Florian Weimer [Thu, 18 Aug 2022 06:49:54 +0000 (08:49 +0200)]
stdio-common: Convert vfprintf and related functions to buffers
vfprintf is entangled with vfwprintf (of course), __printf_fp,
__printf_fphex, __vstrfmon_l_internal, and the strfrom family of
functions. The latter use the internal snprintf functionality,
so vsnprintf is converted as well.
The simples conversion is __printf_fphex, followed by
__vstrfmon_l_internal and __printf_fp, and finally
__vfprintf_internal and __vfwprintf_internal. __vsnprintf_internal
and strfrom* are mostly consuming the new interfaces, so they
are comparatively simple.
__printf_fp is a public symbol, so the FILE *-based interface
had to preserved.
The __printf_fp rewrite does not change the actual binary-to-decimal
conversion algorithm, and digits are still not emitted directly to
the target buffer. However, the staging buffer now uses bytes
instead of wide characters, and one buffer copy is eliminated.
The changes are at least performance-neutral in my testing.
Floating point printing and snprintf improved measurably, so that
this Lua script
runs about 5% faster for me. To preserve fprintf performance for
a simple "%d" format, this commit has some logic changes under
LABEL (unsigned_number) to avoid additional function calls. There
are certainly some very easy performance improvements here: binary,
octal and hexadecimal formatting can easily avoid the temporary work
buffer (the number of digits can be computed ahead-of-time using one
of the __builtin_clz* built-ins). Decimal formatting can use a
specialized version of _itoa_word for base 10.
The existing (inconsistent) width handling between strfmon and printf
is preserved here. __print_fp_buffer_1 would have to use
__translated_number_width to achieve ISO conformance for printf.
Test expectations in libio/tst-vtables-common.c are adjusted because
the internal staging buffer merges all virtual function calls into
one.
In general, stack buffer usage is greatly reduced, particularly for
unbuffered input streams. __printf_fp can still use a large buffer
in binary128 mode for %g, though.
Florian Weimer [Thu, 18 Aug 2022 06:49:54 +0000 (08:49 +0200)]
stdio-common: Add __printf_function_invoke
And __wprintf_function_invoke. These functions will be used to
to call registered printf specifier callbacks on printf buffers
after vfprintf and vfwprintf have been converted to buffers. The new
implementation avoids alloca/variable length arrays.
Florian Weimer [Thu, 18 Aug 2022 06:49:54 +0000 (08:49 +0200)]
stdio-common: Introduce buffers for implementing printf
These buffers will eventually be used instead of FILE * objects
to implement printf functions. The multibyte buffer is struct
__printf_buffer, the wide buffer is struct __wprintf_buffer.
To enable writing type-generic code, the header files
printf_buffer-char.h and printf_buffer-wchar_t.h define the
Xprintf macro differently, enabling Xprintf (buffer) to stand
for __printf_buffer and __wprintf_buffer as appropriate. For
common cases, macros like Xprintf_buffer are provided as a more
syntactically convenient shortcut.
Buffer-specific flush callbacks are implemented with a switch
statement instead of a function pointer, to avoid hardening issues
similar to those of libio vtables. struct __printf_buffer_as_file
is needed to support custom printf specifiers because the public
interface for that requires passing a FILE *, which is why there
is a trapdoor back from these buffers to FILE * streams.
Since the immediate user of these interfaces knows when processing
has finished, there is no flush callback for the end of processing,
only a flush callback for the intermediate buffer flush.
nscd: Use 64 bit time_t on libc nscd routines (BZ# 29402)
Although the nscd module is built with 64 bit time_t, the routines
linked direct to libc.so need to use the internal symbols. Reviewed-by: DJ Delorie <dj@redhat.com>
realloc: Return unchanged if request is within usable size
If there is enough space in the chunk to satisfy the new size, return
the old pointer as is, thus avoiding any locks or reallocations. The
only real place this has a benefit is in large chunks that tend to get
satisfied with mmap, since there is a large enough spare size (up to a
page) for it to matter. For allocations on heap, the extra size is
typically barely a few bytes (up to 15) and it's unlikely that it would
make much difference in performance.
Also added a smoke test to ensure that the old pointer is returned
unchanged if the new size to realloc is within usable size of the old
pointer.
Signed-off-by: Siddhesh Poyarekar <siddhesh@sourceware.org> Reviewed-by: DJ Delorie <dj@redhat.com>
The generic (sysdeps/unix/sysv/linux/generic/bits/typesizes.h) and
default (bits/typesizes.h) differs in two fields:
bits/typesizes.h Linux generic
__NLINK_T_TYPE __UWORD_TYPE __U32_TYPE
__BLKSIZE_T_TYPE __SLONGWORD_TYPE __S32_TYPE
Sinceit leads to different C++ mangling names, the default typesize.h
is copied for the requires archtiectures and the generic is make the
default Linux one. Reviewed-by: Carlos O'Donell <carlos@redhat.com> Tested-by: Carlos O'Donell <carlos@redhat.com>
It is currently used for csky, arc, nios2, and or1k. Newer 64 bit
architecture, like riscv32 and loongarch, reimplement it to override
F_GETLK64/F_SETLK64/F_SETLKW64. Reviewed-by: Carlos O'Donell <carlos@redhat.com> Tested-by: Carlos O'Donell <carlos@redhat.com>
The includes chain is added on each architecture sysdep.h and
the __NR__llseek hack is moved to lseek.c and lseek64.c. Reviewed-by: Carlos O'Donell <carlos@redhat.com> Tested-by: Carlos O'Donell <carlos@redhat.com>
This is similar to other LFS consolidation, where the non-LFS is only
built if __OFF_T_MATCHES_OFF64_T is not defined and the LFS version
is aliased to non-LFS name if __OFF_T_MATCHES_OFF64_T is defined.
For non-LFS variant, use sendfile syscall if defined, otherwise use
sendfile64 plus the offset overflow check (as generic implementation).
Checked on x86_64-linux-gnu. Reviewed-by: Carlos O'Donell <carlos@redhat.com> Tested-by: Carlos O'Donell <carlos@redhat.com>
Use inotify_init syscall if defined, otherwise use inotify_init1. Reviewed-by: Carlos O'Donell <carlos@redhat.com> Tested-by: Carlos O'Donell <carlos@redhat.com>
Use epoll_create syscall if defined, otherwise use epoll_create1. Reviewed-by: Florian Weimer <fweimer@redhat.com> Tested-by: Carlos O'Donell <carlos@redhat.com>
Use the generic implementation as the default, since the syscall
is supported by all architectures.
Also cleanup some headers and remove the INTERNAL_SYSCALL_ERROR_P
usage (the INTERNAL_SYSCALL_CALL macro already returns an negative
value if an error occurs).
Xing Li [Tue, 29 Nov 2022 11:24:43 +0000 (19:24 +0800)]
linux: Use long int for syscall return value
The linux syscall ABI returns long, so the generic syscall code for
linux should use long for the return value.
This fixes the truncation of the return value of the syscall function
when that does not fit into an int. Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
The assembler is not issued directly, but rather always through CC
wrapper. The binutils version check if done with LD instead. Reviewed-by: Carlos O'Donell <carlos@redhat.com>
When using --with-binutils, the configure might not use the specified
linker or assembler while checking for expected support. Move the
tools check early, before any compiler usage test. Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Sergey Bugaev [Fri, 2 Dec 2022 13:55:58 +0000 (16:55 +0300)]
hurd: Make getrandom cache the server port
Previously, getrandom would, each time it's called, traverse the file
system to find /dev/urandom, fetch some random data from it, then throw
away that port. This is quite slow, while calls to getrandom are
genrally expected to be fast.
Additionally, this means that getrandom can not work when /dev/urandom
is unavailable, such as inside a chroot that lacks one. User programs
expect calls to getrandom to work inside a chroot if they first call
getrandom outside of the chroot.
In particular, this is known to break the OpenSSH server, and in that
case the issue is exacerbated by the API of arc4random, which prevents
it from properly reporting errors, forcing glibc to abort on failure.
This causes sshd to just die once it tries to generate a random number.
Caching the random server port, in a manner similar to how socket
server ports are cached, both improves the performance and works around
the chroot issue.
Tested on i686-gnu with the following program:
pthread_barrier_t barrier;
void *worker(void*) {
pthread_barrier_wait(&barrier);
uint32_t sum = 0;
for (int i = 0; i < 10000; i++) {
sum += arc4random();
}
return (void *)(uintptr_t) sum;
}
for (int i = 0; i < THREAD_COUNT; i++) {
pthread_create(&threads[i], NULL, worker, NULL);
}
for (int i = 0; i < THREAD_COUNT; i++) {
void *retval;
pthread_join(threads[i], &retval);
printf("Thread %i: %lu\n", i, (unsigned long)(uintptr_t) retval);
}
In my totally unscientific benchmark, with this patch, this completes
in about 7 seconds, whereas previously it took about 50 seconds. This
program was also used to test that getrandom () doesn't explode if the
random server dies, but instead reopens the /dev/urandom anew. I have
also verified that with this patch, OpenSSH can once again accept
connections properly.
H.J. Lu [Fri, 2 Dec 2022 00:36:02 +0000 (16:36 -0800)]
x86-64 strncpy: Properly handle the length parameter [BZ# 29839]
On x32, the size_t parameter may be passed in the lower 32 bits of a
64-bit register with the non-zero upper 32 bits. The string/memory
functions written in assembly can only use the lower 32 bits of a
64-bit register as length or must clear the upper 32 bits before using
the full 64-bit register for length.
This pach fixes strncpy for x32. Tested on x86-64 and x32. On x86-64,
libc.so is the same with and without the fix. Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>
H.J. Lu [Fri, 2 Dec 2022 00:37:11 +0000 (16:37 -0800)]
x86-64 strncat: Properly handle the length parameter [BZ# 24097]
On x32, the size_t parameter may be passed in the lower 32 bits of a
64-bit register with the non-zero upper 32 bits. The string/memory
functions written in assembly can only use the lower 32 bits of a
64-bit register as length or must clear the upper 32 bits before using
the full 64-bit register for length.
This pach fixes strncat for x32. Tested on x86-64 and x32. On x86-64,
libc.so is the same with and without the fix. Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>
Shahab Vahedi [Sun, 27 Nov 2022 17:38:25 +0000 (18:38 +0100)]
ARC: update definitions in elf/elf.h
While porting ARCv2 to elfutils [1], it was brought up that the
necessary changes to the project's libelf/elf.h must come from
glibc, because they sync it from glibc [2]. Therefore, this patch
is to update ARC entries in elf/elf.h.
The majority of the update is about adding new definitions,
specially for the relocations. However, there is one rename, one
deletion, and one change:
- R_ARC_JUMP_SLOT renamed to R_ARC_JMP_SLOT to match binutils.
- R_ARC_B26 removed because it is unused and deprecated.
- R_ARC_TLS_DTPOFF_S9 changed from 0x4a to the correct value 0x49.
Finally, a specific SHT class for ARC has been added to glibcelf.py.
Else, it would result in a collision:
_register_elf_h(Sht, ranges=True,
File "/src/glibc/scripts/glibcelf.py", line x, in _register_elf_h
raise ValueError('duplicate value {}: {}, {}'.format(
ValueError: duplicate value 1879048193:
SHT_ARC_ATTRIBUTES, SHT_X86_64_UNWIND
Shahab Vahedi [Sun, 27 Nov 2022 17:38:24 +0000 (18:38 +0100)]
scripts: Add "|" operator support to glibcpp's parsing
From the tests point of view, this is a necessary step for another
patch [1] and allows parsing macros such as "#define A | B". Without
it, a few tests [2] choke when the other patch [1] is applied:
/src/glibc/scripts/../elf/elf.h:4167: error: uninterpretable macro
token sequence: ( EF_ARC_MACH_MSK | EF_ARC_OSABI_MSK )
Traceback (most recent call last):
File "/src/glibc/elf/tst-glibcelf.py", line 23, in <module>
import glibcelf
File "/src/glibc/scripts/glibcelf.py", line 226, in <module>
_elf_h = _parse_elf_h()
^^^^^^^^^^^^^^
File "/src/glibc/scripts/glibcelf.py", line 223, in _parse_elf_h
raise IOError('parse error in elf.h')
OSError: parse error in elf.h
[1] ARC: update definitions in elf/elf.h
https://sourceware.org/pipermail/libc-alpha/2022-November/143503.html
[2]
tst-glibcelf, tst-relro-ldso, and tst-relro-libc
Apply asm redirections in syslog.h before first use [BZ #27087]
Similar to d0fa09a770, but for syslog.h when _FORTIFY_SOURCE > 0.
Fixes [BZ #27087] by applying long double-related asm redirections
before using functions in bits/syslog.h.
Xiaolin Tang [Wed, 23 Nov 2022 03:49:24 +0000 (11:49 +0800)]
LoongArch: Add support for scalbn[f]
Add inline assembler for the scalbn functions. Passes GLIBC regression.
GCC 13, LoongArch support ___builtin_scalbn{,f} with -fno-math-errno,
but only "libm" can use -fno-math-errno in GLIBC, and scalbn is in libc
instead of libm because __printf_fp calls it.
Xiaolin Tang [Wed, 23 Nov 2022 03:45:00 +0000 (11:45 +0800)]
Use GCC builtins for logb functions if desired.
This patch is using the corresponding GCC builtin for logbf, logb,
logbl and logbf128 if the USE_FUNCTION_BUILTIN macros are defined to one
in math-use-builtins-function.h.
Xiaolin Tang [Wed, 23 Nov 2022 03:44:58 +0000 (11:44 +0800)]
Use GCC builtins for llrint functions if desired.
This patch is using the corresponding GCC builtin for llrintf, llrint,
llrintl and llrintf128 if the USE_FUNCTION_BUILTIN macros are defined to one
in math-use-builtins-function.h.
Xiaolin Tang [Wed, 23 Nov 2022 03:44:56 +0000 (11:44 +0800)]
Use GCC builtins for lrint functions if desired.
This patch is using the corresponding GCC builtin for lrintf, lrint,
lrintl and lrintf128 if the USE_FUNCTION_BUILTIN macros are defined to one
in math-use-builtins-function.h.
i386: Avoid rely on linker optimization to avoid relocation
lld does not implement all the linker optimization to avoid the GOT
relocation as done by binutils (bfd/elf32-i386.c:elf_i386_convert_load_reloc).
The current 'movl main@GOT(%ebx), %eax' will then create a GOT
relocation when building with lld, which make static-pie status to
not being able to start the provided main function.
The change uses a __wrap_main local symbol, which in turn calls main
(similar as used by aarch64 and s390x).
Checked on i686-linux-gnu with binutils and lld. Reviewed-by: Fangrui Song <maskray@google.com>
Florian Weimer [Tue, 8 Nov 2022 13:15:02 +0000 (14:15 +0100)]
Linux: Support __IPC_64 in sysvctl *ctl command arguments (bug 29771)
Old applications pass __IPC_64 as part of the command argument because
old glibc did not check for unknown commands, and passed through the
arguments directly to the kernel, without adding __IPC_64.
Applications need to continue doing that for old glibc compatibility,
so this commit enables this approach in current glibc.
For msgctl and shmctl, if no translation is required, make
direct system calls, as we did before the time64 changes. If
translation is required, mask __IPC_64 from the command argument.
For semctl, the union-in-vararg argument handling means that
translation is needed on all architectures.
Zong Li [Wed, 9 Nov 2022 14:40:59 +0000 (11:40 -0300)]
riscv: Get level 3 cache's information
RISC-V architecture extends the cache information for level 3 cache
in AUX vector in Linux v.6.1-rc1. This patch supports sysconf to get
the level 3 cache information.
наб [Wed, 9 Nov 2022 12:48:46 +0000 (13:48 +0100)]
iconvdata/tst-table-charmap.sh: remove handling of old, borrowed format
This "Old POSIX/DKUUG borrowed format" handling is original to the file
and doesn't seem to have ever been used, i.e. id/t-t-c doesn't seem to
have ever been called with argv[1] == POSIX.
Upcoming is a POSIX charmap, which would inadvertently trigger this.
Fangrui Song [Wed, 9 Nov 2022 04:56:59 +0000 (20:56 -0800)]
Makerules: Generate shlib.lds with -fuse-ld=bfd
lld does not dump a linker script with --verbose (it does not use a
linker script driven design and lots of linker processing is not
serializable as a linker script anyway). With the default
--with-default-link=no build, $@T is empty and makes `test -s $@T` fail.
Just dump the linker script with -fuse-ld=bfd. lld since 15
(https://reviews.llvm.org/D124656) supports custom RELRO sections in the
GNU ld dumped linker script.
Performance Changes:
Times are from N = 10 runs of the benchmark suite and are reported
as geometric mean of all ratios of New Implementation / Best Old
Implementation. Best Old Implementation was determined with the
highest ISA implementation.
Code Size Changes:
This change increase the size of libc.so by ~5.5kb bytes. For
reference the patch optimizing the normal strcpy family functions
decreases libc.so by ~5.2kb.
Full check passes on x86-64 and build succeeds for all ISA levels w/
and w/o multiarch.
Performance Changes:
Times are from N = 10 runs of the benchmark suite and are reported
as geometric mean of all ratios of New Implementation / Best Old
Implementation. Best Old Implementation was determined with the
highest ISA implementation.
Code Size Changes:
This change increase the size of libc.so by ~6.3kb bytes. For
reference the patch optimizing the normal strcpy family functions
decreases libc.so by ~5.7kb.
Full check passes on x86-64 and build succeeds for all ISA levels w/
and w/o multiarch.
Noah Goldstein [Wed, 9 Nov 2022 01:38:39 +0000 (17:38 -0800)]
x86: Optimize and shrink st{r|p}{n}{cat|cpy}-avx2 functions
Optimizations are:
1. Use more overlapping stores to avoid branches.
2. Reduce how unrolled the aligning copies are (this is more of a
code-size save, its a negative for some sizes in terms of
perf).
3. For st{r|p}n{cat|cpy} re-order the branches to minimize the
number that are taken.
Performance Changes:
Times are from N = 10 runs of the benchmark suite and are
reported as geometric mean of all ratios of
New Implementation / Old Implementation.
Noah Goldstein [Wed, 9 Nov 2022 01:38:38 +0000 (17:38 -0800)]
x86: Optimize and shrink st{r|p}{n}{cat|cpy}-evex functions
Optimizations are:
1. Use more overlapping stores to avoid branches.
2. Reduce how unrolled the aligning copies are (this is more of a
code-size save, its a negative for some sizes in terms of
perf).
3. Improve the loop a bit (similiar to what we do in strlen with
2x vpminu + kortest instead of 3x vpminu + kmov + test).
4. For st{r|p}n{cat|cpy} re-order the branches to minimize the
number that are taken.
Performance Changes:
Times are from N = 10 runs of the benchmark suite and are
reported as geometric mean of all ratios of
New Implementation / Old Implementation.
I couldn't find a way to merge them without making the
ifdefs incredibly difficult to follow.
2. All implementations can be made evex512 by including
"x86-evex512-vecs.h" at the top.
3. All implementations have an optional define:
`USE_EVEX_MASKED_STORE`
Setting to one uses evex-masked stores for handling short
strings. This saves code size and branches. It's disabled
for all implementations are the moment as there are some
serious drawbacks to masked stores in certain cases, but
that may be fixed on future architectures.
Full check passes on x86-64 and build succeeds for all ISA levels w/
and w/o multiarch.
Noah Goldstein [Sat, 29 Oct 2022 20:19:59 +0000 (15:19 -0500)]
x86: Use VMM API in memcmpeq-evex.S and minor changes
Changes to generated code are:
1. In a few places use `vpcmpeqb` instead of `vpcmpneq` to save a
byte of code size.
2. Add a branch for length <= (VEC_SIZE * 6) as opposed to doing
the entire block of [VEC_SIZE * 4 + 1, VEC_SIZE * 8] in a
single basic-block (the space to add the extra branch without
changing code size is bought with the above change).
Change (2) has roughly a 20-25% speedup for sizes in [VEC_SIZE * 4 +
1, VEC_SIZE * 6] and negligible to no-cost for [VEC_SIZE * 6 + 1,
VEC_SIZE * 8]
Linux: Add ppoll fortify symbol for 64 bit time_t (BZ# 29746)
Similar to ppoll, the poll.h header needs to redirect the poll call
to a proper fortified ppoll with 64 bit time_t support.
The implementation is straightforward, just need to add a similar
check as __poll_chk and call the 64 bit time_t ppoll version. The
debug fortify tests are also extended to cover 64 bit time_t for
affected ABIs.
Unfortunately it requires an aditional symbol, which makes backport
tricky. One possibility is to add a static inline version if compiler
supports is and call abort instead of __chk_fail, so fortified version
will call __poll64 in the end.
Another possibility is to just remove the fortify support for
_TIME_BITS=64.
For clang the redeclaration after the first use, the visibility attribute
is silently ignored (symbol is STV_DEFAULT) while the asm label attribute
causes an error.
With clang if the main file is <stdin>, the first non-main-file
dependency is not listed in the -MP output. Although it was fixed
on clang-16 [1], this change adds portability for older version.
Florian Weimer [Fri, 4 Nov 2022 06:43:59 +0000 (07:43 +0100)]
posix: Make posix_spawn extensions available by default
Some sources merely include <spawn.h> without -D_GNU_SOURCE and expect
declarations for posix_spawn_file_actions_addchdir_np to be available.
For consistency, declare posix_spawn_file_actions_addfchdir_np,
posix_spawn_file_actions_addclosefrom_np,
posix_spawn_file_actions_addtcsetpgrp_np as well.
Sunil K Pandey [Tue, 9 Aug 2022 14:57:29 +0000 (07:57 -0700)]
x86_64: Implement evex512 version of strrchr and wcsrchr
Changes from v1:
Use vec api for register.
Replace VPCMP with VPCMPEQ
Restructure and remove 1 unconditional jump.
Change page cross logic to use sall.
This patch implements following evex512 version of string functions.
evex512 version takes up to 30% less cycle as compared to evex,
depending on length and alignment.
- strrchr function using 512 bit vectors.
- wcsrchr function using 512 bit vectors.
Florian Weimer [Thu, 3 Nov 2022 16:28:03 +0000 (17:28 +0100)]
elf: Introduce <dl-call_tls_init_tp.h> and call_tls_init_tp (bug 29249)
This makes it more likely that the compiler can compute the strlen
argument in _startup_fatal at compile time, which is required to
avoid a dependency on strlen this early during process startup.
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
Without this change, parse failures result in an exception:
Traceback (most recent call last):
File "tst-glibcelf.py", line 23, in <module>
import glibcelf
File "/path/to/git/scripts/glibcelf.py", line 226, in <module>
_elf_h = _parse_elf_h()
File "/path/to/git/scripts/glibcelf.py", line 221, in _parse_elf_h
result = glibcpp.macro_eval(glibcpp.macro_definitions(tokens), reporter)
File "/path/to/git/scripts/glibcpp.py", line 379, in macro_eval
reporter.error(md.line, 'macro {} redefined'.format(md.name))
File "/path/to/git/scripts/glibcelf.py", line 214, in error
errors += 1
UnboundLocalError: local variable 'errors' referenced before assignment
Florian Weimer [Thu, 3 Nov 2022 08:39:31 +0000 (09:39 +0100)]
elf: Rework exception handling in the dynamic loader [BZ #25486]
The old exception handling implementation used function interposition
to replace the dynamic loader implementation (no TLS support) with the
libc implementation (TLS support). This results in problems if the
link order between the dynamic loader and libc is reversed (bug 25486).
The new implementation moves the entire implementation of the
exception handling functions back into the dynamic loader, using
THREAD_GETMEM and THREAD_SETMEM for thread-local data support.
These depends on Hurd support for these macros, added in commit b65a82e4e757c1e6cb7073916 ("hurd: Add THREAD_GET/SETMEM/_NC").
One small obstacle is that the exception handling facilities are used
before the TCB has been set up, so a check is needed if the TCB is
available. If not, a regular global variable is used to store the
exception handling information.
Also rename dl-error.c to dl-catch.c, to avoid confusion with the
dlerror function.
git://sourceware.org / glibc.git / log