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 1208962 : static size_t sys_iconv(void *cd,
160 : const char **inbuf, size_t *inbytesleft,
161 : char **outbuf, size_t *outbytesleft)
162 : {
163 1208962 : size_t ret = iconv((iconv_t)cd,
164 : discard_const_p(char *, inbuf), inbytesleft,
165 : outbuf, outbytesleft);
166 1208962 : if (ret == (size_t)-1) iconv(cd, NULL, NULL, NULL, NULL);
167 1208962 : return ret;
168 : }
169 : #endif
170 :
171 : #ifdef HAVE_ICU_I18N
172 0 : static size_t sys_uconv(void *cd,
173 : const char **inbuf,
174 : size_t *inbytesleft,
175 : char **outbuf,
176 : size_t *outbytesleft)
177 0 : {
178 0 : UTransliterator *t = (UTransliterator *)cd;
179 0 : size_t bufsize = *inbytesleft * 2;
180 0 : UChar ustr[bufsize];
181 0 : UChar *up = NULL;
182 0 : char *p = NULL;
183 : int32_t ustrlen;
184 : int32_t limit;
185 : int32_t converted_len;
186 : size_t inbuf_consumed;
187 : size_t outbut_consumed;
188 : UErrorCode ue;
189 :
190 : /* Convert from UTF8 to UCS2 */
191 0 : ue = 0;
192 0 : up = u_strFromUTF8(ustr, /* dst */
193 : bufsize, /* dst buflen */
194 : &converted_len, /* dst written */
195 : *inbuf, /* src */
196 0 : *inbytesleft, /* src length */
197 : &ue);
198 0 : if (up == NULL || U_FAILURE(ue)) {
199 0 : return -1;
200 : }
201 0 : 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 0 : inbuf_consumed = converted_len;
213 :
214 : /*
215 : * The following transliteration function takes two parameters, the
216 : * lenght 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 0 : limit = converted_len;
221 :
222 : /* Inplace transliteration */
223 0 : utrans_transUChars(t,
224 : ustr, /* text */
225 : &converted_len, /* text length */
226 : bufsize, /* text buflen */
227 : 0, /* start */
228 : &limit, /* limit */
229 : &ue);
230 0 : if (U_FAILURE(ue)) {
231 0 : return -1;
232 : }
233 0 : 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 0 : ustrlen = converted_len;
244 :
245 : /* Convert from UCS2 back to UTF8 */
246 0 : ue = 0;
247 0 : p = u_strToUTF8(*outbuf, /* dst */
248 0 : *outbytesleft, /* dst buflen */
249 : &converted_len, /* dst required length */
250 : ustr, /* src */
251 : ustrlen, /* src length */
252 : &ue);
253 0 : if (p == NULL || U_FAILURE(ue)) {
254 0 : return -1;
255 : }
256 :
257 0 : outbut_consumed = converted_len;
258 0 : if (converted_len > *outbytesleft) {
259 : /*
260 : * The caller's result buffer is too small...
261 : */
262 0 : outbut_consumed = *outbytesleft;
263 : }
264 :
265 0 : *inbuf += inbuf_consumed;
266 0 : *inbytesleft -= inbuf_consumed;
267 0 : *outbuf += outbut_consumed;
268 0 : *outbytesleft -= outbut_consumed;
269 :
270 0 : return converted_len;
271 : }
272 : #endif
273 :
274 : /**
275 : * This is a simple portable iconv() implementaion.
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 21306752 : _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 21306752 : if (cd->direct) {
286 18158316 : 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 9251588 : while (*inbytesleft > 0) {
299 3234992 : char *bufp1 = cvtbuf;
300 3234992 : const char *bufp2 = cvtbuf;
301 3234992 : int saved_errno = errno;
302 3234992 : bool pull_failed = false;
303 3234992 : bufsize = SMB_ICONV_BUFSIZE;
304 :
305 3234992 : if (cd->pull(cd->cd_pull,
306 : inbuf, inbytesleft, &bufp1, &bufsize) == -1
307 86556 : && errno != E2BIG) {
308 0 : saved_errno = errno;
309 0 : pull_failed = true;
310 : }
311 :
312 3234992 : bufsize = SMB_ICONV_BUFSIZE - bufsize;
313 :
314 3234992 : if (cd->push(cd->cd_push,
315 : &bufp2, &bufsize,
316 : outbuf, outbytesleft) == -1) {
317 8 : return -1;
318 3234988 : } else if (pull_failed) {
319 : /* We want the pull errno if possible */
320 0 : errno = saved_errno;
321 0 : return -1;
322 : }
323 : }
324 : }
325 :
326 3148432 : return 0;
327 : }
328 :
329 171006 : static bool is_utf16(const char *name)
330 : {
331 342012 : return strcasecmp(name, "UCS-2LE") == 0 ||
332 171006 : strcasecmp(name, "UTF-16LE") == 0;
333 : }
334 :
335 42474 : 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 42474 : if (hwd->direct == sys_uconv) {
343 0 : utrans_close(hwd->cd_direct);
344 0 : return 0;
345 : }
346 : #endif
347 : #ifdef HAVE_NATIVE_ICONV
348 42474 : if (hwd->cd_pull != NULL && hwd->cd_pull != (iconv_t)-1)
349 5076 : iconv_close(hwd->cd_pull);
350 42474 : if (hwd->cd_push != NULL && hwd->cd_push != (iconv_t)-1)
351 1129 : iconv_close(hwd->cd_push);
352 42474 : if (hwd->cd_direct != NULL && hwd->cd_direct != (iconv_t)-1)
353 0 : iconv_close(hwd->cd_direct);
354 : #endif
355 :
356 42474 : return 0;
357 : }
358 :
359 69410 : _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 : smb_iconv_t ret;
363 69410 : const struct charset_functions *from=NULL, *to=NULL;
364 : int i;
365 :
366 69410 : ret = (smb_iconv_t)talloc_named(mem_ctx,
367 : sizeof(*ret),
368 : "iconv(%s,%s)", tocode, fromcode);
369 69410 : if (!ret) {
370 0 : errno = ENOMEM;
371 0 : return (smb_iconv_t)-1;
372 : }
373 69410 : memset(ret, 0, sizeof(*ret));
374 69410 : talloc_set_destructor(ret, smb_iconv_t_destructor);
375 :
376 : /* check for the simplest null conversion */
377 69410 : if (strcmp(fromcode, tocode) == 0) {
378 234 : ret->direct = iconv_copy;
379 234 : return ret;
380 : }
381 :
382 : /* check if we have a builtin function for this conversion */
383 899288 : for (i=0;i<ARRAY_SIZE(builtin_functions);i++) {
384 830112 : if (strcasecmp(fromcode, builtin_functions[i].name) == 0) {
385 57959 : if (use_builtin_handlers || builtin_functions[i].samba_internal_charset) {
386 57959 : from = &builtin_functions[i];
387 : }
388 : }
389 830112 : if (strcasecmp(tocode, builtin_functions[i].name) == 0) {
390 66205 : if (use_builtin_handlers || builtin_functions[i].samba_internal_charset) {
391 66205 : 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 69176 : if (from == NULL) {
403 11217 : ret->cd_pull = iconv_open("UTF-16LE", fromcode);
404 11217 : if (ret->cd_pull == (iconv_t)-1)
405 0 : ret->cd_pull = iconv_open("UCS-2LE", fromcode);
406 11217 : if (ret->cd_pull != (iconv_t)-1) {
407 11217 : ret->pull = sys_iconv;
408 : }
409 : }
410 :
411 69176 : if (to == NULL) {
412 2971 : ret->cd_push = iconv_open(tocode, "UTF-16LE");
413 2971 : if (ret->cd_push == (iconv_t)-1)
414 0 : ret->cd_push = iconv_open(tocode, "UCS-2LE");
415 2971 : if (ret->cd_push != (iconv_t)-1) {
416 2971 : ret->push = sys_iconv;
417 : }
418 : }
419 : #endif
420 :
421 : #ifdef HAVE_ICU_I18N
422 69176 : if (strcasecmp(fromcode, "UTF8-NFD") == 0 &&
423 0 : strcasecmp(tocode, "UTF8-NFC") == 0)
424 : {
425 : U_STRING_DECL(t, "any-nfc", 7);
426 0 : UErrorCode ue = 0;
427 :
428 : U_STRING_INIT(t, "any-nfc", 7);
429 :
430 0 : ret->cd_direct = utrans_openU(t,
431 : strlen("any-nfc"),
432 : UTRANS_FORWARD,
433 : NULL,
434 : 0,
435 : NULL,
436 : &ue);
437 0 : if (U_FAILURE(ue)) {
438 0 : return (smb_iconv_t)-1;
439 : }
440 0 : ret->direct = sys_uconv;
441 0 : return ret;
442 : }
443 :
444 69176 : if (strcasecmp(fromcode, "UTF8-NFC") == 0 &&
445 0 : strcasecmp(tocode, "UTF8-NFD") == 0)
446 : {
447 : U_STRING_DECL(tname, "any-nfd", 7);
448 0 : UErrorCode ue = 0;
449 :
450 : U_STRING_INIT(tname, "any-nfd", 7);
451 :
452 0 : ret->cd_direct = utrans_openU(tname,
453 : 7,
454 : UTRANS_FORWARD,
455 : NULL,
456 : 0,
457 : NULL,
458 : &ue);
459 0 : if (U_FAILURE(ue)) {
460 0 : return (smb_iconv_t)-1;
461 : }
462 0 : ret->direct = sys_uconv;
463 0 : return ret;
464 : }
465 : #endif
466 :
467 69176 : if (ret->pull == NULL && from == NULL) {
468 0 : goto failed;
469 : }
470 :
471 69176 : if (ret->push == NULL && to == NULL) {
472 0 : goto failed;
473 : }
474 :
475 : /* check for conversion to/from ucs2 */
476 69176 : if (is_utf16(fromcode) && to) {
477 23598 : ret->direct = to->push;
478 23598 : return ret;
479 : }
480 45578 : if (is_utf16(tocode) && from) {
481 17452 : ret->direct = from->pull;
482 17452 : return ret;
483 : }
484 :
485 : #ifdef HAVE_NATIVE_ICONV
486 28126 : if (is_utf16(fromcode)) {
487 0 : ret->direct = sys_iconv;
488 0 : ret->cd_direct = ret->cd_push;
489 0 : ret->cd_push = NULL;
490 0 : return ret;
491 : }
492 28126 : if (is_utf16(tocode)) {
493 0 : ret->direct = sys_iconv;
494 0 : ret->cd_direct = ret->cd_pull;
495 0 : ret->cd_pull = NULL;
496 0 : return ret;
497 : }
498 : #endif
499 :
500 : /* the general case has to go via a buffer */
501 28126 : if (!ret->pull) ret->pull = from->pull;
502 28126 : if (!ret->push) ret->push = to->push;
503 28126 : 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 0 : _PUBLIC_ smb_iconv_t smb_iconv_open(const char *tocode, const char *fromcode)
515 : {
516 0 : return smb_iconv_open_ex(NULL, tocode, fromcode, true);
517 : }
518 :
519 : /*
520 : simple iconv_close() wrapper
521 : */
522 42474 : _PUBLIC_ int smb_iconv_close(smb_iconv_t cd)
523 : {
524 42474 : talloc_free(cd);
525 42474 : 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 0 : static size_t ascii_pull(void *cd, const char **inbuf, size_t *inbytesleft,
544 : char **outbuf, size_t *outbytesleft)
545 : {
546 0 : while (*inbytesleft >= 1 && *outbytesleft >= 2) {
547 0 : if (((*inbuf)[0] & 0x7F) != (*inbuf)[0]) {
548 : /* If this is multi-byte, then it isn't legal ASCII */
549 0 : errno = EILSEQ;
550 0 : return -1;
551 : }
552 0 : (*outbuf)[0] = (*inbuf)[0];
553 0 : (*outbuf)[1] = 0;
554 0 : (*inbytesleft) -= 1;
555 0 : (*outbytesleft) -= 2;
556 0 : (*inbuf) += 1;
557 0 : (*outbuf) += 2;
558 : }
559 :
560 0 : 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 0 : static size_t ascii_push(void *cd, const char **inbuf, size_t *inbytesleft,
576 : char **outbuf, size_t *outbytesleft)
577 : {
578 0 : int ir_count=0;
579 :
580 0 : while (*inbytesleft >= 2 && *outbytesleft >= 1) {
581 0 : if (((*inbuf)[0] & 0x7F) != (*inbuf)[0] ||
582 0 : (*inbuf)[1] != 0) {
583 : /* If this is multi-byte, then it isn't legal ASCII */
584 0 : errno = EILSEQ;
585 0 : return -1;
586 : }
587 0 : (*outbuf)[0] = (*inbuf)[0];
588 0 : (*inbytesleft) -= 2;
589 0 : (*outbytesleft) -= 1;
590 0 : (*inbuf) += 2;
591 0 : (*outbuf) += 1;
592 : }
593 :
594 0 : if (*inbytesleft == 1) {
595 0 : errno = EINVAL;
596 0 : return -1;
597 : }
598 :
599 0 : 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 0 : static size_t latin1_pull(void *cd, const char **inbuf, size_t *inbytesleft,
615 : char **outbuf, size_t *outbytesleft)
616 : {
617 0 : while (*inbytesleft >= 1 && *outbytesleft >= 2) {
618 0 : (*outbuf)[0] = (*inbuf)[0];
619 0 : (*outbuf)[1] = 0;
620 0 : (*inbytesleft) -= 1;
621 0 : (*outbytesleft) -= 2;
622 0 : (*inbuf) += 1;
623 0 : (*outbuf) += 2;
624 : }
625 :
626 0 : if (*inbytesleft > 0) {
627 0 : errno = E2BIG;
628 0 : return -1;
629 : }
630 :
631 0 : 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 0 : static size_t latin1_push(void *cd, const char **inbuf, size_t *inbytesleft,
642 : char **outbuf, size_t *outbytesleft)
643 : {
644 0 : int ir_count=0;
645 :
646 0 : while (*inbytesleft >= 2 && *outbytesleft >= 1) {
647 0 : (*outbuf)[0] = (*inbuf)[0];
648 0 : 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 0 : (*inbytesleft) -= 2;
654 0 : (*outbytesleft) -= 1;
655 0 : (*inbuf) += 2;
656 0 : (*outbuf) += 1;
657 : }
658 :
659 0 : if (*inbytesleft == 1) {
660 0 : errno = EINVAL;
661 0 : return -1;
662 : }
663 :
664 0 : if (*inbytesleft > 1) {
665 0 : errno = E2BIG;
666 0 : return -1;
667 : }
668 :
669 0 : 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 : 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 : 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 45579 : static size_t iconv_swab(void *cd, const char **inbuf, size_t *inbytesleft,
759 : char **outbuf, size_t *outbytesleft)
760 : {
761 : int n;
762 :
763 45579 : n = MIN(*inbytesleft, *outbytesleft);
764 :
765 45579 : swab(*inbuf, *outbuf, (n&~1));
766 45579 : if (n&1) {
767 0 : (*outbuf)[n-1] = 0;
768 : }
769 :
770 45579 : (*inbytesleft) -= n;
771 45579 : (*outbytesleft) -= n;
772 45579 : (*inbuf) += n;
773 45579 : (*outbuf) += n;
774 :
775 45579 : if (*inbytesleft > 0) {
776 182 : errno = E2BIG;
777 182 : return -1;
778 : }
779 :
780 45397 : return 0;
781 : }
782 :
783 :
784 2578 : static size_t iconv_copy(void *cd, const char **inbuf, size_t *inbytesleft,
785 : char **outbuf, size_t *outbytesleft)
786 : {
787 : int n;
788 :
789 2578 : n = MIN(*inbytesleft, *outbytesleft);
790 :
791 2578 : memmove(*outbuf, *inbuf, n);
792 :
793 2578 : (*inbytesleft) -= n;
794 2578 : (*outbytesleft) -= n;
795 2578 : (*inbuf) += n;
796 2578 : (*outbuf) += n;
797 :
798 2578 : if (*inbytesleft > 0) {
799 0 : errno = E2BIG;
800 0 : return -1;
801 : }
802 :
803 2578 : return 0;
804 : }
805 :
806 : /*
807 : this takes a UTF8 sequence and produces a UTF16 sequence
808 : */
809 10414160 : static size_t utf8_pull(void *cd, const char **inbuf, size_t *inbytesleft,
810 : char **outbuf, size_t *outbytesleft)
811 : {
812 10414160 : size_t in_left=*inbytesleft, out_left=*outbytesleft;
813 10414160 : const uint8_t *c = (const uint8_t *)*inbuf;
814 10414160 : uint8_t *uc = (uint8_t *)*outbuf;
815 :
816 310948960 : while (in_left >= 1 && out_left >= 2) {
817 291075236 : if ((c[0] & 0x80) == 0) {
818 283961581 : uc[0] = c[0];
819 283961581 : uc[1] = 0;
820 283961581 : c += 1;
821 283961581 : in_left -= 1;
822 283961581 : out_left -= 2;
823 283961581 : uc += 2;
824 283961581 : continue;
825 : }
826 :
827 7113655 : if ((c[0] & 0xe0) == 0xc0) {
828 9572968 : if (in_left < 2 ||
829 4979750 : (c[1] & 0xc0) != 0x80) {
830 48 : errno = EILSEQ;
831 48 : goto error;
832 : }
833 4979702 : uc[1] = (c[0]>>2) & 0x7;
834 4979702 : uc[0] = (c[0]<<6) | (c[1]&0x3f);
835 4979702 : if (uc[1] == 0 && uc[0] < 0x80) {
836 : /* this should have been a single byte */
837 0 : errno = EILSEQ;
838 0 : goto error;
839 : }
840 4979702 : c += 2;
841 4979702 : in_left -= 2;
842 4979702 : out_left -= 2;
843 4979702 : uc += 2;
844 4979702 : continue;
845 : }
846 :
847 2133905 : if ((c[0] & 0xf0) == 0xe0) {
848 : unsigned int codepoint;
849 4121870 : if (in_left < 3 ||
850 4121834 : (c[1] & 0xc0) != 0x80 ||
851 2133611 : (c[2] & 0xc0) != 0x80) {
852 72 : errno = EILSEQ;
853 72 : goto error;
854 : }
855 6109913 : codepoint = ((c[2] & 0x3f) |
856 4121762 : ((c[1] & 0x3f) << 6) |
857 2133611 : ((c[0] & 0x0f) << 12));
858 :
859 2133611 : if (codepoint < 0x800) {
860 : /* this should be a 1 or 2 byte sequence */
861 0 : errno = EILSEQ;
862 0 : goto error;
863 : }
864 2133611 : uc[0] = codepoint & 0xff;
865 2133611 : uc[1] = codepoint >> 8;
866 2133611 : c += 3;
867 2133611 : in_left -= 3;
868 2133611 : out_left -= 2;
869 2133611 : uc += 2;
870 2133611 : continue;
871 : }
872 :
873 222 : if ((c[0] & 0xf8) == 0xf0) {
874 : unsigned int codepoint;
875 52 : if (in_left < 4 ||
876 52 : (c[1] & 0xc0) != 0x80 ||
877 52 : (c[2] & 0xc0) != 0x80 ||
878 26 : (c[3] & 0xc0) != 0x80) {
879 0 : errno = EILSEQ;
880 0 : goto error;
881 : }
882 26 : codepoint =
883 52 : (c[3]&0x3f) |
884 52 : ((c[2]&0x3f)<<6) |
885 52 : ((c[1]&0x3f)<<12) |
886 26 : ((c[0]&0x7)<<18);
887 26 : if (codepoint < 0x10000) {
888 : /* reject UTF-8 characters that are not
889 : minimally packed */
890 0 : errno = EILSEQ;
891 0 : goto error;
892 : }
893 :
894 26 : codepoint -= 0x10000;
895 :
896 26 : if (out_left < 4) {
897 12 : errno = E2BIG;
898 12 : goto error;
899 : }
900 :
901 14 : uc[0] = (codepoint>>10) & 0xFF;
902 14 : uc[1] = (codepoint>>18) | 0xd8;
903 14 : uc[2] = codepoint & 0xFF;
904 14 : uc[3] = ((codepoint>>8) & 0x3) | 0xdc;
905 14 : c += 4;
906 14 : in_left -= 4;
907 14 : out_left -= 4;
908 14 : uc += 4;
909 14 : continue;
910 : }
911 :
912 : /* we don't handle 5 byte sequences */
913 196 : errno = EINVAL;
914 196 : goto error;
915 : }
916 :
917 10413832 : if (in_left > 0) {
918 481079 : errno = E2BIG;
919 481079 : goto error;
920 : }
921 :
922 9932753 : *inbytesleft = in_left;
923 9932753 : *outbytesleft = out_left;
924 9932753 : *inbuf = (const char *)c;
925 9932753 : *outbuf = (char *)uc;
926 9932753 : return 0;
927 :
928 481407 : error:
929 481407 : *inbytesleft = in_left;
930 481407 : *outbytesleft = out_left;
931 481407 : *inbuf = (const char *)c;
932 481407 : *outbuf = (char *)uc;
933 481407 : return -1;
934 : }
935 :
936 :
937 : /*
938 : this takes a UTF16 sequence and produces a UTF8 sequence
939 : */
940 12942500 : static size_t utf8_push(void *cd, const char **inbuf, size_t *inbytesleft,
941 : char **outbuf, size_t *outbytesleft)
942 : {
943 12942500 : size_t in_left=*inbytesleft, out_left=*outbytesleft;
944 12942500 : uint8_t *c = (uint8_t *)*outbuf;
945 12942500 : const uint8_t *uc = (const uint8_t *)*inbuf;
946 :
947 895311112 : while (in_left >= 2 && out_left >= 1) {
948 : unsigned int codepoint;
949 :
950 870048312 : if (uc[1] == 0 && !(uc[0] & 0x80)) {
951 : /* simplest case */
952 864856948 : c[0] = uc[0];
953 864856948 : in_left -= 2;
954 864856948 : out_left -= 1;
955 864856948 : uc += 2;
956 864856948 : c += 1;
957 864856948 : continue;
958 : }
959 :
960 5191364 : if ((uc[1]&0xf8) == 0) {
961 : /* next simplest case */
962 3456290 : if (out_left < 2) {
963 0 : errno = E2BIG;
964 0 : goto error;
965 : }
966 3456290 : c[0] = 0xc0 | (uc[0]>>6) | (uc[1]<<2);
967 3456290 : c[1] = 0x80 | (uc[0] & 0x3f);
968 3456290 : in_left -= 2;
969 3456290 : out_left -= 2;
970 3456290 : uc += 2;
971 3456290 : c += 2;
972 3456290 : continue;
973 : }
974 :
975 1735074 : if ((uc[1] & 0xfc) == 0xdc) {
976 6 : errno = EILSEQ;
977 : #ifndef HAVE_ICONV_ERRNO_ILLEGAL_MULTIBYTE
978 : if (in_left < 4) {
979 : errno = EINVAL;
980 : }
981 : #endif
982 6 : goto error;
983 : }
984 :
985 1735068 : if ((uc[1] & 0xfc) != 0xd8) {
986 1735046 : codepoint = uc[0] | (uc[1]<<8);
987 1735046 : if (out_left < 3) {
988 0 : errno = E2BIG;
989 0 : goto error;
990 : }
991 1735046 : c[0] = 0xe0 | (codepoint >> 12);
992 1735046 : c[1] = 0x80 | ((codepoint >> 6) & 0x3f);
993 1735046 : c[2] = 0x80 | (codepoint & 0x3f);
994 :
995 1735046 : in_left -= 2;
996 1735046 : out_left -= 3;
997 1735046 : uc += 2;
998 1735046 : c += 3;
999 1735046 : continue;
1000 : }
1001 :
1002 : /* its the first part of a 4 byte sequence */
1003 22 : if (in_left < 4) {
1004 0 : errno = EINVAL;
1005 0 : goto error;
1006 : }
1007 22 : if ((uc[3] & 0xfc) != 0xdc) {
1008 2 : errno = EILSEQ;
1009 2 : goto error;
1010 : }
1011 56 : codepoint = 0x10000 + (uc[2] | ((uc[3] & 0x3)<<8) |
1012 38 : (uc[0]<<10) | ((uc[1] & 0x3)<<18));
1013 :
1014 20 : if (out_left < 4) {
1015 0 : errno = E2BIG;
1016 0 : goto error;
1017 : }
1018 20 : c[0] = 0xf0 | (codepoint >> 18);
1019 20 : c[1] = 0x80 | ((codepoint >> 12) & 0x3f);
1020 20 : c[2] = 0x80 | ((codepoint >> 6) & 0x3f);
1021 20 : c[3] = 0x80 | (codepoint & 0x3f);
1022 :
1023 20 : in_left -= 4;
1024 20 : out_left -= 4;
1025 20 : uc += 4;
1026 20 : c += 4;
1027 : }
1028 :
1029 12942492 : if (in_left == 1) {
1030 0 : errno = EINVAL;
1031 0 : goto error;
1032 : }
1033 :
1034 12942492 : if (in_left > 1) {
1035 0 : errno = E2BIG;
1036 0 : goto error;
1037 : }
1038 :
1039 12942492 : *inbytesleft = in_left;
1040 12942492 : *outbytesleft = out_left;
1041 12942492 : *inbuf = (const char *)uc;
1042 12942492 : *outbuf = (char *)c;
1043 :
1044 12942492 : return 0;
1045 :
1046 8 : error:
1047 8 : *inbytesleft = in_left;
1048 8 : *outbytesleft = out_left;
1049 8 : *inbuf = (const char *)uc;
1050 8 : *outbuf = (char *)c;
1051 8 : return -1;
1052 : }
1053 :
1054 :
1055 : /*
1056 : this takes a UTF16 munged sequence, modifies it according to the
1057 : string2key rules, and produces a UTF16 sequence
1058 :
1059 : The rules are:
1060 :
1061 : 1) any 0x0000 characters are mapped to 0x0001
1062 :
1063 : 2) convert any instance of 0xD800 - 0xDBFF (high surrogate)
1064 : without an immediately following 0xDC00 - 0x0xDFFF (low surrogate) to
1065 : U+FFFD (OBJECT REPLACEMENT CHARACTER).
1066 :
1067 : 3) the same for any low surrogate that was not preceded by a high surrogate.
1068 :
1069 : */
1070 14521 : static size_t utf16_munged_pull(void *cd, const char **inbuf, size_t *inbytesleft,
1071 : char **outbuf, size_t *outbytesleft)
1072 : {
1073 14521 : size_t in_left=*inbytesleft, out_left=*outbytesleft;
1074 14521 : uint8_t *c = (uint8_t *)*outbuf;
1075 14521 : const uint8_t *uc = (const uint8_t *)*inbuf;
1076 :
1077 422043 : while (in_left >= 2 && out_left >= 2) {
1078 397374 : unsigned int codepoint = uc[0] | (uc[1]<<8);
1079 :
1080 397374 : if (codepoint == 0) {
1081 3 : codepoint = 1;
1082 : }
1083 :
1084 397374 : if ((codepoint & 0xfc00) == 0xd800) {
1085 : /* a high surrogate */
1086 : unsigned int codepoint2;
1087 796 : if (in_left < 4) {
1088 6 : codepoint = 0xfffd;
1089 6 : goto codepoint16;
1090 : }
1091 790 : codepoint2 = uc[2] | (uc[3]<<8);
1092 790 : if ((codepoint2 & 0xfc00) != 0xdc00) {
1093 : /* high surrogate not followed by low
1094 : surrogate: convert to 0xfffd */
1095 782 : codepoint = 0xfffd;
1096 782 : goto codepoint16;
1097 : }
1098 8 : if (out_left < 4) {
1099 0 : errno = E2BIG;
1100 0 : goto error;
1101 : }
1102 8 : memcpy(c, uc, 4);
1103 8 : in_left -= 4;
1104 8 : out_left -= 4;
1105 8 : uc += 4;
1106 8 : c += 4;
1107 8 : continue;
1108 : }
1109 :
1110 396578 : if ((codepoint & 0xfc00) == 0xdc00) {
1111 : /* low surrogate not preceded by high
1112 : surrogate: convert to 0xfffd */
1113 784 : codepoint = 0xfffd;
1114 : }
1115 :
1116 711642 : codepoint16:
1117 397366 : c[0] = codepoint & 0xFF;
1118 397366 : c[1] = (codepoint>>8) & 0xFF;
1119 :
1120 397366 : in_left -= 2;
1121 397366 : out_left -= 2;
1122 397366 : uc += 2;
1123 397366 : c += 2;
1124 397366 : continue;
1125 : }
1126 :
1127 14521 : if (in_left == 1) {
1128 0 : errno = EINVAL;
1129 0 : goto error;
1130 : }
1131 :
1132 14521 : if (in_left > 1) {
1133 0 : errno = E2BIG;
1134 0 : goto error;
1135 : }
1136 :
1137 14521 : *inbytesleft = in_left;
1138 14521 : *outbytesleft = out_left;
1139 14521 : *inbuf = (const char *)uc;
1140 14521 : *outbuf = (char *)c;
1141 :
1142 14521 : return 0;
1143 :
1144 0 : error:
1145 0 : *inbytesleft = in_left;
1146 0 : *outbytesleft = out_left;
1147 0 : *inbuf = (const char *)uc;
1148 0 : *outbuf = (char *)c;
1149 0 : return -1;
1150 : }
1151 :
1152 :
1153 :
|
