> Ok, we probably want to free firmware image from memory.... Or whatever else a driver may want to do... There are other reasons a driver may rely on userland, or may want to do GFP_KERNEL blocking allocations. Remember that those will possibly block forever too once the swap device is frozen. > Well, I could define two more events, and possibly kill event = ON; do > you really thing so many drivers will care? A few will, I suspect DRI for example will in the long run since it may have to allocate a significant amount of memory. APM emulation is more like a subsystem, so it would rather hook at a different level. We probably want to provide a global function saying something like suspend_in_progress() too, just in case a driver want to WARN/BUG_ON, or actually test if it can do something without having to track the pre-suspend and post-resume events. Ben.
