Re: [PATCH 2/2] usb: dwc3: imx8mp: disable auto suspend for host role

[Thinh Nguyen <Thinh.Nguyen@synopsys.com>]

On Tue, Nov 25, 2025, Xu Yang wrote:
> On Fri, Nov 21, 2025 at 12:37:47AM +0000, Thinh Nguyen wrote:
> > On Mon, Nov 17, 2025, Xu Yang wrote:
> > > On Thu, Nov 13, 2025 at 11:05:43PM +0000, Thinh Nguyen wrote:
> > > > On Wed, Nov 05, 2025, Xu Yang wrote:
> > > > > Do dwc3 core auto suspend enable for device and disable for host
> > > > > , this can make sure dwc3 core device auto suspend setting is
> > > > > correct all the time, the background of disable dwc3 core device
> > > > > auto suspend is to make its parent device suspend immediately
> > > > > (so wakeup enable can be enabled) after xhci-plat device suspended,
> > > > > for device mode, we keep the dwc3 core device auto suspend is to
> > > > > give some wait for gadget to be enumerated.
> > > > > 
> > > > > Signed-off-by: Xu Yang <xu.yang_2@nxp.com>
> > > > > ---
> > > > >  drivers/usb/dwc3/dwc3-imx8mp.c | 31 +++++++++++++++++++++++++++++++
> > > > >  1 file changed, 31 insertions(+)
> > > > > 
> > > > > diff --git a/drivers/usb/dwc3/dwc3-imx8mp.c b/drivers/usb/dwc3/dwc3-imx8mp.c
> > > > > index bce6af82f54c..80a431cb6fed 100644
> > > > > --- a/drivers/usb/dwc3/dwc3-imx8mp.c
> > > > > +++ b/drivers/usb/dwc3/dwc3-imx8mp.c
> > > > > @@ -158,11 +158,31 @@ static irqreturn_t dwc3_imx8mp_interrupt(int irq, void *_dwc3_imx)
> > > > >  	return IRQ_HANDLED;
> > > > >  }
> > > > >  
> > > > > +static void dwc3_imx_pre_set_role(struct dwc3 *dwc, enum usb_role role)
> > > > > +{
> > > > > +	if (role == USB_ROLE_HOST)
> > > > > +		/*
> > > > > +		 * For xhci host, we need disable dwc core auto
> > > > > +		 * suspend, because during this auto suspend delay(5s),
> > > > > +		 * xhci host RUN_STOP is cleared and wakeup is not
> > > > > +		 * enabled, if device is inserted, xhci host can't
> > > > > +		 * response the connection.
> > > > > +		 */
> > > > > +		pm_runtime_dont_use_autosuspend(dwc->dev);
> > > > > +	else
> > > > > +		pm_runtime_use_autosuspend(dwc->dev);
> > > > 
> > > > Would this override the user setting everytime there's a role switch?
> > > 
> > > From what I know, the user can't control whether to enable or disable
> > > autosuspend feature. So there should be no overriding problem. When
> > > user change autosuspend_delay_ms, the delay setting will be kept
> > > everytime there's a role switch.
> > 
> > Should we set pm_runtime_mark_last_busy()?
> 
> I think we needn't explicitly call pm_runtime_mark_last_busy().
> Since commit "08071e64cb64 PM: runtime: Mark last busy stamp in pm_runtime_autosuspend()",
> pm_runtime_autosuspend() will do that thing automatically once dwc3 device is idle.

Ok.

> 
> > 
> > Also, if pm_runtime_use_autosuspend() is set, make sure to unwind with
> > pm_runtime_dont_use_autosuspend() on remove.
> > 
> > > 
> > > > 
> > > > > +}
> > > > > +
> > > > > +struct dwc3_glue_ops dwc3_imx_glue_ops = {
> > > > > +	.pre_set_role   = dwc3_imx_pre_set_role,
> > > > > +};
> > > > > +
> > > > >  static int dwc3_imx8mp_probe(struct platform_device *pdev)
> > > > >  {
> > > > >  	struct device		*dev = &pdev->dev;
> > > > >  	struct device_node	*node = dev->of_node;
> > > > >  	struct dwc3_imx8mp	*dwc3_imx;
> > > > > +	struct dwc3		*dwc3;
> > > > >  	struct resource		*res;
> > > > >  	int			err, irq;
> > > > >  
> > > > > @@ -240,6 +260,17 @@ static int dwc3_imx8mp_probe(struct platform_device *pdev)
> > > > >  		goto depopulate;
> > > > >  	}
> > > > >  
> > > > > +	dwc3 = platform_get_drvdata(dwc3_imx->dwc3);
> > > > 
> > > > It's confusing how dwc3_imx->dwc3 is also named dwc3...
> > > 
> > > I will rename it later.
> > > 
> > > > 
> > > > > +	if (!dwc3) {
> > > > > +		err = dev_err_probe(dev, -EPROBE_DEFER, "failed to get dwc3 platform data\n");
> > > > > +		goto depopulate;
> > > > > +	}
> > > > > +
> > > > > +	dwc3->glue_ops = &dwc3_imx_glue_ops;
> > > > 
> > > > If you want to use glue_ops, please use the new flatten model. I
> > > > don't want dwc3 to be initialized again after of_platform_populate().
> > > 
> > > I understand the new flatten model is a more suitable way. Considering that
> > > many i.MX devices are using this legacy unflatten model, do you mind allow
> > > this change firstly and we will switch to the new flatten model in the future?
> > > 
> > 
> > Do you have plans for this transition? Perhaps this is a good time to
> > initiate the process so to avoid having and maintaining this improper
> > code while we still don't have it yet.
> 
> OK. It's a trend to use new flatten model. I do have plan for it. I'll
> prepare for it later.
> 

Ok. Can we add a note that this is a temporary solution until we switch
to using the new model in the commit message.

Thanks,
Thinh