Add DiscardableMemoryAllocator to work around FD limit issue.

base/memory/discardable_memory_allocator_android_unittest.cc

+// Copyright 2013 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/memory/discardable_memory_allocator_android.h"
+
+#include <sys/types.h>
+#include <unistd.h>
+
+#include "base/memory/discardable_memory.h"
+#include "base/memory/scoped_ptr.h"
+#include "base/strings/string_number_conversions.h"
+#include "base/strings/string_split.h"
+#include "base/strings/stringprintf.h"
+#include "build/build_config.h"
+#include "testing/gtest/include/gtest/gtest.h"
+
+namespace base {
+namespace internal {
+
+const char kAllocatorName[] = "allocator-for-testing";
+
+const size_t kPageSize = 4096;
+const size_t kMinAshmemRegionSize =
+    DiscardableMemoryAllocator::kMinAshmemRegionSize;
+
+class DiscardableMemoryAllocatorTest : public testing::Test {
+ protected:
+  DiscardableMemoryAllocatorTest() : allocator_(kAllocatorName) {}
+
+  DiscardableMemoryAllocator allocator_;
+};
+
+void WriteToDiscardableMemory(DiscardableMemory* memory, size_t size) {
+  // Write to the first and the last pages only to avoid paging in up to 64
+  // MBytes.
+  static_cast<char*>(memory->Memory())[0] = 'a';
+  static_cast<char*>(memory->Memory())[size - 1] = 'a';
+}
+
+TEST_F(DiscardableMemoryAllocatorTest, Basic) {
+  const size_t size = 128;
+  scoped_ptr<DiscardableMemory> memory(allocator_.Allocate(size));
+  ASSERT_TRUE(memory);
+  WriteToDiscardableMemory(memory.get(), size);
+}
+
+TEST_F(DiscardableMemoryAllocatorTest, LargeAllocation) {
+  const size_t size = 64 * 1024 * 1024;
+  scoped_ptr<DiscardableMemory> memory(allocator_.Allocate(size));
+  ASSERT_TRUE(memory);
+  WriteToDiscardableMemory(memory.get(), size);
+}
+
+TEST_F(DiscardableMemoryAllocatorTest, ChunksArePageAligned) {
+  scoped_ptr<DiscardableMemory> memory(allocator_.Allocate(kPageSize));
+  ASSERT_TRUE(memory);
+  EXPECT_EQ(0U, reinterpret_cast<uint64_t>(memory->Memory()) % kPageSize);
+  WriteToDiscardableMemory(memory.get(), kPageSize);
+}
+
+TEST_F(DiscardableMemoryAllocatorTest, AllocateFreeAllocate) {
+  scoped_ptr<DiscardableMemory> memory(allocator_.Allocate(kPageSize));
+  ASSERT_TRUE(memory);
+  void* const address = memory->Memory();
+  memory->Unlock();  // Tests that the recycled chunk is being locked correctly.
+  memory.reset();
+  memory = allocator_.Allocate(kPageSize);
+  ASSERT_TRUE(memory);
+  // The previously freed chunk should be reused.
+  EXPECT_EQ(address, memory->Memory());
+  WriteToDiscardableMemory(memory.get(), kPageSize);
+}
+
+TEST_F(DiscardableMemoryAllocatorTest, FreeingWholeAshmemRegionClosesAshmem) {
+  scoped_ptr<DiscardableMemory> memory(
+      allocator_.Allocate(kMinAshmemRegionSize));
+  ASSERT_TRUE(memory);
+  const int kMagic = 0xdeadbeef;
+  *static_cast<int*>(memory->Memory()) = kMagic;
+  memory.reset();
+  // The previous ashmem region should have been closed thus it should not be
+  // recycled.
+  memory = allocator_.Allocate(kPageSize);
+  ASSERT_TRUE(memory);
+  EXPECT_NE(kMagic, *static_cast<const int*>(memory->Memory()));
+}
+
+TEST_F(DiscardableMemoryAllocatorTest, AllocateUsesBestFitAlgorithm) {
+  scoped_ptr<DiscardableMemory> memory1(allocator_.Allocate(3 * kPageSize));
+  ASSERT_TRUE(memory1);
+  scoped_ptr<DiscardableMemory> memory2(allocator_.Allocate(2 * kPageSize));
+  ASSERT_TRUE(memory2);
+  scoped_ptr<DiscardableMemory> memory3(allocator_.Allocate(1 * kPageSize));
+  ASSERT_TRUE(memory3);
+  void* const address_3 = memory3->Memory();
+  memory1.reset();
+  // Don't free |memory2| to avoid merging the 3 blocks together.
+  memory3.reset();
+  memory1 = allocator_.Allocate(1 * kPageSize);
+  ASSERT_TRUE(memory1);
+  // The chunk whose size is closest to the requested size should be recycled.
+  EXPECT_EQ(address_3, memory1->Memory());
+  WriteToDiscardableMemory(memory1.get(), kPageSize);
+}
+
+TEST_F(DiscardableMemoryAllocatorTest, MergeFreeChunks) {
+  scoped_ptr<DiscardableMemory> memory1(allocator_.Allocate(kPageSize));
+  ASSERT_TRUE(memory1);
+  scoped_ptr<DiscardableMemory> memory2(allocator_.Allocate(kPageSize));
+  ASSERT_TRUE(memory2);
+  scoped_ptr<DiscardableMemory> memory3(allocator_.Allocate(kPageSize));
+  ASSERT_TRUE(memory3);
+  void* const memory1_address = memory1->Memory();
+  memory1.reset();
+  memory3.reset();
+  // Freeing |memory2| (located between memory1 and memory3) should merge the
+  // three free blocks together.
+  memory2.reset();
+  memory1.reset(allocator_.Allocate(3 * kPageSize).release());
+  EXPECT_EQ(memory1_address, memory1->Memory());
+}
+
+TEST_F(DiscardableMemoryAllocatorTest,
+     TooLargeFreeChunksDontCauseTooMuchFragmentationWhenRecycled) {
+  // Keep |memory_1| below allocated so that the ashmem region doesn't get
+  // closed when |memory_2| is deleted.
+  scoped_ptr<DiscardableMemory> memory_1(allocator_.Allocate(64 * 1024));
+  ASSERT_TRUE(memory_1);
+  scoped_ptr<DiscardableMemory> memory_2(allocator_.Allocate(32 * 1024));
+  ASSERT_TRUE(memory_2);
+  void* const address = memory_2->Memory();
+  memory_2.reset();
+  const size_t size = 16 * 1024;
+  memory_2 = allocator_.Allocate(size);
+  ASSERT_TRUE(memory_2);
+  EXPECT_EQ(address, memory_2->Memory());
+  WriteToDiscardableMemory(memory_2.get(), size);
+  scoped_ptr<DiscardableMemory> memory_3(allocator_.Allocate(size));
+  // The unused tail (16 KBytes large) of the previously freed chunk should be
+  // recycled.
+  EXPECT_EQ(static_cast<char*>(address) + size, memory_3->Memory());
+  WriteToDiscardableMemory(memory_3.get(), size);
+}
+
+TEST_F(DiscardableMemoryAllocatorTest, UseMultipleAshmemRegions) {
+  // Leave one page untouched at the end of the ashmem region.
+  const size_t size = kMinAshmemRegionSize - kPageSize;
+  scoped_ptr<DiscardableMemory> memory1(allocator_.Allocate(size));
+  ASSERT_TRUE(memory1);
+  WriteToDiscardableMemory(memory1.get(), size);
+
+  scoped_ptr<DiscardableMemory> memory2(
+      allocator_.Allocate(kMinAshmemRegionSize));
+  ASSERT_TRUE(memory2);
+  WriteToDiscardableMemory(memory2.get(), kMinAshmemRegionSize);
+  // The last page of the first ashmem region should be used for this
+  // allocation.
+  scoped_ptr<DiscardableMemory> memory3(allocator_.Allocate(kPageSize));
+  ASSERT_TRUE(memory3);
+  WriteToDiscardableMemory(memory3.get(), kPageSize);
+  EXPECT_EQ(memory3->Memory(), static_cast<char*>(memory1->Memory()) + size);
+}
+

[willchan no longer on Chromium, 2013/11/19 02:19:43]

I'm still going through this, but is there a test for something like this.
1) Allocate 16KB chunk (triggers an ashmem region of 32KB)
2) Allocate 16KB chunk (uses the rest of the ashmem region of 32KB)
3) Free the first chunk.
4) Allocate a 8KB chunk. This splits the free chunk into two. Now we have 8KB
(used), 8KB (free), 16KB (used)
5) Free the latter 16KB chunk

At this point, what's the state like?
8KB (used), 24KB (free)?

I'm still trying to grok the code, but it looks like 8KB (used), 8KB (free),
16KB (free) from what I can tell. Is that right?

[Philippe, 2013/11/19 15:43:00]

Wow, amazing catch! I added your unit test and made it pass. This implied quite
a lot of changes. Let me know if you can come up with a better approach than
what I did.

+}  // namespace internal
+}  // namespace base