Re: [PATCH 1/4] powerpc/fsl_soc.c: prepare for addition of mpc5121 USB code

[Grant Likely <grant.likely@secretlab.ca>]

On Thu, May 6, 2010 at 1:18 PM, Anatolij Gustschin <agust@denx.de> wrote:
> Hi Grant,
>
> On Tue, 27 Apr 2010 10:51:21 -0600
> Grant Likely <grant.likely@secretlab.ca> wrote:
>
>> On Tue, Apr 27, 2010 at 10:11 AM, Anatolij Gustschin <agust@denx.de> wro=
te:
>> > Factor out common code for registering a FSL EHCI platform
>> > device into new fsl_usb2_register_device() function. This
>> > is done to avoid code duplication while adding code for
>> > instantiating of MPC5121 dual role USB platform devices.
>> > Then, the subsequent patch can use
>> > for_each_compatible_node(np, NULL, "fsl,mpc5121-usb2-dr") {
>> > =A0 =A0 =A0 =A0...
>> > =A0 =A0 =A0 =A0fsl_usb2_register_device();
>> > }
>> >
>> > Signed-off-by: Anatolij Gustschin <agust@denx.de>
>> > Cc: Kumar Gala <galak@kernel.crashing.org>
>> > Cc: Grant Likely <grant.likely@secretlab.ca>
>> > ---
>> > =A0arch/powerpc/sysdev/fsl_soc.c | =A0231 +++++++++++++++++++---------=
------------
>>
>> Hi Anatolij,
>>
>> Thanks for this work. =A0However, I've got concerns.
>>
>> Forgive me for ragging on code that you didn't write, but this
>> fsl_soc.c code for registering the USB device really doesn't belong
>> here anymore. =A0It should be part of the drivers/usb/host/ehci-fsl.c
>> and the driver should do of-style binding (Which should be a lot
>> easier if I manage to get the merge of platform bus and of_platform
>> bus into 2.6.35).
>
> Now I moved the USB devices registration code to
> drivers/usb/host/ehci-fsl.c and added of-style binding there. It
> works for one platform driver based on your test-devicetree branch.
> It seems I can't bind more than one driver to the device described
> in the tree. But I need to bind at least 2 drivers, ehci-hcd and
> fsl-usb2-udc. For USB OTG support I need additional one to be bound
> to the same USB dual role device (also tree different drivers
> simultaneously).
> I also can't register UDC device in the ehci-fsl.c since registering
> of the UDC device should also be possible independent of the EHCI-HDC
> driver (for USB device support we do not need host controller driver).
> Is it possible to bind more than one driver to the same device
> simultaneously? If so, how can I implement this?

No, the linux driver model can bind exactly one driver to a struct
device.  However, that doesn't mean you can't have multiple struct
devices (in whatever form they come) to tell the Linux kernel about
all the details of a single hardware device.

>> This patch series makes the fsl_soc.c code even more complicated, and
>> scatters what is essentially driver code over even more places in the
>> arch/powerpc tree. =A0I'm really not keen on it being merged in this
>> form.
>
> Now I see one reason why it has been originally implemented
> this way.

Should be a solvable problem.  The fsl_soc.c stuff was originally
written simply because the infrastructure wasn't there for doing it
any other way.  We're long past that point now.  I don't have time to
dig into the details now (merge window and all), but ping me in a few
weeks and I'll take another look to see if I can help you.

g.