Re: [PATCH edk2-platforms 1/2] Silicon/AMD/Styx: Make PcdSataPortMode 32 bits

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

On 20 August 2017 at 19:08, Alan Ott <alan@softiron.com> wrote:
> On 08/20/2017 01:46 PM, Ard Biesheuvel wrote:
>>
>> On 19 August 2017 at 22:41, Alan Ott <alan@softiron.com> wrote:
>>>
>>> Extra bits are needed to accomodate all 14 SATA ports
>>>
>>> Signed-off-by: Alan Ott <alan@softiron.com>
>>> Contributed-under: TianoCore Contribution Agreement 1.0
>>> ---
>>>  Silicon/AMD/Styx/AmdStyx.dec                                  | 2 +-
>>>  Silicon/AMD/Styx/Drivers/StyxSataPlatformDxe/InitController.c | 4 ++--
>>>  2 files changed, 3 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/Silicon/AMD/Styx/AmdStyx.dec b/Silicon/AMD/Styx/AmdStyx.dec
>>> index ddd5bf4..c6eebe6 100644
>>> --- a/Silicon/AMD/Styx/AmdStyx.dec
>>> +++ b/Silicon/AMD/Styx/AmdStyx.dec
>>> @@ -54,7 +54,7 @@
>>>
>>> gAmdStyxTokenSpaceGuid.PcdSata0CtrlAxiSlvPort|0xE0300000|UINT32|0x00020000
>>>    gAmdStyxTokenSpaceGuid.PcdSata0PortCount|8|UINT8|0x00020001
>>>    gAmdStyxTokenSpaceGuid.PcdSataPi|0xFF|UINT32|0x00020002
>>> -  gAmdStyxTokenSpaceGuid.PcdSataPortMode|0|UINT16|0x00020003
>>> +  gAmdStyxTokenSpaceGuid.PcdSataPortMode|0|UINT32|0x00020003
>>>    gAmdStyxTokenSpaceGuid.PcdSataPortMpsp|TRUE|BOOLEAN|0x00020004
>>>    gAmdStyxTokenSpaceGuid.PcdSataSmpsSupport|FALSE|BOOLEAN|0x00020005
>>>    gAmdStyxTokenSpaceGuid.PcdSataSssSupport|TRUE|BOOLEAN|0x00020006
>>> diff --git
>>> a/Silicon/AMD/Styx/Drivers/StyxSataPlatformDxe/InitController.c
>>> b/Silicon/AMD/Styx/Drivers/StyxSataPlatformDxe/InitController.c
>>> index 1958d91..78c6819 100644
>>> --- a/Silicon/AMD/Styx/Drivers/StyxSataPlatformDxe/InitController.c
>>> +++ b/Silicon/AMD/Styx/Drivers/StyxSataPlatformDxe/InitController.c
>>> @@ -110,8 +110,8 @@ InitializeSataController (
>>>    SataChPerSerdes = FixedPcdGet8 (PcdSataNumChPerSerdes);
>>>
>>>    for (PortNum = 0; PortNum < SataPortCount; PortNum += SataChPerSerdes)
>>> {
>>> -    EvenPort = (UINT32)(FixedPcdGet16 (PcdSataPortMode) >> (PortNum *
>>> 2)) & 3;
>>> -    OddPort = (UINT32)(FixedPcdGet16 (PcdSataPortMode) >> ((PortNum+1) *
>>> 2)) & 3;
>>> +    EvenPort = (UINT32)(FixedPcdGet32 (PcdSataPortMode) >> (PortNum *
>>> 2)) & 3;
>>> +    OddPort = (UINT32)(FixedPcdGet32 (PcdSataPortMode) >> ((PortNum+1) *
>>> 2)) & 3;
>>>      SataPhyInit ((StartPort + PortNum) / SataChPerSerdes, EvenPort,
>>> OddPort);
>>>    }
>>>
>>
>> This doesn't look right to me. PortNum will always be in the set { 0,
>> 2, 4, 6 } due to SataPortCount being equal to either the port count of
>> sata 0 or of sata 1. So the top 16 bits of the new PCD are never
>> referenced, and the port mode for sata 0 ends up being applied to sata
>> 1.
>>
>> Better use
>>
>> for (PortNum = StartPort; PortNum < StartPort + SataPortCount; ...)
>>   ..etc..
>>
>
> I agree. This has apparently been broken all along. I'll fix and resubmit.
>
>>
>> Also, this code relies rather heavily on SataChPerSerdes == 2, so we
>> might want to add an ASSERT () for that.
>>
>
> I can do that too.
>
> Alan.
>