wok-current view linux/stuff/001-squashfs-decompressors-add-xz-decompressor-module.patch @ rev 9354
Up: phpmyadmin to 3.3.10.
| author | Christopher Rogers <slaxemulator@gmail.com> |
|---|---|
| date | Sat Mar 19 21:41:08 2011 +0000 (2011-03-19) |
| parents | |
| children |
line source
1 From: Lasse Collin <lasse.collin@tukaani.org>
2 Date: Thu, 2 Dec 2010 19:14:19 +0000 (+0200)
3 Subject: Decompressors: Add XZ decompressor module
4 X-Git-Url: http://git.kernel.org/?p=linux%2Fkernel%2Fgit%2Fpkl%2Fsquashfs-xz.git;a=commitdiff_plain;h=3dbc3fe7878e53b43064a12d4ab31ca4c18ce85f
6 Decompressors: Add XZ decompressor module
8 In userspace, the .lzma format has become mostly a legacy
9 file format that got superseded by the .xz format. Similarly,
10 LZMA Utils was superseded by XZ Utils.
12 These patches add support for XZ decompression into
13 the kernel. Most of the code is as is from XZ Embedded
14 <http://tukaani.org/xz/embedded.html>. It was written for
15 the Linux kernel but is usable in other projects too.
17 Advantages of XZ over the current LZMA code in the kernel:
18 - Nice API that can be used by other kernel modules; it's
19 not limited to kernel, initramfs, and initrd decompression.
20 - Integrity check support (CRC32)
21 - BCJ filters improve compression of executable code on
22 certain architectures. These together with LZMA2 can
23 produce a few percent smaller kernel or Squashfs images
24 than plain LZMA without making the decompression slower.
26 This patch: Add the main decompression code (xz_dec), testing
27 module (xz_dec_test), wrapper script (xz_wrap.sh) for the xz
28 command line tool, and documentation. The xz_dec module is
29 enough to have a usable XZ decompressor e.g. for Squashfs.
31 Signed-off-by: Lasse Collin <lasse.collin@tukaani.org>
32 ---
34 diff --git a/Documentation/xz.txt b/Documentation/xz.txt
35 new file mode 100644
36 index 0000000..68329ac
37 --- /dev/null
38 +++ b/Documentation/xz.txt
39 @@ -0,0 +1,122 @@
40 +
41 +XZ data compression in Linux
42 +============================
43 +
44 +Introduction
45 +
46 + XZ is a general purpose data compression format with high compression
47 + ratio and relatively fast decompression. The primary compression
48 + algorithm (filter) is LZMA2. Additional filters can be used to improve
49 + compression ratio even further. E.g. Branch/Call/Jump (BCJ) filters
50 + improve compression ratio of executable data.
51 +
52 + The XZ decompressor in Linux is called XZ Embedded. It supports
53 + the LZMA2 filter and optionally also BCJ filters. CRC32 is supported
54 + for integrity checking. The home page of XZ Embedded is at
55 + <http://tukaani.org/xz/embedded.html>, where you can find the
56 + latest version and also information about using the code outside
57 + the Linux kernel.
58 +
59 + For userspace, XZ Utils provide a zlib-like compression library
60 + and a gzip-like command line tool. XZ Utils can be downloaded from
61 + <http://tukaani.org/xz/>.
62 +
63 +XZ related components in the kernel
64 +
65 + The xz_dec module provides XZ decompressor with single-call (buffer
66 + to buffer) and multi-call (stateful) APIs. The usage of the xz_dec
67 + module is documented in include/linux/xz.h.
68 +
69 + The xz_dec_test module is for testing xz_dec. xz_dec_test is not
70 + useful unless you are hacking the XZ decompressor. xz_dec_test
71 + allocates a char device major dynamically to which one can write
72 + .xz files from userspace. The decompressed output is thrown away.
73 + Keep an eye on dmesg to see diagnostics printed by xz_dec_test.
74 + See the xz_dec_test source code for the details.
75 +
76 + For decompressing the kernel image, initramfs, and initrd, there
77 + is a wrapper function in lib/decompress_unxz.c. Its API is the
78 + same as in other decompress_*.c files, which is defined in
79 + include/linux/decompress/generic.h.
80 +
81 + scripts/xz_wrap.sh is a wrapper for the xz command line tool found
82 + from XZ Utils. The wrapper sets compression options to values suitable
83 + for compressing the kernel image.
84 +
85 + For kernel makefiles, two commands are provided for use with
86 + $(call if_needed). The kernel image should be compressed with
87 + $(call if_needed,xzkern) which will use a BCJ filter and a big LZMA2
88 + dictionary. It will also append a four-byte trailer containing the
89 + uncompressed size of the file, which is needed by the boot code.
90 + Other things should be compressed with $(call if_needed,xzmisc)
91 + which will use no BCJ filter and 1 MiB LZMA2 dictionary.
92 +
93 +Notes on compression options
94 +
95 + Since the XZ Embedded supports only streams with no integrity check or
96 + CRC32, make sure that you don't use some other integrity check type
97 + when encoding files that are supposed to be decoded by the kernel. With
98 + liblzma, you need to use either LZMA_CHECK_NONE or LZMA_CHECK_CRC32
99 + when encoding. With the xz command line tool, use --check=none or
100 + --check=crc32.
101 +
102 + Using CRC32 is strongly recommended unless there is some other layer
103 + which will verify the integrity of the uncompressed data anyway.
104 + Double checking the integrity would probably be waste of CPU cycles.
105 + Note that the headers will always have a CRC32 which will be validated
106 + by the decoder; you can only change the integrity check type (or
107 + disable it) for the actual uncompressed data.
108 +
109 + In userspace, LZMA2 is typically used with dictionary sizes of several
110 + megabytes. The decoder needs to have the dictionary in RAM, thus big
111 + dictionaries cannot be used for files that are intended to be decoded
112 + by the kernel. 1 MiB is probably the maximum reasonable dictionary
113 + size for in-kernel use (maybe more is OK for initramfs). The presets
114 + in XZ Utils may not be optimal when creating files for the kernel,
115 + so don't hesitate to use custom settings. Example:
116 +
117 + xz --check=crc32 --lzma2=dict=512KiB inputfile
118 +
119 + An exception to above dictionary size limitation is when the decoder
120 + is used in single-call mode. Decompressing the kernel itself is an
121 + example of this situation. In single-call mode, the memory usage
122 + doesn't depend on the dictionary size, and it is perfectly fine to
123 + use a big dictionary: for maximum compression, the dictionary should
124 + be at least as big as the uncompressed data itself.
125 +
126 +Future plans
127 +
128 + Creating a limited XZ encoder may be considered if people think it is
129 + useful. LZMA2 is slower to compress than e.g. Deflate or LZO even at
130 + the fastest settings, so it isn't clear if LZMA2 encoder is wanted
131 + into the kernel.
132 +
133 + Support for limited random-access reading is planned for the
134 + decompression code. I don't know if it could have any use in the
135 + kernel, but I know that it would be useful in some embedded projects
136 + outside the Linux kernel.
137 +
138 +Conformance to the .xz file format specification
139 +
140 + There are a couple of corner cases where things have been simplified
141 + at expense of detecting errors as early as possible. These should not
142 + matter in practice all, since they don't cause security issues. But
143 + it is good to know this if testing the code e.g. with the test files
144 + from XZ Utils.
145 +
146 +Reporting bugs
147 +
148 + Before reporting a bug, please check that it's not fixed already
149 + at upstream. See <http://tukaani.org/xz/embedded.html> to get the
150 + latest code.
151 +
152 + Report bugs to <lasse.collin@tukaani.org> or visit #tukaani on
153 + Freenode and talk to Larhzu. I don't actively read LKML or other
154 + kernel-related mailing lists, so if there's something I should know,
155 + you should email to me personally or use IRC.
156 +
157 + Don't bother Igor Pavlov with questions about the XZ implementation
158 + in the kernel or about XZ Utils. While these two implementations
159 + include essential code that is directly based on Igor Pavlov's code,
160 + these implementations aren't maintained nor supported by him.
161 +
162 diff --git a/include/linux/xz.h b/include/linux/xz.h
163 new file mode 100644
164 index 0000000..64cffa6
165 --- /dev/null
166 +++ b/include/linux/xz.h
167 @@ -0,0 +1,264 @@
168 +/*
169 + * XZ decompressor
170 + *
171 + * Authors: Lasse Collin <lasse.collin@tukaani.org>
172 + * Igor Pavlov <http://7-zip.org/>
173 + *
174 + * This file has been put into the public domain.
175 + * You can do whatever you want with this file.
176 + */
177 +
178 +#ifndef XZ_H
179 +#define XZ_H
180 +
181 +#ifdef __KERNEL__
182 +# include <linux/stddef.h>
183 +# include <linux/types.h>
184 +#else
185 +# include <stddef.h>
186 +# include <stdint.h>
187 +#endif
188 +
189 +/* In Linux, this is used to make extern functions static when needed. */
190 +#ifndef XZ_EXTERN
191 +# define XZ_EXTERN extern
192 +#endif
193 +
194 +/**
195 + * enum xz_mode - Operation mode
196 + *
197 + * @XZ_SINGLE: Single-call mode. This uses less RAM than
198 + * than multi-call modes, because the LZMA2
199 + * dictionary doesn't need to be allocated as
200 + * part of the decoder state. All required data
201 + * structures are allocated at initialization,
202 + * so xz_dec_run() cannot return XZ_MEM_ERROR.
203 + * @XZ_PREALLOC: Multi-call mode with preallocated LZMA2
204 + * dictionary buffer. All data structures are
205 + * allocated at initialization, so xz_dec_run()
206 + * cannot return XZ_MEM_ERROR.
207 + * @XZ_DYNALLOC: Multi-call mode. The LZMA2 dictionary is
208 + * allocated once the required size has been
209 + * parsed from the stream headers. If the
210 + * allocation fails, xz_dec_run() will return
211 + * XZ_MEM_ERROR.
212 + *
213 + * It is possible to enable support only for a subset of the above
214 + * modes at compile time by defining XZ_DEC_SINGLE, XZ_DEC_PREALLOC,
215 + * or XZ_DEC_DYNALLOC. The xz_dec kernel module is always compiled
216 + * with support for all operation modes, but the preboot code may
217 + * be built with fewer features to minimize code size.
218 + */
219 +enum xz_mode {
220 + XZ_SINGLE,
221 + XZ_PREALLOC,
222 + XZ_DYNALLOC
223 +};
224 +
225 +/**
226 + * enum xz_ret - Return codes
227 + * @XZ_OK: Everything is OK so far. More input or more
228 + * output space is required to continue. This
229 + * return code is possible only in multi-call mode
230 + * (XZ_PREALLOC or XZ_DYNALLOC).
231 + * @XZ_STREAM_END: Operation finished successfully.
232 + * @XZ_UNSUPPORTED_CHECK: Integrity check type is not supported. Decoding
233 + * is still possible in multi-call mode by simply
234 + * calling xz_dec_run() again.
235 + * Note that this return value is used only if
236 + * XZ_DEC_ANY_CHECK was defined at build time,
237 + * which is not used in the kernel. Unsupported
238 + * check types return XZ_OPTIONS_ERROR if
239 + * XZ_DEC_ANY_CHECK was not defined at build time.
240 + * @XZ_MEM_ERROR: Allocating memory failed. This return code is
241 + * possible only if the decoder was initialized
242 + * with XZ_DYNALLOC. The amount of memory that was
243 + * tried to be allocated was no more than the
244 + * dict_max argument given to xz_dec_init().
245 + * @XZ_MEMLIMIT_ERROR: A bigger LZMA2 dictionary would be needed than
246 + * allowed by the dict_max argument given to
247 + * xz_dec_init(). This return value is possible
248 + * only in multi-call mode (XZ_PREALLOC or
249 + * XZ_DYNALLOC); the single-call mode (XZ_SINGLE)
250 + * ignores the dict_max argument.
251 + * @XZ_FORMAT_ERROR: File format was not recognized (wrong magic
252 + * bytes).
253 + * @XZ_OPTIONS_ERROR: This implementation doesn't support the requested
254 + * compression options. In the decoder this means
255 + * that the header CRC32 matches, but the header
256 + * itself specifies something that we don't support.
257 + * @XZ_DATA_ERROR: Compressed data is corrupt.
258 + * @XZ_BUF_ERROR: Cannot make any progress. Details are slightly
259 + * different between multi-call and single-call
260 + * mode; more information below.
261 + *
262 + * In multi-call mode, XZ_BUF_ERROR is returned when two consecutive calls
263 + * to XZ code cannot consume any input and cannot produce any new output.
264 + * This happens when there is no new input available, or the output buffer
265 + * is full while at least one output byte is still pending. Assuming your
266 + * code is not buggy, you can get this error only when decoding a compressed
267 + * stream that is truncated or otherwise corrupt.
268 + *
269 + * In single-call mode, XZ_BUF_ERROR is returned only when the output buffer
270 + * is too small or the compressed input is corrupt in a way that makes the
271 + * decoder produce more output than the caller expected. When it is
272 + * (relatively) clear that the compressed input is truncated, XZ_DATA_ERROR
273 + * is used instead of XZ_BUF_ERROR.
274 + */
275 +enum xz_ret {
276 + XZ_OK,
277 + XZ_STREAM_END,
278 + XZ_UNSUPPORTED_CHECK,
279 + XZ_MEM_ERROR,
280 + XZ_MEMLIMIT_ERROR,
281 + XZ_FORMAT_ERROR,
282 + XZ_OPTIONS_ERROR,
283 + XZ_DATA_ERROR,
284 + XZ_BUF_ERROR
285 +};
286 +
287 +/**
288 + * struct xz_buf - Passing input and output buffers to XZ code
289 + * @in: Beginning of the input buffer. This may be NULL if and only
290 + * if in_pos is equal to in_size.
291 + * @in_pos: Current position in the input buffer. This must not exceed
292 + * in_size.
293 + * @in_size: Size of the input buffer
294 + * @out: Beginning of the output buffer. This may be NULL if and only
295 + * if out_pos is equal to out_size.
296 + * @out_pos: Current position in the output buffer. This must not exceed
297 + * out_size.
298 + * @out_size: Size of the output buffer
299 + *
300 + * Only the contents of the output buffer from out[out_pos] onward, and
301 + * the variables in_pos and out_pos are modified by the XZ code.
302 + */
303 +struct xz_buf {
304 + const uint8_t *in;
305 + size_t in_pos;
306 + size_t in_size;
307 +
308 + uint8_t *out;
309 + size_t out_pos;
310 + size_t out_size;
311 +};
312 +
313 +/**
314 + * struct xz_dec - Opaque type to hold the XZ decoder state
315 + */
316 +struct xz_dec;
317 +
318 +/**
319 + * xz_dec_init() - Allocate and initialize a XZ decoder state
320 + * @mode: Operation mode
321 + * @dict_max: Maximum size of the LZMA2 dictionary (history buffer) for
322 + * multi-call decoding. This is ignored in single-call mode
323 + * (mode == XZ_SINGLE). LZMA2 dictionary is always 2^n bytes
324 + * or 2^n + 2^(n-1) bytes (the latter sizes are less common
325 + * in practice), so other values for dict_max don't make sense.
326 + * In the kernel, dictionary sizes of 64 KiB, 128 KiB, 256 KiB,
327 + * 512 KiB, and 1 MiB are probably the only reasonable values,
328 + * except for kernel and initramfs images where a bigger
329 + * dictionary can be fine and useful.
330 + *
331 + * Single-call mode (XZ_SINGLE): xz_dec_run() decodes the whole stream at
332 + * once. The caller must provide enough output space or the decoding will
333 + * fail. The output space is used as the dictionary buffer, which is why
334 + * there is no need to allocate the dictionary as part of the decoder's
335 + * internal state.
336 + *
337 + * Because the output buffer is used as the workspace, streams encoded using
338 + * a big dictionary are not a problem in single-call mode. It is enough that
339 + * the output buffer is big enough to hold the actual uncompressed data; it
340 + * can be smaller than the dictionary size stored in the stream headers.
341 + *
342 + * Multi-call mode with preallocated dictionary (XZ_PREALLOC): dict_max bytes
343 + * of memory is preallocated for the LZMA2 dictionary. This way there is no
344 + * risk that xz_dec_run() could run out of memory, since xz_dec_run() will
345 + * never allocate any memory. Instead, if the preallocated dictionary is too
346 + * small for decoding the given input stream, xz_dec_run() will return
347 + * XZ_MEMLIMIT_ERROR. Thus, it is important to know what kind of data will be
348 + * decoded to avoid allocating excessive amount of memory for the dictionary.
349 + *
350 + * Multi-call mode with dynamically allocated dictionary (XZ_DYNALLOC):
351 + * dict_max specifies the maximum allowed dictionary size that xz_dec_run()
352 + * may allocate once it has parsed the dictionary size from the stream
353 + * headers. This way excessive allocations can be avoided while still
354 + * limiting the maximum memory usage to a sane value to prevent running the
355 + * system out of memory when decompressing streams from untrusted sources.
356 + *
357 + * On success, xz_dec_init() returns a pointer to struct xz_dec, which is
358 + * ready to be used with xz_dec_run(). If memory allocation fails,
359 + * xz_dec_init() returns NULL.
360 + */
361 +XZ_EXTERN struct xz_dec *xz_dec_init(enum xz_mode mode, uint32_t dict_max);
362 +
363 +/**
364 + * xz_dec_run() - Run the XZ decoder
365 + * @s: Decoder state allocated using xz_dec_init()
366 + * @b: Input and output buffers
367 + *
368 + * The possible return values depend on build options and operation mode.
369 + * See enum xz_ret for details.
370 + *
371 + * Note that if an error occurs in single-call mode (return value is not
372 + * XZ_STREAM_END), b->in_pos and b->out_pos are not modified and the
373 + * contents of the output buffer from b->out[b->out_pos] onward are
374 + * undefined. This is true even after XZ_BUF_ERROR, because with some filter
375 + * chains, there may be a second pass over the output buffer, and this pass
376 + * cannot be properly done if the output buffer is truncated. Thus, you
377 + * cannot give the single-call decoder a too small buffer and then expect to
378 + * get that amount valid data from the beginning of the stream. You must use
379 + * the multi-call decoder if you don't want to uncompress the whole stream.
380 + */
381 +XZ_EXTERN enum xz_ret xz_dec_run(struct xz_dec *s, struct xz_buf *b);
382 +
383 +/**
384 + * xz_dec_reset() - Reset an already allocated decoder state
385 + * @s: Decoder state allocated using xz_dec_init()
386 + *
387 + * This function can be used to reset the multi-call decoder state without
388 + * freeing and reallocating memory with xz_dec_end() and xz_dec_init().
389 + *
390 + * In single-call mode, xz_dec_reset() is always called in the beginning of
391 + * xz_dec_run(). Thus, explicit call to xz_dec_reset() is useful only in
392 + * multi-call mode.
393 + */
394 +XZ_EXTERN void xz_dec_reset(struct xz_dec *s);
395 +
396 +/**
397 + * xz_dec_end() - Free the memory allocated for the decoder state
398 + * @s: Decoder state allocated using xz_dec_init(). If s is NULL,
399 + * this function does nothing.
400 + */
401 +XZ_EXTERN void xz_dec_end(struct xz_dec *s);
402 +
403 +/*
404 + * Standalone build (userspace build or in-kernel build for boot time use)
405 + * needs a CRC32 implementation. For normal in-kernel use, kernel's own
406 + * CRC32 module is used instead, and users of this module don't need to
407 + * care about the functions below.
408 + */
409 +#ifndef XZ_INTERNAL_CRC32
410 +# ifdef __KERNEL__
411 +# define XZ_INTERNAL_CRC32 0
412 +# else
413 +# define XZ_INTERNAL_CRC32 1
414 +# endif
415 +#endif
416 +
417 +#if XZ_INTERNAL_CRC32
418 +/*
419 + * This must be called before any other xz_* function to initialize
420 + * the CRC32 lookup table.
421 + */
422 +XZ_EXTERN void xz_crc32_init(void);
423 +
424 +/*
425 + * Update CRC32 value using the polynomial from IEEE-802.3. To start a new
426 + * calculation, the third argument must be zero. To continue the calculation,
427 + * the previously returned value is passed as the third argument.
428 + */
429 +XZ_EXTERN uint32_t xz_crc32(const uint8_t *buf, size_t size, uint32_t crc);
430 +#endif
431 +#endif
432 diff --git a/lib/Kconfig b/lib/Kconfig
433 index fa9bf2c..6090314 100644
434 --- a/lib/Kconfig
435 +++ b/lib/Kconfig
436 @@ -106,6 +106,8 @@ config LZO_COMPRESS
437 config LZO_DECOMPRESS
438 tristate
440 +source "lib/xz/Kconfig"
441 +
442 #
443 # These all provide a common interface (hence the apparent duplication with
444 # ZLIB_INFLATE; DECOMPRESS_GZIP is just a wrapper.)
445 diff --git a/lib/Makefile b/lib/Makefile
446 index e6a3763..f2f98dd 100644
447 --- a/lib/Makefile
448 +++ b/lib/Makefile
449 @@ -69,6 +69,7 @@ obj-$(CONFIG_ZLIB_DEFLATE) += zlib_deflate/
450 obj-$(CONFIG_REED_SOLOMON) += reed_solomon/
451 obj-$(CONFIG_LZO_COMPRESS) += lzo/
452 obj-$(CONFIG_LZO_DECOMPRESS) += lzo/
453 +obj-$(CONFIG_XZ_DEC) += xz/
454 obj-$(CONFIG_RAID6_PQ) += raid6/
456 lib-$(CONFIG_DECOMPRESS_GZIP) += decompress_inflate.o
457 diff --git a/lib/xz/Kconfig b/lib/xz/Kconfig
458 new file mode 100644
459 index 0000000..e3b6e18
460 --- /dev/null
461 +++ b/lib/xz/Kconfig
462 @@ -0,0 +1,59 @@
463 +config XZ_DEC
464 + tristate "XZ decompression support"
465 + select CRC32
466 + help
467 + LZMA2 compression algorithm and BCJ filters are supported using
468 + the .xz file format as the container. For integrity checking,
469 + CRC32 is supported. See Documentation/xz.txt for more information.
470 +
471 +config XZ_DEC_X86
472 + bool "x86 BCJ filter decoder" if EMBEDDED
473 + default y
474 + depends on XZ_DEC
475 + select XZ_DEC_BCJ
476 +
477 +config XZ_DEC_POWERPC
478 + bool "PowerPC BCJ filter decoder" if EMBEDDED
479 + default y
480 + depends on XZ_DEC
481 + select XZ_DEC_BCJ
482 +
483 +config XZ_DEC_IA64
484 + bool "IA-64 BCJ filter decoder" if EMBEDDED
485 + default y
486 + depends on XZ_DEC
487 + select XZ_DEC_BCJ
488 +
489 +config XZ_DEC_ARM
490 + bool "ARM BCJ filter decoder" if EMBEDDED
491 + default y
492 + depends on XZ_DEC
493 + select XZ_DEC_BCJ
494 +
495 +config XZ_DEC_ARMTHUMB
496 + bool "ARM-Thumb BCJ filter decoder" if EMBEDDED
497 + default y
498 + depends on XZ_DEC
499 + select XZ_DEC_BCJ
500 +
501 +config XZ_DEC_SPARC
502 + bool "SPARC BCJ filter decoder" if EMBEDDED
503 + default y
504 + depends on XZ_DEC
505 + select XZ_DEC_BCJ
506 +
507 +config XZ_DEC_BCJ
508 + bool
509 + default n
510 +
511 +config XZ_DEC_TEST
512 + tristate "XZ decompressor tester"
513 + default n
514 + depends on XZ_DEC
515 + help
516 + This allows passing .xz files to the in-kernel XZ decoder via
517 + a character special file. It calculates CRC32 of the decompressed
518 + data and writes diagnostics to the system log.
519 +
520 + Unless you are developing the XZ decoder, you don't need this
521 + and should say N.
522 diff --git a/lib/xz/Makefile b/lib/xz/Makefile
523 new file mode 100644
524 index 0000000..a7fa769
525 --- /dev/null
526 +++ b/lib/xz/Makefile
527 @@ -0,0 +1,5 @@
528 +obj-$(CONFIG_XZ_DEC) += xz_dec.o
529 +xz_dec-y := xz_dec_syms.o xz_dec_stream.o xz_dec_lzma2.o
530 +xz_dec-$(CONFIG_XZ_DEC_BCJ) += xz_dec_bcj.o
531 +
532 +obj-$(CONFIG_XZ_DEC_TEST) += xz_dec_test.o
533 diff --git a/lib/xz/xz_crc32.c b/lib/xz/xz_crc32.c
534 new file mode 100644
535 index 0000000..34532d1
536 --- /dev/null
537 +++ b/lib/xz/xz_crc32.c
538 @@ -0,0 +1,59 @@
539 +/*
540 + * CRC32 using the polynomial from IEEE-802.3
541 + *
542 + * Authors: Lasse Collin <lasse.collin@tukaani.org>
543 + * Igor Pavlov <http://7-zip.org/>
544 + *
545 + * This file has been put into the public domain.
546 + * You can do whatever you want with this file.
547 + */
548 +
549 +/*
550 + * This is not the fastest implementation, but it is pretty compact.
551 + * The fastest versions of xz_crc32() on modern CPUs without hardware
552 + * accelerated CRC instruction are 3-5 times as fast as this version,
553 + * but they are bigger and use more memory for the lookup table.
554 + */
555 +
556 +#include "xz_private.h"
557 +
558 +/*
559 + * STATIC_RW_DATA is used in the pre-boot environment on some architectures.
560 + * See <linux/decompress/mm.h> for details.
561 + */
562 +#ifndef STATIC_RW_DATA
563 +# define STATIC_RW_DATA static
564 +#endif
565 +
566 +STATIC_RW_DATA uint32_t xz_crc32_table[256];
567 +
568 +XZ_EXTERN void xz_crc32_init(void)
569 +{
570 + const uint32_t poly = 0xEDB88320;
571 +
572 + uint32_t i;
573 + uint32_t j;
574 + uint32_t r;
575 +
576 + for (i = 0; i < 256; ++i) {
577 + r = i;
578 + for (j = 0; j < 8; ++j)
579 + r = (r >> 1) ^ (poly & ~((r & 1) - 1));
580 +
581 + xz_crc32_table[i] = r;
582 + }
583 +
584 + return;
585 +}
586 +
587 +XZ_EXTERN uint32_t xz_crc32(const uint8_t *buf, size_t size, uint32_t crc)
588 +{
589 + crc = ~crc;
590 +
591 + while (size != 0) {
592 + crc = xz_crc32_table[*buf++ ^ (crc & 0xFF)] ^ (crc >> 8);
593 + --size;
594 + }
595 +
596 + return ~crc;
597 +}
598 diff --git a/lib/xz/xz_dec_bcj.c b/lib/xz/xz_dec_bcj.c
599 new file mode 100644
600 index 0000000..e51e255
601 --- /dev/null
602 +++ b/lib/xz/xz_dec_bcj.c
603 @@ -0,0 +1,561 @@
604 +/*
605 + * Branch/Call/Jump (BCJ) filter decoders
606 + *
607 + * Authors: Lasse Collin <lasse.collin@tukaani.org>
608 + * Igor Pavlov <http://7-zip.org/>
609 + *
610 + * This file has been put into the public domain.
611 + * You can do whatever you want with this file.
612 + */
613 +
614 +#include "xz_private.h"
615 +
616 +/*
617 + * The rest of the file is inside this ifdef. It makes things a little more
618 + * convenient when building without support for any BCJ filters.
619 + */
620 +#ifdef XZ_DEC_BCJ
621 +
622 +struct xz_dec_bcj {
623 + /* Type of the BCJ filter being used */
624 + enum {
625 + BCJ_X86 = 4, /* x86 or x86-64 */
626 + BCJ_POWERPC = 5, /* Big endian only */
627 + BCJ_IA64 = 6, /* Big or little endian */
628 + BCJ_ARM = 7, /* Little endian only */
629 + BCJ_ARMTHUMB = 8, /* Little endian only */
630 + BCJ_SPARC = 9 /* Big or little endian */
631 + } type;
632 +
633 + /*
634 + * Return value of the next filter in the chain. We need to preserve
635 + * this information across calls, because we must not call the next
636 + * filter anymore once it has returned XZ_STREAM_END.
637 + */
638 + enum xz_ret ret;
639 +
640 + /* True if we are operating in single-call mode. */
641 + bool single_call;
642 +
643 + /*
644 + * Absolute position relative to the beginning of the uncompressed
645 + * data (in a single .xz Block). We care only about the lowest 32
646 + * bits so this doesn't need to be uint64_t even with big files.
647 + */
648 + uint32_t pos;
649 +
650 + /* x86 filter state */
651 + uint32_t x86_prev_mask;
652 +
653 + /* Temporary space to hold the variables from struct xz_buf */
654 + uint8_t *out;
655 + size_t out_pos;
656 + size_t out_size;
657 +
658 + struct {
659 + /* Amount of already filtered data in the beginning of buf */
660 + size_t filtered;
661 +
662 + /* Total amount of data currently stored in buf */
663 + size_t size;
664 +
665 + /*
666 + * Buffer to hold a mix of filtered and unfiltered data. This
667 + * needs to be big enough to hold Alignment + 2 * Look-ahead:
668 + *
669 + * Type Alignment Look-ahead
670 + * x86 1 4
671 + * PowerPC 4 0
672 + * IA-64 16 0
673 + * ARM 4 0
674 + * ARM-Thumb 2 2
675 + * SPARC 4 0
676 + */
677 + uint8_t buf[16];
678 + } temp;
679 +};
680 +
681 +#ifdef XZ_DEC_X86
682 +/*
683 + * This is used to test the most significant byte of a memory address
684 + * in an x86 instruction.
685 + */
686 +static inline int bcj_x86_test_msbyte(uint8_t b)
687 +{
688 + return b == 0x00 || b == 0xFF;
689 +}
690 +
691 +static size_t bcj_x86(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
692 +{
693 + static const bool mask_to_allowed_status[8]
694 + = { true, true, true, false, true, false, false, false };
695 +
696 + static const uint8_t mask_to_bit_num[8] = { 0, 1, 2, 2, 3, 3, 3, 3 };
697 +
698 + size_t i;
699 + size_t prev_pos = (size_t)-1;
700 + uint32_t prev_mask = s->x86_prev_mask;
701 + uint32_t src;
702 + uint32_t dest;
703 + uint32_t j;
704 + uint8_t b;
705 +
706 + if (size <= 4)
707 + return 0;
708 +
709 + size -= 4;
710 + for (i = 0; i < size; ++i) {
711 + if ((buf[i] & 0xFE) != 0xE8)
712 + continue;
713 +
714 + prev_pos = i - prev_pos;
715 + if (prev_pos > 3) {
716 + prev_mask = 0;
717 + } else {
718 + prev_mask = (prev_mask << (prev_pos - 1)) & 7;
719 + if (prev_mask != 0) {
720 + b = buf[i + 4 - mask_to_bit_num[prev_mask]];
721 + if (!mask_to_allowed_status[prev_mask]
722 + || bcj_x86_test_msbyte(b)) {
723 + prev_pos = i;
724 + prev_mask = (prev_mask << 1) | 1;
725 + continue;
726 + }
727 + }
728 + }
729 +
730 + prev_pos = i;
731 +
732 + if (bcj_x86_test_msbyte(buf[i + 4])) {
733 + src = get_unaligned_le32(buf + i + 1);
734 + while (true) {
735 + dest = src - (s->pos + (uint32_t)i + 5);
736 + if (prev_mask == 0)
737 + break;
738 +
739 + j = mask_to_bit_num[prev_mask] * 8;
740 + b = (uint8_t)(dest >> (24 - j));
741 + if (!bcj_x86_test_msbyte(b))
742 + break;
743 +
744 + src = dest ^ (((uint32_t)1 << (32 - j)) - 1);
745 + }
746 +
747 + dest &= 0x01FFFFFF;
748 + dest |= (uint32_t)0 - (dest & 0x01000000);
749 + put_unaligned_le32(dest, buf + i + 1);
750 + i += 4;
751 + } else {
752 + prev_mask = (prev_mask << 1) | 1;
753 + }
754 + }
755 +
756 + prev_pos = i - prev_pos;
757 + s->x86_prev_mask = prev_pos > 3 ? 0 : prev_mask << (prev_pos - 1);
758 + return i;
759 +}
760 +#endif
761 +
762 +#ifdef XZ_DEC_POWERPC
763 +static size_t bcj_powerpc(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
764 +{
765 + size_t i;
766 + uint32_t instr;
767 +
768 + for (i = 0; i + 4 <= size; i += 4) {
769 + instr = get_unaligned_be32(buf + i);
770 + if ((instr & 0xFC000003) == 0x48000001) {
771 + instr &= 0x03FFFFFC;
772 + instr -= s->pos + (uint32_t)i;
773 + instr &= 0x03FFFFFC;
774 + instr |= 0x48000001;
775 + put_unaligned_be32(instr, buf + i);
776 + }
777 + }
778 +
779 + return i;
780 +}
781 +#endif
782 +
783 +#ifdef XZ_DEC_IA64
784 +static size_t bcj_ia64(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
785 +{
786 + static const uint8_t branch_table[32] = {
787 + 0, 0, 0, 0, 0, 0, 0, 0,
788 + 0, 0, 0, 0, 0, 0, 0, 0,
789 + 4, 4, 6, 6, 0, 0, 7, 7,
790 + 4, 4, 0, 0, 4, 4, 0, 0
791 + };
792 +
793 + /*
794 + * The local variables take a little bit stack space, but it's less
795 + * than what LZMA2 decoder takes, so it doesn't make sense to reduce
796 + * stack usage here without doing that for the LZMA2 decoder too.
797 + */
798 +
799 + /* Loop counters */
800 + size_t i;
801 + size_t j;
802 +
803 + /* Instruction slot (0, 1, or 2) in the 128-bit instruction word */
804 + uint32_t slot;
805 +
806 + /* Bitwise offset of the instruction indicated by slot */
807 + uint32_t bit_pos;
808 +
809 + /* bit_pos split into byte and bit parts */
810 + uint32_t byte_pos;
811 + uint32_t bit_res;
812 +
813 + /* Address part of an instruction */
814 + uint32_t addr;
815 +
816 + /* Mask used to detect which instructions to convert */
817 + uint32_t mask;
818 +
819 + /* 41-bit instruction stored somewhere in the lowest 48 bits */
820 + uint64_t instr;
821 +
822 + /* Instruction normalized with bit_res for easier manipulation */
823 + uint64_t norm;
824 +
825 + for (i = 0; i + 16 <= size; i += 16) {
826 + mask = branch_table[buf[i] & 0x1F];
827 + for (slot = 0, bit_pos = 5; slot < 3; ++slot, bit_pos += 41) {
828 + if (((mask >> slot) & 1) == 0)
829 + continue;
830 +
831 + byte_pos = bit_pos >> 3;
832 + bit_res = bit_pos & 7;
833 + instr = 0;
834 + for (j = 0; j < 6; ++j)
835 + instr |= (uint64_t)(buf[i + j + byte_pos])
836 + << (8 * j);
837 +
838 + norm = instr >> bit_res;
839 +
840 + if (((norm >> 37) & 0x0F) == 0x05
841 + && ((norm >> 9) & 0x07) == 0) {
842 + addr = (norm >> 13) & 0x0FFFFF;
843 + addr |= ((uint32_t)(norm >> 36) & 1) << 20;
844 + addr <<= 4;
845 + addr -= s->pos + (uint32_t)i;
846 + addr >>= 4;
847 +
848 + norm &= ~((uint64_t)0x8FFFFF << 13);
849 + norm |= (uint64_t)(addr & 0x0FFFFF) << 13;
850 + norm |= (uint64_t)(addr & 0x100000)
851 + << (36 - 20);
852 +
853 + instr &= (1 << bit_res) - 1;
854 + instr |= norm << bit_res;
855 +
856 + for (j = 0; j < 6; j++)
857 + buf[i + j + byte_pos]
858 + = (uint8_t)(instr >> (8 * j));
859 + }
860 + }
861 + }
862 +
863 + return i;
864 +}
865 +#endif
866 +
867 +#ifdef XZ_DEC_ARM
868 +static size_t bcj_arm(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
869 +{
870 + size_t i;
871 + uint32_t addr;
872 +
873 + for (i = 0; i + 4 <= size; i += 4) {
874 + if (buf[i + 3] == 0xEB) {
875 + addr = (uint32_t)buf[i] | ((uint32_t)buf[i + 1] << 8)
876 + | ((uint32_t)buf[i + 2] << 16);
877 + addr <<= 2;
878 + addr -= s->pos + (uint32_t)i + 8;
879 + addr >>= 2;
880 + buf[i] = (uint8_t)addr;
881 + buf[i + 1] = (uint8_t)(addr >> 8);
882 + buf[i + 2] = (uint8_t)(addr >> 16);
883 + }
884 + }
885 +
886 + return i;
887 +}
888 +#endif
889 +
890 +#ifdef XZ_DEC_ARMTHUMB
891 +static size_t bcj_armthumb(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
892 +{
893 + size_t i;
894 + uint32_t addr;
895 +
896 + for (i = 0; i + 4 <= size; i += 2) {
897 + if ((buf[i + 1] & 0xF8) == 0xF0
898 + && (buf[i + 3] & 0xF8) == 0xF8) {
899 + addr = (((uint32_t)buf[i + 1] & 0x07) << 19)
900 + | ((uint32_t)buf[i] << 11)
901 + | (((uint32_t)buf[i + 3] & 0x07) << 8)
902 + | (uint32_t)buf[i + 2];
903 + addr <<= 1;
904 + addr -= s->pos + (uint32_t)i + 4;
905 + addr >>= 1;
906 + buf[i + 1] = (uint8_t)(0xF0 | ((addr >> 19) & 0x07));
907 + buf[i] = (uint8_t)(addr >> 11);
908 + buf[i + 3] = (uint8_t)(0xF8 | ((addr >> 8) & 0x07));
909 + buf[i + 2] = (uint8_t)addr;
910 + i += 2;
911 + }
912 + }
913 +
914 + return i;
915 +}
916 +#endif
917 +
918 +#ifdef XZ_DEC_SPARC
919 +static size_t bcj_sparc(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
920 +{
921 + size_t i;
922 + uint32_t instr;
923 +
924 + for (i = 0; i + 4 <= size; i += 4) {
925 + instr = get_unaligned_be32(buf + i);
926 + if ((instr >> 22) == 0x100 || (instr >> 22) == 0x1FF) {
927 + instr <<= 2;
928 + instr -= s->pos + (uint32_t)i;
929 + instr >>= 2;
930 + instr = ((uint32_t)0x40000000 - (instr & 0x400000))
931 + | 0x40000000 | (instr & 0x3FFFFF);
932 + put_unaligned_be32(instr, buf + i);
933 + }
934 + }
935 +
936 + return i;
937 +}
938 +#endif
939 +
940 +/*
941 + * Apply the selected BCJ filter. Update *pos and s->pos to match the amount
942 + * of data that got filtered.
943 + *
944 + * NOTE: This is implemented as a switch statement to avoid using function
945 + * pointers, which could be problematic in the kernel boot code, which must
946 + * avoid pointers to static data (at least on x86).
947 + */
948 +static void bcj_apply(struct xz_dec_bcj *s,
949 + uint8_t *buf, size_t *pos, size_t size)
950 +{
951 + size_t filtered;
952 +
953 + buf += *pos;
954 + size -= *pos;
955 +
956 + switch (s->type) {
957 +#ifdef XZ_DEC_X86
958 + case BCJ_X86:
959 + filtered = bcj_x86(s, buf, size);
960 + break;
961 +#endif
962 +#ifdef XZ_DEC_POWERPC
963 + case BCJ_POWERPC:
964 + filtered = bcj_powerpc(s, buf, size);
965 + break;
966 +#endif
967 +#ifdef XZ_DEC_IA64
968 + case BCJ_IA64:
969 + filtered = bcj_ia64(s, buf, size);
970 + break;
971 +#endif
972 +#ifdef XZ_DEC_ARM
973 + case BCJ_ARM:
974 + filtered = bcj_arm(s, buf, size);
975 + break;
976 +#endif
977 +#ifdef XZ_DEC_ARMTHUMB
978 + case BCJ_ARMTHUMB:
979 + filtered = bcj_armthumb(s, buf, size);
980 + break;
981 +#endif
982 +#ifdef XZ_DEC_SPARC
983 + case BCJ_SPARC:
984 + filtered = bcj_sparc(s, buf, size);
985 + break;
986 +#endif
987 + default:
988 + /* Never reached but silence compiler warnings. */
989 + filtered = 0;
990 + break;
991 + }
992 +
993 + *pos += filtered;
994 + s->pos += filtered;
995 +}
996 +
997 +/*
998 + * Flush pending filtered data from temp to the output buffer.
999 + * Move the remaining mixture of possibly filtered and unfiltered
1000 + * data to the beginning of temp.
1001 + */
1002 +static void bcj_flush(struct xz_dec_bcj *s, struct xz_buf *b)
1003 +{
1004 + size_t copy_size;
1005 +
1006 + copy_size = min_t(size_t, s->temp.filtered, b->out_size - b->out_pos);
1007 + memcpy(b->out + b->out_pos, s->temp.buf, copy_size);
1008 + b->out_pos += copy_size;
1009 +
1010 + s->temp.filtered -= copy_size;
1011 + s->temp.size -= copy_size;
1012 + memmove(s->temp.buf, s->temp.buf + copy_size, s->temp.size);
1013 +}
1014 +
1015 +/*
1016 + * The BCJ filter functions are primitive in sense that they process the
1017 + * data in chunks of 1-16 bytes. To hide this issue, this function does
1018 + * some buffering.
1019 + */
1020 +XZ_EXTERN enum xz_ret xz_dec_bcj_run(struct xz_dec_bcj *s,
1021 + struct xz_dec_lzma2 *lzma2,
1022 + struct xz_buf *b)
1023 +{
1024 + size_t out_start;
1025 +
1026 + /*
1027 + * Flush pending already filtered data to the output buffer. Return
1028 + * immediatelly if we couldn't flush everything, or if the next
1029 + * filter in the chain had already returned XZ_STREAM_END.
1030 + */
1031 + if (s->temp.filtered > 0) {
1032 + bcj_flush(s, b);
1033 + if (s->temp.filtered > 0)
1034 + return XZ_OK;
1035 +
1036 + if (s->ret == XZ_STREAM_END)
1037 + return XZ_STREAM_END;
1038 + }
1039 +
1040 + /*
1041 + * If we have more output space than what is currently pending in
1042 + * temp, copy the unfiltered data from temp to the output buffer
1043 + * and try to fill the output buffer by decoding more data from the
1044 + * next filter in the chain. Apply the BCJ filter on the new data
1045 + * in the output buffer. If everything cannot be filtered, copy it
1046 + * to temp and rewind the output buffer position accordingly.
1047 + */
1048 + if (s->temp.size < b->out_size - b->out_pos) {
1049 + out_start = b->out_pos;
1050 + memcpy(b->out + b->out_pos, s->temp.buf, s->temp.size);
1051 + b->out_pos += s->temp.size;
1052 +
1053 + s->ret = xz_dec_lzma2_run(lzma2, b);
1054 + if (s->ret != XZ_STREAM_END
1055 + && (s->ret != XZ_OK || s->single_call))
1056 + return s->ret;
1057 +
1058 + bcj_apply(s, b->out, &out_start, b->out_pos);
1059 +
1060 + /*
1061 + * As an exception, if the next filter returned XZ_STREAM_END,
1062 + * we can do that too, since the last few bytes that remain
1063 + * unfiltered are meant to remain unfiltered.
1064 + */
1065 + if (s->ret == XZ_STREAM_END)
1066 + return XZ_STREAM_END;
1067 +
1068 + s->temp.size = b->out_pos - out_start;
1069 + b->out_pos -= s->temp.size;
1070 + memcpy(s->temp.buf, b->out + b->out_pos, s->temp.size);
1071 + }
1072 +
1073 + /*
1074 + * If we have unfiltered data in temp, try to fill by decoding more
1075 + * data from the next filter. Apply the BCJ filter on temp. Then we
1076 + * hopefully can fill the actual output buffer by copying filtered
1077 + * data from temp. A mix of filtered and unfiltered data may be left
1078 + * in temp; it will be taken care on the next call to this function.
1079 + */
1080 + if (s->temp.size > 0) {
1081 + /* Make b->out{,_pos,_size} temporarily point to s->temp. */
1082 + s->out = b->out;
1083 + s->out_pos = b->out_pos;
1084 + s->out_size = b->out_size;
1085 + b->out = s->temp.buf;
1086 + b->out_pos = s->temp.size;
1087 + b->out_size = sizeof(s->temp.buf);
1088 +
1089 + s->ret = xz_dec_lzma2_run(lzma2, b);
1090 +
1091 + s->temp.size = b->out_pos;
1092 + b->out = s->out;
1093 + b->out_pos = s->out_pos;
1094 + b->out_size = s->out_size;
1095 +
1096 + if (s->ret != XZ_OK && s->ret != XZ_STREAM_END)
1097 + return s->ret;
1098 +
1099 + bcj_apply(s, s->temp.buf, &s->temp.filtered, s->temp.size);
1100 +
1101 + /*
1102 + * If the next filter returned XZ_STREAM_END, we mark that
1103 + * everything is filtered, since the last unfiltered bytes
1104 + * of the stream are meant to be left as is.
1105 + */
1106 + if (s->ret == XZ_STREAM_END)
1107 + s->temp.filtered = s->temp.size;
1108 +
1109 + bcj_flush(s, b);
1110 + if (s->temp.filtered > 0)
1111 + return XZ_OK;
1112 + }
1113 +
1114 + return s->ret;
1115 +}
1116 +
1117 +XZ_EXTERN struct xz_dec_bcj *xz_dec_bcj_create(bool single_call)
1118 +{
1119 + struct xz_dec_bcj *s = kmalloc(sizeof(*s), GFP_KERNEL);
1120 + if (s != NULL)
1121 + s->single_call = single_call;
1122 +
1123 + return s;
1124 +}
1125 +
1126 +XZ_EXTERN enum xz_ret xz_dec_bcj_reset(struct xz_dec_bcj *s, uint8_t id)
1127 +{
1128 + switch (id) {
1129 +#ifdef XZ_DEC_X86
1130 + case BCJ_X86:
1131 +#endif
1132 +#ifdef XZ_DEC_POWERPC
1133 + case BCJ_POWERPC:
1134 +#endif
1135 +#ifdef XZ_DEC_IA64
1136 + case BCJ_IA64:
1137 +#endif
1138 +#ifdef XZ_DEC_ARM
1139 + case BCJ_ARM:
1140 +#endif
1141 +#ifdef XZ_DEC_ARMTHUMB
1142 + case BCJ_ARMTHUMB:
1143 +#endif
1144 +#ifdef XZ_DEC_SPARC
1145 + case BCJ_SPARC:
1146 +#endif
1147 + break;
1148 +
1149 + default:
1150 + /* Unsupported Filter ID */
1151 + return XZ_OPTIONS_ERROR;
1152 + }
1153 +
1154 + s->type = id;
1155 + s->ret = XZ_OK;
1156 + s->pos = 0;
1157 + s->x86_prev_mask = 0;
1158 + s->temp.filtered = 0;
1159 + s->temp.size = 0;
1160 +
1161 + return XZ_OK;
1162 +}
1163 +
1164 +#endif
1165 diff --git a/lib/xz/xz_dec_lzma2.c b/lib/xz/xz_dec_lzma2.c
1166 new file mode 100644
1167 index 0000000..ea5fa4f
1168 --- /dev/null
1169 +++ b/lib/xz/xz_dec_lzma2.c
1170 @@ -0,0 +1,1171 @@
1171 +/*
1172 + * LZMA2 decoder
1173 + *
1174 + * Authors: Lasse Collin <lasse.collin@tukaani.org>
1175 + * Igor Pavlov <http://7-zip.org/>
1176 + *
1177 + * This file has been put into the public domain.
1178 + * You can do whatever you want with this file.
1179 + */
1180 +
1181 +#include "xz_private.h"
1182 +#include "xz_lzma2.h"
1183 +
1184 +/*
1185 + * Range decoder initialization eats the first five bytes of each LZMA chunk.
1186 + */
1187 +#define RC_INIT_BYTES 5
1188 +
1189 +/*
1190 + * Minimum number of usable input buffer to safely decode one LZMA symbol.
1191 + * The worst case is that we decode 22 bits using probabilities and 26
1192 + * direct bits. This may decode at maximum of 20 bytes of input. However,
1193 + * lzma_main() does an extra normalization before returning, thus we
1194 + * need to put 21 here.
1195 + */
1196 +#define LZMA_IN_REQUIRED 21
1197 +
1198 +/*
1199 + * Dictionary (history buffer)
1200 + *
1201 + * These are always true:
1202 + * start <= pos <= full <= end
1203 + * pos <= limit <= end
1204 + *
1205 + * In multi-call mode, also these are true:
1206 + * end == size
1207 + * size <= size_max
1208 + * allocated <= size
1209 + *
1210 + * Most of these variables are size_t to support single-call mode,
1211 + * in which the dictionary variables address the actual output
1212 + * buffer directly.
1213 + */
1214 +struct dictionary {
1215 + /* Beginning of the history buffer */
1216 + uint8_t *buf;
1217 +
1218 + /* Old position in buf (before decoding more data) */
1219 + size_t start;
1220 +
1221 + /* Position in buf */
1222 + size_t pos;
1223 +
1224 + /*
1225 + * How full dictionary is. This is used to detect corrupt input that
1226 + * would read beyond the beginning of the uncompressed stream.
1227 + */
1228 + size_t full;
1229 +
1230 + /* Write limit; we don't write to buf[limit] or later bytes. */
1231 + size_t limit;
1232 +
1233 + /*
1234 + * End of the dictionary buffer. In multi-call mode, this is
1235 + * the same as the dictionary size. In single-call mode, this
1236 + * indicates the size of the output buffer.
1237 + */
1238 + size_t end;
1239 +
1240 + /*
1241 + * Size of the dictionary as specified in Block Header. This is used
1242 + * together with "full" to detect corrupt input that would make us
1243 + * read beyond the beginning of the uncompressed stream.
1244 + */
1245 + uint32_t size;
1246 +
1247 + /*
1248 + * Maximum allowed dictionary size in multi-call mode.
1249 + * This is ignored in single-call mode.
1250 + */
1251 + uint32_t size_max;
1252 +
1253 + /*
1254 + * Amount of memory currently allocated for the dictionary.
1255 + * This is used only with XZ_DYNALLOC. (With XZ_PREALLOC,
1256 + * size_max is always the same as the allocated size.)
1257 + */
1258 + uint32_t allocated;
1259 +
1260 + /* Operation mode */
1261 + enum xz_mode mode;
1262 +};
1263 +
1264 +/* Range decoder */
1265 +struct rc_dec {
1266 + uint32_t range;
1267 + uint32_t code;
1268 +
1269 + /*
1270 + * Number of initializing bytes remaining to be read
1271 + * by rc_read_init().
1272 + */
1273 + uint32_t init_bytes_left;
1274 +
1275 + /*
1276 + * Buffer from which we read our input. It can be either
1277 + * temp.buf or the caller-provided input buffer.
1278 + */
1279 + const uint8_t *in;
1280 + size_t in_pos;
1281 + size_t in_limit;
1282 +};
1283 +
1284 +/* Probabilities for a length decoder. */
1285 +struct lzma_len_dec {
1286 + /* Probability of match length being at least 10 */
1287 + uint16_t choice;
1288 +
1289 + /* Probability of match length being at least 18 */
1290 + uint16_t choice2;
1291 +
1292 + /* Probabilities for match lengths 2-9 */
1293 + uint16_t low[POS_STATES_MAX][LEN_LOW_SYMBOLS];
1294 +
1295 + /* Probabilities for match lengths 10-17 */
1296 + uint16_t mid[POS_STATES_MAX][LEN_MID_SYMBOLS];
1297 +
1298 + /* Probabilities for match lengths 18-273 */
1299 + uint16_t high[LEN_HIGH_SYMBOLS];
1300 +};
1301 +
1302 +struct lzma_dec {
1303 + /* Distances of latest four matches */
1304 + uint32_t rep0;
1305 + uint32_t rep1;
1306 + uint32_t rep2;
1307 + uint32_t rep3;
1308 +
1309 + /* Types of the most recently seen LZMA symbols */
1310 + enum lzma_state state;
1311 +
1312 + /*
1313 + * Length of a match. This is updated so that dict_repeat can
1314 + * be called again to finish repeating the whole match.
1315 + */
1316 + uint32_t len;
1317 +
1318 + /*
1319 + * LZMA properties or related bit masks (number of literal
1320 + * context bits, a mask dervied from the number of literal
1321 + * position bits, and a mask dervied from the number
1322 + * position bits)
1323 + */
1324 + uint32_t lc;
1325 + uint32_t literal_pos_mask; /* (1 << lp) - 1 */
1326 + uint32_t pos_mask; /* (1 << pb) - 1 */
1327 +
1328 + /* If 1, it's a match. Otherwise it's a single 8-bit literal. */
1329 + uint16_t is_match[STATES][POS_STATES_MAX];
1330 +
1331 + /* If 1, it's a repeated match. The distance is one of rep0 .. rep3. */
1332 + uint16_t is_rep[STATES];
1333 +
1334 + /*
1335 + * If 0, distance of a repeated match is rep0.
1336 + * Otherwise check is_rep1.
1337 + */
1338 + uint16_t is_rep0[STATES];
1339 +
1340 + /*
1341 + * If 0, distance of a repeated match is rep1.
1342 + * Otherwise check is_rep2.
1343 + */
1344 + uint16_t is_rep1[STATES];
1345 +
1346 + /* If 0, distance of a repeated match is rep2. Otherwise it is rep3. */
1347 + uint16_t is_rep2[STATES];
1348 +
1349 + /*
1350 + * If 1, the repeated match has length of one byte. Otherwise
1351 + * the length is decoded from rep_len_decoder.
1352 + */
1353 + uint16_t is_rep0_long[STATES][POS_STATES_MAX];
1354 +
1355 + /*
1356 + * Probability tree for the highest two bits of the match
1357 + * distance. There is a separate probability tree for match
1358 + * lengths of 2 (i.e. MATCH_LEN_MIN), 3, 4, and [5, 273].
1359 + */
1360 + uint16_t dist_slot[DIST_STATES][DIST_SLOTS];
1361 +
1362 + /*
1363 + * Probility trees for additional bits for match distance
1364 + * when the distance is in the range [4, 127].
1365 + */
1366 + uint16_t dist_special[FULL_DISTANCES - DIST_MODEL_END];
1367 +
1368 + /*
1369 + * Probability tree for the lowest four bits of a match
1370 + * distance that is equal to or greater than 128.
1371 + */
1372 + uint16_t dist_align[ALIGN_SIZE];
1373 +
1374 + /* Length of a normal match */
1375 + struct lzma_len_dec match_len_dec;
1376 +
1377 + /* Length of a repeated match */
1378 + struct lzma_len_dec rep_len_dec;
1379 +
1380 + /* Probabilities of literals */
1381 + uint16_t literal[LITERAL_CODERS_MAX][LITERAL_CODER_SIZE];
1382 +};
1383 +
1384 +struct lzma2_dec {
1385 + /* Position in xz_dec_lzma2_run(). */
1386 + enum lzma2_seq {
1387 + SEQ_CONTROL,
1388 + SEQ_UNCOMPRESSED_1,
1389 + SEQ_UNCOMPRESSED_2,
1390 + SEQ_COMPRESSED_0,
1391 + SEQ_COMPRESSED_1,
1392 + SEQ_PROPERTIES,
1393 + SEQ_LZMA_PREPARE,
1394 + SEQ_LZMA_RUN,
1395 + SEQ_COPY
1396 + } sequence;
1397 +
1398 + /* Next position after decoding the compressed size of the chunk. */
1399 + enum lzma2_seq next_sequence;
1400 +
1401 + /* Uncompressed size of LZMA chunk (2 MiB at maximum) */
1402 + uint32_t uncompressed;
1403 +
1404 + /*
1405 + * Compressed size of LZMA chunk or compressed/uncompressed
1406 + * size of uncompressed chunk (64 KiB at maximum)
1407 + */
1408 + uint32_t compressed;
1409 +
1410 + /*
1411 + * True if dictionary reset is needed. This is false before
1412 + * the first chunk (LZMA or uncompressed).
1413 + */
1414 + bool need_dict_reset;
1415 +
1416 + /*
1417 + * True if new LZMA properties are needed. This is false
1418 + * before the first LZMA chunk.
1419 + */
1420 + bool need_props;
1421 +};
1422 +
1423 +struct xz_dec_lzma2 {
1424 + /*
1425 + * The order below is important on x86 to reduce code size and
1426 + * it shouldn't hurt on other platforms. Everything up to and
1427 + * including lzma.pos_mask are in the first 128 bytes on x86-32,
1428 + * which allows using smaller instructions to access those
1429 + * variables. On x86-64, fewer variables fit into the first 128
1430 + * bytes, but this is still the best order without sacrificing
1431 + * the readability by splitting the structures.
1432 + */
1433 + struct rc_dec rc;
1434 + struct dictionary dict;
1435 + struct lzma2_dec lzma2;
1436 + struct lzma_dec lzma;
1437 +
1438 + /*
1439 + * Temporary buffer which holds small number of input bytes between
1440 + * decoder calls. See lzma2_lzma() for details.
1441 + */
1442 + struct {
1443 + uint32_t size;
1444 + uint8_t buf[3 * LZMA_IN_REQUIRED];
1445 + } temp;
1446 +};
1447 +
1448 +/**************
1449 + * Dictionary *
1450 + **************/
1451 +
1452 +/*
1453 + * Reset the dictionary state. When in single-call mode, set up the beginning
1454 + * of the dictionary to point to the actual output buffer.
1455 + */
1456 +static void dict_reset(struct dictionary *dict, struct xz_buf *b)
1457 +{
1458 + if (DEC_IS_SINGLE(dict->mode)) {
1459 + dict->buf = b->out + b->out_pos;
1460 + dict->end = b->out_size - b->out_pos;
1461 + }
1462 +
1463 + dict->start = 0;
1464 + dict->pos = 0;
1465 + dict->limit = 0;
1466 + dict->full = 0;
1467 +}
1468 +
1469 +/* Set dictionary write limit */
1470 +static void dict_limit(struct dictionary *dict, size_t out_max)
1471 +{
1472 + if (dict->end - dict->pos <= out_max)
1473 + dict->limit = dict->end;
1474 + else
1475 + dict->limit = dict->pos + out_max;
1476 +}
1477 +
1478 +/* Return true if at least one byte can be written into the dictionary. */
1479 +static inline bool dict_has_space(const struct dictionary *dict)
1480 +{
1481 + return dict->pos < dict->limit;
1482 +}
1483 +
1484 +/*
1485 + * Get a byte from the dictionary at the given distance. The distance is
1486 + * assumed to valid, or as a special case, zero when the dictionary is
1487 + * still empty. This special case is needed for single-call decoding to
1488 + * avoid writing a '\0' to the end of the destination buffer.
1489 + */
1490 +static inline uint32_t dict_get(const struct dictionary *dict, uint32_t dist)
1491 +{
1492 + size_t offset = dict->pos - dist - 1;
1493 +
1494 + if (dist >= dict->pos)
1495 + offset += dict->end;
1496 +
1497 + return dict->full > 0 ? dict->buf[offset] : 0;
1498 +}
1499 +
1500 +/*
1501 + * Put one byte into the dictionary. It is assumed that there is space for it.
1502 + */
1503 +static inline void dict_put(struct dictionary *dict, uint8_t byte)
1504 +{
1505 + dict->buf[dict->pos++] = byte;
1506 +
1507 + if (dict->full < dict->pos)
1508 + dict->full = dict->pos;
1509 +}
1510 +
1511 +/*
1512 + * Repeat given number of bytes from the given distance. If the distance is
1513 + * invalid, false is returned. On success, true is returned and *len is
1514 + * updated to indicate how many bytes were left to be repeated.
1515 + */
1516 +static bool dict_repeat(struct dictionary *dict, uint32_t *len, uint32_t dist)
1517 +{
1518 + size_t back;
1519 + uint32_t left;
1520 +
1521 + if (dist >= dict->full || dist >= dict->size)
1522 + return false;
1523 +
1524 + left = min_t(size_t, dict->limit - dict->pos, *len);
1525 + *len -= left;
1526 +
1527 + back = dict->pos - dist - 1;
1528 + if (dist >= dict->pos)
1529 + back += dict->end;
1530 +
1531 + do {
1532 + dict->buf[dict->pos++] = dict->buf[back++];
1533 + if (back == dict->end)
1534 + back = 0;
1535 + } while (--left > 0);
1536 +
1537 + if (dict->full < dict->pos)
1538 + dict->full = dict->pos;
1539 +
1540 + return true;
1541 +}
1542 +
1543 +/* Copy uncompressed data as is from input to dictionary and output buffers. */
1544 +static void dict_uncompressed(struct dictionary *dict, struct xz_buf *b,
1545 + uint32_t *left)
1546 +{
1547 + size_t copy_size;
1548 +
1549 + while (*left > 0 && b->in_pos < b->in_size
1550 + && b->out_pos < b->out_size) {
1551 + copy_size = min(b->in_size - b->in_pos,
1552 + b->out_size - b->out_pos);
1553 + if (copy_size > dict->end - dict->pos)
1554 + copy_size = dict->end - dict->pos;
1555 + if (copy_size > *left)
1556 + copy_size = *left;
1557 +
1558 + *left -= copy_size;
1559 +
1560 + memcpy(dict->buf + dict->pos, b->in + b->in_pos, copy_size);
1561 + dict->pos += copy_size;
1562 +
1563 + if (dict->full < dict->pos)
1564 + dict->full = dict->pos;
1565 +
1566 + if (DEC_IS_MULTI(dict->mode)) {
1567 + if (dict->pos == dict->end)
1568 + dict->pos = 0;
1569 +
1570 + memcpy(b->out + b->out_pos, b->in + b->in_pos,
1571 + copy_size);
1572 + }
1573 +
1574 + dict->start = dict->pos;
1575 +
1576 + b->out_pos += copy_size;
1577 + b->in_pos += copy_size;
1578 + }
1579 +}
1580 +
1581 +/*
1582 + * Flush pending data from dictionary to b->out. It is assumed that there is
1583 + * enough space in b->out. This is guaranteed because caller uses dict_limit()
1584 + * before decoding data into the dictionary.
1585 + */
1586 +static uint32_t dict_flush(struct dictionary *dict, struct xz_buf *b)
1587 +{
1588 + size_t copy_size = dict->pos - dict->start;
1589 +
1590 + if (DEC_IS_MULTI(dict->mode)) {
1591 + if (dict->pos == dict->end)
1592 + dict->pos = 0;
1593 +
1594 + memcpy(b->out + b->out_pos, dict->buf + dict->start,
1595 + copy_size);
1596 + }
1597 +
1598 + dict->start = dict->pos;
1599 + b->out_pos += copy_size;
1600 + return copy_size;
1601 +}
1602 +
1603 +/*****************
1604 + * Range decoder *
1605 + *****************/
1606 +
1607 +/* Reset the range decoder. */
1608 +static void rc_reset(struct rc_dec *rc)
1609 +{
1610 + rc->range = (uint32_t)-1;
1611 + rc->code = 0;
1612 + rc->init_bytes_left = RC_INIT_BYTES;
1613 +}
1614 +
1615 +/*
1616 + * Read the first five initial bytes into rc->code if they haven't been
1617 + * read already. (Yes, the first byte gets completely ignored.)
1618 + */
1619 +static bool rc_read_init(struct rc_dec *rc, struct xz_buf *b)
1620 +{
1621 + while (rc->init_bytes_left > 0) {
1622 + if (b->in_pos == b->in_size)
1623 + return false;
1624 +
1625 + rc->code = (rc->code << 8) + b->in[b->in_pos++];
1626 + --rc->init_bytes_left;
1627 + }
1628 +
1629 + return true;
1630 +}
1631 +
1632 +/* Return true if there may not be enough input for the next decoding loop. */
1633 +static inline bool rc_limit_exceeded(const struct rc_dec *rc)
1634 +{
1635 + return rc->in_pos > rc->in_limit;
1636 +}
1637 +
1638 +/*
1639 + * Return true if it is possible (from point of view of range decoder) that
1640 + * we have reached the end of the LZMA chunk.
1641 + */
1642 +static inline bool rc_is_finished(const struct rc_dec *rc)
1643 +{
1644 + return rc->code == 0;
1645 +}
1646 +
1647 +/* Read the next input byte if needed. */
1648 +static __always_inline void rc_normalize(struct rc_dec *rc)
1649 +{
1650 + if (rc->range < RC_TOP_VALUE) {
1651 + rc->range <<= RC_SHIFT_BITS;
1652 + rc->code = (rc->code << RC_SHIFT_BITS) + rc->in[rc->in_pos++];
1653 + }
1654 +}
1655 +
1656 +/*
1657 + * Decode one bit. In some versions, this function has been splitted in three
1658 + * functions so that the compiler is supposed to be able to more easily avoid
1659 + * an extra branch. In this particular version of the LZMA decoder, this
1660 + * doesn't seem to be a good idea (tested with GCC 3.3.6, 3.4.6, and 4.3.3
1661 + * on x86). Using a non-splitted version results in nicer looking code too.
1662 + *
1663 + * NOTE: This must return an int. Do not make it return a bool or the speed
1664 + * of the code generated by GCC 3.x decreases 10-15 %. (GCC 4.3 doesn't care,
1665 + * and it generates 10-20 % faster code than GCC 3.x from this file anyway.)
1666 + */
1667 +static __always_inline int rc_bit(struct rc_dec *rc, uint16_t *prob)
1668 +{
1669 + uint32_t bound;
1670 + int bit;
1671 +
1672 + rc_normalize(rc);
1673 + bound = (rc->range >> RC_BIT_MODEL_TOTAL_BITS) * *prob;
1674 + if (rc->code < bound) {
1675 + rc->range = bound;
1676 + *prob += (RC_BIT_MODEL_TOTAL - *prob) >> RC_MOVE_BITS;
1677 + bit = 0;
1678 + } else {
1679 + rc->range -= bound;
1680 + rc->code -= bound;
1681 + *prob -= *prob >> RC_MOVE_BITS;
1682 + bit = 1;
1683 + }
1684 +
1685 + return bit;
1686 +}
1687 +
1688 +/* Decode a bittree starting from the most significant bit. */
1689 +static __always_inline uint32_t rc_bittree(struct rc_dec *rc,
1690 + uint16_t *probs, uint32_t limit)
1691 +{
1692 + uint32_t symbol = 1;
1693 +
1694 + do {
1695 + if (rc_bit(rc, &probs[symbol]))
1696 + symbol = (symbol << 1) + 1;
1697 + else
1698 + symbol <<= 1;
1699 + } while (symbol < limit);
1700 +
1701 + return symbol;
1702 +}
1703 +
1704 +/* Decode a bittree starting from the least significant bit. */
1705 +static __always_inline void rc_bittree_reverse(struct rc_dec *rc,
1706 + uint16_t *probs,
1707 + uint32_t *dest, uint32_t limit)
1708 +{
1709 + uint32_t symbol = 1;
1710 + uint32_t i = 0;
1711 +
1712 + do {
1713 + if (rc_bit(rc, &probs[symbol])) {
1714 + symbol = (symbol << 1) + 1;
1715 + *dest += 1 << i;
1716 + } else {
1717 + symbol <<= 1;
1718 + }
1719 + } while (++i < limit);
1720 +}
1721 +
1722 +/* Decode direct bits (fixed fifty-fifty probability) */
1723 +static inline void rc_direct(struct rc_dec *rc, uint32_t *dest, uint32_t limit)
1724 +{
1725 + uint32_t mask;
1726 +
1727 + do {
1728 + rc_normalize(rc);
1729 + rc->range >>= 1;
1730 + rc->code -= rc->range;
1731 + mask = (uint32_t)0 - (rc->code >> 31);
1732 + rc->code += rc->range & mask;
1733 + *dest = (*dest << 1) + (mask + 1);
1734 + } while (--limit > 0);
1735 +}
1736 +
1737 +/********
1738 + * LZMA *
1739 + ********/
1740 +
1741 +/* Get pointer to literal coder probability array. */
1742 +static uint16_t *lzma_literal_probs(struct xz_dec_lzma2 *s)
1743 +{
1744 + uint32_t prev_byte = dict_get(&s->dict, 0);
1745 + uint32_t low = prev_byte >> (8 - s->lzma.lc);
1746 + uint32_t high = (s->dict.pos & s->lzma.literal_pos_mask) << s->lzma.lc;
1747 + return s->lzma.literal[low + high];
1748 +}
1749 +
1750 +/* Decode a literal (one 8-bit byte) */
1751 +static void lzma_literal(struct xz_dec_lzma2 *s)
1752 +{
1753 + uint16_t *probs;
1754 + uint32_t symbol;
1755 + uint32_t match_byte;
1756 + uint32_t match_bit;
1757 + uint32_t offset;
1758 + uint32_t i;
1759 +
1760 + probs = lzma_literal_probs(s);
1761 +
1762 + if (lzma_state_is_literal(s->lzma.state)) {
1763 + symbol = rc_bittree(&s->rc, probs, 0x100);
1764 + } else {
1765 + symbol = 1;
1766 + match_byte = dict_get(&s->dict, s->lzma.rep0) << 1;
1767 + offset = 0x100;
1768 +
1769 + do {
1770 + match_bit = match_byte & offset;
1771 + match_byte <<= 1;
1772 + i = offset + match_bit + symbol;
1773 +
1774 + if (rc_bit(&s->rc, &probs[i])) {
1775 + symbol = (symbol << 1) + 1;
1776 + offset &= match_bit;
1777 + } else {
1778 + symbol <<= 1;
1779 + offset &= ~match_bit;
1780 + }
1781 + } while (symbol < 0x100);
1782 + }
1783 +
1784 + dict_put(&s->dict, (uint8_t)symbol);
1785 + lzma_state_literal(&s->lzma.state);
1786 +}
1787 +
1788 +/* Decode the length of the match into s->lzma.len. */
1789 +static void lzma_len(struct xz_dec_lzma2 *s, struct lzma_len_dec *l,
1790 + uint32_t pos_state)
1791 +{
1792 + uint16_t *probs;
1793 + uint32_t limit;
1794 +
1795 + if (!rc_bit(&s->rc, &l->choice)) {
1796 + probs = l->low[pos_state];
1797 + limit = LEN_LOW_SYMBOLS;
1798 + s->lzma.len = MATCH_LEN_MIN;
1799 + } else {
1800 + if (!rc_bit(&s->rc, &l->choice2)) {
1801 + probs = l->mid[pos_state];
1802 + limit = LEN_MID_SYMBOLS;
1803 + s->lzma.len = MATCH_LEN_MIN + LEN_LOW_SYMBOLS;
1804 + } else {
1805 + probs = l->high;
1806 + limit = LEN_HIGH_SYMBOLS;
1807 + s->lzma.len = MATCH_LEN_MIN + LEN_LOW_SYMBOLS
1808 + + LEN_MID_SYMBOLS;
1809 + }
1810 + }
1811 +
1812 + s->lzma.len += rc_bittree(&s->rc, probs, limit) - limit;
1813 +}
1814 +
1815 +/* Decode a match. The distance will be stored in s->lzma.rep0. */
1816 +static void lzma_match(struct xz_dec_lzma2 *s, uint32_t pos_state)
1817 +{
1818 + uint16_t *probs;
1819 + uint32_t dist_slot;
1820 + uint32_t limit;
1821 +
1822 + lzma_state_match(&s->lzma.state);
1823 +
1824 + s->lzma.rep3 = s->lzma.rep2;
1825 + s->lzma.rep2 = s->lzma.rep1;
1826 + s->lzma.rep1 = s->lzma.rep0;
1827 +
1828 + lzma_len(s, &s->lzma.match_len_dec, pos_state);
1829 +
1830 + probs = s->lzma.dist_slot[lzma_get_dist_state(s->lzma.len)];
1831 + dist_slot = rc_bittree(&s->rc, probs, DIST_SLOTS) - DIST_SLOTS;
1832 +
1833 + if (dist_slot < DIST_MODEL_START) {
1834 + s->lzma.rep0 = dist_slot;
1835 + } else {
1836 + limit = (dist_slot >> 1) - 1;
1837 + s->lzma.rep0 = 2 + (dist_slot & 1);
1838 +
1839 + if (dist_slot < DIST_MODEL_END) {
1840 + s->lzma.rep0 <<= limit;
1841 + probs = s->lzma.dist_special + s->lzma.rep0
1842 + - dist_slot - 1;
1843 + rc_bittree_reverse(&s->rc, probs,
1844 + &s->lzma.rep0, limit);
1845 + } else {
1846 + rc_direct(&s->rc, &s->lzma.rep0, limit - ALIGN_BITS);
1847 + s->lzma.rep0 <<= ALIGN_BITS;
1848 + rc_bittree_reverse(&s->rc, s->lzma.dist_align,
1849 + &s->lzma.rep0, ALIGN_BITS);
1850 + }
1851 + }
1852 +}
1853 +
1854 +/*
1855 + * Decode a repeated match. The distance is one of the four most recently
1856 + * seen matches. The distance will be stored in s->lzma.rep0.
1857 + */
1858 +static void lzma_rep_match(struct xz_dec_lzma2 *s, uint32_t pos_state)
1859 +{
1860 + uint32_t tmp;
1861 +
1862 + if (!rc_bit(&s->rc, &s->lzma.is_rep0[s->lzma.state])) {
1863 + if (!rc_bit(&s->rc, &s->lzma.is_rep0_long[
1864 + s->lzma.state][pos_state])) {
1865 + lzma_state_short_rep(&s->lzma.state);
1866 + s->lzma.len = 1;
1867 + return;
1868 + }
1869 + } else {
1870 + if (!rc_bit(&s->rc, &s->lzma.is_rep1[s->lzma.state])) {
1871 + tmp = s->lzma.rep1;
1872 + } else {
1873 + if (!rc_bit(&s->rc, &s->lzma.is_rep2[s->lzma.state])) {
1874 + tmp = s->lzma.rep2;
1875 + } else {
1876 + tmp = s->lzma.rep3;
1877 + s->lzma.rep3 = s->lzma.rep2;
1878 + }
1879 +
1880 + s->lzma.rep2 = s->lzma.rep1;
1881 + }
1882 +
1883 + s->lzma.rep1 = s->lzma.rep0;
1884 + s->lzma.rep0 = tmp;
1885 + }
1886 +
1887 + lzma_state_long_rep(&s->lzma.state);
1888 + lzma_len(s, &s->lzma.rep_len_dec, pos_state);
1889 +}
1890 +
1891 +/* LZMA decoder core */
1892 +static bool lzma_main(struct xz_dec_lzma2 *s)
1893 +{
1894 + uint32_t pos_state;
1895 +
1896 + /*
1897 + * If the dictionary was reached during the previous call, try to
1898 + * finish the possibly pending repeat in the dictionary.
1899 + */
1900 + if (dict_has_space(&s->dict) && s->lzma.len > 0)
1901 + dict_repeat(&s->dict, &s->lzma.len, s->lzma.rep0);
1902 +
1903 + /*
1904 + * Decode more LZMA symbols. One iteration may consume up to
1905 + * LZMA_IN_REQUIRED - 1 bytes.
1906 + */
1907 + while (dict_has_space(&s->dict) && !rc_limit_exceeded(&s->rc)) {
1908 + pos_state = s->dict.pos & s->lzma.pos_mask;
1909 +
1910 + if (!rc_bit(&s->rc, &s->lzma.is_match[
1911 + s->lzma.state][pos_state])) {
1912 + lzma_literal(s);
1913 + } else {
1914 + if (rc_bit(&s->rc, &s->lzma.is_rep[s->lzma.state]))
1915 + lzma_rep_match(s, pos_state);
1916 + else
1917 + lzma_match(s, pos_state);
1918 +
1919 + if (!dict_repeat(&s->dict, &s->lzma.len, s->lzma.rep0))
1920 + return false;
1921 + }
1922 + }
1923 +
1924 + /*
1925 + * Having the range decoder always normalized when we are outside
1926 + * this function makes it easier to correctly handle end of the chunk.
1927 + */
1928 + rc_normalize(&s->rc);
1929 +
1930 + return true;
1931 +}
1932 +
1933 +/*
1934 + * Reset the LZMA decoder and range decoder state. Dictionary is nore reset
1935 + * here, because LZMA state may be reset without resetting the dictionary.
1936 + */
1937 +static void lzma_reset(struct xz_dec_lzma2 *s)
1938 +{
1939 + uint16_t *probs;
1940 + size_t i;
1941 +
1942 + s->lzma.state = STATE_LIT_LIT;
1943 + s->lzma.rep0 = 0;
1944 + s->lzma.rep1 = 0;
1945 + s->lzma.rep2 = 0;
1946 + s->lzma.rep3 = 0;
1947 +
1948 + /*
1949 + * All probabilities are initialized to the same value. This hack
1950 + * makes the code smaller by avoiding a separate loop for each
1951 + * probability array.
1952 + *
1953 + * This could be optimized so that only that part of literal
1954 + * probabilities that are actually required. In the common case
1955 + * we would write 12 KiB less.
1956 + */
1957 + probs = s->lzma.is_match[0];
1958 + for (i = 0; i < PROBS_TOTAL; ++i)
1959 + probs[i] = RC_BIT_MODEL_TOTAL / 2;
1960 +
1961 + rc_reset(&s->rc);
1962 +}
1963 +
1964 +/*
1965 + * Decode and validate LZMA properties (lc/lp/pb) and calculate the bit masks
1966 + * from the decoded lp and pb values. On success, the LZMA decoder state is
1967 + * reset and true is returned.
1968 + */
1969 +static bool lzma_props(struct xz_dec_lzma2 *s, uint8_t props)
1970 +{
1971 + if (props > (4 * 5 + 4) * 9 + 8)
1972 + return false;
1973 +
1974 + s->lzma.pos_mask = 0;
1975 + while (props >= 9 * 5) {
1976 + props -= 9 * 5;
1977 + ++s->lzma.pos_mask;
1978 + }
1979 +
1980 + s->lzma.pos_mask = (1 << s->lzma.pos_mask) - 1;
1981 +
1982 + s->lzma.literal_pos_mask = 0;
1983 + while (props >= 9) {
1984 + props -= 9;
1985 + ++s->lzma.literal_pos_mask;
1986 + }
1987 +
1988 + s->lzma.lc = props;
1989 +
1990 + if (s->lzma.lc + s->lzma.literal_pos_mask > 4)
1991 + return false;
1992 +
1993 + s->lzma.literal_pos_mask = (1 << s->lzma.literal_pos_mask) - 1;
1994 +
1995 + lzma_reset(s);
1996 +
1997 + return true;
1998 +}
1999 +
2000 +/*********
2001 + * LZMA2 *
2002 + *********/
2003 +
2004 +/*
2005 + * The LZMA decoder assumes that if the input limit (s->rc.in_limit) hasn't
2006 + * been exceeded, it is safe to read up to LZMA_IN_REQUIRED bytes. This
2007 + * wrapper function takes care of making the LZMA decoder's assumption safe.
2008 + *
2009 + * As long as there is plenty of input left to be decoded in the current LZMA
2010 + * chunk, we decode directly from the caller-supplied input buffer until
2011 + * there's LZMA_IN_REQUIRED bytes left. Those remaining bytes are copied into
2012 + * s->temp.buf, which (hopefully) gets filled on the next call to this
2013 + * function. We decode a few bytes from the temporary buffer so that we can
2014 + * continue decoding from the caller-supplied input buffer again.
2015 + */
2016 +static bool lzma2_lzma(struct xz_dec_lzma2 *s, struct xz_buf *b)
2017 +{
2018 + size_t in_avail;
2019 + uint32_t tmp;
2020 +
2021 + in_avail = b->in_size - b->in_pos;
2022 + if (s->temp.size > 0 || s->lzma2.compressed == 0) {
2023 + tmp = 2 * LZMA_IN_REQUIRED - s->temp.size;
2024 + if (tmp > s->lzma2.compressed - s->temp.size)
2025 + tmp = s->lzma2.compressed - s->temp.size;
2026 + if (tmp > in_avail)
2027 + tmp = in_avail;
2028 +
2029 + memcpy(s->temp.buf + s->temp.size, b->in + b->in_pos, tmp);
2030 +
2031 + if (s->temp.size + tmp == s->lzma2.compressed) {
2032 + memzero(s->temp.buf + s->temp.size + tmp,
2033 + sizeof(s->temp.buf)
2034 + - s->temp.size - tmp);
2035 + s->rc.in_limit = s->temp.size + tmp;
2036 + } else if (s->temp.size + tmp < LZMA_IN_REQUIRED) {
2037 + s->temp.size += tmp;
2038 + b->in_pos += tmp;
2039 + return true;
2040 + } else {
2041 + s->rc.in_limit = s->temp.size + tmp - LZMA_IN_REQUIRED;
2042 + }
2043 +
2044 + s->rc.in = s->temp.buf;
2045 + s->rc.in_pos = 0;
2046 +
2047 + if (!lzma_main(s) || s->rc.in_pos > s->temp.size + tmp)
2048 + return false;
2049 +
2050 + s->lzma2.compressed -= s->rc.in_pos;
2051 +
2052 + if (s->rc.in_pos < s->temp.size) {
2053 + s->temp.size -= s->rc.in_pos;
2054 + memmove(s->temp.buf, s->temp.buf + s->rc.in_pos,
2055 + s->temp.size);
2056 + return true;
2057 + }
2058 +
2059 + b->in_pos += s->rc.in_pos - s->temp.size;
2060 + s->temp.size = 0;
2061 + }
2062 +
2063 + in_avail = b->in_size - b->in_pos;
2064 + if (in_avail >= LZMA_IN_REQUIRED) {
2065 + s->rc.in = b->in;
2066 + s->rc.in_pos = b->in_pos;
2067 +
2068 + if (in_avail >= s->lzma2.compressed + LZMA_IN_REQUIRED)
2069 + s->rc.in_limit = b->in_pos + s->lzma2.compressed;
2070 + else
2071 + s->rc.in_limit = b->in_size - LZMA_IN_REQUIRED;
2072 +
2073 + if (!lzma_main(s))
2074 + return false;
2075 +
2076 + in_avail = s->rc.in_pos - b->in_pos;
2077 + if (in_avail > s->lzma2.compressed)
2078 + return false;
2079 +
2080 + s->lzma2.compressed -= in_avail;
2081 + b->in_pos = s->rc.in_pos;
2082 + }
2083 +
2084 + in_avail = b->in_size - b->in_pos;
2085 + if (in_avail < LZMA_IN_REQUIRED) {
2086 + if (in_avail > s->lzma2.compressed)
2087 + in_avail = s->lzma2.compressed;
2088 +
2089 + memcpy(s->temp.buf, b->in + b->in_pos, in_avail);
2090 + s->temp.size = in_avail;
2091 + b->in_pos += in_avail;
2092 + }
2093 +
2094 + return true;
2095 +}
2096 +
2097 +/*
2098 + * Take care of the LZMA2 control layer, and forward the job of actual LZMA
2099 + * decoding or copying of uncompressed chunks to other functions.
2100 + */
2101 +XZ_EXTERN enum xz_ret xz_dec_lzma2_run(struct xz_dec_lzma2 *s,
2102 + struct xz_buf *b)
2103 +{
2104 + uint32_t tmp;
2105 +
2106 + while (b->in_pos < b->in_size || s->lzma2.sequence == SEQ_LZMA_RUN) {
2107 + switch (s->lzma2.sequence) {
2108 + case SEQ_CONTROL:
2109 + /*
2110 + * LZMA2 control byte
2111 + *
2112 + * Exact values:
2113 + * 0x00 End marker
2114 + * 0x01 Dictionary reset followed by
2115 + * an uncompressed chunk
2116 + * 0x02 Uncompressed chunk (no dictionary reset)
2117 + *
2118 + * Highest three bits (s->control & 0xE0):
2119 + * 0xE0 Dictionary reset, new properties and state
2120 + * reset, followed by LZMA compressed chunk
2121 + * 0xC0 New properties and state reset, followed
2122 + * by LZMA compressed chunk (no dictionary
2123 + * reset)
2124 + * 0xA0 State reset using old properties,
2125 + * followed by LZMA compressed chunk (no
2126 + * dictionary reset)
2127 + * 0x80 LZMA chunk (no dictionary or state reset)
2128 + *
2129 + * For LZMA compressed chunks, the lowest five bits
2130 + * (s->control & 1F) are the highest bits of the
2131 + * uncompressed size (bits 16-20).
2132 + *
2133 + * A new LZMA2 stream must begin with a dictionary
2134 + * reset. The first LZMA chunk must set new
2135 + * properties and reset the LZMA state.
2136 + *
2137 + * Values that don't match anything described above
2138 + * are invalid and we return XZ_DATA_ERROR.
2139 + */
2140 + tmp = b->in[b->in_pos++];
2141 +
2142 + if (tmp >= 0xE0 || tmp == 0x01) {
2143 + s->lzma2.need_props = true;
2144 + s->lzma2.need_dict_reset = false;
2145 + dict_reset(&s->dict, b);
2146 + } else if (s->lzma2.need_dict_reset) {
2147 + return XZ_DATA_ERROR;
2148 + }
2149 +
2150 + if (tmp >= 0x80) {
2151 + s->lzma2.uncompressed = (tmp & 0x1F) << 16;
2152 + s->lzma2.sequence = SEQ_UNCOMPRESSED_1;
2153 +
2154 + if (tmp >= 0xC0) {
2155 + /*
2156 + * When there are new properties,
2157 + * state reset is done at
2158 + * SEQ_PROPERTIES.
2159 + */
2160 + s->lzma2.need_props = false;
2161 + s->lzma2.next_sequence
2162 + = SEQ_PROPERTIES;
2163 +
2164 + } else if (s->lzma2.need_props) {
2165 + return XZ_DATA_ERROR;
2166 +
2167 + } else {
2168 + s->lzma2.next_sequence
2169 + = SEQ_LZMA_PREPARE;
2170 + if (tmp >= 0xA0)
2171 + lzma_reset(s);
2172 + }
2173 + } else {
2174 + if (tmp == 0x00)
2175 + return XZ_STREAM_END;
2176 +
2177 + if (tmp > 0x02)
2178 + return XZ_DATA_ERROR;
2179 +
2180 + s->lzma2.sequence = SEQ_COMPRESSED_0;
2181 + s->lzma2.next_sequence = SEQ_COPY;
2182 + }
2183 +
2184 + break;
2185 +
2186 + case SEQ_UNCOMPRESSED_1:
2187 + s->lzma2.uncompressed
2188 + += (uint32_t)b->in[b->in_pos++] << 8;
2189 + s->lzma2.sequence = SEQ_UNCOMPRESSED_2;
2190 + break;
2191 +
2192 + case SEQ_UNCOMPRESSED_2:
2193 + s->lzma2.uncompressed
2194 + += (uint32_t)b->in[b->in_pos++] + 1;
2195 + s->lzma2.sequence = SEQ_COMPRESSED_0;
2196 + break;
2197 +
2198 + case SEQ_COMPRESSED_0:
2199 + s->lzma2.compressed
2200 + = (uint32_t)b->in[b->in_pos++] << 8;
2201 + s->lzma2.sequence = SEQ_COMPRESSED_1;
2202 + break;
2203 +
2204 + case SEQ_COMPRESSED_1:
2205 + s->lzma2.compressed
2206 + += (uint32_t)b->in[b->in_pos++] + 1;
2207 + s->lzma2.sequence = s->lzma2.next_sequence;
2208 + break;
2209 +
2210 + case SEQ_PROPERTIES:
2211 + if (!lzma_props(s, b->in[b->in_pos++]))
2212 + return XZ_DATA_ERROR;
2213 +
2214 + s->lzma2.sequence = SEQ_LZMA_PREPARE;
2215 +
2216 + case SEQ_LZMA_PREPARE:
2217 + if (s->lzma2.compressed < RC_INIT_BYTES)
2218 + return XZ_DATA_ERROR;
2219 +
2220 + if (!rc_read_init(&s->rc, b))
2221 + return XZ_OK;
2222 +
2223 + s->lzma2.compressed -= RC_INIT_BYTES;
2224 + s->lzma2.sequence = SEQ_LZMA_RUN;
2225 +
2226 + case SEQ_LZMA_RUN:
2227 + /*
2228 + * Set dictionary limit to indicate how much we want
2229 + * to be encoded at maximum. Decode new data into the
2230 + * dictionary. Flush the new data from dictionary to
2231 + * b->out. Check if we finished decoding this chunk.
2232 + * In case the dictionary got full but we didn't fill
2233 + * the output buffer yet, we may run this loop
2234 + * multiple times without changing s->lzma2.sequence.
2235 + */
2236 + dict_limit(&s->dict, min_t(size_t,
2237 + b->out_size - b->out_pos,
2238 + s->lzma2.uncompressed));
2239 + if (!lzma2_lzma(s, b))
2240 + return XZ_DATA_ERROR;
2241 +
2242 + s->lzma2.uncompressed -= dict_flush(&s->dict, b);
2243 +
2244 + if (s->lzma2.uncompressed == 0) {
2245 + if (s->lzma2.compressed > 0 || s->lzma.len > 0
2246 + || !rc_is_finished(&s->rc))
2247 + return XZ_DATA_ERROR;
2248 +
2249 + rc_reset(&s->rc);
2250 + s->lzma2.sequence = SEQ_CONTROL;
2251 +
2252 + } else if (b->out_pos == b->out_size
2253 + || (b->in_pos == b->in_size
2254 + && s->temp.size
2255 + < s->lzma2.compressed)) {
2256 + return XZ_OK;
2257 + }
2258 +
2259 + break;
2260 +
2261 + case SEQ_COPY:
2262 + dict_uncompressed(&s->dict, b, &s->lzma2.compressed);
2263 + if (s->lzma2.compressed > 0)
2264 + return XZ_OK;
2265 +
2266 + s->lzma2.sequence = SEQ_CONTROL;
2267 + break;
2268 + }
2269 + }
2270 +
2271 + return XZ_OK;
2272 +}
2273 +
2274 +XZ_EXTERN struct xz_dec_lzma2 *xz_dec_lzma2_create(enum xz_mode mode,
2275 + uint32_t dict_max)
2276 +{
2277 + struct xz_dec_lzma2 *s = kmalloc(sizeof(*s), GFP_KERNEL);
2278 + if (s == NULL)
2279 + return NULL;
2280 +
2281 + s->dict.mode = mode;
2282 + s->dict.size_max = dict_max;
2283 +
2284 + if (DEC_IS_PREALLOC(mode)) {
2285 + s->dict.buf = vmalloc(dict_max);
2286 + if (s->dict.buf == NULL) {
2287 + kfree(s);
2288 + return NULL;
2289 + }
2290 + } else if (DEC_IS_DYNALLOC(mode)) {
2291 + s->dict.buf = NULL;
2292 + s->dict.allocated = 0;
2293 + }
2294 +
2295 + return s;
2296 +}
2297 +
2298 +XZ_EXTERN enum xz_ret xz_dec_lzma2_reset(struct xz_dec_lzma2 *s, uint8_t props)
2299 +{
2300 + /* This limits dictionary size to 3 GiB to keep parsing simpler. */
2301 + if (props > 39)
2302 + return XZ_OPTIONS_ERROR;
2303 +
2304 + s->dict.size = 2 + (props & 1);
2305 + s->dict.size <<= (props >> 1) + 11;
2306 +
2307 + if (DEC_IS_MULTI(s->dict.mode)) {
2308 + if (s->dict.size > s->dict.size_max)
2309 + return XZ_MEMLIMIT_ERROR;
2310 +
2311 + s->dict.end = s->dict.size;
2312 +
2313 + if (DEC_IS_DYNALLOC(s->dict.mode)) {
2314 + if (s->dict.allocated < s->dict.size) {
2315 + vfree(s->dict.buf);
2316 + s->dict.buf = vmalloc(s->dict.size);
2317 + if (s->dict.buf == NULL) {
2318 + s->dict.allocated = 0;
2319 + return XZ_MEM_ERROR;
2320 + }
2321 + }
2322 + }
2323 + }
2324 +
2325 + s->lzma.len = 0;
2326 +
2327 + s->lzma2.sequence = SEQ_CONTROL;
2328 + s->lzma2.need_dict_reset = true;
2329 +
2330 + s->temp.size = 0;
2331 +
2332 + return XZ_OK;
2333 +}
2334 +
2335 +XZ_EXTERN void xz_dec_lzma2_end(struct xz_dec_lzma2 *s)
2336 +{
2337 + if (DEC_IS_MULTI(s->dict.mode))
2338 + vfree(s->dict.buf);
2339 +
2340 + kfree(s);
2341 +}
2342 diff --git a/lib/xz/xz_dec_stream.c b/lib/xz/xz_dec_stream.c
2343 new file mode 100644
2344 index 0000000..ac809b1
2345 --- /dev/null
2346 +++ b/lib/xz/xz_dec_stream.c
2347 @@ -0,0 +1,821 @@
2348 +/*
2349 + * .xz Stream decoder
2350 + *
2351 + * Author: Lasse Collin <lasse.collin@tukaani.org>
2352 + *
2353 + * This file has been put into the public domain.
2354 + * You can do whatever you want with this file.
2355 + */
2356 +
2357 +#include "xz_private.h"
2358 +#include "xz_stream.h"
2359 +
2360 +/* Hash used to validate the Index field */
2361 +struct xz_dec_hash {
2362 + vli_type unpadded;
2363 + vli_type uncompressed;
2364 + uint32_t crc32;
2365 +};
2366 +
2367 +struct xz_dec {
2368 + /* Position in dec_main() */
2369 + enum {
2370 + SEQ_STREAM_HEADER,
2371 + SEQ_BLOCK_START,
2372 + SEQ_BLOCK_HEADER,
2373 + SEQ_BLOCK_UNCOMPRESS,
2374 + SEQ_BLOCK_PADDING,
2375 + SEQ_BLOCK_CHECK,
2376 + SEQ_INDEX,
2377 + SEQ_INDEX_PADDING,
2378 + SEQ_INDEX_CRC32,
2379 + SEQ_STREAM_FOOTER
2380 + } sequence;
2381 +
2382 + /* Position in variable-length integers and Check fields */
2383 + uint32_t pos;
2384 +
2385 + /* Variable-length integer decoded by dec_vli() */
2386 + vli_type vli;
2387 +
2388 + /* Saved in_pos and out_pos */
2389 + size_t in_start;
2390 + size_t out_start;
2391 +
2392 + /* CRC32 value in Block or Index */
2393 + uint32_t crc32;
2394 +
2395 + /* Type of the integrity check calculated from uncompressed data */
2396 + enum xz_check check_type;
2397 +
2398 + /* Operation mode */
2399 + enum xz_mode mode;
2400 +
2401 + /*
2402 + * True if the next call to xz_dec_run() is allowed to return
2403 + * XZ_BUF_ERROR.
2404 + */
2405 + bool allow_buf_error;
2406 +
2407 + /* Information stored in Block Header */
2408 + struct {
2409 + /*
2410 + * Value stored in the Compressed Size field, or
2411 + * VLI_UNKNOWN if Compressed Size is not present.
2412 + */
2413 + vli_type compressed;
2414 +
2415 + /*
2416 + * Value stored in the Uncompressed Size field, or
2417 + * VLI_UNKNOWN if Uncompressed Size is not present.
2418 + */
2419 + vli_type uncompressed;
2420 +
2421 + /* Size of the Block Header field */
2422 + uint32_t size;
2423 + } block_header;
2424 +
2425 + /* Information collected when decoding Blocks */
2426 + struct {
2427 + /* Observed compressed size of the current Block */
2428 + vli_type compressed;
2429 +
2430 + /* Observed uncompressed size of the current Block */
2431 + vli_type uncompressed;
2432 +
2433 + /* Number of Blocks decoded so far */
2434 + vli_type count;
2435 +
2436 + /*
2437 + * Hash calculated from the Block sizes. This is used to
2438 + * validate the Index field.
2439 + */
2440 + struct xz_dec_hash hash;
2441 + } block;
2442 +
2443 + /* Variables needed when verifying the Index field */
2444 + struct {
2445 + /* Position in dec_index() */
2446 + enum {
2447 + SEQ_INDEX_COUNT,
2448 + SEQ_INDEX_UNPADDED,
2449 + SEQ_INDEX_UNCOMPRESSED
2450 + } sequence;
2451 +
2452 + /* Size of the Index in bytes */
2453 + vli_type size;
2454 +
2455 + /* Number of Records (matches block.count in valid files) */
2456 + vli_type count;
2457 +
2458 + /*
2459 + * Hash calculated from the Records (matches block.hash in
2460 + * valid files).
2461 + */
2462 + struct xz_dec_hash hash;
2463 + } index;
2464 +
2465 + /*
2466 + * Temporary buffer needed to hold Stream Header, Block Header,
2467 + * and Stream Footer. The Block Header is the biggest (1 KiB)
2468 + * so we reserve space according to that. buf[] has to be aligned
2469 + * to a multiple of four bytes; the size_t variables before it
2470 + * should guarantee this.
2471 + */
2472 + struct {
2473 + size_t pos;
2474 + size_t size;
2475 + uint8_t buf[1024];
2476 + } temp;
2477 +
2478 + struct xz_dec_lzma2 *lzma2;
2479 +
2480 +#ifdef XZ_DEC_BCJ
2481 + struct xz_dec_bcj *bcj;
2482 + bool bcj_active;
2483 +#endif
2484 +};
2485 +
2486 +#ifdef XZ_DEC_ANY_CHECK
2487 +/* Sizes of the Check field with different Check IDs */
2488 +static const uint8_t check_sizes[16] = {
2489 + 0,
2490 + 4, 4, 4,
2491 + 8, 8, 8,
2492 + 16, 16, 16,
2493 + 32, 32, 32,
2494 + 64, 64, 64
2495 +};
2496 +#endif
2497 +
2498 +/*
2499 + * Fill s->temp by copying data starting from b->in[b->in_pos]. Caller
2500 + * must have set s->temp.pos to indicate how much data we are supposed
2501 + * to copy into s->temp.buf. Return true once s->temp.pos has reached
2502 + * s->temp.size.
2503 + */
2504 +static bool fill_temp(struct xz_dec *s, struct xz_buf *b)
2505 +{
2506 + size_t copy_size = min_t(size_t,
2507 + b->in_size - b->in_pos, s->temp.size - s->temp.pos);
2508 +
2509 + memcpy(s->temp.buf + s->temp.pos, b->in + b->in_pos, copy_size);
2510 + b->in_pos += copy_size;
2511 + s->temp.pos += copy_size;
2512 +
2513 + if (s->temp.pos == s->temp.size) {
2514 + s->temp.pos = 0;
2515 + return true;
2516 + }
2517 +
2518 + return false;
2519 +}
2520 +
2521 +/* Decode a variable-length integer (little-endian base-128 encoding) */
2522 +static enum xz_ret dec_vli(struct xz_dec *s, const uint8_t *in,
2523 + size_t *in_pos, size_t in_size)
2524 +{
2525 + uint8_t byte;
2526 +
2527 + if (s->pos == 0)
2528 + s->vli = 0;
2529 +
2530 + while (*in_pos < in_size) {
2531 + byte = in[*in_pos];
2532 + ++*in_pos;
2533 +
2534 + s->vli |= (vli_type)(byte & 0x7F) << s->pos;
2535 +
2536 + if ((byte & 0x80) == 0) {
2537 + /* Don't allow non-minimal encodings. */
2538 + if (byte == 0 && s->pos != 0)
2539 + return XZ_DATA_ERROR;
2540 +
2541 + s->pos = 0;
2542 + return XZ_STREAM_END;
2543 + }
2544 +
2545 + s->pos += 7;
2546 + if (s->pos == 7 * VLI_BYTES_MAX)
2547 + return XZ_DATA_ERROR;
2548 + }
2549 +
2550 + return XZ_OK;
2551 +}
2552 +
2553 +/*
2554 + * Decode the Compressed Data field from a Block. Update and validate
2555 + * the observed compressed and uncompressed sizes of the Block so that
2556 + * they don't exceed the values possibly stored in the Block Header
2557 + * (validation assumes that no integer overflow occurs, since vli_type
2558 + * is normally uint64_t). Update the CRC32 if presence of the CRC32
2559 + * field was indicated in Stream Header.
2560 + *
2561 + * Once the decoding is finished, validate that the observed sizes match
2562 + * the sizes possibly stored in the Block Header. Update the hash and
2563 + * Block count, which are later used to validate the Index field.
2564 + */
2565 +static enum xz_ret dec_block(struct xz_dec *s, struct xz_buf *b)
2566 +{
2567 + enum xz_ret ret;
2568 +
2569 + s->in_start = b->in_pos;
2570 + s->out_start = b->out_pos;
2571 +
2572 +#ifdef XZ_DEC_BCJ
2573 + if (s->bcj_active)
2574 + ret = xz_dec_bcj_run(s->bcj, s->lzma2, b);
2575 + else
2576 +#endif
2577 + ret = xz_dec_lzma2_run(s->lzma2, b);
2578 +
2579 + s->block.compressed += b->in_pos - s->in_start;
2580 + s->block.uncompressed += b->out_pos - s->out_start;
2581 +
2582 + /*
2583 + * There is no need to separately check for VLI_UNKNOWN, since
2584 + * the observed sizes are always smaller than VLI_UNKNOWN.
2585 + */
2586 + if (s->block.compressed > s->block_header.compressed
2587 + || s->block.uncompressed
2588 + > s->block_header.uncompressed)
2589 + return XZ_DATA_ERROR;
2590 +
2591 + if (s->check_type == XZ_CHECK_CRC32)
2592 + s->crc32 = xz_crc32(b->out + s->out_start,
2593 + b->out_pos - s->out_start, s->crc32);
2594 +
2595 + if (ret == XZ_STREAM_END) {
2596 + if (s->block_header.compressed != VLI_UNKNOWN
2597 + && s->block_header.compressed
2598 + != s->block.compressed)
2599 + return XZ_DATA_ERROR;
2600 +
2601 + if (s->block_header.uncompressed != VLI_UNKNOWN
2602 + && s->block_header.uncompressed
2603 + != s->block.uncompressed)
2604 + return XZ_DATA_ERROR;
2605 +
2606 + s->block.hash.unpadded += s->block_header.size
2607 + + s->block.compressed;
2608 +
2609 +#ifdef XZ_DEC_ANY_CHECK
2610 + s->block.hash.unpadded += check_sizes[s->check_type];
2611 +#else
2612 + if (s->check_type == XZ_CHECK_CRC32)
2613 + s->block.hash.unpadded += 4;
2614 +#endif
2615 +
2616 + s->block.hash.uncompressed += s->block.uncompressed;
2617 + s->block.hash.crc32 = xz_crc32(
2618 + (const uint8_t *)&s->block.hash,
2619 + sizeof(s->block.hash), s->block.hash.crc32);
2620 +
2621 + ++s->block.count;
2622 + }
2623 +
2624 + return ret;
2625 +}
2626 +
2627 +/* Update the Index size and the CRC32 value. */
2628 +static void index_update(struct xz_dec *s, const struct xz_buf *b)
2629 +{
2630 + size_t in_used = b->in_pos - s->in_start;
2631 + s->index.size += in_used;
2632 + s->crc32 = xz_crc32(b->in + s->in_start, in_used, s->crc32);
2633 +}
2634 +
2635 +/*
2636 + * Decode the Number of Records, Unpadded Size, and Uncompressed Size
2637 + * fields from the Index field. That is, Index Padding and CRC32 are not
2638 + * decoded by this function.
2639 + *
2640 + * This can return XZ_OK (more input needed), XZ_STREAM_END (everything
2641 + * successfully decoded), or XZ_DATA_ERROR (input is corrupt).
2642 + */
2643 +static enum xz_ret dec_index(struct xz_dec *s, struct xz_buf *b)
2644 +{
2645 + enum xz_ret ret;
2646 +
2647 + do {
2648 + ret = dec_vli(s, b->in, &b->in_pos, b->in_size);
2649 + if (ret != XZ_STREAM_END) {
2650 + index_update(s, b);
2651 + return ret;
2652 + }
2653 +
2654 + switch (s->index.sequence) {
2655 + case SEQ_INDEX_COUNT:
2656 + s->index.count = s->vli;
2657 +
2658 + /*
2659 + * Validate that the Number of Records field
2660 + * indicates the same number of Records as
2661 + * there were Blocks in the Stream.
2662 + */
2663 + if (s->index.count != s->block.count)
2664 + return XZ_DATA_ERROR;
2665 +
2666 + s->index.sequence = SEQ_INDEX_UNPADDED;
2667 + break;
2668 +
2669 + case SEQ_INDEX_UNPADDED:
2670 + s->index.hash.unpadded += s->vli;
2671 + s->index.sequence = SEQ_INDEX_UNCOMPRESSED;
2672 + break;
2673 +
2674 + case SEQ_INDEX_UNCOMPRESSED:
2675 + s->index.hash.uncompressed += s->vli;
2676 + s->index.hash.crc32 = xz_crc32(
2677 + (const uint8_t *)&s->index.hash,
2678 + sizeof(s->index.hash),
2679 + s->index.hash.crc32);
2680 + --s->index.count;
2681 + s->index.sequence = SEQ_INDEX_UNPADDED;
2682 + break;
2683 + }
2684 + } while (s->index.count > 0);
2685 +
2686 + return XZ_STREAM_END;
2687 +}
2688 +
2689 +/*
2690 + * Validate that the next four input bytes match the value of s->crc32.
2691 + * s->pos must be zero when starting to validate the first byte.
2692 + */
2693 +static enum xz_ret crc32_validate(struct xz_dec *s, struct xz_buf *b)
2694 +{
2695 + do {
2696 + if (b->in_pos == b->in_size)
2697 + return XZ_OK;
2698 +
2699 + if (((s->crc32 >> s->pos) & 0xFF) != b->in[b->in_pos++])
2700 + return XZ_DATA_ERROR;
2701 +
2702 + s->pos += 8;
2703 +
2704 + } while (s->pos < 32);
2705 +
2706 + s->crc32 = 0;
2707 + s->pos = 0;
2708 +
2709 + return XZ_STREAM_END;
2710 +}
2711 +
2712 +#ifdef XZ_DEC_ANY_CHECK
2713 +/*
2714 + * Skip over the Check field when the Check ID is not supported.
2715 + * Returns true once the whole Check field has been skipped over.
2716 + */
2717 +static bool check_skip(struct xz_dec *s, struct xz_buf *b)
2718 +{
2719 + while (s->pos < check_sizes[s->check_type]) {
2720 + if (b->in_pos == b->in_size)
2721 + return false;
2722 +
2723 + ++b->in_pos;
2724 + ++s->pos;
2725 + }
2726 +
2727 + s->pos = 0;
2728 +
2729 + return true;
2730 +}
2731 +#endif
2732 +
2733 +/* Decode the Stream Header field (the first 12 bytes of the .xz Stream). */
2734 +static enum xz_ret dec_stream_header(struct xz_dec *s)
2735 +{
2736 + if (!memeq(s->temp.buf, HEADER_MAGIC, HEADER_MAGIC_SIZE))
2737 + return XZ_FORMAT_ERROR;
2738 +
2739 + if (xz_crc32(s->temp.buf + HEADER_MAGIC_SIZE, 2, 0)
2740 + != get_le32(s->temp.buf + HEADER_MAGIC_SIZE + 2))
2741 + return XZ_DATA_ERROR;
2742 +
2743 + if (s->temp.buf[HEADER_MAGIC_SIZE] != 0)
2744 + return XZ_OPTIONS_ERROR;
2745 +
2746 + /*
2747 + * Of integrity checks, we support only none (Check ID = 0) and
2748 + * CRC32 (Check ID = 1). However, if XZ_DEC_ANY_CHECK is defined,
2749 + * we will accept other check types too, but then the check won't
2750 + * be verified and a warning (XZ_UNSUPPORTED_CHECK) will be given.
2751 + */
2752 + s->check_type = s->temp.buf[HEADER_MAGIC_SIZE + 1];
2753 +
2754 +#ifdef XZ_DEC_ANY_CHECK
2755 + if (s->check_type > XZ_CHECK_MAX)
2756 + return XZ_OPTIONS_ERROR;
2757 +
2758 + if (s->check_type > XZ_CHECK_CRC32)
2759 + return XZ_UNSUPPORTED_CHECK;
2760 +#else
2761 + if (s->check_type > XZ_CHECK_CRC32)
2762 + return XZ_OPTIONS_ERROR;
2763 +#endif
2764 +
2765 + return XZ_OK;
2766 +}
2767 +
2768 +/* Decode the Stream Footer field (the last 12 bytes of the .xz Stream) */
2769 +static enum xz_ret dec_stream_footer(struct xz_dec *s)
2770 +{
2771 + if (!memeq(s->temp.buf + 10, FOOTER_MAGIC, FOOTER_MAGIC_SIZE))
2772 + return XZ_DATA_ERROR;
2773 +
2774 + if (xz_crc32(s->temp.buf + 4, 6, 0) != get_le32(s->temp.buf))
2775 + return XZ_DATA_ERROR;
2776 +
2777 + /*
2778 + * Validate Backward Size. Note that we never added the size of the
2779 + * Index CRC32 field to s->index.size, thus we use s->index.size / 4
2780 + * instead of s->index.size / 4 - 1.
2781 + */
2782 + if ((s->index.size >> 2) != get_le32(s->temp.buf + 4))
2783 + return XZ_DATA_ERROR;
2784 +
2785 + if (s->temp.buf[8] != 0 || s->temp.buf[9] != s->check_type)
2786 + return XZ_DATA_ERROR;
2787 +
2788 + /*
2789 + * Use XZ_STREAM_END instead of XZ_OK to be more convenient
2790 + * for the caller.
2791 + */
2792 + return XZ_STREAM_END;
2793 +}
2794 +
2795 +/* Decode the Block Header and initialize the filter chain. */
2796 +static enum xz_ret dec_block_header(struct xz_dec *s)
2797 +{
2798 + enum xz_ret ret;
2799 +
2800 + /*
2801 + * Validate the CRC32. We know that the temp buffer is at least
2802 + * eight bytes so this is safe.
2803 + */
2804 + s->temp.size -= 4;
2805 + if (xz_crc32(s->temp.buf, s->temp.size, 0)
2806 + != get_le32(s->temp.buf + s->temp.size))
2807 + return XZ_DATA_ERROR;
2808 +
2809 + s->temp.pos = 2;
2810 +
2811 + /*
2812 + * Catch unsupported Block Flags. We support only one or two filters
2813 + * in the chain, so we catch that with the same test.
2814 + */
2815 +#ifdef XZ_DEC_BCJ
2816 + if (s->temp.buf[1] & 0x3E)
2817 +#else
2818 + if (s->temp.buf[1] & 0x3F)
2819 +#endif
2820 + return XZ_OPTIONS_ERROR;
2821 +
2822 + /* Compressed Size */
2823 + if (s->temp.buf[1] & 0x40) {
2824 + if (dec_vli(s, s->temp.buf, &s->temp.pos, s->temp.size)
2825 + != XZ_STREAM_END)
2826 + return XZ_DATA_ERROR;
2827 +
2828 + s->block_header.compressed = s->vli;
2829 + } else {
2830 + s->block_header.compressed = VLI_UNKNOWN;
2831 + }
2832 +
2833 + /* Uncompressed Size */
2834 + if (s->temp.buf[1] & 0x80) {
2835 + if (dec_vli(s, s->temp.buf, &s->temp.pos, s->temp.size)
2836 + != XZ_STREAM_END)
2837 + return XZ_DATA_ERROR;
2838 +
2839 + s->block_header.uncompressed = s->vli;
2840 + } else {
2841 + s->block_header.uncompressed = VLI_UNKNOWN;
2842 + }
2843 +
2844 +#ifdef XZ_DEC_BCJ
2845 + /* If there are two filters, the first one must be a BCJ filter. */
2846 + s->bcj_active = s->temp.buf[1] & 0x01;
2847 + if (s->bcj_active) {
2848 + if (s->temp.size - s->temp.pos < 2)
2849 + return XZ_OPTIONS_ERROR;
2850 +
2851 + ret = xz_dec_bcj_reset(s->bcj, s->temp.buf[s->temp.pos++]);
2852 + if (ret != XZ_OK)
2853 + return ret;
2854 +
2855 + /*
2856 + * We don't support custom start offset,
2857 + * so Size of Properties must be zero.
2858 + */
2859 + if (s->temp.buf[s->temp.pos++] != 0x00)
2860 + return XZ_OPTIONS_ERROR;
2861 + }
2862 +#endif
2863 +
2864 + /* Valid Filter Flags always take at least two bytes. */
2865 + if (s->temp.size - s->temp.pos < 2)
2866 + return XZ_DATA_ERROR;
2867 +
2868 + /* Filter ID = LZMA2 */
2869 + if (s->temp.buf[s->temp.pos++] != 0x21)
2870 + return XZ_OPTIONS_ERROR;
2871 +
2872 + /* Size of Properties = 1-byte Filter Properties */
2873 + if (s->temp.buf[s->temp.pos++] != 0x01)
2874 + return XZ_OPTIONS_ERROR;
2875 +
2876 + /* Filter Properties contains LZMA2 dictionary size. */
2877 + if (s->temp.size - s->temp.pos < 1)
2878 + return XZ_DATA_ERROR;
2879 +
2880 + ret = xz_dec_lzma2_reset(s->lzma2, s->temp.buf[s->temp.pos++]);
2881 + if (ret != XZ_OK)
2882 + return ret;
2883 +
2884 + /* The rest must be Header Padding. */
2885 + while (s->temp.pos < s->temp.size)
2886 + if (s->temp.buf[s->temp.pos++] != 0x00)
2887 + return XZ_OPTIONS_ERROR;
2888 +
2889 + s->temp.pos = 0;
2890 + s->block.compressed = 0;
2891 + s->block.uncompressed = 0;
2892 +
2893 + return XZ_OK;
2894 +}
2895 +
2896 +static enum xz_ret dec_main(struct xz_dec *s, struct xz_buf *b)
2897 +{
2898 + enum xz_ret ret;
2899 +
2900 + /*
2901 + * Store the start position for the case when we are in the middle
2902 + * of the Index field.
2903 + */
2904 + s->in_start = b->in_pos;
2905 +
2906 + while (true) {
2907 + switch (s->sequence) {
2908 + case SEQ_STREAM_HEADER:
2909 + /*
2910 + * Stream Header is copied to s->temp, and then
2911 + * decoded from there. This way if the caller
2912 + * gives us only little input at a time, we can
2913 + * still keep the Stream Header decoding code
2914 + * simple. Similar approach is used in many places
2915 + * in this file.
2916 + */
2917 + if (!fill_temp(s, b))
2918 + return XZ_OK;
2919 +
2920 + /*
2921 + * If dec_stream_header() returns
2922 + * XZ_UNSUPPORTED_CHECK, it is still possible
2923 + * to continue decoding if working in multi-call
2924 + * mode. Thus, update s->sequence before calling
2925 + * dec_stream_header().
2926 + */
2927 + s->sequence = SEQ_BLOCK_START;
2928 +
2929 + ret = dec_stream_header(s);
2930 + if (ret != XZ_OK)
2931 + return ret;
2932 +
2933 + case SEQ_BLOCK_START:
2934 + /* We need one byte of input to continue. */
2935 + if (b->in_pos == b->in_size)
2936 + return XZ_OK;
2937 +
2938 + /* See if this is the beginning of the Index field. */
2939 + if (b->in[b->in_pos] == 0) {
2940 + s->in_start = b->in_pos++;
2941 + s->sequence = SEQ_INDEX;
2942 + break;
2943 + }
2944 +
2945 + /*
2946 + * Calculate the size of the Block Header and
2947 + * prepare to decode it.
2948 + */
2949 + s->block_header.size
2950 + = ((uint32_t)b->in[b->in_pos] + 1) * 4;
2951 +
2952 + s->temp.size = s->block_header.size;
2953 + s->temp.pos = 0;
2954 + s->sequence = SEQ_BLOCK_HEADER;
2955 +
2956 + case SEQ_BLOCK_HEADER:
2957 + if (!fill_temp(s, b))
2958 + return XZ_OK;
2959 +
2960 + ret = dec_block_header(s);
2961 + if (ret != XZ_OK)
2962 + return ret;
2963 +
2964 + s->sequence = SEQ_BLOCK_UNCOMPRESS;
2965 +
2966 + case SEQ_BLOCK_UNCOMPRESS:
2967 + ret = dec_block(s, b);
2968 + if (ret != XZ_STREAM_END)
2969 + return ret;
2970 +
2971 + s->sequence = SEQ_BLOCK_PADDING;
2972 +
2973 + case SEQ_BLOCK_PADDING:
2974 + /*
2975 + * Size of Compressed Data + Block Padding
2976 + * must be a multiple of four. We don't need
2977 + * s->block.compressed for anything else
2978 + * anymore, so we use it here to test the size
2979 + * of the Block Padding field.
2980 + */
2981 + while (s->block.compressed & 3) {
2982 + if (b->in_pos == b->in_size)
2983 + return XZ_OK;
2984 +
2985 + if (b->in[b->in_pos++] != 0)
2986 + return XZ_DATA_ERROR;
2987 +
2988 + ++s->block.compressed;
2989 + }
2990 +
2991 + s->sequence = SEQ_BLOCK_CHECK;
2992 +
2993 + case SEQ_BLOCK_CHECK:
2994 + if (s->check_type == XZ_CHECK_CRC32) {
2995 + ret = crc32_validate(s, b);
2996 + if (ret != XZ_STREAM_END)
2997 + return ret;
2998 + }
2999 +#ifdef XZ_DEC_ANY_CHECK
3000 + else if (!check_skip(s, b)) {
3001 + return XZ_OK;
3002 + }
3003 +#endif
3004 +
3005 + s->sequence = SEQ_BLOCK_START;
3006 + break;
3007 +
3008 + case SEQ_INDEX:
3009 + ret = dec_index(s, b);
3010 + if (ret != XZ_STREAM_END)
3011 + return ret;
3012 +
3013 + s->sequence = SEQ_INDEX_PADDING;
3014 +
3015 + case SEQ_INDEX_PADDING:
3016 + while ((s->index.size + (b->in_pos - s->in_start))
3017 + & 3) {
3018 + if (b->in_pos == b->in_size) {
3019 + index_update(s, b);
3020 + return XZ_OK;
3021 + }
3022 +
3023 + if (b->in[b->in_pos++] != 0)
3024 + return XZ_DATA_ERROR;
3025 + }
3026 +
3027 + /* Finish the CRC32 value and Index size. */
3028 + index_update(s, b);
3029 +
3030 + /* Compare the hashes to validate the Index field. */
3031 + if (!memeq(&s->block.hash, &s->index.hash,
3032 + sizeof(s->block.hash)))
3033 + return XZ_DATA_ERROR;
3034 +
3035 + s->sequence = SEQ_INDEX_CRC32;
3036 +
3037 + case SEQ_INDEX_CRC32:
3038 + ret = crc32_validate(s, b);
3039 + if (ret != XZ_STREAM_END)
3040 + return ret;
3041 +
3042 + s->temp.size = STREAM_HEADER_SIZE;
3043 + s->sequence = SEQ_STREAM_FOOTER;
3044 +
3045 + case SEQ_STREAM_FOOTER:
3046 + if (!fill_temp(s, b))
3047 + return XZ_OK;
3048 +
3049 + return dec_stream_footer(s);
3050 + }
3051 + }
3052 +
3053 + /* Never reached */
3054 +}
3055 +
3056 +/*
3057 + * xz_dec_run() is a wrapper for dec_main() to handle some special cases in
3058 + * multi-call and single-call decoding.
3059 + *
3060 + * In multi-call mode, we must return XZ_BUF_ERROR when it seems clear that we
3061 + * are not going to make any progress anymore. This is to prevent the caller
3062 + * from calling us infinitely when the input file is truncated or otherwise
3063 + * corrupt. Since zlib-style API allows that the caller fills the input buffer
3064 + * only when the decoder doesn't produce any new output, we have to be careful
3065 + * to avoid returning XZ_BUF_ERROR too easily: XZ_BUF_ERROR is returned only
3066 + * after the second consecutive call to xz_dec_run() that makes no progress.
3067 + *
3068 + * In single-call mode, if we couldn't decode everything and no error
3069 + * occurred, either the input is truncated or the output buffer is too small.
3070 + * Since we know that the last input byte never produces any output, we know
3071 + * that if all the input was consumed and decoding wasn't finished, the file
3072 + * must be corrupt. Otherwise the output buffer has to be too small or the
3073 + * file is corrupt in a way that decoding it produces too big output.
3074 + *
3075 + * If single-call decoding fails, we reset b->in_pos and b->out_pos back to
3076 + * their original values. This is because with some filter chains there won't
3077 + * be any valid uncompressed data in the output buffer unless the decoding
3078 + * actually succeeds (that's the price to pay of using the output buffer as
3079 + * the workspace).
3080 + */
3081 +XZ_EXTERN enum xz_ret xz_dec_run(struct xz_dec *s, struct xz_buf *b)
3082 +{
3083 + size_t in_start;
3084 + size_t out_start;
3085 + enum xz_ret ret;
3086 +
3087 + if (DEC_IS_SINGLE(s->mode))
3088 + xz_dec_reset(s);
3089 +
3090 + in_start = b->in_pos;
3091 + out_start = b->out_pos;
3092 + ret = dec_main(s, b);
3093 +
3094 + if (DEC_IS_SINGLE(s->mode)) {
3095 + if (ret == XZ_OK)
3096 + ret = b->in_pos == b->in_size
3097 + ? XZ_DATA_ERROR : XZ_BUF_ERROR;
3098 +
3099 + if (ret != XZ_STREAM_END) {
3100 + b->in_pos = in_start;
3101 + b->out_pos = out_start;
3102 + }
3103 +
3104 + } else if (ret == XZ_OK && in_start == b->in_pos
3105 + && out_start == b->out_pos) {
3106 + if (s->allow_buf_error)
3107 + ret = XZ_BUF_ERROR;
3108 +
3109 + s->allow_buf_error = true;
3110 + } else {
3111 + s->allow_buf_error = false;
3112 + }
3113 +
3114 + return ret;
3115 +}
3116 +
3117 +XZ_EXTERN struct xz_dec *xz_dec_init(enum xz_mode mode, uint32_t dict_max)
3118 +{
3119 + struct xz_dec *s = kmalloc(sizeof(*s), GFP_KERNEL);
3120 + if (s == NULL)
3121 + return NULL;
3122 +
3123 + s->mode = mode;
3124 +
3125 +#ifdef XZ_DEC_BCJ
3126 + s->bcj = xz_dec_bcj_create(DEC_IS_SINGLE(mode));
3127 + if (s->bcj == NULL)
3128 + goto error_bcj;
3129 +#endif
3130 +
3131 + s->lzma2 = xz_dec_lzma2_create(mode, dict_max);
3132 + if (s->lzma2 == NULL)
3133 + goto error_lzma2;
3134 +
3135 + xz_dec_reset(s);
3136 + return s;
3137 +
3138 +error_lzma2:
3139 +#ifdef XZ_DEC_BCJ
3140 + xz_dec_bcj_end(s->bcj);
3141 +error_bcj:
3142 +#endif
3143 + kfree(s);
3144 + return NULL;
3145 +}
3146 +
3147 +XZ_EXTERN void xz_dec_reset(struct xz_dec *s)
3148 +{
3149 + s->sequence = SEQ_STREAM_HEADER;
3150 + s->allow_buf_error = false;
3151 + s->pos = 0;
3152 + s->crc32 = 0;
3153 + memzero(&s->block, sizeof(s->block));
3154 + memzero(&s->index, sizeof(s->index));
3155 + s->temp.pos = 0;
3156 + s->temp.size = STREAM_HEADER_SIZE;
3157 +}
3158 +
3159 +XZ_EXTERN void xz_dec_end(struct xz_dec *s)
3160 +{
3161 + if (s != NULL) {
3162 + xz_dec_lzma2_end(s->lzma2);
3163 +#ifdef XZ_DEC_BCJ
3164 + xz_dec_bcj_end(s->bcj);
3165 +#endif
3166 + kfree(s);
3167 + }
3168 +}
3169 diff --git a/lib/xz/xz_dec_syms.c b/lib/xz/xz_dec_syms.c
3170 new file mode 100644
3171 index 0000000..32eb3c0
3172 --- /dev/null
3173 +++ b/lib/xz/xz_dec_syms.c
3174 @@ -0,0 +1,26 @@
3175 +/*
3176 + * XZ decoder module information
3177 + *
3178 + * Author: Lasse Collin <lasse.collin@tukaani.org>
3179 + *
3180 + * This file has been put into the public domain.
3181 + * You can do whatever you want with this file.
3182 + */
3183 +
3184 +#include <linux/module.h>
3185 +#include <linux/xz.h>
3186 +
3187 +EXPORT_SYMBOL(xz_dec_init);
3188 +EXPORT_SYMBOL(xz_dec_reset);
3189 +EXPORT_SYMBOL(xz_dec_run);
3190 +EXPORT_SYMBOL(xz_dec_end);
3191 +
3192 +MODULE_DESCRIPTION("XZ decompressor");
3193 +MODULE_VERSION("1.0");
3194 +MODULE_AUTHOR("Lasse Collin <lasse.collin@tukaani.org> and Igor Pavlov");
3195 +
3196 +/*
3197 + * This code is in the public domain, but in Linux it's simplest to just
3198 + * say it's GPL and consider the authors as the copyright holders.
3199 + */
3200 +MODULE_LICENSE("GPL");
3201 diff --git a/lib/xz/xz_dec_test.c b/lib/xz/xz_dec_test.c
3202 new file mode 100644
3203 index 0000000..da28a19
3204 --- /dev/null
3205 +++ b/lib/xz/xz_dec_test.c
3206 @@ -0,0 +1,220 @@
3207 +/*
3208 + * XZ decoder tester
3209 + *
3210 + * Author: Lasse Collin <lasse.collin@tukaani.org>
3211 + *
3212 + * This file has been put into the public domain.
3213 + * You can do whatever you want with this file.
3214 + */
3215 +
3216 +#include <linux/kernel.h>
3217 +#include <linux/module.h>
3218 +#include <linux/fs.h>
3219 +#include <linux/uaccess.h>
3220 +#include <linux/crc32.h>
3221 +#include <linux/xz.h>
3222 +
3223 +/* Maximum supported dictionary size */
3224 +#define DICT_MAX (1 << 20)
3225 +
3226 +/* Device name to pass to register_chrdev(). */
3227 +#define DEVICE_NAME "xz_dec_test"
3228 +
3229 +/* Dynamically allocated device major number */
3230 +static int device_major;
3231 +
3232 +/*
3233 + * We reuse the same decoder state, and thus can decode only one
3234 + * file at a time.
3235 + */
3236 +static bool device_is_open;
3237 +
3238 +/* XZ decoder state */
3239 +static struct xz_dec *state;
3240 +
3241 +/*
3242 + * Return value of xz_dec_run(). We need to avoid calling xz_dec_run() after
3243 + * it has returned XZ_STREAM_END, so we make this static.
3244 + */
3245 +static enum xz_ret ret;
3246 +
3247 +/*
3248 + * Input and output buffers. The input buffer is used as a temporary safe
3249 + * place for the data coming from the userspace.
3250 + */
3251 +static uint8_t buffer_in[1024];
3252 +static uint8_t buffer_out[1024];
3253 +
3254 +/*
3255 + * Structure to pass the input and output buffers to the XZ decoder.
3256 + * A few of the fields are never modified so we initialize them here.
3257 + */
3258 +static struct xz_buf buffers = {
3259 + .in = buffer_in,
3260 + .out = buffer_out,
3261 + .out_size = sizeof(buffer_out)
3262 +};
3263 +
3264 +/*
3265 + * CRC32 of uncompressed data. This is used to give the user a simple way
3266 + * to check that the decoder produces correct output.
3267 + */
3268 +static uint32_t crc;
3269 +
3270 +static int xz_dec_test_open(struct inode *i, struct file *f)
3271 +{
3272 + if (device_is_open)
3273 + return -EBUSY;
3274 +
3275 + device_is_open = true;
3276 +
3277 + xz_dec_reset(state);
3278 + ret = XZ_OK;
3279 + crc = 0xFFFFFFFF;
3280 +
3281 + buffers.in_pos = 0;
3282 + buffers.in_size = 0;
3283 + buffers.out_pos = 0;
3284 +
3285 + printk(KERN_INFO DEVICE_NAME ": opened\n");
3286 + return 0;
3287 +}
3288 +
3289 +static int xz_dec_test_release(struct inode *i, struct file *f)
3290 +{
3291 + device_is_open = false;
3292 +
3293 + if (ret == XZ_OK)
3294 + printk(KERN_INFO DEVICE_NAME ": input was truncated\n");
3295 +
3296 + printk(KERN_INFO DEVICE_NAME ": closed\n");
3297 + return 0;
3298 +}
3299 +
3300 +/*
3301 + * Decode the data given to us from the userspace. CRC32 of the uncompressed
3302 + * data is calculated and is printed at the end of successful decoding. The
3303 + * uncompressed data isn't stored anywhere for further use.
3304 + *
3305 + * The .xz file must have exactly one Stream and no Stream Padding. The data
3306 + * after the first Stream is considered to be garbage.
3307 + */
3308 +static ssize_t xz_dec_test_write(struct file *file, const char __user *buf,
3309 + size_t size, loff_t *pos)
3310 +{
3311 + size_t remaining;
3312 +
3313 + if (ret != XZ_OK) {
3314 + if (size > 0)
3315 + printk(KERN_INFO DEVICE_NAME ": %zu bytes of "
3316 + "garbage at the end of the file\n",
3317 + size);
3318 +
3319 + return -ENOSPC;
3320 + }
3321 +
3322 + printk(KERN_INFO DEVICE_NAME ": decoding %zu bytes of input\n",
3323 + size);
3324 +
3325 + remaining = size;
3326 + while ((remaining > 0 || buffers.out_pos == buffers.out_size)
3327 + && ret == XZ_OK) {
3328 + if (buffers.in_pos == buffers.in_size) {
3329 + buffers.in_pos = 0;
3330 + buffers.in_size = min(remaining, sizeof(buffer_in));
3331 + if (copy_from_user(buffer_in, buf, buffers.in_size))
3332 + return -EFAULT;
3333 +
3334 + buf += buffers.in_size;
3335 + remaining -= buffers.in_size;
3336 + }
3337 +
3338 + buffers.out_pos = 0;
3339 + ret = xz_dec_run(state, &buffers);
3340 + crc = crc32(crc, buffer_out, buffers.out_pos);
3341 + }
3342 +
3343 + switch (ret) {
3344 + case XZ_OK:
3345 + printk(KERN_INFO DEVICE_NAME ": XZ_OK\n");
3346 + return size;
3347 +
3348 + case XZ_STREAM_END:
3349 + printk(KERN_INFO DEVICE_NAME ": XZ_STREAM_END, "
3350 + "CRC32 = 0x%08X\n", ~crc);
3351 + return size - remaining - (buffers.in_size - buffers.in_pos);
3352 +
3353 + case XZ_MEMLIMIT_ERROR:
3354 + printk(KERN_INFO DEVICE_NAME ": XZ_MEMLIMIT_ERROR\n");
3355 + break;
3356 +
3357 + case XZ_FORMAT_ERROR:
3358 + printk(KERN_INFO DEVICE_NAME ": XZ_FORMAT_ERROR\n");
3359 + break;
3360 +
3361 + case XZ_OPTIONS_ERROR:
3362 + printk(KERN_INFO DEVICE_NAME ": XZ_OPTIONS_ERROR\n");
3363 + break;
3364 +
3365 + case XZ_DATA_ERROR:
3366 + printk(KERN_INFO DEVICE_NAME ": XZ_DATA_ERROR\n");
3367 + break;
3368 +
3369 + case XZ_BUF_ERROR:
3370 + printk(KERN_INFO DEVICE_NAME ": XZ_BUF_ERROR\n");
3371 + break;
3372 +
3373 + default:
3374 + printk(KERN_INFO DEVICE_NAME ": Bug detected!\n");
3375 + break;
3376 + }
3377 +
3378 + return -EIO;
3379 +}
3380 +
3381 +/* Allocate the XZ decoder state and register the character device. */
3382 +static int __init xz_dec_test_init(void)
3383 +{
3384 + static const struct file_operations fileops = {
3385 + .owner = THIS_MODULE,
3386 + .open = &xz_dec_test_open,
3387 + .release = &xz_dec_test_release,
3388 + .write = &xz_dec_test_write
3389 + };
3390 +
3391 + state = xz_dec_init(XZ_PREALLOC, DICT_MAX);
3392 + if (state == NULL)
3393 + return -ENOMEM;
3394 +
3395 + device_major = register_chrdev(0, DEVICE_NAME, &fileops);
3396 + if (device_major < 0) {
3397 + xz_dec_end(state);
3398 + return device_major;
3399 + }
3400 +
3401 + printk(KERN_INFO DEVICE_NAME ": module loaded\n");
3402 + printk(KERN_INFO DEVICE_NAME ": Create a device node with "
3403 + "'mknod " DEVICE_NAME " c %d 0' and write .xz files "
3404 + "to it.\n", device_major);
3405 + return 0;
3406 +}
3407 +
3408 +static void __exit xz_dec_test_exit(void)
3409 +{
3410 + unregister_chrdev(device_major, DEVICE_NAME);
3411 + xz_dec_end(state);
3412 + printk(KERN_INFO DEVICE_NAME ": module unloaded\n");
3413 +}
3414 +
3415 +module_init(xz_dec_test_init);
3416 +module_exit(xz_dec_test_exit);
3417 +
3418 +MODULE_DESCRIPTION("XZ decompressor tester");
3419 +MODULE_VERSION("1.0");
3420 +MODULE_AUTHOR("Lasse Collin <lasse.collin@tukaani.org>");
3421 +
3422 +/*
3423 + * This code is in the public domain, but in Linux it's simplest to just
3424 + * say it's GPL and consider the authors as the copyright holders.
3425 + */
3426 +MODULE_LICENSE("GPL");
3427 diff --git a/lib/xz/xz_lzma2.h b/lib/xz/xz_lzma2.h
3428 new file mode 100644
3429 index 0000000..071d67b
3430 --- /dev/null
3431 +++ b/lib/xz/xz_lzma2.h
3432 @@ -0,0 +1,204 @@
3433 +/*
3434 + * LZMA2 definitions
3435 + *
3436 + * Authors: Lasse Collin <lasse.collin@tukaani.org>
3437 + * Igor Pavlov <http://7-zip.org/>
3438 + *
3439 + * This file has been put into the public domain.
3440 + * You can do whatever you want with this file.
3441 + */
3442 +
3443 +#ifndef XZ_LZMA2_H
3444 +#define XZ_LZMA2_H
3445 +
3446 +/* Range coder constants */
3447 +#define RC_SHIFT_BITS 8
3448 +#define RC_TOP_BITS 24
3449 +#define RC_TOP_VALUE (1 << RC_TOP_BITS)
3450 +#define RC_BIT_MODEL_TOTAL_BITS 11
3451 +#define RC_BIT_MODEL_TOTAL (1 << RC_BIT_MODEL_TOTAL_BITS)
3452 +#define RC_MOVE_BITS 5
3453 +
3454 +/*
3455 + * Maximum number of position states. A position state is the lowest pb
3456 + * number of bits of the current uncompressed offset. In some places there
3457 + * are different sets of probabilities for different position states.
3458 + */
3459 +#define POS_STATES_MAX (1 << 4)
3460 +
3461 +/*
3462 + * This enum is used to track which LZMA symbols have occurred most recently
3463 + * and in which order. This information is used to predict the next symbol.
3464 + *
3465 + * Symbols:
3466 + * - Literal: One 8-bit byte
3467 + * - Match: Repeat a chunk of data at some distance
3468 + * - Long repeat: Multi-byte match at a recently seen distance
3469 + * - Short repeat: One-byte repeat at a recently seen distance
3470 + *
3471 + * The symbol names are in from STATE_oldest_older_previous. REP means
3472 + * either short or long repeated match, and NONLIT means any non-literal.
3473 + */
3474 +enum lzma_state {
3475 + STATE_LIT_LIT,
3476 + STATE_MATCH_LIT_LIT,
3477 + STATE_REP_LIT_LIT,
3478 + STATE_SHORTREP_LIT_LIT,
3479 + STATE_MATCH_LIT,
3480 + STATE_REP_LIT,
3481 + STATE_SHORTREP_LIT,
3482 + STATE_LIT_MATCH,
3483 + STATE_LIT_LONGREP,
3484 + STATE_LIT_SHORTREP,
3485 + STATE_NONLIT_MATCH,
3486 + STATE_NONLIT_REP
3487 +};
3488 +
3489 +/* Total number of states */
3490 +#define STATES 12
3491 +
3492 +/* The lowest 7 states indicate that the previous state was a literal. */
3493 +#define LIT_STATES 7
3494 +
3495 +/* Indicate that the latest symbol was a literal. */
3496 +static inline void lzma_state_literal(enum lzma_state *state)
3497 +{
3498 + if (*state <= STATE_SHORTREP_LIT_LIT)
3499 + *state = STATE_LIT_LIT;
3500 + else if (*state <= STATE_LIT_SHORTREP)
3501 + *state -= 3;
3502 + else
3503 + *state -= 6;
3504 +}
3505 +
3506 +/* Indicate that the latest symbol was a match. */
3507 +static inline void lzma_state_match(enum lzma_state *state)
3508 +{
3509 + *state = *state < LIT_STATES ? STATE_LIT_MATCH : STATE_NONLIT_MATCH;
3510 +}
3511 +
3512 +/* Indicate that the latest state was a long repeated match. */
3513 +static inline void lzma_state_long_rep(enum lzma_state *state)
3514 +{
3515 + *state = *state < LIT_STATES ? STATE_LIT_LONGREP : STATE_NONLIT_REP;
3516 +}
3517 +
3518 +/* Indicate that the latest symbol was a short match. */
3519 +static inline void lzma_state_short_rep(enum lzma_state *state)
3520 +{
3521 + *state = *state < LIT_STATES ? STATE_LIT_SHORTREP : STATE_NONLIT_REP;
3522 +}
3523 +
3524 +/* Test if the previous symbol was a literal. */
3525 +static inline bool lzma_state_is_literal(enum lzma_state state)
3526 +{
3527 + return state < LIT_STATES;
3528 +}
3529 +
3530 +/* Each literal coder is divided in three sections:
3531 + * - 0x001-0x0FF: Without match byte
3532 + * - 0x101-0x1FF: With match byte; match bit is 0
3533 + * - 0x201-0x2FF: With match byte; match bit is 1
3534 + *
3535 + * Match byte is used when the previous LZMA symbol was something else than
3536 + * a literal (that is, it was some kind of match).
3537 + */
3538 +#define LITERAL_CODER_SIZE 0x300
3539 +
3540 +/* Maximum number of literal coders */
3541 +#define LITERAL_CODERS_MAX (1 << 4)
3542 +
3543 +/* Minimum length of a match is two bytes. */
3544 +#define MATCH_LEN_MIN 2
3545 +
3546 +/* Match length is encoded with 4, 5, or 10 bits.
3547 + *
3548 + * Length Bits
3549 + * 2-9 4 = Choice=0 + 3 bits
3550 + * 10-17 5 = Choice=1 + Choice2=0 + 3 bits
3551 + * 18-273 10 = Choice=1 + Choice2=1 + 8 bits
3552 + */
3553 +#define LEN_LOW_BITS 3
3554 +#define LEN_LOW_SYMBOLS (1 << LEN_LOW_BITS)
3555 +#define LEN_MID_BITS 3
3556 +#define LEN_MID_SYMBOLS (1 << LEN_MID_BITS)
3557 +#define LEN_HIGH_BITS 8
3558 +#define LEN_HIGH_SYMBOLS (1 << LEN_HIGH_BITS)
3559 +#define LEN_SYMBOLS (LEN_LOW_SYMBOLS + LEN_MID_SYMBOLS + LEN_HIGH_SYMBOLS)
3560 +
3561 +/*
3562 + * Maximum length of a match is 273 which is a result of the encoding
3563 + * described above.
3564 + */
3565 +#define MATCH_LEN_MAX (MATCH_LEN_MIN + LEN_SYMBOLS - 1)
3566 +
3567 +/*
3568 + * Different sets of probabilities are used for match distances that have
3569 + * very short match length: Lengths of 2, 3, and 4 bytes have a separate
3570 + * set of probabilities for each length. The matches with longer length
3571 + * use a shared set of probabilities.
3572 + */
3573 +#define DIST_STATES 4
3574 +
3575 +/*
3576 + * Get the index of the appropriate probability array for decoding
3577 + * the distance slot.
3578 + */
3579 +static inline uint32_t lzma_get_dist_state(uint32_t len)
3580 +{
3581 + return len < DIST_STATES + MATCH_LEN_MIN
3582 + ? len - MATCH_LEN_MIN : DIST_STATES - 1;
3583 +}
3584 +
3585 +/*
3586 + * The highest two bits of a 32-bit match distance are encoded using six bits.
3587 + * This six-bit value is called a distance slot. This way encoding a 32-bit
3588 + * value takes 6-36 bits, larger values taking more bits.
3589 + */
3590 +#define DIST_SLOT_BITS 6
3591 +#define DIST_SLOTS (1 << DIST_SLOT_BITS)
3592 +
3593 +/* Match distances up to 127 are fully encoded using probabilities. Since
3594 + * the highest two bits (distance slot) are always encoded using six bits,
3595 + * the distances 0-3 don't need any additional bits to encode, since the
3596 + * distance slot itself is the same as the actual distance. DIST_MODEL_START
3597 + * indicates the first distance slot where at least one additional bit is
3598 + * needed.
3599 + */
3600 +#define DIST_MODEL_START 4
3601 +
3602 +/*
3603 + * Match distances greater than 127 are encoded in three pieces:
3604 + * - distance slot: the highest two bits
3605 + * - direct bits: 2-26 bits below the highest two bits
3606 + * - alignment bits: four lowest bits
3607 + *
3608 + * Direct bits don't use any probabilities.
3609 + *
3610 + * The distance slot value of 14 is for distances 128-191.
3611 + */
3612 +#define DIST_MODEL_END 14
3613 +
3614 +/* Distance slots that indicate a distance <= 127. */
3615 +#define FULL_DISTANCES_BITS (DIST_MODEL_END / 2)
3616 +#define FULL_DISTANCES (1 << FULL_DISTANCES_BITS)
3617 +
3618 +/*
3619 + * For match distances greater than 127, only the highest two bits and the
3620 + * lowest four bits (alignment) is encoded using probabilities.
3621 + */
3622 +#define ALIGN_BITS 4
3623 +#define ALIGN_SIZE (1 << ALIGN_BITS)
3624 +#define ALIGN_MASK (ALIGN_SIZE - 1)
3625 +
3626 +/* Total number of all probability variables */
3627 +#define PROBS_TOTAL (1846 + LITERAL_CODERS_MAX * LITERAL_CODER_SIZE)
3628 +
3629 +/*
3630 + * LZMA remembers the four most recent match distances. Reusing these
3631 + * distances tends to take less space than re-encoding the actual
3632 + * distance value.
3633 + */
3634 +#define REPS 4
3635 +
3636 +#endif
3637 diff --git a/lib/xz/xz_private.h b/lib/xz/xz_private.h
3638 new file mode 100644
3639 index 0000000..a65633e
3640 --- /dev/null
3641 +++ b/lib/xz/xz_private.h
3642 @@ -0,0 +1,156 @@
3643 +/*
3644 + * Private includes and definitions
3645 + *
3646 + * Author: Lasse Collin <lasse.collin@tukaani.org>
3647 + *
3648 + * This file has been put into the public domain.
3649 + * You can do whatever you want with this file.
3650 + */
3651 +
3652 +#ifndef XZ_PRIVATE_H
3653 +#define XZ_PRIVATE_H
3654 +
3655 +#ifdef __KERNEL__
3656 +# include <linux/xz.h>
3657 +# include <asm/byteorder.h>
3658 +# include <asm/unaligned.h>
3659 + /* XZ_PREBOOT may be defined only via decompress_unxz.c. */
3660 +# ifndef XZ_PREBOOT
3661 +# include <linux/slab.h>
3662 +# include <linux/vmalloc.h>
3663 +# include <linux/string.h>
3664 +# ifdef CONFIG_XZ_DEC_X86
3665 +# define XZ_DEC_X86
3666 +# endif
3667 +# ifdef CONFIG_XZ_DEC_POWERPC
3668 +# define XZ_DEC_POWERPC
3669 +# endif
3670 +# ifdef CONFIG_XZ_DEC_IA64
3671 +# define XZ_DEC_IA64
3672 +# endif
3673 +# ifdef CONFIG_XZ_DEC_ARM
3674 +# define XZ_DEC_ARM
3675 +# endif
3676 +# ifdef CONFIG_XZ_DEC_ARMTHUMB
3677 +# define XZ_DEC_ARMTHUMB
3678 +# endif
3679 +# ifdef CONFIG_XZ_DEC_SPARC
3680 +# define XZ_DEC_SPARC
3681 +# endif
3682 +# define memeq(a, b, size) (memcmp(a, b, size) == 0)
3683 +# define memzero(buf, size) memset(buf, 0, size)
3684 +# endif
3685 +# define get_le32(p) le32_to_cpup((const uint32_t *)(p))
3686 +#else
3687 + /*
3688 + * For userspace builds, use a separate header to define the required
3689 + * macros and functions. This makes it easier to adapt the code into
3690 + * different environments and avoids clutter in the Linux kernel tree.
3691 + */
3692 +# include "xz_config.h"
3693 +#endif
3694 +
3695 +/* If no specific decoding mode is requested, enable support for all modes. */
3696 +#if !defined(XZ_DEC_SINGLE) && !defined(XZ_DEC_PREALLOC) \
3697 + && !defined(XZ_DEC_DYNALLOC)
3698 +# define XZ_DEC_SINGLE
3699 +# define XZ_DEC_PREALLOC
3700 +# define XZ_DEC_DYNALLOC
3701 +#endif
3702 +
3703 +/*
3704 + * The DEC_IS_foo(mode) macros are used in "if" statements. If only some
3705 + * of the supported modes are enabled, these macros will evaluate to true or
3706 + * false at compile time and thus allow the compiler to omit unneeded code.
3707 + */
3708 +#ifdef XZ_DEC_SINGLE
3709 +# define DEC_IS_SINGLE(mode) ((mode) == XZ_SINGLE)
3710 +#else
3711 +# define DEC_IS_SINGLE(mode) (false)
3712 +#endif
3713 +
3714 +#ifdef XZ_DEC_PREALLOC
3715 +# define DEC_IS_PREALLOC(mode) ((mode) == XZ_PREALLOC)
3716 +#else
3717 +# define DEC_IS_PREALLOC(mode) (false)
3718 +#endif
3719 +
3720 +#ifdef XZ_DEC_DYNALLOC
3721 +# define DEC_IS_DYNALLOC(mode) ((mode) == XZ_DYNALLOC)
3722 +#else
3723 +# define DEC_IS_DYNALLOC(mode) (false)
3724 +#endif
3725 +
3726 +#if !defined(XZ_DEC_SINGLE)
3727 +# define DEC_IS_MULTI(mode) (true)
3728 +#elif defined(XZ_DEC_PREALLOC) || defined(XZ_DEC_DYNALLOC)
3729 +# define DEC_IS_MULTI(mode) ((mode) != XZ_SINGLE)
3730 +#else
3731 +# define DEC_IS_MULTI(mode) (false)
3732 +#endif
3733 +
3734 +/*
3735 + * If any of the BCJ filter decoders are wanted, define XZ_DEC_BCJ.
3736 + * XZ_DEC_BCJ is used to enable generic support for BCJ decoders.
3737 + */
3738 +#ifndef XZ_DEC_BCJ
3739 +# if defined(XZ_DEC_X86) || defined(XZ_DEC_POWERPC) \
3740 + || defined(XZ_DEC_IA64) || defined(XZ_DEC_ARM) \
3741 + || defined(XZ_DEC_ARM) || defined(XZ_DEC_ARMTHUMB) \
3742 + || defined(XZ_DEC_SPARC)
3743 +# define XZ_DEC_BCJ
3744 +# endif
3745 +#endif
3746 +
3747 +/*
3748 + * Allocate memory for LZMA2 decoder. xz_dec_lzma2_reset() must be used
3749 + * before calling xz_dec_lzma2_run().
3750 + */
3751 +XZ_EXTERN struct xz_dec_lzma2 *xz_dec_lzma2_create(enum xz_mode mode,
3752 + uint32_t dict_max);
3753 +
3754 +/*
3755 + * Decode the LZMA2 properties (one byte) and reset the decoder. Return
3756 + * XZ_OK on success, XZ_MEMLIMIT_ERROR if the preallocated dictionary is not
3757 + * big enough, and XZ_OPTIONS_ERROR if props indicates something that this
3758 + * decoder doesn't support.
3759 + */
3760 +XZ_EXTERN enum xz_ret xz_dec_lzma2_reset(struct xz_dec_lzma2 *s,
3761 + uint8_t props);
3762 +
3763 +/* Decode raw LZMA2 stream from b->in to b->out. */
3764 +XZ_EXTERN enum xz_ret xz_dec_lzma2_run(struct xz_dec_lzma2 *s,
3765 + struct xz_buf *b);
3766 +
3767 +/* Free the memory allocated for the LZMA2 decoder. */
3768 +XZ_EXTERN void xz_dec_lzma2_end(struct xz_dec_lzma2 *s);
3769 +
3770 +#ifdef XZ_DEC_BCJ
3771 +/*
3772 + * Allocate memory for BCJ decoders. xz_dec_bcj_reset() must be used before
3773 + * calling xz_dec_bcj_run().
3774 + */
3775 +XZ_EXTERN struct xz_dec_bcj *xz_dec_bcj_create(bool single_call);
3776 +
3777 +/*
3778 + * Decode the Filter ID of a BCJ filter. This implementation doesn't
3779 + * support custom start offsets, so no decoding of Filter Properties
3780 + * is needed. Returns XZ_OK if the given Filter ID is supported.
3781 + * Otherwise XZ_OPTIONS_ERROR is returned.
3782 + */
3783 +XZ_EXTERN enum xz_ret xz_dec_bcj_reset(struct xz_dec_bcj *s, uint8_t id);
3784 +
3785 +/*
3786 + * Decode raw BCJ + LZMA2 stream. This must be used only if there actually is
3787 + * a BCJ filter in the chain. If the chain has only LZMA2, xz_dec_lzma2_run()
3788 + * must be called directly.
3789 + */
3790 +XZ_EXTERN enum xz_ret xz_dec_bcj_run(struct xz_dec_bcj *s,
3791 + struct xz_dec_lzma2 *lzma2,
3792 + struct xz_buf *b);
3793 +
3794 +/* Free the memory allocated for the BCJ filters. */
3795 +#define xz_dec_bcj_end(s) kfree(s)
3796 +#endif
3797 +
3798 +#endif
3799 diff --git a/lib/xz/xz_stream.h b/lib/xz/xz_stream.h
3800 new file mode 100644
3801 index 0000000..66cb5a7
3802 --- /dev/null
3803 +++ b/lib/xz/xz_stream.h
3804 @@ -0,0 +1,62 @@
3805 +/*
3806 + * Definitions for handling the .xz file format
3807 + *
3808 + * Author: Lasse Collin <lasse.collin@tukaani.org>
3809 + *
3810 + * This file has been put into the public domain.
3811 + * You can do whatever you want with this file.
3812 + */
3813 +
3814 +#ifndef XZ_STREAM_H
3815 +#define XZ_STREAM_H
3816 +
3817 +#if defined(__KERNEL__) && !XZ_INTERNAL_CRC32
3818 +# include <linux/crc32.h>
3819 +# undef crc32
3820 +# define xz_crc32(buf, size, crc) \
3821 + (~crc32_le(~(uint32_t)(crc), buf, size))
3822 +#endif
3823 +
3824 +/*
3825 + * See the .xz file format specification at
3826 + * http://tukaani.org/xz/xz-file-format.txt
3827 + * to understand the container format.
3828 + */
3829 +
3830 +#define STREAM_HEADER_SIZE 12
3831 +
3832 +#define HEADER_MAGIC "\3757zXZ"
3833 +#define HEADER_MAGIC_SIZE 6
3834 +
3835 +#define FOOTER_MAGIC "YZ"
3836 +#define FOOTER_MAGIC_SIZE 2
3837 +
3838 +/*
3839 + * Variable-length integer can hold a 63-bit unsigned integer or a special
3840 + * value indicating that the value is unknown.
3841 + *
3842 + * Experimental: vli_type can be defined to uint32_t to save a few bytes
3843 + * in code size (no effect on speed). Doing so limits the uncompressed and
3844 + * compressed size of the file to less than 256 MiB and may also weaken
3845 + * error detection slightly.
3846 + */
3847 +typedef uint64_t vli_type;
3848 +
3849 +#define VLI_MAX ((vli_type)-1 / 2)
3850 +#define VLI_UNKNOWN ((vli_type)-1)
3851 +
3852 +/* Maximum encoded size of a VLI */
3853 +#define VLI_BYTES_MAX (sizeof(vli_type) * 8 / 7)
3854 +
3855 +/* Integrity Check types */
3856 +enum xz_check {
3857 + XZ_CHECK_NONE = 0,
3858 + XZ_CHECK_CRC32 = 1,
3859 + XZ_CHECK_CRC64 = 4,
3860 + XZ_CHECK_SHA256 = 10
3861 +};
3862 +
3863 +/* Maximum possible Check ID */
3864 +#define XZ_CHECK_MAX 15
3865 +
3866 +#endif
3867 diff --git a/scripts/Makefile.lib b/scripts/Makefile.lib
3868 index 54fd1b7..b862007 100644
3869 --- a/scripts/Makefile.lib
3870 +++ b/scripts/Makefile.lib
3871 @@ -246,6 +246,34 @@ cmd_lzo = (cat $(filter-out FORCE,$^) | \
3872 lzop -9 && $(call size_append, $(filter-out FORCE,$^))) > $@ || \
3873 (rm -f $@ ; false)
3875 +# XZ
3876 +# ---------------------------------------------------------------------------
3877 +# Use xzkern to compress the kernel image and xzmisc to compress other things.
3878 +#
3879 +# xzkern uses a big LZMA2 dictionary since it doesn't increase memory usage
3880 +# of the kernel decompressor. A BCJ filter is used if it is available for
3881 +# the target architecture. xzkern also appends uncompressed size of the data
3882 +# using size_append. The .xz format has the size information available at
3883 +# the end of the file too, but it's in more complex format and it's good to
3884 +# avoid changing the part of the boot code that reads the uncompressed size.
3885 +# Note that the bytes added by size_append will make the xz tool think that
3886 +# the file is corrupt. This is expected.
3887 +#
3888 +# xzmisc doesn't use size_append, so it can be used to create normal .xz
3889 +# files. xzmisc uses smaller LZMA2 dictionary than xzkern, because a very
3890 +# big dictionary would increase the memory usage too much in the multi-call
3891 +# decompression mode. A BCJ filter isn't used either.
3892 +quiet_cmd_xzkern = XZKERN $@
3893 +cmd_xzkern = (cat $(filter-out FORCE,$^) | \
3894 + sh $(srctree)/scripts/xz_wrap.sh && \
3895 + $(call size_append, $(filter-out FORCE,$^))) > $@ || \
3896 + (rm -f $@ ; false)
3897 +
3898 +quiet_cmd_xzmisc = XZMISC $@
3899 +cmd_xzmisc = (cat $(filter-out FORCE,$^) | \
3900 + xz --check=crc32 --lzma2=dict=1MiB) > $@ || \
3901 + (rm -f $@ ; false)
3902 +
3903 # misc stuff
3904 # ---------------------------------------------------------------------------
3905 quote:="
3906 diff --git a/scripts/xz_wrap.sh b/scripts/xz_wrap.sh
3907 new file mode 100644
3908 index 0000000..17a5798
3909 --- /dev/null
3910 +++ b/scripts/xz_wrap.sh
3911 @@ -0,0 +1,23 @@
3912 +#!/bin/sh
3913 +#
3914 +# This is a wrapper for xz to compress the kernel image using appropriate
3915 +# compression options depending on the architecture.
3916 +#
3917 +# Author: Lasse Collin <lasse.collin@tukaani.org>
3918 +#
3919 +# This file has been put into the public domain.
3920 +# You can do whatever you want with this file.
3921 +#
3922 +
3923 +BCJ=
3924 +LZMA2OPTS=
3925 +
3926 +case $ARCH in
3927 + x86|x86_64) BCJ=--x86 ;;
3928 + powerpc) BCJ=--powerpc ;;
3929 + ia64) BCJ=--ia64; LZMA2OPTS=pb=4 ;;
3930 + arm) BCJ=--arm ;;
3931 + sparc) BCJ=--sparc ;;
3932 +esac
3933 +
3934 +exec xz --check=crc32 $BCJ --lzma2=$LZMA2OPTS,dict=32MiB