Optimize andes_clear_page() and andes_copy_page() with prefetch
[linux-2.6/linux-mips.git] / lib / inflate.c
blob7b707075a2f76df68990cf0bec010cb85254ddc0
1 #define DEBG(x)
2 #define DEBG1(x)
3 /* inflate.c -- Not copyrighted 1992 by Mark Adler
4 version c10p1, 10 January 1993 */
6 /*
7 * Adapted for booting Linux by Hannu Savolainen 1993
8 * based on gzip-1.0.3
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 embeded systems.
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.
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.
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.
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.
62 Notes beyond the 1.93a appnote.txt:
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.
106 #ifdef RCSID
107 static char rcsid[] = "#Id: inflate.c,v 0.14 1993/06/10 13:27:04 jloup Exp #";
108 #endif
110 #ifndef STATIC
112 #if defined(STDC_HEADERS) || defined(HAVE_STDLIB_H)
113 # include <sys/types.h>
114 # include <stdlib.h>
115 #endif
117 #include "gzip.h"
118 #define STATIC
119 #endif /* !STATIC */
121 #define slide window
123 /* Huffman code lookup table entry--this entry is four bytes for machines
124 that have 16-bit pointers (e.g. PC's in the small or medium model).
125 Valid extra bits are 0..13. e == 15 is EOB (end of block), e == 16
126 means that v is a literal, 16 < e < 32 means that v is a pointer to
127 the next table, which codes e - 16 bits, and lastly e == 99 indicates
128 an unused code. If a code with e == 99 is looked up, this implies an
129 error in the data. */
130 struct huft {
131 uch e; /* number of extra bits or operation */
132 uch b; /* number of bits in this code or subcode */
133 union {
134 ush n; /* literal, length base, or distance base */
135 struct huft *t; /* pointer to next level of table */
136 } v;
140 /* Function prototypes */
141 STATIC int huft_build OF((unsigned *, unsigned, unsigned,
142 const ush *, const ush *, struct huft **, int *));
143 STATIC int huft_free OF((struct huft *));
144 STATIC int inflate_codes OF((struct huft *, struct huft *, int, int));
145 STATIC int inflate_stored OF((void));
146 STATIC int inflate_fixed OF((void));
147 STATIC int inflate_dynamic OF((void));
148 STATIC int inflate_block OF((int *));
149 STATIC int inflate OF((void));
152 /* The inflate algorithm uses a sliding 32 K byte window on the uncompressed
153 stream to find repeated byte strings. This is implemented here as a
154 circular buffer. The index is updated simply by incrementing and then
155 ANDing with 0x7fff (32K-1). */
156 /* It is left to other modules to supply the 32 K area. It is assumed
157 to be usable as if it were declared "uch slide[32768];" or as just
158 "uch *slide;" and then malloc'ed in the latter case. The definition
159 must be in unzip.h, included above. */
160 /* unsigned wp; current position in slide */
161 #define wp outcnt
162 #define flush_output(w) (wp=(w),flush_window())
164 /* Tables for deflate from PKZIP's appnote.txt. */
165 static const unsigned border[] = { /* Order of the bit length code lengths */
166 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
167 static const ush cplens[] = { /* Copy lengths for literal codes 257..285 */
168 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
169 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
170 /* note: see note #13 above about the 258 in this list. */
171 static const ush cplext[] = { /* Extra bits for literal codes 257..285 */
172 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
173 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */
174 static const ush cpdist[] = { /* Copy offsets for distance codes 0..29 */
175 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
176 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
177 8193, 12289, 16385, 24577};
178 static const ush cpdext[] = { /* Extra bits for distance codes */
179 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
180 7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
181 12, 12, 13, 13};
185 /* Macros for inflate() bit peeking and grabbing.
186 The usage is:
188 NEEDBITS(j)
189 x = b & mask_bits[j];
190 DUMPBITS(j)
192 where NEEDBITS makes sure that b has at least j bits in it, and
193 DUMPBITS removes the bits from b. The macros use the variable k
194 for the number of bits in b. Normally, b and k are register
195 variables for speed, and are initialized at the beginning of a
196 routine that uses these macros from a global bit buffer and count.
198 If we assume that EOB will be the longest code, then we will never
199 ask for bits with NEEDBITS that are beyond the end of the stream.
200 So, NEEDBITS should not read any more bytes than are needed to
201 meet the request. Then no bytes need to be "returned" to the buffer
202 at the end of the last block.
204 However, this assumption is not true for fixed blocks--the EOB code
205 is 7 bits, but the other literal/length codes can be 8 or 9 bits.
206 (The EOB code is shorter than other codes because fixed blocks are
207 generally short. So, while a block always has an EOB, many other
208 literal/length codes have a significantly lower probability of
209 showing up at all.) However, by making the first table have a
210 lookup of seven bits, the EOB code will be found in that first
211 lookup, and so will not require that too many bits be pulled from
212 the stream.
215 STATIC ulg bb; /* bit buffer */
216 STATIC unsigned bk; /* bits in bit buffer */
218 STATIC const ush mask_bits[] = {
219 0x0000,
220 0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff,
221 0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff
224 #define NEXTBYTE() (uch)get_byte()
225 #define NEEDBITS(n) {while(k<(n)){b|=((ulg)NEXTBYTE())<<k;k+=8;}}
226 #define DUMPBITS(n) {b>>=(n);k-=(n);}
230 Huffman code decoding is performed using a multi-level table lookup.
231 The fastest way to decode is to simply build a lookup table whose
232 size is determined by the longest code. However, the time it takes
233 to build this table can also be a factor if the data being decoded
234 is not very long. The most common codes are necessarily the
235 shortest codes, so those codes dominate the decoding time, and hence
236 the speed. The idea is you can have a shorter table that decodes the
237 shorter, more probable codes, and then point to subsidiary tables for
238 the longer codes. The time it costs to decode the longer codes is
239 then traded against the time it takes to make longer tables.
241 This results of this trade are in the variables lbits and dbits
242 below. lbits is the number of bits the first level table for literal/
243 length codes can decode in one step, and dbits is the same thing for
244 the distance codes. Subsequent tables are also less than or equal to
245 those sizes. These values may be adjusted either when all of the
246 codes are shorter than that, in which case the longest code length in
247 bits is used, or when the shortest code is *longer* than the requested
248 table size, in which case the length of the shortest code in bits is
249 used.
251 There are two different values for the two tables, since they code a
252 different number of possibilities each. The literal/length table
253 codes 286 possible values, or in a flat code, a little over eight
254 bits. The distance table codes 30 possible values, or a little less
255 than five bits, flat. The optimum values for speed end up being
256 about one bit more than those, so lbits is 8+1 and dbits is 5+1.
257 The optimum values may differ though from machine to machine, and
258 possibly even between compilers. Your mileage may vary.
262 STATIC const int lbits = 9; /* bits in base literal/length lookup table */
263 STATIC const int dbits = 6; /* bits in base distance lookup table */
266 /* If BMAX needs to be larger than 16, then h and x[] should be ulg. */
267 #define BMAX 16 /* maximum bit length of any code (16 for explode) */
268 #define N_MAX 288 /* maximum number of codes in any set */
271 STATIC unsigned hufts; /* track memory usage */
274 STATIC int huft_build(b, n, s, d, e, t, m)
275 unsigned *b; /* code lengths in bits (all assumed <= BMAX) */
276 unsigned n; /* number of codes (assumed <= N_MAX) */
277 unsigned s; /* number of simple-valued codes (0..s-1) */
278 const ush *d; /* list of base values for non-simple codes */
279 const ush *e; /* list of extra bits for non-simple codes */
280 struct huft **t; /* result: starting table */
281 int *m; /* maximum lookup bits, returns actual */
282 /* Given a list of code lengths and a maximum table size, make a set of
283 tables to decode that set of codes. Return zero on success, one if
284 the given code set is incomplete (the tables are still built in this
285 case), two if the input is invalid (all zero length codes or an
286 oversubscribed set of lengths), and three if not enough memory. */
288 unsigned a; /* counter for codes of length k */
289 unsigned c[BMAX+1]; /* bit length count table */
290 unsigned f; /* i repeats in table every f entries */
291 int g; /* maximum code length */
292 int h; /* table level */
293 register unsigned i; /* counter, current code */
294 register unsigned j; /* counter */
295 register int k; /* number of bits in current code */
296 int l; /* bits per table (returned in m) */
297 register unsigned *p; /* pointer into c[], b[], or v[] */
298 register struct huft *q; /* points to current table */
299 struct huft r; /* table entry for structure assignment */
300 struct huft *u[BMAX]; /* table stack */
301 unsigned v[N_MAX]; /* values in order of bit length */
302 register int w; /* bits before this table == (l * h) */
303 unsigned x[BMAX+1]; /* bit offsets, then code stack */
304 unsigned *xp; /* pointer into x */
305 int y; /* number of dummy codes added */
306 unsigned z; /* number of entries in current table */
308 DEBG("huft1 ");
310 /* Generate counts for each bit length */
311 memzero(c, sizeof(c));
312 p = b; i = n;
313 do {
314 Tracecv(*p, (stderr, (n-i >= ' ' && n-i <= '~' ? "%c %d\n" : "0x%x %d\n"),
315 n-i, *p));
316 c[*p]++; /* assume all entries <= BMAX */
317 p++; /* Can't combine with above line (Solaris bug) */
318 } while (--i);
319 if (c[0] == n) /* null input--all zero length codes */
321 *t = (struct huft *)NULL;
322 *m = 0;
323 return 0;
326 DEBG("huft2 ");
328 /* Find minimum and maximum length, bound *m by those */
329 l = *m;
330 for (j = 1; j <= BMAX; j++)
331 if (c[j])
332 break;
333 k = j; /* minimum code length */
334 if ((unsigned)l < j)
335 l = j;
336 for (i = BMAX; i; i--)
337 if (c[i])
338 break;
339 g = i; /* maximum code length */
340 if ((unsigned)l > i)
341 l = i;
342 *m = l;
344 DEBG("huft3 ");
346 /* Adjust last length count to fill out codes, if needed */
347 for (y = 1 << j; j < i; j++, y <<= 1)
348 if ((y -= c[j]) < 0)
349 return 2; /* bad input: more codes than bits */
350 if ((y -= c[i]) < 0)
351 return 2;
352 c[i] += y;
354 DEBG("huft4 ");
356 /* Generate starting offsets into the value table for each length */
357 x[1] = j = 0;
358 p = c + 1; xp = x + 2;
359 while (--i) { /* note that i == g from above */
360 *xp++ = (j += *p++);
363 DEBG("huft5 ");
365 /* Make a table of values in order of bit lengths */
366 p = b; i = 0;
367 do {
368 if ((j = *p++) != 0)
369 v[x[j]++] = i;
370 } while (++i < n);
372 DEBG("h6 ");
374 /* Generate the Huffman codes and for each, make the table entries */
375 x[0] = i = 0; /* first Huffman code is zero */
376 p = v; /* grab values in bit order */
377 h = -1; /* no tables yet--level -1 */
378 w = -l; /* bits decoded == (l * h) */
379 u[0] = (struct huft *)NULL; /* just to keep compilers happy */
380 q = (struct huft *)NULL; /* ditto */
381 z = 0; /* ditto */
382 DEBG("h6a ");
384 /* go through the bit lengths (k already is bits in shortest code) */
385 for (; k <= g; k++)
387 DEBG("h6b ");
388 a = c[k];
389 while (a--)
391 DEBG("h6b1 ");
392 /* here i is the Huffman code of length k bits for value *p */
393 /* make tables up to required level */
394 while (k > w + l)
396 DEBG1("1 ");
397 h++;
398 w += l; /* previous table always l bits */
400 /* compute minimum size table less than or equal to l bits */
401 z = (z = g - w) > (unsigned)l ? l : z; /* upper limit on table size */
402 if ((f = 1 << (j = k - w)) > a + 1) /* try a k-w bit table */
403 { /* too few codes for k-w bit table */
404 DEBG1("2 ");
405 f -= a + 1; /* deduct codes from patterns left */
406 xp = c + k;
407 while (++j < z) /* try smaller tables up to z bits */
409 if ((f <<= 1) <= *++xp)
410 break; /* enough codes to use up j bits */
411 f -= *xp; /* else deduct codes from patterns */
414 DEBG1("3 ");
415 z = 1 << j; /* table entries for j-bit table */
417 /* allocate and link in new table */
418 if ((q = (struct huft *)malloc((z + 1)*sizeof(struct huft))) ==
419 (struct huft *)NULL)
421 if (h)
422 huft_free(u[0]);
423 return 3; /* not enough memory */
425 DEBG1("4 ");
426 hufts += z + 1; /* track memory usage */
427 *t = q + 1; /* link to list for huft_free() */
428 *(t = &(q->v.t)) = (struct huft *)NULL;
429 u[h] = ++q; /* table starts after link */
431 DEBG1("5 ");
432 /* connect to last table, if there is one */
433 if (h)
435 x[h] = i; /* save pattern for backing up */
436 r.b = (uch)l; /* bits to dump before this table */
437 r.e = (uch)(16 + j); /* bits in this table */
438 r.v.t = q; /* pointer to this table */
439 j = i >> (w - l); /* (get around Turbo C bug) */
440 u[h-1][j] = r; /* connect to last table */
442 DEBG1("6 ");
444 DEBG("h6c ");
446 /* set up table entry in r */
447 r.b = (uch)(k - w);
448 if (p >= v + n)
449 r.e = 99; /* out of values--invalid code */
450 else if (*p < s)
452 r.e = (uch)(*p < 256 ? 16 : 15); /* 256 is end-of-block code */
453 r.v.n = (ush)(*p); /* simple code is just the value */
454 p++; /* one compiler does not like *p++ */
456 else
458 r.e = (uch)e[*p - s]; /* non-simple--look up in lists */
459 r.v.n = d[*p++ - s];
461 DEBG("h6d ");
463 /* fill code-like entries with r */
464 f = 1 << (k - w);
465 for (j = i >> w; j < z; j += f)
466 q[j] = r;
468 /* backwards increment the k-bit code i */
469 for (j = 1 << (k - 1); i & j; j >>= 1)
470 i ^= j;
471 i ^= j;
473 /* backup over finished tables */
474 while ((i & ((1 << w) - 1)) != x[h])
476 h--; /* don't need to update q */
477 w -= l;
479 DEBG("h6e ");
481 DEBG("h6f ");
484 DEBG("huft7 ");
486 /* Return true (1) if we were given an incomplete table */
487 return y != 0 && g != 1;
492 STATIC int huft_free(t)
493 struct huft *t; /* table to free */
494 /* Free the malloc'ed tables built by huft_build(), which makes a linked
495 list of the tables it made, with the links in a dummy first entry of
496 each table. */
498 register struct huft *p, *q;
501 /* Go through linked list, freeing from the malloced (t[-1]) address. */
502 p = t;
503 while (p != (struct huft *)NULL)
505 q = (--p)->v.t;
506 free((char*)p);
507 p = q;
509 return 0;
513 STATIC int inflate_codes(tl, td, bl, bd)
514 struct huft *tl, *td; /* literal/length and distance decoder tables */
515 int bl, bd; /* number of bits decoded by tl[] and td[] */
516 /* inflate (decompress) the codes in a deflated (compressed) block.
517 Return an error code or zero if it all goes ok. */
519 register unsigned e; /* table entry flag/number of extra bits */
520 unsigned n, d; /* length and index for copy */
521 unsigned w; /* current window position */
522 struct huft *t; /* pointer to table entry */
523 unsigned ml, md; /* masks for bl and bd bits */
524 register ulg b; /* bit buffer */
525 register unsigned k; /* number of bits in bit buffer */
528 /* make local copies of globals */
529 b = bb; /* initialize bit buffer */
530 k = bk;
531 w = wp; /* initialize window position */
533 /* inflate the coded data */
534 ml = mask_bits[bl]; /* precompute masks for speed */
535 md = mask_bits[bd];
536 for (;;) /* do until end of block */
538 NEEDBITS((unsigned)bl)
539 if ((e = (t = tl + ((unsigned)b & ml))->e) > 16)
540 do {
541 if (e == 99)
542 return 1;
543 DUMPBITS(t->b)
544 e -= 16;
545 NEEDBITS(e)
546 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
547 DUMPBITS(t->b)
548 if (e == 16) /* then it's a literal */
550 slide[w++] = (uch)t->v.n;
551 Tracevv((stderr, "%c", slide[w-1]));
552 if (w == WSIZE)
554 flush_output(w);
555 w = 0;
558 else /* it's an EOB or a length */
560 /* exit if end of block */
561 if (e == 15)
562 break;
564 /* get length of block to copy */
565 NEEDBITS(e)
566 n = t->v.n + ((unsigned)b & mask_bits[e]);
567 DUMPBITS(e);
569 /* decode distance of block to copy */
570 NEEDBITS((unsigned)bd)
571 if ((e = (t = td + ((unsigned)b & md))->e) > 16)
572 do {
573 if (e == 99)
574 return 1;
575 DUMPBITS(t->b)
576 e -= 16;
577 NEEDBITS(e)
578 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
579 DUMPBITS(t->b)
580 NEEDBITS(e)
581 d = w - t->v.n - ((unsigned)b & mask_bits[e]);
582 DUMPBITS(e)
583 Tracevv((stderr,"\\[%d,%d]", w-d, n));
585 /* do the copy */
586 do {
587 n -= (e = (e = WSIZE - ((d &= WSIZE-1) > w ? d : w)) > n ? n : e);
588 #if !defined(NOMEMCPY) && !defined(DEBUG)
589 if (w - d >= e) /* (this test assumes unsigned comparison) */
591 memcpy(slide + w, slide + d, e);
592 w += e;
593 d += e;
595 else /* do it slow to avoid memcpy() overlap */
596 #endif /* !NOMEMCPY */
597 do {
598 slide[w++] = slide[d++];
599 Tracevv((stderr, "%c", slide[w-1]));
600 } while (--e);
601 if (w == WSIZE)
603 flush_output(w);
604 w = 0;
606 } while (n);
611 /* restore the globals from the locals */
612 wp = w; /* restore global window pointer */
613 bb = b; /* restore global bit buffer */
614 bk = k;
616 /* done */
617 return 0;
622 STATIC int inflate_stored()
623 /* "decompress" an inflated type 0 (stored) block. */
625 unsigned n; /* number of bytes in block */
626 unsigned w; /* current window position */
627 register ulg b; /* bit buffer */
628 register unsigned k; /* number of bits in bit buffer */
630 DEBG("<stor");
632 /* make local copies of globals */
633 b = bb; /* initialize bit buffer */
634 k = bk;
635 w = wp; /* initialize window position */
638 /* go to byte boundary */
639 n = k & 7;
640 DUMPBITS(n);
643 /* get the length and its complement */
644 NEEDBITS(16)
645 n = ((unsigned)b & 0xffff);
646 DUMPBITS(16)
647 NEEDBITS(16)
648 if (n != (unsigned)((~b) & 0xffff))
649 return 1; /* error in compressed data */
650 DUMPBITS(16)
653 /* read and output the compressed data */
654 while (n--)
656 NEEDBITS(8)
657 slide[w++] = (uch)b;
658 if (w == WSIZE)
660 flush_output(w);
661 w = 0;
663 DUMPBITS(8)
667 /* restore the globals from the locals */
668 wp = w; /* restore global window pointer */
669 bb = b; /* restore global bit buffer */
670 bk = k;
672 DEBG(">");
673 return 0;
678 STATIC int inflate_fixed()
679 /* decompress an inflated type 1 (fixed Huffman codes) block. We should
680 either replace this with a custom decoder, or at least precompute the
681 Huffman tables. */
683 int i; /* temporary variable */
684 struct huft *tl; /* literal/length code table */
685 struct huft *td; /* distance code table */
686 int bl; /* lookup bits for tl */
687 int bd; /* lookup bits for td */
688 unsigned l[288]; /* length list for huft_build */
690 DEBG("<fix");
692 /* set up literal table */
693 for (i = 0; i < 144; i++)
694 l[i] = 8;
695 for (; i < 256; i++)
696 l[i] = 9;
697 for (; i < 280; i++)
698 l[i] = 7;
699 for (; i < 288; i++) /* make a complete, but wrong code set */
700 l[i] = 8;
701 bl = 7;
702 if ((i = huft_build(l, 288, 257, cplens, cplext, &tl, &bl)) != 0)
703 return i;
706 /* set up distance table */
707 for (i = 0; i < 30; i++) /* make an incomplete code set */
708 l[i] = 5;
709 bd = 5;
710 if ((i = huft_build(l, 30, 0, cpdist, cpdext, &td, &bd)) > 1)
712 huft_free(tl);
714 DEBG(">");
715 return i;
719 /* decompress until an end-of-block code */
720 if (inflate_codes(tl, td, bl, bd))
721 return 1;
724 /* free the decoding tables, return */
725 huft_free(tl);
726 huft_free(td);
727 return 0;
732 STATIC int inflate_dynamic()
733 /* decompress an inflated type 2 (dynamic Huffman codes) block. */
735 int i; /* temporary variables */
736 unsigned j;
737 unsigned l; /* last length */
738 unsigned m; /* mask for bit lengths table */
739 unsigned n; /* number of lengths to get */
740 struct huft *tl; /* literal/length code table */
741 struct huft *td; /* distance code table */
742 int bl; /* lookup bits for tl */
743 int bd; /* lookup bits for td */
744 unsigned nb; /* number of bit length codes */
745 unsigned nl; /* number of literal/length codes */
746 unsigned nd; /* number of distance codes */
747 #ifdef PKZIP_BUG_WORKAROUND
748 unsigned ll[288+32]; /* literal/length and distance code lengths */
749 #else
750 unsigned ll[286+30]; /* literal/length and distance code lengths */
751 #endif
752 register ulg b; /* bit buffer */
753 register unsigned k; /* number of bits in bit buffer */
755 DEBG("<dyn");
757 /* make local bit buffer */
758 b = bb;
759 k = bk;
762 /* read in table lengths */
763 NEEDBITS(5)
764 nl = 257 + ((unsigned)b & 0x1f); /* number of literal/length codes */
765 DUMPBITS(5)
766 NEEDBITS(5)
767 nd = 1 + ((unsigned)b & 0x1f); /* number of distance codes */
768 DUMPBITS(5)
769 NEEDBITS(4)
770 nb = 4 + ((unsigned)b & 0xf); /* number of bit length codes */
771 DUMPBITS(4)
772 #ifdef PKZIP_BUG_WORKAROUND
773 if (nl > 288 || nd > 32)
774 #else
775 if (nl > 286 || nd > 30)
776 #endif
777 return 1; /* bad lengths */
779 DEBG("dyn1 ");
781 /* read in bit-length-code lengths */
782 for (j = 0; j < nb; j++)
784 NEEDBITS(3)
785 ll[border[j]] = (unsigned)b & 7;
786 DUMPBITS(3)
788 for (; j < 19; j++)
789 ll[border[j]] = 0;
791 DEBG("dyn2 ");
793 /* build decoding table for trees--single level, 7 bit lookup */
794 bl = 7;
795 if ((i = huft_build(ll, 19, 19, NULL, NULL, &tl, &bl)) != 0)
797 if (i == 1)
798 huft_free(tl);
799 return i; /* incomplete code set */
802 DEBG("dyn3 ");
804 /* read in literal and distance code lengths */
805 n = nl + nd;
806 m = mask_bits[bl];
807 i = l = 0;
808 while ((unsigned)i < n)
810 NEEDBITS((unsigned)bl)
811 j = (td = tl + ((unsigned)b & m))->b;
812 DUMPBITS(j)
813 j = td->v.n;
814 if (j < 16) /* length of code in bits (0..15) */
815 ll[i++] = l = j; /* save last length in l */
816 else if (j == 16) /* repeat last length 3 to 6 times */
818 NEEDBITS(2)
819 j = 3 + ((unsigned)b & 3);
820 DUMPBITS(2)
821 if ((unsigned)i + j > n)
822 return 1;
823 while (j--)
824 ll[i++] = l;
826 else if (j == 17) /* 3 to 10 zero length codes */
828 NEEDBITS(3)
829 j = 3 + ((unsigned)b & 7);
830 DUMPBITS(3)
831 if ((unsigned)i + j > n)
832 return 1;
833 while (j--)
834 ll[i++] = 0;
835 l = 0;
837 else /* j == 18: 11 to 138 zero length codes */
839 NEEDBITS(7)
840 j = 11 + ((unsigned)b & 0x7f);
841 DUMPBITS(7)
842 if ((unsigned)i + j > n)
843 return 1;
844 while (j--)
845 ll[i++] = 0;
846 l = 0;
850 DEBG("dyn4 ");
852 /* free decoding table for trees */
853 huft_free(tl);
855 DEBG("dyn5 ");
857 /* restore the global bit buffer */
858 bb = b;
859 bk = k;
861 DEBG("dyn5a ");
863 /* build the decoding tables for literal/length and distance codes */
864 bl = lbits;
865 if ((i = huft_build(ll, nl, 257, cplens, cplext, &tl, &bl)) != 0)
867 DEBG("dyn5b ");
868 if (i == 1) {
869 error(" incomplete literal tree\n");
870 huft_free(tl);
872 return i; /* incomplete code set */
874 DEBG("dyn5c ");
875 bd = dbits;
876 if ((i = huft_build(ll + nl, nd, 0, cpdist, cpdext, &td, &bd)) != 0)
878 DEBG("dyn5d ");
879 if (i == 1) {
880 error(" incomplete distance tree\n");
881 #ifdef PKZIP_BUG_WORKAROUND
882 i = 0;
884 #else
885 huft_free(td);
887 huft_free(tl);
888 return i; /* incomplete code set */
889 #endif
892 DEBG("dyn6 ");
894 /* decompress until an end-of-block code */
895 if (inflate_codes(tl, td, bl, bd))
896 return 1;
898 DEBG("dyn7 ");
900 /* free the decoding tables, return */
901 huft_free(tl);
902 huft_free(td);
904 DEBG(">");
905 return 0;
910 STATIC int inflate_block(e)
911 int *e; /* last block flag */
912 /* decompress an inflated block */
914 unsigned t; /* block type */
915 register ulg b; /* bit buffer */
916 register unsigned k; /* number of bits in bit buffer */
918 DEBG("<blk");
920 /* make local bit buffer */
921 b = bb;
922 k = bk;
925 /* read in last block bit */
926 NEEDBITS(1)
927 *e = (int)b & 1;
928 DUMPBITS(1)
931 /* read in block type */
932 NEEDBITS(2)
933 t = (unsigned)b & 3;
934 DUMPBITS(2)
937 /* restore the global bit buffer */
938 bb = b;
939 bk = k;
941 /* inflate that block type */
942 if (t == 2)
943 return inflate_dynamic();
944 if (t == 0)
945 return inflate_stored();
946 if (t == 1)
947 return inflate_fixed();
949 DEBG(">");
951 /* bad block type */
952 return 2;
957 STATIC int inflate()
958 /* decompress an inflated entry */
960 int e; /* last block flag */
961 int r; /* result code */
962 unsigned h; /* maximum struct huft's malloc'ed */
963 void *ptr;
965 /* initialize window, bit buffer */
966 wp = 0;
967 bk = 0;
968 bb = 0;
971 /* decompress until the last block */
972 h = 0;
973 do {
974 hufts = 0;
975 gzip_mark(&ptr);
976 if ((r = inflate_block(&e)) != 0) {
977 gzip_release(&ptr);
978 return r;
980 gzip_release(&ptr);
981 if (hufts > h)
982 h = hufts;
983 } while (!e);
985 /* Undo too much lookahead. The next read will be byte aligned so we
986 * can discard unused bits in the last meaningful byte.
988 while (bk >= 8) {
989 bk -= 8;
990 inptr--;
993 /* flush out slide */
994 flush_output(wp);
997 /* return success */
998 #ifdef DEBUG
999 fprintf(stderr, "<%u> ", h);
1000 #endif /* DEBUG */
1001 return 0;
1004 /**********************************************************************
1006 * The following are support routines for inflate.c
1008 **********************************************************************/
1010 static ulg crc_32_tab[256];
1011 static ulg crc; /* initialized in makecrc() so it'll reside in bss */
1012 #define CRC_VALUE (crc ^ 0xffffffffL)
1015 * Code to compute the CRC-32 table. Borrowed from
1016 * gzip-1.0.3/makecrc.c.
1019 static void
1020 makecrc(void)
1022 /* Not copyrighted 1990 Mark Adler */
1024 unsigned long c; /* crc shift register */
1025 unsigned long e; /* polynomial exclusive-or pattern */
1026 int i; /* counter for all possible eight bit values */
1027 int k; /* byte being shifted into crc apparatus */
1029 /* terms of polynomial defining this crc (except x^32): */
1030 static const int p[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26};
1032 /* Make exclusive-or pattern from polynomial */
1033 e = 0;
1034 for (i = 0; i < sizeof(p)/sizeof(int); i++)
1035 e |= 1L << (31 - p[i]);
1037 crc_32_tab[0] = 0;
1039 for (i = 1; i < 256; i++)
1041 c = 0;
1042 for (k = i | 256; k != 1; k >>= 1)
1044 c = c & 1 ? (c >> 1) ^ e : c >> 1;
1045 if (k & 1)
1046 c ^= e;
1048 crc_32_tab[i] = c;
1051 /* this is initialized here so this code could reside in ROM */
1052 crc = (ulg)0xffffffffL; /* shift register contents */
1055 /* gzip flag byte */
1056 #define ASCII_FLAG 0x01 /* bit 0 set: file probably ASCII text */
1057 #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
1058 #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
1059 #define ORIG_NAME 0x08 /* bit 3 set: original file name present */
1060 #define COMMENT 0x10 /* bit 4 set: file comment present */
1061 #define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */
1062 #define RESERVED 0xC0 /* bit 6,7: reserved */
1065 * Do the uncompression!
1067 static int gunzip(void)
1069 uch flags;
1070 unsigned char magic[2]; /* magic header */
1071 char method;
1072 ulg orig_crc = 0; /* original crc */
1073 ulg orig_len = 0; /* original uncompressed length */
1074 int res;
1076 magic[0] = (unsigned char)get_byte();
1077 magic[1] = (unsigned char)get_byte();
1078 method = (unsigned char)get_byte();
1080 if (magic[0] != 037 ||
1081 ((magic[1] != 0213) && (magic[1] != 0236))) {
1082 error("bad gzip magic numbers");
1083 return -1;
1086 /* We only support method #8, DEFLATED */
1087 if (method != 8) {
1088 error("internal error, invalid method");
1089 return -1;
1092 flags = (uch)get_byte();
1093 if ((flags & ENCRYPTED) != 0) {
1094 error("Input is encrypted\n");
1095 return -1;
1097 if ((flags & CONTINUATION) != 0) {
1098 error("Multi part input\n");
1099 return -1;
1101 if ((flags & RESERVED) != 0) {
1102 error("Input has invalid flags\n");
1103 return -1;
1105 (ulg)get_byte(); /* Get timestamp */
1106 ((ulg)get_byte()) << 8;
1107 ((ulg)get_byte()) << 16;
1108 ((ulg)get_byte()) << 24;
1110 (void)get_byte(); /* Ignore extra flags for the moment */
1111 (void)get_byte(); /* Ignore OS type for the moment */
1113 if ((flags & EXTRA_FIELD) != 0) {
1114 unsigned len = (unsigned)get_byte();
1115 len |= ((unsigned)get_byte())<<8;
1116 while (len--) (void)get_byte();
1119 /* Get original file name if it was truncated */
1120 if ((flags & ORIG_NAME) != 0) {
1121 /* Discard the old name */
1122 while (get_byte() != 0) /* null */ ;
1125 /* Discard file comment if any */
1126 if ((flags & COMMENT) != 0) {
1127 while (get_byte() != 0) /* null */ ;
1130 /* Decompress */
1131 if ((res = inflate())) {
1132 switch (res) {
1133 case 0:
1134 break;
1135 case 1:
1136 error("invalid compressed format (err=1)");
1137 break;
1138 case 2:
1139 error("invalid compressed format (err=2)");
1140 break;
1141 case 3:
1142 error("out of memory");
1143 break;
1144 default:
1145 error("invalid compressed format (other)");
1147 return -1;
1150 /* Get the crc and original length */
1151 /* crc32 (see algorithm.doc)
1152 * uncompressed input size modulo 2^32
1154 orig_crc = (ulg) get_byte();
1155 orig_crc |= (ulg) get_byte() << 8;
1156 orig_crc |= (ulg) get_byte() << 16;
1157 orig_crc |= (ulg) get_byte() << 24;
1159 orig_len = (ulg) get_byte();
1160 orig_len |= (ulg) get_byte() << 8;
1161 orig_len |= (ulg) get_byte() << 16;
1162 orig_len |= (ulg) get_byte() << 24;
1164 /* Validate decompression */
1165 if (orig_crc != CRC_VALUE) {
1166 error("crc error");
1167 return -1;
1169 if (orig_len != bytes_out) {
1170 error("length error");
1171 return -1;
1173 return 0;