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.
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.
lzma_raw_encoder() and lzma_raw_encoder_init() used "options" as the
parameter name instead of "filters" (used by the declaration). "filters"
is more clear since the parameter represents the list of filters passed
to the raw encoder, each of which contains filter options.
In ELF shared libs:
-fvisibility=hidden affects definitions of symbols but not
declarations.[*] This doesn't affect direct calls to functions
inside liblzma as a linker can replace a call to lzma_foo@plt
with a call directly to lzma_foo when -fvisibility=hidden is used.
[*] It has to be like this because otherwise every installed
header file would need to explictly set the symbol visibility
to default.
When accessing extern variables that aren't defined in the
same translation unit, compiler assumes that the variable has
the default visibility and thus indirection is needed. Unlike
function calls, linker cannot optimize this.
Using __attribute__((__visibility__("hidden"))) with the extern
variable declarations tells the compiler that indirection isn't
needed because the definition is in the same shared library.
About 15+ years ago, someone told me that it would be good if
the CRC tables would be defined in the same translation unit
as the C code of the CRC functions. While I understood that it
could help a tiny amount, I didn't want to change the code because
a separate translation unit for the CRC tables was needed for the
x86 assembly code anyway. But when visibility attributes are
supported, simply marking the extern declaration with the
hidden attribute will get identical result. When there are only
a few affected variables, this is trivial to do. I wish I had
understood this back then already.
xrealloc() is obviously incorrect, modern GCC docs even
mention realloc() as an example where this attribute
cannot be used.
liblzma's lzma_alloc() and lzma_alloc_zero() would be
correct uses most of the time but custom allocators
may use a memory pool or otherwise hold the pointer
so aliasing issues could happen in theory.
The xstrdup() case likely was correct but I removed it anyway.
Now there are no __malloc__ attributes left in the code.
The allocations aren't in hot paths so this should make
no practical difference.
The argument to vli_ceil4() should always guarantee the return value
is also a valid lzma_vli. Thus the highest three valid lzma_vli values
are invalid arguments. All uses of the function ensure this so the
assert is updated to match this.
This was not a security bug since there was no path to overflow
UINT64_MAX in lzma_index_append() or when it calls index_file_size().
The bug was discovered by a failing assert() in vli_ceil4() when called
from index_file_size() when unpadded_sum (the sum of the compressed size
of current Stream and the unpadded_size parameter) exceeds LZMA_VLI_MAX.
Previously, the unpadded_size parameter was checked to be not greater
than UNPADDED_SIZE_MAX, but no check was done once compressed_base was
added.
This could not have caused an integer overflow in index_file_size() when
called by lzma_index_append(). The calculation for file_size breaks down
into the sum of:
- Compressed base from all previous Streams
- 2 * LZMA_STREAM_HEADER_SIZE (size of the current Streams header and
footer)
- stream_padding (can be set by lzma_index_stream_padding())
- Compressed base from the current Stream
- Unpadded size (parameter to lzma_index_append())
The sum of everything except for Unpadded size must be less than
LZMA_VLI_MAX. This is guarenteed by overflow checks in the functions
that can set these values including lzma_index_stream_padding(),
lzma_index_append(), and lzma_index_cat(). The maximum value for
Unpadded size is enforced by lzma_index_append() to be less than or
equal UNPADDED_SIZE_MAX. Thus, the sum cannot exceed UINT64_MAX since
LZMA_VLI_MAX is half of UINT64_MAX.
Thanks to Joona Kannisto for reporting this.
Commit 78704f36e7 added an empty
initializer {} to prevent a warning. The empty initializer is a GNU
extension and results in a build failure on MSVC. The -wpedantic flag
warns about empty initializers.
This change only impacts the compiler warning since it was impossible
for the wait_abs struct in stream_encode_mt() to be used before it was
initialized since mythread_condtime_set() will always be called before
mythread_cond_timedwait().
Since the mythread.h code is different between the POSIX and
Windows versions, this warning was only present on Windows builds.
Thanks to Arthur S for reporting the warning and providing an initial
patch.
In lzma_memcmplen(), the <intrin.h> header file is only included if
_MSC_VER and _M_X64 are both defined but _BitScanForward64() was
previously used if _M_X64 was defined. GCC for MSYS2 defines _M_X64 but
not _MSC_VER so _BitScanForward64() was used without including
<intrin.h>.
Now, lzma_memcmplen() will use __builtin_ctzll() for MSYS2 GCC builds as
expected.
The lzma_mt_block_size() was previously just an internal function for
the multithreaded .xz encoder. It is used to provide a recommended Block
size for a given filter chain.
This function is helpful to determine the maximum Block size for the
multithreaded .xz encoder when one wants to change the filters between
blocks. Then, this determined Block size can be provided to
lzma_stream_encoder_mt() in the lzma_mt options parameter when
intializing the coder. This requires one to know all the filter chains
they are using before starting to encode (or at least the filter chain
that will need the largest Block size), but that isn't a bad limitation.
Legacy Windows did not need to #include <intrin.h> to use the MSVC
intrinsics. Newer versions likely just issue a warning, but the MSVC
documentation says to include the header file for the intrinsics we use.
GCC and Clang can "pretend" to be MSVC on Windows, so extra checks are
needed in tuklib_integer.h to only include <intrin.h> when it will is
actually needed.
In the C99 and C17 standards, section 6.5.6 paragraph 8 means that
adding 0 to a null pointer is undefined behavior. As of writing,
"clang -fsanitize=undefined" (Clang 15) diagnoses this. However,
I'm not aware of any compiler that would take advantage of this
when optimizing (Clang 15 included). It's good to avoid this anyway
since compilers might some day infer that pointer arithmetic implies
that the pointer is not NULL. That is, the following foo() would then
unconditionally return 0, even for foo(NULL, 0):
void bar(char *a, char *b);
int foo(char *a, size_t n)
{
bar(a, a + n);
return a == NULL;
}
In contrast to C, C++ explicitly allows null pointer + 0. So if
the above is compiled as C++ then there is no undefined behavior
in the foo(NULL, 0) call.
To me it seems that changing the C standard would be the sane
thing to do (just add one sentence) as it would ensure that a huge
amount of old code won't break in the future. Based on web searches
it seems that a large number of codebases (where null pointer + 0
occurs) are being fixed instead to be future-proof in case compilers
will some day optimize based on it (like making the above foo(NULL, 0)
return 0) which in the worst case will cause security bugs.
Some projects don't plan to change it. For example, gnulib and thus
many GNU tools currently require that null pointer + 0 is defined:
https://lists.gnu.org/archive/html/bug-gnulib/2021-11/msg00000.htmlhttps://www.gnu.org/software/gnulib/manual/html_node/Other-portability-assumptions.html
In XZ Utils null pointer + 0 issue should be fixed after this
commit. This adds a few if-statements and thus branches to avoid
null pointer + 0. These check for size > 0 instead of ptr != NULL
because this way bugs where size > 0 && ptr == NULL will likely
get caught quickly. None of them are in hot spots so it shouldn't
matter for performance.
A little less readable version would be replacing
ptr + offset
with
offset != 0 ? ptr + offset : ptr
or creating a macro for it:
#define my_ptr_add(ptr, offset) \
((offset) != 0 ? ((ptr) + (offset)) : (ptr))
Checking for offset != 0 instead of ptr != NULL allows GCC >= 8.1,
Clang >= 7, and Clang-based ICX to optimize it to the very same code
as ptr + offset. That is, it won't create a branch. So for hot code
this could be a good solution to avoid null pointer + 0. Unfortunately
other compilers like ICC 2021 or MSVC 19.33 (VS2022) will create a
branch from my_ptr_add().
Thanks to Marcin Kowalczyk for reporting the problem:
https://github.com/tukaani-project/xz/issues/36
The bug is only a problem in applications that do not properly terminate
the filters[] array with LZMA_VLI_UNKNOWN or have more than
LZMA_FILTERS_MAX filters. This bug does not affect xz.
This is similar to 2ce4f36f17.
The actual initialization of the variables is done inside
mythread_sync() macro. Clang doesn't seem to see that
the initialization code inside the macro is always executed.
common/index.h is needed by liblzma internally and tests. common.h will
include and define many things that are not needed by the tests. Also,
this prevents include order problems because common.h will redefine
LZMA_API resulting in a warning.
Using return_if_error on lzma_lzma_lclppb_encode was improper because
return_if_error is expecting an lzma_ret value, but
lzma_lzma_lclppb_encode returns a boolean. This could result in
lzma_microlzma_encoder, which would be misleading for applications.
The API docs gave an impression that such checks are done
but they actually weren't done. In practice it made little
difference since the calling code has a bug if these are NULL.
Thanks to Jia Tan for the original patch that checked for
block->filters == NULL.
If someone sets up Clang to define __GNUC__ to 10 or greater
then symvers broke. __has_attribute is supported by such GCC
and Clang versions that don't support __symver__ so this should
be much better and simpler way to detect if __symver__ is
actually supported.
Thanks to Tomasz Gajc for the bug report.
lzma_str_to_filters() uses static error messages which makes
them not very precise. It tells the position in the string
where an error occurred though which helps quite a bit if
applications take advantage of it. Dynamic error messages can
be added later with a new flag if it seems important enough.
Some file formats need support for LZMA1 streams that don't use
the end of payload marker (EOPM) alias end of stream (EOS) marker.
So far liblzma API has supported decompressing such streams via
lzma_alone_decoder() when .lzma header specifies a known
uncompressed size. Encoding support hasn't been available in the API.
Instead of adding a new LZMA1-only API for this purpose, this commit
adds a new filter ID for use with raw encoder and decoder. The main
benefit of this approach is that then also filter chains are possible,
for example, if someone wants to implement support for .7z files that
use the x86 BCJ filter with LZMA1 (not BCJ2 as that isn't supported
in liblzma).
It not only makes no sense to put symbol versions into a static library
but it can also cause breakage.
By default Libtool #defines PIC if building a shared library and
doesn't define it for static libraries. This is documented in the
Libtool manual. It can be overriden using --with-pic or --without-pic.
configure.ac detects if --with-pic or --without-pic is used and then
gives an error if neither --disable-shared nor --disable-static was
used at the same time. Thus, in normal situations it works to build
both shared and static library at the same time on GNU/Linux,
only --with-pic or --without-pic requires that only one type of
library is built.
Thanks to John Paul Adrian Glaubitz from Debian for reporting
the problem that occurred on ia64:
https://www.mail-archive.com/xz-devel@tukaani.org/msg00610.html
lzma_filters_free() sets the options to NULL and ids to
LZMA_VLI_UNKNOWN so there is no need to do it by caller;
the filter arrays will always be left in a safe state.
Also use memcpy() instead of a loop to copy a filter chain
when it is known to be safe to copy LZMA_FILTERS_MAX + 1
(even if the elements past the terminator might be uninitialized).
This time it can happen when lzma_stream_encoder_mt() is used
to reinitialize an existing multi-threaded Stream encoder
and one of 1-4 tiny allocations in lzma_filters_copy() fail.
It's very similar to the previous bug
10430fbf38, happening with
an array of lzma_filter structures whose old options are freed
but the replacement never arrives due to a memory allocation
failure in lzma_filters_copy().
The documentation mentions that lzma_block_encoder() supports
LZMA_SYNC_FLUSH but it was never added to supported_actions[]
in the internal structure. Because of this, LZMA_SYNC_FLUSH could
not be used with the Block encoder unless it was the next coder
after something like stream_encoder() or stream_encoder_mt().
Jia Tan
Lasse Collin
Dimitri Papadopoulos Orfanos