Re: [linux-usb-devel] Re: [linux-pm] potential pitfall? changing configuration while PC in hibernate (fwd)

[David Brownell <david-b@pacbell.net>]

On Wednesday 25 May 2005 4:47 pm, Andrew Morton wrote:
> Michal Semler <cijoml@volny.cz> wrote:
> >
> > It is real stupid, that Linux kernel freezes when simply during hibernate I 
> > change Bluetooth dongle to USB mouse in my USB port.

That is, you unplug the dongle -- breaking the USB power session -- and then
plug in a mouse?  And the kernel does something other than disconnecting the
dongle, and then enumerate a new mouse?  That'd be a bug.  Somewhere.  It'd
probably help a lot if the bluetooth USB driver had a suspend() method.

It's not clear to me what you mean by "hibernate".  If that's an "everything
powered off" pseudo-suspend state, then the dongle should always be getting
disconnected, and USB always re-enumerated by Linux.  Only "real suspend"
states, like suspend-to-RAM, can normally maintain USB power sessions.  And
even then, the device drivers need to be prepared to maintain them ... and
do things like stop active I/O requests, and resume them later.

You said this was an all-Intel system.  So which host controller driver is
in use here ... UHCI?  EHCI?  With RC5 and CONFIG_USB_DEBUG, what do the
USB debug messages say is happening?  Does CONFIG_USB_SUSPEND change things?
Try this with purely modular USB for now.

- Dave