Hi Rafael, > Speaking of btusb, I'm obseving a reproducible breakage of suspend and > hibernation (for hibernation it's 100% reproducible, for suspend to > RAM about > 50%) on my hp nx6325 with the current -git kernel. > > The system crashes as soon as user space is thawed after a resume > and it's > sufficient to switch the bluetooth subsystem off to avoid this. > After such a > crash I'm usually unable to get any information from the box (from > the stack > traces I've seen once after such a crash it looks like interrupt > handlers are > involved somehow). > > Since the bluetooth adapter in this box is apparently handled by > btusb, I'm > suspecting the problem may be related to it or to the USB subsystem > (ehci-hcd > in particular). I'm going to debug it further, but if you have any > ideas to > try, I'll appreciate any help. the current version has no suspend/resume callbacks. However in that case the USB subystem should take of it, but I might be wrong here. So in case of btusb and not using it, we only have interrupt URBs in fly. However these might get screwed up. So does unloading the driver before suspend fixes it? What about the latest patch from Oliver. It will add the suspend/resume callbacks that are needed. Oliver, can you come up with a small test patch, that just kills all URB when suspend and submits the interrupt ones at resume. Such a patch might have to be merged for 2.6.27 if it fixes this problem. Regards Marcel _______________________________________________ linux-pm mailing list linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/linux-pm
