Hi Florian,
On 6/26/24 14:31, Florian Fainelli wrote:
> Hi,
>
> On 26/06/2024 11:45, Stanimir Varbanov wrote:
>> This patchset aims to add bare minimum support for bcm2712
>> in brcmstb PCIe driver needed to support the peripherals from
>> RP1 south-bridge found in RPi5. In order to support RP1
>> PCIe endpoint peripherals a new interrupt controller is added.
>> The interrupt controller supports 64 interrupt sources which
>> are enough to handle 61 RP1 peripherals.
>>
>> Patch 1 is adding DT binding schema for the MIP interrupt
>> controller, patch 2 is adding relevant changes for PCIe
>> bcm2712 in yaml. Patch 3 adds MIP intterrupt cotroller driver.
>> Patches 4 and 5 are preparations for adding bcm2712 support in 6.
>> The last patch updates bcm2712 .dsti by adding pcie DT nodes.
>>
>> Few concerns about the implementation:
>> - the connection between MIP interrupt-controller and PCIe RC is
>> done through BAR1. The PCIe driver is parsing the msi_parent
>> DT property in order to obtain few private DT properties like
>> "brcm,msi-pci-addr" and "reg". IMO this looks hackish but I failed
>> to find something better. Ideas?
>>
>> - in downstream RPi kernel "ranges" and "dma-ranges" DT properties
>> are under an axi {} simple-bus node even that PCIe block is on CPU
>> MMIO bus. I tried to merge axi {} in soc {} and the result could be
>> seen on the last patch in this series, but I'm still not sure that
>> it looks good enough.
>>
>> This series has been functionally tested on OpenSUSE Tumbleweed with
>> downstream RP1 south-bridge PCIe endpoint driver implementation as
>> MFD by using ethernet which is part of it.
>>
>> The series is based on Andrea's "Add minimal boot support for
>> Raspberry Pi 5"
>> series.
>>
>> Comments are welcome!
>
> We are just about submitting support for 7712 which is the sister chip
> of 2712 and requires similar, if not identical types of changes to
> pcie-brcmstb.c, would you mind reviewing that patch series when it gets
> posted by Jim in the next few days, and base yours upon that one? It
Sure, I'll review bcm7712 series and rebase bcm2712 on top of it.
> does separate changes in a more atomic and a more reviewer friendly
> rather than having one big commit modifying pcie-brcmstb.c
Agreed.
>
> Thanks.
~Stan
