X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=drivers%2Fpci%2Fintel-iommu.c;h=8d6159426311bda6a5bcab491b282e5427d6eff2;hb=11bd04f6f35621193311c32e0721142b073a7794;hp=c9272a1fb691ddbdf270be449680f1c465d2f3ad;hpb=2ebe31513fcbe7a781f27002f065b50ae195022f;p=safe%2Fjmp%2Flinux-2.6 diff --git a/drivers/pci/intel-iommu.c b/drivers/pci/intel-iommu.c index c9272a1..8d61594 100644 --- a/drivers/pci/intel-iommu.c +++ b/drivers/pci/intel-iommu.c @@ -37,6 +37,7 @@ #include #include #include +#include #include #include #include @@ -47,6 +48,7 @@ #define IS_GFX_DEVICE(pdev) ((pdev->class >> 16) == PCI_BASE_CLASS_DISPLAY) #define IS_ISA_DEVICE(pdev) ((pdev->class >> 8) == PCI_CLASS_BRIDGE_ISA) +#define IS_AZALIA(pdev) ((pdev)->vendor == 0x8086 && (pdev)->device == 0x3a3e) #define IOAPIC_RANGE_START (0xfee00000) #define IOAPIC_RANGE_END (0xfeefffff) @@ -93,6 +95,7 @@ static inline unsigned long virt_to_dma_pfn(void *p) /* global iommu list, set NULL for ignored DMAR units */ static struct intel_iommu **g_iommus; +static void __init check_tylersburg_isoch(void); static int rwbf_quirk; /* @@ -735,7 +738,7 @@ static struct dma_pte *pfn_to_dma_pte(struct dmar_domain *domain, return NULL; domain_flush_cache(domain, tmp_page, VTD_PAGE_SIZE); - pteval = (virt_to_dma_pfn(tmp_page) << VTD_PAGE_SHIFT) | DMA_PTE_READ | DMA_PTE_WRITE; + pteval = ((uint64_t)virt_to_dma_pfn(tmp_page) << VTD_PAGE_SHIFT) | DMA_PTE_READ | DMA_PTE_WRITE; if (cmpxchg64(&pte->val, 0ULL, pteval)) { /* Someone else set it while we were thinking; use theirs. */ free_pgtable_page(tmp_page); @@ -785,9 +788,10 @@ static void dma_pte_clear_range(struct dmar_domain *domain, BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width); BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width); + BUG_ON(start_pfn > last_pfn); /* we don't need lock here; nobody else touches the iova range */ - while (start_pfn <= last_pfn) { + do { first_pte = pte = dma_pfn_level_pte(domain, start_pfn, 1); if (!pte) { start_pfn = align_to_level(start_pfn + 1, 2); @@ -801,7 +805,8 @@ static void dma_pte_clear_range(struct dmar_domain *domain, domain_flush_cache(domain, first_pte, (void *)pte - (void *)first_pte); - } + + } while (start_pfn && start_pfn <= last_pfn); } /* free page table pages. last level pte should already be cleared */ @@ -817,6 +822,7 @@ static void dma_pte_free_pagetable(struct dmar_domain *domain, BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width); BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width); + BUG_ON(start_pfn > last_pfn); /* We don't need lock here; nobody else touches the iova range */ level = 2; @@ -827,7 +833,7 @@ static void dma_pte_free_pagetable(struct dmar_domain *domain, if (tmp + level_size(level) - 1 > last_pfn) return; - while (tmp + level_size(level) - 1 <= last_pfn) { + do { first_pte = pte = dma_pfn_level_pte(domain, tmp, level); if (!pte) { tmp = align_to_level(tmp + 1, level + 1); @@ -846,7 +852,7 @@ static void dma_pte_free_pagetable(struct dmar_domain *domain, domain_flush_cache(domain, first_pte, (void *)pte - (void *)first_pte); - } + } while (tmp && tmp + level_size(level) - 1 <= last_pfn); level++; } /* free pgd */ @@ -1605,7 +1611,7 @@ domain_context_mapping(struct dmar_domain *domain, struct pci_dev *pdev, return ret; parent = parent->bus->self; } - if (tmp->is_pcie) /* this is a PCIE-to-PCI bridge */ + if (pci_is_pcie(tmp)) /* this is a PCIE-to-PCI bridge */ return domain_context_mapping_one(domain, pci_domain_nr(tmp->subordinate), tmp->subordinate->number, 0, @@ -1645,7 +1651,7 @@ static int domain_context_mapped(struct pci_dev *pdev) return ret; parent = parent->bus->self; } - if (tmp->is_pcie) + if (pci_is_pcie(tmp)) return device_context_mapped(iommu, tmp->subordinate->number, 0); else @@ -1815,7 +1821,7 @@ static struct dmar_domain *get_domain_for_dev(struct pci_dev *pdev, int gaw) dev_tmp = pci_find_upstream_pcie_bridge(pdev); if (dev_tmp) { - if (dev_tmp->is_pcie) { + if (pci_is_pcie(dev_tmp)) { bus = dev_tmp->subordinate->number; devfn = 0; } else { @@ -1930,6 +1936,9 @@ error: } static int iommu_identity_mapping; +#define IDENTMAP_ALL 1 +#define IDENTMAP_GFX 2 +#define IDENTMAP_AZALIA 4 static int iommu_domain_identity_map(struct dmar_domain *domain, unsigned long long start, @@ -2147,8 +2156,14 @@ static int domain_add_dev_info(struct dmar_domain *domain, static int iommu_should_identity_map(struct pci_dev *pdev, int startup) { - if (iommu_identity_mapping == 2) - return IS_GFX_DEVICE(pdev); + if ((iommu_identity_mapping & IDENTMAP_AZALIA) && IS_AZALIA(pdev)) + return 1; + + if ((iommu_identity_mapping & IDENTMAP_GFX) && IS_GFX_DEVICE(pdev)) + return 1; + + if (!(iommu_identity_mapping & IDENTMAP_ALL)) + return 0; /* * We want to start off with all devices in the 1:1 domain, and @@ -2167,7 +2182,7 @@ static int iommu_should_identity_map(struct pci_dev *pdev, int startup) * the 1:1 domain, just in _case_ one of their siblings turns out * not to be able to map all of memory. */ - if (!pdev->is_pcie) { + if (!pci_is_pcie(pdev)) { if (!pci_is_root_bus(pdev->bus)) return 0; if (pdev->class >> 8 == PCI_CLASS_BRIDGE_PCI) @@ -2328,11 +2343,14 @@ int __init init_dmars(void) } if (iommu_pass_through) - iommu_identity_mapping = 1; + iommu_identity_mapping |= IDENTMAP_ALL; + #ifdef CONFIG_DMAR_BROKEN_GFX_WA - else - iommu_identity_mapping = 2; + iommu_identity_mapping |= IDENTMAP_GFX; #endif + + check_tylersburg_isoch(); + /* * If pass through is not set or not enabled, setup context entries for * identity mappings for rmrr, gfx, and isa and may fall back to static @@ -2400,11 +2418,12 @@ int __init init_dmars(void) iommu->flush.flush_context(iommu, 0, 0, 0, DMA_CCMD_GLOBAL_INVL); iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH); - iommu_disable_protect_mem_regions(iommu); ret = iommu_enable_translation(iommu); if (ret) goto error; + + iommu_disable_protect_mem_regions(iommu); } return 0; @@ -2645,10 +2664,9 @@ static void flush_unmaps(void) unsigned long mask; struct iova *iova = deferred_flush[i].iova[j]; - mask = (iova->pfn_hi - iova->pfn_lo + 1) << PAGE_SHIFT; - mask = ilog2(mask >> VTD_PAGE_SHIFT); + mask = ilog2(mm_to_dma_pfn(iova->pfn_hi - iova->pfn_lo + 1)); iommu_flush_dev_iotlb(deferred_flush[i].domain[j], - iova->pfn_lo << PAGE_SHIFT, mask); + (uint64_t)iova->pfn_lo << PAGE_SHIFT, mask); __free_iova(&deferred_flush[i].domain[j]->iovad, iova); } deferred_flush[i].next = 0; @@ -2749,7 +2767,15 @@ static void *intel_alloc_coherent(struct device *hwdev, size_t size, size = PAGE_ALIGN(size); order = get_order(size); - flags &= ~(GFP_DMA | GFP_DMA32); + + if (!iommu_no_mapping(hwdev)) + flags &= ~(GFP_DMA | GFP_DMA32); + else if (hwdev->coherent_dma_mask < dma_get_required_mask(hwdev)) { + if (hwdev->coherent_dma_mask < DMA_BIT_MASK(32)) + flags |= GFP_DMA; + else + flags |= GFP_DMA32; + } vaddr = (void *)__get_free_pages(flags, order); if (!vaddr) @@ -3064,8 +3090,8 @@ static int init_iommu_hw(void) DMA_CCMD_GLOBAL_INVL); iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH); - iommu_disable_protect_mem_regions(iommu); iommu_enable_translation(iommu); + iommu_disable_protect_mem_regions(iommu); } return 0; @@ -3189,21 +3215,58 @@ static int __init init_iommu_sysfs(void) } #endif /* CONFIG_PM */ +/* + * Here we only respond to action of unbound device from driver. + * + * Added device is not attached to its DMAR domain here yet. That will happen + * when mapping the device to iova. + */ +static int device_notifier(struct notifier_block *nb, + unsigned long action, void *data) +{ + struct device *dev = data; + struct pci_dev *pdev = to_pci_dev(dev); + struct dmar_domain *domain; + + domain = find_domain(pdev); + if (!domain) + return 0; + + if (action == BUS_NOTIFY_UNBOUND_DRIVER && !iommu_pass_through) + domain_remove_one_dev_info(domain, pdev); + + return 0; +} + +static struct notifier_block device_nb = { + .notifier_call = device_notifier, +}; + int __init intel_iommu_init(void) { int ret = 0; + int force_on = 0; + + /* VT-d is required for a TXT/tboot launch, so enforce that */ + force_on = tboot_force_iommu(); - if (dmar_table_init()) + if (dmar_table_init()) { + if (force_on) + panic("tboot: Failed to initialize DMAR table\n"); return -ENODEV; + } - if (dmar_dev_scope_init()) + if (dmar_dev_scope_init()) { + if (force_on) + panic("tboot: Failed to initialize DMAR device scope\n"); return -ENODEV; + } /* * Check the need for DMA-remapping initialization now. * Above initialization will also be used by Interrupt-remapping. */ - if (no_iommu || swiotlb || dmar_disabled) + if (no_iommu || dmar_disabled) return -ENODEV; iommu_init_mempool(); @@ -3213,6 +3276,8 @@ int __init intel_iommu_init(void) ret = init_dmars(); if (ret) { + if (force_on) + panic("tboot: Failed to initialize DMARs\n"); printk(KERN_ERR "IOMMU: dmar init failed\n"); put_iova_domain(&reserved_iova_list); iommu_exit_mempool(); @@ -3222,13 +3287,17 @@ int __init intel_iommu_init(void) "PCI-DMA: Intel(R) Virtualization Technology for Directed I/O\n"); init_timer(&unmap_timer); - force_iommu = 1; +#ifdef CONFIG_SWIOTLB + swiotlb = 0; +#endif dma_ops = &intel_dma_ops; init_iommu_sysfs(); register_iommu(&intel_iommu_ops); + bus_register_notifier(&pci_bus_type, &device_nb); + return 0; } @@ -3250,7 +3319,7 @@ static void iommu_detach_dependent_devices(struct intel_iommu *iommu, parent->devfn); parent = parent->bus->self; } - if (tmp->is_pcie) /* this is a PCIE-to-PCI bridge */ + if (pci_is_pcie(tmp)) /* this is a PCIE-to-PCI bridge */ iommu_detach_dev(iommu, tmp->subordinate->number, 0); else /* this is a legacy PCI bridge */ @@ -3654,3 +3723,61 @@ static void __devinit quirk_iommu_rwbf(struct pci_dev *dev) } DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2a40, quirk_iommu_rwbf); + +/* On Tylersburg chipsets, some BIOSes have been known to enable the + ISOCH DMAR unit for the Azalia sound device, but not give it any + TLB entries, which causes it to deadlock. Check for that. We do + this in a function called from init_dmars(), instead of in a PCI + quirk, because we don't want to print the obnoxious "BIOS broken" + message if VT-d is actually disabled. +*/ +static void __init check_tylersburg_isoch(void) +{ + struct pci_dev *pdev; + uint32_t vtisochctrl; + + /* If there's no Azalia in the system anyway, forget it. */ + pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x3a3e, NULL); + if (!pdev) + return; + pci_dev_put(pdev); + + /* System Management Registers. Might be hidden, in which case + we can't do the sanity check. But that's OK, because the + known-broken BIOSes _don't_ actually hide it, so far. */ + pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x342e, NULL); + if (!pdev) + return; + + if (pci_read_config_dword(pdev, 0x188, &vtisochctrl)) { + pci_dev_put(pdev); + return; + } + + pci_dev_put(pdev); + + /* If Azalia DMA is routed to the non-isoch DMAR unit, fine. */ + if (vtisochctrl & 1) + return; + + /* Drop all bits other than the number of TLB entries */ + vtisochctrl &= 0x1c; + + /* If we have the recommended number of TLB entries (16), fine. */ + if (vtisochctrl == 0x10) + return; + + /* Zero TLB entries? You get to ride the short bus to school. */ + if (!vtisochctrl) { + WARN(1, "Your BIOS is broken; DMA routed to ISOCH DMAR unit but no TLB space.\n" + "BIOS vendor: %s; Ver: %s; Product Version: %s\n", + dmi_get_system_info(DMI_BIOS_VENDOR), + dmi_get_system_info(DMI_BIOS_VERSION), + dmi_get_system_info(DMI_PRODUCT_VERSION)); + iommu_identity_mapping |= IDENTMAP_AZALIA; + return; + } + + printk(KERN_WARNING "DMAR: Recommended TLB entries for ISOCH unit is 16; your BIOS set %d\n", + vtisochctrl); +}