2 Unix SMB/CIFS implementation.
3 Character set conversion Extensions
4 Copyright (C) Igor Vergeichik <iverg@mail.ru> 2001
5 Copyright (C) Andrew Tridgell 2001
6 Copyright (C) Simo Sorce 2001
7 Copyright (C) Jelmer Vernooij 2007
9 This program is free software; you can redistribute it and/or modify
10 it under the terms of the GNU General Public License as published by
11 the Free Software Foundation; either version 3 of the License, or
12 (at your option) any later version.
14 This program is distributed in the hope that it will be useful,
15 but WITHOUT ANY WARRANTY; without even the implied warranty of
16 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 GNU General Public License for more details.
19 You should have received a copy of the GNU General Public License
20 along with this program. If not, see <http://www.gnu.org/licenses/>.
24 #include "lib/util/charset/charset.h"
25 #include "system/locale.h"
26 #include "dynconfig/dynconfig.h"
34 * @brief Unicode string manipulation
37 /* these 2 tables define the unicode case handling. They are loaded
38 at startup either via mmap() or read() from the lib directory */
39 static void *upcase_table;
40 static void *lowcase_table;
43 /*******************************************************************
44 load the case handling tables
46 This is the function that should be called from library code.
47 ********************************************************************/
48 void load_case_tables_library(void)
52 mem_ctx = talloc_init("load_case_tables");
54 smb_panic("No memory for case_tables");
56 upcase_table = map_file(talloc_asprintf(mem_ctx, "%s/upcase.dat", get_dyn_CODEPAGEDIR()), 0x20000);
57 lowcase_table = map_file(talloc_asprintf(mem_ctx, "%s/lowcase.dat", get_dyn_CODEPAGEDIR()), 0x20000);
59 if (upcase_table == NULL) {
60 DEBUG(1, ("Failed to load upcase.dat, will use lame ASCII-only case sensitivity rules\n"));
61 upcase_table = (void *)-1;
63 if (lowcase_table == NULL) {
64 DEBUG(1, ("Failed to load lowcase.dat, will use lame ASCII-only case sensitivity rules\n"));
65 lowcase_table = (void *)-1;
69 /*******************************************************************
70 load the case handling tables
72 This MUST only be called from main() in application code, never from a
73 library. We don't know if the calling program has already done
74 setlocale() to another value, and can't tell if they have.
75 ********************************************************************/
76 void load_case_tables(void)
78 /* This is a useful global hook where we can ensure that the
79 * locale is set from the environment. This is needed so that
80 * we can use LOCALE as a codepage */
82 setlocale(LC_ALL, "");
84 load_case_tables_library();
88 Convert a codepoint_t to upper case.
90 _PUBLIC_ codepoint_t toupper_m(codepoint_t val)
95 if (upcase_table == NULL) {
96 load_case_tables_library();
98 if (upcase_table == (void *)-1) {
101 if (val & 0xFFFF0000) {
104 return SVAL(upcase_table, val*2);
108 Convert a codepoint_t to lower case.
110 _PUBLIC_ codepoint_t tolower_m(codepoint_t val)
115 if (lowcase_table == NULL) {
116 load_case_tables_library();
118 if (lowcase_table == (void *)-1) {
121 if (val & 0xFFFF0000) {
124 return SVAL(lowcase_table, val*2);
128 If we upper cased this character, would we get the same character?
130 _PUBLIC_ bool islower_m(codepoint_t val)
132 return (toupper_m(val) != val);
136 If we lower cased this character, would we get the same character?
138 _PUBLIC_ bool isupper_m(codepoint_t val)
140 return (tolower_m(val) != val);
144 compare two codepoints case insensitively
146 _PUBLIC_ int codepoint_cmpi(codepoint_t c1, codepoint_t c2)
149 toupper_m(c1) == toupper_m(c2)) {
156 struct smb_iconv_handle {
157 TALLOC_CTX *child_ctx;
158 const char *unix_charset;
159 const char *dos_charset;
160 const char *display_charset;
161 bool use_builtin_handlers;
162 smb_iconv_t conv_handles[NUM_CHARSETS][NUM_CHARSETS];
165 struct smb_iconv_handle *global_iconv_handle = NULL;
167 struct smb_iconv_handle *get_iconv_handle(void)
169 if (global_iconv_handle == NULL)
170 global_iconv_handle = smb_iconv_handle_reinit(talloc_autofree_context(),
171 "ASCII", "UTF-8", true, NULL);
172 return global_iconv_handle;
175 struct smb_iconv_handle *get_iconv_testing_handle(TALLOC_CTX *mem_ctx,
176 const char *dos_charset,
177 const char *unix_charset,
178 bool use_builtin_handlers)
180 return smb_iconv_handle_reinit(mem_ctx,
181 dos_charset, unix_charset, use_builtin_handlers, NULL);
185 * Return the name of a charset to give to iconv().
187 const char *charset_name(struct smb_iconv_handle *ic, charset_t ch)
190 case CH_UTF16: return "UTF-16LE";
191 case CH_UNIX: return ic->unix_charset;
192 case CH_DOS: return ic->dos_charset;
193 case CH_UTF8: return "UTF8";
194 case CH_UTF16BE: return "UTF-16BE";
195 case CH_UTF16MUNGED: return "UTF16_MUNGED";
202 re-initialize iconv conversion descriptors
204 static int close_iconv_handle(struct smb_iconv_handle *data)
207 for (c1=0;c1<NUM_CHARSETS;c1++) {
208 for (c2=0;c2<NUM_CHARSETS;c2++) {
209 if (data->conv_handles[c1][c2] != NULL) {
210 if (data->conv_handles[c1][c2] != (smb_iconv_t)-1) {
211 smb_iconv_close(data->conv_handles[c1][c2]);
213 data->conv_handles[c1][c2] = NULL;
222 the old_ic is passed in here as the smb_iconv_handle structure
223 is used as a global pointer in some places (eg. python modules). We
224 don't want to invalidate those global pointers, but we do want to
225 update them with the right charset information when loadparm
226 runs. To do that we need to re-use the structure pointer, but
227 re-fill the elements in the structure with the updated values
229 _PUBLIC_ struct smb_iconv_handle *smb_iconv_handle_reinit(TALLOC_CTX *mem_ctx,
230 const char *dos_charset,
231 const char *unix_charset,
232 bool use_builtin_handlers,
233 struct smb_iconv_handle *old_ic)
235 struct smb_iconv_handle *ret;
237 if (old_ic != NULL) {
239 close_iconv_handle(ret);
240 talloc_free(ret->child_ctx);
243 ret = talloc_zero(mem_ctx, struct smb_iconv_handle);
249 /* we use a child context to allow us to free all ptrs without
250 freeing the structure itself */
251 ret->child_ctx = talloc_new(ret);
252 if (ret->child_ctx == NULL) {
256 talloc_set_destructor(ret, close_iconv_handle);
258 if (strcasecmp(dos_charset, "UTF8") == 0 || strcasecmp(dos_charset, "UTF-8") == 0) {
259 DEBUG(0,("ERROR: invalid DOS charset: 'dos charset' must not be UTF8, using (default value) CP850 instead\n"));
260 dos_charset = "CP850";
263 ret->dos_charset = talloc_strdup(ret->child_ctx, dos_charset);
264 ret->unix_charset = talloc_strdup(ret->child_ctx, unix_charset);
265 ret->use_builtin_handlers = use_builtin_handlers;
271 on-demand initialisation of conversion handles
273 smb_iconv_t get_conv_handle(struct smb_iconv_handle *ic,
274 charset_t from, charset_t to)
278 if (ic->conv_handles[from][to]) {
279 return ic->conv_handles[from][to];
282 n1 = charset_name(ic, from);
283 n2 = charset_name(ic, to);
285 ic->conv_handles[from][to] = smb_iconv_open_ex(ic, n2, n1,
286 ic->use_builtin_handlers);
288 if (ic->conv_handles[from][to] == (smb_iconv_t)-1) {
289 if ((from == CH_DOS || to == CH_DOS) &&
290 strcasecmp(charset_name(ic, CH_DOS), "ASCII") != 0) {
291 DEBUG(0,("dos charset '%s' unavailable - using ASCII\n",
292 charset_name(ic, CH_DOS)));
293 ic->dos_charset = "ASCII";
295 n1 = charset_name(ic, from);
296 n2 = charset_name(ic, to);
298 ic->conv_handles[from][to] =
299 smb_iconv_open_ex(ic, n2, n1, ic->use_builtin_handlers);
303 return ic->conv_handles[from][to];
307 * Return the unicode codepoint for the next character in the input
308 * string in the given src_charset.
309 * The unicode codepoint (codepoint_t) is an unsinged 32 bit value.
311 * Also return the number of bytes consumed (which tells the caller
312 * how many bytes to skip to get to the next src_charset-character).
314 * This is implemented (in the non-ascii-case) by first converting the
315 * next character in the input string to UTF16_LE and then calculating
316 * the unicode codepoint from that.
318 * Return INVALID_CODEPOINT if the next character cannot be converted.
320 _PUBLIC_ codepoint_t next_codepoint_handle_ext(
321 struct smb_iconv_handle *ic,
322 const char *str, size_t len,
323 charset_t src_charset,
324 size_t *bytes_consumed)
326 /* it cannot occupy more than 4 bytes in UTF16 format */
328 smb_iconv_t descriptor;
335 if (((str[0] & 0x80) == 0) && (src_charset == CH_DOS ||
336 src_charset == CH_UNIX ||
337 src_charset == CH_UTF8)) {
339 return (codepoint_t)str[0];
343 * we assume that no multi-byte character can take more than 5 bytes.
344 * This is OK as we only support codepoints up to 1M (U+100000)
346 ilen_orig = MIN(len, 5);
349 descriptor = get_conv_handle(ic, src_charset, CH_UTF16);
350 if (descriptor == (smb_iconv_t)-1) {
352 return INVALID_CODEPOINT;
356 * this looks a little strange, but it is needed to cope with
357 * codepoints above 64k (U+1000) which are encoded as per RFC2781.
360 outbuf = (char *)buf;
361 smb_iconv(descriptor, &str, &ilen, &outbuf, &olen);
364 outbuf = (char *)buf;
365 smb_iconv(descriptor, &str, &ilen, &outbuf, &olen);
367 /* we didn't convert any bytes */
369 return INVALID_CODEPOINT;
376 *bytes_consumed = ilen_orig - ilen;
379 return (codepoint_t)SVAL(buf, 0);
382 /* decode a 4 byte UTF16 character manually */
383 return (codepoint_t)0x10000 +
384 (buf[2] | ((buf[3] & 0x3)<<8) |
385 (buf[0]<<10) | ((buf[1] & 0x3)<<18));
388 /* no other length is valid */
389 return INVALID_CODEPOINT;
393 return the unicode codepoint for the next multi-byte CH_UNIX character
396 also return the number of bytes consumed (which tells the caller
397 how many bytes to skip to get to the next CH_UNIX character)
399 return INVALID_CODEPOINT if the next character cannot be converted
401 _PUBLIC_ codepoint_t next_codepoint_handle(struct smb_iconv_handle *ic,
402 const char *str, size_t *size)
405 * We assume that no multi-byte character can take more than 5 bytes
406 * thus avoiding walking all the way down a long string. This is OK as
407 * Unicode codepoints only go up to (U+10ffff), which can always be
408 * encoded in 4 bytes or less.
410 return next_codepoint_handle_ext(ic, str, strnlen(str, 5), CH_UNIX,
415 push a single codepoint into a CH_UNIX string the target string must
416 be able to hold the full character, which is guaranteed if it is at
417 least 5 bytes in size. The caller may pass less than 5 bytes if they
418 are sure the character will fit (for example, you can assume that
419 uppercase/lowercase of a character will not add more than 1 byte)
421 return the number of bytes occupied by the CH_UNIX character, or
424 _PUBLIC_ ssize_t push_codepoint_handle(struct smb_iconv_handle *ic,
425 char *str, codepoint_t c)
427 smb_iconv_t descriptor;
437 descriptor = get_conv_handle(ic,
439 if (descriptor == (smb_iconv_t)-1) {
448 smb_iconv(descriptor, &inbuf, &ilen, &str, &olen);
457 buf[0] = (c>>10) & 0xFF;
458 buf[1] = (c>>18) | 0xd8;
460 buf[3] = ((c>>8) & 0x3) | 0xdc;
466 smb_iconv(descriptor, &inbuf, &ilen, &str, &olen);
473 _PUBLIC_ codepoint_t next_codepoint_ext(const char *str, size_t len,
474 charset_t src_charset, size_t *size)
476 return next_codepoint_handle_ext(get_iconv_handle(), str, len,
480 _PUBLIC_ codepoint_t next_codepoint(const char *str, size_t *size)
482 return next_codepoint_handle(get_iconv_handle(), str, size);
485 _PUBLIC_ ssize_t push_codepoint(char *str, codepoint_t c)
487 return push_codepoint_handle(get_iconv_handle(), str, c);