1 /* $FreeBSD$ */ 2 /* $NetBSD: unpack.c,v 1.3 2017/08/04 07:27:08 mrg Exp $ */ 3 4 /*- 5 * SPDX-License-Identifier: BSD-2-Clause 6 * 7 * Copyright (c) 2009 Xin LI <delphij@FreeBSD.org> 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28 * SUCH DAMAGE. 29 */ 30 31 /* This file is #included by gzip.c */ 32 33 /* 34 * pack(1) file format: 35 * 36 * The first 7 bytes is the header: 37 * 00, 01 - Signature (US, RS), we already validated it earlier. 38 * 02..05 - Uncompressed size 39 * 06 - Level for the huffman tree (<=24) 40 * 41 * pack(1) will then store symbols (leaf) nodes count in each huffman 42 * tree levels, each level would consume 1 byte (See [1]). 43 * 44 * After the symbol count table, there is the symbol table, storing 45 * symbols represented by corresponding leaf node. EOB is not being 46 * explicitly transmitted (not necessary anyway) in the symbol table. 47 * 48 * Compressed data goes after the symbol table. 49 * 50 * NOTES 51 * 52 * [1] If we count EOB into the symbols, that would mean that we will 53 * have at most 256 symbols in the huffman tree. pack(1) rejects empty 54 * file and files that just repeats one character, which means that we 55 * will have at least 2 symbols. Therefore, pack(1) would reduce the 56 * last level symbol count by 2 which makes it a number in 57 * range [0..254], so all levels' symbol count would fit into 1 byte. 58 */ 59 60 #define PACK_HEADER_LENGTH 7 61 #define HTREE_MAXLEVEL 24 62 63 /* 64 * unpack descriptor 65 * 66 * Represent the huffman tree in a similar way that pack(1) would 67 * store in a packed file. We store all symbols in a linear table, 68 * and store pointers to each level's first symbol. In addition to 69 * that, maintain two counts for each level: inner nodes count and 70 * leaf nodes count. 71 */ 72 typedef struct { 73 int symbol_size; /* Size of the symbol table */ 74 int treelevels; /* Levels for the huffman tree */ 75 76 int *symbolsin; /* Table of leaf symbols count in each 77 * level */ 78 int *inodesin; /* Table of internal nodes count in 79 * each level */ 80 81 char *symbol; /* The symbol table */ 82 char *symbol_eob; /* Pointer to the EOB symbol */ 83 char **tree; /* Decoding huffman tree (pointers to 84 * first symbol of each tree level */ 85 86 off_t uncompressed_size; /* Uncompressed size */ 87 FILE *fpIn; /* Input stream */ 88 FILE *fpOut; /* Output stream */ 89 } unpack_descriptor_t; 90 91 /* 92 * Release resource allocated to an unpack descriptor. 93 * 94 * Caller is responsible to make sure that all of these pointers are 95 * initialized (in our case, they all point to valid memory block). 96 * We don't zero out pointers here because nobody else would ever 97 * reference the memory block without scrubbing them. 98 */ 99 static void 100 unpack_descriptor_fini(unpack_descriptor_t *unpackd) 101 { 102 103 free(unpackd->symbolsin); 104 free(unpackd->inodesin); 105 free(unpackd->symbol); 106 free(unpackd->tree); 107 108 fclose(unpackd->fpIn); 109 fclose(unpackd->fpOut); 110 } 111 112 /* 113 * Recursively fill the internal node count table 114 */ 115 static void 116 unpackd_fill_inodesin(const unpack_descriptor_t *unpackd, int level) 117 { 118 119 /* 120 * The internal nodes would be 1/2 of total internal nodes and 121 * leaf nodes in the next level. For the last level there 122 * would be no internal node by definition. 123 */ 124 if (level < unpackd->treelevels) { 125 unpackd_fill_inodesin(unpackd, level + 1); 126 unpackd->inodesin[level] = (unpackd->inodesin[level + 1] + 127 unpackd->symbolsin[level + 1]) / 2; 128 } else 129 unpackd->inodesin[level] = 0; 130 } 131 132 /* 133 * Update counter for accepted bytes 134 */ 135 static void 136 accepted_bytes(off_t *bytes_in, off_t newbytes) 137 { 138 139 if (bytes_in != NULL) 140 (*bytes_in) += newbytes; 141 } 142 143 /* 144 * Read file header and construct the tree. Also, prepare the buffered I/O 145 * for decode routine. 146 * 147 * Return value is uncompressed size. 148 */ 149 static void 150 unpack_parse_header(int in, int out, char *pre, size_t prelen, off_t *bytes_in, 151 unpack_descriptor_t *unpackd) 152 { 153 unsigned char hdr[PACK_HEADER_LENGTH]; /* buffer for header */ 154 ssize_t bytesread; /* Bytes read from the file */ 155 int i, j, thisbyte; 156 157 /* Prepend the header buffer if we already read some data */ 158 if (prelen != 0) 159 memcpy(hdr, pre, prelen); 160 161 /* Read in and fill the rest bytes of header */ 162 bytesread = read(in, hdr + prelen, PACK_HEADER_LENGTH - prelen); 163 if (bytesread < 0) 164 maybe_err("Error reading pack header"); 165 infile_newdata(bytesread); 166 167 accepted_bytes(bytes_in, PACK_HEADER_LENGTH); 168 169 /* Obtain uncompressed length (bytes 2,3,4,5) */ 170 unpackd->uncompressed_size = 0; 171 for (i = 2; i <= 5; i++) { 172 unpackd->uncompressed_size <<= 8; 173 unpackd->uncompressed_size |= hdr[i]; 174 } 175 176 /* Get the levels of the tree */ 177 unpackd->treelevels = hdr[6]; 178 if (unpackd->treelevels > HTREE_MAXLEVEL || unpackd->treelevels < 1) 179 maybe_errx("Huffman tree has insane levels"); 180 181 /* Let libc take care for buffering from now on */ 182 if ((unpackd->fpIn = fdopen(in, "r")) == NULL) 183 maybe_err("Can not fdopen() input stream"); 184 if ((unpackd->fpOut = fdopen(out, "w")) == NULL) 185 maybe_err("Can not fdopen() output stream"); 186 187 /* Allocate for the tables of bounds and the tree itself */ 188 unpackd->inodesin = 189 calloc(unpackd->treelevels, sizeof(*(unpackd->inodesin))); 190 unpackd->symbolsin = 191 calloc(unpackd->treelevels, sizeof(*(unpackd->symbolsin))); 192 unpackd->tree = 193 calloc(unpackd->treelevels, (sizeof(*(unpackd->tree)))); 194 if (unpackd->inodesin == NULL || unpackd->symbolsin == NULL || 195 unpackd->tree == NULL) 196 maybe_err("calloc"); 197 198 /* We count from 0 so adjust to match array upper bound */ 199 unpackd->treelevels--; 200 201 /* Read the levels symbol count table and calculate total */ 202 unpackd->symbol_size = 1; /* EOB */ 203 for (i = 0; i <= unpackd->treelevels; i++) { 204 if ((thisbyte = fgetc(unpackd->fpIn)) == EOF) 205 maybe_err("File appears to be truncated"); 206 unpackd->symbolsin[i] = (unsigned char)thisbyte; 207 unpackd->symbol_size += unpackd->symbolsin[i]; 208 } 209 accepted_bytes(bytes_in, unpackd->treelevels); 210 if (unpackd->symbol_size > 256) 211 maybe_errx("Bad symbol table"); 212 infile_newdata(unpackd->treelevels); 213 214 /* Allocate for the symbol table, point symbol_eob at the beginning */ 215 unpackd->symbol_eob = unpackd->symbol = calloc(1, unpackd->symbol_size); 216 if (unpackd->symbol == NULL) 217 maybe_err("calloc"); 218 219 /* 220 * Read in the symbol table, which contain [2, 256] symbols. 221 * In order to fit the count in one byte, pack(1) would offset 222 * it by reducing 2 from the actual number from the last level. 223 * 224 * We adjust the last level's symbol count by 1 here, because 225 * the EOB symbol is not being transmitted explicitly. Another 226 * adjustment would be done later afterward. 227 */ 228 unpackd->symbolsin[unpackd->treelevels]++; 229 for (i = 0; i <= unpackd->treelevels; i++) { 230 unpackd->tree[i] = unpackd->symbol_eob; 231 for (j = 0; j < unpackd->symbolsin[i]; j++) { 232 if ((thisbyte = fgetc(unpackd->fpIn)) == EOF) 233 maybe_errx("Symbol table truncated"); 234 *unpackd->symbol_eob++ = (char)thisbyte; 235 } 236 infile_newdata(unpackd->symbolsin[i]); 237 accepted_bytes(bytes_in, unpackd->symbolsin[i]); 238 } 239 240 /* Now, take account for the EOB symbol as well */ 241 unpackd->symbolsin[unpackd->treelevels]++; 242 243 /* 244 * The symbolsin table has been constructed now. 245 * Calculate the internal nodes count table based on it. 246 */ 247 unpackd_fill_inodesin(unpackd, 0); 248 } 249 250 /* 251 * Decode huffman stream, based on the huffman tree. 252 */ 253 static void 254 unpack_decode(const unpack_descriptor_t *unpackd, off_t *bytes_in) 255 { 256 int thislevel, thiscode, thisbyte, inlevelindex; 257 int i; 258 off_t bytes_out = 0; 259 const char *thissymbol; /* The symbol pointer decoded from stream */ 260 261 /* 262 * Decode huffman. Fetch every bytes from the file, get it 263 * into 'thiscode' bit-by-bit, then output the symbol we got 264 * when one has been found. 265 * 266 * Assumption: sizeof(int) > ((max tree levels + 1) / 8). 267 * bad things could happen if not. 268 */ 269 thislevel = 0; 270 thiscode = thisbyte = 0; 271 272 while ((thisbyte = fgetc(unpackd->fpIn)) != EOF) { 273 accepted_bytes(bytes_in, 1); 274 infile_newdata(1); 275 check_siginfo(); 276 277 /* 278 * Split one bit from thisbyte, from highest to lowest, 279 * feed the bit into thiscode, until we got a symbol from 280 * the tree. 281 */ 282 for (i = 7; i >= 0; i--) { 283 thiscode = (thiscode << 1) | ((thisbyte >> i) & 1); 284 285 /* Did we got a symbol? (referencing leaf node) */ 286 if (thiscode >= unpackd->inodesin[thislevel]) { 287 inlevelindex = 288 thiscode - unpackd->inodesin[thislevel]; 289 if (inlevelindex > unpackd->symbolsin[thislevel]) 290 maybe_errx("File corrupt"); 291 292 thissymbol = 293 &(unpackd->tree[thislevel][inlevelindex]); 294 if ((thissymbol == unpackd->symbol_eob) && 295 (bytes_out == unpackd->uncompressed_size)) 296 goto finished; 297 298 fputc((*thissymbol), unpackd->fpOut); 299 bytes_out++; 300 301 /* Prepare for next input */ 302 thislevel = 0; thiscode = 0; 303 } else { 304 thislevel++; 305 if (thislevel > unpackd->treelevels) 306 maybe_errx("File corrupt"); 307 } 308 } 309 } 310 311 finished: 312 if (bytes_out != unpackd->uncompressed_size) 313 maybe_errx("Premature EOF"); 314 } 315 316 /* Handler for pack(1)'ed file */ 317 static off_t 318 unpack(int in, int out, char *pre, size_t prelen, off_t *bytes_in) 319 { 320 unpack_descriptor_t unpackd; 321 322 in = dup(in); 323 if (in == -1) 324 maybe_err("dup"); 325 out = dup(out); 326 if (out == -1) 327 maybe_err("dup"); 328 329 unpack_parse_header(in, out, pre, prelen, bytes_in, &unpackd); 330 unpack_decode(&unpackd, bytes_in); 331 unpack_descriptor_fini(&unpackd); 332 333 /* If we reached here, the unpack was successful */ 334 return (unpackd.uncompressed_size); 335 } 336