3 /* inflate.c -- Not copyrighted 1992 by Mark Adler
4 version c10p1, 10 January 1993 */
7 * Adapted for booting Linux by Hannu Savolainen 1993
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
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
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.
107 static char rcsid[] = "#Id: inflate.c,v 0.14 1993/06/10 13:27:04 jloup Exp #";
112 #if defined(STDC_HEADERS) || defined(HAVE_STDLIB_H)
113 # include <sys/types.h>
119 #endif /* !FUNC_STATIC */
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. */
131 uch e; /* number of extra bits or operation */
132 uch b; /* number of bits in this code or subcode */
134 ush n; /* literal, length base, or distance base */
135 struct huft *t; /* pointer to next level of table */
140 /* Function prototypes */
141 FUNC_STATIC int huft_build OF((unsigned *, unsigned, unsigned,
142 const ush *, const ush *, struct huft **, int *));
143 FUNC_STATIC int huft_free OF((struct huft *));
144 FUNC_STATIC int inflate_codes OF((struct huft *, struct huft *, int, int));
145 FUNC_STATIC int inflate_stored OF((void));
146 FUNC_STATIC int inflate_fixed OF((void));
147 FUNC_STATIC int inflate_dynamic OF((void));
148 FUNC_STATIC int inflate_block OF((int *));
149 FUNC_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 */
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,
185 /* Macros for inflate() bit peeking and grabbing.
189 x = b & mask_bits[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
215 FUNC_STATIC ulg bb; /* bit buffer */
216 FUNC_STATIC unsigned bk; /* bits in bit buffer */
218 FUNC_STATIC const ush mask_bits[] = {
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
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 FUNC_STATIC const int lbits = 9; /* bits in base literal/length lookup table */
263 FUNC_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 FUNC_STATIC unsigned hufts; /* track memory usage */
274 FUNC_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 */
310 /* Generate counts for each bit length */
311 memzero(c, sizeof(c));
315 Tracecv(*p, (stderr, (n-i >= ' ' && n-i <= '~' ? "%c %d\n" : "0x%x %d\n"),
317 c[*p]++; /* assume all entries <= BMAX */
318 p++; /* Can't combine with above line (Solaris bug) */
320 if (c[0] == n) /* null input--all zero length codes */
322 *t = (struct huft *)NULL;
329 /* Find minimum and maximum length, bound *m by those */
331 for (j = 1; j <= BMAX; j++)
334 k = j; /* minimum code length */
337 for (i = BMAX; i; i--)
340 g = i; /* maximum code length */
347 /* Adjust last length count to fill out codes, if needed */
348 for (y = 1 << j; j < i; j++, y <<= 1)
350 return 2; /* bad input: more codes than bits */
357 /* Generate starting offsets into the value table for each length */
359 p = c + 1; xp = x + 2;
360 while (--i) { /* note that i == g from above */
366 /* Make a table of values in order of bit lengths */
375 /* Generate the Huffman codes and for each, make the table entries */
376 x[0] = i = 0; /* first Huffman code is zero */
377 p = v; /* grab values in bit order */
378 h = -1; /* no tables yet--level -1 */
379 w = -l; /* bits decoded == (l * h) */
380 u[0] = (struct huft *)NULL; /* just to keep compilers happy */
381 q = (struct huft *)NULL; /* ditto */
385 /* go through the bit lengths (k already is bits in shortest code) */
393 /* here i is the Huffman code of length k bits for value *p */
394 /* make tables up to required level */
399 w += l; /* previous table always l bits */
401 /* compute minimum size table less than or equal to l bits */
402 z = (z = g - w) > (unsigned)l ? l : z; /* upper limit on table size */
403 if ((f = 1 << (j = k - w)) > a + 1) /* try a k-w bit table */
404 { /* too few codes for k-w bit table */
406 f -= a + 1; /* deduct codes from patterns left */
408 while (++j < z) /* try smaller tables up to z bits */
410 if ((f <<= 1) <= *++xp)
411 break; /* enough codes to use up j bits */
412 f -= *xp; /* else deduct codes from patterns */
416 z = 1 << j; /* table entries for j-bit table */
418 /* allocate and link in new table */
419 if ((q = (struct huft *)gzip_malloc((z + 1)*sizeof(struct huft))) ==
424 return 3; /* not enough memory */
427 hufts += z + 1; /* track memory usage */
428 *t = q + 1; /* link to list for huft_free() */
429 *(t = &(q->v.t)) = (struct huft *)NULL;
430 u[h] = ++q; /* table starts after link */
433 /* connect to last table, if there is one */
436 x[h] = i; /* save pattern for backing up */
437 r.b = (uch)l; /* bits to dump before this table */
438 r.e = (uch)(16 + j); /* bits in this table */
439 r.v.t = q; /* pointer to this table */
440 j = i >> (w - l); /* (get around Turbo C bug) */
441 u[h-1][j] = r; /* connect to last table */
447 /* set up table entry in r */
450 r.e = 99; /* out of values--invalid code */
453 r.e = (uch)(*p < 256 ? 16 : 15); /* 256 is end-of-block code */
454 r.v.n = (ush)(*p); /* simple code is just the value */
455 p++; /* one compiler does not like *p++ */
459 r.e = (uch)e[*p - s]; /* non-simple--look up in lists */
464 /* fill code-like entries with r */
466 for (j = i >> w; j < z; j += f)
469 /* backwards increment the k-bit code i */
470 for (j = 1 << (k - 1); i & j; j >>= 1)
474 /* backup over finished tables */
475 while ((i & ((1 << w) - 1)) != x[h])
477 h--; /* don't need to update q */
487 /* Return true (1) if we were given an incomplete table */
488 return y != 0 && g != 1;
493 FUNC_STATIC int huft_free(t)
494 struct huft *t; /* table to free */
495 /* Free the malloc'ed tables built by huft_build(), which makes a linked
496 list of the tables it made, with the links in a dummy first entry of
499 register struct huft *p, *q;
502 /* Go through linked list, freeing from the malloced (t[-1]) address. */
504 while (p != (struct huft *)NULL)
514 FUNC_STATIC int inflate_codes(tl, td, bl, bd)
515 struct huft *tl, *td; /* literal/length and distance decoder tables */
516 int bl, bd; /* number of bits decoded by tl[] and td[] */
517 /* inflate (decompress) the codes in a deflated (compressed) block.
518 Return an error code or zero if it all goes ok. */
520 register unsigned e; /* table entry flag/number of extra bits */
521 unsigned n, d; /* length and index for copy */
522 unsigned w; /* current window position */
523 struct huft *t; /* pointer to table entry */
524 unsigned ml, md; /* masks for bl and bd bits */
525 register ulg b; /* bit buffer */
526 register unsigned k; /* number of bits in bit buffer */
529 /* make local copies of globals */
530 b = bb; /* initialize bit buffer */
532 w = wp; /* initialize window position */
534 /* inflate the coded data */
535 ml = mask_bits[bl]; /* precompute masks for speed */
537 for (;;) /* do until end of block */
539 NEEDBITS((unsigned)bl)
540 if ((e = (t = tl + ((unsigned)b & ml))->e) > 16)
547 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
549 if (e == 16) /* then it's a literal */
551 slide[w++] = (uch)t->v.n;
552 Tracevv((stderr, "%c", slide[w-1]));
559 else /* it's an EOB or a length */
561 /* exit if end of block */
565 /* get length of block to copy */
567 n = t->v.n + ((unsigned)b & mask_bits[e]);
570 /* decode distance of block to copy */
571 NEEDBITS((unsigned)bd)
572 if ((e = (t = td + ((unsigned)b & md))->e) > 16)
579 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
582 d = w - t->v.n - ((unsigned)b & mask_bits[e]);
584 Tracevv((stderr,"\\[%d,%d]", w-d, n));
588 n -= (e = (e = WSIZE - ((d &= WSIZE-1) > w ? d : w)) > n ? n : e);
589 #if !defined(NOMEMCPY) && !defined(INFLATE_DEBUG)
590 if (w - d >= e) /* (this test assumes unsigned comparison) */
592 memcpy(slide + w, slide + d, e);
596 else /* do it slow to avoid memcpy() overlap */
597 #endif /* !NOMEMCPY */
599 slide[w++] = slide[d++];
600 Tracevv((stderr, "%c", slide[w-1]));
612 /* restore the globals from the locals */
613 wp = w; /* restore global window pointer */
614 bb = b; /* restore global bit buffer */
623 FUNC_STATIC int inflate_stored()
624 /* "decompress" an inflated type 0 (stored) block. */
626 unsigned n; /* number of bytes in block */
627 unsigned w; /* current window position */
628 register ulg b; /* bit buffer */
629 register unsigned k; /* number of bits in bit buffer */
633 /* make local copies of globals */
634 b = bb; /* initialize bit buffer */
636 w = wp; /* initialize window position */
639 /* go to byte boundary */
644 /* get the length and its complement */
646 n = ((unsigned)b & 0xffff);
649 if (n != (unsigned)((~b) & 0xffff))
650 return 1; /* error in compressed data */
654 /* read and output the compressed data */
668 /* restore the globals from the locals */
669 wp = w; /* restore global window pointer */
670 bb = b; /* restore global bit buffer */
679 FUNC_STATIC int inflate_fixed()
680 /* decompress an inflated type 1 (fixed Huffman codes) block. We should
681 either replace this with a custom decoder, or at least precompute the
684 int i; /* temporary variable */
685 struct huft *tl; /* literal/length code table */
686 struct huft *td; /* distance code table */
687 int bl; /* lookup bits for tl */
688 int bd; /* lookup bits for td */
689 unsigned l[288]; /* length list for huft_build */
693 /* set up literal table */
694 for (i = 0; i < 144; i++)
700 for (; i < 288; i++) /* make a complete, but wrong code set */
703 if ((i = huft_build(l, 288, 257, cplens, cplext, &tl, &bl)) != 0)
707 /* set up distance table */
708 for (i = 0; i < 30; i++) /* make an incomplete code set */
711 if ((i = huft_build(l, 30, 0, cpdist, cpdext, &td, &bd)) > 1)
720 /* decompress until an end-of-block code */
721 if (inflate_codes(tl, td, bl, bd))
725 /* free the decoding tables, return */
733 FUNC_STATIC int inflate_dynamic()
734 /* decompress an inflated type 2 (dynamic Huffman codes) block. */
736 int i; /* temporary variables */
738 unsigned l; /* last length */
739 unsigned m; /* mask for bit lengths table */
740 unsigned n; /* number of lengths to get */
741 struct huft *tl; /* literal/length code table */
742 struct huft *td; /* distance code table */
743 int bl; /* lookup bits for tl */
744 int bd; /* lookup bits for td */
745 unsigned nb; /* number of bit length codes */
746 unsigned nl; /* number of literal/length codes */
747 unsigned nd; /* number of distance codes */
748 #ifdef PKZIP_BUG_WORKAROUND
749 unsigned ll[288+32]; /* literal/length and distance code lengths */
751 unsigned ll[286+30]; /* literal/length and distance code lengths */
753 register ulg b; /* bit buffer */
754 register unsigned k; /* number of bits in bit buffer */
758 /* make local bit buffer */
763 /* read in table lengths */
765 nl = 257 + ((unsigned)b & 0x1f); /* number of literal/length codes */
768 nd = 1 + ((unsigned)b & 0x1f); /* number of distance codes */
771 nb = 4 + ((unsigned)b & 0xf); /* number of bit length codes */
773 #ifdef PKZIP_BUG_WORKAROUND
774 if (nl > 288 || nd > 32)
776 if (nl > 286 || nd > 30)
778 return 1; /* bad lengths */
782 /* read in bit-length-code lengths */
783 for (j = 0; j < nb; j++)
786 ll[border[j]] = (unsigned)b & 7;
794 /* build decoding table for trees--single level, 7 bit lookup */
796 if ((i = huft_build(ll, 19, 19, NULL, NULL, &tl, &bl)) != 0)
800 return i; /* incomplete code set */
805 /* read in literal and distance code lengths */
809 while ((unsigned)i < n)
811 NEEDBITS((unsigned)bl)
812 j = (td = tl + ((unsigned)b & m))->b;
815 if (j < 16) /* length of code in bits (0..15) */
816 ll[i++] = l = j; /* save last length in l */
817 else if (j == 16) /* repeat last length 3 to 6 times */
820 j = 3 + ((unsigned)b & 3);
822 if ((unsigned)i + j > n)
827 else if (j == 17) /* 3 to 10 zero length codes */
830 j = 3 + ((unsigned)b & 7);
832 if ((unsigned)i + j > n)
838 else /* j == 18: 11 to 138 zero length codes */
841 j = 11 + ((unsigned)b & 0x7f);
843 if ((unsigned)i + j > n)
853 /* free decoding table for trees */
858 /* restore the global bit buffer */
864 /* build the decoding tables for literal/length and distance codes */
866 if ((i = huft_build(ll, nl, 257, cplens, cplext, &tl, &bl)) != 0)
870 error(" incomplete literal tree\n");
873 return i; /* incomplete code set */
877 if ((i = huft_build(ll + nl, nd, 0, cpdist, cpdext, &td, &bd)) != 0)
881 error(" incomplete distance tree\n");
882 #ifdef PKZIP_BUG_WORKAROUND
889 return i; /* incomplete code set */
895 /* decompress until an end-of-block code */
896 if (inflate_codes(tl, td, bl, bd))
901 /* free the decoding tables, return */
911 FUNC_STATIC int inflate_block(e)
912 int *e; /* last block flag */
913 /* decompress an inflated block */
915 unsigned t; /* block type */
916 register ulg b; /* bit buffer */
917 register unsigned k; /* number of bits in bit buffer */
921 /* make local bit buffer */
926 /* read in last block bit */
932 /* read in block type */
938 /* restore the global bit buffer */
942 /* inflate that block type */
944 return inflate_dynamic();
946 return inflate_stored();
948 return inflate_fixed();
958 FUNC_STATIC int inflate()
959 /* decompress an inflated entry */
961 int e; /* last block flag */
962 int r; /* result code */
963 unsigned h; /* maximum struct huft's malloc'ed */
965 /* initialize window, bit buffer */
971 /* decompress until the last block */
975 if ((r = inflate_block(&e)) != 0) {
982 /* Undo too much lookahead. The next read will be byte aligned so we
983 * can discard unused bits in the last meaningful byte.
990 /* flush out slide */
1001 #endif /* INFLATE_DEBUG */
1005 /**********************************************************************
1007 * The following are support routines for inflate.c
1009 **********************************************************************/
1011 static ulg crc_32_tab[256];
1012 static ulg crc; /* initialized in makecrc() so it'll reside in bss */
1013 #define CRC_VALUE (crc ^ 0xffffffffUL)
1016 * Code to compute the CRC-32 table. Borrowed from
1017 * gzip-1.0.3/makecrc.c.
1023 /* Not copyrighted 1990 Mark Adler */
1025 unsigned long c; /* crc shift register */
1026 unsigned long e; /* polynomial exclusive-or pattern */
1027 int i; /* counter for all possible eight bit values */
1028 int k; /* byte being shifted into crc apparatus */
1030 /* terms of polynomial defining this crc (except x^32): */
1031 static const int p[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26};
1033 /* Make exclusive-or pattern from polynomial */
1035 for (i = 0; i < sizeof(p)/sizeof(int); i++)
1036 e |= 1L << (31 - p[i]);
1040 for (i = 1; i < 256; i++)
1043 for (k = i | 256; k != 1; k >>= 1)
1045 c = c & 1 ? (c >> 1) ^ e : c >> 1;
1052 /* this is initialized here so this code could reside in ROM */
1053 crc = (ulg)0xffffffffUL; /* shift register contents */
1056 /* gzip flag byte */
1057 #define ASCII_FLAG 0x01 /* bit 0 set: file probably ASCII text */
1058 #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
1059 #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
1060 #define ORIG_NAME 0x08 /* bit 3 set: original file name present */
1061 #define COMMENT 0x10 /* bit 4 set: file comment present */
1062 #define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */
1063 #define RESERVED 0xC0 /* bit 6,7: reserved */
1066 * check for valid gzip signature
1068 * 0 : valid gzip archive
1069 * -1: invalid gzip archive
1072 gzip_probe(uch *buf, unsigned long size)
1074 /* FIXME this is wrong here! FIXME!!!!*/
1075 if (size < 4) return -1;
1077 if (buf[0] != 037 ||
1078 ((buf[1] != 0213) && (buf[1] != 0236))) return -1;
1080 /* We only support method #8, DEFLATED */
1081 if (buf[2] != 8) return -1;
1083 if ((buf[3] & ENCRYPTED) != 0) return -1;
1085 if ((buf[3] & CONTINUATION) != 0) return -1;
1087 if ((buf[3] & RESERVED) != 0) return -1;
1094 * Do the uncompression!
1096 static int gunzip(void)
1099 unsigned char magic[2]; /* magic header */
1101 ulg orig_crc = 0; /* original crc */
1102 ulg orig_len = 0; /* original uncompressed length */
1105 magic[0] = (unsigned char)get_byte();
1106 magic[1] = (unsigned char)get_byte();
1107 method = (unsigned char)get_byte();
1109 if (magic[0] != 037 ||
1110 ((magic[1] != 0213) && (magic[1] != 0236))) {
1111 error("bad gzip magic numbers");
1115 /* We only support method #8, DEFLATED */
1117 error("internal error, invalid method");
1121 flags = (uch)get_byte();
1122 if ((flags & ENCRYPTED) != 0) {
1123 error("Input is encrypted\n");
1126 if ((flags & CONTINUATION) != 0) {
1127 error("Multi part input\n");
1130 if ((flags & RESERVED) != 0) {
1131 error("Input has invalid flags\n");
1134 (ulg)get_byte(); /* Get timestamp */
1135 ((ulg)get_byte()) << 8;
1136 ((ulg)get_byte()) << 16;
1137 ((ulg)get_byte()) << 24;
1139 (void)get_byte(); /* Ignore extra flags for the moment */
1140 (void)get_byte(); /* Ignore OS type for the moment */
1142 if ((flags & EXTRA_FIELD) != 0) {
1143 unsigned len = (unsigned)get_byte();
1144 len |= ((unsigned)get_byte())<<8;
1145 while (len--) (void)get_byte();
1148 /* Get original file name if it was truncated */
1149 if ((flags & ORIG_NAME) != 0) {
1150 /* Discard the old name */
1151 while (get_byte() != 0) /* null */ ;
1154 /* Discard file comment if any */
1155 if ((flags & COMMENT) != 0) {
1156 while (get_byte() != 0) /* null */ ;
1160 if ((res = inflate())) {
1165 error("invalid compressed format (err=1)");
1168 error("invalid compressed format (err=2)");
1171 error("out of memory");
1174 error("invalid compressed format (other)");
1179 /* Get the crc and original length */
1180 /* crc32 (see algorithm.doc)
1181 * uncompressed input size modulo 2^32
1183 orig_crc = (ulg) get_byte();
1184 orig_crc |= (ulg) get_byte() << 8;
1185 orig_crc |= (ulg) get_byte() << 16;
1186 orig_crc |= (ulg) get_byte() << 24;
1188 orig_len = (ulg) get_byte();
1189 orig_len |= (ulg) get_byte() << 8;
1190 orig_len |= (ulg) get_byte() << 16;
1191 orig_len |= (ulg) get_byte() << 24;
1193 /* Validate decompression */
1194 if (orig_crc != CRC_VALUE) {
1198 if (orig_len != bytes_out) {
1199 error("length error");