Merge branches 'x86/mmio', 'x86/delay', 'x86/idle', 'x86/oprofile', 'x86/debug', 'x86/ptrace' and 'x86/amd-iommu' into x86/devel

Documentation/kernel-parameters.txt

@@ -271,6 +271,17 @@ and is between 256 and 4096 characters. It is defined in the file
 	aic79xx=	[HW,SCSI]
 			See Documentation/scsi/aic79xx.txt.
 
+	amd_iommu=	[HW,X86-84]
+			Pass parameters to the AMD IOMMU driver in the system.
+			Possible values are:
+			isolate - enable device isolation (each device, as far
+			          as possible, will get its own protection
+			          domain)
+	amd_iommu_size= [HW,X86-64]
+			Define the size of the aperture for the AMD IOMMU
+			driver. Possible values are:
+			'32M', '64M' (default), '128M', '256M', '512M', '1G'
+
 	amijoy.map=	[HW,JOY] Amiga joystick support
 			Map of devices attached to JOY0DAT and JOY1DAT
 			Format: <a>,<b>

MAINTAINERS

@@ -376,6 +376,12 @@ L:	linux-geode@lists.infradead.org (moderated for non-subscribers)
 W:	http://www.amd.com/us-en/ConnectivitySolutions/TechnicalResources/0,,50_2334_2452_11363,00.html
 S:	Supported
 
+AMD IOMMU (AMD-VI)
+P:	Joerg Roedel
+M:	joerg.roedel@amd.com
+L:	iommu@lists.linux-foundation.org
+S:	Supported
+
 AMS (Apple Motion Sensor) DRIVER
 P:	Stelian Pop
 M:	stelian@popies.net

arch/x86/Kconfig

@@ -533,6 +533,21 @@ config CALGARY_IOMMU_ENABLED_BY_DEFAULT
 	  Calgary anyway, pass 'iommu=calgary' on the kernel command line.
 	  If unsure, say Y.
 
+config AMD_IOMMU
+	bool "AMD IOMMU support"
+	select SWIOTLB
+	depends on X86_64 && PCI && ACPI
+	help
+	  With this option you can enable support for AMD IOMMU hardware in
+	  your system. An IOMMU is a hardware component which provides
+	  remapping of DMA memory accesses from devices. With an AMD IOMMU you
+	  can isolate the the DMA memory of different devices and protect the
+	  system from misbehaving device drivers or hardware.
+
+	  You can find out if your system has an AMD IOMMU if you look into
+	  your BIOS for an option to enable it or if you have an IVRS ACPI
+	  table.
+
 # need this always selected by IOMMU for the VIA workaround
 config SWIOTLB
 	bool

arch/x86/kernel/Makefile

@@ -99,6 +99,7 @@ ifeq ($(CONFIG_X86_64),y)
 
         obj-$(CONFIG_GART_IOMMU)	+= pci-gart_64.o aperture_64.o
         obj-$(CONFIG_CALGARY_IOMMU)	+= pci-calgary_64.o tce_64.o
+        obj-$(CONFIG_AMD_IOMMU)		+= amd_iommu_init.o amd_iommu.o
         obj-$(CONFIG_SWIOTLB)		+= pci-swiotlb_64.o
 
         obj-$(CONFIG_PCI_MMCONFIG)	+= mmconf-fam10h_64.o

arch/x86/kernel/amd_iommu.c

@@ -0,0 +1,962 @@
+/*
+ * Copyright (C) 2007-2008 Advanced Micro Devices, Inc.
+ * Author: Joerg Roedel <joerg.roedel@amd.com>
+ *         Leo Duran <leo.duran@amd.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ */
+
+#include <linux/pci.h>
+#include <linux/gfp.h>
+#include <linux/bitops.h>
+#include <linux/scatterlist.h>
+#include <linux/iommu-helper.h>
+#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))
+
+#define to_pages(addr, size) \
+	 (round_up(((addr) & ~PAGE_MASK) + (size), PAGE_SIZE) >> PAGE_SHIFT)
+
+static DEFINE_RWLOCK(amd_iommu_devtable_lock);
+
+struct command {
+	u32 data[4];
+};
+
+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;
+	u8 *target;
+
+	tail = readl(iommu->mmio_base + MMIO_CMD_TAIL_OFFSET);
+	target = (iommu->cmd_buf + tail);
+	memcpy_toio(target, cmd, sizeof(*cmd));
+	tail = (tail + sizeof(*cmd)) % iommu->cmd_buf_size;
+	head = readl(iommu->mmio_base + MMIO_CMD_HEAD_OFFSET);
+	if (tail == head)
+		return -ENOMEM;
+	writel(tail, iommu->mmio_base + MMIO_CMD_TAIL_OFFSET);
+
+	return 0;
+}
+
+static int iommu_queue_command(struct amd_iommu *iommu, struct command *cmd)
+{
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&iommu->lock, flags);
+	ret = __iommu_queue_command(iommu, cmd);
+	spin_unlock_irqrestore(&iommu->lock, flags);
+
+	return ret;
+}
+
+static int iommu_completion_wait(struct amd_iommu *iommu)
+{
+	int ret;
+	struct command cmd;
+	volatile u64 ready = 0;
+	unsigned long ready_phys = virt_to_phys(&ready);
+
+	memset(&cmd, 0, sizeof(cmd));
+	cmd.data[0] = LOW_U32(ready_phys) | CMD_COMPL_WAIT_STORE_MASK;
+	cmd.data[1] = HIGH_U32(ready_phys);
+	cmd.data[2] = 1; /* value written to 'ready' */
+	CMD_SET_TYPE(&cmd, CMD_COMPL_WAIT);
+
+	iommu->need_sync = 0;
+
+	ret = iommu_queue_command(iommu, &cmd);
+
+	if (ret)
+		return ret;
+
+	while (!ready)
+		cpu_relax();
+
+	return 0;
+}
+
+static int iommu_queue_inv_dev_entry(struct amd_iommu *iommu, u16 devid)
+{
+	struct command cmd;
+
+	BUG_ON(iommu == NULL);
+
+	memset(&cmd, 0, sizeof(cmd));
+	CMD_SET_TYPE(&cmd, CMD_INV_DEV_ENTRY);
+	cmd.data[0] = devid;
+
+	iommu->need_sync = 1;
+
+	return iommu_queue_command(iommu, &cmd);
+}
+
+static int iommu_queue_inv_iommu_pages(struct amd_iommu *iommu,
+		u64 address, u16 domid, int pde, int s)
+{
+	struct command cmd;
+
+	memset(&cmd, 0, sizeof(cmd));
+	address &= PAGE_MASK;
+	CMD_SET_TYPE(&cmd, CMD_INV_IOMMU_PAGES);
+	cmd.data[1] |= domid;
+	cmd.data[2] = LOW_U32(address);
+	cmd.data[3] = HIGH_U32(address);
+	if (s)
+		cmd.data[2] |= CMD_INV_IOMMU_PAGES_SIZE_MASK;
+	if (pde)
+		cmd.data[2] |= CMD_INV_IOMMU_PAGES_PDE_MASK;
+
+	iommu->need_sync = 1;
+
+	return iommu_queue_command(iommu, &cmd);
+}
+
+static int iommu_flush_pages(struct amd_iommu *iommu, u16 domid,
+		u64 address, size_t size)
+{
+	int s = 0;
+	unsigned pages = to_pages(address, size);
+
+	address &= PAGE_MASK;
+
+	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;
+}
+
+static int iommu_map(struct protection_domain *dom,
+		     unsigned long bus_addr,
+		     unsigned long phys_addr,
+		     int prot)
+{
+	u64 __pte, *pte, *page;
+
+	bus_addr  = PAGE_ALIGN(bus_addr);
+	phys_addr = PAGE_ALIGN(bus_addr);
+
+	/* only support 512GB address spaces for now */
+	if (bus_addr > IOMMU_MAP_SIZE_L3 || !(prot & IOMMU_PROT_MASK))
+		return -EINVAL;
+
+	pte = &dom->pt_root[IOMMU_PTE_L2_INDEX(bus_addr)];
+
+	if (!IOMMU_PTE_PRESENT(*pte)) {
+		page = (u64 *)get_zeroed_page(GFP_KERNEL);
+		if (!page)
+			return -ENOMEM;
+		*pte = IOMMU_L2_PDE(virt_to_phys(page));
+	}
+
+	pte = IOMMU_PTE_PAGE(*pte);
+	pte = &pte[IOMMU_PTE_L1_INDEX(bus_addr)];
+
+	if (!IOMMU_PTE_PRESENT(*pte)) {
+		page = (u64 *)get_zeroed_page(GFP_KERNEL);
+		if (!page)
+			return -ENOMEM;
+		*pte = IOMMU_L1_PDE(virt_to_phys(page));
+	}
+
+	pte = IOMMU_PTE_PAGE(*pte);
+	pte = &pte[IOMMU_PTE_L0_INDEX(bus_addr)];
+
+	if (IOMMU_PTE_PRESENT(*pte))
+		return -EBUSY;
+
+	__pte = phys_addr | IOMMU_PTE_P;
+	if (prot & IOMMU_PROT_IR)
+		__pte |= IOMMU_PTE_IR;
+	if (prot & IOMMU_PROT_IW)
+		__pte |= IOMMU_PTE_IW;
+
+	*pte = __pte;
+
+	return 0;
+}
+
+static int iommu_for_unity_map(struct amd_iommu *iommu,
+			       struct unity_map_entry *entry)
+{
+	u16 bdf, i;
+
+	for (i = entry->devid_start; i <= entry->devid_end; ++i) {
+		bdf = amd_iommu_alias_table[i];
+		if (amd_iommu_rlookup_table[bdf] == iommu)
+			return 1;
+	}
+
+	return 0;
+}
+
+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;
+}
+
+static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
+			     struct unity_map_entry *e)
+{
+	u64 addr;
+	int ret;
+
+	for (addr = e->address_start; addr < e->address_end;
+	     addr += PAGE_SIZE) {
+		ret = iommu_map(&dma_dom->domain, addr, addr, e->prot);
+		if (ret)
+			return ret;
+		/*
+		 * if unity mapping is in aperture range mark the page
+		 * as allocated in the aperture
+		 */
+		if (addr < dma_dom->aperture_size)
+			__set_bit(addr >> PAGE_SHIFT, dma_dom->bitmap);
+	}
+
+	return 0;
+}
+
+static int init_unity_mappings_for_device(struct dma_ops_domain *dma_dom,
+					  u16 devid)
+{
+	struct unity_map_entry *e;
+	int ret;
+
+	list_for_each_entry(e, &amd_iommu_unity_map, list) {
+		if (!(devid >= e->devid_start && devid <= e->devid_end))
+			continue;
+		ret = dma_ops_unity_map(dma_dom, e);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static unsigned long dma_mask_to_pages(unsigned long mask)
+{
+	return (mask >> PAGE_SHIFT) +
+		(PAGE_ALIGN(mask & ~PAGE_MASK) >> PAGE_SHIFT);
+}
+
+static unsigned long dma_ops_alloc_addresses(struct device *dev,
+					     struct dma_ops_domain *dom,
+					     unsigned int pages)
+{
+	unsigned long limit = dma_mask_to_pages(*dev->dma_mask);
+	unsigned long address;
+	unsigned long size = dom->aperture_size >> PAGE_SHIFT;
+	unsigned long boundary_size;
+
+	boundary_size = ALIGN(dma_get_seg_boundary(dev) + 1,
+			PAGE_SIZE) >> PAGE_SHIFT;
+	limit = limit < size ? limit : size;
+
+	if (dom->next_bit >= limit)
+		dom->next_bit = 0;
+
+	address = iommu_area_alloc(dom->bitmap, limit, dom->next_bit, pages,
+			0 , boundary_size, 0);
+	if (address == -1)
+		address = iommu_area_alloc(dom->bitmap, limit, 0, pages,
+				0, boundary_size, 0);
+
+	if (likely(address != -1)) {
+		dom->next_bit = address + pages;
+		address <<= PAGE_SHIFT;
+	} else
+		address = bad_dma_address;
+
+	WARN_ON((address + (PAGE_SIZE*pages)) > dom->aperture_size);
+
+	return address;
+}
+
+static void dma_ops_free_addresses(struct dma_ops_domain *dom,
+				   unsigned long address,
+				   unsigned int pages)
+{
+	address >>= PAGE_SHIFT;
+	iommu_area_free(dom->bitmap, address, pages);
+}
+
+static u16 domain_id_alloc(void)
+{
+	unsigned long flags;
+	int id;
+
+	write_lock_irqsave(&amd_iommu_devtable_lock, flags);
+	id = find_first_zero_bit(amd_iommu_pd_alloc_bitmap, MAX_DOMAIN_ID);
+	BUG_ON(id == 0);
+	if (id > 0 && id < MAX_DOMAIN_ID)
+		__set_bit(id, amd_iommu_pd_alloc_bitmap);
+	else
+		id = 0;
+	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+
+	return id;
+}
+
+static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
+				      unsigned long start_page,
+				      unsigned int pages)
+{
+	unsigned int last_page = dom->aperture_size >> PAGE_SHIFT;
+
+	if (start_page + pages > last_page)
+		pages = last_page - start_page;
+
+	set_bit_string(dom->bitmap, start_page, pages);
+}
+
+static void dma_ops_free_pagetable(struct dma_ops_domain *dma_dom)
+{
+	int i, j;
+	u64 *p1, *p2, *p3;
+
+	p1 = dma_dom->domain.pt_root;
+
+	if (!p1)
+		return;
+
+	for (i = 0; i < 512; ++i) {
+		if (!IOMMU_PTE_PRESENT(p1[i]))
+			continue;
+
+		p2 = IOMMU_PTE_PAGE(p1[i]);
+		for (j = 0; j < 512; ++i) {
+			if (!IOMMU_PTE_PRESENT(p2[j]))
+				continue;
+			p3 = IOMMU_PTE_PAGE(p2[j]);
+			free_page((unsigned long)p3);
+		}
+
+		free_page((unsigned long)p2);
+	}
+
+	free_page((unsigned long)p1);
+}
+
+static void dma_ops_domain_free(struct dma_ops_domain *dom)
+{
+	if (!dom)
+		return;
+
+	dma_ops_free_pagetable(dom);
+
+	kfree(dom->pte_pages);
+
+	kfree(dom->bitmap);
+
+	kfree(dom);
+}
+
+static struct dma_ops_domain *dma_ops_domain_alloc(struct amd_iommu *iommu,
+						   unsigned order)
+{
+	struct dma_ops_domain *dma_dom;
+	unsigned i, num_pte_pages;
+	u64 *l2_pde;
+	u64 address;
+
+	/*
+	 * Currently the DMA aperture must be between 32 MB and 1GB in size
+	 */
+	if ((order < 25) || (order > 30))
+		return NULL;
+
+	dma_dom = kzalloc(sizeof(struct dma_ops_domain), GFP_KERNEL);
+	if (!dma_dom)
+		return NULL;
+
+	spin_lock_init(&dma_dom->domain.lock);
+
+	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;
+	dma_dom->domain.pt_root = (void *)get_zeroed_page(GFP_KERNEL);
+	dma_dom->domain.priv = dma_dom;
+	if (!dma_dom->domain.pt_root)
+		goto free_dma_dom;
+	dma_dom->aperture_size = (1ULL << order);
+	dma_dom->bitmap = kzalloc(dma_dom->aperture_size / (PAGE_SIZE * 8),
+				  GFP_KERNEL);
+	if (!dma_dom->bitmap)
+		goto free_dma_dom;
+	/*
+	 * mark the first page as allocated so we never return 0 as
+	 * a valid dma-address. So we can use 0 as error value
+	 */
+	dma_dom->bitmap[0] = 1;
+	dma_dom->next_bit = 0;
+
+	if (iommu->exclusion_start &&
+	    iommu->exclusion_start < dma_dom->aperture_size) {
+		unsigned long startpage = iommu->exclusion_start >> PAGE_SHIFT;
+		int pages = to_pages(iommu->exclusion_start,
+				iommu->exclusion_length);
+		dma_ops_reserve_addresses(dma_dom, startpage, pages);
+	}
+
+	num_pte_pages = dma_dom->aperture_size / (PAGE_SIZE * 512);
+	dma_dom->pte_pages = kzalloc(num_pte_pages * sizeof(void *),
+			GFP_KERNEL);
+	if (!dma_dom->pte_pages)
+		goto free_dma_dom;
+
+	l2_pde = (u64 *)get_zeroed_page(GFP_KERNEL);
+	if (l2_pde == NULL)
+		goto free_dma_dom;
+
+	dma_dom->domain.pt_root[0] = IOMMU_L2_PDE(virt_to_phys(l2_pde));
+
+	for (i = 0; i < num_pte_pages; ++i) {
+		dma_dom->pte_pages[i] = (u64 *)get_zeroed_page(GFP_KERNEL);
+		if (!dma_dom->pte_pages[i])
+			goto free_dma_dom;
+		address = virt_to_phys(dma_dom->pte_pages[i]);
+		l2_pde[i] = IOMMU_L1_PDE(address);
+	}
+
+	return dma_dom;
+
+free_dma_dom:
+	dma_ops_domain_free(dma_dom);
+
+	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;
+}

arch/x86/kernel/amd_iommu_init.c

@@ -0,0 +1,875 @@
+/*
+ * Copyright (C) 2007-2008 Advanced Micro Devices, Inc.
+ * Author: Joerg Roedel <joerg.roedel@amd.com>
+ *         Leo Duran <leo.duran@amd.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ */
+
+#include <linux/pci.h>
+#include <linux/acpi.h>
+#include <linux/gfp.h>
+#include <linux/list.h>
+#include <linux/sysdev.h>
+#include <asm/pci-direct.h>
+#include <asm/amd_iommu_types.h>
+#include <asm/amd_iommu.h>
+#include <asm/gart.h>
+
+/*
+ * definitions for the ACPI scanning code
+ */
+#define UPDATE_LAST_BDF(x) do {\
+	if ((x) > amd_iommu_last_bdf) \
+		amd_iommu_last_bdf = (x); \
+	} while (0);
+
+#define DEVID(bus, devfn) (((bus) << 8) | (devfn))
+#define PCI_BUS(x) (((x) >> 8) & 0xff)
+#define IVRS_HEADER_LENGTH 48
+#define TBL_SIZE(x) (1 << (PAGE_SHIFT + get_order(amd_iommu_last_bdf * (x))))
+
+#define ACPI_IVHD_TYPE                  0x10
+#define ACPI_IVMD_TYPE_ALL              0x20
+#define ACPI_IVMD_TYPE                  0x21
+#define ACPI_IVMD_TYPE_RANGE            0x22
+
+#define IVHD_DEV_ALL                    0x01
+#define IVHD_DEV_SELECT                 0x02
+#define IVHD_DEV_SELECT_RANGE_START     0x03
+#define IVHD_DEV_RANGE_END              0x04
+#define IVHD_DEV_ALIAS                  0x42
+#define IVHD_DEV_ALIAS_RANGE            0x43
+#define IVHD_DEV_EXT_SELECT             0x46
+#define IVHD_DEV_EXT_SELECT_RANGE       0x47
+
+#define IVHD_FLAG_HT_TUN_EN             0x00
+#define IVHD_FLAG_PASSPW_EN             0x01
+#define IVHD_FLAG_RESPASSPW_EN          0x02
+#define IVHD_FLAG_ISOC_EN               0x03
+
+#define IVMD_FLAG_EXCL_RANGE            0x08
+#define IVMD_FLAG_UNITY_MAP             0x01
+
+#define ACPI_DEVFLAG_INITPASS           0x01
+#define ACPI_DEVFLAG_EXTINT             0x02
+#define ACPI_DEVFLAG_NMI                0x04
+#define ACPI_DEVFLAG_SYSMGT1            0x10
+#define ACPI_DEVFLAG_SYSMGT2            0x20
+#define ACPI_DEVFLAG_LINT0              0x40
+#define ACPI_DEVFLAG_LINT1              0x80
+#define ACPI_DEVFLAG_ATSDIS             0x10000000
+
+struct ivhd_header {
+	u8 type;
+	u8 flags;
+	u16 length;
+	u16 devid;
+	u16 cap_ptr;
+	u64 mmio_phys;
+	u16 pci_seg;
+	u16 info;
+	u32 reserved;
+} __attribute__((packed));
+
+struct ivhd_entry {
+	u8 type;
+	u16 devid;
+	u8 flags;
+	u32 ext;
+} __attribute__((packed));
+
+struct ivmd_header {
+	u8 type;
+	u8 flags;
+	u16 length;
+	u16 devid;
+	u16 aux;
+	u64 resv;
+	u64 range_start;
+	u64 range_length;
+} __attribute__((packed));
+
+static int __initdata amd_iommu_detected;
+
+u16 amd_iommu_last_bdf;
+struct list_head amd_iommu_unity_map;
+unsigned amd_iommu_aperture_order = 26;
+int amd_iommu_isolate;
+
+struct list_head amd_iommu_list;
+struct dev_table_entry *amd_iommu_dev_table;
+u16 *amd_iommu_alias_table;
+struct amd_iommu **amd_iommu_rlookup_table;
+struct protection_domain **amd_iommu_pd_table;
+unsigned long *amd_iommu_pd_alloc_bitmap;
+
+static u32 dev_table_size;
+static u32 alias_table_size;
+static u32 rlookup_table_size;
+
+static void __init iommu_set_exclusion_range(struct amd_iommu *iommu)
+{
+	u64 start = iommu->exclusion_start & PAGE_MASK;
+	u64 limit = (start + iommu->exclusion_length) & PAGE_MASK;
+	u64 entry;
+
+	if (!iommu->exclusion_start)
+		return;
+
+	entry = start | MMIO_EXCL_ENABLE_MASK;
+	memcpy_toio(iommu->mmio_base + MMIO_EXCL_BASE_OFFSET,
+			&entry, sizeof(entry));
+
+	entry = limit;
+	memcpy_toio(iommu->mmio_base + MMIO_EXCL_LIMIT_OFFSET,
+			&entry, sizeof(entry));
+}
+
+static void __init iommu_set_device_table(struct amd_iommu *iommu)
+{
+	u32 entry;
+
+	BUG_ON(iommu->mmio_base == NULL);
+
+	entry = virt_to_phys(amd_iommu_dev_table);
+	entry |= (dev_table_size >> 12) - 1;
+	memcpy_toio(iommu->mmio_base + MMIO_DEV_TABLE_OFFSET,
+			&entry, sizeof(entry));
+}
+
+static void __init iommu_feature_enable(struct amd_iommu *iommu, u8 bit)
+{
+	u32 ctrl;
+
+	ctrl = readl(iommu->mmio_base + MMIO_CONTROL_OFFSET);
+	ctrl |= (1 << bit);
+	writel(ctrl, iommu->mmio_base + MMIO_CONTROL_OFFSET);
+}
+
+static void __init iommu_feature_disable(struct amd_iommu *iommu, u8 bit)
+{
+	u32 ctrl;
+
+	ctrl = (u64)readl(iommu->mmio_base + MMIO_CONTROL_OFFSET);
+	ctrl &= ~(1 << bit);
+	writel(ctrl, iommu->mmio_base + MMIO_CONTROL_OFFSET);
+}
+
+void __init iommu_enable(struct amd_iommu *iommu)
+{
+	printk(KERN_INFO "AMD IOMMU: Enabling IOMMU at ");
+	print_devid(iommu->devid, 0);
+	printk(" cap 0x%hx\n", iommu->cap_ptr);
+
+	iommu_feature_enable(iommu, CONTROL_IOMMU_EN);
+}
+
+static u8 * __init iommu_map_mmio_space(u64 address)
+{
+	u8 *ret;
+
+	if (!request_mem_region(address, MMIO_REGION_LENGTH, "amd_iommu"))
+		return NULL;
+
+	ret = ioremap_nocache(address, MMIO_REGION_LENGTH);
+	if (ret != NULL)
+		return ret;
+
+	release_mem_region(address, MMIO_REGION_LENGTH);
+
+	return NULL;
+}
+
+static void __init iommu_unmap_mmio_space(struct amd_iommu *iommu)
+{
+	if (iommu->mmio_base)
+		iounmap(iommu->mmio_base);
+	release_mem_region(iommu->mmio_phys, MMIO_REGION_LENGTH);
+}
+
+static int __init find_last_devid_on_pci(int bus, int dev, int fn, int cap_ptr)
+{
+	u32 cap;
+
+	cap = read_pci_config(bus, dev, fn, cap_ptr+MMIO_RANGE_OFFSET);
+	UPDATE_LAST_BDF(DEVID(MMIO_GET_BUS(cap), MMIO_GET_LD(cap)));
+
+	return 0;
+}
+
+static int __init find_last_devid_from_ivhd(struct ivhd_header *h)
+{
+	u8 *p = (void *)h, *end = (void *)h;
+	struct ivhd_entry *dev;
+
+	p += sizeof(*h);
+	end += h->length;
+
+	find_last_devid_on_pci(PCI_BUS(h->devid),
+			PCI_SLOT(h->devid),
+			PCI_FUNC(h->devid),
+			h->cap_ptr);
+
+	while (p < end) {
+		dev = (struct ivhd_entry *)p;
+		switch (dev->type) {
+		case IVHD_DEV_SELECT:
+		case IVHD_DEV_RANGE_END:
+		case IVHD_DEV_ALIAS:
+		case IVHD_DEV_EXT_SELECT:
+			UPDATE_LAST_BDF(dev->devid);
+			break;
+		default:
+			break;
+		}
+		p += 0x04 << (*p >> 6);
+	}
+
+	WARN_ON(p != end);
+
+	return 0;
+}
+
+static int __init find_last_devid_acpi(struct acpi_table_header *table)
+{
+	int i;
+	u8 checksum = 0, *p = (u8 *)table, *end = (u8 *)table;
+	struct ivhd_header *h;
+
+	/*
+	 * Validate checksum here so we don't need to do it when
+	 * we actually parse the table
+	 */
+	for (i = 0; i < table->length; ++i)
+		checksum += p[i];
+	if (checksum != 0)
+		/* ACPI table corrupt */
+		return -ENODEV;
+
+	p += IVRS_HEADER_LENGTH;
+
+	end += table->length;
+	while (p < end) {
+		h = (struct ivhd_header *)p;
+		switch (h->type) {
+		case ACPI_IVHD_TYPE:
+			find_last_devid_from_ivhd(h);
+			break;
+		default:
+			break;
+		}
+		p += h->length;
+	}
+	WARN_ON(p != end);
+
+	return 0;
+}
+
+static u8 * __init alloc_command_buffer(struct amd_iommu *iommu)
+{
+	u8 *cmd_buf = (u8 *)__get_free_pages(GFP_KERNEL,
+			get_order(CMD_BUFFER_SIZE));
+	u64 entry = 0;
+
+	if (cmd_buf == NULL)
+		return NULL;
+
+	iommu->cmd_buf_size = CMD_BUFFER_SIZE;
+
+	memset(cmd_buf, 0, CMD_BUFFER_SIZE);
+
+	entry = (u64)virt_to_phys(cmd_buf);
+	entry |= MMIO_CMD_SIZE_512;
+	memcpy_toio(iommu->mmio_base + MMIO_CMD_BUF_OFFSET,
+			&entry, sizeof(entry));
+
+	iommu_feature_enable(iommu, CONTROL_CMDBUF_EN);
+
+	return cmd_buf;
+}
+
+static void __init free_command_buffer(struct amd_iommu *iommu)
+{
+	if (iommu->cmd_buf)
+		free_pages((unsigned long)iommu->cmd_buf,
+				get_order(CMD_BUFFER_SIZE));
+}
+
+static void set_dev_entry_bit(u16 devid, u8 bit)
+{
+	int i = (bit >> 5) & 0x07;
+	int _bit = bit & 0x1f;
+
+	amd_iommu_dev_table[devid].data[i] |= (1 << _bit);
+}
+
+static void __init set_dev_entry_from_acpi(u16 devid, u32 flags, u32 ext_flags)
+{
+	if (flags & ACPI_DEVFLAG_INITPASS)
+		set_dev_entry_bit(devid, DEV_ENTRY_INIT_PASS);
+	if (flags & ACPI_DEVFLAG_EXTINT)
+		set_dev_entry_bit(devid, DEV_ENTRY_EINT_PASS);
+	if (flags & ACPI_DEVFLAG_NMI)
+		set_dev_entry_bit(devid, DEV_ENTRY_NMI_PASS);
+	if (flags & ACPI_DEVFLAG_SYSMGT1)
+		set_dev_entry_bit(devid, DEV_ENTRY_SYSMGT1);
+	if (flags & ACPI_DEVFLAG_SYSMGT2)
+		set_dev_entry_bit(devid, DEV_ENTRY_SYSMGT2);
+	if (flags & ACPI_DEVFLAG_LINT0)
+		set_dev_entry_bit(devid, DEV_ENTRY_LINT0_PASS);
+	if (flags & ACPI_DEVFLAG_LINT1)
+		set_dev_entry_bit(devid, DEV_ENTRY_LINT1_PASS);
+}
+
+static void __init set_iommu_for_device(struct amd_iommu *iommu, u16 devid)
+{
+	amd_iommu_rlookup_table[devid] = iommu;
+}
+
+static void __init set_device_exclusion_range(u16 devid, struct ivmd_header *m)
+{
+	struct amd_iommu *iommu = amd_iommu_rlookup_table[devid];
+
+	if (!(m->flags & IVMD_FLAG_EXCL_RANGE))
+		return;
+
+	if (iommu) {
+		set_dev_entry_bit(m->devid, DEV_ENTRY_EX);
+		iommu->exclusion_start = m->range_start;
+		iommu->exclusion_length = m->range_length;
+	}
+}
+
+static void __init init_iommu_from_pci(struct amd_iommu *iommu)
+{
+	int bus = PCI_BUS(iommu->devid);
+	int dev = PCI_SLOT(iommu->devid);
+	int fn  = PCI_FUNC(iommu->devid);
+	int cap_ptr = iommu->cap_ptr;
+	u32 range;
+
+	iommu->cap = read_pci_config(bus, dev, fn, cap_ptr+MMIO_CAP_HDR_OFFSET);
+
+	range = read_pci_config(bus, dev, fn, cap_ptr+MMIO_RANGE_OFFSET);
+	iommu->first_device = DEVID(MMIO_GET_BUS(range), MMIO_GET_FD(range));
+	iommu->last_device = DEVID(MMIO_GET_BUS(range), MMIO_GET_LD(range));
+}
+
+static void __init init_iommu_from_acpi(struct amd_iommu *iommu,
+					struct ivhd_header *h)
+{
+	u8 *p = (u8 *)h;
+	u8 *end = p, flags = 0;
+	u16 dev_i, devid = 0, devid_start = 0, devid_to = 0;
+	u32 ext_flags = 0;
+	bool alias = 0;
+	struct ivhd_entry *e;
+
+	/*
+	 * First set the recommended feature enable bits from ACPI
+	 * into the IOMMU control registers
+	 */
+	h->flags & IVHD_FLAG_HT_TUN_EN ?
+		iommu_feature_enable(iommu, CONTROL_HT_TUN_EN) :
+		iommu_feature_disable(iommu, CONTROL_HT_TUN_EN);
+
+	h->flags & IVHD_FLAG_PASSPW_EN ?
+		iommu_feature_enable(iommu, CONTROL_PASSPW_EN) :
+		iommu_feature_disable(iommu, CONTROL_PASSPW_EN);
+
+	h->flags & IVHD_FLAG_RESPASSPW_EN ?
+		iommu_feature_enable(iommu, CONTROL_RESPASSPW_EN) :
+		iommu_feature_disable(iommu, CONTROL_RESPASSPW_EN);
+
+	h->flags & IVHD_FLAG_ISOC_EN ?
+		iommu_feature_enable(iommu, CONTROL_ISOC_EN) :
+		iommu_feature_disable(iommu, CONTROL_ISOC_EN);
+
+	/*
+	 * make IOMMU memory accesses cache coherent
+	 */
+	iommu_feature_enable(iommu, CONTROL_COHERENT_EN);
+
+	/*
+	 * Done. Now parse the device entries
+	 */
+	p += sizeof(struct ivhd_header);
+	end += h->length;
+
+	while (p < end) {
+		e = (struct ivhd_entry *)p;
+		switch (e->type) {
+		case IVHD_DEV_ALL:
+			for (dev_i = iommu->first_device;
+					dev_i <= iommu->last_device; ++dev_i)
+				set_dev_entry_from_acpi(dev_i, e->flags, 0);
+			break;
+		case IVHD_DEV_SELECT:
+			devid = e->devid;
+			set_dev_entry_from_acpi(devid, e->flags, 0);
+			break;
+		case IVHD_DEV_SELECT_RANGE_START:
+			devid_start = e->devid;
+			flags = e->flags;
+			ext_flags = 0;
+			alias = 0;
+			break;
+		case IVHD_DEV_ALIAS:
+			devid = e->devid;
+			devid_to = e->ext >> 8;
+			set_dev_entry_from_acpi(devid, e->flags, 0);
+			amd_iommu_alias_table[devid] = devid_to;
+			break;
+		case IVHD_DEV_ALIAS_RANGE:
+			devid_start = e->devid;
+			flags = e->flags;
+			devid_to = e->ext >> 8;
+			ext_flags = 0;
+			alias = 1;
+			break;
+		case IVHD_DEV_EXT_SELECT:
+			devid = e->devid;
+			set_dev_entry_from_acpi(devid, e->flags, e->ext);
+			break;
+		case IVHD_DEV_EXT_SELECT_RANGE:
+			devid_start = e->devid;
+			flags = e->flags;
+			ext_flags = e->ext;
+			alias = 0;
+			break;
+		case IVHD_DEV_RANGE_END:
+			devid = e->devid;
+			for (dev_i = devid_start; dev_i <= devid; ++dev_i) {
+				if (alias)
+					amd_iommu_alias_table[dev_i] = devid_to;
+				set_dev_entry_from_acpi(
+						amd_iommu_alias_table[dev_i],
+						flags, ext_flags);
+			}
+			break;
+		default:
+			break;
+		}
+
+		p += 0x04 << (e->type >> 6);
+	}
+}
+
+static int __init init_iommu_devices(struct amd_iommu *iommu)
+{
+	u16 i;
+
+	for (i = iommu->first_device; i <= iommu->last_device; ++i)
+		set_iommu_for_device(iommu, i);
+
+	return 0;
+}
+
+static void __init free_iommu_one(struct amd_iommu *iommu)
+{
+	free_command_buffer(iommu);
+	iommu_unmap_mmio_space(iommu);
+}
+
+static void __init free_iommu_all(void)
+{
+	struct amd_iommu *iommu, *next;
+
+	list_for_each_entry_safe(iommu, next, &amd_iommu_list, list) {
+		list_del(&iommu->list);
+		free_iommu_one(iommu);
+		kfree(iommu);
+	}
+}
+
+static int __init init_iommu_one(struct amd_iommu *iommu, struct ivhd_header *h)
+{
+	spin_lock_init(&iommu->lock);
+	list_add_tail(&iommu->list, &amd_iommu_list);
+
+	/*
+	 * Copy data from ACPI table entry to the iommu struct
+	 */
+	iommu->devid = h->devid;
+	iommu->cap_ptr = h->cap_ptr;
+	iommu->mmio_phys = h->mmio_phys;
+	iommu->mmio_base = iommu_map_mmio_space(h->mmio_phys);
+	if (!iommu->mmio_base)
+		return -ENOMEM;
+
+	iommu_set_device_table(iommu);
+	iommu->cmd_buf = alloc_command_buffer(iommu);
+	if (!iommu->cmd_buf)
+		return -ENOMEM;
+
+	init_iommu_from_pci(iommu);
+	init_iommu_from_acpi(iommu, h);
+	init_iommu_devices(iommu);
+
+	return 0;
+}
+
+static int __init init_iommu_all(struct acpi_table_header *table)
+{
+	u8 *p = (u8 *)table, *end = (u8 *)table;
+	struct ivhd_header *h;
+	struct amd_iommu *iommu;
+	int ret;
+
+	INIT_LIST_HEAD(&amd_iommu_list);
+
+	end += table->length;
+	p += IVRS_HEADER_LENGTH;
+
+	while (p < end) {
+		h = (struct ivhd_header *)p;
+		switch (*p) {
+		case ACPI_IVHD_TYPE:
+			iommu = kzalloc(sizeof(struct amd_iommu), GFP_KERNEL);
+			if (iommu == NULL)
+				return -ENOMEM;
+			ret = init_iommu_one(iommu, h);
+			if (ret)
+				return ret;
+			break;
+		default:
+			break;
+		}
+		p += h->length;
+
+	}
+	WARN_ON(p != end);
+
+	return 0;
+}
+
+static void __init free_unity_maps(void)
+{
+	struct unity_map_entry *entry, *next;
+
+	list_for_each_entry_safe(entry, next, &amd_iommu_unity_map, list) {
+		list_del(&entry->list);
+		kfree(entry);
+	}
+}
+
+static int __init init_exclusion_range(struct ivmd_header *m)
+{
+	int i;
+
+	switch (m->type) {
+	case ACPI_IVMD_TYPE:
+		set_device_exclusion_range(m->devid, m);
+		break;
+	case ACPI_IVMD_TYPE_ALL:
+		for (i = 0; i < amd_iommu_last_bdf; ++i)
+			set_device_exclusion_range(i, m);
+		break;
+	case ACPI_IVMD_TYPE_RANGE:
+		for (i = m->devid; i <= m->aux; ++i)
+			set_device_exclusion_range(i, m);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int __init init_unity_map_range(struct ivmd_header *m)
+{
+	struct unity_map_entry *e = 0;
+
+	e = kzalloc(sizeof(*e), GFP_KERNEL);
+	if (e == NULL)
+		return -ENOMEM;
+
+	switch (m->type) {
+	default:
+	case ACPI_IVMD_TYPE:
+		e->devid_start = e->devid_end = m->devid;
+		break;
+	case ACPI_IVMD_TYPE_ALL:
+		e->devid_start = 0;
+		e->devid_end = amd_iommu_last_bdf;
+		break;
+	case ACPI_IVMD_TYPE_RANGE:
+		e->devid_start = m->devid;
+		e->devid_end = m->aux;
+		break;
+	}
+	e->address_start = PAGE_ALIGN(m->range_start);
+	e->address_end = e->address_start + PAGE_ALIGN(m->range_length);
+	e->prot = m->flags >> 1;
+
+	list_add_tail(&e->list, &amd_iommu_unity_map);
+
+	return 0;
+}
+
+static int __init init_memory_definitions(struct acpi_table_header *table)
+{
+	u8 *p = (u8 *)table, *end = (u8 *)table;
+	struct ivmd_header *m;
+
+	INIT_LIST_HEAD(&amd_iommu_unity_map);
+
+	end += table->length;
+	p += IVRS_HEADER_LENGTH;
+
+	while (p < end) {
+		m = (struct ivmd_header *)p;
+		if (m->flags & IVMD_FLAG_EXCL_RANGE)
+			init_exclusion_range(m);
+		else if (m->flags & IVMD_FLAG_UNITY_MAP)
+			init_unity_map_range(m);
+
+		p += m->length;
+	}
+
+	return 0;
+}
+
+static void __init enable_iommus(void)
+{
+	struct amd_iommu *iommu;
+
+	list_for_each_entry(iommu, &amd_iommu_list, list) {
+		iommu_set_exclusion_range(iommu);
+		iommu_enable(iommu);
+	}
+}
+
+/*
+ * Suspend/Resume support
+ * disable suspend until real resume implemented
+ */
+
+static int amd_iommu_resume(struct sys_device *dev)
+{
+	return 0;
+}
+
+static int amd_iommu_suspend(struct sys_device *dev, pm_message_t state)
+{
+	return -EINVAL;
+}
+
+static struct sysdev_class amd_iommu_sysdev_class = {
+	.name = "amd_iommu",
+	.suspend = amd_iommu_suspend,
+	.resume = amd_iommu_resume,
+};
+
+static struct sys_device device_amd_iommu = {
+	.id = 0,
+	.cls = &amd_iommu_sysdev_class,
+};
+
+int __init amd_iommu_init(void)
+{
+	int i, ret = 0;
+
+
+	if (no_iommu) {
+		printk(KERN_INFO "AMD IOMMU disabled by kernel command line\n");
+		return 0;
+	}
+
+	if (!amd_iommu_detected)
+		return -ENODEV;
+
+	/*
+	 * First parse ACPI tables to find the largest Bus/Dev/Func
+	 * we need to handle. Upon this information the shared data
+	 * structures for the IOMMUs in the system will be allocated
+	 */
+	if (acpi_table_parse("IVRS", find_last_devid_acpi) != 0)
+		return -ENODEV;
+
+	dev_table_size     = TBL_SIZE(DEV_TABLE_ENTRY_SIZE);
+	alias_table_size   = TBL_SIZE(ALIAS_TABLE_ENTRY_SIZE);
+	rlookup_table_size = TBL_SIZE(RLOOKUP_TABLE_ENTRY_SIZE);
+
+	ret = -ENOMEM;
+
+	/* Device table - directly used by all IOMMUs */
+	amd_iommu_dev_table = (void *)__get_free_pages(GFP_KERNEL,
+				      get_order(dev_table_size));
+	if (amd_iommu_dev_table == NULL)
+		goto out;
+
+	/*
+	 * Alias table - map PCI Bus/Dev/Func to Bus/Dev/Func the
+	 * IOMMU see for that device
+	 */
+	amd_iommu_alias_table = (void *)__get_free_pages(GFP_KERNEL,
+			get_order(alias_table_size));
+	if (amd_iommu_alias_table == NULL)
+		goto free;
+
+	/* IOMMU rlookup table - find the IOMMU for a specific device */
+	amd_iommu_rlookup_table = (void *)__get_free_pages(GFP_KERNEL,
+			get_order(rlookup_table_size));
+	if (amd_iommu_rlookup_table == NULL)
+		goto free;
+
+	/*
+	 * Protection Domain table - maps devices to protection domains
+	 * This table has the same size as the rlookup_table
+	 */
+	amd_iommu_pd_table = (void *)__get_free_pages(GFP_KERNEL,
+				     get_order(rlookup_table_size));
+	if (amd_iommu_pd_table == NULL)
+		goto free;
+
+	amd_iommu_pd_alloc_bitmap = (void *)__get_free_pages(GFP_KERNEL,
+					    get_order(MAX_DOMAIN_ID/8));
+	if (amd_iommu_pd_alloc_bitmap == NULL)
+		goto free;
+
+	/*
+	 * memory is allocated now; initialize the device table with all zeroes
+	 * and let all alias entries point to itself
+	 */
+	memset(amd_iommu_dev_table, 0, dev_table_size);
+	for (i = 0; i < amd_iommu_last_bdf; ++i)
+		amd_iommu_alias_table[i] = i;
+
+	memset(amd_iommu_pd_table, 0, rlookup_table_size);
+	memset(amd_iommu_pd_alloc_bitmap, 0, MAX_DOMAIN_ID / 8);
+
+	/*
+	 * never allocate domain 0 because its used as the non-allocated and
+	 * error value placeholder
+	 */
+	amd_iommu_pd_alloc_bitmap[0] = 1;
+
+	/*
+	 * now the data structures are allocated and basically initialized
+	 * start the real acpi table scan
+	 */
+	ret = -ENODEV;
+	if (acpi_table_parse("IVRS", init_iommu_all) != 0)
+		goto free;
+
+	if (acpi_table_parse("IVRS", init_memory_definitions) != 0)
+		goto free;
+
+	ret = amd_iommu_init_dma_ops();
+	if (ret)
+		goto free;
+
+	ret = sysdev_class_register(&amd_iommu_sysdev_class);
+	if (ret)
+		goto free;
+
+	ret = sysdev_register(&device_amd_iommu);
+	if (ret)
+		goto free;
+
+	enable_iommus();
+
+	printk(KERN_INFO "AMD IOMMU: aperture size is %d MB\n",
+			(1 << (amd_iommu_aperture_order-20)));
+
+	printk(KERN_INFO "AMD IOMMU: device isolation ");
+	if (amd_iommu_isolate)
+		printk("enabled\n");
+	else
+		printk("disabled\n");
+
+out:
+	return ret;
+
+free:
+	if (amd_iommu_pd_alloc_bitmap)
+		free_pages((unsigned long)amd_iommu_pd_alloc_bitmap, 1);
+
+	if (amd_iommu_pd_table)
+		free_pages((unsigned long)amd_iommu_pd_table,
+				get_order(rlookup_table_size));
+
+	if (amd_iommu_rlookup_table)
+		free_pages((unsigned long)amd_iommu_rlookup_table,
+				get_order(rlookup_table_size));
+
+	if (amd_iommu_alias_table)
+		free_pages((unsigned long)amd_iommu_alias_table,
+				get_order(alias_table_size));
+
+	if (amd_iommu_dev_table)
+		free_pages((unsigned long)amd_iommu_dev_table,
+				get_order(dev_table_size));
+
+	free_iommu_all();
+
+	free_unity_maps();
+
+	goto out;
+}
+
+static int __init early_amd_iommu_detect(struct acpi_table_header *table)
+{
+	return 0;
+}
+
+void __init amd_iommu_detect(void)
+{
+	if (swiotlb || no_iommu || iommu_detected)
+		return;
+
+	if (acpi_table_parse("IVRS", early_amd_iommu_detect) == 0) {
+		iommu_detected = 1;
+		amd_iommu_detected = 1;
+#ifdef CONFIG_GART_IOMMU
+		gart_iommu_aperture_disabled = 1;
+		gart_iommu_aperture = 0;
+#endif
+	}
+}
+
+static int __init parse_amd_iommu_options(char *str)
+{
+	for (; *str; ++str) {
+		if (strcmp(str, "isolate") == 0)
+			amd_iommu_isolate = 1;
+	}
+
+	return 1;
+}
+
+static int __init parse_amd_iommu_size_options(char *str)
+{
+	for (; *str; ++str) {
+		if (strcmp(str, "32M") == 0)
+			amd_iommu_aperture_order = 25;
+		if (strcmp(str, "64M") == 0)
+			amd_iommu_aperture_order = 26;
+		if (strcmp(str, "128M") == 0)
+			amd_iommu_aperture_order = 27;
+		if (strcmp(str, "256M") == 0)
+			amd_iommu_aperture_order = 28;
+		if (strcmp(str, "512M") == 0)
+			amd_iommu_aperture_order = 29;
+		if (strcmp(str, "1G") == 0)
+			amd_iommu_aperture_order = 30;
+	}
+
+	return 1;
+}
+
+__setup("amd_iommu=", parse_amd_iommu_options);
+__setup("amd_iommu_size=", parse_amd_iommu_size_options);

arch/x86/kernel/entry_64.S

@@ -925,11 +925,11 @@ error_kernelspace:
 	   iret run with kernel gs again, so don't set the user space flag.
 	   B stepping K8s sometimes report an truncated RIP for IRET 
 	   exceptions returning to compat mode. Check for these here too. */
-	leaq irq_return(%rip),%rbp
-	cmpq %rbp,RIP(%rsp) 
+	leaq irq_return(%rip),%rcx
+	cmpq %rcx,RIP(%rsp)
 	je   error_swapgs
-	movl %ebp,%ebp	/* zero extend */
-	cmpq %rbp,RIP(%rsp) 
+	movl %ecx,%ecx	/* zero extend */
+	cmpq %rcx,RIP(%rsp)
 	je   error_swapgs
 	cmpq $gs_change,RIP(%rsp)
         je   error_swapgs

arch/x86/kernel/i387.c

@@ -162,7 +162,7 @@ int xfpregs_get(struct task_struct *target, const struct user_regset *regset,
 	int ret;
 
 	if (!cpu_has_fxsr)
-		return -EIO;
+		return -ENODEV;
 
 	ret = init_fpu(target);
 	if (ret)
@@ -179,7 +179,7 @@ int xfpregs_set(struct task_struct *target, const struct user_regset *regset,
 	int ret;
 
 	if (!cpu_has_fxsr)
-		return -EIO;
+		return -ENODEV;
 
 	ret = init_fpu(target);
 	if (ret)

arch/x86/kernel/pci-dma.c

@@ -7,6 +7,7 @@
 #include <asm/dma.h>
 #include <asm/gart.h>
 #include <asm/calgary.h>
+#include <asm/amd_iommu.h>
 
 int forbid_dac __read_mostly;
 EXPORT_SYMBOL(forbid_dac);
@@ -123,6 +124,8 @@ void __init pci_iommu_alloc(void)
 
 	detect_intel_iommu();
 
+	amd_iommu_detect();
+
 #ifdef CONFIG_SWIOTLB
 	pci_swiotlb_init();
 #endif
@@ -503,6 +506,8 @@ static int __init pci_iommu_init(void)
 
 	intel_iommu_init();
 
+	amd_iommu_init();
+
 #ifdef CONFIG_GART_IOMMU
 	gart_iommu_init();
 #endif

arch/x86/kernel/ptrace.c

@@ -943,13 +943,13 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
 		return copy_regset_to_user(child, &user_x86_32_view,
 					   REGSET_XFP,
 					   0, sizeof(struct user_fxsr_struct),
-					   datap);
+					   datap) ? -EIO : 0;
 
 	case PTRACE_SETFPXREGS:	/* Set the child extended FPU state. */
 		return copy_regset_from_user(child, &user_x86_32_view,
 					     REGSET_XFP,
 					     0, sizeof(struct user_fxsr_struct),
-					     datap);
+					     datap) ? -EIO : 0;
 #endif
 
 #if defined CONFIG_X86_32 || defined CONFIG_IA32_EMULATION

arch/x86/kernel/time_64.c

@@ -123,6 +123,8 @@ void __init time_init(void)
 		(boot_cpu_data.x86_vendor == X86_VENDOR_AMD))
 		cpu_khz = calculate_cpu_khz();
 
+	lpj_fine = ((unsigned long)tsc_khz * 1000)/HZ;
+
 	if (unsynchronized_tsc())
 		mark_tsc_unstable("TSCs unsynchronized");
 

arch/x86/kernel/tsc_32.c

@@ -1,6 +1,7 @@
 #include <linux/sched.h>
 #include <linux/clocksource.h>
 #include <linux/workqueue.h>
+#include <linux/delay.h>
 #include <linux/cpufreq.h>
 #include <linux/jiffies.h>
 #include <linux/init.h>
@@ -403,6 +404,7 @@ static inline void check_geode_tsc_reliable(void) { }
 void __init tsc_init(void)
 {
 	int cpu;
+	u64 lpj;
 
 	if (!cpu_has_tsc || tsc_disabled > 0)
 		return;
@@ -415,6 +417,10 @@ void __init tsc_init(void)
 		return;
 	}
 
+	lpj = ((u64)tsc_khz * 1000);
+	do_div(lpj, HZ);
+	lpj_fine = lpj;
+
 	/* now allow native_sched_clock() to use rdtsc */
 	tsc_disabled = 0;
 

arch/x86/lib/delay_32.c

@@ -3,6 +3,7 @@
  *
  *	Copyright (C) 1993 Linus Torvalds
  *	Copyright (C) 1997 Martin Mares <mj@atrey.karlin.mff.cuni.cz>
+ *	Copyright (C) 2008 Jiri Hladky <hladky _dot_ jiri _at_ gmail _dot_ com>
  *
  *	The __delay function must _NOT_ be inlined as its execution time
  *	depends wildly on alignment on many x86 processors. The additional
@@ -28,16 +29,22 @@
 /* simple loop based delay: */
 static void delay_loop(unsigned long loops)
 {
-	int d0;
-
 	__asm__ __volatile__(
-		"\tjmp 1f\n"
-		".align 16\n"
-		"1:\tjmp 2f\n"
-		".align 16\n"
-		"2:\tdecl %0\n\tjns 2b"
-		:"=&a" (d0)
-		:"0" (loops));
+		"	test %0,%0	\n"
+		"	jz 3f		\n"
+		"	jmp 1f		\n"
+
+		".align 16		\n"
+		"1:	jmp 2f		\n"
+
+		".align 16		\n"
+		"2:	decl %0		\n"
+		"	jnz 2b		\n"
+		"3:	decl %0		\n"
+
+		: /* we don't need output */
+		:"a" (loops)
+	);
 }
 
 /* TSC based delay: */

arch/x86/mm/fault.c

@@ -392,11 +392,7 @@ static void show_fault_oops(struct pt_regs *regs, unsigned long error_code,
 		printk(KERN_CONT "NULL pointer dereference");
 	else
 		printk(KERN_CONT "paging request");
-#ifdef CONFIG_X86_32
-	printk(KERN_CONT " at %08lx\n", address);
-#else
-	printk(KERN_CONT " at %016lx\n", address);
-#endif
+	printk(KERN_CONT " at %p\n", (void *) address);
 	printk(KERN_ALERT "IP:");
 	printk_address(regs->ip, 1);
 	dump_pagetable(address);
@@ -796,14 +792,10 @@ bad_area_nosemaphore:
 		if (show_unhandled_signals && unhandled_signal(tsk, SIGSEGV) &&
 		    printk_ratelimit()) {
 			printk(
-#ifdef CONFIG_X86_32
-			"%s%s[%d]: segfault at %lx ip %08lx sp %08lx error %lx",
-#else
-			"%s%s[%d]: segfault at %lx ip %lx sp %lx error %lx",
-#endif
+			"%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
 			task_pid_nr(tsk) > 1 ? KERN_INFO : KERN_EMERG,
-			tsk->comm, task_pid_nr(tsk), address, regs->ip,
-			regs->sp, error_code);
+			tsk->comm, task_pid_nr(tsk), address,
+			(void *) regs->ip, (void *) regs->sp, error_code);
 			print_vma_addr(" in ", regs->ip);
 			printk("\n");
 		}

arch/x86/oprofile/nmi_int.c

@@ -269,12 +269,13 @@ static void nmi_cpu_shutdown(void *dummy)
 
 static void nmi_shutdown(void)
 {
-	struct op_msrs *msrs = &__get_cpu_var(cpu_msrs);
+	struct op_msrs *msrs = &get_cpu_var(cpu_msrs);
 	nmi_enabled = 0;
 	on_each_cpu(nmi_cpu_shutdown, NULL, 0, 1);
 	unregister_die_notifier(&profile_exceptions_nb);
 	model->shutdown(msrs);
 	free_msrs();
+	put_cpu_var(cpu_msrs);
 }
 
 static void nmi_cpu_start(void *dummy)

include/asm-x86/amd_iommu.h

@@ -0,0 +1,32 @@
+/*
+ * Copyright (C) 2007-2008 Advanced Micro Devices, Inc.
+ * Author: Joerg Roedel <joerg.roedel@amd.com>
+ *         Leo Duran <leo.duran@amd.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ */
+
+#ifndef _ASM_X86_AMD_IOMMU_H
+#define _ASM_X86_AMD_IOMMU_H
+
+#ifdef CONFIG_AMD_IOMMU
+extern int amd_iommu_init(void);
+extern int amd_iommu_init_dma_ops(void);
+extern void amd_iommu_detect(void);
+#else
+static inline int amd_iommu_init(void) { return -ENODEV; }
+static inline void amd_iommu_detect(void) { }
+#endif
+
+#endif

include/asm-x86/amd_iommu_types.h

@@ -0,0 +1,244 @@
+/*
+ * Copyright (C) 2007-2008 Advanced Micro Devices, Inc.
+ * Author: Joerg Roedel <joerg.roedel@amd.com>
+ *         Leo Duran <leo.duran@amd.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ */
+
+#ifndef __AMD_IOMMU_TYPES_H__
+#define __AMD_IOMMU_TYPES_H__
+
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/spinlock.h>
+
+/*
+ * some size calculation constants
+ */
+#define DEV_TABLE_ENTRY_SIZE		256
+#define ALIAS_TABLE_ENTRY_SIZE		2
+#define RLOOKUP_TABLE_ENTRY_SIZE	(sizeof(void *))
+
+/* helper macros */
+#define LOW_U32(x) ((x) & ((1ULL << 32)-1))
+#define HIGH_U32(x) (LOW_U32((x) >> 32))
+
+/* Length of the MMIO region for the AMD IOMMU */
+#define MMIO_REGION_LENGTH       0x4000
+
+/* Capability offsets used by the driver */
+#define MMIO_CAP_HDR_OFFSET	0x00
+#define MMIO_RANGE_OFFSET	0x0c
+
+/* Masks, shifts and macros to parse the device range capability */
+#define MMIO_RANGE_LD_MASK	0xff000000
+#define MMIO_RANGE_FD_MASK	0x00ff0000
+#define MMIO_RANGE_BUS_MASK	0x0000ff00
+#define MMIO_RANGE_LD_SHIFT	24
+#define MMIO_RANGE_FD_SHIFT	16
+#define MMIO_RANGE_BUS_SHIFT	8
+#define MMIO_GET_LD(x)  (((x) & MMIO_RANGE_LD_MASK) >> MMIO_RANGE_LD_SHIFT)
+#define MMIO_GET_FD(x)  (((x) & MMIO_RANGE_FD_MASK) >> MMIO_RANGE_FD_SHIFT)
+#define MMIO_GET_BUS(x) (((x) & MMIO_RANGE_BUS_MASK) >> MMIO_RANGE_BUS_SHIFT)
+
+/* Flag masks for the AMD IOMMU exclusion range */
+#define MMIO_EXCL_ENABLE_MASK 0x01ULL
+#define MMIO_EXCL_ALLOW_MASK  0x02ULL
+
+/* Used offsets into the MMIO space */
+#define MMIO_DEV_TABLE_OFFSET   0x0000
+#define MMIO_CMD_BUF_OFFSET     0x0008
+#define MMIO_EVT_BUF_OFFSET     0x0010
+#define MMIO_CONTROL_OFFSET     0x0018
+#define MMIO_EXCL_BASE_OFFSET   0x0020
+#define MMIO_EXCL_LIMIT_OFFSET  0x0028
+#define MMIO_CMD_HEAD_OFFSET	0x2000
+#define MMIO_CMD_TAIL_OFFSET	0x2008
+#define MMIO_EVT_HEAD_OFFSET	0x2010
+#define MMIO_EVT_TAIL_OFFSET	0x2018
+#define MMIO_STATUS_OFFSET	0x2020
+
+/* feature control bits */
+#define CONTROL_IOMMU_EN        0x00ULL
+#define CONTROL_HT_TUN_EN       0x01ULL
+#define CONTROL_EVT_LOG_EN      0x02ULL
+#define CONTROL_EVT_INT_EN      0x03ULL
+#define CONTROL_COMWAIT_EN      0x04ULL
+#define CONTROL_PASSPW_EN       0x08ULL
+#define CONTROL_RESPASSPW_EN    0x09ULL
+#define CONTROL_COHERENT_EN     0x0aULL
+#define CONTROL_ISOC_EN         0x0bULL
+#define CONTROL_CMDBUF_EN       0x0cULL
+#define CONTROL_PPFLOG_EN       0x0dULL
+#define CONTROL_PPFINT_EN       0x0eULL
+
+/* command specific defines */
+#define CMD_COMPL_WAIT          0x01
+#define CMD_INV_DEV_ENTRY       0x02
+#define CMD_INV_IOMMU_PAGES     0x03
+
+#define CMD_COMPL_WAIT_STORE_MASK	0x01
+#define CMD_INV_IOMMU_PAGES_SIZE_MASK	0x01
+#define CMD_INV_IOMMU_PAGES_PDE_MASK	0x02
+
+#define CMD_INV_IOMMU_ALL_PAGES_ADDRESS	0x7fffffffffffffffULL
+
+/* macros and definitions for device table entries */
+#define DEV_ENTRY_VALID         0x00
+#define DEV_ENTRY_TRANSLATION   0x01
+#define DEV_ENTRY_IR            0x3d
+#define DEV_ENTRY_IW            0x3e
+#define DEV_ENTRY_EX            0x67
+#define DEV_ENTRY_SYSMGT1       0x68
+#define DEV_ENTRY_SYSMGT2       0x69
+#define DEV_ENTRY_INIT_PASS     0xb8
+#define DEV_ENTRY_EINT_PASS     0xb9
+#define DEV_ENTRY_NMI_PASS      0xba
+#define DEV_ENTRY_LINT0_PASS    0xbe
+#define DEV_ENTRY_LINT1_PASS    0xbf
+
+/* constants to configure the command buffer */
+#define CMD_BUFFER_SIZE    8192
+#define CMD_BUFFER_ENTRIES 512
+#define MMIO_CMD_SIZE_SHIFT 56
+#define MMIO_CMD_SIZE_512 (0x9ULL << MMIO_CMD_SIZE_SHIFT)
+
+#define PAGE_MODE_1_LEVEL 0x01
+#define PAGE_MODE_2_LEVEL 0x02
+#define PAGE_MODE_3_LEVEL 0x03
+
+#define IOMMU_PDE_NL_0   0x000ULL
+#define IOMMU_PDE_NL_1   0x200ULL
+#define IOMMU_PDE_NL_2   0x400ULL
+#define IOMMU_PDE_NL_3   0x600ULL
+
+#define IOMMU_PTE_L2_INDEX(address) (((address) >> 30) & 0x1ffULL)
+#define IOMMU_PTE_L1_INDEX(address) (((address) >> 21) & 0x1ffULL)
+#define IOMMU_PTE_L0_INDEX(address) (((address) >> 12) & 0x1ffULL)
+
+#define IOMMU_MAP_SIZE_L1 (1ULL << 21)
+#define IOMMU_MAP_SIZE_L2 (1ULL << 30)
+#define IOMMU_MAP_SIZE_L3 (1ULL << 39)
+
+#define IOMMU_PTE_P  (1ULL << 0)
+#define IOMMU_PTE_U  (1ULL << 59)
+#define IOMMU_PTE_FC (1ULL << 60)
+#define IOMMU_PTE_IR (1ULL << 61)
+#define IOMMU_PTE_IW (1ULL << 62)
+
+#define IOMMU_L1_PDE(address) \
+	((address) | IOMMU_PDE_NL_1 | IOMMU_PTE_P | IOMMU_PTE_IR | IOMMU_PTE_IW)
+#define IOMMU_L2_PDE(address) \
+	((address) | IOMMU_PDE_NL_2 | IOMMU_PTE_P | IOMMU_PTE_IR | IOMMU_PTE_IW)
+
+#define IOMMU_PAGE_MASK (((1ULL << 52) - 1) & ~0xfffULL)
+#define IOMMU_PTE_PRESENT(pte) ((pte) & IOMMU_PTE_P)
+#define IOMMU_PTE_PAGE(pte) (phys_to_virt((pte) & IOMMU_PAGE_MASK))
+#define IOMMU_PTE_MODE(pte) (((pte) >> 9) & 0x07)
+
+#define IOMMU_PROT_MASK 0x03
+#define IOMMU_PROT_IR 0x01
+#define IOMMU_PROT_IW 0x02
+
+/* IOMMU capabilities */
+#define IOMMU_CAP_IOTLB   24
+#define IOMMU_CAP_NPCACHE 26
+
+#define MAX_DOMAIN_ID 65536
+
+struct protection_domain {
+	spinlock_t lock;
+	u16 id;
+	int mode;
+	u64 *pt_root;
+	void *priv;
+};
+
+struct dma_ops_domain {
+	struct list_head list;
+	struct protection_domain domain;
+	unsigned long aperture_size;
+	unsigned long next_bit;
+	unsigned long *bitmap;
+	u64 **pte_pages;
+};
+
+struct amd_iommu {
+	struct list_head list;
+	spinlock_t lock;
+
+	u16 devid;
+	u16 cap_ptr;
+
+	u64 mmio_phys;
+	u8 *mmio_base;
+	u32 cap;
+	u16 first_device;
+	u16 last_device;
+	u64 exclusion_start;
+	u64 exclusion_length;
+
+	u8 *cmd_buf;
+	u32 cmd_buf_size;
+
+	int need_sync;
+
+	struct dma_ops_domain *default_dom;
+};
+
+extern struct list_head amd_iommu_list;
+
+struct dev_table_entry {
+	u32 data[8];
+};
+
+struct unity_map_entry {
+	struct list_head list;
+	u16 devid_start;
+	u16 devid_end;
+	u64 address_start;
+	u64 address_end;
+	int prot;
+};
+
+extern struct list_head amd_iommu_unity_map;
+
+/* data structures for device handling */
+extern struct dev_table_entry *amd_iommu_dev_table;
+extern u16 *amd_iommu_alias_table;
+extern struct amd_iommu **amd_iommu_rlookup_table;
+
+extern unsigned amd_iommu_aperture_order;
+
+extern u16 amd_iommu_last_bdf;
+
+/* data structures for protection domain handling */
+extern struct protection_domain **amd_iommu_pd_table;
+extern unsigned long *amd_iommu_pd_alloc_bitmap;
+
+extern int amd_iommu_isolate;
+
+static inline void print_devid(u16 devid, int nl)
+{
+	int bus = devid >> 8;
+	int dev = devid >> 3 & 0x1f;
+	int fn  = devid & 0x07;
+
+	printk("%02x:%02x.%x", bus, dev, fn);
+	if (nl)
+		printk("\n");
+}
+
+#endif

include/asm-x86/cpufeature.h

@@ -106,6 +106,7 @@
 /* More extended AMD flags: CPUID level 0x80000001, ecx, word 6 */
 #define X86_FEATURE_LAHF_LM	(6*32+ 0) /* LAHF/SAHF in long mode */
 #define X86_FEATURE_CMP_LEGACY	(6*32+ 1) /* If yes HyperThreading not valid */
+#define X86_FEATURE_IBS		(6*32+ 10) /* Instruction Based Sampling */
 
 /*
  * Auxiliary flags: Linux defined - For features scattered in various

include/asm-x86/gart.h

@@ -22,8 +22,9 @@ extern int gart_iommu_aperture_allowed;
 extern int gart_iommu_aperture_disabled;
 extern int fix_aperture;
 #else
-#define gart_iommu_aperture 0
-#define gart_iommu_aperture_allowed 0
+#define gart_iommu_aperture		0
+#define gart_iommu_aperture_allowed	0
+#define gart_iommu_aperture_disabled	1
 
 static inline void early_gart_iommu_check(void)
 {

include/asm-x86/io.h

@@ -3,6 +3,62 @@
 
 #define ARCH_HAS_IOREMAP_WC
 
+#include <linux/compiler.h>
+
+#define build_mmio_read(name, size, type, reg, barrier) \
+static inline type name(const volatile void __iomem *addr) \
+{ type ret; asm volatile("mov" size " %1,%0":"=" reg (ret) \
+:"m" (*(volatile type __force *)addr) barrier); return ret; }
+
+#define build_mmio_write(name, size, type, reg, barrier) \
+static inline void name(type val, volatile void __iomem *addr) \
+{ asm volatile("mov" size " %0,%1": :reg (val), \
+"m" (*(volatile type __force *)addr) barrier); }
+
+build_mmio_read(readb, "b", unsigned char, "q", :"memory")
+build_mmio_read(readw, "w", unsigned short, "r", :"memory")
+build_mmio_read(readl, "l", unsigned int, "r", :"memory")
+
+build_mmio_read(__readb, "b", unsigned char, "q", )
+build_mmio_read(__readw, "w", unsigned short, "r", )
+build_mmio_read(__readl, "l", unsigned int, "r", )
+
+build_mmio_write(writeb, "b", unsigned char, "q", :"memory")
+build_mmio_write(writew, "w", unsigned short, "r", :"memory")
+build_mmio_write(writel, "l", unsigned int, "r", :"memory")
+
+build_mmio_write(__writeb, "b", unsigned char, "q", )
+build_mmio_write(__writew, "w", unsigned short, "r", )
+build_mmio_write(__writel, "l", unsigned int, "r", )
+
+#define readb_relaxed(a) __readb(a)
+#define readw_relaxed(a) __readw(a)
+#define readl_relaxed(a) __readl(a)
+#define __raw_readb __readb
+#define __raw_readw __readw
+#define __raw_readl __readl
+
+#define __raw_writeb __writeb
+#define __raw_writew __writew
+#define __raw_writel __writel
+
+#define mmiowb() barrier()
+
+#ifdef CONFIG_X86_64
+build_mmio_read(readq, "q", unsigned long, "r", :"memory")
+build_mmio_read(__readq, "q", unsigned long, "r", )
+build_mmio_write(writeq, "q", unsigned long, "r", :"memory")
+build_mmio_write(__writeq, "q", unsigned long, "r", )
+
+#define readq_relaxed(a) __readq(a)
+#define __raw_readq __readq
+#define __raw_writeq writeq
+
+/* Let people know we have them */
+#define readq readq
+#define writeq writeq
+#endif
+
 #ifdef CONFIG_X86_32
 # include "io_32.h"
 #else

include/asm-x86/io_32.h

@@ -149,55 +149,6 @@ extern void __iomem *fix_ioremap(unsigned idx, unsigned long phys);
 #define virt_to_bus virt_to_phys
 #define bus_to_virt phys_to_virt
 
-/*
- * readX/writeX() are used to access memory mapped devices. On some
- * architectures the memory mapped IO stuff needs to be accessed
- * differently. On the x86 architecture, we just read/write the
- * memory location directly.
- */
-
-static inline unsigned char readb(const volatile void __iomem *addr)
-{
-	return *(volatile unsigned char __force *)addr;
-}
-
-static inline unsigned short readw(const volatile void __iomem *addr)
-{
-	return *(volatile unsigned short __force *)addr;
-}
-
-static inline unsigned int readl(const volatile void __iomem *addr)
-{
-	return *(volatile unsigned int __force *) addr;
-}
-
-#define readb_relaxed(addr) readb(addr)
-#define readw_relaxed(addr) readw(addr)
-#define readl_relaxed(addr) readl(addr)
-#define __raw_readb readb
-#define __raw_readw readw
-#define __raw_readl readl
-
-static inline void writeb(unsigned char b, volatile void __iomem *addr)
-{
-	*(volatile unsigned char __force *)addr = b;
-}
-
-static inline void writew(unsigned short b, volatile void __iomem *addr)
-{
-	*(volatile unsigned short __force *)addr = b;
-}
-
-static inline void writel(unsigned int b, volatile void __iomem *addr)
-{
-	*(volatile unsigned int __force *)addr = b;
-}
-#define __raw_writeb writeb
-#define __raw_writew writew
-#define __raw_writel writel
-
-#define mmiowb()
-
 static inline void
 memset_io(volatile void __iomem *addr, unsigned char val, int count)
 {

include/asm-x86/io_64.h

@@ -204,77 +204,6 @@ extern void __iomem *fix_ioremap(unsigned idx, unsigned long phys);
 #define virt_to_bus virt_to_phys
 #define bus_to_virt phys_to_virt
 
-/*
- * readX/writeX() are used to access memory mapped devices. On some
- * architectures the memory mapped IO stuff needs to be accessed
- * differently. On the x86 architecture, we just read/write the
- * memory location directly.
- */
-
-static inline __u8 __readb(const volatile void __iomem *addr)
-{
-	return *(__force volatile __u8 *)addr;
-}
-
-static inline __u16 __readw(const volatile void __iomem *addr)
-{
-	return *(__force volatile __u16 *)addr;
-}
-
-static __always_inline __u32 __readl(const volatile void __iomem *addr)
-{
-	return *(__force volatile __u32 *)addr;
-}
-
-static inline __u64 __readq(const volatile void __iomem *addr)
-{
-	return *(__force volatile __u64 *)addr;
-}
-
-#define readb(x) __readb(x)
-#define readw(x) __readw(x)
-#define readl(x) __readl(x)
-#define readq(x) __readq(x)
-#define readb_relaxed(a) readb(a)
-#define readw_relaxed(a) readw(a)
-#define readl_relaxed(a) readl(a)
-#define readq_relaxed(a) readq(a)
-#define __raw_readb readb
-#define __raw_readw readw
-#define __raw_readl readl
-#define __raw_readq readq
-
-#define mmiowb()
-
-static inline void __writel(__u32 b, volatile void __iomem *addr)
-{
-	*(__force volatile __u32 *)addr = b;
-}
-
-static inline void __writeq(__u64 b, volatile void __iomem *addr)
-{
-	*(__force volatile __u64 *)addr = b;
-}
-
-static inline void __writeb(__u8 b, volatile void __iomem *addr)
-{
-	*(__force volatile __u8 *)addr = b;
-}
-
-static inline void __writew(__u16 b, volatile void __iomem *addr)
-{
-	*(__force volatile __u16 *)addr = b;
-}
-
-#define writeq(val, addr) __writeq((val), (addr))
-#define writel(val, addr) __writel((val), (addr))
-#define writew(val, addr) __writew((val), (addr))
-#define writeb(val, addr) __writeb((val), (addr))
-#define __raw_writeb writeb
-#define __raw_writew writew
-#define __raw_writel writel
-#define __raw_writeq writeq
-
 void __memcpy_fromio(void *, unsigned long, unsigned);
 void __memcpy_toio(unsigned long, const void *, unsigned);
 

include/linux/delay.h

@@ -41,6 +41,7 @@ static inline void ndelay(unsigned long x)
 #define ndelay(x) ndelay(x)
 #endif
 
+extern unsigned long lpj_fine;
 void calibrate_delay(void);
 void msleep(unsigned int msecs);
 unsigned long msleep_interruptible(unsigned int msecs);

init/calibrate.c

@@ -8,7 +8,9 @@
 #include <linux/delay.h>
 #include <linux/init.h>
 #include <linux/timex.h>
+#include <linux/smp.h>
 
+unsigned long lpj_fine;
 unsigned long preset_lpj;
 static int __init lpj_setup(char *str)
 {
@@ -33,9 +35,9 @@ static unsigned long __cpuinit calibrate_delay_direct(void)
 	unsigned long pre_start, start, post_start;
 	unsigned long pre_end, end, post_end;
 	unsigned long start_jiffies;
-	unsigned long tsc_rate_min, tsc_rate_max;
-	unsigned long good_tsc_sum = 0;
-	unsigned long good_tsc_count = 0;
+	unsigned long timer_rate_min, timer_rate_max;
+	unsigned long good_timer_sum = 0;
+	unsigned long good_timer_count = 0;
 	int i;
 
 	if (read_current_timer(&pre_start) < 0 )
@@ -79,22 +81,24 @@ static unsigned long __cpuinit calibrate_delay_direct(void)
 		}
 		read_current_timer(&post_end);
 
-		tsc_rate_max = (post_end - pre_start) / DELAY_CALIBRATION_TICKS;
-		tsc_rate_min = (pre_end - post_start) / DELAY_CALIBRATION_TICKS;
+		timer_rate_max = (post_end - pre_start) /
+					DELAY_CALIBRATION_TICKS;
+		timer_rate_min = (pre_end - post_start) /
+					DELAY_CALIBRATION_TICKS;
 
 		/*
-	 	 * If the upper limit and lower limit of the tsc_rate is
+		 * If the upper limit and lower limit of the timer_rate is
 		 * >= 12.5% apart, redo calibration.
 		 */
 		if (pre_start != 0 && pre_end != 0 &&
-		    (tsc_rate_max - tsc_rate_min) < (tsc_rate_max >> 3)) {
-			good_tsc_count++;
-			good_tsc_sum += tsc_rate_max;
+		    (timer_rate_max - timer_rate_min) < (timer_rate_max >> 3)) {
+			good_timer_count++;
+			good_timer_sum += timer_rate_max;
 		}
 	}
 
-	if (good_tsc_count)
-		return (good_tsc_sum/good_tsc_count);
+	if (good_timer_count)
+		return (good_timer_sum/good_timer_count);
 
 	printk(KERN_WARNING "calibrate_delay_direct() failed to get a good "
 	       "estimate for loops_per_jiffy.\nProbably due to long platform interrupts. Consider using \"lpj=\" boot option.\n");
@@ -108,6 +112,10 @@ static unsigned long __cpuinit calibrate_delay_direct(void) {return 0;}
  * This is the number of bits of precision for the loops_per_jiffy.  Each
  * bit takes on average 1.5/HZ seconds.  This (like the original) is a little
  * better than 1%
+ * For the boot cpu we can skip the delay calibration and assign it a value
+ * calculated based on the timer frequency.
+ * For the rest of the CPUs we cannot assume that the timer frequency is same as
+ * the cpu frequency, hence do the calibration for those.
  */
 #define LPS_PREC 8
 
@@ -118,20 +126,20 @@ void __cpuinit calibrate_delay(void)
 
 	if (preset_lpj) {
 		loops_per_jiffy = preset_lpj;
-		printk("Calibrating delay loop (skipped)... "
-			"%lu.%02lu BogoMIPS preset\n",
-			loops_per_jiffy/(500000/HZ),
-			(loops_per_jiffy/(5000/HZ)) % 100);
+		printk(KERN_INFO
+			"Calibrating delay loop (skipped) preset value.. ");
+	} else if ((smp_processor_id() == 0) && lpj_fine) {
+		loops_per_jiffy = lpj_fine;
+		printk(KERN_INFO
+			"Calibrating delay loop (skipped), "
+			"value calculated using timer frequency.. ");
 	} else if ((loops_per_jiffy = calibrate_delay_direct()) != 0) {
-		printk("Calibrating delay using timer specific routine.. ");
-		printk("%lu.%02lu BogoMIPS (lpj=%lu)\n",
-			loops_per_jiffy/(500000/HZ),
-			(loops_per_jiffy/(5000/HZ)) % 100,
-			loops_per_jiffy);
+		printk(KERN_INFO
+			"Calibrating delay using timer specific routine.. ");
 	} else {
 		loops_per_jiffy = (1<<12);
 
-		printk(KERN_DEBUG "Calibrating delay loop... ");
+		printk(KERN_INFO "Calibrating delay loop... ");
 		while ((loops_per_jiffy <<= 1) != 0) {
 			/* wait for "start of" clock tick */
 			ticks = jiffies;
@@ -161,12 +169,8 @@ void __cpuinit calibrate_delay(void)
 			if (jiffies != ticks)	/* longer than 1 tick */
 				loops_per_jiffy &= ~loopbit;
 		}
-
-		/* Round the value and print it */
-		printk("%lu.%02lu BogoMIPS (lpj=%lu)\n",
-			loops_per_jiffy/(500000/HZ),
-			(loops_per_jiffy/(5000/HZ)) % 100,
-			loops_per_jiffy);
 	}
-
+	printk(KERN_INFO "%lu.%02lu BogoMIPS (lpj=%lu)\n",
+			loops_per_jiffy/(500000/HZ),
+			(loops_per_jiffy/(5000/HZ)) % 100, loops_per_jiffy);
 }