[v2,1/2] ARM: mx27: Add VPU support

mx27 has a VPU (Video Processing Unit) block that allows doing H264/MPEG4
decoding and encoding in hardware.

Add support for it in a common SoC file instead of per board file.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
---
Javier,

I removed dma_declare_coherent_memory call, as per Sascha's suggestion on my
first submission of adding mx27pdk VPU support. Hope this is ok with you.

Changes since v1:
- Call imx27_add_coda() on a common mx27 location instead of per board.
 arch/arm/mach-imx/mach-imx27_visstrim_m10.c |   18 ------------------
 arch/arm/mach-imx/mm-imx27.c                |    3 ++-
 2 files changed, 2 insertions(+), 19 deletions(-)

On Wed, Nov 14, 2012 at 04:20:36PM -0200, Fabio Estevam wrote:
> mx27 has a VPU (Video Processing Unit) block that allows doing H264/MPEG4
> decoding and encoding in hardware.
> 
> Add support for it in a common SoC file instead of per board file.
> 
> Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
> ---
> Javier,
> 
> I removed dma_declare_coherent_memory call, as per Sascha's suggestion on my
> first submission of adding mx27pdk VPU support. Hope this is ok with you.
> 
> Changes since v1:
> - Call imx27_add_coda() on a common mx27 location instead of per board.
>  arch/arm/mach-imx/mach-imx27_visstrim_m10.c |   18 ------------------
>  arch/arm/mach-imx/mm-imx27.c                |    3 ++-
>  2 files changed, 2 insertions(+), 19 deletions(-)
> 
> diff --git a/arch/arm/mach-imx/mach-imx27_visstrim_m10.c b/arch/arm/mach-imx/mach-imx27_visstrim_m10.c
> index b744226..ee9cbff 100644
> --- a/arch/arm/mach-imx/mach-imx27_visstrim_m10.c
> +++ b/arch/arm/mach-imx/mach-imx27_visstrim_m10.c
> @@ -434,23 +434,6 @@ static const struct imx_ssi_platform_data visstrim_m10_ssi_pdata __initconst = {
>  	.flags			= IMX_SSI_DMA | IMX_SSI_SYN,
>  };
>  
> -/* coda */
> -
> -static void __init visstrim_coda_init(void)
> -{
> -	struct platform_device *pdev;
> -	int dma;
> -
> -	pdev = imx27_add_coda();
> -	dma = dma_declare_coherent_memory(&pdev->dev,
> -					  mx2_camera_base + MX2_CAMERA_BUF_SIZE,
> -					  mx2_camera_base + MX2_CAMERA_BUF_SIZE,
> -					  MX2_CAMERA_BUF_SIZE,
> -					  DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE);
> -	if (!(dma & DMA_MEMORY_MAP))
> -		return;
> -}
> -
>  /* DMA deinterlace */
>  static struct platform_device visstrim_deinterlace = {
>  	.name = "m2m-deinterlace",
> @@ -548,7 +531,6 @@ static void __init visstrim_m10_board_init(void)
>  	gpio_led_register_device(0, &visstrim_m10_led_data);
>  	visstrim_deinterlace_init();
>  	visstrim_camera_init();
> -	visstrim_coda_init();
>  }
>  
>  static void __init visstrim_m10_timer_init(void)
> diff --git a/arch/arm/mach-imx/mm-imx27.c b/arch/arm/mach-imx/mm-imx27.c
> index 4f1be65..fddadb1 100644
> --- a/arch/arm/mach-imx/mm-imx27.c
> +++ b/arch/arm/mach-imx/mm-imx27.c
> @@ -25,9 +25,9 @@
>  #include <asm/mach/map.h>
>  
>  #include "common.h"
> -#include "devices/devices-common.h"
>  #include "hardware.h"
>  #include "iomux-v1.h"
> +#include "devices-imx27.h"
>  
>  /* MX27 memory map definition */
>  static struct map_desc imx27_io_desc[] __initdata = {
> @@ -98,4 +98,5 @@ void __init imx27_soc_init(void)
>  	/* imx27 has the imx21 type audmux */
>  	platform_device_register_simple("imx21-audmux", 0, imx27_audmux_res,
>  					ARRAY_SIZE(imx27_audmux_res));
> +	imx27_add_coda();

As Eric mentioned this may lead to problems on the i.MX27L without VPU.
Do you have the possibility to test this?

Sascha

On Thu, Nov 15, 2012 at 6:12 AM, Sascha Hauer <s.hauer@pengutronix.de> wrote:

> As Eric mentioned this may lead to problems on the i.MX27L without VPU.
> Do you have the possibility to test this?

I do not have any board with a MX27L variant. Maybe we could try
reading the SREV info register to distinguish between MX27 and MX27L.
MX27RM does not provide any info on this though.

If this does not work, then maybe we should register VPU as per board
like the v1 patch did.

Regards,

Fabio Estevam

On Thu, Nov 15, 2012 at 11:38:01AM -0200, Fabio Estevam wrote:
> On Thu, Nov 15, 2012 at 6:12 AM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> 
> > As Eric mentioned this may lead to problems on the i.MX27L without VPU.
> > Do you have the possibility to test this?
> 
> I do not have any board with a MX27L variant. Maybe we could try
> reading the SREV info register to distinguish between MX27 and MX27L.
> MX27RM does not provide any info on this though.

If that works I would prefer this, but I also have no i.MX27L to test
it. Otherwise doing it like v1 is the best we can do.
You should probably put the changes to Javiers board into a separate
patch so that Javier can Ack it indepently of what we are coming up
with.

Sascha