#define LOOP_TIMEOUT 100000
+/*
+ * This bitmap is used to advertise the page sizes our hardware support
+ * to the IOMMU core, which will then use this information to split
+ * physically contiguous memory regions it is mapping into page sizes
+ * that we support.
+ *
+ * Traditionally the IOMMU core just handed us the mappings directly,
+ * after making sure the size is an order of a 4KiB page and that the
+ * mapping has natural alignment.
+ *
+ * To retain this behavior, we currently advertise that we support
+ * all page sizes that are an order of 4KiB.
+ *
+ * If at some point we'd like to utilize the IOMMU core's new behavior,
+ * we could change this to advertise the real page sizes we support.
+ */
+#define AMD_IOMMU_PGSIZES (~0xFFFUL)
+
static DEFINE_RWLOCK(amd_iommu_devtable_lock);
/* A list of preallocated protection domains */
{
static const int caps[] = {
PCI_EXT_CAP_ID_ATS,
- PCI_PRI_CAP,
- PCI_PASID_CAP,
+ PCI_EXT_CAP_ID_PRI,
+ PCI_EXT_CAP_ID_PASID,
};
int i, pos;
return true;
}
+static bool pdev_pri_erratum(struct pci_dev *pdev, u32 erratum)
+{
+ struct iommu_dev_data *dev_data;
+
+ dev_data = get_dev_data(&pdev->dev);
+
+ return dev_data->errata & (1 << erratum) ? true : false;
+}
+
/*
* In this function the list of preallocated protection domains is traversed to
* find the domain for a specific device
DECLARE_STATS_COUNTER(domain_flush_all);
DECLARE_STATS_COUNTER(alloced_io_mem);
DECLARE_STATS_COUNTER(total_map_requests);
+DECLARE_STATS_COUNTER(complete_ppr);
+DECLARE_STATS_COUNTER(invalidate_iotlb);
+DECLARE_STATS_COUNTER(invalidate_iotlb_all);
+DECLARE_STATS_COUNTER(pri_requests);
+
static struct dentry *stats_dir;
static struct dentry *de_fflush;
amd_iommu_stats_add(&domain_flush_all);
amd_iommu_stats_add(&alloced_io_mem);
amd_iommu_stats_add(&total_map_requests);
+ amd_iommu_stats_add(&complete_ppr);
+ amd_iommu_stats_add(&invalidate_iotlb);
+ amd_iommu_stats_add(&invalidate_iotlb_all);
+ amd_iommu_stats_add(&pri_requests);
}
#endif
volatile u64 *raw;
int i;
+ INC_STATS_COUNTER(pri_requests);
+
raw = (u64 *)(iommu->ppr_log + head);
/*
CMD_SET_TYPE(cmd, CMD_INV_IOTLB_PAGES);
}
+static void build_complete_ppr(struct iommu_cmd *cmd, u16 devid, int pasid,
+ int status, int tag, bool gn)
+{
+ memset(cmd, 0, sizeof(*cmd));
+
+ cmd->data[0] = devid;
+ if (gn) {
+ cmd->data[1] = pasid & PASID_MASK;
+ cmd->data[2] = CMD_INV_IOMMU_PAGES_GN_MASK;
+ }
+ cmd->data[3] = tag & 0x1ff;
+ cmd->data[3] |= (status & PPR_STATUS_MASK) << PPR_STATUS_SHIFT;
+
+ CMD_SET_TYPE(cmd, CMD_COMPLETE_PPR);
+}
+
static void build_inv_all(struct iommu_cmd *cmd)
{
memset(cmd, 0, sizeof(*cmd));
domain->pt_root = NULL;
}
+static void free_gcr3_tbl_level1(u64 *tbl)
+{
+ u64 *ptr;
+ int i;
+
+ for (i = 0; i < 512; ++i) {
+ if (!(tbl[i] & GCR3_VALID))
+ continue;
+
+ ptr = __va(tbl[i] & PAGE_MASK);
+
+ free_page((unsigned long)ptr);
+ }
+}
+
+static void free_gcr3_tbl_level2(u64 *tbl)
+{
+ u64 *ptr;
+ int i;
+
+ for (i = 0; i < 512; ++i) {
+ if (!(tbl[i] & GCR3_VALID))
+ continue;
+
+ ptr = __va(tbl[i] & PAGE_MASK);
+
+ free_gcr3_tbl_level1(ptr);
+ }
+}
+
static void free_gcr3_table(struct protection_domain *domain)
{
+ if (domain->glx == 2)
+ free_gcr3_tbl_level2(domain->gcr3_tbl);
+ else if (domain->glx == 1)
+ free_gcr3_tbl_level1(domain->gcr3_tbl);
+ else if (domain->glx != 0)
+ BUG();
+
free_page((unsigned long)domain->gcr3_tbl);
}
pci_disable_pasid(pdev);
}
+/* FIXME: Change generic reset-function to do the same */
+static int pri_reset_while_enabled(struct pci_dev *pdev)
+{
+ u16 control;
+ int pos;
+
+ pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
+ if (!pos)
+ return -EINVAL;
+
+ pci_read_config_word(pdev, pos + PCI_PRI_CTRL, &control);
+ control |= PCI_PRI_CTRL_RESET;
+ pci_write_config_word(pdev, pos + PCI_PRI_CTRL, control);
+
+ return 0;
+}
+
static int pdev_iommuv2_enable(struct pci_dev *pdev)
{
- int ret;
+ bool reset_enable;
+ int reqs, ret;
+
+ /* FIXME: Hardcode number of outstanding requests for now */
+ reqs = 32;
+ if (pdev_pri_erratum(pdev, AMD_PRI_DEV_ERRATUM_LIMIT_REQ_ONE))
+ reqs = 1;
+ reset_enable = pdev_pri_erratum(pdev, AMD_PRI_DEV_ERRATUM_ENABLE_RESET);
/* Only allow access to user-accessible pages */
ret = pci_enable_pasid(pdev, 0);
if (ret)
goto out_err;
- /* FIXME: Hardcode number of outstanding requests for now */
- ret = pci_enable_pri(pdev, 32);
+ /* Enable PRI */
+ ret = pci_enable_pri(pdev, reqs);
if (ret)
goto out_err;
+ if (reset_enable) {
+ ret = pri_reset_while_enabled(pdev);
+ if (ret)
+ goto out_err;
+ }
+
ret = pci_enable_ats(pdev, PAGE_SHIFT);
if (ret)
goto out_err;
return ret;
}
+/* FIXME: Move this to PCI code */
+#define PCI_PRI_TLP_OFF (1 << 2)
+
+bool pci_pri_tlp_required(struct pci_dev *pdev)
+{
+ u16 control;
+ int pos;
+
+ pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
+ if (!pos)
+ return false;
+
+ pci_read_config_word(pdev, pos + PCI_PRI_CTRL, &control);
+
+ return (control & PCI_PRI_TLP_OFF) ? true : false;
+}
+
/*
* If a device is not yet associated with a domain, this function does
* assigns it visible for the hardware
dev_data->ats.enabled = true;
dev_data->ats.qdep = pci_ats_queue_depth(pdev);
+ dev_data->pri_tlp = pci_pri_tlp_required(pdev);
} else if (amd_iommu_iotlb_sup &&
pci_enable_ats(pdev, PAGE_SHIFT) == 0) {
dev_data->ats.enabled = true;
if (!domain->pt_root)
goto out_free;
+ domain->iommu_domain = dom;
+
dom->priv = domain;
return 0;
}
static int amd_iommu_map(struct iommu_domain *dom, unsigned long iova,
- phys_addr_t paddr, int gfp_order, int iommu_prot)
+ phys_addr_t paddr, size_t page_size, int iommu_prot)
{
- unsigned long page_size = 0x1000UL << gfp_order;
struct protection_domain *domain = dom->priv;
int prot = 0;
int ret;
return ret;
}
-static int amd_iommu_unmap(struct iommu_domain *dom, unsigned long iova,
- int gfp_order)
+static size_t amd_iommu_unmap(struct iommu_domain *dom, unsigned long iova,
+ size_t page_size)
{
struct protection_domain *domain = dom->priv;
- unsigned long page_size, unmap_size;
+ size_t unmap_size;
if (domain->mode == PAGE_MODE_NONE)
return -EINVAL;
- page_size = 0x1000UL << gfp_order;
-
mutex_lock(&domain->api_lock);
unmap_size = iommu_unmap_page(domain, iova, page_size);
mutex_unlock(&domain->api_lock);
domain_flush_tlb_pde(domain);
- return get_order(unmap_size);
+ return unmap_size;
}
static phys_addr_t amd_iommu_iova_to_phys(struct iommu_domain *dom,
return 0;
}
+static int amd_iommu_device_group(struct device *dev, unsigned int *groupid)
+{
+ struct iommu_dev_data *dev_data = dev->archdata.iommu;
+ struct pci_dev *pdev = to_pci_dev(dev);
+ u16 devid;
+
+ if (!dev_data)
+ return -ENODEV;
+
+ if (pdev->is_virtfn || !iommu_group_mf)
+ devid = dev_data->devid;
+ else
+ devid = calc_devid(pdev->bus->number,
+ PCI_DEVFN(PCI_SLOT(pdev->devfn), 0));
+
+ *groupid = amd_iommu_alias_table[devid];
+
+ return 0;
+}
+
static struct iommu_ops amd_iommu_ops = {
.domain_init = amd_iommu_domain_init,
.domain_destroy = amd_iommu_domain_destroy,
.unmap = amd_iommu_unmap,
.iova_to_phys = amd_iommu_iova_to_phys,
.domain_has_cap = amd_iommu_domain_has_cap,
+ .device_group = amd_iommu_device_group,
+ .pgsize_bitmap = AMD_IOMMU_PGSIZES,
};
/*****************************************************************************
attach_device(&dev->dev, pt_domain);
}
+ amd_iommu_stats_init();
+
pr_info("AMD-Vi: Initialized for Passthrough Mode\n");
return 0;
static int __amd_iommu_flush_page(struct protection_domain *domain, int pasid,
u64 address)
{
+ INC_STATS_COUNTER(invalidate_iotlb);
+
return __flush_pasid(domain, pasid, address, false);
}
static int __amd_iommu_flush_tlb(struct protection_domain *domain, int pasid)
{
+ INC_STATS_COUNTER(invalidate_iotlb_all);
+
return __flush_pasid(domain, pasid, CMD_INV_IOMMU_ALL_PAGES_ADDRESS,
true);
}
}
EXPORT_SYMBOL(amd_iommu_flush_tlb);
+static u64 *__get_gcr3_pte(u64 *root, int level, int pasid, bool alloc)
+{
+ int index;
+ u64 *pte;
+
+ while (true) {
+
+ index = (pasid >> (9 * level)) & 0x1ff;
+ pte = &root[index];
+
+ if (level == 0)
+ break;
+
+ if (!(*pte & GCR3_VALID)) {
+ if (!alloc)
+ return NULL;
+
+ root = (void *)get_zeroed_page(GFP_ATOMIC);
+ if (root == NULL)
+ return NULL;
+
+ *pte = __pa(root) | GCR3_VALID;
+ }
+
+ root = __va(*pte & PAGE_MASK);
+
+ level -= 1;
+ }
+
+ return pte;
+}
+
+static int __set_gcr3(struct protection_domain *domain, int pasid,
+ unsigned long cr3)
+{
+ u64 *pte;
+
+ if (domain->mode != PAGE_MODE_NONE)
+ return -EINVAL;
+
+ pte = __get_gcr3_pte(domain->gcr3_tbl, domain->glx, pasid, true);
+ if (pte == NULL)
+ return -ENOMEM;
+
+ *pte = (cr3 & PAGE_MASK) | GCR3_VALID;
+
+ return __amd_iommu_flush_tlb(domain, pasid);
+}
+
+static int __clear_gcr3(struct protection_domain *domain, int pasid)
+{
+ u64 *pte;
+
+ if (domain->mode != PAGE_MODE_NONE)
+ return -EINVAL;
+
+ pte = __get_gcr3_pte(domain->gcr3_tbl, domain->glx, pasid, false);
+ if (pte == NULL)
+ return 0;
+
+ *pte = 0;
+
+ return __amd_iommu_flush_tlb(domain, pasid);
+}
+
+int amd_iommu_domain_set_gcr3(struct iommu_domain *dom, int pasid,
+ unsigned long cr3)
+{
+ struct protection_domain *domain = dom->priv;
+ unsigned long flags;
+ int ret;
+
+ spin_lock_irqsave(&domain->lock, flags);
+ ret = __set_gcr3(domain, pasid, cr3);
+ spin_unlock_irqrestore(&domain->lock, flags);
+
+ return ret;
+}
+EXPORT_SYMBOL(amd_iommu_domain_set_gcr3);
+
+int amd_iommu_domain_clear_gcr3(struct iommu_domain *dom, int pasid)
+{
+ struct protection_domain *domain = dom->priv;
+ unsigned long flags;
+ int ret;
+
+ spin_lock_irqsave(&domain->lock, flags);
+ ret = __clear_gcr3(domain, pasid);
+ spin_unlock_irqrestore(&domain->lock, flags);
+
+ return ret;
+}
+EXPORT_SYMBOL(amd_iommu_domain_clear_gcr3);
+
+int amd_iommu_complete_ppr(struct pci_dev *pdev, int pasid,
+ int status, int tag)
+{
+ struct iommu_dev_data *dev_data;
+ struct amd_iommu *iommu;
+ struct iommu_cmd cmd;
+
+ INC_STATS_COUNTER(complete_ppr);
+
+ dev_data = get_dev_data(&pdev->dev);
+ iommu = amd_iommu_rlookup_table[dev_data->devid];
+
+ build_complete_ppr(&cmd, dev_data->devid, pasid, status,
+ tag, dev_data->pri_tlp);
+
+ return iommu_queue_command(iommu, &cmd);
+}
+EXPORT_SYMBOL(amd_iommu_complete_ppr);
+
+struct iommu_domain *amd_iommu_get_v2_domain(struct pci_dev *pdev)
+{
+ struct protection_domain *domain;
+
+ domain = get_domain(&pdev->dev);
+ if (IS_ERR(domain))
+ return NULL;
+
+ /* Only return IOMMUv2 domains */
+ if (!(domain->flags & PD_IOMMUV2_MASK))
+ return NULL;
+
+ return domain->iommu_domain;
+}
+EXPORT_SYMBOL(amd_iommu_get_v2_domain);
+
+void amd_iommu_enable_device_erratum(struct pci_dev *pdev, u32 erratum)
+{
+ struct iommu_dev_data *dev_data;
+
+ if (!amd_iommu_v2_supported())
+ return;
+
+ dev_data = get_dev_data(&pdev->dev);
+ dev_data->errata |= (1 << erratum);
+}
+EXPORT_SYMBOL(amd_iommu_enable_device_erratum);
+
+int amd_iommu_device_info(struct pci_dev *pdev,
+ struct amd_iommu_device_info *info)
+{
+ int max_pasids;
+ int pos;
+
+ if (pdev == NULL || info == NULL)
+ return -EINVAL;
+
+ if (!amd_iommu_v2_supported())
+ return -EINVAL;
+
+ memset(info, 0, sizeof(*info));
+
+ pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ATS);
+ if (pos)
+ info->flags |= AMD_IOMMU_DEVICE_FLAG_ATS_SUP;
+
+ pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
+ if (pos)
+ info->flags |= AMD_IOMMU_DEVICE_FLAG_PRI_SUP;
+
+ pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PASID);
+ if (pos) {
+ int features;
+
+ max_pasids = 1 << (9 * (amd_iommu_max_glx_val + 1));
+ max_pasids = min(max_pasids, (1 << 20));
+
+ info->flags |= AMD_IOMMU_DEVICE_FLAG_PASID_SUP;
+ info->max_pasids = min(pci_max_pasids(pdev), max_pasids);
+
+ features = pci_pasid_features(pdev);
+ if (features & PCI_PASID_CAP_EXEC)
+ info->flags |= AMD_IOMMU_DEVICE_FLAG_EXEC_SUP;
+ if (features & PCI_PASID_CAP_PRIV)
+ info->flags |= AMD_IOMMU_DEVICE_FLAG_PRIV_SUP;
+ }
+
+ return 0;
+}
+EXPORT_SYMBOL(amd_iommu_device_info);