6 * Copyright (C) 2012, 2013 Eric Biggers
8 * This file is part of wimlib, a library for working with WIM files.
10 * wimlib is free software; you can redistribute it and/or modify it under the
11 * terms of the GNU General Public License as published by the Free
12 * Software Foundation; either version 3 of the License, or (at your option)
15 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
16 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
17 * A PARTICULAR PURPOSE. See the GNU General Public License for more
20 * You should have received a copy of the GNU General Public License
21 * along with wimlib; if not, see http://www.gnu.org/licenses/.
29 #include "wimlib/encoding.h"
30 #include "wimlib/endianness.h"
31 #include "wimlib/error.h"
32 #include "wimlib/list.h"
33 #include "wimlib/util.h"
41 bool wimlib_mbs_is_utf8 = !TCHAR_IS_UTF16LE;
43 /* List of iconv_t conversion descriptors for a specific character conversion.
44 * The idea is that it is not thread-safe to have just one conversion
45 * descriptor, but it also is inefficient to open a new conversion descriptor to
46 * convert every string. Both these problems can be solved by maintaining a
47 * list of conversion descriptors; then, a thread can use an existing conversion
48 * descriptor if available. */
49 struct iconv_list_head {
50 const char *from_encoding;
51 const char *to_encoding;
52 struct list_head list;
53 pthread_mutex_t mutex;
58 struct list_head list;
59 struct iconv_list_head *head;
62 #define ICONV_LIST(name, from, to) \
63 struct iconv_list_head name = { \
64 .from_encoding = from, \
66 .list = LIST_HEAD_INIT(name.list), \
67 .mutex = PTHREAD_MUTEX_INITIALIZER, \
71 get_iconv(struct iconv_list_head *head)
77 pthread_mutex_lock(&head->mutex);
78 if (list_empty(&head->list)) {
79 cd = iconv_open(head->to_encoding, head->from_encoding);
80 if (cd == (iconv_t)-1) {
81 ERROR_WITH_ERRNO("Failed to open iconv from %s to %s",
82 head->from_encoding, head->to_encoding);
85 i = MALLOC(sizeof(struct iconv_node));
96 i = container_of(head->list.next, struct iconv_node, list);
97 list_del(head->list.next);
100 pthread_mutex_unlock(&head->mutex);
105 put_iconv(iconv_t *cd)
107 int errno_save = errno;
108 struct iconv_node *i = container_of(cd, struct iconv_node, cd);
109 struct iconv_list_head *head = i->head;
111 pthread_mutex_lock(&head->mutex);
112 list_add(&i->list, &head->list);
113 pthread_mutex_unlock(&head->mutex);
117 /* Prevent printing an error message if a character conversion error occurs
118 * while printing an error message. (This variable is not per-thread but it
119 * doesn't matter too much since it's just the error messages.) */
120 static bool error_message_being_printed = false;
122 #define DEFINE_CHAR_CONVERSION_FUNCTIONS(varname1, longname1, chartype1,\
123 varname2, longname2, chartype2,\
124 earlyreturn_on_utf8_locale, \
126 worst_case_len_expr, \
130 static ICONV_LIST(iconv_##varname1##_to_##varname2, \
131 longname1, longname2); \
134 varname1##_to_##varname2##_nbytes(const chartype1 *in, size_t in_nbytes,\
135 size_t *out_nbytes_ret) \
137 iconv_t *cd = get_iconv(&iconv_##varname1##_to_##varname2); \
139 return WIMLIB_ERR_ICONV_NOT_AVAILABLE; \
141 /* Worst case length */ \
142 chartype2 buf[worst_case_len_expr]; \
143 char *inbuf = (char*)in; \
144 size_t inbytesleft = in_nbytes; \
145 char *outbuf = (char*)buf; \
146 size_t outbytesleft = sizeof(buf); \
150 len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft); \
151 if (len == (size_t)-1) { \
152 if (!error_message_being_printed) { \
153 error_message_being_printed = true; \
155 error_message_being_printed = false; \
159 *out_nbytes_ret = sizeof(buf) - outbytesleft; \
167 varname1##_to_##varname2##_buf(const chartype1 *in, size_t in_nbytes, \
170 iconv_t *cd = get_iconv(&iconv_##varname1##_to_##varname2); \
172 return WIMLIB_ERR_ICONV_NOT_AVAILABLE; \
174 char *inbuf = (char*)in; \
175 size_t inbytesleft = in_nbytes; \
176 char *outbuf = (char*)out; \
177 const size_t LARGE_NUMBER = 1000000000; \
178 size_t outbytesleft = LARGE_NUMBER; \
182 len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft); \
183 if (len == (size_t)-1) { \
184 if (!error_message_being_printed) { \
185 error_message_being_printed = true; \
187 error_message_being_printed = false; \
191 out[(LARGE_NUMBER-outbytesleft)/sizeof(chartype2)] = 0; \
199 varname1##_to_##varname2(const chartype1 *in, size_t in_nbytes, \
200 chartype2 **out_ret, \
201 size_t *out_nbytes_ret) \
207 if (earlyreturn_on_utf8_locale && wimlib_mbs_is_utf8) { \
209 /* Out same as in */ \
210 out = MALLOC(in_nbytes + sizeof(chartype2)); \
212 return WIMLIB_ERR_NOMEM; \
213 memcpy(out, in, in_nbytes); \
214 out[in_nbytes / sizeof(chartype2)] = 0; \
216 *out_nbytes_ret = in_nbytes; \
220 ret = varname1##_to_##varname2##_nbytes(in, in_nbytes, \
225 out = MALLOC(out_nbytes + sizeof(chartype2)); \
227 return WIMLIB_ERR_NOMEM; \
229 ret = varname1##_to_##varname2##_buf(in, in_nbytes, out); \
234 *out_nbytes_ret = out_nbytes; \
239 #if !TCHAR_IS_UTF16LE
243 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", tchar,
244 utf16le, "UTF-16LE", utf16lechar,
248 WIMLIB_ERR_INVALID_UTF8_STRING,
249 ERROR_WITH_ERRNO("Failed to convert UTF-8 string "
250 "to UTF-16LE string!"),
253 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf16le, "UTF-16LE", utf16lechar,
254 utf8, "UTF-8", tchar,
258 WIMLIB_ERR_INVALID_UTF16_STRING,
259 ERROR_WITH_ERRNO("Failed to convert UTF-16LE string "
263 DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "", tchar,
264 utf16le, "UTF-16LE", utf16lechar,
266 return utf8_to_utf16le(in, in_nbytes, out_ret, out_nbytes_ret),
268 WIMLIB_ERR_INVALID_MULTIBYTE_STRING,
269 ERROR_WITH_ERRNO("Failed to convert multibyte "
270 "string \"%"TS"\" to UTF-16LE string!", in);
271 ERROR("If the data you provided was UTF-8, please make sure "
272 "the character encoding\n"
273 " of your current locale is UTF-8."),
276 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf16le, "UTF-16LE", utf16lechar,
279 return utf16le_to_utf8(in, in_nbytes, out_ret, out_nbytes_ret),
281 WIMLIB_ERR_UNICODE_STRING_NOT_REPRESENTABLE,
282 ERROR("Failed to convert UTF-16LE string to "
283 "multibyte string!");
284 ERROR("This may be because the UTF-16LE string "
285 "could not be represented\n"
286 " in your locale's character encoding."),
290 /* tchar to UTF-8 and back */
294 DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "UTF-16LE", tchar,
299 WIMLIB_ERR_INVALID_UTF16_STRING,
300 ERROR_WITH_ERRNO("Failed to convert UTF-16LE "
301 "string \"%"TS"\" to UTF-8 string!", in),
304 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", char,
305 tstr, "UTF-16LE", tchar,
309 WIMLIB_ERR_INVALID_UTF8_STRING,
310 ERROR_WITH_ERRNO("Failed to convert UTF-8 string "
311 "to UTF-16LE string!"),
317 DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "", tchar,
322 WIMLIB_ERR_INVALID_MULTIBYTE_STRING,
323 ERROR_WITH_ERRNO("Failed to convert multibyte "
324 "string \"%"TS"\" to UTF-8 string!", in);
325 ERROR("If the data you provided was UTF-8, please make sure "
327 " encoding of your current locale is UTF-8."),
330 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", char,
335 WIMLIB_ERR_UNICODE_STRING_NOT_REPRESENTABLE,
336 ERROR("Failed to convert UTF-8 string to "
337 "multibyte string!");
338 ERROR("This may be because the UTF-8 data "
339 "could not be represented\n"
340 " in your locale's character encoding."),
345 tstr_to_utf8_simple(const tchar *tstr, char **out)
348 return tstr_to_utf8(tstr, tstrlen(tstr) * sizeof(tchar),
353 utf8_to_tstr_simple(const char *utf8str, tchar **out)
356 return utf8_to_tstr(utf8str, strlen(utf8str), out, &out_nbytes);
360 iconv_cleanup(struct iconv_list_head *head)
362 pthread_mutex_destroy(&head->mutex);
363 while (!list_empty(&head->list)) {
364 struct iconv_node *i;
366 i = container_of(head->list.next, struct iconv_node, list);
374 iconv_global_cleanup(void)
376 iconv_cleanup(&iconv_utf8_to_tstr);
377 iconv_cleanup(&iconv_tstr_to_utf8);
378 #if !TCHAR_IS_UTF16LE
379 iconv_cleanup(&iconv_utf16le_to_tstr);
380 iconv_cleanup(&iconv_tstr_to_utf16le);
381 iconv_cleanup(&iconv_utf16le_to_utf8);
382 iconv_cleanup(&iconv_utf8_to_utf16le);
386 /* Upper case table --- Borrowed from NTFS-3g
388 * Copyright (c) 2000-2004 Anton Altaparmakov
389 * Copyright (c) 2002-2009 Szabolcs Szakacsits
390 * Copyright (c) 2008-2011 Jean-Pierre Andre
391 * Copyright (c) 2008 Bernhard Kaindl
393 * License is GPLv2 or later.
397 * ntfs_upcase_table_build - build the default upcase table for NTFS
398 * @uc: destination buffer where to store the built table
399 * @uc_len: size of destination buffer in bytes
401 * ntfs_upcase_table_build() builds the default upcase table for NTFS and
402 * stores it in the caller supplied buffer @uc of size @uc_len.
404 * Note, @uc_len must be at least 128kiB in size or bad things will happen!
407 ntfs_upcase_table_build(utf16lechar *uc, u32 uc_len)
410 * This is the table as defined by Vista
413 * "Start" is inclusive and "End" is exclusive, every value has the
414 * value of "Add" added to it.
416 static int uc_run_table[][3] = { /* Start, End, Add */
417 {0x0061, 0x007b, -32}, {0x00e0, 0x00f7, -32}, {0x00f8, 0x00ff, -32},
418 {0x0256, 0x0258, -205}, {0x028a, 0x028c, -217}, {0x037b, 0x037e, 130},
419 {0x03ac, 0x03ad, -38}, {0x03ad, 0x03b0, -37}, {0x03b1, 0x03c2, -32},
420 {0x03c2, 0x03c3, -31}, {0x03c3, 0x03cc, -32}, {0x03cc, 0x03cd, -64},
421 {0x03cd, 0x03cf, -63}, {0x0430, 0x0450, -32}, {0x0450, 0x0460, -80},
422 {0x0561, 0x0587, -48}, {0x1f00, 0x1f08, 8}, {0x1f10, 0x1f16, 8},
423 {0x1f20, 0x1f28, 8}, {0x1f30, 0x1f38, 8}, {0x1f40, 0x1f46, 8},
424 {0x1f51, 0x1f52, 8}, {0x1f53, 0x1f54, 8}, {0x1f55, 0x1f56, 8},
425 {0x1f57, 0x1f58, 8}, {0x1f60, 0x1f68, 8}, {0x1f70, 0x1f72, 74},
426 {0x1f72, 0x1f76, 86}, {0x1f76, 0x1f78, 100}, {0x1f78, 0x1f7a, 128},
427 {0x1f7a, 0x1f7c, 112}, {0x1f7c, 0x1f7e, 126}, {0x1f80, 0x1f88, 8},
428 {0x1f90, 0x1f98, 8}, {0x1fa0, 0x1fa8, 8}, {0x1fb0, 0x1fb2, 8},
429 {0x1fb3, 0x1fb4, 9}, {0x1fcc, 0x1fcd, -9}, {0x1fd0, 0x1fd2, 8},
430 {0x1fe0, 0x1fe2, 8}, {0x1fe5, 0x1fe6, 7}, {0x1ffc, 0x1ffd, -9},
431 {0x2170, 0x2180, -16}, {0x24d0, 0x24ea, -26}, {0x2c30, 0x2c5f, -48},
432 {0x2d00, 0x2d26, -7264}, {0xff41, 0xff5b, -32}, {0}
435 * "Start" is exclusive and "End" is inclusive, every second value is
436 * decremented by one.
438 static int uc_dup_table[][2] = { /* Start, End */
439 {0x0100, 0x012f}, {0x0132, 0x0137}, {0x0139, 0x0149}, {0x014a, 0x0178},
440 {0x0179, 0x017e}, {0x01a0, 0x01a6}, {0x01b3, 0x01b7}, {0x01cd, 0x01dd},
441 {0x01de, 0x01ef}, {0x01f4, 0x01f5}, {0x01f8, 0x01f9}, {0x01fa, 0x0220},
442 {0x0222, 0x0234}, {0x023b, 0x023c}, {0x0241, 0x0242}, {0x0246, 0x024f},
443 {0x03d8, 0x03ef}, {0x03f7, 0x03f8}, {0x03fa, 0x03fb}, {0x0460, 0x0481},
444 {0x048a, 0x04bf}, {0x04c1, 0x04c4}, {0x04c5, 0x04c8}, {0x04c9, 0x04ce},
445 {0x04ec, 0x04ed}, {0x04d0, 0x04eb}, {0x04ee, 0x04f5}, {0x04f6, 0x0513},
446 {0x1e00, 0x1e95}, {0x1ea0, 0x1ef9}, {0x2183, 0x2184}, {0x2c60, 0x2c61},
447 {0x2c67, 0x2c6c}, {0x2c75, 0x2c76}, {0x2c80, 0x2ce3}, {0}
450 * Set the Unicode character at offset "Offset" to "Value". Note,
451 * "Value" is host endian.
453 static int uc_byte_table[][2] = { /* Offset, Value */
454 {0x00ff, 0x0178}, {0x0180, 0x0243}, {0x0183, 0x0182}, {0x0185, 0x0184},
455 {0x0188, 0x0187}, {0x018c, 0x018b}, {0x0192, 0x0191}, {0x0195, 0x01f6},
456 {0x0199, 0x0198}, {0x019a, 0x023d}, {0x019e, 0x0220}, {0x01a8, 0x01a7},
457 {0x01ad, 0x01ac}, {0x01b0, 0x01af}, {0x01b9, 0x01b8}, {0x01bd, 0x01bc},
458 {0x01bf, 0x01f7}, {0x01c6, 0x01c4}, {0x01c9, 0x01c7}, {0x01cc, 0x01ca},
459 {0x01dd, 0x018e}, {0x01f3, 0x01f1}, {0x023a, 0x2c65}, {0x023e, 0x2c66},
460 {0x0253, 0x0181}, {0x0254, 0x0186}, {0x0259, 0x018f}, {0x025b, 0x0190},
461 {0x0260, 0x0193}, {0x0263, 0x0194}, {0x0268, 0x0197}, {0x0269, 0x0196},
462 {0x026b, 0x2c62}, {0x026f, 0x019c}, {0x0272, 0x019d}, {0x0275, 0x019f},
463 {0x027d, 0x2c64}, {0x0280, 0x01a6}, {0x0283, 0x01a9}, {0x0288, 0x01ae},
464 {0x0289, 0x0244}, {0x028c, 0x0245}, {0x0292, 0x01b7}, {0x03f2, 0x03f9},
465 {0x04cf, 0x04c0}, {0x1d7d, 0x2c63}, {0x214e, 0x2132}, {0}
470 memset((char*)uc, 0, uc_len);
474 for (i = 0; (u32)i < uc_len; i++)
475 uc[i] = cpu_to_le16(i);
476 for (r = 0; uc_run_table[r][0]; r++) {
477 off = uc_run_table[r][2];
478 for (i = uc_run_table[r][0]; i < uc_run_table[r][1]; i++)
479 uc[i] = cpu_to_le16(i + off);
481 for (r = 0; uc_dup_table[r][0]; r++)
482 for (i = uc_dup_table[r][0]; i < uc_dup_table[r][1]; i += 2)
483 uc[i + 1] = cpu_to_le16(i);
484 for (r = 0; uc_byte_table[r][0]; r++) {
485 k = uc_byte_table[r][1];
486 uc[uc_byte_table[r][0]] = cpu_to_le16(k);
490 static utf16lechar upcase[65536];
495 ntfs_upcase_table_build(upcase, ARRAY_LEN(upcase));
498 /* Compare UTF-16LE strings case-sensitively (%ignore_case == false) or
499 * case-insensitively (%ignore_case == true).
501 * This is implemented using the default upper-case table used by NTFS. It does
502 * not handle all possible cases allowed by UTF-16LE. For example, different
503 * normalizations of the same sequence of "characters" are not considered equal.
504 * It hopefully does the right thing most of the time though. */
506 cmp_utf16le_strings(const utf16lechar *s1, size_t n1,
507 const utf16lechar *s2, size_t n2,
510 size_t n = min(n1, n2);
513 for (size_t i = 0; i < n; i++) {
514 u16 c1 = le16_to_cpu(upcase[le16_to_cpu(s1[i])]);
515 u16 c2 = le16_to_cpu(upcase[le16_to_cpu(s2[i])]);
517 return (c1 < c2) ? -1 : 1;
520 for (size_t i = 0; i < n; i++) {
521 u16 c1 = le16_to_cpu(s1[i]);
522 u16 c2 = le16_to_cpu(s2[i]);
524 return (c1 < c2) ? -1 : 1;
529 return (n1 < n2) ? -1 : 1;
532 /* Duplicates a string of system-dependent encoding into a UTF-16LE string and
533 * returns the string and its length, in bytes, in the pointer arguments. Frees
534 * any existing string at the return location before overwriting it. */
536 get_utf16le_string(const tchar *name, utf16lechar **name_utf16le_ret,
537 u16 *name_utf16le_nbytes_ret)
539 utf16lechar *name_utf16le;
540 size_t name_utf16le_nbytes;
543 name_utf16le_nbytes = tstrlen(name) * sizeof(utf16lechar);
544 name_utf16le = MALLOC(name_utf16le_nbytes + sizeof(utf16lechar));
545 if (name_utf16le == NULL)
546 return WIMLIB_ERR_NOMEM;
547 memcpy(name_utf16le, name, name_utf16le_nbytes + sizeof(utf16lechar));
551 ret = tstr_to_utf16le(name, tstrlen(name), &name_utf16le,
552 &name_utf16le_nbytes);
554 if (name_utf16le_nbytes > 0xffff) {
556 ERROR("Multibyte string \"%"TS"\" is too long!", name);
557 ret = WIMLIB_ERR_INVALID_UTF8_STRING;
562 FREE(*name_utf16le_ret);
563 *name_utf16le_ret = name_utf16le;
564 *name_utf16le_nbytes_ret = name_utf16le_nbytes;