Re: [PATCH v2 3/7] OvmfPkg: introduce PciCapPciIoLib

[Ard Biesheuvel <ard.biesheuvel@linaro.org>]

On 24 May 2018 at 16:50, Laszlo Ersek <lersek@redhat.com> wrote:
> On 05/24/18 10:13, Ard Biesheuvel wrote:
>> On 23 May 2018 at 22:21, Laszlo Ersek <lersek@redhat.com> wrote:
>>> Add a library class, and a UEFI_DRIVER lib instance, that are layered on
>>> top of PciCapLib, and allow clients to plug an EFI_PCI_IO_PROTOCOL backend
>>> into PciCapLib, for config space access.
>>>
>>> (Side note:
>>>
>>
>> Again, please retain the below.
>
> Will do.
>
>>> +STATIC
>>> +EFI_STATUS
>>> +ProtoDevTransferConfig (
>>> +  IN     EFI_PCI_IO_PROTOCOL        *PciIo,
>>> +  IN     EFI_PCI_IO_PROTOCOL_CONFIG TransferFunction,
>>> +  IN     UINT16                     ConfigOffset,
>>> +  IN OUT UINT8                      *Buffer,
>>> +  IN     UINT16                     Size
>>> +  )
>>> +{
>>> +  while (Size > 0) {
>>> +    EFI_PCI_IO_PROTOCOL_WIDTH Width;
>>> +    UINT16                    Count;
>>> +    EFI_STATUS                Status;
>>> +    UINT16                    Progress;
>>> +
>>> +    //
>>> +    // Pick the largest access size that is allowed by the remaining transfer
>>> +    // Size and by the alignment of ConfigOffset.
>>> +    //
>>> +    // When the largest access size is available, transfer as many bytes as
>>> +    // possible in one iteration of the loop. Otherwise, transfer only one
>>> +    // unit, to improve the alignment.
>>> +    //
>>> +    if (Size >= BIT2 && (ConfigOffset & (BIT2 - 1)) == 0) {
>>
>> Ugh. Just use '4' or sizeof(UINT32).
>>
>>> +      Width = EfiPciIoWidthUint32;
>>> +      Count = Size >> Width;
>>> +    } else if (Size >= BIT1 && (ConfigOffset & (BIT1 - 1)) == 0) {
>>> +      Width = EfiPciIoWidthUint16;
>>> +      Count = 1;
>>> +    } else {
>>> +      Width = EfiPciIoWidthUint8;
>>> +      Count = 1;
>>> +    }
>
> I used "BITx" and "(BITx - 1)" for consistency, and because they seemed
> to convey the idea well (namely, shifting down the alignment).
>
> I'm fine replacing "BIT2" with "4", but then I believe I should also
> replace "(BIT2 - 1)" with "3". Similarly, replace "BIT1" with "2", and
> "(BIT1 -1)" with 1.
>
> Do you prefer the current code or the decimal constants?
>

IMHO, BITx is for bitmasks, not for numerical constants.

So yes, if you think (as do I) that sizeof(UINTnn) is too wordy, just
use the plain numbers please.