[PATCH V3 0/3] Add hardware I/O coherency support for Armada 370/XP

[PATCH V3 0/3] Add hardware I/O coherency support for Armada 370/XP

[Gregory CLEMENT]

The purpose of this patch set is to add hardware I/O Coherency support
for Armada 370 and Armada XP. Theses SoCs come with an unit called
coherency fabric. A beginning of the support for this unit have been
introduced with the SMP patch set. This series extend this support:
the coherency fabric unit allows to use the Armada XP and the Armada
370 as nearly coherent architectures.

The third patches enables this new feature and register our own set
of DMA ops, to benefit this hardware enhancement.

The first patches exports dma operation functions needed by to
register our own set of dma ops.

The second patch introduces a new flag for the address decoding
configuration in order to be able to set the memory windows as
shared memory.

This series depend on the SMP patch set (V5 was posted few minutes
earlier)

The git branch called HWIOCC-for-3.8-V3 is also available at
https://github.com/MISL-EBU-System-SW/mainline-public.git.

Changelog:
V2 -> V3:
- Rebased on to ArmadaXP-SMP-for-3.8-V5
- Use the coherent version of dma ops for .alloc() and .free()

V1 -> V2:
- Rebased on to v3.7-rc5
- Added a new patch to exports the dma ops functions
- Renamed the function for a more generic name mvebu_hwcc
- removed the non SMP case during init
- spelling and wording issues
- updating the binding documentation for coherency fabric

Gregory CLEMENT (3):
  arm: dma mapping: Export dma ops functions
  arm: plat-orion: Add coherency attribute when setup mbus target
  arm: mvebu: Add hardware I/O Coherency support

 .../devicetree/bindings/arm/coherency-fabric.txt   |    9 ++-
 arch/arm/boot/dts/armada-370-xp.dtsi               |    3 +-
 arch/arm/include/asm/dma-mapping.h                 |   48 +++++++++++++
 arch/arm/mach-mvebu/addr-map.c                     |    3 +
 arch/arm/mach-mvebu/coherency.c                    |   73 ++++++++++++++++++++
 arch/arm/mm/dma-mapping.c                          |   25 ++-----
 arch/arm/plat-orion/addr-map.c                     |    4 ++
 arch/arm/plat-orion/include/plat/addr-map.h        |    1 +
 8 files changed, 143 insertions(+), 23 deletions(-)

-- 
1.7.9.5

[PATCH V3 1/3] arm: dma mapping: Export dma ops functions

[Gregory CLEMENT]

Expose the DMA operations functions. Until now only the dma_ops
structs in a whole or some dma operation were exposed. This patch
exposes all the dma coherents operations. They can be reused when an
architecture or a driver need to create its own set of dma_operation.

Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
---
 arch/arm/include/asm/dma-mapping.h |   48 ++++++++++++++++++++++++++++++++++++
 arch/arm/mm/dma-mapping.c          |   25 ++++---------------
 2 files changed, 53 insertions(+), 20 deletions(-)

diff --git a/arch/arm/include/asm/dma-mapping.h b/arch/arm/include/asm/dma-mapping.h
index 2300484..b12d7c0 100644
--- a/arch/arm/include/asm/dma-mapping.h
+++ b/arch/arm/include/asm/dma-mapping.h
@@ -112,6 +112,54 @@ static inline void dma_free_noncoherent(struct device *dev, size_t size,
 extern int dma_supported(struct device *dev, u64 mask);
 
 /**
+ * arm_dma_map_page - map a portion of a page for streaming DMA
+ * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
+ * @page: page that buffer resides in
+ * @offset: offset into page for start of buffer
+ * @size: size of buffer to map
+ * @dir: DMA transfer direction
+ *
+ * Ensure that any data held in the cache is appropriately discarded
+ * or written back.
+ *
+ * The device owns this memory once this call has completed.  The CPU
+ * can regain ownership by calling dma_unmap_page().
+ */
+extern dma_addr_t arm_dma_map_page(struct device *dev, struct page *page,
+				    unsigned long offset, size_t size,
+				    enum dma_data_direction dir,
+				    struct dma_attrs *attrs);
+
+/**
+ * arm_dma_unmap_page - unmap a buffer previously mapped through dma_map_page()
+ * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
+ * @handle: DMA address of buffer
+ * @size: size of buffer (same as passed to dma_map_page)
+ * @dir: DMA transfer direction (same as passed to dma_map_page)
+ *
+ * Unmap a page streaming mode DMA translation.  The handle and size
+ * must match what was provided in the previous dma_map_page() call.
+ * All other usages are undefined.
+ *
+ * After this call, reads by the CPU to the buffer are guaranteed to see
+ * whatever the device wrote there.
+ */
+extern void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
+				size_t size, enum dma_data_direction dir,
+				struct dma_attrs *attrs);
+
+extern void arm_dma_sync_single_for_cpu(struct device *dev,
+					dma_addr_t handle, size_t size,
+					enum dma_data_direction dir);
+
+extern void arm_dma_sync_single_for_device(struct device *dev,
+					dma_addr_t handle, size_t size,
+					enum dma_data_direction dir);
+
+extern int arm_dma_set_mask(struct device *dev, u64 dma_mask);
+
+
+/**
  * arm_dma_alloc - allocate consistent memory for DMA
  * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
  * @size: required memory size
diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
index 58bc3e4..dbb67ce 100644
--- a/arch/arm/mm/dma-mapping.c
+++ b/arch/arm/mm/dma-mapping.c
@@ -56,20 +56,13 @@ static void __dma_page_dev_to_cpu(struct page *, unsigned long,
 		size_t, enum dma_data_direction);
 
 /**
- * arm_dma_map_page - map a portion of a page for streaming DMA
- * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
- * @page: page that buffer resides in
- * @offset: offset into page for start of buffer
- * @size: size of buffer to map
- * @dir: DMA transfer direction
- *
  * Ensure that any data held in the cache is appropriately discarded
  * or written back.
  *
  * The device owns this memory once this call has completed.  The CPU
  * can regain ownership by calling dma_unmap_page().
  */
-static dma_addr_t arm_dma_map_page(struct device *dev, struct page *page,
+dma_addr_t arm_dma_map_page(struct device *dev, struct page *page,
 	     unsigned long offset, size_t size, enum dma_data_direction dir,
 	     struct dma_attrs *attrs)
 {
@@ -86,12 +79,6 @@ static dma_addr_t arm_coherent_dma_map_page(struct device *dev, struct page *pag
 }
 
 /**
- * arm_dma_unmap_page - unmap a buffer previously mapped through dma_map_page()
- * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
- * @handle: DMA address of buffer
- * @size: size of buffer (same as passed to dma_map_page)
- * @dir: DMA transfer direction (same as passed to dma_map_page)
- *
  * Unmap a page streaming mode DMA translation.  The handle and size
  * must match what was provided in the previous dma_map_page() call.
  * All other usages are undefined.
@@ -99,7 +86,7 @@ static dma_addr_t arm_coherent_dma_map_page(struct device *dev, struct page *pag
  * After this call, reads by the CPU to the buffer are guaranteed to see
  * whatever the device wrote there.
  */
-static void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
+void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
 		size_t size, enum dma_data_direction dir,
 		struct dma_attrs *attrs)
 {
@@ -108,7 +95,7 @@ static void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
 				      handle & ~PAGE_MASK, size, dir);
 }
 
-static void arm_dma_sync_single_for_cpu(struct device *dev,
+void arm_dma_sync_single_for_cpu(struct device *dev,
 		dma_addr_t handle, size_t size, enum dma_data_direction dir)
 {
 	unsigned int offset = handle & (PAGE_SIZE - 1);
@@ -116,7 +103,7 @@ static void arm_dma_sync_single_for_cpu(struct device *dev,
 	__dma_page_dev_to_cpu(page, offset, size, dir);
 }
 
-static void arm_dma_sync_single_for_device(struct device *dev,
+void arm_dma_sync_single_for_device(struct device *dev,
 		dma_addr_t handle, size_t size, enum dma_data_direction dir)
 {
 	unsigned int offset = handle & (PAGE_SIZE - 1);
@@ -124,8 +111,6 @@ static void arm_dma_sync_single_for_device(struct device *dev,
 	__dma_page_cpu_to_dev(page, offset, size, dir);
 }
 
-static int arm_dma_set_mask(struct device *dev, u64 dma_mask);
-
 struct dma_map_ops arm_dma_ops = {
 	.alloc			= arm_dma_alloc,
 	.free			= arm_dma_free,
@@ -971,7 +956,7 @@ int dma_supported(struct device *dev, u64 mask)
 }
 EXPORT_SYMBOL(dma_supported);
 
-static int arm_dma_set_mask(struct device *dev, u64 dma_mask)
+int arm_dma_set_mask(struct device *dev, u64 dma_mask)
 {
 	if (!dev->dma_mask || !dma_supported(dev, dma_mask))
 		return -EIO;
-- 
1.7.9.5

[PATCH V3 1/3] arm: dma mapping: Export dma ops functions

[Marek Szyprowski]

Hello,

On 11/20/2012 10:56 PM, Gregory CLEMENT wrote:
> Expose the DMA operations functions. Until now only the dma_ops
> structs in a whole or some dma operation were exposed. This patch
> exposes all the dma coherents operations. They can be reused when an
> architecture or a driver need to create its own set of dma_operation.
>
> Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>

Besides arm_dma_set_mask() function I see no reason to export the other
arm dma related functions.

> ---
>   arch/arm/include/asm/dma-mapping.h |   48 ++++++++++++++++++++++++++++++++++++
>   arch/arm/mm/dma-mapping.c          |   25 ++++---------------
>   2 files changed, 53 insertions(+), 20 deletions(-)
>
> diff --git a/arch/arm/include/asm/dma-mapping.h b/arch/arm/include/asm/dma-mapping.h
> index 2300484..b12d7c0 100644
> --- a/arch/arm/include/asm/dma-mapping.h
> +++ b/arch/arm/include/asm/dma-mapping.h
> @@ -112,6 +112,54 @@ static inline void dma_free_noncoherent(struct device *dev, size_t size,
>   extern int dma_supported(struct device *dev, u64 mask);
>   
>   /**
> + * arm_dma_map_page - map a portion of a page for streaming DMA
> + * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
> + * @page: page that buffer resides in
> + * @offset: offset into page for start of buffer
> + * @size: size of buffer to map
> + * @dir: DMA transfer direction
> + *
> + * Ensure that any data held in the cache is appropriately discarded
> + * or written back.
> + *
> + * The device owns this memory once this call has completed.  The CPU
> + * can regain ownership by calling dma_unmap_page().
> + */
> +extern dma_addr_t arm_dma_map_page(struct device *dev, struct page *page,
> +				    unsigned long offset, size_t size,
> +				    enum dma_data_direction dir,
> +				    struct dma_attrs *attrs);
> +
> +/**
> + * arm_dma_unmap_page - unmap a buffer previously mapped through dma_map_page()
> + * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
> + * @handle: DMA address of buffer
> + * @size: size of buffer (same as passed to dma_map_page)
> + * @dir: DMA transfer direction (same as passed to dma_map_page)
> + *
> + * Unmap a page streaming mode DMA translation.  The handle and size
> + * must match what was provided in the previous dma_map_page() call.
> + * All other usages are undefined.
> + *
> + * After this call, reads by the CPU to the buffer are guaranteed to see
> + * whatever the device wrote there.
> + */
> +extern void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
> +				size_t size, enum dma_data_direction dir,
> +				struct dma_attrs *attrs);
> +
> +extern void arm_dma_sync_single_for_cpu(struct device *dev,
> +					dma_addr_t handle, size_t size,
> +					enum dma_data_direction dir);
> +
> +extern void arm_dma_sync_single_for_device(struct device *dev,
> +					dma_addr_t handle, size_t size,
> +					enum dma_data_direction dir);
> +
> +extern int arm_dma_set_mask(struct device *dev, u64 dma_mask);
> +
> +
> +/**
>    * arm_dma_alloc - allocate consistent memory for DMA
>    * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
>    * @size: required memory size
> diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
> index 58bc3e4..dbb67ce 100644
> --- a/arch/arm/mm/dma-mapping.c
> +++ b/arch/arm/mm/dma-mapping.c
> @@ -56,20 +56,13 @@ static void __dma_page_dev_to_cpu(struct page *, unsigned long,
>   		size_t, enum dma_data_direction);
>   
>   /**
> - * arm_dma_map_page - map a portion of a page for streaming DMA
> - * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
> - * @page: page that buffer resides in
> - * @offset: offset into page for start of buffer
> - * @size: size of buffer to map
> - * @dir: DMA transfer direction
> - *
>    * Ensure that any data held in the cache is appropriately discarded
>    * or written back.
>    *
>    * The device owns this memory once this call has completed.  The CPU
>    * can regain ownership by calling dma_unmap_page().
>    */
> -static dma_addr_t arm_dma_map_page(struct device *dev, struct page *page,
> +dma_addr_t arm_dma_map_page(struct device *dev, struct page *page,
>   	     unsigned long offset, size_t size, enum dma_data_direction dir,
>   	     struct dma_attrs *attrs)
>   {
> @@ -86,12 +79,6 @@ static dma_addr_t arm_coherent_dma_map_page(struct device *dev, struct page *pag
>   }
>   
>   /**
> - * arm_dma_unmap_page - unmap a buffer previously mapped through dma_map_page()
> - * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
> - * @handle: DMA address of buffer
> - * @size: size of buffer (same as passed to dma_map_page)
> - * @dir: DMA transfer direction (same as passed to dma_map_page)
> - *
>    * Unmap a page streaming mode DMA translation.  The handle and size
>    * must match what was provided in the previous dma_map_page() call.
>    * All other usages are undefined.
> @@ -99,7 +86,7 @@ static dma_addr_t arm_coherent_dma_map_page(struct device *dev, struct page *pag
>    * After this call, reads by the CPU to the buffer are guaranteed to see
>    * whatever the device wrote there.
>    */
> -static void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
> +void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
>   		size_t size, enum dma_data_direction dir,
>   		struct dma_attrs *attrs)
>   {
> @@ -108,7 +95,7 @@ static void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
>   				      handle & ~PAGE_MASK, size, dir);
>   }
>   
> -static void arm_dma_sync_single_for_cpu(struct device *dev,
> +void arm_dma_sync_single_for_cpu(struct device *dev,
>   		dma_addr_t handle, size_t size, enum dma_data_direction dir)
>   {
>   	unsigned int offset = handle & (PAGE_SIZE - 1);
> @@ -116,7 +103,7 @@ static void arm_dma_sync_single_for_cpu(struct device *dev,
>   	__dma_page_dev_to_cpu(page, offset, size, dir);
>   }
>   
> -static void arm_dma_sync_single_for_device(struct device *dev,
> +void arm_dma_sync_single_for_device(struct device *dev,
>   		dma_addr_t handle, size_t size, enum dma_data_direction dir)
>   {
>   	unsigned int offset = handle & (PAGE_SIZE - 1);
> @@ -124,8 +111,6 @@ static void arm_dma_sync_single_for_device(struct device *dev,
>   	__dma_page_cpu_to_dev(page, offset, size, dir);
>   }
>   
> -static int arm_dma_set_mask(struct device *dev, u64 dma_mask);
> -
>   struct dma_map_ops arm_dma_ops = {
>   	.alloc			= arm_dma_alloc,
>   	.free			= arm_dma_free,
> @@ -971,7 +956,7 @@ int dma_supported(struct device *dev, u64 mask)
>   }
>   EXPORT_SYMBOL(dma_supported);
>   
> -static int arm_dma_set_mask(struct device *dev, u64 dma_mask)
> +int arm_dma_set_mask(struct device *dev, u64 dma_mask)
>   {
>   	if (!dev->dma_mask || !dma_supported(dev, dma_mask))
>   		return -EIO;

Best regards
-- 
Marek Szyprowski
Samsung Poland R&D Center

[PATCH V3 1/3] arm: dma mapping: Export dma ops functions

[Gregory CLEMENT]

On 11/21/2012 08:06 AM, Marek Szyprowski wrote:
> Hello,
Hello,

> 
> On 11/20/2012 10:56 PM, Gregory CLEMENT wrote:
>> Expose the DMA operations functions. Until now only the dma_ops
>> structs in a whole or some dma operation were exposed. This patch
>> exposes all the dma coherents operations. They can be reused when an
>> architecture or a driver need to create its own set of dma_operation.
>>
>> Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
> 
> Besides arm_dma_set_mask() function I see no reason to export the other
> arm dma related functions.

The idea was to let other people use the arm dma related functions,
for their own dma ops. But for the mvebu machines we only need
arm_dma_set_mask() indeed.

So you prefer that I only expose arm_dma_set_mask() and let future
user expose other function if they need it, right?

Thanks,
Gregory
> 
>> ---
>>   arch/arm/include/asm/dma-mapping.h |   48 ++++++++++++++++++++++++++++++++++++
>>   arch/arm/mm/dma-mapping.c          |   25 ++++---------------
>>   2 files changed, 53 insertions(+), 20 deletions(-)
>>
>> diff --git a/arch/arm/include/asm/dma-mapping.h b/arch/arm/include/asm/dma-mapping.h
>> index 2300484..b12d7c0 100644
>> --- a/arch/arm/include/asm/dma-mapping.h
>> +++ b/arch/arm/include/asm/dma-mapping.h
>> @@ -112,6 +112,54 @@ static inline void dma_free_noncoherent(struct device *dev, size_t size,
>>   extern int dma_supported(struct device *dev, u64 mask);
>>   
>>   /**
>> + * arm_dma_map_page - map a portion of a page for streaming DMA
>> + * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
>> + * @page: page that buffer resides in
>> + * @offset: offset into page for start of buffer
>> + * @size: size of buffer to map
>> + * @dir: DMA transfer direction
>> + *
>> + * Ensure that any data held in the cache is appropriately discarded
>> + * or written back.
>> + *
>> + * The device owns this memory once this call has completed.  The CPU
>> + * can regain ownership by calling dma_unmap_page().
>> + */
>> +extern dma_addr_t arm_dma_map_page(struct device *dev, struct page *page,
>> +				    unsigned long offset, size_t size,
>> +				    enum dma_data_direction dir,
>> +				    struct dma_attrs *attrs);
>> +
>> +/**
>> + * arm_dma_unmap_page - unmap a buffer previously mapped through dma_map_page()
>> + * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
>> + * @handle: DMA address of buffer
>> + * @size: size of buffer (same as passed to dma_map_page)
>> + * @dir: DMA transfer direction (same as passed to dma_map_page)
>> + *
>> + * Unmap a page streaming mode DMA translation.  The handle and size
>> + * must match what was provided in the previous dma_map_page() call.
>> + * All other usages are undefined.
>> + *
>> + * After this call, reads by the CPU to the buffer are guaranteed to see
>> + * whatever the device wrote there.
>> + */
>> +extern void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
>> +				size_t size, enum dma_data_direction dir,
>> +				struct dma_attrs *attrs);
>> +
>> +extern void arm_dma_sync_single_for_cpu(struct device *dev,
>> +					dma_addr_t handle, size_t size,
>> +					enum dma_data_direction dir);
>> +
>> +extern void arm_dma_sync_single_for_device(struct device *dev,
>> +					dma_addr_t handle, size_t size,
>> +					enum dma_data_direction dir);
>> +
>> +extern int arm_dma_set_mask(struct device *dev, u64 dma_mask);
>> +
>> +
>> +/**
>>    * arm_dma_alloc - allocate consistent memory for DMA
>>    * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
>>    * @size: required memory size
>> diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
>> index 58bc3e4..dbb67ce 100644
>> --- a/arch/arm/mm/dma-mapping.c
>> +++ b/arch/arm/mm/dma-mapping.c
>> @@ -56,20 +56,13 @@ static void __dma_page_dev_to_cpu(struct page *, unsigned long,
>>   		size_t, enum dma_data_direction);
>>   
>>   /**
>> - * arm_dma_map_page - map a portion of a page for streaming DMA
>> - * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
>> - * @page: page that buffer resides in
>> - * @offset: offset into page for start of buffer
>> - * @size: size of buffer to map
>> - * @dir: DMA transfer direction
>> - *
>>    * Ensure that any data held in the cache is appropriately discarded
>>    * or written back.
>>    *
>>    * The device owns this memory once this call has completed.  The CPU
>>    * can regain ownership by calling dma_unmap_page().
>>    */
>> -static dma_addr_t arm_dma_map_page(struct device *dev, struct page *page,
>> +dma_addr_t arm_dma_map_page(struct device *dev, struct page *page,
>>   	     unsigned long offset, size_t size, enum dma_data_direction dir,
>>   	     struct dma_attrs *attrs)
>>   {
>> @@ -86,12 +79,6 @@ static dma_addr_t arm_coherent_dma_map_page(struct device *dev, struct page *pag
>>   }
>>   
>>   /**
>> - * arm_dma_unmap_page - unmap a buffer previously mapped through dma_map_page()
>> - * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
>> - * @handle: DMA address of buffer
>> - * @size: size of buffer (same as passed to dma_map_page)
>> - * @dir: DMA transfer direction (same as passed to dma_map_page)
>> - *
>>    * Unmap a page streaming mode DMA translation.  The handle and size
>>    * must match what was provided in the previous dma_map_page() call.
>>    * All other usages are undefined.
>> @@ -99,7 +86,7 @@ static dma_addr_t arm_coherent_dma_map_page(struct device *dev, struct page *pag
>>    * After this call, reads by the CPU to the buffer are guaranteed to see
>>    * whatever the device wrote there.
>>    */
>> -static void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
>> +void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
>>   		size_t size, enum dma_data_direction dir,
>>   		struct dma_attrs *attrs)
>>   {
>> @@ -108,7 +95,7 @@ static void arm_dma_unmap_page(struct device *dev, dma_addr_t handle,
>>   				      handle & ~PAGE_MASK, size, dir);
>>   }
>>   
>> -static void arm_dma_sync_single_for_cpu(struct device *dev,
>> +void arm_dma_sync_single_for_cpu(struct device *dev,
>>   		dma_addr_t handle, size_t size, enum dma_data_direction dir)
>>   {
>>   	unsigned int offset = handle & (PAGE_SIZE - 1);
>> @@ -116,7 +103,7 @@ static void arm_dma_sync_single_for_cpu(struct device *dev,
>>   	__dma_page_dev_to_cpu(page, offset, size, dir);
>>   }
>>   
>> -static void arm_dma_sync_single_for_device(struct device *dev,
>> +void arm_dma_sync_single_for_device(struct device *dev,
>>   		dma_addr_t handle, size_t size, enum dma_data_direction dir)
>>   {
>>   	unsigned int offset = handle & (PAGE_SIZE - 1);
>> @@ -124,8 +111,6 @@ static void arm_dma_sync_single_for_device(struct device *dev,
>>   	__dma_page_cpu_to_dev(page, offset, size, dir);
>>   }
>>   
>> -static int arm_dma_set_mask(struct device *dev, u64 dma_mask);
>> -
>>   struct dma_map_ops arm_dma_ops = {
>>   	.alloc			= arm_dma_alloc,
>>   	.free			= arm_dma_free,
>> @@ -971,7 +956,7 @@ int dma_supported(struct device *dev, u64 mask)
>>   }
>>   EXPORT_SYMBOL(dma_supported);
>>   
>> -static int arm_dma_set_mask(struct device *dev, u64 dma_mask)
>> +int arm_dma_set_mask(struct device *dev, u64 dma_mask)
>>   {
>>   	if (!dev->dma_mask || !dma_supported(dev, dma_mask))
>>   		return -EIO;
> 
> Best regards
> 


-- 
Gregory Clement, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[PATCH V3 1/3] arm: dma mapping: Export dma ops functions

[Marek Szyprowski]

Hello,

On 11/21/2012 9:05 AM, Gregory CLEMENT wrote:
> On 11/21/2012 08:06 AM, Marek Szyprowski wrote:
> > On 11/20/2012 10:56 PM, Gregory CLEMENT wrote:
> >> Expose the DMA operations functions. Until now only the dma_ops
> >> structs in a whole or some dma operation were exposed. This patch
> >> exposes all the dma coherents operations. They can be reused when an
> >> architecture or a driver need to create its own set of dma_operation.
> >>
> >> Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
> >
> > Besides arm_dma_set_mask() function I see no reason to export the other
> > arm dma related functions.
>
> The idea was to let other people use the arm dma related functions,
> for their own dma ops. But for the mvebu machines we only need
> arm_dma_set_mask() indeed.
>
> So you prefer that I only expose arm_dma_set_mask() and let future
> user expose other function if they need it, right?

I would prefer to avoid exporting functions which are not used anywhere
else. This improves readability of the code and simply forces others to
think twice before they use some static function and check if their use
case is really correct.

Best regards
-- 
Marek Szyprowski
Samsung Poland R&D Center

[PATCH V3 1/3] arm: dma mapping: Export dma ops functions

[Gregory CLEMENT]

On 11/21/2012 09:12 AM, Marek Szyprowski wrote:
> Hello,
> 
> On 11/21/2012 9:05 AM, Gregory CLEMENT wrote:
>> On 11/21/2012 08:06 AM, Marek Szyprowski wrote:
>>> On 11/20/2012 10:56 PM, Gregory CLEMENT wrote:
>>>> Expose the DMA operations functions. Until now only the dma_ops
>>>> structs in a whole or some dma operation were exposed. This patch
>>>> exposes all the dma coherents operations. They can be reused when an
>>>> architecture or a driver need to create its own set of dma_operation.
>>>>
>>>> Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
>>>
>>> Besides arm_dma_set_mask() function I see no reason to export the other
>>> arm dma related functions.
>>
>> The idea was to let other people use the arm dma related functions,
>> for their own dma ops. But for the mvebu machines we only need
>> arm_dma_set_mask() indeed.
>>
>> So you prefer that I only expose arm_dma_set_mask() and let future
>> user expose other function if they need it, right?
> 
> I would prefer to avoid exporting functions which are not used anywhere
> else. This improves readability of the code and simply forces others to
> think twice before they use some static function and check if their use
> case is really correct.

OK so I will send a new version in a few minutes.

Do  you think these last changes will be enough for getting your Acked-by?

Thanks,
Greogry

[PATCH V3 1/3] arm: dma mapping: Export dma ops functions

[Marek Szyprowski]

Hello,

On 11/21/2012 9:17 AM, Gregory CLEMENT wrote:
> On 11/21/2012 09:12 AM, Marek Szyprowski wrote:
> > Hello,
> >
> > On 11/21/2012 9:05 AM, Gregory CLEMENT wrote:
> >> On 11/21/2012 08:06 AM, Marek Szyprowski wrote:
> >>> On 11/20/2012 10:56 PM, Gregory CLEMENT wrote:
> >>>> Expose the DMA operations functions. Until now only the dma_ops
> >>>> structs in a whole or some dma operation were exposed. This patch
> >>>> exposes all the dma coherents operations. They can be reused when an
> >>>> architecture or a driver need to create its own set of dma_operation.
> >>>>
> >>>> Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
> >>>
> >>> Besides arm_dma_set_mask() function I see no reason to export the other
> >>> arm dma related functions.
> >>
> >> The idea was to let other people use the arm dma related functions,
> >> for their own dma ops. But for the mvebu machines we only need
> >> arm_dma_set_mask() indeed.
> >>
> >> So you prefer that I only expose arm_dma_set_mask() and let future
> >> user expose other function if they need it, right?
> >
> > I would prefer to avoid exporting functions which are not used anywhere
> > else. This improves readability of the code and simply forces others to
> > think twice before they use some static function and check if their use
> > case is really correct.
>
> OK so I will send a new version in a few minutes.
>
> Do  you think these last changes will be enough for getting your Acked-by?

Yes, I see no other issues right now.

Best regards
-- 
Marek Szyprowski
Samsung Poland R&D Center

[PATCH V3 2/3] arm: plat-orion: Add coherency attribute when setup mbus target

[Gregory CLEMENT]

Recent SoC such as Armada 370/XP came with the possibility to deal
with the I/O coherency by hardware. In this case the transaction
attribute of the window must be flagged as "Shared transaction". Once
this flag is set, then the transactions will be forced to be sent
through the coherency block, in other case transaction is driven
directly to DRAM.

Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Reviewed-by: Yehuda Yitschak <yehuday@marvell.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 arch/arm/plat-orion/addr-map.c              |    4 ++++
 arch/arm/plat-orion/include/plat/addr-map.h |    1 +
 2 files changed, 5 insertions(+)

diff --git a/arch/arm/plat-orion/addr-map.c b/arch/arm/plat-orion/addr-map.c
index a7b8060..febe386 100644
--- a/arch/arm/plat-orion/addr-map.c
+++ b/arch/arm/plat-orion/addr-map.c
@@ -42,6 +42,8 @@ EXPORT_SYMBOL_GPL(mv_mbus_dram_info);
 #define WIN_REMAP_LO_OFF	0x0008
 #define WIN_REMAP_HI_OFF	0x000c
 
+#define ATTR_HW_COHERENCY	(0x1 << 4)
+
 /*
  * Default implementation
  */
@@ -163,6 +165,8 @@ void __init orion_setup_cpu_mbus_target(const struct orion_addr_map_cfg *cfg,
 			w = &orion_mbus_dram_info.cs[cs++];
 			w->cs_index = i;
 			w->mbus_attr = 0xf & ~(1 << i);
+			if (cfg->hw_io_coherency)
+				w->mbus_attr |= ATTR_HW_COHERENCY;
 			w->base = base & 0xffff0000;
 			w->size = (size | 0x0000ffff) + 1;
 		}
diff --git a/arch/arm/plat-orion/include/plat/addr-map.h b/arch/arm/plat-orion/include/plat/addr-map.h
index ec63e4a..b76c065 100644
--- a/arch/arm/plat-orion/include/plat/addr-map.h
+++ b/arch/arm/plat-orion/include/plat/addr-map.h
@@ -17,6 +17,7 @@ struct orion_addr_map_cfg {
 	const int num_wins;	/* Total number of windows */
 	const int remappable_wins;
 	void __iomem *bridge_virt_base;
+	int hw_io_coherency;
 
 	/* If NULL, the default cpu_win_can_remap will be used, using
 	   the value in remappable_wins */
-- 
1.7.9.5

[PATCH V3 3/3] arm: mvebu: Add hardware I/O Coherency support

[Gregory CLEMENT]

Armada 370 and XP come with an unit called coherency fabric. This unit
allows to use the Armada 370/XP as a nearly coherent architecture. The
coherency mechanism uses snoop filters to ensure the coherency between
caches, DRAM and devices. This mechanism needs a synchronization
barrier which guarantees that all the memory writes initiated by the
devices have reached their target and do not reside in intermediate
write buffers. That's why the architecture is not totally coherent and
we need to provide our own functions for some DMA operations.

Beside the use of the coherency fabric, the device units will have to
set the attribute flag of the decoding address window to select the
accurate coherency process for the memory transaction. This is done
each device driver programs the DRAM address windows. The value of the
attribute set by the driver is retrieved through the
orion_addr_map_cfg struct filled during the early initialization of
the platform.

Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Reviewed-by: Yehuda Yitschak <yehuday@marvell.com>
---
 .../devicetree/bindings/arm/coherency-fabric.txt   |    9 ++-
 arch/arm/boot/dts/armada-370-xp.dtsi               |    3 +-
 arch/arm/mach-mvebu/addr-map.c                     |    3 +
 arch/arm/mach-mvebu/coherency.c                    |   73 ++++++++++++++++++++
 4 files changed, 85 insertions(+), 3 deletions(-)

diff --git a/Documentation/devicetree/bindings/arm/coherency-fabric.txt b/Documentation/devicetree/bindings/arm/coherency-fabric.txt
index 2bfbf67..17d8cd1 100644
--- a/Documentation/devicetree/bindings/arm/coherency-fabric.txt
+++ b/Documentation/devicetree/bindings/arm/coherency-fabric.txt
@@ -5,12 +5,17 @@ Available on Marvell SOCs: Armada 370 and Armada XP
 Required properties:
 
 - compatible: "marvell,coherency-fabric"
-- reg: Should contain,coherency fabric registers location and length.
+
+- reg: Should contain coherency fabric registers location and
+  length. First pair for the coherency fabric registers, second pair
+  for the per-CPU fabric registers registers.
 
 Example:
 
 coherency-fabric at d0020200 {
 	compatible = "marvell,coherency-fabric";
-	reg = <0xd0020200 0xb0>;
+	reg = <0xd0020200 0xb0>,
+		<0xd0021810 0x1c>;
+
 };
 
diff --git a/arch/arm/boot/dts/armada-370-xp.dtsi b/arch/arm/boot/dts/armada-370-xp.dtsi
index b0d075b..98a6b26 100644
--- a/arch/arm/boot/dts/armada-370-xp.dtsi
+++ b/arch/arm/boot/dts/armada-370-xp.dtsi
@@ -38,7 +38,8 @@
 
 	coherency-fabric at d0020200 {
 		compatible = "marvell,coherency-fabric";
-		reg = <0xd0020200 0xb0>;
+		reg = <0xd0020200 0xb0>,
+		      <0xd0021810 0x1c>;
 	};
 
 	soc {
diff --git a/arch/arm/mach-mvebu/addr-map.c b/arch/arm/mach-mvebu/addr-map.c
index fe454a4..595f6b7 100644
--- a/arch/arm/mach-mvebu/addr-map.c
+++ b/arch/arm/mach-mvebu/addr-map.c
@@ -108,6 +108,9 @@ static int __init armada_setup_cpu_mbus(void)
 
 	addr_map_cfg.bridge_virt_base = mbus_unit_addr_decoding_base;
 
+	if (of_find_compatible_node(NULL, NULL, "marvell,coherency-fabric"))
+		addr_map_cfg.hw_io_coherency = 1;
+
 	/*
 	 * Disable, clear and configure windows.
 	 */
diff --git a/arch/arm/mach-mvebu/coherency.c b/arch/arm/mach-mvebu/coherency.c
index 1bc02d0..9413bd5 100644
--- a/arch/arm/mach-mvebu/coherency.c
+++ b/arch/arm/mach-mvebu/coherency.c
@@ -22,6 +22,8 @@
 #include <linux/of_address.h>
 #include <linux/io.h>
 #include <linux/smp.h>
+#include <linux/dma-mapping.h>
+#include <linux/platform_device.h>
 #include <asm/smp_plat.h>
 #include "armada-370-xp.h"
 
@@ -32,10 +34,13 @@
  * value matching its virtual mapping
  */
 static void __iomem *coherency_base = ARMADA_370_XP_REGS_VIRT_BASE + 0x20200;
+static void __iomem *coherency_cpu_base;
 
 /* Coherency fabric registers */
 #define COHERENCY_FABRIC_CFG_OFFSET		   0x4
 
+#define IO_SYNC_BARRIER_CTL_OFFSET		   0x0
+
 static struct of_device_id of_coherency_table[] = {
 	{.compatible = "marvell,coherency-fabric"},
 	{ /* end of list */ },
@@ -66,6 +71,70 @@ int set_cpu_coherent(unsigned int hw_cpu_id, int smp_group_id)
 	return 0;
 }
 
+static inline void mvebu_hwcc_sync_io_barrier(void)
+{
+	writel(0x1, coherency_cpu_base + IO_SYNC_BARRIER_CTL_OFFSET);
+	while (readl(coherency_cpu_base + IO_SYNC_BARRIER_CTL_OFFSET) & 0x1);
+}
+
+static dma_addr_t mvebu_hwcc_dma_map_page(struct device *dev, struct page *page,
+				  unsigned long offset, size_t size,
+				  enum dma_data_direction dir,
+				  struct dma_attrs *attrs)
+{
+	if (dir != DMA_TO_DEVICE)
+		mvebu_hwcc_sync_io_barrier();
+	return pfn_to_dma(dev, page_to_pfn(page)) + offset;
+}
+
+
+static void mvebu_hwcc_dma_unmap_page(struct device *dev, dma_addr_t dma_handle,
+			      size_t size, enum dma_data_direction dir,
+			      struct dma_attrs *attrs)
+{
+	if (dir != DMA_TO_DEVICE)
+		mvebu_hwcc_sync_io_barrier();
+}
+
+static void mvebu_hwcc_dma_sync(struct device *dev, dma_addr_t dma_handle,
+			size_t size, enum dma_data_direction dir)
+{
+	if (dir != DMA_TO_DEVICE)
+		mvebu_hwcc_sync_io_barrier();
+}
+
+static struct dma_map_ops mvebu_hwcc_dma_ops = {
+	.alloc			= arm_dma_alloc,
+	.free			= arm_dma_free,
+	.mmap			= arm_dma_mmap,
+	.map_page		= mvebu_hwcc_dma_map_page,
+	.unmap_page		= mvebu_hwcc_dma_unmap_page,
+	.get_sgtable		= arm_dma_get_sgtable,
+	.map_sg			= arm_dma_map_sg,
+	.unmap_sg		= arm_dma_unmap_sg,
+	.sync_single_for_cpu	= mvebu_hwcc_dma_sync,
+	.sync_single_for_device	= mvebu_hwcc_dma_sync,
+	.sync_sg_for_cpu	= arm_dma_sync_sg_for_cpu,
+	.sync_sg_for_device	= arm_dma_sync_sg_for_device,
+	.set_dma_mask		= arm_dma_set_mask,
+};
+
+static int mvebu_hwcc_platform_notifier(struct notifier_block *nb,
+				       unsigned long event, void *__dev)
+{
+	struct device *dev = __dev;
+
+	if (event != BUS_NOTIFY_ADD_DEVICE)
+		return NOTIFY_DONE;
+	set_dma_ops(dev, &mvebu_hwcc_dma_ops);
+
+	return NOTIFY_OK;
+}
+
+static struct notifier_block mvebu_hwcc_platform_nb = {
+	.notifier_call = mvebu_hwcc_platform_notifier,
+};
+
 int __init coherency_init(void)
 {
 	struct device_node *np;
@@ -74,6 +143,10 @@ int __init coherency_init(void)
 	if (np) {
 		pr_info("Initializing Coherency fabric\n");
 		coherency_base = of_iomap(np, 0);
+		coherency_cpu_base = of_iomap(np, 1);
+		set_cpu_coherent(cpu_logical_map(smp_processor_id()), 0);
+		bus_register_notifier(&platform_bus_type,
+					&mvebu_hwcc_platform_nb);
 	}
 
 	return 0;
-- 
1.7.9.5