On Po 14-03-05 10:59:25, David Brownell wrote: > On Monday 14 March 2005 10:44 am, Pavel Machek wrote: > > Hi! > > > > > > I'll really need to make pci_choose_state > > > > "NOP" in the current code, so that it can be safely added to the > > > > existing drivers. > > > > > > There's no point in adding NOPs. Why bother? > > > > > > Remember, the original idea behind pci_choose_state() was that > > > PCI drivers for PM-aware devices could use that instead of their > > > own dumb mappings between system suspend states (S0, S1, S2, S3, > > > S4, etc) and PCI states ... which in too many cases was just to > > > re-use the same numeric value. That's _never_ going to be a NOP, > > > unless Linux only ever supports PCI D3hot (ACPI D2) and S3; and > > > likewise will never be needed for PM-stupid devices, or for those > > > drivers with actual intelligence (e.g. Ben's video examples). > > > > When I added pci_choose_state to drivers, I commented it as "no code > > changes". I was wrong. > > That's sort of what I said. If it's a NOP, why bother? The > idea behind such an API was to support S1/S2/S3 suspend levels, > where D3hot may be sub-optimal. That's not NOP-ville. And it > probably ought to consult ACPI tables on some systems... Because pm_message_t is not compatible with pci_power_t, you need something to convert. And that something is pci_choose_state. If more magic is needed, we'll need new function. > > pci_choose_state() needs to return d3hot even > > for pmsg_freeze, because that's what old code did, and I did not audit > > all the drivers. > > Seems like a problem to me. Do S1/S2/S3 transitions even > need a "freeze" transition? I thought we'd agreed they don't; That's not the problem. Old code put devices into D3hot in swsusp "freeze" case. We'll need to do the same now, slowly auditing the drivers and removing that unneccessary D0->D3hot->D0 transition. Pavel -- People were complaining that M$ turns users into beta-testers... ...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!
