Segmentation Fault using primusrun on Linux #2546
Comments
|
Do you get any validation errors if you enable the validation layer? |
|
Also see next comment. I think these are spurious. I turned on "VK_LAYER_KHRONOS_validation" I get (short version): vkGetPhysicalDeviceProperties2: pProperties->pNext chain includes a structure with unknown VkStructureType, with the following two VkStructureTypes: VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROVOKING_VERTEX_PROPERTIES_EXT (1000254002) Full messages: primusrun ./target/debug/triangle List of Vulkan debugging layers available to use: Stack trace: #0 0x00007fffefc0f004 in ?? () from /usr/lib/x86_64-linux-gnu/libGLX_nvidia.so.0 |
|
I think those message may be spurious the the crash: If I turn on the layers: "VK_LAYER_KHRONOS_validation","VK_LAYER_PRIMUS_PrimusVK" primusrun ./target/debug/triangle gives the same messages, but then runs correctly. With those same layers, if I then run without the 'primusrun', it fails to find a rendering GPU: ./target/debug/triangle |
Machine also has: Intel(R) HD Graphics 4600 (HSW GT2) (type: IntegratedGpu)
To reproduce the error:
Checkout the master branch from github , build it with 'cargo build' and:
Output:
MESA-INTEL: warning: Haswell Vulkan support is incomplete
Using device: NVIDIA GeForce GT 730M (type: DiscreteGpu)
Segmentation fault
Stack trace:
#0 0x00007fffefc0f004 in ?? () from /usr/lib/x86_64-linux-gnu/libGLX_nvidia.so.0
#1 0x00007fffee783b31 in ?? () from /usr/lib/x86_64-linux-gnu/libnvidia-glcore.so.470.256.02
#2 0x00007fffee7823ea in ?? () from /usr/lib/x86_64-linux-gnu/libnvidia-glcore.so.470.256.02
#3 0x00007fffee7826c8 in ?? () from /usr/lib/x86_64-linux-gnu/libnvidia-glcore.so.470.256.02
#4 0x00007fffee78411e in ?? () from /usr/lib/x86_64-linux-gnu/libnvidia-glcore.so.470.256.02
#5 0x00007fffee7789b5 in ?? () from /usr/lib/x86_64-linux-gnu/libnvidia-glcore.so.470.256.02
#6 0x00007fffee783fff in ?? () from /usr/lib/x86_64-linux-gnu/libnvidia-glcore.so.470.256.02
#7 0x00007ffff72a189c in ?? () from /usr/lib/x86_64-linux-gnu/libvulkan.so.1
#8 0x000055555600716a in vulkano::swapchain::Swapchain::new_inner_unchecked (device=0x5555571c7880, surface=0x55555690d330, create_info=0x7fffffff80a0,
old_swapchain=...) at vulkano/src/swapchain/mod.rs:1117
#9 0x0000555555709f5f in vulkano::swapchain::Swapchain::new_unchecked (device=..., surface=..., create_info=...) at vulkano/src/swapchain/mod.rs:432
#10 0x000055555570a421 in vulkano::swapchain::Swapchain::new (device=..., surface=..., create_info=...) at vulkano/src/swapchain/mod.rs:421
#11 0x000055555578654b in triangle::main () at examples/triangle/main.rs:209
print create_info_vk
$1 = ash::vk::definitions::SwapchainCreateInfoKHR {s_type: ash::vk::enums::StructureType (1000001000), p_next: 0x0, flags: ash::vk::bitflags::SwapchainCreateFlagsKHR (0), surface: ash::vk::definitions::SurfaceKHR (93825018532320), min_image_count: 2, image_format: ash::vk::enums::Format (44), image_color_space: ash::vk::enums::ColorSpaceKHR (0), image_extent: ash::vk::definitions::Extent2D {width: 800, height: 600}, image_array_layers: 1, image_usage: ash::vk::bitflags::ImageUsageFlags (16), image_sharing_mode: ash::vk::enums::SharingMode (0), queue_family_index_count: 0, p_queue_family_indices: 0x0, pre_transform: ash::vk::bitflags::SurfaceTransformFlagsKHR (1), composite_alpha: ash::vk::bitflags::CompositeAlphaFlagsKHR (1), present_mode: ash::vk::enums::PresentModeKHR (2), clipped: 1, old_swapchain: ash::vk::definitions::SwapchainKHR (0), _marker: core::marker::PhantomData<&()>}
SurfaceKHR (93825018532320) is SurfaceKHR (0x555556e691e0)
For comparison, it does work correctly on the internal Intel(R) HD Graphics 4600 (HSW GT2). With break mod.rs:1117:
print create_info_vk
$1 = ash::vk::definitions::SwapchainCreateInfoKHR {s_type: ash::vk::enums::StructureType (1000001000), p_next: 0x0, flags: ash::vk::bitflags::SwapchainCreateFlagsKHR (0), surface: ash::vk::definitions::SurfaceKHR (93825014261712), min_image_count: 3, image_format: ash::vk::enums::Format (50), image_color_space: ash::vk::enums::ColorSpaceKHR (0), image_extent: ash::vk::definitions::Extent2D {width: 800, height: 600}, image_array_layers: 1, image_usage: ash::vk::bitflags::ImageUsageFlags (16), image_sharing_mode: ash::vk::enums::SharingMode (0), queue_family_index_count: 0, p_queue_family_indices: 0x0, pre_transform: ash::vk::bitflags::SurfaceTransformFlagsKHR (1), composite_alpha: ash::vk::bitflags::CompositeAlphaFlagsKHR (1), present_mode: ash::vk::enums::PresentModeKHR (2), clipped: 1, old_swapchain: ash::vk::definitions::SwapchainKHR (0), _marker: core::marker::PhantomData<&()>}
It also works correctly in a C++ implementation with primusrun (so it's not just an Nvidia driver issue):
print createInfo
{sType = VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR, pNext = 0x0, flags = 0, surface = 0x20000000002, minImageCount = 4,
imageFormat = VK_FORMAT_B8G8R8A8_SRGB, imageColorSpace = VK_COLOR_SPACE_SRGB_NONLINEAR_KHR, imageExtent = {width = 1200, height = 800}, imageArrayLayers = 1,
imageUsage = 16, imageSharingMode = VK_SHARING_MODE_EXCLUSIVE, queueFamilyIndexCount = 0, pQueueFamilyIndices = 0x0,
preTransform = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR, compositeAlpha = VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR, presentMode = VK_PRESENT_MODE_FIFO_KHR,
clipped = 1, oldSwapchain = 0x0}
I can provide any more information that is requested.
I think it's odd that surface is a pointer into main memory in the vulkano version, and a special value in the C++ version - that was the only major difference that I could find (the different minImageCount doesn't matter).
The text was updated successfully, but these errors were encountered: