inflate.c revision 50477
1/* 2 * Most parts of this file are not covered by: 3 * ---------------------------------------------------------------------------- 4 * "THE BEER-WARE LICENSE" (Revision 42): 5 * <phk@login.dknet.dk> wrote this file. As long as you retain this notice you 6 * can do whatever you want with this stuff. If we meet some day, and you think 7 * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp 8 * ---------------------------------------------------------------------------- 9 * 10 * $FreeBSD: head/sys/kern/inflate.c 50477 1999-08-28 01:08:13Z peter $ 11 * 12 * 13 */ 14 15#include <sys/param.h> 16#include <sys/inflate.h> 17#ifdef KERNEL 18#include <sys/systm.h> 19#include <sys/kernel.h> 20#endif 21#include <sys/malloc.h> 22 23#ifdef KERNEL 24static MALLOC_DEFINE(M_GZIP, "Gzip trees", "Gzip trees"); 25#endif 26 27/* needed to make inflate() work */ 28#define uch u_char 29#define ush u_short 30#define ulg u_long 31 32/* Stuff to make inflate() work */ 33#ifdef KERNEL 34#define memzero(dest,len) bzero(dest,len) 35#endif 36#define NOMEMCPY 37#ifdef KERNEL 38#define FPRINTF printf 39#else 40extern void putstr (char *); 41#define FPRINTF putstr 42#endif 43 44#define FLUSH(x,y) { \ 45 int foo = (*x->gz_output)(x->gz_private,x->gz_slide,y); \ 46 if (foo) \ 47 return foo; \ 48 } 49 50static const int qflag = 0; 51 52#ifndef KERNEL /* want to use this file in kzip also */ 53extern unsigned char *kzipmalloc (int); 54extern void kzipfree (void*); 55#define malloc(x, y, z) kzipmalloc((x)) 56#define free(x, y) kzipfree((x)) 57#endif 58 59/* 60 * This came from unzip-5.12. I have changed it the flow to pass 61 * a structure pointer around, thus hopefully making it re-entrant. 62 * Poul-Henning 63 */ 64 65/* inflate.c -- put in the public domain by Mark Adler 66 version c14o, 23 August 1994 */ 67 68/* You can do whatever you like with this source file, though I would 69 prefer that if you modify it and redistribute it that you include 70 comments to that effect with your name and the date. Thank you. 71 72 History: 73 vers date who what 74 ---- --------- -------------- ------------------------------------ 75 a ~~ Feb 92 M. Adler used full (large, one-step) lookup table 76 b1 21 Mar 92 M. Adler first version with partial lookup tables 77 b2 21 Mar 92 M. Adler fixed bug in fixed-code blocks 78 b3 22 Mar 92 M. Adler sped up match copies, cleaned up some 79 b4 25 Mar 92 M. Adler added prototypes; removed window[] (now 80 is the responsibility of unzip.h--also 81 changed name to slide[]), so needs diffs 82 for unzip.c and unzip.h (this allows 83 compiling in the small model on MSDOS); 84 fixed cast of q in huft_build(); 85 b5 26 Mar 92 M. Adler got rid of unintended macro recursion. 86 b6 27 Mar 92 M. Adler got rid of nextbyte() routine. fixed 87 bug in inflate_fixed(). 88 c1 30 Mar 92 M. Adler removed lbits, dbits environment variables. 89 changed BMAX to 16 for explode. Removed 90 OUTB usage, and replaced it with flush()-- 91 this was a 20% speed improvement! Added 92 an explode.c (to replace unimplod.c) that 93 uses the huft routines here. Removed 94 register union. 95 c2 4 Apr 92 M. Adler fixed bug for file sizes a multiple of 32k. 96 c3 10 Apr 92 M. Adler reduced memory of code tables made by 97 huft_build significantly (factor of two to 98 three). 99 c4 15 Apr 92 M. Adler added NOMEMCPY do kill use of memcpy(). 100 worked around a Turbo C optimization bug. 101 c5 21 Apr 92 M. Adler added the GZ_WSIZE #define to allow reducing 102 the 32K window size for specialized 103 applications. 104 c6 31 May 92 M. Adler added some typecasts to eliminate warnings 105 c7 27 Jun 92 G. Roelofs added some more typecasts (444: MSC bug). 106 c8 5 Oct 92 J-l. Gailly added ifdef'd code to deal with PKZIP bug. 107 c9 9 Oct 92 M. Adler removed a memory error message (~line 416). 108 c10 17 Oct 92 G. Roelofs changed ULONG/UWORD/byte to ulg/ush/uch, 109 removed old inflate, renamed inflate_entry 110 to inflate, added Mark's fix to a comment. 111 c10.5 14 Dec 92 M. Adler fix up error messages for incomplete trees. 112 c11 2 Jan 93 M. Adler fixed bug in detection of incomplete 113 tables, and removed assumption that EOB is 114 the longest code (bad assumption). 115 c12 3 Jan 93 M. Adler make tables for fixed blocks only once. 116 c13 5 Jan 93 M. Adler allow all zero length codes (pkzip 2.04c 117 outputs one zero length code for an empty 118 distance tree). 119 c14 12 Mar 93 M. Adler made inflate.c standalone with the 120 introduction of inflate.h. 121 c14b 16 Jul 93 G. Roelofs added (unsigned) typecast to w at 470. 122 c14c 19 Jul 93 J. Bush changed v[N_MAX], l[288], ll[28x+3x] arrays 123 to static for Amiga. 124 c14d 13 Aug 93 J-l. Gailly de-complicatified Mark's c[*p++]++ thing. 125 c14e 8 Oct 93 G. Roelofs changed memset() to memzero(). 126 c14f 22 Oct 93 G. Roelofs renamed quietflg to qflag; made Trace() 127 conditional; added inflate_free(). 128 c14g 28 Oct 93 G. Roelofs changed l/(lx+1) macro to pointer (Cray bug) 129 c14h 7 Dec 93 C. Ghisler huft_build() optimizations. 130 c14i 9 Jan 94 A. Verheijen set fixed_t{d,l} to NULL after freeing; 131 G. Roelofs check NEXTBYTE macro for GZ_EOF. 132 c14j 23 Jan 94 G. Roelofs removed Ghisler "optimizations"; ifdef'd 133 GZ_EOF check. 134 c14k 27 Feb 94 G. Roelofs added some typecasts to avoid warnings. 135 c14l 9 Apr 94 G. Roelofs fixed split comments on preprocessor lines 136 to avoid bug in Encore compiler. 137 c14m 7 Jul 94 P. Kienitz modified to allow assembler version of 138 inflate_codes() (define ASM_INFLATECODES) 139 c14n 22 Jul 94 G. Roelofs changed fprintf to FPRINTF for DLL versions 140 c14o 23 Aug 94 C. Spieler added a newline to a debug statement; 141 G. Roelofs added another typecast to avoid MSC warning 142 */ 143 144 145/* 146 Inflate deflated (PKZIP's method 8 compressed) data. The compression 147 method searches for as much of the current string of bytes (up to a 148 length of 258) in the previous 32K bytes. If it doesn't find any 149 matches (of at least length 3), it codes the next byte. Otherwise, it 150 codes the length of the matched string and its distance backwards from 151 the current position. There is a single Huffman code that codes both 152 single bytes (called "literals") and match lengths. A second Huffman 153 code codes the distance information, which follows a length code. Each 154 length or distance code actually represents a base value and a number 155 of "extra" (sometimes zero) bits to get to add to the base value. At 156 the end of each deflated block is a special end-of-block (EOB) literal/ 157 length code. The decoding process is basically: get a literal/length 158 code; if EOB then done; if a literal, emit the decoded byte; if a 159 length then get the distance and emit the referred-to bytes from the 160 sliding window of previously emitted data. 161 162 There are (currently) three kinds of inflate blocks: stored, fixed, and 163 dynamic. The compressor outputs a chunk of data at a time and decides 164 which method to use on a chunk-by-chunk basis. A chunk might typically 165 be 32K to 64K, uncompressed. If the chunk is uncompressible, then the 166 "stored" method is used. In this case, the bytes are simply stored as 167 is, eight bits per byte, with none of the above coding. The bytes are 168 preceded by a count, since there is no longer an EOB code. 169 170 If the data is compressible, then either the fixed or dynamic methods 171 are used. In the dynamic method, the compressed data is preceded by 172 an encoding of the literal/length and distance Huffman codes that are 173 to be used to decode this block. The representation is itself Huffman 174 coded, and so is preceded by a description of that code. These code 175 descriptions take up a little space, and so for small blocks, there is 176 a predefined set of codes, called the fixed codes. The fixed method is 177 used if the block ends up smaller that way (usually for quite small 178 chunks); otherwise the dynamic method is used. In the latter case, the 179 codes are customized to the probabilities in the current block and so 180 can code it much better than the pre-determined fixed codes can. 181 182 The Huffman codes themselves are decoded using a mutli-level table 183 lookup, in order to maximize the speed of decoding plus the speed of 184 building the decoding tables. See the comments below that precede the 185 lbits and dbits tuning parameters. 186 */ 187 188 189/* 190 Notes beyond the 1.93a appnote.txt: 191 192 1. Distance pointers never point before the beginning of the output 193 stream. 194 2. Distance pointers can point back across blocks, up to 32k away. 195 3. There is an implied maximum of 7 bits for the bit length table and 196 15 bits for the actual data. 197 4. If only one code exists, then it is encoded using one bit. (Zero 198 would be more efficient, but perhaps a little confusing.) If two 199 codes exist, they are coded using one bit each (0 and 1). 200 5. There is no way of sending zero distance codes--a dummy must be 201 sent if there are none. (History: a pre 2.0 version of PKZIP would 202 store blocks with no distance codes, but this was discovered to be 203 too harsh a criterion.) Valid only for 1.93a. 2.04c does allow 204 zero distance codes, which is sent as one code of zero bits in 205 length. 206 6. There are up to 286 literal/length codes. Code 256 represents the 207 end-of-block. Note however that the static length tree defines 208 288 codes just to fill out the Huffman codes. Codes 286 and 287 209 cannot be used though, since there is no length base or extra bits 210 defined for them. Similarily, there are up to 30 distance codes. 211 However, static trees define 32 codes (all 5 bits) to fill out the 212 Huffman codes, but the last two had better not show up in the data. 213 7. Unzip can check dynamic Huffman blocks for complete code sets. 214 The exception is that a single code would not be complete (see #4). 215 8. The five bits following the block type is really the number of 216 literal codes sent minus 257. 217 9. Length codes 8,16,16 are interpreted as 13 length codes of 8 bits 218 (1+6+6). Therefore, to output three times the length, you output 219 three codes (1+1+1), whereas to output four times the same length, 220 you only need two codes (1+3). Hmm. 221 10. In the tree reconstruction algorithm, Code = Code + Increment 222 only if BitLength(i) is not zero. (Pretty obvious.) 223 11. Correction: 4 Bits: # of Bit Length codes - 4 (4 - 19) 224 12. Note: length code 284 can represent 227-258, but length code 285 225 really is 258. The last length deserves its own, short code 226 since it gets used a lot in very redundant files. The length 227 258 is special since 258 - 3 (the min match length) is 255. 228 13. The literal/length and distance code bit lengths are read as a 229 single stream of lengths. It is possible (and advantageous) for 230 a repeat code (16, 17, or 18) to go across the boundary between 231 the two sets of lengths. 232 */ 233 234 235#define PKZIP_BUG_WORKAROUND /* PKZIP 1.93a problem--live with it */ 236 237/* 238 inflate.h must supply the uch slide[GZ_WSIZE] array and the NEXTBYTE, 239 FLUSH() and memzero macros. If the window size is not 32K, it 240 should also define GZ_WSIZE. If INFMOD is defined, it can include 241 compiled functions to support the NEXTBYTE and/or FLUSH() macros. 242 There are defaults for NEXTBYTE and FLUSH() below for use as 243 examples of what those functions need to do. Normally, you would 244 also want FLUSH() to compute a crc on the data. inflate.h also 245 needs to provide these typedefs: 246 247 typedef unsigned char uch; 248 typedef unsigned short ush; 249 typedef unsigned long ulg; 250 251 This module uses the external functions malloc() and free() (and 252 probably memset() or bzero() in the memzero() macro). Their 253 prototypes are normally found in <string.h> and <stdlib.h>. 254 */ 255#define INFMOD /* tell inflate.h to include code to be 256 * compiled */ 257 258/* Huffman code lookup table entry--this entry is four bytes for machines 259 that have 16-bit pointers (e.g. PC's in the small or medium model). 260 Valid extra bits are 0..13. e == 15 is EOB (end of block), e == 16 261 means that v is a literal, 16 < e < 32 means that v is a pointer to 262 the next table, which codes e - 16 bits, and lastly e == 99 indicates 263 an unused code. If a code with e == 99 is looked up, this implies an 264 error in the data. */ 265struct huft { 266 uch e; /* number of extra bits or operation */ 267 uch b; /* number of bits in this code or subcode */ 268 union { 269 ush n; /* literal, length base, or distance 270 * base */ 271 struct huft *t; /* pointer to next level of table */ 272 } v; 273}; 274 275 276/* Function prototypes */ 277static int huft_build __P((struct inflate *, unsigned *, unsigned, unsigned, const ush *, const ush *, struct huft **, int *)); 278static int huft_free __P((struct inflate *, struct huft *)); 279static int inflate_codes __P((struct inflate *, struct huft *, struct huft *, int, int)); 280static int inflate_stored __P((struct inflate *)); 281static int xinflate __P((struct inflate *)); 282static int inflate_fixed __P((struct inflate *)); 283static int inflate_dynamic __P((struct inflate *)); 284static int inflate_block __P((struct inflate *, int *)); 285 286/* The inflate algorithm uses a sliding 32K byte window on the uncompressed 287 stream to find repeated byte strings. This is implemented here as a 288 circular buffer. The index is updated simply by incrementing and then 289 and'ing with 0x7fff (32K-1). */ 290/* It is left to other modules to supply the 32K area. It is assumed 291 to be usable as if it were declared "uch slide[32768];" or as just 292 "uch *slide;" and then malloc'ed in the latter case. The definition 293 must be in unzip.h, included above. */ 294 295 296/* Tables for deflate from PKZIP's appnote.txt. */ 297 298/* Order of the bit length code lengths */ 299static const unsigned border[] = { 300 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}; 301 302static const ush cplens[] = { /* Copy lengths for literal codes 257..285 */ 303 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, 304 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0}; 305 /* note: see note #13 above about the 258 in this list. */ 306 307static const ush cplext[] = { /* Extra bits for literal codes 257..285 */ 308 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 309 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */ 310 311static const ush cpdist[] = { /* Copy offsets for distance codes 0..29 */ 312 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 313 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145, 314 8193, 12289, 16385, 24577}; 315 316static const ush cpdext[] = { /* Extra bits for distance codes */ 317 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 318 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 319 12, 12, 13, 13}; 320 321/* And'ing with mask[n] masks the lower n bits */ 322static const ush mask[] = { 323 0x0000, 324 0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff, 325 0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff 326}; 327 328 329/* Macros for inflate() bit peeking and grabbing. 330 The usage is: 331 332 NEEDBITS(glbl,j) 333 x = b & mask[j]; 334 DUMPBITS(j) 335 336 where NEEDBITS makes sure that b has at least j bits in it, and 337 DUMPBITS removes the bits from b. The macros use the variable k 338 for the number of bits in b. Normally, b and k are register 339 variables for speed, and are initialized at the begining of a 340 routine that uses these macros from a global bit buffer and count. 341 342 In order to not ask for more bits than there are in the compressed 343 stream, the Huffman tables are constructed to only ask for just 344 enough bits to make up the end-of-block code (value 256). Then no 345 bytes need to be "returned" to the buffer at the end of the last 346 block. See the huft_build() routine. 347 */ 348 349/* 350 * The following 2 were global variables. 351 * They are now fields of the inflate structure. 352 */ 353 354#define NEEDBITS(glbl,n) { \ 355 while(k<(n)) { \ 356 int c=(*glbl->gz_input)(glbl->gz_private); \ 357 if(c==GZ_EOF) \ 358 return 1; \ 359 b|=((ulg)c)<<k; \ 360 k+=8; \ 361 } \ 362 } 363 364#define DUMPBITS(n) {b>>=(n);k-=(n);} 365 366/* 367 Huffman code decoding is performed using a multi-level table lookup. 368 The fastest way to decode is to simply build a lookup table whose 369 size is determined by the longest code. However, the time it takes 370 to build this table can also be a factor if the data being decoded 371 is not very long. The most common codes are necessarily the 372 shortest codes, so those codes dominate the decoding time, and hence 373 the speed. The idea is you can have a shorter table that decodes the 374 shorter, more probable codes, and then point to subsidiary tables for 375 the longer codes. The time it costs to decode the longer codes is 376 then traded against the time it takes to make longer tables. 377 378 This results of this trade are in the variables lbits and dbits 379 below. lbits is the number of bits the first level table for literal/ 380 length codes can decode in one step, and dbits is the same thing for 381 the distance codes. Subsequent tables are also less than or equal to 382 those sizes. These values may be adjusted either when all of the 383 codes are shorter than that, in which case the longest code length in 384 bits is used, or when the shortest code is *longer* than the requested 385 table size, in which case the length of the shortest code in bits is 386 used. 387 388 There are two different values for the two tables, since they code a 389 different number of possibilities each. The literal/length table 390 codes 286 possible values, or in a flat code, a little over eight 391 bits. The distance table codes 30 possible values, or a little less 392 than five bits, flat. The optimum values for speed end up being 393 about one bit more than those, so lbits is 8+1 and dbits is 5+1. 394 The optimum values may differ though from machine to machine, and 395 possibly even between compilers. Your mileage may vary. 396 */ 397 398static const int lbits = 9; /* bits in base literal/length lookup table */ 399static const int dbits = 6; /* bits in base distance lookup table */ 400 401 402/* If BMAX needs to be larger than 16, then h and x[] should be ulg. */ 403#define BMAX 16 /* maximum bit length of any code (16 for 404 * explode) */ 405#define N_MAX 288 /* maximum number of codes in any set */ 406 407/* Given a list of code lengths and a maximum table size, make a set of 408 tables to decode that set of codes. Return zero on success, one if 409 the given code set is incomplete (the tables are still built in this 410 case), two if the input is invalid (all zero length codes or an 411 oversubscribed set of lengths), and three if not enough memory. 412 The code with value 256 is special, and the tables are constructed 413 so that no bits beyond that code are fetched when that code is 414 decoded. */ 415static int 416huft_build(glbl, b, n, s, d, e, t, m) 417 struct inflate *glbl; 418 unsigned *b; /* code lengths in bits (all assumed <= BMAX) */ 419 unsigned n; /* number of codes (assumed <= N_MAX) */ 420 unsigned s; /* number of simple-valued codes (0..s-1) */ 421 const ush *d; /* list of base values for non-simple codes */ 422 const ush *e; /* list of extra bits for non-simple codes */ 423 struct huft **t; /* result: starting table */ 424 int *m; /* maximum lookup bits, returns actual */ 425{ 426 unsigned a; /* counter for codes of length k */ 427 unsigned c[BMAX + 1]; /* bit length count table */ 428 unsigned el; /* length of EOB code (value 256) */ 429 unsigned f; /* i repeats in table every f entries */ 430 int g; /* maximum code length */ 431 int h; /* table level */ 432 register unsigned i; /* counter, current code */ 433 register unsigned j; /* counter */ 434 register int k; /* number of bits in current code */ 435 int lx[BMAX + 1]; /* memory for l[-1..BMAX-1] */ 436 int *l = lx + 1; /* stack of bits per table */ 437 register unsigned *p; /* pointer into c[], b[], or v[] */ 438 register struct huft *q;/* points to current table */ 439 struct huft r; /* table entry for structure assignment */ 440 struct huft *u[BMAX];/* table stack */ 441 unsigned v[N_MAX]; /* values in order of bit length */ 442 register int w; /* bits before this table == (l * h) */ 443 unsigned x[BMAX + 1]; /* bit offsets, then code stack */ 444 unsigned *xp; /* pointer into x */ 445 int y; /* number of dummy codes added */ 446 unsigned z; /* number of entries in current table */ 447 448 /* Generate counts for each bit length */ 449 el = n > 256 ? b[256] : BMAX; /* set length of EOB code, if any */ 450#ifdef KERNEL 451 memzero((char *) c, sizeof(c)); 452#else 453 for (i = 0; i < BMAX+1; i++) 454 c [i] = 0; 455#endif 456 p = b; 457 i = n; 458 do { 459 c[*p]++; 460 p++; /* assume all entries <= BMAX */ 461 } while (--i); 462 if (c[0] == n) { /* null input--all zero length codes */ 463 *t = (struct huft *) NULL; 464 *m = 0; 465 return 0; 466 } 467 /* Find minimum and maximum length, bound *m by those */ 468 for (j = 1; j <= BMAX; j++) 469 if (c[j]) 470 break; 471 k = j; /* minimum code length */ 472 if ((unsigned) *m < j) 473 *m = j; 474 for (i = BMAX; i; i--) 475 if (c[i]) 476 break; 477 g = i; /* maximum code length */ 478 if ((unsigned) *m > i) 479 *m = i; 480 481 /* Adjust last length count to fill out codes, if needed */ 482 for (y = 1 << j; j < i; j++, y <<= 1) 483 if ((y -= c[j]) < 0) 484 return 2; /* bad input: more codes than bits */ 485 if ((y -= c[i]) < 0) 486 return 2; 487 c[i] += y; 488 489 /* Generate starting offsets into the value table for each length */ 490 x[1] = j = 0; 491 p = c + 1; 492 xp = x + 2; 493 while (--i) { /* note that i == g from above */ 494 *xp++ = (j += *p++); 495 } 496 497 /* Make a table of values in order of bit lengths */ 498 p = b; 499 i = 0; 500 do { 501 if ((j = *p++) != 0) 502 v[x[j]++] = i; 503 } while (++i < n); 504 505 /* Generate the Huffman codes and for each, make the table entries */ 506 x[0] = i = 0; /* first Huffman code is zero */ 507 p = v; /* grab values in bit order */ 508 h = -1; /* no tables yet--level -1 */ 509 w = l[-1] = 0; /* no bits decoded yet */ 510 u[0] = (struct huft *) NULL; /* just to keep compilers happy */ 511 q = (struct huft *) NULL; /* ditto */ 512 z = 0; /* ditto */ 513 514 /* go through the bit lengths (k already is bits in shortest code) */ 515 for (; k <= g; k++) { 516 a = c[k]; 517 while (a--) { 518 /* 519 * here i is the Huffman code of length k bits for 520 * value *p 521 */ 522 /* make tables up to required level */ 523 while (k > w + l[h]) { 524 w += l[h++]; /* add bits already decoded */ 525 526 /* 527 * compute minimum size table less than or 528 * equal to *m bits 529 */ 530 z = (z = g - w) > (unsigned) *m ? *m : z; /* upper limit */ 531 if ((f = 1 << (j = k - w)) > a + 1) { /* try a k-w bit table *//* t 532 * oo few codes for k-w 533 * bit table */ 534 f -= a + 1; /* deduct codes from 535 * patterns left */ 536 xp = c + k; 537 while (++j < z) { /* try smaller tables up 538 * to z bits */ 539 if ((f <<= 1) <= *++xp) 540 break; /* enough codes to use 541 * up j bits */ 542 f -= *xp; /* else deduct codes 543 * from patterns */ 544 } 545 } 546 if ((unsigned) w + j > el && (unsigned) w < el) 547 j = el - w; /* make EOB code end at 548 * table */ 549 z = 1 << j; /* table entries for j-bit 550 * table */ 551 l[h] = j; /* set table size in stack */ 552 553 /* allocate and link in new table */ 554 if ((q = (struct huft *) malloc((z + 1) * sizeof(struct huft), M_GZIP, M_WAITOK)) == 555 (struct huft *) NULL) { 556 if (h) 557 huft_free(glbl, u[0]); 558 return 3; /* not enough memory */ 559 } 560 glbl->gz_hufts += z + 1; /* track memory usage */ 561 *t = q + 1; /* link to list for 562 * huft_free() */ 563 *(t = &(q->v.t)) = (struct huft *) NULL; 564 u[h] = ++q; /* table starts after link */ 565 566 /* connect to last table, if there is one */ 567 if (h) { 568 x[h] = i; /* save pattern for 569 * backing up */ 570 r.b = (uch) l[h - 1]; /* bits to dump before 571 * this table */ 572 r.e = (uch) (16 + j); /* bits in this table */ 573 r.v.t = q; /* pointer to this table */ 574 j = (i & ((1 << w) - 1)) >> (w - l[h - 1]); 575 u[h - 1][j] = r; /* connect to last table */ 576 } 577 } 578 579 /* set up table entry in r */ 580 r.b = (uch) (k - w); 581 if (p >= v + n) 582 r.e = 99; /* out of values--invalid 583 * code */ 584 else if (*p < s) { 585 r.e = (uch) (*p < 256 ? 16 : 15); /* 256 is end-of-block 586 * code */ 587 r.v.n = *p++; /* simple code is just the 588 * value */ 589 } else { 590 r.e = (uch) e[*p - s]; /* non-simple--look up 591 * in lists */ 592 r.v.n = d[*p++ - s]; 593 } 594 595 /* fill code-like entries with r */ 596 f = 1 << (k - w); 597 for (j = i >> w; j < z; j += f) 598 q[j] = r; 599 600 /* backwards increment the k-bit code i */ 601 for (j = 1 << (k - 1); i & j; j >>= 1) 602 i ^= j; 603 i ^= j; 604 605 /* backup over finished tables */ 606 while ((i & ((1 << w) - 1)) != x[h]) 607 w -= l[--h]; /* don't need to update q */ 608 } 609 } 610 611 /* return actual size of base table */ 612 *m = l[0]; 613 614 /* Return true (1) if we were given an incomplete table */ 615 return y != 0 && g != 1; 616} 617 618static int 619huft_free(glbl, t) 620 struct inflate *glbl; 621 struct huft *t; /* table to free */ 622/* Free the malloc'ed tables built by huft_build(), which makes a linked 623 list of the tables it made, with the links in a dummy first entry of 624 each table. */ 625{ 626 register struct huft *p, *q; 627 628 /* Go through linked list, freeing from the malloced (t[-1]) address. */ 629 p = t; 630 while (p != (struct huft *) NULL) { 631 q = (--p)->v.t; 632 free(p, M_GZIP); 633 p = q; 634 } 635 return 0; 636} 637 638/* inflate (decompress) the codes in a deflated (compressed) block. 639 Return an error code or zero if it all goes ok. */ 640static int 641inflate_codes(glbl, tl, td, bl, bd) 642 struct inflate *glbl; 643 struct huft *tl, *td;/* literal/length and distance decoder tables */ 644 int bl, bd; /* number of bits decoded by tl[] and td[] */ 645{ 646 register unsigned e; /* table entry flag/number of extra bits */ 647 unsigned n, d; /* length and index for copy */ 648 unsigned w; /* current window position */ 649 struct huft *t; /* pointer to table entry */ 650 unsigned ml, md; /* masks for bl and bd bits */ 651 register ulg b; /* bit buffer */ 652 register unsigned k; /* number of bits in bit buffer */ 653 654 /* make local copies of globals */ 655 b = glbl->gz_bb; /* initialize bit buffer */ 656 k = glbl->gz_bk; 657 w = glbl->gz_wp; /* initialize window position */ 658 659 /* inflate the coded data */ 660 ml = mask[bl]; /* precompute masks for speed */ 661 md = mask[bd]; 662 while (1) { /* do until end of block */ 663 NEEDBITS(glbl, (unsigned) bl) 664 if ((e = (t = tl + ((unsigned) b & ml))->e) > 16) 665 do { 666 if (e == 99) 667 return 1; 668 DUMPBITS(t->b) 669 e -= 16; 670 NEEDBITS(glbl, e) 671 } while ((e = (t = t->v.t + ((unsigned) b & mask[e]))->e) > 16); 672 DUMPBITS(t->b) 673 if (e == 16) { /* then it's a literal */ 674 glbl->gz_slide[w++] = (uch) t->v.n; 675 if (w == GZ_WSIZE) { 676 FLUSH(glbl, w); 677 w = 0; 678 } 679 } else { /* it's an EOB or a length */ 680 /* exit if end of block */ 681 if (e == 15) 682 break; 683 684 /* get length of block to copy */ 685 NEEDBITS(glbl, e) 686 n = t->v.n + ((unsigned) b & mask[e]); 687 DUMPBITS(e); 688 689 /* decode distance of block to copy */ 690 NEEDBITS(glbl, (unsigned) bd) 691 if ((e = (t = td + ((unsigned) b & md))->e) > 16) 692 do { 693 if (e == 99) 694 return 1; 695 DUMPBITS(t->b) 696 e -= 16; 697 NEEDBITS(glbl, e) 698 } while ((e = (t = t->v.t + ((unsigned) b & mask[e]))->e) > 16); 699 DUMPBITS(t->b) 700 NEEDBITS(glbl, e) 701 d = w - t->v.n - ((unsigned) b & mask[e]); 702 DUMPBITS(e) 703 /* do the copy */ 704 do { 705 n -= (e = (e = GZ_WSIZE - ((d &= GZ_WSIZE - 1) > w ? d : w)) > n ? n : e); 706#ifndef NOMEMCPY 707 if (w - d >= e) { /* (this test assumes 708 * unsigned comparison) */ 709 memcpy(glbl->gz_slide + w, glbl->gz_slide + d, e); 710 w += e; 711 d += e; 712 } else /* do it slow to avoid memcpy() 713 * overlap */ 714#endif /* !NOMEMCPY */ 715 do { 716 glbl->gz_slide[w++] = glbl->gz_slide[d++]; 717 } while (--e); 718 if (w == GZ_WSIZE) { 719 FLUSH(glbl, w); 720 w = 0; 721 } 722 } while (n); 723 } 724 } 725 726 /* restore the globals from the locals */ 727 glbl->gz_wp = w; /* restore global window pointer */ 728 glbl->gz_bb = b; /* restore global bit buffer */ 729 glbl->gz_bk = k; 730 731 /* done */ 732 return 0; 733} 734 735/* "decompress" an inflated type 0 (stored) block. */ 736static int 737inflate_stored(glbl) 738 struct inflate *glbl; 739{ 740 unsigned n; /* number of bytes in block */ 741 unsigned w; /* current window position */ 742 register ulg b; /* bit buffer */ 743 register unsigned k; /* number of bits in bit buffer */ 744 745 /* make local copies of globals */ 746 b = glbl->gz_bb; /* initialize bit buffer */ 747 k = glbl->gz_bk; 748 w = glbl->gz_wp; /* initialize window position */ 749 750 /* go to byte boundary */ 751 n = k & 7; 752 DUMPBITS(n); 753 754 /* get the length and its complement */ 755 NEEDBITS(glbl, 16) 756 n = ((unsigned) b & 0xffff); 757 DUMPBITS(16) 758 NEEDBITS(glbl, 16) 759 if (n != (unsigned) ((~b) & 0xffff)) 760 return 1; /* error in compressed data */ 761 DUMPBITS(16) 762 /* read and output the compressed data */ 763 while (n--) { 764 NEEDBITS(glbl, 8) 765 glbl->gz_slide[w++] = (uch) b; 766 if (w == GZ_WSIZE) { 767 FLUSH(glbl, w); 768 w = 0; 769 } 770 DUMPBITS(8) 771 } 772 773 /* restore the globals from the locals */ 774 glbl->gz_wp = w; /* restore global window pointer */ 775 glbl->gz_bb = b; /* restore global bit buffer */ 776 glbl->gz_bk = k; 777 return 0; 778} 779 780/* decompress an inflated type 1 (fixed Huffman codes) block. We should 781 either replace this with a custom decoder, or at least precompute the 782 Huffman tables. */ 783static int 784inflate_fixed(glbl) 785 struct inflate *glbl; 786{ 787 /* if first time, set up tables for fixed blocks */ 788 if (glbl->gz_fixed_tl == (struct huft *) NULL) { 789 int i; /* temporary variable */ 790 static unsigned l[288]; /* length list for huft_build */ 791 792 /* literal table */ 793 for (i = 0; i < 144; i++) 794 l[i] = 8; 795 for (; i < 256; i++) 796 l[i] = 9; 797 for (; i < 280; i++) 798 l[i] = 7; 799 for (; i < 288; i++) /* make a complete, but wrong code 800 * set */ 801 l[i] = 8; 802 glbl->gz_fixed_bl = 7; 803 if ((i = huft_build(glbl, l, 288, 257, cplens, cplext, 804 &glbl->gz_fixed_tl, &glbl->gz_fixed_bl)) != 0) { 805 glbl->gz_fixed_tl = (struct huft *) NULL; 806 return i; 807 } 808 /* distance table */ 809 for (i = 0; i < 30; i++) /* make an incomplete code 810 * set */ 811 l[i] = 5; 812 glbl->gz_fixed_bd = 5; 813 if ((i = huft_build(glbl, l, 30, 0, cpdist, cpdext, 814 &glbl->gz_fixed_td, &glbl->gz_fixed_bd)) > 1) { 815 huft_free(glbl, glbl->gz_fixed_tl); 816 glbl->gz_fixed_tl = (struct huft *) NULL; 817 return i; 818 } 819 } 820 /* decompress until an end-of-block code */ 821 return inflate_codes(glbl, glbl->gz_fixed_tl, glbl->gz_fixed_td, glbl->gz_fixed_bl, glbl->gz_fixed_bd) != 0; 822} 823 824/* decompress an inflated type 2 (dynamic Huffman codes) block. */ 825static int 826inflate_dynamic(glbl) 827 struct inflate *glbl; 828{ 829 int i; /* temporary variables */ 830 unsigned j; 831 unsigned l; /* last length */ 832 unsigned m; /* mask for bit lengths table */ 833 unsigned n; /* number of lengths to get */ 834 struct huft *tl; /* literal/length code table */ 835 struct huft *td; /* distance code table */ 836 int bl; /* lookup bits for tl */ 837 int bd; /* lookup bits for td */ 838 unsigned nb; /* number of bit length codes */ 839 unsigned nl; /* number of literal/length codes */ 840 unsigned nd; /* number of distance codes */ 841#ifdef PKZIP_BUG_WORKAROUND 842 unsigned ll[288 + 32]; /* literal/length and distance code 843 * lengths */ 844#else 845 unsigned ll[286 + 30]; /* literal/length and distance code 846 * lengths */ 847#endif 848 register ulg b; /* bit buffer */ 849 register unsigned k; /* number of bits in bit buffer */ 850 851 /* make local bit buffer */ 852 b = glbl->gz_bb; 853 k = glbl->gz_bk; 854 855 /* read in table lengths */ 856 NEEDBITS(glbl, 5) 857 nl = 257 + ((unsigned) b & 0x1f); /* number of 858 * literal/length codes */ 859 DUMPBITS(5) 860 NEEDBITS(glbl, 5) 861 nd = 1 + ((unsigned) b & 0x1f); /* number of distance codes */ 862 DUMPBITS(5) 863 NEEDBITS(glbl, 4) 864 nb = 4 + ((unsigned) b & 0xf); /* number of bit length codes */ 865 DUMPBITS(4) 866#ifdef PKZIP_BUG_WORKAROUND 867 if (nl > 288 || nd > 32) 868#else 869 if (nl > 286 || nd > 30) 870#endif 871 return 1; /* bad lengths */ 872 /* read in bit-length-code lengths */ 873 for (j = 0; j < nb; j++) { 874 NEEDBITS(glbl, 3) 875 ll[border[j]] = (unsigned) b & 7; 876 DUMPBITS(3) 877 } 878 for (; j < 19; j++) 879 ll[border[j]] = 0; 880 881 /* build decoding table for trees--single level, 7 bit lookup */ 882 bl = 7; 883 if ((i = huft_build(glbl, ll, 19, 19, NULL, NULL, &tl, &bl)) != 0) { 884 if (i == 1) 885 huft_free(glbl, tl); 886 return i; /* incomplete code set */ 887 } 888 /* read in literal and distance code lengths */ 889 n = nl + nd; 890 m = mask[bl]; 891 i = l = 0; 892 while ((unsigned) i < n) { 893 NEEDBITS(glbl, (unsigned) bl) 894 j = (td = tl + ((unsigned) b & m))->b; 895 DUMPBITS(j) 896 j = td->v.n; 897 if (j < 16) /* length of code in bits (0..15) */ 898 ll[i++] = l = j; /* save last length in l */ 899 else if (j == 16) { /* repeat last length 3 to 6 times */ 900 NEEDBITS(glbl, 2) 901 j = 3 + ((unsigned) b & 3); 902 DUMPBITS(2) 903 if ((unsigned) i + j > n) 904 return 1; 905 while (j--) 906 ll[i++] = l; 907 } else if (j == 17) { /* 3 to 10 zero length codes */ 908 NEEDBITS(glbl, 3) 909 j = 3 + ((unsigned) b & 7); 910 DUMPBITS(3) 911 if ((unsigned) i + j > n) 912 return 1; 913 while (j--) 914 ll[i++] = 0; 915 l = 0; 916 } else { /* j == 18: 11 to 138 zero length codes */ 917 NEEDBITS(glbl, 7) 918 j = 11 + ((unsigned) b & 0x7f); 919 DUMPBITS(7) 920 if ((unsigned) i + j > n) 921 return 1; 922 while (j--) 923 ll[i++] = 0; 924 l = 0; 925 } 926 } 927 928 /* free decoding table for trees */ 929 huft_free(glbl, tl); 930 931 /* restore the global bit buffer */ 932 glbl->gz_bb = b; 933 glbl->gz_bk = k; 934 935 /* build the decoding tables for literal/length and distance codes */ 936 bl = lbits; 937 i = huft_build(glbl, ll, nl, 257, cplens, cplext, &tl, &bl); 938 if (i != 0) { 939 if (i == 1 && !qflag) { 940 FPRINTF("(incomplete l-tree) "); 941 huft_free(glbl, tl); 942 } 943 return i; /* incomplete code set */ 944 } 945 bd = dbits; 946 i = huft_build(glbl, ll + nl, nd, 0, cpdist, cpdext, &td, &bd); 947 if (i != 0) { 948 if (i == 1 && !qflag) { 949 FPRINTF("(incomplete d-tree) "); 950#ifdef PKZIP_BUG_WORKAROUND 951 i = 0; 952 } 953#else 954 huft_free(glbl, td); 955 } 956 huft_free(glbl, tl); 957 return i; /* incomplete code set */ 958#endif 959 } 960 /* decompress until an end-of-block code */ 961 if (inflate_codes(glbl, tl, td, bl, bd)) 962 return 1; 963 964 /* free the decoding tables, return */ 965 huft_free(glbl, tl); 966 huft_free(glbl, td); 967 return 0; 968} 969 970/* decompress an inflated block */ 971static int 972inflate_block(glbl, e) 973 struct inflate *glbl; 974 int *e; /* last block flag */ 975{ 976 unsigned t; /* block type */ 977 register ulg b; /* bit buffer */ 978 register unsigned k; /* number of bits in bit buffer */ 979 980 /* make local bit buffer */ 981 b = glbl->gz_bb; 982 k = glbl->gz_bk; 983 984 /* read in last block bit */ 985 NEEDBITS(glbl, 1) 986 * e = (int) b & 1; 987 DUMPBITS(1) 988 /* read in block type */ 989 NEEDBITS(glbl, 2) 990 t = (unsigned) b & 3; 991 DUMPBITS(2) 992 /* restore the global bit buffer */ 993 glbl->gz_bb = b; 994 glbl->gz_bk = k; 995 996 /* inflate that block type */ 997 if (t == 2) 998 return inflate_dynamic(glbl); 999 if (t == 0) 1000 return inflate_stored(glbl); 1001 if (t == 1) 1002 return inflate_fixed(glbl); 1003 /* bad block type */ 1004 return 2; 1005} 1006 1007 1008 1009/* decompress an inflated entry */ 1010static int 1011xinflate(glbl) 1012 struct inflate *glbl; 1013{ 1014 int e; /* last block flag */ 1015 int r; /* result code */ 1016 unsigned h; /* maximum struct huft's malloc'ed */ 1017 1018 glbl->gz_fixed_tl = (struct huft *) NULL; 1019 1020 /* initialize window, bit buffer */ 1021 glbl->gz_wp = 0; 1022 glbl->gz_bk = 0; 1023 glbl->gz_bb = 0; 1024 1025 /* decompress until the last block */ 1026 h = 0; 1027 do { 1028 glbl->gz_hufts = 0; 1029 if ((r = inflate_block(glbl, &e)) != 0) 1030 return r; 1031 if (glbl->gz_hufts > h) 1032 h = glbl->gz_hufts; 1033 } while (!e); 1034 1035 /* flush out slide */ 1036 FLUSH(glbl, glbl->gz_wp); 1037 1038 /* return success */ 1039 return 0; 1040} 1041 1042/* Nobody uses this - why not? */ 1043int 1044inflate(glbl) 1045 struct inflate *glbl; 1046{ 1047 int i; 1048#ifdef KERNEL 1049 u_char *p = NULL; 1050 1051 if (!glbl->gz_slide) 1052 p = glbl->gz_slide = malloc(GZ_WSIZE, M_GZIP, M_WAITOK); 1053#endif 1054 if (!glbl->gz_slide) 1055#ifdef KERNEL 1056 return(ENOMEM); 1057#else 1058 return 3; /* kzip expects 3 */ 1059#endif 1060 i = xinflate(glbl); 1061 1062 if (glbl->gz_fixed_td != (struct huft *) NULL) { 1063 huft_free(glbl, glbl->gz_fixed_td); 1064 glbl->gz_fixed_td = (struct huft *) NULL; 1065 } 1066 if (glbl->gz_fixed_tl != (struct huft *) NULL) { 1067 huft_free(glbl, glbl->gz_fixed_tl); 1068 glbl->gz_fixed_tl = (struct huft *) NULL; 1069 } 1070#ifdef KERNEL 1071 if (p == glbl->gz_slide) { 1072 free(glbl->gz_slide, M_GZIP); 1073 glbl->gz_slide = NULL; 1074 } 1075#endif 1076 return i; 1077} 1078/* ----------------------- END INFLATE.C */ 1079