But the kernel patch didnt help much. I am quite sure that CPU was sleeping, but there can be other HW which has to be put to idle mode (USB controller?).

Having good kernel sources would might help here - those sources on github are really crippled because of missing commit history. Is it some legal problem to publish commit history?

I aggree that having this working would be cool - we have nice HW, but it's quite useless for now...