dma,x86,xen: fix dma_get_required_mask() when dma != phys address
Subsystems (6)
cover dma,x86,xen: fix dma_get_required_mask() when dma != phys address
x86 uses the generic dma_get_required_mask() which assumes that DMA address == physical address. For Xen PV guests this is not the case and certain drivers will cause excessive use of the SWIOTLB.
Patch #1 is a straight forward refactoring of a common function.
Patch #2 addes the required arch-specific implementation of dma_get_required_mask() to x86.
David
PATCH 1/2 dma: add dma_get_required_mask_from_max_pfn()
ia64 provides a duplicate of the generic dma_get_required_mask() because it has ARCH_HAS_GET_REQUIRED_MASK. Provide a common dma_get_require_mask_max_pfn() instead.
--- arch/ia64/include/asm/machvec.h | 2 +- arch/ia64/include/asm/machvec_init.h | 1 - arch/ia64/pci/pci.c | 20 -------------------- drivers/base/platform.c | 10 ++++++++-- include/linux/dma-mapping.h | 1 + 5 files changed, 10 insertions(+), 24 deletions(-)
arch/ia64/include/asm/machvec.h +1 -1
@@ -287,7 +287,7 @@ extern struct dma_map_ops *dma_get_ops(struct device *);# define platform_dma_get_ops dma_get_ops#endif #ifndef platform_dma_get_required_mask -# define platform_dma_get_required_mask ia64_dma_get_required_mask +# define platform_dma_get_required_mask dma_get_required_mask_from_max_pfn#endif #ifndef platform_irq_to_vector# define platform_irq_to_vector __ia64_irq_to_vector
arch/ia64/include/asm/machvec_init.h +0 -1
@@ -3,7 +3,6 @@extern ia64_mv_send_ipi_t ia64_send_ipi;extern ia64_mv_global_tlb_purge_t ia64_global_tlb_purge; -extern ia64_mv_dma_get_required_mask ia64_dma_get_required_mask;extern ia64_mv_irq_to_vector __ia64_irq_to_vector;extern ia64_mv_local_vector_to_irq __ia64_local_vector_to_irq;extern ia64_mv_pci_get_legacy_mem_t ia64_pci_get_legacy_mem;
arch/ia64/pci/pci.c +0 -20
@@ -791,26 +791,6 @@ static void __init set_pci_dfl_cacheline_size(void) pci_dfl_cache_line_size = (1 << cci.pcci_line_size) /4 ; } -u64 ia64_dma_get_required_mask (struct device *dev) -{ -u32 low_totalram = ((max_pfn -1 ) <<PAGE_SHIFT ); -u32 high_totalram = ((max_pfn -1 ) >> (32 -PAGE_SHIFT )); -u64 mask; - -if (!high_totalram) { -/* convert to mask just covering totalram */ - low_totalram = (1 << (fls (low_totalram) -1 )); - low_totalram += low_totalram -1 ; - mask = low_totalram; - }else { - high_totalram = (1 << (fls (high_totalram) -1 )); - high_totalram += high_totalram -1 ; - mask = (((u64 )high_totalram) <<32 ) +0xffffffff ; - } -return mask; -} -EXPORT_SYMBOL_GPL (ia64_dma_get_required_mask); -u64 dma_get_required_mask (struct device *dev) {return platform_dma_get_required_mask (dev);
drivers/base/platform.c +8 -2
@@ -942,8 +942,7 @@ int __init platform_bus_init(void)return error; } -#ifndef ARCH_HAS_DMA_GET_REQUIRED_MASK -u64 dma_get_required_mask (struct device *dev) +u64 dma_get_required_mask_from_max_pfn (struct device *dev) {u32 low_totalram = ((max_pfn -1 ) <<PAGE_SHIFT );u32 high_totalram = ((max_pfn -1 ) >> (32 -PAGE_SHIFT ));
@@ -961,6 +960,13 @@ u64 dma_get_required_mask(struct device *dev) }return mask; } +EXPORT_SYMBOL_GPL (dma_get_required_mask_from_max_pfn); + +#ifndef ARCH_HAS_DMA_GET_REQUIRED_MASK +u64 dma_get_required_mask (struct device *dev) +{ +return dma_get_required_mask_from_max_pfn (dev); +}EXPORT_SYMBOL_GPL (dma_get_required_mask);#endif
include/linux/dma-mapping.h +1 -0
@@ -121,6 +121,7 @@ static inline int dma_coerce_mask_and_coherent(struct device *dev, u64 mask)return dma_set_mask_and_coherent (dev, mask); } +extern u64 dma_get_required_mask_from_max_pfn (struct device *dev);extern u64 dma_get_required_mask (struct device *dev);static inline unsigned int dma_get_max_seg_size (struct device *dev)
Signature · 1 line
1.7.10.4
PATCH 1/2 dma: add dma_get_required_mask_from_max_pfn()
In reply to David Vrabelia64 provides a duplicate of the generic dma_get_required_mask() because it has ARCH_HAS_GET_REQUIRED_MASK. Provide a common dma_get_require_mask_max_pfn() instead.
--- arch/ia64/include/asm/machvec.h | 2 +- arch/ia64/include/asm/machvec_init.h | 1 - arch/ia64/pci/pci.c | 20 -------------------- drivers/base/platform.c | 10 ++++++++-- include/linux/dma-mapping.h | 1 + 5 files changed, 10 insertions(+), 24 deletions(-)
arch/ia64/include/asm/machvec.h +1 -1
@@ -287,7 +287,7 @@ extern struct dma_map_ops *dma_get_ops(struct device *);# define platform_dma_get_ops dma_get_ops#endif #ifndef platform_dma_get_required_mask -# define platform_dma_get_required_mask ia64_dma_get_required_mask +# define platform_dma_get_required_mask dma_get_required_mask_from_max_pfn#endif #ifndef platform_irq_to_vector# define platform_irq_to_vector __ia64_irq_to_vector
arch/ia64/include/asm/machvec_init.h +0 -1
@@ -3,7 +3,6 @@extern ia64_mv_send_ipi_t ia64_send_ipi;extern ia64_mv_global_tlb_purge_t ia64_global_tlb_purge; -extern ia64_mv_dma_get_required_mask ia64_dma_get_required_mask;extern ia64_mv_irq_to_vector __ia64_irq_to_vector;extern ia64_mv_local_vector_to_irq __ia64_local_vector_to_irq;extern ia64_mv_pci_get_legacy_mem_t ia64_pci_get_legacy_mem;
arch/ia64/pci/pci.c +0 -20
@@ -791,26 +791,6 @@ static void __init set_pci_dfl_cacheline_size(void) pci_dfl_cache_line_size = (1 << cci.pcci_line_size) /4 ; } -u64 ia64_dma_get_required_mask (struct device *dev) -{ -u32 low_totalram = ((max_pfn -1 ) <<PAGE_SHIFT ); -u32 high_totalram = ((max_pfn -1 ) >> (32 -PAGE_SHIFT )); -u64 mask; - -if (!high_totalram) { -/* convert to mask just covering totalram */ - low_totalram = (1 << (fls (low_totalram) -1 )); - low_totalram += low_totalram -1 ; - mask = low_totalram; - }else { - high_totalram = (1 << (fls (high_totalram) -1 )); - high_totalram += high_totalram -1 ; - mask = (((u64 )high_totalram) <<32 ) +0xffffffff ; - } -return mask; -} -EXPORT_SYMBOL_GPL (ia64_dma_get_required_mask); -u64 dma_get_required_mask (struct device *dev) {return platform_dma_get_required_mask (dev);
drivers/base/platform.c +8 -2
@@ -942,8 +942,7 @@ int __init platform_bus_init(void)return error; } -#ifndef ARCH_HAS_DMA_GET_REQUIRED_MASK -u64 dma_get_required_mask (struct device *dev) +u64 dma_get_required_mask_from_max_pfn (struct device *dev) {u32 low_totalram = ((max_pfn -1 ) <<PAGE_SHIFT );u32 high_totalram = ((max_pfn -1 ) >> (32 -PAGE_SHIFT ));
@@ -961,6 +960,13 @@ u64 dma_get_required_mask(struct device *dev) }return mask; } +EXPORT_SYMBOL_GPL (dma_get_required_mask_from_max_pfn); + +#ifndef ARCH_HAS_DMA_GET_REQUIRED_MASK +u64 dma_get_required_mask (struct device *dev) +{ +return dma_get_required_mask_from_max_pfn (dev); +}EXPORT_SYMBOL_GPL (dma_get_required_mask);#endif
include/linux/dma-mapping.h +1 -0
@@ -121,6 +121,7 @@ static inline int dma_coerce_mask_and_coherent(struct device *dev, u64 mask)return dma_set_mask_and_coherent (dev, mask); } +extern u64 dma_get_required_mask_from_max_pfn (struct device *dev);extern u64 dma_get_required_mask (struct device *dev);static inline unsigned int dma_get_max_seg_size (struct device *dev)
Signature · 1 line
1.7.10.4
PATCH 2/2 x86, xen: correct dma_get_required_mask() for Xen PV guests
In reply to David VrabelOn systems where DMA addresses and physical addresses are not 1:1 (such as Xen PV guests), the generic dma_get_required_mask() will not return the correct mask (since it uses max_pfn).
Some device drivers (such as mptsas, mpt2sas) use dma_get_required_mask() to set device DMA masks to allow them to use only 32-bit DMA addresses in hardware structures. This results in unnecessary use of the SWIOTLB if DMA addresses are more than 32-bits, impacting performance significantly.
Provide an arch-specific dma_get_required_mask() that defaults to the generic dma_get_required_mask_from_pfn().
Under Xen, the required DMA mask can then be set to always 64-bits.
--- arch/x86/include/asm/device.h | 2 ++ arch/x86/kernel/pci-dma.c | 7 +++++++ arch/x86/xen/pci-swiotlb-xen.c | 1 + drivers/xen/swiotlb-xen.c | 7 +++++++ include/xen/swiotlb-xen.h | 4 ++++ 5 files changed, 21 insertions(+)
arch/x86/include/asm/device.h +2 -0
@@ -13,4 +13,6 @@ struct dev_archdata {struct pdev_archdata { }; +#define ARCH_HAS_DMA_GET_REQUIRED_MASK +#endif /* _ASM_X86_DEVICE_H */
arch/x86/kernel/pci-dma.c +7 -0
@@ -135,6 +135,13 @@ void dma_generic_free_coherent(struct device *dev, size_t size, void *vaddr,free_pages ((unsigned long )vaddr,get_order (size)); } +u64 dma_get_required_mask (struct device *dev) +{ +if (dma_ops->get_required_mask) +return dma_ops->get_required_mask (dev); +return dma_get_required_mask_from_max_pfn (dev); +} +/* * See <Documentation/x86/x86_64/boot-options.txt> for the iommu kernel * parameter documentation.
arch/x86/xen/pci-swiotlb-xen.c +1 -0
@@ -31,6 +31,7 @@ static struct dma_map_ops xen_swiotlb_dma_ops = { .map_page = xen_swiotlb_map_page, .unmap_page = xen_swiotlb_unmap_page, .dma_supported = xen_swiotlb_dma_supported, + .get_required_mask = xen_swiotlb_get_required_mask, };/*
drivers/xen/swiotlb-xen.c +7 -0
@@ -683,3 +683,10 @@ xen_swiotlb_set_dma_mask(struct device *dev, u64 dma_mask)return 0 ; }EXPORT_SYMBOL_GPL (xen_swiotlb_set_dma_mask); + +u64 +xen_swiotlb_get_required_mask (struct device *dev) +{ +return DMA_BIT_MASK (64 ); +} +EXPORT_SYMBOL_GPL (xen_swiotlb_get_required_mask);
include/xen/swiotlb-xen.h +4 -0
@@ -58,4 +58,8 @@ xen_swiotlb_dma_supported(struct device *hwdev, u64 mask);extern int xen_swiotlb_set_dma_mask (struct device *dev,u64 dma_mask); + +extern u64 +xen_swiotlb_get_required_mask (struct device *dev); +#endif /* __LINUX_SWIOTLB_XEN_H */
Signature · 1 line
1.7.10.4
PATCH 2/2 x86,xen: correct dma_get_required_mask() for Xen PV guests
In reply to David VrabelOn systems where DMA addresses and physical addresses are not 1:1 (such as Xen PV guests), the generic dma_get_required_mask() will not return the correct mask (since it uses max_pfn).
Some device drivers (such as mptsas, mpt2sas) use dma_get_required_mask() to set device DMA masks to allow them to use only 32-bit DMA addresses in hardware structures. This results in unnecessary use of the SWIOTLB if DMA addresses are more than 32-bits, impacting performance significantly.
Provide an arch-specific dma_get_required_mask() that defaults to the generic dma_get_required_mask_from_pfn().
Under Xen, the required DMA mask can then be set to always 64-bits.
--- arch/x86/include/asm/device.h | 2 ++ arch/x86/kernel/pci-dma.c | 7 +++++++ arch/x86/xen/pci-swiotlb-xen.c | 1 + drivers/xen/swiotlb-xen.c | 7 +++++++ include/xen/swiotlb-xen.h | 4 ++++ 5 files changed, 21 insertions(+)
arch/x86/include/asm/device.h +2 -0
@@ -13,4 +13,6 @@ struct dev_archdata {struct pdev_archdata { }; +#define ARCH_HAS_DMA_GET_REQUIRED_MASK +#endif /* _ASM_X86_DEVICE_H */
arch/x86/kernel/pci-dma.c +7 -0
@@ -135,6 +135,13 @@ void dma_generic_free_coherent(struct device *dev, size_t size, void *vaddr,free_pages ((unsigned long )vaddr,get_order (size)); } +u64 dma_get_required_mask (struct device *dev) +{ +if (dma_ops->get_required_mask) +return dma_ops->get_required_mask (dev); +return dma_get_required_mask_from_max_pfn (dev); +} +/* * See <Documentation/x86/x86_64/boot-options.txt> for the iommu kernel * parameter documentation.
arch/x86/xen/pci-swiotlb-xen.c +1 -0
@@ -31,6 +31,7 @@ static struct dma_map_ops xen_swiotlb_dma_ops = { .map_page = xen_swiotlb_map_page, .unmap_page = xen_swiotlb_unmap_page, .dma_supported = xen_swiotlb_dma_supported, + .get_required_mask = xen_swiotlb_get_required_mask, };/*
drivers/xen/swiotlb-xen.c +7 -0
@@ -683,3 +683,10 @@ xen_swiotlb_set_dma_mask(struct device *dev, u64 dma_mask)return 0 ; }EXPORT_SYMBOL_GPL (xen_swiotlb_set_dma_mask); + +u64 +xen_swiotlb_get_required_mask (struct device *dev) +{ +return DMA_BIT_MASK (64 ); +} +EXPORT_SYMBOL_GPL (xen_swiotlb_get_required_mask);
include/xen/swiotlb-xen.h +4 -0
@@ -58,4 +58,8 @@ xen_swiotlb_dma_supported(struct device *hwdev, u64 mask);extern int xen_swiotlb_set_dma_mask (struct device *dev,u64 dma_mask); + +extern u64 +xen_swiotlb_get_required_mask (struct device *dev); +#endif /* __LINUX_SWIOTLB_XEN_H */
Signature · 1 line
1.7.10.4
Re: [PATCH 0/2] dma, x86, xen: fix dma_get_required_mask() when dma != phys address
In reply to David VrabelQuoted 8 lines · David Vrabel
x86 uses the generic
dma_get_required_mask()which assumes that DMA address == physical address. For Xen PV guests this is not the case and certain drivers will cause excessive use of the SWIOTLB.Patch #1 is a straight forward refactoring of a common function.
Patch #2 addes the required arch-specific implementation of
dma_get_required_mask()to x86.
David,
Thank you for taking a look at that. If you would like, you can tack on Reviewed-by: Konrad Rzeszutek Wilk <[email protected]>
And with the *SWIOTLB maintainer hat on - you can add that tag as well.
Thanks!
Quoted 3 lines
David
Quoted 8 lines · David Vrabel
x86 uses the generic
dma_get_required_mask()which assumes that DMA address == physical address. For Xen PV guests this is not the case and certain drivers will cause excessive use of the SWIOTLB.Patch #1 is a straight forward refactoring of a common function.
Patch #2 addes the required arch-specific implementation of
dma_get_required_mask()to x86.
David,
Thank you for taking a look at that. If you would like, you can tack on Reviewed-by: Konrad Rzeszutek Wilk <[email protected]>
And with the *SWIOTLB maintainer hat on - you can add that tag as well.
Thanks!
Quoted 3 lines
David