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