Merge "Add include guard to fileext.h"

Description: In the kernel the epoll_event structure is packed
in 64 bit kernel builds to allow the structure to be more easily
compatible with 32 bit user space.  As a result, when user space
is 64-bit the structure must be packed as well.

Add unit test to show the ptr alignment issue.

Change-Id: I2c4848d5e38a357219091f350f9b6e3da05090da
Signed-off-by: Philip Hatcher <philip.hatcher@intel.com>
Signed-off-by: Fengwei Yin <fengwei.yin@intel.com>
Reviewed-by: Hazarika, Prodyut <prodyut.hazarika@intel.com>
Tested-by: Hazarika, Prodyut <prodyut.hazarika@intel.com>

ABI-bugs.txt

@@ -8,8 +8,5 @@ KNOWN ABI BUGS
   sigset_t is too small on ARM and x86 (but correct on MIPS), so support
   for real-time signals is broken. http://b/5828899
 
-  Too few TLS slots mean we can't allocate 128 pthread_key_t instances,
-  which POSIX says should be the minimum.
-
   atexit(3) handlers registered by a shared library aren't called on
   dlclose(3); this only affects ARM. http://b/4998315

HACKING.txt

@@ -0,0 +1,165 @@
+Working on bionic
+=================
+
+What are the big pieces of bionic?
+----------------------------------
+
+libc/ --- libc.so, libc.a
+  The C library. Stuff like fopen(3) and kill(2).
+libm/ --- libm.so, libm.a
+  The math library. Traditionally Unix systems kept stuff like sin(3) and
+  cos(3) in a separate library to save space in the days before shared
+  libraries.
+libdl/ --- libdl.so
+  The dynamic linker interface library. This is actually just a bunch of
+  stubs that the dynamic linker replaces with pointers to its own
+  implementation at runtime. This is where stuff like dlopen(3) lives.
+libstdc++/ --- libstdc++.so
+  The C++ ABI support functions. The C++ compiler doesn't know how to
+  implement thread-safe static initialization and the like, so it just calls
+  functions that are supplied by the system. Stuff like __cxa_guard_acquire
+  and __cxa_pure_virtual live here.
+
+linker/ --- /system/bin/linker and /system/bin/linker64
+  The dynamic linker. When you run a dynamically-linked executable, its ELF
+  file has a DT_INTERP entry that says "use the following program to start me".
+  On Android, that's either linker or linker64 (depending on whether it's a
+  32-bit or 64-bit executable). It's responsible for loading the ELF executable
+  into memory and resolving references to symbols (so that when your code tries
+  to jump to fopen(3), say, it lands in the right place).
+
+tests/ --- unit tests
+  The tests/ directory contains unit tests. Roughly arranged as one file per
+  publicly-exported header file.
+benchmarks/ --- benchmarks
+  The benchmarks/ directory contains benchmarks.
+
+
+What's in libc/?
+----------------
+
+libc/
+  arch-arm/
+  arch-arm64/
+  arch-common/
+  arch-mips/
+  arch-mips64/
+  arch-x86/
+  arch-x86_64/
+    # Each architecture has its own subdirectory for stuff that isn't shared
+    # because it's architecture-specific. There will be a .mk file in here that
+    # drags in all the architecture-specific files.
+    bionic/
+      # Every architecture needs a handful of machine-specific assembler files.
+      # They live here.
+    include/
+      machine/
+        # The majority of header files are actually in libc/include/, but many
+        # of them pull in a <machine/something.h> for things like limits,
+        # endianness, and how floating point numbers are represented. Those
+        # headers live here.
+    string/
+      # Most architectures have a handful of optional assembler files
+      # implementing optimized versions of various routines. The <string.h>
+      # functions are particular favorites.
+    syscalls/
+      # The syscalls directories contain script-generated assembler files.
+      # See 'Adding system calls' later.
+
+  include/
+    # The public header files on everyone's include path. These are a mixture of
+    # files written by us and files taken from BSD.
+
+  kernel/
+    # The kernel uapi header files. These are scrubbed copies of the originals
+    # in external/kernel-headers/. These files must not be edited directly. The
+    # generate_uapi_headers.sh script should be used to go from a kernel tree to
+    # external/kernel-headers/ --- this takes care of the architecture-specific
+    # details. The update_all.py script should be used to regenerate bionic's
+    # scrubbed headers from external/kernel-headers/.
+
+  private/
+    # These are private header files meant for use within bionic itself.
+
+  stdio/
+  stdlib/
+  unistd/
+    # These are legacy files of unknown provenance. In the past, bionic was a
+    # mess of random versions of random files from all three of FreeBSD, NetBSD,
+    # and OpenBSD! We've been working to clean that up, but these directories
+    # are basically where all the stuff we haven't got to yet lives.
+  dns/
+    # Contains the DNS resolver (originates from NetBSD code).
+
+  upstream-dlmalloc/
+  upstream-freebsd/
+  upstream-netbsd/
+  upstream-openbsd/
+    # These directories contain unmolested upstream source. Any time we can
+    # just use a BSD implementation of something unmodified, we should.
+    # The structure under these directories mimics the upstream tree,
+    # but there's also...
+    android/
+      include/
+        # This is where we keep the hacks necessary to build BSD source
+        # in our world. The *-compat.h files are automatically included
+        # using -include, but we also provide equivalents for missing
+        # header/source files needed by the BSD implementation.
+
+  bionic/
+    # This is the biggest mess. The C++ files are files we own, typically
+    # because the Linux kernel interface is sufficiently different that we
+    # can't use any of the BSD implementations. The C files are usually
+    # legacy mess that needs to be sorted out, either by replacing it with
+    # current upstream source in one of the upstream directories or by
+    # switching the file to C++ and cleaning it up.
+
+  tools/
+    # Various tools used to maintain bionic.
+
+  tzcode/
+    # A modified superset of the IANA tzcode. Most of the modifications relate
+    # to Android's use of a single file (with corresponding index) to contain
+    # time zone data.
+  zoneinfo/
+    # Android-format time zone data.
+    # See 'Updating tzdata' later.
+
+
+Adding system calls
+-------------------
+
+Adding a system call usually involves:
+
+  1. Add entries to SYSCALLS.TXT.
+     See SYSCALLS.TXT itself for documentation on the format.
+  2. Run the gensyscalls.py script.
+  3. Add constants (and perhaps types) to the appropriate header file.
+     Note that you should check to see whether the constants are already in
+     kernel uapi header files, in which case you just need to make sure that
+     the appropriate POSIX header file in libc/include/ includes the
+     relevant file or files.
+  4. Add function declarations to the appropriate header file.
+  5. Add at least basic tests. Even a test that deliberately supplies
+     an invalid argument helps check that we're generating the right symbol
+     and have the right declaration in the header file. (And strace(1) can
+     confirm that the correct system call is being made.)
+
+
+Updating kernel header files
+----------------------------
+
+As mentioned above, this is currently a two-step process:
+
+  1. Use generate_uapi_headers.sh to go from a Linux source tree to appropriate
+     contents for external/kernel-headers/.
+  2. Run update_all.py to scrub those headers and import them into bionic.
+
+
+Updating tzdata
+---------------
+
+This is fully automated:
+
+  1. Run update-tzdata.py.
+

benchmarks/Android.mk

@@ -0,0 +1,80 @@
+#
+# Copyright (C) 2013 The Android Open Source Project
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+ifneq ($(BUILD_TINY_ANDROID), true)
+
+LOCAL_PATH := $(call my-dir)
+
+# -----------------------------------------------------------------------------
+# Benchmarks.
+# -----------------------------------------------------------------------------
+
+benchmark_c_flags = \
+    -O2 \
+    -Wall -Wextra \
+    -Werror \
+    -fno-builtin \
+    -std=gnu++11 \
+
+benchmark_src_files = \
+    benchmark_main.cpp \
+    math_benchmark.cpp \
+    property_benchmark.cpp \
+    pthread_benchmark.cpp \
+    semaphore_benchmark.cpp \
+    stdio_benchmark.cpp \
+    string_benchmark.cpp \
+    time_benchmark.cpp \
+    unistd_benchmark.cpp \
+
+# Build benchmarks for the device (with bionic's .so). Run with:
+#   adb shell bionic-benchmarks
+include $(CLEAR_VARS)
+LOCAL_MODULE := bionic-benchmarks
+LOCAL_MODULE_STEM_32 := bionic-benchmarks32
+LOCAL_MODULE_STEM_64 := bionic-benchmarks64
+LOCAL_MULTILIB := both
+LOCAL_ADDITIONAL_DEPENDENCIES := $(LOCAL_PATH)/Android.mk
+LOCAL_CFLAGS += $(benchmark_c_flags)
+LOCAL_C_INCLUDES += external/stlport/stlport bionic/ bionic/libstdc++/include
+LOCAL_SHARED_LIBRARIES += libstlport
+LOCAL_SRC_FILES := $(benchmark_src_files)
+include $(BUILD_EXECUTABLE)
+
+ifeq ($(HOST_OS)-$(HOST_ARCH),linux-x86)
+ifeq ($(TARGET_ARCH),x86)
+LINKER = linker
+NATIVE_SUFFIX=32
+else
+LINKER = linker64
+NATIVE_SUFFIX=64
+endif
+
+bionic-benchmarks-run-on-host: bionic-benchmarks $(TARGET_OUT_EXECUTABLES)/$(LINKER) $(TARGET_OUT_EXECUTABLES)/sh
+	if [ ! -d /system -o ! -d /system/bin ]; then \
+	  echo "Attempting to create /system/bin"; \
+	  sudo mkdir -p -m 0777 /system/bin; \
+	fi
+	mkdir -p $(TARGET_OUT_DATA)/local/tmp
+	cp $(TARGET_OUT_EXECUTABLES)/$(LINKER) /system/bin
+	cp $(TARGET_OUT_EXECUTABLES)/sh /system/bin
+	ANDROID_DATA=$(TARGET_OUT_DATA) \
+	ANDROID_ROOT=$(TARGET_OUT) \
+	LD_LIBRARY_PATH=$(TARGET_OUT_SHARED_LIBRARIES) \
+		$(TARGET_OUT_EXECUTABLES)/bionic-benchmarks$(NATIVE_SUFFIX) $(BIONIC_BENCHMARKS_FLAGS)
+endif # linux-x86
+
+endif # !BUILD_TINY_ANDROID

benchmarks/benchmark_main.cpp

@@ -23,6 +23,8 @@
 #include <string>
 #include <map>
 
+#include <inttypes.h>
+
 static int64_t gBytesProcessed;
 static int64_t gBenchmarkTotalTimeNs;
 static int64_t gBenchmarkStartTimeNs;
@@ -162,8 +164,8 @@ void Benchmark::RunWithArg(int arg) {
     snprintf(full_name, sizeof(full_name), "%s", name_);
   }
 
-  printf("%-20s %10lld %10lld%s\n", full_name,
-         static_cast<int64_t>(iterations), gBenchmarkTotalTimeNs/iterations, throughput);
+  printf("%-20s %10d %10" PRId64 "%s\n", full_name,
+         iterations, gBenchmarkTotalTimeNs/iterations, throughput);
   fflush(stdout);
 }
 

benchmarks/math_benchmark.cpp

@@ -0,0 +1,171 @@
+/*
+ * Copyright (C) 2013 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "benchmark.h"
+
+#include <math.h>
+
+// Avoid optimization.
+double d;
+double v;
+
+static void BM_math_sqrt(int iters) {
+  StartBenchmarkTiming();
+
+  d = 0.0;
+  v = 2.0;
+  for (int i = 0; i < iters; ++i) {
+    d += sqrt(v);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_math_sqrt);
+
+static void BM_math_log10(int iters) {
+  StartBenchmarkTiming();
+
+  d = 0.0;
+  v = 1234.0;
+  for (int i = 0; i < iters; ++i) {
+    d += log10(v);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_math_log10);
+
+static void BM_math_logb(int iters) {
+  StartBenchmarkTiming();
+
+  d = 0.0;
+  v = 1234.0;
+  for (int i = 0; i < iters; ++i) {
+    d += logb(v);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_math_logb);
+
+static void BM_math_isinf_NORMAL(int iters) {
+  StartBenchmarkTiming();
+
+  d = 0.0;
+  v = 1234.0; // FP_NORMAL
+  for (int i = 0; i < iters; ++i) {
+    d += (isinf)(v);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_math_isinf_NORMAL);
+
+static void BM_math_isinf_NAN(int iters) {
+  StartBenchmarkTiming();
+
+  d = 0.0;
+  v = nan(""); // FP_NAN
+  for (int i = 0; i < iters; ++i) {
+    d += (isinf)(v);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_math_isinf_NAN);
+
+static void BM_math_isinf_INFINITE(int iters) {
+  StartBenchmarkTiming();
+
+  d = 0.0;
+  v = HUGE_VAL; // FP_INFINITE
+  for (int i = 0; i < iters; ++i) {
+    d += (isinf)(v);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_math_isinf_INFINITE);
+
+static void BM_math_isinf_ZERO(int iters) {
+  StartBenchmarkTiming();
+
+  d = 0.0;
+  v = 0.0; // FP_ZERO
+  for (int i = 0; i < iters; ++i) {
+    d += (isinf)(v);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_math_isinf_ZERO);
+
+
+
+
+
+
+static void BM_math_fpclassify_NORMAL(int iters) {
+  StartBenchmarkTiming();
+
+  d = 0.0;
+  v = 1234.0; // FP_NORMAL
+  for (int i = 0; i < iters; ++i) {
+    d += fpclassify(v);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_math_fpclassify_NORMAL);
+
+static void BM_math_fpclassify_NAN(int iters) {
+  StartBenchmarkTiming();
+
+  d = 0.0;
+  v = nan(""); // FP_NAN
+  for (int i = 0; i < iters; ++i) {
+    d += fpclassify(v);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_math_fpclassify_NAN);
+
+static void BM_math_fpclassify_INFINITE(int iters) {
+  StartBenchmarkTiming();
+
+  d = 0.0;
+  v = HUGE_VAL; // FP_INFINITE
+  for (int i = 0; i < iters; ++i) {
+    d += fpclassify(v);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_math_fpclassify_INFINITE);
+
+static void BM_math_fpclassify_ZERO(int iters) {
+  StartBenchmarkTiming();
+
+  d = 0.0;
+  v = 0.0; // FP_ZERO
+  for (int i = 0; i < iters; ++i) {
+    d += fpclassify(v);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_math_fpclassify_ZERO);

benchmarks/property_benchmark.cpp

@@ -15,6 +15,9 @@
  */
 
 #include "benchmark.h"
+#include <errno.h>
+#include <stdio.h>
+#include <stdlib.h>
 #include <unistd.h>
 
 #define _REALLY_INCLUDE_SYS__SYSTEM_PROPERTIES_H_
@@ -32,10 +35,17 @@ struct LocalPropertyTestState {
     LocalPropertyTestState(int nprops) : nprops(nprops), valid(false) {
         static const char prop_name_chars[] = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ.-_";
 
-        char dir_template[] = "/data/local/tmp/prop-XXXXXX";
+        const char* android_data = getenv("ANDROID_DATA");
+        if (android_data == NULL) {
+          printf("ANDROID_DATA environment variable not set\n");
+          return;
+        }
+        char dir_template[PATH_MAX];
+        snprintf(dir_template, sizeof(dir_template), "%s/local/tmp/prop-XXXXXX", android_data);
         char *dirname = mkdtemp(dir_template);
         if (!dirname) {
-            perror("making temp file for test state failed (is /data/local/tmp writable?)");
+            printf("making temp file for test state failed (is %s/local/tmp writable?): %s\n",
+                   android_data, strerror(errno));
             return;
         }
 

benchmarks/pthread_benchmark.cpp

@@ -0,0 +1,104 @@
+/*
+ * Copyright (C) 2014 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "benchmark.h"
+
+#include <pthread.h>
+
+static void BM_pthread_self(int iters) {
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+    pthread_self();
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_pthread_self);
+
+static void BM_pthread_getspecific(int iters) {
+  StopBenchmarkTiming();
+  pthread_key_t key;
+  pthread_key_create(&key, NULL);
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+    pthread_getspecific(key);
+  }
+
+  StopBenchmarkTiming();
+  pthread_key_delete(key);
+}
+BENCHMARK(BM_pthread_getspecific);
+
+static void DummyPthreadOnceInitFunction() {
+}
+
+static void BM_pthread_once(int iters) {
+  StopBenchmarkTiming();
+  pthread_once_t once = PTHREAD_ONCE_INIT;
+  pthread_once(&once, DummyPthreadOnceInitFunction);
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+    pthread_once(&once, DummyPthreadOnceInitFunction);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_pthread_once);
+
+static void BM_pthread_mutex_lock(int iters) {
+  StopBenchmarkTiming();
+  pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+    pthread_mutex_lock(&mutex);
+    pthread_mutex_unlock(&mutex);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_pthread_mutex_lock);
+
+static void BM_pthread_mutex_lock_ERRORCHECK(int iters) {
+  StopBenchmarkTiming();
+  pthread_mutex_t mutex = PTHREAD_ERRORCHECK_MUTEX_INITIALIZER;
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+    pthread_mutex_lock(&mutex);
+    pthread_mutex_unlock(&mutex);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_pthread_mutex_lock_ERRORCHECK);
+
+static void BM_pthread_mutex_lock_RECURSIVE(int iters) {
+  StopBenchmarkTiming();
+  pthread_mutex_t mutex = PTHREAD_RECURSIVE_MUTEX_INITIALIZER;
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+    pthread_mutex_lock(&mutex);
+    pthread_mutex_unlock(&mutex);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_pthread_mutex_lock_RECURSIVE);

benchmarks/semaphore_benchmark.cpp

@@ -0,0 +1,49 @@
+/*
+ * Copyright (C) 2014 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "benchmark.h"
+
+#include <semaphore.h>
+
+static void BM_semaphore_sem_getvalue(int iters) {
+  StopBenchmarkTiming();
+  sem_t semaphore;
+  sem_init(&semaphore, 1, 1);
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+    int dummy;
+    sem_getvalue(&semaphore, &dummy);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_semaphore_sem_getvalue);
+
+static void BM_semaphore_sem_wait_sem_post(int iters) {
+  StopBenchmarkTiming();
+  sem_t semaphore;
+  sem_init(&semaphore, 1, 1);
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+    sem_wait(&semaphore);
+    sem_post(&semaphore);
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_semaphore_sem_wait_sem_post);

benchmarks/stdio_benchmark.cpp

@@ -0,0 +1,61 @@
+/*
+ * Copyright (C) 2014 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "benchmark.h"
+
+#include <stdio.h>
+
+#define KB 1024
+#define MB 1024*KB
+
+#define AT_COMMON_SIZES \
+    Arg(1)->Arg(2)->Arg(3)->Arg(4)->Arg(8)->Arg(16)->Arg(32)->Arg(64)->Arg(512)-> \
+    Arg(1*KB)->Arg(4*KB)->Arg(8*KB)->Arg(16*KB)->Arg(64*KB)
+
+static void BM_stdio_fread(int iters, int chunk_size) {
+  StopBenchmarkTiming();
+  FILE* fp = fopen("/dev/zero", "rw");
+  char* buf = new char[chunk_size];
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+    fread(buf, chunk_size, 1, fp);
+  }
+
+  StopBenchmarkTiming();
+  SetBenchmarkBytesProcessed(int64_t(iters) * int64_t(chunk_size));
+  delete[] buf;
+  fclose(fp);
+}
+BENCHMARK(BM_stdio_fread)->AT_COMMON_SIZES;
+
+
+static void BM_stdio_fwrite(int iters, int chunk_size) {
+  StopBenchmarkTiming();
+  FILE* fp = fopen("/dev/zero", "rw");
+  char* buf = new char[chunk_size];
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+      fwrite(buf, chunk_size, 1, fp);
+  }
+
+  StopBenchmarkTiming();
+  SetBenchmarkBytesProcessed(int64_t(iters) * int64_t(chunk_size));
+  delete[] buf;
+  fclose(fp);
+}
+BENCHMARK(BM_stdio_fwrite)->AT_COMMON_SIZES;

benchmarks/unistd_benchmark.cpp

@@ -0,0 +1,30 @@
+/*
+ * Copyright (C) 2014 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "benchmark.h"
+
+#include <unistd.h>
+
+static void BM_unistd_getpid(int iters) {
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+    getpid();
+  }
+
+  StopBenchmarkTiming();
+}
+BENCHMARK(BM_unistd_getpid);

libc/SYSCALLS.TXT

@@ -1,326 +1,333 @@
-# This file is used to automatically generate bionic's the system calls stubs.
+# This file is used to automatically generate bionic's system call stubs.
 #
-# Each non comment line has the following format:
+# Each non-blank, non-comment line has the following format:
 #
-# return_type    func_name[:syscall_name[:call_id]]([parameter_list])  (1|-1|"stub")
+# return_type func_name[|alias_list][:syscall_name[:socketcall_id]]([parameter_list]) arch_list
 #
-# Note that:
+# where:
+#       arch_list ::= "all" | arch+
+#       arch      ::= "arm" | "arm64" | "mips" | "mips64" | "x86" | "x86_64"
+#
+# Note:
 #      - syscall_name corresponds to the name of the syscall, which may differ from
 #        the exported function name (example: the exit syscall is implemented by the _exit()
 #        function, which is not the same as the standard C exit() function which calls it)
-#        The call_id parameter, given that func_name and syscall_name have
+#
+#      - alias_list is optional comma separated list of function aliases.
+#
+#      - The call_id parameter, given that func_name and syscall_name have
 #        been provided, allows the user to specify dispatch style syscalls.
 #        For example, socket() syscall on i386 actually becomes:
 #          socketcall(__NR_socket, 1, *(rest of args on stack)).
 #
-#      - each parameter type is assumed to be stored on 32 bits, there is no plan to support
-#        64-bit architectures at the moment
-#
-#      - the final field can be "1", meaning: generate a stub for each architecture,
-#        taking the constants from the kernel header files.
-#
-#      - the final field can be "stub" meaning: do not generate any stubs ---
-#        in this case, a hand-written custom stub must be provided.
-#        TODO: replace this with something like "custom" or "none", or remove
-#        it entirely.
-#
-#      - the final field can be a three-element list of 1s and -1 meaning:
-#        this system call is only available on some of the architectures (1),
-#        and no stub should be generated for those architectures marked with -1.
-#        the order is arm,x86,mips.
-#        TODO: replace this with something more readable like "-arm,-mips" (meaning x86 only).
+#      - Each parameter type is assumed to be stored in 32 bits.
 #
 # This file is processed by a python script named gensyscalls.py.
-#
 
-# process management
-void    _exit:exit_group (int)      1
-void    _exit_thread:exit (int)     1
-pid_t   __fork:fork (void)           1
-pid_t   _waitpid:waitpid (pid_t, int*, int, struct rusage*)   -1,1,1
-int     __waitid:waitid(int, pid_t, struct siginfo_t*, int,void*)          1
-pid_t   wait4(pid_t pid, int *status, int options, struct rusage *rusage)   1
+int     execve(const char*, char* const*, char* const*)  all
 
-# NOTE: this system call is never called directly, but we list it there
-#       to have __NR_clone properly defined.
-#
-pid_t   __sys_clone:clone (int, void*, int*, void*, int*) 1
+uid_t   getuid:getuid32()         arm,x86
+uid_t   getuid:getuid()           arm64,mips,mips64,x86_64
+gid_t   getgid:getgid32()         arm,x86
+gid_t   getgid:getgid()           arm64,mips,mips64,x86_64
+uid_t   geteuid:geteuid32()       arm,x86
+uid_t   geteuid:geteuid()         arm64,mips,mips64,x86_64
+gid_t   getegid:getegid32()       arm,x86
+gid_t   getegid:getegid()         arm64,mips,mips64,x86_64
+uid_t   getresuid:getresuid32(uid_t* ruid, uid_t* euid, uid_t* suid)   arm,x86
+uid_t   getresuid:getresuid(uid_t* ruid, uid_t* euid, uid_t* suid)     arm64,mips,mips64,x86_64
+gid_t   getresgid:getresgid32(gid_t* rgid, gid_t* egid, gid_t* sgid)   arm,x86
+gid_t   getresgid:getresgid(gid_t* rgid, gid_t* egid, gid_t* sgid)     arm64,mips,mips64,x86_64
+pid_t   gettid()                   all
+ssize_t readahead(int, off64_t, size_t)     all
+int     getgroups:getgroups32(int, gid_t*)    arm,x86
+int     getgroups:getgroups(int, gid_t*)      arm64,mips,mips64,x86_64
+pid_t   getpgid(pid_t)             all
+pid_t   getppid()                  all
+pid_t   getsid(pid_t)              all
+pid_t   setsid()                   all
+int     setgid:setgid32(gid_t)     arm,x86
+int     setgid:setgid(gid_t)       arm64,mips,mips64,x86_64
+int     setuid:setuid32(uid_t)    arm,x86
+int     setuid:setuid(uid_t)      arm64,mips,mips64,x86_64
+int     setreuid:setreuid32(uid_t, uid_t)   arm,x86
+int     setreuid:setreuid(uid_t, uid_t)     arm64,mips,mips64,x86_64
+int     setresuid:setresuid32(uid_t, uid_t, uid_t)   arm,x86
+int     setresuid:setresuid(uid_t, uid_t, uid_t)     arm64,mips,mips64,x86_64
+int     setresgid:setresgid32(gid_t, gid_t, gid_t)   arm,x86
+int     setresgid:setresgid(gid_t, gid_t, gid_t)     arm64,mips,mips64,x86_64
+void*   __brk:brk(void*)           all
+int     kill(pid_t, int)           all
+int     tkill(pid_t tid, int sig)  all
+int     tgkill(pid_t tgid, pid_t tid, int sig)  all
+int     __ptrace:ptrace(int request, int pid, void* addr, void* data)  all
+int     __set_thread_area:set_thread_area(void*  user_desc)  mips,mips64,x86
 
-int     execve (const char*, char* const*, char* const*)  1
+# <sys/resource.h>
+int getrusage(int, struct rusage*)  all
+int __getpriority:getpriority(int, int)  all
+int setpriority(int, int, int)   all
+# On LP64, rlimit and rlimit64 are the same.
+# On 32-bit systems we use prlimit64 to implement the rlimit64 functions.
+int getrlimit:ugetrlimit(int, struct rlimit*)  arm,x86
+int getrlimit(int, struct rlimit*)  mips
+int getrlimit|getrlimit64(int, struct rlimit*)  arm64,mips64,x86_64
+int setrlimit(int, const struct rlimit*)  arm,mips,x86
+int setrlimit|setrlimit64(int, const struct rlimit*)  arm64,mips64,x86_64
+int prlimit64|prlimit(pid_t, int, struct rlimit64*, const struct rlimit64*)  arm64,mips64,x86_64
+int prlimit64(pid_t, int, struct rlimit64*, const struct rlimit64*)  arm,mips,x86
 
-int     __setuid:setuid32 (uid_t)    1,1,-1
-int     __setuid:setuid (uid_t)   -1,-1,1
-uid_t   getuid:getuid32 ()         1,1,-1
-uid_t   getuid:getuid ()           -1,-1,1
-gid_t   getgid:getgid32 ()         1,1,-1
-gid_t   getgid:getgid ()           -1,-1,1
-uid_t   geteuid:geteuid32 ()       1,1,-1
-uid_t   geteuid:geteuid ()         -1,-1,1
-gid_t   getegid:getegid32 ()       1,1,-1
-gid_t   getegid:getegid ()         -1,-1,1
-uid_t   getresuid:getresuid32 (uid_t *ruid, uid_t *euid, uid_t *suid)   1,1,-1
-uid_t   getresuid:getresuid (uid_t *ruid, uid_t *euid, uid_t *suid)     -1,-1,1
-gid_t   getresgid:getresgid32 (gid_t *rgid, gid_t *egid, gid_t *sgid)   1,1,-1
-gid_t   getresgid:getresgid (gid_t *rgid, gid_t *egid, gid_t *sgid)     -1,-1,1
-pid_t   gettid()                   1
-ssize_t readahead(int, off64_t, size_t)     1
-int     getgroups:getgroups32(int, gid_t *)    1,1,-1
-int     getgroups:getgroups(int, gid_t *)      -1,-1,1
-pid_t   getpgid(pid_t)             1
-pid_t   getppid()                  1
-pid_t   getsid(pid_t)              1
-pid_t   setsid()                   1
-int     setgid:setgid32(gid_t)     1,1,-1
-int     setgid:setgid(gid_t)       -1,-1,1
-int     seteuid:seteuid32(uid_t)   stub
-int     __setreuid:setreuid32(uid_t, uid_t)   1,1,-1
-int     __setreuid:setreuid(uid_t, uid_t)     -1,-1,1
-int     __setresuid:setresuid32(uid_t, uid_t, uid_t)   1,1,-1
-int     __setresuid:setresuid(uid_t, uid_t, uid_t)     -1,-1,1
-int     setresgid:setresgid32(gid_t, gid_t, gid_t)   1,1,-1
-int     setresgid:setresgid(gid_t, gid_t, gid_t)     -1,-1,1
-void*   __brk:brk(void*)           1
-# see comments in arch-arm/bionic/kill.S to understand why we don't generate an ARM stub for kill/tkill
-int     kill(pid_t, int)           -1,1,1
-int     tkill(pid_t tid, int sig)  -1,1,1
-int     tgkill(pid_t tgid, pid_t tid, int sig)  -1,1,1
-int     __ptrace:ptrace(int request, int pid, void* addr, void* data)  1
-int     __set_thread_area:set_thread_area(void*  user_desc)  -1,1,1
-int     __getpriority:getpriority(int, int)  1
-int     setpriority(int, int, int)   1
-int     setrlimit(int resource, const struct rlimit *rlp)  1
-int     getrlimit:ugetrlimit(int resource, struct rlimit *rlp)  1,1,-1
-int     getrlimit:getrlimit(int resource, struct rlimit *rlp)  -1,-1,1
-int     getrusage(int who, struct rusage*  r_usage)  1
-int     setgroups:setgroups32(int, const gid_t *)   1,1,-1
-int     setgroups:setgroups(int, const gid_t *)     -1,-1,1
-pid_t   getpgrp(void)  stub
-int     setpgid(pid_t, pid_t)  1
-pid_t   vfork(void)  1,-1,-1
-int     setregid:setregid32(gid_t, gid_t)  1,1,-1
-int     setregid:setregid(gid_t, gid_t)    -1,-1,1
-int     chroot(const char *)  1
-# IMPORTANT: Even though <sys/prctl.h> declares prctl(int,...), the syscall stub must take 6 arguments
+int     setgroups:setgroups32(int, const gid_t*)   arm,x86
+int     setgroups:setgroups(int, const gid_t*)     arm64,mips,mips64,x86_64
+int     setpgid(pid_t, pid_t)  all
+pid_t   vfork(void)  arm
+int     setregid:setregid32(gid_t, gid_t)  arm,x86
+int     setregid:setregid(gid_t, gid_t)    arm64,mips,mips64,x86_64
+int     chroot(const char*)  all
+# IMPORTANT: Even though <sys/prctl.h> declares prctl(int, ...), the syscall stub must take 6 arguments
 #            to match the kernel implementation.
-int     prctl(int option, unsigned int arg2, unsigned int arg3, unsigned int arg4, unsigned int arg5) 1
-int     capget(cap_user_header_t header, cap_user_data_t data) 1
-int     capset(cap_user_header_t header, const cap_user_data_t data) 1
-int     sigaltstack(const stack_t*, stack_t*) 1
-int     acct(const char*  filepath)  1
+int     prctl(int option, unsigned int arg2, unsigned int arg3, unsigned int arg4, unsigned int arg5) all
+long    __arch_prctl:arch_prctl(int, unsigned long) x86_64
+int     capget(cap_user_header_t header, cap_user_data_t data) all
+int     capset(cap_user_header_t header, const cap_user_data_t data) all
+int     sigaltstack(const stack_t*, stack_t*) all
+int     acct(const char*  filepath)  all
 
 # file descriptors
-ssize_t     read (int, void*, size_t)        1
-ssize_t     write (int, const void*, size_t)       1
-ssize_t     pread64 (int, void *, size_t, off64_t) 1
-ssize_t     pwrite64 (int, void *, size_t, off64_t) 1
-int         __open:open (const char*, int, mode_t)  1
-int         __openat:openat (int, const char*, int, mode_t) 1
-int         close (int)                      1
-int         creat(const char*, mode_t)       stub
-off_t       lseek(int, off_t, int)           1
-int         __llseek:_llseek (int, unsigned long, unsigned long, loff_t*, int)  1
-pid_t       getpid ()    1
-void *      mmap(void *, size_t, int, int, int, long)  stub
-void *      __mmap2:mmap2(void*, size_t, int, int, int, long)   1
-int         munmap(void *, size_t)  1
-void *      mremap(void *, size_t, size_t, unsigned long)  1
-int         msync(const void *, size_t, int)    1
-int         mprotect(const void *, size_t, int)  1
-int         madvise(const void *, size_t, int)  1
-int         mlock(const void *addr, size_t len)    1
-int         munlock(const void *addr, size_t len)   1
-int         mlockall(int flags)   1
-int         munlockall()   1
-int         mincore(void*  start, size_t  length, unsigned char*  vec)   1
-int         __ioctl:ioctl(int, int, void *)  1
-int         readv(int, const struct iovec *, int)   1
-int         writev(int, const struct iovec *, int)  1
-int         __fcntl:fcntl(int, int, void*)  1
-int         flock(int, int)   1
-int         fchmod(int, mode_t)  1
-int         dup(int)  1
-int         pipe(int *)  1,1,-1
-int         pipe2(int *, int) 1
-int         dup2(int, int)   1
-int         select:_newselect(int, struct fd_set *, struct fd_set *, struct fd_set *, struct timeval *)  1
-int         ftruncate(int, off_t)  1
-int         ftruncate64(int, off64_t) 1
-int         getdents:getdents64(unsigned int, struct dirent *, unsigned int)   1
-int         fsync(int)  1
-int         fdatasync(int) 1
-int         fchown:fchown32(int, uid_t, gid_t)  1,1,-1
-int         fchown:fchown(int, uid_t, gid_t)    -1,-1,1
-void        sync(void)  1
-int         __fcntl64:fcntl64(int, int, void *)  1
-int         __fstatfs64:fstatfs64(int, size_t, struct statfs *)  1
-ssize_t     sendfile(int out_fd, int in_fd, off_t *offset, size_t count)  1
-int         fstatat:fstatat64(int dirfd, const char *path, struct stat *buf, int flags)   1
-int         mkdirat(int dirfd, const char *pathname, mode_t mode)  1
-int         fchownat(int dirfd, const char *path, uid_t owner, gid_t group, int flags)  1
-int         fchmodat(int dirfd, const char *path, mode_t mode, int flags)  1
-int         renameat(int olddirfd, const char *oldpath, int newdirfd, const char *newpath)  1
-int         fsetxattr(int, const char *, const void *, size_t, int) 1
-ssize_t     fgetxattr(int, const char *, void *, size_t) 1
-ssize_t     flistxattr(int, char *, size_t) 1
-int         fremovexattr(int, const char *) 1
+ssize_t     read(int, void*, size_t)        all
+ssize_t     write(int, const void*, size_t)       all
+ssize_t     pread64(int, void*, size_t, off64_t) arm,mips,x86
+ssize_t     pread64|pread(int, void*, size_t, off_t) arm64,mips64,x86_64
+ssize_t     pwrite64(int, void*, size_t, off64_t) arm,mips,x86
+ssize_t     pwrite64|pwrite(int, void*, size_t, off_t) arm64,mips64,x86_64
+int         close(int)                      all
+pid_t       getpid()    all
+int         munmap(void*, size_t)  all
+void*       mremap(void*, size_t, size_t, unsigned long)  all
+int         msync(const void*, size_t, int)    all
+int         mprotect(const void*, size_t, int)  all
+int         madvise(const void*, size_t, int)  all
+int         mlock(const void* addr, size_t len)    all
+int         munlock(const void* addr, size_t len)   all
+int         mlockall(int flags)   all
+int         munlockall()   all
+int         mincore(void*  start, size_t  length, unsigned char*  vec)   all
+int         __ioctl:ioctl(int, int, void*)  all
+int         readv(int, const struct iovec*, int)   all
+int         writev(int, const struct iovec*, int)  all
+int         __fcntl64:fcntl64(int, int, void*)  arm,mips,x86
+int         fcntl(int, int, void*)  arm64,mips64,x86_64
+int         flock(int, int)   all
+int         fchmod(int, mode_t)  all
+int         dup(int)  all
+int         pipe2(int*, int) all
+int         dup3(int, int, int)   all
+int         fsync(int)  all
+int         fdatasync(int) all
+int         fchown:fchown32(int, uid_t, gid_t)  arm,x86
+int         fchown:fchown(int, uid_t, gid_t)    arm64,mips,mips64,x86_64
+void        sync(void)  all
+int         fsetxattr(int, const char*, const void*, size_t, int) all
+ssize_t     fgetxattr(int, const char*, void*, size_t) all
+ssize_t     flistxattr(int, char*, size_t) all
+int         fremovexattr(int, const char*) all
+
+# mips64 doesn't have getdents64 until 3.10 kernels.
+# We need this special-case hack as long as we need to support mips64 on older kernels.
+# The currently-available Debian qemu image is on a 3.2 kernel.
+int getdents:getdents64(unsigned int, struct dirent*, unsigned int)   arm,arm64,mips,x86,x86_64
+int __getdents64:getdents64(unsigned int, struct dirent*, unsigned int)   mips64
+int __getdents:getdents(unsigned int, void*, unsigned int)   mips64
+
+int __openat:openat(int, const char*, int, mode_t) all
+int faccessat(int, const char*, int, int)  all
+int fchmodat(int, const char*, mode_t, int)  all
+int fchownat(int, const char*, uid_t, gid_t, int)  all
+int fstatat64|fstatat:fstatat64(int, const char*, struct stat*, int)   arm,mips,x86
+int fstatat64|fstatat:newfstatat(int, const char*, struct stat*, int)  arm64,mips64,x86_64
+int linkat(int, const char*, int, const char*, int)  all
+int mkdirat(int, const char*, mode_t)  all
+int mknodat(int, const char*, mode_t, dev_t)  all
+int readlinkat(int, const char*, char*, size_t)  all
+int renameat(int, const char*, int, const char*)  all
+int symlinkat(const char*, int, const char*)  all
+int unlinkat(int, const char*, int)   all
+int utimensat(int, const char*, const struct timespec times[2], int)  all
+
+# Paired off_t/off64_t system calls. On 64-bit systems,
+# sizeof(off_t) == sizeof(off64_t), so there we emit two symbols that are
+# aliases. On 32-bit systems, we have two different system calls.
+# That means that every system call in this section should take three lines.
+off_t lseek(int, off_t, int) arm,mips,x86
+int __llseek:_llseek(int, unsigned long, unsigned long, off64_t*, int) arm,mips,x86
+off_t lseek|lseek64(int, off_t, int) arm64,mips64,x86_64
+int ftruncate(int, off_t) arm,mips,x86
+int ftruncate64(int, off64_t) arm,mips,x86
+int ftruncate|ftruncate64(int, off_t) arm64,mips64,x86_64
+ssize_t sendfile(int out_fd, int in_fd, off_t* offset, size_t count) arm,mips,x86
+ssize_t sendfile64(int out_fd, int in_fd, off64_t* offset, size_t count) arm,mips,x86
+ssize_t sendfile|sendfile64(int out_fd, int in_fd, off_t* offset, size_t count) arm64,mips64,x86_64
+int truncate(const char*, off_t) arm,mips,x86
+int truncate64(const char*, off64_t) arm,mips,x86
+int truncate|truncate64(const char*, off_t) arm64,mips64,x86_64
+# (mmap only gets two lines because we only used the 64-bit variant on 32-bit systems.)
+void* __mmap2:mmap2(void*, size_t, int, int, int, long)   arm,mips,x86
+void* mmap|mmap64(void*, size_t, int, int, int, off_t)  arm64,mips64,x86_64
+# (fallocate only gets two lines because there is no 32-bit variant.)
+int fallocate64:fallocate(int, int, off64_t, off64_t) arm,mips,x86
+int fallocate|fallocate64(int, int, off_t, off_t) arm64,mips64,x86_64
+
+int __fstatfs64:fstatfs64(int, size_t, struct statfs*)  arm,mips,x86
+int fstatfs64|fstatfs:fstatfs(int, struct statfs*)  arm64,mips64,x86_64
+int __statfs64:statfs64(const char*, size_t, struct statfs*)  arm,mips,x86
+int statfs64|statfs:statfs(const char*, struct statfs*)  arm64,mips64,x86_64
+
+int     fstat64|fstat:fstat64(int, struct stat*)    arm,mips,x86
+int     fstat64|fstat:fstat(int, struct stat*)    arm64,mips64,x86_64
 
 # file system
-int     link (const char*, const char*)  1
-int     unlink (const char*)             1
-int     unlinkat (int, const char *, int)   1
-int     chdir (const char*)              1
-int     mknod (const char*, mode_t, dev_t)  1
-int     chmod (const char*,mode_t)          1
-int     chown:chown32(const char *, uid_t, gid_t)  1,1,-1
-int     chown:chown(const char *, uid_t, gid_t)    -1,-1,1
-int     lchown:lchown32 (const char*, uid_t, gid_t)  1,1,-1
-int     lchown:lchown (const char*, uid_t, gid_t)  -1,-1,1
-int     mount (const char*, const char*, const char*, unsigned long, const void*)  1
-int     umount(const char*)  stub
-int     umount2 (const char*, int)  1
-int     fstat:fstat64(int, struct stat*)    1
-int     stat:stat64(const char *, struct stat *)  1
-int     lstat:lstat64(const char *, struct stat *)  1
-int     mkdir(const char *, mode_t) 1
-int     readlink(const char *, char *, size_t)  1
-int     rmdir(const char *)  1
-int     rename(const char *, const char *)  1
-int     __getcwd:getcwd(char * buf, size_t size)  1
-int     access(const char *, int)  1
-int     faccessat(int, const char *, int, int)  1
-int     symlink(const char *, const char *)  1
-int     fchdir(int)    1
-int     truncate(const char*, off_t)    1
-int     setxattr(const char *, const char *, const void *, size_t, int) 1
-int     lsetxattr(const char *, const char *, const void *, size_t, int) 1
-ssize_t getxattr(const char *, const char *, void *, size_t) 1
-ssize_t lgetxattr(const char *, const char *, void *, size_t) 1
-ssize_t listxattr(const char *, char *, size_t) 1
-ssize_t llistxattr(const char *, char *, size_t) 1
-int     removexattr(const char *, const char *) 1
-int     lremovexattr(const char *, const char *) 1
-int     __statfs64:statfs64(const char *, size_t, struct statfs *)  1
-long    unshare(unsigned long)  1
-int     swapon(const char *, int) 1
-int     swapoff(const char *) 1
+int     chdir(const char*)              all
+int     mount(const char*, const char*, const char*, unsigned long, const void*)  all
+int     umount2(const char*, int)  all
+int     __getcwd:getcwd(char* buf, size_t size)  all
+int     fchdir(int)    all
+int     setxattr(const char*, const char*, const void*, size_t, int) all
+int     lsetxattr(const char*, const char*, const void*, size_t, int) all
+ssize_t getxattr(const char*, const char*, void*, size_t) all
+ssize_t lgetxattr(const char*, const char*, void*, size_t) all
+ssize_t listxattr(const char*, char*, size_t) all
+ssize_t llistxattr(const char*, char*, size_t) all
+int     removexattr(const char*, const char*) all
+int     lremovexattr(const char*, const char*) all
+int     swapon(const char*, int) all
+int     swapoff(const char*) all
 
 # time
-int           pause ()                       1
-int           gettimeofday(struct timeval*, struct timezone*)       1
-int           settimeofday(const struct timeval*, const struct timezone*)   1
-clock_t       times(struct tms *)       1
-int           nanosleep(const struct timespec *, struct timespec *)   1
-int           clock_gettime(clockid_t clk_id, struct timespec *tp)    1
-int           clock_settime(clockid_t clk_id, const struct timespec *tp)  1
-int           clock_getres(clockid_t clk_id, struct timespec *res)   1
-int           clock_nanosleep(clockid_t clock_id, int flags, const struct timespec *req, struct timespec *rem)  1
-int           getitimer(int, const struct itimerval *)   1
-int           setitimer(int, const struct itimerval *, struct itimerval *)  1
-int           __timer_create:timer_create(clockid_t clockid, struct sigevent *evp, timer_t *timerid)    1
-int           __timer_settime:timer_settime(timer_t, int, const struct itimerspec*, struct itimerspec*) 1
-int           __timer_gettime:timer_gettime(timer_t, struct itimerspec*)                                1
-int           __timer_getoverrun:timer_getoverrun(timer_t)                                              1
-int           __timer_delete:timer_delete(timer_t)                                                      1
-int           utimes(const char*, const struct timeval tvp[2])                          1
-int           utimensat(int, const char *, const struct timespec times[2], int)         1
-int           timerfd_create(clockid_t, int)   1
-int           timerfd_settime(int, int, const struct itimerspec *, struct itimerspec *)   1
-int           timerfd_gettime(int, struct itimerspec *)   1
+int           gettimeofday(struct timeval*, struct timezone*)       all
+int           settimeofday(const struct timeval*, const struct timezone*)   all
+clock_t       times(struct tms*)       all
+int           nanosleep(const struct timespec*, struct timespec*)   all
+int           clock_gettime(clockid_t clk_id, struct timespec* tp)    all
+int           clock_settime(clockid_t clk_id, const struct timespec* tp)  all
+int           clock_getres(clockid_t clk_id, struct timespec* res)   all
+int           clock_nanosleep(clockid_t clock_id, int flags, const struct timespec* req, struct timespec* rem)  all
+int           getitimer(int, const struct itimerval*)   all
+int           setitimer(int, const struct itimerval*, struct itimerval*)  all
+int           __timer_create:timer_create(clockid_t clockid, struct sigevent* evp, __kernel_timer_t* timerid)    all
+int           __timer_settime:timer_settime(__kernel_timer_t, int, const struct itimerspec*, struct itimerspec*) all
+int           __timer_gettime:timer_gettime(__kernel_timer_t, struct itimerspec*)                                all
+int           __timer_getoverrun:timer_getoverrun(__kernel_timer_t)                                              all
+int           __timer_delete:timer_delete(__kernel_timer_t)                                                      all
+int           timerfd_create(clockid_t, int)   all
+int           timerfd_settime(int, int, const struct itimerspec*, struct itimerspec*)   all
+int           timerfd_gettime(int, struct itimerspec*)   all
 
 # signals
-int     sigaction(int, const struct sigaction *, struct sigaction *)  1
-int     sigprocmask(int, const sigset_t *, sigset_t *)  1
-int     __sigsuspend:sigsuspend(int unused1, int unused2, unsigned mask)  1,1,-1
-int     __sigsuspend:sigsuspend(const sigset_t *mask)  -1,-1,1
-int     __rt_sigaction:rt_sigaction (int sig, const struct sigaction *act, struct sigaction *oact, size_t sigsetsize)  1
-int     __rt_sigprocmask:rt_sigprocmask (int  how, const sigset_t *set, sigset_t *oset, size_t sigsetsize)  1
-int     __rt_sigtimedwait:rt_sigtimedwait(const sigset_t *set, struct siginfo_t  *info, struct timespec_t  *timeout, size_t  sigset_size)  1
-int     sigpending(sigset_t *)  1
-int     signalfd4(int fd, const sigset_t *mask, size_t sizemask, int flags)  1
+int     __sigaction:sigaction(int, const struct sigaction*, struct sigaction*)  arm,mips,x86
+int     __rt_sigaction:rt_sigaction(int, const struct sigaction*, struct sigaction*, size_t)  all
+int     __rt_sigpending:rt_sigpending(sigset_t*, size_t)  all
+int     __rt_sigprocmask:rt_sigprocmask(int, const sigset_t*, sigset_t*, size_t)  all
+int     __rt_sigsuspend:rt_sigsuspend(const sigset_t*, size_t)  all
+int     __rt_sigtimedwait:rt_sigtimedwait(const sigset_t*, struct siginfo_t*, struct timespec_t*, size_t)  all
+int     signalfd4(int, const sigset_t*, size_t, int)  all
 
 # sockets
-int           socket(int, int, int)              1,-1,1
-int           socketpair(int, int, int, int*)    1,-1,1
-int           bind(int, struct sockaddr *, int)  1,-1,1
-int           connect(int, struct sockaddr *, socklen_t)   1,-1,1
-int           listen(int, int)                   1,-1,1
-int           accept(int, struct sockaddr *, socklen_t *)  1,-1,1
-int           getsockname(int, struct sockaddr *, socklen_t *)  1,-1,1
-int           getpeername(int, struct sockaddr *, socklen_t *)  1,-1,1
-int           sendto(int, const void *, size_t, int, const struct sockaddr *, socklen_t)  1,-1,1
-int           recvfrom(int, void *, size_t, unsigned int, struct sockaddr *, socklen_t *)  1,-1,1
-int           shutdown(int, int)  1,-1,1
-int           setsockopt(int, int, int, const void *, socklen_t)  1,-1,1
-int           getsockopt(int, int, int, void *, socklen_t *)    1,-1,1
-int           sendmsg(int, const struct msghdr *, unsigned int)  1,-1,1
-int           recvmsg(int, struct msghdr *, unsigned int)   1,-1,1
+int           socket(int, int, int)              arm,arm64,mips,mips64,x86_64
+int           socketpair(int, int, int, int*)    arm,arm64,mips,mips64,x86_64
+int           bind(int, struct sockaddr*, int)  arm,arm64,mips,mips64,x86_64
+int           connect(int, struct sockaddr*, socklen_t)   arm,arm64,mips,mips64,x86_64
+int           listen(int, int)                   arm,arm64,mips,mips64,x86_64
+int           accept(int, struct sockaddr*, socklen_t*)  arm,arm64,mips,mips64,x86_64
+int           accept4(int, struct sockaddr*, socklen_t*, int)  arm,arm64,mips,mips64,x86_64
+int           getsockname(int, struct sockaddr*, socklen_t*)  arm,arm64,mips,mips64,x86_64
+int           getpeername(int, struct sockaddr*, socklen_t*)  arm,arm64,mips,mips64,x86_64
+int           sendto(int, const void*, size_t, int, const struct sockaddr*, socklen_t)  arm,arm64,mips,mips64,x86_64
+int           recvfrom(int, void*, size_t, unsigned int, struct sockaddr*, socklen_t*)  arm,arm64,mips,mips64,x86_64
+int           shutdown(int, int)  arm,arm64,mips,mips64,x86_64
+int           setsockopt(int, int, int, const void*, socklen_t)  arm,arm64,mips,mips64,x86_64
+int           getsockopt(int, int, int, void*, socklen_t*)    arm,arm64,mips,mips64,x86_64
+int           sendmsg(int, const struct msghdr*, unsigned int)  arm,arm64,mips,mips64,x86_64
+int           recvmsg(int, struct msghdr*, unsigned int)   arm,arm64,mips,mips64,x86_64
+int           recvmmsg(int, struct mmsghdr*, unsigned int, int, const struct timespec*)   arm,arm64,mips,mips64,x86_64
+int           sendmmsg(int, struct mmsghdr*, unsigned int, int)   arm,arm64,mips,mips64,x86_64
 
 # sockets for x86. These are done as an "indexed" call to socketcall syscall.
-int           socket:socketcall:1 (int, int, int) -1,1,-1
-int           bind:socketcall:2 (int, struct sockaddr *, int)  -1,1,-1
-int           connect:socketcall:3(int, struct sockaddr *, socklen_t)   -1,1,-1
-int           listen:socketcall:4(int, int)                   -1,1,-1
-int           accept:socketcall:5(int, struct sockaddr *, socklen_t *)  -1,1,-1
-int           getsockname:socketcall:6(int, struct sockaddr *, socklen_t *)  -1,1,-1
-int           getpeername:socketcall:7(int, struct sockaddr *, socklen_t *)  -1,1,-1
-int           socketpair:socketcall:8(int, int, int, int*)    -1,1,-1
-int           sendto:socketcall:11(int, const void *, size_t, int, const struct sockaddr *, socklen_t)  -1,1,-1
-int           recvfrom:socketcall:12(int, void *, size_t, unsigned int, struct sockaddr *, socklen_t *)  -1,1,-1
-int           shutdown:socketcall:13(int, int)  -1,1,-1
-int           setsockopt:socketcall:14(int, int, int, const void *, socklen_t)  -1,1,-1
-int           getsockopt:socketcall:15(int, int, int, void *, socklen_t *)    -1,1,-1
-int           sendmsg:socketcall:16(int, const struct msghdr *, unsigned int)  -1,1,-1
-int           recvmsg:socketcall:17(int, struct msghdr *, unsigned int)   -1,1,-1
+int           socket:socketcall:1(int, int, int) x86
+int           bind:socketcall:2(int, struct sockaddr*, int)  x86
+int           connect:socketcall:3(int, struct sockaddr*, socklen_t)   x86
+int           listen:socketcall:4(int, int)                   x86
+int           accept:socketcall:5(int, struct sockaddr*, socklen_t*)  x86
+int           getsockname:socketcall:6(int, struct sockaddr*, socklen_t*)  x86
+int           getpeername:socketcall:7(int, struct sockaddr*, socklen_t*)  x86
+int           socketpair:socketcall:8(int, int, int, int*)    x86
+int           sendto:socketcall:11(int, const void*, size_t, int, const struct sockaddr*, socklen_t)  x86
+int           recvfrom:socketcall:12(int, void*, size_t, unsigned int, struct sockaddr*, socklen_t*)  x86
+int           shutdown:socketcall:13(int, int)  x86
+int           setsockopt:socketcall:14(int, int, int, const void*, socklen_t)  x86
+int           getsockopt:socketcall:15(int, int, int, void*, socklen_t*)    x86
+int           sendmsg:socketcall:16(int, const struct msghdr*, unsigned int)  x86
+int           recvmsg:socketcall:17(int, struct msghdr*, unsigned int)   x86
+int           accept4:socketcall:18(int, struct sockaddr*, socklen_t*, int)  x86
+int           recvmmsg:socketcall:19(int, struct mmsghdr*, unsigned int, int, const struct timespec*)   x86
+int           sendmmsg:socketcall:20(int, struct mmsghdr*, unsigned int, int)   x86
 
 # scheduler & real-time
-int sched_setscheduler(pid_t pid, int policy, const struct sched_param *param)  1
-int sched_getscheduler(pid_t pid)  1
-int sched_yield(void)  1
-int sched_setparam(pid_t pid, const struct sched_param *param)  1
-int sched_getparam(pid_t pid, struct sched_param *param)  1
-int sched_get_priority_max(int policy)  1
-int sched_get_priority_min(int policy)  1
-int sched_rr_get_interval(pid_t pid, struct timespec *interval)  1
-int sched_setaffinity(pid_t pid, size_t setsize, const cpu_set_t* set) 1
-int __sched_getaffinity:sched_getaffinity(pid_t pid, size_t setsize, cpu_set_t* set)  1
-int __getcpu:getcpu(unsigned *cpu, unsigned *node, void *unused) 1
+int sched_setscheduler(pid_t pid, int policy, const struct sched_param* param)  all
+int sched_getscheduler(pid_t pid)  all
+int sched_yield(void)  all
+int sched_setparam(pid_t pid, const struct sched_param* param)  all
+int sched_getparam(pid_t pid, struct sched_param* param)  all
+int sched_get_priority_max(int policy)  all
+int sched_get_priority_min(int policy)  all
+int sched_rr_get_interval(pid_t pid, struct timespec* interval)  all
+int sched_setaffinity(pid_t pid, size_t setsize, const cpu_set_t* set) all
+int setns(int, int) all
+int unshare(int) all
+int __sched_getaffinity:sched_getaffinity(pid_t pid, size_t setsize, cpu_set_t* set)  all
+int __getcpu:getcpu(unsigned*, unsigned*, void*) all
 
 # io priorities
-int ioprio_set(int which, int who, int ioprio) 1
-int ioprio_get(int which, int who) 1
+int ioprio_set(int which, int who, int ioprio) all
+int ioprio_get(int which, int who) all
 
 # other
-int     uname(struct utsname *)  1
-mode_t  umask(mode_t)  1
-int      __reboot:reboot(int, int, int, void *)  1
-int     __syslog:syslog(int, char *, int)  1
-int     init_module(void *, unsigned long, const char *)  1
-int     delete_module(const char*, unsigned int)   1
-int     klogctl:syslog(int, char *, int)   1
-int     sysinfo(struct sysinfo *)  1
-int     personality(unsigned long)  1
-long    perf_event_open(struct perf_event_attr *attr_uptr, pid_t pid, int cpu, int group_fd, unsigned long flags) 1
+int     uname(struct utsname*)  all
+mode_t  umask(mode_t)  all
+int     __reboot:reboot(int, int, int, void*)  all
+int     __syslog:syslog(int, char*, int)  all
+int     init_module(void*, unsigned long, const char*)  all
+int     delete_module(const char*, unsigned int)   all
+int     klogctl:syslog(int, char*, int)   all
+int     sysinfo(struct sysinfo*)  all
+int     personality(unsigned long)  all
+long    perf_event_open(struct perf_event_attr* attr_uptr, pid_t pid, int cpu, int group_fd, unsigned long flags) all
 
-# futex
-int	futex(void *, int, int, void *, void *, int) 1
+int epoll_create1(int)  all
+int epoll_ctl(int, int op, int, struct epoll_event*)  all
+int __epoll_pwait:epoll_pwait(int, struct epoll_event*, int, int, const sigset_t*, size_t)  all
 
-# epoll
-int     epoll_create(int size)     1
-int     epoll_ctl(int epfd, int op, int fd, struct epoll_event *event)    1
-int     epoll_wait(int epfd, struct epoll_event *events, int max, int timeout)   1
+int eventfd:eventfd2(unsigned int, int)  all
 
-int     inotify_init(void)      1
-int     inotify_add_watch(int, const char *, unsigned int)  1
-int     inotify_rm_watch(int, unsigned int)  1
+void _exit|_Exit:exit_group(int)  all
+void __exit:exit(int)  all
 
-int     poll(struct pollfd *, unsigned int, long)  1
+int futex(void*, int, int, void*, void*, int)  all
 
-int     eventfd:eventfd2(unsigned int, int)  1
+int inotify_init1(int)  all
+int inotify_add_watch(int, const char*, unsigned int)  all
+int inotify_rm_watch(int, unsigned int)  all
 
-# ARM-specific ARM_NR_BASE == 0x0f0000 == 983040
-int     __set_tls:__ARM_NR_set_tls(void*)                                 1,-1,-1
-int     cacheflush:__ARM_NR_cacheflush(long start, long end, long flags)  1,-1,-1
+int __pselect6:pselect6(int, fd_set*, fd_set*, fd_set*, timespec*, void*)  all
+int __ppoll:ppoll(pollfd*, unsigned int, timespec*, const sigset_t*, size_t)  all
+
+int __set_tid_address:set_tid_address(int*)  all
+
+pid_t wait4(pid_t, int*, int, struct rusage*)  all
+int __waitid:waitid(int, pid_t, struct siginfo_t*, int, void*)  all
+
+# ARM-specific
+int     __set_tls:__ARM_NR_set_tls(void*)                                 arm
+int     cacheflush:__ARM_NR_cacheflush(long start, long end, long flags)  arm
 
 # MIPS-specific
-int     _flush_cache:cacheflush(char *addr, const int nbytes, const int op) -1,-1,1
-int     syscall(int number,...) -1,-1,1
+int     _flush_cache:cacheflush(char* addr, const int nbytes, const int op) mips,mips64

libc/arch-arm/arm.mk

@@ -1,13 +1,59 @@
-_LIBC_ARCH_COMMON_SRC_FILES := \
+# arm specific configs
+
+# These are used by the 32-bit targets, but not the 64-bit ones.
+libc_common_src_files_arm := \
+    bionic/legacy_32_bit_support.cpp \
+    bionic/ndk_cruft.cpp \
+    bionic/time64.c \
+
+# These are shared by all the 32-bit targets, but not the 64-bit ones.
+libc_bionic_src_files_arm := \
+    bionic/mmap.cpp
+
+libc_common_src_files_arm += \
+    bionic/index.cpp \
+    bionic/memchr.c \
+    bionic/memmove.c.arm \
+    bionic/memrchr.c \
+    bionic/strchr.cpp \
+    bionic/strnlen.c \
+    bionic/strrchr.cpp \
+    upstream-freebsd/lib/libc/string/wcscat.c \
+    upstream-freebsd/lib/libc/string/wcschr.c \
+    upstream-freebsd/lib/libc/string/wcscmp.c \
+    upstream-freebsd/lib/libc/string/wcscpy.c \
+    upstream-freebsd/lib/libc/string/wcslen.c \
+    upstream-freebsd/lib/libc/string/wcsrchr.c \
+    upstream-freebsd/lib/libc/string/wmemcmp.c \
+    upstream-openbsd/lib/libc/string/bcopy.c \
+    upstream-openbsd/lib/libc/string/strlcat.c \
+    upstream-openbsd/lib/libc/string/strlcpy.c \
+    upstream-openbsd/lib/libc/string/strncat.c \
+    upstream-openbsd/lib/libc/string/strncmp.c \
+    upstream-openbsd/lib/libc/string/strncpy.c \
+
+# The C++ fortify function implementations for which there is an
+# arm assembler version.
+#
+# Fortify implementations of libc functions.
+# libc_common_src_files_arm +=
+#    bionic/__memcpy_chk.cpp \
+#    bionic/__memset_chk.cpp \
+#    bionic/__strcpy_chk.cpp \
+#    bionic/__strcat_chk.cpp \
+
+libc_common_cflags_arm := -DSOFTFLOAT
+
+##########################################
+### CPU specific source files
+libc_bionic_src_files_arm += \
     arch-arm/bionic/abort_arm.S \
     arch-arm/bionic/atomics_arm.c \
-    arch-arm/bionic/clone.S \
+    arch-arm/bionic/__bionic_clone.S \
     arch-arm/bionic/eabi.c \
     arch-arm/bionic/_exit_with_stack_teardown.S \
-    arch-arm/bionic/ffs.S \
     arch-arm/bionic/futex_arm.S \
     arch-arm/bionic/__get_sp.S \
-    arch-arm/bionic/kill.S \
     arch-arm/bionic/libgcc_compat.c \
     arch-arm/bionic/memcmp16.S \
     arch-arm/bionic/memcmp.S \
@@ -15,31 +61,32 @@ _LIBC_ARCH_COMMON_SRC_FILES := \
     arch-arm/bionic/setjmp.S \
     arch-arm/bionic/sigsetjmp.S \
     arch-arm/bionic/syscall.S \
-    arch-arm/bionic/tgkill.S \
-    arch-arm/bionic/tkill.S \
 
-# These are used by the static and dynamic versions of the libc
-# respectively.
-_LIBC_ARCH_STATIC_SRC_FILES := \
-    arch-arm/bionic/exidx_static.c \
-    bionic/dl_iterate_phdr_static.c \
+libc_arch_static_src_files_arm := arch-arm/bionic/exidx_static.c
+libc_arch_dynamic_src_files_arm := arch-arm/bionic/exidx_dynamic.c
 
-_LIBC_ARCH_DYNAMIC_SRC_FILES := \
-    arch-arm/bionic/exidx_dynamic.c \
-
-# Remove the C++ fortify function implementations for which there is an
-# arm assembler version.
-_LIBC_FORTIFY_FILES_TO_REMOVE := \
-    bionic/__memcpy_chk.cpp \
-    bionic/__memset_chk.cpp \
-    bionic/__strcpy_chk.cpp \
-    bionic/__strcat_chk.cpp \
-
-libc_common_src_files := \
-    $(filter-out $(_LIBC_FORTIFY_FILES_TO_REMOVE),$(libc_common_src_files))
-
-ifeq ($(strip $(wildcard bionic/libc/arch-arm/$(TARGET_CPU_VARIANT)/$(TARGET_CPU_VARIANT).mk)),)
-$(error "TARGET_CPU_VARIANT not set or set to an unknown value. Possible values are cortex-a7, cortex-a8, cortex-a9, cortex-a15, krait. Use generic for devices that do not have a CPU similar to any of the supported cpu variants.")
+## CPU variant specific source files
+ifeq ($(strip $(TARGET_$(my_2nd_arch_prefix)CPU_VARIANT)),)
+  $(warning TARGET_$(my_2nd_arch_prefix)ARCH is arm, but TARGET_$(my_2nd_arch_prefix)CPU_VARIANT is not defined)
 endif
+cpu_variant_mk := $(LOCAL_PATH)/arch-arm/$(TARGET_$(my_2nd_arch_prefix)CPU_VARIANT)/$(TARGET_$(my_2nd_arch_prefix)CPU_VARIANT).mk
+ifeq ($(wildcard $(cpu_variant_mk)),)
+$(error "TARGET_$(my_2nd_arch_prefix)CPU_VARIANT not set or set to an unknown value. Possible values are cortex-a7, cortex-a8, cortex-a9, cortex-a15, krait, denver. Use generic for devices that do not have a CPU similar to any of the supported cpu variants.")
+endif
+include $(cpu_variant_mk)
+libc_common_additional_dependencies += $(cpu_variant_mk)
 
-include bionic/libc/arch-arm/$(TARGET_CPU_VARIANT)/$(TARGET_CPU_VARIANT).mk
+cpu_variant_mk :=
+
+
+libc_crt_target_cflags_arm := \
+    -I$(LOCAL_PATH)/arch-arm/include \
+    -mthumb-interwork
+
+libc_crt_target_so_cflags_arm :=
+
+libc_crt_target_crtbegin_file_arm := \
+    $(LOCAL_PATH)/arch-common/bionic/crtbegin.c
+
+libc_crt_target_crtbegin_so_file_arm := \
+    $(LOCAL_PATH)/arch-common/bionic/crtbegin_so.c

libc/arch-arm/bionic/__bionic_clone.S

@@ -0,0 +1,73 @@
+/*
+ * Copyright (C) 2008-2010 The Android Open Source Project
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *  * Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  * Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+ * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
+ * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
+ * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
+ * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
+ * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include <private/bionic_asm.h>
+
+// pid_t __bionic_clone(int flags, void* child_stack, pid_t* parent_tid, void* tls, pid_t* child_tid, int (*fn)(void*), void* arg);
+ENTRY(__bionic_clone)
+    mov     ip, sp
+    # save registers to parent stack
+    stmfd   sp!, {r4, r5, r6, r7}
+    .cfi_def_cfa_offset 16
+    .cfi_rel_offset r4, 0
+    .cfi_rel_offset r5, 4
+    .cfi_rel_offset r6, 8
+    .cfi_rel_offset r7, 12
+
+    # load extra parameters
+    ldmfd   ip, {r4, r5, r6}
+
+    # store 'fn' and 'arg' to the child stack
+    str     r5, [r1, #-4]
+    str     r6, [r1, #-8]
+
+    # System call
+    ldr     r7, =__NR_clone
+    swi     #0
+    movs    r0, r0
+    beq     1f
+
+    # In the parent, reload saved registers then either return or set errno.
+    ldmfd   sp!, {r4, r5, r6, r7}
+    .cfi_def_cfa_offset 0
+    cmn     r0, #(MAX_ERRNO + 1)
+    bxls    lr
+    neg     r0, r0
+    b       __set_errno
+
+1:  # The child.
+    # Re-add the unwind directives that were reset from above.
+    .cfi_def_cfa_offset 16
+    .cfi_rel_offset r4, 0
+    .cfi_rel_offset r5, 4
+    .cfi_rel_offset r6, 8
+    .cfi_rel_offset r7, 12
+    ldr    r0, [sp, #-4]
+    ldr    r1, [sp, #-8]
+    b      __bionic_clone_entry
+END(__bionic_clone)

libc/arch-arm/bionic/__get_sp.S

@@ -26,7 +26,7 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 ENTRY(__get_sp)
   mov r0, sp

libc/arch-arm/bionic/_exit_with_stack_teardown.S

@@ -26,19 +26,16 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
-#include <asm/unistd.h>
+#include <private/bionic_asm.h>
 
-// void _exit_with_stack_teardown(void* stackBase, int stackSize, int retCode)
+// void _exit_with_stack_teardown(void* stackBase, size_t stackSize)
 ENTRY(_exit_with_stack_teardown)
-    mov     lr, r2
-    ldr     r7, =__NR_munmap
-    swi     #0              // the stack is destroyed by this call
-    mov     r0, lr
-    ldr     r7, =__NR_exit
-    swi     #0
+  ldr r7, =__NR_munmap
+  swi #0
+  // If munmap failed, we ignore the failure and exit anyway.
 
-    // exit() should never return, cause a crash if it does
-    mov     r0, #0
-    ldr     r0, [r0]
+  mov r0, #0
+  ldr r7, =__NR_exit
+  swi #0
+  // The exit syscall does not return.
 END(_exit_with_stack_teardown)

libc/arch-arm/bionic/_setjmp.S

@@ -34,7 +34,7 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 #include <machine/setjmp.h>
 #include <machine/cpu-features.h>
 
@@ -107,7 +107,7 @@ ENTRY(_longjmp)
 
 	/* validation failed, die die die. */
 botch:
-	bl	PIC_SYM(_C_LABEL(longjmperror), PLT)
-	bl	PIC_SYM(_C_LABEL(abort), PLT)
+	bl	PIC_SYM(longjmperror, PLT)
+	bl	PIC_SYM(abort, PLT)
 	b	. - 8		/* Cannot get here */
 END(_longjmp)

libc/arch-arm/bionic/abort_arm.S

@@ -26,7 +26,7 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 /*
  * Coding the abort function in assembly so that registers are guaranteed to
@@ -36,7 +36,9 @@
  * sequence when the crash happens.
  */
 ENTRY(abort)
-    .save   {r3, r14}
     stmfd   sp!, {r3, r14}
-    bl      PIC_SYM(_C_LABEL(__libc_android_abort), PLT)
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r3, 0
+    .cfi_rel_offset r14, 4
+    bl      PIC_SYM(__libc_android_abort, PLT)
 END(abort)

libc/arch-arm/bionic/atexit.h

@@ -1,35 +0,0 @@
-/*
- * Copyright (C) 2012 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-extern void *__dso_handle;
-
-__attribute__ ((visibility ("hidden")))
-int atexit(void (*func)(void))
-{
-  return (__cxa_atexit((void (*)(void *))func, (void *)0, &__dso_handle));
-}

libc/arch-arm/bionic/atexit_legacy.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012 The Android Open Source Project
+ * Copyright (C) 2012 The Android Open Source Project
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -27,9 +27,10 @@
  */
 
 #include <sys/types.h>
-#include <private/libc_logging.h>
 #include <stdio.h>
 
+#include "private/libc_logging.h"
+
 /*
  * This source file should only be included by libc.so, its purpose is
  * to support legacy ARM binaries by exporting a publicly visible

libc/arch-arm/bionic/clone.S

@@ -1,111 +0,0 @@
-/*
- * Copyright (C) 2008-2010 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-#include <linux/err.h>
-#include <machine/asm.h>
-#include <asm/unistd.h>
-
-// int  __pthread_clone(void* (*fn)(void*), void* child_stack, int flags, void* arg);
-ENTRY(__pthread_clone)
-    # Push 'fn' and 'arg' onto 'child_stack'.
-    stmdb   r1!, {r0, r3}
-
-    # The sys_clone system call only takes two arguments: 'flags' and 'child_stack'.
-    # 'child_stack' is already in r1, but we need to move 'flags' into position.
-    mov     r0, r2
-
-    # System call.
-    mov     ip, r7
-    ldr     r7, =__NR_clone
-    swi     #0
-
-    # Child?
-    movs    r0, r0
-    beq     1f
-
-    # Parent.
-    mov     r7, ip
-    cmn     r0, #(MAX_ERRNO + 1)
-    bxls    lr
-    neg     r0, r0
-    b       __set_errno
-
-1:  # Child.
-    # Pop 'fn' and 'arg' back off the stack and call __thread_entry.
-    pop     {r0, r1}
-    # __thread_entry also needs our stack pointer.
-    mov     r2, sp
-    b       __thread_entry
-END(__pthread_clone)
-
-
-    #
-    # This function is defined as:
-    #
-    #   pid_t  __bionic_clone( int  flags, void *child_stack,
-    #                          pid_t *pid, void *tls, pid_t *ctid,
-    #                          int  (*fn)(void *), void* arg );
-    #
-    # NOTE: This is not the same signature as the glibc
-    #       __clone function. Placing 'fn' and 'arg'
-    #       at the end of the parameter list makes the
-    #       implementation much simpler.
-    #
-
-ENTRY(__bionic_clone)
-    mov     ip, sp
-    .save   {r4, r5, r6, r7}
-
-    # save registers to parent stack
-    stmfd   sp!, {r4, r5, r6, r7}
-
-    # load extra parameters
-    ldmfd   ip, {r4, r5, r6}
-
-    # store 'fn' and 'arg' to the child stack
-    str     r5, [r1, #-4]
-    str     r6, [r1, #-8]
-
-    # System call
-    ldr     r7, =__NR_clone
-    swi     #0
-    movs    r0, r0
-    beq     1f
-
-    # In the parent, reload saved registers then either return or set errno.
-    ldmfd   sp!, {r4, r5, r6, r7}
-    cmn     r0, #(MAX_ERRNO + 1)
-    bxls    lr
-    neg     r0, r0
-    b       __set_errno
-
-1:  # The child.
-    ldr    r0, [sp, #-4]
-    ldr    r1, [sp, #-8]
-    b      __bionic_clone_entry
-END(__bionic_clone)

libc/arch-arm/bionic/crtbegin.c

@@ -1,53 +0,0 @@
-/*
- * Copyright (C) 2012 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-#include "../../bionic/libc_init_common.h"
-#include <stddef.h>
-#include <stdint.h>
-
-__attribute__ ((section (".preinit_array")))
-void (*__PREINIT_ARRAY__)(void) = (void (*)(void)) -1;
-
-__attribute__ ((section (".init_array")))
-void (*__INIT_ARRAY__)(void) = (void (*)(void)) -1;
-
-__attribute__ ((section (".fini_array")))
-void (*__FINI_ARRAY__)(void) = (void (*)(void)) -1;
-
-__LIBC_HIDDEN__ void _start() {
-  structors_array_t array;
-  array.preinit_array = &__PREINIT_ARRAY__;
-  array.init_array = &__INIT_ARRAY__;
-  array.fini_array = &__FINI_ARRAY__;
-
-  void* raw_args = (void*) ((uintptr_t) __builtin_frame_address(0) + sizeof(void*));
-  __libc_init(raw_args, NULL, &main, &array);
-}
-
-#include "__dso_handle.h"
-#include "atexit.h"

libc/arch-arm/bionic/crtbegin_so.c

@@ -1,58 +0,0 @@
-/*
- * Copyright (C) 2012 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-extern void __cxa_finalize(void *);
-extern void *__dso_handle;
-
-__attribute__((visibility("hidden"),destructor))
-void __on_dlclose() {
-  __cxa_finalize(&__dso_handle);
-}
-
-/* CRT_LEGACY_WORKAROUND should only be defined when building
- * this file as part of the platform's C library.
- *
- * The C library already defines a function named 'atexit()'
- * for backwards compatibility with older NDK-generated binaries.
- *
- * For newer ones, 'atexit' is actually embedded in the C
- * runtime objects that are linked into the final ELF
- * binary (shared library or executable), and will call
- * __cxa_atexit() in order to un-register any atexit()
- * handler when a library is unloaded.
- *
- * This function must be global *and* hidden. Only the
- * code inside the same ELF binary should be able to access it.
- */
-
-#ifdef CRT_LEGACY_WORKAROUND
-#include "__dso_handle.h"
-#else
-#include "__dso_handle_so.h"
-#include "atexit.h"
-#endif

libc/arch-arm/bionic/crtend.S

@@ -1,40 +0,0 @@
-/*
- * Copyright (C) 2008 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-	
-	.section .preinit_array, "aw"
-	.long 0
-
-	.section .init_array, "aw"
-	.long 0
-
-	.section .fini_array, "aw"
-	.long 0
-
-#if defined(__linux__) && defined(__ELF__)
-	.section .note.GNU-stack,"",%progbits
-#endif

libc/arch-arm/bionic/crtend_so.S

@@ -1,31 +0,0 @@
-/*
- * Copyright (C) 2010 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-#if defined(__linux__) && defined(__ELF__)
-	.section .note.GNU-stack,"",%progbits
-#endif

libc/arch-arm/bionic/ffs.S

@@ -1,41 +0,0 @@
-/*	$NetBSD: ffs.S,v 1.5 2003/04/05 23:08:52 bjh21 Exp $	*/
-/*
- * Copyright (c) 2001 Christopher Gilbert
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- * 1. Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in the
- *    documentation and/or other materials provided with the distribution.
- * 3. The name of the company nor the name of the author may be used to
- *    endorse or promote products derived from this software without specific
- *    prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
- * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
- * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
- * IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
- * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
- * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
- * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
- * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
- * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
- * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-#include <machine/asm.h>
-#include <machine/cpu-features.h>
-
-ENTRY(ffs)
-	/* Standard trick to isolate bottom bit in r0 or 0 if r0 = 0 on entry */
- 	rsb     r1, r0, #0
- 	ands    r0, r0, r1
-	clzne	r0, r0
-	rsbne	r0, r0, #32
-	bx		lr
-END(ffs)

libc/arch-arm/bionic/futex_arm.S

@@ -26,13 +26,12 @@
  * SUCH DAMAGE.
  */
 
-#include <asm/unistd.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 #define FUTEX_WAIT 0
 #define FUTEX_WAKE 1
 
-// __futex_syscall3(*ftx, op, val)
+// int __futex_syscall3(volatile void* ftx, int op, int count)
 ENTRY(__futex_syscall3)
     mov     ip, r7
     ldr     r7, =__NR_futex
@@ -41,12 +40,12 @@ ENTRY(__futex_syscall3)
     bx      lr
 END(__futex_syscall3)
 
-// __futex_syscall4(*ftx, op, val, *timespec)
+// int __futex_syscall4(volatile void* ftx, int op, int val, const struct timespec* timeout)
 ENTRY(__futex_syscall4)
     b __futex_syscall3
 END(__futex_syscall4)
 
-// __futex_wait(*ftx, val, *timespec)
+// int __futex_wait(volatile void* ftx, int val, const struct timespec* timeout)
 ENTRY(__futex_wait)
     mov     ip, r7
     mov     r3, r2
@@ -58,7 +57,7 @@ ENTRY(__futex_wait)
     bx      lr
 END(__futex_wait)
 
-// __futex_wake(*ftx, counter)
+// int __futex_wake(volatile void* ftx, int count)
 ENTRY(__futex_wake)
     mov     ip, r7
     mov     r2, r1

libc/arch-arm/bionic/kill.S

@@ -1,50 +0,0 @@
-/*
- * Copyright (C) 2008 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-#include <linux/err.h>
-#include <asm/unistd.h>
-#include <machine/asm.h>
-
-/* unlike our auto-generated syscall stubs, this code saves lr
-   on the stack, as well as a few other registers. this makes
-   our stack unwinder happy, when we generate debug stack
-   traces after the C library or other parts of the system
-   abort due to a fatal runtime error (e.g. detection
-   of a corrupted malloc heap).
-*/
-
-ENTRY(kill)
-    stmfd   sp!, {r4-r7, ip, lr}
-    ldr     r7, =__NR_kill
-    swi     #0
-    ldmfd   sp!, {r4-r7, ip, lr}
-    cmn     r0, #(MAX_ERRNO + 1)
-    bxls    lr
-    neg     r0, r0
-    b       __set_errno
-END(kill)

libc/arch-arm/bionic/libgcc_compat.c

@@ -1,170 +1,159 @@
-/*
- * Copyright (C) 2008 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
+/* Generated by genlibgcc_compat.py */
 
-/* This file contains dummy references to libgcc.a functions to force the
- * dynamic linker to copy their definition into the final libc.so binary.
- *
- * They are required to ensure backwards binary compatibility with
- * libc.so provided by the platform and binaries built with the NDK or
- * different versions/configurations of toolchains.
- *
- * Now, for a more elaborate description of the issue:
- *
- * libgcc.a is a compiler-specific library containing various helper
- * functions used to implement certain operations that are not necessarily
- * supported by the target CPU. For example, integer division doesn't have a
- * corresponding CPU instruction on ARMv5, and is instead implemented in the
- * compiler-generated machine code as a call to an __idiv helper function.
- *
- * Normally, one has to place libgcc.a in the link command used to generate
- * target binaries (shared libraries and executables) after all objects and
- * static libraries, but before dependent shared libraries, i.e. something
- * like:
- *         gcc <options> -o libfoo.so  foo.a libgcc.a -lc -lm
- *
- * This ensures that any helper function needed by the code in foo.a is copied
- * into the final libfoo.so. However, doing so will link a bunch of other __cxa
- * functions from libgcc.a into each .so and executable, causing 4k+ increase
- * in every binary. Therefore the Android platform build system has been
- * using this instead:
- *
- *         gcc <options> -o libfoo.so foo.a -lc -lm libgcc.a
- *
- * The problem with this is that if one helper function needed by foo.a has
- * already been copied into libc.so or libm.so, then nothing will be copied
- * into libfoo.so. Instead, a symbol import definition will be added to it
- * so libfoo.so can directly call the one in libc.so at runtime.
- *
- * When refreshing toolchains for new versions or using different architecture
- * flags, the set of helper functions copied to libc.so may change, which
- * resulted in some native shared libraries generated with the NDK or prebuilts
- * from vendors to fail to load properly.
- *
- * The NDK has been fixed after 1.6_r1 to use the correct link command, so
- * any native shared library generated with it should now be safe from that
- * problem. On the other hand, existing shared libraries distributed with
- * applications that were generated with a previous version of the NDK
- * still need all 1.5/1.6 helper functions in libc.so and libm.so
- *
- * After 3.2, the toolchain was updated again, adding __aeabi_f2uiz to the
- * list of requirements. Technically, this is due to mis-linked NDK libraries
- * but it is easier to add a single function here than asking several app
- * developers to fix their build.
- *
- * The __aeabi_idiv function is added to the list since cortex-a15 supports
- * HW idiv instructions so the system libc.so doesn't pull in the reference to
- * __aeabi_idiv but legacy libraries built against cortex-a9 targets still need
- * it.
- *
- * Final note: some of the functions below should really be in libm.so to
- *             completely reflect the state of 1.5/1.6 system images. However,
- *             since libm.so depends on libc.so, it's easier to put all of
- *             these in libc.so instead, since the dynamic linker will always
- *             search in libc.so before libm.so for dependencies.
- */
+extern char __adddf3;
+extern char __addsf3;
+extern char __aeabi_cdcmpeq;
+extern char __aeabi_cdcmple;
+extern char __aeabi_cdrcmple;
+extern char __aeabi_d2f;
+extern char __aeabi_d2iz;
+extern char __aeabi_dadd;
+extern char __aeabi_dcmpeq;
+extern char __aeabi_dcmpge;
+extern char __aeabi_dcmpgt;
+extern char __aeabi_dcmple;
+extern char __aeabi_dcmplt;
+extern char __aeabi_dcmpun;
+extern char __aeabi_ddiv;
+extern char __aeabi_dmul;
+extern char __aeabi_drsub;
+extern char __aeabi_dsub;
+extern char __aeabi_f2d;
+extern char __aeabi_f2iz;
+extern char __aeabi_f2uiz;
+extern char __aeabi_fadd;
+extern char __aeabi_fcmpun;
+extern char __aeabi_fdiv;
+extern char __aeabi_fmul;
+extern char __aeabi_frsub;
+extern char __aeabi_fsub;
+extern char __aeabi_i2d;
+extern char __aeabi_i2f;
+extern char __aeabi_idiv;
+extern char __aeabi_idivmod;
+extern char __aeabi_l2d;
+extern char __aeabi_l2f;
+extern char __aeabi_lasr;
+extern char __aeabi_ldivmod;
+extern char __aeabi_llsl;
+extern char __aeabi_llsr;
+extern char __aeabi_lmul;
+extern char __aeabi_ui2d;
+extern char __aeabi_ui2f;
+extern char __aeabi_uidiv;
+extern char __aeabi_uidivmod;
+extern char __aeabi_ul2d;
+extern char __aeabi_ul2f;
+extern char __aeabi_uldivmod;
+extern char __aeabi_unwind_cpp_pr0;
+extern char __aeabi_unwind_cpp_pr1;
+extern char __cmpdf2;
+extern char __divdf3;
+extern char __divsf3;
+extern char __eqdf2;
+extern char __extendsfdf2;
+extern char __fixdfsi;
+extern char __fixsfsi;
+extern char __floatdidf;
+extern char __floatdisf;
+extern char __floatsidf;
+extern char __floatsisf;
+extern char __floatundidf;
+extern char __floatundisf;
+extern char __floatunsidf;
+extern char __floatunsisf;
+extern char __gedf2;
+extern char __gtdf2;
+extern char __ledf2;
+extern char __ltdf2;
+extern char __muldf3;
+extern char __muldi3;
+extern char __mulsf3;
+extern char __nedf2;
+extern char __popcount_tab;
+extern char __popcountsi2;
+extern char __subdf3;
+extern char __subsf3;
+extern char __truncdfsf2;
+extern char __unorddf2;
+extern char __unordsf2;
 
-#define   COMPAT_FUNCTIONS_LIST \
-    XX(__adddf3)             \
-    XX(__addsf3)             \
-    XX(__aeabi_cdcmpeq)      \
-    XX(__aeabi_cdcmple)      \
-    XX(__aeabi_cdrcmple)     \
-    XX(__aeabi_d2f)          \
-    XX(__aeabi_d2iz)         \
-    XX(__aeabi_dadd)         \
-    XX(__aeabi_dcmpeq)       \
-    XX(__aeabi_dcmpge)       \
-    XX(__aeabi_dcmpgt)       \
-    XX(__aeabi_dcmple)       \
-    XX(__aeabi_dcmplt)       \
-    XX(__aeabi_dcmpun)       \
-    XX(__aeabi_ddiv)         \
-    XX(__aeabi_dmul)         \
-    XX(__aeabi_drsub)        \
-    XX(__aeabi_dsub)         \
-    XX(__aeabi_f2d)          \
-    XX(__aeabi_f2iz)         \
-    XX(__aeabi_f2uiz)        \
-    XX(__aeabi_fadd)         \
-    XX(__aeabi_fcmpun)       \
-    XX(__aeabi_fdiv)         \
-    XX(__aeabi_fmul)         \
-    XX(__aeabi_frsub)        \
-    XX(__aeabi_fsub)         \
-    XX(__aeabi_i2d)          \
-    XX(__aeabi_i2f)          \
-    XX(__aeabi_idiv)         \
-    XX(__aeabi_l2d)          \
-    XX(__aeabi_l2f)          \
-    XX(__aeabi_lasr)         \
-    XX(__aeabi_lmul)         \
-    XX(__aeabi_llsl)         \
-    XX(__aeabi_llsr)         \
-    XX(__aeabi_ui2d)         \
-    XX(__aeabi_ui2f)         \
-    XX(__aeabi_ul2d)         \
-    XX(__aeabi_ul2f)         \
-    XX(__cmpdf2)             \
-    XX(__divdf3)             \
-    XX(__divsf3)             \
-    XX(__eqdf2)              \
-    XX(__extendsfdf2)        \
-    XX(__fixdfsi)            \
-    XX(__fixsfsi)            \
-    XX(__floatdidf)          \
-    XX(__floatdisf)          \
-    XX(__floatsidf)          \
-    XX(__floatsisf)          \
-    XX(__floatundidf)        \
-    XX(__floatundisf)        \
-    XX(__floatunsidf)        \
-    XX(__floatunsisf)        \
-    XX(__gedf2)              \
-    XX(__gtdf2)              \
-    XX(__ledf2)              \
-    XX(__ltdf2)              \
-    XX(__muldf3)             \
-    XX(__muldi3)             \
-    XX(__mulsf3)             \
-    XX(__nedf2)              \
-    XX(__subdf3)             \
-    XX(__subsf3)             \
-    XX(__truncdfsf2)         \
-    XX(__unorddf2)           \
-    XX(__unordsf2)           \
-
-#define  XX(f)    extern void f(void);
-COMPAT_FUNCTIONS_LIST
-#undef XX
-
-void  __bionic_libgcc_compat_hooks(void)
-{
-#define XX(f)    f();
-COMPAT_FUNCTIONS_LIST
-#undef XX
-}
+void* __bionic_libgcc_compat_symbols[] = {
+    &__adddf3,
+    &__addsf3,
+    &__aeabi_cdcmpeq,
+    &__aeabi_cdcmple,
+    &__aeabi_cdrcmple,
+    &__aeabi_d2f,
+    &__aeabi_d2iz,
+    &__aeabi_dadd,
+    &__aeabi_dcmpeq,
+    &__aeabi_dcmpge,
+    &__aeabi_dcmpgt,
+    &__aeabi_dcmple,
+    &__aeabi_dcmplt,
+    &__aeabi_dcmpun,
+    &__aeabi_ddiv,
+    &__aeabi_dmul,
+    &__aeabi_drsub,
+    &__aeabi_dsub,
+    &__aeabi_f2d,
+    &__aeabi_f2iz,
+    &__aeabi_f2uiz,
+    &__aeabi_fadd,
+    &__aeabi_fcmpun,
+    &__aeabi_fdiv,
+    &__aeabi_fmul,
+    &__aeabi_frsub,
+    &__aeabi_fsub,
+    &__aeabi_i2d,
+    &__aeabi_i2f,
+    &__aeabi_idiv,
+    &__aeabi_idivmod,
+    &__aeabi_l2d,
+    &__aeabi_l2f,
+    &__aeabi_lasr,
+    &__aeabi_ldivmod,
+    &__aeabi_llsl,
+    &__aeabi_llsr,
+    &__aeabi_lmul,
+    &__aeabi_ui2d,
+    &__aeabi_ui2f,
+    &__aeabi_uidiv,
+    &__aeabi_uidivmod,
+    &__aeabi_ul2d,
+    &__aeabi_ul2f,
+    &__aeabi_uldivmod,
+    &__aeabi_unwind_cpp_pr0,
+    &__aeabi_unwind_cpp_pr1,
+    &__cmpdf2,
+    &__divdf3,
+    &__divsf3,
+    &__eqdf2,
+    &__extendsfdf2,
+    &__fixdfsi,
+    &__fixsfsi,
+    &__floatdidf,
+    &__floatdisf,
+    &__floatsidf,
+    &__floatsisf,
+    &__floatundidf,
+    &__floatundisf,
+    &__floatunsidf,
+    &__floatunsisf,
+    &__gedf2,
+    &__gtdf2,
+    &__ledf2,
+    &__ltdf2,
+    &__muldf3,
+    &__muldi3,
+    &__mulsf3,
+    &__nedf2,
+    &__popcount_tab,
+    &__popcountsi2,
+    &__subdf3,
+    &__subsf3,
+    &__truncdfsf2,
+    &__unorddf2,
+    &__unordsf2,
+};

libc/arch-arm/bionic/memcmp.S

@@ -27,7 +27,7 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 
 #ifdef HAVE_32_BYTE_CACHE_LINE
@@ -107,9 +107,11 @@ ENTRY(memcmp)
         bmi        10f
 #endif
 
-        .save {r4, lr}
         /* save registers */
         stmfd       sp!, {r4, lr}
+        .cfi_def_cfa_offset 8
+        .cfi_rel_offset r4, 0
+        .cfi_rel_offset lr, 4
 
         /* since r0 hold the result, move the first source
          * pointer somewhere else

libc/arch-arm/bionic/memcmp16.S

@@ -27,20 +27,20 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 /*
  * Optimized memcmp16() for ARM9.
  * This would not be optimal on XScale or ARM11, where more prefetching
- * and use of PLD will be needed.
+ * and use of pld will be needed.
  * The 2 major optimzations here are
  * (1) The main loop compares 16 bytes at a time
  * (2) The loads are scheduled in a way they won't stall
  */
 
 ENTRY(__memcmp16)
-        PLD         (r0, #0)
-        PLD         (r1, #0)
+        pld         [r0, #0]
+        pld         [r1, #0]
 
         /* take of the case where length is nul or the buffers are the same */
         cmp         r0, r1
@@ -62,8 +62,8 @@ ENTRY(__memcmp16)
         bpl         0f
 
         /* small blocks (less then 12 words) */
-        PLD         (r0, #32)
-        PLD         (r1, #32)
+        pld         [r0, #32]
+        pld         [r1, #32]
 
 1:      ldrh        r0, [r3], #2
         ldrh        ip, [r1], #2
@@ -74,9 +74,11 @@ ENTRY(__memcmp16)
         bx          lr
 
 
-        .save {r4, lr}
         /* save registers */
 0:      stmfd       sp!, {r4, lr}
+        .cfi_def_cfa_offset 8
+        .cfi_rel_offset r4, 0
+        .cfi_rel_offset lr, 4
 
         /* align first pointer to word boundary */
         tst         r3, #2
@@ -111,8 +113,8 @@ ENTRY(__memcmp16)
         bmi         1f
 
 0:
-        PLD         (r3, #64)
-        PLD         (r1, #64)
+        pld         [r3, #64]
+        pld         [r1, #64]
         ldr         r0, [r3], #4
         ldr         lr, [r1, #4]!
         eors        r0, r0, ip
@@ -193,8 +195,8 @@ ENTRY(__memcmp16)
         sub         r2, r2, #8
 
 6:
-        PLD         (r3, #64)
-        PLD         (r1, #64)
+        pld         [r3, #64]
+        pld         [r1, #64]
         mov         ip, lr, lsr #16
         ldr         lr, [r1], #4
         ldr         r0, [r3], #4

libc/arch-arm/bionic/memcpy.S

@@ -27,7 +27,7 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 #if defined(__ARM_NEON__) && !defined(ARCH_ARM_USE_NON_NEON_MEMCPY)
 
@@ -352,9 +352,9 @@ ENTRY(memcpy)
 
         // preload the destination because we'll align it to a cache line
         // with small writes. Also start the source "pump".
-        PLD         (r0, #0)
-        PLD         (r1, #0)
-        PLD         (r1, #32)
+        pld         [r0, #0]
+        pld         [r1, #0]
+        pld         [r1, #32]
 
 		/* it simplifies things to take care of len<4 early */
 		cmp			r2, #4
@@ -442,7 +442,7 @@ cached_aligned32:
         add         r12, r12, #64
 
 1:      ldmia       r1!, { r4-r11 }
-        PLD         (r12, #64)
+        pld         [r12, #64]
         subs        r2, r2, #32
 
         // NOTE: if r12 is more than 64 ahead of r1, the following ldrhi
@@ -563,7 +563,7 @@ loop16:
         ldr         r12, [r1], #4
 1:      mov         r4, r12
 		ldmia		r1!, {   r5,r6,r7,  r8,r9,r10,r11}
-        PLD         (r1, #64)
+        pld         [r1, #64]
         subs        r2, r2, #32
         ldrhs       r12, [r1], #4
 		orr			r3, r3, r4,		lsl #16
@@ -590,7 +590,7 @@ loop8:
         ldr         r12, [r1], #4
 1:      mov         r4, r12
 		ldmia		r1!, {   r5,r6,r7,  r8,r9,r10,r11}
-        PLD         (r1, #64)
+        pld         [r1, #64]
 		subs		r2, r2, #32
         ldrhs       r12, [r1], #4
 		orr			r3, r3, r4,		lsl #24
@@ -617,7 +617,7 @@ loop24:
         ldr         r12, [r1], #4
 1:      mov         r4, r12
 		ldmia		r1!, {   r5,r6,r7,  r8,r9,r10,r11}
-        PLD         (r1, #64)
+        pld         [r1, #64]
 		subs		r2, r2, #32
         ldrhs       r12, [r1], #4
 		orr			r3, r3, r4,		lsl #8

libc/arch-arm/bionic/memcpy.a15.S

@@ -1,406 +0,0 @@
-/*
- * Copyright (c) 2013 ARM Ltd
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- * 1. Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in the
- *    documentation and/or other materials provided with the distribution.
- * 3. The name of the company may not be used to endorse or promote
- *    products derived from this software without specific prior written
- *    permission.
- *
- * THIS SOFTWARE IS PROVIDED BY ARM LTD ``AS IS'' AND ANY EXPRESS OR IMPLIED
- * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
- * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
- * IN NO EVENT SHALL ARM LTD BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
- * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
- * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
- * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
- * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
- * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-    /* Prototype: void *memcpy (void *dst, const void *src, size_t count).  */
-
-        /* Use the version of memcpy implemented using LDRD and STRD.
-           This version is tuned for Cortex-A15.
-           This might not be the best for other ARMv7-A CPUs,
-           but there is no predefine to distinguish between
-           different CPUs in the same architecture,
-           and this version is better than the plain memcpy provided in newlib.
-
-           Therefore, we use this version for all ARMv7-A CPUS.  */
-
-        /* To make the same code compile for both ARM and Thumb instruction
-       sets, switch to unified syntax at the beginning of this function.
-           However, by using the same code, we may be missing optimization
-       opportunities.  For instance, in LDRD/STRD instructions, the first
-       destination register must be even and the second consecutive in
-       ARM state, but not in Thumb state.  */
-
-#include <machine/cpu-features.h>
-#include <machine/asm.h>
-
-        .syntax         unified
-
-ENTRY(memcpy)
-
-       /* Assumes that n >= 0, and dst, src are valid pointers.
-          If there is at least 8 bytes to copy, use LDRD/STRD.
-          If src and dst are misaligned with different offsets,
-          first copy byte by byte until dst is aligned,
-          and then copy using LDRD/STRD and shift if needed.
-          When less than 8 left, copy a word and then byte by byte.  */
-
-       /* Save registers (r0 holds the return value):
-          optimized push {r0, r4, r5, r6, r7, lr}.
-          To try and improve performance, stack layout changed,
-          i.e., not keeping the stack looking like users expect
-          (highest numbered register at highest address).  */
-        .save   {r0, lr}
-        push    {r0, lr}
-        .save   {r4, r5}
-        strd    r4, r5, [sp, #-8]!
-        .save   {r6, r7}
-        strd    r6, r7, [sp, #-8]!
-
-       /* TODO: Add debug frame directives.
-          We don't need exception unwind directives, because the code below
-      does not throw any exceptions and does not call any other functions.
-          Generally, newlib functions like this lack debug information for
-      assembler source.  */
-
-        /* Get copying of tiny blocks out of the way first.  */
-        /* Is there at least 4 bytes to copy?  */
-        subs    r2, r2, #4
-        blt     copy_less_than_4                 /* If n < 4.  */
-
-        /* Check word alignment.  */
-        ands    ip, r0, #3                       /* ip = last 2 bits of dst.  */
-        bne     dst_not_word_aligned             /* If dst is not word-aligned.  */
-
-        /* Get here if dst is word-aligned.  */
-        ands    ip, r1, #3                      /* ip = last 2 bits of src.  */
-        bne     src_not_word_aligned            /* If src is not word-aligned.  */
-word_aligned:
-        /* Get here if source and dst both are word-aligned.
-           The number of bytes remaining to copy is r2+4.  */
-
-        /* Is there is at least 64 bytes to copy?  */
-        subs    r2, r2, #60
-        blt     copy_less_than_64                /* If r2 + 4 < 64.  */
-
-        /* First, align the destination buffer to 8-bytes,
-           to make sure double loads and stores don't cross cache line boundary,
-           as they are then more expensive even if the data is in the cache
-           (require two load/store issue cycles instead of one).
-           If only one of the buffers is not 8-bytes aligned,
-           then it's more important to align dst than src,
-           because there is more penalty for stores
-           than loads that cross cacheline boundary.
-           This check and realignment are only worth doing
-           if there is a lot to copy.  */
-
-        /* Get here if dst is word aligned,
-           i.e., the 2 least significant bits are 0.
-           If dst is not 2w aligned (i.e., the 3rd bit is not set in dst),
-           then copy 1 word (4 bytes).  */
-        ands    r3, r0, #4
-        beq     11f                  /* If dst already two-word aligned.  */
-        ldr     r3, [r1], #4
-        str     r3, [r0], #4
-        subs    r2, r2, #4
-        blt     copy_less_than_64
-
-11:
-        /* TODO: Align to cacheline (useful for PLD optimization).  */
-
-        /* Every loop iteration copies 64 bytes.  */
-1:
-        .irp    offset, #0, #8, #16, #24, #32, #40, #48, #56
-        ldrd    r4, r5, [r1, \offset]
-        strd    r4, r5, [r0, \offset]
-        .endr
-
-        add     r0, r0, #64
-        add     r1, r1, #64
-        subs    r2, r2, #64
-        bge     1b                            /* If there is more to copy.  */
-
-copy_less_than_64:
-
-        /* Get here if less than 64 bytes to copy, -64 <= r2 < 0.
-           Restore the count if there is more than 7 bytes to copy.  */
-        adds    r2, r2, #56
-        blt     copy_less_than_8
-
-        /* Copy 8 bytes at a time.  */
-2:
-        ldrd    r4, r5, [r1], #8
-        strd    r4, r5, [r0], #8
-        subs    r2, r2, #8
-        bge     2b                            /* If there is more to copy.  */
-
-copy_less_than_8:
-
-        /* Get here if less than 8 bytes to copy, -8 <= r2 < 0.
-           Check if there is more to copy.  */
-        cmn     r2, #8
-        beq     return                          /* If r2 + 8 == 0.  */
-
-        /* Restore the count if there is more than 3 bytes to copy.  */
-        adds    r2, r2, #4
-        blt     copy_less_than_4
-
-        /* Copy 4 bytes.  */
-        ldr     r3, [r1], #4
-        str     r3, [r0], #4
-
-copy_less_than_4:
-        /* Get here if less than 4 bytes to copy, -4 <= r2 < 0.  */
-
-        /* Restore the count, check if there is more to copy.  */
-        adds    r2, r2, #4
-        beq     return                          /* If r2 == 0.  */
-
-        /* Get here with r2 is in {1,2,3}={01,10,11}.  */
-        /* Logical shift left r2, insert 0s, update flags.  */
-        lsls    r2, r2, #31
-
-        /* Copy byte by byte.
-           Condition ne means the last bit of r2 is 0.
-           Condition cs means the second to last bit of r2 is set,
-           i.e., r2 is 1 or 3.  */
-        itt     ne
-        ldrbne  r3, [r1], #1
-        strbne  r3, [r0], #1
-
-        itttt   cs
-        ldrbcs  r4, [r1], #1
-        ldrbcs  r5, [r1]
-        strbcs  r4, [r0], #1
-        strbcs  r5, [r0]
-
-return:
-        /* Restore registers: optimized pop {r0, r4, r5, r6, r7, pc}   */
-        /* This is the only return point of memcpy.  */
-        ldrd r6, r7, [sp], #8
-        ldrd r4, r5, [sp], #8
-        pop {r0, pc}
-
-#ifndef __ARM_FEATURE_UNALIGNED
-
-       /* The following assembly macro implements misaligned copy in software.
-          Assumes that dst is word aligned, src is at offset "pull" bits from
-      word, push = 32 - pull, and the number of bytes that remain to copy
-      is r2 + 4, r2 >= 0.  */
-
-       /* In the code below, r2 is the number of bytes that remain to be
-      written.  The number of bytes read is always larger, because we have
-      partial words in the shift queue.  */
-
-        .macro  miscopy pull push shiftleft shiftright
-
-        /* Align src to the previous word boundary.  */
-        bic     r1, r1, #3
-
-        /* Initialize the shift queue.  */
-        ldr     r5, [r1], #4                   /* Load a word from source.  */
-
-        subs    r2, r2, #4
-        blt     6f          /* Go to misaligned copy of less than 8 bytes.  */
-
-       /* Get here if there is more than 8 bytes to copy.
-          The number of bytes to copy is r2+8, r2 >= 0.  */
-
-       subs     r2, r2, #56
-       blt      4f         /* Go to misaligned copy of less than 64 bytes.  */
-
-3:
-       /* Get here if there is more than 64 bytes to copy.
-          The number of bytes to copy is r2+64, r2 >= 0.  */
-
-       /* Copy 64 bytes in every iteration.
-          Use a partial word from the shift queue.  */
-        .irp    offset, #0, #8, #16, #24, #32, #40, #48, #56
-        mov     r6, r5, \shiftleft #\pull
-        ldrd    r4, r5, [r1, \offset]
-        orr     r6, r6, r4, \shiftright #\push
-        mov     r7, r4, \shiftleft #\pull
-        orr     r7, r7, r5, \shiftright #\push
-        strd    r6, r7, [r0, \offset]
-        .endr
-
-        add     r1, r1, #64
-        add     r0, r0, #64
-        subs    r2, r2, #64
-        bge     3b
-
-4:
-       /* Get here if there is less than 64 bytes to copy (-64 <= r2 < 0)
-      and they are misaligned.  */
-
-       /* Restore the count if there is more than 7 bytes to copy.  */
-        adds    r2, r2, #56
-
-        blt     6f          /* Go to misaligned copy of less than 8 bytes.  */
-
-5:
-        /* Copy 8 bytes at a time.
-           Use a partial word from the shift queue.  */
-        mov     r6, r5, \shiftleft #\pull
-        ldrd    r4, r5, [r1], #8
-        orr     r6, r6, r4, \shiftright #\push
-        mov     r7, r4, \shiftleft #\pull
-        orr     r7, r7, r5, \shiftright #\push
-        strd    r6, r7, [r0], #8
-
-        subs    r2, r2, #8
-        bge     5b                        /* If there is more to copy.  */
-
-6:
-        /* Get here if there less than 8 bytes to copy (-8 <= r2 < 0)
-           and they are misaligned.  */
-
-        /* Check if there is more to copy.  */
-        cmn     r2, #8
-        beq     return
-
-        /* Check if there is less than 4 bytes to copy.  */
-        cmn     r2, #4
-
-        itt     lt
-        /* Restore src offset from word-align.  */
-        sublt   r1, r1, #(\push / 8)
-        blt     copy_less_than_4
-
-        /* Use a partial word from the shift queue.  */
-        mov     r3, r5, \shiftleft #\pull
-        /* Load a word from src, but without writeback
-           (this word is not fully written to dst).  */
-        ldr     r5, [r1]
-
-        /* Restore src offset from word-align.  */
-        add     r1, r1, #(\pull / 8)
-
-        /* Shift bytes to create one dst word and store it.  */
-        orr     r3, r3, r5, \shiftright #\push
-        str     r3, [r0], #4
-
-        /* Use single byte copying of the remaining bytes.  */
-        b       copy_less_than_4
-
-        .endm
-
-#endif /* not __ARM_FEATURE_UNALIGNED  */
-
-dst_not_word_aligned:
-
-       /* Get here when dst is not aligned and ip has the last 2 bits of dst,
-          i.e., ip is the offset of dst from word.
-          The number of bytes that remains to copy is r2 + 4,
-          i.e., there are at least 4 bytes to copy.
-          Write a partial word (0 to 3 bytes), such that dst becomes
-      word-aligned.  */
-
-       /* If dst is at ip bytes offset from a word (with 0 < ip < 4),
-          then there are (4 - ip) bytes to fill up to align dst to the next
-      word.  */
-        rsb     ip, ip, #4                        /* ip = #4 - ip.  */
-        cmp     ip, #2
-
-       /* Copy byte by byte with conditionals.  */
-        itt     gt
-        ldrbgt  r3, [r1], #1
-        strbgt  r3, [r0], #1
-
-        itt     ge
-        ldrbge  r4, [r1], #1
-        strbge  r4, [r0], #1
-
-        ldrb    lr, [r1], #1
-        strb    lr, [r0], #1
-
-       /* Update the count.
-          ip holds the number of bytes we have just copied.  */
-        subs    r2, r2, ip                        /* r2 = r2 - ip.  */
-        blt     copy_less_than_4                  /* If r2 < ip.  */
-
-       /* Get here if there are more than 4 bytes to copy.
-          Check if src is aligned.  If beforehand src and dst were not word
-      aligned but congruent (same offset), then now they are both
-      word-aligned, and we can copy the rest efficiently (without
-      shifting).  */
-        ands    ip, r1, #3                    /* ip = last 2 bits of src.  */
-        beq     word_aligned                  /* If r1 is word-aligned.  */
-
-src_not_word_aligned:
-       /* Get here when src is not word-aligned, but dst is word-aligned.
-          The number of bytes that remains to copy is r2+4.  */
-
-#ifdef __ARM_FEATURE_UNALIGNED
-       /* Copy word by word using LDR when alignment can be done in hardware,
-          i.e., SCTLR.A is set, supporting unaligned access in LDR and STR.  */
-        subs    r2, r2, #60
-        blt     8f
-
-7:
-        /* Copy 64 bytes in every loop iteration.  */
-        .irp    offset, #0, #4, #8, #12, #16, #20, #24, #28, #32, #36, #40, #44, #48, #52, #56, #60
-        ldr     r3, [r1, \offset]
-        str     r3, [r0, \offset]
-        .endr
-
-        add     r0, r0, #64
-        add     r1, r1, #64
-        subs    r2, r2, #64
-        bge     7b
-
-8:
-        /* Get here if less than 64 bytes to copy, -64 <= r2 < 0.
-           Check if there is more than 3 bytes to copy.  */
-        adds    r2, r2, #60
-        blt     copy_less_than_4
-
-9:
-       /* Get here if there is less than 64 but at least 4 bytes to copy,
-          where the number of bytes to copy is r2+4.  */
-        ldr     r3, [r1], #4
-        str     r3, [r0], #4
-        subs    r2, r2, #4
-        bge     9b
-
-        b       copy_less_than_4
-
-#else /* not __ARM_FEATURE_UNALIGNED  */
-
-       /* ip has last 2 bits of src,
-          i.e., ip is the offset of src from word, and ip > 0.
-          Compute shifts needed to copy from src to dst.  */
-        cmp     ip, #2
-        beq     miscopy_16_16             /* If ip == 2.  */
-        bge     miscopy_24_8              /* If ip == 3.  */
-
-        /* Get here if ip == 1.  */
-
-        /* Endian independent macros for shifting bytes within registers.  */
-
-#ifndef __ARMEB__
-miscopy_8_24:   miscopy pull=8 push=24 shiftleft=lsr shiftright=lsl
-miscopy_16_16:  miscopy pull=16 push=16 shiftleft=lsr shiftright=lsl
-miscopy_24_8:   miscopy pull=24 push=8 shiftleft=lsr shiftright=lsl
-#else  /* not __ARMEB__ */
-miscopy_8_24:   miscopy pull=8 push=24 shiftleft=lsl shiftright=lsr
-miscopy_16_16:  miscopy pull=16 push=16 shiftleft=lsl shiftright=lsr
-miscopy_24_8:   miscopy pull=24 push=8 shiftleft=lsl shiftright=lsr
-#endif  /* not __ARMEB__ */
-
-#endif  /* not __ARM_FEATURE_UNALIGNED  */
-
-END(memcpy)

libc/arch-arm/bionic/memcpy.a9.S

@@ -44,7 +44,7 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 	.syntax unified
 	/* This implementation requires ARM state.  */

libc/arch-arm/bionic/memset.S

@@ -1,200 +0,0 @@
-/*
- * Copyright (C) 2008 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-#include <machine/cpu-features.h>
-#include <machine/asm.h>
-
-		/*
-		 * Optimized memset() for ARM.
-         *
-         * memset() returns its first argument.
-		 */
-
-#if defined(__ARM_NEON__)
-    .fpu    neon
-#endif
-
-ENTRY(bzero)
-        mov     r2, r1
-        mov     r1, #0
-        // Fall through to memset...
-END(bzero)
-
-ENTRY(memset)
-#if defined(__ARM_NEON__)
-
-#ifdef  NEON_MEMSET_DIVIDER
-        cmp         r2, #NEON_MEMSET_DIVIDER
-        bhi         11f
-#endif
-        .save       {r0}
-        stmfd       sp!, {r0}
-
-        vdup.8      q0, r1
-
-#ifndef NEON_UNALIGNED_ACCESS
-        /* do we have at least 16-bytes to write (needed for alignment below) */
-        cmp         r2, #16
-        blo         3f
-
-        /* align destination to 16 bytes for the write-buffer */
-        rsb         r3, r0, #0
-        ands        r3, r3, #0xF
-        beq         2f
-
-        /* write up to 15-bytes (count in r3) */
-        sub         r2, r2, r3
-        movs        ip, r3, lsl #31
-        strmib      r1, [r0], #1
-        strcsb      r1, [r0], #1
-        strcsb      r1, [r0], #1
-        movs        ip, r3, lsl #29
-        bge         1f
-
-        // writes 4 bytes, 32-bits aligned
-        vst1.32     {d0[0]}, [r0, :32]!
-1:      bcc         2f
-
-        // writes 8 bytes, 64-bits aligned
-        vst1.8      {d0}, [r0, :64]!
-2:
-#endif
-        /* make sure we have at least 32 bytes to write */
-        subs        r2, r2, #32
-        blo         2f
-        vmov        q1, q0
-
-1:      /* The main loop writes 32 bytes at a time */
-        subs        r2, r2, #32
-#ifndef NEON_UNALIGNED_ACCESS
-        vst1.8      {d0 - d3}, [r0, :128]!
-#else
-        vst1.8      {d0 - d3}, [r0]!
-#endif
-        bhs         1b
-
-2:      /* less than 32 left */
-        add         r2, r2, #32
-        tst         r2, #0x10
-        beq         3f
-
-        // writes 16 bytes, 128-bits aligned
-#ifndef NEON_UNALIGNED_ACCESS
-        vst1.8      {d0, d1}, [r0, :128]!
-#else
-        vst1.8      {d0, d1}, [r0]!
-#endif
-3:      /* write up to 15-bytes (count in r2) */
-        movs        ip, r2, lsl #29
-        bcc         1f
-        vst1.8      {d0}, [r0]!
-1:      bge         2f
-        vst1.32     {d0[0]}, [r0]!
-2:      movs        ip, r2, lsl #31
-        strmib      r1, [r0], #1
-        strcsb      r1, [r0], #1
-        strcsb      r1, [r0], #1
-        ldmfd       sp!, {r0}
-        bx          lr
-11:
-#endif
-
-        /*
-         * Optimized memset() for ARM.
-         *
-         * memset() returns its first argument.
-         */
-
-		/* compute the offset to align the destination
-		 * offset = (4-(src&3))&3 = -src & 3
-		 */
-
-        .save       {r0, r4-r7, lr}
-		stmfd		sp!, {r0, r4-r7, lr}
-		rsb			r3, r0, #0
-		ands		r3, r3, #3
-        cmp         r3, r2
-        movhi       r3, r2
-
-        /* splat r1 */
-        mov         r1, r1, lsl #24
-        orr         r1, r1, r1, lsr #8
-        orr         r1, r1, r1, lsr #16
-
-		movs		r12, r3, lsl #31
-		strcsb		r1, [r0], #1    /* can't use strh (alignment unknown) */
-		strcsb		r1, [r0], #1
-		strmib		r1, [r0], #1
-		subs		r2, r2, r3
-        ldmlsfd     sp!, {r0, r4-r7, lr}   /* return */
-        bxls        lr
-
-		/* align the destination to a cache-line */
-        mov         r12, r1
-        mov         lr, r1
-        mov         r4, r1
-        mov         r5, r1
-        mov         r6, r1
-        mov         r7, r1
-
-		rsb         r3, r0, #0
-		ands		r3, r3, #0x1C
-		beq         3f
-		cmp         r3, r2
-		andhi		r3, r2, #0x1C
-		sub         r2, r2, r3
-
-		/* conditionally writes 0 to 7 words (length in r3) */
-		movs		r3, r3, lsl #28
-		stmcsia		r0!, {r1, lr}
-		stmcsia		r0!, {r1, lr}
-		stmmiia		r0!, {r1, lr}
-		movs		r3, r3, lsl #2
-        strcs       r1, [r0], #4
-
-3:
-        subs        r2, r2, #32
-        mov         r3, r1
-        bmi         2f
-1:      subs        r2, r2, #32
-        stmia		r0!, {r1,r3,r4,r5,r6,r7,r12,lr}
-        bhs         1b
-2:      add         r2, r2, #32
-
-		/* conditionally stores 0 to 31 bytes */
-		movs		r2, r2, lsl #28
-		stmcsia		r0!, {r1,r3,r12,lr}
-		stmmiia		r0!, {r1, lr}
-		movs		r2, r2, lsl #2
-        strcs       r1, [r0], #4
-		strmih		r1, [r0], #2
-		movs		r2, r2, lsl #2
-		strcsb		r1, [r0]
-        ldmfd		sp!, {r0, r4-r7, lr}
-        bx          lr
-END(memset)

libc/arch-arm/bionic/setjmp.S

@@ -34,7 +34,7 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 #include <machine/setjmp.h>
 #include <machine/cpu-features.h>
 
@@ -51,12 +51,16 @@
 ENTRY(setjmp)
 	/* Block all signals and retrieve the old signal mask */
 	stmfd	sp!, {r0, r14}
+	.cfi_def_cfa_offset 8
+	.cfi_rel_offset r0, 0
+	.cfi_rel_offset r14, 4
 	mov	r0, #0x00000000
 
-	bl	PIC_SYM(_C_LABEL(sigblock), PLT)
+	bl	PIC_SYM(sigblock, PLT)
 	mov	r1, r0
 
 	ldmfd	sp!, {r0, r14}
+	.cfi_def_cfa_offset 0
 
 	/* Store signal mask */
 	str	r1, [r0, #(_JB_SIGMASK * 4)]
@@ -96,13 +100,20 @@ ENTRY(longjmp)
 
 	/* Set signal mask */
 	stmfd	sp!, {r0, r1, r14}
+	.cfi_def_cfa_offset 12
+	.cfi_rel_offset r0, 0
+	.cfi_rel_offset r1, 4
+	.cfi_rel_offset r14, 8
 	sub	sp, sp, #4	/* align the stack */
+	.cfi_adjust_cfa_offset 4
 
 	mov	r0, r2
-	bl	PIC_SYM(_C_LABEL(sigsetmask), PLT)
+	bl	PIC_SYM(sigsetmask, PLT)
 
 	add	sp, sp, #4	/* unalign the stack */
+	.cfi_adjust_cfa_offset -4
 	ldmfd	sp!, {r0, r1, r14}
+	.cfi_def_cfa_offset 0
 
 #ifdef __ARM_HAVE_VFP
 	/* Restore floating-point registers */
@@ -136,7 +147,7 @@ ENTRY(longjmp)
 
 	/* validation failed, die die die. */
 botch:
-	bl	PIC_SYM(_C_LABEL(longjmperror), PLT)
-	bl	PIC_SYM(_C_LABEL(abort), PLT)
+	bl	PIC_SYM(longjmperror, PLT)
+	bl	PIC_SYM(abort, PLT)
 	b	. - 8		/* Cannot get here */
 END(longjmp)

libc/arch-arm/bionic/sigsetjmp.S

@@ -35,7 +35,7 @@
 
 #define _ALIGN_TEXT .align 0
 
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 #include <machine/setjmp.h>
 
 /*
@@ -50,8 +50,8 @@
 
 ENTRY(sigsetjmp)
 	teq	r1, #0
-	beq	PIC_SYM(_C_LABEL(_setjmp), PLT)
-	b	PIC_SYM(_C_LABEL(setjmp), PLT)
+	beq	PIC_SYM(_setjmp, PLT)
+	b	PIC_SYM(setjmp, PLT)
 END(sigsetjmp)
 
 .L_setjmp_magic:
@@ -61,6 +61,6 @@ ENTRY(siglongjmp)
 	ldr	r2, .L_setjmp_magic
 	ldr	r3, [r0]
 	teq	r2, r3
-	beq	PIC_SYM(_C_LABEL(_longjmp), PLT)
-	b	PIC_SYM(_C_LABEL(longjmp), PLT)
+	beq	PIC_SYM(_longjmp, PLT)
+	b	PIC_SYM(longjmp, PLT)
 END(siglongjmp)

libc/arch-arm/bionic/strcmp.S

@@ -28,7 +28,7 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 	.text
 
@@ -52,8 +52,8 @@
 #define magic2(REG) REG, lsl #7
 
 ENTRY(strcmp)
-	PLD(r0, #0)
-	PLD(r1, #0)
+	pld	[r0, #0]
+	pld	[r1, #0]
 	eor	r2, r0, r1
 	tst	r2, #3
 
@@ -88,8 +88,8 @@ ENTRY(strcmp)
 	orr	r4, r4, r4, lsl #16
 	.p2align	2
 4:
-	PLD(r0, #8)
-	PLD(r1, #8)
+	pld	[r0, #8]
+	pld	[r1, #8]
 	sub	r2, ip, magic1(r4)
 	cmp	ip, r3
 	itttt	eq

libc/arch-arm/bionic/strcmp.a15.S

@@ -1,787 +0,0 @@
-/*
- * Copyright (c) 2013 ARM Ltd
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- * 1. Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in the
- *    documentation and/or other materials provided with the distribution.
- * 3. The name of the company may not be used to endorse or promote
- *    products derived from this software without specific prior written
- *    permission.
- *
- * THIS SOFTWARE IS PROVIDED BY ARM LTD ``AS IS'' AND ANY EXPRESS OR IMPLIED
- * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
- * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
- * IN NO EVENT SHALL ARM LTD BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
- * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
- * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
- * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
- * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
- * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#include "arm_asm.h"
-
-#ifdef __ARMEB__
-#define S2LOMEM lsl
-#define S2LOMEMEQ lsleq
-#define S2HIMEM lsr
-#define MSB 0x000000ff
-#define LSB 0xff000000
-#define BYTE0_OFFSET 24
-#define BYTE1_OFFSET 16
-#define BYTE2_OFFSET 8
-#define BYTE3_OFFSET 0
-#else /* not  __ARMEB__ */
-#define S2LOMEM lsr
-#define S2LOMEMEQ lsreq
-#define S2HIMEM lsl
-#define BYTE0_OFFSET 0
-#define BYTE1_OFFSET 8
-#define BYTE2_OFFSET 16
-#define BYTE3_OFFSET 24
-#define MSB 0xff000000
-#define LSB 0x000000ff
-#endif /* not  __ARMEB__ */
-
-.syntax         unified
-
-#if defined (__thumb__)
-        .thumb
-        .thumb_func
-#endif
-        .global strcmp
-        .type   strcmp, %function
-strcmp:
-
-#if (defined (__thumb__) && !defined (__thumb2__))
-1:
-        ldrb    r2, [r0]
-        ldrb    r3, [r1]
-        adds    r0, r0, #1
-        adds    r1, r1, #1
-        cmp     r2, #0
-        beq     2f
-        cmp     r2, r3
-        beq     1b
-2:
-        subs    r0, r2, r3
-        bx      lr
-#elif (defined (__OPTIMIZE_SIZE__) || defined (PREFER_SIZE_OVER_SPEED))
-1:
-        ldrb    r2, [r0], #1
-        ldrb    r3, [r1], #1
-        cmp     r2, #1
-        it      cs
-        cmpcs   r2, r3
-        beq     1b
-        subs    r0, r2, r3
-        RETURN
-
-
-#elif (defined (_ISA_THUMB_2) || defined (_ISA_ARM_6))
-      /* Use LDRD whenever possible.  */
-
-/* The main thing to look out for when comparing large blocks is that
-   the loads do not cross a page boundary when loading past the index
-   of the byte with the first difference or the first string-terminator.
-
-   For example, if the strings are identical and the string-terminator
-   is at index k, byte by byte comparison will not load beyond address
-   s1+k and s2+k; word by word comparison may load up to 3 bytes beyond
-   k; double word - up to 7 bytes.  If the load of these bytes crosses
-   a page boundary, it might cause a memory fault (if the page is not mapped)
-   that would not have happened in byte by byte comparison.
-
-   If an address is (double) word aligned, then a load of a (double) word
-   from that address will not cross a page boundary.
-   Therefore, the algorithm below considers word and double-word alignment
-   of strings separately.  */
-
-/* High-level description of the algorithm.
-
-   * The fast path: if both strings are double-word aligned,
-     use LDRD to load two words from each string in every loop iteration.
-   * If the strings have the same offset from a word boundary,
-     use LDRB to load and compare byte by byte until
-     the first string is aligned to a word boundary (at most 3 bytes).
-     This is optimized for quick return on short unaligned strings.
-   * If the strings have the same offset from a double-word boundary,
-     use LDRD to load two words from each string in every loop iteration, as in the fast path.
-   * If the strings do not have the same offset from a double-word boundary,
-     load a word from the second string before the loop to initialize the queue.
-     Use LDRD to load two words from every string in every loop iteration.
-     Inside the loop, load the second word from the second string only after comparing
-     the first word, using the queued value, to guarantee safety across page boundaries.
-   * If the strings do not have the same offset from a word boundary,
-     use LDR and a shift queue. Order of loads and comparisons matters,
-     similarly to the previous case.
-
-   * Use UADD8 and SEL to compare words, and use REV and CLZ to compute the return value.
-   * The only difference between ARM and Thumb modes is the use of CBZ instruction.
-   * The only difference between big and little endian is the use of REV in little endian
-     to compute the return value, instead of MOV.
-   * No preload. [TODO.]
-*/
-
-        .macro m_cbz reg label
-#ifdef __thumb2__
-        cbz     \reg, \label
-#else   /* not defined __thumb2__ */
-        cmp     \reg, #0
-        beq     \label
-#endif /* not defined __thumb2__ */
-        .endm /* m_cbz */
-
-        .macro m_cbnz reg label
-#ifdef __thumb2__
-        cbnz    \reg, \label
-#else   /* not defined __thumb2__ */
-        cmp     \reg, #0
-        bne     \label
-#endif /* not defined __thumb2__ */
-        .endm /* m_cbnz */
-
-        .macro  init
-        /* Macro to save temporary registers and prepare magic values.  */
-        subs    sp, sp, #16
-        strd    r4, r5, [sp, #8]
-        strd    r6, r7, [sp]
-        mvn     r6, #0  /* all F */
-        mov     r7, #0  /* all 0 */
-        .endm   /* init */
-
-        .macro  magic_compare_and_branch w1 w2 label
-        /* Macro to compare registers w1 and w2 and conditionally branch to label.  */
-        cmp     \w1, \w2        /* Are w1 and w2 the same?  */
-        magic_find_zero_bytes \w1
-        it      eq
-        cmpeq   ip, #0          /* Is there a zero byte in w1?  */
-        bne     \label
-        .endm /* magic_compare_and_branch */
-
-        .macro  magic_find_zero_bytes w1
-        /* Macro to find all-zero bytes in w1, result is in ip.  */
-#if (defined (__ARM_FEATURE_DSP))
-        uadd8   ip, \w1, r6
-        sel     ip, r7, r6
-#else /* not defined (__ARM_FEATURE_DSP) */
-        /* __ARM_FEATURE_DSP is not defined for some Cortex-M processors.
-        Coincidently, these processors only have Thumb-2 mode, where we can use the
-        the (large) magic constant available directly as an immediate in instructions.
-        Note that we cannot use the magic constant in ARM mode, where we need
-        to create the constant in a register.  */
-        sub     ip, \w1, #0x01010101
-        bic     ip, ip, \w1
-        and     ip, ip, #0x80808080
-#endif /* not defined (__ARM_FEATURE_DSP) */
-        .endm /* magic_find_zero_bytes */
-
-        .macro  setup_return w1 w2
-#ifdef __ARMEB__
-        mov     r1, \w1
-        mov     r2, \w2
-#else /* not  __ARMEB__ */
-        rev     r1, \w1
-        rev     r2, \w2
-#endif /* not  __ARMEB__ */
-        .endm /* setup_return */
-
-        /*
-        optpld r0, #0
-        optpld r1, #0
-        */
-
-        /* Are both strings double-word aligned?  */
-        orr     ip, r0, r1
-        tst     ip, #7
-        bne     do_align
-
-        /* Fast path.  */
-        init
-
-doubleword_aligned:
-
-        /* Get here when the strings to compare are double-word aligned.  */
-        /* Compare two words in every iteration.  */
-        .p2align        2
-2:
-        /*
-        optpld r0, #16
-        optpld r1, #16
-        */
-
-        /* Load the next double-word from each string.  */
-        ldrd    r2, r3, [r0], #8
-        ldrd    r4, r5, [r1], #8
-
-        magic_compare_and_branch w1=r2, w2=r4, label=return_24
-        magic_compare_and_branch w1=r3, w2=r5, label=return_35
-        b       2b
-
-do_align:
-        /* Is the first string word-aligned?  */
-        ands    ip, r0, #3
-        beq     word_aligned_r0
-
-        /* Fast compare byte by byte until the first string is word-aligned.  */
-        /* The offset of r0 from a word boundary is in ip. Thus, the number of bytes
-        to read until the next word boudnary is 4-ip.  */
-        bic     r0, r0, #3
-        ldr     r2, [r0], #4
-        lsls    ip, ip, #31
-        beq     byte2
-        bcs     byte3
-
-byte1:
-        ldrb    ip, [r1], #1
-        uxtb    r3, r2, ror #BYTE1_OFFSET
-        subs    ip, r3, ip
-        bne     fast_return
-        m_cbz   reg=r3, label=fast_return
-
-byte2:
-        ldrb    ip, [r1], #1
-        uxtb    r3, r2, ror #BYTE2_OFFSET
-        subs    ip, r3, ip
-        bne     fast_return
-        m_cbz   reg=r3, label=fast_return
-
-byte3:
-        ldrb    ip, [r1], #1
-        uxtb    r3, r2, ror #BYTE3_OFFSET
-        subs    ip, r3, ip
-        bne     fast_return
-        m_cbnz  reg=r3, label=word_aligned_r0
-
-fast_return:
-        mov     r0, ip
-        bx      lr
-
-word_aligned_r0:
-        init
-        /* The first string is word-aligned.  */
-        /* Is the second string word-aligned?  */
-        ands    ip, r1, #3
-        bne     strcmp_unaligned
-
-word_aligned:
-        /* The strings are word-aligned. */
-        /* Is the first string double-word aligned?  */
-        tst     r0, #4
-        beq     doubleword_aligned_r0
-
-        /* If r0 is not double-word aligned yet, align it by loading
-        and comparing the next word from each string.  */
-        ldr     r2, [r0], #4
-        ldr     r4, [r1], #4
-        magic_compare_and_branch w1=r2 w2=r4 label=return_24
-
-doubleword_aligned_r0:
-        /* Get here when r0 is double-word aligned.  */
-        /* Is r1 doubleword_aligned?  */
-        tst     r1, #4
-        beq     doubleword_aligned
-
-        /* Get here when the strings to compare are word-aligned,
-        r0 is double-word aligned, but r1 is not double-word aligned.  */
-
-        /* Initialize the queue.  */
-        ldr     r5, [r1], #4
-
-        /* Compare two words in every iteration.  */
-        .p2align        2
-3:
-        /*
-        optpld r0, #16
-        optpld r1, #16
-        */
-
-        /* Load the next double-word from each string and compare.  */
-        ldrd    r2, r3, [r0], #8
-        magic_compare_and_branch w1=r2 w2=r5 label=return_25
-        ldrd    r4, r5, [r1], #8
-        magic_compare_and_branch w1=r3 w2=r4 label=return_34
-        b       3b
-
-        .macro miscmp_word offsetlo offsethi
-        /* Macro to compare misaligned strings.  */
-        /* r0, r1 are word-aligned, and at least one of the strings
-        is not double-word aligned.  */
-        /* Compare one word in every loop iteration.  */
-        /* OFFSETLO is the original bit-offset of r1 from a word-boundary,
-        OFFSETHI is 32 - OFFSETLO (i.e., offset from the next word).  */
-
-        /* Initialize the shift queue.  */
-        ldr     r5, [r1], #4
-
-        /* Compare one word from each string in every loop iteration.  */
-        .p2align        2
-7:
-        ldr     r3, [r0], #4
-        S2LOMEM r5, r5, #\offsetlo
-        magic_find_zero_bytes w1=r3
-        cmp     r7, ip, S2HIMEM #\offsetlo
-        and     r2, r3, r6, S2LOMEM #\offsetlo
-        it      eq
-        cmpeq   r2, r5
-        bne     return_25
-        ldr     r5, [r1], #4
-        cmp     ip, #0
-        eor r3, r2, r3
-        S2HIMEM r2, r5, #\offsethi
-        it      eq
-        cmpeq   r3, r2
-        bne     return_32
-        b       7b
-        .endm /* miscmp_word */
-
-strcmp_unaligned:
-        /* r0 is word-aligned, r1 is at offset ip from a word.  */
-        /* Align r1 to the (previous) word-boundary.  */
-        bic     r1, r1, #3
-
-        /* Unaligned comparison word by word using LDRs. */
-        cmp     ip, #2
-        beq     miscmp_word_16                    /* If ip == 2.  */
-        bge     miscmp_word_24                    /* If ip == 3.  */
-        miscmp_word offsetlo=8 offsethi=24        /* If ip == 1.  */
-miscmp_word_16:  miscmp_word offsetlo=16 offsethi=16
-miscmp_word_24:  miscmp_word offsetlo=24 offsethi=8
-
-
-return_32:
-        setup_return w1=r3, w2=r2
-        b       do_return
-return_34:
-        setup_return w1=r3, w2=r4
-        b       do_return
-return_25:
-        setup_return w1=r2, w2=r5
-        b       do_return
-return_35:
-        setup_return w1=r3, w2=r5
-        b       do_return
-return_24:
-        setup_return w1=r2, w2=r4
-
-do_return:
-
-#ifdef __ARMEB__
-        mov     r0, ip
-#else /* not  __ARMEB__ */
-        rev     r0, ip
-#endif /* not  __ARMEB__ */
-
-        /* Restore temporaries early, before computing the return value.  */
-        ldrd    r6, r7, [sp]
-        ldrd    r4, r5, [sp, #8]
-        adds    sp, sp, #16
-
-        /* There is a zero or a different byte between r1 and r2.  */
-        /* r0 contains a mask of all-zero bytes in r1.  */
-        /* Using r0 and not ip here because cbz requires low register.  */
-        m_cbz   reg=r0, label=compute_return_value
-        clz     r0, r0
-        /* r0 contains the number of bits on the left of the first all-zero byte in r1.  */
-        rsb     r0, r0, #24
-        /* Here, r0 contains the number of bits on the right of the first all-zero byte in r1.  */
-        lsr     r1, r1, r0
-        lsr     r2, r2, r0
-
-compute_return_value:
-        movs    r0, #1
-        cmp     r1, r2
-        /* The return value is computed as follows.
-        If r1>r2 then (C==1 and Z==0) and LS doesn't hold and r0 is #1 at return.
-        If r1<r2 then (C==0 and Z==0) and we execute SBC with carry_in=0,
-        which means r0:=r0-r0-1 and r0 is #-1 at return.
-        If r1=r2 then (C==1 and Z==1) and we execute SBC with carry_in=1,
-        which means r0:=r0-r0 and r0 is #0 at return.
-        (C==0 and Z==1) cannot happen because the carry bit is "not borrow".  */
-        it      ls
-        sbcls   r0, r0, r0
-        bx      lr
-
-
-#else   /* !(defined (_ISA_THUMB_2) || defined (_ISA_ARM_6)
-             defined (__OPTIMIZE_SIZE__) || defined (PREFER_SIZE_OVER_SPEED) ||
-             (defined (__thumb__) && !defined (__thumb2__))) */
-
-        /* Use LDR whenever possible. */
-
-#ifdef __thumb2__
-#define magic1(REG) 0x01010101
-#define magic2(REG) 0x80808080
-#else
-#define magic1(REG) REG
-#define magic2(REG) REG, lsl #7
-#endif
-
-        optpld  r0
-        optpld  r1
-        eor     r2, r0, r1
-        tst     r2, #3
-        /* Strings not at same byte offset from a word boundary.  */
-        bne     strcmp_unaligned
-        ands    r2, r0, #3
-        bic     r0, r0, #3
-        bic     r1, r1, #3
-        ldr     ip, [r0], #4
-        it      eq
-        ldreq   r3, [r1], #4
-        beq     1f
-        /* Although s1 and s2 have identical initial alignment, they are
-        not currently word aligned.  Rather than comparing bytes,
-    make sure that any bytes fetched from before the addressed
-    bytes are forced to 0xff.  Then they will always compare
-    equal.  */
-        eor     r2, r2, #3
-        lsl     r2, r2, #3
-        mvn     r3, MSB
-        S2LOMEM        r2, r3, r2
-        ldr     r3, [r1], #4
-        orr     ip, ip, r2
-        orr     r3, r3, r2
-1:
-#ifndef __thumb2__
-              /* Load the 'magic' constant 0x01010101.  */
-        str     r4, [sp, #-4]!
-        mov     r4, #1
-        orr     r4, r4, r4, lsl #8
-        orr     r4, r4, r4, lsl #16
-#endif
-        .p2align        2
-4:
-        optpld  r0, #8
-        optpld  r1, #8
-        sub     r2, ip, magic1(r4)
-        cmp     ip, r3
-        itttt   eq
-        /* check for any zero bytes in first word */
-        biceq   r2, r2, ip
-        tsteq   r2, magic2(r4)
-        ldreq   ip, [r0], #4
-        ldreq   r3, [r1], #4
-        beq     4b
-2:
-        /* There's a zero or a different byte in the word */
-        S2HIMEM  r0, ip, #24
-        S2LOMEM  ip, ip, #8
-        cmp     r0, #1
-        it      cs
-        cmpcs   r0, r3, S2HIMEM #24
-        it      eq
-        S2LOMEMEQ r3, r3, #8
-        beq     2b
-        /* On a big-endian machine, r0 contains the desired byte in bits
-        0-7; on a little-endian machine they are in bits 24-31.  In
-        both cases the other bits in r0 are all zero.  For r3 the
-        interesting byte is at the other end of the word, but the
-        other bits are not necessarily zero.  We need a signed result
-        representing the differnece in the unsigned bytes, so for the
-        little-endian case we can't just shift the interesting bits
-        up.  */
-#ifdef __ARMEB__
-        sub     r0, r0, r3, lsr #24
-#else
-        and     r3, r3, #255
-#ifdef __thumb2__
-        /* No RSB instruction in Thumb2 */
-        lsr     r0, r0, #24
-        sub     r0, r0, r3
-#else
-        rsb     r0, r3, r0, lsr #24
-#endif
-#endif
-#ifndef __thumb2__
-        ldr     r4, [sp], #4
-#endif
-        RETURN
-
-
-strcmp_unaligned:
-
-#if 0
-        /* The assembly code below is based on the following alogrithm.  */
-#ifdef __ARMEB__
-#define RSHIFT <<
-#define LSHIFT >>
-#else
-#define RSHIFT >>
-#define LSHIFT <<
-#endif
-
-#define body(shift)                         \
-  mask = 0xffffffffU RSHIFT shift;                  \
-  w1 = *wp1++;                              \
-  w2 = *wp2++;                              \
-  do                                    \
-    {                                   \
-      t1 = w1 & mask;                           \
-      if (__builtin_expect(t1 != w2 RSHIFT shift, 0))           \
-    {                               \
-      w2 RSHIFT= shift;                     \
-      break;                            \
-    }                               \
-      if (__builtin_expect(((w1 - b1) & ~w1) & (b1 << 7), 0))       \
-    {                               \
-      /* See comment in assembler below re syndrome on big-endian */\
-      if ((((w1 - b1) & ~w1) & (b1 << 7)) & mask)           \
-        w2 RSHIFT= shift;                       \
-      else                              \
-        {                               \
-          w2 = *wp2;                        \
-          t1 = w1 RSHIFT (32 - shift);              \
-          w2 = (w2 LSHIFT (32 - shift)) RSHIFT (32 - shift);    \
-        }                               \
-      break;                            \
-    }                               \
-      w2 = *wp2++;                          \
-      t1 ^= w1;                             \
-      if (__builtin_expect(t1 != w2 LSHIFT (32 - shift), 0))        \
-    {                               \
-      t1 = w1 >> (32 - shift);                  \
-      w2 = (w2 << (32 - shift)) RSHIFT (32 - shift);        \
-      break;                            \
-    }                               \
-      w1 = *wp1++;                          \
-    } while (1)
-
-  const unsigned* wp1;
-  const unsigned* wp2;
-  unsigned w1, w2;
-  unsigned mask;
-  unsigned shift;
-  unsigned b1 = 0x01010101;
-  char c1, c2;
-  unsigned t1;
-
-  while (((unsigned) s1) & 3)
-    {
-      c1 = *s1++;
-      c2 = *s2++;
-      if (c1 == 0 || c1 != c2)
-    return c1 - (int)c2;
-    }
-  wp1 = (unsigned*) (((unsigned)s1) & ~3);
-  wp2 = (unsigned*) (((unsigned)s2) & ~3);
-  t1 = ((unsigned) s2) & 3;
-  if (t1 == 1)
-    {
-      body(8);
-    }
-  else if (t1 == 2)
-    {
-      body(16);
-    }
-  else
-    {
-      body (24);
-    }
-
-  do
-    {
-#ifdef __ARMEB__
-      c1 = (char) t1 >> 24;
-      c2 = (char) w2 >> 24;
-#else /* not  __ARMEB__ */
-      c1 = (char) t1;
-      c2 = (char) w2;
-#endif /* not  __ARMEB__ */
-      t1 RSHIFT= 8;
-      w2 RSHIFT= 8;
-    } while (c1 != 0 && c1 == c2);
-  return c1 - c2;
-#endif /* 0 */
-
-
-        wp1 .req r0
-        wp2 .req r1
-        b1  .req r2
-        w1  .req r4
-        w2  .req r5
-        t1  .req ip
-        @ r3 is scratch
-
-        /* First of all, compare bytes until wp1(sp1) is word-aligned. */
-1:
-        tst     wp1, #3
-        beq     2f
-        ldrb    r2, [wp1], #1
-        ldrb    r3, [wp2], #1
-        cmp     r2, #1
-        it      cs
-        cmpcs   r2, r3
-        beq     1b
-        sub     r0, r2, r3
-        RETURN
-
-2:
-        str     r5, [sp, #-4]!
-        str     r4, [sp, #-4]!
-        //stmfd   sp!, {r4, r5}
-        mov     b1, #1
-        orr     b1, b1, b1, lsl #8
-        orr     b1, b1, b1, lsl #16
-
-        and     t1, wp2, #3
-        bic     wp2, wp2, #3
-        ldr     w1, [wp1], #4
-        ldr     w2, [wp2], #4
-        cmp     t1, #2
-        beq     2f
-        bhi     3f
-
-        /* Critical inner Loop: Block with 3 bytes initial overlap */
-        .p2align        2
-1:
-        bic     t1, w1, MSB
-        cmp     t1, w2, S2LOMEM #8
-        sub     r3, w1, b1
-        bic     r3, r3, w1
-        bne     4f
-        ands    r3, r3, b1, lsl #7
-        it      eq
-        ldreq   w2, [wp2], #4
-        bne     5f
-        eor     t1, t1, w1
-        cmp     t1, w2, S2HIMEM #24
-        bne     6f
-        ldr     w1, [wp1], #4
-        b       1b
-4:
-        S2LOMEM        w2, w2, #8
-        b       8f
-
-5:
-#ifdef __ARMEB__
-        /* The syndrome value may contain false ones if the string ends
-        with the bytes 0x01 0x00 */
-        tst     w1, #0xff000000
-        itt     ne
-        tstne   w1, #0x00ff0000
-        tstne   w1, #0x0000ff00
-        beq     7f
-#else
-        bics    r3, r3, #0xff000000
-        bne     7f
-#endif
-        ldrb    w2, [wp2]
-        S2LOMEM  t1, w1, #24
-#ifdef __ARMEB__
-        lsl     w2, w2, #24
-#endif
-        b       8f
-
-6:
-        S2LOMEM  t1, w1, #24
-        and     w2, w2, LSB
-        b       8f
-
-        /* Critical inner Loop: Block with 2 bytes initial overlap */
-        .p2align        2
-2:
-        S2HIMEM  t1, w1, #16
-        sub     r3, w1, b1
-        S2LOMEM  t1, t1, #16
-        bic     r3, r3, w1
-        cmp     t1, w2, S2LOMEM #16
-        bne     4f
-        ands    r3, r3, b1, lsl #7
-        it      eq
-        ldreq   w2, [wp2], #4
-        bne     5f
-        eor     t1, t1, w1
-        cmp     t1, w2, S2HIMEM #16
-        bne     6f
-        ldr     w1, [wp1], #4
-        b       2b
-
-5:
-#ifdef __ARMEB__
-        /* The syndrome value may contain false ones if the string ends
-        with the bytes 0x01 0x00 */
-        tst     w1, #0xff000000
-        it      ne
-        tstne   w1, #0x00ff0000
-        beq     7f
-#else
-        lsls    r3, r3, #16
-        bne     7f
-#endif
-        ldrh    w2, [wp2]
-        S2LOMEM  t1, w1, #16
-#ifdef __ARMEB__
-        lsl     w2, w2, #16
-#endif
-        b       8f
-
-6:
-        S2HIMEM  w2, w2, #16
-        S2LOMEM  t1, w1, #16
-4:
-        S2LOMEM  w2, w2, #16
-        b       8f
-
-        /* Critical inner Loop: Block with 1 byte initial overlap */
-        .p2align        2
-3:
-        and     t1, w1, LSB
-        cmp     t1, w2, S2LOMEM #24
-        sub     r3, w1, b1
-        bic     r3, r3, w1
-        bne     4f
-        ands    r3, r3, b1, lsl #7
-        it      eq
-        ldreq   w2, [wp2], #4
-        bne     5f
-        eor     t1, t1, w1
-        cmp     t1, w2, S2HIMEM #8
-        bne     6f
-        ldr     w1, [wp1], #4
-        b       3b
-4:
-        S2LOMEM  w2, w2, #24
-        b       8f
-5:
-        /* The syndrome value may contain false ones if the string ends
-        with the bytes 0x01 0x00 */
-        tst     w1, LSB
-        beq     7f
-        ldr     w2, [wp2], #4
-6:
-        S2LOMEM  t1, w1, #8
-        bic     w2, w2, MSB
-        b       8f
-7:
-        mov     r0, #0
-        //ldmfd   sp!, {r4, r5}
-        ldr     r4, [sp], #4
-        ldr     r5, [sp], #4
-        RETURN
-8:
-        and     r2, t1, LSB
-        and     r0, w2, LSB
-        cmp     r0, #1
-        it      cs
-        cmpcs   r0, r2
-        itt     eq
-        S2LOMEMEQ        t1, t1, #8
-        S2LOMEMEQ        w2, w2, #8
-        beq     8b
-        sub     r0, r2, r0
-        //ldmfd   sp!, {r4, r5}
-        ldr     r4, [sp], #4
-        ldr     r5, [sp], #4
-        RETURN
-
-#endif /* !(defined (_ISA_THUMB_2) || defined (_ISA_ARM_6)
-            defined (__OPTIMIZE_SIZE__) || defined (PREFER_SIZE_OVER_SPEED) ||
-            (defined (__thumb__) && !defined (__thumb2__))) */

libc/arch-arm/bionic/syscall.S

@@ -26,12 +26,16 @@
  * SUCH DAMAGE.
  */
 
-#include <linux/err.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 ENTRY(syscall)
     mov     ip, sp
     stmfd   sp!, {r4, r5, r6, r7}
+    .cfi_def_cfa_offset 16
+    .cfi_rel_offset r4, 0
+    .cfi_rel_offset r5, 4
+    .cfi_rel_offset r6, 8
+    .cfi_rel_offset r7, 12
     mov     r7, r0
     mov     r0, r1
     mov     r1, r2
@@ -39,6 +43,7 @@ ENTRY(syscall)
     ldmfd   ip, {r3, r4, r5, r6}
     swi     #0
     ldmfd   sp!, {r4, r5, r6, r7}
+    .cfi_def_cfa_offset 0
     cmn     r0, #(MAX_ERRNO + 1)
     bxls    lr
     neg     r0, r0

libc/arch-arm/bionic/tgkill.S

@@ -1,51 +0,0 @@
-/*
- * Copyright (C) 2008 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-#include <linux/err.h>
-#include <asm/unistd.h>
-#include <machine/asm.h>
-
-/* unlike our auto-generated syscall stubs, this code saves lr
-   on the stack, as well as a few other registers. this makes
-   our stack unwinder happy, when we generate debug stack
-   traces after the C library or other parts of the system
-   abort due to a fatal runtime error (e.g. detection
-   of a corrupted malloc heap).
-*/
-
-ENTRY(tgkill)
-    .save   {r4-r7, ip, lr}
-    stmfd   sp!, {r4-r7, ip, lr}
-    ldr     r7, =__NR_tgkill
-    swi     #0
-    ldmfd   sp!, {r4-r7, ip, lr}
-    cmn     r0, #(MAX_ERRNO + 1)
-    bxls    lr
-    neg     r0, r0
-    b       __set_errno
-END(tgkill)

libc/arch-arm/bionic/tkill.S

@@ -1,50 +0,0 @@
-/*
- * Copyright (C) 2008 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-
-#include <linux/err.h>
-#include <asm/unistd.h>
-#include <machine/asm.h>
-
-/* unlike our auto-generated syscall stubs, this code saves lr
-   on the stack, as well as a few other registers. this makes
-   our stack unwinder happy, when we generate debug stack
-   traces after the C library or other parts of the system
-   abort due to a fatal runtime error (e.g. detection
-   of a corrupted malloc heap).
-*/
-
-ENTRY(tkill)
-    stmfd   sp!, {r4-r7, ip, lr}
-    ldr     r7, =__NR_tkill
-    swi     #0
-    ldmfd   sp!, {r4-r7, ip, lr}
-    cmn     r0, #(MAX_ERRNO + 1)
-    bxls    lr
-    neg     r0, r0
-    b       __set_errno
-END(tkill)

libc/arch-arm/cortex-a15/bionic/__strcat_chk.S

@@ -26,8 +26,8 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
     .syntax unified
 
@@ -38,16 +38,17 @@
 // Check that the two lengths together don't exceed the threshold, then
 // do a memcpy of the data.
 ENTRY(__strcat_chk)
-    .cfi_startproc
     pld     [r0, #0]
     push    {r0, lr}
+    .save   {r0, lr}
     .cfi_def_cfa_offset 8
     .cfi_rel_offset r0, 0
     .cfi_rel_offset lr, 4
     push    {r4, r5}
+    .save   {r4, r5}
     .cfi_adjust_cfa_offset 8
     .cfi_rel_offset r4, 0
-    .cfi_rel_offset r5, 0
+    .cfi_rel_offset r5, 4
 
     mov     lr, r2
 
@@ -178,7 +179,7 @@ ENTRY(__strcat_chk)
 .L_strlen_done:
     add     r2, r3, r4
     cmp     r2, lr
-    bgt     .L_fortify_check_failed
+    bhi     __strcat_chk_failed
 
     // Set up the registers for the memcpy code.
     mov     r1, r5
@@ -186,13 +187,20 @@ ENTRY(__strcat_chk)
     mov     r2, r4
     add     r0, r0, r3
     pop     {r4, r5}
-    .cfi_adjust_cfa_offset -8
-    .cfi_restore r4
-    .cfi_restore r5
+END(__strcat_chk)
 
-    #include "memcpy_base.S"
+#define MEMCPY_BASE         __strcat_chk_memcpy_base
+#define MEMCPY_BASE_ALIGNED __strcat_chk_memcpy_base_aligned
 
-.L_fortify_check_failed:
+#include "memcpy_base.S"
+
+ENTRY_PRIVATE(__strcat_chk_failed)
+    .save   {r0, lr}
+    .save   {r4, r5}
+
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
     .cfi_adjust_cfa_offset 8
     .cfi_rel_offset r4, 0
     .cfi_rel_offset r5, 4
@@ -206,10 +214,8 @@ error_code:
     .word   BIONIC_EVENT_STRCAT_BUFFER_OVERFLOW
 error_message:
     .word   error_string-(1b+4)
-
-    .cfi_endproc
-END(__strcat_chk)
+END(__strcat_chk_failed)
 
     .data
 error_string:
-    .string "strcat buffer overflow"
+    .string "strcat: prevented write past end of buffer"

libc/arch-arm/cortex-a15/bionic/__strcpy_chk.S

@@ -26,8 +26,8 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
     .syntax unified
 
@@ -37,9 +37,9 @@
 // Get the length of the source string first, then do a memcpy of the data
 // instead of a strcpy.
 ENTRY(__strcpy_chk)
-    .cfi_startproc
     pld     [r0, #0]
     push    {r0, lr}
+    .save   {r0, lr}
     .cfi_def_cfa_offset 8
     .cfi_rel_offset r0, 0
     .cfi_rel_offset lr, 4
@@ -150,14 +150,22 @@ ENTRY(__strcpy_chk)
     pld     [r1, #64]
     ldr     r0, [sp]
     cmp     r3, lr
-    bge     .L_fortify_check_failed
+    bhs     __strcpy_chk_failed
 
     // Add 1 for copy length to get the string terminator.
     add     r2, r3, #1
+END(__strcpy_chk)
 
-    #include "memcpy_base.S"
+#define MEMCPY_BASE         __strcpy_chk_memcpy_base
+#define MEMCPY_BASE_ALIGNED __strcpy_chk_memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY_PRIVATE(__strcpy_chk_failed)
+    .save   {r0, lr}
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
 
-.L_fortify_check_failed:
     ldr     r0, error_message
     ldr     r1, error_code
 1:
@@ -167,10 +175,8 @@ error_code:
     .word   BIONIC_EVENT_STRCPY_BUFFER_OVERFLOW
 error_message:
     .word   error_string-(1b+4)
-
-    .cfi_endproc
-END(__strcpy_chk)
+END(__strcpy_chk_failed)
 
     .data
 error_string:
-    .string "strcpy buffer overflow"
+    .string "strcpy: prevented write past end of buffer"

libc/arch-arm/cortex-a15/bionic/memcpy.S

@@ -55,38 +55,37 @@
 
 // Prototype: void *memcpy (void *dst, const void *src, size_t count).
 
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
         .text
         .syntax unified
         .fpu    neon
 
 ENTRY(__memcpy_chk)
-        .cfi_startproc
         cmp     r2, r3
-        bgt     __memcpy_chk_fail
+        bhi     __memcpy_chk_fail
 
         // Fall through to memcpy...
-        .cfi_endproc
 END(__memcpy_chk)
 
 ENTRY(memcpy)
-        .cfi_startproc
         pld     [r1, #64]
         push    {r0, lr}
+        .save   {r0, lr}
         .cfi_def_cfa_offset 8
         .cfi_rel_offset r0, 0
         .cfi_rel_offset lr, 4
-
-        #include "memcpy_base.S"
-        .cfi_endproc
 END(memcpy)
 
-        .cfi_startproc
-__memcpy_chk_fail:
+#define MEMCPY_BASE         __memcpy_base
+#define MEMCPY_BASE_ALIGNED __memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY_PRIVATE(__memcpy_chk_fail)
         // Preserve lr for backtrace.
         push    {lr}
+        .save   {lr}
         .cfi_def_cfa_offset 4
         .cfi_rel_offset lr, 0
 
@@ -99,8 +98,8 @@ error_code:
         .word   BIONIC_EVENT_MEMCPY_BUFFER_OVERFLOW
 error_message:
         .word   error_string-(1b+8)
-        .cfi_endproc
+END(__memcpy_chk_fail)
 
         .data
 error_string:
-        .string "memcpy buffer overflow"
+        .string "memcpy: prevented write past end of buffer"

libc/arch-arm/cortex-a15/bionic/memcpy_base.S

@@ -53,6 +53,12 @@
  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+ENTRY_PRIVATE(MEMCPY_BASE)
+        .save   {r0, lr}
+        .cfi_def_cfa_offset 8
+        .cfi_rel_offset r0, 0
+        .cfi_rel_offset lr, 4
+
         // Assumes that n >= 0, and dst, src are valid pointers.
         // For any sizes less than 832 use the neon code that doesn't
         // care about the src alignment. This avoids any checks
@@ -67,8 +73,10 @@
         cmp     r2, #16
         blo     .L_copy_less_than_16_unknown_align
 
-        cmp     r2, #832
-        bge     .L_check_alignment
+        // TODO: The aligned copy code is extremely slow copying some large
+        //       buffers so always go through the unaligned path for now.
+        //cmp     r2, #832
+        //bge     .L_check_alignment
 
 .L_copy_unknown_alignment:
         // Unknown alignment of src and dst.
@@ -161,21 +169,32 @@
         eor     r3, r0, r1
         ands    r3, r3, #0x3
         bne     .L_copy_unknown_alignment
+END(MEMCPY_BASE)
+
+ENTRY_PRIVATE(MEMCPY_BASE_ALIGNED)
+        .save   {r0, lr}
+        .cfi_def_cfa_offset 8
+        .cfi_rel_offset r0, 0
+        .cfi_rel_offset lr, 4
 
         // To try and improve performance, stack layout changed,
         // i.e., not keeping the stack looking like users expect
         // (highest numbered register at highest address).
-        // TODO: Add debug frame directives.
-        // We don't need exception unwind directives, because the code below
-        // does not throw any exceptions and does not call any other functions.
-        // Generally, newlib functions like this lack debug information for
-        // assembler source.
-        .save   {r4, r5}
         strd    r4, r5, [sp, #-8]!
-        .save   {r6, r7}
+        .save   {r4, r5}
+        .cfi_adjust_cfa_offset 8
+        .cfi_rel_offset r4, 0
+        .cfi_rel_offset r5, 4
         strd    r6, r7, [sp, #-8]!
-        .save   {r8, r9}
+        .save   {r6, r7}
+        .cfi_adjust_cfa_offset 8
+        .cfi_rel_offset r6, 0
+        .cfi_rel_offset r7, 0
         strd    r8, r9, [sp, #-8]!
+        .save   {r8, r9}
+        .cfi_adjust_cfa_offset 8
+        .cfi_rel_offset r8, 0
+        .cfi_rel_offset r9, 4
 
         // Optimized for already aligned dst code.
         ands    ip, r0, #3
@@ -301,3 +320,4 @@
 
         // Src is guaranteed to be at least word aligned by this point.
         b       .L_word_aligned
+END(MEMCPY_BASE_ALIGNED)

libc/arch-arm/cortex-a15/bionic/memset.S

@@ -27,8 +27,8 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
         /*
          * Optimized memset() for ARM.
@@ -40,11 +40,11 @@
         .syntax     unified
 
 ENTRY(__memset_chk)
-        .cfi_startproc
         cmp         r2, r3
         bls         .L_done
 
         // Preserve lr for backtrace.
+        .save       {lr}
         push        {lr}
         .cfi_def_cfa_offset 4
         .cfi_rel_offset lr, 0
@@ -58,21 +58,17 @@ error_code:
         .word       BIONIC_EVENT_MEMSET_BUFFER_OVERFLOW
 error_message:
         .word       error_string-(1b+8)
-
-        .cfi_endproc
 END(__memset_chk)
 
 ENTRY(bzero)
-        .cfi_startproc
         mov         r2, r1
         mov         r1, #0
 .L_done:
         // Fall through to memset...
-        .cfi_endproc
 END(bzero)
 
 ENTRY(memset)
-        .cfi_startproc
+        .save       {r0}
         stmfd       sp!, {r0}
         .cfi_def_cfa_offset 4
         .cfi_rel_offset r0, 0
@@ -190,9 +186,8 @@ ENTRY(memset)
         strbcs      r1, [r0], #1
         ldmfd       sp!, {r0}
         bx          lr
-        .cfi_endproc
 END(memset)
 
         .data
 error_string:
-        .string     "memset buffer overflow"
+        .string     "memset: prevented write past end of buffer"

libc/arch-arm/cortex-a15/bionic/strcat.S

@@ -53,7 +53,7 @@
  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
     .syntax unified
 

libc/arch-arm/cortex-a15/bionic/strcmp.S

@@ -27,7 +27,7 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 #ifdef __ARMEB__
 #define S2LOMEM lsl
@@ -123,8 +123,13 @@ ENTRY(strcmp)
         .macro  init
         /* Macro to save temporary registers and prepare magic values.  */
         subs    sp, sp, #16
+        .cfi_def_cfa_offset 16
         strd    r4, r5, [sp, #8]
+        .cfi_rel_offset r4, 0
+        .cfi_rel_offset r5, 4
         strd    r6, r7, [sp]
+        .cfi_rel_offset r6, 8
+        .cfi_rel_offset r7, 12
         mvn     r6, #0  /* all F */
         mov     r7, #0  /* all 0 */
         .endm   /* init */
@@ -140,19 +145,8 @@ ENTRY(strcmp)
 
         .macro  magic_find_zero_bytes w1
         /* Macro to find all-zero bytes in w1, result is in ip.  */
-#if (defined (__ARM_FEATURE_DSP))
         uadd8   ip, \w1, r6
         sel     ip, r7, r6
-#else /* not defined (__ARM_FEATURE_DSP) */
-        /* __ARM_FEATURE_DSP is not defined for some Cortex-M processors.
-        Coincidently, these processors only have Thumb-2 mode, where we can use the
-        the (large) magic constant available directly as an immediate in instructions.
-        Note that we cannot use the magic constant in ARM mode, where we need
-        to create the constant in a register.  */
-        sub     ip, \w1, #0x01010101
-        bic     ip, ip, \w1
-        and     ip, ip, #0x80808080
-#endif /* not defined (__ARM_FEATURE_DSP) */
         .endm /* magic_find_zero_bytes */
 
         .macro  setup_return w1 w2
@@ -171,12 +165,13 @@ ENTRY(strcmp)
         /* Are both strings double-word aligned?  */
         orr     ip, r0, r1
         tst     ip, #7
-        bne     do_align
+        bne     .L_do_align
 
         /* Fast path.  */
+        .save   {r4-r7}
         init
 
-doubleword_aligned:
+.L_doubleword_aligned:
 
         /* Get here when the strings to compare are double-word aligned.  */
         /* Compare two words in every iteration.  */
@@ -189,14 +184,14 @@ doubleword_aligned:
         ldrd    r2, r3, [r0], #8
         ldrd    r4, r5, [r1], #8
 
-        magic_compare_and_branch w1=r2, w2=r4, label=return_24
-        magic_compare_and_branch w1=r3, w2=r5, label=return_35
+        magic_compare_and_branch w1=r2, w2=r4, label=.L_return_24
+        magic_compare_and_branch w1=r3, w2=r5, label=.L_return_35
         b       2b
 
-do_align:
+.L_do_align:
         /* Is the first string word-aligned?  */
         ands    ip, r0, #3
-        beq     word_aligned_r0
+        beq     .L_word_aligned_r0
 
         /* Fast compare byte by byte until the first string is word-aligned.  */
         /* The offset of r0 from a word boundary is in ip. Thus, the number of bytes
@@ -204,58 +199,58 @@ do_align:
         bic     r0, r0, #3
         ldr     r2, [r0], #4
         lsls    ip, ip, #31
-        beq     byte2
-        bcs     byte3
+        beq     .L_byte2
+        bcs     .L_byte3
 
-byte1:
+.L_byte1:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE1_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbz   reg=r3, label=fast_return
+        bne     .L_fast_return
+        m_cbz   reg=r3, label=.L_fast_return
 
-byte2:
+.L_byte2:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE2_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbz   reg=r3, label=fast_return
+        bne     .L_fast_return
+        m_cbz   reg=r3, label=.L_fast_return
 
-byte3:
+.L_byte3:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE3_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbnz  reg=r3, label=word_aligned_r0
+        bne     .L_fast_return
+        m_cbnz  reg=r3, label=.L_word_aligned_r0
 
-fast_return:
+.L_fast_return:
         mov     r0, ip
         bx      lr
 
-word_aligned_r0:
+.L_word_aligned_r0:
         init
         /* The first string is word-aligned.  */
         /* Is the second string word-aligned?  */
         ands    ip, r1, #3
-        bne     strcmp_unaligned
+        bne     .L_strcmp_unaligned
 
-word_aligned:
+.L_word_aligned:
         /* The strings are word-aligned. */
         /* Is the first string double-word aligned?  */
         tst     r0, #4
-        beq     doubleword_aligned_r0
+        beq     .L_doubleword_aligned_r0
 
         /* If r0 is not double-word aligned yet, align it by loading
         and comparing the next word from each string.  */
         ldr     r2, [r0], #4
         ldr     r4, [r1], #4
-        magic_compare_and_branch w1=r2 w2=r4 label=return_24
+        magic_compare_and_branch w1=r2 w2=r4 label=.L_return_24
 
-doubleword_aligned_r0:
+.L_doubleword_aligned_r0:
         /* Get here when r0 is double-word aligned.  */
         /* Is r1 doubleword_aligned?  */
         tst     r1, #4
-        beq     doubleword_aligned
+        beq     .L_doubleword_aligned
 
         /* Get here when the strings to compare are word-aligned,
         r0 is double-word aligned, but r1 is not double-word aligned.  */
@@ -271,9 +266,9 @@ doubleword_aligned_r0:
 
         /* Load the next double-word from each string and compare.  */
         ldrd    r2, r3, [r0], #8
-        magic_compare_and_branch w1=r2 w2=r5 label=return_25
+        magic_compare_and_branch w1=r2 w2=r5 label=.L_return_25
         ldrd    r4, r5, [r1], #8
-        magic_compare_and_branch w1=r3 w2=r4 label=return_34
+        magic_compare_and_branch w1=r3 w2=r4 label=.L_return_34
         b       3b
 
         .macro miscmp_word offsetlo offsethi
@@ -297,47 +292,47 @@ doubleword_aligned_r0:
         and     r2, r3, r6, S2LOMEM #\offsetlo
         it      eq
         cmpeq   r2, r5
-        bne     return_25
+        bne     .L_return_25
         ldr     r5, [r1], #4
         cmp     ip, #0
         eor r3, r2, r3
         S2HIMEM r2, r5, #\offsethi
         it      eq
         cmpeq   r3, r2
-        bne     return_32
+        bne     .L_return_32
         b       7b
         .endm /* miscmp_word */
 
-strcmp_unaligned:
+.L_strcmp_unaligned:
         /* r0 is word-aligned, r1 is at offset ip from a word.  */
         /* Align r1 to the (previous) word-boundary.  */
         bic     r1, r1, #3
 
         /* Unaligned comparison word by word using LDRs. */
         cmp     ip, #2
-        beq     miscmp_word_16                    /* If ip == 2.  */
-        bge     miscmp_word_24                    /* If ip == 3.  */
+        beq     .L_miscmp_word_16                 /* If ip == 2.  */
+        bge     .L_miscmp_word_24                 /* If ip == 3.  */
         miscmp_word offsetlo=8 offsethi=24        /* If ip == 1.  */
-miscmp_word_16:  miscmp_word offsetlo=16 offsethi=16
-miscmp_word_24:  miscmp_word offsetlo=24 offsethi=8
+.L_miscmp_word_16:  miscmp_word offsetlo=16 offsethi=16
+.L_miscmp_word_24:  miscmp_word offsetlo=24 offsethi=8
 
 
-return_32:
+.L_return_32:
         setup_return w1=r3, w2=r2
-        b       do_return
-return_34:
+        b       .L_do_return
+.L_return_34:
         setup_return w1=r3, w2=r4
-        b       do_return
-return_25:
+        b       .L_do_return
+.L_return_25:
         setup_return w1=r2, w2=r5
-        b       do_return
-return_35:
+        b       .L_do_return
+.L_return_35:
         setup_return w1=r3, w2=r5
-        b       do_return
-return_24:
+        b       .L_do_return
+.L_return_24:
         setup_return w1=r2, w2=r4
 
-do_return:
+.L_do_return:
 
 #ifdef __ARMEB__
         mov     r0, ip
@@ -349,11 +344,16 @@ do_return:
         ldrd    r6, r7, [sp]
         ldrd    r4, r5, [sp, #8]
         adds    sp, sp, #16
+        .cfi_def_cfa_offset 0
+        .cfi_restore r4
+        .cfi_restore r5
+        .cfi_restore r6
+        .cfi_restore r7
 
         /* There is a zero or a different byte between r1 and r2.  */
         /* r0 contains a mask of all-zero bytes in r1.  */
         /* Using r0 and not ip here because cbz requires low register.  */
-        m_cbz   reg=r0, label=compute_return_value
+        m_cbz   reg=r0, label=.L_compute_return_value
         clz     r0, r0
         /* r0 contains the number of bits on the left of the first all-zero byte in r1.  */
         rsb     r0, r0, #24
@@ -361,7 +361,7 @@ do_return:
         lsr     r1, r1, r0
         lsr     r2, r2, r0
 
-compute_return_value:
+.L_compute_return_value:
         movs    r0, #1
         cmp     r1, r2
         /* The return value is computed as follows.

libc/arch-arm/cortex-a15/bionic/strcpy.S

@@ -53,7 +53,7 @@
  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
     .syntax unified
 

libc/arch-arm/cortex-a15/bionic/strlen.S

@@ -53,7 +53,7 @@
  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
     .syntax unified
 

libc/arch-arm/cortex-a15/cortex-a15.mk

@@ -1,10 +1,9 @@
-$(call libc-add-cpu-variant-src,MEMCPY,arch-arm/cortex-a15/bionic/memcpy.S)
-$(call libc-add-cpu-variant-src,MEMSET,arch-arm/cortex-a15/bionic/memset.S)
-$(call libc-add-cpu-variant-src,STRCAT,arch-arm/cortex-a15/bionic/strcat.S)
-$(call libc-add-cpu-variant-src,STRCMP,arch-arm/cortex-a15/bionic/strcmp.S)
-$(call libc-add-cpu-variant-src,STRCPY,arch-arm/cortex-a15/bionic/strcpy.S)
-$(call libc-add-cpu-variant-src,STRLEN,arch-arm/cortex-a15/bionic/strlen.S)
-$(call libc-add-cpu-variant-src,__STRCAT_CHK,arch-arm/cortex-a15/bionic/__strcat_chk.S)
-$(call libc-add-cpu-variant-src,__STRCPY_CHK,arch-arm/cortex-a15/bionic/__strcpy_chk.S)
-
-include bionic/libc/arch-arm/generic/generic.mk
+libc_bionic_src_files_arm += \
+    arch-arm/cortex-a15/bionic/memcpy.S \
+    arch-arm/cortex-a15/bionic/memset.S \
+    arch-arm/cortex-a15/bionic/strcat.S \
+    arch-arm/cortex-a15/bionic/strcmp.S \
+    arch-arm/cortex-a15/bionic/strcpy.S \
+    arch-arm/cortex-a15/bionic/strlen.S \
+    arch-arm/cortex-a15/bionic/__strcat_chk.S \
+    arch-arm/cortex-a15/bionic/__strcpy_chk.S \

libc/arch-arm/cortex-a9/bionic/__strcat_chk.S

@@ -26,8 +26,8 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
     .syntax unified
     .fpu    neon
@@ -38,13 +38,14 @@
 // Check that the two lengths together don't exceed the threshold, then
 // do a memcpy of the data.
 ENTRY(__strcat_chk)
-    .cfi_startproc
     pld     [r0, #0]
     push    {r0, lr}
+    .save   {r0, lr}
     .cfi_def_cfa_offset 8
     .cfi_rel_offset r0, 0
     .cfi_rel_offset lr, 4
     push    {r4, r5}
+    .save   {r4, r5}
     .cfi_adjust_cfa_offset 8
     .cfi_rel_offset r4, 0
     .cfi_rel_offset r5, 4
@@ -181,7 +182,7 @@ ENTRY(__strcat_chk)
 .L_strlen_done:
     add     r2, r3, r4
     cmp     r2, lr
-    bgt     .L_fortify_check_failed
+    bhi     __strcat_chk_fail
 
     // Set up the registers for the memcpy code.
     mov     r1, r5
@@ -189,13 +190,20 @@ ENTRY(__strcat_chk)
     mov     r2, r4
     add     r0, r0, r3
     pop     {r4, r5}
-    .cfi_adjust_cfa_offset -8
-    .cfi_restore r4
-    .cfi_restore r5
 
-    #include "memcpy_base.S"
+    // Fall through into the memcpy_base function.
+END(__strcat_chk)
 
-.L_fortify_check_failed:
+#define MEMCPY_BASE         __strcat_chk_memcpy_base
+#define MEMCPY_BASE_ALIGNED __strcat_chk_memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY_PRIVATE(__strcat_chk_fail)
+    .save   {r0, lr}
+    .save   {r4, r5}
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
     .cfi_adjust_cfa_offset 8
     .cfi_rel_offset r4, 0
     .cfi_rel_offset r5, 4
@@ -209,10 +217,8 @@ error_code:
     .word   BIONIC_EVENT_STRCAT_BUFFER_OVERFLOW
 error_message:
     .word   error_string-(1b+4)
-
-    .cfi_endproc
-END(__strcat_chk)
+END(__strcat_chk_fail)
 
     .data
 error_string:
-    .string "strcat buffer overflow"
+    .string "strcat: prevented write past end of buffer"

libc/arch-arm/cortex-a9/bionic/__strcpy_chk.S

@@ -26,8 +26,8 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
     .syntax unified
     .fpu    neon
@@ -37,9 +37,9 @@
 // Get the length of the source string first, then do a memcpy of the data
 // instead of a strcpy.
 ENTRY(__strcpy_chk)
-    .cfi_startproc
     pld     [r0, #0]
     push    {r0, lr}
+    .save   {r0, lr}
     .cfi_def_cfa_offset 8
     .cfi_rel_offset r0, 0
     .cfi_rel_offset lr, 4
@@ -152,27 +152,36 @@ ENTRY(__strcpy_chk)
     pld     [r1, #64]
     ldr     r0, [sp]
     cmp     r3, lr
-    bge     .L_fortify_check_failed
+    bhs     __strcpy_chk_fail
 
     // Add 1 for copy length to get the string terminator.
     add     r2, r3, #1
 
-    #include "memcpy_base.S"
+    // Fall through into the memcpy_base function.
+END(__strcpy_chk)
+
+#define MEMCPY_BASE         __strcpy_chk_memcpy_base
+#define MEMCPY_BASE_ALIGNED __strcpy_chk_memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY_PRIVATE(__strcpy_chk_fail)
+    .save   {r0, lr}
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
 
-.L_fortify_check_failed:
     ldr     r0, error_message
     ldr     r1, error_code
 1:
     add     r0, pc
     bl      __fortify_chk_fail
+
 error_code:
     .word   BIONIC_EVENT_STRCPY_BUFFER_OVERFLOW
 error_message:
     .word   error_string-(1b+4)
-
-    .cfi_endproc
-END(__strcpy_chk)
+END(__strcpy_chk_fail)
 
     .data
 error_string:
-    .string "strcpy buffer overflow"
+    .string "strcpy: prevented write past end of buffer"

libc/arch-arm/cortex-a9/bionic/memcpy.S

@@ -26,8 +26,8 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
 /*
  * This code assumes it is running on a processor that supports all arm v7
@@ -41,31 +41,30 @@
         .thumb_func
 
 ENTRY(__memcpy_chk)
-        .cfi_startproc
         cmp         r2, r3
-        bgt         __memcpy_chk_fail
+        bhi         __memcpy_chk_fail
 
         // Fall through to memcpy...
-        .cfi_endproc
 END(__memcpy_chk)
 
 ENTRY(memcpy)
-        .cfi_startproc
         pld     [r1, #0]
         stmfd   sp!, {r0, lr}
+        .save   {r0, lr}
         .cfi_def_cfa_offset 8
         .cfi_rel_offset r0, 0
         .cfi_rel_offset lr, 4
         pld     [r1, #64]
-
-        #include "memcpy_base.S"
-        .cfi_endproc
 END(memcpy)
 
-        .cfi_startproc
-__memcpy_chk_fail:
+#define MEMCPY_BASE         __memcpy_base
+#define MEMCPY_BASE_ALIGNED __memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY_PRIVATE(__memcpy_chk_fail)
         // Preserve lr for backtrace.
         push    {lr}
+        .save   {lr}
         .cfi_def_cfa_offset 4
         .cfi_rel_offset lr, 0
 
@@ -78,8 +77,8 @@ error_code:
         .word   BIONIC_EVENT_MEMCPY_BUFFER_OVERFLOW
 error_message:
         .word   error_string-(1b+4)
-        .cfi_endproc
+END(__memcpy_chk_fail)
 
         .data
 error_string:
-        .string     "memcpy buffer overflow"
+        .string     "memcpy: prevented write past end of buffer"

libc/arch-arm/cortex-a9/bionic/memcpy_base.S

@@ -32,15 +32,20 @@
  * cache line.
  */
 
+ENTRY_PRIVATE(MEMCPY_BASE)
+        .save       {r0, lr}
+        .cfi_def_cfa_offset 8
+        .cfi_rel_offset r0, 0
+        .cfi_rel_offset lr, 4
+
         // Check so divider is at least 16 bytes, needed for alignment code.
         cmp         r2, #16
         blo         5f
 
-
         /* check if buffers are aligned. If so, run arm-only version */
         eor         r3, r0, r1
         ands        r3, r3, #0x3
-        beq         11f
+        beq         __memcpy_base_aligned
 
         /* Check the upper size limit for Neon unaligned memory access in memcpy */
         cmp         r2, #224
@@ -131,9 +136,23 @@
 
         ldmfd       sp!, {r0, lr}
         bx          lr
-11:
+END(MEMCPY_BASE)
+
+ENTRY_PRIVATE(MEMCPY_BASE_ALIGNED)
+        .save       {r0, lr}
+        .cfi_def_cfa_offset 8
+        .cfi_rel_offset r0, 0
+        .cfi_rel_offset lr, 4
+
         /* Simple arm-only copy loop to handle aligned copy operations */
-        stmfd       sp!, {r4, r5, r6, r7, r8}
+        stmfd       sp!, {r4-r8}
+        .save       {r4-r8}
+        .cfi_adjust_cfa_offset 20
+        .cfi_rel_offset r4, 0
+        .cfi_rel_offset r5, 4
+        .cfi_rel_offset r6, 8
+        .cfi_rel_offset r7, 12
+        .cfi_rel_offset r8, 16
         pld         [r1, #(32 * 4)]
 
         /* Check alignment */
@@ -202,5 +221,6 @@
         ldrbne      r3, [r1]                /*  last byte  */
         strbne      r3, [r0]
 6:
-        ldmfd       sp!, {r4, r5, r6, r7, r8}
+        ldmfd       sp!, {r4-r8}
         ldmfd       sp!, {r0, pc}
+END(MEMCPY_BASE_ALIGNED)

libc/arch-arm/cortex-a9/bionic/memset.S

@@ -26,9 +26,8 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/cpu-features.h>
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
 /*
  * This code assumes it is running on a processor that supports all arm v7
@@ -38,12 +37,12 @@
     .fpu    neon
 
 ENTRY(__memset_chk)
-        .cfi_startproc
         cmp         r2, r3
         bls         .L_done
 
         // Preserve lr for backtrace.
         push        {lr}
+        .save       {lr}
         .cfi_def_cfa_offset 4
         .cfi_rel_offset lr, 0
 
@@ -56,28 +55,24 @@ error_code:
         .word       BIONIC_EVENT_MEMSET_BUFFER_OVERFLOW
 error_message:
         .word       error_string-(1b+8)
-
-        .cfi_endproc
 END(__memset_chk)
 
 ENTRY(bzero)
-        .cfi_startproc
         mov     r2, r1
         mov     r1, #0
 
 .L_done:
         // Fall through to memset...
-        .cfi_endproc
 END(bzero)
 
 /* memset() returns its first argument.  */
 ENTRY(memset)
-        .cfi_startproc
-        # The neon memset only wins for less than 132.
+        // The neon memset only wins for less than 132.
         cmp         r2, #132
-        bhi         11f
+        bhi         __memset_large_copy
 
         stmfd       sp!, {r0}
+        .save       {r0}
         .cfi_def_cfa_offset 4
         .cfi_rel_offset r0, 0
 
@@ -112,12 +107,14 @@ ENTRY(memset)
         strcsb      r1, [r0], #1
         ldmfd       sp!, {r0}
         bx          lr
-11:
+END(memset)
+
+ENTRY_PRIVATE(__memset_large_copy)
         /* compute the offset to align the destination
          * offset = (4-(src&3))&3 = -src & 3
          */
-
         stmfd       sp!, {r0, r4-r7, lr}
+        .save       {r0, r4-r7, lr}
         .cfi_def_cfa_offset 24
         .cfi_rel_offset r0, 0
         .cfi_rel_offset r4, 4
@@ -187,9 +184,8 @@ ENTRY(memset)
         strcsb      r1, [r0]
         ldmfd       sp!, {r0, r4-r7, lr}
         bx          lr
-        .cfi_endproc
-END(memset)
+END(__memset_large_copy)
 
         .data
 error_string:
-        .string     "memset buffer overflow"
+        .string     "memset: prevented write past end of buffer"

libc/arch-arm/cortex-a9/bionic/strcat.S

@@ -53,7 +53,7 @@
  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
     .syntax unified
 

libc/arch-arm/cortex-a9/bionic/strcmp.S

@@ -27,7 +27,7 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 #ifdef __ARMEB__
 #define S2LOMEM lsl
@@ -123,8 +123,13 @@ ENTRY(strcmp)
         .macro  init
         /* Macro to save temporary registers and prepare magic values.  */
         subs    sp, sp, #16
+        .cfi_def_cfa_offset 16
         strd    r4, r5, [sp, #8]
+        .cfi_rel_offset r4, 0
+        .cfi_rel_offset r5, 4
         strd    r6, r7, [sp]
+        .cfi_rel_offset r6, 8
+        .cfi_rel_offset r7, 12
         mvn     r6, #0  /* all F */
         mov     r7, #0  /* all 0 */
         .endm   /* init */
@@ -140,19 +145,8 @@ ENTRY(strcmp)
 
         .macro  magic_find_zero_bytes w1
         /* Macro to find all-zero bytes in w1, result is in ip.  */
-#if (defined (__ARM_FEATURE_DSP))
         uadd8   ip, \w1, r6
         sel     ip, r7, r6
-#else /* not defined (__ARM_FEATURE_DSP) */
-        /* __ARM_FEATURE_DSP is not defined for some Cortex-M processors.
-        Coincidently, these processors only have Thumb-2 mode, where we can use the
-        the (large) magic constant available directly as an immediate in instructions.
-        Note that we cannot use the magic constant in ARM mode, where we need
-        to create the constant in a register.  */
-        sub     ip, \w1, #0x01010101
-        bic     ip, ip, \w1
-        and     ip, ip, #0x80808080
-#endif /* not defined (__ARM_FEATURE_DSP) */
         .endm /* magic_find_zero_bytes */
 
         .macro  setup_return w1 w2
@@ -171,12 +165,13 @@ ENTRY(strcmp)
         /* Are both strings double-word aligned?  */
         orr     ip, r0, r1
         tst     ip, #7
-        bne     do_align
+        bne     .L_do_align
 
         /* Fast path.  */
+        .save   {r4-r7}
         init
 
-doubleword_aligned:
+.L_doubleword_aligned:
 
         /* Get here when the strings to compare are double-word aligned.  */
         /* Compare two words in every iteration.  */
@@ -189,14 +184,14 @@ doubleword_aligned:
         ldrd    r2, r3, [r0], #8
         ldrd    r4, r5, [r1], #8
 
-        magic_compare_and_branch w1=r2, w2=r4, label=return_24
-        magic_compare_and_branch w1=r3, w2=r5, label=return_35
+        magic_compare_and_branch w1=r2, w2=r4, label=.L_return_24
+        magic_compare_and_branch w1=r3, w2=r5, label=.L_return_35
         b       2b
 
-do_align:
+.L_do_align:
         /* Is the first string word-aligned?  */
         ands    ip, r0, #3
-        beq     word_aligned_r0
+        beq     .L_word_aligned_r0
 
         /* Fast compare byte by byte until the first string is word-aligned.  */
         /* The offset of r0 from a word boundary is in ip. Thus, the number of bytes
@@ -204,58 +199,58 @@ do_align:
         bic     r0, r0, #3
         ldr     r2, [r0], #4
         lsls    ip, ip, #31
-        beq     byte2
-        bcs     byte3
+        beq     .L_byte2
+        bcs     .L_byte3
 
-byte1:
+.L_byte1:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE1_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbz   reg=r3, label=fast_return
+        bne     .L_fast_return
+        m_cbz   reg=r3, label=.L_fast_return
 
-byte2:
+.L_byte2:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE2_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbz   reg=r3, label=fast_return
+        bne     .L_fast_return
+        m_cbz   reg=r3, label=.L_fast_return
 
-byte3:
+.L_byte3:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE3_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbnz  reg=r3, label=word_aligned_r0
+        bne     .L_fast_return
+        m_cbnz  reg=r3, label=.L_word_aligned_r0
 
-fast_return:
+.L_fast_return:
         mov     r0, ip
         bx      lr
 
-word_aligned_r0:
+.L_word_aligned_r0:
         init
         /* The first string is word-aligned.  */
         /* Is the second string word-aligned?  */
         ands    ip, r1, #3
-        bne     strcmp_unaligned
+        bne     .L_strcmp_unaligned
 
-word_aligned:
+.L_word_aligned:
         /* The strings are word-aligned. */
         /* Is the first string double-word aligned?  */
         tst     r0, #4
-        beq     doubleword_aligned_r0
+        beq     .L_doubleword_aligned_r0
 
         /* If r0 is not double-word aligned yet, align it by loading
         and comparing the next word from each string.  */
         ldr     r2, [r0], #4
         ldr     r4, [r1], #4
-        magic_compare_and_branch w1=r2 w2=r4 label=return_24
+        magic_compare_and_branch w1=r2 w2=r4 label=.L_return_24
 
-doubleword_aligned_r0:
+.L_doubleword_aligned_r0:
         /* Get here when r0 is double-word aligned.  */
         /* Is r1 doubleword_aligned?  */
         tst     r1, #4
-        beq     doubleword_aligned
+        beq     .L_doubleword_aligned
 
         /* Get here when the strings to compare are word-aligned,
         r0 is double-word aligned, but r1 is not double-word aligned.  */
@@ -271,9 +266,9 @@ doubleword_aligned_r0:
 
         /* Load the next double-word from each string and compare.  */
         ldrd    r2, r3, [r0], #8
-        magic_compare_and_branch w1=r2 w2=r5 label=return_25
+        magic_compare_and_branch w1=r2 w2=r5 label=.L_return_25
         ldrd    r4, r5, [r1], #8
-        magic_compare_and_branch w1=r3 w2=r4 label=return_34
+        magic_compare_and_branch w1=r3 w2=r4 label=.L_return_34
         b       3b
 
         .macro miscmp_word offsetlo offsethi
@@ -297,33 +292,33 @@ doubleword_aligned_r0:
         and     r2, r3, r6, S2LOMEM #\offsetlo
         it      eq
         cmpeq   r2, r5
-        bne     return_25
+        bne     .L_return_25
         ldr     r5, [r1], #4
         cmp     ip, #0
         eor r3, r2, r3
         S2HIMEM r2, r5, #\offsethi
         it      eq
         cmpeq   r3, r2
-        bne     return_32
+        bne     .L_return_32
         b       7b
         .endm /* miscmp_word */
 
-return_32:
+.L_return_32:
         setup_return w1=r3, w2=r2
-        b       do_return
-return_34:
+        b       .L_do_return
+.L_return_34:
         setup_return w1=r3, w2=r4
-        b       do_return
-return_25:
+        b       .L_do_return
+.L_return_25:
         setup_return w1=r2, w2=r5
-        b       do_return
-return_35:
+        b       .L_do_return
+.L_return_35:
         setup_return w1=r3, w2=r5
-        b       do_return
-return_24:
+        b       .L_do_return
+.L_return_24:
         setup_return w1=r2, w2=r4
 
-do_return:
+.L_do_return:
 
 #ifdef __ARMEB__
         mov     r0, ip
@@ -335,11 +330,16 @@ do_return:
         ldrd    r6, r7, [sp]
         ldrd    r4, r5, [sp, #8]
         adds    sp, sp, #16
+        .cfi_def_cfa_offset 0
+        .cfi_restore r4
+        .cfi_restore r5
+        .cfi_restore r6
+        .cfi_restore r7
 
         /* There is a zero or a different byte between r1 and r2.  */
         /* r0 contains a mask of all-zero bytes in r1.  */
         /* Using r0 and not ip here because cbz requires low register.  */
-        m_cbz   reg=r0, label=compute_return_value
+        m_cbz   reg=r0, label=.L_compute_return_value
         clz     r0, r0
         /* r0 contains the number of bits on the left of the first all-zero byte in r1.  */
         rsb     r0, r0, #24
@@ -347,7 +347,7 @@ do_return:
         lsr     r1, r1, r0
         lsr     r2, r2, r0
 
-compute_return_value:
+.L_compute_return_value:
         movs    r0, #1
         cmp     r1, r2
         /* The return value is computed as follows.
@@ -367,7 +367,7 @@ compute_return_value:
      * bionic/libc/arch-arm/cortex-a15/bionic/strcmp.S for the unedited
      * version of the code.
      */
-strcmp_unaligned:
+.L_strcmp_unaligned:
 	wp1 .req r0
 	wp2 .req r1
 	b1  .req r2
@@ -520,6 +520,11 @@ strcmp_unaligned:
     ldrd    r6, r7, [sp]
     ldrd    r4, r5, [sp, #8]
     adds    sp, sp, #16
+    .cfi_def_cfa_offset 0
+    .cfi_restore r4
+    .cfi_restore r5
+    .cfi_restore r6
+    .cfi_restore r7
 
 	bx	lr
 

libc/arch-arm/cortex-a9/bionic/strcpy.S

@@ -53,7 +53,7 @@
  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
     .syntax unified
 

libc/arch-arm/cortex-a9/bionic/strlen.S

@@ -53,7 +53,7 @@
  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
     .syntax unified
 

libc/arch-arm/cortex-a9/cortex-a9.mk

@@ -1,10 +1,9 @@
-$(call libc-add-cpu-variant-src,MEMCPY,arch-arm/cortex-a9/bionic/memcpy.S)
-$(call libc-add-cpu-variant-src,MEMSET,arch-arm/cortex-a9/bionic/memset.S)
-$(call libc-add-cpu-variant-src,STRCAT,arch-arm/cortex-a9/bionic/strcat.S)
-$(call libc-add-cpu-variant-src,STRCMP,arch-arm/cortex-a9/bionic/strcmp.S)
-$(call libc-add-cpu-variant-src,STRCPY,arch-arm/cortex-a9/bionic/strcpy.S)
-$(call libc-add-cpu-variant-src,STRLEN,arch-arm/cortex-a9/bionic/strlen.S)
-$(call libc-add-cpu-variant-src,__STRCAT_CHK,arch-arm/cortex-a9/bionic/__strcat_chk.S)
-$(call libc-add-cpu-variant-src,__STRCPY_CHK,arch-arm/cortex-a9/bionic/__strcpy_chk.S)
-
-include bionic/libc/arch-arm/generic/generic.mk
+libc_bionic_src_files_arm += \
+    arch-arm/cortex-a9/bionic/memcpy.S \
+    arch-arm/cortex-a9/bionic/memset.S \
+    arch-arm/cortex-a9/bionic/strcat.S \
+    arch-arm/cortex-a9/bionic/strcmp.S \
+    arch-arm/cortex-a9/bionic/strcpy.S \
+    arch-arm/cortex-a9/bionic/strlen.S \
+    arch-arm/cortex-a9/bionic/__strcat_chk.S \
+    arch-arm/cortex-a9/bionic/__strcpy_chk.S \

libc/arch-arm/denver/bionic/__strcat_chk.S

@@ -0,0 +1,221 @@
+/*
+ * Copyright (C) 2013 The Android Open Source Project
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *  * Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  * Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+ * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
+ * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
+ * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
+ * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
+ * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
+
+    .syntax unified
+
+    .thumb
+    .thumb_func
+
+// Get the length of src string, then get the source of the dst string.
+// Check that the two lengths together don't exceed the threshold, then
+// do a memcpy of the data.
+ENTRY(__strcat_chk)
+    pld     [r0, #0]
+    push    {r0, lr}
+    .save   {r0, lr}
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
+    push    {r4, r5}
+    .save   {r4, r5}
+    .cfi_adjust_cfa_offset 8
+    .cfi_rel_offset r4, 0
+    .cfi_rel_offset r5, 4
+
+    mov     lr, r2
+
+    // Save the dst register to r5
+    mov     r5, r0
+
+    // Zero out r4
+    eor     r4, r4, r4
+
+    // r1 contains the address of the string to count.
+.L_strlen_start:
+    mov     r0, r1
+    ands    r3, r1, #7
+    beq     .L_mainloop
+
+    // Align to a double word (64 bits).
+    rsb     r3, r3, #8
+    lsls    ip, r3, #31
+    beq     .L_align_to_32
+
+    ldrb    r2, [r1], #1
+    cbz     r2, .L_update_count_and_finish
+
+.L_align_to_32:
+    bcc     .L_align_to_64
+    ands    ip, r3, #2
+    beq     .L_align_to_64
+
+    ldrb    r2, [r1], #1
+    cbz     r2, .L_update_count_and_finish
+    ldrb    r2, [r1], #1
+    cbz     r2, .L_update_count_and_finish
+
+.L_align_to_64:
+    tst     r3, #4
+    beq     .L_mainloop
+    ldr     r3, [r1], #4
+
+    sub     ip, r3, #0x01010101
+    bic     ip, ip, r3
+    ands    ip, ip, #0x80808080
+    bne     .L_zero_in_second_register
+
+    .p2align 2
+.L_mainloop:
+    ldrd    r2, r3, [r1], #8
+
+    pld     [r1, #64]
+
+    sub     ip, r2, #0x01010101
+    bic     ip, ip, r2
+    ands    ip, ip, #0x80808080
+    bne     .L_zero_in_first_register
+
+    sub     ip, r3, #0x01010101
+    bic     ip, ip, r3
+    ands    ip, ip, #0x80808080
+    bne     .L_zero_in_second_register
+    b       .L_mainloop
+
+.L_update_count_and_finish:
+    sub     r3, r1, r0
+    sub     r3, r3, #1
+    b       .L_finish
+
+.L_zero_in_first_register:
+    sub     r3, r1, r0
+    lsls    r2, ip, #17
+    bne     .L_sub8_and_finish
+    bcs     .L_sub7_and_finish
+    lsls    ip, ip, #1
+    bne     .L_sub6_and_finish
+
+    sub     r3, r3, #5
+    b       .L_finish
+
+.L_sub8_and_finish:
+    sub     r3, r3, #8
+    b       .L_finish
+
+.L_sub7_and_finish:
+    sub     r3, r3, #7
+    b       .L_finish
+
+.L_sub6_and_finish:
+    sub     r3, r3, #6
+    b       .L_finish
+
+.L_zero_in_second_register:
+    sub     r3, r1, r0
+    lsls    r2, ip, #17
+    bne     .L_sub4_and_finish
+    bcs     .L_sub3_and_finish
+    lsls    ip, ip, #1
+    bne     .L_sub2_and_finish
+
+    sub     r3, r3, #1
+    b       .L_finish
+
+.L_sub4_and_finish:
+    sub     r3, r3, #4
+    b       .L_finish
+
+.L_sub3_and_finish:
+    sub     r3, r3, #3
+    b       .L_finish
+
+.L_sub2_and_finish:
+    sub     r3, r3, #2
+
+.L_finish:
+    cmp     r4, #0
+    bne     .L_strlen_done
+
+    // Time to get the dst string length.
+    mov     r1, r5
+
+    // Save the original source address to r5.
+    mov     r5, r0
+
+    // Save the current length (adding 1 for the terminator).
+    add     r4, r3, #1
+    b       .L_strlen_start
+
+    // r0 holds the pointer to the dst string.
+    // r3 holds the dst string length.
+    // r4 holds the src string length + 1.
+.L_strlen_done:
+    add     r2, r3, r4
+    cmp     r2, lr
+    bhi     __strcat_chk_failed
+
+    // Set up the registers for the memcpy code.
+    mov     r1, r5
+    pld     [r1, #64]
+    mov     r2, r4
+    add     r0, r0, r3
+    pop     {r4, r5}
+END(__strcat_chk)
+
+#define MEMCPY_BASE         __strcat_chk_memcpy_base
+#define MEMCPY_BASE_ALIGNED __strcat_chk_memcpy_base_aligned
+
+#include "memcpy_base.S"
+
+ENTRY_PRIVATE(__strcat_chk_failed)
+    .save   {r0, lr}
+    .save   {r4, r5}
+
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
+    .cfi_adjust_cfa_offset 8
+    .cfi_rel_offset r4, 0
+    .cfi_rel_offset r5, 4
+
+    ldr     r0, error_message
+    ldr     r1, error_code
+1:
+    add     r0, pc
+    bl      __fortify_chk_fail
+error_code:
+    .word   BIONIC_EVENT_STRCAT_BUFFER_OVERFLOW
+error_message:
+    .word   error_string-(1b+4)
+END(__strcat_chk_failed)
+
+    .data
+error_string:
+    .string "strcat: prevented write past end of buffer"

libc/arch-arm/denver/bionic/__strcpy_chk.S

@@ -0,0 +1,182 @@
+/*
+ * Copyright (C) 2013 The Android Open Source Project
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *  * Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  * Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+ * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
+ * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
+ * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
+ * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
+ * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
+
+    .syntax unified
+
+    .thumb
+    .thumb_func
+
+// Get the length of the source string first, then do a memcpy of the data
+// instead of a strcpy.
+ENTRY(__strcpy_chk)
+    pld     [r0, #0]
+    push    {r0, lr}
+    .save   {r0, lr}
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
+
+    mov     lr, r2
+    mov     r0, r1
+
+    ands    r3, r1, #7
+    beq     .L_mainloop
+
+    // Align to a double word (64 bits).
+    rsb     r3, r3, #8
+    lsls    ip, r3, #31
+    beq     .L_align_to_32
+
+    ldrb    r2, [r0], #1
+    cbz     r2, .L_update_count_and_finish
+
+.L_align_to_32:
+    bcc     .L_align_to_64
+    ands    ip, r3, #2
+    beq     .L_align_to_64
+
+    ldrb    r2, [r0], #1
+    cbz     r2, .L_update_count_and_finish
+    ldrb    r2, [r0], #1
+    cbz     r2, .L_update_count_and_finish
+
+.L_align_to_64:
+    tst     r3, #4
+    beq     .L_mainloop
+    ldr     r3, [r0], #4
+
+    sub     ip, r3, #0x01010101
+    bic     ip, ip, r3
+    ands    ip, ip, #0x80808080
+    bne     .L_zero_in_second_register
+
+    .p2align 2
+.L_mainloop:
+    ldrd    r2, r3, [r0], #8
+
+    pld     [r0, #64]
+
+    sub     ip, r2, #0x01010101
+    bic     ip, ip, r2
+    ands    ip, ip, #0x80808080
+    bne     .L_zero_in_first_register
+
+    sub     ip, r3, #0x01010101
+    bic     ip, ip, r3
+    ands    ip, ip, #0x80808080
+    bne     .L_zero_in_second_register
+    b       .L_mainloop
+
+.L_update_count_and_finish:
+    sub     r3, r0, r1
+    sub     r3, r3, #1
+    b       .L_check_size
+
+.L_zero_in_first_register:
+    sub     r3, r0, r1
+    lsls    r2, ip, #17
+    bne     .L_sub8_and_finish
+    bcs     .L_sub7_and_finish
+    lsls    ip, ip, #1
+    bne     .L_sub6_and_finish
+
+    sub     r3, r3, #5
+    b       .L_check_size
+
+.L_sub8_and_finish:
+    sub     r3, r3, #8
+    b       .L_check_size
+
+.L_sub7_and_finish:
+    sub     r3, r3, #7
+    b       .L_check_size
+
+.L_sub6_and_finish:
+    sub     r3, r3, #6
+    b       .L_check_size
+
+.L_zero_in_second_register:
+    sub     r3, r0, r1
+    lsls    r2, ip, #17
+    bne     .L_sub4_and_finish
+    bcs     .L_sub3_and_finish
+    lsls    ip, ip, #1
+    bne     .L_sub2_and_finish
+
+    sub     r3, r3, #1
+    b       .L_check_size
+
+.L_sub4_and_finish:
+    sub     r3, r3, #4
+    b       .L_check_size
+
+.L_sub3_and_finish:
+    sub     r3, r3, #3
+    b       .L_check_size
+
+.L_sub2_and_finish:
+    sub     r3, r3, #2
+
+.L_check_size:
+    pld     [r1, #0]
+    pld     [r1, #64]
+    ldr     r0, [sp]
+    cmp     r3, lr
+    bhs     __strcpy_chk_failed
+
+    // Add 1 for copy length to get the string terminator.
+    add     r2, r3, #1
+END(__strcpy_chk)
+
+#define MEMCPY_BASE         __strcpy_chk_memcpy_base
+#define MEMCPY_BASE_ALIGNED __strcpy_chk_memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY_PRIVATE(__strcpy_chk_failed)
+    .save   {r0, lr}
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
+
+    ldr     r0, error_message
+    ldr     r1, error_code
+1:
+    add     r0, pc
+    bl      __fortify_chk_fail
+error_code:
+    .word   BIONIC_EVENT_STRCPY_BUFFER_OVERFLOW
+error_message:
+    .word   error_string-(1b+4)
+END(__strcpy_chk_failed)
+
+    .data
+error_string:
+    .string "strcpy: prevented write past end of buffer"

libc/arch-arm/denver/bionic/memcpy.S

@@ -0,0 +1,105 @@
+/*
+ * Copyright (C) 2008 The Android Open Source Project
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *  * Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  * Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+ * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
+ * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
+ * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
+ * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
+ * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+/*
+ * Copyright (c) 2013 ARM Ltd
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. The name of the company may not be used to endorse or promote
+ *    products derived from this software without specific prior written
+ *    permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY ARM LTD ``AS IS'' AND ANY EXPRESS OR IMPLIED
+ * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL ARM LTD BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
+ * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+ * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+// Prototype: void *memcpy (void *dst, const void *src, size_t count).
+
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
+
+        .text
+        .syntax unified
+        .fpu    neon
+
+ENTRY(__memcpy_chk)
+        cmp     r2, r3
+        bhi     __memcpy_chk_fail
+
+        // Fall through to memcpy...
+END(__memcpy_chk)
+
+ENTRY(memcpy)
+        pld     [r1, #64]
+        push    {r0, lr}
+        .save   {r0, lr}
+        .cfi_def_cfa_offset 8
+        .cfi_rel_offset r0, 0
+        .cfi_rel_offset lr, 4
+END(memcpy)
+
+#define MEMCPY_BASE         __memcpy_base
+#define MEMCPY_BASE_ALIGNED __memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY_PRIVATE(__memcpy_chk_fail)
+        // Preserve lr for backtrace.
+        push    {lr}
+        .save   {lr}
+        .cfi_def_cfa_offset 4
+        .cfi_rel_offset lr, 0
+
+        ldr     r0, error_message
+        ldr     r1, error_code
+1:
+        add     r0, pc
+        bl      __fortify_chk_fail
+error_code:
+        .word   BIONIC_EVENT_MEMCPY_BUFFER_OVERFLOW
+error_message:
+        .word   error_string-(1b+8)
+END(__memcpy_chk_fail)
+
+        .data
+error_string:
+        .string "memcpy: prevented write past end of buffer"

libc/arch-arm/denver/bionic/memcpy_base.S

@@ -0,0 +1,234 @@
+/*
+ * Copyright (C) 2008 The Android Open Source Project
+ * All rights reserved.
+ * Copyright (c) 2013-2014, NVIDIA Corporation.  All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *  * Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  * Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+ * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
+ * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
+ * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
+ * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
+ * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#define CACHE_LINE_SIZE         (64)
+#define PREFETCH_DISTANCE       (CACHE_LINE_SIZE*6)
+
+ENTRY_PRIVATE(MEMCPY_BASE)
+        .cfi_def_cfa_offset 8
+        .cfi_rel_offset r0, 0
+        .cfi_rel_offset lr, 4
+
+        cmp         r2, #0
+        beq         .L_memcpy_done
+        cmp         r0, r1
+        beq         .L_memcpy_done
+
+        /* preload next cache line */
+        pld         [r1, #CACHE_LINE_SIZE*1]
+
+        /* Deal with very small blocks (< 32bytes) asap */
+        cmp         r2, #32
+        blo         .L_memcpy_lt_32bytes
+        /* no need to align if len < 128 bytes */
+        cmp         r2, #128
+        blo         .L_memcpy_lt_128bytes
+
+        /* large copy, align dest to 64 byte boundry */
+        pld         [r1, #CACHE_LINE_SIZE*2]
+        rsb         r3, r0, #0
+        ands        r3, r3, #0x3F
+        pld         [r1, #CACHE_LINE_SIZE*3]
+        beq         .L_memcpy_dispatch
+        sub         r2, r2, r3
+        /* copy 1 byte */
+        movs        ip, r3, lsl #31
+        itt         mi
+        ldrbmi      ip, [r1], #1
+        strbmi      ip, [r0], #1
+        /* copy 2 bytes */
+        itt         cs
+        ldrhcs      ip, [r1], #2
+        strhcs      ip, [r0], #2
+        /* copy 4 bytes */
+        movs        ip, r3, lsl #29
+        itt         mi
+        ldrmi       ip, [r1], #4
+        strmi       ip, [r0], #4
+        /* copy 8 bytes */
+        bcc         1f
+        vld1.8      {d0}, [r1]!
+        vst1.8      {d0}, [r0, :64]!
+1:      /* copy 16 bytes */
+        movs        ip, r3, lsl #27
+        bpl         1f
+        vld1.8      {q0}, [r1]!
+        vst1.8      {q0}, [r0, :128]!
+1:      /* copy 32 bytes */
+        bcc         .L_memcpy_dispatch
+        vld1.8      {q0, q1}, [r1]!
+        vst1.8      {q0, q1}, [r0, :256]!
+
+.L_memcpy_dispatch:
+        // pre-decrement by 128 to detect nearly-done condition easily, but
+        // also need to check if we have less than 128 bytes left at this
+        // point due to alignment code above
+        subs        r2, r2, #128
+        blo         .L_memcpy_lt_128presub
+
+        // Denver does better if both source and dest are aligned so
+        // we'll special-case that even though the code is virually identical
+        tst         r1, #0xF
+        bne         .L_memcpy_neon_unalign_src_pld
+
+        // DRAM memcpy should be throttled slightly to get full bandwidth
+        //
+        cmp         r2, #32768
+        bhi         .L_memcpy_neon_unalign_src_pld
+        .align      4
+1:
+        /* copy 128 bytes in each loop */
+        subs        r2, r2, #128
+
+        /* preload a cache line */
+        pld         [r1, #PREFETCH_DISTANCE]
+        /* copy a cache line */
+        vld1.8      {q0, q1}, [r1, :128]!
+        vst1.8      {q0, q1}, [r0, :256]!
+        vld1.8      {q0, q1}, [r1, :128]!
+        vst1.8      {q0, q1}, [r0, :256]!
+        /* preload a cache line */
+        pld         [r1, #PREFETCH_DISTANCE]
+        /* copy a cache line */
+        vld1.8      {q0, q1}, [r1, :128]!
+        vst1.8      {q0, q1}, [r0, :256]!
+        vld1.8      {q0, q1}, [r1, :128]!
+        vst1.8      {q0, q1}, [r0, :256]!
+
+        bhs         1b
+        adds        r2, r2, #128
+        bne         .L_memcpy_lt_128bytes_align
+        pop         {r0, pc}
+
+        .align      4
+.L_memcpy_neon_unalign_src_pld:
+1:
+        /* copy 128 bytes in each loop */
+        subs        r2, r2, #128
+
+        /* preload a cache line */
+        pld         [r1, #PREFETCH_DISTANCE]
+        /* copy a cache line */
+        vld1.8      {q0, q1}, [r1]!
+        vst1.8      {q0, q1}, [r0, :256]!
+        vld1.8      {q0, q1}, [r1]!
+        vst1.8      {q0, q1}, [r0, :256]!
+        /* preload a cache line */
+        pld         [r1, #PREFETCH_DISTANCE]
+        /* copy a cache line */
+        vld1.8      {q0, q1}, [r1]!
+        vst1.8      {q0, q1}, [r0, :256]!
+        vld1.8      {q0, q1}, [r1]!
+        vst1.8      {q0, q1}, [r0, :256]!
+
+        bhs         1b
+        adds        r2, r2, #128
+        bne         .L_memcpy_lt_128bytes_align
+        pop         {r0, pc}
+
+.L_memcpy_lt_128presub:
+        add         r2, r2, #128
+.L_memcpy_lt_128bytes_align:
+        /* copy 64 bytes */
+        movs        ip, r2, lsl #26
+        bcc         1f
+        vld1.8      {q0, q1}, [r1]!
+        vst1.8      {q0, q1}, [r0, :256]!
+        vld1.8      {q0, q1}, [r1]!
+        vst1.8      {q0, q1}, [r0, :256]!
+1:      /* copy 32 bytes */
+        bpl         1f
+        vld1.8      {q0, q1}, [r1]!
+        vst1.8      {q0, q1}, [r0, :256]!
+1:      /* copy 16 bytes */
+        movs        ip, r2, lsl #28
+        bcc         1f
+        vld1.8      {q0}, [r1]!
+        vst1.8      {q0}, [r0, :128]!
+1:      /* copy 8 bytes */
+        bpl         1f
+        vld1.8      {d0}, [r1]!
+        vst1.8      {d0}, [r0, :64]!
+1:      /* copy 4 bytes */
+        tst         r2, #4
+        itt         ne
+        ldrne       ip, [r1], #4
+        strne       ip, [r0], #4
+        /* copy 2 bytes */
+        movs        ip, r2, lsl #31
+        itt         cs
+        ldrhcs      ip, [r1], #2
+        strhcs      ip, [r0], #2
+        /* copy 1 byte */
+        itt         mi
+        ldrbmi      ip, [r1]
+        strbmi      ip, [r0]
+
+        pop         {r0, pc}
+
+.L_memcpy_lt_128bytes:
+        /* copy 64 bytes */
+        movs        ip, r2, lsl #26
+        bcc         1f
+        vld1.8      {q0, q1}, [r1]!
+        vst1.8      {q0, q1}, [r0]!
+        vld1.8      {q0, q1}, [r1]!
+        vst1.8      {q0, q1}, [r0]!
+1:      /* copy 32 bytes */
+        bpl	    .L_memcpy_lt_32bytes
+        vld1.8      {q0, q1}, [r1]!
+        vst1.8      {q0, q1}, [r0]!
+.L_memcpy_lt_32bytes:
+        /* copy 16 bytes */
+        movs        ip, r2, lsl #28
+        bcc         1f
+        vld1.8      {q0}, [r1]!
+        vst1.8      {q0}, [r0]!
+1:      /* copy 8 bytes */
+        bpl         1f
+        vld1.8      {d0}, [r1]!
+        vst1.8      {d0}, [r0]!
+1:      /* copy 4 bytes */
+        tst         r2, #4
+        itt         ne
+        ldrne       ip, [r1], #4
+        strne       ip, [r0], #4
+        /* copy 2 bytes */
+        movs        ip, r2, lsl #31
+        itt         cs
+        ldrhcs      ip, [r1], #2
+        strhcs      ip, [r0], #2
+        /* copy 1 byte */
+        itt         mi
+        ldrbmi      ip, [r1]
+        strbmi      ip, [r0]
+
+.L_memcpy_done:
+        pop         {r0, pc}
+END(MEMCPY_BASE)

libc/arch-arm/denver/bionic/memset.S

@@ -0,0 +1,207 @@
+/*
+ * Copyright (C) 2013 The Android Open Source Project
+ * Copyright (c) 2014, NVIDIA CORPORATION.  All rights reserved.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *  * Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *  * Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+ * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
+ * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
+ * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
+ * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
+ * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include <machine/cpu-features.h>
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
+
+        /*
+         * Optimized memset() for ARM.
+         *
+         * memset() returns its first argument.
+         */
+
+        .fpu        neon
+        .syntax     unified
+
+ENTRY(__memset_chk)
+        cmp         r2, r3
+        bls         .L_done
+
+        // Preserve lr for backtrace.
+        push        {lr}
+        .cfi_def_cfa_offset 4
+        .cfi_rel_offset lr, 0
+
+
+        ldr         r0, error_message
+        ldr         r1, error_code
+1:
+        add         r0, pc
+        bl          __fortify_chk_fail
+error_code:
+        .word       BIONIC_EVENT_MEMSET_BUFFER_OVERFLOW
+error_message:
+        .word       error_string-(1b+8)
+END(__memset_chk)
+
+ENTRY(bzero)
+        mov         r2, r1
+        mov         r1, #0
+.L_done:
+        // Fall through to memset...
+END(bzero)
+
+ENTRY(memset)
+        pldw        [r0]
+        mov         r3, r0
+
+        // Duplicate the low byte of r1
+        mov         r1, r1, lsl #24
+        orr         r1, r1, r1, lsr #8
+        orr         r1, r1, r1, lsr #16
+
+        cmp         r2, #16
+        blo         .L_less_than_16
+
+        // This section handles regions 16 bytes or larger
+        //
+        // Use aligned vst1.8 and vstm when possible.  Register values will be:
+        //   ip is scratch
+        //   q0, q1, and r1 contain the memset value
+        //   r2 is the number of bytes to set
+        //   r3 is the advancing destination pointer
+        vdup.32     q0, r1
+
+        ands        ip, r3, 0xF
+        beq         .L_memset_aligned
+
+        // Align dest pointer to 16-byte boundary.
+        pldw        [r0, #64]
+        rsb         ip, ip, #16
+
+        // Pre-adjust the byte count to reflect post-aligment value.  Expecting
+        // 8-byte alignment to be rather common so we special case that one.
+        sub         r2, r2, ip
+
+        /* set 1 byte */
+        tst         ip, #1
+        it          ne
+        strbne      r1, [r3], #1
+        /* set 2 bytes */
+        tst         ip, #2
+        it          ne
+        strhne      r1, [r3], #2
+        /* set 4 bytes */
+        movs        ip, ip, lsl #29
+        it          mi
+        strmi       r1, [r3], #4
+        /* set 8 bytes */
+        itt         cs
+        strcs       r1, [r3], #4
+        strcs       r1, [r3], #4
+
+.L_memset_aligned:
+        // Destination is now 16-byte aligned.  Determine how to handle
+        // remaining bytes.
+        vmov        q1, q0
+        cmp         r2, #128
+        blo         .L_less_than_128
+
+        // We need to set a larger block of memory.  Use four Q regs to
+        // set a full cache line in one instruction.  Pre-decrement
+        // r2 to simplify end-of-loop detection
+        vmov        q2, q0
+        vmov        q3, q0
+        pldw        [r0, #128]
+        sub         r2, r2, #128
+        .align 4
+.L_memset_loop_128:
+        pldw        [r3, #192]
+        vstm        r3!, {q0, q1, q2, q3}
+        vstm        r3!, {q0, q1, q2, q3}
+        subs        r2, r2, #128
+        bhs         .L_memset_loop_128
+
+        // Un-bias r2 so it contains the number of bytes left.  Early
+        // exit if we are done.
+        adds        r2, r2, #128
+        beq         2f
+
+        .align 4
+.L_less_than_128:
+        // set 64 bytes
+        movs        ip, r2, lsl #26
+        bcc         1f
+        vst1.8      {q0, q1}, [r3, :128]!
+        vst1.8      {q0, q1}, [r3, :128]!
+        beq         2f
+1:
+        // set 32 bytes
+        bpl         1f
+        vst1.8      {q0, q1}, [r3, :128]!
+1:
+        // set 16 bytes
+        movs        ip, r2, lsl #28
+        bcc         1f
+        vst1.8      {q0}, [r3, :128]!
+        beq         2f
+1:
+        // set 8 bytes
+        bpl         1f
+        vst1.8      {d0}, [r3, :64]!
+1:
+        // set 4 bytes
+        tst         r2, #4
+        it          ne
+        strne       r1, [r3], #4
+1:
+        // set 2 bytes
+        movs        ip, r2, lsl #31
+        it          cs
+        strhcs      r1, [r3], #2
+        // set 1 byte
+        it          mi
+        strbmi      r1, [r3]
+2:
+        bx          lr
+
+.L_less_than_16:
+        // Store up to 15 bytes without worrying about byte alignment
+        movs        ip, r2, lsl #29
+        bcc         1f
+        str         r1, [r3], #4
+        str         r1, [r3], #4
+        beq         2f
+1:
+        it          mi
+        strmi       r1, [r3], #4
+        movs        ip, r2, lsl #31
+        it          mi
+        strbmi      r1, [r3], #1
+        itt         cs
+        strbcs      r1, [r3], #1
+        strbcs      r1, [r3]
+2:
+        bx          lr
+END(memset)
+
+        .data
+error_string:
+        .string     "memset: prevented write past end of buffer"

libc/arch-arm/denver/denver.mk

@@ -0,0 +1,12 @@
+libc_bionic_src_files_arm += \
+    arch-arm/denver/bionic/memcpy.S \
+    arch-arm/denver/bionic/memset.S \
+    arch-arm/denver/bionic/__strcat_chk.S \
+    arch-arm/denver/bionic/__strcpy_chk.S
+
+# Use cortex-a15 versions of strcat/strcpy/strlen.
+libc_bionic_src_files_arm += \
+    arch-arm/cortex-a15/bionic/strcat.S \
+    arch-arm/cortex-a15/bionic/strcpy.S \
+    arch-arm/cortex-a15/bionic/strlen.S \
+    arch-arm/cortex-a15/bionic/strcmp.S

libc/arch-arm/generic/bionic/memcpy.S

@@ -27,8 +27,8 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
         /*
          * Optimized memcpy() for ARM.
@@ -57,9 +57,9 @@ ENTRY(memcpy)
 
         // preload the destination because we'll align it to a cache line
         // with small writes. Also start the source "pump".
-        PLD         (r0, #0)
-        PLD         (r1, #0)
-        PLD         (r1, #32)
+        pld         [r0, #0]
+        pld         [r1, #0]
+        pld         [r1, #32]
 
         /* it simplifies things to take care of len<4 early */
         cmp         r2, #4
@@ -147,7 +147,7 @@ cached_aligned32:
         add         r12, r12, #64
 
 1:      ldmia       r1!, { r4-r11 }
-        PLD         (r12, #64)
+        pld         [r12, #64]
         subs        r2, r2, #32
 
         // NOTE: if r12 is more than 64 ahead of r1, the following ldrhi
@@ -268,7 +268,7 @@ loop16:
         ldr         r12, [r1], #4
 1:      mov         r4, r12
         ldmia       r1!, {   r5,r6,r7,  r8,r9,r10,r11}
-        PLD         (r1, #64)
+        pld         [r1, #64]
         subs        r2, r2, #32
         ldrhs       r12, [r1], #4
         orr         r3, r3, r4,     lsl #16
@@ -295,7 +295,7 @@ loop8:
         ldr         r12, [r1], #4
 1:      mov         r4, r12
         ldmia       r1!, {   r5,r6,r7,  r8,r9,r10,r11}
-        PLD         (r1, #64)
+        pld         [r1, #64]
         subs        r2, r2, #32
         ldrhs       r12, [r1], #4
         orr         r3, r3, r4,     lsl #24
@@ -322,7 +322,7 @@ loop24:
         ldr         r12, [r1], #4
 1:      mov         r4, r12
         ldmia       r1!, {   r5,r6,r7,  r8,r9,r10,r11}
-        PLD         (r1, #64)
+        pld         [r1, #64]
         subs        r2, r2, #32
         ldrhs       r12, [r1], #4
         orr         r3, r3, r4,     lsl #8
@@ -401,4 +401,4 @@ END(memcpy)
 
         .data
 error_string:
-        .string     "memcpy buffer overflow"
+        .string     "memcpy: prevented write past end of buffer"

libc/arch-arm/generic/bionic/memset.S

@@ -26,8 +26,8 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
         /*
          * Optimized memset() for ARM.
@@ -130,4 +130,4 @@ END(memset)
 
         .data
 error_string:
-        .string     "memset buffer overflow"
+        .string     "memset: prevented write past end of buffer"

libc/arch-arm/generic/bionic/strcmp.S

@@ -28,7 +28,7 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 	.text
 
@@ -52,8 +52,8 @@
 #define magic2(REG) REG, lsl #7
 
 ENTRY(strcmp)
-	PLD(r0, #0)
-	PLD(r1, #0)
+	pld	[r0, #0]
+	pld	[r1, #0]
 	eor	r2, r0, r1
 	tst	r2, #3
 
@@ -88,8 +88,8 @@ ENTRY(strcmp)
 	orr	r4, r4, r4, lsl #16
 	.p2align	2
 4:
-	PLD(r0, #8)
-	PLD(r1, #8)
+	pld	[r0, #8]
+	pld	[r1, #8]
 	sub	r2, ip, magic1(r4)
 	cmp	ip, r3
 	itttt	eq

libc/arch-arm/generic/bionic/strcpy.S

@@ -30,10 +30,10 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 ENTRY(strcpy)
-	PLD(r1, #0)
+	pld	[r1, #0]
 	eor	r2, r0, r1
 	mov	ip, r0
 	tst	r2, #3
@@ -62,7 +62,7 @@ ENTRY(strcpy)
 	  load stalls.  */
 	.p2align 2
 2:
-	PLD(r1, #8)
+	pld	[r1, #8]
 	ldr	r4, [r1], #4
 	sub	r2, r3, r5
 	bics	r2, r2, r3

libc/arch-arm/generic/bionic/strlen.c

@@ -63,9 +63,7 @@ size_t strlen(const char *s)
         "ldr     %[v], [%[s]], #4           \n"
         "sub     %[l], %[l], %[s]           \n"
         "0:                                 \n"
-#if __ARM_HAVE_PLD
         "pld     [%[s], #64]                \n"
-#endif
         "sub     %[t], %[v], %[mask], lsr #7\n"
         "and     %[t], %[t], %[mask]        \n"
         "bics    %[t], %[t], %[v]           \n"

libc/arch-arm/generic/generic.mk

@@ -1,8 +1,9 @@
-$(call libc-add-cpu-variant-src,MEMCPY,arch-arm/generic/bionic/memcpy.S)
-$(call libc-add-cpu-variant-src,MEMSET,arch-arm/generic/bionic/memset.S)
-$(call libc-add-cpu-variant-src,STRCAT,string/strcat.c)
-$(call libc-add-cpu-variant-src,STRCMP,arch-arm/generic/bionic/strcmp.S)
-$(call libc-add-cpu-variant-src,STRCPY,arch-arm/generic/bionic/strcpy.S)
-$(call libc-add-cpu-variant-src,STRLEN,arch-arm/generic/bionic/strlen.c)
-$(call libc-add-cpu-variant-src,__STRCAT_CHK,bionic/__strcat_chk.cpp)
-$(call libc-add-cpu-variant-src,__STRCPY_CHK,bionic/__strcpy_chk.cpp)
+libc_bionic_src_files_arm += \
+    arch-arm/generic/bionic/memcpy.S \
+    arch-arm/generic/bionic/memset.S \
+    arch-arm/generic/bionic/strcmp.S \
+    arch-arm/generic/bionic/strcpy.S \
+    arch-arm/generic/bionic/strlen.c \
+    bionic/__strcat_chk.cpp \
+    bionic/__strcpy_chk.cpp \
+    upstream-openbsd/lib/libc/string/strcat.c \

libc/arch-arm/include/machine/_types.h

@@ -1,110 +0,0 @@
-/*	$OpenBSD: _types.h,v 1.3 2006/02/14 18:12:58 miod Exp $	*/
-
-/*-
- * Copyright (c) 1990, 1993
- *	The Regents of the University of California.  All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- * 1. Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in the
- *    documentation and/or other materials provided with the distribution.
- * 3. Neither the name of the University nor the names of its contributors
- *    may be used to endorse or promote products derived from this software
- *    without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
- * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
- * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
- * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
- * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
- * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
- * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- *
- *	@(#)types.h	8.3 (Berkeley) 1/5/94
- *	@(#)ansi.h	8.2 (Berkeley) 1/4/94
- */
-
-#ifndef _ARM__TYPES_H_
-#define _ARM__TYPES_H_
-
-/* 7.18.1.1 Exact-width integer types */
-typedef	__signed char		__int8_t;
-typedef	unsigned char		__uint8_t;
-typedef	short			__int16_t;
-typedef	unsigned short		__uint16_t;
-typedef	int			__int32_t;
-typedef	unsigned int		__uint32_t;
-/* LONGLONG */
-typedef	long long		__int64_t;
-/* LONGLONG */
-typedef	unsigned long long	__uint64_t;
-
-/* 7.18.1.2 Minimum-width integer types */
-typedef	__int8_t		__int_least8_t;
-typedef	__uint8_t		__uint_least8_t;
-typedef	__int16_t		__int_least16_t;
-typedef	__uint16_t		__uint_least16_t;
-typedef	__int32_t		__int_least32_t;
-typedef	__uint32_t		__uint_least32_t;
-typedef	__int64_t		__int_least64_t;
-typedef	__uint64_t		__uint_least64_t;
-
-/* 7.18.1.3 Fastest minimum-width integer types */
-typedef	__int32_t		__int_fast8_t;
-typedef	__uint32_t		__uint_fast8_t;
-typedef	__int32_t		__int_fast16_t;
-typedef	__uint32_t		__uint_fast16_t;
-typedef	__int32_t		__int_fast32_t;
-typedef	__uint32_t		__uint_fast32_t;
-typedef	__int64_t		__int_fast64_t;
-typedef	__uint64_t		__uint_fast64_t;
-
-/* 7.18.1.4 Integer types capable of holding object pointers */
-typedef	int 			__intptr_t;
-typedef	unsigned int 		__uintptr_t;
-
-/* 7.18.1.5 Greatest-width integer types */
-typedef	__int64_t		__intmax_t;
-typedef	__uint64_t		__uintmax_t;
-
-/* Register size */
-typedef __int32_t		__register_t;
-
-/* VM system types */
-typedef unsigned long		__vaddr_t;
-typedef unsigned long		__paddr_t;
-typedef unsigned long		__vsize_t;
-typedef unsigned long		__psize_t;
-
-/* Standard system types */
-typedef int			__clock_t;
-typedef int			__clockid_t;
-typedef double			__double_t;
-typedef float			__float_t;
-typedef long			__ptrdiff_t;
-typedef	int			__time_t;
-typedef int			__timer_t;
-#if defined(__GNUC__) && __GNUC__ >= 3
-typedef	__builtin_va_list	__va_list;
-#else
-typedef	char *			__va_list;
-#endif
-
-/* Wide character support types */
-#ifndef __cplusplus
-typedef	int			__wchar_t;
-#endif
-typedef int			__wint_t;
-typedef	int			__rune_t;
-typedef	void *			__wctrans_t;
-typedef	void *			__wctype_t;
-
-#endif	/* _ARM__TYPES_H_ */

libc/arch-arm/include/machine/asm.h

@@ -38,107 +38,22 @@
 #ifndef _ARM32_ASM_H_
 #define _ARM32_ASM_H_
 
-#ifdef __ELF__
-# define _C_LABEL(x)	x
-#else
-# ifdef __STDC__
-#  define _C_LABEL(x)	_ ## x
-# else
-#  define _C_LABEL(x)	_/**/x
-# endif
-#endif
-#define	_ASM_LABEL(x)	x
-
-#ifdef __STDC__
-# define __CONCAT(x,y)	x ## y
-# define __STRING(x)	#x
-#else
-# define __CONCAT(x,y)	x/**/y
-# define __STRING(x)	"x"
-#endif
-
 #ifndef _ALIGN_TEXT
 # define _ALIGN_TEXT .align 0
 #endif
 
-/*
- * gas/arm uses @ as a single comment character and thus cannot be used here
- * Instead it recognised the # instead of an @ symbols in .type directives
- * We define a couple of macros so that assembly code will not be dependant
- * on one or the other.
- */
-#define _ASM_TYPE_FUNCTION	#function
-#define _ASM_TYPE_OBJECT	#object
-#define _ENTRY(x) \
-	.text; _ALIGN_TEXT; .globl x; .type x,_ASM_TYPE_FUNCTION; x: .fnstart
+#undef __bionic_asm_custom_entry
+#undef __bionic_asm_custom_end
+#define __bionic_asm_custom_entry(f) .fnstart
+#define __bionic_asm_custom_end(f) .fnend
 
-#define _ASM_SIZE(x)	.size x, .-x;
-
-#define _END(x) \
-	.fnend; \
-	_ASM_SIZE(x)
-
-#ifdef GPROF
-# ifdef __ELF__
-#  define _PROF_PROLOGUE	\
-	mov ip, lr; bl __mcount
-# else
-#  define _PROF_PROLOGUE	\
-	mov ip,lr; bl mcount
-# endif
-#else
-# define _PROF_PROLOGUE
-#endif
-
-#define	ENTRY(y)	_ENTRY(_C_LABEL(y)); _PROF_PROLOGUE
-#define	ENTRY_NP(y)	_ENTRY(_C_LABEL(y))
-#define	END(y)		_END(_C_LABEL(y))
-#define	ASENTRY(y)	_ENTRY(_ASM_LABEL(y)); _PROF_PROLOGUE
-#define	ASENTRY_NP(y)	_ENTRY(_ASM_LABEL(y))
-#define	ASEND(y)	_END(_ASM_LABEL(y))
-
-#ifdef __ELF__
-#define ENTRY_PRIVATE(y)  ENTRY(y); .hidden _C_LABEL(y)
-#else
-#define ENTRY_PRIVATE(y)  ENTRY(y)
-#endif
-
-#define	ASMSTR		.asciz
+#undef __bionic_asm_function_type
+#define __bionic_asm_function_type #function
 
 #if defined(__ELF__) && defined(PIC)
-#ifdef __STDC__
-#define	PIC_SYM(x,y)	x ## ( ## y ## )
+#define PIC_SYM(x,y) x ## ( ## y ## )
 #else
-#define	PIC_SYM(x,y)	x/**/(/**/y/**/)
+#define PIC_SYM(x,y) x
 #endif
-#else
-#define	PIC_SYM(x,y)	x
-#endif
-
-#ifdef __ELF__
-#define RCSID(x)	.section ".ident"; .asciz x
-#else
-#define RCSID(x)	.text; .asciz x
-#endif
-
-#ifdef __ELF__
-#define	WEAK_ALIAS(alias,sym)						\
-	.weak alias;							\
-	alias = sym
-#endif
-
-#ifdef __STDC__
-#define	WARN_REFERENCES(sym,msg)					\
-	.stabs msg ## ,30,0,0,0 ;					\
-	.stabs __STRING(_C_LABEL(sym)) ## ,1,0,0,0
-#elif defined(__ELF__)
-#define	WARN_REFERENCES(sym,msg)					\
-	.stabs msg,30,0,0,0 ;						\
-	.stabs __STRING(sym),1,0,0,0
-#else
-#define	WARN_REFERENCES(sym,msg)					\
-	.stabs msg,30,0,0,0 ;						\
-	.stabs __STRING(_/**/sym),1,0,0,0
-#endif /* __STDC__ */
 
 #endif /* !_ARM_ASM_H_ */

libc/arch-arm/include/machine/cdefs.h

@@ -1,19 +0,0 @@
-/*	$OpenBSD: cdefs.h,v 1.2 2005/11/24 20:46:44 deraadt Exp $	*/
-
-#ifndef	_MACHINE_CDEFS_H_
-#define	_MACHINE_CDEFS_H_
-
-#if defined(lint)
-#define __indr_reference(sym,alias)	__lint_equal__(sym,alias)
-#define __warn_references(sym,msg)
-#define __weak_alias(alias,sym)		__lint_equal__(sym,alias)
-#elif defined(__GNUC__) && defined(__STDC__)
-#define __weak_alias(alias,sym)					\
-	__asm__(".weak " __STRING(alias) " ; " __STRING(alias)	\
-	    " = " __STRING(sym));
-#define	__warn_references(sym,msg)				\
-	__asm__(".section .gnu.warning." __STRING(sym)		\
-	    " ; .ascii \"" msg "\" ; .text");
-#endif
-
-#endif /* !_MACHINE_CDEFS_H_ */

libc/arch-arm/include/machine/cpu-features.h

@@ -34,133 +34,29 @@
  *
  * This is done to abstract us from the various ARM Architecture
  * quirks and alphabet soup.
- *
- * IMPORTANT: We have no intention to support anything below an ARMv4T !
  */
 
 /* __ARM_ARCH__ is a number corresponding to the ARM revision
- * we're going to support
- *
- * it looks like our toolchain doesn't define __ARM_ARCH__
+ * we're going to support. Our toolchain doesn't define __ARM_ARCH__
  * so try to guess it.
- *
- *
- *
  */
 #ifndef __ARM_ARCH__
-
 #  if defined __ARM_ARCH_7__   || defined __ARM_ARCH_7A__ || \
-      defined __ARM_ARCH_7R__  || defined __ARM_ARCH_7M__
-
+        defined __ARM_ARCH_7R__  || defined __ARM_ARCH_7M__
 #    define __ARM_ARCH__ 7
-
 #  elif defined __ARM_ARCH_6__   || defined __ARM_ARCH_6J__ || \
-      defined __ARM_ARCH_6K__  || defined __ARM_ARCH_6Z__ || \
-      defined __ARM_ARCH_6KZ__ || defined __ARM_ARCH_6T2__
-#
+        defined __ARM_ARCH_6K__  || defined __ARM_ARCH_6Z__ || \
+        defined __ARM_ARCH_6KZ__ || defined __ARM_ARCH_6T2__
 #    define __ARM_ARCH__ 6
-#
-#  elif defined __ARM_ARCH_5__ || defined __ARM_ARCH_5T__ || \
-        defined __ARM_ARCH_5TE__ || defined __ARM_ARCH_5TEJ__
-#
-#    define __ARM_ARCH__ 5
-#
-#  elif defined __ARM_ARCH_4T__
-#
-#    define __ARM_ARCH__ 4
-#
-#  elif defined __ARM_ARCH_4__
-#    error ARMv4 is not supported, please use ARMv4T at a minimum
 #  else
 #    error Unknown or unsupported ARM architecture
 #  endif
 #endif
 
-/* experimental feature used to check that our ARMv4 workarounds
- * work correctly without a real ARMv4 machine */
-#ifdef BIONIC_EXPERIMENTAL_FORCE_ARMV4
-#  undef  __ARM_ARCH__
-#  define __ARM_ARCH__  4
-#endif
-
-/* define __ARM_HAVE_5TE if we have the ARMv5TE instructions */
-#if __ARM_ARCH__ > 5
-#  define  __ARM_HAVE_5TE  1
-#elif __ARM_ARCH__ == 5
-#  if defined __ARM_ARCH_5TE__ || defined __ARM_ARCH_5TEJ__
-#    define __ARM_HAVE_5TE  1
-#  endif
-#endif
-
-/* instructions introduced in ARMv5 */
-#if __ARM_ARCH__ >= 5
-#  define  __ARM_HAVE_BLX  1
-#  define  __ARM_HAVE_CLZ  1
-#  define  __ARM_HAVE_LDC2 1
-#  define  __ARM_HAVE_MCR2 1
-#  define  __ARM_HAVE_MRC2 1
-#  define  __ARM_HAVE_STC2 1
-#endif
-
-/* ARMv5TE introduces a few instructions */
-#if __ARM_HAVE_5TE
-#  define  __ARM_HAVE_PLD   1
-#  define  __ARM_HAVE_MCRR  1
-#  define  __ARM_HAVE_MRRC  1
-#endif
-
 /* define __ARM_HAVE_HALFWORD_MULTIPLY when half-word multiply instructions
  * this means variants of: smul, smulw, smla, smlaw, smlal
  */
-#if __ARM_HAVE_5TE
-#  define  __ARM_HAVE_HALFWORD_MULTIPLY  1
-#endif
-
-/* define __ARM_HAVE_PAIR_LOAD_STORE when 64-bit memory loads and stored
- * into/from a pair of 32-bit registers is supported throuhg 'ldrd' and 'strd'
- */
-#if __ARM_HAVE_5TE
-#  define  __ARM_HAVE_PAIR_LOAD_STORE 1
-#endif
-
-/* define __ARM_HAVE_SATURATED_ARITHMETIC is you have the saturated integer
- * arithmetic instructions: qdd, qdadd, qsub, qdsub
- */
-#if __ARM_HAVE_5TE
-#  define  __ARM_HAVE_SATURATED_ARITHMETIC 1
-#endif
-
-/* define __ARM_HAVE_PC_INTERWORK when a direct assignment to the
- * pc register will switch into thumb/ARM mode depending on bit 0
- * of the new instruction address. Before ARMv5, this was not the
- * case, and you have to write:
- *
- *     mov  r0, [<some address>]
- *     bx   r0
- *
- * instead of:
- *
- *     ldr  pc, [<some address>]
- *
- * note that this affects any instruction that explicitly changes the
- * value of the pc register, including ldm { ...,pc } or 'add pc, #offset'
- */
-#if __ARM_ARCH__ >= 5
-#  define __ARM_HAVE_PC_INTERWORK
-#endif
-
-/* define __ARM_HAVE_LDREX_STREX for ARMv6 and ARMv7 architecture to be
- * used in replacement of deprecated swp instruction
- */
-#if __ARM_ARCH__ >= 6
-#  define __ARM_HAVE_LDREX_STREX
-#endif
-
-/* define __ARM_HAVE_DMB for ARMv7 architecture
- */
-#if __ARM_ARCH__ >= 7
-#  define __ARM_HAVE_DMB
-#endif
+#define  __ARM_HAVE_HALFWORD_MULTIPLY  1
 
 /* define __ARM_HAVE_LDREXD for ARMv7 architecture
  * (also present in ARMv6K, but not implemented in ARMv7-M, neither of which
@@ -184,18 +80,4 @@
 #  define __ARM_HAVE_NEON
 #endif
 
-/* Assembly-only macros */
-#ifdef __ASSEMBLY__
-
-/* define a handy PLD(address) macro since the cache preload
- * is an optional opcode
- */
-#if __ARM_HAVE_PLD
-#  define  PLD(reg,offset)    pld    [reg, offset]
-#else
-#  define  PLD(reg,offset)    /* nothing */
-#endif
-
-#endif /* ! __ASSEMBLY__ */
-
 #endif /* _ARM_MACHINE_CPU_FEATURES_H */

libc/arch-arm/include/machine/endian.h

@@ -33,15 +33,6 @@
 
 #ifdef __GNUC__
 
-/*
- * REV and REV16 weren't available on ARM5 or ARM4.
- * We don't include <machine/cpu-features.h> because it pollutes the
- * namespace with macros like PLD.
- */
-#if !defined __ARM_ARCH_5__ && !defined __ARM_ARCH_5T__ && \
-    !defined __ARM_ARCH_5TE__ && !defined __ARM_ARCH_5TEJ__ && \
-    !defined __ARM_ARCH_4T__ && !defined __ARM_ARCH_4__
-
 /* According to RealView Assembler User's Guide, REV and REV16 are available
  * in Thumb code and 16-bit instructions when used in Thumb-2 code.
  *
@@ -55,13 +46,13 @@
  */
 #define __swap16md(x) ({                                        \
     register u_int16_t _x = (x);                                \
-    __asm volatile ("rev16 %0, %0" : "+l" (_x));                \
+    __asm__ __volatile__("rev16 %0, %0" : "+l" (_x));           \
     _x;                                                         \
 })
 
 #define __swap32md(x) ({                                        \
     register u_int32_t _x = (x);                                \
-    __asm volatile ("rev %0, %0" : "+l" (_x));                  \
+    __asm__ __volatile__("rev %0, %0" : "+l" (_x));             \
     _x;                                                         \
 })
 
@@ -74,7 +65,6 @@
 /* Tell sys/endian.h we have MD variants of the swap macros.  */
 #define MD_SWAP
 
-#endif  /* __ARM_ARCH__ */
 #endif  /* __GNUC__ */
 
 #if defined(__ARMEB__)

libc/arch-arm/include/machine/ieee.h

@@ -1,191 +0,0 @@
-/*	$OpenBSD: ieee.h,v 1.1 2004/02/01 05:09:49 drahn Exp $	*/
-/*	$NetBSD: ieee.h,v 1.2 2001/02/21 17:43:50 bjh21 Exp $	*/
-
-/*
- * Copyright (c) 1992, 1993
- *	The Regents of the University of California.  All rights reserved.
- *
- * This software was developed by the Computer Systems Engineering group
- * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
- * contributed to Berkeley.
- *
- * All advertising materials mentioning features or use of this software
- * must display the following acknowledgement:
- *	This product includes software developed by the University of
- *	California, Lawrence Berkeley Laboratory.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- * 1. Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in the
- *    documentation and/or other materials provided with the distribution.
- * 3. All advertising materials mentioning features or use of this software
- *    must display the following acknowledgement:
- *	This product includes software developed by the University of
- *	California, Berkeley and its contributors.
- * 4. Neither the name of the University nor the names of its contributors
- *    may be used to endorse or promote products derived from this software
- *    without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
- * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
- * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
- * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
- * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
- * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
- * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- *
- *	@(#)ieee.h	8.1 (Berkeley) 6/11/93
- */
-
-/*
- * ieee.h defines the machine-dependent layout of the machine's IEEE
- * floating point.
- */
-
-/*
- * Define the number of bits in each fraction and exponent.
- *
- *		     k	         k+1
- * Note that  1.0 x 2  == 0.1 x 2      and that denorms are represented
- *
- *					  (-exp_bias+1)
- * as fractions that look like 0.fffff x 2             .  This means that
- *
- *			 -126
- * the number 0.10000 x 2    , for instance, is the same as the normalized
- *
- *		-127			   -128
- * float 1.0 x 2    .  Thus, to represent 2    , we need one leading zero
- *
- *				  -129
- * in the fraction; to represent 2    , we need two, and so on.  This
- *
- *						     (-exp_bias-fracbits+1)
- * implies that the smallest denormalized number is 2
- *
- * for whichever format we are talking about: for single precision, for
- *
- *						-126		-149
- * instance, we get .00000000000000000000001 x 2    , or 1.0 x 2    , and
- *
- * -149 == -127 - 23 + 1.
- */
-
-/*
- * The ARM has two sets of FP data formats.  The FPA supports 32-bit, 64-bit
- * and 96-bit IEEE formats, with the words in big-endian order.  VFP supports
- * 32-bin and 64-bit IEEE formats with the words in the CPU's native byte
- * order.
- *
- * The FPA also has two packed decimal formats, but we ignore them here.
- */
-
-#define	SNG_EXPBITS	8
-#define	SNG_FRACBITS	23
-
-#define	DBL_EXPBITS	11
-#define	DBL_FRACBITS	52
-
-#ifndef __VFP_FP__
-#define	E80_EXPBITS	15
-#define	E80_FRACBITS	64
-
-#define	EXT_EXPBITS	15
-#define	EXT_FRACBITS	112
-#endif
-
-struct ieee_single {
-	u_int	sng_frac:23;
-	u_int	sng_exponent:8;
-	u_int	sng_sign:1;
-};
-
-#ifdef __VFP_FP__
-struct ieee_double {
-#ifdef __ARMEB__
-	u_int	dbl_sign:1;
-	u_int	dbl_exp:11;
-	u_int	dbl_frach:20;
-	u_int	dbl_fracl;
-#else /* !__ARMEB__ */
-	u_int	dbl_fracl;
-	u_int	dbl_frach:20;
-	u_int	dbl_exp:11;
-	u_int	dbl_sign:1;
-#endif /* !__ARMEB__ */
-};
-#else /* !__VFP_FP__ */
-struct ieee_double {
-	u_int	dbl_frach:20;
-	u_int	dbl_exp:11;
-	u_int	dbl_sign:1;
-	u_int	dbl_fracl;
-};
-
-union ieee_double_u {
-	double                  dblu_d;
-	struct ieee_double      dblu_dbl;
-};
-
-
-struct ieee_e80 {
-	u_int	e80_exp:15;
-	u_int	e80_zero:16;
-	u_int	e80_sign:1;
-	u_int	e80_frach:31;
-	u_int	e80_j:1;
-	u_int	e80_fracl;
-};
-
-struct ieee_ext {
-	u_int	ext_frach:16;
-	u_int	ext_exp:15;
-	u_int	ext_sign:1;
-	u_int	ext_frachm;
-	u_int	ext_fraclm;
-	u_int	ext_fracl;
-};
-#endif /* !__VFP_FP__ */
-
-/*
- * Floats whose exponent is in [1..INFNAN) (of whatever type) are
- * `normal'.  Floats whose exponent is INFNAN are either Inf or NaN.
- * Floats whose exponent is zero are either zero (iff all fraction
- * bits are zero) or subnormal values.
- *
- * A NaN is a `signalling NaN' if its QUIETNAN bit is clear in its
- * high fraction; if the bit is set, it is a `quiet NaN'.
- */
-#define	SNG_EXP_INFNAN	255
-#define	DBL_EXP_INFNAN	2047
-#ifndef __VFP_FP__
-#define	E80_EXP_INFNAN	32767
-#define	EXT_EXP_INFNAN	32767
-#endif /* !__VFP_FP__ */
-
-#if 0
-#define	SNG_QUIETNAN	(1 << 22)
-#define	DBL_QUIETNAN	(1 << 19)
-#ifndef __VFP_FP__
-#define	E80_QUIETNAN	(1 << 15)
-#define	EXT_QUIETNAN	(1 << 15)
-#endif /* !__VFP_FP__ */
-#endif
-
-/*
- * Exponent biases.
- */
-#define	SNG_EXP_BIAS	127
-#define	DBL_EXP_BIAS	1023
-#ifndef __VFP_FP__
-#define	E80_EXP_BIAS	16383
-#define	EXT_EXP_BIAS	16383
-#endif /* !__VFP_FP__ */

libc/arch-arm/include/machine/internal_types.h

@@ -1,10 +0,0 @@
-/* $OpenBSD: internal_types.h,v 1.2 2004/05/06 15:53:39 drahn Exp $ */
-/* Public domain */
-#ifndef _ARM_INTERNAL_TYPES_H_
-#define _ARM_INTERNAL_TYPES_H_
-
-#ifdef __CHAR_UNSIGNED__
-#define __machine_has_unsigned_chars
-#endif
-
-#endif

libc/arch-arm/include/machine/kernel.h

@@ -1,42 +0,0 @@
-/*
- * Copyright (C) 2008 The Android Open Source Project
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *  * Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- *  * Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in
- *    the documentation and/or other materials provided with the
- *    distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
- * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
- * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
- * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
- * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
- * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
- * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
- * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
- * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-#ifndef _ARCH_ARM_KERNEL_H
-#define _ARCH_ARM_KERNEL_H
-
-/* this file contains kernel-specific definitions that were optimized out of
-   our processed kernel headers, but still useful nonetheless... */
-
-typedef unsigned long   __kernel_blkcnt_t;
-typedef unsigned long   __kernel_blksize_t;
-
-/* these aren't really defined by the kernel headers though... */
-typedef unsigned long   __kernel_fsblkcnt_t;
-typedef unsigned long   __kernel_fsfilcnt_t;
-typedef unsigned int    __kernel_id_t;
-
-#endif /* _ARCH_ARM_KERNEL_H */

libc/arch-arm/include/machine/limits.h

@@ -1,63 +0,0 @@
-/*	$OpenBSD: limits.h,v 1.3 2006/01/06 22:48:46 millert Exp $	*/
-/*	$NetBSD: limits.h,v 1.4 2003/04/28 23:16:18 bjh21 Exp $	*/
-
-/*
- * Copyright (c) 1988 The Regents of the University of California.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- * 1. Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in the
- *    documentation and/or other materials provided with the distribution.
- * 3. Neither the name of the University nor the names of its contributors
- *    may be used to endorse or promote products derived from this software
- *    without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
- * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
- * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
- * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
- * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
- * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
- * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- *
- *	from: @(#)limits.h	7.2 (Berkeley) 6/28/90
- */
-
-#ifndef	_ARM32_LIMITS_H_
-#define	_ARM32_LIMITS_H_
-
-#include <sys/cdefs.h>
-
-#define	MB_LEN_MAX	1		/* no multibyte characters */
-
-#ifndef	SIZE_MAX
-#define	SIZE_MAX	UINT_MAX	/* max value for a size_t */
-#endif
-#ifndef SSIZE_MAX
-#define	SSIZE_MAX	INT_MAX		/* max value for a ssize_t */
-#endif
-
-#if __BSD_VISIBLE
-#define	SIZE_T_MAX	UINT_MAX	/* max value for a size_t (historic) */
-
-#define	UQUAD_MAX	0xffffffffffffffffULL		/* max unsigned quad */
-#define	QUAD_MAX	0x7fffffffffffffffLL		/* max signed quad */
-#define	QUAD_MIN	(-0x7fffffffffffffffLL-1)	/* min signed quad */
-
-#endif /* __BSD_VISIBLE */
-
-#define LONGLONG_BIT    64
-#define LONGLONG_MIN    (-9223372036854775807LL-1)
-#define LONGLONG_MAX    9223372036854775807LL
-#define ULONGLONG_MAX   18446744073709551615ULL
-
-#endif	/* _ARM32_LIMITS_H_ */

libc/arch-arm/krait/bionic/__strcat_chk.S

@@ -26,8 +26,8 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
     .syntax unified
 
@@ -38,13 +38,14 @@
 // Check that the two lengths together don't exceed the threshold, then
 // do a memcpy of the data.
 ENTRY(__strcat_chk)
-    .cfi_startproc
     pld     [r0, #0]
     push    {r0, lr}
+    .save   {r0, lr}
     .cfi_def_cfa_offset 8
     .cfi_rel_offset r0, 0
     .cfi_rel_offset lr, 4
     push    {r4, r5}
+    .save   {r4, r5}
     .cfi_adjust_cfa_offset 8
     .cfi_rel_offset r4, 0
     .cfi_rel_offset r5, 4
@@ -178,7 +179,7 @@ ENTRY(__strcat_chk)
 .L_strlen_done:
     add     r2, r3, r4
     cmp     r2, lr
-    bgt     .L_fortify_check_failed
+    bhi     __strcat_chk_failed
 
     // Set up the registers for the memcpy code.
     mov     r1, r5
@@ -186,13 +187,18 @@ ENTRY(__strcat_chk)
     mov     r2, r4
     add     r0, r0, r3
     pop     {r4, r5}
-    .cfi_adjust_cfa_offset -8
-    .cfi_restore r4
-    .cfi_restore r5
+END(__strcat_chk)
 
-    #include "memcpy_base.S"
+#define MEMCPY_BASE         __strcat_chk_memcpy_base
+#define MEMCPY_BASE_ALIGNED __strcat_chk_memcpy_base_aligned
+#include "memcpy_base.S"
 
-.L_fortify_check_failed:
+ENTRY_PRIVATE(__strcat_chk_failed)
+    .save   {r0, lr}
+    .save   {r4, r5}
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
     .cfi_adjust_cfa_offset 8
     .cfi_rel_offset r4, 0
     .cfi_rel_offset r5, 4
@@ -206,10 +212,8 @@ error_code:
     .word   BIONIC_EVENT_STRCAT_BUFFER_OVERFLOW
 error_message:
     .word   error_string-(1b+4)
-
-    .cfi_endproc
-END(__strcat_chk)
+END(__strcat_chk_failed)
 
     .data
 error_string:
-    .string "strcat buffer overflow"
+    .string "strcat: prevented write past end of buffer"

libc/arch-arm/krait/bionic/__strcpy_chk.S

@@ -26,8 +26,8 @@
  * SUCH DAMAGE.
  */
 
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
     .syntax unified
 
@@ -37,9 +37,9 @@
 // Get the length of the source string first, then do a memcpy of the data
 // instead of a strcpy.
 ENTRY(__strcpy_chk)
-    .cfi_startproc
     pld     [r0, #0]
     push    {r0, lr}
+    .save   {r0, lr}
     .cfi_def_cfa_offset 8
     .cfi_rel_offset r0, 0
     .cfi_rel_offset lr, 4
@@ -150,14 +150,22 @@ ENTRY(__strcpy_chk)
     pld     [r1, #64]
     ldr     r0, [sp]
     cmp     r3, lr
-    bge     .L_fortify_check_failed
+    bhs     __strcpy_chk_failed
 
     // Add 1 for copy length to get the string terminator.
     add     r2, r3, #1
+END(__strcpy_chk)
 
-    #include "memcpy_base.S"
+#define MEMCPY_BASE         __strcpy_chk_memcpy_base
+#define MEMCPY_BASE_ALIGNED __strcpy_chk_memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY_PRIVATE(__strcpy_chk_failed)
+    .save   {r0, lr}
+    .cfi_def_cfa_offset 8
+    .cfi_rel_offset r0, 0
+    .cfi_rel_offset lr, 4
 
-.L_fortify_check_failed:
     ldr     r0, error_message
     ldr     r1, error_code
 1:
@@ -167,9 +175,8 @@ error_code:
     .word   BIONIC_EVENT_STRCPY_BUFFER_OVERFLOW
 error_message:
     .word   error_string-(1b+4)
-    .cfi_endproc
-END(__strcpy_chk)
+END(__strcpy_chk_failed)
 
     .data
 error_string:
-    .string "strcpy buffer overflow"
+    .string "strcpy: prevented write past end of buffer"

libc/arch-arm/krait/bionic/memcpy.S

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2008 The Android Open Source Project
+ * Copyright (C) 2013 The Android Open Source Project
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -28,8 +28,8 @@
 
 /* Assumes neon instructions and a cache line size of 32 bytes. */
 
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
 /*
  * This code assumes it is running on a processor that supports all arm v7
@@ -44,30 +44,29 @@
         .thumb_func
 
 ENTRY(__memcpy_chk)
-        .cfi_startproc
         cmp         r2, r3
-        bgt         __memcpy_chk_fail
+        bhi         __memcpy_chk_fail
 
         // Fall through to memcpy...
-        .cfi_endproc
 END(__memcpy_chk)
 
 ENTRY(memcpy)
-        .cfi_startproc
         pld     [r1, #64]
         stmfd   sp!, {r0, lr}
+        .save   {r0, lr}
         .cfi_def_cfa_offset 8
         .cfi_rel_offset r0, 0
         .cfi_rel_offset lr, 4
-
-        #include "memcpy_base.S"
-        .cfi_endproc
 END(memcpy)
 
-        .cfi_startproc
-__memcpy_chk_fail:
+#define MEMCPY_BASE         __memcpy_base
+#define MEMCPY_BASE_ALIGNED __memcpy_base_aligned
+#include "memcpy_base.S"
+
+ENTRY_PRIVATE(__memcpy_chk_fail)
         // Preserve lr for backtrace.
         push    {lr}
+        .save   {lr}
         .cfi_def_cfa_offset 4
         .cfi_rel_offset lr, 0
 
@@ -80,8 +79,8 @@ error_code:
         .word   BIONIC_EVENT_MEMCPY_BUFFER_OVERFLOW
 error_message:
         .word   error_string-(1b+4)
-        .cfi_endproc
+END(__memcpy_chk_fail)
 
         .data
 error_string:
-        .string     "memcpy buffer overflow"
+        .string     "memcpy: prevented write past end of buffer"

libc/arch-arm/krait/bionic/memcpy_base.S

@@ -35,6 +35,12 @@
 
 // Assumes neon instructions and a cache line size of 32 bytes.
 
+ENTRY_PRIVATE(MEMCPY_BASE)
+        .save {r0, lr}
+        .cfi_def_cfa_offset 8
+        .cfi_rel_offset r0, 0
+        .cfi_rel_offset lr, 4
+
         /* do we have at least 16-bytes to copy (needed for alignment below) */
         cmp         r2, #16
         blo         5f
@@ -72,7 +78,7 @@
 1:      /* The main loop copies 64 bytes at a time */
         vld1.8      {d0  - d3},   [r1]!
         vld1.8      {d4  - d7},   [r1]!
-        pld         [r1, #(32*2)]
+        pld         [r1, #(32*8)]
         subs        r2, r2, #64
         vst1.8      {d0  - d3},   [r0, :128]!
         vst1.8      {d4  - d7},   [r0, :128]!
@@ -115,3 +121,4 @@
 
         ldmfd       sp!, {r0, lr}
         bx          lr
+END(MEMCPY_BASE)

libc/arch-arm/krait/bionic/memset.S

@@ -27,8 +27,8 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
-#include "libc_events.h"
+#include <private/bionic_asm.h>
+#include <private/libc_events.h>
 
 /*
  * This code assumes it is running on a processor that supports all arm v7
@@ -39,11 +39,11 @@
     .fpu    neon
 
 ENTRY(__memset_chk)
-        .cfi_startproc
         cmp         r2, r3
         bls         .L_done
 
         // Preserve lr for backtrace.
+        .save       {lr}
         push        {lr}
         .cfi_def_cfa_offset 4
         .cfi_rel_offset lr, 0
@@ -57,23 +57,19 @@ error_code:
         .word       BIONIC_EVENT_MEMSET_BUFFER_OVERFLOW
 error_message:
         .word       error_string-(1b+8)
-
-        .cfi_endproc
 END(__memset_chk)
 
 ENTRY(bzero)
-        .cfi_startproc
         mov     r2, r1
         mov     r1, #0
 
 .L_done:
         // Fall through to memset...
-        .cfi_endproc
 END(bzero)
 
 /* memset() returns its first argument.  */
 ENTRY(memset)
-        .cfi_startproc
+        .save       {r0}
         stmfd       sp!, {r0}
         .cfi_def_cfa_offset 4
         .cfi_rel_offset r0, 0
@@ -109,9 +105,8 @@ ENTRY(memset)
         strcsb      r1, [r0], #1
         ldmfd       sp!, {r0}
         bx          lr
-        .cfi_endproc
 END(memset)
 
         .data
 error_string:
-        .string     "memset buffer overflow"
+        .string     "memset: prevented write past end of buffer"

libc/arch-arm/krait/bionic/strcmp.S

@@ -27,7 +27,7 @@
  */
 
 #include <machine/cpu-features.h>
-#include <machine/asm.h>
+#include <private/bionic_asm.h>
 
 #ifdef __ARMEB__
 #define S2LOMEM lsl
@@ -123,8 +123,13 @@ ENTRY(strcmp)
         .macro  init
         /* Macro to save temporary registers and prepare magic values.  */
         subs    sp, sp, #16
+        .cfi_def_cfa_offset 16
         strd    r4, r5, [sp, #8]
+        .cfi_rel_offset r4, 0
+        .cfi_rel_offset r5, 4
         strd    r6, r7, [sp]
+        .cfi_rel_offset r6, 8
+        .cfi_rel_offset r7, 12
         mvn     r6, #0  /* all F */
         mov     r7, #0  /* all 0 */
         .endm   /* init */
@@ -140,19 +145,8 @@ ENTRY(strcmp)
 
         .macro  magic_find_zero_bytes w1
         /* Macro to find all-zero bytes in w1, result is in ip.  */
-#if (defined (__ARM_FEATURE_DSP))
         uadd8   ip, \w1, r6
         sel     ip, r7, r6
-#else /* not defined (__ARM_FEATURE_DSP) */
-        /* __ARM_FEATURE_DSP is not defined for some Cortex-M processors.
-        Coincidently, these processors only have Thumb-2 mode, where we can use the
-        the (large) magic constant available directly as an immediate in instructions.
-        Note that we cannot use the magic constant in ARM mode, where we need
-        to create the constant in a register.  */
-        sub     ip, \w1, #0x01010101
-        bic     ip, ip, \w1
-        and     ip, ip, #0x80808080
-#endif /* not defined (__ARM_FEATURE_DSP) */
         .endm /* magic_find_zero_bytes */
 
         .macro  setup_return w1 w2
@@ -171,12 +165,13 @@ ENTRY(strcmp)
         /* Are both strings double-word aligned?  */
         orr     ip, r0, r1
         tst     ip, #7
-        bne     do_align
+        bne     .L_do_align
 
         /* Fast path.  */
+        .save   {r4-r7}
         init
 
-doubleword_aligned:
+.L_doubleword_aligned:
 
         /* Get here when the strings to compare are double-word aligned.  */
         /* Compare two words in every iteration.  */
@@ -189,14 +184,14 @@ doubleword_aligned:
         ldrd    r2, r3, [r0], #8
         ldrd    r4, r5, [r1], #8
 
-        magic_compare_and_branch w1=r2, w2=r4, label=return_24
-        magic_compare_and_branch w1=r3, w2=r5, label=return_35
+        magic_compare_and_branch w1=r2, w2=r4, label=.L_return_24
+        magic_compare_and_branch w1=r3, w2=r5, label=.L_return_35
         b       2b
 
-do_align:
+.L_do_align:
         /* Is the first string word-aligned?  */
         ands    ip, r0, #3
-        beq     word_aligned_r0
+        beq     .L_word_aligned_r0
 
         /* Fast compare byte by byte until the first string is word-aligned.  */
         /* The offset of r0 from a word boundary is in ip. Thus, the number of bytes
@@ -204,58 +199,58 @@ do_align:
         bic     r0, r0, #3
         ldr     r2, [r0], #4
         lsls    ip, ip, #31
-        beq     byte2
-        bcs     byte3
+        beq     .L_byte2
+        bcs     .L_byte3
 
-byte1:
+.L_byte1:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE1_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbz   reg=r3, label=fast_return
+        bne     .L_fast_return
+        m_cbz   reg=r3, label=.L_fast_return
 
-byte2:
+.L_byte2:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE2_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbz   reg=r3, label=fast_return
+        bne     .L_fast_return
+        m_cbz   reg=r3, label=.L_fast_return
 
-byte3:
+.L_byte3:
         ldrb    ip, [r1], #1
         uxtb    r3, r2, ror #BYTE3_OFFSET
         subs    ip, r3, ip
-        bne     fast_return
-        m_cbnz  reg=r3, label=word_aligned_r0
+        bne     .L_fast_return
+        m_cbnz  reg=r3, label=.L_word_aligned_r0
 
-fast_return:
+.L_fast_return:
         mov     r0, ip
         bx      lr
 
-word_aligned_r0:
+.L_word_aligned_r0:
         init
         /* The first string is word-aligned.  */
         /* Is the second string word-aligned?  */
         ands    ip, r1, #3
-        bne     strcmp_unaligned
+        bne     .L_strcmp_unaligned
 
-word_aligned:
+.L_word_aligned:
         /* The strings are word-aligned. */
         /* Is the first string double-word aligned?  */
         tst     r0, #4
-        beq     doubleword_aligned_r0
+        beq     .L_doubleword_aligned_r0
 
         /* If r0 is not double-word aligned yet, align it by loading
         and comparing the next word from each string.  */
         ldr     r2, [r0], #4
         ldr     r4, [r1], #4
-        magic_compare_and_branch w1=r2 w2=r4 label=return_24
+        magic_compare_and_branch w1=r2 w2=r4 label=.L_return_24
 
-doubleword_aligned_r0:
+.L_doubleword_aligned_r0:
         /* Get here when r0 is double-word aligned.  */
         /* Is r1 doubleword_aligned?  */
         tst     r1, #4
-        beq     doubleword_aligned
+        beq     .L_doubleword_aligned
 
         /* Get here when the strings to compare are word-aligned,
         r0 is double-word aligned, but r1 is not double-word aligned.  */
@@ -271,9 +266,9 @@ doubleword_aligned_r0:
 
         /* Load the next double-word from each string and compare.  */
         ldrd    r2, r3, [r0], #8
-        magic_compare_and_branch w1=r2 w2=r5 label=return_25
+        magic_compare_and_branch w1=r2 w2=r5 label=.L_return_25
         ldrd    r4, r5, [r1], #8
-        magic_compare_and_branch w1=r3 w2=r4 label=return_34
+        magic_compare_and_branch w1=r3 w2=r4 label=.L_return_34
         b       3b
 
         .macro miscmp_word offsetlo offsethi
@@ -297,46 +292,46 @@ doubleword_aligned_r0:
         and     r2, r3, r6, S2LOMEM #\offsetlo
         it      eq
         cmpeq   r2, r5
-        bne     return_25
+        bne     .L_return_25
         ldr     r5, [r1], #4
         cmp     ip, #0
         eor r3, r2, r3
         S2HIMEM r2, r5, #\offsethi
         it      eq
         cmpeq   r3, r2
-        bne     return_32
+        bne     .L_return_32
         b       7b
         .endm /* miscmp_word */
 
-strcmp_unaligned:
+.L_strcmp_unaligned:
         /* r0 is word-aligned, r1 is at offset ip from a word.  */
         /* Align r1 to the (previous) word-boundary.  */
         bic     r1, r1, #3
 
         /* Unaligned comparison word by word using LDRs. */
         cmp     ip, #2
-        beq     miscmp_word_16                    /* If ip == 2.  */
-        bge     miscmp_word_24                    /* If ip == 3.  */
+        beq     .L_miscmp_word_16                 /* If ip == 2.  */
+        bge     .L_miscmp_word_24                 /* If ip == 3.  */
         miscmp_word offsetlo=8 offsethi=24        /* If ip == 1.  */
-miscmp_word_24:  miscmp_word offsetlo=24 offsethi=8
+.L_miscmp_word_24:  miscmp_word offsetlo=24 offsethi=8
 
 
-return_32:
+.L_return_32:
         setup_return w1=r3, w2=r2
-        b       do_return
-return_34:
+        b       .L_do_return
+.L_return_34:
         setup_return w1=r3, w2=r4
-        b       do_return
-return_25:
+        b       .L_do_return
+.L_return_25:
         setup_return w1=r2, w2=r5
-        b       do_return
-return_35:
+        b       .L_do_return
+.L_return_35:
         setup_return w1=r3, w2=r5
-        b       do_return
-return_24:
+        b       .L_do_return
+.L_return_24:
         setup_return w1=r2, w2=r4
 
-do_return:
+.L_do_return:
 
 #ifdef __ARMEB__
         mov     r0, ip
@@ -348,11 +343,16 @@ do_return:
         ldrd    r6, r7, [sp]
         ldrd    r4, r5, [sp, #8]
         adds    sp, sp, #16
+        .cfi_def_cfa_offset 0
+        .cfi_restore r4
+        .cfi_restore r5
+        .cfi_restore r6
+        .cfi_restore r7
 
         /* There is a zero or a different byte between r1 and r2.  */
         /* r0 contains a mask of all-zero bytes in r1.  */
         /* Using r0 and not ip here because cbz requires low register.  */
-        m_cbz   reg=r0, label=compute_return_value
+        m_cbz   reg=r0, label=.L_compute_return_value
         clz     r0, r0
         /* r0 contains the number of bits on the left of the first all-zero byte in r1.  */
         rsb     r0, r0, #24
@@ -360,7 +360,7 @@ do_return:
         lsr     r1, r1, r0
         lsr     r2, r2, r0
 
-compute_return_value:
+.L_compute_return_value:
         movs    r0, #1
         cmp     r1, r2
         /* The return value is computed as follows.
@@ -380,7 +380,7 @@ compute_return_value:
      * previous version. See bionic/libc/arch-arm/cortex-a15/bionic/strcmp.S
      * for the unedited version of this code.
      */
-miscmp_word_16:
+.L_miscmp_word_16:
 	wp1 .req r0
 	wp2 .req r1
 	b1  .req r2
@@ -453,6 +453,11 @@ miscmp_word_16:
     ldrd    r6, r7, [sp]
     ldrd    r4, r5, [sp, #8]
     adds    sp, sp, #16
+    .cfi_def_cfa_offset 0
+    .cfi_restore r4
+    .cfi_restore r5
+    .cfi_restore r6
+    .cfi_restore r7
 
 	bx	lr
 
@@ -472,6 +477,11 @@ miscmp_word_16:
     ldrd    r6, r7, [sp]
     ldrd    r4, r5, [sp, #8]
     adds    sp, sp, #16
+    .cfi_def_cfa_offset 0
+    .cfi_restore r4
+    .cfi_restore r5
+    .cfi_restore r6
+    .cfi_restore r7
 
 	bx	lr
 END(strcmp)

libc/arch-arm/krait/krait.mk

@@ -1,11 +1,12 @@
-$(call libc-add-cpu-variant-src,MEMCPY,arch-arm/krait/bionic/memcpy.S)
-$(call libc-add-cpu-variant-src,MEMSET,arch-arm/krait/bionic/memset.S)
-$(call libc-add-cpu-variant-src,STRCMP,arch-arm/krait/bionic/strcmp.S)
-$(call libc-add-cpu-variant-src,__STRCAT_CHK,arch-arm/krait/bionic/__strcat_chk.S)
-$(call libc-add-cpu-variant-src,__STRCPY_CHK,arch-arm/krait/bionic/__strcpy_chk.S)
-# Use cortex-a15 versions of strcat/strcpy/strlen.
-$(call libc-add-cpu-variant-src,STRCAT,arch-arm/cortex-a15/bionic/strcat.S)
-$(call libc-add-cpu-variant-src,STRCPY,arch-arm/cortex-a15/bionic/strcpy.S)
-$(call libc-add-cpu-variant-src,STRLEN,arch-arm/cortex-a15/bionic/strlen.S)
+libc_bionic_src_files_arm += \
+    arch-arm/krait/bionic/memcpy.S \
+    arch-arm/krait/bionic/memset.S \
+    arch-arm/krait/bionic/strcmp.S \
+    arch-arm/krait/bionic/__strcat_chk.S \
+    arch-arm/krait/bionic/__strcpy_chk.S \
 
-include bionic/libc/arch-arm/generic/generic.mk
+# Use cortex-a15 versions of strcat/strcpy/strlen.
+libc_bionic_src_files_arm += \
+    arch-arm/cortex-a15/bionic/strcat.S \
+    arch-arm/cortex-a15/bionic/strcpy.S \
+    arch-arm/cortex-a15/bionic/strlen.S \

libc/arch-arm/syscalls.mk

@@ -1,208 +0,0 @@
-# auto-generated by gensyscalls.py, do not touch
-syscall_src := 
-syscall_src += arch-arm/syscalls/_exit.S
-syscall_src += arch-arm/syscalls/_exit_thread.S
-syscall_src += arch-arm/syscalls/__fork.S
-syscall_src += arch-arm/syscalls/__waitid.S
-syscall_src += arch-arm/syscalls/wait4.S
-syscall_src += arch-arm/syscalls/__sys_clone.S
-syscall_src += arch-arm/syscalls/execve.S
-syscall_src += arch-arm/syscalls/__setuid.S
-syscall_src += arch-arm/syscalls/getuid.S
-syscall_src += arch-arm/syscalls/getgid.S
-syscall_src += arch-arm/syscalls/geteuid.S
-syscall_src += arch-arm/syscalls/getegid.S
-syscall_src += arch-arm/syscalls/getresuid.S
-syscall_src += arch-arm/syscalls/getresgid.S
-syscall_src += arch-arm/syscalls/gettid.S
-syscall_src += arch-arm/syscalls/readahead.S
-syscall_src += arch-arm/syscalls/getgroups.S
-syscall_src += arch-arm/syscalls/getpgid.S
-syscall_src += arch-arm/syscalls/getppid.S
-syscall_src += arch-arm/syscalls/getsid.S
-syscall_src += arch-arm/syscalls/setsid.S
-syscall_src += arch-arm/syscalls/setgid.S
-syscall_src += arch-arm/syscalls/__setreuid.S
-syscall_src += arch-arm/syscalls/__setresuid.S
-syscall_src += arch-arm/syscalls/setresgid.S
-syscall_src += arch-arm/syscalls/__brk.S
-syscall_src += arch-arm/syscalls/__ptrace.S
-syscall_src += arch-arm/syscalls/__getpriority.S
-syscall_src += arch-arm/syscalls/setpriority.S
-syscall_src += arch-arm/syscalls/setrlimit.S
-syscall_src += arch-arm/syscalls/getrlimit.S
-syscall_src += arch-arm/syscalls/getrusage.S
-syscall_src += arch-arm/syscalls/setgroups.S
-syscall_src += arch-arm/syscalls/setpgid.S
-syscall_src += arch-arm/syscalls/vfork.S
-syscall_src += arch-arm/syscalls/setregid.S
-syscall_src += arch-arm/syscalls/chroot.S
-syscall_src += arch-arm/syscalls/prctl.S
-syscall_src += arch-arm/syscalls/capget.S
-syscall_src += arch-arm/syscalls/capset.S
-syscall_src += arch-arm/syscalls/sigaltstack.S
-syscall_src += arch-arm/syscalls/acct.S
-syscall_src += arch-arm/syscalls/read.S
-syscall_src += arch-arm/syscalls/write.S
-syscall_src += arch-arm/syscalls/pread64.S
-syscall_src += arch-arm/syscalls/pwrite64.S
-syscall_src += arch-arm/syscalls/__open.S
-syscall_src += arch-arm/syscalls/__openat.S
-syscall_src += arch-arm/syscalls/close.S
-syscall_src += arch-arm/syscalls/lseek.S
-syscall_src += arch-arm/syscalls/__llseek.S
-syscall_src += arch-arm/syscalls/getpid.S
-syscall_src += arch-arm/syscalls/__mmap2.S
-syscall_src += arch-arm/syscalls/munmap.S
-syscall_src += arch-arm/syscalls/mremap.S
-syscall_src += arch-arm/syscalls/msync.S
-syscall_src += arch-arm/syscalls/mprotect.S
-syscall_src += arch-arm/syscalls/madvise.S
-syscall_src += arch-arm/syscalls/mlock.S
-syscall_src += arch-arm/syscalls/munlock.S
-syscall_src += arch-arm/syscalls/mlockall.S
-syscall_src += arch-arm/syscalls/munlockall.S
-syscall_src += arch-arm/syscalls/mincore.S
-syscall_src += arch-arm/syscalls/__ioctl.S
-syscall_src += arch-arm/syscalls/readv.S
-syscall_src += arch-arm/syscalls/writev.S
-syscall_src += arch-arm/syscalls/__fcntl.S
-syscall_src += arch-arm/syscalls/flock.S
-syscall_src += arch-arm/syscalls/fchmod.S
-syscall_src += arch-arm/syscalls/dup.S
-syscall_src += arch-arm/syscalls/pipe.S
-syscall_src += arch-arm/syscalls/pipe2.S
-syscall_src += arch-arm/syscalls/dup2.S
-syscall_src += arch-arm/syscalls/select.S
-syscall_src += arch-arm/syscalls/ftruncate.S
-syscall_src += arch-arm/syscalls/ftruncate64.S
-syscall_src += arch-arm/syscalls/getdents.S
-syscall_src += arch-arm/syscalls/fsync.S
-syscall_src += arch-arm/syscalls/fdatasync.S
-syscall_src += arch-arm/syscalls/fchown.S
-syscall_src += arch-arm/syscalls/sync.S
-syscall_src += arch-arm/syscalls/__fcntl64.S
-syscall_src += arch-arm/syscalls/__fstatfs64.S
-syscall_src += arch-arm/syscalls/sendfile.S
-syscall_src += arch-arm/syscalls/fstatat.S
-syscall_src += arch-arm/syscalls/mkdirat.S
-syscall_src += arch-arm/syscalls/fchownat.S
-syscall_src += arch-arm/syscalls/fchmodat.S
-syscall_src += arch-arm/syscalls/renameat.S
-syscall_src += arch-arm/syscalls/fsetxattr.S
-syscall_src += arch-arm/syscalls/fgetxattr.S
-syscall_src += arch-arm/syscalls/flistxattr.S
-syscall_src += arch-arm/syscalls/fremovexattr.S
-syscall_src += arch-arm/syscalls/link.S
-syscall_src += arch-arm/syscalls/unlink.S
-syscall_src += arch-arm/syscalls/unlinkat.S
-syscall_src += arch-arm/syscalls/chdir.S
-syscall_src += arch-arm/syscalls/mknod.S
-syscall_src += arch-arm/syscalls/chmod.S
-syscall_src += arch-arm/syscalls/chown.S
-syscall_src += arch-arm/syscalls/lchown.S
-syscall_src += arch-arm/syscalls/mount.S
-syscall_src += arch-arm/syscalls/umount2.S
-syscall_src += arch-arm/syscalls/fstat.S
-syscall_src += arch-arm/syscalls/stat.S
-syscall_src += arch-arm/syscalls/lstat.S
-syscall_src += arch-arm/syscalls/mkdir.S
-syscall_src += arch-arm/syscalls/readlink.S
-syscall_src += arch-arm/syscalls/rmdir.S
-syscall_src += arch-arm/syscalls/rename.S
-syscall_src += arch-arm/syscalls/__getcwd.S
-syscall_src += arch-arm/syscalls/access.S
-syscall_src += arch-arm/syscalls/faccessat.S
-syscall_src += arch-arm/syscalls/symlink.S
-syscall_src += arch-arm/syscalls/fchdir.S
-syscall_src += arch-arm/syscalls/truncate.S
-syscall_src += arch-arm/syscalls/setxattr.S
-syscall_src += arch-arm/syscalls/lsetxattr.S
-syscall_src += arch-arm/syscalls/getxattr.S
-syscall_src += arch-arm/syscalls/lgetxattr.S
-syscall_src += arch-arm/syscalls/listxattr.S
-syscall_src += arch-arm/syscalls/llistxattr.S
-syscall_src += arch-arm/syscalls/removexattr.S
-syscall_src += arch-arm/syscalls/lremovexattr.S
-syscall_src += arch-arm/syscalls/__statfs64.S
-syscall_src += arch-arm/syscalls/unshare.S
-syscall_src += arch-arm/syscalls/swapon.S
-syscall_src += arch-arm/syscalls/swapoff.S
-syscall_src += arch-arm/syscalls/pause.S
-syscall_src += arch-arm/syscalls/gettimeofday.S
-syscall_src += arch-arm/syscalls/settimeofday.S
-syscall_src += arch-arm/syscalls/times.S
-syscall_src += arch-arm/syscalls/nanosleep.S
-syscall_src += arch-arm/syscalls/clock_gettime.S
-syscall_src += arch-arm/syscalls/clock_settime.S
-syscall_src += arch-arm/syscalls/clock_getres.S
-syscall_src += arch-arm/syscalls/clock_nanosleep.S
-syscall_src += arch-arm/syscalls/getitimer.S
-syscall_src += arch-arm/syscalls/setitimer.S
-syscall_src += arch-arm/syscalls/__timer_create.S
-syscall_src += arch-arm/syscalls/__timer_settime.S
-syscall_src += arch-arm/syscalls/__timer_gettime.S
-syscall_src += arch-arm/syscalls/__timer_getoverrun.S
-syscall_src += arch-arm/syscalls/__timer_delete.S
-syscall_src += arch-arm/syscalls/utimes.S
-syscall_src += arch-arm/syscalls/utimensat.S
-syscall_src += arch-arm/syscalls/timerfd_create.S
-syscall_src += arch-arm/syscalls/timerfd_settime.S
-syscall_src += arch-arm/syscalls/timerfd_gettime.S
-syscall_src += arch-arm/syscalls/sigaction.S
-syscall_src += arch-arm/syscalls/sigprocmask.S
-syscall_src += arch-arm/syscalls/__sigsuspend.S
-syscall_src += arch-arm/syscalls/__rt_sigaction.S
-syscall_src += arch-arm/syscalls/__rt_sigprocmask.S
-syscall_src += arch-arm/syscalls/__rt_sigtimedwait.S
-syscall_src += arch-arm/syscalls/sigpending.S
-syscall_src += arch-arm/syscalls/signalfd4.S
-syscall_src += arch-arm/syscalls/socket.S
-syscall_src += arch-arm/syscalls/socketpair.S
-syscall_src += arch-arm/syscalls/bind.S
-syscall_src += arch-arm/syscalls/connect.S
-syscall_src += arch-arm/syscalls/listen.S
-syscall_src += arch-arm/syscalls/accept.S
-syscall_src += arch-arm/syscalls/getsockname.S
-syscall_src += arch-arm/syscalls/getpeername.S
-syscall_src += arch-arm/syscalls/sendto.S
-syscall_src += arch-arm/syscalls/recvfrom.S
-syscall_src += arch-arm/syscalls/shutdown.S
-syscall_src += arch-arm/syscalls/setsockopt.S
-syscall_src += arch-arm/syscalls/getsockopt.S
-syscall_src += arch-arm/syscalls/sendmsg.S
-syscall_src += arch-arm/syscalls/recvmsg.S
-syscall_src += arch-arm/syscalls/sched_setscheduler.S
-syscall_src += arch-arm/syscalls/sched_getscheduler.S
-syscall_src += arch-arm/syscalls/sched_yield.S
-syscall_src += arch-arm/syscalls/sched_setparam.S
-syscall_src += arch-arm/syscalls/sched_getparam.S
-syscall_src += arch-arm/syscalls/sched_get_priority_max.S
-syscall_src += arch-arm/syscalls/sched_get_priority_min.S
-syscall_src += arch-arm/syscalls/sched_rr_get_interval.S
-syscall_src += arch-arm/syscalls/sched_setaffinity.S
-syscall_src += arch-arm/syscalls/__sched_getaffinity.S
-syscall_src += arch-arm/syscalls/__getcpu.S
-syscall_src += arch-arm/syscalls/ioprio_set.S
-syscall_src += arch-arm/syscalls/ioprio_get.S
-syscall_src += arch-arm/syscalls/uname.S
-syscall_src += arch-arm/syscalls/umask.S
-syscall_src += arch-arm/syscalls/__reboot.S
-syscall_src += arch-arm/syscalls/__syslog.S
-syscall_src += arch-arm/syscalls/init_module.S
-syscall_src += arch-arm/syscalls/delete_module.S
-syscall_src += arch-arm/syscalls/klogctl.S
-syscall_src += arch-arm/syscalls/sysinfo.S
-syscall_src += arch-arm/syscalls/personality.S
-syscall_src += arch-arm/syscalls/perf_event_open.S
-syscall_src += arch-arm/syscalls/futex.S
-syscall_src += arch-arm/syscalls/epoll_create.S
-syscall_src += arch-arm/syscalls/epoll_ctl.S
-syscall_src += arch-arm/syscalls/epoll_wait.S
-syscall_src += arch-arm/syscalls/inotify_init.S
-syscall_src += arch-arm/syscalls/inotify_add_watch.S
-syscall_src += arch-arm/syscalls/inotify_rm_watch.S
-syscall_src += arch-arm/syscalls/poll.S
-syscall_src += arch-arm/syscalls/eventfd.S
-syscall_src += arch-arm/syscalls/__set_tls.S
-syscall_src += arch-arm/syscalls/cacheflush.S

libc/arch-arm/syscalls/__brk.S

@@ -1,7 +1,6 @@
-/* autogenerated by gensyscalls.py */
-#include <asm/unistd.h>
-#include <linux/err.h>
-#include <machine/asm.h>
+/* Generated by gensyscalls.py. Do not edit. */
+
+#include <private/bionic_asm.h>
 
 ENTRY(__brk)
     mov     ip, r7

libc/arch-arm/syscalls/__epoll_pwait.S

@@ -0,0 +1,22 @@
+/* Generated by gensyscalls.py. Do not edit. */
+
+#include <private/bionic_asm.h>
+
+ENTRY(__epoll_pwait)
+    mov     ip, sp
+    stmfd   sp!, {r4, r5, r6, r7}
+    .cfi_def_cfa_offset 16
+    .cfi_rel_offset r4, 0
+    .cfi_rel_offset r5, 4
+    .cfi_rel_offset r6, 8
+    .cfi_rel_offset r7, 12
+    ldmfd   ip, {r4, r5, r6}
+    ldr     r7, =__NR_epoll_pwait
+    swi     #0
+    ldmfd   sp!, {r4, r5, r6, r7}
+    .cfi_def_cfa_offset 0
+    cmn     r0, #(MAX_ERRNO + 1)
+    bxls    lr
+    neg     r0, r0
+    b       __set_errno
+END(__epoll_pwait)

libc/arch-arm/syscalls/__exit.S

@@ -0,0 +1,14 @@
+/* Generated by gensyscalls.py. Do not edit. */
+
+#include <private/bionic_asm.h>
+
+ENTRY(__exit)
+    mov     ip, r7
+    ldr     r7, =__NR_exit
+    swi     #0
+    mov     r7, ip
+    cmn     r0, #(MAX_ERRNO + 1)
+    bxls    lr
+    neg     r0, r0
+    b       __set_errno
+END(__exit)

libc/arch-arm/syscalls/__fcntl.S

@@ -1,15 +0,0 @@
-/* autogenerated by gensyscalls.py */
-#include <asm/unistd.h>
-#include <linux/err.h>
-#include <machine/asm.h>
-
-ENTRY(__fcntl)
-    mov     ip, r7
-    ldr     r7, =__NR_fcntl
-    swi     #0
-    mov     r7, ip
-    cmn     r0, #(MAX_ERRNO + 1)
-    bxls    lr
-    neg     r0, r0
-    b       __set_errno
-END(__fcntl)

libc/arch-arm/syscalls/__fcntl64.S

@@ -1,7 +1,6 @@
-/* autogenerated by gensyscalls.py */
-#include <asm/unistd.h>
-#include <linux/err.h>
-#include <machine/asm.h>
+/* Generated by gensyscalls.py. Do not edit. */
+
+#include <private/bionic_asm.h>
 
 ENTRY(__fcntl64)
     mov     ip, r7