dma-mapping: merge direct and noncoherent ops

All the cache maintainance is already stubbed out when not enabled,
but merging the two allows us to nicely handle the case where
cache maintainance is required for some devices, but not others.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Paul Burton <paul.burton@mips.com> # MIPS parts

arch/arc/Kconfig

@@ -17,7 +17,7 @@ config ARC
 	select BUILDTIME_EXTABLE_SORT
 	select CLONE_BACKWARDS
 	select COMMON_CLK
-	select DMA_NONCOHERENT_OPS
+	select DMA_DIRECT_OPS
 	select DMA_NONCOHERENT_MMAP
 	select GENERIC_ATOMIC64 if !ISA_ARCV2 || !(ARC_HAS_LL64 && ARC_HAS_LLSC)
 	select GENERIC_CLOCKEVENTS

arch/arc/mm/dma.c

@@ -167,7 +167,7 @@ void arch_sync_dma_for_cpu(struct device *dev, phys_addr_t paddr,
 }
 
 /*
- * Plug in coherent or noncoherent dma ops
+ * Plug in direct dma map ops.
  */
 void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size,
 			const struct iommu_ops *iommu, bool coherent)
@@ -175,13 +175,11 @@ void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size,
 	/*
 	 * IOC hardware snoops all DMA traffic keeping the caches consistent
 	 * with memory - eliding need for any explicit cache maintenance of
-	 * DMA buffers - so we can use dma_direct cache ops.
+	 * DMA buffers.
 	 */
-	if (is_isa_arcv2() && ioc_enable && coherent) {
-		set_dma_ops(dev, &dma_direct_ops);
-		dev_info(dev, "use dma_direct_ops cache ops\n");
-	} else {
-		set_dma_ops(dev, &dma_noncoherent_ops);
-		dev_info(dev, "use dma_noncoherent_ops cache ops\n");
-	}
+	if (is_isa_arcv2() && ioc_enable && coherent)
+		dev->dma_coherent = true;
+
+	dev_info(dev, "use %sncoherent DMA ops\n",
+		 dev->dma_coherent ? "" : "non");
 }

arch/arm/mm/dma-mapping-nommu.c

@@ -47,7 +47,8 @@ static void *arm_nommu_dma_alloc(struct device *dev, size_t size,
 	 */
 
 	if (attrs & DMA_ATTR_NON_CONSISTENT)
-		return dma_direct_alloc(dev, size, dma_handle, gfp, attrs);
+		return dma_direct_alloc_pages(dev, size, dma_handle, gfp,
+				attrs);
 
 	ret = dma_alloc_from_global_coherent(size, dma_handle);
 
@@ -70,7 +71,7 @@ static void arm_nommu_dma_free(struct device *dev, size_t size,
 			       unsigned long attrs)
 {
 	if (attrs & DMA_ATTR_NON_CONSISTENT) {
-		dma_direct_free(dev, size, cpu_addr, dma_addr, attrs);
+		dma_direct_free_pages(dev, size, cpu_addr, dma_addr, attrs);
 	} else {
 		int ret = dma_release_from_global_coherent(get_order(size),
 							   cpu_addr);

arch/c6x/Kconfig

@@ -9,7 +9,7 @@ config C6X
 	select ARCH_HAS_SYNC_DMA_FOR_CPU
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
 	select CLKDEV_LOOKUP
-	select DMA_NONCOHERENT_OPS
+	select DMA_DIRECT_OPS
 	select GENERIC_ATOMIC64
 	select GENERIC_IRQ_SHOW
 	select HAVE_ARCH_TRACEHOOK

arch/hexagon/Kconfig

@@ -30,7 +30,7 @@ config HEXAGON
 	select GENERIC_CLOCKEVENTS_BROADCAST
 	select MODULES_USE_ELF_RELA
 	select GENERIC_CPU_DEVICES
-	select DMA_NONCOHERENT_OPS
+	select DMA_DIRECT_OPS
 	---help---
 	  Qualcomm Hexagon is a processor architecture designed for high
 	  performance and low power across a wide variety of applications.

arch/m68k/Kconfig

@@ -26,7 +26,7 @@ config M68K
 	select MODULES_USE_ELF_RELA
 	select OLD_SIGSUSPEND3
 	select OLD_SIGACTION
-	select DMA_NONCOHERENT_OPS if HAS_DMA
+	select DMA_DIRECT_OPS if HAS_DMA
 	select HAVE_MEMBLOCK
 	select ARCH_DISCARD_MEMBLOCK
 	select NO_BOOTMEM

arch/microblaze/Kconfig

@@ -11,7 +11,7 @@ config MICROBLAZE
 	select TIMER_OF
 	select CLONE_BACKWARDS3
 	select COMMON_CLK
-	select DMA_NONCOHERENT_OPS
+	select DMA_DIRECT_OPS
 	select DMA_NONCOHERENT_MMAP
 	select GENERIC_ATOMIC64
 	select GENERIC_CLOCKEVENTS

arch/mips/Kconfig

@@ -1121,7 +1121,6 @@ config DMA_NONCOHERENT
 	select NEED_DMA_MAP_STATE
 	select DMA_NONCOHERENT_MMAP
 	select DMA_NONCOHERENT_CACHE_SYNC
-	select DMA_NONCOHERENT_OPS
 
 config SYS_HAS_EARLY_PRINTK
 	bool

arch/mips/include/asm/dma-mapping.h

@@ -12,8 +12,6 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 	return &jazz_dma_ops;
 #elif defined(CONFIG_SWIOTLB)
 	return &swiotlb_dma_ops;
-#elif defined(CONFIG_DMA_NONCOHERENT_OPS)
-	return &dma_noncoherent_ops;
 #else
 	return &dma_direct_ops;
 #endif

arch/mips/jazz/jazzdma.c

@@ -564,13 +564,13 @@ static void *jazz_dma_alloc(struct device *dev, size_t size,
 {
 	void *ret;
 
-	ret = dma_direct_alloc(dev, size, dma_handle, gfp, attrs);
+	ret = dma_direct_alloc_pages(dev, size, dma_handle, gfp, attrs);
 	if (!ret)
 		return NULL;
 
 	*dma_handle = vdma_alloc(virt_to_phys(ret), size);
 	if (*dma_handle == VDMA_ERROR) {
-		dma_direct_free(dev, size, ret, *dma_handle, attrs);
+		dma_direct_free_pages(dev, size, ret, *dma_handle, attrs);
 		return NULL;
 	}
 
@@ -587,7 +587,7 @@ static void jazz_dma_free(struct device *dev, size_t size, void *vaddr,
 	vdma_free(dma_handle);
 	if (!(attrs & DMA_ATTR_NON_CONSISTENT))
 		vaddr = (void *)CAC_ADDR((unsigned long)vaddr);
-	return dma_direct_free(dev, size, vaddr, dma_handle, attrs);
+	dma_direct_free_pages(dev, size, vaddr, dma_handle, attrs);
 }
 
 static dma_addr_t jazz_dma_map_page(struct device *dev, struct page *page,

arch/mips/mm/dma-noncoherent.c

@@ -29,9 +29,6 @@
  */
 static inline bool cpu_needs_post_dma_flush(struct device *dev)
 {
-	if (dev_is_dma_coherent(dev))
-		return false;
-
 	switch (boot_cpu_type()) {
 	case CPU_R10000:
 	case CPU_R12000:
@@ -52,11 +49,8 @@ void *arch_dma_alloc(struct device *dev, size_t size,
 {
 	void *ret;
 
-	ret = dma_direct_alloc(dev, size, dma_handle, gfp, attrs);
-	if (!ret)
-		return NULL;
-
-	if (!dev_is_dma_coherent(dev) && !(attrs & DMA_ATTR_NON_CONSISTENT)) {
+	ret = dma_direct_alloc_pages(dev, size, dma_handle, gfp, attrs);
+	if (!ret && !(attrs & DMA_ATTR_NON_CONSISTENT)) {
 		dma_cache_wback_inv((unsigned long) ret, size);
 		ret = (void *)UNCAC_ADDR(ret);
 	}
@@ -67,9 +61,9 @@ void *arch_dma_alloc(struct device *dev, size_t size,
 void arch_dma_free(struct device *dev, size_t size, void *cpu_addr,
 		dma_addr_t dma_addr, unsigned long attrs)
 {
-	if (!(attrs & DMA_ATTR_NON_CONSISTENT) && !dev_is_dma_coherent(dev))
+	if (!(attrs & DMA_ATTR_NON_CONSISTENT))
 		cpu_addr = (void *)CAC_ADDR((unsigned long)cpu_addr);
-	dma_direct_free(dev, size, cpu_addr, dma_addr, attrs);
+	dma_direct_free_pages(dev, size, cpu_addr, dma_addr, attrs);
 }
 
 int arch_dma_mmap(struct device *dev, struct vm_area_struct *vma,
@@ -78,16 +72,11 @@ int arch_dma_mmap(struct device *dev, struct vm_area_struct *vma,
 {
 	unsigned long user_count = vma_pages(vma);
 	unsigned long count = PAGE_ALIGN(size) >> PAGE_SHIFT;
-	unsigned long addr = (unsigned long)cpu_addr;
+	unsigned long addr = CAC_ADDR((unsigned long)cpu_addr);
 	unsigned long off = vma->vm_pgoff;
-	unsigned long pfn;
+	unsigned long pfn = page_to_pfn(virt_to_page((void *)addr));
 	int ret = -ENXIO;
 
-	if (!dev_is_dma_coherent(dev))
-		addr = CAC_ADDR(addr);
-
-	pfn = page_to_pfn(virt_to_page((void *)addr));
-
 	if (attrs & DMA_ATTR_WRITE_COMBINE)
 		vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
 	else
@@ -167,8 +156,7 @@ static inline void dma_sync_phys(phys_addr_t paddr, size_t size,
 void arch_sync_dma_for_device(struct device *dev, phys_addr_t paddr,
 		size_t size, enum dma_data_direction dir)
 {
-	if (!dev_is_dma_coherent(dev))
-		dma_sync_phys(paddr, size, dir);
+	dma_sync_phys(paddr, size, dir);
 }
 
 void arch_sync_dma_for_cpu(struct device *dev, phys_addr_t paddr,
@@ -183,6 +171,5 @@ void arch_dma_cache_sync(struct device *dev, void *vaddr, size_t size,
 {
 	BUG_ON(direction == DMA_NONE);
 
-	if (!dev_is_dma_coherent(dev))
-		dma_sync_virt(vaddr, size, direction);
+	dma_sync_virt(vaddr, size, direction);
 }

arch/nds32/Kconfig

@@ -11,7 +11,7 @@ config NDS32
 	select CLKSRC_MMIO
 	select CLONE_BACKWARDS
 	select COMMON_CLK
-	select DMA_NONCOHERENT_OPS
+	select DMA_DIRECT_OPS
 	select GENERIC_ATOMIC64
 	select GENERIC_CPU_DEVICES
 	select GENERIC_CLOCKEVENTS

arch/nios2/Kconfig

@@ -4,7 +4,7 @@ config NIOS2
 	select ARCH_HAS_SYNC_DMA_FOR_CPU
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
 	select ARCH_NO_SWAP
-	select DMA_NONCOHERENT_OPS
+	select DMA_DIRECT_OPS
 	select TIMER_OF
 	select GENERIC_ATOMIC64
 	select GENERIC_CLOCKEVENTS

arch/openrisc/Kconfig

@@ -7,7 +7,7 @@
 config OPENRISC
 	def_bool y
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
-	select DMA_NONCOHERENT_OPS
+	select DMA_DIRECT_OPS
 	select OF
 	select OF_EARLY_FLATTREE
 	select IRQ_DOMAIN

arch/parisc/Kconfig

@@ -186,7 +186,7 @@ config PA11
 	depends on PA7000 || PA7100LC || PA7200 || PA7300LC
 	select ARCH_HAS_SYNC_DMA_FOR_CPU
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
-	select DMA_NONCOHERENT_OPS
+	select DMA_DIRECT_OPS
 	select DMA_NONCOHERENT_CACHE_SYNC
 
 config PREFETCH

arch/parisc/kernel/setup.c

@@ -102,7 +102,7 @@ void __init dma_ops_init(void)
 	case pcxl: /* falls through */
 	case pcxs:
 	case pcxt:
-		hppa_dma_ops = &dma_noncoherent_ops;
+		hppa_dma_ops = &dma_direct_ops;
 		break;
 	default:
 		break;

arch/sh/Kconfig

@@ -7,6 +7,7 @@ config SUPERH
 	select ARCH_NO_COHERENT_DMA_MMAP if !MMU
 	select HAVE_PATA_PLATFORM
 	select CLKDEV_LOOKUP
+	select DMA_DIRECT_OPS
 	select HAVE_IDE if HAS_IOPORT_MAP
 	select HAVE_MEMBLOCK
 	select HAVE_MEMBLOCK_NODE_MAP
@@ -158,13 +159,11 @@ config SWAP_IO_SPACE
 	bool
 
 config DMA_COHERENT
-	select DMA_DIRECT_OPS
 	bool
 
 config DMA_NONCOHERENT
 	def_bool !DMA_COHERENT
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
-	select DMA_NONCOHERENT_OPS
 
 config PGTABLE_LEVELS
 	default 3 if X2TLB

arch/sparc/Kconfig

@@ -51,7 +51,7 @@ config SPARC
 config SPARC32
 	def_bool !64BIT
 	select ARCH_HAS_SYNC_DMA_FOR_CPU
-	select DMA_NONCOHERENT_OPS
+	select DMA_DIRECT_OPS
 	select GENERIC_ATOMIC64
 	select CLZ_TAB
 	select HAVE_UID16

arch/sparc/include/asm/dma-mapping.h

@@ -14,11 +14,11 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 {
 #ifdef CONFIG_SPARC_LEON
 	if (sparc_cpu_model == sparc_leon)
-		return &dma_noncoherent_ops;
+		return &dma_direct_ops;
 #endif
 #if defined(CONFIG_SPARC32) && defined(CONFIG_PCI)
 	if (bus == &pci_bus_type)
-		return &dma_noncoherent_ops;
+		return &dma_direct_ops;
 #endif
 	return dma_ops;
 }

arch/x86/kernel/amd_gart_64.c

@@ -482,7 +482,7 @@ gart_alloc_coherent(struct device *dev, size_t size, dma_addr_t *dma_addr,
 {
 	void *vaddr;
 
-	vaddr = dma_direct_alloc(dev, size, dma_addr, flag, attrs);
+	vaddr = dma_direct_alloc_pages(dev, size, dma_addr, flag, attrs);
 	if (!vaddr ||
 	    !force_iommu || dev->coherent_dma_mask <= DMA_BIT_MASK(24))
 		return vaddr;
@@ -494,7 +494,7 @@ gart_alloc_coherent(struct device *dev, size_t size, dma_addr_t *dma_addr,
 		goto out_free;
 	return vaddr;
 out_free:
-	dma_direct_free(dev, size, vaddr, *dma_addr, attrs);
+	dma_direct_free_pages(dev, size, vaddr, *dma_addr, attrs);
 	return NULL;
 }
 
@@ -504,7 +504,7 @@ gart_free_coherent(struct device *dev, size_t size, void *vaddr,
 		   dma_addr_t dma_addr, unsigned long attrs)
 {
 	gart_unmap_page(dev, dma_addr, size, DMA_BIDIRECTIONAL, 0);
-	dma_direct_free(dev, size, vaddr, dma_addr, attrs);
+	dma_direct_free_pages(dev, size, vaddr, dma_addr, attrs);
 }
 
 static int gart_mapping_error(struct device *dev, dma_addr_t dma_addr)