From mboxrd@z Thu Jan  1 00:00:00 1970
From: aisheng.dong@freescale.com (Dong Aisheng)
Date: Tue, 19 Jun 2012 21:01:46 +0800
Subject: [PATCH 16/16] ARM: imx: enable SPARSE_IRQ for imx platform
In-Reply-To: <20120619074258.GF21951@S2101-09.ap.freescale.net>
References: <1339653587-4832-1-git-send-email-shawn.guo@linaro.org>
 <1339653587-4832-17-git-send-email-shawn.guo@linaro.org>
 <20120618084850.GG25447@shlinux2.ap.freescale.net>
 <20120618150422.GG19249@S2101-09.ap.freescale.net>
 <20120619061637.GB10387@shlinux2.ap.freescale.net>
 <20120619064742.GE21951@S2101-09.ap.freescale.net>
 <20120619072110.GC10387@shlinux2.ap.freescale.net>
 <20120619074258.GF21951@S2101-09.ap.freescale.net>
Message-ID: <20120619130145.GD10387@shlinux2.ap.freescale.net>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Tue, Jun 19, 2012 at 03:43:00PM +0800, Shawn Guo wrote:
> On Tue, Jun 19, 2012 at 03:21:11PM +0800, Dong Aisheng wrote:
> > On Tue, Jun 19, 2012 at 02:47:44PM +0800, Shawn Guo wrote:
> > > On Tue, Jun 19, 2012 at 02:16:38PM +0800, Dong Aisheng wrote:
> > > > > > Ideally i would see we keep the code as before that still using hw irqs
> > > > > > for device resource definition, but convert to linux virt irq in a standard
> > > > > > irqdomain map way when adding devices by calling imx_add_platform_device.
> > > > > > 
> > > > > What's the point of hiding this irq number conversion in
> > > > > imx_add_platform_device?  
> > > > It's standard and safe way.
> > > > 
> > > Why do you think it's standard?
> > Why do you think using irq_find_mapping to get the linux virt irq is not
> > standard way?
> > 
> I never said using irq_find_mapping to get the linux irq is not
> standard way.  What I said is using it during adding platform_device
> is nothing standard.
> 
> > > Care to show me a couple of examples
> > > that make this conversion when adding platform device?
> > > 
> > I did not search any example, currently it's just my idea based on my understanding
> > on irq domain design. Maybe i should give a patch to describe my idea,
> > then we can discuss on the patch.
> > 
> I understand your idea, and I just do not think it's a good/right one.
> 
Can you tell the reasons?

> > Or did you see any example on using shift way to define device irq resource?
> > Then i can look at it.
> > 
> http://thread.gmane.org/gmane.linux.ports.arm.kernel/171075
> 
Hmm, i didn't know the history
Do you know the reason why we decide to do that way?

> > > > > The irq number used in resource definition
> > > > > should simply just be Linux irq.  Doing what you suggest here will
> > > > No, it's should be hw irq.
> > > 
> > > You are simply wrong here.  Nothing more to respond on this.
> > >
> > If using linux virt irq, how do we avoid the issues i said in my last reply?
> > 
> Do not make up any issue.  I'm only interesting in the practical issues.
> 
Aren't they really exist potential issues?
Shouldn't we take care of it?

I'm afraid the solution you adopted is a bit weak and we may need change it again and again
in the future. The irq number defined definitely should be stable and not easily change
since they're hw properties.

I made a patch for the method i prefered, will send out in this thread for discuss.

Regards
Dong Aisheng