| Message ID | 20180925162231.4354-4-logang@deltatee.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Copy Offload in NVMe Fabrics with P2P PCI Memory | expand |
On Tue, 2018-09-25 at 10:22 -0600, Logan Gunthorpe wrote: > +int pci_p2pdma_map_sg(struct device *dev, struct scatterlist *sg, int nents, > + enum dma_data_direction dir) > +{ > + struct dev_pagemap *pgmap; > + struct scatterlist *s; > + phys_addr_t paddr; > + int i; > + > + /* > + * p2pdma mappings are not compatible with devices that use > + * dma_virt_ops. If the upper layers do the right thing > + * this should never happen because it will be prevented > + * by the check in pci_p2pdma_add_client() > + */ > + if (WARN_ON_ONCE(IS_ENABLED(CONFIG_DMA_VIRT_OPS) && > + dev->dma_ops == &dma_virt_ops)) > + return 0; Are you assuming that the compiler will optimize out the dev->dma_ops == &dma_virt_ops test if CONFIG_DMA_VIRT_OPS=n such that no reference to the dma_virt_ops symbol appears in the object file? Are you sure all compilers and compiler versions that are used to build the Linux kernel will do that? Thanks, Bart.
On 2018-09-25 11:33 a.m., Bart Van Assche wrote: >> + /* >> + * p2pdma mappings are not compatible with devices that use >> + * dma_virt_ops. If the upper layers do the right thing >> + * this should never happen because it will be prevented >> + * by the check in pci_p2pdma_add_client() >> + */ >> + if (WARN_ON_ONCE(IS_ENABLED(CONFIG_DMA_VIRT_OPS) && >> + dev->dma_ops == &dma_virt_ops)) >> + return 0; > > Are you assuming that the compiler will optimize out the dev->dma_ops > == &dma_virt_ops test if CONFIG_DMA_VIRT_OPS=n such that no reference to the > dma_virt_ops symbol appears in the object file? Are you sure all compilers > and compiler versions that are used to build the Linux kernel will do that? Hmm, I can't say for certain but I would definitely expect any sane compiler to do so. It's a fairly obvious optimization and the kbuild robot is happy with it. I just did a test compile on 4.9 (which is the earliest compiler version I have readily available) and it worked fine. The minimum gcc the kernel supports is 4.6 right now. I've also done a similar test on godbolt.org[1] and can't find any compiler version that doesn't work. Logan [1] https://godbolt.org/z/d5Mhee
diff --git a/drivers/pci/p2pdma.c b/drivers/pci/p2pdma.c index d61389945da2..bbbf86165428 100644 --- a/drivers/pci/p2pdma.c +++ b/drivers/pci/p2pdma.c @@ -190,6 +190,8 @@ int pci_p2pdma_add_resource(struct pci_dev *pdev, int bar, size_t size, pgmap->res.flags = pci_resource_flags(pdev, bar); pgmap->ref = &pdev->p2pdma->devmap_ref; pgmap->type = MEMORY_DEVICE_PCI_P2PDMA; + pgmap->pci_p2pdma_bus_offset = pci_bus_address(pdev, bar) - + pci_resource_start(pdev, bar); addr = devm_memremap_pages(&pdev->dev, pgmap); if (IS_ERR(addr)) { @@ -816,3 +818,44 @@ void pci_p2pmem_publish(struct pci_dev *pdev, bool publish) pdev->p2pdma->p2pmem_published = publish; } EXPORT_SYMBOL_GPL(pci_p2pmem_publish); + +/** + * pci_p2pdma_map_sg - map a PCI peer-to-peer scatterlist for DMA + * @dev: device doing the DMA request + * @sg: scatter list to map + * @nents: elements in the scatterlist + * @dir: DMA direction + * + * Scatterlists mapped with this function should not be unmapped in any way. + * + * Returns the number of SG entries mapped or 0 on error. + */ +int pci_p2pdma_map_sg(struct device *dev, struct scatterlist *sg, int nents, + enum dma_data_direction dir) +{ + struct dev_pagemap *pgmap; + struct scatterlist *s; + phys_addr_t paddr; + int i; + + /* + * p2pdma mappings are not compatible with devices that use + * dma_virt_ops. If the upper layers do the right thing + * this should never happen because it will be prevented + * by the check in pci_p2pdma_add_client() + */ + if (WARN_ON_ONCE(IS_ENABLED(CONFIG_DMA_VIRT_OPS) && + dev->dma_ops == &dma_virt_ops)) + return 0; + + for_each_sg(sg, s, nents, i) { + pgmap = sg_page(s)->pgmap; + paddr = sg_phys(s); + + s->dma_address = paddr - pgmap->pci_p2pdma_bus_offset; + sg_dma_len(s) = s->length; + } + + return nents; +} +EXPORT_SYMBOL_GPL(pci_p2pdma_map_sg); diff --git a/include/linux/memremap.h b/include/linux/memremap.h index 9553370ebdad..0ac69ddf5fc4 100644 --- a/include/linux/memremap.h +++ b/include/linux/memremap.h @@ -125,6 +125,7 @@ struct dev_pagemap { struct device *dev; void *data; enum memory_type type; + u64 pci_p2pdma_bus_offset; }; #ifdef CONFIG_ZONE_DEVICE diff --git a/include/linux/pci-p2pdma.h b/include/linux/pci-p2pdma.h index 127e95f50491..f039c48990a5 100644 --- a/include/linux/pci-p2pdma.h +++ b/include/linux/pci-p2pdma.h @@ -35,6 +35,8 @@ struct scatterlist *pci_p2pmem_alloc_sgl(struct pci_dev *pdev, unsigned int *nents, u32 length); void pci_p2pmem_free_sgl(struct pci_dev *pdev, struct scatterlist *sgl); void pci_p2pmem_publish(struct pci_dev *pdev, bool publish); +int pci_p2pdma_map_sg(struct device *dev, struct scatterlist *sg, int nents, + enum dma_data_direction dir); #else /* CONFIG_PCI_P2PDMA */ static inline int pci_p2pdma_add_resource(struct pci_dev *pdev, int bar, size_t size, u64 offset) @@ -97,5 +99,10 @@ static inline void pci_p2pmem_free_sgl(struct pci_dev *pdev, static inline void pci_p2pmem_publish(struct pci_dev *pdev, bool publish) { } +static inline int pci_p2pdma_map_sg(struct device *dev, + struct scatterlist *sg, int nents, enum dma_data_direction dir) +{ + return 0; +} #endif /* CONFIG_PCI_P2PDMA */ #endif /* _LINUX_PCI_P2P_H */