#include <linux/iommu.h>
#include <linux/intel-iommu.h>
#include <linux/sysdev.h>
+#include <linux/tboot.h>
#include <asm/cacheflush.h>
#include <asm/iommu.h>
#include "pci.h"
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.
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++;
}
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);
}
+/* 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)
}
while (nr_pages--) {
+ uint64_t tmp;
+
if (!sg_res) {
- sg_res = (sg->offset + sg->length + VTD_PAGE_SIZE - 1) >> VTD_PAGE_SHIFT;
+ 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;
/* We don't need lock here, nobody else
* touches the iova range
*/
- if (unlikely(dma_pte_addr(pte))) {
+ 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)\n",
- iov_pfn, pte->val);
+ 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->val = pteval;
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;
return 0;
}
+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 iommu_prepare_static_identity_mapping(void)
{
struct pci_dev *pdev = NULL;
return -EFAULT;
for_each_pci_dev(pdev) {
- printk(KERN_INFO "IOMMU: identity mapping for device %s\n",
- pci_name(pdev));
+ if (iommu_should_identity_map(pdev, 1)) {
+ 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;
+ 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;
+ }
}
return 0;
* identity mapping if iommu_identity_mapping is set.
*/
if (!iommu_pass_through) {
+#ifdef CONFIG_DMAR_BROKEN_GFX_WA
+ if (!iommu_identity_mapping)
+ iommu_identity_mapping = 2;
+#endif
if (iommu_identity_mapping)
iommu_prepare_static_identity_mapping();
/*
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;
-}
-
+/* 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)
}
/* 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);
+ if (ret)
+ return 0;
+ ret = domain_context_mapping(si_domain, pdev, 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, 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);
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);
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, 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;
start_vpfn = mm_to_dma_pfn(iova->pfn_lo);
- ret = domain_sg_mapping(domain, start_vpfn, sglist, mm_to_dma_pfn(size), prot);
+ ret = domain_sg_mapping(domain, start_vpfn, sglist, size, prot);
if (unlikely(ret)) {
/* clear the page */
dma_pte_clear_range(domain, start_vpfn,
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.
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();
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 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