Re: [PATCH v2 02/13] dt-bindings: phy: Add binding for QCS615 standalone QMP DP PHY | linux-phy

quoted

On 8/1/2025 1:13 AM, Dmitry Baryshkov wrote:
On 31/07/2025 08:06, Xiangxu Yin wrote:
On 7/31/2025 2:35 AM, Dmitry Baryshkov wrote:
On Wed, Jul 30, 2025 at 04:53:16PM +0800, Xiangxu Yin wrote:
On 7/22/2025 8:41 PM, Dmitry Baryshkov wrote:
On Tue, Jul 22, 2025 at 08:05:06PM +0800, Xiangxu Yin wrote:
On 7/22/2025 4:38 PM, Dmitry Baryshkov wrote:
On Tue, Jul 22, 2025 at 03:22:03PM +0800, Xiangxu Yin wrote:
Introduce device tree binding documentation for the Qualcomm QMP DP PHY
on QCS615 SoCs. This PHY supports DisplayPort functionality and is
designed to operate independently from the USB3 PHY.

Unlike combo PHYs found on other platforms, the QCS615 DP PHY is
standalone and does not support USB/DP multiplexing. The binding
describes the required clocks, resets, TCSR configuration, and clock/PHY
cells for proper integration.
Simply put: no, this is not correct. Even if you go to the SM6150 block
diagram, it points out that DP uses the USB3 PHY, not a separate DP PHY.

I thought that we have discussed it beforehand.

I can quote my comment from the previous thread:

No. It means replacing extending existing entries with bigger reg and
#phy-cells = <1>. The driver must keep working with old node definitions
as is to ensure backwards compatibility. New nodes should make it
register two PHYs (USB3 and DP). On the driver side modify generic code
paths, all platforms supported by the driver should be able to support
USB3+DP combination.
Looking at the hardware memory maps:

MSM8998: USB3 PHY regs at 0xc010000, DP PHY regs at 0xc011000
SDM660: USB3 PHY regs at 0xc010000, DP PHY regs at 0xc011000
QCM2290: USB3 PHY regs at 0x1615000, DP PHY regs at 0x1616000
SM6115: USB3 PHY regs at 0x1615000, DP PHY regs at 0x1616000

Now:
SM6150: USB3 PHY regs at 0x88e6000
         USB3 PHY regs at 0x88e8000, DP PHY regs at 0x88e9000

I do not know, why msm-4.14 didn't describe second USB3 PHY. Maybe you
can comment on it.

But based on that list, the only special case that we need to handle is
the first USB3 PHY, which doesn't have a corresponding DP PHY block. But
it will be handled anyway by the code that implements support for the
existing DT entries. All other hardware blocks are combo USB+DP PHYs.

Having all of that in mind, please, for v3 patchset implement USB+DP
support in the phy-qcom-qmp-usbc driver and add the following logic
that also was requested in v1 review:

Not quite. Both USB3 and DP drivers should be calling power_on / _off.
If USB3 is on, powering on DP PHY should fail. Vice versa, if DP is on,
powering on USB should fail.
I think our understanding might not be fully aligned.
I did not write this. Please fix your mailer to quote messages properly.
As you are using Thunderbird, I'm not sure where the issue comes from.

Also please fix it to wrap your responses somwhere logically.

Perhaps this is because I didn’t accurately update the mutual exclusion relationships and test results for the different PHYs.
Let me clarify my latest findings and explain why I believe these are separate PHYs that require mutual exclusion via TCSR.

1. About the TCSR DP_PHYMODE Registers

MSM8998/SDM660:
    Only one TCSR_USB3_DP_PHYMODE register at 0x1FCB248.
QCM2290/SM6115:
    TCSR_USB3_0_DP_PHYMODE at 0x3CB248
    TCSR_USB3_1_DP_PHYMODE at 0x3CB24C
SM6150:
    TCSR_USB3_0_DP_PHYMODE at 0x1FCB248
    TCSR_USB3_1_DP_PHYMODE at 0x1FCB24C
SM6150 has two different sets of output pins, so the first register
covers first set of SS lanes (which are routed to the documented SS
PHY), the second register covers the second set of SS lanes (which are
routed to the DP and secondary USB PHY).

I can only assume that the same configuration was supposed to be
applicable to QCM2290 / SM6115, but was later removed / disabled, while
the registers were kept in the TCSR block.

Even though MSM8998, SDM660, QCM2290, and SM6115 all have one USB3 PHY and one DP PHY, the TCSR DP_PHYMODE register configuration is different on each platform.

Additionally, I found some interesting register documentation for QCM2290/SM6115:
    TCSR_USB3_0_DP_PHYMODE: “In kamorta this one is for mobile usb. DP not supported.”
    TCSR_USB3_1_DP_PHYMODE: “DP mode supported for Auto usb in kamorta.”
I think the reason for having two different TCSR registers is to allow both the USB3.0 and DP PHYs to be useds at the same time in certain product configurations.
Sure. One for the first PHY (USB), one for the second PHY (USB+DP).
If you check the memory map, you will find the second VLS CLAMP register
for the second USB PHY.

2. SM6150 Test Results
When TCSR_DP_PHYMODE_0 is switched to DP, the USB3 primary PHY cannot work, and the DP PHY is also not functional (possibly due to clock lack or other configuration mismatch with this TCSR setting).
When TCSR_DP_PHYMODE_1 is switched to DP, both the USB3 primary PHY and the DP PHY work normally.
I think "why msm-4.14 didn't describe second USB3 PHY", because TCSR_DP_PHYMODE_1 always works in DP mode.
https://android.googlesource.com/kernel/msm/+/af03eef7d4c3cbd1fe26c67d4f1915b05d0c1488/drivers/gpu/drm/msm/dp/dp_catalog_v200.c
Here it still programs the TCSR register.

Based on these info, I believe these are separate PHYs, and only the
TCSR DP_PHYMODE registers determine which USB3/DP PHYs are paired or
mutually exclusive. This is why I have maintained separate private
data for each PHY and implemented Power on mutex control via TCSR,
rather than using a qmp_combo-like structure.
Still, no. Check the block diagram of SM6150.

Given the above, do you think we still need to force USB and DP to be strictly bound together like a combo PHY?
Yes.
I checked the related PHY series and block diagrams again.

PRI and SEC go to different nodes based on the SoC design, and there are two types of configurations: USB3-only and USB3+DP pairing.

Before proceed the v3 patchset, I’d like to double-confirm whether the following structure is what you expect:

usb_qmpphy_1: phy@88e6000 {
     compatible = "qcom,sm6150-qmp-usb3-prim-phy"; <== rename to PRIM
No, we already have a compatible name and DT schema for this device.
Then current compatible name is "qcom,qcs615-qmp-usb3-phy" and shall we need update to "qcom,sm6150-qmp-usb3-phy"?
Why? You _already_ have a compatible string. You don't need to change it just to follow the SoC name. 
Ok, but just to confirm — in this case, the USB3-DP PHY would use "qcom,sm6150-qmp-usb3-dp-phy" while the USB3-only PHY still uses "qcom,qcs615-qmp-usb3-phy"?

Since both PHYs are on the same SoC, would it make sense to keep the naming consistent and use "qcom,qcs615-..." for both? 

     ...
     qcom,tcsr-reg = <&tcsr 0xb244>, <&tcsr 0xb248>;
     qcom,tcsr-names = "vls_clamp", "dp_phy_mode";
No need for qcom,tcsr-names. Second TCSR register should be optional in
the driver.
Ok.

          #clock-cells = <1>;
     #phy-cells = <1>;
#clock-cells = <0>;
#phy-cells = <0>;

     ...
};

usb_qmpphy_2: phy@88e8000 {
     compatible = "qcom,sm6150-qmp-usb3dp-sec-phy"; <== SEC SS, use usb3dp to indicate DP capability
qcom,sm6150-qmp-usb3-dp-phy
Ok, but for this part, shall we update dt-binding in "qcom,msm8998-qmp-usb3-phy.yaml" or create a new one?
I think (I might be wrong here) new ones is a better fit. We'll migrate the rest of PHYs to new bindings later on. 
Ok, I’ll keep the USB3-DP PHY node definition in a separate YAML file in v3.

     reg = <0x0 0x088e8000 0x0 0x2000>; <== SS2 base address and offset define in driver config

     clocks = <&gcc GCC_AHB2PHY_WEST_CLK>,
             <&gcc GCC_USB3_SEC_CLKREF_CLK>; <== This SoC has no USB3.0 SEC SS clk
     clock-names = "cfg_ahb",
                 "ref";
     clock-output-names = "dp_phy_link_clk",
                     "dp_phy_vco_div_clk";
No need to, the driver can generate names on its own.
Ok.

                          resets = <&gcc GCC_USB3PHY_PHY_SEC_BCR >,
          <&gcc GCC_USB3_DP_PHY_SEC_BCR>;
     reset-names = "phy", "phy_phy";
"phy_phy", "dp_phy". Is there no GCC_USB3_PHY_SEC_BCR?
There are only GCC_USB2_PHY_SEC_BCR and GCC_USB3PHY_PHY_SEC_BCR, no GCC_USB3_PHY_SEC_BCR.
Ack.

     qcom,tcsr-reg = <&tcsr 0xbff0>, <&tcsr 0xb24c>;
     qcom,tcsr-names = "vls_clamp", "dp_phy_mode"; <== added for backward compatibility with legacy configs that only had vls_clamp
No need for qcom,tcsr-names, correct otherwise.

     #clock-cells = <1>;
     #phy-cells = <1>;

     status = "disabled";
};

Signed-off-by: Xiangxu Yin <redacted>
---
  .../bindings/phy/qcom,qcs615-qmp-dp-phy.yaml       | 111 +++++++++++++++++++++
  1 file changed, 111 insertions(+)

-- 
linux-phy mailing list
linux-phy@lists.infradead.org
https://lists.infradead.org/mailman/listinfo/linux-phy

`h`	back out one level
`j`	next message in thread
`k`	previous message in thread
`l`	drill in
`Esc`	close help / fold thread tree
`?`	toggle this help