[Xenomai-core] xenomai and framebuffer-ipu on target iMx35

[Xenomai-core] xenomai and framebuffer-ipu on target iMx35

[François Beaufort]

[-- Attachment #1: Type: text/plain, Size: 2149 bytes --]


Hello,

I have to port Linux-xenomai on target Phytec PhyCARD-M iMx35 (ARM 11 core) and I need your help.

I reconfigured the kernel from the configuration provided by Phytec :

    - CONFIG_PREEMPT = y
    - CONFIG_VGA_CONSOLE = n
    - CONFIG_MX3_IPU = n
    - CONFIG_FB_MX3 = n
    - CONFIG_USB_ISP1760 = n

So, Linux-xenomai runs on the target, but I can't have pictures on the display (the configuration of the framebuffer is not set : CONFIG_MX3_IPU and CONFIG_FB_MX3).

I would like to enable xenomai and the framebuffer.

When I patch Linux with adeos-ipipe, disable xenomai and enable ipu (image processing unit), the kernel doesn't start :

        booting kernel of type uimage from /dev/ram0.kernel
           Verifying Checksum ... OK
           Image Name:   Linux-2.6.35.9
           Created:      2011-04-20   7:14:44 UTC
           Image Type:   ARM Linux Kernel Image (uncompressed)
           Data Size:    2446512 Bytes =  2.3 MB
           Load Address: 80008000
           Entry Point:  80008000
        OK
        commandline: console=ttymxc0,115200 video=mx3fb:Primeview-PD050VL1 ip=192.168.0.195::192.168.0.1:255.255.255.0::: root=/dev/nfs   nfsroot=192.168.0.84:/opt/francois/nfs_imx35_xenom;
        arch_number: 2595

        Starting kernel ...

        Linux version 2.6.35.9 (francois@domain.hid) (gcc version 4.3.2 (OSELAS.Toolchain-1.99.3) ) #2 PREEMPT Wed Apr 20 09:14:32 CEST 2011
        CPU: ARMv6-compatible processor [4117b363] revision 3 (ARMv6TEJ), cr=00c5387f
        CPU: VIPT nonaliasing data cache, VIPT nonaliasing instruction c


I think there is a conflict between ipipe and ipu but I don't know why.

IPU depends on dma engine.

When I read the /proc files, I can't see IRQ conflicts nor memory conflicts.

The ipu driver consists of two C files : ipu_idmac.c and ipu_irq.c. 
When I disable ipu_idma.c, I have 4 more IRQs (CONFIG_MX3_IPU_IRQS = 4) and Linux-ipipe (and xenomai) starts but I can't have pictures on the display.

I can't debug with printk because the printk calls modify the behaviour.

Do you have any idea ?

Thank you. 		 	   		  

[-- Attachment #2: Type: text/html, Size: 3131 bytes --]

Re: [Xenomai-core] xenomai and framebuffer-ipu on target iMx35

[Gilles Chanteperdrix]

François Beaufort wrote:
> Hello,
> 
> I have to port Linux-xenomai on target Phytec PhyCARD-M iMx35 (ARM 11 core) and I need your help.
> 
> I reconfigured the kernel from the configuration provided by Phytec :
> 
>     - CONFIG_PREEMPT = y
>     - CONFIG_VGA_CONSOLE = n
>     - CONFIG_MX3_IPU = n
>     - CONFIG_FB_MX3 = n
>     - CONFIG_USB_ISP1760 = n
> 
> So, Linux-xenomai runs on the target, but I can't have pictures on the display (the configuration of the framebuffer is not set : CONFIG_MX3_IPU and CONFIG_FB_MX3).
> 
> I would like to enable xenomai and the framebuffer.
> 
> When I patch Linux with adeos-ipipe, disable xenomai and enable ipu (image processing unit), the kernel doesn't start :
> 
>         booting kernel of type uimage from /dev/ram0.kernel
>            Verifying Checksum ... OK
>            Image Name:   Linux-2.6.35.9
>            Created:      2011-04-20   7:14:44 UTC
>            Image Type:   ARM Linux Kernel Image (uncompressed)
>            Data Size:    2446512 Bytes =  2.3 MB
>            Load Address: 80008000
>            Entry Point:  80008000
>         OK
>         commandline: console=ttymxc0,115200 video=mx3fb:Primeview-PD050VL1 ip=192.168.0.195::192.168.0.1:255.255.255.0::: root=/dev/nfs   nfsroot=192.168.0.84:/opt/francois/nfs_imx35_xenom;
>         arch_number: 2595
> 
>         Starting kernel ...
> 
>         Linux version 2.6.35.9 (francois@domain.hid) (gcc version 4.3.2 (OSELAS.Toolchain-1.99.3) ) #2 PREEMPT Wed Apr 20 09:14:32 CEST 2011
>         CPU: ARMv6-compatible processor [4117b363] revision 3 (ARMv6TEJ), cr=00c5387f
>         CPU: VIPT nonaliasing data cache, VIPT nonaliasing instruction c
> 
> 
> I think there is a conflict between ipipe and ipu but I don't know why.
> 
> IPU depends on dma engine.
> 
> When I read the /proc files, I can't see IRQ conflicts nor memory conflicts.
> 
> The ipu driver consists of two C files : ipu_idmac.c and ipu_irq.c. 
> When I disable ipu_idma.c, I have 4 more IRQs (CONFIG_MX3_IPU_IRQS = 4) and Linux-ipipe (and xenomai) starts but I can't have pictures on the display.
> 
> I can't debug with printk because the printk calls modify the behaviour.

How so?

> 
> Do you have any idea ?

First thing to confirm: does the same configuration which does not work
with I-pipe enabled work with I-pipe disable. I mean, the exact same
configuration except for CONFIG_IPIPE?

Second thing: if the IPU dma engine, whatever, uses an interrupt, can
you compare the counters in /proc/interrupts for this interrupt with and
without CONFIG_IPIPE enabled?


-- 
					    Gilles.

Re: [Xenomai-help] [Xenomai-core] xenomai and framebuffer-ipu on target iMx35

[Gilles Chanteperdrix]

François Beaufort wrote:
> I can't show you  /proc/interrupts when ipu and I-pipe are enabled
> because the system doesn't start in this case.

Ok. I failed to get that from your first mail. I do not know either if
you are using a vanilla kernel, or a patched one?

If a patched one, then you have to verify that generic_handle_irq is
replaced with ipipe_handle_chained_irq for multiplexed GPIO irqs.

Note that I CCd the Xenomai-help mailing list, next private mail will
receive no answer.

-- 
                                                                Gilles.

Re: [Xenomai-help] [Xenomai-core] xenomai and framebuffer-ipu on target iMx35

[François Beaufort]

[-- Attachment #1: Type: text/plain, Size: 1263 bytes --]




> Date: Wed, 20 Apr 2011 12:54:00 +0200
> From: gilles.chanteperdrix@xenomai.org
> To: francois.beaufort@domain.hid
> CC: xenomai@xenomai.org
> Subject: Re: [Xenomai-core] xenomai and framebuffer-ipu on target iMx35
> 
> François Beaufort wrote:
> > I can't show you  /proc/interrupts when ipu and I-pipe are enabled
> > because the system doesn't start in this case.
> 
> Ok. I failed to get that from your first mail. I do not know either if
> you are using a vanilla kernel, or a patched one?
> 
> If a patched one, then you have to verify that generic_handle_irq is
> replaced with ipipe_handle_chained_irq for multiplexed GPIO irqs.
> 
> Note that I CCd the Xenomai-help mailing list, next private mail will
> receive no answer.
> 
> -- 
>                                                                 Gilles.
> 

Thank you for your answers.

I can now start Linux-xenomai with framebuffer and ipu.

I use a kernel patched by the Phytec BSP patches and next by the adeos pipe.

I have modified the file linux-2.6.35.9/drivers/dma/ipu/ipu_irq.c :

lines 305 and 343 :

-  generic_handle_irq(irq);
+ ipipe_handle_chained_irq(irq);


This file is not modified by any patch.

Thank you.

François.

 		 	   		  

[-- Attachment #2: Type: text/html, Size: 1626 bytes --]

Re: [Xenomai-help] [Xenomai-core] xenomai and framebuffer-ipu on target iMx35

[Gilles Chanteperdrix]

François Beaufort wrote:
> 
> 
>> Date: Wed, 20 Apr 2011 12:54:00 +0200
>> From: gilles.chanteperdrix@xenomai.org
>> To: francois.beaufort@domain.hid
>> CC: xenomai@xenomai.org
>> Subject: Re: [Xenomai-core] xenomai and framebuffer-ipu on target iMx35
>>
>> François Beaufort wrote:
>>> I can't show you  /proc/interrupts when ipu and I-pipe are enabled
>>> because the system doesn't start in this case.
>> Ok. I failed to get that from your first mail. I do not know either if
>> you are using a vanilla kernel, or a patched one?
>>
>> If a patched one, then you have to verify that generic_handle_irq is
>> replaced with ipipe_handle_chained_irq for multiplexed GPIO irqs.
>>
>> Note that I CCd the Xenomai-help mailing list, next private mail will
>> receive no answer.
>>
>> -- 
>>                                                                 Gilles.
>>
> 
> Thank you for your answers.
> 
> I can now start Linux-xenomai with framebuffer and ipu.
> 
> I use a kernel patched by the Phytec BSP patches and next by the adeos pipe.
> 
> I have modified the file linux-2.6.35.9/drivers/dma/ipu/ipu_irq.c :
> 
> lines 305 and 343 :
> 
> -  generic_handle_irq(irq);
> + ipipe_handle_chained_irq(irq);
> 
> 
> This file is not modified by any patch.

You have to replace any call to generic_handle_irq in the Phytec patch
by a call to ipipe_handle_chained_irq, otherwise, you will get this
issue again with other multiplexed irqs.

-- 
					    Gilles.