On Tue, 2004-11-02 at 10:53 +1100, Benjamin Herrenschmidt wrote: > On Tue, 2004-11-02 at 00:54 +0100, Pavel Machek wrote: > > > I believe it would be nicer to SUSPEND, because apm bios may fail to put > > devices into powersave... > > Pavel > > PS: The mail with list of states was sent before I received this one... > > This is tricky... APM will eventually try to use some devices after we > do our suspend pass, for example, it will try to suspend the video chip > into "APM" mode (which in most case for ATIs is D2 state). The problem > is that if the device is already suspended, the BIOS may well crash > trying to tap it's registers... > > I'm not sure what is the proper policy with APM, but then, I have the > APM maintainer not far from my desk so ... :) Ok, I talked to him (Stephen Rothwell) and he confirms. APM wants us to freeze the drivers, but not change their power state (yah ! yet another use for freeze :) They tried in the early day of the PM tree to call that for APM but it broke... because the drivers were suspending the HW, which the BIOS doesn't like at all. That is typically an example where the pm_message would contain a state of PM_STATE_FREEZE, and the flags would additionally tell it's an APM suspend (for the 0.0001% of drivers that may actually care). Ben.