vfork: Use clone if arch does not have the vfork syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

utime: Use utimensat if arch does not have the utime syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

inotify_init: Use inotify_init1 if inotify_init syscall is not defined

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

epoll: Use epoll_pwait if arch does not have the epoll_wait syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

epoll: Use epoll_create1 if arch does not have the epoll_create syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

pipe: Use pipe2 if arch does not have the pipe syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

utimes: Use utimensat if arch does not have the utimes syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

symlink: Use symlinkat if arch does not have the symlink syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

readlink: Use readlinkat if arch does not have the readlink syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

rename: Use renameat if arch does not have the rename syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

mkdir: Use mkdirat if arch does not have the mkdir syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

lchown: Use fchownat if arch does not have the lchown syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

rmdir: Use unlinkat if arch does not have the rmdir syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

chown: Use fchownat if arch does not have the chown syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

mknod: Use mknodat if arch does not have the mknod syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

unlink: Use unlinkat if arch does not have the unlink syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

link: Use linkat if arch does not have the link syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

access: Use faccessat if arch does not have the access syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

chmod: Use fchmodat if arch does not have the chmod syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

dup2: Use dup3 if arch does not have the dup2 syscall

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

Makefile.in: Use non-legacy interfaces for new architectures

New architectures, which don't define ARCH_HAS_DEPRECATED syscalls, need
to use the new interfaces from libc/sysdeps/linux/generic-syscalls/bits/

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

Config.in: Introduce symbol for arches with deprecated syscalls

Linuxthreads(old and new) need deprecated syscalls to build. Existing
architectures support these system calls but new architectures don't.
This symbol has no functional change apart from hidding the Linuxthreads
symbols from arches that don't support them.

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc/sysdeps: Add new interfaces for new architectures

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc: add non standard execvpe function

Signed-off-by: Henning Heinold <heinold@inf.fu-berlin.de>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc/{arm, i386}/Makefile.arch: Don't build 64-bit code if LFS is disabled

Fixes the following compilation error:

_lfs_64.h: # error Do not include this header in files not built when LFS is disabled

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

stdlib: Fix fndecl of mkostemp, mkostemp64

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

test: run test{,_glibc} adjacent

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc: add posix_fallocate()

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

buildsys: switch warning into a TODO

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

buildsys: Fix kbuild-style switch fallout

regressions introduced in 1b3025b7352f5e432ffa1c7adc57085ac9092b77
"buildsys: switch libc to kbuild-style"

unistd: Handle !UCLIBC_HAS_GET{, SUB}OPT cases

libc/libc_so.a(getsubopt-susv3.os): In function getsubopt':
getsubopt-susv3.c:(.text+0x0): multiple definition of getsubopt'
libc/libc_so.a(getsubopt.os):getsubopt.c:(.text+0x0):
first defined here.

The solution is to filter out the correct files when
UCLIBC_HAS_GETOPT and UCLIBC_HAS_GETSUBOPT are not defined.

inet: Move INET_RPC_{DIR, OUT} before CSRC_ALL

libc/libc_so.a(rpc_thread.oS): In function __rpc_thread_svc_max_pollfd':
rpc_thread.c:(.text+0x148): undefined reference to svc_max_pollfd'
libc/libc_so.a(rpc_thread.oS): In function __rpc_thread_svc_pollfd':
rpc_thread.c:(.text+0x184): undefined reference to svc_pollfd'

The CSRC_ALL variable uses the INET_RPC_DIR variable which is
initialized later on. We fix this problem by moving the
INET_RCP_{DIR,OUT} definitions before we use them in CSRC_ALL

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

mman: rename MAP_UNINITIALIZE to MAP_UNINITIALIZED

The name was changed to include a trailing 'D' when it went into the
kernel.

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

buildsys: switch libc to kbuild-style

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libpthread: Fix handling of SUBARCH includes

TARGET_SUBARCH can only ever be set if TARGET_ARCH is set.
Only include SUBARCH includes if SUBARCH is non-empty.

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc: atexit: reuse free slots at the end of exit functions table

Continuosly dlopen and dlclose of shared object will cause a memory leak
in atexit function. This fix reuse free slots at the end of the list.

For further detail see https://bugs.busybox.net/show_bug.cgi?id=2455

Signed-off-by: Ronald Wahl <ronald.wahl@raritan.com>
Tested-by: Filippo Arcidiacono <filippo.arcidiacono@st.com>
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>

libc/inet: pull in stddef.h for NULL

Fixes a build error:
In file included from libc/inet/recv.c:8:0:
libc/inet/socketcalls.c: In function '__recv_nocancel':
libc/inet/socketcalls.c:203:57: error: 'NULL' undeclared (first use in this function)

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

libc/sysdeps: sync bits/in.h with glibc

URL: https://bugs.busybox.net/show_bug.cgi?id=5888
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

buildsys: gen_bits_syscall_h: do not leave undefined SYS_xxx around

If we end up doing '#undef __NR_xxx', we don't want to leave the
corresponding SYS_xxx symbol defined.  So undef it too.

For example, with the ARM EABI layer, we have a bunch of legacy
syscalls that we define early on and then later undefine (such
as __NR_utime).  But we left SYS_utime defined so code that tests
for that define before using it would be broken (since it'd be
defined to a non-existent symbol).

URL: https://bugs.gentoo.org/425006
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

buildsys: gen_bits_syscall_h: clean up script

Just style changes here to make it more readable.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

buildsys: gen_bits_syscall_h: remove top_builddir references

This script hasn't used this var in a while (it writes to stdout), so
drop all mention/usage of it.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

buildsys: allow the toolchain to search its dirs for kernel headers

Setting KERNEL_HEADERS up to the exact path all the time is a pita,
especially when the toolchain is often times already configured to
do the right thing.  So if the user has set that to "", don't force
any specific paths.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

buildsys: clean up path scrubbing a bit

This just makes it easier to read -- no functional changes.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

libc/sysdeps: dup3: fix typo in comment

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

libc: fix typo in statvfs

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc: fix typo in statvfs

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

mount.h: update

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

statfs: towards _STATFS_F_FLAGS

preparatory work.

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

statfs: support f_frsize

closes bugzilla #5834

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

socket.h: pull socket_type.h from eglibc

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

config: improve help of UCLIBC_HAS_RESOLVER_SUPPORT

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

buildsys: link libgcc_eh if DODEBUG

with -O0 we (e.g. lockf) might end up with references to
_Unwind_Resume, so pull in gcc_eh in this case..

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

getpwnam: hide relocation

The non-reentrant version of getpwnam is used in the RPC code (for
!HAS_REENTRANT_RPC)

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

LT.old: reinstate build

As noted by Florian Fainelli:
  LD libpthread-0.9.34-git.so
libpthread/linuxthreads.old/libpthread_so.a(wrapsyscall.oS): In function `fcntl64':
wrapsyscall.c:(.text+0xd8): undefined reference to `__libc_fcntl64'
libpthread/linuxthreads.old/libpthread_so.a(wrapsyscall.oS): In function `lseek64':
wrapsyscall.c:(.text+0x190): undefined reference to `__libc_lseek64'
libpthread/linuxthreads.old/libpthread_so.a(wrapsyscall.oS): In function `pread64':
wrapsyscall.c:(.text+0x39c): undefined reference to `__libc_pread64'

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

buildsys: use kbuild style, fix omission

forgot this hunk in d0234b16d61f06b2a3fd6241eff8c81250283a60

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

buildsys: use kbuild style

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

buildsys: remove superfluous default n

kconfig's default defaults to n so spelling out default n is redundant

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc: add {get,set,swap,make}context user context manipulation functions

Add the obsolescent SUSv3 family of user context manipulating functions
for arm, i386, mips, x86_64.

Signed-off-by: Timon ter Braak <timonterbraak@gmail.com>
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
Signed-off-by: Natanael Copa <ncopa@alpinelinux.org>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

dl: fix dlsym lookups with RTLD_NEXT

The current code for dlsym() when invoked with RTLD_NEXT lookup
searches for the module where it's being called from, and executes the
_dl_find_hash only for the next module in the chain. However, if the
looked symbol is not there, the rest of the modules are not checked.

Generally this is not a problem as symbols are merged for the parent
modules; so this affects only RTLD_NEXT.

This patch adds a loop iterating through all the following modules.

Signed-off-by: Timo Teräs <timo.teras@iki.fi>
Reviewed-by: Filippo ARCIDIACONO <filippo.arcidiacono@st.com>
Tested-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

buildsys: prevent make from searching for .config

Fixes:
make distclean
mkdir /.config
make -C test/API clean
Reading makefile `../../.config' (search path) (don't care) (no ~ expansion)...
make: *** ../../.config: Is a directory.  Stop.

write(1, "Reading makefile `../../.config'"..., 80) = 80
open("../../.config", O_RDONLY)         = -1 ENOENT (No such file or directory)
open("/usr/include/../../.config", O_RDONLY) = 6
fstat(6, {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
read(6, 0x7f2fe3bfc000, 4096)           = -1 EISDIR (Is a directory)

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

arm: workaround GCC PR target/53735 with fork()

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc: handle sync_file_range

Add cancellation and support arm, ppc64, mips32.
Compile-tested.

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

arm: clone: restore stack pointer just after return from syscall

If the syscall returns with an error the stack pointer and r4 register
are not restored because the instruction 'ldmnefd sp!, {r4}' is executed
after branching to '__error' label.
This bug has been spotted out by running './utstest clone 5' from LTP
built with -fstack-protector-all compiler flag as log below:

root@cortex-a9:/usr/tests/ltp/testcases/bin# ./utstest clone 5
stack smashing detected: ./utstest terminated()

Regression introduced by commit e58798e107d652644629a1daaa95d76430808d53

Signed-off-by: Filippo Arcidiacono <filippo.arcidiacono@st.com>
Signed-off-by: Giuseppe Di Giore <giuseppe.di-giore@st.com>
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>

xtensa: fix setjmp that didn't save all registers correctly

Setjmp was only saving the lower order of registers (a0-a3) correctly, but
not the higher ones (a4-a8/a12).
The change also includes additional information, and renames many of the
registers, so that setjmp and longjmp look more like the inverse of
each other.

Signed-off-by: Chris Zankel <chris@zankel.net>

sched.h: include stdlib.h for malloc/free

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

update ptrace.h to latest from glibc

Update libc/sysdeps/linux/common/sys/ptrace.h to latest from glibc's
sysdeps/unix/sysv/linux/sys/ptrace.h.

This adds definitions for operations:
 - PTRACE_GETREGSET
 - PTRACE_SETREGSET
 - PTRACE_SEIZE
 - PTRACE_INTERRUPT
 - PTRACE_LISTEN

And adds flags:
 - PTRACE_SEIZE_DEVEL

And adds event codes:
 - PTRACE_EVENT_SECCOMP

This is to allow access to the generic interface for accessing
architecture specific regsets using the corresponding NT_* types,
required for new Linux kernel architecture ports.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

libc/stdlib: localize __unix_grantpt

When ASSUME_DEVPTS is disabled, we end up exporting __unix_grantpt and
get a reloc to it from grantpt.  Mark it static to fix all of that.

Reported-by: Michael Deutschmann <michael@talamasca.ocis.net>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

drop support for pre ISO-C compilers

This drops __signed, __volatile, and __const.  Only the latter was
used in the code base, and for uClibc, not consistently.  Much of
the code used plain "const" which meant "__const" was useless.

Really, the point of this is to stay in sync with what glibc did.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

Replace FSF snail mail address with URLs

This matches a similar change made to glibc.

No functional changes here.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

AVR32: declare missing result variable

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

nptl: fix gcc noreturn warning

Current builds show the warning:
  CC libpthread/nptl/pthread_create.os
libpthread/nptl/pthread_create.c: In function 'start_thread':
libpthread/nptl/pthread_create.c:418:3: warning: function declared
'noreturn' has a 'return' statement [enabled by default]
libpthread/nptl/pthread_create.c:418:3: warning: 'noreturn' function
does return [enabled by default]

Looking at glibc, it does not mark this function as noreturn.  Testing
code size, gcc is smart enough to detect the code never actually returns
regardless of the attribute.  So drop it to keep in sync with glibc and
to avoid the warnings.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

nptl: fix gcc noreturn warning

Current builds show the warning:
  CC libpthread/nptl/sysdeps/unix/sysv/linux/mq_notify.os
libpthread/nptl/sysdeps/unix/sysv/linux/mq_notify.c: In function 'helper_thread':
libpthread/nptl/sysdeps/unix/sysv/linux/mq_notify.c:142:3: warning:
function declared 'noreturn' has a 'return' statement [enabled by default]

Looking at glibc, it does not mark this function as noreturn.  Testing
code size, gcc is smart enough to detect the while(1){} never actually
returns regardless of the attribute.  So drop it to keep in sync with
glibc and to avoid the warning.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ldso: include dlfcn.h for RTLD_NODELETE

Building with NPTL enabled and shared library support disabled we hit:
In file included from libpthread/nptl/sysdeps/generic/dl-tls.c:30:0:
./ldso/include/dl-elf.h: In function '__dl_parse_dynamic_info':
./ldso/include/dl-elf.h:173:20: error: 'RTLD_NODELETE' undeclared (first use in this function)
./ldso/include/dl-elf.h:173:20: note: each undeclared identifier is reported only once for each function it appears in
make: *** [libpthread/nptl/sysdeps/generic/dl-tls.os] Error 1

A previous commit (f26c5f6952ce9bf8edec9c1571c47addb1bcc442) touched
on a similar issue, but added the include to the incorrect location.

Reported-by: Christophe Lyon <christophe.lyon@st.com> [arm nommu]
Reported-by: Daniel Beecham <daniel@lunix.se> [static x86_64]
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

string/microblaze: fix preprocessor style

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

libc/sysdeps: dup3: check for __NR_dup3 since older kernels/arches lack it

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

include/elf.h: update for ELFOSABI_* changes.

ELFOSABI_GNU replaces ELFOSABI_LINUX, the latter is kept as a compatibility
alias, and ELFOSABI_HURD is removed.  See the table on
<http://www.sco.com/developers/gabi/latest/ch4.eheader.html#osabi> for
reference.

Signed-off-by: Thomas Schwinge <thomas@codesourcery.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc/stdlib: add mkostemp helpers

Some projects (like udev) are starting to use this.

Imported from glibc.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

libc/sysdeps: add dup3 syscall wrapper

Some projects (like udev) are starting to use this.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

include/: ignore more symlinks

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

test/math: add libm-test-ulps-xtensa file

Add ulps file for Xtensa for the math test.

Signed-off-by: Chris Zankel <chris@zankel.net>

string/microblaze: Fix for little-endian

Fix the asm-optimised memcpy and memmove so they
work for little-endian as well as big-endian.

Testing has shown no issues, but I am not a microblaze
asm expert so YMMV.

Signed-off-by: Steve Bennett <steveb@workware.net.au>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

microblaze/bits/setjmp.h: Fix stray #endif

Introduced in c53099f7

Signed-off-by: Steve Bennett <steveb@workware.net.au>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

xtensa: add missing atomic intrinsics

Add more atomic intrinsics. These are mostly non 32-bit versions, which
are not support by Xtensa. This file needs some more clean-up and
consolidation.

Signed-off-by: Chris Zankel <chris@zankel.net>

xtensa: sysdep.h needs to include sys/syscall.h

The Xtensa version of sysdep.h didn't include sys/syscall.h. This header
file is, for example, needed by not-cancel.h, which uses the
INTERNAL_SYSCALL macro.

Signed-off-by: Chris Zankel <chris@zankel.net>

xtensa: add a stub user.h file

Xtensa only supports the ELF binary format, so the user.h file isn't
needed. However, it's included by other files, so provide a stub user.h.

Signed-off-by: Chris Zankel <chris@zankel.net>

ldd: ELFCLASSM was not defined for the Xtensa architecture

Signed-off-by: Chris Zankel <chris@zankel.net>

xtensa: fix a misnumbered register in sigsetjmp

Signed-off-by: Chris Zankel <chris@zankel.net>

xtensa: add a set of atomic intrinsics

Add a new file that provides various atomic intrinsics, which use the
conditional store instruction.

Signed-off-by: Chris Zankel <chris@zankel.net>

xtensa: use atomic instructions instead of a syscall

Replace system calls with atomic instructions for 'compare and swap'
in linuxthreads.old.

Signed-off-by: Chris Zankel <chris@zankel.net>

xtensa: add sigcontext.h

Signed-off-by: Chris Zankel <chris@zankel.net>

_scanf.c: Implement 'm' modifier for 'c' and '[' conversions.

The current code implements the 'm' modifier only for 's'
conversions and would cause a segfault if it was used for 'c'
or '[' conversions. This patch extends the code to cover these
cases too.

Signed-off-by: Will Newton <will.newton@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

utils: Fix missing semicolon

Thanks to Manuel Rüger for spotting this one.

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

__uclibc_main: remove stray trailing space

test post-receive hook

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

test: Fix some warnings in inet

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

test: handle !DO_C99_MATH

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc/sysdeps: add __kernel_long and __kernel_ulong

Linux 3.4 added __kernel_long_t and __kernel_ulong_t and various
exported header files were updated to use these new types. Add the
definitions for __kernel_long_t and __kernel_ulong_t to the relevant
kernel_types.h headers.

This change was automated with the following scriptlet

  git grep --name-only 'typedef.*__kernel_old_dev_t' \
    | xargs sed -i '/typedef.*__kernel_old_dev_t/ a\
  typedef long\t\t__kernel_long_t;\
  typedef unsigned long\t__kernel_ulong_t;'

Whitespace in arm, avr32, hppa, sparc was then manually fixed up.

Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
--
Here's a cleaned up patch which should get the whitespace right. I'm a
bit iffy about the sparc changes they make sense to me but it's not a
platform I have access to.

I can break this up per arch or per maintainer if requested.

 libc/sysdeps/linux/alpha/bits/kernel_types.h      |    2 ++
 libc/sysdeps/linux/arm/bits/kernel_types.h        |    2 ++
 libc/sysdeps/linux/avr32/bits/kernel_types.h      |    2 ++
 libc/sysdeps/linux/bfin/bits/kernel_types.h       |    2 ++
 libc/sysdeps/linux/c6x/bits/kernel_types.h        |    2 ++
 libc/sysdeps/linux/cris/bits/kernel_types.h       |    2 ++
 libc/sysdeps/linux/e1/bits/kernel_types.h         |    2 ++
 libc/sysdeps/linux/h8300/bits/kernel_types.h      |    2 ++
 libc/sysdeps/linux/hppa/bits/kernel_types.h       |    2 ++
 libc/sysdeps/linux/i386/bits/kernel_types.h       |    2 ++
 libc/sysdeps/linux/ia64/bits/kernel_types.h       |    2 ++
 libc/sysdeps/linux/m68k/bits/kernel_types.h       |    2 ++
 libc/sysdeps/linux/microblaze/bits/kernel_types.h |    2 ++
 libc/sysdeps/linux/mips/bits/kernel_types.h       |    4 ++++
 libc/sysdeps/linux/nios2/bits/kernel_types.h      |    2 ++
 libc/sysdeps/linux/powerpc/bits/kernel_types.h    |    4 ++++
 libc/sysdeps/linux/sh/bits/kernel_types.h         |    2 ++
 libc/sysdeps/linux/sh64/bits/kernel_types.h       |    2 ++
 libc/sysdeps/linux/sparc/bits/kernel_types.h      |    4 ++++
 libc/sysdeps/linux/v850/bits/kernel_types.h       |    2 ++
 libc/sysdeps/linux/x86_64/bits/kernel_types.h     |    2 ++
 libc/sysdeps/linux/xtensa/bits/kernel_types.h     |    2 ++
 22 files changed, 50 insertions(+)
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

librt: re-add SIGCANCEL to the list of blocked signal in helper thread

Indeed if the libpthread is before the libc in the library look up
the SIGCANCEL is removed from the list of the blocked signal by
sigfillset func, this can produce the handler not properly called.
This commit revert what Denys modified in commit
162cfaea20d807f0ae329efe39292a9b22593b41.

Signed-off-by: Filippo Arcidiacono <filippo.arcidiacono@st.com>
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>

i386/bits/syscalls.h: allow immediate values as 6th syscall arg

Allow use of immedate values as the 6th syscall argument. Otherwise we must
store the arg on memory. This gives gcc more options to optimize better.

This also works around an issue with posix_fallocate.

Signed-off-by: Natanael Copa <ncopa@alpinelinux.org>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

bits/time.h: sync with glibc 2.16

CLOCK_MONOTONIC_RAW is available since 2.6.28
(2d42244ae71d: clocksource: introduce CLOCK_MONOTONIC_RAW), and
CLOCK_*_COARSE since 2.6.32 (da15cfdae033: time: Introduce
CLOCK_REALTIME_COARSE).

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc/inet/resolv.c: Use IE model for TLS data.

All other TLS data in uClibc is explicitly using the IE model,
so use it for __resp too. This generates smaller and faster
code.

Signed-off-by: Will Newton <will.newton@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc/inet/resolv.c: Remove obsolete comment.

The definition of __resp has been hidden for some years (commit
22de495d). Remove the comment suggesting that it is not.

Signed-off-by: Will Newton <will.newton@imgtec.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

Accept either __NR__lseek or __NR_lseek

In llseek.c, we already allow both variants; the same needs to be done here
to avoid duplicate definitions of lseek64_nocancel.

Signed-off-by: Bernd Schmidt <bernds@codesourcery.com>