From: Christoph Hellwig hch@lst.de

commit 7ec916f82c48dcfc115eee2e3e0e6d400e310fc5 upstream.

This commit may cause a less than required dma mask to be used for some allocations, which apparently leads to module load failures for iwlwifi sometimes.

This reverts commit d657c5c73ca987214a6f9436e435b34fc60f332a.

Signed-off-by: Christoph Hellwig hch@lst.de Reported-by: Fabio Coatti fabio.coatti@gmail.com Tested-by: Fabio Coatti fabio.coatti@gmail.com --- Backporting this and submitting this to stable@, because without it, ordinary WiFi is broken on a fairly vanilla Thinkpad P50, on all 4.17 kernels.

drivers/iommu/Kconfig | 1 - drivers/iommu/intel-iommu.c | 62 +++++++++++++++++++++++++++---------- 2 files changed, 46 insertions(+), 17 deletions(-)

diff --git a/drivers/iommu/Kconfig b/drivers/iommu/Kconfig index b38798cc5288..f3a21343e636 100644 --- a/drivers/iommu/Kconfig +++ b/drivers/iommu/Kconfig @@ -142,7 +142,6 @@ config DMAR_TABLE config INTEL_IOMMU bool "Support for Intel IOMMU using DMA Remapping Devices" depends on PCI_MSI && ACPI && (X86 || IA64_GENERIC) - select DMA_DIRECT_OPS select IOMMU_API select IOMMU_IOVA select DMAR_TABLE diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c index 749d8f235346..6392a4964fc5 100644 --- a/drivers/iommu/intel-iommu.c +++ b/drivers/iommu/intel-iommu.c @@ -31,7 +31,6 @@ #include <linux/pci.h> #include <linux/dmar.h> #include <linux/dma-mapping.h> -#include <linux/dma-direct.h> #include <linux/mempool.h> #include <linux/memory.h> #include <linux/cpu.h> @@ -3709,30 +3708,61 @@ static void *intel_alloc_coherent(struct device *dev, size_t size, dma_addr_t *dma_handle, gfp_t flags, unsigned long attrs) { - void *vaddr; + struct page *page = NULL; + int order;

- vaddr = dma_direct_alloc(dev, size, dma_handle, flags, attrs); - if (iommu_no_mapping(dev) || !vaddr) - return vaddr; + size = PAGE_ALIGN(size); + order = get_order(size);

- *dma_handle = __intel_map_single(dev, virt_to_phys(vaddr), - PAGE_ALIGN(size), DMA_BIDIRECTIONAL, - dev->coherent_dma_mask); - if (!*dma_handle) - goto out_free_pages; - return vaddr; + if (!iommu_no_mapping(dev)) + flags &= ~(GFP_DMA | GFP_DMA32); + else if (dev->coherent_dma_mask < dma_get_required_mask(dev)) { + if (dev->coherent_dma_mask < DMA_BIT_MASK(32)) + flags |= GFP_DMA; + else + flags |= GFP_DMA32; + } + + if (gfpflags_allow_blocking(flags)) { + unsigned int count = size >> PAGE_SHIFT; + + page = dma_alloc_from_contiguous(dev, count, order, flags); + if (page && iommu_no_mapping(dev) && + page_to_phys(page) + size > dev->coherent_dma_mask) { + dma_release_from_contiguous(dev, page, count); + page = NULL; + } + } + + if (!page) + page = alloc_pages(flags, order); + if (!page) + return NULL; + memset(page_address(page), 0, size); + + *dma_handle = __intel_map_single(dev, page_to_phys(page), size, + DMA_BIDIRECTIONAL, + dev->coherent_dma_mask); + if (*dma_handle) + return page_address(page); + if (!dma_release_from_contiguous(dev, page, size >> PAGE_SHIFT)) + __free_pages(page, order);

-out_free_pages: - dma_direct_free(dev, size, vaddr, *dma_handle, attrs); return NULL; }

static void intel_free_coherent(struct device *dev, size_t size, void *vaddr, dma_addr_t dma_handle, unsigned long attrs) { - if (!iommu_no_mapping(dev)) - intel_unmap(dev, dma_handle, PAGE_ALIGN(size)); - dma_direct_free(dev, size, vaddr, dma_handle, attrs); + int order; + struct page *page = virt_to_page(vaddr); + + size = PAGE_ALIGN(size); + order = get_order(size); + + intel_unmap(dev, dma_handle, size); + if (!dma_release_from_contiguous(dev, page, size >> PAGE_SHIFT)) + __free_pages(page, order); }

static void intel_unmap_sg(struct device *dev, struct scatterlist *sglist,