[linux-pm] [patch 2.6.25-rc6 1/7] crosslink ACPI and "real" device nodes

[David Brownell]

On Friday 21 March 2008, Zhao Yakui wrote:
> On Fri, 2008-03-21 at 00:31 -0700, David Brownell wrote:
> > > > The PNP device gets a "firmware_node" link pointing to the ACPI device.
> > > > The ACPI device has a "physical_node" link pointing to the PNP device.
> > > > Linux drivers currently bind only to the "physical" device nodes.
> > >  
> > > Very good idea. 
> > > But maybe there is a lot of ACPI devices on the laptops. And we take a
> > > little care about the association between the acpi device and "real"
> > > device.
> > 
> > Are you suggesting that the ACPI nodes shouldn't exist at all?
> > Or that something is wrong with how they're set up or used?
> 
> No. The ACPI nodes should exist. What I said is whether it is necessary
> to create the link for all the ACPI devices between the ACPI device and
> "real" node device.  Of course it is also OK if link is created for the
> ACPI device with the ability to wake the sleeping system. 

The confusion exists for all ACPI device nodes that
mirror "real" device nodes (like PNP or PCI devices).

It's *not* limited to wake-capable devices.


> > For now, there's some confusion.  Devices listed in ACPI tables
> > have one or two extra sysfs device nodes.  I think *something*
> > should help sort out the confusion ...

(Minor correction:  *most* devices listed in ACPI tables have
the problem of extra sysfs nodes.  A few don't; like buttons.)


> Sorry. What I means is that the link(point to the ACPI device with the
> ability to wake the sleeping system) is created in the /sys/power/. 
> After doing so, we can easily check which device has the ability to wake
> the sleeping system. 

There's no need for symlinks to do that ... or ACPI.

I've been meaning to repost my script that scans sysfs
for the wakeup-capable devices ... I updated it a short
while ago to work right without "legacy" sysfs nodes.

- Dave