Re: Followup : EfiTestManagedDevice() issue in driver GetControllerName() function

["David A. Van Arnem" <dvanarnem@cmlab.biz>]

Hi Jeff,

On 02/21/2017 03:10 PM, Jeff Westfahl wrote:
> Hello David,
> 
> It sounds like you are trying to get a list of handles to your specific
> devices, correct?

Yes, exactly.

> 
> One way to do this is to create a new protocol for your devices and
> create an instance of this protocol when your driver attaches to a PCI
> device. Then you can just retrieve all handles that match your new
> protocol.
> 
> Creating a new protocol for this use case could be as simple as just
> allocating a unique GUID, using uuidgen or something similar. You don't
> necessarily need to have data or functions associated with your protocol
> for this purpose, although that may prove useful.

This sounds like a simple method to accomplish what I need, thanks.  I
assume I would install the custom protocol in the same place I install
the Driver Binding and Component Name(2) implementations?

Regards,
David

> 
> I wouldn't consider myself an expert of any kind on UEFI drivers, so if
> others see problems with my suggestions or have better ideas, please
> chime in.
> 
> Regards,
> Jeff
> 
> On Tue, 21 Feb 2017, David A. Van Arnem wrote:
> 
>> Hello again,
>>
>> I think the problem lies in my application, as upon further
>> investigation, all the handles that publish the Component Name 2
>> protocol appear to be failing the EfiTestManagedDevice() check (they all
>> return EFI_UNSUPPORTED).  My eventual goal is, given a buffer of handles
>> that publish EFI_PCI_IO_PROTOCOL, I would like to locate my specific
>> device.  I was going to do a string comparison of the ControllerName,
>> but this doesn't seem to be the right approach.  Is there a recommended
>> way to accomplish this?
>>
>> Regards,
>> David
>>
>> On 02/21/2017 01:07 PM, David A. Van Arnem wrote:
>>> Hi all,
>>>
>>> I am developing a UEFI_DRIVER for a custom PCIe device.  I've followed
>>> the UEFI Driver Model, and implemented the Driver Binding, Component
>>> Name, and Component Name 2 protocols for my driver.  I'm trying to
>>> retrieve the Controller Name from this driver in a UEFI_APPLICATION, but
>>> I've run into an issue.
>>>
>>> In the driver's Start() function, I open the EFI_PCI_IO_PROTOCOL on
>>> ControllerHandle using the following:
>>>
>>> Status = gBS->OpenProtocol (
>>>             ControllerHandle,
>>>             &gEfiPciIoProtocolGuid,
>>>             (VOID **)&PciIoProtocol,
>>>             This->DriverBindingHandle,
>>>             ControllerHandle,
>>>             EFI_OPEN_PROTOCOL_BY_DRIVER
>>>     );
>>>
>>> And leave it open when Start() exits, unless some other part of the
>>> Start() function fails, which doesn't appear to be happening based on my
>>> debug messages.
>>>
>>> Following some of the examples in edk2, in my Component Name's
>>> GetControllerName() function, I used EfiTestManagedDevice() to check
>>> that the ControllerHandle opened EFI_PCI_IO_PROTOCOL:
>>>
>>> Status = EfiTestManagedDevice (
>>>             ControllerHandle,
>>>             g<devicename>DriverBinding.DriverBindingHandle,
>>>             &gEfiPciIoProtocolGuid
>>>     );
>>>
>>> Finally, to get the Controller Name in the application, I first get a
>>> buffer of handles that publish the ComponentName2 protocol, then walk
>>> through each and call GetControllerName().
>>>
>>> However, the EfiTestManagedDevice() check in my driver is returning
>>> EFI_UNSUPPORTED, indicating the protocol is not supported on the handle.
>>>  Furthermore, after using "load <drivername>.efi" from the shell,
>>> openinfo on the handle returns:
>>>
>>> Handle 188 (6E16A798)
>>> ComponentName2
>>> ComponentName
>>> DriverBinding
>>>   Drv[01] Ctrl[ ] Cnt(200) HandProt  Image(<null string>)
>>> ImageDevicePath
>>>   Drv[187] Ctrl[ ] Cnt(01) GetProt  Image(<null string>)
>>> LoadedImage
>>>   Drv[01] Ctrl[ ] Cnt(09) HandProt  Image(<null string>)
>>>
>>> I feel like maybe I'm missing a step here.  Should I see PciIo in
>>> openinfo?  Do I need to open a "child" instance of the ControllerHandle
>>> in my application to get EfiTestManagedDevice() to pass?  I'm not sure
>>> if that's the right terminology, I'm having a little difficulty
>>> understanding the "child" concept in this context.  Any tips
>>> appreciated.
>>>
>>
>> -- 
>> Regards,
>> David Van Arnem
>> Development Engineer IV
>> Computer Measurement Laboratory, LLC
>>
>> _______________________________________________
>> edk2-devel mailing list
>> edk2-devel@lists.01.org
>> https://lists.01.org/mailman/listinfo/edk2-devel
>>

-- 
Regards,
David Van Arnem
Development Engineer IV
Computer Measurement Laboratory, LLC