On Sat, 2015-06-27 at 08:29 +0200, Jiri Kosina wrote:

On Fri, 26 Jun 2015, Greg Kroah-Hartman wrote:

Last time we were testing this, autosuspend for USB HID devices was quite 
a disaster.

Do you have any idea whether udev developers tested the "autosuspend on by 
default for USB HID devices" on reasonable set of devices?

The culrpits that I remember from top of my head (it's been long time 
ago):

- the LEDs for suspended device go off. This is very confusing at least on 
  keyboards, and brings really bad user experience

That is a bug. hidinput_count_leds() is supposed to prevent that.
What did you test?

- many keyboards were losing first keystroke when waking up from suspend. 
  We've been debugging this with Alan, and never root-caused it to a 
  problem in our code, it seems to be the property of the HW

- mice don't wake up from movement alone.

And again I would state that we don't get enough information
from user space. Hardware seems to be designed for sleeping
while a screen saver is on. In kernel space we just get a binary
input desired or not desired from open/close.
That is insufficient.

	Regards
		Oliver