This patches consolidates all the powerpc ceil{f} implementations on
the generic sysdeps/powerpc/fpu/s_ceil{f}. The generic implementation
uses either the compiler builts for ISA 2.03+ (which generates the frip
instruction) or a generic implementation which uses FP only operations.
It adds a generic implementation (round_to_integer.h) which is shared
with other rounding to integer routines. The resulting code should be
similar in term os performance to previous assembly one.
The IFUNC organization for powerpc64 is also change to be enabled only
for powerpc64 and not for powerpc64le (since minium ISA of 2.08 does not
require the fallback generic implementation).
Checked on powerpc-linux-gnu (built without --with-cpu, with
--with-cpu=power4 and with --with-cpu=power5+ and --disable-multi-arch),
powerpc64-linux-gnu (built without --with-cp and with --with-cpu=power5+
and --disable-multi-arch).
This patch removes the POWER4 optimized mpa optimization used currently
on all powerpc targets. In fact for newer chips, GCC generates *worse*
code than generic implementation as below. One possibilty would to
add ifunc variants for the mpa routines (as x86_64), but it will add
complexity only for older chips (and one would need to check if
power5, power5+, and power6 do benefict from this optimization),
and only for specific implementation (since most used one such
as sin, cos, exp, pow where optimized to avoid calling the slow
multiprecision path).
Checked on powerpc-linux-gnu (built without --with-cpu, with
--with-cpu=power4 and with --with-cpu=power5+ and --disable-multi-arch),
powerpc64-linux-gnu (built without --with-cp and with --with-cpu=power5+
and --disable-multi-arch).
Except the following functions, NPTL implementation assume sem_t
argument (or other arguments) are not NULL, so they would benefit
from having the nonnull attribute.
- sem_close(): can cope with a NULL sem_t and return -1 with error EINVAL;
- sem_destroy(): does nothing at all
Makeconfig: Move -Wl,-rpath-link options before library references
Previously, the -Wl,-rpath-link options came after the libraries
injected using LDLIBS-* variables on the link editor command line for
main programs. As a result, it could happen that installed libraries
that reference glibc libraries used the installed glibc from the system
directories, instead of the glibc from the build tree. This can lead to
link failures if the wrong version of libpthread.so.0 is used, for
instance, due to differences in the internal GLIBC_PRIVATE interfaces,
as seen with memusagestat and -lgd after commit f9b645b4b0a10c43753296ce3fa40053fa44606a ("memusagestat: use local glibc
when linking [BZ #18465]").
The isolation is necessarily imperfect because these installed
libraries are linked against the installed glibc in the system
directories. However, in most cases, the built glibc will be newer
than the installed glibc, and this link is permitted because of the
ABI backwards compatibility glibc provides.
This change is needed to add linker flags which come very early in the
command linke (before LDFLAGS) and are not applied to test programs
(only to installed programs).
While working on enabling D front-end (GDC) in GCC we noticed that
druntime was segfaulting if it is linked dynamically. This was tracked
to DL_RO_DYN_SECTION.
DL_RO_DYN_SECTION lines seem to be copied from MIPS file (which is the
only user of it), but the comment doesn't apply to RISC-V. There is no
such requirement in RISC-V ABI.
[BZ#24484]
* sysdeps/riscv/ldsodefs.h: Remove DL_RO_DYN_SECTION as it is not
required by RISC-V ABI.
Extend BIND_NOW to installed programs with --enable-bind-now
Commit 2d6ab5df3b675e96ee587ae6a8c2ce004c6b1ba9 ("Document and fix
--enable-bind-now [BZ #21015]") extended BIND_NOW to all installed
shared objects. This change also covers installed programs.
Reduce the total time taken by benchtests. The malloc thread test takes 4
minutes to run which is significantly more than most other tests. Reduce
this to a more reasonable 40 seconds. The math tests take 10 seconds each,
eventhough all they do is loop on the same input. Anything more than 1
second runtime is way overkill, so set the limit to 1 second.
* benchtests/Makefile (BENCH_DURATION): Set to 1 second.
* benchtests/bench-malloc-thread.c (BENCH_DURATION): Set to 10 seconds.
Mike Frysinger [Wed, 24 Apr 2019 11:32:22 +0000 (13:32 +0200)]
memusagestat: use local glibc when linking [BZ #18465]
The memusagestat is the only binary that has its own link line which
causes it to be linked against the existing installed C library. It
has been this way since it was originally committed in 1999, but I
don't see any reason as to why. Since we want all the programs we
build locally to be against the new copy of glibc, change the build
to be like all other programs.
Since 9182aa67994 (Fix vDSO l_name for GDB's, BZ#387) the initial link_map
for executable itself and loader will have both l_name and l_libname->name
holding the same value due:
Since the value at ln.name (l_libname->name) will be the same as previously
read. The straightforward fix is just avoid the check and read the new list
entry.
I checked also against binaries issues with old loaders with fix for BZ#387,
and pldd could dump the shared objects.
Checked on x86_64-linux-gnu, i686-linux-gnu, aarch64-linux-gnu, and
powerpc64le-linux-gnu.
[BZ #18035]
* elf/Makefile (tests-container): Add tst-pldd.
* elf/pldd-xx.c: Use _Static_assert in of pldd_assert.
(E(find_maps)): Avoid use alloca, use default read file operations
instead of explicit LFS names, and fix infinite loop.
* elf/pldd.c: Explicit set _FILE_OFFSET_BITS, cleanup headers.
(get_process_info): Use _Static_assert instead of assert, use default
directory operations instead of explicit LFS names, and free some
leadek pointers.
* elf/tst-pldd.c: New file.
malloc: make malloc fail with requests larger than PTRDIFF_MAX (BZ#23741)
As discussed previously on libc-alpha [1], this patch follows up the idea
and add both the __attribute_alloc_size__ on malloc functions (malloc,
calloc, realloc, reallocarray, valloc, pvalloc, and memalign) and limit
maximum requested allocation size to up PTRDIFF_MAX (taking into
consideration internal padding and alignment).
This aligns glibc with gcc expected size defined by default warning
-Walloc-size-larger-than value which warns for allocation larger than
PTRDIFF_MAX. It also aligns with gcc expectation regarding libc and
expected size, such as described in PR#67999 [2] and previously discussed
ISO C11 issues [3] on libc-alpha.
From the RFC thread [4] and previous discussion, it seems that consensus
is only to limit such requested size for malloc functions, not the system
allocation one (mmap, sbrk, etc.).
The implementation changes checked_request2size to check for both overflow
and maximum object size up to PTRDIFF_MAX. No additional checks are done
on sysmalloc, so it can still issue mmap with values larger than
PTRDIFF_T depending on the requested size.
The __attribute_alloc_size__ is for functions that return a pointer only,
which means it cannot be applied to posix_memalign (see remarks in GCC
PR#87683 [5]). The runtimes checks to limit maximum requested allocation
size does applies to posix_memalign.
[BZ #23741]
* malloc/hooks.c (malloc_check, realloc_check): Use
__builtin_add_overflow on overflow check and adapt to
checked_request2size change.
* malloc/malloc.c (__libc_malloc, __libc_realloc, _mid_memalign,
__libc_pvalloc, __libc_calloc, _int_memalign): Limit maximum
allocation size to PTRDIFF_MAX.
(REQUEST_OUT_OF_RANGE): Remove macro.
(checked_request2size): Change to inline function and limit maximum
requested size to PTRDIFF_MAX.
(__libc_malloc, __libc_realloc, _int_malloc, _int_memalign): Limit
maximum allocation size to PTRDIFF_MAX.
(_mid_memalign): Use _int_memalign call for overflow check.
(__libc_pvalloc): Use __builtin_add_overflow on overflow check.
(__libc_calloc): Use __builtin_mul_overflow for overflow check and
limit maximum requested size to PTRDIFF_MAX.
* malloc/malloc.h (malloc, calloc, realloc, reallocarray, memalign,
valloc, pvalloc): Add __attribute_alloc_size__.
* stdlib/stdlib.h (malloc, realloc, reallocarray, valloc): Likewise.
* malloc/tst-malloc-too-large.c (do_test): Add check for allocation
larger than PTRDIFF_MAX.
* malloc/tst-memalign.c (do_test): Disable -Walloc-size-larger-than=
around tests of malloc with negative sizes.
* malloc/tst-posix_memalign.c (do_test): Likewise.
* malloc/tst-pvalloc.c (do_test): Likewise.
* malloc/tst-valloc.c (do_test): Likewise.
* malloc/tst-reallocarray.c (do_test): Replace call to reallocarray
with resulting size allocation larger than PTRDIFF_MAX with
reallocarray_nowarn.
(reallocarray_nowarn): New function.
* NEWS: Mention the malloc function semantic change.
This patch just refactor the assembly implementation to use compiler
builtins instead.
Checked on powerpc-linux-gnu (built without --with-cpu, with
--with-cpu=power4 and with --with-cpu=power5+ and --disable-multi-arch),
powerpc64-linux-gnu (built without --with-cp and with --with-cpu=power5+
and --disable-multi-arch).
Since be2e25bbd78f9fdf the generic ieee754 implementation uses
compiler builtin which generates fabs{f} for all supported targets.
Checked on powerpc-linux-gnu (built without --with-cpu, with
--with-cpu=power4 and with --with-cpu=power5+ and --disable-multi-arch),
powerpc64-linux-gnu (built without --with-cp and with --with-cpu=power5+
and --disable-multi-arch).
mips: Remove rt_sigreturn usage on context function
Similar to powerpc, mips also issues rt_sigreturn for setcontext
case the v0 value saved is not the one set by setcontext or
makecontext. As for powerpc, it is intention is no really supported
since setcontext is not async-signal-safe.
Checked the context tests on mips64-linux-gnu and mips-linux-gnu.
powerpc: Remove rt_sigreturn usage on context function
As described in a recent glibc thread [1], the rt_sigreturn syscall
on setcontext and swapcontext is not used on default use and its
intention is no really supported since neither setcontext nor
swapcontext are async-signal-safe.
Checked on powerpc64-linux-gnu and powerpc-linux-gnu.
Its API is similar to support_capture_subprocess, but rather creates a
new process based on the input path and arguments. Under the hoods it
uses posix_spawn to create the new process.
It also allows the use of other support_capture_* functions to check
for expected results and free the resources.
Checked on x86_64-linux-gnu.
* support/Makefile (libsupport-routines): Add support_subprocess,
xposix_spawn, xposix_spawn_file_actions_addclose, and
xposix_spawn_file_actions_adddup2.
(tst-support_capture_subprocess-ARGS): New rule.
* support/capture_subprocess.h (support_capture_subprogram): New
prototype.
* support/support_capture_subprocess.c (support_capture_subprocess):
Refactor to use support_subprocess and support_capture_poll.
(support_capture_subprogram): New function.
* support/tst-support_capture_subprocess.c (write_mode_to_str,
str_to_write_mode, test_common, parse_int, handle_restart,
do_subprocess, do_subprogram, do_multiple_tests): New functions.
(do_test): Add support_capture_subprogram tests.
* support/subprocess.h: New file.
* support/support_subprocess.c: Likewise.
* support/xposix_spawn.c: Likewise.
* support/xposix_spawn_file_actions_addclose.c: Likewise.
* support/xposix_spawn_file_actions_adddup2.c: Likewise.
* support/xspawn.h: Likewise.
Mike Gerow [Wed, 17 Apr 2019 09:45:34 +0000 (11:45 +0200)]
stdlib/tst-secure-getenv: handle >64 groups
This test would fail unnecessarily if the user running it had more than
64 groups since getgroups returns EINVAL if the size provided is less
than the number of supplementary group IDs. Instead dynamically
determine the number of supplementary groups the user has.
nss_dns: Do not replace root domain with empty string
The purpose of the bp[0] == '.' check is unclear. Only the root domain
starts with '.'. The empty string is accepted as a domain name in many
places, denoting the root, but using it implicitly is confusing.
alloc_buffer: Return unqualified pointer type in alloc_buffer_next
alloc_buffer_next is useful for peeking to the remaining part of the
buffer and update it, with subsequent allocation (once the length
is known) using alloc_buffer_alloc_bytes. This is not as robust
as the other interfaces, but it allows using alloc_buffer with
string-writing interfaces such as snprintf and ns_name_ntop.
Carlos O'Donell [Mon, 8 Apr 2019 21:35:05 +0000 (17:35 -0400)]
malloc: Set and reset all hooks for tracing (Bug 16573)
If an error occurs during the tracing operation, particularly during a
call to lock_and_info() which calls _dl_addr, we may end up calling back
into the malloc-subsystem and relock the loader lock and deadlock. For
all intents and purposes the call to _dl_addr can call any of the malloc
family API functions and so we should disable all tracing before calling
such loader functions. This is similar to the strategy that the new
malloc tracer takes when calling the real malloc, namely that all
tracing ceases at the boundary to the real function and any faults at
that point are the purvue of the library (though the new tracer does
this on a per-thread basis in an MT-safe fashion). Since the new tracer
and the hook deprecation are not yet complete we must fix these issues
where we can.
Tested on x86_64 with no regressions.
Co-authored-by: Kwok Cheung Yeung <kcy@codesourcery.com> Reviewed-by: DJ Delorie <dj@redhat.com>
Replace slow byte-oriented tests in several string benchmarks with the
generic implementations from the string/ directory so the comparisons
are more realistic and useful.
* benchtests/bench-stpcpy.c (SIMPLE_STPCPY): Remove function.
(generic_stpcpy): New function.
* benchtests/bench-stpncpy.c (SIMPLE_STPNCPY): Remove function.
(generic_stpncpy): New function.
* benchtests/bench-strcat.c (SIMPLE_STRCAT): Remove function.
(generic_strcat): New function.
* benchtests/bench-strcpy.c (SIMPLE_STRCPY): Remove function.
(generic_strcpy): New function.
* benchtests/bench-strncat.c (SIMPLE_STRNCAT): Remove function.
(STUPID_STRNCAT): Remove function.
(generic_strncat): New function.
* benchtests/bench-strncpy.c (SIMPLE_STRNCPY): Remove function.
(STUPID_STRNCPY): Remove function.
(generic_strncpy): New function.
* benchtests/bench-strnlen.c (SIMPLE_STRNLEN): Remove function.
(generic_strnlen): New function.
(memchr_strnlen): New function.
* benchtests/bench-strlen.c (generic_strlen): Define for WIDE.
(memchr_strlen): Likewise.
Improve bench-strstr by using an extract from the manual as the input
to make the test more realistic. Use the same input for both found and
fail cases rather than using a memset of '0' for most of the string,
which measures performance of strchr rather than strstr. Add result
checking to catch potential errors. Remove the repeated tests at slightly
different alignments and add more large needle and haystack testcases.
Replace stupid_strstr with an efficient basic implementation. Add the
Two-way implementation to simplify comparisons with much faster generic
implementations.
* benchtests/bench-strstr.c (input): Add realistic input text.
(stupid_strstr): Remove function.
(basic_strstr): Add function.
(twoway_strstr): Add function.
(do_one_test): Add result checking.
(do_test): Use new input text. Remove accidental early matches.
(test_main): Improve range of tests, reduce unaligned cases.
Improve bench-memmem by replacing simple_memmem with a more efficient
implementation. Add the Two-way implementation to enable direct comparison
with the optimized memmem.
resolv: Remove support for RES_USE_INET6 and the inet6 option
This functionality was deprecated in glibc 2.25.
This commit only includes the core changes to remove the
functionality. It does not remove the RES_USE_INET6 handling in the
individual NSS service modules and the res_use_inet6 function.
These changes will happen in future commits.
aarch64: thunderx2 memcpy implementation cleanup and streamlining
Here is the updated patch for improving the long unaligned
code path (the one using "ext" instruction).
1. Always taken conditional branch at the beginning is
removed.
2. Epilogue code is placed after the end of the loop to
reduce the number of branches.
3. The redundant "mov" instructions inside the loop are
gone due to the changed order of the registers in the "ext"
instructions inside the loop, the prologue has additional
"ext" instruction.
4.Updating count in the prologue was hoisted out as
it is the same update for each prologue.
5. Invariant code of the loop epilogue was hoisted out.
6. As the current size of the ext chunk is exactly 16
instructions long "nop" was added at the beginning
of the code sequence so that the loop entry for all the
chunks be aligned.
* sysdeps/aarch64/multiarch/memcpy_thunderx2.S: Cleanup branching
and remove redundant code.
This patch removes the power6 wcsrchr optimization and uses generic
implementation instead. Currently, both power6 and power7 IFUNC variant
resulting binary are essentially the same and the generic implementation
with unrolling loop set to 8 also results in similar performance.
This patch removes the power6 wcschr optimization and uses generic
implementation instead. Currently, both power6 and power7 IFUNC variant
resulting binary are essentially the same and the generic implementation
with unrolling loop set to 8 also results in similar performance.
This patch removes the power6 wcscpy optimization and uses generic
implementation instead. Currently, both power6 and power7 IFUNC variant
resulting binary are essentially the same and the generic implementation
with unrolling loop set to 8 also results in similar performance.
DJ Delorie [Wed, 3 Apr 2019 21:49:43 +0000 (17:49 -0400)]
time/tst-strftime3.c: Disable snprintf warning.
snprintf will only truncate the output if the data its given
is corrupted, but a truncated buffer will not match the
"pristine" data's buffer, which is all we need. So just
disable the warning via the DIAG macros.
ja_JP locale: Add entry for the new Japanese era [BZ #22964]
The Japanese era name will be changed on May 1, 2019. The Japanese
government made a preliminary announcement on April 1, 2019.
The glibc ja_JP locale must be updated to include the new era name for
strftime's alternative year format support.
Checked on x86_64-linux-gnu.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
ChangeLog:
[BZ #22964]
* localedata/locales/ja_JP (LC_TIME): Add entry for the new Japanese
era.
* time/tst-strftime2.c (dates): Add 2019-04-30 and 2019-05-01.
(mkreftable): Add rules for the new Japanese era and the new dates.
Co-authored-by: Rafal Luzynski <digitalfreak@lingonborough.com> Reviewed-by: Carlos O'Donell <carlos@redhat.com>
ChangeLog:
[BZ #24293]
* time/Makefile (LOCALES): Add zh_TW.UTF-8, cmn_TW.UTF-8,
hak_TW.UTF-8, nan_TW.UTF-8, and lzh_TW.UTF-8.
* time/tst-strftime2.c (locales): Likewise.
(dates): Add 1910-04-01, 1911-12-31, 1912-01-01, 1913-04-01,
2010-04-01, and 2011-04-01.
(mkreftable): Add rules for the new locales and the new dates.
time/tst-strftime2.c: Make the file easier to maintain
Express the years as full Gregorian years (e.g., 1988 instead of 88)
and months with natural numbers (1-12 rather than 0-11).
Compare actual dates rather than indexes when selecting the era name.
Declare the local variable era as a string character pointer rather
than an array of chars where the actual string is copied which might
lead to potential buffer overflows in future.
Co-authored-by: Rafal Luzynski <digitalfreak@lingonborough.com> Reviewed-by: Carlos O'Donell <carlos@redhat.com>
ChangeLog:
* time/tst-strftime2.c (date_t): Explicitly define the type.
(dates): Use natural month and year numbers to express a date.
(is_before): New function to compare dates.
(mkreftable): Minor improvements to simplify maintenance.
(do_test): Reflect the changes in dates array.
DJ Delorie [Wed, 27 Mar 2019 21:44:51 +0000 (17:44 -0400)]
Fix strptime era handling, add more strftime tests [BZ #24394]
Test the transition points between all the currently listed Japanese
era name changes. This includes testing the transition between the
first year date and the second year date. This test will help test
the upcoming Japanese era name change.
Also fixes a fencepost error where the era name isn't properly parsed
by strptime in the last (partial) year of the era.
Example: if an era change happens in Feb 1990, and again in Aug 1995,
that's 5.5 years long, but the 0.5 year wasn't accounted for.
alpha: Improve sysdeps/alpha/divqu.S and sysdeps/alpha/remqu.S
* sysdeps/alpha/divqu.S (__divqu): Move save of $f0 and excb after
conditional branch to DIVBYZERO. Fix unwind info.
* sysdeps/alpha/remqu.S (__remqu): Move saves of $f0, $f1, $f2 and
excb after conditional branch to $powerof2. Add missing unop
instructions and .align directives and reorder instructions to
match __divqu.
alpha: Do not redefine __NR_shmat or __NR_osf_shmat
Fixes build using v5.1-rc1 headers.
The kernel has cleaned up how these are defined. Previous behavior
was to define __NR_osf_shmat as 209 and not define __NR_shmat.
Current behavior is to define __NR_shmat as 209 and then define
__NR_osf_shmat as __NR_shmat.
* sysdeps/unix/sysv/linux/alpha/kernel-features.h (__NR_shmat):
Do not redefine.
* sysdeps/unix/sysv/linux/alpha/sysdep.h (__NR_osf_shmat):
Do not redefine.
RISC-V: Fix `test' operand error with soft-float ABI being configured
Fix a:
.../sysdeps/unix/sysv/linux/riscv/configure: line 181: test: =: unary operator expected
message produced by the RISC-V configure fragment with the soft-float
ABI selected, caused by $libc_cv_riscv_float_abi evaluating to nil in
the invocation of `test $libc_cv_riscv_float_abi = no'.
* sysdeps/unix/sysv/linux/riscv/configure.ac: Quote
$libc_cv_riscv_float_abi in `test' invocation.
* sysdeps/unix/sysv/linux/riscv/configure: Regenerate.
Paul A. Clarke [Sat, 30 Mar 2019 00:16:34 +0000 (19:16 -0500)]
[powerpc] Use __builtin_{mffs,mtfsf}
Replace inline asm uses of the "mffs" and "mtfsf" instructions with
the analogous GCC builtins.
__builtin_mffs and __builtin_mtfsf are both available in GCC 5 and above.
Given the minimum GCC level for GLibC is now GCC 6.2, it is safe to use
these builtins without restriction.
This patch enable the builtin usage for clang for the C99 functions
fpclassify, isfinite, isnormal, isnan, isinf, and sigbit. This allows
clang optimize the calls on frontend instead of call the appropriate
glibc symbols.
Checked on aarch64-linux-gnu and x86_64-linux-gnu. I checked the supported
version for each builtin based on released version from clang/llvm.
* math/math.h (fpclassify, isfinite, isnormal, isnan): Use builtin for
clang 2.8.
(signbit): Use builtin for clang 3.3.
(isinf): Use builtin for clang 3.7.
This file is not used anywhere since removal of {k,e}_rem_pio2f.c
(commit ca3aac57efa89).
Checked with a build for powerpc-linux-gnu (with --with-cpu=power4
and --with-cpu=power7), powerpc64-linux-gnu (with --with-cpu=power4
and --with-cpu=power7), and powerpc64le-linux (with --with-cpu=power8).
Mike Crowe [Mon, 25 Mar 2019 12:28:18 +0000 (09:28 -0300)]
nptl: Convert tst-sem5 & tst-sem13 to use libsupport
Checked on x86_64-linux-gnu and i686-linux-gnu.
* nptl/tst-sem5.c: Remove unused headers. Add <support/check.h>.
(do_test) Use libsupport test macros rather than hand-coded
conditionals and error messages. Ensure that sem_init returns zero
rather than not -1. Use <support/test-driver.c> rather than
test-skeleton.c.
* nptl/tst-sem13.c: Add <support/check.h>. (do_test) Use libsupport
test macros rather than hand-coded conditionals and error messages.
Use <support/test-driver.c> rather than test-skeleton.c.
Wilco Dijkstra [Tue, 29 Jan 2019 17:43:45 +0000 (17:43 +0000)]
Add generic hp-timing support
Add missing generic hp_timing support. It uses clock_gettime (CLOCK_MONOTONIC)
which has unspecified starting time, nano-second accuracy, and should faster on
architectures that implementes the symbol as vDSO.
Checked on aarch64-linux-gnu, x86_64-linux-gnu, and i686-linux-gnu. I also
checked the builds for all afected ABIs.
This patch refactor how hp-timing is used on loader code for statistics
report. The HP_TIMING_AVAIL and HP_SMALL_TIMING_AVAIL are removed and
HP_TIMING_INLINE is used instead to check for hp-timing avaliability.
For alpha, which only defines HP_SMALL_TIMING_AVAIL, the HP_TIMING_INLINE
is set iff for IS_IN(rtld).
Checked on aarch64-linux-gnu, x86_64-linux-gnu, and i686-linux-gnu. I also
checked the builds for all afected ABIs.
This patch removes the HP_TIMING_BITS usage for fast random bits and replace
with clock_gettime (CLOCK_MONOTONIC). It has unspecified starting time and
nano-second accuracy, so its randomness is significantly better than
gettimeofday.
Althoug it should incur in more overhead (specially for architecture that
support hp-timing), the symbol is also common implemented as a vDSO.
Checked on aarch64-linux-gnu, x86_64-linux-gnu, and i686-linux-gnu. I also
checked on a i686-gnu build.
* include/random-bits.h: New file.
* resolv/res_mkquery.c [HP_TIMING_AVAIL] (RANDOM_BITS,
(__res_context_mkquery): Remove usage hp-timing usage and replace with
random_bits.
* resolv/res_send.c [HP_TIMING_AVAIL] (nameserver_offset): Likewise.
* sysdeps/posix/tempname.c [HP_TIMING_AVAIL] (__gen_tempname):
Likewise.
With clock_getres, clock_gettime, and clock_settime refactor to remove the
generic CLOCK_PROCESS_CPUTIME_ID and CLOCK_THREAD_CPUTIME_ID support through
hp-timing, there is no usage of internal __get_clockfreq. This patch removes
both generic and Linux implementation..
Checked with a build against aarch64-linux-gnu, i686-linux-gnu, ia64-linux-gnu,
sparc64-linux-gnu, powerpc-linux-gnu-power4.
The Linux 3.2 clock_getres kernel code (kernel/posix-cpu-timers.c)
issued for clock_getres CLOCK_PROCESS_CPUTIME_ID (process_cpu_clock_getres)
and CLOCK_THREAD_CPUTIME_ID (thread_cpu_clock_getres) call
posix_cpu_clock_getres. And it fails on check_clock only if an invalid
clock is used (not the case) or if we pass an invalid the pid/tid in
29 msb of clock_id (not the case either).
This patch assumes that clock_getres syscall always support
CLOCK_PROCESS_CPUTIME_ID and CLOCK_THREAD_CPUTIME_ID, so there is no need
to fallback to hp-timing support for _SC_MONOTONIC_CLOCK neither to issue
the syscall to certify the clock_id is supported bt the kernel. This
allows simplify the sysconf support to always use the syscall.
it also removes ia64 itc drift check and assume kernel handles it correctly.
Checked on aarch64-linux-gnu, x86_64-linux-gnu, and i686-linux-gnu.
* sysdeps/unix/sysv/linux/ia64/has_cpuclock.c: Remove file.
* sysdeps/unix/sysv/linux/ia64/sysconf.c: Likewise.
* sysdeps/unix/sysv/linux/sysconf.c (has_cpuclock): Remove function.
(__sysconf): Assume kernel support for _SC_MONOTONIC_CLOCK,
_SC_CPUTIME, and _SC_THREAD_CPUTIME.
This patch removes CLOCK_THREAD_CPUTIME_ID and CLOCK_PROCESS_CPUTIME_ID support
from clock_gettime and clock_settime generic implementation. For Linux, kernel
already provides supports through the syscall and Hurd HTL lacks
__pthread_clock_gettime and __pthread_clock_settime internal implementation.
As described in clock_gettime man-page [1] on 'Historical note for SMP
system', implementing CLOCK_{THREAD,PROCESS}_CPUTIME_ID with timer registers
is error-prone and susceptible to timing and accurary issues that the libc
can not deal without kernel support.
This allows removes unused code which, however, still incur in some runtime
overhead in thread creation (the struct pthread cpuclock_offset
initialization).
If hurd eventually wants to support them it should either either implement as
a kernel facility (or something related due its architecture) or in system
specific implementation.
Checked on aarch64-linux-gnu, x86_64-linux-gnu, and i686-linux-gnu. I also
checked on a i686-gnu build.
Stefan Liebler [Fri, 22 Mar 2019 10:14:09 +0000 (11:14 +0100)]
S390: Add arch13 memmem ifunc variant.
This patch introduces the new arch13 ifunc variant for memmem.
For needles longer than 9 bytes it is relying on the common-code
implementation. For shorter needles it is using the new vstrs instruction
which is able to search a substring within a vector register.
ChangeLog:
* sysdeps/s390/Makefile (sysdep_routines): Add memmem-arch13.
* sysdeps/s390/ifunc-memmem.h (HAVE_MEMMEM_ARCH13, MEMMEM_ARCH13,
MEMMEM_Z13_ONLY_USED_AS_FALLBACK, HAVE_MEMMEM_IFUNC_AND_ARCH13_SUPPORT):
New defines.
* sysdeps/s390/memmem-arch13.S: New file.
* sysdeps/s390/memmem-vx.c: Omit GI symbol for z13 memmem ifunc variant
if it is only used as fallback.
* sysdeps/s390/memmem.c (memmem): Add arch13 variant in ifunc selector.
* sysdeps/s390/multiarch/ifunc-impl-list.c
(__libc_ifunc_impl_list): Add ifunc variant for arch13 memmem.
Stefan Liebler [Fri, 22 Mar 2019 10:14:08 +0000 (11:14 +0100)]
S390: Add arch13 strstr ifunc variant.
This patch introduces the new arch13 ifunc variant for strstr.
For needles longer than 9 charachters it is relying on the common-code
implementation. For shorter needles it is using the new vstrs instruction
which is able to search a substring within a vector register.
ChangeLog:
* sysdeps/s390/Makefile (sysdep_routines): Add strstr-arch13.
* sysdeps/s390/ifunc-strstr.h (HAVE_STRSTR_ARCH13, STRSTR_ARCH13,
STRSTR_Z13_ONLY_USED_AS_FALLBACK, HAVE_STRSTR_IFUNC_AND_ARCH13_SUPPORT):
New defines.
* sysdeps/s390/multiarch/ifunc-impl-list.c
(__libc_ifunc_impl_list): Add ifunc variant for arch13 strstr.
* sysdeps/s390/strstr-arch13.S: New file.
* sysdeps/s390/strstr-vx.c: Omit GI symbol for z13 strstr ifunc variant
if it is only used as fallback.
* sysdeps/s390/strstr.c (strstr): Add arch13 variant in ifunc selector.
Stefan Liebler [Fri, 22 Mar 2019 10:14:08 +0000 (11:14 +0100)]
S390: Add arch13 memmove ifunc variant.
This patch introduces the new arch13 ifunc variant for memmove.
For the forward or non-overlapping case it is just using memcpy.
For the backward case it relies on the new instruction mvcrl.
The instruction copies up to 256 bytes at once.
In case of an overlap, it copies the bytes like copying them
one by one starting from right to left.
ChangeLog:
* sysdeps/s390/ifunc-memcpy.h (HAVE_MEMMOVE_ARCH13, MEMMOVE_ARCH13
HAVE_MEMMOVE_IFUNC_AND_ARCH13_SUPPORT): New defines.
* sysdeps/s390/memcpy-z900.S: Add arch13 memmove implementation.
* sysdeps/s390/memmove.c (memmove): Add arch13 variant in
ifunc selector.
* sysdeps/s390/multiarch/ifunc-impl-list.c
(__libc_ifunc_impl_list): Add ifunc variant for arch13 memmove.
* sysdeps/s390/multiarch/ifunc-resolve.h (S390_STFLE_BITS_ARCH13_MIE3,
S390_IS_ARCH13_MIE3): New defines.
Stefan Liebler [Fri, 22 Mar 2019 10:14:08 +0000 (11:14 +0100)]
S390: Add configure check to detect support for arch13.
Add two configure checks which detect if arch13 is supported
by the assembler at all - by explicitely setting the machine -
and if it is supported with default settings.
ChangeLog:
* config.h.in (HAVE_S390_MIN_ARCH13_ZARCH_ASM_SUPPORT,
HAVE_S390_ARCH13_ASM_SUPPORT): New undefine.
* sysdeps/s390/configure.ac: Add checks for arch13 support.
* sysdeps/s390/configure: Regenerated.
Fix parentheses error in iconvconfig.c and ld-collate.c [BZ #24372]
When -Werror=parentheses is in use, iconvconfig.c builds fail with:
iconvconfig.c: In function ‘write_output’:
iconvconfig.c:1084:34: error: suggest parentheses around ‘+’ inside ‘>>’ [-Werror=parentheses]
hash_size = next_prime (nnames + nnames >> 1);
~~~~~~~^~~~~~~~
This patch adds parentheses to the expression. Not where suggested by
the compiler warning, but where it produces the expected result, i.e.:
where it has the effect of multiplying nnames by 1.5.
DJ Delorie [Thu, 21 Mar 2019 03:56:59 +0000 (23:56 -0400)]
iconv, localedef: avoid floating point rounding differences [BZ #24372]
Two cases of "int * 1.4" may result in imprecise results, which
in at least one case resulted in i686 and x86-64 producing
different locale files. This replaced that floating point multiply
with integer operations. While the hash table margin is increased
from 40% to 50%, testing shows only 2% increase in overall size
of the locale archive.
mansayk [Sat, 16 Mar 2019 08:05:22 +0000 (11:05 +0300)]
tt_RU: Fix orthographic mistakes in day and abday sections [BZ #24296]
This commit fixes some errors and converts all weekday names to lowercase.
The content is synchronized with CLDR-34 now, but trailing dots are removed
from abday values in order to maintain consistency with the previous values
and with many other locales which do the same.
[BZ #24296]
* localedata/locales/tt_RU (day): Update from CLDR-34, fix errors.
(abday): Likewise, but remove the trailing dots.
Joseph Myers [Tue, 19 Mar 2019 14:02:17 +0000 (14:02 +0000)]
Add AArch64 HWCAPs from Linux 5.0.
This patch adds new AArch64 HWCAPs from Linux 5.0 to the AArch64
bits/hwcap.h and dl-procinfo.c.
Tested (compilation only) with build-many-glibcs.py for
aarch64-linux-gnu.
* sysdeps/unix/sysv/linux/aarch64/bits/hwcap.h (HWCAP_SB): New
macro.
(HWCAP_PACA): Likewise.
(HWCAP_PACG): Likewise.
* sysdeps/unix/sysv/linux/aarch64/dl-procinfo.c (_DL_HWCAP_COUNT):
Increase to 32.
(_dl_aarch64_cap_flags): Add new entries for new HWCAPs.
Joseph Myers [Fri, 15 Mar 2019 18:18:40 +0000 (18:18 +0000)]
Update syscall-names.list for Linux 5.0.
This patch updates sysdeps/unix/sysv/linux/syscall-names.list for
Linux 5.0. Based on testing with build-many-glibcs.py, the only new
entry needed is for old_getpagesize (a newly added __NR_* name for an
old syscall on ia64). (Because 5.0 changes how syscall tables are
handled in the kernel, checking diffs wasn't a useful way of looking
for new syscalls in 5.0 as most of the syscall tables were moved to
the new representation without actually adding any syscalls to them.)
Tested with build-many-glibcs.py.
* sysdeps/unix/sysv/linux/syscall-names.list: Update kernel
version to 5.0.
(old_getpagesize): New syscall.
Adam Maris [Thu, 14 Mar 2019 20:51:16 +0000 (16:51 -0400)]
malloc: Check for large bin list corruption when inserting unsorted chunk
Fixes bug 24216. This patch adds security checks for bk and bk_nextsize pointers
of chunks in large bin when inserting chunk from unsorted bin. It was possible
to write the pointer to victim (newly inserted chunk) to arbitrary memory
locations if bk or bk_nextsize pointers of the next large bin chunk
got corrupted.
The stub implementations are turned into compat symbols.
Linux actually has two reserved system call numbers (for getpmsg
and putpmsg), but these system calls have never been implemented,
and there are no plans to implement them, so this patch replaces
the wrappers with the generic stubs.
According to <https://bugzilla.redhat.com/show_bug.cgi?id=436349>,
the presence of the XSI STREAMS declarations is a minor portability
hazard because they are not actually implemented.
This commit does not change the TIRPC support code in
sunrpc/rpc_svcout.c. It uses additional XTI functionality and
therefore never worked with glibc.
Florian Weimer [Thu, 14 Mar 2019 14:02:52 +0000 (15:02 +0100)]
nss: Fix tst-nss-files-alias-truncated for default --as-needed linking
Linking to the NSS module directly does not work if the linker defaults
to --as-needed because it will remove the apparently unused DSO
reference and not generate a DT_NEEDED entry. Use an explicit dlopen
call, like in the other chroot tests involving NSS modules.
Zack Weinberg [Thu, 14 Mar 2019 13:44:22 +0000 (09:44 -0400)]
scripts/check-obsolete-constructs.py: Process all headers as UTF-8.
A few of our installed headers contain UTF-8 in comments.
check-obsolete-constructs opened files without explicitly specifying
their encoding, so it would barf on these headers if “make check” was
run in a non-UTF-8 locale.
* scripts/check-obsolete-constructs.py (HeaderChecker.check):
Specify encoding="utf-8" when opening headers to check.
Joseph Myers [Wed, 13 Mar 2019 22:03:37 +0000 (22:03 +0000)]
Use Linux 5.0 in build-many-glibcs.py.
This patch makes build-many-glibcs.py use Linux 5.0 in place of 4.20
(now that the test change required to avoid false positives with ulong
in kernel headers has been committed). This includes adjusting the
logic to compute a tarball URL to handle different major version
numbers (rather than changing the path to hardcode v5.x in place of
v4.x, as someone might still wish to check out a v4.x version).
Tested that build-many-glibcs.py successfully checks out Linux 5.0
sources after this patch.
* scripts/build-many-glibcs.py (Context.checkout): Default Linux
version to 5.0.
(Context.checkout_tar): Handle variable major version for Linux
kernel.
Zack Weinberg [Mon, 11 Mar 2019 14:59:45 +0000 (10:59 -0400)]
Move inttypes.h and stdint.h to stdlib.
inttypes.h and stdint.h are in sysdeps/generic, but there are no other
versions of these headers anywhere in the source tree, so they aren’t
actually system-dependent. Move them to the subdirectory that
installs them (stdlib).
Reviewed-by: Joseph Myers <joseph@codesourcery.com>
* sysdeps/generic/inttypes.h, sysdeps/generic/stdint.h:
Move to stdlib.
* include/inttypes.h: Adjust to match.
* include/stdint.h: New wrapper.
Zack Weinberg [Mon, 11 Mar 2019 14:59:27 +0000 (10:59 -0400)]
Use a proper C tokenizer to implement the obsolete typedefs test.
The test for obsolete typedefs in installed headers was implemented
using grep, and could therefore get false positives on e.g. “ulong”
in a comment. It was also scanning all of the headers included by
our headers, and therefore testing headers we don’t control, e.g.
Linux kernel headers.
This patch splits the obsolete-typedef test from
scripts/check-installed-headers.sh to a separate program,
scripts/check-obsolete-constructs.py. Being implemented in Python,
it is feasible to make it tokenize C accurately enough to avoid false
positives on the contents of comments and strings. It also only
examines $(headers) in each subdirectory--all the headers we install,
but not any external dependencies of those headers. Headers whose
installed name starts with finclude/ are ignored, on the assumption
that they contain Fortran.
It is also feasible to make the new test understand the difference
between _defining_ the obsolete typedefs and _using_ the obsolete
typedefs, which means posix/{bits,sys}/types.h no longer need to be
exempted. This uncovered an actual bug in bits/types.h: __quad_t and
__u_quad_t were being used to define __S64_TYPE, __U64_TYPE,
__SQUAD_TYPE and __UQUAD_TYPE. These are changed to __int64_t and
__uint64_t respectively. This is a safe change, despite the comments
in bits/types.h claiming a difference between __quad_t and __int64_t,
because those comments are incorrect. In all current ABIs, both
__quad_t and __int64_t are ‘long’ when ‘long’ is a 64-bit type, and
‘long long’ when ‘long’ is a 32-bit type, and similarly for __u_quad_t
and __uint64_t. (Changing the types to be what the comments say they
are would be an ABI break, as it affects C++ name mangling.) This
patch includes a minimal change to make the comments not completely
wrong.
sys/types.h was defining the legacy BSD u_intN_t typedefs using a
construct that was not necessarily consistent with how the C99 uintN_t
typedefs are defined, and is also too complicated for the new script to
understand (it lexes C relatively accurately, but it does not attempt
to expand preprocessor macros, nor does it do any actual parsing).
This patch cuts all of that out and uses bits/types.h's __uintN_t typedefs
to define u_intN_t instead. This is verified to not change the ABI on
any supported architecture, via the c++-types test, which means u_intN_t
and uintN_t were, in fact, consistent on all supported architectures.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
* scripts/check-obsolete-constructs.py: New test script.
* scripts/check-installed-headers.sh: Remove tests for
obsolete typedefs, superseded by check-obsolete-constructs.py.
* Rules: Run scripts/check-obsolete-constructs.py over $(headers)
as a special test. Update commentary.
* posix/bits/types.h (__SQUAD_TYPE, __S64_TYPE): Define as __int64_t.
(__UQUAD_TYPE, __U64_TYPE): Define as __uint64_t.
Update commentary.
* posix/sys/types.h (__u_intN_t): Remove.
(u_int8_t): Typedef using __uint8_t.
(u_int16_t): Typedef using __uint16_t.
(u_int32_t): Typedef using __uint32_t.
(u_int64_t): Typedef using __uint64_t.
Stefan Liebler [Wed, 13 Mar 2019 09:45:35 +0000 (10:45 +0100)]
Fix output of LD_SHOW_AUXV=1.
Starting with commit 1616d034b61622836d3a36af53dcfca7624c844e
the output was corrupted on some platforms as _dl_procinfo
was called for every auxv entry and on some architectures like s390
all entries were represented as "AT_HWCAP".
This patch is removing the condition and let _dl_procinfo decide if
an entry is printed in a platform specific or generic way.
This patch also adjusts all _dl_procinfo implementations which assumed
that they are only called for AT_HWCAP or AT_HWCAP2. They are now just
returning a non-zero-value for entries which are not handled platform
specifc.
ChangeLog:
* elf/dl-sysdep.c (_dl_show_auxv): Remove condition and always
call _dl_procinfo.
* sysdeps/unix/sysv/linux/s390/dl-procinfo.h (_dl_procinfo):
Ignore types other than AT_HWCAP.
* sysdeps/sparc/dl-procinfo.h (_dl_procinfo): Likewise.
* sysdeps/unix/sysv/linux/i386/dl-procinfo.h (_dl_procinfo):
Likewise.
* sysdeps/powerpc/dl-procinfo.h (_dl_procinfo): Adjust comment
in the case of falling back to generic output mechanism.
* sysdeps/unix/sysv/linux/arm/dl-procinfo.h (_dl_procinfo):
Likewise.
Florian Weimer [Tue, 12 Mar 2019 10:40:47 +0000 (11:40 +0100)]
resolv: Enable full ICMP errors for UDP DNS sockets [BZ #24047]
The Linux kernel suppresses some ICMP error messages by default for
UDP sockets. This commit enables full ICMP error reporting,
hopefully resulting in faster failover to working name servers.
git://sourceware.org / glibc.git / log