When copying with "rep movsb", if the distance between source and
destination is N*4GB + [1..63] with N >= 0, performance may be very
slow. This patch updates memmove-vec-unaligned-erms.S for AVX and
AVX512 versions with the distance in RCX:
cmpl $63, %ecx
// Don't use "rep movsb" if ECX <= 63
jbe L(Don't use rep movsb")
Use "rep movsb"
Benchtests data with bench-memcpy, bench-memcpy-large, bench-memcpy-random
and bench-memcpy-walk on Skylake, Ice Lake and Tiger Lake show that its
performance impact is within noise range as "rep movsb" is only used for
data size >= 4KB.
After sp is updated, the CFA offset should be set before next instruction.
Tested in glibc-2.28:
Thread 2 "xxxxxxx" hit Breakpoint 1, _dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:149
149 stp x1, x2, [sp, #-32]!
Missing separate debuginfos, use: dnf debuginfo-install libgcc-7.3.0-20190804.h24.aarch64
(gdb) bt
#0 _dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:149
#1 0x0000ffffbe4fbb44 in OurFunction (threadId=3194870184)
at /home/test/test_function.c:30
#2 0x0000000000400c08 in initaaa () at thread.c:58
#3 0x0000000000400c50 in thread_proc (param=0x0) at thread.c:71
#4 0x0000ffffbf6918bc in start_thread (arg=0xfffffffff29f) at pthread_create.c:486
#5 0x0000ffffbf5669ec in thread_start () at ../sysdeps/unix/sysv/linux/aarch64/clone.S:78
(gdb) ni
_dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:150
150 stp x3, x4, [sp, #16]
(gdb) bt
#0 _dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:150
#1 0x0000ffffbe4fbb44 in OurFunction (threadId=3194870184)
at /home/test/test_function.c:30
#2 0x0000000000000000 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)
(gdb) ni
_dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:157
157 mrs x4, tpidr_el0
(gdb) bt
#0 _dl_tlsdesc_dynamic () at ../sysdeps/aarch64/dl-tlsdesc.S:157
#1 0x0000ffffbe4fbb44 in OurFunction (threadId=3194870184)
at /home/test/test_function.c:30
#2 0x0000000000400c08 in initaaa () at thread.c:58
#3 0x0000000000400c50 in thread_proc (param=0x0) at thread.c:71
#4 0x0000ffffbf6918bc in start_thread (arg=0xfffffffff29f) at pthread_create.c:486
#5 0x0000ffffbf5669ec in thread_start () at ../sysdeps/unix/sysv/linux/aarch64/clone.S:78
Signed-off-by: liqingqing <liqingqing3@huawei.com>
Signed-off-by: Shuo Wang <wangshuo47@huawei.com>
Make the tests use TEST_COND_intel96 to decide on whether to build the
unnormal tests instead of the macro in nan-pseudo-number.h and then
drop the header inclusion. This unbreaks test runs on all
architectures that do not have ldbl-96.
Also drop the HANDLE_PSEUDO_NUMBERS macro since it is not used
anywhere.
I've updated copyright dates in glibc for 2021. This is the patch for
the changes not generated by scripts/update-copyrights and subsequent
build / regeneration of generated files. As well as the usual annual
updates, mainly dates in --version output (minus csu/version.c which
previously had to be handled manually but is now successfully updated
by update-copyrights), there is a small change to the copyright notice
in NEWS which should let NEWS get updated automatically next year.
Please remember to include 2021 in the dates for any new files added
in future (which means updating any existing uncommitted patches you
have that add new files to use the new copyright dates in them).
I used these shell commands:
../glibc/scripts/update-copyrights $PWD/../gnulib/build-aux/update-copyright
(cd ../glibc && git commit -am"[this commit message]")
and then ignored the output, which consisted lines saying "FOO: warning:
copyright statement not found" for each of 6694 files FOO.
I then removed trailing white space from benchtests/bench-pthread-locks.c
and iconvdata/tst-iconv-big5-hkscs-to-2ucs4.c, to work around this
diagnostic from Savannah:
remote: *** pre-commit check failed ...
remote: *** error: lines with trailing whitespace found
remote: error: hook declined to update refs/heads/master
DELOUSE was added to asm code to make them compatible with non-LP64
ABIs, but it is an unfortunate name and the code was not compatible
with ABIs where pointer and size_t are different. Glibc currently
only supports the LP64 ABI so these macros are not really needed or
tested, but for now the name is changed to be more meaningful instead
of removing them completely.
Some DELOUSE macros were dropped: clone, strlen and strnlen used it
unnecessarily.
The out of tree ILP32 patches are currently not maintained and will
likely need a rework to rebase them on top of the time64 changes.
clone already uses r31 to temporarily save input arguments before doing the
syscall, so we use a different register to read from the TCB. We can also avoid
allocating another stack frame, which is not needed since we can simply extend
the usage of the red zone.
Tested-by: Lucas A. M. Magalhães <lamm@linux.ibm.com>
Reviewed-by: Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
Linux kernel v5.9 added support for system calls using the scv
instruction for POWER9 and later. The new codepath provides better
performance (see below) if compared to using sc. For the
foreseeable future, both sc and scv mechanisms will co-exist, so this
patch enables glibc to do a runtime check and use scv when it is
available.
Before issuing the system call to the kernel, we check hwcap2 in the TCB
for PPC_FEATURE2_SCV to see if scv is supported by the kernel. If not,
we fallback to sc and keep the old behavior.
The kernel implements a different error return convention for scv, so
when returning from a system call we need to handle the return value
differently depending on the instruction we used to enter the kernel.
For syscalls implemented in ASM, entry and exit are implemented by
different macros (PSEUDO and PSEUDO_RET, resp.), which may be used in
sequence (e.g. for templated syscalls) or with other instructions in
between (e.g. clone). To avoid accessing the TCB a second time on
PSEUDO_RET to check which instruction we used, the value read from
hwcap2 is cached on a non-volatile register.
This is not needed when using INTERNAL_SYSCALL macro, since entry and
exit are bundled into the same inline asm directive.
The dynamic loader may issue syscalls before the TCB has been setup
so it always uses sc with no extra checks. For the static case, there
is no compile-time way to determine if we are inside startup code,
so we also check the value of the thread pointer before effectively
accessing the TCB. For such situations in which the availability of
scv cannot be determined, sc is always used.
Support for scv in syscalls implemented in their own ASM file (clone and
vfork) will be added later. For now simply use sc as before.
Average performance over 1M calls for each syscall "type":
- stat: C wrapper calling INTERNAL_SYSCALL
- getpid: templated ASM syscall
- syscall: call to gettid using syscall function
Standard:
stat : 1.573445 us / ~3619 cycles
getpid : 0.164986 us / ~379 cycles
syscall : 0.162743 us / ~374 cycles
With scv:
stat : 1.537049 us / ~3535 cycles <~ -84 cycles / -2.32%
getpid : 0.109923 us / ~253 cycles <~ -126 cycles / -33.25%
syscall : 0.116410 us / ~268 cycles <~ -106 cycles / -28.34%
Tested on powerpc, powerpc64, powerpc64le (with and without scv)
Tested-by: Lucas A. M. Magalhães <lamm@linux.ibm.com>
Reviewed-by: Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
Add support to treat pseudo-numbers specially and implement x86
version to consider all of them as signaling.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
With xmknod wrapper functions removed (589260cef8), the mknod functions
are now properly exported, and version is done using symbols versioning
instead of the extra _MKNOD_* argument.
It also allows us to consolidate Linux and Hurd mknod implementation.
Reviewed-by: Lukasz Majewski <lukma@denx.de>
With xstat wrapper functions removed (8ed005daf0), the stat functions
are now properly exported, and version is done using symbols versioning
instead of the extra _STAT_* argument.
Reviewed-by: Lukasz Majewski <lukma@denx.de>
The new __proc_waitid RPC now expects WEXITED to be passed, allowing to
properly implement waitid, and thus define the missing W* macros
(according to FreeBSD values).
Instead of having the arch-specific trampoline setup code detect whether
preemption happened or not, we'd rather pass it the sigaction. In the
future, this may also allow to change sa_flags from post_signal().
Do not attempt to fix the significand top bit in long double input
received in printf. The code should never reach here because isnan
should now detect unnormals as NaN. This is already a NOP for glibc
since it uses the gcc __builtin_isnan, which detects unnormals as NaN.
Reviewed-by: Florian Weimer <fweimer@redhat.com>
This syncs up isnanl behaviour with gcc. Also move the isnanl
implementation to sysdeps/x86 and remove the sysdeps/x86_64 version.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Also move sysdeps/i386/fpu/s_fpclassifyl.c to
sysdeps/x86/fpu/s_fpclassifyl.c and remove
sysdeps/x86_64/fpu/s_fpclassifyl.c
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Add Intel Linear Address Masking (LAM) support to <sys/platform/x86.h>.
HAS_CPU_FEATURE (LAM) can be used to detect if LAM is enabled in CPU.
LAM modifies the checking that is applied to 64-bit linear addresses,
allowing software to use of the untranslated address bits for metadata.
Add various defines and stubs for enabling MTE on AArch64 sysv-like
systems such as Linux. The HWCAP feature bit is copied over in the
same way as other feature bits. Similarly we add a new wrapper header
for mman.h to define the PROT_MTE flag that can be used with mmap and
related functions.
We add a new field to struct cpu_features that can be used, for
example, to check whether or not certain ifunc'd routines should be
bound to MTE-safe versions.
Finally, if we detect that MTE should be enabled (ie via the glibc
tunable); we enable MTE during startup as required.
Support in the Linux kernel was added in version 5.10.
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
Older versions of the Linux kernel headers obviously lack support for
memory tagging, but we still want to be able to build in support when
using those (obviously it can't be enabled on such systems).
The linux kernel extensions are made to the platform-independent
header (linux/prctl.h), so this patch takes a similar approach.
This patch adds the basic support for memory tagging.
Various flavours are supported, particularly being able to turn on
tagged memory at run-time: this allows the same code to be used on
systems where memory tagging support is not present without neededing
a separate build of glibc. Also, depending on whether the kernel
supports it, the code will use mmap for the default arena if morecore
does not, or cannot support tagged memory (on AArch64 it is not
available).
All the hooks use function pointers to allow this to work without
needing ifuncs.
Reviewed-by: DJ Delorie <dj@redhat.com>
This is clever, but it confuses downstream detection in at least zstd
and GNOME's glib. zstd has preprocessor tests for the 'st_mtime' macro,
which is not provided by the path using the anonymous union; glib checks
for the presence of 'st_mtimensec' in struct stat but then tries to
access that field in struct statx (which might be a bug on its own).
Checked with a build for alpha-linux-gnu.
setjmp() uses C code to store current registers into jmp_buf
environment. -fstack-protector-all places canary into setjmp()
prologue and clobbers 'a5' before it gets saved.
The change inhibits stack canary injection to avoid clobber.
When SA_SIGINFO is available, sysdeps/posix/s?profil.c use it, so we have to
fix the __profil_counter function accordingly, using sigcontextinfo.h's
sigcontext_get_pc.
SA_SIGINFO is actually just another way of expressing what we were
already passing over with struct sigcontext. This just introduces the
SIGINFO interface and fixes the posix values when that interface is
requested by the application.
asin and acos have slow paths for rounding the last bit that cause some
calls to be 500-1500x slower than average calls.
These slow paths are rare, a test of a trillion (1.000.000.000.000)
random inputs between -1 and 1 showed 32870 slow calls for acos and 4473
for asin, with most occurrences between -1.0 .. -0.9 and 0.9 .. 1.0.
The slow paths claim correct rounding and use __sin32() and __cos32()
(which compare two result candidates and return the closest one) as the
final step, with the second result candidate (res1) having a small offset
applied from res. This suggests that res and res1 are intended to be 1
ULP apart (which makes sense for rounding), barring bugs, allowing us to
pick either one and still remain within 1 ULP of the exact result.
Remove the slow paths as the accuracy is better than 1 ULP even without
them, which is enough for glibc.
Also remove code comments claiming correctly rounded results.
After slow path removal, checking the accuracy of 14.400.000.000 random
asin() and acos() inputs showed only three incorrectly rounded
(error > 0.5 ULP) results:
- asin(-0x1.ee2b43286db75p-1) (0.500002 ULP, same as before)
- asin(-0x1.f692ba202abcp-4) (0.500003 ULP, same as before)
- asin(-0x1.9915e876fc062p-1) (0.50000000001 ULP, previously exact)
The first two had the same error even before this commit, and they did
not use the slow path at all.
Checking 4934 known randomly found previously-slow-path asin inputs
shows 25 calls with incorrectly rounded results, with a maximum error of
0.500000002 ULP (for 0x1.fcd5742999ab8p-1). The previous slow-path code
rounded all these inputs correctly (error < 0.5 ULP).
The observed average speed increase was 130x.
Checking 36240 known randomly found previously-slow-path acos inputs
shows 42 calls with incorrectly rounded results, with a maximum error of
0.500000008 ULP (for 0x1.f63845056f35ep-1). The previous "exact"
slow-path code showed 34 calls with incorrectly rounded results, with the
same maximum error of 0.500000008 ULP (for 0x1.f63845056f35ep-1).
The observed average speed increase was 130x.
The functions could likely be trimmed more while keeping acceptable
accuracy, but this at least gets rid of the egregiously slow cases.
Tested on x86_64.
This patch updates the kernel version in the test tst-mman-consts.py
to 5.10. (There are no new MAP_* constants covered by this test in
5.10 that need any other header changes.)
Tested with build-many-glibcs.py.
GCC 6.5 fails to correctly build ldconfig with recent ld.so.cache
commits, e.g.:
785969a047
elf: Implement a string table for ldconfig, with tail merging
If glibc is build with gcc 6.5.0:
__builtin_add_overflow is used in
<glibc>/elf/stringtable.c:stringtable_finalize()
which leads to ldconfig failing with "String table is too large".
This is also recognizable in following tests:
FAIL: elf/tst-glibc-hwcaps-cache
FAIL: elf/tst-glibc-hwcaps-prepend-cache
FAIL: elf/tst-ldconfig-X
FAIL: elf/tst-ldconfig-bad-aux-cache
FAIL: elf/tst-ldconfig-ld_so_conf-update
FAIL: elf/tst-stringtable
See gcc "Bug 98269 - gcc 6.5.0 __builtin_add_overflow() with small
uint32_t values incorrectly detects overflow"
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98269)
Change sbrk to fail for !__libc_initial (in the generic
implementation). As a result, sbrk is (relatively) safe to use
for the __libc_initial case (from the main libc). It is therefore
no longer necessary to avoid using it in that case (or updating the
brk cache), and the __libc_initial flag does not need to be updated
as part of dlmopen or static dlopen.
As before, direct brk system calls on Linux may lead to memory
corruption.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Linux 5.10 has one new syscall, process_madvise. Update
syscall-names.list and regenerate the arch-syscall.h headers with
build-many-glibcs.py update-syscalls.
Tested with build-many-glibcs.py.
