Oliver: There are a couple of significant differences between the existing USB runtime PM code and the new framework. The first has to do with intf->pm_usage_cnt. In the framework each struct device already has its own PM usage counter, so keeping our own would be redundant. On the other hand, without it we have to depend on drivers to make sure that they correctly balance autopm_get and autopm_put calls (they won't be able to change the usage counter directly). In particular, their total usage change from binding to unbinding must be exactly zero. I'm not sure the existing code in the drivers does this. The second has to do with the power/level attribute. Currently the framework doesn't define any runtime PM attributes, but it may well add its own level attribute eventually. When it does, there won't be anything corresponding to level = off. Should we start phasing this feature out now? (This is one reason for adding the "safely remove hardware" API; at the moment suspending a port is the closest userspace can come to disabling it.) Here's a third matter (not directly related to the new framework but connected with runtime PM): Should we add rudimentary autosuspend support back into usb-storage? By default it will remain disabled, of course. But there are many cases where it could be very helpful to users -- think of low-power systems with USB flash drives. Such users could enable it manually. 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
