On Monday 30 May 2005 8:27 am, Alan Stern wrote: > On Sat, 28 May 2005, David Brownell wrote: > > > > But don't you think we should still call hub_quiesce(), to turn off the > > > status URB? > > > > That would make good sense, yes. For root hubs, all that interacts > > with the timer processing too ... which runs orthogonally to the > > resume and khubd tasks (see below). > > > > That suggests an interesting experiment: do the latest MM patches, > > which I think have your root hub timer fixes, change this behavior? > > Either with or without USB_SUSPEND enabled. > > It's doubtful whether those timer changes will make any difference. Last > I looked the EHCI and OHCI drivers were still using the "legacy" mode.
Nonetheless those changes do affect things in those areas, and some of those changes could matter. And given how many other things changed in those as-yet-not-mainstream patches, I still think it'd be interesting; it's not as if these aren't all interrelated! > There shouldn't be any trouble switching them to "polled" mode, and the > best thing would be to go to "interrupt" mode. The infrastructure, > including the new hub_irq_enable() callback, is all set; it has been in > Greg's kernels for some time. It's just waiting for you to make the > necessary changes to the HCDs. Note that if you decide to use "polled" > mode then the HCD is responsible for turning off the status timer in the > suspend routine and turning it back on in the resume routine. I'm not planning on touching that until after that hits the main kernel tree; I don't have much time to pick up the pieces if/when things break due to the "last patch breaks the camel's back" syndrome. :) > There are a few other things also awaiting your action or approval: the > remove-the-recursion-from-suspend patch and the root-hub-lost-power patch > (although there was still a little confusion about how EHCI should handle > that -- the code near the start of ehci_resume is confusing). Same thing. Though I do have a slightly updated version of my version of that recursion patch, now that the "detach_state" thing and its deadlock potential is gone, and that'd seem good to merge sooner rather than later. - Dave ------------------------------------------------------- This SF.Net email is sponsored by Yahoo. Introducing Yahoo! Search Developer Network - Create apps using Yahoo! Search APIs Find out how you can build Yahoo! directly into your own Applications - visit http://developer.yahoo.net/?fr=offad-ysdn-ostg-q22005 _______________________________________________ linux-usb-devel@lists.sourceforge.net To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
