Re: Wake on USB

[Dale P. <DEP...@ed...>]

On 03/28/11 12:53, Paul Bender wrote:
> On 3/27/2011 6:03 PM, Dale Pontius wrote:
>> On 03/25/11 18:41, Jarod Wilson wrote:
>> <snip>
>>> The main goal behind this release is compatibility with all the
>>> lirc code that has been merged into the Linux kernel as of
>>> around 2.6.36 and on. I've also removed several drivers that I
>>> consider completely obsolete and will never update ever again,
>>> as they've been superseded by upstream kernel drivers. There will
>>> be more chopped out of 0.9.1 as well. :)
>>
>> Perhaps this is a good place/opportunity to ask this question.  I'm
>> using an mceusb IR receiver/remote on a dedicated mythfrontend machine.
>>   To make it more appliance-like with faster powerup I suspend instead of
>> powering off.
>>
>> I haven't been able to wake-on-USB with any kernel after the 2.6.32
>> series.  I'm currently running 2.6.32-gentoo-r27, and it and every
>> kernel before that properly does wake-on-USB.  No kernel beginning with
>> 2.6.33 does, and that includes testing 2.6.38 this afternoon.  It also
>> includes both Gentoo and vanilla kernels.
>>
>> I ask here because you're about to remove the lirc drivers that I'm
>> using, and I haven't been able to get myself up to 2.6.36 yet, though
>> I'd really like to.  I recognize that this issue is not core to lirc,
>> though in the lirc community people may be more aware of this kind of
>> problem and how to fix it.
>>
>> Everything in userspace remains unchanged between the various kernels,
>> and when I build 2.6.33+ kernels I start with my 2.6.32 config and tweak
>> the new stuff, but the old stuff is still essentially there.
>>
>> One note...  My mceusb IR receiver has an LED that blinks when it
>> receives a signal.  When the system is suspended and I press the power
>> button, that LED comes on solid until the system has wakened.  When
>> suspended with a 2.6.33+ kernel the LED doesn't come on, doesn't even
>> blink when the power button is pressed.  I can wake the system with the
>> front-panel power button, in any case.
> 
> You need to be sure that the power is provided to the device. In the 
> past, this was done through /proc/acpi/wakeup. However, with more recent 
> kernels, this is done through /sys/devices/*/wakeup. I use the following 
> script to enable wakeup when the remote is detected by udev:
> 
> #!/bin/sh
> ################################################################################
> # wakeup_enable
> ################################################################################
> 
> devpath=`readlink -f "/sys/${DEVPATH}"`
> while echo "${devpath}" | /bin/grep -q '^/sys/devices/' > /dev/null 2>&1 
> ; do
>      if test -e "${devpath}/power/wakeup" ; then
>          state=`cat "${devpath}/power/wakeup"`
>          if test "x${state}" = "xdisabled" ; then
>              echo "enabled" > "${devpath}/power/wakeup"
>          fi
>      fi
>      devpath=`dirname "${devpath}"`
> done
> 
> exit 0
> 
This didn't make it into last night's post here, but I did go looking
through /sys/devices/pci/pciXXXXX/power/wakeup and found that it did
indeed appear to be enabled.  Adding to search terms a bit I found that
this has been discussed, with almost identical symptoms to mine, here:
http://comments.gmane.org/gmane.linux.usb.general/32115
Unfortunately I didn't find any resolution on the thread, and it was
some time back.

Leaves me wondering 2 things...
1 - How does your udev script know what device it's enabling wakeup for?
2 - The USB list seemed to suggest "fuller enablement".  Do I need to go
into /sys/bus/usb/****/wakeup, as well?  My mythbox is off/suspended and
the family watching TV at the moment, so it's not convenient to look.
They also talked about "kernel drivers" being involved.  Don't the LIRC
drivers count on this?

I'm in a bit of a Catch-22 here.  I can't move past 2.6.32 until I can
get the silly thing to wake-on-USB, and I can't go with in-kernel
drivers until I get past 2.6.32.  Switching to in-kernel drivers calls
for rebuilding LIRC with different options, so it's no longer boot back
and forth, but boot+rebuild LIRC for any experimentation.

Dale Pontius