Hi Neil,

On Mon Jan 12, 2026 at 9:26 AM CET, Neil Armstrong wrote:
> On 1/7/26 14:53, Neil Armstrong wrote:
>> Hi,
>> 
>> On 1/7/26 09:05, Luca Weiss wrote:
>>> Add the nodes for the UFS PHY and UFS host controller, along with the
>>> ICE used for UFS.
>>>
>>> Signed-off-by: Luca Weiss <[email protected]>
>>> ---
>>>   arch/arm64/boot/dts/qcom/milos.dtsi | 127 
>>> +++++++++++++++++++++++++++++++++++-
>>>   1 file changed, 124 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/arch/arm64/boot/dts/qcom/milos.dtsi 
>>> b/arch/arm64/boot/dts/qcom/milos.dtsi
>>> index e1a51d43943f..0f69deabb60c 100644
>>> --- a/arch/arm64/boot/dts/qcom/milos.dtsi
>>> +++ b/arch/arm64/boot/dts/qcom/milos.dtsi
>>> @@ -797,9 +797,9 @@ gcc: clock-controller@100000 {
>>>                    <&sleep_clk>,
>>>                    <0>, /* pcie_0_pipe_clk */
>>>                    <0>, /* pcie_1_pipe_clk */
>>> -                 <0>, /* ufs_phy_rx_symbol_0_clk */
>>> -                 <0>, /* ufs_phy_rx_symbol_1_clk */
>>> -                 <0>, /* ufs_phy_tx_symbol_0_clk */
>>> +                 <&ufs_mem_phy 0>,
>>> +                 <&ufs_mem_phy 1>,
>>> +                 <&ufs_mem_phy 2>,
>>>                    <0>; /* usb3_phy_wrapper_gcc_usb30_pipe_clk */
>>>               #clock-cells = <1>;
>>> @@ -1151,6 +1151,127 @@ aggre2_noc: interconnect@1700000 {
>>>               qcom,bcm-voters = <&apps_bcm_voter>;
>>>           };
>>> +        ufs_mem_phy: phy@1d80000 {
>>> +            compatible = "qcom,milos-qmp-ufs-phy";
>>> +            reg = <0x0 0x01d80000 0x0 0x2000>;
>>> +
>>> +            clocks = <&rpmhcc RPMH_CXO_CLK>,
>>> +                 <&gcc GCC_UFS_PHY_PHY_AUX_CLK>,
>>> +                 <&tcsr TCSR_UFS_CLKREF_EN>;
>>> +            clock-names = "ref",
>>> +                      "ref_aux",
>>> +                      "qref";
>>> +
>>> +            resets = <&ufs_mem_hc 0>;
>>> +            reset-names = "ufsphy";
>>> +
>>> +            power-domains = <&gcc UFS_MEM_PHY_GDSC>;
>>> +
>>> +            #clock-cells = <1>;
>>> +            #phy-cells = <0>;
>>> +
>>> +            status = "disabled";
>>> +        };
>>> +
>>> +        ufs_mem_hc: ufshc@1d84000 {
>>> +            compatible = "qcom,milos-ufshc", "qcom,ufshc", "jedec,ufs-2.0";
>>> +            reg = <0x0 0x01d84000 0x0 0x3000>;
>>> +
>>> +            interrupts = <GIC_SPI 265 IRQ_TYPE_LEVEL_HIGH 0>;
>>> +
>>> +            clocks = <&gcc GCC_UFS_PHY_AXI_CLK>,
>>> +                 <&gcc GCC_AGGRE_UFS_PHY_AXI_CLK>,
>>> +                 <&gcc GCC_UFS_PHY_AHB_CLK>,
>>> +                 <&gcc GCC_UFS_PHY_UNIPRO_CORE_CLK>,
>>> +                 <&tcsr TCSR_UFS_PAD_CLKREF_EN>,
>>> +                 <&gcc GCC_UFS_PHY_TX_SYMBOL_0_CLK>,
>>> +                 <&gcc GCC_UFS_PHY_RX_SYMBOL_0_CLK>,
>>> +                 <&gcc GCC_UFS_PHY_RX_SYMBOL_1_CLK>;
>>> +            clock-names = "core_clk",
>>> +                      "bus_aggr_clk",
>>> +                      "iface_clk",
>>> +                      "core_clk_unipro",
>>> +                      "ref_clk",
>>> +                      "tx_lane0_sync_clk",
>>> +                      "rx_lane0_sync_clk",
>>> +                      "rx_lane1_sync_clk";
>>> +
>>> +            resets = <&gcc GCC_UFS_PHY_BCR>;
>>> +            reset-names = "rst";
>>> +
>>> +            interconnects = <&aggre1_noc MASTER_UFS_MEM QCOM_ICC_TAG_ALWAYS
>>> +                     &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>,
>>> +                    <&gem_noc MASTER_APPSS_PROC QCOM_ICC_TAG_ACTIVE_ONLY
>>> +                     &cnoc_cfg SLAVE_UFS_MEM_CFG QCOM_ICC_TAG_ACTIVE_ONLY>;
>>> +            interconnect-names = "ufs-ddr",
>>> +                         "cpu-ufs";
>>> +
>>> +            power-domains = <&gcc UFS_PHY_GDSC>;
>>> +            required-opps = <&rpmhpd_opp_nom>;
>>> +
>>> +            operating-points-v2 = <&ufs_opp_table>;
>>> +
>>> +            iommus = <&apps_smmu 0x60 0>;
>> 
>> dma-coherent ?


Given that downstream volcano.dtsi has dma-coherent in the ufshc@1d84000
node, looks like this is missing in my patch.

>> 
>> and no MCQ support ?

Not sure, I could only find one reference to MCQ on createpoint for
milos, but given there's no mcq_sqd/mcq_vs reg defined downstream, and I
couldn't find anything for the same register values in the .FLAT file, I
don't think Milos has MCQ? Feel free to prove me wrong though.

>
> So, people just ignore my comment ?
>
> Milos is based on SM8550, so it should have dma-coherent, for the MCQ
> I hope they used the fixed added to the SM8650 UFS controller for MCQ.

Not sure what this should mean regarding MCQ...

Regards
Luca

>
> Neil
>
>> 
>> <snip>
>> 
>> Thanks,
>> Neil


Reply via email to