From: Felix Fietkau Date: Thu, 15 Aug 2013 09:23:07 +0000 (+0000) Subject: kernel: improve dma ops inlining patches X-Git-Url: http://git.lede-project.org./?a=commitdiff_plain;h=e27f67b30a5c1a838fd4dc7762f43d41361ae4e2;p=openwrt%2Fstaging%2Fstintel.git kernel: improve dma ops inlining patches Signed-off-by: Felix Fietkau SVN-Revision: 37790 --- diff --git a/target/linux/generic/patches-3.10/130-mips_remove_plat_dma_functions.patch b/target/linux/generic/patches-3.10/130-mips_remove_plat_dma_functions.patch index 76a3584274..958841da1a 100644 --- a/target/linux/generic/patches-3.10/130-mips_remove_plat_dma_functions.patch +++ b/target/linux/generic/patches-3.10/130-mips_remove_plat_dma_functions.patch @@ -1,35 +1,20 @@ +From 151c4e4a06b0b8d16c2fd392bb0e33868b12357f Mon Sep 17 00:00:00 2001 From: Felix Fietkau -Subject: [PATCH 1/2] MIPS: remove unnecessary platform dma helper functions +Date: Mon, 12 Aug 2013 12:45:52 +0200 +Subject: [PATCH] MIPS: remove unnecessary platform dma helper functions Signed-off-by: Felix Fietkau --- ---- a/arch/mips/mm/dma-default.c -+++ b/arch/mips/mm/dma-default.c -@@ -289,7 +289,6 @@ static void mips_dma_sync_single_for_cpu - static void mips_dma_sync_single_for_device(struct device *dev, - dma_addr_t dma_handle, size_t size, enum dma_data_direction direction) - { -- plat_extra_sync_for_device(dev); - if (!plat_device_is_coherent(dev)) - __dma_sync(dma_addr_to_page(dev, dma_handle), - dma_handle & ~PAGE_MASK, size, direction); -@@ -323,7 +322,7 @@ static void mips_dma_sync_sg_for_device( - - int mips_dma_mapping_error(struct device *dev, dma_addr_t dma_addr) - { -- return plat_dma_mapping_error(dev, dma_addr); -+ return 0; - } - - int mips_dma_supported(struct device *dev, u64 mask) -@@ -336,7 +335,6 @@ void dma_cache_sync(struct device *dev, - { - BUG_ON(direction == DMA_NONE); - -- plat_extra_sync_for_device(dev); - if (!plat_device_is_coherent(dev)) - __dma_sync_virtual(vaddr, size, direction); - } + arch/mips/include/asm/mach-cavium-octeon/dma-coherence.h | 12 ------------ + arch/mips/include/asm/mach-generic/dma-coherence.h | 10 ---------- + arch/mips/include/asm/mach-ip27/dma-coherence.h | 10 ---------- + arch/mips/include/asm/mach-ip32/dma-coherence.h | 11 ----------- + arch/mips/include/asm/mach-jazz/dma-coherence.h | 10 ---------- + arch/mips/include/asm/mach-loongson/dma-coherence.h | 10 ---------- + arch/mips/include/asm/mach-powertv/dma-coherence.h | 10 ---------- + arch/mips/mm/dma-default.c | 4 +--- + 8 files changed, 1 insertion(+), 76 deletions(-) + --- a/arch/mips/include/asm/mach-cavium-octeon/dma-coherence.h +++ b/arch/mips/include/asm/mach-cavium-octeon/dma-coherence.h @@ -46,22 +46,11 @@ static inline int plat_dma_supported(str @@ -170,3 +155,30 @@ Signed-off-by: Felix Fietkau static inline int plat_device_is_coherent(struct device *dev) { return 0; +--- a/arch/mips/mm/dma-default.c ++++ b/arch/mips/mm/dma-default.c +@@ -289,7 +289,6 @@ static void mips_dma_sync_single_for_cpu + static void mips_dma_sync_single_for_device(struct device *dev, + dma_addr_t dma_handle, size_t size, enum dma_data_direction direction) + { +- plat_extra_sync_for_device(dev); + if (!plat_device_is_coherent(dev)) + __dma_sync(dma_addr_to_page(dev, dma_handle), + dma_handle & ~PAGE_MASK, size, direction); +@@ -323,7 +322,7 @@ static void mips_dma_sync_sg_for_device( + + int mips_dma_mapping_error(struct device *dev, dma_addr_t dma_addr) + { +- return plat_dma_mapping_error(dev, dma_addr); ++ return 0; + } + + int mips_dma_supported(struct device *dev, u64 mask) +@@ -336,7 +335,6 @@ void dma_cache_sync(struct device *dev, + { + BUG_ON(direction == DMA_NONE); + +- plat_extra_sync_for_device(dev); + if (!plat_device_is_coherent(dev)) + __dma_sync_virtual(vaddr, size, direction); + } diff --git a/target/linux/generic/patches-3.10/131-improve_noncoherent_dma_checks.patch b/target/linux/generic/patches-3.10/131-improve_noncoherent_dma_checks.patch new file mode 100644 index 0000000000..c18e7a3e21 --- /dev/null +++ b/target/linux/generic/patches-3.10/131-improve_noncoherent_dma_checks.patch @@ -0,0 +1,89 @@ +From d593f8fc627f8cdaee9c14e4d22b0770a09baaf1 Mon Sep 17 00:00:00 2001 +From: Felix Fietkau +Date: Thu, 15 Aug 2013 10:47:47 +0200 +Subject: [PATCH] MIPS: improve checks for noncoherent DMA + +Only one MIPS development board actually supports enabling/disabling DMA +coherency at runtime, so it's not a good idea to push the overhead of +checking that configuration setting onto every other supported target as +well. + +Signed-off-by: Felix Fietkau +--- + arch/mips/Kconfig | 6 +++++- + arch/mips/include/asm/dma-coherence.h | 7 +++++++ + arch/mips/include/asm/mach-generic/dma-coherence.h | 4 ---- + arch/mips/mm/dma-default.c | 2 ++ + 4 files changed, 14 insertions(+), 5 deletions(-) + +--- a/arch/mips/Kconfig ++++ b/arch/mips/Kconfig +@@ -296,7 +296,7 @@ config MIPS_MALTA + select CEVT_R4K + select CSRC_R4K + select CSRC_GIC +- select DMA_NONCOHERENT ++ select DMA_MAYBE_COHERENT + select GENERIC_ISA_DMA + select HAVE_PCSPKR_PLATFORM + select IRQ_CPU +@@ -947,6 +947,10 @@ config FW_CFE + config ARCH_DMA_ADDR_T_64BIT + def_bool (HIGHMEM && 64BIT_PHYS_ADDR) || 64BIT + ++config DMA_MAYBE_COHERENT ++ select DMA_NONCOHERENT ++ bool ++ + config DMA_COHERENT + bool + +--- a/arch/mips/include/asm/dma-coherence.h ++++ b/arch/mips/include/asm/dma-coherence.h +@@ -9,7 +9,14 @@ + #ifndef __ASM_DMA_COHERENCE_H + #define __ASM_DMA_COHERENCE_H + ++#ifdef CONFIG_DMA_MAYBE_COHERENT + extern int coherentio; + extern int hw_coherentio; ++#elif defined(CONFIG_DMA_COHERENT) ++#define coherentio 1 ++#else ++#define coherentio 0 ++#endif ++#define hw_coherentio 0 + + #endif +--- a/arch/mips/include/asm/mach-generic/dma-coherence.h ++++ b/arch/mips/include/asm/mach-generic/dma-coherence.h +@@ -49,11 +49,7 @@ static inline int plat_dma_supported(str + + static inline int plat_device_is_coherent(struct device *dev) + { +-#ifdef CONFIG_DMA_COHERENT +- return 1; +-#else + return coherentio; +-#endif + } + + #endif /* __ASM_MACH_GENERIC_DMA_COHERENCE_H */ +--- a/arch/mips/mm/dma-default.c ++++ b/arch/mips/mm/dma-default.c +@@ -22,6 +22,7 @@ + + #include + ++#ifdef CONFIG_DMA_MAYBE_COHERENT + int coherentio = 0; /* User defined DMA coherency from command line. */ + EXPORT_SYMBOL_GPL(coherentio); + int hw_coherentio = 0; /* Actual hardware supported DMA coherency setting. */ +@@ -41,6 +42,7 @@ static int __init setnocoherentio(char * + return 0; + } + early_param("nocoherentio", setnocoherentio); ++#endif + + static inline struct page *dma_addr_to_page(struct device *dev, + dma_addr_t dma_addr) diff --git a/target/linux/generic/patches-3.10/131-mips_inline_dma_ops.patch b/target/linux/generic/patches-3.10/131-mips_inline_dma_ops.patch deleted file mode 100644 index 7ab008086d..0000000000 --- a/target/linux/generic/patches-3.10/131-mips_inline_dma_ops.patch +++ /dev/null @@ -1,661 +0,0 @@ -From: Felix Fietkau -Subject: [PATCH 2/2] MIPS: partially inline dma ops - -Several DMA ops are no-op on many platforms, and the indirection through -the mips_dma_map_ops function table is causing the compiler to emit -unnecessary code. - -Inlining visibly improves network performance in my tests (on a 24Kc -based system), and also slightly reduces code size of a few drivers. - -Signed-off-by: Felix Fietkau ---- ---- a/arch/mips/Kconfig -+++ b/arch/mips/Kconfig -@@ -1426,6 +1426,7 @@ config CPU_CAVIUM_OCTEON - select LIBFDT - select USE_OF - select USB_EHCI_BIG_ENDIAN_MMIO -+ select SYS_HAS_DMA_OPS - help - The Cavium Octeon processor is a highly integrated chip containing - many ethernet hardware widgets for networking tasks. The processor -@@ -1646,6 +1647,9 @@ config SYS_HAS_CPU_XLR - config SYS_HAS_CPU_XLP - bool - -+config SYS_HAS_DMA_OPS -+ bool -+ - # - # CPU may reorder R->R, R->W, W->R, W->W - # Reordering beyond LL and SC is handled in WEAK_REORDERING_BEYOND_LLSC ---- a/arch/mips/include/asm/dma-mapping.h -+++ b/arch/mips/include/asm/dma-mapping.h -@@ -1,6 +1,12 @@ - #ifndef _ASM_DMA_MAPPING_H - #define _ASM_DMA_MAPPING_H - -+#include -+#include -+#include -+#include -+#include -+ - #include - #include - #include -@@ -10,14 +16,47 @@ - #include - #endif - --extern struct dma_map_ops *mips_dma_map_ops; -+void __dma_sync(struct page *page, unsigned long offset, size_t size, -+ enum dma_data_direction direction); -+void *mips_dma_alloc_coherent(struct device *dev, size_t size, -+ dma_addr_t *dma_handle, gfp_t gfp, -+ struct dma_attrs *attrs); -+void mips_dma_free_coherent(struct device *dev, size_t size, void *vaddr, -+ dma_addr_t dma_handle, struct dma_attrs *attrs); - - static inline struct dma_map_ops *get_dma_ops(struct device *dev) - { -+#ifdef CONFIG_SYS_HAS_DMA_OPS - if (dev && dev->archdata.dma_ops) - return dev->archdata.dma_ops; - else - return mips_dma_map_ops; -+#else -+ return NULL; -+#endif -+} -+ -+/* -+ * Warning on the terminology - Linux calls an uncached area coherent; -+ * MIPS terminology calls memory areas with hardware maintained coherency -+ * coherent. -+ */ -+ -+static inline int cpu_is_noncoherent_r10000(struct device *dev) -+{ -+#ifndef CONFIG_SYS_HAS_CPU_R10000 -+ return 0; -+#endif -+ return !plat_device_is_coherent(dev) && -+ (current_cpu_type() == CPU_R10000 || -+ current_cpu_type() == CPU_R12000); -+} -+ -+static inline struct page *dma_addr_to_page(struct device *dev, -+ dma_addr_t dma_addr) -+{ -+ return pfn_to_page( -+ plat_dma_addr_to_phys(dev, dma_addr) >> PAGE_SHIFT); - } - - static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size) -@@ -30,12 +69,309 @@ static inline bool dma_capable(struct de - - static inline void dma_mark_clean(void *addr, size_t size) {} - --#include -+static inline dma_addr_t dma_map_single_attrs(struct device *dev, void *ptr, -+ size_t size, -+ enum dma_data_direction dir, -+ struct dma_attrs *attrs) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ unsigned long offset = (unsigned long)ptr & ~PAGE_MASK; -+ struct page *page = virt_to_page(ptr); -+ dma_addr_t addr; -+ -+ kmemcheck_mark_initialized(ptr, size); -+ BUG_ON(!valid_dma_direction(dir)); -+ if (ops) { -+ addr = ops->map_page(dev, page, offset, size, dir, attrs); -+ } else { -+ if (!plat_device_is_coherent(dev)) -+ __dma_sync(page, offset, size, dir); -+ -+ addr = plat_map_dma_mem_page(dev, page) + offset; -+ } -+ debug_dma_map_page(dev, page, offset, size, dir, addr, true); -+ return addr; -+} -+ -+static inline void dma_unmap_single_attrs(struct device *dev, dma_addr_t addr, -+ size_t size, -+ enum dma_data_direction dir, -+ struct dma_attrs *attrs) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ -+ BUG_ON(!valid_dma_direction(dir)); -+ if (ops) { -+ ops->unmap_page(dev, addr, size, dir, attrs); -+ } else { -+ if (cpu_is_noncoherent_r10000(dev)) -+ __dma_sync(dma_addr_to_page(dev, addr), -+ addr & ~PAGE_MASK, size, dir); -+ -+ plat_unmap_dma_mem(dev, addr, size, dir); -+ } -+ debug_dma_unmap_page(dev, addr, size, dir, true); -+} -+ -+static inline int dma_map_sg_attrs(struct device *dev, struct scatterlist *sg, -+ int nents, enum dma_data_direction dir, -+ struct dma_attrs *attrs) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ int i, ents; -+ struct scatterlist *s; -+ -+ for_each_sg(sg, s, nents, i) -+ kmemcheck_mark_initialized(sg_virt(s), s->length); -+ BUG_ON(!valid_dma_direction(dir)); -+ if (ops) { -+ ents = ops->map_sg(dev, sg, nents, dir, attrs); -+ } else { -+ for_each_sg(sg, s, nents, i) { -+ struct page *page = sg_page(s); -+ -+ if (!plat_device_is_coherent(dev)) -+ __dma_sync(page, s->offset, s->length, dir); -+ s->dma_address = -+ plat_map_dma_mem_page(dev, page) + s->offset; -+ } -+ ents = nents; -+ } -+ debug_dma_map_sg(dev, sg, nents, ents, dir); -+ -+ return ents; -+} -+ -+static inline void dma_unmap_sg_attrs(struct device *dev, struct scatterlist *sg, -+ int nents, enum dma_data_direction dir, -+ struct dma_attrs *attrs) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ struct scatterlist *s; -+ int i; -+ -+ BUG_ON(!valid_dma_direction(dir)); -+ debug_dma_unmap_sg(dev, sg, nents, dir); -+ if (ops) { -+ ops->unmap_sg(dev, sg, nents, dir, attrs); -+ return; -+ } -+ -+ for_each_sg(sg, s, nents, i) { -+ if (!plat_device_is_coherent(dev) && dir != DMA_TO_DEVICE) -+ __dma_sync(sg_page(s), s->offset, s->length, dir); -+ plat_unmap_dma_mem(dev, s->dma_address, s->length, dir); -+ } -+} -+ -+static inline dma_addr_t dma_map_page(struct device *dev, struct page *page, -+ size_t offset, size_t size, -+ enum dma_data_direction dir) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ dma_addr_t addr; -+ -+ kmemcheck_mark_initialized(page_address(page) + offset, size); -+ BUG_ON(!valid_dma_direction(dir)); -+ if (ops) { -+ addr = ops->map_page(dev, page, offset, size, dir, NULL); -+ } else { -+ if (!plat_device_is_coherent(dev)) -+ __dma_sync(page, offset, size, dir); -+ -+ addr = plat_map_dma_mem_page(dev, page) + offset; -+ } -+ debug_dma_map_page(dev, page, offset, size, dir, addr, false); -+ -+ return addr; -+} -+ -+static inline void dma_unmap_page(struct device *dev, dma_addr_t addr, -+ size_t size, enum dma_data_direction dir) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ -+ BUG_ON(!valid_dma_direction(dir)); -+ if (ops) { -+ ops->unmap_page(dev, addr, size, dir, NULL); -+ } else { -+ if (cpu_is_noncoherent_r10000(dev)) -+ __dma_sync(dma_addr_to_page(dev, addr), -+ addr & ~PAGE_MASK, size, dir); -+ -+ plat_unmap_dma_mem(dev, addr, size, dir); -+ } -+ debug_dma_unmap_page(dev, addr, size, dir, false); -+} -+ -+static inline void dma_sync_single_for_cpu(struct device *dev, dma_addr_t addr, -+ size_t size, -+ enum dma_data_direction dir) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ -+ BUG_ON(!valid_dma_direction(dir)); -+ if (ops) -+ ops->sync_single_for_cpu(dev, addr, size, dir); -+ else if (cpu_is_noncoherent_r10000(dev)) -+ __dma_sync(dma_addr_to_page(dev, addr), -+ addr & ~PAGE_MASK, size, dir); -+ debug_dma_sync_single_for_cpu(dev, addr, size, dir); -+} -+ -+static inline void dma_sync_single_for_device(struct device *dev, -+ dma_addr_t addr, size_t size, -+ enum dma_data_direction dir) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ -+ BUG_ON(!valid_dma_direction(dir)); -+ if (ops) -+ ops->sync_single_for_device(dev, addr, size, dir); -+ else if (!plat_device_is_coherent(dev)) -+ __dma_sync(dma_addr_to_page(dev, addr), -+ addr & ~PAGE_MASK, size, dir); -+ debug_dma_sync_single_for_device(dev, addr, size, dir); -+} -+ -+static inline void dma_sync_single_range_for_cpu(struct device *dev, -+ dma_addr_t addr, -+ unsigned long offset, -+ 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) -+ ops->sync_single_for_cpu(dev, addr + offset, size, dir); -+ else if (cpu_is_noncoherent_r10000(dev)) -+ __dma_sync(dma_addr_to_page(dev, addr + offset), -+ (addr + offset) & ~PAGE_MASK, size, dir); -+ debug_dma_sync_single_range_for_cpu(dev, addr, offset, size, dir); -+} -+ -+static inline void dma_sync_single_range_for_device(struct device *dev, -+ dma_addr_t addr, -+ unsigned long offset, -+ 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) -+ ops->sync_single_for_device(dev, addr + offset, size, dir); -+ else if (!plat_device_is_coherent(dev)) -+ __dma_sync(dma_addr_to_page(dev, addr + offset), -+ (addr + offset) & ~PAGE_MASK, size, dir); -+ debug_dma_sync_single_range_for_device(dev, addr, offset, size, dir); -+} -+ -+static inline void -+dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sg, -+ int nelems, enum dma_data_direction dir) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ struct scatterlist *s; -+ int i; -+ -+ BUG_ON(!valid_dma_direction(dir)); -+ if (ops) -+ ops->sync_sg_for_cpu(dev, sg, nelems, dir); -+ else if (cpu_is_noncoherent_r10000(dev)) { -+ for_each_sg(sg, s, nelems, i) -+ __dma_sync(sg_page(s), s->offset, s->length, dir); -+ } -+ debug_dma_sync_sg_for_cpu(dev, sg, nelems, dir); -+} -+ -+static inline void -+dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg, -+ int nelems, enum dma_data_direction dir) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ struct scatterlist *s; -+ int i; -+ -+ BUG_ON(!valid_dma_direction(dir)); -+ if (ops) -+ ops->sync_sg_for_device(dev, sg, nelems, dir); -+ else if (!plat_device_is_coherent(dev)) { -+ for_each_sg(sg, s, nelems, i) -+ __dma_sync(sg_page(s), s->offset, s->length, dir); -+ } -+ debug_dma_sync_sg_for_device(dev, sg, nelems, dir); -+ -+} -+ -+#define dma_map_single(d, a, s, r) dma_map_single_attrs(d, a, s, r, NULL) -+#define dma_unmap_single(d, a, s, r) dma_unmap_single_attrs(d, a, s, r, NULL) -+#define dma_map_sg(d, s, n, r) dma_map_sg_attrs(d, s, n, r, NULL) -+#define dma_unmap_sg(d, s, n, r) dma_unmap_sg_attrs(d, s, n, r, NULL) -+ -+extern int dma_common_mmap(struct device *dev, struct vm_area_struct *vma, -+ void *cpu_addr, dma_addr_t dma_addr, size_t size); -+ -+/** -+ * dma_mmap_attrs - map a coherent DMA allocation into user space -+ * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices -+ * @vma: vm_area_struct describing requested user mapping -+ * @cpu_addr: kernel CPU-view address returned from dma_alloc_attrs -+ * @handle: device-view address returned from dma_alloc_attrs -+ * @size: size of memory originally requested in dma_alloc_attrs -+ * @attrs: attributes of mapping properties requested in dma_alloc_attrs -+ * -+ * Map a coherent DMA buffer previously allocated by dma_alloc_attrs -+ * into user space. The coherent DMA buffer must not be freed by the -+ * driver until the user space mapping has been released. -+ */ -+static inline int -+dma_mmap_attrs(struct device *dev, struct vm_area_struct *vma, void *cpu_addr, -+ dma_addr_t dma_addr, size_t size, struct dma_attrs *attrs) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ BUG_ON(!ops); -+ if (ops && ops->mmap) -+ return ops->mmap(dev, vma, cpu_addr, dma_addr, size, attrs); -+ return dma_common_mmap(dev, vma, cpu_addr, dma_addr, size); -+} -+ -+#define dma_mmap_coherent(d, v, c, h, s) dma_mmap_attrs(d, v, c, h, s, NULL) -+ -+static inline int dma_mmap_writecombine(struct device *dev, struct vm_area_struct *vma, -+ void *cpu_addr, dma_addr_t dma_addr, size_t size) -+{ -+ DEFINE_DMA_ATTRS(attrs); -+ dma_set_attr(DMA_ATTR_WRITE_COMBINE, &attrs); -+ return dma_mmap_attrs(dev, vma, cpu_addr, dma_addr, size, &attrs); -+} -+ -+int -+dma_common_get_sgtable(struct device *dev, struct sg_table *sgt, -+ void *cpu_addr, dma_addr_t dma_addr, size_t size); -+ -+static inline int -+dma_get_sgtable_attrs(struct device *dev, struct sg_table *sgt, void *cpu_addr, -+ dma_addr_t dma_addr, size_t size, struct dma_attrs *attrs) -+{ -+ struct dma_map_ops *ops = get_dma_ops(dev); -+ BUG_ON(!ops); -+ if (ops && ops->get_sgtable) -+ return ops->get_sgtable(dev, sgt, cpu_addr, dma_addr, size, -+ attrs); -+ return dma_common_get_sgtable(dev, sgt, cpu_addr, dma_addr, size); -+} -+ -+#define dma_get_sgtable(d, t, v, h, s) dma_get_sgtable_attrs(d, t, v, h, s, NULL) -+ - - static inline int dma_supported(struct device *dev, u64 mask) - { - struct dma_map_ops *ops = get_dma_ops(dev); -- return ops->dma_supported(dev, mask); -+ if (ops) -+ return ops->dma_supported(dev, mask); -+ return plat_dma_supported(dev, mask); - } - - static inline int dma_mapping_error(struct device *dev, u64 mask) -@@ -43,7 +379,9 @@ static inline int dma_mapping_error(stru - struct dma_map_ops *ops = get_dma_ops(dev); - - debug_dma_mapping_error(dev, mask); -- return ops->mapping_error(dev, mask); -+ if (ops) -+ return ops->mapping_error(dev, mask); -+ return 0; - } - - static inline int -@@ -69,7 +407,11 @@ static inline void *dma_alloc_attrs(stru - void *ret; - struct dma_map_ops *ops = get_dma_ops(dev); - -- ret = ops->alloc(dev, size, dma_handle, gfp, attrs); -+ if (ops) -+ ret = ops->alloc(dev, size, dma_handle, gfp, attrs); -+ else -+ ret = mips_dma_alloc_coherent(dev, size, dma_handle, gfp, -+ attrs); - - debug_dma_alloc_coherent(dev, size, *dma_handle, ret); - -@@ -84,7 +426,10 @@ static inline void dma_free_attrs(struct - { - struct dma_map_ops *ops = get_dma_ops(dev); - -- ops->free(dev, size, vaddr, dma_handle, attrs); -+ if (ops) -+ ops->free(dev, size, vaddr, dma_handle, attrs); -+ else -+ mips_dma_free_coherent(dev, size, vaddr, dma_handle, attrs); - - debug_dma_free_coherent(dev, size, vaddr, dma_handle); - } ---- a/arch/mips/mm/dma-default.c -+++ b/arch/mips/mm/dma-default.c -@@ -42,26 +42,6 @@ static int __init setnocoherentio(char * - } - early_param("nocoherentio", setnocoherentio); - --static inline struct page *dma_addr_to_page(struct device *dev, -- dma_addr_t dma_addr) --{ -- return pfn_to_page( -- plat_dma_addr_to_phys(dev, dma_addr) >> PAGE_SHIFT); --} -- --/* -- * Warning on the terminology - Linux calls an uncached area coherent; -- * MIPS terminology calls memory areas with hardware maintained coherency -- * coherent. -- */ -- --static inline int cpu_is_noncoherent_r10000(struct device *dev) --{ -- return !plat_device_is_coherent(dev) && -- (current_cpu_type() == CPU_R10000 || -- current_cpu_type() == CPU_R12000); --} -- - static gfp_t massage_gfp_flags(const struct device *dev, gfp_t gfp) - { - gfp_t dma_flag; -@@ -117,8 +97,9 @@ void *dma_alloc_noncoherent(struct devic - } - EXPORT_SYMBOL(dma_alloc_noncoherent); - --static void *mips_dma_alloc_coherent(struct device *dev, size_t size, -- dma_addr_t * dma_handle, gfp_t gfp, struct dma_attrs *attrs) -+void *mips_dma_alloc_coherent(struct device *dev, size_t size, -+ dma_addr_t *dma_handle, gfp_t gfp, -+ struct dma_attrs *attrs) - { - void *ret; - -@@ -142,6 +123,7 @@ static void *mips_dma_alloc_coherent(str - - return ret; - } -+EXPORT_SYMBOL(mips_dma_alloc_coherent); - - - void dma_free_noncoherent(struct device *dev, size_t size, void *vaddr, -@@ -152,8 +134,8 @@ void dma_free_noncoherent(struct device - } - EXPORT_SYMBOL(dma_free_noncoherent); - --static void mips_dma_free_coherent(struct device *dev, size_t size, void *vaddr, -- dma_addr_t dma_handle, struct dma_attrs *attrs) -+void mips_dma_free_coherent(struct device *dev, size_t size, void *vaddr, -+ dma_addr_t dma_handle, struct dma_attrs *attrs) - { - unsigned long addr = (unsigned long) vaddr; - int order = get_order(size); -@@ -168,6 +150,7 @@ static void mips_dma_free_coherent(struc - - free_pages(addr, get_order(size)); - } -+EXPORT_SYMBOL(mips_dma_free_coherent); - - static inline void __dma_sync_virtual(void *addr, size_t size, - enum dma_data_direction direction) -@@ -196,8 +179,8 @@ static inline void __dma_sync_virtual(vo - * If highmem is not configured then the bulk of this loop gets - * optimized out. - */ --static inline void __dma_sync(struct page *page, -- unsigned long offset, size_t size, enum dma_data_direction direction) -+void __dma_sync(struct page *page, unsigned long offset, size_t size, -+ enum dma_data_direction direction) - { - size_t left = size; - -@@ -226,109 +209,7 @@ static inline void __dma_sync(struct pag - left -= len; - } while (left); - } -- --static void mips_dma_unmap_page(struct device *dev, dma_addr_t dma_addr, -- size_t size, enum dma_data_direction direction, struct dma_attrs *attrs) --{ -- if (cpu_is_noncoherent_r10000(dev)) -- __dma_sync(dma_addr_to_page(dev, dma_addr), -- dma_addr & ~PAGE_MASK, size, direction); -- -- plat_unmap_dma_mem(dev, dma_addr, size, direction); --} -- --static int mips_dma_map_sg(struct device *dev, struct scatterlist *sg, -- int nents, enum dma_data_direction direction, struct dma_attrs *attrs) --{ -- int i; -- -- for (i = 0; i < nents; i++, sg++) { -- if (!plat_device_is_coherent(dev)) -- __dma_sync(sg_page(sg), sg->offset, sg->length, -- direction); -- sg->dma_address = plat_map_dma_mem_page(dev, sg_page(sg)) + -- sg->offset; -- } -- -- return nents; --} -- --static dma_addr_t mips_dma_map_page(struct device *dev, struct page *page, -- unsigned long offset, size_t size, enum dma_data_direction direction, -- struct dma_attrs *attrs) --{ -- if (!plat_device_is_coherent(dev)) -- __dma_sync(page, offset, size, direction); -- -- return plat_map_dma_mem_page(dev, page) + offset; --} -- --static void mips_dma_unmap_sg(struct device *dev, struct scatterlist *sg, -- int nhwentries, enum dma_data_direction direction, -- struct dma_attrs *attrs) --{ -- int i; -- -- for (i = 0; i < nhwentries; i++, sg++) { -- if (!plat_device_is_coherent(dev) && -- direction != DMA_TO_DEVICE) -- __dma_sync(sg_page(sg), sg->offset, sg->length, -- direction); -- plat_unmap_dma_mem(dev, sg->dma_address, sg->length, direction); -- } --} -- --static void mips_dma_sync_single_for_cpu(struct device *dev, -- dma_addr_t dma_handle, size_t size, enum dma_data_direction direction) --{ -- if (cpu_is_noncoherent_r10000(dev)) -- __dma_sync(dma_addr_to_page(dev, dma_handle), -- dma_handle & ~PAGE_MASK, size, direction); --} -- --static void mips_dma_sync_single_for_device(struct device *dev, -- dma_addr_t dma_handle, size_t size, enum dma_data_direction direction) --{ -- if (!plat_device_is_coherent(dev)) -- __dma_sync(dma_addr_to_page(dev, dma_handle), -- dma_handle & ~PAGE_MASK, size, direction); --} -- --static void mips_dma_sync_sg_for_cpu(struct device *dev, -- struct scatterlist *sg, int nelems, enum dma_data_direction direction) --{ -- int i; -- -- /* Make sure that gcc doesn't leave the empty loop body. */ -- for (i = 0; i < nelems; i++, sg++) { -- if (cpu_is_noncoherent_r10000(dev)) -- __dma_sync(sg_page(sg), sg->offset, sg->length, -- direction); -- } --} -- --static void mips_dma_sync_sg_for_device(struct device *dev, -- struct scatterlist *sg, int nelems, enum dma_data_direction direction) --{ -- int i; -- -- /* Make sure that gcc doesn't leave the empty loop body. */ -- for (i = 0; i < nelems; i++, sg++) { -- if (!plat_device_is_coherent(dev)) -- __dma_sync(sg_page(sg), sg->offset, sg->length, -- direction); -- } --} -- --int mips_dma_mapping_error(struct device *dev, dma_addr_t dma_addr) --{ -- return 0; --} -- --int mips_dma_supported(struct device *dev, u64 mask) --{ -- return plat_dma_supported(dev, mask); --} -+EXPORT_SYMBOL(__dma_sync); - - void dma_cache_sync(struct device *dev, void *vaddr, size_t size, - enum dma_data_direction direction) -@@ -341,23 +222,10 @@ void dma_cache_sync(struct device *dev, - - EXPORT_SYMBOL(dma_cache_sync); - --static struct dma_map_ops mips_default_dma_map_ops = { -- .alloc = mips_dma_alloc_coherent, -- .free = mips_dma_free_coherent, -- .map_page = mips_dma_map_page, -- .unmap_page = mips_dma_unmap_page, -- .map_sg = mips_dma_map_sg, -- .unmap_sg = mips_dma_unmap_sg, -- .sync_single_for_cpu = mips_dma_sync_single_for_cpu, -- .sync_single_for_device = mips_dma_sync_single_for_device, -- .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 --}; -- --struct dma_map_ops *mips_dma_map_ops = &mips_default_dma_map_ops; -+#ifdef CONFIG_SYS_HAS_DMA_OPS -+struct dma_map_ops *mips_dma_map_ops = NULL; - EXPORT_SYMBOL(mips_dma_map_ops); -+#endif - - #define PREALLOC_DMA_DEBUG_ENTRIES (1 << 16) - diff --git a/target/linux/generic/patches-3.10/132-mips_inline_dma_ops.patch b/target/linux/generic/patches-3.10/132-mips_inline_dma_ops.patch new file mode 100644 index 0000000000..b72bb620fd --- /dev/null +++ b/target/linux/generic/patches-3.10/132-mips_inline_dma_ops.patch @@ -0,0 +1,677 @@ +From 2c58080407554e1bac8fd50d23cb02420524caed Mon Sep 17 00:00:00 2001 +From: Felix Fietkau +Date: Mon, 12 Aug 2013 12:50:22 +0200 +Subject: [PATCH] MIPS: partially inline dma ops + +Several DMA ops are no-op on many platforms, and the indirection through +the mips_dma_map_ops function table is causing the compiler to emit +unnecessary code. + +Inlining visibly improves network performance in my tests (on a 24Kc +based system), and also slightly reduces code size of a few drivers. + +Signed-off-by: Felix Fietkau +--- + arch/mips/Kconfig | 4 + + arch/mips/include/asm/dma-mapping.h | 360 +++++++++++++++++++++++++++++++++++- + arch/mips/mm/dma-default.c | 163 ++-------------- + 3 files changed, 373 insertions(+), 154 deletions(-) + +--- a/arch/mips/Kconfig ++++ b/arch/mips/Kconfig +@@ -1430,6 +1430,7 @@ config CPU_CAVIUM_OCTEON + select LIBFDT + select USE_OF + select USB_EHCI_BIG_ENDIAN_MMIO ++ select SYS_HAS_DMA_OPS + help + The Cavium Octeon processor is a highly integrated chip containing + many ethernet hardware widgets for networking tasks. The processor +@@ -1650,6 +1651,9 @@ config SYS_HAS_CPU_XLR + config SYS_HAS_CPU_XLP + bool + ++config SYS_HAS_DMA_OPS ++ bool ++ + # + # CPU may reorder R->R, R->W, W->R, W->W + # Reordering beyond LL and SC is handled in WEAK_REORDERING_BEYOND_LLSC +--- a/arch/mips/include/asm/dma-mapping.h ++++ b/arch/mips/include/asm/dma-mapping.h +@@ -1,6 +1,12 @@ + #ifndef _ASM_DMA_MAPPING_H + #define _ASM_DMA_MAPPING_H + ++#include ++#include ++#include ++#include ++#include ++ + #include + #include + #include +@@ -10,14 +16,47 @@ + #include + #endif + +-extern struct dma_map_ops *mips_dma_map_ops; ++void __dma_sync(struct page *page, unsigned long offset, size_t size, ++ enum dma_data_direction direction); ++void *mips_dma_alloc_coherent(struct device *dev, size_t size, ++ dma_addr_t *dma_handle, gfp_t gfp, ++ struct dma_attrs *attrs); ++void mips_dma_free_coherent(struct device *dev, size_t size, void *vaddr, ++ dma_addr_t dma_handle, struct dma_attrs *attrs); + + static inline struct dma_map_ops *get_dma_ops(struct device *dev) + { ++#ifdef CONFIG_SYS_HAS_DMA_OPS + if (dev && dev->archdata.dma_ops) + return dev->archdata.dma_ops; + else + return mips_dma_map_ops; ++#else ++ return NULL; ++#endif ++} ++ ++/* ++ * Warning on the terminology - Linux calls an uncached area coherent; ++ * MIPS terminology calls memory areas with hardware maintained coherency ++ * coherent. ++ */ ++ ++static inline int cpu_is_noncoherent_r10000(struct device *dev) ++{ ++#ifndef CONFIG_SYS_HAS_CPU_R10000 ++ return 0; ++#endif ++ return !plat_device_is_coherent(dev) && ++ (current_cpu_type() == CPU_R10000 || ++ current_cpu_type() == CPU_R12000); ++} ++ ++static inline struct page *dma_addr_to_page(struct device *dev, ++ dma_addr_t dma_addr) ++{ ++ return pfn_to_page( ++ plat_dma_addr_to_phys(dev, dma_addr) >> PAGE_SHIFT); + } + + static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size) +@@ -30,12 +69,309 @@ static inline bool dma_capable(struct de + + static inline void dma_mark_clean(void *addr, size_t size) {} + +-#include ++static inline dma_addr_t dma_map_single_attrs(struct device *dev, void *ptr, ++ size_t size, ++ enum dma_data_direction dir, ++ struct dma_attrs *attrs) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ unsigned long offset = (unsigned long)ptr & ~PAGE_MASK; ++ struct page *page = virt_to_page(ptr); ++ dma_addr_t addr; ++ ++ kmemcheck_mark_initialized(ptr, size); ++ BUG_ON(!valid_dma_direction(dir)); ++ if (ops) { ++ addr = ops->map_page(dev, page, offset, size, dir, attrs); ++ } else { ++ if (!plat_device_is_coherent(dev)) ++ __dma_sync(page, offset, size, dir); ++ ++ addr = plat_map_dma_mem_page(dev, page) + offset; ++ } ++ debug_dma_map_page(dev, page, offset, size, dir, addr, true); ++ return addr; ++} ++ ++static inline void dma_unmap_single_attrs(struct device *dev, dma_addr_t addr, ++ size_t size, ++ enum dma_data_direction dir, ++ struct dma_attrs *attrs) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ ++ BUG_ON(!valid_dma_direction(dir)); ++ if (ops) { ++ ops->unmap_page(dev, addr, size, dir, attrs); ++ } else { ++ if (cpu_is_noncoherent_r10000(dev)) ++ __dma_sync(dma_addr_to_page(dev, addr), ++ addr & ~PAGE_MASK, size, dir); ++ ++ plat_unmap_dma_mem(dev, addr, size, dir); ++ } ++ debug_dma_unmap_page(dev, addr, size, dir, true); ++} ++ ++static inline int dma_map_sg_attrs(struct device *dev, struct scatterlist *sg, ++ int nents, enum dma_data_direction dir, ++ struct dma_attrs *attrs) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ int i, ents; ++ struct scatterlist *s; ++ ++ for_each_sg(sg, s, nents, i) ++ kmemcheck_mark_initialized(sg_virt(s), s->length); ++ BUG_ON(!valid_dma_direction(dir)); ++ if (ops) { ++ ents = ops->map_sg(dev, sg, nents, dir, attrs); ++ } else { ++ for_each_sg(sg, s, nents, i) { ++ struct page *page = sg_page(s); ++ ++ if (!plat_device_is_coherent(dev)) ++ __dma_sync(page, s->offset, s->length, dir); ++ s->dma_address = ++ plat_map_dma_mem_page(dev, page) + s->offset; ++ } ++ ents = nents; ++ } ++ debug_dma_map_sg(dev, sg, nents, ents, dir); ++ ++ return ents; ++} ++ ++static inline void dma_unmap_sg_attrs(struct device *dev, struct scatterlist *sg, ++ int nents, enum dma_data_direction dir, ++ struct dma_attrs *attrs) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ struct scatterlist *s; ++ int i; ++ ++ BUG_ON(!valid_dma_direction(dir)); ++ debug_dma_unmap_sg(dev, sg, nents, dir); ++ if (ops) { ++ ops->unmap_sg(dev, sg, nents, dir, attrs); ++ return; ++ } ++ ++ for_each_sg(sg, s, nents, i) { ++ if (!plat_device_is_coherent(dev) && dir != DMA_TO_DEVICE) ++ __dma_sync(sg_page(s), s->offset, s->length, dir); ++ plat_unmap_dma_mem(dev, s->dma_address, s->length, dir); ++ } ++} ++ ++static inline dma_addr_t dma_map_page(struct device *dev, struct page *page, ++ size_t offset, size_t size, ++ enum dma_data_direction dir) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ dma_addr_t addr; ++ ++ kmemcheck_mark_initialized(page_address(page) + offset, size); ++ BUG_ON(!valid_dma_direction(dir)); ++ if (ops) { ++ addr = ops->map_page(dev, page, offset, size, dir, NULL); ++ } else { ++ if (!plat_device_is_coherent(dev)) ++ __dma_sync(page, offset, size, dir); ++ ++ addr = plat_map_dma_mem_page(dev, page) + offset; ++ } ++ debug_dma_map_page(dev, page, offset, size, dir, addr, false); ++ ++ return addr; ++} ++ ++static inline void dma_unmap_page(struct device *dev, dma_addr_t addr, ++ size_t size, enum dma_data_direction dir) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ ++ BUG_ON(!valid_dma_direction(dir)); ++ if (ops) { ++ ops->unmap_page(dev, addr, size, dir, NULL); ++ } else { ++ if (cpu_is_noncoherent_r10000(dev)) ++ __dma_sync(dma_addr_to_page(dev, addr), ++ addr & ~PAGE_MASK, size, dir); ++ ++ plat_unmap_dma_mem(dev, addr, size, dir); ++ } ++ debug_dma_unmap_page(dev, addr, size, dir, false); ++} ++ ++static inline void dma_sync_single_for_cpu(struct device *dev, dma_addr_t addr, ++ size_t size, ++ enum dma_data_direction dir) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ ++ BUG_ON(!valid_dma_direction(dir)); ++ if (ops) ++ ops->sync_single_for_cpu(dev, addr, size, dir); ++ else if (cpu_is_noncoherent_r10000(dev)) ++ __dma_sync(dma_addr_to_page(dev, addr), ++ addr & ~PAGE_MASK, size, dir); ++ debug_dma_sync_single_for_cpu(dev, addr, size, dir); ++} ++ ++static inline void dma_sync_single_for_device(struct device *dev, ++ dma_addr_t addr, size_t size, ++ enum dma_data_direction dir) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ ++ BUG_ON(!valid_dma_direction(dir)); ++ if (ops) ++ ops->sync_single_for_device(dev, addr, size, dir); ++ else if (!plat_device_is_coherent(dev)) ++ __dma_sync(dma_addr_to_page(dev, addr), ++ addr & ~PAGE_MASK, size, dir); ++ debug_dma_sync_single_for_device(dev, addr, size, dir); ++} ++ ++static inline void dma_sync_single_range_for_cpu(struct device *dev, ++ dma_addr_t addr, ++ unsigned long offset, ++ 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) ++ ops->sync_single_for_cpu(dev, addr + offset, size, dir); ++ else if (cpu_is_noncoherent_r10000(dev)) ++ __dma_sync(dma_addr_to_page(dev, addr + offset), ++ (addr + offset) & ~PAGE_MASK, size, dir); ++ debug_dma_sync_single_range_for_cpu(dev, addr, offset, size, dir); ++} ++ ++static inline void dma_sync_single_range_for_device(struct device *dev, ++ dma_addr_t addr, ++ unsigned long offset, ++ 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) ++ ops->sync_single_for_device(dev, addr + offset, size, dir); ++ else if (!plat_device_is_coherent(dev)) ++ __dma_sync(dma_addr_to_page(dev, addr + offset), ++ (addr + offset) & ~PAGE_MASK, size, dir); ++ debug_dma_sync_single_range_for_device(dev, addr, offset, size, dir); ++} ++ ++static inline void ++dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sg, ++ int nelems, enum dma_data_direction dir) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ struct scatterlist *s; ++ int i; ++ ++ BUG_ON(!valid_dma_direction(dir)); ++ if (ops) ++ ops->sync_sg_for_cpu(dev, sg, nelems, dir); ++ else if (cpu_is_noncoherent_r10000(dev)) { ++ for_each_sg(sg, s, nelems, i) ++ __dma_sync(sg_page(s), s->offset, s->length, dir); ++ } ++ debug_dma_sync_sg_for_cpu(dev, sg, nelems, dir); ++} ++ ++static inline void ++dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg, ++ int nelems, enum dma_data_direction dir) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ struct scatterlist *s; ++ int i; ++ ++ BUG_ON(!valid_dma_direction(dir)); ++ if (ops) ++ ops->sync_sg_for_device(dev, sg, nelems, dir); ++ else if (!plat_device_is_coherent(dev)) { ++ for_each_sg(sg, s, nelems, i) ++ __dma_sync(sg_page(s), s->offset, s->length, dir); ++ } ++ debug_dma_sync_sg_for_device(dev, sg, nelems, dir); ++ ++} ++ ++#define dma_map_single(d, a, s, r) dma_map_single_attrs(d, a, s, r, NULL) ++#define dma_unmap_single(d, a, s, r) dma_unmap_single_attrs(d, a, s, r, NULL) ++#define dma_map_sg(d, s, n, r) dma_map_sg_attrs(d, s, n, r, NULL) ++#define dma_unmap_sg(d, s, n, r) dma_unmap_sg_attrs(d, s, n, r, NULL) ++ ++extern int dma_common_mmap(struct device *dev, struct vm_area_struct *vma, ++ void *cpu_addr, dma_addr_t dma_addr, size_t size); ++ ++/** ++ * dma_mmap_attrs - map a coherent DMA allocation into user space ++ * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices ++ * @vma: vm_area_struct describing requested user mapping ++ * @cpu_addr: kernel CPU-view address returned from dma_alloc_attrs ++ * @handle: device-view address returned from dma_alloc_attrs ++ * @size: size of memory originally requested in dma_alloc_attrs ++ * @attrs: attributes of mapping properties requested in dma_alloc_attrs ++ * ++ * Map a coherent DMA buffer previously allocated by dma_alloc_attrs ++ * into user space. The coherent DMA buffer must not be freed by the ++ * driver until the user space mapping has been released. ++ */ ++static inline int ++dma_mmap_attrs(struct device *dev, struct vm_area_struct *vma, void *cpu_addr, ++ dma_addr_t dma_addr, size_t size, struct dma_attrs *attrs) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ BUG_ON(!ops); ++ if (ops && ops->mmap) ++ return ops->mmap(dev, vma, cpu_addr, dma_addr, size, attrs); ++ return dma_common_mmap(dev, vma, cpu_addr, dma_addr, size); ++} ++ ++#define dma_mmap_coherent(d, v, c, h, s) dma_mmap_attrs(d, v, c, h, s, NULL) ++ ++static inline int dma_mmap_writecombine(struct device *dev, struct vm_area_struct *vma, ++ void *cpu_addr, dma_addr_t dma_addr, size_t size) ++{ ++ DEFINE_DMA_ATTRS(attrs); ++ dma_set_attr(DMA_ATTR_WRITE_COMBINE, &attrs); ++ return dma_mmap_attrs(dev, vma, cpu_addr, dma_addr, size, &attrs); ++} ++ ++int ++dma_common_get_sgtable(struct device *dev, struct sg_table *sgt, ++ void *cpu_addr, dma_addr_t dma_addr, size_t size); ++ ++static inline int ++dma_get_sgtable_attrs(struct device *dev, struct sg_table *sgt, void *cpu_addr, ++ dma_addr_t dma_addr, size_t size, struct dma_attrs *attrs) ++{ ++ struct dma_map_ops *ops = get_dma_ops(dev); ++ BUG_ON(!ops); ++ if (ops && ops->get_sgtable) ++ return ops->get_sgtable(dev, sgt, cpu_addr, dma_addr, size, ++ attrs); ++ return dma_common_get_sgtable(dev, sgt, cpu_addr, dma_addr, size); ++} ++ ++#define dma_get_sgtable(d, t, v, h, s) dma_get_sgtable_attrs(d, t, v, h, s, NULL) ++ + + static inline int dma_supported(struct device *dev, u64 mask) + { + struct dma_map_ops *ops = get_dma_ops(dev); +- return ops->dma_supported(dev, mask); ++ if (ops) ++ return ops->dma_supported(dev, mask); ++ return plat_dma_supported(dev, mask); + } + + static inline int dma_mapping_error(struct device *dev, u64 mask) +@@ -43,7 +379,9 @@ static inline int dma_mapping_error(stru + struct dma_map_ops *ops = get_dma_ops(dev); + + debug_dma_mapping_error(dev, mask); +- return ops->mapping_error(dev, mask); ++ if (ops) ++ return ops->mapping_error(dev, mask); ++ return 0; + } + + static inline int +@@ -69,7 +407,11 @@ static inline void *dma_alloc_attrs(stru + void *ret; + struct dma_map_ops *ops = get_dma_ops(dev); + +- ret = ops->alloc(dev, size, dma_handle, gfp, attrs); ++ if (ops) ++ ret = ops->alloc(dev, size, dma_handle, gfp, attrs); ++ else ++ ret = mips_dma_alloc_coherent(dev, size, dma_handle, gfp, ++ attrs); + + debug_dma_alloc_coherent(dev, size, *dma_handle, ret); + +@@ -84,7 +426,10 @@ static inline void dma_free_attrs(struct + { + struct dma_map_ops *ops = get_dma_ops(dev); + +- ops->free(dev, size, vaddr, dma_handle, attrs); ++ if (ops) ++ ops->free(dev, size, vaddr, dma_handle, attrs); ++ else ++ mips_dma_free_coherent(dev, size, vaddr, dma_handle, attrs); + + debug_dma_free_coherent(dev, size, vaddr, dma_handle); + } +--- a/arch/mips/mm/dma-default.c ++++ b/arch/mips/mm/dma-default.c +@@ -24,7 +24,7 @@ + + #ifdef CONFIG_DMA_MAYBE_COHERENT + int coherentio = 0; /* User defined DMA coherency from command line. */ +-EXPORT_SYMBOL_GPL(coherentio); ++EXPORT_SYMBOL(coherentio); + int hw_coherentio = 0; /* Actual hardware supported DMA coherency setting. */ + + static int __init setcoherentio(char *str) +@@ -44,26 +44,6 @@ static int __init setnocoherentio(char * + early_param("nocoherentio", setnocoherentio); + #endif + +-static inline struct page *dma_addr_to_page(struct device *dev, +- dma_addr_t dma_addr) +-{ +- return pfn_to_page( +- plat_dma_addr_to_phys(dev, dma_addr) >> PAGE_SHIFT); +-} +- +-/* +- * Warning on the terminology - Linux calls an uncached area coherent; +- * MIPS terminology calls memory areas with hardware maintained coherency +- * coherent. +- */ +- +-static inline int cpu_is_noncoherent_r10000(struct device *dev) +-{ +- return !plat_device_is_coherent(dev) && +- (current_cpu_type() == CPU_R10000 || +- current_cpu_type() == CPU_R12000); +-} +- + static gfp_t massage_gfp_flags(const struct device *dev, gfp_t gfp) + { + gfp_t dma_flag; +@@ -119,8 +99,9 @@ void *dma_alloc_noncoherent(struct devic + } + EXPORT_SYMBOL(dma_alloc_noncoherent); + +-static void *mips_dma_alloc_coherent(struct device *dev, size_t size, +- dma_addr_t * dma_handle, gfp_t gfp, struct dma_attrs *attrs) ++void *mips_dma_alloc_coherent(struct device *dev, size_t size, ++ dma_addr_t *dma_handle, gfp_t gfp, ++ struct dma_attrs *attrs) + { + void *ret; + +@@ -144,6 +125,7 @@ static void *mips_dma_alloc_coherent(str + + return ret; + } ++EXPORT_SYMBOL(mips_dma_alloc_coherent); + + + void dma_free_noncoherent(struct device *dev, size_t size, void *vaddr, +@@ -154,8 +136,8 @@ void dma_free_noncoherent(struct device + } + EXPORT_SYMBOL(dma_free_noncoherent); + +-static void mips_dma_free_coherent(struct device *dev, size_t size, void *vaddr, +- dma_addr_t dma_handle, struct dma_attrs *attrs) ++void mips_dma_free_coherent(struct device *dev, size_t size, void *vaddr, ++ dma_addr_t dma_handle, struct dma_attrs *attrs) + { + unsigned long addr = (unsigned long) vaddr; + int order = get_order(size); +@@ -170,6 +152,7 @@ static void mips_dma_free_coherent(struc + + free_pages(addr, get_order(size)); + } ++EXPORT_SYMBOL(mips_dma_free_coherent); + + static inline void __dma_sync_virtual(void *addr, size_t size, + enum dma_data_direction direction) +@@ -198,8 +181,8 @@ static inline void __dma_sync_virtual(vo + * If highmem is not configured then the bulk of this loop gets + * optimized out. + */ +-static inline void __dma_sync(struct page *page, +- unsigned long offset, size_t size, enum dma_data_direction direction) ++void __dma_sync(struct page *page, unsigned long offset, size_t size, ++ enum dma_data_direction direction) + { + size_t left = size; + +@@ -228,109 +211,7 @@ static inline void __dma_sync(struct pag + left -= len; + } while (left); + } +- +-static void mips_dma_unmap_page(struct device *dev, dma_addr_t dma_addr, +- size_t size, enum dma_data_direction direction, struct dma_attrs *attrs) +-{ +- if (cpu_is_noncoherent_r10000(dev)) +- __dma_sync(dma_addr_to_page(dev, dma_addr), +- dma_addr & ~PAGE_MASK, size, direction); +- +- plat_unmap_dma_mem(dev, dma_addr, size, direction); +-} +- +-static int mips_dma_map_sg(struct device *dev, struct scatterlist *sg, +- int nents, enum dma_data_direction direction, struct dma_attrs *attrs) +-{ +- int i; +- +- for (i = 0; i < nents; i++, sg++) { +- if (!plat_device_is_coherent(dev)) +- __dma_sync(sg_page(sg), sg->offset, sg->length, +- direction); +- sg->dma_address = plat_map_dma_mem_page(dev, sg_page(sg)) + +- sg->offset; +- } +- +- return nents; +-} +- +-static dma_addr_t mips_dma_map_page(struct device *dev, struct page *page, +- unsigned long offset, size_t size, enum dma_data_direction direction, +- struct dma_attrs *attrs) +-{ +- if (!plat_device_is_coherent(dev)) +- __dma_sync(page, offset, size, direction); +- +- return plat_map_dma_mem_page(dev, page) + offset; +-} +- +-static void mips_dma_unmap_sg(struct device *dev, struct scatterlist *sg, +- int nhwentries, enum dma_data_direction direction, +- struct dma_attrs *attrs) +-{ +- int i; +- +- for (i = 0; i < nhwentries; i++, sg++) { +- if (!plat_device_is_coherent(dev) && +- direction != DMA_TO_DEVICE) +- __dma_sync(sg_page(sg), sg->offset, sg->length, +- direction); +- plat_unmap_dma_mem(dev, sg->dma_address, sg->length, direction); +- } +-} +- +-static void mips_dma_sync_single_for_cpu(struct device *dev, +- dma_addr_t dma_handle, size_t size, enum dma_data_direction direction) +-{ +- if (cpu_is_noncoherent_r10000(dev)) +- __dma_sync(dma_addr_to_page(dev, dma_handle), +- dma_handle & ~PAGE_MASK, size, direction); +-} +- +-static void mips_dma_sync_single_for_device(struct device *dev, +- dma_addr_t dma_handle, size_t size, enum dma_data_direction direction) +-{ +- if (!plat_device_is_coherent(dev)) +- __dma_sync(dma_addr_to_page(dev, dma_handle), +- dma_handle & ~PAGE_MASK, size, direction); +-} +- +-static void mips_dma_sync_sg_for_cpu(struct device *dev, +- struct scatterlist *sg, int nelems, enum dma_data_direction direction) +-{ +- int i; +- +- /* Make sure that gcc doesn't leave the empty loop body. */ +- for (i = 0; i < nelems; i++, sg++) { +- if (cpu_is_noncoherent_r10000(dev)) +- __dma_sync(sg_page(sg), sg->offset, sg->length, +- direction); +- } +-} +- +-static void mips_dma_sync_sg_for_device(struct device *dev, +- struct scatterlist *sg, int nelems, enum dma_data_direction direction) +-{ +- int i; +- +- /* Make sure that gcc doesn't leave the empty loop body. */ +- for (i = 0; i < nelems; i++, sg++) { +- if (!plat_device_is_coherent(dev)) +- __dma_sync(sg_page(sg), sg->offset, sg->length, +- direction); +- } +-} +- +-int mips_dma_mapping_error(struct device *dev, dma_addr_t dma_addr) +-{ +- return 0; +-} +- +-int mips_dma_supported(struct device *dev, u64 mask) +-{ +- return plat_dma_supported(dev, mask); +-} ++EXPORT_SYMBOL(__dma_sync); + + void dma_cache_sync(struct device *dev, void *vaddr, size_t size, + enum dma_data_direction direction) +@@ -343,23 +224,10 @@ void dma_cache_sync(struct device *dev, + + EXPORT_SYMBOL(dma_cache_sync); + +-static struct dma_map_ops mips_default_dma_map_ops = { +- .alloc = mips_dma_alloc_coherent, +- .free = mips_dma_free_coherent, +- .map_page = mips_dma_map_page, +- .unmap_page = mips_dma_unmap_page, +- .map_sg = mips_dma_map_sg, +- .unmap_sg = mips_dma_unmap_sg, +- .sync_single_for_cpu = mips_dma_sync_single_for_cpu, +- .sync_single_for_device = mips_dma_sync_single_for_device, +- .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 +-}; +- +-struct dma_map_ops *mips_dma_map_ops = &mips_default_dma_map_ops; ++#ifdef CONFIG_SYS_HAS_DMA_OPS ++struct dma_map_ops *mips_dma_map_ops = NULL; + EXPORT_SYMBOL(mips_dma_map_ops); ++#endif + + #define PREALLOC_DMA_DEBUG_ENTRIES (1 << 16) +