Xargon/Duckstation
1
0
Fork 0
mirror of https://github.com/RetroDECK/Duckstation.git synced 2024-11-26 23:55:40 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

GPU/HW: Fix crash with oversized writes and sw readback

Browse source
This commit is contained in:
Connor McLaughlin 2021-05-20 14:01:47 +10:00
parent 5f2a340953
commit 245dd5b27a
5 changed files with 100 additions and 77 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

129
src/core/gpu_hw.cpp
View file

@ -1092,30 +1092,6 @@ void GPU_HW::ResetBatchVertexDepth()
m_current_depth = 1;
}
void GPU_HW::FillBackendCommandParameters(GPUBackendCommand* cmd) const
{
cmd->params.bits = 0;
cmd->params.check_mask_before_draw = m_GPUSTAT.check_mask_before_draw;
cmd->params.set_mask_while_drawing = m_GPUSTAT.set_mask_while_drawing;
cmd->params.active_line_lsb = m_crtc_state.active_line_lsb;
cmd->params.interlaced_rendering = m_GPUSTAT.SkipDrawingToActiveField();
}
void GPU_HW::FillDrawCommand(GPUBackendDrawCommand* cmd, GPURenderCommand rc) const
{
FillBackendCommandParameters(cmd);
cmd->rc.bits = rc.bits;
cmd->draw_mode.bits = m_draw_mode.mode_reg.bits;
cmd->palette.bits = m_draw_mode.palette_reg;
cmd->window = m_draw_mode.texture_window;
}
void GPU_HW::HandleVRAMReadWithSoftwareRenderer(u32 x, u32 y, u32 width, u32 height)
{
DebugAssert(m_sw_renderer);
m_sw_renderer->Sync(false);
}
void GPU_HW::UpdateSoftwareRenderer(bool copy_vram_from_hw)
{
const bool current_enabled = (m_sw_renderer != nullptr);
@ -1154,22 +1130,75 @@ void GPU_HW::UpdateSoftwareRenderer(bool copy_vram_from_hw)
m_vram_ptr = m_sw_renderer->GetVRAM();
}
void GPU_HW::FillBackendCommandParameters(GPUBackendCommand* cmd) const
{
cmd->params.bits = 0;
cmd->params.check_mask_before_draw = m_GPUSTAT.check_mask_before_draw;
cmd->params.set_mask_while_drawing = m_GPUSTAT.set_mask_while_drawing;
cmd->params.active_line_lsb = m_crtc_state.active_line_lsb;
cmd->params.interlaced_rendering = m_GPUSTAT.SkipDrawingToActiveField();
}
void GPU_HW::FillDrawCommand(GPUBackendDrawCommand* cmd, GPURenderCommand rc) const
{
FillBackendCommandParameters(cmd);
cmd->rc.bits = rc.bits;
cmd->draw_mode.bits = m_draw_mode.mode_reg.bits;
cmd->palette.bits = m_draw_mode.palette_reg;
cmd->window = m_draw_mode.texture_window;
}
void GPU_HW::ReadSoftwareRendererVRAM(u32 x, u32 y, u32 width, u32 height)
{
DebugAssert(m_sw_renderer);
m_sw_renderer->Sync(false);
}
void GPU_HW::UpdateSoftwareRendererVRAM(u32 x, u32 y, u32 width, u32 height, const void* data, bool set_mask,
bool check_mask)
{
const u32 num_words = width * height;
GPUBackendUpdateVRAMCommand* cmd = m_sw_renderer->NewUpdateVRAMCommand(num_words);
FillBackendCommandParameters(cmd);
cmd->params.set_mask_while_drawing = set_mask;
cmd->params.check_mask_before_draw = check_mask;
cmd->x = static_cast<u16>(x);
cmd->y = static_cast<u16>(y);
cmd->width = static_cast<u16>(width);
cmd->height = static_cast<u16>(height);
std::memcpy(cmd->data, data, sizeof(u16) * num_words);
m_sw_renderer->PushCommand(cmd);
}
void GPU_HW::FillSoftwareRendererVRAM(u32 x, u32 y, u32 width, u32 height, u32 color)
{
GPUBackendFillVRAMCommand* cmd = m_sw_renderer->NewFillVRAMCommand();
FillBackendCommandParameters(cmd);
cmd->x = static_cast<u16>(x);
cmd->y = static_cast<u16>(y);
cmd->width = static_cast<u16>(width);
cmd->height = static_cast<u16>(height);
cmd->color = color;
m_sw_renderer->PushCommand(cmd);
}
void GPU_HW::CopySoftwareRendererVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height)
{
GPUBackendCopyVRAMCommand* cmd = m_sw_renderer->NewCopyVRAMCommand();
FillBackendCommandParameters(cmd);
cmd->src_x = static_cast<u16>(src_x);
cmd->src_y = static_cast<u16>(src_y);
cmd->dst_x = static_cast<u16>(dst_x);
cmd->dst_y = static_cast<u16>(dst_y);
cmd->width = static_cast<u16>(width);
cmd->height = static_cast<u16>(height);
m_sw_renderer->PushCommand(cmd);
}
void GPU_HW::FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color)
{
IncludeVRAMDirtyRectangle(
Common::Rectangle<u32>::FromExtents(x, y, width, height).Clamped(0, 0, VRAM_WIDTH, VRAM_HEIGHT));
if (m_sw_renderer)
{
GPUBackendFillVRAMCommand* cmd = m_sw_renderer->NewFillVRAMCommand();
FillBackendCommandParameters(cmd);
cmd->x = static_cast<u16>(x);
cmd->y = static_cast<u16>(y);
cmd->width = static_cast<u16>(width);
cmd->height = static_cast<u16>(height);
cmd->color = color;
m_sw_renderer->PushCommand(cmd);
}
}
void GPU_HW::UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* data, bool set_mask, bool check_mask)
@ -1182,21 +1211,6 @@ void GPU_HW::UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* data, b
// set new vertex counter since we want this to take into consideration previous masked pixels
m_current_depth++;
}
if (m_sw_renderer)
{
const u32 num_words = width * height;
GPUBackendUpdateVRAMCommand* cmd = m_sw_renderer->NewUpdateVRAMCommand(num_words);
FillBackendCommandParameters(cmd);
cmd->params.set_mask_while_drawing = set_mask;
cmd->params.check_mask_before_draw = check_mask;
cmd->x = static_cast<u16>(x);
cmd->y = static_cast<u16>(y);
cmd->width = static_cast<u16>(width);
cmd->height = static_cast<u16>(height);
std::memcpy(cmd->data, data, sizeof(u16) * num_words);
m_sw_renderer->PushCommand(cmd);
}
}
void GPU_HW::CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height)
@ -1209,19 +1223,6 @@ void GPU_HW::CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32
// set new vertex counter since we want this to take into consideration previous masked pixels
m_current_depth++;
}
if (m_sw_renderer)
{
GPUBackendCopyVRAMCommand* cmd = m_sw_renderer->NewCopyVRAMCommand();
FillBackendCommandParameters(cmd);
cmd->src_x = static_cast<u16>(src_x);
cmd->src_y = static_cast<u16>(src_y);
cmd->dst_x = static_cast<u16>(dst_x);
cmd->dst_y = static_cast<u16>(dst_y);
cmd->width = static_cast<u16>(width);
cmd->height = static_cast<u16>(height);
m_sw_renderer->PushCommand(cmd);
}
}
void GPU_HW::DispatchRenderCommand()

6
src/core/gpu_hw.h
View file

@ -262,8 +262,12 @@ protected:
void FillBackendCommandParameters(GPUBackendCommand* cmd) const;
void FillDrawCommand(GPUBackendDrawCommand* cmd, GPURenderCommand rc) const;
void HandleVRAMReadWithSoftwareRenderer(u32 x, u32 y, u32 width, u32 height);
void UpdateSoftwareRenderer(bool copy_vram_from_hw);
void ReadSoftwareRendererVRAM(u32 x, u32 y, u32 width, u32 height);
void UpdateSoftwareRendererVRAM(u32 x, u32 y, u32 width, u32 height, const void* data, bool set_mask,
bool check_mask);
void FillSoftwareRendererVRAM(u32 x, u32 y, u32 width, u32 height, u32 color);
void CopySoftwareRendererVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height);
void FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color) override;
void UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* data, bool set_mask, bool check_mask) override;

14
src/core/gpu_hw_d3d11.cpp
View file

@ -949,7 +949,7 @@ void GPU_HW_D3D11::ReadVRAM(u32 x, u32 y, u32 width, u32 height)
{
if (IsUsingSoftwareRendererForReadbacks())
{
HandleVRAMReadWithSoftwareRenderer(x, y, width, height);
ReadSoftwareRendererVRAM(x, y, width, height);
return;
}
@ -988,6 +988,9 @@ void GPU_HW_D3D11::ReadVRAM(u32 x, u32 y, u32 width, u32 height)
void GPU_HW_D3D11::FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color)
{
if (IsUsingSoftwareRendererForReadbacks())
FillSoftwareRendererVRAM(x, y, width, height, color);
if ((x + width) > VRAM_WIDTH || (y + height) > VRAM_HEIGHT)
{
// CPU round trip if oversized for now.
@ -1015,6 +1018,9 @@ void GPU_HW_D3D11::FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color)
void GPU_HW_D3D11::UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* data, bool set_mask, bool check_mask)
{
if (IsUsingSoftwareRendererForReadbacks())
UpdateSoftwareRendererVRAM(x, y, width, height, data, set_mask, check_mask);
const Common::Rectangle<u32> bounds = GetVRAMTransferBounds(x, y, width, height);
GPU_HW::UpdateVRAM(bounds.left, bounds.top, bounds.GetWidth(), bounds.GetHeight(), data, set_mask, check_mask);
@ -1050,11 +1056,11 @@ void GPU_HW_D3D11::UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* d
void GPU_HW_D3D11::CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height)
{
if (IsUsingSoftwareRendererForReadbacks())
CopySoftwareRendererVRAM(src_x, src_y, dst_x, dst_y, width, height);
if (UseVRAMCopyShader(src_x, src_y, dst_x, dst_y, width, height) || IsUsingMultisampling())
{
if (IsUsingSoftwareRendererForReadbacks())
GPU_HW::CopyVRAM(src_x, src_y, dst_x, dst_y, width, height);
const Common::Rectangle<u32> src_bounds = GetVRAMTransferBounds(src_x, src_y, width, height);
const Common::Rectangle<u32> dst_bounds = GetVRAMTransferBounds(dst_x, dst_y, width, height);
if (m_vram_dirty_rect.Intersects(src_bounds))

14
src/core/gpu_hw_opengl.cpp
View file

@ -984,7 +984,7 @@ void GPU_HW_OpenGL::ReadVRAM(u32 x, u32 y, u32 width, u32 height)
{
if (IsUsingSoftwareRendererForReadbacks())
{
HandleVRAMReadWithSoftwareRenderer(x, y, width, height);
ReadSoftwareRendererVRAM(x, y, width, height);
return;
}
@ -1019,6 +1019,9 @@ void GPU_HW_OpenGL::ReadVRAM(u32 x, u32 y, u32 width, u32 height)
void GPU_HW_OpenGL::FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color)
{
if (IsUsingSoftwareRendererForReadbacks())
FillSoftwareRendererVRAM(x, y, width, height, color);
if ((x + width) > VRAM_WIDTH || (y + height) > VRAM_HEIGHT)
{
// CPU round trip if oversized for now.
@ -1066,6 +1069,9 @@ void GPU_HW_OpenGL::FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color)
void GPU_HW_OpenGL::UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* data, bool set_mask, bool check_mask)
{
if (IsUsingSoftwareRendererForReadbacks())
UpdateSoftwareRendererVRAM(x, y, width, height, data, set_mask, check_mask);
const Common::Rectangle<u32> bounds = GetVRAMTransferBounds(x, y, width, height);
GPU_HW::UpdateVRAM(bounds.left, bounds.top, bounds.GetWidth(), bounds.GetHeight(), data, set_mask, check_mask);
@ -1182,15 +1188,15 @@ void GPU_HW_OpenGL::UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void*
void GPU_HW_OpenGL::CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height)
{
if (IsUsingSoftwareRendererForReadbacks())
CopySoftwareRendererVRAM(src_x, src_y, dst_x, dst_y, width, height);
const Common::Rectangle<u32> dst_bounds = GetVRAMTransferBounds(dst_x, dst_y, width, height);
const Common::Rectangle<u32> src_bounds = GetVRAMTransferBounds(src_x, src_y, width, height);
const bool src_dirty = m_vram_dirty_rect.Intersects(src_bounds);
if (UseVRAMCopyShader(src_x, src_y, dst_x, dst_y, width, height))
{
if (IsUsingSoftwareRendererForReadbacks())
GPU_HW::CopyVRAM(src_x, src_y, dst_x, dst_y, width, height);
if (src_dirty)
UpdateVRAMReadTexture();
IncludeVRAMDirtyRectangle(dst_bounds);

14
src/core/gpu_hw_vulkan.cpp
View file

@ -1440,7 +1440,7 @@ void GPU_HW_Vulkan::ReadVRAM(u32 x, u32 y, u32 width, u32 height)
{
if (IsUsingSoftwareRendererForReadbacks())
{
HandleVRAMReadWithSoftwareRenderer(x, y, width, height);
ReadSoftwareRendererVRAM(x, y, width, height);
return;
}
@ -1489,6 +1489,9 @@ void GPU_HW_Vulkan::ReadVRAM(u32 x, u32 y, u32 width, u32 height)
void GPU_HW_Vulkan::FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color)
{
if (IsUsingSoftwareRendererForReadbacks())
FillSoftwareRendererVRAM(x, y, width, height, color);
if ((x + width) > VRAM_WIDTH || (y + height) > VRAM_HEIGHT)
{
// CPU round trip if oversized for now.
@ -1522,6 +1525,9 @@ void GPU_HW_Vulkan::FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color)
void GPU_HW_Vulkan::UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* data, bool set_mask, bool check_mask)
{
if (IsUsingSoftwareRendererForReadbacks())
UpdateSoftwareRendererVRAM(x, y, width, height, data, set_mask, check_mask);
const Common::Rectangle<u32> bounds = GetVRAMTransferBounds(x, y, width, height);
GPU_HW::UpdateVRAM(bounds.left, bounds.top, bounds.GetWidth(), bounds.GetHeight(), data, set_mask, check_mask);
@ -1576,11 +1582,11 @@ void GPU_HW_Vulkan::UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void*
void GPU_HW_Vulkan::CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height)
{
if (IsUsingSoftwareRendererForReadbacks())
CopySoftwareRendererVRAM(src_x, src_y, dst_x, dst_y, width, height);
if (UseVRAMCopyShader(src_x, src_y, dst_x, dst_y, width, height) || IsUsingMultisampling())
{
if (IsUsingSoftwareRendererForReadbacks())
GPU_HW::CopyVRAM(src_x, src_y, dst_x, dst_y, width, height);
const Common::Rectangle<u32> src_bounds = GetVRAMTransferBounds(src_x, src_y, width, height);
const Common::Rectangle<u32> dst_bounds = GetVRAMTransferBounds(dst_x, dst_y, width, height);
if (m_vram_dirty_rect.Intersects(src_bounds))