[PATCH] PM: allow runtime_suspend methods to call pm_schedule_suspend

[PATCH] PM: allow runtime_suspend methods to call pm_schedule_suspend

[Alan Stern]

This patch (as1361) changes the runtime PM interface slightly; it
allows suspend requests to be scheduled while the runtime_suspend
method is running.  If the method succeeds then the scheduled request
is cancelled, whereas if the method fails then an idle notification is
sent only if no request was scheduled.

Being able to schedule suspend requests from within a runtime_suspend
method is useful for drivers that need to test for idleness and
suspend the device all while holding a single spinlock, or for drivers
that want to check for idleness by polling.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>

---

Index: usb-2.6/drivers/base/power/runtime.c
===================================================================
--- usb-2.6.orig/drivers/base/power/runtime.c
+++ usb-2.6/drivers/base/power/runtime.c
@@ -229,14 +229,16 @@ int __pm_runtime_suspend(struct device *
 
 	if (retval) {
 		dev->power.runtime_status = RPM_ACTIVE;
-		pm_runtime_cancel_pending(dev);
-
 		if (retval == -EAGAIN || retval == -EBUSY) {
-			notify = true;
+			if (dev->power.timer_expires == 0)
+				notify = true;
 			dev->power.runtime_error = 0;
+		} else {
+			pm_runtime_cancel_pending(dev);
 		}
 	} else {
 		dev->power.runtime_status = RPM_SUSPENDED;
+		pm_runtime_deactivate_timer(dev);
 
 		if (dev->parent) {
 			parent = dev->parent;
@@ -659,8 +661,6 @@ int pm_schedule_suspend(struct device *d
 
 	if (dev->power.runtime_status == RPM_SUSPENDED)
 		retval = 1;
-	else if (dev->power.runtime_status == RPM_SUSPENDING)
-		retval = -EINPROGRESS;
 	else if (atomic_read(&dev->power.usage_count) > 0
 	    || dev->power.disable_depth > 0)
 		retval = -EAGAIN;

Re: [PATCH] PM: allow runtime_suspend methods to call pm_schedule_suspend

[Rafael J. Wysocki]

On Friday 19 March 2010, Alan Stern wrote:
> This patch (as1361) changes the runtime PM interface slightly; it
> allows suspend requests to be scheduled while the runtime_suspend
> method is running.  If the method succeeds then the scheduled request
> is cancelled, whereas if the method fails then an idle notification is
> sent only if no request was scheduled.
> 
> Being able to schedule suspend requests from within a runtime_suspend
> method is useful for drivers that need to test for idleness and
> suspend the device all while holding a single spinlock, or for drivers
> that want to check for idleness by polling.
> 
> Signed-off-by: Alan Stern <stern@rowland.harvard.edu>

Looks good.

Do you need it in 2.6.34 or is it sufficient to queue it up for 2.6.35 (I'd prefer so)?

Rafael


> ---
> 
> Index: usb-2.6/drivers/base/power/runtime.c
> ===================================================================
> --- usb-2.6.orig/drivers/base/power/runtime.c
> +++ usb-2.6/drivers/base/power/runtime.c
> @@ -229,14 +229,16 @@ int __pm_runtime_suspend(struct device *
>  
>  	if (retval) {
>  		dev->power.runtime_status = RPM_ACTIVE;
> -		pm_runtime_cancel_pending(dev);
> -
>  		if (retval == -EAGAIN || retval == -EBUSY) {
> -			notify = true;
> +			if (dev->power.timer_expires == 0)
> +				notify = true;
>  			dev->power.runtime_error = 0;
> +		} else {
> +			pm_runtime_cancel_pending(dev);
>  		}
>  	} else {
>  		dev->power.runtime_status = RPM_SUSPENDED;
> +		pm_runtime_deactivate_timer(dev);
>  
>  		if (dev->parent) {
>  			parent = dev->parent;
> @@ -659,8 +661,6 @@ int pm_schedule_suspend(struct device *d
>  
>  	if (dev->power.runtime_status == RPM_SUSPENDED)
>  		retval = 1;
> -	else if (dev->power.runtime_status == RPM_SUSPENDING)
> -		retval = -EINPROGRESS;
>  	else if (atomic_read(&dev->power.usage_count) > 0
>  	    || dev->power.disable_depth > 0)
>  		retval = -EAGAIN;

Re: [PATCH] PM: allow runtime_suspend methods to call pm_schedule_suspend

[Alan Stern]

On Fri, 19 Mar 2010, Rafael J. Wysocki wrote:

> On Friday 19 March 2010, Alan Stern wrote:
> > This patch (as1361) changes the runtime PM interface slightly; it
> > allows suspend requests to be scheduled while the runtime_suspend
> > method is running.  If the method succeeds then the scheduled request
> > is cancelled, whereas if the method fails then an idle notification is
> > sent only if no request was scheduled.
> > 
> > Being able to schedule suspend requests from within a runtime_suspend
> > method is useful for drivers that need to test for idleness and
> > suspend the device all while holding a single spinlock, or for drivers
> > that want to check for idleness by polling.
> > 
> > Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
> 
> Looks good.
> 
> Do you need it in 2.6.34 or is it sufficient to queue it up for 2.6.35 (I'd prefer so)?

2.6.35 is good enough.

Alan Stern