Re: btusb hibernation/suspend breakage in current -git

[Marcel Holtmann <marcel@holtmann.org>]

Hi Rafael,

> > > > > > Good. Can you test what happens if you unplug the device while suspended
> > > > > > and hibernated?
> > > > > 
> > > > > It's built-in, I can't unplug it. :-)
> > > > 
> > > > Maybe you can disable it in the BIOS, but this might change the DSDT / other
> > > > system configuration, so it might break resume in other ways :-(
> > > 
> > > There is a switch that's supposed to disable the radio (rfkill or something).
> > > I used it to switch the radio off while the box was waking up from hibernation
> > > and kbluetooth didn't find the adapter after the resume.  After I've pressed
> > > the "radio off" button again, the bluetooth appears to be functional again.
> > > 
> > > However, this "radio off" button is shared between bluetooth and wireless
> > > (b43) and there are some surprising interactions.  Nothing seems to be broken,
> > > though.
> > 
> > This doesn't explain the original failure. Can you comment out the support
> > for suspend/resume in the driver and try again?
> 
> With that commented out, I'm able to reproduce the failure.  With the original
> patch, I'm not.

I've never seen any issues with the suspend/resume and btusb, but I
must admit that I am using an X61 and in that case pm-utils has a magic
hack to disable Bluetooth before suspend and this means a clean
disconnect from the USB bus.

Anyway, killing all URBs in-fly on suspend and bringing up the interrupt
one on resume should do the right thing. However we have to check if we
not just better resume all URBs and let the Bluetooth core handle lost
connection during the suspended time.

Regards

Marcel