1 files changed, 199 insertions, 50 deletions

diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
index 0eca38125..c941e9304 100644
--- a/arch/arm/mm/dma-mapping.c
+++ b/arch/arm/mm/dma-mapping.c
@@ -42,6 +42,55 @@
 #include "dma.h"
 #include "mm.h"
 
+struct arm_dma_alloc_args {
+	struct device *dev;
+	size_t size;
+	gfp_t gfp;
+	pgprot_t prot;
+	const void *caller;
+	bool want_vaddr;
+};
+
+struct arm_dma_free_args {
+	struct device *dev;
+	size_t size;
+	void *cpu_addr;
+	struct page *page;
+	bool want_vaddr;
+};
+
+struct arm_dma_allocator {
+	void *(*alloc)(struct arm_dma_alloc_args *args,
+		       struct page **ret_page);
+	void (*free)(struct arm_dma_free_args *args);
+};
+
+struct arm_dma_buffer {
+	struct list_head list;
+	void *virt;
+	struct arm_dma_allocator *allocator;
+};
+
+static LIST_HEAD(arm_dma_bufs);
+static DEFINE_SPINLOCK(arm_dma_bufs_lock);
+
+static struct arm_dma_buffer *arm_dma_buffer_find(void *virt)
+{
+	struct arm_dma_buffer *buf, *found = NULL;
+	unsigned long flags;
+
+	spin_lock_irqsave(&arm_dma_bufs_lock, flags);
+	list_for_each_entry(buf, &arm_dma_bufs, list) {
+		if (buf->virt == virt) {
+			list_del(&buf->list);
+			found = buf;
+			break;
+		}
+	}
+	spin_unlock_irqrestore(&arm_dma_bufs_lock, flags);
+	return found;
+}
+
 /*
  * The DMA API is built upon the notion of "buffer ownership".  A buffer
  * is either exclusively owned by the CPU (and therefore may be accessed
@@ -592,7 +641,7 @@ static inline pgprot_t __get_dma_pgprot(struct dma_attrs *attrs, pgprot_t prot)
 #define __alloc_remap_buffer(dev, size, gfp, prot, ret, c, wv)	NULL
 #define __alloc_from_pool(size, ret_page)			NULL
 #define __alloc_from_contiguous(dev, size, prot, ret, c, wv)	NULL
-#define __free_from_pool(cpu_addr, size)			0
+#define __free_from_pool(cpu_addr, size)			do { } while (0)
 #define __free_from_contiguous(dev, page, cpu_addr, size, wv)	do { } while (0)
 #define __dma_free_remap(cpu_addr, size)			do { } while (0)
 
@@ -610,7 +659,78 @@ static void *__alloc_simple_buffer(struct device *dev, size_t size, gfp_t gfp,
 	return page_address(page);
 }
 
+static void *simple_allocator_alloc(struct arm_dma_alloc_args *args,
+				    struct page **ret_page)
+{
+	return __alloc_simple_buffer(args->dev, args->size, args->gfp,
+				     ret_page);
+}
+
+static void simple_allocator_free(struct arm_dma_free_args *args)
+{
+	__dma_free_buffer(args->page, args->size);
+}
 
+static struct arm_dma_allocator simple_allocator = {
+	.alloc = simple_allocator_alloc,
+	.free = simple_allocator_free,
+};
+
+static void *cma_allocator_alloc(struct arm_dma_alloc_args *args,
+				 struct page **ret_page)
+{
+	return __alloc_from_contiguous(args->dev, args->size, args->prot,
+				       ret_page, args->caller,
+				       args->want_vaddr);
+}
+
+static void cma_allocator_free(struct arm_dma_free_args *args)
+{
+	__free_from_contiguous(args->dev, args->page, args->cpu_addr,
+			       args->size, args->want_vaddr);
+}
+
+static struct arm_dma_allocator cma_allocator = {
+	.alloc = cma_allocator_alloc,
+	.free = cma_allocator_free,
+};
+
+static void *pool_allocator_alloc(struct arm_dma_alloc_args *args,
+				  struct page **ret_page)
+{
+	return __alloc_from_pool(args->size, ret_page);
+}
+
+static void pool_allocator_free(struct arm_dma_free_args *args)
+{
+	__free_from_pool(args->cpu_addr, args->size);
+}
+
+static struct arm_dma_allocator pool_allocator = {
+	.alloc = pool_allocator_alloc,
+	.free = pool_allocator_free,
+};
+
+static void *remap_allocator_alloc(struct arm_dma_alloc_args *args,
+				   struct page **ret_page)
+{
+	return __alloc_remap_buffer(args->dev, args->size, args->gfp,
+				    args->prot, ret_page, args->caller,
+				    args->want_vaddr);
+}
+
+static void remap_allocator_free(struct arm_dma_free_args *args)
+{
+	if (args->want_vaddr)
+		__dma_free_remap(args->cpu_addr, args->size);
+
+	__dma_free_buffer(args->page, args->size);
+}
+
+static struct arm_dma_allocator remap_allocator = {
+	.alloc = remap_allocator_alloc,
+	.free = remap_allocator_free,
+};
 
 static void *__dma_alloc(struct device *dev, size_t size, dma_addr_t *handle,
 			 gfp_t gfp, pgprot_t prot, bool is_coherent,
@@ -619,7 +739,16 @@ static void *__dma_alloc(struct device *dev, size_t size, dma_addr_t *handle,
 	u64 mask = get_coherent_dma_mask(dev);
 	struct page *page = NULL;
 	void *addr;
-	bool want_vaddr;
+	bool allowblock, cma;
+	struct arm_dma_buffer *buf;
+	struct arm_dma_alloc_args args = {
+		.dev = dev,
+		.size = PAGE_ALIGN(size),
+		.gfp = gfp,
+		.prot = prot,
+		.caller = caller,
+		.want_vaddr = !dma_get_attr(DMA_ATTR_NO_KERNEL_MAPPING, attrs),
+	};
 
 #ifdef CONFIG_DMA_API_DEBUG
 	u64 limit = (mask + 1) & ~mask;
@@ -633,6 +762,11 @@ static void *__dma_alloc(struct device *dev, size_t size, dma_addr_t *handle,
 	if (!mask)
 		return NULL;
 
+	buf = kzalloc(sizeof(*buf),
+		      gfp & ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM));
+	if (!buf)
+		return NULL;
+
 	if (mask < 0xffffffffULL)
 		gfp |= GFP_DMA;
 
@@ -644,28 +778,37 @@ static void *__dma_alloc(struct device *dev, size_t size, dma_addr_t *handle,
 	 * platform; see CONFIG_HUGETLBFS.
 	 */
 	gfp &= ~(__GFP_COMP);
+	args.gfp = gfp;
 
 	*handle = DMA_ERROR_CODE;
-	size = PAGE_ALIGN(size);
-	want_vaddr = !dma_get_attr(DMA_ATTR_NO_KERNEL_MAPPING, attrs);
-
-	if (nommu())
-		addr = __alloc_simple_buffer(dev, size, gfp, &page);
-	else if (dev_get_cma_area(dev) && (gfp & __GFP_DIRECT_RECLAIM))
-		addr = __alloc_from_contiguous(dev, size, prot, &page,
-					       caller, want_vaddr);
-	else if (is_coherent)
-		addr = __alloc_simple_buffer(dev, size, gfp, &page);
-	else if (!gfpflags_allow_blocking(gfp))
-		addr = __alloc_from_pool(size, &page);
+	allowblock = gfpflags_allow_blocking(gfp);
+	cma = allowblock ? dev_get_cma_area(dev) : false;
+
+	if (cma)
+		buf->allocator = &cma_allocator;
+	else if (nommu() || is_coherent)
+		buf->allocator = &simple_allocator;
+	else if (allowblock)
+		buf->allocator = &remap_allocator;
 	else
-		addr = __alloc_remap_buffer(dev, size, gfp, prot, &page,
-					    caller, want_vaddr);
+		buf->allocator = &pool_allocator;
+
+	addr = buf->allocator->alloc(&args, &page);
+
+	if (page) {
+		unsigned long flags;
 
-	if (page)
 		*handle = pfn_to_dma(dev, page_to_pfn(page));
+		buf->virt = args.want_vaddr ? addr : page;
+
+		spin_lock_irqsave(&arm_dma_bufs_lock, flags);
+		list_add(&buf->list, &arm_dma_bufs);
+		spin_unlock_irqrestore(&arm_dma_bufs_lock, flags);
+	} else {
+		kfree(buf);
+	}
 
-	return want_vaddr ? addr : page;
+	return args.want_vaddr ? addr : page;
 }
 
 /*
@@ -741,25 +884,21 @@ static void __arm_dma_free(struct device *dev, size_t size, void *cpu_addr,
 			   bool is_coherent)
 {
 	struct page *page = pfn_to_page(dma_to_pfn(dev, handle));
-	bool want_vaddr = !dma_get_attr(DMA_ATTR_NO_KERNEL_MAPPING, attrs);
-
-	size = PAGE_ALIGN(size);
-
-	if (nommu()) {
-		__dma_free_buffer(page, size);
-	} else if (!is_coherent && __free_from_pool(cpu_addr, size)) {
+	struct arm_dma_buffer *buf;
+	struct arm_dma_free_args args = {
+		.dev = dev,
+		.size = PAGE_ALIGN(size),
+		.cpu_addr = cpu_addr,
+		.page = page,
+		.want_vaddr = !dma_get_attr(DMA_ATTR_NO_KERNEL_MAPPING, attrs),
+	};
+
+	buf = arm_dma_buffer_find(cpu_addr);
+	if (WARN(!buf, "Freeing invalid buffer %p\n", cpu_addr))
 		return;
-	} else if (!dev_get_cma_area(dev)) {
-		if (want_vaddr && !is_coherent)
-			__dma_free_remap(cpu_addr, size);
-		__dma_free_buffer(page, size);
-	} else {
-		/*
-		 * Non-atomic allocations cannot be freed with IRQs disabled
-		 */
-		WARN_ON(irqs_disabled());
-		__free_from_contiguous(dev, page, cpu_addr, size, want_vaddr);
-	}
+
+	buf->allocator->free(&args);
+	kfree(buf);
 }
 
 void arm_dma_free(struct device *dev, size_t size, void *cpu_addr,
@@ -1122,6 +1261,9 @@ static inline void __free_iova(struct dma_iommu_mapping *mapping,
 	spin_unlock_irqrestore(&mapping->lock, flags);
 }
 
+/* We'll try 2M, 1M, 64K, and finally 4K; array must end with 0! */
+static const int iommu_order_array[] = { 9, 8, 4, 0 };
+
 static struct page **__iommu_alloc_buffer(struct device *dev, size_t size,
 					  gfp_t gfp, struct dma_attrs *attrs)
 {
@@ -1129,6 +1271,7 @@ static struct page **__iommu_alloc_buffer(struct device *dev, size_t size,
 	int count = size >> PAGE_SHIFT;
 	int array_size = count * sizeof(struct page *);
 	int i = 0;
+	int order_idx = 0;
 
 	if (array_size <= PAGE_SIZE)
 		pages = kzalloc(array_size, GFP_KERNEL);
@@ -1154,6 +1297,10 @@ static struct page **__iommu_alloc_buffer(struct device *dev, size_t size,
 		return pages;
 	}
 
+	/* Go straight to 4K chunks if caller says it's OK. */
+	if (dma_get_attr(DMA_ATTR_ALLOC_SINGLE_PAGES, attrs))
+		order_idx = ARRAY_SIZE(iommu_order_array) - 1;
+
 	/*
 	 * IOMMU can map any pages, so himem can also be used here
 	 */
@@ -1162,22 +1309,24 @@ static struct page **__iommu_alloc_buffer(struct device *dev, size_t size,
 	while (count) {
 		int j, order;
 
-		for (order = __fls(count); order > 0; --order) {
-			/*
-			 * We do not want OOM killer to be invoked as long
-			 * as we can fall back to single pages, so we force
-			 * __GFP_NORETRY for orders higher than zero.
-			 */
-			pages[i] = alloc_pages(gfp | __GFP_NORETRY, order);
-			if (pages[i])
-				break;
+		order = iommu_order_array[order_idx];
+
+		/* Drop down when we get small */
+		if (__fls(count) < order) {
+			order_idx++;
+			continue;
 		}
 
-		if (!pages[i]) {
-			/*
-			 * Fall back to single page allocation.
-			 * Might invoke OOM killer as last resort.
-			 */
+		if (order) {
+			/* See if it's easy to allocate a high-order chunk */
+			pages[i] = alloc_pages(gfp | __GFP_NORETRY, order);
+
+			/* Go down a notch at first sign of pressure */
+			if (!pages[i]) {
+				order_idx++;
+				continue;
+			}
+		} else {
 			pages[i] = alloc_pages(gfp, 0);
 			if (!pages[i])
 				goto error;