1 /*
2  * .xz Stream decoder
3  *
4  * Author: Lasse Collin <lasse.collin@tukaani.org>
5  *
6  * This file has been put into the public domain.
7  * You can do whatever you want with this file.
8  */
9 
10 #include "xz_private.h"
11 #include "xz_stream.h"
12 
13 /* Hash used to validate the Index field */
14 struct xz_dec_hash {
15 	vli_type unpadded;
16 	vli_type uncompressed;
17 	uint32_t crc32;
18 };
19 
20 struct xz_dec {
21 	/* Position in dec_main() */
22 	enum {
23 		SEQ_STREAM_HEADER,
24 		SEQ_BLOCK_START,
25 		SEQ_BLOCK_HEADER,
26 		SEQ_BLOCK_UNCOMPRESS,
27 		SEQ_BLOCK_PADDING,
28 		SEQ_BLOCK_CHECK,
29 		SEQ_INDEX,
30 		SEQ_INDEX_PADDING,
31 		SEQ_INDEX_CRC32,
32 		SEQ_STREAM_FOOTER
33 	} sequence;
34 
35 	/* Position in variable-length integers and Check fields */
36 	uint32_t pos;
37 
38 	/* Variable-length integer decoded by dec_vli() */
39 	vli_type vli;
40 
41 	/* Saved in_pos and out_pos */
42 	size_t in_start;
43 	size_t out_start;
44 
45 	/* CRC32 value in Block or Index */
46 	uint32_t crc32;
47 
48 	/* Type of the integrity check calculated from uncompressed data */
49 	enum xz_check check_type;
50 
51 	/* Operation mode */
52 	enum xz_mode mode;
53 
54 	/*
55 	 * True if the next call to xz_dec_run() is allowed to return
56 	 * XZ_BUF_ERROR.
57 	 */
58 	bool allow_buf_error;
59 
60 	/* Information stored in Block Header */
61 	struct {
62 		/*
63 		 * Value stored in the Compressed Size field, or
64 		 * VLI_UNKNOWN if Compressed Size is not present.
65 		 */
66 		vli_type compressed;
67 
68 		/*
69 		 * Value stored in the Uncompressed Size field, or
70 		 * VLI_UNKNOWN if Uncompressed Size is not present.
71 		 */
72 		vli_type uncompressed;
73 
74 		/* Size of the Block Header field */
75 		uint32_t size;
76 	} block_header;
77 
78 	/* Information collected when decoding Blocks */
79 	struct {
80 		/* Observed compressed size of the current Block */
81 		vli_type compressed;
82 
83 		/* Observed uncompressed size of the current Block */
84 		vli_type uncompressed;
85 
86 		/* Number of Blocks decoded so far */
87 		vli_type count;
88 
89 		/*
90 		 * Hash calculated from the Block sizes. This is used to
91 		 * validate the Index field.
92 		 */
93 		struct xz_dec_hash hash;
94 	} block;
95 
96 	/* Variables needed when verifying the Index field */
97 	struct {
98 		/* Position in dec_index() */
99 		enum {
100 			SEQ_INDEX_COUNT,
101 			SEQ_INDEX_UNPADDED,
102 			SEQ_INDEX_UNCOMPRESSED
103 		} sequence;
104 
105 		/* Size of the Index in bytes */
106 		vli_type size;
107 
108 		/* Number of Records (matches block.count in valid files) */
109 		vli_type count;
110 
111 		/*
112 		 * Hash calculated from the Records (matches block.hash in
113 		 * valid files).
114 		 */
115 		struct xz_dec_hash hash;
116 	} index;
117 
118 	/*
119 	 * Temporary buffer needed to hold Stream Header, Block Header,
120 	 * and Stream Footer. The Block Header is the biggest (1 KiB)
121 	 * so we reserve space according to that. buf[] has to be aligned
122 	 * to a multiple of four bytes; the size_t variables before it
123 	 * should guarantee this.
124 	 */
125 	struct {
126 		size_t pos;
127 		size_t size;
128 		uint8_t buf[1024];
129 	} temp;
130 
131 	struct xz_dec_lzma2 *lzma2;
132 
133 #ifdef XZ_DEC_BCJ
134 	struct xz_dec_bcj *bcj;
135 	bool bcj_active;
136 #endif
137 };
138 
139 #ifdef XZ_DEC_ANY_CHECK
140 /* Sizes of the Check field with different Check IDs */
141 static const uint8_t check_sizes[16] = {
142 	0,
143 	4, 4, 4,
144 	8, 8, 8,
145 	16, 16, 16,
146 	32, 32, 32,
147 	64, 64, 64
148 };
149 #endif
150 
151 /*
152  * Fill s->temp by copying data starting from b->in[b->in_pos]. Caller
153  * must have set s->temp.pos to indicate how much data we are supposed
154  * to copy into s->temp.buf. Return true once s->temp.pos has reached
155  * s->temp.size.
156  */
fill_temp(struct xz_dec * s,struct xz_buf * b)157 static bool XZ_FUNC fill_temp(struct xz_dec *s, struct xz_buf *b)
158 {
159 	size_t copy_size = min_t(size_t,
160 			b->in_size - b->in_pos, s->temp.size - s->temp.pos);
161 
162 	memcpy(s->temp.buf + s->temp.pos, b->in + b->in_pos, copy_size);
163 	b->in_pos += copy_size;
164 	s->temp.pos += copy_size;
165 
166 	if (s->temp.pos == s->temp.size) {
167 		s->temp.pos = 0;
168 		return true;
169 	}
170 
171 	return false;
172 }
173 
174 /* Decode a variable-length integer (little-endian base-128 encoding) */
dec_vli(struct xz_dec * s,const uint8_t * in,size_t * in_pos,size_t in_size)175 static enum xz_ret XZ_FUNC dec_vli(struct xz_dec *s,
176 		const uint8_t *in, size_t *in_pos, size_t in_size)
177 {
178 	uint8_t byte;
179 
180 	if (s->pos == 0)
181 		s->vli = 0;
182 
183 	while (*in_pos < in_size) {
184 		byte = in[*in_pos];
185 		++*in_pos;
186 
187 		s->vli |= (vli_type)(byte & 0x7F) << s->pos;
188 
189 		if ((byte & 0x80) == 0) {
190 			/* Don't allow non-minimal encodings. */
191 			if (byte == 0 && s->pos != 0)
192 				return XZ_DATA_ERROR;
193 
194 			s->pos = 0;
195 			return XZ_STREAM_END;
196 		}
197 
198 		s->pos += 7;
199 		if (s->pos == 7 * VLI_BYTES_MAX)
200 			return XZ_DATA_ERROR;
201 	}
202 
203 	return XZ_OK;
204 }
205 
206 /*
207  * Decode the Compressed Data field from a Block. Update and validate
208  * the observed compressed and uncompressed sizes of the Block so that
209  * they don't exceed the values possibly stored in the Block Header
210  * (validation assumes that no integer overflow occurs, since vli_type
211  * is normally uint64_t). Update the CRC32 if presence of the CRC32
212  * field was indicated in Stream Header.
213  *
214  * Once the decoding is finished, validate that the observed sizes match
215  * the sizes possibly stored in the Block Header. Update the hash and
216  * Block count, which are later used to validate the Index field.
217  */
dec_block(struct xz_dec * s,struct xz_buf * b)218 static enum xz_ret XZ_FUNC dec_block(struct xz_dec *s, struct xz_buf *b)
219 {
220 	enum xz_ret ret;
221 
222 	s->in_start = b->in_pos;
223 	s->out_start = b->out_pos;
224 
225 #ifdef XZ_DEC_BCJ
226 	if (s->bcj_active)
227 		ret = xz_dec_bcj_run(s->bcj, s->lzma2, b);
228 	else
229 #endif
230 		ret = xz_dec_lzma2_run(s->lzma2, b);
231 
232 	s->block.compressed += b->in_pos - s->in_start;
233 	s->block.uncompressed += b->out_pos - s->out_start;
234 
235 	/*
236 	 * There is no need to separately check for VLI_UNKNOWN, since
237 	 * the observed sizes are always smaller than VLI_UNKNOWN.
238 	 */
239 	if (s->block.compressed > s->block_header.compressed
240 			|| s->block.uncompressed
241 				> s->block_header.uncompressed)
242 		return XZ_DATA_ERROR;
243 
244 	if (s->check_type == XZ_CHECK_CRC32)
245 		s->crc32 = xz_crc32(b->out + s->out_start,
246 				b->out_pos - s->out_start, s->crc32);
247 
248 	if (ret == XZ_STREAM_END) {
249 		if (s->block_header.compressed != VLI_UNKNOWN
250 				&& s->block_header.compressed
251 					!= s->block.compressed)
252 			return XZ_DATA_ERROR;
253 
254 		if (s->block_header.uncompressed != VLI_UNKNOWN
255 				&& s->block_header.uncompressed
256 					!= s->block.uncompressed)
257 			return XZ_DATA_ERROR;
258 
259 		s->block.hash.unpadded += s->block_header.size
260 				+ s->block.compressed;
261 
262 #ifdef XZ_DEC_ANY_CHECK
263 		s->block.hash.unpadded += check_sizes[s->check_type];
264 #else
265 		if (s->check_type == XZ_CHECK_CRC32)
266 			s->block.hash.unpadded += 4;
267 #endif
268 
269 		s->block.hash.uncompressed += s->block.uncompressed;
270 		s->block.hash.crc32 = xz_crc32(
271 				(const uint8_t *)&s->block.hash,
272 				sizeof(s->block.hash), s->block.hash.crc32);
273 
274 		++s->block.count;
275 	}
276 
277 	return ret;
278 }
279 
280 /* Update the Index size and the CRC32 value. */
index_update(struct xz_dec * s,const struct xz_buf * b)281 static void XZ_FUNC index_update(struct xz_dec *s, const struct xz_buf *b)
282 {
283 	size_t in_used = b->in_pos - s->in_start;
284 	s->index.size += in_used;
285 	s->crc32 = xz_crc32(b->in + s->in_start, in_used, s->crc32);
286 }
287 
288 /*
289  * Decode the Number of Records, Unpadded Size, and Uncompressed Size
290  * fields from the Index field. That is, Index Padding and CRC32 are not
291  * decoded by this function.
292  *
293  * This can return XZ_OK (more input needed), XZ_STREAM_END (everything
294  * successfully decoded), or XZ_DATA_ERROR (input is corrupt).
295  */
dec_index(struct xz_dec * s,struct xz_buf * b)296 static enum xz_ret XZ_FUNC dec_index(struct xz_dec *s, struct xz_buf *b)
297 {
298 	enum xz_ret ret;
299 
300 	do {
301 		ret = dec_vli(s, b->in, &b->in_pos, b->in_size);
302 		if (ret != XZ_STREAM_END) {
303 			index_update(s, b);
304 			return ret;
305 		}
306 
307 		switch (s->index.sequence) {
308 		case SEQ_INDEX_COUNT:
309 			s->index.count = s->vli;
310 
311 			/*
312 			 * Validate that the Number of Records field
313 			 * indicates the same number of Records as
314 			 * there were Blocks in the Stream.
315 			 */
316 			if (s->index.count != s->block.count)
317 				return XZ_DATA_ERROR;
318 
319 			s->index.sequence = SEQ_INDEX_UNPADDED;
320 			break;
321 
322 		case SEQ_INDEX_UNPADDED:
323 			s->index.hash.unpadded += s->vli;
324 			s->index.sequence = SEQ_INDEX_UNCOMPRESSED;
325 			break;
326 
327 		case SEQ_INDEX_UNCOMPRESSED:
328 			s->index.hash.uncompressed += s->vli;
329 			s->index.hash.crc32 = xz_crc32(
330 					(const uint8_t *)&s->index.hash,
331 					sizeof(s->index.hash),
332 					s->index.hash.crc32);
333 			--s->index.count;
334 			s->index.sequence = SEQ_INDEX_UNPADDED;
335 			break;
336 		}
337 	} while (s->index.count > 0);
338 
339 	return XZ_STREAM_END;
340 }
341 
342 /*
343  * Validate that the next four input bytes match the value of s->crc32.
344  * s->pos must be zero when starting to validate the first byte.
345  */
crc32_validate(struct xz_dec * s,struct xz_buf * b)346 static enum xz_ret XZ_FUNC crc32_validate(struct xz_dec *s, struct xz_buf *b)
347 {
348 	do {
349 		if (b->in_pos == b->in_size)
350 			return XZ_OK;
351 
352 		if (((s->crc32 >> s->pos) & 0xFF) != b->in[b->in_pos++])
353 			return XZ_DATA_ERROR;
354 
355 		s->pos += 8;
356 	} while (s->pos < 32);
357 
358 	s->crc32 = 0;
359 	s->pos = 0;
360 
361 	return XZ_STREAM_END;
362 }
363 
364 #ifdef XZ_DEC_ANY_CHECK
365 /*
366  * Skip over the Check field when the Check ID is not supported.
367  * Returns true once the whole Check field has been skipped over.
368  */
check_skip(struct xz_dec * s,struct xz_buf * b)369 static bool XZ_FUNC check_skip(struct xz_dec *s, struct xz_buf *b)
370 {
371 	while (s->pos < check_sizes[s->check_type]) {
372 		if (b->in_pos == b->in_size)
373 			return false;
374 
375 		++b->in_pos;
376 		++s->pos;
377 	}
378 
379 	s->pos = 0;
380 
381 	return true;
382 }
383 #endif
384 
385 /* Decode the Stream Header field (the first 12 bytes of the .xz Stream). */
dec_stream_header(struct xz_dec * s)386 static enum xz_ret XZ_FUNC dec_stream_header(struct xz_dec *s)
387 {
388 	if (!memeq(s->temp.buf, HEADER_MAGIC, HEADER_MAGIC_SIZE))
389 		return XZ_FORMAT_ERROR;
390 
391 	if (xz_crc32(s->temp.buf + HEADER_MAGIC_SIZE, 2, 0)
392 			!= get_le32(s->temp.buf + HEADER_MAGIC_SIZE + 2))
393 		return XZ_DATA_ERROR;
394 
395 	if (s->temp.buf[HEADER_MAGIC_SIZE] != 0)
396 		return XZ_OPTIONS_ERROR;
397 
398 	/*
399 	 * Of integrity checks, we support only none (Check ID = 0) and
400 	 * CRC32 (Check ID = 1). However, if XZ_DEC_ANY_CHECK is defined,
401 	 * we will accept other check types too, but then the check won't
402 	 * be verified and a warning (XZ_UNSUPPORTED_CHECK) will be given.
403 	 */
404 	s->check_type = s->temp.buf[HEADER_MAGIC_SIZE + 1];
405 
406 #ifdef XZ_DEC_ANY_CHECK
407 	if (s->check_type > XZ_CHECK_MAX)
408 		return XZ_OPTIONS_ERROR;
409 
410 	if (s->check_type > XZ_CHECK_CRC32)
411 		return XZ_UNSUPPORTED_CHECK;
412 #else
413 	if (s->check_type > XZ_CHECK_CRC32)
414 		return XZ_OPTIONS_ERROR;
415 #endif
416 
417 	return XZ_OK;
418 }
419 
420 /* Decode the Stream Footer field (the last 12 bytes of the .xz Stream) */
dec_stream_footer(struct xz_dec * s)421 static enum xz_ret XZ_FUNC dec_stream_footer(struct xz_dec *s)
422 {
423 	if (!memeq(s->temp.buf + 10, FOOTER_MAGIC, FOOTER_MAGIC_SIZE))
424 		return XZ_DATA_ERROR;
425 
426 	if (xz_crc32(s->temp.buf + 4, 6, 0) != get_le32(s->temp.buf))
427 		return XZ_DATA_ERROR;
428 
429 	/*
430 	 * Validate Backward Size. Note that we never added the size of the
431 	 * Index CRC32 field to s->index.size, thus we use s->index.size / 4
432 	 * instead of s->index.size / 4 - 1.
433 	 */
434 	if ((s->index.size >> 2) != get_le32(s->temp.buf + 4))
435 		return XZ_DATA_ERROR;
436 
437 	if (s->temp.buf[8] != 0 || s->temp.buf[9] != s->check_type)
438 		return XZ_DATA_ERROR;
439 
440 	/*
441 	 * Use XZ_STREAM_END instead of XZ_OK to be more convenient
442 	 * for the caller.
443 	 */
444 	return XZ_STREAM_END;
445 }
446 
447 /* Decode the Block Header and initialize the filter chain. */
dec_block_header(struct xz_dec * s)448 static enum xz_ret XZ_FUNC dec_block_header(struct xz_dec *s)
449 {
450 	enum xz_ret ret;
451 
452 	/*
453 	 * Validate the CRC32. We know that the temp buffer is at least
454 	 * eight bytes so this is safe.
455 	 */
456 	s->temp.size -= 4;
457 	if (xz_crc32(s->temp.buf, s->temp.size, 0)
458 			!= get_le32(s->temp.buf + s->temp.size))
459 		return XZ_DATA_ERROR;
460 
461 	s->temp.pos = 2;
462 
463 	/*
464 	 * Catch unsupported Block Flags. We support only one or two filters
465 	 * in the chain, so we catch that with the same test.
466 	 */
467 #ifdef XZ_DEC_BCJ
468 	if (s->temp.buf[1] & 0x3E)
469 #else
470 	if (s->temp.buf[1] & 0x3F)
471 #endif
472 		return XZ_OPTIONS_ERROR;
473 
474 	/* Compressed Size */
475 	if (s->temp.buf[1] & 0x40) {
476 		if (dec_vli(s, s->temp.buf, &s->temp.pos, s->temp.size)
477 					!= XZ_STREAM_END)
478 			return XZ_DATA_ERROR;
479 
480 		s->block_header.compressed = s->vli;
481 	} else {
482 		s->block_header.compressed = VLI_UNKNOWN;
483 	}
484 
485 	/* Uncompressed Size */
486 	if (s->temp.buf[1] & 0x80) {
487 		if (dec_vli(s, s->temp.buf, &s->temp.pos, s->temp.size)
488 				!= XZ_STREAM_END)
489 			return XZ_DATA_ERROR;
490 
491 		s->block_header.uncompressed = s->vli;
492 	} else {
493 		s->block_header.uncompressed = VLI_UNKNOWN;
494 	}
495 
496 #ifdef XZ_DEC_BCJ
497 	/* If there are two filters, the first one must be a BCJ filter. */
498 	s->bcj_active = s->temp.buf[1] & 0x01;
499 	if (s->bcj_active) {
500 		if (s->temp.size - s->temp.pos < 2)
501 			return XZ_OPTIONS_ERROR;
502 
503 		ret = xz_dec_bcj_reset(s->bcj, s->temp.buf[s->temp.pos++]);
504 		if (ret != XZ_OK)
505 			return ret;
506 
507 		/*
508 		 * We don't support custom start offset,
509 		 * so Size of Properties must be zero.
510 		 */
511 		if (s->temp.buf[s->temp.pos++] != 0x00)
512 			return XZ_OPTIONS_ERROR;
513 	}
514 #endif
515 
516 	/* Valid Filter Flags always take at least two bytes. */
517 	if (s->temp.size - s->temp.pos < 2)
518 		return XZ_DATA_ERROR;
519 
520 	/* Filter ID = LZMA2 */
521 	if (s->temp.buf[s->temp.pos++] != 0x21)
522 		return XZ_OPTIONS_ERROR;
523 
524 	/* Size of Properties = 1-byte Filter Properties */
525 	if (s->temp.buf[s->temp.pos++] != 0x01)
526 		return XZ_OPTIONS_ERROR;
527 
528 	/* Filter Properties contains LZMA2 dictionary size. */
529 	if (s->temp.size - s->temp.pos < 1)
530 		return XZ_DATA_ERROR;
531 
532 	ret = xz_dec_lzma2_reset(s->lzma2, s->temp.buf[s->temp.pos++]);
533 	if (ret != XZ_OK)
534 		return ret;
535 
536 	/* The rest must be Header Padding. */
537 	while (s->temp.pos < s->temp.size)
538 		if (s->temp.buf[s->temp.pos++] != 0x00)
539 			return XZ_OPTIONS_ERROR;
540 
541 	s->temp.pos = 0;
542 	s->block.compressed = 0;
543 	s->block.uncompressed = 0;
544 
545 	return XZ_OK;
546 }
547 
dec_main(struct xz_dec * s,struct xz_buf * b)548 static NOINLINE enum xz_ret XZ_FUNC dec_main(struct xz_dec *s, struct xz_buf *b)
549 {
550 	enum xz_ret ret;
551 
552 	/*
553 	 * Store the start position for the case when we are in the middle
554 	 * of the Index field.
555 	 */
556 	s->in_start = b->in_pos;
557 
558 	while (true) {
559 		switch (s->sequence) {
560 		case SEQ_STREAM_HEADER:
561 			/*
562 			 * Stream Header is copied to s->temp, and then
563 			 * decoded from there. This way if the caller
564 			 * gives us only little input at a time, we can
565 			 * still keep the Stream Header decoding code
566 			 * simple. Similar approach is used in many places
567 			 * in this file.
568 			 */
569 			if (!fill_temp(s, b))
570 				return XZ_OK;
571 
572 			/*
573 			 * If dec_stream_header() returns
574 			 * XZ_UNSUPPORTED_CHECK, it is still possible
575 			 * to continue decoding if working in multi-call
576 			 * mode. Thus, update s->sequence before calling
577 			 * dec_stream_header().
578 			 */
579 			s->sequence = SEQ_BLOCK_START;
580 
581 			ret = dec_stream_header(s);
582 			if (ret != XZ_OK)
583 				return ret;
584 
585 		case SEQ_BLOCK_START:
586 			/* We need one byte of input to continue. */
587 			if (b->in_pos == b->in_size)
588 				return XZ_OK;
589 
590 			/* See if this is the beginning of the Index field. */
591 			if (b->in[b->in_pos] == 0) {
592 				s->in_start = b->in_pos++;
593 				s->sequence = SEQ_INDEX;
594 				break;
595 			}
596 
597 			/*
598 			 * Calculate the size of the Block Header and
599 			 * prepare to decode it.
600 			 */
601 			s->block_header.size
602 				= ((uint32_t)b->in[b->in_pos] + 1) * 4;
603 
604 			s->temp.size = s->block_header.size;
605 			s->temp.pos = 0;
606 			s->sequence = SEQ_BLOCK_HEADER;
607 
608 		case SEQ_BLOCK_HEADER:
609 			if (!fill_temp(s, b))
610 				return XZ_OK;
611 
612 			ret = dec_block_header(s);
613 			if (ret != XZ_OK)
614 				return ret;
615 
616 			s->sequence = SEQ_BLOCK_UNCOMPRESS;
617 
618 		case SEQ_BLOCK_UNCOMPRESS:
619 			ret = dec_block(s, b);
620 			if (ret != XZ_STREAM_END)
621 				return ret;
622 
623 			s->sequence = SEQ_BLOCK_PADDING;
624 
625 		case SEQ_BLOCK_PADDING:
626 			/*
627 			 * Size of Compressed Data + Block Padding
628 			 * must be a multiple of four. We don't need
629 			 * s->block.compressed for anything else
630 			 * anymore, so we use it here to test the size
631 			 * of the Block Padding field.
632 			 */
633 			while (s->block.compressed & 3) {
634 				if (b->in_pos == b->in_size)
635 					return XZ_OK;
636 
637 				if (b->in[b->in_pos++] != 0)
638 					return XZ_DATA_ERROR;
639 
640 				++s->block.compressed;
641 			}
642 
643 			s->sequence = SEQ_BLOCK_CHECK;
644 
645 		case SEQ_BLOCK_CHECK:
646 			if (s->check_type == XZ_CHECK_CRC32) {
647 				ret = crc32_validate(s, b);
648 				if (ret != XZ_STREAM_END)
649 					return ret;
650 			}
651 #ifdef XZ_DEC_ANY_CHECK
652 			else if (!check_skip(s, b)) {
653 				return XZ_OK;
654 			}
655 #endif
656 
657 			s->sequence = SEQ_BLOCK_START;
658 			break;
659 
660 		case SEQ_INDEX:
661 			ret = dec_index(s, b);
662 			if (ret != XZ_STREAM_END)
663 				return ret;
664 
665 			s->sequence = SEQ_INDEX_PADDING;
666 
667 		case SEQ_INDEX_PADDING:
668 			while ((s->index.size + (b->in_pos - s->in_start))
669 					& 3) {
670 				if (b->in_pos == b->in_size) {
671 					index_update(s, b);
672 					return XZ_OK;
673 				}
674 
675 				if (b->in[b->in_pos++] != 0)
676 					return XZ_DATA_ERROR;
677 			}
678 
679 			/* Finish the CRC32 value and Index size. */
680 			index_update(s, b);
681 
682 			/* Compare the hashes to validate the Index field. */
683 			if (!memeq(&s->block.hash, &s->index.hash,
684 					sizeof(s->block.hash)))
685 				return XZ_DATA_ERROR;
686 
687 			s->sequence = SEQ_INDEX_CRC32;
688 
689 		case SEQ_INDEX_CRC32:
690 			ret = crc32_validate(s, b);
691 			if (ret != XZ_STREAM_END)
692 				return ret;
693 
694 			s->temp.size = STREAM_HEADER_SIZE;
695 			s->sequence = SEQ_STREAM_FOOTER;
696 
697 		case SEQ_STREAM_FOOTER:
698 			if (!fill_temp(s, b))
699 				return XZ_OK;
700 
701 			return dec_stream_footer(s);
702 		}
703 	}
704 
705 	/* Never reached */
706 }
707 
708 /*
709  * xz_dec_run() is a wrapper for dec_main() to handle some special cases in
710  * multi-call and single-call decoding.
711  *
712  * In multi-call mode, we must return XZ_BUF_ERROR when it seems clear that we
713  * are not going to make any progress anymore. This is to prevent the caller
714  * from calling us infinitely when the input file is truncated or otherwise
715  * corrupt. Since zlib-style API allows that the caller fills the input buffer
716  * only when the decoder doesn't produce any new output, we have to be careful
717  * to avoid returning XZ_BUF_ERROR too easily: XZ_BUF_ERROR is returned only
718  * after the second consecutive call to xz_dec_run() that makes no progress.
719  *
720  * In single-call mode, if we couldn't decode everything and no error
721  * occurred, either the input is truncated or the output buffer is too small.
722  * Since we know that the last input byte never produces any output, we know
723  * that if all the input was consumed and decoding wasn't finished, the file
724  * must be corrupt. Otherwise the output buffer has to be too small or the
725  * file is corrupt in a way that decoding it produces too big output.
726  *
727  * If single-call decoding fails, we reset b->in_pos and b->out_pos back to
728  * their original values. This is because with some filter chains there won't
729  * be any valid uncompressed data in the output buffer unless the decoding
730  * actually succeeds (that's the price to pay of using the output buffer as
731  * the workspace).
732  */
xz_dec_run(struct xz_dec * s,struct xz_buf * b)733 XZ_EXTERN enum xz_ret XZ_FUNC xz_dec_run(struct xz_dec *s, struct xz_buf *b)
734 {
735 	size_t in_start;
736 	size_t out_start;
737 	enum xz_ret ret;
738 
739 	if (DEC_IS_SINGLE(s->mode))
740 		xz_dec_reset(s);
741 
742 	in_start = b->in_pos;
743 	out_start = b->out_pos;
744 	ret = dec_main(s, b);
745 
746 	if (DEC_IS_SINGLE(s->mode)) {
747 		if (ret == XZ_OK)
748 			ret = b->in_pos == b->in_size
749 					? XZ_DATA_ERROR : XZ_BUF_ERROR;
750 
751 		if (ret != XZ_STREAM_END) {
752 			b->in_pos = in_start;
753 			b->out_pos = out_start;
754 		}
755 	} else if (ret == XZ_OK && in_start == b->in_pos
756 			&& out_start == b->out_pos) {
757 		if (s->allow_buf_error)
758 			ret = XZ_BUF_ERROR;
759 
760 		s->allow_buf_error = true;
761 	} else {
762 		s->allow_buf_error = false;
763 	}
764 
765 	return ret;
766 }
767 
xz_dec_init(enum xz_mode mode,uint32_t dict_max)768 XZ_EXTERN struct xz_dec * XZ_FUNC xz_dec_init(
769 		enum xz_mode mode, uint32_t dict_max)
770 {
771 	struct xz_dec *s = kmalloc(sizeof(*s), GFP_KERNEL);
772 	if (s == NULL)
773 		return NULL;
774 
775 	s->mode = mode;
776 
777 #ifdef XZ_DEC_BCJ
778 	s->bcj = xz_dec_bcj_create(DEC_IS_SINGLE(mode));
779 	if (s->bcj == NULL)
780 		goto error_bcj;
781 #endif
782 
783 	s->lzma2 = xz_dec_lzma2_create(mode, dict_max);
784 	if (s->lzma2 == NULL)
785 		goto error_lzma2;
786 
787 	xz_dec_reset(s);
788 	return s;
789 
790 error_lzma2:
791 #ifdef XZ_DEC_BCJ
792 	xz_dec_bcj_end(s->bcj);
793 error_bcj:
794 #endif
795 	kfree(s);
796 	return NULL;
797 }
798 
xz_dec_reset(struct xz_dec * s)799 XZ_EXTERN void XZ_FUNC xz_dec_reset(struct xz_dec *s)
800 {
801 	s->sequence = SEQ_STREAM_HEADER;
802 	s->allow_buf_error = false;
803 	s->pos = 0;
804 	s->crc32 = 0;
805 	memzero(&s->block, sizeof(s->block));
806 	memzero(&s->index, sizeof(s->index));
807 	s->temp.pos = 0;
808 	s->temp.size = STREAM_HEADER_SIZE;
809 }
810 
xz_dec_end(struct xz_dec * s)811 XZ_EXTERN void XZ_FUNC xz_dec_end(struct xz_dec *s)
812 {
813 	if (s != NULL) {
814 		xz_dec_lzma2_end(s->lzma2);
815 #ifdef XZ_DEC_BCJ
816 		xz_dec_bcj_end(s->bcj);
817 #endif
818 		kfree(s);
819 	}
820 }
821