[PATCH v1 0/9] OMAP: DMA: hwmod and DMA as platform device

[PATCH v1 0/9] OMAP: DMA: hwmod and DMA as platform device

[G, Manjunath Kondaiah]

Hi Tony,

* G, Manjunath Kondaiah <manjugk@ti.com> [2010-12-04 09:13:57 +0530]:

> Patch series to convert DMA library into platform driver using platform
> device model and adapting hwmod for omap2+.
> 
> The original patch series :
> http://comments.gmane.org/gmane.linux.ports.arm.omap/46953
> has been split into two patch series based on suggestion from Tony.
> (https://patchwork.kernel.org/patch/375831/)
> 
> The first series will prepare existing DMA library for DMA hwmod and 
> converting the same into platform driver.
> 
> The second series will have:
> arch/arm/mach-omap1/dma.c       omap1 specific platform init
> arch/arm/mach-omap2/dma.c       omap2+ specific platform init
> drivers/dma/omap-dma.c          driver using dmaengine.c
> 
> Patch series1 Design:
> 1. The low level read/write macros are converted into static inline functions
> so that, these functions can be moved to respective mach-omap driver
> files later.
> (Thanks to Tony and Kevin for their suggestions on handling all omap register
> offset without adding extra enums)
> 2. Implements generic errata handling for all OMAP DMA errata.
> 3. DMA hwmod data is updated for respective hwmod db files.
> 4. The DMA library is split into two layers.
>    a. The generic code is retained in plat-omap/dma.c
>    b. The machine specific init code is moved to
>       respective mach-omap dma files.
> 
> Minimal machine specific code is moved to respective mach-omap dma files with 
> this series. Rest of code movement and API cleanup's are handled in second
> series.
> 
> Patch series applies on top of latest linux omap master branch:
> *************************************************************************
> git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap-2.6.git
> Branch: master
> commit a04fd22204b13ce34a3f8a8157f83c44d64f8da9
> Merge: e941bb0 afd2d11
> Author: Tony Lindgren <tony@atomide.com>
> 
>     Linux-omap rebuilt: Merged in usb patches for testing
> *************************************************************************

If there are no further issues, can you pls merge this series for
testing. I can send pull request if it is convinient.

As per the alignment with original series(v5), all the fixes done for
OMAP1 are pulled into this patch series. OMAP1 and OMAP2+ will build and
boots with published test reports.

-Manjunath

[PATCH v1 0/9] OMAP: DMA: hwmod and DMA as platform device

[Tony Lindgren]

* G, Manjunath Kondaiah <manjugk@ti.com> [101203 19:33]:
> *************************************************************************
> 
> Test Report:
> ************
> Build:
> omap2plus_defconfig    : Success
> omap_osk_5912_defconfig: Success
> 
> Boot:
> OMAP3530Beagle         : Success
> OMAP4430Blaze(ES2.1)   : Success

Can you please make sure that memory to memory DMA test
runs OK on OSK5912?x

AFAIK, Kevin tried it a few days ago and it hung.

To me that sounds like the only remaining issue merging these
in for testing.

Regards,

Tony

[PATCH v1 0/9] OMAP: DMA: hwmod and DMA as platform device

[Kevin Hilman]

Tony Lindgren <tony@atomide.com> writes:

> * G, Manjunath Kondaiah <manjugk@ti.com> [101203 19:33]:
>> *************************************************************************
>> 
>> Test Report:
>> ************
>> Build:
>> omap2plus_defconfig    : Success
>> omap_osk_5912_defconfig: Success
>> 
>> Boot:
>> OMAP3530Beagle         : Success
>> OMAP4430Blaze(ES2.1)   : Success
>
> Can you please make sure that memory to memory DMA test
> runs OK on OSK5912?x
>
> AFAIK, Kevin tried it a few days ago and it hung.

Yes, it hung, but I didn't have time to debug the root cause until
today.

The reason is that the OMAP1 DMA IRQ numbers used in mach-omap1/dma.c
have no basis in reality. So doing a memory-to-memory transfer using
*all* channels results in spurious interrupts, then hangs.

Will post details in reply to the OMAP1 patch (7/9). 

Kevin

[PATCH v1 7/9] OMAP1: DMA: Implement in platform device model

[Kevin Hilman]

Manjunath,

"G, Manjunath Kondaiah" <manjugk@ti.com> writes:

> Implement OMAP1 DMA as platform device and add support for
> registering through platform device layer using resource
> structures.
>
> Signed-off-by: G, Manjunath Kondaiah <manjugk@ti.com>
> Cc: Benoit Cousson <b-cousson@ti.com>
> Cc: Kevin Hilman <khilman@deeprootsystems.com>
> Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>

Using a memory-to-memory DMA test which exercises all available
channels, this hangs on OMAP1 (OMAP5912/OSK.)

The root cause is that the interrupt numbers used here are very
different from those of the original code it is replacing.  Details
below...

> ---
>  arch/arm/mach-omap1/dma.c |  179 +++++++++++++++++++++++++++++++++++++++++++++
>  1 files changed, 179 insertions(+), 0 deletions(-)
>  create mode 100644 arch/arm/mach-omap1/dma.c
>
> diff --git a/arch/arm/mach-omap1/dma.c b/arch/arm/mach-omap1/dma.c
> new file mode 100644
> index 0000000..b56ee21
> --- /dev/null
> +++ b/arch/arm/mach-omap1/dma.c
> @@ -0,0 +1,179 @@
> +/*
> + * OMAP1/OMAP7xx - specific DMA driver
> + *
> + * Copyright (C) 2003 - 2008 Nokia Corporation
> + * Author: Juha Yrj?l? <juha.yrjola@nokia.com>
> + * DMA channel linking for 1610 by Samuel Ortiz <samuel.ortiz@nokia.com>
> + * Graphics DMA and LCD DMA graphics tranformations
> + * by Imre Deak <imre.deak@nokia.com>
> + * OMAP2/3 support Copyright (C) 2004-2007 Texas Instruments, Inc.
> + * Some functions based on earlier dma-omap.c Copyright (C) 2001 RidgeRun, Inc.
> + *
> + * Copyright (C) 2010 Texas Instruments Incorporated - http://www.ti.com/
> + * Converted DMA library into platform driver
> + *                   - G, Manjunath Kondaiah <manjugk@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
> + * published by the Free Software Foundation.
> + */
> +
> +#include <linux/err.h>
> +#include <linux/io.h>
> +#include <linux/slab.h>
> +#include <linux/module.h>
> +#include <linux/init.h>
> +#include <linux/device.h>
> +
> +#include <plat/dma.h>
> +#include <plat/tc.h>
> +#include <plat/irqs.h>
> +
> +#define OMAP1_DMA_BASE			(0xfffed800)
> +
> +static struct resource res[] __initdata = {
> +	[0] = {
> +		.start	= OMAP1_DMA_BASE,
> +		.end	= OMAP1_DMA_BASE + SZ_2K - 1,
> +		.flags	= IORESOURCE_MEM,
> +	},
> +	[1] = {
> +		.name   = "0",
> +		.start  = INT_DMA_CH0_6,
> +		.flags  = IORESOURCE_IRQ,
> +	},

Minor nit: to keep the index numbers and the 'name' fields aligned, and
less confusing, you could move the memory resource to the end.  Either
that, or just drop the hard-coded index values.

> +	[2] = {
> +		.name   = "1",
> +		.start  = INT_DMA_CH1_7,
> +		.flags  = IORESOURCE_IRQ,
> +	},
> +	[3] = {
> +		.name   = "2",
> +		.start  = INT_DMA_CH2_8,
> +		.flags  = IORESOURCE_IRQ,
> +	},
> +	[4] = {
> +		.name   = "3",
> +		.start  = INT_DMA_CH3,
> +		.flags  = IORESOURCE_IRQ,
> +	},
> +	[5] = {
> +		.name   = "4",
> +		.start  = INT_DMA_CH4,
> +		.flags  = IORESOURCE_IRQ,
> +	},
> +	[6] = {
> +		.name   = "5",
> +		.start  = INT_DMA_CH5,
> +		.flags  = IORESOURCE_IRQ,
> +	},

Here's the array of interrupt numbers from the original code (which you
remove in PATCH 9/9)

static const u8 omap1_dma_irq[OMAP1_LOGICAL_DMA_CH_COUNT] = {
	INT_DMA_CH0_6, INT_DMA_CH1_7, INT_DMA_CH2_8, INT_DMA_CH3,
	INT_DMA_CH4, INT_DMA_CH5, INT_1610_DMA_CH6, INT_1610_DMA_CH7,
	INT_1610_DMA_CH8, INT_1610_DMA_CH9, INT_1610_DMA_CH10,
	INT_1610_DMA_CH11, INT_1610_DMA_CH12, INT_1610_DMA_CH13,
	INT_1610_DMA_CH14, INT_1610_DMA_CH15, INT_DMA_LCD
};

Up until this point in the patch, the interrupt numbers match up
with the previous code.

> +	[7] = {
> +		.name   = "6",
> +		.start  = INT_DMA_LCD,
> +		.flags  = IORESOURCE_IRQ,
> +	},

Starting here though, you use INT_DMA_LCD for LCH 6, where previous code
uses INT_1610_DMA_CH6.

> +	/* irq's for omap16xx and omap7xx */
> +	[8] = {
> +		.name   = "7",
> +		.start  = 53 + IH2_BASE,
> +		.flags  = IORESOURCE_IRQ,
> +	},

Here, there are a couple problems.

First, exising #defines for interrupt numbers are no longer used used,
and second, this is the wrong value.  Looking at plat/irqs.h:

#define INT_1610_DMA_CH6	(53 + IH2_BASE)
#define INT_1610_DMA_CH7	(54 + IH2_BASE)
[...]
#define INT_1610_DMA_CH15	(62 + IH2_BASE)

You can see that the IRQ value for LCH 7 in this patch is actually the
value for LCH 6.  The same "one off" problem exists for the rest of the
channels below up through LCH 15.

> +	[9] = {
> +		.name   = "8",
> +		.start  = 54 + IH2_BASE,
> +		.flags  = IORESOURCE_IRQ,
> +	},
> +	[10] = {
> +		.name  = "9",
> +		.start = 55 + IH2_BASE,
> +		.flags = IORESOURCE_IRQ,
> +	},
> +	[11] = {
> +		.name  = "10",
> +		.start = 56 + IH2_BASE,
> +		.flags = IORESOURCE_IRQ,
> +	},
> +	[12] = {
> +		.name  = "11",
> +		.start = 57 + IH2_BASE,
> +		.flags = IORESOURCE_IRQ,
> +	},
> +	[13] = {
> +		.name  = "12",
> +		.start = 58 + IH2_BASE,
> +		.flags = IORESOURCE_IRQ,
> +	},
> +	[14] = {
> +		.name  = "13",
> +		.start = 59 + IH2_BASE,
> +		.flags = IORESOURCE_IRQ,
> +	},
> +	[15] = {
> +		.name  = "14",
> +		.start = 60 + IH2_BASE,
> +		.flags = IORESOURCE_IRQ,
> +	},
> +	[16] = {
> +		.name  = "15",
> +		.start = 61 + IH2_BASE,
> +		.flags = IORESOURCE_IRQ,
> +	},
> +	[17] = {
> +		.name  = "16",
> +		.start = 62 + IH2_BASE,
> +		.flags = IORESOURCE_IRQ,
> +	},

And based on the original code, this last one should be INT_LCD_DMA.

Using the patch below which fixes up the interrupt numbers to match the
original code, the memory-to-memory test is now working on OSK.

Kevin

diff --git a/arch/arm/mach-omap1/dma.c b/arch/arm/mach-omap1/dma.c
index 17814e0..d855934 100644
--- a/arch/arm/mach-omap1/dma.c
+++ b/arch/arm/mach-omap1/dma.c
@@ -121,58 +121,58 @@ static struct resource res[] __initdata = {
 	/* Handled in lcd_dma.c */
 	[7] = {
 		.name   = "6",
-		.start  = INT_DMA_LCD,
+		.start  = INT_1610_DMA_CH6,
 		.flags  = IORESOURCE_IRQ,
 	},
 	/* irq's for omap16xx and omap7xx */
 	[8] = {
 		.name   = "7",
-		.start  = 53 + IH2_BASE,
+		.start  = INT_1610_DMA_CH7,
 		.flags  = IORESOURCE_IRQ,
 	},
 	[9] = {
 		.name   = "8",
-		.start  = 54 + IH2_BASE,
+		.start  = INT_1610_DMA_CH8,
 		.flags  = IORESOURCE_IRQ,
 	},
 	[10] = {
 		.name  = "9",
-		.start = 55 + IH2_BASE,
+		.start = INT_1610_DMA_CH9,
 		.flags = IORESOURCE_IRQ,
 	},
 	[11] = {
 		.name  = "10",
-		.start = 56 + IH2_BASE,
+		.start = INT_1610_DMA_CH10,
 		.flags = IORESOURCE_IRQ,
 	},
 	[12] = {
 		.name  = "11",
-		.start = 57 + IH2_BASE,
+		.start = INT_1610_DMA_CH11,
 		.flags = IORESOURCE_IRQ,
 	},
 	[13] = {
 		.name  = "12",
-		.start = 58 + IH2_BASE,
+		.start = INT_1610_DMA_CH12,
 		.flags = IORESOURCE_IRQ,
 	},
 	[14] = {
 		.name  = "13",
-		.start = 59 + IH2_BASE,
+		.start = INT_1610_DMA_CH13,
 		.flags = IORESOURCE_IRQ,
 	},
 	[15] = {
 		.name  = "14",
-		.start = 60 + IH2_BASE,
+		.start = INT_1610_DMA_CH14,
 		.flags = IORESOURCE_IRQ,
 	},
 	[16] = {
 		.name  = "15",
-		.start = 61 + IH2_BASE,
+		.start = INT_1610_DMA_CH15,
 		.flags = IORESOURCE_IRQ,
 	},
 	[17] = {
 		.name  = "16",
-		.start = 62 + IH2_BASE,
+		.start = INT_DMA_LCD,
 		.flags = IORESOURCE_IRQ,
 	},
 };

[PATCH v1 7/9] OMAP1: DMA: Implement in platform device model

[Kevin Hilman]

Kevin Hilman <khilman@deeprootsystems.com> writes:

[...]

>
> Using the patch below which fixes up the interrupt numbers to match the
> original code, the memory-to-memory test is now working on OSK.

Tony, if you prefer just to fold this into the original, feel free to
add

[khilman at deeprootsystems.com]: fixed IRQ numbers to match original code
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>


> diff --git a/arch/arm/mach-omap1/dma.c b/arch/arm/mach-omap1/dma.c
> index 17814e0..d855934 100644
> --- a/arch/arm/mach-omap1/dma.c
> +++ b/arch/arm/mach-omap1/dma.c
> @@ -121,58 +121,58 @@ static struct resource res[] __initdata = {
>  	/* Handled in lcd_dma.c */
>  	[7] = {
>  		.name   = "6",
> -		.start  = INT_DMA_LCD,
> +		.start  = INT_1610_DMA_CH6,
>  		.flags  = IORESOURCE_IRQ,
>  	},
>  	/* irq's for omap16xx and omap7xx */
>  	[8] = {
>  		.name   = "7",
> -		.start  = 53 + IH2_BASE,
> +		.start  = INT_1610_DMA_CH7,
>  		.flags  = IORESOURCE_IRQ,
>  	},
>  	[9] = {
>  		.name   = "8",
> -		.start  = 54 + IH2_BASE,
> +		.start  = INT_1610_DMA_CH8,
>  		.flags  = IORESOURCE_IRQ,
>  	},
>  	[10] = {
>  		.name  = "9",
> -		.start = 55 + IH2_BASE,
> +		.start = INT_1610_DMA_CH9,
>  		.flags = IORESOURCE_IRQ,
>  	},
>  	[11] = {
>  		.name  = "10",
> -		.start = 56 + IH2_BASE,
> +		.start = INT_1610_DMA_CH10,
>  		.flags = IORESOURCE_IRQ,
>  	},
>  	[12] = {
>  		.name  = "11",
> -		.start = 57 + IH2_BASE,
> +		.start = INT_1610_DMA_CH11,
>  		.flags = IORESOURCE_IRQ,
>  	},
>  	[13] = {
>  		.name  = "12",
> -		.start = 58 + IH2_BASE,
> +		.start = INT_1610_DMA_CH12,
>  		.flags = IORESOURCE_IRQ,
>  	},
>  	[14] = {
>  		.name  = "13",
> -		.start = 59 + IH2_BASE,
> +		.start = INT_1610_DMA_CH13,
>  		.flags = IORESOURCE_IRQ,
>  	},
>  	[15] = {
>  		.name  = "14",
> -		.start = 60 + IH2_BASE,
> +		.start = INT_1610_DMA_CH14,
>  		.flags = IORESOURCE_IRQ,
>  	},
>  	[16] = {
>  		.name  = "15",
> -		.start = 61 + IH2_BASE,
> +		.start = INT_1610_DMA_CH15,
>  		.flags = IORESOURCE_IRQ,
>  	},
>  	[17] = {
>  		.name  = "16",
> -		.start = 62 + IH2_BASE,
> +		.start = INT_DMA_LCD,
>  		.flags = IORESOURCE_IRQ,
>  	},
>  };

[PATCH v1 7/9] OMAP1: DMA: Implement in platform device model

[Tony Lindgren]

* Kevin Hilman <khilman@deeprootsystems.com> [101214 16:55]:
> Kevin Hilman <khilman@deeprootsystems.com> writes:
> 
> [...]
> 
> >
> > Using the patch below which fixes up the interrupt numbers to match the
> > original code, the memory-to-memory test is now working on OSK.
> 
> Tony, if you prefer just to fold this into the original, feel free to
> add
> 
> [khilman at deeprootsystems.com]: fixed IRQ numbers to match original code
> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>

OK thanks, will fold it in and let's merge them into linux-omap
master branch for few days for testing. Then if everything looks
good, I'll queue these for the 2.6.38 next merge window.

Regards,

Tony

[PATCH v1 0/9] OMAP: DMA: hwmod and DMA as platform device

[Kevin Hilman]

"G, Manjunath Kondaiah" <manjugk@ti.com> writes:

> Patch series to convert DMA library into platform driver using platform
> device model and adapting hwmod for omap2+.

After the OMAP1 fixup for PATCH 7/9,

Acked-by: Kevin Hilman <khilman@deeprootsystems.com>

and

Tested-by: Kevin Hilman <khilman@deeprootsystems.com>

on omap1611/OSK, omap3430/n900 and omap4430/panda

Tested using a memory-to-memory DMA test which tests all available
channels, and tests linking all the channels together.

I resurected an older DMA test that I wrote sometime around 2004, and
lost somehow (thankfully, Tony still had a copy.)  I've updated it
slightly and created a git repo for it now that it is known working on
OMAP1, 2, 3 and 4.

       git://gitorious.org/omap-test/dmatest.git

It's a bit hacky as you have to manually edit the Makefile to build an
out of tree module, but it does work.

Kevin

> The original patch series :
> http://comments.gmane.org/gmane.linux.ports.arm.omap/46953
> has been split into two patch series based on suggestion from Tony.
> (https://patchwork.kernel.org/patch/375831/)
>
> The first series will prepare existing DMA library for DMA hwmod and 
> converting the same into platform driver.
>
> The second series will have:
> arch/arm/mach-omap1/dma.c       omap1 specific platform init
> arch/arm/mach-omap2/dma.c       omap2+ specific platform init
> drivers/dma/omap-dma.c          driver using dmaengine.c
>
> Patch series1 Design:
> 1. The low level read/write macros are converted into static inline functions
> so that, these functions can be moved to respective mach-omap driver
> files later.
> (Thanks to Tony and Kevin for their suggestions on handling all omap register
> offset without adding extra enums)
> 2. Implements generic errata handling for all OMAP DMA errata.
> 3. DMA hwmod data is updated for respective hwmod db files.
> 4. The DMA library is split into two layers.
>    a. The generic code is retained in plat-omap/dma.c
>    b. The machine specific init code is moved to
>       respective mach-omap dma files.
>
> Minimal machine specific code is moved to respective mach-omap dma files with 
> this series. Rest of code movement and API cleanup's are handled in second
> series.
>
> Patch series applies on top of latest linux omap master branch:
> *************************************************************************
> git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap-2.6.git
> Branch: master
> commit a04fd22204b13ce34a3f8a8157f83c44d64f8da9
> Merge: e941bb0 afd2d11
> Author: Tony Lindgren <tony@atomide.com>
>
>     Linux-omap rebuilt: Merged in usb patches for testing
> *************************************************************************
>
> Test Report:
> ************
> Build:
> omap2plus_defconfig    : Success
> omap_osk_5912_defconfig: Success
>
> Boot:
> OMAP3530Beagle         : Success
> OMAP4430Blaze(ES2.1)   : Success
> OMAP1(OSK5912)         : Success
>
> Test cases executed:
> 1. All applicable TI DMA tests which are located at:
> git://dev.omapzoom.org/pub/scm/richo/device_driver_test.git
> Branch: master
>
> Report can be accessed at: 
> Beagle board: http://pastebin.com/sDUChNLr
>
> The original patch series and change history can be found at:
> http://permalink.gmane.org/gmane.linux.ports.arm.omap/46953#
>
> Benoit Cousson (1):
>   OMAP4: hwmod data: add system DMA
>
> G, Manjunath Kondaiah (8):
>   OMAP: DMA: Replace read/write macros with functions
>   OMAP: DMA: Introduce errata handling feature
>   OMAP2420: hwmod data: add system DMA
>   OMAP2430: hwmod data: add system DMA
>   OMAP3: hwmod data: add system DMA
>   OMAP1: DMA: Implement in platform device model
>   OMAP2+: DMA: hwmod: Device registration
>   OMAP: DMA: Convert DMA library into platform driver
>
>  arch/arm/mach-omap1/Makefile               |    2 +-
>  arch/arm/mach-omap1/dma.c                  |  390 ++++++++++++++++
>  arch/arm/mach-omap2/Makefile               |    2 +-
>  arch/arm/mach-omap2/dma.c                  |  297 ++++++++++++
>  arch/arm/mach-omap2/omap_hwmod_2420_data.c |   87 ++++
>  arch/arm/mach-omap2/omap_hwmod_2430_data.c |   87 ++++
>  arch/arm/mach-omap2/omap_hwmod_3xxx_data.c |   97 ++++
>  arch/arm/mach-omap2/omap_hwmod_44xx_data.c |  101 ++++
>  arch/arm/plat-omap/dma.c                   |  697 ++++++++++++----------------
>  arch/arm/plat-omap/include/plat/dma.h      |  232 ++++------
>  10 files changed, 1452 insertions(+), 540 deletions(-)
>  create mode 100644 arch/arm/mach-omap1/dma.c
>  create mode 100644 arch/arm/mach-omap2/dma.c
>
> Cc: Benoit Cousson <b-cousson@ti.com>
> Cc: Kevin Hilman <khilman@deeprootsystems.com>
> Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>

[PATCH v1 0/9] OMAP: DMA: hwmod and DMA as platform device

[Tony Lindgren]

* Kevin Hilman <khilman@deeprootsystems.com> [101214 17:37]:
> "G, Manjunath Kondaiah" <manjugk@ti.com> writes:
> 
> > Patch series to convert DMA library into platform driver using platform
> > device model and adapting hwmod for omap2+.
> 
> After the OMAP1 fixup for PATCH 7/9,
> 
> Acked-by: Kevin Hilman <khilman@deeprootsystems.com>

I guess here you mean Signed-off-by for this patch since
you wrote some code? Assuming Signed-off-by, let me know
if you want to change that.
 
> and
> 
> Tested-by: Kevin Hilman <khilman@deeprootsystems.com>

Adding this to the others in the series.

Regards,

Tony

[PATCH v1 0/9] OMAP: DMA: hwmod and DMA as platform device

[G, Manjunath Kondaiah]

Kevin/Tony,

On Tue, Dec 14, 2010 at 05:37:14PM -0800, Kevin Hilman wrote:
> "G, Manjunath Kondaiah" <manjugk@ti.com> writes:
> 
> > Patch series to convert DMA library into platform driver using platform
> > device model and adapting hwmod for omap2+.
> 
> After the OMAP1 fixup for PATCH 7/9,
> 
> Acked-by: Kevin Hilman <khilman@deeprootsystems.com>
> 
> and
> 
> Tested-by: Kevin Hilman <khilman@deeprootsystems.com>
> 
> on omap1611/OSK, omap3430/n900 and omap4430/panda
> 
> Tested using a memory-to-memory DMA test which tests all available
> channels, and tests linking all the channels together.
> 
> I resurected an older DMA test that I wrote sometime around 2004, and
> lost somehow (thankfully, Tony still had a copy.)  I've updated it
> slightly and created a git repo for it now that it is known working on
> OMAP1, 2, 3 and 4.
> 
>        git://gitorious.org/omap-test/dmatest.git

Nice. TI too has exhaustive memory to memory test cases available at
public git repo(includes other driver test cases as well):

git://dev.omapzoom.org/pub/scm/richo/device_driver_test.git

> 
> It's a bit hacky as you have to manually edit the Makefile to build an
> out of tree module, but it does work.

Thanks for fixing OMAP1 irq issues and testing the same. I hope we will
have OSK5912 for future patches and we will test all upstream patches on
OMAP1 too.

-Manjunath

[PATCH v1 7/9] OMAP1: DMA: Implement in platform device model

[G, Manjunath Kondaiah]

On Tue, Dec 14, 2010 at 05:04:00PM -0800, Tony Lindgren wrote:
> * Kevin Hilman <khilman@deeprootsystems.com> [101214 16:55]:
> > Kevin Hilman <khilman@deeprootsystems.com> writes:
> > 
> > [...]
> > 
> > >
> > > Using the patch below which fixes up the interrupt numbers to match the
> > > original code, the memory-to-memory test is now working on OSK.
> > 
> > Tony, if you prefer just to fold this into the original, feel free to
> > add
> > 
> > [khilman at deeprootsystems.com]: fixed IRQ numbers to match original code
> > Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
> 
> OK thanks, will fold it in and let's merge them into linux-omap
> master branch for few days for testing. Then if everything looks
> good, I'll queue these for the 2.6.38 next merge window.

Thanks Tony.

-Manjunath

[PATCH v1 0/9] OMAP: DMA: hwmod and DMA as platform device

[Kevin Hilman]

Tony Lindgren <tony@atomide.com> writes:

> * Kevin Hilman <khilman@deeprootsystems.com> [101214 17:37]:
>> "G, Manjunath Kondaiah" <manjugk@ti.com> writes:
>> 
>> > Patch series to convert DMA library into platform driver using platform
>> > device model and adapting hwmod for omap2+.
>> 
>> After the OMAP1 fixup for PATCH 7/9,
>> 
>> Acked-by: Kevin Hilman <khilman@deeprootsystems.com>
>
> I guess here you mean Signed-off-by for this patch since
> you wrote some code? Assuming Signed-off-by, let me know
> if you want to change that.

I only added my s-o-b to PATCH 7/9 since that's the only one I added
code to.

Kevin

>> and
>> 
>> Tested-by: Kevin Hilman <khilman@deeprootsystems.com>
>
> Adding this to the others in the series.
>
> Regards,
>
> Tony

[PATCH v1 0/9] OMAP: DMA: hwmod and DMA as platform device

[Tony Lindgren]

* Kevin Hilman <khilman@deeprootsystems.com> [101214 18:02]:
> Tony Lindgren <tony@atomide.com> writes:
> 
> > * Kevin Hilman <khilman@deeprootsystems.com> [101214 17:37]:
> >> "G, Manjunath Kondaiah" <manjugk@ti.com> writes:
> >> 
> >> > Patch series to convert DMA library into platform driver using platform
> >> > device model and adapting hwmod for omap2+.
> >> 
> >> After the OMAP1 fixup for PATCH 7/9,
> >> 
> >> Acked-by: Kevin Hilman <khilman@deeprootsystems.com>
> >
> > I guess here you mean Signed-off-by for this patch since
> > you wrote some code? Assuming Signed-off-by, let me know
> > if you want to change that.
> 
> I only added my s-o-b to PATCH 7/9 since that's the only one I added
> code to.

Sorry I misunderstood. Adding your Signed-off-by for 7/9, and Acked-by
and Tested-by for others.

Thanks,

Tony

[PATCH v1 3/9] OMAP2420: hwmod data: add system DMA

[Paul Walmsley]

Manju

On Sat, 4 Dec 2010, G, Manjunath Kondaiah wrote:

> Add OMAP2420 DMA hwmod data and also add required
> DMA device attributes.
> 
> Signed-off-by: G, Manjunath Kondaiah <manjugk@ti.com>
> Cc: Benoit Cousson <b-cousson@ti.com>
> Cc: Kevin Hilman <khilman@deeprootsystems.com>
> Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>
>
> +/* dma_system -> L3 */
> +static struct omap_hwmod_ocp_if omap2420_dma_system__l3 = {
> +	.master		= &omap2420_dma_system_hwmod,
> +	.slave		= &omap2420_l3_main_hwmod,
> +	.clk		= "l3_div_ck",

This clock does not exist on OMAP2420.  Did you test this patch on 2420?

> +/* l4_cfg -> dma_system */
> +static struct omap_hwmod_ocp_if omap2420_l4_core__dma_system = {
> +	.master		= &omap2420_l4_core_hwmod,
> +	.slave		= &omap2420_dma_system_hwmod,
> +	.clk		= "l4_div_ck",

Nor does this clock exist on OMAP2420.

> +static struct omap_hwmod omap2420_dma_system_hwmod = {
> +	.name		= "dma",
> +	.class		= &omap2420_dma_hwmod_class,
> +	.mpu_irqs	= omap2420_dma_system_irqs,
> +	.mpu_irqs_cnt	= ARRAY_SIZE(omap2420_dma_system_irqs),
> +	.main_clk	= "l3_div_ck",

And neither does this one.

Please fix these and test on OMAP2420 before sending the fixed patches.


- Paul

[PATCH v1 4/9] OMAP2430: hwmod data: add system DMA

[Paul Walmsley]

On Sat, 4 Dec 2010, G, Manjunath Kondaiah wrote:

> Add OMAP2430 DMA hwmod data and also add required
> DMA device attributes.

...

> +/* dma_system -> L3 */
> +static struct omap_hwmod_ocp_if omap2430_dma_system__l3 = {
> +	.master		= &omap2430_dma_system_hwmod,
> +	.slave		= &omap2430_l3_main_hwmod,
> +	.clk		= "l3_div_ck",

This clock does not exist on OMAP2430.  Did you test this on OMAP2430?

> +/* l4_cfg -> dma_system */
> +static struct omap_hwmod_ocp_if omap2430_l4_core__dma_system = {
> +	.master		= &omap2430_l4_core_hwmod,
> +	.slave		= &omap2430_dma_system_hwmod,
> +	.clk		= "l4_div_ck",

This clock also does not exist on OMAP2430.

> +static struct omap_hwmod omap2430_dma_system_hwmod = {
> +	.name		= "dma",
> +	.class		= &omap2430_dma_hwmod_class,
> +	.mpu_irqs	= omap2430_dma_system_irqs,
> +	.mpu_irqs_cnt	= ARRAY_SIZE(omap2430_dma_system_irqs),
> +	.main_clk	= "l3_div_ck",

Nor does this one.

Please fix these and test on OMAP2430 before resending.


- Paul

[PATCH v1 7/9] OMAP1: DMA: Implement in platform device model

[Tony Lindgren]

* G, Manjunath Kondaiah <manjugk@ti.com> [101214 18:01]:
> On Tue, Dec 14, 2010 at 05:04:00PM -0800, Tony Lindgren wrote:
> > 
> > OK thanks, will fold it in and let's merge them into linux-omap
> > master branch for few days for testing. Then if everything looks
> > good, I'll queue these for the 2.6.38 next merge window.

Oops, not quite ready yet, see the mail from Paul. Waiting for
those fixes.

Tony

[PATCH v1 4/9] OMAP2430: hwmod data: add system DMA

[G, Manjunath Kondaiah]

On Tue, Dec 14, 2010 at 07:27:33PM -0700, Paul Walmsley wrote:
> On Sat, 4 Dec 2010, G, Manjunath Kondaiah wrote:
> 
> > Add OMAP2430 DMA hwmod data and also add required
> > DMA device attributes.
> 
> ...
> 
> > +/* dma_system -> L3 */
> > +static struct omap_hwmod_ocp_if omap2430_dma_system__l3 = {
> > +	.master		= &omap2430_dma_system_hwmod,
> > +	.slave		= &omap2430_l3_main_hwmod,
> > +	.clk		= "l3_div_ck",
> 
> This clock does not exist on OMAP2430.  Did you test this on OMAP2430?

Yes. I have tested this on SDP2430. For confirmation, I tested again now.
It boots up without any issues and all DMA test cases are passing.

> 
> > +/* l4_cfg -> dma_system */
> > +static struct omap_hwmod_ocp_if omap2430_l4_core__dma_system = {
> > +	.master		= &omap2430_l4_core_hwmod,
> > +	.slave		= &omap2430_dma_system_hwmod,
> > +	.clk		= "l4_div_ck",
> 
> This clock also does not exist on OMAP2430.

ok. I will replace with "sdma_ick" here.

> 
> > +static struct omap_hwmod omap2430_dma_system_hwmod = {
> > +	.name		= "dma",
> > +	.class		= &omap2430_dma_hwmod_class,
> > +	.mpu_irqs	= omap2430_dma_system_irqs,
> > +	.mpu_irqs_cnt	= ARRAY_SIZE(omap2430_dma_system_irqs),
> > +	.main_clk	= "l3_div_ck",
> 
> Nor does this one.

ok. will be replaced with "sdma_fck"

> 
> Please fix these and test on OMAP2430 before resending.

Thanks. I will test again with above changes on SDP2430.

-Manjunath

[PATCH v1 3/9] OMAP2420: hwmod data: add system DMA

[G, Manjunath Kondaiah]

On Tue, Dec 14, 2010 at 07:25:23PM -0700, Paul Walmsley wrote:
> Manju
> 
> On Sat, 4 Dec 2010, G, Manjunath Kondaiah wrote:
> 
> > Add OMAP2420 DMA hwmod data and also add required
> > DMA device attributes.
> > 
> > Signed-off-by: G, Manjunath Kondaiah <manjugk@ti.com>
> > Cc: Benoit Cousson <b-cousson@ti.com>
> > Cc: Kevin Hilman <khilman@deeprootsystems.com>
> > Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>
> >
> > +/* dma_system -> L3 */
> > +static struct omap_hwmod_ocp_if omap2420_dma_system__l3 = {
> > +	.master		= &omap2420_dma_system_hwmod,
> > +	.slave		= &omap2420_l3_main_hwmod,
> > +	.clk		= "l3_div_ck",
> 
> This clock does not exist on OMAP2420.  Did you test this patch on 2420?

ok. will be replaced with "sdma_ick". I don't have 2420 for testing. I
tested it on 2430. My understanding is that, DMA clock interface is same
for 2420 and 2430. Correct me if am wrong.

> 
> > +/* l4_cfg -> dma_system */
> > +static struct omap_hwmod_ocp_if omap2420_l4_core__dma_system = {
> > +	.master		= &omap2420_l4_core_hwmod,
> > +	.slave		= &omap2420_dma_system_hwmod,
> > +	.clk		= "l4_div_ck",
> 
> Nor does this clock exist on OMAP2420.

ok. Will replace with "sdma_ick"
> 
> > +static struct omap_hwmod omap2420_dma_system_hwmod = {
> > +	.name		= "dma",
> > +	.class		= &omap2420_dma_hwmod_class,
> > +	.mpu_irqs	= omap2420_dma_system_irqs,
> > +	.mpu_irqs_cnt	= ARRAY_SIZE(omap2420_dma_system_irqs),
> > +	.main_clk	= "l3_div_ck",
> 
> And neither does this one.

This will be replaced with "sdma_fck"
> 
> Please fix these and test on OMAP2420 before sending the fixed patches.

I don't have setup to test this. I will do the changes and test it on
2430SDP which has similar change.


-Manjunath

[PATCH v1 3/9] OMAP2420: hwmod data: add system DMA]

[G, Manjunath Kondaiah]

On Wed, Dec 15, 2010 at 04:39:17PM +0530, G, Manjunath Kondaiah wrote:
> On Tue, Dec 14, 2010 at 07:25:23PM -0700, Paul Walmsley wrote:
> > Manju
> > 
> > On Sat, 4 Dec 2010, G, Manjunath Kondaiah wrote:
> > 
> > > Add OMAP2420 DMA hwmod data and also add required
> > > DMA device attributes.
> > > 
> > > Signed-off-by: G, Manjunath Kondaiah <manjugk@ti.com>
> > > Cc: Benoit Cousson <b-cousson@ti.com>
> > > Cc: Kevin Hilman <khilman@deeprootsystems.com>
> > > Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>
> > >
> > > +/* dma_system -> L3 */
> > > +static struct omap_hwmod_ocp_if omap2420_dma_system__l3 = {
> > > +	.master		= &omap2420_dma_system_hwmod,
> > > +	.slave		= &omap2420_l3_main_hwmod,
> > > +	.clk		= "l3_div_ck",
> > 
> > This clock does not exist on OMAP2420.  Did you test this patch on 2420?
> 
> ok. will be replaced with "sdma_ick". I don't have 2420 for testing. I
> tested it on 2430. My understanding is that, DMA clock interface is same
> for 2420 and 2430. Correct me if am wrong.

Looking into omap3/omap4 structures, this entry takes core l3 clock.

This should be "sdma_fck" since it's parent is core_l3_ck.

Pls confirm.

-Manjunath

[PATCH v1 4/9] OMAP2430: hwmod data: add system DMA

[Paul Walmsley]

On Wed, 15 Dec 2010, G, Manjunath Kondaiah wrote:

> On Tue, Dec 14, 2010 at 07:27:33PM -0700, Paul Walmsley wrote:
> > On Sat, 4 Dec 2010, G, Manjunath Kondaiah wrote:
> > 
> > > Add OMAP2430 DMA hwmod data and also add required
> > > DMA device attributes.
> > 
> > ...
> > 
> > > +/* dma_system -> L3 */
> > > +static struct omap_hwmod_ocp_if omap2430_dma_system__l3 = {
> > > +	.master		= &omap2430_dma_system_hwmod,
> > > +	.slave		= &omap2430_l3_main_hwmod,
> > > +	.clk		= "l3_div_ck",
> > 
> > This clock does not exist on OMAP2430.  Did you test this on OMAP2430?
> 
> Yes. I have tested this on SDP2430. For confirmation, I tested again now.
> It boots up without any issues and all DMA test cases are passing.

Don't you see warnings in the boot messages that this clock cannot be 
found?



- Paul

[PATCH v1 3/9] OMAP2420: hwmod data: add system DMA]

[Paul Walmsley]

On Wed, 15 Dec 2010, G, Manjunath Kondaiah wrote:

> On Wed, Dec 15, 2010 at 04:39:17PM +0530, G, Manjunath Kondaiah wrote:
> > On Tue, Dec 14, 2010 at 07:25:23PM -0700, Paul Walmsley wrote:
> > > On Sat, 4 Dec 2010, G, Manjunath Kondaiah wrote:
> > > 
> > > > Add OMAP2420 DMA hwmod data and also add required
> > > > DMA device attributes.
> > > > 
> > > > Signed-off-by: G, Manjunath Kondaiah <manjugk@ti.com>
> > > > Cc: Benoit Cousson <b-cousson@ti.com>
> > > > Cc: Kevin Hilman <khilman@deeprootsystems.com>
> > > > Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>
> > > >
> > > > +/* dma_system -> L3 */
> > > > +static struct omap_hwmod_ocp_if omap2420_dma_system__l3 = {
> > > > +	.master		= &omap2420_dma_system_hwmod,
> > > > +	.slave		= &omap2420_l3_main_hwmod,
> > > > +	.clk		= "l3_div_ck",
> > > 
> > > This clock does not exist on OMAP2420.  Did you test this patch on 2420?
> > 
> > ok. will be replaced with "sdma_ick". I don't have 2420 for testing. I
> > tested it on 2430. My understanding is that, DMA clock interface is same
> > for 2420 and 2430. Correct me if am wrong.
> 
> Looking into omap3/omap4 structures, this entry takes core l3 clock.
> 
> This should be "sdma_fck" since it's parent is core_l3_ck.

The OMAP2xxx SDMA data can use either sdma_fck or core_l3_ck for its 
functional clock.  At some point we will probably remove sdma_fck from the 
2xxx data.  That's a relic of the pre-clkdev days, so you might as well 
just use core_l3_ck.  Maybe just stick with sdma_ick for the interface 
clock for right now.


- Paul

[PATCH v1 4/9] OMAP2430: hwmod data: add system DMA

[G, Manjunath Kondaiah]

On Wed, Dec 15, 2010 at 08:14:49AM -0700, Paul Walmsley wrote:
> On Wed, 15 Dec 2010, G, Manjunath Kondaiah wrote:
> 
> > On Tue, Dec 14, 2010 at 07:27:33PM -0700, Paul Walmsley wrote:
> > > On Sat, 4 Dec 2010, G, Manjunath Kondaiah wrote:
> > > 
> > > > Add OMAP2430 DMA hwmod data and also add required
> > > > DMA device attributes.
> > > 
> > > ...
> > > 
> > > > +/* dma_system -> L3 */
> > > > +static struct omap_hwmod_ocp_if omap2430_dma_system__l3 = {
> > > > +	.master		= &omap2430_dma_system_hwmod,
> > > > +	.slave		= &omap2430_l3_main_hwmod,
> > > > +	.clk		= "l3_div_ck",
> > > 
> > > This clock does not exist on OMAP2430.  Did you test this on OMAP2430?
> > 
> > Yes. I have tested this on SDP2430. For confirmation, I tested again now.
> > It boots up without any issues and all DMA test cases are passing.
> 
> Don't you see warnings in the boot messages that this clock cannot be 
> found?

No. I don't see any warnings in boot log. May missing during early
prints?

-Manjunath