Forum Discussion
Nullpointer crash during xrCreateSession inside oculus drivers (LibOVRRT64_1.dll) using OpenXR on PC
Hi,
I've integrated OpenXR in my custom game engine. It works fine when I use my Oculus headset using SteamVR as the OpenXR runtime.
However, when using Oculus as the OpenXR runtime, I get a null pointer crash deep inside LibOVRRT64_1.dll, which is being called by OpenXR SDK at xrCreateSession(), which was called by my program.
Here's the relevant part of the callstack:
call stack
Who can I contact at Oculus to help solve this driver crash? Do you want a coredump of the crash?
Thanks!
6 Replies
Thanks for reporting stenyak.
I don't know if a core dump would help very much unless you can supply symbols for your binary too?
Otherwise, maybe you let me know the data supplied in the XrSessionCreateInfo parameter?Ok, just in case, here's the api dump of both xrCreateInstance, and the (crashing) xrCreateSession:
XrResult xrCreateInstance
const XrInstanceCreateInfo* info = 0x00000002250fe208
XrStructureType info->type = 3
const void * info->next = 0x0000000000000000
XrInstanceCreateFlags info->createFlags = 0
XrApplicationInfo info->applicationInfo = 0x00000002250fe220
char* info->applicationInfo->applicationName = redacted
uint32_t info->applicationInfo->applicationVersion = 0x493e0
char* info->applicationInfo->engineName = redacted
uint32_t info->applicationInfo->engineVersion = 0x493e0
XrVersion info->applicationInfo->apiVersion = 0x1000000000017
uint32_t info->enabledApiLayerCount = 0x2
const char* const* info->enabledApiLayerNames = 0x00000002250FE778
const char* const* info->enabledApiLayerNames[0] = XR_APILAYER_LUNARG_core_validation
const char* const* info->enabledApiLayerNames[1] = XR_APILAYER_LUNARG_api_dump
uint32_t info->enabledExtensionCount = 0x1
const char* const* info->enabledExtensionNames = 0x0000017490BA9080
const char* const* info->enabledExtensionNames[0] = XR_KHR_vulkan_enable
XrInstance* instance = 0x00000002250fdab8XrResult xrCreateSession
XrInstance instance = 00007B4400000001
const XrSessionCreateInfo* createInfo = 0x00000002250feac8
XrStructureType createInfo->type = XR_TYPE_SESSION_CREATE_INFO
const void * createInfo->next = 0x00000002250fec90
const XrGraphicsBindingVulkanKHR* createInfo->next = 0x00000002250fec90
XrStructureType createInfo->next->type = XR_TYPE_GRAPHICS_BINDING_VULKAN_KHR
const void * createInfo->next->next = 0x0000000000000000
VkInstance createInfo->next->instance = 000001743FEE10D0
VkPhysicalDevice createInfo->next->physicalDevice = 0000017443EECA20
VkDevice createInfo->next->device = 000001744400E1A0
uint32_t createInfo->next->queueFamilyIndex = 0x0
uint32_t createInfo->next->queueIndex = 0x0
XrSessionCreateFlags createInfo->createFlags = 0
XrSystemId createInfo->systemId = 14
XrSession* session = 0000017444036EC8It's a bit surprising that `queueFamilyIndex` and `queueIndex` are 0 but that shouldn't cause a NULL pointer dereference. I don't think there is much that I can do without a crash dump (with pdb).
For anybody who comes to this issue in future - stenyak and I investigated this issue and found it to be a problem in the application code. The crash happened in the Vulkan Loader because the data from `xrGetVulkanDeviceExtensionsKHR` was not being used when creating the VkDevice.
Thanks,
JohnThanks John, was about to come here to post the solution myself!
So I was indeed not loading all the required Vulkan extensions, which made the runtime unable to work as intended, leading to that crash.
