intel-iommu.c

linux 内核源代码

/*
 * Copyright (c) 2006, Intel Corporation.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope 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.
 *
 * Copyright (C) Ashok Raj <[e-mail]>
 * Copyright (C) Shaohua Li <[e-mail]>
 * Copyright (C) Anil S Keshavamurthy <[e-mail]>
 */

#include <linux/init.h>
#include <linux/bitmap.h>
#include <linux/slab.h>
#include <linux/irq.h>
#include <linux/interrupt.h>
#include <linux/sysdev.h>
#include <linux/spinlock.h>
#include <linux/pci.h>
#include <linux/dmar.h>
#include <linux/dma-mapping.h>
#include <linux/mempool.h>
#include "iova.h"
#include "intel-iommu.h"
#include <asm/proto.h> /* force_iommu in this header in x86-64*/
#include <asm/cacheflush.h>
#include <asm/gart.h>
#include "pci.h"

#define IS_GFX_DEVICE(pdev) ((pdev->class >> 16) == PCI_BASE_CLASS_DISPLAY)
#define IS_ISA_DEVICE(pdev) ((pdev->class >> 8) == PCI_CLASS_BRIDGE_ISA)

#define IOAPIC_RANGE_START	(0xfee00000)
#define IOAPIC_RANGE_END	(0xfeefffff)
#define IOVA_START_ADDR		(0x1000)

#define DEFAULT_DOMAIN_ADDRESS_WIDTH 48

#define DMAR_OPERATION_TIMEOUT (HZ*60) /* 1m */

#define DOMAIN_MAX_ADDR(gaw) ((((u64)1) << gaw) - 1)

static void domain_remove_dev_info(struct dmar_domain *domain);

static int dmar_disabled;
static int __initdata dmar_map_gfx = 1;
static int dmar_forcedac;

#define DUMMY_DEVICE_DOMAIN_INFO ((struct device_domain_info *)(-1))
static DEFINE_SPINLOCK(device_domain_lock);
static LIST_HEAD(device_domain_list);

static int __init intel_iommu_setup(char *str)
{
	if (!str)
		return -EINVAL;
	while (*str) {
		if (!strncmp(str, "off", 3)) {
			dmar_disabled = 1;
			printk(KERN_INFO"Intel-IOMMU: disabled\n");
		} else if (!strncmp(str, "igfx_off", 8)) {
			dmar_map_gfx = 0;
			printk(KERN_INFO
				"Intel-IOMMU: disable GFX device mapping\n");
		} else if (!strncmp(str, "forcedac", 8)) {
			printk (KERN_INFO
				"Intel-IOMMU: Forcing DAC for PCI devices\n");
			dmar_forcedac = 1;
		}

		str += strcspn(str, ",");
		while (*str == ',')
			str++;
	}
	return 0;
}
__setup("intel_iommu=", intel_iommu_setup);

static struct kmem_cache *iommu_domain_cache;
static struct kmem_cache *iommu_devinfo_cache;
static struct kmem_cache *iommu_iova_cache;

static inline void *iommu_kmem_cache_alloc(struct kmem_cache *cachep)
{
	unsigned int flags;
	void *vaddr;

	/* trying to avoid low memory issues */
	flags = current->flags & PF_MEMALLOC;
	current->flags |= PF_MEMALLOC;
	vaddr = kmem_cache_alloc(cachep, GFP_ATOMIC);
	current->flags &= (~PF_MEMALLOC | flags);
	return vaddr;
}


static inline void *alloc_pgtable_page(void)
{
	unsigned int flags;
	void *vaddr;

	/* trying to avoid low memory issues */
	flags = current->flags & PF_MEMALLOC;
	current->flags |= PF_MEMALLOC;
	vaddr = (void *)get_zeroed_page(GFP_ATOMIC);
	current->flags &= (~PF_MEMALLOC | flags);
	return vaddr;
}

static inline void free_pgtable_page(void *vaddr)
{
	free_page((unsigned long)vaddr);
}

static inline void *alloc_domain_mem(void)
{
	return iommu_kmem_cache_alloc(iommu_domain_cache);
}

static inline void free_domain_mem(void *vaddr)
{
	kmem_cache_free(iommu_domain_cache, vaddr);
}

static inline void * alloc_devinfo_mem(void)
{
	return iommu_kmem_cache_alloc(iommu_devinfo_cache);
}

static inline void free_devinfo_mem(void *vaddr)
{
	kmem_cache_free(iommu_devinfo_cache, vaddr);
}

struct iova *alloc_iova_mem(void)
{
	return iommu_kmem_cache_alloc(iommu_iova_cache);
}

void free_iova_mem(struct iova *iova)
{
	kmem_cache_free(iommu_iova_cache, iova);
}

static inline void __iommu_flush_cache(
	struct intel_iommu *iommu, void *addr, int size)
{
	if (!ecap_coherent(iommu->ecap))
		clflush_cache_range(addr, size);
}

/* Gets context entry for a given bus and devfn */
static struct context_entry * device_to_context_entry(struct intel_iommu *iommu,
		u8 bus, u8 devfn)
{
	struct root_entry *root;
	struct context_entry *context;
	unsigned long phy_addr;
	unsigned long flags;

	spin_lock_irqsave(&iommu->lock, flags);
	root = &iommu->root_entry[bus];
	context = get_context_addr_from_root(root);
	if (!context) {
		context = (struct context_entry *)alloc_pgtable_page();
		if (!context) {
			spin_unlock_irqrestore(&iommu->lock, flags);
			return NULL;
		}
		__iommu_flush_cache(iommu, (void *)context, PAGE_SIZE_4K);
		phy_addr = virt_to_phys((void *)context);
		set_root_value(root, phy_addr);
		set_root_present(root);
		__iommu_flush_cache(iommu, root, sizeof(*root));
	}
	spin_unlock_irqrestore(&iommu->lock, flags);
	return &context[devfn];
}

static int device_context_mapped(struct intel_iommu *iommu, u8 bus, u8 devfn)
{
	struct root_entry *root;
	struct context_entry *context;
	int ret;
	unsigned long flags;

	spin_lock_irqsave(&iommu->lock, flags);
	root = &iommu->root_entry[bus];
	context = get_context_addr_from_root(root);
	if (!context) {
		ret = 0;
		goto out;
	}
	ret = context_present(context[devfn]);
out:
	spin_unlock_irqrestore(&iommu->lock, flags);
	return ret;
}

static void clear_context_table(struct intel_iommu *iommu, u8 bus, u8 devfn)
{
	struct root_entry *root;
	struct context_entry *context;
	unsigned long flags;

	spin_lock_irqsave(&iommu->lock, flags);
	root = &iommu->root_entry[bus];
	context = get_context_addr_from_root(root);
	if (context) {
		context_clear_entry(context[devfn]);
		__iommu_flush_cache(iommu, &context[devfn], \
			sizeof(*context));
	}
	spin_unlock_irqrestore(&iommu->lock, flags);
}

static void free_context_table(struct intel_iommu *iommu)
{
	struct root_entry *root;
	int i;
	unsigned long flags;
	struct context_entry *context;

	spin_lock_irqsave(&iommu->lock, flags);
	if (!iommu->root_entry) {
		goto out;
	}
	for (i = 0; i < ROOT_ENTRY_NR; i++) {
		root = &iommu->root_entry[i];
		context = get_context_addr_from_root(root);
		if (context)
			free_pgtable_page(context);
	}
	free_pgtable_page(iommu->root_entry);
	iommu->root_entry = NULL;
out:
	spin_unlock_irqrestore(&iommu->lock, flags);
}

/* page table handling */
#define LEVEL_STRIDE		(9)
#define LEVEL_MASK		(((u64)1 << LEVEL_STRIDE) - 1)

static inline int agaw_to_level(int agaw)
{
	return agaw + 2;
}

static inline int agaw_to_width(int agaw)
{
	return 30 + agaw * LEVEL_STRIDE;

}

static inline int width_to_agaw(int width)
{
	return (width - 30) / LEVEL_STRIDE;
}

static inline unsigned int level_to_offset_bits(int level)
{
	return (12 + (level - 1) * LEVEL_STRIDE);
}

static inline int address_level_offset(u64 addr, int level)
{
	return ((addr >> level_to_offset_bits(level)) & LEVEL_MASK);
}

static inline u64 level_mask(int level)
{
	return ((u64)-1 << level_to_offset_bits(level));
}

static inline u64 level_size(int level)
{
	return ((u64)1 << level_to_offset_bits(level));
}

static inline u64 align_to_level(u64 addr, int level)
{
	return ((addr + level_size(level) - 1) & level_mask(level));
}

static struct dma_pte * addr_to_dma_pte(struct dmar_domain *domain, u64 addr)
{
	int addr_width = agaw_to_width(domain->agaw);
	struct dma_pte *parent, *pte = NULL;
	int level = agaw_to_level(domain->agaw);
	int offset;
	unsigned long flags;

	BUG_ON(!domain->pgd);

	addr &= (((u64)1) << addr_width) - 1;
	parent = domain->pgd;

	spin_lock_irqsave(&domain->mapping_lock, flags);
	while (level > 0) {
		void *tmp_page;

		offset = address_level_offset(addr, level);
		pte = &parent[offset];
		if (level == 1)
			break;

		if (!dma_pte_present(*pte)) {
			tmp_page = alloc_pgtable_page();

			if (!tmp_page) {
				spin_unlock_irqrestore(&domain->mapping_lock,
					flags);
				return NULL;
			}
			__iommu_flush_cache(domain->iommu, tmp_page,
					PAGE_SIZE_4K);
			dma_set_pte_addr(*pte, virt_to_phys(tmp_page));
			/*
			 * high level table always sets r/w, last level page
			 * table control read/write
			 */
			dma_set_pte_readable(*pte);
			dma_set_pte_writable(*pte);
			__iommu_flush_cache(domain->iommu, pte, sizeof(*pte));
		}
		parent = phys_to_virt(dma_pte_addr(*pte));
		level--;
	}

	spin_unlock_irqrestore(&domain->mapping_lock, flags);
	return pte;
}

/* return address's pte at specific level */
static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr,
		int level)
{
	struct dma_pte *parent, *pte = NULL;
	int total = agaw_to_level(domain->agaw);
	int offset;

	parent = domain->pgd;
	while (level <= total) {
		offset = address_level_offset(addr, total);
		pte = &parent[offset];
		if (level == total)
			return pte;

		if (!dma_pte_present(*pte))
			break;
		parent = phys_to_virt(dma_pte_addr(*pte));
		total--;
	}
	return NULL;
}

/* clear one page's page table */
static void dma_pte_clear_one(struct dmar_domain *domain, u64 addr)
{
	struct dma_pte *pte = NULL;

	/* get last level pte */
	pte = dma_addr_level_pte(domain, addr, 1);

	if (pte) {
		dma_clear_pte(*pte);
		__iommu_flush_cache(domain->iommu, pte, sizeof(*pte));
	}
}

/* clear last level pte, a tlb flush should be followed */
static void dma_pte_clear_range(struct dmar_domain *domain, u64 start, u64 end)
{
	int addr_width = agaw_to_width(domain->agaw);

	start &= (((u64)1) << addr_width) - 1;
	end &= (((u64)1) << addr_width) - 1;
	/* in case it's partial page */
	start = PAGE_ALIGN_4K(start);
	end &= PAGE_MASK_4K;

	/* we don't need lock here, nobody else touches the iova range */
	while (start < end) {
		dma_pte_clear_one(domain, start);
		start += PAGE_SIZE_4K;
	}
}

/* free page table pages. last level pte should already be cleared */
static void dma_pte_free_pagetable(struct dmar_domain *domain,
	u64 start, u64 end)
{
	int addr_width = agaw_to_width(domain->agaw);
	struct dma_pte *pte;
	int total = agaw_to_level(domain->agaw);
	int level;
	u64 tmp;

	start &= (((u64)1) << addr_width) - 1;
	end &= (((u64)1) << addr_width) - 1;

	/* we don't need lock here, nobody else touches the iova range */
	level = 2;
	while (level <= total) {
		tmp = align_to_level(start, level);
		if (tmp >= end || (tmp + level_size(level) > end))
			return;

		while (tmp < end) {
			pte = dma_addr_level_pte(domain, tmp, level);
			if (pte) {
				free_pgtable_page(
					phys_to_virt(dma_pte_addr(*pte)));
				dma_clear_pte(*pte);
				__iommu_flush_cache(domain->iommu,
						pte, sizeof(*pte));
			}
			tmp += level_size(level);
		}
		level++;
	}
	/* free pgd */
	if (start == 0 && end >= ((((u64)1) << addr_width) - 1)) {
		free_pgtable_page(domain->pgd);
		domain->pgd = NULL;
	}
}

/* iommu handling */
static int iommu_alloc_root_entry(struct intel_iommu *iommu)
{
	struct root_entry *root;
	unsigned long flags;

	root = (struct root_entry *)alloc_pgtable_page();
	if (!root)
		return -ENOMEM;

	__iommu_flush_cache(iommu, root, PAGE_SIZE_4K);

	spin_lock_irqsave(&iommu->lock, flags);
	iommu->root_entry = root;
	spin_unlock_irqrestore(&iommu->lock, flags);

	return 0;
}

#define IOMMU_WAIT_OP(iommu, offset, op, cond, sts) \
{\
	unsigned long start_time = jiffies;\
	while (1) {\
		sts = op (iommu->reg + offset);\
		if (cond)\
			break;\
		if (time_after(jiffies, start_time + DMAR_OPERATION_TIMEOUT))\
			panic("DMAR hardware is malfunctioning\n");\
		cpu_relax();\
	}\
}

static void iommu_set_root_entry(struct intel_iommu *iommu)
{
	void *addr;
	u32 cmd, sts;
	unsigned long flag;

	addr = iommu->root_entry;

	spin_lock_irqsave(&iommu->register_lock, flag);
	dmar_writeq(iommu->reg + DMAR_RTADDR_REG, virt_to_phys(addr));

	cmd = iommu->gcmd | DMA_GCMD_SRTP;
	writel(cmd, iommu->reg + DMAR_GCMD_REG);

	/* Make sure hardware complete it */
	IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
		readl, (sts & DMA_GSTS_RTPS), sts);

	spin_unlock_irqrestore(&iommu->register_lock, flag);
}

static void iommu_flush_write_buffer(struct intel_iommu *iommu)
{
	u32 val;
	unsigned long flag;

	if (!cap_rwbf(iommu->cap))
		return;
	val = iommu->gcmd | DMA_GCMD_WBF;

	spin_lock_irqsave(&iommu->register_lock, flag);
	writel(val, iommu->reg + DMAR_GCMD_REG);

	/* Make sure hardware complete it */
	IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
			readl, (!(val & DMA_GSTS_WBFS)), val);

	spin_unlock_irqrestore(&iommu->register_lock, flag);
}

/* return value determine if we need a write buffer flush */
static int __iommu_flush_context(struct intel_iommu *iommu,
	u16 did, u16 source_id, u8 function_mask, u64 type,
	int non_present_entry_flush)
{
	u64 val = 0;
	unsigned long flag;

	/*
	 * In the non-present entry flush case, if hardware doesn't cache
	 * non-present entry we do nothing and if hardware cache non-present
	 * entry, we flush entries of domain 0 (the domain id is used to cache
	 * any non-present entries)
	 */
	if (non_present_entry_flush) {
		if (!cap_caching_mode(iommu->cap))
			return 1;
		else
			did = 0;
	}

	switch (type) {
	case DMA_CCMD_GLOBAL_INVL:
		val = DMA_CCMD_GLOBAL_INVL;
		break;
	case DMA_CCMD_DOMAIN_INVL:
		val = DMA_CCMD_DOMAIN_INVL|DMA_CCMD_DID(did);
		break;
	case DMA_CCMD_DEVICE_INVL:
		val = DMA_CCMD_DEVICE_INVL|DMA_CCMD_DID(did)
			| DMA_CCMD_SID(source_id) | DMA_CCMD_FM(function_mask);
		break;
	default:
		BUG();
	}
	val |= DMA_CCMD_ICC;

	spin_lock_irqsave(&iommu->register_lock, flag);
	dmar_writeq(iommu->reg + DMAR_CCMD_REG, val);

	/* Make sure hardware complete it */
	IOMMU_WAIT_OP(iommu, DMAR_CCMD_REG,
		dmar_readq, (!(val & DMA_CCMD_ICC)), val);

	spin_unlock_irqrestore(&iommu->register_lock, flag);

	/* flush context entry will implictly flush write buffer */
	return 0;
}

static int inline iommu_flush_context_global(struct intel_iommu *iommu,
	int non_present_entry_flush)
{
	return __iommu_flush_context(iommu, 0, 0, 0, DMA_CCMD_GLOBAL_INVL,
		non_present_entry_flush);
}

static int inline iommu_flush_context_domain(struct intel_iommu *iommu, u16 did,