[REGRESSION] Potential problems for DT-described PCIe nodes on i.MX

[Sherry Sun]

> Subject: [REGRESSION] Potential problems for DT-described PCIe nodes on
> i.MX
>
> Hi,
>
> Linux commits:
>
> 9dc39532172c ("ARM: dts: imx6qdl: Add Root Port node and PERST
> property") 37f53f50b36f ("arm64: dts: imx8mm: Add Root Port node and
> PERST property") 396fbe9a028d ("arm64: dts: imx8mp: Add Root Port node
> and PERST property")
>
> Both in v7.2-rc1 introduce potential regressions for i.MX boards that already
> described a root port under a different node name, e.g.:
>
> host at 0:   ZII RDU2
> root at 0,0: GE Boards
> pcie at 0,0: Gateworks boards
> root at 0,0: SolidRun Hummingboard
>
> All these nodes may become stale now as the SoC DTSI have introduced their
> own pcie at 0.
>
> At least on the ZII board, barebox fixes up a MAC address into the DT node,
> which is no longer used if the kernel decides to use the newly added
> duplicate pcie at 0 node instead.
>
> Fix is probably merging the existing properties and child nodes to be under
> the new &pcie_port0/&pcie0_port0 and drop the board-level Root Port node.
>

Hi Ahmad, thanks for the report.
Agree for the fix plan -- merge the existing properties and child nodes to the new pcie_port0.
I will send a patch set to fix above impacted dts/dtsi files.

Best Regards
Sherry


NXP Confidential