1 #ifndef DEFLATE_H_
2 #define DEFLATE_H_
3 /* deflate.h -- internal compression state
4 * Copyright (C) 1995-2016 Jean-loup Gailly
5 * For conditions of distribution and use, see copyright notice in zlib.h
6 */
7
8 /* WARNING: this file should *not* be used by applications. It is
9 part of the implementation of the compression library and is
10 subject to change. Applications should only use zlib.h.
11 */
12
13 #include "zutil.h"
14 #include "zendian.h"
15
16 /* define NO_GZIP when compiling if you want to disable gzip header and
17 trailer creation by deflate(). NO_GZIP would be used to avoid linking in
18 the crc code when it is not needed. For shared libraries, gzip encoding
19 should be left enabled. */
20 #ifndef NO_GZIP
21 # define GZIP
22 #endif
23
24 /* ===========================================================================
25 * Internal compression state.
26 */
27
28 #define LENGTH_CODES 29
29 /* number of length codes, not counting the special END_BLOCK code */
30
31 #define LITERALS 256
32 /* number of literal bytes 0..255 */
33
34 #define L_CODES (LITERALS+1+LENGTH_CODES)
35 /* number of Literal or Length codes, including the END_BLOCK code */
36
37 #define D_CODES 30
38 /* number of distance codes */
39
40 #define BL_CODES 19
41 /* number of codes used to transfer the bit lengths */
42
43 #define HEAP_SIZE (2*L_CODES+1)
44 /* maximum heap size */
45
46 #define MAX_BITS 15
47 /* All codes must not exceed MAX_BITS bits */
48
49 #define BIT_BUF_SIZE 64
50 /* size of bit buffer in bi_buf */
51
52 #define END_BLOCK 256
53 /* end of block literal code */
54
55 #define INIT_STATE 42 /* zlib header -> BUSY_STATE */
56 #ifdef GZIP
57 # define GZIP_STATE 57 /* gzip header -> BUSY_STATE | EXTRA_STATE */
58 #endif
59 #define EXTRA_STATE 69 /* gzip extra block -> NAME_STATE */
60 #define NAME_STATE 73 /* gzip file name -> COMMENT_STATE */
61 #define COMMENT_STATE 91 /* gzip comment -> HCRC_STATE */
62 #define HCRC_STATE 103 /* gzip header CRC -> BUSY_STATE */
63 #define BUSY_STATE 113 /* deflate -> FINISH_STATE */
64 #define FINISH_STATE 666 /* stream complete */
65 /* Stream status */
66
67 #define HASH_BITS 16u /* log2(HASH_SIZE) */
68 #define HASH_SIZE 65536u /* number of elements in hash table */
69 #define HASH_MASK (HASH_SIZE - 1u) /* HASH_SIZE-1 */
70
71
72 /* Data structure describing a single value and its code string. */
73 typedef struct ct_data_s {
74 union {
75 uint16_t freq; /* frequency count */
76 uint16_t code; /* bit string */
77 } fc;
78 union {
79 uint16_t dad; /* father node in Huffman tree */
80 uint16_t len; /* length of bit string */
81 } dl;
82 } ct_data;
83
84 #define Freq fc.freq
85 #define Code fc.code
86 #define Dad dl.dad
87 #define Len dl.len
88
89 typedef struct static_tree_desc_s static_tree_desc;
90
91 typedef struct tree_desc_s {
92 ct_data *dyn_tree; /* the dynamic tree */
93 int max_code; /* largest code with non zero frequency */
94 const static_tree_desc *stat_desc; /* the corresponding static tree */
95 } tree_desc;
96
97 typedef uint16_t Pos;
98
99 /* A Pos is an index in the character window. We use short instead of int to
100 * save space in the various tables.
101 */
102
103 typedef struct internal_state {
104 PREFIX3(stream) *strm; /* pointer back to this zlib stream */
105 unsigned char *pending_buf; /* output still pending */
106 unsigned char *pending_out; /* next pending byte to output to the stream */
107 uint32_t pending_buf_size; /* size of pending_buf */
108 uint32_t pending; /* nb of bytes in the pending buffer */
109 int wrap; /* bit 0 true for zlib, bit 1 true for gzip */
110 uint32_t gzindex; /* where in extra, name, or comment */
111 PREFIX(gz_headerp) gzhead; /* gzip header information to write */
112 int status; /* as the name implies */
113 int last_flush; /* value of flush param for previous deflate call */
114 int reproducible; /* Whether reproducible compression results are required. */
115
116 int block_open;
117 /* Whether or not a block is currently open for the QUICK deflation scheme.
118 * This is set to 1 if there is an active block, or 0 if the block was just closed.
119 */
120
121 /* used by deflate.c: */
122
123 unsigned int w_size; /* LZ77 window size (32K by default) */
124 unsigned int w_bits; /* log2(w_size) (8..16) */
125 unsigned int w_mask; /* w_size - 1 */
126 unsigned int lookahead; /* number of valid bytes ahead in window */
127
128 unsigned int high_water;
129 /* High water mark offset in window for initialized bytes -- bytes above
130 * this are set to zero in order to avoid memory check warnings when
131 * longest match routines access bytes past the input. This is then
132 * updated to the new high water mark.
133 */
134
135 unsigned int window_size;
136 /* Actual size of window: 2*wSize, except when the user input buffer
137 * is directly used as sliding window.
138 */
139
140 unsigned char *window;
141 /* Sliding window. Input bytes are read into the second half of the window,
142 * and move to the first half later to keep a dictionary of at least wSize
143 * bytes. With this organization, matches are limited to a distance of
144 * wSize-MAX_MATCH bytes, but this ensures that IO is always
145 * performed with a length multiple of the block size. Also, it limits
146 * the window size to 64K, which is quite useful on MSDOS.
147 * To do: use the user input buffer as sliding window.
148 */
149
150 Pos *prev;
151 /* Link to older string with same hash index. To limit the size of this
152 * array to 64K, this link is maintained only for the last 32K strings.
153 * An index in this array is thus a window index modulo 32K.
154 */
155
156 Pos *head; /* Heads of the hash chains or 0. */
157
158 int block_start;
159 /* Window position at the beginning of the current output block. Gets
160 * negative when the window is moved backwards.
161 */
162
163 unsigned int match_length; /* length of best match */
164 Pos prev_match; /* previous match */
165 int match_available; /* set if previous match exists */
166 unsigned int strstart; /* start of string to insert */
167 unsigned int match_start; /* start of matching string */
168
169 unsigned int prev_length;
170 /* Length of the best match at previous step. Matches not greater than this
171 * are discarded. This is used in the lazy match evaluation.
172 */
173
174 unsigned int max_chain_length;
175 /* To speed up deflation, hash chains are never searched beyond this length.
176 * A higher limit improves compression ratio but degrades the speed.
177 */
178
179 unsigned int max_lazy_match;
180 /* Attempt to find a better match only when the current match is strictly smaller
181 * than this value. This mechanism is used only for compression levels >= 4.
182 */
183 # define max_insert_length max_lazy_match
184 /* Insert new strings in the hash table only if the match length is not
185 * greater than this length. This saves time but degrades compression.
186 * max_insert_length is used only for compression levels <= 3.
187 */
188
189 int level; /* compression level (1..9) */
190 int strategy; /* favor or force Huffman coding*/
191
192 unsigned int good_match;
193 /* Use a faster search when the previous match is longer than this */
194
195 int nice_match; /* Stop searching when current match exceeds this */
196
197 #if defined(__x86_64__) || defined(_M_X64) || defined(__i386) || defined(_M_IX86)
198 /* Only used if X86_PCLMULQDQ_CRC is defined */
199 unsigned crc0[4 * 5];
200 #endif
201
202 /* used by trees.c: */
203 /* Didn't use ct_data typedef below to suppress compiler warning */
204 struct ct_data_s dyn_ltree[HEAP_SIZE]; /* literal and length tree */
205 struct ct_data_s dyn_dtree[2*D_CODES+1]; /* distance tree */
206 struct ct_data_s bl_tree[2*BL_CODES+1]; /* Huffman tree for bit lengths */
207
208 struct tree_desc_s l_desc; /* desc. for literal tree */
209 struct tree_desc_s d_desc; /* desc. for distance tree */
210 struct tree_desc_s bl_desc; /* desc. for bit length tree */
211
212 uint16_t bl_count[MAX_BITS+1];
213 /* number of codes at each bit length for an optimal tree */
214
215 int heap[2*L_CODES+1]; /* heap used to build the Huffman trees */
216 int heap_len; /* number of elements in the heap */
217 int heap_max; /* element of largest frequency */
218 /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
219 * The same heap array is used to build all trees.
220 */
221
222 unsigned char depth[2*L_CODES+1];
223 /* Depth of each subtree used as tie breaker for trees of equal frequency
224 */
225
226 unsigned int lit_bufsize;
227 /* Size of match buffer for literals/lengths. There are 4 reasons for
228 * limiting lit_bufsize to 64K:
229 * - frequencies can be kept in 16 bit counters
230 * - if compression is not successful for the first block, all input
231 * data is still in the window so we can still emit a stored block even
232 * when input comes from standard input. (This can also be done for
233 * all blocks if lit_bufsize is not greater than 32K.)
234 * - if compression is not successful for a file smaller than 64K, we can
235 * even emit a stored file instead of a stored block (saving 5 bytes).
236 * This is applicable only for zip (not gzip or zlib).
237 * - creating new Huffman trees less frequently may not provide fast
238 * adaptation to changes in the input data statistics. (Take for
239 * example a binary file with poorly compressible code followed by
240 * a highly compressible string table.) Smaller buffer sizes give
241 * fast adaptation but have of course the overhead of transmitting
242 * trees more frequently.
243 * - I can't count above 4
244 */
245
246 unsigned char *sym_buf; /* buffer for distances and literals/lengths */
247 unsigned int sym_next; /* running index in sym_buf */
248 unsigned int sym_end; /* symbol table full when sym_next reaches this */
249
250 unsigned long opt_len; /* bit length of current block with optimal trees */
251 unsigned long static_len; /* bit length of current block with static trees */
252 unsigned int matches; /* number of string matches in current block */
253 unsigned int insert; /* bytes at end of window left to insert */
254
255 /* compressed_len and bits_sent are only used if ZLIB_DEBUG is defined */
256 unsigned long compressed_len; /* total bit length of compressed file mod 2^32 */
257 unsigned long bits_sent; /* bit length of compressed data sent mod 2^32 */
258
259 /* Reserved for future use and alignment purposes */
260 char *reserved_p;
261
262 uint64_t bi_buf;
263 /* Output buffer. bits are inserted starting at the bottom (least significant bits). */
264
265 int32_t bi_valid;
266 /* Number of valid bits in bi_buf. All bits above the last valid bit are always zero. */
267
268 /* Reserved for future use and alignment purposes */
269 int32_t reserved[11];
270 } ALIGNED_(8) deflate_state;
271
272 typedef enum {
273 need_more, /* block not completed, need more input or more output */
274 block_done, /* block flush performed */
275 finish_started, /* finish started, need only more output at next deflate */
276 finish_done /* finish done, accept no more input or output */
277 } block_state;
278
279 /* Output a byte on the stream.
280 * IN assertion: there is enough room in pending_buf.
281 */
282 #define put_byte(s, c) { \
283 s->pending_buf[s->pending++] = (unsigned char)(c); \
284 }
285
286 /* ===========================================================================
287 * Output a short LSB first on the stream.
288 * IN assertion: there is enough room in pending_buf.
289 */
put_short(deflate_state * s,uint16_t w)290 static inline void put_short(deflate_state *s, uint16_t w) {
291 #if defined(UNALIGNED_OK)
292 *(uint16_t *)(&s->pending_buf[s->pending]) = w;
293 s->pending += 2;
294 #else
295 put_byte(s, (w & 0xff));
296 put_byte(s, ((w >> 8) & 0xff));
297 #endif
298 }
299
300 /* ===========================================================================
301 * Output a short MSB first on the stream.
302 * IN assertion: there is enough room in pending_buf.
303 */
put_short_msb(deflate_state * s,uint16_t w)304 static inline void put_short_msb(deflate_state *s, uint16_t w) {
305 put_byte(s, ((w >> 8) & 0xff));
306 put_byte(s, (w & 0xff));
307 }
308
309 /* ===========================================================================
310 * Output a 32-bit unsigned int LSB first on the stream.
311 * IN assertion: there is enough room in pending_buf.
312 */
put_uint32(deflate_state * s,uint32_t dw)313 static inline void put_uint32(deflate_state *s, uint32_t dw) {
314 #if defined(UNALIGNED_OK)
315 *(uint32_t *)(&s->pending_buf[s->pending]) = dw;
316 s->pending += 4;
317 #else
318 put_byte(s, (dw & 0xff));
319 put_byte(s, ((dw >> 8) & 0xff));
320 put_byte(s, ((dw >> 16) & 0xff));
321 put_byte(s, ((dw >> 24) & 0xff));
322 #endif
323 }
324
325 /* ===========================================================================
326 * Output a 32-bit unsigned int MSB first on the stream.
327 * IN assertion: there is enough room in pending_buf.
328 */
put_uint32_msb(deflate_state * s,uint32_t dw)329 static inline void put_uint32_msb(deflate_state *s, uint32_t dw) {
330 #if defined(UNALIGNED_OK)
331 *(uint32_t *)(&s->pending_buf[s->pending]) = ZSWAP32(dw);
332 s->pending += 4;
333 #else
334 put_byte(s, ((dw >> 24) & 0xff));
335 put_byte(s, ((dw >> 16) & 0xff));
336 put_byte(s, ((dw >> 8) & 0xff));
337 put_byte(s, (dw & 0xff));
338 #endif
339 }
340
341 /* ===========================================================================
342 * Output a 64-bit unsigned int LSB first on the stream.
343 * IN assertion: there is enough room in pending_buf.
344 */
put_uint64(deflate_state * s,uint64_t lld)345 static inline void put_uint64(deflate_state *s, uint64_t lld) {
346 #if defined(UNALIGNED64_OK)
347 *(uint64_t *)(&s->pending_buf[s->pending]) = lld;
348 s->pending += 8;
349 #elif defined(UNALIGNED_OK)
350 *(uint32_t *)(&s->pending_buf[s->pending]) = lld & 0xffffffff;
351 s->pending += 4;
352 *(uint32_t *)(&s->pending_buf[s->pending]) = (lld >> 32) & 0xffffffff;
353 s->pending += 4;
354 #else
355 put_byte(s, (lld & 0xff));
356 put_byte(s, ((lld >> 8) & 0xff));
357 put_byte(s, ((lld >> 16) & 0xff));
358 put_byte(s, ((lld >> 24) & 0xff));
359 put_byte(s, ((lld >> 32) & 0xff));
360 put_byte(s, ((lld >> 40) & 0xff));
361 put_byte(s, ((lld >> 48) & 0xff));
362 put_byte(s, ((lld >> 56) & 0xff));
363 #endif
364 }
365
366 #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
367 /* Minimum amount of lookahead, except at the end of the input file.
368 * See deflate.c for comments about the MIN_MATCH+1.
369 */
370
371 #define MAX_DIST(s) ((s)->w_size-MIN_LOOKAHEAD)
372 /* In order to simplify the code, particularly on 16 bit machines, match
373 * distances are limited to MAX_DIST instead of WSIZE.
374 */
375
376 #define WIN_INIT MAX_MATCH
377 /* Number of bytes after end of data in window to initialize in order to avoid
378 memory checker errors from longest match routines */
379
380
381 void Z_INTERNAL fill_window(deflate_state *s);
382 void Z_INTERNAL slide_hash_c(deflate_state *s);
383
384 /* in trees.c */
385 void Z_INTERNAL zng_tr_init(deflate_state *s);
386 void Z_INTERNAL zng_tr_flush_block(deflate_state *s, char *buf, uint32_t stored_len, int last);
387 void Z_INTERNAL zng_tr_flush_bits(deflate_state *s);
388 void Z_INTERNAL zng_tr_align(deflate_state *s);
389 void Z_INTERNAL zng_tr_stored_block(deflate_state *s, char *buf, uint32_t stored_len, int last);
390 unsigned Z_INTERNAL bi_reverse(unsigned code, int len);
391 void Z_INTERNAL flush_pending(PREFIX3(streamp) strm);
392 #define d_code(dist) ((dist) < 256 ? zng_dist_code[dist] : zng_dist_code[256+((dist)>>7)])
393 /* Mapping from a distance to a distance code. dist is the distance - 1 and
394 * must not have side effects. zng_dist_code[256] and zng_dist_code[257] are never
395 * used.
396 */
397
398 /* Bit buffer and compress bits calculation debugging */
399 #ifdef ZLIB_DEBUG
400 # define cmpr_bits_add(s, len) s->compressed_len += (len)
401 # define cmpr_bits_align(s) s->compressed_len = (s->compressed_len + 7) & ~7L
402 # define sent_bits_add(s, bits) s->bits_sent += (bits)
403 # define sent_bits_align(s) s->bits_sent = (s->bits_sent + 7) & ~7L
404 #else
405 # define cmpr_bits_add(s, len) Z_UNUSED(len)
406 # define cmpr_bits_align(s)
407 # define sent_bits_add(s, bits) Z_UNUSED(bits)
408 # define sent_bits_align(s)
409 #endif
410
411 #endif /* DEFLATE_H_ */
412