Re: [PATCH] PNPACPI: add support for remote wakeup

[Bjorn Helgaas <bjorn.helgaas@hp.com>]

On Monday 08 March 2010 02:50:50 pm Alan Stern wrote:
> This patch (as1354) adds remote-wakeup support to the pnpacpi driver.
> The new can_wakeup method also allows other PNP protocol drivers
> (pnpbios or iaspnp) to add wakeup support, but I don't know enough
> about how they work to actually do it.
> 
> Signed-off-by: Alan Stern <stern@rowland.harvard.edu>

We usually send PNP patches through the ACPI tree, so I added a CC
to linux-acpi.

> --- usb-2.6.orig/drivers/pnp/core.c
> +++ usb-2.6/drivers/pnp/core.c
> @@ -164,6 +164,9 @@ int __pnp_add_device(struct pnp_dev *dev
>  	list_add_tail(&dev->global_list, &pnp_global);
>  	list_add_tail(&dev->protocol_list, &dev->protocol->devices);
>  	spin_unlock(&pnp_lock);
> +	if (dev->protocol->can_wakeup)
> +		device_set_wakeup_capable(&dev->dev,
> +				dev->protocol->can_wakeup(dev));

I also added Rafael because he added code in acpi_bind_one() that
does the same thing.  I think the struct dev there will be the same
one as &dev->dev here: we build both an acpi_device and a pnp_dev,
and they refer to the same struct device.

However, I think we still need your patch because acpi_bind_one()
is only used for PCI devices, so it looks like there's currently
no way to use acpi_pm_device_sleep_wake() for non-PCI devices.

Reviewed-by: Bjorn Helgaas <bjorn.helgaas@hp.com>

>  	return device_register(&dev->dev);
>  }
>  
> Index: usb-2.6/drivers/pnp/pnpacpi/core.c
> ===================================================================
> --- usb-2.6.orig/drivers/pnp/pnpacpi/core.c
> +++ usb-2.6/drivers/pnp/pnpacpi/core.c
> @@ -121,17 +121,37 @@ static int pnpacpi_disable_resources(str
>  }
>  
>  #ifdef CONFIG_ACPI_SLEEP
> +static bool pnpacpi_can_wakeup(struct pnp_dev *dev)
> +{
> +	struct acpi_device *acpi_dev = dev->data;
> +	acpi_handle handle = acpi_dev->handle;
> +
> +	return acpi_bus_can_wakeup(handle);
> +}
> +
>  static int pnpacpi_suspend(struct pnp_dev *dev, pm_message_t state)
>  {
>  	struct acpi_device *acpi_dev = dev->data;
>  	acpi_handle handle = acpi_dev->handle;
>  	int power_state;
>  
> +	if (device_can_wakeup(&dev->dev)) {
> +		int rc = acpi_pm_device_sleep_wake(&dev->dev,
> +				device_may_wakeup(&dev->dev));
> +
> +		if (rc)
> +			return rc;
> +	}
>  	power_state = acpi_pm_device_sleep_state(&dev->dev, NULL);
>  	if (power_state < 0)
>  		power_state = (state.event == PM_EVENT_ON) ?
>  				ACPI_STATE_D0 : ACPI_STATE_D3;
>  
> +	/* acpi_bus_set_power() often fails (keyboard port can't be
> +	 * powered-down?), and in any case, our return value is ignored
> +	 * by pnp_bus_suspend().  Hence we don't revert the wakeup
> +	 * setting if the set_power fails.
> +	 */
>  	return acpi_bus_set_power(handle, power_state);
>  }
>  
> @@ -140,6 +160,8 @@ static int pnpacpi_resume(struct pnp_dev
>  	struct acpi_device *acpi_dev = dev->data;
>  	acpi_handle handle = acpi_dev->handle;
>  
> +	if (device_may_wakeup(&dev->dev))
> +		acpi_pm_device_sleep_wake(&dev->dev, false);
>  	return acpi_bus_set_power(handle, ACPI_STATE_D0);
>  }
>  #endif
> @@ -150,6 +172,7 @@ struct pnp_protocol pnpacpi_protocol = {
>  	.set	 = pnpacpi_set_resources,
>  	.disable = pnpacpi_disable_resources,
>  #ifdef CONFIG_ACPI_SLEEP
> +	.can_wakeup = pnpacpi_can_wakeup,
>  	.suspend = pnpacpi_suspend,
>  	.resume = pnpacpi_resume,
>  #endif
> 
>