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>
106 #ifdef NO_INFLATE_MALLOC
107 #include <linux/slab.h>
111 static char rcsid
[] = "#Id: inflate.c,v 0.14 1993/06/10 13:27:04 jloup Exp #";
116 #if defined(STDC_HEADERS) || defined(HAVE_STDLIB_H)
117 # include <sys/types.h>
131 /* Huffman code lookup table entry--this entry is four bytes for machines
132 that have 16-bit pointers (e.g. PC's in the small or medium model).
133 Valid extra bits are 0..13. e == 15 is EOB (end of block), e == 16
134 means that v is a literal, 16 < e < 32 means that v is a pointer to
135 the next table, which codes e - 16 bits, and lastly e == 99 indicates
136 an unused code. If a code with e == 99 is looked up, this implies an
137 error in the data. */
139 uch e
; /* number of extra bits or operation */
140 uch b
; /* number of bits in this code or subcode */
142 ush n
; /* literal, length base, or distance base */
143 struct huft
*t
; /* pointer to next level of table */
148 /* Function prototypes */
149 STATIC
int INIT huft_build
OF((unsigned *, unsigned, unsigned,
150 const ush
*, const ush
*, struct huft
**, int *));
151 STATIC
int INIT huft_free
OF((struct huft
*));
152 STATIC
int INIT inflate_codes
OF((struct huft
*, struct huft
*, int, int));
153 STATIC
int INIT inflate_stored
OF((void));
154 STATIC
int INIT inflate_fixed
OF((void));
155 STATIC
int INIT inflate_dynamic
OF((void));
156 STATIC
int INIT inflate_block
OF((int *));
157 STATIC
int INIT inflate
OF((void));
160 /* The inflate algorithm uses a sliding 32 K byte window on the uncompressed
161 stream to find repeated byte strings. This is implemented here as a
162 circular buffer. The index is updated simply by incrementing and then
163 ANDing with 0x7fff (32K-1). */
164 /* It is left to other modules to supply the 32 K area. It is assumed
165 to be usable as if it were declared "uch slide[32768];" or as just
166 "uch *slide;" and then malloc'ed in the latter case. The definition
167 must be in unzip.h, included above. */
168 /* unsigned wp; current position in slide */
170 #define flush_output(w) (wp=(w),flush_window())
172 /* Tables for deflate from PKZIP's appnote.txt. */
173 static const unsigned border
[] = { /* Order of the bit length code lengths */
174 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
175 static const ush cplens
[] = { /* Copy lengths for literal codes 257..285 */
176 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
177 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
178 /* note: see note #13 above about the 258 in this list. */
179 static const ush cplext
[] = { /* Extra bits for literal codes 257..285 */
180 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
181 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */
182 static const ush cpdist
[] = { /* Copy offsets for distance codes 0..29 */
183 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
184 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
185 8193, 12289, 16385, 24577};
186 static const ush cpdext
[] = { /* Extra bits for distance codes */
187 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
188 7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
193 /* Macros for inflate() bit peeking and grabbing.
197 x = b & mask_bits[j];
200 where NEEDBITS makes sure that b has at least j bits in it, and
201 DUMPBITS removes the bits from b. The macros use the variable k
202 for the number of bits in b. Normally, b and k are register
203 variables for speed, and are initialized at the beginning of a
204 routine that uses these macros from a global bit buffer and count.
206 If we assume that EOB will be the longest code, then we will never
207 ask for bits with NEEDBITS that are beyond the end of the stream.
208 So, NEEDBITS should not read any more bytes than are needed to
209 meet the request. Then no bytes need to be "returned" to the buffer
210 at the end of the last block.
212 However, this assumption is not true for fixed blocks--the EOB code
213 is 7 bits, but the other literal/length codes can be 8 or 9 bits.
214 (The EOB code is shorter than other codes because fixed blocks are
215 generally short. So, while a block always has an EOB, many other
216 literal/length codes have a significantly lower probability of
217 showing up at all.) However, by making the first table have a
218 lookup of seven bits, the EOB code will be found in that first
219 lookup, and so will not require that too many bits be pulled from
223 STATIC ulg bb
; /* bit buffer */
224 STATIC
unsigned bk
; /* bits in bit buffer */
226 STATIC
const ush mask_bits
[] = {
228 0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff,
229 0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff
232 #define NEXTBYTE() ({ int v = get_byte(); if (v < 0) goto underrun; (uch)v; })
233 #define NEEDBITS(n) {while(k<(n)){b|=((ulg)NEXTBYTE())<<k;k+=8;}}
234 #define DUMPBITS(n) {b>>=(n);k-=(n);}
236 #ifndef NO_INFLATE_MALLOC
237 /* A trivial malloc implementation, adapted from
238 * malloc by Hannu Savolainen 1993 and Matthias Urlichs 1994
241 static unsigned long malloc_ptr
;
242 static int malloc_count
;
244 static void *malloc(int size
)
249 error("Malloc error");
251 malloc_ptr
= free_mem_ptr
;
253 malloc_ptr
= (malloc_ptr
+ 3) & ~3; /* Align */
255 p
= (void *)malloc_ptr
;
258 if (free_mem_end_ptr
&& malloc_ptr
>= free_mem_end_ptr
)
259 error("Out of memory");
265 static void free(void *where
)
269 malloc_ptr
= free_mem_ptr
;
272 #define malloc(a) kmalloc(a, GFP_KERNEL)
273 #define free(a) kfree(a)
277 Huffman code decoding is performed using a multi-level table lookup.
278 The fastest way to decode is to simply build a lookup table whose
279 size is determined by the longest code. However, the time it takes
280 to build this table can also be a factor if the data being decoded
281 is not very long. The most common codes are necessarily the
282 shortest codes, so those codes dominate the decoding time, and hence
283 the speed. The idea is you can have a shorter table that decodes the
284 shorter, more probable codes, and then point to subsidiary tables for
285 the longer codes. The time it costs to decode the longer codes is
286 then traded against the time it takes to make longer tables.
288 This results of this trade are in the variables lbits and dbits
289 below. lbits is the number of bits the first level table for literal/
290 length codes can decode in one step, and dbits is the same thing for
291 the distance codes. Subsequent tables are also less than or equal to
292 those sizes. These values may be adjusted either when all of the
293 codes are shorter than that, in which case the longest code length in
294 bits is used, or when the shortest code is *longer* than the requested
295 table size, in which case the length of the shortest code in bits is
298 There are two different values for the two tables, since they code a
299 different number of possibilities each. The literal/length table
300 codes 286 possible values, or in a flat code, a little over eight
301 bits. The distance table codes 30 possible values, or a little less
302 than five bits, flat. The optimum values for speed end up being
303 about one bit more than those, so lbits is 8+1 and dbits is 5+1.
304 The optimum values may differ though from machine to machine, and
305 possibly even between compilers. Your mileage may vary.
309 STATIC
const int lbits
= 9; /* bits in base literal/length lookup table */
310 STATIC
const int dbits
= 6; /* bits in base distance lookup table */
313 /* If BMAX needs to be larger than 16, then h and x[] should be ulg. */
314 #define BMAX 16 /* maximum bit length of any code (16 for explode) */
315 #define N_MAX 288 /* maximum number of codes in any set */
318 STATIC
unsigned hufts
; /* track memory usage */
321 STATIC
int INIT
huft_build(
322 unsigned *b
, /* code lengths in bits (all assumed <= BMAX) */
323 unsigned n
, /* number of codes (assumed <= N_MAX) */
324 unsigned s
, /* number of simple-valued codes (0..s-1) */
325 const ush
*d
, /* list of base values for non-simple codes */
326 const ush
*e
, /* list of extra bits for non-simple codes */
327 struct huft
**t
, /* result: starting table */
328 int *m
/* maximum lookup bits, returns actual */
330 /* Given a list of code lengths and a maximum table size, make a set of
331 tables to decode that set of codes. Return zero on success, one if
332 the given code set is incomplete (the tables are still built in this
333 case), two if the input is invalid (all zero length codes or an
334 oversubscribed set of lengths), and three if not enough memory. */
336 unsigned a
; /* counter for codes of length k */
337 unsigned f
; /* i repeats in table every f entries */
338 int g
; /* maximum code length */
339 int h
; /* table level */
340 register unsigned i
; /* counter, current code */
341 register unsigned j
; /* counter */
342 register int k
; /* number of bits in current code */
343 int l
; /* bits per table (returned in m) */
344 register unsigned *p
; /* pointer into c[], b[], or v[] */
345 register struct huft
*q
; /* points to current table */
346 struct huft r
; /* table entry for structure assignment */
347 register int w
; /* bits before this table == (l * h) */
348 unsigned *xp
; /* pointer into x */
349 int y
; /* number of dummy codes added */
350 unsigned z
; /* number of entries in current table */
352 unsigned c
[BMAX
+1]; /* bit length count table */
353 struct huft
*u
[BMAX
]; /* table stack */
354 unsigned v
[N_MAX
]; /* values in order of bit length */
355 unsigned x
[BMAX
+1]; /* bit offsets, then code stack */
363 stk
= malloc(sizeof(*stk
));
365 return 3; /* out of memory */
372 /* Generate counts for each bit length */
373 memzero(stk
->c
, sizeof(stk
->c
));
376 Tracecv(*p
, (stderr
, (n
-i
>= ' ' && n
-i
<= '~' ? "%c %d\n" : "0x%x %d\n"),
378 c
[*p
]++; /* assume all entries <= BMAX */
379 p
++; /* Can't combine with above line (Solaris bug) */
381 if (c
[0] == n
) /* null input--all zero length codes */
383 *t
= (struct huft
*)NULL
;
391 /* Find minimum and maximum length, bound *m by those */
393 for (j
= 1; j
<= BMAX
; j
++)
396 k
= j
; /* minimum code length */
399 for (i
= BMAX
; i
; i
--)
402 g
= i
; /* maximum code length */
409 /* Adjust last length count to fill out codes, if needed */
410 for (y
= 1 << j
; j
< i
; j
++, y
<<= 1)
411 if ((y
-= c
[j
]) < 0) {
412 ret
= 2; /* bad input: more codes than bits */
415 if ((y
-= c
[i
]) < 0) {
423 /* Generate starting offsets into the value table for each length */
425 p
= c
+ 1; xp
= x
+ 2;
426 while (--i
) { /* note that i == g from above */
432 /* Make a table of values in order of bit lengths */
438 n
= x
[g
]; /* set n to length of v */
442 /* Generate the Huffman codes and for each, make the table entries */
443 x
[0] = i
= 0; /* first Huffman code is zero */
444 p
= v
; /* grab values in bit order */
445 h
= -1; /* no tables yet--level -1 */
446 w
= -l
; /* bits decoded == (l * h) */
447 u
[0] = (struct huft
*)NULL
; /* just to keep compilers happy */
448 q
= (struct huft
*)NULL
; /* ditto */
452 /* go through the bit lengths (k already is bits in shortest code) */
460 /* here i is the Huffman code of length k bits for value *p */
461 /* make tables up to required level */
466 w
+= l
; /* previous table always l bits */
468 /* compute minimum size table less than or equal to l bits */
469 z
= (z
= g
- w
) > (unsigned)l
? l
: z
; /* upper limit on table size */
470 if ((f
= 1 << (j
= k
- w
)) > a
+ 1) /* try a k-w bit table */
471 { /* too few codes for k-w bit table */
473 f
-= a
+ 1; /* deduct codes from patterns left */
476 while (++j
< z
) /* try smaller tables up to z bits */
478 if ((f
<<= 1) <= *++xp
)
479 break; /* enough codes to use up j bits */
480 f
-= *xp
; /* else deduct codes from patterns */
484 z
= 1 << j
; /* table entries for j-bit table */
486 /* allocate and link in new table */
487 if ((q
= (struct huft
*)malloc((z
+ 1)*sizeof(struct huft
))) ==
492 ret
= 3; /* not enough memory */
496 hufts
+= z
+ 1; /* track memory usage */
497 *t
= q
+ 1; /* link to list for huft_free() */
498 *(t
= &(q
->v
.t
)) = (struct huft
*)NULL
;
499 u
[h
] = ++q
; /* table starts after link */
502 /* connect to last table, if there is one */
505 x
[h
] = i
; /* save pattern for backing up */
506 r
.b
= (uch
)l
; /* bits to dump before this table */
507 r
.e
= (uch
)(16 + j
); /* bits in this table */
508 r
.v
.t
= q
; /* pointer to this table */
509 j
= i
>> (w
- l
); /* (get around Turbo C bug) */
510 u
[h
-1][j
] = r
; /* connect to last table */
516 /* set up table entry in r */
519 r
.e
= 99; /* out of values--invalid code */
522 r
.e
= (uch
)(*p
< 256 ? 16 : 15); /* 256 is end-of-block code */
523 r
.v
.n
= (ush
)(*p
); /* simple code is just the value */
524 p
++; /* one compiler does not like *p++ */
528 r
.e
= (uch
)e
[*p
- s
]; /* non-simple--look up in lists */
533 /* fill code-like entries with r */
535 for (j
= i
>> w
; j
< z
; j
+= f
)
538 /* backwards increment the k-bit code i */
539 for (j
= 1 << (k
- 1); i
& j
; j
>>= 1)
543 /* backup over finished tables */
544 while ((i
& ((1 << w
) - 1)) != x
[h
])
546 h
--; /* don't need to update q */
556 /* Return true (1) if we were given an incomplete table */
557 ret
= y
!= 0 && g
!= 1;
566 STATIC
int INIT
huft_free(
567 struct huft
*t
/* table to free */
569 /* Free the malloc'ed tables built by huft_build(), which makes a linked
570 list of the tables it made, with the links in a dummy first entry of
573 register struct huft
*p
, *q
;
576 /* Go through linked list, freeing from the malloced (t[-1]) address. */
578 while (p
!= (struct huft
*)NULL
)
588 STATIC
int INIT
inflate_codes(
589 struct huft
*tl
, /* literal/length decoder tables */
590 struct huft
*td
, /* distance decoder tables */
591 int bl
, /* number of bits decoded by tl[] */
592 int bd
/* number of bits decoded by td[] */
594 /* inflate (decompress) the codes in a deflated (compressed) block.
595 Return an error code or zero if it all goes ok. */
597 register unsigned e
; /* table entry flag/number of extra bits */
598 unsigned n
, d
; /* length and index for copy */
599 unsigned w
; /* current window position */
600 struct huft
*t
; /* pointer to table entry */
601 unsigned ml
, md
; /* masks for bl and bd bits */
602 register ulg b
; /* bit buffer */
603 register unsigned k
; /* number of bits in bit buffer */
606 /* make local copies of globals */
607 b
= bb
; /* initialize bit buffer */
609 w
= wp
; /* initialize window position */
611 /* inflate the coded data */
612 ml
= mask_bits
[bl
]; /* precompute masks for speed */
614 for (;;) /* do until end of block */
616 NEEDBITS((unsigned)bl
)
617 if ((e
= (t
= tl
+ ((unsigned)b
& ml
))->e
) > 16)
624 } while ((e
= (t
= t
->v
.t
+ ((unsigned)b
& mask_bits
[e
]))->e
) > 16);
626 if (e
== 16) /* then it's a literal */
628 slide
[w
++] = (uch
)t
->v
.n
;
629 Tracevv((stderr
, "%c", slide
[w
-1]));
636 else /* it's an EOB or a length */
638 /* exit if end of block */
642 /* get length of block to copy */
644 n
= t
->v
.n
+ ((unsigned)b
& mask_bits
[e
]);
647 /* decode distance of block to copy */
648 NEEDBITS((unsigned)bd
)
649 if ((e
= (t
= td
+ ((unsigned)b
& md
))->e
) > 16)
656 } while ((e
= (t
= t
->v
.t
+ ((unsigned)b
& mask_bits
[e
]))->e
) > 16);
659 d
= w
- t
->v
.n
- ((unsigned)b
& mask_bits
[e
]);
661 Tracevv((stderr
,"\\[%d,%d]", w
-d
, n
));
665 n
-= (e
= (e
= WSIZE
- ((d
&= WSIZE
-1) > w
? d
: w
)) > n
? n
: e
);
666 #if !defined(NOMEMCPY) && !defined(DEBUG)
667 if (w
- d
>= e
) /* (this test assumes unsigned comparison) */
669 memcpy(slide
+ w
, slide
+ d
, e
);
673 else /* do it slow to avoid memcpy() overlap */
674 #endif /* !NOMEMCPY */
676 slide
[w
++] = slide
[d
++];
677 Tracevv((stderr
, "%c", slide
[w
-1]));
689 /* restore the globals from the locals */
690 wp
= w
; /* restore global window pointer */
691 bb
= b
; /* restore global bit buffer */
698 return 4; /* Input underrun */
703 STATIC
int INIT
inflate_stored(void)
704 /* "decompress" an inflated type 0 (stored) block. */
706 unsigned n
; /* number of bytes in block */
707 unsigned w
; /* current window position */
708 register ulg b
; /* bit buffer */
709 register unsigned k
; /* number of bits in bit buffer */
713 /* make local copies of globals */
714 b
= bb
; /* initialize bit buffer */
716 w
= wp
; /* initialize window position */
719 /* go to byte boundary */
724 /* get the length and its complement */
726 n
= ((unsigned)b
& 0xffff);
729 if (n
!= (unsigned)((~b
) & 0xffff))
730 return 1; /* error in compressed data */
734 /* read and output the compressed data */
748 /* restore the globals from the locals */
749 wp
= w
; /* restore global window pointer */
750 bb
= b
; /* restore global bit buffer */
757 return 4; /* Input underrun */
762 * We use `noinline' here to prevent gcc-3.5 from using too much stack space
764 STATIC
int noinline INIT
inflate_fixed(void)
765 /* decompress an inflated type 1 (fixed Huffman codes) block. We should
766 either replace this with a custom decoder, or at least precompute the
769 int i
; /* temporary variable */
770 struct huft
*tl
; /* literal/length code table */
771 struct huft
*td
; /* distance code table */
772 int bl
; /* lookup bits for tl */
773 int bd
; /* lookup bits for td */
774 unsigned *l
; /* length list for huft_build */
778 l
= malloc(sizeof(*l
) * 288);
780 return 3; /* out of memory */
782 /* set up literal table */
783 for (i
= 0; i
< 144; i
++)
789 for (; i
< 288; i
++) /* make a complete, but wrong code set */
792 if ((i
= huft_build(l
, 288, 257, cplens
, cplext
, &tl
, &bl
)) != 0) {
797 /* set up distance table */
798 for (i
= 0; i
< 30; i
++) /* make an incomplete code set */
801 if ((i
= huft_build(l
, 30, 0, cpdist
, cpdext
, &td
, &bd
)) > 1)
811 /* decompress until an end-of-block code */
812 if (inflate_codes(tl
, td
, bl
, bd
)) {
817 /* free the decoding tables, return */
826 * We use `noinline' here to prevent gcc-3.5 from using too much stack space
828 STATIC
int noinline INIT
inflate_dynamic(void)
829 /* decompress an inflated type 2 (dynamic Huffman codes) block. */
831 int i
; /* temporary variables */
833 unsigned l
; /* last length */
834 unsigned m
; /* mask for bit lengths table */
835 unsigned n
; /* number of lengths to get */
836 struct huft
*tl
; /* literal/length code table */
837 struct huft
*td
; /* distance code table */
838 int bl
; /* lookup bits for tl */
839 int bd
; /* lookup bits for td */
840 unsigned nb
; /* number of bit length codes */
841 unsigned nl
; /* number of literal/length codes */
842 unsigned nd
; /* number of distance codes */
843 unsigned *ll
; /* literal/length and distance code lengths */
844 register ulg b
; /* bit buffer */
845 register unsigned k
; /* number of bits in bit buffer */
850 #ifdef PKZIP_BUG_WORKAROUND
851 ll
= malloc(sizeof(*ll
) * (288+32)); /* literal/length and distance code lengths */
853 ll
= malloc(sizeof(*ll
) * (286+30)); /* literal/length and distance code lengths */
859 /* make local bit buffer */
864 /* read in table lengths */
866 nl
= 257 + ((unsigned)b
& 0x1f); /* number of literal/length codes */
869 nd
= 1 + ((unsigned)b
& 0x1f); /* number of distance codes */
872 nb
= 4 + ((unsigned)b
& 0xf); /* number of bit length codes */
874 #ifdef PKZIP_BUG_WORKAROUND
875 if (nl
> 288 || nd
> 32)
877 if (nl
> 286 || nd
> 30)
880 ret
= 1; /* bad lengths */
886 /* read in bit-length-code lengths */
887 for (j
= 0; j
< nb
; j
++)
890 ll
[border
[j
]] = (unsigned)b
& 7;
898 /* build decoding table for trees--single level, 7 bit lookup */
900 if ((i
= huft_build(ll
, 19, 19, NULL
, NULL
, &tl
, &bl
)) != 0)
904 ret
= i
; /* incomplete code set */
910 /* read in literal and distance code lengths */
914 while ((unsigned)i
< n
)
916 NEEDBITS((unsigned)bl
)
917 j
= (td
= tl
+ ((unsigned)b
& m
))->b
;
920 if (j
< 16) /* length of code in bits (0..15) */
921 ll
[i
++] = l
= j
; /* save last length in l */
922 else if (j
== 16) /* repeat last length 3 to 6 times */
925 j
= 3 + ((unsigned)b
& 3);
927 if ((unsigned)i
+ j
> n
) {
934 else if (j
== 17) /* 3 to 10 zero length codes */
937 j
= 3 + ((unsigned)b
& 7);
939 if ((unsigned)i
+ j
> n
) {
947 else /* j == 18: 11 to 138 zero length codes */
950 j
= 11 + ((unsigned)b
& 0x7f);
952 if ((unsigned)i
+ j
> n
) {
964 /* free decoding table for trees */
969 /* restore the global bit buffer */
975 /* build the decoding tables for literal/length and distance codes */
977 if ((i
= huft_build(ll
, nl
, 257, cplens
, cplext
, &tl
, &bl
)) != 0)
981 error("incomplete literal tree");
984 ret
= i
; /* incomplete code set */
989 if ((i
= huft_build(ll
+ nl
, nd
, 0, cpdist
, cpdext
, &td
, &bd
)) != 0)
993 error("incomplete distance tree");
994 #ifdef PKZIP_BUG_WORKAROUND
1001 ret
= i
; /* incomplete code set */
1008 /* decompress until an end-of-block code */
1009 if (inflate_codes(tl
, td
, bl
, bd
)) {
1016 /* free the decoding tables, return */
1027 ret
= 4; /* Input underrun */
1033 STATIC
int INIT
inflate_block(
1034 int *e
/* last block flag */
1036 /* decompress an inflated block */
1038 unsigned t
; /* block type */
1039 register ulg b
; /* bit buffer */
1040 register unsigned k
; /* number of bits in bit buffer */
1044 /* make local bit buffer */
1049 /* read in last block bit */
1055 /* read in block type */
1057 t
= (unsigned)b
& 3;
1061 /* restore the global bit buffer */
1065 /* inflate that block type */
1067 return inflate_dynamic();
1069 return inflate_stored();
1071 return inflate_fixed();
1075 /* bad block type */
1079 return 4; /* Input underrun */
1084 STATIC
int INIT
inflate(void)
1085 /* decompress an inflated entry */
1087 int e
; /* last block flag */
1088 int r
; /* result code */
1089 unsigned h
; /* maximum struct huft's malloc'ed */
1091 /* initialize window, bit buffer */
1097 /* decompress until the last block */
1101 #ifdef ARCH_HAS_DECOMP_WDOG
1104 r
= inflate_block(&e
);
1111 /* Undo too much lookahead. The next read will be byte aligned so we
1112 * can discard unused bits in the last meaningful byte.
1119 /* flush out slide */
1123 /* return success */
1125 fprintf(stderr
, "<%u> ", h
);
1130 /**********************************************************************
1132 * The following are support routines for inflate.c
1134 **********************************************************************/
1136 static ulg crc_32_tab
[256];
1137 static ulg crc
; /* initialized in makecrc() so it'll reside in bss */
1138 #define CRC_VALUE (crc ^ 0xffffffffUL)
1141 * Code to compute the CRC-32 table. Borrowed from
1142 * gzip-1.0.3/makecrc.c.
1148 /* Not copyrighted 1990 Mark Adler */
1150 unsigned long c
; /* crc shift register */
1151 unsigned long e
; /* polynomial exclusive-or pattern */
1152 int i
; /* counter for all possible eight bit values */
1153 int k
; /* byte being shifted into crc apparatus */
1155 /* terms of polynomial defining this crc (except x^32): */
1156 static const int p
[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26};
1158 /* Make exclusive-or pattern from polynomial */
1160 for (i
= 0; i
< sizeof(p
)/sizeof(int); i
++)
1161 e
|= 1L << (31 - p
[i
]);
1165 for (i
= 1; i
< 256; i
++)
1168 for (k
= i
| 256; k
!= 1; k
>>= 1)
1170 c
= c
& 1 ? (c
>> 1) ^ e
: c
>> 1;
1177 /* this is initialized here so this code could reside in ROM */
1178 crc
= (ulg
)0xffffffffUL
; /* shift register contents */
1181 /* gzip flag byte */
1182 #define ASCII_FLAG 0x01 /* bit 0 set: file probably ASCII text */
1183 #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
1184 #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
1185 #define ORIG_NAME 0x08 /* bit 3 set: original file name present */
1186 #define COMMENT 0x10 /* bit 4 set: file comment present */
1187 #define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */
1188 #define RESERVED 0xC0 /* bit 6,7: reserved */
1191 * Do the uncompression!
1193 static int INIT
gunzip(void)
1196 unsigned char magic
[2]; /* magic header */
1198 ulg orig_crc
= 0; /* original crc */
1199 ulg orig_len
= 0; /* original uncompressed length */
1202 magic
[0] = NEXTBYTE();
1203 magic
[1] = NEXTBYTE();
1204 method
= NEXTBYTE();
1206 if (magic
[0] != 037 ||
1207 ((magic
[1] != 0213) && (magic
[1] != 0236))) {
1208 error("bad gzip magic numbers");
1212 /* We only support method #8, DEFLATED */
1214 error("internal error, invalid method");
1218 flags
= (uch
)get_byte();
1219 if ((flags
& ENCRYPTED
) != 0) {
1220 error("Input is encrypted");
1223 if ((flags
& CONTINUATION
) != 0) {
1224 error("Multi part input");
1227 if ((flags
& RESERVED
) != 0) {
1228 error("Input has invalid flags");
1231 NEXTBYTE(); /* Get timestamp */
1236 (void)NEXTBYTE(); /* Ignore extra flags for the moment */
1237 (void)NEXTBYTE(); /* Ignore OS type for the moment */
1239 if ((flags
& EXTRA_FIELD
) != 0) {
1240 unsigned len
= (unsigned)NEXTBYTE();
1241 len
|= ((unsigned)NEXTBYTE())<<8;
1242 while (len
--) (void)NEXTBYTE();
1245 /* Get original file name if it was truncated */
1246 if ((flags
& ORIG_NAME
) != 0) {
1247 /* Discard the old name */
1248 while (NEXTBYTE() != 0) /* null */ ;
1251 /* Discard file comment if any */
1252 if ((flags
& COMMENT
) != 0) {
1253 while (NEXTBYTE() != 0) /* null */ ;
1257 if ((res
= inflate())) {
1262 error("invalid compressed format (err=1)");
1265 error("invalid compressed format (err=2)");
1268 error("out of memory");
1271 error("out of input data");
1274 error("invalid compressed format (other)");
1279 /* Get the crc and original length */
1280 /* crc32 (see algorithm.doc)
1281 * uncompressed input size modulo 2^32
1283 orig_crc
= (ulg
) NEXTBYTE();
1284 orig_crc
|= (ulg
) NEXTBYTE() << 8;
1285 orig_crc
|= (ulg
) NEXTBYTE() << 16;
1286 orig_crc
|= (ulg
) NEXTBYTE() << 24;
1288 orig_len
= (ulg
) NEXTBYTE();
1289 orig_len
|= (ulg
) NEXTBYTE() << 8;
1290 orig_len
|= (ulg
) NEXTBYTE() << 16;
1291 orig_len
|= (ulg
) NEXTBYTE() << 24;
1293 /* Validate decompression */
1294 if (orig_crc
!= CRC_VALUE
) {
1298 if (orig_len
!= bytes_out
) {
1299 error("length error");
1304 underrun
: /* NEXTBYTE() goto's here if needed */
1305 error("out of input data");
This page took 0.058713 seconds and 5 git commands to generate.