Re: [PATCH edk2-platforms v2 0/7] SynQuacer ACPI support

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

On 1 March 2018 at 11:23, Graeme Gregory (Linaro)
<graeme.gregory@linaro.org> wrote:
> On Wed, Feb 28, 2018 at 07:24:14PM +0000, Ard Biesheuvel wrote:
>> This implements ACPI support for the SynQuacer platforms.
>>
>> Changes since v1:
>> - improve commit log (#1, #2)
>> - replace bare numbers with symbolic constants (#2)
>> - add Leif's R-b (#4)
>> - add patches #6 and #7
>>
>> Note that supporting ACPI on this SoC is non-trivial, due to the quirky
>> DesignWare RCs and the pre-ITS that sits between the PCIe RCs and the GICv3.
>> However, the most important issue has been addressed by modifying the static
>> SMMU mapping that sits between the CPUs and the PCIe config space, working
>> around the ghosting issue that occurs on these RCs, due the complete lack of
>> type 0 config TLP filtering by the [non-existent] root port. (This was tested
>> using the 20180226-LB1.1-ACPI-ramfw.bin SCP firmware image, which is not [yet]
>> installed by default on DeveloperBox hardware)
>>
>> That leaves the MSI issue, which is worked around by limiting MSI support to a
>> single RC. In the presented configuration, this is RC #1, which connects to the
>> x16 slot [and nothing else] on the DeveloperBox PCB. The onboard PCIe devices
>> (XHCI + SATA) work without problem using wired interrupts only, and so RC #0
>> has MSI support disabled. This means cards that require MSI support should
>> be inserted into the x16 slot, which is likely to be the preferred slot in
>> such cases anwyay (e.g., when using NVME or high end networking plugin cards)
>>
>> Patch #1 fixes a minor issue in the slot-to-BDF mapping.
>>
>> Patch #2 modifies the static PCIe window configuration so it can be described
>> using ACPI as well as DT.
>>
>> Patch #3 introduces the static ACPI tables that describe the fixed platform
>> devices and peripherals to the OS.
>>
>> Patch #4 adds a menu option to the platform driver to make ACPI vs DT user
>> selectable.
>>
>> Patch #5 adds support for describing the eMMC controller using a SSDT table
>> which is only installed if eMMC support is enabled.
>>
>> Patch #6 adds a _STA method implementation to the PCIe RC devices so that
>> they are only exposed to the OS when running on a platform that has one of
>> the several ECAM workarounds enabled. Otherwise, we can still boot via ACPI
>> using platform devices, but the PCIe RCs are unavailable.
>>
>> Patch #7 extends the _STA method for PCI0 to take the presence detect GPIO
>> into account. This is necessary because on the SynQuacer evaluation board,
>> any attempt to access the device registers will lock up the system if no
>> card is inserted into the slot.
>>
>> Note that driver support for the eMMC and network controller only landed in
>> v4.15, but when using a SATA driver and a plugin network card that does have
>> driver support, these patches should allow the SynQuacer based platforms to
>> boot stock Debian Stretch/Fedora/Centos etc installers.
>>
>
> The ACPI parts look good to me.
>
> Reviewed-by: Graeme Gregory <graeme.gregory@linaro.org>
>

Cheers.

I'll hold off on merging this until Heyi's PciHostBridgeDxe patches
are in (which are in pretty good shape now)