Bug 4180

There is an auto off kernel parameter which may help here:

usbcore.autosuspend=-1

Can you try that?

Stoicescu Cornel,

I am not seeing this issue with latest branches. Can you try with the latest branches and see if you see the issue?

I am still seeing this issue in:
dylan:
meta-intel:

Here is some useful output: 

root@sugarbay:/# tail /sys/bus/usb/devices/usb2/power/autosuspend
0
root@sugarbay:/# tail /sys/bus/usb/devices/usb2/power/autosuspend_delay_ms
0
root@sugarbay:/# tail /sys/bus/usb/devices/usb1/power/autosuspend_delay_ms
0
root@sugarbay:/# tail /sys/bus/usb/devices/usb1/power/autosuspend
0
root@sugarbay:/# tail /sys/module/usbcore/parameters/autosuspend
2

As per documentation: 
"
Each USB device directory will contain a file named power/autosuspend.  This file holds the time (in seconds) the device must be idle before it will be autosuspended. 0 means the device will be autosuspended as soon as possible.  Negative values will prevent the device from being autosuspended at all, and writing a negative value will resume the device if it is already suspended.
"

Doing this solves the problem: 
echo "-1" > /sys/module/usbcore/parameters/autosuspend

I forgot to add the commits:

dylan:87eb70243749e0e2431f98770746fa42002f4441
meta-intel:76f70c337e5daa9593153dfdf5107aec08610a44

So it sounds like the autosuspend is working as advertised.

The question then is what changed the settings and is it the same on other BSPs including core BSPs.

I will get this information during the current full pass.

This issue is present on sato-sdk-FRI2, lsb-sdk-jasperforest, sato-sdk-sugarbay

Also I noticed /sys/module/usbcore/parameters/autosuspend is set to 2 in both Ubuntu and Fedora and the issue is not present there; therefore that might not be the source of the problem.

The real issue here might be that way autosuspend is done. I think it should not miss inputs on behalf of autosuspend, thus not making it a problem.

All the images you mention have -sdk in common.  Is that a requirement, or just what you tested i.e. does it happen with just -sato?

With core-image-sato on crownbay, I don't see this problem, and the settings there are the same as in comment 3.

The target milestone says 1.5, but this was mentioned in yesterday's staff meeting as something that should be fixed for 1.4?

Has anyone tried the kernel parameter from comment 1? If that works, this can be easily documented for the release.

This appears to be a behavior preference as opposed to a bug, I don't see this as critical for 1.4. So unless RP or Saul would like to override that, I'd like to see us test the kernel param and document for 1.4.

I'm having USB problems with my sugarbay, so can't even reproduce the problem at this point.

I will test this on my sugarbay here with the kernel parameter: 

usbcore.autosuspend=-1

The USb device inactive issue does go away with the kernel parameter usbcore.autosuspend=-1 on sugarbay hardware

Fixed in linux-yocto-3.8 kernel repository:

commit 5995fa5838b843b5393e3a6cfdee8b508cb95ba8
Author: Nitin A Kamble <nitin.a.kamble@intel.com>
Date:   Fri Aug 9 00:57:33 2013 +0000

    meta: features/power/intel.cfg
    
    Disable USB autosuspend configuration. This is causing unusable
    keyboards & mice issues for many Intel BSPs.
    Also by default this config is disabled in the kernel.
    
    Fixes bug:
    [YOCTO #4992]
    
    Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
    Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>

VERIFIED in:
poky master:a02280f99f05b1c21de2a78de9222a2f62a0dabc