[Xenomai] Any idea how to fix ipipe failing to compile with ARM LPAE enabled?

[Xenomai] Any idea how to fix ipipe failing to compile with ARM LPAE enabled?

[Lennart Sorensen]

Any idea how to fix ipipe to not cause a linking error when LPAE is enabled on arm?

I see the function __ipipe_pin_range_globally is inside an ifdef block that says essentially:

#if !IPIPE
define __ipipe_pin_range_globally as do nothing
#else
#if !LPAE
define __ipipe_pin_range_globally function
#endif
#endif

So if IPIPE and LPAE are both true, then the function is never defined.

This is in arch/arm/mm/fault.c

-- 
Len Sorensen

Re: [Xenomai] Any idea how to fix ipipe failing to compile with ARM LPAE enabled?

[Gilles Chanteperdrix]

On 08/22/2014 11:31 PM, Lennart Sorensen wrote:
> Any idea how to fix ipipe to not cause a linking error when LPAE is enabled on arm?
> 
> I see the function __ipipe_pin_range_globally is inside an ifdef block that says essentially:
> 
> #if !IPIPE
> define __ipipe_pin_range_globally as do nothing
> #else
> #if !LPAE
> define __ipipe_pin_range_globally function
> #endif
> #endif
> 
> So if IPIPE and LPAE are both true, then the function is never defined.
> 
> This is in arch/arm/mm/fault.c
> 
I do not believe this is intentional. That being said, the Linux code in
the #ifdef is not reassuring: exception handlers are not registered when
the kernel is compiled with LPAE, which could mean that exceptions are
used for LPAE when compiled with LPAE. If this is the case, then LPAE
should not be enabled, otherwise your applications will not stay in
primary mode when these faults happen. How much RAM do you have, can you
really test LPAE ?

-- 
                                                                Gilles.

Re: [Xenomai] Any idea how to fix ipipe failing to compile with ARM LPAE enabled?

[Gilles Chanteperdrix]

On 08/22/2014 11:36 PM, Gilles Chanteperdrix wrote:
> On 08/22/2014 11:31 PM, Lennart Sorensen wrote:
>> Any idea how to fix ipipe to not cause a linking error when LPAE is enabled on arm?
>>
>> I see the function __ipipe_pin_range_globally is inside an ifdef block that says essentially:
>>
>> #if !IPIPE
>> define __ipipe_pin_range_globally as do nothing
>> #else
>> #if !LPAE
>> define __ipipe_pin_range_globally function
>> #endif
>> #endif
>>
>> So if IPIPE and LPAE are both true, then the function is never defined.
>>
>> This is in arch/arm/mm/fault.c
>>
> I do not believe this is intentional. That being said, the Linux code in
> the #ifdef is not reassuring: exception handlers are not registered when
> the kernel is compiled with LPAE, which could mean that exceptions are
> used for LPAE when compiled with LPAE. If this is the case, then LPAE
> should not be enabled, otherwise your applications will not stay in
> primary mode when these faults happen. How much RAM do you have, can you
> really test LPAE ?

No, I do not think that LPAE uses exceptions. You can try moving the
#endif at the bottom of the file to above the #ifdef CONFIG_IPIPE.

> 


-- 
                                                                Gilles.

Re: [Xenomai] Any idea how to fix ipipe failing to compile with ARM LPAE enabled?

[Lennart Sorensen]

On Fri, Aug 22, 2014 at 11:36:33PM +0200, Gilles Chanteperdrix wrote:
> On 08/22/2014 11:31 PM, Lennart Sorensen wrote:
> > Any idea how to fix ipipe to not cause a linking error when LPAE is enabled on arm?
> > 
> > I see the function __ipipe_pin_range_globally is inside an ifdef block that says essentially:
> > 
> > #if !IPIPE
> > define __ipipe_pin_range_globally as do nothing
> > #else
> > #if !LPAE
> > define __ipipe_pin_range_globally function
> > #endif
> > #endif
> > 
> > So if IPIPE and LPAE are both true, then the function is never defined.
> > 
> > This is in arch/arm/mm/fault.c
> > 
> I do not believe this is intentional. That being said, the Linux code in
> the #ifdef is not reassuring: exception handlers are not registered when
> the kernel is compiled with LPAE, which could mean that exceptions are
> used for LPAE when compiled with LPAE. If this is the case, then LPAE
> should not be enabled, otherwise your applications will not stay in
> primary mode when these faults happen. How much RAM do you have, can you
> really test LPAE ?

LPAE is required on arm to enable KVM, which we want to experiment with.
Don't need it for the ram quantity.

-- 
Len Sorensen

Re: [Xenomai] Any idea how to fix ipipe failing to compile with ARM LPAE enabled?

[Lennart Sorensen]

> No, I do not think that LPAE uses exceptions. You can try moving the
> #endif at the bottom of the file to above the #ifdef CONFIG_IPIPE.

I did try that, and it did compile.  Never did boot, although I think I
may need to enable hypervisor mode in the boot loader before it will boot.
I will experiment some more.

-- 
Len Sorensen

Re: [Xenomai] Any idea how to fix ipipe failing to compile with ARM LPAE enabled?

[Gilles Chanteperdrix]

On 08/25/2014 04:29 AM, Lennart Sorensen wrote:
>> No, I do not think that LPAE uses exceptions. You can try moving the
>> #endif at the bottom of the file to above the #ifdef CONFIG_IPIPE.
> 
> I did try that, and it did compile.  Never did boot, although I think I
> may need to enable hypervisor mode in the boot loader before it will boot.
> I will experiment some more.

Does the kernel configured without LPAE boot?

-- 
                                                                Gilles.

Re: [Xenomai] Any idea how to fix ipipe failing to compile with ARM LPAE enabled?

[Lennart Sorensen]

On Mon, Aug 25, 2014 at 10:28:22AM +0200, Gilles Chanteperdrix wrote:
> Does the kernel configured without LPAE boot?

Yes.  If I enable KVM (and hence LPAE), I get no console output at all.
I don't think I tried LPAE without KVM yet.  Maybe I should.  I think
I will finish moving from 3.12 to 3.14 first though.

-- 
Len Sorensen