From: FUJITA Tomonori This patch converts gart IOMMU to use the IOMMU helper functions. The IOMMU doesn't allocate a memory area spanning LLD's segment boundary anymore. Signed-off-by: FUJITA Tomonori Cc: Jeff Garzik Cc: James Bottomley Cc: Jens Axboe Cc: Thomas Gleixner Cc: Ingo Molnar Cc: Muli Ben-Yehuda Cc: Andi Kleen Signed-off-by: Andrew Morton --- arch/x86/Kconfig | 2 - arch/x86/kernel/pci-gart_64.c | 41 +++++++++++++++++++------------- 2 files changed, 26 insertions(+), 17 deletions(-) diff -puN arch/x86/Kconfig~iommu-sg-x86-convert-gart-iommu-to-use-the-iommu-helper arch/x86/Kconfig --- a/arch/x86/Kconfig~iommu-sg-x86-convert-gart-iommu-to-use-the-iommu-helper +++ a/arch/x86/Kconfig @@ -467,7 +467,7 @@ config CALGARY_IOMMU_ENABLED_BY_DEFAULT If unsure, say Y. config IOMMU_HELPER - def_bool CALGARY_IOMMU + def_bool (CALGARY_IOMMU || GART_IOMMU) # need this always selected by IOMMU for the VIA workaround config SWIOTLB diff -puN arch/x86/kernel/pci-gart_64.c~iommu-sg-x86-convert-gart-iommu-to-use-the-iommu-helper arch/x86/kernel/pci-gart_64.c --- a/arch/x86/kernel/pci-gart_64.c~iommu-sg-x86-convert-gart-iommu-to-use-the-iommu-helper +++ a/arch/x86/kernel/pci-gart_64.c @@ -25,6 +25,7 @@ #include #include #include +#include #include #include #include @@ -82,17 +83,24 @@ AGPEXTERN __u32 *agp_gatt_table; static unsigned long next_bit; /* protected by iommu_bitmap_lock */ static int need_flush; /* global flush state. set for each gart wrap */ -static unsigned long alloc_iommu(int size) +static unsigned long alloc_iommu(struct device *dev, int size) { unsigned long offset, flags; + unsigned long boundary_size; + unsigned long base_index; + + base_index = ALIGN(iommu_bus_base & dma_get_seg_boundary(dev), + PAGE_SIZE) >> PAGE_SHIFT; + boundary_size = ALIGN(dma_get_seg_boundary(dev) + 1, + PAGE_SIZE) >> PAGE_SHIFT; spin_lock_irqsave(&iommu_bitmap_lock, flags); - offset = find_next_zero_string(iommu_gart_bitmap, next_bit, - iommu_pages, size); + offset = iommu_area_alloc(iommu_gart_bitmap, iommu_pages, next_bit, + size, base_index, boundary_size, 0); if (offset == -1) { need_flush = 1; - offset = find_next_zero_string(iommu_gart_bitmap, 0, - iommu_pages, size); + offset = iommu_area_alloc(iommu_gart_bitmap, iommu_pages, 0, + size, base_index, boundary_size, 0); } if (offset != -1) { set_bit_string(iommu_gart_bitmap, offset, size); @@ -114,7 +122,7 @@ static void free_iommu(unsigned long off unsigned long flags; spin_lock_irqsave(&iommu_bitmap_lock, flags); - __clear_bit_string(iommu_gart_bitmap, offset, size); + iommu_area_free(iommu_gart_bitmap, offset, size); spin_unlock_irqrestore(&iommu_bitmap_lock, flags); } @@ -235,7 +243,7 @@ static dma_addr_t dma_map_area(struct de size_t size, int dir) { unsigned long npages = to_pages(phys_mem, size); - unsigned long iommu_page = alloc_iommu(npages); + unsigned long iommu_page = alloc_iommu(dev, npages); int i; if (iommu_page == -1) { @@ -355,10 +363,11 @@ static int dma_map_sg_nonforce(struct de } /* Map multiple scatterlist entries continuous into the first. */ -static int __dma_map_cont(struct scatterlist *start, int nelems, - struct scatterlist *sout, unsigned long pages) +static int __dma_map_cont(struct device *dev, struct scatterlist *start, + int nelems, struct scatterlist *sout, + unsigned long pages) { - unsigned long iommu_start = alloc_iommu(pages); + unsigned long iommu_start = alloc_iommu(dev, pages); unsigned long iommu_page = iommu_start; struct scatterlist *s; int i; @@ -394,8 +403,8 @@ static int __dma_map_cont(struct scatter } static inline int -dma_map_cont(struct scatterlist *start, int nelems, struct scatterlist *sout, - unsigned long pages, int need) +dma_map_cont(struct device *dev, struct scatterlist *start, int nelems, + struct scatterlist *sout, unsigned long pages, int need) { if (!need) { BUG_ON(nelems != 1); @@ -403,7 +412,7 @@ dma_map_cont(struct scatterlist *start, sout->dma_length = start->length; return 0; } - return __dma_map_cont(start, nelems, sout, pages); + return __dma_map_cont(dev, start, nelems, sout, pages); } /* @@ -449,8 +458,8 @@ gart_map_sg(struct device *dev, struct s if (!iommu_merge || !nextneed || !need || s->offset || (s->length + seg_size > max_seg_size) || (ps->offset + ps->length) % PAGE_SIZE) { - if (dma_map_cont(start_sg, i - start, sgmap, - pages, need) < 0) + if (dma_map_cont(dev, start_sg, i - start, + sgmap, pages, need) < 0) goto error; out++; seg_size = 0; @@ -466,7 +475,7 @@ gart_map_sg(struct device *dev, struct s pages += to_pages(s->offset, s->length); ps = s; } - if (dma_map_cont(start_sg, i - start, sgmap, pages, need) < 0) + if (dma_map_cont(dev, start_sg, i - start, sgmap, pages, need) < 0) goto error; out++; flush_gart(); _