vmd: Stop overriding dma_map_ops
Subsystems (5)
PATCH 1/1 vmd: Stop overriding dma_map_ops
With a little tweak to the intel-iommu code we should be able to work around the VMD mess for the requester IDs without having to create giant amounts of boilerplate DMA ops wrapping code. The other advantage of this scheme is that we can respect the real DMA masks for the actual devices, and I bet it will only be a matter of time until we'll see the first DMA challeneged NVMe devices.
The only downside is that we can't offer vmd as a module given that intel-iommu calls into it. But the driver only has about 700 lines of code, so this should not be a major issue.
This also removes the leftover bits of the X86_DEV_DMA_OPS dma_map_ops registry.
--- arch/x86/Kconfig | 3 - arch/x86/include/asm/device.h | 10 --- arch/x86/include/asm/pci.h | 2 + arch/x86/pci/common.c | 38 --------- drivers/iommu/intel-iommu.c | 33 +++++--- drivers/pci/controller/Kconfig | 6 +- drivers/pci/controller/vmd.c | 139 +-------------------------------- 7 files changed, 27 insertions(+), 204 deletions(-)
arch/x86/Kconfig +0 -3
@@ -2905,9 +2905,6 @@ config HAVE_ATOMIC_IOMAP def_bool y depends on X86_32 -config X86_DEV_DMA_OPS - bool - source "drivers/firmware/Kconfig" source "arch/x86/kvm/Kconfig"
arch/x86/include/asm/device.h +0 -10
@@ -11,16 +11,6 @@ struct dev_archdata {#endif }; -#if defined(CONFIG_X86_DEV_DMA_OPS ) && defined(CONFIG_PCI_DOMAINS ) -struct dma_domain { -struct list_head node; -const struct dma_map_ops *dma_ops; -int domain_nr; -}; -void add_dma_domain (struct dma_domain *domain); -void del_dma_domain (struct dma_domain *domain); -#endif -struct pdev_archdata { };
arch/x86/include/asm/pci.h +2 -0
@@ -73,6 +73,8 @@ static inline bool is_vmd(struct pci_bus *bus)#endif } +struct device *to_vmd_dev (struct device *dev); +/* Can be used to override the logic in pci_scan_bus for skipping already-configured bus numbers - to be used for buggy BIOSes or architectures with incomplete PCI setup by the loader */
arch/x86/pci/common.c +0 -38
@@ -625,43 +625,6 @@ unsigned int pcibios_assign_all_busses(void)return (pci_probe &PCI_ASSIGN_ALL_BUSSES ) ?1 :0 ; } -#if defined(CONFIG_X86_DEV_DMA_OPS ) && defined(CONFIG_PCI_DOMAINS ) -static LIST_HEAD (dma_domain_list); -static DEFINE_SPINLOCK (dma_domain_list_lock); - -void add_dma_domain (struct dma_domain *domain) -{ -spin_lock (&dma_domain_list_lock); -list_add (&domain->node, &dma_domain_list); -spin_unlock (&dma_domain_list_lock); -} -EXPORT_SYMBOL_GPL (add_dma_domain); - -void del_dma_domain (struct dma_domain *domain) -{ -spin_lock (&dma_domain_list_lock); -list_del (&domain->node); -spin_unlock (&dma_domain_list_lock); -} -EXPORT_SYMBOL_GPL (del_dma_domain); - -static void set_dma_domain_ops (struct pci_dev *pdev) -{ -struct dma_domain *domain; - -spin_lock (&dma_domain_list_lock); -list_for_each_entry (domain, &dma_domain_list, node) { -if (pci_domain_nr (pdev->bus) == domain->domain_nr) { - pdev->dev.dma_ops = domain->dma_ops; -break ; - } - } -spin_unlock (&dma_domain_list_lock); -} -#else -static void set_dma_domain_ops (struct pci_dev *pdev) {} -#endif -static void set_dev_domain_options (struct pci_dev *pdev) {if (is_vmd (pdev->bus))
@@ -697,7 +660,6 @@ int pcibios_add_device(struct pci_dev *dev) pa_data = data->next;memunmap (data); } -set_dma_domain_ops (dev);set_dev_domain_options (dev);return 0 ; }
drivers/iommu/intel-iommu.c +23 -10
@@ -3416,9 +3416,14 @@ static struct dmar_domain *get_private_domain_for_dev(struct device *dev)return domain; } -/* Check if the dev needs to go through non-identity map and unmap process.*/ -static bool iommu_need_mapping (struct device *dev) +/* +* Check if the dev needs to go through non-identity map and unmap process. +* Also where needed replace dev with the actual device used for the mapping +* process. +*/ +static bool iommu_need_mapping (struct device **devp) { +struct device *dev = *devp;int ret;if (iommu_dummy (dev))
@@ -3456,6 +3461,14 @@ static bool iommu_need_mapping(struct device *dev)dev_info (dev,"32bit DMA uses non-identity mapping\n" ); } +/* +* For VMD we need to use the VMD devices for mapping requests instead +* of the actual device to get the proper PCIe requester ID. +*/ +#ifdef CONFIG_VMD +if (dev_is_pci (dev) &&is_vmd (to_pci_dev (dev)->bus)) + *devp =to_vmd_dev (dev); +#endif return true; }
@@ -3520,7 +3533,7 @@ static dma_addr_t intel_map_page(struct device *dev, struct page *page, enum dma_data_direction dir, unsigned long attrs) { - if (iommu_need_mapping(dev)) + if (iommu_need_mapping(&dev)) return __intel_map_single(dev, page_to_phys(page) + offset, size, dir, *dev->dma_mask);return dma_direct_map_page (dev, page, offset, size, dir, attrs);
@@ -3530,7 +3543,7 @@ static dma_addr_t intel_map_resource(struct device *dev, phys_addr_t phys_addr, size_t size, enum dma_data_direction dir, unsigned long attrs) { - if (iommu_need_mapping(dev)) + if (iommu_need_mapping(&dev)) return __intel_map_single(dev, phys_addr, size, dir, *dev->dma_mask);return dma_direct_map_resource (dev, phys_addr, size, dir, attrs);
@@ -3585,7 +3598,7 @@ static void intel_unmap_page(struct device *dev, dma_addr_t dev_addr, size_t size, enum dma_data_direction dir, unsigned long attrs) { - if (iommu_need_mapping(dev)) + if (iommu_need_mapping(&dev)) intel_unmap(dev, dev_addr, size); elsedma_direct_unmap_page (dev ,dev_addr ,size ,dir ,attrs );
@@ -3594,7 +3607,7 @@ static void intel_unmap_page(struct device *dev, dma_addr_t dev_addr,static void intel_unmap_resource (struct device *dev,dma_addr_t dev_addr,size_t size,enum dma_data_direction dir,unsigned long attrs) { -if (iommu_need_mapping (dev)) +if (iommu_need_mapping (&dev))intel_unmap (dev, dev_addr, size); }
@@ -3605,7 +3618,7 @@ static void *intel_alloc_coherent(struct device *dev, size_t size,struct page *page =NULL ;int order; -if (!iommu_need_mapping (dev)) +if (!iommu_need_mapping (&dev))return dma_direct_alloc (dev, size, dma_handle, flags, attrs); size =PAGE_ALIGN (size);
@@ -3641,7 +3654,7 @@ static void intel_free_coherent(struct device *dev, size_t size, void *vaddr,int order;struct page *page =virt_to_page (vaddr); -if (!iommu_need_mapping (dev)) +if (!iommu_need_mapping (&dev))return dma_direct_free (dev, size, vaddr, dma_handle, attrs); size =PAGE_ALIGN (size);
@@ -3661,7 +3674,7 @@ static void intel_unmap_sg(struct device *dev, struct scatterlist *sglist,struct scatterlist *sg;int i; -if (!iommu_need_mapping (dev)) +if (!iommu_need_mapping (&dev))return dma_direct_unmap_sg (dev, sglist, nelems, dir, attrs);for_each_sg (sglist, sg, nelems, i) {
@@ -3685,7 +3698,7 @@ static int intel_map_sg(struct device *dev, struct scatterlist *sglist, int nelestruct intel_iommu *iommu;BUG_ON (dir ==DMA_NONE ); -if (!iommu_need_mapping (dev)) +if (!iommu_need_mapping (&dev))return dma_direct_map_sg (dev, sglist, nelems, dir, attrs); domain =find_domain (dev);
drivers/pci/controller/Kconfig +1 -5
@@ -267,8 +267,7 @@ config PCIE_TANGO_SMP8759 config VMD depends on PCI_MSI && X86_64 && SRCU - select X86_DEV_DMA_OPS - tristate "Intel Volume Management Device Driver" + bool "Intel Volume Management Device Driver" ---help--- Adds support for the Intel Volume Management Device (VMD). VMD is a secondary PCI host bridge that allows PCI Express root ports,
@@ -278,8 +277,5 @@ config VMD single domain. If you know your system provides one of these and has devices attached to it, say Y; if you are not sure, say N. - To compile this driver as a module, choose M here: the - module will be called vmd. - source "drivers/pci/controller/dwc/Kconfig" endmenu
drivers/pci/controller/vmd.c +1 -138
169 lines not on this page · raw message
Signature · 1 line
2.20.1
Quoted 10 lines · Christoph Hellwig
With a little tweak to the intel-iommu code we should be able to work around the VMD mess for the requester IDs without having to create giant amounts of boilerplate DMA ops wrapping code. The other advantage of this scheme is that we can respect the real DMA masks for the actual devices, and I bet it will only be a matter of time until we'll see the first DMA challeneged NVMe devices.
The only downside is that we can't offer vmd as a module given that intel-iommu calls into it. But the driver only has about 700 lines of code, so this should not be a major issue.
If we're going to remove its ability to be a module, and given its small size, could we make this default =y?
Otherwise we risk breaking platforms which have it enabled with OSVs who miss enabling it
Quoted 6 lines
This also removes the leftover bits of the
X86_DEV_DMA_OPSdma_map_opsregistry.Signed-off-by Christoph Hellwig <[email protected]>
lgtm otherwise
Quoted 8 lines · Jonathan
Quoted 7 lines · Christoph Hellwig
With a little tweak to the intel-iommu code we should be able to work around the VMD mess for the requester IDs without having to create giant amounts of boilerplate DMA ops wrapping code. The other advantage of this scheme is that we can respect the real DMA masks for the actual devices, and I bet it will only be a matter of time until we'll see the first DMA challeneged NVMe devices.
Quoted 8 lines
Quoted 3 lines
The only downside is that we can't offer vmd as a module given that intel-iommu calls into it. But the driver only has about 700 lines of code, so this should not be a major issue.
If we're going to remove its ability to be a module, and given its small size, could we make this default =y?
Otherwise we risk breaking platforms which have it enabled with OSVs who miss enabling it
Can we keep this as a module if we stick the remapping struct device in pci_sysdata instead of going through the vmd driver to get it?
Otherwise, very happy to see this dma wrapping go away.