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