xref: /titanic_44/usr/src/cmd/ssh/libssh/common/g11n.c (revision e824d57f8160a27ac5e650005c7a4f037109c2be)
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  * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
22  * Use is subject to license terms.
23  */
24 
25 #pragma ident	"%Z%%M%	%I%	%E% SMI"
26 
27 #include <errno.h>
28 #include <locale.h>
29 #include <langinfo.h>
30 #include <iconv.h>
31 #include <ctype.h>
32 #include <strings.h>
33 #include <string.h>
34 #include <stdio.h>
35 #include <stdlib.h>
36 #include "includes.h"
37 #include "xmalloc.h"
38 #include "xlist.h"
39 
40 #ifdef MIN
41 #undef MIN
42 #endif /* MIN */
43 
44 #define	MIN(x, y)	((x) < (y) ? (x) : (y))
45 
46 #define	LOCALE_PATH	"/usr/bin/locale"
47 
48 /* two-char country code, '-' and two-char region code */
49 #define	LANGTAG_MAX	5
50 
51 static uchar_t *do_iconv(iconv_t cd, uint_t *mul_ptr, const void *buf,
52     uint_t len, uint_t *outlen, int *err, uchar_t **err_str);
53 
54 static int locale_cmp(const void *d1, const void *d2);
55 static char *g11n_locale2langtag(char *locale);
56 
57 uint_t g11n_validate_ascii(const char *str, uint_t len, uchar_t **error_str);
58 uint_t g11n_validate_utf8(const uchar_t *str, uint_t len, uchar_t **error_str);
59 
60 static char *
61 g11n_locale2langtag(char *locale)
62 {
63 	char *langtag;
64 
65 	/* base cases */
66 	if (!locale || !*locale)
67 		return (NULL);
68 
69 	if (strcmp(locale, "POSIX") == 0 || strcmp(locale, "C") == 0)
70 		return ("i-default");
71 
72 	/* punt for language codes which are not exactly 2 letters */
73 	if (strlen(locale) < 2 ||
74 	    !isalpha(locale[0]) ||
75 	    !isalpha(locale[1]) ||
76 	    (locale[2] != '\0' &&
77 	    locale[2] != '_' &&
78 	    locale[2] != '.' &&
79 	    locale[2] != '@'))
80 		return (NULL);
81 
82 
83 	/* we have a primary language sub-tag */
84 	langtag = (char *)xmalloc(LANGTAG_MAX + 1);
85 
86 	strncpy(langtag, locale, 2);
87 	langtag[2] = '\0';
88 
89 	/* do we have country sub-tag? For example: cs_CZ */
90 	if (locale[2] == '_') {
91 		if (strlen(locale) < 5 ||
92 		    !isalpha(locale[3]) ||
93 		    !isalpha(locale[4]) ||
94 		    (locale[5] != '\0' && (locale[5] != '.' &&
95 		    locale[5] != '@'))) {
96 			return (langtag);
97 		}
98 
99 		/* example: create cs-CZ from cs_CZ */
100 		if (snprintf(langtag, 6, "%.*s-%.*s", 2, locale, 2,
101 		    locale + 3) == 5)
102 			return (langtag);
103 	}
104 
105 	/* in all other cases we just use the primary language sub-tag */
106 	return (langtag);
107 }
108 
109 uint_t
110 g11n_langtag_is_default(char *langtag)
111 {
112 	return (strcmp(langtag, "i-default") == 0);
113 }
114 
115 /*
116  * This lang tag / locale matching function works only for two-character
117  * language primary sub-tags and two-character country sub-tags.
118  */
119 uint_t
120 g11n_langtag_matches_locale(char *langtag, char *locale)
121 {
122 	/* match "i-default" to the process' current locale if possible */
123 	if (g11n_langtag_is_default(langtag)) {
124 		if (strcasecmp(locale, "POSIX") == 0 ||
125 		    strcasecmp(locale, "C") == 0)
126 			return (1);
127 		else
128 			return (0);
129 	}
130 
131 	/*
132 	 * locale must be at least 2 chars long and the lang part must be
133 	 * exactly two characters
134 	 */
135 	if (strlen(locale) < 2 ||
136 	    (!isalpha(locale[0]) || !isalpha(locale[1]) ||
137 	    (locale[2] != '\0' && locale[2] != '_' &&
138 	    locale[2] != '.' && locale[2] != '@')))
139 		return (0);
140 
141 	/* same thing with the langtag */
142 	if (strlen(langtag) < 2 ||
143 	    (!isalpha(langtag[0]) || !isalpha(langtag[1]) ||
144 	    (langtag[2] != '\0' && langtag[2] != '-')))
145 		return (0);
146 
147 	/* primary language sub-tag and the locale's language part must match */
148 	if (strncasecmp(langtag, locale, 2) != 0)
149 		return (0);
150 
151 	/*
152 	 * primary language sub-tag and the locale's language match, now
153 	 * fuzzy check country part
154 	 */
155 
156 	/* neither langtag nor locale have more than one component */
157 	if (langtag[2] == '\0' &&
158 	    (locale[2] == '\0' || locale[2] == '.' || locale[2] == '@'))
159 		return (2);
160 
161 	/* langtag has only one sub-tag... */
162 	if (langtag[2] == '\0')
163 		return (1);
164 
165 	/* locale has no country code... */
166 	if (locale[2] == '\0' || locale[2] == '.' || locale[2] == '@')
167 		return (1);
168 
169 	/* langtag has more than one subtag and the locale has a country code */
170 
171 	/* ignore second subtag if not two chars */
172 	if (strlen(langtag) < 5)
173 		return (1);
174 
175 	if (!isalpha(langtag[3]) || !isalpha(langtag[4]) ||
176 	    (langtag[5] != '\0' && langtag[5] != '-'))
177 		return (1);
178 
179 	/* ignore rest of locale if there is no two-character country part */
180 	if (strlen(locale) < 5)
181 		return (1);
182 
183 	if (locale[2] != '_' || !isalpha(locale[3]) || !isalpha(locale[4]) ||
184 	    (locale[5] != '\0' && locale[5] != '.' && locale[5] != '@'))
185 		return (1);
186 
187 	/* if the country part matches, return 2 */
188 	if (strncasecmp(&langtag[3], &locale[3], 2) == 0)
189 		return (2);
190 
191 	return (1);
192 }
193 
194 char *
195 g11n_getlocale()
196 {
197 	/* we have one text domain - always set it */
198 	(void) textdomain(TEXT_DOMAIN);
199 
200 	/* if the locale is not set, set it from the env vars */
201 	if (!setlocale(LC_MESSAGES, NULL))
202 		(void) setlocale(LC_MESSAGES, "");
203 
204 	return (setlocale(LC_MESSAGES, NULL));
205 }
206 
207 void
208 g11n_setlocale(int category, const char *locale)
209 {
210 	char *curr;
211 
212 	/* we have one text domain - always set it */
213 	(void) textdomain(TEXT_DOMAIN);
214 
215 	if (!locale)
216 		return;
217 
218 	if (*locale && ((curr = setlocale(category, NULL))) &&
219 	    strcmp(curr, locale) == 0)
220 		return;
221 
222 	/* if <category> is bogus, setlocale() will do nothing */
223 	(void) setlocale(category, locale);
224 }
225 
226 char **
227 g11n_getlocales()
228 {
229 	FILE *locale_out;
230 	uint_t n_elems, list_size, long_line = 0;
231 	char **list;
232 	char locale[64];	/* 64 bytes is plenty for locale names */
233 
234 	if ((locale_out = popen(LOCALE_PATH " -a", "r")) == NULL)
235 		return (NULL);
236 
237 	/*
238 	 * start with enough room for 65 locales - that's a lot fewer than
239 	 * all the locales available for installation, but a lot more than
240 	 * what most users will need and install
241 	 */
242 	n_elems = 0;
243 	list_size = 192;
244 	list = (char **) xmalloc(sizeof (char *) * (list_size + 1));
245 	memset(list, 0, sizeof (char *) * (list_size + 1));
246 
247 	while (fgets(locale, sizeof (locale), locale_out)) {
248 		/* skip long locale names (if any) */
249 		if (!strchr(locale, '\n')) {
250 			long_line = 1;
251 			continue;
252 		} else if (long_line) {
253 			long_line = 0;
254 			continue;
255 		}
256 
257 		if (strncmp(locale, "iso_8859", 8) == 0)
258 			/* ignore locale names like "iso_8859-1" */
259 			continue;
260 
261 		if (n_elems == list_size) {
262 			list_size *= 2;
263 			list = (char **)xrealloc((void *) list,
264 			    (list_size + 1) * sizeof (char *));
265 			memset(&list[n_elems + 1], 0,
266 			    sizeof (char *) * (list_size - n_elems + 1));
267 		}
268 
269 		*(strchr(locale, '\n')) = '\0';	/* remove the trailing \n */
270 		list[n_elems++] = xstrdup(locale);
271 	}
272 
273 	list[n_elems] = NULL;
274 	(void) pclose(locale_out);
275 
276 	qsort(list, n_elems - 1, sizeof (char *), locale_cmp);
277 	return (list);
278 }
279 
280 char *
281 g11n_getlangs()
282 {
283 	char *locale;
284 
285 	if (getenv("SSH_LANGS"))
286 		return (xstrdup(getenv("SSH_LANGS")));
287 
288 	locale = g11n_getlocale();
289 
290 	if (!locale || !*locale)
291 		return (xstrdup("i-default"));
292 
293 	return (g11n_locale2langtag(locale));
294 }
295 
296 char *
297 g11n_locales2langs(char **locale_set)
298 {
299 	char **p, **r, **q;
300 	char *langtag;
301 	int locales, skip;
302 
303 	for (locales = 0, p = locale_set; p && *p; p++)
304 		locales++;
305 
306 	r = (char **)xmalloc((locales + 1) * sizeof (char *));
307 	memset(r, 0, (locales + 1) * sizeof (char *));
308 
309 	for (p = locale_set; p && *p && ((p - locale_set) <= locales); p++) {
310 		skip = 0;
311 		if ((langtag = g11n_locale2langtag(*p)) == NULL)
312 			continue;
313 		for (q = r; (q - r) < locales; q++) {
314 			if (!*q)
315 				break;
316 			if (*q && strcmp(*q, langtag) == 0)
317 				skip = 1;
318 		}
319 		if (!skip)
320 			*(q++) = langtag;
321 		*q = NULL;
322 	}
323 
324 	return (xjoin(r, ','));
325 }
326 
327 static int
328 sortcmp(const void *d1, const void *d2)
329 {
330 	char *s1 = *(char **)d1;
331 	char *s2 = *(char **)d2;
332 
333 	return (strcmp(s1, s2));
334 }
335 
336 int
337 g11n_langtag_match(char *langtag1, char *langtag2)
338 {
339 	int len1, len2;
340 	char c1, c2;
341 
342 	len1 = (strchr(langtag1, '-')) ?
343 		(strchr(langtag1, '-') - langtag1)
344 		: strlen(langtag1);
345 
346 	len2 = (strchr(langtag2, '-')) ?
347 		(strchr(langtag2, '-') - langtag2)
348 		: strlen(langtag2);
349 
350 	/* no match */
351 	if (len1 != len2 || strncmp(langtag1, langtag2, len1) != 0)
352 		return (0);
353 
354 	c1 = *(langtag1 + len1);
355 	c2 = *(langtag2 + len2);
356 
357 	/* no country sub-tags - exact match */
358 	if (c1 == '\0' && c2 == '\0')
359 		return (2);
360 
361 	/* one langtag has a country sub-tag, the other doesn't */
362 	if (c1 == '\0' || c2 == '\0')
363 		return (1);
364 
365 	/* can't happen - both langtags have a country sub-tag */
366 	if (c1 != '-' || c2 != '-')
367 		return (1);
368 
369 	/* compare country subtags */
370 	langtag1 = langtag1 + len1 + 1;
371 	langtag2 = langtag2 + len2 + 1;
372 
373 	len1 = (strchr(langtag1, '-')) ?
374 	    (strchr(langtag1, '-') - langtag1) : strlen(langtag1);
375 
376 	len2 = (strchr(langtag2, '-')) ?
377 	    (strchr(langtag2, '-') - langtag2) : strlen(langtag2);
378 
379 	if (len1 != len2 || strncmp(langtag1, langtag2, len1) != 0)
380 		return (1);
381 
382 	/* country tags matched - exact match */
383 	return (2);
384 }
385 
386 char *
387 g11n_langtag_set_intersect(char *set1, char *set2)
388 {
389 	char **list1, **list2, **list3, **p, **q, **r;
390 	char *set3, *lang_subtag;
391 	uint_t n1, n2, n3;
392 	uint_t do_append;
393 
394 	list1 = xsplit(set1, ',');
395 	list2 = xsplit(set2, ',');
396 
397 	for (n1 = 0, p = list1; p && *p; p++, n1++)
398 		;
399 	for (n2 = 0, p = list2; p && *p; p++, n2++)
400 		;
401 
402 	list3 = (char **) xmalloc(sizeof (char *) * (n1 + n2 + 1));
403 	*list3 = NULL;
404 
405 	/*
406 	 * we must not sort the user langtags - sorting or not the server's
407 	 * should not affect the outcome
408 	 */
409 	qsort(list2, n2, sizeof (char *), sortcmp);
410 
411 	for (n3 = 0, p = list1; p && *p; p++) {
412 		do_append = 0;
413 		for (q = list2; q && *q; q++) {
414 			if (g11n_langtag_match(*p, *q) != 2) continue;
415 			/* append element */
416 			for (r = list3; (r - list3) <= (n1 + n2); r++) {
417 				do_append = 1;
418 				if (!*r)
419 					break;
420 				if (strcmp(*p, *r) == 0) {
421 					do_append = 0;
422 					break;
423 				}
424 			}
425 			if (do_append && n3 <= (n1 + n2)) {
426 				list3[n3++] = xstrdup(*p);
427 				list3[n3] = NULL;
428 			}
429 		}
430 	}
431 
432 	for (p = list1; p && *p; p++) {
433 		do_append = 0;
434 		for (q = list2; q && *q; q++) {
435 			if (g11n_langtag_match(*p, *q) != 1)
436 				continue;
437 
438 			/* append element */
439 			lang_subtag = xstrdup(*p);
440 			if (strchr(lang_subtag, '-'))
441 				*(strchr(lang_subtag, '-')) = '\0';
442 			for (r = list3; (r - list3) <= (n1 + n2); r++) {
443 				do_append = 1;
444 				if (!*r)
445 					break;
446 				if (strcmp(lang_subtag, *r) == 0) {
447 					do_append = 0;
448 					break;
449 				}
450 			}
451 			if (do_append && n3 <= (n1 + n2)) {
452 				list3[n3++] = lang_subtag;
453 				list3[n3] = NULL;
454 			} else
455 				xfree(lang_subtag);
456 		}
457 	}
458 
459 	set3 = xjoin(list3, ',');
460 	xfree_split_list(list1);
461 	xfree_split_list(list2);
462 	xfree_split_list(list3);
463 
464 	return (set3);
465 }
466 
467 char *
468 g11n_clnt_langtag_negotiate(char *clnt_langtags, char *srvr_langtags)
469 {
470 	char *list, *result;
471 	char **xlist;
472 
473 	/* g11n_langtag_set_intersect uses xmalloc - should not return NULL */
474 	list = g11n_langtag_set_intersect(clnt_langtags, srvr_langtags);
475 
476 	if (!list)
477 		return (NULL);
478 
479 	xlist = xsplit(list, ',');
480 
481 	xfree(list);
482 
483 	if (!xlist || !*xlist)
484 		return (NULL);
485 
486 	result = xstrdup(*xlist);
487 	xfree_split_list(xlist);
488 
489 	return (result);
490 }
491 
492 /*
493  * Compare locales, preferring UTF-8 codesets to others, otherwise doing
494  * a stright strcmp()
495  */
496 static int
497 locale_cmp(const void *d1, const void *d2)
498 {
499 	char *dot_ptr;
500 	char *s1 = *(char **)d1;
501 	char *s2 = *(char **)d2;
502 	int s1_is_utf8 = 0;
503 	int s2_is_utf8 = 0;
504 
505 	/* check if s1 is a UTF-8 locale */
506 	if (((dot_ptr = strchr((char *)s1, '.')) != NULL) &&
507 	    (*dot_ptr != '\0') && (strncmp(dot_ptr + 1, "UTF-8", 5) == 0) &&
508 	    (*(dot_ptr + 6) == '\0' || *(dot_ptr + 6) == '@')) {
509 		s1_is_utf8++;
510 	}
511 
512 	/* check if s2 is a UTF-8 locale */
513 	if (((dot_ptr = strchr((char *)s2, '.')) != NULL) &&
514 	    (*dot_ptr != '\0') && (strncmp(dot_ptr + 1, "UTF-8", 5) == 0) &&
515 	    (*(dot_ptr + 6) == '\0' || *(dot_ptr + 6) == '@')) {
516 		s2_is_utf8++;
517 	}
518 
519 	/* prefer UTF-8 locales */
520 	if (s1_is_utf8 && !s2_is_utf8)
521 		return (-1);
522 
523 	if (s2_is_utf8 && !s1_is_utf8)
524 		return (1);
525 
526 	/* prefer any locale over the default locales */
527 	if (strcmp(s1, "C") == 0 || strcmp(s1, "POSIX") == 0 ||
528 	    strcmp(s1, "common") == 0) {
529 		if (strcmp(s2, "C") != 0 && strcmp(s2, "POSIX") != 0 &&
530 		    strcmp(s2, "common") != 0)
531 			return (1);
532 	}
533 
534 	if (strcmp(s2, "C") == 0 || strcmp(s2, "POSIX") == 0 ||
535 	    strcmp(s2, "common") == 0) {
536 		if (strcmp(s1, "C") != 0 &&
537 		    strcmp(s1, "POSIX") != 0 &&
538 		    strcmp(s1, "common") != 0)
539 			return (-1);
540 	}
541 
542 	return (strcmp(s1, s2));
543 }
544 
545 
546 char **
547 g11n_langtag_set_locale_set_intersect(char *langtag_set, char **locale_set)
548 {
549 	char **langtag_list, **result, **p, **q, **r;
550 	char *s;
551 	uint_t do_append, n_langtags, n_locales, n_results, max_results;
552 
553 	/* count lang tags and locales */
554 	for (n_locales = 0, p = locale_set; p && *p; p++)
555 		n_locales++;
556 
557 	n_langtags = ((s = langtag_set) != NULL && *s && *s != ',') ? 1 : 0;
558 	/* count the number of langtags */
559 	for (; s = strchr(s, ','); s++, n_langtags++)
560 		;
561 
562 	qsort(locale_set, n_locales, sizeof (char *), locale_cmp);
563 
564 	langtag_list = xsplit(langtag_set, ',');
565 	for (n_langtags = 0, p = langtag_list; p && *p; p++, n_langtags++)
566 		;
567 
568 	max_results = MIN(n_locales, n_langtags) * 2;
569 	result = (char **) xmalloc(sizeof (char *) * (max_results + 1));
570 	*result = NULL;
571 	n_results = 0;
572 
573 	/* more specific matches first */
574 	for (p = langtag_list; p && *p; p++) {
575 		do_append = 0;
576 		for (q = locale_set; q && *q; q++) {
577 			if (g11n_langtag_matches_locale(*p, *q) == 2) {
578 				do_append = 1;
579 				for (r = result; (r - result) <=
580 				    MIN(n_locales, n_langtags); r++) {
581 					if (!*r)
582 						break;
583 					if (strcmp(*q, *r) == 0) {
584 						do_append = 0;
585 						break;
586 					}
587 				}
588 				if (do_append && n_results < max_results) {
589 					result[n_results++] = xstrdup(*q);
590 					result[n_results] = NULL;
591 				}
592 				break;
593 			}
594 		}
595 	}
596 
597 	for (p = langtag_list; p && *p; p++) {
598 		do_append = 0;
599 		for (q = locale_set; q && *q; q++) {
600 			if (g11n_langtag_matches_locale(*p, *q) == 1) {
601 				do_append = 1;
602 				for (r = result; (r - result) <=
603 				    MIN(n_locales, n_langtags); r++) {
604 					if (!*r)
605 						break;
606 					if (strcmp(*q, *r) == 0) {
607 						do_append = 0;
608 						break;
609 					}
610 				}
611 				if (do_append && n_results < max_results) {
612 					result[n_results++] = xstrdup(*q);
613 					result[n_results] = NULL;
614 				}
615 				break;
616 			}
617 		}
618 	}
619 
620 	xfree_split_list(langtag_list);
621 
622 	return (result);
623 }
624 
625 char *
626 g11n_srvr_locale_negotiate(char *clnt_langtags, char **srvr_locales)
627 {
628 	char **results, *result = NULL;
629 
630 	if ((results = g11n_langtag_set_locale_set_intersect(clnt_langtags,
631 	    srvr_locales ?  srvr_locales : g11n_getlocales())) == NULL)
632 		return (NULL);
633 
634 	if (*results != NULL)
635 		result = xstrdup(*results);
636 
637 	xfree_split_list(results);
638 
639 	return (result);
640 }
641 
642 
643 /*
644  * Functions for validating ASCII and UTF-8 strings
645  *
646  * The error_str parameter is an optional pointer to a char variable
647  * where to store a string suitable for use with error() or fatal() or
648  * friends.
649  *
650  * The return value is 0 if success, EILSEQ or EINVAL.
651  *
652  */
653 uint_t
654 g11n_validate_ascii(const char *str, uint_t len, uchar_t **error_str)
655 {
656 	uchar_t *p;
657 
658 	for (p = (uchar_t *)str; p && *p && (!(*p & 0x80)); p++)
659 		;
660 
661 	if (len && ((p - (uchar_t *)str) != len))
662 		return (EILSEQ);
663 
664 	return (0);
665 }
666 
667 uint_t
668 g11n_validate_utf8(const uchar_t *str, uint_t len, uchar_t **error_str)
669 {
670 	uchar_t *p;
671 	uint_t c, l;
672 
673 	if (len == 0)
674 		len = strlen((const char *)str);
675 
676 	for (p = (uchar_t *)str; p && (p - str < len) && *p; ) {
677 		/* 8-bit chars begin a UTF-8 sequence */
678 		if (*p & 0x80) {
679 			/* get sequence length and sanity check first byte */
680 			if (*p < 0xc0)
681 				return (EILSEQ);
682 			else if (*p < 0xe0)
683 				l = 2;
684 			else if (*p < 0xf0)
685 				l = 3;
686 			else if (*p < 0xf8)
687 				l = 4;
688 			else if (*p < 0xfc)
689 				l = 5;
690 			else if (*p < 0xfe)
691 				l = 6;
692 			else
693 				return (EILSEQ);
694 
695 			if ((p + l - str) >= len)
696 				return (EILSEQ);
697 
698 			/* overlong detection - build codepoint */
699 			c = *p & 0x3f;
700 			/* shift c bits from first byte */
701 			c = c << (6 * (l - 1));
702 
703 			if (l > 1) {
704 				if (*(p + 1) && ((*(p + 1) & 0xc0) == 0x80))
705 					c = c | ((*(p + 1) & 0x3f) <<
706 					    (6 * (l - 2)));
707 				else
708 					return (EILSEQ);
709 
710 				if (c < 0x80)
711 					return (EILSEQ);
712 			}
713 
714 			if (l > 2) {
715 				if (*(p + 2) && ((*(p + 2) & 0xc0) == 0x80))
716 					c = c | ((*(p + 2) & 0x3f) <<
717 					    (6 * (l - 3)));
718 				else
719 					return (EILSEQ);
720 
721 				if (c < 0x800)
722 					return (EILSEQ);
723 			}
724 
725 			if (l > 3) {
726 				if (*(p + 3) && ((*(p + 3) & 0xc0) == 0x80))
727 					c = c | ((*(p + 3) & 0x3f) <<
728 					    (6 * (l - 4)));
729 				else
730 					return (EILSEQ);
731 
732 				if (c < 0x10000)
733 					return (EILSEQ);
734 			}
735 
736 			if (l > 4) {
737 				if (*(p + 4) && ((*(p + 4) & 0xc0) == 0x80))
738 					c = c | ((*(p + 4) & 0x3f) <<
739 					    (6 * (l - 5)));
740 				else
741 					return (EILSEQ);
742 
743 				if (c < 0x200000)
744 					return (EILSEQ);
745 			}
746 
747 			if (l > 5) {
748 				if (*(p + 5) && ((*(p + 5) & 0xc0) == 0x80))
749 					c = c | (*(p + 5) & 0x3f);
750 				else
751 					return (EILSEQ);
752 
753 				if (c < 0x4000000)
754 					return (EILSEQ);
755 			}
756 
757 			/*
758 			 * check for UTF-16 surrogates ifs other illegal
759 			 * UTF-8 * points
760 			 */
761 			if (((c <= 0xdfff) && (c >= 0xd800)) ||
762 			    (c == 0xfffe) || (c == 0xffff))
763 				return (EILSEQ);
764 			p += l;
765 		}
766 		/* 7-bit chars are fine */
767 		else
768 			p++;
769 	}
770 	return (0);
771 }
772 
773 /*
774  * Functions for converting to ASCII or UTF-8 from the local codeset
775  * Functions for converting from ASCII or UTF-8 to the local codeset
776  *
777  * The error_str parameter is an optional pointer to a char variable
778  * where to store a string suitable for use with error() or fatal() or
779  * friends.
780  *
781  * The err parameter is an optional pointer to an integer where 0
782  * (success) or EILSEQ or EINVAL will be stored (failure).
783  *
784  * These functions return NULL if the conversion fails.
785  *
786  */
787 uchar_t *
788 g11n_convert_from_ascii(const char *str, int *err_ptr, uchar_t **error_str)
789 {
790 	static uint_t initialized = 0;
791 	static uint_t do_convert = 0;
792 	iconv_t cd;
793 	int err;
794 
795 	if (!initialized) {
796 		/*
797 		 * iconv_open() fails if the to/from codesets are the
798 		 * same, and there are aliases of codesets to boot...
799 		 */
800 		if (strcmp("646", nl_langinfo(CODESET)) == 0 ||
801 			strcmp("ASCII",  nl_langinfo(CODESET)) == 0 ||
802 			strcmp("US-ASCII",  nl_langinfo(CODESET)) == 0) {
803 			initialized = 1;
804 			do_convert = 0;
805 		} else {
806 			cd = iconv_open(nl_langinfo(CODESET), "646");
807 			if (cd == (iconv_t)-1) {
808 				if (err_ptr)
809 					*err_ptr = errno;
810 				if (error_str)
811 					*error_str = (uchar_t *)"Cannot "
812 					    "convert ASCII strings to the local"
813 					    " codeset";
814 			}
815 			initialized = 1;
816 			do_convert = 1;
817 		}
818 	}
819 
820 	if (!do_convert) {
821 		if ((err = g11n_validate_ascii(str, 0, error_str))) {
822 			if (err_ptr)
823 				*err_ptr = err;
824 			return (NULL);
825 		} else
826 			return ((uchar_t *)xstrdup(str));
827 	}
828 
829 	return (do_iconv(cd, NULL, str, 0, NULL, err_ptr, error_str));
830 }
831 
832 uchar_t *
833 g11n_convert_from_utf8(const uchar_t *str, int *err_ptr, uchar_t **error_str)
834 {
835 	static uint_t initialized = 0;
836 	static uint_t do_convert = 0;
837 	iconv_t cd;
838 	int err;
839 
840 	if (!initialized) {
841 		/*
842 		 * iconv_open() fails if the to/from codesets are the
843 		 * same, and there are aliases of codesets to boot...
844 		 */
845 		if (strcmp("UTF-8", nl_langinfo(CODESET)) == 0 ||
846 		    strcmp("UTF8",  nl_langinfo(CODESET)) == 0) {
847 			initialized = 1;
848 			do_convert = 0;
849 		} else {
850 			cd = iconv_open(nl_langinfo(CODESET), "UTF-8");
851 			if (cd == (iconv_t)-1) {
852 				if (err_ptr)
853 					*err_ptr = errno;
854 				if (error_str)
855 					*error_str = (uchar_t *)"Cannot "
856 					    "convert UTF-8 strings to the "
857 					    "local codeset";
858 			}
859 			initialized = 1;
860 			do_convert = 1;
861 		}
862 	}
863 
864 	if (!do_convert) {
865 		if ((err = g11n_validate_utf8(str, 0, error_str))) {
866 			if (err_ptr)
867 				*err_ptr = err;
868 			return (NULL);
869 		} else
870 			return ((uchar_t *)xstrdup((char *)str));
871 	}
872 
873 	return (do_iconv(cd, NULL, str, 0, NULL, err_ptr, error_str));
874 }
875 
876 char *
877 g11n_convert_to_ascii(const uchar_t *str, int *err_ptr, uchar_t **error_str)
878 {
879 	static uint_t initialized = 0;
880 	static uint_t do_convert = 0;
881 	iconv_t cd;
882 
883 	if (!initialized) {
884 		/*
885 		 * iconv_open() fails if the to/from codesets are the
886 		 * same, and there are aliases of codesets to boot...
887 		 */
888 		if (strcmp("646", nl_langinfo(CODESET)) == 0 ||
889 		    strcmp("ASCII",  nl_langinfo(CODESET)) == 0 ||
890 		    strcmp("US-ASCII",  nl_langinfo(CODESET)) == 0) {
891 			initialized = 1;
892 			do_convert = 0;
893 		} else {
894 			cd = iconv_open("646", nl_langinfo(CODESET));
895 			if (cd == (iconv_t)-1) {
896 				if (err_ptr)
897 					*err_ptr = errno;
898 				if (error_str)
899 					*error_str = (uchar_t *)"Cannot "
900 					    "convert UTF-8 strings to the "
901 					    "local codeset";
902 			}
903 			initialized = 1;
904 			do_convert = 1;
905 		}
906 	}
907 
908 	if (!do_convert)
909 		return (xstrdup((char *)str));
910 
911 	return ((char *)do_iconv(cd, NULL, str, 0, NULL, err_ptr, error_str));
912 }
913 
914 uchar_t *
915 g11n_convert_to_utf8(const uchar_t *str, int *err_ptr, uchar_t **error_str)
916 {
917 	static uint_t initialized = 0;
918 	static uint_t do_convert = 0;
919 	iconv_t cd;
920 
921 	if (!initialized) {
922 		/*
923 		 * iconv_open() fails if the to/from codesets are the
924 		 * same, and there are aliases of codesets to boot...
925 		 */
926 		if (strcmp("UTF-8", nl_langinfo(CODESET)) == 0 ||
927 		    strcmp("UTF8",  nl_langinfo(CODESET)) == 0) {
928 			initialized = 1;
929 			do_convert = 0;
930 		} else {
931 			cd = iconv_open("UTF-8", nl_langinfo(CODESET));
932 			if (cd == (iconv_t)-1) {
933 				if (err_ptr)
934 					*err_ptr = errno;
935 				if (error_str)
936 					*error_str = (uchar_t *)"Cannot "
937 					    "convert UTF-8 strings to the "
938 					    "local codeset";
939 			}
940 			initialized = 1;
941 			do_convert = 1;
942 		}
943 	}
944 
945 	if (!do_convert)
946 		return ((uchar_t *)xstrdup((char *)str));
947 
948 	return (do_iconv(cd, NULL, str, 0, NULL, err_ptr, error_str));
949 }
950 
951 
952 /*
953  * Wrapper around iconv()
954  *
955  * The caller is responsible for freeing the result and for handling
956  * (errno && errno != E2BIG) (i.e., EILSEQ, EINVAL, EBADF).
957  */
958 static uchar_t *
959 do_iconv(iconv_t cd, uint_t *mul_ptr, const void *buf, uint_t len,
960     uint_t *outlen, int *err, uchar_t **err_str)
961 {
962 	size_t inbytesleft, outbytesleft, converted_size;
963 	char *outbuf;
964 	uchar_t *converted;
965 	const char *inbuf;
966 	uint_t mul = 0;
967 
968 	if (!buf || !(*(char *)buf))
969 		return (NULL);
970 
971 	if (len == 0)
972 		len = strlen(buf);
973 
974 	/* reset conversion descriptor */
975 	/* XXX Do we need initial shift sequences for UTF-8??? */
976 	(void) iconv(cd, NULL, &inbytesleft, &outbuf, &outbytesleft);
977 	inbuf = (const char *) buf;
978 
979 	if (mul_ptr)
980 		mul = *mul_ptr;
981 
982 	converted_size = (len << mul);
983 	outbuf = (char *)xmalloc(converted_size + 1); /* for null */
984 	converted = (uchar_t *)outbuf;
985 	outbytesleft = len;
986 
987 	do {
988 		if (iconv(cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft) ==
989 		    (size_t)-1) {
990 			if (errno == E2BIG) {
991 				/* UTF-8 codepoints are at most 8 bytes long */
992 				if (mul > 2) {
993 					if (err_str)
994 						*err_str = (uchar_t *)
995 						    "Conversion to UTF-8 failed"
996 						    " due to preposterous space"
997 						    " requirements";
998 					if (err)
999 						*err = EILSEQ;
1000 					return (NULL);
1001 				}
1002 
1003 				/*
1004 				 * re-alloc output and ensure that the outbuf
1005 				 * and outbytesleft values are adjusted
1006 				 */
1007 				converted = xrealloc(converted,
1008 				    converted_size << 1 + 1);
1009 				outbuf = (char *)converted + converted_size -
1010 				    outbytesleft;
1011 				converted_size = (len << ++(mul));
1012 				outbytesleft = converted_size - outbytesleft;
1013 			} else {
1014 				/*
1015 				 * let the caller deal with iconv() errors,
1016 				 * probably by calling fatal(); xfree() does
1017 				 * not set errno
1018 				 */
1019 				if (err)
1020 					*err = errno;
1021 				xfree(converted);
1022 				return (NULL);
1023 			}
1024 		}
1025 	} while (inbytesleft);
1026 
1027 	*outbuf = '\0'; /* ensure null-termination */
1028 	if (outlen)
1029 		*outlen = converted_size - outbytesleft;
1030 	if (mul_ptr)
1031 		*mul_ptr = mul;
1032 
1033 	return (converted);
1034 }
1035