Commit | Line | Data |
---|---|---|
c988f1f4 | 1 | /* $Cambridge: exim/src/src/exim_dbmbuild.c,v 1.2 2005/01/04 10:00:42 ph10 Exp $ */ |
059ec3d9 PH |
2 | |
3 | /************************************************* | |
4 | * Exim - an Internet mail transport agent * | |
5 | *************************************************/ | |
6 | ||
c988f1f4 | 7 | /* Copyright (c) University of Cambridge 1995 - 2005 */ |
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 | |
61 | exists" when you try to open a db file. */ | |
62 | ||
63 | #if defined(USE_DB) && defined(DB_VERSION_STRING) | |
64 | void | |
65 | dbfn_bdb_error_callback(const char *pfx, char *msg) | |
66 | { | |
67 | pfx = pfx; | |
68 | printf("Berkeley DB error: %s\n", msg); | |
69 | } | |
70 | #endif | |
71 | ||
72 | ||
73 | ||
74 | /************************************************* | |
75 | * Interpret escape sequence * | |
76 | *************************************************/ | |
77 | ||
78 | /* This function is copied from the main Exim code. | |
79 | ||
80 | Arguments: | |
81 | pp points a pointer to the initiating "\" in the string; | |
82 | the pointer gets updated to point to the final character | |
83 | Returns: the value of the character escape | |
84 | */ | |
85 | ||
86 | int | |
87 | string_interpret_escape(uschar **pp) | |
88 | { | |
89 | int ch; | |
90 | uschar *p = *pp; | |
91 | ch = *(++p); | |
92 | if (isdigit(ch) && ch != '8' && ch != '9') | |
93 | { | |
94 | ch -= '0'; | |
95 | if (isdigit(p[1]) && p[1] != '8' && p[1] != '9') | |
96 | { | |
97 | ch = ch * 8 + *(++p) - '0'; | |
98 | if (isdigit(p[1]) && p[1] != '8' && p[1] != '9') | |
99 | ch = ch * 8 + *(++p) - '0'; | |
100 | } | |
101 | } | |
102 | else switch(ch) | |
103 | { | |
104 | case 'n': ch = '\n'; break; | |
105 | case 'r': ch = '\r'; break; | |
106 | case 't': ch = '\t'; break; | |
107 | case 'x': | |
108 | ch = 0; | |
109 | if (isxdigit(p[1])) | |
110 | { | |
111 | ch = ch * 16 + | |
112 | Ustrchr(hex_digits, tolower(*(++p))) - hex_digits; | |
113 | if (isxdigit(p[1])) ch = ch * 16 + | |
114 | Ustrchr(hex_digits, tolower(*(++p))) - hex_digits; | |
115 | } | |
116 | break; | |
117 | } | |
118 | *pp = p; | |
119 | return ch; | |
120 | } | |
121 | ||
122 | ||
123 | /************************************************* | |
124 | * Main Program * | |
125 | *************************************************/ | |
126 | ||
127 | int main(int argc, char **argv) | |
128 | { | |
129 | int started; | |
130 | int count = 0; | |
131 | int dupcount = 0; | |
132 | int yield = 0; | |
133 | int arg = 1; | |
134 | int add_zero = 1; | |
135 | BOOL lowercase = TRUE; | |
136 | BOOL warn = TRUE; | |
137 | BOOL duperr = TRUE; | |
138 | BOOL lastdup = FALSE; | |
139 | #if !defined (USE_DB) && !defined(USE_TDB) && !defined(USE_GDBM) | |
140 | int is_db = 0; | |
141 | struct stat statbuf; | |
142 | #endif | |
143 | FILE *f; | |
144 | EXIM_DB *d; | |
145 | EXIM_DATUM key, content; | |
146 | uschar *bptr; | |
147 | uschar keybuffer[256]; | |
148 | uschar temp_dbmname[256]; | |
149 | uschar real_dbmname[256]; | |
150 | uschar *buffer = malloc(max_outsize); | |
151 | uschar *line = malloc(max_insize); | |
152 | ||
153 | while (argc > 1) | |
154 | { | |
155 | if (Ustrcmp(argv[arg], "-nolc") == 0) lowercase = FALSE; | |
156 | else if (Ustrcmp(argv[arg], "-nowarn") == 0) warn = FALSE; | |
157 | else if (Ustrcmp(argv[arg], "-lastdup") == 0) lastdup = TRUE; | |
158 | else if (Ustrcmp(argv[arg], "-noduperr") == 0) duperr = FALSE; | |
159 | else if (Ustrcmp(argv[arg], "-nozero") == 0) add_zero = 0; | |
160 | else break; | |
161 | arg++; | |
162 | argc--; | |
163 | } | |
164 | ||
165 | if (argc != 3) | |
166 | { | |
167 | printf("usage: exim_dbmbuild [-nolc] <source file> <dbm base name>\n"); | |
168 | exit(1); | |
169 | } | |
170 | ||
171 | if (Ustrcmp(argv[arg], "-") == 0) f = stdin; else | |
172 | { | |
173 | f = fopen(argv[arg], "rb"); | |
174 | if (f == NULL) | |
175 | { | |
176 | printf("exim_dbmbuild: unable to open %s: %s\n", argv[arg], strerror(errno)); | |
177 | exit(1); | |
178 | } | |
179 | } | |
180 | ||
181 | /* By default Berkeley db does not put extensions on... which | |
182 | can be painful! */ | |
183 | ||
184 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) | |
185 | if (Ustrcmp(argv[arg], argv[arg+1]) == 0) | |
186 | { | |
187 | printf("exim_dbmbuild: input and output filenames are the same\n"); | |
188 | exit(1); | |
189 | } | |
190 | #endif | |
191 | ||
192 | Ustrcpy(temp_dbmname, argv[arg+1]); | |
193 | Ustrcat(temp_dbmname, ".dbmbuild_temp"); | |
194 | ||
195 | /* It is apparently necessary to open with O_RDWR for this to work | |
196 | with gdbm-1.7.3, though no reading is actually going to be done. */ | |
197 | ||
198 | EXIM_DBOPEN(temp_dbmname, O_RDWR|O_CREAT|O_EXCL, 0644, &d); | |
199 | ||
200 | if (d == NULL) | |
201 | { | |
202 | printf("exim_dbmbuild: unable to create %s: %s\n", temp_dbmname, | |
203 | strerror(errno)); | |
204 | fclose(f); | |
205 | exit(1); | |
206 | } | |
207 | ||
208 | /* Unless using native db calls, see if we have created <name>.db; if not, | |
209 | assume .dir & .pag */ | |
210 | ||
211 | #if !defined(USE_DB) && !defined(USE_TDB) && !defined(USE_GDBM) | |
212 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
213 | is_db = Ustat(real_dbmname, &statbuf) == 0; | |
214 | #endif | |
215 | ||
216 | /* Now do the business */ | |
217 | ||
218 | bptr = buffer; | |
219 | started = 0; | |
220 | ||
221 | while (Ufgets(line, max_insize, f) != NULL) | |
222 | { | |
223 | uschar *p; | |
224 | int len = Ustrlen(line); | |
225 | ||
226 | p = line + len; | |
227 | ||
228 | if (len >= max_insize - 1 && p[-1] != '\n') | |
229 | { | |
230 | printf("Overlong line read: max permitted length is %d\n", max_insize - 1); | |
231 | return 1; | |
232 | } | |
233 | ||
234 | if (line[0] == '#') continue; | |
235 | while (p > line && isspace(p[-1])) p--; | |
236 | *p = 0; | |
237 | if (line[0] == 0) continue; | |
238 | ||
239 | /* A continuation line is valid only if there was a previous first | |
240 | line. */ | |
241 | ||
242 | if (isspace(line[0])) | |
243 | { | |
244 | uschar *s = line; | |
245 | if (!started) | |
246 | { | |
247 | printf("Unexpected continuation line ignored\n%s\n\n", line); | |
248 | continue; | |
249 | } | |
250 | while (isspace(*s)) s++; | |
251 | *(--s) = ' '; | |
252 | ||
253 | if (bptr - buffer + p - s >= max_outsize - 1) | |
254 | { | |
255 | printf("Continued set of lines is too long: max permitted length is %d\n", | |
256 | max_outsize -1); | |
257 | return 1; | |
258 | } | |
259 | ||
260 | Ustrcpy(bptr, s); | |
261 | bptr += p - s; | |
262 | } | |
263 | ||
264 | /* A first line must have a name followed by a colon or whitespace or | |
265 | end of line, but first finish with a previous line. The key is lower | |
266 | cased by default - this is what the newaliases program for sendmail does. | |
267 | However, there's an option not to do this. */ | |
268 | ||
269 | else | |
270 | { | |
271 | int i, rc; | |
272 | uschar *s = line; | |
273 | uschar *keystart; | |
274 | ||
275 | if (started) | |
276 | { | |
277 | EXIM_DATUM_INIT(content); | |
278 | EXIM_DATUM_DATA(content) = CS buffer; | |
279 | EXIM_DATUM_SIZE(content) = bptr - buffer + add_zero; | |
280 | ||
281 | switch(rc = EXIM_DBPUTB(d, key, content)) | |
282 | { | |
283 | case EXIM_DBPUTB_OK: | |
284 | count++; | |
285 | break; | |
286 | ||
287 | case EXIM_DBPUTB_DUP: | |
288 | if (warn) fprintf(stderr, "** Duplicate key \"%s\"\n", | |
289 | keybuffer); | |
290 | dupcount++; | |
291 | if(duperr) yield = 1; | |
292 | if (lastdup) EXIM_DBPUT(d, key, content); | |
293 | break; | |
294 | ||
295 | default: | |
296 | fprintf(stderr, "Error %d while writing key %s: errno=%d\n", rc, | |
297 | keybuffer, errno); | |
298 | yield = 2; | |
299 | goto TIDYUP; | |
300 | } | |
301 | ||
302 | bptr = buffer; | |
303 | } | |
304 | ||
305 | EXIM_DATUM_INIT(key); | |
306 | EXIM_DATUM_DATA(key) = CS keybuffer; | |
307 | ||
308 | /* Deal with quoted keys. Escape sequences always make one character | |
309 | out of several, so we can re-build in place. */ | |
310 | ||
311 | if (*s == '\"') | |
312 | { | |
313 | uschar *t = s++; | |
314 | keystart = t; | |
315 | while (*s != 0 && *s != '\"') | |
316 | { | |
317 | if (*s == '\\') *t++ = string_interpret_escape(&s); | |
318 | else *t++ = *s; | |
319 | s++; | |
320 | } | |
321 | if (*s != 0) s++; /* Past terminating " */ | |
322 | EXIM_DATUM_SIZE(key) = t - keystart + add_zero; | |
323 | } | |
324 | else | |
325 | { | |
326 | keystart = s; | |
327 | while (*s != 0 && *s != ':' && !isspace(*s)) s++; | |
328 | EXIM_DATUM_SIZE(key) = s - keystart + add_zero; | |
329 | } | |
330 | ||
331 | if (EXIM_DATUM_SIZE(key) > 256) | |
332 | { | |
333 | printf("Keys longer than 255 characters cannot be handled\n"); | |
334 | started = 0; | |
335 | yield = 2; | |
336 | goto TIDYUP; | |
337 | } | |
338 | ||
339 | if (lowercase) | |
340 | { | |
341 | for (i = 0; i < EXIM_DATUM_SIZE(key) - add_zero; i++) | |
342 | keybuffer[i] = tolower(keystart[i]); | |
343 | } | |
344 | else | |
345 | { | |
346 | for (i = 0; i < EXIM_DATUM_SIZE(key) - add_zero; i++) | |
347 | keybuffer[i] = keystart[i]; | |
348 | } | |
349 | ||
350 | keybuffer[i] = 0; | |
351 | started = 1; | |
352 | ||
353 | while (isspace(*s))s++; | |
354 | if (*s == ':') | |
355 | { | |
356 | s++; | |
357 | while (isspace(*s))s++; | |
358 | } | |
359 | if (*s != 0) | |
360 | { | |
361 | Ustrcpy(bptr, s); | |
362 | bptr += p - s; | |
363 | } | |
364 | else buffer[0] = 0; | |
365 | } | |
366 | } | |
367 | ||
368 | if (started) | |
369 | { | |
370 | int rc; | |
371 | EXIM_DATUM_INIT(content); | |
372 | EXIM_DATUM_DATA(content) = CS buffer; | |
373 | EXIM_DATUM_SIZE(content) = bptr - buffer + add_zero; | |
374 | ||
375 | switch(rc = EXIM_DBPUTB(d, key, content)) | |
376 | { | |
377 | case EXIM_DBPUTB_OK: | |
378 | count++; | |
379 | break; | |
380 | ||
381 | case EXIM_DBPUTB_DUP: | |
382 | if (warn) fprintf(stderr, "** Duplicate key \"%s\"\n", keybuffer); | |
383 | dupcount++; | |
384 | if (duperr) yield = 1; | |
385 | if (lastdup) EXIM_DBPUT(d, key, content); | |
386 | break; | |
387 | ||
388 | default: | |
389 | fprintf(stderr, "Error %d while writing key %s: errno=%d\n", rc, | |
390 | keybuffer, errno); | |
391 | yield = 2; | |
392 | break; | |
393 | } | |
394 | } | |
395 | ||
396 | /* Close files, rename or abandon the temporary files, and exit */ | |
397 | ||
398 | TIDYUP: | |
399 | ||
400 | EXIM_DBCLOSE(d); | |
401 | fclose(f); | |
402 | ||
403 | /* If successful, output the number of entries and rename the temporary | |
404 | files. */ | |
405 | ||
406 | if (yield == 0 || yield == 1) | |
407 | { | |
408 | printf("%d entr%s written\n", count, (count == 1)? "y" : "ies"); | |
409 | if (dupcount > 0) | |
410 | { | |
411 | printf("%d duplicate key%s \n", dupcount, (dupcount > 1)? "s" : ""); | |
412 | } | |
413 | ||
414 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) | |
415 | Ustrcpy(real_dbmname, temp_dbmname); | |
416 | Ustrcpy(buffer, argv[arg+1]); | |
417 | if (Urename(real_dbmname, buffer) != 0) | |
418 | { | |
419 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
420 | return 1; | |
421 | } | |
422 | #else | |
423 | ||
424 | /* Rename a single .db file */ | |
425 | ||
426 | if (is_db) | |
427 | { | |
428 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
429 | sprintf(CS buffer, "%s.db", argv[arg+1]); | |
430 | if (Urename(real_dbmname, buffer) != 0) | |
431 | { | |
432 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
433 | return 1; | |
434 | } | |
435 | } | |
436 | ||
437 | /* Rename .dir and .pag files */ | |
438 | ||
439 | else | |
440 | { | |
441 | sprintf(CS real_dbmname, "%s.dir", temp_dbmname); | |
442 | sprintf(CS buffer, "%s.dir", argv[arg+1]); | |
443 | if (Urename(real_dbmname, buffer) != 0) | |
444 | { | |
445 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
446 | return 1; | |
447 | } | |
448 | ||
449 | sprintf(CS real_dbmname, "%s.pag", temp_dbmname); | |
450 | sprintf(CS buffer, "%s.pag", argv[arg+1]); | |
451 | if (Urename(real_dbmname, buffer) != 0) | |
452 | { | |
453 | printf("Unable to rename %s as %s\n", real_dbmname, buffer); | |
454 | return 1; | |
455 | } | |
456 | } | |
457 | ||
458 | #endif /* USE_DB || USE_TDB || USE_GDBM */ | |
459 | } | |
460 | ||
461 | /* Otherwise unlink the temporary files. */ | |
462 | ||
463 | else | |
464 | { | |
465 | printf("dbmbuild abandoned\n"); | |
466 | #if defined(USE_DB) || defined(USE_TDB) || defined(USE_GDBM) | |
467 | Uunlink(temp_dbmname); | |
468 | #else | |
469 | if (is_db) | |
470 | { | |
471 | sprintf(CS real_dbmname, "%s.db", temp_dbmname); | |
472 | Uunlink(real_dbmname); | |
473 | } | |
474 | else | |
475 | { | |
476 | sprintf(CS real_dbmname, "%s.dir", temp_dbmname); | |
477 | Uunlink(real_dbmname); | |
478 | sprintf(CS real_dbmname, "%s.pag", temp_dbmname); | |
479 | Uunlink(real_dbmname); | |
480 | } | |
481 | #endif /* USE_DB || USE_TDB */ | |
482 | } | |
483 | ||
484 | return yield; | |
485 | } | |
486 | ||
487 | /* End of exim_dbmbuild.c */ |