Re: [PATCH 1/3] arm64: dts: qcom: eliza: Describe the ADSP and USB related nodes

[Konrad Dybcio]

On 4/28/26 7:46 AM, Krishna Kurapati PSSNV wrote:
> On Wed, Apr 22, 2026 at 3:55 PM Abel Vesa <abel.vesa@xxxxxxxxxxxxxxxx> wrote:
>>
>> On 26-04-22 12:09:31, Konrad Dybcio wrote:
>>> On 4/22/26 11:41 AM, Abel Vesa wrote:
>>>> On 26-03-31 15:37:08, Konrad Dybcio wrote:
>>>>> On 3/31/26 12:37 PM, Abel Vesa wrote:
>>>>>> Describe the ADSP remoteproc node along with its dependencies, including
>>>>>> the IPCC mailbox, AOSS QMP and SMP2P links used for communication.
>>>>>>
>>>>>> The Eliza SoC features a USB 3.1 Gen 2 controller connected to a QMP
>>>>>> combo PHY and an SNPS eUSB2 PHY. Describe them.
>>>>>>
>>>>>> Signed-off-by: Abel Vesa <abel.vesa@xxxxxxxxxxxxxxxx>
>>>>>> ---
>>>>>
>>>>> [...]
>>>>>
>>>>>> +         usb_hsphy: phy@88e3000 {
>>>>>> +                 compatible = "qcom,eliza-snps-eusb2-phy",
>>>>>> +                              "qcom,sm8550-snps-eusb2-phy";
>>>>>> +                 reg = <0x0 0x088e3000 0x0 0x154>;
>>>>>> +                 #phy-cells = <0>;
>>>>>> +
>>>>>> +                 clocks = <&rpmhcc RPMH_CXO_CLK>;
>>>>>
>>>>> This is TCSR_USB2_CLKREF_EN
>>>>
>>>> Good point. Will fix.
>>>>
>>>>>
>>>>>
>>>>>> +         usb: usb@a600000 {
>>>>>> +                 compatible = "qcom,eliza-dwc3", "qcom,snps-dwc3";
>>>>>
>>>>> Does the device suspend and resume successfully?
>>>>
>>>> Well, tested with pm_test devices and it does suspend and resume
>>>> successfully, but there is this:
>>>>
>>>> [   54.584126] dwc3-qcom a600000.usb: port-1 HS-PHY not in L2
>>>>
>>>> But if I'm not mistaken, this is valid accross all SNPS eUSB2 PHYs, on
>>>> all platforms that have them.
>>>
>>> Well it's not fatal, but ideally this wouldn't be there. Maybe you're missing
>>> some DWC quirk in the list, although it seems pretty long already. Perhaps
>>> Wesley would know more.
>>
>> + Wesley
>>
> 
> As per HPG and downstream, this is what needs to be done while entering suspend:
> 
> 1. Clear PWR_EVNT_LPM_IN_L2_MASK bit of pwr_evnt_irq_stat_reg
> 2. Clear PWR_EVNT_LPM_OUT_L2_MASK bit of pwr_evnt_irq_stat_reg
> 3. Set the following bits in the pwr_evnt_irq_stat_reg:
>  a)  DWC3_GUSB2PHYCFG_ENBLSLPM  and  DWC3_GUSB2PHYCFG_SUSPHY

In case that's related, most platforms (including this one), set
snps,dis_enblslpm_quirk which prevents the first bit from being set

Likewise, snps,dis_u2_susphy_quirk for the second one

(although it looks like setting these bits is currently
unconditional upon suspend in HOST mode?)


As for the sequence you mentioned, I believe the diff below should be
OK - although it _really_ just adds some delay vs the current state,
since the bits are cleared in the resume call

diff --git a/drivers/usb/dwc3/dwc3-qcom.c b/drivers/usb/dwc3/dwc3-qcom.c
index f43f73ac36ff..e7b1775b7207 100644
--- a/drivers/usb/dwc3/dwc3-qcom.c
+++ b/drivers/usb/dwc3/dwc3-qcom.c
@@ -12,6 +12,7 @@
 #include <linux/module.h>
 #include <linux/kernel.h>
 #include <linux/interconnect.h>
+#include <linux/iopoll.h>
 #include <linux/platform_device.h>
 #include <linux/phy/phy.h>
 #include <linux/usb/of.h>
@@ -344,10 +345,18 @@ static int dwc3_qcom_suspend(struct dwc3_qcom *qcom, bool wakeup)
                return 0;
 
        for (i = 0; i < qcom->num_ports; i++) {
-               val = readl(qcom->qscratch_base + pwr_evnt_irq_stat_reg[i]);
-               if (!(val & PWR_EVNT_LPM_IN_L2_MASK))
+               /* Wait for the PHYs to go into L2 */
+               ret = readl_poll_timeout(qcom->qscratch_base + pwr_evnt_irq_stat_reg[i],
+                                        val, val & PWR_EVNT_LPM_IN_L2_MASK,
+                                        10, 3 * USEC_PER_MSEC);
+
+               if (ret == -ETIMEDOUT)
                        dev_err(qcom->dev, "port-%d HS-PHY not in L2\n", i + 1);
        }
+
+       /* Clear L2 event bit */
+       writel(PWR_EVNT_LPM_IN_L2_MASK, qcom->qscratch_base + pwr_evnt_irq_stat_reg[i]);
+
        clk_bulk_disable_unprepare(qcom->num_clocks, qcom->clks);
 
        ret = dwc3_qcom_interconnect_disable(qcom);
@@ -752,6 +761,13 @@ static int dwc3_qcom_pm_suspend(struct device *dev)
        bool wakeup = device_may_wakeup(dev);
        int ret;
 
+       /* Clear previous L2 events */
+       for (int i = 0; i < qcom->num_ports; i++) {
+               writel(PWR_EVNT_LPM_IN_L2_MASK | PWR_EVNT_LPM_OUT_L2_MASK,
+                      qcom->qscratch_base + pwr_evnt_irq_stat_reg[i]);
+       }
+
        ret = dwc3_pm_suspend(&qcom->dwc);
        if (ret)
                return ret;


Konrad