1 // SPDX-License-Identifier: 0BSD
9 #include "xz_private.h"
10 #include "xz_stream.h"
12 /* Hash used to validate the Index field */
15 vli_type uncompressed;
20 /* Position in dec_main() */
34 /* Position in variable-length integers and Check fields */
37 /* Variable-length integer decoded by dec_vli() */
40 /* Saved in_pos and out_pos */
44 /* CRC32 value in Block or Index */
47 /* Type of the integrity check calculated from uncompressed data */
48 enum xz_check check_type;
54 * True if the next call to xz_dec_run() is allowed to return
59 /* Information stored in Block Header */
62 * Value stored in the Compressed Size field, or
63 * VLI_UNKNOWN if Compressed Size is not present.
68 * Value stored in the Uncompressed Size field, or
69 * VLI_UNKNOWN if Uncompressed Size is not present.
71 vli_type uncompressed;
73 /* Size of the Block Header field */
77 /* Information collected when decoding Blocks */
79 /* Observed compressed size of the current Block */
82 /* Observed uncompressed size of the current Block */
83 vli_type uncompressed;
85 /* Number of Blocks decoded so far */
89 * Hash calculated from the Block sizes. This is used to
90 * validate the Index field.
92 struct xz_dec_hash hash;
95 /* Variables needed when verifying the Index field */
97 /* Position in dec_index() */
101 SEQ_INDEX_UNCOMPRESSED
104 /* Size of the Index in bytes */
107 /* Number of Records (matches block.count in valid files) */
111 * Hash calculated from the Records (matches block.hash in
114 struct xz_dec_hash hash;
118 * Temporary buffer needed to hold Stream Header, Block Header,
119 * and Stream Footer. The Block Header is the biggest (1 KiB)
120 * so we reserve space according to that. buf[] has to be aligned
121 * to a multiple of four bytes; the size_t variables before it
122 * should guarantee this.
130 struct xz_dec_lzma2 *lzma2;
133 struct xz_dec_bcj *bcj;
138 #ifdef XZ_DEC_ANY_CHECK
139 /* Sizes of the Check field with different Check IDs */
140 static const uint8_t check_sizes[16] = {
151 * Fill s->temp by copying data starting from b->in[b->in_pos]. Caller
152 * must have set s->temp.pos to indicate how much data we are supposed
153 * to copy into s->temp.buf. Return true once s->temp.pos has reached
156 static bool fill_temp(struct xz_dec *s, struct xz_buf *b)
158 size_t copy_size = min_t(size_t,
159 b->in_size - b->in_pos, s->temp.size - s->temp.pos);
161 memcpy(s->temp.buf + s->temp.pos, b->in + b->in_pos, copy_size);
162 b->in_pos += copy_size;
163 s->temp.pos += copy_size;
165 if (s->temp.pos == s->temp.size) {
173 /* Decode a variable-length integer (little-endian base-128 encoding) */
174 static enum xz_ret dec_vli(struct xz_dec *s, const uint8_t *in,
175 size_t *in_pos, size_t in_size)
182 while (*in_pos < in_size) {
186 s->vli |= (vli_type)(byte & 0x7F) << s->pos;
188 if ((byte & 0x80) == 0) {
189 /* Don't allow non-minimal encodings. */
190 if (byte == 0 && s->pos != 0)
191 return XZ_DATA_ERROR;
194 return XZ_STREAM_END;
198 if (s->pos == 7 * VLI_BYTES_MAX)
199 return XZ_DATA_ERROR;
206 * Decode the Compressed Data field from a Block. Update and validate
207 * the observed compressed and uncompressed sizes of the Block so that
208 * they don't exceed the values possibly stored in the Block Header
209 * (validation assumes that no integer overflow occurs, since vli_type
210 * is normally uint64_t). Update the CRC32 if presence of the CRC32
211 * field was indicated in Stream Header.
213 * Once the decoding is finished, validate that the observed sizes match
214 * the sizes possibly stored in the Block Header. Update the hash and
215 * Block count, which are later used to validate the Index field.
217 static enum xz_ret dec_block(struct xz_dec *s, struct xz_buf *b)
221 s->in_start = b->in_pos;
222 s->out_start = b->out_pos;
226 ret = xz_dec_bcj_run(s->bcj, s->lzma2, b);
229 ret = xz_dec_lzma2_run(s->lzma2, b);
231 s->block.compressed += b->in_pos - s->in_start;
232 s->block.uncompressed += b->out_pos - s->out_start;
235 * There is no need to separately check for VLI_UNKNOWN, since
236 * the observed sizes are always smaller than VLI_UNKNOWN.
238 if (s->block.compressed > s->block_header.compressed
239 || s->block.uncompressed
240 > s->block_header.uncompressed)
241 return XZ_DATA_ERROR;
243 if (s->check_type == XZ_CHECK_CRC32)
244 s->crc32 = xz_crc32(b->out + s->out_start,
245 b->out_pos - s->out_start, s->crc32);
247 if (ret == XZ_STREAM_END) {
248 if (s->block_header.compressed != VLI_UNKNOWN
249 && s->block_header.compressed
250 != s->block.compressed)
251 return XZ_DATA_ERROR;
253 if (s->block_header.uncompressed != VLI_UNKNOWN
254 && s->block_header.uncompressed
255 != s->block.uncompressed)
256 return XZ_DATA_ERROR;
258 s->block.hash.unpadded += s->block_header.size
259 + s->block.compressed;
261 #ifdef XZ_DEC_ANY_CHECK
262 s->block.hash.unpadded += check_sizes[s->check_type];
264 if (s->check_type == XZ_CHECK_CRC32)
265 s->block.hash.unpadded += 4;
268 s->block.hash.uncompressed += s->block.uncompressed;
269 s->block.hash.crc32 = xz_crc32(
270 (const uint8_t *)&s->block.hash,
271 sizeof(s->block.hash), s->block.hash.crc32);
279 /* Update the Index size and the CRC32 value. */
280 static void index_update(struct xz_dec *s, const struct xz_buf *b)
282 size_t in_used = b->in_pos - s->in_start;
283 s->index.size += in_used;
284 s->crc32 = xz_crc32(b->in + s->in_start, in_used, s->crc32);
288 * Decode the Number of Records, Unpadded Size, and Uncompressed Size
289 * fields from the Index field. That is, Index Padding and CRC32 are not
290 * decoded by this function.
292 * This can return XZ_OK (more input needed), XZ_STREAM_END (everything
293 * successfully decoded), or XZ_DATA_ERROR (input is corrupt).
295 static enum xz_ret dec_index(struct xz_dec *s, struct xz_buf *b)
300 ret = dec_vli(s, b->in, &b->in_pos, b->in_size);
301 if (ret != XZ_STREAM_END) {
306 switch (s->index.sequence) {
307 case SEQ_INDEX_COUNT:
308 s->index.count = s->vli;
311 * Validate that the Number of Records field
312 * indicates the same number of Records as
313 * there were Blocks in the Stream.
315 if (s->index.count != s->block.count)
316 return XZ_DATA_ERROR;
318 s->index.sequence = SEQ_INDEX_UNPADDED;
321 case SEQ_INDEX_UNPADDED:
322 s->index.hash.unpadded += s->vli;
323 s->index.sequence = SEQ_INDEX_UNCOMPRESSED;
326 case SEQ_INDEX_UNCOMPRESSED:
327 s->index.hash.uncompressed += s->vli;
328 s->index.hash.crc32 = xz_crc32(
329 (const uint8_t *)&s->index.hash,
330 sizeof(s->index.hash),
331 s->index.hash.crc32);
333 s->index.sequence = SEQ_INDEX_UNPADDED;
336 } while (s->index.count > 0);
338 return XZ_STREAM_END;
342 * Validate that the next four input bytes match the value of s->crc32.
343 * s->pos must be zero when starting to validate the first byte.
345 static enum xz_ret crc32_validate(struct xz_dec *s, struct xz_buf *b)
348 if (b->in_pos == b->in_size)
351 if (((s->crc32 >> s->pos) & 0xFF) != b->in[b->in_pos++])
352 return XZ_DATA_ERROR;
356 } while (s->pos < 32);
361 return XZ_STREAM_END;
364 #ifdef XZ_DEC_ANY_CHECK
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.
369 static bool check_skip(struct xz_dec *s, struct xz_buf *b)
371 while (s->pos < check_sizes[s->check_type]) {
372 if (b->in_pos == b->in_size)
385 /* Decode the Stream Header field (the first 12 bytes of the .xz Stream). */
386 static enum xz_ret dec_stream_header(struct xz_dec *s)
388 if (!memeq(s->temp.buf, HEADER_MAGIC, HEADER_MAGIC_SIZE))
389 return XZ_FORMAT_ERROR;
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;
395 if (s->temp.buf[HEADER_MAGIC_SIZE] != 0)
396 return XZ_OPTIONS_ERROR;
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.
404 if (s->temp.buf[HEADER_MAGIC_SIZE + 1] > XZ_CHECK_MAX)
405 return XZ_OPTIONS_ERROR;
407 s->check_type = s->temp.buf[HEADER_MAGIC_SIZE + 1];
409 #ifdef XZ_DEC_ANY_CHECK
410 if (s->check_type > XZ_CHECK_CRC32)
411 return XZ_UNSUPPORTED_CHECK;
413 if (s->check_type > XZ_CHECK_CRC32)
414 return XZ_OPTIONS_ERROR;
420 /* Decode the Stream Footer field (the last 12 bytes of the .xz Stream) */
421 static enum xz_ret dec_stream_footer(struct xz_dec *s)
423 if (!memeq(s->temp.buf + 10, FOOTER_MAGIC, FOOTER_MAGIC_SIZE))
424 return XZ_DATA_ERROR;
426 if (xz_crc32(s->temp.buf + 4, 6, 0) != get_le32(s->temp.buf))
427 return XZ_DATA_ERROR;
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.
434 if ((s->index.size >> 2) != get_le32(s->temp.buf + 4))
435 return XZ_DATA_ERROR;
437 if (s->temp.buf[8] != 0 || s->temp.buf[9] != s->check_type)
438 return XZ_DATA_ERROR;
441 * Use XZ_STREAM_END instead of XZ_OK to be more convenient
444 return XZ_STREAM_END;
447 /* Decode the Block Header and initialize the filter chain. */
448 static enum xz_ret dec_block_header(struct xz_dec *s)
453 * Validate the CRC32. We know that the temp buffer is at least
454 * eight bytes so this is safe.
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;
464 * Catch unsupported Block Flags. We support only one or two filters
465 * in the chain, so we catch that with the same test.
468 if (s->temp.buf[1] & 0x3E)
470 if (s->temp.buf[1] & 0x3F)
472 return XZ_OPTIONS_ERROR;
474 /* Compressed Size */
475 if (s->temp.buf[1] & 0x40) {
476 if (dec_vli(s, s->temp.buf, &s->temp.pos, s->temp.size)
478 return XZ_DATA_ERROR;
480 s->block_header.compressed = s->vli;
482 s->block_header.compressed = VLI_UNKNOWN;
485 /* Uncompressed Size */
486 if (s->temp.buf[1] & 0x80) {
487 if (dec_vli(s, s->temp.buf, &s->temp.pos, s->temp.size)
489 return XZ_DATA_ERROR;
491 s->block_header.uncompressed = s->vli;
493 s->block_header.uncompressed = VLI_UNKNOWN;
497 /* If there are two filters, the first one must be a BCJ filter. */
498 s->bcj_active = s->temp.buf[1] & 0x01;
500 if (s->temp.size - s->temp.pos < 2)
501 return XZ_OPTIONS_ERROR;
503 ret = xz_dec_bcj_reset(s->bcj, s->temp.buf[s->temp.pos++]);
508 * We don't support custom start offset,
509 * so Size of Properties must be zero.
511 if (s->temp.buf[s->temp.pos++] != 0x00)
512 return XZ_OPTIONS_ERROR;
516 /* Valid Filter Flags always take at least two bytes. */
517 if (s->temp.size - s->temp.pos < 2)
518 return XZ_DATA_ERROR;
520 /* Filter ID = LZMA2 */
521 if (s->temp.buf[s->temp.pos++] != 0x21)
522 return XZ_OPTIONS_ERROR;
524 /* Size of Properties = 1-byte Filter Properties */
525 if (s->temp.buf[s->temp.pos++] != 0x01)
526 return XZ_OPTIONS_ERROR;
528 /* Filter Properties contains LZMA2 dictionary size. */
529 if (s->temp.size - s->temp.pos < 1)
530 return XZ_DATA_ERROR;
532 ret = xz_dec_lzma2_reset(s->lzma2, s->temp.buf[s->temp.pos++]);
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;
542 s->block.compressed = 0;
543 s->block.uncompressed = 0;
548 static enum xz_ret dec_main(struct xz_dec *s, struct xz_buf *b)
553 * Store the start position for the case when we are in the middle
554 * of the Index field.
556 s->in_start = b->in_pos;
559 switch (s->sequence) {
560 case SEQ_STREAM_HEADER:
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
569 if (!fill_temp(s, b))
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().
579 s->sequence = SEQ_BLOCK_START;
581 ret = dec_stream_header(s);
587 case SEQ_BLOCK_START:
588 /* We need one byte of input to continue. */
589 if (b->in_pos == b->in_size)
592 /* See if this is the beginning of the Index field. */
593 if (b->in[b->in_pos] == 0) {
594 s->in_start = b->in_pos++;
595 s->sequence = SEQ_INDEX;
600 * Calculate the size of the Block Header and
601 * prepare to decode it.
604 = ((uint32_t)b->in[b->in_pos] + 1) * 4;
606 s->temp.size = s->block_header.size;
608 s->sequence = SEQ_BLOCK_HEADER;
612 case SEQ_BLOCK_HEADER:
613 if (!fill_temp(s, b))
616 ret = dec_block_header(s);
620 s->sequence = SEQ_BLOCK_UNCOMPRESS;
624 case SEQ_BLOCK_UNCOMPRESS:
625 ret = dec_block(s, b);
626 if (ret != XZ_STREAM_END)
629 s->sequence = SEQ_BLOCK_PADDING;
633 case SEQ_BLOCK_PADDING:
635 * Size of Compressed Data + Block Padding
636 * must be a multiple of four. We don't need
637 * s->block.compressed for anything else
638 * anymore, so we use it here to test the size
639 * of the Block Padding field.
641 while (s->block.compressed & 3) {
642 if (b->in_pos == b->in_size)
645 if (b->in[b->in_pos++] != 0)
646 return XZ_DATA_ERROR;
648 ++s->block.compressed;
651 s->sequence = SEQ_BLOCK_CHECK;
655 case SEQ_BLOCK_CHECK:
656 if (s->check_type == XZ_CHECK_CRC32) {
657 ret = crc32_validate(s, b);
658 if (ret != XZ_STREAM_END)
661 #ifdef XZ_DEC_ANY_CHECK
662 else if (!check_skip(s, b)) {
667 s->sequence = SEQ_BLOCK_START;
671 ret = dec_index(s, b);
672 if (ret != XZ_STREAM_END)
675 s->sequence = SEQ_INDEX_PADDING;
679 case SEQ_INDEX_PADDING:
680 while ((s->index.size + (b->in_pos - s->in_start))
682 if (b->in_pos == b->in_size) {
687 if (b->in[b->in_pos++] != 0)
688 return XZ_DATA_ERROR;
691 /* Finish the CRC32 value and Index size. */
694 /* Compare the hashes to validate the Index field. */
695 if (!memeq(&s->block.hash, &s->index.hash,
696 sizeof(s->block.hash)))
697 return XZ_DATA_ERROR;
699 s->sequence = SEQ_INDEX_CRC32;
703 case SEQ_INDEX_CRC32:
704 ret = crc32_validate(s, b);
705 if (ret != XZ_STREAM_END)
708 s->temp.size = STREAM_HEADER_SIZE;
709 s->sequence = SEQ_STREAM_FOOTER;
713 case SEQ_STREAM_FOOTER:
714 if (!fill_temp(s, b))
717 return dec_stream_footer(s);
725 * xz_dec_run() is a wrapper for dec_main() to handle some special cases in
726 * multi-call and single-call decoding.
728 * In multi-call mode, we must return XZ_BUF_ERROR when it seems clear that we
729 * are not going to make any progress anymore. This is to prevent the caller
730 * from calling us infinitely when the input file is truncated or otherwise
731 * corrupt. Since zlib-style API allows that the caller fills the input buffer
732 * only when the decoder doesn't produce any new output, we have to be careful
733 * to avoid returning XZ_BUF_ERROR too easily: XZ_BUF_ERROR is returned only
734 * after the second consecutive call to xz_dec_run() that makes no progress.
736 * In single-call mode, if we couldn't decode everything and no error
737 * occurred, either the input is truncated or the output buffer is too small.
738 * Since we know that the last input byte never produces any output, we know
739 * that if all the input was consumed and decoding wasn't finished, the file
740 * must be corrupt. Otherwise the output buffer has to be too small or the
741 * file is corrupt in a way that decoding it produces too big output.
743 * If single-call decoding fails, we reset b->in_pos and b->out_pos back to
744 * their original values. This is because with some filter chains there won't
745 * be any valid uncompressed data in the output buffer unless the decoding
746 * actually succeeds (that's the price to pay of using the output buffer as
749 enum xz_ret xz_dec_run(struct xz_dec *s, struct xz_buf *b)
755 if (DEC_IS_SINGLE(s->mode))
758 in_start = b->in_pos;
759 out_start = b->out_pos;
760 ret = dec_main(s, b);
762 if (DEC_IS_SINGLE(s->mode)) {
764 ret = b->in_pos == b->in_size
765 ? XZ_DATA_ERROR : XZ_BUF_ERROR;
767 if (ret != XZ_STREAM_END) {
768 b->in_pos = in_start;
769 b->out_pos = out_start;
772 } else if (ret == XZ_OK && in_start == b->in_pos
773 && out_start == b->out_pos) {
774 if (s->allow_buf_error)
777 s->allow_buf_error = true;
779 s->allow_buf_error = false;
785 struct xz_dec *xz_dec_init(enum xz_mode mode, uint32_t dict_max)
787 struct xz_dec *s = kmalloc(sizeof(*s), GFP_KERNEL);
794 s->bcj = xz_dec_bcj_create(DEC_IS_SINGLE(mode));
799 s->lzma2 = xz_dec_lzma2_create(mode, dict_max);
800 if (s->lzma2 == NULL)
808 xz_dec_bcj_end(s->bcj);
815 void xz_dec_reset(struct xz_dec *s)
817 s->sequence = SEQ_STREAM_HEADER;
818 s->allow_buf_error = false;
821 memzero(&s->block, sizeof(s->block));
822 memzero(&s->index, sizeof(s->index));
824 s->temp.size = STREAM_HEADER_SIZE;
827 void xz_dec_end(struct xz_dec *s)
830 xz_dec_lzma2_end(s->lzma2);
832 xz_dec_bcj_end(s->bcj);