From aa6f5ffbdba45aa8e19e5048648fc6c7b25376d3 Mon Sep 17 00:00:00 2001 From: merge Date: Thu, 22 Jan 2009 13:55:32 +0000 Subject: MERGE-via-pending-tracking-hist-MERGE-via-stable-tracking-MERGE-via-mokopatches-tracking-fix-stray-endmenu-patch-1232632040-1232632141 pending-tracking-hist top was MERGE-via-stable-tracking-MERGE-via-mokopatches-tracking-fix-stray-endmenu-patch-1232632040-1232632141 / fdf777a63bcb59e0dfd78bfe2c6242e01f6d4eb9 ... parent commitmessage: From: merge MERGE-via-stable-tracking-hist-MERGE-via-mokopatches-tracking-fix-stray-endmenu-patch-1232632040 stable-tracking-hist top was MERGE-via-mokopatches-tracking-fix-stray-endmenu-patch-1232632040 / 90463bfd2d5a3c8b52f6e6d71024a00e052b0ced ... parent commitmessage: From: merge MERGE-via-mokopatches-tracking-hist-fix-stray-endmenu-patch mokopatches-tracking-hist top was fix-stray-endmenu-patch / 3630e0be570de8057e7f8d2fe501ed353cdf34e6 ... parent commitmessage: From: Andy Green fix-stray-endmenu.patch Signed-off-by: Andy Green --- drivers/pci/intel-iommu.c | 951 +++++++++++++++++++++++++++++++++++++++------- 1 file changed, 823 insertions(+), 128 deletions(-) (limited to 'drivers/pci/intel-iommu.c') diff --git a/drivers/pci/intel-iommu.c b/drivers/pci/intel-iommu.c index a2692724b68..3dfecb20d5e 100644 --- a/drivers/pci/intel-iommu.c +++ b/drivers/pci/intel-iommu.c @@ -27,7 +27,6 @@ #include #include #include -#include #include #include #include @@ -35,6 +34,7 @@ #include #include #include +#include #include #include #include @@ -54,6 +54,195 @@ #define DOMAIN_MAX_ADDR(gaw) ((((u64)1) << gaw) - 1) +#define IOVA_PFN(addr) ((addr) >> PAGE_SHIFT) +#define DMA_32BIT_PFN IOVA_PFN(DMA_32BIT_MASK) +#define DMA_64BIT_PFN IOVA_PFN(DMA_64BIT_MASK) + +/* global iommu list, set NULL for ignored DMAR units */ +static struct intel_iommu **g_iommus; + +/* + * 0: Present + * 1-11: Reserved + * 12-63: Context Ptr (12 - (haw-1)) + * 64-127: Reserved + */ +struct root_entry { + u64 val; + u64 rsvd1; +}; +#define ROOT_ENTRY_NR (VTD_PAGE_SIZE/sizeof(struct root_entry)) +static inline bool root_present(struct root_entry *root) +{ + return (root->val & 1); +} +static inline void set_root_present(struct root_entry *root) +{ + root->val |= 1; +} +static inline void set_root_value(struct root_entry *root, unsigned long value) +{ + root->val |= value & VTD_PAGE_MASK; +} + +static inline struct context_entry * +get_context_addr_from_root(struct root_entry *root) +{ + return (struct context_entry *) + (root_present(root)?phys_to_virt( + root->val & VTD_PAGE_MASK) : + NULL); +} + +/* + * low 64 bits: + * 0: present + * 1: fault processing disable + * 2-3: translation type + * 12-63: address space root + * high 64 bits: + * 0-2: address width + * 3-6: aval + * 8-23: domain id + */ +struct context_entry { + u64 lo; + u64 hi; +}; + +static inline bool context_present(struct context_entry *context) +{ + return (context->lo & 1); +} +static inline void context_set_present(struct context_entry *context) +{ + context->lo |= 1; +} + +static inline void context_set_fault_enable(struct context_entry *context) +{ + context->lo &= (((u64)-1) << 2) | 1; +} + +#define CONTEXT_TT_MULTI_LEVEL 0 + +static inline void context_set_translation_type(struct context_entry *context, + unsigned long value) +{ + context->lo &= (((u64)-1) << 4) | 3; + context->lo |= (value & 3) << 2; +} + +static inline void context_set_address_root(struct context_entry *context, + unsigned long value) +{ + context->lo |= value & VTD_PAGE_MASK; +} + +static inline void context_set_address_width(struct context_entry *context, + unsigned long value) +{ + context->hi |= value & 7; +} + +static inline void context_set_domain_id(struct context_entry *context, + unsigned long value) +{ + context->hi |= (value & ((1 << 16) - 1)) << 8; +} + +static inline void context_clear_entry(struct context_entry *context) +{ + context->lo = 0; + context->hi = 0; +} + +/* + * 0: readable + * 1: writable + * 2-6: reserved + * 7: super page + * 8-11: available + * 12-63: Host physcial address + */ +struct dma_pte { + u64 val; +}; + +static inline void dma_clear_pte(struct dma_pte *pte) +{ + pte->val = 0; +} + +static inline void dma_set_pte_readable(struct dma_pte *pte) +{ + pte->val |= DMA_PTE_READ; +} + +static inline void dma_set_pte_writable(struct dma_pte *pte) +{ + pte->val |= DMA_PTE_WRITE; +} + +static inline void dma_set_pte_prot(struct dma_pte *pte, unsigned long prot) +{ + pte->val = (pte->val & ~3) | (prot & 3); +} + +static inline u64 dma_pte_addr(struct dma_pte *pte) +{ + return (pte->val & VTD_PAGE_MASK); +} + +static inline void dma_set_pte_addr(struct dma_pte *pte, u64 addr) +{ + pte->val |= (addr & VTD_PAGE_MASK); +} + +static inline bool dma_pte_present(struct dma_pte *pte) +{ + return (pte->val & 3) != 0; +} + +/* devices under the same p2p bridge are owned in one domain */ +#define DOMAIN_FLAG_P2P_MULTIPLE_DEVICES (1 << 0) + +/* domain represents a virtual machine, more than one devices + * across iommus may be owned in one domain, e.g. kvm guest. + */ +#define DOMAIN_FLAG_VIRTUAL_MACHINE (1 << 1) + +struct dmar_domain { + int id; /* domain id */ + unsigned long iommu_bmp; /* bitmap of iommus this domain uses*/ + + struct list_head devices; /* all devices' list */ + struct iova_domain iovad; /* iova's that belong to this domain */ + + struct dma_pte *pgd; /* virtual address */ + spinlock_t mapping_lock; /* page table lock */ + int gaw; /* max guest address width */ + + /* adjusted guest address width, 0 is level 2 30-bit */ + int agaw; + + int flags; /* flags to find out type of domain */ + + int iommu_coherency;/* indicate coherency of iommu access */ + int iommu_count; /* reference count of iommu */ + spinlock_t iommu_lock; /* protect iommu set in domain */ + u64 max_addr; /* maximum mapped address */ +}; + +/* PCI domain-device relationship */ +struct device_domain_info { + struct list_head link; /* link to domain siblings */ + struct list_head global; /* link to global list */ + u8 bus; /* PCI bus numer */ + u8 devfn; /* PCI devfn number */ + struct pci_dev *dev; /* it's NULL for PCIE-to-PCI bridge */ + struct dmar_domain *domain; /* pointer to domain */ +}; static void flush_unmaps_timeout(unsigned long data); @@ -88,6 +277,8 @@ static int intel_iommu_strict; static DEFINE_SPINLOCK(device_domain_lock); static LIST_HEAD(device_domain_list); +static struct iommu_ops intel_iommu_ops; + static int __init intel_iommu_setup(char *str) { if (!str) @@ -184,6 +375,88 @@ void free_iova_mem(struct iova *iova) kmem_cache_free(iommu_iova_cache, iova); } + +static inline int width_to_agaw(int width); + +/* calculate agaw for each iommu. + * "SAGAW" may be different across iommus, use a default agaw, and + * get a supported less agaw for iommus that don't support the default agaw. + */ +int iommu_calculate_agaw(struct intel_iommu *iommu) +{ + unsigned long sagaw; + int agaw = -1; + + sagaw = cap_sagaw(iommu->cap); + for (agaw = width_to_agaw(DEFAULT_DOMAIN_ADDRESS_WIDTH); + agaw >= 0; agaw--) { + if (test_bit(agaw, &sagaw)) + break; + } + + return agaw; +} + +/* in native case, each domain is related to only one iommu */ +static struct intel_iommu *domain_get_iommu(struct dmar_domain *domain) +{ + int iommu_id; + + BUG_ON(domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE); + + iommu_id = find_first_bit(&domain->iommu_bmp, g_num_of_iommus); + if (iommu_id < 0 || iommu_id >= g_num_of_iommus) + return NULL; + + return g_iommus[iommu_id]; +} + +/* "Coherency" capability may be different across iommus */ +static void domain_update_iommu_coherency(struct dmar_domain *domain) +{ + int i; + + domain->iommu_coherency = 1; + + i = find_first_bit(&domain->iommu_bmp, g_num_of_iommus); + for (; i < g_num_of_iommus; ) { + if (!ecap_coherent(g_iommus[i]->ecap)) { + domain->iommu_coherency = 0; + break; + } + i = find_next_bit(&domain->iommu_bmp, g_num_of_iommus, i+1); + } +} + +static struct intel_iommu *device_to_iommu(u8 bus, u8 devfn) +{ + struct dmar_drhd_unit *drhd = NULL; + int i; + + for_each_drhd_unit(drhd) { + if (drhd->ignored) + continue; + + for (i = 0; i < drhd->devices_cnt; i++) + if (drhd->devices[i] && + drhd->devices[i]->bus->number == bus && + drhd->devices[i]->devfn == devfn) + return drhd->iommu; + + if (drhd->include_all) + return drhd->iommu; + } + + return NULL; +} + +static void domain_flush_cache(struct dmar_domain *domain, + void *addr, int size) +{ + if (!domain->iommu_coherency) + clflush_cache_range(addr, size); +} + /* Gets context entry for a given bus and devfn */ static struct context_entry * device_to_context_entry(struct intel_iommu *iommu, u8 bus, u8 devfn) @@ -226,7 +499,7 @@ static int device_context_mapped(struct intel_iommu *iommu, u8 bus, u8 devfn) ret = 0; goto out; } - ret = context_present(context[devfn]); + ret = context_present(&context[devfn]); out: spin_unlock_irqrestore(&iommu->lock, flags); return ret; @@ -242,7 +515,7 @@ static void clear_context_table(struct intel_iommu *iommu, u8 bus, u8 devfn) root = &iommu->root_entry[bus]; context = get_context_addr_from_root(root); if (context) { - context_clear_entry(context[devfn]); + context_clear_entry(&context[devfn]); __iommu_flush_cache(iommu, &context[devfn], \ sizeof(*context)); } @@ -339,7 +612,7 @@ static struct dma_pte * addr_to_dma_pte(struct dmar_domain *domain, u64 addr) if (level == 1) break; - if (!dma_pte_present(*pte)) { + if (!dma_pte_present(pte)) { tmp_page = alloc_pgtable_page(); if (!tmp_page) { @@ -347,18 +620,17 @@ static struct dma_pte * addr_to_dma_pte(struct dmar_domain *domain, u64 addr) flags); return NULL; } - __iommu_flush_cache(domain->iommu, tmp_page, - PAGE_SIZE); - dma_set_pte_addr(*pte, virt_to_phys(tmp_page)); + domain_flush_cache(domain, tmp_page, PAGE_SIZE); + dma_set_pte_addr(pte, virt_to_phys(tmp_page)); /* * high level table always sets r/w, last level page * table control read/write */ - dma_set_pte_readable(*pte); - dma_set_pte_writable(*pte); - __iommu_flush_cache(domain->iommu, pte, sizeof(*pte)); + dma_set_pte_readable(pte); + dma_set_pte_writable(pte); + domain_flush_cache(domain, pte, sizeof(*pte)); } - parent = phys_to_virt(dma_pte_addr(*pte)); + parent = phys_to_virt(dma_pte_addr(pte)); level--; } @@ -381,9 +653,9 @@ static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr, if (level == total) return pte; - if (!dma_pte_present(*pte)) + if (!dma_pte_present(pte)) break; - parent = phys_to_virt(dma_pte_addr(*pte)); + parent = phys_to_virt(dma_pte_addr(pte)); total--; } return NULL; @@ -398,8 +670,8 @@ static void dma_pte_clear_one(struct dmar_domain *domain, u64 addr) pte = dma_addr_level_pte(domain, addr, 1); if (pte) { - dma_clear_pte(*pte); - __iommu_flush_cache(domain->iommu, pte, sizeof(*pte)); + dma_clear_pte(pte); + domain_flush_cache(domain, pte, sizeof(*pte)); } } @@ -445,10 +717,9 @@ static void dma_pte_free_pagetable(struct dmar_domain *domain, pte = dma_addr_level_pte(domain, tmp, level); if (pte) { free_pgtable_page( - phys_to_virt(dma_pte_addr(*pte))); - dma_clear_pte(*pte); - __iommu_flush_cache(domain->iommu, - pte, sizeof(*pte)); + phys_to_virt(dma_pte_addr(pte))); + dma_clear_pte(pte); + domain_flush_cache(domain, pte, sizeof(*pte)); } tmp += level_size(level); } @@ -950,17 +1221,28 @@ static int iommu_init_domains(struct intel_iommu *iommu) static void domain_exit(struct dmar_domain *domain); +static void vm_domain_exit(struct dmar_domain *domain); void free_dmar_iommu(struct intel_iommu *iommu) { struct dmar_domain *domain; int i; + unsigned long flags; i = find_first_bit(iommu->domain_ids, cap_ndoms(iommu->cap)); for (; i < cap_ndoms(iommu->cap); ) { domain = iommu->domains[i]; clear_bit(i, iommu->domain_ids); - domain_exit(domain); + + spin_lock_irqsave(&domain->iommu_lock, flags); + if (--domain->iommu_count == 0) { + if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) + vm_domain_exit(domain); + else + domain_exit(domain); + } + spin_unlock_irqrestore(&domain->iommu_lock, flags); + i = find_next_bit(iommu->domain_ids, cap_ndoms(iommu->cap), i+1); } @@ -978,6 +1260,17 @@ void free_dmar_iommu(struct intel_iommu *iommu) kfree(iommu->domains); kfree(iommu->domain_ids); + g_iommus[iommu->seq_id] = NULL; + + /* if all iommus are freed, free g_iommus */ + for (i = 0; i < g_num_of_iommus; i++) { + if (g_iommus[i]) + break; + } + + if (i == g_num_of_iommus) + kfree(g_iommus); + /* free context mapping */ free_context_table(iommu); } @@ -1006,7 +1299,9 @@ static struct dmar_domain * iommu_alloc_domain(struct intel_iommu *iommu) set_bit(num, iommu->domain_ids); domain->id = num; - domain->iommu = iommu; + memset(&domain->iommu_bmp, 0, sizeof(unsigned long)); + set_bit(iommu->seq_id, &domain->iommu_bmp); + domain->flags = 0; iommu->domains[num] = domain; spin_unlock_irqrestore(&iommu->lock, flags); @@ -1016,10 +1311,13 @@ static struct dmar_domain * iommu_alloc_domain(struct intel_iommu *iommu) static void iommu_free_domain(struct dmar_domain *domain) { unsigned long flags; + struct intel_iommu *iommu; - spin_lock_irqsave(&domain->iommu->lock, flags); - clear_bit(domain->id, domain->iommu->domain_ids); - spin_unlock_irqrestore(&domain->iommu->lock, flags); + iommu = domain_get_iommu(domain); + + spin_lock_irqsave(&iommu->lock, flags); + clear_bit(domain->id, iommu->domain_ids); + spin_unlock_irqrestore(&iommu->lock, flags); } static struct iova_domain reserved_iova_list; @@ -1094,11 +1392,12 @@ static int domain_init(struct dmar_domain *domain, int guest_width) init_iova_domain(&domain->iovad, DMA_32BIT_PFN); spin_lock_init(&domain->mapping_lock); + spin_lock_init(&domain->iommu_lock); domain_reserve_special_ranges(domain); /* calculate AGAW */ - iommu = domain->iommu; + iommu = domain_get_iommu(domain); if (guest_width > cap_mgaw(iommu->cap)) guest_width = cap_mgaw(iommu->cap); domain->gaw = guest_width; @@ -1115,6 +1414,13 @@ static int domain_init(struct dmar_domain *domain, int guest_width) domain->agaw = agaw; INIT_LIST_HEAD(&domain->devices); + if (ecap_coherent(iommu->ecap)) + domain->iommu_coherency = 1; + else + domain->iommu_coherency = 0; + + domain->iommu_count = 1; + /* always allocate the top pgd */ domain->pgd = (struct dma_pte *)alloc_pgtable_page(); if (!domain->pgd) @@ -1151,28 +1457,82 @@ static int domain_context_mapping_one(struct dmar_domain *domain, u8 bus, u8 devfn) { struct context_entry *context; - struct intel_iommu *iommu = domain->iommu; unsigned long flags; + struct intel_iommu *iommu; + struct dma_pte *pgd; + unsigned long num; + unsigned long ndomains; + int id; + int agaw; pr_debug("Set context mapping for %02x:%02x.%d\n", bus, PCI_SLOT(devfn), PCI_FUNC(devfn)); BUG_ON(!domain->pgd); + + iommu = device_to_iommu(bus, devfn); + if (!iommu) + return -ENODEV; + context = device_to_context_entry(iommu, bus, devfn); if (!context) return -ENOMEM; spin_lock_irqsave(&iommu->lock, flags); - if (context_present(*context)) { + if (context_present(context)) { spin_unlock_irqrestore(&iommu->lock, flags); return 0; } - context_set_domain_id(*context, domain->id); - context_set_address_width(*context, domain->agaw); - context_set_address_root(*context, virt_to_phys(domain->pgd)); - context_set_translation_type(*context, CONTEXT_TT_MULTI_LEVEL); - context_set_fault_enable(*context); - context_set_present(*context); - __iommu_flush_cache(iommu, context, sizeof(*context)); + id = domain->id; + pgd = domain->pgd; + + if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) { + int found = 0; + + /* find an available domain id for this device in iommu */ + ndomains = cap_ndoms(iommu->cap); + num = find_first_bit(iommu->domain_ids, ndomains); + for (; num < ndomains; ) { + if (iommu->domains[num] == domain) { + id = num; + found = 1; + break; + } + num = find_next_bit(iommu->domain_ids, + cap_ndoms(iommu->cap), num+1); + } + + if (found == 0) { + num = find_first_zero_bit(iommu->domain_ids, ndomains); + if (num >= ndomains) { + spin_unlock_irqrestore(&iommu->lock, flags); + printk(KERN_ERR "IOMMU: no free domain ids\n"); + return -EFAULT; + } + + set_bit(num, iommu->domain_ids); + iommu->domains[num] = domain; + id = num; + } + + /* Skip top levels of page tables for + * iommu which has less agaw than default. + */ + for (agaw = domain->agaw; agaw != iommu->agaw; agaw--) { + pgd = phys_to_virt(dma_pte_addr(pgd)); + if (!dma_pte_present(pgd)) { + spin_unlock_irqrestore(&iommu->lock, flags); + return -ENOMEM; + } + } + } + + context_set_domain_id(context, id); + context_set_address_width(context, iommu->agaw); + context_set_address_root(context, virt_to_phys(pgd)); + context_set_translation_type(context, CONTEXT_TT_MULTI_LEVEL); + context_set_fault_enable(context); + context_set_present(context); + domain_flush_cache(domain, context, sizeof(*context)); /* it's a non-present to present mapping */ if (iommu->flush.flush_context(iommu, domain->id, @@ -1183,6 +1543,13 @@ static int domain_context_mapping_one(struct dmar_domain *domain, iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_DSI_FLUSH, 0); spin_unlock_irqrestore(&iommu->lock, flags); + + spin_lock_irqsave(&domain->iommu_lock, flags); + if (!test_and_set_bit(iommu->seq_id, &domain->iommu_bmp)) { + domain->iommu_count++; + domain_update_iommu_coherency(domain); + } + spin_unlock_irqrestore(&domain->iommu_lock, flags); return 0; } @@ -1218,13 +1585,17 @@ domain_context_mapping(struct dmar_domain *domain, struct pci_dev *pdev) tmp->bus->number, tmp->devfn); } -static int domain_context_mapped(struct dmar_domain *domain, - struct pci_dev *pdev) +static int domain_context_mapped(struct pci_dev *pdev) { int ret; struct pci_dev *tmp, *parent; + struct intel_iommu *iommu; + + iommu = device_to_iommu(pdev->bus->number, pdev->devfn); + if (!iommu) + return -ENODEV; - ret = device_context_mapped(domain->iommu, + ret = device_context_mapped(iommu, pdev->bus->number, pdev->devfn); if (!ret) return ret; @@ -1235,17 +1606,17 @@ static int domain_context_mapped(struct dmar_domain *domain, /* Secondary interface's bus number and devfn 0 */ parent = pdev->bus->self; while (parent != tmp) { - ret = device_context_mapped(domain->iommu, parent->bus->number, + ret = device_context_mapped(iommu, parent->bus->number, parent->devfn); if (!ret) return ret; parent = parent->bus->self; } if (tmp->is_pcie) - return device_context_mapped(domain->iommu, + return device_context_mapped(iommu, tmp->subordinate->number, 0); else - return device_context_mapped(domain->iommu, + return device_context_mapped(iommu, tmp->bus->number, tmp->devfn); } @@ -1273,22 +1644,25 @@ domain_page_mapping(struct dmar_domain *domain, dma_addr_t iova, /* We don't need lock here, nobody else * touches the iova range */ - BUG_ON(dma_pte_addr(*pte)); - dma_set_pte_addr(*pte, start_pfn << VTD_PAGE_SHIFT); - dma_set_pte_prot(*pte, prot); - __iommu_flush_cache(domain->iommu, pte, sizeof(*pte)); + BUG_ON(dma_pte_addr(pte)); + dma_set_pte_addr(pte, start_pfn << VTD_PAGE_SHIFT); + dma_set_pte_prot(pte, prot); + domain_flush_cache(domain, pte, sizeof(*pte)); start_pfn++; index++; } return 0; } -static void detach_domain_for_dev(struct dmar_domain *domain, u8 bus, u8 devfn) +static void iommu_detach_dev(struct intel_iommu *iommu, u8 bus, u8 devfn) { - clear_context_table(domain->iommu, bus, devfn); - domain->iommu->flush.flush_context(domain->iommu, 0, 0, 0, + if (!iommu) + return; + + clear_context_table(iommu, bus, devfn); + iommu->flush.flush_context(iommu, 0, 0, 0, DMA_CCMD_GLOBAL_INVL, 0); - domain->iommu->flush.flush_iotlb(domain->iommu, 0, 0, 0, + iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH, 0); } @@ -1296,6 +1670,7 @@ static void domain_remove_dev_info(struct dmar_domain *domain) { struct device_domain_info *info; unsigned long flags; + struct intel_iommu *iommu; spin_lock_irqsave(&device_domain_lock, flags); while (!list_empty(&domain->devices)) { @@ -1307,7 +1682,8 @@ static void domain_remove_dev_info(struct dmar_domain *domain) info->dev->dev.archdata.iommu = NULL; spin_unlock_irqrestore(&device_domain_lock, flags); - detach_domain_for_dev(info->domain, info->bus, info->devfn); + iommu = device_to_iommu(info->bus, info->devfn); + iommu_detach_dev(iommu, info->bus, info->devfn); free_devinfo_mem(info); spin_lock_irqsave(&device_domain_lock, flags); @@ -1400,7 +1776,7 @@ static struct dmar_domain *get_domain_for_dev(struct pci_dev *pdev, int gaw) info->dev = NULL; info->domain = domain; /* This domain is shared by devices under p2p bridge */ - domain->flags |= DOMAIN_FLAG_MULTIPLE_DEVICES; + domain->flags |= DOMAIN_FLAG_P2P_MULTIPLE_DEVICES; /* pcie-to-pci bridge already has a domain, uses it */ found = NULL; @@ -1563,6 +1939,11 @@ static void __init iommu_prepare_gfx_mapping(void) printk(KERN_ERR "IOMMU: mapping reserved region failed\n"); } } +#else /* !CONFIG_DMAR_GFX_WA */ +static inline void iommu_prepare_gfx_mapping(void) +{ + return; +} #endif #ifdef CONFIG_DMAR_FLOPPY_WA @@ -1590,7 +1971,7 @@ static inline void iommu_prepare_isa(void) } #endif /* !CONFIG_DMAR_FLPY_WA */ -int __init init_dmars(void) +static int __init init_dmars(void) { struct dmar_drhd_unit *drhd; struct dmar_rmrr_unit *rmrr; @@ -1613,9 +1994,18 @@ int __init init_dmars(void) */ } + g_iommus = kcalloc(g_num_of_iommus, sizeof(struct intel_iommu *), + GFP_KERNEL); + if (!g_iommus) { + printk(KERN_ERR "Allocating global iommu array failed\n"); + ret = -ENOMEM; + goto error; + } + deferred_flush = kzalloc(g_num_of_iommus * sizeof(struct deferred_flush_tables), GFP_KERNEL); if (!deferred_flush) { + kfree(g_iommus); ret = -ENOMEM; goto error; } @@ -1625,6 +2015,7 @@ int __init init_dmars(void) continue; iommu = drhd->iommu; + g_iommus[iommu->seq_id] = iommu; ret = iommu_init_domains(iommu); if (ret) @@ -1655,12 +2046,14 @@ int __init init_dmars(void) iommu->flush.flush_context = __iommu_flush_context; iommu->flush.flush_iotlb = __iommu_flush_iotlb; printk(KERN_INFO "IOMMU 0x%Lx: using Register based " - "invalidation\n", drhd->reg_base_addr); + "invalidation\n", + (unsigned long long)drhd->reg_base_addr); } else { iommu->flush.flush_context = qi_flush_context; iommu->flush.flush_iotlb = qi_flush_iotlb; printk(KERN_INFO "IOMMU 0x%Lx: using Queued " - "invalidation\n", drhd->reg_base_addr); + "invalidation\n", + (unsigned long long)drhd->reg_base_addr); } } @@ -1735,6 +2128,7 @@ error: iommu = drhd->iommu; free_iommu(iommu); } + kfree(g_iommus); return ret; } @@ -1803,7 +2197,7 @@ get_valid_domain_for_dev(struct pci_dev *pdev) } /* make sure context mapping is ok */ - if (unlikely(!domain_context_mapped(domain, pdev))) { + if (unlikely(!domain_context_mapped(pdev))) { ret = domain_context_mapping(domain, pdev); if (ret) { printk(KERN_ERR @@ -1825,6 +2219,7 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr, struct iova *iova; int prot = 0; int ret; + struct intel_iommu *iommu; BUG_ON(dir == DMA_NONE); if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO) @@ -1834,6 +2229,7 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr, if (!domain) return 0; + iommu = domain_get_iommu(domain); size = aligned_size((u64)paddr, size); iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask); @@ -1847,7 +2243,7 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr, * mappings.. */ if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL || \ - !cap_zlr(domain->iommu->cap)) + !cap_zlr(iommu->cap)) prot |= DMA_PTE_READ; if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL) prot |= DMA_PTE_WRITE; @@ -1863,10 +2259,10 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr, goto error; /* it's a non-present to present mapping */ - ret = iommu_flush_iotlb_psi(domain->iommu, domain->id, + ret = iommu_flush_iotlb_psi(iommu, domain->id, start_paddr, size >> VTD_PAGE_SHIFT, 1); if (ret) - iommu_flush_write_buffer(domain->iommu); + iommu_flush_write_buffer(iommu); return start_paddr + ((u64)paddr & (~PAGE_MASK)); @@ -1893,10 +2289,11 @@ static void flush_unmaps(void) /* just flush them all */ for (i = 0; i < g_num_of_iommus; i++) { - if (deferred_flush[i].next) { - struct intel_iommu *iommu = - deferred_flush[i].domain[0]->iommu; + struct intel_iommu *iommu = g_iommus[i]; + if (!iommu) + continue; + if (deferred_flush[i].next) { iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH, 0); for (j = 0; j < deferred_flush[i].next; j++) { @@ -1923,12 +2320,14 @@ static void add_unmap(struct dmar_domain *dom, struct iova *iova) { unsigned long flags; int next, iommu_id; + struct intel_iommu *iommu; spin_lock_irqsave(&async_umap_flush_lock, flags); if (list_size == HIGH_WATER_MARK) flush_unmaps(); - iommu_id = dom->iommu->seq_id; + iommu = domain_get_iommu(dom); + iommu_id = iommu->seq_id; next = deferred_flush[iommu_id].next; deferred_flush[iommu_id].domain[next] = dom; @@ -1950,12 +2349,15 @@ void intel_unmap_single(struct device *dev, dma_addr_t dev_addr, size_t size, struct dmar_domain *domain; unsigned long start_addr; struct iova *iova; + struct intel_iommu *iommu; if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO) return; domain = find_domain(pdev); BUG_ON(!domain); + iommu = domain_get_iommu(domain); + iova = find_iova(&domain->iovad, IOVA_PFN(dev_addr)); if (!iova) return; @@ -1971,9 +2373,9 @@ void intel_unmap_single(struct device *dev, dma_addr_t dev_addr, size_t size, /* free page tables */ dma_pte_free_pagetable(domain, start_addr, start_addr + size); if (intel_iommu_strict) { - if (iommu_flush_iotlb_psi(domain->iommu, + if (iommu_flush_iotlb_psi(iommu, domain->id, start_addr, size >> VTD_PAGE_SHIFT, 0)) - iommu_flush_write_buffer(domain->iommu); + iommu_flush_write_buffer(iommu); /* free iova */ __free_iova(&domain->iovad, iova); } else { @@ -2034,11 +2436,15 @@ void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist, size_t size = 0; void *addr; struct scatterlist *sg; + struct intel_iommu *iommu; if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO) return; domain = find_domain(pdev); + BUG_ON(!domain); + + iommu = domain_get_iommu(domain); iova = find_iova(&domain->iovad, IOVA_PFN(sglist[0].dma_address)); if (!iova) @@ -2055,9 +2461,9 @@ void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist, /* free page tables */ dma_pte_free_pagetable(domain, start_addr, start_addr + size); - if (iommu_flush_iotlb_psi(domain->iommu, domain->id, start_addr, + if (iommu_flush_iotlb_psi(iommu, domain->id, start_addr, size >> VTD_PAGE_SHIFT, 0)) - iommu_flush_write_buffer(domain->iommu); + iommu_flush_write_buffer(iommu); /* free iova */ __free_iova(&domain->iovad, iova); @@ -2091,6 +2497,7 @@ int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems, int ret; struct scatterlist *sg; unsigned long start_addr; + struct intel_iommu *iommu; BUG_ON(dir == DMA_NONE); if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO) @@ -2100,6 +2507,8 @@ int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems, if (!domain) return 0; + iommu = domain_get_iommu(domain); + for_each_sg(sglist, sg, nelems, i) { addr = SG_ENT_VIRT_ADDRESS(sg); addr = (void *)virt_to_phys(addr); @@ -2117,7 +2526,7 @@ int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems, * mappings.. */ if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL || \ - !cap_zlr(domain->iommu->cap)) + !cap_zlr(iommu->cap)) prot |= DMA_PTE_READ; if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL) prot |= DMA_PTE_WRITE; @@ -2149,9 +2558,9 @@ int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems, } /* it's a non-present to present mapping */ - if (iommu_flush_iotlb_psi(domain->iommu, domain->id, + if (iommu_flush_iotlb_psi(iommu, domain->id, start_addr, offset >> VTD_PAGE_SHIFT, 1)) - iommu_flush_write_buffer(domain->iommu); + iommu_flush_write_buffer(iommu); return nelems; } @@ -2323,10 +2732,220 @@ int __init intel_iommu_init(void) init_timer(&unmap_timer); force_iommu = 1; dma_ops = &intel_dma_ops; + + register_iommu(&intel_iommu_ops); + return 0; } -void intel_iommu_domain_exit(struct dmar_domain *domain) +static int vm_domain_add_dev_info(struct dmar_domain *domain, + struct pci_dev *pdev) +{ + struct device_domain_info *info; + unsigned long flags; + + info = alloc_devinfo_mem(); + if (!info) + return -ENOMEM; + + info->bus = pdev->bus->number; + info->devfn = pdev->devfn; + info->dev = pdev; + info->domain = domain; + + spin_lock_irqsave(&device_domain_lock, flags); + list_add(&info->link, &domain->devices); + list_add(&info->global, &device_domain_list); + pdev->dev.archdata.iommu = info; + spin_unlock_irqrestore(&device_domain_lock, flags); + + return 0; +} + +static void vm_domain_remove_one_dev_info(struct dmar_domain *domain, + struct pci_dev *pdev) +{ + struct device_domain_info *info; + struct intel_iommu *iommu; + unsigned long flags; + int found = 0; + struct list_head *entry, *tmp; + + iommu = device_to_iommu(pdev->bus->number, pdev->devfn); + if (!iommu) + return; + + spin_lock_irqsave(&device_domain_lock, flags); + list_for_each_safe(entry, tmp, &domain->devices) { + info = list_entry(entry, struct device_domain_info, link); + if (info->bus == pdev->bus->number && + info->devfn == pdev->devfn) { + list_del(&info->link); + list_del(&info->global); + if (info->dev) + info->dev->dev.archdata.iommu = NULL; + spin_unlock_irqrestore(&device_domain_lock, flags); + + iommu_detach_dev(iommu, info->bus, info->devfn); + free_devinfo_mem(info); + + spin_lock_irqsave(&device_domain_lock, flags); + + if (found) + break; + else + continue; + } + + /* if there is no other devices under the same iommu + * owned by this domain, clear this iommu in iommu_bmp + * update iommu count and coherency + */ + if (device_to_iommu(info->bus, info->devfn) == iommu) + found = 1; + } + + if (found == 0) { + unsigned long tmp_flags; + spin_lock_irqsave(&domain->iommu_lock, tmp_flags); + clear_bit(iommu->seq_id, &domain->iommu_bmp); + domain->iommu_count--; + domain_update_iommu_coherency(domain); + spin_unlock_irqrestore(&domain->iommu_lock, tmp_flags); + } + + spin_unlock_irqrestore(&device_domain_lock, flags); +} + +static void vm_domain_remove_all_dev_info(struct dmar_domain *domain) +{ + struct device_domain_info *info; + struct intel_iommu *iommu; + unsigned long flags1, flags2; + + spin_lock_irqsave(&device_domain_lock, flags1); + while (!list_empty(&domain->devices)) { + info = list_entry(domain->devices.next, + struct device_domain_info, link); + list_del(&info->link); + list_del(&info->global); + if (info->dev) + info->dev->dev.archdata.iommu = NULL; + + spin_unlock_irqrestore(&device_domain_lock, flags1); + + iommu = device_to_iommu(info->bus, info->devfn); + iommu_detach_dev(iommu, info->bus, info->devfn); + + /* clear this iommu in iommu_bmp, update iommu count + * and coherency + */ + spin_lock_irqsave(&domain->iommu_lock, flags2); + if (test_and_clear_bit(iommu->seq_id, + &domain->iommu_bmp)) { + domain->iommu_count--; + domain_update_iommu_coherency(domain); + } + spin_unlock_irqrestore(&domain->iommu_lock, flags2); + + free_devinfo_mem(info); + spin_lock_irqsave(&device_domain_lock, flags1); + } + spin_unlock_irqrestore(&device_domain_lock, flags1); +} + +/* domain id for virtual machine, it won't be set in context */ +static unsigned long vm_domid; + +static int vm_domain_min_agaw(struct dmar_domain *domain) +{ + int i; + int min_agaw = domain->agaw; + + i = find_first_bit(&domain->iommu_bmp, g_num_of_iommus); + for (; i < g_num_of_iommus; ) { + if (min_agaw > g_iommus[i]->agaw) + min_agaw = g_iommus[i]->agaw; + + i = find_next_bit(&domain->iommu_bmp, g_num_of_iommus, i+1); + } + + return min_agaw; +} + +static struct dmar_domain *iommu_alloc_vm_domain(void) +{ + struct dmar_domain *domain; + + domain = alloc_domain_mem(); + if (!domain) + return NULL; + + domain->id = vm_domid++; + memset(&domain->iommu_bmp, 0, sizeof(unsigned long)); + domain->flags = DOMAIN_FLAG_VIRTUAL_MACHINE; + + return domain; +} + +static int vm_domain_init(struct dmar_domain *domain, int guest_width) +{ + int adjust_width; + + init_iova_domain(&domain->iovad, DMA_32BIT_PFN); + spin_lock_init(&domain->mapping_lock); + spin_lock_init(&domain->iommu_lock); + + domain_reserve_special_ranges(domain); + + /* calculate AGAW */ + domain->gaw = guest_width; + adjust_width = guestwidth_to_adjustwidth(guest_width); + domain->agaw = width_to_agaw(adjust_width); + + INIT_LIST_HEAD(&domain->devices); + + domain->iommu_count = 0; + domain->iommu_coherency = 0; + domain->max_addr = 0; + + /* always allocate the top pgd */ + domain->pgd = (struct dma_pte *)alloc_pgtable_page(); + if (!domain->pgd) + return -ENOMEM; + domain_flush_cache(domain, domain->pgd, PAGE_SIZE); + return 0; +} + +static void iommu_free_vm_domain(struct dmar_domain *domain) +{ + unsigned long flags; + struct dmar_drhd_unit *drhd; + struct intel_iommu *iommu; + unsigned long i; + unsigned long ndomains; + + for_each_drhd_unit(drhd) { + if (drhd->ignored) + continue; + iommu = drhd->iommu; + + ndomains = cap_ndoms(iommu->cap); + i = find_first_bit(iommu->domain_ids, ndomains); + for (; i < ndomains; ) { + if (iommu->domains[i] == domain) { + spin_lock_irqsave(&iommu->lock, flags); + clear_bit(i, iommu->domain_ids); + iommu->domains[i] = NULL; + spin_unlock_irqrestore(&iommu->lock, flags); + break; + } + i = find_next_bit(iommu->domain_ids, ndomains, i+1); + } + } +} + +static void vm_domain_exit(struct dmar_domain *domain) { u64 end; @@ -2334,6 +2953,9 @@ void intel_iommu_domain_exit(struct dmar_domain *domain) if (!domain) return; + vm_domain_remove_all_dev_info(domain); + /* destroy iovas */ + put_iova_domain(&domain->iovad); end = DOMAIN_MAX_ADDR(domain->gaw); end = end & (~VTD_PAGE_MASK); @@ -2343,94 +2965,167 @@ void intel_iommu_domain_exit(struct dmar_domain *domain) /* free page tables */ dma_pte_free_pagetable(domain, 0, end); - iommu_free_domain(domain); + iommu_free_vm_domain(domain); free_domain_mem(domain); } -EXPORT_SYMBOL_GPL(intel_iommu_domain_exit); -struct dmar_domain *intel_iommu_domain_alloc(struct pci_dev *pdev) +static int intel_iommu_domain_init(struct iommu_domain *domain) { - struct dmar_drhd_unit *drhd; - struct dmar_domain *domain; - struct intel_iommu *iommu; - - drhd = dmar_find_matched_drhd_unit(pdev); - if (!drhd) { - printk(KERN_ERR "intel_iommu_domain_alloc: drhd == NULL\n"); - return NULL; - } + struct dmar_domain *dmar_domain; - iommu = drhd->iommu; - if (!iommu) { - printk(KERN_ERR - "intel_iommu_domain_alloc: iommu == NULL\n"); - return NULL; - } - domain = iommu_alloc_domain(iommu); - if (!domain) { + dmar_domain = iommu_alloc_vm_domain(); + if (!dmar_domain) { printk(KERN_ERR - "intel_iommu_domain_alloc: domain == NULL\n"); - return NULL; + "intel_iommu_domain_init: dmar_domain == NULL\n"); + return -ENOMEM; } - if (domain_init(domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) { + if (vm_domain_init(dmar_domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) { printk(KERN_ERR - "intel_iommu_domain_alloc: domain_init() failed\n"); - intel_iommu_domain_exit(domain); - return NULL; + "intel_iommu_domain_init() failed\n"); + vm_domain_exit(dmar_domain); + return -ENOMEM; } - return domain; + domain->priv = dmar_domain; + + return 0; } -EXPORT_SYMBOL_GPL(intel_iommu_domain_alloc); -int intel_iommu_context_mapping( - struct dmar_domain *domain, struct pci_dev *pdev) +static void intel_iommu_domain_destroy(struct iommu_domain *domain) { - int rc; - rc = domain_context_mapping(domain, pdev); - return rc; + struct dmar_domain *dmar_domain = domain->priv; + + domain->priv = NULL; + vm_domain_exit(dmar_domain); } -EXPORT_SYMBOL_GPL(intel_iommu_context_mapping); -int intel_iommu_page_mapping( - struct dmar_domain *domain, dma_addr_t iova, - u64 hpa, size_t size, int prot) +static int intel_iommu_attach_device(struct iommu_domain *domain, + struct device *dev) { - int rc; - rc = domain_page_mapping(domain, iova, hpa, size, prot); - return rc; + struct dmar_domain *dmar_domain = domain->priv; + struct pci_dev *pdev = to_pci_dev(dev); + struct intel_iommu *iommu; + int addr_width; + u64 end; + int ret; + + /* normally pdev is not mapped */ + if (unlikely(domain_context_mapped(pdev))) { + struct dmar_domain *old_domain; + + old_domain = find_domain(pdev); + if (old_domain) { + if (dmar_domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) + vm_domain_remove_one_dev_info(old_domain, pdev); + else + domain_remove_dev_info(old_domain); + } + } + + iommu = device_to_iommu(pdev->bus->number, pdev->devfn); + if (!iommu) + return -ENODEV; + + /* check if this iommu agaw is sufficient for max mapped address */ + addr_width = agaw_to_width(iommu->agaw); + end = DOMAIN_MAX_ADDR(addr_width); + end = end & VTD_PAGE_MASK; + if (end < dmar_domain->max_addr) { + printk(KERN_ERR "%s: iommu agaw (%d) is not " + "sufficient for the mapped address (%llx)\n", + __func__, iommu->agaw, dmar_domain->max_addr); + return -EFAULT; + } + + ret = domain_context_mapping(dmar_domain, pdev); + if (ret) + return ret; + + ret = vm_domain_add_dev_info(dmar_domain, pdev); + return ret; } -EXPORT_SYMBOL_GPL(intel_iommu_page_mapping); -void intel_iommu_detach_dev(struct dmar_domain *domain, u8 bus, u8 devfn) +static void intel_iommu_detach_device(struct iommu_domain *domain, + struct device *dev) { - detach_domain_for_dev(domain, bus, devfn); + struct dmar_domain *dmar_domain = domain->priv; + struct pci_dev *pdev = to_pci_dev(dev); + + vm_domain_remove_one_dev_info(dmar_domain, pdev); } -EXPORT_SYMBOL_GPL(intel_iommu_detach_dev); -struct dmar_domain * -intel_iommu_find_domain(struct pci_dev *pdev) +static int intel_iommu_map_range(struct iommu_domain *domain, + unsigned long iova, phys_addr_t hpa, + size_t size, int iommu_prot) { - return find_domain(pdev); + struct dmar_domain *dmar_domain = domain->priv; + u64 max_addr; + int addr_width; + int prot = 0; + int ret; + + if (iommu_prot & IOMMU_READ) + prot |= DMA_PTE_READ; + if (iommu_prot & IOMMU_WRITE) + prot |= DMA_PTE_WRITE; + + max_addr = (iova & VTD_PAGE_MASK) + VTD_PAGE_ALIGN(size); + if (dmar_domain->max_addr < max_addr) { + int min_agaw; + u64 end; + + /* check if minimum agaw is sufficient for mapped address */ + min_agaw = vm_domain_min_agaw(dmar_domain); + addr_width = agaw_to_width(min_agaw); + end = DOMAIN_MAX_ADDR(addr_width); + end = end & VTD_PAGE_MASK; + if (end < max_addr) { + printk(KERN_ERR "%s: iommu agaw (%d) is not " + "sufficient for the mapped address (%llx)\n", + __func__, min_agaw, max_addr); + return -EFAULT; + } + dmar_domain->max_addr = max_addr; + } + + ret = domain_page_mapping(dmar_domain, iova, hpa, size, prot); + return ret; } -EXPORT_SYMBOL_GPL(intel_iommu_find_domain); -int intel_iommu_found(void) +static void intel_iommu_unmap_range(struct iommu_domain *domain, + unsigned long iova, size_t size) { - return g_num_of_iommus; + struct dmar_domain *dmar_domain = domain->priv; + dma_addr_t base; + + /* The address might not be aligned */ + base = iova & VTD_PAGE_MASK; + size = VTD_PAGE_ALIGN(size); + dma_pte_clear_range(dmar_domain, base, base + size); + + if (dmar_domain->max_addr == base + size) + dmar_domain->max_addr = base; } -EXPORT_SYMBOL_GPL(intel_iommu_found); -u64 intel_iommu_iova_to_pfn(struct dmar_domain *domain, u64 iova) +static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain, + unsigned long iova) { + struct dmar_domain *dmar_domain = domain->priv; struct dma_pte *pte; - u64 pfn; - - pfn = 0; - pte = addr_to_dma_pte(domain, iova); + u64 phys = 0; + pte = addr_to_dma_pte(dmar_domain, iova); if (pte) - pfn = dma_pte_addr(*pte); + phys = dma_pte_addr(pte); - return pfn >> VTD_PAGE_SHIFT; + return phys; } -EXPORT_SYMBOL_GPL(intel_iommu_iova_to_pfn); + +static struct iommu_ops intel_iommu_ops = { + .domain_init = intel_iommu_domain_init, + .domain_destroy = intel_iommu_domain_destroy, + .attach_dev = intel_iommu_attach_device, + .detach_dev = intel_iommu_detach_device, + .map = intel_iommu_map_range, + .unmap = intel_iommu_unmap_range, + .iova_to_phys = intel_iommu_iova_to_phys, +}; -- cgit v1.2.3