WDDM initialization problem for windows7
in [Device Drivers]
|
From: nan_hsing Chang on 1 Sep 2009 02:41 I write the display WDDM driver for windwos7. Windwos stopped my driver during initialization. The DxgkDdiStartDevice callback has be called, then DxgkDdiQueryChildRelations and dxgkQueryAdapterInfo. In dxgkQueryAdapterInfo function, I fill DXGK_DRIVERCAPS structure and return STATUS_SUCCESS. But it can't call DxgkddiCreateDevice, it call DxgkStopDevice instead. At last, my driver has been removed. Could someone give me suggestion? Should I report some caps in the DXGK_DRIVERCAPS structure for windows7. Or My user mode driver can't be load. I find my user mode driver can't be call and I don't know why.
From: Ivan Brugiolo [MSFT] on 1 Sep 2009 12:12 Is this a real device backed by a real piece of hardware, or, some virtual-bus enumerated device ? DxgKrnl.sys is extremely picky about hardware resource requirements for your device, and, it requires a real hardware interrput (CmResourceTypeInterrupt) and a piece of memory (CmResourceTypeMemory), and it parses the CM_RESOURCE_LIST to register the interrput for you. That said, if you made-it past DxgkDdiStartDevice, it looks like that part is OK. DxgkDdiQueryAdapterInfo is supposed to called 3 times during initialization: One for DXGKQAITYPE_DRIVERCAPS Twice for DXGKQAITYPE_QUERYSEGMENT. QuerySegment is again really picky about flags, since it tries to make sense of Memory-Space Segments, Aperture Segments and (for legacy AGP) AGP-Aperture segments. The general expectations for your exercise are: -1- You get past DxgkDdiQueryAdapterInfo, and, the Video-Memory-Manager gets intialized for your DXGADAPTER -2- Required registry information is retrieved and parsed correctly -3- Monitor Modes are requested for non hot-pluggable DxgkChild -4- DxgkDdiEnumVidPnCofuncModality is called hundreds to thousands of times, until the topology for that adapter is `constrained` and `made functional`. If the Gdi view of the VidPn Source is part of the desktop, then: -5- The Cdd kernel-mode DXGDEVICE is created -6- Adapter affine allocations (or Stardard Allocations) Shared and Shadow are created -7- Allocations are pinned, and your DMA completion routine will confirm Paging-Buffers -8- You will get some Gdi `Present` (or Blt) from the Shadow to the Shared If the desktop compositor is in a configuration where it can compose all of the Gdi views, or if there is a DX application, then you may see your User-mode driver to be loaded, and possibly invoked. -- This posting is provided "AS IS" with no warranties, and confers no rights. Use of any included script samples are subject to the terms specified at http://www.microsoft.com/info/cpyright.htm "nan_hsing Chang" <nhchang03(a)gmail.com> wrote in message news:7dee95a0-9b81-46c0-99bb-f2c8be96b46f(a)z4g2000prh.googlegroups.com... >I write the display WDDM driver for windwos7. Windwos stopped my > driver during initialization. The DxgkDdiStartDevice callback has be > called, then DxgkDdiQueryChildRelations and dxgkQueryAdapterInfo. In > dxgkQueryAdapterInfo function, I fill DXGK_DRIVERCAPS structure and > return STATUS_SUCCESS. But it can't call DxgkddiCreateDevice, it call > DxgkStopDevice instead. At last, my driver has been removed. > > Could someone give me suggestion? Should I report some caps in the > DXGK_DRIVERCAPS structure for windows7. Or My user mode driver can't > be load. I find my user mode driver can't be call and I don't know why.
From: nan_hsing Chang on 2 Sep 2009 04:03 Thank for your replay. But I meet a strange situation. Original, I build my driver in WDK 6001.18001 and the function flow is as you says that DxgkDdiQueryAdapterInfo is called 3 times. But when I build in WDK 7600.16385.0, DxgkDdiQueryAdapterInfo is called only one time for DXGKQAITYPE_DRIVERCAPS, then the DxgkStopDevice is called, then DxgkRemoveDevice. Do I miss some settings for WDK 7600 environment? Because I can't find video driver sample(ex, r2000) in WDK 7600, I can't try sample code. BTW, I write display driver for a real device. Ivan Brugiolo [MSFT] wrote: > Is this a real device backed by a real piece of hardware, or, some > virtual-bus enumerated device ? > DxgKrnl.sys is extremely picky about hardware resource requirements > for your device, and, it requires a real hardware interrput > (CmResourceTypeInterrupt) and a piece of memory (CmResourceTypeMemory), > and it parses the CM_RESOURCE_LIST to register the interrput for you. > > That said, if you made-it past DxgkDdiStartDevice, > it looks like that part is OK. > > DxgkDdiQueryAdapterInfo is supposed to called 3 times during initialization: > One for DXGKQAITYPE_DRIVERCAPS > Twice for DXGKQAITYPE_QUERYSEGMENT. > QuerySegment is again really picky about flags, since it tries to make sense > of Memory-Space Segments, Aperture Segments > and (for legacy AGP) AGP-Aperture segments. > > The general expectations for your exercise are: > > -1- You get past DxgkDdiQueryAdapterInfo, and, > the Video-Memory-Manager gets intialized for your DXGADAPTER > -2- Required registry information is retrieved and parsed correctly > -3- Monitor Modes are requested for non hot-pluggable DxgkChild > -4- DxgkDdiEnumVidPnCofuncModality is called hundreds to thousands of times, > until the topology for that adapter is `constrained` and `made functional`. > > If the Gdi view of the VidPn Source is part of the desktop, then: > > -5- The Cdd kernel-mode DXGDEVICE is created > -6- Adapter affine allocations (or Stardard Allocations) Shared and Shadow > are created > -7- Allocations are pinned, and your DMA completion routine will confirm > Paging-Buffers > -8- You will get some Gdi `Present` (or Blt) from the Shadow to the Shared > > If the desktop compositor is in a configuration where it can compose > all of the Gdi views, or if there is a DX application, then you may see > your User-mode driver to be loaded, and possibly invoked. > > -- > This posting is provided "AS IS" with no warranties, and confers no rights. > Use of any included script samples are subject to the terms specified at > http://www.microsoft.com/info/cpyright.htm > > > "nan_hsing Chang" <nhchang03(a)gmail.com> wrote in message > news:7dee95a0-9b81-46c0-99bb-f2c8be96b46f(a)z4g2000prh.googlegroups.com... > >I write the display WDDM driver for windwos7. Windwos stopped my > > driver during initialization. The DxgkDdiStartDevice callback has be > > called, then DxgkDdiQueryChildRelations and dxgkQueryAdapterInfo. In > > dxgkQueryAdapterInfo function, I fill DXGK_DRIVERCAPS structure and > > return STATUS_SUCCESS. But it can't call DxgkddiCreateDevice, it call > > DxgkStopDevice instead. At last, my driver has been removed. > > > > Could someone give me suggestion? Should I report some caps in the > > DXGK_DRIVERCAPS structure for windows7. Or My user mode driver can't > > be load. I find my user mode driver can't be call and I don't know why.
From: Ivan Brugiolo [MSFT] on 2 Sep 2009 05:26 What do you pass as DXGKDDI_INTERFACE_xxx in DRIVER_INITIALIZATION_DATA ? There is a coupling between that value and what the DRIVERCAPS expexts. Also, what do you expect as the result your exercise ? Compile a WDDM 1.0 driver with the Win7 compiler ? Compile a Win7 only WDDM 1.1 driver ? A combined 1.0/1.1 package ? I believe that you are mixing Win7 only #defines unforced by the build environment with Legacy - Vista typedefs and other declarations. I would recommend stepping out of your miniport function, and, see in the disassembly where DXGADAPTER::Initialize returns STATUS_REVISION_MISMATCH. switch(pQueryAdapterInfo->Type) { case DXGKQAITYPE_DRIVERCAPS: { DXGK_DRIVERCAPS * AdapterCaps = (DXGK_DRIVERCAPS *)pQueryAdapterInfo->pOutputData; AdapterCaps->HighestAcceptableAddress.QuadPart = 0xFFFFFFFFFFF; //PHYSICAL_ADDRESS // // DMA buffer splitting support // AdapterCaps->MaxAllocationListSlotId = 2; AdapterCaps->ApertureSegmentCommitLimit = 0x10000000; // 256Meg AdapterCaps->MaxPointerWidth = 0x40; // 64 pixel AdapterCaps->MaxPointerHeight = 0x40; // 64 pixel //DXGK_POINTERFLAGS AdapterCaps->PointerCaps.Color = TRUE; AdapterCaps->PointerCaps.Reserved = 0; AdapterCaps->InterruptMessageNumber = 0; AdapterCaps->NumberOfSwizzlingRanges = 0x2; AdapterCaps->MaxOverlays = 0x0; //DXGK_GAMMARAMPCAPS AdapterCaps->GammaRampCaps.Gamma_Rgb256x3x16 = TRUE; //DXGK_PRESENTATIONCAPS AdapterCaps->PresentationCaps.Value = 0; AdapterCaps->PresentationCaps.SupportKernelModeCommandBuffer = TRUE; AdapterCaps->PresentationCaps.DriverSupportsCddDwmInterop = TRUE; AdapterCaps->PresentationCaps.SupportAllBltRops = TRUE; AdapterCaps->MaxQueuedFlipOnVSync = 0x1; //DXGK_FLIPCAPS AdapterCaps->FlipCaps.FlipOnVSyncWithNoWait = TRUE; AdapterCaps->FlipCaps.FlipOnVSyncMmIo = TRUE; //DXGK_VIDSCHCAPS AdapterCaps->SchedulingCaps.MultiEngineAware = FALSE; AdapterCaps->SchedulingCaps.VSyncPowerSaveAware = TRUE; //DXGK_VIDMMCAPS AdapterCaps->MemoryManagementCaps.OutOfOrderLock = FALSE; AdapterCaps->MemoryManagementCaps.PagingNode = 0; //DXGK_GPUENGINETOPOLOGY AdapterCaps->GpuEngineTopology.NbAsymetricProcessingNodes = 1; //DXGK_WDDMVERSION AdapterCaps->WDDMVersion = DXGKDDI_WDDMv1; } break; -- -- This posting is provided "AS IS" with no warranties, and confers no rights. Use of any included script samples are subject to the terms specified at http://www.microsoft.com/info/cpyright.htm "nan_hsing Chang" <nhchang03(a)gmail.com> wrote in message news:80c8282d-ded9-4193-8df8-f3dbab98cc61(a)m3g2000pri.googlegroups.com... > Thank for your replay. > > But I meet a strange situation. Original, I build my driver in WDK > 6001.18001 and the function flow is as you says that > DxgkDdiQueryAdapterInfo is called 3 times. But when I build in WDK > 7600.16385.0, DxgkDdiQueryAdapterInfo is called only one time for > DXGKQAITYPE_DRIVERCAPS, then the DxgkStopDevice is called, then > DxgkRemoveDevice. > > Do I miss some settings for WDK 7600 environment? Because I can't find > video driver sample(ex, r2000) in WDK 7600, I can't try sample code. > > BTW, I write display driver for a real device. > > > Ivan Brugiolo [MSFT] wrote: >> Is this a real device backed by a real piece of hardware, or, some >> virtual-bus enumerated device ? >> DxgKrnl.sys is extremely picky about hardware resource requirements >> for your device, and, it requires a real hardware interrput >> (CmResourceTypeInterrupt) and a piece of memory (CmResourceTypeMemory), >> and it parses the CM_RESOURCE_LIST to register the interrput for you. >> >> That said, if you made-it past DxgkDdiStartDevice, >> it looks like that part is OK. >> >> DxgkDdiQueryAdapterInfo is supposed to called 3 times during >> initialization: >> One for DXGKQAITYPE_DRIVERCAPS >> Twice for DXGKQAITYPE_QUERYSEGMENT. >> QuerySegment is again really picky about flags, since it tries to make >> sense >> of Memory-Space Segments, Aperture Segments >> and (for legacy AGP) AGP-Aperture segments. >> >> The general expectations for your exercise are: >> >> -1- You get past DxgkDdiQueryAdapterInfo, and, >> the Video-Memory-Manager gets intialized for your DXGADAPTER >> -2- Required registry information is retrieved and parsed correctly >> -3- Monitor Modes are requested for non hot-pluggable DxgkChild >> -4- DxgkDdiEnumVidPnCofuncModality is called hundreds to thousands of >> times, >> until the topology for that adapter is `constrained` and `made >> functional`. >> >> If the Gdi view of the VidPn Source is part of the desktop, then: >> >> -5- The Cdd kernel-mode DXGDEVICE is created >> -6- Adapter affine allocations (or Stardard Allocations) Shared and >> Shadow >> are created >> -7- Allocations are pinned, and your DMA completion routine will confirm >> Paging-Buffers >> -8- You will get some Gdi `Present` (or Blt) from the Shadow to the >> Shared >> >> If the desktop compositor is in a configuration where it can compose >> all of the Gdi views, or if there is a DX application, then you may see >> your User-mode driver to be loaded, and possibly invoked. >> >> -- >> This posting is provided "AS IS" with no warranties, and confers no >> rights. >> Use of any included script samples are subject to the terms specified at >> http://www.microsoft.com/info/cpyright.htm >> >> >> "nan_hsing Chang" <nhchang03(a)gmail.com> wrote in message >> news:7dee95a0-9b81-46c0-99bb-f2c8be96b46f(a)z4g2000prh.googlegroups.com... >> >I write the display WDDM driver for windwos7. Windwos stopped my >> > driver during initialization. The DxgkDdiStartDevice callback has be >> > called, then DxgkDdiQueryChildRelations and dxgkQueryAdapterInfo. In >> > dxgkQueryAdapterInfo function, I fill DXGK_DRIVERCAPS structure and >> > return STATUS_SUCCESS. But it can't call DxgkddiCreateDevice, it call >> > DxgkStopDevice instead. At last, my driver has been removed. >> > >> > Could someone give me suggestion? Should I report some caps in the >> > DXGK_DRIVERCAPS structure for windows7. Or My user mode driver can't >> > be load. I find my user mode driver can't be call and I don't know why.
|
Pages: 1 Prev: Confusing INFGate.EXE Errors Next: Vista WHQL PnP failure question |