merge in mnc-release history after reset to mnc-dev

Changes affecting future time stamps

    North Korea switches to +0830 on 2015-08-15.  (Thanks to Steffen Thorsen.)
    The abbreviation remains "KST".  (Thanks to Robert Elz.)

    Uruguay no longer observes DST.  (Thanks to Steffen Thorsen
    and Pablo Camargo.)

  Changes affecting past and future time stamps

    Moldova starts and ends DST at 00:00 UTC, not at 01:00 UTC.
    (Thanks to Roman Tudos.)

Bug: 23480367
(cherry-picked from commit 6e7bc700e2)

Change-Id: I4d82604366e9e30a2100e09433e2772fe6957fff

ABI-bugs.txt

@@ -1,12 +0,0 @@
-KNOWN ABI BUGS
---------------
-
-  time_t is 32-bit. http://b/5819737
-
-  off_t is 32-bit. There is off64_t, but no _FILE_OFFSET_BITS support.
-
-  sigset_t is too small on ARM and x86 (but correct on MIPS), so support
-  for real-time signals is broken. http://b/5828899
-
-  atexit(3) handlers registered by a shared library aren't called on
-  dlclose(3); this only affects ARM. http://b/4998315

CPPLINT.cfg

@@ -0,0 +1,2 @@
+set noparent
+filter=-build/header_guard,-runtime/int,-readability/function

README.md

@@ -208,19 +208,17 @@ The tests are all built from the tests/ directory.
 The host tests require that you have `lunch`ed either an x86 or x86_64 target.
 
     $ mma
-    # 64-bit tests for 64-bit targets, 32-bit otherwise.
-    $ mm bionic-unit-tests-run-on-host
-    # Only exists for 64-bit targets.
     $ mm bionic-unit-tests-run-on-host32
+    $ mm bionic-unit-tests-run-on-host64  # For 64-bit *targets* only.
 
 ### Against glibc
 
 As a way to check that our tests do in fact test the correct behavior (and not
 just the behavior we think is correct), it is possible to run the tests against
-the host's glibc.
+the host's glibc. The executables are already in your path.
 
     $ mma
-    $ bionic-unit-tests-glibc32 # already in your path
+    $ bionic-unit-tests-glibc32
     $ bionic-unit-tests-glibc64
 
 
@@ -256,3 +254,20 @@ First, build and run the host tests as usual (see above).
     $ genhtml -o covreport coverage.info # or lcov --list coverage.info
 
 The coverage report is now available at `covreport/index.html`.
+
+
+LP32 ABI bugs
+-------------
+
+This probably belongs in the NDK documentation rather than here, but these
+are the known ABI bugs in LP32:
+
+ * `time_t` is 32-bit. <http://b/5819737>
+
+ * `off_t` is 32-bit. There is `off64_t`, but no `_FILE_OFFSET_BITS` support.
+   Many of the `off64_t` functions are missing in older releases, and
+   stdio uses 32-bit offsets, so there's no way to fully implement
+   `_FILE_OFFSET_BITS`.
+
+ * `sigset_t` is too small on ARM and x86 (but correct on MIPS), so support
+   for real-time signals is broken. <http://b/5828899>

benchmarks/Android.mk

@@ -14,24 +14,58 @@
 # limitations under the License.
 #
 
-ifneq ($(BUILD_TINY_ANDROID), true)
-
 LOCAL_PATH := $(call my-dir)
 
+# -----------------------------------------------------------------------------
+# Benchmarks library, usable by projects outside this directory.
+# -----------------------------------------------------------------------------
+
+benchmark_cflags := \
+    -O2 \
+    -fno-builtin \
+    -Wall \
+    -Wextra \
+    -Werror \
+    -Wunused \
+
+benchmark_cppflags := \
+    -std=gnu++11 \
+
+benchmarklib_src_files := \
+    Benchmark.cpp \
+    utils.cpp \
+    main.cpp \
+
+include $(CLEAR_VARS)
+LOCAL_MODULE := libbenchmark
+LOCAL_CFLAGS := $(benchmark_cflags)
+LOCAL_CPPFLAGS := $(benchmark_cppflags)
+LOCAL_SRC_FILES := $(benchmarklib_src_files)
+LOCAL_C_INCLUDES := $(benchmark_c_includes)
+LOCAL_STATIC_LIBRARIES := libbase
+include $(BUILD_STATIC_LIBRARY)
+
+# Only supported on linux systems.
+ifeq ($(HOST_OS),linux)
+
+include $(CLEAR_VARS)
+LOCAL_MODULE := libbenchmark
+LOCAL_CFLAGS := $(benchmark_cflags)
+LOCAL_CPPFLAGS := $(benchmark_cppflags)
+LOCAL_SRC_FILES := $(benchmarklib_src_files)
+LOCAL_C_INCLUDES := $(benchmark_c_includes)
+LOCAL_MULTILIB := both
+LOCAL_STATIC_LIBRARIES := libbase
+include $(BUILD_HOST_STATIC_LIBRARY)
+
+endif
+
 # -----------------------------------------------------------------------------
 # Benchmarks.
 # -----------------------------------------------------------------------------
-
-benchmark_c_flags = \
-    -O2 \
-    -Wall -Wextra -Wunused \
-    -Werror \
-    -fno-builtin \
-    -std=gnu++11 \
-
-benchmark_src_files = \
-    benchmark_main.cpp \
+benchmark_src_files := \
     math_benchmark.cpp \
+    property_benchmark.cpp \
     pthread_benchmark.cpp \
     semaphore_benchmark.cpp \
     stdio_benchmark.cpp \
@@ -47,10 +81,10 @@ 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_SRC_FILES := $(benchmark_src_files) property_benchmark.cpp
-LOCAL_CXX_STL := libc++
+LOCAL_CFLAGS := $(benchmark_cflags)
+LOCAL_CPPFLAGS := $(benchmark_cppflags)
+LOCAL_SRC_FILES := $(benchmark_src_files)
+LOCAL_STATIC_LIBRARIES := libbenchmark libbase
 include $(BUILD_EXECUTABLE)
 
 # We don't build a static benchmark executable because it's not usually
@@ -59,19 +93,24 @@ include $(BUILD_EXECUTABLE)
 # performance of the old release's libc, and a static benchmark isn't
 # going to let you do that.
 
+# Only supported on linux systems.
+ifeq ($(HOST_OS),linux)
+
 # Build benchmarks for the host (against glibc!). Run with:
 include $(CLEAR_VARS)
 LOCAL_MODULE := bionic-benchmarks-glibc
 LOCAL_MODULE_STEM_32 := bionic-benchmarks-glibc32
 LOCAL_MODULE_STEM_64 := bionic-benchmarks-glibc64
 LOCAL_MULTILIB := both
-LOCAL_ADDITIONAL_DEPENDENCIES := $(LOCAL_PATH)/Android.mk
-LOCAL_CFLAGS += $(benchmark_c_flags)
-LOCAL_LDFLAGS += -lrt
+LOCAL_CFLAGS := $(benchmark_cflags)
+LOCAL_CPPFLAGS := $(benchmark_cppflags)
+LOCAL_LDFLAGS := -lrt
 LOCAL_SRC_FILES := $(benchmark_src_files)
-LOCAL_CXX_STL := libc++
+LOCAL_STATIC_LIBRARIES := libbenchmark libbase
 include $(BUILD_HOST_EXECUTABLE)
 
+endif
+
 ifeq ($(HOST_OS)-$(HOST_ARCH),$(filter $(HOST_OS)-$(HOST_ARCH),linux-x86 linux-x86_64))
 include $(LOCAL_PATH)/../build/run-on-host.mk
 
@@ -90,5 +129,3 @@ bionic-benchmarks-run-on-host64: bionic-benchmarks bionic-prepare-run-on-host
 endif
 
 endif
-
-endif # !BUILD_TINY_ANDROID

benchmarks/Benchmark.cpp

@@ -0,0 +1,159 @@
+/*
+ * Copyright (C) 2012 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 <inttypes.h>
+#include <regex.h>
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <time.h>
+
+#include <string>
+#include <vector>
+
+#include <base/stringprintf.h>
+
+#include <benchmark/Benchmark.h>
+
+#include "utils.h"
+
+namespace testing {
+
+static uint64_t NanoTime() {
+  struct timespec t;
+  t.tv_sec = t.tv_nsec = 0;
+  clock_gettime(CLOCK_MONOTONIC, &t);
+  return static_cast<uint64_t>(t.tv_sec) * 1000000000LL + t.tv_nsec;
+}
+
+bool Benchmark::header_printed_;
+
+std::vector<Benchmark*>& Benchmark::List() {
+  static std::vector<Benchmark*> list;
+  return list;
+}
+
+int Benchmark::MaxNameColumnWidth() {
+  size_t max = 20;
+  for (auto& benchmark : List()) {
+    max = std::max(max, benchmark->NameColumnWidth());
+  }
+  return static_cast<int>(max);
+}
+
+size_t Benchmark::RunAll(std::vector<regex_t*>& regs) {
+  size_t benchmarks_run = 0;
+  header_printed_ = false;
+  for (auto& benchmark : List()) {
+    benchmarks_run += benchmark->RunAllArgs(regs);
+  }
+  return benchmarks_run;
+}
+
+void Benchmark::PrintHeader() {
+  if (!header_printed_) {
+    printf("%-*s %10s %10s\n", MaxNameColumnWidth(), "", "iterations", "ns/op");
+    header_printed_ = true;
+  }
+}
+
+template <typename T>
+bool BenchmarkT<T>::ShouldRun(std::vector<regex_t*>& regs, T arg) {
+  if (regs.empty()) {
+    return true;
+  }
+
+  for (const auto& re : regs) {
+    if (regexec(re, GetNameStr(arg).c_str(), 0, NULL, 0) != REG_NOMATCH) {
+      return true;
+    }
+  }
+  return false;
+}
+
+void Benchmark::StopBenchmarkTiming() {
+  if (start_time_ns_ != 0) {
+    total_time_ns_ += NanoTime() - start_time_ns_;
+  }
+  start_time_ns_ = 0;
+}
+
+void Benchmark::StartBenchmarkTiming() {
+  if (start_time_ns_ == 0) {
+    start_time_ns_ = NanoTime();
+  }
+}
+
+std::string BenchmarkWithoutArg::GetNameStr(void*) {
+  return Name();
+}
+
+template <>
+std::string BenchmarkWithArg<int>::GetNameStr(int arg) {
+  return Name() + "/" + PrettyInt(arg, 2);
+}
+
+template <>
+std::string BenchmarkWithArg<double>::GetNameStr(double arg) {
+  return Name() + "/" + android::base::StringPrintf("%0.6f", arg);
+}
+
+template<typename T>
+void BenchmarkT<T>::RunWithArg(T arg) {
+  int new_iterations = 1;
+  int iterations;
+  while (new_iterations < 1e8) {
+    bytes_processed_ = 0;
+    total_time_ns_ = 0;
+    start_time_ns_ = 0;
+
+    iterations = new_iterations;
+    RunIterations(iterations, arg);
+    if (total_time_ns_ >= 1e9) {
+      break;
+    }
+
+    if (total_time_ns_/iterations == 0) {
+      new_iterations = 1e9;
+    } else {
+      new_iterations = 1e9/ (total_time_ns_/iterations);
+    }
+    new_iterations = std::max(iterations + 1,
+                          std::min(new_iterations + new_iterations/2, 100*iterations));
+
+    new_iterations = Round(new_iterations);
+  }
+
+  printf("%-*s %10s %10" PRId64, MaxNameColumnWidth(), GetNameStr(arg).c_str(),
+         PrettyInt(iterations, 10).c_str(), total_time_ns_/iterations);
+
+  if (total_time_ns_ > 0 && bytes_processed_ > 0) {
+    double gib_processed = static_cast<double>(bytes_processed_)/1e9;
+    double seconds = static_cast<double>(total_time_ns_)/1e9;
+    printf(" %8.3f GiB/s", gib_processed/seconds);
+  }
+  printf("\n");
+  fflush(stdout);
+}
+
+template class BenchmarkT<int>;
+template class BenchmarkT<double>;
+template class BenchmarkT<void*>;
+
+template class BenchmarkWithArg<int>;
+template class BenchmarkWithArg<double>;
+
+}  // namespace testing

benchmarks/benchmark.h

@@ -1,61 +0,0 @@
-/*
- * Copyright (C) 2012 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 <stdint.h>
-
-#include <vector>
-
-namespace testing {
-
-class Benchmark {
- public:
-  Benchmark(const char* name, void (*fn)(int)) {
-    Register(name, fn, NULL);
-  }
-
-  Benchmark(const char* name, void (*fn_range)(int, int)) {
-    Register(name, NULL, fn_range);
-  }
-
-  Benchmark* Arg(int x);
-
-  const char* Name();
-
-  bool ShouldRun(int argc, char* argv[]);
-  void Run();
-
- private:
-  const char* name_;
-
-  void (*fn_)(int);
-  void (*fn_range_)(int, int);
-
-  std::vector<int> args_;
-
-  void Register(const char* name, void (*fn)(int), void (*fn_range)(int, int));
-  void RunRepeatedlyWithArg(int iterations, int arg);
-  void RunWithArg(int arg);
-};
-
-}  // namespace testing
-
-void SetBenchmarkBytesProcessed(int64_t);
-void StopBenchmarkTiming();
-void StartBenchmarkTiming();
-
-#define BENCHMARK(f) \
-    static ::testing::Benchmark* _benchmark_##f __attribute__((unused)) = \
-        (new ::testing::Benchmark(#f, f))

benchmarks/benchmark/Benchmark.h

@@ -0,0 +1,169 @@
+/*
+ * Copyright (C) 2012 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.
+ */
+
+#ifndef BENCHMARKS_BENCHMARK_H_
+#define BENCHMARKS_BENCHMARK_H_
+
+#include <regex.h>
+#include <stdint.h>
+
+#include <string>
+#include <vector>
+
+namespace testing {
+
+class Benchmark {
+public:
+  Benchmark() {
+    List().push_back(this);
+  }
+  virtual ~Benchmark() {}
+
+  virtual std::string Name() = 0;
+
+  virtual size_t RunAllArgs(std::vector<regex_t*>&) = 0;
+
+  void SetBenchmarkBytesProcessed(uint64_t bytes) { bytes_processed_ += bytes; }
+  void StopBenchmarkTiming();
+  void StartBenchmarkTiming();
+
+  // Run all of the benchmarks that have registered.
+  static size_t RunAll(std::vector<regex_t*>&);
+
+  static std::vector<Benchmark*>& List();
+
+  static int MaxNameColumnWidth();
+
+protected:
+  virtual size_t NameColumnWidth() = 0;
+
+  uint64_t bytes_processed_;
+  uint64_t total_time_ns_;
+  uint64_t start_time_ns_;
+
+  static bool header_printed_;
+
+  static void PrintHeader();
+};
+
+template <typename T>
+class BenchmarkT : public Benchmark {
+public:
+  BenchmarkT() {}
+  virtual ~BenchmarkT() {}
+
+protected:
+  bool ShouldRun(std::vector<regex_t*>&, T arg);
+  void RunWithArg(T arg);
+  virtual void RunIterations(int, T) = 0;
+  virtual std::string GetNameStr(T) = 0;
+};
+
+class BenchmarkWithoutArg : public BenchmarkT<void*> {
+public:
+  BenchmarkWithoutArg() {}
+  virtual ~BenchmarkWithoutArg() {}
+
+protected:
+  virtual size_t RunAllArgs(std::vector<regex_t*>& regs) override {
+    size_t benchmarks_run = 0;
+    if (BenchmarkT<void*>::ShouldRun(regs, nullptr)) {
+      PrintHeader();
+      RunWithArg(nullptr);
+      benchmarks_run++;
+    }
+    return benchmarks_run;
+  }
+
+  virtual void RunIterations(int iters, void*) override {
+    Run(iters);
+  }
+
+  virtual void Run(int) = 0;
+
+  virtual size_t NameColumnWidth() override {
+    return Name().size();
+  }
+
+  virtual std::string GetNameStr(void *) override;
+};
+
+template<typename T>
+class BenchmarkWithArg : public BenchmarkT<T> {
+public:
+  BenchmarkWithArg() {}
+  virtual ~BenchmarkWithArg() {}
+
+  BenchmarkWithArg* Arg(T arg) {
+    args_.push_back(arg);
+    return this;
+  }
+
+protected:
+  virtual size_t NameColumnWidth() override {
+    size_t max = 0;
+    for (const auto& arg : args_) {
+      max = std::max(max, GetNameStr(arg).size());
+    }
+    return max;
+  }
+
+  std::string GetNameStr(T arg) override;
+
+  virtual size_t RunAllArgs(std::vector<regex_t*>& regs) override {
+    size_t benchmarks_run = 0;
+    for (T& arg : args_) {
+      if (BenchmarkT<T>::ShouldRun(regs, arg)) {
+        Benchmark::PrintHeader();
+        BenchmarkT<T>::RunWithArg(arg);
+        benchmarks_run++;
+      }
+    }
+    return benchmarks_run;
+  }
+
+  virtual void RunIterations(int iters, T arg) override {
+    Run(iters, arg);
+  }
+
+  virtual void Run(int iters, T arg) = 0;
+
+private:
+  std::vector<T> args_;
+};
+
+}  // namespace testing
+
+#define BENCHMARK_START(f, super_class) \
+  class f : public super_class { \
+  public: \
+    f() {} \
+    virtual ~f() {} \
+    virtual std::string Name() override { return #f; } \
+
+#define BENCHMARK_NO_ARG(f) \
+  BENCHMARK_START(f, ::testing::BenchmarkWithoutArg) \
+    virtual void Run(int) override; \
+  }; \
+  static ::testing::Benchmark* __benchmark_##f = new f()
+
+#define BENCHMARK_WITH_ARG(f, arg_type) \
+  BENCHMARK_START(f, ::testing::BenchmarkWithArg<arg_type>) \
+    virtual void Run(int, arg_type) override; \
+  }; \
+  static ::testing::BenchmarkWithArg<arg_type>* __benchmark_##f = (new f())
+
+#endif  // BENCHMARKS_BENCHMARK_H_

benchmarks/benchmark_main.cpp

@@ -1,227 +0,0 @@
-/*
- * Copyright (C) 2012 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 <regex.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <time.h>
-
-#include <string>
-#include <map>
-
-#include <inttypes.h>
-
-static int64_t g_bytes_processed;
-static int64_t g_benchmark_total_time_ns;
-static int64_t g_benchmark_start_time_ns;
-
-typedef std::map<std::string, ::testing::Benchmark*> BenchmarkMap;
-typedef BenchmarkMap::iterator BenchmarkMapIt;
-static BenchmarkMap g_benchmarks;
-static int g_name_column_width = 20;
-
-static int Round(int n) {
-  int base = 1;
-  while (base*10 < n) {
-    base *= 10;
-  }
-  if (n < 2*base) {
-    return 2*base;
-  }
-  if (n < 5*base) {
-    return 5*base;
-  }
-  return 10*base;
-}
-
-static int64_t NanoTime() {
-  struct timespec t;
-  t.tv_sec = t.tv_nsec = 0;
-  clock_gettime(CLOCK_MONOTONIC, &t);
-  return static_cast<int64_t>(t.tv_sec) * 1000000000LL + t.tv_nsec;
-}
-
-namespace testing {
-
-Benchmark* Benchmark::Arg(int arg) {
-  args_.push_back(arg);
-  return this;
-}
-
-const char* Benchmark::Name() {
-  return name_;
-}
-
-bool Benchmark::ShouldRun(int argc, char* argv[]) {
-  if (argc == 1) {
-    return true;  // With no arguments, we run all benchmarks.
-  }
-  // Otherwise, we interpret each argument as a regular expression and
-  // see if any of our benchmarks match.
-  for (int i = 1; i < argc; i++) {
-    regex_t re;
-    if (regcomp(&re, argv[i], 0) != 0) {
-      fprintf(stderr, "couldn't compile \"%s\" as a regular expression!\n", argv[i]);
-      exit(EXIT_FAILURE);
-    }
-    int match = regexec(&re, name_, 0, NULL, 0);
-    regfree(&re);
-    if (match != REG_NOMATCH) {
-      return true;
-    }
-  }
-  return false;
-}
-
-void Benchmark::Register(const char* name, void (*fn)(int), void (*fn_range)(int, int)) {
-  name_ = name;
-  fn_ = fn;
-  fn_range_ = fn_range;
-
-  if (fn_ == NULL && fn_range_ == NULL) {
-    fprintf(stderr, "%s: missing function\n", name_);
-    exit(EXIT_FAILURE);
-  }
-
-  g_benchmarks.insert(std::make_pair(name, this));
-}
-
-void Benchmark::Run() {
-  if (fn_ != NULL) {
-    RunWithArg(0);
-  } else {
-    if (args_.empty()) {
-      fprintf(stderr, "%s: no args!\n", name_);
-      exit(EXIT_FAILURE);
-    }
-    for (size_t i = 0; i < args_.size(); ++i) {
-      RunWithArg(args_[i]);
-    }
-  }
-}
-
-void Benchmark::RunRepeatedlyWithArg(int iterations, int arg) {
-  g_bytes_processed = 0;
-  g_benchmark_total_time_ns = 0;
-  g_benchmark_start_time_ns = NanoTime();
-  if (fn_ != NULL) {
-    fn_(iterations);
-  } else {
-    fn_range_(iterations, arg);
-  }
-  if (g_benchmark_start_time_ns != 0) {
-    g_benchmark_total_time_ns += NanoTime() - g_benchmark_start_time_ns;
-  }
-}
-
-void Benchmark::RunWithArg(int arg) {
-  // run once in case it's expensive
-  int iterations = 1;
-  RunRepeatedlyWithArg(iterations, arg);
-  while (g_benchmark_total_time_ns < 1e9 && iterations < 1e9) {
-    int last = iterations;
-    if (g_benchmark_total_time_ns/iterations == 0) {
-      iterations = 1e9;
-    } else {
-      iterations = 1e9 / (g_benchmark_total_time_ns/iterations);
-    }
-    iterations = std::max(last + 1, std::min(iterations + iterations/2, 100*last));
-    iterations = Round(iterations);
-    RunRepeatedlyWithArg(iterations, arg);
-  }
-
-  char throughput[100];
-  throughput[0] = '\0';
-  if (g_benchmark_total_time_ns > 0 && g_bytes_processed > 0) {
-    double mib_processed = static_cast<double>(g_bytes_processed)/1e6;
-    double seconds = static_cast<double>(g_benchmark_total_time_ns)/1e9;
-    snprintf(throughput, sizeof(throughput), " %8.2f MiB/s", mib_processed/seconds);
-  }
-
-  char full_name[100];
-  if (fn_range_ != NULL) {
-    if (arg >= (1<<20)) {
-      snprintf(full_name, sizeof(full_name), "%s/%dM", name_, arg/(1<<20));
-    } else if (arg >= (1<<10)) {
-      snprintf(full_name, sizeof(full_name), "%s/%dK", name_, arg/(1<<10));
-    } else {
-      snprintf(full_name, sizeof(full_name), "%s/%d", name_, arg);
-    }
-  } else {
-    snprintf(full_name, sizeof(full_name), "%s", name_);
-  }
-
-  printf("%-*s %10d %10" PRId64 "%s\n", g_name_column_width, full_name,
-         iterations, g_benchmark_total_time_ns/iterations, throughput);
-  fflush(stdout);
-}
-
-}  // namespace testing
-
-void SetBenchmarkBytesProcessed(int64_t x) {
-  g_bytes_processed = x;
-}
-
-void StopBenchmarkTiming() {
-  if (g_benchmark_start_time_ns != 0) {
-    g_benchmark_total_time_ns += NanoTime() - g_benchmark_start_time_ns;
-  }
-  g_benchmark_start_time_ns = 0;
-}
-
-void StartBenchmarkTiming() {
-  if (g_benchmark_start_time_ns == 0) {
-    g_benchmark_start_time_ns = NanoTime();
-  }
-}
-
-int main(int argc, char* argv[]) {
-  if (g_benchmarks.empty()) {
-    fprintf(stderr, "No benchmarks registered!\n");
-    exit(EXIT_FAILURE);
-  }
-
-  for (BenchmarkMapIt it = g_benchmarks.begin(); it != g_benchmarks.end(); ++it) {
-    int name_width = static_cast<int>(strlen(it->second->Name()));
-    g_name_column_width = std::max(g_name_column_width, name_width);
-  }
-
-  bool need_header = true;
-  for (BenchmarkMapIt it = g_benchmarks.begin(); it != g_benchmarks.end(); ++it) {
-    ::testing::Benchmark* b = it->second;
-    if (b->ShouldRun(argc, argv)) {
-      if (need_header) {
-        printf("%-*s %10s %10s\n", g_name_column_width, "", "iterations", "ns/op");
-        fflush(stdout);
-        need_header = false;
-      }
-      b->Run();
-    }
-  }
-
-  if (need_header) {
-    fprintf(stderr, "No matching benchmarks!\n");
-    fprintf(stderr, "Available benchmarks:\n");
-    for (BenchmarkMapIt it = g_benchmarks.begin(); it != g_benchmarks.end(); ++it) {
-      fprintf(stderr, "  %s\n", it->second->Name());
-    }
-    exit(EXIT_FAILURE);
-  }
-
-  return 0;
-}

benchmarks/main.cpp

@@ -0,0 +1,61 @@
+/*
+ * Copyright (C) 2012 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 <regex.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include <vector>
+
+#include <benchmark/Benchmark.h>
+
+int main(int argc, char* argv[]) {
+  if (::testing::Benchmark::List().empty()) {
+    fprintf(stderr, "No benchmarks registered!\n");
+    exit(EXIT_FAILURE);
+  }
+
+  std::vector<regex_t*> regs;
+  for (int i = 1; i < argc; i++) {
+    regex_t* re = new regex_t;
+    int errcode = regcomp(re, argv[i], 0);
+    if (errcode != 0) {
+      size_t errbuf_size = regerror(errcode, re, NULL, 0);
+      if (errbuf_size > 0) {
+        char* errbuf = new char[errbuf_size];
+        regerror(errcode, re, errbuf, errbuf_size);
+        fprintf(stderr, "Couldn't compile \"%s\" as a regular expression: %s\n",
+                argv[i], errbuf);
+      } else {
+        fprintf(stderr, "Unknown compile error for \"%s\" as a regular expression!\n", argv[i]);
+      }
+      exit(EXIT_FAILURE);
+    }
+    regs.push_back(re);
+  }
+
+  if (::testing::Benchmark::RunAll(regs) == 0) {
+    fprintf(stderr, "No matching benchmarks!\n");
+    fprintf(stderr, "Available benchmarks:\n");
+    for (const auto& benchmark : ::testing::Benchmark::List()) {
+      fprintf(stderr, "  %s\n", benchmark->Name().c_str());
+    }
+    exit(EXIT_FAILURE);
+  }
+
+  return 0;
+}

benchmarks/math_benchmark.cpp

@@ -14,16 +14,20 @@
  * limitations under the License.
  */
 
-#include "benchmark.h"
-
 #include <fenv.h>
 #include <math.h>
 
+#include <benchmark/Benchmark.h>
+
+#define AT_COMMON_VALS \
+    Arg(1234.0)->Arg(nan(""))->Arg(HUGE_VAL)->Arg(0.0)
+
 // Avoid optimization.
 volatile double d;
 volatile double v;
 
-static void BM_math_sqrt(int iters) {
+BENCHMARK_NO_ARG(BM_math_sqrt);
+void BM_math_sqrt::Run(int iters) {
   StartBenchmarkTiming();
 
   d = 0.0;
@@ -34,9 +38,9 @@ static void BM_math_sqrt(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_math_sqrt);
 
-static void BM_math_log10(int iters) {
+BENCHMARK_NO_ARG(BM_math_log10);
+void BM_math_log10::Run(int iters) {
   StartBenchmarkTiming();
 
   d = 0.0;
@@ -47,9 +51,9 @@ static void BM_math_log10(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_math_log10);
 
-static void BM_math_logb(int iters) {
+BENCHMARK_NO_ARG(BM_math_logb);
+void BM_math_logb::Run(int iters) {
   StartBenchmarkTiming();
 
   d = 0.0;
@@ -60,61 +64,22 @@ static void BM_math_logb(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_math_logb);
 
-static void BM_math_isinf_NORMAL(int iters) {
+BENCHMARK_WITH_ARG(BM_math_isinf, double)->AT_COMMON_VALS;
+void BM_math_isinf::Run(int iters, double value) {
   StartBenchmarkTiming();
 
   d = 0.0;
-  v = 1234.0; // FP_NORMAL
+  v = value;
   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_sin_fast(int iters) {
+BENCHMARK_NO_ARG(BM_math_sin_fast);
+void BM_math_sin_fast::Run(int iters) {
   StartBenchmarkTiming();
 
   d = 1.0;
@@ -124,9 +89,9 @@ static void BM_math_sin_fast(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_math_sin_fast);
 
-static void BM_math_sin_feupdateenv(int iters) {
+BENCHMARK_NO_ARG(BM_math_sin_feupdateenv);
+void BM_math_sin_feupdateenv::Run(int iters) {
   StartBenchmarkTiming();
 
   d = 1.0;
@@ -140,9 +105,9 @@ static void BM_math_sin_feupdateenv(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_math_sin_feupdateenv);
 
-static void BM_math_sin_fesetenv(int iters) {
+BENCHMARK_NO_ARG(BM_math_sin_fesetenv);
+void BM_math_sin_fesetenv::Run(int iters) {
   StartBenchmarkTiming();
 
   d = 1.0;
@@ -156,56 +121,16 @@ static void BM_math_sin_fesetenv(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_math_sin_fesetenv);
 
-static void BM_math_fpclassify_NORMAL(int iters) {
+BENCHMARK_WITH_ARG(BM_math_fpclassify, double)->AT_COMMON_VALS;
+void BM_math_fpclassify::Run(int iters, double value) {
   StartBenchmarkTiming();
 
   d = 0.0;
-  v = 1234.0; // FP_NORMAL
+  v = value;
   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

@@ -14,19 +14,21 @@
  * limitations under the License.
  */
 
-#include "benchmark.h"
 #include <errno.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <unistd.h>
 
+#include <string>
+
+#if defined(__BIONIC__)
+
 #define _REALLY_INCLUDE_SYS__SYSTEM_PROPERTIES_H_
 #include <sys/_system_properties.h>
 
-#include <vector>
-#include <string>
+#include <benchmark/Benchmark.h>
 
-extern void *__system_property_area__;
+extern void* __system_property_area__;
 
 // Do not exceed 512, that is about the largest number of properties
 // that can be created with the current property area size.
@@ -44,7 +46,7 @@ struct LocalPropertyTestState {
     }
     char dir_template[PATH_MAX];
     snprintf(dir_template, sizeof(dir_template), "%s/local/tmp/prop-XXXXXX", android_data);
-        char *dirname = mkdtemp(dir_template);
+    char* dirname = mkdtemp(dir_template);
     if (!dirname) {
       printf("making temp file for test state failed (is %s/local/tmp writable?): %s\n",
              android_data, strerror(errno));
@@ -126,20 +128,20 @@ struct LocalPropertyTestState {
   }
 public:
   const int nprops;
-    char **names;
-    int *name_lens;
-    char **values;
-    int *value_lens;
+  char** names;
+  int* name_lens;
+  char** values;
+  int* value_lens;
   bool valid;
 
 private:
   std::string pa_dirname;
   std::string pa_filename;
-    void *old_pa;
+  void* old_pa;
 };
 
-static void BM_property_get(int iters, int nprops)
-{
+BENCHMARK_WITH_ARG(BM_property_get, int)->TEST_NUM_PROPS;
+void BM_property_get::Run(int iters, int nprops) {
   StopBenchmarkTiming();
 
   LocalPropertyTestState pa(nprops);
@@ -157,10 +159,9 @@ static void BM_property_get(int iters, int nprops)
   }
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_property_get)->TEST_NUM_PROPS;
 
-static void BM_property_find(int iters, int nprops)
-{
+BENCHMARK_WITH_ARG(BM_property_find, int)->TEST_NUM_PROPS;
+void BM_property_find::Run(int iters, int nprops) {
   StopBenchmarkTiming();
 
   LocalPropertyTestState pa(nprops);
@@ -177,10 +178,9 @@ static void BM_property_find(int iters, int nprops)
   }
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_property_find)->TEST_NUM_PROPS;
 
-static void BM_property_read(int iters, int nprops)
-{
+BENCHMARK_WITH_ARG(BM_property_read, int)->TEST_NUM_PROPS;
+void BM_property_read::Run(int iters, int nprops) {
   StopBenchmarkTiming();
 
   LocalPropertyTestState pa(nprops);
@@ -204,10 +204,9 @@ static void BM_property_read(int iters, int nprops)
 
   delete[] pinfo;
 }
-BENCHMARK(BM_property_read)->TEST_NUM_PROPS;
 
-static void BM_property_serial(int iters, int nprops)
-{
+BENCHMARK_WITH_ARG(BM_property_serial, int)->TEST_NUM_PROPS;
+void BM_property_serial::Run(int iters, int nprops) {
   StopBenchmarkTiming();
 
   LocalPropertyTestState pa(nprops);
@@ -230,4 +229,5 @@ static void BM_property_serial(int iters, int nprops)
 
   delete[] pinfo;
 }
-BENCHMARK(BM_property_serial)->TEST_NUM_PROPS;
+
+#endif  // __BIONIC__

benchmarks/pthread_benchmark.cpp

@@ -14,14 +14,15 @@
  * limitations under the License.
  */
 
-#include "benchmark.h"
-
 #include <pthread.h>
 
+#include <benchmark/Benchmark.h>
+
 // Stop GCC optimizing out our pure function.
 /* Must not be static! */ pthread_t (*pthread_self_fp)() = pthread_self;
 
-static void BM_pthread_self(int iters) {
+BENCHMARK_NO_ARG(BM_pthread_self);
+void BM_pthread_self::Run(int iters) {
   StartBenchmarkTiming();
 
   for (int i = 0; i < iters; ++i) {
@@ -30,9 +31,9 @@ static void BM_pthread_self(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_pthread_self);
 
-static void BM_pthread_getspecific(int iters) {
+BENCHMARK_NO_ARG(BM_pthread_getspecific);
+void BM_pthread_getspecific::Run(int iters) {
   StopBenchmarkTiming();
   pthread_key_t key;
   pthread_key_create(&key, NULL);
@@ -45,12 +46,27 @@ static void BM_pthread_getspecific(int iters) {
   StopBenchmarkTiming();
   pthread_key_delete(key);
 }
-BENCHMARK(BM_pthread_getspecific);
+
+BENCHMARK_NO_ARG(BM_pthread_setspecific);
+void BM_pthread_setspecific::Run(int iters) {
+  StopBenchmarkTiming();
+  pthread_key_t key;
+  pthread_key_create(&key, NULL);
+  StartBenchmarkTiming();
+
+  for (int i = 0; i < iters; ++i) {
+    pthread_setspecific(key, NULL);
+  }
+
+  StopBenchmarkTiming();
+  pthread_key_delete(key);
+}
 
 static void DummyPthreadOnceInitFunction() {
 }
 
-static void BM_pthread_once(int iters) {
+BENCHMARK_NO_ARG(BM_pthread_once);
+void BM_pthread_once::Run(int iters) {
   StopBenchmarkTiming();
   pthread_once_t once = PTHREAD_ONCE_INIT;
   pthread_once(&once, DummyPthreadOnceInitFunction);
@@ -62,9 +78,9 @@ static void BM_pthread_once(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_pthread_once);
 
-static void BM_pthread_mutex_lock(int iters) {
+BENCHMARK_NO_ARG(BM_pthread_mutex_lock);
+void BM_pthread_mutex_lock::Run(int iters) {
   StopBenchmarkTiming();
   pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
   StartBenchmarkTiming();
@@ -76,9 +92,9 @@ static void BM_pthread_mutex_lock(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_pthread_mutex_lock);
 
-static void BM_pthread_mutex_lock_ERRORCHECK(int iters) {
+BENCHMARK_NO_ARG(BM_pthread_mutex_lock_ERRORCHECK);
+void BM_pthread_mutex_lock_ERRORCHECK::Run(int iters) {
   StopBenchmarkTiming();
   pthread_mutex_t mutex = PTHREAD_ERRORCHECK_MUTEX_INITIALIZER_NP;
   StartBenchmarkTiming();
@@ -90,9 +106,9 @@ static void BM_pthread_mutex_lock_ERRORCHECK(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_pthread_mutex_lock_ERRORCHECK);
 
-static void BM_pthread_mutex_lock_RECURSIVE(int iters) {
+BENCHMARK_NO_ARG(BM_pthread_mutex_lock_RECURSIVE);
+void BM_pthread_mutex_lock_RECURSIVE::Run(int iters) {
   StopBenchmarkTiming();
   pthread_mutex_t mutex = PTHREAD_RECURSIVE_MUTEX_INITIALIZER_NP;
   StartBenchmarkTiming();
@@ -104,9 +120,9 @@ static void BM_pthread_mutex_lock_RECURSIVE(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_pthread_mutex_lock_RECURSIVE);
 
-static void BM_pthread_rw_lock_read(int iters) {
+BENCHMARK_NO_ARG(BM_pthread_rwlock_read);
+void BM_pthread_rwlock_read::Run(int iters) {
   StopBenchmarkTiming();
   pthread_rwlock_t lock;
   pthread_rwlock_init(&lock, NULL);
@@ -120,9 +136,9 @@ static void BM_pthread_rw_lock_read(int iters) {
   StopBenchmarkTiming();
   pthread_rwlock_destroy(&lock);
 }
-BENCHMARK(BM_pthread_rw_lock_read);
 
-static void BM_pthread_rw_lock_write(int iters) {
+BENCHMARK_NO_ARG(BM_pthread_rwlock_write);
+void BM_pthread_rwlock_write::Run(int iters) {
   StopBenchmarkTiming();
   pthread_rwlock_t lock;
   pthread_rwlock_init(&lock, NULL);
@@ -136,4 +152,82 @@ static void BM_pthread_rw_lock_write(int iters) {
   StopBenchmarkTiming();
   pthread_rwlock_destroy(&lock);
 }
-BENCHMARK(BM_pthread_rw_lock_write);
+
+static void* IdleThread(void*) {
+  return NULL;
+}
+
+BENCHMARK_NO_ARG(BM_pthread_create);
+void BM_pthread_create::Run(int iters) {
+  StopBenchmarkTiming();
+  pthread_t thread;
+
+  for (int i = 0; i < iters; ++i) {
+    StartBenchmarkTiming();
+    pthread_create(&thread, NULL, IdleThread, NULL);
+    StopBenchmarkTiming();
+    pthread_join(thread, NULL);
+  }
+}
+
+static void* RunThread(void* arg) {
+  ::testing::Benchmark* benchmark = reinterpret_cast<::testing::Benchmark*>(arg);
+  benchmark->StopBenchmarkTiming();
+  return NULL;
+}
+
+BENCHMARK_NO_ARG(BM_pthread_create_and_run);
+void BM_pthread_create_and_run::Run(int iters) {
+  StopBenchmarkTiming();
+  pthread_t thread;
+
+  for (int i = 0; i < iters; ++i) {
+    StartBenchmarkTiming();
+    pthread_create(&thread, NULL, RunThread, this);
+    pthread_join(thread, NULL);
+  }
+}
+
+static void* ExitThread(void* arg) {
+  ::testing::Benchmark* benchmark = reinterpret_cast<::testing::Benchmark*>(arg);
+  benchmark->StartBenchmarkTiming();
+  pthread_exit(NULL);
+}
+
+BENCHMARK_NO_ARG(BM_pthread_exit_and_join);
+void BM_pthread_exit_and_join::Run(int iters) {
+  StopBenchmarkTiming();
+  pthread_t thread;
+
+  for (int i = 0; i < iters; ++i) {
+    pthread_create(&thread, NULL, ExitThread, this);
+    pthread_join(thread, NULL);
+    StopBenchmarkTiming();
+  }
+}
+
+BENCHMARK_NO_ARG(BM_pthread_key_create);
+void BM_pthread_key_create::Run(int iters) {
+  StopBenchmarkTiming();
+  pthread_key_t key;
+
+  for (int i = 0; i < iters; ++i) {
+    StartBenchmarkTiming();
+    pthread_key_create(&key, NULL);
+    StopBenchmarkTiming();
+    pthread_key_delete(key);
+  }
+}
+
+BENCHMARK_NO_ARG(BM_pthread_key_delete);
+void BM_pthread_key_delete::Run(int iters) {
+  StopBenchmarkTiming();
+  pthread_key_t key;
+
+  for (int i = 0; i < iters; ++i) {
+    pthread_key_create(&key, NULL);
+    StartBenchmarkTiming();
+    pthread_key_delete(key);
+    StopBenchmarkTiming();
+  }
+}

benchmarks/semaphore_benchmark.cpp

@@ -14,11 +14,15 @@
  * limitations under the License.
  */
 
-#include "benchmark.h"
-
+#include <pthread.h>
 #include <semaphore.h>
+#include <stdatomic.h>
+#include <stdio.h>
 
-static void BM_semaphore_sem_getvalue(int iters) {
+#include <benchmark/Benchmark.h>
+
+BENCHMARK_NO_ARG(BM_semaphore_sem_getvalue);
+void BM_semaphore_sem_getvalue::Run(int iters) {
   StopBenchmarkTiming();
   sem_t semaphore;
   sem_init(&semaphore, 1, 1);
@@ -31,9 +35,9 @@ static void BM_semaphore_sem_getvalue(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_semaphore_sem_getvalue);
 
-static void BM_semaphore_sem_wait_sem_post(int iters) {
+BENCHMARK_NO_ARG(BM_semaphore_sem_wait_sem_post);
+void BM_semaphore_sem_wait_sem_post::Run(int iters) {
   StopBenchmarkTiming();
   sem_t semaphore;
   sem_init(&semaphore, 1, 1);
@@ -46,4 +50,94 @@ static void BM_semaphore_sem_wait_sem_post(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_semaphore_sem_wait_sem_post);
+
+/*
+ *    This test reports the overhead of the underlying futex wake syscall on
+ * the producer. It does not report the overhead from issuing the wake to the
+ * point where the posted consumer thread wakes up. It suffers from
+ * clock_gettime syscall overhead. Lock the CPU speed for consistent results
+ * as we may not reach >50% cpu utilization.
+ *
+ *    We will run a background thread that catches the sem_post wakeup and
+ * loops immediately returning back to sleep in sem_wait for the next one. This
+ * thread is run with policy SCHED_OTHER (normal policy), a middle policy.
+ *
+ *    The primary thread will run at SCHED_IDLE (lowest priority policy) when
+ * monitoring the background thread to detect when it hits sem_wait sleep. It
+ * will do so with no clock running. Once we are ready, we will switch to
+ * SCHED_FIFO (highest priority policy) to time the act of running sem_post
+ * with the benchmark clock running. This ensures nothing else in the system
+ * can preempt our timed activity, including the background thread. We are
+ * also protected with the scheduling policy of letting a process hit a
+ * resource limit rather than get hit with a context switch.
+ *
+ *    The background thread will start executing either on another CPU, or
+ * after we back down from SCHED_FIFO, but certainly not in the context of
+ * the timing of the sem_post.
+ */
+static atomic_int BM_semaphore_sem_post_running;
+
+static void *BM_semaphore_sem_post_start_thread(void *obj) {
+    sem_t *semaphore = reinterpret_cast<sem_t *>(obj);
+
+    while ((BM_semaphore_sem_post_running > 0) && !sem_wait(semaphore)) {
+        ;
+    }
+    BM_semaphore_sem_post_running = -1;
+    return NULL;
+}
+
+BENCHMARK_NO_ARG(BM_semaphore_sem_post);
+void BM_semaphore_sem_post::Run(int iters) {
+  StopBenchmarkTiming();
+
+  sem_t semaphore;
+  sem_init(&semaphore, 0, 0);
+
+  pthread_attr_t attr;
+  pthread_attr_init(&attr);
+  BM_semaphore_sem_post_running = 1;
+  struct sched_param param = { 0, };
+  pthread_attr_setschedparam(&attr, &param);
+  pthread_attr_setschedpolicy(&attr, SCHED_OTHER);
+  pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
+  pthread_t pthread;
+  pthread_create(&pthread, &attr, BM_semaphore_sem_post_start_thread, &semaphore);
+  pthread_attr_destroy(&attr);
+
+  sched_setscheduler((pid_t)0, SCHED_IDLE, &param);
+  for (int i = 0; i < iters; ++i) {
+    int trys = 3, dummy = 0;
+    do {
+      if (BM_semaphore_sem_post_running < 0) {
+        sched_setscheduler((pid_t)0, SCHED_OTHER, &param);
+        fprintf(stderr, "BM_semaphore_sem_post: start_thread died unexpectedly\n");
+        return;
+      }
+      sched_yield();
+      sem_getvalue(&semaphore, &dummy);
+      if (dummy < 0) {  // POSIX.1-2001 possibility 1
+        break;
+      }
+      if (dummy == 0) { // POSIX.1-2001 possibility 2
+        --trys;
+      }
+    } while (trys);
+    param.sched_priority = 1;
+    sched_setscheduler((pid_t)0, SCHED_FIFO, &param);
+    StartBenchmarkTiming();
+    sem_post(&semaphore);
+    StopBenchmarkTiming(); // Remember to subtract clock syscall overhead
+    param.sched_priority = 0;
+    sched_setscheduler((pid_t)0, SCHED_IDLE, &param);
+  }
+  sched_setscheduler((pid_t)0, SCHED_OTHER, &param);
+
+  if (BM_semaphore_sem_post_running > 0) {
+    BM_semaphore_sem_post_running = 0;
+  }
+  do {
+    sem_post(&semaphore);
+    sched_yield();
+  } while (!BM_semaphore_sem_post_running);
+}

benchmarks/stdio_benchmark.cpp

@@ -14,9 +14,10 @@
  * limitations under the License.
  */
 
-#include "benchmark.h"
-
 #include <stdio.h>
+#include <stdio_ext.h>
+
+#include <benchmark/Benchmark.h>
 
 #define KB 1024
 #define MB 1024*KB
@@ -26,11 +27,12 @@
     Arg(1*KB)->Arg(4*KB)->Arg(8*KB)->Arg(16*KB)->Arg(64*KB)
 
 template <typename Fn>
-static void ReadWriteTest(int iters, int chunk_size, Fn f, bool buffered) {
-  StopBenchmarkTiming();
+void ReadWriteTest(::testing::Benchmark* benchmark, int iters, int chunk_size, Fn f, bool buffered) {
+  benchmark->StopBenchmarkTiming();
   FILE* fp = fopen("/dev/zero", "rw");
+  __fsetlocking(fp, FSETLOCKING_BYCALLER);
   char* buf = new char[chunk_size];
-  StartBenchmarkTiming();
+  benchmark->StartBenchmarkTiming();
 
   if (!buffered) {
     setvbuf(fp, 0, _IONBF, 0);
@@ -40,28 +42,52 @@ static void ReadWriteTest(int iters, int chunk_size, Fn f, bool buffered) {
     f(buf, chunk_size, 1, fp);
   }
 
-  StopBenchmarkTiming();
-  SetBenchmarkBytesProcessed(int64_t(iters) * int64_t(chunk_size));
+  benchmark->StopBenchmarkTiming();
+  benchmark->SetBenchmarkBytesProcessed(int64_t(iters) * int64_t(chunk_size));
   delete[] buf;
   fclose(fp);
 }
 
-static void BM_stdio_fread(int iters, int chunk_size) {
-  ReadWriteTest(iters, chunk_size, fread, true);
+BENCHMARK_WITH_ARG(BM_stdio_fread, int)->AT_COMMON_SIZES;
+void BM_stdio_fread::Run(int iters, int chunk_size) {
+  ReadWriteTest(this, iters, chunk_size, fread, true);
 }
-BENCHMARK(BM_stdio_fread)->AT_COMMON_SIZES;
 
-static void BM_stdio_fwrite(int iters, int chunk_size) {
-  ReadWriteTest(iters, chunk_size, fwrite, true);
+BENCHMARK_WITH_ARG(BM_stdio_fwrite, int)->AT_COMMON_SIZES;
+void BM_stdio_fwrite::Run(int iters, int chunk_size) {
+  ReadWriteTest(this, iters, chunk_size, fwrite, true);
 }
-BENCHMARK(BM_stdio_fwrite)->AT_COMMON_SIZES;
 
-static void BM_stdio_fread_unbuffered(int iters, int chunk_size) {
-  ReadWriteTest(iters, chunk_size, fread, false);
+BENCHMARK_WITH_ARG(BM_stdio_fread_unbuffered, int)->AT_COMMON_SIZES;
+void BM_stdio_fread_unbuffered::Run(int iters, int chunk_size) {
+  ReadWriteTest(this, iters, chunk_size, fread, false);
 }
-BENCHMARK(BM_stdio_fread_unbuffered)->AT_COMMON_SIZES;
 
-static void BM_stdio_fwrite_unbuffered(int iters, int chunk_size) {
-  ReadWriteTest(iters, chunk_size, fwrite, false);
+BENCHMARK_WITH_ARG(BM_stdio_fwrite_unbuffered, int)->AT_COMMON_SIZES;
+void BM_stdio_fwrite_unbuffered::Run(int iters, int chunk_size) {
+  ReadWriteTest(this, iters, chunk_size, fwrite, false);
+}
+
+static void FopenFgetsFclose(int iters, bool no_locking) {
+  char buf[1024];
+  for (int i = 0; i < iters; ++i) {
+    FILE* fp = fopen("/proc/version", "re");
+    if (no_locking) __fsetlocking(fp, FSETLOCKING_BYCALLER);
+    fgets(buf, sizeof(buf), fp);
+    fclose(fp);
+  }
+}
+
+BENCHMARK_NO_ARG(BM_stdio_fopen_fgets_fclose_locking);
+void BM_stdio_fopen_fgets_fclose_locking::Run(int iters) {
+  StartBenchmarkTiming();
+  FopenFgetsFclose(iters, false);
+  StopBenchmarkTiming();
+}
+
+BENCHMARK_NO_ARG(BM_stdio_fopen_fgets_fclose_no_locking);
+void BM_stdio_fopen_fgets_fclose_no_locking::Run(int iters) {
+  StartBenchmarkTiming();
+  FopenFgetsFclose(iters, true);
+  StopBenchmarkTiming();
 }
-BENCHMARK(BM_stdio_fwrite_unbuffered)->AT_COMMON_SIZES;

benchmarks/string_benchmark.cpp

@@ -14,10 +14,11 @@
  * limitations under the License.
  */
 
-#include "benchmark.h"
-
+#include <stdint.h>
 #include <string.h>
 
+#include <benchmark/Benchmark.h>
+
 #define KB 1024
 #define MB 1024*KB
 
@@ -26,7 +27,8 @@
 
 // TODO: test unaligned operation too? (currently everything will be 8-byte aligned by malloc.)
 
-static void BM_string_memcmp(int iters, int nbytes) {
+BENCHMARK_WITH_ARG(BM_string_memcmp, int)->AT_COMMON_SIZES;
+void BM_string_memcmp::Run(int iters, int nbytes) {
   StopBenchmarkTiming();
   char* src = new char[nbytes]; char* dst = new char[nbytes];
   memset(src, 'x', nbytes);
@@ -39,13 +41,13 @@ static void BM_string_memcmp(int iters, int nbytes) {
   }
 
   StopBenchmarkTiming();
-  SetBenchmarkBytesProcessed(int64_t(iters) * int64_t(nbytes));
+  SetBenchmarkBytesProcessed(uint64_t(iters) * uint64_t(nbytes));
   delete[] src;
   delete[] dst;
 }
-BENCHMARK(BM_string_memcmp)->AT_COMMON_SIZES;
 
-static void BM_string_memcpy(int iters, int nbytes) {
+BENCHMARK_WITH_ARG(BM_string_memcpy, int)->AT_COMMON_SIZES;
+void BM_string_memcpy::Run(int iters, int nbytes) {
   StopBenchmarkTiming();
   char* src = new char[nbytes]; char* dst = new char[nbytes];
   memset(src, 'x', nbytes);
@@ -56,13 +58,13 @@ static void BM_string_memcpy(int iters, int nbytes) {
   }
 
   StopBenchmarkTiming();
-  SetBenchmarkBytesProcessed(int64_t(iters) * int64_t(nbytes));
+  SetBenchmarkBytesProcessed(uint64_t(iters) * uint64_t(nbytes));
   delete[] src;
   delete[] dst;
 }
-BENCHMARK(BM_string_memcpy)->AT_COMMON_SIZES;
 
-static void BM_string_memmove(int iters, int nbytes) {
+BENCHMARK_WITH_ARG(BM_string_memmove, int)->AT_COMMON_SIZES;
+void BM_string_memmove::Run(int iters, int nbytes) {
   StopBenchmarkTiming();
   char* buf = new char[nbytes + 64];
   memset(buf, 'x', nbytes + 64);
@@ -73,12 +75,12 @@ static void BM_string_memmove(int iters, int nbytes) {
   }
 
   StopBenchmarkTiming();
-  SetBenchmarkBytesProcessed(int64_t(iters) * int64_t(nbytes));
+  SetBenchmarkBytesProcessed(uint64_t(iters) * uint64_t(nbytes));
   delete[] buf;
 }
-BENCHMARK(BM_string_memmove)->AT_COMMON_SIZES;
 
-static void BM_string_memset(int iters, int nbytes) {
+BENCHMARK_WITH_ARG(BM_string_memset, int)->AT_COMMON_SIZES;
+void BM_string_memset::Run(int iters, int nbytes) {
   StopBenchmarkTiming();
   char* dst = new char[nbytes];
   StartBenchmarkTiming();
@@ -88,12 +90,12 @@ static void BM_string_memset(int iters, int nbytes) {
   }
 
   StopBenchmarkTiming();
-  SetBenchmarkBytesProcessed(int64_t(iters) * int64_t(nbytes));
+  SetBenchmarkBytesProcessed(uint64_t(iters) * uint64_t(nbytes));
   delete[] dst;
 }
-BENCHMARK(BM_string_memset)->AT_COMMON_SIZES;
 
-static void BM_string_strlen(int iters, int nbytes) {
+BENCHMARK_WITH_ARG(BM_string_strlen, int)->AT_COMMON_SIZES;
+void BM_string_strlen::Run(int iters, int nbytes) {
   StopBenchmarkTiming();
   char* s = new char[nbytes];
   memset(s, 'x', nbytes);
@@ -106,7 +108,6 @@ static void BM_string_strlen(int iters, int nbytes) {
   }
 
   StopBenchmarkTiming();
-  SetBenchmarkBytesProcessed(int64_t(iters) * int64_t(nbytes));
+  SetBenchmarkBytesProcessed(uint64_t(iters) * uint64_t(nbytes));
   delete[] s;
 }
-BENCHMARK(BM_string_strlen)->AT_COMMON_SIZES;

benchmarks/time_benchmark.cpp

@@ -14,14 +14,14 @@
  * limitations under the License.
  */
 
-#include "benchmark.h"
-
-#include <unistd.h>
 #include <sys/syscall.h>
 #include <sys/time.h>
 #include <time.h>
 
-static void BM_time_clock_gettime(int iters) {
+#include <benchmark/Benchmark.h>
+
+BENCHMARK_NO_ARG(BM_time_clock_gettime);
+void BM_time_clock_gettime::Run(int iters) {
   StartBenchmarkTiming();
 
   timespec t;
@@ -31,9 +31,9 @@ static void BM_time_clock_gettime(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_time_clock_gettime);
 
-static void BM_time_clock_gettime_syscall(int iters) {
+BENCHMARK_NO_ARG(BM_time_clock_gettime_syscall);
+void BM_time_clock_gettime_syscall::Run(int iters) {
   StartBenchmarkTiming();
 
   timespec t;
@@ -43,9 +43,9 @@ static void BM_time_clock_gettime_syscall(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_time_clock_gettime_syscall);
 
-static void BM_time_gettimeofday(int iters) {
+BENCHMARK_NO_ARG(BM_time_gettimeofday);
+void BM_time_gettimeofday::Run(int iters) {
   StartBenchmarkTiming();
 
   timeval tv;
@@ -55,9 +55,9 @@ static void BM_time_gettimeofday(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_time_gettimeofday);
 
-static void BM_time_gettimeofday_syscall(int iters) {
+BENCHMARK_NO_ARG(BM_time_gettimeofday_syscall);
+void BM_time_gettimeofday_syscall::Run(int iters) {
   StartBenchmarkTiming();
 
   timeval tv;
@@ -67,9 +67,9 @@ static void BM_time_gettimeofday_syscall(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_time_gettimeofday_syscall);
 
-static void BM_time_time(int iters) {
+BENCHMARK_NO_ARG(BM_time_time);
+void BM_time_time::Run(int iters) {
   StartBenchmarkTiming();
 
   for (int i = 0; i < iters; ++i) {
@@ -78,4 +78,3 @@ static void BM_time_time(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_time_time);

benchmarks/unistd_benchmark.cpp

@@ -14,12 +14,13 @@
  * limitations under the License.
  */
 
-#include "benchmark.h"
-
 #include <sys/syscall.h>
 #include <unistd.h>
 
-static void BM_unistd_getpid(int iters) {
+#include <benchmark/Benchmark.h>
+
+BENCHMARK_NO_ARG(BM_unistd_getpid);
+void BM_unistd_getpid::Run(int iters) {
   StartBenchmarkTiming();
 
   for (int i = 0; i < iters; ++i) {
@@ -28,9 +29,9 @@ static void BM_unistd_getpid(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_unistd_getpid);
 
-static void BM_unistd_getpid_syscall(int iters) {
+BENCHMARK_NO_ARG(BM_unistd_getpid_syscall);
+void BM_unistd_getpid_syscall::Run(int iters) {
   StartBenchmarkTiming();
 
   for (int i = 0; i < iters; ++i) {
@@ -39,14 +40,14 @@ static void BM_unistd_getpid_syscall(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_unistd_getpid_syscall);
 
 #if defined(__BIONIC__)
 
 // Stop GCC optimizing out our pure function.
 /* Must not be static! */ pid_t (*gettid_fp)() = gettid;
 
-static void BM_unistd_gettid(int iters) {
+BENCHMARK_NO_ARG(BM_unistd_gettid);
+void BM_unistd_gettid::Run(int iters) {
   StartBenchmarkTiming();
 
   for (int i = 0; i < iters; ++i) {
@@ -55,11 +56,11 @@ static void BM_unistd_gettid(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_unistd_gettid);
 
 #endif
 
-static void BM_unistd_gettid_syscall(int iters) {
+BENCHMARK_NO_ARG(BM_unistd_gettid_syscall);
+void BM_unistd_gettid_syscall::Run(int iters) {
   StartBenchmarkTiming();
 
   for (int i = 0; i < iters; ++i) {
@@ -68,4 +69,3 @@ static void BM_unistd_gettid_syscall(int iters) {
 
   StopBenchmarkTiming();
 }
-BENCHMARK(BM_unistd_gettid_syscall);

benchmarks/utils.cpp

@@ -0,0 +1,81 @@
+/*
+ * Copyright (C) 2012 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 <inttypes.h>
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+#include <string>
+
+#include "utils.h"
+
+int Round(int n) {
+  int base = 1;
+  while (base*10 < n) {
+    base *= 10;
+  }
+  if (n < 2*base) {
+    return 2*base;
+  }
+  if (n < 5*base) {
+    return 5*base;
+  }
+  return 10*base;
+}
+
+// Similar to the code in art, but supporting both binary and decimal prefixes.
+std::string PrettyInt(long value, size_t base) {
+  if (base != 2 && base != 10) abort();
+
+  uint64_t count = static_cast<uint64_t>(value);
+  bool negative_number = false;
+  if (value < 0) {
+    negative_number = true;
+    count = static_cast<uint64_t>(-value);
+  }
+
+  // The byte thresholds at which we display amounts. A count is displayed
+  // in unit U when kUnitThresholds[U] <= bytes < kUnitThresholds[U+1].
+  static const uint64_t kUnitThresholds2[] = {
+    1024*1024*1024 /* Gi */, 2*1024*1024 /* Mi */, 3*1024 /* Ki */, 0,
+  };
+  static const uint64_t kUnitThresholds10[] = {
+    1000*1000*1000 /* G */, 2*1000*1000 /* M */, 3*1000 /* k */, 0,
+  };
+  static const uint64_t kAmountPerUnit2[] = { 1024*1024*1024, 1024*1024, 1024, 1 };
+  static const uint64_t kAmountPerUnit10[] = { 1000*1000*1000, 1000*1000, 1000, 1 };
+  static const char* const kUnitStrings2[] = { "Gi", "Mi", "Ki", "" };
+  static const char* const kUnitStrings10[] = { "G", "M", "k", "" };
+
+  // Which set are we using?
+  const uint64_t* kUnitThresholds = ((base == 2) ? kUnitThresholds2 : kUnitThresholds10);
+  const uint64_t* kAmountPerUnit = ((base == 2) ? kAmountPerUnit2 : kAmountPerUnit10);
+  const char* const* kUnitStrings = ((base == 2) ? kUnitStrings2 : kUnitStrings10);
+
+  size_t i = 0;
+  for (; kUnitThresholds[i] != 0; ++i) {
+    if (count >= kUnitThresholds[i]) {
+      break;
+    }
+  }
+  char* s = NULL;
+  asprintf(&s, "%s%" PRId64 "%s", (negative_number ? "-" : ""),
+           count / kAmountPerUnit[i], kUnitStrings[i]);
+  std::string result(s);
+  free(s);
+  return result;
+}

benchmarks/utils.h

@@ -0,0 +1,26 @@
+/*
+ * Copyright (C) 2012 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.
+ */
+
+#ifndef BENCHMARKS_UTILS_H
+#define BENCHMARKS_UTILS_H
+
+#include <stddef.h>
+#include <string>
+
+int Round(int n);
+std::string PrettyInt(long value, size_t base);
+
+#endif  // BENCHMARKS_UTILS_H

libc/Android.mk

@@ -61,14 +61,24 @@ libc_common_src_files := \
     stdio/fread.c \
     stdio/snprintf.c\
     stdio/sprintf.c \
+    stdio/stdio.c \
     stdio/stdio_ext.cpp \
+    stdlib/atexit.c \
+    stdlib/exit.c \
 
 # Fortify implementations of libc functions.
 libc_common_src_files += \
     bionic/__FD_chk.cpp \
     bionic/__fgets_chk.cpp \
+    bionic/__memchr_chk.cpp \
     bionic/__memmove_chk.cpp \
+    bionic/__memrchr_chk.cpp \
+    bionic/__poll_chk.cpp \
+    bionic/__pread64_chk.cpp \
+    bionic/__pread_chk.cpp \
     bionic/__read_chk.cpp \
+    bionic/__readlink_chk.cpp \
+    bionic/__readlinkat_chk.cpp \
     bionic/__recvfrom_chk.cpp \
     bionic/__stpcpy_chk.cpp \
     bionic/__stpncpy_chk.cpp \
@@ -83,7 +93,7 @@ libc_common_src_files += \
     bionic/__vsnprintf_chk.cpp \
     bionic/__vsprintf_chk.cpp \
 
-libc_bionic_src_files := \
+libc_bionic_ndk_src_files := \
     bionic/abort.cpp \
     bionic/accept.cpp \
     bionic/accept4.cpp \
@@ -102,6 +112,7 @@ libc_bionic_src_files := \
     bionic/clock_getcpuclockid.cpp \
     bionic/clock_nanosleep.cpp \
     bionic/clone.cpp \
+    bionic/close.cpp \
     bionic/__cmsg_nxthdr.cpp \
     bionic/connect.cpp \
     bionic/ctype.cpp \
@@ -111,16 +122,20 @@ libc_bionic_src_files := \
     bionic/epoll_pwait.cpp \
     bionic/epoll_wait.cpp \
     bionic/__errno.cpp \
+    bionic/error.cpp \
     bionic/eventfd_read.cpp \
     bionic/eventfd_write.cpp \
+    bionic/faccessat.cpp \
+    bionic/fchmod.cpp \
+    bionic/fchmodat.cpp \
     bionic/ffs.cpp \
+    bionic/fgetxattr.cpp \
     bionic/flockfile.cpp \
-    bionic/fork.cpp \
     bionic/fpclassify.cpp \
+    bionic/fsetxattr.cpp \
+    bionic/ftruncate.cpp \
     bionic/futimens.cpp \
-    bionic/getauxval.cpp \
     bionic/getcwd.cpp \
-    bionic/getentropy_linux.c \
     bionic/gethostname.cpp \
     bionic/getpgrp.cpp \
     bionic/getpid.cpp \
@@ -141,6 +156,7 @@ libc_bionic_src_files := \
     bionic/mbrtoc16.cpp \
     bionic/mbrtoc32.cpp \
     bionic/mbstate.cpp \
+    bionic/mempcpy.cpp \
     bionic/mkdir.cpp \
     bionic/mkfifo.cpp \
     bionic/mknod.cpp \
@@ -155,27 +171,6 @@ libc_bionic_src_files := \
     bionic/posix_fallocate.cpp \
     bionic/posix_madvise.cpp \
     bionic/posix_timers.cpp \
-    bionic/pthread_atfork.cpp \
-    bionic/pthread_attr.cpp \
-    bionic/pthread_cond.cpp \
-    bionic/pthread_create.cpp \
-    bionic/pthread_detach.cpp \
-    bionic/pthread_equal.cpp \
-    bionic/pthread_exit.cpp \
-    bionic/pthread_getcpuclockid.cpp \
-    bionic/pthread_getschedparam.cpp \
-    bionic/pthread_gettid_np.cpp \
-    bionic/pthread_internals.cpp \
-    bionic/pthread_join.cpp \
-    bionic/pthread_key.cpp \
-    bionic/pthread_kill.cpp \
-    bionic/pthread_mutex.cpp \
-    bionic/pthread_once.cpp \
-    bionic/pthread_rwlock.cpp \
-    bionic/pthread_self.cpp \
-    bionic/pthread_setname_np.cpp \
-    bionic/pthread_setschedparam.cpp \
-    bionic/pthread_sigmask.cpp \
     bionic/ptrace.cpp \
     bionic/pty.cpp \
     bionic/raise.cpp \
@@ -218,7 +213,6 @@ libc_bionic_src_files := \
     bionic/strtold.cpp \
     bionic/stubs.cpp \
     bionic/symlink.cpp \
-    bionic/sysconf.cpp \
     bionic/sysinfo.cpp \
     bionic/syslog.cpp \
     bionic/sys_siglist.c \
@@ -231,10 +225,27 @@ libc_bionic_src_files := \
     bionic/umount.cpp \
     bionic/unlink.cpp \
     bionic/utimes.cpp \
-    bionic/vdso.cpp \
     bionic/wait.cpp \
     bionic/wchar.cpp \
     bionic/wctype.cpp \
+    bionic/wmempcpy.cpp \
+
+libc_bionic_src_files :=
+
+# The fork implementation depends on pthread data, so we can't include it in
+# libc_ndk.a.
+libc_bionic_src_files += bionic/fork.cpp
+
+# The data that backs getauxval is initialized in the libc init functions which
+# are invoked by the linker. If this file is included in libc_ndk.a, only one of
+# the copies of the global data will be initialized, resulting in nullptr
+# dereferences.
+libc_bionic_src_files += bionic/getauxval.cpp
+
+# These three require getauxval, which isn't available on older platforms.
+libc_bionic_src_files += bionic/getentropy_linux.c
+libc_bionic_src_files += bionic/sysconf.cpp
+libc_bionic_src_files += bionic/vdso.cpp
 
 libc_cxa_src_files := \
     bionic/__cxa_guard.cpp \
@@ -269,7 +280,6 @@ libc_upstream_freebsd_src_files := \
     upstream-freebsd/lib/libc/string/wcsspn.c \
     upstream-freebsd/lib/libc/string/wcstok.c \
     upstream-freebsd/lib/libc/string/wmemchr.c \
-    upstream-freebsd/lib/libc/string/wmemcpy.c \
     upstream-freebsd/lib/libc/string/wmemset.c \
 
 libc_upstream_netbsd_src_files := \
@@ -333,10 +343,13 @@ libc_upstream_openbsd_gdtoa_src_files_64 := \
     $(libc_upstream_openbsd_gdtoa_src_files) \
     upstream-openbsd/lib/libc/gdtoa/strtorQ.c \
 
+# These two depend on getentropy_linux.cpp, which isn't in libc_ndk.a.
 libc_upstream_openbsd_src_files := \
-    upstream-openbsd/lib/libc/compat-43/killpg.c \
     upstream-openbsd/lib/libc/crypt/arc4random.c \
     upstream-openbsd/lib/libc/crypt/arc4random_uniform.c \
+
+libc_upstream_openbsd_ndk_src_files := \
+    upstream-openbsd/lib/libc/compat-43/killpg.c \
     upstream-openbsd/lib/libc/gen/alarm.c \
     upstream-openbsd/lib/libc/gen/ctype_.c \
     upstream-openbsd/lib/libc/gen/daemon.c \
@@ -450,7 +463,6 @@ libc_upstream_openbsd_src_files := \
     upstream-openbsd/lib/libc/stdio/setbuffer.c \
     upstream-openbsd/lib/libc/stdio/setvbuf.c \
     upstream-openbsd/lib/libc/stdio/sscanf.c \
-    upstream-openbsd/lib/libc/stdio/stdio.c \
     upstream-openbsd/lib/libc/stdio/swprintf.c \
     upstream-openbsd/lib/libc/stdio/swscanf.c \
     upstream-openbsd/lib/libc/stdio/tempnam.c \
@@ -476,11 +488,9 @@ libc_upstream_openbsd_src_files := \
     upstream-openbsd/lib/libc/stdio/wprintf.c \
     upstream-openbsd/lib/libc/stdio/wscanf.c \
     upstream-openbsd/lib/libc/stdio/wsetup.c \
-    upstream-openbsd/lib/libc/stdlib/atexit.c \
     upstream-openbsd/lib/libc/stdlib/atoi.c \
     upstream-openbsd/lib/libc/stdlib/atol.c \
     upstream-openbsd/lib/libc/stdlib/atoll.c \
-    upstream-openbsd/lib/libc/stdlib/exit.c \
     upstream-openbsd/lib/libc/stdlib/getenv.c \
     upstream-openbsd/lib/libc/stdlib/insque.c \
     upstream-openbsd/lib/libc/stdlib/lsearch.c \
@@ -505,13 +515,59 @@ libc_upstream_openbsd_src_files := \
     upstream-openbsd/lib/libc/string/strspn.c \
     upstream-openbsd/lib/libc/string/strstr.c \
     upstream-openbsd/lib/libc/string/strtok.c \
+    upstream-openbsd/lib/libc/string/wmemcpy.c \
     upstream-openbsd/lib/libc/string/wcslcpy.c \
     upstream-openbsd/lib/libc/string/wcsstr.c \
     upstream-openbsd/lib/libc/string/wcswidth.c \
 
+libc_pthread_src_files := \
+    bionic/pthread_atfork.cpp \
+    bionic/pthread_attr.cpp \
+    bionic/pthread_cond.cpp \
+    bionic/pthread_create.cpp \
+    bionic/pthread_detach.cpp \
+    bionic/pthread_equal.cpp \
+    bionic/pthread_exit.cpp \
+    bionic/pthread_getcpuclockid.cpp \
+    bionic/pthread_getschedparam.cpp \
+    bionic/pthread_gettid_np.cpp \
+    bionic/pthread_internal.cpp \
+    bionic/pthread_join.cpp \
+    bionic/pthread_key.cpp \
+    bionic/pthread_kill.cpp \
+    bionic/pthread_mutex.cpp \
+    bionic/pthread_once.cpp \
+    bionic/pthread_rwlock.cpp \
+    bionic/pthread_self.cpp \
+    bionic/pthread_setname_np.cpp \
+    bionic/pthread_setschedparam.cpp \
+    bionic/pthread_sigmask.cpp \
+
+libc_thread_atexit_impl_src_files := \
+    bionic/__cxa_thread_atexit_impl.cpp \
+
 libc_arch_static_src_files := \
     bionic/dl_iterate_phdr_static.cpp \
 
+# Various kinds of cruft.
+# ========================================================
+libc_common_src_files += \
+    bionic/ndk_cruft.cpp \
+
+libc_bionic_ndk_src_files_32 += \
+    bionic/mmap.cpp \
+
+libc_common_src_files_32 += \
+    bionic/legacy_32_bit_support.cpp \
+    bionic/time64.c \
+
+libc_netbsd_src_files_32 += \
+    upstream-netbsd/common/lib/libc/hash/sha1/sha1.c \
+
+libc_openbsd_src_files_32 += \
+    upstream-openbsd/lib/libc/stdio/putw.c \
+
+
 # Define some common cflags
 # ========================================================
 libc_common_cflags := \
@@ -568,12 +624,13 @@ libc_common_c_includes += \
     $(LOCAL_PATH)/stdio   \
 
 # ========================================================
-# Add in the arch-specific flags.
+# Add in the arch or 32-bit specific flags
 # Must be called with $(eval).
 # $(1): the LOCAL_ variable name
 # $(2): the bionic variable name to pull in
 define patch-up-arch-specific-flags
 $(1)_$(TARGET_ARCH) += $($(2)_$(TARGET_ARCH))
+$(1)_32 += $($(2)_32)
 ifdef TARGET_2ND_ARCH
 $(1)_$(TARGET_2ND_ARCH) += $($(2)_$(TARGET_2ND_ARCH))
 endif
@@ -672,15 +729,16 @@ LOCAL_CFLAGS += \
     -DINET6 \
     -fvisibility=hidden \
     -Wno-unused-parameter \
-    -I$(LOCAL_PATH)/dns/include \
-    -I$(LOCAL_PATH)/private \
-    -I$(LOCAL_PATH)/upstream-netbsd/lib/libc/include \
-    -I$(LOCAL_PATH)/upstream-netbsd/android/include \
     -include netbsd-compat.h \
 
 LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
 LOCAL_CPPFLAGS := $(libc_common_cppflags)
-LOCAL_C_INCLUDES := $(libc_common_c_includes)
+LOCAL_C_INCLUDES := $(libc_common_c_includes) \
+    $(LOCAL_PATH)/dns/include \
+    $(LOCAL_PATH)/private \
+    $(LOCAL_PATH)/upstream-netbsd/lib/libc/include \
+    $(LOCAL_PATH)/upstream-netbsd/android/include \
+
 LOCAL_MODULE := libc_dns
 LOCAL_CLANG := $(use_clang)
 LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
@@ -706,13 +764,14 @@ LOCAL_SRC_FILES := $(libc_upstream_freebsd_src_files)
 LOCAL_CFLAGS := \
     $(libc_common_cflags) \
     -Wno-sign-compare -Wno-uninitialized \
-    -I$(LOCAL_PATH)/upstream-freebsd/android/include \
-    -I$(LOCAL_PATH)/upstream-freebsd/lib/libc/include \
     -include freebsd-compat.h \
 
 LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
 LOCAL_CPPFLAGS := $(libc_common_cppflags)
-LOCAL_C_INCLUDES := $(libc_common_c_includes)
+LOCAL_C_INCLUDES := $(libc_common_c_includes) \
+    $(LOCAL_PATH)/upstream-freebsd/android/include \
+    $(LOCAL_PATH)/upstream-freebsd/lib/libc/include \
+
 LOCAL_MODULE := libc_freebsd
 LOCAL_CLANG := $(use_clang)
 LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
@@ -740,13 +799,14 @@ LOCAL_CFLAGS := \
     $(libc_common_cflags) \
     -Wno-sign-compare -Wno-uninitialized \
     -DPOSIX_MISTAKE \
-    -I$(LOCAL_PATH)/upstream-netbsd/android/include \
-    -I$(LOCAL_PATH)/upstream-netbsd/lib/libc/include \
     -include netbsd-compat.h \
 
 LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
 LOCAL_CPPFLAGS := $(libc_common_cppflags)
-LOCAL_C_INCLUDES := $(libc_common_c_includes)
+LOCAL_C_INCLUDES := $(libc_common_c_includes) \
+    $(LOCAL_PATH)/upstream-netbsd/android/include \
+    $(LOCAL_PATH)/upstream-netbsd/lib/libc/include \
+
 LOCAL_MODULE := libc_netbsd
 LOCAL_CLANG := $(use_clang)
 LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
@@ -760,6 +820,46 @@ $(eval $(call patch-up-arch-specific-flags,LOCAL_SRC_FILES,libc_netbsd_src_files
 include $(BUILD_STATIC_LIBRARY)
 
 
+# ========================================================
+# libc_openbsd_ndk.a - upstream OpenBSD C library code
+# that can be safely included in the libc_ndk.a (doesn't
+# contain any troublesome global data or constructors).
+# ========================================================
+#
+# These files are built with the openbsd-compat.h header file
+# automatically included.
+
+include $(CLEAR_VARS)
+
+LOCAL_SRC_FILES := $(libc_upstream_openbsd_ndk_src_files)
+LOCAL_CLANG := $(use_clang)
+
+LOCAL_CFLAGS := \
+    $(libc_common_cflags) \
+    -Wno-sign-compare \
+    -Wno-uninitialized \
+    -Wno-unused-parameter \
+    -include openbsd-compat.h \
+
+LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
+LOCAL_CPPFLAGS := $(libc_common_cppflags)
+LOCAL_C_INCLUDES := $(libc_common_c_includes) \
+    $(LOCAL_PATH)/private \
+    $(LOCAL_PATH)/upstream-openbsd/android/include \
+    $(LOCAL_PATH)/upstream-openbsd/lib/libc/include \
+    $(LOCAL_PATH)/upstream-openbsd/lib/libc/gdtoa/ \
+
+LOCAL_MODULE := libc_openbsd_ndk
+LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
+LOCAL_CXX_STL := none
+LOCAL_SYSTEM_SHARED_LIBRARIES :=
+LOCAL_ADDRESS_SANITIZER := false
+LOCAL_NATIVE_COVERAGE := $(bionic_coverage)
+
+$(eval $(call patch-up-arch-specific-flags,LOCAL_CFLAGS,libc_common_cflags))
+include $(BUILD_STATIC_LIBRARY)
+
+
 # ========================================================
 # libc_openbsd.a - upstream OpenBSD C library code
 # ========================================================
@@ -770,27 +870,23 @@ include $(BUILD_STATIC_LIBRARY)
 include $(CLEAR_VARS)
 
 LOCAL_SRC_FILES := $(libc_upstream_openbsd_src_files)
-ifneq (,$(filter $(TARGET_ARCH),x86 x86_64))
-  # Clang has wrong long double size or LDBL_MANT_DIG, http://b/17163651.
-  LOCAL_CLANG := false
-else
-  LOCAL_CLANG := $(use_clang)
-endif
+LOCAL_CLANG := $(use_clang)
 
 LOCAL_CFLAGS := \
     $(libc_common_cflags) \
     -Wno-sign-compare \
     -Wno-uninitialized \
     -Wno-unused-parameter \
-    -I$(LOCAL_PATH)/private \
-    -I$(LOCAL_PATH)/upstream-openbsd/android/include \
-    -I$(LOCAL_PATH)/upstream-openbsd/lib/libc/include \
-    -I$(LOCAL_PATH)/upstream-openbsd/lib/libc/gdtoa/ \
     -include openbsd-compat.h \
 
 LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
 LOCAL_CPPFLAGS := $(libc_common_cppflags)
-LOCAL_C_INCLUDES := $(libc_common_c_includes)
+LOCAL_C_INCLUDES := $(libc_common_c_includes) \
+    $(LOCAL_PATH)/private \
+    $(LOCAL_PATH)/upstream-openbsd/android/include \
+    $(LOCAL_PATH)/upstream-openbsd/lib/libc/include \
+    $(LOCAL_PATH)/upstream-openbsd/lib/libc/gdtoa/ \
+
 LOCAL_MODULE := libc_openbsd
 LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
 LOCAL_CXX_STL := none
@@ -814,25 +910,21 @@ include $(CLEAR_VARS)
 
 LOCAL_SRC_FILES_32 := $(libc_upstream_openbsd_gdtoa_src_files_32)
 LOCAL_SRC_FILES_64 := $(libc_upstream_openbsd_gdtoa_src_files_64)
-ifneq (,$(filter $(TARGET_ARCH),x86 x86_64))
-  # Clang has wrong long double size or LDBL_MANT_DIG, http://b/17163651.
-  LOCAL_CLANG := false
-else
-  LOCAL_CLANG := $(use_clang)
-endif
+LOCAL_CLANG := $(use_clang)
 
 LOCAL_CFLAGS := \
     $(libc_common_cflags) \
     -Wno-sign-compare -Wno-uninitialized \
     -fvisibility=hidden \
-    -I$(LOCAL_PATH)/private \
-    -I$(LOCAL_PATH)/upstream-openbsd/android/include \
-    -I$(LOCAL_PATH)/upstream-openbsd/lib/libc/include \
     -include openbsd-compat.h \
 
 LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
 LOCAL_CPPFLAGS := $(libc_common_cppflags)
-LOCAL_C_INCLUDES := $(libc_common_c_includes)
+LOCAL_C_INCLUDES := $(libc_common_c_includes) \
+    $(LOCAL_PATH)/private \
+    $(LOCAL_PATH)/upstream-openbsd/android/include \
+    $(LOCAL_PATH)/upstream-openbsd/lib/libc/include \
+
 LOCAL_MODULE := libc_gdtoa
 LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
 LOCAL_CXX_STL := none
@@ -854,15 +946,8 @@ LOCAL_SRC_FILES := $(libc_bionic_src_files)
 LOCAL_CFLAGS := $(libc_common_cflags) \
     -Wframe-larger-than=2048 \
 
-# ssse3-strcmp-slm.S does not compile with Clang.
-LOCAL_CLANG_ASFLAGS_x86_64 += -no-integrated-as
-
-# memcpy.S, memchr.S, etc. do not compile with Clang.
-LOCAL_CLANG_ASFLAGS_arm += -no-integrated-as
-LOCAL_CLANG_ASFLAGS_arm64 += -no-integrated-as
-
 LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
-LOCAL_CPPFLAGS := $(libc_common_cppflags)
+LOCAL_CPPFLAGS := $(libc_common_cppflags) -Wold-style-cast
 LOCAL_C_INCLUDES := $(libc_common_c_includes) bionic/libstdc++/include
 LOCAL_MODULE := libc_bionic
 LOCAL_CLANG := $(use_clang)
@@ -877,6 +962,80 @@ $(eval $(call patch-up-arch-specific-flags,LOCAL_SRC_FILES,libc_bionic_src_files
 include $(BUILD_STATIC_LIBRARY)
 
 
+# ========================================================
+# libc_bionic_ndk.a - The portions of libc_bionic that can
+# be safely used in libc_ndk.a (no troublesome global data
+# or constructors).
+# ========================================================
+
+include $(CLEAR_VARS)
+
+LOCAL_SRC_FILES := $(libc_bionic_ndk_src_files)
+LOCAL_CFLAGS := $(libc_common_cflags) \
+    -Wframe-larger-than=2048 \
+
+LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
+LOCAL_CPPFLAGS := $(libc_common_cppflags) -Wold-style-cast
+LOCAL_C_INCLUDES := $(libc_common_c_includes) bionic/libstdc++/include
+LOCAL_MODULE := libc_bionic_ndk
+LOCAL_CLANG := $(use_clang)
+LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
+LOCAL_CXX_STL := none
+LOCAL_SYSTEM_SHARED_LIBRARIES :=
+LOCAL_ADDRESS_SANITIZER := false
+LOCAL_NATIVE_COVERAGE := $(bionic_coverage)
+
+$(eval $(call patch-up-arch-specific-flags,LOCAL_CFLAGS,libc_common_cflags))
+$(eval $(call patch-up-arch-specific-flags,LOCAL_SRC_FILES,libc_bionic_ndk_src_files))
+include $(BUILD_STATIC_LIBRARY)
+
+include $(CLEAR_VARS)
+LOCAL_SRC_FILES := $(libc_thread_atexit_impl_src_files)
+LOCAL_CFLAGS := $(libc_common_cflags) -Wframe-larger-than=2048
+
+LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
+LOCAL_CPPFLAGS := $(libc_common_cppflags) -Wold-style-cast
+LOCAL_C_INCLUDES := $(libc_common_c_includes)
+LOCAL_MODULE := libc_thread_atexit_impl
+# TODO: Clang tries to use __tls_get_addr which is not supported yet
+# remove after it is implemented.
+LOCAL_CLANG := false
+LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
+LOCAL_CXX_STL := none
+LOCAL_SYSTEM_SHARED_LIBRARIES :=
+LOCAL_ADDRESS_SANITIZER := false
+LOCAL_NATIVE_COVERAGE := $(bionic_coverage)
+
+include $(BUILD_STATIC_LIBRARY)
+
+# ========================================================
+# libc_pthread.a - pthreads parts that previously lived in
+# libc_bionic.a. Relocated to their own library because
+# they can't be included in libc_ndk.a (as they layout of
+# pthread_t has changed over the years and has ABI
+# compatibility issues).
+# ========================================================
+
+include $(CLEAR_VARS)
+
+LOCAL_SRC_FILES := $(libc_pthread_src_files)
+LOCAL_CFLAGS := $(libc_common_cflags) \
+    -Wframe-larger-than=2048 \
+
+LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
+LOCAL_CPPFLAGS := $(libc_common_cppflags) -Wold-style-cast
+LOCAL_C_INCLUDES := $(libc_common_c_includes)
+LOCAL_MODULE := libc_pthread
+LOCAL_CLANG := $(use_clang)
+LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
+LOCAL_CXX_STL := none
+LOCAL_SYSTEM_SHARED_LIBRARIES :=
+LOCAL_ADDRESS_SANITIZER := false
+LOCAL_NATIVE_COVERAGE := $(bionic_coverage)
+
+include $(BUILD_STATIC_LIBRARY)
+
+
 # ========================================================
 # libc_cxa.a - Things traditionally in libstdc++
 # ========================================================
@@ -896,7 +1055,14 @@ LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
 LOCAL_CXX_STL := none
 LOCAL_SYSTEM_SHARED_LIBRARIES :=
 LOCAL_ADDRESS_SANITIZER := false
-LOCAL_NATIVE_COVERAGE := $(bionic_coverage)
+# b/17574078: Need to disable coverage until we have a prebuilt libprofile_rt.
+# Since this is a static library built with clang, it needs to link
+# libprofile_rt when it is linked into the final binary. Since the final binary
+# is built with GCC, it won't link libprofile_rt. We can't very easily just add
+# libprofile_rt to all link lines the way we've done for libgcov because
+# libprofile_rt isn't prebuilt, and it would be tricky to write a rule that
+# would make sure libprofile_rt is built.
+LOCAL_NATIVE_COVERAGE := false
 
 include $(BUILD_STATIC_LIBRARY)
 
@@ -943,6 +1109,73 @@ LOCAL_NATIVE_COVERAGE := $(bionic_coverage)
 
 include $(BUILD_STATIC_LIBRARY)
 
+# ========================================================
+# libc_ndk.a
+# Compatibility library for the NDK. This library contains
+# all the parts of libc that are safe to statically link.
+# We can't safely statically link things that can only run
+# on a certain version of the OS. Examples include
+# anything that talks to netd (a large portion of the DNS
+# code) and anything that is dependent on the layout of a
+# data structure that has changed across releases (such as
+# pthread_t).
+# ========================================================
+
+include $(CLEAR_VARS)
+
+LOCAL_MODULE := libc_ndk
+LOCAL_CLANG := $(use_clang)
+LOCAL_ASFLAGS := $(LOCAL_CFLAGS)
+LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
+LOCAL_CFLAGS := $(libc_common_cflags) -fvisibility=hidden -O0
+LOCAL_CPPFLAGS := $(libc_common_cppflags)
+LOCAL_C_INCLUDES := $(libc_common_c_includes)
+LOCAL_ADDRESS_SANITIZER := false
+LOCAL_NATIVE_COVERAGE := $(bionic_coverage)
+LOCAL_SYSTEM_SHARED_LIBRARIES :=
+
+LOCAL_SRC_FILES := \
+    $(libc_common_src_files) \
+    $(libc_arch_dynamic_src_files) \
+    $(libc_ndk_stub_src_files) \
+    bionic/malloc_debug_common.cpp \
+
+LOCAL_SRC_FILES_arm += \
+    arch-common/bionic/crtbegin_so.c \
+    arch-arm/bionic/atexit_legacy.c \
+    arch-common/bionic/crtend_so.S \
+
+LOCAL_CFLAGS := $(libc_common_cflags) \
+    -DLIBC_STATIC \
+
+LOCAL_WHOLE_STATIC_LIBRARIES := \
+    libc_bionic_ndk \
+    libc_cxa \
+    libc_freebsd \
+    libc_gdtoa \
+    libc_malloc \
+    libc_netbsd \
+    libc_openbsd_ndk \
+    libc_stack_protector \
+    libc_syscalls \
+    libc_tzcode \
+    libm \
+
+LOCAL_WHOLE_STATIC_LIBRARIES_arm := libc_aeabi
+LOCAL_CXX_STL := none
+
+ifneq ($(MALLOC_IMPL),dlmalloc)
+LOCAL_WHOLE_STATIC_LIBRARIES += libjemalloc
+endif
+
+$(eval $(call patch-up-arch-specific-flags,LOCAL_CFLAGS,libc_common_cflags))
+$(eval $(call patch-up-arch-specific-flags,LOCAL_SRC_FILES,libc_common_src_files))
+$(eval $(call patch-up-arch-specific-flags,LOCAL_SRC_FILES,libc_arch_dynamic_src_files))
+$(eval $(call patch-up-arch-specific-flags,LOCAL_ASFLAGS,LOCAL_CFLAGS))
+
+LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
+include $(BUILD_STATIC_LIBRARY)
+
 # ========================================================
 # libc_common.a
 # ========================================================
@@ -959,6 +1192,7 @@ LOCAL_CLANG := $(use_clang)
 LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
 LOCAL_WHOLE_STATIC_LIBRARIES := \
     libc_bionic \
+    libc_bionic_ndk \
     libc_cxa \
     libc_dns \
     libc_freebsd \
@@ -966,16 +1200,15 @@ LOCAL_WHOLE_STATIC_LIBRARIES := \
     libc_malloc \
     libc_netbsd \
     libc_openbsd \
+    libc_openbsd_ndk \
+    libc_pthread \
     libc_stack_protector \
     libc_syscalls \
+    libc_thread_atexit_impl \
     libc_tzcode \
 
 LOCAL_WHOLE_STATIC_LIBRARIES_arm := libc_aeabi
 
-ifneq ($(MALLOC_IMPL),dlmalloc)
-LOCAL_WHOLE_STATIC_LIBRARIES += libjemalloc
-endif
-
 LOCAL_CXX_STL := none
 LOCAL_SYSTEM_SHARED_LIBRARIES :=
 
@@ -1004,7 +1237,6 @@ include $(CLEAR_VARS)
 
 LOCAL_SRC_FILES := \
     $(libc_arch_static_src_files) \
-    $(libc_static_common_src_files) \
     bionic/libc_init_static.cpp
 
 LOCAL_C_INCLUDES := $(libc_common_c_includes)
@@ -1056,7 +1288,6 @@ include $(CLEAR_VARS)
 
 LOCAL_SRC_FILES := \
     $(libc_arch_static_src_files) \
-    $(libc_static_common_src_files) \
     bionic/malloc_debug_common.cpp \
     bionic/libc_init_static.cpp \
 
@@ -1070,6 +1301,11 @@ LOCAL_MODULE := libc
 LOCAL_CLANG := $(use_clang)
 LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
 LOCAL_WHOLE_STATIC_LIBRARIES := libc_common
+
+ifneq ($(MALLOC_IMPL),dlmalloc)
+LOCAL_WHOLE_STATIC_LIBRARIES += libjemalloc
+endif
+
 LOCAL_CXX_STL := none
 LOCAL_SYSTEM_SHARED_LIBRARIES :=
 LOCAL_ADDRESS_SANITIZER := false
@@ -1088,23 +1324,31 @@ include $(CLEAR_VARS)
 LOCAL_CFLAGS := $(libc_common_cflags)
 LOCAL_CONLYFLAGS := $(libc_common_conlyflags)
 LOCAL_CPPFLAGS := $(libc_common_cppflags)
+
 LOCAL_C_INCLUDES := $(libc_common_c_includes)
 LOCAL_SRC_FILES := \
+    arch-common/bionic/crtbegin_so.c \
+    arch-common/bionic/crtbrand.S \
     $(libc_arch_dynamic_src_files) \
-    $(libc_static_common_src_files) \
     bionic/malloc_debug_common.cpp \
     bionic/libc_init_dynamic.cpp \
     bionic/NetdClient.cpp \
+    arch-common/bionic/crtend_so.S \
 
 LOCAL_MODULE := libc
 LOCAL_CLANG := $(use_clang)
-LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
 LOCAL_REQUIRED_MODULES := tzdata
+LOCAL_ADDITIONAL_DEPENDENCIES := \
+    $(libc_common_additional_dependencies) \
+    $(LOCAL_PATH)/libc.map \
 
 # Leave the symbols in the shared library so that stack unwinders can produce
 # meaningful name resolution.
 LOCAL_STRIP_MODULE := keep_symbols
 
+# Do not pack libc.so relocations; see http://b/20645321 for details.
+LOCAL_PACK_MODULE_RELOCATIONS := false
+
 # WARNING: The only library libc.so should depend on is libdl.so!  If you add other libraries,
 # make sure to add -Wl,--exclude-libs=libgcc.a to the LOCAL_LDFLAGS for those libraries.  This
 # ensures that symbols that are pulled into those new libraries from libgcc.a are not declared
@@ -1115,26 +1359,37 @@ LOCAL_STRIP_MODULE := keep_symbols
 
 LOCAL_SHARED_LIBRARIES := libdl
 LOCAL_WHOLE_STATIC_LIBRARIES := libc_common
+
+ifneq ($(MALLOC_IMPL),dlmalloc)
+LOCAL_WHOLE_STATIC_LIBRARIES += libjemalloc
+endif
+
 LOCAL_CXX_STL := none
 LOCAL_SYSTEM_SHARED_LIBRARIES :=
 
+# Don't re-export new/delete and friends, even if the compiler really wants to.
+LOCAL_LDFLAGS := -Wl,--version-script,$(LOCAL_PATH)/libc.map
+
 # We'd really like to do this for all architectures, but since this wasn't done
 # before, these symbols must continue to be exported on LP32 for binary
 # compatibility.
-LOCAL_LDFLAGS_arm64 := -Wl,--exclude-libs,libgcc.a
-LOCAL_LDFLAGS_mips64 := -Wl,--exclude-libs,libgcc.a
-LOCAL_LDFLAGS_x86_64 := -Wl,--exclude-libs,libgcc.a
+LOCAL_LDFLAGS_64 := -Wl,--exclude-libs,libgcc.a
+
+# TODO: This is to work around b/19059885. Remove after root cause is fixed
+LOCAL_LDFLAGS_arm := -Wl,--hash-style=both
+LOCAL_LDFLAGS_x86 := -Wl,--hash-style=both
 
 $(eval $(call patch-up-arch-specific-flags,LOCAL_CFLAGS,libc_common_cflags))
 $(eval $(call patch-up-arch-specific-flags,LOCAL_SRC_FILES,libc_arch_dynamic_src_files))
-$(eval $(call patch-up-arch-specific-flags,LOCAL_SRC_FILES,libc_static_common_src_files))
+
+LOCAL_NO_CRT := true
+LOCAL_ASFLAGS += $(libc_crt_target_cflags)
+
 # special for arm
-LOCAL_NO_CRT_arm := true
 LOCAL_CFLAGS_arm += -DCRT_LEGACY_WORKAROUND
 LOCAL_SRC_FILES_arm += \
-    arch-common/bionic/crtbegin_so.c \
-    arch-arm/bionic/atexit_legacy.c \
-    arch-common/bionic/crtend_so.S
+    arch-arm/bionic/atexit_legacy.c
+
 LOCAL_ADDRESS_SANITIZER := false
 LOCAL_NATIVE_COVERAGE := $(bionic_coverage)
 
@@ -1171,16 +1426,22 @@ LOCAL_SRC_FILES := \
 
 LOCAL_MODULE := libc_malloc_debug_leak
 LOCAL_CLANG := $(use_clang)
-LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
+LOCAL_ADDITIONAL_DEPENDENCIES := \
+    $(libc_common_additional_dependencies) \
+    $(LOCAL_PATH)/version_script.txt \
 
 LOCAL_SHARED_LIBRARIES := libc libdl
 LOCAL_CXX_STL := none
 LOCAL_SYSTEM_SHARED_LIBRARIES :=
 # Only need this for arm since libc++ uses its own unwind code that
 # doesn't mix with the other default unwind code.
-LOCAL_STATIC_LIBRARIES_arm := libunwind_llvm libc++abi
+LOCAL_STATIC_LIBRARIES_arm := libunwind_llvm
+LOCAL_STATIC_LIBRARIES += libc++abi
 LOCAL_ALLOW_UNDEFINED_SYMBOLS := true
 
+# Don't re-export new/delete and friends, even if the compiler really wants to.
+LOCAL_LDFLAGS := -Wl,--version-script,$(LOCAL_PATH)/version_script.txt
+
 # Don't install on release build
 LOCAL_MODULE_TAGS := eng debug
 LOCAL_ADDRESS_SANITIZER := false
@@ -1210,12 +1471,17 @@ LOCAL_SRC_FILES := \
 
 LOCAL_MODULE := libc_malloc_debug_qemu
 LOCAL_CLANG := $(use_clang)
-LOCAL_ADDITIONAL_DEPENDENCIES := $(libc_common_additional_dependencies)
+LOCAL_ADDITIONAL_DEPENDENCIES := \
+    $(libc_common_additional_dependencies) \
+    $(LOCAL_PATH)/version_script.txt \
 
 LOCAL_SHARED_LIBRARIES := libc libdl
 LOCAL_CXX_STL := none
 LOCAL_SYSTEM_SHARED_LIBRARIES :=
 
+# Don't re-export new/delete and friends, even if the compiler really wants to.
+LOCAL_LDFLAGS := -Wl,--version-script,$(LOCAL_PATH)/version_script.txt
+
 # Don't install on release build
 LOCAL_MODULE_TAGS := eng debug
 LOCAL_ADDRESS_SANITIZER := false
@@ -1239,6 +1505,11 @@ include $(CLEAR_VARS)
 LOCAL_C_INCLUDES := $(libc_common_c_includes) bionic/libstdc++/include
 LOCAL_CFLAGS := $(libc_common_cflags)
 LOCAL_CPPFLAGS := $(libc_common_cppflags)
+
+# TODO: This is to work around b/19059885. Remove after root cause is fixed
+LOCAL_LDFLAGS_arm := -Wl,--hash-style=both
+LOCAL_LDFLAGS_x86 := -Wl,--hash-style=both
+
 LOCAL_SRC_FILES := $(libstdcxx_common_src_files)
 LOCAL_MODULE:= libstdc++
 LOCAL_ADDITIONAL_DEPENDENCIES := $(LOCAL_PATH)/Android.mk

libc/NOTICE

@@ -307,34 +307,6 @@ SUCH DAMAGE.
 
 -------------------------------------------------------------------
 
-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.
-
--------------------------------------------------------------------
-
 Copyright (C) 2009 The Android Open Source Project
 
 Licensed under the Apache License, Version 2.0 (the "License");
@@ -454,22 +426,6 @@ Android adaptation and tweak by Jim Huang <jserv@0xlab.org>.
 
 -------------------------------------------------------------------
 
-Copyright (C) 2011 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.
-
--------------------------------------------------------------------
-
 Copyright (C) 2011 The Android Open Source Project
 All rights reserved.
 
@@ -688,6 +644,50 @@ SUCH DAMAGE.
 
 -------------------------------------------------------------------
 
+Copyright (C) 2015 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.
+
+-------------------------------------------------------------------
+
+Copyright (C) 2015 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) 1980, 1983, 1988, 1993
    The Regents of the University of California.  All rights reserved.
 
@@ -1679,35 +1679,6 @@ SUCH DAMAGE.
 
 -------------------------------------------------------------------
 
-Copyright (c) 1990 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.
-
--------------------------------------------------------------------
-
 Copyright (c) 1990 The Regents of the University of California.
 All rights reserved.
 
@@ -2539,33 +2510,6 @@ SUCH DAMAGE.
 
 -------------------------------------------------------------------
 
-Copyright (c) 1993 Christopher G. Demetriou
-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 author may not 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 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) 1993 Martin Birgmeier
 All rights reserved.
 
@@ -2606,33 +2550,6 @@ ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 -------------------------------------------------------------------
 
-Copyright (c) 1995, 1996 Carnegie-Mellon University.
-All rights reserved.
-
-Author: Chris G. Demetriou
-
-Permission to use, copy, modify and distribute this software and
-its documentation is hereby granted, provided that both the copyright
-notice and this permission notice appear in all copies of the
-software, derivative works or modified versions, and any portions
-thereof, and that both notices appear in supporting documentation.
-
-CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
-CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
-FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
-
-Carnegie Mellon requests users of this software to return to
-
- Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
- School of Computer Science
- Carnegie Mellon University
- Pittsburgh PA 15213-3890
-
-any improvements or extensions that they make and grant Carnegie the
-rights to redistribute these changes.
-
--------------------------------------------------------------------
-
 Copyright (c) 1996 by Internet Software Consortium.
 
 Permission to use, copy, modify, and distribute this software for any
@@ -2685,63 +2602,6 @@ OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 
 -------------------------------------------------------------------
 
-Copyright (c) 1996-2004 Per Fogelstrom, Opsycon AB
-
-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.
-
-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 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) 1997 Mark Brinicombe
-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. All advertising materials mentioning features or use of this software
-   must display the following acknowledgement:
-   This product includes software developed by Mark Brinicombe
-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 AUTHOR 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 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.
-
--------------------------------------------------------------------
-
 Copyright (c) 1997 Mark Brinicombe
 Copyright (c) 2010 Android Open Source Project.
 All rights reserved.
@@ -3365,34 +3225,6 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 -------------------------------------------------------------------
 
-Copyright (c) 2002 Opsycon AB  (www.opsycon.se / www.opsycon.com)
-
-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 Opsycon AB 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 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 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) 2002 The NetBSD Foundation, Inc.
 All rights reserved.
 
@@ -3936,6 +3768,22 @@ OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 
 -------------------------------------------------------------------
 
+Copyright (c) 2007 Todd C. Miller <Todd.Miller@courtesan.com>
+
+Permission to use, copy, modify, and distribute this software for any
+purpose with or without fee is hereby granted, provided that the above
+copyright notice and this permission notice appear in all copies.
+
+THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+
+-------------------------------------------------------------------
+
 Copyright (c) 2007-2008  Michael G Schwern
 
 This software originally derived from Paul Sheer's pivotal_gmtime_r.c.
@@ -4025,6 +3873,22 @@ SUCH DAMAGE.
 
 -------------------------------------------------------------------
 
+Copyright (c) 2008 Otto Moerbeek <otto@drijf.net>
+
+Permission to use, copy, modify, and distribute this software for any
+purpose with or without fee is hereby granted, provided that the above
+copyright notice and this permission notice appear in all copies.
+
+THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+
+-------------------------------------------------------------------
+
 Copyright (c) 2008 Todd C. Miller <millert@openbsd.org>
 
 Permission to use, copy, modify, and distribute this software for any
@@ -4611,35 +4475,32 @@ OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 
 -------------------------------------------------------------------
 
-Copyright (c) 2013, Linaro Limited
-   All rights reserved.
+Copyright (c) 2013 The NetBSD Foundation, Inc.
+All rights reserved.
 
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions
-   are met:
+This code is derived from software contributed to The NetBSD Foundation
+by Christos Zoulas.
 
-      * Redistributions of source code must retain the above copyright
+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.
-
-      * Redistributions in binary form must reproduce the above copyright
+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.
 
-      * Neither the name of Linaro Limited 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 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
-   HOLDER 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.
+THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. 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 FOUNDATION 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.
 
 -------------------------------------------------------------------
 
@@ -4861,31 +4722,6 @@ THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 -------------------------------------------------------------------
 
-Copyright 2000 David E. O'Brien, John D. Polstra.
-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.
-
-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 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 2008  Android Open Source Project (source port randomization)
 Copyright (c) 1985, 1989, 1993
    The Regents of the University of California.  All rights reserved.

libc/SYSCALLS.TXT

@@ -95,7 +95,7 @@ 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
+int         ___close:close(int)  all
 pid_t       __getpid:getpid()  all
 int         munmap(void*, size_t)  all
 void*       mremap(void*, size_t, size_t, unsigned long)  all
@@ -113,7 +113,7 @@ 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         ___fchmod:fchmod(int, mode_t)  all
 int         dup(int)  all
 int         pipe2(int*, int) all
 int         dup3(int, int, int)   all
@@ -122,19 +122,19 @@ 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
+int         ___fsetxattr:fsetxattr(int, const char*, const void*, size_t, int) all
+ssize_t     ___fgetxattr:fgetxattr(int, const char*, void*, size_t) all
 ssize_t     flistxattr(int, char*, size_t) all
 int         fremovexattr(int, const char*) all
 
 int __getdents64:getdents64(unsigned int, struct dirent*, unsigned int)   arm,arm64,mips,mips64,x86,x86_64
 
 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 ___faccessat:faccessat(int, const char*, int)  all
+int ___fchmodat:fchmodat(int, const char*, mode_t)  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 fstatat64|fstatat:newfstatat(int, const char*, struct stat*, int)  arm64,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
@@ -151,7 +151,6 @@ int utimensat(int, const char*, const struct timespec times[2], int)  all
 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
@@ -174,12 +173,12 @@ int __fadvise64:fadvise64_64(int, off64_t, off64_t, int) x86
 int __fadvise64:fadvise64(int, off64_t, off64_t, int) arm64,mips,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 __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 __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
+int     fstat64|fstat:fstat(int, struct stat*)    arm64,x86_64
 
 # file system
 int     chdir(const char*)              all
@@ -204,7 +203,7 @@ clock_t       times(struct tms*)       all
 int           nanosleep(const struct timespec*, struct timespec*)   all
 int           clock_settime(clockid_t, const struct timespec*)  all
 int           clock_getres(clockid_t, struct timespec*)   all
-int           __clock_nanosleep:clock_nanosleep(clockid_t, int, const struct timespec*, struct timespec*)  all
+int           ___clock_nanosleep:clock_nanosleep(clockid_t, int, const struct timespec*, struct timespec*)  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
@@ -223,7 +222,7 @@ 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     __rt_sigqueueinfo:rt_sigqueueinfo(pid_t, int, siginfo_t*)  all
+int     ___rt_sigqueueinfo:rt_sigqueueinfo(pid_t, int, siginfo_t*)  all
 int     __signalfd4:signalfd4(int, const sigset_t*, size_t, int)  all
 
 # sockets
@@ -309,6 +308,9 @@ int inotify_rm_watch(int, unsigned int)  all
 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
 
+ssize_t process_vm_readv(pid_t, const struct iovec*, unsigned long, const struct iovec*, unsigned long, unsigned long)  all
+ssize_t process_vm_writev(pid_t, const struct iovec*, unsigned long, const struct iovec*, unsigned long, unsigned long)  all
+
 int __set_tid_address:set_tid_address(int*)  all
 
 int setfsgid(gid_t)  all

libc/arch-arm/arm.mk

@@ -1,30 +1,10 @@
 # 32-bit arm.
 
-#
-# Various kinds of LP32 cruft.
-#
-
-libc_bionic_src_files_arm += \
-    bionic/mmap.cpp \
-
-libc_common_src_files_arm += \
-    bionic/legacy_32_bit_support.cpp \
-    bionic/ndk_cruft.cpp \
-    bionic/time64.c \
-
-libc_netbsd_src_files_arm += \
-    upstream-netbsd/common/lib/libc/hash/sha1/sha1.c \
-
-libc_openbsd_src_files_arm += \
-    upstream-openbsd/lib/libc/stdio/putw.c \
-
 #
 # Default implementations of functions that are commonly optimized.
 #
 
 libc_bionic_src_files_arm += \
-    bionic/memchr.c \
-    bionic/memrchr.c \
     bionic/strchr.cpp \
     bionic/strnlen.c \
     bionic/strrchr.cpp \
@@ -40,7 +20,8 @@ libc_freebsd_src_files_arm += \
     upstream-freebsd/lib/libc/string/wmemmove.c \
 
 libc_openbsd_src_files_arm += \
-    upstream-openbsd/lib/libc/string/bcopy.c \
+    upstream-openbsd/lib/libc/string/memchr.c \
+    upstream-openbsd/lib/libc/string/memrchr.c \
     upstream-openbsd/lib/libc/string/stpncpy.c \
     upstream-openbsd/lib/libc/string/strlcat.c \
     upstream-openbsd/lib/libc/string/strlcpy.c \
@@ -58,7 +39,6 @@ libc_bionic_src_files_arm += \
     arch-arm/bionic/__bionic_clone.S \
     arch-arm/bionic/_exit_with_stack_teardown.S \
     arch-arm/bionic/libgcc_compat.c \
-    arch-arm/bionic/memcmp.S \
     arch-arm/bionic/__restore.S \
     arch-arm/bionic/setjmp.S \
     arch-arm/bionic/syscall.S \

libc/arch-arm/bionic/__bionic_clone.S

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2008-2010 The Android Open Source Project
+ * Copyright (C) 2008 The Android Open Source Project
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without

libc/arch-arm/bionic/memcpy.S

@@ -1,686 +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 <private/bionic_asm.h>
-
-#if defined(__ARM_NEON__) && !defined(ARCH_ARM_USE_NON_NEON_MEMCPY)
-
-        .text
-        .fpu    neon
-
-#ifdef HAVE_32_BYTE_CACHE_LINE
-/* a prefetch distance of 2 cache-lines */
-#define CACHE_LINE_SIZE     32
-#else
-/* a prefetch distance of 4 cache-lines works best experimentally */
-#define CACHE_LINE_SIZE     64
-#endif
-
-ENTRY(memcpy)
-        .save       {r0, lr}
-        /* start preloading as early as possible */
-        pld         [r1, #(CACHE_LINE_SIZE * 0)]
-        stmfd       sp!, {r0, lr}
-        pld         [r1, #(CACHE_LINE_SIZE * 1)]
-
-/* If Neon supports unaligned access then remove the align code,
- * unless a size limit has been specified.
- */
-#ifndef NEON_UNALIGNED_ACCESS
-        /* do we have at least 16-bytes to copy (needed for alignment below) */
-        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
-
-        /* align destination to cache-line for the write-buffer */
-        rsb         r3, r0, #0
-        ands        r3, r3, #0xF
-        beq         2f
-
-        /* copy up to 15-bytes (count in r3) */
-        sub         r2, r2, r3
-        movs        ip, r3, lsl #31
-        ldrmib      lr, [r1], #1
-        strmib      lr, [r0], #1
-        ldrcsb      ip, [r1], #1
-        ldrcsb      lr, [r1], #1
-        strcsb      ip, [r0], #1
-        strcsb      lr, [r0], #1
-        movs        ip, r3, lsl #29
-        bge         1f
-        // copies 4 bytes, destination 32-bits aligned
-        vld4.8      {d0[0], d1[0], d2[0], d3[0]}, [r1]!
-        vst4.8      {d0[0], d1[0], d2[0], d3[0]}, [r0, :32]!
-1:      bcc         2f
-        // copies 8 bytes, destination 64-bits aligned
-        vld1.8      {d0}, [r1]!
-        vst1.8      {d0}, [r0, :64]!
-2:
-        /* preload immediately the next cache line, which we may need */
-        pld         [r1, #(CACHE_LINE_SIZE * 0)]
-        pld         [r1, #(CACHE_LINE_SIZE * 1)]
-
-#ifdef HAVE_32_BYTE_CACHE_LINE
-        /* make sure we have at least 32 bytes to copy */
-        subs        r2, r2, #32
-        blo         4f
-
-        /* preload all the cache lines we need.
-         * NOTE: the number of pld below depends on PREFETCH_DISTANCE,
-         * ideally would would increase the distance in the main loop to
-         * avoid the goofy code below. In practice this doesn't seem to make
-         * a big difference.
-         */
-        pld         [r1, #(PREFETCH_DISTANCE)]
-
-1:      /* The main loop copies 32 bytes at a time */
-        vld1.8      {d0  - d3},   [r1]!
-        pld         [r1, #(PREFETCH_DISTANCE)]
-        subs        r2, r2, #32
-        vst1.8      {d0  - d3},   [r0, :128]!
-        bhs         1b
-#else
-        /* make sure we have at least 64 bytes to copy */
-        subs        r2, r2, #64
-        blo         2f
-
-        /* preload all the cache lines we need. */
-        pld         [r1, #(CACHE_LINE_SIZE * 2)]
-        pld         [r1, #(CACHE_LINE_SIZE * 3)]
-
-1:      /* The main loop copies 64 bytes at a time */
-        vld1.8      {d0 - d3}, [r1]!
-        vld1.8      {d4 - d7}, [r1]!
-#ifdef  HAVE_32_BYTE_CACHE_LINE
-        pld         [r1, #(CACHE_LINE_SIZE * 2)]
-        pld         [r1, #(CACHE_LINE_SIZE * 3)]
-#else
-        pld         [r1, #(CACHE_LINE_SIZE * 3)]
-#endif
-        subs        r2, r2, #64
-        vst1.8      {d0 - d3}, [r0, :128]!
-        vst1.8      {d4 - d7}, [r0, :128]!
-        bhs         1b
-
-2:      /* fix-up the remaining count and make sure we have >= 32 bytes left */
-        add         r2, r2, #64
-        subs        r2, r2, #32
-        blo         4f
-
-3:      /* 32 bytes at a time. These cache lines were already preloaded */
-        vld1.8      {d0 - d3}, [r1]!
-        subs        r2, r2, #32
-        vst1.8      {d0 - d3}, [r0, :128]!
-        bhs         3b
-#endif
-4:      /* less than 32 left */
-        add         r2, r2, #32
-        tst         r2, #0x10
-        beq         5f
-        // copies 16 bytes, 128-bits aligned
-        vld1.8      {d0, d1}, [r1]!
-        vst1.8      {d0, d1}, [r0, :128]!
-5:      /* copy up to 15-bytes (count in r2) */
-        movs        ip, r2, lsl #29
-        bcc         1f
-        vld1.8      {d0}, [r1]!
-        vst1.8      {d0}, [r0]!
-1:      bge         2f
-        vld4.8      {d0[0], d1[0], d2[0], d3[0]}, [r1]!
-        vst4.8      {d0[0], d1[0], d2[0], d3[0]}, [r0]!
-2:      movs        ip, r2, lsl #31
-        ldrmib      r3, [r1], #1
-        ldrcsb      ip, [r1], #1
-        ldrcsb      lr, [r1], #1
-        strmib      r3, [r0], #1
-        strcsb      ip, [r0], #1
-        strcsb      lr, [r0], #1
-
-        ldmfd       sp!, {r0, lr}
-        bx          lr
-
-#else   /* NEON_UNALIGNED_ACCESS */
-
-        // Check so divider is at least 16 bytes, needed for alignment code.
-        cmp         r2, #16
-        blo         5f
-
-#ifdef NEON_MEMCPY_ALIGNMENT_DIVIDER
-        /* Check the upper size limit for Neon unaligned memory access in memcpy */
-#if NEON_MEMCPY_ALIGNMENT_DIVIDER >= 16
-        cmp         r2, #NEON_MEMCPY_ALIGNMENT_DIVIDER
-        blo         3f
-#endif
-        /* check if buffers are aligned. If so, run arm-only version */
-        eor         r3, r0, r1
-        ands        r3, r3, #0x3
-        beq         11f
-
-        /* align destination to 16 bytes for the write-buffer */
-        rsb         r3, r0, #0
-        ands        r3, r3, #0xF
-        beq         3f
-
-        /* copy up to 15-bytes (count in r3) */
-        sub         r2, r2, r3
-        movs        ip, r3, lsl #31
-        ldrmib      lr, [r1], #1
-        strmib      lr, [r0], #1
-        ldrcsb      ip, [r1], #1
-        ldrcsb      lr, [r1], #1
-        strcsb      ip, [r0], #1
-        strcsb      lr, [r0], #1
-        movs        ip, r3, lsl #29
-        bge         1f
-        // copies 4 bytes, destination 32-bits aligned
-        vld1.32     {d0[0]}, [r1]!
-        vst1.32     {d0[0]}, [r0, :32]!
-1:      bcc         2f
-        // copies 8 bytes, destination 64-bits aligned
-        vld1.8      {d0}, [r1]!
-        vst1.8      {d0}, [r0, :64]!
-2:
-        /* preload immediately the next cache line, which we may need */
-        pld         [r1, #(CACHE_LINE_SIZE * 0)]
-        pld         [r1, #(CACHE_LINE_SIZE * 1)]
-3:
-#endif
-        /* make sure we have at least 64 bytes to copy */
-        subs        r2, r2, #64
-        blo         2f
-
-        /* preload all the cache lines we need */
-        pld         [r1, #(CACHE_LINE_SIZE * 2)]
-        pld         [r1, #(CACHE_LINE_SIZE * 3)]
-
-1:      /* The main loop copies 64 bytes at a time */
-        vld1.8      {d0 - d3}, [r1]!
-        vld1.8      {d4 - d7}, [r1]!
-#ifdef  HAVE_32_BYTE_CACHE_LINE
-        pld         [r1, #(CACHE_LINE_SIZE * 2)]
-        pld         [r1, #(CACHE_LINE_SIZE * 3)]
-#else
-        pld         [r1, #(CACHE_LINE_SIZE * 3)]
-#endif
-        subs        r2, r2, #64
-        vst1.8      {d0 - d3}, [r0]!
-        vst1.8      {d4 - d7}, [r0]!
-        bhs         1b
-
-2:      /* fix-up the remaining count and make sure we have >= 32 bytes left */
-        add         r2, r2, #64
-        subs        r2, r2, #32
-        blo         4f
-
-3:      /* 32 bytes at a time. These cache lines were already preloaded */
-        vld1.8      {d0 - d3}, [r1]!
-        subs        r2, r2, #32
-        vst1.8      {d0 - d3}, [r0]!
-        bhs         3b
-
-4:      /* less than 32 left */
-        add         r2, r2, #32
-        tst         r2, #0x10
-        beq         5f
-        // copies 16 bytes, 128-bits aligned
-        vld1.8      {d0, d1}, [r1]!
-        vst1.8      {d0, d1}, [r0]!
-5:      /* copy up to 15-bytes (count in r2) */
-        movs        ip, r2, lsl #29
-        bcc         1f
-        vld1.8      {d0}, [r1]!
-        vst1.8      {d0}, [r0]!
-1:      bge         2f
-        vld1.32     {d0[0]}, [r1]!
-        vst1.32     {d0[0]}, [r0]!
-2:      movs        ip, r2, lsl #31
-        ldrmib      r3, [r1], #1
-        ldrcsb      ip, [r1], #1
-        ldrcsb      lr, [r1], #1
-        strmib      r3, [r0], #1
-        strcsb      ip, [r0], #1
-        strcsb      lr, [r0], #1
-
-        ldmfd       sp!, {r0, lr}
-        bx          lr
-#endif  /* NEON_UNALIGNED_ACCESS */
-11:
-        /* Simple arm-only copy loop to handle aligned copy operations */
-        stmfd       sp!, {r4, r5, r6, r7, r8}
-        pld         [r1, #(CACHE_LINE_SIZE * 2)]
-
-        /* Check alignment */
-        rsb         r3, r1, #0
-        ands        r3, #3
-        beq         2f
-
-        /* align source to 32 bits. We need to insert 2 instructions between
-         * a ldr[b|h] and str[b|h] because byte and half-word instructions
-         * stall 2 cycles.
-         */
-        movs        r12, r3, lsl #31
-        sub         r2, r2, r3      /* we know that r3 <= r2 because r2 >= 4 */
-        ldrmib      r3, [r1], #1
-        ldrcsb      r4, [r1], #1
-        ldrcsb      r5, [r1], #1
-        strmib      r3, [r0], #1
-        strcsb      r4, [r0], #1
-        strcsb      r5, [r0], #1
-2:
-        subs        r2, #32
-        blt         5f
-        pld         [r1, #(CACHE_LINE_SIZE * 3)]
-3:      /* Main copy loop, copying 32 bytes at a time */
-        pld         [r1, #(CACHE_LINE_SIZE * 4)]
-        ldmia       r1!, {r3, r4, r5, r6, r7, r8, r12, lr}
-        subs        r2, r2, #32
-        stmia       r0!, {r3, r4, r5, r6, r7, r8, r12, lr}
-        bge         3b
-5:      /* Handle any remaining bytes */
-        adds        r2, #32
-        beq         6f
-
-        movs        r12, r2, lsl #28
-        ldmcsia     r1!, {r3, r4, r5, r6}   /* 16 bytes */
-        ldmmiia     r1!, {r7, r8}           /*  8 bytes */
-        stmcsia     r0!, {r3, r4, r5, r6}
-        stmmiia     r0!, {r7, r8}
-        movs        r12, r2, lsl #30
-        ldrcs       r3, [r1], #4            /*  4 bytes */
-        ldrmih      r4, [r1], #2            /*  2 bytes */
-        strcs       r3, [r0], #4
-        strmih      r4, [r0], #2
-        tst         r2, #0x1
-        ldrneb      r3, [r1]                /*  last byte  */
-        strneb      r3, [r0]
-6:
-        ldmfd       sp!, {r4, r5, r6, r7, r8}
-        ldmfd       sp!, {r0, pc}
-END(memcpy)
-
-
-#else   /* __ARM_ARCH__ < 7 */
-
-
-		/*
-		 * Optimized memcpy() for ARM.
-         *
-		 * note that memcpy() always returns the destination pointer,
-		 * so we have to preserve R0.
-		 */
-
-ENTRY(memcpy)
-		/* The stack must always be 64-bits aligned to be compliant with the
-		 * ARM ABI. Since we have to save R0, we might as well save R4
-		 * which we can use for better pipelining of the reads below
-		 */
-        .save       {r0, r4, lr}
-        stmfd       sp!, {r0, r4, lr}
-        /* Making room for r5-r11 which will be spilled later */
-        .pad        #28
-        sub         sp, sp, #28
-
-        // 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]
-
-		/* it simplifies things to take care of len<4 early */
-		cmp			r2, #4
-		blo			copy_last_3_and_return
-
-		/* compute the offset to align the source
-		 * offset = (4-(src&3))&3 = -src & 3
-		 */
-		rsb			r3, r1, #0
-		ands		r3, r3, #3
-		beq			src_aligned
-
-		/* align source to 32 bits. We need to insert 2 instructions between
-		 * a ldr[b|h] and str[b|h] because byte and half-word instructions
-		 * stall 2 cycles.
-		 */
-		movs		r12, r3, lsl #31
-		sub			r2, r2, r3		/* we know that r3 <= r2 because r2 >= 4 */
-		ldrmib		r3, [r1], #1
-		ldrcsb		r4, [r1], #1
-		ldrcsb		r12,[r1], #1
-        strmib		r3, [r0], #1
-		strcsb		r4, [r0], #1
-		strcsb		r12,[r0], #1
-
-src_aligned:
-
-		/* see if src and dst are aligned together (congruent) */
-		eor			r12, r0, r1
-		tst			r12, #3
-		bne			non_congruent
-
-        /* Use post-incriment mode for stm to spill r5-r11 to reserved stack
-         * frame. Don't update sp.
-         */
-        stmea		sp, {r5-r11}
-
-		/* align the destination to a cache-line */
-		rsb         r3, r0, #0
-		ands		r3, r3, #0x1C
-		beq         congruent_aligned32
-		cmp         r3, r2
-		andhi		r3, r2, #0x1C
-
-		/* conditionnaly copies 0 to 7 words (length in r3) */
-		movs		r12, r3, lsl #28
-		ldmcsia		r1!, {r4, r5, r6, r7}	/* 16 bytes */
-		ldmmiia		r1!, {r8, r9}			/*  8 bytes */
-		stmcsia		r0!, {r4, r5, r6, r7}
-		stmmiia		r0!, {r8, r9}
-		tst         r3, #0x4
-		ldrne		r10,[r1], #4			/*  4 bytes */
-		strne		r10,[r0], #4
-		sub         r2, r2, r3
-
-congruent_aligned32:
-		/*
-		 * here source is aligned to 32 bytes.
-		 */
-
-cached_aligned32:
-        subs        r2, r2, #32
-        blo         less_than_32_left
-
-        /*
-         * We preload a cache-line up to 64 bytes ahead. On the 926, this will
-         * stall only until the requested world is fetched, but the linefill
-         * continues in the the background.
-         * While the linefill is going, we write our previous cache-line
-         * into the write-buffer (which should have some free space).
-         * When the linefill is done, the writebuffer will
-         * start dumping its content into memory
-         *
-         * While all this is going, we then load a full cache line into
-         * 8 registers, this cache line should be in the cache by now
-         * (or partly in the cache).
-         *
-         * This code should work well regardless of the source/dest alignment.
-         *
-         */
-
-        // Align the preload register to a cache-line because the cpu does
-        // "critical word first" (the first word requested is loaded first).
-        bic         r12, r1, #0x1F
-        add         r12, r12, #64
-
-1:      ldmia       r1!, { r4-r11 }
-        pld         [r12, #64]
-        subs        r2, r2, #32
-
-        // NOTE: if r12 is more than 64 ahead of r1, the following ldrhi
-        // for ARM9 preload will not be safely guarded by the preceding subs.
-        // When it is safely guarded the only possibility to have SIGSEGV here
-        // is because the caller overstates the length.
-        ldrhi       r3, [r12], #32      /* cheap ARM9 preload */
-        stmia       r0!, { r4-r11 }
-		bhs         1b
-
-        add         r2, r2, #32
-
-
-
-
-less_than_32_left:
-		/*
-		 * less than 32 bytes left at this point (length in r2)
-		 */
-
-		/* skip all this if there is nothing to do, which should
-		 * be a common case (if not executed the code below takes
-		 * about 16 cycles)
-		 */
-		tst			r2, #0x1F
-		beq			1f
-
-		/* conditionnaly copies 0 to 31 bytes */
-		movs		r12, r2, lsl #28
-		ldmcsia		r1!, {r4, r5, r6, r7}	/* 16 bytes */
-		ldmmiia		r1!, {r8, r9}			/*  8 bytes */
-		stmcsia		r0!, {r4, r5, r6, r7}
-		stmmiia		r0!, {r8, r9}
-		movs		r12, r2, lsl #30
-		ldrcs		r3, [r1], #4			/*  4 bytes */
-		ldrmih		r4, [r1], #2			/*  2 bytes */
-		strcs		r3, [r0], #4
-		strmih		r4, [r0], #2
-		tst         r2, #0x1
-		ldrneb		r3, [r1]				/*  last byte  */
-		strneb		r3, [r0]
-
-		/* we're done! restore everything and return */
-1:		ldmfd		sp!, {r5-r11}
-		ldmfd		sp!, {r0, r4, lr}
-		bx			lr
-
-		/********************************************************************/
-
-non_congruent:
-		/*
-		 * here source is aligned to 4 bytes
-		 * but destination is not.
-		 *
-		 * in the code below r2 is the number of bytes read
-		 * (the number of bytes written is always smaller, because we have
-		 * partial words in the shift queue)
-		 */
-		cmp			r2, #4
-		blo			copy_last_3_and_return
-
-        /* Use post-incriment mode for stm to spill r5-r11 to reserved stack
-         * frame. Don't update sp.
-         */
-        stmea		sp, {r5-r11}
-
-		/* compute shifts needed to align src to dest */
-		rsb			r5, r0, #0
-		and			r5, r5, #3			/* r5 = # bytes in partial words */
-		mov			r12, r5, lsl #3		/* r12 = right */
-		rsb			lr, r12, #32		/* lr = left  */
-
-		/* read the first word */
-		ldr			r3, [r1], #4
-		sub			r2, r2, #4
-
-		/* write a partial word (0 to 3 bytes), such that destination
-		 * becomes aligned to 32 bits (r5 = nb of words to copy for alignment)
-		 */
-		movs		r5, r5, lsl #31
-		strmib		r3, [r0], #1
-		movmi		r3, r3, lsr #8
-		strcsb		r3, [r0], #1
-		movcs		r3, r3, lsr #8
-		strcsb		r3, [r0], #1
-		movcs		r3, r3, lsr #8
-
-		cmp			r2, #4
-		blo			partial_word_tail
-
-		/* Align destination to 32 bytes (cache line boundary) */
-1:		tst			r0, #0x1c
-		beq			2f
-		ldr			r5, [r1], #4
-		sub         r2, r2, #4
-		orr			r4, r3, r5,		lsl lr
-		mov			r3, r5,			lsr r12
-		str			r4, [r0], #4
-        cmp         r2, #4
-		bhs			1b
-		blo			partial_word_tail
-
-		/* copy 32 bytes at a time */
-2:		subs		r2, r2, #32
-		blo			less_than_thirtytwo
-
-		/* Use immediate mode for the shifts, because there is an extra cycle
-		 * for register shifts, which could account for up to 50% of
-		 * performance hit.
-		 */
-
-        cmp			r12, #24
-		beq			loop24
-		cmp			r12, #8
-		beq			loop8
-
-loop16:
-        ldr         r12, [r1], #4
-1:      mov         r4, r12
-		ldmia		r1!, {   r5,r6,r7,  r8,r9,r10,r11}
-        pld         [r1, #64]
-        subs        r2, r2, #32
-        ldrhs       r12, [r1], #4
-		orr			r3, r3, r4,		lsl #16
-		mov			r4, r4,			lsr #16
-		orr			r4, r4, r5,		lsl #16
-		mov			r5, r5,			lsr #16
-		orr			r5, r5, r6,		lsl #16
-		mov			r6, r6,			lsr #16
-		orr			r6, r6, r7,		lsl #16
-		mov			r7, r7,			lsr #16
-		orr			r7, r7, r8,		lsl #16
-		mov			r8, r8,			lsr #16
-		orr			r8, r8, r9,		lsl #16
-		mov			r9, r9,			lsr #16
-		orr			r9, r9, r10,	lsl #16
-		mov			r10, r10,		lsr #16
-		orr			r10, r10, r11,	lsl #16
-		stmia		r0!, {r3,r4,r5,r6, r7,r8,r9,r10}
-		mov			r3, r11,		lsr #16
-		bhs			1b
-		b			less_than_thirtytwo
-
-loop8:
-        ldr         r12, [r1], #4
-1:      mov         r4, r12
-		ldmia		r1!, {   r5,r6,r7,  r8,r9,r10,r11}
-        pld         [r1, #64]
-		subs		r2, r2, #32
-        ldrhs       r12, [r1], #4
-		orr			r3, r3, r4,		lsl #24
-		mov			r4, r4,			lsr #8
-		orr			r4, r4, r5,		lsl #24
-		mov			r5, r5,			lsr #8
-		orr			r5, r5, r6,		lsl #24
-		mov			r6, r6,			lsr #8
-		orr			r6, r6, r7,		lsl #24
-		mov			r7, r7,			lsr #8
-		orr			r7, r7, r8,		lsl #24
-		mov			r8, r8,			lsr #8
-		orr			r8, r8, r9,		lsl #24
-		mov			r9, r9,			lsr #8
-		orr			r9, r9, r10,	lsl #24
-		mov			r10, r10,		lsr #8
-		orr			r10, r10, r11,	lsl #24
-		stmia		r0!, {r3,r4,r5,r6, r7,r8,r9,r10}
-		mov			r3, r11,		lsr #8
-		bhs			1b
-		b			less_than_thirtytwo
-
-loop24:
-        ldr         r12, [r1], #4
-1:      mov         r4, r12
-		ldmia		r1!, {   r5,r6,r7,  r8,r9,r10,r11}
-        pld         [r1, #64]
-		subs		r2, r2, #32
-        ldrhs       r12, [r1], #4
-		orr			r3, r3, r4,		lsl #8
-		mov			r4, r4,			lsr #24
-		orr			r4, r4, r5,		lsl #8
-		mov			r5, r5,			lsr #24
-		orr			r5, r5, r6,		lsl #8
-		mov			r6, r6,			lsr #24
-		orr			r6, r6, r7,		lsl #8
-		mov			r7, r7,			lsr #24
-		orr			r7, r7, r8,		lsl #8
-		mov			r8, r8,			lsr #24
-		orr			r8, r8, r9,		lsl #8
-		mov			r9, r9,			lsr #24
-		orr			r9, r9, r10,	lsl #8
-		mov			r10, r10,		lsr #24
-		orr			r10, r10, r11,	lsl #8
-		stmia		r0!, {r3,r4,r5,r6, r7,r8,r9,r10}
-		mov			r3, r11,		lsr #24
-		bhs			1b
-
-
-less_than_thirtytwo:
-		/* copy the last 0 to 31 bytes of the source */
-		rsb			r12, lr, #32		/* we corrupted r12, recompute it  */
-		add			r2, r2, #32
-		cmp			r2, #4
-		blo			partial_word_tail
-
-1:		ldr			r5, [r1], #4
-		sub         r2, r2, #4
-		orr			r4, r3, r5,		lsl lr
-		mov			r3,	r5,			lsr r12
-		str			r4, [r0], #4
-        cmp         r2, #4
-		bhs			1b
-
-partial_word_tail:
-		/* we have a partial word in the input buffer */
-		movs		r5, lr, lsl #(31-3)
-		strmib		r3, [r0], #1
-		movmi		r3, r3, lsr #8
-		strcsb		r3, [r0], #1
-		movcs		r3, r3, lsr #8
-		strcsb		r3, [r0], #1
-
-		/* Refill spilled registers from the stack. Don't update sp. */
-		ldmfd		sp, {r5-r11}
-
-copy_last_3_and_return:
-		movs		r2, r2, lsl #31	/* copy remaining 0, 1, 2 or 3 bytes */
-		ldrmib		r2, [r1], #1
-		ldrcsb		r3, [r1], #1
-		ldrcsb		r12,[r1]
-		strmib		r2, [r0], #1
-		strcsb		r3, [r0], #1
-		strcsb		r12,[r0]
-
-        /* we're done! restore sp and spilled registers and return */
-        add         sp,  sp, #28
-		ldmfd		sp!, {r0, r4, lr}
-		bx			lr
-END(memcpy)
-
-
-#endif    /* __ARM_ARCH__ < 7 */

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

@@ -1,614 +0,0 @@
-/* Copyright (c) 2013, Linaro Limited
-   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.
-
-      * Neither the name of Linaro Limited 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 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
-   HOLDER 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.
-
- */
-
-/*
-   This memcpy routine is optimised for Cortex-A15 cores and takes advantage
-   of VFP or NEON when built with the appropriate flags.
-
-   Assumptions:
-
-    ARMv6 (ARMv7-a if using Neon)
-    ARM state
-    Unaligned accesses
-    LDRD/STRD support unaligned word accesses
-
- */
-
-#include <machine/cpu-features.h>
-#include <private/bionic_asm.h>
-
-	.syntax unified
-	/* This implementation requires ARM state.  */
-	.arm
-
-#ifdef __ARM_NEON__
-
-	.fpu	neon
-	.arch	armv7-a
-# define FRAME_SIZE	4
-# define USE_VFP
-# define USE_NEON
-
-#elif !defined (__SOFTFP__)
-
-	.arch	armv6
-	.fpu	vfpv2
-# define FRAME_SIZE	32
-# define USE_VFP
-
-#else
-	.arch	armv6
-# define FRAME_SIZE    32
-
-#endif
-
-/* Old versions of GAS incorrectly implement the NEON align semantics.  */
-#ifdef BROKEN_ASM_NEON_ALIGN
-#define ALIGN(addr, align) addr,:align
-#else
-#define ALIGN(addr, align) addr:align
-#endif
-
-#define PC_OFFSET	8	/* PC pipeline compensation.  */
-#define INSN_SIZE	4
-
-/* Call parameters.  */
-#define dstin	r0
-#define src	r1
-#define count	r2
-
-/* Locals.  */
-#define tmp1	r3
-#define dst	ip
-#define tmp2	r10
-
-#ifndef USE_NEON
-/* For bulk copies using GP registers.  */
-#define	A_l	r2		/* Call-clobbered.  */
-#define	A_h	r3		/* Call-clobbered.  */
-#define	B_l	r4
-#define	B_h	r5
-#define	C_l	r6
-#define	C_h	r7
-#define	D_l	r8
-#define	D_h	r9
-#endif
-
-/* Number of lines ahead to pre-fetch data.  If you change this the code
-   below will need adjustment to compensate.  */
-
-#define prefetch_lines	5
-
-#ifdef USE_VFP
-	.macro	cpy_line_vfp vreg, base
-	vstr	\vreg, [dst, #\base]
-	vldr	\vreg, [src, #\base]
-	vstr	d0, [dst, #\base + 8]
-	vldr	d0, [src, #\base + 8]
-	vstr	d1, [dst, #\base + 16]
-	vldr	d1, [src, #\base + 16]
-	vstr	d2, [dst, #\base + 24]
-	vldr	d2, [src, #\base + 24]
-	vstr	\vreg, [dst, #\base + 32]
-	vldr	\vreg, [src, #\base + prefetch_lines * 64 - 32]
-	vstr	d0, [dst, #\base + 40]
-	vldr	d0, [src, #\base + 40]
-	vstr	d1, [dst, #\base + 48]
-	vldr	d1, [src, #\base + 48]
-	vstr	d2, [dst, #\base + 56]
-	vldr	d2, [src, #\base + 56]
-	.endm
-
-	.macro	cpy_tail_vfp vreg, base
-	vstr	\vreg, [dst, #\base]
-	vldr	\vreg, [src, #\base]
-	vstr	d0, [dst, #\base + 8]
-	vldr	d0, [src, #\base + 8]
-	vstr	d1, [dst, #\base + 16]
-	vldr	d1, [src, #\base + 16]
-	vstr	d2, [dst, #\base + 24]
-	vldr	d2, [src, #\base + 24]
-	vstr	\vreg, [dst, #\base + 32]
-	vstr	d0, [dst, #\base + 40]
-	vldr	d0, [src, #\base + 40]
-	vstr	d1, [dst, #\base + 48]
-	vldr	d1, [src, #\base + 48]
-	vstr	d2, [dst, #\base + 56]
-	vldr	d2, [src, #\base + 56]
-	.endm
-#endif
-
-	.p2align 6
-ENTRY(memcpy)
-
-	mov	dst, dstin	/* Preserve dstin, we need to return it.  */
-	cmp	count, #64
-	bge	.Lcpy_not_short
-	/* Deal with small copies quickly by dropping straight into the
-	   exit block.  */
-
-.Ltail63unaligned:
-#ifdef USE_NEON
-	and	tmp1, count, #0x38
-	rsb	tmp1, tmp1, #(56 - PC_OFFSET + INSN_SIZE)
-	add	pc, pc, tmp1
-	vld1.8	{d0}, [src]!	/* 14 words to go.  */
-	vst1.8	{d0}, [dst]!
-	vld1.8	{d0}, [src]!	/* 12 words to go.  */
-	vst1.8	{d0}, [dst]!
-	vld1.8	{d0}, [src]!	/* 10 words to go.  */
-	vst1.8	{d0}, [dst]!
-	vld1.8	{d0}, [src]!	/* 8 words to go.  */
-	vst1.8	{d0}, [dst]!
-	vld1.8	{d0}, [src]!	/* 6 words to go.  */
-	vst1.8	{d0}, [dst]!
-	vld1.8	{d0}, [src]!	/* 4 words to go.  */
-	vst1.8	{d0}, [dst]!
-	vld1.8	{d0}, [src]!	/* 2 words to go.  */
-	vst1.8	{d0}, [dst]!
-
-	tst	count, #4
-	ldrne	tmp1, [src], #4
-	strne	tmp1, [dst], #4
-#else
-	/* Copy up to 15 full words of data.  May not be aligned.  */
-	/* Cannot use VFP for unaligned data.  */
-	and	tmp1, count, #0x3c
-	add	dst, dst, tmp1
-	add	src, src, tmp1
-	rsb	tmp1, tmp1, #(60 - PC_OFFSET/2 + INSN_SIZE/2)
-	/* Jump directly into the sequence below at the correct offset.  */
-	add	pc, pc, tmp1, lsl #1
-
-	ldr	tmp1, [src, #-60]	/* 15 words to go.  */
-	str	tmp1, [dst, #-60]
-
-	ldr	tmp1, [src, #-56]	/* 14 words to go.  */
-	str	tmp1, [dst, #-56]
-	ldr	tmp1, [src, #-52]
-	str	tmp1, [dst, #-52]
-
-	ldr	tmp1, [src, #-48]	/* 12 words to go.  */
-	str	tmp1, [dst, #-48]
-	ldr	tmp1, [src, #-44]
-	str	tmp1, [dst, #-44]
-
-	ldr	tmp1, [src, #-40]	/* 10 words to go.  */
-	str	tmp1, [dst, #-40]
-	ldr	tmp1, [src, #-36]
-	str	tmp1, [dst, #-36]
-
-	ldr	tmp1, [src, #-32]	/* 8 words to go.  */
-	str	tmp1, [dst, #-32]
-	ldr	tmp1, [src, #-28]
-	str	tmp1, [dst, #-28]
-
-	ldr	tmp1, [src, #-24]	/* 6 words to go.  */
-	str	tmp1, [dst, #-24]
-	ldr	tmp1, [src, #-20]
-	str	tmp1, [dst, #-20]
-
-	ldr	tmp1, [src, #-16]	/* 4 words to go.  */
-	str	tmp1, [dst, #-16]
-	ldr	tmp1, [src, #-12]
-	str	tmp1, [dst, #-12]
-
-	ldr	tmp1, [src, #-8]	/* 2 words to go.  */
-	str	tmp1, [dst, #-8]
-	ldr	tmp1, [src, #-4]
-	str	tmp1, [dst, #-4]
-#endif
-
-	lsls	count, count, #31
-	ldrhcs	tmp1, [src], #2
-	ldrbne	src, [src]		/* Src is dead, use as a scratch.  */
-	strhcs	tmp1, [dst], #2
-	strbne	src, [dst]
-	bx	lr
-
-.Lcpy_not_short:
-	/* At least 64 bytes to copy, but don't know the alignment yet.  */
-	str	tmp2, [sp, #-FRAME_SIZE]!
-	and	tmp2, src, #7
-	and	tmp1, dst, #7
-	cmp	tmp1, tmp2
-	bne	.Lcpy_notaligned
-
-#ifdef USE_VFP
-	/* Magic dust alert!  Force VFP on Cortex-A9.  Experiments show
-	   that the FP pipeline is much better at streaming loads and
-	   stores.  This is outside the critical loop.  */
-	vmov.f32	s0, s0
-#endif
-
-	/* SRC and DST have the same mutual 32-bit alignment, but we may
-	   still need to pre-copy some bytes to get to natural alignment.
-	   We bring DST into full 64-bit alignment.  */
-	lsls	tmp2, dst, #29
-	beq	1f
-	rsbs	tmp2, tmp2, #0
-	sub	count, count, tmp2, lsr #29
-	ldrmi	tmp1, [src], #4
-	strmi	tmp1, [dst], #4
-	lsls	tmp2, tmp2, #2
-	ldrhcs	tmp1, [src], #2
-	ldrbne	tmp2, [src], #1
-	strhcs	tmp1, [dst], #2
-	strbne	tmp2, [dst], #1
-
-1:
-	subs	tmp2, count, #64	/* Use tmp2 for count.  */
-	blt	.Ltail63aligned
-
-	cmp	tmp2, #512
-	bge	.Lcpy_body_long
-
-.Lcpy_body_medium:			/* Count in tmp2.  */
-#ifdef USE_VFP
-1:
-	vldr	d0, [src, #0]
-	subs	tmp2, tmp2, #64
-	vldr	d1, [src, #8]
-	vstr	d0, [dst, #0]
-	vldr	d0, [src, #16]
-	vstr	d1, [dst, #8]
-	vldr	d1, [src, #24]
-	vstr	d0, [dst, #16]
-	vldr	d0, [src, #32]
-	vstr	d1, [dst, #24]
-	vldr	d1, [src, #40]
-	vstr	d0, [dst, #32]
-	vldr	d0, [src, #48]
-	vstr	d1, [dst, #40]
-	vldr	d1, [src, #56]
-	vstr	d0, [dst, #48]
-	add	src, src, #64
-	vstr	d1, [dst, #56]
-	add	dst, dst, #64
-	bge	1b
-	tst	tmp2, #0x3f
-	beq	.Ldone
-
-.Ltail63aligned:			/* Count in tmp2.  */
-	and	tmp1, tmp2, #0x38
-	add	dst, dst, tmp1
-	add	src, src, tmp1
-	rsb	tmp1, tmp1, #(56 - PC_OFFSET + INSN_SIZE)
-	add	pc, pc, tmp1
-
-	vldr	d0, [src, #-56]	/* 14 words to go.  */
-	vstr	d0, [dst, #-56]
-	vldr	d0, [src, #-48]	/* 12 words to go.  */
-	vstr	d0, [dst, #-48]
-	vldr	d0, [src, #-40]	/* 10 words to go.  */
-	vstr	d0, [dst, #-40]
-	vldr	d0, [src, #-32]	/* 8 words to go.  */
-	vstr	d0, [dst, #-32]
-	vldr	d0, [src, #-24]	/* 6 words to go.  */
-	vstr	d0, [dst, #-24]
-	vldr	d0, [src, #-16]	/* 4 words to go.  */
-	vstr	d0, [dst, #-16]
-	vldr	d0, [src, #-8]	/* 2 words to go.  */
-	vstr	d0, [dst, #-8]
-#else
-	sub	src, src, #8
-	sub	dst, dst, #8
-1:
-	ldrd	A_l, A_h, [src, #8]
-	strd	A_l, A_h, [dst, #8]
-	ldrd	A_l, A_h, [src, #16]
-	strd	A_l, A_h, [dst, #16]
-	ldrd	A_l, A_h, [src, #24]
-	strd	A_l, A_h, [dst, #24]
-	ldrd	A_l, A_h, [src, #32]
-	strd	A_l, A_h, [dst, #32]
-	ldrd	A_l, A_h, [src, #40]
-	strd	A_l, A_h, [dst, #40]
-	ldrd	A_l, A_h, [src, #48]
-	strd	A_l, A_h, [dst, #48]
-	ldrd	A_l, A_h, [src, #56]
-	strd	A_l, A_h, [dst, #56]
-	ldrd	A_l, A_h, [src, #64]!
-	strd	A_l, A_h, [dst, #64]!
-	subs	tmp2, tmp2, #64
-	bge	1b
-	tst	tmp2, #0x3f
-	bne	1f
-	ldr	tmp2,[sp], #FRAME_SIZE
-	bx	lr
-1:
-	add	src, src, #8
-	add	dst, dst, #8
-
-.Ltail63aligned:			/* Count in tmp2.  */
-	/* Copy up to 7 d-words of data.  Similar to Ltail63unaligned, but
-	   we know that the src and dest are 32-bit aligned so we can use
-	   LDRD/STRD to improve efficiency.  */
-	/* TMP2 is now negative, but we don't care about that.  The bottom
-	   six bits still tell us how many bytes are left to copy.  */
-
-	and	tmp1, tmp2, #0x38
-	add	dst, dst, tmp1
-	add	src, src, tmp1
-	rsb	tmp1, tmp1, #(56 - PC_OFFSET + INSN_SIZE)
-	add	pc, pc, tmp1
-	ldrd	A_l, A_h, [src, #-56]	/* 14 words to go.  */
-	strd	A_l, A_h, [dst, #-56]
-	ldrd	A_l, A_h, [src, #-48]	/* 12 words to go.  */
-	strd	A_l, A_h, [dst, #-48]
-	ldrd	A_l, A_h, [src, #-40]	/* 10 words to go.  */
-	strd	A_l, A_h, [dst, #-40]
-	ldrd	A_l, A_h, [src, #-32]	/* 8 words to go.  */
-	strd	A_l, A_h, [dst, #-32]
-	ldrd	A_l, A_h, [src, #-24]	/* 6 words to go.  */
-	strd	A_l, A_h, [dst, #-24]
-	ldrd	A_l, A_h, [src, #-16]	/* 4 words to go.  */
-	strd	A_l, A_h, [dst, #-16]
-	ldrd	A_l, A_h, [src, #-8]	/* 2 words to go.  */
-	strd	A_l, A_h, [dst, #-8]
-
-#endif
-	tst	tmp2, #4
-	ldrne	tmp1, [src], #4
-	strne	tmp1, [dst], #4
-	lsls	tmp2, tmp2, #31		/* Count (tmp2) now dead. */
-	ldrhcs	tmp1, [src], #2
-	ldrbne	tmp2, [src]
-	strhcs	tmp1, [dst], #2
-	strbne	tmp2, [dst]
-
-.Ldone:
-	ldr	tmp2, [sp], #FRAME_SIZE
-	bx	lr
-
-.Lcpy_body_long:			/* Count in tmp2.  */
-
-	/* Long copy.  We know that there's at least (prefetch_lines * 64)
-	   bytes to go.  */
-#ifdef USE_VFP
-	/* Don't use PLD.  Instead, read some data in advance of the current
-	   copy position into a register.  This should act like a PLD
-	   operation but we won't have to repeat the transfer.  */
-
-	vldr	d3, [src, #0]
-	vldr	d4, [src, #64]
-	vldr	d5, [src, #128]
-	vldr	d6, [src, #192]
-	vldr	d7, [src, #256]
-
-	vldr	d0, [src, #8]
-	vldr	d1, [src, #16]
-	vldr	d2, [src, #24]
-	add	src, src, #32
-
-	subs	tmp2, tmp2, #prefetch_lines * 64 * 2
-	blt	2f
-1:
-	cpy_line_vfp	d3, 0
-	cpy_line_vfp	d4, 64
-	cpy_line_vfp	d5, 128
-	add	dst, dst, #3 * 64
-	add	src, src, #3 * 64
-	cpy_line_vfp	d6, 0
-	cpy_line_vfp	d7, 64
-	add	dst, dst, #2 * 64
-	add	src, src, #2 * 64
-	subs	tmp2, tmp2, #prefetch_lines * 64
-	bge	1b
-
-2:
-	cpy_tail_vfp	d3, 0
-	cpy_tail_vfp	d4, 64
-	cpy_tail_vfp	d5, 128
-	add	src, src, #3 * 64
-	add	dst, dst, #3 * 64
-	cpy_tail_vfp	d6, 0
-	vstr	d7, [dst, #64]
-	vldr	d7, [src, #64]
-	vstr	d0, [dst, #64 + 8]
-	vldr	d0, [src, #64 + 8]
-	vstr	d1, [dst, #64 + 16]
-	vldr	d1, [src, #64 + 16]
-	vstr	d2, [dst, #64 + 24]
-	vldr	d2, [src, #64 + 24]
-	vstr	d7, [dst, #64 + 32]
-	add	src, src, #96
-	vstr	d0, [dst, #64 + 40]
-	vstr	d1, [dst, #64 + 48]
-	vstr	d2, [dst, #64 + 56]
-	add	dst, dst, #128
-	add	tmp2, tmp2, #prefetch_lines * 64
-	b	.Lcpy_body_medium
-#else
-	/* Long copy.  Use an SMS style loop to maximize the I/O
-	   bandwidth of the core.  We don't have enough spare registers
-	   to synthesise prefetching, so use PLD operations.  */
-	/* Pre-bias src and dst.  */
-	sub	src, src, #8
-	sub	dst, dst, #8
-	pld	[src, #8]
-	pld	[src, #72]
-	subs	tmp2, tmp2, #64
-	pld	[src, #136]
-	ldrd	A_l, A_h, [src, #8]
-	strd	B_l, B_h, [sp, #8]
-	ldrd	B_l, B_h, [src, #16]
-	strd	C_l, C_h, [sp, #16]
-	ldrd	C_l, C_h, [src, #24]
-	strd	D_l, D_h, [sp, #24]
-	pld	[src, #200]
-	ldrd	D_l, D_h, [src, #32]!
-	b	1f
-	.p2align	6
-2:
-	pld	[src, #232]
-	strd	A_l, A_h, [dst, #40]
-	ldrd	A_l, A_h, [src, #40]
-	strd	B_l, B_h, [dst, #48]
-	ldrd	B_l, B_h, [src, #48]
-	strd	C_l, C_h, [dst, #56]
-	ldrd	C_l, C_h, [src, #56]
-	strd	D_l, D_h, [dst, #64]!
-	ldrd	D_l, D_h, [src, #64]!
-	subs	tmp2, tmp2, #64
-1:
-	strd	A_l, A_h, [dst, #8]
-	ldrd	A_l, A_h, [src, #8]
-	strd	B_l, B_h, [dst, #16]
-	ldrd	B_l, B_h, [src, #16]
-	strd	C_l, C_h, [dst, #24]
-	ldrd	C_l, C_h, [src, #24]
-	strd	D_l, D_h, [dst, #32]
-	ldrd	D_l, D_h, [src, #32]
-	bcs	2b
-	/* Save the remaining bytes and restore the callee-saved regs.  */
-	strd	A_l, A_h, [dst, #40]
-	add	src, src, #40
-	strd	B_l, B_h, [dst, #48]
-	ldrd	B_l, B_h, [sp, #8]
-	strd	C_l, C_h, [dst, #56]
-	ldrd	C_l, C_h, [sp, #16]
-	strd	D_l, D_h, [dst, #64]
-	ldrd	D_l, D_h, [sp, #24]
-	add	dst, dst, #72
-	tst	tmp2, #0x3f
-	bne	.Ltail63aligned
-	ldr	tmp2, [sp], #FRAME_SIZE
-	bx	lr
-#endif
-
-.Lcpy_notaligned:
-	pld	[src]
-	pld	[src, #64]
-	/* There's at least 64 bytes to copy, but there is no mutual
-	   alignment.  */
-	/* Bring DST to 64-bit alignment.  */
-	lsls	tmp2, dst, #29
-	pld	[src, #(2 * 64)]
-	beq	1f
-	rsbs	tmp2, tmp2, #0
-	sub	count, count, tmp2, lsr #29
-	ldrmi	tmp1, [src], #4
-	strmi	tmp1, [dst], #4
-	lsls	tmp2, tmp2, #2
-	ldrbne	tmp1, [src], #1
-	ldrhcs	tmp2, [src], #2
-	strbne	tmp1, [dst], #1
-	strhcs	tmp2, [dst], #2
-1:
-	pld	[src, #(3 * 64)]
-	subs	count, count, #64
-	ldrmi	tmp2, [sp], #FRAME_SIZE
-	bmi	.Ltail63unaligned
-	pld	[src, #(4 * 64)]
-
-#ifdef USE_NEON
-	vld1.8	{d0-d3}, [src]!
-	vld1.8	{d4-d7}, [src]!
-	subs	count, count, #64
-	bmi	2f
-1:
-	pld	[src, #(4 * 64)]
-	vst1.8	{d0-d3}, [ALIGN (dst, 64)]!
-	vld1.8	{d0-d3}, [src]!
-	vst1.8	{d4-d7}, [ALIGN (dst, 64)]!
-	vld1.8	{d4-d7}, [src]!
-	subs	count, count, #64
-	bpl	1b
-2:
-	vst1.8	{d0-d3}, [ALIGN (dst, 64)]!
-	vst1.8	{d4-d7}, [ALIGN (dst, 64)]!
-	ands	count, count, #0x3f
-#else
-	/* Use an SMS style loop to maximize the I/O bandwidth.  */
-	sub	src, src, #4
-	sub	dst, dst, #8
-	subs	tmp2, count, #64	/* Use tmp2 for count.  */
-	ldr	A_l, [src, #4]
-	ldr	A_h, [src, #8]
-	strd	B_l, B_h, [sp, #8]
-	ldr	B_l, [src, #12]
-	ldr	B_h, [src, #16]
-	strd	C_l, C_h, [sp, #16]
-	ldr	C_l, [src, #20]
-	ldr	C_h, [src, #24]
-	strd	D_l, D_h, [sp, #24]
-	ldr	D_l, [src, #28]
-	ldr	D_h, [src, #32]!
-	b	1f
-	.p2align	6
-2:
-	pld	[src, #(5 * 64) - (32 - 4)]
-	strd	A_l, A_h, [dst, #40]
-	ldr	A_l, [src, #36]
-	ldr	A_h, [src, #40]
-	strd	B_l, B_h, [dst, #48]
-	ldr	B_l, [src, #44]
-	ldr	B_h, [src, #48]
-	strd	C_l, C_h, [dst, #56]
-	ldr	C_l, [src, #52]
-	ldr	C_h, [src, #56]
-	strd	D_l, D_h, [dst, #64]!
-	ldr	D_l, [src, #60]
-	ldr	D_h, [src, #64]!
-	subs	tmp2, tmp2, #64
-1:
-	strd	A_l, A_h, [dst, #8]
-	ldr	A_l, [src, #4]
-	ldr	A_h, [src, #8]
-	strd	B_l, B_h, [dst, #16]
-	ldr	B_l, [src, #12]
-	ldr	B_h, [src, #16]
-	strd	C_l, C_h, [dst, #24]
-	ldr	C_l, [src, #20]
-	ldr	C_h, [src, #24]
-	strd	D_l, D_h, [dst, #32]
-	ldr	D_l, [src, #28]
-	ldr	D_h, [src, #32]
-	bcs	2b
-
-	/* Save the remaining bytes and restore the callee-saved regs.  */
-	strd	A_l, A_h, [dst, #40]
-	add	src, src, #36
-	strd	B_l, B_h, [dst, #48]
-	ldrd	B_l, B_h, [sp, #8]
-	strd	C_l, C_h, [dst, #56]
-	ldrd	C_l, C_h, [sp, #16]
-	strd	D_l, D_h, [dst, #64]
-	ldrd	D_l, D_h, [sp, #24]
-	add	dst, dst, #72
-	ands	count, tmp2, #0x3f
-#endif
-	ldr	tmp2, [sp], #FRAME_SIZE
-	bne	.Ltail63unaligned
-	bx	lr
-END(memcpy)

libc/arch-arm/bionic/setjmp.S

@@ -169,7 +169,5 @@ ENTRY(siglongjmp)
   bx lr
 END(siglongjmp)
 
-  .globl longjmp
-  .equ longjmp, siglongjmp
-  .globl _longjmp
-  .equ _longjmp, siglongjmp
+ALIAS_SYMBOL(longjmp, siglongjmp)
+ALIAS_SYMBOL(_longjmp, siglongjmp)

libc/arch-arm/bionic/strcmp.S

@@ -1,317 +0,0 @@
-/*
- * Copyright (c) 2011 The Android Open Source Project
- * Copyright (c) 2008 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 <machine/cpu-features.h>
-#include <private/bionic_asm.h>
-
-	.text
-
-#ifdef __ARMEB__
-#define SHFT2LSB lsl
-#define SHFT2LSBEQ lsleq
-#define SHFT2MSB lsr
-#define SHFT2MSBEQ lsreq
-#define MSB 0x000000ff
-#define LSB 0xff000000
-#else
-#define SHFT2LSB lsr
-#define SHFT2LSBEQ lsreq
-#define SHFT2MSB lsl
-#define SHFT2MSBEQ lsleq
-#define MSB 0xff000000
-#define LSB 0x000000ff
-#endif
-
-#define magic1(REG) REG
-#define magic2(REG) REG, lsl #7
-
-ENTRY(strcmp)
-	pld	[r0, #0]
-	pld	[r1, #0]
-	eor	r2, r0, r1
-	tst	r2, #3
-
-	/* Strings not at same byte offset from a word boundary.  */
-	bne	.Lstrcmp_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
-	SHFT2LSB	r2, r3, r2
-	ldr	r3, [r1], #4
-	orr	ip, ip, r2
-	orr	r3, r3, r2
-1:
-	/* Load the 'magic' constant 0x01010101. */
-	str	r4, [sp, #-4]!
-	mov	r4, #1
-	orr	r4, r4, r4, lsl #8
-	orr	r4, r4, r4, lsl #16
-	.p2align	2
-4:
-	pld	[r0, #8]
-	pld	[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 */
-	SHFT2MSB	r0, ip, #24
-	SHFT2LSB	ip, ip, #8
-	cmp	r0, #1
-	it	cs
-	cmpcs	r0, r3, SHFT2MSB #24
-	it	eq
-	SHFT2LSBEQ 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
-	/* No RSB instruction in Thumb2 */
-#ifdef __thumb2__
-	lsr	r0, r0, #24
-	sub	r0, r0, r3
-#else
-	rsb	r0, r3, r0, lsr #24
-#endif
-#endif
-	ldr	r4, [sp], #4
-	bx	lr
-
-.Lstrcmp_unaligned:
-	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
-	bx	lr
-
-2:
-	str	r5, [sp, #-4]!
-	str	r4, [sp, #-4]!
-	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, SHFT2LSB #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, SHFT2MSB #24
-	bne	6f
-	ldr	w1, [wp1], #4
-	b	1b
-4:
-	SHFT2LSB	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]
-	SHFT2LSB	t1, w1, #24
-#ifdef __ARMEB__
-	lsl	w2, w2, #24
-#endif
-	b	8f
-
-6:
-	SHFT2LSB	t1, w1, #24
-	and	w2, w2, #LSB
-	b	8f
-
-	/* Critical inner Loop: Block with 2 bytes initial overlap */
-	.p2align	2
-2:
-	SHFT2MSB	t1, w1, #16
-	sub	r3, w1, b1
-	SHFT2LSB	t1, t1, #16
-	bic	r3, r3, w1
-	cmp	t1, w2, SHFT2LSB #16
-	bne	4f
-	ands	r3, r3, b1, lsl #7
-	it	eq
-	ldreq	w2, [wp2], #4
-	bne	5f
-	eor	t1, t1, w1
-	cmp	t1, w2, SHFT2MSB #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]
-	SHFT2LSB	t1, w1, #16
-#ifdef __ARMEB__
-	lsl	w2, w2, #16
-#endif
-	b	8f
-
-6:
-	SHFT2MSB	w2, w2, #16
-	SHFT2LSB	t1, w1, #16
-4:
-	SHFT2LSB	w2, w2, #16
-	b	8f
-
-	/* Critical inner Loop: Block with 1 byte initial overlap */
-	.p2align	2
-3:
-	and	t1, w1, #LSB
-	cmp	t1, w2, SHFT2LSB #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, SHFT2MSB #8
-	bne	6f
-	ldr	w1, [wp1], #4
-	b	3b
-4:
-	SHFT2LSB	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:
-	SHFT2LSB	t1, w1, #8
-	bic	w2, w2, #MSB
-	b	8f
-7:
-	mov	r0, #0
-	ldr	r4, [sp], #4
-	ldr	r5, [sp], #4
-	bx	lr
-
-8:
-	and	r2, t1, #LSB
-	and	r0, w2, #LSB
-	cmp	r0, #1
-	it	cs
-	cmpcs	r0, r2
-	itt	eq
-	SHFT2LSBEQ	t1, t1, #8
-	SHFT2LSBEQ	w2, w2, #8
-	beq	8b
-	sub	r0, r2, r0
-	ldr	r4, [sp], #4
-	ldr	r5, [sp], #4
-	bx	lr
-END(strcmp)

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

@@ -8,4 +8,9 @@ libc_bionic_src_files_arm += \
     arch-arm/cortex-a15/bionic/strcpy.S \
     arch-arm/cortex-a15/bionic/__strcpy_chk.S \
     arch-arm/cortex-a15/bionic/strlen.S \
-    bionic/memmove.c \
+
+libc_bionic_src_files_arm += \
+    arch-arm/generic/bionic/memcmp.S \
+
+libc_bionic_src_files_arm += \
+    arch-arm/denver/bionic/memmove.S \

libc/arch-arm/cortex-a53/cortex-a53.mk

@@ -0,0 +1 @@
+include bionic/libc/arch-arm/cortex-a7/cortex-a7.mk

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

@@ -35,6 +35,7 @@
  */
 
     .fpu    neon
+    .syntax unified
 
 ENTRY(__memset_chk)
         cmp         r2, r3
@@ -100,9 +101,9 @@ ENTRY(memset)
 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
+        strbmi      r1, [r0], #1
+        strbcs      r1, [r0], #1
+        strbcs      r1, [r0], #1
         ldmfd       sp!, {r0}
         bx          lr
 END(memset)
@@ -131,11 +132,11 @@ ENTRY_PRIVATE(__memset_large_copy)
         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
+        strbcs      r1, [r0], #1    /* can't use strh (alignment unknown) */
+        strbcs      r1, [r0], #1
+        strbmi      r1, [r0], #1
         subs        r2, r2, r3
-        ldmlsfd     sp!, {r0, r4-r7, lr}   /* return */
+        popls       {r0, r4-r7, lr}   /* return */
         bxls        lr
 
         /* align the destination to a cache-line */
@@ -155,9 +156,9 @@ ENTRY_PRIVATE(__memset_large_copy)
 
         /* 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}
+        stmcs       r0!, {r1, lr}
+        stmcs       r0!, {r1, lr}
+        stmmi       r0!, {r1, lr}
         movs        r3, r3, lsl #2
         strcs       r1, [r0], #4
 
@@ -172,13 +173,13 @@ ENTRY_PRIVATE(__memset_large_copy)
 
         /* conditionally stores 0 to 31 bytes */
         movs        r2, r2, lsl #28
-        stmcsia     r0!, {r1,r3,r12,lr}
-        stmmiia     r0!, {r1, lr}
+        stmcs       r0!, {r1,r3,r12,lr}
+        stmmi       r0!, {r1, lr}
         movs        r2, r2, lsl #2
         strcs       r1, [r0], #4
-        strmih      r1, [r0], #2
+        strhmi      r1, [r0], #2
         movs        r2, r2, lsl #2
-        strcsb      r1, [r0]
+        strbcs      r1, [r0]
         ldmfd       sp!, {r0, r4-r7, lr}
         bx          lr
 END(__memset_large_copy)

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

@@ -8,4 +8,9 @@ libc_bionic_src_files_arm += \
     arch-arm/cortex-a9/bionic/strcpy.S \
     arch-arm/cortex-a9/bionic/__strcpy_chk.S \
     arch-arm/cortex-a9/bionic/strlen.S \
-    bionic/memmove.c \
+
+libc_bionic_src_files_arm += \
+    arch-arm/generic/bionic/memcmp.S \
+
+libc_bionic_src_files_arm += \
+    arch-arm/denver/bionic/memmove.S \

libc/arch-arm/denver/denver.mk

@@ -1,4 +1,5 @@
 libc_bionic_src_files_arm += \
+    arch-arm/generic/bionic/memcmp.S \
     arch-arm/denver/bionic/memcpy.S \
     arch-arm/denver/bionic/memmove.S \
     arch-arm/denver/bionic/memset.S \

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

@@ -40,6 +40,8 @@
  * Optimized memcmp() for Cortex-A9.
  */
 
+.syntax unified
+
 ENTRY(memcmp)
         pld         [r0, #(CACHE_LINE_SIZE * 0)]
         pld         [r0, #(CACHE_LINE_SIZE * 1)]
@@ -161,25 +163,25 @@ ENTRY(memcmp)
         eors        r0, r0, ip
         ldreq       r0, [r4], #4
         ldreq       ip, [r1, #4]!
-        eoreqs      r0, r0, lr
+        eorseq      r0, r0, lr
         ldreq       r0, [r4], #4
         ldreq       lr, [r1, #4]!
-        eoreqs      r0, r0, ip
+        eorseq      r0, r0, ip
         ldreq       r0, [r4], #4
         ldreq       ip, [r1, #4]!
-        eoreqs      r0, r0, lr
+        eorseq      r0, r0, lr
         ldreq       r0, [r4], #4
         ldreq       lr, [r1, #4]!
-        eoreqs      r0, r0, ip
+        eorseq      r0, r0, ip
         ldreq       r0, [r4], #4
         ldreq       ip, [r1, #4]!
-        eoreqs      r0, r0, lr
+        eorseq      r0, r0, lr
         ldreq       r0, [r4], #4
         ldreq       lr, [r1, #4]!
-        eoreqs      r0, r0, ip
+        eorseq      r0, r0, ip
         ldreq       r0, [r4], #4
         ldreq       ip, [r1, #4]!
-        eoreqs      r0, r0, lr
+        eorseq      r0, r0, lr
         bne         2f
         subs        r2, r2, #32
         bhs         0b
@@ -263,17 +265,17 @@ ENTRY(memcmp)
         ldreq       lr, [r1], #4
         ldreq       r0, [r4], #4
         orreq       ip, ip, lr, lsl #16
-        eoreqs      r0, r0, ip
+        eorseq      r0, r0, ip
         moveq       ip, lr, lsr #16
         ldreq       lr, [r1], #4
         ldreq       r0, [r4], #4
         orreq       ip, ip, lr, lsl #16
-        eoreqs      r0, r0, ip
+        eorseq      r0, r0, ip
         moveq       ip, lr, lsr #16
         ldreq       lr, [r1], #4
         ldreq       r0, [r4], #4
         orreq       ip, ip, lr, lsl #16
-        eoreqs      r0, r0, ip
+        eorseq      r0, r0, ip
         bne         7f
         subs        r2, r2, #16
         bhs         6b
@@ -317,7 +319,7 @@ ENTRY(memcmp)
         ldreq       r7, [r1], #4
         ldreq       r0, [r4], #4
         orreq       ip, ip, r7, lsl r6
-        eoreqs      r0, r0, ip
+        eorseq      r0, r0, ip
         bne         7f
         subs        r2, r2, #8
         bhs         6b

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

@@ -37,6 +37,8 @@
          * so we have to preserve R0.
          */
 
+         .syntax unified
+
 ENTRY(__memcpy_chk)
         cmp         r2, r3
         bhi         __memcpy_chk_fail
@@ -81,12 +83,12 @@ ENTRY(memcpy)
          */
         movs        r12, r3, lsl #31
         sub         r2, r2, r3      /* we know that r3 <= r2 because r2 >= 4 */
-        ldrmib      r3, [r1], #1
-        ldrcsb      r4, [r1], #1
-        ldrcsb      r12,[r1], #1
-        strmib      r3, [r0], #1
-        strcsb      r4, [r0], #1
-        strcsb      r12,[r0], #1
+        ldrbmi      r3, [r1], #1
+        ldrbcs      r4, [r1], #1
+        ldrbcs      r12,[r1], #1
+        strbmi      r3, [r0], #1
+        strbcs      r4, [r0], #1
+        strbcs      r12,[r0], #1
 
 .Lsrc_aligned:
 
@@ -109,10 +111,10 @@ ENTRY(memcpy)
 
         /* conditionally copies 0 to 7 words (length in r3) */
         movs        r12, r3, lsl #28
-        ldmcsia     r1!, {r4, r5, r6, r7}   /* 16 bytes */
-        ldmmiia     r1!, {r8, r9}           /*  8 bytes */
-        stmcsia     r0!, {r4, r5, r6, r7}
-        stmmiia     r0!, {r8, r9}
+        ldmcs       r1!, {r4, r5, r6, r7}   /* 16 bytes */
+        ldmmi       r1!, {r8, r9}           /*  8 bytes */
+        stmcs       r0!, {r4, r5, r6, r7}
+        stmmi       r0!, {r8, r9}
         tst         r3, #0x4
         ldrne       r10,[r1], #4            /*  4 bytes */
         strne       r10,[r0], #4
@@ -177,18 +179,18 @@ ENTRY(memcpy)
 
         /* conditionnaly copies 0 to 31 bytes */
         movs        r12, r2, lsl #28
-        ldmcsia     r1!, {r4, r5, r6, r7}   /* 16 bytes */
-        ldmmiia     r1!, {r8, r9}           /*  8 bytes */
-        stmcsia     r0!, {r4, r5, r6, r7}
-        stmmiia     r0!, {r8, r9}
+        ldmcs       r1!, {r4, r5, r6, r7}   /* 16 bytes */
+        ldmmi       r1!, {r8, r9}           /*  8 bytes */
+        stmcs       r0!, {r4, r5, r6, r7}
+        stmmi       r0!, {r8, r9}
         movs        r12, r2, lsl #30
         ldrcs       r3, [r1], #4            /*  4 bytes */
-        ldrmih      r4, [r1], #2            /*  2 bytes */
+        ldrhmi      r4, [r1], #2            /*  2 bytes */
         strcs       r3, [r0], #4
-        strmih      r4, [r0], #2
+        strhmi      r4, [r0], #2
         tst         r2, #0x1
-        ldrneb      r3, [r1]                /*  last byte  */
-        strneb      r3, [r0]
+        ldrbne      r3, [r1]                /*  last byte  */
+        strbne      r3, [r0]
 
         /* we're done! restore everything and return */
 1:      ldmfd       sp!, {r5-r11}
@@ -228,11 +230,11 @@ ENTRY(memcpy)
          * becomes aligned to 32 bits (r5 = nb of words to copy for alignment)
          */
         movs        r5, r5, lsl #31
-        strmib      r3, [r0], #1
+        strbmi      r3, [r0], #1
         movmi       r3, r3, lsr #8
-        strcsb      r3, [r0], #1
+        strbcs      r3, [r0], #1
         movcs       r3, r3, lsr #8
-        strcsb      r3, [r0], #1
+        strbcs      r3, [r0], #1
         movcs       r3, r3, lsr #8
 
         cmp         r2, #4
@@ -363,23 +365,23 @@ ENTRY(memcpy)
 .Lpartial_word_tail:
         /* we have a partial word in the input buffer */
         movs        r5, lr, lsl #(31-3)
-        strmib      r3, [r0], #1
+        strbmi      r3, [r0], #1
         movmi       r3, r3, lsr #8
-        strcsb      r3, [r0], #1
+        strbcs      r3, [r0], #1
         movcs       r3, r3, lsr #8
-        strcsb      r3, [r0], #1
+        strbcs      r3, [r0], #1
 
         /* Refill spilled registers from the stack. Don't update sp. */
         ldmfd       sp, {r5-r11}
 
 .Lcopy_last_3_and_return:
         movs        r2, r2, lsl #31 /* copy remaining 0, 1, 2 or 3 bytes */
-        ldrmib      r2, [r1], #1
-        ldrcsb      r3, [r1], #1
-        ldrcsb      r12,[r1]
-        strmib      r2, [r0], #1
-        strcsb      r3, [r0], #1
-        strcsb      r12,[r0]
+        ldrbmi      r2, [r1], #1
+        ldrbcs      r3, [r1], #1
+        ldrbcs      r12,[r1]
+        strbmi      r2, [r0], #1
+        strbcs      r3, [r0], #1
+        strbcs      r12,[r0]
 
         /* we're done! restore sp and spilled registers and return */
         add         sp,  sp, #28

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

@@ -35,6 +35,8 @@
          * memset() returns its first argument.
          */
 
+         .syntax unified
+
 ENTRY(__memset_chk)
         cmp         r2, r3
         bls         done
@@ -76,11 +78,11 @@ ENTRY(memset)
         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
+        strbcs      r1, [r0], #1    /* can't use strh (alignment unknown) */
+        strbcs      r1, [r0], #1
+        strbmi      r1, [r0], #1
         subs        r2, r2, r3
-        ldmlsfd     sp!, {r0, r4-r7, lr}    /* return */
+        popls       {r0, r4-r7, lr}    /* return */
         bxls        lr
 
         /* align the destination to a cache-line */
@@ -100,9 +102,9 @@ ENTRY(memset)
 
         /* 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}
+        stmcs       r0!, {r1, lr}
+        stmcs       r0!, {r1, lr}
+        stmmi       r0!, {r1, lr}
         movs        r3, r3, lsl #2
         strcs       r1, [r0], #4
 
@@ -117,13 +119,13 @@ ENTRY(memset)
 
         /* conditionally stores 0 to 31 bytes */
         movs        r2, r2, lsl #28
-        stmcsia     r0!, {r1,r3,r12,lr}
-        stmmiia     r0!, {r1, lr}
+        stmcs       r0!, {r1,r3,r12,lr}
+        stmmi       r0!, {r1, lr}
         movs        r2, r2, lsl #2
         strcs       r1, [r0], #4
-        strmih      r1, [r0], #2
+        strhmi      r1, [r0], #2
         movs        r2, r2, lsl #2
-        strcsb      r1, [r0]
+        strbcs      r1, [r0]
         ldmfd       sp!, {r0, r4-r7, lr}
         bx          lr
 END(memset)

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

@@ -32,6 +32,8 @@
 #include <machine/cpu-features.h>
 #include <private/bionic_asm.h>
 
+.syntax unified
+
 ENTRY(strcpy)
 	pld	[r1, #0]
 	eor	r2, r0, r1
@@ -108,15 +110,15 @@ ENTRY(strcpy)
 #ifdef __ARMEB__
 	tst	r2, #0xff00
 	iteet	ne
-	strneh	r2, [ip], #2
+	strhne	r2, [ip], #2
 	lsreq	r2, r2, #8
-	streqb	r2, [ip]
+	strbeq	r2, [ip]
 	tstne	r2, #0xff
 #else
 	tst	r2, #0xff
 	itet	ne
-	strneh	r2, [ip], #2
-	streqb	r2, [ip]
+	strhne	r2, [ip], #2
+	strbeq	r2, [ip]
 	tstne	r2, #0xff00
 #endif
 	bne	5b

libc/arch-arm/generic/generic.mk

@@ -1,11 +1,14 @@
 libc_bionic_src_files_arm += \
+    arch-arm/generic/bionic/memcmp.S \
     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/memmove.c \
     bionic/__strcat_chk.cpp \
     bionic/__strcpy_chk.cpp \
+
+libc_openbsd_src_files_arm += \
+    upstream-openbsd/lib/libc/string/memmove.c \
     upstream-openbsd/lib/libc/string/stpcpy.c \
     upstream-openbsd/lib/libc/string/strcat.c \

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

@@ -102,6 +102,8 @@
 
 /* 112-127 are reserved for private experiments. */
 
+#define R_ARM_IRELATIVE   160
+
 #define R_ARM_RXPC25		249
 #define R_ARM_RSBREL32		250
 #define R_ARM_THM_RPC22		251

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

@@ -37,6 +37,7 @@
  */
 
     .fpu    neon
+    .syntax unified
 
 ENTRY(__memset_chk)
         cmp         r2, r3
@@ -98,9 +99,9 @@ ENTRY(memset)
 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
+        strbmi      r1, [r0], #1
+        strbcs      r1, [r0], #1
+        strbcs      r1, [r0], #1
         ldmfd       sp!, {r0}
         bx          lr
 END(memset)

libc/arch-arm/krait/krait.mk

@@ -11,4 +11,9 @@ 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 \
-    bionic/memmove.c \
+
+libc_bionic_src_files_arm += \
+    arch-arm/generic/bionic/memcmp.S \
+
+libc_bionic_src_files_arm += \
+    arch-arm/denver/bionic/memmove.S \

libc/arch-arm/syscalls/___clock_nanosleep.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(__clock_nanosleep)
+ENTRY(___clock_nanosleep)
     mov     ip, r7
     ldr     r7, =__NR_clock_nanosleep
     swi     #0
@@ -11,4 +11,5 @@ ENTRY(__clock_nanosleep)
     bxls    lr
     neg     r0, r0
     b       __set_errno_internal
-END(__clock_nanosleep)
+END(___clock_nanosleep)
+.hidden ___clock_nanosleep

libc/arch-arm/syscalls/___close.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(close)
+ENTRY(___close)
     mov     ip, r7
     ldr     r7, =__NR_close
     swi     #0
@@ -11,4 +11,5 @@ ENTRY(close)
     bxls    lr
     neg     r0, r0
     b       __set_errno_internal
-END(close)
+END(___close)
+.hidden ___close

libc/arch-arm/syscalls/___faccessat.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(faccessat)
+ENTRY(___faccessat)
     mov     ip, r7
     ldr     r7, =__NR_faccessat
     swi     #0
@@ -11,4 +11,5 @@ ENTRY(faccessat)
     bxls    lr
     neg     r0, r0
     b       __set_errno_internal
-END(faccessat)
+END(___faccessat)
+.hidden ___faccessat

libc/arch-arm/syscalls/___fchmod.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(fchmod)
+ENTRY(___fchmod)
     mov     ip, r7
     ldr     r7, =__NR_fchmod
     swi     #0
@@ -11,4 +11,5 @@ ENTRY(fchmod)
     bxls    lr
     neg     r0, r0
     b       __set_errno_internal
-END(fchmod)
+END(___fchmod)
+.hidden ___fchmod

libc/arch-arm/syscalls/___fchmodat.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(fchmodat)
+ENTRY(___fchmodat)
     mov     ip, r7
     ldr     r7, =__NR_fchmodat
     swi     #0
@@ -11,4 +11,5 @@ ENTRY(fchmodat)
     bxls    lr
     neg     r0, r0
     b       __set_errno_internal
-END(fchmodat)
+END(___fchmodat)
+.hidden ___fchmodat

libc/arch-arm/syscalls/___fgetxattr.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(fgetxattr)
+ENTRY(___fgetxattr)
     mov     ip, r7
     ldr     r7, =__NR_fgetxattr
     swi     #0
@@ -11,4 +11,5 @@ ENTRY(fgetxattr)
     bxls    lr
     neg     r0, r0
     b       __set_errno_internal
-END(fgetxattr)
+END(___fgetxattr)
+.hidden ___fgetxattr

libc/arch-arm/syscalls/___fsetxattr.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(fsetxattr)
+ENTRY(___fsetxattr)
     mov     ip, sp
     stmfd   sp!, {r4, r5, r6, r7}
     .cfi_def_cfa_offset 16
@@ -19,4 +19,5 @@ ENTRY(fsetxattr)
     bxls    lr
     neg     r0, r0
     b       __set_errno_internal
-END(fsetxattr)
+END(___fsetxattr)
+.hidden ___fsetxattr

libc/arch-arm/syscalls/___rt_sigqueueinfo.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(__rt_sigqueueinfo)
+ENTRY(___rt_sigqueueinfo)
     mov     ip, r7
     ldr     r7, =__NR_rt_sigqueueinfo
     swi     #0
@@ -11,4 +11,5 @@ ENTRY(__rt_sigqueueinfo)
     bxls    lr
     neg     r0, r0
     b       __set_errno_internal
-END(__rt_sigqueueinfo)
+END(___rt_sigqueueinfo)
+.hidden ___rt_sigqueueinfo

libc/arch-arm/syscalls/_exit.S

@@ -13,5 +13,4 @@ ENTRY(_exit)
     b       __set_errno_internal
 END(_exit)
 
-    .globl _Exit
-    .equ _Exit, _exit
+ALIAS_SYMBOL(_Exit, _exit)

libc/arch-arm/syscalls/fstat64.S

@@ -13,5 +13,4 @@ ENTRY(fstat64)
     b       __set_errno_internal
 END(fstat64)
 
-    .globl fstat
-    .equ fstat, fstat64
+ALIAS_SYMBOL(fstat, fstat64)

libc/arch-arm/syscalls/fstatat64.S

@@ -13,5 +13,4 @@ ENTRY(fstatat64)
     b       __set_errno_internal
 END(fstatat64)
 
-    .globl fstatat
-    .equ fstatat, fstatat64
+ALIAS_SYMBOL(fstatat, fstatat64)

libc/arch-arm/syscalls/ftruncate.S

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

libc/arch-arm/syscalls/process_vm_readv.S

@@ -0,0 +1,22 @@
+/* Generated by gensyscalls.py. Do not edit. */
+
+#include <private/bionic_asm.h>
+
+ENTRY(process_vm_readv)
+    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_process_vm_readv
+    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_internal
+END(process_vm_readv)

libc/arch-arm/syscalls/process_vm_writev.S

@@ -0,0 +1,22 @@
+/* Generated by gensyscalls.py. Do not edit. */
+
+#include <private/bionic_asm.h>
+
+ENTRY(process_vm_writev)
+    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_process_vm_writev
+    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_internal
+END(process_vm_writev)

libc/arch-arm64/arm64.mk

@@ -8,7 +8,6 @@ libc_bionic_src_files_arm64 += \
     bionic/__memset_chk.cpp \
     bionic/__strcpy_chk.cpp \
     bionic/__strcat_chk.cpp \
-    bionic/memrchr.c \
     bionic/strrchr.cpp \
 
 libc_freebsd_src_files_arm64 += \
@@ -21,6 +20,7 @@ libc_freebsd_src_files_arm64 += \
     upstream-freebsd/lib/libc/string/wmemcmp.c \
 
 libc_openbsd_src_files_arm64 += \
+    upstream-openbsd/lib/libc/string/memrchr.c \
     upstream-openbsd/lib/libc/string/stpncpy.c \
     upstream-openbsd/lib/libc/string/strcat.c \
     upstream-openbsd/lib/libc/string/strlcat.c \
@@ -59,6 +59,6 @@ ifeq ($(wildcard $(cpu_variant_mk)),)
 $(error "TARGET_CPU_VARIANT not set or set to an unknown value. Possible values are generic, denver64. 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_variank_mk)
+libc_common_additional_dependencies += $(cpu_variant_mk)
 
 cpu_variant_mk :=

libc/arch-arm64/bionic/crtbegin.c

@@ -67,3 +67,4 @@ __asm__ (
 
 #include "../../arch-common/bionic/__dso_handle.h"
 #include "../../arch-common/bionic/atexit.h"
+#include "../../arch-common/bionic/pthread_atfork.h"

libc/arch-arm64/bionic/setjmp.S

@@ -49,8 +49,8 @@
 #define _JB_X19         (_JB_X20_X21 + 2)
 #define _JB_D14_D15     (_JB_X19 + 1)
 #define _JB_D12_D13     (_JB_D14_D15 + 2)
-#define _JB_D10_D11     (_JB_D12_D13 + 1)
-#define _JB_D8_D9       (_JB_D10_D11 + 1)
+#define _JB_D10_D11     (_JB_D12_D13 + 2)
+#define _JB_D8_D9       (_JB_D10_D11 + 2)
 
 ENTRY(setjmp)
   mov w1, #1
@@ -146,7 +146,5 @@ ENTRY(siglongjmp)
   ret
 END(siglongjmp)
 
-  .globl longjmp
-  .equ longjmp, siglongjmp
-  .globl _longjmp
-  .equ _longjmp, siglongjmp
+ALIAS_SYMBOL(longjmp, siglongjmp)
+ALIAS_SYMBOL(_longjmp, siglongjmp)

libc/arch-arm64/cortex-a53/cortex-a53.mk

@@ -0,0 +1 @@
+include bionic/libc/arch-arm64/generic/generic.mk

libc/arch-arm64/generic/bionic/memchr.S

@@ -101,7 +101,7 @@ ENTRY(memchr)
 	and	vhas_chr2.16b, vhas_chr2.16b, vrepmask.16b
 	addp	vend.16b, vhas_chr1.16b, vhas_chr2.16b		/* 256->128 */
 	addp	vend.16b, vend.16b, vend.16b			/* 128->64 */
-	mov	synd, vend.2d[0]
+	mov	synd, vend.d[0]
 	/* Clear the soff*2 lower bits */
 	lsl	tmp, soff, #1
 	lsr	synd, synd, tmp
@@ -121,7 +121,7 @@ ENTRY(memchr)
 	/* Use a fast check for the termination condition */
 	orr	vend.16b, vhas_chr1.16b, vhas_chr2.16b
 	addp	vend.2d, vend.2d, vend.2d
-	mov	synd, vend.2d[0]
+	mov	synd, vend.d[0]
 	/* We're not out of data, loop if we haven't found the character */
 	cbz	synd, .Lloop
 
@@ -131,7 +131,7 @@ ENTRY(memchr)
 	and	vhas_chr2.16b, vhas_chr2.16b, vrepmask.16b
 	addp	vend.16b, vhas_chr1.16b, vhas_chr2.16b		/* 256->128 */
 	addp	vend.16b, vend.16b, vend.16b			/* 128->64 */
-	mov	synd, vend.2d[0]
+	mov	synd, vend.d[0]
 	/* Only do the clear for the last possible block */
 	b.hi	.Ltail
 

libc/arch-arm64/generic/bionic/strchr.S

@@ -109,7 +109,7 @@ ENTRY(strchr)
 	addp	vend1.16b, vend1.16b, vend2.16b		// 128->64
 	lsr	tmp1, tmp3, tmp1
 
-	mov	tmp3, vend1.2d[0]
+	mov	tmp3, vend1.d[0]
 	bic	tmp1, tmp3, tmp1	// Mask padding bits.
 	cbnz	tmp1, .Ltail
 
@@ -124,7 +124,7 @@ ENTRY(strchr)
 	orr	vend2.16b, vhas_nul2.16b, vhas_chr2.16b
 	orr	vend1.16b, vend1.16b, vend2.16b
 	addp	vend1.2d, vend1.2d, vend1.2d
-	mov	tmp1, vend1.2d[0]
+	mov	tmp1, vend1.d[0]
 	cbz	tmp1, .Lloop
 
 	/* Termination condition found.  Now need to establish exactly why
@@ -138,7 +138,7 @@ ENTRY(strchr)
 	addp	vend1.16b, vend1.16b, vend2.16b		// 256->128
 	addp	vend1.16b, vend1.16b, vend2.16b		// 128->64
 
-	mov	tmp1, vend1.2d[0]
+	mov	tmp1, vend1.d[0]
 .Ltail:
 	/* Count the trailing zeros, by bit reversing...  */
 	rbit	tmp1, tmp1

libc/arch-arm64/syscalls/___clock_nanosleep.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(__clock_nanosleep)
+ENTRY(___clock_nanosleep)
     mov     x8, __NR_clock_nanosleep
     svc     #0
 
@@ -11,5 +11,5 @@ ENTRY(__clock_nanosleep)
     b.hi    __set_errno_internal
 
     ret
-END(__clock_nanosleep)
-.hidden __clock_nanosleep
+END(___clock_nanosleep)
+.hidden ___clock_nanosleep

libc/arch-arm64/syscalls/___close.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(close)
+ENTRY(___close)
     mov     x8, __NR_close
     svc     #0
 
@@ -11,4 +11,5 @@ ENTRY(close)
     b.hi    __set_errno_internal
 
     ret
-END(close)
+END(___close)
+.hidden ___close

libc/arch-arm64/syscalls/___faccessat.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(faccessat)
+ENTRY(___faccessat)
     mov     x8, __NR_faccessat
     svc     #0
 
@@ -11,4 +11,5 @@ ENTRY(faccessat)
     b.hi    __set_errno_internal
 
     ret
-END(faccessat)
+END(___faccessat)
+.hidden ___faccessat

libc/arch-arm64/syscalls/___fchmod.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(fchmod)
+ENTRY(___fchmod)
     mov     x8, __NR_fchmod
     svc     #0
 
@@ -11,4 +11,5 @@ ENTRY(fchmod)
     b.hi    __set_errno_internal
 
     ret
-END(fchmod)
+END(___fchmod)
+.hidden ___fchmod

libc/arch-arm64/syscalls/___fchmodat.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(fchmodat)
+ENTRY(___fchmodat)
     mov     x8, __NR_fchmodat
     svc     #0
 
@@ -11,4 +11,5 @@ ENTRY(fchmodat)
     b.hi    __set_errno_internal
 
     ret
-END(fchmodat)
+END(___fchmodat)
+.hidden ___fchmodat

libc/arch-arm64/syscalls/___fgetxattr.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(fgetxattr)
+ENTRY(___fgetxattr)
     mov     x8, __NR_fgetxattr
     svc     #0
 
@@ -11,4 +11,5 @@ ENTRY(fgetxattr)
     b.hi    __set_errno_internal
 
     ret
-END(fgetxattr)
+END(___fgetxattr)
+.hidden ___fgetxattr

libc/arch-arm64/syscalls/___fsetxattr.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(fsetxattr)
+ENTRY(___fsetxattr)
     mov     x8, __NR_fsetxattr
     svc     #0
 
@@ -11,4 +11,5 @@ ENTRY(fsetxattr)
     b.hi    __set_errno_internal
 
     ret
-END(fsetxattr)
+END(___fsetxattr)
+.hidden ___fsetxattr

libc/arch-arm64/syscalls/___rt_sigqueueinfo.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(__rt_sigqueueinfo)
+ENTRY(___rt_sigqueueinfo)
     mov     x8, __NR_rt_sigqueueinfo
     svc     #0
 
@@ -11,5 +11,5 @@ ENTRY(__rt_sigqueueinfo)
     b.hi    __set_errno_internal
 
     ret
-END(__rt_sigqueueinfo)
-.hidden __rt_sigqueueinfo
+END(___rt_sigqueueinfo)
+.hidden ___rt_sigqueueinfo

libc/arch-arm64/syscalls/__fstatfs.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(fstatfs64)
+ENTRY(__fstatfs)
     mov     x8, __NR_fstatfs
     svc     #0
 
@@ -11,7 +11,5 @@ ENTRY(fstatfs64)
     b.hi    __set_errno_internal
 
     ret
-END(fstatfs64)
-
-    .globl fstatfs
-    .equ fstatfs, fstatfs64
+END(__fstatfs)
+.hidden __fstatfs

libc/arch-arm64/syscalls/__statfs.S

@@ -2,7 +2,7 @@
 
 #include <private/bionic_asm.h>
 
-ENTRY(statfs64)
+ENTRY(__statfs)
     mov     x8, __NR_statfs
     svc     #0
 
@@ -11,7 +11,5 @@ ENTRY(statfs64)
     b.hi    __set_errno_internal
 
     ret
-END(statfs64)
-
-    .globl statfs
-    .equ statfs, statfs64
+END(__statfs)
+.hidden __statfs

libc/arch-arm64/syscalls/_exit.S

@@ -13,5 +13,4 @@ ENTRY(_exit)
     ret
 END(_exit)
 
-    .globl _Exit
-    .equ _Exit, _exit
+ALIAS_SYMBOL(_Exit, _exit)

libc/arch-arm64/syscalls/fallocate.S

@@ -13,5 +13,4 @@ ENTRY(fallocate)
     ret
 END(fallocate)
 
-    .globl fallocate64
-    .equ fallocate64, fallocate
+ALIAS_SYMBOL(fallocate64, fallocate)

libc/arch-arm64/syscalls/fstat64.S

@@ -13,5 +13,4 @@ ENTRY(fstat64)
     ret
 END(fstat64)
 
-    .globl fstat
-    .equ fstat, fstat64
+ALIAS_SYMBOL(fstat, fstat64)

libc/arch-arm64/syscalls/fstatat64.S

@@ -13,5 +13,4 @@ ENTRY(fstatat64)
     ret
 END(fstatat64)
 
-    .globl fstatat
-    .equ fstatat, fstatat64
+ALIAS_SYMBOL(fstatat, fstatat64)

libc/arch-arm64/syscalls/ftruncate.S

@@ -13,5 +13,4 @@ ENTRY(ftruncate)
     ret
 END(ftruncate)
 
-    .globl ftruncate64
-    .equ ftruncate64, ftruncate
+ALIAS_SYMBOL(ftruncate64, ftruncate)

libc/arch-arm64/syscalls/getrlimit.S

@@ -13,5 +13,4 @@ ENTRY(getrlimit)
     ret
 END(getrlimit)
 
-    .globl getrlimit64
-    .equ getrlimit64, getrlimit
+ALIAS_SYMBOL(getrlimit64, getrlimit)

libc/arch-arm64/syscalls/lseek.S

@@ -13,5 +13,4 @@ ENTRY(lseek)
     ret
 END(lseek)
 
-    .globl lseek64
-    .equ lseek64, lseek
+ALIAS_SYMBOL(lseek64, lseek)

libc/arch-arm64/syscalls/mmap.S

@@ -13,5 +13,4 @@ ENTRY(mmap)
     ret
 END(mmap)
 
-    .globl mmap64
-    .equ mmap64, mmap
+ALIAS_SYMBOL(mmap64, mmap)

libc/arch-arm64/syscalls/pread64.S

@@ -13,5 +13,4 @@ ENTRY(pread64)
     ret
 END(pread64)
 
-    .globl pread
-    .equ pread, pread64
+ALIAS_SYMBOL(pread, pread64)

libc/arch-arm64/syscalls/prlimit64.S

@@ -13,5 +13,4 @@ ENTRY(prlimit64)
     ret
 END(prlimit64)
 
-    .globl prlimit
-    .equ prlimit, prlimit64
+ALIAS_SYMBOL(prlimit, prlimit64)

libc/arch-arm64/syscalls/process_vm_readv.S

@@ -0,0 +1,14 @@
+/* Generated by gensyscalls.py. Do not edit. */
+
+#include <private/bionic_asm.h>
+
+ENTRY(process_vm_readv)
+    mov     x8, __NR_process_vm_readv
+    svc     #0
+
+    cmn     x0, #(MAX_ERRNO + 1)
+    cneg    x0, x0, hi
+    b.hi    __set_errno_internal
+
+    ret
+END(process_vm_readv)

libc/arch-arm64/syscalls/process_vm_writev.S

@@ -0,0 +1,14 @@
+/* Generated by gensyscalls.py. Do not edit. */
+
+#include <private/bionic_asm.h>
+
+ENTRY(process_vm_writev)
+    mov     x8, __NR_process_vm_writev
+    svc     #0
+
+    cmn     x0, #(MAX_ERRNO + 1)
+    cneg    x0, x0, hi
+    b.hi    __set_errno_internal
+
+    ret
+END(process_vm_writev)

libc/arch-arm64/syscalls/pwrite64.S

@@ -13,5 +13,4 @@ ENTRY(pwrite64)
     ret
 END(pwrite64)
 
-    .globl pwrite
-    .equ pwrite, pwrite64
+ALIAS_SYMBOL(pwrite, pwrite64)

libc/arch-arm64/syscalls/sendfile.S

@@ -13,5 +13,4 @@ ENTRY(sendfile)
     ret
 END(sendfile)
 
-    .globl sendfile64
-    .equ sendfile64, sendfile
+ALIAS_SYMBOL(sendfile64, sendfile)

libc/arch-arm64/syscalls/setrlimit.S

@@ -13,5 +13,4 @@ ENTRY(setrlimit)
     ret
 END(setrlimit)
 
-    .globl setrlimit64
-    .equ setrlimit64, setrlimit
+ALIAS_SYMBOL(setrlimit64, setrlimit)

libc/arch-arm64/syscalls/truncate.S

@@ -13,5 +13,4 @@ ENTRY(truncate)
     ret
 END(truncate)
 
-    .globl truncate64
-    .equ truncate64, truncate
+ALIAS_SYMBOL(truncate64, truncate)

libc/arch-common/bionic/asm_multiarch.h

@@ -28,9 +28,9 @@
 
 #ifdef __LP64__
 # define ASM_PTR_SIZE(x) .quad x
-# define ASM_ALIGN(x)
+# define ASM_ALIGN_TO_PTR_SIZE .balign 8
 #else
 # define ASM_PTR_SIZE(x) .long x
-# define ASM_ALIGN(x)    .align x
+# define ASM_ALIGN_TO_PTR_SIZE .balign 4
 #endif
 

libc/arch-common/bionic/crtbegin.c

@@ -59,6 +59,7 @@ void _start() {
 
 #include "__dso_handle.h"
 #include "atexit.h"
+#include "pthread_atfork.h"
 #ifdef __i386__
 # include "../../arch-x86/bionic/__stack_chk_fail_local.h"
 #endif

libc/arch-common/bionic/crtbegin_so.c

@@ -56,6 +56,7 @@ void __on_dlclose() {
 # include "__dso_handle_so.h"
 # include "atexit.h"
 #endif
+#include "pthread_atfork.h"
 #ifdef __i386__
 # include "../../arch-x86/bionic/__stack_chk_fail_local.h"
 #endif

libc/arch-common/bionic/crtbrand.S

@@ -0,0 +1,39 @@
+/*
+ * Copyright (C) 2015 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	.note.android.ident,"a",%note
+	.align	2
+	.type	abitag, %object
+abitag:
+	.long	2f-1f			/* int32_t namesz */
+	.long	3f-2f			/* int32_t descsz */
+	.long	1			/* int32_t type */
+1:	.ascii	"Android\0"		/* char name[] */
+2:	.long	PLATFORM_SDK_VERSION	/* int32_t android_api */
+3:
+	.size	abitag, .-abitag

libc/arch-common/bionic/crtend.S

@@ -29,12 +29,15 @@
 #include "asm_multiarch.h"
 
 	.section .preinit_array, "aw"
+	ASM_ALIGN_TO_PTR_SIZE
 	ASM_PTR_SIZE(0)
 
 	.section .init_array, "aw"
+	ASM_ALIGN_TO_PTR_SIZE
 	ASM_PTR_SIZE(0)
 
 	.section .fini_array, "aw"
+	ASM_ALIGN_TO_PTR_SIZE
 	ASM_PTR_SIZE(0)
 
 #if defined(__linux__) && defined(__ELF__)
@@ -42,7 +45,9 @@
 #endif
 #if defined(__i386__) || defined(__x86_64__)
 	.section	.eh_frame,"a",@progbits
-	ASM_ALIGN(4)
+#if defined(__i386__)
+	.balign 4
+#endif
 	.type	__FRAME_END__, @object
 	.size	__FRAME_END__, 4
 __FRAME_END__:

libc/arch-common/bionic/crtend_so.S

@@ -26,22 +26,14 @@
  * SUCH DAMAGE.
  */
 
-#include "asm_multiarch.h"
-
-#ifndef __arm__
-	.section .init_array, "aw"
-	ASM_PTR_SIZE(0)
-
-	.section .fini_array, "aw"
-	ASM_PTR_SIZE(0)
-#endif
-
 #if defined(__linux__) && defined(__ELF__)
 	.section .note.GNU-stack,"",%progbits
 #endif
 #if defined(__i386__) || defined(__x86_64__)
 	.section	.eh_frame,"a",@progbits
-	ASM_ALIGN(4)
+#if defined(__i386__)
+	.balign 4
+#endif
 	.type	__FRAME_END__, @object
 	.size	__FRAME_END__, 4
 __FRAME_END__:

libc/arch-common/bionic/pthread_atfork.h

@@ -0,0 +1,29 @@
+/*
+ * Copyright (C) 2015 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.
+ */
+
+extern void* __dso_handle;
+
+extern int __register_atfork(void (*prepare)(void), void (*parent)(void), void (*child)(void), void* dso);
+
+#ifndef _LIBC
+// Libc used to export this in previous versions, therefore it needs
+// to remain global for binary compatibility.
+__attribute__ ((visibility ("hidden")))
+#endif
+int pthread_atfork(void (*prepare)(void), void (*parent)(void), void (*child)(void)) {
+  return __register_atfork(prepare, parent, child, &__dso_handle);
+}
+

libc/arch-mips/bionic/crtbegin.c

@@ -92,3 +92,4 @@ __asm__ (
 
 #include "../../arch-common/bionic/__dso_handle.h"
 #include "../../arch-common/bionic/atexit.h"
+#include "../../arch-common/bionic/pthread_atfork.h"

libc/arch-mips/bionic/setjmp.S

@@ -1,3 +1,30 @@
+/*
+ * Copyright (C) 2014 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) 2001-2002 Opsycon AB  (www.opsycon.se / www.opsycon.com)
  *
@@ -94,23 +121,31 @@
 #include <private/bionic_asm.h>
 #include <machine/setjmp.h>
 
-/* On Mips32, jmpbuf begins with optional 4-byte filler so that
- *  all saved FP regs are aligned on 8-byte boundary, despite this whole
- *  struct being mis-declared to users as an array of (4-byte) longs.
- *  All the following offsets are then from the rounded-up base addr
+/* jmpbuf is declared to users as an array of longs, which is only
+ * 4-byte aligned in 32-bit builds.  The Mips jmpbuf begins with a
+ * dynamically-sized 0- or 4-byte unused filler so that double-prec FP regs
+ * are saved to 8-byte-aligned mem cells.
+ * All the following jmpbuf offsets are from the rounded-DOWN base addr.
  */
 
 /* Fields of same size on all MIPS abis: */
-#define	SC_MAGIC        (0*4)		/* 4 bytes, identify jmpbuf */
-#define	SC_MASK		(1*4)		/* 4 bytes, saved signal mask */
-#define	SC_FPSR		(2*4)		/* 4 bytes, floating point control/status reg */
-/*     	filler2		(3*4)		   4 bytes, pad to 8-byte boundary */
+/*     	field:		byte offset:	size:						*/
+/*     	dynam filler	(0*4)		   0-4 bytes of rounddown filler, DON'T TOUCH!!
+						often overlays user storage!!		*/
+#define	SC_MAGIC_OFFSET	(1*4)		/* 4 bytes, identify jmpbuf, first actual field */
+#define	SC_FLAG_OFFSET	(2*4)		/* 4 bytes, savesigs flag */
+#define	SC_FPSR_OFFSET	(3*4)		/* 4 bytes, floating point control/status reg */
+/* following fields are 8-byte aligned */
+#define	SC_MASK_OFFSET	(4*4)		/* 16 bytes, mips32/mips64 version of sigset_t */
+#define	SC_SPARE_OFFSET	(8*4)		/* 8 bytes, reserved for future uses */
 
 /* Registers that are 4-byte on mips32 o32, and 8-byte on mips64 n64 abi */
-#define	SC_REGS_SAVED	12		/* ra,gp,sp,s0-s8 */
-#define	SC_REGS		(4*4)		/* SC_REGS_SAVED*REGSZ bytes */
+#define	SC_REGS_OFFSET	(10*4)		/* SC_REGS_BYTES */
+#define	SC_REGS_SAVED	12 /*regs*/	/* ra,s0-s8,gp,sp */
+#define	SC_REGS_BYTES   (SC_REGS_SAVED*REGSZ)
+#define	SC_REGS		SC_REGS_OFFSET
 
-/* Floating pt registers are 8-bytes on all abis,
+/* Double floating pt registers are 8-bytes on all abis,
  * but the number of saved fp regs varies for o32/n32 versus n64 abis:
  */
 
@@ -120,22 +155,20 @@
 #define	SC_FPREGS_SAVED	6  /* even fp regs f20,f22,f24,f26,f28,f30 */
 #endif
 
-#define	SC_FPREGS	(SC_REGS + SC_REGS_SAVED*REGSZ)  /* SC_FPREGS_SAVED*REGSZ_FP bytes */
+#define	SC_FPREGS_OFFSET (SC_REGS_OFFSET + SC_REGS_BYTES)  /* SC_FPREGS_BYTES */
+#define	SC_FPREGS_BYTES (SC_FPREGS_SAVED*REGSZ_FP)
+#define	SC_FPREGS	SC_FPREGS_OFFSET
 
-#define	SC_BYTES	(SC_FPREGS + SC_FPREGS_SAVED*REGSZ_FP)
-#define	SC_LONGS	(SC_BYTES/REGSZ)
+#define	SC_TOTAL_BYTES	(SC_FPREGS_OFFSET + SC_FPREGS_BYTES)
+#define	SC_TOTAL_LONGS	(SC_TOTAL_BYTES/REGSZ)
 
-#ifdef __LP64__
-/* SC_LONGS is 22, so _JBLEN should be 22 or larger */
-#else
-/* SC_LONGS is 28, but must also allocate dynamic-roundup filler.
-   so _JBLEN should be 29 or larger */
+#if SC_TOTAL_LONGS > _JBLEN
+#error _JBLEN is too small
 #endif
 
 /*
- * _setjmp, _longjmp (restoring signal state)
  *
- *  GPOFF and FRAMESIZE must be the same for both _setjmp and _longjmp!
+ *  GPOFF and FRAMESIZE must be the same for all setjmp/longjmp routines
  *
  */
 
@@ -145,30 +178,33 @@ A0OFF= FRAMESZ-3*REGSZ
 GPOFF= FRAMESZ-2*REGSZ
 RAOFF= FRAMESZ-1*REGSZ
 
-NON_LEAF(setjmp, FRAMESZ, ra)
+NON_LEAF(sigsetjmp, FRAMESZ, ra)
 	.mask	0x80000000, RAOFF
 	PTR_SUBU sp, FRAMESZ			# allocate stack frame
-	SETUP_GP64(GPOFF, setjmp)
+	SETUP_GP64(GPOFF, sigsetjmp)
 	SAVE_GP(GPOFF)
 	.set	reorder
 
+setjmp_common:
 #ifndef __LP64__
-	addiu   a0, 7				# roundup jmpbuf addr to 8-byte boundary
 	li	t0, ~7
-	and     a0, t0
+	and	a0, t0				# round jmpbuf addr DOWN to 8-byte boundary
 #endif
+	sw	a1, SC_FLAG_OFFSET(a0)		# save savesigs flag
+	beqz	a1, 1f				# do saving of signal mask?
 
-	REG_S	ra, RAOFF(sp)			# save state
+	REG_S	ra, RAOFF(sp)			# spill state
 	REG_S	a0, A0OFF(sp)
-	move	a0, zero			# get current signal mask
-	jal	sigblock
+	# call sigprocmask(int how ignored, sigset_t* null, sigset_t* SC_MASK(a0)):
+	LA	a2, SC_MASK_OFFSET(a0)		# gets current signal mask
+	li	a0, 0				# how; ignored when new mask is null
+	li	a1, 0				# null new mask
+	jal	sigprocmask			# get current signal mask
 	REG_L	a0, A0OFF(sp)
 	REG_L	ra, RAOFF(sp)
-
-	REG_S	v0, SC_MASK(a0)			# save sc_mask = sigblock(0)
-
+1:
 	li	v0, 0xACEDBADE			# sigcontext magic number
-	sw	v0, SC_MAGIC(a0)
+	sw	v0, SC_MAGIC_OFFSET(a0)
 	# callee-saved long-sized regs:
 	REG_S	ra, SC_REGS+0*REGSZ(a0)
 	REG_S	s0, SC_REGS+1*REGSZ(a0)
@@ -181,9 +217,9 @@ NON_LEAF(setjmp, FRAMESZ, ra)
 	REG_S	s7, SC_REGS+8*REGSZ(a0)
 	REG_S	s8, SC_REGS+9*REGSZ(a0)
 	REG_L	v0, GPOFF(sp)
-	REG_S	v0, SC_REGS+10*REGSZ(a0)
+	REG_S	v0, SC_REGS+10*REGSZ(a0)	# save gp
 	PTR_ADDU v0, sp, FRAMESZ
-	REG_S	v0, SC_REGS+11*REGSZ(a0)
+	REG_S	v0, SC_REGS+11*REGSZ(a0)	# save orig sp
 
 	cfc1	v0, $31
 
@@ -199,7 +235,7 @@ NON_LEAF(setjmp, FRAMESZ, ra)
 	s.d	$f31, SC_FPREGS+7*REGSZ_FP(a0)
 #else
 	# callee-saved fp regs on mips o32 ABI are
-	#   the even-numbered fp regs $f20,$f22,...$f30
+	#   the even-numbered double fp regs $f20,$f22,...$f30
 	s.d	$f20, SC_FPREGS+0*REGSZ_FP(a0)
 	s.d	$f22, SC_FPREGS+1*REGSZ_FP(a0)
 	s.d	$f24, SC_FPREGS+2*REGSZ_FP(a0)
@@ -207,37 +243,68 @@ NON_LEAF(setjmp, FRAMESZ, ra)
 	s.d	$f28, SC_FPREGS+4*REGSZ_FP(a0)
 	s.d	$f30, SC_FPREGS+5*REGSZ_FP(a0)
 #endif
-	sw	v0, SC_FPSR(a0)
+	sw	v0, SC_FPSR_OFFSET(a0)
 	move	v0, zero
 	RESTORE_GP64
 	PTR_ADDU sp, FRAMESZ
 	j	ra
-END(setjmp)
+END(sigsetjmp)
 
-NON_LEAF(longjmp, FRAMESZ, ra)
+
+# Alternate entry points:
+
+NON_LEAF(setjmp, FRAMESZ, ra)
 	.mask	0x80000000, RAOFF
 	PTR_SUBU sp, FRAMESZ
-	SETUP_GP64(GPOFF, longjmp)
+	SETUP_GP64(GPOFF, setjmp)		# can't share sigsetjmp's gp code
+	SAVE_GP(GPOFF)
+	.set	reorder
+
+	li	a1, 1				# save/restore signals state
+	b	setjmp_common			# tail call
+END(setjmp)
+
+
+NON_LEAF(_setjmp, FRAMESZ, ra)
+	.mask	0x80000000, RAOFF
+	PTR_SUBU sp, FRAMESZ
+	SETUP_GP64(GPOFF, _setjmp)		# can't share sigsetjmp's gp code
+	SAVE_GP(GPOFF)
+	.set	reorder
+
+	li	a1, 0				# don't save/restore signals
+	b	setjmp_common			# tail call
+END(_setjmp)
+
+
+NON_LEAF(siglongjmp, FRAMESZ, ra)
+	.mask	0x80000000, RAOFF
+	PTR_SUBU sp, FRAMESZ
+	SETUP_GP64(GPOFF, siglongjmp)
 	SAVE_GP(GPOFF)
 	.set	reorder
 
 #ifndef __LP64__
-	addiu	a0, 7				# roundup jmpbuf addr to 8-byte boundary
 	li	t0, ~7
-	and	a0, t0
+	and	a0, t0				# round jmpbuf addr DOWN to 8-byte boundary
 #endif
-
-	REG_S	a1, A1OFF(sp)
-	REG_S	a0, A0OFF(sp)
-	lw	a0, SC_MASK(a0)
-	jal	sigsetmask
-	REG_L	a0, A0OFF(sp)
-	REG_L	a1, A1OFF(sp)
-
-	lw	v0, SC_MAGIC(a0)
+	lw	v0, SC_MAGIC_OFFSET(a0)
 	li	t0, 0xACEDBADE
 	bne	v0, t0, longjmp_botch		# jump if error
 
+	lw	t0, SC_FLAG_OFFSET(a0)		# get savesigs flag
+	beqz	t0, 1f				# restore signal mask?
+
+	REG_S	a1, A1OFF(sp)			# temp spill
+	REG_S	a0, A0OFF(sp)
+        # call sigprocmask(int how SIG_SETMASK, sigset_t* SC_MASK(a0), sigset_t* null):
+	LA	a1, SC_MASK_OFFSET(a0)		# signals being restored
+	li	a0, 3				# mips SIG_SETMASK
+	li	a2, 0				# null
+	jal	sigprocmask			# restore signal mask
+	REG_L	a0, A0OFF(sp)
+	REG_L	a1, A1OFF(sp)
+1:
 	# callee-saved long-sized regs:
 	REG_L	ra, SC_REGS+0*REGSZ(a0)
 	REG_L	s0, SC_REGS+1*REGSZ(a0)
@@ -252,8 +319,8 @@ NON_LEAF(longjmp, FRAMESZ, ra)
 	REG_L	gp, SC_REGS+10*REGSZ(a0)
 	REG_L	sp, SC_REGS+11*REGSZ(a0)
 
-	lw	v0, SC_FPSR(a0)
-	ctc1	v0, $31
+	lw	v0, SC_FPSR_OFFSET(a0)
+	ctc1	v0, $31			# restore old fr mode before fp values
 #ifdef __LP64__
 	# callee-saved fp regs on mips n64 ABI are $f24..$f31
 	l.d	$f24, SC_FPREGS+0*REGSZ_FP(a0)
@@ -266,7 +333,7 @@ NON_LEAF(longjmp, FRAMESZ, ra)
 	l.d	$f31, SC_FPREGS+7*REGSZ_FP(a0)
 #else
 	# callee-saved fp regs on mips o32 ABI are
-	#   the even-numbered fp regs $f20,$f22,...$f30
+	#   the even-numbered double fp regs $f20,$f22,...$f30
 	l.d	$f20, SC_FPREGS+0*REGSZ_FP(a0)
 	l.d	$f22, SC_FPREGS+1*REGSZ_FP(a0)
 	l.d	$f24, SC_FPREGS+2*REGSZ_FP(a0)
@@ -278,192 +345,12 @@ NON_LEAF(longjmp, FRAMESZ, ra)
 	li	a1, 1			# never return 0!
 1:
 	move	v0, a1
-	j	ra
+	j	ra			# return to setjmp call site
 
 longjmp_botch:
 	jal	longjmperror
 	jal	abort
-	RESTORE_GP64
-	PTR_ADDU sp, FRAMESZ
-END(longjmp)
-
-
-/*
- * _setjmp, _longjmp (not restoring signal state)
- *
- *  GPOFF and FRAMESIZE must be the same for both _setjmp and _longjmp!
- *
- */
-
-FRAMESZ= MKFSIZ(0,4)
-GPOFF= FRAMESZ-2*REGSZ
-
-LEAF(_setjmp, FRAMESZ)
-	PTR_SUBU sp, FRAMESZ
-	SETUP_GP64(GPOFF, _setjmp)
-	SAVE_GP(GPOFF)
-	.set	reorder
-
-#ifndef __LP64__
-	addiu   a0, 7				# roundup jmpbuf addr to 8-byte boundary
-	li      t0, ~7
-	and     a0, t0
-#endif
-
-	# SC_MASK is unused here
-
-	li	v0, 0xACEDBADE			# sigcontext magic number
-	sw	v0, SC_MAGIC(a0)
-	# callee-saved long-sized regs:
-	REG_S	ra, SC_REGS+0*REGSZ(a0)
-	REG_S	s0, SC_REGS+1*REGSZ(a0)
-	REG_S	s1, SC_REGS+2*REGSZ(a0)
-	REG_S	s2, SC_REGS+3*REGSZ(a0)
-	REG_S	s3, SC_REGS+4*REGSZ(a0)
-	REG_S	s4, SC_REGS+5*REGSZ(a0)
-	REG_S	s5, SC_REGS+6*REGSZ(a0)
-	REG_S	s6, SC_REGS+7*REGSZ(a0)
-	REG_S	s7, SC_REGS+8*REGSZ(a0)
-	REG_S	s8, SC_REGS+9*REGSZ(a0)
-	REG_L	v0, GPOFF(sp)
-	REG_S	v0, SC_REGS+10*REGSZ(a0)
-	PTR_ADDU v0, sp, FRAMESZ
-	REG_S	v0, SC_REGS+11*REGSZ(a0)
-
-	cfc1	v0, $31
-
-#ifdef __LP64__
-	# callee-saved fp regs on mips n64 ABI are $f24..$f31
-	s.d	$f24, SC_FPREGS+0*REGSZ_FP(a0)
-	s.d	$f25, SC_FPREGS+1*REGSZ_FP(a0)
-	s.d	$f26, SC_FPREGS+2*REGSZ_FP(a0)
-	s.d	$f27, SC_FPREGS+3*REGSZ_FP(a0)
-	s.d	$f28, SC_FPREGS+4*REGSZ_FP(a0)
-	s.d	$f29, SC_FPREGS+5*REGSZ_FP(a0)
-	s.d	$f30, SC_FPREGS+6*REGSZ_FP(a0)
-	s.d	$f31, SC_FPREGS+7*REGSZ_FP(a0)
-#else
-	# callee-saved fp regs on mips o32 ABI are
-	#   the even-numbered fp regs $f20,$f22,...$f30
-	s.d	$f20, SC_FPREGS+0*REGSZ_FP(a0)
-	s.d	$f22, SC_FPREGS+1*REGSZ_FP(a0)
-	s.d	$f24, SC_FPREGS+2*REGSZ_FP(a0)
-	s.d	$f26, SC_FPREGS+3*REGSZ_FP(a0)
-	s.d	$f28, SC_FPREGS+4*REGSZ_FP(a0)
-	s.d	$f30, SC_FPREGS+5*REGSZ_FP(a0)
-#endif
-	sw	v0, SC_FPSR(a0)
-	move	v0, zero
-	RESTORE_GP64
-	PTR_ADDU sp, FRAMESZ
-	j	ra
-END(_setjmp)
-
-
-LEAF(_longjmp, FRAMESZ)
-	PTR_SUBU sp, FRAMESZ
-	SETUP_GP64(GPOFF, _longjmp)
-	SAVE_GP(GPOFF)
-	.set	reorder
-
-#ifndef __LP64__
-	addiu	a0, 7				# roundup jmpbuf addr to 8-byte boundary
-	li      t0, ~7
-	and	a0, t0
-#endif
-
-	# SC_MASK is unused here
-
-	lw	v0, SC_MAGIC(a0)
-	li	t0, 0xACEDBADE
-	bne	v0, t0, _longjmp_botch			# jump if error
-
-	# callee-saved long-sized regs:
-	REG_L	ra, SC_REGS+0*REGSZ(a0)
-	REG_L	s0, SC_REGS+1*REGSZ(a0)
-	REG_L	s1, SC_REGS+2*REGSZ(a0)
-	REG_L	s2, SC_REGS+3*REGSZ(a0)
-	REG_L	s3, SC_REGS+4*REGSZ(a0)
-	REG_L	s4, SC_REGS+5*REGSZ(a0)
-	REG_L	s5, SC_REGS+6*REGSZ(a0)
-	REG_L	s6, SC_REGS+7*REGSZ(a0)
-	REG_L	s7, SC_REGS+8*REGSZ(a0)
-	REG_L	s8, SC_REGS+9*REGSZ(a0)
-	REG_L	gp, SC_REGS+10*REGSZ(a0)
-	REG_L	sp, SC_REGS+11*REGSZ(a0)
-
-	lw	v0, SC_FPSR(a0)
-	ctc1	v0, $31
-#ifdef __LP64__
-	# callee-saved fp regs on mips n64 ABI are $f24..$f31
-	l.d	$f24, SC_FPREGS+0*REGSZ_FP(a0)
-	l.d	$f25, SC_FPREGS+1*REGSZ_FP(a0)
-	l.d	$f26, SC_FPREGS+2*REGSZ_FP(a0)
-	l.d	$f27, SC_FPREGS+3*REGSZ_FP(a0)
-	l.d	$f28, SC_FPREGS+4*REGSZ_FP(a0)
-	l.d	$f29, SC_FPREGS+5*REGSZ_FP(a0)
-	l.d	$f30, SC_FPREGS+6*REGSZ_FP(a0)
-	l.d	$f31, SC_FPREGS+7*REGSZ_FP(a0)
-#else
-	# callee-saved fp regs on mips o32 ABI are
-	#   the even-numbered fp regs $f20,$f22,...$f30
-	l.d	$f20, SC_FPREGS+0*REGSZ_FP(a0)
-	l.d	$f22, SC_FPREGS+1*REGSZ_FP(a0)
-	l.d	$f24, SC_FPREGS+2*REGSZ_FP(a0)
-	l.d	$f26, SC_FPREGS+3*REGSZ_FP(a0)
-	l.d	$f28, SC_FPREGS+4*REGSZ_FP(a0)
-	l.d	$f30, SC_FPREGS+5*REGSZ_FP(a0)
-#endif
-	bne	a1, zero, 1f
-	li	a1, 1			# never return 0!
-1:
-	move	v0, a1
-	j	ra
-
-_longjmp_botch:
-	jal	longjmperror
-	jal	abort
-	RESTORE_GP64
-	PTR_ADDU sp, FRAMESZ
-END(_longjmp)
-
-/*
- * trampolines for sigsetjmp and  siglongjmp save and restore mask.
- *
- */
-FRAMESZ= MKFSIZ(1,1)
-GPOFF= FRAMESZ-2*REGSZ
-
-LEAF(sigsetjmp, FRAMESZ)
-	PTR_SUBU sp, FRAMESZ
-	SETUP_GP64(GPOFF, sigsetjmp)
-	.set	reorder
-	sw	a1, _JBLEN*REGSZ(a0)		# save "savemask"
-	bne	a1, 0x0, 1f			# do saving of signal mask?
-	LA	t9, _setjmp
-	RESTORE_GP64
-	PTR_ADDU sp, FRAMESZ
-	jr t9
-
-1:	LA	t9, setjmp
-	RESTORE_GP64
-	PTR_ADDU sp, FRAMESZ
-	jr t9
-END(sigsetjmp)
-
-LEAF(siglongjmp, FRAMESZ)
-	PTR_SUBU sp, FRAMESZ
-	SETUP_GP64(GPOFF, siglongjmp)
-	.set	reorder
-	lw	t0, _JBLEN*REGSZ(a0)		# get "savemask"
-	bne	t0, 0x0, 1f			# restore signal mask?
-	LA	t9, _longjmp
-	RESTORE_GP64
-	PTR_ADDU sp, FRAMESZ
-	jr	t9
-1:
-	LA	t9, longjmp
-	RESTORE_GP64
-	PTR_ADDU sp, FRAMESZ
-	jr	t9
 END(siglongjmp)
+
+ALIAS_SYMBOL(longjmp, siglongjmp)
+ALIAS_SYMBOL(_longjmp, siglongjmp)

libc/arch-mips/include/machine/elf_machdep.h

@@ -121,6 +121,7 @@
 #define	DT_MIPS_GOTSYM		0x70000013	/* first dynamic sym in got */
 #define DT_MIPS_HIPAGENO	0x70000014
 #define	DT_MIPS_RLD_MAP		0x70000016	/* address of loader map */
+#define DT_MIPS_RLD_MAP2	0x70000035	/* offset of loader map, used for PIE */
 
 /*
  * ELF Flags

libc/arch-mips/include/machine/setjmp.h

@@ -6,9 +6,10 @@
 #define _MIPS_SETJMP_H_
 
 #ifdef __LP64__
-#define	_JBLEN	22		/* size, in 8-byte longs, of a mips64 jmp_buf */
+#define	_JBLEN	25	/* size, in 8-byte longs, of a mips64 jmp_buf/sigjmp_buf */
 #else
-#define	_JBLEN	29		/* size, in 4-byte longs, of a mips32 jmp_buf */
+#define	_JBLEN	157	/* historical size, in 4-byte longs, of a mips32 jmp_buf */
+			/* actual used size is 34 */
 #endif
 
 #endif /* !_MIPS_SETJMP_H_ */

libc/arch-mips/mips.mk

@@ -1,36 +1,18 @@
 # 32-bit mips.
 
-#
-# Various kinds of LP32 cruft.
-#
-
-libc_bionic_src_files_mips += \
-    bionic/mmap.cpp \
-
-libc_common_src_files_mips += \
-    bionic/legacy_32_bit_support.cpp \
-    bionic/ndk_cruft.cpp \
-    bionic/time64.c \
-
-libc_netbsd_src_files_mips += \
-    upstream-netbsd/common/lib/libc/hash/sha1/sha1.c \
-
-libc_openbsd_src_files_mips += \
-    upstream-openbsd/lib/libc/stdio/putw.c \
-
 #
 # Default implementations of functions that are commonly optimized.
 #
 
 libc_bionic_src_files_mips += \
+    arch-mips/string/memcmp.c \
+    arch-mips/string/memcpy.S \
+    arch-mips/string/memset.S \
+    arch-mips/string/strcmp.S \
     bionic/__memcpy_chk.cpp \
     bionic/__memset_chk.cpp \
     bionic/__strcpy_chk.cpp \
     bionic/__strcat_chk.cpp \
-    bionic/memchr.c \
-    bionic/memcmp.c \
-    bionic/memmove.c \
-    bionic/memrchr.c \
     bionic/strchr.cpp \
     bionic/strnlen.c \
     bionic/strrchr.cpp \
@@ -46,11 +28,12 @@ libc_freebsd_src_files_mips += \
     upstream-freebsd/lib/libc/string/wmemmove.c \
 
 libc_openbsd_src_files_mips += \
-    upstream-openbsd/lib/libc/string/bcopy.c \
+    upstream-openbsd/lib/libc/string/memchr.c \
+    upstream-openbsd/lib/libc/string/memmove.c \
+    upstream-openbsd/lib/libc/string/memrchr.c \
     upstream-openbsd/lib/libc/string/stpcpy.c \
     upstream-openbsd/lib/libc/string/stpncpy.c \
     upstream-openbsd/lib/libc/string/strcat.c \
-    upstream-openbsd/lib/libc/string/strcmp.c \
     upstream-openbsd/lib/libc/string/strcpy.c \
     upstream-openbsd/lib/libc/string/strlcat.c \
     upstream-openbsd/lib/libc/string/strlcpy.c \
@@ -73,17 +56,11 @@ libc_bionic_src_files_mips += \
 
 ifndef ARCH_MIPS_REV6
 libc_bionic_src_files_mips += \
-    arch-mips/string/memcpy.S \
-    arch-mips/string/memset.S \
     arch-mips/string/mips_strlen.c \
 
 else
 libc_bionic_src_files_mips += \
-    bionic/memcpy.cpp \
-    bionic/memset.c \
-
-libc_openbsd_src_files_mips += \
-    upstream-openbsd/lib/libc/string/strlen.c \
+    arch-mips/string/strlen.c \
 
 endif