On 01/22/2012 10:39 PM, Mark Brown wrote: > On Sun, Jan 22, 2012 at 06:48:36PM +0100, Sylwester Nawrocki wrote: >> On 01/22/2012 06:27 PM, Bill Gatliff wrote: > >>> I for one would rather see in-kernel drivers that require it, and then > >> In fact we have to deal with the opposite now, as some existing drivers >> have been used for multiple generations of SoC, where almost unchanged >> device IPs are deployed. Those drivers were originally written for the >> simplest SoCs. > > TBH I think most of the devices for which people are running these days > will be able to get some win from the system wide stuff - the WFI modes > aren't exactly the latest thing in hardware terms, it's just been a long > road to getting them supported. Infrastructure like Mark's PM QoS work > and Raphael's PM domains work has really helped a lot here. Indeed, now that I checked some earliest SoC TRMs even the simplest systems have some sort of intermediate Idle states. >>> patches in mailing lists that show how to un-require it. Make it more >>> painful to avoid PM_RUNTIME, and less painful to use it. > >> Yeah, makes a lot of sense. With new code there is no issue, only the code >> with long history is sort of troublesome. > > It's mostly a transition management issue I think. When I repost I'll > add an additional patch on top which moves the clock gating into the > runtime PM callbacks, that way the decision on that doesn't block the > system wide work. Sounds like a great resolution to the problem, thanks for considering it. And sorry for disturbing. Let's see what's Kukjin's opinion on that. -- To unsubscribe from this list: send the line "unsubscribe linux-i2c" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
