On Mon, 15 Oct 2012 12:32:05 +1000 Dave Airlie <airlied at gmail.com> wrote: > > The console lock can be contended, so rather than prevent other drivers > > after us from being held up, queue the console suspend into the global > > work queue that can happen anytime. I've measured this to take around > > 200ms on my T420. Combined with the ring freq/turbo change, we should > > save almost 1/2 a second on resume. > > Did you measure booting with quiet? I can't think what else could be > contending on this lock other than printk, now maybe we should reduce > the printk noise! > > This will make debugging things even worse than they are, since you > don't see any printks on the screen until after we do this. I thought I measured with quiet, but I'll double check. Not sure about debugging; it's pretty bad to begin with. I generally use netconsole or the "stare really hard at the code" method when it comes to suspend/resume or init time anyway. But yeah, some investigation into the console lock contention is probably a good idea at any rate. A typical distro config probably also has some other fbcon driver and probably vgacon built-in too, and I haven't checked what they do. Jesse
