Sebastian Huber [Thu, 22 Jun 2017 07:22:47 +0000 (09:22 +0200)]
Introduce _REENT_GLOBAL_STDIO_STREAMS
In Newlib, the stdio streams are defined to thread-specific pointers
_reent::_stdin, _reent::_stdout and _reent::_stderr. In case
_REENT_SMALL is not defined, then these pointers are initialized via
_REENT_INIT_PTR() or _REENT_INIT_PTR_ZEROED() to thread-specific FILE
objects provided via _reent::__sf[3]. There are two problems with this
(at least in case of RTEMS).
(1) The thread-specific FILE objects are closed by _reclaim_reent().
This leads to problems with language run-time libraries that provide
wrappers to the C/POSIX stdio streams (e.g. C++ and Ada), since they
use the thread-specific FILE objects of the initialization thread. In
case the initialization thread is deleted, then they use freed memory.
(2) Since thread-specific FILE objects are used with a common output
device via file descriptors 0, 1 and 2, the locking at FILE object level
cannot ensure atomicity of the output, e.g. a call to printf().
Introduce a new Newlib configuration option _REENT_GLOBAL_STDIO_STREAMS
to enable the use of global stdio FILE objects.
As a side-effect this reduces the size of struct _reent by more than
50%.
The _REENT_GLOBAL_STDIO_STREAMS should not be used without
_STDIO_CLOSE_PER_REENT_STD_STREAMS.
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Wilco Dijkstra [Thu, 29 Jun 2017 14:32:09 +0000 (14:32 +0000)]
Optimized memcmp
This is an optimized memcmp for AArch64. This is a complete rewrite
using a different algorithm. The previous version split into cases
where both inputs were aligned, the inputs were mutually aligned and
unaligned using a byte loop. The new version combines all these cases,
while small inputs of less than 8 bytes are handled separately.
This allows the main code to be sped up using unaligned loads since
there are now at least 8 bytes to be compared. After the first 8 bytes,
align the first input. This ensures each iteration does at most one
unaligned access and mutually aligned inputs behave as aligned.
After the main loop, process the last 8 bytes using unaligned accesses.
This improves performance of (mutually) aligned cases by 25% and
unaligned by >500% (yes >6 times faster) on large inputs.
Sebastian Pop [Fri, 23 Jun 2017 20:23:09 +0000 (15:23 -0500)]
aarch64: optimize the unaligned case of memcmp
This brings to newlib a performance improvement that we developed in Bionic
libc. That change has been submitted for review to Bionic libc:
https://android-review.googlesource.com/418279
A similar patch has been submitted for review in glibc:
https://sourceware.org/ml/libc-alpha/2017-06/msg01143.html
Patch written by Vikas Sinha and Sebastian Pop.
The performance was measured on the bionic-benchmarks on a hikey (aarch64 8xA53)
board. There was no performance change to the existing benchmark
and a performance improvement on the new benchmark for memcmp
on the unaligned side. The new benchmark has been submitted for
review at https://android-review.googlesource.com/414860
The overall performance improves by 18% for the small data set 8
and the performance improves by 450% for the large data set 64k.
The base is with the libc from /system/lib64. The bionic libc
with this patch is in /data.
Fujii Hironori [Tue, 20 Jun 2017 11:17:09 +0000 (13:17 +0200)]
cygwin: regtool: encode error messages correctly
Error messages of regtool can't be read, which are encoded in,
for instance, SHIFT_JIS in Japanese Windows. Fix by using
wide chars instead of multibyte.
Commit 8a3b3bb4d7224d419cc1a4af60ccf7e70edc876b changed the guard on
some functions from _POSIX_THREADS to __POSIX_VISIBLE. As a consequence,
some use of siginfo_t and pthread_t became visible under configurations
where _POSIX_THREADS is unset but __POSIX_VISIBLE is. Build then fails
because the definition of those types are still unavailable.
This commit make those type definition visible for __POSIX_VISIBLE
configurations. This requires moving the siginfo_t definition out of the
RTEMS specific definitions in sys/signal.h while still guarding it
against cygwin case.
Erik M. Bray [Thu, 15 Jun 2017 13:30:08 +0000 (15:30 +0200)]
Ensure that send() interrupted by a signal returns sucessfully
When SA_RESTART is not set on a socket, a blocking send() that is
interrupted mid-transition by a signal should return success (and
report just how many bytes were actually transmitted).
The err variable used here was not always guaranteed to be set
correctly in the loop, so better to just remove it and call
WSAGetLastError() explicitly.
Yaakov Selkowitz [Wed, 14 Jun 2017 15:20:05 +0000 (10:20 -0500)]
Export XSI sigpause
There are two common sigpause variants, both of which take an int argument.
If you request _XOPEN_SOURCE or _GNU_SOURCE, you get the System V version,
which removes the given signal from the process's signal mask; otherwise
you get the BSD version, which sets the process's signal mask to the given
value.
Yaakov Selkowitz [Tue, 13 Jun 2017 19:51:18 +0000 (14:51 -0500)]
stdio.h: guard function macros with !__cplusplus
While POSIX allows these functions to also be defined as macros in C, in
C++ this is not allowed, and prevents these names (particularly feof) from
being used in a custom namespace.
Corinna Vinschen [Wed, 14 Jun 2017 11:22:56 +0000 (13:22 +0200)]
cygwin: readdir: don't lookup mount target inodes
So far Cygwin's readdir returned the inode number of a mount target
in d_ino, rather than the actual inode number of the mount point in
the underlying filesystem. This not only results in a performance
hit if the mount target is a remote FS, it is also not done on other
POSIX systems.
Remove the code evaluating the mount target inode number.
Joe Lowe [Tue, 13 Jun 2017 18:12:50 +0000 (11:12 -0700)]
readdir() with mount point dentry, return mount point INO
This patch fixes a minor compatibility issue w/ cygwin mount point handling in
readdir(), compared to equivalent behavior of Linux and MacOS. dentry.d_ino
should indicate the INO of the mount point itself, not the target volume root
folder.
Changed return type from readdir_check_reparse_point to uint8_t, to avoid
unnecessarily being implicitly cast to and from a signed int.
Renamed a related local variable "attr" to "oattr" that was eclipsing a member
variable with the same name.
Sebastian Huber [Mon, 12 Jun 2017 06:38:56 +0000 (08:38 +0200)]
Remove FreeBSD specifics from RTEMS <arpa/inet.h>
For whatever reason FreeBSD renames several functions provided by
<arpa/inet.h> and uses weak references to provide the standard function
names. This causes problems on targets lacking proper support for weak
references. We do not need this function renaming on RTEMS.
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Silviu Baranga [Mon, 5 Jun 2017 08:54:42 +0000 (09:54 +0100)]
Don't overread or write memory returned by _DTOA_R
Don't over-read memory returned by _DTOA_R, and never write to it
since the result might be a string literal.
For example, when doing:
swprintf(tt, 20, L"%.*f", 6, 0.0);
we will get back "0".
Instead, write the result returned by _DTOA_R to the output buffer.
After this, write the 0 chars directly to the the output buffer
(if there are any). This also has the (marginal) advantage that
we read/write less memory overall.
cygwin: Fix lrint{f,l} to return a 64 bit long on x86_64
Mingw-w64 (where the code has been taken from) has 4 byte longs
independently of the architecture but x86_64 Cygwin has 64 bit longs.
So use fistpll instead of fistpl on x86_64 Cygwin.
Workaround a bug (or undocumented behaviour) in LCMapStringW:
It's documented(*) that the cchDest parameter is a byte count with
LCMAP_SORTKEY, but a character count otherwise. But the docs don't
state what happens if you combine LCMAP_SORTKEY with LCMAP_BYTEREV.
Tests indicate that LCMAP_SORTKEY treats cchDest as byte count, but
then LCMAP_BYTEREV treats it as char count in the same call. So the
latter swaps twice as much bytes in the destination buffer than the
byte count it returns, which potentially results in writing past the
end of the given output buffer.
Solution: Don't specify LCMAP_BYTEREV in the LCMapStringW(LCMAP_SORTKEY)
call, rather byte swap afterwards.
Sebastian Huber [Mon, 22 May 2017 07:34:31 +0000 (09:34 +0200)]
Move ARM access.c from machine to sys
The implementation of the POSIX access() function is nothing machine
specific like memcpy(), etc. Move it back to the system domain. This
avoids problems due to the include search order of the Newlib/GCC build
which picks up machine includes before system includes.
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
When simulating arm code, the target program startup code (crt0) uses
semihosting invocations to get the command line from the simulator. The
simulator returns the command line and its size into the area passed in
parameter. (ARM 32-bit specifications :
http://infocenter.arm.com/help/topic/com.arm.doc.dui0058d/DUI0058.pdf
chapter "5.4.19 SYS_GET_CMDLINE").
The memory area pointed by the semihosting register argument is located
in .text section (usually not writtable (RX)).
If we run this code on a simulator that respects this rights properties
(qemu user-mode for instance), the command line will not be written to
the .text program memory, in particular the length of the string. The
program runs with an empty command line. This problem hasn't been seen
earlier probably because qemu user-mode is not so much used, but this can
happen with another simulator that refuse to write in a read-only segment.
With this modification, the command line can be correctly passed to the
target program.
Changes:
- libgloss/arm/crt0.S : Arguments passed to the AngelSWI_Reason_GetCmdLine
semihosting invocation are placed into .data section instead of .text
- libgloss/aarch64/crt0.S : Idem for aarch64 AngelSVC_Reason_GetCmdLine
semihosting.
cygwin: pinfo: do not wait for setting ppid on a transitional procinfo
This leads to excessive lag when stracing processes if the inferior
process checks the process table. The reason is that ppid isn't set
in the procinfo memory of the dynamically loading strace itself.
cygwin: try unprivileged symlink creation on W10 1703 and later
Add new SYMBOLIC_LINK_FLAG_ALLOW_UNPRIVILEGED_CREATE flag to
CreateSymbolicLinkW call when running on W10 1703 or later.
Don't do that on older versions to avoid ERROR_INVALID_PARAMETER.
Preliminary, needs testing. There's an off-chance that the
flag results in the same ERROR_INVALID_PARAMETER on 1703 if the
developer settings are not enabled.
RtlGetNtVersionNumbers returns the build number with some upper bits
set for no apparent reason. The fact that RtlGetNtVersionNumbers is
undocumented doesn't exactly help.
Just filter out the upper WORD for now. If build numbers are in
danger to become 6 digit numbers, re-evaluate.
Christian Franke [Sat, 22 Apr 2017 12:50:58 +0000 (14:50 +0200)]
Fix stat.st_blocks for files compressed with CompactOS method
Always retrieve FileCompressionInformation for non-empty
files if FileStandardInformation returns 0 allocated blocks.
This fixes stat.st_blocks for files compressed with CompactOS method.
Signed-off-by: Christian Franke <franke@computer.org>