Re: [Xenomai] unsupported ARM9 machine

[Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>]

On 06/22/2012 03:07 PM, Marcin Kuśka wrote:
> 2012/6/13 Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>
> 
>>
>> For instance, the fact that you are compiling both
>> arch/arm/common/timer-sp.c and smp_twd.c will lead to a conflict the way
>> timer-sp.c is currently made. omap4 has the same problem since for
>> instance arch/arm/mach-omap2/timer.c is compiled both in the omap3 and
>> the omap4 case. So, just look at how it is done, and do it the same way.
>>
>> Another simpler solution is to start from the ipipe core-3.2 branch,
>> which supports being built for several SOCs. However, it will require
>> upgrading to xenomai 2.6 branch head.
>>
>> --
>>                                                                Gilles.
>>
>>
> Ok I checked the arch/arm/common/timer-sp.c (i assume this file implements
> the timer for vexpress) and all the__ipipe_mach_timerint;
> __ipipe_mach_timerstolen; and so on, are there and there is nothing to be
> done there. I don't know if i understand you right. Which part should I do
> the same as with omap4?

The problem is that if you are compiling with CONFIG_SMP for a cortex
a9, there is also an __ipipe_mach_timerint, __ipipe_mach_timerstolen,
etc... in the generic code for cortex a9. So, you get multiple definitions.

omap4 has exactly the same problem, because it compiles some code which
is common with omap3, already modified for CONFIG_IPIPE.

So what you have to do is solve the multiple definitions for vexpress
the way they are solved for the omap4/omap3.

> 
> 
> About that simpler solution using ipipe core-3.2 branch. There is no ipipe
> core-3.2 at the http://download.gna.org/adeos/patches/v3.x/arm/ site or am
> I looking for wrong site?

Yes, I am talking about a branch in the adeos ipipe git repository. The
patch is not released yet. But anyway, for what you have to do, working
with git is much simpler than working with patches.

-- 
					    Gilles.