#include <asm/proto.h>
#include <asm/gart.h>
#include <asm/amd_iommu_types.h>
+#include <asm/amd_iommu.h>
#define CMD_SET_TYPE(cmd, t) ((cmd)->data[1] |= ((t) << 28))
static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
struct unity_map_entry *e);
+static int iommu_has_npcache(struct amd_iommu *iommu)
+{
+ return iommu->cap & IOMMU_CAP_NPCACHE;
+}
+
static int __iommu_queue_command(struct amd_iommu *iommu, struct command *cmd)
{
u32 tail, head;
static int iommu_flush_pages(struct amd_iommu *iommu, u16 domid,
u64 address, size_t size)
{
- int i;
+ int s = 0;
unsigned pages = to_pages(address, size);
address &= PAGE_MASK;
- for (i = 0; i < pages; ++i) {
- iommu_queue_inv_iommu_pages(iommu, address, domid, 0, 0);
- address += PAGE_SIZE;
+ if (pages > 1) {
+ /*
+ * If we have to flush more than one page, flush all
+ * TLB entries for this domain
+ */
+ address = CMD_INV_IOMMU_ALL_PAGES_ADDRESS;
+ s = 1;
}
+ iommu_queue_inv_iommu_pages(iommu, address, domid, 0, s);
+
return 0;
}
0, boundary_size, 0);
if (likely(address != -1)) {
- set_bit_string(dom->bitmap, address, pages);
dom->next_bit = address + pages;
address <<= PAGE_SHIFT;
} else
return NULL;
}
+static struct protection_domain *domain_for_device(u16 devid)
+{
+ struct protection_domain *dom;
+ unsigned long flags;
+
+ read_lock_irqsave(&amd_iommu_devtable_lock, flags);
+ dom = amd_iommu_pd_table[devid];
+ read_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+
+ return dom;
+}
+
+static void set_device_domain(struct amd_iommu *iommu,
+ struct protection_domain *domain,
+ u16 devid)
+{
+ unsigned long flags;
+
+ u64 pte_root = virt_to_phys(domain->pt_root);
+
+ pte_root |= (domain->mode & 0x07) << 9;
+ pte_root |= IOMMU_PTE_IR | IOMMU_PTE_IW | IOMMU_PTE_P | 2;
+
+ write_lock_irqsave(&amd_iommu_devtable_lock, flags);
+ amd_iommu_dev_table[devid].data[0] = pte_root;
+ amd_iommu_dev_table[devid].data[1] = pte_root >> 32;
+ amd_iommu_dev_table[devid].data[2] = domain->id;
+
+ amd_iommu_pd_table[devid] = domain;
+ write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+
+ iommu_queue_inv_dev_entry(iommu, devid);
+
+ iommu->need_sync = 1;
+}
+
+static int get_device_resources(struct device *dev,
+ struct amd_iommu **iommu,
+ struct protection_domain **domain,
+ u16 *bdf)
+{
+ struct dma_ops_domain *dma_dom;
+ struct pci_dev *pcidev;
+ u16 _bdf;
+
+ BUG_ON(!dev || dev->bus != &pci_bus_type || !dev->dma_mask);
+
+ pcidev = to_pci_dev(dev);
+ _bdf = (pcidev->bus->number << 8) | pcidev->devfn;
+
+ if (_bdf >= amd_iommu_last_bdf) {
+ *iommu = NULL;
+ *domain = NULL;
+ *bdf = 0xffff;
+ return 0;
+ }
+
+ *bdf = amd_iommu_alias_table[_bdf];
+
+ *iommu = amd_iommu_rlookup_table[*bdf];
+ if (*iommu == NULL)
+ return 0;
+ dma_dom = (*iommu)->default_dom;
+ *domain = domain_for_device(*bdf);
+ if (*domain == NULL) {
+ *domain = &dma_dom->domain;
+ set_device_domain(*iommu, *domain, *bdf);
+ printk(KERN_INFO "AMD IOMMU: Using protection domain %d for "
+ "device ", (*domain)->id);
+ print_devid(_bdf, 1);
+ }
+
+ return 1;
+}
+
+static dma_addr_t dma_ops_domain_map(struct amd_iommu *iommu,
+ struct dma_ops_domain *dom,
+ unsigned long address,
+ phys_addr_t paddr,
+ int direction)
+{
+ u64 *pte, __pte;
+
+ WARN_ON(address > dom->aperture_size);
+
+ paddr &= PAGE_MASK;
+
+ pte = dom->pte_pages[IOMMU_PTE_L1_INDEX(address)];
+ pte += IOMMU_PTE_L0_INDEX(address);
+
+ __pte = paddr | IOMMU_PTE_P | IOMMU_PTE_FC;
+
+ if (direction == DMA_TO_DEVICE)
+ __pte |= IOMMU_PTE_IR;
+ else if (direction == DMA_FROM_DEVICE)
+ __pte |= IOMMU_PTE_IW;
+ else if (direction == DMA_BIDIRECTIONAL)
+ __pte |= IOMMU_PTE_IR | IOMMU_PTE_IW;
+
+ WARN_ON(*pte);
+
+ *pte = __pte;
+
+ return (dma_addr_t)address;
+}
+
+static void dma_ops_domain_unmap(struct amd_iommu *iommu,
+ struct dma_ops_domain *dom,
+ unsigned long address)
+{
+ u64 *pte;
+
+ if (address >= dom->aperture_size)
+ return;
+
+ WARN_ON(address & 0xfffULL || address > dom->aperture_size);
+
+ pte = dom->pte_pages[IOMMU_PTE_L1_INDEX(address)];
+ pte += IOMMU_PTE_L0_INDEX(address);
+
+ WARN_ON(!*pte);
+
+ *pte = 0ULL;
+}
+
+static dma_addr_t __map_single(struct device *dev,
+ struct amd_iommu *iommu,
+ struct dma_ops_domain *dma_dom,
+ phys_addr_t paddr,
+ size_t size,
+ int dir)
+{
+ dma_addr_t offset = paddr & ~PAGE_MASK;
+ dma_addr_t address, start;
+ unsigned int pages;
+ int i;
+
+ pages = to_pages(paddr, size);
+ paddr &= PAGE_MASK;
+
+ address = dma_ops_alloc_addresses(dev, dma_dom, pages);
+ if (unlikely(address == bad_dma_address))
+ goto out;
+
+ start = address;
+ for (i = 0; i < pages; ++i) {
+ dma_ops_domain_map(iommu, dma_dom, start, paddr, dir);
+ paddr += PAGE_SIZE;
+ start += PAGE_SIZE;
+ }
+ address += offset;
+
+out:
+ return address;
+}
+
+static void __unmap_single(struct amd_iommu *iommu,
+ struct dma_ops_domain *dma_dom,
+ dma_addr_t dma_addr,
+ size_t size,
+ int dir)
+{
+ dma_addr_t i, start;
+ unsigned int pages;
+
+ if ((dma_addr == 0) || (dma_addr + size > dma_dom->aperture_size))
+ return;
+
+ pages = to_pages(dma_addr, size);
+ dma_addr &= PAGE_MASK;
+ start = dma_addr;
+
+ for (i = 0; i < pages; ++i) {
+ dma_ops_domain_unmap(iommu, dma_dom, start);
+ start += PAGE_SIZE;
+ }
+
+ dma_ops_free_addresses(dma_dom, dma_addr, pages);
+}
+
+static dma_addr_t map_single(struct device *dev, phys_addr_t paddr,
+ size_t size, int dir)
+{
+ unsigned long flags;
+ struct amd_iommu *iommu;
+ struct protection_domain *domain;
+ u16 devid;
+ dma_addr_t addr;
+
+ get_device_resources(dev, &iommu, &domain, &devid);
+
+ if (iommu == NULL || domain == NULL)
+ return (dma_addr_t)paddr;
+
+ spin_lock_irqsave(&domain->lock, flags);
+ addr = __map_single(dev, iommu, domain->priv, paddr, size, dir);
+ if (addr == bad_dma_address)
+ goto out;
+
+ if (iommu_has_npcache(iommu))
+ iommu_flush_pages(iommu, domain->id, addr, size);
+
+ if (iommu->need_sync)
+ iommu_completion_wait(iommu);
+
+out:
+ spin_unlock_irqrestore(&domain->lock, flags);
+
+ return addr;
+}
+
+static void unmap_single(struct device *dev, dma_addr_t dma_addr,
+ size_t size, int dir)
+{
+ unsigned long flags;
+ struct amd_iommu *iommu;
+ struct protection_domain *domain;
+ u16 devid;
+
+ if (!get_device_resources(dev, &iommu, &domain, &devid))
+ return;
+
+ spin_lock_irqsave(&domain->lock, flags);
+
+ __unmap_single(iommu, domain->priv, dma_addr, size, dir);
+
+ iommu_flush_pages(iommu, domain->id, dma_addr, size);
+
+ if (iommu->need_sync)
+ iommu_completion_wait(iommu);
+
+ spin_unlock_irqrestore(&domain->lock, flags);
+}
+
+static int map_sg_no_iommu(struct device *dev, struct scatterlist *sglist,
+ int nelems, int dir)
+{
+ struct scatterlist *s;
+ int i;
+
+ for_each_sg(sglist, s, nelems, i) {
+ s->dma_address = (dma_addr_t)sg_phys(s);
+ s->dma_length = s->length;
+ }
+
+ return nelems;
+}
+
+static int map_sg(struct device *dev, struct scatterlist *sglist,
+ int nelems, int dir)
+{
+ unsigned long flags;
+ struct amd_iommu *iommu;
+ struct protection_domain *domain;
+ u16 devid;
+ int i;
+ struct scatterlist *s;
+ phys_addr_t paddr;
+ int mapped_elems = 0;
+
+ get_device_resources(dev, &iommu, &domain, &devid);
+
+ if (!iommu || !domain)
+ return map_sg_no_iommu(dev, sglist, nelems, dir);
+
+ spin_lock_irqsave(&domain->lock, flags);
+
+ for_each_sg(sglist, s, nelems, i) {
+ paddr = sg_phys(s);
+
+ s->dma_address = __map_single(dev, iommu, domain->priv,
+ paddr, s->length, dir);
+
+ if (s->dma_address) {
+ s->dma_length = s->length;
+ mapped_elems++;
+ } else
+ goto unmap;
+ if (iommu_has_npcache(iommu))
+ iommu_flush_pages(iommu, domain->id, s->dma_address,
+ s->dma_length);
+ }
+
+ if (iommu->need_sync)
+ iommu_completion_wait(iommu);
+
+out:
+ spin_unlock_irqrestore(&domain->lock, flags);
+
+ return mapped_elems;
+unmap:
+ for_each_sg(sglist, s, mapped_elems, i) {
+ if (s->dma_address)
+ __unmap_single(iommu, domain->priv, s->dma_address,
+ s->dma_length, dir);
+ s->dma_address = s->dma_length = 0;
+ }
+
+ mapped_elems = 0;
+
+ goto out;
+}
+
+static void unmap_sg(struct device *dev, struct scatterlist *sglist,
+ int nelems, int dir)
+{
+ unsigned long flags;
+ struct amd_iommu *iommu;
+ struct protection_domain *domain;
+ struct scatterlist *s;
+ u16 devid;
+ int i;
+
+ if (!get_device_resources(dev, &iommu, &domain, &devid))
+ return;
+
+ spin_lock_irqsave(&domain->lock, flags);
+
+ for_each_sg(sglist, s, nelems, i) {
+ __unmap_single(iommu, domain->priv, s->dma_address,
+ s->dma_length, dir);
+ iommu_flush_pages(iommu, domain->id, s->dma_address,
+ s->dma_length);
+ s->dma_address = s->dma_length = 0;
+ }
+
+ if (iommu->need_sync)
+ iommu_completion_wait(iommu);
+
+ spin_unlock_irqrestore(&domain->lock, flags);
+}
+
+static void *alloc_coherent(struct device *dev, size_t size,
+ dma_addr_t *dma_addr, gfp_t flag)
+{
+ unsigned long flags;
+ void *virt_addr;
+ struct amd_iommu *iommu;
+ struct protection_domain *domain;
+ u16 devid;
+ phys_addr_t paddr;
+
+ virt_addr = (void *)__get_free_pages(flag, get_order(size));
+ if (!virt_addr)
+ return 0;
+
+ memset(virt_addr, 0, size);
+ paddr = virt_to_phys(virt_addr);
+
+ get_device_resources(dev, &iommu, &domain, &devid);
+
+ if (!iommu || !domain) {
+ *dma_addr = (dma_addr_t)paddr;
+ return virt_addr;
+ }
+
+ spin_lock_irqsave(&domain->lock, flags);
+
+ *dma_addr = __map_single(dev, iommu, domain->priv, paddr,
+ size, DMA_BIDIRECTIONAL);
+
+ if (*dma_addr == bad_dma_address) {
+ free_pages((unsigned long)virt_addr, get_order(size));
+ virt_addr = NULL;
+ goto out;
+ }
+
+ if (iommu_has_npcache(iommu))
+ iommu_flush_pages(iommu, domain->id, *dma_addr, size);
+
+ if (iommu->need_sync)
+ iommu_completion_wait(iommu);
+
+out:
+ spin_unlock_irqrestore(&domain->lock, flags);
+
+ return virt_addr;
+}
+
+static void free_coherent(struct device *dev, size_t size,
+ void *virt_addr, dma_addr_t dma_addr)
+{
+ unsigned long flags;
+ struct amd_iommu *iommu;
+ struct protection_domain *domain;
+ u16 devid;
+
+ get_device_resources(dev, &iommu, &domain, &devid);
+
+ if (!iommu || !domain)
+ goto free_mem;
+
+ spin_lock_irqsave(&domain->lock, flags);
+
+ __unmap_single(iommu, domain->priv, dma_addr, size, DMA_BIDIRECTIONAL);
+ iommu_flush_pages(iommu, domain->id, dma_addr, size);
+
+ if (iommu->need_sync)
+ iommu_completion_wait(iommu);
+
+ spin_unlock_irqrestore(&domain->lock, flags);
+
+free_mem:
+ free_pages((unsigned long)virt_addr, get_order(size));
+}
+
+/*
+ * If the driver core informs the DMA layer if a driver grabs a device
+ * we don't need to preallocate the protection domains anymore.
+ * For now we have to.
+ */
+void prealloc_protection_domains(void)
+{
+ struct pci_dev *dev = NULL;
+ struct dma_ops_domain *dma_dom;
+ struct amd_iommu *iommu;
+ int order = amd_iommu_aperture_order;
+ u16 devid;
+
+ while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
+ devid = (dev->bus->number << 8) | dev->devfn;
+ if (devid >= amd_iommu_last_bdf)
+ continue;
+ devid = amd_iommu_alias_table[devid];
+ if (domain_for_device(devid))
+ continue;
+ iommu = amd_iommu_rlookup_table[devid];
+ if (!iommu)
+ continue;
+ dma_dom = dma_ops_domain_alloc(iommu, order);
+ if (!dma_dom)
+ continue;
+ init_unity_mappings_for_device(dma_dom, devid);
+ set_device_domain(iommu, &dma_dom->domain, devid);
+ printk(KERN_INFO "AMD IOMMU: Allocated domain %d for device ",
+ dma_dom->domain.id);
+ print_devid(devid, 1);
+ }
+}
+
+static struct dma_mapping_ops amd_iommu_dma_ops = {
+ .alloc_coherent = alloc_coherent,
+ .free_coherent = free_coherent,
+ .map_single = map_single,
+ .unmap_single = unmap_single,
+ .map_sg = map_sg,
+ .unmap_sg = unmap_sg,
+};
+
+int __init amd_iommu_init_dma_ops(void)
+{
+ struct amd_iommu *iommu;
+ int order = amd_iommu_aperture_order;
+ int ret;
+
+ list_for_each_entry(iommu, &amd_iommu_list, list) {
+ iommu->default_dom = dma_ops_domain_alloc(iommu, order);
+ if (iommu->default_dom == NULL)
+ return -ENOMEM;
+ ret = iommu_init_unity_mappings(iommu);
+ if (ret)
+ goto free_domains;
+ }
+
+ if (amd_iommu_isolate)
+ prealloc_protection_domains();
+
+ iommu_detected = 1;
+ force_iommu = 1;
+ bad_dma_address = 0;
+#ifdef CONFIG_GART_IOMMU
+ gart_iommu_aperture_disabled = 1;
+ gart_iommu_aperture = 0;
+#endif
+
+ dma_ops = &amd_iommu_dma_ops;
+
+ return 0;
+
+free_domains:
+
+ list_for_each_entry(iommu, &amd_iommu_list, list) {
+ if (iommu->default_dom)
+ dma_ops_domain_free(iommu->default_dom);
+ }
+
+ return ret;
+}