xref: /dragonfly/contrib/xz/src/liblzma/lz/lz_decoder.h (revision 9348a738)
1 ///////////////////////////////////////////////////////////////////////////////
2 //
3 /// \file       lz_decoder.h
4 /// \brief      LZ out window
5 ///
6 //  Authors:    Igor Pavlov
7 //              Lasse Collin
8 //
9 //  This file has been put into the public domain.
10 //  You can do whatever you want with this file.
11 //
12 ///////////////////////////////////////////////////////////////////////////////
13 
14 #ifndef LZMA_LZ_DECODER_H
15 #define LZMA_LZ_DECODER_H
16 
17 #include "common.h"
18 
19 
20 typedef struct {
21 	/// Pointer to the dictionary buffer. It can be an allocated buffer
22 	/// internal to liblzma, or it can a be a buffer given by the
23 	/// application when in single-call mode (not implemented yet).
24 	uint8_t *buf;
25 
26 	/// Write position in dictionary. The next byte will be written to
27 	/// buf[pos].
28 	size_t pos;
29 
30 	/// Indicates how full the dictionary is. This is used by
31 	/// dict_is_distance_valid() to detect corrupt files that would
32 	/// read beyond the beginning of the dictionary.
33 	size_t full;
34 
35 	/// Write limit
36 	size_t limit;
37 
38 	/// Size of the dictionary
39 	size_t size;
40 
41 	/// True when dictionary should be reset before decoding more data.
42 	bool need_reset;
43 
44 } lzma_dict;
45 
46 
47 typedef struct {
48 	size_t dict_size;
49 	const uint8_t *preset_dict;
50 	size_t preset_dict_size;
51 } lzma_lz_options;
52 
53 
54 typedef struct {
55 	/// Data specific to the LZ-based decoder
56 	lzma_coder *coder;
57 
58 	/// Function to decode from in[] to *dict
59 	lzma_ret (*code)(lzma_coder *restrict coder,
60 			lzma_dict *restrict dict, const uint8_t *restrict in,
61 			size_t *restrict in_pos, size_t in_size);
62 
63 	void (*reset)(lzma_coder *coder, const void *options);
64 
65 	/// Set the uncompressed size
66 	void (*set_uncompressed)(lzma_coder *coder,
67 			lzma_vli uncompressed_size);
68 
69 	/// Free allocated resources
70 	void (*end)(lzma_coder *coder, const lzma_allocator *allocator);
71 
72 } lzma_lz_decoder;
73 
74 
75 #define LZMA_LZ_DECODER_INIT \
76 	(lzma_lz_decoder){ \
77 		.coder = NULL, \
78 		.code = NULL, \
79 		.reset = NULL, \
80 		.set_uncompressed = NULL, \
81 		.end = NULL, \
82 	}
83 
84 
85 extern lzma_ret lzma_lz_decoder_init(lzma_next_coder *next,
86 		const lzma_allocator *allocator,
87 		const lzma_filter_info *filters,
88 		lzma_ret (*lz_init)(lzma_lz_decoder *lz,
89 			const lzma_allocator *allocator, const void *options,
90 			lzma_lz_options *lz_options));
91 
92 extern uint64_t lzma_lz_decoder_memusage(size_t dictionary_size);
93 
94 extern void lzma_lz_decoder_uncompressed(
95 		lzma_coder *coder, lzma_vli uncompressed_size);
96 
97 
98 //////////////////////
99 // Inline functions //
100 //////////////////////
101 
102 /// Get a byte from the history buffer.
103 static inline uint8_t
104 dict_get(const lzma_dict *const dict, const uint32_t distance)
105 {
106 	return dict->buf[dict->pos - distance - 1
107 			+ (distance < dict->pos ? 0 : dict->size)];
108 }
109 
110 
111 /// Test if dictionary is empty.
112 static inline bool
113 dict_is_empty(const lzma_dict *const dict)
114 {
115 	return dict->full == 0;
116 }
117 
118 
119 /// Validate the match distance
120 static inline bool
121 dict_is_distance_valid(const lzma_dict *const dict, const size_t distance)
122 {
123 	return dict->full > distance;
124 }
125 
126 
127 /// Repeat *len bytes at distance.
128 static inline bool
129 dict_repeat(lzma_dict *dict, uint32_t distance, uint32_t *len)
130 {
131 	// Don't write past the end of the dictionary.
132 	const size_t dict_avail = dict->limit - dict->pos;
133 	uint32_t left = my_min(dict_avail, *len);
134 	*len -= left;
135 
136 	// Repeat a block of data from the history. Because memcpy() is faster
137 	// than copying byte by byte in a loop, the copying process gets split
138 	// into three cases.
139 	if (distance < left) {
140 		// Source and target areas overlap, thus we can't use
141 		// memcpy() nor even memmove() safely.
142 		do {
143 			dict->buf[dict->pos] = dict_get(dict, distance);
144 			++dict->pos;
145 		} while (--left > 0);
146 
147 	} else if (distance < dict->pos) {
148 		// The easiest and fastest case
149 		memcpy(dict->buf + dict->pos,
150 				dict->buf + dict->pos - distance - 1,
151 				left);
152 		dict->pos += left;
153 
154 	} else {
155 		// The bigger the dictionary, the more rare this
156 		// case occurs. We need to "wrap" the dict, thus
157 		// we might need two memcpy() to copy all the data.
158 		assert(dict->full == dict->size);
159 		const uint32_t copy_pos
160 				= dict->pos - distance - 1 + dict->size;
161 		uint32_t copy_size = dict->size - copy_pos;
162 
163 		if (copy_size < left) {
164 			memmove(dict->buf + dict->pos, dict->buf + copy_pos,
165 					copy_size);
166 			dict->pos += copy_size;
167 			copy_size = left - copy_size;
168 			memcpy(dict->buf + dict->pos, dict->buf, copy_size);
169 			dict->pos += copy_size;
170 		} else {
171 			memmove(dict->buf + dict->pos, dict->buf + copy_pos,
172 					left);
173 			dict->pos += left;
174 		}
175 	}
176 
177 	// Update how full the dictionary is.
178 	if (dict->full < dict->pos)
179 		dict->full = dict->pos;
180 
181 	return unlikely(*len != 0);
182 }
183 
184 
185 /// Puts one byte into the dictionary. Returns true if the dictionary was
186 /// already full and the byte couldn't be added.
187 static inline bool
188 dict_put(lzma_dict *dict, uint8_t byte)
189 {
190 	if (unlikely(dict->pos == dict->limit))
191 		return true;
192 
193 	dict->buf[dict->pos++] = byte;
194 
195 	if (dict->pos > dict->full)
196 		dict->full = dict->pos;
197 
198 	return false;
199 }
200 
201 
202 /// Copies arbitrary amount of data into the dictionary.
203 static inline void
204 dict_write(lzma_dict *restrict dict, const uint8_t *restrict in,
205 		size_t *restrict in_pos, size_t in_size,
206 		size_t *restrict left)
207 {
208 	// NOTE: If we are being given more data than the size of the
209 	// dictionary, it could be possible to optimize the LZ decoder
210 	// so that not everything needs to go through the dictionary.
211 	// This shouldn't be very common thing in practice though, and
212 	// the slowdown of one extra memcpy() isn't bad compared to how
213 	// much time it would have taken if the data were compressed.
214 
215 	if (in_size - *in_pos > *left)
216 		in_size = *in_pos + *left;
217 
218 	*left -= lzma_bufcpy(in, in_pos, in_size,
219 			dict->buf, &dict->pos, dict->limit);
220 
221 	if (dict->pos > dict->full)
222 		dict->full = dict->pos;
223 
224 	return;
225 }
226 
227 
228 static inline void
229 dict_reset(lzma_dict *dict)
230 {
231 	dict->need_reset = true;
232 	return;
233 }
234 
235 #endif
236