RE: [PATCH 2/5 v11] arm: omap: usb: ehci and ohci hwmod structures for omap3

[Partha Basak <p-basak2@ti.com>]

>-----Original Message-----
>From: Felipe Balbi [mailto:balbi@ti.com]
>Sent: Tuesday, September 27, 2011 6:55 PM
>To: Munegowda, Keshava
>Cc: t-kristo@ti.com; Paul Walmsley; Cousson, Benoit; Basak, Partha;
>Balbi, Felipe; parthab@india.ti.com; linux-usb@vger.kernel.org; linux-
>omap@vger.kernel.org; linux-kernel@vger.kernel.org; Gadiyar, Anand;
>sameo@linux.intel.com; tony@atomide.com; Hilman, Kevin;
>johnstul@us.ibm.com; Sripathy, Vishwanath
>Subject: Re: [PATCH 2/5 v11] arm: omap: usb: ehci and ohci hwmod
>structures for omap3
>
>Hi,
>
>On Tue, Sep 27, 2011 at 06:48:35PM +0530, Munegowda, Keshava wrote:
>> > So, you would need a mechanism to do something like this:
>> >
>> > pad a or b wakeup detected -> irq0
>> > pad c or d wakeup detected -> irq1?
>>
>> yes, if get something like this , its perfect.
>
>can't you have different IRQs for each pad ? I mean, allocate one
>irq_desc for each pad and let drivers request a pad/pin as an IRQ
>source. Then, when you detect a pad wakeup, you can:
>
>unsigned	pad_irq = pad_number - pad->irq_base;
>
>handle_nested_thread(pad_irq);
>
>this will make use of threaded IRQ handlers even. Could it be something
>like that ?

Felipe, your suggestion would mean more design change from the existing
implementation of Tero.

I would propose something like what Tero said initially:
For each mux-info have an associated irq handler.
So, say pads a..d form mux info1. This gets associated to irq_handler1.
Similarly, say pads e..h form mux info2. This gets associated to
irq_handler2.
Both get associated to the same uhh_hwmod. Now, when chain handler scans
for wakeup sources,
it scans both mux-info1 & mux-info2.
If at-least one pad in mux-info1 is woken up, irqhandler1 is called & same
for irqhandler2.
This mechanism would need multiple mux-infos to be attached to the same
hwmod.

So, fundamentally, if we are in alignment, can we go ahead now to collapse
the ehci & ohci hwmods into one?

>
>--
>balbi