gecko/gfx/layers/ipc/ISurfaceAllocator.cpp

329 lines
10 KiB
C++

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*-
* vim: sw=2 ts=8 et :
*/
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#include "ISurfaceAllocator.h"
#include <sys/types.h> // for int32_t
#include "gfx2DGlue.h" // for IntSize
#include "gfxASurface.h" // for gfxASurface, etc
#include "gfxPlatform.h" // for gfxPlatform, gfxImageFormat
#include "gfxSharedImageSurface.h" // for gfxSharedImageSurface
#include "mozilla/Assertions.h" // for MOZ_ASSERT, etc
#include "mozilla/Atomics.h" // for PrimitiveIntrinsics
#include "mozilla/ipc/SharedMemory.h" // for SharedMemory, etc
#include "mozilla/layers/LayersSurfaces.h" // for SurfaceDescriptor, etc
#include "ShadowLayerUtils.h"
#include "mozilla/mozalloc.h" // for operator delete[], etc
#include "nsAutoPtr.h" // for nsRefPtr, getter_AddRefs, etc
#include "nsDebug.h" // for NS_RUNTIMEABORT
#include "nsXULAppAPI.h" // for XRE_GetProcessType, etc
#ifdef DEBUG
#include "prenv.h"
#endif
using namespace mozilla::ipc;
namespace mozilla {
namespace layers {
NS_IMPL_ISUPPORTS1(GfxMemoryImageReporter, nsIMemoryReporter)
mozilla::Atomic<size_t> GfxMemoryImageReporter::sAmount(0);
mozilla::ipc::SharedMemory::SharedMemoryType OptimalShmemType()
{
return mozilla::ipc::SharedMemory::TYPE_BASIC;
}
bool
IsSurfaceDescriptorValid(const SurfaceDescriptor& aSurface)
{
return aSurface.type() != SurfaceDescriptor::T__None &&
aSurface.type() != SurfaceDescriptor::Tnull_t;
}
ISurfaceAllocator::~ISurfaceAllocator()
{
// Check if we're not leaking..
MOZ_ASSERT(mUsedShmems.empty());
}
void
ISurfaceAllocator::Finalize()
{
ShrinkShmemSectionHeap();
}
bool
ISurfaceAllocator::AllocSharedImageSurface(const gfx::IntSize& aSize,
gfxContentType aContent,
gfxSharedImageSurface** aBuffer)
{
mozilla::ipc::SharedMemory::SharedMemoryType shmemType = OptimalShmemType();
gfxImageFormat format = gfxPlatform::GetPlatform()->OptimalFormatForContent(aContent);
nsRefPtr<gfxSharedImageSurface> back =
gfxSharedImageSurface::CreateUnsafe(this,
gfx::ThebesIntSize(aSize),
format,
shmemType);
if (!back)
return false;
*aBuffer = nullptr;
back.swap(*aBuffer);
return true;
}
bool
ISurfaceAllocator::AllocSurfaceDescriptor(const gfx::IntSize& aSize,
gfxContentType aContent,
SurfaceDescriptor* aBuffer)
{
return AllocSurfaceDescriptorWithCaps(aSize, aContent, DEFAULT_BUFFER_CAPS, aBuffer);
}
bool
ISurfaceAllocator::AllocSurfaceDescriptorWithCaps(const gfx::IntSize& aSize,
gfxContentType aContent,
uint32_t aCaps,
SurfaceDescriptor* aBuffer)
{
bool tryPlatformSurface = true;
#ifdef DEBUG
tryPlatformSurface = !PR_GetEnv("MOZ_LAYERS_FORCE_SHMEM_SURFACES");
#endif
if (tryPlatformSurface &&
PlatformAllocSurfaceDescriptor(aSize, aContent, aCaps, aBuffer)) {
return true;
}
if (XRE_GetProcessType() == GeckoProcessType_Default) {
gfxImageFormat format =
gfxPlatform::GetPlatform()->OptimalFormatForContent(aContent);
int32_t stride = gfxASurface::FormatStrideForWidth(format, aSize.width);
uint8_t *data = new (std::nothrow) uint8_t[stride * aSize.height];
if (!data) {
return false;
}
GfxMemoryImageReporter::DidAlloc(data);
#ifdef XP_MACOSX
// Workaround a bug in Quartz where drawing an a8 surface to another a8
// surface with OPERATOR_SOURCE still requires the destination to be clear.
if (format == gfxImageFormat::A8) {
memset(data, 0, stride * aSize.height);
}
#endif
*aBuffer = MemoryImage((uintptr_t)data, aSize, stride, format);
return true;
}
nsRefPtr<gfxSharedImageSurface> buffer;
if (!AllocSharedImageSurface(aSize, aContent,
getter_AddRefs(buffer))) {
return false;
}
*aBuffer = buffer->GetShmem();
return true;
}
/* static */ bool
ISurfaceAllocator::IsShmem(SurfaceDescriptor* aSurface)
{
return aSurface && (aSurface->type() == SurfaceDescriptor::TShmem ||
aSurface->type() == SurfaceDescriptor::TYCbCrImage ||
aSurface->type() == SurfaceDescriptor::TRGBImage);
}
void
ISurfaceAllocator::DestroySharedSurface(SurfaceDescriptor* aSurface)
{
MOZ_ASSERT(aSurface);
if (!aSurface) {
return;
}
if (!IPCOpen()) {
return;
}
if (PlatformDestroySharedSurface(aSurface)) {
return;
}
switch (aSurface->type()) {
case SurfaceDescriptor::TShmem:
DeallocShmem(aSurface->get_Shmem());
break;
case SurfaceDescriptor::TYCbCrImage:
DeallocShmem(aSurface->get_YCbCrImage().data());
break;
case SurfaceDescriptor::TRGBImage:
DeallocShmem(aSurface->get_RGBImage().data());
break;
case SurfaceDescriptor::TSurfaceDescriptorD3D9:
case SurfaceDescriptor::TSurfaceDescriptorDIB:
case SurfaceDescriptor::TSurfaceDescriptorD3D10:
break;
case SurfaceDescriptor::TMemoryImage:
GfxMemoryImageReporter::WillFree((uint8_t*)aSurface->get_MemoryImage().data());
delete [] (uint8_t*)aSurface->get_MemoryImage().data();
break;
case SurfaceDescriptor::Tnull_t:
case SurfaceDescriptor::T__None:
break;
default:
NS_RUNTIMEABORT("surface type not implemented!");
}
*aSurface = SurfaceDescriptor();
}
#if !defined(MOZ_HAVE_PLATFORM_SPECIFIC_LAYER_BUFFERS)
bool
ISurfaceAllocator::PlatformAllocSurfaceDescriptor(const gfx::IntSize&,
gfxContentType,
uint32_t,
SurfaceDescriptor*)
{
return false;
}
#endif
// XXX - We should actually figure out the minimum shmem allocation size on
// a certain platform and use that.
const uint32_t sShmemPageSize = 4096;
const uint32_t sSupportedBlockSize = 4;
enum AllocationStatus
{
STATUS_ALLOCATED,
STATUS_FREED
};
struct ShmemSectionHeapHeader
{
Atomic<uint32_t> mTotalBlocks;
Atomic<uint32_t> mAllocatedBlocks;
};
struct ShmemSectionHeapAllocation
{
Atomic<uint32_t> mStatus;
uint32_t mSize;
};
bool
ISurfaceAllocator::AllocShmemSection(size_t aSize, mozilla::layers::ShmemSection* aShmemSection)
{
// For now we only support sizes of 4. If we want to support different sizes
// some more complicated bookkeeping should be added.
MOZ_ASSERT(aSize == sSupportedBlockSize);
MOZ_ASSERT(aShmemSection);
uint32_t allocationSize = (aSize + sizeof(ShmemSectionHeapAllocation));
for (size_t i = 0; i < mUsedShmems.size(); i++) {
ShmemSectionHeapHeader* header = mUsedShmems[i].get<ShmemSectionHeapHeader>();
if ((header->mAllocatedBlocks + 1) * allocationSize + sizeof(ShmemSectionHeapHeader) < sShmemPageSize) {
aShmemSection->shmem() = mUsedShmems[i];
MOZ_ASSERT(mUsedShmems[i].IsWritable());
break;
}
}
if (!aShmemSection->shmem().IsWritable()) {
ipc::Shmem tmp;
if (!AllocUnsafeShmem(sShmemPageSize, ipc::SharedMemory::TYPE_BASIC, &tmp)) {
return false;
}
ShmemSectionHeapHeader* header = tmp.get<ShmemSectionHeapHeader>();
header->mTotalBlocks = 0;
header->mAllocatedBlocks = 0;
mUsedShmems.push_back(tmp);
aShmemSection->shmem() = tmp;
}
MOZ_ASSERT(aShmemSection->shmem().IsWritable());
ShmemSectionHeapHeader* header = aShmemSection->shmem().get<ShmemSectionHeapHeader>();
uint8_t* heap = aShmemSection->shmem().get<uint8_t>() + sizeof(ShmemSectionHeapHeader);
ShmemSectionHeapAllocation* allocHeader = nullptr;
if (header->mTotalBlocks > header->mAllocatedBlocks) {
// Search for the first available block.
for (size_t i = 0; i < header->mTotalBlocks; i++) {
allocHeader = reinterpret_cast<ShmemSectionHeapAllocation*>(heap);
if (allocHeader->mStatus == STATUS_FREED) {
break;
}
heap += allocationSize;
}
MOZ_ASSERT(allocHeader && allocHeader->mStatus == STATUS_FREED);
MOZ_ASSERT(allocHeader->mSize == sSupportedBlockSize);
} else {
heap += header->mTotalBlocks * allocationSize;
header->mTotalBlocks++;
allocHeader = reinterpret_cast<ShmemSectionHeapAllocation*>(heap);
allocHeader->mSize = aSize;
}
MOZ_ASSERT(allocHeader);
header->mAllocatedBlocks++;
allocHeader->mStatus = STATUS_ALLOCATED;
aShmemSection->size() = aSize;
aShmemSection->offset() = (heap + sizeof(ShmemSectionHeapAllocation)) - aShmemSection->shmem().get<uint8_t>();
ShrinkShmemSectionHeap();
return true;
}
void
ISurfaceAllocator::FreeShmemSection(mozilla::layers::ShmemSection& aShmemSection)
{
MOZ_ASSERT(aShmemSection.size() == sSupportedBlockSize);
MOZ_ASSERT(aShmemSection.offset() < sShmemPageSize - sSupportedBlockSize);
ShmemSectionHeapAllocation* allocHeader =
reinterpret_cast<ShmemSectionHeapAllocation*>(aShmemSection.shmem().get<char>() +
aShmemSection.offset() -
sizeof(ShmemSectionHeapAllocation));
MOZ_ASSERT(allocHeader->mSize == aShmemSection.size());
DebugOnly<bool> success = allocHeader->mStatus.compareExchange(STATUS_ALLOCATED, STATUS_FREED);
// If this fails something really weird is going on.
MOZ_ASSERT(success);
ShmemSectionHeapHeader* header = aShmemSection.shmem().get<ShmemSectionHeapHeader>();
header->mAllocatedBlocks--;
ShrinkShmemSectionHeap();
}
void
ISurfaceAllocator::ShrinkShmemSectionHeap()
{
for (size_t i = 0; i < mUsedShmems.size(); i++) {
ShmemSectionHeapHeader* header = mUsedShmems[i].get<ShmemSectionHeapHeader>();
if (header->mAllocatedBlocks == 0) {
DeallocShmem(mUsedShmems[i]);
// We don't particularly care about order, move the last one in the array
// to this position.
mUsedShmems[i] = mUsedShmems[mUsedShmems.size() - 1];
mUsedShmems.pop_back();
i--;
break;
}
}
}
} // namespace
} // namespace