992e088869e05ab48329466324a64e9e3d023a13
[exim.git] / src / src / spool_in.c
1 /*************************************************
2 * Exim - an Internet mail transport agent *
3 *************************************************/
4
5 /* Copyright (c) University of Cambridge 1995 - 2015 */
6 /* See the file NOTICE for conditions of use and distribution. */
7
8 /* Functions for reading spool files. When compiling for a utility (eximon),
9 not all are needed, and some functionality can be cut out. */
10
11
12 #include "exim.h"
13
14
15
16 #ifndef COMPILE_UTILITY
17 /*************************************************
18 * Open and lock data file *
19 *************************************************/
20
21 /* The data file is the one that is used for locking, because the header file
22 can get replaced during delivery because of header rewriting. The file has
23 to opened with write access so that we can get an exclusive lock, but in
24 fact it won't be written to. Just in case there's a major disaster (e.g.
25 overwriting some other file descriptor with the value of this one), open it
26 with append.
27
28 Argument: the id of the message
29 Returns: TRUE if file successfully opened and locked
30
31 Side effect: deliver_datafile is set to the fd of the open file.
32 */
33
34 BOOL
35 spool_open_datafile(uschar *id)
36 {
37 int i;
38 struct stat statbuf;
39 flock_t lock_data;
40 uschar spoolname[256];
41
42 /* If split_spool_directory is set, first look for the file in the appropriate
43 sub-directory of the input directory. If it is not found there, try the input
44 directory itself, to pick up leftovers from before the splitting. If split_
45 spool_directory is not set, first look in the main input directory. If it is
46 not found there, try the split sub-directory, in case it is left over from a
47 splitting state. */
48
49 for (i = 0; i < 2; i++)
50 {
51 int save_errno;
52 message_subdir[0] = (split_spool_directory == (i == 0))? id[5] : 0;
53 sprintf(CS spoolname, "%s/input/%s/%s-D", spool_directory, message_subdir, id);
54 deliver_datafile = Uopen(spoolname, O_RDWR | O_APPEND, 0);
55 if (deliver_datafile >= 0) break;
56 save_errno = errno;
57 if (errno == ENOENT)
58 {
59 if (i == 0) continue;
60 if (!queue_running)
61 log_write(0, LOG_MAIN, "Spool file %s-D not found", id);
62 }
63 else log_write(0, LOG_MAIN, "Spool error for %s: %s", spoolname,
64 strerror(errno));
65 errno = save_errno;
66 return FALSE;
67 }
68
69 /* File is open and message_subdir is set. Set the close-on-exec flag, and lock
70 the file. We lock only the first line of the file (containing the message ID)
71 because this apparently is needed for running Exim under Cygwin. If the entire
72 file is locked in one process, a sub-process cannot access it, even when passed
73 an open file descriptor (at least, I think that's the Cygwin story). On real
74 Unix systems it doesn't make any difference as long as Exim is consistent in
75 what it locks. */
76
77 (void)fcntl(deliver_datafile, F_SETFD, fcntl(deliver_datafile, F_GETFD) |
78 FD_CLOEXEC);
79
80 lock_data.l_type = F_WRLCK;
81 lock_data.l_whence = SEEK_SET;
82 lock_data.l_start = 0;
83 lock_data.l_len = SPOOL_DATA_START_OFFSET;
84
85 if (fcntl(deliver_datafile, F_SETLK, &lock_data) < 0)
86 {
87 log_write(L_skip_delivery,
88 LOG_MAIN,
89 "Spool file is locked (another process is handling this message)");
90 (void)close(deliver_datafile);
91 deliver_datafile = -1;
92 errno = 0;
93 return FALSE;
94 }
95
96 /* Get the size of the data; don't include the leading filename line
97 in the count, but add one for the newline before the data. */
98
99 if (fstat(deliver_datafile, &statbuf) == 0)
100 {
101 message_body_size = statbuf.st_size - SPOOL_DATA_START_OFFSET;
102 message_size = message_body_size + 1;
103 }
104
105 return TRUE;
106 }
107 #endif /* COMPILE_UTILITY */
108
109
110
111 /*************************************************
112 * Read non-recipients tree from spool file *
113 *************************************************/
114
115 /* The tree of non-recipients is written to the spool file in a form that
116 makes it easy to read back into a tree. The format is as follows:
117
118 . Each node is preceded by two letter(Y/N) indicating whether it has left
119 or right children. There's one space after the two flags, before the name.
120
121 . The left subtree (if any) then follows, then the right subtree (if any).
122
123 This function is entered with the next input line in the buffer. Note we must
124 save the right flag before recursing with the same buffer.
125
126 Once the tree is read, we re-construct the balance fields by scanning the tree.
127 I forgot to write them out originally, and the compatible fix is to do it this
128 way. This initial local recursing function does the necessary.
129
130 Arguments:
131 node tree node
132
133 Returns: maximum depth below the node, including the node itself
134 */
135
136 static int
137 count_below(tree_node *node)
138 {
139 int nleft, nright;
140 if (node == NULL) return 0;
141 nleft = count_below(node->left);
142 nright = count_below(node->right);
143 node->balance = (nleft > nright)? 1 : ((nright > nleft)? 2 : 0);
144 return 1 + ((nleft > nright)? nleft : nright);
145 }
146
147 /* This is the real function...
148
149 Arguments:
150 connect pointer to the root of the tree
151 f FILE to read data from
152 buffer contains next input line; further lines read into it
153 buffer_size size of the buffer
154
155 Returns: FALSE on format error
156 */
157
158 static BOOL
159 read_nonrecipients_tree(tree_node **connect, FILE *f, uschar *buffer,
160 int buffer_size)
161 {
162 tree_node *node;
163 int n = Ustrlen(buffer);
164 BOOL right = buffer[1] == 'Y';
165
166 if (n < 5) return FALSE; /* malformed line */
167 buffer[n-1] = 0; /* Remove \n */
168 node = store_get(sizeof(tree_node) + n - 3);
169 *connect = node;
170 Ustrcpy(node->name, buffer + 3);
171 node->data.ptr = NULL;
172
173 if (buffer[0] == 'Y')
174 {
175 if (Ufgets(buffer, buffer_size, f) == NULL ||
176 !read_nonrecipients_tree(&node->left, f, buffer, buffer_size))
177 return FALSE;
178 }
179 else node->left = NULL;
180
181 if (right)
182 {
183 if (Ufgets(buffer, buffer_size, f) == NULL ||
184 !read_nonrecipients_tree(&node->right, f, buffer, buffer_size))
185 return FALSE;
186 }
187 else node->right = NULL;
188
189 (void) count_below(*connect);
190 return TRUE;
191 }
192
193
194
195
196 /*************************************************
197 * Read spool header file *
198 *************************************************/
199
200 /* This function reads a spool header file and places the data into the
201 appropriate global variables. The header portion is always read, but header
202 structures are built only if read_headers is set true. It isn't, for example,
203 while generating -bp output.
204
205 It may be possible for blocks of nulls (binary zeroes) to get written on the
206 end of a file if there is a system crash during writing. It was observed on an
207 earlier version of Exim that omitted to fsync() the files - this is thought to
208 have been the cause of that incident, but in any case, this code must be robust
209 against such an event, and if such a file is encountered, it must be treated as
210 malformed.
211
212 Arguments:
213 name name of the header file, including the -H
214 read_headers TRUE if in-store header structures are to be built
215 subdir_set TRUE is message_subdir is already set
216
217 Returns: spool_read_OK success
218 spool_read_notopen open failed
219 spool_read_enverror error in the envelope portion
220 spool_read_hdrdrror error in the header portion
221 */
222
223 int
224 spool_read_header(uschar *name, BOOL read_headers, BOOL subdir_set)
225 {
226 FILE *f = NULL;
227 int n;
228 int rcount = 0;
229 long int uid, gid;
230 BOOL inheader = FALSE;
231 uschar *p;
232
233 /* Reset all the global variables to their default values. However, there is
234 one exception. DO NOT change the default value of dont_deliver, because it may
235 be forced by an external setting. */
236
237 acl_var_c = acl_var_m = NULL;
238 authenticated_id = NULL;
239 authenticated_sender = NULL;
240 allow_unqualified_recipient = FALSE;
241 allow_unqualified_sender = FALSE;
242 body_linecount = 0;
243 body_zerocount = 0;
244 deliver_firsttime = FALSE;
245 deliver_freeze = FALSE;
246 deliver_frozen_at = 0;
247 deliver_manual_thaw = FALSE;
248 /* dont_deliver must NOT be reset */
249 header_list = header_last = NULL;
250 host_lookup_deferred = FALSE;
251 host_lookup_failed = FALSE;
252 interface_address = NULL;
253 interface_port = 0;
254 local_error_message = FALSE;
255 local_scan_data = NULL;
256 max_received_linelength = 0;
257 message_linecount = 0;
258 received_protocol = NULL;
259 received_count = 0;
260 recipients_list = NULL;
261 sender_address = NULL;
262 sender_fullhost = NULL;
263 sender_helo_name = NULL;
264 sender_host_address = NULL;
265 sender_host_name = NULL;
266 sender_host_port = 0;
267 sender_host_authenticated = NULL;
268 sender_ident = NULL;
269 sender_local = FALSE;
270 sender_set_untrusted = FALSE;
271 smtp_active_hostname = primary_hostname;
272 tree_nonrecipients = NULL;
273
274 #ifdef EXPERIMENTAL_BRIGHTMAIL
275 bmi_run = 0;
276 bmi_verdicts = NULL;
277 #endif
278
279 #ifndef DISABLE_DKIM
280 dkim_signers = NULL;
281 dkim_disable_verify = FALSE;
282 dkim_collect_input = FALSE;
283 #endif
284
285 #ifdef SUPPORT_TLS
286 tls_in.certificate_verified = FALSE;
287 # ifdef EXPERIMENTAL_DANE
288 tls_in.dane_verified = FALSE;
289 # endif
290 tls_in.cipher = NULL;
291 # ifndef COMPILE_UTILITY /* tls support fns not built in */
292 tls_free_cert(&tls_in.ourcert);
293 tls_free_cert(&tls_in.peercert);
294 # endif
295 tls_in.peerdn = NULL;
296 tls_in.sni = NULL;
297 tls_in.ocsp = OCSP_NOT_REQ;
298 #endif
299
300 #ifdef WITH_CONTENT_SCAN
301 spam_bar = NULL;
302 spam_score = NULL;
303 spam_score_int = NULL;
304 #endif
305
306 #if defined(SUPPORT_I18N) && !defined(COMPILE_UTILITY)
307 message_smtputf8 = FALSE;
308 message_utf8_downconvert = 0;
309 #endif
310
311 dsn_ret = 0;
312 dsn_envid = NULL;
313
314 /* Generate the full name and open the file. If message_subdir is already
315 set, just look in the given directory. Otherwise, look in both the split
316 and unsplit directories, as for the data file above. */
317
318 for (n = 0; n < 2; n++)
319 {
320 if (!subdir_set)
321 message_subdir[0] = (split_spool_directory == (n == 0))? name[5] : 0;
322 sprintf(CS big_buffer, "%s/input/%s/%s", spool_directory, message_subdir,
323 name);
324 f = Ufopen(big_buffer, "rb");
325 if (f != NULL) break;
326 if (n != 0 || subdir_set || errno != ENOENT) return spool_read_notopen;
327 }
328
329 errno = 0;
330
331 #ifndef COMPILE_UTILITY
332 DEBUG(D_deliver) debug_printf("reading spool file %s\n", name);
333 #endif /* COMPILE_UTILITY */
334
335 /* The first line of a spool file contains the message id followed by -H (i.e.
336 the file name), in order to make the file self-identifying. */
337
338 if (Ufgets(big_buffer, big_buffer_size, f) == NULL) goto SPOOL_READ_ERROR;
339 if (Ustrlen(big_buffer) != MESSAGE_ID_LENGTH + 3 ||
340 Ustrncmp(big_buffer, name, MESSAGE_ID_LENGTH + 2) != 0)
341 goto SPOOL_FORMAT_ERROR;
342
343 /* The next three lines in the header file are in a fixed format. The first
344 contains the login, uid, and gid of the user who caused the file to be written.
345 There are known cases where a negative gid is used, so we allow for both
346 negative uids and gids. The second contains the mail address of the message's
347 sender, enclosed in <>. The third contains the time the message was received,
348 and the number of warning messages for delivery delays that have been sent. */
349
350 if (Ufgets(big_buffer, big_buffer_size, f) == NULL) goto SPOOL_READ_ERROR;
351
352 p = big_buffer + Ustrlen(big_buffer);
353 while (p > big_buffer && isspace(p[-1])) p--;
354 *p = 0;
355 if (!isdigit(p[-1])) goto SPOOL_FORMAT_ERROR;
356 while (p > big_buffer && (isdigit(p[-1]) || '-' == p[-1])) p--;
357 gid = Uatoi(p);
358 if (p <= big_buffer || *(--p) != ' ') goto SPOOL_FORMAT_ERROR;
359 *p = 0;
360 if (!isdigit(p[-1])) goto SPOOL_FORMAT_ERROR;
361 while (p > big_buffer && (isdigit(p[-1]) || '-' == p[-1])) p--;
362 uid = Uatoi(p);
363 if (p <= big_buffer || *(--p) != ' ') goto SPOOL_FORMAT_ERROR;
364 *p = 0;
365
366 originator_login = string_copy(big_buffer);
367 originator_uid = (uid_t)uid;
368 originator_gid = (gid_t)gid;
369
370 /* envelope from */
371 if (Ufgets(big_buffer, big_buffer_size, f) == NULL) goto SPOOL_READ_ERROR;
372 n = Ustrlen(big_buffer);
373 if (n < 3 || big_buffer[0] != '<' || big_buffer[n-2] != '>')
374 goto SPOOL_FORMAT_ERROR;
375
376 sender_address = store_get(n-2);
377 Ustrncpy(sender_address, big_buffer+1, n-3);
378 sender_address[n-3] = 0;
379
380 /* time */
381 if (Ufgets(big_buffer, big_buffer_size, f) == NULL) goto SPOOL_READ_ERROR;
382 if (sscanf(CS big_buffer, "%d %d", &received_time, &warning_count) != 2)
383 goto SPOOL_FORMAT_ERROR;
384
385 message_age = time(NULL) - received_time;
386
387 #ifndef COMPILE_UTILITY
388 DEBUG(D_deliver) debug_printf("user=%s uid=%ld gid=%ld sender=%s\n",
389 originator_login, (long int)originator_uid, (long int)originator_gid,
390 sender_address);
391 #endif /* COMPILE_UTILITY */
392
393 /* Now there may be a number of optional lines, each starting with "-". If you
394 add a new setting here, make sure you set the default above.
395
396 Because there are now quite a number of different possibilities, we use a
397 switch on the first character to avoid too many failing tests. Thanks to Nico
398 Erfurth for the patch that implemented this. I have made it even more efficient
399 by not re-scanning the first two characters.
400
401 To allow new versions of Exim that add additional flags to interwork with older
402 versions that do not understand them, just ignore any lines starting with "-"
403 that we don't recognize. Otherwise it wouldn't be possible to back off a new
404 version that left new-style flags written on the spool. */
405
406 p = big_buffer + 2;
407 for (;;)
408 {
409 int len;
410 if (Ufgets(big_buffer, big_buffer_size, f) == NULL) goto SPOOL_READ_ERROR;
411 if (big_buffer[0] != '-') break;
412 while ( (len = Ustrlen(big_buffer)) == big_buffer_size-1
413 && big_buffer[len-1] != '\n'
414 )
415 { /* buffer not big enough for line; certs make this possible */
416 uschar * buf;
417 if (big_buffer_size >= BIG_BUFFER_SIZE*4) goto SPOOL_READ_ERROR;
418 buf = store_get_perm(big_buffer_size *= 2);
419 memcpy(buf, big_buffer, --len);
420 big_buffer = buf;
421 if (Ufgets(big_buffer+len, big_buffer_size-len, f) == NULL)
422 goto SPOOL_READ_ERROR;
423 }
424 big_buffer[len-1] = 0;
425
426 switch(big_buffer[1])
427 {
428 case 'a':
429
430 /* Nowadays we use "-aclc" and "-aclm" for the different types of ACL
431 variable, because Exim allows any number of them, with arbitrary names.
432 The line in the spool file is "-acl[cm] <name> <length>". The name excludes
433 the c or m. */
434
435 if (Ustrncmp(p, "clc ", 4) == 0 ||
436 Ustrncmp(p, "clm ", 4) == 0)
437 {
438 uschar *name, *endptr;
439 int count;
440 tree_node *node;
441 endptr = Ustrchr(big_buffer + 6, ' ');
442 if (endptr == NULL) goto SPOOL_FORMAT_ERROR;
443 name = string_sprintf("%c%.*s", big_buffer[4], endptr - big_buffer - 6,
444 big_buffer + 6);
445 if (sscanf(CS endptr, " %d", &count) != 1) goto SPOOL_FORMAT_ERROR;
446 node = acl_var_create(name);
447 node->data.ptr = store_get(count + 1);
448 if (fread(node->data.ptr, 1, count+1, f) < count) goto SPOOL_READ_ERROR;
449 ((uschar*)node->data.ptr)[count] = 0;
450 }
451
452 else if (Ustrcmp(p, "llow_unqualified_recipient") == 0)
453 allow_unqualified_recipient = TRUE;
454 else if (Ustrcmp(p, "llow_unqualified_sender") == 0)
455 allow_unqualified_sender = TRUE;
456
457 else if (Ustrncmp(p, "uth_id", 6) == 0)
458 authenticated_id = string_copy(big_buffer + 9);
459 else if (Ustrncmp(p, "uth_sender", 10) == 0)
460 authenticated_sender = string_copy(big_buffer + 13);
461 else if (Ustrncmp(p, "ctive_hostname", 14) == 0)
462 smtp_active_hostname = string_copy(big_buffer + 17);
463
464 /* For long-term backward compatibility, we recognize "-acl", which was
465 used before the number of ACL variables changed from 10 to 20. This was
466 before the subsequent change to an arbitrary number of named variables.
467 This code is retained so that upgrades from very old versions can still
468 handle old-format spool files. The value given after "-acl" is a number
469 that is 0-9 for connection variables, and 10-19 for message variables. */
470
471 else if (Ustrncmp(p, "cl ", 3) == 0)
472 {
473 int index, count;
474 uschar name[20]; /* Need plenty of space for %d format */
475 tree_node *node;
476 if ( sscanf(CS big_buffer + 5, "%d %d", &index, &count) != 2
477 || index >= 20
478 )
479 goto SPOOL_FORMAT_ERROR;
480 if (index < 10)
481 (void) string_format(name, sizeof(name), "%c%d", 'c', index);
482 else
483 (void) string_format(name, sizeof(name), "%c%d", 'm', index - 10);
484 node = acl_var_create(name);
485 node->data.ptr = store_get(count + 1);
486 if (fread(node->data.ptr, 1, count+1, f) < count) goto SPOOL_READ_ERROR;
487 ((uschar*)node->data.ptr)[count] = 0;
488 }
489 break;
490
491 case 'b':
492 if (Ustrncmp(p, "ody_linecount", 13) == 0)
493 body_linecount = Uatoi(big_buffer + 15);
494 else if (Ustrncmp(p, "ody_zerocount", 13) == 0)
495 body_zerocount = Uatoi(big_buffer + 15);
496 #ifdef EXPERIMENTAL_BRIGHTMAIL
497 else if (Ustrncmp(p, "mi_verdicts ", 12) == 0)
498 bmi_verdicts = string_copy(big_buffer + 14);
499 #endif
500 break;
501
502 case 'd':
503 if (Ustrcmp(p, "eliver_firsttime") == 0)
504 deliver_firsttime = TRUE;
505 /* Check if the dsn flags have been set in the header file */
506 else if (Ustrncmp(p, "sn_ret", 6) == 0)
507 dsn_ret= atoi(CS big_buffer + 8);
508 else if (Ustrncmp(p, "sn_envid", 8) == 0)
509 dsn_envid = string_copy(big_buffer + 11);
510 break;
511
512 case 'f':
513 if (Ustrncmp(p, "rozen", 5) == 0)
514 {
515 deliver_freeze = TRUE;
516 sscanf(CS big_buffer+7, TIME_T_FMT, &deliver_frozen_at);
517 }
518 break;
519
520 case 'h':
521 if (Ustrcmp(p, "ost_lookup_deferred") == 0)
522 host_lookup_deferred = TRUE;
523 else if (Ustrcmp(p, "ost_lookup_failed") == 0)
524 host_lookup_failed = TRUE;
525 else if (Ustrncmp(p, "ost_auth", 8) == 0)
526 sender_host_authenticated = string_copy(big_buffer + 11);
527 else if (Ustrncmp(p, "ost_name", 8) == 0)
528 sender_host_name = string_copy(big_buffer + 11);
529 else if (Ustrncmp(p, "elo_name", 8) == 0)
530 sender_helo_name = string_copy(big_buffer + 11);
531
532 /* We now record the port number after the address, separated by a
533 dot. For compatibility during upgrading, do nothing if there
534 isn't a value (it gets left at zero). */
535
536 else if (Ustrncmp(p, "ost_address", 11) == 0)
537 {
538 sender_host_port = host_address_extract_port(big_buffer + 14);
539 sender_host_address = string_copy(big_buffer + 14);
540 }
541 break;
542
543 case 'i':
544 if (Ustrncmp(p, "nterface_address", 16) == 0)
545 {
546 interface_port = host_address_extract_port(big_buffer + 19);
547 interface_address = string_copy(big_buffer + 19);
548 }
549 else if (Ustrncmp(p, "dent", 4) == 0)
550 sender_ident = string_copy(big_buffer + 7);
551 break;
552
553 case 'l':
554 if (Ustrcmp(p, "ocal") == 0) sender_local = TRUE;
555 else if (Ustrcmp(big_buffer, "-localerror") == 0)
556 local_error_message = TRUE;
557 else if (Ustrncmp(p, "ocal_scan ", 10) == 0)
558 local_scan_data = string_copy(big_buffer + 12);
559 break;
560
561 case 'm':
562 if (Ustrcmp(p, "anual_thaw") == 0) deliver_manual_thaw = TRUE;
563 else if (Ustrncmp(p, "ax_received_linelength", 22) == 0)
564 max_received_linelength = Uatoi(big_buffer + 24);
565 break;
566
567 case 'N':
568 if (*p == 0) dont_deliver = TRUE; /* -N */
569 break;
570
571 case 'r':
572 if (Ustrncmp(p, "eceived_protocol", 16) == 0)
573 received_protocol = string_copy(big_buffer + 19);
574 break;
575
576 case 's':
577 if (Ustrncmp(p, "ender_set_untrusted", 19) == 0)
578 sender_set_untrusted = TRUE;
579 #ifdef WITH_CONTENT_SCAN
580 else if (Ustrncmp(p, "pam_bar ", 8) == 0)
581 spam_bar = string_copy(big_buffer + 10);
582 else if (Ustrncmp(p, "pam_score ", 10) == 0)
583 spam_score = string_copy(big_buffer + 12);
584 else if (Ustrncmp(p, "pam_score_int ", 14) == 0)
585 spam_score_int = string_copy(big_buffer + 16);
586 #endif
587 #if defined(SUPPORT_I18N) && !defined(COMPILE_UTILITY)
588 else if (Ustrncmp(p, "mtputf8", 7) == 0)
589 message_smtputf8 = TRUE;
590 #endif
591 break;
592
593 #ifdef SUPPORT_TLS
594 case 't':
595 if (Ustrncmp(p, "ls_certificate_verified", 23) == 0)
596 tls_in.certificate_verified = TRUE;
597 else if (Ustrncmp(p, "ls_cipher", 9) == 0)
598 tls_in.cipher = string_copy(big_buffer + 12);
599 # ifndef COMPILE_UTILITY /* tls support fns not built in */
600 else if (Ustrncmp(p, "ls_ourcert", 10) == 0)
601 (void) tls_import_cert(big_buffer + 13, &tls_in.ourcert);
602 else if (Ustrncmp(p, "ls_peercert", 11) == 0)
603 (void) tls_import_cert(big_buffer + 14, &tls_in.peercert);
604 # endif
605 else if (Ustrncmp(p, "ls_peerdn", 9) == 0)
606 tls_in.peerdn = string_unprinting(string_copy(big_buffer + 12));
607 else if (Ustrncmp(p, "ls_sni", 6) == 0)
608 tls_in.sni = string_unprinting(string_copy(big_buffer + 9));
609 else if (Ustrncmp(p, "ls_ocsp", 7) == 0)
610 tls_in.ocsp = big_buffer[10] - '0';
611 break;
612 #endif
613
614 #if defined(SUPPORT_I18N) && !defined(COMPILE_UTILITY)
615 case 'u':
616 if (Ustrncmp(p, "tf8_downcvt", 11) == 0)
617 message_utf8_downconvert = 1;
618 else if (Ustrncmp(p, "tf8_optdowncvt", 15) == 0)
619 message_utf8_downconvert = -1;
620 break;
621 #endif
622
623 default: /* Present because some compilers complain if all */
624 break; /* possibilities are not covered. */
625 }
626 }
627
628 /* Build sender_fullhost if required */
629
630 #ifndef COMPILE_UTILITY
631 host_build_sender_fullhost();
632 #endif /* COMPILE_UTILITY */
633
634 #ifndef COMPILE_UTILITY
635 DEBUG(D_deliver)
636 debug_printf("sender_local=%d ident=%s\n", sender_local,
637 (sender_ident == NULL)? US"unset" : sender_ident);
638 #endif /* COMPILE_UTILITY */
639
640 /* We now have the tree of addresses NOT to deliver to, or a line
641 containing "XX", indicating no tree. */
642
643 if (Ustrncmp(big_buffer, "XX\n", 3) != 0 &&
644 !read_nonrecipients_tree(&tree_nonrecipients, f, big_buffer, big_buffer_size))
645 goto SPOOL_FORMAT_ERROR;
646
647 #ifndef COMPILE_UTILITY
648 DEBUG(D_deliver)
649 {
650 debug_printf("Non-recipients:\n");
651 debug_print_tree(tree_nonrecipients);
652 }
653 #endif /* COMPILE_UTILITY */
654
655 /* After reading the tree, the next line has not yet been read into the
656 buffer. It contains the count of recipients which follow on separate lines. */
657
658 if (Ufgets(big_buffer, big_buffer_size, f) == NULL) goto SPOOL_READ_ERROR;
659 if (sscanf(CS big_buffer, "%d", &rcount) != 1) goto SPOOL_FORMAT_ERROR;
660
661 #ifndef COMPILE_UTILITY
662 DEBUG(D_deliver) debug_printf("recipients_count=%d\n", rcount);
663 #endif /* COMPILE_UTILITY */
664
665 recipients_list_max = rcount;
666 recipients_list = store_get(rcount * sizeof(recipient_item));
667
668 for (recipients_count = 0; recipients_count < rcount; recipients_count++)
669 {
670 int nn;
671 int pno = -1;
672 int dsn_flags = 0;
673 uschar *orcpt = NULL;
674 uschar *errors_to = NULL;
675 uschar *p;
676
677 if (Ufgets(big_buffer, big_buffer_size, f) == NULL) goto SPOOL_READ_ERROR;
678 nn = Ustrlen(big_buffer);
679 if (nn < 2) goto SPOOL_FORMAT_ERROR;
680
681 /* Remove the newline; this terminates the address if there is no additional
682 data on the line. */
683
684 p = big_buffer + nn - 1;
685 *p-- = 0;
686
687 /* Look back from the end of the line for digits and special terminators.
688 Since an address must end with a domain, we can tell that extra data is
689 present by the presence of the terminator, which is always some character
690 that cannot exist in a domain. (If I'd thought of the need for additional
691 data early on, I'd have put it at the start, with the address at the end. As
692 it is, we have to operate backwards. Addresses are permitted to contain
693 spaces, you see.)
694
695 This code has to cope with various versions of this data that have evolved
696 over time. In all cases, the line might just contain an address, with no
697 additional data. Otherwise, the possibilities are as follows:
698
699 Exim 3 type: <address><space><digits>,<digits>,<digits>
700
701 The second set of digits is the parent number for one_time addresses. The
702 other values were remnants of earlier experiments that were abandoned.
703
704 Exim 4 first type: <address><space><digits>
705
706 The digits are the parent number for one_time addresses.
707
708 Exim 4 new type: <address><space><data>#<type bits>
709
710 The type bits indicate what the contents of the data are.
711
712 Bit 01 indicates that, reading from right to left, the data
713 ends with <errors_to address><space><len>,<pno> where pno is
714 the parent number for one_time addresses, and len is the length
715 of the errors_to address (zero meaning none).
716
717 Bit 02 indicates that, again reading from right to left, the data continues
718 with orcpt len(orcpt),dsn_flags
719 */
720
721 while (isdigit(*p)) p--;
722
723 /* Handle Exim 3 spool files */
724
725 if (*p == ',')
726 {
727 int dummy;
728 while (isdigit(*(--p)) || *p == ',');
729 if (*p == ' ')
730 {
731 *p++ = 0;
732 (void)sscanf(CS p, "%d,%d", &dummy, &pno);
733 }
734 }
735
736 /* Handle early Exim 4 spool files */
737
738 else if (*p == ' ')
739 {
740 *p++ = 0;
741 (void)sscanf(CS p, "%d", &pno);
742 }
743
744 /* Handle current format Exim 4 spool files */
745
746 else if (*p == '#')
747 {
748 int flags;
749
750 #if !defined (COMPILE_UTILITY)
751 DEBUG(D_deliver) debug_printf("**** SPOOL_IN - Exim 4 standard format spoolfile\n");
752 #endif
753
754 (void)sscanf(CS p+1, "%d", &flags);
755
756 if ((flags & 0x01) != 0) /* one_time data exists */
757 {
758 int len;
759 while (isdigit(*(--p)) || *p == ',' || *p == '-');
760 (void)sscanf(CS p+1, "%d,%d", &len, &pno);
761 *p = 0;
762 if (len > 0)
763 {
764 p -= len;
765 errors_to = string_copy(p);
766 }
767 }
768
769 *(--p) = 0; /* Terminate address */
770 if ((flags & 0x02) != 0) /* one_time data exists */
771 {
772 int len;
773 while (isdigit(*(--p)) || *p == ',' || *p == '-');
774 (void)sscanf(CS p+1, "%d,%d", &len, &dsn_flags);
775 *p = 0;
776 if (len > 0)
777 {
778 p -= len;
779 orcpt = string_copy(p);
780 }
781 }
782
783 *(--p) = 0; /* Terminate address */
784 }
785 #if !defined(COMPILE_UTILITY)
786 else
787 { DEBUG(D_deliver) debug_printf("**** SPOOL_IN - No additional fields\n"); }
788
789 if ((orcpt != NULL) || (dsn_flags != 0))
790 {
791 DEBUG(D_deliver) debug_printf("**** SPOOL_IN - address: |%s| orcpt: |%s| dsn_flags: %d\n",
792 big_buffer, orcpt, dsn_flags);
793 }
794 if (errors_to != NULL)
795 {
796 DEBUG(D_deliver) debug_printf("**** SPOOL_IN - address: |%s| errorsto: |%s|\n",
797 big_buffer, errors_to);
798 }
799 #endif
800
801 recipients_list[recipients_count].address = string_copy(big_buffer);
802 recipients_list[recipients_count].pno = pno;
803 recipients_list[recipients_count].errors_to = errors_to;
804 recipients_list[recipients_count].orcpt = orcpt;
805 recipients_list[recipients_count].dsn_flags = dsn_flags;
806 }
807
808 /* The remainder of the spool header file contains the headers for the message,
809 separated off from the previous data by a blank line. Each header is preceded
810 by a count of its length and either a certain letter (for various identified
811 headers), space (for a miscellaneous live header) or an asterisk (for a header
812 that has been rewritten). Count the Received: headers. We read the headers
813 always, in order to check on the format of the file, but only create a header
814 list if requested to do so. */
815
816 inheader = TRUE;
817 if (Ufgets(big_buffer, big_buffer_size, f) == NULL) goto SPOOL_READ_ERROR;
818 if (big_buffer[0] != '\n') goto SPOOL_FORMAT_ERROR;
819
820 while ((n = fgetc(f)) != EOF)
821 {
822 header_line *h;
823 uschar flag[4];
824 int i;
825
826 if (!isdigit(n)) goto SPOOL_FORMAT_ERROR;
827 if(ungetc(n, f) == EOF || fscanf(f, "%d%c ", &n, flag) == EOF)
828 goto SPOOL_READ_ERROR;
829 if (flag[0] != '*') message_size += n; /* Omit non-transmitted headers */
830
831 if (read_headers)
832 {
833 h = store_get(sizeof(header_line));
834 h->next = NULL;
835 h->type = flag[0];
836 h->slen = n;
837 h->text = store_get(n+1);
838
839 if (h->type == htype_received) received_count++;
840
841 if (header_list == NULL) header_list = h;
842 else header_last->next = h;
843 header_last = h;
844
845 for (i = 0; i < n; i++)
846 {
847 int c = fgetc(f);
848 if (c == 0 || c == EOF) goto SPOOL_FORMAT_ERROR;
849 if (c == '\n' && h->type != htype_old) message_linecount++;
850 h->text[i] = c;
851 }
852 h->text[i] = 0;
853 }
854
855 /* Not requiring header data, just skip through the bytes */
856
857 else for (i = 0; i < n; i++)
858 {
859 int c = fgetc(f);
860 if (c == 0 || c == EOF) goto SPOOL_FORMAT_ERROR;
861 }
862 }
863
864 /* We have successfully read the data in the header file. Update the message
865 line count by adding the body linecount to the header linecount. Close the file
866 and give a positive response. */
867
868 #ifndef COMPILE_UTILITY
869 DEBUG(D_deliver) debug_printf("body_linecount=%d message_linecount=%d\n",
870 body_linecount, message_linecount);
871 #endif /* COMPILE_UTILITY */
872
873 message_linecount += body_linecount;
874
875 fclose(f);
876 return spool_read_OK;
877
878
879 /* There was an error reading the spool or there was missing data,
880 or there was a format error. A "read error" with no errno means an
881 unexpected EOF, which we treat as a format error. */
882
883 SPOOL_READ_ERROR:
884 if (errno != 0)
885 {
886 n = errno;
887
888 #ifndef COMPILE_UTILITY
889 DEBUG(D_any) debug_printf("Error while reading spool file %s\n", name);
890 #endif /* COMPILE_UTILITY */
891
892 fclose(f);
893 errno = n;
894 return inheader? spool_read_hdrerror : spool_read_enverror;
895 }
896
897 SPOOL_FORMAT_ERROR:
898
899 #ifndef COMPILE_UTILITY
900 DEBUG(D_any) debug_printf("Format error in spool file %s\n", name);
901 #endif /* COMPILE_UTILITY */
902
903 fclose(f);
904 errno = ERRNO_SPOOLFORMAT;
905 return inheader? spool_read_hdrerror : spool_read_enverror;
906 }
907
908 /* vi: aw ai sw=2
909 */
910 /* End of spool_in.c */