1235 (con.get_grey()); // want to use it for greylisting

1236 priv.want_dccbulk |= (priv.dccifd) && // have dcc interface and

1237 (con.get_bulk() != 0); // want to use it for bulk detection

1238 }

1239 if (st == white) {

1240 priv.have_whites = true;

1241 }

1242 return SMFIS_CONTINUE;

1243 }

1244

1245 sfsistat mlfi_header(SMFICTX* ctx, char* headerf, char* headerv)

1246 {

1247 mlfiPriv &priv = *MLFIPRIV;

1248 // headers that avoid autowhitelisting

1249 if (((strcasecmp(headerf, "precedence") == 0) && (strcasecmp(headerv, "bulk") == 0)) ||

1250 ((strcasecmp(headerf, "content-type") == 0) && (strncasecmp(headerv, "multipart/report", 16) == 0))) {

1251 priv.allow_autowhitelisting = false;

1252 }

1253

1254 // other headers are only needed for content filtering

1255 if (priv.authenticated) return SMFIS_CONTINUE;

1256 if (priv.only_whites) return SMFIS_CONTINUE;

1257 if (priv.want_spamassassin) priv.assassin->mlfi_header(headerf, headerv);

1258 if (priv.want_dccgrey || priv.want_dccbulk) priv.dccifd->mlfi_header(ctx, headerf, headerv);

1259 return SMFIS_CONTINUE;

1260 }

1261

1262 sfsistat mlfi_eoh(SMFICTX* ctx)

1263 {

1264 mlfiPriv &priv = *MLFIPRIV;

1265 // delayed autowhitelisting

1266 while (!priv.delayer.empty()) {

1267 DELAYWHITEP dwp = priv.delayer.front();

1268 const char *loto = dwp->get_loto();

1269 if (priv.allow_autowhitelisting) {

1270 WHITELISTERP w = dwp->get_w();

1271 CONTEXTP con2 = dwp->get_con();

1272 if (debug_syslog > 1) {

1273 char buf[maxlen];

1274 char msg[maxlen];

1275 snprintf(msg, sizeof(msg), "whitelist reply from <%s> in context %s", loto, con2->get_full_name(buf,maxlen));

1276 my_syslog(&priv, msg);

1277 }

1278 w->sent(loto); // don't free it, the whitelister takes ownership of the string

1279 }

1280 else {

1281 if (debug_syslog > 1) {

1282 char msg[maxlen];

1283 snprintf(msg, sizeof(msg), "avoid whitelist reply from <%s> for outgoing auto-responder", loto);

1284 my_syslog(&priv, msg);

1285 }

1286 if (loto) free((void*)loto);// or we free it here

1287 }

1288 delete dwp;

1289 priv.delayer.pop_front();

1290 }

1291 // content filtering

1292 if (priv.authenticated) return SMFIS_CONTINUE;

1293 if (priv.only_whites) return SMFIS_CONTINUE;

1294 if (priv.want_spamassassin) priv.assassin->mlfi_eoh();

1295 if (priv.want_dccgrey || priv.want_dccbulk) priv.dccifd->mlfi_eoh();

1296 return SMFIS_CONTINUE;

1297 }

1298

1299 sfsistat mlfi_body(SMFICTX *ctx, u_char *data, size_t len)

1300 {

1301 mlfiPriv &priv = *MLFIPRIV;

1302 if (priv.authenticated) return SMFIS_CONTINUE;

1303 if (priv.only_whites) return SMFIS_CONTINUE;

1304 if (priv.want_spamassassin) priv.assassin->mlfi_body(data, len);

1305 if (priv.want_dccgrey || priv.want_dccbulk) priv.dccifd->mlfi_body(data, len);

1306 priv.scanner->scan(data, len);

1307 return SMFIS_CONTINUE;

1308 }

1309

1310 sfsistat mlfi_eom(SMFICTX *ctx)

1311 {

1312 sfsistat rc;

1313 mlfiPriv &priv = *MLFIPRIV;

1314 const char *host = NULL;

1315 int ip;

1316 // process end of message

1317 priv.eom = true;

1318 if (priv.authenticated || priv.only_whites) rc = SMFIS_CONTINUE;

1319 else {

1320 // assert env_to not empty, it contains the

1321 // non-whitelisted folks that want content filtering

1322 int score = (priv.want_spamassassin) ? priv.assassin->mlfi_eom() : 0;

1323 bool grey = false;

1324 int bulk = 0;

1325 if (priv.want_dccgrey || priv.want_dccbulk) priv.dccifd->mlfi_eom(grey, bulk);

1326

1327 char buf[maxlen];

1328 string msg;

1329 string_set alive;

1330 bool random = false;

1331 int limit = 0;

1332 for (context_map::iterator i=priv.env_to.begin(); i!=priv.env_to.end(); i++) {

1333 const char *rcpt = (*i).first;

1334 CONTEXT &con = *((*i).second);

1335 if (!con.acceptable_content(*priv.memory, score, bulk, msg)) {

1336 // bad html tags or excessive hosts or

1337 // high spam assassin score or dcc bulk threshold exceedeed

1338 smfi_delrcpt(ctx, (char*)rcpt);

1339 }

1340 else {

1341 alive.insert(rcpt);

1342 random |= con.get_host_random();

1343 limit = max(limit, con.get_host_limit());

1344 }

1345 }

1346 bool rejecting = alive.empty(); // if alive is empty, we must have set msg above in acceptable_content()

1347 if (!rejecting) {

1348 const char *fmt;

1349 const char *found;

1350 if (check_hosts(priv, random, limit, fmt, host, ip, found)) {

1351 if (found) {

1352 // uribl style

1353 snprintf(buf, sizeof(buf), fmt, host, found);

1354 }

1355 else {

1356 // dnsbl style

1357 char adr[sizeof "255.255.255.255 "];

1358 adr[0] = '\0';

1359 inet_ntop(AF_INET, (const u_char *)&ip, adr, sizeof(adr));

1360 snprintf(buf, sizeof(buf), fmt, host, adr);

1361 }

1362 msg = string(buf);

1363 rejecting = true;

1364 }

1365 }

1366 if (!rejecting) {

1367 if (priv.want_dccgrey && grey) {

1368 smfi_setreply(ctx, (char*)"452", (char*)"4.2.1", (char*)"temporary greylist embargoed");

1369 rc = SMFIS_TEMPFAIL;

1370 }

1371 else rc = SMFIS_CONTINUE;

1372 }

1373 else if (!priv.have_whites) {

1374 // can reject the entire message

1375 snprintf(buf, sizeof(buf), "%s", msg.c_str());

1376 smfi_setreply(ctx, (char*)"550", (char*)"5.7.1", buf);

1377 rc = SMFIS_REJECT;

1378 }

1379 else {

1380 // need to accept it but remove the recipients that don't want it

1381 for (string_set::iterator i=alive.begin(); i!=alive.end(); i++) {

1382 const char *rcpt = *i;

1383 smfi_delrcpt(ctx, (char*)rcpt);

1384 }

1385 rc = SMFIS_CONTINUE;

1386 }

1387 }

1388 // reset for a new message on the same connection

1389 mlfi_abort(ctx);

1390 return rc;

1391 }

1392

1393 sfsistat mlfi_abort(SMFICTX *ctx)

1394 {

1395 mlfiPriv &priv = *MLFIPRIV;

1396 priv.reset();

1397 return SMFIS_CONTINUE;

1398 }

1399

1400 sfsistat mlfi_close(SMFICTX *ctx)

1401 {

1402 mlfiPriv *priv = MLFIPRIV;

1403 if (!priv) return SMFIS_CONTINUE;

1404 delete priv;

1405 smfi_setpriv(ctx, NULL);

1406 return SMFIS_CONTINUE;

1407 }

1408

1409 struct smfiDesc smfilter =

1410 {

1411 (char*)"DNSBL", // filter name

1412 SMFI_VERSION, // version code -- do not change

1413 SMFIF_DELRCPT, // flags

1414 mlfi_connect, // connection info filter

1415 mlfi_helo, // SMTP HELO command filter

1416 mlfi_envfrom, // envelope sender filter

1417 mlfi_envrcpt, // envelope recipient filter

1418 mlfi_header, // header filter

1419 mlfi_eoh, // end of header

1420 mlfi_body, // body block filter

1421 mlfi_eom, // end of message

1422 mlfi_abort, // message aborted

1423 mlfi_close, // connection cleanup

1424 };

1425

1426

1427 ////////////////////////////////////////////////

1428 // reload the config

1429 //

1430 CONFIG* new_conf();

1431 CONFIG* new_conf() {

1432 CONFIG *newc = new CONFIG;

1433 pthread_mutex_lock(&config_mutex);

1434 newc->generation = generation++;

1435 pthread_mutex_unlock(&config_mutex);

1436 if (debug_syslog) {

1437 char buf[maxlen];

1438 snprintf(buf, sizeof(buf), "loading configuration generation %d", newc->generation);

1439 my_syslog(buf);

1440 }

1441 if (load_conf(*newc, "dnsbl.conf")) {

1442 newc->load_time = time(NULL);

1443 return newc;

1444 }

1445 delete newc;

1446 return NULL;

1447 }

1448

1449

1450 ////////////////////////////////////////////////

1451 // thread to watch the old config files for changes

1452 // and reload when needed.

1453 // we also clear the SMTP AUTH recipient counts hourly

1454 //

1455 extern "C" {void* config_loader(void *arg);}

1456 void* config_loader(void *arg) {

1457 int loop = 0;

1458 while (loader_run) {

1459 sleep(180); // look for modifications every 3 minutes

1460 if (!loader_run) break;

1461 loop++;

1462 if (loop == 20) {

1463 // three minutes thru each loop, 20 loops per hour

1464 // clear the recipient counts

1465 pthread_mutex_lock(&rate_mutex);

1466 for (rcpt_rates::iterator i=rcpt_counts.begin(); i!=rcpt_counts.end(); i++) {

1467 (*i).second = 0;

1468 }

1469 pthread_mutex_unlock(&rate_mutex);

1470 loop = 0;

1471 }

1472 CONFIG &dc = *config;

1473 time_t then = dc.load_time;

1474 struct stat st;

1475 bool reload = false;

1476 for (string_set::iterator i=dc.config_files.begin(); i!=dc.config_files.end(); i++) {

1477 const char *fn = *i;

1478 if (stat(fn, &st)) reload = true; // file disappeared

1479 else if (st.st_mtime > then) reload = true; // file modified

1480 if (reload) break;

1481 }

1482 if (reload) {

1483 CONFIG *newc = new_conf();

1484 if (newc) {

1485 // replace the global config pointer

1486 pthread_mutex_lock(&config_mutex);

1487 CONFIG *pc = config;

1488 bool last = pc && (!pc->reference_count);

1489 config = newc;

1490 pthread_mutex_unlock(&config_mutex);

1491 if (last) delete pc; // there were no references to this config

1492 }

1493 else {

1494 // failed to load new config

1495 my_syslog("failed to load new configuration");

1496 system("echo 'failed to load new dnsbl configuration from /etc/dnsbl' | mail -s 'error in /etc/dnsbl configuration' root");

1497 // update the load time on the current config to prevent complaining every 3 minutes

1498 dc.load_time = time(NULL);

1499 }

1500 }

1501 }

1502 return NULL;

1503 }

1504

1505

1506 void usage(const char *prog);

1507 void usage(const char *prog)

1508 {

1509 fprintf(stderr, "Usage: %s [-d [level]] [-c] [-s] [-e from|to] [-b dccifd-addr] -r port -p sm-sock-addr [-t timeout]\n", prog);

1510 fprintf(stderr, "where dccifd_addr is for the connection to dccifd\n");

1511 fprintf(stderr, " and should be local-domain-socket-file-name\n");

1512 fprintf(stderr, "where port is for the connection to our own dns resolver processes\n");

1513 fprintf(stderr, " and should be local-domain-socket-file-name\n");

1514 fprintf(stderr, "where sm-sock-addr is for the connection to sendmail\n");

1515 fprintf(stderr, " and should be one of\n");

1516 fprintf(stderr, " inet:port@ip-address\n");

1517 fprintf(stderr, " local:local-domain-socket-file-name\n");

1518 fprintf(stderr, "-c will load and dump the config to stdout\n");

1519 fprintf(stderr, "-s will stress test the config loading code by repeating the load/free cycle\n");

1520 fprintf(stderr, " in an infinte loop.\n");

1521 fprintf(stderr, "-d will set the syslog message level, currently 0 to 3\n");

1522 fprintf(stderr, "-e will print the results of looking up the from and to addresses in the\n");

1523 fprintf(stderr, " current config. The | character is used to separate the from and to\n");

1524 fprintf(stderr, " addresses in the argument to the -e switch\n");

1525 }

1526

1527

1528

1529 void setup_socket(const char *sock);

1530 void setup_socket(const char *sock) {

1531 unlink(sock);

1532 }

1533

1534

1535 /*

1536 * The signal handler function -- only gets called when a SIGCHLD

1537 * is received, ie when a child terminates

1538 */

1539 void sig_chld(int signo)

1540 {

1541 int status;

1542 /* Wait for any child without blocking */

1543 while (waitpid(-1, &status, WNOHANG) > 0) {

1544 // ignore child exit status, we only do this to cleanup zombies

1545 }

1546 }

1547

1548

1549 int main(int argc, char**argv)

1550 {

1551 token_init();

1552 bool check = false;

1553 bool stress = false;

1554 bool setconn = false;

1555 bool setreso = false;

1556 const char *email = NULL;

1557 int c;

1558 const char *args = "b:r:p:t:e:d:chs";

1559 extern char *optarg;

1560

1561 // setup srs coding detection

1562 if (regcomp(&srs_pattern, "^srs(0|1)=([^=]*)=([^=]*)=([^=]*)=([^@]*)@", REG_ICASE | REG_EXTENDED)) {

1563 printf("cannot compile regex pattern to find srs coding in mail addresses\n");

1564 exit(3);

1565 }

1566

1567 // Process command line options

1568 while ((c = getopt(argc, argv, args)) != -1) {

1569 switch (c) {

1570 case 'b':

1571 if (optarg == NULL || *optarg == '\0') {

1572 fprintf(stderr, "Illegal dccifd socket: %s\n", optarg);

1573 exit(EX_USAGE);

1574 }

1575 dccifd_port = strdup(optarg);

1576 break;

1577

1578 case 'r':

1579 if (optarg == NULL || *optarg == '\0') {

1580 fprintf(stderr, "Illegal resolver socket: %s\n", optarg);

1581 exit(EX_USAGE);

1582 }

1583 resolver_port = strdup(optarg);

1584 setup_socket(resolver_port);

1585 setreso = true;

1586 break;

1587

1588 case 'p':

1589 if (optarg == NULL || *optarg == '\0') {

1590 fprintf(stderr, "Illegal sendmail socket: %s\n", optarg);

1591 exit(EX_USAGE);

1592 }

1593 if (smfi_setconn(optarg) == MI_FAILURE) {

1594 fprintf(stderr, "smfi_setconn failed\n");

1595 exit(EX_SOFTWARE);

1596 }

1597 if (strncasecmp(optarg, "unix:", 5) == 0) setup_socket(optarg + 5);

1598 else if (strncasecmp(optarg, "local:", 6) == 0) setup_socket(optarg + 6);

1599 setconn = true;

1600 break;

1601

1602 case 't':

1603 if (optarg == NULL || *optarg == '\0') {

1604 fprintf(stderr, "Illegal timeout: %s\n", optarg);

1605 exit(EX_USAGE);

1606 }

1607 if (smfi_settimeout(atoi(optarg)) == MI_FAILURE) {

1608 fprintf(stderr, "smfi_settimeout failed\n");

1609 exit(EX_SOFTWARE);

1610 }

1611 break;

1612

1613 case 'e':

1614 if (email) free((void*)email);

1615 email = strdup(optarg);

1616 break;

1617

1618 case 'c':

1619 check = true;

1620 break;

1621

1622 case 's':

1623 stress = true;

1624 break;

1625

1626 case 'd':

1627 if (optarg == NULL || *optarg == '\0') debug_syslog = 1;

1628 else debug_syslog = atoi(optarg);

1629 break;

1630

1631 case 'h':

1632 default:

1633 usage(argv[0]);

1634 exit(EX_USAGE);

1635 }

1636 }

1637

1638 if (check) {

1639 use_syslog = false;

1640 debug_syslog = 10;

1641 CONFIG *conf = new_conf();

1642 if (conf) {

1643 conf->dump();

1644 delete conf;

1645 clear_strings(); // for valgrind checking

1646 return 0;

1647 }

1648 else {

1649 return 1; // config failed to load

1650 }

1651 }

1652

1653 if (stress) {

1654 fprintf(stdout, "stress testing\n");

1655 while (1) {

1656 for (int i=0; i<10; i++) {

1657 CONFIG *conf = new_conf();

1658 if (conf) delete conf;

1659 }

1660 fprintf(stdout, ".");

1661 fflush(stdout);

1662 sleep(1);

1663 }

1664 }

1665

1666 if (email) {

1667 char *x = strchr(email, '|');

1668 if (x) {

1669 *x = '\0';

1670 const char *from = to_lower_string(email);

1671 const char *to = to_lower_string(x+1);

1672 use_syslog = false;

1673 CONFIG *conf = new_conf();

1674 if (conf) {

1675 CONTEXTP con = conf->find_context(to);

1676 char buf[maxlen];

1677 fprintf(stdout, "envelope to <%s> finds context %s\n", to, con->get_full_name(buf,maxlen));

1678 CONTEXTP fc = con->find_context(from);

1679 fprintf(stdout, "envelope from <%s> finds context %s\n", from, fc->get_full_name(buf,maxlen));

1680 const char *st = fc->find_from(from);

1681 fprintf(stdout, "envelope from <%s> finds status %s\n", from, st);

1682 bool self = (strcmp(from, to) == 0);

1683 if ((st == token_white) && self) fprintf(stdout, "ignore self whitelisting\n");

1684 delete conf;

1685 }

1686 }

1687 return 0;

1688 }

1689

1690 if (!setconn) {

1691 fprintf(stderr, "%s: Missing required -p argument\n", argv[0]);

1692 usage(argv[0]);

1693 exit(EX_USAGE);

1694 }

1695

1696 if (!setreso) {

1697 fprintf(stderr, "%s: Missing required -r argument\n", argv[0]);

1698 usage(argv[0]);

1699 exit(EX_USAGE);

1700 }

1701

1702 if (smfi_register(smfilter) == MI_FAILURE) {

1703 fprintf(stderr, "smfi_register failed\n");

1704 exit(EX_UNAVAILABLE);

1705 }

1706

1707 // switch to background mode

1708 if (daemon(1,0) < 0) {

1709 fprintf(stderr, "daemon() call failed\n");

1710 exit(EX_UNAVAILABLE);

1711 }

1712

1713 // write the pid

1714 const char *pidpath = "/var/run/dnsbl.pid";

1715 unlink(pidpath);

1716 FILE *f = fopen(pidpath, "w");

1717 if (f) {

1718 #ifdef linux

1719 // from a comment in the DCC source code:

1720 // Linux threads are broken. Signals given the

1721 // original process are delivered to only the

1722 // thread that happens to have that PID. The

1723 // sendmail libmilter thread that needs to hear

1724 // SIGINT and other signals does not, and that breaks

1725 // scripts that need to stop milters.

1726 // However, signaling the process group works.

1727 fprintf(f, "-%d\n", (u_int)getpgrp());

1728 #else

1729 fprintf(f, "%d\n", (u_int)getpid());

1730 #endif

1731 fclose(f);

1732 }

1733

1734 // initialize the thread sync objects

1735 pthread_mutex_init(&config_mutex, 0);

1736 pthread_mutex_init(&syslog_mutex, 0);

1737 pthread_mutex_init(&resolve_mutex, 0);

1738 pthread_mutex_init(&fd_pool_mutex, 0);

1739 pthread_mutex_init(&verifier_mutex, 0);

1740 pthread_mutex_init(&whitelister_mutex, 0);

1741

1742 // drop root privs

1743 struct passwd *pw = getpwnam("dnsbl");

1744 if (pw) {

1745 if (setgid(pw->pw_gid) == -1) {

1746 my_syslog("failed to switch to group dnsbl");

1747 }

1748 if (setuid(pw->pw_uid) == -1) {

1749 my_syslog("failed to switch to user dnsbl");

1750 }

1751 }

1752

1753 // load the initial config

1754 config = new_conf();

1755 if (!config) {

1756 my_syslog("failed to load initial configuration, quitting");

1757 exit(1);

1758 }

1759

1760 // fork off the resolver listener process

1761 pid_t child = fork();

1762 if (child < 0) {

1763 my_syslog("failed to create resolver listener process");

1764 exit(0);

1765 }

1766 if (child == 0) {

1767 // we are the child - dns resolver listener process

1768 resolver_socket = socket(AF_UNIX, SOCK_STREAM, 0);

1769 if (resolver_socket < 0) {

1770 my_syslog("child failed to create resolver socket");

1771 exit(0); // failed

1772 }

1773 sockaddr_un server;

1774 memset(&server, '\0', sizeof(server));

1775 server.sun_family = AF_UNIX;

1776 strncpy(server.sun_path, resolver_port, sizeof(server.sun_path)-1);

1777 //try to bind the address to the socket.

1778 if (bind(resolver_socket, (sockaddr *)&server, sizeof(server)) < 0) {

1779 // bind failed

1780 shutdown(resolver_socket, SHUT_RDWR);

1781 close(resolver_socket);

1782 my_syslog("child failed to bind resolver socket");

1783 exit(0); // failed

1784 }

1785 //listen on the socket.

1786 if (listen(resolver_socket, 10) < 0) {

1787 // listen failed

1788 shutdown(resolver_socket, SHUT_RDWR);

1789 close(resolver_socket);

1790 my_syslog("child failed to listen to resolver socket");

1791 exit(0); // failed

1792 }

1793 // setup sigchld handler to prevent zombies

1794 struct sigaction act;

1795 act.sa_handler = sig_chld; // Assign sig_chld as our SIGCHLD handler

1796 sigemptyset(&act.sa_mask); // We don't want to block any other signals in this example

1797 act.sa_flags = SA_NOCLDSTOP; // only want children that have terminated

1798 if (sigaction(SIGCHLD, &act, NULL) < 0) {

1799 my_syslog("child failed to setup SIGCHLD handler");

1800 exit(0); // failed

1801 }

1802 while (true) {

1803 sockaddr_un client;

1804 socklen_t clientlen = sizeof(client);

1805 int s = accept(resolver_socket, (sockaddr *)&client, &clientlen);

1806 if (s > 0) {

1807 // accept worked, it did not get cancelled before we could accept it

1808 // fork off a process to handle this connection

1809 int newchild = fork();

1810 if (newchild == 0) {

1811 // this is the worker process

1812 // child does not need the listening socket

1813 close(resolver_socket);

1814 #ifdef NS_PACKETSZ

1815 res_init();

1816 _res.retry = 2;

1817 _res.retrans = RES_TIMEOUT;

1818 #endif

1819 process_resolver_requests(s);

1820 exit(0);

1821 }

1822 else {

1823 // this is the parent

1824 // parent does not need the accepted socket

1825 close(s);

1826 }

1827 }

1828 }

1829 exit(0); // make sure we don't fall thru.

1830 }

1831 else {

1832 sleep(2); // allow child to get started

1833 }

1834

1835 // only create threads after the fork() in daemon

1836 pthread_t tid;

1837 if (pthread_create(&tid, 0, config_loader, 0))

1838 my_syslog("failed to create config loader thread");

1839 if (pthread_detach(tid))

1840 my_syslog("failed to detach config loader thread");

1841

1842 if (pthread_create(&tid, 0, verify_closer, 0))

1843 my_syslog("failed to create verify closer thread");

1844 if (pthread_detach(tid))

1845 my_syslog("failed to detach verify closer thread");

1846

1847 if (pthread_create(&tid, 0, whitelister_writer, 0))

1848 my_syslog("failed to create autowhite writer thread");

1849 if (pthread_detach(tid))

1850 my_syslog("failed to detach autowhite writer thread");

1851

1852 time_t starting = time(NULL);

1853 int rc = smfi_main();

1854 if ((rc != MI_SUCCESS) && (time(NULL) > starting+5*60)) {

1855 my_syslog("trying to restart after smfi_main()");

1856 loader_run = false; // eventually the config loader thread will terminate

1857 execvp(argv[0], argv);

1858 }

1859 exit((rc == MI_SUCCESS) ? 0 : EX_UNAVAILABLE);

1860 }

1861