Re: [PATCH RFC] PM / Runtime: Rework the "runtime idle" helper routine

[Alan Stern <stern@rowland.harvard.edu>]

On Wed, 29 May 2013, Rafael J. Wysocki wrote:

> From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> 
> The "runtime idle" helper routine, rpm_idle(), currently ignores
> return values from .runtime_idle() callbacks executed by it.
> 
> However, it turns out that many subsystems use the generic idle
> callback routine pm_generic_runtime_idle() which checks the return
> value of the driver's callback and executes pm_runtime_suspend() for
> the device unless that value is different from 0.  If that logic is
> moved to rpm_idle() instead, pm_generic_runtime_idle() can be dropped
> and its users will not need any .runtime_idle() callbacks any more.

Since you're making this change, wouldn't it be a good idea to adopt
Mika's original suggestion and turn on the RPM_AUTO bit in rpmflags
when the use_autosuspend flag is set?

> Moreover, the PCI subsystem's .runtime_idle() routine,
> pci_pm_runtime_idle(), works in analogy with the generic one and if
> rpm_idle() calls rpm_suspend() after 0 has been returned by the
> .runtime_idle() callback executed by it, that routine will not be
> necessary any more and may be dropped.

See below.

What about cases where the runtime-idle callback does
rpm_schedule_suspend or rpm_request_suspend?  You'd have to make sure
that it returns -EBUSY in such cases.  Did you audit for this?

> Index: linux-pm/Documentation/power/runtime_pm.txt
> ===================================================================
> --- linux-pm.orig/Documentation/power/runtime_pm.txt
> +++ linux-pm/Documentation/power/runtime_pm.txt
> @@ -660,11 +660,6 @@ Subsystems may wish to conserve code spa
>  management callbacks provided by the PM core, defined in
>  driver/base/power/generic_ops.c:
>  
> -  int pm_generic_runtime_idle(struct device *dev);
> -    - invoke the ->runtime_idle() callback provided by the driver of this
> -      device, if defined, and call pm_runtime_suspend() for this device if the
> -      return value is 0 or the callback is not defined
> -

The documentation for the runtime-idle callback needs to be updated too.

> Index: linux-pm/drivers/pci/pci-driver.c
> ===================================================================
> --- linux-pm.orig/drivers/pci/pci-driver.c
> +++ linux-pm/drivers/pci/pci-driver.c
> @@ -1046,32 +1046,6 @@ static int pci_pm_runtime_resume(struct
>  	return rc;
>  }
>  
> -static int pci_pm_runtime_idle(struct device *dev)
> -{
> -	struct pci_dev *pci_dev = to_pci_dev(dev);
> -	const struct dev_pm_ops *pm = dev->driver ? dev->driver->pm : NULL;
> -
> -	/*
> -	 * If pci_dev->driver is not set (unbound), the device should
> -	 * always remain in D0 regardless of the runtime PM status
> -	 */
> -	if (!pci_dev->driver)
> -		goto out;
> -
> -	if (!pm)
> -		return -ENOSYS;
> -
> -	if (pm->runtime_idle) {
> -		int ret = pm->runtime_idle(dev);
> -		if (ret)
> -			return ret;
> -	}
> -
> -out:
> -	pm_runtime_suspend(dev);
> -	return 0;
> -}

This may not be a safe change, because now the behavior is different
in the case where dev->driver is set but pci_dev->driver isn't.  The
difference is that you will now call the driver's runtime-idle
handler, whereas the existing code doesn't.

In fact, this may turn out to be a more widespread problem.  
dev->driver gets set before the probe routine is called, and it gets
cleared after the remove routine is called.  A runtime PM callback to
the driver during these windows isn't a good idea.  Erasing subsystems'
runtime_idle handlers, as this patch does, makes it impossible for the
subsystems to protect against this.

The patch also needs to update
drivers/usb/core/driver.c:usb_runtime_idle().  If you include Mika's
suggestion, the routine can be removed entirely.

Alan Stern