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@fluxnic.net>, 1999/04/14 :
11 * Little mods for all variable to reside either into rodata or bss segments
12 * by marking constant variables with 'const' and initializing all the others
13 * at run-time only. This allows for the kernel uncompressor to run
14 * directly from Flash or ROM memory on embedded systems.
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.
105 #include <linux/compiler.h>
108 static char rcsid[] = "#Id: inflate.c,v 0.14 1993/06/10 13:27:04 jloup Exp #";
113 #if defined(STDC_HEADERS) || defined(HAVE_STDLIB_H)
114 # include <sys/types.h>
128 /* Huffman code lookup table entry--this entry is four bytes for machines
129 that have 16-bit pointers (e.g. PC's in the small or medium model).
130 Valid extra bits are 0..13. e == 15 is EOB (end of block), e == 16
131 means that v is a literal, 16 < e < 32 means that v is a pointer to
132 the next table, which codes e - 16 bits, and lastly e == 99 indicates
133 an unused code. If a code with e == 99 is looked up, this implies an
134 error in the data. */
136 uch e; /* number of extra bits or operation */
137 uch b; /* number of bits in this code or subcode */
139 ush n; /* literal, length base, or distance base */
140 struct huft *t; /* pointer to next level of table */
145 /* Function prototypes */
146 STATIC int INIT huft_build OF((unsigned *, unsigned, unsigned,
147 const ush *, const ush *, struct huft **, int *));
148 STATIC int INIT huft_free OF((struct huft *));
149 STATIC int INIT inflate_codes OF((struct huft *, struct huft *, int, int));
150 STATIC int INIT inflate_stored OF((void));
151 STATIC int INIT inflate_fixed OF((void));
152 STATIC int INIT inflate_dynamic OF((void));
153 STATIC int INIT inflate_block OF((int *));
154 STATIC int INIT inflate OF((void));
157 /* The inflate algorithm uses a sliding 32 K byte window on the uncompressed
158 stream to find repeated byte strings. This is implemented here as a
159 circular buffer. The index is updated simply by incrementing and then
160 ANDing with 0x7fff (32K-1). */
161 /* It is left to other modules to supply the 32 K area. It is assumed
162 to be usable as if it were declared "uch slide[32768];" or as just
163 "uch *slide;" and then malloc'ed in the latter case. The definition
164 must be in unzip.h, included above. */
165 /* unsigned wp; current position in slide */
167 #define flush_output(w) (wp=(w),flush_window())
169 /* Tables for deflate from PKZIP's appnote.txt. */
170 static const unsigned border[] = { /* Order of the bit length code lengths */
171 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
172 static const ush cplens[] = { /* Copy lengths for literal codes 257..285 */
173 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
174 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
175 /* note: see note #13 above about the 258 in this list. */
176 static const ush cplext[] = { /* Extra bits for literal codes 257..285 */
177 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
178 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */
179 static const ush cpdist[] = { /* Copy offsets for distance codes 0..29 */
180 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
181 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
182 8193, 12289, 16385, 24577};
183 static const ush cpdext[] = { /* Extra bits for distance codes */
184 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
185 7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
190 /* Macros for inflate() bit peeking and grabbing.
194 x = b & mask_bits[j];
197 where NEEDBITS makes sure that b has at least j bits in it, and
198 DUMPBITS removes the bits from b. The macros use the variable k
199 for the number of bits in b. Normally, b and k are register
200 variables for speed, and are initialized at the beginning of a
201 routine that uses these macros from a global bit buffer and count.
203 If we assume that EOB will be the longest code, then we will never
204 ask for bits with NEEDBITS that are beyond the end of the stream.
205 So, NEEDBITS should not read any more bytes than are needed to
206 meet the request. Then no bytes need to be "returned" to the buffer
207 at the end of the last block.
209 However, this assumption is not true for fixed blocks--the EOB code
210 is 7 bits, but the other literal/length codes can be 8 or 9 bits.
211 (The EOB code is shorter than other codes because fixed blocks are
212 generally short. So, while a block always has an EOB, many other
213 literal/length codes have a significantly lower probability of
214 showing up at all.) However, by making the first table have a
215 lookup of seven bits, the EOB code will be found in that first
216 lookup, and so will not require that too many bits be pulled from
220 STATIC ulg bb; /* bit buffer */
221 STATIC unsigned bk; /* bits in bit buffer */
223 STATIC const ush mask_bits[] = {
225 0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff,
226 0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff
229 #define NEXTBYTE() ({ int v = get_byte(); if (v < 0) goto underrun; (uch)v; })
230 #define NEEDBITS(n) {while(k<(n)){b|=((ulg)NEXTBYTE())<<k;k+=8;}}
231 #define DUMPBITS(n) {b>>=(n);k-=(n);}
233 #ifndef NO_INFLATE_MALLOC
234 /* A trivial malloc implementation, adapted from
235 * malloc by Hannu Savolainen 1993 and Matthias Urlichs 1994
238 static unsigned long malloc_ptr;
239 static int malloc_count;
241 static void *malloc(int size)
246 error("Malloc error");
248 malloc_ptr = free_mem_ptr;
250 malloc_ptr = (malloc_ptr + 3) & ~3; /* Align */
252 p = (void *)malloc_ptr;
255 if (free_mem_end_ptr && malloc_ptr >= free_mem_end_ptr)
256 error("Out of memory");
262 static void free(void *where)
266 malloc_ptr = free_mem_ptr;
269 #define malloc(a) kmalloc(a, GFP_KERNEL)
270 #define free(a) kfree(a)
274 Huffman code decoding is performed using a multi-level table lookup.
275 The fastest way to decode is to simply build a lookup table whose
276 size is determined by the longest code. However, the time it takes
277 to build this table can also be a factor if the data being decoded
278 is not very long. The most common codes are necessarily the
279 shortest codes, so those codes dominate the decoding time, and hence
280 the speed. The idea is you can have a shorter table that decodes the
281 shorter, more probable codes, and then point to subsidiary tables for
282 the longer codes. The time it costs to decode the longer codes is
283 then traded against the time it takes to make longer tables.
285 This results of this trade are in the variables lbits and dbits
286 below. lbits is the number of bits the first level table for literal/
287 length codes can decode in one step, and dbits is the same thing for
288 the distance codes. Subsequent tables are also less than or equal to
289 those sizes. These values may be adjusted either when all of the
290 codes are shorter than that, in which case the longest code length in
291 bits is used, or when the shortest code is *longer* than the requested
292 table size, in which case the length of the shortest code in bits is
295 There are two different values for the two tables, since they code a
296 different number of possibilities each. The literal/length table
297 codes 286 possible values, or in a flat code, a little over eight
298 bits. The distance table codes 30 possible values, or a little less
299 than five bits, flat. The optimum values for speed end up being
300 about one bit more than those, so lbits is 8+1 and dbits is 5+1.
301 The optimum values may differ though from machine to machine, and
302 possibly even between compilers. Your mileage may vary.
306 STATIC const int lbits = 9; /* bits in base literal/length lookup table */
307 STATIC const int dbits = 6; /* bits in base distance lookup table */
310 /* If BMAX needs to be larger than 16, then h and x[] should be ulg. */
311 #define BMAX 16 /* maximum bit length of any code (16 for explode) */
312 #define N_MAX 288 /* maximum number of codes in any set */
315 STATIC unsigned hufts; /* track memory usage */
318 STATIC int INIT huft_build(
319 unsigned *b, /* code lengths in bits (all assumed <= BMAX) */
320 unsigned n, /* number of codes (assumed <= N_MAX) */
321 unsigned s, /* number of simple-valued codes (0..s-1) */
322 const ush *d, /* list of base values for non-simple codes */
323 const ush *e, /* list of extra bits for non-simple codes */
324 struct huft **t, /* result: starting table */
325 int *m /* maximum lookup bits, returns actual */
327 /* Given a list of code lengths and a maximum table size, make a set of
328 tables to decode that set of codes. Return zero on success, one if
329 the given code set is incomplete (the tables are still built in this
330 case), two if the input is invalid (all zero length codes or an
331 oversubscribed set of lengths), and three if not enough memory. */
333 unsigned a; /* counter for codes of length k */
334 unsigned f; /* i repeats in table every f entries */
335 int g; /* maximum code length */
336 int h; /* table level */
337 register unsigned i; /* counter, current code */
338 register unsigned j; /* counter */
339 register int k; /* number of bits in current code */
340 int l; /* bits per table (returned in m) */
341 register unsigned *p; /* pointer into c[], b[], or v[] */
342 register struct huft *q; /* points to current table */
343 struct huft r; /* table entry for structure assignment */
344 register int w; /* bits before this table == (l * h) */
345 unsigned *xp; /* pointer into x */
346 int y; /* number of dummy codes added */
347 unsigned z; /* number of entries in current table */
349 unsigned c[BMAX+1]; /* bit length count table */
350 struct huft *u[BMAX]; /* table stack */
351 unsigned v[N_MAX]; /* values in order of bit length */
352 unsigned x[BMAX+1]; /* bit offsets, then code stack */
360 stk = malloc(sizeof(*stk));
362 return 3; /* out of memory */
369 /* Generate counts for each bit length */
370 memzero(stk->c, sizeof(stk->c));
373 Tracecv(*p, (stderr, (n-i >= ' ' && n-i <= '~' ? "%c %d\n" : "0x%x %d\n"),
375 c[*p]++; /* assume all entries <= BMAX */
376 p++; /* Can't combine with above line (Solaris bug) */
378 if (c[0] == n) /* null input--all zero length codes */
380 *t = (struct huft *)NULL;
388 /* Find minimum and maximum length, bound *m by those */
390 for (j = 1; j <= BMAX; j++)
393 k = j; /* minimum code length */
396 for (i = BMAX; i; i--)
399 g = i; /* maximum code length */
406 /* Adjust last length count to fill out codes, if needed */
407 for (y = 1 << j; j < i; j++, y <<= 1)
408 if ((y -= c[j]) < 0) {
409 ret = 2; /* bad input: more codes than bits */
412 if ((y -= c[i]) < 0) {
420 /* Generate starting offsets into the value table for each length */
422 p = c + 1; xp = x + 2;
423 while (--i) { /* note that i == g from above */
429 /* Make a table of values in order of bit lengths */
435 n = x[g]; /* set n to length of v */
439 /* Generate the Huffman codes and for each, make the table entries */
440 x[0] = i = 0; /* first Huffman code is zero */
441 p = v; /* grab values in bit order */
442 h = -1; /* no tables yet--level -1 */
443 w = -l; /* bits decoded == (l * h) */
444 u[0] = (struct huft *)NULL; /* just to keep compilers happy */
445 q = (struct huft *)NULL; /* ditto */
449 /* go through the bit lengths (k already is bits in shortest code) */
457 /* here i is the Huffman code of length k bits for value *p */
458 /* make tables up to required level */
463 w += l; /* previous table always l bits */
465 /* compute minimum size table less than or equal to l bits */
466 z = (z = g - w) > (unsigned)l ? l : z; /* upper limit on table size */
467 if ((f = 1 << (j = k - w)) > a + 1) /* try a k-w bit table */
468 { /* too few codes for k-w bit table */
470 f -= a + 1; /* deduct codes from patterns left */
473 while (++j < z) /* try smaller tables up to z bits */
475 if ((f <<= 1) <= *++xp)
476 break; /* enough codes to use up j bits */
477 f -= *xp; /* else deduct codes from patterns */
481 z = 1 << j; /* table entries for j-bit table */
483 /* allocate and link in new table */
484 if ((q = (struct huft *)malloc((z + 1)*sizeof(struct huft))) ==
489 ret = 3; /* not enough memory */
493 hufts += z + 1; /* track memory usage */
494 *t = q + 1; /* link to list for huft_free() */
495 *(t = &(q->v.t)) = (struct huft *)NULL;
496 u[h] = ++q; /* table starts after link */
499 /* connect to last table, if there is one */
502 x[h] = i; /* save pattern for backing up */
503 r.b = (uch)l; /* bits to dump before this table */
504 r.e = (uch)(16 + j); /* bits in this table */
505 r.v.t = q; /* pointer to this table */
506 j = i >> (w - l); /* (get around Turbo C bug) */
507 u[h-1][j] = r; /* connect to last table */
513 /* set up table entry in r */
516 r.e = 99; /* out of values--invalid code */
519 r.e = (uch)(*p < 256 ? 16 : 15); /* 256 is end-of-block code */
520 r.v.n = (ush)(*p); /* simple code is just the value */
521 p++; /* one compiler does not like *p++ */
525 r.e = (uch)e[*p - s]; /* non-simple--look up in lists */
530 /* fill code-like entries with r */
532 for (j = i >> w; j < z; j += f)
535 /* backwards increment the k-bit code i */
536 for (j = 1 << (k - 1); i & j; j >>= 1)
540 /* backup over finished tables */
541 while ((i & ((1 << w) - 1)) != x[h])
543 h--; /* don't need to update q */
553 /* Return true (1) if we were given an incomplete table */
554 ret = y != 0 && g != 1;
563 STATIC int INIT huft_free(
564 struct huft *t /* table to free */
566 /* Free the malloc'ed tables built by huft_build(), which makes a linked
567 list of the tables it made, with the links in a dummy first entry of
570 register struct huft *p, *q;
573 /* Go through linked list, freeing from the malloced (t[-1]) address. */
575 while (p != (struct huft *)NULL)
585 STATIC int INIT inflate_codes(
586 struct huft *tl, /* literal/length decoder tables */
587 struct huft *td, /* distance decoder tables */
588 int bl, /* number of bits decoded by tl[] */
589 int bd /* number of bits decoded by td[] */
591 /* inflate (decompress) the codes in a deflated (compressed) block.
592 Return an error code or zero if it all goes ok. */
594 register unsigned e; /* table entry flag/number of extra bits */
595 unsigned n, d; /* length and index for copy */
596 unsigned w; /* current window position */
597 struct huft *t; /* pointer to table entry */
598 unsigned ml, md; /* masks for bl and bd bits */
599 register ulg b; /* bit buffer */
600 register unsigned k; /* number of bits in bit buffer */
603 /* make local copies of globals */
604 b = bb; /* initialize bit buffer */
606 w = wp; /* initialize window position */
608 /* inflate the coded data */
609 ml = mask_bits[bl]; /* precompute masks for speed */
611 for (;;) /* do until end of block */
613 NEEDBITS((unsigned)bl)
614 if ((e = (t = tl + ((unsigned)b & ml))->e) > 16)
621 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
623 if (e == 16) /* then it's a literal */
625 slide[w++] = (uch)t->v.n;
626 Tracevv((stderr, "%c", slide[w-1]));
633 else /* it's an EOB or a length */
635 /* exit if end of block */
639 /* get length of block to copy */
641 n = t->v.n + ((unsigned)b & mask_bits[e]);
644 /* decode distance of block to copy */
645 NEEDBITS((unsigned)bd)
646 if ((e = (t = td + ((unsigned)b & md))->e) > 16)
653 } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
656 d = w - t->v.n - ((unsigned)b & mask_bits[e]);
658 Tracevv((stderr,"\\[%d,%d]", w-d, n));
662 n -= (e = (e = WSIZE - ((d &= WSIZE-1) > w ? d : w)) > n ? n : e);
663 #if !defined(NOMEMCPY) && !defined(DEBUG)
664 if (w - d >= e) /* (this test assumes unsigned comparison) */
666 memcpy(slide + w, slide + d, e);
670 else /* do it slow to avoid memcpy() overlap */
671 #endif /* !NOMEMCPY */
673 slide[w++] = slide[d++];
674 Tracevv((stderr, "%c", slide[w-1]));
686 /* restore the globals from the locals */
687 wp = w; /* restore global window pointer */
688 bb = b; /* restore global bit buffer */
695 return 4; /* Input underrun */
700 STATIC int INIT inflate_stored(void)
701 /* "decompress" an inflated type 0 (stored) block. */
703 unsigned n; /* number of bytes in block */
704 unsigned w; /* current window position */
705 register ulg b; /* bit buffer */
706 register unsigned k; /* number of bits in bit buffer */
710 /* make local copies of globals */
711 b = bb; /* initialize bit buffer */
713 w = wp; /* initialize window position */
716 /* go to byte boundary */
721 /* get the length and its complement */
723 n = ((unsigned)b & 0xffff);
726 if (n != (unsigned)((~b) & 0xffff))
727 return 1; /* error in compressed data */
731 /* read and output the compressed data */
745 /* restore the globals from the locals */
746 wp = w; /* restore global window pointer */
747 bb = b; /* restore global bit buffer */
754 return 4; /* Input underrun */
759 * We use `noinline' here to prevent gcc-3.5 from using too much stack space
761 STATIC int noinline INIT inflate_fixed(void)
762 /* decompress an inflated type 1 (fixed Huffman codes) block. We should
763 either replace this with a custom decoder, or at least precompute the
766 int i; /* temporary variable */
767 struct huft *tl; /* literal/length code table */
768 struct huft *td; /* distance code table */
769 int bl; /* lookup bits for tl */
770 int bd; /* lookup bits for td */
771 unsigned *l; /* length list for huft_build */
775 l = malloc(sizeof(*l) * 288);
777 return 3; /* out of memory */
779 /* set up literal table */
780 for (i = 0; i < 144; i++)
786 for (; i < 288; i++) /* make a complete, but wrong code set */
789 if ((i = huft_build(l, 288, 257, cplens, cplext, &tl, &bl)) != 0) {
794 /* set up distance table */
795 for (i = 0; i < 30; i++) /* make an incomplete code set */
798 if ((i = huft_build(l, 30, 0, cpdist, cpdext, &td, &bd)) > 1)
808 /* decompress until an end-of-block code */
809 if (inflate_codes(tl, td, bl, bd)) {
814 /* free the decoding tables, return */
823 * We use `noinline' here to prevent gcc-3.5 from using too much stack space
825 STATIC int noinline INIT inflate_dynamic(void)
826 /* decompress an inflated type 2 (dynamic Huffman codes) block. */
828 int i; /* temporary variables */
830 unsigned l; /* last length */
831 unsigned m; /* mask for bit lengths table */
832 unsigned n; /* number of lengths to get */
833 struct huft *tl; /* literal/length code table */
834 struct huft *td; /* distance code table */
835 int bl; /* lookup bits for tl */
836 int bd; /* lookup bits for td */
837 unsigned nb; /* number of bit length codes */
838 unsigned nl; /* number of literal/length codes */
839 unsigned nd; /* number of distance codes */
840 unsigned *ll; /* literal/length and distance code lengths */
841 register ulg b; /* bit buffer */
842 register unsigned k; /* number of bits in bit buffer */
847 #ifdef PKZIP_BUG_WORKAROUND
848 ll = malloc(sizeof(*ll) * (288+32)); /* literal/length and distance code lengths */
850 ll = malloc(sizeof(*ll) * (286+30)); /* literal/length and distance code lengths */
856 /* make local bit buffer */
861 /* read in table lengths */
863 nl = 257 + ((unsigned)b & 0x1f); /* number of literal/length codes */
866 nd = 1 + ((unsigned)b & 0x1f); /* number of distance codes */
869 nb = 4 + ((unsigned)b & 0xf); /* number of bit length codes */
871 #ifdef PKZIP_BUG_WORKAROUND
872 if (nl > 288 || nd > 32)
874 if (nl > 286 || nd > 30)
877 ret = 1; /* bad lengths */
883 /* read in bit-length-code lengths */
884 for (j = 0; j < nb; j++)
887 ll[border[j]] = (unsigned)b & 7;
895 /* build decoding table for trees--single level, 7 bit lookup */
897 if ((i = huft_build(ll, 19, 19, NULL, NULL, &tl, &bl)) != 0)
901 ret = i; /* incomplete code set */
907 /* read in literal and distance code lengths */
911 while ((unsigned)i < n)
913 NEEDBITS((unsigned)bl)
914 j = (td = tl + ((unsigned)b & m))->b;
917 if (j < 16) /* length of code in bits (0..15) */
918 ll[i++] = l = j; /* save last length in l */
919 else if (j == 16) /* repeat last length 3 to 6 times */
922 j = 3 + ((unsigned)b & 3);
924 if ((unsigned)i + j > n) {
931 else if (j == 17) /* 3 to 10 zero length codes */
934 j = 3 + ((unsigned)b & 7);
936 if ((unsigned)i + j > n) {
944 else /* j == 18: 11 to 138 zero length codes */
947 j = 11 + ((unsigned)b & 0x7f);
949 if ((unsigned)i + j > n) {
961 /* free decoding table for trees */
966 /* restore the global bit buffer */
972 /* build the decoding tables for literal/length and distance codes */
974 if ((i = huft_build(ll, nl, 257, cplens, cplext, &tl, &bl)) != 0)
978 error("incomplete literal tree");
981 ret = i; /* incomplete code set */
986 if ((i = huft_build(ll + nl, nd, 0, cpdist, cpdext, &td, &bd)) != 0)
990 error("incomplete distance tree");
991 #ifdef PKZIP_BUG_WORKAROUND
998 ret = i; /* incomplete code set */
1005 /* decompress until an end-of-block code */
1006 if (inflate_codes(tl, td, bl, bd)) {
1013 /* free the decoding tables, return */
1024 ret = 4; /* Input underrun */
1030 STATIC int INIT inflate_block(
1031 int *e /* last block flag */
1033 /* decompress an inflated block */
1035 unsigned t; /* block type */
1036 register ulg b; /* bit buffer */
1037 register unsigned k; /* number of bits in bit buffer */
1041 /* make local bit buffer */
1046 /* read in last block bit */
1052 /* read in block type */
1054 t = (unsigned)b & 3;
1058 /* restore the global bit buffer */
1062 /* inflate that block type */
1064 return inflate_dynamic();
1066 return inflate_stored();
1068 return inflate_fixed();
1072 /* bad block type */
1076 return 4; /* Input underrun */
1081 STATIC int INIT inflate(void)
1082 /* decompress an inflated entry */
1084 int e; /* last block flag */
1085 int r; /* result code */
1086 unsigned h; /* maximum struct huft's malloc'ed */
1088 /* initialize window, bit buffer */
1094 /* decompress until the last block */
1098 #ifdef ARCH_HAS_DECOMP_WDOG
1101 r = inflate_block(&e);
1108 /* Undo too much lookahead. The next read will be byte aligned so we
1109 * can discard unused bits in the last meaningful byte.
1116 /* flush out slide */
1120 /* return success */
1122 fprintf(stderr, "<%u> ", h);
1127 /**********************************************************************
1129 * The following are support routines for inflate.c
1131 **********************************************************************/
1133 static ulg crc_32_tab[256];
1134 static ulg crc; /* initialized in makecrc() so it'll reside in bss */
1135 #define CRC_VALUE (crc ^ 0xffffffffUL)
1138 * Code to compute the CRC-32 table. Borrowed from
1139 * gzip-1.0.3/makecrc.c.
1145 /* Not copyrighted 1990 Mark Adler */
1147 unsigned long c; /* crc shift register */
1148 unsigned long e; /* polynomial exclusive-or pattern */
1149 int i; /* counter for all possible eight bit values */
1150 int k; /* byte being shifted into crc apparatus */
1152 /* terms of polynomial defining this crc (except x^32): */
1153 static const int p[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26};
1155 /* Make exclusive-or pattern from polynomial */
1157 for (i = 0; i < sizeof(p)/sizeof(int); i++)
1158 e |= 1L << (31 - p[i]);
1162 for (i = 1; i < 256; i++)
1165 for (k = i | 256; k != 1; k >>= 1)
1167 c = c & 1 ? (c >> 1) ^ e : c >> 1;
1174 /* this is initialized here so this code could reside in ROM */
1175 crc = (ulg)0xffffffffUL; /* shift register contents */
1178 /* gzip flag byte */
1179 #define ASCII_FLAG 0x01 /* bit 0 set: file probably ASCII text */
1180 #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
1181 #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
1182 #define ORIG_NAME 0x08 /* bit 3 set: original file name present */
1183 #define COMMENT 0x10 /* bit 4 set: file comment present */
1184 #define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */
1185 #define RESERVED 0xC0 /* bit 6,7: reserved */
1188 * Do the uncompression!
1190 static int INIT gunzip(void)
1193 unsigned char magic[2]; /* magic header */
1195 ulg orig_crc = 0; /* original crc */
1196 ulg orig_len = 0; /* original uncompressed length */
1199 magic[0] = NEXTBYTE();
1200 magic[1] = NEXTBYTE();
1201 method = NEXTBYTE();
1203 if (magic[0] != 037 ||
1204 ((magic[1] != 0213) && (magic[1] != 0236))) {
1205 error("bad gzip magic numbers");
1209 /* We only support method #8, DEFLATED */
1211 error("internal error, invalid method");
1215 flags = (uch)get_byte();
1216 if ((flags & ENCRYPTED) != 0) {
1217 error("Input is encrypted");
1220 if ((flags & CONTINUATION) != 0) {
1221 error("Multi part input");
1224 if ((flags & RESERVED) != 0) {
1225 error("Input has invalid flags");
1228 NEXTBYTE(); /* Get timestamp */
1233 (void)NEXTBYTE(); /* Ignore extra flags for the moment */
1234 (void)NEXTBYTE(); /* Ignore OS type for the moment */
1236 if ((flags & EXTRA_FIELD) != 0) {
1237 unsigned len = (unsigned)NEXTBYTE();
1238 len |= ((unsigned)NEXTBYTE())<<8;
1239 while (len--) (void)NEXTBYTE();
1242 /* Get original file name if it was truncated */
1243 if ((flags & ORIG_NAME) != 0) {
1244 /* Discard the old name */
1245 while (NEXTBYTE() != 0) /* null */ ;
1248 /* Discard file comment if any */
1249 if ((flags & COMMENT) != 0) {
1250 while (NEXTBYTE() != 0) /* null */ ;
1254 if ((res = inflate())) {
1259 error("invalid compressed format (err=1)");
1262 error("invalid compressed format (err=2)");
1265 error("out of memory");
1268 error("out of input data");
1271 error("invalid compressed format (other)");
1276 /* Get the crc and original length */
1277 /* crc32 (see algorithm.doc)
1278 * uncompressed input size modulo 2^32
1280 orig_crc = (ulg) NEXTBYTE();
1281 orig_crc |= (ulg) NEXTBYTE() << 8;
1282 orig_crc |= (ulg) NEXTBYTE() << 16;
1283 orig_crc |= (ulg) NEXTBYTE() << 24;
1285 orig_len = (ulg) NEXTBYTE();
1286 orig_len |= (ulg) NEXTBYTE() << 8;
1287 orig_len |= (ulg) NEXTBYTE() << 16;
1288 orig_len |= (ulg) NEXTBYTE() << 24;
1290 /* Validate decompression */
1291 if (orig_crc != CRC_VALUE) {
1295 if (orig_len != bytes_out) {
1296 error("length error");
1301 underrun: /* NEXTBYTE() goto's here if needed */
1302 error("out of input data");