NVHPC compiler has several issues that make it impossible to
build liblzma:
- the compiler cannot handle unions that contain pointers that
are not the first members;
- the compiler cannot handle the assembler code in range_decoder.h
(LZMA_RANGE_DECODER_CONFIG has to be set to zero);
- the compiler fails to produce valid code for delta_decode if the
vectorization is enabled, which results in failed tests.
This introduces NVHPC-specific workarounds that address the issues.
With GCC and a certain combination of flags, Valgrind will falsely
trigger an invalid write. This appears to be due to the omission of
instructions to properly save, set up, and restore the frame pointer.
The IFUNC resolver is a leaf function since it only calls a function
that is inlined. So sometimes GCC omits the frame pointer instructions
in the resolver unless this optimization is explictly disabled.
This fixes https://bugzilla.redhat.com/show_bug.cgi?id=2267598.
Now that multi threaded encoding is the default, users do not need to
see a warning message everytime the number of threads is reduced. On
some machines, this could happen very often. It is not unreasonable for
users to need to set double verbose mode to see this kind of
information.
To see these warning messages -vv or --verbose --verbose must be passed
to set xz into the highest possible verbosity mode.
These warnings had caused automated testing frameworks to fail when they
expected no output to stderr.
Thanks to Sebastian Andrzej Siewior for reporting this and for the
initial version of the patch.
The previous Linux Landlock feature test assumed that having the
linux/landlock.h header file was enough. The new feature tests also
requires that prctl() and the required Landlock system calls are
supported.
Like 5.5.0alpha, 5.7.0alpha won't be released, it's just to mark that
the branch is not stable.
Once again there is no API/ABI stability for new features in devel
versions. The major soname won't be bumped even if API/ABI of new
features breaks between devel releases.
If xz is given a directory, it should look like this:
$ xz /usr/bin
xz: /usr/bin: Is a directory, skipping
The Landlock rules didn't allow opening directories for reading:
$ xz /usr/bin
xz: /usr/bin: Permission denied
The simplest fix was to allow opening directories for reading.
While it's a bit silly to allow it solely for the error message,
it shouldn't make the sandbox significantly weaker.
The single-file use case (like when called from GNU tar) is
still as strict as possible: all Landlock restrictions are
enabled before (de)compression starts.
This makes these sandboxing methods stricter when no files are
created or deleted. That is, it's a middle ground between the
initial sandbox and the strictest single-file-to-stdout sandbox:
this allows opening files for reading but output has to go to stdout.
Linux 6.7 added support for ABI version 4 which restricts
TCP connections which xz won't need and thus those can be
forbidden now. Since the ABI version is handled at runtime,
supporting version 4 won't cause any compatibility issues.
Note that new enough kernel headers are required to get
version 4 support enabled at build time.
Landlock is now always used just like pledge(2) is: first in more
permissive mode and later (under certain common conditions) in
a strict mode that doesn't allow opening more files.
I put pledge(2) first in sandbox.c because it's the simplest API
to use and still somewhat fine-grained for basic applications.
So it's the simplest thing to understand for anyone reading sandbox.c.
Also explicitly initialize progress_automatic to make it clear
that it can be read before message_init() sets it. Static variable
was initialized to false by default already so this is only for
clarity.
GCC docs promise that it works and a few other compilers do
too. Clang/LLVM is documented source code only but unsurprisingly
it behaves the same as others on x86-64 at least. But the
certainly-portable way is good enough here so use that.
The x32 port has a x86-64 ABI in term of all registers but uses only
32bit pointer like x86-32. The assembly optimisation fails to compile on
x32. Given the state of x32 I suggest to exclude it from the
optimisation rather than trying to fix it.
Signed-off-by: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
It's used only for basic bittrees and fixed-size reverse bittree
because those showed a clear benefit on x86-64 with GCC and Clang.
The other methods were more mixed and thus are commented out but
they should be tested on other archs.
Now extra buffer space is reserved so that repeating bytes for
any single match will never need to copy from two places (both
the beginning and the end of the buffer). This simplifies
dict_repeat() and helps a little with speed.
This seems to reduce .lzma decompression time about 2 %, so
with .xz and CRC it could be slightly less. The small things
add up still.
It's not completely obvious if this is better in the decoder.
It should be good if compiler can avoid creating a branch
(like using CMOV on x86).
This also makes lzma_encoder.c use the new macros.
The new decoder resumes the first decoder loop in the Resumable mode.
Then, the code executes in Non-resumable mode until it detects that it
cannot guarantee to have enough input/output to decode another symbol.
The Resumable mode is how the decoder has always worked. Before decoding
every input bit, it checks if there is enough space and will save its
location to be resumed later. When the decoder has more input/output,
it jumps back to the correct sequence in the Resumable mode code.
When the input/output buffers are large, the Resumable mode is much
slower than the Non-resumable because it has more branches and is harder
for the compiler to optimize since it is in a large switch block.
Early benchmarking shows significant time improvement (8-10% on gcc and
clang x86) by using the Non-resumable code as much as possible.
The new "safe" range decoder mode is the same as old range decoder, but
now the default behavior of the range decoder will not check if there is
enough input or output to complete the operation. When the buffers are
close to fully consumed, the "safe" operations must be used instead. This
will improve speed because it will reduce the number of branches needed
for most of the range decoder operations.
The main reason is a kind of silly one:
xz-man.pot contains strings from all man pages in XZ Utils.
The man pages of xzdiff, xzgrep, and xzmore were under GPLv2
and the rest under 0BSD. Thus xz-man.pot contained strings
under two licences. po4a creates the translated man pages
from the combined 0BSD+GPLv2 xz-man.pot.
I haven't liked this mixing in xz-man.pot but the
Translation Project requires that all man pages must be
in the same .pot file. So a separate xz-man-gpl.pot
wasn't an option.
Since these man pages are short, rewriting them was quick enough.
Now xz-man.pot is entirely under 0BSD and marking the per-file
licenses is simpler.
As a bonus, some wording hopefully is now slightly better
although it's perhaps a matter of taste.
NOTE: In xzgrep.1, the EXIT STATUS section was written by me
in the commit d796b6d7fd so that's
why that section could be taken as is from the old xzgrep.1.
Perhaps the generated files aren't even copyrightable but
using the same license for them as for the rest of the liblzma
keeps things more consistent for tools that look for license info.
The initial commit 5d018dc035
in 2007 had a comment in sha256.c that the code is based on
Crypto++ Library 5.5.1. In 2009 the Authors list in sha256.c
and the AUTHORS file was updated with information that the
code had come from Crypto++ but via 7-Zip. I know I had viewed
7-Zip's SHA-256 code but back then the C code has been identical
enough with Crypto++, so I don't why I thought the author info
would need that extra step via 7-Zip for this single file.
Another error is that I had mixed sha.* and shacal2.* files
when checking for author info in Crypto++. The shacal2.* files
aren't related to liblzma's sha256.c and thus Kevin Springle's
code in Crypto++ isn't either.
If liblzma is configured with --disable-clmul-crc
CFLAGS="-msse4.1 -mpclmul", then it will fail to compile because the
generic version must be used but the CRC tables were not included.
The code was using HAVE_FUNC_ATTRIBUTE_IFUNC instead of CRC_USE_IFUNC.
With ARM64, ifunc is incompatible because it requires non-inline
function calls for runtime detection.
Even though the proper name for the architecture is aarch64, this
project uses ARM64 throughout. So the rename is for consistency.
Additionally, crc32_arm64.h was slightly refactored for the following
changes:
* Added MSVC, FreeBSD, and macOS support in
is_arch_extension_supported().
* crc32_arch_optimized() now checks the size when aligning the
buffer.
* crc32_arch_optimized() loop conditions were slightly modified to
avoid both decrementing the size and incrementing the buffer
pointer.
* Use the intrinsic wrappers defined in <arm_acle.h> because GCC and
Clang name them differently.
* Minor spacing and comment changes.
The CRC_GENERIC is now split into CRC32_GENERIC and CRC64_GENERIC, since
the ARM64 optimizations will be different between CRC32 and CRC64.
For the same reason, CRC_ARCH_OPTIMIZED is split into
CRC32_ARCH_OPTIMIZED and CRC64_ARCH_OPTIMIZED.
ifunc will only be used with x86-64 CLMUL because the runtime detection
methods needed with ARM64 are not compatible with ifunc.
The CRC32 instructions in ARM64 can calculate the CRC32 result
for 8 bytes in a single operation, making the use of ARM64
instructions much faster compared to the general CRC32 algorithm.
Optimized CRC32 will be enabled if ARM64 has CRC extension
running on Linux.
Signed-off-by: Chenxi Mao <chenxi.mao2013@gmail.com>
This hopefully does more good than bad:
+ It's faster by default.
+ Only the threaded compressor creates files that
can be decompressed in threaded mode.
- Compression ratio is worse, usually not too much though.
When it matters, -T1 must be used.
- Memory usage increases.
- Scripts that assume single-threaded mode but don't use -T1 will
possibly use too much resources, for example, if they run
multiple xz processes in parallel to compress multiple files.
- Output from single-threaded and multi-threaded compressors
differ but such changes could happen for other reasons too
(they just haven't happened since 5.0.0).
Not all RISC-V processors support fast unaligned access so
it's better to read only one byte in the main loop. This can
be faster even on x86-64 when compared to reading 32 bits at
a time as half the time the address is only 16-bit aligned.
The downside is larger code size on archs that do support
fast unaligned access.
Version 5.6.0 will be shown, even though upcoming alphas and betas
will be able to support this filter. 5.6.0 looks nicer in the output and
people shouldn't be encouraged to use an unstable version in production
in any way.
The new Filter ID is 0x0B.
Thanks to Chien Wong <m@xv97.com> for the initial version of the Filter,
the xz CLI updates, and the Autotools build system modifications.
Thanks to Igor Pavlov for his many contributions to the design of
the filter.
Now crc_simd_body() in crc_x86_clmul.h is only called once
in a translation unit, we no longer need to be so cautious
about ensuring the always-inline behavior.
CRC_CLMUL was split to CRC_ARCH_OPTIMIZED and CRC_X86_CLMUL.
CRC_ARCH_OPTIMIZED is defined when an arch-optimized version is used.
Currently the x86 CLMUL implementations are the only arch-optimized
versions, and these also use the CRC_x86_CLMUL macro to tell when
crc_x86_clmul.h needs to be included.
is_clmul_supported() was renamed to is_arch_extension_supported().
crc32_clmul() and crc64_clmul() were renamed to
crc32_arch_optimized() and crc64_arch_optimized().
This way the names make sense with arch-specific non-CLMUL
implementations as well.
A CLMUL-only build will have the crcxx_clmul() inlined into
lzma_crcxx(). Previously a jump to the extern lzma_crcxx_clmul()
was needed. Notes about shared liblzma on ELF platforms:
- On platforms that support ifunc and -fvisibility=hidden, this
was silly because CLMUL-only build would have that single extra
jump instruction of extra overhead.
- On platforms that support neither -fvisibility=hidden nor linker
version script (liblzma*.map), jumping to lzma_crcxx_clmul()
would go via PLT so a few more instructions of overhead (still
not a big issue but silly nevertheless).
There was a downside with static liblzma too: if an application only
needs lzma_crc64(), static linking would make the linker include the
CLMUL code for both CRC32 and CRC64 from crc_x86_clmul.o even though
the CRC32 code wouldn't be needed, thus increasing code size of the
executable (assuming that -ffunction-sections isn't used).
Also, now compilers are likely to inline crc_simd_body()
even if they don't support the always_inline attribute
(or MSVC's __forceinline). Quite possibly all compilers
that build the code do support such an attribute. But now
it likely isn't a problem even if the attribute wasn't supported.
Now all x86-specific stuff is in crc_x86_clmul.h. If other archs
The other archs can then have their own headers with their own
is_clmul_supported() and crcxx_clmul().
Another bonus is that the build system doesn't need to care if
crc_clmul.c is needed.
is_clmul_supported() stays as inline function as it's not needed
when doing a CLMUL-only build (avoids a warning about unused function).
It requires fast unaligned access to 64-bit integers
and a fast instruction to count leading zeros in
a 64-bit integer (__builtin_ctzll()). This perhaps
should be enabled on some other archs too.
Thanks to Chenxi Mao for the original patch:
https://github.com/tukaani-project/xz/pull/75 (the first commit)
According to the numbers there, this may improve encoding
speed by about 3-5 %.
This enables the 8-byte method on MSVC ARM64 too which
should work but wasn't tested.
A very strict sandbox is used when the last file is decompressed. The
likely most common use case of xzdec is to decompress a single file.
The Pledge sandbox is applied to the entire process with slightly more
relaxed promises, until the last file is processed.
Thanks to Christian Weisgerber for the initial patch adding Pledge
sandboxing.
This fixes the recent change to lzma_lz_encoder that used memzero
instead of the NULL constant. On some compilers the NULL constant
(always 0) may not equal the NULL pointer (this only needs to guarentee
to not point to valid memory address).
Later code compares the pointers to the NULL pointer so we must
initialize them with the NULL pointer instead of 0 to guarentee
code correctness.
The first member of lzma_lz_encoder doesn't necessarily need to be set
to NULL since it will always be set before anything tries to use it.
However the function pointer members must be set to NULL since other
functions rely on this NULL value to determine if this behavior is
supported or not.
This fixes a somewhat serious bug, where the options_update() and
set_out_limit() function pointers are not set to NULL. This seems to
have been forgotten since these function pointers were added many years
after the original two (code() and end()).
The problem is that by not setting this to NULL we are relying on the
memory allocation to zero things out if lzma_filters_update() is called
on a LZMA1 encoder. The function pointer for set_out_limit() is less
serious because there is not an API function that could call this in an
incorrect way. set_out_limit() is only called by the MicroLZMA encoder,
which must use LZMA1 where set_out_limit() is always set. Its currently
not possible to call set_out_limit() on an LZMA2 encoder at this time.
So calling lzma_filters_update() on an LZMA1 encoder had undefined
behavior since its possible that memory could be manipulated so the
options_update member pointed to a different instruction sequence.
This is unlikely to be a bug in an existing application since it relies
on calling lzma_filters_update() on an LZMA1 encoder in the first place.
For instance, it does not affect xz because lzma_filters_update() can
only be used when encoding to the .xz format.
This is fixed by using memzero() to set all members of lzma_lz_encoder
to NULL after it is allocated. This ensures this mistake will not occur
here in the future if any additional function pointers are added.
Lasse Collin
Sergey Kosukhin
Jia Tan
Chien Wong
Sebastian Andrzej Siewior
Chenxi Mao