1 | ///////////////////////////////////////////////////////////////////////////////
|
---|
2 | //
|
---|
3 | /// \file stream_decoder.c
|
---|
4 | /// \brief Decodes .xz Streams
|
---|
5 | //
|
---|
6 | // Author: Lasse Collin
|
---|
7 | //
|
---|
8 | // This file has been put into the public domain.
|
---|
9 | // You can do whatever you want with this file.
|
---|
10 | //
|
---|
11 | ///////////////////////////////////////////////////////////////////////////////
|
---|
12 |
|
---|
13 | #include "stream_decoder.h"
|
---|
14 | #include "block_decoder.h"
|
---|
15 | #include "index.h"
|
---|
16 |
|
---|
17 |
|
---|
18 | typedef struct {
|
---|
19 | enum {
|
---|
20 | SEQ_STREAM_HEADER,
|
---|
21 | SEQ_BLOCK_HEADER,
|
---|
22 | SEQ_BLOCK_INIT,
|
---|
23 | SEQ_BLOCK_RUN,
|
---|
24 | SEQ_INDEX,
|
---|
25 | SEQ_STREAM_FOOTER,
|
---|
26 | SEQ_STREAM_PADDING,
|
---|
27 | } sequence;
|
---|
28 |
|
---|
29 | /// Block decoder
|
---|
30 | lzma_next_coder block_decoder;
|
---|
31 |
|
---|
32 | /// Block options decoded by the Block Header decoder and used by
|
---|
33 | /// the Block decoder.
|
---|
34 | lzma_block block_options;
|
---|
35 |
|
---|
36 | /// Stream Flags from Stream Header
|
---|
37 | lzma_stream_flags stream_flags;
|
---|
38 |
|
---|
39 | /// Index is hashed so that it can be compared to the sizes of Blocks
|
---|
40 | /// with O(1) memory usage.
|
---|
41 | lzma_index_hash *index_hash;
|
---|
42 |
|
---|
43 | /// Memory usage limit
|
---|
44 | uint64_t memlimit;
|
---|
45 |
|
---|
46 | /// Amount of memory actually needed (only an estimate)
|
---|
47 | uint64_t memusage;
|
---|
48 |
|
---|
49 | /// If true, LZMA_NO_CHECK is returned if the Stream has
|
---|
50 | /// no integrity check.
|
---|
51 | bool tell_no_check;
|
---|
52 |
|
---|
53 | /// If true, LZMA_UNSUPPORTED_CHECK is returned if the Stream has
|
---|
54 | /// an integrity check that isn't supported by this liblzma build.
|
---|
55 | bool tell_unsupported_check;
|
---|
56 |
|
---|
57 | /// If true, LZMA_GET_CHECK is returned after decoding Stream Header.
|
---|
58 | bool tell_any_check;
|
---|
59 |
|
---|
60 | /// If true, we will tell the Block decoder to skip calculating
|
---|
61 | /// and verifying the integrity check.
|
---|
62 | bool ignore_check;
|
---|
63 |
|
---|
64 | /// If true, we will decode concatenated Streams that possibly have
|
---|
65 | /// Stream Padding between or after them. LZMA_STREAM_END is returned
|
---|
66 | /// once the application isn't giving us any new input (LZMA_FINISH),
|
---|
67 | /// and we aren't in the middle of a Stream, and possible
|
---|
68 | /// Stream Padding is a multiple of four bytes.
|
---|
69 | bool concatenated;
|
---|
70 |
|
---|
71 | /// When decoding concatenated Streams, this is true as long as we
|
---|
72 | /// are decoding the first Stream. This is needed to avoid misleading
|
---|
73 | /// LZMA_FORMAT_ERROR in case the later Streams don't have valid magic
|
---|
74 | /// bytes.
|
---|
75 | bool first_stream;
|
---|
76 |
|
---|
77 | /// Write position in buffer[] and position in Stream Padding
|
---|
78 | size_t pos;
|
---|
79 |
|
---|
80 | /// Buffer to hold Stream Header, Block Header, and Stream Footer.
|
---|
81 | /// Block Header has biggest maximum size.
|
---|
82 | uint8_t buffer[LZMA_BLOCK_HEADER_SIZE_MAX];
|
---|
83 | } lzma_stream_coder;
|
---|
84 |
|
---|
85 |
|
---|
86 | static lzma_ret
|
---|
87 | stream_decoder_reset(lzma_stream_coder *coder, const lzma_allocator *allocator)
|
---|
88 | {
|
---|
89 | // Initialize the Index hash used to verify the Index.
|
---|
90 | coder->index_hash = lzma_index_hash_init(coder->index_hash, allocator);
|
---|
91 | if (coder->index_hash == NULL)
|
---|
92 | return LZMA_MEM_ERROR;
|
---|
93 |
|
---|
94 | // Reset the rest of the variables.
|
---|
95 | coder->sequence = SEQ_STREAM_HEADER;
|
---|
96 | coder->pos = 0;
|
---|
97 |
|
---|
98 | return LZMA_OK;
|
---|
99 | }
|
---|
100 |
|
---|
101 |
|
---|
102 | static lzma_ret
|
---|
103 | stream_decode(void *coder_ptr, const lzma_allocator *allocator,
|
---|
104 | const uint8_t *restrict in, size_t *restrict in_pos,
|
---|
105 | size_t in_size, uint8_t *restrict out,
|
---|
106 | size_t *restrict out_pos, size_t out_size, lzma_action action)
|
---|
107 | {
|
---|
108 | lzma_stream_coder *coder = coder_ptr;
|
---|
109 |
|
---|
110 | // When decoding the actual Block, it may be able to produce more
|
---|
111 | // output even if we don't give it any new input.
|
---|
112 | while (true)
|
---|
113 | switch (coder->sequence) {
|
---|
114 | case SEQ_STREAM_HEADER: {
|
---|
115 | // Copy the Stream Header to the internal buffer.
|
---|
116 | lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos,
|
---|
117 | LZMA_STREAM_HEADER_SIZE);
|
---|
118 |
|
---|
119 | // Return if we didn't get the whole Stream Header yet.
|
---|
120 | if (coder->pos < LZMA_STREAM_HEADER_SIZE)
|
---|
121 | return LZMA_OK;
|
---|
122 |
|
---|
123 | coder->pos = 0;
|
---|
124 |
|
---|
125 | // Decode the Stream Header.
|
---|
126 | const lzma_ret ret = lzma_stream_header_decode(
|
---|
127 | &coder->stream_flags, coder->buffer);
|
---|
128 | if (ret != LZMA_OK)
|
---|
129 | return ret == LZMA_FORMAT_ERROR && !coder->first_stream
|
---|
130 | ? LZMA_DATA_ERROR : ret;
|
---|
131 |
|
---|
132 | // If we are decoding concatenated Streams, and the later
|
---|
133 | // Streams have invalid Header Magic Bytes, we give
|
---|
134 | // LZMA_DATA_ERROR instead of LZMA_FORMAT_ERROR.
|
---|
135 | coder->first_stream = false;
|
---|
136 |
|
---|
137 | // Copy the type of the Check so that Block Header and Block
|
---|
138 | // decoders see it.
|
---|
139 | coder->block_options.check = coder->stream_flags.check;
|
---|
140 |
|
---|
141 | // Even if we return LZMA_*_CHECK below, we want
|
---|
142 | // to continue from Block Header decoding.
|
---|
143 | coder->sequence = SEQ_BLOCK_HEADER;
|
---|
144 |
|
---|
145 | // Detect if there's no integrity check or if it is
|
---|
146 | // unsupported if those were requested by the application.
|
---|
147 | if (coder->tell_no_check && coder->stream_flags.check
|
---|
148 | == LZMA_CHECK_NONE)
|
---|
149 | return LZMA_NO_CHECK;
|
---|
150 |
|
---|
151 | if (coder->tell_unsupported_check
|
---|
152 | && !lzma_check_is_supported(
|
---|
153 | coder->stream_flags.check))
|
---|
154 | return LZMA_UNSUPPORTED_CHECK;
|
---|
155 |
|
---|
156 | if (coder->tell_any_check)
|
---|
157 | return LZMA_GET_CHECK;
|
---|
158 | }
|
---|
159 |
|
---|
160 | // Fall through
|
---|
161 |
|
---|
162 | case SEQ_BLOCK_HEADER: {
|
---|
163 | if (*in_pos >= in_size)
|
---|
164 | return LZMA_OK;
|
---|
165 |
|
---|
166 | if (coder->pos == 0) {
|
---|
167 | // Detect if it's Index.
|
---|
168 | if (in[*in_pos] == INDEX_INDICATOR) {
|
---|
169 | coder->sequence = SEQ_INDEX;
|
---|
170 | break;
|
---|
171 | }
|
---|
172 |
|
---|
173 | // Calculate the size of the Block Header. Note that
|
---|
174 | // Block Header decoder wants to see this byte too
|
---|
175 | // so don't advance *in_pos.
|
---|
176 | coder->block_options.header_size
|
---|
177 | = lzma_block_header_size_decode(
|
---|
178 | in[*in_pos]);
|
---|
179 | }
|
---|
180 |
|
---|
181 | // Copy the Block Header to the internal buffer.
|
---|
182 | lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos,
|
---|
183 | coder->block_options.header_size);
|
---|
184 |
|
---|
185 | // Return if we didn't get the whole Block Header yet.
|
---|
186 | if (coder->pos < coder->block_options.header_size)
|
---|
187 | return LZMA_OK;
|
---|
188 |
|
---|
189 | coder->pos = 0;
|
---|
190 | coder->sequence = SEQ_BLOCK_INIT;
|
---|
191 | }
|
---|
192 |
|
---|
193 | // Fall through
|
---|
194 |
|
---|
195 | case SEQ_BLOCK_INIT: {
|
---|
196 | // Checking memusage and doing the initialization needs
|
---|
197 | // its own sequence point because we need to be able to
|
---|
198 | // retry if we return LZMA_MEMLIMIT_ERROR.
|
---|
199 |
|
---|
200 | // Version 1 is needed to support the .ignore_check option.
|
---|
201 | coder->block_options.version = 1;
|
---|
202 |
|
---|
203 | // Set up a buffer to hold the filter chain. Block Header
|
---|
204 | // decoder will initialize all members of this array so
|
---|
205 | // we don't need to do it here.
|
---|
206 | lzma_filter filters[LZMA_FILTERS_MAX + 1];
|
---|
207 | coder->block_options.filters = filters;
|
---|
208 |
|
---|
209 | // Decode the Block Header.
|
---|
210 | return_if_error(lzma_block_header_decode(&coder->block_options,
|
---|
211 | allocator, coder->buffer));
|
---|
212 |
|
---|
213 | // If LZMA_IGNORE_CHECK was used, this flag needs to be set.
|
---|
214 | // It has to be set after lzma_block_header_decode() because
|
---|
215 | // it always resets this to false.
|
---|
216 | coder->block_options.ignore_check = coder->ignore_check;
|
---|
217 |
|
---|
218 | // Check the memory usage limit.
|
---|
219 | const uint64_t memusage = lzma_raw_decoder_memusage(filters);
|
---|
220 | lzma_ret ret;
|
---|
221 |
|
---|
222 | if (memusage == UINT64_MAX) {
|
---|
223 | // One or more unknown Filter IDs.
|
---|
224 | ret = LZMA_OPTIONS_ERROR;
|
---|
225 | } else {
|
---|
226 | // Now we can set coder->memusage since we know that
|
---|
227 | // the filter chain is valid. We don't want
|
---|
228 | // lzma_memusage() to return UINT64_MAX in case of
|
---|
229 | // invalid filter chain.
|
---|
230 | coder->memusage = memusage;
|
---|
231 |
|
---|
232 | if (memusage > coder->memlimit) {
|
---|
233 | // The chain would need too much memory.
|
---|
234 | ret = LZMA_MEMLIMIT_ERROR;
|
---|
235 | } else {
|
---|
236 | // Memory usage is OK.
|
---|
237 | // Initialize the Block decoder.
|
---|
238 | ret = lzma_block_decoder_init(
|
---|
239 | &coder->block_decoder,
|
---|
240 | allocator,
|
---|
241 | &coder->block_options);
|
---|
242 | }
|
---|
243 | }
|
---|
244 |
|
---|
245 | // Free the allocated filter options since they are needed
|
---|
246 | // only to initialize the Block decoder.
|
---|
247 | lzma_filters_free(filters, allocator);
|
---|
248 | coder->block_options.filters = NULL;
|
---|
249 |
|
---|
250 | // Check if memory usage calculation and Block decoder
|
---|
251 | // initialization succeeded.
|
---|
252 | if (ret != LZMA_OK)
|
---|
253 | return ret;
|
---|
254 |
|
---|
255 | coder->sequence = SEQ_BLOCK_RUN;
|
---|
256 | }
|
---|
257 |
|
---|
258 | // Fall through
|
---|
259 |
|
---|
260 | case SEQ_BLOCK_RUN: {
|
---|
261 | const lzma_ret ret = coder->block_decoder.code(
|
---|
262 | coder->block_decoder.coder, allocator,
|
---|
263 | in, in_pos, in_size, out, out_pos, out_size,
|
---|
264 | action);
|
---|
265 |
|
---|
266 | if (ret != LZMA_STREAM_END)
|
---|
267 | return ret;
|
---|
268 |
|
---|
269 | // Block decoded successfully. Add the new size pair to
|
---|
270 | // the Index hash.
|
---|
271 | return_if_error(lzma_index_hash_append(coder->index_hash,
|
---|
272 | lzma_block_unpadded_size(
|
---|
273 | &coder->block_options),
|
---|
274 | coder->block_options.uncompressed_size));
|
---|
275 |
|
---|
276 | coder->sequence = SEQ_BLOCK_HEADER;
|
---|
277 | break;
|
---|
278 | }
|
---|
279 |
|
---|
280 | case SEQ_INDEX: {
|
---|
281 | // If we don't have any input, don't call
|
---|
282 | // lzma_index_hash_decode() since it would return
|
---|
283 | // LZMA_BUF_ERROR, which we must not do here.
|
---|
284 | if (*in_pos >= in_size)
|
---|
285 | return LZMA_OK;
|
---|
286 |
|
---|
287 | // Decode the Index and compare it to the hash calculated
|
---|
288 | // from the sizes of the Blocks (if any).
|
---|
289 | const lzma_ret ret = lzma_index_hash_decode(coder->index_hash,
|
---|
290 | in, in_pos, in_size);
|
---|
291 | if (ret != LZMA_STREAM_END)
|
---|
292 | return ret;
|
---|
293 |
|
---|
294 | coder->sequence = SEQ_STREAM_FOOTER;
|
---|
295 | }
|
---|
296 |
|
---|
297 | // Fall through
|
---|
298 |
|
---|
299 | case SEQ_STREAM_FOOTER: {
|
---|
300 | // Copy the Stream Footer to the internal buffer.
|
---|
301 | lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos,
|
---|
302 | LZMA_STREAM_HEADER_SIZE);
|
---|
303 |
|
---|
304 | // Return if we didn't get the whole Stream Footer yet.
|
---|
305 | if (coder->pos < LZMA_STREAM_HEADER_SIZE)
|
---|
306 | return LZMA_OK;
|
---|
307 |
|
---|
308 | coder->pos = 0;
|
---|
309 |
|
---|
310 | // Decode the Stream Footer. The decoder gives
|
---|
311 | // LZMA_FORMAT_ERROR if the magic bytes don't match,
|
---|
312 | // so convert that return code to LZMA_DATA_ERROR.
|
---|
313 | lzma_stream_flags footer_flags;
|
---|
314 | const lzma_ret ret = lzma_stream_footer_decode(
|
---|
315 | &footer_flags, coder->buffer);
|
---|
316 | if (ret != LZMA_OK)
|
---|
317 | return ret == LZMA_FORMAT_ERROR
|
---|
318 | ? LZMA_DATA_ERROR : ret;
|
---|
319 |
|
---|
320 | // Check that Index Size stored in the Stream Footer matches
|
---|
321 | // the real size of the Index field.
|
---|
322 | if (lzma_index_hash_size(coder->index_hash)
|
---|
323 | != footer_flags.backward_size)
|
---|
324 | return LZMA_DATA_ERROR;
|
---|
325 |
|
---|
326 | // Compare that the Stream Flags fields are identical in
|
---|
327 | // both Stream Header and Stream Footer.
|
---|
328 | return_if_error(lzma_stream_flags_compare(
|
---|
329 | &coder->stream_flags, &footer_flags));
|
---|
330 |
|
---|
331 | if (!coder->concatenated)
|
---|
332 | return LZMA_STREAM_END;
|
---|
333 |
|
---|
334 | coder->sequence = SEQ_STREAM_PADDING;
|
---|
335 | }
|
---|
336 |
|
---|
337 | // Fall through
|
---|
338 |
|
---|
339 | case SEQ_STREAM_PADDING:
|
---|
340 | assert(coder->concatenated);
|
---|
341 |
|
---|
342 | // Skip over possible Stream Padding.
|
---|
343 | while (true) {
|
---|
344 | if (*in_pos >= in_size) {
|
---|
345 | // Unless LZMA_FINISH was used, we cannot
|
---|
346 | // know if there's more input coming later.
|
---|
347 | if (action != LZMA_FINISH)
|
---|
348 | return LZMA_OK;
|
---|
349 |
|
---|
350 | // Stream Padding must be a multiple of
|
---|
351 | // four bytes.
|
---|
352 | return coder->pos == 0
|
---|
353 | ? LZMA_STREAM_END
|
---|
354 | : LZMA_DATA_ERROR;
|
---|
355 | }
|
---|
356 |
|
---|
357 | // If the byte is not zero, it probably indicates
|
---|
358 | // beginning of a new Stream (or the file is corrupt).
|
---|
359 | if (in[*in_pos] != 0x00)
|
---|
360 | break;
|
---|
361 |
|
---|
362 | ++*in_pos;
|
---|
363 | coder->pos = (coder->pos + 1) & 3;
|
---|
364 | }
|
---|
365 |
|
---|
366 | // Stream Padding must be a multiple of four bytes (empty
|
---|
367 | // Stream Padding is OK).
|
---|
368 | if (coder->pos != 0) {
|
---|
369 | ++*in_pos;
|
---|
370 | return LZMA_DATA_ERROR;
|
---|
371 | }
|
---|
372 |
|
---|
373 | // Prepare to decode the next Stream.
|
---|
374 | return_if_error(stream_decoder_reset(coder, allocator));
|
---|
375 | break;
|
---|
376 |
|
---|
377 | default:
|
---|
378 | assert(0);
|
---|
379 | return LZMA_PROG_ERROR;
|
---|
380 | }
|
---|
381 |
|
---|
382 | // Never reached
|
---|
383 | }
|
---|
384 |
|
---|
385 |
|
---|
386 | static void
|
---|
387 | stream_decoder_end(void *coder_ptr, const lzma_allocator *allocator)
|
---|
388 | {
|
---|
389 | lzma_stream_coder *coder = coder_ptr;
|
---|
390 | lzma_next_end(&coder->block_decoder, allocator);
|
---|
391 | lzma_index_hash_end(coder->index_hash, allocator);
|
---|
392 | lzma_free(coder, allocator);
|
---|
393 | return;
|
---|
394 | }
|
---|
395 |
|
---|
396 |
|
---|
397 | static lzma_check
|
---|
398 | stream_decoder_get_check(const void *coder_ptr)
|
---|
399 | {
|
---|
400 | const lzma_stream_coder *coder = coder_ptr;
|
---|
401 | return coder->stream_flags.check;
|
---|
402 | }
|
---|
403 |
|
---|
404 |
|
---|
405 | static lzma_ret
|
---|
406 | stream_decoder_memconfig(void *coder_ptr, uint64_t *memusage,
|
---|
407 | uint64_t *old_memlimit, uint64_t new_memlimit)
|
---|
408 | {
|
---|
409 | lzma_stream_coder *coder = coder_ptr;
|
---|
410 |
|
---|
411 | *memusage = coder->memusage;
|
---|
412 | *old_memlimit = coder->memlimit;
|
---|
413 |
|
---|
414 | if (new_memlimit != 0) {
|
---|
415 | if (new_memlimit < coder->memusage)
|
---|
416 | return LZMA_MEMLIMIT_ERROR;
|
---|
417 |
|
---|
418 | coder->memlimit = new_memlimit;
|
---|
419 | }
|
---|
420 |
|
---|
421 | return LZMA_OK;
|
---|
422 | }
|
---|
423 |
|
---|
424 |
|
---|
425 | extern lzma_ret
|
---|
426 | lzma_stream_decoder_init(
|
---|
427 | lzma_next_coder *next, const lzma_allocator *allocator,
|
---|
428 | uint64_t memlimit, uint32_t flags)
|
---|
429 | {
|
---|
430 | lzma_next_coder_init(&lzma_stream_decoder_init, next, allocator);
|
---|
431 |
|
---|
432 | if (flags & ~LZMA_SUPPORTED_FLAGS)
|
---|
433 | return LZMA_OPTIONS_ERROR;
|
---|
434 |
|
---|
435 | lzma_stream_coder *coder = next->coder;
|
---|
436 | if (coder == NULL) {
|
---|
437 | coder = lzma_alloc(sizeof(lzma_stream_coder), allocator);
|
---|
438 | if (coder == NULL)
|
---|
439 | return LZMA_MEM_ERROR;
|
---|
440 |
|
---|
441 | next->coder = coder;
|
---|
442 | next->code = &stream_decode;
|
---|
443 | next->end = &stream_decoder_end;
|
---|
444 | next->get_check = &stream_decoder_get_check;
|
---|
445 | next->memconfig = &stream_decoder_memconfig;
|
---|
446 |
|
---|
447 | coder->block_decoder = LZMA_NEXT_CODER_INIT;
|
---|
448 | coder->index_hash = NULL;
|
---|
449 | }
|
---|
450 |
|
---|
451 | coder->memlimit = my_max(1, memlimit);
|
---|
452 | coder->memusage = LZMA_MEMUSAGE_BASE;
|
---|
453 | coder->tell_no_check = (flags & LZMA_TELL_NO_CHECK) != 0;
|
---|
454 | coder->tell_unsupported_check
|
---|
455 | = (flags & LZMA_TELL_UNSUPPORTED_CHECK) != 0;
|
---|
456 | coder->tell_any_check = (flags & LZMA_TELL_ANY_CHECK) != 0;
|
---|
457 | coder->ignore_check = (flags & LZMA_IGNORE_CHECK) != 0;
|
---|
458 | coder->concatenated = (flags & LZMA_CONCATENATED) != 0;
|
---|
459 | coder->first_stream = true;
|
---|
460 |
|
---|
461 | return stream_decoder_reset(coder, allocator);
|
---|
462 | }
|
---|
463 |
|
---|
464 |
|
---|
465 | extern LZMA_API(lzma_ret)
|
---|
466 | lzma_stream_decoder(lzma_stream *strm, uint64_t memlimit, uint32_t flags)
|
---|
467 | {
|
---|
468 | lzma_next_strm_init(lzma_stream_decoder_init, strm, memlimit, flags);
|
---|
469 |
|
---|
470 | strm->internal->supported_actions[LZMA_RUN] = true;
|
---|
471 | strm->internal->supported_actions[LZMA_FINISH] = true;
|
---|
472 |
|
---|
473 | return LZMA_OK;
|
---|
474 | }
|
---|