Matthew Garrett wrote: > On Fri, Jun 26, 2009 at 09:23:12AM -0700, Kok, Auke wrote: >> Alberto Gonzalez wrote: >>> Thanks for the link. Interestingly Matthew Garrett says there that powertop >>> is >>> actually doing it in the wrong way (USB autosuspend), so maybe this is a >>> bug >>> in powertop (bug in the sense of using a wrong/old implementation)? >> Matthew knows where to find me and how to write patches, so I assume he's got >> something coming for me then... > > To be honest, most of my work in this field has been trying to ensure > that we can just remove all the hints from powertop :) It used to be > that powertop just read the autosuspend file to check whether > autosuspend was enabled - that seems to have been fixed now, but it > still suggests booting with "usbcore.autosuspend=1" to enable > autosuspend. That only changes the default delay, it doesn't actually > enable it. There's no general kernel option to enable autosuspend, > mostly because whether you want it or not is a somewhat difficult > question. As the original poster noted, enabling autosuspend on input > devices will generally result in unhappiness - many keyboards will drop > the first keystroke, and mice will only wake up in response to a button > press and not movement. > >> That's the file that powertop touches when you press 'U'. powertop also sets >> power/level to "auto", meaning the kernel can suspend/resume the device when >> it >> wants. >> >> if someone has patches on how to improve this, I'll gladly take them... > > I'd actually suggest that powertop stop making this suggestion. The > kernel and userspace will be updated as hardware is verified, but > there's no real way powertop itself can decide whether it's safe to > enable USB autosuspend or not.
are you saying that we need to just remove the parts which write to power/autosuspend only? or do we currently (new kernels etc) want to stop setting power/level to "auto" as well? This still leaves the kernel in charge. I'll definately remove the suggestion for the kernel commandline. Auke > _______________________________________________ Power mailing list [email protected] http://www.bughost.org/mailman/listinfo/power
