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]; | |
059ec3d9 PH |
154 | uschar *buffer = malloc(max_outsize); |
155 | uschar *line = malloc(max_insize); | |
156 | ||
157 | while (argc > 1) | |
158 | { | |
159 | if (Ustrcmp(argv[arg], "-nolc") == 0) lowercase = FALSE; | |
160 | else if (Ustrcmp(argv[arg], "-nowarn") == 0) warn = FALSE; | |
161 | else if (Ustrcmp(argv[arg], "-lastdup") == 0) lastdup = TRUE; | |
162 | else if (Ustrcmp(argv[arg], "-noduperr") == 0) duperr = FALSE; | |
163 | else if (Ustrcmp(argv[arg], "-nozero") == 0) add_zero = 0; | |
164 | else break; | |
165 | arg++; | |
166 | argc--; | |
167 | } | |
168 | ||
169 | if (argc != 3) | |
170 | { | |
171 | printf("usage: exim_dbmbuild [-nolc] <source file> <dbm base name>\n"); | |
172 | exit(1); | |
173 | } | |
174 | ||
175 | if (Ustrcmp(argv[arg], "-") == 0) f = stdin; else | |
176 | { | |
177 | f = fopen(argv[arg], "rb"); | |
178 | if (f == NULL) | |
179 | { | |
180 | printf("exim_dbmbuild: unable to open %s: %s\n", argv[arg], strerror(errno)); | |
181 | exit(1); | |
182 | } | |
183 | } | |
184 | ||
185 | /* By default Berkeley db does not put extensions on... which | |
186 | can be painful! */ | |
187 | ||
188 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) | |
189 | if (Ustrcmp(argv[arg], argv[arg+1]) == 0) | |
190 | { | |
191 | printf("exim_dbmbuild: input and output filenames are the same\n"); | |
192 | exit(1); | |
193 | } | |
194 | #endif | |
195 | ||
a0540757 PH |
196 | /* Check length of filename; allow for adding .dbmbuild_temp and .db or |
197 | .dir/.pag later. */ | |
198 | ||
199 | if (strlen(argv[arg+1]) > sizeof(temp_dbmname) - 20) | |
200 | { | |
201 | printf("exim_dbmbuild: output filename is ridiculously long\n"); | |
202 | exit(1); | |
203 | } | |
204 | ||
059ec3d9 PH |
205 | Ustrcpy(temp_dbmname, argv[arg+1]); |
206 | Ustrcat(temp_dbmname, ".dbmbuild_temp"); | |
207 | ||
208 | /* It is apparently necessary to open with O_RDWR for this to work | |
209 | with gdbm-1.7.3, though no reading is actually going to be done. */ | |
210 | ||
211 | EXIM_DBOPEN(temp_dbmname, O_RDWR|O_CREAT|O_EXCL, 0644, &d); | |
212 | ||
213 | if (d == NULL) | |
214 | { | |
215 | printf("exim_dbmbuild: unable to create %s: %s\n", temp_dbmname, | |
216 | strerror(errno)); | |
f1e894f3 | 217 | (void)fclose(f); |
059ec3d9 PH |
218 | exit(1); |
219 | } | |
220 | ||
221 | /* Unless using native db calls, see if we have created <name>.db; if not, | |
222 | assume .dir & .pag */ | |
223 | ||
224 | #if !defined(USE_DB) && !defined(USE_TDB) && !defined(USE_GDBM) | |
225 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
226 | is_db = Ustat(real_dbmname, &statbuf) == 0; | |
227 | #endif | |
228 | ||
229 | /* Now do the business */ | |
230 | ||
231 | bptr = buffer; | |
232 | started = 0; | |
233 | ||
234 | while (Ufgets(line, max_insize, f) != NULL) | |
235 | { | |
236 | uschar *p; | |
237 | int len = Ustrlen(line); | |
238 | ||
239 | p = line + len; | |
240 | ||
241 | if (len >= max_insize - 1 && p[-1] != '\n') | |
242 | { | |
243 | printf("Overlong line read: max permitted length is %d\n", max_insize - 1); | |
aa2b5c79 PH |
244 | yield = 2; |
245 | goto TIDYUP; | |
059ec3d9 PH |
246 | } |
247 | ||
248 | if (line[0] == '#') continue; | |
249 | while (p > line && isspace(p[-1])) p--; | |
250 | *p = 0; | |
251 | if (line[0] == 0) continue; | |
252 | ||
253 | /* A continuation line is valid only if there was a previous first | |
254 | line. */ | |
255 | ||
256 | if (isspace(line[0])) | |
257 | { | |
258 | uschar *s = line; | |
259 | if (!started) | |
260 | { | |
261 | printf("Unexpected continuation line ignored\n%s\n\n", line); | |
262 | continue; | |
263 | } | |
264 | while (isspace(*s)) s++; | |
265 | *(--s) = ' '; | |
266 | ||
267 | if (bptr - buffer + p - s >= max_outsize - 1) | |
268 | { | |
269 | printf("Continued set of lines is too long: max permitted length is %d\n", | |
270 | max_outsize -1); | |
aa2b5c79 PH |
271 | yield = 2; |
272 | goto TIDYUP; | |
059ec3d9 PH |
273 | } |
274 | ||
275 | Ustrcpy(bptr, s); | |
276 | bptr += p - s; | |
277 | } | |
278 | ||
279 | /* A first line must have a name followed by a colon or whitespace or | |
280 | end of line, but first finish with a previous line. The key is lower | |
281 | cased by default - this is what the newaliases program for sendmail does. | |
282 | However, there's an option not to do this. */ | |
283 | ||
284 | else | |
285 | { | |
286 | int i, rc; | |
287 | uschar *s = line; | |
288 | uschar *keystart; | |
289 | ||
290 | if (started) | |
291 | { | |
292 | EXIM_DATUM_INIT(content); | |
293 | EXIM_DATUM_DATA(content) = CS buffer; | |
294 | EXIM_DATUM_SIZE(content) = bptr - buffer + add_zero; | |
295 | ||
296 | switch(rc = EXIM_DBPUTB(d, key, content)) | |
297 | { | |
298 | case EXIM_DBPUTB_OK: | |
299 | count++; | |
300 | break; | |
301 | ||
302 | case EXIM_DBPUTB_DUP: | |
303 | if (warn) fprintf(stderr, "** Duplicate key \"%s\"\n", | |
304 | keybuffer); | |
305 | dupcount++; | |
306 | if(duperr) yield = 1; | |
307 | if (lastdup) EXIM_DBPUT(d, key, content); | |
308 | break; | |
309 | ||
310 | default: | |
311 | fprintf(stderr, "Error %d while writing key %s: errno=%d\n", rc, | |
312 | keybuffer, errno); | |
313 | yield = 2; | |
314 | goto TIDYUP; | |
315 | } | |
316 | ||
317 | bptr = buffer; | |
318 | } | |
319 | ||
320 | EXIM_DATUM_INIT(key); | |
321 | EXIM_DATUM_DATA(key) = CS keybuffer; | |
322 | ||
323 | /* Deal with quoted keys. Escape sequences always make one character | |
324 | out of several, so we can re-build in place. */ | |
325 | ||
326 | if (*s == '\"') | |
327 | { | |
328 | uschar *t = s++; | |
329 | keystart = t; | |
330 | while (*s != 0 && *s != '\"') | |
331 | { | |
55414b25 | 332 | if (*s == '\\') *t++ = string_interpret_escape((const uschar **)&s); |
059ec3d9 PH |
333 | else *t++ = *s; |
334 | s++; | |
335 | } | |
336 | if (*s != 0) s++; /* Past terminating " */ | |
337 | EXIM_DATUM_SIZE(key) = t - keystart + add_zero; | |
338 | } | |
339 | else | |
340 | { | |
341 | keystart = s; | |
342 | while (*s != 0 && *s != ':' && !isspace(*s)) s++; | |
343 | EXIM_DATUM_SIZE(key) = s - keystart + add_zero; | |
344 | } | |
345 | ||
346 | if (EXIM_DATUM_SIZE(key) > 256) | |
347 | { | |
348 | printf("Keys longer than 255 characters cannot be handled\n"); | |
349 | started = 0; | |
350 | yield = 2; | |
351 | goto TIDYUP; | |
352 | } | |
353 | ||
354 | if (lowercase) | |
355 | { | |
356 | for (i = 0; i < EXIM_DATUM_SIZE(key) - add_zero; i++) | |
357 | keybuffer[i] = tolower(keystart[i]); | |
358 | } | |
359 | else | |
360 | { | |
361 | for (i = 0; i < EXIM_DATUM_SIZE(key) - add_zero; i++) | |
362 | keybuffer[i] = keystart[i]; | |
363 | } | |
364 | ||
365 | keybuffer[i] = 0; | |
366 | started = 1; | |
367 | ||
368 | while (isspace(*s))s++; | |
369 | if (*s == ':') | |
370 | { | |
371 | s++; | |
372 | while (isspace(*s))s++; | |
373 | } | |
374 | if (*s != 0) | |
375 | { | |
376 | Ustrcpy(bptr, s); | |
377 | bptr += p - s; | |
378 | } | |
379 | else buffer[0] = 0; | |
380 | } | |
381 | } | |
382 | ||
383 | if (started) | |
384 | { | |
385 | int rc; | |
386 | EXIM_DATUM_INIT(content); | |
387 | EXIM_DATUM_DATA(content) = CS buffer; | |
388 | EXIM_DATUM_SIZE(content) = bptr - buffer + add_zero; | |
389 | ||
390 | switch(rc = EXIM_DBPUTB(d, key, content)) | |
391 | { | |
392 | case EXIM_DBPUTB_OK: | |
393 | count++; | |
394 | break; | |
395 | ||
396 | case EXIM_DBPUTB_DUP: | |
397 | if (warn) fprintf(stderr, "** Duplicate key \"%s\"\n", keybuffer); | |
398 | dupcount++; | |
399 | if (duperr) yield = 1; | |
400 | if (lastdup) EXIM_DBPUT(d, key, content); | |
401 | break; | |
402 | ||
403 | default: | |
404 | fprintf(stderr, "Error %d while writing key %s: errno=%d\n", rc, | |
405 | keybuffer, errno); | |
406 | yield = 2; | |
407 | break; | |
408 | } | |
409 | } | |
410 | ||
411 | /* Close files, rename or abandon the temporary files, and exit */ | |
412 | ||
413 | TIDYUP: | |
414 | ||
415 | EXIM_DBCLOSE(d); | |
f1e894f3 | 416 | (void)fclose(f); |
059ec3d9 PH |
417 | |
418 | /* If successful, output the number of entries and rename the temporary | |
419 | files. */ | |
420 | ||
421 | if (yield == 0 || yield == 1) | |
422 | { | |
423 | printf("%d entr%s written\n", count, (count == 1)? "y" : "ies"); | |
424 | if (dupcount > 0) | |
425 | { | |
426 | printf("%d duplicate key%s \n", dupcount, (dupcount > 1)? "s" : ""); | |
427 | } | |
428 | ||
429 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) | |
430 | Ustrcpy(real_dbmname, temp_dbmname); | |
431 | Ustrcpy(buffer, argv[arg+1]); | |
432 | if (Urename(real_dbmname, buffer) != 0) | |
433 | { | |
434 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
435 | return 1; | |
436 | } | |
437 | #else | |
438 | ||
439 | /* Rename a single .db file */ | |
440 | ||
441 | if (is_db) | |
442 | { | |
443 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
444 | sprintf(CS buffer, "%s.db", argv[arg+1]); | |
445 | if (Urename(real_dbmname, buffer) != 0) | |
446 | { | |
447 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
448 | return 1; | |
449 | } | |
450 | } | |
451 | ||
452 | /* Rename .dir and .pag files */ | |
453 | ||
454 | else | |
455 | { | |
456 | sprintf(CS real_dbmname, "%s.dir", temp_dbmname); | |
457 | sprintf(CS buffer, "%s.dir", argv[arg+1]); | |
458 | if (Urename(real_dbmname, buffer) != 0) | |
459 | { | |
460 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
461 | return 1; | |
462 | } | |
463 | ||
464 | sprintf(CS real_dbmname, "%s.pag", temp_dbmname); | |
465 | sprintf(CS buffer, "%s.pag", argv[arg+1]); | |
466 | if (Urename(real_dbmname, buffer) != 0) | |
467 | { | |
468 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
469 | return 1; | |
470 | } | |
471 | } | |
472 | ||
473 | #endif /* USE_DB || USE_TDB || USE_GDBM */ | |
474 | } | |
475 | ||
476 | /* Otherwise unlink the temporary files. */ | |
477 | ||
478 | else | |
479 | { | |
480 | printf("dbmbuild abandoned\n"); | |
481 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) | |
482 | Uunlink(temp_dbmname); | |
483 | #else | |
484 | if (is_db) | |
485 | { | |
486 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
487 | Uunlink(real_dbmname); | |
488 | } | |
489 | else | |
490 | { | |
491 | sprintf(CS real_dbmname, "%s.dir", temp_dbmname); | |
492 | Uunlink(real_dbmname); | |
493 | sprintf(CS real_dbmname, "%s.pag", temp_dbmname); | |
494 | Uunlink(real_dbmname); | |
495 | } | |
496 | #endif /* USE_DB || USE_TDB */ | |
497 | } | |
498 | ||
499 | return yield; | |
500 | } | |
501 | ||
502 | /* End of exim_dbmbuild.c */ |