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 0; 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); 377 378 DEBG("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 */ 388 DEBG("h6a "); 389 390 /* go through the bit lengths (k already is bits in shortest code) */ 391 for (; k <= g; k++) 392 { 393 DEBG("h6b "); 394 a = c[k]; 395 while (a--) 396 { 397 DEBG("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 { 402 DEBG1("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 */ 410 DEBG1("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 } 420 DEBG1("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 } 431 DEBG1("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 */ 436 437 DEBG1("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 } 448 DEBG1("6 "); 449 } 450 DEBG("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 } 467 DEBG("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 } 485 DEBG("h6e "); 486 } 487 DEBG("h6f "); 488 } 489 490 DEBG("huft7 "); 491 492 /* Return true (1) if we were given an incomplete table */ 493 return y != 0 && g != 1; 494 } 495 496 497 498 STATIC int INIT huft_free( 499 struct huft *t /* table to free */ 500 ) 501 /* Free the malloc'ed tables built by huft_build(), which makes a linked 502 list of the tables it made, with the links in a dummy first entry of 503 each table. */ 504 { 505 register struct huft *p, *q; 506 507 508 /* Go through linked list, freeing from the malloced (t[-1]) address. */ 509 p = t; 510 while (p != (struct huft *)NULL) 511 { 512 q = (--p)->v.t; 513 free((char*)p); 514 p = q; 515 } 516 return 0; 517 } 518 519 520 STATIC int INIT inflate_codes( 521 struct huft *tl, /* literal/length decoder tables */ 522 struct huft *td, /* distance decoder tables */ 523 int bl, /* number of bits decoded by tl[] */ 524 int bd /* number of bits decoded by td[] */ 525 ) 526 /* inflate (decompress) the codes in a deflated (compressed) block. 527 Return an error code or zero if it all goes ok. */ 528 { 529 register unsigned e; /* table entry flag/number of extra bits */ 530 unsigned n, d; /* length and index for copy */ 531 unsigned w; /* current window position */ 532 struct huft *t; /* pointer to table entry */ 533 unsigned ml, md; /* masks for bl and bd bits */ 534 register ulg b; /* bit buffer */ 535 register unsigned k; /* number of bits in bit buffer */ 536 537 538 /* make local copies of globals */ 539 b = bb; /* initialize bit buffer */ 540 k = bk; 541 w = wp; /* initialize window position */ 542 543 /* inflate the coded data */ 544 ml = mask_bits[bl]; /* precompute masks for speed */ 545 md = mask_bits[bd]; 546 for (;;) /* do until end of block */ 547 { 548 NEEDBITS((unsigned)bl) 549 if ((e = (t = tl + ((unsigned)b & ml))->e) > 16) 550 do { 551 if (e == 99) 552 return 1; 553 DUMPBITS(t->b) 554 e -= 16; 555 NEEDBITS(e) 556 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16); 557 DUMPBITS(t->b) 558 if (e == 16) /* then it's a literal */ 559 { 560 slide[w++] = (uch)t->v.n; 561 Tracevv((stderr, "%c", slide[w-1])); 562 if (w == WSIZE) 563 { 564 flush_output(w); 565 w = 0; 566 } 567 } 568 else /* it's an EOB or a length */ 569 { 570 /* exit if end of block */ 571 if (e == 15) 572 break; 573 574 /* get length of block to copy */ 575 NEEDBITS(e) 576 n = t->v.n + ((unsigned)b & mask_bits[e]); 577 DUMPBITS(e); 578 579 /* decode distance of block to copy */ 580 NEEDBITS((unsigned)bd) 581 if ((e = (t = td + ((unsigned)b & md))->e) > 16) 582 do { 583 if (e == 99) 584 return 1; 585 DUMPBITS(t->b) 586 e -= 16; 587 NEEDBITS(e) 588 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16); 589 DUMPBITS(t->b) 590 NEEDBITS(e) 591 d = w - t->v.n - ((unsigned)b & mask_bits[e]); 592 DUMPBITS(e) 593 Tracevv((stderr,"\\[%d,%d]", w-d, n)); 594 595 /* do the copy */ 596 do { 597 n -= (e = (e = WSIZE - ((d &= WSIZE-1) > w ? d : w)) > n ? n : e); 598 #if !defined(NOMEMCPY) && !defined(DEBUG) 599 if (w - d >= e) /* (this test assumes unsigned comparison) */ 600 { 601 memcpy(slide + w, slide + d, e); 602 w += e; 603 d += e; 604 } 605 else /* do it slow to avoid memcpy() overlap */ 606 #endif /* !NOMEMCPY */ 607 do { 608 slide[w++] = slide[d++]; 609 Tracevv((stderr, "%c", slide[w-1])); 610 } while (--e); 611 if (w == WSIZE) 612 { 613 flush_output(w); 614 w = 0; 615 } 616 } while (n); 617 } 618 } 619 620 621 /* restore the globals from the locals */ 622 wp = w; /* restore global window pointer */ 623 bb = b; /* restore global bit buffer */ 624 bk = k; 625 626 /* done */ 627 return 0; 628 629 underrun: 630 return 4; /* Input underrun */ 631 } 632 633 634 635 STATIC int INIT inflate_stored(void) 636 /* "decompress" an inflated type 0 (stored) block. */ 637 { 638 unsigned n; /* number of bytes in block */ 639 unsigned w; /* current window position */ 640 register ulg b; /* bit buffer */ 641 register unsigned k; /* number of bits in bit buffer */ 642 643 DEBG("<stor"); 644 645 /* make local copies of globals */ 646 b = bb; /* initialize bit buffer */ 647 k = bk; 648 w = wp; /* initialize window position */ 649 650 651 /* go to byte boundary */ 652 n = k & 7; 653 DUMPBITS(n); 654 655 656 /* get the length and its complement */ 657 NEEDBITS(16) 658 n = ((unsigned)b & 0xffff); 659 DUMPBITS(16) 660 NEEDBITS(16) 661 if (n != (unsigned)((~b) & 0xffff)) 662 return 1; /* error in compressed data */ 663 DUMPBITS(16) 664 665 666 /* read and output the compressed data */ 667 while (n--) 668 { 669 NEEDBITS(8) 670 slide[w++] = (uch)b; 671 if (w == WSIZE) 672 { 673 flush_output(w); 674 w = 0; 675 } 676 DUMPBITS(8) 677 } 678 679 680 /* restore the globals from the locals */ 681 wp = w; /* restore global window pointer */ 682 bb = b; /* restore global bit buffer */ 683 bk = k; 684 685 DEBG(">"); 686 return 0; 687 688 underrun: 689 return 4; /* Input underrun */ 690 } 691 692 693 /* 694 * We use `noinline' here to prevent gcc-3.5 from using too much stack space 695 */ 696 STATIC int noinline INIT inflate_fixed(void) 697 /* decompress an inflated type 1 (fixed Huffman codes) block. We should 698 either replace this with a custom decoder, or at least precompute the 699 Huffman tables. */ 700 { 701 int i; /* temporary variable */ 702 struct huft *tl; /* literal/length code table */ 703 struct huft *td; /* distance code table */ 704 int bl; /* lookup bits for tl */ 705 int bd; /* lookup bits for td */ 706 unsigned l[288]; /* length list for huft_build */ 707 708 DEBG("<fix"); 709 710 /* set up literal table */ 711 for (i = 0; i < 144; i++) 712 l[i] = 8; 713 for (; i < 256; i++) 714 l[i] = 9; 715 for (; i < 280; i++) 716 l[i] = 7; 717 for (; i < 288; i++) /* make a complete, but wrong code set */ 718 l[i] = 8; 719 bl = 7; 720 if ((i = huft_build(l, 288, 257, cplens, cplext, &tl, &bl)) != 0) 721 return i; 722 723 724 /* set up distance table */ 725 for (i = 0; i < 30; i++) /* make an incomplete code set */ 726 l[i] = 5; 727 bd = 5; 728 if ((i = huft_build(l, 30, 0, cpdist, cpdext, &td, &bd)) > 1) 729 { 730 huft_free(tl); 731 732 DEBG(">"); 733 return i; 734 } 735 736 737 /* decompress until an end-of-block code */ 738 if (inflate_codes(tl, td, bl, bd)) 739 return 1; 740 741 742 /* free the decoding tables, return */ 743 huft_free(tl); 744 huft_free(td); 745 return 0; 746 } 747 748 749 /* 750 * We use `noinline' here to prevent gcc-3.5 from using too much stack space 751 */ 752 STATIC int noinline INIT inflate_dynamic(void) 753 /* decompress an inflated type 2 (dynamic Huffman codes) block. */ 754 { 755 int i; /* temporary variables */ 756 unsigned j; 757 unsigned l; /* last length */ 758 unsigned m; /* mask for bit lengths table */ 759 unsigned n; /* number of lengths to get */ 760 struct huft *tl; /* literal/length code table */ 761 struct huft *td; /* distance code table */ 762 int bl; /* lookup bits for tl */ 763 int bd; /* lookup bits for td */ 764 unsigned nb; /* number of bit length codes */ 765 unsigned nl; /* number of literal/length codes */ 766 unsigned nd; /* number of distance codes */ 767 #ifdef PKZIP_BUG_WORKAROUND 768 unsigned ll[288+32]; /* literal/length and distance code lengths */ 769 #else 770 unsigned ll[286+30]; /* literal/length and distance code lengths */ 771 #endif 772 register ulg b; /* bit buffer */ 773 register unsigned k; /* number of bits in bit buffer */ 774 775 DEBG("<dyn"); 776 777 /* make local bit buffer */ 778 b = bb; 779 k = bk; 780 781 782 /* read in table lengths */ 783 NEEDBITS(5) 784 nl = 257 + ((unsigned)b & 0x1f); /* number of literal/length codes */ 785 DUMPBITS(5) 786 NEEDBITS(5) 787 nd = 1 + ((unsigned)b & 0x1f); /* number of distance codes */ 788 DUMPBITS(5) 789 NEEDBITS(4) 790 nb = 4 + ((unsigned)b & 0xf); /* number of bit length codes */ 791 DUMPBITS(4) 792 #ifdef PKZIP_BUG_WORKAROUND 793 if (nl > 288 || nd > 32) 794 #else 795 if (nl > 286 || nd > 30) 796 #endif 797 return 1; /* bad lengths */ 798 799 DEBG("dyn1 "); 800 801 /* read in bit-length-code lengths */ 802 for (j = 0; j < nb; j++) 803 { 804 NEEDBITS(3) 805 ll[border[j]] = (unsigned)b & 7; 806 DUMPBITS(3) 807 } 808 for (; j < 19; j++) 809 ll[border[j]] = 0; 810 811 DEBG("dyn2 "); 812 813 /* build decoding table for trees--single level, 7 bit lookup */ 814 bl = 7; 815 if ((i = huft_build(ll, 19, 19, NULL, NULL, &tl, &bl)) != 0) 816 { 817 if (i == 1) 818 huft_free(tl); 819 return i; /* incomplete code set */ 820 } 821 822 DEBG("dyn3 "); 823 824 /* read in literal and distance code lengths */ 825 n = nl + nd; 826 m = mask_bits[bl]; 827 i = l = 0; 828 while ((unsigned)i < n) 829 { 830 NEEDBITS((unsigned)bl) 831 j = (td = tl + ((unsigned)b & m))->b; 832 DUMPBITS(j) 833 j = td->v.n; 834 if (j < 16) /* length of code in bits (0..15) */ 835 ll[i++] = l = j; /* save last length in l */ 836 else if (j == 16) /* repeat last length 3 to 6 times */ 837 { 838 NEEDBITS(2) 839 j = 3 + ((unsigned)b & 3); 840 DUMPBITS(2) 841 if ((unsigned)i + j > n) 842 return 1; 843 while (j--) 844 ll[i++] = l; 845 } 846 else if (j == 17) /* 3 to 10 zero length codes */ 847 { 848 NEEDBITS(3) 849 j = 3 + ((unsigned)b & 7); 850 DUMPBITS(3) 851 if ((unsigned)i + j > n) 852 return 1; 853 while (j--) 854 ll[i++] = 0; 855 l = 0; 856 } 857 else /* j == 18: 11 to 138 zero length codes */ 858 { 859 NEEDBITS(7) 860 j = 11 + ((unsigned)b & 0x7f); 861 DUMPBITS(7) 862 if ((unsigned)i + j > n) 863 return 1; 864 while (j--) 865 ll[i++] = 0; 866 l = 0; 867 } 868 } 869 870 DEBG("dyn4 "); 871 872 /* free decoding table for trees */ 873 huft_free(tl); 874 875 DEBG("dyn5 "); 876 877 /* restore the global bit buffer */ 878 bb = b; 879 bk = k; 880 881 DEBG("dyn5a "); 882 883 /* build the decoding tables for literal/length and distance codes */ 884 bl = lbits; 885 if ((i = huft_build(ll, nl, 257, cplens, cplext, &tl, &bl)) != 0) 886 { 887 DEBG("dyn5b "); 888 if (i == 1) { 889 error("incomplete literal tree"); 890 huft_free(tl); 891 } 892 return i; /* incomplete code set */ 893 } 894 DEBG("dyn5c "); 895 bd = dbits; 896 if ((i = huft_build(ll + nl, nd, 0, cpdist, cpdext, &td, &bd)) != 0) 897 { 898 DEBG("dyn5d "); 899 if (i == 1) { 900 error("incomplete distance tree"); 901 #ifdef PKZIP_BUG_WORKAROUND 902 i = 0; 903 } 904 #else 905 huft_free(td); 906 } 907 huft_free(tl); 908 return i; /* incomplete code set */ 909 #endif 910 } 911 912 DEBG("dyn6 "); 913 914 /* decompress until an end-of-block code */ 915 if (inflate_codes(tl, td, bl, bd)) 916 return 1; 917 918 DEBG("dyn7 "); 919 920 /* free the decoding tables, return */ 921 huft_free(tl); 922 huft_free(td); 923 924 DEBG(">"); 925 return 0; 926 927 underrun: 928 return 4; /* Input underrun */ 929 } 930 931 932 933 STATIC int INIT inflate_block( 934 int *e /* last block flag */ 935 ) 936 /* decompress an inflated block */ 937 { 938 unsigned t; /* block type */ 939 register ulg b; /* bit buffer */ 940 register unsigned k; /* number of bits in bit buffer */ 941 942 DEBG("<blk"); 943 944 /* make local bit buffer */ 945 b = bb; 946 k = bk; 947 948 949 /* read in last block bit */ 950 NEEDBITS(1) 951 *e = (int)b & 1; 952 DUMPBITS(1) 953 954 955 /* read in block type */ 956 NEEDBITS(2) 957 t = (unsigned)b & 3; 958 DUMPBITS(2) 959 960 961 /* restore the global bit buffer */ 962 bb = b; 963 bk = k; 964 965 /* inflate that block type */ 966 if (t == 2) 967 return inflate_dynamic(); 968 if (t == 0) 969 return inflate_stored(); 970 if (t == 1) 971 return inflate_fixed(); 972 973 DEBG(">"); 974 975 /* bad block type */ 976 return 2; 977 978 underrun: 979 return 4; /* Input underrun */ 980 } 981 982 983 984 STATIC int INIT inflate(void) 985 /* decompress an inflated entry */ 986 { 987 int e; /* last block flag */ 988 int r; /* result code */ 989 unsigned h; /* maximum struct huft's malloc'ed */ 990 void *ptr; 991 992 /* initialize window, bit buffer */ 993 wp = 0; 994 bk = 0; 995 bb = 0; 996 997 998 /* decompress until the last block */ 999 h = 0; 1000 do { 1001 hufts = 0; 1002 gzip_mark(&ptr); 1003 if ((r = inflate_block(&e)) != 0) { 1004 gzip_release(&ptr); 1005 return r; 1006 } 1007 gzip_release(&ptr); 1008 if (hufts > h) 1009 h = hufts; 1010 } while (!e); 1011 1012 /* Undo too much lookahead. The next read will be byte aligned so we 1013 * can discard unused bits in the last meaningful byte. 1014 */ 1015 while (bk >= 8) { 1016 bk -= 8; 1017 inptr--; 1018 } 1019 1020 /* flush out slide */ 1021 flush_output(wp); 1022 1023 1024 /* return success */ 1025 #ifdef DEBUG 1026 fprintf(stderr, "<%u> ", h); 1027 #endif /* DEBUG */ 1028 return 0; 1029 } 1030 1031 /********************************************************************** 1032 * 1033 * The following are support routines for inflate.c 1034 * 1035 **********************************************************************/ 1036 1037 static ulg crc_32_tab[256]; 1038 static ulg crc; /* initialized in makecrc() so it'll reside in bss */ 1039 #define CRC_VALUE (crc ^ 0xffffffffUL) 1040 1041 /* 1042 * Code to compute the CRC-32 table. Borrowed from 1043 * gzip-1.0.3/makecrc.c. 1044 */ 1045 1046 static void INIT 1047 makecrc(void) 1048 { 1049 /* Not copyrighted 1990 Mark Adler */ 1050 1051 unsigned long c; /* crc shift register */ 1052 unsigned long e; /* polynomial exclusive-or pattern */ 1053 int i; /* counter for all possible eight bit values */ 1054 int k; /* byte being shifted into crc apparatus */ 1055 1056 /* terms of polynomial defining this crc (except x^32): */ 1057 static const int p[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26}; 1058 1059 /* Make exclusive-or pattern from polynomial */ 1060 e = 0; 1061 for (i = 0; i < sizeof(p)/sizeof(int); i++) 1062 e |= 1L << (31 - p[i]); 1063 1064 crc_32_tab[0] = 0; 1065 1066 for (i = 1; i < 256; i++) 1067 { 1068 c = 0; 1069 for (k = i | 256; k != 1; k >>= 1) 1070 { 1071 c = c & 1 ? (c >> 1) ^ e : c >> 1; 1072 if (k & 1) 1073 c ^= e; 1074 } 1075 crc_32_tab[i] = c; 1076 } 1077 1078 /* this is initialized here so this code could reside in ROM */ 1079 crc = (ulg)0xffffffffUL; /* shift register contents */ 1080 } 1081 1082 /* gzip flag byte */ 1083 #define ASCII_FLAG 0x01 /* bit 0 set: file probably ASCII text */ 1084 #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */ 1085 #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */ 1086 #define ORIG_NAME 0x08 /* bit 3 set: original file name present */ 1087 #define COMMENT 0x10 /* bit 4 set: file comment present */ 1088 #define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */ 1089 #define RESERVED 0xC0 /* bit 6,7: reserved */ 1090 1091 /* 1092 * Do the uncompression! 1093 */ 1094 static int INIT gunzip(void) 1095 { 1096 uch flags; 1097 unsigned char magic[2]; /* magic header */ 1098 char method; 1099 ulg orig_crc = 0; /* original crc */ 1100 ulg orig_len = 0; /* original uncompressed length */ 1101 int res; 1102 1103 magic[0] = NEXTBYTE(); 1104 magic[1] = NEXTBYTE(); 1105 method = NEXTBYTE(); 1106 1107 if (magic[0] != 037 || 1108 ((magic[1] != 0213) && (magic[1] != 0236))) { 1109 error("bad gzip magic numbers"); 1110 return -1; 1111 } 1112 1113 /* We only support method #8, DEFLATED */ 1114 if (method != 8) { 1115 error("internal error, invalid method"); 1116 return -1; 1117 } 1118 1119 flags = (uch)get_byte(); 1120 if ((flags & ENCRYPTED) != 0) { 1121 error("Input is encrypted"); 1122 return -1; 1123 } 1124 if ((flags & CONTINUATION) != 0) { 1125 error("Multi part input"); 1126 return -1; 1127 } 1128 if ((flags & RESERVED) != 0) { 1129 error("Input has invalid flags"); 1130 return -1; 1131 } 1132 NEXTBYTE(); /* Get timestamp */ 1133 NEXTBYTE(); 1134 NEXTBYTE(); 1135 NEXTBYTE(); 1136 1137 (void)NEXTBYTE(); /* Ignore extra flags for the moment */ 1138 (void)NEXTBYTE(); /* Ignore OS type for the moment */ 1139 1140 if ((flags & EXTRA_FIELD) != 0) { 1141 unsigned len = (unsigned)NEXTBYTE(); 1142 len |= ((unsigned)NEXTBYTE())<<8; 1143 while (len--) (void)NEXTBYTE(); 1144 } 1145 1146 /* Get original file name if it was truncated */ 1147 if ((flags & ORIG_NAME) != 0) { 1148 /* Discard the old name */ 1149 while (NEXTBYTE() != 0) /* null */ ; 1150 } 1151 1152 /* Discard file comment if any */ 1153 if ((flags & COMMENT) != 0) { 1154 while (NEXTBYTE() != 0) /* null */ ; 1155 } 1156 1157 /* Decompress */ 1158 if ((res = inflate())) { 1159 switch (res) { 1160 case 0: 1161 break; 1162 case 1: 1163 error("invalid compressed format (err=1)"); 1164 break; 1165 case 2: 1166 error("invalid compressed format (err=2)"); 1167 break; 1168 case 3: 1169 error("out of memory"); 1170 break; 1171 case 4: 1172 error("out of input data"); 1173 break; 1174 default: 1175 error("invalid compressed format (other)"); 1176 } 1177 return -1; 1178 } 1179 1180 /* Get the crc and original length */ 1181 /* crc32 (see algorithm.doc) 1182 * uncompressed input size modulo 2^32 1183 */ 1184 orig_crc = (ulg) NEXTBYTE(); 1185 orig_crc |= (ulg) NEXTBYTE() << 8; 1186 orig_crc |= (ulg) NEXTBYTE() << 16; 1187 orig_crc |= (ulg) NEXTBYTE() << 24; 1188 1189 orig_len = (ulg) NEXTBYTE(); 1190 orig_len |= (ulg) NEXTBYTE() << 8; 1191 orig_len |= (ulg) NEXTBYTE() << 16; 1192 orig_len |= (ulg) NEXTBYTE() << 24; 1193 1194 /* Validate decompression */ 1195 if (orig_crc != CRC_VALUE) { 1196 error("crc error"); 1197 return -1; 1198 } 1199 if (orig_len != bytes_out) { 1200 error("length error"); 1201 return -1; 1202 } 1203 return 0; 1204 1205 underrun: /* NEXTBYTE() goto's here if needed */ 1206 error("out of input data"); 1207 return -1; 1208 } 1209 1210 1211