Merge tag 'dma-mapping-5.4' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping updates from Christoph Hellwig:

 - add dma-mapping and block layer helpers to take care of IOMMU merging
   for mmc plus subsequent fixups (Yoshihiro Shimoda)

 - rework handling of the pgprot bits for remapping (me)

 - take care of the dma direct infrastructure for swiotlb-xen (me)

 - improve the dma noncoherent remapping infrastructure (me)

 - better defaults for ->mmap, ->get_sgtable and ->get_required_mask
   (me)

 - cleanup mmaping of coherent DMA allocations (me)

 - various misc cleanups (Andy Shevchenko, me)

* tag 'dma-mapping-5.4' of git://git.infradead.org/users/hch/dma-mapping: (41 commits)
  mmc: renesas_sdhi_internal_dmac: Add MMC_CAP2_MERGE_CAPABLE
  mmc: queue: Fix bigger segments usage
  arm64: use asm-generic/dma-mapping.h
  swiotlb-xen: merge xen_unmap_single into xen_swiotlb_unmap_page
  swiotlb-xen: simplify cache maintainance
  swiotlb-xen: use the same foreign page check everywhere
  swiotlb-xen: remove xen_swiotlb_dma_mmap and xen_swiotlb_dma_get_sgtable
  xen: remove the exports for xen_{create,destroy}_contiguous_region
  xen/arm: remove xen_dma_ops
  xen/arm: simplify dma_cache_maint
  xen/arm: use dev_is_dma_coherent
  xen/arm: consolidate page-coherent.h
  xen/arm: use dma-noncoherent.h calls for xen-swiotlb cache maintainance
  arm: remove wrappers for the generic dma remap helpers
  dma-mapping: introduce a dma_common_find_pages helper
  dma-mapping: always use VM_DMA_COHERENT for generic DMA remap
  vmalloc: lift the arm flag for coherent mappings to common code
  dma-mapping: provide a better default ->get_required_mask
  dma-mapping: remove the dma_declare_coherent_memory export
  remoteproc: don't allow modular build
  ...

Documentation/DMA-API.txt

@@ -204,6 +204,14 @@ Returns the maximum size of a mapping for the device. The size parameter
 of the mapping functions like dma_map_single(), dma_map_page() and
 others should not be larger than the returned value.
 
+::
+
+	unsigned long
+	dma_get_merge_boundary(struct device *dev);
+
+Returns the DMA merge boundary. If the device cannot merge any the DMA address
+segments, the function returns 0.
+
 Part Id - Streaming DMA mappings
 --------------------------------
 
@@ -595,17 +603,6 @@ For reasons of efficiency, most platforms choose to track the declared
 region only at the granularity of a page.  For smaller allocations,
 you should use the dma_pool() API.
 
-::
-
-	void
-	dma_release_declared_memory(struct device *dev)
-
-Remove the memory region previously declared from the system.  This
-API performs *no* in-use checking for this region and will return
-unconditionally having removed all the required structures.  It is the
-driver's job to ensure that no parts of this memory region are
-currently in use.
-
 Part III - Debug drivers use of the DMA-API
 -------------------------------------------
 

Documentation/x86/x86_64/boot-options.rst

@@ -230,7 +230,7 @@ IOMMU (input/output memory management unit)
 ===========================================
 Multiple x86-64 PCI-DMA mapping implementations exist, for example:
 
-   1. <lib/dma-direct.c>: use no hardware/software IOMMU at all
+   1. <kernel/dma/direct.c>: use no hardware/software IOMMU at all
       (e.g. because you have < 3 GB memory).
       Kernel boot message: "PCI-DMA: Disabling IOMMU"
 

arch/Kconfig

@@ -793,9 +793,6 @@ config COMPAT_32BIT_TIME
 	  This is relevant on all 32-bit architectures, and 64-bit architectures
 	  as part of compat syscall handling.
 
-config ARCH_NO_COHERENT_DMA_MMAP
-	bool
-
 config ARCH_NO_PREEMPT
 	bool
 

arch/alpha/kernel/pci_iommu.c

@@ -955,5 +955,7 @@ const struct dma_map_ops alpha_pci_ops = {
 	.map_sg			= alpha_pci_map_sg,
 	.unmap_sg		= alpha_pci_unmap_sg,
 	.dma_supported		= alpha_pci_supported,
+	.mmap			= dma_common_mmap,
+	.get_sgtable		= dma_common_get_sgtable,
 };
 EXPORT_SYMBOL(alpha_pci_ops);

arch/arc/mm/dma.c

@@ -104,9 +104,3 @@ void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size,
 	dev_info(dev, "use %scoherent DMA ops\n",
 		 dev->dma_coherent ? "" : "non");
 }
-
-static int __init atomic_pool_init(void)
-{
-	return dma_atomic_pool_init(GFP_KERNEL, pgprot_noncached(PAGE_KERNEL));
-}
-postcore_initcall(atomic_pool_init);

arch/arm/Kconfig

@@ -8,7 +8,7 @@ config ARM
 	select ARCH_HAS_DEBUG_VIRTUAL if MMU
 	select ARCH_HAS_DEVMEM_IS_ALLOWED
 	select ARCH_HAS_DMA_COHERENT_TO_PFN if SWIOTLB
-	select ARCH_HAS_DMA_MMAP_PGPROT if SWIOTLB
+	select ARCH_HAS_DMA_WRITE_COMBINE if !ARM_DMA_MEM_BUFFERABLE
 	select ARCH_HAS_ELF_RANDOMIZE
 	select ARCH_HAS_FORTIFY_SOURCE
 	select ARCH_HAS_KEEPINITRD

arch/arm/include/asm/device.h

@@ -14,9 +14,6 @@ struct dev_archdata {
 #endif
 #ifdef CONFIG_ARM_DMA_USE_IOMMU
 	struct dma_iommu_mapping	*mapping;
-#endif
-#ifdef CONFIG_XEN
-	const struct dma_map_ops *dev_dma_ops;
 #endif
 	unsigned int dma_coherent:1;
 	unsigned int dma_ops_setup:1;

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

@@ -91,12 +91,6 @@ static inline dma_addr_t virt_to_dma(struct device *dev, void *addr)
 }
 #endif
 
-/* do not use this function in a driver */
-static inline bool is_device_dma_coherent(struct device *dev)
-{
-	return dev->archdata.dma_coherent;
-}
-
 /**
  * arm_dma_alloc - allocate consistent memory for DMA
  * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices

arch/arm/include/asm/pgtable-nommu.h

@@ -62,7 +62,6 @@ typedef pte_t *pte_addr_t;
  */
 #define pgprot_noncached(prot)	(prot)
 #define pgprot_writecombine(prot) (prot)
-#define pgprot_dmacoherent(prot) (prot)
 #define pgprot_device(prot)	(prot)
 
 

arch/arm/include/asm/xen/page-coherent.h

@@ -1,95 +1,2 @@
 /* SPDX-License-Identifier: GPL-2.0 */
-#ifndef _ASM_ARM_XEN_PAGE_COHERENT_H
-#define _ASM_ARM_XEN_PAGE_COHERENT_H
-
-#include <linux/dma-mapping.h>
-#include <asm/page.h>
 #include <xen/arm/page-coherent.h>
-
-static inline const struct dma_map_ops *xen_get_dma_ops(struct device *dev)
-{
-	if (dev && dev->archdata.dev_dma_ops)
-		return dev->archdata.dev_dma_ops;
-	return get_arch_dma_ops(NULL);
-}
-
-static inline void *xen_alloc_coherent_pages(struct device *hwdev, size_t size,
-		dma_addr_t *dma_handle, gfp_t flags, unsigned long attrs)
-{
-	return xen_get_dma_ops(hwdev)->alloc(hwdev, size, dma_handle, flags, attrs);
-}
-
-static inline void xen_free_coherent_pages(struct device *hwdev, size_t size,
-		void *cpu_addr, dma_addr_t dma_handle, unsigned long attrs)
-{
-	xen_get_dma_ops(hwdev)->free(hwdev, size, cpu_addr, dma_handle, attrs);
-}
-
-static inline void xen_dma_map_page(struct device *hwdev, struct page *page,
-	     dma_addr_t dev_addr, unsigned long offset, size_t size,
-	     enum dma_data_direction dir, unsigned long attrs)
-{
-	unsigned long page_pfn = page_to_xen_pfn(page);
-	unsigned long dev_pfn = XEN_PFN_DOWN(dev_addr);
-	unsigned long compound_pages =
-		(1<<compound_order(page)) * XEN_PFN_PER_PAGE;
-	bool local = (page_pfn <= dev_pfn) &&
-		(dev_pfn - page_pfn < compound_pages);
-
-	/*
-	 * Dom0 is mapped 1:1, while the Linux page can span across
-	 * multiple Xen pages, it's not possible for it to contain a
-	 * mix of local and foreign Xen pages. So if the first xen_pfn
-	 * == mfn the page is local otherwise it's a foreign page
-	 * grant-mapped in dom0. If the page is local we can safely
-	 * call the native dma_ops function, otherwise we call the xen
-	 * specific function.
-	 */
-	if (local)
-		xen_get_dma_ops(hwdev)->map_page(hwdev, page, offset, size, dir, attrs);
-	else
-		__xen_dma_map_page(hwdev, page, dev_addr, offset, size, dir, attrs);
-}
-
-static inline void xen_dma_unmap_page(struct device *hwdev, dma_addr_t handle,
-		size_t size, enum dma_data_direction dir, unsigned long attrs)
-{
-	unsigned long pfn = PFN_DOWN(handle);
-	/*
-	 * Dom0 is mapped 1:1, while the Linux page can be spanned accross
-	 * multiple Xen page, it's not possible to have a mix of local and
-	 * foreign Xen page. Dom0 is mapped 1:1, so calling pfn_valid on a
-	 * foreign mfn will always return false. If the page is local we can
-	 * safely call the native dma_ops function, otherwise we call the xen
-	 * specific function.
-	 */
-	if (pfn_valid(pfn)) {
-		if (xen_get_dma_ops(hwdev)->unmap_page)
-			xen_get_dma_ops(hwdev)->unmap_page(hwdev, handle, size, dir, attrs);
-	} else
-		__xen_dma_unmap_page(hwdev, handle, size, dir, attrs);
-}
-
-static inline void xen_dma_sync_single_for_cpu(struct device *hwdev,
-		dma_addr_t handle, size_t size, enum dma_data_direction dir)
-{
-	unsigned long pfn = PFN_DOWN(handle);
-	if (pfn_valid(pfn)) {
-		if (xen_get_dma_ops(hwdev)->sync_single_for_cpu)
-			xen_get_dma_ops(hwdev)->sync_single_for_cpu(hwdev, handle, size, dir);
-	} else
-		__xen_dma_sync_single_for_cpu(hwdev, handle, size, dir);
-}
-
-static inline void xen_dma_sync_single_for_device(struct device *hwdev,
-		dma_addr_t handle, size_t size, enum dma_data_direction dir)
-{
-	unsigned long pfn = PFN_DOWN(handle);
-	if (pfn_valid(pfn)) {
-		if (xen_get_dma_ops(hwdev)->sync_single_for_device)
-			xen_get_dma_ops(hwdev)->sync_single_for_device(hwdev, handle, size, dir);
-	} else
-		__xen_dma_sync_single_for_device(hwdev, handle, size, dir);
-}
-
-#endif /* _ASM_ARM_XEN_PAGE_COHERENT_H */

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

@@ -68,8 +68,9 @@ static int arm_nommu_dma_mmap(struct device *dev, struct vm_area_struct *vma,
 
 	if (dma_mmap_from_global_coherent(vma, cpu_addr, size, &ret))
 		return ret;
-
-	return dma_common_mmap(dev, vma, cpu_addr, dma_addr, size, attrs);
+	if (dma_mmap_from_dev_coherent(dev, vma, cpu_addr, size, &ret))
+		return ret;
+	return -ENXIO;
 }
 
 

arch/arm/mm/dma-mapping.c

@@ -14,6 +14,7 @@
 #include <linux/list.h>
 #include <linux/init.h>
 #include <linux/device.h>
+#include <linux/dma-direct.h>
 #include <linux/dma-mapping.h>
 #include <linux/dma-noncoherent.h>
 #include <linux/dma-contiguous.h>
@@ -35,6 +36,7 @@
 #include <asm/mach/map.h>
 #include <asm/system_info.h>
 #include <asm/dma-contiguous.h>
+#include <xen/swiotlb-xen.h>
 
 #include "dma.h"
 #include "mm.h"
@@ -192,6 +194,7 @@ const struct dma_map_ops arm_dma_ops = {
 	.sync_sg_for_cpu	= arm_dma_sync_sg_for_cpu,
 	.sync_sg_for_device	= arm_dma_sync_sg_for_device,
 	.dma_supported		= arm_dma_supported,
+	.get_required_mask	= dma_direct_get_required_mask,
 };
 EXPORT_SYMBOL(arm_dma_ops);
 
@@ -212,6 +215,7 @@ const struct dma_map_ops arm_coherent_dma_ops = {
 	.map_sg			= arm_dma_map_sg,
 	.map_resource		= dma_direct_map_resource,
 	.dma_supported		= arm_dma_supported,
+	.get_required_mask	= dma_direct_get_required_mask,
 };
 EXPORT_SYMBOL(arm_coherent_dma_ops);
 
@@ -336,25 +340,6 @@ static void *__alloc_remap_buffer(struct device *dev, size_t size, gfp_t gfp,
 				 pgprot_t prot, struct page **ret_page,
 				 const void *caller, bool want_vaddr);
 
-static void *
-__dma_alloc_remap(struct page *page, size_t size, gfp_t gfp, pgprot_t prot,
-	const void *caller)
-{
-	/*
-	 * DMA allocation can be mapped to user space, so lets
-	 * set VM_USERMAP flags too.
-	 */
-	return dma_common_contiguous_remap(page, size,
-			VM_ARM_DMA_CONSISTENT | VM_USERMAP,
-			prot, caller);
-}
-
-static void __dma_free_remap(void *cpu_addr, size_t size)
-{
-	dma_common_free_remap(cpu_addr, size,
-			VM_ARM_DMA_CONSISTENT | VM_USERMAP);
-}
-
 #define DEFAULT_DMA_COHERENT_POOL_SIZE	SZ_256K
 static struct gen_pool *atomic_pool __ro_after_init;
 
@@ -510,7 +495,7 @@ static void *__alloc_remap_buffer(struct device *dev, size_t size, gfp_t gfp,
 	if (!want_vaddr)
 		goto out;
 
-	ptr = __dma_alloc_remap(page, size, gfp, prot, caller);
+	ptr = dma_common_contiguous_remap(page, size, prot, caller);
 	if (!ptr) {
 		__dma_free_buffer(page, size);
 		return NULL;
@@ -577,7 +562,7 @@ static void *__alloc_from_contiguous(struct device *dev, size_t size,
 		goto out;
 
 	if (PageHighMem(page)) {
-		ptr = __dma_alloc_remap(page, size, GFP_KERNEL, prot, caller);
+		ptr = dma_common_contiguous_remap(page, size, prot, caller);
 		if (!ptr) {
 			dma_release_from_contiguous(dev, page, count);
 			return NULL;
@@ -597,7 +582,7 @@ static void __free_from_contiguous(struct device *dev, struct page *page,
 {
 	if (want_vaddr) {
 		if (PageHighMem(page))
-			__dma_free_remap(cpu_addr, size);
+			dma_common_free_remap(cpu_addr, size);
 		else
 			__dma_remap(page, size, PAGE_KERNEL);
 	}
@@ -689,7 +674,7 @@ static void *remap_allocator_alloc(struct arm_dma_alloc_args *args,
 static void remap_allocator_free(struct arm_dma_free_args *args)
 {
 	if (args->want_vaddr)
-		__dma_free_remap(args->cpu_addr, args->size);
+		dma_common_free_remap(args->cpu_addr, args->size);
 
 	__dma_free_buffer(args->page, args->size);
 }
@@ -877,17 +862,6 @@ static void arm_coherent_dma_free(struct device *dev, size_t size, void *cpu_add
 	__arm_dma_free(dev, size, cpu_addr, handle, attrs, true);
 }
 
-/*
- * The whole dma_get_sgtable() idea is fundamentally unsafe - it seems
- * that the intention is to allow exporting memory allocated via the
- * coherent DMA APIs through the dma_buf API, which only accepts a
- * scattertable.  This presents a couple of problems:
- * 1. Not all memory allocated via the coherent DMA APIs is backed by
- *    a struct page
- * 2. Passing coherent DMA memory into the streaming APIs is not allowed
- *    as we will try to flush the memory through a different alias to that
- *    actually being used (and the flushes are redundant.)
- */
 int arm_dma_get_sgtable(struct device *dev, struct sg_table *sgt,
 		 void *cpu_addr, dma_addr_t handle, size_t size,
 		 unsigned long attrs)
@@ -1132,10 +1106,6 @@ static const struct dma_map_ops *arm_get_dma_map_ops(bool coherent)
 	 * 32-bit DMA.
 	 * Use the generic dma-direct / swiotlb ops code in that case, as that
 	 * handles bounce buffering for us.
-	 *
-	 * Note: this checks CONFIG_ARM_LPAE instead of CONFIG_SWIOTLB as the
-	 * latter is also selected by the Xen code, but that code for now relies
-	 * on non-NULL dev_dma_ops.  To be cleaned up later.
 	 */
 	if (IS_ENABLED(CONFIG_ARM_LPAE))
 		return NULL;
@@ -1372,17 +1342,6 @@ static int __iommu_free_buffer(struct device *dev, struct page **pages,
 	return 0;
 }
 
-/*
- * Create a CPU mapping for a specified pages
- */
-static void *
-__iommu_alloc_remap(struct page **pages, size_t size, gfp_t gfp, pgprot_t prot,
-		    const void *caller)
-{
-	return dma_common_pages_remap(pages, size,
-			VM_ARM_DMA_CONSISTENT | VM_USERMAP, prot, caller);
-}
-
 /*
  * Create a mapping in device IO address space for specified pages
  */
@@ -1455,18 +1414,13 @@ static struct page **__atomic_get_pages(void *addr)
 
 static struct page **__iommu_get_pages(void *cpu_addr, unsigned long attrs)
 {
-	struct vm_struct *area;
-
 	if (__in_atomic_pool(cpu_addr, PAGE_SIZE))
 		return __atomic_get_pages(cpu_addr);
 
 	if (attrs & DMA_ATTR_NO_KERNEL_MAPPING)
 		return cpu_addr;
 
-	area = find_vm_area(cpu_addr);
-	if (area && (area->flags & VM_ARM_DMA_CONSISTENT))
-		return area->pages;
-	return NULL;
+	return dma_common_find_pages(cpu_addr);
 }
 
 static void *__iommu_alloc_simple(struct device *dev, size_t size, gfp_t gfp,
@@ -1539,7 +1493,7 @@ static void *__arm_iommu_alloc_attrs(struct device *dev, size_t size,
 	if (attrs & DMA_ATTR_NO_KERNEL_MAPPING)
 		return pages;
 
-	addr = __iommu_alloc_remap(pages, size, gfp, prot,
+	addr = dma_common_pages_remap(pages, size, prot,
 				   __builtin_return_address(0));
 	if (!addr)
 		goto err_mapping;
@@ -1622,10 +1576,8 @@ void __arm_iommu_free_attrs(struct device *dev, size_t size, void *cpu_addr,
 		return;
 	}
 
-	if ((attrs & DMA_ATTR_NO_KERNEL_MAPPING) == 0) {
-		dma_common_free_remap(cpu_addr, size,
-			VM_ARM_DMA_CONSISTENT | VM_USERMAP);
-	}
+	if ((attrs & DMA_ATTR_NO_KERNEL_MAPPING) == 0)
+		dma_common_free_remap(cpu_addr, size);
 
 	__iommu_remove_mapping(dev, handle, size);
 	__iommu_free_buffer(dev, pages, size, attrs);
@@ -2363,10 +2315,8 @@ void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size,
 	set_dma_ops(dev, dma_ops);
 
 #ifdef CONFIG_XEN
-	if (xen_initial_domain()) {
-		dev->archdata.dev_dma_ops = dev->dma_ops;
-		dev->dma_ops = xen_dma_ops;
-	}
+	if (xen_initial_domain())
+		dev->dma_ops = &xen_swiotlb_dma_ops;
 #endif
 	dev->archdata.dma_ops_setup = true;
 }
@@ -2402,12 +2352,6 @@ long arch_dma_coherent_to_pfn(struct device *dev, void *cpu_addr,
 	return dma_to_pfn(dev, dma_addr);
 }
 
-pgprot_t arch_dma_mmap_pgprot(struct device *dev, pgprot_t prot,
-		unsigned long attrs)
-{
-	return __get_dma_pgprot(attrs, prot);
-}
-
 void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
 		gfp_t gfp, unsigned long attrs)
 {

arch/arm/mm/mm.h

@@ -70,9 +70,6 @@ extern void __flush_dcache_page(struct address_space *mapping, struct page *page
 #define VM_ARM_MTYPE(mt)		((mt) << 20)
 #define VM_ARM_MTYPE_MASK	(0x1f << 20)
 
-/* consistent regions used by dma_alloc_attrs() */
-#define VM_ARM_DMA_CONSISTENT	0x20000000
-
 
 struct static_vm {
 	struct vm_struct vm;

arch/arm/xen/mm.c

@@ -1,6 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0-only
 #include <linux/cpu.h>
-#include <linux/dma-mapping.h>
+#include <linux/dma-noncoherent.h>
 #include <linux/gfp.h>
 #include <linux/highmem.h>
 #include <linux/export.h>
@@ -35,105 +35,56 @@ unsigned long xen_get_swiotlb_free_pages(unsigned int order)
 	return __get_free_pages(flags, order);
 }
 
-enum dma_cache_op {
-       DMA_UNMAP,
-       DMA_MAP,
-};
 static bool hypercall_cflush = false;
 
-/* functions called by SWIOTLB */
-
-static void dma_cache_maint(dma_addr_t handle, unsigned long offset,
-	size_t size, enum dma_data_direction dir, enum dma_cache_op op)
+/* buffers in highmem or foreign pages cannot cross page boundaries */
+static void dma_cache_maint(dma_addr_t handle, size_t size, u32 op)
 {
 	struct gnttab_cache_flush cflush;
-	unsigned long xen_pfn;
-	size_t left = size;
 
-	xen_pfn = (handle >> XEN_PAGE_SHIFT) + offset / XEN_PAGE_SIZE;
-	offset %= XEN_PAGE_SIZE;
+	cflush.a.dev_bus_addr = handle & XEN_PAGE_MASK;
+	cflush.offset = xen_offset_in_page(handle);
+	cflush.op = op;
 
 	do {
-		size_t len = left;
-	
-		/* buffers in highmem or foreign pages cannot cross page
-		 * boundaries */
-		if (len + offset > XEN_PAGE_SIZE)
-			len = XEN_PAGE_SIZE - offset;
+		if (size + cflush.offset > XEN_PAGE_SIZE)
+			cflush.length = XEN_PAGE_SIZE - cflush.offset;
+		else
+			cflush.length = size;
 
-		cflush.op = 0;
-		cflush.a.dev_bus_addr = xen_pfn << XEN_PAGE_SHIFT;
-		cflush.offset = offset;
-		cflush.length = len;
+		HYPERVISOR_grant_table_op(GNTTABOP_cache_flush, &cflush, 1);
 
-		if (op == DMA_UNMAP && dir != DMA_TO_DEVICE)
-			cflush.op = GNTTAB_CACHE_INVAL;
-		if (op == DMA_MAP) {
-			if (dir == DMA_FROM_DEVICE)
-				cflush.op = GNTTAB_CACHE_INVAL;
-			else
-				cflush.op = GNTTAB_CACHE_CLEAN;
-		}
-		if (cflush.op)
-			HYPERVISOR_grant_table_op(GNTTABOP_cache_flush, &cflush, 1);
-
-		offset = 0;
-		xen_pfn++;
-		left -= len;
-	} while (left);
+		cflush.offset = 0;
+		cflush.a.dev_bus_addr += cflush.length;
+		size -= cflush.length;
+	} while (size);
 }
 
-static void __xen_dma_page_dev_to_cpu(struct device *hwdev, dma_addr_t handle,
-		size_t size, enum dma_data_direction dir)
+/*
+ * Dom0 is mapped 1:1, and while the Linux page can span across multiple Xen
+ * pages, it is not possible for it to contain a mix of local and foreign Xen
+ * pages.  Calling pfn_valid on a foreign mfn will always return false, so if
+ * pfn_valid returns true the pages is local and we can use the native
+ * dma-direct functions, otherwise we call the Xen specific version.
+ */
+void xen_dma_sync_for_cpu(struct device *dev, dma_addr_t handle,
+		phys_addr_t paddr, size_t size, enum dma_data_direction dir)
 {
-	dma_cache_maint(handle & PAGE_MASK, handle & ~PAGE_MASK, size, dir, DMA_UNMAP);
+	if (pfn_valid(PFN_DOWN(handle)))
+		arch_sync_dma_for_cpu(dev, paddr, size, dir);
+	else if (dir != DMA_TO_DEVICE)
+		dma_cache_maint(handle, size, GNTTAB_CACHE_INVAL);
 }
 
-static void __xen_dma_page_cpu_to_dev(struct device *hwdev, dma_addr_t handle,
-		size_t size, enum dma_data_direction dir)
+void xen_dma_sync_for_device(struct device *dev, dma_addr_t handle,
+		phys_addr_t paddr, size_t size, enum dma_data_direction dir)
 {
-	dma_cache_maint(handle & PAGE_MASK, handle & ~PAGE_MASK, size, dir, DMA_MAP);
-}
-
-void __xen_dma_map_page(struct device *hwdev, struct page *page,
-	     dma_addr_t dev_addr, unsigned long offset, size_t size,
-	     enum dma_data_direction dir, unsigned long attrs)
-{
-	if (is_device_dma_coherent(hwdev))
-		return;
-	if (attrs & DMA_ATTR_SKIP_CPU_SYNC)
-		return;
-
-	__xen_dma_page_cpu_to_dev(hwdev, dev_addr, size, dir);
-}
-
-void __xen_dma_unmap_page(struct device *hwdev, dma_addr_t handle,
-		size_t size, enum dma_data_direction dir,
-		unsigned long attrs)
-
-{
-	if (is_device_dma_coherent(hwdev))
-		return;
-	if (attrs & DMA_ATTR_SKIP_CPU_SYNC)
-		return;
-
-	__xen_dma_page_dev_to_cpu(hwdev, handle, size, dir);
-}
-
-void __xen_dma_sync_single_for_cpu(struct device *hwdev,
-		dma_addr_t handle, size_t size, enum dma_data_direction dir)
-{
-	if (is_device_dma_coherent(hwdev))
-		return;
-	__xen_dma_page_dev_to_cpu(hwdev, handle, size, dir);
-}
-
-void __xen_dma_sync_single_for_device(struct device *hwdev,
-		dma_addr_t handle, size_t size, enum dma_data_direction dir)
-{
-	if (is_device_dma_coherent(hwdev))
-		return;
-	__xen_dma_page_cpu_to_dev(hwdev, handle, size, dir);
+	if (pfn_valid(PFN_DOWN(handle)))
+		arch_sync_dma_for_device(dev, paddr, size, dir);
+	else if (dir == DMA_FROM_DEVICE)
+		dma_cache_maint(handle, size, GNTTAB_CACHE_INVAL);
+	else
+		dma_cache_maint(handle, size, GNTTAB_CACHE_CLEAN);
 }
 
 bool xen_arch_need_swiotlb(struct device *dev,
@@ -159,7 +110,7 @@ bool xen_arch_need_swiotlb(struct device *dev,
 	 * memory and we are not able to flush the cache.
 	 */
 	return (!hypercall_cflush && (xen_pfn != bfn) &&
-		!is_device_dma_coherent(dev));
+		!dev_is_dma_coherent(dev));
 }
 
 int xen_create_contiguous_region(phys_addr_t pstart, unsigned int order,
@@ -173,16 +124,11 @@ int xen_create_contiguous_region(phys_addr_t pstart, unsigned int order,
 	*dma_handle = pstart;
 	return 0;
 }
-EXPORT_SYMBOL_GPL(xen_create_contiguous_region);
 
 void xen_destroy_contiguous_region(phys_addr_t pstart, unsigned int order)
 {
 	return;
 }
-EXPORT_SYMBOL_GPL(xen_destroy_contiguous_region);
-
-const struct dma_map_ops *xen_dma_ops;
-EXPORT_SYMBOL(xen_dma_ops);
 
 int __init xen_mm_init(void)
 {
@@ -190,7 +136,6 @@ int __init xen_mm_init(void)
 	if (!xen_initial_domain())
 		return 0;
 	xen_swiotlb_init(1, false);
-	xen_dma_ops = &xen_swiotlb_dma_ops;
 
 	cflush.op = 0;
 	cflush.a.dev_bus_addr = 0;

arch/arm64/Kconfig

@@ -13,7 +13,6 @@ config ARM64
 	select ARCH_HAS_DEBUG_VIRTUAL
 	select ARCH_HAS_DEVMEM_IS_ALLOWED
 	select ARCH_HAS_DMA_COHERENT_TO_PFN
-	select ARCH_HAS_DMA_MMAP_PGPROT
 	select ARCH_HAS_DMA_PREP_COHERENT
 	select ARCH_HAS_ACPI_TABLE_UPGRADE if ACPI
 	select ARCH_HAS_ELF_RANDOMIZE

arch/arm64/include/asm/Kbuild

@@ -4,6 +4,7 @@ generic-y += delay.h
 generic-y += div64.h
 generic-y += dma.h
 generic-y += dma-contiguous.h
+generic-y += dma-mapping.h
 generic-y += early_ioremap.h
 generic-y += emergency-restart.h
 generic-y += hw_irq.h

arch/arm64/include/asm/dma-mapping.h

@@ -1,28 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0-only */
-/*
- * Copyright (C) 2012 ARM Ltd.
- */
-#ifndef __ASM_DMA_MAPPING_H
-#define __ASM_DMA_MAPPING_H
-
-#include <linux/types.h>
-#include <linux/vmalloc.h>
-
-#include <xen/xen.h>
-#include <asm/xen/hypervisor.h>
-
-static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
-{
-	return NULL;
-}
-
-/*
- * Do not use this function in a driver, it is only provided for
- * arch/arm/mm/xen.c, which is used by arm64 as well.
- */
-static inline bool is_device_dma_coherent(struct device *dev)
-{
-	return dev->dma_coherent;
-}
-
-#endif	/* __ASM_DMA_MAPPING_H */

arch/arm64/include/asm/pgtable.h

@@ -437,6 +437,18 @@ static inline pmd_t pmd_mkdevmap(pmd_t pmd)
 	__pgprot_modify(prot, PTE_ATTRINDX_MASK, PTE_ATTRINDX(MT_NORMAL_NC) | PTE_PXN | PTE_UXN)
 #define pgprot_device(prot) \
 	__pgprot_modify(prot, PTE_ATTRINDX_MASK, PTE_ATTRINDX(MT_DEVICE_nGnRE) | PTE_PXN | PTE_UXN)
+/*
+ * DMA allocations for non-coherent devices use what the Arm architecture calls
+ * "Normal non-cacheable" memory, which permits speculation, unaligned accesses
+ * and merging of writes.  This is different from "Device-nGnR[nE]" memory which
+ * is intended for MMIO and thus forbids speculation, preserves access size,
+ * requires strict alignment and can also force write responses to come from the
+ * endpoint.
+ */
+#define pgprot_dmacoherent(prot) \
+	__pgprot_modify(prot, PTE_ATTRINDX_MASK, \
+			PTE_ATTRINDX(MT_NORMAL_NC) | PTE_PXN | PTE_UXN)
+
 #define __HAVE_PHYS_MEM_ACCESS_PROT
 struct file;
 extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,

arch/arm64/include/asm/xen/page-coherent.h

@@ -1,77 +1,2 @@
 /* SPDX-License-Identifier: GPL-2.0 */
-#ifndef _ASM_ARM64_XEN_PAGE_COHERENT_H
-#define _ASM_ARM64_XEN_PAGE_COHERENT_H
-
-#include <linux/dma-mapping.h>
-#include <asm/page.h>
 #include <xen/arm/page-coherent.h>
-
-static inline void *xen_alloc_coherent_pages(struct device *hwdev, size_t size,
-		dma_addr_t *dma_handle, gfp_t flags, unsigned long attrs)
-{
-	return dma_direct_alloc(hwdev, size, dma_handle, flags, attrs);
-}
-
-static inline void xen_free_coherent_pages(struct device *hwdev, size_t size,
-		void *cpu_addr, dma_addr_t dma_handle, unsigned long attrs)
-{
-	dma_direct_free(hwdev, size, cpu_addr, dma_handle, attrs);
-}
-
-static inline void xen_dma_sync_single_for_cpu(struct device *hwdev,
-		dma_addr_t handle, size_t size, enum dma_data_direction dir)
-{
-	unsigned long pfn = PFN_DOWN(handle);
-
-	if (pfn_valid(pfn))
-		dma_direct_sync_single_for_cpu(hwdev, handle, size, dir);
-	else
-		__xen_dma_sync_single_for_cpu(hwdev, handle, size, dir);
-}
-
-static inline void xen_dma_sync_single_for_device(struct device *hwdev,
-		dma_addr_t handle, size_t size, enum dma_data_direction dir)
-{
-	unsigned long pfn = PFN_DOWN(handle);
-	if (pfn_valid(pfn))
-		dma_direct_sync_single_for_device(hwdev, handle, size, dir);
-	else
-		__xen_dma_sync_single_for_device(hwdev, handle, size, dir);
-}
-
-static inline void xen_dma_map_page(struct device *hwdev, struct page *page,
-	     dma_addr_t dev_addr, unsigned long offset, size_t size,
-	     enum dma_data_direction dir, unsigned long attrs)
-{
-	unsigned long page_pfn = page_to_xen_pfn(page);
-	unsigned long dev_pfn = XEN_PFN_DOWN(dev_addr);
-	unsigned long compound_pages =
-		(1<<compound_order(page)) * XEN_PFN_PER_PAGE;
-	bool local = (page_pfn <= dev_pfn) &&
-		(dev_pfn - page_pfn < compound_pages);
-
-	if (local)
-		dma_direct_map_page(hwdev, page, offset, size, dir, attrs);
-	else
-		__xen_dma_map_page(hwdev, page, dev_addr, offset, size, dir, attrs);
-}
-
-static inline void xen_dma_unmap_page(struct device *hwdev, dma_addr_t handle,
-		size_t size, enum dma_data_direction dir, unsigned long attrs)
-{
-	unsigned long pfn = PFN_DOWN(handle);
-	/*
-	 * Dom0 is mapped 1:1, while the Linux page can be spanned accross
-	 * multiple Xen page, it's not possible to have a mix of local and
-	 * foreign Xen page. Dom0 is mapped 1:1, so calling pfn_valid on a
-	 * foreign mfn will always return false. If the page is local we can
-	 * safely call the native dma_ops function, otherwise we call the xen
-	 * specific function.
-	 */
-	if (pfn_valid(pfn))
-		dma_direct_unmap_page(hwdev, handle, size, dir, attrs);
-	else
-		__xen_dma_unmap_page(hwdev, handle, size, dir, attrs);
-}
-
-#endif /* _ASM_ARM64_XEN_PAGE_COHERENT_H */

arch/arm64/mm/dma-mapping.c

@@ -8,15 +8,11 @@
 #include <linux/cache.h>
 #include <linux/dma-noncoherent.h>
 #include <linux/dma-iommu.h>
+#include <xen/xen.h>
+#include <xen/swiotlb-xen.h>
 
 #include <asm/cacheflush.h>
 
-pgprot_t arch_dma_mmap_pgprot(struct device *dev, pgprot_t prot,
-		unsigned long attrs)
-{
-	return pgprot_writecombine(prot);
-}
-
 void arch_sync_dma_for_device(struct device *dev, phys_addr_t paddr,
 		size_t size, enum dma_data_direction dir)
 {
@@ -34,12 +30,6 @@ void arch_dma_prep_coherent(struct page *page, size_t size)
 	__dma_flush_area(page_address(page), size);
 }
 
-static int __init arm64_dma_init(void)
-{
-	return dma_atomic_pool_init(GFP_DMA32, __pgprot(PROT_NORMAL_NC));
-}
-arch_initcall(arm64_dma_init);
-
 #ifdef CONFIG_IOMMU_DMA
 void arch_teardown_dma_ops(struct device *dev)
 {
@@ -64,6 +54,6 @@ void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size,
 
 #ifdef CONFIG_XEN
 	if (xen_initial_domain())
-		dev->dma_ops = xen_dma_ops;
+		dev->dma_ops = &xen_swiotlb_dma_ops;
 #endif
 }