libatalk/unicode/utf8.c

   1 /*
   2    Unix SMB/CIFS implementation.
   3    minimal iconv implementation
   4    Copyright (C) Andrew Tridgell 2001
   5    Copyright (C) Jelmer Vernooij 2002,2003
   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 2 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, write to the Free Software
  19    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  20
  21    From samba 3.0 beta and GNU libiconv-1.8
  22    It's bad but most of the time we can't use libc iconv service:
  23    - it doesn't round trip for most encoding
  24    - it doesn't know about Apple extension
  25 */
  26
  27 #ifdef HAVE_CONFIG_H
  28 #include "config.h"
  29 #endif /* HAVE_CONFIG_H */
  30 #include <stdlib.h>
  31 #include <errno.h>
  32 #include <arpa/inet.h>
  33
  34 #include <atalk/unicode.h>
  35 #include <atalk/logger.h>
  36 #include <atalk/unicode.h>
  37 #include <atalk/byteorder.h>
  38
  39 /* Given a trailing UTF-8 byte, get the contribution from it to
  40  * the Unicode scalar value for a particular bit shift amount
  41  */
  42 #define GETUCVAL(utf8_trailbyte,shift)  ((unsigned int) (( utf8_trailbyte & 0x3F) << shift))
  43
  44 /* Given a unicode scalar, get a trail UTF-8 byte for a particular bit shift amount */
  45 #define GETUTF8TRAILBYTE(uc,shift)      ((char)( 0x80 | ((uc >> shift) & 0x3F) ) )
  46
  47
  48
  49 static size_t   utf8_pull(void *,char **, size_t *, char **, size_t *);
  50 static size_t   utf8_push(void *,char **, size_t *, char **, size_t *);
  51
  52 struct charset_functions charset_utf8 =
  53 {
  54         "UTF8",
  55         0x08000103,
  56         utf8_pull,
  57         utf8_push,
  58         CHARSET_VOLUME | CHARSET_MULTIBYTE | CHARSET_PRECOMPOSED,
  59         NULL,
  60         NULL, NULL
  61 };
  62
  63 struct charset_functions charset_utf8_mac =
  64 {
  65         "UTF8-MAC",
  66         0x08000103,
  67         utf8_pull,
  68         utf8_push,
  69         CHARSET_VOLUME | CHARSET_CLIENT | CHARSET_MULTIBYTE | CHARSET_DECOMPOSED,
  70         NULL,
  71         NULL, NULL
  72 };
  73
  74 /* ------------------- Convert from UTF-8 to UTF-16 -------------------*/
  75 static size_t utf8_pull(void *cd _U_, char **inbuf, size_t *inbytesleft,
  76                          char **outbuf, size_t *outbytesleft)
  77 {
  78         ucs2_t uc = 0;
  79         unsigned int codepoint;
  80         int len;
  81
  82         while (*inbytesleft >= 1 && *outbytesleft >= 2) {
  83                 unsigned char *c = (unsigned char *)*inbuf;
  84                 len = 1;
  85
  86                 /* Arrange conditionals in the order of most frequent occurrence
  87                  * for users of Latin-based chars */
  88                 if ((c[0] & 0x80) == 0) {
  89                         uc = c[0];
  90                 } else if ((c[0] & 0xe0) == 0xc0) {
  91                         if (*inbytesleft < 2) {
  92                                 LOG(log_debug, logtype_default, "short utf8 char");
  93                                 goto badseq;
  94                         }
  95                         uc = (ucs2_t) (((c[0] & 0x1f) << 6) | GETUCVAL(c[1],0)) ;
  96                         len = 2;
  97                 } else if ((c[0] & 0xf0) == 0xe0) {
  98                         if (*inbytesleft < 3) {
  99                                 LOG(log_debug, logtype_default, "short utf8 char");
 100                                 goto badseq;
 101                         }
 102                         uc = (ucs2_t) (((c[0] & 0x0f) << 12) | GETUCVAL(c[1],6) | GETUCVAL(c[2],0)) ;
 103                         len = 3;
 104                 } else if ((c[0] & 0xf8) == 0xf0) {
 105                         /* 4 bytes, which happens for surrogate pairs only */
 106                         if (*inbytesleft < 4) {
 107                                 LOG(log_debug, logtype_default, "short utf8 char");
 108                                 goto badseq;
 109                         }
 110                         if (*outbytesleft < 4) {
 111                                 LOG(log_debug, logtype_default, "short ucs-2 write");
 112                                 errno = E2BIG;
 113                                 return -1;
 114                         }
 115                         codepoint = ((c[0] & 0x07) << 18) | GETUCVAL(c[1],12) |
 116                                 GETUCVAL(c[2],6) |  GETUCVAL(c[3],0);
 117                         SSVAL(*outbuf,0,(((codepoint - 0x10000) >> 10) + 0xD800)); /* hi  */
 118                         SSVAL(*outbuf,2,(0xDC00 + (codepoint & 0x03FF)));          /* low */
 119                         len = 4;
 120                         (*inbuf)  += 4;
 121                         (*inbytesleft)  -= 4;
 122                         (*outbytesleft) -= 4;
 123                         (*outbuf) += 4;
 124                         continue;
 125                 }
 126                 else {
 127                         errno = EINVAL;
 128                         return -1;
 129                 }
 130
 131                 SSVAL(*outbuf,0,uc);
 132                 (*inbuf)  += len;
 133                 (*inbytesleft)  -= len;
 134                 (*outbytesleft) -= 2;
 135                 (*outbuf) += 2;
 136         }
 137
 138         if (*inbytesleft > 0) {
 139                 errno = E2BIG;
 140                 return -1;
 141         }
 142
 143         return 0;
 144
 145 badseq:
 146         errno = EINVAL;
 147         return -1;
 148 }
 149
 150 /* --------------------- Convert from UTF-16 to UTF-8 -----------*/
 151 static size_t utf8_push(void *cd _U_, char **inbuf, size_t *inbytesleft,
 152                          char **outbuf, size_t *outbytesleft)
 153 {
 154         ucs2_t uc=0;
 155         ucs2_t hi, low;
 156         unsigned int codepoint;
 157         int olen, ilen;
 158
 159         while (*inbytesleft >= 2 && *outbytesleft >= 1) {
 160                 unsigned char *c = (unsigned char *)*outbuf;
 161                 uc = SVAL((*inbuf),0);
 162                 olen=1;
 163                 ilen=2;
 164
 165                 /* Arrange conditionals in the order of most frequent occurrence for
 166                    users of Latin-based chars */
 167                 if (uc < 0x80) {
 168                         c[0] = uc;
 169                 } else if (uc < 0x800) {
 170                         if (*outbytesleft < 2) {
 171                                 LOG(log_debug, logtype_default, "short utf8 write");
 172                                 goto toobig;
 173                         }
 174                         c[1] = GETUTF8TRAILBYTE(uc, 0);
 175                         c[0] = (char)(0xc0 | ((uc >> 6) & 0x1f));
 176                         olen = 2;
 177                 }
 178                 else if ( uc >= 0x202a && uc <= 0x202e ) {
 179                         /* ignore bidi hint characters */
 180                         olen = 0;
 181                 }
 182                 /*
 183                  * A 2-byte uc value represents a stand-alone Unicode character if
 184                  *     0 <= uc < 0xd800 or 0xdfff < uc <= 0xffff.
 185                  * If  0xd800 <= uc <= 0xdfff, uc itself does not represent a Unicode character.
 186                  * Rather, it is just part of a surrogate pair.  A surrogate pair consists of
 187                  * a high surrogate in the range [0xd800 ... 0xdbff] and a low surrogate in the
 188                  * range [0xdc00 ... 0xdfff].  Together the pair maps to a single Unicode character
 189                  * whose scalar value is 64K or larger.  It is this scalar value that is transformed
 190                  * to UTF-8, not the individual surrogates.
 191                  *
 192                  * See www.unicode.org/faq/utf_bom.html for more info.
 193                  */
 194
 195                 else if ( 0xd800 <= uc && uc <= 0xdfff) {
 196                         /* surrogate - needs 4 bytes from input and 4 bytes for output to UTF-8 */
 197                         if (*outbytesleft < 4) {
 198                                 LOG(log_debug, logtype_default, "short utf8 write");
 199                                 goto toobig;
 200                         }
 201                         if (*inbytesleft < 4) {
 202                                 errno = EINVAL;
 203                                 return -1;
 204                         }
 205                         hi =  SVAL((*inbuf),0);
 206                         low = SVAL((*inbuf),2);
 207                         if ( 0xd800 <= hi && hi <= 0xdbff && 0xdc00 <= low && low <= 0xdfff) {
 208                                 codepoint = ((hi - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
 209                                 c[3] = GETUTF8TRAILBYTE(codepoint, 0);
 210                                 c[2] = GETUTF8TRAILBYTE(codepoint, 6);
 211                                 c[1] = GETUTF8TRAILBYTE(codepoint, 12);
 212                                 c[0] = (char)(0xf0 | ((codepoint >> 18) & 0x07));
 213                                 ilen = olen = 4;
 214                         } else { /* invalid values for surrogate */
 215                                 errno = EINVAL;
 216                                 return -1;
 217                         }
 218                 } else {
 219                         if (*outbytesleft < 3) {
 220                                 LOG(log_debug, logtype_default, "short utf8 write");
 221                                 goto toobig;
 222                         }
 223                         c[2] = GETUTF8TRAILBYTE(uc, 0);
 224                         c[1] = GETUTF8TRAILBYTE(uc, 6);
 225                         c[0] = (char)(0xe0 | ((uc >> 12) & 0x0f));
 226                         olen = 3;
 227                 }
 228
 229                 (*inbytesleft)  -= ilen;
 230                 (*outbytesleft) -= olen;
 231                 (*inbuf)  += ilen;
 232                 (*outbuf) += olen;
 233         }
 234
 235         if (*inbytesleft == 1) {
 236                 errno = EINVAL;
 237                 return -1;
 238         }
 239
 240         if (*inbytesleft > 1) {
 241                 errno = E2BIG;
 242                 return -1;
 243         }
 244
 245         return 0;
 246
 247 toobig:
 248         errno = E2BIG;
 249         return -1;
 250 }