Re: [RFC] filling in ACPI files in sysfs

[Alex Williamson <alex.williamson-VXdhtT5mjnY@public.gmane.org>]

On Thu, 2004-04-08 at 02:14, Paul Ionescu wrote:
> Hi Alex,
> 
> I attached a file containing what I have in
> /sys/firmware/acpi with your second patch.

   Thanks, looks like your namespace is more complete than on my old
omnibook.

> I also tried the docking station with:
> echo 1  >
> /sys/firmware/acpi/namespace/ACPI/_SB/PCI0/PCI1/DOCK/_DCK
> 
> and 
> 
> echo 0 >
> /sys/firmware/acpi/namespace/ACPI/_SB/PCI0/PCI1/DOCK/_DCK
> 
> with this commands I semi turned on/off the doking
> station. I say semi because is not fully on/off.
> When I turn it on (echo 1 > _DCK), the docking station
> fan starts, and the LED indicating the usage of
> docking station turns on, but I dont see the new
> devices from the docking station (2 new usb, 2 new
> cardbus, 1 new ide, 1 new pci etc). I suspect this is
> not your patch fault, but the way ACPI handles this
> event. It should add the new description tables of
> ACPI devices on the fly not only at boot time. When it
> boots without docking station ACPI has no way to know
> what will be in my docking station, so it has to add
> new devices after inserting and activating the
> docking.

   Yes, as far as I'm aware, we don't try to re-parse acpi namespace
after such events.  I'm not trying to solve that problem though.  This
is also just a raw interface to the acpi methods, so undocking w/o
handling the things that will get removed may be unsafe.  I envision it
being an interface by which hotplug could evaluate, via the _EJD
methods, what devices will go away and do smart things prior to the
actual eject.  On my docking station, there's no obvious effect of
calling the _DCK method, but calling _EJ0 pops it out.  Also, I see all
3 possible batteries under /proc/acpi/battery (even though only 1 is
plugged in), so I doubt this patch is changing that behavior at all.

> Now, when I turn it off (with echo 0 > _DCK), the fan
> stops, but the ACTIVE led is still on, and the
> electromechanical eject was not activated.
> 
> Is this the way I was supposed to test docking ?


   Try the _EJ0 method.  Assuming all device that you'll lose are
quiesced and removed (probably everything that lists _SB.PCI0.PCI1.DOCK
in their _EJD method), I believe you need to do the following sequence:

echo 0 > _DCK
echo 1 > _EJ0 

And after you re-dock, you should probably echo 1 > _DCK.  Thanks,

	Alex

-- 
Alex Williamson                             HP Linux & Open Source Lab



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click