On Wed, 1 Aug 2007 16:27:48 -0400 (EDT), Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote: > In fact, I would go even farther. Let's document the ordering given by > dpm_active (by default, in order of registration) so that drivers can > depend on it, and let's add special routines to change the order for > the few cases that require it -- under driver control. Driver authors must definitely be made aware of this, and they can only do something if they are given some control. Implicit assumptions mean waiting for breakage to happen. > > A "move to the end" routine would solve the EHCI issue. It ought to > solve the problems associated with device_move() also, provided the > device being moved doesn't have any children. The device drivers need to be given more control; they could just walk the children list in this case then. > > Cornelia and Marcel, is that the case? When a ccw_device or RFCOMM TTY > is reparented under a new device, do any children get moved along with > it? Hm, device_move() just drags around the children with it (and ccw_devices may have children as added by the driver; zfcp does that, for example). _______________________________________________ linux-pm mailing list linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/linux-pm
