avr32: add varargs handling of prctl syscall

prctl is defined to use varargs in the header file, hence it needs varargs
specific handling in the source. This patch properly handles the variodic
argument before the syscall is passed to the kernel for the AVR32 architecture.

Signed-off-by: Hans-Christian Egtvedt <hans-christian.egtvedt@atmel.com>

Merge remote branch 'origin/master' into nptl

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

Make _dl_dprintf buf non-static to avoid multithreading bugs.

Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

Merge commit 'origin/master' into nptl

Conflicts:
libc/signal/sigpause.c
libc/string/x86_64/memset.S

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

malloc: fix race condition and other bugs in the no-mmu malloc

Fixes multiple race conditions on mmb list. This was done by
making the mmb_heap_lock into a recursive lock and making the
regular heap_lock extend to cover the mmb heap handling.

Also move the new_mmb allocation up to before the mmb list is
iterated through to find the insertion point. When the mmb_heap
also runs out and needs to be extended when the regular heap is
just extended, the mmb list could be messed up.

Signed-off-by: Freeman Wang <xwang@ubicom.com>
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

x86_64: fix multiple definition of chk functions

also enable __chk_fail and only try to call it when SSP is on

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

check if USE_TLS is defined before use

This patch will convert all the #ifdef USE_TLS and #if USE_TLS to
 #if defined(USE_TLS) && USE_TLS.

By checking if the USE_TLS is defined before checking its value will result in
correct behavior for architectures not defining this config symbol.

Signed-off-by: Hans-Christian Egtvedt <hans-christian.egtvedt@atmel.com>
Acked-by: Carmelo AMOROSO <carmelo.amoroso@st.com>

check if USE_TLS is defined before use

This patch will convert all the #ifdef USE_TLS and #if USE_TLS to
 #if defined(USE_TLS) && USE_TLS.

By checking if the USE_TLS is defined before checking its value will result in
correct behavior for architectures not defining this config symbol.

Signed-off-by: Hans-Christian Egtvedt <hans-christian.egtvedt@atmel.com>
Acked-by: Carmelo AMOROSO <carmelo.amoroso@st.com>

x86_64: fix up the rest of the includes and aliases for nptl

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

libc: Remove pointless header string.h

It is already include from the internal header sigset-cvt-mask.h

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

libc: Add missing header for memset

Silence compiler warning by adding the missing string.h header
to include memset prototype.
Do some code tidy-up removing commented out libc_hidden_proto macros
in the meanwhile.

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

x86_64: add missing includes to sem_*.S

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

libc: Add missing header for memset

Silence compiler warning by adding the missing string.h header
to include memset prototype.
Do some code tidy-up removing commented out libc_hidden_proto macros
in the meanwhile.

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

fix multiple definitions of pthread_once for x86_64

also add a missing weak for clone

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

fix compiling test cases in static mode

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

x86_64: no waitpid syscall exists, so use wait4

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

build: Do not define -DPIC longer through build system

Now all code relies upon __PIC__ flag defined by compiler, so
it does not need to explicitly define the PIC macro through build system

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

build: Get rids of PIC macro using compiler flag __PIC__ instead

Based on Peter Mazinger's comments on a recent commit, I decided
to get rids of all occurrences of PIC changing them to __PIC__

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

build: Get rids of PIC macro using compiler flag __PIC__ instead

Based on Peter Mazinger's comments on a recent commit, I decided
to get rids of all occurrences of PIC changing them to __PIC__

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

Merge remote branch 'origin/master' into nptl

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

fix typo that sneaked in with 9a7506

Signed-off-by: Natanael Copa <natanael.copa@gmail.com>
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>

sh: fix endianess and optimise the SH4 memcpy

This patch fixes the big-endian code and adds a new optimization
only for little endian mode.
This optimization is based on prefetching and 64bit data transfer via FPU.
Tests shows that

----------------------------------------
 Memory bandwidth    |        Gain
                     | sh4-300 | sh4-200
----------------------------------------
 512 bytes to 16KiB  | ~20%    | ~25%
 from 32KiB to 16MiB | ~190%   | ~5%
----------------------------------------

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

syslog: use send(MSG_NOSIGNAL) instead of write, thus no need to handle SIGPIPE

Size changes by this and previous change:

text           data     bss     dec     hex filename
1151             13       2    1166     48e libc/misc/syslog/syslog.o
1093             10       2    1105     451 libc/misc/syslog/syslog.o
1047             10       2    1059     423 libc/misc/syslog/syslog.o

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

syslog: fix openlog(xx, LOG_KERN) and optimize a bit

The fix:

logfac == 0 in openlog(xx, logfac) is allowed now.
Corresponding internal openlog() call in vsyslog()
uses explicit LOG_USER in order to set it as a default
facility.

Optimizations:

mylock is not recursive now, since a single intenal call
of openlog is converted to a call to openlog_internal
which assumes that lock is already taken. No recursive
locking is possible now.

LogFacility is reduced to byte.

cache static LogFile in auto variable fd (smaller code).

vsyslog with bogus pri parameter wouldn't lock/unlock
and mess with signals - it will just return at once.

pass NULL as ident string in internal openlog call
- same effect as passing LogTag but smaller code.

comment out "if (LogTag)" checks - it is never NULL.

use the same struct sigaction for setting new sigaction
and for saving old one - saves ~32 bytes of stack.

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

Include bits/libc-lock.h to get __libc_once_define

Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

Go back to using nptl_arch_headers

this is part of the revert of 641e7e36ad0a9

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

Revert "nptl: make build depend directly on nptl headers"

This reverts commit 641e7e36ad0a9b2899b04f256c5b243ada7da279.

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

Include bits/libc-lock.h to get __libc_once_define

Signed-off-by: Khem Raj <raj.khem@gmail.com>
Acked-by: Austin Foxley <austinf@cetoncorp.com>

get x86_64 nptl building

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

nptl: fix even more old style declarations

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

nptl: fix a few more old style declerations

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

nptl: fix old style function declerations

also fix a few shadowed local warnings

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

nptl: make build depend directly on nptl headers

gets rid of nptl_headers,nptl_arch_headers,nptl_linux_headers targets

also, clean up extra slashes in i386/sparc and unify the implementation
of nptl header generation between all archs

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

warn about old-style crap in extra-annoyance-mode

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

test_dlopen: Remove extra -lpthread from LD_FLAGS for dltest{2}

dltest and dltest2 do not refer any symbols implemented
in pthread library, so do not explicitly link them with pthread.
It is required only for the shared objects that are loaded
via dlopen by those tests.

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

test_dlopen: Remove extra -lpthread from LD_FLAGS for dltest{2}

dltest and dltest2 do not refer any symbols implemented
in pthread library, so do not explicitly link them with pthread.
It is required only for the shared objects that are loaded
via dlopen by those tests.

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

i386 nptl: add missing weak_alias for pthread_cond_signal

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

ldso_tls: Refetch dtv from memory if THREAD_DTV has changed

_dl_update_slotinfo might change THREAD_DTV () (if it needs to reallocate it),
but the caller (__tls_get_addr) doesn't refetch dtv from memory, it uses its
cached copy. This may crash (if dtv[GET_ADDR_MODULE] is off the cliff, or
might read uninitialized memory and return it.
Typically dtv[GET_ADDR_MODULE].pointer.val is NULL and so __tls_get_addr
returns NULL + offset_within_PT_TLS. The next time __tls_get_addr is called
for the same library it will return correct address as _dl_update_slotinfo
won't need to be called.

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

test_tls: Add two new tests for tls tst-tls{16,17} taken from glibc

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

test_tls: Rename tst-tls16 to tst-tls-at-ctor

Rename tst-tls16 to tst-tls-at-ctor (including main and related
shared object) to avoid name clashing with new tests ported from
glibc.

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

build: Move test build flags in Makefile.in

Move test build flags in Makefile.in
Makefile should be used only as a container for other sub Makefiles

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

sh: Add a prototype for the gcc __set_fpscr internal function

The sh port uses the gcc __set_fpscr internal function, but neither gcc
nor glibc creates a prototype for it.  This leads to a bunch of random
warnings about implicit decls during the build.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>

sh: Add a prototype for the gcc __set_fpscr internal function

The sh port uses the gcc __set_fpscr internal function, but neither gcc
nor glibc creates a prototype for it.  This leads to a bunch of random
warnings about implicit decls during the build.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Carmelo Amoroso <carmelo.amoroso@st.com>

uClibc_ctype.h: fix inverted check for susv4 macro

Was causing build failures if ctype tables were disabled and susv4 was on

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

uClibc_ctype.h: fix inverted check for susv4 macro

Was causing build failures if ctype tables were disabled and susv4 was on

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

getaddrinfo.c: include tls.h before resolv.h

This is needed because tls.h ends up includeing nptl/descr.h
which needs the full res_state definition. resolv.h also ends up
including tls.h, but _without_ the full res_state struct it errors.

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

test/.gitignore: ignore a missed nptl test binary

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

test/.gitignore: Ignore a few more test binaries

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

test/.gitignore: Ignore a few more test binaries

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

test/Rules.mak: disable -z,defs for now

* A bunch of the nptl tests don't compile when -z,defs is on
* also turn on -std=gnu99 for the HOSTCC (_glibc) compiles

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

test/nptl: rework tst-tls3 to link with -z,now

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

test/: compile fixes for nptl and tls tests

 * Add a GLIBC_TESTS_DISABLED var because many of these tests
   need internal uClibc stuff. This disables the HOSTCC build of
   these tests
 * Fix up tls test macros for x86
 * Fix the linking of the tls tests

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

test/nptl: don't use c99 features if you don't need to

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

test: nptl and tls tests need main include dir

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

test/dlopen: fix build for libtest.so

needs -lpthread, recent addition of -z,defs and -z,now exposed this

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

test/dlopen: Match the function name that libtest actually sets

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

test/pthread: fix build of cancellation-points test

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

test/stat: fix compiling the memcmp-stat test when __USE_MISC is defined

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

test/dlopen: fix build for libtest.so

needs -lpthread, recent addition of -z,defs and -z,now exposed this

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

test/dlopen: Match the function name that libtest actually sets

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

test/pthread: fix build of cancellation-points test

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

test/stat: fix compiling the memcmp-stat test when __USE_MISC is defined

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

generate nptl headers without a recursive make call

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

don't use __i686 in symbol names as it seems gcc can define it

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

Merge remote branch 'origin/master' into nptl

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

PIC fixes for hardened

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>

define local stack_chk_guard for nptl

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

ldso: initialize stack_chk_guard after TLS is initialized

stack_chk_guard is on thread local storage so we need to init TLS before
we can init stack_chk_guard.

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

Add update instructions for x86_64 WRT utmp format

Commit 983c4d5188f8a7eb00766e3b3e89b218899a3a08 changed the utmp format
on x86_64 hosts to be 32bit compatible. Mention this fact and hint
x86_64 users that they will have to wipe their utmp file upon upgrade.

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

make x86_64 struct utmp glibc-compatible

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

try to probe if stat.t_mtim.tv_nsec is available

Do not treat tv_nsec mismatches as errors on filesystems without support for it.

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

correct comment

and disable built-in rules

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

Robustify setting silent mode

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

sync confname, environments with glibc

Plus related synch.
Add a testcase for the sysconf variables based on the one from glibc

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

ignore some more testcase binaries

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

call nanosleep directly

usleep is a SUSv3 function.
TODO: nanosleep moved from SUSv3 Timers to SUSv4 Base

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

conditional cancellation-point test of usleep

usleep is a SUSv3 function

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

pass more ld flags

add -z defs,now

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

ether_*() depend on socket or IP support

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

avoid warning about unused variable

If LDSO_SEARCH_INTERP_PATH is off then ptmp wasn't used.

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

in6_{pktinfo,mtuinfo}: provide type if IPV6 is on

Disable some IPv6 header defines if IPv6 is off.

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

nommu: use MAP_UNINITIALIZE for mallocs

Now that the kernel supports MAP_UNINITIALIZE, have the malloc places use
it to get real uninitialized memory on no-mmu systems.  This avoids a lot
of normally useless overhead involved in zeroing out all of the memory
(sometimes multiple times).

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

bits/mmap.h: unify ala linux asm-generic efforts

Most ports have the same exact mman bit defines, so let's unify things
like the linux kernel has with the asm-generic efforts.

A few ports are left behind as they are non-trivial to merge -- the arch
maintainers can tackle it if they care.

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

futimens: Ducument why we need manual hidden_proto

Also, remove unneeded __need_timespec (It cuts off a tiny bit of
work for CPP but is of course not strictly needed).

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

ldso: let people disable the implicit path search

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

Merge remote branch 'origin/master' into nptl_merge

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

Revert "libc: add hidden calls to pthread cleanup funcs"

This reverts commit e0ac4efbdb498319f03a2a95d75d061ab6c68491.

Was causing segfaults on every pthread linked executable

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

sh: Add new optimisation to the SH4 memcpy

This optimization is based on prefetching and 64bit data transfer via FPU
(only for the little endianess)

Tests shows that:

  ----------------------------------------
  Memory bandwidth    |        Gain
                      | sh4-300 | sh4-200
  ----------------------------------------
  512 bytes to 16KiB  | ~20%    | ~25%
  from 32KiB to 16MiB | ~190%   | ~5%
  ----------------------------------------

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

Merge remote branch 'origin/master' into nptl_merge

Conflicts:
Rules.mak
libc/misc/sysvipc/msgq.c
test/Rules.mak

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

make tls and nptl test build

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

Look at HAVE_SHARED

We do not have UCLIBC_STATIC (anymore) but !HAVE_SHARED

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

remove wait callbacks for non-cancelation funcs

No need to install a cancellation handler, these are no cancellation
point.

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

include string.h for memset()

This avoids a warnings about implicit declaration of function 'memset'.

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

Makefile: fix indentation

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

install_headers.sh: make more user friendly

Make it easier to run this by hand and don't abort when recursive chown
and chmod fail as these often aren't due to uClibc settings.

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

remove incorrect text from abspath option

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

install shared libs with +x perms

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

realclean: wipe kconfig objects

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

silence rule overriding

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

futimens: add function

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