Re: [edk2-devel] [edk2-platforms][PATCH 1/1] Platform/RaspberryPi: Revert don't connect all devices on an ordinary boot

["Pete Batard" <pete@akeo.ie>]

NB: This patch is now superseded by 
https://edk2.groups.io/g/devel/message/60941

On 2020.06.08 18:24, Pete Batard via groups.io wrote:
> On 2020.06.08 18:10, Ard Biesheuvel wrote:
>> On 6/8/20 7:01 PM, Pete Batard wrote:
>>> On 2020.06.08 17:43, Ard Biesheuvel wrote:
>>>> On 6/8/20 5:51 PM, Pete Batard wrote:
>>>>> This reverts commit c8000ecccc83b728baf04ced2fedb870bc3bc1b3 on 
>>>>> account
>>>>> of USB keyboard devices not being detected until after the timeout to
>>>>> enter the Setup menu or run the UEFI Shell has expired.
>>>>>
>>>>> The core of the issue is that, when EfiBootManagerConnectAll() is not
>>>>> invoked, PCI bus scanning only starts after the BDS wait period has
>>>>> elapsed, which effectively takes away the ability to use an USB device
>>>>> to enact Boot Device Selection or interact with the Setup menu.
>>>>>
>>>>> Signed-off-by: Pete Batard <pete@akeo.ie>
>>>>
>>>> Could we move the call to
>>>>
>>>> EfiEventGroupSignal (&gEfiEndOfDxeEventGroupGuid);
>>>>
>>>> to the start of PlatformBootManagerBeforeConsole() instead?
>>>
>>> Moving EfiEventGroupSignal(&gEfiEndOfDxeEventGroupGuid) early into 
>>> PlatformBootManagerBeforeConsole() doesn't seem to have any effect, 
>>> as far as I could test.
>>>
>>> We can however move the EfiBootManagerConnectAll() there if that's 
>>> what you are asking.
>>>
>>
>> No, the whole idea was to get rid of EfiBootManagerConnectAll().
>>
>> You are saying the PCI bus scanning starts after the BDS wait period, 
>> right? This is strange, I wouldn't expect that. That is why I 
>> suggested moving the EfiEventGroupSignal() call earlier.
>>
>> Looking at the ArmPkg version, it appears we need a sequence like
>>
>> EfiEventGroupSignal (&gEfiEndOfDxeEventGroupGuid);
>>
>> EfiBootManagerDispatchDeferredImages ();
>>
>> FilterAndProcess (&gEfiPciRootBridgeIoProtocolGuid, NULL, Connect);
>>
>> and that last call is missing from the RPi4 implementation.
>>
>> After those calls, all the PCI I/O handles should be enumerated.
> 
> Yup, that seems to work.
> 
> I need to test a bit more and clean up, then I'll send a patch to 
> supersede this one.
> 
> Regards,
> 
> /Pete
> 
>>
>>
>>>
>>>>
>>>>
>>>>
>>>>
>>>>> ---
>>>>>   Platform/RaspberryPi/Library/PlatformBootManagerLib/PlatformBm.c 
>>>>> | 5 +++++
>>>>>   1 file changed, 5 insertions(+)
>>>>>
>>>>> diff --git 
>>>>> a/Platform/RaspberryPi/Library/PlatformBootManagerLib/PlatformBm.c 
>>>>> b/Platform/RaspberryPi/Library/PlatformBootManagerLib/PlatformBm.c
>>>>> index 253614a646c1..2bd625ad7e7c 100644
>>>>> --- a/Platform/RaspberryPi/Library/PlatformBootManagerLib/PlatformBm.c
>>>>> +++ b/Platform/RaspberryPi/Library/PlatformBootManagerLib/PlatformBm.c
>>>>> @@ -655,6 +655,11 @@ PlatformBootManagerAfterConsole (
>>>>>       Print (BOOT_PROMPT);
>>>>>     }
>>>>> +  //
>>>>> +  // Connect the rest of the devices.
>>>>> +  //
>>>>> +  EfiBootManagerConnectAll ();
>>>>> +
>>>>>     Status = gBS->LocateProtocol (&gEsrtManagementProtocolGuid, 
>>>>> NULL, (VOID**)&EsrtManagement);
>>>>>     if (!EFI_ERROR (Status)) {
>>>>>       EsrtManagement->SyncEsrtFmp ();
>>>>>
>>>>
>>>
>>
> 
> 
> 
>