shm-directory: Truncated struct member name length
The struct shmdir_name in include/shm-directory.h has name field to
contains the full path of the POSIX IPC object (shm and sem).
The size was previously set to sizeof (SHMDIR) + 4 + NAME_MAX, where 4
bytes were reserved for the optional "sem." prefix.
This led to incorrect execution of the __shm_get_name function
in posix/shm-directory.c which is used accross in shm_[open/unlink] and
sem_[open/unlink] functions.
For shm_[open/unlink]:
This is because the name field was large enough to hold 268 characters
(255 + 4 + 9) instead of the maximum allowed 263 characters (255 + 9).
This caused the __shm_get_name to not throw ENAMETOOLONG error when the
name length exceeded NAME_MAX (255) upto 259 characters.
For sem_[open/unlink]:
Similarly, the __shm_get_name incorrectly returned success for names of
length 255 instead of 251 (255 - 4).
This was overlooked as finally these functions throw the correct
ENAMETOOLONG error; which was thrown by the openat syscall, which is
called later in the shm_* and sem_* functions.
This patch corrects the size of name field in struct shmdir_name to
sizeof (SHMDIR) + NAME_MAX. The __shm_get_name function return
ENAMETOOLONG if alloc_buffer_has_failed returns true (which only happens
when copy length > alloc_buffer_size (buffer)).
Relevant runtime monitoring were done in gdb to confirm the same.
Joseph Myers [Wed, 1 Oct 2025 15:15:15 +0000 (15:15 +0000)]
Add once_flag, ONCE_FLAG_INIT and call_once to stdlib.h for C23
C23 adds once_flag, ONCE_FLAG_INIT and call_once to stdlib.h (in C11
they were only in threads.h, in C23 they are in both headers; this
change came from N2840). Implement this change, with a
bits/types/once_flag.h header for the common type and initializer
definitions.
Note that there's an omnibus bug (bug 33001) that covers more than
just these missing definitions.
This doesn't seem a significant enough feature to be worth mentioning
in NEWS.
ISO C is not concerned with whether functions are in libc or
libpthread, but POSIX links this to what header they are declared in,
so functions declared in stdlib.h are supposed to be in libc.
However, the current edition of POSIX is based on C17; hopefully Hurd
glibc will have completed the merge of libpthread into libc (in
particular, moving call_once) well before a future edition of POSIX
based on C23 (or a later version of ISO C) is released.
Joseph Myers [Wed, 1 Oct 2025 15:14:09 +0000 (15:14 +0000)]
Implement C23 memset_explicit (bug 32378)
Add the C23 memset_explicit function to glibc. Everything here is
closely based on the approach taken for explicit_bzero. This includes
the bits that relate to internal uses of explicit_bzero within glibc
(although we don't currently have any such internal uses of
memset_explicit), and also includes the nonnull attribute (when we
move to nonnull_if_nonzero for various functions following C2y, this
function should be included in that change).
The function is declared both for __USE_MISC and for __GLIBC_USE (ISOC23)
(so by default not just for compilers defaulting to C23 mode).
manual: Fix missing declaration in inetcli example.
Previously this file failed to compile with the following errors:
$ gcc manual/examples/inetcli.c
manual/examples/inetcli.c: In function ‘write_to_server’:
manual/examples/inetcli.c:36:37: error: implicit declaration of function ‘strlen’ [-Wimplicit-function-declaration]
36 | nbytes = write (filedes, MESSAGE, strlen (MESSAGE) + 1);
| ^~~~~~
manual/examples/inetcli.c:26:1: note: include ‘<string.h>’ or provide a declaration of ‘strlen’
25 | #include <netdb.h>
+++ |+#include <string.h>
26 |
manual/examples/inetcli.c:36:37: warning: incompatible implicit declaration of built-in function ‘strlen’ [-Wbuiltin-declaration-mismatch]
36 | nbytes = write (filedes, MESSAGE, strlen (MESSAGE) + 1);
| ^~~~~~
manual/examples/inetcli.c:36:37: note: include ‘<string.h>’ or provide a declaration of ‘strlen’
Previously this file failed to compile with the following errors:
$ gcc manual/examples/inetsrv.c
manual/examples/inetsrv.c: In function ‘main’:
manual/examples/inetsrv.c:97:31: error: passing argument 3 of ‘accept’ from incompatible pointer type [-Wincompatible-pointer-types]
97 | &size);
| ^~~~~
| |
| size_t * {aka long unsigned int *}
In file included from manual/examples/inetsrv.c:23:
/usr/include/sys/socket.h:307:42: note: expected ‘socklen_t * restrict’ {aka ‘unsigned int * restrict’} but argument is of type ‘size_t *’ {aka ‘long unsigned int *’}
307 | socklen_t *__restrict __addr_len);
| ~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~
manual/examples/inetsrv.c:105:26: error: implicit declaration of function ‘inet_ntoa’ [-Wimplicit-function-declaration]
105 | inet_ntoa (clientname.sin_addr),
Previously this file failed to compile with the following errors:
$ gcc manual/examples/filesrv.c
manual/examples/filesrv.c: In function ‘main’:
manual/examples/filesrv.c:37:3: error: implicit declaration of function ‘unlink’ [-Wimplicit-function-declaration]
37 | unlink (SERVER);
| ^~~~~~
manual/examples/filesrv.c:40:10: error: implicit declaration of function ‘make_named_socket’ [-Wimplicit-function-declaration]
40 | sock = make_named_socket (SERVER);
| ^~~~~~~~~~~~~~~~~
manual/examples/filesrv.c:46:54: error: passing argument 6 of ‘recvfrom’ from incompatible pointer type [-Wincompatible-pointer-types]
46 | (struct sockaddr *) & name, &size);
| ^~~~~
| |
| size_t * {aka long unsigned int *}
In file included from manual/examples/filesrv.c:21:
/usr/include/sys/socket.h:165:48: note: expected ‘socklen_t * restrict’ {aka ‘unsigned int * restrict’} but argument is of type ‘size_t *’ {aka ‘long unsigned int *’}
165 | socklen_t *__restrict __addr_len);
| ~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~
This patch fixes the missing declaration for unlink and uses
'socklen_t *' for the fourth argument of recv from. The
make_named_socket function is defined in the manual.
manual: Fix missing declaration in setjmp example.
Previously this file would fail to compile with the following error:
$ gcc manual/examples/setjmp.c
manual/examples/setjmp.c: In function ‘main’:
manual/examples/setjmp.c:37:7: error: implicit declaration of function ‘do_command’ [-Wimplicit-function-declaration]
37 | do_command ();
| ^~~~~~~~~~
manual/examples/setjmp.c: At top level:
manual/examples/setjmp.c:42:1: warning: conflicting types for ‘do_command’; have ‘void(void)’
42 | do_command (void)
| ^~~~~~~~~~
manual/examples/setjmp.c:37:7: note: previous implicit declaration of ‘do_command’ with type ‘void(void)’
37 | do_command ();
| ^~~~~~~~~~
manual: Fix missing declaration in strdupa example.
Without _GNU_SOURCE defined this file fails to compile with the
following error:
$ gcc manual/examples/strdupa.c
manual/examples/strdupa.c: In function ‘main’:
manual/examples/strdupa.c:27:19: error: implicit declaration of function ‘strdupa’; did you mean ‘strdup’? [-Wimplicit-function-declaration]
27 | char *wr_path = strdupa (path);
| ^~~~~~~
| strdup
manual/examples/strdupa.c:27:19: error: initialization of ‘char *’ from ‘int’ makes pointer from integer without a cast [-Wint-conversion]
Previously this file would fail to compile with the following error:
$ gcc manual/examples/memopen.c
manual/examples/memopen.c: In function ‘main’:
manual/examples/memopen.c:28:30: error: implicit declaration of function ‘strlen’ [-Wimplicit-function-declaration]
28 | stream = fmemopen (buffer, strlen (buffer), "r");
| ^~~~~~
manual/examples/memopen.c:19:1: note: include ‘<string.h>’ or provide a declaration of ‘strlen’
18 | #include <stdio.h>
+++ |+#include <string.h>
19 |
manual/examples/memopen.c:28:30: warning: incompatible implicit declaration of built-in function ‘strlen’ [-Wbuiltin-declaration-mismatch]
28 | stream = fmemopen (buffer, strlen (buffer), "r");
| ^~~~~~
manual/examples/memopen.c:28:30: note: include ‘<string.h>’ or provide a declaration of ‘strlen’
Without _GNU_SOURCE defined this file fails to compile with the
following error:
$ gcc manual/examples/twalk.c
manual/examples/twalk.c: In function ‘twalk’:
manual/examples/twalk.c:55:3: error: implicit declaration of function ‘twalk_r’; did you mean ‘twalk’? [-Wimplicit-function-declaration]
55 | twalk_r (root, twalk_with_twalk_r_action, &closure);
| ^~~~~~~
| twalk
Previously this file would fail to compile with the following error:
$ gcc manual/examples/sigusr.c
manual/examples/sigusr.c: In function ‘child_function’:
manual/examples/sigusr.c:46:3: error: implicit declaration of function ‘exit’ [-Wimplicit-function-declaration]
46 | exit (0);
| ^~~~
manual/examples/sigusr.c:23:1: note: include ‘<stdlib.h>’ or provide a declaration of ‘exit’
22 | #include <unistd.h>
+++ |+#include <stdlib.h>
23 | /*@end group*/
manual/examples/sigusr.c:46:3: warning: incompatible implicit declaration of built-in function ‘exit’ [-Wbuiltin-declaration-mismatch]
46 | exit (0);
| ^~~~
manual/examples/sigusr.c:46:3: note: include ‘<stdlib.h>’ or provide a declaration of ‘exit’
manual: Fix missing includes in the mbstouwcs example.
Previously this file would fail to compile with the following error:
$ gcc manual/examples/mbstouwcs.c
manual/examples/mbstouwcs.c: In function ‘mbstouwcs’:
manual/examples/mbstouwcs.c:34:11: error: ‘errno’ undeclared (first use in this function)
34 | errno = EILSEQ;
| ^~~~~
manual/examples/mbstouwcs.c:5:1: note: ‘errno’ is defined in header ‘<errno.h>’; this is probably fixable by adding ‘#include <errno.h>’
4 | #include <wchar.h>
+++ |+#include <errno.h>
5 |
manual/examples/mbstouwcs.c:34:11: note: each undeclared identifier is reported only once for each function it appears in
34 | errno = EILSEQ;
| ^~~~~
manual/examples/mbstouwcs.c:34:19: error: ‘EILSEQ’ undeclared (first use in this function)
34 | errno = EILSEQ;
| ^~~~~~
manual/examples/mbstouwcs.c:47:20: error: implicit declaration of function ‘towupper’ [-Wimplicit-function-declaration]
47 | *wcp++ = towupper (wc);
| ^~~~~~~~
manual/examples/mbstouwcs.c:5:1: note: include ‘<wctype.h>’ or provide a declaration of ‘towupper’
4 | #include <wchar.h>
+++ |+#include <wctype.h>
5 |
manual/examples/mbstouwcs.c:47:20: warning: incompatible implicit declaration of built-in function ‘towupper’ [-Wbuiltin-declaration-mismatch]
47 | *wcp++ = towupper (wc);
| ^~~~~~~~
manual/examples/mbstouwcs.c:47:20: note: include ‘<wctype.h>’ or provide a declaration of ‘towupper’
manual: Fix missing include in group and user database example.
Previously this file would fail to compile with the following error:
$ gcc manual/examples/db.c
db.c: In function ‘main’:
db.c:37:7: error: implicit declaration of function ‘printf’ [-Wimplicit-function-declaration]
37 | printf ("Couldn't find out about user %d.\n", (int) me);
| ^~~~~~
db.c:23:1: note: include ‘<stdio.h>’ or provide a declaration of ‘printf’
22 | #include <stdlib.h>
+++ |+#include <stdio.h>
23 |
db.c:37:7: warning: incompatible implicit declaration of built-in function ‘printf’ [-Wbuiltin-declaration-mismatch]
37 | printf ("Couldn't find out about user %d.\n", (int) me);
| ^~~~~~
db.c:37:7: note: include ‘<stdio.h>’ or provide a declaration of ‘printf’
db.c:42:3: warning: incompatible implicit declaration of built-in function ‘printf’ [-Wbuiltin-declaration-mismatch]
42 | printf ("I am %s.\n", my_passwd->pw_gecos);
| ^~~~~~
db.c:42:3: note: include ‘<stdio.h>’ or provide a declaration of ‘printf’
Linux: Fix tst-copy_file_range-large test on recent kernels [BZ #33498]
Instead of a negative return value the fixed FUSE copy_file_range will
silently truncate the size to UINT_MAX & PAGE_MASK [1]. Allow that value
to be returned as well.
fpu_control.h is an installed header so a wider range of compiler versions
(including ones older than GCC 9) are relevant with it than are relevant
for building glibc.
Jovan Dmitrovic [Wed, 3 Sep 2025 13:53:37 +0000 (13:53 +0000)]
mips: Remove strcmp.S
Testing strcmp on MIPS hardware shows that strcmp.S performs worse
than the combination of using the generic strcmp.c implementation
alongside -funroll-loops.
Suggested-by: Joseph Myers <josmyers@redhat.com> Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
It now calls __libc_assert, which contains similar logic. The assert
call only requires memory allocation for the message translation, so
test-assert2.c is adapted to handle it.
It also removes the fxprintf from assert/assert_perror; although it
is not 100% backwards-compatible (write message only if there is a
file descriptor associated with the stderr). It now writes bytes
directly without going through the wide stream state.
nptl: Fix MADV_GUARD_INSTALL logic for thread without guard page (BZ 33356)
The main issue is that setup_stack_prot fails to account for cases where
the cached thread stack lacks a guard page, which can cause madvise to
fail. Update the logic to also handle whether MADV_GUARD_INSTALL is
supported when resizing the guard page.
Checked on x86_64-linux-gnu with 6.8.0 and 6.15 kernels.
x86: Use "%v" to emit VEX encoded instructions for AVX targets
Legacy encodings of SSE instructions incur AVX-SSE domain transition
penalties on some Intel microarchitectures (e.g. Haswell, Broadwell).
Using the VEX forms avoids these penatlies and keeps all instructions
in the VEX decode domain. Use "%v" sequence to emit the "v" prefix
for opcodes when compiling with -mavx.
GCC now accept plain variable names as valid lvalues for "m"
constraints, automatically spilling locals to memory if necessary.
The long-standing "*&" pattern was originally used as a defensive
workaround for older compiler versions that rejected operands
such as:
asm ("incl %0" : "+m"(x));
with errors like "memory input is not directly addressable".
Modern compilers (GCC >= 9) reliably generate correct code
without the workaround, and the resulting assembly is identical.
Diego Nieto Cid [Fri, 15 Aug 2025 01:57:30 +0000 (02:57 +0100)]
hurd: implement RLIMIT_AS against Mach RPCs
Check for VM limit RPCs
* config.h.in: add #undef for HAVE_MACH_VM_GET_SIZE_LIMIT and
HAVE_MACH_VM_SET_SIZE_LIMIT.
* sysdeps/mach/configure.ac: use mach_RPC_CHECK to check for
vm_set_size_limit and vm_get_size_limit RPCs in gnumach.defs.
* sysdeps/mach/configure: regenerate file.
Use vm_get_size_limit to initialize RLIMIT_AS
* hurd/hurdrlimit.c(init_rlimit): use vm_get_size_limit to initialize
RLIMIT_AS entry of the _hurd_rlimits array.
Notify the kernel of the new VM size limits
* sysdeps/mach/hurd/setrlimit.c: use the vm_set_size_limit RPC,
if available, to notify the kernel of the new limits. Retry RPC
calls if they were interrupted by a signal.
Message-ID: <03fb90a795b354a366ee73f56f73e6ad22a86cda.1755220108.git.dnietoc@gmail.com>
Samuel Thibault [Sun, 21 Sep 2025 21:45:40 +0000 (23:45 +0200)]
hurd: catch SIGSEGV on returning from signal handler
On stack overflow typically, we may not actually have room on the stack to
trampoline back from the signal handler. We have to detect this before
locking the ss, otherwise the signal thread will be stuck on taking the
ss lock while trying to post SIGSEGV.
Remove support for obsolete dumped heaps. Dumping heaps was discontinued
8 years ago, however loading a dumped heap is still supported. This blocks
changes and improvements of the malloc data structures - hence it is time
to remove this. Ancient binaries that still call malloc_set_state will now
get the -1 error code. Update tst-mallocstate.c to just check for this.
The realpath call may trigger OOM termination of the test process
under difficult-to-predict circumstances. (It depends on available
RAM and swap.) Therefore, instruct the test driver to ignore
an OOM process termination during the realpath call.
support: Add support_accept_oom to heuristically support OOM errors
Some tests may trigger the kernel OOM handler under conditions
which are difficult to predict (depending on available RAM and
swap space). If we can determine specific regions which might
do this and this does not contradict the test object, the
functions support_accept_oom (true) and support_accept_oom (false)
can be called at the start and end, and the test driver will
ignore SIGKILL signals.
The "unused" variable could be use unitialized, which is an issue if ldd
is ran with "-u". Fix that by defining the variable to an empty value,
just like it is already done for the bind_now, warn and verbose
variables.
Reported-by: Johan Palmqvist <johan.palmqvist@gmail.com> Reviewed-by: Sam James <sam@gentoo.org>
which warns ‘-pg’ without ‘-mfentry’, when glibc is configured with
--disable-default-pie, GCC 16 fails to compile .op files and gmon tests
with error:
cc1: error: ‘-pg’ without ‘-mfentry’ may be unreliable with shrink wrapping [-Werror]
Compile .op files and gmon tests with -mfentry if it is supported by
CC/TEST_CC and glibc is configured with --disable-default-pie. This
fixes BZ #33376.
Signed-off-by: H.J. Lu <hjl.tools@gmail.com> Reviewed-by: Joseph Myers <josmyers@redhat.com>
i386: Use __seg_gs qualifier to cast access to TCB in THREAD_GSCOPE_RESET_FLAG()
Use the __seg_gs named address space qualifier to cast access to the
gscope_flag in the TCB as a %gs: prefixed address. This enables the
use of the "m" operand constraint, which informs the compiler about
memory access in the inline assembly.
x86_64: Use __seg_fs qualifier to cast access to TCB in THREAD_GSCOPE_RESET_FLAG()
Use the __seg_fs named address space qualifier to cast access to the
gscope_flag in the TCB as a %fs: prefixed address. This enables the
use of the "m" operand constraint, which informs the compiler about
memory access in the inline assembly.
Joseph Myers [Wed, 3 Sep 2025 16:30:11 +0000 (16:30 +0000)]
Fix RISC-V soft-float _FPU_SETCW for GCC 16 set-but-not-used warnings
The soft-float RISC-V definition of _FPU_SETCW results in a
-Werror=unused-but-set-variable= build failure with GCC mainline (in
math/setfpucw.c) because it does not use the dummy cw variable.
Change it to (void) (cw) as on other architectures to avoid this build
failure.
Tested with build-many-glibcs.py (compilers) for
riscv64-linux-gnu-rv64imac-lp64, which previously failed. Reviewed-by: Sam James <sam@gentoo.org>
nss: Group merge does not react to ERANGE during merge (bug 33361)
The break statement in CHECK_MERGE is expected to exit the surrounding
while loop, not the do-while loop with in the macro. Remove the
do-while loop from the macro. It is not needed to turn the macro
expansion into a single statement due to the way CHECK_MERGE is used
(and the statement expression would cover this anyway).
caiyinyu [Fri, 3 Jan 2025 07:23:46 +0000 (15:23 +0800)]
LoongArch: Use the generic shmlab.h.
The shmlba.h file related to the LoongArch architecture was introduced
in commit 3eed5f3a1ee to address the mismatch in the SHMLBA definition
between glibc and the kernel. See [1]. The SHMLBA definition was later
updated in commit d23b77953f5a. See [2]. Now, we adopt the definition
from the common layer.
LoongArch: Fix inconsistency in SHMLBA macro values between glibc and kernel
The LoongArch glibc was using the value of the SHMLBA macro from common code,
which is __getpagesize() (16k), but this was inconsistent with the value of
the SHMLBA macro in the kernel, which is SZ_64K (64k). This caused several
shmat-related tests in LTP (Linux Test Project) to fail. This commit fixes
the issue by ensuring that the glibc's SHMLBA macro value matches the value
used in the kernel like other architectures.
LoongArch has hardware page coloring for L1 Cache, so we don't have
cache aliases. But SFB (Store Fill Buffer) still has aliases. So we
define SHMLBA to SZ_64K previously. But there are losts of applications
use PAGE_SIZE rather than SHMLBA to mmap() file pages and shared pages.
Of course we can fix them one by one, but not easy.
On the other hand, we can simply disable SFB for 4KB page size to fix
cache alias (there will be performance decrease, but acceptable), and
in future we will fix SFB in hardware. So we can safely define SHMLBA to
PAGE_SIZE (use the generic shmparam.h) to make life easier.
alpha: Fix missing inexact-flag raising for lround/lrint
The l*[rint|round]f implements uses alpha 'cvtst/s', 'addt/suc',
adn 'cvttq/svd' which are not not fully IEEE compliant w.r.t
inexact-flag raising.. Use the software fallback implementation
instead.
Checked on alpha-linux-gnu.
Tested-by: Michael Cree <mcree@orcon.net.nz> Reviewed-by: Carlos O'Donell <carlos@redhat.com>
To avoid linknamespace issues on old standards. It is required
if the fallback fma implementation is used if/when it is also
used internally for other implementation. Reviewed-by: DJ Delorie <dj@redhat.com>
To avoid linknamespace issues on old standards. It is required
if the fallback fma implementation is used if/when it is also
used internally for other implementation. Reviewed-by: DJ Delorie <dj@redhat.com>
To avoid linknamespace issues on old standards. It is required
if the fallback fma implementation is used if/when it is also
used internally for other implementation. Reviewed-by: DJ Delorie <dj@redhat.com>
To avoid linknamespace issues on old standards. It is required
if the fallback fma implementation is used if/when it is also
used internally for other implementation. Reviewed-by: DJ Delorie <dj@redhat.com>
To avoid linknamespace issues on old standards. It is required
if the fallback fma implementation is used if/when it is also
used internally for other implementation. Reviewed-by: DJ Delorie <dj@redhat.com>
To avoid linknamespace issues on old standards. It is required
if the fallback fma implementation is used if/when it is also
used internally for other implementation. Reviewed-by: DJ Delorie <dj@redhat.com>
To avoid linknamespace issues on old standards. It is required
if the fallback fma implementation is used if/when it is also
used internally for other implementation. Reviewed-by: DJ Delorie <dj@redhat.com>
Mike FABIAN [Mon, 18 Aug 2025 07:24:06 +0000 (09:24 +0200)]
Update to Unicode 17.0.0 [BZ #33289]
Unicode 17.0.0 Support: Character encoding, character type info, and
transliteration tables are all updated to Unicode 17.0.0, using
the generator scripts contributed by Mike FABIAN (Red Hat).
Changes in CHARMAP and WIDTH:
Total added characters in newly generated CHARMAP: 4803
Total removed characters in newly generated WIDTH: 0
Total changed characters in newly generated WIDTH: 0
Total added characters in newly generated WIDTH: 4512
Some combining characters and other non-spacing marks have been added
with WIDTH 0. Lots of characters have been added with WIDTH 2, most of
them are CJK Ideographs plus a few Tangut characters and 7 emoji.
Changes in ctype:
alpha: Added 4672 characters in new ctype which were not in old ctype
combining: Added 42 characters in new ctype which were not in old ctype
combining_level3: Added 8 characters in new ctype which were not in old ctype
graph: Added 4803 characters in new ctype which were not in old ctype
lower: Missing: ʕ 0x295 LATIN LETTER PHARYNGEAL VOICED FRICATIVE
lower: Added 27 characters in new ctype which were not in old ctype
print: Added 4803 characters in new ctype which were not in old ctype
punct: Added 131 characters in new ctype which were not in old ctype
tolower: Added 28 characters in new ctype which were not in old ctype
totitle: Added 28 characters in new ctype which were not in old ctype
toupper: Added 28 characters in new ctype which were not in old ctype
upper: Added 28 characters in new ctype which were not in old ctype
Nothing suspicious in the additions.
About the character removed from lower:
ʕ 0x295 LATIN LETTER PHARYNGEAL VOICED FRICATIVE
In UnicodeData.txt it changed from 'Ll' (Letter Lowercase) to 'Lo' (Letter Other):
-0295;LATIN LETTER PHARYNGEAL VOICED FRICATIVE;Ll;0;L;;;;;N;LATIN LETTER REVERSED GLOTTAL STOP;;;;
+0295;LATIN LETTER PHARYNGEAL VOICED FRICATIVE;Lo;0;L;;;;;N;LATIN LETTER REVERSED GLOTTAL STOP;;;;
Requires Neon (aka. Advanced SIMD). Looks up 16 characters at a time,
for a 2-3x perfomance improvement, and a ~30% speedup on the strtok &
strsep benchtests, as tested on Cortex A-{53,72}.
Remove all unused atomics. Replace uses of catomic_increment and
catomic_decrement with atomic_fetch_add_relaxed which maps to a standard
compiler builtin. Relaxed memory ordering is correct for simple counters
since they only need atomicity.
Remove obsolete catomic_* locking primitives which don't map
to standard compiler builtins.
There are still a couple of places in the tree that uses them
(malloc/arena.c and malloc/malloc.c).
x86 didn't define __arch_c_compare_and_exchange_bool_* primitives
so fallback code used __arch_c_compare_and_exchange_val_* primitives
instead. This resulted in unoptimal code for
catomic_compare_and_exchange_bool_acq where superfluous
CMP was emitted after CMPXCHG, e.g. in arena_get2:
OTOH, catomic_decrement does not fallback to atomic_fetch_add (, -1)
builtin but to the cmpxchg loop, so the generated code in arena_get2
regresses a bit, from using LOCK DECQ insn:
Depending on the target processor, the compiler may emit either
'LOCK ADD/SUB $1, m' or 'INC/DEC $1, m' instruction, due to partial
flag register stall issue.
atomic: Use builtin atomics with USE_ATOMIC_COMPILER_BUILTINS
Use builtin atomics for atomic_compare_and_exchange_* and
atomic_exchange_and_add if USE_ATOMIC_COMPILER_BUILTINS is enabled.
This allows removing target atomic-machine.h headers.
H.J. Lu [Tue, 9 Sep 2025 02:49:24 +0000 (19:49 -0700)]
x86: Include <bits/stdlib-bsearch.h> in dl-cacheinfo.h
On x86-64, when glibc is configured with --enable-stack-protector=all
and compiled with -Os, ld.so crashes very early:
(gdb) r --direct
Starting program: /export/build/gnu/tools-build/glibc-gitlab/build-x86_64-linux/string/test-memswap --direct
Program received signal SIGSEGV, Segmentation fault.
0x00007ffff7f41b0a in bsearch (__key=__key@entry=0x7fffffffda28,
__base=__base@entry=0x7ffff7fca140 <intel_02_known>,
__nmemb=__nmemb@entry=68, __size=__size@entry=8,
__compar=__compar@entry=0x7ffff7f3b691 <intel_02_known_compare>)
at ../bits/stdlib-bsearch.h:22
22 {
(gdb) disass
Dump of assembler code for function bsearch:
0x00007ffff7f41af0 <+0>: push %r15
0x00007ffff7f41af2 <+2>: mov %rcx,%r15
0x00007ffff7f41af5 <+5>: push %r14
0x00007ffff7f41af7 <+7>: push %r13
0x00007ffff7f41af9 <+9>: mov %rsi,%r13
0x00007ffff7f41afc <+12>: push %r12
0x00007ffff7f41afe <+14>: mov %rdi,%r12
0x00007ffff7f41b01 <+17>: push %rbp
0x00007ffff7f41b02 <+18>: mov %rdx,%rbp
0x00007ffff7f41b05 <+21>: push %rbx
0x00007ffff7f41b06 <+22>: sub $0x18,%rsp
=> 0x00007ffff7f41b0a <+26>: mov %fs:0x28,%r14
^^^^^^^^^^^^^^^^^^^^^^^^^^^^ We can't use stack protector at this point.
0x00007ffff7f41b13 <+35>: mov %r14,0x8(%rsp)
0x00007ffff7f41b18 <+40>: mov %r8,%r14
0x00007ffff7f41b1b <+43>: test %rbp,%rbp
0x00007ffff7f41b1e <+46>: je 0x7ffff7f41b48 <bsearch+88>
0x00007ffff7f41b20 <+48>: mov %rbp,%rbx
0x00007ffff7f41b23 <+51>: mov %r12,%rdi
0x00007ffff7f41b26 <+54>: shr $1,%rbx
0x00007ffff7f41b29 <+57>: imul %r15,%rbx
0x00007ffff7f41b2d <+61>: add %r13,%rbx
0x00007ffff7f41b30 <+64>: mov %rbx,%rsi
(gdb) bt
#0 0x00007ffff7f41b0a in bsearch (__key=__key@entry=0x7fffffffda28,
__base=__base@entry=0x7ffff7fca140 <intel_02_known>,
__nmemb=__nmemb@entry=68, __size=__size@entry=8,
__compar=__compar@entry=0x7ffff7f3b691 <intel_02_known_compare>)
at ../bits/stdlib-bsearch.h:22
#1 0x00007ffff7f3c1be in intel_check_word (name=188, value=1979933440,
has_level_2=has_level_2@entry=0x7fffffffda7f,
no_level_2_or_3=no_level_2_or_3@entry=0x7fffffffda7e,
cpu_features=<optimized out>) at ../sysdeps/x86/dl-cacheinfo.h:217
#2 0x00007ffff7f3c29f in handle_intel (name=name@entry=188,
cpu_features=<optimized out>) at ../sysdeps/x86/dl-cacheinfo.h:279
#3 0x00007ffff7f3ccf9 in dl_init_cacheinfo (cpu_features=<optimized out>)
at ../sysdeps/x86/dl-cacheinfo.h:852
#4 init_cpu_features (cpu_features=<optimized out>)
at ../sysdeps/x86/cpu-features.c:1153
#5 0x00007ffff7f3d6f9 in __libc_start_main_impl (main=0x7ffff7f396dc <main>,
argc=2, argv=0x7fffffffdbe8, init=<optimized out>, fini=<optimized out>,
rtld_fini=0x0, stack_end=0x7fffffffdbd8) at ../csu/libc-start.c:269
#6 0x00007ffff7f39901 in _start () at ../sysdeps/x86_64/start.S:115
(gdb)
The problem is that since __USE_EXTERN_INLINES isn't defined with -Os,
the inline bsearch in <bits/stdlib-bsearch.h> isn't available and the
external bsearch is compiled with stack protector. Include
<bits/stdlib-bsearch.h> in dl-cacheinfo.h fixed BZ #33374.
Signed-off-by: H.J. Lu <hjl.tools@gmail.com> Reviewed-by: Sam James <sam@gentoo.org>
projects / thirdparty / glibc.git / log
