From mboxrd@z Thu Jan  1 00:00:00 1970
From: sylvain.rochet@finsecur.com (Sylvain Rochet)
Date: Sat, 7 Feb 2015 20:37:23 +0100
Subject: [PATCHv6 4/5] USB: gadget: atmel_usba_udc: Prepare for IRQ
 single edge support
In-Reply-To: <54D3A63B.6040209@atmel.com>
References: <1421945805-31129-1-git-send-email-sylvain.rochet@finsecur.com>
 <1421945805-31129-5-git-send-email-sylvain.rochet@finsecur.com>
 <20150122181422.1ba305f1@bbrezillon> <54D3A63B.6040209@atmel.com>
Message-ID: <20150207193723.GA32213@gradator.net>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

Hello Nicolas,


On Thu, Feb 05, 2015 at 06:19:55PM +0100, Nicolas Ferre wrote:
> Le 22/01/2015 18:14, Boris Brezillon a ?crit :
> > On Thu, 22 Jan 2015 17:56:44 +0100
> > Sylvain Rochet <sylvain.rochet@finsecur.com> wrote:
> > 
> > > -static const struct usba_udc_errata at91sam9g45_errata = {
> > > +static const struct usba_udc_caps at91sam9g45_caps = {
> > >  	.pulse_bias = at91sam9g45_pulse_bias,
> > > +	.irq_single_edge_support = true,
> 
> Nope! at91sam9g45 doesn't have this property. You'll have to create
> another compatible string and capabilities structure
> ("atmel,at91sam9x5-udc")

Oops.


> But still, I don't know if it's the proper approach. The possibility to
> trigger an IRQ on both edges or a single edge is a capacity of the gpio
> controller, not the USBA IP. So, it's a little bit strange to have this
> capability here.

I agree.


> I don't know if it's actually feasible but trying to configure the IRQ
> on a single edge, testing if it's accepted by the GPIO irq controller
> and if not, falling back to a "both edge" pattern. Doesn't it look like
> a way to workaround this issue nicely? Can you give it a try?

Tried, it works, but it displays the following message from 
__irq_set_trigger() [1] during devm_request_threaded_irq(?, 
IRQF_TRIGGER_RISING, ?) on boards which does not support single-edge 
IRQ:

genirq: Setting trigger mode 1 for irq 176 failed (gpio_irq_type+0x0/0x34)

Is it acceptable ?
If not, is udc->caps->irq_single_edge_support boolean acceptable ?
If not, I am ok to drop the feature, this is only a bonus.


Sylvain


[1] http://lxr.free-electrons.com/source/kernel/irq/manage.c#L619