Commit | Line | Data |
---|---|---|
059ec3d9 PH |
1 | /************************************************* |
2 | * Exim - an Internet mail transport agent * | |
3 | *************************************************/ | |
4 | ||
3386088d | 5 | /* Copyright (c) University of Cambridge 1995 - 2015 */ |
059ec3d9 PH |
6 | /* See the file NOTICE for conditions of use and distribution. */ |
7 | ||
8 | ||
9 | /* A small freestanding program to build dbm databases from serial input. For | |
10 | alias files, this program fulfils the function of the newaliases program used | |
11 | by other mailers, but it can be used for other dbm data files too. It operates | |
12 | by writing a new file or files, and then renaming; otherwise old entries can | |
13 | never get flushed out. | |
14 | ||
15 | This program is clever enough to cope with ndbm, which creates two files called | |
16 | <name>.dir and <name>.pag, or with db, which creates a single file called | |
17 | <name>.db. If native db is in use (USE_DB defined) or tdb is in use (USE_TDB | |
18 | defined) there is no extension to the output filename. This is also handled. If | |
19 | there are any other variants, the program won't cope. | |
20 | ||
21 | The first argument to the program is the name of the serial file; the second | |
22 | is the base name for the DBM file(s). When native db is in use, these must be | |
23 | different. | |
24 | ||
25 | Input lines beginning with # are ignored, as are blank lines. Entries begin | |
26 | with a key terminated by a colon or end of line or whitespace and continue with | |
27 | indented lines. Keys may be quoted if they contain colons or whitespace or # | |
28 | characters. */ | |
29 | ||
30 | ||
31 | #include "exim.h" | |
32 | ||
33 | ||
34 | #define max_insize 20000 | |
35 | #define max_outsize 100000 | |
36 | ||
37 | /* This is global because it's defined in the headers and compilers grumble | |
38 | if it is made static. */ | |
39 | ||
1ba28e2b | 40 | const uschar *hex_digits = CUS"0123456789abcdef"; |
059ec3d9 PH |
41 | |
42 | ||
43 | #ifdef STRERROR_FROM_ERRLIST | |
44 | /* Some old-fashioned systems still around (e.g. SunOS4) don't have strerror() | |
45 | in their libraries, but can provide the same facility by this simple | |
46 | alternative function. */ | |
47 | ||
48 | char * | |
49 | strerror(int n) | |
50 | { | |
51 | if (n < 0 || n >= sys_nerr) return "unknown error number"; | |
52 | return sys_errlist[n]; | |
53 | } | |
54 | #endif /* STRERROR_FROM_ERRLIST */ | |
55 | ||
56 | ||
57 | /* For Berkeley DB >= 2, we can define a function to be called in case of DB | |
58 | errors. This should help with debugging strange DB problems, e.g. getting "File | |
1f922db1 | 59 | exists" when you try to open a db file. The API changed at release 4.3. */ |
059ec3d9 PH |
60 | |
61 | #if defined(USE_DB) && defined(DB_VERSION_STRING) | |
62 | void | |
1f922db1 PH |
63 | #if DB_VERSION_MAJOR > 4 || (DB_VERSION_MAJOR == 4 && DB_VERSION_MINOR >= 3) |
64 | dbfn_bdb_error_callback(const DB_ENV *dbenv, const char *pfx, const char *msg) | |
65 | { | |
66 | dbenv = dbenv; | |
67 | #else | |
059ec3d9 PH |
68 | dbfn_bdb_error_callback(const char *pfx, char *msg) |
69 | { | |
1f922db1 | 70 | #endif |
059ec3d9 PH |
71 | pfx = pfx; |
72 | printf("Berkeley DB error: %s\n", msg); | |
73 | } | |
74 | #endif | |
75 | ||
76 | ||
77 | ||
78 | /************************************************* | |
79 | * Interpret escape sequence * | |
80 | *************************************************/ | |
81 | ||
82 | /* This function is copied from the main Exim code. | |
83 | ||
84 | Arguments: | |
85 | pp points a pointer to the initiating "\" in the string; | |
86 | the pointer gets updated to point to the final character | |
87 | Returns: the value of the character escape | |
88 | */ | |
89 | ||
90 | int | |
55414b25 | 91 | string_interpret_escape(const uschar **pp) |
059ec3d9 PH |
92 | { |
93 | int ch; | |
55414b25 | 94 | const uschar *p = *pp; |
059ec3d9 PH |
95 | ch = *(++p); |
96 | if (isdigit(ch) && ch != '8' && ch != '9') | |
97 | { | |
98 | ch -= '0'; | |
99 | if (isdigit(p[1]) && p[1] != '8' && p[1] != '9') | |
100 | { | |
101 | ch = ch * 8 + *(++p) - '0'; | |
102 | if (isdigit(p[1]) && p[1] != '8' && p[1] != '9') | |
103 | ch = ch * 8 + *(++p) - '0'; | |
104 | } | |
105 | } | |
106 | else switch(ch) | |
107 | { | |
108 | case 'n': ch = '\n'; break; | |
109 | case 'r': ch = '\r'; break; | |
110 | case 't': ch = '\t'; break; | |
111 | case 'x': | |
112 | ch = 0; | |
113 | if (isxdigit(p[1])) | |
114 | { | |
115 | ch = ch * 16 + | |
116 | Ustrchr(hex_digits, tolower(*(++p))) - hex_digits; | |
117 | if (isxdigit(p[1])) ch = ch * 16 + | |
118 | Ustrchr(hex_digits, tolower(*(++p))) - hex_digits; | |
119 | } | |
120 | break; | |
121 | } | |
122 | *pp = p; | |
123 | return ch; | |
124 | } | |
125 | ||
126 | ||
127 | /************************************************* | |
128 | * Main Program * | |
129 | *************************************************/ | |
130 | ||
131 | int main(int argc, char **argv) | |
132 | { | |
133 | int started; | |
134 | int count = 0; | |
135 | int dupcount = 0; | |
136 | int yield = 0; | |
137 | int arg = 1; | |
138 | int add_zero = 1; | |
139 | BOOL lowercase = TRUE; | |
140 | BOOL warn = TRUE; | |
141 | BOOL duperr = TRUE; | |
142 | BOOL lastdup = FALSE; | |
143 | #if !defined (USE_DB) && !defined(USE_TDB) && !defined(USE_GDBM) | |
144 | int is_db = 0; | |
145 | struct stat statbuf; | |
146 | #endif | |
147 | FILE *f; | |
148 | EXIM_DB *d; | |
149 | EXIM_DATUM key, content; | |
150 | uschar *bptr; | |
151 | uschar keybuffer[256]; | |
a0540757 PH |
152 | uschar temp_dbmname[512]; |
153 | uschar real_dbmname[512]; | |
cfb9cf20 | 154 | uschar dirname[512]; |
059ec3d9 PH |
155 | uschar *buffer = malloc(max_outsize); |
156 | uschar *line = malloc(max_insize); | |
157 | ||
158 | while (argc > 1) | |
159 | { | |
160 | if (Ustrcmp(argv[arg], "-nolc") == 0) lowercase = FALSE; | |
161 | else if (Ustrcmp(argv[arg], "-nowarn") == 0) warn = FALSE; | |
162 | else if (Ustrcmp(argv[arg], "-lastdup") == 0) lastdup = TRUE; | |
163 | else if (Ustrcmp(argv[arg], "-noduperr") == 0) duperr = FALSE; | |
164 | else if (Ustrcmp(argv[arg], "-nozero") == 0) add_zero = 0; | |
165 | else break; | |
166 | arg++; | |
167 | argc--; | |
168 | } | |
169 | ||
170 | if (argc != 3) | |
171 | { | |
172 | printf("usage: exim_dbmbuild [-nolc] <source file> <dbm base name>\n"); | |
173 | exit(1); | |
174 | } | |
175 | ||
176 | if (Ustrcmp(argv[arg], "-") == 0) f = stdin; else | |
177 | { | |
178 | f = fopen(argv[arg], "rb"); | |
179 | if (f == NULL) | |
180 | { | |
181 | printf("exim_dbmbuild: unable to open %s: %s\n", argv[arg], strerror(errno)); | |
182 | exit(1); | |
183 | } | |
184 | } | |
185 | ||
186 | /* By default Berkeley db does not put extensions on... which | |
187 | can be painful! */ | |
188 | ||
189 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) | |
190 | if (Ustrcmp(argv[arg], argv[arg+1]) == 0) | |
191 | { | |
192 | printf("exim_dbmbuild: input and output filenames are the same\n"); | |
193 | exit(1); | |
194 | } | |
195 | #endif | |
196 | ||
a0540757 PH |
197 | /* Check length of filename; allow for adding .dbmbuild_temp and .db or |
198 | .dir/.pag later. */ | |
199 | ||
200 | if (strlen(argv[arg+1]) > sizeof(temp_dbmname) - 20) | |
201 | { | |
202 | printf("exim_dbmbuild: output filename is ridiculously long\n"); | |
203 | exit(1); | |
204 | } | |
205 | ||
059ec3d9 PH |
206 | Ustrcpy(temp_dbmname, argv[arg+1]); |
207 | Ustrcat(temp_dbmname, ".dbmbuild_temp"); | |
208 | ||
cfb9cf20 JH |
209 | Ustrcpy(dirname, temp_dbmname); |
210 | if ((bptr = Ustrrchr(dirname, '/'))) | |
211 | *bptr = '\0'; | |
212 | ||
059ec3d9 PH |
213 | /* It is apparently necessary to open with O_RDWR for this to work |
214 | with gdbm-1.7.3, though no reading is actually going to be done. */ | |
215 | ||
cfb9cf20 | 216 | EXIM_DBOPEN(temp_dbmname, dirname, O_RDWR|O_CREAT|O_EXCL, 0644, &d); |
059ec3d9 PH |
217 | |
218 | if (d == NULL) | |
219 | { | |
220 | printf("exim_dbmbuild: unable to create %s: %s\n", temp_dbmname, | |
221 | strerror(errno)); | |
f1e894f3 | 222 | (void)fclose(f); |
059ec3d9 PH |
223 | exit(1); |
224 | } | |
225 | ||
226 | /* Unless using native db calls, see if we have created <name>.db; if not, | |
227 | assume .dir & .pag */ | |
228 | ||
229 | #if !defined(USE_DB) && !defined(USE_TDB) && !defined(USE_GDBM) | |
230 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
231 | is_db = Ustat(real_dbmname, &statbuf) == 0; | |
232 | #endif | |
233 | ||
234 | /* Now do the business */ | |
235 | ||
236 | bptr = buffer; | |
237 | started = 0; | |
238 | ||
239 | while (Ufgets(line, max_insize, f) != NULL) | |
240 | { | |
241 | uschar *p; | |
242 | int len = Ustrlen(line); | |
243 | ||
244 | p = line + len; | |
245 | ||
246 | if (len >= max_insize - 1 && p[-1] != '\n') | |
247 | { | |
248 | printf("Overlong line read: max permitted length is %d\n", max_insize - 1); | |
aa2b5c79 PH |
249 | yield = 2; |
250 | goto TIDYUP; | |
059ec3d9 PH |
251 | } |
252 | ||
253 | if (line[0] == '#') continue; | |
254 | while (p > line && isspace(p[-1])) p--; | |
255 | *p = 0; | |
256 | if (line[0] == 0) continue; | |
257 | ||
258 | /* A continuation line is valid only if there was a previous first | |
259 | line. */ | |
260 | ||
261 | if (isspace(line[0])) | |
262 | { | |
263 | uschar *s = line; | |
264 | if (!started) | |
265 | { | |
266 | printf("Unexpected continuation line ignored\n%s\n\n", line); | |
267 | continue; | |
268 | } | |
269 | while (isspace(*s)) s++; | |
270 | *(--s) = ' '; | |
271 | ||
272 | if (bptr - buffer + p - s >= max_outsize - 1) | |
273 | { | |
274 | printf("Continued set of lines is too long: max permitted length is %d\n", | |
275 | max_outsize -1); | |
aa2b5c79 PH |
276 | yield = 2; |
277 | goto TIDYUP; | |
059ec3d9 PH |
278 | } |
279 | ||
280 | Ustrcpy(bptr, s); | |
281 | bptr += p - s; | |
282 | } | |
283 | ||
284 | /* A first line must have a name followed by a colon or whitespace or | |
285 | end of line, but first finish with a previous line. The key is lower | |
286 | cased by default - this is what the newaliases program for sendmail does. | |
287 | However, there's an option not to do this. */ | |
288 | ||
289 | else | |
290 | { | |
291 | int i, rc; | |
292 | uschar *s = line; | |
293 | uschar *keystart; | |
294 | ||
295 | if (started) | |
296 | { | |
297 | EXIM_DATUM_INIT(content); | |
298 | EXIM_DATUM_DATA(content) = CS buffer; | |
299 | EXIM_DATUM_SIZE(content) = bptr - buffer + add_zero; | |
300 | ||
301 | switch(rc = EXIM_DBPUTB(d, key, content)) | |
302 | { | |
303 | case EXIM_DBPUTB_OK: | |
304 | count++; | |
305 | break; | |
306 | ||
307 | case EXIM_DBPUTB_DUP: | |
308 | if (warn) fprintf(stderr, "** Duplicate key \"%s\"\n", | |
309 | keybuffer); | |
310 | dupcount++; | |
311 | if(duperr) yield = 1; | |
312 | if (lastdup) EXIM_DBPUT(d, key, content); | |
313 | break; | |
314 | ||
315 | default: | |
316 | fprintf(stderr, "Error %d while writing key %s: errno=%d\n", rc, | |
317 | keybuffer, errno); | |
318 | yield = 2; | |
319 | goto TIDYUP; | |
320 | } | |
321 | ||
322 | bptr = buffer; | |
323 | } | |
324 | ||
325 | EXIM_DATUM_INIT(key); | |
326 | EXIM_DATUM_DATA(key) = CS keybuffer; | |
327 | ||
328 | /* Deal with quoted keys. Escape sequences always make one character | |
329 | out of several, so we can re-build in place. */ | |
330 | ||
331 | if (*s == '\"') | |
332 | { | |
333 | uschar *t = s++; | |
334 | keystart = t; | |
335 | while (*s != 0 && *s != '\"') | |
336 | { | |
55414b25 | 337 | if (*s == '\\') *t++ = string_interpret_escape((const uschar **)&s); |
059ec3d9 PH |
338 | else *t++ = *s; |
339 | s++; | |
340 | } | |
341 | if (*s != 0) s++; /* Past terminating " */ | |
342 | EXIM_DATUM_SIZE(key) = t - keystart + add_zero; | |
343 | } | |
344 | else | |
345 | { | |
346 | keystart = s; | |
347 | while (*s != 0 && *s != ':' && !isspace(*s)) s++; | |
348 | EXIM_DATUM_SIZE(key) = s - keystart + add_zero; | |
349 | } | |
350 | ||
351 | if (EXIM_DATUM_SIZE(key) > 256) | |
352 | { | |
353 | printf("Keys longer than 255 characters cannot be handled\n"); | |
354 | started = 0; | |
355 | yield = 2; | |
356 | goto TIDYUP; | |
357 | } | |
358 | ||
359 | if (lowercase) | |
360 | { | |
361 | for (i = 0; i < EXIM_DATUM_SIZE(key) - add_zero; i++) | |
362 | keybuffer[i] = tolower(keystart[i]); | |
363 | } | |
364 | else | |
365 | { | |
366 | for (i = 0; i < EXIM_DATUM_SIZE(key) - add_zero; i++) | |
367 | keybuffer[i] = keystart[i]; | |
368 | } | |
369 | ||
370 | keybuffer[i] = 0; | |
371 | started = 1; | |
372 | ||
373 | while (isspace(*s))s++; | |
374 | if (*s == ':') | |
375 | { | |
376 | s++; | |
377 | while (isspace(*s))s++; | |
378 | } | |
379 | if (*s != 0) | |
380 | { | |
381 | Ustrcpy(bptr, s); | |
382 | bptr += p - s; | |
383 | } | |
384 | else buffer[0] = 0; | |
385 | } | |
386 | } | |
387 | ||
388 | if (started) | |
389 | { | |
390 | int rc; | |
391 | EXIM_DATUM_INIT(content); | |
392 | EXIM_DATUM_DATA(content) = CS buffer; | |
393 | EXIM_DATUM_SIZE(content) = bptr - buffer + add_zero; | |
394 | ||
395 | switch(rc = EXIM_DBPUTB(d, key, content)) | |
396 | { | |
397 | case EXIM_DBPUTB_OK: | |
398 | count++; | |
399 | break; | |
400 | ||
401 | case EXIM_DBPUTB_DUP: | |
402 | if (warn) fprintf(stderr, "** Duplicate key \"%s\"\n", keybuffer); | |
403 | dupcount++; | |
404 | if (duperr) yield = 1; | |
405 | if (lastdup) EXIM_DBPUT(d, key, content); | |
406 | break; | |
407 | ||
408 | default: | |
409 | fprintf(stderr, "Error %d while writing key %s: errno=%d\n", rc, | |
410 | keybuffer, errno); | |
411 | yield = 2; | |
412 | break; | |
413 | } | |
414 | } | |
415 | ||
416 | /* Close files, rename or abandon the temporary files, and exit */ | |
417 | ||
418 | TIDYUP: | |
419 | ||
420 | EXIM_DBCLOSE(d); | |
f1e894f3 | 421 | (void)fclose(f); |
059ec3d9 PH |
422 | |
423 | /* If successful, output the number of entries and rename the temporary | |
424 | files. */ | |
425 | ||
426 | if (yield == 0 || yield == 1) | |
427 | { | |
428 | printf("%d entr%s written\n", count, (count == 1)? "y" : "ies"); | |
429 | if (dupcount > 0) | |
430 | { | |
431 | printf("%d duplicate key%s \n", dupcount, (dupcount > 1)? "s" : ""); | |
432 | } | |
433 | ||
434 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) | |
435 | Ustrcpy(real_dbmname, temp_dbmname); | |
436 | Ustrcpy(buffer, argv[arg+1]); | |
437 | if (Urename(real_dbmname, buffer) != 0) | |
438 | { | |
439 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
440 | return 1; | |
441 | } | |
442 | #else | |
443 | ||
444 | /* Rename a single .db file */ | |
445 | ||
446 | if (is_db) | |
447 | { | |
448 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
449 | sprintf(CS buffer, "%s.db", argv[arg+1]); | |
450 | if (Urename(real_dbmname, buffer) != 0) | |
451 | { | |
452 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
453 | return 1; | |
454 | } | |
455 | } | |
456 | ||
457 | /* Rename .dir and .pag files */ | |
458 | ||
459 | else | |
460 | { | |
461 | sprintf(CS real_dbmname, "%s.dir", temp_dbmname); | |
462 | sprintf(CS buffer, "%s.dir", argv[arg+1]); | |
463 | if (Urename(real_dbmname, buffer) != 0) | |
464 | { | |
465 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
466 | return 1; | |
467 | } | |
468 | ||
469 | sprintf(CS real_dbmname, "%s.pag", temp_dbmname); | |
470 | sprintf(CS buffer, "%s.pag", argv[arg+1]); | |
471 | if (Urename(real_dbmname, buffer) != 0) | |
472 | { | |
473 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
474 | return 1; | |
475 | } | |
476 | } | |
477 | ||
478 | #endif /* USE_DB || USE_TDB || USE_GDBM */ | |
479 | } | |
480 | ||
481 | /* Otherwise unlink the temporary files. */ | |
482 | ||
483 | else | |
484 | { | |
485 | printf("dbmbuild abandoned\n"); | |
d88f0784 JH |
486 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) |
487 | /* We created it, so safe to delete despite the name coming from outside */ | |
6e92b3ae | 488 | /* coverity[tainted_string] */ |
059ec3d9 | 489 | Uunlink(temp_dbmname); |
d88f0784 | 490 | #else |
059ec3d9 PH |
491 | if (is_db) |
492 | { | |
493 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
494 | Uunlink(real_dbmname); | |
495 | } | |
496 | else | |
497 | { | |
498 | sprintf(CS real_dbmname, "%s.dir", temp_dbmname); | |
499 | Uunlink(real_dbmname); | |
500 | sprintf(CS real_dbmname, "%s.pag", temp_dbmname); | |
501 | Uunlink(real_dbmname); | |
502 | } | |
d88f0784 | 503 | #endif /* USE_DB || USE_TDB */ |
059ec3d9 PH |
504 | } |
505 | ||
506 | return yield; | |
507 | } | |
508 | ||
509 | /* End of exim_dbmbuild.c */ |