On Thursday 28 October 2004 14:00, Pavel Machek wrote:
> >
> > Currently that state comes in two parts:
> >
> > - Hardware ... look at registers, see if "reset" path is right
> > - Software ... maybe registers aren't sufficient for this hw
> >
> > If software state was needed, it had to have been set up
> > earlier before that system sleep transition. I still don't
> > see a need for more resume() state.
>
> You are suggesting
>
> foo_resume(void)
foo_resume(device) ... not foo->resume(void) C++ style!
> {
> if (!ask_hardware_if_it_is_initialized())
> initialize_hw();
Not exactly; if BIOS initialized it, we have to reinit.
It has to be in a state compatible with how Linux left
the device.
> }
>
> while I see some value in
>
> foo_resume(power_t state)
foo_resume(device, somekinda_power_state)
> {
> if (STATE_LOST(state))
> initialize_hw();
That seems like the "software" case I gave, in
which case device->driver_data can hold the answer
to STATE_LOST. It'll have known it in advance,
maybe even based on board-specific logic.
> }
>
> It is only speed optimalization, but I undestand why someone would
> like to do it.
I'm just pushing back on core API changes, to
make sure I understand the anticipated benefit.
They all have a drawback: lots of driver changes!
- Dave
Pavel
> --
> People were complaining that M$ turns users into beta-testers...
> ...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!
>
