On Sat, 23 Mar 2013, Lan, Tianyu wrote: > On 2013/3/23 1:35, Alan Stern wrote:> Tianyu: > > > > Did you ever test the port power-off mechanism with system suspend? > > Right now it doesn't seem like it would work, because it relies on > > runtime PM to turn off the port power, and runtime PM doesn't operate > > normally during a system suspend. > > > > Alan Stern > > > > Hi Alan: > You are right. Current the port power-off can't work during > system suspend. I have already worked on a patch of adding usb > port system pm support. But there is a still problem that how to > control power off mechanism from kernel side. Before, we have a solution > for runtime pm that increase/decrease port dev's usage count. > http://marc.info/?l=linux-usb&m=135772122031446&w=2 > But now, we have another user, System pm. So I think we should find > a solution to cover these two cases. Maybe add a new pm qos request > for NO_POWER_OFF flag? I'd like to see your opinion. Yes, I think you need to copy the test that's in usb_port_suspend. Better yet, move that test to a separate function in port.c and call it from both places. > I have a patch that can work and it still needs to add some condition > checks in the usb_port_system_suspend(). Dev's system wakeup disable > and persist enable. ... > +static int usb_port_system_suspend(struct device *dev) > +{ > + struct usb_port *port_dev = to_usb_port(dev); > + int retval; > + > + if (port_dev->child) > + device_pm_wait_for_dev(dev, &port_dev->child->dev); > + > + retval = usb_port_runtime_suspend(dev); > + if (retval < 0 && retval != -EAGAIN) > + return retval; > + > + return 0; This should always return 0. Ignore any errors from usb_port_runtime_suspend. > +} > + > +static int usb_port_system_resume(struct device *dev) > +{ > + return usb_port_runtime_resume(dev); > +} > + These new functions should be protected by #ifdef CONFIG_PM_SLEEP. > static const struct dev_pm_ops usb_port_pm_ops = { > + .suspend = usb_port_system_suspend, > + .resume = usb_port_system_resume, And the same goes for these entries. The runtime routines now need to be protected by CONFIG_PM, since they can be used if either PM_RUNTIME or PM_SLEEP is enabled. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
