1 #define DEBG(x) 2 #define DEBG1(x) 3 /* inflate.c -- Not copyrighted 1992 by Mark Adler 4 version c10p1, 10 January 1993 */ 5 6 /* 7 * Adapted for booting Linux by Hannu Savolainen 1993 8 * based on gzip-1.0.3 9 * 10 * Nicolas Pitre <nico@cam.org>, 1999/04/14 : 11 * Little mods for all variable to reside either into rodata or bss segments 12 * by marking constant variables with 'const' and initializing all the others 13 * at run-time only. This allows for the kernel uncompressor to run 14 * directly from Flash or ROM memory on embedded systems. 15 */ 16 17 /* 18 Inflate deflated (PKZIP's method 8 compressed) data. The compression 19 method searches for as much of the current string of bytes (up to a 20 length of 258) in the previous 32 K bytes. If it doesn't find any 21 matches (of at least length 3), it codes the next byte. Otherwise, it 22 codes the length of the matched string and its distance backwards from 23 the current position. There is a single Huffman code that codes both 24 single bytes (called "literals") and match lengths. A second Huffman 25 code codes the distance information, which follows a length code. Each 26 length or distance code actually represents a base value and a number 27 of "extra" (sometimes zero) bits to get to add to the base value. At 28 the end of each deflated block is a special end-of-block (EOB) literal/ 29 length code. The decoding process is basically: get a literal/length 30 code; if EOB then done; if a literal, emit the decoded byte; if a 31 length then get the distance and emit the referred-to bytes from the 32 sliding window of previously emitted data. 33 34 There are (currently) three kinds of inflate blocks: stored, fixed, and 35 dynamic. The compressor deals with some chunk of data at a time, and 36 decides which method to use on a chunk-by-chunk basis. A chunk might 37 typically be 32 K or 64 K. If the chunk is incompressible, then the 38 "stored" method is used. In this case, the bytes are simply stored as 39 is, eight bits per byte, with none of the above coding. The bytes are 40 preceded by a count, since there is no longer an EOB code. 41 42 If the data is compressible, then either the fixed or dynamic methods 43 are used. In the dynamic method, the compressed data is preceded by 44 an encoding of the literal/length and distance Huffman codes that are 45 to be used to decode this block. The representation is itself Huffman 46 coded, and so is preceded by a description of that code. These code 47 descriptions take up a little space, and so for small blocks, there is 48 a predefined set of codes, called the fixed codes. The fixed method is 49 used if the block codes up smaller that way (usually for quite small 50 chunks), otherwise the dynamic method is used. In the latter case, the 51 codes are customized to the probabilities in the current block, and so 52 can code it much better than the pre-determined fixed codes. 53 54 The Huffman codes themselves are decoded using a multi-level table 55 lookup, in order to maximize the speed of decoding plus the speed of 56 building the decoding tables. See the comments below that precede the 57 lbits and dbits tuning parameters. 58 */ 59 60 61 /* 62 Notes beyond the 1.93a appnote.txt: 63 64 1. Distance pointers never point before the beginning of the output 65 stream. 66 2. Distance pointers can point back across blocks, up to 32k away. 67 3. There is an implied maximum of 7 bits for the bit length table and 68 15 bits for the actual data. 69 4. If only one code exists, then it is encoded using one bit. (Zero 70 would be more efficient, but perhaps a little confusing.) If two 71 codes exist, they are coded using one bit each (0 and 1). 72 5. There is no way of sending zero distance codes--a dummy must be 73 sent if there are none. (History: a pre 2.0 version of PKZIP would 74 store blocks with no distance codes, but this was discovered to be 75 too harsh a criterion.) Valid only for 1.93a. 2.04c does allow 76 zero distance codes, which is sent as one code of zero bits in 77 length. 78 6. There are up to 286 literal/length codes. Code 256 represents the 79 end-of-block. Note however that the static length tree defines 80 288 codes just to fill out the Huffman codes. Codes 286 and 287 81 cannot be used though, since there is no length base or extra bits 82 defined for them. Similarly, there are up to 30 distance codes. 83 However, static trees define 32 codes (all 5 bits) to fill out the 84 Huffman codes, but the last two had better not show up in the data. 85 7. Unzip can check dynamic Huffman blocks for complete code sets. 86 The exception is that a single code would not be complete (see #4). 87 8. The five bits following the block type is really the number of 88 literal codes sent minus 257. 89 9. Length codes 8,16,16 are interpreted as 13 length codes of 8 bits 90 (1+6+6). Therefore, to output three times the length, you output 91 three codes (1+1+1), whereas to output four times the same length, 92 you only need two codes (1+3). Hmm. 93 10. In the tree reconstruction algorithm, Code = Code + Increment 94 only if BitLength(i) is not zero. (Pretty obvious.) 95 11. Correction: 4 Bits: # of Bit Length codes - 4 (4 - 19) 96 12. Note: length code 284 can represent 227-258, but length code 285 97 really is 258. The last length deserves its own, short code 98 since it gets used a lot in very redundant files. The length 99 258 is special since 258 - 3 (the min match length) is 255. 100 13. The literal/length and distance code bit lengths are read as a 101 single stream of lengths. It is possible (and advantageous) for 102 a repeat code (16, 17, or 18) to go across the boundary between 103 the two sets of lengths. 104 */ 105 #include <linux/compiler.h> 106 107 #ifdef RCSID 108 static char rcsid[] = "#Id: inflate.c,v 0.14 1993/06/10 13:27:04 jloup Exp #"; 109 #endif 110 111 #ifndef STATIC 112 113 #if defined(STDC_HEADERS) || defined(HAVE_STDLIB_H) 114 # include <sys/types.h> 115 # include <stdlib.h> 116 #endif 117 118 #include "gzip.h" 119 #define STATIC 120 #endif /* !STATIC */ 121 122 #ifndef INIT 123 #define INIT 124 #endif 125 126 #define slide window 127 128 /* Huffman code lookup table entry--this entry is four bytes for machines 129 that have 16-bit pointers (e.g. PC's in the small or medium model). 130 Valid extra bits are 0..13. e == 15 is EOB (end of block), e == 16 131 means that v is a literal, 16 < e < 32 means that v is a pointer to 132 the next table, which codes e - 16 bits, and lastly e == 99 indicates 133 an unused code. If a code with e == 99 is looked up, this implies an 134 error in the data. */ 135 struct huft { 136 uch e; /* number of extra bits or operation */ 137 uch b; /* number of bits in this code or subcode */ 138 union { 139 ush n; /* literal, length base, or distance base */ 140 struct huft *t; /* pointer to next level of table */ 141 } v; 142 }; 143 144 145 /* Function prototypes */ 146 STATIC int INIT huft_build OF((unsigned *, unsigned, unsigned, 147 const ush *, const ush *, struct huft **, int *)); 148 STATIC int INIT huft_free OF((struct huft *)); 149 STATIC int INIT inflate_codes OF((struct huft *, struct huft *, int, int)); 150 STATIC int INIT inflate_stored OF((void)); 151 STATIC int INIT inflate_fixed OF((void)); 152 STATIC int INIT inflate_dynamic OF((void)); 153 STATIC int INIT inflate_block OF((int *)); 154 STATIC int INIT inflate OF((void)); 155 156 157 /* The inflate algorithm uses a sliding 32 K byte window on the uncompressed 158 stream to find repeated byte strings. This is implemented here as a 159 circular buffer. The index is updated simply by incrementing and then 160 ANDing with 0x7fff (32K-1). */ 161 /* It is left to other modules to supply the 32 K area. It is assumed 162 to be usable as if it were declared "uch slide[32768];" or as just 163 "uch *slide;" and then malloc'ed in the latter case. The definition 164 must be in unzip.h, included above. */ 165 /* unsigned wp; current position in slide */ 166 #define wp outcnt 167 #define flush_output(w) (wp=(w),flush_window()) 168 169 /* Tables for deflate from PKZIP's appnote.txt. */ 170 static const unsigned border[] = { /* Order of the bit length code lengths */ 171 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}; 172 static const ush cplens[] = { /* Copy lengths for literal codes 257..285 */ 173 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, 174 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0}; 175 /* note: see note #13 above about the 258 in this list. */ 176 static const ush cplext[] = { /* Extra bits for literal codes 257..285 */ 177 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 178 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */ 179 static const ush cpdist[] = { /* Copy offsets for distance codes 0..29 */ 180 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 181 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145, 182 8193, 12289, 16385, 24577}; 183 static const ush cpdext[] = { /* Extra bits for distance codes */ 184 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 185 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 186 12, 12, 13, 13}; 187 188 189 190 /* Macros for inflate() bit peeking and grabbing. 191 The usage is: 192 193 NEEDBITS(j) 194 x = b & mask_bits[j]; 195 DUMPBITS(j) 196 197 where NEEDBITS makes sure that b has at least j bits in it, and 198 DUMPBITS removes the bits from b. The macros use the variable k 199 for the number of bits in b. Normally, b and k are register 200 variables for speed, and are initialized at the beginning of a 201 routine that uses these macros from a global bit buffer and count. 202 203 If we assume that EOB will be the longest code, then we will never 204 ask for bits with NEEDBITS that are beyond the end of the stream. 205 So, NEEDBITS should not read any more bytes than are needed to 206 meet the request. Then no bytes need to be "returned" to the buffer 207 at the end of the last block. 208 209 However, this assumption is not true for fixed blocks--the EOB code 210 is 7 bits, but the other literal/length codes can be 8 or 9 bits. 211 (The EOB code is shorter than other codes because fixed blocks are 212 generally short. So, while a block always has an EOB, many other 213 literal/length codes have a significantly lower probability of 214 showing up at all.) However, by making the first table have a 215 lookup of seven bits, the EOB code will be found in that first 216 lookup, and so will not require that too many bits be pulled from 217 the stream. 218 */ 219 220 STATIC ulg bb; /* bit buffer */ 221 STATIC unsigned bk; /* bits in bit buffer */ 222 223 STATIC const ush mask_bits[] = { 224 0x0000, 225 0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff, 226 0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff 227 }; 228 229 #define NEXTBYTE() ({ int v = get_byte(); if (v < 0) goto underrun; (uch)v; }) 230 #define NEEDBITS(n) {while(k<(n)){b|=((ulg)NEXTBYTE())<<k;k+=8;}} 231 #define DUMPBITS(n) {b>>=(n);k-=(n);} 232 233 234 /* 235 Huffman code decoding is performed using a multi-level table lookup. 236 The fastest way to decode is to simply build a lookup table whose 237 size is determined by the longest code. However, the time it takes 238 to build this table can also be a factor if the data being decoded 239 is not very long. The most common codes are necessarily the 240 shortest codes, so those codes dominate the decoding time, and hence 241 the speed. The idea is you can have a shorter table that decodes the 242 shorter, more probable codes, and then point to subsidiary tables for 243 the longer codes. The time it costs to decode the longer codes is 244 then traded against the time it takes to make longer tables. 245 246 This results of this trade are in the variables lbits and dbits 247 below. lbits is the number of bits the first level table for literal/ 248 length codes can decode in one step, and dbits is the same thing for 249 the distance codes. Subsequent tables are also less than or equal to 250 those sizes. These values may be adjusted either when all of the 251 codes are shorter than that, in which case the longest code length in 252 bits is used, or when the shortest code is *longer* than the requested 253 table size, in which case the length of the shortest code in bits is 254 used. 255 256 There are two different values for the two tables, since they code a 257 different number of possibilities each. The literal/length table 258 codes 286 possible values, or in a flat code, a little over eight 259 bits. The distance table codes 30 possible values, or a little less 260 than five bits, flat. The optimum values for speed end up being 261 about one bit more than those, so lbits is 8+1 and dbits is 5+1. 262 The optimum values may differ though from machine to machine, and 263 possibly even between compilers. Your mileage may vary. 264 */ 265 266 267 STATIC const int lbits = 9; /* bits in base literal/length lookup table */ 268 STATIC const int dbits = 6; /* bits in base distance lookup table */ 269 270 271 /* If BMAX needs to be larger than 16, then h and x[] should be ulg. */ 272 #define BMAX 16 /* maximum bit length of any code (16 for explode) */ 273 #define N_MAX 288 /* maximum number of codes in any set */ 274 275 276 STATIC unsigned hufts; /* track memory usage */ 277 278 279 STATIC int INIT huft_build( 280 unsigned *b, /* code lengths in bits (all assumed <= BMAX) */ 281 unsigned n, /* number of codes (assumed <= N_MAX) */ 282 unsigned s, /* number of simple-valued codes (0..s-1) */ 283 const ush *d, /* list of base values for non-simple codes */ 284 const ush *e, /* list of extra bits for non-simple codes */ 285 struct huft **t, /* result: starting table */ 286 int *m /* maximum lookup bits, returns actual */ 287 ) 288 /* Given a list of code lengths and a maximum table size, make a set of 289 tables to decode that set of codes. Return zero on success, one if 290 the given code set is incomplete (the tables are still built in this 291 case), two if the input is invalid (all zero length codes or an 292 oversubscribed set of lengths), and three if not enough memory. */ 293 { 294 unsigned a; /* counter for codes of length k */ 295 unsigned c[BMAX+1]; /* bit length count table */ 296 unsigned f; /* i repeats in table every f entries */ 297 int g; /* maximum code length */ 298 int h; /* table level */ 299 register unsigned i; /* counter, current code */ 300 register unsigned j; /* counter */ 301 register int k; /* number of bits in current code */ 302 int l; /* bits per table (returned in m) */ 303 register unsigned *p; /* pointer into c[], b[], or v[] */ 304 register struct huft *q; /* points to current table */ 305 struct huft r; /* table entry for structure assignment */ 306 struct huft *u[BMAX]; /* table stack */ 307 unsigned v[N_MAX]; /* values in order of bit length */ 308 register int w; /* bits before this table == (l * h) */ 309 unsigned x[BMAX+1]; /* bit offsets, then code stack */ 310 unsigned *xp; /* pointer into x */ 311 int y; /* number of dummy codes added */ 312 unsigned z; /* number of entries in current table */ 313 314 DEBG("huft1 "); 315 316 /* Generate counts for each bit length */ 317 memzero(c, sizeof(c)); 318 p = b; i = n; 319 do { 320 Tracecv(*p, (stderr, (n-i >= ' ' && n-i <= '~' ? "%c %d\n" : "0x%x %d\n"), 321 n-i, *p)); 322 c[*p]++; /* assume all entries <= BMAX */ 323 p++; /* Can't combine with above line (Solaris bug) */ 324 } while (--i); 325 if (c[0] == n) /* null input--all zero length codes */ 326 { 327 *t = (struct huft *)NULL; 328 *m = 0; 329 return 2; 330 } 331 332 DEBG("huft2 "); 333 334 /* Find minimum and maximum length, bound *m by those */ 335 l = *m; 336 for (j = 1; j <= BMAX; j++) 337 if (c[j]) 338 break; 339 k = j; /* minimum code length */ 340 if ((unsigned)l < j) 341 l = j; 342 for (i = BMAX; i; i--) 343 if (c[i]) 344 break; 345 g = i; /* maximum code length */ 346 if ((unsigned)l > i) 347 l = i; 348 *m = l; 349 350 DEBG("huft3 "); 351 352 /* Adjust last length count to fill out codes, if needed */ 353 for (y = 1 << j; j < i; j++, y <<= 1) 354 if ((y -= c[j]) < 0) 355 return 2; /* bad input: more codes than bits */ 356 if ((y -= c[i]) < 0) 357 return 2; 358 c[i] += y; 359 360 DEBG("huft4 "); 361 362 /* Generate starting offsets into the value table for each length */ 363 x[1] = j = 0; 364 p = c + 1; xp = x + 2; 365 while (--i) { /* note that i == g from above */ 366 *xp++ = (j += *p++); 367 } 368 369 DEBG("huft5 "); 370 371 /* Make a table of values in order of bit lengths */ 372 p = b; i = 0; 373 do { 374 if ((j = *p++) != 0) 375 v[x[j]++] = i; 376 } while (++i < n); 377 n = x[g]; /* set n to length of v */ 378 379 DEBG("h6 "); 380 381 /* Generate the Huffman codes and for each, make the table entries */ 382 x[0] = i = 0; /* first Huffman code is zero */ 383 p = v; /* grab values in bit order */ 384 h = -1; /* no tables yet--level -1 */ 385 w = -l; /* bits decoded == (l * h) */ 386 u[0] = (struct huft *)NULL; /* just to keep compilers happy */ 387 q = (struct huft *)NULL; /* ditto */ 388 z = 0; /* ditto */ 389 DEBG("h6a "); 390 391 /* go through the bit lengths (k already is bits in shortest code) */ 392 for (; k <= g; k++) 393 { 394 DEBG("h6b "); 395 a = c[k]; 396 while (a--) 397 { 398 DEBG("h6b1 "); 399 /* here i is the Huffman code of length k bits for value *p */ 400 /* make tables up to required level */ 401 while (k > w + l) 402 { 403 DEBG1("1 "); 404 h++; 405 w += l; /* previous table always l bits */ 406 407 /* compute minimum size table less than or equal to l bits */ 408 z = (z = g - w) > (unsigned)l ? l : z; /* upper limit on table size */ 409 if ((f = 1 << (j = k - w)) > a + 1) /* try a k-w bit table */ 410 { /* too few codes for k-w bit table */ 411 DEBG1("2 "); 412 f -= a + 1; /* deduct codes from patterns left */ 413 xp = c + k; 414 if (j < z) 415 while (++j < z) /* try smaller tables up to z bits */ 416 { 417 if ((f <<= 1) <= *++xp) 418 break; /* enough codes to use up j bits */ 419 f -= *xp; /* else deduct codes from patterns */ 420 } 421 } 422 DEBG1("3 "); 423 z = 1 << j; /* table entries for j-bit table */ 424 425 /* allocate and link in new table */ 426 if ((q = (struct huft *)malloc((z + 1)*sizeof(struct huft))) == 427 (struct huft *)NULL) 428 { 429 if (h) 430 huft_free(u[0]); 431 return 3; /* not enough memory */ 432 } 433 DEBG1("4 "); 434 hufts += z + 1; /* track memory usage */ 435 *t = q + 1; /* link to list for huft_free() */ 436 *(t = &(q->v.t)) = (struct huft *)NULL; 437 u[h] = ++q; /* table starts after link */ 438 439 DEBG1("5 "); 440 /* connect to last table, if there is one */ 441 if (h) 442 { 443 x[h] = i; /* save pattern for backing up */ 444 r.b = (uch)l; /* bits to dump before this table */ 445 r.e = (uch)(16 + j); /* bits in this table */ 446 r.v.t = q; /* pointer to this table */ 447 j = i >> (w - l); /* (get around Turbo C bug) */ 448 u[h-1][j] = r; /* connect to last table */ 449 } 450 DEBG1("6 "); 451 } 452 DEBG("h6c "); 453 454 /* set up table entry in r */ 455 r.b = (uch)(k - w); 456 if (p >= v + n) 457 r.e = 99; /* out of values--invalid code */ 458 else if (*p < s) 459 { 460 r.e = (uch)(*p < 256 ? 16 : 15); /* 256 is end-of-block code */ 461 r.v.n = (ush)(*p); /* simple code is just the value */ 462 p++; /* one compiler does not like *p++ */ 463 } 464 else 465 { 466 r.e = (uch)e[*p - s]; /* non-simple--look up in lists */ 467 r.v.n = d[*p++ - s]; 468 } 469 DEBG("h6d "); 470 471 /* fill code-like entries with r */ 472 f = 1 << (k - w); 473 for (j = i >> w; j < z; j += f) 474 q[j] = r; 475 476 /* backwards increment the k-bit code i */ 477 for (j = 1 << (k - 1); i & j; j >>= 1) 478 i ^= j; 479 i ^= j; 480 481 /* backup over finished tables */ 482 while ((i & ((1 << w) - 1)) != x[h]) 483 { 484 h--; /* don't need to update q */ 485 w -= l; 486 } 487 DEBG("h6e "); 488 } 489 DEBG("h6f "); 490 } 491 492 DEBG("huft7 "); 493 494 /* Return true (1) if we were given an incomplete table */ 495 return y != 0 && g != 1; 496 } 497 498 499 500 STATIC int INIT huft_free( 501 struct huft *t /* table to free */ 502 ) 503 /* Free the malloc'ed tables built by huft_build(), which makes a linked 504 list of the tables it made, with the links in a dummy first entry of 505 each table. */ 506 { 507 register struct huft *p, *q; 508 509 510 /* Go through linked list, freeing from the malloced (t[-1]) address. */ 511 p = t; 512 while (p != (struct huft *)NULL) 513 { 514 q = (--p)->v.t; 515 free((char*)p); 516 p = q; 517 } 518 return 0; 519 } 520 521 522 STATIC int INIT inflate_codes( 523 struct huft *tl, /* literal/length decoder tables */ 524 struct huft *td, /* distance decoder tables */ 525 int bl, /* number of bits decoded by tl[] */ 526 int bd /* number of bits decoded by td[] */ 527 ) 528 /* inflate (decompress) the codes in a deflated (compressed) block. 529 Return an error code or zero if it all goes ok. */ 530 { 531 register unsigned e; /* table entry flag/number of extra bits */ 532 unsigned n, d; /* length and index for copy */ 533 unsigned w; /* current window position */ 534 struct huft *t; /* pointer to table entry */ 535 unsigned ml, md; /* masks for bl and bd bits */ 536 register ulg b; /* bit buffer */ 537 register unsigned k; /* number of bits in bit buffer */ 538 539 540 /* make local copies of globals */ 541 b = bb; /* initialize bit buffer */ 542 k = bk; 543 w = wp; /* initialize window position */ 544 545 /* inflate the coded data */ 546 ml = mask_bits[bl]; /* precompute masks for speed */ 547 md = mask_bits[bd]; 548 for (;;) /* do until end of block */ 549 { 550 NEEDBITS((unsigned)bl) 551 if ((e = (t = tl + ((unsigned)b & ml))->e) > 16) 552 do { 553 if (e == 99) 554 return 1; 555 DUMPBITS(t->b) 556 e -= 16; 557 NEEDBITS(e) 558 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16); 559 DUMPBITS(t->b) 560 if (e == 16) /* then it's a literal */ 561 { 562 slide[w++] = (uch)t->v.n; 563 Tracevv((stderr, "%c", slide[w-1])); 564 if (w == WSIZE) 565 { 566 flush_output(w); 567 w = 0; 568 } 569 } 570 else /* it's an EOB or a length */ 571 { 572 /* exit if end of block */ 573 if (e == 15) 574 break; 575 576 /* get length of block to copy */ 577 NEEDBITS(e) 578 n = t->v.n + ((unsigned)b & mask_bits[e]); 579 DUMPBITS(e); 580 581 /* decode distance of block to copy */ 582 NEEDBITS((unsigned)bd) 583 if ((e = (t = td + ((unsigned)b & md))->e) > 16) 584 do { 585 if (e == 99) 586 return 1; 587 DUMPBITS(t->b) 588 e -= 16; 589 NEEDBITS(e) 590 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16); 591 DUMPBITS(t->b) 592 NEEDBITS(e) 593 d = w - t->v.n - ((unsigned)b & mask_bits[e]); 594 DUMPBITS(e) 595 Tracevv((stderr,"\\[%d,%d]", w-d, n)); 596 597 /* do the copy */ 598 do { 599 n -= (e = (e = WSIZE - ((d &= WSIZE-1) > w ? d : w)) > n ? n : e); 600 #if !defined(NOMEMCPY) && !defined(DEBUG) 601 if (w - d >= e) /* (this test assumes unsigned comparison) */ 602 { 603 memcpy(slide + w, slide + d, e); 604 w += e; 605 d += e; 606 } 607 else /* do it slow to avoid memcpy() overlap */ 608 #endif /* !NOMEMCPY */ 609 do { 610 slide[w++] = slide[d++]; 611 Tracevv((stderr, "%c", slide[w-1])); 612 } while (--e); 613 if (w == WSIZE) 614 { 615 flush_output(w); 616 w = 0; 617 } 618 } while (n); 619 } 620 } 621 622 623 /* restore the globals from the locals */ 624 wp = w; /* restore global window pointer */ 625 bb = b; /* restore global bit buffer */ 626 bk = k; 627 628 /* done */ 629 return 0; 630 631 underrun: 632 return 4; /* Input underrun */ 633 } 634 635 636 637 STATIC int INIT inflate_stored(void) 638 /* "decompress" an inflated type 0 (stored) block. */ 639 { 640 unsigned n; /* number of bytes in block */ 641 unsigned w; /* current window position */ 642 register ulg b; /* bit buffer */ 643 register unsigned k; /* number of bits in bit buffer */ 644 645 DEBG("<stor"); 646 647 /* make local copies of globals */ 648 b = bb; /* initialize bit buffer */ 649 k = bk; 650 w = wp; /* initialize window position */ 651 652 653 /* go to byte boundary */ 654 n = k & 7; 655 DUMPBITS(n); 656 657 658 /* get the length and its complement */ 659 NEEDBITS(16) 660 n = ((unsigned)b & 0xffff); 661 DUMPBITS(16) 662 NEEDBITS(16) 663 if (n != (unsigned)((~b) & 0xffff)) 664 return 1; /* error in compressed data */ 665 DUMPBITS(16) 666 667 668 /* read and output the compressed data */ 669 while (n--) 670 { 671 NEEDBITS(8) 672 slide[w++] = (uch)b; 673 if (w == WSIZE) 674 { 675 flush_output(w); 676 w = 0; 677 } 678 DUMPBITS(8) 679 } 680 681 682 /* restore the globals from the locals */ 683 wp = w; /* restore global window pointer */ 684 bb = b; /* restore global bit buffer */ 685 bk = k; 686 687 DEBG(">"); 688 return 0; 689 690 underrun: 691 return 4; /* Input underrun */ 692 } 693 694 695 /* 696 * We use `noinline' here to prevent gcc-3.5 from using too much stack space 697 */ 698 STATIC int noinline INIT inflate_fixed(void) 699 /* decompress an inflated type 1 (fixed Huffman codes) block. We should 700 either replace this with a custom decoder, or at least precompute the 701 Huffman tables. */ 702 { 703 int i; /* temporary variable */ 704 struct huft *tl; /* literal/length code table */ 705 struct huft *td; /* distance code table */ 706 int bl; /* lookup bits for tl */ 707 int bd; /* lookup bits for td */ 708 unsigned l[288]; /* length list for huft_build */ 709 710 DEBG("<fix"); 711 712 /* set up literal table */ 713 for (i = 0; i < 144; i++) 714 l[i] = 8; 715 for (; i < 256; i++) 716 l[i] = 9; 717 for (; i < 280; i++) 718 l[i] = 7; 719 for (; i < 288; i++) /* make a complete, but wrong code set */ 720 l[i] = 8; 721 bl = 7; 722 if ((i = huft_build(l, 288, 257, cplens, cplext, &tl, &bl)) != 0) 723 return i; 724 725 726 /* set up distance table */ 727 for (i = 0; i < 30; i++) /* make an incomplete code set */ 728 l[i] = 5; 729 bd = 5; 730 if ((i = huft_build(l, 30, 0, cpdist, cpdext, &td, &bd)) > 1) 731 { 732 huft_free(tl); 733 734 DEBG(">"); 735 return i; 736 } 737 738 739 /* decompress until an end-of-block code */ 740 if (inflate_codes(tl, td, bl, bd)) 741 return 1; 742 743 744 /* free the decoding tables, return */ 745 huft_free(tl); 746 huft_free(td); 747 return 0; 748 } 749 750 751 /* 752 * We use `noinline' here to prevent gcc-3.5 from using too much stack space 753 */ 754 STATIC int noinline INIT inflate_dynamic(void) 755 /* decompress an inflated type 2 (dynamic Huffman codes) block. */ 756 { 757 int i; /* temporary variables */ 758 unsigned j; 759 unsigned l; /* last length */ 760 unsigned m; /* mask for bit lengths table */ 761 unsigned n; /* number of lengths to get */ 762 struct huft *tl; /* literal/length code table */ 763 struct huft *td; /* distance code table */ 764 int bl; /* lookup bits for tl */ 765 int bd; /* lookup bits for td */ 766 unsigned nb; /* number of bit length codes */ 767 unsigned nl; /* number of literal/length codes */ 768 unsigned nd; /* number of distance codes */ 769 #ifdef PKZIP_BUG_WORKAROUND 770 unsigned ll[288+32]; /* literal/length and distance code lengths */ 771 #else 772 unsigned ll[286+30]; /* literal/length and distance code lengths */ 773 #endif 774 register ulg b; /* bit buffer */ 775 register unsigned k; /* number of bits in bit buffer */ 776 777 DEBG("<dyn"); 778 779 /* make local bit buffer */ 780 b = bb; 781 k = bk; 782 783 784 /* read in table lengths */ 785 NEEDBITS(5) 786 nl = 257 + ((unsigned)b & 0x1f); /* number of literal/length codes */ 787 DUMPBITS(5) 788 NEEDBITS(5) 789 nd = 1 + ((unsigned)b & 0x1f); /* number of distance codes */ 790 DUMPBITS(5) 791 NEEDBITS(4) 792 nb = 4 + ((unsigned)b & 0xf); /* number of bit length codes */ 793 DUMPBITS(4) 794 #ifdef PKZIP_BUG_WORKAROUND 795 if (nl > 288 || nd > 32) 796 #else 797 if (nl > 286 || nd > 30) 798 #endif 799 return 1; /* bad lengths */ 800 801 DEBG("dyn1 "); 802 803 /* read in bit-length-code lengths */ 804 for (j = 0; j < nb; j++) 805 { 806 NEEDBITS(3) 807 ll[border[j]] = (unsigned)b & 7; 808 DUMPBITS(3) 809 } 810 for (; j < 19; j++) 811 ll[border[j]] = 0; 812 813 DEBG("dyn2 "); 814 815 /* build decoding table for trees--single level, 7 bit lookup */ 816 bl = 7; 817 if ((i = huft_build(ll, 19, 19, NULL, NULL, &tl, &bl)) != 0) 818 { 819 if (i == 1) 820 huft_free(tl); 821 return i; /* incomplete code set */ 822 } 823 824 DEBG("dyn3 "); 825 826 /* read in literal and distance code lengths */ 827 n = nl + nd; 828 m = mask_bits[bl]; 829 i = l = 0; 830 while ((unsigned)i < n) 831 { 832 NEEDBITS((unsigned)bl) 833 j = (td = tl + ((unsigned)b & m))->b; 834 DUMPBITS(j) 835 j = td->v.n; 836 if (j < 16) /* length of code in bits (0..15) */ 837 ll[i++] = l = j; /* save last length in l */ 838 else if (j == 16) /* repeat last length 3 to 6 times */ 839 { 840 NEEDBITS(2) 841 j = 3 + ((unsigned)b & 3); 842 DUMPBITS(2) 843 if ((unsigned)i + j > n) 844 return 1; 845 while (j--) 846 ll[i++] = l; 847 } 848 else if (j == 17) /* 3 to 10 zero length codes */ 849 { 850 NEEDBITS(3) 851 j = 3 + ((unsigned)b & 7); 852 DUMPBITS(3) 853 if ((unsigned)i + j > n) 854 return 1; 855 while (j--) 856 ll[i++] = 0; 857 l = 0; 858 } 859 else /* j == 18: 11 to 138 zero length codes */ 860 { 861 NEEDBITS(7) 862 j = 11 + ((unsigned)b & 0x7f); 863 DUMPBITS(7) 864 if ((unsigned)i + j > n) 865 return 1; 866 while (j--) 867 ll[i++] = 0; 868 l = 0; 869 } 870 } 871 872 DEBG("dyn4 "); 873 874 /* free decoding table for trees */ 875 huft_free(tl); 876 877 DEBG("dyn5 "); 878 879 /* restore the global bit buffer */ 880 bb = b; 881 bk = k; 882 883 DEBG("dyn5a "); 884 885 /* build the decoding tables for literal/length and distance codes */ 886 bl = lbits; 887 if ((i = huft_build(ll, nl, 257, cplens, cplext, &tl, &bl)) != 0) 888 { 889 DEBG("dyn5b "); 890 if (i == 1) { 891 error("incomplete literal tree"); 892 huft_free(tl); 893 } 894 return i; /* incomplete code set */ 895 } 896 DEBG("dyn5c "); 897 bd = dbits; 898 if ((i = huft_build(ll + nl, nd, 0, cpdist, cpdext, &td, &bd)) != 0) 899 { 900 DEBG("dyn5d "); 901 if (i == 1) { 902 error("incomplete distance tree"); 903 #ifdef PKZIP_BUG_WORKAROUND 904 i = 0; 905 } 906 #else 907 huft_free(td); 908 } 909 huft_free(tl); 910 return i; /* incomplete code set */ 911 #endif 912 } 913 914 DEBG("dyn6 "); 915 916 /* decompress until an end-of-block code */ 917 if (inflate_codes(tl, td, bl, bd)) 918 return 1; 919 920 DEBG("dyn7 "); 921 922 /* free the decoding tables, return */ 923 huft_free(tl); 924 huft_free(td); 925 926 DEBG(">"); 927 return 0; 928 929 underrun: 930 return 4; /* Input underrun */ 931 } 932 933 934 935 STATIC int INIT inflate_block( 936 int *e /* last block flag */ 937 ) 938 /* decompress an inflated block */ 939 { 940 unsigned t; /* block type */ 941 register ulg b; /* bit buffer */ 942 register unsigned k; /* number of bits in bit buffer */ 943 944 DEBG("<blk"); 945 946 /* make local bit buffer */ 947 b = bb; 948 k = bk; 949 950 951 /* read in last block bit */ 952 NEEDBITS(1) 953 *e = (int)b & 1; 954 DUMPBITS(1) 955 956 957 /* read in block type */ 958 NEEDBITS(2) 959 t = (unsigned)b & 3; 960 DUMPBITS(2) 961 962 963 /* restore the global bit buffer */ 964 bb = b; 965 bk = k; 966 967 /* inflate that block type */ 968 if (t == 2) 969 return inflate_dynamic(); 970 if (t == 0) 971 return inflate_stored(); 972 if (t == 1) 973 return inflate_fixed(); 974 975 DEBG(">"); 976 977 /* bad block type */ 978 return 2; 979 980 underrun: 981 return 4; /* Input underrun */ 982 } 983 984 985 986 STATIC int INIT inflate(void) 987 /* decompress an inflated entry */ 988 { 989 int e; /* last block flag */ 990 int r; /* result code */ 991 unsigned h; /* maximum struct huft's malloc'ed */ 992 void *ptr; 993 994 /* initialize window, bit buffer */ 995 wp = 0; 996 bk = 0; 997 bb = 0; 998 999 1000 /* decompress until the last block */ 1001 h = 0; 1002 do { 1003 hufts = 0; 1004 gzip_mark(&ptr); 1005 if ((r = inflate_block(&e)) != 0) { 1006 gzip_release(&ptr); 1007 return r; 1008 } 1009 gzip_release(&ptr); 1010 if (hufts > h) 1011 h = hufts; 1012 } while (!e); 1013 1014 /* Undo too much lookahead. The next read will be byte aligned so we 1015 * can discard unused bits in the last meaningful byte. 1016 */ 1017 while (bk >= 8) { 1018 bk -= 8; 1019 inptr--; 1020 } 1021 1022 /* flush out slide */ 1023 flush_output(wp); 1024 1025 1026 /* return success */ 1027 #ifdef DEBUG 1028 fprintf(stderr, "<%u> ", h); 1029 #endif /* DEBUG */ 1030 return 0; 1031 } 1032 1033 /********************************************************************** 1034 * 1035 * The following are support routines for inflate.c 1036 * 1037 **********************************************************************/ 1038 1039 static ulg crc_32_tab[256]; 1040 static ulg crc; /* initialized in makecrc() so it'll reside in bss */ 1041 #define CRC_VALUE (crc ^ 0xffffffffUL) 1042 1043 /* 1044 * Code to compute the CRC-32 table. Borrowed from 1045 * gzip-1.0.3/makecrc.c. 1046 */ 1047 1048 static void INIT 1049 makecrc(void) 1050 { 1051 /* Not copyrighted 1990 Mark Adler */ 1052 1053 unsigned long c; /* crc shift register */ 1054 unsigned long e; /* polynomial exclusive-or pattern */ 1055 int i; /* counter for all possible eight bit values */ 1056 int k; /* byte being shifted into crc apparatus */ 1057 1058 /* terms of polynomial defining this crc (except x^32): */ 1059 static const int p[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26}; 1060 1061 /* Make exclusive-or pattern from polynomial */ 1062 e = 0; 1063 for (i = 0; i < sizeof(p)/sizeof(int); i++) 1064 e |= 1L << (31 - p[i]); 1065 1066 crc_32_tab[0] = 0; 1067 1068 for (i = 1; i < 256; i++) 1069 { 1070 c = 0; 1071 for (k = i | 256; k != 1; k >>= 1) 1072 { 1073 c = c & 1 ? (c >> 1) ^ e : c >> 1; 1074 if (k & 1) 1075 c ^= e; 1076 } 1077 crc_32_tab[i] = c; 1078 } 1079 1080 /* this is initialized here so this code could reside in ROM */ 1081 crc = (ulg)0xffffffffUL; /* shift register contents */ 1082 } 1083 1084 /* gzip flag byte */ 1085 #define ASCII_FLAG 0x01 /* bit 0 set: file probably ASCII text */ 1086 #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */ 1087 #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */ 1088 #define ORIG_NAME 0x08 /* bit 3 set: original file name present */ 1089 #define COMMENT 0x10 /* bit 4 set: file comment present */ 1090 #define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */ 1091 #define RESERVED 0xC0 /* bit 6,7: reserved */ 1092 1093 /* 1094 * Do the uncompression! 1095 */ 1096 static int INIT gunzip(void) 1097 { 1098 uch flags; 1099 unsigned char magic[2]; /* magic header */ 1100 char method; 1101 ulg orig_crc = 0; /* original crc */ 1102 ulg orig_len = 0; /* original uncompressed length */ 1103 int res; 1104 1105 magic[0] = NEXTBYTE(); 1106 magic[1] = NEXTBYTE(); 1107 method = NEXTBYTE(); 1108 1109 if (magic[0] != 037 || 1110 ((magic[1] != 0213) && (magic[1] != 0236))) { 1111 error("bad gzip magic numbers"); 1112 return -1; 1113 } 1114 1115 /* We only support method #8, DEFLATED */ 1116 if (method != 8) { 1117 error("internal error, invalid method"); 1118 return -1; 1119 } 1120 1121 flags = (uch)get_byte(); 1122 if ((flags & ENCRYPTED) != 0) { 1123 error("Input is encrypted"); 1124 return -1; 1125 } 1126 if ((flags & CONTINUATION) != 0) { 1127 error("Multi part input"); 1128 return -1; 1129 } 1130 if ((flags & RESERVED) != 0) { 1131 error("Input has invalid flags"); 1132 return -1; 1133 } 1134 NEXTBYTE(); /* Get timestamp */ 1135 NEXTBYTE(); 1136 NEXTBYTE(); 1137 NEXTBYTE(); 1138 1139 (void)NEXTBYTE(); /* Ignore extra flags for the moment */ 1140 (void)NEXTBYTE(); /* Ignore OS type for the moment */ 1141 1142 if ((flags & EXTRA_FIELD) != 0) { 1143 unsigned len = (unsigned)NEXTBYTE(); 1144 len |= ((unsigned)NEXTBYTE())<<8; 1145 while (len--) (void)NEXTBYTE(); 1146 } 1147 1148 /* Get original file name if it was truncated */ 1149 if ((flags & ORIG_NAME) != 0) { 1150 /* Discard the old name */ 1151 while (NEXTBYTE() != 0) /* null */ ; 1152 } 1153 1154 /* Discard file comment if any */ 1155 if ((flags & COMMENT) != 0) { 1156 while (NEXTBYTE() != 0) /* null */ ; 1157 } 1158 1159 /* Decompress */ 1160 if ((res = inflate())) { 1161 switch (res) { 1162 case 0: 1163 break; 1164 case 1: 1165 error("invalid compressed format (err=1)"); 1166 break; 1167 case 2: 1168 error("invalid compressed format (err=2)"); 1169 break; 1170 case 3: 1171 error("out of memory"); 1172 break; 1173 case 4: 1174 error("out of input data"); 1175 break; 1176 default: 1177 error("invalid compressed format (other)"); 1178 } 1179 return -1; 1180 } 1181 1182 /* Get the crc and original length */ 1183 /* crc32 (see algorithm.doc) 1184 * uncompressed input size modulo 2^32 1185 */ 1186 orig_crc = (ulg) NEXTBYTE(); 1187 orig_crc |= (ulg) NEXTBYTE() << 8; 1188 orig_crc |= (ulg) NEXTBYTE() << 16; 1189 orig_crc |= (ulg) NEXTBYTE() << 24; 1190 1191 orig_len = (ulg) NEXTBYTE(); 1192 orig_len |= (ulg) NEXTBYTE() << 8; 1193 orig_len |= (ulg) NEXTBYTE() << 16; 1194 orig_len |= (ulg) NEXTBYTE() << 24; 1195 1196 /* Validate decompression */ 1197 if (orig_crc != CRC_VALUE) { 1198 error("crc error"); 1199 return -1; 1200 } 1201 if (orig_len != bytes_out) { 1202 error("length error"); 1203 return -1; 1204 } 1205 return 0; 1206 1207 underrun: /* NEXTBYTE() goto's here if needed */ 1208 error("out of input data"); 1209 return -1; 1210 } 1211 1212 1213