Hi Felipe,
Hi Martin,

I saw the issue reported ([balbi-usb:testing/next 2/17] 
drivers/usb/dwc2/drd.c:80:38: error: no member named 'test_mode' in 
'struct dwc2_hsotg'). I prepare a V2 fixing it + adressing Martin's 
comments.

Regards,
Amelie

On 7/19/20 9:56 PM, Martin Blumenstingl wrote:

Hello Amelie,

sorry for the late reply

On Wed, Jul 8, 2020 at 6:00 PM Amelie DELAUNAY [off-list ref] 
wrote:
[...]

quoted

Could you please test with:

static int dwc2_drd_role_sw_set(struct device *dev, enum usb_role role)
{
        struct dwc2_hsotg *hsotg = dev_get_drvdata(dev);
        unsigned long flags;
        int already = 0;

        /* Skip session not in line with dr_mode */
        if ((role == USB_ROLE_DEVICE && hsotg->dr_mode == USB_DR_MODE_HOST) ||
            (role == USB_ROLE_HOST && hsotg->dr_mode == USB_DR_MODE_PERIPHERAL))
                return -EINVAL;

        /* Skip session if core is in test mode */
        if (role == USB_ROLE_NONE && hsotg->test_mode) {
                dev_dbg(hsotg->dev, "Core is in test mode\n");
                return -EBUSY;
        }

        spin_lock_irqsave(&hsotg->lock, flags);

        if (role == USB_ROLE_HOST) {
                already = dwc2_ovr_avalid(hsotg, true);
        } else if (role == USB_ROLE_DEVICE) {
                already = dwc2_ovr_bvalid(hsotg, true);
                /* This clear DCTL.SFTDISCON bit */
                dwc2_hsotg_core_connect(hsotg);
        } else {
                if (dwc2_is_device_mode(hsotg)) {
                    if (!dwc2_ovr_bvalid(hsotg, false))
                        /* This set DCTL.SFTDISCON bit */
                        dwc2_hsotg_core_disconnect(hsotg);
                } else {
                        dwc2_ovr_avalid(hsotg, false);
                }
        }

        spin_unlock_irqrestore(&hsotg->lock, flags);

        if (!already &&
            role != USB_ROLE_NONE && hsotg->dr_mode == USB_DR_MODE_OTG)
                /* This will raise a Connector ID Status Change Interrupt */
                dwc2_force_mode(hsotg, role == USB_ROLE_HOST);

        dev_dbg(hsotg->dev, "%s-session valid\n",
                role == USB_ROLE_NONE ? "No" :
                role == USB_ROLE_HOST ? "A" : "B");

        return 0;
}


dwc2_force_mode is called outside the spin_lock_irqsave so the kernel
should not complain. I've tested on my setup and the behavior seems the
same.

this one is looking good - the previous kernel warnings are now gone!
thank you very much


Best regards,
Martin