Line data Source code
1 : /*
2 : Unix SMB/CIFS implementation.
3 : minimal iconv implementation
4 : Copyright (C) Andrew Tridgell 2001
5 : Copyright (C) Jelmer Vernooij 2002
6 :
7 : This program is free software; you can redistribute it and/or modify
8 : it under the terms of the GNU General Public License as published by
9 : the Free Software Foundation; either version 3 of the License, or
10 : (at your option) any later version.
11 :
12 : This program is distributed in the hope that it will be useful,
13 : but WITHOUT ANY WARRANTY; without even the implied warranty of
14 : MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 : GNU General Public License for more details.
16 :
17 : You should have received a copy of the GNU General Public License
18 : along with this program. If not, see <http://www.gnu.org/licenses/>.
19 : */
20 :
21 : #include "replace.h"
22 : #include "system/iconv.h"
23 : #include "system/filesys.h"
24 : #include "lib/util/byteorder.h"
25 : #include "lib/util/dlinklist.h"
26 : #include "lib/util/charset/charset.h"
27 : #include "lib/util/charset/charset_proto.h"
28 :
29 : #ifdef HAVE_ICU_I18N
30 : #include <unicode/ustring.h>
31 : #include <unicode/utrans.h>
32 : #endif
33 :
34 : #ifdef strcasecmp
35 : #undef strcasecmp
36 : #endif
37 :
38 : /**
39 : * @file
40 : *
41 : * @brief Samba wrapper/stub for iconv character set conversion.
42 : *
43 : * iconv is the XPG2 interface for converting between character
44 : * encodings. This file provides a Samba wrapper around it, and also
45 : * a simple reimplementation that is used if the system does not
46 : * implement iconv.
47 : *
48 : * Samba only works with encodings that are supersets of ASCII: ascii
49 : * characters like whitespace can be tested for directly, multibyte
50 : * sequences start with a byte with the high bit set, and strings are
51 : * terminated by a nul byte.
52 : *
53 : * Note that the only function provided by iconv is conversion between
54 : * characters. It doesn't directly support operations like
55 : * uppercasing or comparison. We have to convert to UTF-16LE and
56 : * compare there.
57 : *
58 : * @sa Samba Developers Guide
59 : **/
60 :
61 : static size_t ascii_pull (void *,const char **, size_t *, char **, size_t *);
62 : static size_t ascii_push (void *,const char **, size_t *, char **, size_t *);
63 : static size_t latin1_pull(void *,const char **, size_t *, char **, size_t *);
64 : static size_t latin1_push(void *,const char **, size_t *, char **, size_t *);
65 : static size_t utf8_pull (void *,const char **, size_t *, char **, size_t *);
66 : static size_t utf8_push (void *,const char **, size_t *, char **, size_t *);
67 : static size_t utf16_munged_pull(void *,const char **, size_t *, char **, size_t *);
68 : static size_t ucs2hex_pull(void *,const char **, size_t *, char **, size_t *);
69 : static size_t ucs2hex_push(void *,const char **, size_t *, char **, size_t *);
70 : static size_t iconv_copy (void *,const char **, size_t *, char **, size_t *);
71 : static size_t iconv_swab (void *,const char **, size_t *, char **, size_t *);
72 :
73 : static const struct charset_functions builtin_functions[] = {
74 : /* windows is closest to UTF-16 */
75 : {
76 : .name = "UCS-2LE",
77 : .pull = iconv_copy,
78 : .push = iconv_copy
79 : },
80 : {
81 : .name = "UTF-16LE",
82 : .pull = iconv_copy,
83 : .push = iconv_copy
84 : },
85 : {
86 : .name = "UCS-2BE",
87 : .pull = iconv_swab,
88 : .push = iconv_swab
89 : },
90 : {
91 : .name = "UTF-16BE",
92 : .pull = iconv_swab,
93 : .push = iconv_swab
94 : },
95 :
96 : /* we include the UTF-8 alias to cope with differing locale settings */
97 : {
98 : .name = "UTF8",
99 : .pull = utf8_pull,
100 : .push = utf8_push
101 : },
102 : {
103 : .name = "UTF-8",
104 : .pull = utf8_pull,
105 : .push = utf8_push
106 : },
107 :
108 : /* this handles the munging needed for String2Key */
109 : {
110 : .name = "UTF16_MUNGED",
111 : .pull = utf16_munged_pull,
112 : .push = iconv_copy,
113 : .samba_internal_charset = true
114 : },
115 :
116 : {
117 : .name = "ASCII",
118 : .pull = ascii_pull,
119 : .push = ascii_push
120 : },
121 : {
122 : .name = "646",
123 : .pull = ascii_pull,
124 : .push = ascii_push
125 : },
126 : {
127 : .name = "ISO-8859-1",
128 : .pull = latin1_pull,
129 : .push = latin1_push
130 : },
131 : #ifdef DEVELOPER
132 : {
133 : .name = "WEIRD",
134 : .pull = weird_pull,
135 : .push = weird_push,
136 : .samba_internal_charset = true
137 : },
138 : #endif
139 : #ifdef DARWINOS
140 : {
141 : .name = "MACOSXFS",
142 : .pull = macosxfs_encoding_pull,
143 : .push = macosxfs_encoding_push,
144 : .samba_internal_charset = true
145 : },
146 : #endif
147 : {
148 : .name = "UCS2-HEX",
149 : .pull = ucs2hex_pull,
150 : .push = ucs2hex_push,
151 : .samba_internal_charset = true
152 : }
153 : };
154 :
155 : #ifdef HAVE_NATIVE_ICONV
156 : /* if there was an error then reset the internal state,
157 : this ensures that we don't have a shift state remaining for
158 : character sets like SJIS */
159 5243916 : static size_t sys_iconv(void *cd,
160 : const char **inbuf, size_t *inbytesleft,
161 : char **outbuf, size_t *outbytesleft)
162 : {
163 5243916 : size_t ret = iconv((iconv_t)cd,
164 : discard_const_p(char *, inbuf), inbytesleft,
165 : outbuf, outbytesleft);
166 5243916 : if (ret == (size_t)-1) iconv(cd, NULL, NULL, NULL, NULL);
167 5243916 : return ret;
168 : }
169 : #endif
170 :
171 : #ifdef HAVE_ICU_I18N
172 74 : static size_t sys_uconv(void *cd,
173 : const char **inbuf,
174 : size_t *inbytesleft,
175 : char **outbuf,
176 : size_t *outbytesleft)
177 74 : {
178 74 : UTransliterator *t = (UTransliterator *)cd;
179 74 : size_t bufsize = *inbytesleft * 2;
180 74 : UChar ustr[bufsize];
181 74 : UChar *up = NULL;
182 74 : char *p = NULL;
183 14 : int32_t ustrlen;
184 14 : int32_t limit;
185 14 : int32_t converted_len;
186 14 : size_t inbuf_consumed;
187 14 : size_t outbut_consumed;
188 14 : UErrorCode ue;
189 :
190 : /* Convert from UTF8 to UCS2 */
191 74 : ue = 0;
192 74 : up = u_strFromUTF8(ustr, /* dst */
193 : bufsize, /* dst buflen */
194 : &converted_len, /* dst written */
195 : *inbuf, /* src */
196 60 : *inbytesleft, /* src length */
197 : &ue);
198 74 : if (up == NULL || U_FAILURE(ue)) {
199 0 : return -1;
200 : }
201 74 : if (converted_len > bufsize) {
202 : /*
203 : * u_strFromUTF8() returns the required size in
204 : * converted_len. In theory this should never overflow as the
205 : * ustr[] array is allocated with a size twice as big as
206 : * inbytesleft and converted_len should be equal to inbytesleft,
207 : * but you never know...
208 : */
209 0 : errno = EOVERFLOW;
210 0 : return -1;
211 : }
212 74 : inbuf_consumed = converted_len;
213 :
214 : /*
215 : * The following transliteration function takes two parameters, the
216 : * length of the text to be converted (converted_len) and a limit which
217 : * may be smaller then converted_len. We just set limit to converted_len
218 : * and also ignore the value returned in limit.
219 : */
220 74 : limit = converted_len;
221 :
222 : /* Inplace transliteration */
223 74 : utrans_transUChars(t,
224 : ustr, /* text */
225 : &converted_len, /* text length */
226 : bufsize, /* text buflen */
227 : 0, /* start */
228 : &limit, /* limit */
229 : &ue);
230 74 : if (U_FAILURE(ue)) {
231 0 : return -1;
232 : }
233 74 : if (converted_len > bufsize) {
234 : /*
235 : * In theory this should never happen as the ustr[] array is
236 : * allocated with a size twice as big as inbytesleft and
237 : * converted_len should be equal to inbytesleft, but you never
238 : * know...
239 : */
240 0 : errno = EOVERFLOW;
241 0 : return -1;
242 : }
243 74 : ustrlen = converted_len;
244 :
245 : /* Convert from UCS2 back to UTF8 */
246 74 : ue = 0;
247 88 : p = u_strToUTF8(*outbuf, /* dst */
248 74 : *outbytesleft, /* dst buflen */
249 : &converted_len, /* dst required length */
250 : ustr, /* src */
251 : ustrlen, /* src length */
252 : &ue);
253 74 : if (p == NULL || U_FAILURE(ue)) {
254 0 : return -1;
255 : }
256 :
257 68 : outbut_consumed = converted_len;
258 68 : if (converted_len > *outbytesleft) {
259 : /*
260 : * The caller's result buffer is too small...
261 : */
262 0 : outbut_consumed = *outbytesleft;
263 : }
264 :
265 68 : *inbuf += inbuf_consumed;
266 68 : *inbytesleft -= inbuf_consumed;
267 68 : *outbuf += outbut_consumed;
268 68 : *outbytesleft -= outbut_consumed;
269 :
270 68 : return converted_len;
271 : }
272 : #endif
273 :
274 : /**
275 : * This is a simple portable iconv() implementation.
276 : *
277 : * It only knows about a very small number of character sets - just
278 : * enough that Samba works on systems that don't have iconv.
279 : **/
280 151154346 : _PUBLIC_ size_t smb_iconv(smb_iconv_t cd,
281 : const char **inbuf, size_t *inbytesleft,
282 : char **outbuf, size_t *outbytesleft)
283 : {
284 : /* in many cases we can go direct */
285 151154346 : if (cd->direct) {
286 38571565 : return cd->direct(cd->cd_direct,
287 : inbuf, inbytesleft, outbuf, outbytesleft);
288 : }
289 :
290 : /* otherwise we have to do it chunks at a time */
291 : {
292 : #ifndef SMB_ICONV_BUFSIZE
293 : #define SMB_ICONV_BUFSIZE 2048
294 : #endif
295 : size_t bufsize;
296 : char cvtbuf[SMB_ICONV_BUFSIZE];
297 :
298 225393302 : while (*inbytesleft > 0) {
299 112810561 : char *bufp1 = cvtbuf;
300 112810561 : const char *bufp2 = cvtbuf;
301 112810561 : int saved_errno = errno;
302 112810561 : bool pull_failed = false;
303 112810561 : bufsize = SMB_ICONV_BUFSIZE;
304 :
305 112810561 : if (cd->pull(cd->cd_pull,
306 : inbuf, inbytesleft, &bufp1, &bufsize) == -1
307 227790 : && errno != E2BIG) {
308 10 : saved_errno = errno;
309 10 : pull_failed = true;
310 : }
311 :
312 112810561 : bufsize = SMB_ICONV_BUFSIZE - bufsize;
313 :
314 112810561 : if (cd->push(cd->cd_push,
315 : &bufp2, &bufsize,
316 : outbuf, outbytesleft) == -1) {
317 40 : return -1;
318 112810531 : } else if (pull_failed) {
319 : /* We want the pull errno if possible */
320 10 : errno = saved_errno;
321 10 : return -1;
322 : }
323 : }
324 : }
325 :
326 112582741 : return 0;
327 : }
328 :
329 7422153 : static bool is_utf16(const char *name)
330 : {
331 8198863 : return strcasecmp(name, "UCS-2LE") == 0 ||
332 7422153 : strcasecmp(name, "UTF-16LE") == 0;
333 : }
334 :
335 4254984 : static int smb_iconv_t_destructor(smb_iconv_t hwd)
336 : {
337 : #ifdef HAVE_ICU_I18N
338 : /*
339 : * This has to come first, as the cd_direct member won't be an iconv
340 : * handle and must not be passed to iconv_close().
341 : */
342 4254984 : if (hwd->direct == sys_uconv) {
343 40 : utrans_close(hwd->cd_direct);
344 40 : return 0;
345 : }
346 : #endif
347 : #ifdef HAVE_NATIVE_ICONV
348 4254944 : if (hwd->cd_pull != NULL && hwd->cd_pull != (iconv_t)-1)
349 28299 : iconv_close(hwd->cd_pull);
350 4254944 : if (hwd->cd_push != NULL && hwd->cd_push != (iconv_t)-1)
351 11678 : iconv_close(hwd->cd_push);
352 4254944 : if (hwd->cd_direct != NULL && hwd->cd_direct != (iconv_t)-1)
353 1000038 : iconv_close(hwd->cd_direct);
354 : #endif
355 :
356 161659 : return 0;
357 : }
358 :
359 4386502 : _PUBLIC_ smb_iconv_t smb_iconv_open_ex(TALLOC_CTX *mem_ctx, const char *tocode,
360 : const char *fromcode, bool use_builtin_handlers)
361 : {
362 4096479 : smb_iconv_t ret;
363 4386502 : const struct charset_functions *from=NULL, *to=NULL;
364 4096479 : int i;
365 :
366 4386502 : ret = (smb_iconv_t)talloc_named(mem_ctx,
367 : sizeof(*ret),
368 : "iconv(%s,%s)", tocode, fromcode);
369 4386502 : if (!ret) {
370 0 : errno = ENOMEM;
371 0 : return (smb_iconv_t)-1;
372 : }
373 4386502 : memset(ret, 0, sizeof(*ret));
374 4386502 : talloc_set_destructor(ret, smb_iconv_t_destructor);
375 :
376 : /* check for the simplest null conversion */
377 4386502 : if (strcmp(fromcode, tocode) == 0) {
378 275 : ret->direct = iconv_copy;
379 275 : return ret;
380 : }
381 :
382 : /* check if we have a builtin function for this conversion */
383 57020951 : for (i=0;i<ARRAY_SIZE(builtin_functions);i++) {
384 52634724 : if (strcasecmp(fromcode, builtin_functions[i].name) == 0) {
385 3831037 : if (use_builtin_handlers || builtin_functions[i].samba_internal_charset) {
386 3831009 : from = &builtin_functions[i];
387 : }
388 : }
389 52634724 : if (strcasecmp(tocode, builtin_functions[i].name) == 0) {
390 3862113 : if (use_builtin_handlers || builtin_functions[i].samba_internal_charset) {
391 3862092 : to = &builtin_functions[i];
392 : }
393 : }
394 : }
395 :
396 : #ifdef HAVE_NATIVE_ICONV
397 : /* the from and to variables indicate a samba module or
398 : * internal conversion, ret->pull and ret->push are
399 : * initialised only in this block for iconv based
400 : * conversions */
401 :
402 4386227 : if (from == NULL) {
403 555218 : ret->cd_pull = iconv_open("UTF-16LE", fromcode);
404 555218 : if (ret->cd_pull == (iconv_t)-1)
405 46 : ret->cd_pull = iconv_open("UCS-2LE", fromcode);
406 555218 : if (ret->cd_pull != (iconv_t)-1) {
407 555172 : ret->pull = sys_iconv;
408 : }
409 : }
410 :
411 4386227 : if (to == NULL) {
412 524135 : ret->cd_push = iconv_open(tocode, "UTF-16LE");
413 524135 : if (ret->cd_push == (iconv_t)-1)
414 46 : ret->cd_push = iconv_open(tocode, "UCS-2LE");
415 524135 : if (ret->cd_push != (iconv_t)-1) {
416 524089 : ret->push = sys_iconv;
417 : }
418 : }
419 : #endif
420 :
421 : #ifdef HAVE_ICU_I18N
422 4386227 : if (strcasecmp(fromcode, "UTF8-NFD") == 0 &&
423 22 : strcasecmp(tocode, "UTF8-NFC") == 0)
424 : {
425 2 : U_STRING_DECL(t, "any-nfc", 7);
426 22 : UErrorCode ue = 0;
427 :
428 2 : U_STRING_INIT(t, "any-nfc", 7);
429 :
430 22 : ret->cd_direct = utrans_openU(t,
431 : strlen("any-nfc"),
432 : UTRANS_FORWARD,
433 : NULL,
434 : 0,
435 : NULL,
436 : &ue);
437 22 : if (U_FAILURE(ue)) {
438 0 : return (smb_iconv_t)-1;
439 : }
440 22 : ret->direct = sys_uconv;
441 22 : return ret;
442 : }
443 :
444 4386205 : if (strcasecmp(fromcode, "UTF8-NFC") == 0 &&
445 24 : strcasecmp(tocode, "UTF8-NFD") == 0)
446 : {
447 4 : U_STRING_DECL(tname, "any-nfd", 7);
448 24 : UErrorCode ue = 0;
449 :
450 4 : U_STRING_INIT(tname, "any-nfd", 7);
451 :
452 24 : ret->cd_direct = utrans_openU(tname,
453 : 7,
454 : UTRANS_FORWARD,
455 : NULL,
456 : 0,
457 : NULL,
458 : &ue);
459 24 : if (U_FAILURE(ue)) {
460 0 : return (smb_iconv_t)-1;
461 : }
462 24 : ret->direct = sys_uconv;
463 24 : return ret;
464 : }
465 : #endif
466 :
467 4386181 : if (ret->pull == NULL && from == NULL) {
468 0 : goto failed;
469 : }
470 :
471 4386181 : if (ret->push == NULL && to == NULL) {
472 0 : goto failed;
473 : }
474 :
475 : /* check for conversion to/from ucs2 */
476 4386181 : if (is_utf16(fromcode) && to) {
477 1633859 : ret->direct = to->push;
478 1633859 : return ret;
479 : }
480 2752322 : if (is_utf16(tocode) && from) {
481 1608261 : ret->direct = from->pull;
482 1608261 : return ret;
483 : }
484 :
485 : #ifdef HAVE_NATIVE_ICONV
486 1144061 : if (is_utf16(fromcode)) {
487 500024 : ret->direct = sys_iconv;
488 500024 : ret->cd_direct = ret->cd_push;
489 500024 : ret->cd_push = NULL;
490 500024 : return ret;
491 : }
492 644037 : if (is_utf16(tocode)) {
493 500014 : ret->direct = sys_iconv;
494 500014 : ret->cd_direct = ret->cd_pull;
495 500014 : ret->cd_pull = NULL;
496 500014 : return ret;
497 : }
498 : #endif
499 :
500 : /* the general case has to go via a buffer */
501 144023 : if (!ret->pull) ret->pull = from->pull;
502 144023 : if (!ret->push) ret->push = to->push;
503 141819 : return ret;
504 :
505 0 : failed:
506 0 : talloc_free(ret);
507 0 : errno = EINVAL;
508 0 : return (smb_iconv_t)-1;
509 : }
510 :
511 : /*
512 : simple iconv_open() wrapper
513 : */
514 6 : _PUBLIC_ smb_iconv_t smb_iconv_open(const char *tocode, const char *fromcode)
515 : {
516 6 : return smb_iconv_open_ex(NULL, tocode, fromcode, true);
517 : }
518 :
519 : /*
520 : simple iconv_close() wrapper
521 : */
522 4254942 : _PUBLIC_ int smb_iconv_close(smb_iconv_t cd)
523 : {
524 4254942 : talloc_free(cd);
525 4254942 : return 0;
526 : }
527 :
528 :
529 : /**********************************************************************
530 : the following functions implement the builtin character sets in Samba
531 : and also the "test" character sets that are designed to test
532 : multi-byte character set support for english users
533 : ***********************************************************************/
534 :
535 : /*
536 : this takes an ASCII sequence and produces a UTF16 sequence
537 :
538 : The first 127 codepoints of latin1 matches the first 127 codepoints
539 : of unicode, and so can be put into the first byte of UTF16LE
540 :
541 : */
542 :
543 5 : static size_t ascii_pull(void *cd, const char **inbuf, size_t *inbytesleft,
544 : char **outbuf, size_t *outbytesleft)
545 : {
546 26 : while (*inbytesleft >= 1 && *outbytesleft >= 2) {
547 25 : if (((*inbuf)[0] & 0x7F) != (*inbuf)[0]) {
548 : /* If this is multi-byte, then it isn't legal ASCII */
549 4 : errno = EILSEQ;
550 4 : return -1;
551 : }
552 21 : (*outbuf)[0] = (*inbuf)[0];
553 21 : (*outbuf)[1] = 0;
554 21 : (*inbytesleft) -= 1;
555 21 : (*outbytesleft) -= 2;
556 21 : (*inbuf) += 1;
557 21 : (*outbuf) += 2;
558 : }
559 :
560 1 : if (*inbytesleft > 0) {
561 0 : errno = E2BIG;
562 0 : return -1;
563 : }
564 :
565 0 : return 0;
566 : }
567 :
568 : /*
569 : this takes a UTF16 sequence and produces an ASCII sequence
570 :
571 : The first 127 codepoints of ASCII matches the first 127 codepoints
572 : of unicode, and so can be read directly from the first byte of UTF16LE
573 :
574 : */
575 4 : static size_t ascii_push(void *cd, const char **inbuf, size_t *inbytesleft,
576 : char **outbuf, size_t *outbytesleft)
577 : {
578 4 : int ir_count=0;
579 :
580 5 : while (*inbytesleft >= 2 && *outbytesleft >= 1) {
581 4 : if (((*inbuf)[0] & 0x7F) != (*inbuf)[0] ||
582 1 : (*inbuf)[1] != 0) {
583 : /* If this is multi-byte, then it isn't legal ASCII */
584 3 : errno = EILSEQ;
585 3 : return -1;
586 : }
587 1 : (*outbuf)[0] = (*inbuf)[0];
588 1 : (*inbytesleft) -= 2;
589 1 : (*outbytesleft) -= 1;
590 1 : (*inbuf) += 2;
591 1 : (*outbuf) += 1;
592 : }
593 :
594 1 : if (*inbytesleft == 1) {
595 0 : errno = EINVAL;
596 0 : return -1;
597 : }
598 :
599 1 : if (*inbytesleft > 1) {
600 0 : errno = E2BIG;
601 0 : return -1;
602 : }
603 :
604 0 : return ir_count;
605 : }
606 :
607 : /*
608 : this takes a latin1/ISO-8859-1 sequence and produces a UTF16 sequence
609 :
610 : The first 256 codepoints of latin1 matches the first 256 codepoints
611 : of unicode, and so can be put into the first byte of UTF16LE
612 :
613 : */
614 151 : static size_t latin1_pull(void *cd, const char **inbuf, size_t *inbytesleft,
615 : char **outbuf, size_t *outbytesleft)
616 : {
617 2965 : while (*inbytesleft >= 1 && *outbytesleft >= 2) {
618 2814 : (*outbuf)[0] = (*inbuf)[0];
619 2814 : (*outbuf)[1] = 0;
620 2814 : (*inbytesleft) -= 1;
621 2814 : (*outbytesleft) -= 2;
622 2814 : (*inbuf) += 1;
623 2814 : (*outbuf) += 2;
624 : }
625 :
626 151 : if (*inbytesleft > 0) {
627 2 : errno = E2BIG;
628 2 : return -1;
629 : }
630 :
631 140 : return 0;
632 : }
633 :
634 : /*
635 : this takes a UTF16 sequence and produces a latin1/ISO-8859-1 sequence
636 :
637 : The first 256 codepoints of latin1 matches the first 256 codepoints
638 : of unicode, and so can be read directly from the first byte of UTF16LE
639 :
640 : */
641 211 : static size_t latin1_push(void *cd, const char **inbuf, size_t *inbytesleft,
642 : char **outbuf, size_t *outbytesleft)
643 : {
644 211 : int ir_count=0;
645 :
646 2293 : while (*inbytesleft >= 2 && *outbytesleft >= 1) {
647 2082 : (*outbuf)[0] = (*inbuf)[0];
648 2082 : if ((*inbuf)[1] != 0) {
649 : /* If this is multi-byte, then it isn't legal latin1 */
650 0 : errno = EILSEQ;
651 0 : return -1;
652 : }
653 2082 : (*inbytesleft) -= 2;
654 2082 : (*outbytesleft) -= 1;
655 2082 : (*inbuf) += 2;
656 2082 : (*outbuf) += 1;
657 : }
658 :
659 211 : if (*inbytesleft == 1) {
660 0 : errno = EINVAL;
661 0 : return -1;
662 : }
663 :
664 211 : if (*inbytesleft > 1) {
665 1 : errno = E2BIG;
666 1 : return -1;
667 : }
668 :
669 204 : return ir_count;
670 : }
671 :
672 0 : static size_t ucs2hex_pull(void *cd, const char **inbuf, size_t *inbytesleft,
673 : char **outbuf, size_t *outbytesleft)
674 : {
675 0 : while (*inbytesleft >= 1 && *outbytesleft >= 2) {
676 0 : uint8_t hi = 0, lo = 0;
677 0 : bool ok;
678 :
679 0 : if ((*inbuf)[0] != '@') {
680 : /* seven bit ascii case */
681 0 : (*outbuf)[0] = (*inbuf)[0];
682 0 : (*outbuf)[1] = 0;
683 0 : (*inbytesleft) -= 1;
684 0 : (*outbytesleft) -= 2;
685 0 : (*inbuf) += 1;
686 0 : (*outbuf) += 2;
687 0 : continue;
688 : }
689 : /* it's a hex character */
690 0 : if (*inbytesleft < 5) {
691 0 : errno = EINVAL;
692 0 : return -1;
693 : }
694 :
695 0 : ok = hex_byte(&(*inbuf)[1], &hi) && hex_byte(&(*inbuf)[3], &lo);
696 0 : if (!ok) {
697 0 : errno = EILSEQ;
698 0 : return -1;
699 : }
700 :
701 0 : (*outbuf)[0] = lo;
702 0 : (*outbuf)[1] = hi;
703 0 : (*inbytesleft) -= 5;
704 0 : (*outbytesleft) -= 2;
705 0 : (*inbuf) += 5;
706 0 : (*outbuf) += 2;
707 : }
708 :
709 0 : if (*inbytesleft > 0) {
710 0 : errno = E2BIG;
711 0 : return -1;
712 : }
713 :
714 0 : return 0;
715 : }
716 :
717 0 : static size_t ucs2hex_push(void *cd, const char **inbuf, size_t *inbytesleft,
718 : char **outbuf, size_t *outbytesleft)
719 : {
720 0 : while (*inbytesleft >= 2 && *outbytesleft >= 1) {
721 0 : char buf[6];
722 :
723 0 : if ((*inbuf)[1] == 0 &&
724 0 : ((*inbuf)[0] & 0x80) == 0 &&
725 0 : (*inbuf)[0] != '@') {
726 0 : (*outbuf)[0] = (*inbuf)[0];
727 0 : (*inbytesleft) -= 2;
728 0 : (*outbytesleft) -= 1;
729 0 : (*inbuf) += 2;
730 0 : (*outbuf) += 1;
731 0 : continue;
732 : }
733 0 : if (*outbytesleft < 5) {
734 0 : errno = E2BIG;
735 0 : return -1;
736 : }
737 0 : snprintf(buf, 6, "@%04x", SVAL(*inbuf, 0));
738 0 : memcpy(*outbuf, buf, 5);
739 0 : (*inbytesleft) -= 2;
740 0 : (*outbytesleft) -= 5;
741 0 : (*inbuf) += 2;
742 0 : (*outbuf) += 5;
743 : }
744 :
745 0 : if (*inbytesleft == 1) {
746 0 : errno = EINVAL;
747 0 : return -1;
748 : }
749 :
750 0 : if (*inbytesleft > 1) {
751 0 : errno = E2BIG;
752 0 : return -1;
753 : }
754 :
755 0 : return 0;
756 : }
757 :
758 76296 : static size_t iconv_swab(void *cd, const char **inbuf, size_t *inbytesleft,
759 : char **outbuf, size_t *outbytesleft)
760 : {
761 8282 : int n;
762 :
763 76296 : n = MIN(*inbytesleft, *outbytesleft);
764 :
765 76296 : swab(*inbuf, *outbuf, (n&~1));
766 76296 : if (n&1) {
767 0 : (*outbuf)[n-1] = 0;
768 : }
769 :
770 76296 : (*inbytesleft) -= n;
771 76296 : (*outbytesleft) -= n;
772 76296 : (*inbuf) += n;
773 76296 : (*outbuf) += n;
774 :
775 76296 : if (*inbytesleft > 0) {
776 182 : errno = E2BIG;
777 182 : return -1;
778 : }
779 :
780 67832 : return 0;
781 : }
782 :
783 :
784 4981 : static size_t iconv_copy(void *cd, const char **inbuf, size_t *inbytesleft,
785 : char **outbuf, size_t *outbytesleft)
786 : {
787 31 : int n;
788 :
789 4981 : n = MIN(*inbytesleft, *outbytesleft);
790 :
791 4981 : memmove(*outbuf, *inbuf, n);
792 :
793 4981 : (*inbytesleft) -= n;
794 4981 : (*outbytesleft) -= n;
795 4981 : (*inbuf) += n;
796 4981 : (*outbuf) += n;
797 :
798 4981 : if (*inbytesleft > 0) {
799 0 : errno = E2BIG;
800 0 : return -1;
801 : }
802 :
803 4950 : return 0;
804 : }
805 :
806 : /*
807 : this takes a UTF8 sequence and produces a UTF16 sequence
808 : */
809 127029561 : static size_t utf8_pull(void *cd, const char **inbuf, size_t *inbytesleft,
810 : char **outbuf, size_t *outbytesleft)
811 : {
812 127029561 : size_t in_left=*inbytesleft, out_left=*outbytesleft;
813 127029561 : const uint8_t *c = (const uint8_t *)*inbuf;
814 127029561 : uint8_t *uc = (uint8_t *)*outbuf;
815 :
816 2414935796 : while (in_left >= 1 && out_left >= 2) {
817 2288962512 : if ((c[0] & 0x80) == 0) {
818 2271685016 : uc[0] = c[0];
819 2271685016 : uc[1] = 0;
820 2271685016 : c += 1;
821 2271685016 : in_left -= 1;
822 2271685016 : out_left -= 2;
823 2271685016 : uc += 2;
824 2271685016 : continue;
825 : }
826 :
827 17277496 : if ((c[0] & 0xe0) == 0xc0) {
828 7157087 : if (in_left < 2 ||
829 7157085 : (c[1] & 0xc0) != 0x80) {
830 242 : errno = EILSEQ;
831 242 : goto error;
832 : }
833 7156845 : uc[1] = (c[0]>>2) & 0x7;
834 7156845 : uc[0] = (c[0]<<6) | (c[1]&0x3f);
835 7156845 : if (uc[1] == 0 && uc[0] < 0x80) {
836 : /* this should have been a single byte */
837 1 : errno = EILSEQ;
838 1 : goto error;
839 : }
840 7156844 : c += 2;
841 7156844 : in_left -= 2;
842 7156844 : out_left -= 2;
843 7156844 : uc += 2;
844 7156844 : continue;
845 : }
846 :
847 10120409 : if ((c[0] & 0xf0) == 0xe0) {
848 3900308 : unsigned int codepoint;
849 6982279 : if (in_left < 3 ||
850 6982275 : (c[1] & 0xc0) != 0x80 ||
851 6981913 : (c[2] & 0xc0) != 0x80) {
852 366 : errno = EILSEQ;
853 366 : goto error;
854 : }
855 6981913 : codepoint = ((c[2] & 0x3f) |
856 6981913 : ((c[1] & 0x3f) << 6) |
857 6981913 : ((c[0] & 0x0f) << 12));
858 :
859 6981913 : if (codepoint < 0x800) {
860 : /* this should be a 1 or 2 byte sequence */
861 1 : errno = EILSEQ;
862 1 : goto error;
863 : }
864 6981912 : if (codepoint >= 0xd800 && codepoint <= 0xdfff) {
865 : /*
866 : * This is an invalid codepoint, per
867 : * RFC3629, as it encodes part of a
868 : * UTF-16 surrogate pair for a
869 : * character over U+10000, which ought
870 : * to have been encoded as a four byte
871 : * utf-8 sequence.
872 : *
873 : * Prior to Vista, Windows might
874 : * sometimes produce invalid strings
875 : * where a utf-16 sequence containing
876 : * surrogate pairs was converted
877 : * "verbatim" into utf-8, instead of
878 : * encoding the actual codepoint. This
879 : * format is sometimes called "WTF-8".
880 : *
881 : * If we were to support that, we'd
882 : * have a branch here for the case
883 : * where the codepoint is between
884 : * 0xd800 and 0xdbff (a "high
885 : * surrogate"), and read a *six*
886 : * character sequence from there which
887 : * would include a low surrogate. But
888 : * that would undermine the
889 : * hard-learnt principle that each
890 : * character should only have one
891 : * encoding.
892 : */
893 0 : errno = EILSEQ;
894 0 : goto error;
895 : }
896 :
897 6981912 : uc[0] = codepoint & 0xff;
898 6981912 : uc[1] = codepoint >> 8;
899 6981912 : c += 3;
900 6981912 : in_left -= 3;
901 6981912 : out_left -= 2;
902 6981912 : uc += 2;
903 6981912 : continue;
904 : }
905 :
906 3138130 : if ((c[0] & 0xf8) == 0xf0) {
907 2956344 : unsigned int codepoint;
908 3137288 : if (in_left < 4 ||
909 3137288 : (c[1] & 0xc0) != 0x80 ||
910 3137288 : (c[2] & 0xc0) != 0x80 ||
911 3137288 : (c[3] & 0xc0) != 0x80) {
912 0 : errno = EILSEQ;
913 0 : goto error;
914 : }
915 3137288 : codepoint =
916 3137288 : (c[3]&0x3f) |
917 3137288 : ((c[2]&0x3f)<<6) |
918 3137288 : ((c[1]&0x3f)<<12) |
919 3137288 : ((c[0]&0x7)<<18);
920 3137288 : if (codepoint < 0x10000) {
921 : /* reject UTF-8 characters that are not
922 : minimally packed */
923 1 : errno = EILSEQ;
924 1 : goto error;
925 : }
926 3137287 : if (codepoint > 0x10ffff) {
927 : /*
928 : * Unicode stops at 0x10ffff, and if
929 : * we ignore that, we'll end up
930 : * encoding the wrong characters in
931 : * the surrogate pair.
932 : */
933 0 : errno = EILSEQ;
934 0 : goto error;
935 : }
936 :
937 3137287 : codepoint -= 0x10000;
938 :
939 3137287 : if (out_left < 4) {
940 1054824 : errno = E2BIG;
941 1054824 : goto error;
942 : }
943 :
944 2082463 : uc[0] = (codepoint>>10) & 0xFF;
945 2082463 : uc[1] = (codepoint>>18) | 0xd8;
946 2082463 : uc[2] = codepoint & 0xFF;
947 2082463 : uc[3] = ((codepoint>>8) & 0x3) | 0xdc;
948 2082463 : c += 4;
949 2082463 : in_left -= 4;
950 2082463 : out_left -= 4;
951 2082463 : uc += 4;
952 2082463 : continue;
953 : }
954 :
955 : /* we don't handle 5 byte sequences */
956 842 : errno = EINVAL;
957 842 : goto error;
958 : }
959 :
960 125973284 : if (in_left > 0) {
961 2045706 : errno = E2BIG;
962 2045706 : goto error;
963 : }
964 :
965 123927578 : *inbytesleft = in_left;
966 123927578 : *outbytesleft = out_left;
967 123927578 : *inbuf = (const char *)c;
968 123927578 : *outbuf = (char *)uc;
969 123927578 : return 0;
970 :
971 3101983 : error:
972 3101983 : *inbytesleft = in_left;
973 3101983 : *outbytesleft = out_left;
974 3101983 : *inbuf = (const char *)c;
975 3101983 : *outbuf = (char *)uc;
976 3101983 : return -1;
977 : }
978 :
979 :
980 : /*
981 : this takes a UTF16 sequence and produces a UTF8 sequence
982 : */
983 131814489 : static size_t utf8_push(void *cd, const char **inbuf, size_t *inbytesleft,
984 : char **outbuf, size_t *outbytesleft)
985 : {
986 131814489 : size_t in_left=*inbytesleft, out_left=*outbytesleft;
987 131814489 : uint8_t *c = (uint8_t *)*outbuf;
988 131814489 : const uint8_t *uc = (const uint8_t *)*inbuf;
989 :
990 3369428225 : while (in_left >= 2 && out_left >= 1) {
991 72779490 : unsigned int codepoint;
992 :
993 3238014015 : if (uc[1] == 0 && !(uc[0] & 0x80)) {
994 : /* simplest case */
995 3225121256 : c[0] = uc[0];
996 3225121256 : in_left -= 2;
997 3225121256 : out_left -= 1;
998 3225121256 : uc += 2;
999 3225121256 : c += 1;
1000 3225121256 : continue;
1001 : }
1002 :
1003 12892759 : if ((uc[1]&0xf8) == 0) {
1004 : /* next simplest case */
1005 4594518 : if (out_left < 2) {
1006 4 : errno = E2BIG;
1007 4 : goto error;
1008 : }
1009 4594514 : c[0] = 0xc0 | (uc[0]>>6) | (uc[1]<<2);
1010 4594514 : c[1] = 0x80 | (uc[0] & 0x3f);
1011 4594514 : in_left -= 2;
1012 4594514 : out_left -= 2;
1013 4594514 : uc += 2;
1014 4594514 : c += 2;
1015 4594514 : continue;
1016 : }
1017 :
1018 8298241 : if ((uc[1] & 0xfc) == 0xdc) {
1019 304715 : errno = EILSEQ;
1020 : #ifndef HAVE_ICONV_ERRNO_ILLEGAL_MULTIBYTE
1021 : if (in_left < 4) {
1022 : errno = EINVAL;
1023 : }
1024 : #endif
1025 304715 : goto error;
1026 : }
1027 :
1028 7993526 : if ((uc[1] & 0xfc) != 0xd8) {
1029 5907291 : codepoint = uc[0] | (uc[1]<<8);
1030 5907291 : if (out_left < 3) {
1031 3 : errno = E2BIG;
1032 3 : goto error;
1033 : }
1034 5907288 : c[0] = 0xe0 | (codepoint >> 12);
1035 5907288 : c[1] = 0x80 | ((codepoint >> 6) & 0x3f);
1036 5907288 : c[2] = 0x80 | (codepoint & 0x3f);
1037 :
1038 5907288 : in_left -= 2;
1039 5907288 : out_left -= 3;
1040 5907288 : uc += 2;
1041 5907288 : c += 3;
1042 5907288 : continue;
1043 : }
1044 :
1045 : /* its the first part of a 4 byte sequence */
1046 2086235 : if (in_left < 4) {
1047 3606 : errno = EINVAL;
1048 3606 : goto error;
1049 : }
1050 2082629 : if ((uc[3] & 0xfc) != 0xdc) {
1051 91951 : errno = EILSEQ;
1052 91951 : goto error;
1053 : }
1054 1990678 : codepoint = 0x10000 + (uc[2] | ((uc[3] & 0x3)<<8) |
1055 1990678 : (uc[0]<<10) | ((uc[1] & 0x3)<<18));
1056 :
1057 1990678 : if (out_left < 4) {
1058 0 : errno = E2BIG;
1059 0 : goto error;
1060 : }
1061 1990678 : c[0] = 0xf0 | (codepoint >> 18);
1062 1990678 : c[1] = 0x80 | ((codepoint >> 12) & 0x3f);
1063 1990678 : c[2] = 0x80 | ((codepoint >> 6) & 0x3f);
1064 1990678 : c[3] = 0x80 | (codepoint & 0x3f);
1065 :
1066 1990678 : in_left -= 4;
1067 1990678 : out_left -= 4;
1068 1990678 : uc += 4;
1069 1990678 : c += 4;
1070 : }
1071 :
1072 131414210 : if (in_left == 1) {
1073 50804 : errno = EINVAL;
1074 50804 : goto error;
1075 : }
1076 :
1077 131363406 : if (in_left > 1) {
1078 15 : errno = E2BIG;
1079 15 : goto error;
1080 : }
1081 :
1082 131363391 : *inbytesleft = in_left;
1083 131363391 : *outbytesleft = out_left;
1084 131363391 : *inbuf = (const char *)uc;
1085 131363391 : *outbuf = (char *)c;
1086 :
1087 131363391 : return 0;
1088 :
1089 451098 : error:
1090 451098 : *inbytesleft = in_left;
1091 451098 : *outbytesleft = out_left;
1092 451098 : *inbuf = (const char *)uc;
1093 451098 : *outbuf = (char *)c;
1094 451098 : return -1;
1095 : }
1096 :
1097 :
1098 : /*
1099 : this takes a UTF16 munged sequence, modifies it according to the
1100 : string2key rules, and produces a UTF16 sequence
1101 :
1102 : The rules are:
1103 :
1104 : 1) any 0x0000 characters are mapped to 0x0001
1105 :
1106 : 2) convert any instance of 0xD800 - 0xDBFF (high surrogate)
1107 : without an immediately following 0xDC00 - 0x0xDFFF (low surrogate) to
1108 : U+FFFD (OBJECT REPLACEMENT CHARACTER).
1109 :
1110 : 3) the same for any low surrogate that was not preceded by a high surrogate.
1111 :
1112 : */
1113 22999 : static size_t utf16_munged_pull(void *cd, const char **inbuf, size_t *inbytesleft,
1114 : char **outbuf, size_t *outbytesleft)
1115 : {
1116 22999 : size_t in_left=*inbytesleft, out_left=*outbytesleft;
1117 22999 : uint8_t *c = (uint8_t *)*outbuf;
1118 22999 : const uint8_t *uc = (const uint8_t *)*inbuf;
1119 :
1120 701872 : while (in_left >= 2 && out_left >= 2) {
1121 678873 : unsigned int codepoint = uc[0] | (uc[1]<<8);
1122 :
1123 678873 : if (codepoint == 0) {
1124 2 : codepoint = 1;
1125 : }
1126 :
1127 678873 : if ((codepoint & 0xfc00) == 0xd800) {
1128 : /* a high surrogate */
1129 115 : unsigned int codepoint2;
1130 1231 : if (in_left < 4) {
1131 2 : codepoint = 0xfffd;
1132 2 : goto codepoint16;
1133 : }
1134 1229 : codepoint2 = uc[2] | (uc[3]<<8);
1135 1229 : if ((codepoint2 & 0xfc00) != 0xdc00) {
1136 : /* high surrogate not followed by low
1137 : surrogate: convert to 0xfffd */
1138 1214 : codepoint = 0xfffd;
1139 1214 : goto codepoint16;
1140 : }
1141 15 : if (out_left < 4) {
1142 0 : errno = E2BIG;
1143 0 : goto error;
1144 : }
1145 15 : memcpy(c, uc, 4);
1146 15 : in_left -= 4;
1147 15 : out_left -= 4;
1148 15 : uc += 4;
1149 15 : c += 4;
1150 15 : continue;
1151 : }
1152 :
1153 677642 : if ((codepoint & 0xfc00) == 0xdc00) {
1154 : /* low surrogate not preceded by high
1155 : surrogate: convert to 0xfffd */
1156 1169 : codepoint = 0xfffd;
1157 : }
1158 :
1159 676473 : codepoint16:
1160 678858 : c[0] = codepoint & 0xFF;
1161 678858 : c[1] = (codepoint>>8) & 0xFF;
1162 :
1163 678858 : in_left -= 2;
1164 678858 : out_left -= 2;
1165 678858 : uc += 2;
1166 678858 : c += 2;
1167 678858 : continue;
1168 : }
1169 :
1170 22999 : if (in_left == 1) {
1171 0 : errno = EINVAL;
1172 0 : goto error;
1173 : }
1174 :
1175 22999 : if (in_left > 1) {
1176 0 : errno = E2BIG;
1177 0 : goto error;
1178 : }
1179 :
1180 22999 : *inbytesleft = in_left;
1181 22999 : *outbytesleft = out_left;
1182 22999 : *inbuf = (const char *)uc;
1183 22999 : *outbuf = (char *)c;
1184 :
1185 22999 : return 0;
1186 :
1187 0 : error:
1188 0 : *inbytesleft = in_left;
1189 0 : *outbytesleft = out_left;
1190 0 : *inbuf = (const char *)uc;
1191 0 : *outbuf = (char *)c;
1192 0 : return -1;
1193 : }
1194 :
1195 :
1196 :
|
