xref: /freebsd/contrib/xz/src/liblzma/simple/simple_coder.c (revision cdebaff820b2a4915a16cedfd511823d78aab171)
1 ///////////////////////////////////////////////////////////////////////////////
2 //
3 /// \file       simple_coder.c
4 /// \brief      Wrapper for simple filters
5 ///
6 /// Simple filters don't change the size of the data i.e. number of bytes
7 /// in equals the number of bytes out.
8 //
9 //  Author:     Lasse Collin
10 //
11 //  This file has been put into the public domain.
12 //  You can do whatever you want with this file.
13 //
14 ///////////////////////////////////////////////////////////////////////////////
15 
16 #include "simple_private.h"
17 
18 
19 /// Copied or encodes/decodes more data to out[].
20 static lzma_ret
21 copy_or_code(lzma_simple_coder *coder, const lzma_allocator *allocator,
22 		const uint8_t *restrict in, size_t *restrict in_pos,
23 		size_t in_size, uint8_t *restrict out,
24 		size_t *restrict out_pos, size_t out_size, lzma_action action)
25 {
26 	assert(!coder->end_was_reached);
27 
28 	if (coder->next.code == NULL) {
29 		lzma_bufcpy(in, in_pos, in_size, out, out_pos, out_size);
30 
31 		// Check if end of stream was reached.
32 		if (coder->is_encoder && action == LZMA_FINISH
33 				&& *in_pos == in_size)
34 			coder->end_was_reached = true;
35 
36 	} else {
37 		// Call the next coder in the chain to provide us some data.
38 		const lzma_ret ret = coder->next.code(
39 				coder->next.coder, allocator,
40 				in, in_pos, in_size,
41 				out, out_pos, out_size, action);
42 
43 		if (ret == LZMA_STREAM_END) {
44 			assert(!coder->is_encoder
45 					|| action == LZMA_FINISH);
46 			coder->end_was_reached = true;
47 
48 		} else if (ret != LZMA_OK) {
49 			return ret;
50 		}
51 	}
52 
53 	return LZMA_OK;
54 }
55 
56 
57 static size_t
58 call_filter(lzma_simple_coder *coder, uint8_t *buffer, size_t size)
59 {
60 	const size_t filtered = coder->filter(coder->simple,
61 			coder->now_pos, coder->is_encoder,
62 			buffer, size);
63 	coder->now_pos += filtered;
64 	return filtered;
65 }
66 
67 
68 static lzma_ret
69 simple_code(void *coder_ptr, const lzma_allocator *allocator,
70 		const uint8_t *restrict in, size_t *restrict in_pos,
71 		size_t in_size, uint8_t *restrict out,
72 		size_t *restrict out_pos, size_t out_size, lzma_action action)
73 {
74 	lzma_simple_coder *coder = coder_ptr;
75 
76 	// TODO: Add partial support for LZMA_SYNC_FLUSH. We can support it
77 	// in cases when the filter is able to filter everything. With most
78 	// simple filters it can be done at offset that is a multiple of 2,
79 	// 4, or 16. With x86 filter, it needs good luck, and thus cannot
80 	// be made to work predictably.
81 	if (action == LZMA_SYNC_FLUSH)
82 		return LZMA_OPTIONS_ERROR;
83 
84 	// Flush already filtered data from coder->buffer[] to out[].
85 	if (coder->pos < coder->filtered) {
86 		lzma_bufcpy(coder->buffer, &coder->pos, coder->filtered,
87 				out, out_pos, out_size);
88 
89 		// If we couldn't flush all the filtered data, return to
90 		// application immediately.
91 		if (coder->pos < coder->filtered)
92 			return LZMA_OK;
93 
94 		if (coder->end_was_reached) {
95 			assert(coder->filtered == coder->size);
96 			return LZMA_STREAM_END;
97 		}
98 	}
99 
100 	// If we get here, there is no filtered data left in the buffer.
101 	coder->filtered = 0;
102 
103 	assert(!coder->end_was_reached);
104 
105 	// If there is more output space left than there is unfiltered data
106 	// in coder->buffer[], flush coder->buffer[] to out[], and copy/code
107 	// more data to out[] hopefully filling it completely. Then filter
108 	// the data in out[]. This step is where most of the data gets
109 	// filtered if the buffer sizes used by the application are reasonable.
110 	const size_t out_avail = out_size - *out_pos;
111 	const size_t buf_avail = coder->size - coder->pos;
112 	if (out_avail > buf_avail || buf_avail == 0) {
113 		// Store the old position so that we know from which byte
114 		// to start filtering.
115 		const size_t out_start = *out_pos;
116 
117 		// Flush data from coder->buffer[] to out[], but don't reset
118 		// coder->pos and coder->size yet. This way the coder can be
119 		// restarted if the next filter in the chain returns e.g.
120 		// LZMA_MEM_ERROR.
121 		memcpy(out + *out_pos, coder->buffer + coder->pos, buf_avail);
122 		*out_pos += buf_avail;
123 
124 		// Copy/Encode/Decode more data to out[].
125 		{
126 			const lzma_ret ret = copy_or_code(coder, allocator,
127 					in, in_pos, in_size,
128 					out, out_pos, out_size, action);
129 			assert(ret != LZMA_STREAM_END);
130 			if (ret != LZMA_OK)
131 				return ret;
132 		}
133 
134 		// Filter out[].
135 		const size_t size = *out_pos - out_start;
136 		const size_t filtered = call_filter(
137 				coder, out + out_start, size);
138 
139 		const size_t unfiltered = size - filtered;
140 		assert(unfiltered <= coder->allocated / 2);
141 
142 		// Now we can update coder->pos and coder->size, because
143 		// the next coder in the chain (if any) was successful.
144 		coder->pos = 0;
145 		coder->size = unfiltered;
146 
147 		if (coder->end_was_reached) {
148 			// The last byte has been copied to out[] already.
149 			// They are left as is.
150 			coder->size = 0;
151 
152 		} else if (unfiltered > 0) {
153 			// There is unfiltered data left in out[]. Copy it to
154 			// coder->buffer[] and rewind *out_pos appropriately.
155 			*out_pos -= unfiltered;
156 			memcpy(coder->buffer, out + *out_pos, unfiltered);
157 		}
158 	} else if (coder->pos > 0) {
159 		memmove(coder->buffer, coder->buffer + coder->pos, buf_avail);
160 		coder->size -= coder->pos;
161 		coder->pos = 0;
162 	}
163 
164 	assert(coder->pos == 0);
165 
166 	// If coder->buffer[] isn't empty, try to fill it by copying/decoding
167 	// more data. Then filter coder->buffer[] and copy the successfully
168 	// filtered data to out[]. It is probable, that some filtered and
169 	// unfiltered data will be left to coder->buffer[].
170 	if (coder->size > 0) {
171 		{
172 			const lzma_ret ret = copy_or_code(coder, allocator,
173 					in, in_pos, in_size,
174 					coder->buffer, &coder->size,
175 					coder->allocated, action);
176 			assert(ret != LZMA_STREAM_END);
177 			if (ret != LZMA_OK)
178 				return ret;
179 		}
180 
181 		coder->filtered = call_filter(
182 				coder, coder->buffer, coder->size);
183 
184 		// Everything is considered to be filtered if coder->buffer[]
185 		// contains the last bytes of the data.
186 		if (coder->end_was_reached)
187 			coder->filtered = coder->size;
188 
189 		// Flush as much as possible.
190 		lzma_bufcpy(coder->buffer, &coder->pos, coder->filtered,
191 				out, out_pos, out_size);
192 	}
193 
194 	// Check if we got everything done.
195 	if (coder->end_was_reached && coder->pos == coder->size)
196 		return LZMA_STREAM_END;
197 
198 	return LZMA_OK;
199 }
200 
201 
202 static void
203 simple_coder_end(void *coder_ptr, const lzma_allocator *allocator)
204 {
205 	lzma_simple_coder *coder = coder_ptr;
206 	lzma_next_end(&coder->next, allocator);
207 	lzma_free(coder->simple, allocator);
208 	lzma_free(coder, allocator);
209 	return;
210 }
211 
212 
213 static lzma_ret
214 simple_coder_update(void *coder_ptr, const lzma_allocator *allocator,
215 		const lzma_filter *filters_null lzma_attribute((__unused__)),
216 		const lzma_filter *reversed_filters)
217 {
218 	lzma_simple_coder *coder = coder_ptr;
219 
220 	// No update support, just call the next filter in the chain.
221 	return lzma_next_filter_update(
222 			&coder->next, allocator, reversed_filters + 1);
223 }
224 
225 
226 extern lzma_ret
227 lzma_simple_coder_init(lzma_next_coder *next, const lzma_allocator *allocator,
228 		const lzma_filter_info *filters,
229 		size_t (*filter)(void *simple, uint32_t now_pos,
230 			bool is_encoder, uint8_t *buffer, size_t size),
231 		size_t simple_size, size_t unfiltered_max,
232 		uint32_t alignment, bool is_encoder)
233 {
234 	// Allocate memory for the lzma_simple_coder structure if needed.
235 	lzma_simple_coder *coder = next->coder;
236 	if (coder == NULL) {
237 		// Here we allocate space also for the temporary buffer. We
238 		// need twice the size of unfiltered_max, because then it
239 		// is always possible to filter at least unfiltered_max bytes
240 		// more data in coder->buffer[] if it can be filled completely.
241 		coder = lzma_alloc(sizeof(lzma_simple_coder)
242 				+ 2 * unfiltered_max, allocator);
243 		if (coder == NULL)
244 			return LZMA_MEM_ERROR;
245 
246 		next->coder = coder;
247 		next->code = &simple_code;
248 		next->end = &simple_coder_end;
249 		next->update = &simple_coder_update;
250 
251 		coder->next = LZMA_NEXT_CODER_INIT;
252 		coder->filter = filter;
253 		coder->allocated = 2 * unfiltered_max;
254 
255 		// Allocate memory for filter-specific data structure.
256 		if (simple_size > 0) {
257 			coder->simple = lzma_alloc(simple_size, allocator);
258 			if (coder->simple == NULL)
259 				return LZMA_MEM_ERROR;
260 		} else {
261 			coder->simple = NULL;
262 		}
263 	}
264 
265 	if (filters[0].options != NULL) {
266 		const lzma_options_bcj *simple = filters[0].options;
267 		coder->now_pos = simple->start_offset;
268 		if (coder->now_pos & (alignment - 1))
269 			return LZMA_OPTIONS_ERROR;
270 	} else {
271 		coder->now_pos = 0;
272 	}
273 
274 	// Reset variables.
275 	coder->is_encoder = is_encoder;
276 	coder->end_was_reached = false;
277 	coder->pos = 0;
278 	coder->filtered = 0;
279 	coder->size = 0;
280 
281 	return lzma_next_filter_init(&coder->next, allocator, filters + 1);
282 }
283