[Xenomai-help] Build and run xenomai 2.5.5.2 - linux-2.6.33 on imx27

[Xenomai-help] Build and run xenomai 2.5.5.2 - linux-2.6.33 on imx27

[Fabrice Gasnier]

Dear all,

I'm currently evaluating xenomai 2.5.5.2 on a i.MX27 platform with 
adeos-ipipe-2.6.33-arm-1.18-01.patch.


I've encountred two issues for the time being  (same with 
adeos-ipipe-2.6.33-arm-1.18-00.patch).

1°/ First one happens when building a 2.6.33.3 kernel (after xenomai 
prepare/configure/compile stage):

CC      arch/arm/plat-mxc/time.o
arch/arm/plat-mxc/time.c: In function 'mxc_timer_init':
arch/arm/plat-mxc/time.c:400: error: 'TIM1_BASE_ADDR' undeclared (first 
use in this function)
arch/arm/plat-mxc/time.c:400: error: (Each undeclared identifier is 
reported only once
arch/arm/plat-mxc/time.c:400: error: for each function it appears in.)
make[1]: *** [arch/arm/plat-mxc/time.o] Erreur 1
make: *** [arch/arm/plat-mxc] Erreur 2


I first tried 2.6.31 kernel that was compiling (patched with 
adeos-ipipe-2.6.31-arm-1.16-02.patch).
Looking for differences lead me to patch my kernel with:

Index: linux-2.6.33.3/arch/arm/plat-mxc/time.c
===================================================================
--- linux-2.6.33.3.orig/arch/arm/plat-mxc/time.c    2011-02-28 
11:44:02.000000000 +0100
+++ linux-2.6.33.3/arch/arm/plat-mxc/time.c    2011-02-28 
11:48:19.000000000 +0100
@@ -395,12 +395,17 @@
      tsc_info.freq = clk_get_rate(timer_clk);
      mxc_min_delay = ((__ipipe_cpu_freq + 500000) / 1000000) ?: 1;

-    if (cpu_is_mx1() || cpu_is_mx2()) {
-#if defined(CONFIG_ARCH_MX1) || defined(CONFIG_ARCH_MX2)
+    if (cpu_is_mx1()) {
+#ifdef CONFIG_ARCH_MX1
          tsc_info.u.counter_paddr = (TIM1_BASE_ADDR + MX1_2_TCN);
          tsc_info.counter_vaddr =
              (unsigned long)(timer_base + MX1_2_TCN);
  #endif
+    } else if (cpu_is_mx2()) {
+#ifdef CONFIG_ARCH_MX2
+        tsc_info.u.counter_paddr = (GPT1_BASE_ADDR + MX1_2_TCN);
+        tsc_info.counter_vaddr = (unsigned long)(timer_base + MX1_2_TCN);
+#endif
      } else if (cpu_is_mx3()) {
  #ifdef CONFIG_ARCH_MX3
          tsc_info.u.counter_paddr = (GPT1_BASE_ADDR + MX3_TCN);


2°/ Second issue comes when trying to launch xeno-test:
[...]
root@domain.hid:~ /usr/xenomai/share/xenomai/testsuite/latency/run -- -sh 
-T 120 -t0
*
*
* Type ^C to stop this application.
*
*
== Sampling period: 1000 us
== Test mode: periodic user-mode task
== All results in microseconds
warming up...
Unhandled fault: external abort on non-linefetch (0x008) at 0x40021010
Bus error
[...]


Looking at that thread:
http://www.mail-archive.com/xenomai@xenomai.org
It appears a similar issue has already been fixed on an i.MX25 by 
clearing supervisor's protection bit.


So, I tried the following, that seem to fix the issue as well:
Index: linux-2.6.33.3/arch/arm/mach-mx2/devices.c
===================================================================
--- linux-2.6.33.3.orig/arch/arm/mach-mx2/devices.c    2011-03-01 
10:51:53.000000000 +0100
+++ linux-2.6.33.3/arch/arm/mach-mx2/devices.c    2011-03-02 
17:44:08.000000000 +0100
@@ -680,3 +680,16 @@
  {
      return mxc_gpio_init(imx_gpio_ports, ARRAY_SIZE(imx_gpio_ports));
  }
+
+#ifdef CONFIG_IPIPE
+static int post_cpu_init(void)
+{
+    /* Allow userspace access through the
+     * peripheral access register (PAR).
+     */
+    if (cpu_is_mx27 ())
+        ipipe_mach_allow_uaccess(AIPI_BASE_ADDR, 3);
+    return 0;
+}
+postcore_initcall(post_cpu_init);
+#endif /* CONFIG_IPIPE */
Index: linux-2.6.33.3/arch/arm/plat-mxc/cpu.c
===================================================================
--- linux-2.6.33.3.orig/arch/arm/plat-mxc/cpu.c    2011-03-01 
15:11:57.000000000 +0100
+++ linux-2.6.33.3/arch/arm/plat-mxc/cpu.c    2011-03-02 
09:38:28.000000000 +0100
@@ -1,5 +1,6 @@
  #include <linux/module.h>
  #include <linux/io.h>
+#include <mach/hardware.h>

  unsigned int __mxc_cpu_type;
  EXPORT_SYMBOL(__mxc_cpu_type);
@@ -34,4 +35,16 @@
      aips_reg &= 0x00FFFFFF;
      __raw_writel(aips_reg, aips2 + 0x50);
  }
+
+#if defined(CONFIG_ARCH_MX1) || defined(CONFIG_ARCH_MX2)
+void ipipe_mach_allow_uaccess(unsigned long aipi_base_addr, unsigned 
long bit_num)
+{
+    u32 tmp;
+
+    tmp = __raw_readl(IO_ADDRESS(aipi_base_addr + 0x00000008));
+    __raw_writel(tmp & ~(1 << bit_num), IO_ADDRESS(aipi_base_addr + 
0x00000008));
+
+}
+#endif
+
  #endif /* CONFIG_IPIPE */
Index: linux-2.6.33.3/arch/arm/plat-mxc/include/mach/common.h
===================================================================
--- linux-2.6.33.3.orig/arch/arm/plat-mxc/include/mach/common.h    
2011-03-01 15:14:35.000000000 +0100
+++ linux-2.6.33.3/arch/arm/plat-mxc/include/mach/common.h    2011-03-02 
09:47:34.000000000 +0100
@@ -47,6 +47,9 @@

  #ifdef CONFIG_IPIPE
  void ipipe_mach_allow_hwtimer_uaccess(unsigned long aips1, unsigned 
long aips2);
+#if defined(CONFIG_ARCH_MX1) || defined(CONFIG_ARCH_MX2)
+void ipipe_mach_allow_uaccess(unsigned long aipi_base_addr, unsigned 
long bit_num);
+#endif
  #endif

  #endif



Could someone advise me on the modifications brought here?
Do these modifications look like ok to you?

Thanks in advance,

Fabrice

Re: [Xenomai-help] Build and run xenomai 2.5.5.2 - linux-2.6.33 on imx27

[Gilles Chanteperdrix]

Fabrice Gasnier wrote:
> Dear all,
> 
> I'm currently evaluating xenomai 2.5.5.2 on a i.MX27 platform with 
> adeos-ipipe-2.6.33-arm-1.18-01.patch.
> 
> 
> I've encountred two issues for the time being  (same with 
> adeos-ipipe-2.6.33-arm-1.18-00.patch).
> 
> 1°/ First one happens when building a 2.6.33.3 kernel (after xenomai 
> prepare/configure/compile stage):
> 
> CC      arch/arm/plat-mxc/time.o
> arch/arm/plat-mxc/time.c: In function 'mxc_timer_init':
> arch/arm/plat-mxc/time.c:400: error: 'TIM1_BASE_ADDR' undeclared (first 
> use in this function)
> arch/arm/plat-mxc/time.c:400: error: (Each undeclared identifier is 
> reported only once
> arch/arm/plat-mxc/time.c:400: error: for each function it appears in.)
> make[1]: *** [arch/arm/plat-mxc/time.o] Erreur 1
> make: *** [arch/arm/plat-mxc] Erreur 2
> 
> 
> I first tried 2.6.31 kernel that was compiling (patched with 
> adeos-ipipe-2.6.31-arm-1.16-02.patch).
> Looking for differences lead me to patch my kernel with:
> 
> Index: linux-2.6.33.3/arch/arm/plat-mxc/time.c
> ===================================================================
> --- linux-2.6.33.3.orig/arch/arm/plat-mxc/time.c    2011-02-28 
> 11:44:02.000000000 +0100
> +++ linux-2.6.33.3/arch/arm/plat-mxc/time.c    2011-02-28 
> 11:48:19.000000000 +0100
> @@ -395,12 +395,17 @@
>       tsc_info.freq = clk_get_rate(timer_clk);
>       mxc_min_delay = ((__ipipe_cpu_freq + 500000) / 1000000) ?: 1;
> 
> -    if (cpu_is_mx1() || cpu_is_mx2()) {
> -#if defined(CONFIG_ARCH_MX1) || defined(CONFIG_ARCH_MX2)
> +    if (cpu_is_mx1()) {
> +#ifdef CONFIG_ARCH_MX1
>           tsc_info.u.counter_paddr = (TIM1_BASE_ADDR + MX1_2_TCN);
>           tsc_info.counter_vaddr =
>               (unsigned long)(timer_base + MX1_2_TCN);
>   #endif
> +    } else if (cpu_is_mx2()) {
> +#ifdef CONFIG_ARCH_MX2
> +        tsc_info.u.counter_paddr = (GPT1_BASE_ADDR + MX1_2_TCN);
> +        tsc_info.counter_vaddr = (unsigned long)(timer_base + MX1_2_TCN);
> +#endif
>       } else if (cpu_is_mx3()) {
>   #ifdef CONFIG_ARCH_MX3
>           tsc_info.u.counter_paddr = (GPT1_BASE_ADDR + MX3_TCN);

Ok. I'll take this one.

>   #ifdef CONFIG_IPIPE
>   void ipipe_mach_allow_hwtimer_uaccess(unsigned long aips1, unsigned 
> long aips2);
> +#if defined(CONFIG_ARCH_MX1) || defined(CONFIG_ARCH_MX2)
> +void ipipe_mach_allow_uaccess(unsigned long aipi_base_addr, unsigned 
> long bit_num);
> +#endif
>   #endif

Look the function above, you are adding a function which is already
there, please reuse the same function, instead of duplicating it...

-- 
					    Gilles.

Re: [Xenomai-help] Build and run xenomai 2.5.5.2 - linux-2.6.33 on imx27

[Fabrice Gasnier]

Gilles Chanteperdrix wrote:
> Fabrice Gasnier wrote:
>    
>> Dear all,
>>
>> I'm currently evaluating xenomai 2.5.5.2 on a i.MX27 platform with
>> adeos-ipipe-2.6.33-arm-1.18-01.patch.
>>
>>
>> I've encountred two issues for the time being  (same with
>> adeos-ipipe-2.6.33-arm-1.18-00.patch).
>>
>> 1°/ First one happens when building a 2.6.33.3 kernel (after xenomai
>> prepare/configure/compile stage):
>>
>> CC      arch/arm/plat-mxc/time.o
>> arch/arm/plat-mxc/time.c: In function 'mxc_timer_init':
>> arch/arm/plat-mxc/time.c:400: error: 'TIM1_BASE_ADDR' undeclared (first
>> use in this function)
>> arch/arm/plat-mxc/time.c:400: error: (Each undeclared identifier is
>> reported only once
>> arch/arm/plat-mxc/time.c:400: error: for each function it appears in.)
>> make[1]: *** [arch/arm/plat-mxc/time.o] Erreur 1
>> make: *** [arch/arm/plat-mxc] Erreur 2
>>
>>
>> I first tried 2.6.31 kernel that was compiling (patched with
>> adeos-ipipe-2.6.31-arm-1.16-02.patch).
>> Looking for differences lead me to patch my kernel with:
>>
>> Index: linux-2.6.33.3/arch/arm/plat-mxc/time.c
>> ===================================================================
>> --- linux-2.6.33.3.orig/arch/arm/plat-mxc/time.c    2011-02-28
>> 11:44:02.000000000 +0100
>> +++ linux-2.6.33.3/arch/arm/plat-mxc/time.c    2011-02-28
>> 11:48:19.000000000 +0100
>> @@ -395,12 +395,17 @@
>>        tsc_info.freq = clk_get_rate(timer_clk);
>>        mxc_min_delay = ((__ipipe_cpu_freq + 500000) / 1000000) ?: 1;
>>
>> -    if (cpu_is_mx1() || cpu_is_mx2()) {
>> -#if defined(CONFIG_ARCH_MX1) || defined(CONFIG_ARCH_MX2)
>> +    if (cpu_is_mx1()) {
>> +#ifdef CONFIG_ARCH_MX1
>>            tsc_info.u.counter_paddr = (TIM1_BASE_ADDR + MX1_2_TCN);
>>            tsc_info.counter_vaddr =
>>                (unsigned long)(timer_base + MX1_2_TCN);
>>    #endif
>> +    } else if (cpu_is_mx2()) {
>> +#ifdef CONFIG_ARCH_MX2
>> +        tsc_info.u.counter_paddr = (GPT1_BASE_ADDR + MX1_2_TCN);
>> +        tsc_info.counter_vaddr = (unsigned long)(timer_base + MX1_2_TCN);
>> +#endif
>>        } else if (cpu_is_mx3()) {
>>    #ifdef CONFIG_ARCH_MX3
>>            tsc_info.u.counter_paddr = (GPT1_BASE_ADDR + MX3_TCN);
>>      
> Ok. I'll take this one.
>
>    
>>    #ifdef CONFIG_IPIPE
>>    void ipipe_mach_allow_hwtimer_uaccess(unsigned long aips1, unsigned
>> long aips2);
>> +#if defined(CONFIG_ARCH_MX1) || defined(CONFIG_ARCH_MX2)
>> +void ipipe_mach_allow_uaccess(unsigned long aipi_base_addr, unsigned
>> long bit_num);
>> +#endif
>>    #endif
>>      
> Look the function above, you are adding a function which is already
> there, please reuse the same function, instead of duplicating it...
>
>    
Not exactly the same, I'd be pleased to use it, but I didn't manage to 
get it working (yet?). It hangs everything when I try to use it.
It seems there are differences beetween existing routine (used by 
i.MX25) and what is needed by i.MX27.
Having a look at datasheet for both products shows i.MX25 has an "AIPS" 
where i.MX27 has an "AIPI".
Mapping in both case seem to be slightly different (registers seems 
different). I'm quite confused.
This is why i created quite similar routine ...

In i.MX25 case:
__raw_writel(0x0, aips1 + 0x40);
__raw_writel(0x0, aips1 + 0x44);
__raw_writel(0x0, aips1 + 0x48);
__raw_writel(0x0, aips1 + 0x4C);


In i.MX27 case:
__raw_writel(tmp & ~(1 << bit_num), IO_ADDRESS(aipi_base_addr + 
0x00000008));

This second alternative seem to work properly.


Please advise.

Thanks.

Fabrice

Re: [Xenomai-help] Build and run xenomai 2.5.5.2 - linux-2.6.33 on imx27

[Gilles Chanteperdrix]

Fabrice Gasnier wrote:
> Gilles Chanteperdrix wrote:
>> Fabrice Gasnier wrote:
>>    
>>> Dear all,
>>>
>>> I'm currently evaluating xenomai 2.5.5.2 on a i.MX27 platform with
>>> adeos-ipipe-2.6.33-arm-1.18-01.patch.
>>>
>>>
>>> I've encountred two issues for the time being  (same with
>>> adeos-ipipe-2.6.33-arm-1.18-00.patch).
>>>
>>> 1°/ First one happens when building a 2.6.33.3 kernel (after xenomai
>>> prepare/configure/compile stage):
>>>
>>> CC      arch/arm/plat-mxc/time.o
>>> arch/arm/plat-mxc/time.c: In function 'mxc_timer_init':
>>> arch/arm/plat-mxc/time.c:400: error: 'TIM1_BASE_ADDR' undeclared (first
>>> use in this function)
>>> arch/arm/plat-mxc/time.c:400: error: (Each undeclared identifier is
>>> reported only once
>>> arch/arm/plat-mxc/time.c:400: error: for each function it appears in.)
>>> make[1]: *** [arch/arm/plat-mxc/time.o] Erreur 1
>>> make: *** [arch/arm/plat-mxc] Erreur 2
>>>
>>>
>>> I first tried 2.6.31 kernel that was compiling (patched with
>>> adeos-ipipe-2.6.31-arm-1.16-02.patch).
>>> Looking for differences lead me to patch my kernel with:
>>>
>>> Index: linux-2.6.33.3/arch/arm/plat-mxc/time.c
>>> ===================================================================
>>> --- linux-2.6.33.3.orig/arch/arm/plat-mxc/time.c    2011-02-28
>>> 11:44:02.000000000 +0100
>>> +++ linux-2.6.33.3/arch/arm/plat-mxc/time.c    2011-02-28
>>> 11:48:19.000000000 +0100
>>> @@ -395,12 +395,17 @@
>>>        tsc_info.freq = clk_get_rate(timer_clk);
>>>        mxc_min_delay = ((__ipipe_cpu_freq + 500000) / 1000000) ?: 1;
>>>
>>> -    if (cpu_is_mx1() || cpu_is_mx2()) {
>>> -#if defined(CONFIG_ARCH_MX1) || defined(CONFIG_ARCH_MX2)
>>> +    if (cpu_is_mx1()) {
>>> +#ifdef CONFIG_ARCH_MX1
>>>            tsc_info.u.counter_paddr = (TIM1_BASE_ADDR + MX1_2_TCN);
>>>            tsc_info.counter_vaddr =
>>>                (unsigned long)(timer_base + MX1_2_TCN);
>>>    #endif
>>> +    } else if (cpu_is_mx2()) {
>>> +#ifdef CONFIG_ARCH_MX2
>>> +        tsc_info.u.counter_paddr = (GPT1_BASE_ADDR + MX1_2_TCN);
>>> +        tsc_info.counter_vaddr = (unsigned long)(timer_base + MX1_2_TCN);
>>> +#endif
>>>        } else if (cpu_is_mx3()) {
>>>    #ifdef CONFIG_ARCH_MX3
>>>            tsc_info.u.counter_paddr = (GPT1_BASE_ADDR + MX3_TCN);
>>>      
>> Ok. I'll take this one.
>>
>>    
>>>    #ifdef CONFIG_IPIPE
>>>    void ipipe_mach_allow_hwtimer_uaccess(unsigned long aips1, unsigned
>>> long aips2);
>>> +#if defined(CONFIG_ARCH_MX1) || defined(CONFIG_ARCH_MX2)
>>> +void ipipe_mach_allow_uaccess(unsigned long aipi_base_addr, unsigned
>>> long bit_num);
>>> +#endif
>>>    #endif
>>>      
>> Look the function above, you are adding a function which is already
>> there, please reuse the same function, instead of duplicating it...
>>
>>    
> Not exactly the same, I'd be pleased to use it, but I didn't manage to 
> get it working (yet?). It hangs everything when I try to use it.
> It seems there are differences beetween existing routine (used by 
> i.MX25) and what is needed by i.MX27.
> Having a look at datasheet for both products shows i.MX25 has an "AIPS" 
> where i.MX27 has an "AIPI".
> Mapping in both case seem to be slightly different (registers seems 
> different). I'm quite confused.
> This is why i created quite similar routine ...
> 
> In i.MX25 case:
> __raw_writel(0x0, aips1 + 0x40);
> __raw_writel(0x0, aips1 + 0x44);
> __raw_writel(0x0, aips1 + 0x48);
> __raw_writel(0x0, aips1 + 0x4C);
> 
> 
> In i.MX27 case:
> __raw_writel(tmp & ~(1 << bit_num), IO_ADDRESS(aipi_base_addr + 
> 0x00000008));
> 
> This second alternative seem to work properly.

Just put an if in the function... if (cpu_is_mx27()) etc... Remember
that the MXC code is supposed to be able to run on several boards at the
same time.


-- 
					    Gilles.

Re: [Xenomai-help] Build and run xenomai 2.5.5.2 - linux-2.6.33 on imx27

[Gilles Chanteperdrix]

Fabrice Gasnier wrote:
> Dear all,
> 
> I'm currently evaluating xenomai 2.5.5.2 on a i.MX27 platform with 
> adeos-ipipe-2.6.33-arm-1.18-01.patch.

Could you try the following patch? Instead of
adeos-ipipe-2.6.33-arm-1.18-01.patch?

http://www.xenomai.org/~gch/pub/adeos-ipipe-2.6.33-arm-1.18-02-rc1.patch

It fixes things a bit differently.

Thanks in advance.

-- 
                                                                Gilles.

Re: [Xenomai-help] Build and run xenomai 2.5.5.2 - linux-2.6.33 on imx27

[Fabrice Gasnier]

Hi,

Off course, I would try it. I cannot get access to it (403 Forbidden, 
You don't have permission to access 
/~gch/pub/adeos-ipipe-2.6.33-arm-1.18-02-rc1.patch on this server.)


On 07/03/2011 22:28, Gilles Chanteperdrix wrote:
> Fabrice Gasnier wrote:
>    
>> Dear all,
>>
>> I'm currently evaluating xenomai 2.5.5.2 on a i.MX27 platform with
>> adeos-ipipe-2.6.33-arm-1.18-01.patch.
>>      
> Could you try the following patch? Instead of
> adeos-ipipe-2.6.33-arm-1.18-01.patch?
>
> http://www.xenomai.org/~gch/pub/adeos-ipipe-2.6.33-arm-1.18-02-rc1.patch
>
> It fixes things a bit differently.
>
> Thanks in advance.
>
>    

Re: [Xenomai-help] Build and run xenomai 2.5.5.2 - linux-2.6.33 on imx27

[Gilles Chanteperdrix]

Fabrice Gasnier wrote:
> Hi,
> 
> Off course, I would try it. I cannot get access to it (403 Forbidden, 
> You don't have permission to access 
> /~gch/pub/adeos-ipipe-2.6.33-arm-1.18-02-rc1.patch on this server.)
> 

Sorry, please try again...

-- 
                                                                Gilles.

Re: [Xenomai-help] Build and run xenomai 2.5.5.2 - linux-2.6.33 on imx27

[Gilles Chanteperdrix]

Fabrice Gasnier wrote:
> Hi,
> 
> Off course, I would try it. I cannot get access to it (403 Forbidden, 
> You don't have permission to access 
> /~gch/pub/adeos-ipipe-2.6.33-arm-1.18-02-rc1.patch on this server.)

Since you are at it, could you try:

http://www.xenomai.org/~gch/pub/adeos-ipipe-2.6.35-arm-1.18-01-rc1.patch
?

-- 
                                                                Gilles.

Re: [Xenomai-help] Build and run xenomai 2.5.5.2 - linux-2.6.33 on imx27

[Fabrice Gasnier]

Fine,

This patch fixes both issues on 2.6.33 kernel!
Thanks.
I'll let you know in case I detect something else.

Regarding 2.6.35 patch, I need to check more things as I don't use it 
for the moment. (no valuable config / board support yet).
I quiclky tried it, but cannot boot it (yet). I'll give it some more 
tries and keep you informed.

Thanks again.
Fabrice

On 07/03/2011 22:28, Gilles Chanteperdrix wrote:
> Fabrice Gasnier wrote:
>    
>> Dear all,
>>
>> I'm currently evaluating xenomai 2.5.5.2 on a i.MX27 platform with
>> adeos-ipipe-2.6.33-arm-1.18-01.patch.
>>      
> Could you try the following patch? Instead of
> adeos-ipipe-2.6.33-arm-1.18-01.patch?
>
> http://www.xenomai.org/~gch/pub/adeos-ipipe-2.6.33-arm-1.18-02-rc1.patch
>
> It fixes things a bit differently.
>
> Thanks in advance.
>
>    

Re: [Xenomai-help] Build and run xenomai 2.5.5.2 - linux-2.6.33 on imx27

[Gilles Chanteperdrix]

Fabrice Gasnier wrote:
> Fine,
> 
> This patch fixes both issues on 2.6.33 kernel!
> Thanks.
> I'll let you know in case I detect something else.
> 
> Regarding 2.6.35 patch, I need to check more things as I don't use it 
> for the moment. (no valuable config / board support yet).
> I quiclky tried it, but cannot boot it (yet). I'll give it some more 
> tries and keep you informed.
> 
> Thanks again.
> Fabrice

Ok. Forget it, the code in the 2.6.35 patch is the same anyway, and was
build-tested on imx27. 2.5.6 will be released soon with both these patches.


-- 
					    Gilles.