Hi, On Fri, Jun 17, 2011 at 09:44:52AM -0700, Sarah Sharp wrote: > One more question occurred to me. We currently have good runtime PM > support for PCI host controllers. If you add a platform device as > you've proposed, how does that effect the power management for the PCI > host controller? Will it still work as it does today? How will the it should still work the same way. PM is also done in "layers" (child device has to be suspended for parent to be suspended) which means that we can put all xhci specific pm functions in xhci.c and all pci-specific pm functions in xhci-pci.c. IOW, if we need to save context due to hibernation, xhci.c will save context only for the xhci register and xhci-pci.c will handle the PCI specific part. > non-PCI host controller's power management work? non-PCI will also use runtime PM. In fact, this will be one push for the other embedded platforms to move to runtime PM, which will help a lot on ARM consolidation effort. Basically, we will assume xhci.c is already taken care of, and on e.g. OMAP5 we will only care about the OMAP-specific "wrapper" around the xHCI IP (which is an OMAP-specific integration context). I hope this clears out. Let me know if it doesn't. English fails sometimes ;-) -- balbi
Attachment:
signature.asc
Description: Digital signature
