On 3/14/25 5:57 PM, Bjorn Andersson wrote: > On Thu, Mar 06, 2025 at 04:46:09PM +0530, Manikanta Mylavarapu wrote: >> Add PCIe0, PCIe1, PCIe2, PCIe3 (and corresponding PHY) devices >> found on IPQ5424 platform. The PCIe0 & PCIe1 are 1-lane Gen3 >> host whereas PCIe2 & PCIe3 are 2-lane Gen3 host. >> >> Reviewed-by: Konrad Dybcio <konrad.dybcio@xxxxxxxxxxxxxxxx> >> Signed-off-by: Manikanta Mylavarapu <quic_mmanikan@xxxxxxxxxxx> > > When validating this against linux-next DT bindings I get: > > arch/arm64/boot/dts/qcom/ipq5424-rdp466.dtb: pcie@f8000: reg: [[0, 1015808, 0, 12288], [0, 1073741824, 0, 3868], [0, 1073745696, 0, 168], [0, 1073745920, 0, 4096], [0, 1074790400, 0, 4096], [0, 1040384, 0, 4096]] is too long > arch/arm64/boot/dts/qcom/ipq5424-rdp466.dtb: pcie@f8000: reg-names:0: 'dbi' was expected > arch/arm64/boot/dts/qcom/ipq5424-rdp466.dtb: pcie@f8000: reg-names:1: 'elbi' was expected > arch/arm64/boot/dts/qcom/ipq5424-rdp466.dtb: pcie@f8000: reg-names:2: 'atu' was expected > arch/arm64/boot/dts/qcom/ipq5424-rdp466.dtb: pcie@f8000: reg-names:3: 'parf' was expected > arch/arm64/boot/dts/qcom/ipq5424-rdp466.dtb: pcie@f8000: reg-names: ['parf', 'dbi', 'elbi', 'atu', 'config', 'mhi'] is too long > > Are we still missing something? Yes. There's a dt-bindings mess regarding a number of similar ipq platforms that has been impacted by a recent round of reviews that resulted in reverts. The merged bindings for this platform were created based on ipq9574, which used some crazy order for reg entries (i.e. PARF not being first). That went unnoticed for a couple of them and when one of the ipq platforms patches were under review, the sender was asked to align with the other qc platforms. Then, a series changing existing bindings with users was submitted, which was partially merged (and then subsequently reverted), so some patches are based on that, while others are based on the state of -next. Now this v4 puts parf as the first region, but the v1 bindings were merged a couple months ago, before the discrepancy was discussed. Konrad
