>-----Original Message----- >From: Alan Stern [mailto:stern@xxxxxxxxxxxxxxxxxxx] >> >> This isn't *really* about saving power in the individual device; it's >> more about stopping the device generating events that disrupt the rest >> of the system. Suspending the device can be one way of doing that and >> is useful if it can be done but is not really the immediate goal here. > >Runtime PM is _not_ a reliable way of preventing a device from >generating events. It is meant for saving energy, nothing else. > >If that's what this is about, then the answer is simple: Don't use >runtime PM to try to suppress events. > I think it is unfair to discriminate here about saving power and turning off events. I believe putting the device into power save leads to, if device permits preventing generating events is valid to hook up run time PM. >-----Original Message----- >From: Alan Stern [mailto:stern@xxxxxxxxxxxxxxxxxxx] > >I still disagree. The user should be able to tell the driver that he's >not going to be using the keypad, touch panel, or whatever in the near >future... but it's up the driver to decide whether or not to go to low >power. Why cannot the driver decide to return failure in such a case? But why stop the user space from requesting a power save mode? >> Hmm. I was looking at the case with touch screen. I set my display backlight >> to go off after say 30 seconds; my touch screen idle begins only when the >display >> goes off; meaning I still expect the touch screen to respond to any events before >> the timeout. This in turn requires that the timeouts for the backlight and the touch >> be synchronized, failing which the user might interpret the touch to be non- >functional >> in case the touch timeout and backlight timeout vary. Am I wrong to think that >> such inter-dependency between devices should be avoided? > >Although such dependencies do sometimes make things awkward, >occasionally you need to have them. Hmmm...I am not convinced that such sync'ed timeouts between devices are necessary; I still prefer it would be better if a single user space monitor could make sure that all listed devices are communicated for a power save rather than introduce dependencies between a touch and the display device. I was thinking about an another case too where device autosuspend could create issues: Imagine a slider cover for the camera which enables the camera and the live playback on a screen; In such a case you have now not 2, but 3 components to sync up: the auto idle for touch, auto idle for camera; auto idle for the display. Cheers! _______________________________________________ linux-pm mailing list linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/linux-pm
