Commit | Line | Data |
---|---|---|
184e8823 | 1 | /* $Cambridge: exim/src/src/exim_dbmbuild.c,v 1.8 2007/01/08 10:50:18 ph10 Exp $ */ |
059ec3d9 PH |
2 | |
3 | /************************************************* | |
4 | * Exim - an Internet mail transport agent * | |
5 | *************************************************/ | |
6 | ||
184e8823 | 7 | /* Copyright (c) University of Cambridge 1995 - 2007 */ |
059ec3d9 PH |
8 | /* See the file NOTICE for conditions of use and distribution. */ |
9 | ||
10 | ||
11 | /* A small freestanding program to build dbm databases from serial input. For | |
12 | alias files, this program fulfils the function of the newaliases program used | |
13 | by other mailers, but it can be used for other dbm data files too. It operates | |
14 | by writing a new file or files, and then renaming; otherwise old entries can | |
15 | never get flushed out. | |
16 | ||
17 | This program is clever enough to cope with ndbm, which creates two files called | |
18 | <name>.dir and <name>.pag, or with db, which creates a single file called | |
19 | <name>.db. If native db is in use (USE_DB defined) or tdb is in use (USE_TDB | |
20 | defined) there is no extension to the output filename. This is also handled. If | |
21 | there are any other variants, the program won't cope. | |
22 | ||
23 | The first argument to the program is the name of the serial file; the second | |
24 | is the base name for the DBM file(s). When native db is in use, these must be | |
25 | different. | |
26 | ||
27 | Input lines beginning with # are ignored, as are blank lines. Entries begin | |
28 | with a key terminated by a colon or end of line or whitespace and continue with | |
29 | indented lines. Keys may be quoted if they contain colons or whitespace or # | |
30 | characters. */ | |
31 | ||
32 | ||
33 | #include "exim.h" | |
34 | ||
35 | ||
36 | #define max_insize 20000 | |
37 | #define max_outsize 100000 | |
38 | ||
39 | /* This is global because it's defined in the headers and compilers grumble | |
40 | if it is made static. */ | |
41 | ||
42 | uschar *hex_digits = US"0123456789abcdef"; | |
43 | ||
44 | ||
45 | #ifdef STRERROR_FROM_ERRLIST | |
46 | /* Some old-fashioned systems still around (e.g. SunOS4) don't have strerror() | |
47 | in their libraries, but can provide the same facility by this simple | |
48 | alternative function. */ | |
49 | ||
50 | char * | |
51 | strerror(int n) | |
52 | { | |
53 | if (n < 0 || n >= sys_nerr) return "unknown error number"; | |
54 | return sys_errlist[n]; | |
55 | } | |
56 | #endif /* STRERROR_FROM_ERRLIST */ | |
57 | ||
58 | ||
59 | /* For Berkeley DB >= 2, we can define a function to be called in case of DB | |
60 | errors. This should help with debugging strange DB problems, e.g. getting "File | |
1f922db1 | 61 | exists" when you try to open a db file. The API changed at release 4.3. */ |
059ec3d9 PH |
62 | |
63 | #if defined(USE_DB) && defined(DB_VERSION_STRING) | |
64 | void | |
1f922db1 PH |
65 | #if DB_VERSION_MAJOR > 4 || (DB_VERSION_MAJOR == 4 && DB_VERSION_MINOR >= 3) |
66 | dbfn_bdb_error_callback(const DB_ENV *dbenv, const char *pfx, const char *msg) | |
67 | { | |
68 | dbenv = dbenv; | |
69 | #else | |
059ec3d9 PH |
70 | dbfn_bdb_error_callback(const char *pfx, char *msg) |
71 | { | |
1f922db1 | 72 | #endif |
059ec3d9 PH |
73 | pfx = pfx; |
74 | printf("Berkeley DB error: %s\n", msg); | |
75 | } | |
76 | #endif | |
77 | ||
78 | ||
79 | ||
80 | /************************************************* | |
81 | * Interpret escape sequence * | |
82 | *************************************************/ | |
83 | ||
84 | /* This function is copied from the main Exim code. | |
85 | ||
86 | Arguments: | |
87 | pp points a pointer to the initiating "\" in the string; | |
88 | the pointer gets updated to point to the final character | |
89 | Returns: the value of the character escape | |
90 | */ | |
91 | ||
92 | int | |
93 | string_interpret_escape(uschar **pp) | |
94 | { | |
95 | int ch; | |
96 | uschar *p = *pp; | |
97 | ch = *(++p); | |
98 | if (isdigit(ch) && ch != '8' && ch != '9') | |
99 | { | |
100 | ch -= '0'; | |
101 | if (isdigit(p[1]) && p[1] != '8' && p[1] != '9') | |
102 | { | |
103 | ch = ch * 8 + *(++p) - '0'; | |
104 | if (isdigit(p[1]) && p[1] != '8' && p[1] != '9') | |
105 | ch = ch * 8 + *(++p) - '0'; | |
106 | } | |
107 | } | |
108 | else switch(ch) | |
109 | { | |
110 | case 'n': ch = '\n'; break; | |
111 | case 'r': ch = '\r'; break; | |
112 | case 't': ch = '\t'; break; | |
113 | case 'x': | |
114 | ch = 0; | |
115 | if (isxdigit(p[1])) | |
116 | { | |
117 | ch = ch * 16 + | |
118 | Ustrchr(hex_digits, tolower(*(++p))) - hex_digits; | |
119 | if (isxdigit(p[1])) ch = ch * 16 + | |
120 | Ustrchr(hex_digits, tolower(*(++p))) - hex_digits; | |
121 | } | |
122 | break; | |
123 | } | |
124 | *pp = p; | |
125 | return ch; | |
126 | } | |
127 | ||
128 | ||
129 | /************************************************* | |
130 | * Main Program * | |
131 | *************************************************/ | |
132 | ||
133 | int main(int argc, char **argv) | |
134 | { | |
135 | int started; | |
136 | int count = 0; | |
137 | int dupcount = 0; | |
138 | int yield = 0; | |
139 | int arg = 1; | |
140 | int add_zero = 1; | |
141 | BOOL lowercase = TRUE; | |
142 | BOOL warn = TRUE; | |
143 | BOOL duperr = TRUE; | |
144 | BOOL lastdup = FALSE; | |
145 | #if !defined (USE_DB) && !defined(USE_TDB) && !defined(USE_GDBM) | |
146 | int is_db = 0; | |
147 | struct stat statbuf; | |
148 | #endif | |
149 | FILE *f; | |
150 | EXIM_DB *d; | |
151 | EXIM_DATUM key, content; | |
152 | uschar *bptr; | |
153 | uschar keybuffer[256]; | |
a0540757 PH |
154 | uschar temp_dbmname[512]; |
155 | uschar real_dbmname[512]; | |
059ec3d9 PH |
156 | uschar *buffer = malloc(max_outsize); |
157 | uschar *line = malloc(max_insize); | |
158 | ||
159 | while (argc > 1) | |
160 | { | |
161 | if (Ustrcmp(argv[arg], "-nolc") == 0) lowercase = FALSE; | |
162 | else if (Ustrcmp(argv[arg], "-nowarn") == 0) warn = FALSE; | |
163 | else if (Ustrcmp(argv[arg], "-lastdup") == 0) lastdup = TRUE; | |
164 | else if (Ustrcmp(argv[arg], "-noduperr") == 0) duperr = FALSE; | |
165 | else if (Ustrcmp(argv[arg], "-nozero") == 0) add_zero = 0; | |
166 | else break; | |
167 | arg++; | |
168 | argc--; | |
169 | } | |
170 | ||
171 | if (argc != 3) | |
172 | { | |
173 | printf("usage: exim_dbmbuild [-nolc] <source file> <dbm base name>\n"); | |
174 | exit(1); | |
175 | } | |
176 | ||
177 | if (Ustrcmp(argv[arg], "-") == 0) f = stdin; else | |
178 | { | |
179 | f = fopen(argv[arg], "rb"); | |
180 | if (f == NULL) | |
181 | { | |
182 | printf("exim_dbmbuild: unable to open %s: %s\n", argv[arg], strerror(errno)); | |
183 | exit(1); | |
184 | } | |
185 | } | |
186 | ||
187 | /* By default Berkeley db does not put extensions on... which | |
188 | can be painful! */ | |
189 | ||
190 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) | |
191 | if (Ustrcmp(argv[arg], argv[arg+1]) == 0) | |
192 | { | |
193 | printf("exim_dbmbuild: input and output filenames are the same\n"); | |
194 | exit(1); | |
195 | } | |
196 | #endif | |
197 | ||
a0540757 PH |
198 | /* Check length of filename; allow for adding .dbmbuild_temp and .db or |
199 | .dir/.pag later. */ | |
200 | ||
201 | if (strlen(argv[arg+1]) > sizeof(temp_dbmname) - 20) | |
202 | { | |
203 | printf("exim_dbmbuild: output filename is ridiculously long\n"); | |
204 | exit(1); | |
205 | } | |
206 | ||
059ec3d9 PH |
207 | Ustrcpy(temp_dbmname, argv[arg+1]); |
208 | Ustrcat(temp_dbmname, ".dbmbuild_temp"); | |
209 | ||
210 | /* It is apparently necessary to open with O_RDWR for this to work | |
211 | with gdbm-1.7.3, though no reading is actually going to be done. */ | |
212 | ||
213 | EXIM_DBOPEN(temp_dbmname, O_RDWR|O_CREAT|O_EXCL, 0644, &d); | |
214 | ||
215 | if (d == NULL) | |
216 | { | |
217 | printf("exim_dbmbuild: unable to create %s: %s\n", temp_dbmname, | |
218 | strerror(errno)); | |
f1e894f3 | 219 | (void)fclose(f); |
059ec3d9 PH |
220 | exit(1); |
221 | } | |
222 | ||
223 | /* Unless using native db calls, see if we have created <name>.db; if not, | |
224 | assume .dir & .pag */ | |
225 | ||
226 | #if !defined(USE_DB) && !defined(USE_TDB) && !defined(USE_GDBM) | |
227 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
228 | is_db = Ustat(real_dbmname, &statbuf) == 0; | |
229 | #endif | |
230 | ||
231 | /* Now do the business */ | |
232 | ||
233 | bptr = buffer; | |
234 | started = 0; | |
235 | ||
236 | while (Ufgets(line, max_insize, f) != NULL) | |
237 | { | |
238 | uschar *p; | |
239 | int len = Ustrlen(line); | |
240 | ||
241 | p = line + len; | |
242 | ||
243 | if (len >= max_insize - 1 && p[-1] != '\n') | |
244 | { | |
245 | printf("Overlong line read: max permitted length is %d\n", max_insize - 1); | |
aa2b5c79 PH |
246 | yield = 2; |
247 | goto TIDYUP; | |
059ec3d9 PH |
248 | } |
249 | ||
250 | if (line[0] == '#') continue; | |
251 | while (p > line && isspace(p[-1])) p--; | |
252 | *p = 0; | |
253 | if (line[0] == 0) continue; | |
254 | ||
255 | /* A continuation line is valid only if there was a previous first | |
256 | line. */ | |
257 | ||
258 | if (isspace(line[0])) | |
259 | { | |
260 | uschar *s = line; | |
261 | if (!started) | |
262 | { | |
263 | printf("Unexpected continuation line ignored\n%s\n\n", line); | |
264 | continue; | |
265 | } | |
266 | while (isspace(*s)) s++; | |
267 | *(--s) = ' '; | |
268 | ||
269 | if (bptr - buffer + p - s >= max_outsize - 1) | |
270 | { | |
271 | printf("Continued set of lines is too long: max permitted length is %d\n", | |
272 | max_outsize -1); | |
aa2b5c79 PH |
273 | yield = 2; |
274 | goto TIDYUP; | |
059ec3d9 PH |
275 | } |
276 | ||
277 | Ustrcpy(bptr, s); | |
278 | bptr += p - s; | |
279 | } | |
280 | ||
281 | /* A first line must have a name followed by a colon or whitespace or | |
282 | end of line, but first finish with a previous line. The key is lower | |
283 | cased by default - this is what the newaliases program for sendmail does. | |
284 | However, there's an option not to do this. */ | |
285 | ||
286 | else | |
287 | { | |
288 | int i, rc; | |
289 | uschar *s = line; | |
290 | uschar *keystart; | |
291 | ||
292 | if (started) | |
293 | { | |
294 | EXIM_DATUM_INIT(content); | |
295 | EXIM_DATUM_DATA(content) = CS buffer; | |
296 | EXIM_DATUM_SIZE(content) = bptr - buffer + add_zero; | |
297 | ||
298 | switch(rc = EXIM_DBPUTB(d, key, content)) | |
299 | { | |
300 | case EXIM_DBPUTB_OK: | |
301 | count++; | |
302 | break; | |
303 | ||
304 | case EXIM_DBPUTB_DUP: | |
305 | if (warn) fprintf(stderr, "** Duplicate key \"%s\"\n", | |
306 | keybuffer); | |
307 | dupcount++; | |
308 | if(duperr) yield = 1; | |
309 | if (lastdup) EXIM_DBPUT(d, key, content); | |
310 | break; | |
311 | ||
312 | default: | |
313 | fprintf(stderr, "Error %d while writing key %s: errno=%d\n", rc, | |
314 | keybuffer, errno); | |
315 | yield = 2; | |
316 | goto TIDYUP; | |
317 | } | |
318 | ||
319 | bptr = buffer; | |
320 | } | |
321 | ||
322 | EXIM_DATUM_INIT(key); | |
323 | EXIM_DATUM_DATA(key) = CS keybuffer; | |
324 | ||
325 | /* Deal with quoted keys. Escape sequences always make one character | |
326 | out of several, so we can re-build in place. */ | |
327 | ||
328 | if (*s == '\"') | |
329 | { | |
330 | uschar *t = s++; | |
331 | keystart = t; | |
332 | while (*s != 0 && *s != '\"') | |
333 | { | |
334 | if (*s == '\\') *t++ = string_interpret_escape(&s); | |
335 | else *t++ = *s; | |
336 | s++; | |
337 | } | |
338 | if (*s != 0) s++; /* Past terminating " */ | |
339 | EXIM_DATUM_SIZE(key) = t - keystart + add_zero; | |
340 | } | |
341 | else | |
342 | { | |
343 | keystart = s; | |
344 | while (*s != 0 && *s != ':' && !isspace(*s)) s++; | |
345 | EXIM_DATUM_SIZE(key) = s - keystart + add_zero; | |
346 | } | |
347 | ||
348 | if (EXIM_DATUM_SIZE(key) > 256) | |
349 | { | |
350 | printf("Keys longer than 255 characters cannot be handled\n"); | |
351 | started = 0; | |
352 | yield = 2; | |
353 | goto TIDYUP; | |
354 | } | |
355 | ||
356 | if (lowercase) | |
357 | { | |
358 | for (i = 0; i < EXIM_DATUM_SIZE(key) - add_zero; i++) | |
359 | keybuffer[i] = tolower(keystart[i]); | |
360 | } | |
361 | else | |
362 | { | |
363 | for (i = 0; i < EXIM_DATUM_SIZE(key) - add_zero; i++) | |
364 | keybuffer[i] = keystart[i]; | |
365 | } | |
366 | ||
367 | keybuffer[i] = 0; | |
368 | started = 1; | |
369 | ||
370 | while (isspace(*s))s++; | |
371 | if (*s == ':') | |
372 | { | |
373 | s++; | |
374 | while (isspace(*s))s++; | |
375 | } | |
376 | if (*s != 0) | |
377 | { | |
378 | Ustrcpy(bptr, s); | |
379 | bptr += p - s; | |
380 | } | |
381 | else buffer[0] = 0; | |
382 | } | |
383 | } | |
384 | ||
385 | if (started) | |
386 | { | |
387 | int rc; | |
388 | EXIM_DATUM_INIT(content); | |
389 | EXIM_DATUM_DATA(content) = CS buffer; | |
390 | EXIM_DATUM_SIZE(content) = bptr - buffer + add_zero; | |
391 | ||
392 | switch(rc = EXIM_DBPUTB(d, key, content)) | |
393 | { | |
394 | case EXIM_DBPUTB_OK: | |
395 | count++; | |
396 | break; | |
397 | ||
398 | case EXIM_DBPUTB_DUP: | |
399 | if (warn) fprintf(stderr, "** Duplicate key \"%s\"\n", keybuffer); | |
400 | dupcount++; | |
401 | if (duperr) yield = 1; | |
402 | if (lastdup) EXIM_DBPUT(d, key, content); | |
403 | break; | |
404 | ||
405 | default: | |
406 | fprintf(stderr, "Error %d while writing key %s: errno=%d\n", rc, | |
407 | keybuffer, errno); | |
408 | yield = 2; | |
409 | break; | |
410 | } | |
411 | } | |
412 | ||
413 | /* Close files, rename or abandon the temporary files, and exit */ | |
414 | ||
415 | TIDYUP: | |
416 | ||
417 | EXIM_DBCLOSE(d); | |
f1e894f3 | 418 | (void)fclose(f); |
059ec3d9 PH |
419 | |
420 | /* If successful, output the number of entries and rename the temporary | |
421 | files. */ | |
422 | ||
423 | if (yield == 0 || yield == 1) | |
424 | { | |
425 | printf("%d entr%s written\n", count, (count == 1)? "y" : "ies"); | |
426 | if (dupcount > 0) | |
427 | { | |
428 | printf("%d duplicate key%s \n", dupcount, (dupcount > 1)? "s" : ""); | |
429 | } | |
430 | ||
431 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) | |
432 | Ustrcpy(real_dbmname, temp_dbmname); | |
433 | Ustrcpy(buffer, argv[arg+1]); | |
434 | if (Urename(real_dbmname, buffer) != 0) | |
435 | { | |
436 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
437 | return 1; | |
438 | } | |
439 | #else | |
440 | ||
441 | /* Rename a single .db file */ | |
442 | ||
443 | if (is_db) | |
444 | { | |
445 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
446 | sprintf(CS buffer, "%s.db", argv[arg+1]); | |
447 | if (Urename(real_dbmname, buffer) != 0) | |
448 | { | |
449 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
450 | return 1; | |
451 | } | |
452 | } | |
453 | ||
454 | /* Rename .dir and .pag files */ | |
455 | ||
456 | else | |
457 | { | |
458 | sprintf(CS real_dbmname, "%s.dir", temp_dbmname); | |
459 | sprintf(CS buffer, "%s.dir", argv[arg+1]); | |
460 | if (Urename(real_dbmname, buffer) != 0) | |
461 | { | |
462 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
463 | return 1; | |
464 | } | |
465 | ||
466 | sprintf(CS real_dbmname, "%s.pag", temp_dbmname); | |
467 | sprintf(CS buffer, "%s.pag", argv[arg+1]); | |
468 | if (Urename(real_dbmname, buffer) != 0) | |
469 | { | |
470 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
471 | return 1; | |
472 | } | |
473 | } | |
474 | ||
475 | #endif /* USE_DB || USE_TDB || USE_GDBM */ | |
476 | } | |
477 | ||
478 | /* Otherwise unlink the temporary files. */ | |
479 | ||
480 | else | |
481 | { | |
482 | printf("dbmbuild abandoned\n"); | |
483 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) | |
484 | Uunlink(temp_dbmname); | |
485 | #else | |
486 | if (is_db) | |
487 | { | |
488 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
489 | Uunlink(real_dbmname); | |
490 | } | |
491 | else | |
492 | { | |
493 | sprintf(CS real_dbmname, "%s.dir", temp_dbmname); | |
494 | Uunlink(real_dbmname); | |
495 | sprintf(CS real_dbmname, "%s.pag", temp_dbmname); | |
496 | Uunlink(real_dbmname); | |
497 | } | |
498 | #endif /* USE_DB || USE_TDB */ | |
499 | } | |
500 | ||
501 | return yield; | |
502 | } | |
503 | ||
504 | /* End of exim_dbmbuild.c */ |