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 #include <linux/slab.h>
109 static char rcsid
[] = "#Id: inflate.c,v 0.14 1993/06/10 13:27:04 jloup Exp #";
114 #if defined(STDC_HEADERS) || defined(HAVE_STDLIB_H)
115 # include <sys/types.h>
129 /* Huffman code lookup table entry--this entry is four bytes for machines
130 that have 16-bit pointers (e.g. PC's in the small or medium model).
131 Valid extra bits are 0..13. e == 15 is EOB (end of block), e == 16
132 means that v is a literal, 16 < e < 32 means that v is a pointer to
133 the next table, which codes e - 16 bits, and lastly e == 99 indicates
134 an unused code. If a code with e == 99 is looked up, this implies an
135 error in the data. */
137 uch e
; /* number of extra bits or operation */
138 uch b
; /* number of bits in this code or subcode */
140 ush n
; /* literal, length base, or distance base */
141 struct huft
*t
; /* pointer to next level of table */
146 /* Function prototypes */
147 STATIC
int INIT huft_build
OF((unsigned *, unsigned, unsigned,
148 const ush
*, const ush
*, struct huft
**, int *));
149 STATIC
int INIT huft_free
OF((struct huft
*));
150 STATIC
int INIT inflate_codes
OF((struct huft
*, struct huft
*, int, int));
151 STATIC
int INIT inflate_stored
OF((void));
152 STATIC
int INIT inflate_fixed
OF((void));
153 STATIC
int INIT inflate_dynamic
OF((void));
154 STATIC
int INIT inflate_block
OF((int *));
155 STATIC
int INIT inflate
OF((void));
158 /* The inflate algorithm uses a sliding 32 K byte window on the uncompressed
159 stream to find repeated byte strings. This is implemented here as a
160 circular buffer. The index is updated simply by incrementing and then
161 ANDing with 0x7fff (32K-1). */
162 /* It is left to other modules to supply the 32 K area. It is assumed
163 to be usable as if it were declared "uch slide[32768];" or as just
164 "uch *slide;" and then malloc'ed in the latter case. The definition
165 must be in unzip.h, included above. */
166 /* unsigned wp; current position in slide */
168 #define flush_output(w) (wp=(w),flush_window())
170 /* Tables for deflate from PKZIP's appnote.txt. */
171 static const unsigned border
[] = { /* Order of the bit length code lengths */
172 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
173 static const ush cplens
[] = { /* Copy lengths for literal codes 257..285 */
174 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
175 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
176 /* note: see note #13 above about the 258 in this list. */
177 static const ush cplext
[] = { /* Extra bits for literal codes 257..285 */
178 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
179 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */
180 static const ush cpdist
[] = { /* Copy offsets for distance codes 0..29 */
181 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
182 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
183 8193, 12289, 16385, 24577};
184 static const ush cpdext
[] = { /* Extra bits for distance codes */
185 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
186 7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
191 /* Macros for inflate() bit peeking and grabbing.
195 x = b & mask_bits[j];
198 where NEEDBITS makes sure that b has at least j bits in it, and
199 DUMPBITS removes the bits from b. The macros use the variable k
200 for the number of bits in b. Normally, b and k are register
201 variables for speed, and are initialized at the beginning of a
202 routine that uses these macros from a global bit buffer and count.
204 If we assume that EOB will be the longest code, then we will never
205 ask for bits with NEEDBITS that are beyond the end of the stream.
206 So, NEEDBITS should not read any more bytes than are needed to
207 meet the request. Then no bytes need to be "returned" to the buffer
208 at the end of the last block.
210 However, this assumption is not true for fixed blocks--the EOB code
211 is 7 bits, but the other literal/length codes can be 8 or 9 bits.
212 (The EOB code is shorter than other codes because fixed blocks are
213 generally short. So, while a block always has an EOB, many other
214 literal/length codes have a significantly lower probability of
215 showing up at all.) However, by making the first table have a
216 lookup of seven bits, the EOB code will be found in that first
217 lookup, and so will not require that too many bits be pulled from
221 STATIC ulg bb
; /* bit buffer */
222 STATIC
unsigned bk
; /* bits in bit buffer */
224 STATIC
const ush mask_bits
[] = {
226 0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff,
227 0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff
230 #define NEXTBYTE() ({ int v = get_byte(); if (v < 0) goto underrun; (uch)v; })
231 #define NEEDBITS(n) {while(k<(n)){b|=((ulg)NEXTBYTE())<<k;k+=8;}}
232 #define DUMPBITS(n) {b>>=(n);k-=(n);}
234 #ifndef NO_INFLATE_MALLOC
235 /* A trivial malloc implementation, adapted from
236 * malloc by Hannu Savolainen 1993 and Matthias Urlichs 1994
239 static unsigned long malloc_ptr
;
240 static int malloc_count
;
242 static void *malloc(int size
)
247 error("Malloc error");
249 malloc_ptr
= free_mem_ptr
;
251 malloc_ptr
= (malloc_ptr
+ 3) & ~3; /* Align */
253 p
= (void *)malloc_ptr
;
256 if (free_mem_end_ptr
&& malloc_ptr
>= free_mem_end_ptr
)
257 error("Out of memory");
263 static void free(void *where
)
267 malloc_ptr
= free_mem_ptr
;
270 #define malloc(a) kmalloc(a, GFP_KERNEL)
271 #define free(a) kfree(a)
275 Huffman code decoding is performed using a multi-level table lookup.
276 The fastest way to decode is to simply build a lookup table whose
277 size is determined by the longest code. However, the time it takes
278 to build this table can also be a factor if the data being decoded
279 is not very long. The most common codes are necessarily the
280 shortest codes, so those codes dominate the decoding time, and hence
281 the speed. The idea is you can have a shorter table that decodes the
282 shorter, more probable codes, and then point to subsidiary tables for
283 the longer codes. The time it costs to decode the longer codes is
284 then traded against the time it takes to make longer tables.
286 This results of this trade are in the variables lbits and dbits
287 below. lbits is the number of bits the first level table for literal/
288 length codes can decode in one step, and dbits is the same thing for
289 the distance codes. Subsequent tables are also less than or equal to
290 those sizes. These values may be adjusted either when all of the
291 codes are shorter than that, in which case the longest code length in
292 bits is used, or when the shortest code is *longer* than the requested
293 table size, in which case the length of the shortest code in bits is
296 There are two different values for the two tables, since they code a
297 different number of possibilities each. The literal/length table
298 codes 286 possible values, or in a flat code, a little over eight
299 bits. The distance table codes 30 possible values, or a little less
300 than five bits, flat. The optimum values for speed end up being
301 about one bit more than those, so lbits is 8+1 and dbits is 5+1.
302 The optimum values may differ though from machine to machine, and
303 possibly even between compilers. Your mileage may vary.
307 STATIC
const int lbits
= 9; /* bits in base literal/length lookup table */
308 STATIC
const int dbits
= 6; /* bits in base distance lookup table */
311 /* If BMAX needs to be larger than 16, then h and x[] should be ulg. */
312 #define BMAX 16 /* maximum bit length of any code (16 for explode) */
313 #define N_MAX 288 /* maximum number of codes in any set */
316 STATIC
unsigned hufts
; /* track memory usage */
319 STATIC
int INIT
huft_build(
320 unsigned *b
, /* code lengths in bits (all assumed <= BMAX) */
321 unsigned n
, /* number of codes (assumed <= N_MAX) */
322 unsigned s
, /* number of simple-valued codes (0..s-1) */
323 const ush
*d
, /* list of base values for non-simple codes */
324 const ush
*e
, /* list of extra bits for non-simple codes */
325 struct huft
**t
, /* result: starting table */
326 int *m
/* maximum lookup bits, returns actual */
328 /* Given a list of code lengths and a maximum table size, make a set of
329 tables to decode that set of codes. Return zero on success, one if
330 the given code set is incomplete (the tables are still built in this
331 case), two if the input is invalid (all zero length codes or an
332 oversubscribed set of lengths), and three if not enough memory. */
334 unsigned a
; /* counter for codes of length k */
335 unsigned f
; /* i repeats in table every f entries */
336 int g
; /* maximum code length */
337 int h
; /* table level */
338 register unsigned i
; /* counter, current code */
339 register unsigned j
; /* counter */
340 register int k
; /* number of bits in current code */
341 int l
; /* bits per table (returned in m) */
342 register unsigned *p
; /* pointer into c[], b[], or v[] */
343 register struct huft
*q
; /* points to current table */
344 struct huft r
; /* table entry for structure assignment */
345 register int w
; /* bits before this table == (l * h) */
346 unsigned *xp
; /* pointer into x */
347 int y
; /* number of dummy codes added */
348 unsigned z
; /* number of entries in current table */
350 unsigned c
[BMAX
+1]; /* bit length count table */
351 struct huft
*u
[BMAX
]; /* table stack */
352 unsigned v
[N_MAX
]; /* values in order of bit length */
353 unsigned x
[BMAX
+1]; /* bit offsets, then code stack */
361 stk
= malloc(sizeof(*stk
));
363 return 3; /* out of memory */
370 /* Generate counts for each bit length */
371 memzero(stk
->c
, sizeof(stk
->c
));
374 Tracecv(*p
, (stderr
, (n
-i
>= ' ' && n
-i
<= '~' ? "%c %d\n" : "0x%x %d\n"),
376 c
[*p
]++; /* assume all entries <= BMAX */
377 p
++; /* Can't combine with above line (Solaris bug) */
379 if (c
[0] == n
) /* null input--all zero length codes */
381 *t
= (struct huft
*)NULL
;
389 /* Find minimum and maximum length, bound *m by those */
391 for (j
= 1; j
<= BMAX
; j
++)
394 k
= j
; /* minimum code length */
397 for (i
= BMAX
; i
; i
--)
400 g
= i
; /* maximum code length */
407 /* Adjust last length count to fill out codes, if needed */
408 for (y
= 1 << j
; j
< i
; j
++, y
<<= 1)
409 if ((y
-= c
[j
]) < 0) {
410 ret
= 2; /* bad input: more codes than bits */
413 if ((y
-= c
[i
]) < 0) {
421 /* Generate starting offsets into the value table for each length */
423 p
= c
+ 1; xp
= x
+ 2;
424 while (--i
) { /* note that i == g from above */
430 /* Make a table of values in order of bit lengths */
436 n
= x
[g
]; /* set n to length of v */
440 /* Generate the Huffman codes and for each, make the table entries */
441 x
[0] = i
= 0; /* first Huffman code is zero */
442 p
= v
; /* grab values in bit order */
443 h
= -1; /* no tables yet--level -1 */
444 w
= -l
; /* bits decoded == (l * h) */
445 u
[0] = (struct huft
*)NULL
; /* just to keep compilers happy */
446 q
= (struct huft
*)NULL
; /* ditto */
450 /* go through the bit lengths (k already is bits in shortest code) */
458 /* here i is the Huffman code of length k bits for value *p */
459 /* make tables up to required level */
464 w
+= l
; /* previous table always l bits */
466 /* compute minimum size table less than or equal to l bits */
467 z
= (z
= g
- w
) > (unsigned)l
? l
: z
; /* upper limit on table size */
468 if ((f
= 1 << (j
= k
- w
)) > a
+ 1) /* try a k-w bit table */
469 { /* too few codes for k-w bit table */
471 f
-= a
+ 1; /* deduct codes from patterns left */
474 while (++j
< z
) /* try smaller tables up to z bits */
476 if ((f
<<= 1) <= *++xp
)
477 break; /* enough codes to use up j bits */
478 f
-= *xp
; /* else deduct codes from patterns */
482 z
= 1 << j
; /* table entries for j-bit table */
484 /* allocate and link in new table */
485 if ((q
= (struct huft
*)malloc((z
+ 1)*sizeof(struct huft
))) ==
490 ret
= 3; /* not enough memory */
494 hufts
+= z
+ 1; /* track memory usage */
495 *t
= q
+ 1; /* link to list for huft_free() */
496 *(t
= &(q
->v
.t
)) = (struct huft
*)NULL
;
497 u
[h
] = ++q
; /* table starts after link */
500 /* connect to last table, if there is one */
503 x
[h
] = i
; /* save pattern for backing up */
504 r
.b
= (uch
)l
; /* bits to dump before this table */
505 r
.e
= (uch
)(16 + j
); /* bits in this table */
506 r
.v
.t
= q
; /* pointer to this table */
507 j
= i
>> (w
- l
); /* (get around Turbo C bug) */
508 u
[h
-1][j
] = r
; /* connect to last table */
514 /* set up table entry in r */
517 r
.e
= 99; /* out of values--invalid code */
520 r
.e
= (uch
)(*p
< 256 ? 16 : 15); /* 256 is end-of-block code */
521 r
.v
.n
= (ush
)(*p
); /* simple code is just the value */
522 p
++; /* one compiler does not like *p++ */
526 r
.e
= (uch
)e
[*p
- s
]; /* non-simple--look up in lists */
531 /* fill code-like entries with r */
533 for (j
= i
>> w
; j
< z
; j
+= f
)
536 /* backwards increment the k-bit code i */
537 for (j
= 1 << (k
- 1); i
& j
; j
>>= 1)
541 /* backup over finished tables */
542 while ((i
& ((1 << w
) - 1)) != x
[h
])
544 h
--; /* don't need to update q */
554 /* Return true (1) if we were given an incomplete table */
555 ret
= y
!= 0 && g
!= 1;
564 STATIC
int INIT
huft_free(
565 struct huft
*t
/* table to free */
567 /* Free the malloc'ed tables built by huft_build(), which makes a linked
568 list of the tables it made, with the links in a dummy first entry of
571 register struct huft
*p
, *q
;
574 /* Go through linked list, freeing from the malloced (t[-1]) address. */
576 while (p
!= (struct huft
*)NULL
)
586 STATIC
int INIT
inflate_codes(
587 struct huft
*tl
, /* literal/length decoder tables */
588 struct huft
*td
, /* distance decoder tables */
589 int bl
, /* number of bits decoded by tl[] */
590 int bd
/* number of bits decoded by td[] */
592 /* inflate (decompress) the codes in a deflated (compressed) block.
593 Return an error code or zero if it all goes ok. */
595 register unsigned e
; /* table entry flag/number of extra bits */
596 unsigned n
, d
; /* length and index for copy */
597 unsigned w
; /* current window position */
598 struct huft
*t
; /* pointer to table entry */
599 unsigned ml
, md
; /* masks for bl and bd bits */
600 register ulg b
; /* bit buffer */
601 register unsigned k
; /* number of bits in bit buffer */
604 /* make local copies of globals */
605 b
= bb
; /* initialize bit buffer */
607 w
= wp
; /* initialize window position */
609 /* inflate the coded data */
610 ml
= mask_bits
[bl
]; /* precompute masks for speed */
612 for (;;) /* do until end of block */
614 NEEDBITS((unsigned)bl
)
615 if ((e
= (t
= tl
+ ((unsigned)b
& ml
))->e
) > 16)
622 } while ((e
= (t
= t
->v
.t
+ ((unsigned)b
& mask_bits
[e
]))->e
) > 16);
624 if (e
== 16) /* then it's a literal */
626 slide
[w
++] = (uch
)t
->v
.n
;
627 Tracevv((stderr
, "%c", slide
[w
-1]));
634 else /* it's an EOB or a length */
636 /* exit if end of block */
640 /* get length of block to copy */
642 n
= t
->v
.n
+ ((unsigned)b
& mask_bits
[e
]);
645 /* decode distance of block to copy */
646 NEEDBITS((unsigned)bd
)
647 if ((e
= (t
= td
+ ((unsigned)b
& md
))->e
) > 16)
654 } while ((e
= (t
= t
->v
.t
+ ((unsigned)b
& mask_bits
[e
]))->e
) > 16);
657 d
= w
- t
->v
.n
- ((unsigned)b
& mask_bits
[e
]);
659 Tracevv((stderr
,"\\[%d,%d]", w
-d
, n
));
663 n
-= (e
= (e
= WSIZE
- ((d
&= WSIZE
-1) > w
? d
: w
)) > n
? n
: e
);
664 #if !defined(NOMEMCPY) && !defined(DEBUG)
665 if (w
- d
>= e
) /* (this test assumes unsigned comparison) */
667 memcpy(slide
+ w
, slide
+ d
, e
);
671 else /* do it slow to avoid memcpy() overlap */
672 #endif /* !NOMEMCPY */
674 slide
[w
++] = slide
[d
++];
675 Tracevv((stderr
, "%c", slide
[w
-1]));
687 /* restore the globals from the locals */
688 wp
= w
; /* restore global window pointer */
689 bb
= b
; /* restore global bit buffer */
696 return 4; /* Input underrun */
701 STATIC
int INIT
inflate_stored(void)
702 /* "decompress" an inflated type 0 (stored) block. */
704 unsigned n
; /* number of bytes in block */
705 unsigned w
; /* current window position */
706 register ulg b
; /* bit buffer */
707 register unsigned k
; /* number of bits in bit buffer */
711 /* make local copies of globals */
712 b
= bb
; /* initialize bit buffer */
714 w
= wp
; /* initialize window position */
717 /* go to byte boundary */
722 /* get the length and its complement */
724 n
= ((unsigned)b
& 0xffff);
727 if (n
!= (unsigned)((~b
) & 0xffff))
728 return 1; /* error in compressed data */
732 /* read and output the compressed data */
746 /* restore the globals from the locals */
747 wp
= w
; /* restore global window pointer */
748 bb
= b
; /* restore global bit buffer */
755 return 4; /* Input underrun */
760 * We use `noinline' here to prevent gcc-3.5 from using too much stack space
762 STATIC
int noinline INIT
inflate_fixed(void)
763 /* decompress an inflated type 1 (fixed Huffman codes) block. We should
764 either replace this with a custom decoder, or at least precompute the
767 int i
; /* temporary variable */
768 struct huft
*tl
; /* literal/length code table */
769 struct huft
*td
; /* distance code table */
770 int bl
; /* lookup bits for tl */
771 int bd
; /* lookup bits for td */
772 unsigned *l
; /* length list for huft_build */
776 l
= malloc(sizeof(*l
) * 288);
778 return 3; /* out of memory */
780 /* set up literal table */
781 for (i
= 0; i
< 144; i
++)
787 for (; i
< 288; i
++) /* make a complete, but wrong code set */
790 if ((i
= huft_build(l
, 288, 257, cplens
, cplext
, &tl
, &bl
)) != 0) {
795 /* set up distance table */
796 for (i
= 0; i
< 30; i
++) /* make an incomplete code set */
799 if ((i
= huft_build(l
, 30, 0, cpdist
, cpdext
, &td
, &bd
)) > 1)
809 /* decompress until an end-of-block code */
810 if (inflate_codes(tl
, td
, bl
, bd
)) {
815 /* free the decoding tables, return */
824 * We use `noinline' here to prevent gcc-3.5 from using too much stack space
826 STATIC
int noinline INIT
inflate_dynamic(void)
827 /* decompress an inflated type 2 (dynamic Huffman codes) block. */
829 int i
; /* temporary variables */
831 unsigned l
; /* last length */
832 unsigned m
; /* mask for bit lengths table */
833 unsigned n
; /* number of lengths to get */
834 struct huft
*tl
; /* literal/length code table */
835 struct huft
*td
; /* distance code table */
836 int bl
; /* lookup bits for tl */
837 int bd
; /* lookup bits for td */
838 unsigned nb
; /* number of bit length codes */
839 unsigned nl
; /* number of literal/length codes */
840 unsigned nd
; /* number of distance codes */
841 unsigned *ll
; /* literal/length and distance code lengths */
842 register ulg b
; /* bit buffer */
843 register unsigned k
; /* number of bits in bit buffer */
848 #ifdef PKZIP_BUG_WORKAROUND
849 ll
= malloc(sizeof(*ll
) * (288+32)); /* literal/length and distance code lengths */
851 ll
= malloc(sizeof(*ll
) * (286+30)); /* literal/length and distance code lengths */
857 /* make local bit buffer */
862 /* read in table lengths */
864 nl
= 257 + ((unsigned)b
& 0x1f); /* number of literal/length codes */
867 nd
= 1 + ((unsigned)b
& 0x1f); /* number of distance codes */
870 nb
= 4 + ((unsigned)b
& 0xf); /* number of bit length codes */
872 #ifdef PKZIP_BUG_WORKAROUND
873 if (nl
> 288 || nd
> 32)
875 if (nl
> 286 || nd
> 30)
878 ret
= 1; /* bad lengths */
884 /* read in bit-length-code lengths */
885 for (j
= 0; j
< nb
; j
++)
888 ll
[border
[j
]] = (unsigned)b
& 7;
896 /* build decoding table for trees--single level, 7 bit lookup */
898 if ((i
= huft_build(ll
, 19, 19, NULL
, NULL
, &tl
, &bl
)) != 0)
902 ret
= i
; /* incomplete code set */
908 /* read in literal and distance code lengths */
912 while ((unsigned)i
< n
)
914 NEEDBITS((unsigned)bl
)
915 j
= (td
= tl
+ ((unsigned)b
& m
))->b
;
918 if (j
< 16) /* length of code in bits (0..15) */
919 ll
[i
++] = l
= j
; /* save last length in l */
920 else if (j
== 16) /* repeat last length 3 to 6 times */
923 j
= 3 + ((unsigned)b
& 3);
925 if ((unsigned)i
+ j
> n
) {
932 else if (j
== 17) /* 3 to 10 zero length codes */
935 j
= 3 + ((unsigned)b
& 7);
937 if ((unsigned)i
+ j
> n
) {
945 else /* j == 18: 11 to 138 zero length codes */
948 j
= 11 + ((unsigned)b
& 0x7f);
950 if ((unsigned)i
+ j
> n
) {
962 /* free decoding table for trees */
967 /* restore the global bit buffer */
973 /* build the decoding tables for literal/length and distance codes */
975 if ((i
= huft_build(ll
, nl
, 257, cplens
, cplext
, &tl
, &bl
)) != 0)
979 error("incomplete literal tree");
982 ret
= i
; /* incomplete code set */
987 if ((i
= huft_build(ll
+ nl
, nd
, 0, cpdist
, cpdext
, &td
, &bd
)) != 0)
991 error("incomplete distance tree");
992 #ifdef PKZIP_BUG_WORKAROUND
999 ret
= i
; /* incomplete code set */
1006 /* decompress until an end-of-block code */
1007 if (inflate_codes(tl
, td
, bl
, bd
)) {
1014 /* free the decoding tables, return */
1025 ret
= 4; /* Input underrun */
1031 STATIC
int INIT
inflate_block(
1032 int *e
/* last block flag */
1034 /* decompress an inflated block */
1036 unsigned t
; /* block type */
1037 register ulg b
; /* bit buffer */
1038 register unsigned k
; /* number of bits in bit buffer */
1042 /* make local bit buffer */
1047 /* read in last block bit */
1053 /* read in block type */
1055 t
= (unsigned)b
& 3;
1059 /* restore the global bit buffer */
1063 /* inflate that block type */
1065 return inflate_dynamic();
1067 return inflate_stored();
1069 return inflate_fixed();
1073 /* bad block type */
1077 return 4; /* Input underrun */
1082 STATIC
int INIT
inflate(void)
1083 /* decompress an inflated entry */
1085 int e
; /* last block flag */
1086 int r
; /* result code */
1087 unsigned h
; /* maximum struct huft's malloc'ed */
1089 /* initialize window, bit buffer */
1095 /* decompress until the last block */
1099 #ifdef ARCH_HAS_DECOMP_WDOG
1102 r
= inflate_block(&e
);
1109 /* Undo too much lookahead. The next read will be byte aligned so we
1110 * can discard unused bits in the last meaningful byte.
1117 /* flush out slide */
1121 /* return success */
1123 fprintf(stderr
, "<%u> ", h
);
1128 /**********************************************************************
1130 * The following are support routines for inflate.c
1132 **********************************************************************/
1134 static ulg crc_32_tab
[256];
1135 static ulg crc
; /* initialized in makecrc() so it'll reside in bss */
1136 #define CRC_VALUE (crc ^ 0xffffffffUL)
1139 * Code to compute the CRC-32 table. Borrowed from
1140 * gzip-1.0.3/makecrc.c.
1146 /* Not copyrighted 1990 Mark Adler */
1148 unsigned long c
; /* crc shift register */
1149 unsigned long e
; /* polynomial exclusive-or pattern */
1150 int i
; /* counter for all possible eight bit values */
1151 int k
; /* byte being shifted into crc apparatus */
1153 /* terms of polynomial defining this crc (except x^32): */
1154 static const int p
[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26};
1156 /* Make exclusive-or pattern from polynomial */
1158 for (i
= 0; i
< sizeof(p
)/sizeof(int); i
++)
1159 e
|= 1L << (31 - p
[i
]);
1163 for (i
= 1; i
< 256; i
++)
1166 for (k
= i
| 256; k
!= 1; k
>>= 1)
1168 c
= c
& 1 ? (c
>> 1) ^ e
: c
>> 1;
1175 /* this is initialized here so this code could reside in ROM */
1176 crc
= (ulg
)0xffffffffUL
; /* shift register contents */
1179 /* gzip flag byte */
1180 #define ASCII_FLAG 0x01 /* bit 0 set: file probably ASCII text */
1181 #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
1182 #define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
1183 #define ORIG_NAME 0x08 /* bit 3 set: original file name present */
1184 #define COMMENT 0x10 /* bit 4 set: file comment present */
1185 #define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */
1186 #define RESERVED 0xC0 /* bit 6,7: reserved */
1189 * Do the uncompression!
1191 static int INIT
gunzip(void)
1194 unsigned char magic
[2]; /* magic header */
1196 ulg orig_crc
= 0; /* original crc */
1197 ulg orig_len
= 0; /* original uncompressed length */
1200 magic
[0] = NEXTBYTE();
1201 magic
[1] = NEXTBYTE();
1202 method
= NEXTBYTE();
1204 if (magic
[0] != 037 ||
1205 ((magic
[1] != 0213) && (magic
[1] != 0236))) {
1206 error("bad gzip magic numbers");
1210 /* We only support method #8, DEFLATED */
1212 error("internal error, invalid method");
1216 flags
= (uch
)get_byte();
1217 if ((flags
& ENCRYPTED
) != 0) {
1218 error("Input is encrypted");
1221 if ((flags
& CONTINUATION
) != 0) {
1222 error("Multi part input");
1225 if ((flags
& RESERVED
) != 0) {
1226 error("Input has invalid flags");
1229 NEXTBYTE(); /* Get timestamp */
1234 (void)NEXTBYTE(); /* Ignore extra flags for the moment */
1235 (void)NEXTBYTE(); /* Ignore OS type for the moment */
1237 if ((flags
& EXTRA_FIELD
) != 0) {
1238 unsigned len
= (unsigned)NEXTBYTE();
1239 len
|= ((unsigned)NEXTBYTE())<<8;
1240 while (len
--) (void)NEXTBYTE();
1243 /* Get original file name if it was truncated */
1244 if ((flags
& ORIG_NAME
) != 0) {
1245 /* Discard the old name */
1246 while (NEXTBYTE() != 0) /* null */ ;
1249 /* Discard file comment if any */
1250 if ((flags
& COMMENT
) != 0) {
1251 while (NEXTBYTE() != 0) /* null */ ;
1255 if ((res
= inflate())) {
1260 error("invalid compressed format (err=1)");
1263 error("invalid compressed format (err=2)");
1266 error("out of memory");
1269 error("out of input data");
1272 error("invalid compressed format (other)");
1277 /* Get the crc and original length */
1278 /* crc32 (see algorithm.doc)
1279 * uncompressed input size modulo 2^32
1281 orig_crc
= (ulg
) NEXTBYTE();
1282 orig_crc
|= (ulg
) NEXTBYTE() << 8;
1283 orig_crc
|= (ulg
) NEXTBYTE() << 16;
1284 orig_crc
|= (ulg
) NEXTBYTE() << 24;
1286 orig_len
= (ulg
) NEXTBYTE();
1287 orig_len
|= (ulg
) NEXTBYTE() << 8;
1288 orig_len
|= (ulg
) NEXTBYTE() << 16;
1289 orig_len
|= (ulg
) NEXTBYTE() << 24;
1291 /* Validate decompression */
1292 if (orig_crc
!= CRC_VALUE
) {
1296 if (orig_len
!= bytes_out
) {
1297 error("length error");
1302 underrun
: /* NEXTBYTE() goto's here if needed */
1303 error("out of input data");