[PATCH 2/3] usb: dwc3: add may_lose_power flag

[Xu Yang]

On Wed, Feb 04, 2026 at 01:52:15AM +0000, Thinh Nguyen wrote:
> On Tue, Feb 03, 2026, Peng Fan wrote:
> > On Tue, Feb 03, 2026 at 12:24:08AM +0000, Thinh Nguyen wrote:
> > >On Mon, Feb 02, 2026, Xu Yang wrote:
> > >> The current design assumes that the controller remains powered
> > >> when wakeup is enabled. However, some SoCs provide wakeup
> > >> capability even when the controller itself is powered down, using
> > >> separate dedicated wakeup logic. This allows additional power
> > >> savings, but requires the controller to be fully re‑initialized
> > >> after system resume.
> > >> 
> > >> To support these SoCs, introduce a flag to track the controller’s
> > >> power state and use it throughout the suspend/resume flow.
> > >> 
> > >> Signed-off-by: Xu Yang <xu.yang_2 at nxp.com>
> > >> ---
> > >>  drivers/usb/dwc3/core.c | 9 +++++++--
> > >>  drivers/usb/dwc3/core.h | 2 ++
> > >>  drivers/usb/dwc3/glue.h | 3 +++
> > >>  3 files changed, 12 insertions(+), 2 deletions(-)
> > >> 
> > >> diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c
> > >> index c07ffe82c85049364c38c7ba152aab0ff764d95e..9d4326da5ec7669fa714707fb24556723cab51b8 100644
> > >> --- a/drivers/usb/dwc3/core.c
> > >> +++ b/drivers/usb/dwc3/core.c
> > >> @@ -2314,6 +2314,9 @@ int dwc3_core_probe(const struct dwc3_probe_data *data)
> > >>  			goto err_exit_debugfs;
> > >>  	}
> > >>  
> > >> +	if (data->core_may_lose_power)
> > >
> > >Can this be passed down as part of the dwc3_properties within the
> > >probe_data? I'm trying to consolidate all the dwc3 properties to one
> > >place.
> > >
> > >> +		dwc->may_lose_power = true;
> > >> +
> > >>  	pm_runtime_put(dev);
> > >>  
> > >>  	dma_set_max_seg_size(dev, UINT_MAX);
> > >> @@ -2462,7 +2465,8 @@ static int dwc3_suspend_common(struct dwc3 *dwc, pm_message_t msg)
> > >>  		dwc3_core_exit(dwc);
> > >>  		break;
> > >>  	case DWC3_GCTL_PRTCAP_HOST:
> > >> -		if (!PMSG_IS_AUTO(msg) && !device_may_wakeup(dwc->dev)) {
> > >> +		if (!PMSG_IS_AUTO(msg) &&
> > >> +		    (!device_may_wakeup(dwc->dev) || dwc->may_lose_power)) {
> > >>  			dwc3_core_exit(dwc);
> > >>  			break;
> > >>  		}
> > >> @@ -2525,7 +2529,8 @@ static int dwc3_resume_common(struct dwc3 *dwc, pm_message_t msg)
> > >>  		dwc3_gadget_resume(dwc);
> > >>  		break;
> > >>  	case DWC3_GCTL_PRTCAP_HOST:
> > >> -		if (!PMSG_IS_AUTO(msg) && !device_may_wakeup(dwc->dev)) {
> > >> +		if (!PMSG_IS_AUTO(msg) &&
> > >> +		    (!device_may_wakeup(dwc->dev) || dwc->may_lose_power)) {
> > >>  			ret = dwc3_core_init_for_resume(dwc);
> > >>  			if (ret)
> > >>  				return ret;
> > >> diff --git a/drivers/usb/dwc3/core.h b/drivers/usb/dwc3/core.h
> > >> index 08cc6f2b5c23631a752c77fd7394e5876c929f0a..5b1358f36490a001bc9e68139224f7be70a57995 100644
> > >> --- a/drivers/usb/dwc3/core.h
> > >> +++ b/drivers/usb/dwc3/core.h
> > >> @@ -1117,6 +1117,7 @@ struct dwc3_glue_ops {
> > >>   * @usb3_lpm_capable: set if hadrware supports Link Power Management
> > >>   * @usb2_lpm_disable: set to disable usb2 lpm for host
> > >>   * @usb2_gadget_lpm_disable: set to disable usb2 lpm for gadget
> > >> + * @may_lose_power: set to indicate the core may lose power during pm suspend
> > >>   * @disable_scramble_quirk: set if we enable the disable scramble quirk
> > >>   * @u2exit_lfps_quirk: set if we enable u2exit lfps quirk
> > >>   * @u2ss_inp3_quirk: set if we enable P3 OK for U2/SS Inactive quirk
> > >> @@ -1369,6 +1370,7 @@ struct dwc3 {
> > >>  	unsigned		usb3_lpm_capable:1;
> > >>  	unsigned		usb2_lpm_disable:1;
> > >>  	unsigned		usb2_gadget_lpm_disable:1;
> > >> +	unsigned		may_lose_power:1;
> > >
> > >This name sounds like a quirk of a broken SoC.
> > >
> > >Perhaps rename this to something such as power_lost_on_suspend or
> > >needs_full_reinit?
> > 
> > How about out_band_wakeup?
> > 
> 
> IMHO, it's not as clear. It doesn't describe the problem or the action
> to take. Can we use needs_full_reinit?

Thanks for the suggestion.

Agree with Thinh. And checking out_band_wakeup flag during system suspend 
implies that a wakeup event was triggered.

Sure, will user needs_full_reinit.

Thanks,
Xu Yang