#include <linux/iommu.h>
#include <linux/intel-iommu.h>
#include <linux/sysdev.h>
+#include <linux/dmi.h>
#include <asm/cacheflush.h>
#include <asm/iommu.h>
#include "pci.h"
#define MAX_AGAW_WIDTH 64
-#define DOMAIN_MAX_ADDR(gaw) ((((u64)1) << gaw) - 1)
-#define DOMAIN_MAX_PFN(gaw) ((((u64)1) << (gaw-VTD_PAGE_SHIFT)) - 1)
+#define __DOMAIN_MAX_PFN(gaw) ((((uint64_t)1) << (gaw-VTD_PAGE_SHIFT)) - 1)
+#define __DOMAIN_MAX_ADDR(gaw) ((((uint64_t)1) << gaw) - 1)
+
+/* We limit DOMAIN_MAX_PFN to fit in an unsigned long, and DOMAIN_MAX_ADDR
+ to match. That way, we can use 'unsigned long' for PFNs with impunity. */
+#define DOMAIN_MAX_PFN(gaw) ((unsigned long) min_t(uint64_t, \
+ __DOMAIN_MAX_PFN(gaw), (unsigned long)-1))
+#define DOMAIN_MAX_ADDR(gaw) (((uint64_t)__DOMAIN_MAX_PFN(gaw)) << VTD_PAGE_SHIFT)
#define IOVA_PFN(addr) ((addr) >> PAGE_SHIFT)
#define DMA_32BIT_PFN IOVA_PFN(DMA_BIT_MASK(32))
static inline u64 dma_pte_addr(struct dma_pte *pte)
{
- return (pte->val & VTD_PAGE_MASK);
+#ifdef CONFIG_64BIT
+ return pte->val & VTD_PAGE_MASK;
+#else
+ /* Must have a full atomic 64-bit read */
+ return __cmpxchg64(pte, 0ULL, 0ULL) & VTD_PAGE_MASK;
+#endif
}
static inline void dma_set_pte_pfn(struct dma_pte *pte, unsigned long pfn)
return (pte->val & 3) != 0;
}
+static inline int first_pte_in_page(struct dma_pte *pte)
+{
+ return !((unsigned long)pte & ~VTD_PAGE_MASK);
+}
+
/*
* This domain is a statically identity mapping domain.
* 1. This domain creats a static 1:1 mapping to all usable memory.
* 2. It maps to each iommu if successful.
* 3. Each iommu mapps to this domain if successful.
*/
-struct dmar_domain *si_domain;
+static struct dmar_domain *si_domain;
+static int hw_pass_through = 1;
/* devices under the same p2p bridge are owned in one domain */
#define DOMAIN_FLAG_P2P_MULTIPLE_DEVICES (1 << 0)
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 */
struct dma_pte *parent, *pte = NULL;
int level = agaw_to_level(domain->agaw);
int offset;
- unsigned long flags;
BUG_ON(!domain->pgd);
BUG_ON(addr_width < BITS_PER_LONG && pfn >> addr_width);
parent = domain->pgd;
- spin_lock_irqsave(&domain->mapping_lock, flags);
while (level > 0) {
void *tmp_page;
break;
if (!dma_pte_present(pte)) {
+ uint64_t pteval;
+
tmp_page = alloc_pgtable_page();
- if (!tmp_page) {
- spin_unlock_irqrestore(&domain->mapping_lock,
- flags);
+ if (!tmp_page)
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;
+ if (cmpxchg64(&pte->val, 0ULL, pteval)) {
+ /* Someone else set it while we were thinking; use theirs. */
+ free_pgtable_page(tmp_page);
+ } else {
+ dma_pte_addr(pte);
+ domain_flush_cache(domain, pte, sizeof(*pte));
}
- domain_flush_cache(domain, tmp_page, PAGE_SIZE);
- dma_set_pte_pfn(pte, virt_to_dma_pfn(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);
- domain_flush_cache(domain, pte, sizeof(*pte));
}
parent = phys_to_virt(dma_pte_addr(pte));
level--;
}
- spin_unlock_irqrestore(&domain->mapping_lock, flags);
return pte;
}
start_pfn = align_to_level(start_pfn + 1, 2);
continue;
}
- while (start_pfn <= last_pfn &&
- (unsigned long)pte >> VTD_PAGE_SHIFT ==
- (unsigned long)first_pte >> VTD_PAGE_SHIFT) {
+ do {
dma_clear_pte(pte);
start_pfn++;
pte++;
- }
+ } while (start_pfn <= last_pfn && !first_pte_in_page(pte));
+
domain_flush_cache(domain, first_pte,
(void *)pte - (void *)first_pte);
}
unsigned long last_pfn)
{
int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
- struct dma_pte *pte;
+ struct dma_pte *first_pte, *pte;
int total = agaw_to_level(domain->agaw);
int level;
unsigned long tmp;
BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
- /* we don't need lock here, nobody else touches the iova range */
+ /* We don't need lock here; nobody else touches the iova range */
level = 2;
while (level <= total) {
tmp = align_to_level(start_pfn, level);
- /* Only clear this pte/pmd if we're asked to clear its
- _whole_ range */
+ /* If we can't even clear one PTE at this level, we're done */
if (tmp + level_size(level) - 1 > last_pfn)
return;
- while (tmp <= last_pfn) {
- pte = dma_pfn_level_pte(domain, tmp, level);
- if (pte) {
- free_pgtable_page(
- phys_to_virt(dma_pte_addr(pte)));
- dma_clear_pte(pte);
- domain_flush_cache(domain, pte, sizeof(*pte));
+ while (tmp + level_size(level) - 1 <= last_pfn) {
+ first_pte = pte = dma_pfn_level_pte(domain, tmp, level);
+ if (!pte) {
+ tmp = align_to_level(tmp + 1, level + 1);
+ continue;
}
- tmp += level_size(level);
+ do {
+ if (dma_pte_present(pte)) {
+ free_pgtable_page(phys_to_virt(dma_pte_addr(pte)));
+ dma_clear_pte(pte);
+ }
+ pte++;
+ tmp += level_size(level);
+ } while (!first_pte_in_page(pte) &&
+ tmp + level_size(level) - 1 <= last_pfn);
+
+ domain_flush_cache(domain, first_pte,
+ (void *)pte - (void *)first_pte);
+
}
level++;
}
pr_debug("Number of Domains supportd <%ld>\n", ndomains);
nlongs = BITS_TO_LONGS(ndomains);
+ spin_lock_init(&iommu->lock);
+
/* TBD: there might be 64K domains,
* consider other allocation for future chip
*/
GFP_KERNEL);
if (!iommu->domains) {
printk(KERN_ERR "Allocating domain array failed\n");
- kfree(iommu->domain_ids);
return -ENOMEM;
}
- spin_lock_init(&iommu->lock);
-
/*
* if Caching mode is set, then invalid translations are tagged
* with domainid 0. Hence we need to pre-allocate it.
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);
+ if ((iommu->domains) && (iommu->domain_ids)) {
+ 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);
+
+ 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);
- 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);
+ i = find_next_bit(iommu->domain_ids,
+ cap_ndoms(iommu->cap), i+1);
}
- spin_unlock_irqrestore(&domain->iommu_lock, flags);
-
- i = find_next_bit(iommu->domain_ids,
- cap_ndoms(iommu->cap), i+1);
}
if (iommu->gcmd & DMA_GCMD_TE)
}
static struct iova_domain reserved_iova_list;
-static struct lock_class_key reserved_alloc_key;
static struct lock_class_key reserved_rbtree_key;
static void dmar_init_reserved_ranges(void)
init_iova_domain(&reserved_iova_list, DMA_32BIT_PFN);
- lockdep_set_class(&reserved_iova_list.iova_alloc_lock,
- &reserved_alloc_key);
lockdep_set_class(&reserved_iova_list.iova_rbtree_lock,
&reserved_rbtree_key);
unsigned long sagaw;
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);
}
set_bit(num, iommu->domain_ids);
- set_bit(iommu->seq_id, &domain->iommu_bmp);
iommu->domains[num] = domain;
id = num;
}
tmp->devfn);
}
-static int domain_pfn_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
- unsigned long phys_pfn, unsigned long nr_pages,
- int prot)
+/* Returns a number of VTD pages, but aligned to MM page size */
+static inline unsigned long aligned_nrpages(unsigned long host_addr,
+ size_t size)
+{
+ host_addr &= ~PAGE_MASK;
+ return PAGE_ALIGN(host_addr + size) >> VTD_PAGE_SHIFT;
+}
+
+static int __domain_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+ struct scatterlist *sg, unsigned long phys_pfn,
+ unsigned long nr_pages, int prot)
{
struct dma_pte *first_pte = NULL, *pte = NULL;
+ phys_addr_t uninitialized_var(pteval);
int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
+ unsigned long sg_res;
BUG_ON(addr_width < BITS_PER_LONG && (iov_pfn + nr_pages - 1) >> addr_width);
prot &= DMA_PTE_READ | DMA_PTE_WRITE | DMA_PTE_SNP;
+ if (sg)
+ sg_res = 0;
+ else {
+ sg_res = nr_pages + 1;
+ pteval = ((phys_addr_t)phys_pfn << VTD_PAGE_SHIFT) | prot;
+ }
+
while (nr_pages--) {
+ uint64_t tmp;
+
+ if (!sg_res) {
+ sg_res = aligned_nrpages(sg->offset, sg->length);
+ sg->dma_address = ((dma_addr_t)iov_pfn << VTD_PAGE_SHIFT) + sg->offset;
+ sg->dma_length = sg->length;
+ pteval = page_to_phys(sg_page(sg)) | prot;
+ }
if (!pte) {
first_pte = pte = pfn_to_dma_pte(domain, iov_pfn);
if (!pte)
/* We don't need lock here, nobody else
* touches the iova range
*/
- BUG_ON(dma_pte_addr(pte));
- pte->val = (phys_pfn << VTD_PAGE_SHIFT) | prot;
+ tmp = cmpxchg64_local(&pte->val, 0ULL, pteval);
+ if (tmp) {
+ static int dumps = 5;
+ printk(KERN_CRIT "ERROR: DMA PTE for vPFN 0x%lx already set (to %llx not %llx)\n",
+ iov_pfn, tmp, (unsigned long long)pteval);
+ if (dumps) {
+ dumps--;
+ debug_dma_dump_mappings(NULL);
+ }
+ WARN_ON(1);
+ }
pte++;
- if (!nr_pages ||
- (unsigned long)pte >> VTD_PAGE_SHIFT !=
- (unsigned long)first_pte >> VTD_PAGE_SHIFT) {
+ if (!nr_pages || first_pte_in_page(pte)) {
domain_flush_cache(domain, first_pte,
(void *)pte - (void *)first_pte);
pte = NULL;
}
iov_pfn++;
- phys_pfn++;
+ pteval += VTD_PAGE_SIZE;
+ sg_res--;
+ if (!sg_res)
+ sg = sg_next(sg);
}
return 0;
}
+static inline int domain_sg_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+ struct scatterlist *sg, unsigned long nr_pages,
+ int prot)
+{
+ return __domain_mapping(domain, iov_pfn, sg, 0, nr_pages, prot);
+}
+
+static inline int domain_pfn_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+ unsigned long phys_pfn, unsigned long nr_pages,
+ int prot)
+{
+ return __domain_mapping(domain, iov_pfn, NULL, phys_pfn, nr_pages, prot);
+}
+
static void iommu_detach_dev(struct intel_iommu *iommu, u8 bus, u8 devfn)
{
if (!iommu)
struct dmar_domain *domain;
int ret;
- printk(KERN_INFO
- "IOMMU: Setting identity map for device %s [0x%Lx - 0x%Lx]\n",
- pci_name(pdev), start, end);
-
domain = get_domain_for_dev(pdev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
if (!domain)
return -ENOMEM;
+ /* For _hardware_ passthrough, don't bother. But for software
+ passthrough, we do it anyway -- it may indicate a memory
+ range which is reserved in E820, so which didn't get set
+ up to start with in si_domain */
+ if (domain == si_domain && hw_pass_through) {
+ printk("Ignoring identity map for HW passthrough device %s [0x%Lx - 0x%Lx]\n",
+ pci_name(pdev), start, end);
+ return 0;
+ }
+
+ printk(KERN_INFO
+ "IOMMU: Setting identity map for device %s [0x%Lx - 0x%Lx]\n",
+ pci_name(pdev), start, end);
+
+ if (end >> agaw_to_width(domain->agaw)) {
+ WARN(1, "Your BIOS is broken; RMRR exceeds permitted address width (%d bits)\n"
+ "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
+ agaw_to_width(domain->agaw),
+ dmi_get_system_info(DMI_BIOS_VENDOR),
+ dmi_get_system_info(DMI_BIOS_VERSION),
+ dmi_get_system_info(DMI_PRODUCT_VERSION));
+ ret = -EIO;
+ goto error;
+ }
+
ret = iommu_domain_identity_map(domain, start, end);
if (ret)
goto error;
}
#endif /* !CONFIG_DMAR_FLPY_WA */
-/* Initialize each context entry as pass through.*/
-static int __init init_context_pass_through(void)
-{
- struct pci_dev *pdev = NULL;
- struct dmar_domain *domain;
- int ret;
-
- for_each_pci_dev(pdev) {
- domain = get_domain_for_dev(pdev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
- ret = domain_context_mapping(domain, pdev,
- CONTEXT_TT_PASS_THROUGH);
- if (ret)
- return ret;
- }
- return 0;
-}
-
static int md_domain_init(struct dmar_domain *domain, int guest_width);
static int __init si_domain_work_fn(unsigned long start_pfn,
}
-static int si_domain_init(void)
+static int __init si_domain_init(int hw)
{
struct dmar_drhd_unit *drhd;
struct intel_iommu *iommu;
si_domain->flags = DOMAIN_FLAG_STATIC_IDENTITY;
+ if (hw)
+ return 0;
+
for_each_online_node(nid) {
work_with_active_regions(nid, si_domain_work_fn, &ret);
if (ret)
}
static int domain_add_dev_info(struct dmar_domain *domain,
- struct pci_dev *pdev)
+ struct pci_dev *pdev,
+ int translation)
{
struct device_domain_info *info;
unsigned long flags;
+ int ret;
info = alloc_devinfo_mem();
if (!info)
return -ENOMEM;
+ ret = domain_context_mapping(domain, pdev, translation);
+ if (ret) {
+ free_devinfo_mem(info);
+ return ret;
+ }
+
info->segment = pci_domain_nr(pdev->bus);
info->bus = pdev->bus->number;
info->devfn = pdev->devfn;
return 0;
}
-static int iommu_prepare_static_identity_mapping(void)
+static int iommu_should_identity_map(struct pci_dev *pdev, int startup)
+{
+ if (iommu_identity_mapping == 2)
+ return IS_GFX_DEVICE(pdev);
+
+ /*
+ * We want to start off with all devices in the 1:1 domain, and
+ * take them out later if we find they can't access all of memory.
+ *
+ * However, we can't do this for PCI devices behind bridges,
+ * because all PCI devices behind the same bridge will end up
+ * with the same source-id on their transactions.
+ *
+ * Practically speaking, we can't change things around for these
+ * devices at run-time, because we can't be sure there'll be no
+ * DMA transactions in flight for any of their siblings.
+ *
+ * So PCI devices (unless they're on the root bus) as well as
+ * their parent PCI-PCI or PCIe-PCI bridges must be left _out_ of
+ * 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_root_bus(pdev->bus))
+ return 0;
+ if (pdev->class >> 8 == PCI_CLASS_BRIDGE_PCI)
+ return 0;
+ } else if (pdev->pcie_type == PCI_EXP_TYPE_PCI_BRIDGE)
+ return 0;
+
+ /*
+ * At boot time, we don't yet know if devices will be 64-bit capable.
+ * Assume that they will -- if they turn out not to be, then we can
+ * take them out of the 1:1 domain later.
+ */
+ if (!startup)
+ return pdev->dma_mask > DMA_BIT_MASK(32);
+
+ return 1;
+}
+
+static int __init iommu_prepare_static_identity_mapping(int hw)
{
struct pci_dev *pdev = NULL;
int ret;
- ret = si_domain_init();
+ ret = si_domain_init(hw);
if (ret)
return -EFAULT;
for_each_pci_dev(pdev) {
- printk(KERN_INFO "IOMMU: identity mapping for device %s\n",
- pci_name(pdev));
-
- ret = domain_context_mapping(si_domain, pdev,
- CONTEXT_TT_MULTI_LEVEL);
- if (ret)
- return ret;
- ret = domain_add_dev_info(si_domain, pdev);
- if (ret)
- return ret;
+ if (iommu_should_identity_map(pdev, 1)) {
+ printk(KERN_INFO "IOMMU: %s identity mapping for device %s\n",
+ hw ? "hardware" : "software", pci_name(pdev));
+
+ ret = domain_add_dev_info(si_domain, pdev,
+ hw ? CONTEXT_TT_PASS_THROUGH :
+ CONTEXT_TT_MULTI_LEVEL);
+ if (ret)
+ return ret;
+ }
}
return 0;
struct pci_dev *pdev;
struct intel_iommu *iommu;
int i, ret;
- int pass_through = 1;
-
- /*
- * In case pass through can not be enabled, iommu tries to use identity
- * mapping.
- */
- if (iommu_pass_through)
- iommu_identity_mapping = 1;
/*
* for each drhd
deferred_flush = kzalloc(g_num_of_iommus *
sizeof(struct deferred_flush_tables), GFP_KERNEL);
if (!deferred_flush) {
- kfree(g_iommus);
ret = -ENOMEM;
goto error;
}
goto error;
}
if (!ecap_pass_through(iommu->ecap))
- pass_through = 0;
+ hw_pass_through = 0;
}
- if (iommu_pass_through)
- if (!pass_through) {
- printk(KERN_INFO
- "Pass Through is not supported by hardware.\n");
- iommu_pass_through = 0;
- }
/*
* Start from the sane iommu hardware state.
}
}
+ if (iommu_pass_through)
+ iommu_identity_mapping = 1;
+#ifdef CONFIG_DMAR_BROKEN_GFX_WA
+ else
+ iommu_identity_mapping = 2;
+#endif
/*
- * If pass through is set and enabled, context entries of all pci
- * devices are intialized by pass through translation type.
+ * 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
+ * identity mapping if iommu_identity_mapping is set.
*/
- if (iommu_pass_through) {
- ret = init_context_pass_through();
+ if (iommu_identity_mapping) {
+ ret = iommu_prepare_static_identity_mapping(hw_pass_through);
if (ret) {
- printk(KERN_ERR "IOMMU: Pass through init failed.\n");
- iommu_pass_through = 0;
+ printk(KERN_CRIT "Failed to setup IOMMU pass-through\n");
+ goto error;
}
}
-
/*
- * 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
- * identity mapping if iommu_identity_mapping is set.
+ * For each rmrr
+ * for each dev attached to rmrr
+ * do
+ * locate drhd for dev, alloc domain for dev
+ * allocate free domain
+ * allocate page table entries for rmrr
+ * if context not allocated for bus
+ * allocate and init context
+ * set present in root table for this bus
+ * init context with domain, translation etc
+ * endfor
+ * endfor
*/
- if (!iommu_pass_through) {
- if (iommu_identity_mapping)
- iommu_prepare_static_identity_mapping();
- /*
- * For each rmrr
- * for each dev attached to rmrr
- * do
- * locate drhd for dev, alloc domain for dev
- * allocate free domain
- * allocate page table entries for rmrr
- * if context not allocated for bus
- * allocate and init context
- * set present in root table for this bus
- * init context with domain, translation etc
- * endfor
- * endfor
- */
- printk(KERN_INFO "IOMMU: Setting RMRR:\n");
- for_each_rmrr_units(rmrr) {
- for (i = 0; i < rmrr->devices_cnt; i++) {
- pdev = rmrr->devices[i];
- /*
- * some BIOS lists non-exist devices in DMAR
- * table.
- */
- if (!pdev)
- continue;
- ret = iommu_prepare_rmrr_dev(rmrr, pdev);
- if (ret)
- printk(KERN_ERR
- "IOMMU: mapping reserved region failed\n");
- }
+ printk(KERN_INFO "IOMMU: Setting RMRR:\n");
+ for_each_rmrr_units(rmrr) {
+ for (i = 0; i < rmrr->devices_cnt; i++) {
+ pdev = rmrr->devices[i];
+ /*
+ * some BIOS lists non-exist devices in DMAR
+ * table.
+ */
+ if (!pdev)
+ continue;
+ ret = iommu_prepare_rmrr_dev(rmrr, pdev);
+ if (ret)
+ printk(KERN_ERR
+ "IOMMU: mapping reserved region failed\n");
}
-
- iommu_prepare_isa();
}
+ iommu_prepare_isa();
+
/*
* for each drhd
* enable fault log
return ret;
}
-static inline unsigned long aligned_nrpages(unsigned long host_addr,
- size_t size)
-{
- host_addr &= ~PAGE_MASK;
- host_addr += size + PAGE_SIZE - 1;
-
- return host_addr >> VTD_PAGE_SHIFT;
-}
-
-struct iova *
-iommu_alloc_iova(struct dmar_domain *domain, size_t size, u64 end)
-{
- struct iova *piova;
-
- /* Make sure it's in range */
- end = min_t(u64, DOMAIN_MAX_ADDR(domain->gaw), end);
- if (!size || (IOVA_START_ADDR + size > end))
- return NULL;
-
- piova = alloc_iova(&domain->iovad,
- size >> PAGE_SHIFT, IOVA_PFN(end), 1);
- return piova;
-}
-
-static struct iova *
-__intel_alloc_iova(struct device *dev, struct dmar_domain *domain,
- size_t size, u64 dma_mask)
+/* This takes a number of _MM_ pages, not VTD pages */
+static struct iova *intel_alloc_iova(struct device *dev,
+ struct dmar_domain *domain,
+ unsigned long nrpages, uint64_t dma_mask)
{
struct pci_dev *pdev = to_pci_dev(dev);
struct iova *iova = NULL;
- if (dma_mask <= DMA_BIT_MASK(32) || dmar_forcedac)
- iova = iommu_alloc_iova(domain, size, dma_mask);
- else {
+ /* Restrict dma_mask to the width that the iommu can handle */
+ dma_mask = min_t(uint64_t, DOMAIN_MAX_ADDR(domain->gaw), dma_mask);
+
+ if (!dmar_forcedac && dma_mask > DMA_BIT_MASK(32)) {
/*
* First try to allocate an io virtual address in
* DMA_BIT_MASK(32) and if that fails then try allocating
* from higher range
*/
- iova = iommu_alloc_iova(domain, size, DMA_BIT_MASK(32));
- if (!iova)
- iova = iommu_alloc_iova(domain, size, dma_mask);
- }
-
- if (!iova) {
- printk(KERN_ERR"Allocating iova for %s failed", pci_name(pdev));
+ iova = alloc_iova(&domain->iovad, nrpages,
+ IOVA_PFN(DMA_BIT_MASK(32)), 1);
+ if (iova)
+ return iova;
+ }
+ iova = alloc_iova(&domain->iovad, nrpages, IOVA_PFN(dma_mask), 1);
+ if (unlikely(!iova)) {
+ printk(KERN_ERR "Allocating %ld-page iova for %s failed",
+ nrpages, pci_name(pdev));
return NULL;
}
return iova;
}
-static struct dmar_domain *
-get_valid_domain_for_dev(struct pci_dev *pdev)
+static struct dmar_domain *__get_valid_domain_for_dev(struct pci_dev *pdev)
{
struct dmar_domain *domain;
int ret;
return domain;
}
+static inline struct dmar_domain *get_valid_domain_for_dev(struct pci_dev *dev)
+{
+ struct device_domain_info *info;
+
+ /* No lock here, assumes no domain exit in normal case */
+ info = dev->dev.archdata.iommu;
+ if (likely(info))
+ return info->domain;
+
+ return __get_valid_domain_for_dev(dev);
+}
+
static int iommu_dummy(struct pci_dev *pdev)
{
return pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO;
}
/* Check if the pdev needs to go through non-identity map and unmap process.*/
-static int iommu_no_mapping(struct pci_dev *pdev)
+static int iommu_no_mapping(struct device *dev)
{
+ struct pci_dev *pdev;
int found;
+ if (unlikely(dev->bus != &pci_bus_type))
+ return 1;
+
+ pdev = to_pci_dev(dev);
+ if (iommu_dummy(pdev))
+ return 1;
+
if (!iommu_identity_mapping)
- return iommu_dummy(pdev);
+ return 0;
found = identity_mapping(pdev);
if (found) {
- if (pdev->dma_mask > DMA_BIT_MASK(32))
+ if (iommu_should_identity_map(pdev, 0))
return 1;
else {
/*
* In case of a detached 64 bit DMA device from vm, the device
* is put into si_domain for identity mapping.
*/
- if (pdev->dma_mask > DMA_BIT_MASK(32)) {
+ if (iommu_should_identity_map(pdev, 0)) {
int ret;
- ret = domain_add_dev_info(si_domain, pdev);
+ ret = domain_add_dev_info(si_domain, pdev,
+ hw_pass_through ?
+ CONTEXT_TT_PASS_THROUGH :
+ CONTEXT_TT_MULTI_LEVEL);
if (!ret) {
printk(KERN_INFO "64bit %s uses identity mapping\n",
pci_name(pdev));
}
}
- return iommu_dummy(pdev);
+ return 0;
}
static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
int prot = 0;
int ret;
struct intel_iommu *iommu;
+ unsigned long paddr_pfn = paddr >> PAGE_SHIFT;
BUG_ON(dir == DMA_NONE);
- if (iommu_no_mapping(pdev))
+ if (iommu_no_mapping(hwdev))
return paddr;
domain = get_valid_domain_for_dev(pdev);
iommu = domain_get_iommu(domain);
size = aligned_nrpages(paddr, size);
- iova = __intel_alloc_iova(hwdev, domain, size << VTD_PAGE_SHIFT, pdev->dma_mask);
+ iova = intel_alloc_iova(hwdev, domain, dma_to_mm_pfn(size),
+ pdev->dma_mask);
if (!iova)
goto error;
* is not a big problem
*/
ret = domain_pfn_mapping(domain, mm_to_dma_pfn(iova->pfn_lo),
- paddr >> VTD_PAGE_SHIFT, size, prot);
+ mm_to_dma_pfn(paddr_pfn), size, prot);
if (ret)
goto error;
struct iova *iova;
struct intel_iommu *iommu;
- if (iommu_no_mapping(pdev))
+ if (iommu_no_mapping(dev))
return;
domain = find_domain(pdev);
iommu = domain_get_iommu(domain);
iova = find_iova(&domain->iovad, IOVA_PFN(dev_addr));
- if (!iova)
+ if (WARN_ONCE(!iova, "Driver unmaps unmatched page at PFN %llx\n",
+ (unsigned long long)dev_addr))
return;
start_pfn = mm_to_dma_pfn(iova->pfn_lo);
}
}
-static void intel_unmap_single(struct device *dev, dma_addr_t dev_addr, size_t size,
- int dir)
-{
- intel_unmap_page(dev, dev_addr, size, dir, NULL);
-}
-
static void *intel_alloc_coherent(struct device *hwdev, size_t size,
dma_addr_t *dma_handle, gfp_t flags)
{
size = PAGE_ALIGN(size);
order = get_order(size);
- intel_unmap_single(hwdev, dma_handle, size, DMA_BIDIRECTIONAL);
+ intel_unmap_page(hwdev, dma_handle, size, DMA_BIDIRECTIONAL, NULL);
free_pages((unsigned long)vaddr, order);
}
struct iova *iova;
struct intel_iommu *iommu;
- if (iommu_no_mapping(pdev))
+ if (iommu_no_mapping(hwdev))
return;
domain = find_domain(pdev);
iommu = domain_get_iommu(domain);
iova = find_iova(&domain->iovad, IOVA_PFN(sglist[0].dma_address));
- if (!iova)
+ if (WARN_ONCE(!iova, "Driver unmaps unmatched sglist at PFN %llx\n",
+ (unsigned long long)sglist[0].dma_address))
return;
start_pfn = mm_to_dma_pfn(iova->pfn_lo);
/* free page tables */
dma_pte_free_pagetable(domain, start_pfn, last_pfn);
- iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
- (last_pfn - start_pfn + 1));
-
- /* free iova */
- __free_iova(&domain->iovad, iova);
+ if (intel_iommu_strict) {
+ iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
+ last_pfn - start_pfn + 1);
+ /* free iova */
+ __free_iova(&domain->iovad, iova);
+ } else {
+ add_unmap(domain, iova);
+ /*
+ * queue up the release of the unmap to save the 1/6th of the
+ * cpu used up by the iotlb flush operation...
+ */
+ }
}
static int intel_nontranslate_map_sg(struct device *hddev,
struct intel_iommu *iommu;
BUG_ON(dir == DMA_NONE);
- if (iommu_no_mapping(pdev))
+ if (iommu_no_mapping(hwdev))
return intel_nontranslate_map_sg(hwdev, sglist, nelems, dir);
domain = get_valid_domain_for_dev(pdev);
for_each_sg(sglist, sg, nelems, i)
size += aligned_nrpages(sg->offset, sg->length);
- iova = __intel_alloc_iova(hwdev, domain, size << VTD_PAGE_SHIFT,
- pdev->dma_mask);
+ iova = intel_alloc_iova(hwdev, domain, dma_to_mm_pfn(size),
+ pdev->dma_mask);
if (!iova) {
sglist->dma_length = 0;
return 0;
prot |= DMA_PTE_WRITE;
start_vpfn = mm_to_dma_pfn(iova->pfn_lo);
- offset_pfn = 0;
- for_each_sg(sglist, sg, nelems, i) {
- int nr_pages = aligned_nrpages(sg->offset, sg->length);
- ret = domain_pfn_mapping(domain, start_vpfn + offset_pfn,
- page_to_dma_pfn(sg_page(sg)),
- nr_pages, prot);
- if (ret) {
- /* clear the page */
- dma_pte_clear_range(domain, start_vpfn,
- start_vpfn + offset_pfn);
- /* free page tables */
- dma_pte_free_pagetable(domain, start_vpfn,
- start_vpfn + offset_pfn);
- /* free iova */
- __free_iova(&domain->iovad, iova);
- return 0;
- }
- sg->dma_address = ((dma_addr_t)(start_vpfn + offset_pfn)
- << VTD_PAGE_SHIFT) + sg->offset;
- sg->dma_length = sg->length;
- offset_pfn += nr_pages;
+
+ ret = domain_sg_mapping(domain, start_vpfn, sglist, size, prot);
+ if (unlikely(ret)) {
+ /* clear the page */
+ dma_pte_clear_range(domain, start_vpfn,
+ start_vpfn + size - 1);
+ /* free page tables */
+ dma_pte_free_pagetable(domain, start_vpfn,
+ start_vpfn + size - 1);
+ /* free iova */
+ __free_iova(&domain->iovad, iova);
+ return 0;
}
/* it's a non-present to present mapping. Only flush if caching mode */
* Check the need for DMA-remapping initialization now.
* Above initialization will also be used by Interrupt-remapping.
*/
- if (no_iommu || (swiotlb && !iommu_pass_through) || dmar_disabled)
+ if (no_iommu || swiotlb || dmar_disabled)
return -ENODEV;
iommu_init_mempool();
init_timer(&unmap_timer);
force_iommu = 1;
-
- if (!iommu_pass_through) {
- printk(KERN_INFO
- "Multi-level page-table translation for DMAR.\n");
- dma_ops = &intel_dma_ops;
- } else
- printk(KERN_INFO
- "DMAR: Pass through translation for DMAR.\n");
+ dma_ops = &intel_dma_ops;
init_iommu_sysfs();
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);
domain->iommu_count = 0;
domain->iommu_coherency = 0;
+ domain->iommu_snooping = 0;
domain->max_addr = 0;
/* always allocate the top pgd */
struct intel_iommu *iommu;
int addr_width;
u64 end;
- int ret;
/* normally pdev is not mapped */
if (unlikely(domain_context_mapped(pdev))) {
return -EFAULT;
}
- ret = domain_add_dev_info(dmar_domain, pdev);
- if (ret)
- return ret;
-
- ret = domain_context_mapping(dmar_domain, pdev, CONTEXT_TT_MULTI_LEVEL);
- return ret;
+ return domain_add_dev_info(dmar_domain, pdev, CONTEXT_TT_MULTI_LEVEL);
}
static void intel_iommu_detach_device(struct iommu_domain *domain,
{
struct dmar_domain *dmar_domain = domain->priv;
+ if (!size)
+ return;
+
dma_pte_clear_range(dmar_domain, iova >> VTD_PAGE_SHIFT,
(iova + size - 1) >> VTD_PAGE_SHIFT);
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff