On Fri, 2005-03-11 at 11:42 -0500, Alan Stern wrote: > On Fri, 11 Mar 2005, Pavel Machek wrote: > > > > This is why I'm in favor of power domains (power parents and children > > > that can be independent of physical device relations). Power domains > > > are common for internal devices, not just external devices. Still, if > > > you disable a physical parent, then the device is no longer operational. > > > Disable the usb host controller, and the external disk drive cannot be > > > used (in fact data may be lost). So it's necessary to stop every child > > > device first, and generally also power it off if we can control the > > > power source. > > > > I just don't see it like that. > > > > Remember old scsi disks where formatting took half an hour? You may do > > something like > > > > invoke scsi format operation > > power down usb controller > > <half an hour later> > > power up usb controller > > get the results > > > > Now, that was probably bad example, but I do think some devices might > > do usefull work even with their parents disabled. Yes, but at a logical (class) level the parent device would never be disabled, right? I guess it would be possible to model such a situation if we take advantage of the driver class design. I hadn't considered it before, but I'll try to keep it in mind. It would be interesting to see how common it might be. > > There's another aspect to it as well. You can suspend the USB host > controller and then _still_ use the external disk, if the host controller > is automatically resumed when the next disk I/O request arrives. Whereas > if you would stop the disk and its driver, the next I/O request might > simply languish in a plugged queue. I suppose this would depend on > whether the disk driver was smart enough to leave its queue unplugged > under the right circumstances, so perhaps this isn't a great example. > (And you're right about the possibility of data being lost -- although > it's pretty unlikely -- which makes this an even worse example.) > > Alan Stern Yes, so I think if we ensure that the parent device is enabled at a logical level, and that the power source is active, it would be possible to do such a thing. I'm guessing this is something not to worry about too much now; do you think so? Thanks, Adam
