[PATCH] Pre 3430 patch for io.c map table cleanup

[PATCH] Pre 3430 patch for io.c map table cleanup

[Syed Mohammed, Khasim]

Cleanup map descriptor table in io.c for 2420 and 2430

Signed-off-by: Syed Mohammed Khasim  <x0khasim@ti.com>

Files:
 arch/arm/mach-omap2/io.c

=======================================================================

--- linux-omap/arch/arm/mach-omap2/io.c	2007-05-30 14:02:43.000000000 -0500
+++ lin_for_mmc/arch/arm/mach-omap2/io.c	2007-06-14 15:40:42.000000000 -0500
@@ -4,8 +4,11 @@
  * OMAP2 I/O mapping code
  *
  * Copyright (C) 2005 Nokia Corporation
- * Author: Juha Yrjola<juha.yrjola@nokia.com>
- * Updated map desc to add 2430 support : <x0khasim@ti.com>
+ * Copyright (C) 2007 Texas Instruments
+ *
+ * Author:
+ *	Juha Yrjola <juha.yrjola@nokia.com>
+ *	Syed Khasim <x0khasim@ti.com>
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License version 2 as
@@ -35,7 +38,9 @@ extern void omapfb_reserve_sdram(void);
  * The machine specific code may provide the extra mapping besides the
  * default mapping provided here.
  */
-static struct map_desc omap2_io_desc[] __initdata = {
+
+#ifdef CONFIG_ARCH_OMAP24XX
+static struct map_desc omap24xx_io_desc[] __initdata = {
 	{
 		.virtual	= L3_24XX_VIRT,
 		.pfn		= __phys_to_pfn(L3_24XX_PHYS),
@@ -48,7 +53,34 @@ static struct map_desc omap2_io_desc[] _
 		.length		= L4_24XX_SIZE,
 		.type		= MT_DEVICE
 	},
+};
+
+#ifdef CONFIG_ARCH_OMAP2420
+static struct map_desc omap242x_io_desc[] __initdata = {
+	{
+		.virtual	= DSP_MEM_24XX_VIRT,
+		.pfn		= __phys_to_pfn(DSP_MEM_24XX_PHYS),
+		.length		= DSP_MEM_24XX_SIZE,
+		.type		= MT_DEVICE
+	},
+	{
+		.virtual	= DSP_IPI_24XX_VIRT,
+		.pfn		= __phys_to_pfn(DSP_IPI_24XX_PHYS),
+		.length		= DSP_IPI_24XX_SIZE,
+		.type		= MT_DEVICE
+	},
+	{
+		.virtual	= DSP_MMU_24XX_VIRT,
+		.pfn		= __phys_to_pfn(DSP_MMU_24XX_PHYS),
+		.length		= DSP_MMU_24XX_SIZE,
+		.type		= MT_DEVICE
+	},
+};
+
+#endif
+
 #ifdef CONFIG_ARCH_OMAP2430
+static struct map_desc omap243x_io_desc[] __initdata = {
 	{
 		.virtual	= L4_WK_243X_VIRT,
 		.pfn		= __phys_to_pfn(L4_WK_243X_PHYS),
@@ -73,33 +105,19 @@ static struct map_desc omap2_io_desc[] _
 		.length		= OMAP243X_SMS_SIZE,
 		.type		= MT_DEVICE
 	},
+};
 #endif
-	{
-		.virtual	= DSP_MEM_24XX_VIRT,
-		.pfn		= __phys_to_pfn(DSP_MEM_24XX_PHYS),
-		.length		= DSP_MEM_24XX_SIZE,
-		.type		= MT_DEVICE
-	},
-#ifdef CONFIG_ARCH_OMAP2420
-	{
-		.virtual	= DSP_IPI_24XX_VIRT,
-		.pfn		= __phys_to_pfn(DSP_IPI_24XX_PHYS),
-		.length		= DSP_IPI_24XX_SIZE,
-		.type		= MT_DEVICE
-	},
 #endif
-	{
-		.virtual	= DSP_MMU_24XX_VIRT,
-		.pfn		= __phys_to_pfn(DSP_MMU_24XX_PHYS),
-		.length		= DSP_MMU_24XX_SIZE,
-		.type		= MT_DEVICE
-	},
-};
 
 void __init omap2_map_common_io(void)
 {
-	iotable_init(omap2_io_desc, ARRAY_SIZE(omap2_io_desc));
-
+#if defined(CONFIG_ARCH_OMAP2420)
+	iotable_init(omap24xx_io_desc, ARRAY_SIZE(omap24xx_io_desc));
+	iotable_init(omap242x_io_desc, ARRAY_SIZE(omap242x_io_desc));
+#elif defined(CONFIG_ARCH_OMAP2430)
+	iotable_init(omap24xx_io_desc, ARRAY_SIZE(omap24xx_io_desc));
+	iotable_init(omap243x_io_desc, ARRAY_SIZE(omap243x_io_desc));
+#endif
 	/* Normally devicemaps_init() would flush caches and tlb after
 	 * mdesc->map_io(), but we must also do it here because of the CPU
 	 * revision check below.

Re: [PATCH] Pre 3430 patch for io.c map table cleanup

[Tony Lindgren]

* Syed Mohammed, Khasim <x0khasim@ti.com> [070614 14:16]:
> Cleanup map descriptor table in io.c for 2420 and 2430
> 
> Signed-off-by: Syed Mohammed Khasim  <x0khasim@ti.com>
> 
> Files:
>  arch/arm/mach-omap2/io.c
> 
> =======================================================================
> 
> --- linux-omap/arch/arm/mach-omap2/io.c	2007-05-30 14:02:43.000000000 -0500
> +++ lin_for_mmc/arch/arm/mach-omap2/io.c	2007-06-14 15:40:42.000000000 -0500
>  
>  void __init omap2_map_common_io(void)
>  {
> -	iotable_init(omap2_io_desc, ARRAY_SIZE(omap2_io_desc));
> -
> +#if defined(CONFIG_ARCH_OMAP2420)
> +	iotable_init(omap24xx_io_desc, ARRAY_SIZE(omap24xx_io_desc));
> +	iotable_init(omap242x_io_desc, ARRAY_SIZE(omap242x_io_desc));
> +#elif defined(CONFIG_ARCH_OMAP2430)
> +	iotable_init(omap24xx_io_desc, ARRAY_SIZE(omap24xx_io_desc));
> +	iotable_init(omap243x_io_desc, ARRAY_SIZE(omap243x_io_desc));
> +#endif
>  	/* Normally devicemaps_init() would flush caches and tlb after
>  	 * mdesc->map_io(), but we must also do it here because of the CPU
>  	 * revision check below.

You don't need the #defines here. This adds yet another artificial
barrier to compiling in both.

Use cpu_is_omap34xx() et al to init them.

Tony

RE: [PATCH] Pre 3430 patch for io.c map table cleanup

[Syed Mohammed, Khasim]

Hi Tony,

>> +#if defined(CONFIG_ARCH_OMAP2420)
>> +	iotable_init(omap24xx_io_desc, ARRAY_SIZE(omap24xx_io_desc));
>> +	iotable_init(omap242x_io_desc, ARRAY_SIZE(omap242x_io_desc));
>> +#elif defined(CONFIG_ARCH_OMAP2430)
>> +	iotable_init(omap24xx_io_desc, ARRAY_SIZE(omap24xx_io_desc));
>> +	iotable_init(omap243x_io_desc, ARRAY_SIZE(omap243x_io_desc));
>> +#endif
>>  	/* Normally devicemaps_init() would flush caches and tlb after
>>  	 * mdesc->map_io(), but we must also do it here because of the CPU
>>  	 * revision check below.
>
>You don't need the #defines here. This adds yet another artificial
>barrier to compiling in both.
>
>Use cpu_is_omap34xx() et al to init them.

We had discussion on this topic and I had communicated why defines are required.

Let me know your views on the same.

Regards,
Khasim

Mail attached below:
======================================================================
-----Original Message-----
From: linux-omap-open-source-bounces@linux.omap.com [mailto:linux-omap-open-source-bounces@linux.omap.com] On Behalf Of Syed Mohammed, Khasim
Sent: Wednesday, May 30, 2007 5:33 PM
To: Tony Lindgren
Cc: Linux OMAP
Subject: RE: [RFC] io.c for OMAP3430

Hi Tony,

>	if (cpu_is_omap24xx()) {
>		iotable_init(omap24xx_io_desc, ARRAY_SIZE(omap24xx_io_desc));
>		if (cpu_is_omap242x()) {
>			iotable_init(omap242x_io_desc,
>ARRAY_SIZE(omap242x_io_desc));
>		} else if (cpu_is_omap243x()) {
>			iotable_init(omap243x_io_desc,
>ARRAY_SIZE(omap243x_io_desc));
>		}
>	} else if (cpu_is_omap34xx()) {
>		iotable_init(omap34xx_io_desc, ARRAY_SIZE(omap34xx_io_desc));
>	}

We cannot put this kind of check in common_io function. I was trying to validate 2420 / 2430 platforms and I found that "cpu_is_" function calls returning "0" (when "1" was expected). 

Basically the omap2_check_revision function which sets the system_rev for handling the cpu_is macro, gets called after common_io. So putting "cpu_is_" checks in common_io will not work.

I tried making omap2_check_revision as arch_init and sub_sysinit call but it didn't help, the first call for common_io happens from board file's map_io() and we can't avoid the same.

The other possibility is to explicitly call omap2_check_revision from board file's _map_io() functions. But this will have to be updated in many board files.

Any suggestions are appreciated. 

For now I can post a separate patch with #ifdef's instead of if ( cpu_is_ ) later we can work on the correct approach.

Others please review and comment on the approach.

Regards,
Khasim
_______________________________________________

Re: [PATCH] Pre 3430 patch for io.c map table cleanup

[Tony Lindgren]

* Syed Mohammed, Khasim <x0khasim@ti.com> [070621 06:13]:
> Hi Tony,
> 
> >> +#if defined(CONFIG_ARCH_OMAP2420)
> >> +	iotable_init(omap24xx_io_desc, ARRAY_SIZE(omap24xx_io_desc));
> >> +	iotable_init(omap242x_io_desc, ARRAY_SIZE(omap242x_io_desc));
> >> +#elif defined(CONFIG_ARCH_OMAP2430)
> >> +	iotable_init(omap24xx_io_desc, ARRAY_SIZE(omap24xx_io_desc));
> >> +	iotable_init(omap243x_io_desc, ARRAY_SIZE(omap243x_io_desc));
> >> +#endif
> >>  	/* Normally devicemaps_init() would flush caches and tlb after
> >>  	 * mdesc->map_io(), but we must also do it here because of the CPU
> >>  	 * revision check below.
> >
> >You don't need the #defines here. This adds yet another artificial
> >barrier to compiling in both.
> >
> >Use cpu_is_omap34xx() et al to init them.
> 
> We had discussion on this topic and I had communicated why defines are required.

Oh yeah, that's right, sorry I forgot about the cpu detection issue.

> Let me know your views on the same.

Yes, I'll apply your current patch and we'll deal with the cpu
detection issues later on.

Tony