Stack overflow on the Mac after upgrading to 4.8

Development Rendering
, , , , ,
1

Hi,

We noticed that the game started crashing in one of the testing level after upgrading to 4.8 with a stack overflow in the rendering thread.

The callstack looks something like this:

#0	0x000000011ab9fef2 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:396
#1	0x000000011ab9fed8 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:394
#2	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#3	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#4	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#5	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#6	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#7	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#8	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#9	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#10	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#11	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#12	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#13	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#14	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#15	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#16	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#17	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#18	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#19	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#20	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#21	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#22	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#23	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#24	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#25	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#26	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#27	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#28	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#29	0x000000011ab9bd39 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:242
#30	0x000000011ab9b783 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:232
#31	0x000000011ab9b783 in FTextureLayout3d::AddSurfaceInner(int, unsigned int, unsigned int, unsigned int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:232
#32	0x000000011a9e5d33 in FTextureLayout3d::AddElement(unsigned int&, unsigned int&, unsigned int&, unsigned int, unsigned int, unsigned int) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Engine/Public/TextureLayout3d.h:70
#33	0x000000011a9447e0 in FIndirectLightingCache::AllocateBlock(int, FIntVector&) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Renderer/Private/IndirectLightingCache.cpp:202
#34	0x000000011a94662b in FIndirectLightingCache::CreateAllocation(int, FBoxSphereBounds const&, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Renderer/Private/IndirectLightingCache.cpp:312
#35	0x000000011a950563 in FIndirectLightingCache::UpdateCacheAllocation(FBoxSphereBounds const&, int, bool, FIndirectLightingCacheAllocation*&, TMap<FIntVector, FBlockUpdateInfo, FDefaultSetAllocator, TDefaultMapKeyFuncs<FIntVector, FBlockUpdateInfo, false> >&, TArray<FIndirectLightingCacheAllocation*, FDefaultAllocator>&) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Renderer/Private/IndirectLightingCache.cpp:463
#36	0x000000011a94d97b in FIndirectLightingCache::UpdateCachePrimitive(FScene*, FPrimitiveSceneInfo*, bool, bool, TMap<FIntVector, FBlockUpdateInfo, FDefaultSetAllocator, TDefaultMapKeyFuncs<FIntVector, FBlockUpdateInfo, false> >&, TArray<FIndirectLightingCacheAllocation*, FDefaultAllocator>&) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Renderer/Private/IndirectLightingCache.cpp:514
#37	0x000000011a94a162 in FIndirectLightingCache::UpdateCache(FScene*, FSceneRenderer&, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Renderer/Private/IndirectLightingCache.cpp:395
#38	0x000000011b0e63a1 in FSceneRenderer::PostVisibilityFrameSetup() at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Renderer/Private/SceneVisibility.cpp:2158
#39	0x000000011b0ecfda in FDeferredShadingSceneRenderer::InitViews(FRHICommandListImmediate&) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Renderer/Private/SceneVisibility.cpp:2327
#40	0x000000011a553342 in FDeferredShadingSceneRenderer::Render(FRHICommandListImmediate&) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Renderer/Private/DeferredShadingRenderer.cpp:708
#41	0x000000011af0406c in RenderViewFamily_RenderThread(FRHICommandListImmediate&, FSceneRenderer*) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Renderer/Private/SceneRendering.cpp:1350
#42	0x000000011adb11e4 in FRendererModule::BeginRenderingViewFamily(FCanvas*, FSceneViewFamily*)::EURCMacro_FDrawSceneCommand::DoTask(ENamedThreads::Type, TRefCountPtr<FGraphEvent> const&) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Renderer/Private/SceneRendering.cpp:1441
#43	0x000000011adb3ed7 in TGraphTask<FRendererModule::BeginRenderingViewFamily(FCanvas*, FSceneViewFamily*)::EURCMacro_FDrawSceneCommand>::ExecuteTask(TArray<FBaseGraphTask*, FDefaultAllocator>&, ENamedThreads::Type) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Core/Public/Async/TaskGraphInterfaces.h:753
#44	0x000000010041a170 in FBaseGraphTask::Execute(TArray<FBaseGraphTask*, FDefaultAllocator>&, ENamedThreads::Type) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Core/Public/Async/TaskGraphInterfaces.h:308
#45	0x00000001004103a9 in FTaskThread::ProcessTasks(int, bool) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Core/Private/Async/TaskGraph.cpp:428
#46	0x0000000100409ed9 in FTaskThread::ProcessTasksUntilQuit(int) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Core/Private/Async/TaskGraph.cpp:271
#47	0x00000001003fadfd in FTaskGraphImplementation::ProcessThreadUntilRequestReturn(ENamedThreads::Type) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Core/Private/Async/TaskGraph.cpp:946
#48	0x000000010c2c912f in RenderingThreadMain(FEvent*) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/RenderCore/Private/RenderingThread.cpp:283
#49	0x000000010c3ac3ed in FRenderingThread::Run() at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/RenderCore/Private/RenderingThread.cpp:404
#50	0x000000010042575d in FRunnableThreadPThread::Run() at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Core/Private/HAL/PThreadRunnableThread.cpp:26
#51	0x00000001003e4879 in FRunnableThreadPThread::_ThreadProc(void*) at /Users/wanghao/work/UnrealEngine/Engine/Source/Runtime/Core/Private/GenericPlatform/../HAL/PThreadRunnableThread.h:167
#52	0x00007fff974c0268 in _pthread_body ()
#53	0x00007fff974c01e5 in _pthread_start ()
#54	0x00007fff974be41d in thread_start ()

Increasing the stack size to 4Mb helps but we’re not sure if it’s the right thing to do.

Is this something that Epic knows/is addressing internally?

2

Note that so far this crash only happens in the debug build.

3

I’m also curious where did the 2Mb default stack size come from? Or is that just a number that “seems to work”?

This is in MacPlatformRunnableThread.h:

virtual int GetDefaultStackSize() override
{
	// default is 512 KB, we need more
	return 2 * 1024 * 1024;
}
4

So, we’re also seeing similar stack overflow crash in the game thread in another level. Has something major changed in 4.8 with regard to stack usage?

5

Seems like the huge stack frames are due to temp storage for debug logs coming out of check/log/assert macros. Some LinkerLoad and other functions rely heavily on recursion, and end up allocating hundreds of kilobytes of stack per recursion.

Looks like there were a couple of changes to the logging, but I don’t have time to dig into the details. Epic?

6

We’re working on a proper fix, but in a meantime please increase the default stack size to 4 MB in GetDefaultStackSize() in MacPlatformRunnableThread.h

7

Thanks, that helped for our render thread issues, but we also had to bump the game thread to 16 MiB in CocoaThread.cpp, RunGameThread(…)
[GameThread setStackSize:N]