On Sat, 22 Nov 2008 08:18:02 -0800 "Yu, Fenghua" <fenghua.yu@xxxxxxxxx> wrote: > > >+int intel_dma_mapping_error(struct device *hwdev, dma_addr_t dma_addr) > >+{ > >+ return !dma_addr; > >+} > > The implementation of inte_dma_mapping_error is incomplete. > > Dma_addr is valid only when it belongs to hwdev's domain. This is all about VT-d isolation fundamental. > > +int intel_dma_mapping_error(struct device *hwdev, dma_addr_t dma_addr) > +{ > + struct pci_dev *pdev = to_pci_dev(hwdev); > + struct dmar_domain *domain; > + struct iova *iova; > + > + domain = find_domain(pdev); > + > + iova = find_iova(&domain->iovad, IOVA_PFN(dma_addr)); > + if (!iova) > + return -EINVAL; > + return 0; > +} Hmm, intel_dma_mapping_error is used to test only the value that intel_map_single() returns. If intel_map_single() returns a non-zero dma address, it belongs to hwdev's domain. So intel_dma_mapping_error can simply return 1 (failure) if the dma_addr is zero. Do I misunderstand something? -- To unsubscribe from this list: send the line "unsubscribe linux-ia64" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
![]() |