On Fri, 20 Jul 2007, Steven M. Bellovin wrote:
>> With my current patches, I whine and complain at boot if a driver
>> doesn't support power management. Once I get a bit further, I plan on
>> making the "put system to sleep" path depend on every single driver
>> instance supporting it or the request to go to sleep will be denied.
>
> *Please* print a message about such drivers...
>
> Better yet -- that's the sort of thing that can be detected
> quasi-statically.  At boot time, try to send a no-op message to all
> such hooks, so people are informed up front that they can't suspend the
> machine, and which driver is at fault.  (I've long had a habit of
> running near-GENERIC kernels -- that way, I'll know what to exclude...)
>
> I say "quasi" because of dynamically loaded drivers, but those can be
> probed at attach time.

You will currently see one or two messages from drivers at attach time 
(this includes devices plugged in after the system boots):

  1. Driver does not implement powerhooks or devpm handler:

     isa0: WARNING: power management not supported

  2. Driver implements powerhooks, not yet converted to devpm:

     wpi0: WARNING: powerhook_establish is deprecated; please see devpm(4)
     wpi0: WARNING: power management not supported

Is this acceptable?

Cheers,
Jared