[U-Boot] [PATCH v2 14/18] efi_loader: provide links between devices EFI handles

[Heinrich Schuchardt]

On 01/18/2018 07:13 PM, Alexander Graf wrote:
> 
> 
> On 18.01.18 17:18, Heinrich Schuchardt wrote:
>> On 01/18/2018 05:09 PM, Alexander Graf wrote:
>>>
>>>
>>> On 17.01.18 20:16, Heinrich Schuchardt wrote:
>>>> U-Boot devices and EFI handles can be related, e.g. an
>>>> IDE disk relates to a handle with the EFI_BLOCK_IO_PROTOCOL.
>>>> Provide pointers to store these links.
>>>>
>>>> Signed-off-by: Heinrich Schuchardt <xypron.glpk at gmx.de>
>>>> ---
>>>> v2
>>>> 	no change
>>>> ---
>>>>  include/dm/device.h           | 4 ++++
>>>>  include/efi_loader.h          | 2 ++
>>>>  lib/efi_loader/efi_boottime.c | 1 +
>>>>  3 files changed, 7 insertions(+)
>>>>
>>>> diff --git a/include/dm/device.h b/include/dm/device.h
>>>> index 813e49f330..e5c54fe7b6 100644
>>>> --- a/include/dm/device.h
>>>> +++ b/include/dm/device.h
>>>> @@ -11,6 +11,7 @@
>>>>  #ifndef _DM_DEVICE_H
>>>>  #define _DM_DEVICE_H
>>>>  
>>>> +#include <efi_loader.h>
>>>>  #include <dm/ofnode.h>
>>>>  #include <dm/uclass-id.h>
>>>>  #include <fdtdec.h>
>>>> @@ -144,6 +145,9 @@ struct udevice {
>>>>  	uint32_t flags;
>>>>  	int req_seq;
>>>>  	int seq;
>>>> +#ifdef EFI_LOADER
>>>> +	efi_handle_t handle;
>>>> +#endif
>>>
>>> I fail to find where you actually make use of the handle inside
> 
> Care to answer here too? :)

The changes in include/dm/device.h are not needed. I will revert these.

Maybe in future we will have to back link devices to handles but not now.

Regards

Heinrich

> 
>>>
>>>>  #ifdef CONFIG_DEVRES
>>>>  	struct list_head devres_head;
>>>>  #endif
>>>> diff --git a/include/efi_loader.h b/include/efi_loader.h
>>>> index 4060348695..711c901eda 100644
>>>> --- a/include/efi_loader.h
>>>> +++ b/include/efi_loader.h
>>>> @@ -139,6 +139,8 @@ struct efi_object {
>>>>  	struct list_head protocols;
>>>>  	/* The object spawner can either use this for data or as identifier */
>>>>  	void *handle;
>>>> +	/* Device */
>>>> +	struct udevice *dev;
>>>>  };
>>>>  
>>>>  /**
>>>> diff --git a/lib/efi_loader/efi_boottime.c b/lib/efi_loader/efi_boottime.c
>>>> index 5a3349ecb2..4b3b63e39a 100644
>>>> --- a/lib/efi_loader/efi_boottime.c
>>>> +++ b/lib/efi_loader/efi_boottime.c
>>>> @@ -362,6 +362,7 @@ efi_status_t efi_create_handle(efi_handle_t *handle)
>>>>  			      (void **)&obj);
>>>>  	if (r != EFI_SUCCESS)
>>>>  		return r;
>>>> +	obj->dev = NULL;
>>>
>>> How about we just zero initialize the whole struct?
>>
>> All other fields are initialized in efi_add_handle().
>> So why invest more CPU cycles?
> 
> I'm mostly concerned that we get into a situation where people don't
> fully grasp the flow of what gets initialized where and nasty bugs happen.
> 
> So I guess we should either initialize obj->dev in efi_add_handle() or
> fully zero initialize obj, like we do for most other callers of
> efi_add_handle().
> 
> 
> Alex
>