xref: /titanic_51/usr/src/lib/pam_modules/authtok_check/packer.c (revision 4a7ceb24cfcc0a97f96d86cfe5852ae445b50e57)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 #include "packer.h"
29 
30 /*
31  * This file steers the creation of the Crack Dictionary Database.
32  * Based on a list of source dictionaries specified by the administrator,
33  * we create the Database by sorting each dictionary (in memory, one at
34  * a time), writing the sorted result to a temporary file, and merging
35  * all the temporary files into the Database.
36  *
37  * The current implementation has a number of limitations
38  *   - each single source dictionary has to fit in memory
39  *   - each single source dictionary has to be smaller than 2GByte
40  *   - each single source dictionary can only hold up to 4GB words
41  * None of these seem real, practical, problems to me.
42  *
43  * All of this is meant to be run by one thread per host. The caller is
44  * responsible for locking things appropriately (as make_dict_database
45  * in dict.c does).
46  */
47 
48 #include <stdio.h>
49 #include <stdlib.h>
50 #include <unistd.h>
51 #include <ctype.h>
52 #include <string.h>
53 #include <errno.h>
54 #include <sys/stat.h>
55 #include <fcntl.h>
56 
57 /* Stuff used for sorting the dictionary */
58 static char	*buf;		/* used to hold the source dictionary */
59 static uint_t	*offsets;	/* array of word-offsets into "buf" */
60 static uint_t	off_idx = 0;	/* first free index in offsets array */
61 static size_t	off_size = 0;	/* offsets array size */
62 
63 /* stuff to keep track of the temporary files */
64 #define	FNAME_TEMPLATE	"/var/tmp/authtok_check.XXXXXX"
65 #define	MAXTMP		64
66 static FILE	*tmpfp[MAXTMP];	/* FILE *'s to (unlinked) temporary files */
67 static int	tmpfp_idx = 0;	/* points to first free entry in tmpfp */
68 
69 #define	MODNAME "pam_authtok_check::packer"
70 
71 /*
72  * int writeout(void)
73  *
74  * Write the sorted wordlist to disk. We create a temporary file
75  * (in /var/tmp), and immediately unlink() it. We keep an open
76  * FILE pointer to it in tmpfp[] for later use.
77  *
78  * returns 0 on success, -1 on failure (can't create file/output failure).
79  */
80 int
81 writeout(void)
82 {
83 	int i = 0;
84 	char tmpname[sizeof (FNAME_TEMPLATE)];
85 	int fd;
86 
87 	if (tmpfp_idx == MAXTMP) {
88 		syslog(LOG_ERR, MODNAME ": too many temporary "
89 		    "files (maximum %d exceeded)", MAXTMP);
90 		return (-1);
91 	}
92 
93 	(void) strcpy(tmpname, FNAME_TEMPLATE);
94 	if ((fd = mkstemp(tmpname)) == -1) {
95 		syslog(LOG_ERR, MODNAME ": mkstemp() failed: %s\n",
96 		    strerror(errno));
97 		return (-1);
98 	}
99 	(void) unlink(tmpname);
100 
101 	if ((tmpfp[tmpfp_idx] = fdopen(fd, "w+F")) == NULL) {
102 		syslog(LOG_ERR, MODNAME ": fdopen failed: %s",
103 		    strerror(errno));
104 		(void) close(fd);
105 		return (-1);
106 	}
107 
108 	/* write words to file */
109 	while (i < off_idx) {
110 		if (fprintf(tmpfp[tmpfp_idx], "%s\n", &buf[offsets[i++]]) < 0) {
111 			syslog(LOG_ERR, MODNAME ": write to file failed: %s",
112 			    strerror(errno));
113 			(void) close(fd);
114 			return (-1);
115 		}
116 	}
117 
118 	/* we have one extra tmpfp */
119 	tmpfp_idx++;
120 
121 	return (0);
122 }
123 
124 /*
125  * int insert_word(int off)
126  *
127  * insert an offset into the offsets-array. If the offsets-array is out of
128  * space, we allocate additional space (in CHUNKs)
129  *
130  * returns 0 on success, -1 on failure (out of memory)
131  */
132 int
133 insert_word(int off)
134 {
135 #define	CHUNK 10000
136 
137 	if (off_idx == off_size) {
138 		uint_t *tmp;
139 		off_size += CHUNK;
140 		tmp = realloc(offsets, sizeof (uint_t) * off_size);
141 		if (tmp == NULL) {
142 			syslog(LOG_ERR, MODNAME ": out of memory");
143 			free(offsets);
144 			off_idx = off_size = 0;
145 			offsets = NULL;
146 			return (-1);
147 		}
148 		offsets = tmp;
149 	}
150 
151 	offsets[off_idx++] = off;
152 	return (0);
153 }
154 
155 /*
156  * translate(buf, size)
157  *
158  * perform "tr '[A-Z]' '[a-z]' | tr -cd '\012[a-z][0-9]'" on the
159  * words in "buf" and insert each of them into the offsets-array.
160  * We refrain from using 'isupper' and 'islower' to keep this strictly
161  * ASCII-only, as is the original Cracklib code.
162  *
163  * returns 0 on success, -1 on failure (failure of insert_word)
164  */
165 int
166 translate(char *buf, size_t size)
167 {
168 	char *p, *q, *e;
169 	char c;
170 	int wordstart;
171 
172 	e = &buf[size];
173 
174 	wordstart = 0;
175 	for (p = buf, q = buf; q < e; q++) {
176 		c = *q;
177 		if (c >= 'A' && c <= 'Z') {
178 			*(p++) = tolower(c);
179 		} else if (c == '\n') {
180 			*(p++) = '\0';
181 			/*
182 			 * make sure we only insert words consisting of
183 			 * MAXWORDLEN-1 bytes or less
184 			 */
185 			if (p-&buf[wordstart] > MAXWORDLEN)
186 				buf[wordstart+MAXWORDLEN-1] = '\0';
187 			if (insert_word(wordstart) != 0)
188 				return (-1);
189 			wordstart = p-buf;
190 		} else if ((c >= 'a' && c <= 'z') || (c >= '0' && c <= '9')) {
191 			*(p++) = c;
192 		}
193 	}
194 	return (0);
195 }
196 
197 /*
198  * int compare(a, b)
199  *
200  * helper-routine used for quicksort. we compate two words in the
201  * buffer, one start starts at index "a", and the other one that starts
202  * at index "b"
203  */
204 int
205 compare(const void *a, const void *b)
206 {
207 	int idx_a = *(uint_t *)a, idx_b = *(uint_t *)b;
208 
209 	return (strcmp(&buf[idx_a], &buf[idx_b]));
210 }
211 
212 /*
213  *
214  * int sort_file(fname)
215  *
216  * We sort the file in memory: we read the dictionary file, translate all
217  * newlines to '\0's, all uppercase ASCII characters to lowercase characters
218  * and removing all characters but '[a-z][0-9]'.
219  * We maintain an array of offsets into the buffer where each word starts
220  * and sort this array using qsort().
221  *
222  * This implements the original cracklib code that did an execl of
223  *    sh -c "/usr/bin/cat <list of files> |
224  *       /usr/bin/tr '[A-Z]' '[a-z]' | /usr/bin/tr -cd '\012[a-z][0-9]' |
225  *       sort -o tmfpfile
226  *
227  * returns 0 on success, -1 on failure.
228  */
229 int
230 sort_file(char *fname)
231 {
232 	int fd;
233 	struct stat statbuf;
234 	ssize_t n;
235 	int ret = -1;
236 
237 	if ((fd = open(fname, O_RDONLY)) == -1) {
238 		syslog(LOG_ERR, MODNAME ": failed to open %s: %s",
239 		    fname, strerror(errno));
240 		return (-1);
241 	}
242 
243 	if (fstat(fd, &statbuf) == -1) {
244 		syslog(LOG_ERR, MODNAME ": fstat() failed (%s)",
245 		    strerror(errno));
246 		(void) close(fd);
247 		return (-1);
248 	}
249 	if ((buf = malloc(statbuf.st_size + 1)) == NULL) {
250 		syslog(LOG_ERR, MODNAME ": out of memory");
251 		goto error;
252 	}
253 
254 	n = read(fd, buf, statbuf.st_size);
255 
256 	if (n == -1) {
257 		if (errno == EINVAL)
258 			syslog(LOG_ERR, MODNAME ": %s is too big. "
259 			    "Split the file into smaller files.", fname);
260 		else
261 			syslog(LOG_ERR, MODNAME ": read failed: %s",
262 			    strerror(errno));
263 		goto error;
264 	}
265 
266 	if (translate(buf, n) == 0) {
267 		qsort((void *)offsets, off_idx, sizeof (int), compare);
268 
269 		if (writeout() == 0)
270 			ret = 0;
271 	}
272 
273 error:
274 	(void) close(fd);
275 
276 	if (buf != NULL)
277 		free(buf);
278 	if (offsets != NULL)
279 		free(offsets);
280 	offsets = NULL;
281 	off_size = 0;
282 	off_idx = 0;
283 	return (ret);
284 }
285 
286 /*
287  * We merge the temporary files created by previous calls to sort_file()
288  * and insert the thus sorted words into the cracklib database
289  *
290  * returns 0 on success, -1 on failure.
291  */
292 int
293 merge_files(PWDICT *pwp)
294 {
295 	int ti;
296 	char *words[MAXTMP];
297 	char lastword[MAXWORDLEN];
298 	int choice;
299 
300 	lastword[0] = '\0';
301 
302 	for (ti = 0; ti < tmpfp_idx; ti++)
303 		if ((words[ti] = malloc(MAXWORDLEN)) == NULL) {
304 			while (--ti >= 0)
305 				free(words[ti]);
306 			return (-1);
307 		}
308 
309 	/*
310 	 * we read the first word of each of the temp-files into words[].
311 	 */
312 	for (ti = 0; ti < tmpfp_idx; ti++) {
313 		(void) fseek(tmpfp[ti], 0, SEEK_SET);
314 		(void) fgets(words[ti], MAXWORDLEN, tmpfp[ti]);
315 		words[ti][MAXWORDLEN-1] = '\0';
316 	}
317 
318 	/*
319 	 * next, we emit the word that comes first (lexicographically),
320 	 * and replace that word with a new word from the file it
321 	 * came from. If the file is exhausted, we close the fp and
322 	 * swap the fp with the last fp in tmpfp[].
323 	 * we then decrease tmpfp_idx and continue with what's left until
324 	 * we run out of open FILE pointers.
325 	 */
326 	while (tmpfp_idx != 0) {
327 		choice = 0;
328 
329 		for (ti = 1; ti < tmpfp_idx; ti++)
330 			if (strcmp(words[choice], words[ti]) > 0)
331 				choice = ti;
332 		/* Insert word in Cracklib database */
333 		(void) Chomp(words[choice]);
334 		if (words[choice][0] != '\0' &&
335 		    strcmp(lastword, words[choice]) != 0) {
336 			(void) PutPW(pwp, words[choice]);
337 			(void) strncpy(lastword, words[choice], MAXWORDLEN);
338 		}
339 
340 		if (fgets(words[choice], MAXWORDLEN, tmpfp[choice]) == NULL) {
341 			(void) fclose(tmpfp[choice]);
342 			tmpfp[choice] = tmpfp[tmpfp_idx - 1];
343 			tmpfp_idx--;
344 		} else
345 			words[choice][MAXWORDLEN-1] = '\0';
346 	}
347 	return (0);
348 }
349 
350 /*
351  * int packer(list)
352  *
353  * sort all dictionaries in "list", and feed the words into the Crack
354  * Password Database.
355  *
356  * returns 0 on sucess, -1 on failure.
357  */
358 int
359 packer(char *list, char *path)
360 {
361 	PWDICT *pwp;
362 	char *listcopy, *fname;
363 	int ret = 0;
364 
365 	if ((listcopy = strdup(list)) == NULL) {
366 		syslog(LOG_ERR, MODNAME ": out of memory");
367 		return (-1);
368 	}
369 
370 	if (!(pwp = PWOpen(path, "wF")))
371 		return (-1);
372 
373 	fname = strtok(listcopy, " \t,");
374 	while (ret == 0 && fname != NULL) {
375 		if ((ret = sort_file(fname)) == 0)
376 			fname = strtok(NULL, " \t,");
377 	}
378 	free(listcopy);
379 
380 	if (ret == 0)
381 		ret = merge_files(pwp);
382 
383 	(void) PWClose(pwp);
384 
385 	return (ret);
386 }
387