diff options
| author | shadchin <shadchin@yandex-team.com> | 2023-01-05 21:33:07 +0300 |
|---|---|---|
| committer | shadchin <shadchin@yandex-team.com> | 2023-01-05 21:33:07 +0300 |
| commit | 97c1b4fc326f4a9435bc743e56681bb148b17c37 (patch) | |
| tree | c7d8dd514d722f747f5c6584343aeeddd5d2f8be /contrib/libs/lzma | |
| parent | d7b4c95518652104d51c915f0fd20fd3e954135c (diff) | |
| download | ydb-97c1b4fc326f4a9435bc743e56681bb148b17c37.tar.gz | |
Update contrib/libs/lzma to 5.4.0
Diffstat (limited to 'contrib/libs/lzma')
79 files changed, 9001 insertions, 532 deletions
diff --git a/contrib/libs/lzma/AUTHORS b/contrib/libs/lzma/AUTHORS index bda87975a5..69bbfc3ef6 100644 --- a/contrib/libs/lzma/AUTHORS +++ b/contrib/libs/lzma/AUTHORS @@ -3,15 +3,15 @@ Authors of XZ Utils =================== XZ Utils is developed and maintained by Lasse Collin - <lasse.collin@tukaani.org>. + <lasse.collin@tukaani.org> and Jia Tan <jiat0218@gmail.com>. Major parts of liblzma are based on code written by Igor Pavlov, - specifically the LZMA SDK <http://7-zip.org/sdk.html>. Without + specifically the LZMA SDK <https://7-zip.org/sdk.html>. Without this code, XZ Utils wouldn't exist. The SHA-256 implementation in liblzma is based on the code found from - 7-Zip <http://7-zip.org/>, which has a modified version of the SHA-256 - code found from Crypto++ <http://www.cryptopp.com/>. The SHA-256 code + 7-Zip <https://7-zip.org/>, which has a modified version of the SHA-256 + code found from Crypto++ <https://www.cryptopp.com/>. The SHA-256 code in Crypto++ was written by Kevin Springle and Wei Dai. Some scripts have been adapted from gzip. The original versions @@ -19,6 +19,18 @@ Authors of XZ Utils Andrew Dudman helped adapting the scripts and their man pages for XZ Utils. + The initial version of the threaded .xz decompressor was written + by Sebastian Andrzej Siewior. + + The initial version of the .lz (lzip) decoder was written + by Michał Górny. + + CLMUL-accelerated CRC code was contributed by Ilya Kurdyukov. + + Other authors: + - Jonathan Nieder + - Joachim Henke + The GNU Autotools-based build system contains files from many authors, which I'm not trying to list here. diff --git a/contrib/libs/lzma/CMakeLists.darwin.txt b/contrib/libs/lzma/CMakeLists.darwin.txt index 156016b68a..bb12c2a6ad 100644 --- a/contrib/libs/lzma/CMakeLists.darwin.txt +++ b/contrib/libs/lzma/CMakeLists.darwin.txt @@ -55,6 +55,7 @@ target_sources(contrib-libs-lzma PRIVATE ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/easy_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/easy_encoder_memusage.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/easy_preset.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/file_info.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/filter_buffer_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/filter_buffer_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/filter_common.c @@ -68,15 +69,20 @@ target_sources(contrib-libs-lzma PRIVATE ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/index_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/index_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/index_hash.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/lzip_decoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/microlzma_decoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/microlzma_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/outqueue.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_buffer_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_buffer_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_decoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_decoder_mt.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_encoder_mt.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_flags_common.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_flags_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_flags_encoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/string_conversion.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/vli_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/vli_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/vli_size.c @@ -96,6 +102,7 @@ target_sources(contrib-libs-lzma PRIVATE ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/lzma/lzma_encoder_presets.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/rangecoder/price_table.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/arm.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/arm64.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/armthumb.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/ia64.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/powerpc.c diff --git a/contrib/libs/lzma/CMakeLists.linux-aarch64.txt b/contrib/libs/lzma/CMakeLists.linux-aarch64.txt index c9f72e9ce0..8f0fef4c35 100644 --- a/contrib/libs/lzma/CMakeLists.linux-aarch64.txt +++ b/contrib/libs/lzma/CMakeLists.linux-aarch64.txt @@ -58,6 +58,7 @@ target_sources(contrib-libs-lzma PRIVATE ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/easy_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/easy_encoder_memusage.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/easy_preset.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/file_info.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/filter_buffer_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/filter_buffer_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/filter_common.c @@ -71,15 +72,20 @@ target_sources(contrib-libs-lzma PRIVATE ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/index_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/index_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/index_hash.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/lzip_decoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/microlzma_decoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/microlzma_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/outqueue.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_buffer_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_buffer_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_decoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_decoder_mt.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_encoder_mt.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_flags_common.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_flags_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_flags_encoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/string_conversion.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/vli_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/vli_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/vli_size.c @@ -99,6 +105,7 @@ target_sources(contrib-libs-lzma PRIVATE ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/lzma/lzma_encoder_presets.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/rangecoder/price_table.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/arm.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/arm64.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/armthumb.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/ia64.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/powerpc.c diff --git a/contrib/libs/lzma/CMakeLists.linux.txt b/contrib/libs/lzma/CMakeLists.linux.txt index c9f72e9ce0..8f0fef4c35 100644 --- a/contrib/libs/lzma/CMakeLists.linux.txt +++ b/contrib/libs/lzma/CMakeLists.linux.txt @@ -58,6 +58,7 @@ target_sources(contrib-libs-lzma PRIVATE ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/easy_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/easy_encoder_memusage.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/easy_preset.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/file_info.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/filter_buffer_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/filter_buffer_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/filter_common.c @@ -71,15 +72,20 @@ target_sources(contrib-libs-lzma PRIVATE ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/index_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/index_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/index_hash.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/lzip_decoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/microlzma_decoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/microlzma_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/outqueue.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_buffer_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_buffer_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_decoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_decoder_mt.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_encoder_mt.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_flags_common.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_flags_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/stream_flags_encoder.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/string_conversion.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/vli_decoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/vli_encoder.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/common/vli_size.c @@ -99,6 +105,7 @@ target_sources(contrib-libs-lzma PRIVATE ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/lzma/lzma_encoder_presets.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/rangecoder/price_table.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/arm.c + ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/arm64.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/armthumb.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/ia64.c ${CMAKE_SOURCE_DIR}/contrib/libs/lzma/liblzma/simple/powerpc.c diff --git a/contrib/libs/lzma/INSTALL b/contrib/libs/lzma/INSTALL index a55affb03d..67df18803f 100644 --- a/contrib/libs/lzma/INSTALL +++ b/contrib/libs/lzma/INSTALL @@ -142,12 +142,10 @@ XZ Utils Installation If it is enough to build liblzma (no command line tools): - - There is experimental CMake support. As it is, it should be - good enough to build static liblzma with Visual Studio. - Building liblzma.dll might work too (if it doesn't, it should - be fixed). The CMake support may work with MinGW or MinGW-w64. - Read the comment in the beginning of CMakeLists.txt before - running CMake! + - There is CMake support. It should be good enough to build + static liblzma or liblzma.dll with Visual Studio. The CMake + support may work with MinGW or MinGW-w64. Read the comment + in the beginning of CMakeLists.txt before running CMake! - There are Visual Studio project files under the "windows" directory. See windows/INSTALL-MSVC.txt. In the future the @@ -316,6 +314,18 @@ XZ Utils Installation | xz -v -0 -Csha256 > foo.xz time xz --test foo.xz + --disable-microlzma + Don't build MicroLZMA encoder and decoder. This omits + lzma_microlzma_encoder() and lzma_microlzma_decoder() + API functions from liblzma. These functions are needed + by specific applications only. They were written for + erofs-utils but they may be used by others too. + + --disable-lzip-decoder + Disable decompression support for .lz (lzip) files. + This omits the API function lzma_lzip_decoder() from + liblzma and .lz support from the xz tool. + --disable-xz --disable-xzdec --disable-lzmadec @@ -358,16 +368,45 @@ XZ Utils Installation pre-i686 systems, you may want to disable the assembler code. + --disable-clmul-crc + Disable the use carryless multiplication for CRC + calculation even if compiler support for it is detected. + The code uses runtime detection of SSSE3, SSE4.1, and + CLMUL instructions on x86. On 32-bit x86 this currently + is used only if --disable-assembler is used (this might + be fixed in the future). The code works on E2K too. + + If using compiler options that unconditionally allow the + required extensions (-msse4.1 -mpclmul) then runtime + detection isn't used and the generic code is omitted. + --enable-unaligned-access - Allow liblzma to use unaligned memory access for 16-bit - and 32-bit loads and stores. This should be enabled only - when the hardware supports this, i.e. when unaligned - access is fast. Some operating system kernels emulate - unaligned access, which is extremely slow. This option - shouldn't be used on systems that rely on such emulation. + Allow liblzma to use unaligned memory access for 16-bit, + 32-bit, and 64-bit loads and stores. This should be + enabled only when the hardware supports this, that is, + when unaligned access is fast. Some operating system + kernels emulate unaligned access, which is extremely + slow. This option shouldn't be used on systems that + rely on such emulation. Unaligned access is enabled by default on x86, x86-64, - and big endian PowerPC. + big endian PowerPC, some ARM, and some ARM64 systems. + + --enable-unsafe-type-punning + This enables use of code like + + uint8_t *buf8 = ...; + *(uint32_t *)buf8 = ...; + + which violates strict aliasing rules and may result + in broken code. There should be no need to use this + option with recent GCC or Clang versions on any + arch as just as fast code can be generated in a safe + way too (using __builtin_assume_aligned + memcpy). + + However, this option might improve performance in some + other cases, especially with old compilers (for example, + GCC 3 and early 4.x on x86, GCC < 6 on ARMv6 and ARMv7). --enable-small Reduce the size of liblzma by selecting smaller but @@ -427,7 +466,9 @@ XZ Utils Installation no Disable threading support. This is the same as using --disable-threads. - NOTE: If combined with --enable-small, the + NOTE: If combined with --enable-small + and the compiler doesn't support + __attribute__((__constructor__)), the resulting liblzma won't be thread safe, that is, if a multi-threaded application calls any liblzma functions from more than @@ -447,6 +488,7 @@ XZ Utils Installation auto Look for a supported sandboxing method and use it if found. If no method is found, then sandboxing isn't used. + This is the default. no Disable sandboxing support. @@ -455,6 +497,10 @@ XZ Utils Installation sandboxing. If no Capsicum support is found, configure will give an error. + pledge Use pledge(2) (OpenBSD >= 5.9) for + sandboxing. If pledge(2) isn't found, + configure will give an error. + --enable-symbol-versions Use symbol versioning for liblzma. This is enabled by default on GNU/Linux, other GNU-based systems, and @@ -517,7 +563,7 @@ XZ Utils Installation liblzma, pass --enable-small to configure. - Tell the compiler to optimize for size instead of speed. - E.g. with GCC, put -Os into CFLAGS. + For example, with GCC, put -Os into CFLAGS. - xzdec and lzmadec will never use multithreading capabilities of liblzma. You can avoid dependency on libpthread by passing diff --git a/contrib/libs/lzma/NEWS b/contrib/libs/lzma/NEWS index d3f7d60234..feae8a465e 100644 --- a/contrib/libs/lzma/NEWS +++ b/contrib/libs/lzma/NEWS @@ -2,6 +2,883 @@ XZ Utils Release Notes ====================== +5.4.0 (2022-12-13) + + This bumps the minor version of liblzma because new features were + added. The API and ABI are still backward compatible with liblzma + 5.2.x and 5.0.x. + + Since 5.3.5beta: + + * All fixes from 5.2.10. + + * The ARM64 filter is now stable. The xz option is now --arm64. + Decompression requires XZ Utils 5.4.0. In the future the ARM64 + filter will be supported by XZ for Java, XZ Embedded (including + the version in Linux), LZMA SDK, and 7-Zip. + + * Translations: + + - Updated Catalan, Croatian, German, Romanian, and Turkish + translations. + + - Updated German man page translations. + + - Added Romanian man page translations. + + Summary of new features added in the 5.3.x development releases: + + * liblzma: + + - Added threaded .xz decompressor lzma_stream_decoder_mt(). + It can use multiple threads with .xz files that have multiple + Blocks with size information in Block Headers. The threaded + encoder in xz has always created such files. + + Single-threaded encoder cannot store the size information in + Block Headers even if one used LZMA_FULL_FLUSH to create + multiple Blocks, so this threaded decoder cannot use multiple + threads with such files. + + If there are multiple Streams (concatenated .xz files), one + Stream will be decompressed completely before starting the + next Stream. + + - A new decoder flag LZMA_FAIL_FAST was added. It makes the + threaded decompressor report errors soon instead of first + flushing all pending data before the error location. + + - New Filter IDs: + * LZMA_FILTER_ARM64 is for ARM64 binaries. + * LZMA_FILTER_LZMA1EXT is for raw LZMA1 streams that don't + necessarily use the end marker. + + - Added lzma_str_to_filters(), lzma_str_from_filters(), and + lzma_str_list_filters() to convert a preset or a filter chain + string to a lzma_filter[] and vice versa. These should make + it easier to write applications that allow users to specify + custom compression options. + + - Added lzma_filters_free() which can be convenient for freeing + the filter options in a filter chain (an array of lzma_filter + structures). + + - lzma_file_info_decoder() to makes it a little easier to get + the Index field from .xz files. This helps in getting the + uncompressed file size but an easy-to-use random access + API is still missing which has existed in XZ for Java for + a long time. + + - Added lzma_microlzma_encoder() and lzma_microlzma_decoder(). + It is used by erofs-utils and may be used by others too. + + The MicroLZMA format is a raw LZMA stream (without end marker) + whose first byte (always 0x00) has been replaced with + bitwise-negation of the LZMA properties (lc/lp/pb). It was + created for use in EROFS but may be used in other contexts + as well where it is important to avoid wasting bytes for + stream headers or footers. The format is also supported by + XZ Embedded (the XZ Embedded version in Linux got MicroLZMA + support in Linux 5.16). + + The MicroLZMA encoder API in liblzma can compress into a + fixed-sized output buffer so that as much data is compressed + as can be fit into the buffer while still creating a valid + MicroLZMA stream. This is needed for EROFS. + + - Added lzma_lzip_decoder() to decompress the .lz (lzip) file + format version 0 and the original unextended version 1 files. + Also lzma_auto_decoder() supports .lz files. + + - lzma_filters_update() can now be used with the multi-threaded + encoder (lzma_stream_encoder_mt()) to change the filter chain + after LZMA_FULL_BARRIER or LZMA_FULL_FLUSH. + + - In lzma_options_lzma, allow nice_len = 2 and 3 with the match + finders that require at least 3 or 4. Now it is internally + rounded up if needed. + + - CLMUL-based CRC64 on x86-64 and E2K with runtime processor + detection. On 32-bit x86 it currently isn't available unless + --disable-assembler is used which can make the non-CLMUL + CRC64 slower; this might be fixed in the future. + + - Building with --disable-threads --enable-small + is now thread-safe if the compiler supports + __attribute__((__constructor__)). + + * xz: + + - Using -T0 (--threads=0) will now use multi-threaded encoder + even on a single-core system. This is to ensure that output + from the same xz binary is identical on both single-core and + multi-core systems. + + - --threads=+1 or -T+1 is now a way to put xz into + multi-threaded mode while using only one worker thread. + The + is ignored if the number is not 1. + + - A default soft memory usage limit is now used for compression + when -T0 is used and no explicit limit has been specified. + This soft limit is used to restrict the number of threads + but if the limit is exceeded with even one thread then xz + will continue with one thread using the multi-threaded + encoder and this limit is ignored. If the number of threads + is specified manually then no default limit will be used; + this affects only -T0. + + This change helps on systems that have very many cores and + using all of them for xz makes no sense. Previously xz -T0 + could run out of memory on such systems because it attempted + to reserve memory for too many threads. + + This also helps with 32-bit builds which don't have a large + amount of address space that would be required for many + threads. The default soft limit for -T0 is at most 1400 MiB + on all 32-bit platforms. + + - Previously a low value in --memlimit-compress wouldn't cause + xz to switch from multi-threaded mode to single-threaded mode + if the limit cannot otherwise be met; xz failed instead. Now + xz can switch to single-threaded mode and then, if needed, + scale down the LZMA2 dictionary size too just like it already + did when it was started in single-threaded mode. + + - The option --no-adjust no longer prevents xz from scaling down + the number of threads as that doesn't affect the compressed + output (only performance). Now --no-adjust only prevents + adjustments that affect compressed output, that is, with + --no-adjust xz won't switch from multi-threaded mode to + single-threaded mode and won't scale down the LZMA2 + dictionary size. + + - Added a new option --memlimit-mt-decompress=LIMIT. This is + used to limit the number of decompressor threads (possibly + falling back to single-threaded mode) but it will never make + xz refuse to decompress a file. This has a system-specific + default value because without any limit xz could end up + allocating memory for the whole compressed input file, the + whole uncompressed output file, multiple thread-specific + decompressor instances and so on. Basically xz could + attempt to use an insane amount of memory even with fairly + common files. The system-specific default value is currently + the same as the one used for compression with -T0. + + The new option works together with the existing option + --memlimit-decompress=LIMIT. The old option sets a hard limit + that must not be exceeded (xz will refuse to decompress) + while the new option only restricts the number of threads. + If the limit set with --memlimit-mt-decompress is greater + than the limit set with --memlimit-compress, then the latter + value is used also for --memlimit-mt-decompress. + + - Added new information to the output of xz --info-memory and + new fields to the output of xz --robot --info-memory. + + - In --lzma2=nice=NUMBER allow 2 and 3 with all match finders + now that liblzma handles it. + + - Don't mention endianness for ARM and ARM-Thumb filters in + --long-help. The filters only work for little endian + instruction encoding but modern ARM processors using + big endian data access still use little endian + instruction encoding. So the help text was misleading. + In contrast, the PowerPC filter is only for big endian + 32/64-bit PowerPC code. Little endian PowerPC would need + a separate filter. + + - Added decompression support for the .lz (lzip) file format + version 0 and the original unextended version 1. It is + autodetected by default. See also the option --format on + the xz man page. + + - Sandboxing enabled by default: + * Capsicum (FreeBSD) + * pledge(2) (OpenBSD) + + * Scripts now support the .lz format using xz. + + * A few new tests were added. + + * The liblzma-specific tests are now supported in CMake-based + builds too ("make test"). + + +5.3.5beta (2022-12-01) + + * All fixes from 5.2.9. + + * liblzma: + + - Added new LZMA_FILTER_LZMA1EXT for raw encoder and decoder to + handle raw LZMA1 streams that don't have end of payload marker + (EOPM) alias end of stream (EOS) marker. It can be used in + filter chains, for example, with the x86 BCJ filter. + + - Added lzma_str_to_filters(), lzma_str_from_filters(), and + lzma_str_list_filters() to make it easier for applications + to get custom compression options from a user and convert + it to an array of lzma_filter structures. + + - Added lzma_filters_free(). + + - lzma_filters_update() can now be used with the multi-threaded + encoder (lzma_stream_encoder_mt()) to change the filter chain + after LZMA_FULL_BARRIER or LZMA_FULL_FLUSH. + + - In lzma_options_lzma, allow nice_len = 2 and 3 with the match + finders that require at least 3 or 4. Now it is internally + rounded up if needed. + + - ARM64 filter was modified. It is still experimental. + + - Fixed LTO build with Clang if -fgnuc-version=10 or similar + was used to make Clang look like GCC >= 10. Now it uses + __has_attribute(__symver__) which should be reliable. + + * xz: + + - --threads=+1 or -T+1 is now a way to put xz into multi-threaded + mode while using only one worker thread. + + - In --lzma2=nice=NUMBER allow 2 and 3 with all match finders + now that liblzma handles it. + + * Updated translations: Chinese (simplified), Korean, and Turkish. + + +5.3.4alpha (2022-11-15) + + * All fixes from 5.2.7 and 5.2.8. + + * liblzma: + + - Minor improvements to the threaded decoder. + + - Added CRC64 implementation that uses SSSE3, SSE4.1, and CLMUL + instructions on 32/64-bit x86 and E2K. On 32-bit x86 it's + not enabled unless --disable-assembler is used but then + the non-CLMUL code might be slower. Processor support is + detected at runtime so this is built by default on x86-64 + and E2K. On these platforms, if compiler flags indicate + unconditional CLMUL support (-msse4.1 -mpclmul) then the + generic version is not built, making liblzma 8-9 KiB smaller + compared to having both versions included. + + With extremely compressible files this can make decompression + up to twice as fast but with typical files 5 % improvement + is a more realistic expectation. + + The CLMUL version is slower than the generic version with + tiny inputs (especially at 1-8 bytes per call, but up to + 16 bytes). In normal use in xz this doesn't matter at all. + + - Added an experimental ARM64 filter. This is *not* the final + version! Files created with this experimental version won't + be supported in the future versions! The filter design is + a compromise where improving one use case makes some other + cases worse. + + - Added decompression support for the .lz (lzip) file format + version 0 and the original unextended version 1. See the + API docs of lzma_lzip_decoder() for details. Also + lzma_auto_decoder() supports .lz files. + + - Building with --disable-threads --enable-small + is now thread-safe if the compiler supports + __attribute__((__constructor__)) + + * xz: + + - Added support for OpenBSD's pledge(2) as a sandboxing method. + + - Don't mention endianness for ARM and ARM-Thumb filters in + --long-help. The filters only work for little endian + instruction encoding but modern ARM processors using + big endian data access still use little endian + instruction encoding. So the help text was misleading. + In contrast, the PowerPC filter is only for big endian + 32/64-bit PowerPC code. Little endian PowerPC would need + a separate filter. + + - Added --experimental-arm64. This will be renamed once the + filter is finished. Files created with this experimental + filter will not be supported in the future! + + - Added new fields to the output of xz --robot --info-memory. + + - Added decompression support for the .lz (lzip) file format + version 0 and the original unextended version 1. It is + autodetected by default. See also the option --format on + the xz man page. + + * Scripts now support the .lz format using xz. + + * Build systems: + + - New #defines in config.h: HAVE_ENCODER_ARM64, + HAVE_DECODER_ARM64, HAVE_LZIP_DECODER, HAVE_CPUID_H, + HAVE_FUNC_ATTRIBUTE_CONSTRUCTOR, HAVE_USABLE_CLMUL + + - New configure options: --disable-clmul-crc, + --disable-microlzma, --disable-lzip-decoder, and + 'pledge' is now an option in --enable-sandbox (but + it's autodetected by default anyway). + + - INSTALL was updated to document the new configure options. + + - PACKAGERS now lists also --disable-microlzma and + --disable-lzip-decoder as configure options that must + not be used in builds for non-embedded use. + + * Tests: + + - Fix some of the tests so that they skip instead of fail if + certain features have been disabled with configure options. + It's still not perfect. + + - Other improvements to tests. + + * Updated translations: Croatian, Finnish, Hungarian, Polish, + Romanian, Spanish, Swedish, and Ukrainian. + + +5.3.3alpha (2022-08-22) + + * All fixes from 5.2.6. + + * liblzma: + + - Fixed 32-bit build. + + - Added threaded .xz decompressor lzma_stream_decoder_mt(). + It can use multiple threads with .xz files that have multiple + Blocks with size information in Block Headers. The threaded + encoder in xz has always created such files. + + Single-threaded encoder cannot store the size information in + Block Headers even if one used LZMA_FULL_FLUSH to create + multiple Blocks, so this threaded decoder cannot use multiple + threads with such files. + + If there are multiple Streams (concatenated .xz files), one + Stream will be decompressed completely before starting the + next Stream. + + - A new decoder flag LZMA_FAIL_FAST was added. It makes the + threaded decompressor report errors soon instead of first + flushing all pending data before the error location. + + * xz: + + - Using -T0 (--threads=0) will now use multi-threaded encoder + even on a single-core system. This is to ensure that output + from the same xz binary is identical on both single-core and + multi-core systems. + + - A default soft memory usage limit is now used for compression + when -T0 is used and no explicit limit has been specified. + This soft limit is used to restrict the number of threads + but if the limit is exceeded with even one thread then xz + will continue with one thread using the multi-threaded + encoder and this limit is ignored. If the number of threads + is specified manually then no default limit will be used; + this affects only -T0. + + This change helps on systems that have very many cores and + using all of them for xz makes no sense. Previously xz -T0 + could run out of memory on such systems because it attempted + to reserve memory for too many threads. + + This also helps with 32-bit builds which don't have a large + amount of address space that would be required for many + threads. The default limit is 1400 MiB on all 32-bit + platforms with -T0. + + Now xz -T0 should just work. It might use too few threads + in some cases but at least it shouldn't easily run out of + memory. It's possible that this will be tweaked before 5.4.0. + + - Changes to --memlimit-compress and --no-adjust: + + In single-threaded mode, --memlimit-compress can make xz + scale down the LZMA2 dictionary size to meet the memory usage + limit. This obviously affects the compressed output. However, + if xz was in threaded mode, --memlimit-compress could make xz + reduce the number of threads but it wouldn't make xz switch + from multi-threaded mode to single-threaded mode or scale + down the LZMA2 dictionary size. This seemed illogical. + + Now --memlimit-compress can make xz switch to single-threaded + mode if one thread in multi-threaded mode uses too much + memory. If memory usage is still too high, then the LZMA2 + dictionary size can be scaled down too. + + The option --no-adjust was also changed so that it no longer + prevents xz from scaling down the number of threads as that + doesn't affect compressed output (only performance). After + this commit --no-adjust only prevents adjustments that affect + compressed output, that is, with --no-adjust xz won't switch + from multithreaded mode to single-threaded mode and won't + scale down the LZMA2 dictionary size. + + - Added a new option --memlimit-mt-decompress=LIMIT. This is + used to limit the number of decompressor threads (possibly + falling back to single-threaded mode) but it will never make + xz refuse to decompress a file. This has a system-specific + default value because without any limit xz could end up + allocating memory for the whole compressed input file, the + whole uncompressed output file, multiple thread-specific + decompressor instances and so on. Basically xz could + attempt to use an insane amount of memory even with fairly + common files. + + The new option works together with the existing option + --memlimit-decompress=LIMIT. The old option sets a hard limit + that must not be exceeded (xz will refuse to decompress) + while the new option only restricts the number of threads. + If the limit set with --memlimit-mt-decompress is greater + than the limit set with --memlimit-compress, then the latter + value is used also for --memlimit-mt-decompress. + + * Tests: + + - Added a few more tests. + + - Added tests/code_coverage.sh to create a code coverage report + of the tests. + + * Build systems: + + - Automake's parallel test harness is now used to make tests + finish faster. + + - Added the CMake files to the distribution tarball. These were + supposed to be in 5.2.5 already. + + - Added liblzma tests to the CMake build. + + - Windows: Fix building of liblzma.dll with the included + Visual Studio project files. + + +5.3.2alpha (2021-10-28) + + This release was made on short notice so that recent erofs-utils can + be built with LZMA support without needing a snapshot from xz.git. + Thus many pending things were not included, not even updated + translations (which would need to be updated for the new --list + strings anyway). + + * All fixes from 5.2.5. + + * xz: + + - When copying metadata from the source file to the destination + file, don't try to set the group (GID) if it is already set + correctly. This avoids a failure on OpenBSD (and possibly on + a few other OSes) where files may get created so that their + group doesn't belong to the user, and fchown(2) can fail even + if it needs to do nothing. + + - The --keep option now accepts symlinks, hardlinks, and + setuid, setgid, and sticky files. Previously this required + using --force. + + - Split the long strings used in --list and --info-memory modes + to make them much easier for translators. + + - If built with sandbox support and enabling the sandbox fails, + xz will now immediately exit with exit status of 1. Previously + it would only display a warning if -vv was used. + + - Cap --memlimit-compress to 2000 MiB on MIPS32 because on + MIPS32 userspace processes are limited to 2 GiB of address + space. + + * liblzma: + + - Added lzma_microlzma_encoder() and lzma_microlzma_decoder(). + The API is in lzma/container.h. + + The MicroLZMA format is a raw LZMA stream (without end marker) + whose first byte (always 0x00) has been replaced with + bitwise-negation of the LZMA properties (lc/lp/pb). It was + created for use in EROFS but may be used in other contexts + as well where it is important to avoid wasting bytes for + stream headers or footers. The format is also supported by + XZ Embedded. + + The MicroLZMA encoder API in liblzma can compress into a + fixed-sized output buffer so that as much data is compressed + as can be fit into the buffer while still creating a valid + MicroLZMA stream. This is needed for EROFS. + + - Added fuzzing support. + + - Support Intel Control-flow Enforcement Technology (CET) in + 32-bit x86 assembly files. + + - Visual Studio: Use non-standard _MSVC_LANG to detect C++ + standard version in the lzma.h API header. It's used to + detect when "noexcept" can be used. + + * Scripts: + + - Fix exit status of xzdiff/xzcmp. Exit status could be 2 when + the correct value is 1. + + - Fix exit status of xzgrep. + + - Detect corrupt .bz2 files in xzgrep. + + - Add zstd support to xzgrep and xzdiff/xzcmp. + + - Fix less(1) version detection in xzless. It failed if the + version number from "less -V" contained a dot. + + * Fix typos and technical issues in man pages. + + * Build systems: + + - Windows: Fix building of resource files when config.h isn't + used. CMake + Visual Studio can now build liblzma.dll. + + - Various fixes to the CMake support. It might still need a few + more fixes even for liblzma-only builds. + + +5.3.1alpha (2018-04-29) + + * All fixes from 5.2.4. + + * Add lzma_file_info_decoder() into liblzma and use it in xz to + implement the --list feature. + + * Capsicum sandbox support is enabled by default where available + (FreeBSD >= 10). + + +5.2.10 (2022-12-13) + + * xz: Don't modify argv[] when parsing the --memlimit* and + --block-list command line options. This fixes confusing + arguments in process listing (like "ps auxf"). + + * GNU/Linux only: Use __has_attribute(__symver__) to detect if + that attribute is supported. This fixes build on Mandriva where + Clang is patched to define __GNUC__ to 11 by default (instead + of 4 as used by Clang upstream). + + +5.2.9 (2022-11-30) + + * liblzma: + + - Fixed an infinite loop in LZMA encoder initialization + if dict_size >= 2 GiB. (The encoder only supports up + to 1536 MiB.) + + - Fixed two cases of invalid free() that can happen if + a tiny allocation fails in encoder re-initialization + or in lzma_filters_update(). These bugs had some + similarities with the bug fixed in 5.2.7. + + - Fixed lzma_block_encoder() not allowing the use of + LZMA_SYNC_FLUSH with lzma_code() even though it was + documented to be supported. The sync-flush code in + the Block encoder was already used internally via + lzma_stream_encoder(), so this was just a missing flag + in the lzma_block_encoder() API function. + + - GNU/Linux only: Don't put symbol versions into static + liblzma as it breaks things in some cases (and even if + it didn't break anything, symbol versions in static + libraries are useless anyway). The downside of the fix + is that if the configure options --with-pic or --without-pic + are used then it's not possible to build both shared and + static liblzma at the same time on GNU/Linux anymore; + with those options --disable-static or --disable-shared + must be used too. + + * New email address for bug reports is <xz@tukaani.org> which + forwards messages to Lasse Collin and Jia Tan. + + +5.2.8 (2022-11-13) + + * xz: + + - If xz cannot remove an input file when it should, this + is now treated as a warning (exit status 2) instead of + an error (exit status 1). This matches GNU gzip and it + is more logical as at that point the output file has + already been successfully closed. + + - Fix handling of .xz files with an unsupported check type. + Previously such printed a warning message but then xz + behaved as if an error had occurred (didn't decompress, + exit status 1). Now a warning is printed, decompression + is done anyway, and exit status is 2. This used to work + slightly before 5.0.0. In practice this bug matters only + if xz has been built with some check types disabled. As + instructed in PACKAGERS, such builds should be done in + special situations only. + + - Fix "xz -dc --single-stream tests/files/good-0-empty.xz" + which failed with "Internal error (bug)". That is, + --single-stream was broken if the first .xz stream in + the input file didn't contain any uncompressed data. + + - Fix displaying file sizes in the progress indicator when + working in passthru mode and there are multiple input files. + Just like "gzip -cdf", "xz -cdf" works like "cat" when the + input file isn't a supported compressed file format. In + this case the file size counters weren't reset between + files so with multiple input files the progress indicator + displayed an incorrect (too large) value. + + * liblzma: + + - API docs in lzma/container.h: + * Update the list of decoder flags in the decoder + function docs. + * Explain LZMA_CONCATENATED behavior with .lzma files + in lzma_auto_decoder() docs. + + - OpenBSD: Use HW_NCPUONLINE to detect the number of + available hardware threads in lzma_physmem(). + + - Fix use of wrong macro to detect x86 SSE2 support. + __SSE2_MATH__ was used with GCC/Clang but the correct + one is __SSE2__. The first one means that SSE2 is used + for floating point math which is irrelevant here. + The affected SSE2 code isn't used on x86-64 so this affects + only 32-bit x86 builds that use -msse2 without -mfpmath=sse + (there is no runtime detection for SSE2). It improves LZMA + compression speed (not decompression). + + - Fix the build with Intel C compiler 2021 (ICC, not ICX) + on Linux. It defines __GNUC__ to 10 but doesn't support + the __symver__ attribute introduced in GCC 10. + + * Scripts: Ignore warnings from xz by using --quiet --no-warn. + This is needed if the input .xz files use an unsupported + check type. + + * Translations: + + - Updated Croatian and Turkish translations. + + - One new translations wasn't included because it needed + technical fixes. It will be in upcoming 5.4.0. No new + translations will be added to the 5.2.x branch anymore. + + - Renamed the French man page translation file from + fr_FR.po to fr.po and thus also its install directory + (like /usr/share/man/fr_FR -> .../fr). + + - Man page translations for upcoming 5.4.0 are now handled + in the Translation Project. + + * Update doc/faq.txt a little so it's less out-of-date. + + +5.2.7 (2022-09-30) + + * liblzma: + + - Made lzma_filters_copy() to never modify the destination + array if an error occurs. lzma_stream_encoder() and + lzma_stream_encoder_mt() already assumed this. Before this + change, if a tiny memory allocation in lzma_filters_copy() + failed it would lead to a crash (invalid free() or invalid + memory reads) in the cleanup paths of these two encoder + initialization functions. + + - Added missing integer overflow check to lzma_index_append(). + This affects xz --list and other applications that decode + the Index field from .xz files using lzma_index_decoder(). + Normal decompression of .xz files doesn't call this code + and thus most applications using liblzma aren't affected + by this bug. + + - Single-threaded .xz decoder (lzma_stream_decoder()): If + lzma_code() returns LZMA_MEMLIMIT_ERROR it is now possible + to use lzma_memlimit_set() to increase the limit and continue + decoding. This was supposed to work from the beginning + but there was a bug. With other decoders (.lzma or + threaded .xz decoder) this already worked correctly. + + - Fixed accumulation of integrity check type statistics in + lzma_index_cat(). This bug made lzma_index_checks() return + only the type of the integrity check of the last Stream + when multiple lzma_indexes were concatenated. Most + applications don't use these APIs but in xz it made + xz --list not list all check types from concatenated .xz + files. In xz --list --verbose only the per-file "Check:" + lines were affected and in xz --robot --list only the "file" + line was affected. + + - Added ABI compatibility with executables that were linked + against liblzma in RHEL/CentOS 7 or other liblzma builds + that had copied the problematic patch from RHEL/CentOS 7 + (xz-5.2.2-compat-libs.patch). For the details, see the + comment at the top of src/liblzma/validate_map.sh. + + WARNING: This uses __symver__ attribute with GCC >= 10. + In other cases the traditional __asm__(".symver ...") + is used. Using link-time optimization (LTO, -flto) with + GCC versions older than 10 can silently result in + broken liblzma.so.5 (incorrect symbol versions)! If you + want to use -flto with GCC, you must use GCC >= 10. + LTO with Clang seems to work even with the traditional + __asm__(".symver ...") method. + + * xzgrep: Fixed compatibility with old shells that break if + comments inside command substitutions have apostrophes ('). + This problem was introduced in 5.2.6. + + * Build systems: + + - New #define in config.h: HAVE_SYMBOL_VERSIONS_LINUX + + - Windows: Fixed liblzma.dll build with Visual Studio project + files. It broke in 5.2.6 due to a change that was made to + improve CMake support. + + - Windows: Building liblzma with UNICODE defined should now + work. + + - CMake files are now actually included in the release tarball. + They should have been in 5.2.5 already. + + - Minor CMake fixes and improvements. + + * Added a new translation: Turkish + + +5.2.6 (2022-08-12) + + * xz: + + - The --keep option now accepts symlinks, hardlinks, and + setuid, setgid, and sticky files. Previously this required + using --force. + + - When copying metadata from the source file to the destination + file, don't try to set the group (GID) if it is already set + correctly. This avoids a failure on OpenBSD (and possibly on + a few other OSes) where files may get created so that their + group doesn't belong to the user, and fchown(2) can fail even + if it needs to do nothing. + + - Cap --memlimit-compress to 2000 MiB instead of 4020 MiB on + MIPS32 because on MIPS32 userspace processes are limited + to 2 GiB of address space. + + * liblzma: + + - Fixed a missing error-check in the threaded encoder. If a + small memory allocation fails, a .xz file with an invalid + Index field would be created. Decompressing such a file would + produce the correct output but result in an error at the end. + Thus this is a "mild" data corruption bug. Note that while + a failed memory allocation can trigger the bug, it cannot + cause invalid memory access. + + - The decoder for .lzma files now supports files that have + uncompressed size stored in the header and still use the + end of payload marker (end of stream marker) at the end + of the LZMA stream. Such files are rare but, according to + the documentation in LZMA SDK, they are valid. + doc/lzma-file-format.txt was updated too. + + - Improved 32-bit x86 assembly files: + * Support Intel Control-flow Enforcement Technology (CET) + * Use non-executable stack on FreeBSD. + + - Visual Studio: Use non-standard _MSVC_LANG to detect C++ + standard version in the lzma.h API header. It's used to + detect when "noexcept" can be used. + + * xzgrep: + + - Fixed arbitrary command injection via a malicious filename + (CVE-2022-1271, ZDI-CAN-16587). A standalone patch for + this was released to the public on 2022-04-07. A slight + robustness improvement has been made since then and, if + using GNU or *BSD grep, a new faster method is now used + that doesn't use the old sed-based construct at all. This + also fixes bad output with GNU grep >= 3.5 (2020-09-27) + when xzgrepping binary files. + + This vulnerability was discovered by: + cleemy desu wayo working with Trend Micro Zero Day Initiative + + - Fixed detection of corrupt .bz2 files. + + - Improved error handling to fix exit status in some situations + and to fix handling of signals: in some situations a signal + didn't make xzgrep exit when it clearly should have. It's + possible that the signal handling still isn't quite perfect + but hopefully it's good enough. + + - Documented exit statuses on the man page. + + - xzegrep and xzfgrep now use "grep -E" and "grep -F" instead + of the deprecated egrep and fgrep commands. + + - Fixed parsing of the options -E, -F, -G, -P, and -X. The + problem occurred when multiple options were specied in + a single argument, for example, + + echo foo | xzgrep -Fe foo + + treated foo as a filename because -Fe wasn't correctly + split into -F -e. + + - Added zstd support. + + * xzdiff/xzcmp: + + - Fixed wrong exit status. Exit status could be 2 when the + correct value is 1. + + - Documented on the man page that exit status of 2 is used + for decompression errors. + + - Added zstd support. + + * xzless: + + - Fix less(1) version detection. It failed if the version number + from "less -V" contained a dot. + + * Translations: + + - Added new translations: Catalan, Croatian, Esperanto, + Korean, Portuguese, Romanian, Serbian, Spanish, Swedish, + and Ukrainian + + - Updated the Brazilian Portuguese translation. + + - Added French man page translation. This and the existing + German translation aren't complete anymore because the + English man pages got a few updates and the translators + weren't reached so that they could update their work. + + * Build systems: + + - Windows: Fix building of resource files when config.h isn't + used. CMake + Visual Studio can now build liblzma.dll. + + - Various fixes to the CMake support. Building static or shared + liblzma should work fine in most cases. In contrast, building + the command line tools with CMake is still clearly incomplete + and experimental and should be used for testing only. + + 5.2.5 (2020-03-17) * liblzma: diff --git a/contrib/libs/lzma/README b/contrib/libs/lzma/README index 3f0c38dca6..b9081ed455 100644 --- a/contrib/libs/lzma/README +++ b/contrib/libs/lzma/README @@ -202,9 +202,77 @@ XZ Utils https://translationproject.org/html/translators.html - Several strings will change in a future version of xz so if you - wish to start a new translation, look at the code in the xz git - repostiory instead of a 5.2.x release. + Below are notes and testing instructions specific to xz + translations. + + Testing can be done by installing xz into a temporary directory: + + ./configure --disable-shared --prefix=/tmp/xz-test + # <Edit the .po file in the po directory.> + make -C po update-po + make install + bash debug/translation.bash | less + bash debug/translation.bash | less -S # For --list outputs + + Repeat the above as needed (no need to re-run configure though). + + Note especially the following: + + - The output of --help and --long-help must look nice on + an 80-column terminal. It's OK to add extra lines if needed. + + - In contrast, don't add extra lines to error messages and such. + They are often preceded with e.g. a filename on the same line, + so you have no way to predict where to put a \n. Let the terminal + do the wrapping even if it looks ugly. Adding new lines will be + even uglier in the generic case even if it looks nice in a few + limited examples. + + - Be careful with column alignment in tables and table-like output + (--list, --list --verbose --verbose, --info-memory, --help, and + --long-help): + + * All descriptions of options in --help should start in the + same column (but it doesn't need to be the same column as + in the English messages; just be consistent if you change it). + Check that both --help and --long-help look OK, since they + share several strings. + + * --list --verbose and --info-memory print lines that have + the format "Description: %s". If you need a longer + description, you can put extra space between the colon + and %s. Then you may need to add extra space to other + strings too so that the result as a whole looks good (all + values start at the same column). + + * The columns of the actual tables in --list --verbose --verbose + should be aligned properly. Abbreviate if necessary. It might + be good to keep at least 2 or 3 spaces between column headings + and avoid spaces in the headings so that the columns stand out + better, but this is a matter of opinion. Do what you think + looks best. + + - Be careful to put a period at the end of a sentence when the + original version has it, and don't put it when the original + doesn't have it. Similarly, be careful with \n characters + at the beginning and end of the strings. + + - Read the TRANSLATORS comments that have been extracted from the + source code and included in xz.pot. Some comments suggest + testing with a specific command which needs an .xz file. You + may use e.g. any tests/files/good-*.xz. However, these test + commands are included in translations.bash output, so reading + translations.bash output carefully can be enough. + + - If you find language problems in the original English strings, + feel free to suggest improvements. Ask if something is unclear. + + - The translated messages should be understandable (sometimes this + may be a problem with the original English messages too). Don't + make a direct word-by-word translation from English especially if + the result doesn't sound good in your language. + + Thanks for your help! 5. Other implementations of the .xz format @@ -226,11 +294,10 @@ XZ Utils ---------------------- If you have questions, bug reports, patches etc. related to XZ Utils, - contact Lasse Collin <lasse.collin@tukaani.org> (in Finnish or English). - I'm sometimes slow at replying. If you haven't got a reply within two - weeks, assume that your email has got lost and resend it or use IRC. + the project maintainers Lasse Collin and Jia Tan can be reached via + <xz@tukaani.org>. - You can find me also from #tukaani on Freenode; my nick is Larhzu. - The channel tends to be pretty quiet, so just ask your question and - someone may wake up. + You might find Lasse also from #tukaani on Libera Chat (IRC). + The nick is Larhzu. The channel tends to be pretty quiet, + so just ask your question and someone might wake up. diff --git a/contrib/libs/lzma/common/mythread.h b/contrib/libs/lzma/common/mythread.h index be22654240..4138218360 100644 --- a/contrib/libs/lzma/common/mythread.h +++ b/contrib/libs/lzma/common/mythread.h @@ -370,10 +370,11 @@ typedef struct { BOOL pending_; \ if (!InitOnceBeginInitialize(&once_, 0, &pending_, NULL)) \ abort(); \ - if (pending_) \ + if (pending_) { \ func(); \ - if (!InitOnceComplete(&once, 0, NULL)) \ - abort(); \ + if (!InitOnceComplete(&once, 0, NULL)) \ + abort(); \ + } \ } while (0) #endif diff --git a/contrib/libs/lzma/common/tuklib_common.h b/contrib/libs/lzma/common/tuklib_common.h index 31fbab58b0..b1f531ea4a 100644 --- a/contrib/libs/lzma/common/tuklib_common.h +++ b/contrib/libs/lzma/common/tuklib_common.h @@ -14,7 +14,7 @@ #define TUKLIB_COMMON_H // The config file may be replaced by a package-specific file. -// It should include at least stddef.h, inttypes.h, and limits.h. +// It should include at least stddef.h, stdbool.h, inttypes.h, and limits.h. #include "tuklib_config.h" // TUKLIB_SYMBOL_PREFIX is prefixed to all symbols exported by diff --git a/contrib/libs/lzma/common/tuklib_config.h b/contrib/libs/lzma/common/tuklib_config.h index 549cb24d77..9d470ba732 100644 --- a/contrib/libs/lzma/common/tuklib_config.h +++ b/contrib/libs/lzma/common/tuklib_config.h @@ -1,7 +1,10 @@ +// If config.h isn't available, assume that the headers required by +// tuklib_common.h are available. This is required by crc32_tablegen.c. #ifdef HAVE_CONFIG_H # include "sysdefs.h" #else # include <stddef.h> +# include <stdbool.h> # include <inttypes.h> # include <limits.h> #endif diff --git a/contrib/libs/lzma/common/tuklib_cpucores.c b/contrib/libs/lzma/common/tuklib_cpucores.c index cc968dd25e..bb3f2f752b 100644 --- a/contrib/libs/lzma/common/tuklib_cpucores.c +++ b/contrib/libs/lzma/common/tuklib_cpucores.c @@ -72,7 +72,16 @@ tuklib_cpucores(void) } #elif defined(TUKLIB_CPUCORES_SYSCTL) + // On OpenBSD HW_NCPUONLINE tells the number of processor cores that + // are online so it is preferred over HW_NCPU which also counts cores + // that aren't currently available. The number of cores online is + // often less than HW_NCPU because OpenBSD disables simultaneous + // multi-threading (SMT) by default. +# ifdef HW_NCPUONLINE + int name[2] = { CTL_HW, HW_NCPUONLINE }; +# else int name[2] = { CTL_HW, HW_NCPU }; +# endif int cpus; size_t cpus_size = sizeof(cpus); if (sysctl(name, 2, &cpus, &cpus_size, NULL, 0) != -1 diff --git a/contrib/libs/lzma/common/tuklib_integer.h b/contrib/libs/lzma/common/tuklib_integer.h index 6f44a7a0ac..b58ef68d09 100644 --- a/contrib/libs/lzma/common/tuklib_integer.h +++ b/contrib/libs/lzma/common/tuklib_integer.h @@ -17,8 +17,8 @@ /// - Byte swapping: bswapXX(num) /// - Byte order conversions to/from native (byteswaps if Y isn't /// the native endianness): convXXYe(num) -/// - Unaligned reads (16/32-bit only): readXXYe(ptr) -/// - Unaligned writes (16/32-bit only): writeXXYe(ptr, num) +/// - Unaligned reads: readXXYe(ptr) +/// - Unaligned writes: writeXXYe(ptr, num) /// - Aligned reads: aligned_readXXYe(ptr) /// - Aligned writes: aligned_writeXXYe(ptr, num) /// @@ -343,6 +343,46 @@ read32le(const uint8_t *buf) } +static inline uint64_t +read64be(const uint8_t *buf) +{ +#if defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS) + uint64_t num = read64ne(buf); + return conv64be(num); +#else + uint64_t num = (uint64_t)buf[0] << 56; + num |= (uint64_t)buf[1] << 48; + num |= (uint64_t)buf[2] << 40; + num |= (uint64_t)buf[3] << 32; + num |= (uint64_t)buf[4] << 24; + num |= (uint64_t)buf[5] << 16; + num |= (uint64_t)buf[6] << 8; + num |= (uint64_t)buf[7]; + return num; +#endif +} + + +static inline uint64_t +read64le(const uint8_t *buf) +{ +#if !defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS) + uint64_t num = read64ne(buf); + return conv64le(num); +#else + uint64_t num = (uint64_t)buf[0]; + num |= (uint64_t)buf[1] << 8; + num |= (uint64_t)buf[2] << 16; + num |= (uint64_t)buf[3] << 24; + num |= (uint64_t)buf[4] << 32; + num |= (uint64_t)buf[5] << 40; + num |= (uint64_t)buf[6] << 48; + num |= (uint64_t)buf[7] << 56; + return num; +#endif +} + + // NOTE: Possible byte swapping must be done in a macro to allow the compiler // to optimize byte swapping of constants when using glibc's or *BSD's // byte swapping macros. The actual write is done in an inline function @@ -350,11 +390,13 @@ read32le(const uint8_t *buf) #if defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS) # define write16be(buf, num) write16ne(buf, conv16be(num)) # define write32be(buf, num) write32ne(buf, conv32be(num)) +# define write64be(buf, num) write64ne(buf, conv64be(num)) #endif #if !defined(WORDS_BIGENDIAN) || defined(TUKLIB_FAST_UNALIGNED_ACCESS) # define write16le(buf, num) write16ne(buf, conv16le(num)) # define write32le(buf, num) write32ne(buf, conv32le(num)) +# define write64le(buf, num) write64ne(buf, conv64le(num)) #endif diff --git a/contrib/libs/lzma/common/tuklib_physmem.c b/contrib/libs/lzma/common/tuklib_physmem.c index c499791474..fb11478f24 100644 --- a/contrib/libs/lzma/common/tuklib_physmem.c +++ b/contrib/libs/lzma/common/tuklib_physmem.c @@ -84,7 +84,7 @@ tuklib_physmem(void) // supports reporting values greater than 4 GiB. To keep the // code working also on older Windows versions, use // GlobalMemoryStatusEx() conditionally. - HMODULE kernel32 = GetModuleHandle("kernel32.dll"); + HMODULE kernel32 = GetModuleHandle(TEXT("kernel32.dll")); if (kernel32 != NULL) { typedef BOOL (WINAPI *gmse_type)(LPMEMORYSTATUSEX); gmse_type gmse = (gmse_type)GetProcAddress( diff --git a/contrib/libs/lzma/liblzma/api/lzma.h b/contrib/libs/lzma/liblzma/api/lzma.h index 122dab80d3..8fbd9a874b 100644 --- a/contrib/libs/lzma/liblzma/api/lzma.h +++ b/contrib/libs/lzma/liblzma/api/lzma.h @@ -219,7 +219,8 @@ */ #ifndef lzma_nothrow # if defined(__cplusplus) -# if __cplusplus >= 201103L +# if __cplusplus >= 201103L || (defined(_MSVC_LANG) \ + && _MSVC_LANG >= 201103L) # define lzma_nothrow noexcept # else # define lzma_nothrow throw() diff --git a/contrib/libs/lzma/liblzma/api/lzma/base.h b/contrib/libs/lzma/liblzma/api/lzma/base.h index a6005accc9..ff2d86968e 100644 --- a/contrib/libs/lzma/liblzma/api/lzma/base.h +++ b/contrib/libs/lzma/liblzma/api/lzma/base.h @@ -145,6 +145,12 @@ typedef enum { * specified memory usage limit. To continue decoding, * the memory usage limit has to be increased with * lzma_memlimit_set(). + * + * liblzma 5.2.6 and earlier had a bug in single-threaded .xz + * decoder (lzma_stream_decoder()) which made it impossible + * to continue decoding after LZMA_MEMLIMIT_ERROR even if + * the limit was increased using lzma_memlimit_set(). + * Other decoders worked correctly. */ LZMA_FORMAT_ERROR = 7, @@ -234,6 +240,36 @@ typedef enum { * can be a sign of a bug in liblzma. See the documentation * how to report bugs. */ + + LZMA_SEEK_NEEDED = 12, + /**< + * \brief Request to change the input file position + * + * Some coders can do random access in the input file. The + * initialization functions of these coders take the file size + * as an argument. No other coders can return LZMA_SEEK_NEEDED. + * + * When this value is returned, the application must seek to + * the file position given in lzma_stream.seek_pos. This value + * is guaranteed to never exceed the file size that was + * specified at the coder initialization. + * + * After seeking the application should read new input and + * pass it normally via lzma_stream.next_in and .avail_in. + */ + + /* + * These eumerations may be used internally by liblzma + * but they will never be returned to applications. + */ + LZMA_RET_INTERNAL1 = 101, + LZMA_RET_INTERNAL2 = 102, + LZMA_RET_INTERNAL3 = 103, + LZMA_RET_INTERNAL4 = 104, + LZMA_RET_INTERNAL5 = 105, + LZMA_RET_INTERNAL6 = 106, + LZMA_RET_INTERNAL7 = 107, + LZMA_RET_INTERNAL8 = 108 } lzma_ret; @@ -447,7 +483,7 @@ typedef struct lzma_internal_s lzma_internal; * * The lzma_stream structure is used for * - passing pointers to input and output buffers to liblzma; - * - defining custom memory hander functions; and + * - defining custom memory handler functions; and * - holding a pointer to coder-specific internal data structures. * * Typical usage: @@ -514,7 +550,19 @@ typedef struct { void *reserved_ptr2; void *reserved_ptr3; void *reserved_ptr4; - uint64_t reserved_int1; + + /** + * \brief New seek input position for LZMA_SEEK_NEEDED + * + * When lzma_code() returns LZMA_SEEK_NEEDED, the new input position + * needed by liblzma will be available seek_pos. The value is + * guaranteed to not exceed the file size that was specified when + * this lzma_stream was initialized. + * + * In all other situations the value of this variable is undefined. + */ + uint64_t seek_pos; + uint64_t reserved_int2; size_t reserved_int3; size_t reserved_int4; @@ -649,6 +697,11 @@ extern LZMA_API(uint64_t) lzma_memlimit_get(const lzma_stream *strm) * return LZMA_OK. Later versions treat 0 as if 1 had been specified (so * lzma_memlimit_get() will return 1 even if you specify 0 here). * + * liblzma 5.2.6 and earlier had a bug in single-threaded .xz decoder + * (lzma_stream_decoder()) which made it impossible to continue decoding + * after LZMA_MEMLIMIT_ERROR even if the limit was increased using + * lzma_memlimit_set(). Other decoders worked correctly. + * * \return - LZMA_OK: New memory usage limit successfully set. * - LZMA_MEMLIMIT_ERROR: The new limit is too small. * The limit was not changed. diff --git a/contrib/libs/lzma/liblzma/api/lzma/bcj.h b/contrib/libs/lzma/liblzma/api/lzma/bcj.h index 8e37538ad4..a1738d93b7 100644 --- a/contrib/libs/lzma/liblzma/api/lzma/bcj.h +++ b/contrib/libs/lzma/liblzma/api/lzma/bcj.h @@ -49,9 +49,13 @@ * Filter for SPARC binaries. */ +#define LZMA_FILTER_ARM64 LZMA_VLI_C(0x0A) + /**< + * Filter for ARM64 binaries. + */ /** - * \brief Options for BCJ filters + * \brief Options for BCJ filters (except ARM64) * * The BCJ filters never change the size of the data. Specifying options * for them is optional: if pointer to options is NULL, default value is diff --git a/contrib/libs/lzma/liblzma/api/lzma/block.h b/contrib/libs/lzma/liblzma/api/lzma/block.h index 962f38779c..082e55833e 100644 --- a/contrib/libs/lzma/liblzma/api/lzma/block.h +++ b/contrib/libs/lzma/liblzma/api/lzma/block.h @@ -464,9 +464,6 @@ extern LZMA_API(lzma_ret) lzma_block_encoder( * LZMA_FINISH is not required. It is supported only for convenience. * * \return - LZMA_OK: All good, continue with lzma_code(). - * - LZMA_UNSUPPORTED_CHECK: Initialization was successful, but - * the given Check ID is not supported, thus Check will be - * ignored. * - LZMA_PROG_ERROR * - LZMA_MEM_ERROR */ diff --git a/contrib/libs/lzma/liblzma/api/lzma/container.h b/contrib/libs/lzma/liblzma/api/lzma/container.h index 9fbf4df061..6e432a1f3b 100644 --- a/contrib/libs/lzma/liblzma/api/lzma/container.h +++ b/contrib/libs/lzma/liblzma/api/lzma/container.h @@ -69,7 +69,12 @@ typedef struct { * * Set this to zero if no flags are wanted. * - * No flags are currently supported. + * Encoder: No flags are currently supported. + * + * Decoder: Bitwise-or of zero or more of the decoder flags: + * LZMA_TELL_NO_CHECK, LZMA_TELL_UNSUPPORTED_CHECK, + * LZMA_TELL_ANY_CHECK, LZMA_IGNORE_CHECK, + * LZMA_CONCATENATED, LZMA_FAIL_FAST */ uint32_t flags; @@ -79,7 +84,7 @@ typedef struct { uint32_t threads; /** - * \brief Maximum uncompressed size of a Block + * \brief Encoder only: Maximum uncompressed size of a Block * * The encoder will start a new .xz Block every block_size bytes. * Using LZMA_FULL_FLUSH or LZMA_FULL_BARRIER with lzma_code() @@ -135,7 +140,7 @@ typedef struct { uint32_t timeout; /** - * \brief Compression preset (level and possible flags) + * \brief Encoder only: Compression preset * * The preset is set just like with lzma_easy_encoder(). * The preset is ignored if filters below is non-NULL. @@ -143,7 +148,7 @@ typedef struct { uint32_t preset; /** - * \brief Filter chain (alternative to a preset) + * \brief Encoder only: Filter chain (alternative to a preset) * * If this is NULL, the preset above is used. Otherwise the preset * is ignored and the filter chain specified here is used. @@ -151,7 +156,7 @@ typedef struct { const lzma_filter *filters; /** - * \brief Integrity check type + * \brief Encoder only: Integrity check type * * See check.h for available checks. The xz command line tool * defaults to LZMA_CHECK_CRC64, which is a good choice if you @@ -173,8 +178,50 @@ typedef struct { uint32_t reserved_int2; uint32_t reserved_int3; uint32_t reserved_int4; - uint64_t reserved_int5; - uint64_t reserved_int6; + + /** + * \brief Memory usage limit to reduce the number of threads + * + * Encoder: Ignored. + * + * Decoder: + * + * If the number of threads has been set so high that more than + * memlimit_threading bytes of memory would be needed, the number + * of threads will be reduced so that the memory usage will not exceed + * memlimit_threading bytes. However, if memlimit_threading cannot + * be met even in single-threaded mode, then decoding will continue + * in single-threaded mode and memlimit_threading may be exceeded + * even by a large amount. That is, memlimit_threading will never make + * lzma_code() return LZMA_MEMLIMIT_ERROR. To truly cap the memory + * usage, see memlimit_stop below. + * + * Setting memlimit_threading to UINT64_MAX or a similar huge value + * means that liblzma is allowed to keep the whole compressed file + * and the whole uncompressed file in memory in addition to the memory + * needed by the decompressor data structures used by each thread! + * In other words, a reasonable value limit must be set here or it + * will cause problems sooner or later. If you have no idea what + * a reasonable value could be, try lzma_physmem() / 4 as a starting + * point. Setting this limit will never prevent decompression of + * a file; this will only reduce the number of threads. + * + * If memlimit_threading is greater than memlimit_stop, then the value + * of memlimit_stop will be used for both. + */ + uint64_t memlimit_threading; + + /** + * \brief Memory usage limit that should never be exceeded + * + * Encoder: Ignored. + * + * Decoder: If decompressing will need more than this amount of + * memory even in the single-threaded mode, then lzma_code() will + * return LZMA_MEMLIMIT_ERROR. + */ + uint64_t memlimit_stop; + uint64_t reserved_int7; uint64_t reserved_int8; void *reserved_ptr1; @@ -444,6 +491,60 @@ extern LZMA_API(lzma_ret) lzma_stream_buffer_encode( lzma_nothrow lzma_attr_warn_unused_result; +/** + * \brief MicroLZMA encoder + * + * The MicroLZMA format is a raw LZMA stream whose first byte (always 0x00) + * has been replaced with bitwise-negation of the LZMA properties (lc/lp/pb). + * This encoding ensures that the first byte of MicroLZMA stream is never + * 0x00. There is no end of payload marker and thus the uncompressed size + * must be stored separately. For the best error detection the dictionary + * size should be stored separately as well but alternatively one may use + * the uncompressed size as the dictionary size when decoding. + * + * With the MicroLZMA encoder, lzma_code() behaves slightly unusually. + * The action argument must be LZMA_FINISH and the return value will never be + * LZMA_OK. Thus the encoding is always done with a single lzma_code() after + * the initialization. The benefit of the combination of initialization + * function and lzma_code() is that memory allocations can be re-used for + * better performance. + * + * lzma_code() will try to encode as much input as is possible to fit into + * the given output buffer. If not all input can be encoded, the stream will + * be finished without encoding all the input. The caller must check both + * input and output buffer usage after lzma_code() (total_in and total_out + * in lzma_stream can be convenient). Often lzma_code() can fill the output + * buffer completely if there is a lot of input, but sometimes a few bytes + * may remain unused because the next LZMA symbol would require more space. + * + * lzma_stream.avail_out must be at least 6. Otherwise LZMA_PROG_ERROR + * will be returned. + * + * The LZMA dictionary should be reasonably low to speed up the encoder + * re-initialization. A good value is bigger than the resulting + * uncompressed size of most of the output chunks. For example, if output + * size is 4 KiB, dictionary size of 32 KiB or 64 KiB is good. If the + * data compresses extremely well, even 128 KiB may be useful. + * + * The MicroLZMA format and this encoder variant were made with the EROFS + * file system in mind. This format may be convenient in other embedded + * uses too where many small streams are needed. XZ Embedded includes a + * decoder for this format. + * + * \return - LZMA_STREAM_END: All good. Check the amounts of input used + * and output produced. Store the amount of input used + * (uncompressed size) as it needs to be known to decompress + * the data. + * - LZMA_OPTIONS_ERROR + * - LZMA_MEM_ERROR + * - LZMA_PROG_ERROR: In addition to the generic reasons for this + * error code, this may also be returned if there isn't enough + * output space (6 bytes) to create a valid MicroLZMA stream. + */ +extern LZMA_API(lzma_ret) lzma_microlzma_encoder( + lzma_stream *strm, const lzma_options_lzma *options); + + /************ * Decoding * ************/ @@ -501,8 +602,8 @@ extern LZMA_API(lzma_ret) lzma_stream_buffer_encode( /** * This flag enables decoding of concatenated files with file formats that * allow concatenating compressed files as is. From the formats currently - * supported by liblzma, only the .xz format allows concatenated files. - * Concatenated files are not allowed with the legacy .lzma format. + * supported by liblzma, only the .xz and .lz formats allow concatenated + * files. Concatenated files are not allowed with the legacy .lzma format. * * This flag also affects the usage of the `action' argument for lzma_code(). * When LZMA_CONCATENATED is used, lzma_code() won't return LZMA_STREAM_END @@ -516,6 +617,35 @@ extern LZMA_API(lzma_ret) lzma_stream_buffer_encode( /** + * This flag makes the threaded decoder report errors (like LZMA_DATA_ERROR) + * as soon as they are detected. This saves time when the application has no + * interest in a partially decompressed truncated or corrupt file. Note that + * due to timing randomness, if the same truncated or corrupt input is + * decompressed multiple times with this flag, a different amount of output + * may be produced by different runs, and even the error code might vary. + * + * When using LZMA_FAIL_FAST, it is recommended to use LZMA_FINISH to tell + * the decoder when no more input will be coming because it can help fast + * detection and reporting of truncated files. Note that in this situation + * truncated files might be diagnosed with LZMA_DATA_ERROR instead of + * LZMA_OK or LZMA_BUF_ERROR! + * + * Without this flag the threaded decoder will provide as much output as + * possible at first and then report the pending error. This default behavior + * matches the single-threaded decoder and provides repeatable behavior + * with truncated or corrupt input. There are a few special cases where the + * behavior can still differ like memory allocation failures (LZMA_MEM_ERROR). + * + * Single-threaded decoders currently ignore this flag. + * + * Support for this flag was added in liblzma 5.3.3alpha. Note that in older + * versions this flag isn't supported (LZMA_OPTIONS_ERROR) even by functions + * that ignore this flag in newer liblzma versions. + */ +#define LZMA_FAIL_FAST UINT32_C(0x20) + + +/** * \brief Initialize .xz Stream decoder * * \param strm Pointer to properly prepared lzma_stream @@ -526,7 +656,8 @@ extern LZMA_API(lzma_ret) lzma_stream_buffer_encode( * had been specified. * \param flags Bitwise-or of zero or more of the decoder flags: * LZMA_TELL_NO_CHECK, LZMA_TELL_UNSUPPORTED_CHECK, - * LZMA_TELL_ANY_CHECK, LZMA_CONCATENATED + * LZMA_TELL_ANY_CHECK, LZMA_IGNORE_CHECK, + * LZMA_CONCATENATED, LZMA_FAIL_FAST * * \return - LZMA_OK: Initialization was successful. * - LZMA_MEM_ERROR: Cannot allocate memory. @@ -539,11 +670,50 @@ extern LZMA_API(lzma_ret) lzma_stream_decoder( /** - * \brief Decode .xz Streams and .lzma files with autodetection + * \brief Initialize multithreaded .xz Stream decoder + * + * \param strm Pointer to properly prepared lzma_stream + * \param options Pointer to multithreaded compression options + * + * The decoder can decode multiple Blocks in parallel. This requires that each + * Block Header contains the Compressed Size and Uncompressed size fields + * which are added by the multi-threaded encoder, see lzma_stream_encoder_mt(). + * + * A Stream with one Block will only utilize one thread. A Stream with multiple + * Blocks but without size information in Block Headers will be processed in + * single-threaded mode in the same way as done by lzma_stream_decoder(). + * Concatenated Streams are processed one Stream at a time; no inter-Stream + * parallelization is done. + * + * This function behaves like lzma_stream_decoder() when options->threads == 1 + * and options->memlimit_threading <= 1. + * + * \return - LZMA_OK: Initialization was successful. + * - LZMA_MEM_ERROR: Cannot allocate memory. + * - LZMA_MEMLIMIT_ERROR: Memory usage limit was reached. + * - LZMA_OPTIONS_ERROR: Unsupported flags. + * - LZMA_PROG_ERROR + */ +extern LZMA_API(lzma_ret) lzma_stream_decoder_mt( + lzma_stream *strm, const lzma_mt *options) + lzma_nothrow lzma_attr_warn_unused_result; + + +/** + * \brief Decode .xz, .lzma, and .lz (lzip) files with autodetection + * + * This decoder autodetects between the .xz, .lzma, and .lz file formats, + * and calls lzma_stream_decoder(), lzma_alone_decoder(), or + * lzma_lzip_decoder() once the type of the input file has been detected. * - * This decoder autodetects between the .xz and .lzma file formats, and - * calls lzma_stream_decoder() or lzma_alone_decoder() once the type - * of the input file has been detected. + * Support for .lz was added in 5.4.0. + * + * If the flag LZMA_CONCATENATED is used and the input is a .lzma file: + * For historical reasons concatenated .lzma files aren't supported. + * If there is trailing data after one .lzma stream, lzma_code() will + * return LZMA_DATA_ERROR. (lzma_alone_decoder() doesn't have such a check + * as it doesn't support any decoder flags. It will return LZMA_STREAM_END + * after one .lzma stream.) * * \param strm Pointer to properly prepared lzma_stream * \param memlimit Memory usage limit as bytes. Use UINT64_MAX @@ -551,7 +721,10 @@ extern LZMA_API(lzma_ret) lzma_stream_decoder( * 5.2.3 and earlier don't allow 0 here and return * LZMA_PROG_ERROR; later versions treat 0 as if 1 * had been specified. - * \param flags Bitwise-or of flags, or zero for no flags. + * \param flags Bitwise-or of zero or more of the decoder flags: + * LZMA_TELL_NO_CHECK, LZMA_TELL_UNSUPPORTED_CHECK, + * LZMA_TELL_ANY_CHECK, LZMA_IGNORE_CHECK, + * LZMA_CONCATENATED, LZMA_FAIL_FAST * * \return - LZMA_OK: Initialization was successful. * - LZMA_MEM_ERROR: Cannot allocate memory. @@ -587,6 +760,64 @@ extern LZMA_API(lzma_ret) lzma_alone_decoder( /** + * \brief Initialize .lz (lzip) decoder (a foreign file format) + * + * \param strm Pointer to properly prepared lzma_stream + * \param memlimit Memory usage limit as bytes. Use UINT64_MAX + * to effectively disable the limiter. + * \param flags Bitwise-or of flags, or zero for no flags. + * All decoder flags listed above are supported + * although only LZMA_CONCATENATED and (in very rare + * cases) LZMA_IGNORE_CHECK are actually useful. + * LZMA_TELL_NO_CHECK, LZMA_TELL_UNSUPPORTED_CHECK, + * and LZMA_FAIL_FAST do nothing. LZMA_TELL_ANY_CHECK + * is supported for consistency only as CRC32 is + * always used in the .lz format. + * + * This decoder supports the .lz format version 0 and the unextended .lz + * format version 1: + * + * - Files in the format version 0 were produced by lzip 1.3 and older. + * Such files aren't common but may be found from file archives + * as a few source packages were released in this format. People + * might have old personal files in this format too. Decompression + * support for the format version 0 was removed in lzip 1.18. + * + * - lzip 1.3 added decompression support for .lz format version 1 files. + * Compression support was added in lzip 1.4. In lzip 1.6 the .lz format + * version 1 was extended to support the Sync Flush marker. This extension + * is not supported by liblzma. lzma_code() will return LZMA_DATA_ERROR + * at the location of the Sync Flush marker. In practice files with + * the Sync Flush marker are very rare and thus liblzma can decompress + * almost all .lz files. + * + * Just like with lzma_stream_decoder() for .xz files, LZMA_CONCATENATED + * should be used when decompressing normal standalone .lz files. + * + * The .lz format allows putting non-.lz data at the end of a file after at + * least one valid .lz member. That is, one can append custom data at the end + * of a .lz file and the decoder is required to ignore it. In liblzma this + * is relevant only when LZMA_CONCATENATED is used. In that case lzma_code() + * will return LZMA_STREAM_END and leave lzma_stream.next_in pointing to + * the first byte of the non-.lz data. An exception to this is if the first + * 1-3 bytes of the non-.lz data are identical to the .lz magic bytes + * (0x4C, 0x5A, 0x49, 0x50; "LZIP" in US-ASCII). In such a case the 1-3 bytes + * will have been ignored by lzma_code(). If one wishes to locate the non-.lz + * data reliably, one must ensure that the first byte isn't 0x4C. Actually + * one should ensure that none of the first four bytes of trailing data are + * equal to the magic bytes because lzip >= 1.20 requires it by default. + * + * \return - LZMA_OK: Initialization was successful. + * - LZMA_MEM_ERROR: Cannot allocate memory. + * - LZMA_OPTIONS_ERROR: Unsupported flags + * - LZMA_PROG_ERROR + */ +extern LZMA_API(lzma_ret) lzma_lzip_decoder( + lzma_stream *strm, uint64_t memlimit, uint32_t flags) + lzma_nothrow lzma_attr_warn_unused_result; + + +/** * \brief Single-call .xz Stream decoder * * \param memlimit Pointer to how much memory the decoder is allowed @@ -595,7 +826,8 @@ extern LZMA_API(lzma_ret) lzma_alone_decoder( * returned. * \param flags Bitwise-or of zero or more of the decoder flags: * LZMA_TELL_NO_CHECK, LZMA_TELL_UNSUPPORTED_CHECK, - * LZMA_CONCATENATED. Note that LZMA_TELL_ANY_CHECK + * LZMA_IGNORE_CHECK, LZMA_CONCATENATED, + * LZMA_FAIL_FAST. Note that LZMA_TELL_ANY_CHECK * is not allowed and will return LZMA_PROG_ERROR. * \param allocator lzma_allocator for custom allocator functions. * Set to NULL to use malloc() and free(). @@ -630,3 +862,43 @@ extern LZMA_API(lzma_ret) lzma_stream_buffer_decode( const uint8_t *in, size_t *in_pos, size_t in_size, uint8_t *out, size_t *out_pos, size_t out_size) lzma_nothrow lzma_attr_warn_unused_result; + + +/** + * \brief MicroLZMA decoder + * + * See lzma_microlzma_decoder() for more information. + * + * The lzma_code() usage with this decoder is completely normal. The + * special behavior of lzma_code() applies to lzma_microlzma_encoder() only. + * + * \param strm Pointer to properly prepared lzma_stream + * \param comp_size Compressed size of the MicroLZMA stream. + * The caller must somehow know this exactly. + * \param uncomp_size Uncompressed size of the MicroLZMA stream. + * If the exact uncompressed size isn't known, this + * can be set to a value that is at most as big as + * the exact uncompressed size would be, but then the + * next argument uncomp_size_is_exact must be false. + * \param uncomp_size_is_exact + * If true, uncomp_size must be exactly correct. + * This will improve error detection at the end of + * the stream. If the exact uncompressed size isn't + * known, this must be false. uncomp_size must still + * be at most as big as the exact uncompressed size + * is. Setting this to false when the exact size is + * known will work but error detection at the end of + * the stream will be weaker. + * \param dict_size LZMA dictionary size that was used when + * compressing the data. It is OK to use a bigger + * value too but liblzma will then allocate more + * memory than would actually be required and error + * detection will be slightly worse. (Note that with + * the implementation in XZ Embedded it doesn't + * affect the memory usage if one specifies bigger + * dictionary than actually required.) + */ +extern LZMA_API(lzma_ret) lzma_microlzma_decoder( + lzma_stream *strm, uint64_t comp_size, + uint64_t uncomp_size, lzma_bool uncomp_size_is_exact, + uint32_t dict_size); diff --git a/contrib/libs/lzma/liblzma/api/lzma/filter.h b/contrib/libs/lzma/liblzma/api/lzma/filter.h index 8c85931476..58c83bf7ed 100644 --- a/contrib/libs/lzma/liblzma/api/lzma/filter.h +++ b/contrib/libs/lzma/liblzma/api/lzma/filter.h @@ -108,7 +108,9 @@ extern LZMA_API(lzma_bool) lzma_filter_decoder_is_supported(lzma_vli id) * need to be initialized by the caller in any way. * * If an error occurs, memory possibly already allocated by this function - * is always freed. + * is always freed. liblzma versions older than 5.2.7 may modify the dest + * array and leave its contents in an undefined state if an error occurs. + * liblzma 5.2.7 and newer only modify the dest array when returning LZMA_OK. * * \return - LZMA_OK * - LZMA_MEM_ERROR @@ -118,7 +120,29 @@ extern LZMA_API(lzma_bool) lzma_filter_decoder_is_supported(lzma_vli id) */ extern LZMA_API(lzma_ret) lzma_filters_copy( const lzma_filter *src, lzma_filter *dest, - const lzma_allocator *allocator) lzma_nothrow; + const lzma_allocator *allocator) + lzma_nothrow lzma_attr_warn_unused_result; + + +/** + * \brief Free the options in the array of lzma_filter structures + * + * This frees the filter chain options. The filters array itself is not freed. + * + * The filters array must have at most LZMA_FILTERS_MAX + 1 elements + * including the terminating element which must have .id = LZMA_VLI_UNKNOWN. + * For all elements before the terminating element: + * - options will be freed using the given lzma_allocator or, + * if allocator is NULL, using free(). + * - options will be set to NULL. + * - id will be set to LZMA_VLI_UNKNOWN. + * + * If filters is NULL, this does nothing but remember that this never frees + * the filters array itself. + */ +extern LZMA_API(void) lzma_filters_free( + lzma_filter *filters, const lzma_allocator *allocator) + lzma_nothrow; /** @@ -202,21 +226,27 @@ extern LZMA_API(lzma_ret) lzma_raw_decoder( /** * \brief Update the filter chain in the encoder * - * This function is for advanced users only. This function has two slightly - * different purposes: + * This function may be called after lzma_code() has returned LZMA_STREAM_END + * when LZMA_FULL_BARRIER, LZMA_FULL_FLUSH, or LZMA_SYNC_FLUSH was used: * - * - After LZMA_FULL_FLUSH when using Stream encoder: Set a new filter - * chain, which will be used starting from the next Block. + * - After LZMA_FULL_BARRIER or LZMA_FULL_FLUSH: Single-threaded .xz Stream + * encoder (lzma_stream_encoder()) and (since liblzma 5.4.0) multi-threaded + * Stream encoder (lzma_stream_encoder_mt()) allow setting a new filter + * chain to be used for the next Block(s). * - * - After LZMA_SYNC_FLUSH using Raw, Block, or Stream encoder: Change - * the filter-specific options in the middle of encoding. The actual - * filters in the chain (Filter IDs) cannot be changed. In the future, - * it might become possible to change the filter options without - * using LZMA_SYNC_FLUSH. + * - After LZMA_SYNC_FLUSH: Raw encoder (lzma_raw_encoder()), + * Block encocder (lzma_block_encoder()), and single-threaded .xz Stream + * encoder (lzma_stream_encoder()) allow changing certain filter-specific + * options in the middle of encoding. The actual filters in the chain + * (Filter IDs) must not be changed! Currently only the lc, lp, and pb + * options of LZMA2 (not LZMA1) can be changed this way. * - * While rarely useful, this function may be called also when no data has - * been compressed yet. In that case, this function will behave as if - * LZMA_FULL_FLUSH (Stream encoder) or LZMA_SYNC_FLUSH (Raw or Block + * - In the future some filters might allow changing some of their options + * without any barrier or flushing but currently such filters don't exist. + * + * This function may also be called when no data has been compressed yet + * although this is rarely useful. In that case, this function will behave + * as if LZMA_FULL_FLUSH (Stream encoders) or LZMA_SYNC_FLUSH (Raw or Block * encoder) had been used right before calling this function. * * \return - LZMA_OK @@ -424,3 +454,261 @@ extern LZMA_API(lzma_ret) lzma_filter_flags_decode( lzma_filter *filter, const lzma_allocator *allocator, const uint8_t *in, size_t *in_pos, size_t in_size) lzma_nothrow lzma_attr_warn_unused_result; + + +/*********** + * Strings * + ***********/ + +/** + * \brief Allow or show all filters + * + * By default only the filters supported in the .xz format are accept by + * lzma_str_to_filters() or shown by lzma_str_list_filters(). + */ +#define LZMA_STR_ALL_FILTERS UINT32_C(0x01) + + +/** + * \brief Do not validate the filter chain in lzma_str_to_filters() + * + * By default lzma_str_to_filters() can return an error if the filter chain + * as a whole isn't usable in the .xz format or in the raw encoder or decoder. + * With this flag the validation is skipped (this doesn't affect the handling + * of the individual filter options). + */ +#define LZMA_STR_NO_VALIDATION UINT32_C(0x02) + + +/** + * \brief Stringify encoder options + * + * Show the filter-specific options that the encoder will use. + * This may be useful for verbose diagnostic messages. + * + * Note that if options were decoded from .xz headers then the encoder options + * may be undefined. This flag shouldn't be used in such a situation. + */ +#define LZMA_STR_ENCODER UINT32_C(0x10) + + +/** + * \brief Stringify decoder options + * + * Show the filter-specific options that the decoder will use. + * This may be useful for showing what filter options were decoded + * from file headers. + */ +#define LZMA_STR_DECODER UINT32_C(0x20) + + +/** + * \brief Produce xz-compatible getopt_long() syntax + * + * That is, "delta:dist=2 lzma2:dict=4MiB,pb=1,lp=1" becomes + * "--delta=dist=2 --lzma2=dict=4MiB,pb=1,lp=1". + * + * This syntax is compatible with xz 5.0.0 as long as the filters and + * their options are supported too. + */ +#define LZMA_STR_GETOPT_LONG UINT32_C(0x40) + + +/** + * \brief Use two dashes "--" instead of a space to separate filters + * + * That is, "delta:dist=2 lzma2:pb=1,lp=1" becomes + * "delta:dist=2--lzma2:pb=1,lp=1". This looks slightly odd but this + * kind of strings should be usable on the command line without quoting. + * However, it is possible that future versions with new filter options + * might produce strings that require shell quoting anyway as the exact + * set of possible characters isn't frozen for now. + * + * It is guaranteed that the single quote (') will never be used in + * filter chain strings (even if LZMA_STR_NO_SPACES isn't used). + */ +#define LZMA_STR_NO_SPACES UINT32_C(0x80) + + +/** + * \brief Convert a string to a filter chain + * + * This tries to make it easier to write applications that allow users + * to set custom compression options. This only handles the filter + * configuration (including presets) but not the number of threads, + * block size, check type, or memory limits. + * + * The input string can be either a preset or a filter chain. Presets + * begin with a digit 0-9 and may be followed by zero or more flags + * which are lower-case letters. Currently only "e" is supported, matching + * LZMA_PRESET_EXTREME. For partial xz command line syntax compatibility, + * a preset string may start with a single dash "-". + * + * A filter chain consists of one or more "filtername:opt1=value1,opt2=value2" + * strings separated by one or more spaces. Leading and trailing spaces are + * ignored. All names and values must be lower-case. Extra commas in the + * option list are ignored. The order of filters is significant: when + * encoding, the uncompressed input data goes to the leftmost filter first. + * Normally "lzma2" is the last filter in the chain. + * + * If one wishes to avoid spaces, for example, to avoid shell quoting, + * it is possible to use two dashes "--" instead of spaces to separate + * the filters. + * + * For xz command line compatibility, each filter may be prefixed with + * two dashes "--" and the colon ":" separating the filter name from + * the options may be replaced with an equals sign "=". + * + * By default, only filters that can be used in the .xz format are accepted. + * To allow all filters (LZMA1) use the flag LZMA_STR_ALL_FILTERS. + * + * By default, very basic validation is done for the filter chain as a whole, + * for example, that LZMA2 is only used as the last filter in the chain. + * The validation isn't perfect though and it's possible that this function + * succeeds but using the filter chain for encoding or decoding will still + * result in LZMA_OPTIONS_ERROR. To disable this validation, use the flag + * LZMA_STR_NO_VALIDATION. + * + * The available filter names and their options are available via + * lzma_str_list_filters(). See the xz man page for the description + * of filter names and options. + * + * \param str User-supplied string describing a preset or + * a filter chain. If a default value is needed and + * you don't know what would be good, use "6" since + * that is the default preset in xz too. + * \param error_pos If this isn't NULL, this value will be set on + * both success and on all errors. This tells the + * location of the error in the string. This is + * an int to make it straightforward to use this + * as printf() field width. The value is guaranteed + * to be in the range [0, INT_MAX] even if strlen(str) + * somehow was greater than INT_MAX. + * \param filters An array of lzma_filter structures. There must + * be LZMA_FILTERS_MAX + 1 (that is, five) elements + * in the array. The old contents are ignored so it + * doesn't need to be initialized. This array is + * modified only if this function returns LZMA_OK. + * Once the allocated filter options are no longer + * needed, lzma_filters_free() can be used to free the + * options (it doesn't free the filters array itself). + * \param flags Bitwise-or of zero or more of the flags + * LZMA_STR_ALL_FILTERS and LZMA_STR_NO_VALIDATION. + * \param allocator lzma_allocator for custom allocator functions. + * Set to NULL to use malloc() and free(). + * + * \return On success, NULL is returned. On error, a statically-allocated + * error message is returned which together with the error_pos + * should give some idea what is wrong. + * + * For command line applications, below is an example how an error message + * can be displayed. Note the use of an empty string for the field width. + * If "^" was used there it would create an off-by-one error except at + * the very beginning of the line. + * + * \code{.c} + * const char *str = ...; // From user + * lzma_filter filters[LZMA_FILTERS_MAX + 1]; + * int pos; + * const char *msg = lzma_str_to_filters(str, &pos, filters, 0, NULL); + * if (msg != NULL) { + * printf("%s: Error in XZ compression options:\n", argv[0]); + * printf("%s: %s\n", argv[0], str); + * printf("%s: %*s^\n", argv[0], errpos, ""); + * printf("%s: %s\n", argv[0], msg); + * } + * \endcode + */ +extern LZMA_API(const char *) lzma_str_to_filters( + const char *str, int *error_pos, lzma_filter *filters, + uint32_t flags, const lzma_allocator *allocator) + lzma_nothrow lzma_attr_warn_unused_result; + + +/** + * \brief Convert a filter chain to a string + * + * Use cases: + * + * - Verbose output showing the full encoder options to the user + * (use LZMA_STR_ENCODER in flags) + * + * - Showing the filters and options that are required to decode a file + * (use LZMA_STR_DECODER in flags) + * + * - Showing the filter names without any options in informational messages + * where the technical details aren't important (no flags). In this case + * the .options in the filters array are ignored and may be NULL even if + * a filter has a mandatory options structure. + * + * Note that even if the filter chain was specified using a preset, + * the resulting filter chain isn't reversed to a preset. So if you + * specify "6" to lzma_str_to_filters() then lzma_str_from_filters() + * will produce a string containing "lzma2". + * + * \param str On success *str will be set to point to an + * allocated string describing the given filter + * chain. Old value is ignored. On error *str is + * always set to NULL. + * \param filters Array of 1-4 filters and a terminating element + * with .id = LZMA_VLI_UNKNOWN. + * \param flags Bitwise-or of zero or more of the flags + * LZMA_STR_ENCODER, LZMA_STR_DECODER, + * LZMA_STR_GETOPT_LONG, and LZMA_STR_NO_SPACES. + * \param allocator lzma_allocator for custom allocator functions. + * Set to NULL to use malloc() and free(). + * + * \return - LZMA_OK + * - LZMA_OPTIONS_ERROR: Empty filter chain + * (filters[0].id == LZMA_VLI_UNKNOWN) or the filter chain + * includes a Filter ID that is not supported by this function. + * - LZMA_MEM_ERROR + * - LZMA_PROG_ERROR + */ +extern LZMA_API(lzma_ret) lzma_str_from_filters( + char **str, const lzma_filter *filters, uint32_t flags, + const lzma_allocator *allocator) + lzma_nothrow lzma_attr_warn_unused_result; + + +/** + * \brief List available filters and/or their options (for help message) + * + * If a filter_id is given then only one line is created which contains the + * filter name. If LZMA_STR_ENCODER or LZMA_STR_DECODER is used then the + * options required for encoding or decoding are listed on the same line too. + * + * If filter_id is LZMA_VLI_UNKNOWN then all supported .xz-compatible filters + * are listed: + * + * - If neither LZMA_STR_ENCODER nor LZMA_STR_DECODER is used then + * the supported filter names are listed on a single line separated + * by spaces. + * + * - If LZMA_STR_ENCODER or LZMA_STR_DECODER is used then filters and + * the supported options are listed one filter per line. There won't + * be a '\n' after the last filter. + * + * - If LZMA_STR_ALL_FILTERS is used then the list will include also + * those filters that cannot be used in the .xz format (LZMA1). + * + * \param str On success *str will be set to point to an + * allocated string listing the filters and options. + * Old value is ignored. On error *str is always set + * to NULL. + * \param filter_id Filter ID or LZMA_VLI_UNKNOWN. + * \param flags Bitwise-or of zero or more of the flags + * LZMA_STR_ALL_FILTERS, LZMA_STR_ENCODER, + * LZMA_STR_DECODER, and LZMA_STR_GETOPT_LONG. + * \param allocator lzma_allocator for custom allocator functions. + * Set to NULL to use malloc() and free(). + * + * \return - LZMA_OK + * - LZMA_OPTIONS_ERROR: Unsupported filter_id or flags + * - LZMA_MEM_ERROR + * - LZMA_PROG_ERROR + */ +extern LZMA_API(lzma_ret) lzma_str_list_filters( + char **str, lzma_vli filter_id, uint32_t flags, + const lzma_allocator *allocator) + lzma_nothrow lzma_attr_warn_unused_result; diff --git a/contrib/libs/lzma/liblzma/api/lzma/hardware.h b/contrib/libs/lzma/liblzma/api/lzma/hardware.h index 47481f2581..7040aae45f 100644 --- a/contrib/libs/lzma/liblzma/api/lzma/hardware.h +++ b/contrib/libs/lzma/liblzma/api/lzma/hardware.h @@ -57,7 +57,7 @@ extern LZMA_API(uint64_t) lzma_physmem(void) lzma_nothrow; * If the hardware supports more than one thread per CPU core, the number * of hardware threads is returned if that information is available. * - * \brief On success, the number of available CPU threads or cores is + * \return On success, the number of available CPU threads or cores is * returned. If this information isn't available or an error * occurs, zero is returned. */ diff --git a/contrib/libs/lzma/liblzma/api/lzma/index.h b/contrib/libs/lzma/liblzma/api/lzma/index.h index 3dac6fb85c..144d416615 100644 --- a/contrib/libs/lzma/liblzma/api/lzma/index.h +++ b/contrib/libs/lzma/liblzma/api/lzma/index.h @@ -684,3 +684,69 @@ extern LZMA_API(lzma_ret) lzma_index_buffer_decode(lzma_index **i, uint64_t *memlimit, const lzma_allocator *allocator, const uint8_t *in, size_t *in_pos, size_t in_size) lzma_nothrow; + + +/** + * \brief Initialize a .xz file information decoder + * + * \param strm Pointer to a properly prepared lzma_stream + * \param dest_index Pointer to a pointer where the decoder will put + * the decoded lzma_index. The old value + * of *dest_index is ignored (not freed). + * \param memlimit How much memory the resulting lzma_index is + * allowed to require. Use UINT64_MAX to + * effectively disable the limiter. + * \param file_size Size of the input .xz file + * + * This decoder decodes the Stream Header, Stream Footer, Index, and + * Stream Padding field(s) from the input .xz file and stores the resulting + * combined index in *dest_index. This information can be used to get the + * uncompressed file size with lzma_index_uncompressed_size(*dest_index) or, + * for example, to implement random access reading by locating the Blocks + * in the Streams. + * + * To get the required information from the .xz file, lzma_code() may ask + * the application to seek in the input file by returning LZMA_SEEK_NEEDED + * and having the target file position specified in lzma_stream.seek_pos. + * The number of seeks required depends on the input file and how big buffers + * the application provides. When possible, the decoder will seek backward + * and forward in the given buffer to avoid useless seek requests. Thus, if + * the application provides the whole file at once, no external seeking will + * be required (that is, lzma_code() won't return LZMA_SEEK_NEEDED). + * + * The value in lzma_stream.total_in can be used to estimate how much data + * liblzma had to read to get the file information. However, due to seeking + * and the way total_in is updated, the value of total_in will be somewhat + * inaccurate (a little too big). Thus, total_in is a good estimate but don't + * expect to see the same exact value for the same file if you change the + * input buffer size or switch to a different liblzma version. + * + * Valid `action' arguments to lzma_code() are LZMA_RUN and LZMA_FINISH. + * You only need to use LZMA_RUN; LZMA_FINISH is only supported because it + * might be convenient for some applications. If you use LZMA_FINISH and if + * lzma_code() asks the application to seek, remember to reset `action' back + * to LZMA_RUN unless you hit the end of the file again. + * + * Possible return values from lzma_code(): + * - LZMA_OK: All OK so far, more input needed + * - LZMA_SEEK_NEEDED: Provide more input starting from the absolute + * file position strm->seek_pos + * - LZMA_STREAM_END: Decoding was successful, *dest_index has been set + * - LZMA_FORMAT_ERROR: The input file is not in the .xz format (the + * expected magic bytes were not found from the beginning of the file) + * - LZMA_OPTIONS_ERROR: File looks valid but contains headers that aren't + * supported by this version of liblzma + * - LZMA_DATA_ERROR: File is corrupt + * - LZMA_BUF_ERROR + * - LZMA_MEM_ERROR + * - LZMA_MEMLIMIT_ERROR + * - LZMA_PROG_ERROR + * + * \return - LZMA_OK + * - LZMA_MEM_ERROR + * - LZMA_PROG_ERROR + */ +extern LZMA_API(lzma_ret) lzma_file_info_decoder( + lzma_stream *strm, lzma_index **dest_index, + uint64_t memlimit, uint64_t file_size) + lzma_nothrow; diff --git a/contrib/libs/lzma/liblzma/api/lzma/index_hash.h b/contrib/libs/lzma/liblzma/api/lzma/index_hash.h index 9287f1dfdb..8ff4163365 100644 --- a/contrib/libs/lzma/liblzma/api/lzma/index_hash.h +++ b/contrib/libs/lzma/liblzma/api/lzma/index_hash.h @@ -52,7 +52,7 @@ extern LZMA_API(void) lzma_index_hash_end( /** * \brief Add a new Record to an Index hash * - * \param index Pointer to a lzma_index_hash structure + * \param index_hash Pointer to a lzma_index_hash structure * \param unpadded_size Unpadded Size of a Block * \param uncompressed_size Uncompressed Size of a Block * diff --git a/contrib/libs/lzma/liblzma/api/lzma/lzma12.h b/contrib/libs/lzma/liblzma/api/lzma/lzma12.h index df5f23b61a..d34e78392b 100644 --- a/contrib/libs/lzma/liblzma/api/lzma/lzma12.h +++ b/contrib/libs/lzma/liblzma/api/lzma/lzma12.h @@ -18,18 +18,41 @@ /** - * \brief LZMA1 Filter ID + * \brief LZMA1 Filter ID (for raw encoder/decoder only, not in .xz) * * LZMA1 is the very same thing as what was called just LZMA in LZMA Utils, * 7-Zip, and LZMA SDK. It's called LZMA1 here to prevent developers from * accidentally using LZMA when they actually want LZMA2. - * - * LZMA1 shouldn't be used for new applications unless you _really_ know - * what you are doing. LZMA2 is almost always a better choice. */ #define LZMA_FILTER_LZMA1 LZMA_VLI_C(0x4000000000000001) /** + * \brief LZMA1 Filter ID with extended options (for raw encoder/decoder) + * + * This is like LZMA_FILTER_LZMA1 but with this ID a few extra options + * are supported in the lzma_options_lzma structure: + * + * - A flag to tell the encoder if the end of payload marker (EOPM) alias + * end of stream (EOS) marker must be written at the end of the stream. + * In contrast, LZMA_FILTER_LZMA1 always writes the end marker. + * + * - Decoder needs to be told the uncompressed size of the stream + * or that it is unknown (using the special value UINT64_MAX). + * If the size is known, a flag can be set to allow the presence of + * the end marker anyway. In contrast, LZMA_FILTER_LZMA1 always + * behaves as if the uncompressed size was unknown. + * + * This allows handling file formats where LZMA1 streams are used but where + * the end marker isn't allowed or where it might not (always) be present. + * This extended LZMA1 functionality is provided as a Filter ID for raw + * encoder and decoder instead of adding new encoder and decoder initialization + * functions because this way it is possible to also use extra filters, + * for example, LZMA_FILTER_X86 in a filter chain with LZMA_FILTER_LZMA1EXT, + * which might be needed to handle some file formats. + */ +#define LZMA_FILTER_LZMA1EXT LZMA_VLI_C(0x4000000000000002) + +/** * \brief LZMA2 Filter ID * * Usually you want this instead of LZMA1. Compared to LZMA1, LZMA2 adds @@ -374,6 +397,82 @@ typedef struct { */ uint32_t depth; + /** + * \brief For LZMA_FILTER_LZMA1EXT: Extended flags + * + * This is used only with LZMA_FILTER_LZMA1EXT. + * + * Currently only one flag is supported, LZMA_LZMA1EXT_ALLOW_EOPM: + * + * - Encoder: If the flag is set, then end marker is written just + * like it is with LZMA_FILTER_LZMA1. Without this flag the + * end marker isn't written and the application has to store + * the uncompressed size somewhere outside the compressed stream. + * To decompress streams without the end marker, the appliation + * has to set the correct uncompressed size in ext_size_low and + * ext_size_high. + * + * - Decoder: If the uncompressed size in ext_size_low and + * ext_size_high is set to the special value UINT64_MAX + * (indicating unknown uncompressed size) then this flag is + * ignored and the end marker must always be present, that is, + * the behavior is identical to LZMA_FILTER_LZMA1. + * + * Otherwise, if this flag isn't set, then the input stream + * must not have the end marker; if the end marker is detected + * then it will result in LZMA_DATA_ERROR. This is useful when + * it is known that the stream must not have the end marker and + * strict validation is wanted. + * + * If this flag is set, then it is autodetected if the end marker + * is present after the specified number of uncompressed bytes + * has been decompressed (ext_size_low and ext_size_high). The + * end marker isn't allowed in any other position. This behavior + * is useful when uncompressed size is known but the end marker + * may or may not be present. This is the case, for example, + * in .7z files (valid .7z files that have the end marker in + * LZMA1 streams are rare but they do exist). + */ + uint32_t ext_flags; +# define LZMA_LZMA1EXT_ALLOW_EOPM UINT32_C(0x01) + + /** + * \brief For LZMA_FILTER_LZMA1EXT: Uncompressed size (low bits) + * + * The 64-bit uncompressed size is needed for decompression with + * LZMA_FILTER_LZMA1EXT. The size is ignored by the encoder. + * + * The special value UINT64_MAX indicates that the uncompressed size + * is unknown and that the end of payload marker (also known as + * end of stream marker) must be present to indicate the end of + * the LZMA1 stream. Any other value indicates the expected + * uncompressed size of the LZMA1 stream. (If LZMA1 was used together + * with filters that change the size of the data then the uncompressed + * size of the LZMA1 stream could be different than the final + * uncompressed size of the filtered stream.) + * + * ext_size_low holds the least significant 32 bits of the + * uncompressed size. The most significant 32 bits must be set + * in ext_size_high. The macro lzma_ext_size_set(opt_lzma, u64size) + * can be used to set these members. + * + * The 64-bit uncompressed size is split into two uint32_t variables + * because there were no reserved uint64_t members and using the + * same options structure for LZMA_FILTER_LZMA1, LZMA_FILTER_LZMA1EXT, + * and LZMA_FILTER_LZMA2 was otherwise more convenient than having + * a new options structure for LZMA_FILTER_LZMA1EXT. (Replacing two + * uint32_t members with one uint64_t changes the ABI on some systems + * as the alignment of this struct can increase from 4 bytes to 8.) + */ + uint32_t ext_size_low; + + /** + * \brief For LZMA_FILTER_LZMA1EXT: Uncompressed size (high bits) + * + * This holds the most significant 32 bits of the uncompressed size. + */ + uint32_t ext_size_high; + /* * Reserved space to allow possible future extensions without * breaking the ABI. You should not touch these, because the names @@ -381,9 +480,6 @@ typedef struct { * with the currently supported options, so it is safe to leave these * uninitialized. */ - uint32_t reserved_int1; - uint32_t reserved_int2; - uint32_t reserved_int3; uint32_t reserved_int4; uint32_t reserved_int5; uint32_t reserved_int6; @@ -400,6 +496,19 @@ typedef struct { /** + * \brief Macro to set the 64-bit uncompressed size in ext_size_* + * + * This might be convenient when decoding using LZMA_FILTER_LZMA1EXT. + * This isn't used with LZMA_FILTER_LZMA1 or LZMA_FILTER_LZMA2. + */ +#define lzma_set_ext_size(opt_lzma2, u64size) \ +do { \ + (opt_lzma2).ext_size_low = (uint32_t)(u64size); \ + (opt_lzma2).ext_size_high = (uint32_t)((uint64_t)(u64size) >> 32); \ +} while (0) + + +/** * \brief Set a compression preset to lzma_options_lzma structure * * 0 is the fastest and 9 is the slowest. These match the switches -0 .. -9 diff --git a/contrib/libs/lzma/liblzma/api/lzma/version.h b/contrib/libs/lzma/liblzma/api/lzma/version.h index 2bf3eaed24..7d37130f24 100644 --- a/contrib/libs/lzma/liblzma/api/lzma/version.h +++ b/contrib/libs/lzma/liblzma/api/lzma/version.h @@ -21,8 +21,8 @@ * Version number split into components */ #define LZMA_VERSION_MAJOR 5 -#define LZMA_VERSION_MINOR 2 -#define LZMA_VERSION_PATCH 5 +#define LZMA_VERSION_MINOR 4 +#define LZMA_VERSION_PATCH 0 #define LZMA_VERSION_STABILITY LZMA_VERSION_STABILITY_STABLE #ifndef LZMA_VERSION_COMMIT diff --git a/contrib/libs/lzma/liblzma/api/lzma/vli.h b/contrib/libs/lzma/liblzma/api/lzma/vli.h index 1b7a952a40..7a0a803dfc 100644 --- a/contrib/libs/lzma/liblzma/api/lzma/vli.h +++ b/contrib/libs/lzma/liblzma/api/lzma/vli.h @@ -159,6 +159,8 @@ extern LZMA_API(lzma_ret) lzma_vli_decode(lzma_vli *vli, size_t *vli_pos, /** * \brief Get the number of bytes required to encode a VLI * + * \param vli Integer whose encoded size is to be determined + * * \return Number of bytes on success (1-9). If vli isn't valid, * zero is returned. */ diff --git a/contrib/libs/lzma/liblzma/check/crc64_fast.c b/contrib/libs/lzma/liblzma/check/crc64_fast.c index 8af54cda7b..e3cbf1b1e9 100644 --- a/contrib/libs/lzma/liblzma/check/crc64_fast.c +++ b/contrib/libs/lzma/liblzma/check/crc64_fast.c @@ -3,11 +3,25 @@ /// \file crc64.c /// \brief CRC64 calculation /// -/// Calculate the CRC64 using the slice-by-four algorithm. This is the same -/// idea that is used in crc32_fast.c, but for CRC64 we use only four tables +/// There are two methods in this file. crc64_generic uses the +/// the slice-by-four algorithm. This is the same idea that is +/// used in crc32_fast.c, but for CRC64 we use only four tables /// instead of eight to avoid increasing CPU cache usage. +/// +/// crc64_clmul uses 32/64-bit x86 SSSE3, SSE4.1, and CLMUL instructions. +/// It was derived from +/// https://www.intel.com/content/dam/www/public/us/en/documents/white-papers/fast-crc-computation-generic-polynomials-pclmulqdq-paper.pdf +/// and the public domain code from https://github.com/rawrunprotected/crc +/// (URLs were checked on 2022-11-07). +/// +/// FIXME: Builds for 32-bit x86 use crc64_x86.S by default instead +/// of this file and thus CLMUL version isn't available on 32-bit x86 +/// unless configured with --disable-assembler. Even then the lookup table +/// isn't omitted in crc64_table.c since it doesn't know that assembly +/// code has been disabled. // -// Author: Lasse Collin +// Authors: Lasse Collin +// Ilya Kurdyukov // // This file has been put into the public domain. // You can do whatever you want with this file. @@ -15,6 +29,54 @@ /////////////////////////////////////////////////////////////////////////////// #include "check.h" + +#undef CRC_GENERIC +#undef CRC_CLMUL +#undef CRC_USE_GENERIC_FOR_SMALL_INPUTS + +// If CLMUL cannot be used then only the generic slice-by-four is built. +#if !defined(HAVE_USABLE_CLMUL) +# define CRC_GENERIC 1 + +// If CLMUL is allowed unconditionally in the compiler options then the +// generic version can be omitted. Note that this doesn't work with MSVC +// as I don't know how to detect the features here. +// +// NOTE: Keep this this in sync with crc64_table.c. +#elif (defined(__SSSE3__) && defined(__SSE4_1__) && defined(__PCLMUL__)) \ + || (defined(__e2k__) && __iset__ >= 6) +# define CRC_CLMUL 1 + +// Otherwise build both and detect at runtime which version to use. +#else +# define CRC_GENERIC 1 +# define CRC_CLMUL 1 + +/* + // The generic code is much faster with 1-8-byte inputs and has + // similar performance up to 16 bytes at least in microbenchmarks + // (it depends on input buffer alignment too). If both versions are + // built, this #define will use the generic version for inputs up to + // 16 bytes and CLMUL for bigger inputs. It saves a little in code + // size since the special cases for 0-16-byte inputs will be omitted + // from the CLMUL code. +# define CRC_USE_GENERIC_FOR_SMALL_INPUTS 1 +*/ + +# if defined(_MSC_VER) +# include <intrin.h> +# elif defined(HAVE_CPUID_H) +# include <cpuid.h> +# endif +#endif + + +///////////////////////////////// +// Generic slice-by-four CRC64 // +///////////////////////////////// + +#ifdef CRC_GENERIC + #include "crc_macros.h" @@ -26,8 +88,8 @@ // See the comments in crc32_fast.c. They aren't duplicated here. -extern LZMA_API(uint64_t) -lzma_crc64(const uint8_t *buf, size_t size, uint64_t crc) +static uint64_t +crc64_generic(const uint8_t *buf, size_t size, uint64_t crc) { crc = ~crc; @@ -46,10 +108,11 @@ lzma_crc64(const uint8_t *buf, size_t size, uint64_t crc) while (buf < limit) { #ifdef WORDS_BIGENDIAN - const uint32_t tmp = (crc >> 32) + const uint32_t tmp = (uint32_t)(crc >> 32) ^ aligned_read32ne(buf); #else - const uint32_t tmp = crc ^ aligned_read32ne(buf); + const uint32_t tmp = (uint32_t)crc + ^ aligned_read32ne(buf); #endif buf += 4; @@ -70,3 +133,380 @@ lzma_crc64(const uint8_t *buf, size_t size, uint64_t crc) return ~crc; } +#endif + + +///////////////////// +// x86 CLMUL CRC64 // +///////////////////// + +#ifdef CRC_CLMUL + +#include <immintrin.h> + + +/* +// These functions were used to generate the constants +// at the top of crc64_clmul(). +static uint64_t +calc_lo(uint64_t poly) +{ + uint64_t a = poly; + uint64_t b = 0; + + for (unsigned i = 0; i < 64; ++i) { + b = (b >> 1) | (a << 63); + a = (a >> 1) ^ (a & 1 ? poly : 0); + } + + return b; +} + +static uint64_t +calc_hi(uint64_t poly, uint64_t a) +{ + for (unsigned i = 0; i < 64; ++i) + a = (a >> 1) ^ (a & 1 ? poly : 0); + + return a; +} +*/ + + +#define MASK_L(in, mask, r) \ + r = _mm_shuffle_epi8(in, mask) + +#define MASK_H(in, mask, r) \ + r = _mm_shuffle_epi8(in, _mm_xor_si128(mask, vsign)) + +#define MASK_LH(in, mask, low, high) \ + MASK_L(in, mask, low); \ + MASK_H(in, mask, high) + + +// EDG-based compilers (Intel's classic compiler and compiler for E2K) can +// define __GNUC__ but the attribute must not be used with them. +// The new Clang-based ICX needs the attribute. +// +// NOTE: Build systems check for this too, keep them in sync with this. +#if (defined(__GNUC__) || defined(__clang__)) && !defined(__EDG__) +__attribute__((__target__("ssse3,sse4.1,pclmul"))) +#endif +static uint64_t +crc64_clmul(const uint8_t *buf, size_t size, uint64_t crc) +{ + // The prototypes of the intrinsics use signed types while most of + // the values are treated as unsigned here. These warnings in this + // function have been checked and found to be harmless so silence them. +#if TUKLIB_GNUC_REQ(4, 6) || defined(__clang__) +# pragma GCC diagnostic push +# pragma GCC diagnostic ignored "-Wsign-conversion" +# pragma GCC diagnostic ignored "-Wconversion" +#endif + +#ifndef CRC_USE_GENERIC_FOR_SMALL_INPUTS + // The code assumes that there is at least one byte of input. + if (size == 0) + return crc; +#endif + + // const uint64_t poly = 0xc96c5795d7870f42; // CRC polynomial + const uint64_t p = 0x92d8af2baf0e1e85; // (poly << 1) | 1 + const uint64_t mu = 0x9c3e466c172963d5; // (calc_lo(poly) << 1) | 1 + const uint64_t k2 = 0xdabe95afc7875f40; // calc_hi(poly, 1) + const uint64_t k1 = 0xe05dd497ca393ae4; // calc_hi(poly, k2) + const __m128i vfold0 = _mm_set_epi64x(p, mu); + const __m128i vfold1 = _mm_set_epi64x(k2, k1); + + // Create a vector with 8-bit values 0 to 15. This is used to + // construct control masks for _mm_blendv_epi8 and _mm_shuffle_epi8. + const __m128i vramp = _mm_setr_epi32( + 0x03020100, 0x07060504, 0x0b0a0908, 0x0f0e0d0c); + + // This is used to inverse the control mask of _mm_shuffle_epi8 + // so that bytes that wouldn't be picked with the original mask + // will be picked and vice versa. + const __m128i vsign = _mm_set1_epi8(0x80); + + // Memory addresses A to D and the distances between them: + // + // A B C D + // [skip_start][size][skip_end] + // [ size2 ] + // + // A and D are 16-byte aligned. B and C are 1-byte aligned. + // skip_start and skip_end are 0-15 bytes. size is at least 1 byte. + // + // A = aligned_buf will initially point to this address. + // B = The address pointed by the caller-supplied buf. + // C = buf + size == aligned_buf + size2 + // D = buf + size + skip_end == aligned_buf + size2 + skip_end + const size_t skip_start = (size_t)((uintptr_t)buf & 15); + const size_t skip_end = (size_t)(-(uintptr_t)(buf + size) & 15); + const __m128i *aligned_buf = (const __m128i *)( + (uintptr_t)buf & ~(uintptr_t)15); + + // If size2 <= 16 then the whole input fits into a single 16-byte + // vector. If size2 > 16 then at least two 16-byte vectors must + // be processed. If size2 > 16 && size <= 16 then there is only + // one 16-byte vector's worth of input but it is unaligned in memory. + // + // NOTE: There is no integer overflow here if the arguments are valid. + // If this overflowed, buf + size would too. + size_t size2 = skip_start + size; + + // Masks to be used with _mm_blendv_epi8 and _mm_shuffle_epi8: + // The first skip_start or skip_end bytes in the vectors will have + // the high bit (0x80) set. _mm_blendv_epi8 and _mm_shuffle_epi8 + // will produce zeros for these positions. (Bitwise-xor of these + // masks with vsign will produce the opposite behavior.) + const __m128i mask_start + = _mm_sub_epi8(vramp, _mm_set1_epi8(skip_start)); + const __m128i mask_end = _mm_sub_epi8(vramp, _mm_set1_epi8(skip_end)); + + // Get the first 1-16 bytes into data0. If loading less than 16 bytes, + // the bytes are loaded to the high bits of the vector and the least + // significant positions are filled with zeros. + const __m128i data0 = _mm_blendv_epi8(_mm_load_si128(aligned_buf), + _mm_setzero_si128(), mask_start); + ++aligned_buf; + +#if defined(__i386__) || defined(_M_IX86) + const __m128i initial_crc = _mm_set_epi64x(0, ~crc); +#else + // GCC and Clang would produce good code with _mm_set_epi64x + // but MSVC needs _mm_cvtsi64_si128 on x86-64. + const __m128i initial_crc = _mm_cvtsi64_si128(~crc); +#endif + + __m128i v0, v1, v2, v3; + +#ifndef CRC_USE_GENERIC_FOR_SMALL_INPUTS + if (size <= 16) { + // Right-shift initial_crc by 1-16 bytes based on "size" + // and store the result in v1 (high bytes) and v0 (low bytes). + // + // NOTE: The highest 8 bytes of initial_crc are zeros so + // v1 will be filled with zeros if size >= 8. The highest 8 + // bytes of v1 will always become zeros. + // + // [ v1 ][ v0 ] + // [ initial_crc ] size == 1 + // [ initial_crc ] size == 2 + // [ initial_crc ] size == 15 + // [ initial_crc ] size == 16 (all in v0) + const __m128i mask_low = _mm_add_epi8( + vramp, _mm_set1_epi8(size - 16)); + MASK_LH(initial_crc, mask_low, v0, v1); + + if (size2 <= 16) { + // There are 1-16 bytes of input and it is all + // in data0. Copy the input bytes to v3. If there + // are fewer than 16 bytes, the low bytes in v3 + // will be filled with zeros. That is, the input + // bytes are stored to the same position as + // (part of) initial_crc is in v0. + MASK_L(data0, mask_end, v3); + } else { + // There are 2-16 bytes of input but not all bytes + // are in data0. + const __m128i data1 = _mm_load_si128(aligned_buf); + + // Collect the 2-16 input bytes from data0 and data1 + // to v2 and v3, and bitwise-xor them with the + // low bits of initial_crc in v0. Note that the + // the second xor is below this else-block as it + // is shared with the other branch. + MASK_H(data0, mask_end, v2); + MASK_L(data1, mask_end, v3); + v0 = _mm_xor_si128(v0, v2); + } + + v0 = _mm_xor_si128(v0, v3); + v1 = _mm_alignr_epi8(v1, v0, 8); + } else +#endif + { + const __m128i data1 = _mm_load_si128(aligned_buf); + MASK_LH(initial_crc, mask_start, v0, v1); + v0 = _mm_xor_si128(v0, data0); + v1 = _mm_xor_si128(v1, data1); + +#define FOLD \ + v1 = _mm_xor_si128(v1, _mm_clmulepi64_si128(v0, vfold1, 0x00)); \ + v0 = _mm_xor_si128(v1, _mm_clmulepi64_si128(v0, vfold1, 0x11)); + + while (size2 > 32) { + ++aligned_buf; + size2 -= 16; + FOLD + v1 = _mm_load_si128(aligned_buf); + } + + if (size2 < 32) { + MASK_H(v0, mask_end, v2); + MASK_L(v0, mask_end, v0); + MASK_L(v1, mask_end, v3); + v1 = _mm_or_si128(v2, v3); + } + + FOLD + v1 = _mm_srli_si128(v0, 8); +#undef FOLD + } + + v1 = _mm_xor_si128(_mm_clmulepi64_si128(v0, vfold1, 0x10), v1); + v0 = _mm_clmulepi64_si128(v1, vfold0, 0x00); + v2 = _mm_clmulepi64_si128(v0, vfold0, 0x10); + v0 = _mm_xor_si128(_mm_xor_si128(v2, _mm_slli_si128(v0, 8)), v1); + +#if defined(__i386__) || defined(_M_IX86) + return ~(((uint64_t)(uint32_t)_mm_extract_epi32(v0, 3) << 32) | + (uint64_t)(uint32_t)_mm_extract_epi32(v0, 2)); +#else + return ~(uint64_t)_mm_extract_epi64(v0, 1); +#endif + +#if TUKLIB_GNUC_REQ(4, 6) || defined(__clang__) +# pragma GCC diagnostic pop +#endif +} +#endif + + +//////////////////////// +// Detect CPU support // +//////////////////////// + +#if defined(CRC_GENERIC) && defined(CRC_CLMUL) +static inline bool +is_clmul_supported(void) +{ + int success = 1; + uint32_t r[4]; // eax, ebx, ecx, edx + +#if defined(_MSC_VER) + // This needs <intrin.h> with MSVC. ICC has it as a built-in + // on all platforms. + __cpuid(r, 1); +#elif defined(HAVE_CPUID_H) + // Compared to just using __asm__ to run CPUID, this also checks + // that CPUID is supported and saves and restores ebx as that is + // needed with GCC < 5 with position-independent code (PIC). + success = __get_cpuid(1, &r[0], &r[1], &r[2], &r[3]); +#else + // Just a fallback that shouldn't be needed. + __asm__("cpuid\n\t" + : "=a"(r[0]), "=b"(r[1]), "=c"(r[2]), "=d"(r[3]) + : "a"(1), "c"(0)); +#endif + + // Returns true if these are supported: + // CLMUL (bit 1 in ecx) + // SSSE3 (bit 9 in ecx) + // SSE4.1 (bit 19 in ecx) + const uint32_t ecx_mask = (1 << 1) | (1 << 9) | (1 << 19); + return success && (r[2] & ecx_mask) == ecx_mask; + + // Alternative methods that weren't used: + // - ICC's _may_i_use_cpu_feature: the other methods should work too. + // - GCC >= 6 / Clang / ICX __builtin_cpu_supports("pclmul") + // + // CPUID decding is needed with MSVC anyway and older GCC. This keeps + // the feature checks in the build system simpler too. The nice thing + // about __builtin_cpu_supports would be that it generates very short + // code as is it only reads a variable set at startup but a few bytes + // doesn't matter here. +} + + +#ifdef HAVE_FUNC_ATTRIBUTE_CONSTRUCTOR +# define CRC64_FUNC_INIT +# define CRC64_SET_FUNC_ATTR __attribute__((__constructor__)) +#else +# define CRC64_FUNC_INIT = &crc64_dispatch +# define CRC64_SET_FUNC_ATTR +static uint64_t crc64_dispatch(const uint8_t *buf, size_t size, uint64_t crc); +#endif + + +// Pointer to the the selected CRC64 method. +static uint64_t (*crc64_func)(const uint8_t *buf, size_t size, uint64_t crc) + CRC64_FUNC_INIT; + + +CRC64_SET_FUNC_ATTR +static void +crc64_set_func(void) +{ + crc64_func = is_clmul_supported() ? &crc64_clmul : &crc64_generic; + return; +} + + +#ifndef HAVE_FUNC_ATTRIBUTE_CONSTRUCTOR +static uint64_t +crc64_dispatch(const uint8_t *buf, size_t size, uint64_t crc) +{ + // When __attribute__((__constructor__)) isn't supported, set the + // function pointer without any locking. If multiple threads run + // the detection code in parallel, they will all end up setting + // the pointer to the same value. This avoids the use of + // mythread_once() on every call to lzma_crc64() but this likely + // isn't strictly standards compliant. Let's change it if it breaks. + crc64_set_func(); + return crc64_func(buf, size, crc); +} +#endif +#endif + + +extern LZMA_API(uint64_t) +lzma_crc64(const uint8_t *buf, size_t size, uint64_t crc) +{ +#if defined(CRC_GENERIC) && defined(CRC_CLMUL) + // If CLMUL is available, it is the best for non-tiny inputs, + // being over twice as fast as the generic slice-by-four version. + // However, for size <= 16 it's different. In the extreme case + // of size == 1 the generic version can be five times faster. + // At size >= 8 the CLMUL starts to become reasonable. It + // varies depending on the alignment of buf too. + // + // The above doesn't include the overhead of mythread_once(). + // At least on x86-64 GNU/Linux, pthread_once() is very fast but + // it still makes lzma_crc64(buf, 1, crc) 50-100 % slower. When + // size reaches 12-16 bytes the overhead becomes negligible. + // + // So using the generic version for size <= 16 may give better + // performance with tiny inputs but if such inputs happen rarely + // it's not so obvious because then the lookup table of the + // generic version may not be in the processor cache. +#ifdef CRC_USE_GENERIC_FOR_SMALL_INPUTS + if (size <= 16) + return crc64_generic(buf, size, crc); +#endif + +/* +#ifndef HAVE_FUNC_ATTRIBUTE_CONSTRUCTOR + // See crc64_dispatch(). This would be the alternative which uses + // locking and doesn't use crc64_dispatch(). Note that on Windows + // this method needs Vista threads. + mythread_once(crc64_set_func); +#endif +*/ + + return crc64_func(buf, size, crc); + +#elif defined(CRC_CLMUL) + // If CLMUL is used unconditionally without runtime CPU detection + // then omitting the generic version and its 8 KiB lookup table + // makes the library smaller. + // + // FIXME: Lookup table isn't currently omitted on 32-bit x86, + // see crc64_table.c. + return crc64_clmul(buf, size, crc); + +#else + return crc64_generic(buf, size, crc); +#endif +} diff --git a/contrib/libs/lzma/liblzma/check/crc64_table.c b/contrib/libs/lzma/liblzma/check/crc64_table.c index fa334df321..307846ab14 100644 --- a/contrib/libs/lzma/liblzma/check/crc64_table.c +++ b/contrib/libs/lzma/liblzma/check/crc64_table.c @@ -12,11 +12,24 @@ #include "common.h" + +// FIXME: Compared to crc64_fast.c this has to check for __x86_64__ too +// so that in 32-bit builds crc64_x86.S won't break due to a missing table. +#if (defined(__x86_64__) && defined(__SSSE3__) \ + && defined(__SSE4_1__) && defined(__PCLMUL__)) \ + || (defined(__e2k__) && __iset__ >= 6) +// No table needed but something has to be exported to keep some toolchains +// happy. Also use a declaration to silence compiler warnings. +extern const char lzma_crc64_dummy; +const char lzma_crc64_dummy; + +#else // Having the declaration here silences clang -Wmissing-variable-declarations. extern const uint64_t lzma_crc64_table[4][256]; -#ifdef WORDS_BIGENDIAN -# error #include "crc64_table_be.h" -#else -# include "crc64_table_le.h" +# if defined(WORDS_BIGENDIAN) +# error #include "crc64_table_be.h" +# else +# include "crc64_table_le.h" +# endif #endif diff --git a/contrib/libs/lzma/liblzma/common/alone_decoder.c b/contrib/libs/lzma/liblzma/common/alone_decoder.c index 239b230ef1..1dc85badf9 100644 --- a/contrib/libs/lzma/liblzma/common/alone_decoder.c +++ b/contrib/libs/lzma/liblzma/common/alone_decoder.c @@ -110,12 +110,24 @@ alone_decode(void *coder_ptr, const lzma_allocator *allocator, // Another hack to ditch false positives: Assume that // if the uncompressed size is known, it must be less // than 256 GiB. + // + // FIXME? Without picky we allow > LZMA_VLI_MAX which doesn't + // really matter in this specific situation (> LZMA_VLI_MAX is + // safe in the LZMA decoder) but it's somewhat weird still. if (coder->picky && coder->uncompressed_size != LZMA_VLI_UNKNOWN && coder->uncompressed_size >= (LZMA_VLI_C(1) << 38)) return LZMA_FORMAT_ERROR; + // Use LZMA_FILTER_LZMA1EXT features to specify the + // uncompressed size and that the end marker is allowed + // even when the uncompressed size is known. Both .lzma + // header and LZMA1EXT use UINT64_MAX indicate that size + // is unknown. + coder->options.ext_flags = LZMA_LZMA1EXT_ALLOW_EOPM; + lzma_set_ext_size(coder->options, coder->uncompressed_size); + // Calculate the memory usage so that it is ready // for SEQ_CODER_INIT. coder->memusage = lzma_lzma_decoder_memusage(&coder->options) @@ -132,6 +144,7 @@ alone_decode(void *coder_ptr, const lzma_allocator *allocator, lzma_filter_info filters[2] = { { + .id = LZMA_FILTER_LZMA1EXT, .init = &lzma_lzma_decoder_init, .options = &coder->options, }, { @@ -139,14 +152,8 @@ alone_decode(void *coder_ptr, const lzma_allocator *allocator, } }; - const lzma_ret ret = lzma_next_filter_init(&coder->next, - allocator, filters); - if (ret != LZMA_OK) - return ret; - - // Use a hack to set the uncompressed size. - lzma_lz_decoder_uncompressed(coder->next.coder, - coder->uncompressed_size); + return_if_error(lzma_next_filter_init(&coder->next, + allocator, filters)); coder->sequence = SEQ_CODE; break; diff --git a/contrib/libs/lzma/liblzma/common/alone_encoder.c b/contrib/libs/lzma/liblzma/common/alone_encoder.c index 96c1db70cc..c9392f3769 100644 --- a/contrib/libs/lzma/liblzma/common/alone_encoder.c +++ b/contrib/libs/lzma/liblzma/common/alone_encoder.c @@ -129,6 +129,7 @@ alone_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, // Initialize the LZMA encoder. const lzma_filter_info filters[2] = { { + .id = LZMA_FILTER_LZMA1, .init = &lzma_lzma_encoder_init, .options = (void *)(options), }, { diff --git a/contrib/libs/lzma/liblzma/common/auto_decoder.c b/contrib/libs/lzma/liblzma/common/auto_decoder.c index 6895c7ccf7..2a5c0894d1 100644 --- a/contrib/libs/lzma/liblzma/common/auto_decoder.c +++ b/contrib/libs/lzma/liblzma/common/auto_decoder.c @@ -1,7 +1,7 @@ /////////////////////////////////////////////////////////////////////////////// // /// \file auto_decoder.c -/// \brief Autodetect between .xz Stream and .lzma (LZMA_Alone) formats +/// \brief Autodetect between .xz, .lzma (LZMA_Alone), and .lz (lzip) // // Author: Lasse Collin // @@ -12,10 +12,13 @@ #include "stream_decoder.h" #include "alone_decoder.h" +#ifdef HAVE_LZIP_DECODER +# include "lzip_decoder.h" +#endif typedef struct { - /// Stream decoder or LZMA_Alone decoder + /// .xz Stream decoder, LZMA_Alone decoder, or lzip decoder lzma_next_coder next; uint64_t memlimit; @@ -46,14 +49,22 @@ auto_decode(void *coder_ptr, const lzma_allocator *allocator, // SEQ_CODE even if we return some LZMA_*_CHECK. coder->sequence = SEQ_CODE; - // Detect the file format. For now this is simple, since if - // it doesn't start with 0xFD (the first magic byte of the - // new format), it has to be LZMA_Alone, or something that - // we don't support at all. + // Detect the file format. .xz files start with 0xFD which + // cannot be the first byte of .lzma (LZMA_Alone) format. + // The .lz format starts with 0x4C which could be the + // first byte of a .lzma file but luckily it would mean + // lc/lp/pb being 4/3/1 which liblzma doesn't support because + // lc + lp > 4. So using just 0x4C to detect .lz is OK here. if (in[*in_pos] == 0xFD) { return_if_error(lzma_stream_decoder_init( &coder->next, allocator, coder->memlimit, coder->flags)); +#ifdef HAVE_LZIP_DECODER + } else if (in[*in_pos] == 0x4C) { + return_if_error(lzma_lzip_decoder_init( + &coder->next, allocator, + coder->memlimit, coder->flags)); +#endif } else { return_if_error(lzma_alone_decoder_init(&coder->next, allocator, coder->memlimit, true)); @@ -86,8 +97,8 @@ auto_decode(void *coder_ptr, const lzma_allocator *allocator, // Fall through case SEQ_FINISH: - // When LZMA_DECODE_CONCATENATED was used and we were decoding - // LZMA_Alone file, we need to check check that there is no + // When LZMA_CONCATENATED was used and we were decoding + // a LZMA_Alone file, we need to check that there is no // trailing garbage and wait for LZMA_FINISH. if (*in_pos < in_size) return LZMA_DATA_ERROR; diff --git a/contrib/libs/lzma/liblzma/common/block_buffer_encoder.c b/contrib/libs/lzma/liblzma/common/block_buffer_encoder.c index 39e263aa47..a47342efd0 100644 --- a/contrib/libs/lzma/liblzma/common/block_buffer_encoder.c +++ b/contrib/libs/lzma/liblzma/common/block_buffer_encoder.c @@ -325,6 +325,24 @@ lzma_block_buffer_encode(lzma_block *block, const lzma_allocator *allocator, } +#ifdef HAVE_SYMBOL_VERSIONS_LINUX +// This is for compatibility with binaries linked against liblzma that +// has been patched with xz-5.2.2-compat-libs.patch from RHEL/CentOS 7. +LZMA_SYMVER_API("lzma_block_uncomp_encode@XZ_5.2.2", + lzma_ret, lzma_block_uncomp_encode_522)(lzma_block *block, + const uint8_t *in, size_t in_size, + uint8_t *out, size_t *out_pos, size_t out_size) + lzma_nothrow lzma_attr_warn_unused_result + __attribute__((__alias__("lzma_block_uncomp_encode_52"))); + +LZMA_SYMVER_API("lzma_block_uncomp_encode@@XZ_5.2", + lzma_ret, lzma_block_uncomp_encode_52)(lzma_block *block, + const uint8_t *in, size_t in_size, + uint8_t *out, size_t *out_pos, size_t out_size) + lzma_nothrow lzma_attr_warn_unused_result; + +#define lzma_block_uncomp_encode lzma_block_uncomp_encode_52 +#endif extern LZMA_API(lzma_ret) lzma_block_uncomp_encode(lzma_block *block, const uint8_t *in, size_t in_size, diff --git a/contrib/libs/lzma/liblzma/common/block_decoder.c b/contrib/libs/lzma/liblzma/common/block_decoder.c index 075bd279ff..4827e0f046 100644 --- a/contrib/libs/lzma/liblzma/common/block_decoder.c +++ b/contrib/libs/lzma/liblzma/common/block_decoder.c @@ -40,6 +40,9 @@ typedef struct { /// is unknown. lzma_vli compressed_limit; + /// Maximum allowed Uncompressed Size. + lzma_vli uncompressed_limit; + /// Position when reading the Check field size_t check_pos; @@ -52,21 +55,6 @@ typedef struct { static inline bool -update_size(lzma_vli *size, lzma_vli add, lzma_vli limit) -{ - if (limit > LZMA_VLI_MAX) - limit = LZMA_VLI_MAX; - - if (limit < *size || limit - *size < add) - return true; - - *size += add; - - return false; -} - - -static inline bool is_size_valid(lzma_vli size, lzma_vli reference) { return reference == LZMA_VLI_UNKNOWN || reference == size; @@ -86,21 +74,54 @@ block_decode(void *coder_ptr, const lzma_allocator *allocator, const size_t in_start = *in_pos; const size_t out_start = *out_pos; + // Limit the amount of input and output space that we give + // to the raw decoder based on the information we have + // (or don't have) from Block Header. + const size_t in_stop = *in_pos + (size_t)my_min( + in_size - *in_pos, + coder->compressed_limit - coder->compressed_size); + const size_t out_stop = *out_pos + (size_t)my_min( + out_size - *out_pos, + coder->uncompressed_limit - coder->uncompressed_size); + const lzma_ret ret = coder->next.code(coder->next.coder, - allocator, in, in_pos, in_size, - out, out_pos, out_size, action); + allocator, in, in_pos, in_stop, + out, out_pos, out_stop, action); const size_t in_used = *in_pos - in_start; const size_t out_used = *out_pos - out_start; - // NOTE: We compare to compressed_limit here, which prevents - // the total size of the Block growing past LZMA_VLI_MAX. - if (update_size(&coder->compressed_size, in_used, - coder->compressed_limit) - || update_size(&coder->uncompressed_size, - out_used, - coder->block->uncompressed_size)) - return LZMA_DATA_ERROR; + // Because we have limited the input and output sizes, + // we know that these cannot grow too big or overflow. + coder->compressed_size += in_used; + coder->uncompressed_size += out_used; + + if (ret == LZMA_OK) { + const bool comp_done = coder->compressed_size + == coder->block->compressed_size; + const bool uncomp_done = coder->uncompressed_size + == coder->block->uncompressed_size; + + // If both input and output amounts match the sizes + // in Block Header but we still got LZMA_OK instead + // of LZMA_STREAM_END, the file is broken. + if (comp_done && uncomp_done) + return LZMA_DATA_ERROR; + + // If the decoder has consumed all the input that it + // needs but it still couldn't fill the output buffer + // or return LZMA_STREAM_END, the file is broken. + if (comp_done && *out_pos < out_size) + return LZMA_DATA_ERROR; + + // If the decoder has produced all the output but + // it still didn't return LZMA_STREAM_END or consume + // more input (for example, detecting an end of + // payload marker may need more input but produce + // no output) the file is broken. + if (uncomp_done && *in_pos < in_size) + return LZMA_DATA_ERROR; + } if (!coder->ignore_check) lzma_check_update(&coder->check, coder->block->check, @@ -230,6 +251,14 @@ lzma_block_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, - lzma_check_size(block->check) : block->compressed_size; + // With Uncompressed Size this is simpler. If Block Header lacks + // the size info, then LZMA_VLI_MAX is the maximum possible + // Uncompressed Size. + coder->uncompressed_limit + = block->uncompressed_size == LZMA_VLI_UNKNOWN + ? LZMA_VLI_MAX + : block->uncompressed_size; + // Initialize the check. It's caller's problem if the Check ID is not // supported, and the Block decoder cannot verify the Check field. // Caller can test lzma_check_is_supported(block->check). diff --git a/contrib/libs/lzma/liblzma/common/block_encoder.c b/contrib/libs/lzma/liblzma/common/block_encoder.c index 168846ad68..520ecc5a49 100644 --- a/contrib/libs/lzma/liblzma/common/block_encoder.c +++ b/contrib/libs/lzma/liblzma/common/block_encoder.c @@ -217,6 +217,7 @@ lzma_block_encoder(lzma_stream *strm, lzma_block *block) lzma_next_strm_init(lzma_block_encoder_init, strm, block); strm->internal->supported_actions[LZMA_RUN] = true; + strm->internal->supported_actions[LZMA_SYNC_FLUSH] = true; strm->internal->supported_actions[LZMA_FINISH] = true; return LZMA_OK; diff --git a/contrib/libs/lzma/liblzma/common/block_header_decoder.c b/contrib/libs/lzma/liblzma/common/block_header_decoder.c index 2e1135dd63..c4935dcf46 100644 --- a/contrib/libs/lzma/liblzma/common/block_header_decoder.c +++ b/contrib/libs/lzma/liblzma/common/block_header_decoder.c @@ -14,22 +14,6 @@ #include "check.h" -static void -free_properties(lzma_block *block, const lzma_allocator *allocator) -{ - // Free allocated filter options. The last array member is not - // touched after the initialization in the beginning of - // lzma_block_header_decode(), so we don't need to touch that here. - for (size_t i = 0; i < LZMA_FILTERS_MAX; ++i) { - lzma_free(block->filters[i].options, allocator); - block->filters[i].id = LZMA_VLI_UNKNOWN; - block->filters[i].options = NULL; - } - - return; -} - - extern LZMA_API(lzma_ret) lzma_block_header_decode(lzma_block *block, const lzma_allocator *allocator, const uint8_t *in) @@ -39,6 +23,10 @@ lzma_block_header_decode(lzma_block *block, // are invalid or over 63 bits, or if the header is too small // to contain the claimed information. + // Catch unexpected NULL pointers. + if (block == NULL || block->filters == NULL || in == NULL) + return LZMA_PROG_ERROR; + // Initialize the filter options array. This way the caller can // safely free() the options even if an error occurs in this function. for (size_t i = 0; i <= LZMA_FILTERS_MAX; ++i) { @@ -67,8 +55,11 @@ lzma_block_header_decode(lzma_block *block, const size_t in_size = block->header_size - 4; // Verify CRC32 - if (lzma_crc32(in, in_size, 0) != read32le(in + in_size)) + if (lzma_crc32(in, in_size, 0) != read32le(in + in_size)) { +#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION return LZMA_DATA_ERROR; +#endif + } // Check for unsupported flags. if (in[1] & 0x3C) @@ -104,7 +95,7 @@ lzma_block_header_decode(lzma_block *block, &block->filters[i], allocator, in, &in_pos, in_size); if (ret != LZMA_OK) { - free_properties(block, allocator); + lzma_filters_free(block->filters, allocator); return ret; } } @@ -112,7 +103,7 @@ lzma_block_header_decode(lzma_block *block, // Padding while (in_pos < in_size) { if (in[in_pos++] != 0x00) { - free_properties(block, allocator); + lzma_filters_free(block->filters, allocator); // Possibly some new field present so use // LZMA_OPTIONS_ERROR instead of LZMA_DATA_ERROR. diff --git a/contrib/libs/lzma/liblzma/common/common.c b/contrib/libs/lzma/liblzma/common/common.c index cf714e5e43..a708fdf187 100644 --- a/contrib/libs/lzma/liblzma/common/common.c +++ b/contrib/libs/lzma/liblzma/common/common.c @@ -211,7 +211,6 @@ lzma_code(lzma_stream *strm, lzma_action action) || strm->reserved_ptr2 != NULL || strm->reserved_ptr3 != NULL || strm->reserved_ptr4 != NULL - || strm->reserved_int1 != 0 || strm->reserved_int2 != 0 || strm->reserved_int3 != 0 || strm->reserved_int4 != 0 @@ -299,9 +298,7 @@ lzma_code(lzma_stream *strm, lzma_action action) strm->internal->avail_in = strm->avail_in; - // Cast is needed to silence a warning about LZMA_TIMED_OUT, which - // isn't part of lzma_ret enumeration. - switch ((unsigned int)(ret)) { + switch (ret) { case LZMA_OK: // Don't return LZMA_BUF_ERROR when it happens the first time. // This is to avoid returning LZMA_BUF_ERROR when avail_out @@ -322,6 +319,17 @@ lzma_code(lzma_stream *strm, lzma_action action) ret = LZMA_OK; break; + case LZMA_SEEK_NEEDED: + strm->internal->allow_buf_error = false; + + // If LZMA_FINISH was used, reset it back to the + // LZMA_RUN-based state so that new input can be supplied + // by the application. + if (strm->internal->sequence == ISEQ_FINISH) + strm->internal->sequence = ISEQ_RUN; + + break; + case LZMA_STREAM_END: if (strm->internal->sequence == ISEQ_SYNC_FLUSH || strm->internal->sequence == ISEQ_FULL_FLUSH @@ -366,6 +374,20 @@ lzma_end(lzma_stream *strm) } +#ifdef HAVE_SYMBOL_VERSIONS_LINUX +// This is for compatibility with binaries linked against liblzma that +// has been patched with xz-5.2.2-compat-libs.patch from RHEL/CentOS 7. +LZMA_SYMVER_API("lzma_get_progress@XZ_5.2.2", + void, lzma_get_progress_522)(lzma_stream *strm, + uint64_t *progress_in, uint64_t *progress_out) lzma_nothrow + __attribute__((__alias__("lzma_get_progress_52"))); + +LZMA_SYMVER_API("lzma_get_progress@@XZ_5.2", + void, lzma_get_progress_52)(lzma_stream *strm, + uint64_t *progress_in, uint64_t *progress_out) lzma_nothrow; + +#define lzma_get_progress lzma_get_progress_52 +#endif extern LZMA_API(void) lzma_get_progress(lzma_stream *strm, uint64_t *progress_in, uint64_t *progress_out) diff --git a/contrib/libs/lzma/liblzma/common/common.h b/contrib/libs/lzma/liblzma/common/common.h index b3d3b7a059..11fec52c59 100644 --- a/contrib/libs/lzma/liblzma/common/common.h +++ b/contrib/libs/lzma/liblzma/common/common.h @@ -34,6 +34,59 @@ #include "lzma.h" +// This is for detecting modern GCC and Clang attributes +// like __symver__ in GCC >= 10. +#ifdef __has_attribute +# define lzma_has_attribute(attr) __has_attribute(attr) +#else +# define lzma_has_attribute(attr) 0 +#endif + +// The extra symbol versioning in the C files may only be used when +// building a shared library. If HAVE_SYMBOL_VERSIONS_LINUX is defined +// to 2 then symbol versioning is done only if also PIC is defined. +// By default Libtool defines PIC when building a shared library and +// doesn't define it when building a static library but it can be +// overriden with --with-pic and --without-pic. configure let's rely +// on PIC if neither --with-pic or --without-pic was used. +#if defined(HAVE_SYMBOL_VERSIONS_LINUX) \ + && (HAVE_SYMBOL_VERSIONS_LINUX == 2 && !defined(PIC)) +# undef HAVE_SYMBOL_VERSIONS_LINUX +#endif + +#ifdef HAVE_SYMBOL_VERSIONS_LINUX +// To keep link-time optimization (LTO, -flto) working with GCC, +// the __symver__ attribute must be used instead of __asm__(".symver ..."). +// Otherwise the symbol versions may be lost, resulting in broken liblzma +// that has wrong default versions in the exported symbol list! +// The attribute was added in GCC 10; LTO with older GCC is not supported. +// +// To keep -Wmissing-prototypes happy, use LZMA_SYMVER_API only with function +// declarations (including those with __alias__ attribute) and LZMA_API with +// the function definitions. This means a little bit of silly copy-and-paste +// between declarations and definitions though. +// +// As of GCC 12.2, the __symver__ attribute supports only @ and @@ but the +// very convenient @@@ isn't supported (it's supported by GNU assembler +// since 2000). When using @@ instead of @@@, the internal name must not be +// the same as the external name to avoid problems in some situations. This +// is why "#define foo_52 foo" is needed for the default symbol versions. +// +// __has_attribute is supported before GCC 10 and it is supported in Clang 14 +// too (which doesn't support __symver__) so use it to detect if __symver__ +// is available. This should be far more reliable than looking at compiler +// version macros as nowadays especially __GNUC__ is defined by many compilers. +# if lzma_has_attribute(__symver__) +# define LZMA_SYMVER_API(extnamever, type, intname) \ + extern __attribute__((__symver__(extnamever))) \ + LZMA_API(type) intname +# else +# define LZMA_SYMVER_API(extnamever, type, intname) \ + __asm__(".symver " #intname "," extnamever); \ + extern LZMA_API(type) intname +# endif +#endif + // These allow helping the compiler in some often-executed branches, whose // result is almost always the same. #ifdef __GNUC__ @@ -67,14 +120,15 @@ #define LZMA_FILTER_RESERVED_START (LZMA_VLI_C(1) << 62) -/// Supported flags that can be passed to lzma_stream_decoder() -/// or lzma_auto_decoder(). +/// Supported flags that can be passed to lzma_stream_decoder(), +/// lzma_auto_decoder(), or lzma_stream_decoder_mt(). #define LZMA_SUPPORTED_FLAGS \ ( LZMA_TELL_NO_CHECK \ | LZMA_TELL_UNSUPPORTED_CHECK \ | LZMA_TELL_ANY_CHECK \ | LZMA_IGNORE_CHECK \ - | LZMA_CONCATENATED ) + | LZMA_CONCATENATED \ + | LZMA_FAIL_FAST ) /// Largest valid lzma_action value as unsigned integer. @@ -83,9 +137,12 @@ /// Special return value (lzma_ret) to indicate that a timeout was reached /// and lzma_code() must not return LZMA_BUF_ERROR. This is converted to -/// LZMA_OK in lzma_code(). This is not in the lzma_ret enumeration because -/// there's no need to have it in the public API. -#define LZMA_TIMED_OUT 32 +/// LZMA_OK in lzma_code(). +#define LZMA_TIMED_OUT LZMA_RET_INTERNAL1 + +/// Special return value (lzma_ret) for use in stream_decoder_mt.c to +/// indicate Index was detected instead of a Block Header. +#define LZMA_INDEX_DETECTED LZMA_RET_INTERNAL2 typedef struct lzma_next_coder_s lzma_next_coder; @@ -118,8 +175,11 @@ typedef void (*lzma_end_function)( /// an array of lzma_filter_info structures. This array is used with /// lzma_next_filter_init to initialize the filter chain. struct lzma_filter_info_s { - /// Filter ID. This is used only by the encoder - /// with lzma_filters_update(). + /// Filter ID. This can be used to share the same initiazation + /// function *and* data structures with different Filter IDs + /// (LZMA_FILTER_LZMA1EXT does it), and also by the encoder + /// with lzma_filters_update() if filter chain is updated + /// in the middle of a raw stream or Block (LZMA_SYNC_FLUSH). lzma_vli id; /// Pointer to function used to initialize the filter. @@ -173,6 +233,16 @@ struct lzma_next_coder_s { lzma_ret (*update)(void *coder, const lzma_allocator *allocator, const lzma_filter *filters, const lzma_filter *reversed_filters); + + /// Set how many bytes of output this coder may produce at maximum. + /// On success LZMA_OK must be returned. + /// If the filter chain as a whole cannot support this feature, + /// this must return LZMA_OPTIONS_ERROR. + /// If no input has been given to the coder and the requested limit + /// is too small, this must return LZMA_BUF_ERROR. If input has been + /// seen, LZMA_OK is allowed too. + lzma_ret (*set_out_limit)(void *coder, uint64_t *uncomp_size, + uint64_t out_limit); }; @@ -188,6 +258,7 @@ struct lzma_next_coder_s { .get_check = NULL, \ .memconfig = NULL, \ .update = NULL, \ + .set_out_limit = NULL, \ } diff --git a/contrib/libs/lzma/liblzma/common/file_info.c b/contrib/libs/lzma/liblzma/common/file_info.c new file mode 100644 index 0000000000..a6b7e145ae --- /dev/null +++ b/contrib/libs/lzma/liblzma/common/file_info.c @@ -0,0 +1,855 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file file_info.c +/// \brief Decode .xz file information into a lzma_index structure +// +// Author: Lasse Collin +// +// This file has been put into the public domain. +// You can do whatever you want with this file. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "index_decoder.h" + + +typedef struct { + enum { + SEQ_MAGIC_BYTES, + SEQ_PADDING_SEEK, + SEQ_PADDING_DECODE, + SEQ_FOOTER, + SEQ_INDEX_INIT, + SEQ_INDEX_DECODE, + SEQ_HEADER_DECODE, + SEQ_HEADER_COMPARE, + } sequence; + + /// Absolute position of in[*in_pos] in the file. All code that + /// modifies *in_pos also updates this. seek_to_pos() needs this + /// to determine if we need to request the application to seek for + /// us or if we can do the seeking internally by adjusting *in_pos. + uint64_t file_cur_pos; + + /// This refers to absolute positions of interesting parts of the + /// input file. Sometimes it points to the *beginning* of a specific + /// field and sometimes to the *end* of a field. The current target + /// position at each moment is explained in the comments. + uint64_t file_target_pos; + + /// Size of the .xz file (from the application). + uint64_t file_size; + + /// Index decoder + lzma_next_coder index_decoder; + + /// Number of bytes remaining in the Index field that is currently + /// being decoded. + lzma_vli index_remaining; + + /// The Index decoder will store the decoded Index in this pointer. + lzma_index *this_index; + + /// Amount of Stream Padding in the current Stream. + lzma_vli stream_padding; + + /// The final combined index is collected here. + lzma_index *combined_index; + + /// Pointer from the application where to store the index information + /// after successful decoding. + lzma_index **dest_index; + + /// Pointer to lzma_stream.seek_pos to be used when returning + /// LZMA_SEEK_NEEDED. This is set by seek_to_pos() when needed. + uint64_t *external_seek_pos; + + /// Memory usage limit + uint64_t memlimit; + + /// Stream Flags from the very beginning of the file. + lzma_stream_flags first_header_flags; + + /// Stream Flags from Stream Header of the current Stream. + lzma_stream_flags header_flags; + + /// Stream Flags from Stream Footer of the current Stream. + lzma_stream_flags footer_flags; + + size_t temp_pos; + size_t temp_size; + uint8_t temp[8192]; + +} lzma_file_info_coder; + + +/// Copies data from in[*in_pos] into coder->temp until +/// coder->temp_pos == coder->temp_size. This also keeps coder->file_cur_pos +/// in sync with *in_pos. Returns true if more input is needed. +static bool +fill_temp(lzma_file_info_coder *coder, const uint8_t *restrict in, + size_t *restrict in_pos, size_t in_size) +{ + coder->file_cur_pos += lzma_bufcpy(in, in_pos, in_size, + coder->temp, &coder->temp_pos, coder->temp_size); + return coder->temp_pos < coder->temp_size; +} + + +/// Seeks to the absolute file position specified by target_pos. +/// This tries to do the seeking by only modifying *in_pos, if possible. +/// The main benefit of this is that if one passes the whole file at once +/// to lzma_code(), the decoder will never need to return LZMA_SEEK_NEEDED +/// as all the seeking can be done by adjusting *in_pos in this function. +/// +/// Returns true if an external seek is needed and the caller must return +/// LZMA_SEEK_NEEDED. +static bool +seek_to_pos(lzma_file_info_coder *coder, uint64_t target_pos, + size_t in_start, size_t *in_pos, size_t in_size) +{ + // The input buffer doesn't extend beyond the end of the file. + // This has been checked by file_info_decode() already. + assert(coder->file_size - coder->file_cur_pos >= in_size - *in_pos); + + const uint64_t pos_min = coder->file_cur_pos - (*in_pos - in_start); + const uint64_t pos_max = coder->file_cur_pos + (in_size - *in_pos); + + bool external_seek_needed; + + if (target_pos >= pos_min && target_pos <= pos_max) { + // The requested position is available in the current input + // buffer or right after it. That is, in a corner case we + // end up setting *in_pos == in_size and thus will immediately + // need new input bytes from the application. + *in_pos += (size_t)(target_pos - coder->file_cur_pos); + external_seek_needed = false; + } else { + // Ask the application to seek the input file. + *coder->external_seek_pos = target_pos; + external_seek_needed = true; + + // Mark the whole input buffer as used. This way + // lzma_stream.total_in will have a better estimate + // of the amount of data read. It still won't be perfect + // as the value will depend on the input buffer size that + // the application uses, but it should be good enough for + // those few who want an estimate. + *in_pos = in_size; + } + + // After seeking (internal or external) the current position + // will match the requested target position. + coder->file_cur_pos = target_pos; + + return external_seek_needed; +} + + +/// The caller sets coder->file_target_pos so that it points to the *end* +/// of the desired file position. This function then determines how far +/// backwards from that position we can seek. After seeking fill_temp() +/// can be used to read data into coder->temp. When fill_temp() has finished, +/// coder->temp[coder->temp_size] will match coder->file_target_pos. +/// +/// This also validates that coder->target_file_pos is sane in sense that +/// we aren't trying to seek too far backwards (too close or beyond the +/// beginning of the file). +static lzma_ret +reverse_seek(lzma_file_info_coder *coder, + size_t in_start, size_t *in_pos, size_t in_size) +{ + // Check that there is enough data before the target position + // to contain at least Stream Header and Stream Footer. If there + // isn't, the file cannot be valid. + if (coder->file_target_pos < 2 * LZMA_STREAM_HEADER_SIZE) + return LZMA_DATA_ERROR; + + coder->temp_pos = 0; + + // The Stream Header at the very beginning of the file gets handled + // specially in SEQ_MAGIC_BYTES and thus we will never need to seek + // there. By not seeking to the first LZMA_STREAM_HEADER_SIZE bytes + // we avoid a useless external seek after SEQ_MAGIC_BYTES if the + // application uses an extremely small input buffer and the input + // file is very small. + if (coder->file_target_pos - LZMA_STREAM_HEADER_SIZE + < sizeof(coder->temp)) + coder->temp_size = (size_t)(coder->file_target_pos + - LZMA_STREAM_HEADER_SIZE); + else + coder->temp_size = sizeof(coder->temp); + + // The above if-statements guarantee this. This is important because + // the Stream Header/Footer decoders assume that there's at least + // LZMA_STREAM_HEADER_SIZE bytes in coder->temp. + assert(coder->temp_size >= LZMA_STREAM_HEADER_SIZE); + + if (seek_to_pos(coder, coder->file_target_pos - coder->temp_size, + in_start, in_pos, in_size)) + return LZMA_SEEK_NEEDED; + + return LZMA_OK; +} + + +/// Gets the number of zero-bytes at the end of the buffer. +static size_t +get_padding_size(const uint8_t *buf, size_t buf_size) +{ + size_t padding = 0; + while (buf_size > 0 && buf[--buf_size] == 0x00) + ++padding; + + return padding; +} + + +/// With the Stream Header at the very beginning of the file, LZMA_FORMAT_ERROR +/// is used to tell the application that Magic Bytes didn't match. In other +/// Stream Header/Footer fields (in the middle/end of the file) it could be +/// a bit confusing to return LZMA_FORMAT_ERROR as we already know that there +/// is a valid Stream Header at the beginning of the file. For those cases +/// this function is used to convert LZMA_FORMAT_ERROR to LZMA_DATA_ERROR. +static lzma_ret +hide_format_error(lzma_ret ret) +{ + if (ret == LZMA_FORMAT_ERROR) + ret = LZMA_DATA_ERROR; + + return ret; +} + + +/// Calls the Index decoder and updates coder->index_remaining. +/// This is a separate function because the input can be either directly +/// from the application or from coder->temp. +static lzma_ret +decode_index(lzma_file_info_coder *coder, const lzma_allocator *allocator, + const uint8_t *restrict in, size_t *restrict in_pos, + size_t in_size, bool update_file_cur_pos) +{ + const size_t in_start = *in_pos; + + const lzma_ret ret = coder->index_decoder.code( + coder->index_decoder.coder, + allocator, in, in_pos, in_size, + NULL, NULL, 0, LZMA_RUN); + + coder->index_remaining -= *in_pos - in_start; + + if (update_file_cur_pos) + coder->file_cur_pos += *in_pos - in_start; + + return ret; +} + + +static lzma_ret +file_info_decode(void *coder_ptr, const lzma_allocator *allocator, + const uint8_t *restrict in, size_t *restrict in_pos, + size_t in_size, + uint8_t *restrict out lzma_attribute((__unused__)), + size_t *restrict out_pos lzma_attribute((__unused__)), + size_t out_size lzma_attribute((__unused__)), + lzma_action action lzma_attribute((__unused__))) +{ + lzma_file_info_coder *coder = coder_ptr; + const size_t in_start = *in_pos; + + // If the caller provides input past the end of the file, trim + // the extra bytes from the buffer so that we won't read too far. + assert(coder->file_size >= coder->file_cur_pos); + if (coder->file_size - coder->file_cur_pos < in_size - in_start) + in_size = in_start + + (size_t)(coder->file_size - coder->file_cur_pos); + + while (true) + switch (coder->sequence) { + case SEQ_MAGIC_BYTES: + // Decode the Stream Header at the beginning of the file + // first to check if the Magic Bytes match. The flags + // are stored in coder->first_header_flags so that we + // don't need to seek to it again. + // + // Check that the file is big enough to contain at least + // Stream Header. + if (coder->file_size < LZMA_STREAM_HEADER_SIZE) + return LZMA_FORMAT_ERROR; + + // Read the Stream Header field into coder->temp. + if (fill_temp(coder, in, in_pos, in_size)) + return LZMA_OK; + + // This is the only Stream Header/Footer decoding where we + // want to return LZMA_FORMAT_ERROR if the Magic Bytes don't + // match. Elsewhere it will be converted to LZMA_DATA_ERROR. + return_if_error(lzma_stream_header_decode( + &coder->first_header_flags, coder->temp)); + + // Now that we know that the Magic Bytes match, check the + // file size. It's better to do this here after checking the + // Magic Bytes since this way we can give LZMA_FORMAT_ERROR + // instead of LZMA_DATA_ERROR when the Magic Bytes don't + // match in a file that is too big or isn't a multiple of + // four bytes. + if (coder->file_size > LZMA_VLI_MAX || (coder->file_size & 3)) + return LZMA_DATA_ERROR; + + // Start looking for Stream Padding and Stream Footer + // at the end of the file. + coder->file_target_pos = coder->file_size; + + // Fall through + + case SEQ_PADDING_SEEK: + coder->sequence = SEQ_PADDING_DECODE; + return_if_error(reverse_seek( + coder, in_start, in_pos, in_size)); + + // Fall through + + case SEQ_PADDING_DECODE: { + // Copy to coder->temp first. This keeps the code simpler if + // the application only provides input a few bytes at a time. + if (fill_temp(coder, in, in_pos, in_size)) + return LZMA_OK; + + // Scan the buffer backwards to get the size of the + // Stream Padding field (if any). + const size_t new_padding = get_padding_size( + coder->temp, coder->temp_size); + coder->stream_padding += new_padding; + + // Set the target position to the beginning of Stream Padding + // that has been observed so far. If all Stream Padding has + // been seen, then the target position will be at the end + // of the Stream Footer field. + coder->file_target_pos -= new_padding; + + if (new_padding == coder->temp_size) { + // The whole buffer was padding. Seek backwards in + // the file to get more input. + coder->sequence = SEQ_PADDING_SEEK; + break; + } + + // Size of Stream Padding must be a multiple of 4 bytes. + if (coder->stream_padding & 3) + return LZMA_DATA_ERROR; + + coder->sequence = SEQ_FOOTER; + + // Calculate the amount of non-padding data in coder->temp. + coder->temp_size -= new_padding; + coder->temp_pos = coder->temp_size; + + // We can avoid an external seek if the whole Stream Footer + // is already in coder->temp. In that case SEQ_FOOTER won't + // read more input and will find the Stream Footer from + // coder->temp[coder->temp_size - LZMA_STREAM_HEADER_SIZE]. + // + // Otherwise we will need to seek. The seeking is done so + // that Stream Footer wil be at the end of coder->temp. + // This way it's likely that we also get a complete Index + // field into coder->temp without needing a separate seek + // for that (unless the Index field is big). + if (coder->temp_size < LZMA_STREAM_HEADER_SIZE) + return_if_error(reverse_seek( + coder, in_start, in_pos, in_size)); + } + + // Fall through + + case SEQ_FOOTER: + // Copy the Stream Footer field into coder->temp. + // If Stream Footer was already available in coder->temp + // in SEQ_PADDING_DECODE, then this does nothing. + if (fill_temp(coder, in, in_pos, in_size)) + return LZMA_OK; + + // Make coder->file_target_pos and coder->temp_size point + // to the beginning of Stream Footer and thus to the end + // of the Index field. coder->temp_pos will be updated + // a bit later. + coder->file_target_pos -= LZMA_STREAM_HEADER_SIZE; + coder->temp_size -= LZMA_STREAM_HEADER_SIZE; + + // Decode Stream Footer. + return_if_error(hide_format_error(lzma_stream_footer_decode( + &coder->footer_flags, + coder->temp + coder->temp_size))); + + // Check that we won't seek past the beginning of the file. + // + // LZMA_STREAM_HEADER_SIZE is added because there must be + // space for Stream Header too even though we won't seek + // there before decoding the Index field. + // + // There's no risk of integer overflow here because + // Backward Size cannot be greater than 2^34. + if (coder->file_target_pos < coder->footer_flags.backward_size + + LZMA_STREAM_HEADER_SIZE) + return LZMA_DATA_ERROR; + + // Set the target position to the beginning of the Index field. + coder->file_target_pos -= coder->footer_flags.backward_size; + coder->sequence = SEQ_INDEX_INIT; + + // We can avoid an external seek if the whole Index field is + // already available in coder->temp. + if (coder->temp_size >= coder->footer_flags.backward_size) { + // Set coder->temp_pos to point to the beginning + // of the Index. + coder->temp_pos = coder->temp_size + - coder->footer_flags.backward_size; + } else { + // These are set to zero to indicate that there's no + // useful data (Index or anything else) in coder->temp. + coder->temp_pos = 0; + coder->temp_size = 0; + + // Seek to the beginning of the Index field. + if (seek_to_pos(coder, coder->file_target_pos, + in_start, in_pos, in_size)) + return LZMA_SEEK_NEEDED; + } + + // Fall through + + case SEQ_INDEX_INIT: { + // Calculate the amount of memory already used by the earlier + // Indexes so that we know how big memory limit to pass to + // the Index decoder. + // + // NOTE: When there are multiple Streams, the separate + // lzma_index structures can use more RAM (as measured by + // lzma_index_memused()) than the final combined lzma_index. + // Thus memlimit may need to be slightly higher than the final + // calculated memory usage will be. This is perhaps a bit + // confusing to the application, but I think it shouldn't + // cause problems in practice. + uint64_t memused = 0; + if (coder->combined_index != NULL) { + memused = lzma_index_memused(coder->combined_index); + assert(memused <= coder->memlimit); + if (memused > coder->memlimit) // Extra sanity check + return LZMA_PROG_ERROR; + } + + // Initialize the Index decoder. + return_if_error(lzma_index_decoder_init( + &coder->index_decoder, allocator, + &coder->this_index, + coder->memlimit - memused)); + + coder->index_remaining = coder->footer_flags.backward_size; + coder->sequence = SEQ_INDEX_DECODE; + } + + // Fall through + + case SEQ_INDEX_DECODE: { + // Decode (a part of) the Index. If the whole Index is already + // in coder->temp, read it from there. Otherwise read from + // in[*in_pos] onwards. Note that index_decode() updates + // coder->index_remaining and optionally coder->file_cur_pos. + lzma_ret ret; + if (coder->temp_size != 0) { + assert(coder->temp_size - coder->temp_pos + == coder->index_remaining); + ret = decode_index(coder, allocator, coder->temp, + &coder->temp_pos, coder->temp_size, + false); + } else { + // Don't give the decoder more input than the known + // remaining size of the Index field. + size_t in_stop = in_size; + if (in_size - *in_pos > coder->index_remaining) + in_stop = *in_pos + + (size_t)(coder->index_remaining); + + ret = decode_index(coder, allocator, + in, in_pos, in_stop, true); + } + + switch (ret) { + case LZMA_OK: + // If the Index docoder asks for more input when we + // have already given it as much input as Backward Size + // indicated, the file is invalid. + if (coder->index_remaining == 0) + return LZMA_DATA_ERROR; + + // We cannot get here if we were reading Index from + // coder->temp because when reading from coder->temp + // we give the Index decoder exactly + // coder->index_remaining bytes of input. + assert(coder->temp_size == 0); + + return LZMA_OK; + + case LZMA_STREAM_END: + // If the decoding seems to be successful, check also + // that the Index decoder consumed as much input as + // indicated by the Backward Size field. + if (coder->index_remaining != 0) + return LZMA_DATA_ERROR; + + break; + + default: + return ret; + } + + // Calculate how much the Index tells us to seek backwards + // (relative to the beginning of the Index): Total size of + // all Blocks plus the size of the Stream Header field. + // No integer overflow here because lzma_index_total_size() + // cannot return a value greater than LZMA_VLI_MAX. + const uint64_t seek_amount + = lzma_index_total_size(coder->this_index) + + LZMA_STREAM_HEADER_SIZE; + + // Check that Index is sane in sense that seek_amount won't + // make us seek past the beginning of the file when locating + // the Stream Header. + // + // coder->file_target_pos still points to the beginning of + // the Index field. + if (coder->file_target_pos < seek_amount) + return LZMA_DATA_ERROR; + + // Set the target to the beginning of Stream Header. + coder->file_target_pos -= seek_amount; + + if (coder->file_target_pos == 0) { + // We would seek to the beginning of the file, but + // since we already decoded that Stream Header in + // SEQ_MAGIC_BYTES, we can use the cached value from + // coder->first_header_flags to avoid the seek. + coder->header_flags = coder->first_header_flags; + coder->sequence = SEQ_HEADER_COMPARE; + break; + } + + coder->sequence = SEQ_HEADER_DECODE; + + // Make coder->file_target_pos point to the end of + // the Stream Header field. + coder->file_target_pos += LZMA_STREAM_HEADER_SIZE; + + // If coder->temp_size is non-zero, it points to the end + // of the Index field. Then the beginning of the Index + // field is at coder->temp[coder->temp_size + // - coder->footer_flags.backward_size]. + assert(coder->temp_size == 0 || coder->temp_size + >= coder->footer_flags.backward_size); + + // If coder->temp contained the whole Index, see if it has + // enough data to contain also the Stream Header. If so, + // we avoid an external seek. + // + // NOTE: This can happen only with small .xz files and only + // for the non-first Stream as the Stream Flags of the first + // Stream are cached and already handled a few lines above. + // So this isn't as useful as the other seek-avoidance cases. + if (coder->temp_size != 0 && coder->temp_size + - coder->footer_flags.backward_size + >= seek_amount) { + // Make temp_pos and temp_size point to the *end* of + // Stream Header so that SEQ_HEADER_DECODE will find + // the start of Stream Header from coder->temp[ + // coder->temp_size - LZMA_STREAM_HEADER_SIZE]. + coder->temp_pos = coder->temp_size + - coder->footer_flags.backward_size + - seek_amount + + LZMA_STREAM_HEADER_SIZE; + coder->temp_size = coder->temp_pos; + } else { + // Seek so that Stream Header will be at the end of + // coder->temp. With typical multi-Stream files we + // will usually also get the Stream Footer and Index + // of the *previous* Stream in coder->temp and thus + // won't need a separate seek for them. + return_if_error(reverse_seek(coder, + in_start, in_pos, in_size)); + } + } + + // Fall through + + case SEQ_HEADER_DECODE: + // Copy the Stream Header field into coder->temp. + // If Stream Header was already available in coder->temp + // in SEQ_INDEX_DECODE, then this does nothing. + if (fill_temp(coder, in, in_pos, in_size)) + return LZMA_OK; + + // Make all these point to the beginning of Stream Header. + coder->file_target_pos -= LZMA_STREAM_HEADER_SIZE; + coder->temp_size -= LZMA_STREAM_HEADER_SIZE; + coder->temp_pos = coder->temp_size; + + // Decode the Stream Header. + return_if_error(hide_format_error(lzma_stream_header_decode( + &coder->header_flags, + coder->temp + coder->temp_size))); + + coder->sequence = SEQ_HEADER_COMPARE; + + // Fall through + + case SEQ_HEADER_COMPARE: + // Compare Stream Header against Stream Footer. They must + // match. + return_if_error(lzma_stream_flags_compare( + &coder->header_flags, &coder->footer_flags)); + + // Store the decoded Stream Flags into the Index. Use the + // Footer Flags because it contains Backward Size, although + // it shouldn't matter in practice. + if (lzma_index_stream_flags(coder->this_index, + &coder->footer_flags) != LZMA_OK) + return LZMA_PROG_ERROR; + + // Store also the size of the Stream Padding field. It is + // needed to calculate the offsets of the Streams correctly. + if (lzma_index_stream_padding(coder->this_index, + coder->stream_padding) != LZMA_OK) + return LZMA_PROG_ERROR; + + // Reset it so that it's ready for the next Stream. + coder->stream_padding = 0; + + // Append the earlier decoded Indexes after this_index. + if (coder->combined_index != NULL) + return_if_error(lzma_index_cat(coder->this_index, + coder->combined_index, allocator)); + + coder->combined_index = coder->this_index; + coder->this_index = NULL; + + // If the whole file was decoded, tell the caller that we + // are finished. + if (coder->file_target_pos == 0) { + // The combined index must indicate the same file + // size as was told to us at initialization. + assert(lzma_index_file_size(coder->combined_index) + == coder->file_size); + + // Make the combined index available to + // the application. + *coder->dest_index = coder->combined_index; + coder->combined_index = NULL; + + // Mark the input buffer as used since we may have + // done internal seeking and thus don't know how + // many input bytes were actually used. This way + // lzma_stream.total_in gets a slightly better + // estimate of the amount of input used. + *in_pos = in_size; + return LZMA_STREAM_END; + } + + // We didn't hit the beginning of the file yet, so continue + // reading backwards in the file. If we have unprocessed + // data in coder->temp, use it before requesting more data + // from the application. + // + // coder->file_target_pos, coder->temp_size, and + // coder->temp_pos all point to the beginning of Stream Header + // and thus the end of the previous Stream in the file. + coder->sequence = coder->temp_size > 0 + ? SEQ_PADDING_DECODE : SEQ_PADDING_SEEK; + break; + + default: + assert(0); + return LZMA_PROG_ERROR; + } +} + + +static lzma_ret +file_info_decoder_memconfig(void *coder_ptr, uint64_t *memusage, + uint64_t *old_memlimit, uint64_t new_memlimit) +{ + lzma_file_info_coder *coder = coder_ptr; + + // The memory usage calculation comes from three things: + // + // (1) The Indexes that have already been decoded and processed into + // coder->combined_index. + // + // (2) The latest Index in coder->this_index that has been decoded but + // not yet put into coder->combined_index. + // + // (3) The latest Index that we have started decoding but haven't + // finished and thus isn't available in coder->this_index yet. + // Memory usage and limit information needs to be communicated + // from/to coder->index_decoder. + // + // Care has to be taken to not do both (2) and (3) when calculating + // the memory usage. + uint64_t combined_index_memusage = 0; + uint64_t this_index_memusage = 0; + + // (1) If we have already successfully decoded one or more Indexes, + // get their memory usage. + if (coder->combined_index != NULL) + combined_index_memusage = lzma_index_memused( + coder->combined_index); + + // Choose between (2), (3), or neither. + if (coder->this_index != NULL) { + // (2) The latest Index is available. Use its memory usage. + this_index_memusage = lzma_index_memused(coder->this_index); + + } else if (coder->sequence == SEQ_INDEX_DECODE) { + // (3) The Index decoder is activate and hasn't yet stored + // the new index in coder->this_index. Get the memory usage + // information from the Index decoder. + // + // NOTE: If the Index decoder doesn't yet know how much memory + // it will eventually need, it will return a tiny value here. + uint64_t dummy; + if (coder->index_decoder.memconfig(coder->index_decoder.coder, + &this_index_memusage, &dummy, 0) + != LZMA_OK) { + assert(0); + return LZMA_PROG_ERROR; + } + } + + // Now we know the total memory usage/requirement. If we had neither + // old Indexes nor a new Index, this will be zero which isn't + // acceptable as lzma_memusage() has to return non-zero on success + // and even with an empty .xz file we will end up with a lzma_index + // that takes some memory. + *memusage = combined_index_memusage + this_index_memusage; + if (*memusage == 0) + *memusage = lzma_index_memusage(1, 0); + + *old_memlimit = coder->memlimit; + + // If requested, set a new memory usage limit. + if (new_memlimit != 0) { + if (new_memlimit < *memusage) + return LZMA_MEMLIMIT_ERROR; + + // In the condition (3) we need to tell the Index decoder + // its new memory usage limit. + if (coder->this_index == NULL + && coder->sequence == SEQ_INDEX_DECODE) { + const uint64_t idec_new_memlimit = new_memlimit + - combined_index_memusage; + + assert(this_index_memusage > 0); + assert(idec_new_memlimit > 0); + + uint64_t dummy1; + uint64_t dummy2; + + if (coder->index_decoder.memconfig( + coder->index_decoder.coder, + &dummy1, &dummy2, idec_new_memlimit) + != LZMA_OK) { + assert(0); + return LZMA_PROG_ERROR; + } + } + + coder->memlimit = new_memlimit; + } + + return LZMA_OK; +} + + +static void +file_info_decoder_end(void *coder_ptr, const lzma_allocator *allocator) +{ + lzma_file_info_coder *coder = coder_ptr; + + lzma_next_end(&coder->index_decoder, allocator); + lzma_index_end(coder->this_index, allocator); + lzma_index_end(coder->combined_index, allocator); + + lzma_free(coder, allocator); + return; +} + + +static lzma_ret +lzma_file_info_decoder_init(lzma_next_coder *next, + const lzma_allocator *allocator, uint64_t *seek_pos, + lzma_index **dest_index, + uint64_t memlimit, uint64_t file_size) +{ + lzma_next_coder_init(&lzma_file_info_decoder_init, next, allocator); + + if (dest_index == NULL) + return LZMA_PROG_ERROR; + + lzma_file_info_coder *coder = next->coder; + if (coder == NULL) { + coder = lzma_alloc(sizeof(lzma_file_info_coder), allocator); + if (coder == NULL) + return LZMA_MEM_ERROR; + + next->coder = coder; + next->code = &file_info_decode; + next->end = &file_info_decoder_end; + next->memconfig = &file_info_decoder_memconfig; + + coder->index_decoder = LZMA_NEXT_CODER_INIT; + coder->this_index = NULL; + coder->combined_index = NULL; + } + + coder->sequence = SEQ_MAGIC_BYTES; + coder->file_cur_pos = 0; + coder->file_target_pos = 0; + coder->file_size = file_size; + + lzma_index_end(coder->this_index, allocator); + coder->this_index = NULL; + + lzma_index_end(coder->combined_index, allocator); + coder->combined_index = NULL; + + coder->stream_padding = 0; + + coder->dest_index = dest_index; + coder->external_seek_pos = seek_pos; + + // If memlimit is 0, make it 1 to ensure that lzma_memlimit_get() + // won't return 0 (which would indicate an error). + coder->memlimit = my_max(1, memlimit); + + // Prepare these for reading the first Stream Header into coder->temp. + coder->temp_pos = 0; + coder->temp_size = LZMA_STREAM_HEADER_SIZE; + + return LZMA_OK; +} + + +extern LZMA_API(lzma_ret) +lzma_file_info_decoder(lzma_stream *strm, lzma_index **dest_index, + uint64_t memlimit, uint64_t file_size) +{ + lzma_next_strm_init(lzma_file_info_decoder_init, strm, &strm->seek_pos, + dest_index, memlimit, file_size); + + // We allow LZMA_FINISH in addition to LZMA_RUN for convenience. + // lzma_code() is able to handle the LZMA_FINISH + LZMA_SEEK_NEEDED + // combination in a sane way. Applications still need to be careful + // if they use LZMA_FINISH so that they remember to reset it back + // to LZMA_RUN after seeking if needed. + strm->internal->supported_actions[LZMA_RUN] = true; + strm->internal->supported_actions[LZMA_FINISH] = true; + + return LZMA_OK; +} diff --git a/contrib/libs/lzma/liblzma/common/filter_common.c b/contrib/libs/lzma/liblzma/common/filter_common.c index 9ad5d5d8e2..fa0927cf9b 100644 --- a/contrib/libs/lzma/liblzma/common/filter_common.c +++ b/contrib/libs/lzma/liblzma/common/filter_common.c @@ -42,6 +42,13 @@ static const struct { .last_ok = true, .changes_size = true, }, + { + .id = LZMA_FILTER_LZMA1EXT, + .options_size = sizeof(lzma_options_lzma), + .non_last_ok = false, + .last_ok = true, + .changes_size = true, + }, #endif #if defined(HAVE_ENCODER_LZMA2) || defined(HAVE_DECODER_LZMA2) { @@ -97,6 +104,15 @@ static const struct { .changes_size = false, }, #endif +#if defined(HAVE_ENCODER_ARM64) || defined(HAVE_DECODER_ARM64) + { + .id = LZMA_FILTER_ARM64, + .options_size = sizeof(lzma_options_bcj), + .non_last_ok = true, + .last_ok = false, + .changes_size = false, + }, +#endif #if defined(HAVE_ENCODER_SPARC) || defined(HAVE_DECODER_SPARC) { .id = LZMA_FILTER_SPARC, @@ -122,12 +138,16 @@ static const struct { extern LZMA_API(lzma_ret) -lzma_filters_copy(const lzma_filter *src, lzma_filter *dest, +lzma_filters_copy(const lzma_filter *src, lzma_filter *real_dest, const lzma_allocator *allocator) { - if (src == NULL || dest == NULL) + if (src == NULL || real_dest == NULL) return LZMA_PROG_ERROR; + // Use a temporary destination so that the real destination + // will never be modied if an error occurs. + lzma_filter dest[LZMA_FILTERS_MAX + 1]; + lzma_ret ret; size_t i; for (i = 0; src[i].id != LZMA_VLI_UNKNOWN; ++i) { @@ -173,25 +193,53 @@ lzma_filters_copy(const lzma_filter *src, lzma_filter *dest, } // Terminate the filter array. - assert(i <= LZMA_FILTERS_MAX + 1); + assert(i < LZMA_FILTERS_MAX + 1); dest[i].id = LZMA_VLI_UNKNOWN; dest[i].options = NULL; + // Copy it to the caller-supplied array now that we know that + // no errors occurred. + memcpy(real_dest, dest, (i + 1) * sizeof(lzma_filter)); + return LZMA_OK; error: // Free the options which we have already allocated. - while (i-- > 0) { + while (i-- > 0) lzma_free(dest[i].options, allocator); - dest[i].options = NULL; - } return ret; } -static lzma_ret -validate_chain(const lzma_filter *filters, size_t *count) +extern LZMA_API(void) +lzma_filters_free(lzma_filter *filters, const lzma_allocator *allocator) +{ + if (filters == NULL) + return; + + for (size_t i = 0; filters[i].id != LZMA_VLI_UNKNOWN; ++i) { + if (i == LZMA_FILTERS_MAX) { + // The API says that LZMA_FILTERS_MAX + 1 is the + // maximum allowed size including the terminating + // element. Thus, we should never get here but in + // case there is a bug and we do anyway, don't go + // past the (probable) end of the array. + assert(0); + break; + } + + lzma_free(filters[i].options, allocator); + filters[i].options = NULL; + filters[i].id = LZMA_VLI_UNKNOWN; + } + + return; +} + + +extern lzma_ret +lzma_validate_chain(const lzma_filter *filters, size_t *count) { // There must be at least one filter. if (filters == NULL || filters[0].id == LZMA_VLI_UNKNOWN) @@ -245,7 +293,7 @@ lzma_raw_coder_init(lzma_next_coder *next, const lzma_allocator *allocator, { // Do some basic validation and get the number of filters. size_t count; - return_if_error(validate_chain(options, &count)); + return_if_error(lzma_validate_chain(options, &count)); // Set the filter functions and copy the options pointer. lzma_filter_info filters[LZMA_FILTERS_MAX + 1]; @@ -298,7 +346,7 @@ lzma_raw_coder_memusage(lzma_filter_find coder_find, // The chain has to have at least one filter. { size_t tmp; - if (validate_chain(filters, &tmp) != LZMA_OK) + if (lzma_validate_chain(filters, &tmp) != LZMA_OK) return UINT64_MAX; } diff --git a/contrib/libs/lzma/liblzma/common/filter_common.h b/contrib/libs/lzma/liblzma/common/filter_common.h index 9390305c26..2e47bb69f7 100644 --- a/contrib/libs/lzma/liblzma/common/filter_common.h +++ b/contrib/libs/lzma/liblzma/common/filter_common.h @@ -35,6 +35,9 @@ typedef struct { typedef const lzma_filter_coder *(*lzma_filter_find)(lzma_vli id); +extern lzma_ret lzma_validate_chain(const lzma_filter *filters, size_t *count); + + extern lzma_ret lzma_raw_coder_init( lzma_next_coder *next, const lzma_allocator *allocator, const lzma_filter *filters, diff --git a/contrib/libs/lzma/liblzma/common/filter_decoder.c b/contrib/libs/lzma/liblzma/common/filter_decoder.c index c75b0a89c3..fa53f5bdba 100644 --- a/contrib/libs/lzma/liblzma/common/filter_decoder.c +++ b/contrib/libs/lzma/liblzma/common/filter_decoder.c @@ -50,6 +50,12 @@ static const lzma_filter_decoder decoders[] = { .memusage = &lzma_lzma_decoder_memusage, .props_decode = &lzma_lzma_props_decode, }, + { + .id = LZMA_FILTER_LZMA1EXT, + .init = &lzma_lzma_decoder_init, + .memusage = &lzma_lzma_decoder_memusage, + .props_decode = &lzma_lzma_props_decode, + }, #endif #ifdef HAVE_DECODER_LZMA2 { @@ -99,6 +105,14 @@ static const lzma_filter_decoder decoders[] = { .props_decode = &lzma_simple_props_decode, }, #endif +#ifdef HAVE_DECODER_ARM64 + { + .id = LZMA_FILTER_ARM64, + .init = &lzma_simple_arm64_decoder_init, + .memusage = NULL, + .props_decode = &lzma_simple_props_decode, + }, +#endif #ifdef HAVE_DECODER_SPARC { .id = LZMA_FILTER_SPARC, diff --git a/contrib/libs/lzma/liblzma/common/filter_encoder.c b/contrib/libs/lzma/liblzma/common/filter_encoder.c index c5d8f39721..978b7a6bb5 100644 --- a/contrib/libs/lzma/liblzma/common/filter_encoder.c +++ b/contrib/libs/lzma/liblzma/common/filter_encoder.c @@ -59,7 +59,16 @@ static const lzma_filter_encoder encoders[] = { .id = LZMA_FILTER_LZMA1, .init = &lzma_lzma_encoder_init, .memusage = &lzma_lzma_encoder_memusage, - .block_size = NULL, // FIXME + .block_size = NULL, // Not needed for LZMA1 + .props_size_get = NULL, + .props_size_fixed = 5, + .props_encode = &lzma_lzma_props_encode, + }, + { + .id = LZMA_FILTER_LZMA1EXT, + .init = &lzma_lzma_encoder_init, + .memusage = &lzma_lzma_encoder_memusage, + .block_size = NULL, // Not needed for LZMA1 .props_size_get = NULL, .props_size_fixed = 5, .props_encode = &lzma_lzma_props_encode, @@ -70,7 +79,7 @@ static const lzma_filter_encoder encoders[] = { .id = LZMA_FILTER_LZMA2, .init = &lzma_lzma2_encoder_init, .memusage = &lzma_lzma2_encoder_memusage, - .block_size = &lzma_lzma2_block_size, // FIXME + .block_size = &lzma_lzma2_block_size, .props_size_get = NULL, .props_size_fixed = 1, .props_encode = &lzma_lzma2_props_encode, @@ -126,6 +135,16 @@ static const lzma_filter_encoder encoders[] = { .props_encode = &lzma_simple_props_encode, }, #endif +#ifdef HAVE_ENCODER_ARM64 + { + .id = LZMA_FILTER_ARM64, + .init = &lzma_simple_arm64_encoder_init, + .memusage = NULL, + .block_size = NULL, + .props_size_get = &lzma_simple_props_size, + .props_encode = &lzma_simple_props_encode, + }, +#endif #ifdef HAVE_ENCODER_SPARC { .id = LZMA_FILTER_SPARC, diff --git a/contrib/libs/lzma/liblzma/common/hardware_cputhreads.c b/contrib/libs/lzma/liblzma/common/hardware_cputhreads.c index f468366a60..5d246d2cc0 100644 --- a/contrib/libs/lzma/liblzma/common/hardware_cputhreads.c +++ b/contrib/libs/lzma/liblzma/common/hardware_cputhreads.c @@ -15,6 +15,18 @@ #include "tuklib_cpucores.h" +#ifdef HAVE_SYMBOL_VERSIONS_LINUX +// This is for compatibility with binaries linked against liblzma that +// has been patched with xz-5.2.2-compat-libs.patch from RHEL/CentOS 7. +LZMA_SYMVER_API("lzma_cputhreads@XZ_5.2.2", + uint32_t, lzma_cputhreads_522)(void) lzma_nothrow + __attribute__((__alias__("lzma_cputhreads_52"))); + +LZMA_SYMVER_API("lzma_cputhreads@@XZ_5.2", + uint32_t, lzma_cputhreads_52)(void) lzma_nothrow; + +#define lzma_cputhreads lzma_cputhreads_52 +#endif extern LZMA_API(uint32_t) lzma_cputhreads(void) { diff --git a/contrib/libs/lzma/liblzma/common/index.c b/contrib/libs/lzma/liblzma/common/index.c index a41e8f3308..24ec3c10c2 100644 --- a/contrib/libs/lzma/liblzma/common/index.c +++ b/contrib/libs/lzma/liblzma/common/index.c @@ -656,6 +656,10 @@ lzma_index_append(lzma_index *i, const lzma_allocator *allocator, const uint32_t index_list_size_add = lzma_vli_size(unpadded_size) + lzma_vli_size(uncompressed_size); + // Check that uncompressed size will not overflow. + if (uncompressed_base + uncompressed_size > LZMA_VLI_MAX) + return LZMA_DATA_ERROR; + // Check that the file size will stay within limits. if (index_file_size(s->node.compressed_base, compressed_base + unpadded_size, s->record_count + 1, @@ -767,6 +771,9 @@ extern LZMA_API(lzma_ret) lzma_index_cat(lzma_index *restrict dest, lzma_index *restrict src, const lzma_allocator *allocator) { + if (dest == NULL || src == NULL) + return LZMA_PROG_ERROR; + const lzma_vli dest_file_size = lzma_index_file_size(dest); // Check that we don't exceed the file size limits. @@ -835,6 +842,11 @@ lzma_index_cat(lzma_index *restrict dest, lzma_index *restrict src, } } + // dest->checks includes the check types of all except the last Stream + // in dest. Set the bit for the check type of the last Stream now so + // that it won't get lost when Stream(s) from src are appended to dest. + dest->checks = lzma_index_checks(dest); + // Add all the Streams from src to dest. Update the base offsets // of each Stream from src. const index_cat_info info = { @@ -851,7 +863,7 @@ lzma_index_cat(lzma_index *restrict dest, lzma_index *restrict src, dest->total_size += src->total_size; dest->record_count += src->record_count; dest->index_list_size += src->index_list_size; - dest->checks = lzma_index_checks(dest) | src->checks; + dest->checks |= src->checks; // There's nothing else left in src than the base structure. lzma_free(src, allocator); @@ -1226,7 +1238,7 @@ lzma_index_iter_locate(lzma_index_iter *iter, lzma_vli target) // Use binary search to locate the exact Record. It is the first // Record whose uncompressed_sum is greater than target. - // This is because we want the rightmost Record that fullfills the + // This is because we want the rightmost Record that fulfills the // search criterion. It is possible that there are empty Blocks; // we don't want to return them. size_t left = 0; diff --git a/contrib/libs/lzma/liblzma/common/index_decoder.c b/contrib/libs/lzma/liblzma/common/index_decoder.c index cc07a1b8c5..b268988533 100644 --- a/contrib/libs/lzma/liblzma/common/index_decoder.c +++ b/contrib/libs/lzma/liblzma/common/index_decoder.c @@ -10,7 +10,7 @@ // /////////////////////////////////////////////////////////////////////////////// -#include "index.h" +#include "index_decoder.h" #include "check.h" @@ -180,8 +180,11 @@ index_decode(void *coder_ptr, const lzma_allocator *allocator, return LZMA_OK; if (((coder->crc32 >> (coder->pos * 8)) & 0xFF) - != in[(*in_pos)++]) + != in[(*in_pos)++]) { +#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION return LZMA_DATA_ERROR; +#endif + } } while (++coder->pos < 4); @@ -265,11 +268,11 @@ index_decoder_reset(lzma_index_coder *coder, const lzma_allocator *allocator, } -static lzma_ret -index_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, +extern lzma_ret +lzma_index_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, lzma_index **i, uint64_t memlimit) { - lzma_next_coder_init(&index_decoder_init, next, allocator); + lzma_next_coder_init(&lzma_index_decoder_init, next, allocator); if (i == NULL) return LZMA_PROG_ERROR; @@ -296,7 +299,7 @@ index_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, extern LZMA_API(lzma_ret) lzma_index_decoder(lzma_stream *strm, lzma_index **i, uint64_t memlimit) { - lzma_next_strm_init(index_decoder_init, strm, i, memlimit); + lzma_next_strm_init(lzma_index_decoder_init, strm, i, memlimit); strm->internal->supported_actions[LZMA_RUN] = true; strm->internal->supported_actions[LZMA_FINISH] = true; diff --git a/contrib/libs/lzma/liblzma/common/index_decoder.h b/contrib/libs/lzma/liblzma/common/index_decoder.h new file mode 100644 index 0000000000..1af433b58b --- /dev/null +++ b/contrib/libs/lzma/liblzma/common/index_decoder.h @@ -0,0 +1,24 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file index_decoder.h +/// \brief Decodes the Index field +// +// Author: Lasse Collin +// +// This file has been put into the public domain. +// You can do whatever you want with this file. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef LZMA_INDEX_DECODER_H +#define LZMA_INDEX_DECODER_H + +#include "index.h" + + +extern lzma_ret lzma_index_decoder_init(lzma_next_coder *next, + const lzma_allocator *allocator, + lzma_index **i, uint64_t memlimit); + + +#endif diff --git a/contrib/libs/lzma/liblzma/common/index_hash.c b/contrib/libs/lzma/liblzma/common/index_hash.c index d7a0344b76..34df85d72f 100644 --- a/contrib/libs/lzma/liblzma/common/index_hash.c +++ b/contrib/libs/lzma/liblzma/common/index_hash.c @@ -122,7 +122,7 @@ lzma_index_hash_size(const lzma_index_hash *index_hash) /// Updates the sizes and the hash without any validation. -static lzma_ret +static void hash_append(lzma_index_hash_info *info, lzma_vli unpadded_size, lzma_vli uncompressed_size) { @@ -136,7 +136,7 @@ hash_append(lzma_index_hash_info *info, lzma_vli unpadded_size, lzma_check_update(&info->check, LZMA_CHECK_BEST, (const uint8_t *)(sizes), sizeof(sizes)); - return LZMA_OK; + return; } @@ -152,8 +152,7 @@ lzma_index_hash_append(lzma_index_hash *index_hash, lzma_vli unpadded_size, return LZMA_PROG_ERROR; // Update the hash. - return_if_error(hash_append(&index_hash->blocks, - unpadded_size, uncompressed_size)); + hash_append(&index_hash->blocks, unpadded_size, uncompressed_size); // Validate the properties of *info are still in allowed limits. if (index_hash->blocks.blocks_size > LZMA_VLI_MAX @@ -239,9 +238,9 @@ lzma_index_hash_decode(lzma_index_hash *index_hash, const uint8_t *in, index_hash->sequence = SEQ_UNCOMPRESSED; } else { // Update the hash. - return_if_error(hash_append(&index_hash->records, + hash_append(&index_hash->records, index_hash->unpadded_size, - index_hash->uncompressed_size)); + index_hash->uncompressed_size); // Verify that we don't go over the known sizes. Note // that this validation is simpler than the one used @@ -313,8 +312,11 @@ lzma_index_hash_decode(lzma_index_hash *index_hash, const uint8_t *in, return LZMA_OK; if (((index_hash->crc32 >> (index_hash->pos * 8)) - & 0xFF) != in[(*in_pos)++]) + & 0xFF) != in[(*in_pos)++]) { +#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION return LZMA_DATA_ERROR; +#endif + } } while (++index_hash->pos < 4); diff --git a/contrib/libs/lzma/liblzma/common/lzip_decoder.c b/contrib/libs/lzma/liblzma/common/lzip_decoder.c new file mode 100644 index 0000000000..20794f9466 --- /dev/null +++ b/contrib/libs/lzma/liblzma/common/lzip_decoder.c @@ -0,0 +1,414 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file lzip_decoder.c +/// \brief Decodes .lz (lzip) files +// +// Author: Michał Górny +// Lasse Collin +// +// This file has been put into the public domain. +// You can do whatever you want with this file. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "lzip_decoder.h" +#include "lzma_decoder.h" +#include "check.h" + + +// .lz format version 0 lacks the 64-bit Member size field in the footer. +#define LZIP_V0_FOOTER_SIZE 12 +#define LZIP_V1_FOOTER_SIZE 20 +#define LZIP_FOOTER_SIZE_MAX LZIP_V1_FOOTER_SIZE + +// lc/lp/pb are hardcoded in the .lz format. +#define LZIP_LC 3 +#define LZIP_LP 0 +#define LZIP_PB 2 + + +typedef struct { + enum { + SEQ_ID_STRING, + SEQ_VERSION, + SEQ_DICT_SIZE, + SEQ_CODER_INIT, + SEQ_LZMA_STREAM, + SEQ_MEMBER_FOOTER, + } sequence; + + /// .lz member format version + uint32_t version; + + /// CRC32 of the uncompressed data in the .lz member + uint32_t crc32; + + /// Uncompressed size of the .lz member + uint64_t uncompressed_size; + + /// Compressed size of the .lz member + uint64_t member_size; + + /// Memory usage limit + uint64_t memlimit; + + /// Amount of memory actually needed + uint64_t memusage; + + /// If true, LZMA_GET_CHECK is returned after decoding the header + /// fields. As all files use CRC32 this is redundant but it's + /// implemented anyway since the initialization functions supports + /// all other flags in addition to LZMA_TELL_ANY_CHECK. + bool tell_any_check; + + /// If true, we won't calculate or verify the CRC32 of + /// the uncompressed data. + bool ignore_check; + + /// If true, we will decode concatenated .lz members and stop if + /// non-.lz data is seen after at least one member has been + /// successfully decoded. + bool concatenated; + + /// When decoding concatenated .lz members, this is true as long as + /// we are decoding the first .lz member. This is needed to avoid + /// incorrect LZMA_FORMAT_ERROR in case there is non-.lz data at + /// the end of the file. + bool first_member; + + /// Reading position in the header and footer fields + size_t pos; + + /// Buffer to hold the .lz footer fields + uint8_t buffer[LZIP_FOOTER_SIZE_MAX]; + + /// Options decoded from the .lz header that needed to initialize + /// the LZMA1 decoder. + lzma_options_lzma options; + + /// LZMA1 decoder + lzma_next_coder lzma_decoder; + +} lzma_lzip_coder; + + +static lzma_ret +lzip_decode(void *coder_ptr, const lzma_allocator *allocator, + const uint8_t *restrict in, size_t *restrict in_pos, + size_t in_size, uint8_t *restrict out, + size_t *restrict out_pos, size_t out_size, lzma_action action) +{ + lzma_lzip_coder *coder = coder_ptr; + + while (true) + switch (coder->sequence) { + case SEQ_ID_STRING: { + // The "ID string" or magic bytes are "LZIP" in US-ASCII. + const uint8_t lzip_id_string[4] = { 0x4C, 0x5A, 0x49, 0x50 }; + + while (coder->pos < sizeof(lzip_id_string)) { + if (*in_pos >= in_size) { + // If we are on the 2nd+ concatenated member + // and the input ends before we can read + // the magic bytes, we discard the bytes that + // were already read (up to 3) and finish. + // See the reasoning below. + return !coder->first_member + && action == LZMA_FINISH + ? LZMA_STREAM_END : LZMA_OK; + } + + if (in[*in_pos] != lzip_id_string[coder->pos]) { + // The .lz format allows putting non-.lz data + // at the end of the file. If we have seen + // at least one valid .lz member already, + // then we won't consume the byte at *in_pos + // and will return LZMA_STREAM_END. This way + // apps can easily locate and read the non-.lz + // data after the .lz member(s). + // + // NOTE: If the first 1-3 bytes of the non-.lz + // data match the .lz ID string then the first + // 1-3 bytes of the junk will get ignored by + // us. If apps want to properly locate the + // trailing data they must ensure that the + // first byte of their custom data isn't the + // same as the first byte of .lz ID string. + // With the liblzma API we cannot rewind the + // input position across calls to lzma_code(). + return !coder->first_member + ? LZMA_STREAM_END : LZMA_FORMAT_ERROR; + } + + ++*in_pos; + ++coder->pos; + } + + coder->pos = 0; + + coder->crc32 = 0; + coder->uncompressed_size = 0; + coder->member_size = sizeof(lzip_id_string); + + coder->sequence = SEQ_VERSION; + } + + // Fall through + + case SEQ_VERSION: + if (*in_pos >= in_size) + return LZMA_OK; + + coder->version = in[(*in_pos)++]; + + // We support version 0 and unextended version 1. + if (coder->version > 1) + return LZMA_OPTIONS_ERROR; + + ++coder->member_size; + coder->sequence = SEQ_DICT_SIZE; + + // .lz versions 0 and 1 use CRC32 as the integrity check + // so if the application wanted to know that + // (LZMA_TELL_ANY_CHECK) we can tell it now. + if (coder->tell_any_check) + return LZMA_GET_CHECK; + + // Fall through + + case SEQ_DICT_SIZE: { + if (*in_pos >= in_size) + return LZMA_OK; + + const uint32_t ds = in[(*in_pos)++]; + ++coder->member_size; + + // The five lowest bits are for the base-2 logarithm of + // the dictionary size and the highest three bits are + // the fractional part (0/16 to 7/16) that will be + // substracted to get the final value. + // + // For example, with 0xB5: + // b2log = 21 + // fracnum = 5 + // dict_size = 2^21 - 2^21 * 5 / 16 = 1408 KiB + const uint32_t b2log = ds & 0x1F; + const uint32_t fracnum = ds >> 5; + + // The format versions 0 and 1 allow dictionary size in the + // range [4 KiB, 512 MiB]. + if (b2log < 12 || b2log > 29 || (b2log == 12 && fracnum > 0)) + return LZMA_DATA_ERROR; + + // 2^[b2log] - 2^[b2log] * [fracnum] / 16 + // = 2^[b2log] - [fracnum] * 2^([b2log] - 4) + coder->options.dict_size = (UINT32_C(1) << b2log) + - (fracnum << (b2log - 4)); + + assert(coder->options.dict_size >= 4096); + assert(coder->options.dict_size <= (UINT32_C(512) << 20)); + + coder->options.preset_dict = NULL; + coder->options.lc = LZIP_LC; + coder->options.lp = LZIP_LP; + coder->options.pb = LZIP_PB; + + // Calculate the memory usage. + coder->memusage = lzma_lzma_decoder_memusage(&coder->options) + + LZMA_MEMUSAGE_BASE; + + // Initialization is a separate step because if we return + // LZMA_MEMLIMIT_ERROR we need to be able to restart after + // the memlimit has been increased. + coder->sequence = SEQ_CODER_INIT; + } + + // Fall through + + case SEQ_CODER_INIT: { + if (coder->memusage > coder->memlimit) + return LZMA_MEMLIMIT_ERROR; + + const lzma_filter_info filters[2] = { + { + .id = LZMA_FILTER_LZMA1, + .init = &lzma_lzma_decoder_init, + .options = &coder->options, + }, { + .init = NULL, + } + }; + + return_if_error(lzma_next_filter_init(&coder->lzma_decoder, + allocator, filters)); + + coder->crc32 = 0; + coder->sequence = SEQ_LZMA_STREAM; + } + + // Fall through + + case SEQ_LZMA_STREAM: { + const size_t in_start = *in_pos; + const size_t out_start = *out_pos; + + const lzma_ret ret = coder->lzma_decoder.code( + coder->lzma_decoder.coder, allocator, + in, in_pos, in_size, out, out_pos, out_size, + action); + + const size_t out_used = *out_pos - out_start; + + coder->member_size += *in_pos - in_start; + coder->uncompressed_size += out_used; + + if (!coder->ignore_check) + coder->crc32 = lzma_crc32(out + out_start, out_used, + coder->crc32); + + if (ret != LZMA_STREAM_END) + return ret; + + coder->sequence = SEQ_MEMBER_FOOTER; + } + + // Fall through + + case SEQ_MEMBER_FOOTER: { + // The footer of .lz version 0 lacks the Member size field. + // This is the only difference between version 0 and + // unextended version 1 formats. + const size_t footer_size = coder->version == 0 + ? LZIP_V0_FOOTER_SIZE + : LZIP_V1_FOOTER_SIZE; + + // Copy the CRC32, Data size, and Member size fields to + // the internal buffer. + lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos, + footer_size); + + // Return if we didn't get the whole footer yet. + if (coder->pos < footer_size) + return LZMA_OK; + + coder->pos = 0; + coder->member_size += footer_size; + + // Check that the footer fields match the observed data. + if (!coder->ignore_check + && coder->crc32 != read32le(&coder->buffer[0])) + return LZMA_DATA_ERROR; + + if (coder->uncompressed_size != read64le(&coder->buffer[4])) + return LZMA_DATA_ERROR; + + if (coder->version > 0) { + // .lz version 0 has no Member size field. + if (coder->member_size != read64le(&coder->buffer[12])) + return LZMA_DATA_ERROR; + } + + // Decoding is finished if we weren't requested to decode + // more than one .lz member. + if (!coder->concatenated) + return LZMA_STREAM_END; + + coder->first_member = false; + coder->sequence = SEQ_ID_STRING; + break; + } + + default: + assert(0); + return LZMA_PROG_ERROR; + } + + // Never reached +} + + +static void +lzip_decoder_end(void *coder_ptr, const lzma_allocator *allocator) +{ + lzma_lzip_coder *coder = coder_ptr; + lzma_next_end(&coder->lzma_decoder, allocator); + lzma_free(coder, allocator); + return; +} + + +static lzma_check +lzip_decoder_get_check(const void *coder_ptr lzma_attribute((__unused__))) +{ + return LZMA_CHECK_CRC32; +} + + +static lzma_ret +lzip_decoder_memconfig(void *coder_ptr, uint64_t *memusage, + uint64_t *old_memlimit, uint64_t new_memlimit) +{ + lzma_lzip_coder *coder = coder_ptr; + + *memusage = coder->memusage; + *old_memlimit = coder->memlimit; + + if (new_memlimit != 0) { + if (new_memlimit < coder->memusage) + return LZMA_MEMLIMIT_ERROR; + + coder->memlimit = new_memlimit; + } + + return LZMA_OK; +} + + +extern lzma_ret +lzma_lzip_decoder_init( + lzma_next_coder *next, const lzma_allocator *allocator, + uint64_t memlimit, uint32_t flags) +{ + lzma_next_coder_init(&lzma_lzip_decoder_init, next, allocator); + + if (flags & ~LZMA_SUPPORTED_FLAGS) + return LZMA_OPTIONS_ERROR; + + lzma_lzip_coder *coder = next->coder; + if (coder == NULL) { + coder = lzma_alloc(sizeof(lzma_lzip_coder), allocator); + if (coder == NULL) + return LZMA_MEM_ERROR; + + next->coder = coder; + next->code = &lzip_decode; + next->end = &lzip_decoder_end; + next->get_check = &lzip_decoder_get_check; + next->memconfig = &lzip_decoder_memconfig; + + coder->lzma_decoder = LZMA_NEXT_CODER_INIT; + } + + coder->sequence = SEQ_ID_STRING; + coder->memlimit = my_max(1, memlimit); + coder->memusage = LZMA_MEMUSAGE_BASE; + coder->tell_any_check = (flags & LZMA_TELL_ANY_CHECK) != 0; + coder->ignore_check = (flags & LZMA_IGNORE_CHECK) != 0; + coder->concatenated = (flags & LZMA_CONCATENATED) != 0; + coder->first_member = true; + coder->pos = 0; + + return LZMA_OK; +} + + +extern LZMA_API(lzma_ret) +lzma_lzip_decoder(lzma_stream *strm, uint64_t memlimit, uint32_t flags) +{ + lzma_next_strm_init(lzma_lzip_decoder_init, strm, memlimit, flags); + + strm->internal->supported_actions[LZMA_RUN] = true; + strm->internal->supported_actions[LZMA_FINISH] = true; + + return LZMA_OK; +} diff --git a/contrib/libs/lzma/liblzma/common/lzip_decoder.h b/contrib/libs/lzma/liblzma/common/lzip_decoder.h new file mode 100644 index 0000000000..33a01c352c --- /dev/null +++ b/contrib/libs/lzma/liblzma/common/lzip_decoder.h @@ -0,0 +1,22 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file lzip_decoder.h +/// \brief Decodes .lz (lzip) files +// +// Author: Michał Górny +// +// This file has been put into the public domain. +// You can do whatever you want with this file. +// +/////////////////////////////////////////////////////////////////////////////// + +#ifndef LZMA_LZIP_DECODER_H +#define LZMA_LZIP_DECODER_H + +#include "common.h" + +extern lzma_ret lzma_lzip_decoder_init( + lzma_next_coder *next, const lzma_allocator *allocator, + uint64_t memlimit, uint32_t flags); + +#endif diff --git a/contrib/libs/lzma/liblzma/common/memcmplen.h b/contrib/libs/lzma/liblzma/common/memcmplen.h index dcfd8d6f89..5a481a02c9 100644 --- a/contrib/libs/lzma/liblzma/common/memcmplen.h +++ b/contrib/libs/lzma/liblzma/common/memcmplen.h @@ -51,10 +51,6 @@ lzma_memcmplen(const uint8_t *buf1, const uint8_t *buf2, || (defined(__INTEL_COMPILER) && defined(__x86_64__)) \ || (defined(__INTEL_COMPILER) && defined(_M_X64)) \ || (defined(_MSC_VER) && defined(_M_X64))) - // NOTE: This will use 64-bit unaligned access which - // TUKLIB_FAST_UNALIGNED_ACCESS wasn't meant to permit, but - // it's convenient here at least as long as it's x86-64 only. - // // I keep this x86-64 only for now since that's where I know this // to be a good method. This may be fine on other 64-bit CPUs too. // On big endian one should use xor instead of subtraction and switch @@ -80,12 +76,12 @@ lzma_memcmplen(const uint8_t *buf1, const uint8_t *buf2, #elif defined(TUKLIB_FAST_UNALIGNED_ACCESS) \ && defined(HAVE__MM_MOVEMASK_EPI8) \ - && ((defined(__GNUC__) && defined(__SSE2_MATH__)) \ - || (defined(__INTEL_COMPILER) && defined(__SSE2__)) \ + && (defined(__SSE2__) \ || (defined(_MSC_VER) && defined(_M_IX86_FP) \ && _M_IX86_FP >= 2)) - // NOTE: Like above, this will use 128-bit unaligned access which - // TUKLIB_FAST_UNALIGNED_ACCESS wasn't meant to permit. + // NOTE: This will use 128-bit unaligned access which + // TUKLIB_FAST_UNALIGNED_ACCESS wasn't meant to permit, + // but it's convenient here since this is x86-only. // // SSE2 version for 32-bit and 64-bit x86. On x86-64 the above // version is sometimes significantly faster and sometimes diff --git a/contrib/libs/lzma/liblzma/common/microlzma_decoder.c b/contrib/libs/lzma/liblzma/common/microlzma_decoder.c new file mode 100644 index 0000000000..e473373daa --- /dev/null +++ b/contrib/libs/lzma/liblzma/common/microlzma_decoder.c @@ -0,0 +1,221 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file microlzma_decoder.c +/// \brief Decode MicroLZMA format +// +// Author: Lasse Collin +// +// This file has been put into the public domain. +// You can do whatever you want with this file. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "lzma_decoder.h" +#include "lz_decoder.h" + + +typedef struct { + /// LZMA1 decoder + lzma_next_coder lzma; + + /// Compressed size of the stream as given by the application. + /// This must be exactly correct. + /// + /// This will be decremented when input is read. + uint64_t comp_size; + + /// Uncompressed size of the stream as given by the application. + /// This may be less than the actual uncompressed size if + /// uncomp_size_is_exact is false. + /// + /// This will be decremented when output is produced. + lzma_vli uncomp_size; + + /// LZMA dictionary size as given by the application + uint32_t dict_size; + + /// If true, the exact uncompressed size is known. If false, + /// uncomp_size may be smaller than the real uncompressed size; + /// uncomp_size may never be bigger than the real uncompressed size. + bool uncomp_size_is_exact; + + /// True once the first byte of the MicroLZMA stream + /// has been processed. + bool props_decoded; +} lzma_microlzma_coder; + + +static lzma_ret +microlzma_decode(void *coder_ptr, const lzma_allocator *allocator, + const uint8_t *restrict in, size_t *restrict in_pos, + size_t in_size, uint8_t *restrict out, + size_t *restrict out_pos, size_t out_size, lzma_action action) +{ + lzma_microlzma_coder *coder = coder_ptr; + + // Remember the in start position so that we can update comp_size. + const size_t in_start = *in_pos; + + // Remember the out start position so that we can update uncomp_size. + const size_t out_start = *out_pos; + + // Limit the amount of input so that the decoder won't read more than + // comp_size. This is required when uncomp_size isn't exact because + // in that case the LZMA decoder will try to decode more input even + // when it has no output space (it can be looking for EOPM). + if (in_size - *in_pos > coder->comp_size) + in_size = *in_pos + (size_t)(coder->comp_size); + + // When the exact uncompressed size isn't known, we must limit + // the available output space to prevent the LZMA decoder from + // trying to decode too much. + if (!coder->uncomp_size_is_exact + && out_size - *out_pos > coder->uncomp_size) + out_size = *out_pos + (size_t)(coder->uncomp_size); + + if (!coder->props_decoded) { + // There must be at least one byte of input to decode + // the properties byte. + if (*in_pos >= in_size) + return LZMA_OK; + + lzma_options_lzma options = { + .dict_size = coder->dict_size, + .preset_dict = NULL, + .preset_dict_size = 0, + .ext_flags = 0, // EOPM not allowed when size is known + .ext_size_low = UINT32_MAX, // Unknown size by default + .ext_size_high = UINT32_MAX, + }; + + if (coder->uncomp_size_is_exact) + lzma_set_ext_size(options, coder->uncomp_size); + + // The properties are stored as bitwise-negation + // of the typical encoding. + if (lzma_lzma_lclppb_decode(&options, ~in[*in_pos])) + return LZMA_OPTIONS_ERROR; + + ++*in_pos; + + // Initialize the decoder. + lzma_filter_info filters[2] = { + { + .id = LZMA_FILTER_LZMA1EXT, + .init = &lzma_lzma_decoder_init, + .options = &options, + }, { + .init = NULL, + } + }; + + return_if_error(lzma_next_filter_init(&coder->lzma, + allocator, filters)); + + // Pass one dummy 0x00 byte to the LZMA decoder since that + // is what it expects the first byte to be. + const uint8_t dummy_in = 0; + size_t dummy_in_pos = 0; + if (coder->lzma.code(coder->lzma.coder, allocator, + &dummy_in, &dummy_in_pos, 1, + out, out_pos, out_size, LZMA_RUN) != LZMA_OK) + return LZMA_PROG_ERROR; + + assert(dummy_in_pos == 1); + coder->props_decoded = true; + } + + // The rest is normal LZMA decoding. + lzma_ret ret = coder->lzma.code(coder->lzma.coder, allocator, + in, in_pos, in_size, + out, out_pos, out_size, action); + + // Update the remaining compressed size. + assert(coder->comp_size >= *in_pos - in_start); + coder->comp_size -= *in_pos - in_start; + + if (coder->uncomp_size_is_exact) { + // After successful decompression of the complete stream + // the compressed size must match. + if (ret == LZMA_STREAM_END && coder->comp_size != 0) + ret = LZMA_DATA_ERROR; + } else { + // Update the amount of output remaining. + assert(coder->uncomp_size >= *out_pos - out_start); + coder->uncomp_size -= *out_pos - out_start; + + // - We must not get LZMA_STREAM_END because the stream + // shouldn't have EOPM. + // - We must use uncomp_size to determine when to + // return LZMA_STREAM_END. + if (ret == LZMA_STREAM_END) + ret = LZMA_DATA_ERROR; + else if (coder->uncomp_size == 0) + ret = LZMA_STREAM_END; + } + + return ret; +} + + +static void +microlzma_decoder_end(void *coder_ptr, const lzma_allocator *allocator) +{ + lzma_microlzma_coder *coder = coder_ptr; + lzma_next_end(&coder->lzma, allocator); + lzma_free(coder, allocator); + return; +} + + +static lzma_ret +microlzma_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, + uint64_t comp_size, + uint64_t uncomp_size, bool uncomp_size_is_exact, + uint32_t dict_size) +{ + lzma_next_coder_init(µlzma_decoder_init, next, allocator); + + lzma_microlzma_coder *coder = next->coder; + + if (coder == NULL) { + coder = lzma_alloc(sizeof(lzma_microlzma_coder), allocator); + if (coder == NULL) + return LZMA_MEM_ERROR; + + next->coder = coder; + next->code = µlzma_decode; + next->end = µlzma_decoder_end; + + coder->lzma = LZMA_NEXT_CODER_INIT; + } + + // The public API is uint64_t but the internal LZ decoder API uses + // lzma_vli. + if (uncomp_size > LZMA_VLI_MAX) + return LZMA_OPTIONS_ERROR; + + coder->comp_size = comp_size; + coder->uncomp_size = uncomp_size; + coder->uncomp_size_is_exact = uncomp_size_is_exact; + coder->dict_size = dict_size; + + coder->props_decoded = false; + + return LZMA_OK; +} + + +extern LZMA_API(lzma_ret) +lzma_microlzma_decoder(lzma_stream *strm, uint64_t comp_size, + uint64_t uncomp_size, lzma_bool uncomp_size_is_exact, + uint32_t dict_size) +{ + lzma_next_strm_init(microlzma_decoder_init, strm, comp_size, + uncomp_size, uncomp_size_is_exact, dict_size); + + strm->internal->supported_actions[LZMA_RUN] = true; + strm->internal->supported_actions[LZMA_FINISH] = true; + + return LZMA_OK; +} diff --git a/contrib/libs/lzma/liblzma/common/microlzma_encoder.c b/contrib/libs/lzma/liblzma/common/microlzma_encoder.c new file mode 100644 index 0000000000..d3ef0632dd --- /dev/null +++ b/contrib/libs/lzma/liblzma/common/microlzma_encoder.c @@ -0,0 +1,140 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file microlzma_encoder.c +/// \brief Encode into MicroLZMA format +// +// Author: Lasse Collin +// +// This file has been put into the public domain. +// You can do whatever you want with this file. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "lzma_encoder.h" + + +typedef struct { + /// LZMA1 encoder + lzma_next_coder lzma; + + /// LZMA properties byte (lc/lp/pb) + uint8_t props; +} lzma_microlzma_coder; + + +static lzma_ret +microlzma_encode(void *coder_ptr, const lzma_allocator *allocator, + const uint8_t *restrict in, size_t *restrict in_pos, + size_t in_size, uint8_t *restrict out, + size_t *restrict out_pos, size_t out_size, lzma_action action) +{ + lzma_microlzma_coder *coder = coder_ptr; + + // Remember *out_pos so that we can overwrite the first byte with + // the LZMA properties byte. + const size_t out_start = *out_pos; + + // Remember *in_pos so that we can set it based on how many + // uncompressed bytes were actually encoded. + const size_t in_start = *in_pos; + + // Set the output size limit based on the available output space. + // We know that the encoder supports set_out_limit() so + // LZMA_OPTIONS_ERROR isn't possible. LZMA_BUF_ERROR is possible + // but lzma_code() has an assertion to not allow it to be returned + // from here and I don't want to change that for now, so + // LZMA_BUF_ERROR becomes LZMA_PROG_ERROR. + uint64_t uncomp_size; + if (coder->lzma.set_out_limit(coder->lzma.coder, + &uncomp_size, out_size - *out_pos) != LZMA_OK) + return LZMA_PROG_ERROR; + + // set_out_limit fails if this isn't true. + assert(out_size - *out_pos >= 6); + + // Encode as much as possible. + const lzma_ret ret = coder->lzma.code(coder->lzma.coder, allocator, + in, in_pos, in_size, out, out_pos, out_size, action); + + if (ret != LZMA_STREAM_END) { + if (ret == LZMA_OK) { + assert(0); + return LZMA_PROG_ERROR; + } + + return ret; + } + + // The first output byte is bitwise-negation of the properties byte. + // We know that there is space for this byte because set_out_limit + // and the actual encoding succeeded. + out[out_start] = (uint8_t)(~coder->props); + + // The LZMA encoder likely read more input than it was able to encode. + // Set *in_pos based on uncomp_size. + assert(uncomp_size <= in_size - in_start); + *in_pos = in_start + (size_t)(uncomp_size); + + return ret; +} + + +static void +microlzma_encoder_end(void *coder_ptr, const lzma_allocator *allocator) +{ + lzma_microlzma_coder *coder = coder_ptr; + lzma_next_end(&coder->lzma, allocator); + lzma_free(coder, allocator); + return; +} + + +static lzma_ret +microlzma_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, + const lzma_options_lzma *options) +{ + lzma_next_coder_init(µlzma_encoder_init, next, allocator); + + lzma_microlzma_coder *coder = next->coder; + + if (coder == NULL) { + coder = lzma_alloc(sizeof(lzma_microlzma_coder), allocator); + if (coder == NULL) + return LZMA_MEM_ERROR; + + next->coder = coder; + next->code = µlzma_encode; + next->end = µlzma_encoder_end; + + coder->lzma = LZMA_NEXT_CODER_INIT; + } + + // Encode the properties byte. Bitwise-negation of it will be the + // first output byte. + return_if_error(lzma_lzma_lclppb_encode(options, &coder->props)); + + // Initialize the LZMA encoder. + const lzma_filter_info filters[2] = { + { + .id = LZMA_FILTER_LZMA1, + .init = &lzma_lzma_encoder_init, + .options = (void *)(options), + }, { + .init = NULL, + } + }; + + return lzma_next_filter_init(&coder->lzma, allocator, filters); +} + + +extern LZMA_API(lzma_ret) +lzma_microlzma_encoder(lzma_stream *strm, const lzma_options_lzma *options) +{ + lzma_next_strm_init(microlzma_encoder_init, strm, options); + + strm->internal->supported_actions[LZMA_FINISH] = true; + + return LZMA_OK; + +} diff --git a/contrib/libs/lzma/liblzma/common/outqueue.c b/contrib/libs/lzma/liblzma/common/outqueue.c index 2dc8a38d1b..71e8648a29 100644 --- a/contrib/libs/lzma/liblzma/common/outqueue.c +++ b/contrib/libs/lzma/liblzma/common/outqueue.c @@ -13,84 +13,121 @@ #include "outqueue.h" -/// This is to ease integer overflow checking: We may allocate up to -/// 2 * LZMA_THREADS_MAX buffers and we need some extra memory for other -/// data structures (that's the second /2). -#define BUF_SIZE_MAX (UINT64_MAX / LZMA_THREADS_MAX / 2 / 2) +/// Get the maximum number of buffers that may be allocated based +/// on the number of threads. For now this is twice the number of threads. +/// It's a compromise between RAM usage and keeping the worker threads busy +/// when buffers finish out of order. +#define GET_BUFS_LIMIT(threads) (2 * (threads)) -static lzma_ret -get_options(uint64_t *bufs_alloc_size, uint32_t *bufs_count, - uint64_t buf_size_max, uint32_t threads) +extern uint64_t +lzma_outq_memusage(uint64_t buf_size_max, uint32_t threads) { - if (threads > LZMA_THREADS_MAX || buf_size_max > BUF_SIZE_MAX) - return LZMA_OPTIONS_ERROR; - - // The number of buffers is twice the number of threads. - // This wastes RAM but keeps the threads busy when buffers - // finish out of order. + // This is to ease integer overflow checking: We may allocate up to + // GET_BUFS_LIMIT(LZMA_THREADS_MAX) buffers and we need some extra + // memory for other data structures too (that's the /2). // - // NOTE: If this is changed, update BUF_SIZE_MAX too. - *bufs_count = threads * 2; - *bufs_alloc_size = *bufs_count * buf_size_max; + // lzma_outq_prealloc_buf() will still accept bigger buffers than this. + const uint64_t limit + = UINT64_MAX / GET_BUFS_LIMIT(LZMA_THREADS_MAX) / 2; - return LZMA_OK; + if (threads > LZMA_THREADS_MAX || buf_size_max > limit) + return UINT64_MAX; + + return GET_BUFS_LIMIT(threads) + * lzma_outq_outbuf_memusage(buf_size_max); } -extern uint64_t -lzma_outq_memusage(uint64_t buf_size_max, uint32_t threads) +static void +move_head_to_cache(lzma_outq *outq, const lzma_allocator *allocator) { - uint64_t bufs_alloc_size; - uint32_t bufs_count; + assert(outq->head != NULL); + assert(outq->tail != NULL); + assert(outq->bufs_in_use > 0); - if (get_options(&bufs_alloc_size, &bufs_count, buf_size_max, threads) - != LZMA_OK) - return UINT64_MAX; + lzma_outbuf *buf = outq->head; + outq->head = buf->next; + if (outq->head == NULL) + outq->tail = NULL; - return sizeof(lzma_outq) + bufs_count * sizeof(lzma_outbuf) - + bufs_alloc_size; + if (outq->cache != NULL && outq->cache->allocated != buf->allocated) + lzma_outq_clear_cache(outq, allocator); + + buf->next = outq->cache; + outq->cache = buf; + + --outq->bufs_in_use; + outq->mem_in_use -= lzma_outq_outbuf_memusage(buf->allocated); + + return; +} + + +static void +free_one_cached_buffer(lzma_outq *outq, const lzma_allocator *allocator) +{ + assert(outq->cache != NULL); + + lzma_outbuf *buf = outq->cache; + outq->cache = buf->next; + + --outq->bufs_allocated; + outq->mem_allocated -= lzma_outq_outbuf_memusage(buf->allocated); + + lzma_free(buf, allocator); + return; +} + + +extern void +lzma_outq_clear_cache(lzma_outq *outq, const lzma_allocator *allocator) +{ + while (outq->cache != NULL) + free_one_cached_buffer(outq, allocator); + + return; +} + + +extern void +lzma_outq_clear_cache2(lzma_outq *outq, const lzma_allocator *allocator, + size_t keep_size) +{ + if (outq->cache == NULL) + return; + + // Free all but one. + while (outq->cache->next != NULL) + free_one_cached_buffer(outq, allocator); + + // Free the last one only if its size doesn't equal to keep_size. + if (outq->cache->allocated != keep_size) + free_one_cached_buffer(outq, allocator); + + return; } extern lzma_ret lzma_outq_init(lzma_outq *outq, const lzma_allocator *allocator, - uint64_t buf_size_max, uint32_t threads) + uint32_t threads) { - uint64_t bufs_alloc_size; - uint32_t bufs_count; - - // Set bufs_count and bufs_alloc_size. - return_if_error(get_options(&bufs_alloc_size, &bufs_count, - buf_size_max, threads)); - - // Allocate memory if needed. - if (outq->buf_size_max != buf_size_max - || outq->bufs_allocated != bufs_count) { - lzma_outq_end(outq, allocator); - -#if SIZE_MAX < UINT64_MAX - if (bufs_alloc_size > SIZE_MAX) - return LZMA_MEM_ERROR; -#endif - - outq->bufs = lzma_alloc(bufs_count * sizeof(lzma_outbuf), - allocator); - outq->bufs_mem = lzma_alloc((size_t)(bufs_alloc_size), - allocator); - - if (outq->bufs == NULL || outq->bufs_mem == NULL) { - lzma_outq_end(outq, allocator); - return LZMA_MEM_ERROR; - } - } + if (threads > LZMA_THREADS_MAX) + return LZMA_OPTIONS_ERROR; + + const uint32_t bufs_limit = GET_BUFS_LIMIT(threads); + + // Clear head/tail. + while (outq->head != NULL) + move_head_to_cache(outq, allocator); - // Initialize the rest of the main structure. Initialization of - // outq->bufs[] is done when they are actually needed. - outq->buf_size_max = (size_t)(buf_size_max); - outq->bufs_allocated = bufs_count; - outq->bufs_pos = 0; - outq->bufs_used = 0; + // If new buf_limit is lower than the old one, we may need to free + // a few cached buffers. + while (bufs_limit < outq->bufs_allocated) + free_one_cached_buffer(outq, allocator); + + outq->bufs_limit = bufs_limit; outq->read_pos = 0; return LZMA_OK; @@ -100,33 +137,81 @@ lzma_outq_init(lzma_outq *outq, const lzma_allocator *allocator, extern void lzma_outq_end(lzma_outq *outq, const lzma_allocator *allocator) { - lzma_free(outq->bufs, allocator); - outq->bufs = NULL; - - lzma_free(outq->bufs_mem, allocator); - outq->bufs_mem = NULL; + while (outq->head != NULL) + move_head_to_cache(outq, allocator); + lzma_outq_clear_cache(outq, allocator); return; } -extern lzma_outbuf * -lzma_outq_get_buf(lzma_outq *outq) +extern lzma_ret +lzma_outq_prealloc_buf(lzma_outq *outq, const lzma_allocator *allocator, + size_t size) { // Caller must have checked it with lzma_outq_has_buf(). - assert(outq->bufs_used < outq->bufs_allocated); + assert(outq->bufs_in_use < outq->bufs_limit); + + // If there already is appropriately-sized buffer in the cache, + // we need to do nothing. + if (outq->cache != NULL && outq->cache->allocated == size) + return LZMA_OK; + + if (size > SIZE_MAX - sizeof(lzma_outbuf)) + return LZMA_MEM_ERROR; - // Initialize the new buffer. - lzma_outbuf *buf = &outq->bufs[outq->bufs_pos]; - buf->buf = outq->bufs_mem + outq->bufs_pos * outq->buf_size_max; - buf->size = 0; + const size_t alloc_size = lzma_outq_outbuf_memusage(size); + + // The cache may have buffers but their size is wrong. + lzma_outq_clear_cache(outq, allocator); + + outq->cache = lzma_alloc(alloc_size, allocator); + if (outq->cache == NULL) + return LZMA_MEM_ERROR; + + outq->cache->next = NULL; + outq->cache->allocated = size; + + ++outq->bufs_allocated; + outq->mem_allocated += alloc_size; + + return LZMA_OK; +} + + +extern lzma_outbuf * +lzma_outq_get_buf(lzma_outq *outq, void *worker) +{ + // Caller must have used lzma_outq_prealloc_buf() to ensure these. + assert(outq->bufs_in_use < outq->bufs_limit); + assert(outq->bufs_in_use < outq->bufs_allocated); + assert(outq->cache != NULL); + + lzma_outbuf *buf = outq->cache; + outq->cache = buf->next; + buf->next = NULL; + + if (outq->tail != NULL) { + assert(outq->head != NULL); + outq->tail->next = buf; + } else { + assert(outq->head == NULL); + outq->head = buf; + } + + outq->tail = buf; + + buf->worker = worker; buf->finished = false; + buf->finish_ret = LZMA_STREAM_END; + buf->pos = 0; + buf->decoder_in_pos = 0; - // Update the queue state. - if (++outq->bufs_pos == outq->bufs_allocated) - outq->bufs_pos = 0; + buf->unpadded_size = 0; + buf->uncompressed_size = 0; - ++outq->bufs_used; + ++outq->bufs_in_use; + outq->mem_in_use += lzma_outq_outbuf_memusage(buf->allocated); return buf; } @@ -135,50 +220,68 @@ lzma_outq_get_buf(lzma_outq *outq) extern bool lzma_outq_is_readable(const lzma_outq *outq) { - uint32_t i = outq->bufs_pos - outq->bufs_used; - if (outq->bufs_pos < outq->bufs_used) - i += outq->bufs_allocated; + if (outq->head == NULL) + return false; - return outq->bufs[i].finished; + return outq->read_pos < outq->head->pos || outq->head->finished; } extern lzma_ret -lzma_outq_read(lzma_outq *restrict outq, uint8_t *restrict out, - size_t *restrict out_pos, size_t out_size, +lzma_outq_read(lzma_outq *restrict outq, + const lzma_allocator *restrict allocator, + uint8_t *restrict out, size_t *restrict out_pos, + size_t out_size, lzma_vli *restrict unpadded_size, lzma_vli *restrict uncompressed_size) { // There must be at least one buffer from which to read. - if (outq->bufs_used == 0) + if (outq->bufs_in_use == 0) return LZMA_OK; // Get the buffer. - uint32_t i = outq->bufs_pos - outq->bufs_used; - if (outq->bufs_pos < outq->bufs_used) - i += outq->bufs_allocated; - - lzma_outbuf *buf = &outq->bufs[i]; - - // If it isn't finished yet, we cannot read from it. - if (!buf->finished) - return LZMA_OK; + lzma_outbuf *buf = outq->head; // Copy from the buffer to output. - lzma_bufcpy(buf->buf, &outq->read_pos, buf->size, + // + // FIXME? In threaded decoder it may be bad to do this copy while + // the mutex is being held. + lzma_bufcpy(buf->buf, &outq->read_pos, buf->pos, out, out_pos, out_size); // Return if we didn't get all the data from the buffer. - if (outq->read_pos < buf->size) + if (!buf->finished || outq->read_pos < buf->pos) return LZMA_OK; // The buffer was finished. Tell the caller its size information. - *unpadded_size = buf->unpadded_size; - *uncompressed_size = buf->uncompressed_size; + if (unpadded_size != NULL) + *unpadded_size = buf->unpadded_size; + + if (uncompressed_size != NULL) + *uncompressed_size = buf->uncompressed_size; + + // Remember the return value. + const lzma_ret finish_ret = buf->finish_ret; // Free this buffer for further use. - --outq->bufs_used; + move_head_to_cache(outq, allocator); outq->read_pos = 0; - return LZMA_STREAM_END; + return finish_ret; +} + + +extern void +lzma_outq_enable_partial_output(lzma_outq *outq, + void (*enable_partial_output)(void *worker)) +{ + if (outq->head != NULL && !outq->head->finished + && outq->head->worker != NULL) { + enable_partial_output(outq->head->worker); + + // Set it to NULL since calling it twice is pointless. + outq->head->worker = NULL; + } + + return; } diff --git a/contrib/libs/lzma/liblzma/common/outqueue.h b/contrib/libs/lzma/liblzma/common/outqueue.h index 079634de45..596911e95e 100644 --- a/contrib/libs/lzma/liblzma/common/outqueue.h +++ b/contrib/libs/lzma/liblzma/common/outqueue.h @@ -14,16 +14,36 @@ /// Output buffer for a single thread -typedef struct { - /// Pointer to the output buffer of lzma_outq.buf_size_max bytes - uint8_t *buf; - - /// Amount of data written to buf - size_t size; +typedef struct lzma_outbuf_s lzma_outbuf; +struct lzma_outbuf_s { + /// Pointer to the next buffer. This is used for the cached buffers. + /// The worker thread must not modify this. + lzma_outbuf *next; + + /// This initialized by lzma_outq_get_buf() and + /// is used by lzma_outq_enable_partial_output(). + /// The worker thread must not modify this. + void *worker; + + /// Amount of memory allocated for buf[]. + /// The worker thread must not modify this. + size_t allocated; + + /// Writing position in the worker thread or, in other words, the + /// amount of finished data written to buf[] which can be copied out + /// + /// \note This is read by another thread and thus access + /// to this variable needs a mutex. + size_t pos; - /// Additional size information - lzma_vli unpadded_size; - lzma_vli uncompressed_size; + /// Decompression: Position in the input buffer in the worker thread + /// that matches the output "pos" above. This is used to detect if + /// more output might be possible from the worker thread: if it has + /// consumed all its input, then more output isn't possible. + /// + /// \note This is read by another thread and thus access + /// to this variable needs a mutex. + size_t decoder_in_pos; /// True when no more data will be written into this buffer. /// @@ -31,32 +51,55 @@ typedef struct { /// to this variable needs a mutex. bool finished; -} lzma_outbuf; + /// Return value for lzma_outq_read() when the last byte from + /// a finished buffer has been read. Defaults to LZMA_STREAM_END. + /// This must *not* be LZMA_OK. The idea is to allow a decoder to + /// pass an error code to the main thread, setting the code here + /// together with finished = true. + lzma_ret finish_ret; + + /// Additional size information. lzma_outq_read() may read these + /// when "finished" is true. + lzma_vli unpadded_size; + lzma_vli uncompressed_size; + + /// Buffer of "allocated" bytes + uint8_t buf[]; +}; typedef struct { - /// Array of buffers that are used cyclically. - lzma_outbuf *bufs; + /// Linked list of buffers in use. The next output byte will be + /// read from the head and buffers for the next thread will be + /// appended to the tail. tail->next is always NULL. + lzma_outbuf *head; + lzma_outbuf *tail; - /// Memory allocated for all the buffers - uint8_t *bufs_mem; + /// Number of bytes read from head->buf[] in lzma_outq_read() + size_t read_pos; - /// Amount of buffer space available in each buffer - size_t buf_size_max; + /// Linked list of allocated buffers that aren't currently used. + /// This way buffers of similar size can be reused and don't + /// need to be reallocated every time. For simplicity, all + /// cached buffers in the list have the same allocated size. + lzma_outbuf *cache; - /// Number of buffers allocated - uint32_t bufs_allocated; + /// Total amount of memory allocated for buffers + uint64_t mem_allocated; - /// Position in the bufs array. The next buffer to be taken - /// into use is bufs[bufs_pos]. - uint32_t bufs_pos; + /// Amount of memory used by the buffers that are in use in + /// the head...tail linked list. + uint64_t mem_in_use; - /// Number of buffers in use - uint32_t bufs_used; + /// Number of buffers in use in the head...tail list. If and only if + /// this is zero, the pointers head and tail above are NULL. + uint32_t bufs_in_use; - /// Position in the buffer in lzma_outq_read() - size_t read_pos; + /// Number of buffers allocated (in use + cached) + uint32_t bufs_allocated; + /// Maximum allowed number of allocated buffers + uint32_t bufs_limit; } lzma_outq; @@ -76,32 +119,60 @@ extern uint64_t lzma_outq_memusage(uint64_t buf_size_max, uint32_t threads); /// function knows that there are no previous /// allocations to free. /// \param allocator Pointer to allocator or NULL -/// \param buf_size_max Maximum amount of data that a single buffer -/// in the queue may need to store. /// \param threads Number of buffers that may be in use /// concurrently. Note that more than this number -/// of buffers will actually get allocated to +/// of buffers may actually get allocated to /// improve performance when buffers finish -/// out of order. +/// out of order. The actual maximum number of +/// allocated buffers is derived from the number +/// of threads. /// /// \return - LZMA_OK /// - LZMA_MEM_ERROR /// -extern lzma_ret lzma_outq_init( - lzma_outq *outq, const lzma_allocator *allocator, - uint64_t buf_size_max, uint32_t threads); +extern lzma_ret lzma_outq_init(lzma_outq *outq, + const lzma_allocator *allocator, uint32_t threads); /// \brief Free the memory associated with the output queue extern void lzma_outq_end(lzma_outq *outq, const lzma_allocator *allocator); +/// \brief Free all cached buffers that consume memory but aren't in use +extern void lzma_outq_clear_cache( + lzma_outq *outq, const lzma_allocator *allocator); + + +/// \brief Like lzma_outq_clear_cache() but might keep one buffer +/// +/// One buffer is not freed if its size is equal to keep_size. +/// This is useful if the caller knows that it will soon need a buffer of +/// keep_size bytes. This way it won't be freed and immediately reallocated. +extern void lzma_outq_clear_cache2( + lzma_outq *outq, const lzma_allocator *allocator, + size_t keep_size); + + +/// \brief Preallocate a new buffer into cache +/// +/// Splitting the buffer allocation into a separate function makes it +/// possible to ensure that way lzma_outq_get_buf() cannot fail. +/// If the preallocated buffer isn't actually used (for example, some +/// other error occurs), the caller has to do nothing as the buffer will +/// be used later or cleared from the cache when not needed. +/// +/// \return LZMA_OK on success, LZMA_MEM_ERROR if allocation fails +/// +extern lzma_ret lzma_outq_prealloc_buf( + lzma_outq *outq, const lzma_allocator *allocator, size_t size); + + /// \brief Get a new buffer /// -/// lzma_outq_has_buf() must be used to check that there is a buffer +/// lzma_outq_prealloc_buf() must be used to ensure that there is a buffer /// available before calling lzma_outq_get_buf(). /// -extern lzma_outbuf *lzma_outq_get_buf(lzma_outq *outq); +extern lzma_outbuf *lzma_outq_get_buf(lzma_outq *outq, void *worker); /// \brief Test if there is data ready to be read @@ -126,17 +197,32 @@ extern bool lzma_outq_is_readable(const lzma_outq *outq); /// \return - LZMA: All OK. Either no data was available or the buffer /// being read didn't become empty yet. /// - LZMA_STREAM_END: The buffer being read was finished. -/// *unpadded_size and *uncompressed_size were set. +/// *unpadded_size and *uncompressed_size were set if they +/// were not NULL. /// -/// \note This reads lzma_outbuf.finished variables and thus call -/// to this function needs to be protected with a mutex. +/// \note This reads lzma_outbuf.finished and .pos variables and thus +/// calls to this function need to be protected with a mutex. /// extern lzma_ret lzma_outq_read(lzma_outq *restrict outq, + const lzma_allocator *restrict allocator, uint8_t *restrict out, size_t *restrict out_pos, size_t out_size, lzma_vli *restrict unpadded_size, lzma_vli *restrict uncompressed_size); +/// \brief Enable partial output from a worker thread +/// +/// If the buffer at the head of the output queue isn't finished, +/// this will call enable_partial_output on the worker associated with +/// that output buffer. +/// +/// \note This reads a lzma_outbuf.finished variable and thus +/// calls to this function need to be protected with a mutex. +/// +extern void lzma_outq_enable_partial_output(lzma_outq *outq, + void (*enable_partial_output)(void *worker)); + + /// \brief Test if there is at least one buffer free /// /// This must be used before getting a new buffer with lzma_outq_get_buf(). @@ -144,7 +230,7 @@ extern lzma_ret lzma_outq_read(lzma_outq *restrict outq, static inline bool lzma_outq_has_buf(const lzma_outq *outq) { - return outq->bufs_used < outq->bufs_allocated; + return outq->bufs_in_use < outq->bufs_limit; } @@ -152,5 +238,17 @@ lzma_outq_has_buf(const lzma_outq *outq) static inline bool lzma_outq_is_empty(const lzma_outq *outq) { - return outq->bufs_used == 0; + return outq->bufs_in_use == 0; +} + + +/// \brief Get the amount of memory needed for a single lzma_outbuf +/// +/// \note Caller must check that the argument is significantly less +/// than SIZE_MAX to avoid an integer overflow! +static inline uint64_t +lzma_outq_outbuf_memusage(size_t buf_size) +{ + assert(buf_size <= SIZE_MAX - sizeof(lzma_outbuf)); + return sizeof(lzma_outbuf) + buf_size; } diff --git a/contrib/libs/lzma/liblzma/common/stream_decoder.c b/contrib/libs/lzma/liblzma/common/stream_decoder.c index fdd8ff2f9a..dcf7c1499f 100644 --- a/contrib/libs/lzma/liblzma/common/stream_decoder.c +++ b/contrib/libs/lzma/liblzma/common/stream_decoder.c @@ -18,15 +18,14 @@ typedef struct { enum { SEQ_STREAM_HEADER, SEQ_BLOCK_HEADER, - SEQ_BLOCK, + SEQ_BLOCK_INIT, + SEQ_BLOCK_RUN, SEQ_INDEX, SEQ_STREAM_FOOTER, SEQ_STREAM_PADDING, } sequence; - /// Block or Metadata decoder. This takes little memory and the same - /// data structure can be used to decode every Block Header, so it's - /// a good idea to have a separate lzma_next_coder structure for it. + /// Block decoder lzma_next_coder block_decoder; /// Block options decoded by the Block Header decoder and used by @@ -63,9 +62,9 @@ typedef struct { /// If true, we will decode concatenated Streams that possibly have /// Stream Padding between or after them. LZMA_STREAM_END is returned - /// once the application isn't giving us any new input, and we aren't - /// in the middle of a Stream, and possible Stream Padding is a - /// multiple of four bytes. + /// once the application isn't giving us any new input (LZMA_FINISH), + /// and we aren't in the middle of a Stream, and possible + /// Stream Padding is a multiple of four bytes. bool concatenated; /// When decoding concatenated Streams, this is true as long as we @@ -187,6 +186,15 @@ stream_decode(void *coder_ptr, const lzma_allocator *allocator, return LZMA_OK; coder->pos = 0; + coder->sequence = SEQ_BLOCK_INIT; + } + + // Fall through + + case SEQ_BLOCK_INIT: { + // Checking memusage and doing the initialization needs + // its own sequence point because we need to be able to + // retry if we return LZMA_MEMLIMIT_ERROR. // Version 1 is needed to support the .ignore_check option. coder->block_options.version = 1; @@ -235,22 +243,20 @@ stream_decode(void *coder_ptr, const lzma_allocator *allocator, // Free the allocated filter options since they are needed // only to initialize the Block decoder. - for (size_t i = 0; i < LZMA_FILTERS_MAX; ++i) - lzma_free(filters[i].options, allocator); - + lzma_filters_free(filters, allocator); coder->block_options.filters = NULL; - // Check if memory usage calculation and Block enocoder + // Check if memory usage calculation and Block decoder // initialization succeeded. if (ret != LZMA_OK) return ret; - coder->sequence = SEQ_BLOCK; + coder->sequence = SEQ_BLOCK_RUN; } // Fall through - case SEQ_BLOCK: { + case SEQ_BLOCK_RUN: { const lzma_ret ret = coder->block_decoder.code( coder->block_decoder.coder, allocator, in, in_pos, in_size, out, out_pos, out_size, diff --git a/contrib/libs/lzma/liblzma/common/stream_decoder_mt.c b/contrib/libs/lzma/liblzma/common/stream_decoder_mt.c new file mode 100644 index 0000000000..5733c76489 --- /dev/null +++ b/contrib/libs/lzma/liblzma/common/stream_decoder_mt.c @@ -0,0 +1,2016 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file stream_decoder_mt.c +/// \brief Multithreaded .xz Stream decoder +// +// Authors: Sebastian Andrzej Siewior +// Lasse Collin +// +// This file has been put into the public domain. +// You can do whatever you want with this file. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "common.h" +#include "block_decoder.h" +#include "stream_decoder.h" +#include "index.h" +#include "outqueue.h" + + +typedef enum { + /// Waiting for work. + /// Main thread may change this to THR_RUN or THR_EXIT. + THR_IDLE, + + /// Decoding is in progress. + /// Main thread may change this to THR_STOP or THR_EXIT. + /// The worker thread may change this to THR_IDLE. + THR_RUN, + + /// The main thread wants the thread to stop whatever it was doing + /// but not exit. Main thread may change this to THR_EXIT. + /// The worker thread may change this to THR_IDLE. + THR_STOP, + + /// The main thread wants the thread to exit. + THR_EXIT, + +} worker_state; + + +typedef enum { + /// Partial updates (storing of worker thread progress + /// to lzma_outbuf) are disabled. + PARTIAL_DISABLED, + + /// Main thread requests partial updates to be enabled but + /// no partial update has been done by the worker thread yet. + /// + /// Changing from PARTIAL_DISABLED to PARTIAL_START requires + /// use of the worker-thread mutex. Other transitions don't + /// need a mutex. + PARTIAL_START, + + /// Partial updates are enabled and the worker thread has done + /// at least one partial update. + PARTIAL_ENABLED, + +} partial_update_mode; + + +struct worker_thread { + /// Worker state is protected with our mutex. + worker_state state; + + /// Input buffer that will contain the whole Block except Block Header. + uint8_t *in; + + /// Amount of memory allocated for "in" + size_t in_size; + + /// Number of bytes written to "in" by the main thread + size_t in_filled; + + /// Number of bytes consumed from "in" by the worker thread. + size_t in_pos; + + /// Amount of uncompressed data that has been decoded. This local + /// copy is needed because updating outbuf->pos requires locking + /// the main mutex (coder->mutex). + size_t out_pos; + + /// Pointer to the main structure is needed to (1) lock the main + /// mutex (coder->mutex) when updating outbuf->pos and (2) when + /// putting this thread back to the stack of free threads. + struct lzma_stream_coder *coder; + + /// The allocator is set by the main thread. Since a copy of the + /// pointer is kept here, the application must not change the + /// allocator before calling lzma_end(). + const lzma_allocator *allocator; + + /// Output queue buffer to which the uncompressed data is written. + lzma_outbuf *outbuf; + + /// Amount of compressed data that has already been decompressed. + /// This is updated from in_pos when our mutex is locked. + /// This is size_t, not uint64_t, because per-thread progress + /// is limited to sizes of allocated buffers. + size_t progress_in; + + /// Like progress_in but for uncompressed data. + size_t progress_out; + + /// Updating outbuf->pos requires locking the main mutex + /// (coder->mutex). Since the main thread will only read output + /// from the oldest outbuf in the queue, only the worker thread + /// that is associated with the oldest outbuf needs to update its + /// outbuf->pos. This avoids useless mutex contention that would + /// happen if all worker threads were frequently locking the main + /// mutex to update their outbuf->pos. + /// + /// Only when partial_update is something else than PARTIAL_DISABLED, + /// this worker thread will update outbuf->pos after each call to + /// the Block decoder. + partial_update_mode partial_update; + + /// Block decoder + lzma_next_coder block_decoder; + + /// Thread-specific Block options are needed because the Block + /// decoder modifies the struct given to it at initialization. + lzma_block block_options; + + /// Filter chain memory usage + uint64_t mem_filters; + + /// Next structure in the stack of free worker threads. + struct worker_thread *next; + + mythread_mutex mutex; + mythread_cond cond; + + /// The ID of this thread is used to join the thread + /// when it's not needed anymore. + mythread thread_id; +}; + + +struct lzma_stream_coder { + enum { + SEQ_STREAM_HEADER, + SEQ_BLOCK_HEADER, + SEQ_BLOCK_INIT, + SEQ_BLOCK_THR_INIT, + SEQ_BLOCK_THR_RUN, + SEQ_BLOCK_DIRECT_INIT, + SEQ_BLOCK_DIRECT_RUN, + SEQ_INDEX_WAIT_OUTPUT, + SEQ_INDEX_DECODE, + SEQ_STREAM_FOOTER, + SEQ_STREAM_PADDING, + SEQ_ERROR, + } sequence; + + /// Block decoder + lzma_next_coder block_decoder; + + /// Every Block Header will be decoded into this structure. + /// This is also used to initialize a Block decoder when in + /// direct mode. In threaded mode, a thread-specific copy will + /// be made for decoder initialization because the Block decoder + /// will modify the structure given to it. + lzma_block block_options; + + /// Buffer to hold a filter chain for Block Header decoding and + /// initialization. These are freed after successful Block decoder + /// initialization or at stream_decoder_mt_end(). The thread-specific + /// copy of block_options won't hold a pointer to filters[] after + /// initialization. + lzma_filter filters[LZMA_FILTERS_MAX + 1]; + + /// Stream Flags from Stream Header + lzma_stream_flags stream_flags; + + /// Index is hashed so that it can be compared to the sizes of Blocks + /// with O(1) memory usage. + lzma_index_hash *index_hash; + + + /// Maximum wait time if cannot use all the input and cannot + /// fill the output buffer. This is in milliseconds. + uint32_t timeout; + + + /// Error code from a worker thread. + /// + /// \note Use mutex. + lzma_ret thread_error; + + /// Error code to return after pending output has been copied out. If + /// set in read_output_and_wait(), this is a mirror of thread_error. + /// If set in stream_decode_mt() then it's, for example, error that + /// occurred when decoding Block Header. + lzma_ret pending_error; + + /// Number of threads that will be created at maximum. + uint32_t threads_max; + + /// Number of thread structures that have been initialized from + /// "threads", and thus the number of worker threads actually + /// created so far. + uint32_t threads_initialized; + + /// Array of allocated thread-specific structures. When no threads + /// are in use (direct mode) this is NULL. In threaded mode this + /// points to an array of threads_max number of worker_thread structs. + struct worker_thread *threads; + + /// Stack of free threads. When a thread finishes, it puts itself + /// back into this stack. This starts as empty because threads + /// are created only when actually needed. + /// + /// \note Use mutex. + struct worker_thread *threads_free; + + /// The most recent worker thread to which the main thread writes + /// the new input from the application. + struct worker_thread *thr; + + /// Output buffer queue for decompressed data from the worker threads + /// + /// \note Use mutex with operations that need it. + lzma_outq outq; + + mythread_mutex mutex; + mythread_cond cond; + + + /// Memory usage that will not be exceeded in multi-threaded mode. + /// Single-threaded mode can exceed this even by a large amount. + uint64_t memlimit_threading; + + /// Memory usage limit that should never be exceeded. + /// LZMA_MEMLIMIT_ERROR will be returned if decoding isn't possible + /// even in single-threaded mode without exceeding this limit. + uint64_t memlimit_stop; + + /// Amount of memory in use by the direct mode decoder + /// (coder->block_decoder). In threaded mode this is 0. + uint64_t mem_direct_mode; + + /// Amount of memory needed by the running worker threads. + /// This doesn't include the memory needed by the output buffer. + /// + /// \note Use mutex. + uint64_t mem_in_use; + + /// Amount of memory used by the idle (cached) threads. + /// + /// \note Use mutex. + uint64_t mem_cached; + + + /// Amount of memory needed for the filter chain of the next Block. + uint64_t mem_next_filters; + + /// Amount of memory needed for the thread-specific input buffer + /// for the next Block. + uint64_t mem_next_in; + + /// Amount of memory actually needed to decode the next Block + /// in threaded mode. This is + /// mem_next_filters + mem_next_in + memory needed for lzma_outbuf. + uint64_t mem_next_block; + + + /// Amount of compressed data in Stream Header + Blocks that have + /// already been finished. + /// + /// \note Use mutex. + uint64_t progress_in; + + /// Amount of uncompressed data in Blocks that have already + /// been finished. + /// + /// \note Use mutex. + uint64_t progress_out; + + + /// If true, LZMA_NO_CHECK is returned if the Stream has + /// no integrity check. + bool tell_no_check; + + /// If true, LZMA_UNSUPPORTED_CHECK is returned if the Stream has + /// an integrity check that isn't supported by this liblzma build. + bool tell_unsupported_check; + + /// If true, LZMA_GET_CHECK is returned after decoding Stream Header. + bool tell_any_check; + + /// If true, we will tell the Block decoder to skip calculating + /// and verifying the integrity check. + bool ignore_check; + + /// If true, we will decode concatenated Streams that possibly have + /// Stream Padding between or after them. LZMA_STREAM_END is returned + /// once the application isn't giving us any new input (LZMA_FINISH), + /// and we aren't in the middle of a Stream, and possible + /// Stream Padding is a multiple of four bytes. + bool concatenated; + + /// If true, we will return any errors immediately instead of first + /// producing all output before the location of the error. + bool fail_fast; + + + /// When decoding concatenated Streams, this is true as long as we + /// are decoding the first Stream. This is needed to avoid misleading + /// LZMA_FORMAT_ERROR in case the later Streams don't have valid magic + /// bytes. + bool first_stream; + + /// This is used to track if the previous call to stream_decode_mt() + /// had output space (*out_pos < out_size) and managed to fill the + /// output buffer (*out_pos == out_size). This may be set to true + /// in read_output_and_wait(). This is read and then reset to false + /// at the beginning of stream_decode_mt(). + /// + /// This is needed to support applications that call lzma_code() in + /// such a way that more input is provided only when lzma_code() + /// didn't fill the output buffer completely. Basically, this makes + /// it easier to convert such applications from single-threaded + /// decoder to multi-threaded decoder. + bool out_was_filled; + + /// Write position in buffer[] and position in Stream Padding + size_t pos; + + /// Buffer to hold Stream Header, Block Header, and Stream Footer. + /// Block Header has biggest maximum size. + uint8_t buffer[LZMA_BLOCK_HEADER_SIZE_MAX]; +}; + + +/// Enables updating of outbuf->pos. This is a callback function that is +/// used with lzma_outq_enable_partial_output(). +static void +worker_enable_partial_update(void *thr_ptr) +{ + struct worker_thread *thr = thr_ptr; + + mythread_sync(thr->mutex) { + thr->partial_update = PARTIAL_START; + mythread_cond_signal(&thr->cond); + } +} + + +/// Things do to at THR_STOP or when finishing a Block. +/// This is called with thr->mutex locked. +static void +worker_stop(struct worker_thread *thr) +{ + // Update memory usage counters. + thr->coder->mem_in_use -= thr->in_size; + thr->in_size = 0; // thr->in was freed above. + + thr->coder->mem_in_use -= thr->mem_filters; + thr->coder->mem_cached += thr->mem_filters; + + // Put this thread to the stack of free threads. + thr->next = thr->coder->threads_free; + thr->coder->threads_free = thr; + + mythread_cond_signal(&thr->coder->cond); + return; +} + + +static MYTHREAD_RET_TYPE +worker_decoder(void *thr_ptr) +{ + struct worker_thread *thr = thr_ptr; + size_t in_filled; + partial_update_mode partial_update; + lzma_ret ret; + +next_loop_lock: + + mythread_mutex_lock(&thr->mutex); +next_loop_unlocked: + + if (thr->state == THR_IDLE) { + mythread_cond_wait(&thr->cond, &thr->mutex); + goto next_loop_unlocked; + } + + if (thr->state == THR_EXIT) { + mythread_mutex_unlock(&thr->mutex); + + lzma_free(thr->in, thr->allocator); + lzma_next_end(&thr->block_decoder, thr->allocator); + + mythread_mutex_destroy(&thr->mutex); + mythread_cond_destroy(&thr->cond); + + return MYTHREAD_RET_VALUE; + } + + if (thr->state == THR_STOP) { + thr->state = THR_IDLE; + mythread_mutex_unlock(&thr->mutex); + + mythread_sync(thr->coder->mutex) { + worker_stop(thr); + } + + goto next_loop_lock; + } + + assert(thr->state == THR_RUN); + + // Update progress info for get_progress(). + thr->progress_in = thr->in_pos; + thr->progress_out = thr->out_pos; + + // If we don't have any new input, wait for a signal from the main + // thread except if partial output has just been enabled. In that + // case we will do one normal run so that the partial output info + // gets passed to the main thread. The call to block_decoder.code() + // is useless but harmless as it can occur only once per Block. + in_filled = thr->in_filled; + partial_update = thr->partial_update; + + if (in_filled == thr->in_pos && partial_update != PARTIAL_START) { + mythread_cond_wait(&thr->cond, &thr->mutex); + goto next_loop_unlocked; + } + + mythread_mutex_unlock(&thr->mutex); + + // Pass the input in small chunks to the Block decoder. + // This way we react reasonably fast if we are told to stop/exit, + // and (when partial update is enabled) we tell about our progress + // to the main thread frequently enough. + const size_t chunk_size = 16384; + if ((in_filled - thr->in_pos) > chunk_size) + in_filled = thr->in_pos + chunk_size; + + ret = thr->block_decoder.code( + thr->block_decoder.coder, thr->allocator, + thr->in, &thr->in_pos, in_filled, + thr->outbuf->buf, &thr->out_pos, + thr->outbuf->allocated, LZMA_RUN); + + if (ret == LZMA_OK) { + if (partial_update != PARTIAL_DISABLED) { + // The main thread uses thr->mutex to change from + // PARTIAL_DISABLED to PARTIAL_START. The main thread + // doesn't care about this variable after that so we + // can safely change it here to PARTIAL_ENABLED + // without a mutex. + thr->partial_update = PARTIAL_ENABLED; + + // The main thread is reading decompressed data + // from thr->outbuf. Tell the main thread about + // our progress. + // + // NOTE: It's possible that we consumed input without + // producing any new output so it's possible that + // only in_pos has changed. In case of PARTIAL_START + // it is possible that neither in_pos nor out_pos has + // changed. + mythread_sync(thr->coder->mutex) { + thr->outbuf->pos = thr->out_pos; + thr->outbuf->decoder_in_pos = thr->in_pos; + mythread_cond_signal(&thr->coder->cond); + } + } + + goto next_loop_lock; + } + + // Either we finished successfully (LZMA_STREAM_END) or an error + // occurred. Both cases are handled almost identically. The error + // case requires updating thr->coder->thread_error. + // + // The sizes are in the Block Header and the Block decoder + // checks that they match, thus we know these: + assert(ret != LZMA_STREAM_END || thr->in_pos == thr->in_size); + assert(ret != LZMA_STREAM_END + || thr->out_pos == thr->block_options.uncompressed_size); + + // Free the input buffer. Don't update in_size as we need + // it later to update thr->coder->mem_in_use. + lzma_free(thr->in, thr->allocator); + thr->in = NULL; + + mythread_sync(thr->mutex) { + if (thr->state != THR_EXIT) + thr->state = THR_IDLE; + } + + mythread_sync(thr->coder->mutex) { + // Move our progress info to the main thread. + thr->coder->progress_in += thr->in_pos; + thr->coder->progress_out += thr->out_pos; + thr->progress_in = 0; + thr->progress_out = 0; + + // Mark the outbuf as finished. + thr->outbuf->pos = thr->out_pos; + thr->outbuf->decoder_in_pos = thr->in_pos; + thr->outbuf->finished = true; + thr->outbuf->finish_ret = ret; + thr->outbuf = NULL; + + // If an error occurred, tell it to the main thread. + if (ret != LZMA_STREAM_END + && thr->coder->thread_error == LZMA_OK) + thr->coder->thread_error = ret; + + worker_stop(thr); + } + + goto next_loop_lock; +} + + +/// Tells the worker threads to exit and waits for them to terminate. +static void +threads_end(struct lzma_stream_coder *coder, const lzma_allocator *allocator) +{ + for (uint32_t i = 0; i < coder->threads_initialized; ++i) { + mythread_sync(coder->threads[i].mutex) { + coder->threads[i].state = THR_EXIT; + mythread_cond_signal(&coder->threads[i].cond); + } + } + + for (uint32_t i = 0; i < coder->threads_initialized; ++i) + mythread_join(coder->threads[i].thread_id); + + lzma_free(coder->threads, allocator); + coder->threads_initialized = 0; + coder->threads = NULL; + coder->threads_free = NULL; + + // The threads don't update these when they exit. Do it here. + coder->mem_in_use = 0; + coder->mem_cached = 0; + + return; +} + + +static void +threads_stop(struct lzma_stream_coder *coder) +{ + for (uint32_t i = 0; i < coder->threads_initialized; ++i) { + mythread_sync(coder->threads[i].mutex) { + // The state must be changed conditionally because + // THR_IDLE -> THR_STOP is not a valid state change. + if (coder->threads[i].state != THR_IDLE) { + coder->threads[i].state = THR_STOP; + mythread_cond_signal(&coder->threads[i].cond); + } + } + } + + return; +} + + +/// Initialize a new worker_thread structure and create a new thread. +static lzma_ret +initialize_new_thread(struct lzma_stream_coder *coder, + const lzma_allocator *allocator) +{ + // Allocate the coder->threads array if needed. It's done here instead + // of when initializing the decoder because we don't need this if we + // use the direct mode (we may even free coder->threads in the middle + // of the file if we switch from threaded to direct mode). + if (coder->threads == NULL) { + coder->threads = lzma_alloc( + coder->threads_max * sizeof(struct worker_thread), + allocator); + + if (coder->threads == NULL) + return LZMA_MEM_ERROR; + } + + // Pick a free structure. + assert(coder->threads_initialized < coder->threads_max); + struct worker_thread *thr + = &coder->threads[coder->threads_initialized]; + + if (mythread_mutex_init(&thr->mutex)) + goto error_mutex; + + if (mythread_cond_init(&thr->cond)) + goto error_cond; + + thr->state = THR_IDLE; + thr->in = NULL; + thr->in_size = 0; + thr->allocator = allocator; + thr->coder = coder; + thr->outbuf = NULL; + thr->block_decoder = LZMA_NEXT_CODER_INIT; + thr->mem_filters = 0; + + if (mythread_create(&thr->thread_id, worker_decoder, thr)) + goto error_thread; + + ++coder->threads_initialized; + coder->thr = thr; + + return LZMA_OK; + +error_thread: + mythread_cond_destroy(&thr->cond); + +error_cond: + mythread_mutex_destroy(&thr->mutex); + +error_mutex: + return LZMA_MEM_ERROR; +} + + +static lzma_ret +get_thread(struct lzma_stream_coder *coder, const lzma_allocator *allocator) +{ + // If there is a free structure on the stack, use it. + mythread_sync(coder->mutex) { + if (coder->threads_free != NULL) { + coder->thr = coder->threads_free; + coder->threads_free = coder->threads_free->next; + + // The thread is no longer in the cache so substract + // it from the cached memory usage. Don't add it + // to mem_in_use though; the caller will handle it + // since it knows how much memory it will actually + // use (the filter chain might change). + coder->mem_cached -= coder->thr->mem_filters; + } + } + + if (coder->thr == NULL) { + assert(coder->threads_initialized < coder->threads_max); + + // Initialize a new thread. + return_if_error(initialize_new_thread(coder, allocator)); + } + + coder->thr->in_filled = 0; + coder->thr->in_pos = 0; + coder->thr->out_pos = 0; + + coder->thr->progress_in = 0; + coder->thr->progress_out = 0; + + coder->thr->partial_update = PARTIAL_DISABLED; + + return LZMA_OK; +} + + +static lzma_ret +read_output_and_wait(struct lzma_stream_coder *coder, + const lzma_allocator *allocator, + uint8_t *restrict out, size_t *restrict out_pos, + size_t out_size, + bool *input_is_possible, + bool waiting_allowed, + mythread_condtime *wait_abs, bool *has_blocked) +{ + lzma_ret ret = LZMA_OK; + + mythread_sync(coder->mutex) { + do { + // Get as much output from the queue as is possible + // without blocking. + const size_t out_start = *out_pos; + do { + ret = lzma_outq_read(&coder->outq, allocator, + out, out_pos, out_size, + NULL, NULL); + + // If a Block was finished, tell the worker + // thread of the next Block (if it is still + // running) to start telling the main thread + // when new output is available. + if (ret == LZMA_STREAM_END) + lzma_outq_enable_partial_output( + &coder->outq, + &worker_enable_partial_update); + + // Loop until a Block wasn't finished. + // It's important to loop around even if + // *out_pos == out_size because there could + // be an empty Block that will return + // LZMA_STREAM_END without needing any + // output space. + } while (ret == LZMA_STREAM_END); + + // Check if lzma_outq_read reported an error from + // the Block decoder. + if (ret != LZMA_OK) + break; + + // If the output buffer is now full but it wasn't full + // when this function was called, set out_was_filled. + // This way the next call to stream_decode_mt() knows + // that some output was produced and no output space + // remained in the previous call to stream_decode_mt(). + if (*out_pos == out_size && *out_pos != out_start) + coder->out_was_filled = true; + + // Check if any thread has indicated an error. + if (coder->thread_error != LZMA_OK) { + // If LZMA_FAIL_FAST was used, report errors + // from worker threads immediately. + if (coder->fail_fast) { + ret = coder->thread_error; + break; + } + + // Otherwise set pending_error. The value we + // set here will not actually get used other + // than working as a flag that an error has + // occurred. This is because in SEQ_ERROR + // all output before the error will be read + // first by calling this function, and once we + // reach the location of the (first) error the + // error code from the above lzma_outq_read() + // will be returned to the application. + // + // Use LZMA_PROG_ERROR since the value should + // never leak to the application. It's + // possible that pending_error has already + // been set but that doesn't matter: if we get + // here, pending_error only works as a flag. + coder->pending_error = LZMA_PROG_ERROR; + } + + // Check if decoding of the next Block can be started. + // The memusage of the active threads must be low + // enough, there must be a free buffer slot in the + // output queue, and there must be a free thread + // (that can be either created or an existing one + // reused). + // + // NOTE: This is checked after reading the output + // above because reading the output can free a slot in + // the output queue and also reduce active memusage. + // + // NOTE: If output queue is empty, then input will + // always be possible. + if (input_is_possible != NULL + && coder->memlimit_threading + - coder->mem_in_use + - coder->outq.mem_in_use + >= coder->mem_next_block + && lzma_outq_has_buf(&coder->outq) + && (coder->threads_initialized + < coder->threads_max + || coder->threads_free + != NULL)) { + *input_is_possible = true; + break; + } + + // If the caller doesn't want us to block, return now. + if (!waiting_allowed) + break; + + // This check is needed only when input_is_possible + // is NULL. We must return if we aren't waiting for + // input to become possible and there is no more + // output coming from the queue. + if (lzma_outq_is_empty(&coder->outq)) { + assert(input_is_possible == NULL); + break; + } + + // If there is more data available from the queue, + // our out buffer must be full and we need to return + // so that the application can provide more output + // space. + // + // NOTE: In general lzma_outq_is_readable() can return + // true also when there are no more bytes available. + // This can happen when a Block has finished without + // providing any new output. We know that this is not + // the case because in the beginning of this loop we + // tried to read as much as possible even when we had + // no output space left and the mutex has been locked + // all the time (so worker threads cannot have changed + // anything). Thus there must be actual pending output + // in the queue. + if (lzma_outq_is_readable(&coder->outq)) { + assert(*out_pos == out_size); + break; + } + + // If the application stops providing more input + // in the middle of a Block, there will eventually + // be one worker thread left that is stuck waiting for + // more input (that might never arrive) and a matching + // outbuf which the worker thread cannot finish due + // to lack of input. We must detect this situation, + // otherwise we would end up waiting indefinitely + // (if no timeout is in use) or keep returning + // LZMA_TIMED_OUT while making no progress. Thus, the + // application would never get LZMA_BUF_ERROR from + // lzma_code() which would tell the application that + // no more progress is possible. No LZMA_BUF_ERROR + // means that, for example, truncated .xz files could + // cause an infinite loop. + // + // A worker thread doing partial updates will + // store not only the output position in outbuf->pos + // but also the matching input position in + // outbuf->decoder_in_pos. Here we check if that + // input position matches the amount of input that + // the worker thread has been given (in_filled). + // If so, we must return and not wait as no more + // output will be coming without first getting more + // input to the worker thread. If the application + // keeps calling lzma_code() without providing more + // input, it will eventually get LZMA_BUF_ERROR. + // + // NOTE: We can read partial_update and in_filled + // without thr->mutex as only the main thread + // modifies these variables. decoder_in_pos requires + // coder->mutex which we are already holding. + if (coder->thr != NULL && coder->thr->partial_update + != PARTIAL_DISABLED) { + // There is exactly one outbuf in the queue. + assert(coder->thr->outbuf == coder->outq.head); + assert(coder->thr->outbuf == coder->outq.tail); + + if (coder->thr->outbuf->decoder_in_pos + == coder->thr->in_filled) + break; + } + + // Wait for input or output to become possible. + if (coder->timeout != 0) { + // See the comment in stream_encoder_mt.c + // about why mythread_condtime_set() is used + // like this. + // + // FIXME? + // In contrast to the encoder, this calls + // _condtime_set while the mutex is locked. + if (!*has_blocked) { + *has_blocked = true; + mythread_condtime_set(wait_abs, + &coder->cond, + coder->timeout); + } + + if (mythread_cond_timedwait(&coder->cond, + &coder->mutex, + wait_abs) != 0) { + ret = LZMA_TIMED_OUT; + break; + } + } else { + mythread_cond_wait(&coder->cond, + &coder->mutex); + } + } while (ret == LZMA_OK); + } + + // If we are returning an error, then the application cannot get + // more output from us and thus keeping the threads running is + // useless and waste of CPU time. + if (ret != LZMA_OK && ret != LZMA_TIMED_OUT) + threads_stop(coder); + + return ret; +} + + +static lzma_ret +decode_block_header(struct lzma_stream_coder *coder, + const lzma_allocator *allocator, const uint8_t *restrict in, + size_t *restrict in_pos, size_t in_size) +{ + if (*in_pos >= in_size) + return LZMA_OK; + + if (coder->pos == 0) { + // Detect if it's Index. + if (in[*in_pos] == 0x00) + return LZMA_INDEX_DETECTED; + + // Calculate the size of the Block Header. Note that + // Block Header decoder wants to see this byte too + // so don't advance *in_pos. + coder->block_options.header_size + = lzma_block_header_size_decode( + in[*in_pos]); + } + + // Copy the Block Header to the internal buffer. + lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos, + coder->block_options.header_size); + + // Return if we didn't get the whole Block Header yet. + if (coder->pos < coder->block_options.header_size) + return LZMA_OK; + + coder->pos = 0; + + // Version 1 is needed to support the .ignore_check option. + coder->block_options.version = 1; + + // Block Header decoder will initialize all members of this array + // so we don't need to do it here. + coder->block_options.filters = coder->filters; + + // Decode the Block Header. + return_if_error(lzma_block_header_decode(&coder->block_options, + allocator, coder->buffer)); + + // If LZMA_IGNORE_CHECK was used, this flag needs to be set. + // It has to be set after lzma_block_header_decode() because + // it always resets this to false. + coder->block_options.ignore_check = coder->ignore_check; + + // coder->block_options is ready now. + return LZMA_STREAM_END; +} + + +/// Get the size of the Compressed Data + Block Padding + Check. +static size_t +comp_blk_size(const struct lzma_stream_coder *coder) +{ + return vli_ceil4(coder->block_options.compressed_size) + + lzma_check_size(coder->stream_flags.check); +} + + +/// Returns true if the size (compressed or uncompressed) is such that +/// threaded decompression cannot be used. Sizes that are too big compared +/// to SIZE_MAX must be rejected to avoid integer overflows and truncations +/// when lzma_vli is assigned to a size_t. +static bool +is_direct_mode_needed(lzma_vli size) +{ + return size == LZMA_VLI_UNKNOWN || size > SIZE_MAX / 3; +} + + +static lzma_ret +stream_decoder_reset(struct lzma_stream_coder *coder, + const lzma_allocator *allocator) +{ + // Initialize the Index hash used to verify the Index. + coder->index_hash = lzma_index_hash_init(coder->index_hash, allocator); + if (coder->index_hash == NULL) + return LZMA_MEM_ERROR; + + // Reset the rest of the variables. + coder->sequence = SEQ_STREAM_HEADER; + coder->pos = 0; + + return LZMA_OK; +} + + +static lzma_ret +stream_decode_mt(void *coder_ptr, const lzma_allocator *allocator, + const uint8_t *restrict in, size_t *restrict in_pos, + size_t in_size, + uint8_t *restrict out, size_t *restrict out_pos, + size_t out_size, lzma_action action) +{ + struct lzma_stream_coder *coder = coder_ptr; + + mythread_condtime wait_abs; + bool has_blocked = false; + + // Determine if in SEQ_BLOCK_HEADER and SEQ_BLOCK_THR_RUN we should + // tell read_output_and_wait() to wait until it can fill the output + // buffer (or a timeout occurs). Two conditions must be met: + // + // (1) If the caller provided no new input. The reason for this + // can be, for example, the end of the file or that there is + // a pause in the input stream and more input is available + // a little later. In this situation we should wait for output + // because otherwise we would end up in a busy-waiting loop where + // we make no progress and the application just calls us again + // without providing any new input. This would then result in + // LZMA_BUF_ERROR even though more output would be available + // once the worker threads decode more data. + // + // (2) Even if (1) is true, we will not wait if the previous call to + // this function managed to produce some output and the output + // buffer became full. This is for compatibility with applications + // that call lzma_code() in such a way that new input is provided + // only when the output buffer didn't become full. Without this + // trick such applications would have bad performance (bad + // parallelization due to decoder not getting input fast enough). + // + // NOTE: Such loops might require that timeout is disabled (0) + // if they assume that output-not-full implies that all input has + // been consumed. If and only if timeout is enabled, we may return + // when output isn't full *and* not all input has been consumed. + // + // However, if LZMA_FINISH is used, the above is ignored and we always + // wait (timeout can still cause us to return) because we know that + // we won't get any more input. This matters if the input file is + // truncated and we are doing single-shot decoding, that is, + // timeout = 0 and LZMA_FINISH is used on the first call to + // lzma_code() and the output buffer is known to be big enough + // to hold all uncompressed data: + // + // - If LZMA_FINISH wasn't handled specially, we could return + // LZMA_OK before providing all output that is possible with the + // truncated input. The rest would be available if lzma_code() was + // called again but then it's not single-shot decoding anymore. + // + // - By handling LZMA_FINISH specially here, the first call will + // produce all the output, matching the behavior of the + // single-threaded decoder. + // + // So it's a very specific corner case but also easy to avoid. Note + // that this special handling of LZMA_FINISH has no effect for + // single-shot decoding when the input file is valid (not truncated); + // premature LZMA_OK wouldn't be possible as long as timeout = 0. + const bool waiting_allowed = action == LZMA_FINISH + || (*in_pos == in_size && !coder->out_was_filled); + coder->out_was_filled = false; + + while (true) + switch (coder->sequence) { + case SEQ_STREAM_HEADER: { + // Copy the Stream Header to the internal buffer. + const size_t in_old = *in_pos; + lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos, + LZMA_STREAM_HEADER_SIZE); + coder->progress_in += *in_pos - in_old; + + // Return if we didn't get the whole Stream Header yet. + if (coder->pos < LZMA_STREAM_HEADER_SIZE) + return LZMA_OK; + + coder->pos = 0; + + // Decode the Stream Header. + const lzma_ret ret = lzma_stream_header_decode( + &coder->stream_flags, coder->buffer); + if (ret != LZMA_OK) + return ret == LZMA_FORMAT_ERROR && !coder->first_stream + ? LZMA_DATA_ERROR : ret; + + // If we are decoding concatenated Streams, and the later + // Streams have invalid Header Magic Bytes, we give + // LZMA_DATA_ERROR instead of LZMA_FORMAT_ERROR. + coder->first_stream = false; + + // Copy the type of the Check so that Block Header and Block + // decoders see it. + coder->block_options.check = coder->stream_flags.check; + + // Even if we return LZMA_*_CHECK below, we want + // to continue from Block Header decoding. + coder->sequence = SEQ_BLOCK_HEADER; + + // Detect if there's no integrity check or if it is + // unsupported if those were requested by the application. + if (coder->tell_no_check && coder->stream_flags.check + == LZMA_CHECK_NONE) + return LZMA_NO_CHECK; + + if (coder->tell_unsupported_check + && !lzma_check_is_supported( + coder->stream_flags.check)) + return LZMA_UNSUPPORTED_CHECK; + + if (coder->tell_any_check) + return LZMA_GET_CHECK; + } + + // Fall through + + case SEQ_BLOCK_HEADER: { + const size_t in_old = *in_pos; + const lzma_ret ret = decode_block_header(coder, allocator, + in, in_pos, in_size); + coder->progress_in += *in_pos - in_old; + + if (ret == LZMA_OK) { + // We didn't decode the whole Block Header yet. + // + // Read output from the queue before returning. This + // is important because it is possible that the + // application doesn't have any new input available + // immediately. If we didn't try to copy output from + // the output queue here, lzma_code() could end up + // returning LZMA_BUF_ERROR even though queued output + // is available. + // + // If the lzma_code() call provided at least one input + // byte, only copy as much data from the output queue + // as is available immediately. This way the + // application will be able to provide more input + // without a delay. + // + // On the other hand, if lzma_code() was called with + // an empty input buffer(*), treat it specially: try + // to fill the output buffer even if it requires + // waiting for the worker threads to provide output + // (timeout, if specified, can still cause us to + // return). + // + // - This way the application will be able to get all + // data that can be decoded from the input provided + // so far. + // + // - We avoid both premature LZMA_BUF_ERROR and + // busy-waiting where the application repeatedly + // calls lzma_code() which immediately returns + // LZMA_OK without providing new data. + // + // - If the queue becomes empty, we won't wait + // anything and will return LZMA_OK immediately + // (coder->timeout is completely ignored). + // + // (*) See the comment at the beginning of this + // function how waiting_allowed is determined + // and why there is an exception to the rule + // of "called with an empty input buffer". + assert(*in_pos == in_size); + + // If LZMA_FINISH was used we know that we won't get + // more input, so the file must be truncated if we + // get here. If worker threads don't detect any + // errors, eventually there will be no more output + // while we keep returning LZMA_OK which gets + // converted to LZMA_BUF_ERROR in lzma_code(). + // + // If fail-fast is enabled then we will return + // immediately using LZMA_DATA_ERROR instead of + // LZMA_OK or LZMA_BUF_ERROR. Rationale for the + // error code: + // + // - Worker threads may have a large amount of + // not-yet-decoded input data and we don't + // know for sure if all data is valid. Bad + // data there would result in LZMA_DATA_ERROR + // when fail-fast isn't used. + // + // - Immediate LZMA_BUF_ERROR would be a bit weird + // considering the older liblzma code. lzma_code() + // even has an assertion to prevent coders from + // returning LZMA_BUF_ERROR directly. + // + // The downside of this is that with fail-fast apps + // cannot always distinguish between corrupt and + // truncated files. + if (action == LZMA_FINISH && coder->fail_fast) { + // We won't produce any more output. Stop + // the unfinished worker threads so they + // won't waste CPU time. + threads_stop(coder); + return LZMA_DATA_ERROR; + } + + // read_output_and_wait() will call threads_stop() + // if needed so with that we can use return_if_error. + return_if_error(read_output_and_wait(coder, allocator, + out, out_pos, out_size, + NULL, waiting_allowed, + &wait_abs, &has_blocked)); + + if (coder->pending_error != LZMA_OK) { + coder->sequence = SEQ_ERROR; + break; + } + + return LZMA_OK; + } + + if (ret == LZMA_INDEX_DETECTED) { + coder->sequence = SEQ_INDEX_WAIT_OUTPUT; + break; + } + + // See if an error occurred. + if (ret != LZMA_STREAM_END) { + // NOTE: Here and in all other places where + // pending_error is set, it may overwrite the value + // (LZMA_PROG_ERROR) set by read_output_and_wait(). + // That function might overwrite value set here too. + // These are fine because when read_output_and_wait() + // sets pending_error, it actually works as a flag + // variable only ("some error has occurred") and the + // actual value of pending_error is not used in + // SEQ_ERROR. In such cases SEQ_ERROR will eventually + // get the correct error code from the return value of + // a later read_output_and_wait() call. + coder->pending_error = ret; + coder->sequence = SEQ_ERROR; + break; + } + + // Calculate the memory usage of the filters / Block decoder. + coder->mem_next_filters = lzma_raw_decoder_memusage( + coder->filters); + + if (coder->mem_next_filters == UINT64_MAX) { + // One or more unknown Filter IDs. + coder->pending_error = LZMA_OPTIONS_ERROR; + coder->sequence = SEQ_ERROR; + break; + } + + coder->sequence = SEQ_BLOCK_INIT; + } + + // Fall through + + case SEQ_BLOCK_INIT: { + // Check if decoding is possible at all with the current + // memlimit_stop which we must never exceed. + // + // This needs to be the first thing in SEQ_BLOCK_INIT + // to make it possible to restart decoding after increasing + // memlimit_stop with lzma_memlimit_set(). + if (coder->mem_next_filters > coder->memlimit_stop) { + // Flush pending output before returning + // LZMA_MEMLIMIT_ERROR. If the application doesn't + // want to increase the limit, at least it will get + // all the output possible so far. + return_if_error(read_output_and_wait(coder, allocator, + out, out_pos, out_size, + NULL, true, &wait_abs, &has_blocked)); + + if (!lzma_outq_is_empty(&coder->outq)) + return LZMA_OK; + + return LZMA_MEMLIMIT_ERROR; + } + + // Check if the size information is available in Block Header. + // If it is, check if the sizes are small enough that we don't + // need to worry *too* much about integer overflows later in + // the code. If these conditions are not met, we must use the + // single-threaded direct mode. + if (is_direct_mode_needed(coder->block_options.compressed_size) + || is_direct_mode_needed( + coder->block_options.uncompressed_size)) { + coder->sequence = SEQ_BLOCK_DIRECT_INIT; + break; + } + + // Calculate the amount of memory needed for the input and + // output buffers in threaded mode. + // + // These cannot overflow because we already checked that + // the sizes are small enough using is_direct_mode_needed(). + coder->mem_next_in = comp_blk_size(coder); + const uint64_t mem_buffers = coder->mem_next_in + + lzma_outq_outbuf_memusage( + coder->block_options.uncompressed_size); + + // Add the amount needed by the filters. + // Avoid integer overflows. + if (UINT64_MAX - mem_buffers < coder->mem_next_filters) { + // Use direct mode if the memusage would overflow. + // This is a theoretical case that shouldn't happen + // in practice unless the input file is weird (broken + // or malicious). + coder->sequence = SEQ_BLOCK_DIRECT_INIT; + break; + } + + // Amount of memory needed to decode this Block in + // threaded mode: + coder->mem_next_block = coder->mem_next_filters + mem_buffers; + + // If this alone would exceed memlimit_threading, then we must + // use the single-threaded direct mode. + if (coder->mem_next_block > coder->memlimit_threading) { + coder->sequence = SEQ_BLOCK_DIRECT_INIT; + break; + } + + // Use the threaded mode. Free the direct mode decoder in + // case it has been initialized. + lzma_next_end(&coder->block_decoder, allocator); + coder->mem_direct_mode = 0; + + // Since we already know what the sizes are supposed to be, + // we can already add them to the Index hash. The Block + // decoder will verify the values while decoding. + const lzma_ret ret = lzma_index_hash_append(coder->index_hash, + lzma_block_unpadded_size( + &coder->block_options), + coder->block_options.uncompressed_size); + if (ret != LZMA_OK) { + coder->pending_error = ret; + coder->sequence = SEQ_ERROR; + break; + } + + coder->sequence = SEQ_BLOCK_THR_INIT; + } + + // Fall through + + case SEQ_BLOCK_THR_INIT: { + // We need to wait for a multiple conditions to become true + // until we can initialize the Block decoder and let a worker + // thread decode it: + // + // - Wait for the memory usage of the active threads to drop + // so that starting the decoding of this Block won't make + // us go over memlimit_threading. + // + // - Wait for at least one free output queue slot. + // + // - Wait for a free worker thread. + // + // While we wait, we must copy decompressed data to the out + // buffer and catch possible decoder errors. + // + // read_output_and_wait() does all the above. + bool block_can_start = false; + + return_if_error(read_output_and_wait(coder, allocator, + out, out_pos, out_size, + &block_can_start, true, + &wait_abs, &has_blocked)); + + if (coder->pending_error != LZMA_OK) { + coder->sequence = SEQ_ERROR; + break; + } + + if (!block_can_start) { + // It's not a timeout because return_if_error handles + // it already. Output queue cannot be empty either + // because in that case block_can_start would have + // been true. Thus the output buffer must be full and + // the queue isn't empty. + assert(*out_pos == out_size); + assert(!lzma_outq_is_empty(&coder->outq)); + return LZMA_OK; + } + + // We know that we can start decoding this Block without + // exceeding memlimit_threading. However, to stay below + // memlimit_threading may require freeing some of the + // cached memory. + // + // Get a local copy of variables that require locking the + // mutex. It is fine if the worker threads modify the real + // values after we read these as those changes can only be + // towards more favorable conditions (less memory in use, + // more in cache). + uint64_t mem_in_use; + uint64_t mem_cached; + struct worker_thread *thr = NULL; // Init to silence warning. + + mythread_sync(coder->mutex) { + mem_in_use = coder->mem_in_use; + mem_cached = coder->mem_cached; + thr = coder->threads_free; + } + + // The maximum amount of memory that can be held by other + // threads and cached buffers while allowing us to start + // decoding the next Block. + const uint64_t mem_max = coder->memlimit_threading + - coder->mem_next_block; + + // If the existing allocations are so large that starting + // to decode this Block might exceed memlimit_threads, + // try to free memory from the output queue cache first. + // + // NOTE: This math assumes the worst case. It's possible + // that the limit wouldn't be exceeded if the existing cached + // allocations are reused. + if (mem_in_use + mem_cached + coder->outq.mem_allocated + > mem_max) { + // Clear the outq cache except leave one buffer in + // the cache if its size is correct. That way we + // don't free and almost immediately reallocate + // an identical buffer. + lzma_outq_clear_cache2(&coder->outq, allocator, + coder->block_options.uncompressed_size); + } + + // If there is at least one worker_thread in the cache and + // the existing allocations are so large that starting to + // decode this Block might exceed memlimit_threads, free + // memory by freeing cached Block decoders. + // + // NOTE: The comparison is different here than above. + // Here we don't care about cached buffers in outq anymore + // and only look at memory actually in use. This is because + // if there is something in outq cache, it's a single buffer + // that can be used as is. We ensured this in the above + // if-block. + uint64_t mem_freed = 0; + if (thr != NULL && mem_in_use + mem_cached + + coder->outq.mem_in_use > mem_max) { + // Don't free the first Block decoder if its memory + // usage isn't greater than what this Block will need. + // Typically the same filter chain is used for all + // Blocks so this way the allocations can be reused + // when get_thread() picks the first worker_thread + // from the cache. + if (thr->mem_filters <= coder->mem_next_filters) + thr = thr->next; + + while (thr != NULL) { + lzma_next_end(&thr->block_decoder, allocator); + mem_freed += thr->mem_filters; + thr->mem_filters = 0; + thr = thr->next; + } + } + + // Update the memory usage counters. Note that coder->mem_* + // may have changed since we read them so we must substract + // or add the changes. + mythread_sync(coder->mutex) { + coder->mem_cached -= mem_freed; + + // Memory needed for the filters and the input buffer. + // The output queue takes care of its own counter so + // we don't touch it here. + // + // NOTE: After this, coder->mem_in_use + + // coder->mem_cached might count the same thing twice. + // If so, this will get corrected in get_thread() when + // a worker_thread is picked from coder->free_threads + // and its memory usage is substracted from mem_cached. + coder->mem_in_use += coder->mem_next_in + + coder->mem_next_filters; + } + + // Allocate memory for the output buffer in the output queue. + lzma_ret ret = lzma_outq_prealloc_buf( + &coder->outq, allocator, + coder->block_options.uncompressed_size); + if (ret != LZMA_OK) { + threads_stop(coder); + return ret; + } + + // Set up coder->thr. + ret = get_thread(coder, allocator); + if (ret != LZMA_OK) { + threads_stop(coder); + return ret; + } + + // The new Block decoder memory usage is already counted in + // coder->mem_in_use. Store it in the thread too. + coder->thr->mem_filters = coder->mem_next_filters; + + // Initialize the Block decoder. + coder->thr->block_options = coder->block_options; + ret = lzma_block_decoder_init( + &coder->thr->block_decoder, allocator, + &coder->thr->block_options); + + // Free the allocated filter options since they are needed + // only to initialize the Block decoder. + lzma_filters_free(coder->filters, allocator); + coder->thr->block_options.filters = NULL; + + // Check if memory usage calculation and Block encoder + // initialization succeeded. + if (ret != LZMA_OK) { + coder->pending_error = ret; + coder->sequence = SEQ_ERROR; + break; + } + + // Allocate the input buffer. + coder->thr->in_size = coder->mem_next_in; + coder->thr->in = lzma_alloc(coder->thr->in_size, allocator); + if (coder->thr->in == NULL) { + threads_stop(coder); + return LZMA_MEM_ERROR; + } + + // Get the preallocated output buffer. + coder->thr->outbuf = lzma_outq_get_buf( + &coder->outq, coder->thr); + + // Start the decoder. + mythread_sync(coder->thr->mutex) { + assert(coder->thr->state == THR_IDLE); + coder->thr->state = THR_RUN; + mythread_cond_signal(&coder->thr->cond); + } + + // Enable output from the thread that holds the oldest output + // buffer in the output queue (if such a thread exists). + mythread_sync(coder->mutex) { + lzma_outq_enable_partial_output(&coder->outq, + &worker_enable_partial_update); + } + + coder->sequence = SEQ_BLOCK_THR_RUN; + } + + // Fall through + + case SEQ_BLOCK_THR_RUN: { + if (action == LZMA_FINISH && coder->fail_fast) { + // We know that we won't get more input and that + // the caller wants fail-fast behavior. If we see + // that we don't have enough input to finish this + // Block, return LZMA_DATA_ERROR immediately. + // See SEQ_BLOCK_HEADER for the error code rationale. + const size_t in_avail = in_size - *in_pos; + const size_t in_needed = coder->thr->in_size + - coder->thr->in_filled; + if (in_avail < in_needed) { + threads_stop(coder); + return LZMA_DATA_ERROR; + } + } + + // Copy input to the worker thread. + size_t cur_in_filled = coder->thr->in_filled; + lzma_bufcpy(in, in_pos, in_size, coder->thr->in, + &cur_in_filled, coder->thr->in_size); + + // Tell the thread how much we copied. + mythread_sync(coder->thr->mutex) { + coder->thr->in_filled = cur_in_filled; + + // NOTE: Most of the time we are copying input faster + // than the thread can decode so most of the time + // calling mythread_cond_signal() is useless but + // we cannot make it conditional because thr->in_pos + // is updated without a mutex. And the overhead should + // be very much negligible anyway. + mythread_cond_signal(&coder->thr->cond); + } + + // Read output from the output queue. Just like in + // SEQ_BLOCK_HEADER, we wait to fill the output buffer + // only if waiting_allowed was set to true in the beginning + // of this function (see the comment there). + return_if_error(read_output_and_wait(coder, allocator, + out, out_pos, out_size, + NULL, waiting_allowed, + &wait_abs, &has_blocked)); + + if (coder->pending_error != LZMA_OK) { + coder->sequence = SEQ_ERROR; + break; + } + + // Return if the input didn't contain the whole Block. + if (coder->thr->in_filled < coder->thr->in_size) { + assert(*in_pos == in_size); + return LZMA_OK; + } + + // The whole Block has been copied to the thread-specific + // buffer. Continue from the next Block Header or Index. + coder->thr = NULL; + coder->sequence = SEQ_BLOCK_HEADER; + break; + } + + case SEQ_BLOCK_DIRECT_INIT: { + // Wait for the threads to finish and that all decoded data + // has been copied to the output. That is, wait until the + // output queue becomes empty. + // + // NOTE: No need to check for coder->pending_error as + // we aren't consuming any input until the queue is empty + // and if there is a pending error, read_output_and_wait() + // will eventually return it before the queue is empty. + return_if_error(read_output_and_wait(coder, allocator, + out, out_pos, out_size, + NULL, true, &wait_abs, &has_blocked)); + if (!lzma_outq_is_empty(&coder->outq)) + return LZMA_OK; + + // Free the cached output buffers. + lzma_outq_clear_cache(&coder->outq, allocator); + + // Get rid of the worker threads, including the coder->threads + // array. + threads_end(coder, allocator); + + // Initialize the Block decoder. + const lzma_ret ret = lzma_block_decoder_init( + &coder->block_decoder, allocator, + &coder->block_options); + + // Free the allocated filter options since they are needed + // only to initialize the Block decoder. + lzma_filters_free(coder->filters, allocator); + coder->block_options.filters = NULL; + + // Check if Block decoder initialization succeeded. + if (ret != LZMA_OK) + return ret; + + // Make the memory usage visible to _memconfig(). + coder->mem_direct_mode = coder->mem_next_filters; + + coder->sequence = SEQ_BLOCK_DIRECT_RUN; + } + + // Fall through + + case SEQ_BLOCK_DIRECT_RUN: { + const size_t in_old = *in_pos; + const size_t out_old = *out_pos; + const lzma_ret ret = coder->block_decoder.code( + coder->block_decoder.coder, allocator, + in, in_pos, in_size, out, out_pos, out_size, + action); + coder->progress_in += *in_pos - in_old; + coder->progress_out += *out_pos - out_old; + + if (ret != LZMA_STREAM_END) + return ret; + + // Block decoded successfully. Add the new size pair to + // the Index hash. + return_if_error(lzma_index_hash_append(coder->index_hash, + lzma_block_unpadded_size( + &coder->block_options), + coder->block_options.uncompressed_size)); + + coder->sequence = SEQ_BLOCK_HEADER; + break; + } + + case SEQ_INDEX_WAIT_OUTPUT: + // Flush the output from all worker threads so that we can + // decode the Index without thinking about threading. + return_if_error(read_output_and_wait(coder, allocator, + out, out_pos, out_size, + NULL, true, &wait_abs, &has_blocked)); + + if (!lzma_outq_is_empty(&coder->outq)) + return LZMA_OK; + + coder->sequence = SEQ_INDEX_DECODE; + + // Fall through + + case SEQ_INDEX_DECODE: { + // If we don't have any input, don't call + // lzma_index_hash_decode() since it would return + // LZMA_BUF_ERROR, which we must not do here. + if (*in_pos >= in_size) + return LZMA_OK; + + // Decode the Index and compare it to the hash calculated + // from the sizes of the Blocks (if any). + const size_t in_old = *in_pos; + const lzma_ret ret = lzma_index_hash_decode(coder->index_hash, + in, in_pos, in_size); + coder->progress_in += *in_pos - in_old; + if (ret != LZMA_STREAM_END) + return ret; + + coder->sequence = SEQ_STREAM_FOOTER; + } + + // Fall through + + case SEQ_STREAM_FOOTER: { + // Copy the Stream Footer to the internal buffer. + const size_t in_old = *in_pos; + lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos, + LZMA_STREAM_HEADER_SIZE); + coder->progress_in += *in_pos - in_old; + + // Return if we didn't get the whole Stream Footer yet. + if (coder->pos < LZMA_STREAM_HEADER_SIZE) + return LZMA_OK; + + coder->pos = 0; + + // Decode the Stream Footer. The decoder gives + // LZMA_FORMAT_ERROR if the magic bytes don't match, + // so convert that return code to LZMA_DATA_ERROR. + lzma_stream_flags footer_flags; + const lzma_ret ret = lzma_stream_footer_decode( + &footer_flags, coder->buffer); + if (ret != LZMA_OK) + return ret == LZMA_FORMAT_ERROR + ? LZMA_DATA_ERROR : ret; + + // Check that Index Size stored in the Stream Footer matches + // the real size of the Index field. + if (lzma_index_hash_size(coder->index_hash) + != footer_flags.backward_size) + return LZMA_DATA_ERROR; + + // Compare that the Stream Flags fields are identical in + // both Stream Header and Stream Footer. + return_if_error(lzma_stream_flags_compare( + &coder->stream_flags, &footer_flags)); + + if (!coder->concatenated) + return LZMA_STREAM_END; + + coder->sequence = SEQ_STREAM_PADDING; + } + + // Fall through + + case SEQ_STREAM_PADDING: + assert(coder->concatenated); + + // Skip over possible Stream Padding. + while (true) { + if (*in_pos >= in_size) { + // Unless LZMA_FINISH was used, we cannot + // know if there's more input coming later. + if (action != LZMA_FINISH) + return LZMA_OK; + + // Stream Padding must be a multiple of + // four bytes. + return coder->pos == 0 + ? LZMA_STREAM_END + : LZMA_DATA_ERROR; + } + + // If the byte is not zero, it probably indicates + // beginning of a new Stream (or the file is corrupt). + if (in[*in_pos] != 0x00) + break; + + ++*in_pos; + ++coder->progress_in; + coder->pos = (coder->pos + 1) & 3; + } + + // Stream Padding must be a multiple of four bytes (empty + // Stream Padding is OK). + if (coder->pos != 0) { + ++*in_pos; + ++coder->progress_in; + return LZMA_DATA_ERROR; + } + + // Prepare to decode the next Stream. + return_if_error(stream_decoder_reset(coder, allocator)); + break; + + case SEQ_ERROR: + if (!coder->fail_fast) { + // Let the application get all data before the point + // where the error was detected. This matches the + // behavior of single-threaded use. + // + // FIXME? Some errors (LZMA_MEM_ERROR) don't get here, + // they are returned immediately. Thus in rare cases + // the output will be less than in the single-threaded + // mode. Maybe this doesn't matter much in practice. + return_if_error(read_output_and_wait(coder, allocator, + out, out_pos, out_size, + NULL, true, &wait_abs, &has_blocked)); + + // We get here only if the error happened in the main + // thread, for example, unsupported Block Header. + if (!lzma_outq_is_empty(&coder->outq)) + return LZMA_OK; + } + + // We only get here if no errors were detected by the worker + // threads. Errors from worker threads would have already been + // returned by the call to read_output_and_wait() above. + return coder->pending_error; + + default: + assert(0); + return LZMA_PROG_ERROR; + } + + // Never reached +} + + +static void +stream_decoder_mt_end(void *coder_ptr, const lzma_allocator *allocator) +{ + struct lzma_stream_coder *coder = coder_ptr; + + threads_end(coder, allocator); + lzma_outq_end(&coder->outq, allocator); + + lzma_next_end(&coder->block_decoder, allocator); + lzma_filters_free(coder->filters, allocator); + lzma_index_hash_end(coder->index_hash, allocator); + + lzma_free(coder, allocator); + return; +} + + +static lzma_check +stream_decoder_mt_get_check(const void *coder_ptr) +{ + const struct lzma_stream_coder *coder = coder_ptr; + return coder->stream_flags.check; +} + + +static lzma_ret +stream_decoder_mt_memconfig(void *coder_ptr, uint64_t *memusage, + uint64_t *old_memlimit, uint64_t new_memlimit) +{ + // NOTE: This function gets/sets memlimit_stop. For now, + // memlimit_threading cannot be modified after initialization. + // + // *memusage will include cached memory too. Excluding cached memory + // would be misleading and it wouldn't help the applications to + // know how much memory is actually needed to decompress the file + // because the higher the number of threads and the memlimits are + // the more memory the decoder may use. + // + // Setting a new limit includes the cached memory too and too low + // limits will be rejected. Alternative could be to free the cached + // memory immediately if that helps to bring the limit down but + // the current way is the simplest. It's unlikely that limit needs + // to be lowered in the middle of a file anyway; the typical reason + // to want a new limit is to increase after LZMA_MEMLIMIT_ERROR + // and even such use isn't common. + struct lzma_stream_coder *coder = coder_ptr; + + mythread_sync(coder->mutex) { + *memusage = coder->mem_direct_mode + + coder->mem_in_use + + coder->mem_cached + + coder->outq.mem_allocated; + } + + // If no filter chains are allocated, *memusage may be zero. + // Always return at least LZMA_MEMUSAGE_BASE. + if (*memusage < LZMA_MEMUSAGE_BASE) + *memusage = LZMA_MEMUSAGE_BASE; + + *old_memlimit = coder->memlimit_stop; + + if (new_memlimit != 0) { + if (new_memlimit < *memusage) + return LZMA_MEMLIMIT_ERROR; + + coder->memlimit_stop = new_memlimit; + } + + return LZMA_OK; +} + + +static void +stream_decoder_mt_get_progress(void *coder_ptr, + uint64_t *progress_in, uint64_t *progress_out) +{ + struct lzma_stream_coder *coder = coder_ptr; + + // Lock coder->mutex to prevent finishing threads from moving their + // progress info from the worker_thread structure to lzma_stream_coder. + mythread_sync(coder->mutex) { + *progress_in = coder->progress_in; + *progress_out = coder->progress_out; + + for (size_t i = 0; i < coder->threads_initialized; ++i) { + mythread_sync(coder->threads[i].mutex) { + *progress_in += coder->threads[i].progress_in; + *progress_out += coder->threads[i] + .progress_out; + } + } + } + + return; +} + + +static lzma_ret +stream_decoder_mt_init(lzma_next_coder *next, const lzma_allocator *allocator, + const lzma_mt *options) +{ + struct lzma_stream_coder *coder; + + if (options->threads == 0 || options->threads > LZMA_THREADS_MAX) + return LZMA_OPTIONS_ERROR; + + if (options->flags & ~LZMA_SUPPORTED_FLAGS) + return LZMA_OPTIONS_ERROR; + + lzma_next_coder_init(&stream_decoder_mt_init, next, allocator); + + coder = next->coder; + if (!coder) { + coder = lzma_alloc(sizeof(struct lzma_stream_coder), allocator); + if (coder == NULL) + return LZMA_MEM_ERROR; + + next->coder = coder; + + if (mythread_mutex_init(&coder->mutex)) { + lzma_free(coder, allocator); + return LZMA_MEM_ERROR; + } + + if (mythread_cond_init(&coder->cond)) { + mythread_mutex_destroy(&coder->mutex); + lzma_free(coder, allocator); + return LZMA_MEM_ERROR; + } + + next->code = &stream_decode_mt; + next->end = &stream_decoder_mt_end; + next->get_check = &stream_decoder_mt_get_check; + next->memconfig = &stream_decoder_mt_memconfig; + next->get_progress = &stream_decoder_mt_get_progress; + + coder->filters[0].id = LZMA_VLI_UNKNOWN; + memzero(&coder->outq, sizeof(coder->outq)); + + coder->block_decoder = LZMA_NEXT_CODER_INIT; + coder->mem_direct_mode = 0; + + coder->index_hash = NULL; + coder->threads = NULL; + coder->threads_free = NULL; + coder->threads_initialized = 0; + } + + // Cleanup old filter chain if one remains after unfinished decoding + // of a previous Stream. + lzma_filters_free(coder->filters, allocator); + + // By allocating threads from scratch we can start memory-usage + // accounting from scratch, too. Changes in filter and block sizes may + // affect number of threads. + // + // FIXME? Reusing should be easy but unlike the single-threaded + // decoder, with some types of input file combinations reusing + // could leave quite a lot of memory allocated but unused (first + // file could allocate a lot, the next files could use fewer + // threads and some of the allocations from the first file would not + // get freed unless memlimit_threading forces us to clear caches). + // + // NOTE: The direct mode decoder isn't freed here if one exists. + // It will be reused or freed as needed in the main loop. + threads_end(coder, allocator); + + // All memusage counters start at 0 (including mem_direct_mode). + // The little extra that is needed for the structs in this file + // get accounted well enough by the filter chain memory usage + // which adds LZMA_MEMUSAGE_BASE for each chain. However, + // stream_decoder_mt_memconfig() has to handle this specially so that + // it will never return less than LZMA_MEMUSAGE_BASE as memory usage. + coder->mem_in_use = 0; + coder->mem_cached = 0; + coder->mem_next_block = 0; + + coder->progress_in = 0; + coder->progress_out = 0; + + coder->sequence = SEQ_STREAM_HEADER; + coder->thread_error = LZMA_OK; + coder->pending_error = LZMA_OK; + coder->thr = NULL; + + coder->timeout = options->timeout; + + coder->memlimit_threading = my_max(1, options->memlimit_threading); + coder->memlimit_stop = my_max(1, options->memlimit_stop); + if (coder->memlimit_threading > coder->memlimit_stop) + coder->memlimit_threading = coder->memlimit_stop; + + coder->tell_no_check = (options->flags & LZMA_TELL_NO_CHECK) != 0; + coder->tell_unsupported_check + = (options->flags & LZMA_TELL_UNSUPPORTED_CHECK) != 0; + coder->tell_any_check = (options->flags & LZMA_TELL_ANY_CHECK) != 0; + coder->ignore_check = (options->flags & LZMA_IGNORE_CHECK) != 0; + coder->concatenated = (options->flags & LZMA_CONCATENATED) != 0; + coder->fail_fast = (options->flags & LZMA_FAIL_FAST) != 0; + + coder->first_stream = true; + coder->out_was_filled = false; + coder->pos = 0; + + coder->threads_max = options->threads; + + return_if_error(lzma_outq_init(&coder->outq, allocator, + coder->threads_max)); + + return stream_decoder_reset(coder, allocator); +} + + +extern LZMA_API(lzma_ret) +lzma_stream_decoder_mt(lzma_stream *strm, const lzma_mt *options) +{ + lzma_next_strm_init(stream_decoder_mt_init, strm, options); + + strm->internal->supported_actions[LZMA_RUN] = true; + strm->internal->supported_actions[LZMA_FINISH] = true; + + return LZMA_OK; +} diff --git a/contrib/libs/lzma/liblzma/common/stream_encoder.c b/contrib/libs/lzma/liblzma/common/stream_encoder.c index 858cba473a..ee92046018 100644 --- a/contrib/libs/lzma/liblzma/common/stream_encoder.c +++ b/contrib/libs/lzma/liblzma/common/stream_encoder.c @@ -219,8 +219,7 @@ stream_encoder_end(void *coder_ptr, const lzma_allocator *allocator) lzma_next_end(&coder->index_encoder, allocator); lzma_index_end(coder->index, allocator); - for (size_t i = 0; coder->filters[i].id != LZMA_VLI_UNKNOWN; ++i) - lzma_free(coder->filters[i].options, allocator); + lzma_filters_free(coder->filters, allocator); lzma_free(coder, allocator); return; @@ -233,6 +232,13 @@ stream_encoder_update(void *coder_ptr, const lzma_allocator *allocator, const lzma_filter *reversed_filters) { lzma_stream_coder *coder = coder_ptr; + lzma_ret ret; + + // Make a copy to a temporary buffer first. This way it is easier + // to keep the encoder state unchanged if an error occurs with + // lzma_filters_copy(). + lzma_filter temp[LZMA_FILTERS_MAX + 1]; + return_if_error(lzma_filters_copy(filters, temp, allocator)); if (coder->sequence <= SEQ_BLOCK_INIT) { // There is no incomplete Block waiting to be finished, @@ -240,31 +246,40 @@ stream_encoder_update(void *coder_ptr, const lzma_allocator *allocator, // trying to initialize the Block encoder with the new // chain. This way we detect if the chain is valid. coder->block_encoder_is_initialized = false; - coder->block_options.filters = (lzma_filter *)(filters); - const lzma_ret ret = block_encoder_init(coder, allocator); + coder->block_options.filters = temp; + ret = block_encoder_init(coder, allocator); coder->block_options.filters = coder->filters; if (ret != LZMA_OK) - return ret; + goto error; coder->block_encoder_is_initialized = true; } else if (coder->sequence <= SEQ_BLOCK_ENCODE) { // We are in the middle of a Block. Try to update only // the filter-specific options. - return_if_error(coder->block_encoder.update( + ret = coder->block_encoder.update( coder->block_encoder.coder, allocator, - filters, reversed_filters)); + filters, reversed_filters); + if (ret != LZMA_OK) + goto error; } else { // Trying to update the filter chain when we are already // encoding Index or Stream Footer. - return LZMA_PROG_ERROR; + ret = LZMA_PROG_ERROR; + goto error; } - // Free the copy of the old chain and make a copy of the new chain. - for (size_t i = 0; coder->filters[i].id != LZMA_VLI_UNKNOWN; ++i) - lzma_free(coder->filters[i].options, allocator); + // Free the options of the old chain. + lzma_filters_free(coder->filters, allocator); + + // Copy the new filter chain in place. + memcpy(coder->filters, temp, sizeof(temp)); + + return LZMA_OK; - return lzma_filters_copy(filters, coder->filters, allocator); +error: + lzma_filters_free(temp, allocator); + return ret; } @@ -319,7 +334,7 @@ stream_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, // Initialize the Block encoder. This way we detect unsupported // filter chains when initializing the Stream encoder instead of - // giving an error after Stream Header has already written out. + // giving an error after Stream Header has already been written out. return stream_encoder_update(coder, allocator, filters, NULL); } diff --git a/contrib/libs/lzma/liblzma/common/stream_encoder_mt.c b/contrib/libs/lzma/liblzma/common/stream_encoder_mt.c index 01e4033975..f4497c10b9 100644 --- a/contrib/libs/lzma/liblzma/common/stream_encoder_mt.c +++ b/contrib/libs/lzma/liblzma/common/stream_encoder_mt.c @@ -85,6 +85,11 @@ struct worker_thread_s { /// Compression options for this Block lzma_block block_options; + /// Filter chain for this thread. By copying the filters array + /// to each thread it is possible to change the filter chain + /// between Blocks using lzma_filters_update(). + lzma_filter filters[LZMA_FILTERS_MAX + 1]; + /// Next structure in the stack of free worker threads. worker_thread *next; @@ -109,9 +114,22 @@ struct lzma_stream_coder_s { /// LZMA_FULL_FLUSH or LZMA_FULL_BARRIER is used earlier. size_t block_size; - /// The filter chain currently in use + /// The filter chain to use for the next Block. + /// This can be updated using lzma_filters_update() + /// after LZMA_FULL_BARRIER or LZMA_FULL_FLUSH. lzma_filter filters[LZMA_FILTERS_MAX + 1]; + /// A copy of filters[] will be put here when attempting to get + /// a new worker thread. This will be copied to a worker thread + /// when a thread becomes free and then this cache is marked as + /// empty by setting [0].id = LZMA_VLI_UNKNOWN. Without this cache + /// the filter options from filters[] would get uselessly copied + /// multiple times (allocated and freed) when waiting for a new free + /// worker thread. + /// + /// This is freed if filters[] is updated via lzma_filters_update(). + lzma_filter filters_cache[LZMA_FILTERS_MAX + 1]; + /// Index to hold sizes of the Blocks lzma_index *index; @@ -133,6 +151,9 @@ struct lzma_stream_coder_s { /// Output buffer queue for compressed data lzma_outq outq; + /// How much memory to allocate for each lzma_outbuf.buf + size_t outbuf_alloc_size; + /// Maximum wait time if cannot use all the input and cannot /// fill the output buffer. This is in milliseconds. @@ -196,7 +217,7 @@ worker_error(worker_thread *thr, lzma_ret ret) static worker_state -worker_encode(worker_thread *thr, worker_state state) +worker_encode(worker_thread *thr, size_t *out_pos, worker_state state) { assert(thr->progress_in == 0); assert(thr->progress_out == 0); @@ -205,12 +226,9 @@ worker_encode(worker_thread *thr, worker_state state) thr->block_options = (lzma_block){ .version = 0, .check = thr->coder->stream_flags.check, - .compressed_size = thr->coder->outq.buf_size_max, + .compressed_size = thr->outbuf->allocated, .uncompressed_size = thr->coder->block_size, - - // TODO: To allow changing the filter chain, the filters - // array must be copied to each worker_thread. - .filters = thr->coder->filters, + .filters = thr->filters, }; // Calculate maximum size of the Block Header. This amount is @@ -234,12 +252,12 @@ worker_encode(worker_thread *thr, worker_state state) size_t in_pos = 0; size_t in_size = 0; - thr->outbuf->size = thr->block_options.header_size; - const size_t out_size = thr->coder->outq.buf_size_max; + *out_pos = thr->block_options.header_size; + const size_t out_size = thr->outbuf->allocated; do { mythread_sync(thr->mutex) { - // Store in_pos and out_pos into *thr so that + // Store in_pos and *out_pos into *thr so that // an application may read them via // lzma_get_progress() to get progress information. // @@ -247,7 +265,7 @@ worker_encode(worker_thread *thr, worker_state state) // finishes. Instead, the final values are taken // later from thr->outbuf. thr->progress_in = in_pos; - thr->progress_out = thr->outbuf->size; + thr->progress_out = *out_pos; while (in_size == thr->in_size && thr->state == THR_RUN) @@ -277,8 +295,8 @@ worker_encode(worker_thread *thr, worker_state state) ret = thr->block_encoder.code( thr->block_encoder.coder, thr->allocator, thr->in, &in_pos, in_limit, thr->outbuf->buf, - &thr->outbuf->size, out_size, action); - } while (ret == LZMA_OK && thr->outbuf->size < out_size); + out_pos, out_size, action); + } while (ret == LZMA_OK && *out_pos < out_size); switch (ret) { case LZMA_STREAM_END: @@ -313,10 +331,10 @@ worker_encode(worker_thread *thr, worker_state state) return state; // Do the encoding. This takes care of the Block Header too. - thr->outbuf->size = 0; + *out_pos = 0; ret = lzma_block_uncomp_encode(&thr->block_options, thr->in, in_size, thr->outbuf->buf, - &thr->outbuf->size, out_size); + out_pos, out_size); // It shouldn't fail. if (ret != LZMA_OK) { @@ -367,11 +385,13 @@ worker_start(void *thr_ptr) } } + size_t out_pos = 0; + assert(state != THR_IDLE); assert(state != THR_STOP); if (state <= THR_FINISH) - state = worker_encode(thr, state); + state = worker_encode(thr, &out_pos, state); if (state == THR_EXIT) break; @@ -387,14 +407,17 @@ worker_start(void *thr_ptr) } mythread_sync(thr->coder->mutex) { - // Mark the output buffer as finished if - // no errors occurred. - thr->outbuf->finished = state == THR_FINISH; + // If no errors occurred, make the encoded data + // available to be copied out. + if (state == THR_FINISH) { + thr->outbuf->pos = out_pos; + thr->outbuf->finished = true; + } // Update the main progress info. thr->coder->progress_in += thr->outbuf->uncompressed_size; - thr->coder->progress_out += thr->outbuf->size; + thr->coder->progress_out += out_pos; thr->progress_in = 0; thr->progress_out = 0; @@ -407,6 +430,8 @@ worker_start(void *thr_ptr) } // Exiting, free the resources. + lzma_filters_free(thr->filters, thr->allocator); + mythread_mutex_destroy(&thr->mutex); mythread_cond_destroy(&thr->cond); @@ -490,6 +515,7 @@ initialize_new_thread(lzma_stream_coder *coder, thr->progress_in = 0; thr->progress_out = 0; thr->block_encoder = LZMA_NEXT_CODER_INIT; + thr->filters[0].id = LZMA_VLI_UNKNOWN; if (mythread_create(&thr->thread_id, &worker_start, thr)) goto error_thread; @@ -519,6 +545,18 @@ get_thread(lzma_stream_coder *coder, const lzma_allocator *allocator) if (!lzma_outq_has_buf(&coder->outq)) return LZMA_OK; + // That's also true if we cannot allocate memory for the output + // buffer in the output queue. + return_if_error(lzma_outq_prealloc_buf(&coder->outq, allocator, + coder->outbuf_alloc_size)); + + // Make a thread-specific copy of the filter chain. Put it in + // the cache array first so that if we cannot get a new thread yet, + // the allocation is ready when we try again. + if (coder->filters_cache[0].id == LZMA_VLI_UNKNOWN) + return_if_error(lzma_filters_copy( + coder->filters, coder->filters_cache, allocator)); + // If there is a free structure on the stack, use it. mythread_sync(coder->mutex) { if (coder->threads_free != NULL) { @@ -541,7 +579,16 @@ get_thread(lzma_stream_coder *coder, const lzma_allocator *allocator) mythread_sync(coder->thr->mutex) { coder->thr->state = THR_RUN; coder->thr->in_size = 0; - coder->thr->outbuf = lzma_outq_get_buf(&coder->outq); + coder->thr->outbuf = lzma_outq_get_buf(&coder->outq, NULL); + + // Free the old thread-specific filter options and replace + // them with the already-allocated new options from + // coder->filters_cache[]. Then mark the cache as empty. + lzma_filters_free(coder->thr->filters, allocator); + memcpy(coder->thr->filters, coder->filters_cache, + sizeof(coder->filters_cache)); + coder->filters_cache[0].id = LZMA_VLI_UNKNOWN; + mythread_cond_signal(&coder->thr->cond); } @@ -627,9 +674,13 @@ wait_for_work(lzma_stream_coder *coder, mythread_condtime *wait_abs, // to true here and calculate the absolute time when // we must return if there's nothing to do. // - // The idea of *has_blocked is to avoid unneeded calls - // to mythread_condtime_set(), which may do a syscall - // depending on the operating system. + // This way if we block multiple times for short moments + // less than "timeout" milliseconds, we will return once + // "timeout" amount of time has passed since the *first* + // blocking occurred. If the absolute time was calculated + // again every time we block, "timeout" would effectively + // be meaningless if we never consecutively block longer + // than "timeout" ms. *has_blocked = true; mythread_condtime_set(wait_abs, &coder->cond, coder->timeout); } @@ -704,7 +755,7 @@ stream_encode_mt(void *coder_ptr, const lzma_allocator *allocator, } // Try to read compressed data to out[]. - ret = lzma_outq_read(&coder->outq, + ret = lzma_outq_read(&coder->outq, allocator, out, out_pos, out_size, &unpadded_size, &uncompressed_size); @@ -715,6 +766,10 @@ stream_encode_mt(void *coder_ptr, const lzma_allocator *allocator, ret = lzma_index_append(coder->index, allocator, unpadded_size, uncompressed_size); + if (ret != LZMA_OK) { + threads_stop(coder, false); + return ret; + } // If we didn't fill the output buffer yet, // try to read more data. Maybe the next @@ -724,8 +779,7 @@ stream_encode_mt(void *coder_ptr, const lzma_allocator *allocator, } if (ret != LZMA_OK) { - // coder->thread_error was set or - // lzma_index_append() failed. + // coder->thread_error was set. threads_stop(coder, false); return ret; } @@ -846,8 +900,8 @@ stream_encoder_mt_end(void *coder_ptr, const lzma_allocator *allocator) threads_end(coder, allocator); lzma_outq_end(&coder->outq, allocator); - for (size_t i = 0; coder->filters[i].id != LZMA_VLI_UNKNOWN; ++i) - lzma_free(coder->filters[i].options, allocator); + lzma_filters_free(coder->filters, allocator); + lzma_filters_free(coder->filters_cache, allocator); lzma_next_end(&coder->index_encoder, allocator); lzma_index_end(coder->index, allocator); @@ -860,6 +914,45 @@ stream_encoder_mt_end(void *coder_ptr, const lzma_allocator *allocator) } +static lzma_ret +stream_encoder_mt_update(void *coder_ptr, const lzma_allocator *allocator, + const lzma_filter *filters, + const lzma_filter *reversed_filters + lzma_attribute((__unused__))) +{ + lzma_stream_coder *coder = coder_ptr; + + // Applications shouldn't attempt to change the options when + // we are already encoding the Index or Stream Footer. + if (coder->sequence > SEQ_BLOCK) + return LZMA_PROG_ERROR; + + // For now the threaded encoder doesn't support changing + // the options in the middle of a Block. + if (coder->thr != NULL) + return LZMA_PROG_ERROR; + + // Check if the filter chain seems mostly valid. See the comment + // in stream_encoder_mt_init(). + if (lzma_raw_encoder_memusage(filters) == UINT64_MAX) + return LZMA_OPTIONS_ERROR; + + // Make a copy to a temporary buffer first. This way the encoder + // state stays unchanged if an error occurs in lzma_filters_copy(). + lzma_filter temp[LZMA_FILTERS_MAX + 1]; + return_if_error(lzma_filters_copy(filters, temp, allocator)); + + // Free the options of the old chain as well as the cache. + lzma_filters_free(coder->filters, allocator); + lzma_filters_free(coder->filters_cache, allocator); + + // Copy the new filter chain in place. + memcpy(coder->filters, temp, sizeof(temp)); + + return LZMA_OK; +} + + /// Options handling for lzma_stream_encoder_mt_init() and /// lzma_stream_encoder_mt_memusage() static lzma_ret @@ -951,14 +1044,16 @@ stream_encoder_mt_init(lzma_next_coder *next, const lzma_allocator *allocator, &block_size, &outbuf_size_max)); #if SIZE_MAX < UINT64_MAX - if (block_size > SIZE_MAX) + if (block_size > SIZE_MAX || outbuf_size_max > SIZE_MAX) return LZMA_MEM_ERROR; #endif // Validate the filter chain so that we can give an error in this // function instead of delaying it to the first call to lzma_code(). // The memory usage calculation verifies the filter chain as - // a side effect so we take advantage of that. + // a side effect so we take advantage of that. It's not a perfect + // check though as raw encoder allows LZMA1 too but such problems + // will be caught eventually with Block Header encoder. if (lzma_raw_encoder_memusage(filters) == UINT64_MAX) return LZMA_OPTIONS_ERROR; @@ -998,9 +1093,10 @@ stream_encoder_mt_init(lzma_next_coder *next, const lzma_allocator *allocator, next->code = &stream_encode_mt; next->end = &stream_encoder_mt_end; next->get_progress = &get_progress; -// next->update = &stream_encoder_mt_update; + next->update = &stream_encoder_mt_update; coder->filters[0].id = LZMA_VLI_UNKNOWN; + coder->filters_cache[0].id = LZMA_VLI_UNKNOWN; coder->index_encoder = LZMA_NEXT_CODER_INIT; coder->index = NULL; memzero(&coder->outq, sizeof(coder->outq)); @@ -1012,6 +1108,7 @@ stream_encoder_mt_init(lzma_next_coder *next, const lzma_allocator *allocator, // Basic initializations coder->sequence = SEQ_STREAM_HEADER; coder->block_size = (size_t)(block_size); + coder->outbuf_alloc_size = (size_t)(outbuf_size_max); coder->thread_error = LZMA_OK; coder->thr = NULL; @@ -1041,15 +1138,16 @@ stream_encoder_mt_init(lzma_next_coder *next, const lzma_allocator *allocator, // Output queue return_if_error(lzma_outq_init(&coder->outq, allocator, - outbuf_size_max, options->threads)); + options->threads)); // Timeout coder->timeout = options->timeout; - // Free the old filter chain and copy the new one. - for (size_t i = 0; coder->filters[i].id != LZMA_VLI_UNKNOWN; ++i) - lzma_free(coder->filters[i].options, allocator); + // Free the old filter chain and the cache. + lzma_filters_free(coder->filters, allocator); + lzma_filters_free(coder->filters_cache, allocator); + // Copy the new filter chain. return_if_error(lzma_filters_copy( filters, coder->filters, allocator)); @@ -1075,6 +1173,31 @@ stream_encoder_mt_init(lzma_next_coder *next, const lzma_allocator *allocator, } +#ifdef HAVE_SYMBOL_VERSIONS_LINUX +// These are for compatibility with binaries linked against liblzma that +// has been patched with xz-5.2.2-compat-libs.patch from RHEL/CentOS 7. +// Actually that patch didn't create lzma_stream_encoder_mt@XZ_5.2.2 +// but it has been added here anyway since someone might misread the +// RHEL patch and think both @XZ_5.1.2alpha and @XZ_5.2.2 exist. +LZMA_SYMVER_API("lzma_stream_encoder_mt@XZ_5.1.2alpha", + lzma_ret, lzma_stream_encoder_mt_512a)( + lzma_stream *strm, const lzma_mt *options) + lzma_nothrow lzma_attr_warn_unused_result + __attribute__((__alias__("lzma_stream_encoder_mt_52"))); + +LZMA_SYMVER_API("lzma_stream_encoder_mt@XZ_5.2.2", + lzma_ret, lzma_stream_encoder_mt_522)( + lzma_stream *strm, const lzma_mt *options) + lzma_nothrow lzma_attr_warn_unused_result + __attribute__((__alias__("lzma_stream_encoder_mt_52"))); + +LZMA_SYMVER_API("lzma_stream_encoder_mt@@XZ_5.2", + lzma_ret, lzma_stream_encoder_mt_52)( + lzma_stream *strm, const lzma_mt *options) + lzma_nothrow lzma_attr_warn_unused_result; + +#define lzma_stream_encoder_mt lzma_stream_encoder_mt_52 +#endif extern LZMA_API(lzma_ret) lzma_stream_encoder_mt(lzma_stream *strm, const lzma_mt *options) { @@ -1090,6 +1213,23 @@ lzma_stream_encoder_mt(lzma_stream *strm, const lzma_mt *options) } +#ifdef HAVE_SYMBOL_VERSIONS_LINUX +LZMA_SYMVER_API("lzma_stream_encoder_mt_memusage@XZ_5.1.2alpha", + uint64_t, lzma_stream_encoder_mt_memusage_512a)( + const lzma_mt *options) lzma_nothrow lzma_attr_pure + __attribute__((__alias__("lzma_stream_encoder_mt_memusage_52"))); + +LZMA_SYMVER_API("lzma_stream_encoder_mt_memusage@XZ_5.2.2", + uint64_t, lzma_stream_encoder_mt_memusage_522)( + const lzma_mt *options) lzma_nothrow lzma_attr_pure + __attribute__((__alias__("lzma_stream_encoder_mt_memusage_52"))); + +LZMA_SYMVER_API("lzma_stream_encoder_mt_memusage@@XZ_5.2", + uint64_t, lzma_stream_encoder_mt_memusage_52)( + const lzma_mt *options) lzma_nothrow lzma_attr_pure; + +#define lzma_stream_encoder_mt_memusage lzma_stream_encoder_mt_memusage_52 +#endif // This function name is a monster but it's consistent with the older // monster names. :-( 31 chars is the max that C99 requires so in that // sense it's not too long. ;-) diff --git a/contrib/libs/lzma/liblzma/common/stream_flags_decoder.c b/contrib/libs/lzma/liblzma/common/stream_flags_decoder.c index 4e43e359e1..b8d263ba44 100644 --- a/contrib/libs/lzma/liblzma/common/stream_flags_decoder.c +++ b/contrib/libs/lzma/liblzma/common/stream_flags_decoder.c @@ -39,8 +39,11 @@ lzma_stream_header_decode(lzma_stream_flags *options, const uint8_t *in) const uint32_t crc = lzma_crc32(in + sizeof(lzma_header_magic), LZMA_STREAM_FLAGS_SIZE, 0); if (crc != read32le(in + sizeof(lzma_header_magic) - + LZMA_STREAM_FLAGS_SIZE)) + + LZMA_STREAM_FLAGS_SIZE)) { +#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION return LZMA_DATA_ERROR; +#endif + } // Stream Flags if (stream_flags_decode(options, in + sizeof(lzma_header_magic))) @@ -67,8 +70,11 @@ lzma_stream_footer_decode(lzma_stream_flags *options, const uint8_t *in) // CRC32 const uint32_t crc = lzma_crc32(in + sizeof(uint32_t), sizeof(uint32_t) + LZMA_STREAM_FLAGS_SIZE, 0); - if (crc != read32le(in)) + if (crc != read32le(in)) { +#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION return LZMA_DATA_ERROR; +#endif + } // Stream Flags if (stream_flags_decode(options, in + sizeof(uint32_t) * 2)) diff --git a/contrib/libs/lzma/liblzma/common/string_conversion.c b/contrib/libs/lzma/liblzma/common/string_conversion.c new file mode 100644 index 0000000000..53fdff2a62 --- /dev/null +++ b/contrib/libs/lzma/liblzma/common/string_conversion.c @@ -0,0 +1,1317 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file string_conversion.c +/// \brief Conversion of strings to filter chain and vice versa +// +// Author: Lasse Collin +// +// This file has been put into the public domain. +// You can do whatever you want with this file. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "filter_common.h" + + +///////////////////// +// String building // +///////////////////// + +/// How much memory to allocate for strings. For now, no realloc is used +/// so this needs to be big enough even though there of course is +/// an overflow check still. +/// +/// FIXME? Using a fixed size is wasteful if the application doesn't free +/// the string fairly quickly but this can be improved later if needed. +#define STR_ALLOC_SIZE 800 + + +typedef struct { + char *buf; + size_t pos; +} lzma_str; + + +static lzma_ret +str_init(lzma_str *str, const lzma_allocator *allocator) +{ + str->buf = lzma_alloc(STR_ALLOC_SIZE, allocator); + if (str->buf == NULL) + return LZMA_MEM_ERROR; + + str->pos = 0; + return LZMA_OK; +} + + +static void +str_free(lzma_str *str, const lzma_allocator *allocator) +{ + lzma_free(str->buf, allocator); + return; +} + + +static bool +str_is_full(const lzma_str *str) +{ + return str->pos == STR_ALLOC_SIZE - 1; +} + + +static lzma_ret +str_finish(char **dest, lzma_str *str, const lzma_allocator *allocator) +{ + if (str_is_full(str)) { + // The preallocated buffer was too small. + // This shouldn't happen as STR_ALLOC_SIZE should + // be adjusted if new filters are added. + lzma_free(str->buf, allocator); + *dest = NULL; + assert(0); + return LZMA_PROG_ERROR; + } + + str->buf[str->pos] = '\0'; + *dest = str->buf; + return LZMA_OK; +} + + +static void +str_append_str(lzma_str *str, const char *s) +{ + const size_t len = strlen(s); + const size_t limit = STR_ALLOC_SIZE - 1 - str->pos; + const size_t copy_size = my_min(len, limit); + + memcpy(str->buf + str->pos, s, copy_size); + str->pos += copy_size; + return; +} + + +static void +str_append_u32(lzma_str *str, uint32_t v, bool use_byte_suffix) +{ + if (v == 0) { + str_append_str(str, "0"); + } else { + // NOTE: Don't use plain "B" because xz and the parser in this + // file don't support it and at glance it may look like 8 + // (there cannot be a space before the suffix). + static const char suffixes[4][4] = { "", "KiB", "MiB", "GiB" }; + + size_t suf = 0; + if (use_byte_suffix) { + while ((v & 1023) == 0 + && suf < ARRAY_SIZE(suffixes) - 1) { + v >>= 10; + ++suf; + } + } + + // UINT32_MAX in base 10 would need 10 + 1 bytes. Remember + // that initializing to "" initializes all elements to + // zero so '\0'-termination gets handled by this. + char buf[16] = ""; + size_t pos = sizeof(buf) - 1; + + do { + buf[--pos] = '0' + (v % 10); + v /= 10; + } while (v != 0); + + str_append_str(str, buf + pos); + str_append_str(str, suffixes[suf]); + } + + return; +} + + +////////////////////////////////////////////// +// Parsing and stringification declarations // +////////////////////////////////////////////// + +/// Maximum length for filter and option names. +/// 11 chars + terminating '\0' + sizeof(uint32_t) = 16 bytes +#define NAME_LEN_MAX 11 + + +/// For option_map.flags: Use .u.map to do convert the input value +/// to an integer. Without this flag, .u.range.{min,max} are used +/// as the allowed range for the integer. +#define OPTMAP_USE_NAME_VALUE_MAP 0x01 + +/// For option_map.flags: Allow KiB/MiB/GiB in input string and use them in +/// the stringified output if the value is an exact multiple of these. +/// This is used e.g. for LZMA1/2 dictionary size. +#define OPTMAP_USE_BYTE_SUFFIX 0x02 + +/// For option_map.flags: If the integer value is zero then this option +/// won't be included in the stringified output. It's used e.g. for +/// BCJ filter start offset which usually is zero. +#define OPTMAP_NO_STRFY_ZERO 0x04 + +/// Possible values for option_map.type. Since OPTMAP_TYPE_UINT32 is 0, +/// it doesn't need to be specified in the initializers as it is +/// the implicit value. +enum { + OPTMAP_TYPE_UINT32, + OPTMAP_TYPE_LZMA_MODE, + OPTMAP_TYPE_LZMA_MATCH_FINDER, + OPTMAP_TYPE_LZMA_PRESET, +}; + + +/// This is for mapping string values in options to integers. +/// The last element of an array must have "" as the name. +/// It's used e.g. for match finder names in LZMA1/2. +typedef struct { + const char name[NAME_LEN_MAX + 1]; + const uint32_t value; +} name_value_map; + + +/// Each filter that has options needs an array of option_map structures. +/// The array doesn't need to be terminated as the functions take the +/// length of the array as an argument. +/// +/// When converting a string to filter options structure, option values +/// will be handled in a few different ways: +/// +/// (1) If .type equals OPTMAP_TYPE_LZMA_PRESET then LZMA1/2 preset string +/// is handled specially. +/// +/// (2) If .flags has OPTMAP_USE_NAME_VALUE_MAP set then the string is +/// converted to an integer using the name_value_map pointed by .u.map. +/// The last element in .u.map must have .name = "" as the terminator. +/// +/// (3) Otherwise the string is treated as a non-negative unsigned decimal +/// integer which must be in the range set in .u.range. If .flags has +/// OPTMAP_USE_BYTE_SUFFIX then KiB, MiB, and GiB suffixes are allowed. +/// +/// The integer value from (2) or (3) is then stored to filter_options +/// at the offset specified in .offset using the type specified in .type +/// (default is uint32_t). +/// +/// Stringifying a filter is done by processing a given number of options +/// in oder from the beginning of an option_map array. The integer is +/// read from filter_options at .offset using the type from .type. +/// +/// If the integer is zero and .flags has OPTMAP_NO_STRFY_ZERO then the +/// option is skipped. +/// +/// If .flags has OPTMAP_USE_NAME_VALUE_MAP set then .u.map will be used +/// to convert the option to a string. If the map doesn't contain a string +/// for the integer value then "UNKNOWN" is used. +/// +/// If .flags doesn't have OPTMAP_USE_NAME_VALUE_MAP set then the integer is +/// converted to a decimal value. If OPTMAP_USE_BYTE_SUFFIX is used then KiB, +/// MiB, or GiB suffix is used if the value is an exact multiple of these. +/// Plain "B" suffix is never used. +typedef struct { + char name[NAME_LEN_MAX + 1]; + uint8_t type; + uint8_t flags; + uint16_t offset; + + union { + struct { + uint32_t min; + uint32_t max; + } range; + + const name_value_map *map; + } u; +} option_map; + + +static const char *parse_options(const char **const str, const char *str_end, + void *filter_options, + const option_map *const optmap, const size_t optmap_size); + + +///////// +// BCJ // +///////// + +#if defined(HAVE_ENCODER_X86) \ + || defined(HAVE_DECODER_X86) \ + || defined(HAVE_ENCODER_ARM) \ + || defined(HAVE_DECODER_ARM) \ + || defined(HAVE_ENCODER_ARMTHUMB) \ + || defined(HAVE_DECODER_ARMTHUMB) \ + || defined(HAVE_ENCODER_ARM64) \ + || defined(HAVE_DECODER_ARM64) \ + || defined(HAVE_ENCODER_POWERPC) \ + || defined(HAVE_DECODER_POWERPC) \ + || defined(HAVE_ENCODER_IA64) \ + || defined(HAVE_DECODER_IA64) \ + || defined(HAVE_ENCODER_SPARC) \ + || defined(HAVE_DECODER_SPARC) +static const option_map bcj_optmap[] = { + { + .name = "start", + .flags = OPTMAP_NO_STRFY_ZERO | OPTMAP_USE_BYTE_SUFFIX, + .offset = offsetof(lzma_options_bcj, start_offset), + .u.range.min = 0, + .u.range.max = UINT32_MAX, + } +}; + + +static const char * +parse_bcj(const char **const str, const char *str_end, void *filter_options) +{ + // filter_options was zeroed on allocation and that is enough + // for the default value. + return parse_options(str, str_end, filter_options, + bcj_optmap, ARRAY_SIZE(bcj_optmap)); +} +#endif + + +/////////// +// Delta // +/////////// + +#if defined(HAVE_ENCODER_DELTA) || defined(HAVE_DECODER_DELTA) +static const option_map delta_optmap[] = { + { + .name = "dist", + .offset = offsetof(lzma_options_delta, dist), + .u.range.min = LZMA_DELTA_DIST_MIN, + .u.range.max = LZMA_DELTA_DIST_MAX, + } +}; + + +static const char * +parse_delta(const char **const str, const char *str_end, void *filter_options) +{ + lzma_options_delta *opts = filter_options; + opts->type = LZMA_DELTA_TYPE_BYTE; + opts->dist = LZMA_DELTA_DIST_MIN; + + return parse_options(str, str_end, filter_options, + delta_optmap, ARRAY_SIZE(delta_optmap)); +} +#endif + + +/////////////////// +// LZMA1 & LZMA2 // +/////////////////// + +/// Help string for presets +#define LZMA12_PRESET_STR "0-9[e]" + + +static const char * +parse_lzma12_preset(const char **const str, const char *str_end, + uint32_t *preset) +{ + assert(*str < str_end); + *preset = (uint32_t)(**str - '0'); + + // NOTE: Remember to update LZMA_PRESET_STR if this is modified! + while (++*str < str_end) { + switch (**str) { + case 'e': + *preset |= LZMA_PRESET_EXTREME; + break; + + default: + return "Unsupported preset flag"; + } + } + + return NULL; +} + + +static const char * +set_lzma12_preset(const char **const str, const char *str_end, + void *filter_options) +{ + uint32_t preset; + const char *errmsg = parse_lzma12_preset(str, str_end, &preset); + if (errmsg != NULL) + return errmsg; + + lzma_options_lzma *opts = filter_options; + if (lzma_lzma_preset(opts, preset)) + return "Unsupported preset"; + + return NULL; +} + + +static const name_value_map lzma12_mode_map[] = { + { "fast", LZMA_MODE_FAST }, + { "normal", LZMA_MODE_NORMAL }, + { "", 0 } +}; + + +static const name_value_map lzma12_mf_map[] = { + { "hc3", LZMA_MF_HC3 }, + { "hc4", LZMA_MF_HC4 }, + { "bt2", LZMA_MF_BT2 }, + { "bt3", LZMA_MF_BT3 }, + { "bt4", LZMA_MF_BT4 }, + { "", 0 } +}; + + +static const option_map lzma12_optmap[] = { + { + .name = "preset", + .type = OPTMAP_TYPE_LZMA_PRESET, + }, { + .name = "dict", + .flags = OPTMAP_USE_BYTE_SUFFIX, + .offset = offsetof(lzma_options_lzma, dict_size), + .u.range.min = LZMA_DICT_SIZE_MIN, + // FIXME? The max is really max for encoding but decoding + // would allow 4 GiB - 1 B. + .u.range.max = (UINT32_C(1) << 30) + (UINT32_C(1) << 29), + }, { + .name = "lc", + .offset = offsetof(lzma_options_lzma, lc), + .u.range.min = LZMA_LCLP_MIN, + .u.range.max = LZMA_LCLP_MAX, + }, { + .name = "lp", + .offset = offsetof(lzma_options_lzma, lp), + .u.range.min = LZMA_LCLP_MIN, + .u.range.max = LZMA_LCLP_MAX, + }, { + .name = "pb", + .offset = offsetof(lzma_options_lzma, pb), + .u.range.min = LZMA_PB_MIN, + .u.range.max = LZMA_PB_MAX, + }, { + .name = "mode", + .type = OPTMAP_TYPE_LZMA_MODE, + .flags = OPTMAP_USE_NAME_VALUE_MAP, + .offset = offsetof(lzma_options_lzma, mode), + .u.map = lzma12_mode_map, + }, { + .name = "nice", + .offset = offsetof(lzma_options_lzma, nice_len), + .u.range.min = 2, + .u.range.max = 273, + }, { + .name = "mf", + .type = OPTMAP_TYPE_LZMA_MATCH_FINDER, + .flags = OPTMAP_USE_NAME_VALUE_MAP, + .offset = offsetof(lzma_options_lzma, mf), + .u.map = lzma12_mf_map, + }, { + .name = "depth", + .offset = offsetof(lzma_options_lzma, depth), + .u.range.min = 0, + .u.range.max = UINT32_MAX, + } +}; + + +static const char * +parse_lzma12(const char **const str, const char *str_end, void *filter_options) +{ + lzma_options_lzma *opts = filter_options; + + // It cannot fail. + const bool preset_ret = lzma_lzma_preset(opts, LZMA_PRESET_DEFAULT); + assert(!preset_ret); + (void)preset_ret; + + const char *errmsg = parse_options(str, str_end, filter_options, + lzma12_optmap, ARRAY_SIZE(lzma12_optmap)); + if (errmsg != NULL) + return errmsg; + + if (opts->lc + opts->lp > LZMA_LCLP_MAX) + return "The sum of lc and lp must not exceed 4"; + + return NULL; +} + + +///////////////////////////////////////// +// Generic parsing and stringification // +///////////////////////////////////////// + +static const struct { + /// Name of the filter + char name[NAME_LEN_MAX + 1]; + + /// For lzma_str_to_filters: + /// Size of the filter-specific options structure. + uint32_t opts_size; + + /// Filter ID + lzma_vli id; + + /// For lzma_str_to_filters: + /// Function to parse the filter-specific options. The filter_options + /// will already have been allocated using lzma_alloc_zero(). + const char *(*parse)(const char **str, const char *str_end, + void *filter_options); + + /// For lzma_str_from_filters: + /// If the flag LZMA_STR_ENCODER is used then the first + /// strfy_encoder elements of optmap are stringified. + /// With LZMA_STR_DECODER strfy_decoder is used. + /// Currently encoders use all flags that decoders do but if + /// that changes then this needs to be changed too, for example, + /// add a new OPTMAP flag to skip printing some decoder-only flags. + const option_map *optmap; + uint8_t strfy_encoder; + uint8_t strfy_decoder; + + /// For lzma_str_from_filters: + /// If true, lzma_filter.options is allowed to be NULL. In that case, + /// only the filter name is printed without any options. + bool allow_null; + +} filter_name_map[] = { +#if defined (HAVE_ENCODER_LZMA1) || defined(HAVE_DECODER_LZMA1) + { "lzma1", sizeof(lzma_options_lzma), LZMA_FILTER_LZMA1, + &parse_lzma12, lzma12_optmap, 9, 5, false }, +#endif + +#if defined(HAVE_ENCODER_LZMA2) || defined(HAVE_DECODER_LZMA2) + { "lzma2", sizeof(lzma_options_lzma), LZMA_FILTER_LZMA2, + &parse_lzma12, lzma12_optmap, 9, 2, false }, +#endif + +#if defined(HAVE_ENCODER_X86) || defined(HAVE_DECODER_X86) + { "x86", sizeof(lzma_options_bcj), LZMA_FILTER_X86, + &parse_bcj, bcj_optmap, 1, 1, true }, +#endif + +#if defined(HAVE_ENCODER_ARM) || defined(HAVE_DECODER_ARM) + { "arm", sizeof(lzma_options_bcj), LZMA_FILTER_ARM, + &parse_bcj, bcj_optmap, 1, 1, true }, +#endif + +#if defined(HAVE_ENCODER_ARMTHUMB) || defined(HAVE_DECODER_ARMTHUMB) + { "armthumb", sizeof(lzma_options_bcj), LZMA_FILTER_ARMTHUMB, + &parse_bcj, bcj_optmap, 1, 1, true }, +#endif + +#if defined(HAVE_ENCODER_ARM64) || defined(HAVE_DECODER_ARM64) + { "arm64", sizeof(lzma_options_bcj), LZMA_FILTER_ARM64, + &parse_bcj, bcj_optmap, 1, 1, true }, +#endif + +#if defined(HAVE_ENCODER_POWERPC) || defined(HAVE_DECODER_POWERPC) + { "powerpc", sizeof(lzma_options_bcj), LZMA_FILTER_POWERPC, + &parse_bcj, bcj_optmap, 1, 1, true }, +#endif + +#if defined(HAVE_ENCODER_IA64) || defined(HAVE_DECODER_IA64) + { "ia64", sizeof(lzma_options_bcj), LZMA_FILTER_IA64, + &parse_bcj, bcj_optmap, 1, 1, true }, +#endif + +#if defined(HAVE_ENCODER_SPARC) || defined(HAVE_DECODER_SPARC) + { "sparc", sizeof(lzma_options_bcj), LZMA_FILTER_SPARC, + &parse_bcj, bcj_optmap, 1, 1, true }, +#endif + +#if defined(HAVE_ENCODER_DELTA) || defined(HAVE_DECODER_DELTA) + { "delta", sizeof(lzma_options_delta), LZMA_FILTER_DELTA, + &parse_delta, delta_optmap, 1, 1, false }, +#endif +}; + + +/// Decodes options from a string for one filter (name1=value1,name2=value2). +/// Caller must have allocated memory for filter_options already and set +/// the initial default values. This is called from the filter-specific +/// parse_* functions. +/// +/// The input string starts at *str and the address in str_end is the first +/// char that is not part of the string anymore. So no '\0' terminator is +/// used. *str is advanced everytime something has been decoded successfully. +static const char * +parse_options(const char **const str, const char *str_end, + void *filter_options, + const option_map *const optmap, const size_t optmap_size) +{ + while (*str < str_end && **str != '\0') { + // Each option is of the form name=value. + // Commas (',') separate options. Extra commas are ignored. + // Ignoring extra commas makes it simpler if an optional + // option stored in a shell variable which can be empty. + if (**str == ',') { + ++*str; + continue; + } + + // Find where the next name=value ends. + const size_t str_len = (size_t)(str_end - *str); + const char *name_eq_value_end = memchr(*str, ',', str_len); + if (name_eq_value_end == NULL) + name_eq_value_end = str_end; + + const char *equals_sign = memchr(*str, '=', + (size_t)(name_eq_value_end - *str)); + + // Fail if the '=' wasn't found or the option name is missing + // (the first char is '='). + if (equals_sign == NULL || **str == '=') + return "Options must be 'name=value' pairs separated " + "with commas"; + + // Reject a too long option name so that the memcmp() + // in the loop below won't read past the end of the + // string in optmap[i].name. + const size_t name_len = (size_t)(equals_sign - *str); + if (name_len > NAME_LEN_MAX) + return "Unknown option name"; + + // Find the option name from optmap[]. + size_t i = 0; + while (true) { + if (i == optmap_size) + return "Unknown option name"; + + if (memcmp(*str, optmap[i].name, name_len) == 0 + && optmap[i].name[name_len] == '\0') + break; + + ++i; + } + + // The input string is good at least until the start of + // the option value. + *str = equals_sign + 1; + + // The code assumes that the option value isn't an empty + // string so check it here. + const size_t value_len = (size_t)(name_eq_value_end - *str); + if (value_len == 0) + return "Option value cannot be empty"; + + // LZMA1/2 preset has its own parsing function. + if (optmap[i].type == OPTMAP_TYPE_LZMA_PRESET) { + const char *errmsg = set_lzma12_preset(str, + name_eq_value_end, filter_options); + if (errmsg != NULL) + return errmsg; + + continue; + } + + // It's an integer value. + uint32_t v; + if (optmap[i].flags & OPTMAP_USE_NAME_VALUE_MAP) { + // The integer is picked from a string-to-integer map. + // + // Reject a too long value string so that the memcmp() + // in the loop below won't read past the end of the + // string in optmap[i].u.map[j].name. + if (value_len > NAME_LEN_MAX) + return "Invalid option value"; + + const name_value_map *map = optmap[i].u.map; + size_t j = 0; + while (true) { + // The array is terminated with an empty name. + if (map[j].name[0] == '\0') + return "Invalid option value"; + + if (memcmp(*str, map[j].name, value_len) == 0 + && map[j].name[value_len] + == '\0') { + v = map[j].value; + break; + } + + ++j; + } + } else if (**str < '0' || **str > '9') { + // Note that "max" isn't supported while it is + // supported in xz. It's not useful here. + return "Value is not a non-negative decimal integer"; + } else { + // strtoul() has locale-specific behavior so it cannot + // be relied on to get reproducible results since we + // cannot change the locate in a thread-safe library. + // It also needs '\0'-termination. + // + // Use a temporary pointer so that *str will point + // to the beginning of the value string in case + // an error occurs. + const char *p = *str; + v = 0; + do { + if (v > UINT32_MAX / 10) + return "Value out of range"; + + v *= 10; + + const uint32_t add = (uint32_t)(*p - '0'); + if (UINT32_MAX - add < v) + return "Value out of range"; + + v += add; + ++p; + } while (p < name_eq_value_end + && *p >= '0' && *p <= '9'); + + if (p < name_eq_value_end) { + // Remember this position so that it an be + // used for error messages that are + // specifically about the suffix. (Out of + // range values are about the whole value + // and those error messages point to the + // beginning of the number part, + // not to the suffix.) + const char *multiplier_start = p; + + // If multiplier suffix shouldn't be used + // then don't allow them even if the value + // would stay within limits. This is a somewhat + // unnecessary check but it rejects silly + // things like lzma2:pb=0MiB which xz allows. + if ((optmap[i].flags & OPTMAP_USE_BYTE_SUFFIX) + == 0) { + *str = multiplier_start; + return "This option does not support " + "any integer suffixes"; + } + + uint32_t shift; + + switch (*p) { + case 'k': + case 'K': + shift = 10; + break; + + case 'm': + case 'M': + shift = 20; + break; + + case 'g': + case 'G': + shift = 30; + break; + + default: + *str = multiplier_start; + return "Invalid multiplier suffix " + "(KiB, MiB, or GiB)"; + } + + ++p; + + // Allow "M", "Mi", "MB", "MiB" and the same + // for the other five characters from the + // switch-statement above. All are handled + // as base-2 (perhaps a mistake, perhaps not). + // Note that 'i' and 'B' are case sensitive. + if (p < name_eq_value_end && *p == 'i') + ++p; + + if (p < name_eq_value_end && *p == 'B') + ++p; + + // Now we must have no chars remaining. + if (p < name_eq_value_end) { + *str = multiplier_start; + return "Invalid multiplier suffix " + "(KiB, MiB, or GiB)"; + } + + if (v > (UINT32_MAX >> shift)) + return "Value out of range"; + + v <<= shift; + } + + if (v < optmap[i].u.range.min + || v > optmap[i].u.range.max) + return "Value out of range"; + } + + // Set the value in filter_options. Enums are handled + // specially since the underlying type isn't the same + // as uint32_t on all systems. + void *ptr = (char *)filter_options + optmap[i].offset; + switch (optmap[i].type) { + case OPTMAP_TYPE_LZMA_MODE: + *(lzma_mode *)ptr = (lzma_mode)v; + break; + + case OPTMAP_TYPE_LZMA_MATCH_FINDER: + *(lzma_match_finder *)ptr = (lzma_match_finder)v; + break; + + default: + *(uint32_t *)ptr = v; + break; + } + + // This option has been successfully handled. + *str = name_eq_value_end; + } + + // No errors. + return NULL; +} + + +/// Finds the name of the filter at the beginning of the string and +/// calls filter_name_map[i].parse() to decode the filter-specific options. +/// The caller must have set str_end so that exactly one filter and its +/// options are present without any trailing characters. +static const char * +parse_filter(const char **const str, const char *str_end, lzma_filter *filter, + const lzma_allocator *allocator, bool only_xz) +{ + // Search for a colon or equals sign that would separate the filter + // name from filter options. If neither is found, then the input + // string only contains a filter name and there are no options. + // + // First assume that a colon or equals sign won't be found: + const char *name_end = str_end; + const char *opts_start = str_end; + + for (const char *p = *str; p < str_end; ++p) { + if (*p == ':' || *p == '=') { + name_end = p; + + // Filter options (name1=value1,name2=value2,...) + // begin after the colon or equals sign. + opts_start = p + 1; + break; + } + } + + // Reject a too long filter name so that the memcmp() + // in the loop below won't read past the end of the + // string in filter_name_map[i].name. + const size_t name_len = (size_t)(name_end - *str); + if (name_len > NAME_LEN_MAX) + return "Unknown filter name"; + + for (size_t i = 0; i < ARRAY_SIZE(filter_name_map); ++i) { + if (memcmp(*str, filter_name_map[i].name, name_len) == 0 + && filter_name_map[i].name[name_len] == '\0') { + if (only_xz && filter_name_map[i].id + >= LZMA_FILTER_RESERVED_START) + return "This filter cannot be used in " + "the .xz format"; + + // Allocate the filter-specific options and + // initialize the memory with zeros. + void *options = lzma_alloc_zero( + filter_name_map[i].opts_size, + allocator); + if (options == NULL) + return "Memory allocation failed"; + + // Filter name was found so the input string is good + // at least this far. + *str = opts_start; + + const char *errmsg = filter_name_map[i].parse( + str, str_end, options); + if (errmsg != NULL) { + lzma_free(options, allocator); + return errmsg; + } + + // *filter is modified only when parsing is successful. + filter->id = filter_name_map[i].id; + filter->options = options; + return NULL; + } + } + + return "Unknown filter name"; +} + + +/// Converts the string to a filter chain (array of lzma_filter structures). +/// +/// *str is advanced everytime something has been decoded successfully. +/// This way the caller knows where in the string a possible error occurred. +static const char * +str_to_filters(const char **const str, lzma_filter *filters, uint32_t flags, + const lzma_allocator *allocator) +{ + const char *errmsg; + + // Skip leading spaces. + while (**str == ' ') + ++*str; + + if (**str == '\0') + return "Empty string is not allowed, " + "try \"6\" if a default value is needed"; + + // Detect the type of the string. + // + // A string beginning with a digit or a string beginning with + // one dash and a digit are treated as presets. Trailing spaces + // will be ignored too (leading spaces were already ignored above). + // + // For example, "6", "7 ", "-9e", or " -3 " are treated as presets. + // Strings like "-" or "- " aren't preset. +#define MY_IS_DIGIT(c) ((c) >= '0' && (c) <= '9') + if (MY_IS_DIGIT(**str) || (**str == '-' && MY_IS_DIGIT((*str)[1]))) { + if (**str == '-') + ++*str; + + // Ignore trailing spaces. + const size_t str_len = strlen(*str); + const char *str_end = memchr(*str, ' ', str_len); + if (str_end != NULL) { + // There is at least one trailing space. Check that + // there are no chars other than spaces. + for (size_t i = 1; str_end[i] != '\0'; ++i) + if (str_end[i] != ' ') + return "Unsupported preset"; + } else { + // There are no trailing spaces. Use the whole string. + str_end = *str + str_len; + } + + uint32_t preset; + errmsg = parse_lzma12_preset(str, str_end, &preset); + if (errmsg != NULL) + return errmsg; + + lzma_options_lzma *opts = lzma_alloc(sizeof(*opts), allocator); + if (opts == NULL) + return "Memory allocation failed"; + + if (lzma_lzma_preset(opts, preset)) { + lzma_free(opts, allocator); + return "Unsupported preset"; + } + + filters[0].id = LZMA_FILTER_LZMA2; + filters[0].options = opts; + filters[1].id = LZMA_VLI_UNKNOWN; + filters[1].options = NULL; + + return NULL; + } + + // Not a preset so it must be a filter chain. + // + // If LZMA_STR_ALL_FILTERS isn't used we allow only filters that + // can be used in .xz. + const bool only_xz = (flags & LZMA_STR_ALL_FILTERS) == 0; + + // Use a temporary array so that we don't modify the caller-supplied + // one until we know that no errors occurred. + lzma_filter temp_filters[LZMA_FILTERS_MAX + 1]; + + size_t i = 0; + do { + if (i == LZMA_FILTERS_MAX) { + errmsg = "The maximum number of filters is four"; + goto error; + } + + // Skip "--" if present. + if ((*str)[0] == '-' && (*str)[1] == '-') + *str += 2; + + // Locate the end of "filter:name1=value1,name2=value2", + // stopping at the first "--" or a single space. + const char *filter_end = *str; + while (filter_end[0] != '\0') { + if ((filter_end[0] == '-' && filter_end[1] == '-') + || filter_end[0] == ' ') + break; + + ++filter_end; + } + + // Inputs that have "--" at the end or "-- " in the middle + // will result in an empty filter name. + if (filter_end == *str) { + errmsg = "Filter name is missing"; + goto error; + } + + errmsg = parse_filter(str, filter_end, &temp_filters[i], + allocator, only_xz); + if (errmsg != NULL) + goto error; + + // Skip trailing spaces. + while (**str == ' ') + ++*str; + + ++i; + } while (**str != '\0'); + + // Seems to be good, terminate the array so that + // basic validation can be done. + temp_filters[i].id = LZMA_VLI_UNKNOWN; + temp_filters[i].options = NULL; + + // Do basic validation if the application didn't prohibit it. + if ((flags & LZMA_STR_NO_VALIDATION) == 0) { + size_t dummy; + const lzma_ret ret = lzma_validate_chain(temp_filters, &dummy); + assert(ret == LZMA_OK || ret == LZMA_OPTIONS_ERROR); + if (ret != LZMA_OK) { + errmsg = "Invalid filter chain " + "('lzma2' missing at the end?)"; + goto error; + } + } + + // All good. Copy the filters to the application supplied array. + memcpy(filters, temp_filters, (i + 1) * sizeof(lzma_filter)); + return NULL; + +error: + // Free the filter options that were successfully decoded. + while (i-- > 0) + lzma_free(temp_filters[i].options, allocator); + + return errmsg; +} + + +extern LZMA_API(const char *) +lzma_str_to_filters(const char *str, int *error_pos, lzma_filter *filters, + uint32_t flags, const lzma_allocator *allocator) +{ + if (str == NULL || filters == NULL) + return "Unexpected NULL pointer argument(s) " + "to lzma_str_to_filters()"; + + // Validate the flags. + const uint32_t supported_flags + = LZMA_STR_ALL_FILTERS + | LZMA_STR_NO_VALIDATION; + + if (flags & ~supported_flags) + return "Unsupported flags to lzma_str_to_filters()"; + + const char *used = str; + const char *errmsg = str_to_filters(&used, filters, flags, allocator); + + if (error_pos != NULL) { + const size_t n = (size_t)(used - str); + *error_pos = n > INT_MAX ? INT_MAX : (int)n; + } + + return errmsg; +} + + +/// Converts options of one filter to a string. +/// +/// The caller must have already put the filter name in the destination +/// string. Since it is possible that no options will be needed, the caller +/// won't have put a delimiter character (':' or '=') in the string yet. +/// We will add it if at least one option will be added to the string. +static void +strfy_filter(lzma_str *dest, const char *delimiter, + const option_map *optmap, size_t optmap_count, + const void *filter_options) +{ + for (size_t i = 0; i < optmap_count; ++i) { + // No attempt is made to reverse LZMA1/2 preset. + if (optmap[i].type == OPTMAP_TYPE_LZMA_PRESET) + continue; + + // All options have integer values, some just are mapped + // to a string with a name_value_map. LZMA1/2 preset + // isn't reversed back to preset=PRESET form. + uint32_t v; + const void *ptr + = (const char *)filter_options + optmap[i].offset; + switch (optmap[i].type) { + case OPTMAP_TYPE_LZMA_MODE: + v = *(const lzma_mode *)ptr; + break; + + case OPTMAP_TYPE_LZMA_MATCH_FINDER: + v = *(const lzma_match_finder *)ptr; + break; + + default: + v = *(const uint32_t *)ptr; + break; + } + + // Skip this if this option should be omitted from + // the string when the value is zero. + if (v == 0 && (optmap[i].flags & OPTMAP_NO_STRFY_ZERO)) + continue; + + // Before the first option we add whatever delimiter + // the caller gave us. For later options a comma is used. + str_append_str(dest, delimiter); + delimiter = ","; + + // Add the option name and equals sign. + str_append_str(dest, optmap[i].name); + str_append_str(dest, "="); + + if (optmap[i].flags & OPTMAP_USE_NAME_VALUE_MAP) { + const name_value_map *map = optmap[i].u.map; + size_t j = 0; + while (true) { + if (map[j].name[0] == '\0') { + str_append_str(dest, "UNKNOWN"); + break; + } + + if (map[j].value == v) { + str_append_str(dest, map[j].name); + break; + } + + ++j; + } + } else { + str_append_u32(dest, v, + optmap[i].flags & OPTMAP_USE_BYTE_SUFFIX); + } + } + + return; +} + + +extern LZMA_API(lzma_ret) +lzma_str_from_filters(char **output_str, const lzma_filter *filters, + uint32_t flags, const lzma_allocator *allocator) +{ + // On error *output_str is always set to NULL. + // Do it as the very first step. + if (output_str == NULL) + return LZMA_PROG_ERROR; + + *output_str = NULL; + + if (filters == NULL) + return LZMA_PROG_ERROR; + + // Validate the flags. + const uint32_t supported_flags + = LZMA_STR_ENCODER + | LZMA_STR_DECODER + | LZMA_STR_GETOPT_LONG + | LZMA_STR_NO_SPACES; + + if (flags & ~supported_flags) + return LZMA_OPTIONS_ERROR; + + // There must be at least one filter. + if (filters[0].id == LZMA_VLI_UNKNOWN) + return LZMA_OPTIONS_ERROR; + + // Allocate memory for the output string. + lzma_str dest; + return_if_error(str_init(&dest, allocator)); + + const bool show_opts = (flags & (LZMA_STR_ENCODER | LZMA_STR_DECODER)); + + const char *opt_delim = (flags & LZMA_STR_GETOPT_LONG) ? "=" : ":"; + + for (size_t i = 0; filters[i].id != LZMA_VLI_UNKNOWN; ++i) { + // Don't add a space between filters if the caller + // doesn't want them. + if (i > 0 && !(flags & LZMA_STR_NO_SPACES)) + str_append_str(&dest, " "); + + // Use dashes for xz getopt_long() compatible syntax but also + // use dashes to separate filters when spaces weren't wanted. + if ((flags & LZMA_STR_GETOPT_LONG) + || (i > 0 && (flags & LZMA_STR_NO_SPACES))) + str_append_str(&dest, "--"); + + size_t j = 0; + while (true) { + if (j == ARRAY_SIZE(filter_name_map)) { + // Filter ID in filters[i].id isn't supported. + str_free(&dest, allocator); + return LZMA_OPTIONS_ERROR; + } + + if (filter_name_map[j].id == filters[i].id) { + // Add the filter name. + str_append_str(&dest, filter_name_map[j].name); + + // If only the filter names were wanted then + // skip to the next filter. In this case + // .options is ignored and may be NULL even + // when the filter doesn't allow NULL options. + if (!show_opts) + break; + + if (filters[i].options == NULL) { + if (!filter_name_map[j].allow_null) { + // Filter-specific options + // are missing but with + // this filter the options + // structure is mandatory. + str_free(&dest, allocator); + return LZMA_OPTIONS_ERROR; + } + + // .options is allowed to be NULL. + // There is no need to add any + // options to the string. + break; + } + + // Options structure is available. Add + // the filter options to the string. + const size_t optmap_count + = (flags & LZMA_STR_ENCODER) + ? filter_name_map[j].strfy_encoder + : filter_name_map[j].strfy_decoder; + strfy_filter(&dest, opt_delim, + filter_name_map[j].optmap, + optmap_count, + filters[i].options); + break; + } + + ++j; + } + } + + return str_finish(output_str, &dest, allocator); +} + + +extern LZMA_API(lzma_ret) +lzma_str_list_filters(char **output_str, lzma_vli filter_id, uint32_t flags, + const lzma_allocator *allocator) +{ + // On error *output_str is always set to NULL. + // Do it as the very first step. + if (output_str == NULL) + return LZMA_PROG_ERROR; + + *output_str = NULL; + + // Validate the flags. + const uint32_t supported_flags + = LZMA_STR_ALL_FILTERS + | LZMA_STR_ENCODER + | LZMA_STR_DECODER + | LZMA_STR_GETOPT_LONG; + + if (flags & ~supported_flags) + return LZMA_OPTIONS_ERROR; + + // Allocate memory for the output string. + lzma_str dest; + return_if_error(str_init(&dest, allocator)); + + // If only listing the filter names then separate them with spaces. + // Otherwise use newlines. + const bool show_opts = (flags & (LZMA_STR_ENCODER | LZMA_STR_DECODER)); + const char *filter_delim = show_opts ? "\n" : " "; + + const char *opt_delim = (flags & LZMA_STR_GETOPT_LONG) ? "=" : ":"; + bool first_filter_printed = false; + + for (size_t i = 0; i < ARRAY_SIZE(filter_name_map); ++i) { + // If we are printing only one filter then skip others. + if (filter_id != LZMA_VLI_UNKNOWN + && filter_id != filter_name_map[i].id) + continue; + + // If we are printing only .xz filters then skip the others. + if (filter_name_map[i].id >= LZMA_FILTER_RESERVED_START + && (flags & LZMA_STR_ALL_FILTERS) == 0 + && filter_id == LZMA_VLI_UNKNOWN) + continue; + + // Add a new line if this isn't the first filter being + // written to the string. + if (first_filter_printed) + str_append_str(&dest, filter_delim); + + first_filter_printed = true; + + if (flags & LZMA_STR_GETOPT_LONG) + str_append_str(&dest, "--"); + + str_append_str(&dest, filter_name_map[i].name); + + // If only the filter names were wanted then continue + // to the next filter. + if (!show_opts) + continue; + + const option_map *optmap = filter_name_map[i].optmap; + const char *d = opt_delim; + + const size_t end = (flags & LZMA_STR_ENCODER) + ? filter_name_map[i].strfy_encoder + : filter_name_map[i].strfy_decoder; + + for (size_t j = 0; j < end; ++j) { + // The first option is delimited from the filter + // name using "=" or ":" and the rest of the options + // are separated with ",". + str_append_str(&dest, d); + d = ","; + + // optname=<possible_values> + str_append_str(&dest, optmap[j].name); + str_append_str(&dest, "=<"); + + if (optmap[j].type == OPTMAP_TYPE_LZMA_PRESET) { + // LZMA1/2 preset has its custom help string. + str_append_str(&dest, LZMA12_PRESET_STR); + } else if (optmap[j].flags + & OPTMAP_USE_NAME_VALUE_MAP) { + // Separate the possible option values by "|". + const name_value_map *m = optmap[j].u.map; + for (size_t k = 0; m[k].name[0] != '\0'; ++k) { + if (k > 0) + str_append_str(&dest, "|"); + + str_append_str(&dest, m[k].name); + } + } else { + // Integer range is shown as min-max. + const bool use_byte_suffix = optmap[j].flags + & OPTMAP_USE_BYTE_SUFFIX; + str_append_u32(&dest, optmap[j].u.range.min, + use_byte_suffix); + str_append_str(&dest, "-"); + str_append_u32(&dest, optmap[j].u.range.max, + use_byte_suffix); + } + + str_append_str(&dest, ">"); + } + } + + // If no filters were added to the string then it must be because + // the caller provided an unsupported Filter ID. + if (!first_filter_printed) { + str_free(&dest, allocator); + return LZMA_OPTIONS_ERROR; + } + + return str_finish(output_str, &dest, allocator); +} diff --git a/contrib/libs/lzma/liblzma/lz/lz_decoder.c b/contrib/libs/lzma/liblzma/lz/lz_decoder.c index 09b574388f..06c95c1137 100644 --- a/contrib/libs/lzma/liblzma/lz/lz_decoder.c +++ b/contrib/libs/lzma/liblzma/lz/lz_decoder.c @@ -212,7 +212,8 @@ extern lzma_ret lzma_lz_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, const lzma_filter_info *filters, lzma_ret (*lz_init)(lzma_lz_decoder *lz, - const lzma_allocator *allocator, const void *options, + const lzma_allocator *allocator, + lzma_vli id, const void *options, lzma_lz_options *lz_options)) { // Allocate the base structure if it isn't already allocated. @@ -236,7 +237,7 @@ lzma_lz_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, // us the dictionary size. lzma_lz_options lz_options; return_if_error(lz_init(&coder->lz, allocator, - filters[0].options, &lz_options)); + filters[0].id, filters[0].options, &lz_options)); // If the dictionary size is very small, increase it to 4096 bytes. // This is to prevent constant wrapping of the dictionary, which @@ -301,11 +302,3 @@ lzma_lz_decoder_memusage(size_t dictionary_size) { return sizeof(lzma_coder) + (uint64_t)(dictionary_size); } - - -extern void -lzma_lz_decoder_uncompressed(void *coder_ptr, lzma_vli uncompressed_size) -{ - lzma_coder *coder = coder_ptr; - coder->lz.set_uncompressed(coder->lz.coder, uncompressed_size); -} diff --git a/contrib/libs/lzma/liblzma/lz/lz_decoder.h b/contrib/libs/lzma/liblzma/lz/lz_decoder.h index 754ccf37c6..ad80d4dd0d 100644 --- a/contrib/libs/lzma/liblzma/lz/lz_decoder.h +++ b/contrib/libs/lzma/liblzma/lz/lz_decoder.h @@ -62,8 +62,10 @@ typedef struct { void (*reset)(void *coder, const void *options); - /// Set the uncompressed size - void (*set_uncompressed)(void *coder, lzma_vli uncompressed_size); + /// Set the uncompressed size. If uncompressed_size == LZMA_VLI_UNKNOWN + /// then allow_eopm will always be true. + void (*set_uncompressed)(void *coder, lzma_vli uncompressed_size, + bool allow_eopm); /// Free allocated resources void (*end)(void *coder, const lzma_allocator *allocator); @@ -85,14 +87,12 @@ extern lzma_ret lzma_lz_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, const lzma_filter_info *filters, lzma_ret (*lz_init)(lzma_lz_decoder *lz, - const lzma_allocator *allocator, const void *options, + const lzma_allocator *allocator, + lzma_vli id, const void *options, lzma_lz_options *lz_options)); extern uint64_t lzma_lz_decoder_memusage(size_t dictionary_size); -extern void lzma_lz_decoder_uncompressed( - void *coder, lzma_vli uncompressed_size); - ////////////////////// // Inline functions // diff --git a/contrib/libs/lzma/liblzma/lz/lz_encoder.c b/contrib/libs/lzma/liblzma/lz/lz_encoder.c index ad7a303acb..63d4aa057f 100644 --- a/contrib/libs/lzma/liblzma/lz/lz_encoder.c +++ b/contrib/libs/lzma/liblzma/lz/lz_encoder.c @@ -293,11 +293,15 @@ lz_encoder_prepare(lzma_mf *mf, const lzma_allocator *allocator, return true; } - // Calculate the sizes of mf->hash and mf->son and check that - // nice_len is big enough for the selected match finder. - const uint32_t hash_bytes = lz_options->match_finder & 0x0F; - if (hash_bytes > mf->nice_len) - return true; + // Calculate the sizes of mf->hash and mf->son. + // + // NOTE: Since 5.3.5beta the LZMA encoder ensures that nice_len + // is big enough for the selected match finder. This makes it + // easier for applications as nice_len = 2 will always be accepted + // even though the effective value can be slightly bigger. + const uint32_t hash_bytes + = mf_get_hash_bytes(lz_options->match_finder); + assert(hash_bytes <= mf->nice_len); const bool is_bt = (lz_options->match_finder & 0x10) != 0; uint32_t hs; @@ -521,14 +525,30 @@ lz_encoder_update(void *coder_ptr, const lzma_allocator *allocator, } +static lzma_ret +lz_encoder_set_out_limit(void *coder_ptr, uint64_t *uncomp_size, + uint64_t out_limit) +{ + lzma_coder *coder = coder_ptr; + + // This is supported only if there are no other filters chained. + if (coder->next.code == NULL && coder->lz.set_out_limit != NULL) + return coder->lz.set_out_limit( + coder->lz.coder, uncomp_size, out_limit); + + return LZMA_OPTIONS_ERROR; +} + + extern lzma_ret lzma_lz_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, const lzma_filter_info *filters, lzma_ret (*lz_init)(lzma_lz_encoder *lz, - const lzma_allocator *allocator, const void *options, + const lzma_allocator *allocator, + lzma_vli id, const void *options, lzma_lz_options *lz_options)) { -#ifdef HAVE_SMALL +#if defined(HAVE_SMALL) && !defined(HAVE_FUNC_ATTRIBUTE_CONSTRUCTOR) // We need that the CRC32 table has been initialized. lzma_crc32_init(); #endif @@ -544,6 +564,7 @@ lzma_lz_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, next->code = &lz_encode; next->end = &lz_encoder_end; next->update = &lz_encoder_update; + next->set_out_limit = &lz_encoder_set_out_limit; coder->lz.coder = NULL; coder->lz.code = NULL; @@ -565,7 +586,7 @@ lzma_lz_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, // Initialize the LZ-based encoder. lzma_lz_options lz_options; return_if_error(lz_init(&coder->lz, allocator, - filters[0].options, &lz_options)); + filters[0].id, filters[0].options, &lz_options)); // Setup the size information into coder->mf and deallocate // old buffers if they have wrong size. @@ -585,32 +606,28 @@ lzma_lz_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, extern LZMA_API(lzma_bool) lzma_mf_is_supported(lzma_match_finder mf) { - bool ret = false; - + switch (mf) { #ifdef HAVE_MF_HC3 - if (mf == LZMA_MF_HC3) - ret = true; + case LZMA_MF_HC3: + return true; #endif - #ifdef HAVE_MF_HC4 - if (mf == LZMA_MF_HC4) - ret = true; + case LZMA_MF_HC4: + return true; #endif - #ifdef HAVE_MF_BT2 - if (mf == LZMA_MF_BT2) - ret = true; + case LZMA_MF_BT2: + return true; #endif - #ifdef HAVE_MF_BT3 - if (mf == LZMA_MF_BT3) - ret = true; + case LZMA_MF_BT3: + return true; #endif - #ifdef HAVE_MF_BT4 - if (mf == LZMA_MF_BT4) - ret = true; + case LZMA_MF_BT4: + return true; #endif - - return ret; + default: + return false; + } } diff --git a/contrib/libs/lzma/liblzma/lz/lz_encoder.h b/contrib/libs/lzma/liblzma/lz/lz_encoder.h index 426dcd8a38..7950a2f4ef 100644 --- a/contrib/libs/lzma/liblzma/lz/lz_encoder.h +++ b/contrib/libs/lzma/liblzma/lz/lz_encoder.h @@ -204,6 +204,10 @@ typedef struct { /// Update the options in the middle of the encoding. lzma_ret (*options_update)(void *coder, const lzma_filter *filter); + /// Set maximum allowed output size + lzma_ret (*set_out_limit)(void *coder, uint64_t *uncomp_size, + uint64_t out_limit); + } lzma_lz_encoder; @@ -216,6 +220,15 @@ typedef struct { // are called `read ahead'. +/// Get how many bytes the match finder hashes in its initial step. +/// This is also the minimum nice_len value with the match finder. +static inline uint32_t +mf_get_hash_bytes(lzma_match_finder match_finder) +{ + return (uint32_t)match_finder & 0x0F; +} + + /// Get pointer to the first byte not ran through the match finder static inline const uint8_t * mf_ptr(const lzma_mf *mf) @@ -298,7 +311,8 @@ extern lzma_ret lzma_lz_encoder_init( lzma_next_coder *next, const lzma_allocator *allocator, const lzma_filter_info *filters, lzma_ret (*lz_init)(lzma_lz_encoder *lz, - const lzma_allocator *allocator, const void *options, + const lzma_allocator *allocator, + lzma_vli id, const void *options, lzma_lz_options *lz_options)); diff --git a/contrib/libs/lzma/liblzma/lzma/lzma2_decoder.c b/contrib/libs/lzma/liblzma/lzma/lzma2_decoder.c index cf1b5110ac..567df490ca 100644 --- a/contrib/libs/lzma/liblzma/lzma/lzma2_decoder.c +++ b/contrib/libs/lzma/liblzma/lzma/lzma2_decoder.c @@ -139,7 +139,7 @@ lzma2_decode(void *coder_ptr, lzma_dict *restrict dict, coder->uncompressed_size += in[(*in_pos)++] + 1U; coder->sequence = SEQ_COMPRESSED_0; coder->lzma.set_uncompressed(coder->lzma.coder, - coder->uncompressed_size); + coder->uncompressed_size, false); break; case SEQ_COMPRESSED_0: @@ -226,7 +226,8 @@ lzma2_decoder_end(void *coder_ptr, const lzma_allocator *allocator) static lzma_ret lzma2_decoder_init(lzma_lz_decoder *lz, const lzma_allocator *allocator, - const void *opt, lzma_lz_options *lz_options) + lzma_vli id lzma_attribute((__unused__)), const void *opt, + lzma_lz_options *lz_options) { lzma_lzma2_coder *coder = lz->coder; if (coder == NULL) { diff --git a/contrib/libs/lzma/liblzma/lzma/lzma2_encoder.c b/contrib/libs/lzma/liblzma/lzma/lzma2_encoder.c index 63588ee30c..4b6b23118d 100644 --- a/contrib/libs/lzma/liblzma/lzma/lzma2_encoder.c +++ b/contrib/libs/lzma/liblzma/lzma/lzma2_encoder.c @@ -310,7 +310,8 @@ lzma2_encoder_options_update(void *coder_ptr, const lzma_filter *filter) static lzma_ret lzma2_encoder_init(lzma_lz_encoder *lz, const lzma_allocator *allocator, - const void *options, lzma_lz_options *lz_options) + lzma_vli id lzma_attribute((__unused__)), const void *options, + lzma_lz_options *lz_options) { if (options == NULL) return LZMA_PROG_ERROR; @@ -340,7 +341,7 @@ lzma2_encoder_init(lzma_lz_encoder *lz, const lzma_allocator *allocator, // Initialize LZMA encoder return_if_error(lzma_lzma_encoder_create(&coder->lzma, allocator, - &coder->opt_cur, lz_options)); + LZMA_FILTER_LZMA2, &coder->opt_cur, lz_options)); // Make sure that we will always have enough history available in // case we need to use uncompressed chunks. They are used when the @@ -378,6 +379,9 @@ lzma_lzma2_encoder_memusage(const void *options) extern lzma_ret lzma_lzma2_props_encode(const void *options, uint8_t *out) { + if (options == NULL) + return LZMA_PROG_ERROR; + const lzma_options_lzma *const opt = options; uint32_t d = my_max(opt->dict_size, LZMA_DICT_SIZE_MIN); diff --git a/contrib/libs/lzma/liblzma/lzma/lzma_decoder.c b/contrib/libs/lzma/liblzma/lzma/lzma_decoder.c index e605a0a916..26c148a95e 100644 --- a/contrib/libs/lzma/liblzma/lzma/lzma_decoder.c +++ b/contrib/libs/lzma/liblzma/lzma/lzma_decoder.c @@ -238,6 +238,11 @@ typedef struct { /// payload marker is expected. lzma_vli uncompressed_size; + /// True if end of payload marker (EOPM) is allowed even when + /// uncompressed_size is known; false if EOPM must not be present. + /// This is ignored if uncompressed_size == LZMA_VLI_UNKNOWN. + bool allow_eopm; + //////////////////////////////// // State of incomplete symbol // //////////////////////////////// @@ -343,12 +348,24 @@ lzma_decode(void *coder_ptr, lzma_dict *restrict dictptr, lzma_ret ret = LZMA_OK; - // If uncompressed size is known, there must be no end of payload - // marker. - const bool no_eopm = coder->uncompressed_size - != LZMA_VLI_UNKNOWN; - if (no_eopm && coder->uncompressed_size < dict.limit - dict.pos) + // This is true when the next LZMA symbol is allowed to be EOPM. + // That is, if this is false, then EOPM is considered + // an invalid symbol and we will return LZMA_DATA_ERROR. + // + // EOPM is always required (not just allowed) when + // the uncompressed size isn't known. When uncompressed size + // is known, eopm_is_valid may be set to true later. + bool eopm_is_valid = coder->uncompressed_size == LZMA_VLI_UNKNOWN; + + // If uncompressed size is known and there is enough output space + // to decode all the data, limit the available buffer space so that + // the main loop won't try to decode past the end of the stream. + bool might_finish_without_eopm = false; + if (coder->uncompressed_size != LZMA_VLI_UNKNOWN + && coder->uncompressed_size <= dict.limit - dict.pos) { dict.limit = dict.pos + (size_t)(coder->uncompressed_size); + might_finish_without_eopm = true; + } // The main decoder loop. The "switch" is used to restart the decoder at // correct location. Once restarted, the "switch" is no longer used. @@ -361,8 +378,32 @@ lzma_decode(void *coder_ptr, lzma_dict *restrict dictptr, case SEQ_NORMALIZE: case SEQ_IS_MATCH: - if (unlikely(no_eopm && dict.pos == dict.limit)) - break; + if (unlikely(might_finish_without_eopm + && dict.pos == dict.limit)) { + // In rare cases there is a useless byte that needs + // to be read anyway. + rc_normalize(SEQ_NORMALIZE); + + // If the range decoder state is such that we can + // be at the end of the LZMA stream, then the + // decoding is finished. + if (rc_is_finished(rc)) { + ret = LZMA_STREAM_END; + goto out; + } + + // If the caller hasn't allowed EOPM to be present + // together with known uncompressed size, then the + // LZMA stream is corrupt. + if (!coder->allow_eopm) { + ret = LZMA_DATA_ERROR; + goto out; + } + + // Otherwise continue decoding with the expectation + // that the next LZMA symbol is EOPM. + eopm_is_valid = true; + } rc_if_0(coder->is_match[state][pos_state], SEQ_IS_MATCH) { rc_update_0(coder->is_match[state][pos_state]); @@ -658,11 +699,18 @@ lzma_decode(void *coder_ptr, lzma_dict *restrict dictptr, if (rep0 == UINT32_MAX) { // End of payload marker was - // found. It must not be - // present if uncompressed - // size is known. - if (coder->uncompressed_size - != LZMA_VLI_UNKNOWN) { + // found. It may only be + // present if + // - uncompressed size is + // unknown or + // - after known uncompressed + // size amount of bytes has + // been decompressed and + // caller has indicated + // that EOPM might be used + // (it's not allowed in + // LZMA2). + if (!eopm_is_valid) { ret = LZMA_DATA_ERROR; goto out; } @@ -671,7 +719,9 @@ lzma_decode(void *coder_ptr, lzma_dict *restrict dictptr, // LZMA1 stream with // end-of-payload marker. rc_normalize(SEQ_EOPM); - ret = LZMA_STREAM_END; + ret = rc_is_finished(rc) + ? LZMA_STREAM_END + : LZMA_DATA_ERROR; goto out; } } @@ -793,9 +843,6 @@ lzma_decode(void *coder_ptr, lzma_dict *restrict dictptr, } } - rc_normalize(SEQ_NORMALIZE); - coder->sequence = SEQ_IS_MATCH; - out: // Save state @@ -822,24 +869,21 @@ out: if (coder->uncompressed_size != LZMA_VLI_UNKNOWN) { coder->uncompressed_size -= dict.pos - dict_start; - // Since there cannot be end of payload marker if the - // uncompressed size was known, we check here if we - // finished decoding. + // If we have gotten all the output but the decoder wants + // to write more output, the file is corrupt. There are + // three SEQ values where output is produced. if (coder->uncompressed_size == 0 && ret == LZMA_OK - && coder->sequence != SEQ_NORMALIZE) - ret = coder->sequence == SEQ_IS_MATCH - ? LZMA_STREAM_END : LZMA_DATA_ERROR; + && (coder->sequence == SEQ_LITERAL_WRITE + || coder->sequence == SEQ_SHORTREP + || coder->sequence == SEQ_COPY)) + ret = LZMA_DATA_ERROR; } - // We can do an additional check in the range decoder to catch some - // corrupted files. if (ret == LZMA_STREAM_END) { - if (!rc_is_finished(coder->rc)) - ret = LZMA_DATA_ERROR; - // Reset the range decoder so that it is ready to reinitialize // for a new LZMA2 chunk. rc_reset(coder->rc); + coder->sequence = SEQ_IS_MATCH; } return ret; @@ -848,10 +892,12 @@ out: static void -lzma_decoder_uncompressed(void *coder_ptr, lzma_vli uncompressed_size) +lzma_decoder_uncompressed(void *coder_ptr, lzma_vli uncompressed_size, + bool allow_eopm) { lzma_lzma1_decoder *coder = coder_ptr; coder->uncompressed_size = uncompressed_size; + coder->allow_eopm = allow_eopm; } @@ -940,7 +986,7 @@ lzma_decoder_reset(void *coder_ptr, const void *opt) extern lzma_ret lzma_lzma_decoder_create(lzma_lz_decoder *lz, const lzma_allocator *allocator, - const void *opt, lzma_lz_options *lz_options) + const lzma_options_lzma *options, lzma_lz_options *lz_options) { if (lz->coder == NULL) { lz->coder = lzma_alloc(sizeof(lzma_lzma1_decoder), allocator); @@ -954,7 +1000,6 @@ lzma_lzma_decoder_create(lzma_lz_decoder *lz, const lzma_allocator *allocator, // All dictionary sizes are OK here. LZ decoder will take care of // the special cases. - const lzma_options_lzma *options = opt; lz_options->dict_size = options->dict_size; lz_options->preset_dict = options->preset_dict; lz_options->preset_dict_size = options->preset_dict_size; @@ -968,16 +1013,40 @@ lzma_lzma_decoder_create(lzma_lz_decoder *lz, const lzma_allocator *allocator, /// the LZ initialization). static lzma_ret lzma_decoder_init(lzma_lz_decoder *lz, const lzma_allocator *allocator, - const void *options, lzma_lz_options *lz_options) + lzma_vli id, const void *options, lzma_lz_options *lz_options) { if (!is_lclppb_valid(options)) return LZMA_PROG_ERROR; + lzma_vli uncomp_size = LZMA_VLI_UNKNOWN; + bool allow_eopm = true; + + if (id == LZMA_FILTER_LZMA1EXT) { + const lzma_options_lzma *opt = options; + + // Only one flag is supported. + if (opt->ext_flags & ~LZMA_LZMA1EXT_ALLOW_EOPM) + return LZMA_OPTIONS_ERROR; + + // FIXME? Using lzma_vli instead of uint64_t is weird because + // this has nothing to do with .xz headers and variable-length + // integer encoding. On the other hand, using LZMA_VLI_UNKNOWN + // instead of UINT64_MAX is clearer when unknown size is + // meant. A problem with using lzma_vli is that now we + // allow > LZMA_VLI_MAX which is fine in this file but + // it's still confusing. Note that alone_decoder.c also + // allows > LZMA_VLI_MAX when setting uncompressed size. + uncomp_size = opt->ext_size_low + + ((uint64_t)(opt->ext_size_high) << 32); + allow_eopm = (opt->ext_flags & LZMA_LZMA1EXT_ALLOW_EOPM) != 0 + || uncomp_size == LZMA_VLI_UNKNOWN; + } + return_if_error(lzma_lzma_decoder_create( lz, allocator, options, lz_options)); lzma_decoder_reset(lz->coder, options); - lzma_decoder_uncompressed(lz->coder, LZMA_VLI_UNKNOWN); + lzma_decoder_uncompressed(lz->coder, uncomp_size, allow_eopm); return LZMA_OK; } diff --git a/contrib/libs/lzma/liblzma/lzma/lzma_decoder.h b/contrib/libs/lzma/liblzma/lzma/lzma_decoder.h index fa8ecb23e4..1427bc2461 100644 --- a/contrib/libs/lzma/liblzma/lzma/lzma_decoder.h +++ b/contrib/libs/lzma/liblzma/lzma/lzma_decoder.h @@ -42,7 +42,7 @@ extern bool lzma_lzma_lclppb_decode( /// LZMA2 decoders. extern lzma_ret lzma_lzma_decoder_create( lzma_lz_decoder *lz, const lzma_allocator *allocator, - const void *opt, lzma_lz_options *lz_options); + const lzma_options_lzma *opt, lzma_lz_options *lz_options); /// Gets memory usage without validating lc/lp/pb. This is used by LZMA2 /// decoder, because raw LZMA2 decoding doesn't need lc/lp/pb. diff --git a/contrib/libs/lzma/liblzma/lzma/lzma_encoder.c b/contrib/libs/lzma/liblzma/lzma/lzma_encoder.c index 07d2b87bc6..dc62f44f1b 100644 --- a/contrib/libs/lzma/liblzma/lzma/lzma_encoder.c +++ b/contrib/libs/lzma/liblzma/lzma/lzma_encoder.c @@ -268,6 +268,7 @@ static bool encode_init(lzma_lzma1_encoder *coder, lzma_mf *mf) { assert(mf_position(mf) == 0); + assert(coder->uncomp_size == 0); if (mf->read_pos == mf->read_limit) { if (mf->action == LZMA_RUN) @@ -283,6 +284,7 @@ encode_init(lzma_lzma1_encoder *coder, lzma_mf *mf) mf->read_ahead = 0; rc_bit(&coder->rc, &coder->is_match[0][0], 0); rc_bittree(&coder->rc, coder->literal[0], 8, mf->buffer[0]); + ++coder->uncomp_size; } // Initialization is done (except if empty file). @@ -317,21 +319,28 @@ lzma_lzma_encode(lzma_lzma1_encoder *restrict coder, lzma_mf *restrict mf, if (!coder->is_initialized && !encode_init(coder, mf)) return LZMA_OK; - // Get the lowest bits of the uncompressed offset from the LZ layer. - uint32_t position = mf_position(mf); + // Encode pending output bytes from the range encoder. + // At the start of the stream, encode_init() encodes one literal. + // Later there can be pending output only with LZMA1 because LZMA2 + // ensures that there is always enough output space. Thus when using + // LZMA2, rc_encode() calls in this function will always return false. + if (rc_encode(&coder->rc, out, out_pos, out_size)) { + // We don't get here with LZMA2. + assert(limit == UINT32_MAX); + return LZMA_OK; + } - while (true) { - // Encode pending bits, if any. Calling this before encoding - // the next symbol is needed only with plain LZMA, since - // LZMA2 always provides big enough buffer to flush - // everything out from the range encoder. For the same reason, - // rc_encode() never returns true when this function is used - // as part of LZMA2 encoder. - if (rc_encode(&coder->rc, out, out_pos, out_size)) { - assert(limit == UINT32_MAX); - return LZMA_OK; - } + // If the range encoder was flushed in an earlier call to this + // function but there wasn't enough output buffer space, those + // bytes would have now been encoded by the above rc_encode() call + // and the stream has now been finished. This can only happen with + // LZMA1 as LZMA2 always provides enough output buffer space. + if (coder->is_flushed) { + assert(limit == UINT32_MAX); + return LZMA_STREAM_END; + } + while (true) { // With LZMA2 we need to take care that compressed size of // a chunk doesn't get too big. // FIXME? Check if this could be improved. @@ -365,37 +374,64 @@ lzma_lzma_encode(lzma_lzma1_encoder *restrict coder, lzma_mf *restrict mf, if (coder->fast_mode) lzma_lzma_optimum_fast(coder, mf, &back, &len); else - lzma_lzma_optimum_normal( - coder, mf, &back, &len, position); - - encode_symbol(coder, mf, back, len, position); + lzma_lzma_optimum_normal(coder, mf, &back, &len, + (uint32_t)(coder->uncomp_size)); + + encode_symbol(coder, mf, back, len, + (uint32_t)(coder->uncomp_size)); + + // If output size limiting is active (out_limit != 0), check + // if encoding this LZMA symbol would make the output size + // exceed the specified limit. + if (coder->out_limit != 0 && rc_encode_dummy( + &coder->rc, coder->out_limit)) { + // The most recent LZMA symbol would make the output + // too big. Throw it away. + rc_forget(&coder->rc); + + // FIXME: Tell the LZ layer to not read more input as + // it would be waste of time. This doesn't matter if + // output-size-limited encoding is done with a single + // call though. - position += len; - } - - if (!coder->is_flushed) { - coder->is_flushed = true; - - // We don't support encoding plain LZMA streams without EOPM, - // and LZMA2 doesn't use EOPM at LZMA level. - if (limit == UINT32_MAX) - encode_eopm(coder, position); + break; + } - // Flush the remaining bytes from the range encoder. - rc_flush(&coder->rc); + // This symbol will be encoded so update the uncompressed size. + coder->uncomp_size += len; - // Copy the remaining bytes to the output buffer. If there - // isn't enough output space, we will copy out the remaining - // bytes on the next call to this function by using - // the rc_encode() call in the encoding loop above. + // Encode the LZMA symbol. if (rc_encode(&coder->rc, out, out_pos, out_size)) { + // Once again, this can only happen with LZMA1. assert(limit == UINT32_MAX); return LZMA_OK; } } - // Make it ready for the next LZMA2 chunk. - coder->is_flushed = false; + // Make the uncompressed size available to the application. + if (coder->uncomp_size_ptr != NULL) + *coder->uncomp_size_ptr = coder->uncomp_size; + + // LZMA2 doesn't use EOPM at LZMA level. + // + // Plain LZMA streams without EOPM aren't supported except when + // output size limiting is enabled. + if (coder->use_eopm) + encode_eopm(coder, (uint32_t)(coder->uncomp_size)); + + // Flush the remaining bytes from the range encoder. + rc_flush(&coder->rc); + + // Copy the remaining bytes to the output buffer. If there + // isn't enough output space, we will copy out the remaining + // bytes on the next call to this function. + if (rc_encode(&coder->rc, out, out_pos, out_size)) { + // This cannot happen with LZMA2. + assert(limit == UINT32_MAX); + + coder->is_flushed = true; + return LZMA_OK; + } return LZMA_STREAM_END; } @@ -414,6 +450,23 @@ lzma_encode(void *coder, lzma_mf *restrict mf, } +static lzma_ret +lzma_lzma_set_out_limit( + void *coder_ptr, uint64_t *uncomp_size, uint64_t out_limit) +{ + // Minimum output size is 5 bytes but that cannot hold any output + // so we use 6 bytes. + if (out_limit < 6) + return LZMA_BUF_ERROR; + + lzma_lzma1_encoder *coder = coder_ptr; + coder->out_limit = out_limit; + coder->uncomp_size_ptr = uncomp_size; + coder->use_eopm = false; + return LZMA_OK; +} + + //////////////////// // Initialization // //////////////////// @@ -440,7 +493,8 @@ set_lz_options(lzma_lz_options *lz_options, const lzma_options_lzma *options) lz_options->dict_size = options->dict_size; lz_options->after_size = LOOP_INPUT_MAX; lz_options->match_len_max = MATCH_LEN_MAX; - lz_options->nice_len = options->nice_len; + lz_options->nice_len = my_max(mf_get_hash_bytes(options->mf), + options->nice_len); lz_options->match_finder = options->mf; lz_options->depth = options->depth; lz_options->preset_dict = options->preset_dict; @@ -546,10 +600,13 @@ lzma_lzma_encoder_reset(lzma_lzma1_encoder *coder, extern lzma_ret -lzma_lzma_encoder_create(void **coder_ptr, - const lzma_allocator *allocator, - const lzma_options_lzma *options, lzma_lz_options *lz_options) +lzma_lzma_encoder_create(void **coder_ptr, const lzma_allocator *allocator, + lzma_vli id, const lzma_options_lzma *options, + lzma_lz_options *lz_options) { + assert(id == LZMA_FILTER_LZMA1 || id == LZMA_FILTER_LZMA1EXT + || id == LZMA_FILTER_LZMA2); + // Allocate lzma_lzma1_encoder if it wasn't already allocated. if (*coder_ptr == NULL) { *coder_ptr = lzma_alloc(sizeof(lzma_lzma1_encoder), allocator); @@ -559,10 +616,9 @@ lzma_lzma_encoder_create(void **coder_ptr, lzma_lzma1_encoder *coder = *coder_ptr; - // Set compression mode. We haven't validates the options yet, - // but it's OK here, since nothing bad happens with invalid - // options in the code below, and they will get rejected by - // lzma_lzma_encoder_reset() call at the end of this function. + // Set compression mode. Note that we haven't validated the options + // yet. Invalid options will get rejected by lzma_lzma_encoder_reset() + // call at the end of this function. switch (options->mode) { case LZMA_MODE_FAST: coder->fast_mode = true; @@ -573,6 +629,18 @@ lzma_lzma_encoder_create(void **coder_ptr, // Set dist_table_size. // Round the dictionary size up to next 2^n. + // + // Currently the maximum encoder dictionary size + // is 1.5 GiB due to lz_encoder.c and here we need + // to be below 2 GiB to make the rounded up value + // fit in an uint32_t and avoid an infite while-loop + // (and undefined behavior due to a too large shift). + // So do the same check as in LZ encoder, + // limiting to 1.5 GiB. + if (options->dict_size > (UINT32_C(1) << 30) + + (UINT32_C(1) << 29)) + return LZMA_OPTIONS_ERROR; + uint32_t log_size = 0; while ((UINT32_C(1) << log_size) < options->dict_size) ++log_size; @@ -580,10 +648,14 @@ lzma_lzma_encoder_create(void **coder_ptr, coder->dist_table_size = log_size * 2; // Length encoders' price table size + const uint32_t nice_len = my_max( + mf_get_hash_bytes(options->mf), + options->nice_len); + coder->match_len_encoder.table_size - = options->nice_len + 1 - MATCH_LEN_MIN; + = nice_len + 1 - MATCH_LEN_MIN; coder->rep_len_encoder.table_size - = options->nice_len + 1 - MATCH_LEN_MIN; + = nice_len + 1 - MATCH_LEN_MIN; break; } @@ -598,6 +670,37 @@ lzma_lzma_encoder_create(void **coder_ptr, coder->is_initialized = options->preset_dict != NULL && options->preset_dict_size > 0; coder->is_flushed = false; + coder->uncomp_size = 0; + coder->uncomp_size_ptr = NULL; + + // Output size limitting is disabled by default. + coder->out_limit = 0; + + // Determine if end marker is wanted: + // - It is never used with LZMA2. + // - It is always used with LZMA_FILTER_LZMA1 (unless + // lzma_lzma_set_out_limit() is called later). + // - LZMA_FILTER_LZMA1EXT has a flag for it in the options. + coder->use_eopm = (id == LZMA_FILTER_LZMA1); + if (id == LZMA_FILTER_LZMA1EXT) { + // Check if unsupported flags are present. + if (options->ext_flags & ~LZMA_LZMA1EXT_ALLOW_EOPM) + return LZMA_OPTIONS_ERROR; + + coder->use_eopm = (options->ext_flags + & LZMA_LZMA1EXT_ALLOW_EOPM) != 0; + + // TODO? As long as there are no filters that change the size + // of the data, it is enough to look at lzma_stream.total_in + // after encoding has been finished to know the uncompressed + // size of the LZMA1 stream. But in the future there could be + // filters that change the size of the data and then total_in + // doesn't work as the LZMA1 stream size might be different + // due to another filter in the chain. The problem is simple + // to solve: Add another flag to ext_flags and then set + // coder->uncomp_size_ptr to the address stored in + // lzma_options_lzma.reserved_ptr2 (or _ptr1). + } set_lz_options(lz_options, options); @@ -607,11 +710,12 @@ lzma_lzma_encoder_create(void **coder_ptr, static lzma_ret lzma_encoder_init(lzma_lz_encoder *lz, const lzma_allocator *allocator, - const void *options, lzma_lz_options *lz_options) + lzma_vli id, const void *options, lzma_lz_options *lz_options) { lz->code = &lzma_encode; + lz->set_out_limit = &lzma_lzma_set_out_limit; return lzma_lzma_encoder_create( - &lz->coder, allocator, options, lz_options); + &lz->coder, allocator, id, options, lz_options); } @@ -658,6 +762,9 @@ lzma_lzma_lclppb_encode(const lzma_options_lzma *options, uint8_t *byte) extern lzma_ret lzma_lzma_props_encode(const void *options, uint8_t *out) { + if (options == NULL) + return LZMA_PROG_ERROR; + const lzma_options_lzma *const opt = options; if (lzma_lzma_lclppb_encode(opt, out)) diff --git a/contrib/libs/lzma/liblzma/lzma/lzma_encoder.h b/contrib/libs/lzma/liblzma/lzma/lzma_encoder.h index 6cfdf228bf..84d8c9163f 100644 --- a/contrib/libs/lzma/liblzma/lzma/lzma_encoder.h +++ b/contrib/libs/lzma/liblzma/lzma/lzma_encoder.h @@ -40,7 +40,8 @@ extern bool lzma_lzma_lclppb_encode( /// Initializes raw LZMA encoder; this is used by LZMA2. extern lzma_ret lzma_lzma_encoder_create( void **coder_ptr, const lzma_allocator *allocator, - const lzma_options_lzma *options, lzma_lz_options *lz_options); + lzma_vli id, const lzma_options_lzma *options, + lzma_lz_options *lz_options); /// Resets an already initialized LZMA encoder; this is used by LZMA2. diff --git a/contrib/libs/lzma/liblzma/lzma/lzma_encoder_private.h b/contrib/libs/lzma/liblzma/lzma/lzma_encoder_private.h index 2e34aace16..b228c57761 100644 --- a/contrib/libs/lzma/liblzma/lzma/lzma_encoder_private.h +++ b/contrib/libs/lzma/liblzma/lzma/lzma_encoder_private.h @@ -72,6 +72,18 @@ struct lzma_lzma1_encoder_s { /// Range encoder lzma_range_encoder rc; + /// Uncompressed size (doesn't include possible preset dictionary) + uint64_t uncomp_size; + + /// If non-zero, produce at most this much output. + /// Some input may then be missing from the output. + uint64_t out_limit; + + /// If the above out_limit is non-zero, *uncomp_size_ptr is set to + /// the amount of uncompressed data that we were able to fit + /// in the output buffer. + uint64_t *uncomp_size_ptr; + /// State lzma_lzma_state state; @@ -99,6 +111,9 @@ struct lzma_lzma1_encoder_s { /// have been written to the output buffer yet. bool is_flushed; + /// True if end of payload marker will be written. + bool use_eopm; + uint32_t pos_mask; ///< (1 << pos_bits) - 1 uint32_t literal_context_bits; uint32_t literal_pos_mask; diff --git a/contrib/libs/lzma/liblzma/rangecoder/range_encoder.h b/contrib/libs/lzma/liblzma/rangecoder/range_encoder.h index 1e1c36995b..d794eabbcc 100644 --- a/contrib/libs/lzma/liblzma/rangecoder/range_encoder.h +++ b/contrib/libs/lzma/liblzma/rangecoder/range_encoder.h @@ -19,9 +19,9 @@ /// Maximum number of symbols that can be put pending into lzma_range_encoder -/// structure between calls to lzma_rc_encode(). For LZMA, 52+5 is enough +/// structure between calls to lzma_rc_encode(). For LZMA, 48+5 is enough /// (match with big distance and length followed by range encoder flush). -#define RC_SYMBOLS_MAX 58 +#define RC_SYMBOLS_MAX 53 typedef struct { @@ -30,6 +30,9 @@ typedef struct { uint32_t range; uint8_t cache; + /// Number of bytes written out by rc_encode() -> rc_shift_low() + uint64_t out_total; + /// Number of symbols in the tables size_t count; @@ -58,12 +61,22 @@ rc_reset(lzma_range_encoder *rc) rc->cache_size = 1; rc->range = UINT32_MAX; rc->cache = 0; + rc->out_total = 0; rc->count = 0; rc->pos = 0; } static inline void +rc_forget(lzma_range_encoder *rc) +{ + // This must not be called when rc_encode() is partially done. + assert(rc->pos == 0); + rc->count = 0; +} + + +static inline void rc_bit(lzma_range_encoder *rc, probability *prob, uint32_t bit) { rc->symbols[rc->count] = bit; @@ -132,6 +145,7 @@ rc_shift_low(lzma_range_encoder *rc, out[*out_pos] = rc->cache + (uint8_t)(rc->low >> 32); ++*out_pos; + ++rc->out_total; rc->cache = 0xFF; } while (--rc->cache_size != 0); @@ -146,6 +160,34 @@ rc_shift_low(lzma_range_encoder *rc, } +// NOTE: The last two arguments are uint64_t instead of size_t because in +// the dummy version these refer to the size of the whole range-encoded +// output stream, not just to the currently available output buffer space. +static inline bool +rc_shift_low_dummy(uint64_t *low, uint64_t *cache_size, uint8_t *cache, + uint64_t *out_pos, uint64_t out_size) +{ + if ((uint32_t)(*low) < (uint32_t)(0xFF000000) + || (uint32_t)(*low >> 32) != 0) { + do { + if (*out_pos == out_size) + return true; + + ++*out_pos; + *cache = 0xFF; + + } while (--*cache_size != 0); + + *cache = (*low >> 24) & 0xFF; + } + + ++*cache_size; + *low = (*low & 0x00FFFFFF) << RC_SHIFT_BITS; + + return false; +} + + static inline bool rc_encode(lzma_range_encoder *rc, uint8_t *out, size_t *out_pos, size_t out_size) @@ -222,6 +264,83 @@ rc_encode(lzma_range_encoder *rc, } +static inline bool +rc_encode_dummy(const lzma_range_encoder *rc, uint64_t out_limit) +{ + assert(rc->count <= RC_SYMBOLS_MAX); + + uint64_t low = rc->low; + uint64_t cache_size = rc->cache_size; + uint32_t range = rc->range; + uint8_t cache = rc->cache; + uint64_t out_pos = rc->out_total; + + size_t pos = rc->pos; + + while (true) { + // Normalize + if (range < RC_TOP_VALUE) { + if (rc_shift_low_dummy(&low, &cache_size, &cache, + &out_pos, out_limit)) + return true; + + range <<= RC_SHIFT_BITS; + } + + // This check is here because the normalization above must + // be done before flushing the last bytes. + if (pos == rc->count) + break; + + // Encode a bit + switch (rc->symbols[pos]) { + case RC_BIT_0: { + probability prob = *rc->probs[pos]; + range = (range >> RC_BIT_MODEL_TOTAL_BITS) + * prob; + break; + } + + case RC_BIT_1: { + probability prob = *rc->probs[pos]; + const uint32_t bound = prob * (range + >> RC_BIT_MODEL_TOTAL_BITS); + low += bound; + range -= bound; + break; + } + + case RC_DIRECT_0: + range >>= 1; + break; + + case RC_DIRECT_1: + range >>= 1; + low += range; + break; + + case RC_FLUSH: + default: + assert(0); + break; + } + + ++pos; + } + + // Flush the last bytes. This isn't in rc->symbols[] so we do + // it after the above loop to take into account the size of + // the flushing that will be done at the end of the stream. + for (pos = 0; pos < 5; ++pos) { + if (rc_shift_low_dummy(&low, &cache_size, + &cache, &out_pos, out_limit)) + return true; + } + + return false; +} + + static inline uint64_t rc_pending(const lzma_range_encoder *rc) { diff --git a/contrib/libs/lzma/liblzma/simple/arm.c b/contrib/libs/lzma/liblzma/simple/arm.c index ff5073ae58..6e53970d2f 100644 --- a/contrib/libs/lzma/liblzma/simple/arm.c +++ b/contrib/libs/lzma/liblzma/simple/arm.c @@ -53,6 +53,7 @@ arm_coder_init(lzma_next_coder *next, const lzma_allocator *allocator, } +#ifdef HAVE_ENCODER_ARM extern lzma_ret lzma_simple_arm_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -60,8 +61,10 @@ lzma_simple_arm_encoder_init(lzma_next_coder *next, { return arm_coder_init(next, allocator, filters, true); } +#endif +#ifdef HAVE_DECODER_ARM extern lzma_ret lzma_simple_arm_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -69,3 +72,4 @@ lzma_simple_arm_decoder_init(lzma_next_coder *next, { return arm_coder_init(next, allocator, filters, false); } +#endif diff --git a/contrib/libs/lzma/liblzma/simple/arm64.c b/contrib/libs/lzma/liblzma/simple/arm64.c new file mode 100644 index 0000000000..5e7f26562d --- /dev/null +++ b/contrib/libs/lzma/liblzma/simple/arm64.c @@ -0,0 +1,136 @@ +/////////////////////////////////////////////////////////////////////////////// +// +/// \file arm64.c +/// \brief Filter for ARM64 binaries +/// +/// This converts ARM64 relative addresses in the BL and ADRP immediates +/// to absolute values to increase redundancy of ARM64 code. +/// +/// Converting B or ADR instructions was also tested but it's not useful. +/// A majority of the jumps for the B instruction are very small (+/- 0xFF). +/// These are typical for loops and if-statements. Encoding them to their +/// absolute address reduces redundancy since many of the small relative +/// jump values are repeated, but very few of the absolute addresses are. +// +// Authors: Lasse Collin +// Jia Tan +// +// This file has been put into the public domain. +// You can do whatever you want with this file. +// +/////////////////////////////////////////////////////////////////////////////// + +#include "simple_private.h" + + +static size_t +arm64_code(void *simple lzma_attribute((__unused__)), + uint32_t now_pos, bool is_encoder, + uint8_t *buffer, size_t size) +{ + size_t i; + + // Clang 14.0.6 on x86-64 makes this four times bigger and 40 % slower + // with auto-vectorization that is enabled by default with -O2. + // Such vectorization bloat happens with -O2 when targeting ARM64 too + // but performance hasn't been tested. +#ifdef __clang__ +# pragma clang loop vectorize(disable) +#endif + for (i = 0; i + 4 <= size; i += 4) { + uint32_t pc = (uint32_t)(now_pos + i); + uint32_t instr = read32le(buffer + i); + + if ((instr >> 26) == 0x25) { + // BL instruction: + // The full 26-bit immediate is converted. + // The range is +/-128 MiB. + // + // Using the full range is helps quite a lot with + // big executables. Smaller range would reduce false + // positives in non-code sections of the input though + // so this is a compromise that slightly favors big + // files. With the full range only six bits of the 32 + // need to match to trigger a conversion. + const uint32_t src = instr; + instr = 0x94000000; + + pc >>= 2; + if (!is_encoder) + pc = 0U - pc; + + instr |= (src + pc) & 0x03FFFFFF; + write32le(buffer + i, instr); + + } else if ((instr & 0x9F000000) == 0x90000000) { + // ADRP instruction: + // Only values in the range +/-512 MiB are converted. + // + // Using less than the full +/-4 GiB range reduces + // false positives on non-code sections of the input + // while being excellent for executables up to 512 MiB. + // The positive effect of ADRP conversion is smaller + // than that of BL but it also doesn't hurt so much in + // non-code sections of input because, with +/-512 MiB + // range, nine bits of 32 need to match to trigger a + // conversion (two 10-bit match choices = 9 bits). + const uint32_t src = ((instr >> 29) & 3) + | ((instr >> 3) & 0x001FFFFC); + + // With the addition only one branch is needed to + // check the +/- range. This is usually false when + // processing ARM64 code so branch prediction will + // handle it well in terms of performance. + // + //if ((src & 0x001E0000) != 0 + // && (src & 0x001E0000) != 0x001E0000) + if ((src + 0x00020000) & 0x001C0000) + continue; + + instr &= 0x9000001F; + + pc >>= 12; + if (!is_encoder) + pc = 0U - pc; + + const uint32_t dest = src + pc; + instr |= (dest & 3) << 29; + instr |= (dest & 0x0003FFFC) << 3; + instr |= (0U - (dest & 0x00020000)) & 0x00E00000; + write32le(buffer + i, instr); + } + } + + return i; +} + + +static lzma_ret +arm64_coder_init(lzma_next_coder *next, const lzma_allocator *allocator, + const lzma_filter_info *filters, bool is_encoder) +{ + return lzma_simple_coder_init(next, allocator, filters, + &arm64_code, 0, 4, 4, is_encoder); +} + + +#ifdef HAVE_ENCODER_ARM64 +extern lzma_ret +lzma_simple_arm64_encoder_init(lzma_next_coder *next, + const lzma_allocator *allocator, + const lzma_filter_info *filters) +{ + return arm64_coder_init(next, allocator, filters, true); +} +#endif + + +#ifdef HAVE_DECODER_ARM64 +extern lzma_ret +lzma_simple_arm64_decoder_init(lzma_next_coder *next, + const lzma_allocator *allocator, + const lzma_filter_info *filters) +{ + return arm64_coder_init(next, allocator, filters, false); +} +#endif diff --git a/contrib/libs/lzma/liblzma/simple/armthumb.c b/contrib/libs/lzma/liblzma/simple/armthumb.c index a8da334a04..25d8dbd4f3 100644 --- a/contrib/libs/lzma/liblzma/simple/armthumb.c +++ b/contrib/libs/lzma/liblzma/simple/armthumb.c @@ -58,6 +58,7 @@ armthumb_coder_init(lzma_next_coder *next, const lzma_allocator *allocator, } +#ifdef HAVE_ENCODER_ARMTHUMB extern lzma_ret lzma_simple_armthumb_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -65,8 +66,10 @@ lzma_simple_armthumb_encoder_init(lzma_next_coder *next, { return armthumb_coder_init(next, allocator, filters, true); } +#endif +#ifdef HAVE_DECODER_ARMTHUMB extern lzma_ret lzma_simple_armthumb_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -74,3 +77,4 @@ lzma_simple_armthumb_decoder_init(lzma_next_coder *next, { return armthumb_coder_init(next, allocator, filters, false); } +#endif diff --git a/contrib/libs/lzma/liblzma/simple/ia64.c b/contrib/libs/lzma/liblzma/simple/ia64.c index 6492d0a384..692b0a295e 100644 --- a/contrib/libs/lzma/liblzma/simple/ia64.c +++ b/contrib/libs/lzma/liblzma/simple/ia64.c @@ -94,6 +94,7 @@ ia64_coder_init(lzma_next_coder *next, const lzma_allocator *allocator, } +#ifdef HAVE_ENCODER_IA64 extern lzma_ret lzma_simple_ia64_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -101,8 +102,10 @@ lzma_simple_ia64_encoder_init(lzma_next_coder *next, { return ia64_coder_init(next, allocator, filters, true); } +#endif +#ifdef HAVE_DECODER_IA64 extern lzma_ret lzma_simple_ia64_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -110,3 +113,4 @@ lzma_simple_ia64_decoder_init(lzma_next_coder *next, { return ia64_coder_init(next, allocator, filters, false); } +#endif diff --git a/contrib/libs/lzma/liblzma/simple/powerpc.c b/contrib/libs/lzma/liblzma/simple/powerpc.c index 0b60e9b3fe..3a340fd171 100644 --- a/contrib/libs/lzma/liblzma/simple/powerpc.c +++ b/contrib/libs/lzma/liblzma/simple/powerpc.c @@ -58,6 +58,7 @@ powerpc_coder_init(lzma_next_coder *next, const lzma_allocator *allocator, } +#ifdef HAVE_ENCODER_POWERPC extern lzma_ret lzma_simple_powerpc_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -65,8 +66,10 @@ lzma_simple_powerpc_encoder_init(lzma_next_coder *next, { return powerpc_coder_init(next, allocator, filters, true); } +#endif +#ifdef HAVE_DECODER_POWERPC extern lzma_ret lzma_simple_powerpc_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -74,3 +77,4 @@ lzma_simple_powerpc_decoder_init(lzma_next_coder *next, { return powerpc_coder_init(next, allocator, filters, false); } +#endif diff --git a/contrib/libs/lzma/liblzma/simple/simple_coder.h b/contrib/libs/lzma/liblzma/simple/simple_coder.h index 19c2ee03af..668a5092ad 100644 --- a/contrib/libs/lzma/liblzma/simple/simple_coder.h +++ b/contrib/libs/lzma/liblzma/simple/simple_coder.h @@ -61,6 +61,15 @@ extern lzma_ret lzma_simple_armthumb_decoder_init(lzma_next_coder *next, const lzma_filter_info *filters); +extern lzma_ret lzma_simple_arm64_encoder_init(lzma_next_coder *next, + const lzma_allocator *allocator, + const lzma_filter_info *filters); + +extern lzma_ret lzma_simple_arm64_decoder_init(lzma_next_coder *next, + const lzma_allocator *allocator, + const lzma_filter_info *filters); + + extern lzma_ret lzma_simple_sparc_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, const lzma_filter_info *filters); diff --git a/contrib/libs/lzma/liblzma/simple/sparc.c b/contrib/libs/lzma/liblzma/simple/sparc.c index 74b2655f36..bad8492ebc 100644 --- a/contrib/libs/lzma/liblzma/simple/sparc.c +++ b/contrib/libs/lzma/liblzma/simple/sparc.c @@ -65,6 +65,7 @@ sparc_coder_init(lzma_next_coder *next, const lzma_allocator *allocator, } +#ifdef HAVE_ENCODER_SPARC extern lzma_ret lzma_simple_sparc_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -72,8 +73,10 @@ lzma_simple_sparc_encoder_init(lzma_next_coder *next, { return sparc_coder_init(next, allocator, filters, true); } +#endif +#ifdef HAVE_DECODER_SPARC extern lzma_ret lzma_simple_sparc_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -81,3 +84,4 @@ lzma_simple_sparc_decoder_init(lzma_next_coder *next, { return sparc_coder_init(next, allocator, filters, false); } +#endif diff --git a/contrib/libs/lzma/liblzma/simple/x86.c b/contrib/libs/lzma/liblzma/simple/x86.c index 0e78909ccc..232b29542e 100644 --- a/contrib/libs/lzma/liblzma/simple/x86.c +++ b/contrib/libs/lzma/liblzma/simple/x86.c @@ -141,6 +141,7 @@ x86_coder_init(lzma_next_coder *next, const lzma_allocator *allocator, } +#ifdef HAVE_ENCODER_X86 extern lzma_ret lzma_simple_x86_encoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -148,8 +149,10 @@ lzma_simple_x86_encoder_init(lzma_next_coder *next, { return x86_coder_init(next, allocator, filters, true); } +#endif +#ifdef HAVE_DECODER_X86 extern lzma_ret lzma_simple_x86_decoder_init(lzma_next_coder *next, const lzma_allocator *allocator, @@ -157,3 +160,4 @@ lzma_simple_x86_decoder_init(lzma_next_coder *next, { return x86_coder_init(next, allocator, filters, false); } +#endif |