Merge commit 'origin/master' into nptl

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

nptl: turn pointer demangling off

Currently the stub calls would
demangle the pointer that is never mangled. Alternatively, we could sync
nptl/sysdeps/unix/sysv/linux/libc_pthread_init.c with new glibc version
that properly mangles the pointers, but that would need more validation
to make sure that all struct pthread_function readers do demangling.
uclibc currently does not do pointer mangling everywhere so it's more
consistent to just disable it here too.

Signed-off-by: Timo Teräs <timo.teras@iki.fi>
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

nptl: initialize stdio locking

uClibc requires the threading library to enable locking for
stdio, or the locking is not done at all.

Signed-off-by: Timo Teräs <timo.teras@iki.fi>
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

nptl: mark symbols with libc forwarder hidden

Add attribute_hidden to all symbols having libc forwarder. This prevents
recursive self calls which would happen if libc is before libpthread in
linking order: the forwarder functions would call itself via the function
table, since the libpthread symbols would get overwritten with libc ones.
This has not been a problem in glibc since there these symbols are marked
hidden with linker version-script. Since we don't use one, we need to mark
these explicitly.

Signed-off-by: Timo Teräs <timo.teras@iki.fi>
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

*.o[sS]: depend on pregen

allows 'make realclean && make utils' to behave properly

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

ldso: Add config option for controlling LD_PRELOAD

On hardened system it could be useful to disable the use
of LD_PRELOAD environment variable for preloading shared objects
before the system libraries. So this patch add a config option,
LDSO_PRELOAD_ENV_SUPPORT, to control this behaviour.
It is enabled by default.

Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

test: add API test

part1; Needs improvement:
Think about a sensible way to map feature sets to the mandated syms.
Our LEGACY stuff maps to OB, OBXSI for example.
A "normal" config should satisfy the BASE (i.e. $foo.SUSv4.syms)

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

provide prototype for internal helper

i386 was missing a prototype for __tls_get_addr()

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

temporarily turn on warnings for NPTL

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

silence debug warning

variable 'tmp' set but not used

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

whitespace cleanup

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

Merge remote branch 'origin/master' into nptl

Conflicts:
libc/unistd/confstr.c

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

nptl/sysdeps/unix/sysv/linux/i386: cleanup -DUSE___THREAD

Remove the -DUSE___THREAD from specific assembler files since its
used everywhere anyway.

This is a cleanup after 447a9d1cc181395c3e2ea77ea88e45dee4b30ce6

Signed-off-by: Natanael Copa <natanael.copa@gmail.com>
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

attribute_optimize: fix typo with args

curious how 308f5c6e5fd56ea3d1a5512e34388aad788f1180 ever worked.. :P

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

pthread_mutex_timedlock: always build with optimizations

avoids breakage with DODEBUG and lll_timedlock() constraints

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

nptl/sysdeps/unix/sysv/linux/i386: enable USE___THREAD for assembler

NPTL defines errno as __thread variable. Define USE___THREAD so proper
errno assignment is used in assembly files.

This fixes segfault while building python.

Signed-off-by: Timo Teräs <timo.teras@iki.fi>
Signed-off-by: Natanael Copa <natanael.copa@gmail.com>
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

use user-provided host binaries

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

prettify make clean

forgot extra/config

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

*.i depend on generated headers

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

silence warning about undefined preprocessor token

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

libc_regex: __libc_lock primitives are actually available on uClibc

__libc_lock primitives are actually available on uClibc when threading
support is enable, so in this case they can be used.

It also fixes the following compiler warnings:

In file included from libc/misc/regex/regex.c:55:
libc/misc/regex/regex_internal.h:49:1: warning: "__libc_lock_define" redefined
[SNIP]
libc/misc/regex/regex_internal.h:50:1: warning: "__libc_lock_init" redefined
[SNIP]
libc/misc/regex/regex_internal.h:51:1: warning: "__libc_lock_lock" redefined
[SNIP]
libc/misc/regex/regex_internal.h:52:1: warning: "__libc_lock_unlock" redefined

Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>
(cherry picked from commit 65f9ccdafd008abd9892dfc46fb9737ec4d964c5)

Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>

libc_regex: __libc_lock primitives are actually available on uClibc

__libc_lock primitives are actually available on uClibc when threading
support is enable, so in this case they can be used.

It also fixes the following compiler warnings:

In file included from libc/misc/regex/regex.c:55:
libc/misc/regex/regex_internal.h:49:1: warning: "__libc_lock_define" redefined
[SNIP]
libc/misc/regex/regex_internal.h:50:1: warning: "__libc_lock_init" redefined
[SNIP]
libc/misc/regex/regex_internal.h:51:1: warning: "__libc_lock_lock" redefined
[SNIP]
libc/misc/regex/regex_internal.h:52:1: warning: "__libc_lock_unlock" redefined

Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>

ldso_sh: Provide only one definition for elf_machine_type_class

It doesn't need to provide different definition of elf_machine_type_class
macro depending on TLS support.
It also fixes build for sh architecture when thread support is off:
  CC ldso/ldso/ldso.oS
In file included from ./ldso/ldso/sh/dl-sysdep.h:9,
                 from ./ldso/include/dl-string.h:11,
                 from ./ldso/include/ldso.h:40,
                 from ldso/ldso/ldso.c:33:
./include/tls.h:6:22: error: tls.h: No such file or directory
make: *** [ldso/ldso/ldso.oS] Error 1

Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>

resolv: tentatively fix usage of uninitialized DNS parameters

See "Possible regression from timeout commit for resolv.conf" thread.
Also remove superfluous NULL check.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Fixes NPTL build on x86.

Signed-off-by: Natanael Copa <natanael.copa@gmail.com>
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

.gitignore getconf binaries

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

confstr: properly stringify version parts

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

pthread_{attr_,}{s,g}etaffinity: cleanup symbol names

fixes typo in a1ee5b32fc698297149c770417354441ee62701e which forgot to
rename __pthread_attr_getaffinity_np to pthread_attr_getaffinity_np.
Thanks to Andre DRASZIK for mentioning my omission.

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

utils .gitignore update to include getconf

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

correctly quote version in confstr

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

Merge commit 'origin/master' into nptl

Conflicts:
libc/misc/utmp/utent.c
libc/sysdeps/linux/i386/bits/syscalls.h

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

getconf: print arbitrary GNU_LIBPTHREAD_VERSION string

the NPTL should probably print the version-string of the glibc it's
compatible with and not the uClibc version (check what mysql expects).

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

getconf: move to utils/

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

Fix use-after-free bug in __dns_lookup

If the type of the first answer does not match with the requested type,
then the dotted name was freed. If there are no further answers in
the DNS reply, this pointer was used later on in the same function.
Additionally it is passed to the caller, and caused strange
behaviour.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

i386/bits/syscalls.h: more compact, but stack-hungry syscall code. disabled so far

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

i386/bits/syscalls.h: explain _BITS_SYSCALLS_ASM

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

i386/bits/syscalls.h: generate better code using more restrictive asm constraints

Example: time() doesn't push/pop two registers now, and is smaller:

    <__GI_time>:
    -57                     push   %edi
    -8b 7c 24 08            mov    0x8(%esp),%edi
    -53                     push   %ebx
    -89 fb                  mov    %edi,%ebx
    +8b 4c 24 04            mov    0x4(%esp),%ecx
    +87 cb                  xchg   %ecx,%ebx
     b8 0d 00 00 00         mov    $0xd,%eax
     cd 80                  int    $0x80
    -5b                     pop    %ebx
    -5f                     pop    %edi
    +87 cb                  xchg   %ecx,%ebx
     c3                     ret

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

i386/bits/syscalls.h: explain how it works. no code changes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

libutil/login: was totally broken. fixed

Also made login() and logout() usable with utmpname(non_std_utmp_file)
   text           data     bss     dec     hex filename
    111              0       0     111      6f libutil/login.o
    217              0       0     217      d9 libutil/login.o
   text           data     bss     dec     hex filename
    164              0       0     164      a4 libutil/logout.o
    134              0       0     134      86 libutil/logout.o

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

utent: do not create extra static functions if !THREADS

   text           data     bss     dec     hex filename
    547              8     384     939     3ab libc/misc/utmp/utent.o
    519              8     384     911     38f libc/misc/utmp/utent.o

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

regex_old: convert static flag variable to smallint

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

getutid is not used internally, removing hidden_proto

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

nptl needs stdio futexes enabled.

this got reverted in the merge I just did

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

poll.c: fix incorrect merge

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

Revert "remove config knobs that belong to the nptl branch"

This reverts commit de82658a05de0ace6f3d261dedc358d3a14f3681.

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

Merge commit 'origin/master' into nptl

Conflicts:
Makefile.in
extra/Configs/Config.in
libc/sysdeps/linux/common/bits/kernel-features.h
libc/sysdeps/linux/common/poll.c
libc/sysdeps/linux/common/sysdep.h
libc/sysdeps/linux/sh/sysdep.h

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

bump version to 0.9.32-git

NPTL merge candidate, finally!

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

Release 0.9.31

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

resolv: DEBUG-print nameserver we talk to

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

utils: hide comment if !V

pass CFLAGS-dir to CC-m

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

prettify make clean

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

nios2: Define INTERNAL_SYSCALL_NCS, use common sycall defines

Define the nios2 version of INTERNAL_SYSCALL_NCS and remove all sycall
defines already defined by the common syscall defines.

This will make the master branch of uClibc compile again for nios2.

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

nios2: Update fcntl.h from m68k

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

lift printf field width limit

uClibc mishandles printf field width limits larger than 40959, as a
result of misguided overflow-protection code.  This causes spurious test
failures with GNU coreutils, which depends on "%65536s" and "%20000000f"
working according to spec.

Signed-off-by: Michael Deutschmann <michael@talamasca.ocis.net>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

nios2: Add sys/user.h

This is a fixed version containing copyright information. The patch should now
apply using 'git am'.

linux/user.h is no longer exported during kernel make headers_install

Signed-off-by: Atle Nissestad <atle@nissestad.no>
Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

config: tweak text

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

buildsys: touchup 'make dist' a tiny bit

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

bump version

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

bump version to 0.9.31-rc1

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

buildsys: Fix O= PREFIX= case

make sure that something like:
$ rm -rf /dev/shm/o/* /dev/shm/o/.conf* /dev/shm/DELME
$ time (make defconfig O=/dev/shm/o 2>&1 >& /dev/null && make \
install_dev -j O=/dev/shm/o PREFIX=/dev/shm/DELME 2>&1 >& /dev/null)
works

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

buildsys: fix dir order-only prereq of unifdef

|$(@D) -> zilch, so spell it out

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

add MULTILIB_DIR: Path component for libdirs

forgot to checkin this file

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

config: support make defconfig O=/f/o/o

previously wanted to create the temporary cfg file in the srcdir (which
failed on a RO srcdir)

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

add MULTILIB_DIR: Path component for libdirs

defaults to "lib". Other prominent values include "lib32" or "lib64"

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

install: fix O= PREFIX= install

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

reduce number of mkdir calls

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

remove config knobs that belong to the nptl branch

Fixes bug #1243

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

pull kernel-features.h from NPTL

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

poll: unavailable on linux < 2.2.0

fixes bug #253

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

pregen-headers: Add new target which depends on headers.

* This target is necessary to serialize making of
  headers before pregen-headers and pregen-headers
  before the rest of build

Signed-off-by: Khem Raj <raj.khem@gmail.com>

pregen: Fix the parallel build problem in pregen target

* pregen-headers really need the headers to get themselves generated.
  so we create explicit dependency and invoke make as an action on
  pregen. This fixes the parallel build issue for better.

Signed-off-by: Khem Raj <raj.khem@gmail.com>

get rid of extra variables for nptl pregen altogether

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

order headers before pregen-headers

   until someone gets a chance to fix this for real, this will unbreak
   building from a clean repo, as some of the pregen headers need some
   of the static headers

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

improve parallel make behaviour

 * add library dependencies for libdl,libpthread
 * fix typo in librt/Makefile.in
 * also remove extra trailing slashes on i386, sparc pregen headers

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

initialize new pregen-headers-y variable before including sub makefiles

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

nptl/headers: Convert the makefile targets into variables.

* Introduce pregen-headers-y which is list of generated headers
  during the build and add this to pregen target in top level makefile.
  Also add nptl_headers, nptl_arch_headers, nptl_linux_headers
  to this if compiling with nptl.

* Replace pthread_pregen var with pregen-headers-y.

* Make nptl_headers, nptl_arch_headers, nptl_linux_headers variables
  instead of sub targets.

Signed-off-by: Khem Raj <raj.khem@gmail.com>

Get x86_64 compile to succeed.

* The changes to clone.S needs testing

Signed-off-by: Khem Raj <raj.khem@gmail.com>

sparc: disable cancellable system, as it fails in strange ways right now

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

sparc: also needs generic version of lowlevelrobustlock

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

Add dummy tcb-offsets.h

* This file is used during creation of real tcb-offsets.h
  because lowlevellock.h from i386 also include tcb-offsets.h
  and it is inturn included by tls.h and tls.h is used in
  the dummy C file generated from tcb-offsets.sym, it created
  a catch-22

Signed-off-by: Khem Raj <raj.khem@gmail.com>

nptl_sh: Fix build problem with FUTEX_CLOCK_REALTIME

This patch fixes a build problem into the lowlevellock.S
when support FUTEX_CLOCK_REALTIME (kernel newer than 2.6.29).
__have_futex_clock_realtime symbol is not defined when set
__ASSUME_FUTEX_CLOCK_REALTIME. The same is on i386 arch.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>

x86_64: compile pthread it's own version of __syscall_error

also fix up clone.S from sync

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

more fixes for x86_64 nptl

also cleanup some PIC (should be __PIC__) that snuck in

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

add dwarf defines used in cfi statements

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

update kernel-features.h with futex features

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

use macro to link ldso to libpthread, so it gets the the right name on 64 bit archs

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

fixes to get nptl compiling for x86_64

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

x86_64: finish up tls reloc support in ldso

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

Fixes to get nptl compiling for x86

Signed-off-by: Khem Raj <raj.khem@gmail.com>

Fix typo in generated file names.

Signed-off-by: Khem Raj <raj.khem@gmail.com>

i386 unify sysdep.h

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

x86_64: unify sysdep.h

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

nptl: Fix clean target to remove generated files

Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>

include tcb-offsets.h in sysdep-cancel.h for arm.

Signed-off-by: Khem Raj <raj.khem@gmail.com>

Get new nptl building on SH4

Signed-off-by: Khem Raj <raj.khem@gmail.com>

compile fixes for i386 nptl

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

fix nptl build for sparc

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

move pthread_barrier_{init,destroy} to sysdeps/pthread
for archs that implement their own like sparc and x86_64

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

Fix nptl build for mips

Signed-off-by: Khem Raj <raj.khem@gmail.com>