Currently strsep calls strpbrk is is now a veneer to strcspn. Calling
strcspn directly is faster. Since it handles a delimiter string of size
1 as a special case, this is not needed in strsep itself. Although this
means there is a slightly higher overhead if the delimiter size is 1,
all other cases are slightly faster. The overall performance gain is 5-10%
on AArch64.
The string/bits/string2.h header contains optimizations for constant
delimiters of size 1-3. Benchmarking these showed similar performance for
size 1 (since in all cases strchr/strchrnul is used), while size 2 and 3
can give up to 2x speedup for small input strings. However if these cases
are common it seems much better to add this optimization to strcspn.
So move these header optimizations to string-inlines.c.
Improve the strsep benchmark so that it actually benchmarks something.
The current version contains a delimiter character at every position in the
input string, so there is very little work to do, and the extremely inefficent
simple_strsep implementation appears fastest in every case. The new version
has either no match in the input for the fail case and a match halfway in the
input for the success case. The input is then restored so that each iteration
does exactly the same amount of work. Reduce the number of testcases since
simple_strsep takes a lot of time now.
* benchtests/bench-strsep.c (oldstrsep): Add old implementation.
(do_one_test) Restore original string so iteration works.
* string/string-inlines.c (do_test): Create better input strings.
(test_main) Reduce number of testruns.
* string/string-inlines.c (__old_strsep_1c): New function.
(__old_strsep_2c): Likewise.
(__old_strsep_3c): Likewise.
* string/strsep.c (__strsep): Remove case of small delim string.
Call strcspn directly rather than strpbrk.
* string/bits/string2.h (__strsep): Remove define.
(__strsep_1c): Remove.
(__strsep_2c): Remove.
(__strsep_3c): Remove.
(strsep): Remove.
* sysdeps/unix/sysv/linux/internal_statvfs.c
(__statvfs_getflags): Rename to __strsep.
Commit 7a5e3d9d63 (elf: Assume TLS is
initialized in _dl_map_object_from_fd) removed the last call of
_dl_tls_setup, but did not remove the function itself.
With stack protection enabled, these files have external symbol
references for the first time, so the fact that they are not compiled
with -fPIE and are then linked into a -pie binary starts to hurt.
Stack-protection on .o files in csu/ must be suppressed for the sake of
library startup code. This also suppresses stack-protection in tests
(which are also covered by CFLAGS-.o), though this is neither necessary
nor desirable.
So impose the rule that .o files in csu/ are necessarily C startup code,
and move the few tests in there into misc/ instead.
Texinfo @vindex commands add entries to the Variable and Constant
Macro Index. Similarly, @items in @vtables are automatically indexed.
A number of @tables exist where all @items are @vindexed or all @items
are variables, but not indexed, suggesting an optimization by
converting such @tables to @vtables and dropping the @vindex.
Using a @vtable provides a context for processing @items whereby it
can be known the @items should have header and standards annotations.
This commit converts @tables of such @items to @vtables in order to
establish a framework for automated processing.
A pleasant consequence of these changes is that @items previously
lacking a @vindex are present in the Variable and Constant Macro Index
now. @vindex entries previously detected by summary.awk will still be
detected as @items with appropriate annotations.
The @vtable of the NSS databases is converted to a @table because 1)
those @items are not variables (and will no longer appear in the
Variable and Constant Macro Index) and 2) they do not need header and
standards annotations, so the incorrect context is fixed.
* manual/nss.texi: Change incorrect @vtable to @table.
* manual/arith.texi: Convert @tables of variables to @vtables
and remove unnecessary indexing.
* manual/filesys.texi: Likewise.
* manual/llio.texi: Likewise.
* manual/memory.texi: Likewise.
* manual/process.texi: Likewise.
* manual/resource.texi: Likewise.
* manual/search.texi: Likewise.
* manual/signal.texi: Likewise.
* manual/socket.texi: Likewise.
* manual/stdio.texi: Likewise.
* manual/sysinfo.texi: Likewise.
* manual/syslog.texi: Likewise.
* manual/terminal.texi: Likewise.
* manual/time.texi: Likewise.
* manual/users.texi: Likewise.
TS 18661-1 defines roundeven functions that round a floating-point
number to the nearest integer, in that floating-point type, with ties
rounding to even (whereas the round functions round ties away from
zero). As with other such functions, they raise no exceptions apart
from "invalid" for signaling NaNs. There was a previous user request
for this functionality in glibc in
<https://sourceware.org/ml/libc-help/2015-02/msg00005.html>.
This patch implements these functions for glibc. The implementations
use integer bit-manipulation (or roundeven on the high and low parts,
in the IBM long double case). It's possible that there may be faster
approaches on some architectures (in particular, on AArch64 the frintn
instruction should do exactly what's required); I'll leave it to
architecture maintainers or others interested to implement such
architecture-specific versions if desired. (Where architectures have
instructions to round to nearest integer in the current rounding mode,
implementations saving and restoring the rounding mode - and dealing
with exceptions if those instructions generate "inexact" - are also
possible, though their performance depends on the cost of manipulating
exceptions / rounding mode state.)
Tested for x86_64, x86, mips64 and powerpc.
* math/bits/mathcalls.h [__GLIBC_USE (IEC_60559_BFP_EXT)]
(roundeven): New declaration.
* math/tgmath.h [__GLIBC_USE (IEC_60559_BFP_EXT)] (roundeven): New
macro.
* math/Versions (roundeven): New libm symbol at version
GLIBC_2.25.
(roundevenf): Likewise.
(roundevenl): Likewise.
* math/Makefile (libm-calls): Add s_roundevenF.
* math/libm-test.inc (roundeven_test_data): New array.
(roundeven_test): New function.
(main): Call roundeven_test.
* math/test-tgmath.c (NCALLS): Increase to 134.
(F(compile_test)): Call roundeven.
(F(roundeven)): New function.
* manual/arith.texi (Rounding Functions): Document roundeven,
roundevenf and roundevenl.
* manual/libm-err-tab.pl (@all_functions): Add roundeven.
* include/math.h (roundeven): Use libm_hidden_proto.
* sysdeps/ieee754/dbl-64/s_roundeven.c: New file.
* sysdeps/ieee754/dbl-64/wordsize-64/s_roundeven.c: Likewise.
* sysdeps/ieee754/flt-32/s_roundevenf.c: Likewise.
* sysdeps/ieee754/ldbl-128/s_roundevenl.c: Likewise.
* sysdeps/ieee754/ldbl-128ibm/s_roundevenl.c: Likewise.
* sysdeps/ieee754/ldbl-96/s_roundevenl.c: Likewise.
* sysdeps/ieee754/ldbl-opt/Makefile (libnldbl-calls): Add
roundeven.
(CFLAGS-nldbl-roundeven.c): New variable.
* sysdeps/ieee754/ldbl-opt/nldbl-roundeven.c: New file.
* sysdeps/nacl/libm.abilist: Update.
* sysdeps/unix/sysv/linux/aarch64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/alpha/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/arm/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/hppa/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/i386/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/ia64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/m68k/coldfire/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/m68k/m680x0/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/microblaze/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/mips/mips32/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/mips/mips64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/nios2/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/fpu/libm.abilist:
Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/nofpu/libm.abilist:
Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/libm-le.abilist:
Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/libm.abilist:
Likewise.
* sysdeps/unix/sysv/linux/s390/s390-32/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/s390/s390-64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/sh/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/sparc/sparc32/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/sparc/sparc64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/tile/tilegx/tilegx32/libm.abilist:
Likewise.
* sysdeps/unix/sysv/linux/tile/tilegx/tilegx64/libm.abilist:
Likewise.
* sysdeps/unix/sysv/linux/tile/tilepro/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/x86_64/64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/x86_64/x32/libm.abilist: Likewise.
In order to support float128 tests, the calls to snprintf, which does
not support the type __float128, are replaced with calls to
strfrom{f,d,l}.
Tested for powerpc64le, s390, and x64_64.
This patch decrements the adapt_count while unlocking the futex
instead of before aquiring the futex as it is done on power, too.
Furthermore a transaction is only started if the futex is currently free.
This check is done after starting the transaction, too.
If the futex is not free and the transaction nesting depth is one,
we can simply end the started transaction instead of aborting it.
The implementation of this check was faulty as it always ended the
started transaction. By using the fallback path, the the outermost
transaction was aborted. Now the outermost transaction is aborted
directly.
This patch also adds some commentary and aligns the code in
elision-trylock.c to the code in elision-lock.c as possible.
ChangeLog:
* sysdeps/unix/sysv/linux/s390/lowlevellock.h
(__lll_unlock_elision, lll_unlock_elision): Add adapt_count argument.
* sysdeps/unix/sysv/linux/s390/elision-lock.c:
(__lll_lock_elision): Decrement adapt_count while unlocking
instead of before locking.
* sysdeps/unix/sysv/linux/s390/elision-trylock.c
(__lll_trylock_elision): Likewise.
* sysdeps/unix/sysv/linux/s390/elision-unlock.c:
(__lll_unlock_elision): Likewise.
This patch implements __libc_tbegin_retry macro which is equivalent to
gcc builtin __builtin_tbegin_retry, except the changes which were applied
to __libc_tbegin in the previous patch.
If tbegin aborts with _HTM_TBEGIN_TRANSIENT. Then this macros restores
the fpc, fprs and automatically retries up to retry_cnt tbegins.
Further saving of the state is omitted as it is already saved in the
first round. Before retrying a further transaction, the
transaction-abort-assist instruction is used to support the cpu.
This macro is now used in function __lll_lock_elision.
ChangeLog:
* sysdeps/unix/sysv/linux/s390/htm.h(__libc_tbegin_retry): New macro.
* sysdeps/unix/sysv/linux/s390/elision-lock.c (__lll_lock_elision):
Use __libc_tbegin_retry macro.
This patch defines __libc_tbegin, __libc_tend, __libc_tabort and
__libc_tx_nesting_depth in htm.h which replaces the direct usage of
equivalent gcc builtins.
We have to use an own inline assembly instead of __builtin_tbegin,
as tbegin has to filter program interruptions which can't be done with
the builtin. Before this change, e.g. a segmentation fault within a
transaction, leads to a coredump where the instruction pointer points
behind the tbegin instruction instead of real failing one.
Now the transaction aborts and the code should be reexecuted by the
fallback path without transactions. The segmentation fault will
produce a coredump with the real failing instruction.
The fpc is not saved before starting the transaction. If e.g. the
rounging mode is changed and the transaction is aborting afterwards,
the builtin will not restore the fpc. This is now done with the
__libc_tbegin macro.
Now the call saved fprs have to be saved / restored in the
__libc_tbegin macro. Using the gcc builtin had forced the saving /
restoring of fprs at begin / end of e.g. __lll_lock_elision function.
The new macro saves these fprs before tbegin instruction and only
restores them on a transaction abort. Restoring is not needed on
a successfully started transaction.
The used inline assembly does not clobber the fprs / vrs!
Clobbering the latter ones would force the compiler to save / restore
the call saved fprs as those overlap with the vrs, but they only
need to be restored if the transaction fails. Thus the user of the
tbegin macros has to compile the file / function with -msoft-float.
It prevents gcc from using fprs / vrs.
ChangeLog:
* sysdeps/unix/sysv/linux/s390/Makefile (elision-CFLAGS):
Add -msoft-float.
* sysdeps/unix/sysv/linux/s390/htm.h: New File.
* sysdeps/unix/sysv/linux/s390/elision-lock.c:
Use __libc_t* transaction macros instead of __builtin_t*.
* sysdeps/unix/sysv/linux/s390/elision-trylock.c: Likewise.
* sysdeps/unix/sysv/linux/s390/elision-unlock.c: Likewise.
This uses atomic operations to access lock elision metadata that is accessed
concurrently (ie, adapt_count fields). The size of the data is less than a
word but accessed only with atomic loads and stores.
See also x86 commit ca6e601a9d:
"Use C11-like atomics instead of plain memory accesses in x86 lock elision."
ChangeLog:
* sysdeps/unix/sysv/linux/s390/elision-lock.c
(__lll_lock_elision): Use atomics to load / store adapt_count.
* sysdeps/unix/sysv/linux/s390/elision-trylock.c
(__lll_trylock_elision): Likewise.
Some targets fail to apply dead store elimination to the
memset call in setup_ordinary_clear. Before this commit,
this causes the test case to fail. Instead, the test case
now logs lack of memset elimination as an informational
message.
Assume that Thread 1 waits to acquire a robust mutex using futexes to
block (and thus sets the FUTEX_WAITERS flag), and is unblocked when this
mutex is released. If Thread 2 concurrently acquires the lock and is
killed, Thread 1 can recover from the died owner but fail to restore the
FUTEX_WAITERS flag. This can lead to a Thread 3 that also blocked using
futexes at the same time as Thread 1 to not get woken up because
FUTEX_WAITERS is not set anymore.
The fix for this is to ensure that we continue to preserve the
FUTEX_WAITERS flag whenever we may have set it or shared it with another
thread. This is the same requirement as in the algorithm for normal
mutexes, only that the robust mutexes need additional handling for died
owners and thus preserving the FUTEX_WAITERS flag cannot be done just in
the futex slowpath code.
[BZ #20973]
* nptl/pthread_mutex_lock.c (__pthread_mutex_lock_full): Fix lost
wake-up in robust mutexes.
* nptl/pthread_mutex_timedlock.c (pthread_mutex_timedlock): Likewise.
This patch adds fmaxf and fminf benchtests. It is based on
math/s_fmax_template.c implementation which checks for basically four
different classes:
1. if x is greater or equal than y.
2. if x is less than y.
3. if x or y is signaling.
4. if y is nan.
Cases 1 and 2 are used for default input number (by mixing normal double
numbers and infinity), while case 3 and 4 are used each for on for a
benchmark class.
Checked on x86_64-linux-gnu and powerpc64-linux-gnu.
* benchtests/Makefile (bench-math): Add fminf and fmaxf.
(CFLAGS-bench-fmaxf.c): New rule.
(CFLAGS-bench-fminf.c): Likewise.
* benchtests/fmaxf-inputs: New file.
* benchtests/fminf-inputs: Likewise.
This patch adds fmax and fmin benchtests. It is based math/s_fmax_template.c
implementation which checks for basically four different classes:
1. if x is greater or equal than y.
2. if x is less than y.
3. if x or y is signaling.
4. if y is nan.
Cases 1 and 2 are used for default input number (by mixing normal double
numbers and infinity), while case 3 and 4 are used each for on for a
benchmark class.
Checked on x86_64-linux-gnu and powerpc64-linux-gnu.
* benchtests/Makefile (bench-math): Add fmin and fmax.
(CFLAGS-bench-fmax.c): New rule.
(CFLAGS-bench-fmin.c): New rule.
* benchtests/fmax-inputs: New file.
* benchtests/fmin-inputs: Likewise.
Patch disables Intel TSX on some Haswell processors to avoid TSX
on kernels that weren't updated with the latest microcode package
(which disables broken feature by default).
* sysdeps/x86/cpu-features.c (get_common_indeces): Add
stepping identification.
(init_cpu_features): Add handle of Haswell.
* assert/assert.h (assert): Rewrite assert's definition so that
a s/==/=/ typo, e.g., assert(errno = ENOENT) is not hidden from
gcc's -Wparentheses by assert-added parentheses. The new definition
uses "if (expr) /* empty */; else __assert_fail...", so
gcc -Wall will now detect that type of error in an assert, too.
The __STRICT_ANSI__ disjunct is to make this work also with both
-ansi and -pedantic, which would reject the use of ({...}).
I would have preferred to use __extension__ to mark that, but
doing so would mistakenly suppress warnings about any extension
in the user-supplied "expr".
E.g., "assert ( ({1;}) )" must continue to evoke a warning.
Benchsets in benchtests use test-skeleton, so they too need to be
linked against the new libsupport DSO.
* benchtests/Makefile (binaries-benchset): Depend on libsupport
DSO.
Without -B, python invocations may result in generation of pyc files
for modules within the source tree, which does not work well when the
source tree is read-only.
2016-12-17 Martin Galvan <martingalvan@sourceware.org>
* Rules (python-flags, python-invoke): New.
($(test-printers-out)): Use $(python-flags).
TS 18661-1 says that "Whether a signaling NaN input causes a domain
error is implementation-defined.". Considering it a domain error
would (given glibc's math_errhandling definition) mean setting errno
to EDOM. glibc consistently does not set errno for sNaN inputs
(unless it does so for qNaN as well, i.e. iseqsig), so this patch adds
documentation of the implementation-defined choice not to treat this
case as a domain error.
* manual/arith.texi (Math Error Reporting): Document that sNaN
arguments are not considered domain errors.
explicit_bzero(s, n) is the same as memset(s, 0, n), except that the
compiler is not allowed to delete a call to explicit_bzero even if the
memory pointed to by 's' is dead after the call. Right now, this effect
is achieved externally by having explicit_bzero be a function whose
semantics are unknown to the compiler, and internally, with a no-op
asm statement that clobbers memory. This does mean that small
explicit_bzero operations cannot be expanded inline as small memset
operations can, but on the other hand, small memset operations do get
deleted by the compiler. Hopefully full compiler support for
explicit_bzero will happen relatively soon.
There are two new tests: test-explicit_bzero.c verifies the
visible semantics in the same way as the existing test-bzero.c,
and tst-xbzero-opt.c verifies the not-being-optimized-out property.
The latter is conceptually based on a test written by Matthew Dempsky
for the OpenBSD regression suite.
The crypt() implementation has an immediate use for this new feature.
We avoid having to add a GLIBC_PRIVATE alias for explicit_bzero
by running all of libcrypt's calls through the fortified variant,
__explicit_bzero_chk, which is in the impl namespace anyway. Currently
I'm not aware of anything in libc proper that needs this, but the
glue is all in place if it does become necessary. The legacy DES
implementation wasn't bothering to clear its buffers, so I added that,
mostly for consistency's sake.
* string/explicit_bzero.c: New routine.
* string/test-explicit_bzero.c, string/tst-xbzero-opt.c: New tests.
* string/Makefile (routines, strop-tests, tests): Add them.
* string/test-memset.c: Add ifdeffage for testing explicit_bzero.
* string/string.h [__USE_MISC]: Declare explicit_bzero.
* debug/explicit_bzero_chk.c: New routine.
* debug/Makefile (routines): Add it.
* debug/tst-chk1.c: Test fortification of explicit_bzero.
* string/bits/string3.h: Fortify explicit_bzero.
* manual/string.texi: Document explicit_bzero.
* NEWS: Mention addition of explicit_bzero.
* crypt/crypt-entry.c (__crypt_r): Clear key-dependent intermediate
data before returning, using explicit_bzero.
* crypt/md5-crypt.c (__md5_crypt_r): Likewise.
* crypt/sha256-crypt.c (__sha256_crypt_r): Likewise.
* crypt/sha512-crypt.c (__sha512_crypt_r): Likewise.
* include/string.h: Redirect internal uses of explicit_bzero
to __explicit_bzero_chk[_internal].
* string/Versions [GLIBC_2.25]: Add explicit_bzero.
* debug/Versions [GLIBC_2.25]: Add __explicit_bzero_chk.
* sysdeps/arm/nacl/libc.abilist
* sysdeps/unix/sysv/linux/aarch64/libc.abilist
* sysdeps/unix/sysv/linux/alpha/libc.abilist
* sysdeps/unix/sysv/linux/arm/libc.abilist
* sysdeps/unix/sysv/linux/hppa/libc.abilist
* sysdeps/unix/sysv/linux/i386/libc.abilist
* sysdeps/unix/sysv/linux/ia64/libc.abilist
* sysdeps/unix/sysv/linux/m68k/coldfire/libc.abilist
* sysdeps/unix/sysv/linux/m68k/m680x0/libc.abilist
* sysdeps/unix/sysv/linux/microblaze/libc.abilist
* sysdeps/unix/sysv/linux/mips/mips32/fpu/libc.abilist
* sysdeps/unix/sysv/linux/mips/mips32/nofpu/libc.abilist
* sysdeps/unix/sysv/linux/mips/mips64/n32/libc.abilist
* sysdeps/unix/sysv/linux/mips/mips64/n64/libc.abilist
* sysdeps/unix/sysv/linux/nios2/libc.abilist
* sysdeps/unix/sysv/linux/powerpc/powerpc32/fpu/libc.abilist
* sysdeps/unix/sysv/linux/powerpc/powerpc32/nofpu/libc.abilist
* sysdeps/unix/sysv/linux/powerpc/powerpc64/libc-le.abilist
* sysdeps/unix/sysv/linux/powerpc/powerpc64/libc.abilist
* sysdeps/unix/sysv/linux/s390/s390-32/libc.abilist
* sysdeps/unix/sysv/linux/s390/s390-64/libc.abilist
* sysdeps/unix/sysv/linux/sh/libc.abilist
* sysdeps/unix/sysv/linux/sparc/sparc32/libc.abilist
* sysdeps/unix/sysv/linux/sparc/sparc64/libc.abilist
* sysdeps/unix/sysv/linux/tile/tilegx/tilegx32/libc.abilist
* sysdeps/unix/sysv/linux/tile/tilegx/tilegx64/libc.abilist
* sysdeps/unix/sysv/linux/tile/tilepro/libc.abilist
* sysdeps/unix/sysv/linux/x86_64/64/libc.abilist
* sysdeps/unix/sysv/linux/x86_64/x32/libc.abilist:
Add entries for explicit_bzero and __explicit_bzero_chk.
TS 18661-1 defines a macro FE_SNANS_ALWAYS_SIGNAL in <fenv.h>, to
indicate that the recommended practice regarding sNaNs (that
operations always produce a qNaN output with "invalid" exception, even
in the fmax / fmin / hypot / pow cases where a qNaN input would not
result in qNaN output) is followed.
Now that those functions with C99 special cases for NaNs have been
fixed not to apply those special cases to sNaN, only to qNaN, glibc
follows that recommended practice. This patch makes it define the
corresponding macro.
Since compiler optimizations may affect whether sNaNs behave as
expected and the macro relates to both language and library features,
it is only defined if __SUPPORT_SNAN__ is defined (which GCC defines
for -fsignaling-nans). It is also not defined if FE_INVALID is
undefined, since the recommended practice specifically refers to
raising the "invalid" exception, so it seems inappropriate to define
the macro for soft-float cases without support for exceptions.
(Further refinement would be possible in cases where bits/fenv.h is
shared by configurations both with and without exceptions support.)
Tested for x86_64 and x86, and also did compile-only testing for nios2
to cover the no-exceptions case.
* math/fenv.h
[__GLIBC_USE (IEC_60559_BFP_EXT) && FE_INVALID && __SUPPORT_SNAN__]
(FE_SNANS_ALWAYS_SIGNAL): New macro.
* math/test-fe-snans-always-signal.c: New file.
* math/Makefile (tests): Add test-fe-snans-always-signal.
(CFLAGS-test-fe-snans-always-signal.c): New variable.
* manual/arith.texi (Infinity and NaN): Document
FE_SNANS_ALWAYS_SIGNAL.
When build-many-glibcs.py re-execs itself with execv, any buffered
output on stdout may be lost (in particular, messages intended to go
to a bot's log about the re-exec taking place). This patch makes it
flush stdout before execv, similar to the flush before running a
subprocess from the bot that is done to ensure output appears in the
right order.
* scripts/build-many-glibcs.py (Context.exec_self): Flush stdout
before calling execv.
Current optimized powercp64/power7 memchr uses a strategy to check for
p versus align(p+n) (where 'p' is the input char pointer and n the
maximum size to check for the byte) without taking care for possible
overflow on the pointer addition in case of large 'n'.
It was triggered by 3038145ca2 where default rawmemchr (used to
created ppc64 rawmemchr in ifunc selection) now uses memchr (p, c, (size_t)-1)
on its implementation.
This patch fixes it by implement a satured addition where overflows
sets the maximum pointer size to UINTPTR_MAX.
Checked on powerpc64le-linux-gnu.
[BZ# 20971]
* sysdeps/powerpc/powerpc64/power7/memchr.S (__memchr): Avoid
overflow in pointer addition.
* string/test-memchr.c (do_test): Add an argument to pass as
the size on memchr.
(test_main): Add check for SIZE_MAX.
This patch converts the wrapper scalbln (which set errno directly
rather than doing anything with __kernel_standard) to use the
type-generic template machinery, in the same way that has been done
for ldexp.
Tested for powerpc64le, s390, and x86_64.
Various fmax and fmin function implementations mishandle sNaN
arguments:
(a) When both arguments are NaNs, the return value should be a qNaN,
but sometimes it is an sNaN if at least one argument is an sNaN.
(b) Under TS 18661-1 semantics, if either argument is an sNaN then the
result should be a qNaN (whereas if one argument is a qNaN and the
other is not a NaN, the result should be the non-NaN argument).
Various implementations treat sNaNs like qNaNs here.
This patch fixes the x86 and x86_64 versions (ignoring float and
double for 32-bit x86 given the inability to reliably avoid the sNaN
turning into a qNaN before it gets to the called function). Tests of
sNaN inputs to these functions are added.
Note on architecture versions I haven't changed for this issue:
AArch64 already gets this right (it uses a hardware instruction with
the correct semantics for both quiet and signaling NaNs) and does not
need changes. It's possible Alpha, IA64, SPARC might need changes
(this would be shown by the testsuite if so).
Tested for x86_64 and x86 (both i686 and i586 builds, to cover the
different x86 implementations).
[BZ #20947]
* sysdeps/i386/fpu/s_fmaxl.S (__fmaxl): Add the arguments when
either is a signaling NaN.
* sysdeps/i386/fpu/s_fminl.S (__fminl): Likewise. Make code
follow fmaxl more closely.
* sysdeps/i386/i686/fpu/s_fmaxl.S (__fmaxl): Add the arguments
when either is a signaling NaN.
* sysdeps/i386/i686/fpu/s_fminl.S (__fminl): Likewise.
* sysdeps/x86_64/fpu/s_fmax.S (__fmax): Likewise.
* sysdeps/x86_64/fpu/s_fmaxf.S (__fmaxf): Likewise.
* sysdeps/x86_64/fpu/s_fmaxl.S (__fmaxl): Likewise.
* sysdeps/x86_64/fpu/s_fmin.S (__fmin): Likewise.
* sysdeps/x86_64/fpu/s_fminf.S (__fminf): Likewise.
* sysdeps/x86_64/fpu/s_fminl.S (__fminl): Likewise.
* math/libm-test.inc (fmax_test_data): Add tests of sNaN inputs.
(fmin_test_data): Likewise.
Various fmax and fmin function implementations mishandle sNaN
arguments:
(a) When both arguments are NaNs, the return value should be a qNaN,
but sometimes it is an sNaN if at least one argument is an sNaN.
(b) Under TS 18661-1 semantics, if either argument is an sNaN then the
result should be a qNaN (whereas if one argument is a qNaN and the
other is not a NaN, the result should be the non-NaN argument).
Various implementations treat sNaNs like qNaNs here.
This patch fixes the powerpc versions of these functions (shared by
float and double, 32-bit and 64-bit). The structure of those versions
is that all ordered cases are already handled before anything dealing
with the case where the arguments are unordered; thus, this patch
causes no change to the code executed in the common case (neither
argument a NaN).
Tested for powerpc (32-bit and 64-bit), together with tests to be
added along with the x86_64 / x86 fixes.
[BZ #20947]
* sysdeps/powerpc/fpu/s_fmax.S (__fmax): Add the arguments when
either is a signaling NaN.
* sysdeps/powerpc/fpu/s_fmin.S (__fmin): Likewise.
Various fmax and fmin function implementations mishandle sNaN
arguments:
(a) When both arguments are NaNs, the return value should be a qNaN,
but sometimes it is an sNaN if at least one argument is an sNaN.
(b) Under TS 18661-1 semantics, if either argument is an sNaN then the
result should be a qNaN (whereas if one argument is a qNaN and the
other is not a NaN, the result should be the non-NaN argument).
Various implementations treat sNaNs like qNaNs here.
This patch fixes the generic implementations used in the absence of
architecture-specific versions.
Tested for mips64 and powerpc (together with testcases that I'll add
along with the x86_64 / x86 fixes).
[BZ #20947]
* math/s_fmax_template.c (M_DECL_FUNC (__fmax)): Add the arguments
when either is a signaling NaN.
* math/s_fmin_template.c (M_DECL_FUNC (__fmin)): Likewise.
Information about whether the ABI of long double is the same as that
of double is split between bits/mathdef.h and bits/wordsize.h.
When the ABIs are the same, bits/mathdef.h defines
__NO_LONG_DOUBLE_MATH. In addition, in the case where the same glibc
binary supports both -mlong-double-64 and -mlong-double-128,
bits/wordsize.h defines __LONG_DOUBLE_MATH_OPTIONAL, along with
__NO_LONG_DOUBLE_MATH if this particular compilation is with
-mlong-double-64.
As part of the refactoring I proposed in
<https://sourceware.org/ml/libc-alpha/2016-11/msg00745.html>, this
patch puts all that information in a single header,
bits/long-double.h. It is included from sys/cdefs.h alongside the
include of bits/wordsize.h, so other headers generally do not need to
include bits/long-double.h directly.
Previously, various bits/mathdef.h headers and bits/wordsize.h headers
had this long double information (including implicitly in some
bits/mathdef.h headers through not having the defines present in the
default version). After the patch, it's all in six bits/long-double.h
headers. Furthermore, most of those new headers are not
architecture-specific. Architectures with optional long double all
use the ldbl-opt sysdeps directory, either in the order (ldbl-64-128,
ldbl-opt, ldbl-128) or (ldbl-128ibm, ldbl-opt). Thus a generic header
for the case where long double = double, and headers in ldbl-128,
ldbl-96 and ldbl-opt, suffices to cover every architecture except for
cases where long double properties vary between different ABIs sharing
a set of installed headers; fortunately all the ldbl-opt cases share a
single compiler-predefined macro __LONG_DOUBLE_128__ that can be used
to tell whether this compilation is -mlong-double-64 or
-mlong-double-128.
The two cases where a set of headers is shared between ABIs with
different long double properties, MIPS (o32 has long double = double,
other ABIs use ldbl-128) and SPARC (32-bit has optional long double,
64-bit has required long double), need their own bits/long-double.h
headers.
As with bits/wordsize.h, multiple-include protection for this header
is generally implicit through the include guards on sys/cdefs.h, and
multiple inclusion is harmless in any case. There is one subtlety:
the header must not define __LONG_DOUBLE_MATH_OPTIONAL if
__NO_LONG_DOUBLE_MATH was defined before its inclusion, because doing
so breaks how sysdeps/ieee754/ldbl-opt/nldbl-compat.h defines
__NO_LONG_DOUBLE_MATH itself before including system headers. Subject
to keeping that working, it would be reasonable to move these macros
from defined/undefined #ifdef to always-defined 1/0 #if semantics, but
this patch does not attempt to do so, just rearranges where the macros
are defined.
After this patch, the only use of bits/mathdef.h is the alpha one for
modifying complex function ABIs for old GCC. Thus, all versions of
the header other than the default and alpha versions are removed, as
is the include from math.h.
Tested for x86_64 and x86. Also did compilation-only testing with
build-many-glibcs.py.
* bits/long-double.h: New file.
* sysdeps/ieee754/ldbl-128/bits/long-double.h: Likewise.
* sysdeps/ieee754/ldbl-96/bits/long-double.h: Likewise.
* sysdeps/ieee754/ldbl-opt/bits/long-double.h: Likewise.
* sysdeps/mips/bits/long-double.h: Likewise.
* sysdeps/unix/sysv/linux/sparc/bits/long-double.h: Likewise.
* math/Makefile (headers): Add bits/long-double.h.
* misc/sys/cdefs.h: Include <bits/long-double.h>.
* stdlib/strtold.c: Include <bits/long-double.h> instead of
<bits/wordsize.h>.
* bits/mathdef.h [!_COMPLEX_H]: Do not allow inclusion.
[!__NO_LONG_DOUBLE_MATH]: Remove conditional code.
* math/math.h: Do not include <bits/mathdef.h>.
* sysdeps/aarch64/bits/mathdef.h: Remove file.
* sysdeps/alpha/bits/mathdef.h [!_COMPLEX_H]: Do not allow
inclusion.
* sysdeps/ia64/bits/mathdef.h: Remove file.
* sysdeps/m68k/m680x0/bits/mathdef.h: Likewise.
* sysdeps/mips/bits/mathdef.h: Likewise.
* sysdeps/powerpc/bits/mathdef.h: Likewise.
* sysdeps/s390/bits/mathdef.h: Likewise.
* sysdeps/sparc/bits/mathdef.h: Likewise.
* sysdeps/x86/bits/mathdef.h: Likewise.
* sysdeps/s390/s390-32/bits/wordsize.h
[!__NO_LONG_DOUBLE_MATH && !__LONG_DOUBLE_MATH_OPTIONAL]: Remove
conditional code.
* sysdeps/s390/s390-64/bits/wordsize.h
[!__NO_LONG_DOUBLE_MATH && !__LONG_DOUBLE_MATH_OPTIONAL]:
Likewise.
* sysdeps/unix/sysv/linux/alpha/bits/wordsize.h
[!__NO_LONG_DOUBLE_MATH && !__LONG_DOUBLE_MATH_OPTIONAL]:
Likewise.
* sysdeps/unix/sysv/linux/powerpc/bits/wordsize.h
[!__NO_LONG_DOUBLE_MATH && !__LONG_DOUBLE_MATH_OPTIONAL]:
Likewise.
* sysdeps/unix/sysv/linux/sparc/bits/wordsize.h
[!__NO_LONG_DOUBLE_MATH && !__LONG_DOUBLE_MATH_OPTIONAL]:
Likewise.
This patch makes bits/fcntl-linux.h include <linux/falloc.h> to define
the FALLOC_* flags under __USE_GNU (linux/falloc.h defines only those
bits, nothing else).
Tested for x86_64 and x86.
* sysdeps/unix/sysv/linux/bits/fcntl-linux.h [__USE_GNU]: Include
<linux/falloc.h>.
(FALLOC_FL_KEEP_SIZE): Remove.
(FALLOC_FL_PUNCH_HOLE): Likewise.
(FALLOC_FL_COLLAPSE_RANGE): Likewise.
(FALLOC_FL_ZERO_RANGE): Likewise.
The macro ENTRY in tst-strfrom.h is used to generate the input values for
each floating-point type (float, double, long double). It should append
the parameter LSUF (Literal suffix) to the floating-point number, but is
using CSUF (C function suffix). This patch fixes it.
Tested for powerpc64le and x86_64.
This patch consolidates the Linux renameat implementation on
sysdeps/unix/sysv/linux/renameat.c. The renameat syscall was
deprecated at b0da6d44 for newer architectures, so using the
auto-generation list may generate wrappers that returns ENOSYS.
Current code try to use __NR_renameat and if it is not define
it uses __NR_renameat2.
Checked on x86_64 and aarch64.
* sysdeps/unix/sysv/linux/renameat.c: New file.
* sysdeps/unix/sysv/linux/syscalls.list: Remove renameat.
This patch consolidates the Linux rename implementation on
sysdeps/unix/sysv/linux/rename.c. Current code try to use
__NR_rename if is defined and apply the same strategy for
__NR_renameat and __NR_renameat2.
Check on x86_64 and aarch64.
* sysdeps/unix/sysv/linux/rename.c: New file.
* sysdeps/unix/sysv/linux/generic/rename.c: Remove file.
calls strcspn, call strcspn directly so we get the end of the token without
an extra call to rawmemchr. Also avoid an unnecessary call to strcspn after
the last token by adding an early exit for an empty string. Change strtok
to tailcall strtok_r to avoid unnecessary code duplication.
Remove the special header optimization for strtok_r of a 1-character
constant string - both strspn and strcspn contain optimizations for this
case. Benchmarking this showed similar performance in the worst case,
but up to 5.5x better performance in the "found" case for large inputs.
* benchtests/bench-strtok.c (oldstrtok): Add old implementation.
* string/strtok.c (strtok): Change to tailcall __strtok_r.
* string/strtok_r.c (__strtok_r): Optimize for performance.
* string/string-inlines.c (__old_strtok_r_1c): New function.
* string/bits/string2.h (__strtok_r): Move to string-inlines.c.
This patch converts the wrapper log1p (which set errno directly rather
than doing anything with __kernel_standard) to use the type-generic
template machinery, in the same way that has been done for ilogb.
Tested for powerpc64le, s390, and x86_64.
assembler version by tailcalling memchr with the maximum size.
If a target has an optimized memchr this is significantly faster,
if not, then this makes little difference.
Also optimize the special case of zero to use strlen as this is
typically faster than memchr.
* string/rawmemchr.c (RAWMEMCHR): Use faster memchr/strlen.
This patch updates build-many-glibcs.py to use Linux 4.9 for kernel
headers unless another version is explicitly specified. Note that
when a version changes like this you'll need to use --replace-sources
when updating an existing checkout to tell build-many-glibcs.py it's
OK to delete and replace the sources of a component for which the
version used has changed.
* scripts/build-many-glibcs.py (Context.checkout): Default Linux
kernel version to 4.9.
