• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * XZ decompressor
3  *
4  * Authors: Lasse Collin <lasse.collin@tukaani.org>
5  *          Igor Pavlov <https://7-zip.org/>
6  *
7  * This file has been put into the public domain.
8  * You can do whatever you want with this file.
9  */
10 
11 #ifndef XZ_H
12 #define XZ_H
13 
14 #ifdef __KERNEL__
15 #	include <linux/stddef.h>
16 #	include <linux/types.h>
17 #else
18 #	include <stddef.h>
19 #	include <stdint.h>
20 #endif
21 
22 #ifdef __cplusplus
23 extern "C" {
24 #endif
25 
26 /* In Linux, this is used to make extern functions static when needed. */
27 #ifndef XZ_EXTERN
28 #	define XZ_EXTERN extern
29 #endif
30 
31 /**
32  * enum xz_mode - Operation mode
33  *
34  * @XZ_SINGLE:              Single-call mode. This uses less RAM than
35  *                          multi-call modes, because the LZMA2
36  *                          dictionary doesn't need to be allocated as
37  *                          part of the decoder state. All required data
38  *                          structures are allocated at initialization,
39  *                          so xz_dec_run() cannot return XZ_MEM_ERROR.
40  * @XZ_PREALLOC:            Multi-call mode with preallocated LZMA2
41  *                          dictionary buffer. All data structures are
42  *                          allocated at initialization, so xz_dec_run()
43  *                          cannot return XZ_MEM_ERROR.
44  * @XZ_DYNALLOC:            Multi-call mode. The LZMA2 dictionary is
45  *                          allocated once the required size has been
46  *                          parsed from the stream headers. If the
47  *                          allocation fails, xz_dec_run() will return
48  *                          XZ_MEM_ERROR.
49  *
50  * It is possible to enable support only for a subset of the above
51  * modes at compile time by defining XZ_DEC_SINGLE, XZ_DEC_PREALLOC,
52  * or XZ_DEC_DYNALLOC. The xz_dec kernel module is always compiled
53  * with support for all operation modes, but the preboot code may
54  * be built with fewer features to minimize code size.
55  */
56 enum xz_mode {
57 	XZ_SINGLE,
58 	XZ_PREALLOC,
59 	XZ_DYNALLOC
60 };
61 
62 /**
63  * enum xz_ret - Return codes
64  * @XZ_OK:                  Everything is OK so far. More input or more
65  *                          output space is required to continue. This
66  *                          return code is possible only in multi-call mode
67  *                          (XZ_PREALLOC or XZ_DYNALLOC).
68  * @XZ_STREAM_END:          Operation finished successfully.
69  * @XZ_UNSUPPORTED_CHECK:   Integrity check type is not supported. Decoding
70  *                          is still possible in multi-call mode by simply
71  *                          calling xz_dec_run() again.
72  *                          Note that this return value is used only if
73  *                          XZ_DEC_ANY_CHECK was defined at build time,
74  *                          which is not used in the kernel. Unsupported
75  *                          check types return XZ_OPTIONS_ERROR if
76  *                          XZ_DEC_ANY_CHECK was not defined at build time.
77  * @XZ_MEM_ERROR:           Allocating memory failed. This return code is
78  *                          possible only if the decoder was initialized
79  *                          with XZ_DYNALLOC. The amount of memory that was
80  *                          tried to be allocated was no more than the
81  *                          dict_max argument given to xz_dec_init().
82  * @XZ_MEMLIMIT_ERROR:      A bigger LZMA2 dictionary would be needed than
83  *                          allowed by the dict_max argument given to
84  *                          xz_dec_init(). This return value is possible
85  *                          only in multi-call mode (XZ_PREALLOC or
86  *                          XZ_DYNALLOC); the single-call mode (XZ_SINGLE)
87  *                          ignores the dict_max argument.
88  * @XZ_FORMAT_ERROR:        File format was not recognized (wrong magic
89  *                          bytes).
90  * @XZ_OPTIONS_ERROR:       This implementation doesn't support the requested
91  *                          compression options. In the decoder this means
92  *                          that the header CRC32 matches, but the header
93  *                          itself specifies something that we don't support.
94  * @XZ_DATA_ERROR:          Compressed data is corrupt.
95  * @XZ_BUF_ERROR:           Cannot make any progress. Details are slightly
96  *                          different between multi-call and single-call
97  *                          mode; more information below.
98  *
99  * In multi-call mode, XZ_BUF_ERROR is returned when two consecutive calls
100  * to XZ code cannot consume any input and cannot produce any new output.
101  * This happens when there is no new input available, or the output buffer
102  * is full while at least one output byte is still pending. Assuming your
103  * code is not buggy, you can get this error only when decoding a compressed
104  * stream that is truncated or otherwise corrupt.
105  *
106  * In single-call mode, XZ_BUF_ERROR is returned only when the output buffer
107  * is too small or the compressed input is corrupt in a way that makes the
108  * decoder produce more output than the caller expected. When it is
109  * (relatively) clear that the compressed input is truncated, XZ_DATA_ERROR
110  * is used instead of XZ_BUF_ERROR.
111  */
112 enum xz_ret {
113 	XZ_OK,
114 	XZ_STREAM_END,
115 	XZ_UNSUPPORTED_CHECK,
116 	XZ_MEM_ERROR,
117 	XZ_MEMLIMIT_ERROR,
118 	XZ_FORMAT_ERROR,
119 	XZ_OPTIONS_ERROR,
120 	XZ_DATA_ERROR,
121 	XZ_BUF_ERROR
122 };
123 
124 /**
125  * struct xz_buf - Passing input and output buffers to XZ code
126  * @in:         Beginning of the input buffer. This may be NULL if and only
127  *              if in_pos is equal to in_size.
128  * @in_pos:     Current position in the input buffer. This must not exceed
129  *              in_size.
130  * @in_size:    Size of the input buffer
131  * @out:        Beginning of the output buffer. This may be NULL if and only
132  *              if out_pos is equal to out_size.
133  * @out_pos:    Current position in the output buffer. This must not exceed
134  *              out_size.
135  * @out_size:   Size of the output buffer
136  *
137  * Only the contents of the output buffer from out[out_pos] onward, and
138  * the variables in_pos and out_pos are modified by the XZ code.
139  */
140 struct xz_buf {
141 	const uint8_t *in;
142 	size_t in_pos;
143 	size_t in_size;
144 
145 	uint8_t *out;
146 	size_t out_pos;
147 	size_t out_size;
148 };
149 
150 /**
151  * struct xz_dec - Opaque type to hold the XZ decoder state
152  */
153 struct xz_dec;
154 
155 /**
156  * xz_dec_init() - Allocate and initialize a XZ decoder state
157  * @mode:       Operation mode
158  * @dict_max:   Maximum size of the LZMA2 dictionary (history buffer) for
159  *              multi-call decoding. This is ignored in single-call mode
160  *              (mode == XZ_SINGLE). LZMA2 dictionary is always 2^n bytes
161  *              or 2^n + 2^(n-1) bytes (the latter sizes are less common
162  *              in practice), so other values for dict_max don't make sense.
163  *              In the kernel, dictionary sizes of 64 KiB, 128 KiB, 256 KiB,
164  *              512 KiB, and 1 MiB are probably the only reasonable values,
165  *              except for kernel and initramfs images where a bigger
166  *              dictionary can be fine and useful.
167  *
168  * Single-call mode (XZ_SINGLE): xz_dec_run() decodes the whole stream at
169  * once. The caller must provide enough output space or the decoding will
170  * fail. The output space is used as the dictionary buffer, which is why
171  * there is no need to allocate the dictionary as part of the decoder's
172  * internal state.
173  *
174  * Because the output buffer is used as the workspace, streams encoded using
175  * a big dictionary are not a problem in single-call mode. It is enough that
176  * the output buffer is big enough to hold the actual uncompressed data; it
177  * can be smaller than the dictionary size stored in the stream headers.
178  *
179  * Multi-call mode with preallocated dictionary (XZ_PREALLOC): dict_max bytes
180  * of memory is preallocated for the LZMA2 dictionary. This way there is no
181  * risk that xz_dec_run() could run out of memory, since xz_dec_run() will
182  * never allocate any memory. Instead, if the preallocated dictionary is too
183  * small for decoding the given input stream, xz_dec_run() will return
184  * XZ_MEMLIMIT_ERROR. Thus, it is important to know what kind of data will be
185  * decoded to avoid allocating excessive amount of memory for the dictionary.
186  *
187  * Multi-call mode with dynamically allocated dictionary (XZ_DYNALLOC):
188  * dict_max specifies the maximum allowed dictionary size that xz_dec_run()
189  * may allocate once it has parsed the dictionary size from the stream
190  * headers. This way excessive allocations can be avoided while still
191  * limiting the maximum memory usage to a sane value to prevent running the
192  * system out of memory when decompressing streams from untrusted sources.
193  *
194  * On success, xz_dec_init() returns a pointer to struct xz_dec, which is
195  * ready to be used with xz_dec_run(). If memory allocation fails,
196  * xz_dec_init() returns NULL.
197  */
198 XZ_EXTERN struct xz_dec *xz_dec_init(enum xz_mode mode, uint32_t dict_max);
199 
200 /**
201  * xz_dec_run() - Run the XZ decoder for a single XZ stream
202  * @s:          Decoder state allocated using xz_dec_init()
203  * @b:          Input and output buffers
204  *
205  * The possible return values depend on build options and operation mode.
206  * See enum xz_ret for details.
207  *
208  * Note that if an error occurs in single-call mode (return value is not
209  * XZ_STREAM_END), b->in_pos and b->out_pos are not modified and the
210  * contents of the output buffer from b->out[b->out_pos] onward are
211  * undefined. This is true even after XZ_BUF_ERROR, because with some filter
212  * chains, there may be a second pass over the output buffer, and this pass
213  * cannot be properly done if the output buffer is truncated. Thus, you
214  * cannot give the single-call decoder a too small buffer and then expect to
215  * get that amount valid data from the beginning of the stream. You must use
216  * the multi-call decoder if you don't want to uncompress the whole stream.
217  *
218  * Use xz_dec_run() when XZ data is stored inside some other file format.
219  * The decoding will stop after one XZ stream has been decompresed. To
220  * decompress regular .xz files which might have multiple concatenated
221  * streams, use xz_dec_catrun() instead.
222  */
223 XZ_EXTERN enum xz_ret xz_dec_run(struct xz_dec *s, struct xz_buf *b);
224 
225 /**
226  * xz_dec_catrun() - Run the XZ decoder with support for concatenated streams
227  * @s:          Decoder state allocated using xz_dec_init()
228  * @b:          Input and output buffers
229  * @finish:     This is an int instead of bool to avoid requiring stdbool.h.
230  *              As long as more input might be coming, finish must be false.
231  *              When the caller knows that it has provided all the input to
232  *              the decoder (some possibly still in b->in), it must set finish
233  *              to true. Only when finish is true can this function return
234  *              XZ_STREAM_END to indicate successful decompression of the
235  *              file. In single-call mode (XZ_SINGLE) finish is assumed to
236  *              always be true; the caller-provided value is ignored.
237  *
238  * This is like xz_dec_run() except that this makes it easy to decode .xz
239  * files with multiple streams (multiple .xz files concatenated as is).
240  * The rarely-used Stream Padding feature is supported too, that is, there
241  * can be null bytes after or between the streams. The number of null bytes
242  * must be a multiple of four.
243  *
244  * When finish is false and b->in_pos == b->in_size, it is possible that
245  * XZ_BUF_ERROR isn't returned even when no progress is possible (XZ_OK is
246  * returned instead). This shouldn't matter because in this situation a
247  * reasonable caller will attempt to provide more input or set finish to
248  * true for the next xz_dec_catrun() call anyway.
249  *
250  * For any struct xz_dec that has been initialized for multi-call mode:
251  * Once decoding has been started with xz_dec_run() or xz_dec_catrun(),
252  * the same function must be used until xz_dec_reset() or xz_dec_end().
253  * Switching between the two decoding functions without resetting results
254  * in undefined behavior.
255  *
256  * xz_dec_catrun() is only available if XZ_DEC_CONCATENATED was defined
257  * at compile time.
258  */
259 XZ_EXTERN enum xz_ret xz_dec_catrun(struct xz_dec *s, struct xz_buf *b,
260 				    int finish);
261 
262 /**
263  * xz_dec_reset() - Reset an already allocated decoder state
264  * @s:          Decoder state allocated using xz_dec_init()
265  *
266  * This function can be used to reset the multi-call decoder state without
267  * freeing and reallocating memory with xz_dec_end() and xz_dec_init().
268  *
269  * In single-call mode, xz_dec_reset() is always called in the beginning of
270  * xz_dec_run(). Thus, explicit call to xz_dec_reset() is useful only in
271  * multi-call mode.
272  */
273 XZ_EXTERN void xz_dec_reset(struct xz_dec *s);
274 
275 /**
276  * xz_dec_end() - Free the memory allocated for the decoder state
277  * @s:          Decoder state allocated using xz_dec_init(). If s is NULL,
278  *              this function does nothing.
279  */
280 XZ_EXTERN void xz_dec_end(struct xz_dec *s);
281 
282 /*
283  * Decompressor for MicroLZMA, an LZMA variant with a very minimal header.
284  * See xz_dec_microlzma_alloc() below for details.
285  *
286  * These functions aren't used or available in preboot code and thus aren't
287  * marked with XZ_EXTERN. This avoids warnings about static functions that
288  * are never defined.
289  */
290 /**
291  * struct xz_dec_microlzma - Opaque type to hold the MicroLZMA decoder state
292  */
293 struct xz_dec_microlzma;
294 
295 /**
296  * xz_dec_microlzma_alloc() - Allocate memory for the MicroLZMA decoder
297  * @mode        XZ_SINGLE or XZ_PREALLOC
298  * @dict_size   LZMA dictionary size. This must be at least 4 KiB and
299  *              at most 3 GiB.
300  *
301  * In contrast to xz_dec_init(), this function only allocates the memory
302  * and remembers the dictionary size. xz_dec_microlzma_reset() must be used
303  * before calling xz_dec_microlzma_run().
304  *
305  * The amount of allocated memory is a little less than 30 KiB with XZ_SINGLE.
306  * With XZ_PREALLOC also a dictionary buffer of dict_size bytes is allocated.
307  *
308  * On success, xz_dec_microlzma_alloc() returns a pointer to
309  * struct xz_dec_microlzma. If memory allocation fails or
310  * dict_size is invalid, NULL is returned.
311  *
312  * The compressed format supported by this decoder is a raw LZMA stream
313  * whose first byte (always 0x00) has been replaced with bitwise-negation
314  * of the LZMA properties (lc/lp/pb) byte. For example, if lc/lp/pb is
315  * 3/0/2, the first byte is 0xA2. This way the first byte can never be 0x00.
316  * Just like with LZMA2, lc + lp <= 4 must be true. The LZMA end-of-stream
317  * marker must not be used. The unused values are reserved for future use.
318  * This MicroLZMA header format was created for use in EROFS but may be used
319  * by others too.
320  */
321 extern struct xz_dec_microlzma *xz_dec_microlzma_alloc(enum xz_mode mode,
322 						       uint32_t dict_size);
323 
324 /**
325  * xz_dec_microlzma_reset() - Reset the MicroLZMA decoder state
326  * @s           Decoder state allocated using xz_dec_microlzma_alloc()
327  * @comp_size   Compressed size of the input stream
328  * @uncomp_size Uncompressed size of the input stream. A value smaller
329  *              than the real uncompressed size of the input stream can
330  *              be specified if uncomp_size_is_exact is set to false.
331  *              uncomp_size can never be set to a value larger than the
332  *              expected real uncompressed size because it would eventually
333  *              result in XZ_DATA_ERROR.
334  * @uncomp_size_is_exact  This is an int instead of bool to avoid
335  *              requiring stdbool.h. This should normally be set to true.
336  *              When this is set to false, error detection is weaker.
337  */
338 extern void xz_dec_microlzma_reset(struct xz_dec_microlzma *s,
339 				   uint32_t comp_size, uint32_t uncomp_size,
340 				   int uncomp_size_is_exact);
341 
342 /**
343  * xz_dec_microlzma_run() - Run the MicroLZMA decoder
344  * @s           Decoder state initialized using xz_dec_microlzma_reset()
345  * @b:          Input and output buffers
346  *
347  * This works similarly to xz_dec_run() with a few important differences.
348  * Only the differences are documented here.
349  *
350  * The only possible return values are XZ_OK, XZ_STREAM_END, and
351  * XZ_DATA_ERROR. This function cannot return XZ_BUF_ERROR: if no progress
352  * is possible due to lack of input data or output space, this function will
353  * keep returning XZ_OK. Thus, the calling code must be written so that it
354  * will eventually provide input and output space matching (or exceeding)
355  * comp_size and uncomp_size arguments given to xz_dec_microlzma_reset().
356  * If the caller cannot do this (for example, if the input file is truncated
357  * or otherwise corrupt), the caller must detect this error by itself to
358  * avoid an infinite loop.
359  *
360  * If the compressed data seems to be corrupt, XZ_DATA_ERROR is returned.
361  * This can happen also when incorrect dictionary, uncompressed, or
362  * compressed sizes have been specified.
363  *
364  * With XZ_PREALLOC only: As an extra feature, b->out may be NULL to skip over
365  * uncompressed data. This way the caller doesn't need to provide a temporary
366  * output buffer for the bytes that will be ignored.
367  *
368  * With XZ_SINGLE only: In contrast to xz_dec_run(), the return value XZ_OK
369  * is also possible and thus XZ_SINGLE is actually a limited multi-call mode.
370  * After XZ_OK the bytes decoded so far may be read from the output buffer.
371  * It is possible to continue decoding but the variables b->out and b->out_pos
372  * MUST NOT be changed by the caller. Increasing the value of b->out_size is
373  * allowed to make more output space available; one doesn't need to provide
374  * space for the whole uncompressed data on the first call. The input buffer
375  * may be changed normally like with XZ_PREALLOC. This way input data can be
376  * provided from non-contiguous memory.
377  */
378 extern enum xz_ret xz_dec_microlzma_run(struct xz_dec_microlzma *s,
379 					struct xz_buf *b);
380 
381 /**
382  * xz_dec_microlzma_end() - Free the memory allocated for the decoder state
383  * @s:          Decoder state allocated using xz_dec_microlzma_alloc().
384  *              If s is NULL, this function does nothing.
385  */
386 extern void xz_dec_microlzma_end(struct xz_dec_microlzma *s);
387 
388 /*
389  * Standalone build (userspace build or in-kernel build for boot time use)
390  * needs a CRC32 implementation. For normal in-kernel use, kernel's own
391  * CRC32 module is used instead, and users of this module don't need to
392  * care about the functions below.
393  */
394 #ifndef XZ_INTERNAL_CRC32
395 #	ifdef __KERNEL__
396 #		define XZ_INTERNAL_CRC32 0
397 #	else
398 #		define XZ_INTERNAL_CRC32 1
399 #	endif
400 #endif
401 
402 /*
403  * If CRC64 support has been enabled with XZ_USE_CRC64, a CRC64
404  * implementation is needed too.
405  */
406 #ifndef XZ_USE_CRC64
407 #	undef XZ_INTERNAL_CRC64
408 #	define XZ_INTERNAL_CRC64 0
409 #endif
410 #ifndef XZ_INTERNAL_CRC64
411 #	ifdef __KERNEL__
412 #		error Using CRC64 in the kernel has not been implemented.
413 #	else
414 #		define XZ_INTERNAL_CRC64 1
415 #	endif
416 #endif
417 
418 #if XZ_INTERNAL_CRC32
419 /*
420  * This must be called before any other xz_* function to initialize
421  * the CRC32 lookup table.
422  */
423 XZ_EXTERN void xz_crc32_init(void);
424 
425 /*
426  * Update CRC32 value using the polynomial from IEEE-802.3. To start a new
427  * calculation, the third argument must be zero. To continue the calculation,
428  * the previously returned value is passed as the third argument.
429  */
430 XZ_EXTERN uint32_t xz_crc32(const uint8_t *buf, size_t size, uint32_t crc);
431 #endif
432 
433 #if XZ_INTERNAL_CRC64
434 /*
435  * This must be called before any other xz_* function (except xz_crc32_init())
436  * to initialize the CRC64 lookup table.
437  */
438 XZ_EXTERN void xz_crc64_init(void);
439 
440 /*
441  * Update CRC64 value using the polynomial from ECMA-182. To start a new
442  * calculation, the third argument must be zero. To continue the calculation,
443  * the previously returned value is passed as the third argument.
444  */
445 XZ_EXTERN uint64_t xz_crc64(const uint8_t *buf, size_t size, uint64_t crc);
446 #endif
447 
448 #ifdef __cplusplus
449 }
450 #endif
451 
452 #endif
453