wok-6.x rev 7650

Add squashfs+xz to kernel. This allows for greater compressing then lzma and it will work without cause a kernel opps. As long as the custom code for freeing lzma initramfs stays with lzma initramfs.
author Christopher Rogers <slaxemulator@gmail.com>
date Tue Dec 14 21:45:09 2010 +0000 (2010-12-14)
parents 13e4d4e6fcc1
children 5e64ecd01686
files linux/receipt linux/stuff/001-squashfs-decompressors-add-xz-decompressor-module.patch linux/stuff/002-squashfs-decompressors-add-boot-time-xz-support.patch linux/stuff/003-squashfs-x86-support-xz-compressed-kernel.patch linux/stuff/004-squashfs-add-xz-compression-support.patch linux/stuff/005-squashfs-add-xz-compression-configuration-option.patch linux/stuff/linux-2.6.36-slitaz.config linux/stuff/modules-2.6.36.list
line diff
     1.1 --- a/linux/receipt	Tue Dec 14 21:34:46 2010 +0000
     1.2 +++ b/linux/receipt	Tue Dec 14 21:45:09 2010 +0000
     1.3 @@ -26,13 +26,13 @@
     1.4  	TARBALL=$SOURCES_REPOSITORY/$AUFSDIR.tar.gz
     1.5  	if [ -f $TARBALL ]; then
     1.6  		tar xzf $TARBALL
     1.7 -		cd $AUFSDIR && git checkout origin/aufs2
     1.8 +		cd $AUFSDIR && git checkout origin/aufs2.1-36
     1.9  		cd $WOK/$PACKAGE
    1.10  	else
    1.11  		# Aufs2 from git repository
    1.12  		git clone http://git.c3sl.ufpr.br/pub/scm/aufs/aufs2-standalone.git $AUFSDIR
    1.13  		tar czf $TARBALL $AUFSDIR
    1.14 -		cd $AUFSDIR && git checkout origin/aufs2
    1.15 +		cd $AUFSDIR && git checkout origin/aufs2.1-36
    1.16  		cd $WOK/$PACKAGE
    1.17  	fi
    1.18  	cp -a $AUFSDIR/Documentation $AUFSDIR/fs $AUFSDIR/include $src
    1.19 @@ -64,8 +64,11 @@
    1.20  $PACKAGE-freeinitrd-$VERSION.u
    1.21  aufs2-base.patch
    1.22  aufs2-standalone.patch
    1.23 -aufs2-module-2.6.36.patch
    1.24 -aufs2-2.6.36-fix.patch
    1.25 +001-squashfs-decompressors-add-xz-decompressor-module.patch
    1.26 +002-squashfs-decompressors-add-boot-time-xz-support.patch
    1.27 +003-squashfs-x86-support-xz-compressed-kernel.patch
    1.28 +004-squashfs-add-xz-compression-support.patch
    1.29 +005-squashfs-add-xz-compression-configuration-option.patch
    1.30  EOT
    1.31  	make mrproper
    1.32  	cd Documentation/lguest
     2.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     2.2 +++ b/linux/stuff/001-squashfs-decompressors-add-xz-decompressor-module.patch	Tue Dec 14 21:45:09 2010 +0000
     2.3 @@ -0,0 +1,3934 @@
     2.4 +From: Lasse Collin <lasse.collin@tukaani.org>
     2.5 +Date: Thu, 2 Dec 2010 19:14:19 +0000 (+0200)
     2.6 +Subject: Decompressors: Add XZ decompressor module
     2.7 +X-Git-Url: http://git.kernel.org/?p=linux%2Fkernel%2Fgit%2Fpkl%2Fsquashfs-xz.git;a=commitdiff_plain;h=3dbc3fe7878e53b43064a12d4ab31ca4c18ce85f
     2.8 +
     2.9 +Decompressors: Add XZ decompressor module
    2.10 +
    2.11 +In userspace, the .lzma format has become mostly a legacy
    2.12 +file format that got superseded by the .xz format. Similarly,
    2.13 +LZMA Utils was superseded by XZ Utils.
    2.14 +
    2.15 +These patches add support for XZ decompression into
    2.16 +the kernel. Most of the code is as is from XZ Embedded
    2.17 +<http://tukaani.org/xz/embedded.html>. It was written for
    2.18 +the Linux kernel but is usable in other projects too.
    2.19 +
    2.20 +Advantages of XZ over the current LZMA code in the kernel:
    2.21 +  - Nice API that can be used by other kernel modules; it's
    2.22 +    not limited to kernel, initramfs, and initrd decompression.
    2.23 +  - Integrity check support (CRC32)
    2.24 +  - BCJ filters improve compression of executable code on
    2.25 +    certain architectures. These together with LZMA2 can
    2.26 +    produce a few percent smaller kernel or Squashfs images
    2.27 +    than plain LZMA without making the decompression slower.
    2.28 +
    2.29 +This patch: Add the main decompression code (xz_dec), testing
    2.30 +module (xz_dec_test), wrapper script (xz_wrap.sh) for the xz
    2.31 +command line tool, and documentation. The xz_dec module is
    2.32 +enough to have a usable XZ decompressor e.g. for Squashfs.
    2.33 +
    2.34 +Signed-off-by: Lasse Collin <lasse.collin@tukaani.org>
    2.35 +---
    2.36 +
    2.37 +diff --git a/Documentation/xz.txt b/Documentation/xz.txt
    2.38 +new file mode 100644
    2.39 +index 0000000..68329ac
    2.40 +--- /dev/null
    2.41 ++++ b/Documentation/xz.txt
    2.42 +@@ -0,0 +1,122 @@
    2.43 ++
    2.44 ++XZ data compression in Linux
    2.45 ++============================
    2.46 ++
    2.47 ++Introduction
    2.48 ++
    2.49 ++    XZ is a general purpose data compression format with high compression
    2.50 ++    ratio and relatively fast decompression. The primary compression
    2.51 ++    algorithm (filter) is LZMA2. Additional filters can be used to improve
    2.52 ++    compression ratio even further. E.g. Branch/Call/Jump (BCJ) filters
    2.53 ++    improve compression ratio of executable data.
    2.54 ++
    2.55 ++    The XZ decompressor in Linux is called XZ Embedded. It supports
    2.56 ++    the LZMA2 filter and optionally also BCJ filters. CRC32 is supported
    2.57 ++    for integrity checking. The home page of XZ Embedded is at
    2.58 ++    <http://tukaani.org/xz/embedded.html>, where you can find the
    2.59 ++    latest version and also information about using the code outside
    2.60 ++    the Linux kernel.
    2.61 ++
    2.62 ++    For userspace, XZ Utils provide a zlib-like compression library
    2.63 ++    and a gzip-like command line tool. XZ Utils can be downloaded from
    2.64 ++    <http://tukaani.org/xz/>.
    2.65 ++
    2.66 ++XZ related components in the kernel
    2.67 ++
    2.68 ++    The xz_dec module provides XZ decompressor with single-call (buffer
    2.69 ++    to buffer) and multi-call (stateful) APIs. The usage of the xz_dec
    2.70 ++    module is documented in include/linux/xz.h.
    2.71 ++
    2.72 ++    The xz_dec_test module is for testing xz_dec. xz_dec_test is not
    2.73 ++    useful unless you are hacking the XZ decompressor. xz_dec_test
    2.74 ++    allocates a char device major dynamically to which one can write
    2.75 ++    .xz files from userspace. The decompressed output is thrown away.
    2.76 ++    Keep an eye on dmesg to see diagnostics printed by xz_dec_test.
    2.77 ++    See the xz_dec_test source code for the details.
    2.78 ++
    2.79 ++    For decompressing the kernel image, initramfs, and initrd, there
    2.80 ++    is a wrapper function in lib/decompress_unxz.c. Its API is the
    2.81 ++    same as in other decompress_*.c files, which is defined in
    2.82 ++    include/linux/decompress/generic.h.
    2.83 ++
    2.84 ++    scripts/xz_wrap.sh is a wrapper for the xz command line tool found
    2.85 ++    from XZ Utils. The wrapper sets compression options to values suitable
    2.86 ++    for compressing the kernel image.
    2.87 ++
    2.88 ++    For kernel makefiles, two commands are provided for use with
    2.89 ++    $(call if_needed). The kernel image should be compressed with
    2.90 ++    $(call if_needed,xzkern) which will use a BCJ filter and a big LZMA2
    2.91 ++    dictionary. It will also append a four-byte trailer containing the
    2.92 ++    uncompressed size of the file, which is needed by the boot code.
    2.93 ++    Other things should be compressed with $(call if_needed,xzmisc)
    2.94 ++    which will use no BCJ filter and 1 MiB LZMA2 dictionary.
    2.95 ++
    2.96 ++Notes on compression options
    2.97 ++
    2.98 ++    Since the XZ Embedded supports only streams with no integrity check or
    2.99 ++    CRC32, make sure that you don't use some other integrity check type
   2.100 ++    when encoding files that are supposed to be decoded by the kernel. With
   2.101 ++    liblzma, you need to use either LZMA_CHECK_NONE or LZMA_CHECK_CRC32
   2.102 ++    when encoding. With the xz command line tool, use --check=none or
   2.103 ++    --check=crc32.
   2.104 ++
   2.105 ++    Using CRC32 is strongly recommended unless there is some other layer
   2.106 ++    which will verify the integrity of the uncompressed data anyway.
   2.107 ++    Double checking the integrity would probably be waste of CPU cycles.
   2.108 ++    Note that the headers will always have a CRC32 which will be validated
   2.109 ++    by the decoder; you can only change the integrity check type (or
   2.110 ++    disable it) for the actual uncompressed data.
   2.111 ++
   2.112 ++    In userspace, LZMA2 is typically used with dictionary sizes of several
   2.113 ++    megabytes. The decoder needs to have the dictionary in RAM, thus big
   2.114 ++    dictionaries cannot be used for files that are intended to be decoded
   2.115 ++    by the kernel. 1 MiB is probably the maximum reasonable dictionary
   2.116 ++    size for in-kernel use (maybe more is OK for initramfs). The presets
   2.117 ++    in XZ Utils may not be optimal when creating files for the kernel,
   2.118 ++    so don't hesitate to use custom settings. Example:
   2.119 ++
   2.120 ++        xz --check=crc32 --lzma2=dict=512KiB inputfile
   2.121 ++
   2.122 ++    An exception to above dictionary size limitation is when the decoder
   2.123 ++    is used in single-call mode. Decompressing the kernel itself is an
   2.124 ++    example of this situation. In single-call mode, the memory usage
   2.125 ++    doesn't depend on the dictionary size, and it is perfectly fine to
   2.126 ++    use a big dictionary: for maximum compression, the dictionary should
   2.127 ++    be at least as big as the uncompressed data itself.
   2.128 ++
   2.129 ++Future plans
   2.130 ++
   2.131 ++    Creating a limited XZ encoder may be considered if people think it is
   2.132 ++    useful. LZMA2 is slower to compress than e.g. Deflate or LZO even at
   2.133 ++    the fastest settings, so it isn't clear if LZMA2 encoder is wanted
   2.134 ++    into the kernel.
   2.135 ++
   2.136 ++    Support for limited random-access reading is planned for the
   2.137 ++    decompression code. I don't know if it could have any use in the
   2.138 ++    kernel, but I know that it would be useful in some embedded projects
   2.139 ++    outside the Linux kernel.
   2.140 ++
   2.141 ++Conformance to the .xz file format specification
   2.142 ++
   2.143 ++    There are a couple of corner cases where things have been simplified
   2.144 ++    at expense of detecting errors as early as possible. These should not
   2.145 ++    matter in practice all, since they don't cause security issues. But
   2.146 ++    it is good to know this if testing the code e.g. with the test files
   2.147 ++    from XZ Utils.
   2.148 ++
   2.149 ++Reporting bugs
   2.150 ++
   2.151 ++    Before reporting a bug, please check that it's not fixed already
   2.152 ++    at upstream. See <http://tukaani.org/xz/embedded.html> to get the
   2.153 ++    latest code.
   2.154 ++
   2.155 ++    Report bugs to <lasse.collin@tukaani.org> or visit #tukaani on
   2.156 ++    Freenode and talk to Larhzu. I don't actively read LKML or other
   2.157 ++    kernel-related mailing lists, so if there's something I should know,
   2.158 ++    you should email to me personally or use IRC.
   2.159 ++
   2.160 ++    Don't bother Igor Pavlov with questions about the XZ implementation
   2.161 ++    in the kernel or about XZ Utils. While these two implementations
   2.162 ++    include essential code that is directly based on Igor Pavlov's code,
   2.163 ++    these implementations aren't maintained nor supported by him.
   2.164 ++
   2.165 +diff --git a/include/linux/xz.h b/include/linux/xz.h
   2.166 +new file mode 100644
   2.167 +index 0000000..64cffa6
   2.168 +--- /dev/null
   2.169 ++++ b/include/linux/xz.h
   2.170 +@@ -0,0 +1,264 @@
   2.171 ++/*
   2.172 ++ * XZ decompressor
   2.173 ++ *
   2.174 ++ * Authors: Lasse Collin <lasse.collin@tukaani.org>
   2.175 ++ *          Igor Pavlov <http://7-zip.org/>
   2.176 ++ *
   2.177 ++ * This file has been put into the public domain.
   2.178 ++ * You can do whatever you want with this file.
   2.179 ++ */
   2.180 ++
   2.181 ++#ifndef XZ_H
   2.182 ++#define XZ_H
   2.183 ++
   2.184 ++#ifdef __KERNEL__
   2.185 ++#	include <linux/stddef.h>
   2.186 ++#	include <linux/types.h>
   2.187 ++#else
   2.188 ++#	include <stddef.h>
   2.189 ++#	include <stdint.h>
   2.190 ++#endif
   2.191 ++
   2.192 ++/* In Linux, this is used to make extern functions static when needed. */
   2.193 ++#ifndef XZ_EXTERN
   2.194 ++#	define XZ_EXTERN extern
   2.195 ++#endif
   2.196 ++
   2.197 ++/**
   2.198 ++ * enum xz_mode - Operation mode
   2.199 ++ *
   2.200 ++ * @XZ_SINGLE:              Single-call mode. This uses less RAM than
   2.201 ++ *                          than multi-call modes, because the LZMA2
   2.202 ++ *                          dictionary doesn't need to be allocated as
   2.203 ++ *                          part of the decoder state. All required data
   2.204 ++ *                          structures are allocated at initialization,
   2.205 ++ *                          so xz_dec_run() cannot return XZ_MEM_ERROR.
   2.206 ++ * @XZ_PREALLOC:            Multi-call mode with preallocated LZMA2
   2.207 ++ *                          dictionary buffer. All data structures are
   2.208 ++ *                          allocated at initialization, so xz_dec_run()
   2.209 ++ *                          cannot return XZ_MEM_ERROR.
   2.210 ++ * @XZ_DYNALLOC:            Multi-call mode. The LZMA2 dictionary is
   2.211 ++ *                          allocated once the required size has been
   2.212 ++ *                          parsed from the stream headers. If the
   2.213 ++ *                          allocation fails, xz_dec_run() will return
   2.214 ++ *                          XZ_MEM_ERROR.
   2.215 ++ *
   2.216 ++ * It is possible to enable support only for a subset of the above
   2.217 ++ * modes at compile time by defining XZ_DEC_SINGLE, XZ_DEC_PREALLOC,
   2.218 ++ * or XZ_DEC_DYNALLOC. The xz_dec kernel module is always compiled
   2.219 ++ * with support for all operation modes, but the preboot code may
   2.220 ++ * be built with fewer features to minimize code size.
   2.221 ++ */
   2.222 ++enum xz_mode {
   2.223 ++	XZ_SINGLE,
   2.224 ++	XZ_PREALLOC,
   2.225 ++	XZ_DYNALLOC
   2.226 ++};
   2.227 ++
   2.228 ++/**
   2.229 ++ * enum xz_ret - Return codes
   2.230 ++ * @XZ_OK:                  Everything is OK so far. More input or more
   2.231 ++ *                          output space is required to continue. This
   2.232 ++ *                          return code is possible only in multi-call mode
   2.233 ++ *                          (XZ_PREALLOC or XZ_DYNALLOC).
   2.234 ++ * @XZ_STREAM_END:          Operation finished successfully.
   2.235 ++ * @XZ_UNSUPPORTED_CHECK:   Integrity check type is not supported. Decoding
   2.236 ++ *                          is still possible in multi-call mode by simply
   2.237 ++ *                          calling xz_dec_run() again.
   2.238 ++ *                          Note that this return value is used only if
   2.239 ++ *                          XZ_DEC_ANY_CHECK was defined at build time,
   2.240 ++ *                          which is not used in the kernel. Unsupported
   2.241 ++ *                          check types return XZ_OPTIONS_ERROR if
   2.242 ++ *                          XZ_DEC_ANY_CHECK was not defined at build time.
   2.243 ++ * @XZ_MEM_ERROR:           Allocating memory failed. This return code is
   2.244 ++ *                          possible only if the decoder was initialized
   2.245 ++ *                          with XZ_DYNALLOC. The amount of memory that was
   2.246 ++ *                          tried to be allocated was no more than the
   2.247 ++ *                          dict_max argument given to xz_dec_init().
   2.248 ++ * @XZ_MEMLIMIT_ERROR:      A bigger LZMA2 dictionary would be needed than
   2.249 ++ *                          allowed by the dict_max argument given to
   2.250 ++ *                          xz_dec_init(). This return value is possible
   2.251 ++ *                          only in multi-call mode (XZ_PREALLOC or
   2.252 ++ *                          XZ_DYNALLOC); the single-call mode (XZ_SINGLE)
   2.253 ++ *                          ignores the dict_max argument.
   2.254 ++ * @XZ_FORMAT_ERROR:        File format was not recognized (wrong magic
   2.255 ++ *                          bytes).
   2.256 ++ * @XZ_OPTIONS_ERROR:       This implementation doesn't support the requested
   2.257 ++ *                          compression options. In the decoder this means
   2.258 ++ *                          that the header CRC32 matches, but the header
   2.259 ++ *                          itself specifies something that we don't support.
   2.260 ++ * @XZ_DATA_ERROR:          Compressed data is corrupt.
   2.261 ++ * @XZ_BUF_ERROR:           Cannot make any progress. Details are slightly
   2.262 ++ *                          different between multi-call and single-call
   2.263 ++ *                          mode; more information below.
   2.264 ++ *
   2.265 ++ * In multi-call mode, XZ_BUF_ERROR is returned when two consecutive calls
   2.266 ++ * to XZ code cannot consume any input and cannot produce any new output.
   2.267 ++ * This happens when there is no new input available, or the output buffer
   2.268 ++ * is full while at least one output byte is still pending. Assuming your
   2.269 ++ * code is not buggy, you can get this error only when decoding a compressed
   2.270 ++ * stream that is truncated or otherwise corrupt.
   2.271 ++ *
   2.272 ++ * In single-call mode, XZ_BUF_ERROR is returned only when the output buffer
   2.273 ++ * is too small or the compressed input is corrupt in a way that makes the
   2.274 ++ * decoder produce more output than the caller expected. When it is
   2.275 ++ * (relatively) clear that the compressed input is truncated, XZ_DATA_ERROR
   2.276 ++ * is used instead of XZ_BUF_ERROR.
   2.277 ++ */
   2.278 ++enum xz_ret {
   2.279 ++	XZ_OK,
   2.280 ++	XZ_STREAM_END,
   2.281 ++	XZ_UNSUPPORTED_CHECK,
   2.282 ++	XZ_MEM_ERROR,
   2.283 ++	XZ_MEMLIMIT_ERROR,
   2.284 ++	XZ_FORMAT_ERROR,
   2.285 ++	XZ_OPTIONS_ERROR,
   2.286 ++	XZ_DATA_ERROR,
   2.287 ++	XZ_BUF_ERROR
   2.288 ++};
   2.289 ++
   2.290 ++/**
   2.291 ++ * struct xz_buf - Passing input and output buffers to XZ code
   2.292 ++ * @in:         Beginning of the input buffer. This may be NULL if and only
   2.293 ++ *              if in_pos is equal to in_size.
   2.294 ++ * @in_pos:     Current position in the input buffer. This must not exceed
   2.295 ++ *              in_size.
   2.296 ++ * @in_size:    Size of the input buffer
   2.297 ++ * @out:        Beginning of the output buffer. This may be NULL if and only
   2.298 ++ *              if out_pos is equal to out_size.
   2.299 ++ * @out_pos:    Current position in the output buffer. This must not exceed
   2.300 ++ *              out_size.
   2.301 ++ * @out_size:   Size of the output buffer
   2.302 ++ *
   2.303 ++ * Only the contents of the output buffer from out[out_pos] onward, and
   2.304 ++ * the variables in_pos and out_pos are modified by the XZ code.
   2.305 ++ */
   2.306 ++struct xz_buf {
   2.307 ++	const uint8_t *in;
   2.308 ++	size_t in_pos;
   2.309 ++	size_t in_size;
   2.310 ++
   2.311 ++	uint8_t *out;
   2.312 ++	size_t out_pos;
   2.313 ++	size_t out_size;
   2.314 ++};
   2.315 ++
   2.316 ++/**
   2.317 ++ * struct xz_dec - Opaque type to hold the XZ decoder state
   2.318 ++ */
   2.319 ++struct xz_dec;
   2.320 ++
   2.321 ++/**
   2.322 ++ * xz_dec_init() - Allocate and initialize a XZ decoder state
   2.323 ++ * @mode:       Operation mode
   2.324 ++ * @dict_max:   Maximum size of the LZMA2 dictionary (history buffer) for
   2.325 ++ *              multi-call decoding. This is ignored in single-call mode
   2.326 ++ *              (mode == XZ_SINGLE). LZMA2 dictionary is always 2^n bytes
   2.327 ++ *              or 2^n + 2^(n-1) bytes (the latter sizes are less common
   2.328 ++ *              in practice), so other values for dict_max don't make sense.
   2.329 ++ *              In the kernel, dictionary sizes of 64 KiB, 128 KiB, 256 KiB,
   2.330 ++ *              512 KiB, and 1 MiB are probably the only reasonable values,
   2.331 ++ *              except for kernel and initramfs images where a bigger
   2.332 ++ *              dictionary can be fine and useful.
   2.333 ++ *
   2.334 ++ * Single-call mode (XZ_SINGLE): xz_dec_run() decodes the whole stream at
   2.335 ++ * once. The caller must provide enough output space or the decoding will
   2.336 ++ * fail. The output space is used as the dictionary buffer, which is why
   2.337 ++ * there is no need to allocate the dictionary as part of the decoder's
   2.338 ++ * internal state.
   2.339 ++ *
   2.340 ++ * Because the output buffer is used as the workspace, streams encoded using
   2.341 ++ * a big dictionary are not a problem in single-call mode. It is enough that
   2.342 ++ * the output buffer is big enough to hold the actual uncompressed data; it
   2.343 ++ * can be smaller than the dictionary size stored in the stream headers.
   2.344 ++ *
   2.345 ++ * Multi-call mode with preallocated dictionary (XZ_PREALLOC): dict_max bytes
   2.346 ++ * of memory is preallocated for the LZMA2 dictionary. This way there is no
   2.347 ++ * risk that xz_dec_run() could run out of memory, since xz_dec_run() will
   2.348 ++ * never allocate any memory. Instead, if the preallocated dictionary is too
   2.349 ++ * small for decoding the given input stream, xz_dec_run() will return
   2.350 ++ * XZ_MEMLIMIT_ERROR. Thus, it is important to know what kind of data will be
   2.351 ++ * decoded to avoid allocating excessive amount of memory for the dictionary.
   2.352 ++ *
   2.353 ++ * Multi-call mode with dynamically allocated dictionary (XZ_DYNALLOC):
   2.354 ++ * dict_max specifies the maximum allowed dictionary size that xz_dec_run()
   2.355 ++ * may allocate once it has parsed the dictionary size from the stream
   2.356 ++ * headers. This way excessive allocations can be avoided while still
   2.357 ++ * limiting the maximum memory usage to a sane value to prevent running the
   2.358 ++ * system out of memory when decompressing streams from untrusted sources.
   2.359 ++ *
   2.360 ++ * On success, xz_dec_init() returns a pointer to struct xz_dec, which is
   2.361 ++ * ready to be used with xz_dec_run(). If memory allocation fails,
   2.362 ++ * xz_dec_init() returns NULL.
   2.363 ++ */
   2.364 ++XZ_EXTERN struct xz_dec *xz_dec_init(enum xz_mode mode, uint32_t dict_max);
   2.365 ++
   2.366 ++/**
   2.367 ++ * xz_dec_run() - Run the XZ decoder
   2.368 ++ * @s:          Decoder state allocated using xz_dec_init()
   2.369 ++ * @b:          Input and output buffers
   2.370 ++ *
   2.371 ++ * The possible return values depend on build options and operation mode.
   2.372 ++ * See enum xz_ret for details.
   2.373 ++ *
   2.374 ++ * Note that if an error occurs in single-call mode (return value is not
   2.375 ++ * XZ_STREAM_END), b->in_pos and b->out_pos are not modified and the
   2.376 ++ * contents of the output buffer from b->out[b->out_pos] onward are
   2.377 ++ * undefined. This is true even after XZ_BUF_ERROR, because with some filter
   2.378 ++ * chains, there may be a second pass over the output buffer, and this pass
   2.379 ++ * cannot be properly done if the output buffer is truncated. Thus, you
   2.380 ++ * cannot give the single-call decoder a too small buffer and then expect to
   2.381 ++ * get that amount valid data from the beginning of the stream. You must use
   2.382 ++ * the multi-call decoder if you don't want to uncompress the whole stream.
   2.383 ++ */
   2.384 ++XZ_EXTERN enum xz_ret xz_dec_run(struct xz_dec *s, struct xz_buf *b);
   2.385 ++
   2.386 ++/**
   2.387 ++ * xz_dec_reset() - Reset an already allocated decoder state
   2.388 ++ * @s:          Decoder state allocated using xz_dec_init()
   2.389 ++ *
   2.390 ++ * This function can be used to reset the multi-call decoder state without
   2.391 ++ * freeing and reallocating memory with xz_dec_end() and xz_dec_init().
   2.392 ++ *
   2.393 ++ * In single-call mode, xz_dec_reset() is always called in the beginning of
   2.394 ++ * xz_dec_run(). Thus, explicit call to xz_dec_reset() is useful only in
   2.395 ++ * multi-call mode.
   2.396 ++ */
   2.397 ++XZ_EXTERN void xz_dec_reset(struct xz_dec *s);
   2.398 ++
   2.399 ++/**
   2.400 ++ * xz_dec_end() - Free the memory allocated for the decoder state
   2.401 ++ * @s:          Decoder state allocated using xz_dec_init(). If s is NULL,
   2.402 ++ *              this function does nothing.
   2.403 ++ */
   2.404 ++XZ_EXTERN void xz_dec_end(struct xz_dec *s);
   2.405 ++
   2.406 ++/*
   2.407 ++ * Standalone build (userspace build or in-kernel build for boot time use)
   2.408 ++ * needs a CRC32 implementation. For normal in-kernel use, kernel's own
   2.409 ++ * CRC32 module is used instead, and users of this module don't need to
   2.410 ++ * care about the functions below.
   2.411 ++ */
   2.412 ++#ifndef XZ_INTERNAL_CRC32
   2.413 ++#	ifdef __KERNEL__
   2.414 ++#		define XZ_INTERNAL_CRC32 0
   2.415 ++#	else
   2.416 ++#		define XZ_INTERNAL_CRC32 1
   2.417 ++#	endif
   2.418 ++#endif
   2.419 ++
   2.420 ++#if XZ_INTERNAL_CRC32
   2.421 ++/*
   2.422 ++ * This must be called before any other xz_* function to initialize
   2.423 ++ * the CRC32 lookup table.
   2.424 ++ */
   2.425 ++XZ_EXTERN void xz_crc32_init(void);
   2.426 ++
   2.427 ++/*
   2.428 ++ * Update CRC32 value using the polynomial from IEEE-802.3. To start a new
   2.429 ++ * calculation, the third argument must be zero. To continue the calculation,
   2.430 ++ * the previously returned value is passed as the third argument.
   2.431 ++ */
   2.432 ++XZ_EXTERN uint32_t xz_crc32(const uint8_t *buf, size_t size, uint32_t crc);
   2.433 ++#endif
   2.434 ++#endif
   2.435 +diff --git a/lib/Kconfig b/lib/Kconfig
   2.436 +index fa9bf2c..6090314 100644
   2.437 +--- a/lib/Kconfig
   2.438 ++++ b/lib/Kconfig
   2.439 +@@ -106,6 +106,8 @@ config LZO_COMPRESS
   2.440 + config LZO_DECOMPRESS
   2.441 + 	tristate
   2.442 + 
   2.443 ++source "lib/xz/Kconfig"
   2.444 ++
   2.445 + #
   2.446 + # These all provide a common interface (hence the apparent duplication with
   2.447 + # ZLIB_INFLATE; DECOMPRESS_GZIP is just a wrapper.)
   2.448 +diff --git a/lib/Makefile b/lib/Makefile
   2.449 +index e6a3763..f2f98dd 100644
   2.450 +--- a/lib/Makefile
   2.451 ++++ b/lib/Makefile
   2.452 +@@ -69,6 +69,7 @@ obj-$(CONFIG_ZLIB_DEFLATE) += zlib_deflate/
   2.453 + obj-$(CONFIG_REED_SOLOMON) += reed_solomon/
   2.454 + obj-$(CONFIG_LZO_COMPRESS) += lzo/
   2.455 + obj-$(CONFIG_LZO_DECOMPRESS) += lzo/
   2.456 ++obj-$(CONFIG_XZ_DEC) += xz/
   2.457 + obj-$(CONFIG_RAID6_PQ) += raid6/
   2.458 + 
   2.459 + lib-$(CONFIG_DECOMPRESS_GZIP) += decompress_inflate.o
   2.460 +diff --git a/lib/xz/Kconfig b/lib/xz/Kconfig
   2.461 +new file mode 100644
   2.462 +index 0000000..e3b6e18
   2.463 +--- /dev/null
   2.464 ++++ b/lib/xz/Kconfig
   2.465 +@@ -0,0 +1,59 @@
   2.466 ++config XZ_DEC
   2.467 ++	tristate "XZ decompression support"
   2.468 ++	select CRC32
   2.469 ++	help
   2.470 ++	  LZMA2 compression algorithm and BCJ filters are supported using
   2.471 ++	  the .xz file format as the container. For integrity checking,
   2.472 ++	  CRC32 is supported. See Documentation/xz.txt for more information.
   2.473 ++
   2.474 ++config XZ_DEC_X86
   2.475 ++	bool "x86 BCJ filter decoder" if EMBEDDED
   2.476 ++	default y
   2.477 ++	depends on XZ_DEC
   2.478 ++	select XZ_DEC_BCJ
   2.479 ++
   2.480 ++config XZ_DEC_POWERPC
   2.481 ++	bool "PowerPC BCJ filter decoder" if EMBEDDED
   2.482 ++	default y
   2.483 ++	depends on XZ_DEC
   2.484 ++	select XZ_DEC_BCJ
   2.485 ++
   2.486 ++config XZ_DEC_IA64
   2.487 ++	bool "IA-64 BCJ filter decoder" if EMBEDDED
   2.488 ++	default y
   2.489 ++	depends on XZ_DEC
   2.490 ++	select XZ_DEC_BCJ
   2.491 ++
   2.492 ++config XZ_DEC_ARM
   2.493 ++	bool "ARM BCJ filter decoder" if EMBEDDED
   2.494 ++	default y
   2.495 ++	depends on XZ_DEC
   2.496 ++	select XZ_DEC_BCJ
   2.497 ++
   2.498 ++config XZ_DEC_ARMTHUMB
   2.499 ++	bool "ARM-Thumb BCJ filter decoder" if EMBEDDED
   2.500 ++	default y
   2.501 ++	depends on XZ_DEC
   2.502 ++	select XZ_DEC_BCJ
   2.503 ++
   2.504 ++config XZ_DEC_SPARC
   2.505 ++	bool "SPARC BCJ filter decoder" if EMBEDDED
   2.506 ++	default y
   2.507 ++	depends on XZ_DEC
   2.508 ++	select XZ_DEC_BCJ
   2.509 ++
   2.510 ++config XZ_DEC_BCJ
   2.511 ++	bool
   2.512 ++	default n
   2.513 ++
   2.514 ++config XZ_DEC_TEST
   2.515 ++	tristate "XZ decompressor tester"
   2.516 ++	default n
   2.517 ++	depends on XZ_DEC
   2.518 ++	help
   2.519 ++	  This allows passing .xz files to the in-kernel XZ decoder via
   2.520 ++	  a character special file. It calculates CRC32 of the decompressed
   2.521 ++	  data and writes diagnostics to the system log.
   2.522 ++
   2.523 ++	  Unless you are developing the XZ decoder, you don't need this
   2.524 ++	  and should say N.
   2.525 +diff --git a/lib/xz/Makefile b/lib/xz/Makefile
   2.526 +new file mode 100644
   2.527 +index 0000000..a7fa769
   2.528 +--- /dev/null
   2.529 ++++ b/lib/xz/Makefile
   2.530 +@@ -0,0 +1,5 @@
   2.531 ++obj-$(CONFIG_XZ_DEC) += xz_dec.o
   2.532 ++xz_dec-y := xz_dec_syms.o xz_dec_stream.o xz_dec_lzma2.o
   2.533 ++xz_dec-$(CONFIG_XZ_DEC_BCJ) += xz_dec_bcj.o
   2.534 ++
   2.535 ++obj-$(CONFIG_XZ_DEC_TEST) += xz_dec_test.o
   2.536 +diff --git a/lib/xz/xz_crc32.c b/lib/xz/xz_crc32.c
   2.537 +new file mode 100644
   2.538 +index 0000000..34532d1
   2.539 +--- /dev/null
   2.540 ++++ b/lib/xz/xz_crc32.c
   2.541 +@@ -0,0 +1,59 @@
   2.542 ++/*
   2.543 ++ * CRC32 using the polynomial from IEEE-802.3
   2.544 ++ *
   2.545 ++ * Authors: Lasse Collin <lasse.collin@tukaani.org>
   2.546 ++ *          Igor Pavlov <http://7-zip.org/>
   2.547 ++ *
   2.548 ++ * This file has been put into the public domain.
   2.549 ++ * You can do whatever you want with this file.
   2.550 ++ */
   2.551 ++
   2.552 ++/*
   2.553 ++ * This is not the fastest implementation, but it is pretty compact.
   2.554 ++ * The fastest versions of xz_crc32() on modern CPUs without hardware
   2.555 ++ * accelerated CRC instruction are 3-5 times as fast as this version,
   2.556 ++ * but they are bigger and use more memory for the lookup table.
   2.557 ++ */
   2.558 ++
   2.559 ++#include "xz_private.h"
   2.560 ++
   2.561 ++/*
   2.562 ++ * STATIC_RW_DATA is used in the pre-boot environment on some architectures.
   2.563 ++ * See <linux/decompress/mm.h> for details.
   2.564 ++ */
   2.565 ++#ifndef STATIC_RW_DATA
   2.566 ++#	define STATIC_RW_DATA static
   2.567 ++#endif
   2.568 ++
   2.569 ++STATIC_RW_DATA uint32_t xz_crc32_table[256];
   2.570 ++
   2.571 ++XZ_EXTERN void xz_crc32_init(void)
   2.572 ++{
   2.573 ++	const uint32_t poly = 0xEDB88320;
   2.574 ++
   2.575 ++	uint32_t i;
   2.576 ++	uint32_t j;
   2.577 ++	uint32_t r;
   2.578 ++
   2.579 ++	for (i = 0; i < 256; ++i) {
   2.580 ++		r = i;
   2.581 ++		for (j = 0; j < 8; ++j)
   2.582 ++			r = (r >> 1) ^ (poly & ~((r & 1) - 1));
   2.583 ++
   2.584 ++		xz_crc32_table[i] = r;
   2.585 ++	}
   2.586 ++
   2.587 ++	return;
   2.588 ++}
   2.589 ++
   2.590 ++XZ_EXTERN uint32_t xz_crc32(const uint8_t *buf, size_t size, uint32_t crc)
   2.591 ++{
   2.592 ++	crc = ~crc;
   2.593 ++
   2.594 ++	while (size != 0) {
   2.595 ++		crc = xz_crc32_table[*buf++ ^ (crc & 0xFF)] ^ (crc >> 8);
   2.596 ++		--size;
   2.597 ++	}
   2.598 ++
   2.599 ++	return ~crc;
   2.600 ++}
   2.601 +diff --git a/lib/xz/xz_dec_bcj.c b/lib/xz/xz_dec_bcj.c
   2.602 +new file mode 100644
   2.603 +index 0000000..e51e255
   2.604 +--- /dev/null
   2.605 ++++ b/lib/xz/xz_dec_bcj.c
   2.606 +@@ -0,0 +1,561 @@
   2.607 ++/*
   2.608 ++ * Branch/Call/Jump (BCJ) filter decoders
   2.609 ++ *
   2.610 ++ * Authors: Lasse Collin <lasse.collin@tukaani.org>
   2.611 ++ *          Igor Pavlov <http://7-zip.org/>
   2.612 ++ *
   2.613 ++ * This file has been put into the public domain.
   2.614 ++ * You can do whatever you want with this file.
   2.615 ++ */
   2.616 ++
   2.617 ++#include "xz_private.h"
   2.618 ++
   2.619 ++/*
   2.620 ++ * The rest of the file is inside this ifdef. It makes things a little more
   2.621 ++ * convenient when building without support for any BCJ filters.
   2.622 ++ */
   2.623 ++#ifdef XZ_DEC_BCJ
   2.624 ++
   2.625 ++struct xz_dec_bcj {
   2.626 ++	/* Type of the BCJ filter being used */
   2.627 ++	enum {
   2.628 ++		BCJ_X86 = 4,        /* x86 or x86-64 */
   2.629 ++		BCJ_POWERPC = 5,    /* Big endian only */
   2.630 ++		BCJ_IA64 = 6,       /* Big or little endian */
   2.631 ++		BCJ_ARM = 7,        /* Little endian only */
   2.632 ++		BCJ_ARMTHUMB = 8,   /* Little endian only */
   2.633 ++		BCJ_SPARC = 9       /* Big or little endian */
   2.634 ++	} type;
   2.635 ++
   2.636 ++	/*
   2.637 ++	 * Return value of the next filter in the chain. We need to preserve
   2.638 ++	 * this information across calls, because we must not call the next
   2.639 ++	 * filter anymore once it has returned XZ_STREAM_END.
   2.640 ++	 */
   2.641 ++	enum xz_ret ret;
   2.642 ++
   2.643 ++	/* True if we are operating in single-call mode. */
   2.644 ++	bool single_call;
   2.645 ++
   2.646 ++	/*
   2.647 ++	 * Absolute position relative to the beginning of the uncompressed
   2.648 ++	 * data (in a single .xz Block). We care only about the lowest 32
   2.649 ++	 * bits so this doesn't need to be uint64_t even with big files.
   2.650 ++	 */
   2.651 ++	uint32_t pos;
   2.652 ++
   2.653 ++	/* x86 filter state */
   2.654 ++	uint32_t x86_prev_mask;
   2.655 ++
   2.656 ++	/* Temporary space to hold the variables from struct xz_buf */
   2.657 ++	uint8_t *out;
   2.658 ++	size_t out_pos;
   2.659 ++	size_t out_size;
   2.660 ++
   2.661 ++	struct {
   2.662 ++		/* Amount of already filtered data in the beginning of buf */
   2.663 ++		size_t filtered;
   2.664 ++
   2.665 ++		/* Total amount of data currently stored in buf  */
   2.666 ++		size_t size;
   2.667 ++
   2.668 ++		/*
   2.669 ++		 * Buffer to hold a mix of filtered and unfiltered data. This
   2.670 ++		 * needs to be big enough to hold Alignment + 2 * Look-ahead:
   2.671 ++		 *
   2.672 ++		 * Type         Alignment   Look-ahead
   2.673 ++		 * x86              1           4
   2.674 ++		 * PowerPC          4           0
   2.675 ++		 * IA-64           16           0
   2.676 ++		 * ARM              4           0
   2.677 ++		 * ARM-Thumb        2           2
   2.678 ++		 * SPARC            4           0
   2.679 ++		 */
   2.680 ++		uint8_t buf[16];
   2.681 ++	} temp;
   2.682 ++};
   2.683 ++
   2.684 ++#ifdef XZ_DEC_X86
   2.685 ++/*
   2.686 ++ * This is used to test the most significant byte of a memory address
   2.687 ++ * in an x86 instruction.
   2.688 ++ */
   2.689 ++static inline int bcj_x86_test_msbyte(uint8_t b)
   2.690 ++{
   2.691 ++	return b == 0x00 || b == 0xFF;
   2.692 ++}
   2.693 ++
   2.694 ++static size_t bcj_x86(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
   2.695 ++{
   2.696 ++	static const bool mask_to_allowed_status[8]
   2.697 ++		= { true, true, true, false, true, false, false, false };
   2.698 ++
   2.699 ++	static const uint8_t mask_to_bit_num[8] = { 0, 1, 2, 2, 3, 3, 3, 3 };
   2.700 ++
   2.701 ++	size_t i;
   2.702 ++	size_t prev_pos = (size_t)-1;
   2.703 ++	uint32_t prev_mask = s->x86_prev_mask;
   2.704 ++	uint32_t src;
   2.705 ++	uint32_t dest;
   2.706 ++	uint32_t j;
   2.707 ++	uint8_t b;
   2.708 ++
   2.709 ++	if (size <= 4)
   2.710 ++		return 0;
   2.711 ++
   2.712 ++	size -= 4;
   2.713 ++	for (i = 0; i < size; ++i) {
   2.714 ++		if ((buf[i] & 0xFE) != 0xE8)
   2.715 ++			continue;
   2.716 ++
   2.717 ++		prev_pos = i - prev_pos;
   2.718 ++		if (prev_pos > 3) {
   2.719 ++			prev_mask = 0;
   2.720 ++		} else {
   2.721 ++			prev_mask = (prev_mask << (prev_pos - 1)) & 7;
   2.722 ++			if (prev_mask != 0) {
   2.723 ++				b = buf[i + 4 - mask_to_bit_num[prev_mask]];
   2.724 ++				if (!mask_to_allowed_status[prev_mask]
   2.725 ++						|| bcj_x86_test_msbyte(b)) {
   2.726 ++					prev_pos = i;
   2.727 ++					prev_mask = (prev_mask << 1) | 1;
   2.728 ++					continue;
   2.729 ++				}
   2.730 ++			}
   2.731 ++		}
   2.732 ++
   2.733 ++		prev_pos = i;
   2.734 ++
   2.735 ++		if (bcj_x86_test_msbyte(buf[i + 4])) {
   2.736 ++			src = get_unaligned_le32(buf + i + 1);
   2.737 ++			while (true) {
   2.738 ++				dest = src - (s->pos + (uint32_t)i + 5);
   2.739 ++				if (prev_mask == 0)
   2.740 ++					break;
   2.741 ++
   2.742 ++				j = mask_to_bit_num[prev_mask] * 8;
   2.743 ++				b = (uint8_t)(dest >> (24 - j));
   2.744 ++				if (!bcj_x86_test_msbyte(b))
   2.745 ++					break;
   2.746 ++
   2.747 ++				src = dest ^ (((uint32_t)1 << (32 - j)) - 1);
   2.748 ++			}
   2.749 ++
   2.750 ++			dest &= 0x01FFFFFF;
   2.751 ++			dest |= (uint32_t)0 - (dest & 0x01000000);
   2.752 ++			put_unaligned_le32(dest, buf + i + 1);
   2.753 ++			i += 4;
   2.754 ++		} else {
   2.755 ++			prev_mask = (prev_mask << 1) | 1;
   2.756 ++		}
   2.757 ++	}
   2.758 ++
   2.759 ++	prev_pos = i - prev_pos;
   2.760 ++	s->x86_prev_mask = prev_pos > 3 ? 0 : prev_mask << (prev_pos - 1);
   2.761 ++	return i;
   2.762 ++}
   2.763 ++#endif
   2.764 ++
   2.765 ++#ifdef XZ_DEC_POWERPC
   2.766 ++static size_t bcj_powerpc(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
   2.767 ++{
   2.768 ++	size_t i;
   2.769 ++	uint32_t instr;
   2.770 ++
   2.771 ++	for (i = 0; i + 4 <= size; i += 4) {
   2.772 ++		instr = get_unaligned_be32(buf + i);
   2.773 ++		if ((instr & 0xFC000003) == 0x48000001) {
   2.774 ++			instr &= 0x03FFFFFC;
   2.775 ++			instr -= s->pos + (uint32_t)i;
   2.776 ++			instr &= 0x03FFFFFC;
   2.777 ++			instr |= 0x48000001;
   2.778 ++			put_unaligned_be32(instr, buf + i);
   2.779 ++		}
   2.780 ++	}
   2.781 ++
   2.782 ++	return i;
   2.783 ++}
   2.784 ++#endif
   2.785 ++
   2.786 ++#ifdef XZ_DEC_IA64
   2.787 ++static size_t bcj_ia64(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
   2.788 ++{
   2.789 ++	static const uint8_t branch_table[32] = {
   2.790 ++		0, 0, 0, 0, 0, 0, 0, 0,
   2.791 ++		0, 0, 0, 0, 0, 0, 0, 0,
   2.792 ++		4, 4, 6, 6, 0, 0, 7, 7,
   2.793 ++		4, 4, 0, 0, 4, 4, 0, 0
   2.794 ++	};
   2.795 ++
   2.796 ++	/*
   2.797 ++	 * The local variables take a little bit stack space, but it's less
   2.798 ++	 * than what LZMA2 decoder takes, so it doesn't make sense to reduce
   2.799 ++	 * stack usage here without doing that for the LZMA2 decoder too.
   2.800 ++	 */
   2.801 ++
   2.802 ++	/* Loop counters */
   2.803 ++	size_t i;
   2.804 ++	size_t j;
   2.805 ++
   2.806 ++	/* Instruction slot (0, 1, or 2) in the 128-bit instruction word */
   2.807 ++	uint32_t slot;
   2.808 ++
   2.809 ++	/* Bitwise offset of the instruction indicated by slot */
   2.810 ++	uint32_t bit_pos;
   2.811 ++
   2.812 ++	/* bit_pos split into byte and bit parts */
   2.813 ++	uint32_t byte_pos;
   2.814 ++	uint32_t bit_res;
   2.815 ++
   2.816 ++	/* Address part of an instruction */
   2.817 ++	uint32_t addr;
   2.818 ++
   2.819 ++	/* Mask used to detect which instructions to convert */
   2.820 ++	uint32_t mask;
   2.821 ++
   2.822 ++	/* 41-bit instruction stored somewhere in the lowest 48 bits */
   2.823 ++	uint64_t instr;
   2.824 ++
   2.825 ++	/* Instruction normalized with bit_res for easier manipulation */
   2.826 ++	uint64_t norm;
   2.827 ++
   2.828 ++	for (i = 0; i + 16 <= size; i += 16) {
   2.829 ++		mask = branch_table[buf[i] & 0x1F];
   2.830 ++		for (slot = 0, bit_pos = 5; slot < 3; ++slot, bit_pos += 41) {
   2.831 ++			if (((mask >> slot) & 1) == 0)
   2.832 ++				continue;
   2.833 ++
   2.834 ++			byte_pos = bit_pos >> 3;
   2.835 ++			bit_res = bit_pos & 7;
   2.836 ++			instr = 0;
   2.837 ++			for (j = 0; j < 6; ++j)
   2.838 ++				instr |= (uint64_t)(buf[i + j + byte_pos])
   2.839 ++						<< (8 * j);
   2.840 ++
   2.841 ++			norm = instr >> bit_res;
   2.842 ++
   2.843 ++			if (((norm >> 37) & 0x0F) == 0x05
   2.844 ++					&& ((norm >> 9) & 0x07) == 0) {
   2.845 ++				addr = (norm >> 13) & 0x0FFFFF;
   2.846 ++				addr |= ((uint32_t)(norm >> 36) & 1) << 20;
   2.847 ++				addr <<= 4;
   2.848 ++				addr -= s->pos + (uint32_t)i;
   2.849 ++				addr >>= 4;
   2.850 ++
   2.851 ++				norm &= ~((uint64_t)0x8FFFFF << 13);
   2.852 ++				norm |= (uint64_t)(addr & 0x0FFFFF) << 13;
   2.853 ++				norm |= (uint64_t)(addr & 0x100000)
   2.854 ++						<< (36 - 20);
   2.855 ++
   2.856 ++				instr &= (1 << bit_res) - 1;
   2.857 ++				instr |= norm << bit_res;
   2.858 ++
   2.859 ++				for (j = 0; j < 6; j++)
   2.860 ++					buf[i + j + byte_pos]
   2.861 ++						= (uint8_t)(instr >> (8 * j));
   2.862 ++			}
   2.863 ++		}
   2.864 ++	}
   2.865 ++
   2.866 ++	return i;
   2.867 ++}
   2.868 ++#endif
   2.869 ++
   2.870 ++#ifdef XZ_DEC_ARM
   2.871 ++static size_t bcj_arm(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
   2.872 ++{
   2.873 ++	size_t i;
   2.874 ++	uint32_t addr;
   2.875 ++
   2.876 ++	for (i = 0; i + 4 <= size; i += 4) {
   2.877 ++		if (buf[i + 3] == 0xEB) {
   2.878 ++			addr = (uint32_t)buf[i] | ((uint32_t)buf[i + 1] << 8)
   2.879 ++					| ((uint32_t)buf[i + 2] << 16);
   2.880 ++			addr <<= 2;
   2.881 ++			addr -= s->pos + (uint32_t)i + 8;
   2.882 ++			addr >>= 2;
   2.883 ++			buf[i] = (uint8_t)addr;
   2.884 ++			buf[i + 1] = (uint8_t)(addr >> 8);
   2.885 ++			buf[i + 2] = (uint8_t)(addr >> 16);
   2.886 ++		}
   2.887 ++	}
   2.888 ++
   2.889 ++	return i;
   2.890 ++}
   2.891 ++#endif
   2.892 ++
   2.893 ++#ifdef XZ_DEC_ARMTHUMB
   2.894 ++static size_t bcj_armthumb(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
   2.895 ++{
   2.896 ++	size_t i;
   2.897 ++	uint32_t addr;
   2.898 ++
   2.899 ++	for (i = 0; i + 4 <= size; i += 2) {
   2.900 ++		if ((buf[i + 1] & 0xF8) == 0xF0
   2.901 ++				&& (buf[i + 3] & 0xF8) == 0xF8) {
   2.902 ++			addr = (((uint32_t)buf[i + 1] & 0x07) << 19)
   2.903 ++					| ((uint32_t)buf[i] << 11)
   2.904 ++					| (((uint32_t)buf[i + 3] & 0x07) << 8)
   2.905 ++					| (uint32_t)buf[i + 2];
   2.906 ++			addr <<= 1;
   2.907 ++			addr -= s->pos + (uint32_t)i + 4;
   2.908 ++			addr >>= 1;
   2.909 ++			buf[i + 1] = (uint8_t)(0xF0 | ((addr >> 19) & 0x07));
   2.910 ++			buf[i] = (uint8_t)(addr >> 11);
   2.911 ++			buf[i + 3] = (uint8_t)(0xF8 | ((addr >> 8) & 0x07));
   2.912 ++			buf[i + 2] = (uint8_t)addr;
   2.913 ++			i += 2;
   2.914 ++		}
   2.915 ++	}
   2.916 ++
   2.917 ++	return i;
   2.918 ++}
   2.919 ++#endif
   2.920 ++
   2.921 ++#ifdef XZ_DEC_SPARC
   2.922 ++static size_t bcj_sparc(struct xz_dec_bcj *s, uint8_t *buf, size_t size)
   2.923 ++{
   2.924 ++	size_t i;
   2.925 ++	uint32_t instr;
   2.926 ++
   2.927 ++	for (i = 0; i + 4 <= size; i += 4) {
   2.928 ++		instr = get_unaligned_be32(buf + i);
   2.929 ++		if ((instr >> 22) == 0x100 || (instr >> 22) == 0x1FF) {
   2.930 ++			instr <<= 2;
   2.931 ++			instr -= s->pos + (uint32_t)i;
   2.932 ++			instr >>= 2;
   2.933 ++			instr = ((uint32_t)0x40000000 - (instr & 0x400000))
   2.934 ++					| 0x40000000 | (instr & 0x3FFFFF);
   2.935 ++			put_unaligned_be32(instr, buf + i);
   2.936 ++		}
   2.937 ++	}
   2.938 ++
   2.939 ++	return i;
   2.940 ++}
   2.941 ++#endif
   2.942 ++
   2.943 ++/*
   2.944 ++ * Apply the selected BCJ filter. Update *pos and s->pos to match the amount
   2.945 ++ * of data that got filtered.
   2.946 ++ *
   2.947 ++ * NOTE: This is implemented as a switch statement to avoid using function
   2.948 ++ * pointers, which could be problematic in the kernel boot code, which must
   2.949 ++ * avoid pointers to static data (at least on x86).
   2.950 ++ */
   2.951 ++static void bcj_apply(struct xz_dec_bcj *s,
   2.952 ++		      uint8_t *buf, size_t *pos, size_t size)
   2.953 ++{
   2.954 ++	size_t filtered;
   2.955 ++
   2.956 ++	buf += *pos;
   2.957 ++	size -= *pos;
   2.958 ++
   2.959 ++	switch (s->type) {
   2.960 ++#ifdef XZ_DEC_X86
   2.961 ++	case BCJ_X86:
   2.962 ++		filtered = bcj_x86(s, buf, size);
   2.963 ++		break;
   2.964 ++#endif
   2.965 ++#ifdef XZ_DEC_POWERPC
   2.966 ++	case BCJ_POWERPC:
   2.967 ++		filtered = bcj_powerpc(s, buf, size);
   2.968 ++		break;
   2.969 ++#endif
   2.970 ++#ifdef XZ_DEC_IA64
   2.971 ++	case BCJ_IA64:
   2.972 ++		filtered = bcj_ia64(s, buf, size);
   2.973 ++		break;
   2.974 ++#endif
   2.975 ++#ifdef XZ_DEC_ARM
   2.976 ++	case BCJ_ARM:
   2.977 ++		filtered = bcj_arm(s, buf, size);
   2.978 ++		break;
   2.979 ++#endif
   2.980 ++#ifdef XZ_DEC_ARMTHUMB
   2.981 ++	case BCJ_ARMTHUMB:
   2.982 ++		filtered = bcj_armthumb(s, buf, size);
   2.983 ++		break;
   2.984 ++#endif
   2.985 ++#ifdef XZ_DEC_SPARC
   2.986 ++	case BCJ_SPARC:
   2.987 ++		filtered = bcj_sparc(s, buf, size);
   2.988 ++		break;
   2.989 ++#endif
   2.990 ++	default:
   2.991 ++		/* Never reached but silence compiler warnings. */
   2.992 ++		filtered = 0;
   2.993 ++		break;
   2.994 ++	}
   2.995 ++
   2.996 ++	*pos += filtered;
   2.997 ++	s->pos += filtered;
   2.998 ++}
   2.999 ++
  2.1000 ++/*
  2.1001 ++ * Flush pending filtered data from temp to the output buffer.
  2.1002 ++ * Move the remaining mixture of possibly filtered and unfiltered
  2.1003 ++ * data to the beginning of temp.
  2.1004 ++ */
  2.1005 ++static void bcj_flush(struct xz_dec_bcj *s, struct xz_buf *b)
  2.1006 ++{
  2.1007 ++	size_t copy_size;
  2.1008 ++
  2.1009 ++	copy_size = min_t(size_t, s->temp.filtered, b->out_size - b->out_pos);
  2.1010 ++	memcpy(b->out + b->out_pos, s->temp.buf, copy_size);
  2.1011 ++	b->out_pos += copy_size;
  2.1012 ++
  2.1013 ++	s->temp.filtered -= copy_size;
  2.1014 ++	s->temp.size -= copy_size;
  2.1015 ++	memmove(s->temp.buf, s->temp.buf + copy_size, s->temp.size);
  2.1016 ++}
  2.1017 ++
  2.1018 ++/*
  2.1019 ++ * The BCJ filter functions are primitive in sense that they process the
  2.1020 ++ * data in chunks of 1-16 bytes. To hide this issue, this function does
  2.1021 ++ * some buffering.
  2.1022 ++ */
  2.1023 ++XZ_EXTERN enum xz_ret xz_dec_bcj_run(struct xz_dec_bcj *s,
  2.1024 ++				     struct xz_dec_lzma2 *lzma2,
  2.1025 ++				     struct xz_buf *b)
  2.1026 ++{
  2.1027 ++	size_t out_start;
  2.1028 ++
  2.1029 ++	/*
  2.1030 ++	 * Flush pending already filtered data to the output buffer. Return
  2.1031 ++	 * immediatelly if we couldn't flush everything, or if the next
  2.1032 ++	 * filter in the chain had already returned XZ_STREAM_END.
  2.1033 ++	 */
  2.1034 ++	if (s->temp.filtered > 0) {
  2.1035 ++		bcj_flush(s, b);
  2.1036 ++		if (s->temp.filtered > 0)
  2.1037 ++			return XZ_OK;
  2.1038 ++
  2.1039 ++		if (s->ret == XZ_STREAM_END)
  2.1040 ++			return XZ_STREAM_END;
  2.1041 ++	}
  2.1042 ++
  2.1043 ++	/*
  2.1044 ++	 * If we have more output space than what is currently pending in
  2.1045 ++	 * temp, copy the unfiltered data from temp to the output buffer
  2.1046 ++	 * and try to fill the output buffer by decoding more data from the
  2.1047 ++	 * next filter in the chain. Apply the BCJ filter on the new data
  2.1048 ++	 * in the output buffer. If everything cannot be filtered, copy it
  2.1049 ++	 * to temp and rewind the output buffer position accordingly.
  2.1050 ++	 */
  2.1051 ++	if (s->temp.size < b->out_size - b->out_pos) {
  2.1052 ++		out_start = b->out_pos;
  2.1053 ++		memcpy(b->out + b->out_pos, s->temp.buf, s->temp.size);
  2.1054 ++		b->out_pos += s->temp.size;
  2.1055 ++
  2.1056 ++		s->ret = xz_dec_lzma2_run(lzma2, b);
  2.1057 ++		if (s->ret != XZ_STREAM_END
  2.1058 ++				&& (s->ret != XZ_OK || s->single_call))
  2.1059 ++			return s->ret;
  2.1060 ++
  2.1061 ++		bcj_apply(s, b->out, &out_start, b->out_pos);
  2.1062 ++
  2.1063 ++		/*
  2.1064 ++		 * As an exception, if the next filter returned XZ_STREAM_END,
  2.1065 ++		 * we can do that too, since the last few bytes that remain
  2.1066 ++		 * unfiltered are meant to remain unfiltered.
  2.1067 ++		 */
  2.1068 ++		if (s->ret == XZ_STREAM_END)
  2.1069 ++			return XZ_STREAM_END;
  2.1070 ++
  2.1071 ++		s->temp.size = b->out_pos - out_start;
  2.1072 ++		b->out_pos -= s->temp.size;
  2.1073 ++		memcpy(s->temp.buf, b->out + b->out_pos, s->temp.size);
  2.1074 ++	}
  2.1075 ++
  2.1076 ++	/*
  2.1077 ++	 * If we have unfiltered data in temp, try to fill by decoding more
  2.1078 ++	 * data from the next filter. Apply the BCJ filter on temp. Then we
  2.1079 ++	 * hopefully can fill the actual output buffer by copying filtered
  2.1080 ++	 * data from temp. A mix of filtered and unfiltered data may be left
  2.1081 ++	 * in temp; it will be taken care on the next call to this function.
  2.1082 ++	 */
  2.1083 ++	if (s->temp.size > 0) {
  2.1084 ++		/* Make b->out{,_pos,_size} temporarily point to s->temp. */
  2.1085 ++		s->out = b->out;
  2.1086 ++		s->out_pos = b->out_pos;
  2.1087 ++		s->out_size = b->out_size;
  2.1088 ++		b->out = s->temp.buf;
  2.1089 ++		b->out_pos = s->temp.size;
  2.1090 ++		b->out_size = sizeof(s->temp.buf);
  2.1091 ++
  2.1092 ++		s->ret = xz_dec_lzma2_run(lzma2, b);
  2.1093 ++
  2.1094 ++		s->temp.size = b->out_pos;
  2.1095 ++		b->out = s->out;
  2.1096 ++		b->out_pos = s->out_pos;
  2.1097 ++		b->out_size = s->out_size;
  2.1098 ++
  2.1099 ++		if (s->ret != XZ_OK && s->ret != XZ_STREAM_END)
  2.1100 ++			return s->ret;
  2.1101 ++
  2.1102 ++		bcj_apply(s, s->temp.buf, &s->temp.filtered, s->temp.size);
  2.1103 ++
  2.1104 ++		/*
  2.1105 ++		 * If the next filter returned XZ_STREAM_END, we mark that
  2.1106 ++		 * everything is filtered, since the last unfiltered bytes
  2.1107 ++		 * of the stream are meant to be left as is.
  2.1108 ++		 */
  2.1109 ++		if (s->ret == XZ_STREAM_END)
  2.1110 ++			s->temp.filtered = s->temp.size;
  2.1111 ++
  2.1112 ++		bcj_flush(s, b);
  2.1113 ++		if (s->temp.filtered > 0)
  2.1114 ++			return XZ_OK;
  2.1115 ++	}
  2.1116 ++
  2.1117 ++	return s->ret;
  2.1118 ++}
  2.1119 ++
  2.1120 ++XZ_EXTERN struct xz_dec_bcj *xz_dec_bcj_create(bool single_call)
  2.1121 ++{
  2.1122 ++	struct xz_dec_bcj *s = kmalloc(sizeof(*s), GFP_KERNEL);
  2.1123 ++	if (s != NULL)
  2.1124 ++		s->single_call = single_call;
  2.1125 ++
  2.1126 ++	return s;
  2.1127 ++}
  2.1128 ++
  2.1129 ++XZ_EXTERN enum xz_ret xz_dec_bcj_reset(struct xz_dec_bcj *s, uint8_t id)
  2.1130 ++{
  2.1131 ++	switch (id) {
  2.1132 ++#ifdef XZ_DEC_X86
  2.1133 ++	case BCJ_X86:
  2.1134 ++#endif
  2.1135 ++#ifdef XZ_DEC_POWERPC
  2.1136 ++	case BCJ_POWERPC:
  2.1137 ++#endif
  2.1138 ++#ifdef XZ_DEC_IA64
  2.1139 ++	case BCJ_IA64:
  2.1140 ++#endif
  2.1141 ++#ifdef XZ_DEC_ARM
  2.1142 ++	case BCJ_ARM:
  2.1143 ++#endif
  2.1144 ++#ifdef XZ_DEC_ARMTHUMB
  2.1145 ++	case BCJ_ARMTHUMB:
  2.1146 ++#endif
  2.1147 ++#ifdef XZ_DEC_SPARC
  2.1148 ++	case BCJ_SPARC:
  2.1149 ++#endif
  2.1150 ++		break;
  2.1151 ++
  2.1152 ++	default:
  2.1153 ++		/* Unsupported Filter ID */
  2.1154 ++		return XZ_OPTIONS_ERROR;
  2.1155 ++	}
  2.1156 ++
  2.1157 ++	s->type = id;
  2.1158 ++	s->ret = XZ_OK;
  2.1159 ++	s->pos = 0;
  2.1160 ++	s->x86_prev_mask = 0;
  2.1161 ++	s->temp.filtered = 0;
  2.1162 ++	s->temp.size = 0;
  2.1163 ++
  2.1164 ++	return XZ_OK;
  2.1165 ++}
  2.1166 ++
  2.1167 ++#endif
  2.1168 +diff --git a/lib/xz/xz_dec_lzma2.c b/lib/xz/xz_dec_lzma2.c
  2.1169 +new file mode 100644
  2.1170 +index 0000000..ea5fa4f
  2.1171 +--- /dev/null
  2.1172 ++++ b/lib/xz/xz_dec_lzma2.c
  2.1173 +@@ -0,0 +1,1171 @@
  2.1174 ++/*
  2.1175 ++ * LZMA2 decoder
  2.1176 ++ *
  2.1177 ++ * Authors: Lasse Collin <lasse.collin@tukaani.org>
  2.1178 ++ *          Igor Pavlov <http://7-zip.org/>
  2.1179 ++ *
  2.1180 ++ * This file has been put into the public domain.
  2.1181 ++ * You can do whatever you want with this file.
  2.1182 ++ */
  2.1183 ++
  2.1184 ++#include "xz_private.h"
  2.1185 ++#include "xz_lzma2.h"
  2.1186 ++
  2.1187 ++/*
  2.1188 ++ * Range decoder initialization eats the first five bytes of each LZMA chunk.
  2.1189 ++ */
  2.1190 ++#define RC_INIT_BYTES 5
  2.1191 ++
  2.1192 ++/*
  2.1193 ++ * Minimum number of usable input buffer to safely decode one LZMA symbol.
  2.1194 ++ * The worst case is that we decode 22 bits using probabilities and 26
  2.1195 ++ * direct bits. This may decode at maximum of 20 bytes of input. However,
  2.1196 ++ * lzma_main() does an extra normalization before returning, thus we
  2.1197 ++ * need to put 21 here.
  2.1198 ++ */
  2.1199 ++#define LZMA_IN_REQUIRED 21
  2.1200 ++
  2.1201 ++/*
  2.1202 ++ * Dictionary (history buffer)
  2.1203 ++ *
  2.1204 ++ * These are always true:
  2.1205 ++ *    start <= pos <= full <= end
  2.1206 ++ *    pos <= limit <= end
  2.1207 ++ *
  2.1208 ++ * In multi-call mode, also these are true:
  2.1209 ++ *    end == size
  2.1210 ++ *    size <= size_max
  2.1211 ++ *    allocated <= size
  2.1212 ++ *
  2.1213 ++ * Most of these variables are size_t to support single-call mode,
  2.1214 ++ * in which the dictionary variables address the actual output
  2.1215 ++ * buffer directly.
  2.1216 ++ */
  2.1217 ++struct dictionary {
  2.1218 ++	/* Beginning of the history buffer */
  2.1219 ++	uint8_t *buf;
  2.1220 ++
  2.1221 ++	/* Old position in buf (before decoding more data) */
  2.1222 ++	size_t start;
  2.1223 ++
  2.1224 ++	/* Position in buf */
  2.1225 ++	size_t pos;
  2.1226 ++
  2.1227 ++	/*
  2.1228 ++	 * How full dictionary is. This is used to detect corrupt input that
  2.1229 ++	 * would read beyond the beginning of the uncompressed stream.
  2.1230 ++	 */
  2.1231 ++	size_t full;
  2.1232 ++
  2.1233 ++	/* Write limit; we don't write to buf[limit] or later bytes. */
  2.1234 ++	size_t limit;
  2.1235 ++
  2.1236 ++	/*
  2.1237 ++	 * End of the dictionary buffer. In multi-call mode, this is
  2.1238 ++	 * the same as the dictionary size. In single-call mode, this
  2.1239 ++	 * indicates the size of the output buffer.
  2.1240 ++	 */
  2.1241 ++	size_t end;
  2.1242 ++
  2.1243 ++	/*
  2.1244 ++	 * Size of the dictionary as specified in Block Header. This is used
  2.1245 ++	 * together with "full" to detect corrupt input that would make us
  2.1246 ++	 * read beyond the beginning of the uncompressed stream.
  2.1247 ++	 */
  2.1248 ++	uint32_t size;
  2.1249 ++
  2.1250 ++	/*
  2.1251 ++	 * Maximum allowed dictionary size in multi-call mode.
  2.1252 ++	 * This is ignored in single-call mode.
  2.1253 ++	 */
  2.1254 ++	uint32_t size_max;
  2.1255 ++
  2.1256 ++	/*
  2.1257 ++	 * Amount of memory currently allocated for the dictionary.
  2.1258 ++	 * This is used only with XZ_DYNALLOC. (With XZ_PREALLOC,
  2.1259 ++	 * size_max is always the same as the allocated size.)
  2.1260 ++	 */
  2.1261 ++	uint32_t allocated;
  2.1262 ++
  2.1263 ++	/* Operation mode */
  2.1264 ++	enum xz_mode mode;
  2.1265 ++};
  2.1266 ++
  2.1267 ++/* Range decoder */
  2.1268 ++struct rc_dec {
  2.1269 ++	uint32_t range;
  2.1270 ++	uint32_t code;
  2.1271 ++
  2.1272 ++	/*
  2.1273 ++	 * Number of initializing bytes remaining to be read
  2.1274 ++	 * by rc_read_init().
  2.1275 ++	 */
  2.1276 ++	uint32_t init_bytes_left;
  2.1277 ++
  2.1278 ++	/*
  2.1279 ++	 * Buffer from which we read our input. It can be either
  2.1280 ++	 * temp.buf or the caller-provided input buffer.
  2.1281 ++	 */
  2.1282 ++	const uint8_t *in;
  2.1283 ++	size_t in_pos;
  2.1284 ++	size_t in_limit;
  2.1285 ++};
  2.1286 ++
  2.1287 ++/* Probabilities for a length decoder. */
  2.1288 ++struct lzma_len_dec {
  2.1289 ++	/* Probability of match length being at least 10 */
  2.1290 ++	uint16_t choice;
  2.1291 ++
  2.1292 ++	/* Probability of match length being at least 18 */
  2.1293 ++	uint16_t choice2;
  2.1294 ++
  2.1295 ++	/* Probabilities for match lengths 2-9 */
  2.1296 ++	uint16_t low[POS_STATES_MAX][LEN_LOW_SYMBOLS];
  2.1297 ++
  2.1298 ++	/* Probabilities for match lengths 10-17 */
  2.1299 ++	uint16_t mid[POS_STATES_MAX][LEN_MID_SYMBOLS];
  2.1300 ++
  2.1301 ++	/* Probabilities for match lengths 18-273 */
  2.1302 ++	uint16_t high[LEN_HIGH_SYMBOLS];
  2.1303 ++};
  2.1304 ++
  2.1305 ++struct lzma_dec {
  2.1306 ++	/* Distances of latest four matches */
  2.1307 ++	uint32_t rep0;
  2.1308 ++	uint32_t rep1;
  2.1309 ++	uint32_t rep2;
  2.1310 ++	uint32_t rep3;
  2.1311 ++
  2.1312 ++	/* Types of the most recently seen LZMA symbols */
  2.1313 ++	enum lzma_state state;
  2.1314 ++
  2.1315 ++	/*
  2.1316 ++	 * Length of a match. This is updated so that dict_repeat can
  2.1317 ++	 * be called again to finish repeating the whole match.
  2.1318 ++	 */
  2.1319 ++	uint32_t len;
  2.1320 ++
  2.1321 ++	/*
  2.1322 ++	 * LZMA properties or related bit masks (number of literal
  2.1323 ++	 * context bits, a mask dervied from the number of literal
  2.1324 ++	 * position bits, and a mask dervied from the number
  2.1325 ++	 * position bits)
  2.1326 ++	 */
  2.1327 ++	uint32_t lc;
  2.1328 ++	uint32_t literal_pos_mask; /* (1 << lp) - 1 */
  2.1329 ++	uint32_t pos_mask;         /* (1 << pb) - 1 */
  2.1330 ++
  2.1331 ++	/* If 1, it's a match. Otherwise it's a single 8-bit literal. */
  2.1332 ++	uint16_t is_match[STATES][POS_STATES_MAX];
  2.1333 ++
  2.1334 ++	/* If 1, it's a repeated match. The distance is one of rep0 .. rep3. */
  2.1335 ++	uint16_t is_rep[STATES];
  2.1336 ++
  2.1337 ++	/*
  2.1338 ++	 * If 0, distance of a repeated match is rep0.
  2.1339 ++	 * Otherwise check is_rep1.
  2.1340 ++	 */
  2.1341 ++	uint16_t is_rep0[STATES];
  2.1342 ++
  2.1343 ++	/*
  2.1344 ++	 * If 0, distance of a repeated match is rep1.
  2.1345 ++	 * Otherwise check is_rep2.
  2.1346 ++	 */
  2.1347 ++	uint16_t is_rep1[STATES];
  2.1348 ++
  2.1349 ++	/* If 0, distance of a repeated match is rep2. Otherwise it is rep3. */
  2.1350 ++	uint16_t is_rep2[STATES];
  2.1351 ++
  2.1352 ++	/*
  2.1353 ++	 * If 1, the repeated match has length of one byte. Otherwise
  2.1354 ++	 * the length is decoded from rep_len_decoder.
  2.1355 ++	 */
  2.1356 ++	uint16_t is_rep0_long[STATES][POS_STATES_MAX];
  2.1357 ++
  2.1358 ++	/*
  2.1359 ++	 * Probability tree for the highest two bits of the match
  2.1360 ++	 * distance. There is a separate probability tree for match
  2.1361 ++	 * lengths of 2 (i.e. MATCH_LEN_MIN), 3, 4, and [5, 273].
  2.1362 ++	 */
  2.1363 ++	uint16_t dist_slot[DIST_STATES][DIST_SLOTS];
  2.1364 ++
  2.1365 ++	/*
  2.1366 ++	 * Probility trees for additional bits for match distance
  2.1367 ++	 * when the distance is in the range [4, 127].
  2.1368 ++	 */
  2.1369 ++	uint16_t dist_special[FULL_DISTANCES - DIST_MODEL_END];
  2.1370 ++
  2.1371 ++	/*
  2.1372 ++	 * Probability tree for the lowest four bits of a match
  2.1373 ++	 * distance that is equal to or greater than 128.
  2.1374 ++	 */
  2.1375 ++	uint16_t dist_align[ALIGN_SIZE];
  2.1376 ++
  2.1377 ++	/* Length of a normal match */
  2.1378 ++	struct lzma_len_dec match_len_dec;
  2.1379 ++
  2.1380 ++	/* Length of a repeated match */
  2.1381 ++	struct lzma_len_dec rep_len_dec;
  2.1382 ++
  2.1383 ++	/* Probabilities of literals */
  2.1384 ++	uint16_t literal[LITERAL_CODERS_MAX][LITERAL_CODER_SIZE];
  2.1385 ++};
  2.1386 ++
  2.1387 ++struct lzma2_dec {
  2.1388 ++	/* Position in xz_dec_lzma2_run(). */
  2.1389 ++	enum lzma2_seq {
  2.1390 ++		SEQ_CONTROL,
  2.1391 ++		SEQ_UNCOMPRESSED_1,
  2.1392 ++		SEQ_UNCOMPRESSED_2,
  2.1393 ++		SEQ_COMPRESSED_0,
  2.1394 ++		SEQ_COMPRESSED_1,
  2.1395 ++		SEQ_PROPERTIES,
  2.1396 ++		SEQ_LZMA_PREPARE,
  2.1397 ++		SEQ_LZMA_RUN,
  2.1398 ++		SEQ_COPY
  2.1399 ++	} sequence;
  2.1400 ++
  2.1401 ++	/* Next position after decoding the compressed size of the chunk. */
  2.1402 ++	enum lzma2_seq next_sequence;
  2.1403 ++
  2.1404 ++	/* Uncompressed size of LZMA chunk (2 MiB at maximum) */
  2.1405 ++	uint32_t uncompressed;
  2.1406 ++
  2.1407 ++	/*
  2.1408 ++	 * Compressed size of LZMA chunk or compressed/uncompressed
  2.1409 ++	 * size of uncompressed chunk (64 KiB at maximum)
  2.1410 ++	 */
  2.1411 ++	uint32_t compressed;
  2.1412 ++
  2.1413 ++	/*
  2.1414 ++	 * True if dictionary reset is needed. This is false before
  2.1415 ++	 * the first chunk (LZMA or uncompressed).
  2.1416 ++	 */
  2.1417 ++	bool need_dict_reset;
  2.1418 ++
  2.1419 ++	/*
  2.1420 ++	 * True if new LZMA properties are needed. This is false
  2.1421 ++	 * before the first LZMA chunk.
  2.1422 ++	 */
  2.1423 ++	bool need_props;
  2.1424 ++};
  2.1425 ++
  2.1426 ++struct xz_dec_lzma2 {
  2.1427 ++	/*
  2.1428 ++	 * The order below is important on x86 to reduce code size and
  2.1429 ++	 * it shouldn't hurt on other platforms. Everything up to and
  2.1430 ++	 * including lzma.pos_mask are in the first 128 bytes on x86-32,
  2.1431 ++	 * which allows using smaller instructions to access those
  2.1432 ++	 * variables. On x86-64, fewer variables fit into the first 128
  2.1433 ++	 * bytes, but this is still the best order without sacrificing
  2.1434 ++	 * the readability by splitting the structures.
  2.1435 ++	 */
  2.1436 ++	struct rc_dec rc;
  2.1437 ++	struct dictionary dict;
  2.1438 ++	struct lzma2_dec lzma2;
  2.1439 ++	struct lzma_dec lzma;
  2.1440 ++
  2.1441 ++	/*
  2.1442 ++	 * Temporary buffer which holds small number of input bytes between
  2.1443 ++	 * decoder calls. See lzma2_lzma() for details.
  2.1444 ++	 */
  2.1445 ++	struct {
  2.1446 ++		uint32_t size;
  2.1447 ++		uint8_t buf[3 * LZMA_IN_REQUIRED];
  2.1448 ++	} temp;
  2.1449 ++};
  2.1450 ++
  2.1451 ++/**************
  2.1452 ++ * Dictionary *
  2.1453 ++ **************/
  2.1454 ++
  2.1455 ++/*
  2.1456 ++ * Reset the dictionary state. When in single-call mode, set up the beginning
  2.1457 ++ * of the dictionary to point to the actual output buffer.
  2.1458 ++ */
  2.1459 ++static void dict_reset(struct dictionary *dict, struct xz_buf *b)
  2.1460 ++{
  2.1461 ++	if (DEC_IS_SINGLE(dict->mode)) {
  2.1462 ++		dict->buf = b->out + b->out_pos;
  2.1463 ++		dict->end = b->out_size - b->out_pos;
  2.1464 ++	}
  2.1465 ++
  2.1466 ++	dict->start = 0;
  2.1467 ++	dict->pos = 0;
  2.1468 ++	dict->limit = 0;
  2.1469 ++	dict->full = 0;
  2.1470 ++}
  2.1471 ++
  2.1472 ++/* Set dictionary write limit */
  2.1473 ++static void dict_limit(struct dictionary *dict, size_t out_max)
  2.1474 ++{
  2.1475 ++	if (dict->end - dict->pos <= out_max)
  2.1476 ++		dict->limit = dict->end;
  2.1477 ++	else
  2.1478 ++		dict->limit = dict->pos + out_max;
  2.1479 ++}
  2.1480 ++
  2.1481 ++/* Return true if at least one byte can be written into the dictionary. */
  2.1482 ++static inline bool dict_has_space(const struct dictionary *dict)
  2.1483 ++{
  2.1484 ++	return dict->pos < dict->limit;
  2.1485 ++}
  2.1486 ++
  2.1487 ++/*
  2.1488 ++ * Get a byte from the dictionary at the given distance. The distance is
  2.1489 ++ * assumed to valid, or as a special case, zero when the dictionary is
  2.1490 ++ * still empty. This special case is needed for single-call decoding to
  2.1491 ++ * avoid writing a '\0' to the end of the destination buffer.
  2.1492 ++ */
  2.1493 ++static inline uint32_t dict_get(const struct dictionary *dict, uint32_t dist)
  2.1494 ++{
  2.1495 ++	size_t offset = dict->pos - dist - 1;
  2.1496 ++
  2.1497 ++	if (dist >= dict->pos)
  2.1498 ++		offset += dict->end;
  2.1499 ++
  2.1500 ++	return dict->full > 0 ? dict->buf[offset] : 0;
  2.1501 ++}
  2.1502 ++
  2.1503 ++/*
  2.1504 ++ * Put one byte into the dictionary. It is assumed that there is space for it.
  2.1505 ++ */
  2.1506 ++static inline void dict_put(struct dictionary *dict, uint8_t byte)
  2.1507 ++{
  2.1508 ++	dict->buf[dict->pos++] = byte;
  2.1509 ++
  2.1510 ++	if (dict->full < dict->pos)
  2.1511 ++		dict->full = dict->pos;
  2.1512 ++}
  2.1513 ++
  2.1514 ++/*
  2.1515 ++ * Repeat given number of bytes from the given distance. If the distance is
  2.1516 ++ * invalid, false is returned. On success, true is returned and *len is
  2.1517 ++ * updated to indicate how many bytes were left to be repeated.
  2.1518 ++ */
  2.1519 ++static bool dict_repeat(struct dictionary *dict, uint32_t *len, uint32_t dist)
  2.1520 ++{
  2.1521 ++	size_t back;
  2.1522 ++	uint32_t left;
  2.1523 ++
  2.1524 ++	if (dist >= dict->full || dist >= dict->size)
  2.1525 ++		return false;
  2.1526 ++
  2.1527 ++	left = min_t(size_t, dict->limit - dict->pos, *len);
  2.1528 ++	*len -= left;
  2.1529 ++
  2.1530 ++	back = dict->pos - dist - 1;
  2.1531 ++	if (dist >= dict->pos)
  2.1532 ++		back += dict->end;
  2.1533 ++
  2.1534 ++	do {
  2.1535 ++		dict->buf[dict->pos++] = dict->buf[back++];
  2.1536 ++		if (back == dict->end)
  2.1537 ++			back = 0;
  2.1538 ++	} while (--left > 0);
  2.1539 ++
  2.1540 ++	if (dict->full < dict->pos)
  2.1541 ++		dict->full = dict->pos;
  2.1542 ++
  2.1543 ++	return true;
  2.1544 ++}
  2.1545 ++
  2.1546 ++/* Copy uncompressed data as is from input to dictionary and output buffers. */
  2.1547 ++static void dict_uncompressed(struct dictionary *dict, struct xz_buf *b,
  2.1548 ++			      uint32_t *left)
  2.1549 ++{
  2.1550 ++	size_t copy_size;
  2.1551 ++
  2.1552 ++	while (*left > 0 && b->in_pos < b->in_size
  2.1553 ++			&& b->out_pos < b->out_size) {
  2.1554 ++		copy_size = min(b->in_size - b->in_pos,
  2.1555 ++				b->out_size - b->out_pos);
  2.1556 ++		if (copy_size > dict->end - dict->pos)
  2.1557 ++			copy_size = dict->end - dict->pos;
  2.1558 ++		if (copy_size > *left)
  2.1559 ++			copy_size = *left;
  2.1560 ++
  2.1561 ++		*left -= copy_size;
  2.1562 ++
  2.1563 ++		memcpy(dict->buf + dict->pos, b->in + b->in_pos, copy_size);
  2.1564 ++		dict->pos += copy_size;
  2.1565 ++
  2.1566 ++		if (dict->full < dict->pos)
  2.1567 ++			dict->full = dict->pos;
  2.1568 ++
  2.1569 ++		if (DEC_IS_MULTI(dict->mode)) {
  2.1570 ++			if (dict->pos == dict->end)
  2.1571 ++				dict->pos = 0;
  2.1572 ++
  2.1573 ++			memcpy(b->out + b->out_pos, b->in + b->in_pos,
  2.1574 ++					copy_size);
  2.1575 ++		}
  2.1576 ++
  2.1577 ++		dict->start = dict->pos;
  2.1578 ++
  2.1579 ++		b->out_pos += copy_size;
  2.1580 ++		b->in_pos += copy_size;
  2.1581 ++	}
  2.1582 ++}
  2.1583 ++
  2.1584 ++/*
  2.1585 ++ * Flush pending data from dictionary to b->out. It is assumed that there is
  2.1586 ++ * enough space in b->out. This is guaranteed because caller uses dict_limit()
  2.1587 ++ * before decoding data into the dictionary.
  2.1588 ++ */
  2.1589 ++static uint32_t dict_flush(struct dictionary *dict, struct xz_buf *b)
  2.1590 ++{
  2.1591 ++	size_t copy_size = dict->pos - dict->start;
  2.1592 ++
  2.1593 ++	if (DEC_IS_MULTI(dict->mode)) {
  2.1594 ++		if (dict->pos == dict->end)
  2.1595 ++			dict->pos = 0;
  2.1596 ++
  2.1597 ++		memcpy(b->out + b->out_pos, dict->buf + dict->start,
  2.1598 ++				copy_size);
  2.1599 ++	}
  2.1600 ++
  2.1601 ++	dict->start = dict->pos;
  2.1602 ++	b->out_pos += copy_size;
  2.1603 ++	return copy_size;
  2.1604 ++}
  2.1605 ++
  2.1606 ++/*****************
  2.1607 ++ * Range decoder *
  2.1608 ++ *****************/
  2.1609 ++
  2.1610 ++/* Reset the range decoder. */
  2.1611 ++static void rc_reset(struct rc_dec *rc)
  2.1612 ++{
  2.1613 ++	rc->range = (uint32_t)-1;
  2.1614 ++	rc->code = 0;
  2.1615 ++	rc->init_bytes_left = RC_INIT_BYTES;
  2.1616 ++}
  2.1617 ++
  2.1618 ++/*
  2.1619 ++ * Read the first five initial bytes into rc->code if they haven't been
  2.1620 ++ * read already. (Yes, the first byte gets completely ignored.)
  2.1621 ++ */
  2.1622 ++static bool rc_read_init(struct rc_dec *rc, struct xz_buf *b)
  2.1623 ++{
  2.1624 ++	while (rc->init_bytes_left > 0) {
  2.1625 ++		if (b->in_pos == b->in_size)
  2.1626 ++			return false;
  2.1627 ++
  2.1628 ++		rc->code = (rc->code << 8) + b->in[b->in_pos++];
  2.1629 ++		--rc->init_bytes_left;
  2.1630 ++	}
  2.1631 ++
  2.1632 ++	return true;
  2.1633 ++}
  2.1634 ++
  2.1635 ++/* Return true if there may not be enough input for the next decoding loop. */
  2.1636 ++static inline bool rc_limit_exceeded(const struct rc_dec *rc)
  2.1637 ++{
  2.1638 ++	return rc->in_pos > rc->in_limit;
  2.1639 ++}
  2.1640 ++
  2.1641 ++/*
  2.1642 ++ * Return true if it is possible (from point of view of range decoder) that
  2.1643 ++ * we have reached the end of the LZMA chunk.
  2.1644 ++ */
  2.1645 ++static inline bool rc_is_finished(const struct rc_dec *rc)
  2.1646 ++{
  2.1647 ++	return rc->code == 0;
  2.1648 ++}
  2.1649 ++
  2.1650 ++/* Read the next input byte if needed. */
  2.1651 ++static __always_inline void rc_normalize(struct rc_dec *rc)
  2.1652 ++{
  2.1653 ++	if (rc->range < RC_TOP_VALUE) {
  2.1654 ++		rc->range <<= RC_SHIFT_BITS;
  2.1655 ++		rc->code = (rc->code << RC_SHIFT_BITS) + rc->in[rc->in_pos++];
  2.1656 ++	}
  2.1657 ++}
  2.1658 ++
  2.1659 ++/*
  2.1660 ++ * Decode one bit. In some versions, this function has been splitted in three
  2.1661 ++ * functions so that the compiler is supposed to be able to more easily avoid
  2.1662 ++ * an extra branch. In this particular version of the LZMA decoder, this
  2.1663 ++ * doesn't seem to be a good idea (tested with GCC 3.3.6, 3.4.6, and 4.3.3
  2.1664 ++ * on x86). Using a non-splitted version results in nicer looking code too.
  2.1665 ++ *
  2.1666 ++ * NOTE: This must return an int. Do not make it return a bool or the speed
  2.1667 ++ * of the code generated by GCC 3.x decreases 10-15 %. (GCC 4.3 doesn't care,
  2.1668 ++ * and it generates 10-20 % faster code than GCC 3.x from this file anyway.)
  2.1669 ++ */
  2.1670 ++static __always_inline int rc_bit(struct rc_dec *rc, uint16_t *prob)
  2.1671 ++{
  2.1672 ++	uint32_t bound;
  2.1673 ++	int bit;
  2.1674 ++
  2.1675 ++	rc_normalize(rc);
  2.1676 ++	bound = (rc->range >> RC_BIT_MODEL_TOTAL_BITS) * *prob;
  2.1677 ++	if (rc->code < bound) {
  2.1678 ++		rc->range = bound;
  2.1679 ++		*prob += (RC_BIT_MODEL_TOTAL - *prob) >> RC_MOVE_BITS;
  2.1680 ++		bit = 0;
  2.1681 ++	} else {
  2.1682 ++		rc->range -= bound;
  2.1683 ++		rc->code -= bound;
  2.1684 ++		*prob -= *prob >> RC_MOVE_BITS;
  2.1685 ++		bit = 1;
  2.1686 ++	}
  2.1687 ++
  2.1688 ++	return bit;
  2.1689 ++}
  2.1690 ++
  2.1691 ++/* Decode a bittree starting from the most significant bit. */
  2.1692 ++static __always_inline uint32_t rc_bittree(struct rc_dec *rc,
  2.1693 ++					   uint16_t *probs, uint32_t limit)
  2.1694 ++{
  2.1695 ++	uint32_t symbol = 1;
  2.1696 ++
  2.1697 ++	do {
  2.1698 ++		if (rc_bit(rc, &probs[symbol]))
  2.1699 ++			symbol = (symbol << 1) + 1;
  2.1700 ++		else
  2.1701 ++			symbol <<= 1;
  2.1702 ++	} while (symbol < limit);
  2.1703 ++
  2.1704 ++	return symbol;
  2.1705 ++}
  2.1706 ++
  2.1707 ++/* Decode a bittree starting from the least significant bit. */
  2.1708 ++static __always_inline void rc_bittree_reverse(struct rc_dec *rc,
  2.1709 ++					       uint16_t *probs,
  2.1710 ++					       uint32_t *dest, uint32_t limit)
  2.1711 ++{
  2.1712 ++	uint32_t symbol = 1;
  2.1713 ++	uint32_t i = 0;
  2.1714 ++
  2.1715 ++	do {
  2.1716 ++		if (rc_bit(rc, &probs[symbol])) {
  2.1717 ++			symbol = (symbol << 1) + 1;
  2.1718 ++			*dest += 1 << i;
  2.1719 ++		} else {
  2.1720 ++			symbol <<= 1;
  2.1721 ++		}
  2.1722 ++	} while (++i < limit);
  2.1723 ++}
  2.1724 ++
  2.1725 ++/* Decode direct bits (fixed fifty-fifty probability) */
  2.1726 ++static inline void rc_direct(struct rc_dec *rc, uint32_t *dest, uint32_t limit)
  2.1727 ++{
  2.1728 ++	uint32_t mask;
  2.1729 ++
  2.1730 ++	do {
  2.1731 ++		rc_normalize(rc);
  2.1732 ++		rc->range >>= 1;
  2.1733 ++		rc->code -= rc->range;
  2.1734 ++		mask = (uint32_t)0 - (rc->code >> 31);
  2.1735 ++		rc->code += rc->range & mask;
  2.1736 ++		*dest = (*dest << 1) + (mask + 1);
  2.1737 ++	} while (--limit > 0);
  2.1738 ++}
  2.1739 ++
  2.1740 ++/********
  2.1741 ++ * LZMA *
  2.1742 ++ ********/
  2.1743 ++
  2.1744 ++/* Get pointer to literal coder probability array. */
  2.1745 ++static uint16_t *lzma_literal_probs(struct xz_dec_lzma2 *s)
  2.1746 ++{
  2.1747 ++	uint32_t prev_byte = dict_get(&s->dict, 0);
  2.1748 ++	uint32_t low = prev_byte >> (8 - s->lzma.lc);
  2.1749 ++	uint32_t high = (s->dict.pos & s->lzma.literal_pos_mask) << s->lzma.lc;
  2.1750 ++	return s->lzma.literal[low + high];
  2.1751 ++}
  2.1752 ++
  2.1753 ++/* Decode a literal (one 8-bit byte) */
  2.1754 ++static void lzma_literal(struct xz_dec_lzma2 *s)
  2.1755 ++{
  2.1756 ++	uint16_t *probs;
  2.1757 ++	uint32_t symbol;
  2.1758 ++	uint32_t match_byte;
  2.1759 ++	uint32_t match_bit;
  2.1760 ++	uint32_t offset;
  2.1761 ++	uint32_t i;
  2.1762 ++
  2.1763 ++	probs = lzma_literal_probs(s);
  2.1764 ++
  2.1765 ++	if (lzma_state_is_literal(s->lzma.state)) {
  2.1766 ++		symbol = rc_bittree(&s->rc, probs, 0x100);
  2.1767 ++	} else {
  2.1768 ++		symbol = 1;
  2.1769 ++		match_byte = dict_get(&s->dict, s->lzma.rep0) << 1;
  2.1770 ++		offset = 0x100;
  2.1771 ++
  2.1772 ++		do {
  2.1773 ++			match_bit = match_byte & offset;
  2.1774 ++			match_byte <<= 1;
  2.1775 ++			i = offset + match_bit + symbol;
  2.1776 ++
  2.1777 ++			if (rc_bit(&s->rc, &probs[i])) {
  2.1778 ++				symbol = (symbol << 1) + 1;
  2.1779 ++				offset &= match_bit;
  2.1780 ++			} else {
  2.1781 ++				symbol <<= 1;
  2.1782 ++				offset &= ~match_bit;
  2.1783 ++			}
  2.1784 ++		} while (symbol < 0x100);
  2.1785 ++	}
  2.1786 ++
  2.1787 ++	dict_put(&s->dict, (uint8_t)symbol);
  2.1788 ++	lzma_state_literal(&s->lzma.state);
  2.1789 ++}
  2.1790 ++
  2.1791 ++/* Decode the length of the match into s->lzma.len. */
  2.1792 ++static void lzma_len(struct xz_dec_lzma2 *s, struct lzma_len_dec *l,
  2.1793 ++		     uint32_t pos_state)
  2.1794 ++{
  2.1795 ++	uint16_t *probs;
  2.1796 ++	uint32_t limit;
  2.1797 ++
  2.1798 ++	if (!rc_bit(&s->rc, &l->choice)) {
  2.1799 ++		probs = l->low[pos_state];
  2.1800 ++		limit = LEN_LOW_SYMBOLS;
  2.1801 ++		s->lzma.len = MATCH_LEN_MIN;
  2.1802 ++	} else {
  2.1803 ++		if (!rc_bit(&s->rc, &l->choice2)) {
  2.1804 ++			probs = l->mid[pos_state];
  2.1805 ++			limit = LEN_MID_SYMBOLS;
  2.1806 ++			s->lzma.len = MATCH_LEN_MIN + LEN_LOW_SYMBOLS;
  2.1807 ++		} else {
  2.1808 ++			probs = l->high;
  2.1809 ++			limit = LEN_HIGH_SYMBOLS;
  2.1810 ++			s->lzma.len = MATCH_LEN_MIN + LEN_LOW_SYMBOLS
  2.1811 ++					+ LEN_MID_SYMBOLS;
  2.1812 ++		}
  2.1813 ++	}
  2.1814 ++
  2.1815 ++	s->lzma.len += rc_bittree(&s->rc, probs, limit) - limit;
  2.1816 ++}
  2.1817 ++
  2.1818 ++/* Decode a match. The distance will be stored in s->lzma.rep0. */
  2.1819 ++static void lzma_match(struct xz_dec_lzma2 *s, uint32_t pos_state)
  2.1820 ++{
  2.1821 ++	uint16_t *probs;
  2.1822 ++	uint32_t dist_slot;
  2.1823 ++	uint32_t limit;
  2.1824 ++
  2.1825 ++	lzma_state_match(&s->lzma.state);
  2.1826 ++
  2.1827 ++	s->lzma.rep3 = s->lzma.rep2;
  2.1828 ++	s->lzma.rep2 = s->lzma.rep1;
  2.1829 ++	s->lzma.rep1 = s->lzma.rep0;
  2.1830 ++
  2.1831 ++	lzma_len(s, &s->lzma.match_len_dec, pos_state);
  2.1832 ++
  2.1833 ++	probs = s->lzma.dist_slot[lzma_get_dist_state(s->lzma.len)];
  2.1834 ++	dist_slot = rc_bittree(&s->rc, probs, DIST_SLOTS) - DIST_SLOTS;
  2.1835 ++
  2.1836 ++	if (dist_slot < DIST_MODEL_START) {
  2.1837 ++		s->lzma.rep0 = dist_slot;
  2.1838 ++	} else {
  2.1839 ++		limit = (dist_slot >> 1) - 1;
  2.1840 ++		s->lzma.rep0 = 2 + (dist_slot & 1);
  2.1841 ++
  2.1842 ++		if (dist_slot < DIST_MODEL_END) {
  2.1843 ++			s->lzma.rep0 <<= limit;
  2.1844 ++			probs = s->lzma.dist_special + s->lzma.rep0
  2.1845 ++					- dist_slot - 1;
  2.1846 ++			rc_bittree_reverse(&s->rc, probs,
  2.1847 ++					&s->lzma.rep0, limit);
  2.1848 ++		} else {
  2.1849 ++			rc_direct(&s->rc, &s->lzma.rep0, limit - ALIGN_BITS);
  2.1850 ++			s->lzma.rep0 <<= ALIGN_BITS;
  2.1851 ++			rc_bittree_reverse(&s->rc, s->lzma.dist_align,
  2.1852 ++					&s->lzma.rep0, ALIGN_BITS);
  2.1853 ++		}
  2.1854 ++	}
  2.1855 ++}
  2.1856 ++
  2.1857 ++/*
  2.1858 ++ * Decode a repeated match. The distance is one of the four most recently
  2.1859 ++ * seen matches. The distance will be stored in s->lzma.rep0.
  2.1860 ++ */
  2.1861 ++static void lzma_rep_match(struct xz_dec_lzma2 *s, uint32_t pos_state)
  2.1862 ++{
  2.1863 ++	uint32_t tmp;
  2.1864 ++
  2.1865 ++	if (!rc_bit(&s->rc, &s->lzma.is_rep0[s->lzma.state])) {
  2.1866 ++		if (!rc_bit(&s->rc, &s->lzma.is_rep0_long[
  2.1867 ++				s->lzma.state][pos_state])) {
  2.1868 ++			lzma_state_short_rep(&s->lzma.state);
  2.1869 ++			s->lzma.len = 1;
  2.1870 ++			return;
  2.1871 ++		}
  2.1872 ++	} else {
  2.1873 ++		if (!rc_bit(&s->rc, &s->lzma.is_rep1[s->lzma.state])) {
  2.1874 ++			tmp = s->lzma.rep1;
  2.1875 ++		} else {
  2.1876 ++			if (!rc_bit(&s->rc, &s->lzma.is_rep2[s->lzma.state])) {
  2.1877 ++				tmp = s->lzma.rep2;
  2.1878 ++			} else {
  2.1879 ++				tmp = s->lzma.rep3;
  2.1880 ++				s->lzma.rep3 = s->lzma.rep2;
  2.1881 ++			}
  2.1882 ++
  2.1883 ++			s->lzma.rep2 = s->lzma.rep1;
  2.1884 ++		}
  2.1885 ++
  2.1886 ++		s->lzma.rep1 = s->lzma.rep0;
  2.1887 ++		s->lzma.rep0 = tmp;
  2.1888 ++	}
  2.1889 ++
  2.1890 ++	lzma_state_long_rep(&s->lzma.state);
  2.1891 ++	lzma_len(s, &s->lzma.rep_len_dec, pos_state);
  2.1892 ++}
  2.1893 ++
  2.1894 ++/* LZMA decoder core */
  2.1895 ++static bool lzma_main(struct xz_dec_lzma2 *s)
  2.1896 ++{
  2.1897 ++	uint32_t pos_state;
  2.1898 ++
  2.1899 ++	/*
  2.1900 ++	 * If the dictionary was reached during the previous call, try to
  2.1901 ++	 * finish the possibly pending repeat in the dictionary.
  2.1902 ++	 */
  2.1903 ++	if (dict_has_space(&s->dict) && s->lzma.len > 0)
  2.1904 ++		dict_repeat(&s->dict, &s->lzma.len, s->lzma.rep0);
  2.1905 ++
  2.1906 ++	/*
  2.1907 ++	 * Decode more LZMA symbols. One iteration may consume up to
  2.1908 ++	 * LZMA_IN_REQUIRED - 1 bytes.
  2.1909 ++	 */
  2.1910 ++	while (dict_has_space(&s->dict) && !rc_limit_exceeded(&s->rc)) {
  2.1911 ++		pos_state = s->dict.pos & s->lzma.pos_mask;
  2.1912 ++
  2.1913 ++		if (!rc_bit(&s->rc, &s->lzma.is_match[
  2.1914 ++				s->lzma.state][pos_state])) {
  2.1915 ++			lzma_literal(s);
  2.1916 ++		} else {
  2.1917 ++			if (rc_bit(&s->rc, &s->lzma.is_rep[s->lzma.state]))
  2.1918 ++				lzma_rep_match(s, pos_state);
  2.1919 ++			else
  2.1920 ++				lzma_match(s, pos_state);
  2.1921 ++
  2.1922 ++			if (!dict_repeat(&s->dict, &s->lzma.len, s->lzma.rep0))
  2.1923 ++				return false;
  2.1924 ++		}
  2.1925 ++	}
  2.1926 ++
  2.1927 ++	/*
  2.1928 ++	 * Having the range decoder always normalized when we are outside
  2.1929 ++	 * this function makes it easier to correctly handle end of the chunk.
  2.1930 ++	 */
  2.1931 ++	rc_normalize(&s->rc);
  2.1932 ++
  2.1933 ++	return true;
  2.1934 ++}
  2.1935 ++
  2.1936 ++/*
  2.1937 ++ * Reset the LZMA decoder and range decoder state. Dictionary is nore reset
  2.1938 ++ * here, because LZMA state may be reset without resetting the dictionary.
  2.1939 ++ */
  2.1940 ++static void lzma_reset(struct xz_dec_lzma2 *s)
  2.1941 ++{
  2.1942 ++	uint16_t *probs;
  2.1943 ++	size_t i;
  2.1944 ++
  2.1945 ++	s->lzma.state = STATE_LIT_LIT;
  2.1946 ++	s->lzma.rep0 = 0;
  2.1947 ++	s->lzma.rep1 = 0;
  2.1948 ++	s->lzma.rep2 = 0;
  2.1949 ++	s->lzma.rep3 = 0;
  2.1950 ++
  2.1951 ++	/*
  2.1952 ++	 * All probabilities are initialized to the same value. This hack
  2.1953 ++	 * makes the code smaller by avoiding a separate loop for each
  2.1954 ++	 * probability array.
  2.1955 ++	 *
  2.1956 ++	 * This could be optimized so that only that part of literal
  2.1957 ++	 * probabilities that are actually required. In the common case
  2.1958 ++	 * we would write 12 KiB less.
  2.1959 ++	 */
  2.1960 ++	probs = s->lzma.is_match[0];
  2.1961 ++	for (i = 0; i < PROBS_TOTAL; ++i)
  2.1962 ++		probs[i] = RC_BIT_MODEL_TOTAL / 2;
  2.1963 ++
  2.1964 ++	rc_reset(&s->rc);
  2.1965 ++}
  2.1966 ++
  2.1967 ++/*
  2.1968 ++ * Decode and validate LZMA properties (lc/lp/pb) and calculate the bit masks
  2.1969 ++ * from the decoded lp and pb values. On success, the LZMA decoder state is
  2.1970 ++ * reset and true is returned.
  2.1971 ++ */
  2.1972 ++static bool lzma_props(struct xz_dec_lzma2 *s, uint8_t props)
  2.1973 ++{
  2.1974 ++	if (props > (4 * 5 + 4) * 9 + 8)
  2.1975 ++		return false;
  2.1976 ++
  2.1977 ++	s->lzma.pos_mask = 0;
  2.1978 ++	while (props >= 9 * 5) {
  2.1979 ++		props -= 9 * 5;
  2.1980 ++		++s->lzma.pos_mask;
  2.1981 ++	}
  2.1982 ++
  2.1983 ++	s->lzma.pos_mask = (1 << s->lzma.pos_mask) - 1;
  2.1984 ++
  2.1985 ++	s->lzma.literal_pos_mask = 0;
  2.1986 ++	while (props >= 9) {
  2.1987 ++		props -= 9;
  2.1988 ++		++s->lzma.literal_pos_mask;
  2.1989 ++	}
  2.1990 ++
  2.1991 ++	s->lzma.lc = props;
  2.1992 ++
  2.1993 ++	if (s->lzma.lc + s->lzma.literal_pos_mask > 4)
  2.1994 ++		return false;
  2.1995 ++
  2.1996 ++	s->lzma.literal_pos_mask = (1 << s->lzma.literal_pos_mask) - 1;
  2.1997 ++
  2.1998 ++	lzma_reset(s);
  2.1999 ++
  2.2000 ++	return true;
  2.2001 ++}
  2.2002 ++
  2.2003 ++/*********
  2.2004 ++ * LZMA2 *
  2.2005 ++ *********/
  2.2006 ++
  2.2007 ++/*
  2.2008 ++ * The LZMA decoder assumes that if the input limit (s->rc.in_limit) hasn't
  2.2009 ++ * been exceeded, it is safe to read up to LZMA_IN_REQUIRED bytes. This
  2.2010 ++ * wrapper function takes care of making the LZMA decoder's assumption safe.
  2.2011 ++ *
  2.2012 ++ * As long as there is plenty of input left to be decoded in the current LZMA
  2.2013 ++ * chunk, we decode directly from the caller-supplied input buffer until
  2.2014 ++ * there's LZMA_IN_REQUIRED bytes left. Those remaining bytes are copied into
  2.2015 ++ * s->temp.buf, which (hopefully) gets filled on the next call to this
  2.2016 ++ * function. We decode a few bytes from the temporary buffer so that we can
  2.2017 ++ * continue decoding from the caller-supplied input buffer again.
  2.2018 ++ */
  2.2019 ++static bool lzma2_lzma(struct xz_dec_lzma2 *s, struct xz_buf *b)
  2.2020 ++{
  2.2021 ++	size_t in_avail;
  2.2022 ++	uint32_t tmp;
  2.2023 ++
  2.2024 ++	in_avail = b->in_size - b->in_pos;
  2.2025 ++	if (s->temp.size > 0 || s->lzma2.compressed == 0) {
  2.2026 ++		tmp = 2 * LZMA_IN_REQUIRED - s->temp.size;
  2.2027 ++		if (tmp > s->lzma2.compressed - s->temp.size)
  2.2028 ++			tmp = s->lzma2.compressed - s->temp.size;
  2.2029 ++		if (tmp > in_avail)
  2.2030 ++			tmp = in_avail;
  2.2031 ++
  2.2032 ++		memcpy(s->temp.buf + s->temp.size, b->in + b->in_pos, tmp);
  2.2033 ++
  2.2034 ++		if (s->temp.size + tmp == s->lzma2.compressed) {
  2.2035 ++			memzero(s->temp.buf + s->temp.size + tmp,
  2.2036 ++					sizeof(s->temp.buf)
  2.2037 ++						- s->temp.size - tmp);
  2.2038 ++			s->rc.in_limit = s->temp.size + tmp;
  2.2039 ++		} else if (s->temp.size + tmp < LZMA_IN_REQUIRED) {
  2.2040 ++			s->temp.size += tmp;
  2.2041 ++			b->in_pos += tmp;
  2.2042 ++			return true;
  2.2043 ++		} else {
  2.2044 ++			s->rc.in_limit = s->temp.size + tmp - LZMA_IN_REQUIRED;
  2.2045 ++		}
  2.2046 ++
  2.2047 ++		s->rc.in = s->temp.buf;
  2.2048 ++		s->rc.in_pos = 0;
  2.2049 ++
  2.2050 ++		if (!lzma_main(s) || s->rc.in_pos > s->temp.size + tmp)
  2.2051 ++			return false;
  2.2052 ++
  2.2053 ++		s->lzma2.compressed -= s->rc.in_pos;
  2.2054 ++
  2.2055 ++		if (s->rc.in_pos < s->temp.size) {
  2.2056 ++			s->temp.size -= s->rc.in_pos;
  2.2057 ++			memmove(s->temp.buf, s->temp.buf + s->rc.in_pos,
  2.2058 ++					s->temp.size);
  2.2059 ++			return true;
  2.2060 ++		}
  2.2061 ++
  2.2062 ++		b->in_pos += s->rc.in_pos - s->temp.size;
  2.2063 ++		s->temp.size = 0;
  2.2064 ++	}
  2.2065 ++
  2.2066 ++	in_avail = b->in_size - b->in_pos;
  2.2067 ++	if (in_avail >= LZMA_IN_REQUIRED) {
  2.2068 ++		s->rc.in = b->in;
  2.2069 ++		s->rc.in_pos = b->in_pos;
  2.2070 ++
  2.2071 ++		if (in_avail >= s->lzma2.compressed + LZMA_IN_REQUIRED)
  2.2072 ++			s->rc.in_limit = b->in_pos + s->lzma2.compressed;
  2.2073 ++		else
  2.2074 ++			s->rc.in_limit = b->in_size - LZMA_IN_REQUIRED;
  2.2075 ++
  2.2076 ++		if (!lzma_main(s))
  2.2077 ++			return false;
  2.2078 ++
  2.2079 ++		in_avail = s->rc.in_pos - b->in_pos;
  2.2080 ++		if (in_avail > s->lzma2.compressed)
  2.2081 ++			return false;
  2.2082 ++
  2.2083 ++		s->lzma2.compressed -= in_avail;
  2.2084 ++		b->in_pos = s->rc.in_pos;
  2.2085 ++	}
  2.2086 ++
  2.2087 ++	in_avail = b->in_size - b->in_pos;
  2.2088 ++	if (in_avail < LZMA_IN_REQUIRED) {
  2.2089 ++		if (in_avail > s->lzma2.compressed)
  2.2090 ++			in_avail = s->lzma2.compressed;
  2.2091 ++
  2.2092 ++		memcpy(s->temp.buf, b->in + b->in_pos, in_avail);
  2.2093 ++		s->temp.size = in_avail;
  2.2094 ++		b->in_pos += in_avail;
  2.2095 ++	}
  2.2096 ++
  2.2097 ++	return true;
  2.2098 ++}
  2.2099 ++
  2.2100 ++/*
  2.2101 ++ * Take care of the LZMA2 control layer, and forward the job of actual LZMA
  2.2102 ++ * decoding or copying of uncompressed chunks to other functions.
  2.2103 ++ */
  2.2104 ++XZ_EXTERN enum xz_ret xz_dec_lzma2_run(struct xz_dec_lzma2 *s,
  2.2105 ++				       struct xz_buf *b)
  2.2106 ++{
  2.2107 ++	uint32_t tmp;
  2.2108 ++
  2.2109 ++	while (b->in_pos < b->in_size || s->lzma2.sequence == SEQ_LZMA_RUN) {
  2.2110 ++		switch (s->lzma2.sequence) {
  2.2111 ++		case SEQ_CONTROL:
  2.2112 ++			/*
  2.2113 ++			 * LZMA2 control byte
  2.2114 ++			 *
  2.2115 ++			 * Exact values:
  2.2116 ++			 *   0x00   End marker
  2.2117 ++			 *   0x01   Dictionary reset followed by
  2.2118 ++			 *          an uncompressed chunk
  2.2119 ++			 *   0x02   Uncompressed chunk (no dictionary reset)
  2.2120 ++			 *
  2.2121 ++			 * Highest three bits (s->control & 0xE0):
  2.2122 ++			 *   0xE0   Dictionary reset, new properties and state
  2.2123 ++			 *          reset, followed by LZMA compressed chunk
  2.2124 ++			 *   0xC0   New properties and state reset, followed
  2.2125 ++			 *          by LZMA compressed chunk (no dictionary
  2.2126 ++			 *          reset)
  2.2127 ++			 *   0xA0   State reset using old properties,
  2.2128 ++			 *          followed by LZMA compressed chunk (no
  2.2129 ++			 *          dictionary reset)
  2.2130 ++			 *   0x80   LZMA chunk (no dictionary or state reset)
  2.2131 ++			 *
  2.2132 ++			 * For LZMA compressed chunks, the lowest five bits
  2.2133 ++			 * (s->control & 1F) are the highest bits of the
  2.2134 ++			 * uncompressed size (bits 16-20).
  2.2135 ++			 *
  2.2136 ++			 * A new LZMA2 stream must begin with a dictionary
  2.2137 ++			 * reset. The first LZMA chunk must set new
  2.2138 ++			 * properties and reset the LZMA state.
  2.2139 ++			 *
  2.2140 ++			 * Values that don't match anything described above
  2.2141 ++			 * are invalid and we return XZ_DATA_ERROR.
  2.2142 ++			 */
  2.2143 ++			tmp = b->in[b->in_pos++];
  2.2144 ++
  2.2145 ++			if (tmp >= 0xE0 || tmp == 0x01) {
  2.2146 ++				s->lzma2.need_props = true;
  2.2147 ++				s->lzma2.need_dict_reset = false;
  2.2148 ++				dict_reset(&s->dict, b);
  2.2149 ++			} else if (s->lzma2.need_dict_reset) {
  2.2150 ++				return XZ_DATA_ERROR;
  2.2151 ++			}
  2.2152 ++
  2.2153 ++			if (tmp >= 0x80) {
  2.2154 ++				s->lzma2.uncompressed = (tmp & 0x1F) << 16;
  2.2155 ++				s->lzma2.sequence = SEQ_UNCOMPRESSED_1;
  2.2156 ++
  2.2157 ++				if (tmp >= 0xC0) {
  2.2158 ++					/*
  2.2159 ++					 * When there are new properties,
  2.2160 ++					 * state reset is done at
  2.2161 ++					 * SEQ_PROPERTIES.
  2.2162 ++					 */
  2.2163 ++					s->lzma2.need_props = false;
  2.2164 ++					s->lzma2.next_sequence
  2.2165 ++							= SEQ_PROPERTIES;
  2.2166 ++
  2.2167 ++				} else if (s->lzma2.need_props) {
  2.2168 ++					return XZ_DATA_ERROR;
  2.2169 ++
  2.2170 ++				} else {
  2.2171 ++					s->lzma2.next_sequence
  2.2172 ++							= SEQ_LZMA_PREPARE;
  2.2173 ++					if (tmp >= 0xA0)
  2.2174 ++						lzma_reset(s);
  2.2175 ++				}
  2.2176 ++			} else {
  2.2177 ++				if (tmp == 0x00)
  2.2178 ++					return XZ_STREAM_END;
  2.2179 ++
  2.2180 ++				if (tmp > 0x02)
  2.2181 ++					return XZ_DATA_ERROR;
  2.2182 ++
  2.2183 ++				s->lzma2.sequence = SEQ_COMPRESSED_0;
  2.2184 ++				s->lzma2.next_sequence = SEQ_COPY;
  2.2185 ++			}
  2.2186 ++
  2.2187 ++			break;
  2.2188 ++
  2.2189 ++		case SEQ_UNCOMPRESSED_1:
  2.2190 ++			s->lzma2.uncompressed
  2.2191 ++					+= (uint32_t)b->in[b->in_pos++] << 8;
  2.2192 ++			s->lzma2.sequence = SEQ_UNCOMPRESSED_2;
  2.2193 ++			break;
  2.2194 ++
  2.2195 ++		case SEQ_UNCOMPRESSED_2:
  2.2196 ++			s->lzma2.uncompressed
  2.2197 ++					+= (uint32_t)b->in[b->in_pos++] + 1;
  2.2198 ++			s->lzma2.sequence = SEQ_COMPRESSED_0;
  2.2199 ++			break;
  2.2200 ++
  2.2201 ++		case SEQ_COMPRESSED_0:
  2.2202 ++			s->lzma2.compressed
  2.2203 ++					= (uint32_t)b->in[b->in_pos++] << 8;
  2.2204 ++			s->lzma2.sequence = SEQ_COMPRESSED_1;
  2.2205 ++			break;
  2.2206 ++
  2.2207 ++		case SEQ_COMPRESSED_1:
  2.2208 ++			s->lzma2.compressed
  2.2209 ++					+= (uint32_t)b->in[b->in_pos++] + 1;
  2.2210 ++			s->lzma2.sequence = s->lzma2.next_sequence;
  2.2211 ++			break;
  2.2212 ++
  2.2213 ++		case SEQ_PROPERTIES:
  2.2214 ++			if (!lzma_props(s, b->in[b->in_pos++]))
  2.2215 ++				return XZ_DATA_ERROR;
  2.2216 ++
  2.2217 ++			s->lzma2.sequence = SEQ_LZMA_PREPARE;
  2.2218 ++
  2.2219 ++		case SEQ_LZMA_PREPARE:
  2.2220 ++			if (s->lzma2.compressed < RC_INIT_BYTES)
  2.2221 ++				return XZ_DATA_ERROR;
  2.2222 ++
  2.2223 ++			if (!rc_read_init(&s->rc, b))
  2.2224 ++				return XZ_OK;
  2.2225 ++
  2.2226 ++			s->lzma2.compressed -= RC_INIT_BYTES;
  2.2227 ++			s->lzma2.sequence = SEQ_LZMA_RUN;
  2.2228 ++
  2.2229 ++		case SEQ_LZMA_RUN:
  2.2230 ++			/*
  2.2231 ++			 * Set dictionary limit to indicate how much we want
  2.2232 ++			 * to be encoded at maximum. Decode new data into the
  2.2233 ++			 * dictionary. Flush the new data from dictionary to
  2.2234 ++			 * b->out. Check if we finished decoding this chunk.
  2.2235 ++			 * In case the dictionary got full but we didn't fill
  2.2236 ++			 * the output buffer yet, we may run this loop
  2.2237 ++			 * multiple times without changing s->lzma2.sequence.
  2.2238 ++			 */
  2.2239 ++			dict_limit(&s->dict, min_t(size_t,
  2.2240 ++					b->out_size - b->out_pos,
  2.2241 ++					s->lzma2.uncompressed));
  2.2242 ++			if (!lzma2_lzma(s, b))
  2.2243 ++				return XZ_DATA_ERROR;
  2.2244 ++
  2.2245 ++			s->lzma2.uncompressed -= dict_flush(&s->dict, b);
  2.2246 ++
  2.2247 ++			if (s->lzma2.uncompressed == 0) {
  2.2248 ++				if (s->lzma2.compressed > 0 || s->lzma.len > 0
  2.2249 ++						|| !rc_is_finished(&s->rc))
  2.2250 ++					return XZ_DATA_ERROR;
  2.2251 ++
  2.2252 ++				rc_reset(&s->rc);
  2.2253 ++				s->lzma2.sequence = SEQ_CONTROL;
  2.2254 ++
  2.2255 ++			} else if (b->out_pos == b->out_size
  2.2256 ++					|| (b->in_pos == b->in_size
  2.2257 ++						&& s->temp.size
  2.2258 ++						< s->lzma2.compressed)) {
  2.2259 ++				return XZ_OK;
  2.2260 ++			}
  2.2261 ++
  2.2262 ++			break;
  2.2263 ++
  2.2264 ++		case SEQ_COPY:
  2.2265 ++			dict_uncompressed(&s->dict, b, &s->lzma2.compressed);
  2.2266 ++			if (s->lzma2.compressed > 0)
  2.2267 ++				return XZ_OK;
  2.2268 ++
  2.2269 ++			s->lzma2.sequence = SEQ_CONTROL;
  2.2270 ++			break;
  2.2271 ++		}
  2.2272 ++	}
  2.2273 ++
  2.2274 ++	return XZ_OK;
  2.2275 ++}
  2.2276 ++
  2.2277 ++XZ_EXTERN struct xz_dec_lzma2 *xz_dec_lzma2_create(enum xz_mode mode,
  2.2278 ++						   uint32_t dict_max)
  2.2279 ++{
  2.2280 ++	struct xz_dec_lzma2 *s = kmalloc(sizeof(*s), GFP_KERNEL);
  2.2281 ++	if (s == NULL)
  2.2282 ++		return NULL;
  2.2283 ++
  2.2284 ++	s->dict.mode = mode;
  2.2285 ++	s->dict.size_max = dict_max;
  2.2286 ++
  2.2287 ++	if (DEC_IS_PREALLOC(mode)) {
  2.2288 ++		s->dict.buf = vmalloc(dict_max);
  2.2289 ++		if (s->dict.buf == NULL) {
  2.2290 ++			kfree(s);
  2.2291 ++			return NULL;
  2.2292 ++		}
  2.2293 ++	} else if (DEC_IS_DYNALLOC(mode)) {
  2.2294 ++		s->dict.buf = NULL;
  2.2295 ++		s->dict.allocated = 0;
  2.2296 ++	}
  2.2297 ++
  2.2298 ++	return s;
  2.2299 ++}
  2.2300 ++
  2.2301 ++XZ_EXTERN enum xz_ret xz_dec_lzma2_reset(struct xz_dec_lzma2 *s, uint8_t props)
  2.2302 ++{
  2.2303 ++	/* This limits dictionary size to 3 GiB to keep parsing simpler. */
  2.2304 ++	if (props > 39)
  2.2305 ++		return XZ_OPTIONS_ERROR;
  2.2306 ++
  2.2307 ++	s->dict.size = 2 + (props & 1);
  2.2308 ++	s->dict.size <<= (props >> 1) + 11;
  2.2309 ++
  2.2310 ++	if (DEC_IS_MULTI(s->dict.mode)) {
  2.2311 ++		if (s->dict.size > s->dict.size_max)
  2.2312 ++			return XZ_MEMLIMIT_ERROR;
  2.2313 ++
  2.2314 ++		s->dict.end = s->dict.size;
  2.2315 ++
  2.2316 ++		if (DEC_IS_DYNALLOC(s->dict.mode)) {
  2.2317 ++			if (s->dict.allocated < s->dict.size) {
  2.2318 ++				vfree(s->dict.buf);
  2.2319 ++				s->dict.buf = vmalloc(s->dict.size);
  2.2320 ++				if (s->dict.buf == NULL) {
  2.2321 ++					s->dict.allocated = 0;
  2.2322 ++					return XZ_MEM_ERROR;
  2.2323 ++				}
  2.2324 ++			}
  2.2325 ++		}
  2.2326 ++	}
  2.2327 ++
  2.2328 ++	s->lzma.len = 0;
  2.2329 ++
  2.2330 ++	s->lzma2.sequence = SEQ_CONTROL;
  2.2331 ++	s->lzma2.need_dict_reset = true;
  2.2332 ++
  2.2333 ++	s->temp.size = 0;
  2.2334 ++
  2.2335 ++	return XZ_OK;
  2.2336 ++}
  2.2337 ++
  2.2338 ++XZ_EXTERN void xz_dec_lzma2_end(struct xz_dec_lzma2 *s)
  2.2339 ++{
  2.2340 ++	if (DEC_IS_MULTI(s->dict.mode))
  2.2341 ++		vfree(s->dict.buf);
  2.2342 ++
  2.2343 ++	kfree(s);
  2.2344 ++}
  2.2345 +diff --git a/lib/xz/xz_dec_stream.c b/lib/xz/xz_dec_stream.c
  2.2346 +new file mode 100644
  2.2347 +index 0000000..ac809b1
  2.2348 +--- /dev/null
  2.2349 ++++ b/lib/xz/xz_dec_stream.c
  2.2350 +@@ -0,0 +1,821 @@
  2.2351 ++/*
  2.2352 ++ * .xz Stream decoder
  2.2353 ++ *
  2.2354 ++ * Author: Lasse Collin <lasse.collin@tukaani.org>
  2.2355 ++ *
  2.2356 ++ * This file has been put into the public domain.
  2.2357 ++ * You can do whatever you want with this file.
  2.2358 ++ */
  2.2359 ++
  2.2360 ++#include "xz_private.h"
  2.2361 ++#include "xz_stream.h"
  2.2362 ++
  2.2363 ++/* Hash used to validate the Index field */
  2.2364 ++struct xz_dec_hash {
  2.2365 ++	vli_type unpadded;
  2.2366 ++	vli_type uncompressed;
  2.2367 ++	uint32_t crc32;
  2.2368 ++};
  2.2369 ++
  2.2370 ++struct xz_dec {
  2.2371 ++	/* Position in dec_main() */
  2.2372 ++	enum {
  2.2373 ++		SEQ_STREAM_HEADER,
  2.2374 ++		SEQ_BLOCK_START,
  2.2375 ++		SEQ_BLOCK_HEADER,
  2.2376 ++		SEQ_BLOCK_UNCOMPRESS,
  2.2377 ++		SEQ_BLOCK_PADDING,
  2.2378 ++		SEQ_BLOCK_CHECK,
  2.2379 ++		SEQ_INDEX,
  2.2380 ++		SEQ_INDEX_PADDING,
  2.2381 ++		SEQ_INDEX_CRC32,
  2.2382 ++		SEQ_STREAM_FOOTER
  2.2383 ++	} sequence;
  2.2384 ++
  2.2385 ++	/* Position in variable-length integers and Check fields */
  2.2386 ++	uint32_t pos;
  2.2387 ++
  2.2388 ++	/* Variable-length integer decoded by dec_vli() */
  2.2389 ++	vli_type vli;
  2.2390 ++
  2.2391 ++	/* Saved in_pos and out_pos */
  2.2392 ++	size_t in_start;
  2.2393 ++	size_t out_start;
  2.2394 ++
  2.2395 ++	/* CRC32 value in Block or Index */
  2.2396 ++	uint32_t crc32;
  2.2397 ++
  2.2398 ++	/* Type of the integrity check calculated from uncompressed data */
  2.2399 ++	enum xz_check check_type;
  2.2400 ++
  2.2401 ++	/* Operation mode */
  2.2402 ++	enum xz_mode mode;
  2.2403 ++
  2.2404 ++	/*
  2.2405 ++	 * True if the next call to xz_dec_run() is allowed to return
  2.2406 ++	 * XZ_BUF_ERROR.
  2.2407 ++	 */
  2.2408 ++	bool allow_buf_error;
  2.2409 ++
  2.2410 ++	/* Information stored in Block Header */
  2.2411 ++	struct {
  2.2412 ++		/*
  2.2413 ++		 * Value stored in the Compressed Size field, or
  2.2414 ++		 * VLI_UNKNOWN if Compressed Size is not present.
  2.2415 ++		 */
  2.2416 ++		vli_type compressed;
  2.2417 ++
  2.2418 ++		/*
  2.2419 ++		 * Value stored in the Uncompressed Size field, or
  2.2420 ++		 * VLI_UNKNOWN if Uncompressed Size is not present.
  2.2421 ++		 */
  2.2422 ++		vli_type uncompressed;
  2.2423 ++
  2.2424 ++		/* Size of the Block Header field */
  2.2425 ++		uint32_t size;
  2.2426 ++	} block_header;
  2.2427 ++
  2.2428 ++	/* Information collected when decoding Blocks */
  2.2429 ++	struct {
  2.2430 ++		/* Observed compressed size of the current Block */
  2.2431 ++		vli_type compressed;
  2.2432 ++
  2.2433 ++		/* Observed uncompressed size of the current Block */
  2.2434 ++		vli_type uncompressed;
  2.2435 ++
  2.2436 ++		/* Number of Blocks decoded so far */
  2.2437 ++		vli_type count;
  2.2438 ++
  2.2439 ++		/*
  2.2440 ++		 * Hash calculated from the Block sizes. This is used to
  2.2441 ++		 * validate the Index field.
  2.2442 ++		 */
  2.2443 ++		struct xz_dec_hash hash;
  2.2444 ++	} block;
  2.2445 ++
  2.2446 ++	/* Variables needed when verifying the Index field */
  2.2447 ++	struct {
  2.2448 ++		/* Position in dec_index() */
  2.2449 ++		enum {
  2.2450 ++			SEQ_INDEX_COUNT,
  2.2451 ++			SEQ_INDEX_UNPADDED,
  2.2452 ++			SEQ_INDEX_UNCOMPRESSED
  2.2453 ++		} sequence;
  2.2454 ++
  2.2455 ++		/* Size of the Index in bytes */
  2.2456 ++		vli_type size;
  2.2457 ++
  2.2458 ++		/* Number of Records (matches block.count in valid files) */
  2.2459 ++		vli_type count;
  2.2460 ++
  2.2461 ++		/*
  2.2462 ++		 * Hash calculated from the Records (matches block.hash in
  2.2463 ++		 * valid files).
  2.2464 ++		 */
  2.2465 ++		struct xz_dec_hash hash;
  2.2466 ++	} index;
  2.2467 ++
  2.2468 ++	/*
  2.2469 ++	 * Temporary buffer needed to hold Stream Header, Block Header,
  2.2470 ++	 * and Stream Footer. The Block Header is the biggest (1 KiB)
  2.2471 ++	 * so we reserve space according to that. buf[] has to be aligned
  2.2472 ++	 * to a multiple of four bytes; the size_t variables before it
  2.2473 ++	 * should guarantee this.
  2.2474 ++	 */
  2.2475 ++	struct {
  2.2476 ++		size_t pos;
  2.2477 ++		size_t size;
  2.2478 ++		uint8_t buf[1024];
  2.2479 ++	} temp;
  2.2480 ++
  2.2481 ++	struct xz_dec_lzma2 *lzma2;
  2.2482 ++
  2.2483 ++#ifdef XZ_DEC_BCJ
  2.2484 ++	struct xz_dec_bcj *bcj;
  2.2485 ++	bool bcj_active;
  2.2486 ++#endif
  2.2487 ++};
  2.2488 ++
  2.2489 ++#ifdef XZ_DEC_ANY_CHECK
  2.2490 ++/* Sizes of the Check field with different Check IDs */
  2.2491 ++static const uint8_t check_sizes[16] = {
  2.2492 ++	0,
  2.2493 ++	4, 4, 4,
  2.2494 ++	8, 8, 8,
  2.2495 ++	16, 16, 16,
  2.2496 ++	32, 32, 32,
  2.2497 ++	64, 64, 64
  2.2498 ++};
  2.2499 ++#endif
  2.2500 ++
  2.2501 ++/*
  2.2502 ++ * Fill s->temp by copying data starting from b->in[b->in_pos]. Caller
  2.2503 ++ * must have set s->temp.pos to indicate how much data we are supposed
  2.2504 ++ * to copy into s->temp.buf. Return true once s->temp.pos has reached
  2.2505 ++ * s->temp.size.
  2.2506 ++ */
  2.2507 ++static bool fill_temp(struct xz_dec *s, struct xz_buf *b)
  2.2508 ++{
  2.2509 ++	size_t copy_size = min_t(size_t,
  2.2510 ++			b->in_size - b->in_pos, s->temp.size - s->temp.pos);
  2.2511 ++
  2.2512 ++	memcpy(s->temp.buf + s->temp.pos, b->in + b->in_pos, copy_size);
  2.2513 ++	b->in_pos += copy_size;
  2.2514 ++	s->temp.pos += copy_size;
  2.2515 ++
  2.2516 ++	if (s->temp.pos == s->temp.size) {
  2.2517 ++		s->temp.pos = 0;
  2.2518 ++		return true;
  2.2519 ++	}
  2.2520 ++
  2.2521 ++	return false;
  2.2522 ++}
  2.2523 ++
  2.2524 ++/* Decode a variable-length integer (little-endian base-128 encoding) */
  2.2525 ++static enum xz_ret dec_vli(struct xz_dec *s, const uint8_t *in,
  2.2526 ++			   size_t *in_pos, size_t in_size)
  2.2527 ++{
  2.2528 ++	uint8_t byte;
  2.2529 ++
  2.2530 ++	if (s->pos == 0)
  2.2531 ++		s->vli = 0;
  2.2532 ++
  2.2533 ++	while (*in_pos < in_size) {
  2.2534 ++		byte = in[*in_pos];
  2.2535 ++		++*in_pos;
  2.2536 ++
  2.2537 ++		s->vli |= (vli_type)(byte & 0x7F) << s->pos;
  2.2538 ++
  2.2539 ++		if ((byte & 0x80) == 0) {
  2.2540 ++			/* Don't allow non-minimal encodings. */
  2.2541 ++			if (byte == 0 && s->pos != 0)
  2.2542 ++				return XZ_DATA_ERROR;
  2.2543 ++
  2.2544 ++			s->pos = 0;
  2.2545 ++			return XZ_STREAM_END;
  2.2546 ++		}
  2.2547 ++
  2.2548 ++		s->pos += 7;
  2.2549 ++		if (s->pos == 7 * VLI_BYTES_MAX)
  2.2550 ++			return XZ_DATA_ERROR;
  2.2551 ++	}
  2.2552 ++
  2.2553 ++	return XZ_OK;
  2.2554 ++}
  2.2555 ++
  2.2556 ++/*
  2.2557 ++ * Decode the Compressed Data field from a Block. Update and validate
  2.2558 ++ * the observed compressed and uncompressed sizes of the Block so that
  2.2559 ++ * they don't exceed the values possibly stored in the Block Header
  2.2560 ++ * (validation assumes that no integer overflow occurs, since vli_type
  2.2561 ++ * is normally uint64_t). Update the CRC32 if presence of the CRC32
  2.2562 ++ * field was indicated in Stream Header.
  2.2563 ++ *
  2.2564 ++ * Once the decoding is finished, validate that the observed sizes match
  2.2565 ++ * the sizes possibly stored in the Block Header. Update the hash and
  2.2566 ++ * Block count, which are later used to validate the Index field.
  2.2567 ++ */
  2.2568 ++static enum xz_ret dec_block(struct xz_dec *s, struct xz_buf *b)
  2.2569 ++{
  2.2570 ++	enum xz_ret ret;
  2.2571 ++
  2.2572 ++	s->in_start = b->in_pos;
  2.2573 ++	s->out_start = b->out_pos;
  2.2574 ++
  2.2575 ++#ifdef XZ_DEC_BCJ
  2.2576 ++	if (s->bcj_active)
  2.2577 ++		ret = xz_dec_bcj_run(s->bcj, s->lzma2, b);
  2.2578 ++	else
  2.2579 ++#endif
  2.2580 ++		ret = xz_dec_lzma2_run(s->lzma2, b);
  2.2581 ++
  2.2582 ++	s->block.compressed += b->in_pos - s->in_start;
  2.2583 ++	s->block.uncompressed += b->out_pos - s->out_start;
  2.2584 ++
  2.2585 ++	/*
  2.2586 ++	 * There is no need to separately check for VLI_UNKNOWN, since
  2.2587 ++	 * the observed sizes are always smaller than VLI_UNKNOWN.
  2.2588 ++	 */
  2.2589 ++	if (s->block.compressed > s->block_header.compressed
  2.2590 ++			|| s->block.uncompressed
  2.2591 ++				> s->block_header.uncompressed)
  2.2592 ++		return XZ_DATA_ERROR;
  2.2593 ++
  2.2594 ++	if (s->check_type == XZ_CHECK_CRC32)
  2.2595 ++		s->crc32 = xz_crc32(b->out + s->out_start,
  2.2596 ++				b->out_pos - s->out_start, s->crc32);
  2.2597 ++
  2.2598 ++	if (ret == XZ_STREAM_END) {
  2.2599 ++		if (s->block_header.compressed != VLI_UNKNOWN
  2.2600 ++				&& s->block_header.compressed
  2.2601 ++					!= s->block.compressed)
  2.2602 ++			return XZ_DATA_ERROR;
  2.2603 ++
  2.2604 ++		if (s->block_header.uncompressed != VLI_UNKNOWN
  2.2605 ++				&& s->block_header.uncompressed
  2.2606 ++					!= s->block.uncompressed)
  2.2607 ++			return XZ_DATA_ERROR;
  2.2608 ++
  2.2609 ++		s->block.hash.unpadded += s->block_header.size
  2.2610 ++				+ s->block.compressed;
  2.2611 ++
  2.2612 ++#ifdef XZ_DEC_ANY_CHECK
  2.2613 ++		s->block.hash.unpadded += check_sizes[s->check_type];
  2.2614 ++#else
  2.2615 ++		if (s->check_type == XZ_CHECK_CRC32)
  2.2616 ++			s->block.hash.unpadded += 4;
  2.2617 ++#endif
  2.2618 ++
  2.2619 ++		s->block.hash.uncompressed += s->block.uncompressed;
  2.2620 ++		s->block.hash.crc32 = xz_crc32(
  2.2621 ++				(const uint8_t *)&s->block.hash,
  2.2622 ++				sizeof(s->block.hash), s->block.hash.crc32);
  2.2623 ++
  2.2624 ++		++s->block.count;
  2.2625 ++	}
  2.2626 ++
  2.2627 ++	return ret;
  2.2628 ++}
  2.2629 ++
  2.2630 ++/* Update the Index size and the CRC32 value. */
  2.2631 ++static void index_update(struct xz_dec *s, const struct xz_buf *b)
  2.2632 ++{
  2.2633 ++	size_t in_used = b->in_pos - s->in_start;
  2.2634 ++	s->index.size += in_used;
  2.2635 ++	s->crc32 = xz_crc32(b->in + s->in_start, in_used, s->crc32);
  2.2636 ++}
  2.2637 ++
  2.2638 ++/*
  2.2639 ++ * Decode the Number of Records, Unpadded Size, and Uncompressed Size
  2.2640 ++ * fields from the Index field. That is, Index Padding and CRC32 are not
  2.2641 ++ * decoded by this function.
  2.2642 ++ *
  2.2643 ++ * This can return XZ_OK (more input needed), XZ_STREAM_END (everything
  2.2644 ++ * successfully decoded), or XZ_DATA_ERROR (input is corrupt).
  2.2645 ++ */
  2.2646 ++static enum xz_ret dec_index(struct xz_dec *s, struct xz_buf *b)
  2.2647 ++{
  2.2648 ++	enum xz_ret ret;
  2.2649 ++
  2.2650 ++	do {
  2.2651 ++		ret = dec_vli(s, b->in, &b->in_pos, b->in_size);
  2.2652 ++		if (ret != XZ_STREAM_END) {
  2.2653 ++			index_update(s, b);
  2.2654 ++			return ret;
  2.2655 ++		}
  2.2656 ++
  2.2657 ++		switch (s->index.sequence) {
  2.2658 ++		case SEQ_INDEX_COUNT:
  2.2659 ++			s->index.count = s->vli;
  2.2660 ++
  2.2661 ++			/*
  2.2662 ++			 * Validate that the Number of Records field
  2.2663 ++			 * indicates the same number of Records as
  2.2664 ++			 * there were Blocks in the Stream.
  2.2665 ++			 */
  2.2666 ++			if (s->index.count != s->block.count)
  2.2667 ++				return XZ_DATA_ERROR;
  2.2668 ++
  2.2669 ++			s->index.sequence = SEQ_INDEX_UNPADDED;
  2.2670 ++			break;
  2.2671 ++
  2.2672 ++		case SEQ_INDEX_UNPADDED:
  2.2673 ++			s->index.hash.unpadded += s->vli;
  2.2674 ++			s->index.sequence = SEQ_INDEX_UNCOMPRESSED;
  2.2675 ++			break;
  2.2676 ++
  2.2677 ++		case SEQ_INDEX_UNCOMPRESSED:
  2.2678 ++			s->index.hash.uncompressed += s->vli;
  2.2679 ++			s->index.hash.crc32 = xz_crc32(
  2.2680 ++					(const uint8_t *)&s->index.hash,
  2.2681 ++					sizeof(s->index.hash),
  2.2682 ++					s->index.hash.crc32);
  2.2683 ++			--s->index.count;
  2.2684 ++			s->index.sequence = SEQ_INDEX_UNPADDED;
  2.2685 ++			break;
  2.2686 ++		}
  2.2687 ++	} while (s->index.count > 0);
  2.2688 ++
  2.2689 ++	return XZ_STREAM_END;
  2.2690 ++}
  2.2691 ++
  2.2692 ++/*
  2.2693 ++ * Validate that the next four input bytes match the value of s->crc32.
  2.2694 ++ * s->pos must be zero when starting to validate the first byte.
  2.2695 ++ */
  2.2696 ++static enum xz_ret crc32_validate(struct xz_dec *s, struct xz_buf *b)
  2.2697 ++{
  2.2698 ++	do {
  2.2699 ++		if (b->in_pos == b->in_size)
  2.2700 ++			return XZ_OK;
  2.2701 ++
  2.2702 ++		if (((s->crc32 >> s->pos) & 0xFF) != b->in[b->in_pos++])
  2.2703 ++			return XZ_DATA_ERROR;
  2.2704 ++
  2.2705 ++		s->pos += 8;
  2.2706 ++
  2.2707 ++	} while (s->pos < 32);
  2.2708 ++
  2.2709 ++	s->crc32 = 0;
  2.2710 ++	s->pos = 0;
  2.2711 ++
  2.2712 ++	return XZ_STREAM_END;
  2.2713 ++}
  2.2714 ++
  2.2715 ++#ifdef XZ_DEC_ANY_CHECK
  2.2716 ++/*
  2.2717 ++ * Skip over the Check field when the Check ID is not supported.
  2.2718 ++ * Returns true once the whole Check field has been skipped over.
  2.2719 ++ */
  2.2720 ++static bool check_skip(struct xz_dec *s, struct xz_buf *b)
  2.2721 ++{
  2.2722 ++	while (s->pos < check_sizes[s->check_type]) {
  2.2723 ++		if (b->in_pos == b->in_size)
  2.2724 ++			return false;
  2.2725 ++
  2.2726 ++		++b->in_pos;
  2.2727 ++		++s->pos;
  2.2728 ++	}
  2.2729 ++
  2.2730 ++	s->pos = 0;
  2.2731 ++
  2.2732 ++	return true;
  2.2733 ++}
  2.2734 ++#endif
  2.2735 ++
  2.2736 ++/* Decode the Stream Header field (the first 12 bytes of the .xz Stream). */
  2.2737 ++static enum xz_ret dec_stream_header(struct xz_dec *s)
  2.2738 ++{
  2.2739 ++	if (!memeq(s->temp.buf, HEADER_MAGIC, HEADER_MAGIC_SIZE))
  2.2740 ++		return XZ_FORMAT_ERROR;
  2.2741 ++
  2.2742 ++	if (xz_crc32(s->temp.buf + HEADER_MAGIC_SIZE, 2, 0)
  2.2743 ++			!= get_le32(s->temp.buf + HEADER_MAGIC_SIZE + 2))
  2.2744 ++		return XZ_DATA_ERROR;
  2.2745 ++
  2.2746 ++	if (s->temp.buf[HEADER_MAGIC_SIZE] != 0)
  2.2747 ++		return XZ_OPTIONS_ERROR;
  2.2748 ++
  2.2749 ++	/*
  2.2750 ++	 * Of integrity checks, we support only none (Check ID = 0) and
  2.2751 ++	 * CRC32 (Check ID = 1). However, if XZ_DEC_ANY_CHECK is defined,
  2.2752 ++	 * we will accept other check types too, but then the check won't
  2.2753 ++	 * be verified and a warning (XZ_UNSUPPORTED_CHECK) will be given.
  2.2754 ++	 */
  2.2755 ++	s->check_type = s->temp.buf[HEADER_MAGIC_SIZE + 1];
  2.2756 ++
  2.2757 ++#ifdef XZ_DEC_ANY_CHECK
  2.2758 ++	if (s->check_type > XZ_CHECK_MAX)
  2.2759 ++		return XZ_OPTIONS_ERROR;
  2.2760 ++
  2.2761 ++	if (s->check_type > XZ_CHECK_CRC32)
  2.2762 ++		return XZ_UNSUPPORTED_CHECK;
  2.2763 ++#else
  2.2764 ++	if (s->check_type > XZ_CHECK_CRC32)
  2.2765 ++		return XZ_OPTIONS_ERROR;
  2.2766 ++#endif
  2.2767 ++
  2.2768 ++	return XZ_OK;
  2.2769 ++}
  2.2770 ++
  2.2771 ++/* Decode the Stream Footer field (the last 12 bytes of the .xz Stream) */
  2.2772 ++static enum xz_ret dec_stream_footer(struct xz_dec *s)
  2.2773 ++{
  2.2774 ++	if (!memeq(s->temp.buf + 10, FOOTER_MAGIC, FOOTER_MAGIC_SIZE))
  2.2775 ++		return XZ_DATA_ERROR;
  2.2776 ++
  2.2777 ++	if (xz_crc32(s->temp.buf + 4, 6, 0) != get_le32(s->temp.buf))
  2.2778 ++		return XZ_DATA_ERROR;
  2.2779 ++
  2.2780 ++	/*
  2.2781 ++	 * Validate Backward Size. Note that we never added the size of the
  2.2782 ++	 * Index CRC32 field to s->index.size, thus we use s->index.size / 4
  2.2783 ++	 * instead of s->index.size / 4 - 1.
  2.2784 ++	 */
  2.2785 ++	if ((s->index.size >> 2) != get_le32(s->temp.buf + 4))
  2.2786 ++		return XZ_DATA_ERROR;
  2.2787 ++
  2.2788 ++	if (s->temp.buf[8] != 0 || s->temp.buf[9] != s->check_type)
  2.2789 ++		return XZ_DATA_ERROR;
  2.2790 ++
  2.2791 ++	/*
  2.2792 ++	 * Use XZ_STREAM_END instead of XZ_OK to be more convenient
  2.2793 ++	 * for the caller.
  2.2794 ++	 */
  2.2795 ++	return XZ_STREAM_END;
  2.2796 ++}
  2.2797 ++
  2.2798 ++/* Decode the Block Header and initialize the filter chain. */
  2.2799 ++static enum xz_ret dec_block_header(struct xz_dec *s)
  2.2800 ++{
  2.2801 ++	enum xz_ret ret;
  2.2802 ++
  2.2803 ++	/*
  2.2804 ++	 * Validate the CRC32. We know that the temp buffer is at least
  2.2805 ++	 * eight bytes so this is safe.
  2.2806 ++	 */
  2.2807 ++	s->temp.size -= 4;
  2.2808 ++	if (xz_crc32(s->temp.buf, s->temp.size, 0)
  2.2809 ++			!= get_le32(s->temp.buf + s->temp.size))
  2.2810 ++		return XZ_DATA_ERROR;
  2.2811 ++
  2.2812 ++	s->temp.pos = 2;
  2.2813 ++
  2.2814 ++	/*
  2.2815 ++	 * Catch unsupported Block Flags. We support only one or two filters
  2.2816 ++	 * in the chain, so we catch that with the same test.
  2.2817 ++	 */
  2.2818 ++#ifdef XZ_DEC_BCJ
  2.2819 ++	if (s->temp.buf[1] & 0x3E)
  2.2820 ++#else
  2.2821 ++	if (s->temp.buf[1] & 0x3F)
  2.2822 ++#endif
  2.2823 ++		return XZ_OPTIONS_ERROR;
  2.2824 ++
  2.2825 ++	/* Compressed Size */
  2.2826 ++	if (s->temp.buf[1] & 0x40) {
  2.2827 ++		if (dec_vli(s, s->temp.buf, &s->temp.pos, s->temp.size)
  2.2828 ++					!= XZ_STREAM_END)
  2.2829 ++			return XZ_DATA_ERROR;
  2.2830 ++
  2.2831 ++		s->block_header.compressed = s->vli;
  2.2832 ++	} else {
  2.2833 ++		s->block_header.compressed = VLI_UNKNOWN;
  2.2834 ++	}
  2.2835 ++
  2.2836 ++	/* Uncompressed Size */
  2.2837 ++	if (s->temp.buf[1] & 0x80) {
  2.2838 ++		if (dec_vli(s, s->temp.buf, &s->temp.pos, s->temp.size)
  2.2839 ++				!= XZ_STREAM_END)
  2.2840 ++			return XZ_DATA_ERROR;
  2.2841 ++
  2.2842 ++		s->block_header.uncompressed = s->vli;
  2.2843 ++	} else {
  2.2844 ++		s->block_header.uncompressed = VLI_UNKNOWN;
  2.2845 ++	}
  2.2846 ++
  2.2847 ++#ifdef XZ_DEC_BCJ
  2.2848 ++	/* If there are two filters, the first one must be a BCJ filter. */
  2.2849 ++	s->bcj_active = s->temp.buf[1] & 0x01;
  2.2850 ++	if (s->bcj_active) {
  2.2851 ++		if (s->temp.size - s->temp.pos < 2)
  2.2852 ++			return XZ_OPTIONS_ERROR;
  2.2853 ++
  2.2854 ++		ret = xz_dec_bcj_reset(s->bcj, s->temp.buf[s->temp.pos++]);
  2.2855 ++		if (ret != XZ_OK)
  2.2856 ++			return ret;
  2.2857 ++
  2.2858 ++		/*
  2.2859 ++		 * We don't support custom start offset,
  2.2860 ++		 * so Size of Properties must be zero.
  2.2861 ++		 */
  2.2862 ++		if (s->temp.buf[s->temp.pos++] != 0x00)
  2.2863 ++			return XZ_OPTIONS_ERROR;
  2.2864 ++	}
  2.2865 ++#endif
  2.2866 ++
  2.2867 ++	/* Valid Filter Flags always take at least two bytes. */
  2.2868 ++	if (s->temp.size - s->temp.pos < 2)
  2.2869 ++		return XZ_DATA_ERROR;
  2.2870 ++
  2.2871 ++	/* Filter ID = LZMA2 */
  2.2872 ++	if (s->temp.buf[s->temp.pos++] != 0x21)
  2.2873 ++		return XZ_OPTIONS_ERROR;
  2.2874 ++
  2.2875 ++	/* Size of Properties = 1-byte Filter Properties */
  2.2876 ++	if (s->temp.buf[s->temp.pos++] != 0x01)
  2.2877 ++		return XZ_OPTIONS_ERROR;
  2.2878 ++
  2.2879 ++	/* Filter Properties contains LZMA2 dictionary size. */
  2.2880 ++	if (s->temp.size - s->temp.pos < 1)
  2.2881 ++		return XZ_DATA_ERROR;
  2.2882 ++
  2.2883 ++	ret = xz_dec_lzma2_reset(s->lzma2, s->temp.buf[s->temp.pos++]);
  2.2884 ++	if (ret != XZ_OK)
  2.2885 ++		return ret;
  2.2886 ++
  2.2887 ++	/* The rest must be Header Padding. */
  2.2888 ++	while (s->temp.pos < s->temp.size)
  2.2889 ++		if (s->temp.buf[s->temp.pos++] != 0x00)
  2.2890 ++			return XZ_OPTIONS_ERROR;
  2.2891 ++
  2.2892 ++	s->temp.pos = 0;
  2.2893 ++	s->block.compressed = 0;
  2.2894 ++	s->block.uncompressed = 0;
  2.2895 ++
  2.2896 ++	return XZ_OK;
  2.2897 ++}
  2.2898 ++
  2.2899 ++static enum xz_ret dec_main(struct xz_dec *s, struct xz_buf *b)
  2.2900 ++{
  2.2901 ++	enum xz_ret ret;
  2.2902 ++
  2.2903 ++	/*
  2.2904 ++	 * Store the start position for the case when we are in the middle
  2.2905 ++	 * of the Index field.
  2.2906 ++	 */
  2.2907 ++	s->in_start = b->in_pos;
  2.2908 ++
  2.2909 ++	while (true) {
  2.2910 ++		switch (s->sequence) {
  2.2911 ++		case SEQ_STREAM_HEADER:
  2.2912 ++			/*
  2.2913 ++			 * Stream Header is copied to s->temp, and then
  2.2914 ++			 * decoded from there. This way if the caller
  2.2915 ++			 * gives us only little input at a time, we can
  2.2916 ++			 * still keep the Stream Header decoding code
  2.2917 ++			 * simple. Similar approach is used in many places
  2.2918 ++			 * in this file.
  2.2919 ++			 */
  2.2920 ++			if (!fill_temp(s, b))
  2.2921 ++				return XZ_OK;
  2.2922 ++
  2.2923 ++			/*
  2.2924 ++			 * If dec_stream_header() returns
  2.2925 ++			 * XZ_UNSUPPORTED_CHECK, it is still possible
  2.2926 ++			 * to continue decoding if working in multi-call
  2.2927 ++			 * mode. Thus, update s->sequence before calling
  2.2928 ++			 * dec_stream_header().
  2.2929 ++			 */
  2.2930 ++			s->sequence = SEQ_BLOCK_START;
  2.2931 ++
  2.2932 ++			ret = dec_stream_header(s);
  2.2933 ++			if (ret != XZ_OK)
  2.2934 ++				return ret;
  2.2935 ++
  2.2936 ++		case SEQ_BLOCK_START:
  2.2937 ++			/* We need one byte of input to continue. */
  2.2938 ++			if (b->in_pos == b->in_size)
  2.2939 ++				return XZ_OK;
  2.2940 ++
  2.2941 ++			/* See if this is the beginning of the Index field. */
  2.2942 ++			if (b->in[b->in_pos] == 0) {
  2.2943 ++				s->in_start = b->in_pos++;
  2.2944 ++				s->sequence = SEQ_INDEX;
  2.2945 ++				break;
  2.2946 ++			}
  2.2947 ++
  2.2948 ++			/*
  2.2949 ++			 * Calculate the size of the Block Header and
  2.2950 ++			 * prepare to decode it.
  2.2951 ++			 */
  2.2952 ++			s->block_header.size
  2.2953 ++				= ((uint32_t)b->in[b->in_pos] + 1) * 4;
  2.2954 ++
  2.2955 ++			s->temp.size = s->block_header.size;
  2.2956 ++			s->temp.pos = 0;
  2.2957 ++			s->sequence = SEQ_BLOCK_HEADER;
  2.2958 ++
  2.2959 ++		case SEQ_BLOCK_HEADER:
  2.2960 ++			if (!fill_temp(s, b))
  2.2961 ++				return XZ_OK;
  2.2962 ++
  2.2963 ++			ret = dec_block_header(s);
  2.2964 ++			if (ret != XZ_OK)
  2.2965 ++				return ret;
  2.2966 ++
  2.2967 ++			s->sequence = SEQ_BLOCK_UNCOMPRESS;
  2.2968 ++
  2.2969 ++		case SEQ_BLOCK_UNCOMPRESS:
  2.2970 ++			ret = dec_block(s, b);
  2.2971 ++			if (ret != XZ_STREAM_END)
  2.2972 ++				return ret;
  2.2973 ++
  2.2974 ++			s->sequence = SEQ_BLOCK_PADDING;
  2.2975 ++
  2.2976 ++		case SEQ_BLOCK_PADDING:
  2.2977 ++			/*
  2.2978 ++			 * Size of Compressed Data + Block Padding
  2.2979 ++			 * must be a multiple of four. We don't need
  2.2980 ++			 * s->block.compressed for anything else
  2.2981 ++			 * anymore, so we use it here to test the size
  2.2982 ++			 * of the Block Padding field.
  2.2983 ++			 */
  2.2984 ++			while (s->block.compressed & 3) {
  2.2985 ++				if (b->in_pos == b->in_size)
  2.2986 ++					return XZ_OK;
  2.2987 ++
  2.2988 ++				if (b->in[b->in_pos++] != 0)
  2.2989 ++					return XZ_DATA_ERROR;
  2.2990 ++
  2.2991 ++				++s->block.compressed;
  2.2992 ++			}
  2.2993 ++
  2.2994 ++			s->sequence = SEQ_BLOCK_CHECK;
  2.2995 ++
  2.2996 ++		case SEQ_BLOCK_CHECK:
  2.2997 ++			if (s->check_type == XZ_CHECK_CRC32) {
  2.2998 ++				ret = crc32_validate(s, b);
  2.2999 ++				if (ret != XZ_STREAM_END)
  2.3000 ++					return ret;
  2.3001 ++			}
  2.3002 ++#ifdef XZ_DEC_ANY_CHECK
  2.3003 ++			else if (!check_skip(s, b)) {
  2.3004 ++				return XZ_OK;
  2.3005 ++			}
  2.3006 ++#endif
  2.3007 ++
  2.3008 ++			s->sequence = SEQ_BLOCK_START;
  2.3009 ++			break;
  2.3010 ++
  2.3011 ++		case SEQ_INDEX:
  2.3012 ++			ret = dec_index(s, b);
  2.3013 ++			if (ret != XZ_STREAM_END)
  2.3014 ++				return ret;
  2.3015 ++
  2.3016 ++			s->sequence = SEQ_INDEX_PADDING;
  2.3017 ++
  2.3018 ++		case SEQ_INDEX_PADDING:
  2.3019 ++			while ((s->index.size + (b->in_pos - s->in_start))
  2.3020 ++					& 3) {
  2.3021 ++				if (b->in_pos == b->in_size) {
  2.3022 ++					index_update(s, b);
  2.3023 ++					return XZ_OK;
  2.3024 ++				}
  2.3025 ++
  2.3026 ++				if (b->in[b->in_pos++] != 0)
  2.3027 ++					return XZ_DATA_ERROR;
  2.3028 ++			}
  2.3029 ++
  2.3030 ++			/* Finish the CRC32 value and Index size. */
  2.3031 ++			index_update(s, b);
  2.3032 ++
  2.3033 ++			/* Compare the hashes to validate the Index field. */
  2.3034 ++			if (!memeq(&s->block.hash, &s->index.hash,
  2.3035 ++					sizeof(s->block.hash)))
  2.3036 ++				return XZ_DATA_ERROR;
  2.3037 ++
  2.3038 ++			s->sequence = SEQ_INDEX_CRC32;
  2.3039 ++
  2.3040 ++		case SEQ_INDEX_CRC32:
  2.3041 ++			ret = crc32_validate(s, b);
  2.3042 ++			if (ret != XZ_STREAM_END)
  2.3043 ++				return ret;
  2.3044 ++
  2.3045 ++			s->temp.size = STREAM_HEADER_SIZE;
  2.3046 ++			s->sequence = SEQ_STREAM_FOOTER;
  2.3047 ++
  2.3048 ++		case SEQ_STREAM_FOOTER:
  2.3049 ++			if (!fill_temp(s, b))
  2.3050 ++				return XZ_OK;
  2.3051 ++
  2.3052 ++			return dec_stream_footer(s);
  2.3053 ++		}
  2.3054 ++	}
  2.3055 ++
  2.3056 ++	/* Never reached */
  2.3057 ++}
  2.3058 ++
  2.3059 ++/*
  2.3060 ++ * xz_dec_run() is a wrapper for dec_main() to handle some special cases in
  2.3061 ++ * multi-call and single-call decoding.
  2.3062 ++ *
  2.3063 ++ * In multi-call mode, we must return XZ_BUF_ERROR when it seems clear that we
  2.3064 ++ * are not going to make any progress anymore. This is to prevent the caller
  2.3065 ++ * from calling us infinitely when the input file is truncated or otherwise
  2.3066 ++ * corrupt. Since zlib-style API allows that the caller fills the input buffer
  2.3067 ++ * only when the decoder doesn't produce any new output, we have to be careful
  2.3068 ++ * to avoid returning XZ_BUF_ERROR too easily: XZ_BUF_ERROR is returned only
  2.3069 ++ * after the second consecutive call to xz_dec_run() that makes no progress.
  2.3070 ++ *
  2.3071 ++ * In single-call mode, if we couldn't decode everything and no error
  2.3072 ++ * occurred, either the input is truncated or the output buffer is too small.
  2.3073 ++ * Since we know that the last input byte never produces any output, we know
  2.3074 ++ * that if all the input was consumed and decoding wasn't finished, the file
  2.3075 ++ * must be corrupt. Otherwise the output buffer has to be too small or the
  2.3076 ++ * file is corrupt in a way that decoding it produces too big output.
  2.3077 ++ *
  2.3078 ++ * If single-call decoding fails, we reset b->in_pos and b->out_pos back to
  2.3079 ++ * their original values. This is because with some filter chains there won't
  2.3080 ++ * be any valid uncompressed data in the output buffer unless the decoding
  2.3081 ++ * actually succeeds (that's the price to pay of using the output buffer as
  2.3082 ++ * the workspace).
  2.3083 ++ */
  2.3084 ++XZ_EXTERN enum xz_ret xz_dec_run(struct xz_dec *s, struct xz_buf *b)
  2.3085 ++{
  2.3086 ++	size_t in_start;
  2.3087 ++	size_t out_start;
  2.3088 ++	enum xz_ret ret;
  2.3089 ++
  2.3090 ++	if (DEC_IS_SINGLE(s->mode))
  2.3091 ++		xz_dec_reset(s);
  2.3092 ++
  2.3093 ++	in_start = b->in_pos;
  2.3094 ++	out_start = b->out_pos;
  2.3095 ++	ret = dec_main(s, b);
  2.3096 ++
  2.3097 ++	if (DEC_IS_SINGLE(s->mode)) {
  2.3098 ++		if (ret == XZ_OK)
  2.3099 ++			ret = b->in_pos == b->in_size
  2.3100 ++					? XZ_DATA_ERROR : XZ_BUF_ERROR;
  2.3101 ++
  2.3102 ++		if (ret != XZ_STREAM_END) {
  2.3103 ++			b->in_pos = in_start;
  2.3104 ++			b->out_pos = out_start;
  2.3105 ++		}
  2.3106 ++
  2.3107 ++	} else if (ret == XZ_OK && in_start == b->in_pos
  2.3108 ++			&& out_start == b->out_pos) {
  2.3109 ++		if (s->allow_buf_error)
  2.3110 ++			ret = XZ_BUF_ERROR;
  2.3111 ++
  2.3112 ++		s->allow_buf_error = true;
  2.3113 ++	} else {
  2.3114 ++		s->allow_buf_error = false;
  2.3115 ++	}
  2.3116 ++
  2.3117 ++	return ret;
  2.3118 ++}
  2.3119 ++
  2.3120 ++XZ_EXTERN struct xz_dec *xz_dec_init(enum xz_mode mode, uint32_t dict_max)
  2.3121 ++{
  2.3122 ++	struct xz_dec *s = kmalloc(sizeof(*s), GFP_KERNEL);
  2.3123 ++	if (s == NULL)
  2.3124 ++		return NULL;
  2.3125 ++
  2.3126 ++	s->mode = mode;
  2.3127 ++
  2.3128 ++#ifdef XZ_DEC_BCJ
  2.3129 ++	s->bcj = xz_dec_bcj_create(DEC_IS_SINGLE(mode));
  2.3130 ++	if (s->bcj == NULL)
  2.3131 ++		goto error_bcj;
  2.3132 ++#endif
  2.3133 ++
  2.3134 ++	s->lzma2 = xz_dec_lzma2_create(mode, dict_max);
  2.3135 ++	if (s->lzma2 == NULL)
  2.3136 ++		goto error_lzma2;
  2.3137 ++
  2.3138 ++	xz_dec_reset(s);
  2.3139 ++	return s;
  2.3140 ++
  2.3141 ++error_lzma2:
  2.3142 ++#ifdef XZ_DEC_BCJ
  2.3143 ++	xz_dec_bcj_end(s->bcj);
  2.3144 ++error_bcj:
  2.3145 ++#endif
  2.3146 ++	kfree(s);
  2.3147 ++	return NULL;
  2.3148 ++}
  2.3149 ++
  2.3150 ++XZ_EXTERN void xz_dec_reset(struct xz_dec *s)
  2.3151 ++{
  2.3152 ++	s->sequence = SEQ_STREAM_HEADER;
  2.3153 ++	s->allow_buf_error = false;
  2.3154 ++	s->pos = 0;
  2.3155 ++	s->crc32 = 0;
  2.3156 ++	memzero(&s->block, sizeof(s->block));
  2.3157 ++	memzero(&s->index, sizeof(s->index));
  2.3158 ++	s->temp.pos = 0;
  2.3159 ++	s->temp.size = STREAM_HEADER_SIZE;
  2.3160 ++}
  2.3161 ++
  2.3162 ++XZ_EXTERN void xz_dec_end(struct xz_dec *s)
  2.3163 ++{
  2.3164 ++	if (s != NULL) {
  2.3165 ++		xz_dec_lzma2_end(s->lzma2);
  2.3166 ++#ifdef XZ_DEC_BCJ
  2.3167 ++		xz_dec_bcj_end(s->bcj);
  2.3168 ++#endif
  2.3169 ++		kfree(s);
  2.3170 ++	}
  2.3171 ++}
  2.3172 +diff --git a/lib/xz/xz_dec_syms.c b/lib/xz/xz_dec_syms.c
  2.3173 +new file mode 100644
  2.3174 +index 0000000..32eb3c0
  2.3175 +--- /dev/null
  2.3176 ++++ b/lib/xz/xz_dec_syms.c
  2.3177 +@@ -0,0 +1,26 @@
  2.3178 ++/*
  2.3179 ++ * XZ decoder module information
  2.3180 ++ *
  2.3181 ++ * Author: Lasse Collin <lasse.collin@tukaani.org>
  2.3182 ++ *
  2.3183 ++ * This file has been put into the public domain.
  2.3184 ++ * You can do whatever you want with this file.
  2.3185 ++ */
  2.3186 ++
  2.3187 ++#include <linux/module.h>
  2.3188 ++#include <linux/xz.h>
  2.3189 ++
  2.3190 ++EXPORT_SYMBOL(xz_dec_init);
  2.3191 ++EXPORT_SYMBOL(xz_dec_reset);
  2.3192 ++EXPORT_SYMBOL(xz_dec_run);
  2.3193 ++EXPORT_SYMBOL(xz_dec_end);
  2.3194 ++
  2.3195 ++MODULE_DESCRIPTION("XZ decompressor");
  2.3196 ++MODULE_VERSION("1.0");
  2.3197 ++MODULE_AUTHOR("Lasse Collin <lasse.collin@tukaani.org> and Igor Pavlov");
  2.3198 ++
  2.3199 ++/*
  2.3200 ++ * This code is in the public domain, but in Linux it's simplest to just
  2.3201 ++ * say it's GPL and consider the authors as the copyright holders.
  2.3202 ++ */
  2.3203 ++MODULE_LICENSE("GPL");
  2.3204 +diff --git a/lib/xz/xz_dec_test.c b/lib/xz/xz_dec_test.c
  2.3205 +new file mode 100644
  2.3206 +index 0000000..da28a19
  2.3207 +--- /dev/null
  2.3208 ++++ b/lib/xz/xz_dec_test.c
  2.3209 +@@ -0,0 +1,220 @@
  2.3210 ++/*
  2.3211 ++ * XZ decoder tester
  2.3212 ++ *
  2.3213 ++ * Author: Lasse Collin <lasse.collin@tukaani.org>
  2.3214 ++ *
  2.3215 ++ * This file has been put into the public domain.
  2.3216 ++ * You can do whatever you want with this file.
  2.3217 ++ */
  2.3218 ++
  2.3219 ++#include <linux/kernel.h>
  2.3220 ++#include <linux/module.h>
  2.3221 ++#include <linux/fs.h>
  2.3222 ++#include <linux/uaccess.h>
  2.3223 ++#include <linux/crc32.h>
  2.3224 ++#include <linux/xz.h>
  2.3225 ++
  2.3226 ++/* Maximum supported dictionary size */
  2.3227 ++#define DICT_MAX (1 << 20)
  2.3228 ++
  2.3229 ++/* Device name to pass to register_chrdev(). */
  2.3230 ++#define DEVICE_NAME "xz_dec_test"
  2.3231 ++
  2.3232 ++/* Dynamically allocated device major number */
  2.3233 ++static int device_major;
  2.3234 ++
  2.3235 ++/*
  2.3236 ++ * We reuse the same decoder state, and thus can decode only one
  2.3237 ++ * file at a time.
  2.3238 ++ */
  2.3239 ++static bool device_is_open;
  2.3240 ++
  2.3241 ++/* XZ decoder state */
  2.3242 ++static struct xz_dec *state;
  2.3243 ++
  2.3244 ++/*
  2.3245 ++ * Return value of xz_dec_run(). We need to avoid calling xz_dec_run() after
  2.3246 ++ * it has returned XZ_STREAM_END, so we make this static.
  2.3247 ++ */
  2.3248 ++static enum xz_ret ret;
  2.3249 ++
  2.3250 ++/*
  2.3251 ++ * Input and output buffers. The input buffer is used as a temporary safe
  2.3252 ++ * place for the data coming from the userspace.
  2.3253 ++ */
  2.3254 ++static uint8_t buffer_in[1024];
  2.3255 ++static uint8_t buffer_out[1024];
  2.3256 ++
  2.3257 ++/*
  2.3258 ++ * Structure to pass the input and output buffers to the XZ decoder.
  2.3259 ++ * A few of the fields are never modified so we initialize them here.
  2.3260 ++ */
  2.3261 ++static struct xz_buf buffers = {
  2.3262 ++	.in = buffer_in,
  2.3263 ++	.out = buffer_out,
  2.3264 ++	.out_size = sizeof(buffer_out)
  2.3265 ++};
  2.3266 ++
  2.3267 ++/*
  2.3268 ++ * CRC32 of uncompressed data. This is used to give the user a simple way
  2.3269 ++ * to check that the decoder produces correct output.
  2.3270 ++ */
  2.3271 ++static uint32_t crc;
  2.3272 ++
  2.3273 ++static int xz_dec_test_open(struct inode *i, struct file *f)
  2.3274 ++{
  2.3275 ++	if (device_is_open)
  2.3276 ++		return -EBUSY;
  2.3277 ++
  2.3278 ++	device_is_open = true;
  2.3279 ++
  2.3280 ++	xz_dec_reset(state);
  2.3281 ++	ret = XZ_OK;
  2.3282 ++	crc = 0xFFFFFFFF;
  2.3283 ++
  2.3284 ++	buffers.in_pos = 0;
  2.3285 ++	buffers.in_size = 0;
  2.3286 ++	buffers.out_pos = 0;
  2.3287 ++
  2.3288 ++	printk(KERN_INFO DEVICE_NAME ": opened\n");
  2.3289 ++	return 0;
  2.3290 ++}
  2.3291 ++
  2.3292 ++static int xz_dec_test_release(struct inode *i, struct file *f)
  2.3293 ++{
  2.3294 ++	device_is_open = false;
  2.3295 ++
  2.3296 ++	if (ret == XZ_OK)
  2.3297 ++		printk(KERN_INFO DEVICE_NAME ": input was truncated\n");
  2.3298 ++
  2.3299 ++	printk(KERN_INFO DEVICE_NAME ": closed\n");
  2.3300 ++	return 0;
  2.3301 ++}
  2.3302 ++
  2.3303 ++/*
  2.3304 ++ * Decode the data given to us from the userspace. CRC32 of the uncompressed
  2.3305 ++ * data is calculated and is printed at the end of successful decoding. The
  2.3306 ++ * uncompressed data isn't stored anywhere for further use.
  2.3307 ++ *
  2.3308 ++ * The .xz file must have exactly one Stream and no Stream Padding. The data
  2.3309 ++ * after the first Stream is considered to be garbage.
  2.3310 ++ */
  2.3311 ++static ssize_t xz_dec_test_write(struct file *file, const char __user *buf,
  2.3312 ++				 size_t size, loff_t *pos)
  2.3313 ++{
  2.3314 ++	size_t remaining;
  2.3315 ++
  2.3316 ++	if (ret != XZ_OK) {
  2.3317 ++		if (size > 0)
  2.3318 ++			printk(KERN_INFO DEVICE_NAME ": %zu bytes of "
  2.3319 ++					"garbage at the end of the file\n",
  2.3320 ++					size);
  2.3321 ++
  2.3322 ++		return -ENOSPC;
  2.3323 ++	}
  2.3324 ++
  2.3325 ++	printk(KERN_INFO DEVICE_NAME ": decoding %zu bytes of input\n",
  2.3326 ++			size);
  2.3327 ++
  2.3328 ++	remaining = size;
  2.3329 ++	while ((remaining > 0 || buffers.out_pos == buffers.out_size)
  2.3330 ++			&& ret == XZ_OK) {
  2.3331 ++		if (buffers.in_pos == buffers.in_size) {
  2.3332 ++			buffers.in_pos = 0;
  2.3333 ++			buffers.in_size = min(remaining, sizeof(buffer_in));
  2.3334 ++			if (copy_from_user(buffer_in, buf, buffers.in_size))
  2.3335 ++				return -EFAULT;
  2.3336 ++
  2.3337 ++			buf += buffers.in_size;
  2.3338 ++			remaining -= buffers.in_size;
  2.3339 ++		}
  2.3340 ++
  2.3341 ++		buffers.out_pos = 0;
  2.3342 ++		ret = xz_dec_run(state, &buffers);
  2.3343 ++		crc = crc32(crc, buffer_out, buffers.out_pos);
  2.3344 ++	}
  2.3345 ++
  2.3346 ++	switch (ret) {
  2.3347 ++	case XZ_OK:
  2.3348 ++		printk(KERN_INFO DEVICE_NAME ": XZ_OK\n");
  2.3349 ++		return size;
  2.3350 ++
  2.3351 ++	case XZ_STREAM_END:
  2.3352 ++		printk(KERN_INFO DEVICE_NAME ": XZ_STREAM_END, "
  2.3353 ++				"CRC32 = 0x%08X\n", ~crc);
  2.3354 ++		return size - remaining - (buffers.in_size - buffers.in_pos);
  2.3355 ++
  2.3356 ++	case XZ_MEMLIMIT_ERROR:
  2.3357 ++		printk(KERN_INFO DEVICE_NAME ": XZ_MEMLIMIT_ERROR\n");
  2.3358 ++		break;
  2.3359 ++
  2.3360 ++	case XZ_FORMAT_ERROR:
  2.3361 ++		printk(KERN_INFO DEVICE_NAME ": XZ_FORMAT_ERROR\n");
  2.3362 ++		break;
  2.3363 ++
  2.3364 ++	case XZ_OPTIONS_ERROR:
  2.3365 ++		printk(KERN_INFO DEVICE_NAME ": XZ_OPTIONS_ERROR\n");
  2.3366 ++		break;
  2.3367 ++
  2.3368 ++	case XZ_DATA_ERROR:
  2.3369 ++		printk(KERN_INFO DEVICE_NAME ": XZ_DATA_ERROR\n");
  2.3370 ++		break;
  2.3371 ++
  2.3372 ++	case XZ_BUF_ERROR:
  2.3373 ++		printk(KERN_INFO DEVICE_NAME ": XZ_BUF_ERROR\n");
  2.3374 ++		break;
  2.3375 ++
  2.3376 ++	default:
  2.3377 ++		printk(KERN_INFO DEVICE_NAME ": Bug detected!\n");
  2.3378 ++		break;
  2.3379 ++	}
  2.3380 ++
  2.3381 ++	return -EIO;
  2.3382 ++}
  2.3383 ++
  2.3384 ++/* Allocate the XZ decoder state and register the character device. */
  2.3385 ++static int __init xz_dec_test_init(void)
  2.3386 ++{
  2.3387 ++	static const struct file_operations fileops = {
  2.3388 ++		.owner = THIS_MODULE,
  2.3389 ++		.open = &xz_dec_test_open,
  2.3390 ++		.release = &xz_dec_test_release,
  2.3391 ++		.write = &xz_dec_test_write
  2.3392 ++	};
  2.3393 ++
  2.3394 ++	state = xz_dec_init(XZ_PREALLOC, DICT_MAX);
  2.3395 ++	if (state == NULL)
  2.3396 ++		return -ENOMEM;
  2.3397 ++
  2.3398 ++	device_major = register_chrdev(0, DEVICE_NAME, &fileops);
  2.3399 ++	if (device_major < 0) {
  2.3400 ++		xz_dec_end(state);
  2.3401 ++		return device_major;
  2.3402 ++	}
  2.3403 ++
  2.3404 ++	printk(KERN_INFO DEVICE_NAME ": module loaded\n");
  2.3405 ++	printk(KERN_INFO DEVICE_NAME ": Create a device node with "
  2.3406 ++			"'mknod " DEVICE_NAME " c %d 0' and write .xz files "
  2.3407 ++			"to it.\n", device_major);
  2.3408 ++	return 0;
  2.3409 ++}
  2.3410 ++
  2.3411 ++static void __exit xz_dec_test_exit(void)
  2.3412 ++{
  2.3413 ++	unregister_chrdev(device_major, DEVICE_NAME);
  2.3414 ++	xz_dec_end(state);
  2.3415 ++	printk(KERN_INFO DEVICE_NAME ": module unloaded\n");
  2.3416 ++}
  2.3417 ++
  2.3418 ++module_init(xz_dec_test_init);
  2.3419 ++module_exit(xz_dec_test_exit);
  2.3420 ++
  2.3421 ++MODULE_DESCRIPTION("XZ decompressor tester");
  2.3422 ++MODULE_VERSION("1.0");
  2.3423 ++MODULE_AUTHOR("Lasse Collin <lasse.collin@tukaani.org>");
  2.3424 ++
  2.3425 ++/*
  2.3426 ++ * This code is in the public domain, but in Linux it's simplest to just
  2.3427 ++ * say it's GPL and consider the authors as the copyright holders.
  2.3428 ++ */
  2.3429 ++MODULE_LICENSE("GPL");
  2.3430 +diff --git a/lib/xz/xz_lzma2.h b/lib/xz/xz_lzma2.h
  2.3431 +new file mode 100644
  2.3432 +index 0000000..071d67b
  2.3433 +--- /dev/null
  2.3434 ++++ b/lib/xz/xz_lzma2.h
  2.3435 +@@ -0,0 +1,204 @@
  2.3436 ++/*
  2.3437 ++ * LZMA2 definitions
  2.3438 ++ *
  2.3439 ++ * Authors: Lasse Collin <lasse.collin@tukaani.org>
  2.3440 ++ *          Igor Pavlov <http://7-zip.org/>
  2.3441 ++ *
  2.3442 ++ * This file has been put into the public domain.
  2.3443 ++ * You can do whatever you want with this file.
  2.3444 ++ */
  2.3445 ++
  2.3446 ++#ifndef XZ_LZMA2_H
  2.3447 ++#define XZ_LZMA2_H
  2.3448 ++
  2.3449 ++/* Range coder constants */
  2.3450 ++#define RC_SHIFT_BITS 8
  2.3451 ++#define RC_TOP_BITS 24
  2.3452 ++#define RC_TOP_VALUE (1 << RC_TOP_BITS)
  2.3453 ++#define RC_BIT_MODEL_TOTAL_BITS 11
  2.3454 ++#define RC_BIT_MODEL_TOTAL (1 << RC_BIT_MODEL_TOTAL_BITS)
  2.3455 ++#define RC_MOVE_BITS 5
  2.3456 ++
  2.3457 ++/*
  2.3458 ++ * Maximum number of position states. A position state is the lowest pb
  2.3459 ++ * number of bits of the current uncompressed offset. In some places there
  2.3460 ++ * are different sets of probabilities for different position states.
  2.3461 ++ */
  2.3462 ++#define POS_STATES_MAX (1 << 4)
  2.3463 ++
  2.3464 ++/*
  2.3465 ++ * This enum is used to track which LZMA symbols have occurred most recently
  2.3466 ++ * and in which order. This information is used to predict the next symbol.
  2.3467 ++ *
  2.3468 ++ * Symbols:
  2.3469 ++ *  - Literal: One 8-bit byte
  2.3470 ++ *  - Match: Repeat a chunk of data at some distance
  2.3471 ++ *  - Long repeat: Multi-byte match at a recently seen distance
  2.3472 ++ *  - Short repeat: One-byte repeat at a recently seen distance
  2.3473 ++ *
  2.3474 ++ * The symbol names are in from STATE_oldest_older_previous. REP means
  2.3475 ++ * either short or long repeated match, and NONLIT means any non-literal.
  2.3476 ++ */
  2.3477 ++enum lzma_state {
  2.3478 ++	STATE_LIT_LIT,
  2.3479 ++	STATE_MATCH_LIT_LIT,
  2.3480 ++	STATE_REP_LIT_LIT,
  2.3481 ++	STATE_SHORTREP_LIT_LIT,
  2.3482 ++	STATE_MATCH_LIT,
  2.3483 ++	STATE_REP_LIT,
  2.3484 ++	STATE_SHORTREP_LIT,
  2.3485 ++	STATE_LIT_MATCH,
  2.3486 ++	STATE_LIT_LONGREP,
  2.3487 ++	STATE_LIT_SHORTREP,
  2.3488 ++	STATE_NONLIT_MATCH,
  2.3489 ++	STATE_NONLIT_REP
  2.3490 ++};
  2.3491 ++
  2.3492 ++/* Total number of states */
  2.3493 ++#define STATES 12
  2.3494 ++
  2.3495 ++/* The lowest 7 states indicate that the previous state was a literal. */
  2.3496 ++#define LIT_STATES 7
  2.3497 ++
  2.3498 ++/* Indicate that the latest symbol was a literal. */
  2.3499 ++static inline void lzma_state_literal(enum lzma_state *state)
  2.3500 ++{
  2.3501 ++	if (*state <= STATE_SHORTREP_LIT_LIT)
  2.3502 ++		*state = STATE_LIT_LIT;
  2.3503 ++	else if (*state <= STATE_LIT_SHORTREP)
  2.3504 ++		*state -= 3;
  2.3505 ++	else
  2.3506 ++		*state -= 6;
  2.3507 ++}
  2.3508 ++
  2.3509 ++/* Indicate that the latest symbol was a match. */
  2.3510 ++static inline void lzma_state_match(enum lzma_state *state)
  2.3511 ++{
  2.3512 ++	*state = *state < LIT_STATES ? STATE_LIT_MATCH : STATE_NONLIT_MATCH;
  2.3513 ++}
  2.3514 ++
  2.3515 ++/* Indicate that the latest state was a long repeated match. */
  2.3516 ++static inline void lzma_state_long_rep(enum lzma_state *state)
  2.3517 ++{
  2.3518 ++	*state = *state < LIT_STATES ? STATE_LIT_LONGREP : STATE_NONLIT_REP;
  2.3519 ++}
  2.3520 ++
  2.3521 ++/* Indicate that the latest symbol was a short match. */
  2.3522 ++static inline void lzma_state_short_rep(enum lzma_state *state)
  2.3523 ++{
  2.3524 ++	*state = *state < LIT_STATES ? STATE_LIT_SHORTREP : STATE_NONLIT_REP;
  2.3525 ++}
  2.3526 ++
  2.3527 ++/* Test if the previous symbol was a literal. */
  2.3528 ++static inline bool lzma_state_is_literal(enum lzma_state state)
  2.3529 ++{
  2.3530 ++	return state < LIT_STATES;
  2.3531 ++}
  2.3532 ++
  2.3533 ++/* Each literal coder is divided in three sections:
  2.3534 ++ *   - 0x001-0x0FF: Without match byte
  2.3535 ++ *   - 0x101-0x1FF: With match byte; match bit is 0
  2.3536 ++ *   - 0x201-0x2FF: With match byte; match bit is 1
  2.3537 ++ *
  2.3538 ++ * Match byte is used when the previous LZMA symbol was something else than
  2.3539 ++ * a literal (that is, it was some kind of match).
  2.3540 ++ */
  2.3541 ++#define LITERAL_CODER_SIZE 0x300
  2.3542 ++
  2.3543 ++/* Maximum number of literal coders */
  2.3544 ++#define LITERAL_CODERS_MAX (1 << 4)
  2.3545 ++
  2.3546 ++/* Minimum length of a match is two bytes. */
  2.3547 ++#define MATCH_LEN_MIN 2
  2.3548 ++
  2.3549 ++/* Match length is encoded with 4, 5, or 10 bits.
  2.3550 ++ *
  2.3551 ++ * Length   Bits
  2.3552 ++ *  2-9      4 = Choice=0 + 3 bits
  2.3553 ++ * 10-17     5 = Choice=1 + Choice2=0 + 3 bits
  2.3554 ++ * 18-273   10 = Choice=1 + Choice2=1 + 8 bits
  2.3555 ++ */
  2.3556 ++#define LEN_LOW_BITS 3
  2.3557 ++#define LEN_LOW_SYMBOLS (1 << LEN_LOW_BITS)
  2.3558 ++#define LEN_MID_BITS 3
  2.3559 ++#define LEN_MID_SYMBOLS (1 << LEN_MID_BITS)
  2.3560 ++#define LEN_HIGH_BITS 8
  2.3561 ++#define LEN_HIGH_SYMBOLS (1 << LEN_HIGH_BITS)
  2.3562 ++#define LEN_SYMBOLS (LEN_LOW_SYMBOLS + LEN_MID_SYMBOLS + LEN_HIGH_SYMBOLS)
  2.3563 ++
  2.3564 ++/*
  2.3565 ++ * Maximum length of a match is 273 which is a result of the encoding
  2.3566 ++ * described above.
  2.3567 ++ */
  2.3568 ++#define MATCH_LEN_MAX (MATCH_LEN_MIN + LEN_SYMBOLS - 1)
  2.3569 ++
  2.3570 ++/*
  2.3571 ++ * Different sets of probabilities are used for match distances that have
  2.3572 ++ * very short match length: Lengths of 2, 3, and 4 bytes have a separate
  2.3573 ++ * set of probabilities for each length. The matches with longer length
  2.3574 ++ * use a shared set of probabilities.
  2.3575 ++ */
  2.3576 ++#define DIST_STATES 4
  2.3577 ++
  2.3578 ++/*
  2.3579 ++ * Get the index of the appropriate probability array for decoding
  2.3580 ++ * the distance slot.
  2.3581 ++ */
  2.3582 ++static inline uint32_t lzma_get_dist_state(uint32_t len)
  2.3583 ++{
  2.3584 ++	return len < DIST_STATES + MATCH_LEN_MIN
  2.3585 ++			? len - MATCH_LEN_MIN : DIST_STATES - 1;
  2.3586 ++}
  2.3587 ++
  2.3588 ++/*
  2.3589 ++ * The highest two bits of a 32-bit match distance are encoded using six bits.
  2.3590 ++ * This six-bit value is called a distance slot. This way encoding a 32-bit
  2.3591 ++ * value takes 6-36 bits, larger values taking more bits.
  2.3592 ++ */
  2.3593 ++#define DIST_SLOT_BITS 6
  2.3594 ++#define DIST_SLOTS (1 << DIST_SLOT_BITS)
  2.3595 ++
  2.3596 ++/* Match distances up to 127 are fully encoded using probabilities. Since
  2.3597 ++ * the highest two bits (distance slot) are always encoded using six bits,
  2.3598 ++ * the distances 0-3 don't need any additional bits to encode, since the
  2.3599 ++ * distance slot itself is the same as the actual distance. DIST_MODEL_START
  2.3600 ++ * indicates the first distance slot where at least one additional bit is
  2.3601 ++ * needed.
  2.3602 ++ */
  2.3603 ++#define DIST_MODEL_START 4
  2.3604 ++
  2.3605 ++/*
  2.3606 ++ * Match distances greater than 127 are encoded in three pieces:
  2.3607 ++ *   - distance slot: the highest two bits
  2.3608 ++ *   - direct bits: 2-26 bits below the highest two bits
  2.3609 ++ *   - alignment bits: four lowest bits
  2.3610 ++ *
  2.3611 ++ * Direct bits don't use any probabilities.
  2.3612 ++ *
  2.3613 ++ * The distance slot value of 14 is for distances 128-191.
  2.3614 ++ */
  2.3615 ++#define DIST_MODEL_END 14
  2.3616 ++
  2.3617 ++/* Distance slots that indicate a distance <= 127. */
  2.3618 ++#define FULL_DISTANCES_BITS (DIST_MODEL_END / 2)
  2.3619 ++#define FULL_DISTANCES (1 << FULL_DISTANCES_BITS)
  2.3620 ++
  2.3621 ++/*
  2.3622 ++ * For match distances greater than 127, only the highest two bits and the
  2.3623 ++ * lowest four bits (alignment) is encoded using probabilities.
  2.3624 ++ */
  2.3625 ++#define ALIGN_BITS 4
  2.3626 ++#define ALIGN_SIZE (1 << ALIGN_BITS)
  2.3627 ++#define ALIGN_MASK (ALIGN_SIZE - 1)
  2.3628 ++
  2.3629 ++/* Total number of all probability variables */
  2.3630 ++#define PROBS_TOTAL (1846 + LITERAL_CODERS_MAX * LITERAL_CODER_SIZE)
  2.3631 ++
  2.3632 ++/*
  2.3633 ++ * LZMA remembers the four most recent match distances. Reusing these
  2.3634 ++ * distances tends to take less space than re-encoding the actual
  2.3635 ++ * distance value.
  2.3636 ++ */
  2.3637 ++#define REPS 4
  2.3638 ++
  2.3639 ++#endif
  2.3640 +diff --git a/lib/xz/xz_private.h b/lib/xz/xz_private.h
  2.3641 +new file mode 100644
  2.3642 +index 0000000..a65633e
  2.3643 +--- /dev/null
  2.3644 ++++ b/lib/xz/xz_private.h
  2.3645 +@@ -0,0 +1,156 @@
  2.3646 ++/*
  2.3647 ++ * Private includes and definitions
  2.3648 ++ *
  2.3649 ++ * Author: Lasse Collin <lasse.collin@tukaani.org>
  2.3650 ++ *
  2.3651 ++ * This file has been put into the public domain.
  2.3652 ++ * You can do whatever you want with this file.
  2.3653 ++ */
  2.3654 ++
  2.3655 ++#ifndef XZ_PRIVATE_H
  2.3656 ++#define XZ_PRIVATE_H
  2.3657 ++
  2.3658 ++#ifdef __KERNEL__
  2.3659 ++#	include <linux/xz.h>
  2.3660 ++#	include <asm/byteorder.h>
  2.3661 ++#	include <asm/unaligned.h>
  2.3662 ++	/* XZ_PREBOOT may be defined only via decompress_unxz.c. */
  2.3663 ++#	ifndef XZ_PREBOOT
  2.3664 ++#		include <linux/slab.h>
  2.3665 ++#		include <linux/vmalloc.h>
  2.3666 ++#		include <linux/string.h>
  2.3667 ++#		ifdef CONFIG_XZ_DEC_X86
  2.3668 ++#			define XZ_DEC_X86
  2.3669 ++#		endif
  2.3670 ++#		ifdef CONFIG_XZ_DEC_POWERPC
  2.3671 ++#			define XZ_DEC_POWERPC
  2.3672 ++#		endif
  2.3673 ++#		ifdef CONFIG_XZ_DEC_IA64
  2.3674 ++#			define XZ_DEC_IA64
  2.3675 ++#		endif
  2.3676 ++#		ifdef CONFIG_XZ_DEC_ARM
  2.3677 ++#			define XZ_DEC_ARM
  2.3678 ++#		endif
  2.3679 ++#		ifdef CONFIG_XZ_DEC_ARMTHUMB
  2.3680 ++#			define XZ_DEC_ARMTHUMB
  2.3681 ++#		endif
  2.3682 ++#		ifdef CONFIG_XZ_DEC_SPARC
  2.3683 ++#			define XZ_DEC_SPARC
  2.3684 ++#		endif
  2.3685 ++#		define memeq(a, b, size) (memcmp(a, b, size) == 0)
  2.3686 ++#		define memzero(buf, size) memset(buf, 0, size)
  2.3687 ++#	endif
  2.3688 ++#	define get_le32(p) le32_to_cpup((const uint32_t *)(p))
  2.3689 ++#else
  2.3690 ++	/*
  2.3691 ++	 * For userspace builds, use a separate header to define the required
  2.3692 ++	 * macros and functions. This makes it easier to adapt the code into
  2.3693 ++	 * different environments and avoids clutter in the Linux kernel tree.
  2.3694 ++	 */
  2.3695 ++#	include "xz_config.h"
  2.3696 ++#endif
  2.3697 ++
  2.3698 ++/* If no specific decoding mode is requested, enable support for all modes. */
  2.3699 ++#if !defined(XZ_DEC_SINGLE) && !defined(XZ_DEC_PREALLOC) \
  2.3700 ++		&& !defined(XZ_DEC_DYNALLOC)
  2.3701 ++#	define XZ_DEC_SINGLE
  2.3702 ++#	define XZ_DEC_PREALLOC
  2.3703 ++#	define XZ_DEC_DYNALLOC
  2.3704 ++#endif
  2.3705 ++
  2.3706 ++/*
  2.3707 ++ * The DEC_IS_foo(mode) macros are used in "if" statements. If only some
  2.3708 ++ * of the supported modes are enabled, these macros will evaluate to true or
  2.3709 ++ * false at compile time and thus allow the compiler to omit unneeded code.
  2.3710 ++ */
  2.3711 ++#ifdef XZ_DEC_SINGLE
  2.3712 ++#	define DEC_IS_SINGLE(mode) ((mode) == XZ_SINGLE)
  2.3713 ++#else
  2.3714 ++#	define DEC_IS_SINGLE(mode) (false)
  2.3715 ++#endif
  2.3716 ++
  2.3717 ++#ifdef XZ_DEC_PREALLOC
  2.3718 ++#	define DEC_IS_PREALLOC(mode) ((mode) == XZ_PREALLOC)
  2.3719 ++#else
  2.3720 ++#	define DEC_IS_PREALLOC(mode) (false)
  2.3721 ++#endif
  2.3722 ++
  2.3723 ++#ifdef XZ_DEC_DYNALLOC
  2.3724 ++#	define DEC_IS_DYNALLOC(mode) ((mode) == XZ_DYNALLOC)
  2.3725 ++#else
  2.3726 ++#	define DEC_IS_DYNALLOC(mode) (false)
  2.3727 ++#endif
  2.3728 ++
  2.3729 ++#if !defined(XZ_DEC_SINGLE)
  2.3730 ++#	define DEC_IS_MULTI(mode) (true)
  2.3731 ++#elif defined(XZ_DEC_PREALLOC) || defined(XZ_DEC_DYNALLOC)
  2.3732 ++#	define DEC_IS_MULTI(mode) ((mode) != XZ_SINGLE)
  2.3733 ++#else
  2.3734 ++#	define DEC_IS_MULTI(mode) (false)
  2.3735 ++#endif
  2.3736 ++
  2.3737 ++/*
  2.3738 ++ * If any of the BCJ filter decoders are wanted, define XZ_DEC_BCJ.
  2.3739 ++ * XZ_DEC_BCJ is used to enable generic support for BCJ decoders.
  2.3740 ++ */
  2.3741 ++#ifndef XZ_DEC_BCJ
  2.3742 ++#	if defined(XZ_DEC_X86) || defined(XZ_DEC_POWERPC) \
  2.3743 ++			|| defined(XZ_DEC_IA64) || defined(XZ_DEC_ARM) \
  2.3744 ++			|| defined(XZ_DEC_ARM) || defined(XZ_DEC_ARMTHUMB) \
  2.3745 ++			|| defined(XZ_DEC_SPARC)
  2.3746 ++#		define XZ_DEC_BCJ
  2.3747 ++#	endif
  2.3748 ++#endif
  2.3749 ++
  2.3750 ++/*
  2.3751 ++ * Allocate memory for LZMA2 decoder. xz_dec_lzma2_reset() must be used
  2.3752 ++ * before calling xz_dec_lzma2_run().
  2.3753 ++ */
  2.3754 ++XZ_EXTERN struct xz_dec_lzma2 *xz_dec_lzma2_create(enum xz_mode mode,
  2.3755 ++						   uint32_t dict_max);
  2.3756 ++
  2.3757 ++/*
  2.3758 ++ * Decode the LZMA2 properties (one byte) and reset the decoder. Return
  2.3759 ++ * XZ_OK on success, XZ_MEMLIMIT_ERROR if the preallocated dictionary is not
  2.3760 ++ * big enough, and XZ_OPTIONS_ERROR if props indicates something that this
  2.3761 ++ * decoder doesn't support.
  2.3762 ++ */
  2.3763 ++XZ_EXTERN enum xz_ret xz_dec_lzma2_reset(struct xz_dec_lzma2 *s,
  2.3764 ++					 uint8_t props);
  2.3765 ++
  2.3766 ++/* Decode raw LZMA2 stream from b->in to b->out. */
  2.3767 ++XZ_EXTERN enum xz_ret xz_dec_lzma2_run(struct xz_dec_lzma2 *s,
  2.3768 ++				       struct xz_buf *b);
  2.3769 ++
  2.3770 ++/* Free the memory allocated for the LZMA2 decoder. */
  2.3771 ++XZ_EXTERN void xz_dec_lzma2_end(struct xz_dec_lzma2 *s);
  2.3772 ++
  2.3773 ++#ifdef XZ_DEC_BCJ
  2.3774 ++/*
  2.3775 ++ * Allocate memory for BCJ decoders. xz_dec_bcj_reset() must be used before
  2.3776 ++ * calling xz_dec_bcj_run().
  2.3777 ++ */
  2.3778 ++XZ_EXTERN struct xz_dec_bcj *xz_dec_bcj_create(bool single_call);
  2.3779 ++
  2.3780 ++/*
  2.3781 ++ * Decode the Filter ID of a BCJ filter. This implementation doesn't
  2.3782 ++ * support custom start offsets, so no decoding of Filter Properties
  2.3783 ++ * is needed. Returns XZ_OK if the given Filter ID is supported.
  2.3784 ++ * Otherwise XZ_OPTIONS_ERROR is returned.
  2.3785 ++ */
  2.3786 ++XZ_EXTERN enum xz_ret xz_dec_bcj_reset(struct xz_dec_bcj *s, uint8_t id);
  2.3787 ++
  2.3788 ++/*
  2.3789 ++ * Decode raw BCJ + LZMA2 stream. This must be used only if there actually is
  2.3790 ++ * a BCJ filter in the chain. If the chain has only LZMA2, xz_dec_lzma2_run()
  2.3791 ++ * must be called directly.
  2.3792 ++ */
  2.3793 ++XZ_EXTERN enum xz_ret xz_dec_bcj_run(struct xz_dec_bcj *s,
  2.3794 ++				     struct xz_dec_lzma2 *lzma2,
  2.3795 ++				     struct xz_buf *b);
  2.3796 ++
  2.3797 ++/* Free the memory allocated for the BCJ filters. */
  2.3798 ++#define xz_dec_bcj_end(s) kfree(s)
  2.3799 ++#endif
  2.3800 ++
  2.3801 ++#endif
  2.3802 +diff --git a/lib/xz/xz_stream.h b/lib/xz/xz_stream.h
  2.3803 +new file mode 100644
  2.3804 +index 0000000..66cb5a7
  2.3805 +--- /dev/null
  2.3806 ++++ b/lib/xz/xz_stream.h
  2.3807 +@@ -0,0 +1,62 @@
  2.3808 ++/*
  2.3809 ++ * Definitions for handling the .xz file format
  2.3810 ++ *
  2.3811 ++ * Author: Lasse Collin <lasse.collin@tukaani.org>
  2.3812 ++ *
  2.3813 ++ * This file has been put into the public domain.
  2.3814 ++ * You can do whatever you want with this file.
  2.3815 ++ */
  2.3816 ++
  2.3817 ++#ifndef XZ_STREAM_H
  2.3818 ++#define XZ_STREAM_H
  2.3819 ++
  2.3820 ++#if defined(__KERNEL__) && !XZ_INTERNAL_CRC32
  2.3821 ++#	include <linux/crc32.h>
  2.3822 ++#	undef crc32
  2.3823 ++#	define xz_crc32(buf, size, crc) \
  2.3824 ++		(~crc32_le(~(uint32_t)(crc), buf, size))
  2.3825 ++#endif
  2.3826 ++
  2.3827 ++/*
  2.3828 ++ * See the .xz file format specification at
  2.3829 ++ * http://tukaani.org/xz/xz-file-format.txt
  2.3830 ++ * to understand the container format.
  2.3831 ++ */
  2.3832 ++
  2.3833 ++#define STREAM_HEADER_SIZE 12
  2.3834 ++
  2.3835 ++#define HEADER_MAGIC "\3757zXZ"
  2.3836 ++#define HEADER_MAGIC_SIZE 6
  2.3837 ++
  2.3838 ++#define FOOTER_MAGIC "YZ"
  2.3839 ++#define FOOTER_MAGIC_SIZE 2
  2.3840 ++
  2.3841 ++/*
  2.3842 ++ * Variable-length integer can hold a 63-bit unsigned integer or a special
  2.3843 ++ * value indicating that the value is unknown.
  2.3844 ++ *
  2.3845 ++ * Experimental: vli_type can be defined to uint32_t to save a few bytes
  2.3846 ++ * in code size (no effect on speed). Doing so limits the uncompressed and
  2.3847 ++ * compressed size of the file to less than 256 MiB and may also weaken
  2.3848 ++ * error detection slightly.
  2.3849 ++ */
  2.3850 ++typedef uint64_t vli_type;
  2.3851 ++
  2.3852 ++#define VLI_MAX ((vli_type)-1 / 2)
  2.3853 ++#define VLI_UNKNOWN ((vli_type)-1)
  2.3854 ++
  2.3855 ++/* Maximum encoded size of a VLI */
  2.3856 ++#define VLI_BYTES_MAX (sizeof(vli_type) * 8 / 7)
  2.3857 ++
  2.3858 ++/* Integrity Check types */
  2.3859 ++enum xz_check {
  2.3860 ++	XZ_CHECK_NONE = 0,
  2.3861 ++	XZ_CHECK_CRC32 = 1,
  2.3862 ++	XZ_CHECK_CRC64 = 4,
  2.3863 ++	XZ_CHECK_SHA256 = 10
  2.3864 ++};
  2.3865 ++
  2.3866 ++/* Maximum possible Check ID */
  2.3867 ++#define XZ_CHECK_MAX 15
  2.3868 ++
  2.3869 ++#endif
  2.3870 +diff --git a/scripts/Makefile.lib b/scripts/Makefile.lib
  2.3871 +index 54fd1b7..b862007 100644
  2.3872 +--- a/scripts/Makefile.lib
  2.3873 ++++ b/scripts/Makefile.lib
  2.3874 +@@ -246,6 +246,34 @@ cmd_lzo = (cat $(filter-out FORCE,$^) | \
  2.3875 + 	lzop -9 && $(call size_append, $(filter-out FORCE,$^))) > $@ || \
  2.3876 + 	(rm -f $@ ; false)
  2.3877 + 
  2.3878 ++# XZ
  2.3879 ++# ---------------------------------------------------------------------------
  2.3880 ++# Use xzkern to compress the kernel image and xzmisc to compress other things.
  2.3881 ++#
  2.3882 ++# xzkern uses a big LZMA2 dictionary since it doesn't increase memory usage
  2.3883 ++# of the kernel decompressor. A BCJ filter is used if it is available for
  2.3884 ++# the target architecture. xzkern also appends uncompressed size of the data
  2.3885 ++# using size_append. The .xz format has the size information available at
  2.3886 ++# the end of the file too, but it's in more complex format and it's good to
  2.3887 ++# avoid changing the part of the boot code that reads the uncompressed size.
  2.3888 ++# Note that the bytes added by size_append will make the xz tool think that
  2.3889 ++# the file is corrupt. This is expected.
  2.3890 ++#
  2.3891 ++# xzmisc doesn't use size_append, so it can be used to create normal .xz
  2.3892 ++# files. xzmisc uses smaller LZMA2 dictionary than xzkern, because a very
  2.3893 ++# big dictionary would increase the memory usage too much in the multi-call
  2.3894 ++# decompression mode. A BCJ filter isn't used either.
  2.3895 ++quiet_cmd_xzkern = XZKERN  $@
  2.3896 ++cmd_xzkern = (cat $(filter-out FORCE,$^) | \
  2.3897 ++	sh $(srctree)/scripts/xz_wrap.sh && \
  2.3898 ++	$(call size_append, $(filter-out FORCE,$^))) > $@ || \
  2.3899 ++	(rm -f $@ ; false)
  2.3900 ++
  2.3901 ++quiet_cmd_xzmisc = XZMISC  $@
  2.3902 ++cmd_xzmisc = (cat $(filter-out FORCE,$^) | \
  2.3903 ++	xz --check=crc32 --lzma2=dict=1MiB) > $@ || \
  2.3904 ++	(rm -f $@ ; false)
  2.3905 ++
  2.3906 + # misc stuff
  2.3907 + # ---------------------------------------------------------------------------
  2.3908 + quote:="
  2.3909 +diff --git a/scripts/xz_wrap.sh b/scripts/xz_wrap.sh
  2.3910 +new file mode 100644
  2.3911 +index 0000000..17a5798
  2.3912 +--- /dev/null
  2.3913 ++++ b/scripts/xz_wrap.sh
  2.3914 +@@ -0,0 +1,23 @@
  2.3915 ++#!/bin/sh
  2.3916 ++#
  2.3917 ++# This is a wrapper for xz to compress the kernel image using appropriate
  2.3918 ++# compression options depending on the architecture.
  2.3919 ++#
  2.3920 ++# Author: Lasse Collin <lasse.collin@tukaani.org>
  2.3921 ++#
  2.3922 ++# This file has been put into the public domain.
  2.3923 ++# You can do whatever you want with this file.
  2.3924 ++#
  2.3925 ++
  2.3926 ++BCJ=
  2.3927 ++LZMA2OPTS=
  2.3928 ++
  2.3929 ++case $ARCH in
  2.3930 ++	x86|x86_64)     BCJ=--x86 ;;
  2.3931 ++	powerpc)        BCJ=--powerpc ;;
  2.3932 ++	ia64)           BCJ=--ia64; LZMA2OPTS=pb=4 ;;
  2.3933 ++	arm)            BCJ=--arm ;;
  2.3934 ++	sparc)          BCJ=--sparc ;;
  2.3935 ++esac
  2.3936 ++
  2.3937 ++exec xz --check=crc32 $BCJ --lzma2=$LZMA2OPTS,dict=32MiB
     3.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     3.2 +++ b/linux/stuff/002-squashfs-decompressors-add-boot-time-xz-support.patch	Tue Dec 14 21:45:09 2010 +0000
     3.3 @@ -0,0 +1,638 @@
     3.4 +From: Lasse Collin <lasse.collin@tukaani.org>
     3.5 +Date: Thu, 2 Dec 2010 19:14:37 +0000 (+0200)
     3.6 +Subject: Decompressors: Add boot-time XZ support
     3.7 +X-Git-Url: http://git.kernel.org/?p=linux%2Fkernel%2Fgit%2Fpkl%2Fsquashfs-xz.git;a=commitdiff_plain;h=c64bc9a229b46db75d7761601dd8ca25385a7780
     3.8 +
     3.9 +Decompressors: Add boot-time XZ support
    3.10 +
    3.11 +This implements the API defined in <linux/decompress/generic.h>
    3.12 +which is used for kernel, initramfs, and initrd decompression.
    3.13 +This patch together with the first patch is enough for
    3.14 +XZ-compressed initramfs and initrd; XZ-compressed kernel will
    3.15 +need arch-specific changes.
    3.16 +
    3.17 +In contrast to other initramfs compression methods, support for
    3.18 +XZ-compressed initramfs is not enabled by default in usr/Kconfig.
    3.19 +This is primarily due to the Kconfig options of the xz_dec
    3.20 +module. It can be good to require that xz_dec is enabled
    3.21 +separately so the user can select only the BCJ filters he needs
    3.22 +when EMBEDDED=y.
    3.23 +
    3.24 +The buffering requirements described in decompress_unxz.c are
    3.25 +stricter than with gzip, so the relevant changes should be done
    3.26 +to the arch-specific code when adding support for XZ-compressed
    3.27 +kernel. Similarly, the heap size in arch-specific pre-boot code
    3.28 +may need to be increased (30 KiB is enough).
    3.29 +
    3.30 +The XZ decompressor needs memmove(), memeq() (memcmp() == 0),
    3.31 +and memzero() (memset(ptr, 0, size)), which aren't available in
    3.32 +all arch-specific pre-boot environments. I'm including simple
    3.33 +versions in decompress_unxz.c, but a cleaner solution would
    3.34 +naturally be nicer.
    3.35 +
    3.36 +Signed-off-by: Lasse Collin <lasse.collin@tukaani.org>
    3.37 +---
    3.38 +
    3.39 +diff --git a/include/linux/decompress/unxz.h b/include/linux/decompress/unxz.h
    3.40 +new file mode 100644
    3.41 +index 0000000..41728fc
    3.42 +--- /dev/null
    3.43 ++++ b/include/linux/decompress/unxz.h
    3.44 +@@ -0,0 +1,19 @@
    3.45 ++/*
    3.46 ++ * Wrapper for decompressing XZ-compressed kernel, initramfs, and initrd
    3.47 ++ *
    3.48 ++ * Author: Lasse Collin <lasse.collin@tukaani.org>
    3.49 ++ *
    3.50 ++ * This file has been put into the public domain.
    3.51 ++ * You can do whatever you want with this file.
    3.52 ++ */
    3.53 ++
    3.54 ++#ifndef DECOMPRESS_UNXZ_H
    3.55 ++#define DECOMPRESS_UNXZ_H
    3.56 ++
    3.57 ++int unxz(unsigned char *in, int in_size,
    3.58 ++	 int (*fill)(void *dest, unsigned int size),
    3.59 ++	 int (*flush)(void *src, unsigned int size),
    3.60 ++	 unsigned char *out, int *in_used,
    3.61 ++	 void (*error)(char *x));
    3.62 ++
    3.63 ++#endif
    3.64 +diff --git a/init/Kconfig b/init/Kconfig
    3.65 +index 2de5b1c..d9fbb0f 100644
    3.66 +--- a/init/Kconfig
    3.67 ++++ b/init/Kconfig
    3.68 +@@ -123,13 +123,16 @@ config HAVE_KERNEL_BZIP2
    3.69 + config HAVE_KERNEL_LZMA
    3.70 + 	bool
    3.71 + 
    3.72 ++config HAVE_KERNEL_XZ
    3.73 ++	bool
    3.74 ++
    3.75 + config HAVE_KERNEL_LZO
    3.76 + 	bool
    3.77 + 
    3.78 + choice
    3.79 + 	prompt "Kernel compression mode"
    3.80 + 	default KERNEL_GZIP
    3.81 +-	depends on HAVE_KERNEL_GZIP || HAVE_KERNEL_BZIP2 || HAVE_KERNEL_LZMA || HAVE_KERNEL_LZO
    3.82 ++	depends on HAVE_KERNEL_GZIP || HAVE_KERNEL_BZIP2 || HAVE_KERNEL_LZMA || HAVE_KERNEL_XZ || HAVE_KERNEL_LZO
    3.83 + 	help
    3.84 + 	  The linux kernel is a kind of self-extracting executable.
    3.85 + 	  Several compression algorithms are available, which differ
    3.86 +@@ -174,6 +177,21 @@ config KERNEL_LZMA
    3.87 + 	  two. Compression is slowest.	The kernel size is about 33%
    3.88 + 	  smaller with LZMA in comparison to gzip.
    3.89 + 
    3.90 ++config KERNEL_XZ
    3.91 ++	bool "XZ"
    3.92 ++	depends on HAVE_KERNEL_XZ
    3.93 ++	help
    3.94 ++	  XZ uses the LZMA2 algorithm and instruction set specific
    3.95 ++	  BCJ filters which can improve compression ratio of executable
    3.96 ++	  code. The size of the kernel is about 30% smaller with XZ in
    3.97 ++	  comparison to gzip. On architectures for which there is a BCJ
    3.98 ++	  filter (i386, x86_64, ARM, IA-64, PowerPC, and SPARC), XZ
    3.99 ++	  will create a few percent smaller kernel than plain LZMA.
   3.100 ++
   3.101 ++	  The speed is about the same as with LZMA: The decompression
   3.102 ++	  speed of XZ is better than that of bzip2 but worse than gzip
   3.103 ++	  and LZO. Compression is slow.
   3.104 ++
   3.105 + config KERNEL_LZO
   3.106 + 	bool "LZO"
   3.107 + 	depends on HAVE_KERNEL_LZO
   3.108 +diff --git a/lib/Kconfig b/lib/Kconfig
   3.109 +index 6090314..45960e1 100644
   3.110 +--- a/lib/Kconfig
   3.111 ++++ b/lib/Kconfig
   3.112 +@@ -122,6 +122,9 @@ config DECOMPRESS_BZIP2
   3.113 + config DECOMPRESS_LZMA
   3.114 + 	tristate
   3.115 + 
   3.116 ++config DECOMPRESS_XZ
   3.117 ++	tristate
   3.118 ++
   3.119 + config DECOMPRESS_LZO
   3.120 + 	select LZO_DECOMPRESS
   3.121 + 	tristate
   3.122 +diff --git a/lib/Makefile b/lib/Makefile
   3.123 +index f2f98dd..06e3d8a 100644
   3.124 +--- a/lib/Makefile
   3.125 ++++ b/lib/Makefile
   3.126 +@@ -75,6 +75,7 @@ obj-$(CONFIG_RAID6_PQ) += raid6/
   3.127 + lib-$(CONFIG_DECOMPRESS_GZIP) += decompress_inflate.o
   3.128 + lib-$(CONFIG_DECOMPRESS_BZIP2) += decompress_bunzip2.o
   3.129 + lib-$(CONFIG_DECOMPRESS_LZMA) += decompress_unlzma.o
   3.130 ++lib-$(CONFIG_DECOMPRESS_XZ) += decompress_unxz.o
   3.131 + lib-$(CONFIG_DECOMPRESS_LZO) += decompress_unlzo.o
   3.132 + 
   3.133 + obj-$(CONFIG_TEXTSEARCH) += textsearch.o
   3.134 +diff --git a/lib/decompress.c b/lib/decompress.c
   3.135 +index a760681..3d766b7 100644
   3.136 +--- a/lib/decompress.c
   3.137 ++++ b/lib/decompress.c
   3.138 +@@ -8,6 +8,7 @@
   3.139 + 
   3.140 + #include <linux/decompress/bunzip2.h>
   3.141 + #include <linux/decompress/unlzma.h>
   3.142 ++#include <linux/decompress/unxz.h>
   3.143 + #include <linux/decompress/inflate.h>
   3.144 + #include <linux/decompress/unlzo.h>
   3.145 + 
   3.146 +@@ -23,6 +24,9 @@
   3.147 + #ifndef CONFIG_DECOMPRESS_LZMA
   3.148 + # define unlzma NULL
   3.149 + #endif
   3.150 ++#ifndef CONFIG_DECOMPRESS_XZ
   3.151 ++# define unxz NULL
   3.152 ++#endif
   3.153 + #ifndef CONFIG_DECOMPRESS_LZO
   3.154 + # define unlzo NULL
   3.155 + #endif
   3.156 +@@ -36,6 +40,7 @@ static const struct compress_format {
   3.157 + 	{ {037, 0236}, "gzip", gunzip },
   3.158 + 	{ {0x42, 0x5a}, "bzip2", bunzip2 },
   3.159 + 	{ {0x5d, 0x00}, "lzma", unlzma },
   3.160 ++	{ {0xfd, 0x37}, "xz", unxz },
   3.161 + 	{ {0x89, 0x4c}, "lzo", unlzo },
   3.162 + 	{ {0, 0}, NULL, NULL }
   3.163 + };
   3.164 +diff --git a/lib/decompress_unxz.c b/lib/decompress_unxz.c
   3.165 +new file mode 100644
   3.166 +index 0000000..cecd23d
   3.167 +--- /dev/null
   3.168 ++++ b/lib/decompress_unxz.c
   3.169 +@@ -0,0 +1,397 @@
   3.170 ++/*
   3.171 ++ * Wrapper for decompressing XZ-compressed kernel, initramfs, and initrd
   3.172 ++ *
   3.173 ++ * Author: Lasse Collin <lasse.collin@tukaani.org>
   3.174 ++ *
   3.175 ++ * This file has been put into the public domain.
   3.176 ++ * You can do whatever you want with this file.
   3.177 ++ */
   3.178 ++
   3.179 ++/*
   3.180 ++ * Important notes about in-place decompression
   3.181 ++ *
   3.182 ++ * At least on x86, the kernel is decompressed in place: the compressed data
   3.183 ++ * is placed to the end of the output buffer, and the decompressor overwrites
   3.184 ++ * most of the compressed data. There must be enough safety margin to
   3.185 ++ * guarantee that the write position is always behind the read position.
   3.186 ++ *
   3.187 ++ * The safety margin for XZ with LZMA2 or BCJ+LZMA2 is calculated below.
   3.188 ++ * Note that the margin with XZ is bigger than with Deflate (gzip)!
   3.189 ++ *
   3.190 ++ * The worst case for in-place decompression is that the beginning of
   3.191 ++ * the file is compressed extremely well, and the rest of the file is
   3.192 ++ * uncompressible. Thus, we must look for worst-case expansion when the
   3.193 ++ * compressor is encoding uncompressible data.
   3.194 ++ *
   3.195 ++ * The structure of the .xz file in case of a compresed kernel is as follows.
   3.196 ++ * Sizes (as bytes) of the fields are in parenthesis.
   3.197 ++ *
   3.198 ++ *    Stream Header (12)
   3.199 ++ *    Block Header:
   3.200 ++ *      Block Header (8-12)
   3.201 ++ *      Compressed Data (N)
   3.202 ++ *      Block Padding (0-3)
   3.203 ++ *      CRC32 (4)
   3.204 ++ *    Index (8-20)
   3.205 ++ *    Stream Footer (12)
   3.206 ++ *
   3.207 ++ * Normally there is exactly one Block, but let's assume that there are
   3.208 ++ * 2-4 Blocks just in case. Because Stream Header and also Block Header
   3.209 ++ * of the first Block don't make the decompressor produce any uncompressed
   3.210 ++ * data, we can ignore them from our calculations. Block Headers of possible
   3.211 ++ * additional Blocks have to be taken into account still. With these
   3.212 ++ * assumptions, it is safe to assume that the total header overhead is
   3.213 ++ * less than 128 bytes.
   3.214 ++ *
   3.215 ++ * Compressed Data contains LZMA2 or BCJ+LZMA2 encoded data. Since BCJ
   3.216 ++ * doesn't change the size of the data, it is enough to calculate the
   3.217 ++ * safety margin for LZMA2.
   3.218 ++ *
   3.219 ++ * LZMA2 stores the data in chunks. Each chunk has a header whose size is
   3.220 ++ * a maximum of 6 bytes, but to get round 2^n numbers, let's assume that
   3.221 ++ * the maximum chunk header size is 8 bytes. After the chunk header, there
   3.222 ++ * may be up to 64 KiB of actual payload in the chunk. Often the payload is
   3.223 ++ * quite a bit smaller though; to be safe, let's assume that an average
   3.224 ++ * chunk has only 32 KiB of payload.
   3.225 ++ *
   3.226 ++ * The maximum uncompressed size of the payload is 2 MiB. The minimum
   3.227 ++ * uncompressed size of the payload is in practice never less than the
   3.228 ++ * payload size itself. The LZMA2 format would allow uncompressed size
   3.229 ++ * to be less than the payload size, but no sane compressor creates such
   3.230 ++ * files. LZMA2 supports storing uncompressible data in uncompressed form,
   3.231 ++ * so there's never a need to create payloads whose uncompressed size is
   3.232 ++ * smaller than the compressed size.
   3.233 ++ *
   3.234 ++ * The assumption, that the uncompressed size of the payload is never
   3.235 ++ * smaller than the payload itself, is valid only when talking about
   3.236 ++ * the payload as a whole. It is possible that the payload has parts where
   3.237 ++ * the decompressor consumes more input than it produces output. Calculating
   3.238 ++ * the worst case for this would be tricky. Instead of trying to do that,
   3.239 ++ * let's simply make sure that the decompressor never overwrites any bytes
   3.240 ++ * of the payload which it is currently reading.
   3.241 ++ *
   3.242 ++ * Now we have enough information to calculate the safety margin. We need
   3.243 ++ *   - 128 bytes for the .xz file format headers;
   3.244 ++ *   - 8 bytes per every 32 KiB of uncompressed size (one LZMA2 chunk header
   3.245 ++ *     per chunk, each chunk having average payload size of 32 KiB); and
   3.246 ++ *   - 64 KiB (biggest possible LZMA2 chunk payload size) to make sure that
   3.247 ++ *     the decompressor never overwrites anything from the LZMA2 chunk
   3.248 ++ *     payload it is currently reading.
   3.249 ++ *
   3.250 ++ * We get the following formula:
   3.251 ++ *
   3.252 ++ *    safety_margin = 128 + uncompressed_size * 8 / 32768 + 65536
   3.253 ++ *                  = 128 + (uncompressed_size >> 12) + 65536
   3.254 ++ *
   3.255 ++ * For comparision, according to arch/x86/boot/compressed/misc.c, the
   3.256 ++ * equivalent formula for Deflate is this:
   3.257 ++ *
   3.258 ++ *    safety_margin = 18 + (uncompressed_size >> 12) + 32768
   3.259 ++ *
   3.260 ++ * Thus, when updating Deflate-only in-place kernel decompressor to
   3.261 ++ * support XZ, the fixed overhead has to be increased from 18+32768 bytes
   3.262 ++ * to 128+65536 bytes.
   3.263 ++ */
   3.264 ++
   3.265 ++/*
   3.266 ++ * STATIC is defined to "static" if we are being built for kernel
   3.267 ++ * decompression (pre-boot code). <linux/decompress/mm.h> will define
   3.268 ++ * STATIC to empty if it wasn't already defined. Since we will need to
   3.269 ++ * know later if we are being used for kernel decompression, we define
   3.270 ++ * XZ_PREBOOT here.
   3.271 ++ */
   3.272 ++#ifdef STATIC
   3.273 ++#	define XZ_PREBOOT
   3.274 ++#endif
   3.275 ++#ifdef __KERNEL__
   3.276 ++#	include <linux/decompress/mm.h>
   3.277 ++#endif
   3.278 ++#define XZ_EXTERN STATIC
   3.279 ++
   3.280 ++#ifndef XZ_PREBOOT
   3.281 ++#	include <linux/slab.h>
   3.282 ++#	include <linux/xz.h>
   3.283 ++#else
   3.284 ++/*
   3.285 ++ * Use the internal CRC32 code instead of kernel's CRC32 module, which
   3.286 ++ * is not available in early phase of booting.
   3.287 ++ */
   3.288 ++#define XZ_INTERNAL_CRC32 1
   3.289 ++
   3.290 ++/*
   3.291 ++ * For boot time use, we enable only the BCJ filter of the current
   3.292 ++ * architecture or none if no BCJ filter is available for the architecture.
   3.293 ++ */
   3.294 ++#ifdef CONFIG_X86
   3.295 ++#	define XZ_DEC_X86
   3.296 ++#endif
   3.297 ++#ifdef CONFIG_PPC
   3.298 ++#	define XZ_DEC_POWERPC
   3.299 ++#endif
   3.300 ++#ifdef CONFIG_ARM
   3.301 ++#	define XZ_DEC_ARM
   3.302 ++#endif
   3.303 ++#ifdef CONFIG_IA64
   3.304 ++#	define XZ_DEC_IA64
   3.305 ++#endif
   3.306 ++#ifdef CONFIG_SPARC
   3.307 ++#	define XZ_DEC_SPARC
   3.308 ++#endif
   3.309 ++
   3.310 ++/*
   3.311 ++ * This will get the basic headers so that memeq() and others
   3.312 ++ * can be defined.
   3.313 ++ */
   3.314 ++#include "xz/xz_private.h"
   3.315 ++
   3.316 ++/*
   3.317 ++ * Replace the normal allocation functions with the versions from
   3.318 ++ * <linux/decompress/mm.h>. vfree() needs to support vfree(NULL)
   3.319 ++ * when XZ_DYNALLOC is used, but the pre-boot free() doesn't support it.
   3.320 ++ * Workaround it here because the other decompressors don't need it.
   3.321 ++ */
   3.322 ++#undef kmalloc
   3.323 ++#undef kfree
   3.324 ++#undef vmalloc
   3.325 ++#undef vfree
   3.326 ++#define kmalloc(size, flags) malloc(size)
   3.327 ++#define kfree(ptr) free(ptr)
   3.328 ++#define vmalloc(size) malloc(size)
   3.329 ++#define vfree(ptr) do { if (ptr != NULL) free(ptr); } while (0)
   3.330 ++
   3.331 ++/*
   3.332 ++ * FIXME: Not all basic memory functions are provided in architecture-specific
   3.333 ++ * files (yet). We define our own versions here for now, but this should be
   3.334 ++ * only a temporary solution.
   3.335 ++ *
   3.336 ++ * memeq and memzero are not used much and any remotely sane implementation
   3.337 ++ * is fast enough. memcpy/memmove speed matters in multi-call mode, but
   3.338 ++ * the kernel image is decompressed in single-call mode, in which only
   3.339 ++ * memcpy speed can matter and only if there is a lot of uncompressible data
   3.340 ++ * (LZMA2 stores uncompressible chunks in uncompressed form). Thus, the
   3.341 ++ * functions below should just be kept small; it's probably not worth
   3.342 ++ * optimizing for speed.
   3.343 ++ */
   3.344 ++
   3.345 ++#ifndef memeq
   3.346 ++static bool memeq(const void *a, const void *b, size_t size)
   3.347 ++{
   3.348 ++	const uint8_t *x = a;
   3.349 ++	const uint8_t *y = b;
   3.350 ++	size_t i;
   3.351 ++
   3.352 ++	for (i = 0; i < size; ++i)
   3.353 ++		if (x[i] != y[i])
   3.354 ++			return false;
   3.355 ++
   3.356 ++	return true;
   3.357 ++}
   3.358 ++#endif
   3.359 ++
   3.360 ++#ifndef memzero
   3.361 ++static void memzero(void *buf, size_t size)
   3.362 ++{
   3.363 ++	uint8_t *b = buf;
   3.364 ++	uint8_t *e = b + size;
   3.365 ++
   3.366 ++	while (b != e)
   3.367 ++		*b++ = '\0';
   3.368 ++}
   3.369 ++#endif
   3.370 ++
   3.371 ++#ifndef memmove
   3.372 ++/* Not static to avoid a conflict with the prototype in the Linux headers. */
   3.373 ++void *memmove(void *dest, const void *src, size_t size)
   3.374 ++{
   3.375 ++	uint8_t *d = dest;
   3.376 ++	const uint8_t *s = src;
   3.377 ++	size_t i;
   3.378 ++
   3.379 ++	if (d < s) {
   3.380 ++		for (i = 0; i < size; ++i)
   3.381 ++			d[i] = s[i];
   3.382 ++	} else if (d > s) {
   3.383 ++		i = size;
   3.384 ++		while (i-- > 0)
   3.385 ++			d[i] = s[i];
   3.386 ++	}
   3.387 ++
   3.388 ++	return dest;
   3.389 ++}
   3.390 ++#endif
   3.391 ++
   3.392 ++/*
   3.393 ++ * Since we need memmove anyway, would use it as memcpy too.
   3.394 ++ * Commented out for now to avoid breaking things.
   3.395 ++ */
   3.396 ++/*
   3.397 ++#ifndef memcpy
   3.398 ++#	define memcpy memmove
   3.399 ++#endif
   3.400 ++*/
   3.401 ++
   3.402 ++#include "xz/xz_crc32.c"
   3.403 ++#include "xz/xz_dec_stream.c"
   3.404 ++#include "xz/xz_dec_lzma2.c"
   3.405 ++#include "xz/xz_dec_bcj.c"
   3.406 ++
   3.407 ++#endif /* XZ_PREBOOT */
   3.408 ++
   3.409 ++/* Size of the input and output buffers in multi-call mode */
   3.410 ++#define XZ_IOBUF_SIZE 4096
   3.411 ++
   3.412 ++/*
   3.413 ++ * This function implements the API defined in <linux/decompress/generic.h>.
   3.414 ++ *
   3.415 ++ * This wrapper will automatically choose single-call or multi-call mode
   3.416 ++ * of the native XZ decoder API. The single-call mode can be used only when
   3.417 ++ * both input and output buffers are available as a single chunk, i.e. when
   3.418 ++ * fill() and flush() won't be used.
   3.419 ++ */
   3.420 ++STATIC int INIT unxz(unsigned char *in, int in_size,
   3.421 ++		     int (*fill)(void *dest, unsigned int size),
   3.422 ++		     int (*flush)(void *src, unsigned int size),
   3.423 ++		     unsigned char *out, int *in_used,
   3.424 ++		     void (*error)(char *x))
   3.425 ++{
   3.426 ++	struct xz_buf b;
   3.427 ++	struct xz_dec *s;
   3.428 ++	enum xz_ret ret;
   3.429 ++	bool must_free_in = false;
   3.430 ++
   3.431 ++#if XZ_INTERNAL_CRC32
   3.432 ++	xz_crc32_init();
   3.433 ++#endif
   3.434 ++
   3.435 ++	if (in_used != NULL)
   3.436 ++		*in_used = 0;
   3.437 ++
   3.438 ++	if (fill == NULL && flush == NULL)
   3.439 ++		s = xz_dec_init(XZ_SINGLE, 0);
   3.440 ++	else
   3.441 ++		s = xz_dec_init(XZ_DYNALLOC, (uint32_t)-1);
   3.442 ++
   3.443 ++	if (s == NULL)
   3.444 ++		goto error_alloc_state;
   3.445 ++
   3.446 ++	if (flush == NULL) {
   3.447 ++		b.out = out;
   3.448 ++		b.out_size = (size_t)-1;
   3.449 ++	} else {
   3.450 ++		b.out_size = XZ_IOBUF_SIZE;
   3.451 ++		b.out = malloc(XZ_IOBUF_SIZE);
   3.452 ++		if (b.out == NULL)
   3.453 ++			goto error_alloc_out;
   3.454 ++	}
   3.455 ++
   3.456 ++	if (in == NULL) {
   3.457 ++		must_free_in = true;
   3.458 ++		in = malloc(XZ_IOBUF_SIZE);
   3.459 ++		if (in == NULL)
   3.460 ++			goto error_alloc_in;
   3.461 ++	}
   3.462 ++
   3.463 ++	b.in = in;
   3.464 ++	b.in_pos = 0;
   3.465 ++	b.in_size = in_size;
   3.466 ++	b.out_pos = 0;
   3.467 ++
   3.468 ++	if (fill == NULL && flush == NULL) {
   3.469 ++		ret = xz_dec_run(s, &b);
   3.470 ++	} else {
   3.471 ++		do {
   3.472 ++			if (b.in_pos == b.in_size && fill != NULL) {
   3.473 ++				if (in_used != NULL)
   3.474 ++					*in_used += b.in_pos;
   3.475 ++
   3.476 ++				b.in_pos = 0;
   3.477 ++
   3.478 ++				in_size = fill(in, XZ_IOBUF_SIZE);
   3.479 ++				if (in_size < 0) {
   3.480 ++					/*
   3.481 ++					 * This isn't an optimal error code
   3.482 ++					 * but it probably isn't worth making
   3.483 ++					 * a new one either.
   3.484 ++					 */
   3.485 ++					ret = XZ_BUF_ERROR;
   3.486 ++					break;
   3.487 ++				}
   3.488 ++
   3.489 ++				b.in_size = in_size;
   3.490 ++			}
   3.491 ++
   3.492 ++			ret = xz_dec_run(s, &b);
   3.493 ++
   3.494 ++			if (flush != NULL && (b.out_pos == b.out_size
   3.495 ++					|| (ret != XZ_OK && b.out_pos > 0))) {
   3.496 ++				/*
   3.497 ++				 * Setting ret here may hide an error
   3.498 ++				 * returned by xz_dec_run(), but probably
   3.499 ++				 * it's not too bad.
   3.500 ++				 */
   3.501 ++				if (flush(b.out, b.out_pos) != (int)b.out_pos)
   3.502 ++					ret = XZ_BUF_ERROR;
   3.503 ++
   3.504 ++				b.out_pos = 0;
   3.505 ++			}
   3.506 ++		} while (ret == XZ_OK);
   3.507 ++
   3.508 ++		if (must_free_in)
   3.509 ++			free(in);
   3.510 ++
   3.511 ++		if (flush != NULL)
   3.512 ++			free(b.out);
   3.513 ++	}
   3.514 ++
   3.515 ++	if (in_used != NULL)
   3.516 ++		*in_used += b.in_pos;
   3.517 ++
   3.518 ++	xz_dec_end(s);
   3.519 ++
   3.520 ++	switch (ret) {
   3.521 ++	case XZ_STREAM_END:
   3.522 ++		return 0;
   3.523 ++
   3.524 ++	case XZ_MEM_ERROR:
   3.525 ++		/* This can occur only in multi-call mode. */
   3.526 ++		error("XZ decompressor ran out of memory");
   3.527 ++		break;
   3.528 ++
   3.529 ++	case XZ_FORMAT_ERROR:
   3.530 ++		error("Input is not in the XZ format (wrong magic bytes)");
   3.531 ++		break;
   3.532 ++
   3.533 ++	case XZ_OPTIONS_ERROR:
   3.534 ++		error("Input was encoded with settings that are not "
   3.535 ++				"supported by this XZ decoder");
   3.536 ++		break;
   3.537 ++
   3.538 ++	case XZ_DATA_ERROR:
   3.539 ++	case XZ_BUF_ERROR:
   3.540 ++		error("XZ-compressed data is corrupt");
   3.541 ++		break;
   3.542 ++
   3.543 ++	default:
   3.544 ++		error("Bug in the XZ decompressor");
   3.545 ++		break;
   3.546 ++	}
   3.547 ++
   3.548 ++	return -1;
   3.549 ++
   3.550 ++error_alloc_in:
   3.551 ++	if (flush != NULL)
   3.552 ++		free(b.out);
   3.553 ++
   3.554 ++error_alloc_out:
   3.555 ++	xz_dec_end(s);
   3.556 ++
   3.557 ++error_alloc_state:
   3.558 ++	error("XZ decompressor ran out of memory");
   3.559 ++	return -1;
   3.560 ++}
   3.561 ++
   3.562 ++/*
   3.563 ++ * This macro is used by architecture-specific files to decompress
   3.564 ++ * the kernel image.
   3.565 ++ */
   3.566 ++#define decompress unxz
   3.567 +diff --git a/scripts/gen_initramfs_list.sh b/scripts/gen_initramfs_list.sh
   3.568 +index 5958fff..55caecd 100644
   3.569 +--- a/scripts/gen_initramfs_list.sh
   3.570 ++++ b/scripts/gen_initramfs_list.sh
   3.571 +@@ -243,6 +243,8 @@ case "$arg" in
   3.572 + 		echo "$output_file" | grep -q "\.gz$" && compr="gzip -9 -f"
   3.573 + 		echo "$output_file" | grep -q "\.bz2$" && compr="bzip2 -9 -f"
   3.574 + 		echo "$output_file" | grep -q "\.lzma$" && compr="lzma -9 -f"
   3.575 ++		echo "$output_file" | grep -q "\.xz$" && \
   3.576 ++				compr="xz --check=crc32 --lzma2=dict=1MiB"
   3.577 + 		echo "$output_file" | grep -q "\.lzo$" && compr="lzop -9 -f"
   3.578 + 		echo "$output_file" | grep -q "\.cpio$" && compr="cat"
   3.579 + 		shift
   3.580 +diff --git a/usr/Kconfig b/usr/Kconfig
   3.581 +index e2721f5..9f51a29 100644
   3.582 +--- a/usr/Kconfig
   3.583 ++++ b/usr/Kconfig
   3.584 +@@ -72,6 +72,18 @@ config RD_LZMA
   3.585 + 	  Support loading of a LZMA encoded initial ramdisk or cpio buffer
   3.586 + 	  If unsure, say N.
   3.587 + 
   3.588 ++config RD_XZ
   3.589 ++	bool "Support initial ramdisks compressed using XZ"
   3.590 ++	depends on BLK_DEV_INITRD && XZ_DEC=y
   3.591 ++	select DECOMPRESS_XZ
   3.592 ++	help
   3.593 ++	  Support loading of a XZ encoded initial ramdisk or cpio buffer.
   3.594 ++
   3.595 ++	  If this option is inactive, say Y to "XZ decompression support"
   3.596 ++	  under "Library routines" first.
   3.597 ++
   3.598 ++	  If unsure, say N.
   3.599 ++
   3.600 + config RD_LZO
   3.601 + 	bool "Support initial ramdisks compressed using LZO" if EMBEDDED
   3.602 + 	default !EMBEDDED
   3.603 +@@ -139,6 +151,15 @@ config INITRAMFS_COMPRESSION_LZMA
   3.604 + 	  three. Compression is slowest. The initramfs size is about 33%
   3.605 + 	  smaller with LZMA in comparison to gzip.
   3.606 + 
   3.607 ++config INITRAMFS_COMPRESSION_XZ
   3.608 ++	bool "XZ"
   3.609 ++	depends on RD_XZ
   3.610 ++	help
   3.611 ++	  XZ uses the LZMA2 algorithm. The initramfs size is about 30%
   3.612 ++	  smaller with XZ in comparison to gzip. Decompression speed
   3.613 ++	  is better than that of bzip2 but worse than gzip and LZO.
   3.614 ++	  Compression is slow.
   3.615 ++
   3.616 + config INITRAMFS_COMPRESSION_LZO
   3.617 + 	bool "LZO"
   3.618 + 	depends on RD_LZO
   3.619 +diff --git a/usr/Makefile b/usr/Makefile
   3.620 +index 6b4b6da..5845a13 100644
   3.621 +--- a/usr/Makefile
   3.622 ++++ b/usr/Makefile
   3.623 +@@ -15,6 +15,9 @@ suffix_$(CONFIG_INITRAMFS_COMPRESSION_BZIP2)  = .bz2
   3.624 + # Lzma
   3.625 + suffix_$(CONFIG_INITRAMFS_COMPRESSION_LZMA)   = .lzma
   3.626 + 
   3.627 ++# XZ
   3.628 ++suffix_$(CONFIG_INITRAMFS_COMPRESSION_XZ)     = .xz
   3.629 ++
   3.630 + # Lzo
   3.631 + suffix_$(CONFIG_INITRAMFS_COMPRESSION_LZO)   = .lzo
   3.632 + 
   3.633 +@@ -48,7 +51,7 @@ endif
   3.634 + quiet_cmd_initfs = GEN     $@
   3.635 +       cmd_initfs = $(initramfs) -o $@ $(ramfs-args) $(ramfs-input)
   3.636 + 
   3.637 +-targets := initramfs_data.cpio.gz initramfs_data.cpio.bz2 initramfs_data.cpio.lzma initramfs_data.cpio.lzo initramfs_data.cpio
   3.638 ++targets := initramfs_data.cpio.gz initramfs_data.cpio.bz2 initramfs_data.cpio.lzma initramfs_data.cpio.xz initramfs_data.cpio.lzo initramfs_data.cpio
   3.639 + # do not try to update files included in initramfs
   3.640 + $(deps_initramfs): ;
   3.641 + 
     4.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     4.2 +++ b/linux/stuff/003-squashfs-x86-support-xz-compressed-kernel.patch	Tue Dec 14 21:45:09 2010 +0000
     4.3 @@ -0,0 +1,128 @@
     4.4 +From: Lasse Collin <lasse.collin@tukaani.org>
     4.5 +Date: Thu, 2 Dec 2010 19:14:57 +0000 (+0200)
     4.6 +Subject: x86: Support XZ-compressed kernel
     4.7 +X-Git-Url: http://git.kernel.org/?p=linux%2Fkernel%2Fgit%2Fpkl%2Fsquashfs-xz.git;a=commitdiff_plain;h=d4ad78414e5312126127b1f88cdaf8584af3eef1
     4.8 +
     4.9 +x86: Support XZ-compressed kernel
    4.10 +
    4.11 +This integrates the XZ decompression code to the x86
    4.12 +pre-boot code.
    4.13 +
    4.14 +mkpiggy.c is updated to reserve about 32 KiB more buffer safety
    4.15 +margin for kernel decompression. It is done unconditionally for
    4.16 +all decompressors to keep the code simpler.
    4.17 +
    4.18 +The XZ decompressor needs around 30 KiB of heap, so the heap size
    4.19 +is increased to 32 KiB on both x86-32 and x86-64.
    4.20 +
    4.21 +Documentation/x86/boot.txt is updated to list the XZ magic number.
    4.22 +
    4.23 +With the x86 BCJ filter in XZ, XZ-compressed x86 kernel tends to be
    4.24 +a few percent smaller than the equivalent LZMA-compressed kernel.
    4.25 +
    4.26 +Signed-off-by: Lasse Collin <lasse.collin@tukaani.org>
    4.27 +---
    4.28 +
    4.29 +diff --git a/Documentation/x86/boot.txt b/Documentation/x86/boot.txt
    4.30 +index 30b43e1..3988cde 100644
    4.31 +--- a/Documentation/x86/boot.txt
    4.32 ++++ b/Documentation/x86/boot.txt
    4.33 +@@ -621,9 +621,9 @@ Protocol:	2.08+
    4.34 +   The payload may be compressed. The format of both the compressed and
    4.35 +   uncompressed data should be determined using the standard magic
    4.36 +   numbers.  The currently supported compression formats are gzip
    4.37 +-  (magic numbers 1F 8B or 1F 9E), bzip2 (magic number 42 5A) and LZMA
    4.38 +-  (magic number 5D 00).  The uncompressed payload is currently always ELF
    4.39 +-  (magic number 7F 45 4C 46).
    4.40 ++  (magic numbers 1F 8B or 1F 9E), bzip2 (magic number 42 5A), LZMA
    4.41 ++  (magic number 5D 00), and XZ (magic number FD 37).  The uncompressed
    4.42 ++  payload is currently always ELF (magic number 7F 45 4C 46).
    4.43 +   
    4.44 + Field name:	payload_length
    4.45 + Type:		read
    4.46 +diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
    4.47 +index cea0cd9..f3db0d7 100644
    4.48 +--- a/arch/x86/Kconfig
    4.49 ++++ b/arch/x86/Kconfig
    4.50 +@@ -51,6 +51,7 @@ config X86
    4.51 + 	select HAVE_KERNEL_GZIP
    4.52 + 	select HAVE_KERNEL_BZIP2
    4.53 + 	select HAVE_KERNEL_LZMA
    4.54 ++	select HAVE_KERNEL_XZ
    4.55 + 	select HAVE_KERNEL_LZO
    4.56 + 	select HAVE_HW_BREAKPOINT
    4.57 + 	select HAVE_MIXED_BREAKPOINTS_REGS
    4.58 +diff --git a/arch/x86/boot/compressed/Makefile b/arch/x86/boot/compressed/Makefile
    4.59 +index 0c22955..09664ef 100644
    4.60 +--- a/arch/x86/boot/compressed/Makefile
    4.61 ++++ b/arch/x86/boot/compressed/Makefile
    4.62 +@@ -4,7 +4,7 @@
    4.63 + # create a compressed vmlinux image from the original vmlinux
    4.64 + #
    4.65 + 
    4.66 +-targets := vmlinux.lds vmlinux vmlinux.bin vmlinux.bin.gz vmlinux.bin.bz2 vmlinux.bin.lzma vmlinux.bin.lzo head_$(BITS).o misc.o string.o cmdline.o early_serial_console.o piggy.o
    4.67 ++targets := vmlinux.lds vmlinux vmlinux.bin vmlinux.bin.gz vmlinux.bin.bz2 vmlinux.bin.lzma vmlinux.bin.xz vmlinux.bin.lzo head_$(BITS).o misc.o string.o cmdline.o early_serial_console.o piggy.o
    4.68 + 
    4.69 + KBUILD_CFLAGS := -m$(BITS) -D__KERNEL__ $(LINUX_INCLUDE) -O2
    4.70 + KBUILD_CFLAGS += -fno-strict-aliasing -fPIC
    4.71 +@@ -49,12 +49,15 @@ $(obj)/vmlinux.bin.bz2: $(vmlinux.bin.all-y) FORCE
    4.72 + 	$(call if_changed,bzip2)
    4.73 + $(obj)/vmlinux.bin.lzma: $(vmlinux.bin.all-y) FORCE
    4.74 + 	$(call if_changed,lzma)
    4.75 ++$(obj)/vmlinux.bin.xz: $(vmlinux.bin.all-y) FORCE
    4.76 ++	$(call if_changed,xzkern)
    4.77 + $(obj)/vmlinux.bin.lzo: $(vmlinux.bin.all-y) FORCE
    4.78 + 	$(call if_changed,lzo)
    4.79 + 
    4.80 + suffix-$(CONFIG_KERNEL_GZIP)	:= gz
    4.81 + suffix-$(CONFIG_KERNEL_BZIP2)	:= bz2
    4.82 + suffix-$(CONFIG_KERNEL_LZMA)	:= lzma
    4.83 ++suffix-$(CONFIG_KERNEL_XZ)	:= xz
    4.84 + suffix-$(CONFIG_KERNEL_LZO) 	:= lzo
    4.85 + 
    4.86 + quiet_cmd_mkpiggy = MKPIGGY $@
    4.87 +diff --git a/arch/x86/boot/compressed/misc.c b/arch/x86/boot/compressed/misc.c
    4.88 +index 8f7bef8..6d4cab7 100644
    4.89 +--- a/arch/x86/boot/compressed/misc.c
    4.90 ++++ b/arch/x86/boot/compressed/misc.c
    4.91 +@@ -139,6 +139,10 @@ static int lines, cols;
    4.92 + #include "../../../../lib/decompress_unlzma.c"
    4.93 + #endif
    4.94 + 
    4.95 ++#ifdef CONFIG_KERNEL_XZ
    4.96 ++#include "../../../../lib/decompress_unxz.c"
    4.97 ++#endif
    4.98 ++
    4.99 + #ifdef CONFIG_KERNEL_LZO
   4.100 + #include "../../../../lib/decompress_unlzo.c"
   4.101 + #endif
   4.102 +diff --git a/arch/x86/boot/compressed/mkpiggy.c b/arch/x86/boot/compressed/mkpiggy.c
   4.103 +index 5c22812..646aa78 100644
   4.104 +--- a/arch/x86/boot/compressed/mkpiggy.c
   4.105 ++++ b/arch/x86/boot/compressed/mkpiggy.c
   4.106 +@@ -74,7 +74,7 @@ int main(int argc, char *argv[])
   4.107 + 
   4.108 + 	offs = (olen > ilen) ? olen - ilen : 0;
   4.109 + 	offs += olen >> 12;	/* Add 8 bytes for each 32K block */
   4.110 +-	offs += 32*1024 + 18;	/* Add 32K + 18 bytes slack */
   4.111 ++	offs += 64*1024 + 128;	/* Add 64K + 128 bytes slack */
   4.112 + 	offs = (offs+4095) & ~4095; /* Round to a 4K boundary */
   4.113 + 
   4.114 + 	printf(".section \".rodata..compressed\",\"a\",@progbits\n");
   4.115 +diff --git a/arch/x86/include/asm/boot.h b/arch/x86/include/asm/boot.h
   4.116 +index 3b62ab5..5e1a2ee 100644
   4.117 +--- a/arch/x86/include/asm/boot.h
   4.118 ++++ b/arch/x86/include/asm/boot.h
   4.119 +@@ -32,11 +32,7 @@
   4.120 + #define BOOT_HEAP_SIZE             0x400000
   4.121 + #else /* !CONFIG_KERNEL_BZIP2 */
   4.122 + 
   4.123 +-#ifdef CONFIG_X86_64
   4.124 +-#define BOOT_HEAP_SIZE	0x7000
   4.125 +-#else
   4.126 +-#define BOOT_HEAP_SIZE	0x4000
   4.127 +-#endif
   4.128 ++#define BOOT_HEAP_SIZE	0x8000
   4.129 + 
   4.130 + #endif /* !CONFIG_KERNEL_BZIP2 */
   4.131 + 
     5.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     5.2 +++ b/linux/stuff/004-squashfs-add-xz-compression-support.patch	Tue Dec 14 21:45:09 2010 +0000
     5.3 @@ -0,0 +1,183 @@
     5.4 +From: Phillip Lougher <phillip@lougher.demon.co.uk>
     5.5 +Date: Thu, 9 Dec 2010 02:02:29 +0000 (+0000)
     5.6 +Subject: Squashfs: add XZ compression support
     5.7 +X-Git-Url: http://git.kernel.org/?p=linux%2Fkernel%2Fgit%2Fpkl%2Fsquashfs-xz.git;a=commitdiff_plain;h=d3e6969b9ff1f3a3c6bf3da71433c77046aa80e4
     5.8 +
     5.9 +Squashfs: add XZ compression support
    5.10 +
    5.11 +Add XZ decompressor wrapper code.
    5.12 +
    5.13 +Signed-off-by: Phillip Lougher <phillip@lougher.demon.co.uk>
    5.14 +---
    5.15 +
    5.16 +diff --git a/fs/squashfs/squashfs_fs.h b/fs/squashfs/squashfs_fs.h
    5.17 +index c5137fc..39533fe 100644
    5.18 +--- a/fs/squashfs/squashfs_fs.h
    5.19 ++++ b/fs/squashfs/squashfs_fs.h
    5.20 +@@ -238,6 +238,7 @@ struct meta_index {
    5.21 + #define ZLIB_COMPRESSION	1
    5.22 + #define LZMA_COMPRESSION	2
    5.23 + #define LZO_COMPRESSION		3
    5.24 ++#define XZ_COMPRESSION		4
    5.25 + 
    5.26 + struct squashfs_super_block {
    5.27 + 	__le32			s_magic;
    5.28 +diff --git a/fs/squashfs/xz_wrapper.c b/fs/squashfs/xz_wrapper.c
    5.29 +new file mode 100644
    5.30 +index 0000000..053fe35
    5.31 +--- /dev/null
    5.32 ++++ b/fs/squashfs/xz_wrapper.c
    5.33 +@@ -0,0 +1,153 @@
    5.34 ++/*
    5.35 ++ * Squashfs - a compressed read only filesystem for Linux
    5.36 ++ *
    5.37 ++ * Copyright (c) 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
    5.38 ++ * Phillip Lougher <phillip@lougher.demon.co.uk>
    5.39 ++ *
    5.40 ++ * This program is free software; you can redistribute it and/or
    5.41 ++ * modify it under the terms of the GNU General Public License
    5.42 ++ * as published by the Free Software Foundation; either version 2,
    5.43 ++ * or (at your option) any later version.
    5.44 ++ *
    5.45 ++ * This program is distributed in the hope that it will be useful,
    5.46 ++ * but WITHOUT ANY WARRANTY; without even the implied warranty of
    5.47 ++ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    5.48 ++ * GNU General Public License for more details.
    5.49 ++ *
    5.50 ++ * You should have received a copy of the GNU General Public License
    5.51 ++ * along with this program; if not, write to the Free Software
    5.52 ++ * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
    5.53 ++ *
    5.54 ++ * xz_wrapper.c
    5.55 ++ */
    5.56 ++
    5.57 ++
    5.58 ++#include <linux/mutex.h>
    5.59 ++#include <linux/buffer_head.h>
    5.60 ++#include <linux/slab.h>
    5.61 ++#include <linux/xz.h>
    5.62 ++
    5.63 ++#include "squashfs_fs.h"
    5.64 ++#include "squashfs_fs_sb.h"
    5.65 ++#include "squashfs_fs_i.h"
    5.66 ++#include "squashfs.h"
    5.67 ++#include "decompressor.h"
    5.68 ++
    5.69 ++struct squashfs_xz {
    5.70 ++	struct xz_dec *state;
    5.71 ++	struct xz_buf buf;
    5.72 ++};
    5.73 ++
    5.74 ++static void *squashfs_xz_init(struct squashfs_sb_info *msblk)
    5.75 ++{
    5.76 ++        int block_size = max_t(int, msblk->block_size, SQUASHFS_METADATA_SIZE);
    5.77 ++
    5.78 ++        struct squashfs_xz *stream = kmalloc(sizeof(*stream), GFP_KERNEL);
    5.79 ++        if (stream == NULL)
    5.80 ++                goto failed;
    5.81 ++	stream->state = xz_dec_init(XZ_PREALLOC, block_size);
    5.82 ++	if (stream->state == NULL)
    5.83 ++		goto failed;
    5.84 ++
    5.85 ++	return stream;
    5.86 ++
    5.87 ++failed:
    5.88 ++	ERROR("Failed to allocate xz workspace\n");
    5.89 ++	kfree(stream);
    5.90 ++	return NULL;
    5.91 ++}
    5.92 ++
    5.93 ++
    5.94 ++static void squashfs_xz_free(void *strm)
    5.95 ++{
    5.96 ++	struct squashfs_xz *stream = strm;
    5.97 ++
    5.98 ++	if (stream) {
    5.99 ++		xz_dec_end(stream->state);
   5.100 ++		kfree(stream);
   5.101 ++	}
   5.102 ++}
   5.103 ++
   5.104 ++
   5.105 ++static int squashfs_xz_uncompress(struct squashfs_sb_info *msblk, void **buffer,
   5.106 ++	struct buffer_head **bh, int b, int offset, int length, int srclength,
   5.107 ++	int pages)
   5.108 ++{
   5.109 ++	enum xz_ret xz_err;
   5.110 ++	int avail, total = 0, k = 0, page = 0;
   5.111 ++	struct squashfs_xz *stream = msblk->stream;
   5.112 ++
   5.113 ++	mutex_lock(&msblk->read_data_mutex);
   5.114 ++
   5.115 ++	xz_dec_reset(stream->state);
   5.116 ++	stream->buf.in_pos = 0;
   5.117 ++	stream->buf.in_size = 0;
   5.118 ++	stream->buf.out_pos = 0;
   5.119 ++	stream->buf.out_size = PAGE_CACHE_SIZE;
   5.120 ++	stream->buf.out = buffer[page++];
   5.121 ++
   5.122 ++	do {
   5.123 ++		if (stream->buf.in_pos == stream->buf.in_size && k < b) {
   5.124 ++			avail = min(length, msblk->devblksize - offset);
   5.125 ++			length -= avail;
   5.126 ++			wait_on_buffer(bh[k]);
   5.127 ++			if (!buffer_uptodate(bh[k]))
   5.128 ++				goto release_mutex;
   5.129 ++
   5.130 ++			if (avail == 0) {
   5.131 ++				offset = 0;
   5.132 ++				put_bh(bh[k++]);
   5.133 ++				continue;
   5.134 ++			}
   5.135 ++
   5.136 ++			stream->buf.in = bh[k]->b_data + offset;
   5.137 ++			stream->buf.in_size = avail;
   5.138 ++			stream->buf.in_pos = 0;
   5.139 ++			offset = 0;
   5.140 ++		}
   5.141 ++
   5.142 ++		if (stream->buf.out_pos == stream->buf.out_size
   5.143 ++							&& page < pages) {
   5.144 ++			stream->buf.out = buffer[page++];
   5.145 ++			stream->buf.out_pos = 0;
   5.146 ++			total += PAGE_CACHE_SIZE;
   5.147 ++		}
   5.148 ++
   5.149 ++		xz_err = xz_dec_run(stream->state, &stream->buf);
   5.150 ++
   5.151 ++		if (stream->buf.in_pos == stream->buf.in_size && k < b)
   5.152 ++			put_bh(bh[k++]);
   5.153 ++	} while (xz_err == XZ_OK);
   5.154 ++
   5.155 ++	if (xz_err != XZ_STREAM_END) {
   5.156 ++		ERROR("xz_dec_run error, data probably corrupt\n");
   5.157 ++		goto release_mutex;
   5.158 ++	}
   5.159 ++
   5.160 ++	if (k < b) {
   5.161 ++		ERROR("xz_uncompress error, input remaining\n");
   5.162 ++		goto release_mutex;
   5.163 ++	}
   5.164 ++
   5.165 ++	total += stream->buf.out_pos;
   5.166 ++	mutex_unlock(&msblk->read_data_mutex);
   5.167 ++	return total;
   5.168 ++
   5.169 ++release_mutex:
   5.170 ++	mutex_unlock(&msblk->read_data_mutex);
   5.171 ++
   5.172 ++	for (; k < b; k++)
   5.173 ++		put_bh(bh[k]);
   5.174 ++
   5.175 ++	return -EIO;
   5.176 ++}
   5.177 ++
   5.178 ++const struct squashfs_decompressor squashfs_xz_comp_ops = {
   5.179 ++	.init = squashfs_xz_init,
   5.180 ++	.free = squashfs_xz_free,
   5.181 ++	.decompress = squashfs_xz_uncompress,
   5.182 ++	.id = XZ_COMPRESSION,
   5.183 ++	.name = "xz",
   5.184 ++	.supported = 1
   5.185 ++};
   5.186 ++
     6.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     6.2 +++ b/linux/stuff/005-squashfs-add-xz-compression-configuration-option.patch	Tue Dec 14 21:45:09 2010 +0000
     6.3 @@ -0,0 +1,86 @@
     6.4 +From: Phillip Lougher <phillip@lougher.demon.co.uk>
     6.5 +Date: Thu, 9 Dec 2010 02:08:31 +0000 (+0000)
     6.6 +Subject: Squashfs: Add XZ compression configuration option
     6.7 +X-Git-Url: http://git.kernel.org/?p=linux%2Fkernel%2Fgit%2Fpkl%2Fsquashfs-xz.git;a=commitdiff_plain;h=e23d468968e608de27328888240de27d7582ad52
     6.8 +
     6.9 +Squashfs: Add XZ compression configuration option
    6.10 +
    6.11 +Signed-off-by: Phillip Lougher <phillip@lougher.demon.co.uk>
    6.12 +---
    6.13 +
    6.14 +diff --git a/fs/squashfs/Kconfig b/fs/squashfs/Kconfig
    6.15 +index e5f63da..e96d99a 100644
    6.16 +--- a/fs/squashfs/Kconfig
    6.17 ++++ b/fs/squashfs/Kconfig
    6.18 +@@ -53,6 +53,22 @@ config SQUASHFS_LZO
    6.19 + 
    6.20 + 	  If unsure, say N.
    6.21 + 
    6.22 ++config SQUASHFS_XZ
    6.23 ++	bool "Include support for XZ compressed file systems"
    6.24 ++	depends on SQUASHFS
    6.25 ++	default n
    6.26 ++	select XZ_DEC
    6.27 ++	help
    6.28 ++	  Saying Y here includes support for reading Squashfs file systems
    6.29 ++	  compressed with XZ compresssion.  XZ gives better compression than
    6.30 ++	  the default zlib compression, at the expense of greater CPU and
    6.31 ++	  memory overhead.
    6.32 ++
    6.33 ++	  XZ is not the standard compression used in Squashfs and so most
    6.34 ++	  file systems will be readable without selecting this option.
    6.35 ++
    6.36 ++	  If unsure, say N.
    6.37 ++
    6.38 + config SQUASHFS_EMBEDDED
    6.39 + 	bool "Additional option for memory-constrained systems"
    6.40 + 	depends on SQUASHFS
    6.41 +diff --git a/fs/squashfs/Makefile b/fs/squashfs/Makefile
    6.42 +index 7672bac..cecf2be 100644
    6.43 +--- a/fs/squashfs/Makefile
    6.44 ++++ b/fs/squashfs/Makefile
    6.45 +@@ -7,3 +7,4 @@ squashfs-y += block.o cache.o dir.o export.o file.o fragment.o id.o inode.o
    6.46 + squashfs-y += namei.o super.o symlink.o zlib_wrapper.o decompressor.o
    6.47 + squashfs-$(CONFIG_SQUASHFS_XATTR) += xattr.o xattr_id.o
    6.48 + squashfs-$(CONFIG_SQUASHFS_LZO) += lzo_wrapper.o
    6.49 ++squashfs-$(CONFIG_SQUASHFS_XZ) += xz_wrapper.o
    6.50 +diff --git a/fs/squashfs/decompressor.c b/fs/squashfs/decompressor.c
    6.51 +index 24af9ce..ac333b8 100644
    6.52 +--- a/fs/squashfs/decompressor.c
    6.53 ++++ b/fs/squashfs/decompressor.c
    6.54 +@@ -46,6 +46,12 @@ static const struct squashfs_decompressor squashfs_lzo_unsupported_comp_ops = {
    6.55 + };
    6.56 + #endif
    6.57 + 
    6.58 ++#ifndef CONFIG_SQUASHFS_XZ
    6.59 ++static const struct squashfs_decompressor squashfs_xz_unsupported_comp_ops = {
    6.60 ++	NULL, NULL, NULL, XZ_COMPRESSION, "xz", 0
    6.61 ++};
    6.62 ++#endif
    6.63 ++
    6.64 + static const struct squashfs_decompressor squashfs_unknown_comp_ops = {
    6.65 + 	NULL, NULL, NULL, 0, "unknown", 0
    6.66 + };
    6.67 +@@ -58,6 +64,11 @@ static const struct squashfs_decompressor *decompressor[] = {
    6.68 + #else
    6.69 + 	&squashfs_lzo_unsupported_comp_ops,
    6.70 + #endif
    6.71 ++#ifdef CONFIG_SQUASHFS_XZ
    6.72 ++	&squashfs_xz_comp_ops,
    6.73 ++#else
    6.74 ++	&squashfs_xz_unsupported_comp_ops,
    6.75 ++#endif
    6.76 + 	&squashfs_unknown_comp_ops
    6.77 + };
    6.78 + 
    6.79 +diff --git a/fs/squashfs/squashfs.h b/fs/squashfs/squashfs.h
    6.80 +index 5d45569..1096e2e 100644
    6.81 +--- a/fs/squashfs/squashfs.h
    6.82 ++++ b/fs/squashfs/squashfs.h
    6.83 +@@ -107,3 +107,6 @@ extern const struct squashfs_decompressor squashfs_zlib_comp_ops;
    6.84 + 
    6.85 + /* lzo_wrapper.c */
    6.86 + extern const struct squashfs_decompressor squashfs_lzo_comp_ops;
    6.87 ++
    6.88 ++/* xz_wrapper.c */
    6.89 ++extern const struct squashfs_decompressor squashfs_xz_comp_ops;
     7.1 --- a/linux/stuff/linux-2.6.36-slitaz.config	Tue Dec 14 21:34:46 2010 +0000
     7.2 +++ b/linux/stuff/linux-2.6.36-slitaz.config	Tue Dec 14 21:45:09 2010 +0000
     7.3 @@ -1,7 +1,7 @@
     7.4  #
     7.5  # Automatically generated make config: don't edit
     7.6  # Linux kernel version: 2.6.36
     7.7 -# Fri Nov  5 02:14:33 2010
     7.8 +# Tue Dec 14 11:22:55 2010
     7.9  #
    7.10  # CONFIG_64BIT is not set
    7.11  CONFIG_X86_32=y
    7.12 @@ -74,10 +74,12 @@
    7.13  CONFIG_HAVE_KERNEL_GZIP=y
    7.14  CONFIG_HAVE_KERNEL_BZIP2=y
    7.15  CONFIG_HAVE_KERNEL_LZMA=y
    7.16 +CONFIG_HAVE_KERNEL_XZ=y
    7.17  CONFIG_HAVE_KERNEL_LZO=y
    7.18  # CONFIG_KERNEL_GZIP is not set
    7.19  # CONFIG_KERNEL_BZIP2 is not set
    7.20  CONFIG_KERNEL_LZMA=y
    7.21 +# CONFIG_KERNEL_XZ is not set
    7.22  # CONFIG_KERNEL_LZO is not set
    7.23  CONFIG_SWAP=y
    7.24  CONFIG_SYSVIPC=y
    7.25 @@ -116,6 +118,7 @@
    7.26  CONFIG_RD_GZIP=y
    7.27  CONFIG_RD_BZIP2=y
    7.28  CONFIG_RD_LZMA=y
    7.29 +CONFIG_RD_XZ=y
    7.30  CONFIG_RD_LZO=y
    7.31  CONFIG_CC_OPTIMIZE_FOR_SIZE=y
    7.32  CONFIG_SYSCTL=y
    7.33 @@ -837,7 +840,6 @@
    7.34  CONFIG_ATM_BR2684=m
    7.35  # CONFIG_ATM_BR2684_IPFILTER is not set
    7.36  # CONFIG_L2TP is not set
    7.37 -# CONFIG_L2TP_V3 is not set
    7.38  CONFIG_STP=m
    7.39  CONFIG_BRIDGE=m
    7.40  # CONFIG_BRIDGE_IGMP_SNOOPING is not set
    7.41 @@ -1815,7 +1817,6 @@
    7.42  CONFIG_PPP_MPPE=y
    7.43  CONFIG_PPPOE=y
    7.44  CONFIG_PPPOATM=m
    7.45 -# CONFIG_PPPOL2TP is not set
    7.46  # CONFIG_SLIP is not set
    7.47  CONFIG_SLHC=y
    7.48  # CONFIG_NET_FC is not set
    7.49 @@ -3268,6 +3269,7 @@
    7.50  CONFIG_SQUASHFS=m
    7.51  CONFIG_SQUASHFS_XATTR=y
    7.52  CONFIG_SQUASHFS_LZO=y
    7.53 +CONFIG_SQUASHFS_XZ=y
    7.54  # CONFIG_SQUASHFS_EMBEDDED is not set
    7.55  CONFIG_SQUASHFS_FRAGMENT_CACHE_SIZE=3
    7.56  # CONFIG_VXFS_FS is not set
    7.57 @@ -3278,23 +3280,6 @@
    7.58  # CONFIG_ROMFS_FS is not set
    7.59  # CONFIG_SYSV_FS is not set
    7.60  # CONFIG_UFS_FS is not set
    7.61 -CONFIG_AUFS_FS=m
    7.62 -# CONFIG_AUFS_BRANCH_MAX_127 is not set
    7.63 -# CONFIG_AUFS_BRANCH_MAX_511 is not set
    7.64 -CONFIG_AUFS_BRANCH_MAX_1023=y
    7.65 -# CONFIG_AUFS_BRANCH_MAX_32767 is not set
    7.66 -CONFIG_AUFS_HNOTIFY=y
    7.67 -CONFIG_AUFS_HFSNOTIFY=y
    7.68 -# CONFIG_AUFS_HINOTIFY is not set
    7.69 -# CONFIG_AUFS_EXPORT is not set
    7.70 -# CONFIG_AUFS_RDU is not set
    7.71 -# CONFIG_AUFS_SP_IATTR is not set
    7.72 -CONFIG_AUFS_SHWH=y
    7.73 -CONFIG_AUFS_BR_RAMFS=y
    7.74 -# CONFIG_AUFS_BR_FUSE is not set
    7.75 -# CONFIG_AUFS_BR_HFSPLUS is not set
    7.76 -CONFIG_AUFS_BDEV_LOOP=y
    7.77 -# CONFIG_AUFS_DEBUG is not set
    7.78  CONFIG_NETWORK_FILESYSTEMS=y
    7.79  CONFIG_NFS_FS=y
    7.80  CONFIG_NFS_V3=y
    7.81 @@ -3615,9 +3600,19 @@
    7.82  CONFIG_ZLIB_INFLATE=y
    7.83  CONFIG_ZLIB_DEFLATE=y
    7.84  CONFIG_LZO_DECOMPRESS=y
    7.85 +CONFIG_XZ_DEC=y
    7.86 +CONFIG_XZ_DEC_X86=y
    7.87 +CONFIG_XZ_DEC_POWERPC=y
    7.88 +CONFIG_XZ_DEC_IA64=y
    7.89 +CONFIG_XZ_DEC_ARM=y
    7.90 +CONFIG_XZ_DEC_ARMTHUMB=y
    7.91 +CONFIG_XZ_DEC_SPARC=y
    7.92 +CONFIG_XZ_DEC_BCJ=y
    7.93 +# CONFIG_XZ_DEC_TEST is not set
    7.94  CONFIG_DECOMPRESS_GZIP=y
    7.95  CONFIG_DECOMPRESS_BZIP2=y
    7.96  CONFIG_DECOMPRESS_LZMA=y
    7.97 +CONFIG_DECOMPRESS_XZ=y
    7.98  CONFIG_DECOMPRESS_LZO=y
    7.99  CONFIG_TEXTSEARCH=y
   7.100  CONFIG_TEXTSEARCH_KMP=m
     8.1 --- a/linux/stuff/modules-2.6.36.list	Tue Dec 14 21:34:46 2010 +0000
     8.2 +++ b/linux/stuff/modules-2.6.36.list	Tue Dec 14 21:45:09 2010 +0000
     8.3 @@ -128,6 +128,9 @@
     8.4  net/llc/llc.ko.gz
     8.5  drivers/mmc/core/mmc_core.ko.gz
     8.6  net/rfkill/rfkill.ko.gz
     8.7 +net/sunrpc/auth_gss/auth_rpcgss.ko.gz
     8.8 +net/sunrpc/auth_gss/rpcsec_gss_krb5.ko.gz
     8.9 +net/netfilter/nf_conntrack.ko.gz
    8.10  drivers/input/gameport/gameport.ko.gz
    8.11  drivers/i2c/i2c-core.ko.gz
    8.12  sound/soundcore.ko.gz