/*
- * Copyright (C) 2007-2008 Advanced Micro Devices, Inc.
+ * Copyright (C) 2007-2009 Advanced Micro Devices, Inc.
* Author: Joerg Roedel <joerg.roedel@amd.com>
* Leo Duran <leo.duran@amd.com>
*
*/
#include <linux/pci.h>
-#include <linux/gfp.h>
-#include <linux/bitops.h>
+#include <linux/bitmap.h>
+#include <linux/slab.h>
#include <linux/debugfs.h>
#include <linux/scatterlist.h>
#include <linux/dma-mapping.h>
#include <asm/proto.h>
#include <asm/iommu.h>
#include <asm/gart.h>
+#include <asm/amd_iommu_proto.h>
#include <asm/amd_iommu_types.h>
#include <asm/amd_iommu.h>
static LIST_HEAD(iommu_pd_list);
static DEFINE_SPINLOCK(iommu_pd_list_lock);
-#ifdef CONFIG_IOMMU_API
+/*
+ * Domain for untranslated devices - only allocated
+ * if iommu=pt passed on kernel cmd line.
+ */
+static struct protection_domain *pt_domain;
+
static struct iommu_ops amd_iommu_ops;
-#endif
/*
* general struct to manage commands send to an IOMMU
u32 data[4];
};
-static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
- struct unity_map_entry *e);
-static struct dma_ops_domain *find_protection_domain(u16 devid);
-static u64* alloc_pte(struct protection_domain *dom,
- unsigned long address, u64
- **pte_page, gfp_t gfp);
-static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
- unsigned long start_page,
- unsigned int pages);
+static void reset_iommu_command_buffer(struct amd_iommu *iommu);
+static void update_domain(struct protection_domain *domain);
-#ifndef BUS_NOTIFY_UNBOUND_DRIVER
-#define BUS_NOTIFY_UNBOUND_DRIVER 0x0005
-#endif
+/****************************************************************************
+ *
+ * Helper functions
+ *
+ ****************************************************************************/
+
+static inline u16 get_device_id(struct device *dev)
+{
+ struct pci_dev *pdev = to_pci_dev(dev);
+
+ return calc_devid(pdev->bus->number, pdev->devfn);
+}
+
+static struct iommu_dev_data *get_dev_data(struct device *dev)
+{
+ return dev->archdata.iommu;
+}
+
+/*
+ * In this function the list of preallocated protection domains is traversed to
+ * find the domain for a specific device
+ */
+static struct dma_ops_domain *find_protection_domain(u16 devid)
+{
+ struct dma_ops_domain *entry, *ret = NULL;
+ unsigned long flags;
+ u16 alias = amd_iommu_alias_table[devid];
+ if (list_empty(&iommu_pd_list))
+ return NULL;
+
+ spin_lock_irqsave(&iommu_pd_list_lock, flags);
+
+ list_for_each_entry(entry, &iommu_pd_list, list) {
+ if (entry->target_dev == devid ||
+ entry->target_dev == alias) {
+ ret = entry;
+ break;
+ }
+ }
+
+ spin_unlock_irqrestore(&iommu_pd_list_lock, flags);
+
+ return ret;
+}
+
+/*
+ * This function checks if the driver got a valid device from the caller to
+ * avoid dereferencing invalid pointers.
+ */
+static bool check_device(struct device *dev)
+{
+ u16 devid;
+
+ if (!dev || !dev->dma_mask)
+ return false;
+
+ /* No device or no PCI device */
+ if (!dev || dev->bus != &pci_bus_type)
+ return false;
+
+ devid = get_device_id(dev);
+
+ /* Out of our scope? */
+ if (devid > amd_iommu_last_bdf)
+ return false;
+
+ if (amd_iommu_rlookup_table[devid] == NULL)
+ return false;
+
+ return true;
+}
+
+static int iommu_init_device(struct device *dev)
+{
+ struct iommu_dev_data *dev_data;
+ struct pci_dev *pdev;
+ u16 devid, alias;
+
+ if (dev->archdata.iommu)
+ return 0;
+
+ dev_data = kzalloc(sizeof(*dev_data), GFP_KERNEL);
+ if (!dev_data)
+ return -ENOMEM;
+
+ dev_data->dev = dev;
+
+ devid = get_device_id(dev);
+ alias = amd_iommu_alias_table[devid];
+ pdev = pci_get_bus_and_slot(PCI_BUS(alias), alias & 0xff);
+ if (pdev)
+ dev_data->alias = &pdev->dev;
+
+ atomic_set(&dev_data->bind, 0);
+
+ dev->archdata.iommu = dev_data;
+
+
+ return 0;
+}
+
+static void iommu_uninit_device(struct device *dev)
+{
+ kfree(dev->archdata.iommu);
+}
+
+void __init amd_iommu_uninit_devices(void)
+{
+ struct pci_dev *pdev = NULL;
+
+ for_each_pci_dev(pdev) {
+
+ if (!check_device(&pdev->dev))
+ continue;
+
+ iommu_uninit_device(&pdev->dev);
+ }
+}
+
+int __init amd_iommu_init_devices(void)
+{
+ struct pci_dev *pdev = NULL;
+ int ret = 0;
+
+ for_each_pci_dev(pdev) {
+
+ if (!check_device(&pdev->dev))
+ continue;
+
+ ret = iommu_init_device(&pdev->dev);
+ if (ret)
+ goto out_free;
+ }
+
+ return 0;
+
+out_free:
+
+ amd_iommu_uninit_devices();
+
+ return ret;
+}
#ifdef CONFIG_AMD_IOMMU_STATS
/*
DECLARE_STATS_COUNTER(total_map_requests);
static struct dentry *stats_dir;
-static struct dentry *de_isolate;
static struct dentry *de_fflush;
static void amd_iommu_stats_add(struct __iommu_counter *cnt)
if (stats_dir == NULL)
return;
- de_isolate = debugfs_create_bool("isolation", 0444, stats_dir,
- (u32 *)&amd_iommu_isolate);
-
de_fflush = debugfs_create_bool("fullflush", 0444, stats_dir,
(u32 *)&amd_iommu_unmap_flush);
#endif
-/* returns !0 if the IOMMU is caching non-present entries in its TLB */
-static int iommu_has_npcache(struct amd_iommu *iommu)
-{
- return iommu->cap & (1UL << IOMMU_CAP_NPCACHE);
-}
-
/****************************************************************************
*
* Interrupt handling functions
amd_iommu_dev_table[devid].data[i]);
}
-static void iommu_print_event(void *__evt)
+static void dump_command(unsigned long phys_addr)
+{
+ struct iommu_cmd *cmd = phys_to_virt(phys_addr);
+ int i;
+
+ for (i = 0; i < 4; ++i)
+ pr_err("AMD-Vi: CMD[%d]: %08x\n", i, cmd->data[i]);
+}
+
+static void iommu_print_event(struct amd_iommu *iommu, void *__evt)
{
u32 *event = __evt;
int type = (event[1] >> EVENT_TYPE_SHIFT) & EVENT_TYPE_MASK;
int flags = (event[1] >> EVENT_FLAGS_SHIFT) & EVENT_FLAGS_MASK;
u64 address = (u64)(((u64)event[3]) << 32) | event[2];
- printk(KERN_ERR "AMD IOMMU: Event logged [");
+ printk(KERN_ERR "AMD-Vi: Event logged [");
switch (type) {
case EVENT_TYPE_ILL_DEV:
break;
case EVENT_TYPE_ILL_CMD:
printk("ILLEGAL_COMMAND_ERROR address=0x%016llx]\n", address);
+ iommu->reset_in_progress = true;
+ reset_iommu_command_buffer(iommu);
+ dump_command(address);
break;
case EVENT_TYPE_CMD_HARD_ERR:
printk("COMMAND_HARDWARE_ERROR address=0x%016llx "
tail = readl(iommu->mmio_base + MMIO_EVT_TAIL_OFFSET);
while (head != tail) {
- iommu_print_event(iommu->evt_buf + head);
+ iommu_print_event(iommu, iommu->evt_buf + head);
head = (head + EVENT_ENTRY_SIZE) % iommu->evt_buf_size;
}
writel(status, iommu->mmio_base + MMIO_STATUS_OFFSET);
if (unlikely(i == EXIT_LOOP_COUNT))
- panic("AMD IOMMU: Completion wait loop failed\n");
+ iommu->reset_in_progress = true;
}
/*
out:
spin_unlock_irqrestore(&iommu->lock, flags);
+ if (iommu->reset_in_progress)
+ reset_iommu_command_buffer(iommu);
+
return 0;
}
+static void iommu_flush_complete(struct protection_domain *domain)
+{
+ int i;
+
+ for (i = 0; i < amd_iommus_present; ++i) {
+ if (!domain->dev_iommu[i])
+ continue;
+
+ /*
+ * Devices of this domain are behind this IOMMU
+ * We need to wait for completion of all commands.
+ */
+ iommu_completion_wait(amd_iommus[i]);
+ }
+}
+
/*
* Command send function for invalidating a device table entry
*/
-static int iommu_queue_inv_dev_entry(struct amd_iommu *iommu, u16 devid)
+static int iommu_flush_device(struct device *dev)
{
+ struct amd_iommu *iommu;
struct iommu_cmd cmd;
- int ret;
+ u16 devid;
- BUG_ON(iommu == NULL);
+ devid = get_device_id(dev);
+ iommu = amd_iommu_rlookup_table[devid];
+ /* Build command */
memset(&cmd, 0, sizeof(cmd));
CMD_SET_TYPE(&cmd, CMD_INV_DEV_ENTRY);
cmd.data[0] = devid;
- ret = iommu_queue_command(iommu, &cmd);
-
- return ret;
+ return iommu_queue_command(iommu, &cmd);
}
static void __iommu_build_inv_iommu_pages(struct iommu_cmd *cmd, u64 address,
* It invalidates a single PTE if the range to flush is within a single
* page. Otherwise it flushes the whole TLB of the IOMMU.
*/
-static int iommu_flush_pages(struct amd_iommu *iommu, u16 domid,
- u64 address, size_t size)
+static void __iommu_flush_pages(struct protection_domain *domain,
+ u64 address, size_t size, int pde)
{
- int s = 0;
- unsigned pages = iommu_num_pages(address, size, PAGE_SIZE);
+ int s = 0, i;
+ unsigned long pages = iommu_num_pages(address, size, PAGE_SIZE);
address &= PAGE_MASK;
s = 1;
}
- iommu_queue_inv_iommu_pages(iommu, address, domid, 0, s);
- return 0;
+ for (i = 0; i < amd_iommus_present; ++i) {
+ if (!domain->dev_iommu[i])
+ continue;
+
+ /*
+ * Devices of this domain are behind this IOMMU
+ * We need a TLB flush
+ */
+ iommu_queue_inv_iommu_pages(amd_iommus[i], address,
+ domain->id, pde, s);
+ }
+
+ return;
}
-/* Flush the whole IO/TLB for a given protection domain */
-static void iommu_flush_tlb(struct amd_iommu *iommu, u16 domid)
+static void iommu_flush_pages(struct protection_domain *domain,
+ u64 address, size_t size)
{
- u64 address = CMD_INV_IOMMU_ALL_PAGES_ADDRESS;
-
- INC_STATS_COUNTER(domain_flush_single);
+ __iommu_flush_pages(domain, address, size, 0);
+}
- iommu_queue_inv_iommu_pages(iommu, address, domid, 0, 1);
+/* Flush the whole IO/TLB for a given protection domain */
+static void iommu_flush_tlb(struct protection_domain *domain)
+{
+ __iommu_flush_pages(domain, 0, CMD_INV_IOMMU_ALL_PAGES_ADDRESS, 0);
}
/* Flush the whole IO/TLB for a given protection domain - including PDE */
-static void iommu_flush_tlb_pde(struct amd_iommu *iommu, u16 domid)
+static void iommu_flush_tlb_pde(struct protection_domain *domain)
{
- u64 address = CMD_INV_IOMMU_ALL_PAGES_ADDRESS;
-
- INC_STATS_COUNTER(domain_flush_single);
-
- iommu_queue_inv_iommu_pages(iommu, address, domid, 1, 1);
+ __iommu_flush_pages(domain, 0, CMD_INV_IOMMU_ALL_PAGES_ADDRESS, 1);
}
+
/*
- * This function is used to flush the IO/TLB for a given protection domain
- * on every IOMMU in the system
+ * This function flushes the DTEs for all devices in domain
*/
-static void iommu_flush_domain(u16 domid)
+static void iommu_flush_domain_devices(struct protection_domain *domain)
{
+ struct iommu_dev_data *dev_data;
unsigned long flags;
- struct amd_iommu *iommu;
- struct iommu_cmd cmd;
- INC_STATS_COUNTER(domain_flush_all);
+ spin_lock_irqsave(&domain->lock, flags);
- __iommu_build_inv_iommu_pages(&cmd, CMD_INV_IOMMU_ALL_PAGES_ADDRESS,
- domid, 1, 1);
+ list_for_each_entry(dev_data, &domain->dev_list, list)
+ iommu_flush_device(dev_data->dev);
- for_each_iommu(iommu) {
- spin_lock_irqsave(&iommu->lock, flags);
- __iommu_queue_command(iommu, &cmd);
- __iommu_completion_wait(iommu);
- __iommu_wait_for_completion(iommu);
- spin_unlock_irqrestore(&iommu->lock, flags);
- }
+ spin_unlock_irqrestore(&domain->lock, flags);
}
-void amd_iommu_flush_all_domains(void)
+static void iommu_flush_all_domain_devices(void)
{
- int i;
+ struct protection_domain *domain;
+ unsigned long flags;
- for (i = 1; i < MAX_DOMAIN_ID; ++i) {
- if (!test_bit(i, amd_iommu_pd_alloc_bitmap))
- continue;
- iommu_flush_domain(i);
+ spin_lock_irqsave(&amd_iommu_pd_lock, flags);
+
+ list_for_each_entry(domain, &amd_iommu_pd_list, list) {
+ iommu_flush_domain_devices(domain);
+ iommu_flush_complete(domain);
}
+
+ spin_unlock_irqrestore(&amd_iommu_pd_lock, flags);
}
void amd_iommu_flush_all_devices(void)
{
- struct amd_iommu *iommu;
- int i;
+ iommu_flush_all_domain_devices();
+}
- for (i = 0; i <= amd_iommu_last_bdf; ++i) {
- if (amd_iommu_pd_table[i] == NULL)
- continue;
+/*
+ * This function uses heavy locking and may disable irqs for some time. But
+ * this is no issue because it is only called during resume.
+ */
+void amd_iommu_flush_all_domains(void)
+{
+ struct protection_domain *domain;
+ unsigned long flags;
- iommu = amd_iommu_rlookup_table[i];
- if (!iommu)
- continue;
+ spin_lock_irqsave(&amd_iommu_pd_lock, flags);
- iommu_queue_inv_dev_entry(iommu, i);
- iommu_completion_wait(iommu);
+ list_for_each_entry(domain, &amd_iommu_pd_list, list) {
+ spin_lock(&domain->lock);
+ iommu_flush_tlb_pde(domain);
+ iommu_flush_complete(domain);
+ spin_unlock(&domain->lock);
}
+
+ spin_unlock_irqrestore(&amd_iommu_pd_lock, flags);
+}
+
+static void reset_iommu_command_buffer(struct amd_iommu *iommu)
+{
+ pr_err("AMD-Vi: Resetting IOMMU command buffer\n");
+
+ if (iommu->reset_in_progress)
+ panic("AMD-Vi: ILLEGAL_COMMAND_ERROR while resetting command buffer\n");
+
+ amd_iommu_reset_cmd_buffer(iommu);
+ amd_iommu_flush_all_devices();
+ amd_iommu_flush_all_domains();
+
+ iommu->reset_in_progress = false;
}
/****************************************************************************
****************************************************************************/
/*
+ * This function is used to add another level to an IO page table. Adding
+ * another level increases the size of the address space by 9 bits to a size up
+ * to 64 bits.
+ */
+static bool increase_address_space(struct protection_domain *domain,
+ gfp_t gfp)
+{
+ u64 *pte;
+
+ if (domain->mode == PAGE_MODE_6_LEVEL)
+ /* address space already 64 bit large */
+ return false;
+
+ pte = (void *)get_zeroed_page(gfp);
+ if (!pte)
+ return false;
+
+ *pte = PM_LEVEL_PDE(domain->mode,
+ virt_to_phys(domain->pt_root));
+ domain->pt_root = pte;
+ domain->mode += 1;
+ domain->updated = true;
+
+ return true;
+}
+
+static u64 *alloc_pte(struct protection_domain *domain,
+ unsigned long address,
+ int end_lvl,
+ u64 **pte_page,
+ gfp_t gfp)
+{
+ u64 *pte, *page;
+ int level;
+
+ while (address > PM_LEVEL_SIZE(domain->mode))
+ increase_address_space(domain, gfp);
+
+ level = domain->mode - 1;
+ pte = &domain->pt_root[PM_LEVEL_INDEX(level, address)];
+
+ while (level > end_lvl) {
+ if (!IOMMU_PTE_PRESENT(*pte)) {
+ page = (u64 *)get_zeroed_page(gfp);
+ if (!page)
+ return NULL;
+ *pte = PM_LEVEL_PDE(level, virt_to_phys(page));
+ }
+
+ level -= 1;
+
+ pte = IOMMU_PTE_PAGE(*pte);
+
+ if (pte_page && level == end_lvl)
+ *pte_page = pte;
+
+ pte = &pte[PM_LEVEL_INDEX(level, address)];
+ }
+
+ return pte;
+}
+
+/*
+ * This function checks if there is a PTE for a given dma address. If
+ * there is one, it returns the pointer to it.
+ */
+static u64 *fetch_pte(struct protection_domain *domain,
+ unsigned long address, int map_size)
+{
+ int level;
+ u64 *pte;
+
+ level = domain->mode - 1;
+ pte = &domain->pt_root[PM_LEVEL_INDEX(level, address)];
+
+ while (level > map_size) {
+ if (!IOMMU_PTE_PRESENT(*pte))
+ return NULL;
+
+ level -= 1;
+
+ pte = IOMMU_PTE_PAGE(*pte);
+ pte = &pte[PM_LEVEL_INDEX(level, address)];
+
+ if ((PM_PTE_LEVEL(*pte) == 0) && level != map_size) {
+ pte = NULL;
+ break;
+ }
+ }
+
+ return pte;
+}
+
+/*
* Generic mapping functions. It maps a physical address into a DMA
* address space. It allocates the page table pages if necessary.
* In the future it can be extended to a generic mapping function
static int iommu_map_page(struct protection_domain *dom,
unsigned long bus_addr,
unsigned long phys_addr,
- int prot)
+ int prot,
+ int map_size)
{
u64 __pte, *pte;
bus_addr = PAGE_ALIGN(bus_addr);
phys_addr = PAGE_ALIGN(phys_addr);
- /* only support 512GB address spaces for now */
- if (bus_addr > IOMMU_MAP_SIZE_L3 || !(prot & IOMMU_PROT_MASK))
+ BUG_ON(!PM_ALIGNED(map_size, bus_addr));
+ BUG_ON(!PM_ALIGNED(map_size, phys_addr));
+
+ if (!(prot & IOMMU_PROT_MASK))
return -EINVAL;
- pte = alloc_pte(dom, bus_addr, NULL, GFP_KERNEL);
+ pte = alloc_pte(dom, bus_addr, map_size, NULL, GFP_KERNEL);
if (IOMMU_PTE_PRESENT(*pte))
return -EBUSY;
*pte = __pte;
+ update_domain(dom);
+
return 0;
}
static void iommu_unmap_page(struct protection_domain *dom,
- unsigned long bus_addr)
+ unsigned long bus_addr, int map_size)
{
- u64 *pte;
-
- pte = &dom->pt_root[IOMMU_PTE_L2_INDEX(bus_addr)];
-
- if (!IOMMU_PTE_PRESENT(*pte))
- return;
+ u64 *pte = fetch_pte(dom, bus_addr, map_size);
- pte = IOMMU_PTE_PAGE(*pte);
- pte = &pte[IOMMU_PTE_L1_INDEX(bus_addr)];
-
- if (!IOMMU_PTE_PRESENT(*pte))
- return;
-
- pte = IOMMU_PTE_PAGE(*pte);
- pte = &pte[IOMMU_PTE_L1_INDEX(bus_addr)];
-
- *pte = 0;
+ if (pte)
+ *pte = 0;
}
/*
}
/*
- * Init the unity mappings for a specific IOMMU in the system
- *
- * Basically iterates over all unity mapping entries and applies them to
- * the default domain DMA of that IOMMU if necessary.
- */
-static int iommu_init_unity_mappings(struct amd_iommu *iommu)
-{
- struct unity_map_entry *entry;
- int ret;
-
- list_for_each_entry(entry, &amd_iommu_unity_map, list) {
- if (!iommu_for_unity_map(iommu, entry))
- continue;
- ret = dma_ops_unity_map(iommu->default_dom, entry);
- if (ret)
- return ret;
- }
-
- return 0;
-}
-
-/*
* This function actually applies the mapping to the page table of the
* dma_ops domain.
*/
for (addr = e->address_start; addr < e->address_end;
addr += PAGE_SIZE) {
- ret = iommu_map_page(&dma_dom->domain, addr, addr, e->prot);
+ ret = iommu_map_page(&dma_dom->domain, addr, addr, e->prot,
+ PM_MAP_4k);
if (ret)
return ret;
/*
}
/*
+ * Init the unity mappings for a specific IOMMU in the system
+ *
+ * Basically iterates over all unity mapping entries and applies them to
+ * the default domain DMA of that IOMMU if necessary.
+ */
+static int iommu_init_unity_mappings(struct amd_iommu *iommu)
+{
+ struct unity_map_entry *entry;
+ int ret;
+
+ list_for_each_entry(entry, &amd_iommu_unity_map, list) {
+ if (!iommu_for_unity_map(iommu, entry))
+ continue;
+ ret = dma_ops_unity_map(iommu->default_dom, entry);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
+/*
* Inits the unity mappings required for a specific device
*/
static int init_unity_mappings_for_device(struct dma_ops_domain *dma_dom,
*/
/*
- * This function checks if there is a PTE for a given dma address. If
- * there is one, it returns the pointer to it.
+ * Used to reserve address ranges in the aperture (e.g. for exclusion
+ * ranges.
*/
-static u64* fetch_pte(struct protection_domain *domain,
- unsigned long address)
+static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
+ unsigned long start_page,
+ unsigned int pages)
{
- u64 *pte;
-
- pte = &domain->pt_root[IOMMU_PTE_L2_INDEX(address)];
-
- if (!IOMMU_PTE_PRESENT(*pte))
- return NULL;
-
- pte = IOMMU_PTE_PAGE(*pte);
- pte = &pte[IOMMU_PTE_L1_INDEX(address)];
-
- if (!IOMMU_PTE_PRESENT(*pte))
- return NULL;
+ unsigned int i, last_page = dom->aperture_size >> PAGE_SHIFT;
- pte = IOMMU_PTE_PAGE(*pte);
- pte = &pte[IOMMU_PTE_L0_INDEX(address)];
+ if (start_page + pages > last_page)
+ pages = last_page - start_page;
- return pte;
+ for (i = start_page; i < start_page + pages; ++i) {
+ int index = i / APERTURE_RANGE_PAGES;
+ int page = i % APERTURE_RANGE_PAGES;
+ __set_bit(page, dom->aperture[index]->bitmap);
+ }
}
/*
* aperture in case of dma_ops domain allocation or address allocation
* failure.
*/
-static int alloc_new_range(struct amd_iommu *iommu,
- struct dma_ops_domain *dma_dom,
+static int alloc_new_range(struct dma_ops_domain *dma_dom,
bool populate, gfp_t gfp)
{
int index = dma_dom->aperture_size >> APERTURE_RANGE_SHIFT;
- int i;
+ struct amd_iommu *iommu;
+ unsigned long i;
#ifdef CONFIG_IOMMU_STRESS
populate = false;
u64 *pte, *pte_page;
for (i = 0; i < num_ptes; ++i) {
- pte = alloc_pte(&dma_dom->domain, address,
+ pte = alloc_pte(&dma_dom->domain, address, PM_MAP_4k,
&pte_page, gfp);
if (!pte)
goto out_free;
dma_dom->aperture_size += APERTURE_RANGE_SIZE;
/* Intialize the exclusion range if necessary */
- if (iommu->exclusion_start &&
- iommu->exclusion_start >= dma_dom->aperture[index]->offset &&
- iommu->exclusion_start < dma_dom->aperture_size) {
- unsigned long startpage = iommu->exclusion_start >> PAGE_SHIFT;
- int pages = iommu_num_pages(iommu->exclusion_start,
- iommu->exclusion_length,
- PAGE_SIZE);
- dma_ops_reserve_addresses(dma_dom, startpage, pages);
+ for_each_iommu(iommu) {
+ if (iommu->exclusion_start &&
+ iommu->exclusion_start >= dma_dom->aperture[index]->offset
+ && iommu->exclusion_start < dma_dom->aperture_size) {
+ unsigned long startpage;
+ int pages = iommu_num_pages(iommu->exclusion_start,
+ iommu->exclusion_length,
+ PAGE_SIZE);
+ startpage = iommu->exclusion_start >> PAGE_SHIFT;
+ dma_ops_reserve_addresses(dma_dom, startpage, pages);
+ }
}
/*
for (i = dma_dom->aperture[index]->offset;
i < dma_dom->aperture_size;
i += PAGE_SIZE) {
- u64 *pte = fetch_pte(&dma_dom->domain, i);
+ u64 *pte = fetch_pte(&dma_dom->domain, i, PM_MAP_4k);
if (!pte || !IOMMU_PTE_PRESENT(*pte))
continue;
dma_ops_reserve_addresses(dma_dom, i << PAGE_SHIFT, 1);
}
+ update_domain(&dma_dom->domain);
+
return 0;
out_free:
+ update_domain(&dma_dom->domain);
+
free_page((unsigned long)dma_dom->aperture[index]->bitmap);
kfree(dma_dom->aperture[index]);
}
if (unlikely(address == -1))
- address = bad_dma_address;
+ address = DMA_ERROR_CODE;
WARN_ON((address + (PAGE_SIZE*pages)) > dom->aperture_size);
address = (address % APERTURE_RANGE_SIZE) >> PAGE_SHIFT;
- iommu_area_free(range->bitmap, address, pages);
+ bitmap_clear(range->bitmap, address, pages);
}
*
****************************************************************************/
+/*
+ * This function adds a protection domain to the global protection domain list
+ */
+static void add_domain_to_list(struct protection_domain *domain)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&amd_iommu_pd_lock, flags);
+ list_add(&domain->list, &amd_iommu_pd_list);
+ spin_unlock_irqrestore(&amd_iommu_pd_lock, flags);
+}
+
+/*
+ * This function removes a protection domain to the global
+ * protection domain list
+ */
+static void del_domain_from_list(struct protection_domain *domain)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&amd_iommu_pd_lock, flags);
+ list_del(&domain->list);
+ spin_unlock_irqrestore(&amd_iommu_pd_lock, flags);
+}
+
static u16 domain_id_alloc(void)
{
unsigned long flags;
write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
}
-/*
- * Used to reserve address ranges in the aperture (e.g. for exclusion
- * ranges.
- */
-static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
- unsigned long start_page,
- unsigned int pages)
-{
- unsigned int i, last_page = dom->aperture_size >> PAGE_SHIFT;
-
- if (start_page + pages > last_page)
- pages = last_page - start_page;
-
- for (i = start_page; i < start_page + pages; ++i) {
- int index = i / APERTURE_RANGE_PAGES;
- int page = i % APERTURE_RANGE_PAGES;
- __set_bit(page, dom->aperture[index]->bitmap);
- }
-}
-
static void free_pagetable(struct protection_domain *domain)
{
int i, j;
if (!dom)
return;
+ del_domain_from_list(&dom->domain);
+
free_pagetable(&dom->domain);
for (i = 0; i < APERTURE_MAX_RANGES; ++i) {
* It also intializes the page table and the address allocator data
* structures required for the dma_ops interface
*/
-static struct dma_ops_domain *dma_ops_domain_alloc(struct amd_iommu *iommu)
+static struct dma_ops_domain *dma_ops_domain_alloc(void)
{
struct dma_ops_domain *dma_dom;
dma_dom->domain.id = domain_id_alloc();
if (dma_dom->domain.id == 0)
goto free_dma_dom;
- dma_dom->domain.mode = PAGE_MODE_3_LEVEL;
+ INIT_LIST_HEAD(&dma_dom->domain.dev_list);
+ dma_dom->domain.mode = PAGE_MODE_2_LEVEL;
dma_dom->domain.pt_root = (void *)get_zeroed_page(GFP_KERNEL);
dma_dom->domain.flags = PD_DMA_OPS_MASK;
dma_dom->domain.priv = dma_dom;
dma_dom->need_flush = false;
dma_dom->target_dev = 0xffff;
- if (alloc_new_range(iommu, dma_dom, true, GFP_KERNEL))
+ add_domain_to_list(&dma_dom->domain);
+
+ if (alloc_new_range(dma_dom, true, GFP_KERNEL))
goto free_dma_dom;
/*
return domain->flags & PD_DMA_OPS_MASK;
}
+static void set_dte_entry(u16 devid, struct protection_domain *domain)
+{
+ u64 pte_root = virt_to_phys(domain->pt_root);
+
+ pte_root |= (domain->mode & DEV_ENTRY_MODE_MASK)
+ << DEV_ENTRY_MODE_SHIFT;
+ pte_root |= IOMMU_PTE_IR | IOMMU_PTE_IW | IOMMU_PTE_P | IOMMU_PTE_TV;
+
+ amd_iommu_dev_table[devid].data[2] = domain->id;
+ amd_iommu_dev_table[devid].data[1] = upper_32_bits(pte_root);
+ amd_iommu_dev_table[devid].data[0] = lower_32_bits(pte_root);
+}
+
+static void clear_dte_entry(u16 devid)
+{
+ /* remove entry from the device table seen by the hardware */
+ amd_iommu_dev_table[devid].data[0] = IOMMU_PTE_P | IOMMU_PTE_TV;
+ amd_iommu_dev_table[devid].data[1] = 0;
+ amd_iommu_dev_table[devid].data[2] = 0;
+
+ amd_iommu_apply_erratum_63(devid);
+}
+
+static void do_attach(struct device *dev, struct protection_domain *domain)
+{
+ struct iommu_dev_data *dev_data;
+ struct amd_iommu *iommu;
+ u16 devid;
+
+ devid = get_device_id(dev);
+ iommu = amd_iommu_rlookup_table[devid];
+ dev_data = get_dev_data(dev);
+
+ /* Update data structures */
+ dev_data->domain = domain;
+ list_add(&dev_data->list, &domain->dev_list);
+ set_dte_entry(devid, domain);
+
+ /* Do reference counting */
+ domain->dev_iommu[iommu->index] += 1;
+ domain->dev_cnt += 1;
+
+ /* Flush the DTE entry */
+ iommu_flush_device(dev);
+}
+
+static void do_detach(struct device *dev)
+{
+ struct iommu_dev_data *dev_data;
+ struct amd_iommu *iommu;
+ u16 devid;
+
+ devid = get_device_id(dev);
+ iommu = amd_iommu_rlookup_table[devid];
+ dev_data = get_dev_data(dev);
+
+ /* decrease reference counters */
+ dev_data->domain->dev_iommu[iommu->index] -= 1;
+ dev_data->domain->dev_cnt -= 1;
+
+ /* Update data structures */
+ dev_data->domain = NULL;
+ list_del(&dev_data->list);
+ clear_dte_entry(devid);
+
+ /* Flush the DTE entry */
+ iommu_flush_device(dev);
+}
+
/*
- * Find out the protection domain structure for a given PCI device. This
- * will give us the pointer to the page table root for example.
+ * If a device is not yet associated with a domain, this function does
+ * assigns it visible for the hardware
*/
-static struct protection_domain *domain_for_device(u16 devid)
+static int __attach_device(struct device *dev,
+ struct protection_domain *domain)
{
- struct protection_domain *dom;
- unsigned long flags;
+ struct iommu_dev_data *dev_data, *alias_data;
- read_lock_irqsave(&amd_iommu_devtable_lock, flags);
- dom = amd_iommu_pd_table[devid];
- read_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+ dev_data = get_dev_data(dev);
+ alias_data = get_dev_data(dev_data->alias);
- return dom;
+ if (!alias_data)
+ return -EINVAL;
+
+ /* lock domain */
+ spin_lock(&domain->lock);
+
+ /* Some sanity checks */
+ if (alias_data->domain != NULL &&
+ alias_data->domain != domain)
+ return -EBUSY;
+
+ if (dev_data->domain != NULL &&
+ dev_data->domain != domain)
+ return -EBUSY;
+
+ /* Do real assignment */
+ if (dev_data->alias != dev) {
+ alias_data = get_dev_data(dev_data->alias);
+ if (alias_data->domain == NULL)
+ do_attach(dev_data->alias, domain);
+
+ atomic_inc(&alias_data->bind);
+ }
+
+ if (dev_data->domain == NULL)
+ do_attach(dev, domain);
+
+ atomic_inc(&dev_data->bind);
+
+ /* ready */
+ spin_unlock(&domain->lock);
+
+ return 0;
}
/*
* If a device is not yet associated with a domain, this function does
* assigns it visible for the hardware
*/
-static void attach_device(struct amd_iommu *iommu,
- struct protection_domain *domain,
- u16 devid)
+static int attach_device(struct device *dev,
+ struct protection_domain *domain)
{
unsigned long flags;
- u64 pte_root = virt_to_phys(domain->pt_root);
-
- domain->dev_cnt += 1;
-
- pte_root |= (domain->mode & DEV_ENTRY_MODE_MASK)
- << DEV_ENTRY_MODE_SHIFT;
- pte_root |= IOMMU_PTE_IR | IOMMU_PTE_IW | IOMMU_PTE_P | IOMMU_PTE_TV;
+ int ret;
write_lock_irqsave(&amd_iommu_devtable_lock, flags);
- amd_iommu_dev_table[devid].data[0] = lower_32_bits(pte_root);
- amd_iommu_dev_table[devid].data[1] = upper_32_bits(pte_root);
- amd_iommu_dev_table[devid].data[2] = domain->id;
-
- amd_iommu_pd_table[devid] = domain;
+ ret = __attach_device(dev, domain);
write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
- /*
- * We might boot into a crash-kernel here. The crashed kernel
- * left the caches in the IOMMU dirty. So we have to flush
- * here to evict all dirty stuff.
- */
- iommu_queue_inv_dev_entry(iommu, devid);
- iommu_flush_tlb_pde(iommu, domain->id);
+ /*
+ * We might boot into a crash-kernel here. The crashed kernel
+ * left the caches in the IOMMU dirty. So we have to flush
+ * here to evict all dirty stuff.
+ */
+ iommu_flush_tlb_pde(domain);
+
+ return ret;
}
/*
* Removes a device from a protection domain (unlocked)
*/
-static void __detach_device(struct protection_domain *domain, u16 devid)
+static void __detach_device(struct device *dev)
{
+ struct iommu_dev_data *dev_data = get_dev_data(dev);
+ struct iommu_dev_data *alias_data;
+ struct protection_domain *domain;
+ unsigned long flags;
- /* lock domain */
- spin_lock(&domain->lock);
+ BUG_ON(!dev_data->domain);
- /* remove domain from the lookup table */
- amd_iommu_pd_table[devid] = NULL;
+ domain = dev_data->domain;
- /* remove entry from the device table seen by the hardware */
- amd_iommu_dev_table[devid].data[0] = IOMMU_PTE_P | IOMMU_PTE_TV;
- amd_iommu_dev_table[devid].data[1] = 0;
- amd_iommu_dev_table[devid].data[2] = 0;
+ spin_lock_irqsave(&domain->lock, flags);
+
+ if (dev_data->alias != dev) {
+ alias_data = get_dev_data(dev_data->alias);
+ if (atomic_dec_and_test(&alias_data->bind))
+ do_detach(dev_data->alias);
+ }
- /* decrease reference counter */
- domain->dev_cnt -= 1;
+ if (atomic_dec_and_test(&dev_data->bind))
+ do_detach(dev);
- /* ready */
- spin_unlock(&domain->lock);
+ spin_unlock_irqrestore(&domain->lock, flags);
+
+ /*
+ * If we run in passthrough mode the device must be assigned to the
+ * passthrough domain if it is detached from any other domain.
+ * Make sure we can deassign from the pt_domain itself.
+ */
+ if (iommu_pass_through &&
+ (dev_data->domain == NULL && domain != pt_domain))
+ __attach_device(dev, pt_domain);
}
/*
* Removes a device from a protection domain (with devtable_lock held)
*/
-static void detach_device(struct protection_domain *domain, u16 devid)
+static void detach_device(struct device *dev)
{
unsigned long flags;
/* lock device table */
write_lock_irqsave(&amd_iommu_devtable_lock, flags);
- __detach_device(domain, devid);
+ __detach_device(dev);
write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
}
+/*
+ * Find out the protection domain structure for a given PCI device. This
+ * will give us the pointer to the page table root for example.
+ */
+static struct protection_domain *domain_for_device(struct device *dev)
+{
+ struct protection_domain *dom;
+ struct iommu_dev_data *dev_data, *alias_data;
+ unsigned long flags;
+ u16 devid, alias;
+
+ devid = get_device_id(dev);
+ alias = amd_iommu_alias_table[devid];
+ dev_data = get_dev_data(dev);
+ alias_data = get_dev_data(dev_data->alias);
+ if (!alias_data)
+ return NULL;
+
+ read_lock_irqsave(&amd_iommu_devtable_lock, flags);
+ dom = dev_data->domain;
+ if (dom == NULL &&
+ alias_data->domain != NULL) {
+ __attach_device(dev, alias_data->domain);
+ dom = alias_data->domain;
+ }
+
+ read_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+
+ return dom;
+}
+
static int device_change_notifier(struct notifier_block *nb,
unsigned long action, void *data)
{
struct device *dev = data;
- struct pci_dev *pdev = to_pci_dev(dev);
- u16 devid = calc_devid(pdev->bus->number, pdev->devfn);
+ u16 devid;
struct protection_domain *domain;
struct dma_ops_domain *dma_domain;
struct amd_iommu *iommu;
unsigned long flags;
- if (devid > amd_iommu_last_bdf)
- goto out;
-
- devid = amd_iommu_alias_table[devid];
-
- iommu = amd_iommu_rlookup_table[devid];
- if (iommu == NULL)
- goto out;
-
- domain = domain_for_device(devid);
+ if (!check_device(dev))
+ return 0;
- if (domain && !dma_ops_domain(domain))
- WARN_ONCE(1, "AMD IOMMU WARNING: device %s already bound "
- "to a non-dma-ops domain\n", dev_name(dev));
+ devid = get_device_id(dev);
+ iommu = amd_iommu_rlookup_table[devid];
switch (action) {
case BUS_NOTIFY_UNBOUND_DRIVER:
+
+ domain = domain_for_device(dev);
+
if (!domain)
goto out;
- detach_device(domain, devid);
+ if (iommu_pass_through)
+ break;
+ detach_device(dev);
break;
case BUS_NOTIFY_ADD_DEVICE:
+
+ iommu_init_device(dev);
+
+ domain = domain_for_device(dev);
+
/* allocate a protection domain if a device is added */
dma_domain = find_protection_domain(devid);
if (dma_domain)
goto out;
- dma_domain = dma_ops_domain_alloc(iommu);
+ dma_domain = dma_ops_domain_alloc();
if (!dma_domain)
goto out;
dma_domain->target_dev = devid;
spin_unlock_irqrestore(&iommu_pd_list_lock, flags);
break;
+ case BUS_NOTIFY_DEL_DEVICE:
+
+ iommu_uninit_device(dev);
+
default:
goto out;
}
- iommu_queue_inv_dev_entry(iommu, devid);
+ iommu_flush_device(dev);
iommu_completion_wait(iommu);
out:
.notifier_call = device_change_notifier,
};
+void amd_iommu_init_notifier(void)
+{
+ bus_register_notifier(&pci_bus_type, &device_nb);
+}
+
/*****************************************************************************
*
* The next functions belong to the dma_ops mapping/unmapping code.
*****************************************************************************/
/*
- * This function checks if the driver got a valid device from the caller to
- * avoid dereferencing invalid pointers.
- */
-static bool check_device(struct device *dev)
-{
- if (!dev || !dev->dma_mask)
- return false;
-
- return true;
-}
-
-/*
- * In this function the list of preallocated protection domains is traversed to
- * find the domain for a specific device
- */
-static struct dma_ops_domain *find_protection_domain(u16 devid)
-{
- struct dma_ops_domain *entry, *ret = NULL;
- unsigned long flags;
-
- if (list_empty(&iommu_pd_list))
- return NULL;
-
- spin_lock_irqsave(&iommu_pd_list_lock, flags);
-
- list_for_each_entry(entry, &iommu_pd_list, list) {
- if (entry->target_dev == devid) {
- ret = entry;
- break;
- }
- }
-
- spin_unlock_irqrestore(&iommu_pd_list_lock, flags);
-
- return ret;
-}
-
-/*
* In the dma_ops path we only have the struct device. This function
* finds the corresponding IOMMU, the protection domain and the
* requestor id for a given device.
* If the device is not yet associated with a domain this is also done
* in this function.
*/
-static int get_device_resources(struct device *dev,
- struct amd_iommu **iommu,
- struct protection_domain **domain,
- u16 *bdf)
+static struct protection_domain *get_domain(struct device *dev)
{
+ struct protection_domain *domain;
struct dma_ops_domain *dma_dom;
- struct pci_dev *pcidev;
- u16 _bdf;
-
- *iommu = NULL;
- *domain = NULL;
- *bdf = 0xffff;
+ u16 devid = get_device_id(dev);
- if (dev->bus != &pci_bus_type)
- return 0;
-
- pcidev = to_pci_dev(dev);
- _bdf = calc_devid(pcidev->bus->number, pcidev->devfn);
-
- /* device not translated by any IOMMU in the system? */
- if (_bdf > amd_iommu_last_bdf)
- return 0;
+ if (!check_device(dev))
+ return ERR_PTR(-EINVAL);
- *bdf = amd_iommu_alias_table[_bdf];
+ domain = domain_for_device(dev);
+ if (domain != NULL && !dma_ops_domain(domain))
+ return ERR_PTR(-EBUSY);
- *iommu = amd_iommu_rlookup_table[*bdf];
- if (*iommu == NULL)
- return 0;
- *domain = domain_for_device(*bdf);
- if (*domain == NULL) {
- dma_dom = find_protection_domain(*bdf);
- if (!dma_dom)
- dma_dom = (*iommu)->default_dom;
- *domain = &dma_dom->domain;
- attach_device(*iommu, *domain, *bdf);
- DUMP_printk("Using protection domain %d for device %s\n",
- (*domain)->id, dev_name(dev));
- }
+ if (domain != NULL)
+ return domain;
- if (domain_for_device(_bdf) == NULL)
- attach_device(*iommu, *domain, _bdf);
+ /* Device not bount yet - bind it */
+ dma_dom = find_protection_domain(devid);
+ if (!dma_dom)
+ dma_dom = amd_iommu_rlookup_table[devid]->default_dom;
+ attach_device(dev, &dma_dom->domain);
+ DUMP_printk("Using protection domain %d for device %s\n",
+ dma_dom->domain.id, dev_name(dev));
- return 1;
+ return &dma_dom->domain;
}
-/*
- * If the pte_page is not yet allocated this function is called
- */
-static u64* alloc_pte(struct protection_domain *dom,
- unsigned long address, u64 **pte_page, gfp_t gfp)
+static void update_device_table(struct protection_domain *domain)
{
- u64 *pte, *page;
-
- pte = &dom->pt_root[IOMMU_PTE_L2_INDEX(address)];
-
- if (!IOMMU_PTE_PRESENT(*pte)) {
- page = (u64 *)get_zeroed_page(gfp);
- if (!page)
- return NULL;
- *pte = IOMMU_L2_PDE(virt_to_phys(page));
- }
+ struct iommu_dev_data *dev_data;
- pte = IOMMU_PTE_PAGE(*pte);
- pte = &pte[IOMMU_PTE_L1_INDEX(address)];
-
- if (!IOMMU_PTE_PRESENT(*pte)) {
- page = (u64 *)get_zeroed_page(gfp);
- if (!page)
- return NULL;
- *pte = IOMMU_L1_PDE(virt_to_phys(page));
+ list_for_each_entry(dev_data, &domain->dev_list, list) {
+ u16 devid = get_device_id(dev_data->dev);
+ set_dte_entry(devid, domain);
}
+}
- pte = IOMMU_PTE_PAGE(*pte);
-
- if (pte_page)
- *pte_page = pte;
+static void update_domain(struct protection_domain *domain)
+{
+ if (!domain->updated)
+ return;
- pte = &pte[IOMMU_PTE_L0_INDEX(address)];
+ update_device_table(domain);
+ iommu_flush_domain_devices(domain);
+ iommu_flush_tlb_pde(domain);
- return pte;
+ domain->updated = false;
}
/*
pte = aperture->pte_pages[APERTURE_PAGE_INDEX(address)];
if (!pte) {
- pte = alloc_pte(&dom->domain, address, &pte_page, GFP_ATOMIC);
+ pte = alloc_pte(&dom->domain, address, PM_MAP_4k, &pte_page,
+ GFP_ATOMIC);
aperture->pte_pages[APERTURE_PAGE_INDEX(address)] = pte_page;
} else
- pte += IOMMU_PTE_L0_INDEX(address);
+ pte += PM_LEVEL_INDEX(0, address);
+
+ update_domain(&dom->domain);
return pte;
}
* This is the generic map function. It maps one 4kb page at paddr to
* the given address in the DMA address space for the domain.
*/
-static dma_addr_t dma_ops_domain_map(struct amd_iommu *iommu,
- struct dma_ops_domain *dom,
+static dma_addr_t dma_ops_domain_map(struct dma_ops_domain *dom,
unsigned long address,
phys_addr_t paddr,
int direction)
pte = dma_ops_get_pte(dom, address);
if (!pte)
- return bad_dma_address;
+ return DMA_ERROR_CODE;
__pte = paddr | IOMMU_PTE_P | IOMMU_PTE_FC;
/*
* The generic unmapping function for on page in the DMA address space.
*/
-static void dma_ops_domain_unmap(struct amd_iommu *iommu,
- struct dma_ops_domain *dom,
+static void dma_ops_domain_unmap(struct dma_ops_domain *dom,
unsigned long address)
{
struct aperture_range *aperture;
if (!pte)
return;
- pte += IOMMU_PTE_L0_INDEX(address);
+ pte += PM_LEVEL_INDEX(0, address);
WARN_ON(!*pte);
* Must be called with the domain lock held.
*/
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,
retry:
address = dma_ops_alloc_addresses(dev, dma_dom, pages, align_mask,
dma_mask);
- if (unlikely(address == bad_dma_address)) {
+ if (unlikely(address == DMA_ERROR_CODE)) {
/*
* setting next_address here will let the address
* allocator only scan the new allocated range in the
*/
dma_dom->next_address = dma_dom->aperture_size;
- if (alloc_new_range(iommu, dma_dom, false, GFP_ATOMIC))
+ if (alloc_new_range(dma_dom, false, GFP_ATOMIC))
goto out;
/*
- * aperture was sucessfully enlarged by 128 MB, try
+ * aperture was successfully enlarged by 128 MB, try
* allocation again
*/
goto retry;
start = address;
for (i = 0; i < pages; ++i) {
- ret = dma_ops_domain_map(iommu, dma_dom, start, paddr, dir);
- if (ret == bad_dma_address)
+ ret = dma_ops_domain_map(dma_dom, start, paddr, dir);
+ if (ret == DMA_ERROR_CODE)
goto out_unmap;
paddr += PAGE_SIZE;
ADD_STATS_COUNTER(alloced_io_mem, size);
if (unlikely(dma_dom->need_flush && !amd_iommu_unmap_flush)) {
- iommu_flush_tlb(iommu, dma_dom->domain.id);
+ iommu_flush_tlb(&dma_dom->domain);
dma_dom->need_flush = false;
- } else if (unlikely(iommu_has_npcache(iommu)))
- iommu_flush_pages(iommu, dma_dom->domain.id, address, size);
+ } else if (unlikely(amd_iommu_np_cache))
+ iommu_flush_pages(&dma_dom->domain, address, size);
out:
return address;
for (--i; i >= 0; --i) {
start -= PAGE_SIZE;
- dma_ops_domain_unmap(iommu, dma_dom, start);
+ dma_ops_domain_unmap(dma_dom, start);
}
dma_ops_free_addresses(dma_dom, address, pages);
- return bad_dma_address;
+ return DMA_ERROR_CODE;
}
/*
* Does the reverse of the __map_single function. Must be called with
* the domain lock held too
*/
-static void __unmap_single(struct amd_iommu *iommu,
- struct dma_ops_domain *dma_dom,
+static void __unmap_single(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 == bad_dma_address) ||
+ if ((dma_addr == DMA_ERROR_CODE) ||
(dma_addr + size > dma_dom->aperture_size))
return;
start = dma_addr;
for (i = 0; i < pages; ++i) {
- dma_ops_domain_unmap(iommu, dma_dom, start);
+ dma_ops_domain_unmap(dma_dom, start);
start += PAGE_SIZE;
}
dma_ops_free_addresses(dma_dom, dma_addr, pages);
if (amd_iommu_unmap_flush || dma_dom->need_flush) {
- iommu_flush_pages(iommu, dma_dom->domain.id, dma_addr, size);
+ iommu_flush_pages(&dma_dom->domain, dma_addr, size);
dma_dom->need_flush = false;
}
}
struct dma_attrs *attrs)
{
unsigned long flags;
- struct amd_iommu *iommu;
struct protection_domain *domain;
- u16 devid;
dma_addr_t addr;
u64 dma_mask;
phys_addr_t paddr = page_to_phys(page) + offset;
INC_STATS_COUNTER(cnt_map_single);
- if (!check_device(dev))
- return bad_dma_address;
-
- dma_mask = *dev->dma_mask;
-
- get_device_resources(dev, &iommu, &domain, &devid);
-
- if (iommu == NULL || domain == NULL)
- /* device not handled by any AMD IOMMU */
+ domain = get_domain(dev);
+ if (PTR_ERR(domain) == -EINVAL)
return (dma_addr_t)paddr;
+ else if (IS_ERR(domain))
+ return DMA_ERROR_CODE;
- if (!dma_ops_domain(domain))
- return bad_dma_address;
+ dma_mask = *dev->dma_mask;
spin_lock_irqsave(&domain->lock, flags);
- addr = __map_single(dev, iommu, domain->priv, paddr, size, dir, false,
+
+ addr = __map_single(dev, domain->priv, paddr, size, dir, false,
dma_mask);
- if (addr == bad_dma_address)
+ if (addr == DMA_ERROR_CODE)
goto out;
- iommu_completion_wait(iommu);
+ iommu_flush_complete(domain);
out:
spin_unlock_irqrestore(&domain->lock, flags);
enum dma_data_direction dir, struct dma_attrs *attrs)
{
unsigned long flags;
- struct amd_iommu *iommu;
struct protection_domain *domain;
- u16 devid;
INC_STATS_COUNTER(cnt_unmap_single);
- if (!check_device(dev) ||
- !get_device_resources(dev, &iommu, &domain, &devid))
- /* device not handled by any AMD IOMMU */
- return;
-
- if (!dma_ops_domain(domain))
+ domain = get_domain(dev);
+ if (IS_ERR(domain))
return;
spin_lock_irqsave(&domain->lock, flags);
- __unmap_single(iommu, domain->priv, dma_addr, size, dir);
+ __unmap_single(domain->priv, dma_addr, size, dir);
- iommu_completion_wait(iommu);
+ iommu_flush_complete(domain);
spin_unlock_irqrestore(&domain->lock, flags);
}
struct dma_attrs *attrs)
{
unsigned long flags;
- struct amd_iommu *iommu;
struct protection_domain *domain;
- u16 devid;
int i;
struct scatterlist *s;
phys_addr_t paddr;
INC_STATS_COUNTER(cnt_map_sg);
- if (!check_device(dev))
+ domain = get_domain(dev);
+ if (PTR_ERR(domain) == -EINVAL)
+ return map_sg_no_iommu(dev, sglist, nelems, dir);
+ else if (IS_ERR(domain))
return 0;
dma_mask = *dev->dma_mask;
- get_device_resources(dev, &iommu, &domain, &devid);
-
- if (!iommu || !domain)
- return map_sg_no_iommu(dev, sglist, nelems, dir);
-
- if (!dma_ops_domain(domain))
- return 0;
-
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,
+ s->dma_address = __map_single(dev, domain->priv,
paddr, s->length, dir, false,
dma_mask);
goto unmap;
}
- iommu_completion_wait(iommu);
+ iommu_flush_complete(domain);
out:
spin_unlock_irqrestore(&domain->lock, flags);
unmap:
for_each_sg(sglist, s, mapped_elems, i) {
if (s->dma_address)
- __unmap_single(iommu, domain->priv, s->dma_address,
+ __unmap_single(domain->priv, s->dma_address,
s->dma_length, dir);
s->dma_address = s->dma_length = 0;
}
struct dma_attrs *attrs)
{
unsigned long flags;
- struct amd_iommu *iommu;
struct protection_domain *domain;
struct scatterlist *s;
- u16 devid;
int i;
INC_STATS_COUNTER(cnt_unmap_sg);
- if (!check_device(dev) ||
- !get_device_resources(dev, &iommu, &domain, &devid))
- return;
-
- if (!dma_ops_domain(domain))
+ domain = get_domain(dev);
+ if (IS_ERR(domain))
return;
spin_lock_irqsave(&domain->lock, flags);
for_each_sg(sglist, s, nelems, i) {
- __unmap_single(iommu, domain->priv, s->dma_address,
+ __unmap_single(domain->priv, s->dma_address,
s->dma_length, dir);
s->dma_address = s->dma_length = 0;
}
- iommu_completion_wait(iommu);
+ iommu_flush_complete(domain);
spin_unlock_irqrestore(&domain->lock, flags);
}
{
unsigned long flags;
void *virt_addr;
- struct amd_iommu *iommu;
struct protection_domain *domain;
- u16 devid;
phys_addr_t paddr;
u64 dma_mask = dev->coherent_dma_mask;
INC_STATS_COUNTER(cnt_alloc_coherent);
- if (!check_device(dev))
+ domain = get_domain(dev);
+ if (PTR_ERR(domain) == -EINVAL) {
+ virt_addr = (void *)__get_free_pages(flag, get_order(size));
+ *dma_addr = __pa(virt_addr);
+ return virt_addr;
+ } else if (IS_ERR(domain))
return NULL;
- if (!get_device_resources(dev, &iommu, &domain, &devid))
- flag &= ~(__GFP_DMA | __GFP_HIGHMEM | __GFP_DMA32);
+ dma_mask = dev->coherent_dma_mask;
+ flag &= ~(__GFP_DMA | __GFP_HIGHMEM | __GFP_DMA32);
+ flag |= __GFP_ZERO;
- flag |= __GFP_ZERO;
virt_addr = (void *)__get_free_pages(flag, get_order(size));
if (!virt_addr)
return NULL;
paddr = virt_to_phys(virt_addr);
- if (!iommu || !domain) {
- *dma_addr = (dma_addr_t)paddr;
- return virt_addr;
- }
-
- if (!dma_ops_domain(domain))
- goto out_free;
-
if (!dma_mask)
dma_mask = *dev->dma_mask;
spin_lock_irqsave(&domain->lock, flags);
- *dma_addr = __map_single(dev, iommu, domain->priv, paddr,
+ *dma_addr = __map_single(dev, domain->priv, paddr,
size, DMA_BIDIRECTIONAL, true, dma_mask);
- if (*dma_addr == bad_dma_address) {
+ if (*dma_addr == DMA_ERROR_CODE) {
spin_unlock_irqrestore(&domain->lock, flags);
goto out_free;
}
- iommu_completion_wait(iommu);
+ iommu_flush_complete(domain);
spin_unlock_irqrestore(&domain->lock, flags);
void *virt_addr, dma_addr_t dma_addr)
{
unsigned long flags;
- struct amd_iommu *iommu;
struct protection_domain *domain;
- u16 devid;
INC_STATS_COUNTER(cnt_free_coherent);
- if (!check_device(dev))
- return;
-
- get_device_resources(dev, &iommu, &domain, &devid);
-
- if (!iommu || !domain)
- goto free_mem;
-
- if (!dma_ops_domain(domain))
+ domain = get_domain(dev);
+ if (IS_ERR(domain))
goto free_mem;
spin_lock_irqsave(&domain->lock, flags);
- __unmap_single(iommu, domain->priv, dma_addr, size, DMA_BIDIRECTIONAL);
+ __unmap_single(domain->priv, dma_addr, size, DMA_BIDIRECTIONAL);
- iommu_completion_wait(iommu);
+ iommu_flush_complete(domain);
spin_unlock_irqrestore(&domain->lock, flags);
*/
static int amd_iommu_dma_supported(struct device *dev, u64 mask)
{
- u16 bdf;
- struct pci_dev *pcidev;
-
- /* No device or no PCI device */
- if (!dev || dev->bus != &pci_bus_type)
- return 0;
-
- pcidev = to_pci_dev(dev);
-
- bdf = calc_devid(pcidev->bus->number, pcidev->devfn);
-
- /* Out of our scope? */
- if (bdf > amd_iommu_last_bdf)
- return 0;
-
- return 1;
+ return check_device(dev);
}
/*
{
struct pci_dev *dev = NULL;
struct dma_ops_domain *dma_dom;
- struct amd_iommu *iommu;
u16 devid;
while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
- devid = calc_devid(dev->bus->number, dev->devfn);
- if (devid > amd_iommu_last_bdf)
- continue;
- devid = amd_iommu_alias_table[devid];
- if (domain_for_device(devid))
+
+ /* Do we handle this device? */
+ if (!check_device(&dev->dev))
continue;
- iommu = amd_iommu_rlookup_table[devid];
- if (!iommu)
+
+ /* Is there already any domain for it? */
+ if (domain_for_device(&dev->dev))
continue;
- dma_dom = dma_ops_domain_alloc(iommu);
+
+ devid = get_device_id(&dev->dev);
+
+ dma_dom = dma_ops_domain_alloc();
if (!dma_dom)
continue;
init_unity_mappings_for_device(dma_dom, devid);
dma_dom->target_dev = devid;
+ attach_device(&dev->dev, &dma_dom->domain);
+
list_add_tail(&dma_dom->list, &iommu_pd_list);
}
}
/*
* The function which clues the AMD IOMMU driver into dma_ops.
*/
+
+void __init amd_iommu_init_api(void)
+{
+ register_iommu(&amd_iommu_ops);
+}
+
int __init amd_iommu_init_dma_ops(void)
{
struct amd_iommu *iommu;
* protection domain will be assigned to the default one.
*/
for_each_iommu(iommu) {
- iommu->default_dom = dma_ops_domain_alloc(iommu);
+ iommu->default_dom = dma_ops_domain_alloc();
if (iommu->default_dom == NULL)
return -ENOMEM;
iommu->default_dom->domain.flags |= PD_DEFAULT_MASK;
}
/*
- * If device isolation is enabled, pre-allocate the protection
- * domains for each device.
+ * Pre-allocate the protection domains for each device.
*/
- if (amd_iommu_isolate)
- prealloc_protection_domains();
+ prealloc_protection_domains();
iommu_detected = 1;
- force_iommu = 1;
- bad_dma_address = 0;
+ swiotlb = 0;
#ifdef CONFIG_GART_IOMMU
gart_iommu_aperture_disabled = 1;
gart_iommu_aperture = 0;
/* Make the driver finally visible to the drivers */
dma_ops = &amd_iommu_dma_ops;
- register_iommu(&amd_iommu_ops);
-
- bus_register_notifier(&pci_bus_type, &device_nb);
-
amd_iommu_stats_init();
return 0;
static void cleanup_domain(struct protection_domain *domain)
{
+ struct iommu_dev_data *dev_data, *next;
unsigned long flags;
- u16 devid;
write_lock_irqsave(&amd_iommu_devtable_lock, flags);
- for (devid = 0; devid <= amd_iommu_last_bdf; ++devid)
- if (amd_iommu_pd_table[devid] == domain)
- __detach_device(domain, devid);
+ list_for_each_entry_safe(dev_data, next, &domain->dev_list, list) {
+ struct device *dev = dev_data->dev;
+
+ do_detach(dev);
+ atomic_set(&dev_data->bind, 0);
+ }
write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
}
-static int amd_iommu_domain_init(struct iommu_domain *dom)
+static void protection_domain_free(struct protection_domain *domain)
+{
+ if (!domain)
+ return;
+
+ del_domain_from_list(domain);
+
+ if (domain->id)
+ domain_id_free(domain->id);
+
+ kfree(domain);
+}
+
+static struct protection_domain *protection_domain_alloc(void)
{
struct protection_domain *domain;
domain = kzalloc(sizeof(*domain), GFP_KERNEL);
if (!domain)
- return -ENOMEM;
+ return NULL;
spin_lock_init(&domain->lock);
- domain->mode = PAGE_MODE_3_LEVEL;
domain->id = domain_id_alloc();
if (!domain->id)
+ goto out_err;
+ INIT_LIST_HEAD(&domain->dev_list);
+
+ add_domain_to_list(domain);
+
+ return domain;
+
+out_err:
+ kfree(domain);
+
+ return NULL;
+}
+
+static int amd_iommu_domain_init(struct iommu_domain *dom)
+{
+ struct protection_domain *domain;
+
+ domain = protection_domain_alloc();
+ if (!domain)
goto out_free;
+
+ domain->mode = PAGE_MODE_3_LEVEL;
domain->pt_root = (void *)get_zeroed_page(GFP_KERNEL);
if (!domain->pt_root)
goto out_free;
return 0;
out_free:
- kfree(domain);
+ protection_domain_free(domain);
return -ENOMEM;
}
static void amd_iommu_detach_device(struct iommu_domain *dom,
struct device *dev)
{
- struct protection_domain *domain = dom->priv;
+ struct iommu_dev_data *dev_data = dev->archdata.iommu;
struct amd_iommu *iommu;
- struct pci_dev *pdev;
u16 devid;
- if (dev->bus != &pci_bus_type)
+ if (!check_device(dev))
return;
- pdev = to_pci_dev(dev);
+ devid = get_device_id(dev);
- devid = calc_devid(pdev->bus->number, pdev->devfn);
-
- if (devid > 0)
- detach_device(domain, devid);
+ if (dev_data->domain != NULL)
+ detach_device(dev);
iommu = amd_iommu_rlookup_table[devid];
if (!iommu)
return;
- iommu_queue_inv_dev_entry(iommu, devid);
+ iommu_flush_device(dev);
iommu_completion_wait(iommu);
}
struct device *dev)
{
struct protection_domain *domain = dom->priv;
- struct protection_domain *old_domain;
+ struct iommu_dev_data *dev_data;
struct amd_iommu *iommu;
- struct pci_dev *pdev;
+ int ret;
u16 devid;
- if (dev->bus != &pci_bus_type)
+ if (!check_device(dev))
return -EINVAL;
- pdev = to_pci_dev(dev);
+ dev_data = dev->archdata.iommu;
- devid = calc_devid(pdev->bus->number, pdev->devfn);
-
- if (devid >= amd_iommu_last_bdf ||
- devid != amd_iommu_alias_table[devid])
- return -EINVAL;
+ devid = get_device_id(dev);
iommu = amd_iommu_rlookup_table[devid];
if (!iommu)
return -EINVAL;
- old_domain = domain_for_device(devid);
- if (old_domain)
- detach_device(old_domain, devid);
+ if (dev_data->domain)
+ detach_device(dev);
- attach_device(iommu, domain, devid);
+ ret = attach_device(dev, domain);
iommu_completion_wait(iommu);
- return 0;
+ return ret;
}
static int amd_iommu_map_range(struct iommu_domain *dom,
paddr &= PAGE_MASK;
for (i = 0; i < npages; ++i) {
- ret = iommu_map_page(domain, iova, paddr, prot);
+ ret = iommu_map_page(domain, iova, paddr, prot, PM_MAP_4k);
if (ret)
return ret;
iova &= PAGE_MASK;
for (i = 0; i < npages; ++i) {
- iommu_unmap_page(domain, iova);
+ iommu_unmap_page(domain, iova, PM_MAP_4k);
iova += PAGE_SIZE;
}
- iommu_flush_domain(domain->id);
+ iommu_flush_tlb_pde(domain);
}
static phys_addr_t amd_iommu_iova_to_phys(struct iommu_domain *dom,
phys_addr_t paddr;
u64 *pte;
- pte = &domain->pt_root[IOMMU_PTE_L2_INDEX(iova)];
-
- if (!IOMMU_PTE_PRESENT(*pte))
- return 0;
-
- pte = IOMMU_PTE_PAGE(*pte);
- pte = &pte[IOMMU_PTE_L1_INDEX(iova)];
-
- if (!IOMMU_PTE_PRESENT(*pte))
- return 0;
-
- pte = IOMMU_PTE_PAGE(*pte);
- pte = &pte[IOMMU_PTE_L0_INDEX(iova)];
+ pte = fetch_pte(domain, iova, PM_MAP_4k);
- if (!IOMMU_PTE_PRESENT(*pte))
+ if (!pte || !IOMMU_PTE_PRESENT(*pte))
return 0;
paddr = *pte & IOMMU_PAGE_MASK;
.domain_has_cap = amd_iommu_domain_has_cap,
};
+/*****************************************************************************
+ *
+ * The next functions do a basic initialization of IOMMU for pass through
+ * mode
+ *
+ * In passthrough mode the IOMMU is initialized and enabled but not used for
+ * DMA-API translation.
+ *
+ *****************************************************************************/
+
+int __init amd_iommu_init_passthrough(void)
+{
+ struct amd_iommu *iommu;
+ struct pci_dev *dev = NULL;
+ u16 devid;
+
+ /* allocate passthrough domain */
+ pt_domain = protection_domain_alloc();
+ if (!pt_domain)
+ return -ENOMEM;
+
+ pt_domain->mode |= PAGE_MODE_NONE;
+
+ while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
+
+ if (!check_device(&dev->dev))
+ continue;
+
+ devid = get_device_id(&dev->dev);
+
+ iommu = amd_iommu_rlookup_table[devid];
+ if (!iommu)
+ continue;
+
+ attach_device(&dev->dev, pt_domain);
+ }
+
+ pr_info("AMD-Vi: Initialized for Passthrough Mode\n");
+
+ return 0;
+}
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff