content/browser/gpu/browser_gpu_memory_buffer_manager.cc - chromium/src - Git at Google

 // Copyright 2014 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 "content/browser/gpu/browser_gpu_memory_buffer_manager.h"

 #include "base/atomic_sequence_num.h"
 #include "base/bind.h"
 #include "base/lazy_instance.h"
 #include "base/synchronization/waitable_event.h"
 #include "base/threading/thread_restrictions.h"
 #include "base/trace_event/trace_event.h"
 #include "content/common/gpu/client/gpu_memory_buffer_factory_host.h"
 #include "content/common/gpu/client/gpu_memory_buffer_impl.h"
 #include "content/common/gpu/client/gpu_memory_buffer_impl_shared_memory.h"
 #include "content/public/browser/browser_thread.h"

 namespace content {
 namespace {

 BrowserGpuMemoryBufferManager* g_gpu_memory_buffer_manager = nullptr;

 // Global atomic to generate gpu memory buffer unique IDs.
 base::StaticAtomicSequenceNumber g_next_gpu_memory_buffer_id;

 }  // namespace

 struct BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferRequest {
   AllocateGpuMemoryBufferRequest(const gfx::Size& size,
                                  gfx::GpuMemoryBuffer::Format format,
                                  gfx::GpuMemoryBuffer::Usage usage,
                                  int client_id,
                                  int surface_id)
       : event(true, false),
         size(size),
         format(format),
         usage(usage),
         client_id(client_id),
         surface_id(surface_id) {}
   ~AllocateGpuMemoryBufferRequest() {}
   base::WaitableEvent event;
   gfx::Size size;
   gfx::GpuMemoryBuffer::Format format;
   gfx::GpuMemoryBuffer::Usage usage;
   int client_id;
   int surface_id;
   scoped_ptr<gfx::GpuMemoryBuffer> result;
 };

 BrowserGpuMemoryBufferManager::BrowserGpuMemoryBufferManager(
     GpuMemoryBufferFactoryHost* gpu_memory_buffer_factory_host,
     int gpu_client_id)
     : gpu_memory_buffer_factory_host_(gpu_memory_buffer_factory_host),
       gpu_client_id_(gpu_client_id),
       weak_ptr_factory_(this) {
   DCHECK(!g_gpu_memory_buffer_manager);
   g_gpu_memory_buffer_manager = this;
 }

 BrowserGpuMemoryBufferManager::~BrowserGpuMemoryBufferManager() {
   g_gpu_memory_buffer_manager = nullptr;
 }

 // static
 BrowserGpuMemoryBufferManager* BrowserGpuMemoryBufferManager::current() {
   return g_gpu_memory_buffer_manager;
 }

 scoped_ptr<gfx::GpuMemoryBuffer>
 BrowserGpuMemoryBufferManager::AllocateGpuMemoryBuffer(
     const gfx::Size& size,
     gfx::GpuMemoryBuffer::Format format,
     gfx::GpuMemoryBuffer::Usage usage) {
   return AllocateGpuMemoryBufferCommon(size, format, usage, 0);
 }

 scoped_ptr<gfx::GpuMemoryBuffer>
 BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferForScanout(
     const gfx::Size& size,
     gfx::GpuMemoryBuffer::Format format,
     int32 surface_id) {
   DCHECK_GT(surface_id, 0);
   return AllocateGpuMemoryBufferCommon(
       size, format, gfx::GpuMemoryBuffer::SCANOUT, surface_id);
 }

 scoped_ptr<gfx::GpuMemoryBuffer>
 BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferCommon(
     const gfx::Size& size,
     gfx::GpuMemoryBuffer::Format format,
     gfx::GpuMemoryBuffer::Usage usage,
     int32 surface_id) {
   DCHECK(!BrowserThread::CurrentlyOn(BrowserThread::IO));

   // Fallback to shared memory buffer if |format| and |usage| are not supported
   // by factory.
   if (!gpu_memory_buffer_factory_host_->IsGpuMemoryBufferConfigurationSupported(
           format, usage)) {
     DCHECK(GpuMemoryBufferImplSharedMemory::IsFormatSupported(format));
     DCHECK_EQ(usage, gfx::GpuMemoryBuffer::MAP);
     return GpuMemoryBufferImplSharedMemory::Create(
         g_next_gpu_memory_buffer_id.GetNext(), size, format);
   }

   AllocateGpuMemoryBufferRequest request(size, format, usage, gpu_client_id_,
                                          surface_id);
   BrowserThread::PostTask(
       BrowserThread::IO,
       FROM_HERE,
       base::Bind(&BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferOnIO,
                  base::Unretained(this),  // Safe as we wait for result below.
                  base::Unretained(&request)));

   // We're blocking the UI thread, which is generally undesirable.
   TRACE_EVENT0("browser",
                "BrowserGpuMemoryBufferManager::AllocateGpuMemoryBuffer");
   base::ThreadRestrictions::ScopedAllowWait allow_wait;
   request.event.Wait();
   return request.result.Pass();
 }

 void BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferForChildProcess(
     const gfx::Size& size,
     gfx::GpuMemoryBuffer::Format format,
     gfx::GpuMemoryBuffer::Usage usage,
     base::ProcessHandle child_process_handle,
     int child_client_id,
     const AllocationCallback& callback) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));

   gfx::GpuMemoryBufferId new_id = g_next_gpu_memory_buffer_id.GetNext();

   BufferMap& buffers = clients_[child_client_id];
   DCHECK(buffers.find(new_id) == buffers.end());

   // Fallback to shared memory buffer if |format| and |usage| are not supported
   // by factory.
   if (!gpu_memory_buffer_factory_host_->IsGpuMemoryBufferConfigurationSupported(
           format, usage)) {
     // Early out if we cannot fallback to shared memory buffer.
     if (!GpuMemoryBufferImplSharedMemory::IsFormatSupported(format) ||
         !GpuMemoryBufferImplSharedMemory::IsSizeValidForFormat(size, format) ||
         usage != gfx::GpuMemoryBuffer::MAP) {
       callback.Run(gfx::GpuMemoryBufferHandle());
       return;
     }

     buffers[new_id] = gfx::SHARED_MEMORY_BUFFER;
     callback.Run(GpuMemoryBufferImplSharedMemory::AllocateForChildProcess(
         new_id, size, format, child_process_handle));
     return;
   }

   // Note: Handling of cases where the child process is removed before the
   // allocation completes is less subtle if we set the buffer type to
   // EMPTY_BUFFER here and verify that this has not changed when allocation
   // completes.
   buffers[new_id] = gfx::EMPTY_BUFFER;

   gpu_memory_buffer_factory_host_->CreateGpuMemoryBuffer(
       new_id, size, format, usage, child_client_id, 0,
       base::Bind(&BrowserGpuMemoryBufferManager::
                      GpuMemoryBufferAllocatedForChildProcess,
                  weak_ptr_factory_.GetWeakPtr(), child_client_id, callback));
 }

 gfx::GpuMemoryBuffer*
 BrowserGpuMemoryBufferManager::GpuMemoryBufferFromClientBuffer(
     ClientBuffer buffer) {
   return GpuMemoryBufferImpl::FromClientBuffer(buffer);
 }

 void BrowserGpuMemoryBufferManager::SetDestructionSyncPoint(
     gfx::GpuMemoryBuffer* buffer,
     uint32 sync_point) {
   static_cast<GpuMemoryBufferImpl*>(buffer)
       ->set_destruction_sync_point(sync_point);
 }

 void BrowserGpuMemoryBufferManager::ChildProcessDeletedGpuMemoryBuffer(
     gfx::GpuMemoryBufferId id,
     base::ProcessHandle child_process_handle,
     int child_client_id,
     uint32 sync_point) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   DCHECK(clients_.find(child_client_id) != clients_.end());

   BufferMap& buffers = clients_[child_client_id];

   BufferMap::iterator buffer_it = buffers.find(id);
   if (buffer_it == buffers.end()) {
     LOG(ERROR) << "Invalid GpuMemoryBuffer ID for child process.";
     return;
   }

   // This can happen if a child process managed to trigger a call to this while
   // a buffer is in the process of being allocated.
   if (buffer_it->second == gfx::EMPTY_BUFFER) {
     LOG(ERROR) << "Invalid GpuMemoryBuffer type.";
     return;
   }

   // Buffers allocated using the factory need to be destroyed through the
   // factory.
   if (buffer_it->second != gfx::SHARED_MEMORY_BUFFER) {
     gpu_memory_buffer_factory_host_->DestroyGpuMemoryBuffer(id,
                                                             child_client_id,
                                                             sync_point);
   }

   buffers.erase(buffer_it);
 }

 void BrowserGpuMemoryBufferManager::ProcessRemoved(
     base::ProcessHandle process_handle,
     int client_id) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));

   ClientMap::iterator client_it = clients_.find(client_id);
   if (client_it == clients_.end())
     return;

   for (auto &buffer_it : client_it->second) {
     // This might happen if buffer is currenlty in the process of being
     // allocated. The buffer will in that case be cleaned up when allocation
     // completes.
     if (buffer_it.second == gfx::EMPTY_BUFFER)
       continue;

     // Skip shared memory buffers as they were not allocated using the factory.
     if (buffer_it.second == gfx::SHARED_MEMORY_BUFFER)
       continue;

     gpu_memory_buffer_factory_host_->DestroyGpuMemoryBuffer(buffer_it.first,
                                                             client_id,
                                                             0);
   }

   clients_.erase(client_it);
 }

 void BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferOnIO(
     AllocateGpuMemoryBufferRequest* request) {
   // Note: Unretained is safe as this is only used for synchronous allocation
   // from a non-IO thread.
   gpu_memory_buffer_factory_host_->CreateGpuMemoryBuffer(
       g_next_gpu_memory_buffer_id.GetNext(), request->size, request->format,
       request->usage, request->client_id, request->surface_id,
       base::Bind(&BrowserGpuMemoryBufferManager::GpuMemoryBufferAllocatedOnIO,
                  base::Unretained(this), base::Unretained(request)));
 }

 void BrowserGpuMemoryBufferManager::GpuMemoryBufferAllocatedOnIO(
     AllocateGpuMemoryBufferRequest* request,
     const gfx::GpuMemoryBufferHandle& handle) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));

   // Early out if factory failed to allocate the buffer.
   if (handle.is_null()) {
     request->event.Signal();
     return;
   }

   DCHECK_NE(handle.type, gfx::SHARED_MEMORY_BUFFER);
   request->result = GpuMemoryBufferImpl::CreateFromHandle(
       handle,
       request->size,
       request->format,
       base::Bind(&BrowserGpuMemoryBufferManager::GpuMemoryBufferDeleted,
                  weak_ptr_factory_.GetWeakPtr(),
                  handle.id,
                  request->client_id));
   request->event.Signal();
 }

 void BrowserGpuMemoryBufferManager::GpuMemoryBufferDeleted(
     gfx::GpuMemoryBufferId id,
     int client_id,
     uint32 sync_point) {
   gpu_memory_buffer_factory_host_->DestroyGpuMemoryBuffer(id,
                                                           client_id,
                                                           sync_point);
 }

 void BrowserGpuMemoryBufferManager::GpuMemoryBufferAllocatedForChildProcess(
     int child_client_id,
     const AllocationCallback& callback,
     const gfx::GpuMemoryBufferHandle& handle) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));

   ClientMap::iterator client_it = clients_.find(child_client_id);

   // This can happen if the child process is removed while the buffer is being
   // allocated.
   if (client_it == clients_.end()) {
     if (!handle.is_null()) {
       gpu_memory_buffer_factory_host_->DestroyGpuMemoryBuffer(
           handle.id, child_client_id, 0);
     }
     callback.Run(gfx::GpuMemoryBufferHandle());
     return;
   }

   BufferMap& buffers = client_it->second;

   BufferMap::iterator buffer_it = buffers.find(handle.id);
   DCHECK(buffer_it != buffers.end());
   DCHECK_EQ(buffer_it->second, gfx::EMPTY_BUFFER);

   if (handle.is_null()) {
     buffers.erase(buffer_it);
     callback.Run(gfx::GpuMemoryBufferHandle());
     return;
   }

   // The factory should never return a shared memory backed buffer.
   DCHECK_NE(handle.type, gfx::SHARED_MEMORY_BUFFER);

   // Store the type of this buffer so it can be cleaned up if the child
   // process is removed.
   buffer_it->second = handle.type;

   callback.Run(handle);
 }

 }  // namespace content
	// Copyright 2014 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 "content/browser/gpu/browser_gpu_memory_buffer_manager.h"

	#include "base/atomic_sequence_num.h"
	#include "base/bind.h"
	#include "base/lazy_instance.h"
	#include "base/synchronization/waitable_event.h"
	#include "base/threading/thread_restrictions.h"
	#include "base/trace_event/trace_event.h"
	#include "content/common/gpu/client/gpu_memory_buffer_factory_host.h"
	#include "content/common/gpu/client/gpu_memory_buffer_impl.h"
	#include "content/common/gpu/client/gpu_memory_buffer_impl_shared_memory.h"
	#include "content/public/browser/browser_thread.h"

	namespace content {
	namespace {

	BrowserGpuMemoryBufferManager* g_gpu_memory_buffer_manager = nullptr;

	// Global atomic to generate gpu memory buffer unique IDs.
	base::StaticAtomicSequenceNumber g_next_gpu_memory_buffer_id;

	} // namespace

	struct BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferRequest {
	AllocateGpuMemoryBufferRequest(const gfx::Size& size,
	gfx::GpuMemoryBuffer::Format format,
	gfx::GpuMemoryBuffer::Usage usage,
	int client_id,
	int surface_id)
	: event(true, false),
	size(size),
	format(format),
	usage(usage),
	client_id(client_id),
	surface_id(surface_id) {}
	~AllocateGpuMemoryBufferRequest() {}
	base::WaitableEvent event;
	gfx::Size size;
	gfx::GpuMemoryBuffer::Format format;
	gfx::GpuMemoryBuffer::Usage usage;
	int client_id;
	int surface_id;
	scoped_ptr<gfx::GpuMemoryBuffer> result;
	};

	BrowserGpuMemoryBufferManager::BrowserGpuMemoryBufferManager(
	GpuMemoryBufferFactoryHost* gpu_memory_buffer_factory_host,
	int gpu_client_id)
	: gpu_memory_buffer_factory_host_(gpu_memory_buffer_factory_host),
	gpu_client_id_(gpu_client_id),
	weak_ptr_factory_(this) {
	DCHECK(!g_gpu_memory_buffer_manager);
	g_gpu_memory_buffer_manager = this;
	}

	BrowserGpuMemoryBufferManager::~BrowserGpuMemoryBufferManager() {
	g_gpu_memory_buffer_manager = nullptr;
	}

	// static
	BrowserGpuMemoryBufferManager* BrowserGpuMemoryBufferManager::current() {
	return g_gpu_memory_buffer_manager;
	}

	scoped_ptr<gfx::GpuMemoryBuffer>
	BrowserGpuMemoryBufferManager::AllocateGpuMemoryBuffer(
	const gfx::Size& size,
	gfx::GpuMemoryBuffer::Format format,
	gfx::GpuMemoryBuffer::Usage usage) {
	return AllocateGpuMemoryBufferCommon(size, format, usage, 0);
	}

	scoped_ptr<gfx::GpuMemoryBuffer>
	BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferForScanout(
	const gfx::Size& size,
	gfx::GpuMemoryBuffer::Format format,
	int32 surface_id) {
	DCHECK_GT(surface_id, 0);
	return AllocateGpuMemoryBufferCommon(
	size, format, gfx::GpuMemoryBuffer::SCANOUT, surface_id);
	}

	scoped_ptr<gfx::GpuMemoryBuffer>
	BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferCommon(
	const gfx::Size& size,
	gfx::GpuMemoryBuffer::Format format,
	gfx::GpuMemoryBuffer::Usage usage,
	int32 surface_id) {
	DCHECK(!BrowserThread::CurrentlyOn(BrowserThread::IO));

	// Fallback to shared memory buffer if \|format\| and \|usage\| are not supported
	// by factory.
	if (!gpu_memory_buffer_factory_host_->IsGpuMemoryBufferConfigurationSupported(
	format, usage)) {
	DCHECK(GpuMemoryBufferImplSharedMemory::IsFormatSupported(format));
	DCHECK_EQ(usage, gfx::GpuMemoryBuffer::MAP);
	return GpuMemoryBufferImplSharedMemory::Create(
	g_next_gpu_memory_buffer_id.GetNext(), size, format);
	}

	AllocateGpuMemoryBufferRequest request(size, format, usage, gpu_client_id_,
	surface_id);
	BrowserThread::PostTask(
	BrowserThread::IO,
	FROM_HERE,
	base::Bind(&BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferOnIO,
	base::Unretained(this), // Safe as we wait for result below.
	base::Unretained(&request)));

	// We're blocking the UI thread, which is generally undesirable.
	TRACE_EVENT0("browser",
	"BrowserGpuMemoryBufferManager::AllocateGpuMemoryBuffer");
	base::ThreadRestrictions::ScopedAllowWait allow_wait;
	request.event.Wait();
	return request.result.Pass();
	}

	void BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferForChildProcess(
	const gfx::Size& size,
	gfx::GpuMemoryBuffer::Format format,
	gfx::GpuMemoryBuffer::Usage usage,
	base::ProcessHandle child_process_handle,
	int child_client_id,
	const AllocationCallback& callback) {
	DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));

	gfx::GpuMemoryBufferId new_id = g_next_gpu_memory_buffer_id.GetNext();

	BufferMap& buffers = clients_[child_client_id];
	DCHECK(buffers.find(new_id) == buffers.end());

	// Fallback to shared memory buffer if \|format\| and \|usage\| are not supported
	// by factory.
	if (!gpu_memory_buffer_factory_host_->IsGpuMemoryBufferConfigurationSupported(
	format, usage)) {
	// Early out if we cannot fallback to shared memory buffer.
	if (!GpuMemoryBufferImplSharedMemory::IsFormatSupported(format) \|\|
	!GpuMemoryBufferImplSharedMemory::IsSizeValidForFormat(size, format) \|\|
	usage != gfx::GpuMemoryBuffer::MAP) {
	callback.Run(gfx::GpuMemoryBufferHandle());
	return;
	}

	buffers[new_id] = gfx::SHARED_MEMORY_BUFFER;
	callback.Run(GpuMemoryBufferImplSharedMemory::AllocateForChildProcess(
	new_id, size, format, child_process_handle));
	return;
	}

	// Note: Handling of cases where the child process is removed before the
	// allocation completes is less subtle if we set the buffer type to
	// EMPTY_BUFFER here and verify that this has not changed when allocation
	// completes.
	buffers[new_id] = gfx::EMPTY_BUFFER;

	gpu_memory_buffer_factory_host_->CreateGpuMemoryBuffer(
	new_id, size, format, usage, child_client_id, 0,
	base::Bind(&BrowserGpuMemoryBufferManager::
	GpuMemoryBufferAllocatedForChildProcess,
	weak_ptr_factory_.GetWeakPtr(), child_client_id, callback));
	}

	gfx::GpuMemoryBuffer*
	BrowserGpuMemoryBufferManager::GpuMemoryBufferFromClientBuffer(
	ClientBuffer buffer) {
	return GpuMemoryBufferImpl::FromClientBuffer(buffer);
	}

	void BrowserGpuMemoryBufferManager::SetDestructionSyncPoint(
	gfx::GpuMemoryBuffer* buffer,
	uint32 sync_point) {
	static_cast<GpuMemoryBufferImpl*>(buffer)
	->set_destruction_sync_point(sync_point);
	}

	void BrowserGpuMemoryBufferManager::ChildProcessDeletedGpuMemoryBuffer(
	gfx::GpuMemoryBufferId id,
	base::ProcessHandle child_process_handle,
	int child_client_id,
	uint32 sync_point) {
	DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
	DCHECK(clients_.find(child_client_id) != clients_.end());

	BufferMap& buffers = clients_[child_client_id];

	BufferMap::iterator buffer_it = buffers.find(id);
	if (buffer_it == buffers.end()) {
	LOG(ERROR) << "Invalid GpuMemoryBuffer ID for child process.";
	return;
	}

	// This can happen if a child process managed to trigger a call to this while
	// a buffer is in the process of being allocated.
	if (buffer_it->second == gfx::EMPTY_BUFFER) {
	LOG(ERROR) << "Invalid GpuMemoryBuffer type.";
	return;
	}

	// Buffers allocated using the factory need to be destroyed through the
	// factory.
	if (buffer_it->second != gfx::SHARED_MEMORY_BUFFER) {
	gpu_memory_buffer_factory_host_->DestroyGpuMemoryBuffer(id,
	child_client_id,
	sync_point);
	}

	buffers.erase(buffer_it);
	}

	void BrowserGpuMemoryBufferManager::ProcessRemoved(
	base::ProcessHandle process_handle,
	int client_id) {
	DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));

	ClientMap::iterator client_it = clients_.find(client_id);
	if (client_it == clients_.end())
	return;

	for (auto &buffer_it : client_it->second) {
	// This might happen if buffer is currenlty in the process of being
	// allocated. The buffer will in that case be cleaned up when allocation
	// completes.
	if (buffer_it.second == gfx::EMPTY_BUFFER)
	continue;

	// Skip shared memory buffers as they were not allocated using the factory.
	if (buffer_it.second == gfx::SHARED_MEMORY_BUFFER)
	continue;

	gpu_memory_buffer_factory_host_->DestroyGpuMemoryBuffer(buffer_it.first,
	client_id,
	0);
	}

	clients_.erase(client_it);
	}

	void BrowserGpuMemoryBufferManager::AllocateGpuMemoryBufferOnIO(
	AllocateGpuMemoryBufferRequest* request) {
	// Note: Unretained is safe as this is only used for synchronous allocation
	// from a non-IO thread.
	gpu_memory_buffer_factory_host_->CreateGpuMemoryBuffer(
	g_next_gpu_memory_buffer_id.GetNext(), request->size, request->format,
	request->usage, request->client_id, request->surface_id,
	base::Bind(&BrowserGpuMemoryBufferManager::GpuMemoryBufferAllocatedOnIO,
	base::Unretained(this), base::Unretained(request)));
	}

	void BrowserGpuMemoryBufferManager::GpuMemoryBufferAllocatedOnIO(
	AllocateGpuMemoryBufferRequest* request,
	const gfx::GpuMemoryBufferHandle& handle) {
	DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));

	// Early out if factory failed to allocate the buffer.
	if (handle.is_null()) {
	request->event.Signal();
	return;
	}

	DCHECK_NE(handle.type, gfx::SHARED_MEMORY_BUFFER);
	request->result = GpuMemoryBufferImpl::CreateFromHandle(
	handle,
	request->size,
	request->format,
	base::Bind(&BrowserGpuMemoryBufferManager::GpuMemoryBufferDeleted,
	weak_ptr_factory_.GetWeakPtr(),
	handle.id,
	request->client_id));
	request->event.Signal();
	}

	void BrowserGpuMemoryBufferManager::GpuMemoryBufferDeleted(
	gfx::GpuMemoryBufferId id,
	int client_id,
	uint32 sync_point) {
	gpu_memory_buffer_factory_host_->DestroyGpuMemoryBuffer(id,
	client_id,
	sync_point);
	}

	void BrowserGpuMemoryBufferManager::GpuMemoryBufferAllocatedForChildProcess(
	int child_client_id,
	const AllocationCallback& callback,
	const gfx::GpuMemoryBufferHandle& handle) {
	DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));

	ClientMap::iterator client_it = clients_.find(child_client_id);

	// This can happen if the child process is removed while the buffer is being
	// allocated.
	if (client_it == clients_.end()) {
	if (!handle.is_null()) {
	gpu_memory_buffer_factory_host_->DestroyGpuMemoryBuffer(
	handle.id, child_client_id, 0);
	}
	callback.Run(gfx::GpuMemoryBufferHandle());
	return;
	}

	BufferMap& buffers = client_it->second;

	BufferMap::iterator buffer_it = buffers.find(handle.id);
	DCHECK(buffer_it != buffers.end());
	DCHECK_EQ(buffer_it->second, gfx::EMPTY_BUFFER);

	if (handle.is_null()) {
	buffers.erase(buffer_it);
	callback.Run(gfx::GpuMemoryBufferHandle());
	return;
	}

	// The factory should never return a shared memory backed buffer.
	DCHECK_NE(handle.type, gfx::SHARED_MEMORY_BUFFER);

	// Store the type of this buffer so it can be cleaned up if the child
	// process is removed.
	buffer_it->second = handle.type;

	callback.Run(handle);
	}

	} // namespace content