refactor dma_cache_sync V2

refactor dma_cache_sync V2

[Christoph Hellwig]

The dma_cache_sync routines is used to flush caches for memory returned
by dma_alloc_attrs with the DMA_ATTR_NON_CONSISTENT flag (or previously
from dma_alloc_noncoherent), but the requirements for it seems to be
frequently misunderstood.  dma_cache_sync is documented to be a no-op for
allocations that do not have the DMA_ATTR_NON_CONSISTENT flag set, and
yet a lot of architectures implement it in some way despite not
implementing DMA_ATTR_NON_CONSISTENT.

This series removes a few abuses of dma_cache_sync for non-DMA API
purposes, then changes all remaining architectures that do not implement
DMA_ATTR_NON_CONSISTENT to implement dma_cache_sync as a no-op, and
then adds the struct dma_map_ops indirection we use for all other
DMA mapping operations to dma_cache_sync as well, thus removing all but
two implementations of the function.

Changes since V1:
 - drop the mips fd_cacheflush, merged via maintainer
 - spelling fix in last commit descriptions (thanks Geert)

refactor dma_cache_sync V2

[Christoph Hellwig]

The dma_cache_sync routines is used to flush caches for memory returned
by dma_alloc_attrs with the DMA_ATTR_NON_CONSISTENT flag (or previously
from dma_alloc_noncoherent), but the requirements for it seems to be
frequently misunderstood.  dma_cache_sync is documented to be a no-op for
allocations that do not have the DMA_ATTR_NON_CONSISTENT flag set, and
yet a lot of architectures implement it in some way despite not
implementing DMA_ATTR_NON_CONSISTENT.

This series removes a few abuses of dma_cache_sync for non-DMA API
purposes, then changes all remaining architectures that do not implement
DMA_ATTR_NON_CONSISTENT to implement dma_cache_sync as a no-op, and
then adds the struct dma_map_ops indirection we use for all other
DMA mapping operations to dma_cache_sync as well, thus removing all but
two implementations of the function.

Changes since V1:
 - drop the mips fd_cacheflush, merged via maintainer
 - spelling fix in last commit descriptions (thanks Geert)

refactor dma_cache_sync V2

[Christoph Hellwig]

The dma_cache_sync routines is used to flush caches for memory returned
by dma_alloc_attrs with the DMA_ATTR_NON_CONSISTENT flag (or previously
from dma_alloc_noncoherent), but the requirements for it seems to be
frequently misunderstood.  dma_cache_sync is documented to be a no-op for
allocations that do not have the DMA_ATTR_NON_CONSISTENT flag set, and
yet a lot of architectures implement it in some way despite not
implementing DMA_ATTR_NON_CONSISTENT.

This series removes a few abuses of dma_cache_sync for non-DMA API
purposes, then changes all remaining architectures that do not implement
DMA_ATTR_NON_CONSISTENT to implement dma_cache_sync as a no-op, and
then adds the struct dma_map_ops indirection we use for all other
DMA mapping operations to dma_cache_sync as well, thus removing all but
two implementations of the function.

Changes since V1:
 - drop the mips fd_cacheflush, merged via maintainer
 - spelling fix in last commit descriptions (thanks Geert)

[PATCH 01/11] floppy: consolidate the dummy fd_cacheflush definition

[Christoph Hellwig]

Only mips defines this helper, so remove all the other arch definitions.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/alpha/include/asm/floppy.h    | 2 --
 arch/powerpc/include/asm/floppy.h  | 2 --
 arch/sparc/include/asm/floppy_32.h | 1 -
 arch/sparc/include/asm/floppy_64.h | 1 -
 drivers/block/floppy.c             | 4 ++++
 5 files changed, 4 insertions(+), 6 deletions(-)

diff --git a/arch/alpha/include/asm/floppy.h b/arch/alpha/include/asm/floppy.h
index bae97eb19d26..942924756cf2 100644
--- a/arch/alpha/include/asm/floppy.h
+++ b/arch/alpha/include/asm/floppy.h
@@ -24,7 +24,6 @@
 #define fd_set_dma_count(count) set_dma_count(FLOPPY_DMA,count)
 #define fd_enable_irq()         enable_irq(FLOPPY_IRQ)
 #define fd_disable_irq()        disable_irq(FLOPPY_IRQ)
-#define fd_cacheflush(addr,size) /* nothing */
 #define fd_request_irq()        request_irq(FLOPPY_IRQ, floppy_interrupt,\
 					    0, "floppy", NULL)
 #define fd_free_irq()           free_irq(FLOPPY_IRQ, NULL)
@@ -62,7 +61,6 @@ alpha_fd_dma_setup(char *addr, unsigned long size, int mode, int io)
 	prev_dir = dir;
 
 	fd_clear_dma_ff();
-	fd_cacheflush(addr, size);
 	fd_set_dma_mode(mode);
 	set_dma_addr(FLOPPY_DMA, bus_addr);
 	fd_set_dma_count(size);
diff --git a/arch/powerpc/include/asm/floppy.h b/arch/powerpc/include/asm/floppy.h
index 936a904ae78c..167c44b58848 100644
--- a/arch/powerpc/include/asm/floppy.h
+++ b/arch/powerpc/include/asm/floppy.h
@@ -25,7 +25,6 @@
 #define fd_get_dma_residue()    fd_ops->_get_dma_residue(FLOPPY_DMA)
 #define fd_enable_irq()         enable_irq(FLOPPY_IRQ)
 #define fd_disable_irq()        disable_irq(FLOPPY_IRQ)
-#define fd_cacheflush(addr,size) /* nothing */
 #define fd_free_irq()           free_irq(FLOPPY_IRQ, NULL);
 
 #include <linux/pci.h>
@@ -152,7 +151,6 @@ static int hard_dma_setup(char *addr, unsigned long size, int mode, int io)
 	prev_dir = dir;
 
 	fd_clear_dma_ff();
-	fd_cacheflush(addr, size);
 	fd_set_dma_mode(mode);
 	set_dma_addr(FLOPPY_DMA, bus_addr);
 	fd_set_dma_count(size);
diff --git a/arch/sparc/include/asm/floppy_32.h b/arch/sparc/include/asm/floppy_32.h
index 071b83e52f15..dab58525229e 100644
--- a/arch/sparc/include/asm/floppy_32.h
+++ b/arch/sparc/include/asm/floppy_32.h
@@ -70,7 +70,6 @@ static struct sun_floppy_ops sun_fdops;
 #define fd_set_dma_count(count)   sun_fd_set_dma_count(count)
 #define fd_enable_irq()           /* nothing... */
 #define fd_disable_irq()          /* nothing... */
-#define fd_cacheflush(addr, size) /* nothing... */
 #define fd_request_irq()          sun_fd_request_irq()
 #define fd_free_irq()             /* nothing... */
 #if 0  /* P3: added by Alain, these cause a MMU corruption. 19960524 XXX */
diff --git a/arch/sparc/include/asm/floppy_64.h b/arch/sparc/include/asm/floppy_64.h
index 625756406a7e..a1db35a22c99 100644
--- a/arch/sparc/include/asm/floppy_64.h
+++ b/arch/sparc/include/asm/floppy_64.h
@@ -72,7 +72,6 @@ static struct sun_floppy_ops sun_fdops;
 #define fd_set_dma_addr(addr)     sun_fdops.fd_set_dma_addr(addr)
 #define fd_set_dma_count(count)   sun_fdops.fd_set_dma_count(count)
 #define get_dma_residue(x)        sun_fdops.get_dma_residue()
-#define fd_cacheflush(addr, size) /* nothing... */
 #define fd_request_irq()          sun_fdops.fd_request_irq()
 #define fd_free_irq()             sun_fdops.fd_free_irq()
 #define fd_eject(drive)           sun_fdops.fd_eject(drive)
diff --git a/drivers/block/floppy.c b/drivers/block/floppy.c
index 60c086a53609..a54183935aa1 100644
--- a/drivers/block/floppy.c
+++ b/drivers/block/floppy.c
@@ -275,6 +275,10 @@ static int set_next_request(void);
 #define fd_dma_mem_alloc(size) __get_dma_pages(GFP_KERNEL, get_order(size))
 #endif
 
+#ifndef fd_cacheflush
+#define fd_cacheflush(addr, size) /* nothing... */
+#endif
+
 static inline void fallback_on_nodma_alloc(char **addr, size_t l)
 {
 #ifdef FLOPPY_CAN_FALLBACK_ON_NODMA
-- 
2.14.1

[PATCH 01/11] floppy: consolidate the dummy fd_cacheflush definition

[Christoph Hellwig]

Only mips defines this helper, so remove all the other arch definitions.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/alpha/include/asm/floppy.h    | 2 --
 arch/powerpc/include/asm/floppy.h  | 2 --
 arch/sparc/include/asm/floppy_32.h | 1 -
 arch/sparc/include/asm/floppy_64.h | 1 -
 drivers/block/floppy.c             | 4 ++++
 5 files changed, 4 insertions(+), 6 deletions(-)

diff --git a/arch/alpha/include/asm/floppy.h b/arch/alpha/include/asm/floppy.h
index bae97eb19d26..942924756cf2 100644
--- a/arch/alpha/include/asm/floppy.h
+++ b/arch/alpha/include/asm/floppy.h
@@ -24,7 +24,6 @@
 #define fd_set_dma_count(count) set_dma_count(FLOPPY_DMA,count)
 #define fd_enable_irq()         enable_irq(FLOPPY_IRQ)
 #define fd_disable_irq()        disable_irq(FLOPPY_IRQ)
-#define fd_cacheflush(addr,size) /* nothing */
 #define fd_request_irq()        request_irq(FLOPPY_IRQ, floppy_interrupt,\
 					    0, "floppy", NULL)
 #define fd_free_irq()           free_irq(FLOPPY_IRQ, NULL)
@@ -62,7 +61,6 @@ alpha_fd_dma_setup(char *addr, unsigned long size, int mode, int io)
 	prev_dir = dir;
 
 	fd_clear_dma_ff();
-	fd_cacheflush(addr, size);
 	fd_set_dma_mode(mode);
 	set_dma_addr(FLOPPY_DMA, bus_addr);
 	fd_set_dma_count(size);
diff --git a/arch/powerpc/include/asm/floppy.h b/arch/powerpc/include/asm/floppy.h
index 936a904ae78c..167c44b58848 100644
--- a/arch/powerpc/include/asm/floppy.h
+++ b/arch/powerpc/include/asm/floppy.h
@@ -25,7 +25,6 @@
 #define fd_get_dma_residue()    fd_ops->_get_dma_residue(FLOPPY_DMA)
 #define fd_enable_irq()         enable_irq(FLOPPY_IRQ)
 #define fd_disable_irq()        disable_irq(FLOPPY_IRQ)
-#define fd_cacheflush(addr,size) /* nothing */
 #define fd_free_irq()           free_irq(FLOPPY_IRQ, NULL);
 
 #include <linux/pci.h>
@@ -152,7 +151,6 @@ static int hard_dma_setup(char *addr, unsigned long size, int mode, int io)
 	prev_dir = dir;
 
 	fd_clear_dma_ff();
-	fd_cacheflush(addr, size);
 	fd_set_dma_mode(mode);
 	set_dma_addr(FLOPPY_DMA, bus_addr);
 	fd_set_dma_count(size);
diff --git a/arch/sparc/include/asm/floppy_32.h b/arch/sparc/include/asm/floppy_32.h
index 071b83e52f15..dab58525229e 100644
--- a/arch/sparc/include/asm/floppy_32.h
+++ b/arch/sparc/include/asm/floppy_32.h
@@ -70,7 +70,6 @@ static struct sun_floppy_ops sun_fdops;
 #define fd_set_dma_count(count)   sun_fd_set_dma_count(count)
 #define fd_enable_irq()           /* nothing... */
 #define fd_disable_irq()          /* nothing... */
-#define fd_cacheflush(addr, size) /* nothing... */
 #define fd_request_irq()          sun_fd_request_irq()
 #define fd_free_irq()             /* nothing... */
 #if 0  /* P3: added by Alain, these cause a MMU corruption. 19960524 XXX */
diff --git a/arch/sparc/include/asm/floppy_64.h b/arch/sparc/include/asm/floppy_64.h
index 625756406a7e..a1db35a22c99 100644
--- a/arch/sparc/include/asm/floppy_64.h
+++ b/arch/sparc/include/asm/floppy_64.h
@@ -72,7 +72,6 @@ static struct sun_floppy_ops sun_fdops;
 #define fd_set_dma_addr(addr)     sun_fdops.fd_set_dma_addr(addr)
 #define fd_set_dma_count(count)   sun_fdops.fd_set_dma_count(count)
 #define get_dma_residue(x)        sun_fdops.get_dma_residue()
-#define fd_cacheflush(addr, size) /* nothing... */
 #define fd_request_irq()          sun_fdops.fd_request_irq()
 #define fd_free_irq()             sun_fdops.fd_free_irq()
 #define fd_eject(drive)           sun_fdops.fd_eject(drive)
diff --git a/drivers/block/floppy.c b/drivers/block/floppy.c
index 60c086a53609..a54183935aa1 100644
--- a/drivers/block/floppy.c
+++ b/drivers/block/floppy.c
@@ -275,6 +275,10 @@ static int set_next_request(void);
 #define fd_dma_mem_alloc(size) __get_dma_pages(GFP_KERNEL, get_order(size))
 #endif
 
+#ifndef fd_cacheflush
+#define fd_cacheflush(addr, size) /* nothing... */
+#endif
+
 static inline void fallback_on_nodma_alloc(char **addr, size_t l)
 {
 #ifdef FLOPPY_CAN_FALLBACK_ON_NODMA
-- 
2.14.1

[PATCH 01/11] floppy: consolidate the dummy fd_cacheflush definition

[Christoph Hellwig]

Only mips defines this helper, so remove all the other arch definitions.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 arch/alpha/include/asm/floppy.h    | 2 --
 arch/powerpc/include/asm/floppy.h  | 2 --
 arch/sparc/include/asm/floppy_32.h | 1 -
 arch/sparc/include/asm/floppy_64.h | 1 -
 drivers/block/floppy.c             | 4 ++++
 5 files changed, 4 insertions(+), 6 deletions(-)

diff --git a/arch/alpha/include/asm/floppy.h b/arch/alpha/include/asm/floppy.h
index bae97eb19d26..942924756cf2 100644
--- a/arch/alpha/include/asm/floppy.h
+++ b/arch/alpha/include/asm/floppy.h
@@ -24,7 +24,6 @@
 #define fd_set_dma_count(count) set_dma_count(FLOPPY_DMA,count)
 #define fd_enable_irq()         enable_irq(FLOPPY_IRQ)
 #define fd_disable_irq()        disable_irq(FLOPPY_IRQ)
-#define fd_cacheflush(addr,size) /* nothing */
 #define fd_request_irq()        request_irq(FLOPPY_IRQ, floppy_interrupt,\
 					    0, "floppy", NULL)
 #define fd_free_irq()           free_irq(FLOPPY_IRQ, NULL)
@@ -62,7 +61,6 @@ alpha_fd_dma_setup(char *addr, unsigned long size, int mode, int io)
 	prev_dir = dir;
 
 	fd_clear_dma_ff();
-	fd_cacheflush(addr, size);
 	fd_set_dma_mode(mode);
 	set_dma_addr(FLOPPY_DMA, bus_addr);
 	fd_set_dma_count(size);
diff --git a/arch/powerpc/include/asm/floppy.h b/arch/powerpc/include/asm/floppy.h
index 936a904ae78c..167c44b58848 100644
--- a/arch/powerpc/include/asm/floppy.h
+++ b/arch/powerpc/include/asm/floppy.h
@@ -25,7 +25,6 @@
 #define fd_get_dma_residue()    fd_ops->_get_dma_residue(FLOPPY_DMA)
 #define fd_enable_irq()         enable_irq(FLOPPY_IRQ)
 #define fd_disable_irq()        disable_irq(FLOPPY_IRQ)
-#define fd_cacheflush(addr,size) /* nothing */
 #define fd_free_irq()           free_irq(FLOPPY_IRQ, NULL);
 
 #include <linux/pci.h>
@@ -152,7 +151,6 @@ static int hard_dma_setup(char *addr, unsigned long size, int mode, int io)
 	prev_dir = dir;
 
 	fd_clear_dma_ff();
-	fd_cacheflush(addr, size);
 	fd_set_dma_mode(mode);
 	set_dma_addr(FLOPPY_DMA, bus_addr);
 	fd_set_dma_count(size);
diff --git a/arch/sparc/include/asm/floppy_32.h b/arch/sparc/include/asm/floppy_32.h
index 071b83e52f15..dab58525229e 100644
--- a/arch/sparc/include/asm/floppy_32.h
+++ b/arch/sparc/include/asm/floppy_32.h
@@ -70,7 +70,6 @@ static struct sun_floppy_ops sun_fdops;
 #define fd_set_dma_count(count)   sun_fd_set_dma_count(count)
 #define fd_enable_irq()           /* nothing... */
 #define fd_disable_irq()          /* nothing... */
-#define fd_cacheflush(addr, size) /* nothing... */
 #define fd_request_irq()          sun_fd_request_irq()
 #define fd_free_irq()             /* nothing... */
 #if 0  /* P3: added by Alain, these cause a MMU corruption. 19960524 XXX */
diff --git a/arch/sparc/include/asm/floppy_64.h b/arch/sparc/include/asm/floppy_64.h
index 625756406a7e..a1db35a22c99 100644
--- a/arch/sparc/include/asm/floppy_64.h
+++ b/arch/sparc/include/asm/floppy_64.h
@@ -72,7 +72,6 @@ static struct sun_floppy_ops sun_fdops;
 #define fd_set_dma_addr(addr)     sun_fdops.fd_set_dma_addr(addr)
 #define fd_set_dma_count(count)   sun_fdops.fd_set_dma_count(count)
 #define get_dma_residue(x)        sun_fdops.get_dma_residue()
-#define fd_cacheflush(addr, size) /* nothing... */
 #define fd_request_irq()          sun_fdops.fd_request_irq()
 #define fd_free_irq()             sun_fdops.fd_free_irq()
 #define fd_eject(drive)           sun_fdops.fd_eject(drive)
diff --git a/drivers/block/floppy.c b/drivers/block/floppy.c
index 60c086a53609..a54183935aa1 100644
--- a/drivers/block/floppy.c
+++ b/drivers/block/floppy.c
@@ -275,6 +275,10 @@ static int set_next_request(void);
 #define fd_dma_mem_alloc(size) __get_dma_pages(GFP_KERNEL, get_order(size))
 #endif
 
+#ifndef fd_cacheflush
+#define fd_cacheflush(addr, size) /* nothing... */
+#endif
+
 static inline void fallback_on_nodma_alloc(char **addr, size_t l)
 {
 #ifdef FLOPPY_CAN_FALLBACK_ON_NODMA
-- 
2.14.1

[PATCH 02/11] x86: make dma_cache_sync a no-op

[Christoph Hellwig]

x86 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
make any sense to do any work in dma_cache_sync given that it must be a
no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
---
 arch/x86/include/asm/dma-mapping.h | 1 -
 1 file changed, 1 deletion(-)

diff --git a/arch/x86/include/asm/dma-mapping.h b/arch/x86/include/asm/dma-mapping.h
index 1387dafdba2d..c6d3367be916 100644
--- a/arch/x86/include/asm/dma-mapping.h
+++ b/arch/x86/include/asm/dma-mapping.h
@@ -71,7 +71,6 @@ static inline void
 dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 	enum dma_data_direction dir)
 {
-	flush_write_buffers();
 }
 
 static inline unsigned long dma_alloc_coherent_mask(struct device *dev,
-- 
2.14.1

[PATCH 02/11] x86: make dma_cache_sync a no-op

[Christoph Hellwig]

x86 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
make any sense to do any work in dma_cache_sync given that it must be a
no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
---
 arch/x86/include/asm/dma-mapping.h | 1 -
 1 file changed, 1 deletion(-)

diff --git a/arch/x86/include/asm/dma-mapping.h b/arch/x86/include/asm/dma-mapping.h
index 1387dafdba2d..c6d3367be916 100644
--- a/arch/x86/include/asm/dma-mapping.h
+++ b/arch/x86/include/asm/dma-mapping.h
@@ -71,7 +71,6 @@ static inline void
 dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 	enum dma_data_direction dir)
 {
-	flush_write_buffers();
 }
 
 static inline unsigned long dma_alloc_coherent_mask(struct device *dev,
-- 
2.14.1

[PATCH 02/11] x86: make dma_cache_sync a no-op

[Christoph Hellwig]

x86 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
make any sense to do any work in dma_cache_sync given that it must be a
no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
Reviewed-by: Thomas Gleixner <tglx-hfZtesqFncYOwBW4kG4KsQ@public.gmane.org>
---
 arch/x86/include/asm/dma-mapping.h | 1 -
 1 file changed, 1 deletion(-)

diff --git a/arch/x86/include/asm/dma-mapping.h b/arch/x86/include/asm/dma-mapping.h
index 1387dafdba2d..c6d3367be916 100644
--- a/arch/x86/include/asm/dma-mapping.h
+++ b/arch/x86/include/asm/dma-mapping.h
@@ -71,7 +71,6 @@ static inline void
 dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 	enum dma_data_direction dir)
 {
-	flush_write_buffers();
 }
 
 static inline unsigned long dma_alloc_coherent_mask(struct device *dev,
-- 
2.14.1

[PATCH 03/11] frv: make dma_cache_sync a no-op

[Christoph Hellwig]

frv does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
make any sense to do any work in dma_cache_sync given that it must be a
no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/frv/include/asm/dma-mapping.h | 1 -
 1 file changed, 1 deletion(-)

diff --git a/arch/frv/include/asm/dma-mapping.h b/arch/frv/include/asm/dma-mapping.h
index 354900917585..da0e5c9744c4 100644
--- a/arch/frv/include/asm/dma-mapping.h
+++ b/arch/frv/include/asm/dma-mapping.h
@@ -18,7 +18,6 @@ static inline
 void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		    enum dma_data_direction direction)
 {
-	flush_write_buffers();
 }
 
 #endif  /* _ASM_DMA_MAPPING_H */
-- 
2.14.1

[PATCH 03/11] frv: make dma_cache_sync a no-op

[Christoph Hellwig]

frv does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
make any sense to do any work in dma_cache_sync given that it must be a
no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/frv/include/asm/dma-mapping.h | 1 -
 1 file changed, 1 deletion(-)

diff --git a/arch/frv/include/asm/dma-mapping.h b/arch/frv/include/asm/dma-mapping.h
index 354900917585..da0e5c9744c4 100644
--- a/arch/frv/include/asm/dma-mapping.h
+++ b/arch/frv/include/asm/dma-mapping.h
@@ -18,7 +18,6 @@ static inline
 void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		    enum dma_data_direction direction)
 {
-	flush_write_buffers();
 }
 
 #endif  /* _ASM_DMA_MAPPING_H */
-- 
2.14.1

[PATCH 03/11] frv: make dma_cache_sync a no-op

[Christoph Hellwig]

frv does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
make any sense to do any work in dma_cache_sync given that it must be a
no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 arch/frv/include/asm/dma-mapping.h | 1 -
 1 file changed, 1 deletion(-)

diff --git a/arch/frv/include/asm/dma-mapping.h b/arch/frv/include/asm/dma-mapping.h
index 354900917585..da0e5c9744c4 100644
--- a/arch/frv/include/asm/dma-mapping.h
+++ b/arch/frv/include/asm/dma-mapping.h
@@ -18,7 +18,6 @@ static inline
 void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		    enum dma_data_direction direction)
 {
-	flush_write_buffers();
 }
 
 #endif  /* _ASM_DMA_MAPPING_H */
-- 
2.14.1

[PATCH 04/11] ia64: make dma_cache_sync a no-op

[Christoph Hellwig]

ia64 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
make any sense to do any work in dma_cache_sync given that it must be a
no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/ia64/include/asm/dma-mapping.h | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/arch/ia64/include/asm/dma-mapping.h b/arch/ia64/include/asm/dma-mapping.h
index 3ce5ab4339f3..99dfc1aa9d3c 100644
--- a/arch/ia64/include/asm/dma-mapping.h
+++ b/arch/ia64/include/asm/dma-mapping.h
@@ -48,11 +48,6 @@ static inline void
 dma_cache_sync (struct device *dev, void *vaddr, size_t size,
 	enum dma_data_direction dir)
 {
-	/*
-	 * IA-64 is cache-coherent, so this is mostly a no-op.  However, we do need to
-	 * ensure that dma_cache_sync() enforces order, hence the mb().
-	 */
-	mb();
 }
 
 #endif /* _ASM_IA64_DMA_MAPPING_H */
-- 
2.14.1

[PATCH 04/11] ia64: make dma_cache_sync a no-op

[Christoph Hellwig]

ia64 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
make any sense to do any work in dma_cache_sync given that it must be a
no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/ia64/include/asm/dma-mapping.h | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/arch/ia64/include/asm/dma-mapping.h b/arch/ia64/include/asm/dma-mapping.h
index 3ce5ab4339f3..99dfc1aa9d3c 100644
--- a/arch/ia64/include/asm/dma-mapping.h
+++ b/arch/ia64/include/asm/dma-mapping.h
@@ -48,11 +48,6 @@ static inline void
 dma_cache_sync (struct device *dev, void *vaddr, size_t size,
 	enum dma_data_direction dir)
 {
-	/*
-	 * IA-64 is cache-coherent, so this is mostly a no-op.  However, we do need to
-	 * ensure that dma_cache_sync() enforces order, hence the mb().
-	 */
-	mb();
 }
 
 #endif /* _ASM_IA64_DMA_MAPPING_H */
-- 
2.14.1

[PATCH 04/11] ia64: make dma_cache_sync a no-op

[Christoph Hellwig]

ia64 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
make any sense to do any work in dma_cache_sync given that it must be a
no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 arch/ia64/include/asm/dma-mapping.h | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/arch/ia64/include/asm/dma-mapping.h b/arch/ia64/include/asm/dma-mapping.h
index 3ce5ab4339f3..99dfc1aa9d3c 100644
--- a/arch/ia64/include/asm/dma-mapping.h
+++ b/arch/ia64/include/asm/dma-mapping.h
@@ -48,11 +48,6 @@ static inline void
 dma_cache_sync (struct device *dev, void *vaddr, size_t size,
 	enum dma_data_direction dir)
 {
-	/*
-	 * IA-64 is cache-coherent, so this is mostly a no-op.  However, we do need to
-	 * ensure that dma_cache_sync() enforces order, hence the mb().
-	 */
-	mb();
 }
 
 #endif /* _ASM_IA64_DMA_MAPPING_H */
-- 
2.14.1

[PATCH 05/11] microblaze: make dma_cache_sync a no-op

[Christoph Hellwig]

microblaze does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

This also allows moving __dma_sync out of the microblaze asm/dma-mapping.h
and thus greatly reduce the amount of includes there.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/microblaze/include/asm/dma-mapping.h | 34 -------------------------------
 arch/microblaze/kernel/dma.c              | 17 ++++++++++++++++
 2 files changed, 17 insertions(+), 34 deletions(-)

diff --git a/arch/microblaze/include/asm/dma-mapping.h b/arch/microblaze/include/asm/dma-mapping.h
index e15cd2f76e23..ad448e4aedb6 100644
--- a/arch/microblaze/include/asm/dma-mapping.h
+++ b/arch/microblaze/include/asm/dma-mapping.h
@@ -15,22 +15,6 @@
 #ifndef _ASM_MICROBLAZE_DMA_MAPPING_H
 #define _ASM_MICROBLAZE_DMA_MAPPING_H
 
-/*
- * See Documentation/DMA-API-HOWTO.txt and
- * Documentation/DMA-API.txt for documentation.
- */
-
-#include <linux/types.h>
-#include <linux/cache.h>
-#include <linux/mm.h>
-#include <linux/scatterlist.h>
-#include <linux/dma-debug.h>
-#include <asm/io.h>
-#include <asm/cacheflush.h>
-
-#define __dma_alloc_coherent(dev, gfp, size, handle)	NULL
-#define __dma_free_coherent(size, addr)		((void)0)
-
 /*
  * Available generic sets of operations
  */
@@ -41,27 +25,9 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &dma_direct_ops;
 }
 
-static inline void __dma_sync(unsigned long paddr,
-			      size_t size, enum dma_data_direction direction)
-{
-	switch (direction) {
-	case DMA_TO_DEVICE:
-	case DMA_BIDIRECTIONAL:
-		flush_dcache_range(paddr, paddr + size);
-		break;
-	case DMA_FROM_DEVICE:
-		invalidate_dcache_range(paddr, paddr + size);
-		break;
-	default:
-		BUG();
-	}
-}
-
 static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		enum dma_data_direction direction)
 {
-	BUG_ON(direction = DMA_NONE);
-	__dma_sync(virt_to_phys(vaddr), size, (int)direction);
 }
 
 #endif	/* _ASM_MICROBLAZE_DMA_MAPPING_H */
diff --git a/arch/microblaze/kernel/dma.c b/arch/microblaze/kernel/dma.c
index 94700c5270a9..e52b684f8ea2 100644
--- a/arch/microblaze/kernel/dma.c
+++ b/arch/microblaze/kernel/dma.c
@@ -12,6 +12,7 @@
 #include <linux/dma-debug.h>
 #include <linux/export.h>
 #include <linux/bug.h>
+#include <asm/cacheflush.h>
 
 #define NOT_COHERENT_CACHE
 
@@ -51,6 +52,22 @@ static void dma_direct_free_coherent(struct device *dev, size_t size,
 #endif
 }
 
+static inline void __dma_sync(unsigned long paddr,
+			      size_t size, enum dma_data_direction direction)
+{
+	switch (direction) {
+	case DMA_TO_DEVICE:
+	case DMA_BIDIRECTIONAL:
+		flush_dcache_range(paddr, paddr + size);
+		break;
+	case DMA_FROM_DEVICE:
+		invalidate_dcache_range(paddr, paddr + size);
+		break;
+	default:
+		BUG();
+	}
+}
+
 static int dma_direct_map_sg(struct device *dev, struct scatterlist *sgl,
 			     int nents, enum dma_data_direction direction,
 			     unsigned long attrs)
-- 
2.14.1

[PATCH 05/11] microblaze: make dma_cache_sync a no-op

[Christoph Hellwig]

microblaze does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

This also allows moving __dma_sync out of the microblaze asm/dma-mapping.h
and thus greatly reduce the amount of includes there.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/microblaze/include/asm/dma-mapping.h | 34 -------------------------------
 arch/microblaze/kernel/dma.c              | 17 ++++++++++++++++
 2 files changed, 17 insertions(+), 34 deletions(-)

diff --git a/arch/microblaze/include/asm/dma-mapping.h b/arch/microblaze/include/asm/dma-mapping.h
index e15cd2f76e23..ad448e4aedb6 100644
--- a/arch/microblaze/include/asm/dma-mapping.h
+++ b/arch/microblaze/include/asm/dma-mapping.h
@@ -15,22 +15,6 @@
 #ifndef _ASM_MICROBLAZE_DMA_MAPPING_H
 #define _ASM_MICROBLAZE_DMA_MAPPING_H
 
-/*
- * See Documentation/DMA-API-HOWTO.txt and
- * Documentation/DMA-API.txt for documentation.
- */
-
-#include <linux/types.h>
-#include <linux/cache.h>
-#include <linux/mm.h>
-#include <linux/scatterlist.h>
-#include <linux/dma-debug.h>
-#include <asm/io.h>
-#include <asm/cacheflush.h>
-
-#define __dma_alloc_coherent(dev, gfp, size, handle)	NULL
-#define __dma_free_coherent(size, addr)		((void)0)
-
 /*
  * Available generic sets of operations
  */
@@ -41,27 +25,9 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &dma_direct_ops;
 }
 
-static inline void __dma_sync(unsigned long paddr,
-			      size_t size, enum dma_data_direction direction)
-{
-	switch (direction) {
-	case DMA_TO_DEVICE:
-	case DMA_BIDIRECTIONAL:
-		flush_dcache_range(paddr, paddr + size);
-		break;
-	case DMA_FROM_DEVICE:
-		invalidate_dcache_range(paddr, paddr + size);
-		break;
-	default:
-		BUG();
-	}
-}
-
 static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		enum dma_data_direction direction)
 {
-	BUG_ON(direction == DMA_NONE);
-	__dma_sync(virt_to_phys(vaddr), size, (int)direction);
 }
 
 #endif	/* _ASM_MICROBLAZE_DMA_MAPPING_H */
diff --git a/arch/microblaze/kernel/dma.c b/arch/microblaze/kernel/dma.c
index 94700c5270a9..e52b684f8ea2 100644
--- a/arch/microblaze/kernel/dma.c
+++ b/arch/microblaze/kernel/dma.c
@@ -12,6 +12,7 @@
 #include <linux/dma-debug.h>
 #include <linux/export.h>
 #include <linux/bug.h>
+#include <asm/cacheflush.h>
 
 #define NOT_COHERENT_CACHE
 
@@ -51,6 +52,22 @@ static void dma_direct_free_coherent(struct device *dev, size_t size,
 #endif
 }
 
+static inline void __dma_sync(unsigned long paddr,
+			      size_t size, enum dma_data_direction direction)
+{
+	switch (direction) {
+	case DMA_TO_DEVICE:
+	case DMA_BIDIRECTIONAL:
+		flush_dcache_range(paddr, paddr + size);
+		break;
+	case DMA_FROM_DEVICE:
+		invalidate_dcache_range(paddr, paddr + size);
+		break;
+	default:
+		BUG();
+	}
+}
+
 static int dma_direct_map_sg(struct device *dev, struct scatterlist *sgl,
 			     int nents, enum dma_data_direction direction,
 			     unsigned long attrs)
-- 
2.14.1

[PATCH 05/11] microblaze: make dma_cache_sync a no-op

[Christoph Hellwig]

microblaze does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

This also allows moving __dma_sync out of the microblaze asm/dma-mapping.h
and thus greatly reduce the amount of includes there.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 arch/microblaze/include/asm/dma-mapping.h | 34 -------------------------------
 arch/microblaze/kernel/dma.c              | 17 ++++++++++++++++
 2 files changed, 17 insertions(+), 34 deletions(-)

diff --git a/arch/microblaze/include/asm/dma-mapping.h b/arch/microblaze/include/asm/dma-mapping.h
index e15cd2f76e23..ad448e4aedb6 100644
--- a/arch/microblaze/include/asm/dma-mapping.h
+++ b/arch/microblaze/include/asm/dma-mapping.h
@@ -15,22 +15,6 @@
 #ifndef _ASM_MICROBLAZE_DMA_MAPPING_H
 #define _ASM_MICROBLAZE_DMA_MAPPING_H
 
-/*
- * See Documentation/DMA-API-HOWTO.txt and
- * Documentation/DMA-API.txt for documentation.
- */
-
-#include <linux/types.h>
-#include <linux/cache.h>
-#include <linux/mm.h>
-#include <linux/scatterlist.h>
-#include <linux/dma-debug.h>
-#include <asm/io.h>
-#include <asm/cacheflush.h>
-
-#define __dma_alloc_coherent(dev, gfp, size, handle)	NULL
-#define __dma_free_coherent(size, addr)		((void)0)
-
 /*
  * Available generic sets of operations
  */
@@ -41,27 +25,9 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &dma_direct_ops;
 }
 
-static inline void __dma_sync(unsigned long paddr,
-			      size_t size, enum dma_data_direction direction)
-{
-	switch (direction) {
-	case DMA_TO_DEVICE:
-	case DMA_BIDIRECTIONAL:
-		flush_dcache_range(paddr, paddr + size);
-		break;
-	case DMA_FROM_DEVICE:
-		invalidate_dcache_range(paddr, paddr + size);
-		break;
-	default:
-		BUG();
-	}
-}
-
 static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		enum dma_data_direction direction)
 {
-	BUG_ON(direction == DMA_NONE);
-	__dma_sync(virt_to_phys(vaddr), size, (int)direction);
 }
 
 #endif	/* _ASM_MICROBLAZE_DMA_MAPPING_H */
diff --git a/arch/microblaze/kernel/dma.c b/arch/microblaze/kernel/dma.c
index 94700c5270a9..e52b684f8ea2 100644
--- a/arch/microblaze/kernel/dma.c
+++ b/arch/microblaze/kernel/dma.c
@@ -12,6 +12,7 @@
 #include <linux/dma-debug.h>
 #include <linux/export.h>
 #include <linux/bug.h>
+#include <asm/cacheflush.h>
 
 #define NOT_COHERENT_CACHE
 
@@ -51,6 +52,22 @@ static void dma_direct_free_coherent(struct device *dev, size_t size,
 #endif
 }
 
+static inline void __dma_sync(unsigned long paddr,
+			      size_t size, enum dma_data_direction direction)
+{
+	switch (direction) {
+	case DMA_TO_DEVICE:
+	case DMA_BIDIRECTIONAL:
+		flush_dcache_range(paddr, paddr + size);
+		break;
+	case DMA_FROM_DEVICE:
+		invalidate_dcache_range(paddr, paddr + size);
+		break;
+	default:
+		BUG();
+	}
+}
+
 static int dma_direct_map_sg(struct device *dev, struct scatterlist *sgl,
 			     int nents, enum dma_data_direction direction,
 			     unsigned long attrs)
-- 
2.14.1

[PATCH 06/11] mn10300: make dma_cache_sync a no-op

[Christoph Hellwig]

mn10300 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it must
be a no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/mn10300/include/asm/dma-mapping.h | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/arch/mn10300/include/asm/dma-mapping.h b/arch/mn10300/include/asm/dma-mapping.h
index 737ef574b3ea..dc24163b190f 100644
--- a/arch/mn10300/include/asm/dma-mapping.h
+++ b/arch/mn10300/include/asm/dma-mapping.h
@@ -11,9 +11,6 @@
 #ifndef _ASM_DMA_MAPPING_H
 #define _ASM_DMA_MAPPING_H
 
-#include <asm/cache.h>
-#include <asm/io.h>
-
 extern const struct dma_map_ops mn10300_dma_ops;
 
 static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
@@ -25,7 +22,6 @@ static inline
 void dma_cache_sync(void *vaddr, size_t size,
 		    enum dma_data_direction direction)
 {
-	mn10300_dcache_flush_inv();
 }
 
 #endif
-- 
2.14.1

[PATCH 06/11] mn10300: make dma_cache_sync a no-op

[Christoph Hellwig]

mn10300 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it must
be a no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/mn10300/include/asm/dma-mapping.h | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/arch/mn10300/include/asm/dma-mapping.h b/arch/mn10300/include/asm/dma-mapping.h
index 737ef574b3ea..dc24163b190f 100644
--- a/arch/mn10300/include/asm/dma-mapping.h
+++ b/arch/mn10300/include/asm/dma-mapping.h
@@ -11,9 +11,6 @@
 #ifndef _ASM_DMA_MAPPING_H
 #define _ASM_DMA_MAPPING_H
 
-#include <asm/cache.h>
-#include <asm/io.h>
-
 extern const struct dma_map_ops mn10300_dma_ops;
 
 static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
@@ -25,7 +22,6 @@ static inline
 void dma_cache_sync(void *vaddr, size_t size,
 		    enum dma_data_direction direction)
 {
-	mn10300_dcache_flush_inv();
 }
 
 #endif
-- 
2.14.1

[PATCH 06/11] mn10300: make dma_cache_sync a no-op

[Christoph Hellwig]

mn10300 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it must
be a no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 arch/mn10300/include/asm/dma-mapping.h | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/arch/mn10300/include/asm/dma-mapping.h b/arch/mn10300/include/asm/dma-mapping.h
index 737ef574b3ea..dc24163b190f 100644
--- a/arch/mn10300/include/asm/dma-mapping.h
+++ b/arch/mn10300/include/asm/dma-mapping.h
@@ -11,9 +11,6 @@
 #ifndef _ASM_DMA_MAPPING_H
 #define _ASM_DMA_MAPPING_H
 
-#include <asm/cache.h>
-#include <asm/io.h>
-
 extern const struct dma_map_ops mn10300_dma_ops;
 
 static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
@@ -25,7 +22,6 @@ static inline
 void dma_cache_sync(void *vaddr, size_t size,
 		    enum dma_data_direction direction)
 {
-	mn10300_dcache_flush_inv();
 }
 
 #endif
-- 
2.14.1

[PATCH 07/11] powerpc: make dma_cache_sync a no-op

[Christoph Hellwig]

powerpc does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/powerpc/include/asm/dma-mapping.h | 2 --
 1 file changed, 2 deletions(-)

diff --git a/arch/powerpc/include/asm/dma-mapping.h b/arch/powerpc/include/asm/dma-mapping.h
index eaece3d3e225..320846442bfb 100644
--- a/arch/powerpc/include/asm/dma-mapping.h
+++ b/arch/powerpc/include/asm/dma-mapping.h
@@ -144,8 +144,6 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
 static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		enum dma_data_direction direction)
 {
-	BUG_ON(direction = DMA_NONE);
-	__dma_sync(vaddr, size, (int)direction);
 }
 
 #endif /* __KERNEL__ */
-- 
2.14.1

[PATCH 07/11] powerpc: make dma_cache_sync a no-op

[Christoph Hellwig]

powerpc does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/powerpc/include/asm/dma-mapping.h | 2 --
 1 file changed, 2 deletions(-)

diff --git a/arch/powerpc/include/asm/dma-mapping.h b/arch/powerpc/include/asm/dma-mapping.h
index eaece3d3e225..320846442bfb 100644
--- a/arch/powerpc/include/asm/dma-mapping.h
+++ b/arch/powerpc/include/asm/dma-mapping.h
@@ -144,8 +144,6 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
 static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		enum dma_data_direction direction)
 {
-	BUG_ON(direction == DMA_NONE);
-	__dma_sync(vaddr, size, (int)direction);
 }
 
 #endif /* __KERNEL__ */
-- 
2.14.1

[PATCH 08/11] unicore32: make dma_cache_sync a no-op

[Christoph Hellwig]

unicore32 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/unicore32/include/asm/cacheflush.h  |  9 ---------
 arch/unicore32/include/asm/dma-mapping.h | 17 -----------------
 arch/unicore32/mm/proc-syms.c            |  3 ---
 3 files changed, 29 deletions(-)

diff --git a/arch/unicore32/include/asm/cacheflush.h b/arch/unicore32/include/asm/cacheflush.h
index c0301e6c8b81..a5e08e2d5d6d 100644
--- a/arch/unicore32/include/asm/cacheflush.h
+++ b/arch/unicore32/include/asm/cacheflush.h
@@ -101,15 +101,6 @@ extern void __cpuc_coherent_user_range(unsigned long, unsigned long);
 extern void __cpuc_flush_dcache_area(void *, size_t);
 extern void __cpuc_flush_kern_dcache_area(void *addr, size_t size);
 
-/*
- * These are private to the dma-mapping API.  Do not use directly.
- * Their sole purpose is to ensure that data held in the cache
- * is visible to DMA, or data written by DMA to system memory is
- * visible to the CPU.
- */
-extern void __cpuc_dma_clean_range(unsigned long, unsigned long);
-extern void __cpuc_dma_flush_range(unsigned long, unsigned long);
-
 /*
  * Copy user data from/to a page which is mapped into a different
  * processes address space.  Really, we want to allow our "user
diff --git a/arch/unicore32/include/asm/dma-mapping.h b/arch/unicore32/include/asm/dma-mapping.h
index 518ba5848dd6..e949855bb794 100644
--- a/arch/unicore32/include/asm/dma-mapping.h
+++ b/arch/unicore32/include/asm/dma-mapping.h
@@ -18,9 +18,6 @@
 #include <linux/scatterlist.h>
 #include <linux/swiotlb.h>
 
-#include <asm/memory.h>
-#include <asm/cacheflush.h>
-
 extern const struct dma_map_ops swiotlb_dma_map_ops;
 
 static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
@@ -51,20 +48,6 @@ static inline void dma_mark_clean(void *addr, size_t size) {}
 static inline void dma_cache_sync(struct device *dev, void *vaddr,
 		size_t size, enum dma_data_direction direction)
 {
-	unsigned long start = (unsigned long)vaddr;
-	unsigned long end   = start + size;
-
-	switch (direction) {
-	case DMA_NONE:
-		BUG();
-	case DMA_FROM_DEVICE:
-	case DMA_BIDIRECTIONAL:	/* writeback and invalidate */
-		__cpuc_dma_flush_range(start, end);
-		break;
-	case DMA_TO_DEVICE:		/* writeback only */
-		__cpuc_dma_clean_range(start, end);
-		break;
-	}
 }
 
 #endif /* __KERNEL__ */
diff --git a/arch/unicore32/mm/proc-syms.c b/arch/unicore32/mm/proc-syms.c
index 21c00fc85c99..df215fd6d639 100644
--- a/arch/unicore32/mm/proc-syms.c
+++ b/arch/unicore32/mm/proc-syms.c
@@ -20,6 +20,3 @@ EXPORT_SYMBOL(cpu_dcache_clean_area);
 EXPORT_SYMBOL(cpu_set_pte);
 
 EXPORT_SYMBOL(__cpuc_coherent_kern_range);
-
-EXPORT_SYMBOL(__cpuc_dma_flush_range);
-EXPORT_SYMBOL(__cpuc_dma_clean_range);
-- 
2.14.1

[PATCH 08/11] unicore32: make dma_cache_sync a no-op

[Christoph Hellwig]

unicore32 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/unicore32/include/asm/cacheflush.h  |  9 ---------
 arch/unicore32/include/asm/dma-mapping.h | 17 -----------------
 arch/unicore32/mm/proc-syms.c            |  3 ---
 3 files changed, 29 deletions(-)

diff --git a/arch/unicore32/include/asm/cacheflush.h b/arch/unicore32/include/asm/cacheflush.h
index c0301e6c8b81..a5e08e2d5d6d 100644
--- a/arch/unicore32/include/asm/cacheflush.h
+++ b/arch/unicore32/include/asm/cacheflush.h
@@ -101,15 +101,6 @@ extern void __cpuc_coherent_user_range(unsigned long, unsigned long);
 extern void __cpuc_flush_dcache_area(void *, size_t);
 extern void __cpuc_flush_kern_dcache_area(void *addr, size_t size);
 
-/*
- * These are private to the dma-mapping API.  Do not use directly.
- * Their sole purpose is to ensure that data held in the cache
- * is visible to DMA, or data written by DMA to system memory is
- * visible to the CPU.
- */
-extern void __cpuc_dma_clean_range(unsigned long, unsigned long);
-extern void __cpuc_dma_flush_range(unsigned long, unsigned long);
-
 /*
  * Copy user data from/to a page which is mapped into a different
  * processes address space.  Really, we want to allow our "user
diff --git a/arch/unicore32/include/asm/dma-mapping.h b/arch/unicore32/include/asm/dma-mapping.h
index 518ba5848dd6..e949855bb794 100644
--- a/arch/unicore32/include/asm/dma-mapping.h
+++ b/arch/unicore32/include/asm/dma-mapping.h
@@ -18,9 +18,6 @@
 #include <linux/scatterlist.h>
 #include <linux/swiotlb.h>
 
-#include <asm/memory.h>
-#include <asm/cacheflush.h>
-
 extern const struct dma_map_ops swiotlb_dma_map_ops;
 
 static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
@@ -51,20 +48,6 @@ static inline void dma_mark_clean(void *addr, size_t size) {}
 static inline void dma_cache_sync(struct device *dev, void *vaddr,
 		size_t size, enum dma_data_direction direction)
 {
-	unsigned long start = (unsigned long)vaddr;
-	unsigned long end   = start + size;
-
-	switch (direction) {
-	case DMA_NONE:
-		BUG();
-	case DMA_FROM_DEVICE:
-	case DMA_BIDIRECTIONAL:	/* writeback and invalidate */
-		__cpuc_dma_flush_range(start, end);
-		break;
-	case DMA_TO_DEVICE:		/* writeback only */
-		__cpuc_dma_clean_range(start, end);
-		break;
-	}
 }
 
 #endif /* __KERNEL__ */
diff --git a/arch/unicore32/mm/proc-syms.c b/arch/unicore32/mm/proc-syms.c
index 21c00fc85c99..df215fd6d639 100644
--- a/arch/unicore32/mm/proc-syms.c
+++ b/arch/unicore32/mm/proc-syms.c
@@ -20,6 +20,3 @@ EXPORT_SYMBOL(cpu_dcache_clean_area);
 EXPORT_SYMBOL(cpu_set_pte);
 
 EXPORT_SYMBOL(__cpuc_coherent_kern_range);
-
-EXPORT_SYMBOL(__cpuc_dma_flush_range);
-EXPORT_SYMBOL(__cpuc_dma_clean_range);
-- 
2.14.1

[PATCH 08/11] unicore32: make dma_cache_sync a no-op

[Christoph Hellwig]

unicore32 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 arch/unicore32/include/asm/cacheflush.h  |  9 ---------
 arch/unicore32/include/asm/dma-mapping.h | 17 -----------------
 arch/unicore32/mm/proc-syms.c            |  3 ---
 3 files changed, 29 deletions(-)

diff --git a/arch/unicore32/include/asm/cacheflush.h b/arch/unicore32/include/asm/cacheflush.h
index c0301e6c8b81..a5e08e2d5d6d 100644
--- a/arch/unicore32/include/asm/cacheflush.h
+++ b/arch/unicore32/include/asm/cacheflush.h
@@ -101,15 +101,6 @@ extern void __cpuc_coherent_user_range(unsigned long, unsigned long);
 extern void __cpuc_flush_dcache_area(void *, size_t);
 extern void __cpuc_flush_kern_dcache_area(void *addr, size_t size);
 
-/*
- * These are private to the dma-mapping API.  Do not use directly.
- * Their sole purpose is to ensure that data held in the cache
- * is visible to DMA, or data written by DMA to system memory is
- * visible to the CPU.
- */
-extern void __cpuc_dma_clean_range(unsigned long, unsigned long);
-extern void __cpuc_dma_flush_range(unsigned long, unsigned long);
-
 /*
  * Copy user data from/to a page which is mapped into a different
  * processes address space.  Really, we want to allow our "user
diff --git a/arch/unicore32/include/asm/dma-mapping.h b/arch/unicore32/include/asm/dma-mapping.h
index 518ba5848dd6..e949855bb794 100644
--- a/arch/unicore32/include/asm/dma-mapping.h
+++ b/arch/unicore32/include/asm/dma-mapping.h
@@ -18,9 +18,6 @@
 #include <linux/scatterlist.h>
 #include <linux/swiotlb.h>
 
-#include <asm/memory.h>
-#include <asm/cacheflush.h>
-
 extern const struct dma_map_ops swiotlb_dma_map_ops;
 
 static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
@@ -51,20 +48,6 @@ static inline void dma_mark_clean(void *addr, size_t size) {}
 static inline void dma_cache_sync(struct device *dev, void *vaddr,
 		size_t size, enum dma_data_direction direction)
 {
-	unsigned long start = (unsigned long)vaddr;
-	unsigned long end   = start + size;
-
-	switch (direction) {
-	case DMA_NONE:
-		BUG();
-	case DMA_FROM_DEVICE:
-	case DMA_BIDIRECTIONAL:	/* writeback and invalidate */
-		__cpuc_dma_flush_range(start, end);
-		break;
-	case DMA_TO_DEVICE:		/* writeback only */
-		__cpuc_dma_clean_range(start, end);
-		break;
-	}
 }
 
 #endif /* __KERNEL__ */
diff --git a/arch/unicore32/mm/proc-syms.c b/arch/unicore32/mm/proc-syms.c
index 21c00fc85c99..df215fd6d639 100644
--- a/arch/unicore32/mm/proc-syms.c
+++ b/arch/unicore32/mm/proc-syms.c
@@ -20,6 +20,3 @@ EXPORT_SYMBOL(cpu_dcache_clean_area);
 EXPORT_SYMBOL(cpu_set_pte);
 
 EXPORT_SYMBOL(__cpuc_coherent_kern_range);
-
-EXPORT_SYMBOL(__cpuc_dma_flush_range);
-EXPORT_SYMBOL(__cpuc_dma_clean_range);
-- 
2.14.1

[PATCH 09/11] xtensa: make dma_cache_sync a no-op

[Christoph Hellwig]

xtensa does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/xtensa/include/asm/dma-mapping.h |  6 ++++--
 arch/xtensa/kernel/pci-dma.c          | 23 -----------------------
 2 files changed, 4 insertions(+), 25 deletions(-)

diff --git a/arch/xtensa/include/asm/dma-mapping.h b/arch/xtensa/include/asm/dma-mapping.h
index 269738dc9d1d..353e0314d6ba 100644
--- a/arch/xtensa/include/asm/dma-mapping.h
+++ b/arch/xtensa/include/asm/dma-mapping.h
@@ -23,8 +23,10 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &xtensa_dma_map_ops;
 }
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction direction);
+static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+		    enum dma_data_direction direction)
+{
+}
 
 static inline dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr)
 {
diff --git a/arch/xtensa/kernel/pci-dma.c b/arch/xtensa/kernel/pci-dma.c
index cec86a1c2acc..623720a11143 100644
--- a/arch/xtensa/kernel/pci-dma.c
+++ b/arch/xtensa/kernel/pci-dma.c
@@ -26,29 +26,6 @@
 #include <asm/cacheflush.h>
 #include <asm/io.h>
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction dir)
-{
-	switch (dir) {
-	case DMA_BIDIRECTIONAL:
-		__flush_invalidate_dcache_range((unsigned long)vaddr, size);
-		break;
-
-	case DMA_FROM_DEVICE:
-		__invalidate_dcache_range((unsigned long)vaddr, size);
-		break;
-
-	case DMA_TO_DEVICE:
-		__flush_dcache_range((unsigned long)vaddr, size);
-		break;
-
-	case DMA_NONE:
-		BUG();
-		break;
-	}
-}
-EXPORT_SYMBOL(dma_cache_sync);
-
 static void do_cache_op(dma_addr_t dma_handle, size_t size,
 			void (*fn)(unsigned long, unsigned long))
 {
-- 
2.14.1

[PATCH 09/11] xtensa: make dma_cache_sync a no-op

[Christoph Hellwig]

xtensa does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/xtensa/include/asm/dma-mapping.h |  6 ++++--
 arch/xtensa/kernel/pci-dma.c          | 23 -----------------------
 2 files changed, 4 insertions(+), 25 deletions(-)

diff --git a/arch/xtensa/include/asm/dma-mapping.h b/arch/xtensa/include/asm/dma-mapping.h
index 269738dc9d1d..353e0314d6ba 100644
--- a/arch/xtensa/include/asm/dma-mapping.h
+++ b/arch/xtensa/include/asm/dma-mapping.h
@@ -23,8 +23,10 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &xtensa_dma_map_ops;
 }
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction direction);
+static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+		    enum dma_data_direction direction)
+{
+}
 
 static inline dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr)
 {
diff --git a/arch/xtensa/kernel/pci-dma.c b/arch/xtensa/kernel/pci-dma.c
index cec86a1c2acc..623720a11143 100644
--- a/arch/xtensa/kernel/pci-dma.c
+++ b/arch/xtensa/kernel/pci-dma.c
@@ -26,29 +26,6 @@
 #include <asm/cacheflush.h>
 #include <asm/io.h>
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction dir)
-{
-	switch (dir) {
-	case DMA_BIDIRECTIONAL:
-		__flush_invalidate_dcache_range((unsigned long)vaddr, size);
-		break;
-
-	case DMA_FROM_DEVICE:
-		__invalidate_dcache_range((unsigned long)vaddr, size);
-		break;
-
-	case DMA_TO_DEVICE:
-		__flush_dcache_range((unsigned long)vaddr, size);
-		break;
-
-	case DMA_NONE:
-		BUG();
-		break;
-	}
-}
-EXPORT_SYMBOL(dma_cache_sync);
-
 static void do_cache_op(dma_addr_t dma_handle, size_t size,
 			void (*fn)(unsigned long, unsigned long))
 {
-- 
2.14.1

[PATCH 09/11] xtensa: make dma_cache_sync a no-op

[Christoph Hellwig]

xtensa does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
doesn't make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 arch/xtensa/include/asm/dma-mapping.h |  6 ++++--
 arch/xtensa/kernel/pci-dma.c          | 23 -----------------------
 2 files changed, 4 insertions(+), 25 deletions(-)

diff --git a/arch/xtensa/include/asm/dma-mapping.h b/arch/xtensa/include/asm/dma-mapping.h
index 269738dc9d1d..353e0314d6ba 100644
--- a/arch/xtensa/include/asm/dma-mapping.h
+++ b/arch/xtensa/include/asm/dma-mapping.h
@@ -23,8 +23,10 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &xtensa_dma_map_ops;
 }
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction direction);
+static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+		    enum dma_data_direction direction)
+{
+}
 
 static inline dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr)
 {
diff --git a/arch/xtensa/kernel/pci-dma.c b/arch/xtensa/kernel/pci-dma.c
index cec86a1c2acc..623720a11143 100644
--- a/arch/xtensa/kernel/pci-dma.c
+++ b/arch/xtensa/kernel/pci-dma.c
@@ -26,29 +26,6 @@
 #include <asm/cacheflush.h>
 #include <asm/io.h>
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction dir)
-{
-	switch (dir) {
-	case DMA_BIDIRECTIONAL:
-		__flush_invalidate_dcache_range((unsigned long)vaddr, size);
-		break;
-
-	case DMA_FROM_DEVICE:
-		__invalidate_dcache_range((unsigned long)vaddr, size);
-		break;
-
-	case DMA_TO_DEVICE:
-		__flush_dcache_range((unsigned long)vaddr, size);
-		break;
-
-	case DMA_NONE:
-		BUG();
-		break;
-	}
-}
-EXPORT_SYMBOL(dma_cache_sync);
-
 static void do_cache_op(dma_addr_t dma_handle, size_t size,
 			void (*fn)(unsigned long, unsigned long))
 {
-- 
2.14.1

[PATCH 10/11] sh: make dma_cache_sync a no-op

[Christoph Hellwig]

sh does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

On the other hand sh uses dma_cache_sync internally in the dma_ops
implementation and for the maple bus that does not use the DMA API,
so a the old functionality for dma_cache_sync is still provided under
the name sh_sync_dma_for_device, and without the redundant dev
argument.  While at it two of the syncing dma_ops also go the proper
_for_device postfix.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/include/asm/dma-mapping.h |  9 +++++++--
 arch/sh/kernel/dma-nommu.c        | 17 +++++++++--------
 arch/sh/mm/consistent.c           |  6 +++---
 drivers/sh/maple/maple.c          |  5 ++---
 4 files changed, 21 insertions(+), 16 deletions(-)

diff --git a/arch/sh/include/asm/dma-mapping.h b/arch/sh/include/asm/dma-mapping.h
index 9b06be07db4d..b46194ecef17 100644
--- a/arch/sh/include/asm/dma-mapping.h
+++ b/arch/sh/include/asm/dma-mapping.h
@@ -9,8 +9,10 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return dma_ops;
 }
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction dir);
+static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+		    enum dma_data_direction dir)
+{
+}
 
 /* arch/sh/mm/consistent.c */
 extern void *dma_generic_alloc_coherent(struct device *dev, size_t size,
@@ -20,4 +22,7 @@ extern void dma_generic_free_coherent(struct device *dev, size_t size,
 				      void *vaddr, dma_addr_t dma_handle,
 				      unsigned long attrs);
 
+void sh_sync_dma_for_device(void *vaddr, size_t size,
+	    enum dma_data_direction dir);
+
 #endif /* __ASM_SH_DMA_MAPPING_H */
diff --git a/arch/sh/kernel/dma-nommu.c b/arch/sh/kernel/dma-nommu.c
index d24c707b2181..62b485107eae 100644
--- a/arch/sh/kernel/dma-nommu.c
+++ b/arch/sh/kernel/dma-nommu.c
@@ -9,6 +9,7 @@
  */
 #include <linux/dma-mapping.h>
 #include <linux/io.h>
+#include <asm/cacheflush.h>
 
 static dma_addr_t nommu_map_page(struct device *dev, struct page *page,
 				 unsigned long offset, size_t size,
@@ -20,7 +21,7 @@ static dma_addr_t nommu_map_page(struct device *dev, struct page *page,
 	WARN_ON(size = 0);
 
 	if (!(attrs & DMA_ATTR_SKIP_CPU_SYNC))
-		dma_cache_sync(dev, page_address(page) + offset, size, dir);
+		sh_sync_dma_for_device(page_address(page) + offset, size, dir);
 
 	return addr;
 }
@@ -38,7 +39,7 @@ static int nommu_map_sg(struct device *dev, struct scatterlist *sg,
 		BUG_ON(!sg_page(s));
 
 		if (!(attrs & DMA_ATTR_SKIP_CPU_SYNC))
-			dma_cache_sync(dev, sg_virt(s), s->length, dir);
+			sh_sync_dma_for_device(sg_virt(s), s->length, dir);
 
 		s->dma_address = sg_phys(s);
 		s->dma_length = s->length;
@@ -48,20 +49,20 @@ static int nommu_map_sg(struct device *dev, struct scatterlist *sg,
 }
 
 #ifdef CONFIG_DMA_NONCOHERENT
-static void nommu_sync_single(struct device *dev, dma_addr_t addr,
+static void nommu_sync_single_for_device(struct device *dev, dma_addr_t addr,
 			      size_t size, enum dma_data_direction dir)
 {
-	dma_cache_sync(dev, phys_to_virt(addr), size, dir);
+	sh_sync_dma_for_device(phys_to_virt(addr), size, dir);
 }
 
-static void nommu_sync_sg(struct device *dev, struct scatterlist *sg,
+static void nommu_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
 			  int nelems, enum dma_data_direction dir)
 {
 	struct scatterlist *s;
 	int i;
 
 	for_each_sg(sg, s, nelems, i)
-		dma_cache_sync(dev, sg_virt(s), s->length, dir);
+		sh_sync_dma_for_device(sg_virt(s), s->length, dir);
 }
 #endif
 
@@ -71,8 +72,8 @@ const struct dma_map_ops nommu_dma_ops = {
 	.map_page		= nommu_map_page,
 	.map_sg			= nommu_map_sg,
 #ifdef CONFIG_DMA_NONCOHERENT
-	.sync_single_for_device	= nommu_sync_single,
-	.sync_sg_for_device	= nommu_sync_sg,
+	.sync_single_for_device	= nommu_sync_single_for_device,
+	.sync_sg_for_device	= nommu_sync_sg_for_device,
 #endif
 	.is_phys		= 1,
 };
diff --git a/arch/sh/mm/consistent.c b/arch/sh/mm/consistent.c
index d1275adfa0ef..6ea3aab508f2 100644
--- a/arch/sh/mm/consistent.c
+++ b/arch/sh/mm/consistent.c
@@ -49,7 +49,7 @@ void *dma_generic_alloc_coherent(struct device *dev, size_t size,
 	 * Pages from the page allocator may have data present in
 	 * cache. So flush the cache before using uncached memory.
 	 */
-	dma_cache_sync(dev, ret, size, DMA_BIDIRECTIONAL);
+	sh_sync_dma_for_device(ret, size, DMA_BIDIRECTIONAL);
 
 	ret_nocache = (void __force *)ioremap_nocache(virt_to_phys(ret), size);
 	if (!ret_nocache) {
@@ -78,7 +78,7 @@ void dma_generic_free_coherent(struct device *dev, size_t size,
 	iounmap(vaddr);
 }
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+void sh_sync_dma_for_device(void *vaddr, size_t size,
 		    enum dma_data_direction direction)
 {
 	void *addr;
@@ -100,7 +100,7 @@ void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		BUG();
 	}
 }
-EXPORT_SYMBOL(dma_cache_sync);
+EXPORT_SYMBOL(sh_sync_dma_for_device);
 
 static int __init memchunk_setup(char *str)
 {
diff --git a/drivers/sh/maple/maple.c b/drivers/sh/maple/maple.c
index bec81c2404f7..7525039d812c 100644
--- a/drivers/sh/maple/maple.c
+++ b/drivers/sh/maple/maple.c
@@ -300,7 +300,7 @@ static void maple_send(void)
 	mutex_unlock(&maple_wlist_lock);
 	if (maple_packets > 0) {
 		for (i = 0; i < (1 << MAPLE_DMA_PAGES); i++)
-			dma_cache_sync(0, maple_sendbuf + i * PAGE_SIZE,
+			sh_sync_dma_for_device(maple_sendbuf + i * PAGE_SIZE,
 				       PAGE_SIZE, DMA_BIDIRECTIONAL);
 	}
 
@@ -642,8 +642,7 @@ static void maple_dma_handler(struct work_struct *work)
 		list_for_each_entry_safe(mq, nmq, &maple_sentq, list) {
 			mdev = mq->dev;
 			recvbuf = mq->recvbuf->buf;
-			dma_cache_sync(&mdev->dev, recvbuf, 0x400,
-				DMA_FROM_DEVICE);
+			sh_sync_dma_for_device(recvbuf, 0x400, DMA_FROM_DEVICE);
 			code = recvbuf[0];
 			kfree(mq->sendbuf);
 			list_del_init(&mq->list);
-- 
2.14.1

[PATCH 10/11] sh: make dma_cache_sync a no-op

[Christoph Hellwig]

sh does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
make any sense to do any work in dma_cache_sync given that it
must be a no-op when dma_alloc_attrs returns coherent memory.

On the other hand sh uses dma_cache_sync internally in the dma_ops
implementation and for the maple bus that does not use the DMA API,
so a the old functionality for dma_cache_sync is still provided under
the name sh_sync_dma_for_device, and without the redundant dev
argument.  While at it two of the syncing dma_ops also go the proper
_for_device postfix.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/sh/include/asm/dma-mapping.h |  9 +++++++--
 arch/sh/kernel/dma-nommu.c        | 17 +++++++++--------
 arch/sh/mm/consistent.c           |  6 +++---
 drivers/sh/maple/maple.c          |  5 ++---
 4 files changed, 21 insertions(+), 16 deletions(-)

diff --git a/arch/sh/include/asm/dma-mapping.h b/arch/sh/include/asm/dma-mapping.h
index 9b06be07db4d..b46194ecef17 100644
--- a/arch/sh/include/asm/dma-mapping.h
+++ b/arch/sh/include/asm/dma-mapping.h
@@ -9,8 +9,10 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return dma_ops;
 }
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction dir);
+static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+		    enum dma_data_direction dir)
+{
+}
 
 /* arch/sh/mm/consistent.c */
 extern void *dma_generic_alloc_coherent(struct device *dev, size_t size,
@@ -20,4 +22,7 @@ extern void dma_generic_free_coherent(struct device *dev, size_t size,
 				      void *vaddr, dma_addr_t dma_handle,
 				      unsigned long attrs);
 
+void sh_sync_dma_for_device(void *vaddr, size_t size,
+	    enum dma_data_direction dir);
+
 #endif /* __ASM_SH_DMA_MAPPING_H */
diff --git a/arch/sh/kernel/dma-nommu.c b/arch/sh/kernel/dma-nommu.c
index d24c707b2181..62b485107eae 100644
--- a/arch/sh/kernel/dma-nommu.c
+++ b/arch/sh/kernel/dma-nommu.c
@@ -9,6 +9,7 @@
  */
 #include <linux/dma-mapping.h>
 #include <linux/io.h>
+#include <asm/cacheflush.h>
 
 static dma_addr_t nommu_map_page(struct device *dev, struct page *page,
 				 unsigned long offset, size_t size,
@@ -20,7 +21,7 @@ static dma_addr_t nommu_map_page(struct device *dev, struct page *page,
 	WARN_ON(size == 0);
 
 	if (!(attrs & DMA_ATTR_SKIP_CPU_SYNC))
-		dma_cache_sync(dev, page_address(page) + offset, size, dir);
+		sh_sync_dma_for_device(page_address(page) + offset, size, dir);
 
 	return addr;
 }
@@ -38,7 +39,7 @@ static int nommu_map_sg(struct device *dev, struct scatterlist *sg,
 		BUG_ON(!sg_page(s));
 
 		if (!(attrs & DMA_ATTR_SKIP_CPU_SYNC))
-			dma_cache_sync(dev, sg_virt(s), s->length, dir);
+			sh_sync_dma_for_device(sg_virt(s), s->length, dir);
 
 		s->dma_address = sg_phys(s);
 		s->dma_length = s->length;
@@ -48,20 +49,20 @@ static int nommu_map_sg(struct device *dev, struct scatterlist *sg,
 }
 
 #ifdef CONFIG_DMA_NONCOHERENT
-static void nommu_sync_single(struct device *dev, dma_addr_t addr,
+static void nommu_sync_single_for_device(struct device *dev, dma_addr_t addr,
 			      size_t size, enum dma_data_direction dir)
 {
-	dma_cache_sync(dev, phys_to_virt(addr), size, dir);
+	sh_sync_dma_for_device(phys_to_virt(addr), size, dir);
 }
 
-static void nommu_sync_sg(struct device *dev, struct scatterlist *sg,
+static void nommu_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
 			  int nelems, enum dma_data_direction dir)
 {
 	struct scatterlist *s;
 	int i;
 
 	for_each_sg(sg, s, nelems, i)
-		dma_cache_sync(dev, sg_virt(s), s->length, dir);
+		sh_sync_dma_for_device(sg_virt(s), s->length, dir);
 }
 #endif
 
@@ -71,8 +72,8 @@ const struct dma_map_ops nommu_dma_ops = {
 	.map_page		= nommu_map_page,
 	.map_sg			= nommu_map_sg,
 #ifdef CONFIG_DMA_NONCOHERENT
-	.sync_single_for_device	= nommu_sync_single,
-	.sync_sg_for_device	= nommu_sync_sg,
+	.sync_single_for_device	= nommu_sync_single_for_device,
+	.sync_sg_for_device	= nommu_sync_sg_for_device,
 #endif
 	.is_phys		= 1,
 };
diff --git a/arch/sh/mm/consistent.c b/arch/sh/mm/consistent.c
index d1275adfa0ef..6ea3aab508f2 100644
--- a/arch/sh/mm/consistent.c
+++ b/arch/sh/mm/consistent.c
@@ -49,7 +49,7 @@ void *dma_generic_alloc_coherent(struct device *dev, size_t size,
 	 * Pages from the page allocator may have data present in
 	 * cache. So flush the cache before using uncached memory.
 	 */
-	dma_cache_sync(dev, ret, size, DMA_BIDIRECTIONAL);
+	sh_sync_dma_for_device(ret, size, DMA_BIDIRECTIONAL);
 
 	ret_nocache = (void __force *)ioremap_nocache(virt_to_phys(ret), size);
 	if (!ret_nocache) {
@@ -78,7 +78,7 @@ void dma_generic_free_coherent(struct device *dev, size_t size,
 	iounmap(vaddr);
 }
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+void sh_sync_dma_for_device(void *vaddr, size_t size,
 		    enum dma_data_direction direction)
 {
 	void *addr;
@@ -100,7 +100,7 @@ void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		BUG();
 	}
 }
-EXPORT_SYMBOL(dma_cache_sync);
+EXPORT_SYMBOL(sh_sync_dma_for_device);
 
 static int __init memchunk_setup(char *str)
 {
diff --git a/drivers/sh/maple/maple.c b/drivers/sh/maple/maple.c
index bec81c2404f7..7525039d812c 100644
--- a/drivers/sh/maple/maple.c
+++ b/drivers/sh/maple/maple.c
@@ -300,7 +300,7 @@ static void maple_send(void)
 	mutex_unlock(&maple_wlist_lock);
 	if (maple_packets > 0) {
 		for (i = 0; i < (1 << MAPLE_DMA_PAGES); i++)
-			dma_cache_sync(0, maple_sendbuf + i * PAGE_SIZE,
+			sh_sync_dma_for_device(maple_sendbuf + i * PAGE_SIZE,
 				       PAGE_SIZE, DMA_BIDIRECTIONAL);
 	}
 
@@ -642,8 +642,7 @@ static void maple_dma_handler(struct work_struct *work)
 		list_for_each_entry_safe(mq, nmq, &maple_sentq, list) {
 			mdev = mq->dev;
 			recvbuf = mq->recvbuf->buf;
-			dma_cache_sync(&mdev->dev, recvbuf, 0x400,
-				DMA_FROM_DEVICE);
+			sh_sync_dma_for_device(recvbuf, 0x400, DMA_FROM_DEVICE);
 			code = recvbuf[0];
 			kfree(mq->sendbuf);
 			list_del_init(&mq->list);
-- 
2.14.1

[PATCH 11/11] dma-mapping: turn dma_cache_sync into a dma_map_ops method

[Christoph Hellwig]

After we removed all the dead wood it turns out only two architectures
actually implement dma_cache_sync as a real op: mips and parisc.  Add
a cache_sync method to struct dma_map_ops and implement it for the
mips defualt DMA ops, and the parisc pa11 ops.

Note that arm, arc and openrisc support DMA_ATTR_NON_CONSISTENT, but
never provided a functional dma_cache_sync implementations, which
seems somewhat odd.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/alpha/include/asm/dma-mapping.h      |  2 --
 arch/cris/include/asm/dma-mapping.h       |  6 ------
 arch/frv/include/asm/dma-mapping.h        |  6 ------
 arch/hexagon/include/asm/dma-mapping.h    |  3 ---
 arch/ia64/include/asm/dma-mapping.h       |  6 ------
 arch/m32r/include/asm/dma-mapping.h       |  5 -----
 arch/m68k/include/asm/dma-mapping.h       |  6 ------
 arch/metag/include/asm/dma-mapping.h      | 10 ----------
 arch/microblaze/include/asm/dma-mapping.h |  5 -----
 arch/mips/include/asm/dma-mapping.h       |  3 ---
 arch/mips/mm/dma-default.c                |  7 +++----
 arch/mn10300/include/asm/dma-mapping.h    |  6 ------
 arch/nios2/include/asm/dma-mapping.h      |  9 ---------
 arch/parisc/include/asm/dma-mapping.h     |  8 --------
 arch/parisc/kernel/pci-dma.c              |  8 ++++++++
 arch/powerpc/include/asm/dma-mapping.h    |  5 -----
 arch/s390/include/asm/dma-mapping.h       |  5 -----
 arch/sh/include/asm/dma-mapping.h         |  6 ------
 arch/sparc/include/asm/dma-mapping.h      |  8 --------
 arch/tile/include/asm/dma-mapping.h       |  9 ---------
 arch/unicore32/include/asm/dma-mapping.h  |  5 -----
 arch/x86/include/asm/dma-mapping.h        |  6 ------
 arch/xtensa/include/asm/dma-mapping.h     |  5 -----
 include/linux/dma-mapping.h               | 13 +++++++++++++
 24 files changed, 24 insertions(+), 128 deletions(-)

diff --git a/arch/alpha/include/asm/dma-mapping.h b/arch/alpha/include/asm/dma-mapping.h
index 5d53666935e6..399a4f49355e 100644
--- a/arch/alpha/include/asm/dma-mapping.h
+++ b/arch/alpha/include/asm/dma-mapping.h
@@ -8,6 +8,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return dma_ops;
 }
 
-#define dma_cache_sync(dev, va, size, dir)		  ((void)0)
-
 #endif	/* _ALPHA_DMA_MAPPING_H */
diff --git a/arch/cris/include/asm/dma-mapping.h b/arch/cris/include/asm/dma-mapping.h
index 256169de3743..e30adde42beb 100644
--- a/arch/cris/include/asm/dma-mapping.h
+++ b/arch/cris/include/asm/dma-mapping.h
@@ -16,10 +16,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 }
 #endif
 
-static inline void
-dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-	       enum dma_data_direction direction)
-{
-}
-
 #endif
diff --git a/arch/frv/include/asm/dma-mapping.h b/arch/frv/include/asm/dma-mapping.h
index da0e5c9744c4..da24ae943f02 100644
--- a/arch/frv/include/asm/dma-mapping.h
+++ b/arch/frv/include/asm/dma-mapping.h
@@ -14,10 +14,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &frv_dma_ops;
 }
 
-static inline
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction direction)
-{
-}
-
 #endif  /* _ASM_DMA_MAPPING_H */
diff --git a/arch/hexagon/include/asm/dma-mapping.h b/arch/hexagon/include/asm/dma-mapping.h
index 463dbc18f853..5208de242e79 100644
--- a/arch/hexagon/include/asm/dma-mapping.h
+++ b/arch/hexagon/include/asm/dma-mapping.h
@@ -37,9 +37,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return dma_ops;
 }
 
-extern void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-			   enum dma_data_direction direction);
-
 static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
 {
 	if (!dev->dma_mask)
diff --git a/arch/ia64/include/asm/dma-mapping.h b/arch/ia64/include/asm/dma-mapping.h
index 99dfc1aa9d3c..9e5b5df76ff8 100644
--- a/arch/ia64/include/asm/dma-mapping.h
+++ b/arch/ia64/include/asm/dma-mapping.h
@@ -44,10 +44,4 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
 	return daddr;
 }
 
-static inline void
-dma_cache_sync (struct device *dev, void *vaddr, size_t size,
-	enum dma_data_direction dir)
-{
-}
-
 #endif /* _ASM_IA64_DMA_MAPPING_H */
diff --git a/arch/m32r/include/asm/dma-mapping.h b/arch/m32r/include/asm/dma-mapping.h
index aff3ae8b62f7..9e993daed7a0 100644
--- a/arch/m32r/include/asm/dma-mapping.h
+++ b/arch/m32r/include/asm/dma-mapping.h
@@ -13,11 +13,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &dma_noop_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction direction)
-{
-}
-
 static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
 {
 	if (!dev->dma_mask)
diff --git a/arch/m68k/include/asm/dma-mapping.h b/arch/m68k/include/asm/dma-mapping.h
index 9210e470771b..9a0d559fcc13 100644
--- a/arch/m68k/include/asm/dma-mapping.h
+++ b/arch/m68k/include/asm/dma-mapping.h
@@ -8,10 +8,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
         return &m68k_dma_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction dir)
-{
-	/* we use coherent allocation, so not much to do here. */
-}
-
 #endif  /* _M68K_DMA_MAPPING_H */
diff --git a/arch/metag/include/asm/dma-mapping.h b/arch/metag/include/asm/dma-mapping.h
index ea573be2b6d0..340265dcf839 100644
--- a/arch/metag/include/asm/dma-mapping.h
+++ b/arch/metag/include/asm/dma-mapping.h
@@ -8,14 +8,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &metag_dma_ops;
 }
 
-/*
- * dma_alloc_attrs() always returns non-cacheable memory, so there's no need to
- * do any flushing here.
- */
-static inline void
-dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-	       enum dma_data_direction direction)
-{
-}
-
 #endif
diff --git a/arch/microblaze/include/asm/dma-mapping.h b/arch/microblaze/include/asm/dma-mapping.h
index ad448e4aedb6..6b9ea39405b8 100644
--- a/arch/microblaze/include/asm/dma-mapping.h
+++ b/arch/microblaze/include/asm/dma-mapping.h
@@ -25,9 +25,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &dma_direct_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		enum dma_data_direction direction)
-{
-}
-
 #endif	/* _ASM_MICROBLAZE_DMA_MAPPING_H */
diff --git a/arch/mips/include/asm/dma-mapping.h b/arch/mips/include/asm/dma-mapping.h
index aba71385f9d1..6ea1439430a2 100644
--- a/arch/mips/include/asm/dma-mapping.h
+++ b/arch/mips/include/asm/dma-mapping.h
@@ -26,9 +26,6 @@ static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
 
 static inline void dma_mark_clean(void *addr, size_t size) {}
 
-extern void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-	       enum dma_data_direction direction);
-
 #define arch_setup_dma_ops arch_setup_dma_ops
 static inline void arch_setup_dma_ops(struct device *dev, u64 dma_base,
 				      u64 size, const struct iommu_ops *iommu,
diff --git a/arch/mips/mm/dma-default.c b/arch/mips/mm/dma-default.c
index c01bd20d0208..2e2514e00720 100644
--- a/arch/mips/mm/dma-default.c
+++ b/arch/mips/mm/dma-default.c
@@ -383,7 +383,7 @@ static int mips_dma_supported(struct device *dev, u64 mask)
 	return plat_dma_supported(dev, mask);
 }
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+static void mips_dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 			 enum dma_data_direction direction)
 {
 	BUG_ON(direction = DMA_NONE);
@@ -392,8 +392,6 @@ void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		__dma_sync_virtual(vaddr, size, direction);
 }
 
-EXPORT_SYMBOL(dma_cache_sync);
-
 static const struct dma_map_ops mips_default_dma_map_ops = {
 	.alloc = mips_dma_alloc_coherent,
 	.free = mips_dma_free_coherent,
@@ -407,7 +405,8 @@ static const struct dma_map_ops mips_default_dma_map_ops = {
 	.sync_sg_for_cpu = mips_dma_sync_sg_for_cpu,
 	.sync_sg_for_device = mips_dma_sync_sg_for_device,
 	.mapping_error = mips_dma_mapping_error,
-	.dma_supported = mips_dma_supported
+	.dma_supported = mips_dma_supported,
+	.cache_sync = mips_dma_cache_sync,
 };
 
 const struct dma_map_ops *mips_dma_map_ops = &mips_default_dma_map_ops;
diff --git a/arch/mn10300/include/asm/dma-mapping.h b/arch/mn10300/include/asm/dma-mapping.h
index dc24163b190f..439e474ed6d7 100644
--- a/arch/mn10300/include/asm/dma-mapping.h
+++ b/arch/mn10300/include/asm/dma-mapping.h
@@ -18,10 +18,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &mn10300_dma_ops;
 }
 
-static inline
-void dma_cache_sync(void *vaddr, size_t size,
-		    enum dma_data_direction direction)
-{
-}
-
 #endif
diff --git a/arch/nios2/include/asm/dma-mapping.h b/arch/nios2/include/asm/dma-mapping.h
index f8dc62222741..6ceb92251da0 100644
--- a/arch/nios2/include/asm/dma-mapping.h
+++ b/arch/nios2/include/asm/dma-mapping.h
@@ -17,13 +17,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &nios2_dma_ops;
 }
 
-/*
- * dma_alloc_attrs() always returns non-cacheable memory, so there's no need to
- * do any flushing here.
- */
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction direction)
-{
-}
-
 #endif /* _ASM_NIOS2_DMA_MAPPING_H */
diff --git a/arch/parisc/include/asm/dma-mapping.h b/arch/parisc/include/asm/dma-mapping.h
index 2b16282add69..cb26bbd71d8a 100644
--- a/arch/parisc/include/asm/dma-mapping.h
+++ b/arch/parisc/include/asm/dma-mapping.h
@@ -32,14 +32,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return hppa_dma_ops;
 }
 
-static inline void
-dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-	       enum dma_data_direction direction)
-{
-	if (hppa_dma_ops->sync_single_for_cpu)
-		flush_kernel_dcache_range((unsigned long)vaddr, size);
-}
-
 static inline void *
 parisc_walk_tree(struct device *dev)
 {
diff --git a/arch/parisc/kernel/pci-dma.c b/arch/parisc/kernel/pci-dma.c
index bd4c0a7471d3..f87c34cb3b43 100644
--- a/arch/parisc/kernel/pci-dma.c
+++ b/arch/parisc/kernel/pci-dma.c
@@ -571,6 +571,12 @@ static void pa11_dma_sync_sg_for_device(struct device *dev, struct scatterlist *
 		flush_kernel_vmap_range(sg_virt(sg), sg->length);
 }
 
+static void pa11_dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+	       enum dma_data_direction direction)
+{
+	flush_kernel_dcache_range((unsigned long)vaddr, size);
+}
+
 const struct dma_map_ops pcxl_dma_ops = {
 	.dma_supported =	pa11_dma_supported,
 	.alloc =		pa11_dma_alloc,
@@ -583,6 +589,7 @@ const struct dma_map_ops pcxl_dma_ops = {
 	.sync_single_for_device = pa11_dma_sync_single_for_device,
 	.sync_sg_for_cpu =	pa11_dma_sync_sg_for_cpu,
 	.sync_sg_for_device =	pa11_dma_sync_sg_for_device,
+	.cache_sync =		pa11_dma_cache_sync,
 };
 
 static void *pcx_dma_alloc(struct device *dev, size_t size,
@@ -619,4 +626,5 @@ const struct dma_map_ops pcx_dma_ops = {
 	.sync_single_for_device = pa11_dma_sync_single_for_device,
 	.sync_sg_for_cpu =	pa11_dma_sync_sg_for_cpu,
 	.sync_sg_for_device =	pa11_dma_sync_sg_for_device,
+	.cache_sync =		pa11_dma_cache_sync,
 };
diff --git a/arch/powerpc/include/asm/dma-mapping.h b/arch/powerpc/include/asm/dma-mapping.h
index 320846442bfb..2e43c2ef7632 100644
--- a/arch/powerpc/include/asm/dma-mapping.h
+++ b/arch/powerpc/include/asm/dma-mapping.h
@@ -141,10 +141,5 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
 
 #define ARCH_HAS_DMA_MMAP_COHERENT
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		enum dma_data_direction direction)
-{
-}
-
 #endif /* __KERNEL__ */
 #endif	/* _ASM_DMA_MAPPING_H */
diff --git a/arch/s390/include/asm/dma-mapping.h b/arch/s390/include/asm/dma-mapping.h
index 512ad0eaa11a..b17304b13de5 100644
--- a/arch/s390/include/asm/dma-mapping.h
+++ b/arch/s390/include/asm/dma-mapping.h
@@ -15,11 +15,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &dma_noop_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction direction)
-{
-}
-
 static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
 {
 	if (!dev->dma_mask)
diff --git a/arch/sh/include/asm/dma-mapping.h b/arch/sh/include/asm/dma-mapping.h
index b46194ecef17..e89df111c017 100644
--- a/arch/sh/include/asm/dma-mapping.h
+++ b/arch/sh/include/asm/dma-mapping.h
@@ -9,12 +9,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return dma_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction dir)
-{
-}
-
-/* arch/sh/mm/consistent.c */
 extern void *dma_generic_alloc_coherent(struct device *dev, size_t size,
 					dma_addr_t *dma_addr, gfp_t flag,
 					unsigned long attrs);
diff --git a/arch/sparc/include/asm/dma-mapping.h b/arch/sparc/include/asm/dma-mapping.h
index 60bf1633d554..b298ed45cb23 100644
--- a/arch/sparc/include/asm/dma-mapping.h
+++ b/arch/sparc/include/asm/dma-mapping.h
@@ -5,14 +5,6 @@
 #include <linux/mm.h>
 #include <linux/dma-debug.h>
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction dir)
-{
-	/* Since dma_{alloc,free}_noncoherent() allocated coherent memory, this
-	 * routine can be a nop.
-	 */
-}
-
 extern const struct dma_map_ops *dma_ops;
 extern const struct dma_map_ops pci32_dma_ops;
 
diff --git a/arch/tile/include/asm/dma-mapping.h b/arch/tile/include/asm/dma-mapping.h
index 7061dc8af43a..97ad62878290 100644
--- a/arch/tile/include/asm/dma-mapping.h
+++ b/arch/tile/include/asm/dma-mapping.h
@@ -67,13 +67,4 @@ static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
 #define HAVE_ARCH_DMA_SET_MASK 1
 int dma_set_mask(struct device *dev, u64 mask);
 
-/*
- * dma_alloc_attrs() always returns non-cacheable memory, so there's no need to
- * do any flushing here.
- */
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction direction)
-{
-}
-
 #endif /* _ASM_TILE_DMA_MAPPING_H */
diff --git a/arch/unicore32/include/asm/dma-mapping.h b/arch/unicore32/include/asm/dma-mapping.h
index e949855bb794..ac608c2f6af6 100644
--- a/arch/unicore32/include/asm/dma-mapping.h
+++ b/arch/unicore32/include/asm/dma-mapping.h
@@ -45,10 +45,5 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
 
 static inline void dma_mark_clean(void *addr, size_t size) {}
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr,
-		size_t size, enum dma_data_direction direction)
-{
-}
-
 #endif /* __KERNEL__ */
 #endif
diff --git a/arch/x86/include/asm/dma-mapping.h b/arch/x86/include/asm/dma-mapping.h
index c6d3367be916..c26747340ba5 100644
--- a/arch/x86/include/asm/dma-mapping.h
+++ b/arch/x86/include/asm/dma-mapping.h
@@ -67,12 +67,6 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
 }
 #endif /* CONFIG_X86_DMA_REMAP */
 
-static inline void
-dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-	enum dma_data_direction dir)
-{
-}
-
 static inline unsigned long dma_alloc_coherent_mask(struct device *dev,
 						    gfp_t gfp)
 {
diff --git a/arch/xtensa/include/asm/dma-mapping.h b/arch/xtensa/include/asm/dma-mapping.h
index 353e0314d6ba..153bf2370988 100644
--- a/arch/xtensa/include/asm/dma-mapping.h
+++ b/arch/xtensa/include/asm/dma-mapping.h
@@ -23,11 +23,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &xtensa_dma_map_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction direction)
-{
-}
-
 static inline dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr)
 {
 	return (dma_addr_t)paddr;
diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
index 29ce9815da87..028a375d240d 100644
--- a/include/linux/dma-mapping.h
+++ b/include/linux/dma-mapping.h
@@ -126,6 +126,8 @@ struct dma_map_ops {
 	void (*sync_sg_for_device)(struct device *dev,
 				   struct scatterlist *sg, int nents,
 				   enum dma_data_direction dir);
+	void (*cache_sync)(struct device *dev, void *vaddr, size_t size,
+			enum dma_data_direction direction);
 	int (*mapping_error)(struct device *dev, dma_addr_t dma_addr);
 	int (*dma_supported)(struct device *dev, u64 mask);
 #ifdef ARCH_HAS_DMA_GET_REQUIRED_MASK
@@ -436,6 +438,17 @@ dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
 #define dma_map_page(d, p, o, s, r) dma_map_page_attrs(d, p, o, s, r, 0)
 #define dma_unmap_page(d, a, s, r) dma_unmap_page_attrs(d, a, s, r, 0)
 
+static inline void
+dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+		enum dma_data_direction dir)
+{
+	const struct dma_map_ops *ops = get_dma_ops(dev);
+
+	BUG_ON(!valid_dma_direction(dir));
+	if (ops->cache_sync)
+		ops->cache_sync(dev, vaddr, size, dir);
+}
+
 extern int dma_common_mmap(struct device *dev, struct vm_area_struct *vma,
 			   void *cpu_addr, dma_addr_t dma_addr, size_t size);
 
-- 
2.14.1

[PATCH 11/11] dma-mapping: turn dma_cache_sync into a dma_map_ops method

[Christoph Hellwig]

After we removed all the dead wood it turns out only two architectures
actually implement dma_cache_sync as a real op: mips and parisc.  Add
a cache_sync method to struct dma_map_ops and implement it for the
mips defualt DMA ops, and the parisc pa11 ops.

Note that arm, arc and openrisc support DMA_ATTR_NON_CONSISTENT, but
never provided a functional dma_cache_sync implementations, which
seems somewhat odd.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/alpha/include/asm/dma-mapping.h      |  2 --
 arch/cris/include/asm/dma-mapping.h       |  6 ------
 arch/frv/include/asm/dma-mapping.h        |  6 ------
 arch/hexagon/include/asm/dma-mapping.h    |  3 ---
 arch/ia64/include/asm/dma-mapping.h       |  6 ------
 arch/m32r/include/asm/dma-mapping.h       |  5 -----
 arch/m68k/include/asm/dma-mapping.h       |  6 ------
 arch/metag/include/asm/dma-mapping.h      | 10 ----------
 arch/microblaze/include/asm/dma-mapping.h |  5 -----
 arch/mips/include/asm/dma-mapping.h       |  3 ---
 arch/mips/mm/dma-default.c                |  7 +++----
 arch/mn10300/include/asm/dma-mapping.h    |  6 ------
 arch/nios2/include/asm/dma-mapping.h      |  9 ---------
 arch/parisc/include/asm/dma-mapping.h     |  8 --------
 arch/parisc/kernel/pci-dma.c              |  8 ++++++++
 arch/powerpc/include/asm/dma-mapping.h    |  5 -----
 arch/s390/include/asm/dma-mapping.h       |  5 -----
 arch/sh/include/asm/dma-mapping.h         |  6 ------
 arch/sparc/include/asm/dma-mapping.h      |  8 --------
 arch/tile/include/asm/dma-mapping.h       |  9 ---------
 arch/unicore32/include/asm/dma-mapping.h  |  5 -----
 arch/x86/include/asm/dma-mapping.h        |  6 ------
 arch/xtensa/include/asm/dma-mapping.h     |  5 -----
 include/linux/dma-mapping.h               | 13 +++++++++++++
 24 files changed, 24 insertions(+), 128 deletions(-)

diff --git a/arch/alpha/include/asm/dma-mapping.h b/arch/alpha/include/asm/dma-mapping.h
index 5d53666935e6..399a4f49355e 100644
--- a/arch/alpha/include/asm/dma-mapping.h
+++ b/arch/alpha/include/asm/dma-mapping.h
@@ -8,6 +8,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return dma_ops;
 }
 
-#define dma_cache_sync(dev, va, size, dir)		  ((void)0)
-
 #endif	/* _ALPHA_DMA_MAPPING_H */
diff --git a/arch/cris/include/asm/dma-mapping.h b/arch/cris/include/asm/dma-mapping.h
index 256169de3743..e30adde42beb 100644
--- a/arch/cris/include/asm/dma-mapping.h
+++ b/arch/cris/include/asm/dma-mapping.h
@@ -16,10 +16,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 }
 #endif
 
-static inline void
-dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-	       enum dma_data_direction direction)
-{
-}
-
 #endif
diff --git a/arch/frv/include/asm/dma-mapping.h b/arch/frv/include/asm/dma-mapping.h
index da0e5c9744c4..da24ae943f02 100644
--- a/arch/frv/include/asm/dma-mapping.h
+++ b/arch/frv/include/asm/dma-mapping.h
@@ -14,10 +14,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &frv_dma_ops;
 }
 
-static inline
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction direction)
-{
-}
-
 #endif  /* _ASM_DMA_MAPPING_H */
diff --git a/arch/hexagon/include/asm/dma-mapping.h b/arch/hexagon/include/asm/dma-mapping.h
index 463dbc18f853..5208de242e79 100644
--- a/arch/hexagon/include/asm/dma-mapping.h
+++ b/arch/hexagon/include/asm/dma-mapping.h
@@ -37,9 +37,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return dma_ops;
 }
 
-extern void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-			   enum dma_data_direction direction);
-
 static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
 {
 	if (!dev->dma_mask)
diff --git a/arch/ia64/include/asm/dma-mapping.h b/arch/ia64/include/asm/dma-mapping.h
index 99dfc1aa9d3c..9e5b5df76ff8 100644
--- a/arch/ia64/include/asm/dma-mapping.h
+++ b/arch/ia64/include/asm/dma-mapping.h
@@ -44,10 +44,4 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
 	return daddr;
 }
 
-static inline void
-dma_cache_sync (struct device *dev, void *vaddr, size_t size,
-	enum dma_data_direction dir)
-{
-}
-
 #endif /* _ASM_IA64_DMA_MAPPING_H */
diff --git a/arch/m32r/include/asm/dma-mapping.h b/arch/m32r/include/asm/dma-mapping.h
index aff3ae8b62f7..9e993daed7a0 100644
--- a/arch/m32r/include/asm/dma-mapping.h
+++ b/arch/m32r/include/asm/dma-mapping.h
@@ -13,11 +13,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &dma_noop_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction direction)
-{
-}
-
 static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
 {
 	if (!dev->dma_mask)
diff --git a/arch/m68k/include/asm/dma-mapping.h b/arch/m68k/include/asm/dma-mapping.h
index 9210e470771b..9a0d559fcc13 100644
--- a/arch/m68k/include/asm/dma-mapping.h
+++ b/arch/m68k/include/asm/dma-mapping.h
@@ -8,10 +8,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
         return &m68k_dma_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction dir)
-{
-	/* we use coherent allocation, so not much to do here. */
-}
-
 #endif  /* _M68K_DMA_MAPPING_H */
diff --git a/arch/metag/include/asm/dma-mapping.h b/arch/metag/include/asm/dma-mapping.h
index ea573be2b6d0..340265dcf839 100644
--- a/arch/metag/include/asm/dma-mapping.h
+++ b/arch/metag/include/asm/dma-mapping.h
@@ -8,14 +8,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &metag_dma_ops;
 }
 
-/*
- * dma_alloc_attrs() always returns non-cacheable memory, so there's no need to
- * do any flushing here.
- */
-static inline void
-dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-	       enum dma_data_direction direction)
-{
-}
-
 #endif
diff --git a/arch/microblaze/include/asm/dma-mapping.h b/arch/microblaze/include/asm/dma-mapping.h
index ad448e4aedb6..6b9ea39405b8 100644
--- a/arch/microblaze/include/asm/dma-mapping.h
+++ b/arch/microblaze/include/asm/dma-mapping.h
@@ -25,9 +25,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &dma_direct_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		enum dma_data_direction direction)
-{
-}
-
 #endif	/* _ASM_MICROBLAZE_DMA_MAPPING_H */
diff --git a/arch/mips/include/asm/dma-mapping.h b/arch/mips/include/asm/dma-mapping.h
index aba71385f9d1..6ea1439430a2 100644
--- a/arch/mips/include/asm/dma-mapping.h
+++ b/arch/mips/include/asm/dma-mapping.h
@@ -26,9 +26,6 @@ static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
 
 static inline void dma_mark_clean(void *addr, size_t size) {}
 
-extern void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-	       enum dma_data_direction direction);
-
 #define arch_setup_dma_ops arch_setup_dma_ops
 static inline void arch_setup_dma_ops(struct device *dev, u64 dma_base,
 				      u64 size, const struct iommu_ops *iommu,
diff --git a/arch/mips/mm/dma-default.c b/arch/mips/mm/dma-default.c
index c01bd20d0208..2e2514e00720 100644
--- a/arch/mips/mm/dma-default.c
+++ b/arch/mips/mm/dma-default.c
@@ -383,7 +383,7 @@ static int mips_dma_supported(struct device *dev, u64 mask)
 	return plat_dma_supported(dev, mask);
 }
 
-void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+static void mips_dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 			 enum dma_data_direction direction)
 {
 	BUG_ON(direction == DMA_NONE);
@@ -392,8 +392,6 @@ void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 		__dma_sync_virtual(vaddr, size, direction);
 }
 
-EXPORT_SYMBOL(dma_cache_sync);
-
 static const struct dma_map_ops mips_default_dma_map_ops = {
 	.alloc = mips_dma_alloc_coherent,
 	.free = mips_dma_free_coherent,
@@ -407,7 +405,8 @@ static const struct dma_map_ops mips_default_dma_map_ops = {
 	.sync_sg_for_cpu = mips_dma_sync_sg_for_cpu,
 	.sync_sg_for_device = mips_dma_sync_sg_for_device,
 	.mapping_error = mips_dma_mapping_error,
-	.dma_supported = mips_dma_supported
+	.dma_supported = mips_dma_supported,
+	.cache_sync = mips_dma_cache_sync,
 };
 
 const struct dma_map_ops *mips_dma_map_ops = &mips_default_dma_map_ops;
diff --git a/arch/mn10300/include/asm/dma-mapping.h b/arch/mn10300/include/asm/dma-mapping.h
index dc24163b190f..439e474ed6d7 100644
--- a/arch/mn10300/include/asm/dma-mapping.h
+++ b/arch/mn10300/include/asm/dma-mapping.h
@@ -18,10 +18,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &mn10300_dma_ops;
 }
 
-static inline
-void dma_cache_sync(void *vaddr, size_t size,
-		    enum dma_data_direction direction)
-{
-}
-
 #endif
diff --git a/arch/nios2/include/asm/dma-mapping.h b/arch/nios2/include/asm/dma-mapping.h
index f8dc62222741..6ceb92251da0 100644
--- a/arch/nios2/include/asm/dma-mapping.h
+++ b/arch/nios2/include/asm/dma-mapping.h
@@ -17,13 +17,4 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &nios2_dma_ops;
 }
 
-/*
- * dma_alloc_attrs() always returns non-cacheable memory, so there's no need to
- * do any flushing here.
- */
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction direction)
-{
-}
-
 #endif /* _ASM_NIOS2_DMA_MAPPING_H */
diff --git a/arch/parisc/include/asm/dma-mapping.h b/arch/parisc/include/asm/dma-mapping.h
index 2b16282add69..cb26bbd71d8a 100644
--- a/arch/parisc/include/asm/dma-mapping.h
+++ b/arch/parisc/include/asm/dma-mapping.h
@@ -32,14 +32,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return hppa_dma_ops;
 }
 
-static inline void
-dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-	       enum dma_data_direction direction)
-{
-	if (hppa_dma_ops->sync_single_for_cpu)
-		flush_kernel_dcache_range((unsigned long)vaddr, size);
-}
-
 static inline void *
 parisc_walk_tree(struct device *dev)
 {
diff --git a/arch/parisc/kernel/pci-dma.c b/arch/parisc/kernel/pci-dma.c
index bd4c0a7471d3..f87c34cb3b43 100644
--- a/arch/parisc/kernel/pci-dma.c
+++ b/arch/parisc/kernel/pci-dma.c
@@ -571,6 +571,12 @@ static void pa11_dma_sync_sg_for_device(struct device *dev, struct scatterlist *
 		flush_kernel_vmap_range(sg_virt(sg), sg->length);
 }
 
+static void pa11_dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+	       enum dma_data_direction direction)
+{
+	flush_kernel_dcache_range((unsigned long)vaddr, size);
+}
+
 const struct dma_map_ops pcxl_dma_ops = {
 	.dma_supported =	pa11_dma_supported,
 	.alloc =		pa11_dma_alloc,
@@ -583,6 +589,7 @@ const struct dma_map_ops pcxl_dma_ops = {
 	.sync_single_for_device = pa11_dma_sync_single_for_device,
 	.sync_sg_for_cpu =	pa11_dma_sync_sg_for_cpu,
 	.sync_sg_for_device =	pa11_dma_sync_sg_for_device,
+	.cache_sync =		pa11_dma_cache_sync,
 };
 
 static void *pcx_dma_alloc(struct device *dev, size_t size,
@@ -619,4 +626,5 @@ const struct dma_map_ops pcx_dma_ops = {
 	.sync_single_for_device = pa11_dma_sync_single_for_device,
 	.sync_sg_for_cpu =	pa11_dma_sync_sg_for_cpu,
 	.sync_sg_for_device =	pa11_dma_sync_sg_for_device,
+	.cache_sync =		pa11_dma_cache_sync,
 };
diff --git a/arch/powerpc/include/asm/dma-mapping.h b/arch/powerpc/include/asm/dma-mapping.h
index 320846442bfb..2e43c2ef7632 100644
--- a/arch/powerpc/include/asm/dma-mapping.h
+++ b/arch/powerpc/include/asm/dma-mapping.h
@@ -141,10 +141,5 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
 
 #define ARCH_HAS_DMA_MMAP_COHERENT
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		enum dma_data_direction direction)
-{
-}
-
 #endif /* __KERNEL__ */
 #endif	/* _ASM_DMA_MAPPING_H */
diff --git a/arch/s390/include/asm/dma-mapping.h b/arch/s390/include/asm/dma-mapping.h
index 512ad0eaa11a..b17304b13de5 100644
--- a/arch/s390/include/asm/dma-mapping.h
+++ b/arch/s390/include/asm/dma-mapping.h
@@ -15,11 +15,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &dma_noop_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction direction)
-{
-}
-
 static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
 {
 	if (!dev->dma_mask)
diff --git a/arch/sh/include/asm/dma-mapping.h b/arch/sh/include/asm/dma-mapping.h
index b46194ecef17..e89df111c017 100644
--- a/arch/sh/include/asm/dma-mapping.h
+++ b/arch/sh/include/asm/dma-mapping.h
@@ -9,12 +9,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return dma_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction dir)
-{
-}
-
-/* arch/sh/mm/consistent.c */
 extern void *dma_generic_alloc_coherent(struct device *dev, size_t size,
 					dma_addr_t *dma_addr, gfp_t flag,
 					unsigned long attrs);
diff --git a/arch/sparc/include/asm/dma-mapping.h b/arch/sparc/include/asm/dma-mapping.h
index 60bf1633d554..b298ed45cb23 100644
--- a/arch/sparc/include/asm/dma-mapping.h
+++ b/arch/sparc/include/asm/dma-mapping.h
@@ -5,14 +5,6 @@
 #include <linux/mm.h>
 #include <linux/dma-debug.h>
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction dir)
-{
-	/* Since dma_{alloc,free}_noncoherent() allocated coherent memory, this
-	 * routine can be a nop.
-	 */
-}
-
 extern const struct dma_map_ops *dma_ops;
 extern const struct dma_map_ops pci32_dma_ops;
 
diff --git a/arch/tile/include/asm/dma-mapping.h b/arch/tile/include/asm/dma-mapping.h
index 7061dc8af43a..97ad62878290 100644
--- a/arch/tile/include/asm/dma-mapping.h
+++ b/arch/tile/include/asm/dma-mapping.h
@@ -67,13 +67,4 @@ static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
 #define HAVE_ARCH_DMA_SET_MASK 1
 int dma_set_mask(struct device *dev, u64 mask);
 
-/*
- * dma_alloc_attrs() always returns non-cacheable memory, so there's no need to
- * do any flushing here.
- */
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-				  enum dma_data_direction direction)
-{
-}
-
 #endif /* _ASM_TILE_DMA_MAPPING_H */
diff --git a/arch/unicore32/include/asm/dma-mapping.h b/arch/unicore32/include/asm/dma-mapping.h
index e949855bb794..ac608c2f6af6 100644
--- a/arch/unicore32/include/asm/dma-mapping.h
+++ b/arch/unicore32/include/asm/dma-mapping.h
@@ -45,10 +45,5 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
 
 static inline void dma_mark_clean(void *addr, size_t size) {}
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr,
-		size_t size, enum dma_data_direction direction)
-{
-}
-
 #endif /* __KERNEL__ */
 #endif
diff --git a/arch/x86/include/asm/dma-mapping.h b/arch/x86/include/asm/dma-mapping.h
index c6d3367be916..c26747340ba5 100644
--- a/arch/x86/include/asm/dma-mapping.h
+++ b/arch/x86/include/asm/dma-mapping.h
@@ -67,12 +67,6 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
 }
 #endif /* CONFIG_X86_DMA_REMAP */
 
-static inline void
-dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-	enum dma_data_direction dir)
-{
-}
-
 static inline unsigned long dma_alloc_coherent_mask(struct device *dev,
 						    gfp_t gfp)
 {
diff --git a/arch/xtensa/include/asm/dma-mapping.h b/arch/xtensa/include/asm/dma-mapping.h
index 353e0314d6ba..153bf2370988 100644
--- a/arch/xtensa/include/asm/dma-mapping.h
+++ b/arch/xtensa/include/asm/dma-mapping.h
@@ -23,11 +23,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &xtensa_dma_map_ops;
 }
 
-static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
-		    enum dma_data_direction direction)
-{
-}
-
 static inline dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr)
 {
 	return (dma_addr_t)paddr;
diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
index 29ce9815da87..028a375d240d 100644
--- a/include/linux/dma-mapping.h
+++ b/include/linux/dma-mapping.h
@@ -126,6 +126,8 @@ struct dma_map_ops {
 	void (*sync_sg_for_device)(struct device *dev,
 				   struct scatterlist *sg, int nents,
 				   enum dma_data_direction dir);
+	void (*cache_sync)(struct device *dev, void *vaddr, size_t size,
+			enum dma_data_direction direction);
 	int (*mapping_error)(struct device *dev, dma_addr_t dma_addr);
 	int (*dma_supported)(struct device *dev, u64 mask);
 #ifdef ARCH_HAS_DMA_GET_REQUIRED_MASK
@@ -436,6 +438,17 @@ dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
 #define dma_map_page(d, p, o, s, r) dma_map_page_attrs(d, p, o, s, r, 0)
 #define dma_unmap_page(d, a, s, r) dma_unmap_page_attrs(d, a, s, r, 0)
 
+static inline void
+dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+		enum dma_data_direction dir)
+{
+	const struct dma_map_ops *ops = get_dma_ops(dev);
+
+	BUG_ON(!valid_dma_direction(dir));
+	if (ops->cache_sync)
+		ops->cache_sync(dev, vaddr, size, dir);
+}
+
 extern int dma_common_mmap(struct device *dev, struct vm_area_struct *vma,
 			   void *cpu_addr, dma_addr_t dma_addr, size_t size);
 
-- 
2.14.1

Re: [PATCH 07/11] powerpc: make dma_cache_sync a no-op

[Christophe LEROY]

Le 03/10/2017 à 12:43, Christoph Hellwig a écrit :
> powerpc does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
> doesn't make any sense to do any work in dma_cache_sync given that it
> must be a no-op when dma_alloc_attrs returns coherent memory.
What about arch/powerpc/mm/dma-noncoherent.c ?

Powerpc 8xx doesn't have coherent memory.

Christophe

> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>   arch/powerpc/include/asm/dma-mapping.h | 2 --
>   1 file changed, 2 deletions(-)
> 
> diff --git a/arch/powerpc/include/asm/dma-mapping.h b/arch/powerpc/include/asm/dma-mapping.h
> index eaece3d3e225..320846442bfb 100644
> --- a/arch/powerpc/include/asm/dma-mapping.h
> +++ b/arch/powerpc/include/asm/dma-mapping.h
> @@ -144,8 +144,6 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
>   static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
>   		enum dma_data_direction direction)
>   {
> -	BUG_ON(direction = DMA_NONE);
> -	__dma_sync(vaddr, size, (int)direction);
>   }
>   
>   #endif /* __KERNEL__ */
> 

Re: [PATCH 07/11] powerpc: make dma_cache_sync a no-op

[Christophe LEROY]

Le 03/10/2017 à 12:43, Christoph Hellwig a écrit :
> powerpc does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
> doesn't make any sense to do any work in dma_cache_sync given that it
> must be a no-op when dma_alloc_attrs returns coherent memory.
What about arch/powerpc/mm/dma-noncoherent.c ?

Powerpc 8xx doesn't have coherent memory.

Christophe

> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>   arch/powerpc/include/asm/dma-mapping.h | 2 --
>   1 file changed, 2 deletions(-)
> 
> diff --git a/arch/powerpc/include/asm/dma-mapping.h b/arch/powerpc/include/asm/dma-mapping.h
> index eaece3d3e225..320846442bfb 100644
> --- a/arch/powerpc/include/asm/dma-mapping.h
> +++ b/arch/powerpc/include/asm/dma-mapping.h
> @@ -144,8 +144,6 @@ static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
>   static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
>   		enum dma_data_direction direction)
>   {
> -	BUG_ON(direction == DMA_NONE);
> -	__dma_sync(vaddr, size, (int)direction);
>   }
>   
>   #endif /* __KERNEL__ */
> 

Re: [PATCH 07/11] powerpc: make dma_cache_sync a no-op

[Robin Murphy]

On 03/10/17 12:24, Christophe LEROY wrote:
> 
> 
> Le 03/10/2017 à 12:43, Christoph Hellwig a écrit :
>> powerpc does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
>> doesn't make any sense to do any work in dma_cache_sync given that it
>> must be a no-op when dma_alloc_attrs returns coherent memory.
> What about arch/powerpc/mm/dma-noncoherent.c ?

AFAICS, just like the ARM code from which it is derived, that will
always return a non-cacheable remap of the allocation, which should thus
not require explicit maintenance after CPU accesses. dma_cache_sync()
would only matter if dma_alloc_attrs(..., DMA_ATTR_NON_CONSISTENT) got
special treatment and was allowed to return a cacheable address, but PPC
doesn't even propagate the attrs to its internal __dma_alloc_coherent()
implementations.

Robin.

> Powerpc 8xx doesn't have coherent memory.
> 
> Christophe
> 
>>
>> Signed-off-by: Christoph Hellwig <hch@lst.de>
>> ---
>>   arch/powerpc/include/asm/dma-mapping.h | 2 --
>>   1 file changed, 2 deletions(-)
>>
>> diff --git a/arch/powerpc/include/asm/dma-mapping.h
>> b/arch/powerpc/include/asm/dma-mapping.h
>> index eaece3d3e225..320846442bfb 100644
>> --- a/arch/powerpc/include/asm/dma-mapping.h
>> +++ b/arch/powerpc/include/asm/dma-mapping.h
>> @@ -144,8 +144,6 @@ static inline phys_addr_t dma_to_phys(struct
>> device *dev, dma_addr_t daddr)
>>   static inline void dma_cache_sync(struct device *dev, void *vaddr,
>> size_t size,
>>           enum dma_data_direction direction)
>>   {
>> -    BUG_ON(direction = DMA_NONE);
>> -    __dma_sync(vaddr, size, (int)direction);
>>   }
>>     #endif /* __KERNEL__ */
>>

Re: [PATCH 07/11] powerpc: make dma_cache_sync a no-op

[Robin Murphy]

On 03/10/17 12:24, Christophe LEROY wrote:
> 
> 
> Le 03/10/2017 à 12:43, Christoph Hellwig a écrit :
>> powerpc does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
>> doesn't make any sense to do any work in dma_cache_sync given that it
>> must be a no-op when dma_alloc_attrs returns coherent memory.
> What about arch/powerpc/mm/dma-noncoherent.c ?

AFAICS, just like the ARM code from which it is derived, that will
always return a non-cacheable remap of the allocation, which should thus
not require explicit maintenance after CPU accesses. dma_cache_sync()
would only matter if dma_alloc_attrs(..., DMA_ATTR_NON_CONSISTENT) got
special treatment and was allowed to return a cacheable address, but PPC
doesn't even propagate the attrs to its internal __dma_alloc_coherent()
implementations.

Robin.

> Powerpc 8xx doesn't have coherent memory.
> 
> Christophe
> 
>>
>> Signed-off-by: Christoph Hellwig <hch@lst.de>
>> ---
>>   arch/powerpc/include/asm/dma-mapping.h | 2 --
>>   1 file changed, 2 deletions(-)
>>
>> diff --git a/arch/powerpc/include/asm/dma-mapping.h
>> b/arch/powerpc/include/asm/dma-mapping.h
>> index eaece3d3e225..320846442bfb 100644
>> --- a/arch/powerpc/include/asm/dma-mapping.h
>> +++ b/arch/powerpc/include/asm/dma-mapping.h
>> @@ -144,8 +144,6 @@ static inline phys_addr_t dma_to_phys(struct
>> device *dev, dma_addr_t daddr)
>>   static inline void dma_cache_sync(struct device *dev, void *vaddr,
>> size_t size,
>>           enum dma_data_direction direction)
>>   {
>> -    BUG_ON(direction == DMA_NONE);
>> -    __dma_sync(vaddr, size, (int)direction);
>>   }
>>     #endif /* __KERNEL__ */
>>

Re: [PATCH 07/11] powerpc: make dma_cache_sync a no-op

[Christoph Hellwig]

On Tue, Oct 03, 2017 at 01:24:57PM +0200, Christophe LEROY wrote:
>> powerpc does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
>> doesn't make any sense to do any work in dma_cache_sync given that it
>> must be a no-op when dma_alloc_attrs returns coherent memory.
> What about arch/powerpc/mm/dma-noncoherent.c ?
>
> Powerpc 8xx doesn't have coherent memory.

It doesn't implement the DMA_ATTR_NON_CONSISTENT interface either,
so if it really doesn't have a way to provide dma coherent allocation
(although the code in __dma_alloc_coherent suggests it does provide
dma coherent allocations) I have no idea how it could ever have
worked.

Re: [PATCH 07/11] powerpc: make dma_cache_sync a no-op

[Christoph Hellwig]

On Tue, Oct 03, 2017 at 01:24:57PM +0200, Christophe LEROY wrote:
>> powerpc does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
>> doesn't make any sense to do any work in dma_cache_sync given that it
>> must be a no-op when dma_alloc_attrs returns coherent memory.
> What about arch/powerpc/mm/dma-noncoherent.c ?
>
> Powerpc 8xx doesn't have coherent memory.

It doesn't implement the DMA_ATTR_NON_CONSISTENT interface either,
so if it really doesn't have a way to provide dma coherent allocation
(although the code in __dma_alloc_coherent suggests it does provide
dma coherent allocations) I have no idea how it could ever have
worked.

Re: refactor dma_cache_sync V2

[Robin Murphy]

On 03/10/17 11:43, Christoph Hellwig wrote:
> The dma_cache_sync routines is used to flush caches for memory returned
> by dma_alloc_attrs with the DMA_ATTR_NON_CONSISTENT flag (or previously
> from dma_alloc_noncoherent), but the requirements for it seems to be
> frequently misunderstood.  dma_cache_sync is documented to be a no-op for
> allocations that do not have the DMA_ATTR_NON_CONSISTENT flag set, and
> yet a lot of architectures implement it in some way despite not
> implementing DMA_ATTR_NON_CONSISTENT.
> 
> This series removes a few abuses of dma_cache_sync for non-DMA API
> purposes, then changes all remaining architectures that do not implement
> DMA_ATTR_NON_CONSISTENT to implement dma_cache_sync as a no-op, and
> then adds the struct dma_map_ops indirection we use for all other
> DMA mapping operations to dma_cache_sync as well, thus removing all but
> two implementations of the function.
> 
> Changes since V1:
>  - drop the mips fd_cacheflush, merged via maintainer
>  - spelling fix in last commit descriptions (thanks Geert)
> 

For the series,

Reviewed-by: Robin Murphy <robin.murphy@arm.com>

The MIPS DMA ops are a little fiddly, but I've satisfied myself that
CONFIG_DMA_NONCOHERENT and CONFIG_SWIOTLB are mutually exclusive for
Loongson64 such that patch #11 isn't missing anything.

Robin.

Re: refactor dma_cache_sync V2

[Robin Murphy]

On 03/10/17 11:43, Christoph Hellwig wrote:
> The dma_cache_sync routines is used to flush caches for memory returned
> by dma_alloc_attrs with the DMA_ATTR_NON_CONSISTENT flag (or previously
> from dma_alloc_noncoherent), but the requirements for it seems to be
> frequently misunderstood.  dma_cache_sync is documented to be a no-op for
> allocations that do not have the DMA_ATTR_NON_CONSISTENT flag set, and
> yet a lot of architectures implement it in some way despite not
> implementing DMA_ATTR_NON_CONSISTENT.
> 
> This series removes a few abuses of dma_cache_sync for non-DMA API
> purposes, then changes all remaining architectures that do not implement
> DMA_ATTR_NON_CONSISTENT to implement dma_cache_sync as a no-op, and
> then adds the struct dma_map_ops indirection we use for all other
> DMA mapping operations to dma_cache_sync as well, thus removing all but
> two implementations of the function.
> 
> Changes since V1:
>  - drop the mips fd_cacheflush, merged via maintainer
>  - spelling fix in last commit descriptions (thanks Geert)
> 

For the series,

Reviewed-by: Robin Murphy <robin.murphy@arm.com>

The MIPS DMA ops are a little fiddly, but I've satisfied myself that
CONFIG_DMA_NONCOHERENT and CONFIG_SWIOTLB are mutually exclusive for
Loongson64 such that patch #11 isn't missing anything.

Robin.

RE: [PATCH 02/11] x86: make dma_cache_sync a no-op

[David Laight]

From: Christoph Hellwig
> Sent: 03 October 2017 11:43
> x86 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
> make any sense to do any work in dma_cache_sync given that it must be a
> no-op when dma_alloc_attrs returns coherent memory.

I believe it is just about possible to require an explicit
write flush on x86.
ISTR this can happen with something like write combining.

Whether this can actually happen is the kernel is another matter.

	David

RE: [PATCH 02/11] x86: make dma_cache_sync a no-op

[David Laight]

From: Christoph Hellwig
> Sent: 03 October 2017 11:43
> x86 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
> make any sense to do any work in dma_cache_sync given that it must be a
> no-op when dma_alloc_attrs returns coherent memory.

I believe it is just about possible to require an explicit
write flush on x86.
ISTR this can happen with something like write combining.

Whether this can actually happen is the kernel is another matter.

	David

RE: [PATCH 02/11] x86: make dma_cache_sync a no-op

[Thomas Gleixner]

On Tue, 3 Oct 2017, David Laight wrote:

> From: Christoph Hellwig
> > Sent: 03 October 2017 11:43
> > x86 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
> > make any sense to do any work in dma_cache_sync given that it must be a
> > no-op when dma_alloc_attrs returns coherent memory.
> 
> I believe it is just about possible to require an explicit
> write flush on x86.
> ISTR this can happen with something like write combining.

As the changelog says: x86 only implements dma_alloc_coherent() which as
the name says returns coherent memory, i.e. type WB (write back), which is
not subject to dma_cache_sync() operations.

If the driver converts that memory to WC (write combine) on its own via
PAT/MTRR, then it needs to take care of flushing the write buffer on its
own. It's not convered by this interface.

Thanks,

	tglx

RE: [PATCH 02/11] x86: make dma_cache_sync a no-op

[Thomas Gleixner]

On Tue, 3 Oct 2017, David Laight wrote:

> From: Christoph Hellwig
> > Sent: 03 October 2017 11:43
> > x86 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
> > make any sense to do any work in dma_cache_sync given that it must be a
> > no-op when dma_alloc_attrs returns coherent memory.
> 
> I believe it is just about possible to require an explicit
> write flush on x86.
> ISTR this can happen with something like write combining.

As the changelog says: x86 only implements dma_alloc_coherent() which as
the name says returns coherent memory, i.e. type WB (write back), which is
not subject to dma_cache_sync() operations.

If the driver converts that memory to WC (write combine) on its own via
PAT/MTRR, then it needs to take care of flushing the write buffer on its
own. It's not convered by this interface.

Thanks,

	tglx

RE: [PATCH 02/11] x86: make dma_cache_sync a no-op

[Thomas Gleixner]

On Tue, 3 Oct 2017, David Laight wrote:

> From: Christoph Hellwig
> > Sent: 03 October 2017 11:43
> > x86 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
> > make any sense to do any work in dma_cache_sync given that it must be a
> > no-op when dma_alloc_attrs returns coherent memory.
> 
> I believe it is just about possible to require an explicit
> write flush on x86.
> ISTR this can happen with something like write combining.

As the changelog says: x86 only implements dma_alloc_coherent() which as
the name says returns coherent memory, i.e. type WB (write back), which is
not subject to dma_cache_sync() operations.

If the driver converts that memory to WC (write combine) on its own via
PAT/MTRR, then it needs to take care of flushing the write buffer on its
own. It's not convered by this interface.

Thanks,

	tglx

Re: [PATCH 07/11] powerpc: make dma_cache_sync a no-op

[Christophe LEROY]

Le 03/10/2017 à 13:43, Christoph Hellwig a écrit :
> On Tue, Oct 03, 2017 at 01:24:57PM +0200, Christophe LEROY wrote:
>>> powerpc does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
>>> doesn't make any sense to do any work in dma_cache_sync given that it
>>> must be a no-op when dma_alloc_attrs returns coherent memory.
>> What about arch/powerpc/mm/dma-noncoherent.c ?
>>
>> Powerpc 8xx doesn't have coherent memory.
> 
> It doesn't implement the DMA_ATTR_NON_CONSISTENT interface either,
> so if it really doesn't have a way to provide dma coherent allocation
> (although the code in __dma_alloc_coherent suggests it does provide
> dma coherent allocations) I have no idea how it could ever have
> worked.
> 

Yes indeed it provides coherent memory by allocation non cached memory.

And drivers aiming at using non coherent memory do it by using kmalloc() 
with GFP_DMA then dma_map_single().
Then they use dma_sync_single_for_xxx(), which calls __dma_sync() on the 
8xx and is a nop on other powerpcs.

Christophe

Re: [PATCH 07/11] powerpc: make dma_cache_sync a no-op

[Christophe LEROY]

Le 03/10/2017 à 13:43, Christoph Hellwig a écrit :
> On Tue, Oct 03, 2017 at 01:24:57PM +0200, Christophe LEROY wrote:
>>> powerpc does not implement DMA_ATTR_NON_CONSISTENT allocations, so it
>>> doesn't make any sense to do any work in dma_cache_sync given that it
>>> must be a no-op when dma_alloc_attrs returns coherent memory.
>> What about arch/powerpc/mm/dma-noncoherent.c ?
>>
>> Powerpc 8xx doesn't have coherent memory.
> 
> It doesn't implement the DMA_ATTR_NON_CONSISTENT interface either,
> so if it really doesn't have a way to provide dma coherent allocation
> (although the code in __dma_alloc_coherent suggests it does provide
> dma coherent allocations) I have no idea how it could ever have
> worked.
> 

Yes indeed it provides coherent memory by allocation non cached memory.

And drivers aiming at using non coherent memory do it by using kmalloc() 
with GFP_DMA then dma_map_single().
Then they use dma_sync_single_for_xxx(), which calls __dma_sync() on the 
8xx and is a nop on other powerpcs.

Christophe

RE: [PATCH 04/11] ia64: make dma_cache_sync a no-op

[David Laight]

From: Christoph Hellwig
> Sent: 03 October 2017 11:43
>
> ia64 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
> make any sense to do any work in dma_cache_sync given that it must be a
> no-op when dma_alloc_attrs returns coherent memory.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  arch/ia64/include/asm/dma-mapping.h | 5 -----
>  1 file changed, 5 deletions(-)
> 
> diff --git a/arch/ia64/include/asm/dma-mapping.h b/arch/ia64/include/asm/dma-mapping.h
> index 3ce5ab4339f3..99dfc1aa9d3c 100644
> --- a/arch/ia64/include/asm/dma-mapping.h
> +++ b/arch/ia64/include/asm/dma-mapping.h
> @@ -48,11 +48,6 @@ static inline void
>  dma_cache_sync (struct device *dev, void *vaddr, size_t size,
>  	enum dma_data_direction dir)
>  {
> -	/*
> -	 * IA-64 is cache-coherent, so this is mostly a no-op.  However, we do need to
> -	 * ensure that dma_cache_sync() enforces order, hence the mb().
> -	 */
> -	mb();
>  }

Are you sure about this one?
It looks as though you are doing a mechanical change for all architectures.
Some of them are probably stranger than you realise.

Even with cache coherent memory any cpu 'store/write buffer' may not
be snooped by dma reads.

Something needs to flush the store buffer between the last cpu write
to the dma buffer and the write (probably a device register) that
tells the device it can read the memory.

My guess from the comment is that dma_cache_synch() is expected to
include that barrier - and it might not be anywhere else.

	David

RE: [PATCH 04/11] ia64: make dma_cache_sync a no-op

[David Laight]

From: Christoph Hellwig
> Sent: 03 October 2017 11:43
>
> ia64 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
> make any sense to do any work in dma_cache_sync given that it must be a
> no-op when dma_alloc_attrs returns coherent memory.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  arch/ia64/include/asm/dma-mapping.h | 5 -----
>  1 file changed, 5 deletions(-)
> 
> diff --git a/arch/ia64/include/asm/dma-mapping.h b/arch/ia64/include/asm/dma-mapping.h
> index 3ce5ab4339f3..99dfc1aa9d3c 100644
> --- a/arch/ia64/include/asm/dma-mapping.h
> +++ b/arch/ia64/include/asm/dma-mapping.h
> @@ -48,11 +48,6 @@ static inline void
>  dma_cache_sync (struct device *dev, void *vaddr, size_t size,
>  	enum dma_data_direction dir)
>  {
> -	/*
> -	 * IA-64 is cache-coherent, so this is mostly a no-op.  However, we do need to
> -	 * ensure that dma_cache_sync() enforces order, hence the mb().
> -	 */
> -	mb();
>  }

Are you sure about this one?
It looks as though you are doing a mechanical change for all architectures.
Some of them are probably stranger than you realise.

Even with cache coherent memory any cpu 'store/write buffer' may not
be snooped by dma reads.

Something needs to flush the store buffer between the last cpu write
to the dma buffer and the write (probably a device register) that
tells the device it can read the memory.

My guess from the comment is that dma_cache_synch() is expected to
include that barrier - and it might not be anywhere else.

	David

RE: [PATCH 04/11] ia64: make dma_cache_sync a no-op

[David Laight]

From: Christoph Hellwig
> Sent: 03 October 2017 11:43
>
> ia64 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn'=
t
> make any sense to do any work in dma_cache_sync given that it must be a
> no-op when dma_alloc_attrs returns coherent memory.
>=20
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  arch/ia64/include/asm/dma-mapping.h | 5 -----
>  1 file changed, 5 deletions(-)
>=20
> diff --git a/arch/ia64/include/asm/dma-mapping.h b/arch/ia64/include/asm/=
dma-mapping.h
> index 3ce5ab4339f3..99dfc1aa9d3c 100644
> --- a/arch/ia64/include/asm/dma-mapping.h
> +++ b/arch/ia64/include/asm/dma-mapping.h
> @@ -48,11 +48,6 @@ static inline void
>  dma_cache_sync (struct device *dev, void *vaddr, size_t size,
>  	enum dma_data_direction dir)
>  {
> -	/*
> -	 * IA-64 is cache-coherent, so this is mostly a no-op.  However, we do =
need to
> -	 * ensure that dma_cache_sync() enforces order, hence the mb().
> -	 */
> -	mb();
>  }

Are you sure about this one?
It looks as though you are doing a mechanical change for all architectures.
Some of them are probably stranger than you realise.

Even with cache coherent memory any cpu 'store/write buffer' may not
be snooped by dma reads.

Something needs to flush the store buffer between the last cpu write
to the dma buffer and the write (probably a device register) that
tells the device it can read the memory.

My guess from the comment is that dma_cache_synch() is expected to
include that barrier - and it might not be anywhere else.

	David

RE: [PATCH 04/11] ia64: make dma_cache_sync a no-op

[David Laight]

From: Christoph Hellwig
> Sent: 03 October 2017 11:43
>
> ia64 does not implement DMA_ATTR_NON_CONSISTENT allocations, so it doesn't
> make any sense to do any work in dma_cache_sync given that it must be a
> no-op when dma_alloc_attrs returns coherent memory.
> 
> Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
> ---
>  arch/ia64/include/asm/dma-mapping.h | 5 -----
>  1 file changed, 5 deletions(-)
> 
> diff --git a/arch/ia64/include/asm/dma-mapping.h b/arch/ia64/include/asm/dma-mapping.h
> index 3ce5ab4339f3..99dfc1aa9d3c 100644
> --- a/arch/ia64/include/asm/dma-mapping.h
> +++ b/arch/ia64/include/asm/dma-mapping.h
> @@ -48,11 +48,6 @@ static inline void
>  dma_cache_sync (struct device *dev, void *vaddr, size_t size,
>  	enum dma_data_direction dir)
>  {
> -	/*
> -	 * IA-64 is cache-coherent, so this is mostly a no-op.  However, we do need to
> -	 * ensure that dma_cache_sync() enforces order, hence the mb().
> -	 */
> -	mb();
>  }

Are you sure about this one?
It looks as though you are doing a mechanical change for all architectures.
Some of them are probably stranger than you realise.

Even with cache coherent memory any cpu 'store/write buffer' may not
be snooped by dma reads.

Something needs to flush the store buffer between the last cpu write
to the dma buffer and the write (probably a device register) that
tells the device it can read the memory.

My guess from the comment is that dma_cache_synch() is expected to
include that barrier - and it might not be anywhere else.

	David

Re: [PATCH 04/11] ia64: make dma_cache_sync a no-op

['Christoph Hellwig']

On Wed, Oct 04, 2017 at 08:59:24AM +0000, David Laight wrote:
> Are you sure about this one?

Yes.  And if you are not you haven't read either of the cover letter,
the DMA-API documentation, or the reply from Thomas to your mail
yesterday.

Re: [PATCH 04/11] ia64: make dma_cache_sync a no-op

['Christoph Hellwig']

On Wed, Oct 04, 2017 at 08:59:24AM +0000, David Laight wrote:
> Are you sure about this one?

Yes.  And if you are not you haven't read either of the cover letter,
the DMA-API documentation, or the reply from Thomas to your mail
yesterday.

Re: [PATCH 04/11] ia64: make dma_cache_sync a no-op

['Christoph Hellwig']

On Wed, Oct 04, 2017 at 08:59:24AM +0000, David Laight wrote:
> Are you sure about this one?

Yes.  And if you are not you haven't read either of the cover letter,
the DMA-API documentation, or the reply from Thomas to your mail
yesterday.

Re: refactor dma_cache_sync V2

[Christoph Hellwig]

On Tue, Oct 03, 2017 at 12:49:51PM +0100, Robin Murphy wrote:
> Reviewed-by: Robin Murphy <robin.murphy@arm.com>

Thanks Robin.  I've heard very little from the arch maintainers,
but if people remain silent I will apply the whole series to the
dma-mapping tree in the next days.

Re: refactor dma_cache_sync V2

[Christoph Hellwig]

On Tue, Oct 03, 2017 at 12:49:51PM +0100, Robin Murphy wrote:
> Reviewed-by: Robin Murphy <robin.murphy@arm.com>

Thanks Robin.  I've heard very little from the arch maintainers,
but if people remain silent I will apply the whole series to the
dma-mapping tree in the next days.

Re: refactor dma_cache_sync V2

[Christoph Hellwig]

On Mon, Oct 16, 2017 at 02:44:42PM +0200, Christoph Hellwig wrote:
> On Tue, Oct 03, 2017 at 12:49:51PM +0100, Robin Murphy wrote:
> > Reviewed-by: Robin Murphy <robin.murphy@arm.com>
> 
> Thanks Robin.  I've heard very little from the arch maintainers,
> but if people remain silent I will apply the whole series to the
> dma-mapping tree in the next days.

Applied to the dma-mapping tree for 4.15.

Re: refactor dma_cache_sync V2

[Christoph Hellwig]

On Mon, Oct 16, 2017 at 02:44:42PM +0200, Christoph Hellwig wrote:
> On Tue, Oct 03, 2017 at 12:49:51PM +0100, Robin Murphy wrote:
> > Reviewed-by: Robin Murphy <robin.murphy@arm.com>
> 
> Thanks Robin.  I've heard very little from the arch maintainers,
> but if people remain silent I will apply the whole series to the
> dma-mapping tree in the next days.

Applied to the dma-mapping tree for 4.15.

Re: refactor dma_cache_sync V2

[Christoph Hellwig]

On Mon, Oct 16, 2017 at 02:44:42PM +0200, Christoph Hellwig wrote:
> On Tue, Oct 03, 2017 at 12:49:51PM +0100, Robin Murphy wrote:
> > Reviewed-by: Robin Murphy <robin.murphy-5wv7dgnIgG8@public.gmane.org>
> 
> Thanks Robin.  I've heard very little from the arch maintainers,
> but if people remain silent I will apply the whole series to the
> dma-mapping tree in the next days.

Applied to the dma-mapping tree for 4.15.