Re: [PATCH v5 5/5] usb: host: enable sideband transfer during system sleep

[Greg KH]

On Mon, Oct 14, 2024 at 08:50:29AM +0000, Guan-Yu Lin wrote:
> Sharing a USB controller with another entity via xhci-sideband driver
> creates power management complexities. To prevent the USB controller
> from being inadvertently deactivated while in use by the other entity, a
> usage-count based mechanism is implemented. This allows the system to
> manage power effectively, ensuring the controller remains available
> whenever needed.
> 
> Signed-off-by: Guan-Yu Lin <guanyulin@xxxxxxxxxx>
> ---
>  drivers/usb/core/driver.c         | 10 ++++++++++
>  drivers/usb/core/hcd.c            |  1 +
>  drivers/usb/core/usb.c            |  1 +
>  drivers/usb/dwc3/core.c           | 13 +++++++++++++
>  drivers/usb/dwc3/core.h           |  8 ++++++++
>  drivers/usb/host/xhci-plat.c      | 10 ++++++++++
>  drivers/usb/host/xhci-plat.h      |  7 +++++++
>  sound/usb/qcom/qc_audio_offload.c |  3 +++
>  8 files changed, 53 insertions(+)
> 
> diff --git a/drivers/usb/core/driver.c b/drivers/usb/core/driver.c
> index e713cf9b3dd2..eb85cbb1a2ff 100644
> --- a/drivers/usb/core/driver.c
> +++ b/drivers/usb/core/driver.c
> @@ -1583,6 +1583,11 @@ int usb_suspend(struct device *dev, pm_message_t msg)
>  	struct usb_device	*udev = to_usb_device(dev);
>  	int r;
>  
> +	if (msg.event == PM_EVENT_SUSPEND && usb_sideband_check(udev)) {
> +		dev_dbg(dev, "device accessed via sideband\n");
> +		return 0;
> +	}

What prevents the check from changing state right after you call this?

> +
>  	unbind_no_pm_drivers_interfaces(udev);
>  
>  	/* From now on we are sure all drivers support suspend/resume
> @@ -1619,6 +1624,11 @@ int usb_resume(struct device *dev, pm_message_t msg)
>  	struct usb_device	*udev = to_usb_device(dev);
>  	int			status;
>  
> +	if (msg.event == PM_EVENT_RESUME && usb_sideband_check(udev)) {
> +		dev_dbg(dev, "device accessed via sideband\n");
> +		return 0;
> +	}

Same here, what's keeping the state from changing?


> +
>  	/* For all calls, take the device back to full power and
>  	 * tell the PM core in case it was autosuspended previously.
>  	 * Unbind the interfaces that will need rebinding later,
> diff --git a/drivers/usb/core/hcd.c b/drivers/usb/core/hcd.c
> index 1ff7d901fede..9876b3940281 100644
> --- a/drivers/usb/core/hcd.c
> +++ b/drivers/usb/core/hcd.c
> @@ -2593,6 +2593,7 @@ struct usb_hcd *__usb_create_hcd(const struct hc_driver *driver,
>  	timer_setup(&hcd->rh_timer, rh_timer_func, 0);
>  #ifdef CONFIG_PM
>  	INIT_WORK(&hcd->wakeup_work, hcd_resume_work);
> +	refcount_set(&hcd->sb_usage_count, 0);
>  #endif
>  
>  	INIT_WORK(&hcd->died_work, hcd_died_work);
> diff --git a/drivers/usb/core/usb.c b/drivers/usb/core/usb.c
> index 0b4685aad2d5..d315d066a56b 100644
> --- a/drivers/usb/core/usb.c
> +++ b/drivers/usb/core/usb.c
> @@ -671,6 +671,7 @@ struct usb_device *usb_alloc_dev(struct usb_device *parent,
>  	dev->state = USB_STATE_ATTACHED;
>  	dev->lpm_disable_count = 1;
>  	atomic_set(&dev->urbnum, 0);
> +	refcount_set(&dev->sb_usage_count, 0);
>  
>  	INIT_LIST_HEAD(&dev->ep0.urb_list);
>  	dev->ep0.desc.bLength = USB_DT_ENDPOINT_SIZE;
> diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c
> index 2fdafbcbe44c..18c743ce5ac5 100644
> --- a/drivers/usb/dwc3/core.c
> +++ b/drivers/usb/dwc3/core.c
> @@ -2550,8 +2550,15 @@ static int dwc3_runtime_idle(struct device *dev)
>  static int dwc3_suspend(struct device *dev)
>  {
>  	struct dwc3	*dwc = dev_get_drvdata(dev);
> +	struct platform_device *xhci = dwc->xhci;
>  	int		ret;
>  
> +	if (xhci && xhci_sideband_check(xhci->dev.driver_data)) {

What could go wrong with poking into a random device structure's private
data that you don't know the type of?  :(

> +		dev_dbg(dev, "device accessed via sideband\n");
> +		return 0;

I predict, that if this all does get implemented, we're going to have a
lot of confusion of "why will my devices not go into suspend?"
questions, right?

How does userspace know if a device is controlled by a sideband path or
not?  Is there some sysfs link somewhere, and does any tool show it
anyway?

thanks,

greg k-h