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) Martin Pool 2003
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/>.
25 /* We can parameterize this if someone complains.... JRA. */
27 char lp_failed_convert_char(void)
35 * @brief Character-set conversion routines built on our iconv.
37 * @note Samba's internal character set (at least in the 3.0 series)
38 * is always the same as the one for the Unix filesystem. It is
39 * <b>not</b> necessarily UTF-8 and may be different on machines that
40 * need i18n filenames to be compatible with Unix software. It does
41 * have to be a superset of ASCII. All multibyte sequences must start
42 * with a byte with the high bit set.
48 static smb_iconv_t conv_handles[NUM_CHARSETS][NUM_CHARSETS];
49 static bool conv_silent; /* Should we do a debug if the conversion fails ? */
50 static bool initialized;
53 * Return the name of a charset to give to iconv().
55 static const char *charset_name(charset_t ch)
57 const char *ret = NULL;
59 if (ch == CH_UTF16LE) ret = "UTF-16LE";
60 else if (ch == CH_UTF16BE) ret = "UTF-16BE";
61 else if (ch == CH_UNIX) ret = lp_unix_charset();
62 else if (ch == CH_DOS) ret = lp_dos_charset();
63 else if (ch == CH_DISPLAY) ret = lp_display_charset();
64 else if (ch == CH_UTF8) ret = "UTF8";
66 #if defined(HAVE_NL_LANGINFO) && defined(CODESET)
67 if (ret && !strcmp(ret, "LOCALE")) {
68 const char *ln = NULL;
71 setlocale(LC_ALL, "");
73 ln = nl_langinfo(CODESET);
75 /* Check whether the charset name is supported
77 smb_iconv_t handle = smb_iconv_open(ln,"UCS-2LE");
78 if (handle == (smb_iconv_t) -1) {
79 DEBUG(5,("Locale charset '%s' unsupported, using ASCII instead\n", ln));
82 DEBUG(5,("Substituting charset '%s' for LOCALE\n", ln));
83 smb_iconv_close(handle);
90 if (!ret || !*ret) ret = "ASCII";
94 void lazy_initialize_conv(void)
104 * Destroy global objects allocated by init_iconv()
106 void gfree_charcnv(void)
110 for (c1=0;c1<NUM_CHARSETS;c1++) {
111 for (c2=0;c2<NUM_CHARSETS;c2++) {
112 if ( conv_handles[c1][c2] ) {
113 smb_iconv_close( conv_handles[c1][c2] );
114 conv_handles[c1][c2] = 0;
122 * Initialize iconv conversion descriptors.
124 * This is called the first time it is needed, and also called again
125 * every time the configuration is reloaded, because the charset or
126 * codepage might have changed.
128 void init_iconv(void)
131 bool did_reload = False;
133 /* so that charset_name() works we need to get the UNIX<->UCS2 going
135 if (!conv_handles[CH_UNIX][CH_UTF16LE])
136 conv_handles[CH_UNIX][CH_UTF16LE] = smb_iconv_open(charset_name(CH_UTF16LE), "ASCII");
138 if (!conv_handles[CH_UTF16LE][CH_UNIX])
139 conv_handles[CH_UTF16LE][CH_UNIX] = smb_iconv_open("ASCII", charset_name(CH_UTF16LE));
141 for (c1=0;c1<NUM_CHARSETS;c1++) {
142 for (c2=0;c2<NUM_CHARSETS;c2++) {
143 const char *n1 = charset_name((charset_t)c1);
144 const char *n2 = charset_name((charset_t)c2);
145 if (conv_handles[c1][c2] &&
146 strcmp(n1, conv_handles[c1][c2]->from_name) == 0 &&
147 strcmp(n2, conv_handles[c1][c2]->to_name) == 0)
152 if (conv_handles[c1][c2])
153 smb_iconv_close(conv_handles[c1][c2]);
155 conv_handles[c1][c2] = smb_iconv_open(n2,n1);
156 if (conv_handles[c1][c2] == (smb_iconv_t)-1) {
157 DEBUG(0,("init_iconv: Conversion from %s to %s not supported\n",
158 charset_name((charset_t)c1), charset_name((charset_t)c2)));
159 if (c1 != CH_UTF16LE && c1 != CH_UTF16BE) {
162 if (c2 != CH_UTF16LE && c2 != CH_UTF16BE) {
165 DEBUG(0,("init_iconv: Attempting to replace with conversion from %s to %s\n",
167 conv_handles[c1][c2] = smb_iconv_open(n2,n1);
168 if (!conv_handles[c1][c2]) {
169 DEBUG(0,("init_iconv: Conversion from %s to %s failed", n1, n2));
170 smb_panic("init_iconv: conv_handle initialization failed");
177 /* XXX: Does this really get called every time the dos
178 * codepage changes? */
179 /* XXX: Is the did_reload test too strict? */
187 * Convert string from one encoding to another, making error checking etc
188 * Slow path version - uses (slow) iconv.
190 * @param src pointer to source string (multibyte or singlebyte)
191 * @param srclen length of the source string in bytes
192 * @param dest pointer to destination string (multibyte or singlebyte)
193 * @param destlen maximal length allowed for string
194 * @param allow_bad_conv determines if a "best effort" conversion is acceptable (never returns errors)
195 * @returns the number of bytes occupied in the destination
197 * Ensure the srclen contains the terminating zero.
201 static size_t convert_string_internal(charset_t from, charset_t to,
202 void const *src, size_t srclen,
203 void *dest, size_t destlen, bool allow_bad_conv)
207 const char* inbuf = (const char*)src;
208 char* outbuf = (char*)dest;
209 smb_iconv_t descriptor;
211 lazy_initialize_conv();
213 descriptor = conv_handles[from][to];
215 if (srclen == (size_t)-1) {
216 if (from == CH_UTF16LE || from == CH_UTF16BE) {
217 srclen = (strlen_w((const smb_ucs2_t *)src)+1) * 2;
219 srclen = strlen((const char *)src)+1;
224 if (descriptor == (smb_iconv_t)-1 || descriptor == (smb_iconv_t)0) {
226 DEBUG(0,("convert_string_internal: Conversion not supported.\n"));
235 retval = smb_iconv(descriptor, &inbuf, &i_len, &outbuf, &o_len);
236 if(retval==(size_t)-1) {
237 const char *reason="unknown error";
240 reason="Incomplete multibyte sequence";
242 DEBUG(3,("convert_string_internal: Conversion error: %s(%s)\n",reason,inbuf));
247 reason="No more room";
249 if (from == CH_UNIX) {
250 DEBUG(3,("E2BIG: convert_string(%s,%s): srclen=%u destlen=%u - '%s'\n",
251 charset_name(from), charset_name(to),
252 (unsigned int)srclen, (unsigned int)destlen, (const char *)src));
254 DEBUG(3,("E2BIG: convert_string(%s,%s): srclen=%u destlen=%u\n",
255 charset_name(from), charset_name(to),
256 (unsigned int)srclen, (unsigned int)destlen));
261 reason="Illegal multibyte sequence";
263 DEBUG(3,("convert_string_internal: Conversion error: %s(%s)\n",reason,inbuf));
269 DEBUG(0,("convert_string_internal: Conversion error: %s(%s)\n",reason,inbuf));
272 /* smb_panic(reason); */
274 return destlen-o_len;
279 * Conversion not supported. This is actually an error, but there are so
280 * many misconfigured iconv systems and smb.conf's out there we can't just
281 * fail. Do a very bad conversion instead.... JRA.
285 if (o_len == 0 || i_len == 0)
286 return destlen - o_len;
288 if (((from == CH_UTF16LE)||(from == CH_UTF16BE)) &&
289 ((to != CH_UTF16LE)||(to != CH_UTF16BE))) {
290 /* Can't convert from utf16 any endian to multibyte.
291 Replace with the default fail char.
294 return destlen - o_len;
296 *outbuf = lp_failed_convert_char();
305 if (o_len == 0 || i_len == 0)
306 return destlen - o_len;
308 /* Keep trying with the next char... */
311 } else if (from != CH_UTF16LE && from != CH_UTF16BE && to == CH_UTF16LE) {
312 /* Can't convert to UTF16LE - just widen by adding the
313 default fail char then zero.
316 return destlen - o_len;
318 outbuf[0] = lp_failed_convert_char();
327 if (o_len == 0 || i_len == 0)
328 return destlen - o_len;
330 /* Keep trying with the next char... */
333 } else if (from != CH_UTF16LE && from != CH_UTF16BE &&
334 to != CH_UTF16LE && to != CH_UTF16BE) {
335 /* Failed multibyte to multibyte. Just copy the default fail char and
337 outbuf[0] = lp_failed_convert_char();
345 if (o_len == 0 || i_len == 0)
346 return destlen - o_len;
348 /* Keep trying with the next char... */
352 /* Keep compiler happy.... */
353 return destlen - o_len;
359 * Convert string from one encoding to another, making error checking etc
360 * Fast path version - handles ASCII first.
362 * @param src pointer to source string (multibyte or singlebyte)
363 * @param srclen length of the source string in bytes, or -1 for nul terminated.
364 * @param dest pointer to destination string (multibyte or singlebyte)
365 * @param destlen maximal length allowed for string - *NEVER* -1.
366 * @param allow_bad_conv determines if a "best effort" conversion is acceptable (never returns errors)
367 * @returns the number of bytes occupied in the destination
369 * Ensure the srclen contains the terminating zero.
371 * This function has been hand-tuned to provide a fast path.
372 * Don't change unless you really know what you are doing. JRA.
375 size_t convert_string(charset_t from, charset_t to,
376 void const *src, size_t srclen,
377 void *dest, size_t destlen, bool allow_bad_conv)
380 * NB. We deliberately don't do a strlen here if srclen == -1.
381 * This is very expensive over millions of calls and is taken
382 * care of in the slow path in convert_string_internal. JRA.
386 SMB_ASSERT(destlen != (size_t)-1);
392 if (from != CH_UTF16LE && from != CH_UTF16BE && to != CH_UTF16LE && to != CH_UTF16BE) {
393 const unsigned char *p = (const unsigned char *)src;
394 unsigned char *q = (unsigned char *)dest;
395 size_t slen = srclen;
396 size_t dlen = destlen;
397 unsigned char lastp = '\0';
400 /* If all characters are ascii, fast path here. */
401 while (slen && dlen) {
402 if ((lastp = *p) <= 0x7f) {
404 if (slen != (size_t)-1) {
412 #ifdef BROKEN_UNICODE_COMPOSE_CHARACTERS
415 return retval + convert_string_internal(from, to, p, slen, q, dlen, allow_bad_conv);
420 /* Even if we fast path we should note if we ran out of room. */
421 if (((slen != (size_t)-1) && slen) ||
422 ((slen == (size_t)-1) && lastp)) {
427 } else if (from == CH_UTF16LE && to != CH_UTF16LE) {
428 const unsigned char *p = (const unsigned char *)src;
429 unsigned char *q = (unsigned char *)dest;
431 size_t slen = srclen;
432 size_t dlen = destlen;
433 unsigned char lastp = '\0';
435 /* If all characters are ascii, fast path here. */
436 while (((slen == (size_t)-1) || (slen >= 2)) && dlen) {
437 if (((lastp = *p) <= 0x7f) && (p[1] == 0)) {
439 if (slen != (size_t)-1) {
448 #ifdef BROKEN_UNICODE_COMPOSE_CHARACTERS
451 return retval + convert_string_internal(from, to, p, slen, q, dlen, allow_bad_conv);
456 /* Even if we fast path we should note if we ran out of room. */
457 if (((slen != (size_t)-1) && slen) ||
458 ((slen == (size_t)-1) && lastp)) {
463 } else if (from != CH_UTF16LE && from != CH_UTF16BE && to == CH_UTF16LE) {
464 const unsigned char *p = (const unsigned char *)src;
465 unsigned char *q = (unsigned char *)dest;
467 size_t slen = srclen;
468 size_t dlen = destlen;
469 unsigned char lastp = '\0';
471 /* If all characters are ascii, fast path here. */
472 while (slen && (dlen >= 2)) {
473 if ((lastp = *p) <= 0x7F) {
476 if (slen != (size_t)-1) {
484 #ifdef BROKEN_UNICODE_COMPOSE_CHARACTERS
487 return retval + convert_string_internal(from, to, p, slen, q, dlen, allow_bad_conv);
492 /* Even if we fast path we should note if we ran out of room. */
493 if (((slen != (size_t)-1) && slen) ||
494 ((slen == (size_t)-1) && lastp)) {
501 #ifdef BROKEN_UNICODE_COMPOSE_CHARACTERS
504 return convert_string_internal(from, to, src, srclen, dest, destlen, allow_bad_conv);
508 * Convert between character sets, allocating a new buffer for the result.
510 * @param ctx TALLOC_CTX to use to allocate with. If NULL use malloc.
511 * (this is a bad interface and needs fixing. JRA).
512 * @param srclen length of source buffer.
513 * @param dest always set at least to NULL
514 * @param converted_size set to the size of the allocated buffer on return
516 * @note -1 is not accepted for srclen.
518 * @return true if new buffer was correctly allocated, and string was
521 * Ensure the srclen contains the terminating zero.
523 * I hate the goto's in this function. It's embarressing.....
524 * There has to be a cleaner way to do this. JRA.
527 bool convert_string_allocate(TALLOC_CTX *ctx, charset_t from, charset_t to,
528 void const *src, size_t srclen, void *dst,
529 size_t *converted_size, bool allow_bad_conv)
531 size_t i_len, o_len, destlen = (srclen * 3) / 2;
533 const char *inbuf = (const char *)src;
534 char *outbuf = NULL, *ob = NULL;
535 smb_iconv_t descriptor;
536 void **dest = (void **)dst;
540 if (!converted_size) {
545 if (src == NULL || srclen == (size_t)-1) {
554 lazy_initialize_conv();
556 descriptor = conv_handles[from][to];
558 if (descriptor == (smb_iconv_t)-1 || descriptor == (smb_iconv_t)0) {
560 DEBUG(0,("convert_string_allocate: Conversion not supported.\n"));
567 /* +2 is for ucs2 null termination. */
568 if ((destlen*2)+2 < destlen) {
569 /* wrapped ! abort. */
571 DEBUG(0, ("convert_string_allocate: destlen wrapped !\n"));
577 destlen = destlen * 2;
580 /* +2 is for ucs2 null termination. */
582 ob = (char *)TALLOC_REALLOC(ctx, ob, destlen + 2);
584 ob = (char *)SMB_REALLOC(ob, destlen + 2);
588 DEBUG(0, ("convert_string_allocate: realloc failed!\n"));
598 retval = smb_iconv(descriptor,
601 if(retval == (size_t)-1) {
602 const char *reason="unknown error";
605 reason="Incomplete multibyte sequence";
607 DEBUG(3,("convert_string_allocate: Conversion error: %s(%s)\n",reason,inbuf));
614 reason="Illegal multibyte sequence";
616 DEBUG(3,("convert_string_allocate: Conversion error: %s(%s)\n",reason,inbuf));
622 DEBUG(0,("Conversion error: %s(%s)\n",reason,inbuf));
623 /* smb_panic(reason); */
634 destlen = destlen - o_len;
635 /* Don't shrink unless we're reclaiming a lot of
636 * space. This is in the hot codepath and these
637 * reallocs *cost*. JRA.
640 /* We're shrinking here so we know the +2 is safe from wrap. */
642 ob = (char *)TALLOC_REALLOC(ctx,ob,destlen + 2);
644 ob = (char *)SMB_REALLOC(ob,destlen + 2);
648 if (destlen && !ob) {
649 DEBUG(0, ("convert_string_allocate: out of memory!\n"));
656 /* Must ucs2 null terminate in the extra space we allocated. */
658 ob[destlen+1] = '\0';
660 *converted_size = destlen;
666 * Conversion not supported. This is actually an error, but there are so
667 * many misconfigured iconv systems and smb.conf's out there we can't just
668 * fail. Do a very bad conversion instead.... JRA.
672 if (o_len == 0 || i_len == 0)
675 if (((from == CH_UTF16LE)||(from == CH_UTF16BE)) &&
676 ((to != CH_UTF16LE)||(to != CH_UTF16BE))) {
677 /* Can't convert from utf16 any endian to multibyte.
678 Replace with the default fail char.
685 *outbuf = lp_failed_convert_char();
694 if (o_len == 0 || i_len == 0)
697 /* Keep trying with the next char... */
700 } else if (from != CH_UTF16LE && from != CH_UTF16BE && to == CH_UTF16LE) {
701 /* Can't convert to UTF16LE - just widen by adding the
702 default fail char then zero.
707 outbuf[0] = lp_failed_convert_char();
716 if (o_len == 0 || i_len == 0)
719 /* Keep trying with the next char... */
722 } else if (from != CH_UTF16LE && from != CH_UTF16BE &&
723 to != CH_UTF16LE && to != CH_UTF16BE) {
724 /* Failed multibyte to multibyte. Just copy the default fail char and
726 outbuf[0] = lp_failed_convert_char();
734 if (o_len == 0 || i_len == 0)
737 /* Keep trying with the next char... */
741 /* Keep compiler happy.... */
748 * Convert between character sets, allocating a new buffer using talloc for the result.
750 * @param srclen length of source buffer.
751 * @param dest always set at least to NULL
752 * @parm converted_size set to the number of bytes occupied by the string in
753 * the destination on success.
754 * @note -1 is not accepted for srclen.
756 * @return true if new buffer was correctly allocated, and string was
759 bool convert_string_talloc(TALLOC_CTX *ctx, charset_t from, charset_t to,
760 void const *src, size_t srclen, void *dst,
761 size_t *converted_size, bool allow_bad_conv)
763 void **dest = (void **)dst;
766 return convert_string_allocate(ctx, from, to, src, srclen, dest,
767 converted_size, allow_bad_conv);
770 size_t unix_strupper(const char *src, size_t srclen, char *dest, size_t destlen)
775 if (!push_ucs2_allocate(&buffer, src, &size)) {
779 if (!strupper_w(buffer) && (dest == src)) {
784 size = convert_string(CH_UTF16LE, CH_UNIX, buffer, size, dest, destlen, True);
790 strdup() a unix string to upper case.
793 char *strdup_upper(const char *s)
795 char *out_buffer = SMB_STRDUP(s);
796 const unsigned char *p = (const unsigned char *)s;
797 unsigned char *q = (unsigned char *)out_buffer;
803 /* this is quite a common operation, so we want it to be
804 fast. We optimise for the ascii case, knowing that all our
805 supported multi-byte character sets are ascii-compatible
806 (ie. they match for the first 128 chars) */
811 *q++ = toupper_ascii_fast(*p);
817 size_t converted_size, converted_size2;
818 smb_ucs2_t *buffer = NULL;
820 SAFE_FREE(out_buffer);
821 if (!convert_string_allocate(NULL, CH_UNIX, CH_UTF16LE, s,
823 (void **)(void *)&buffer,
824 &converted_size, True))
831 if (!convert_string_allocate(NULL, CH_UTF16LE, CH_UNIX, buffer,
833 (void **)(void *)&out_buffer,
834 &converted_size2, True))
840 /* Don't need the intermediate buffer
850 talloc_strdup() a unix string to upper case.
853 char *talloc_strdup_upper(TALLOC_CTX *ctx, const char *s)
855 char *out_buffer = talloc_strdup(ctx,s);
856 const unsigned char *p = (const unsigned char *)s;
857 unsigned char *q = (unsigned char *)out_buffer;
863 /* this is quite a common operation, so we want it to be
864 fast. We optimise for the ascii case, knowing that all our
865 supported multi-byte character sets are ascii-compatible
866 (ie. they match for the first 128 chars) */
871 *q++ = toupper_ascii_fast(*p);
877 size_t converted_size, converted_size2;
878 smb_ucs2_t *ubuf = NULL;
880 /* We're not using the ascii buffer above. */
881 TALLOC_FREE(out_buffer);
883 if (!convert_string_talloc(ctx, CH_UNIX, CH_UTF16LE, s,
884 strlen(s)+1, (void *)&ubuf,
885 &converted_size, True))
892 if (!convert_string_talloc(ctx, CH_UTF16LE, CH_UNIX, ubuf,
893 converted_size, (void *)&out_buffer,
894 &converted_size2, True))
900 /* Don't need the intermediate buffer
909 size_t unix_strlower(const char *src, size_t srclen, char *dest, size_t destlen)
912 smb_ucs2_t *buffer = NULL;
914 if (!convert_string_allocate(NULL, CH_UNIX, CH_UTF16LE, src, srclen,
915 (void **)(void *)&buffer, &size,
918 smb_panic("failed to create UCS2 buffer");
920 if (!strlower_w(buffer) && (dest == src)) {
924 size = convert_string(CH_UTF16LE, CH_UNIX, buffer, size, dest, destlen, True);
930 strdup() a unix string to lower case.
933 char *strdup_lower(const char *s)
935 size_t converted_size;
936 smb_ucs2_t *buffer = NULL;
939 if (!push_ucs2_allocate(&buffer, s, &converted_size)) {
945 if (!pull_ucs2_allocate(&out_buffer, buffer, &converted_size)) {
955 char *talloc_strdup_lower(TALLOC_CTX *ctx, const char *s)
957 size_t converted_size;
958 smb_ucs2_t *buffer = NULL;
961 if (!push_ucs2_talloc(ctx, &buffer, s, &converted_size)) {
967 if (!pull_ucs2_talloc(ctx, &out_buffer, buffer, &converted_size)) {
978 size_t ucs2_align(const void *base_ptr, const void *p, int flags)
980 if (flags & (STR_NOALIGN|STR_ASCII))
982 return PTR_DIFF(p, base_ptr) & 1;
987 * Copy a string from a char* unix src to a dos codepage string destination.
989 * @return the number of bytes occupied by the string in the destination.
991 * @param flags can include
993 * <dt>STR_TERMINATE</dt> <dd>means include the null termination</dd>
994 * <dt>STR_UPPER</dt> <dd>means uppercase in the destination</dd>
997 * @param dest_len the maximum length in bytes allowed in the
1000 size_t push_ascii(void *dest, const char *src, size_t dest_len, int flags)
1002 size_t src_len = strlen(src);
1003 char *tmpbuf = NULL;
1006 /* No longer allow a length of -1. */
1007 if (dest_len == (size_t)-1) {
1008 smb_panic("push_ascii - dest_len == -1");
1011 if (flags & STR_UPPER) {
1012 tmpbuf = SMB_STRDUP(src);
1014 smb_panic("malloc fail");
1020 if (flags & (STR_TERMINATE | STR_TERMINATE_ASCII)) {
1024 ret = convert_string(CH_UNIX, CH_DOS, src, src_len, dest, dest_len, True);
1025 if (ret == (size_t)-1 &&
1026 (flags & (STR_TERMINATE | STR_TERMINATE_ASCII))
1028 ((char *)dest)[0] = '\0';
1034 size_t push_ascii_fstring(void *dest, const char *src)
1036 return push_ascii(dest, src, sizeof(fstring), STR_TERMINATE);
1039 /********************************************************************
1040 Push an nstring - ensure null terminated. Written by
1041 moriyama@miraclelinux.com (MORIYAMA Masayuki).
1042 ********************************************************************/
1044 size_t push_ascii_nstring(void *dest, const char *src)
1046 size_t i, buffer_len, dest_len;
1050 if (!push_ucs2_allocate(&buffer, src, &buffer_len)) {
1051 smb_panic("failed to create UCS2 buffer");
1054 /* We're using buffer_len below to count ucs2 characters, not bytes. */
1055 buffer_len /= sizeof(smb_ucs2_t);
1058 for (i = 0; buffer[i] != 0 && (i < buffer_len); i++) {
1059 unsigned char mb[10];
1060 /* Convert one smb_ucs2_t character at a time. */
1061 size_t mb_len = convert_string(CH_UTF16LE, CH_DOS, buffer+i, sizeof(smb_ucs2_t), mb, sizeof(mb), False);
1062 if ((mb_len != (size_t)-1) && (dest_len + mb_len <= MAX_NETBIOSNAME_LEN - 1)) {
1063 memcpy((char *)dest + dest_len, mb, mb_len);
1070 ((char *)dest)[dest_len] = '\0';
1073 conv_silent = False;
1077 /********************************************************************
1078 Push and malloc an ascii string. src and dest null terminated.
1079 ********************************************************************/
1081 bool push_ascii_allocate(char **dest, const char *src, size_t *converted_size)
1083 size_t src_len = strlen(src)+1;
1086 return convert_string_allocate(NULL, CH_UNIX, CH_DOS, src, src_len,
1087 (void **)dest, converted_size, True);
1091 * Copy a string from a dos codepage source to a unix char* destination.
1093 * The resulting string in "dest" is always null terminated.
1095 * @param flags can have:
1097 * <dt>STR_TERMINATE</dt>
1098 * <dd>STR_TERMINATE means the string in @p src
1099 * is null terminated, and src_len is ignored.</dd>
1102 * @param src_len is the length of the source area in bytes.
1103 * @returns the number of bytes occupied by the string in @p src.
1105 size_t pull_ascii(char *dest, const void *src, size_t dest_len, size_t src_len, int flags)
1109 if (dest_len == (size_t)-1) {
1110 /* No longer allow dest_len of -1. */
1111 smb_panic("pull_ascii - invalid dest_len of -1");
1114 if (flags & STR_TERMINATE) {
1115 if (src_len == (size_t)-1) {
1116 src_len = strlen((const char *)src) + 1;
1118 size_t len = strnlen((const char *)src, src_len);
1125 ret = convert_string(CH_DOS, CH_UNIX, src, src_len, dest, dest_len, True);
1126 if (ret == (size_t)-1) {
1131 if (dest_len && ret) {
1132 /* Did we already process the terminating zero ? */
1133 if (dest[MIN(ret-1, dest_len-1)] != 0) {
1134 dest[MIN(ret, dest_len-1)] = 0;
1144 * Copy a string from a dos codepage source to a unix char* destination.
1146 Uses malloc if TALLOC_CTX is NULL (this is a bad interface and
1149 * The resulting string in "dest" is always null terminated.
1151 * @param flags can have:
1153 * <dt>STR_TERMINATE</dt>
1154 * <dd>STR_TERMINATE means the string in @p src
1155 * is null terminated, and src_len is ignored.</dd>
1158 * @param src_len is the length of the source area in bytes.
1159 * @returns the number of bytes occupied by the string in @p src.
1162 static size_t pull_ascii_base_talloc(TALLOC_CTX *ctx,
1172 /* Ensure we never use the braindead "malloc" varient. */
1174 smb_panic("NULL talloc CTX in pull_ascii_base_talloc\n");
1184 if (flags & STR_TERMINATE) {
1185 if (src_len == (size_t)-1) {
1186 src_len = strlen((const char *)src) + 1;
1188 size_t len = strnlen((const char *)src, src_len);
1193 /* Ensure we don't use an insane length from the client. */
1194 if (src_len >= 1024*1024) {
1195 char *msg = talloc_asprintf(ctx,
1196 "Bad src length (%u) in "
1197 "pull_ascii_base_talloc",
1198 (unsigned int)src_len);
1202 /* Can't have an unlimited length
1203 * non STR_TERMINATE'd.
1205 if (src_len == (size_t)-1) {
1211 /* src_len != -1 here. */
1213 if (!convert_string_allocate(ctx, CH_DOS, CH_UNIX, src, src_len, &dest,
1218 if (dest_len && dest) {
1219 /* Did we already process the terminating zero ? */
1220 if (dest[dest_len-1] != 0) {
1221 size_t size = talloc_get_size(dest);
1222 /* Have we got space to append the '\0' ? */
1223 if (size <= dest_len) {
1225 dest = TALLOC_REALLOC_ARRAY(ctx, dest, char,
1229 dest_len = (size_t)-1;
1234 dest[dest_len] = '\0';
1245 size_t pull_ascii_fstring(char *dest, const void *src)
1247 return pull_ascii(dest, src, sizeof(fstring), -1, STR_TERMINATE);
1250 /* When pulling an nstring it can expand into a larger size (dos cp -> utf8). Cope with this. */
1252 size_t pull_ascii_nstring(char *dest, size_t dest_len, const void *src)
1254 return pull_ascii(dest, src, dest_len, sizeof(nstring)-1, STR_TERMINATE);
1258 * Copy a string from a char* src to a unicode destination.
1260 * @returns the number of bytes occupied by the string in the destination.
1262 * @param flags can have:
1265 * <dt>STR_TERMINATE <dd>means include the null termination.
1266 * <dt>STR_UPPER <dd>means uppercase in the destination.
1267 * <dt>STR_NOALIGN <dd>means don't do alignment.
1270 * @param dest_len is the maximum length allowed in the
1274 size_t push_ucs2(const void *base_ptr, void *dest, const char *src, size_t dest_len, int flags)
1280 if (dest_len == (size_t)-1) {
1281 /* No longer allow dest_len of -1. */
1282 smb_panic("push_ucs2 - invalid dest_len of -1");
1285 if (flags & STR_TERMINATE)
1286 src_len = (size_t)-1;
1288 src_len = strlen(src);
1290 if (ucs2_align(base_ptr, dest, flags)) {
1292 dest = (void *)((char *)dest + 1);
1298 /* ucs2 is always a multiple of 2 bytes */
1301 ret = convert_string(CH_UNIX, CH_UTF16LE, src, src_len, dest, dest_len, True);
1302 if (ret == (size_t)-1) {
1303 if ((flags & STR_TERMINATE) &&
1313 if (flags & STR_UPPER) {
1314 smb_ucs2_t *dest_ucs2 = (smb_ucs2_t *)dest;
1317 /* We check for i < (ret / 2) below as the dest string isn't null
1318 terminated if STR_TERMINATE isn't set. */
1320 for (i = 0; i < (ret / 2) && i < (dest_len / 2) && dest_ucs2[i]; i++) {
1321 smb_ucs2_t v = toupper_w(dest_ucs2[i]);
1322 if (v != dest_ucs2[i]) {
1333 * Copy a string from a unix char* src to a UCS2 destination,
1334 * allocating a buffer using talloc().
1336 * @param dest always set at least to NULL
1337 * @parm converted_size set to the number of bytes occupied by the string in
1338 * the destination on success.
1340 * @return true if new buffer was correctly allocated, and string was
1343 bool push_ucs2_talloc(TALLOC_CTX *ctx, smb_ucs2_t **dest, const char *src,
1344 size_t *converted_size)
1346 size_t src_len = strlen(src)+1;
1349 return convert_string_talloc(ctx, CH_UNIX, CH_UTF16LE, src, src_len,
1350 (void **)dest, converted_size, True);
1355 * Copy a string from a unix char* src to a UCS2 destination, allocating a buffer
1357 * @param dest always set at least to NULL
1358 * @parm converted_size set to the number of bytes occupied by the string in
1359 * the destination on success.
1361 * @return true if new buffer was correctly allocated, and string was
1365 bool push_ucs2_allocate(smb_ucs2_t **dest, const char *src,
1366 size_t *converted_size)
1368 size_t src_len = strlen(src)+1;
1371 return convert_string_allocate(NULL, CH_UNIX, CH_UTF16LE, src, src_len,
1372 (void **)dest, converted_size, True);
1376 Copy a string from a char* src to a UTF-8 destination.
1377 Return the number of bytes occupied by the string in the destination
1379 STR_TERMINATE means include the null termination
1380 STR_UPPER means uppercase in the destination
1381 dest_len is the maximum length allowed in the destination. If dest_len
1382 is -1 then no maxiumum is used.
1385 static size_t push_utf8(void *dest, const char *src, size_t dest_len, int flags)
1389 char *tmpbuf = NULL;
1391 if (dest_len == (size_t)-1) {
1392 /* No longer allow dest_len of -1. */
1393 smb_panic("push_utf8 - invalid dest_len of -1");
1396 if (flags & STR_UPPER) {
1397 tmpbuf = strdup_upper(src);
1402 src_len = strlen(src);
1405 src_len = strlen(src);
1406 if (flags & STR_TERMINATE) {
1410 ret = convert_string(CH_UNIX, CH_UTF8, src, src_len, dest, dest_len, True);
1415 size_t push_utf8_fstring(void *dest, const char *src)
1417 return push_utf8(dest, src, sizeof(fstring), STR_TERMINATE);
1421 * Copy a string from a unix char* src to a UTF-8 destination, allocating a buffer using talloc
1423 * @param dest always set at least to NULL
1424 * @parm converted_size set to the number of bytes occupied by the string in
1425 * the destination on success.
1427 * @return true if new buffer was correctly allocated, and string was
1431 bool push_utf8_talloc(TALLOC_CTX *ctx, char **dest, const char *src,
1432 size_t *converted_size)
1434 size_t src_len = strlen(src)+1;
1437 return convert_string_talloc(ctx, CH_UNIX, CH_UTF8, src, src_len,
1438 (void**)dest, converted_size, True);
1442 * Copy a string from a unix char* src to a UTF-8 destination, allocating a buffer
1444 * @param dest always set at least to NULL
1445 * @parm converted_size set to the number of bytes occupied by the string in
1446 * the destination on success.
1448 * @return true if new buffer was correctly allocated, and string was
1452 bool push_utf8_allocate(char **dest, const char *src, size_t *converted_size)
1454 size_t src_len = strlen(src)+1;
1457 return convert_string_allocate(NULL, CH_UNIX, CH_UTF8, src, src_len,
1458 (void **)dest, converted_size, True);
1462 Copy a string from a ucs2 source to a unix char* destination.
1464 STR_TERMINATE means the string in src is null terminated.
1465 STR_NOALIGN means don't try to align.
1466 if STR_TERMINATE is set then src_len is ignored if it is -1.
1467 src_len is the length of the source area in bytes
1468 Return the number of bytes occupied by the string in src.
1469 The resulting string in "dest" is always null terminated.
1472 size_t pull_ucs2(const void *base_ptr, char *dest, const void *src, size_t dest_len, size_t src_len, int flags)
1476 if (dest_len == (size_t)-1) {
1477 /* No longer allow dest_len of -1. */
1478 smb_panic("pull_ucs2 - invalid dest_len of -1");
1482 if (dest && dest_len > 0) {
1488 if (ucs2_align(base_ptr, src, flags)) {
1489 src = (const void *)((const char *)src + 1);
1490 if (src_len != (size_t)-1)
1494 if (flags & STR_TERMINATE) {
1495 /* src_len -1 is the default for null terminated strings. */
1496 if (src_len != (size_t)-1) {
1497 size_t len = strnlen_w((const smb_ucs2_t *)src,
1499 if (len < src_len/2)
1505 /* ucs2 is always a multiple of 2 bytes */
1506 if (src_len != (size_t)-1)
1509 ret = convert_string(CH_UTF16LE, CH_UNIX, src, src_len, dest, dest_len, True);
1510 if (ret == (size_t)-1) {
1515 if (src_len == (size_t)-1)
1518 if (dest_len && ret) {
1519 /* Did we already process the terminating zero ? */
1520 if (dest[MIN(ret-1, dest_len-1)] != 0) {
1521 dest[MIN(ret, dest_len-1)] = 0;
1531 Copy a string from a ucs2 source to a unix char* destination.
1532 Talloc version with a base pointer.
1533 Uses malloc if TALLOC_CTX is NULL (this is a bad interface and
1536 STR_TERMINATE means the string in src is null terminated.
1537 STR_NOALIGN means don't try to align.
1538 if STR_TERMINATE is set then src_len is ignored if it is -1.
1539 src_len is the length of the source area in bytes
1540 Return the number of bytes occupied by the string in src.
1541 The resulting string in "dest" is always null terminated.
1544 size_t pull_ucs2_base_talloc(TALLOC_CTX *ctx,
1545 const void *base_ptr,
1557 /* Ensure we never use the braindead "malloc" varient. */
1559 smb_panic("NULL talloc CTX in pull_ucs2_base_talloc\n");
1567 if (ucs2_align(base_ptr, src, flags)) {
1568 src = (const void *)((const char *)src + 1);
1569 if (src_len != (size_t)-1)
1573 if (flags & STR_TERMINATE) {
1574 /* src_len -1 is the default for null terminated strings. */
1575 if (src_len != (size_t)-1) {
1576 size_t len = strnlen_w((const smb_ucs2_t *)src,
1578 if (len < src_len/2)
1583 * src_len == -1 - alloc interface won't take this
1584 * so we must calculate.
1586 src_len = (strlen_w((const smb_ucs2_t *)src)+1)*sizeof(smb_ucs2_t);
1588 /* Ensure we don't use an insane length from the client. */
1589 if (src_len >= 1024*1024) {
1590 smb_panic("Bad src length in pull_ucs2_base_talloc\n");
1593 /* Can't have an unlimited length
1594 * non STR_TERMINATE'd.
1596 if (src_len == (size_t)-1) {
1602 /* src_len != -1 here. */
1604 /* ucs2 is always a multiple of 2 bytes */
1607 if (!convert_string_talloc(ctx, CH_UTF16LE, CH_UNIX, src, src_len,
1608 (void *)&dest, &dest_len, True)) {
1613 /* Did we already process the terminating zero ? */
1614 if (dest[dest_len-1] != 0) {
1615 size_t size = talloc_get_size(dest);
1616 /* Have we got space to append the '\0' ? */
1617 if (size <= dest_len) {
1619 dest = TALLOC_REALLOC_ARRAY(ctx, dest, char,
1623 dest_len = (size_t)-1;
1628 dest[dest_len] = '\0';
1639 size_t pull_ucs2_fstring(char *dest, const void *src)
1641 return pull_ucs2(NULL, dest, src, sizeof(fstring), -1, STR_TERMINATE);
1645 * Copy a string from a UCS2 src to a unix char * destination, allocating a buffer using talloc
1647 * @param dest always set at least to NULL
1648 * @parm converted_size set to the number of bytes occupied by the string in
1649 * the destination on success.
1651 * @return true if new buffer was correctly allocated, and string was
1655 bool pull_ucs2_talloc(TALLOC_CTX *ctx, char **dest, const smb_ucs2_t *src,
1656 size_t *converted_size)
1658 size_t src_len = (strlen_w(src)+1) * sizeof(smb_ucs2_t);
1661 return convert_string_talloc(ctx, CH_UTF16LE, CH_UNIX, src, src_len,
1662 (void **)dest, converted_size, True);
1666 * Copy a string from a UCS2 src to a unix char * destination, allocating a buffer
1668 * @param dest always set at least to NULL
1669 * @parm converted_size set to the number of bytes occupied by the string in
1670 * the destination on success.
1671 * @return true if new buffer was correctly allocated, and string was
1675 bool pull_ucs2_allocate(char **dest, const smb_ucs2_t *src,
1676 size_t *converted_size)
1678 size_t src_len = (strlen_w(src)+1) * sizeof(smb_ucs2_t);
1681 return convert_string_allocate(NULL, CH_UTF16LE, CH_UNIX, src, src_len,
1682 (void **)dest, converted_size, True);
1686 * Copy a string from a UTF-8 src to a unix char * destination, allocating a buffer using talloc
1688 * @param dest always set at least to NULL
1689 * @parm converted_size set to the number of bytes occupied by the string in
1690 * the destination on success.
1692 * @return true if new buffer was correctly allocated, and string was
1696 bool pull_utf8_talloc(TALLOC_CTX *ctx, char **dest, const char *src,
1697 size_t *converted_size)
1699 size_t src_len = strlen(src)+1;
1702 return convert_string_talloc(ctx, CH_UTF8, CH_UNIX, src, src_len,
1703 (void **)dest, converted_size, True);
1707 * Copy a string from a UTF-8 src to a unix char * destination, allocating a buffer
1709 * @param dest always set at least to NULL
1710 * @parm converted_size set to the number of bytes occupied by the string in
1711 * the destination on success.
1713 * @return true if new buffer was correctly allocated, and string was
1717 bool pull_utf8_allocate(char **dest, const char *src, size_t *converted_size)
1719 size_t src_len = strlen(src)+1;
1722 return convert_string_allocate(NULL, CH_UTF8, CH_UNIX, src, src_len,
1723 (void **)dest, converted_size, True);
1727 * Copy a string from a DOS src to a unix char * destination, allocating a buffer using talloc
1729 * @param dest always set at least to NULL
1730 * @parm converted_size set to the number of bytes occupied by the string in
1731 * the destination on success.
1733 * @return true if new buffer was correctly allocated, and string was
1737 bool pull_ascii_talloc(TALLOC_CTX *ctx, char **dest, const char *src,
1738 size_t *converted_size)
1740 size_t src_len = strlen(src)+1;
1743 return convert_string_talloc(ctx, CH_DOS, CH_UNIX, src, src_len,
1744 (void **)dest, converted_size, True);
1748 Copy a string from a char* src to a unicode or ascii
1749 dos codepage destination choosing unicode or ascii based on the
1750 flags in the SMB buffer starting at base_ptr.
1751 Return the number of bytes occupied by the string in the destination.
1753 STR_TERMINATE means include the null termination.
1754 STR_UPPER means uppercase in the destination.
1755 STR_ASCII use ascii even with unicode packet.
1756 STR_NOALIGN means don't do alignment.
1757 dest_len is the maximum length allowed in the destination. If dest_len
1758 is -1 then no maxiumum is used.
1761 size_t push_string_fn(const char *function, unsigned int line,
1762 const void *base_ptr, uint16 flags2,
1763 void *dest, const char *src,
1764 size_t dest_len, int flags)
1767 /* We really need to zero fill here, not clobber
1768 * region, as we want to ensure that valgrind thinks
1769 * all of the outgoing buffer has been written to
1770 * so a send() or write() won't trap an error.
1774 clobber_region(function, line, dest, dest_len);
1776 memset(dest, '\0', dest_len);
1780 if (!(flags & STR_ASCII) && \
1781 ((flags & STR_UNICODE || \
1782 (flags2 & FLAGS2_UNICODE_STRINGS)))) {
1783 return push_ucs2(base_ptr, dest, src, dest_len, flags);
1785 return push_ascii(dest, src, dest_len, flags);
1790 Copy a string from a unicode or ascii source (depending on
1791 the packet flags) to a char* destination.
1793 STR_TERMINATE means the string in src is null terminated.
1794 STR_UNICODE means to force as unicode.
1795 STR_ASCII use ascii even with unicode packet.
1796 STR_NOALIGN means don't do alignment.
1797 if STR_TERMINATE is set then src_len is ignored is it is -1
1798 src_len is the length of the source area in bytes.
1799 Return the number of bytes occupied by the string in src.
1800 The resulting string in "dest" is always null terminated.
1803 size_t pull_string_fn(const char *function,
1805 const void *base_ptr,
1814 clobber_region(function, line, dest, dest_len);
1817 if ((base_ptr == NULL) && ((flags & (STR_ASCII|STR_UNICODE)) == 0)) {
1818 smb_panic("No base ptr to get flg2 and neither ASCII nor "
1822 if (!(flags & STR_ASCII) && \
1823 ((flags & STR_UNICODE || \
1824 (smb_flags2 & FLAGS2_UNICODE_STRINGS)))) {
1825 return pull_ucs2(base_ptr, dest, src, dest_len, src_len, flags);
1827 return pull_ascii(dest, src, dest_len, src_len, flags);
1831 Copy a string from a unicode or ascii source (depending on
1832 the packet flags) to a char* destination.
1833 Variant that uses talloc.
1835 STR_TERMINATE means the string in src is null terminated.
1836 STR_UNICODE means to force as unicode.
1837 STR_ASCII use ascii even with unicode packet.
1838 STR_NOALIGN means don't do alignment.
1839 if STR_TERMINATE is set then src_len is ignored is it is -1
1840 src_len is the length of the source area in bytes.
1841 Return the number of bytes occupied by the string in src.
1842 The resulting string in "dest" is always null terminated.
1845 size_t pull_string_talloc_fn(const char *function,
1848 const void *base_ptr,
1855 if ((base_ptr == NULL) && ((flags & (STR_ASCII|STR_UNICODE)) == 0)) {
1856 smb_panic("No base ptr to get flg2 and neither ASCII nor "
1860 if (!(flags & STR_ASCII) && \
1861 ((flags & STR_UNICODE || \
1862 (smb_flags2 & FLAGS2_UNICODE_STRINGS)))) {
1863 return pull_ucs2_base_talloc(ctx,
1870 return pull_ascii_base_talloc(ctx,
1878 size_t align_string(const void *base_ptr, const char *p, int flags)
1880 if (!(flags & STR_ASCII) && \
1881 ((flags & STR_UNICODE || \
1882 (SVAL(base_ptr, smb_flg2) & FLAGS2_UNICODE_STRINGS)))) {
1883 return ucs2_align(base_ptr, p, flags);
1889 Return the unicode codepoint for the next multi-byte CH_UNIX character
1890 in the string. The unicode codepoint (codepoint_t) is an unsinged 32 bit value.
1892 Also return the number of bytes consumed (which tells the caller
1893 how many bytes to skip to get to the next CH_UNIX character).
1895 Return INVALID_CODEPOINT if the next character cannot be converted.
1898 codepoint_t next_codepoint(const char *str, size_t *size)
1900 /* It cannot occupy more than 4 bytes in UTF16 format */
1902 smb_iconv_t descriptor;
1908 if ((str[0] & 0x80) == 0) {
1910 return (codepoint_t)str[0];
1913 /* We assume that no multi-byte character can take
1914 more than 5 bytes. This is OK as we only
1915 support codepoints up to 1M */
1917 ilen_orig = strnlen(str, 5);
1920 lazy_initialize_conv();
1922 descriptor = conv_handles[CH_UNIX][CH_UTF16LE];
1923 if (descriptor == (smb_iconv_t)-1 || descriptor == (smb_iconv_t)0) {
1925 return INVALID_CODEPOINT;
1928 /* This looks a little strange, but it is needed to cope
1929 with codepoints above 64k which are encoded as per RFC2781. */
1931 outbuf = (char *)buf;
1932 smb_iconv(descriptor, &str, &ilen, &outbuf, &olen);
1934 /* We failed to convert to a 2 byte character.
1935 See if we can convert to a 4 UTF16-LE byte char encoding.
1938 outbuf = (char *)buf;
1939 smb_iconv(descriptor, &str, &ilen, &outbuf, &olen);
1941 /* We didn't convert any bytes */
1943 return INVALID_CODEPOINT;
1950 *size = ilen_orig - ilen;
1953 /* 2 byte, UTF16-LE encoded value. */
1954 return (codepoint_t)SVAL(buf, 0);
1957 /* Decode a 4 byte UTF16-LE character manually.
1958 See RFC2871 for the encoding machanism.
1960 codepoint_t w1 = SVAL(buf,0) & ~0xD800;
1961 codepoint_t w2 = SVAL(buf,2) & ~0xDC00;
1963 return (codepoint_t)0x10000 +
1967 /* no other length is valid */
1968 return INVALID_CODEPOINT;