/* * * Copyright (C) 2015 Valve Corporation * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), * to deal in the Software without restriction, including without limitation * the rights to use, copy, modify, merge, publish, distribute, sublicense, * and/or sell copies of the Software, and to permit persons to whom the * Software is furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included * in all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * * Author: Courtney Goeltzenleuchter * Author: Jon Ashburn * */ #include "vk_loader_platform.h" #include "loader.h" #include "vulkan/vk_lunarg_debug_report.h" /* * CreateMsgCallback is global and needs to be * applied to all layers and ICDs. * What happens if a layer is enabled on both the instance chain * as well as the device chain and a call to CreateMsgCallback is made? * Do we need to make sure that each layer / driver only gets called once? * Should a layer implementing support for CreateMsgCallback only be allowed (?) * to live on one chain? Or maybe make it the application's responsibility. * If the app enables DRAW_STATE on at both CreateInstance time and CreateDevice * time, CreateMsgCallback will call the DRAW_STATE layer twice. Once via * the instance chain and once via the device chain. * The loader should only return the DEBUG_REPORT extension as supported * for the GetGlobalExtensionSupport call. That should help eliminate one * duplication. * Since the instance chain requires us iterating over the available ICDs * and each ICD will have it's own unique MsgCallback object we need to * track those objects to give back the right one. * This also implies that the loader has to intercept vkDestroyObject and * if the extension is enabled and the object type is a MsgCallback then * we must translate the object into the proper ICD specific ones. * DestroyObject works on a device chain. Should not be what's destroying * the MsgCallback object. That needs to be an instance thing. So, since * we used an instance to create it, we need a custom Destroy that also * takes an instance. That way we can iterate over the ICDs properly. * Example use: * CreateInstance: DEBUG_REPORT * Loader will create instance chain with enabled extensions. * TODO: Should validation layers be enabled here? If not, they will not be in the instance chain. * fn = GetProcAddr(INSTANCE, "vkCreateMsgCallback") -> point to loader's vkCreateMsgCallback * App creates a callback object: fn(..., &MsgCallbackObject1) * Have only established the instance chain so far. Loader will call the instance chain. * Each layer in the instance chain will call down to the next layer, terminating with * the CreateMsgCallback loader terminator function that creates the actual MsgCallbackObject1 object. * The loader CreateMsgCallback terminator will iterate over the ICDs. * Calling each ICD that supports vkCreateMsgCallback and collect answers in icd_msg_callback_map here. * As result is sent back up the chain each layer has opportunity to record the callback operation and * appropriate MsgCallback object. * ... * Any reports matching the flags set in MsgCallbackObject1 will generate the defined callback behavior * in the layer / ICD that initiated that report. * ... * CreateDevice: MemTracker:... * App does not include DEBUG_REPORT as that is a global extension. * TODO: GetExtensionSupport must not report DEBUG_REPORT when using instance. * App MUST include any desired validation layers or they will not participate in the device call chain. * App creates a callback object: fn(..., &MsgCallbackObject2) * Loader's vkCreateMsgCallback is called. * Loader sends call down instance chain - this is a global extension - any validation layer that was * enabled at CreateInstance will be able to register the callback. Loader will iterate over the ICDs and * will record the ICD's version of the MsgCallback2 object here. * ... * Any report will go to the layer's report function and it will check the flags for MsgCallbackObject1 * and MsgCallbackObject2 and take the appropriate action as indicated by the app. * ... * App calls vkDestroyMsgCallback( MsgCallbackObject1 ) * Loader's DestroyMsgCallback is where call starts. DestroyMsgCallback will be sent down instance chain * ending in the loader's DestroyMsgCallback terminator which will iterate over the ICD's destroying each * ICD version of that MsgCallback object and then destroy the loader's version of the object. * Any reports generated after this will only have MsgCallbackObject2 available. */ void debug_report_add_instance_extensions( const struct loader_instance *inst, struct loader_extension_list *ext_list); void debug_report_create_instance( struct loader_instance *ptr_instance, const VkInstanceCreateInfo *pCreateInfo); bool debug_report_instance_gpa( struct loader_instance *ptr_instance, const char* name, void **addr); VKAPI_ATTR VkResult VKAPI_CALL loader_DbgCreateMsgCallback( VkInstance instance, VkFlags msgFlags, const PFN_vkDbgMsgCallback pfnMsgCallback, void* pUserData, VkDbgMsgCallback* pMsgCallback); VKAPI_ATTR VkResult VKAPI_CALL loader_DbgDestroyMsgCallback( VkInstance instance, VkDbgMsgCallback msgCallback);