From: Don Dutile <ddutile@redhat.com>
Date: Tue, 1 Dec 2009 23:42:46 -0500
Subject: [pci] intel-iommu: add 2.6.32-rc4 sw and hw pass-through
Message-id: <4B15A9F6.8090403@redhat.com>
Patchwork-id: 21630
O-Subject: [RHEL5.5 PATCH V2] 2/9: INTEL_IOMMU: Backport 2.6.32-rc4 sw and
hw pass-through
Bugzilla: 516811 518103
RH-Acked-by: Chris Wright <chrisw@redhat.com>
RH-Acked-by: Prarit Bhargava <prarit@redhat.com>
V2 cleaned up space/tabs/indents and removed duplicate pci_is_root_device
that a post rhel5-172 patch added (pci-aer patch?). otherwise, the same.
BZ 516727.
0/9 states build & tests.
This was a set of over 25 commits, so
it seemed ridiculous to list them all.
It's primarily this commit with numerous bug fixes & patches:
commit 19943b0e30b05d42e494ae6fef78156ebc8c637e
Author: David Woodhouse <David.Woodhouse@intel.com>
Date: Tue Aug 4 16:19:20 2009 +0100
intel-iommu: Unify hardware and software passthrough support
This makes the hardware passthrough mode work a lot more like the
software version, so that the behaviour of a kernel with 'iommu=pt'
is the same whether the hardware supports passthrough or not.
In particular:
- We use a single si_domain for the pass-through devices.
- 32-bit devices can be taken out of the pass-through domain so that
they don't have to use swiotlb.
- Devices will work again after being removed from a KVM guest.
- A potential oops on OOM (in init_context_pass_through()) is fixed.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Please review & ACK.
- Don
>From f56dfb24677d857c5e60f692bed824b8e6311129 Mon Sep 17 00:00:00 2001
From: Donald Dutile <ddutile@redhat.com>
Date: Mon, 16 Nov 2009 17:21:32 -0500
Subject: [PATCH 2/9] INTEL_IOMMU: Backport 2.6.32-rc4 sw and hw pass-through
diff --git a/arch/x86_64/kernel/pci-dma.c b/arch/x86_64/kernel/pci-dma.c
index 96d150e..50b27c3 100644
--- a/arch/x86_64/kernel/pci-dma.c
+++ b/arch/x86_64/kernel/pci-dma.c
@@ -40,6 +40,15 @@ int force_iommu __read_mostly= 0;
/* Set this to 1 if there is a HW IOMMU in the system */
int iommu_detected __read_mostly = 0;
+/*
+ * This variable becomes 1 if iommu=pt is passed on the kernel command line.
+ * If this variable is 1, IOMMU implementations do no DMA ranslation for
+ * devices and allow every device to access to whole physical memory. This is
+ * useful if a user want to use an IOMMU only for KVM device assignment to
+ * guests and not for driver dma translation.
+ */
+int iommu_pass_through __read_mostly;
+
/* Dummy device used for NULL arguments (normally ISA). Better would
be probably a smaller DMA mask, but this is bug-to-bug compatible
to i386. */
@@ -160,8 +169,6 @@ dma_alloc_coherent(struct device *dev, size_t size, dma_addr_t *dma_handle,
}
EXPORT_SYMBOL(dma_alloc_coherent);
-extern int iommu_pass_through;
-
/*
* Unmap coherent memory.
* The caller must ensure that the device has finished accessing the mapping.
@@ -273,10 +280,8 @@ __init int iommu_setup(char *p)
#ifdef CONFIG_SWIOTLB
if (!strncmp(p, "soft",4))
swiotlb = 1;
- if (!strncmp(p, "pt", 2)) {
+ if (!strncmp(p, "pt", 2))
iommu_pass_through = 1;
- return 1;
- }
#endif
#ifdef CONFIG_IOMMU
diff --git a/arch/x86_64/kernel/pci-swiotlb.c b/arch/x86_64/kernel/pci-swiotlb.c
index 92fdbf0..4e6167a 100644
--- a/arch/x86_64/kernel/pci-swiotlb.c
+++ b/arch/x86_64/kernel/pci-swiotlb.c
@@ -32,8 +32,7 @@ struct dma_mapping_ops swiotlb_dma_ops = {
void pci_swiotlb_init(void)
{
/* don't initialize swiotlb if iommu=off (no_iommu=1) */
- if ((!iommu_detected && !no_iommu && end_pfn > MAX_DMA32_PFN) ||
- iommu_pass_through)
+ if ((!iommu_detected && !no_iommu && end_pfn > MAX_DMA32_PFN))
swiotlb = 1;
if (swiotlb_force)
swiotlb = 1;
diff --git a/drivers/pci/dmar.c b/drivers/pci/dmar.c
index 0b2ce5b..65513e9 100644
--- a/drivers/pci/dmar.c
+++ b/drivers/pci/dmar.c
@@ -560,9 +560,6 @@ int __init dmar_table_init(void)
printk(KERN_INFO PREFIX "No ATSR found\n");
#endif
-#ifdef CONFIG_INTR_REMAP
- parse_ioapics_under_ir();
-#endif
return 0;
}
@@ -621,20 +618,32 @@ int alloc_iommu(struct dmar_drhd_unit *drhd)
iommu->cap = dmar_readq(iommu->reg + DMAR_CAP_REG);
iommu->ecap = dmar_readq(iommu->reg + DMAR_ECAP_REG);
+ if (iommu->cap == (uint64_t)-1 && iommu->ecap == (uint64_t)-1) {
+ /* Promote an attitude of violence to a BIOS engineer today */
+ printk(KERN_WARNING PREFIX
+ "Your BIOS is broken; DMAR reported at address %llx returns all ones!\n"
+ "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
+ drhd->reg_base_addr,
+ dmi_get_system_info(DMI_BIOS_VENDOR),
+ dmi_get_system_info(DMI_BIOS_VERSION),
+ dmi_get_system_info(DMI_PRODUCT_VERSION));
+ goto err_unmap;
+ }
+
#ifdef CONFIG_DMAR
agaw = iommu_calculate_agaw(iommu);
if (agaw < 0) {
printk(KERN_ERR
"Cannot get a valid agaw for iommu (seq_id = %d)\n",
iommu->seq_id);
- goto error;
+ goto err_unmap;
}
msagaw = iommu_calculate_max_sagaw(iommu);
if (msagaw < 0) {
printk(KERN_ERR
"Cannot get a valid max agaw for iommu (seq_id = %d)\n",
iommu->seq_id);
- goto error;
+ goto err_unmap;
}
#endif
iommu->agaw = agaw;
@@ -654,7 +663,7 @@ int alloc_iommu(struct dmar_drhd_unit *drhd)
}
ver = readl(iommu->reg + DMAR_VER_REG);
- pr_debug("IOMMU %llx: ver %d:%d cap %llx ecap %llx\n",
+ pr_info("IOMMU %llx: ver %d:%d cap %llx ecap %llx\n",
(unsigned long long)drhd->reg_base_addr,
DMAR_VER_MAJOR(ver), DMAR_VER_MINOR(ver),
(unsigned long long)iommu->cap,
@@ -664,7 +673,10 @@ int alloc_iommu(struct dmar_drhd_unit *drhd)
drhd->iommu = iommu;
return 0;
-error:
+
+ err_unmap:
+ iounmap(iommu->reg);
+ error:
kfree(iommu);
return -1;
}
@@ -826,6 +838,7 @@ restart:
cpu_relax();
spin_lock(&qi->q_lock);
}
+
qi->desc_status[index] = QI_DONE;
reclaim_free_desc(qi);
@@ -1086,8 +1099,8 @@ void dmar_msi_unmask(unsigned int irq)
void dmar_msi_mask(unsigned int irq)
{
- struct intel_iommu *iommu = get_irq_data(irq);
unsigned long flag;
+ struct intel_iommu *iommu = get_irq_data(irq);
/* mask it */
spin_lock_irqsave(&iommu->register_lock, flag);
@@ -1176,7 +1189,7 @@ irqreturn_t dmar_fault(int irq, void *dev_id, struct pt_regs *regs)
source_id, guest_addr);
fault_index++;
- if (fault_index > cap_num_fault_regs(iommu->cap))
+ if (fault_index >= cap_num_fault_regs(iommu->cap))
fault_index = 0;
spin_lock_irqsave(&iommu->register_lock, flag);
}
diff --git a/drivers/pci/intel-iommu.c b/drivers/pci/intel-iommu.c
index 88eb575..7898263 100644
--- a/drivers/pci/intel-iommu.c
+++ b/drivers/pci/intel-iommu.c
@@ -37,6 +37,7 @@
#include <linux/iommu.h>
#include <linux/intel-iommu.h>
#include <linux/sysdev.h>
+#include <linux/dmi.h>
#include <asm/cacheflush.h>
#include <asm/proto.h>
#include "pci.h"
@@ -55,11 +56,39 @@
#define MAX_AGAW_WIDTH 64
-#define DOMAIN_MAX_ADDR(gaw) ((((u64)1) << gaw) - 1)
+#define __DOMAIN_MAX_PFN(gaw) ((((uint64_t)1) << (gaw-VTD_PAGE_SHIFT)) - 1)
+#define __DOMAIN_MAX_ADDR(gaw) ((((uint64_t)1) << gaw) - 1)
+
+/* We limit DOMAIN_MAX_PFN to fit in an unsigned long, and DOMAIN_MAX_ADDR
+ to match. That way, we can use 'unsigned long' for PFNs with impunity. */
+#define DOMAIN_MAX_PFN(gaw) ((unsigned long) min_t(uint64_t, \
+ __DOMAIN_MAX_PFN(gaw), (unsigned long)-1))
+#define DOMAIN_MAX_ADDR(gaw) (((uint64_t)__DOMAIN_MAX_PFN(gaw)) << VTD_PAGE_SHIFT)
#define IOVA_PFN(addr) ((addr) >> PAGE_SHIFT)
-#define DMA_32BIT_PFN IOVA_PFN(DMA_32BIT_MASK)
-#define DMA_64BIT_PFN IOVA_PFN(DMA_64BIT_MASK)
+#define DMA_32BIT_PFN IOVA_PFN(DMA_BIT_MASK(32))
+#define DMA_64BIT_PFN IOVA_PFN(DMA_BIT_MASK(64))
+
+
+/* VT-d pages must always be _smaller_ than MM pages. Otherwise things
+ are never going to work. */
+static inline unsigned long dma_to_mm_pfn(unsigned long dma_pfn)
+{
+ return dma_pfn >> (PAGE_SHIFT - VTD_PAGE_SHIFT);
+}
+
+static inline unsigned long mm_to_dma_pfn(unsigned long mm_pfn)
+{
+ return mm_pfn << (PAGE_SHIFT - VTD_PAGE_SHIFT);
+}
+static inline unsigned long page_to_dma_pfn(struct page *pg)
+{
+ return mm_to_dma_pfn(page_to_pfn(pg));
+}
+static inline unsigned long virt_to_dma_pfn(void *p)
+{
+ return page_to_dma_pfn(virt_to_page(p));
+}
/* global iommu list, set NULL for ignored DMAR units */
static struct intel_iommu **g_iommus;
@@ -203,9 +232,9 @@ static inline u64 dma_pte_addr(struct dma_pte *pte)
return (pte->val & VTD_PAGE_MASK);
}
-static inline void dma_set_pte_addr(struct dma_pte *pte, u64 addr)
+static inline void dma_set_pte_pfn(struct dma_pte *pte, unsigned long pfn)
{
- pte->val |= (addr & VTD_PAGE_MASK);
+ pte->val |= (uint64_t)pfn << VTD_PAGE_SHIFT;
}
static inline bool dma_pte_present(struct dma_pte *pte)
@@ -213,6 +242,20 @@ static inline bool dma_pte_present(struct dma_pte *pte)
return (pte->val & 3) != 0;
}
+static inline int first_pte_in_page(struct dma_pte *pte)
+{
+ return !((unsigned long)pte & ~VTD_PAGE_MASK);
+}
+
+/*
+ * This domain is a statically identity mapping domain.
+ * 1. This domain creats a static 1:1 mapping to all usable memory.
+ * 2. It maps to each iommu if successful.
+ * 3. Each iommu mapps to this domain if successful.
+ */
+static struct dmar_domain *si_domain;
+static int hw_pass_through = 1;
+
/* devices under the same p2p bridge are owned in one domain */
#define DOMAIN_FLAG_P2P_MULTIPLE_DEVICES (1 << 0)
@@ -221,6 +264,9 @@ static inline bool dma_pte_present(struct dma_pte *pte)
*/
#define DOMAIN_FLAG_VIRTUAL_MACHINE (1 << 1)
+/* si_domain contains mulitple devices */
+#define DOMAIN_FLAG_STATIC_IDENTITY (1 << 2)
+
struct dmar_domain {
int id; /* domain id */
unsigned long iommu_bmp; /* bitmap of iommus this domain uses*/
@@ -229,7 +275,6 @@ struct dmar_domain {
struct iova_domain iovad; /* iova's that belong to this domain */
struct dma_pte *pgd; /* virtual address */
- spinlock_t mapping_lock; /* page table lock */
int gaw; /* max guest address width */
/* adjusted guest address width, 0 is level 2 30-bit */
@@ -289,7 +334,6 @@ int dmar_disabled = 1;
static int __initdata dmar_map_gfx = 1;
static int dmar_forcedac;
static int intel_iommu_strict;
-int iommu_pass_through;
#define DUMMY_DEVICE_DOMAIN_INFO ((struct device_domain_info *)(-1))
static DEFINE_SPINLOCK(device_domain_lock);
@@ -432,12 +476,14 @@ int iommu_calculate_agaw(struct intel_iommu *iommu)
return __iommu_calculate_agaw(iommu, DEFAULT_DOMAIN_ADDRESS_WIDTH);
}
-/* in native case, each domain is related to only one iommu */
+/* This functionin only returns single iommu in a domain */
static struct intel_iommu *domain_get_iommu(struct dmar_domain *domain)
{
int iommu_id;
+ /* si_domain and vm domain should not get here. */
BUG_ON(domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE);
+ BUG_ON(domain->flags & DOMAIN_FLAG_STATIC_IDENTITY);
iommu_id = find_first_bit(&domain->iommu_bmp, g_num_of_iommus);
if (iommu_id < 0 || iommu_id >= g_num_of_iommus)
@@ -633,80 +679,78 @@ static inline int width_to_agaw(int width)
static inline unsigned int level_to_offset_bits(int level)
{
- return (12 + (level - 1) * LEVEL_STRIDE);
+ return (level - 1) * LEVEL_STRIDE;
}
-static inline int address_level_offset(u64 addr, int level)
+static inline int pfn_level_offset(unsigned long pfn, int level)
{
- return ((addr >> level_to_offset_bits(level)) & LEVEL_MASK);
+ return (pfn >> level_to_offset_bits(level)) & LEVEL_MASK;
}
-static inline u64 level_mask(int level)
+static inline unsigned long level_mask(int level)
{
- return ((u64)-1 << level_to_offset_bits(level));
+ return -1UL << level_to_offset_bits(level);
}
-static inline u64 level_size(int level)
+static inline unsigned long level_size(int level)
{
- return ((u64)1 << level_to_offset_bits(level));
+ return 1UL << level_to_offset_bits(level);
}
-static inline u64 align_to_level(u64 addr, int level)
+static inline unsigned long align_to_level(unsigned long pfn, int level)
{
- return ((addr + level_size(level) - 1) & level_mask(level));
+ return (pfn + level_size(level) - 1) & level_mask(level);
}
-static struct dma_pte * addr_to_dma_pte(struct dmar_domain *domain, u64 addr)
+static struct dma_pte *pfn_to_dma_pte(struct dmar_domain *domain,
+ unsigned long pfn)
{
- int addr_width = agaw_to_width(domain->agaw);
+ int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
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;
+ BUG_ON(addr_width < BITS_PER_LONG && pfn >> addr_width);
parent = domain->pgd;
- spin_lock_irqsave(&domain->mapping_lock, flags);
while (level > 0) {
void *tmp_page;
- offset = address_level_offset(addr, level);
+ offset = pfn_level_offset(pfn, level);
pte = &parent[offset];
if (level == 1)
break;
if (!dma_pte_present(pte)) {
+ uint64_t pteval;
+
tmp_page = alloc_pgtable_page();
- if (!tmp_page) {
- spin_unlock_irqrestore(&domain->mapping_lock,
- flags);
+ if (!tmp_page)
return NULL;
+
+ domain_flush_cache(domain, tmp_page, VTD_PAGE_SIZE);
+ pteval = ((uint64_t)virt_to_dma_pfn(tmp_page) << VTD_PAGE_SHIFT) | DMA_PTE_READ | DMA_PTE_WRITE;
+ if (cmpxchg(&pte->val, 0ULL, pteval)) {
+ /* Someone else set it while we were thinking; use theirs. */
+ free_pgtable_page(tmp_page);
+ } else {
+ dma_pte_addr(pte);
+ domain_flush_cache(domain, pte, sizeof(*pte));
}
- domain_flush_cache(domain, tmp_page, PAGE_SIZE);
- 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);
- domain_flush_cache(domain, 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)
+static struct dma_pte *dma_pfn_level_pte(struct dmar_domain *domain,
+ unsigned long pfn,
+ int level)
{
struct dma_pte *parent, *pte = NULL;
int total = agaw_to_level(domain->agaw);
@@ -714,7 +758,7 @@ static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr,
parent = domain->pgd;
while (level <= total) {
- offset = address_level_offset(addr, total);
+ offset = pfn_level_offset(pfn, total);
pte = &parent[offset];
if (level == total)
return pte;
@@ -727,74 +771,85 @@ static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr,
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);
- domain_flush_cache(domain, 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);
- int npages;
+static void dma_pte_clear_range(struct dmar_domain *domain,
+ unsigned long start_pfn,
+ unsigned long last_pfn)
+{
+ int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
+ struct dma_pte *first_pte, *pte;
+
+ BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
+ BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
+ BUG_ON(start_pfn > last_pfn);
+
+ /* we don't need lock here; nobody else touches the iova range */
+ do {
+ first_pte = pte = dma_pfn_level_pte(domain, start_pfn, 1);
+ if (!pte) {
+ start_pfn = align_to_level(start_pfn + 1, 2);
+ continue;
+ }
+ do {
+ dma_clear_pte(pte);
+ start_pfn++;
+ pte++;
+ } while (start_pfn <= last_pfn && !first_pte_in_page(pte));
- start &= (((u64)1) << addr_width) - 1;
- end &= (((u64)1) << addr_width) - 1;
- /* in case it's partial page */
- start &= PAGE_MASK;
- end = PAGE_ALIGN(end);
- npages = (end - start) / VTD_PAGE_SIZE;
+ domain_flush_cache(domain, first_pte,
+ (void *)pte - (void *)first_pte);
- /* we don't need lock here, nobody else touches the iova range */
- while (npages--) {
- dma_pte_clear_one(domain, start);
- start += VTD_PAGE_SIZE;
- }
+ } while (start_pfn && start_pfn <= last_pfn);
}
/* 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)
+ unsigned long start_pfn,
+ unsigned long last_pfn)
{
- int addr_width = agaw_to_width(domain->agaw);
- struct dma_pte *pte;
+ int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
+ struct dma_pte *first_pte, *pte;
int total = agaw_to_level(domain->agaw);
int level;
- u64 tmp;
+ unsigned long tmp;
- start &= (((u64)1) << addr_width) - 1;
- end &= (((u64)1) << addr_width) - 1;
+ BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
+ BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
+ BUG_ON(start_pfn > last_pfn);
- /* we don't need lock here, nobody else touches the iova range */
+ /* 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))
+ tmp = align_to_level(start_pfn, level);
+
+ /* If we can't even clear one PTE at this level, we're done */
+ if (tmp + level_size(level) - 1 > last_pfn)
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);
- domain_flush_cache(domain, pte, sizeof(*pte));
+ do {
+ first_pte = pte = dma_pfn_level_pte(domain, tmp, level);
+ if (!pte) {
+ tmp = align_to_level(tmp + 1, level + 1);
+ continue;
}
- tmp += level_size(level);
- }
+ do {
+ if (dma_pte_present(pte)) {
+ free_pgtable_page(phys_to_virt(dma_pte_addr(pte)));
+ dma_clear_pte(pte);
+ }
+ pte++;
+ tmp += level_size(level);
+ } while (!first_pte_in_page(pte) &&
+ tmp + level_size(level) - 1 <= last_pfn);
+
+ domain_flush_cache(domain, first_pte,
+ (void *)pte - (void *)first_pte);
+
+ } while (tmp && tmp + level_size(level) - 1 <= last_pfn);
level++;
}
/* free pgd */
- if (start == 0 && end >= ((((u64)1) << addr_width) - 1)) {
+ if (start_pfn == 0 && last_pfn == DOMAIN_MAX_PFN(domain->gaw)) {
free_pgtable_page(domain->pgd);
domain->pgd = NULL;
}
@@ -834,7 +889,7 @@ static void iommu_set_root_entry(struct intel_iommu *iommu)
/* Make sure hardware complete it */
IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
- readl, (sts & DMA_GSTS_RTPS), sts);
+ readl, (sts & DMA_GSTS_RTPS), sts);
spin_unlock_irqrestore(&iommu->register_lock, flag);
}
@@ -893,7 +948,7 @@ static void __iommu_flush_context(struct intel_iommu *iommu,
/* return value determine if we need a write buffer flush */
static void __iommu_flush_iotlb(struct intel_iommu *iommu, u16 did,
- u64 addr, unsigned int size_order, u64 type)
+ u64 addr, unsigned int size_order, u64 type)
{
int tlb_offset = ecap_iotlb_offset(iommu->ecap);
u64 val = 0, val_iva = 0;
@@ -941,7 +996,7 @@ static void __iommu_flush_iotlb(struct intel_iommu *iommu, u16 did,
/* check IOTLB invalidation granularity */
if (DMA_TLB_IAIG(val) == 0)
- printk(KERN_ERR"IOMMU: flush IOTLB failed\n");
+ printk(KERN_ERR "IOMMU: flush IOTLB failed\n");
if (DMA_TLB_IAIG(val) != DMA_TLB_IIRG(type))
pr_debug("IOMMU: tlb flush request %Lx, actual %Lx\n",
(unsigned long long)DMA_TLB_IIRG(type),
@@ -1020,14 +1075,14 @@ static void iommu_flush_dev_iotlb(struct dmar_domain *domain,
}
static void iommu_flush_iotlb_psi(struct intel_iommu *iommu, u16 did,
- u64 addr, unsigned int pages)
+ unsigned long pfn, unsigned int pages)
{
unsigned int mask = ilog2(__roundup_pow_of_two(pages));
+ uint64_t addr = (uint64_t)pfn << VTD_PAGE_SHIFT;
- BUG_ON(addr & (~VTD_PAGE_MASK));
BUG_ON(pages == 0);
- /*
+ /*
* Fallback to domain selective flush if no PSI support or the size is
* too big.
* PSI requires page size to be 2 ^ x, and the base address is naturally
@@ -1035,12 +1090,16 @@ static void iommu_flush_iotlb_psi(struct intel_iommu *iommu, u16 did,
*/
if (!cap_pgsel_inv(iommu->cap) || mask > cap_max_amask_val(iommu->cap))
iommu->flush.flush_iotlb(iommu, did, 0, 0,
- DMA_TLB_DSI_FLUSH);
+ DMA_TLB_DSI_FLUSH);
else
iommu->flush.flush_iotlb(iommu, did, addr, mask,
- DMA_TLB_DSI_FLUSH);
+ DMA_TLB_PSI_FLUSH);
- if (did)
+ /*
+ * In caching mode, domain ID 0 is reserved for non-present to present
+ * mapping flush. Device IOTLB doesn't need to be flushed in this case.
+ */
+ if (!cap_caching_mode(iommu->cap) || did)
iommu_flush_dev_iotlb(iommu->domains[did], addr, mask);
}
@@ -1068,7 +1127,7 @@ static int iommu_enable_translation(struct intel_iommu *iommu)
spin_lock_irqsave(&iommu->register_lock, flags);
iommu->gcmd |= DMA_GCMD_TE;
- writel(iommu->gcmd|DMA_GCMD_TE, iommu->reg + DMAR_GCMD_REG);
+ writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
/* Make sure hardware complete it */
IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
@@ -1105,6 +1164,8 @@ static int iommu_init_domains(struct intel_iommu *iommu)
pr_debug("Number of Domains supportd <%ld>\n", ndomains);
nlongs = BITS_TO_LONGS(ndomains);
+ spin_lock_init(&iommu->lock);
+
/* TBD: there might be 64K domains,
* consider other allocation for future chip
*/
@@ -1117,12 +1178,9 @@ static int iommu_init_domains(struct intel_iommu *iommu)
GFP_KERNEL);
if (!iommu->domains) {
printk(KERN_ERR "Allocating domain array failed\n");
- kfree(iommu->domain_ids);
return -ENOMEM;
}
- spin_lock_init(&iommu->lock);
-
/*
* if Caching mode is set, then invalid translations are tagged
* with domainid 0. Hence we need to pre-allocate it.
@@ -1142,22 +1200,24 @@ void free_dmar_iommu(struct intel_iommu *iommu)
int i;
unsigned long flags;
- i = find_first_bit(iommu->domain_ids, cap_ndoms(iommu->cap));
- for (; i < cap_ndoms(iommu->cap); ) {
- domain = iommu->domains[i];
- clear_bit(i, iommu->domain_ids);
+ if ((iommu->domains) && (iommu->domain_ids)) {
+ i = find_first_bit(iommu->domain_ids, cap_ndoms(iommu->cap));
+ for (; i < cap_ndoms(iommu->cap); ) {
+ domain = iommu->domains[i];
+ clear_bit(i, iommu->domain_ids);
+
+ spin_lock_irqsave(&domain->iommu_lock, flags);
+ if (--domain->iommu_count == 0) {
+ if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE)
+ vm_domain_exit(domain);
+ else
+ domain_exit(domain);
+ }
+ spin_unlock_irqrestore(&domain->iommu_lock, flags);
- spin_lock_irqsave(&domain->iommu_lock, flags);
- if (--domain->iommu_count == 0) {
- if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE)
- vm_domain_exit(domain);
- else
- domain_exit(domain);
+ i = find_next_bit(iommu->domain_ids,
+ cap_ndoms(iommu->cap), i+1);
}
- spin_unlock_irqrestore(&domain->iommu_lock, flags);
-
- i = find_next_bit(iommu->domain_ids,
- cap_ndoms(iommu->cap), i+1);
}
if (iommu->gcmd & DMA_GCMD_TE)
@@ -1188,53 +1248,75 @@ void free_dmar_iommu(struct intel_iommu *iommu)
free_context_table(iommu);
}
-static struct dmar_domain * iommu_alloc_domain(struct intel_iommu *iommu)
+static struct dmar_domain *alloc_domain(void)
{
- unsigned long num;
- unsigned long ndomains;
struct dmar_domain *domain;
- unsigned long flags;
domain = alloc_domain_mem();
if (!domain)
return NULL;
+ memset(&domain->iommu_bmp, 0, sizeof(unsigned long));
+ domain->flags = 0;
+
+ return domain;
+}
+
+static int iommu_attach_domain(struct dmar_domain *domain,
+ struct intel_iommu *iommu)
+{
+ int num;
+ unsigned long ndomains;
+ unsigned long flags;
+
ndomains = cap_ndoms(iommu->cap);
spin_lock_irqsave(&iommu->lock, flags);
+
num = find_first_zero_bit(iommu->domain_ids, ndomains);
if (num >= ndomains) {
spin_unlock_irqrestore(&iommu->lock, flags);
- free_domain_mem(domain);
printk(KERN_ERR "IOMMU: no free domain ids\n");
- return NULL;
+ return -ENOMEM;
}
- set_bit(num, iommu->domain_ids);
domain->id = num;
- memset(&domain->iommu_bmp, 0, sizeof(unsigned long));
+ set_bit(num, iommu->domain_ids);
set_bit(iommu->seq_id, &domain->iommu_bmp);
- domain->flags = 0;
iommu->domains[num] = domain;
spin_unlock_irqrestore(&iommu->lock, flags);
- return domain;
+ return 0;
}
-static void iommu_free_domain(struct dmar_domain *domain)
+static void iommu_detach_domain(struct dmar_domain *domain,
+ struct intel_iommu *iommu)
{
unsigned long flags;
- struct intel_iommu *iommu;
-
- iommu = domain_get_iommu(domain);
+ int num, ndomains;
+ int found = 0;
spin_lock_irqsave(&iommu->lock, flags);
- clear_bit(domain->id, iommu->domain_ids);
+ ndomains = cap_ndoms(iommu->cap);
+ num = find_first_bit(iommu->domain_ids, ndomains);
+ for (; num < ndomains; ) {
+ if (iommu->domains[num] == domain) {
+ found = 1;
+ break;
+ }
+ num = find_next_bit(iommu->domain_ids,
+ cap_ndoms(iommu->cap), num+1);
+ }
+
+ if (found) {
+ clear_bit(num, iommu->domain_ids);
+ clear_bit(iommu->seq_id, &domain->iommu_bmp);
+ iommu->domains[num] = NULL;
+ }
spin_unlock_irqrestore(&iommu->lock, flags);
}
static struct iova_domain reserved_iova_list;
-static struct lock_class_key reserved_alloc_key;
static struct lock_class_key reserved_rbtree_key;
static void dmar_init_reserved_ranges(void)
@@ -1242,12 +1324,9 @@ static void dmar_init_reserved_ranges(void)
struct pci_dev *pdev = NULL;
struct iova *iova;
int i;
- u64 addr, size;
init_iova_domain(&reserved_iova_list, DMA_32BIT_PFN);
- lockdep_set_class(&reserved_iova_list.iova_alloc_lock,
- &reserved_alloc_key);
lockdep_set_class(&reserved_iova_list.iova_rbtree_lock,
&reserved_rbtree_key);
@@ -1265,12 +1344,9 @@ static void dmar_init_reserved_ranges(void)
r = &pdev->resource[i];
if (!r->flags || !(r->flags & IORESOURCE_MEM))
continue;
- addr = r->start;
- addr &= PAGE_MASK;
- size = r->end - addr;
- size = PAGE_ALIGN(size);
- iova = reserve_iova(&reserved_iova_list, IOVA_PFN(addr),
- IOVA_PFN(size + addr) - 1);
+ iova = reserve_iova(&reserved_iova_list,
+ IOVA_PFN(r->start),
+ IOVA_PFN(r->end));
if (!iova)
printk(KERN_ERR "Reserve iova failed\n");
}
@@ -1304,7 +1380,6 @@ static int domain_init(struct dmar_domain *domain, int guest_width)
unsigned long sagaw;
init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
- spin_lock_init(&domain->mapping_lock);
spin_lock_init(&domain->iommu_lock);
domain_reserve_special_ranges(domain);
@@ -1349,7 +1424,8 @@ static int domain_init(struct dmar_domain *domain, int guest_width)
static void domain_exit(struct dmar_domain *domain)
{
- u64 end;
+ struct dmar_drhd_unit *drhd;
+ struct intel_iommu *iommu;
/* Domain 0 is reserved, so dont process it */
if (!domain)
@@ -1358,16 +1434,17 @@ static void domain_exit(struct dmar_domain *domain)
domain_remove_dev_info(domain);
/* destroy iovas */
put_iova_domain(&domain->iovad);
- end = DOMAIN_MAX_ADDR(domain->gaw);
- end = end & (~PAGE_MASK);
/* clear ptes */
- dma_pte_clear_range(domain, 0, end);
+ dma_pte_clear_range(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
/* free page tables */
- dma_pte_free_pagetable(domain, 0, end);
+ dma_pte_free_pagetable(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
+
+ for_each_active_iommu(iommu, drhd)
+ if (test_bit(iommu->seq_id, &domain->iommu_bmp))
+ iommu_detach_domain(domain, iommu);
- iommu_free_domain(domain);
free_domain_mem(domain);
}
@@ -1407,7 +1484,8 @@ static int domain_context_mapping_one(struct dmar_domain *domain, int segment,
id = domain->id;
pgd = domain->pgd;
- if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) {
+ if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE ||
+ domain->flags & DOMAIN_FLAG_STATIC_IDENTITY) {
int found = 0;
/* find an available domain id for this device in iommu */
@@ -1455,12 +1533,11 @@ static int domain_context_mapping_one(struct dmar_domain *domain, int segment,
translation = info ? CONTEXT_TT_DEV_IOTLB :
CONTEXT_TT_MULTI_LEVEL;
}
-
/*
* In pass through mode, AW must be programmed to indicate the largest
* AGAW value supported by hardware. And ASR is ignored by hardware.
*/
- if (unlikely(translation == CONTEXT_TT_PASS_THROUGH))
+ if (unlikely(translation == CONTEXT_TT_PASS_THROUGH))
context_set_address_width(context, iommu->msagaw);
else {
context_set_address_root(context, virt_to_phys(pgd));
@@ -1487,7 +1564,6 @@ static int domain_context_mapping_one(struct dmar_domain *domain, int segment,
} else {
iommu_flush_write_buffer(iommu);
}
-
iommu_enable_dev_iotlb(info);
spin_unlock_irqrestore(&iommu->lock, flags);
@@ -1576,42 +1652,99 @@ static int domain_context_mapped(struct pci_dev *pdev)
tmp->devfn);
}
-static int
-domain_page_mapping(struct dmar_domain *domain, dma_addr_t iova,
- u64 hpa, size_t size, int prot)
+/* Returns a number of VTD pages, but aligned to MM page size */
+static inline unsigned long aligned_nrpages(unsigned long host_addr,
+ size_t size)
{
- u64 start_pfn, end_pfn;
- struct dma_pte *pte;
- int index;
- int addr_width = agaw_to_width(domain->agaw);
+ host_addr &= ~PAGE_MASK;
+ return PAGE_ALIGN(host_addr + size) >> VTD_PAGE_SHIFT;
+}
- hpa &= (((u64)1) << addr_width) - 1;
+static inline struct page *sg_page(struct scatterlist *sg)
+{
+ return sg->page;
+}
+
+static int __domain_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+ struct scatterlist *sg, unsigned long phys_pfn,
+ unsigned long nr_pages, int prot)
+{
+ struct dma_pte *first_pte = NULL, *pte = NULL;
+ phys_addr_t uninitialized_var(pteval);
+ int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
+ unsigned long sg_res;
+
+ BUG_ON(addr_width < BITS_PER_LONG && (iov_pfn + nr_pages - 1) >> addr_width);
if ((prot & (DMA_PTE_READ|DMA_PTE_WRITE)) == 0)
return -EINVAL;
- iova &= PAGE_MASK;
- start_pfn = ((u64)hpa) >> VTD_PAGE_SHIFT;
- end_pfn = (VTD_PAGE_ALIGN(((u64)hpa) + size)) >> VTD_PAGE_SHIFT;
- index = 0;
- while (start_pfn < end_pfn) {
- pte = addr_to_dma_pte(domain, iova + VTD_PAGE_SIZE * index);
- if (!pte)
- return -ENOMEM;
+
+ prot &= DMA_PTE_READ | DMA_PTE_WRITE | DMA_PTE_SNP;
+
+ if (sg)
+ sg_res = 0;
+ else {
+ sg_res = nr_pages + 1;
+ pteval = ((phys_addr_t)phys_pfn << VTD_PAGE_SHIFT) | prot;
+ }
+
+ while (nr_pages--) {
+ uint64_t tmp;
+
+ if (!sg_res) {
+ sg_res = aligned_nrpages(sg->offset, sg->length);
+ sg->dma_address = ((dma_addr_t)iov_pfn << VTD_PAGE_SHIFT) + sg->offset;
+ sg->dma_length = sg->length;
+ pteval = page_to_phys(sg_page(sg)) | prot;
+ }
+ if (!pte) {
+ first_pte = pte = pfn_to_dma_pte(domain, iov_pfn);
+ if (!pte)
+ return -ENOMEM;
+ }
/* We don't need lock here, nobody else
* touches the iova range
*/
- BUG_ON(dma_pte_addr(pte));
- dma_set_pte_addr(pte, start_pfn << VTD_PAGE_SHIFT);
- dma_set_pte_prot(pte, prot);
- if (prot & DMA_PTE_SNP)
- dma_set_pte_snp(pte);
- domain_flush_cache(domain, pte, sizeof(*pte));
- start_pfn++;
- index++;
+ tmp = cmpxchg(&pte->val, 0ULL, pteval);
+ if (tmp) {
+ static int dumps = 5;
+ printk(KERN_CRIT "ERROR: DMA PTE for vPFN 0x%lx already set (to %llx not %llx)\n",
+ iov_pfn, tmp, (unsigned long long)pteval);
+ if (dumps) {
+ dumps--;
+ debug_dma_dump_mappings(NULL);
+ }
+ WARN_ON_ONCE(1);
+ }
+ pte++;
+ if (!nr_pages || first_pte_in_page(pte)) {
+ domain_flush_cache(domain, first_pte,
+ (void *)pte - (void *)first_pte);
+ pte = NULL;
+ }
+ iov_pfn++;
+ pteval += VTD_PAGE_SIZE;
+ sg_res--;
+ if (!sg_res)
+ sg++;
}
return 0;
}
+static inline int domain_sg_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+ struct scatterlist *sg, unsigned long nr_pages,
+ int prot)
+{
+ return __domain_mapping(domain, iov_pfn, sg, 0, nr_pages, prot);
+}
+
+static inline int domain_pfn_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+ unsigned long phys_pfn, unsigned long nr_pages,
+ int prot)
+{
+ return __domain_mapping(domain, iov_pfn, NULL, phys_pfn, nr_pages, prot);
+}
+
static void iommu_detach_dev(struct intel_iommu *iommu, u8 bus, u8 devfn)
{
if (!iommu)
@@ -1676,6 +1809,7 @@ static struct dmar_domain *get_domain_for_dev(struct pci_dev *pdev, int gaw)
unsigned long flags;
int bus = 0, devfn = 0;
int segment;
+ int ret;
domain = find_domain(pdev);
if (domain)
@@ -1708,6 +1842,10 @@ static struct dmar_domain *get_domain_for_dev(struct pci_dev *pdev, int gaw)
}
}
+ domain = alloc_domain();
+ if (!domain)
+ goto error;
+
/* Allocate new domain for the device */
drhd = dmar_find_matched_drhd_unit(pdev);
if (!drhd) {
@@ -1717,9 +1855,11 @@ static struct dmar_domain *get_domain_for_dev(struct pci_dev *pdev, int gaw)
}
iommu = drhd->iommu;
- domain = iommu_alloc_domain(iommu);
- if (!domain)
+ ret = iommu_attach_domain(domain, iommu);
+ if (ret) {
+ domain_exit(domain);
goto error;
+ }
if (domain_init(domain, gaw)) {
domain_exit(domain);
@@ -1793,55 +1933,85 @@ error:
return find_domain(pdev);
}
+static int iommu_identity_mapping;
+
+static int iommu_domain_identity_map(struct dmar_domain *domain,
+ unsigned long long start,
+ unsigned long long end)
+{
+ unsigned long first_vpfn = start >> VTD_PAGE_SHIFT;
+ unsigned long last_vpfn = end >> VTD_PAGE_SHIFT;
+
+ if (!reserve_iova(&domain->iovad, dma_to_mm_pfn(first_vpfn),
+ dma_to_mm_pfn(last_vpfn))) {
+ printk(KERN_ERR "IOMMU: reserve iova failed\n");
+ return -ENOMEM;
+ }
+
+ pr_debug("Mapping reserved region %llx-%llx for domain %d\n",
+ start, end, domain->id);
+ /*
+ * RMRR range might have overlap with physical memory range,
+ * clear it first
+ */
+ dma_pte_clear_range(domain, first_vpfn, last_vpfn);
+
+ return domain_pfn_mapping(domain, first_vpfn, first_vpfn,
+ last_vpfn - first_vpfn + 1,
+ DMA_PTE_READ|DMA_PTE_WRITE);
+}
+
static int iommu_prepare_identity_map(struct pci_dev *pdev,
unsigned long long start,
unsigned long long end)
{
struct dmar_domain *domain;
- unsigned long size;
- unsigned long long base;
int ret;
- printk(KERN_INFO
- "IOMMU: Setting identity map for device %s [0x%Lx - 0x%Lx]\n",
- pci_name(pdev), start, end);
- /* page table init */
domain = get_domain_for_dev(pdev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
if (!domain)
return -ENOMEM;
- /* The address might not be aligned */
- base = start & PAGE_MASK;
- size = end - base;
- size = PAGE_ALIGN(size);
- if (!reserve_iova(&domain->iovad, IOVA_PFN(base),
- IOVA_PFN(base + size) - 1)) {
- printk(KERN_ERR "IOMMU: reserve iova failed\n");
- ret = -ENOMEM;
- goto error;
+ /* For _hardware_ passthrough, don't bother. But for software
+ passthrough, we do it anyway -- it may indicate a memory
+ range which is reserved in E820, so which didn't get set
+ up to start with in si_domain */
+ if (domain == si_domain && hw_pass_through) {
+ printk("Ignoring identity map for HW passthrough device %s [0x%Lx - 0x%Lx]\n",
+ pci_name(pdev), start, end);
+ return 0;
}
- pr_debug("Mapping reserved region %lx@%llx for %s\n",
- size, base, pci_name(pdev));
- /*
- * RMRR range might have overlap with physical memory range,
- * clear it first
- */
- dma_pte_clear_range(domain, base, base + size);
+ printk(KERN_INFO
+ "IOMMU: Setting identity map for device %s [0x%Lx - 0x%Lx]\n",
+ pci_name(pdev), start, end);
+
+ if (end >> agaw_to_width(domain->agaw)) {
+ printk(KERN_WARNING "IOMMU: Your BIOS is broken;"
+ "RMRR exceeds permitted address width (%d bits)\n"
+ "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
+ agaw_to_width(domain->agaw),
+ dmi_get_system_info(DMI_BIOS_VENDOR),
+ dmi_get_system_info(DMI_BIOS_VERSION),
+ dmi_get_system_info(DMI_PRODUCT_VERSION));
+ ret = -EIO;
+ goto error;
+ }
- ret = domain_page_mapping(domain, base, base, size,
- DMA_PTE_READ|DMA_PTE_WRITE);
+ ret = iommu_domain_identity_map(domain, start, end);
if (ret)
goto error;
/* context entry init */
ret = domain_context_mapping(domain, pdev, CONTEXT_TT_MULTI_LEVEL);
- if (!ret)
- return 0;
-error:
+ if (ret)
+ goto error;
+
+ return 0;
+
+ error:
domain_exit(domain);
return ret;
-
}
static inline int iommu_prepare_rmrr_dev(struct dmar_rmrr_unit *rmrr,
@@ -1853,124 +2023,218 @@ static inline int iommu_prepare_rmrr_dev(struct dmar_rmrr_unit *rmrr,
rmrr->end_address + 1);
}
-#ifdef CONFIG_DMAR_GFX_WA
-struct iommu_prepare_data {
+#ifdef CONFIG_DMAR_FLOPPY_WA
+static inline void iommu_prepare_isa(void)
+{
struct pci_dev *pdev;
int ret;
-};
-static int __init iommu_prepare_work_fn(unsigned long start_pfn,
- unsigned long end_pfn, void *datax)
-{
- struct iommu_prepare_data *data;
+ pdev = pci_get_class(PCI_CLASS_BRIDGE_ISA << 8, NULL);
+ if (!pdev)
+ return;
- data = (struct iommu_prepare_data *)datax;
+ printk(KERN_INFO "IOMMU: Prepare 0-16MiB unity mapping for LPC\n");
+ ret = iommu_prepare_identity_map(pdev, 0, 16*1024*1024);
- data->ret = iommu_prepare_identity_map(data->pdev,
- start_pfn<<PAGE_SHIFT, end_pfn<<PAGE_SHIFT);
- return data->ret;
+ if (ret)
+ printk(KERN_ERR "IOMMU: Failed to create 0-16MiB identity map; "
+ "floppy might not work\n");
}
+#else
+static inline void iommu_prepare_isa(void)
+{
+ return;
+}
+#endif /* !CONFIG_DMAR_FLPY_WA */
-static int __init iommu_prepare_with_active_regions(struct pci_dev *pdev)
+static int md_domain_init(struct dmar_domain *domain, int guest_width);
+
+static int __init si_domain_work_fn(unsigned long start_pfn,
+ unsigned long end_pfn)
{
- struct iommu_prepare_data data;
- struct zone *zone;
+ return iommu_domain_identity_map(si_domain,
+ (uint64_t)start_pfn << PAGE_SHIFT,
+ (uint64_t)end_pfn << PAGE_SHIFT);
+}
- data.pdev = pdev;
- data.ret = 0;
+static int __init iommu_prepare_with_active_regions(void)
+{
+ struct zone *zone;
+ int ret;
- for_each_zone(zone) {
- unsigned long start_pfn, end_pfn;
+ for_each_zone(zone) {
+ unsigned long start_pfn, end_pfn;
- start_pfn = zone->zone_start_pfn;
- end_pfn = start_pfn + zone->spanned_pages;
+ start_pfn = zone->zone_start_pfn;
+ end_pfn = start_pfn + zone->spanned_pages;
- iommu_prepare_work_fn(start_pfn, end_pfn, &data);
- if (data.ret)
- return data.ret;
- }
- return data.ret;
+ ret = si_domain_work_fn(start_pfn, end_pfn);
+ if (ret)
+ return ret;
+ }
+ return ret;
}
-static void __init iommu_prepare_gfx_mapping(void)
+static int __init si_domain_init(int hw)
{
- struct pci_dev *pdev = NULL;
- int ret;
+ struct dmar_drhd_unit *drhd;
+ struct intel_iommu *iommu;
+ int ret = 0;
- for_each_pci_dev(pdev) {
- if (pdev->iommu == DUMMY_DEVICE_DOMAIN_INFO ||
- !IS_GFX_DEVICE(pdev))
- continue;
- printk(KERN_INFO "IOMMU: gfx device %s 1-1 mapping\n",
- pci_name(pdev));
- ret = iommu_prepare_with_active_regions(pdev);
- if (ret)
- printk(KERN_ERR "IOMMU: mapping reserved region failed\n");
+ si_domain = alloc_domain();
+ if (!si_domain)
+ return -EFAULT;
+
+ pr_debug("Identity mapping domain is domain %d\n", si_domain->id);
+
+ for_each_active_iommu(iommu, drhd) {
+ ret = iommu_attach_domain(si_domain, iommu);
+ if (ret) {
+ domain_exit(si_domain);
+ return -EFAULT;
+ }
}
+
+ if (md_domain_init(si_domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) {
+ domain_exit(si_domain);
+ return -EFAULT;
+ }
+
+ si_domain->flags = DOMAIN_FLAG_STATIC_IDENTITY;
+
+ if (hw)
+ return 0;
+
+ ret = iommu_prepare_with_active_regions();
+ if (ret)
+ return ret;
+
+ return 0;
}
-#else /* !CONFIG_DMAR_GFX_WA */
-static inline void iommu_prepare_gfx_mapping(void)
+
+static void domain_remove_one_dev_info(struct dmar_domain *domain,
+ struct pci_dev *pdev);
+static int identity_mapping(struct pci_dev *pdev)
{
- return;
+ struct device_domain_info *info;
+
+ if (likely(!iommu_identity_mapping))
+ return 0;
+
+
+ list_for_each_entry(info, &si_domain->devices, link)
+ if (info->dev == pdev)
+ return 1;
+ return 0;
}
-#endif
-#ifdef CONFIG_DMAR_FLOPPY_WA
-static inline void iommu_prepare_isa(void)
+static int domain_add_dev_info(struct dmar_domain *domain,
+ struct pci_dev *pdev,
+ int translation)
{
- struct pci_dev *pdev;
+ struct device_domain_info *info;
+ unsigned long flags;
int ret;
- pdev = pci_get_class(PCI_CLASS_BRIDGE_ISA << 8, NULL);
- if (!pdev)
- return;
+ info = alloc_devinfo_mem();
+ if (!info)
+ return -ENOMEM;
- printk(KERN_INFO "IOMMU: Prepare 0-16M unity mapping for LPC\n");
- ret = iommu_prepare_identity_map(pdev, 0, 16*1024*1024);
+ ret = domain_context_mapping(domain, pdev, translation);
+ if (ret) {
+ free_devinfo_mem(info);
+ return ret;
+ }
- if (ret)
- printk(KERN_ERR "IOMMU: Failed to create 0-64M identity map, "
- "floppy might not work\n");
+ info->segment = pci_domain_nr(pdev->bus);
+ info->bus = pdev->bus->number;
+ info->devfn = pdev->devfn;
+ info->dev = pdev;
+ info->domain = domain;
+ spin_lock_irqsave(&device_domain_lock, flags);
+ list_add(&info->link, &domain->devices);
+ list_add(&info->global, &device_domain_list);
+ pdev->iommu = info;
+ spin_unlock_irqrestore(&device_domain_lock, flags);
+
+ return 0;
}
-#else
-static inline void iommu_prepare_isa(void)
+
+static int iommu_should_identity_map(struct pci_dev *pdev, int startup)
{
- return;
+ if (iommu_identity_mapping == 2)
+ return IS_GFX_DEVICE(pdev);
+
+ /*
+ * We want to start off with all devices in the 1:1 domain, and
+ * take them out later if we find they can't access all of memory.
+ *
+ * However, we can't do this for PCI devices behind bridges,
+ * because all PCI devices behind the same bridge will end up
+ * with the same source-id on their transactions.
+ *
+ * Practically speaking, we can't change things around for these
+ * devices at run-time, because we can't be sure there'll be no
+ * DMA transactions in flight for any of their siblings.
+ *
+ * So PCI devices (unless they're on the root bus) as well as
+ * their parent PCI-PCI or PCIe-PCI bridges must be left _out_ of
+ * the 1:1 domain, just in _case_ one of their siblings turns out
+ * not to be able to map all of memory.
+ */
+ if (!pdev->is_pcie) {
+ if (!pci_is_root_bus(pdev->bus))
+ return 0;
+ if (pdev->class >> 8 == PCI_CLASS_BRIDGE_PCI)
+ return 0;
+ } else if (pdev->pcie_type == PCI_EXP_TYPE_PCI_BRIDGE)
+ return 0;
+
+ /*
+ * At boot time, we don't yet know if devices will be 64-bit capable.
+ * Assume that they will -- if they turn out not to be, then we can
+ * take them out of the 1:1 domain later.
+ */
+ if (!startup)
+ return pdev->dma_mask > DMA_BIT_MASK(32);
+
+ return 1;
}
-#endif /* !CONFIG_DMAR_FLPY_WA */
-/* Initialize each context entry as pass through.*/
-static int __init init_context_pass_through(void)
+static int __init iommu_prepare_static_identity_mapping(int hw)
{
struct pci_dev *pdev = NULL;
- struct dmar_domain *domain;
int ret;
+ ret = si_domain_init(hw);
+ if (ret)
+ return -EFAULT;
+
for_each_pci_dev(pdev) {
- domain = get_domain_for_dev(pdev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
- if (!domain) {
- printk(KERN_ERR
- "Allocating pass-thru domain for %s failed", pci_name(pdev));
- return -ENODEV;
+ if (iommu_should_identity_map(pdev, 1)) {
+ printk(KERN_INFO "IOMMU: %s identity mapping for device %s\n",
+ hw ? "hardware" : "software", pci_name(pdev));
+
+ ret = domain_add_dev_info(si_domain, pdev,
+ hw ? CONTEXT_TT_PASS_THROUGH :
+ CONTEXT_TT_MULTI_LEVEL);
+ if (ret)
+ return ret;
}
- ret = domain_context_mapping(domain, pdev,
- CONTEXT_TT_PASS_THROUGH);
- if (ret)
- return ret;
}
+
return 0;
}
-static int __init init_dmars(void)
+int __init init_dmars(void)
{
struct dmar_drhd_unit *drhd;
struct dmar_rmrr_unit *rmrr;
struct pci_dev *pdev;
struct intel_iommu *iommu;
int i, ret;
- int pass_through = 1;
/*
* for each drhd
@@ -1998,7 +2262,6 @@ static int __init init_dmars(void)
deferred_flush = kzalloc(g_num_of_iommus *
sizeof(struct deferred_flush_tables), GFP_KERNEL);
if (!deferred_flush) {
- kfree(g_iommus);
ret = -ENOMEM;
goto error;
}
@@ -2025,14 +2288,8 @@ static int __init init_dmars(void)
goto error;
}
if (!ecap_pass_through(iommu->ecap))
- pass_through = 0;
+ hw_pass_through = 0;
}
- if (iommu_pass_through)
- if (!pass_through) {
- printk(KERN_INFO
- "Pass Through is not supported by hardware.\n");
- iommu_pass_through = 0;
- }
/*
* Start from the sane iommu hardware state.
@@ -2088,66 +2345,57 @@ static int __init init_dmars(void)
}
}
-#ifdef CONFIG_INTR_REMAP
- if (!intr_remapping_enabled) {
- ret = enable_intr_remapping(0);
- if (ret)
- printk(KERN_ERR
- "IOMMU: enable interrupt remapping failed\n");
- }
+ if (iommu_pass_through)
+ iommu_identity_mapping = 1;
+#ifdef CONFIG_DMAR_BROKEN_GFX_WA
+ else
+ iommu_identity_mapping = 2;
#endif
/*
- * If pass through is set and enabled, context entries of all pci
- * devices are intialized by pass through translation type.
+ * If pass through is not set or not enabled, setup context entries for
+ * identity mappings for rmrr, gfx, and isa and may fall back to static
+ * identity mapping if iommu_identity_mapping is set.
*/
- if (iommu_pass_through) {
- ret = init_context_pass_through();
+ if (iommu_identity_mapping) {
+ ret = iommu_prepare_static_identity_mapping(hw_pass_through);
if (ret) {
- printk(KERN_ERR "IOMMU: Pass through init failed.\n");
- iommu_pass_through = 0;
+ printk(KERN_CRIT "Failed to setup IOMMU pass-through\n");
+ goto error;
}
}
-
/*
- * If pass through is not set or not enabled, setup context entries for
- * identity mappings for rmrr, gfx, and isa.
+ * For each rmrr
+ * for each dev attached to rmrr
+ * do
+ * locate drhd for dev, alloc domain for dev
+ * allocate free domain
+ * allocate page table entries for rmrr
+ * if context not allocated for bus
+ * allocate and init context
+ * set present in root table for this bus
+ * init context with domain, translation etc
+ * endfor
+ * endfor
*/
- if (!iommu_pass_through) {
- /*
- * For each rmrr
- * for each dev attached to rmrr
- * do
- * locate drhd for dev, alloc domain for dev
- * allocate free domain
- * allocate page table entries for rmrr
- * if context not allocated for bus
- * allocate and init context
- * set present in root table for this bus
- * init context with domain, translation etc
- * endfor
- * endfor
- */
- for_each_rmrr_units(rmrr) {
- for (i = 0; i < rmrr->devices_cnt; i++) {
- pdev = rmrr->devices[i];
- /*
- * some BIOS lists non-exist devices in DMAR
- * table.
- */
- if (!pdev)
- continue;
- ret = iommu_prepare_rmrr_dev(rmrr, pdev);
- if (ret)
- printk(KERN_ERR
- "IOMMU: mapping reserved region failed\n");
- }
+ printk(KERN_INFO "IOMMU: Setting RMRR:\n");
+ for_each_rmrr_units(rmrr) {
+ for (i = 0; i < rmrr->devices_cnt; i++) {
+ pdev = rmrr->devices[i];
+ /*
+ * some BIOS lists non-exist devices in DMAR
+ * table.
+ */
+ if (!pdev)
+ continue;
+ ret = iommu_prepare_rmrr_dev(rmrr, pdev);
+ if (ret)
+ printk(KERN_ERR
+ "IOMMU: mapping reserved region failed\n");
}
-
- iommu_prepare_gfx_mapping();
-
- iommu_prepare_isa();
}
+ iommu_prepare_isa();
+
/*
* for each drhd
* enable fault log
@@ -2170,11 +2418,12 @@ static int __init init_dmars(void)
iommu->flush.flush_context(iommu, 0, 0, 0, DMA_CCMD_GLOBAL_INVL);
iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH);
- iommu_disable_protect_mem_regions(iommu);
ret = iommu_enable_translation(iommu);
if (ret)
goto error;
+
+ iommu_disable_protect_mem_regions(iommu);
}
return 0;
@@ -2189,58 +2438,39 @@ error:
return ret;
}
-static inline u64 aligned_size(u64 host_addr, size_t size)
-{
- u64 addr;
- addr = (host_addr & (~PAGE_MASK)) + size;
- return PAGE_ALIGN(addr);
-}
-
-struct iova *
-iommu_alloc_iova(struct dmar_domain *domain, size_t size, u64 end)
-{
- struct iova *piova;
-
- /* Make sure it's in range */
- end = min_t(u64, DOMAIN_MAX_ADDR(domain->gaw), end);
- if (!size || (IOVA_START_ADDR + size > end))
- return NULL;
-
- piova = alloc_iova(&domain->iovad,
- size >> PAGE_SHIFT, IOVA_PFN(end), 1);
- return piova;
-}
-
-static struct iova *
-__intel_alloc_iova(struct device *dev, struct dmar_domain *domain,
- size_t size, u64 dma_mask)
+/* This takes a number of _MM_ pages, not VTD pages */
+static struct iova *intel_alloc_iova(struct device *dev,
+ struct dmar_domain *domain,
+ unsigned long nrpages, uint64_t dma_mask)
{
struct pci_dev *pdev = to_pci_dev(dev);
struct iova *iova = NULL;
- if (dma_mask <= DMA_32BIT_MASK || dmar_forcedac)
- iova = iommu_alloc_iova(domain, size, dma_mask);
- else {
+ /* Restrict dma_mask to the width that the iommu can handle */
+ dma_mask = min_t(uint64_t, DOMAIN_MAX_ADDR(domain->gaw), dma_mask);
+
+ if (!dmar_forcedac && dma_mask > DMA_BIT_MASK(32)) {
/*
* First try to allocate an io virtual address in
- * DMA_32BIT_MASK and if that fails then try allocating
+ * DMA_BIT_MASK(32) and if that fails then try allocating
* from higher range
*/
- iova = iommu_alloc_iova(domain, size, DMA_32BIT_MASK);
- if (!iova)
- iova = iommu_alloc_iova(domain, size, dma_mask);
- }
-
- if (!iova) {
- printk(KERN_ERR"Allocating iova for %s failed", pci_name(pdev));
+ iova = alloc_iova(&domain->iovad, nrpages,
+ IOVA_PFN(DMA_BIT_MASK(32)), 1);
+ if (iova)
+ return iova;
+ }
+ iova = alloc_iova(&domain->iovad, nrpages, IOVA_PFN(dma_mask), 1);
+ if (unlikely(!iova)) {
+ printk(KERN_ERR "Allocating %ld-page iova for %s failed",
+ nrpages, pci_name(pdev));
return NULL;
}
return iova;
}
-static struct dmar_domain *
-get_valid_domain_for_dev(struct pci_dev *pdev)
+static struct dmar_domain *__get_valid_domain_for_dev(struct pci_dev *pdev)
{
struct dmar_domain *domain;
int ret;
@@ -2268,6 +2498,75 @@ get_valid_domain_for_dev(struct pci_dev *pdev)
return domain;
}
+static inline struct dmar_domain *get_valid_domain_for_dev(struct pci_dev *dev)
+{
+ struct device_domain_info *info;
+
+ /* No lock here, assumes no domain exit in normal case */
+ info = dev->iommu;
+ if (likely(info))
+ return info->domain;
+
+ return __get_valid_domain_for_dev(dev);
+}
+
+static int iommu_dummy(struct pci_dev *pdev)
+{
+ return pdev->iommu == DUMMY_DEVICE_DOMAIN_INFO;
+}
+
+/* Check if the pdev needs to go through non-identity map and unmap process.*/
+static int iommu_no_mapping(struct device *dev)
+{
+ struct pci_dev *pdev;
+ int found;
+
+ if (unlikely(dev->bus != &pci_bus_type))
+ return 1;
+
+ pdev = to_pci_dev(dev);
+ if (iommu_dummy(pdev))
+ return 1;
+
+ if (!iommu_identity_mapping)
+ return 0;
+
+ found = identity_mapping(pdev);
+ if (found) {
+ if (iommu_should_identity_map(pdev, 0))
+ return 1;
+ else {
+ /*
+ * 32 bit DMA is removed from si_domain and fall back
+ * to non-identity mapping.
+ */
+ domain_remove_one_dev_info(si_domain, pdev);
+ printk(KERN_INFO "32bit %s uses non-identity mapping\n",
+ pci_name(pdev));
+ return 0;
+ }
+ } else {
+ /*
+ * In case of a detached 64 bit DMA device from vm, the device
+ * is put into si_domain for identity mapping.
+ */
+ if (iommu_should_identity_map(pdev, 0)) {
+ int ret;
+ ret = domain_add_dev_info(si_domain, pdev,
+ hw_pass_through ?
+ CONTEXT_TT_PASS_THROUGH :
+ CONTEXT_TT_MULTI_LEVEL);
+ if (!ret) {
+ printk(KERN_INFO "64bit %s uses identity mapping\n",
+ pci_name(pdev));
+ return 1;
+ }
+ }
+ }
+
+ return 0;
+}
+
static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
size_t size, int dir, u64 dma_mask)
{
@@ -2278,9 +2577,11 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
int prot = 0;
int ret;
struct intel_iommu *iommu;
+ unsigned long paddr_pfn = paddr >> PAGE_SHIFT;
BUG_ON(dir == DMA_NONE);
- if (pdev->iommu == DUMMY_DEVICE_DOMAIN_INFO)
+
+ if (iommu_no_mapping(hwdev))
return paddr;
domain = get_valid_domain_for_dev(pdev);
@@ -2288,14 +2589,13 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
return 0;
iommu = domain_get_iommu(domain);
- size = aligned_size((u64)paddr, size);
+ size = aligned_nrpages(paddr, size);
- iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
+ iova = intel_alloc_iova(hwdev, domain, dma_to_mm_pfn(size),
+ pdev->dma_mask);
if (!iova)
goto error;
- start_paddr = (phys_addr_t)iova->pfn_lo << PAGE_SHIFT;
-
/*
* Check if DMAR supports zero-length reads on write only
* mappings..
@@ -2311,19 +2611,20 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
* might have two guest_addr mapping to the same host paddr, but this
* is not a big problem
*/
- ret = domain_page_mapping(domain, start_paddr,
- ((u64)paddr) & PAGE_MASK, size, prot);
+ ret = domain_pfn_mapping(domain, mm_to_dma_pfn(iova->pfn_lo),
+ mm_to_dma_pfn(paddr_pfn), size, prot);
if (ret)
goto error;
- /* it's a non-present to present mapping nly flush if caching mode */
+ /* it's a non-present to present mapping. Only flush if caching mode */
if (cap_caching_mode(iommu->cap))
- iommu_flush_iotlb_psi(iommu, 0, start_paddr,
- size >> VTD_PAGE_SHIFT);
+ iommu_flush_iotlb_psi(iommu, 0, mm_to_dma_pfn(iova->pfn_lo), size);
else
iommu_flush_write_buffer(iommu);
- return start_paddr + ((u64)paddr & (~PAGE_MASK));
+ start_paddr = (phys_addr_t)iova->pfn_lo << PAGE_SHIFT;
+ start_paddr += paddr & ~PAGE_MASK;
+ return start_paddr;
error:
if (iova)
@@ -2358,8 +2659,13 @@ static void flush_unmaps(void)
iommu->flush.flush_iotlb(iommu, 0, 0, 0,
DMA_TLB_GLOBAL_FLUSH);
for (j = 0; j < deferred_flush[i].next; j++) {
- __free_iova(&deferred_flush[i].domain[j]->iovad,
- deferred_flush[i].iova[j]);
+ unsigned long mask;
+ struct iova *iova = deferred_flush[i].iova[j];
+
+ mask = ilog2(mm_to_dma_pfn(iova->pfn_hi - iova->pfn_lo + 1));
+ iommu_flush_dev_iotlb(deferred_flush[i].domain[j],
+ (uint64_t)iova->pfn_lo << PAGE_SHIFT, mask);
+ __free_iova(&deferred_flush[i].domain[j]->iovad, iova);
}
deferred_flush[i].next = 0;
}
@@ -2402,39 +2708,48 @@ static void add_unmap(struct dmar_domain *dom, struct iova *iova)
spin_unlock_irqrestore(&async_umap_flush_lock, flags);
}
+static int warn_once_unmap_single;
static void intel_unmap_single(struct device *dev, dma_addr_t dev_addr, size_t size,
int dir)
{
struct pci_dev *pdev = to_pci_dev(dev);
struct dmar_domain *domain;
- unsigned long start_addr;
+ unsigned long start_pfn, last_pfn;
struct iova *iova;
struct intel_iommu *iommu;
- if (pdev->iommu == DUMMY_DEVICE_DOMAIN_INFO)
+ if (iommu_no_mapping(dev))
return;
+
domain = find_domain(pdev);
BUG_ON(!domain);
iommu = domain_get_iommu(domain);
iova = find_iova(&domain->iovad, IOVA_PFN(dev_addr));
- if (!iova)
+ if (!iova) {
+ if(!warn_once_unmap_single)
+ printk(KERN_WARNING "IOMMU: "
+ "Driver unmaps unmatched page at PFN %llx\n",
+ (unsigned long long)dev_addr);
return;
+ }
- start_addr = iova->pfn_lo << PAGE_SHIFT;
- size = aligned_size((u64)dev_addr, size);
+ start_pfn = mm_to_dma_pfn(iova->pfn_lo);
+ last_pfn = mm_to_dma_pfn(iova->pfn_hi + 1) - 1;
- pr_debug("Device %s unmapping: %lx@%llx\n",
- pci_name(pdev), size, (unsigned long long)start_addr);
+ pr_debug("Device %s unmapping: pfn %lx-%lx\n",
+ pci_name(pdev), start_pfn, last_pfn);
/* clear the whole page */
- dma_pte_clear_range(domain, start_addr, start_addr + size);
+ dma_pte_clear_range(domain, start_pfn, last_pfn);
+
/* free page tables */
- dma_pte_free_pagetable(domain, start_addr, start_addr + size);
+ dma_pte_free_pagetable(domain, start_pfn, last_pfn);
+
if (intel_iommu_strict) {
- iommu_flush_iotlb_psi(iommu, domain->id, start_addr,
- size >> VTD_PAGE_SHIFT);
+ iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
+ last_pfn - start_pfn + 1);
/* free iova */
__free_iova(&domain->iovad, iova);
} else {
@@ -2447,7 +2762,7 @@ static void intel_unmap_single(struct device *dev, dma_addr_t dev_addr, size_t s
}
static void *intel_alloc_coherent(struct device *hwdev, size_t size,
- dma_addr_t *dma_handle, gfp_t flags)
+ dma_addr_t *dma_handle, gfp_t flags)
{
void *vaddr;
int order;
@@ -2482,25 +2797,17 @@ static void intel_free_coherent(struct device *hwdev, size_t size, void *vaddr,
free_pages((unsigned long)vaddr, order);
}
-static inline struct page *sg_page(struct scatterlist *sg)
-{
- return sg->page;
-}
-
+static int warn_once_unmap_sg;
static void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
int nelems, int dir)
{
- int i;
struct pci_dev *pdev = to_pci_dev(hwdev);
struct dmar_domain *domain;
- unsigned long start_addr;
+ unsigned long start_pfn, last_pfn;
struct iova *iova;
- size_t size = 0;
- phys_addr_t addr;
- struct scatterlist *sg;
struct intel_iommu *iommu;
- if (pdev->iommu == DUMMY_DEVICE_DOMAIN_INFO)
+ if (iommu_no_mapping(hwdev))
return;
domain = find_domain(pdev);
@@ -2509,25 +2816,35 @@ static void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
iommu = domain_get_iommu(domain);
iova = find_iova(&domain->iovad, IOVA_PFN(sglist[0].dma_address));
- if (!iova)
+ if (!iova) {
+ if(!warn_once_unmap_sg)
+ printk(KERN_WARNING "IOMMU: "
+ "Driver unmaps unmatched sglist at PFN %llx\n",
+ (unsigned long long)sglist[0].dma_address);
return;
- for_each_sg(sglist, sg, nelems, i) {
- addr = page_to_phys(sg_page(sg)) + sg->offset;
- size += aligned_size((u64)addr, sg->length);
}
- start_addr = iova->pfn_lo << PAGE_SHIFT;
+ start_pfn = mm_to_dma_pfn(iova->pfn_lo);
+ last_pfn = mm_to_dma_pfn(iova->pfn_hi + 1) - 1;
/* clear the whole page */
- dma_pte_clear_range(domain, start_addr, start_addr + size);
- /* free page tables */
- dma_pte_free_pagetable(domain, start_addr, start_addr + size);
+ dma_pte_clear_range(domain, start_pfn, last_pfn);
- iommu_flush_iotlb_psi(iommu, domain->id, start_addr,
- size >> VTD_PAGE_SHIFT);
+ /* free page tables */
+ dma_pte_free_pagetable(domain, start_pfn, last_pfn);
- /* free iova */
- __free_iova(&domain->iovad, iova);
+ if (intel_iommu_strict) {
+ iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
+ last_pfn - start_pfn + 1);
+ /* free iova */
+ __free_iova(&domain->iovad, iova);
+ } else {
+ add_unmap(domain, iova);
+ /*
+ * queue up the release of the unmap to save the 1/6th of the
+ * cpu used up by the iotlb flush operation...
+ */
+ }
}
static int intel_nontranslate_map_sg(struct device *hddev,
@@ -2547,21 +2864,20 @@ static int intel_nontranslate_map_sg(struct device *hddev,
static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems,
int dir)
{
- phys_addr_t addr;
int i;
struct pci_dev *pdev = to_pci_dev(hwdev);
struct dmar_domain *domain;
size_t size = 0;
int prot = 0;
- size_t offset = 0;
+ size_t offset_pfn = 0;
struct iova *iova = NULL;
int ret;
struct scatterlist *sg;
- unsigned long start_addr;
+ unsigned long start_vpfn;
struct intel_iommu *iommu;
BUG_ON(dir == DMA_NONE);
- if (pdev->iommu == DUMMY_DEVICE_DOMAIN_INFO)
+ if (iommu_no_mapping(hwdev))
return intel_nontranslate_map_sg(hwdev, sglist, nelems, dir);
domain = get_valid_domain_for_dev(pdev);
@@ -2570,12 +2886,11 @@ static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int ne
iommu = domain_get_iommu(domain);
- for_each_sg(sglist, sg, nelems, i) {
- addr = page_to_phys(sg_page(sg)) + sg->offset;
- size += aligned_size((u64)addr, sg->length);
- }
+ for_each_sg(sglist, sg, nelems, i)
+ size += aligned_nrpages(sg->offset, sg->length);
- iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
+ iova = intel_alloc_iova(hwdev, domain, dma_to_mm_pfn(size),
+ pdev->dma_mask);
if (!iova) {
sglist->dma_length = 0;
return 0;
@@ -2591,35 +2906,24 @@ static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int ne
if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL)
prot |= DMA_PTE_WRITE;
- start_addr = iova->pfn_lo << PAGE_SHIFT;
- offset = 0;
- for_each_sg(sglist, sg, nelems, i) {
- addr = page_to_phys(sg_page(sg)) + sg->offset;
- size = aligned_size((u64)addr, sg->length);
- ret = domain_page_mapping(domain, start_addr + offset,
- ((u64)addr) & PAGE_MASK,
- size, prot);
- if (ret) {
- /* clear the page */
- dma_pte_clear_range(domain, start_addr,
- start_addr + offset);
- /* free page tables */
- dma_pte_free_pagetable(domain, start_addr,
- start_addr + offset);
- /* free iova */
- __free_iova(&domain->iovad, iova);
- return 0;
- }
- sg->dma_address = start_addr + offset +
- ((u64)addr & (~PAGE_MASK));
- sg->dma_length = sg->length;
- offset += size;
+ start_vpfn = mm_to_dma_pfn(iova->pfn_lo);
+
+ ret = domain_sg_mapping(domain, start_vpfn, sglist, size, prot);
+ if (unlikely(ret)) {
+ /* clear the page */
+ dma_pte_clear_range(domain, start_vpfn,
+ start_vpfn + size - 1);
+ /* free page tables */
+ dma_pte_free_pagetable(domain, start_vpfn,
+ start_vpfn + size - 1);
+ /* free iova */
+ __free_iova(&domain->iovad, iova);
+ return 0;
}
/* it's a non-present to present mapping. Only flush if caching mode */
if (cap_caching_mode(iommu->cap))
- iommu_flush_iotlb_psi(iommu, 0, start_addr,
- offset >> VTD_PAGE_SHIFT);
+ iommu_flush_iotlb_psi(iommu, 0, start_vpfn, offset_pfn);
else
iommu_flush_write_buffer(iommu);
@@ -2777,8 +3081,8 @@ static int init_iommu_hw(void)
DMA_CCMD_GLOBAL_INVL);
iommu->flush.flush_iotlb(iommu, 0, 0, 0,
DMA_TLB_GLOBAL_FLUSH);
- iommu_disable_protect_mem_regions(iommu);
iommu_enable_translation(iommu);
+ iommu_disable_protect_mem_regions(iommu);
}
return 0;
@@ -2844,7 +3148,7 @@ static int iommu_resume(struct sys_device *dev)
unsigned long flag;
if (init_iommu_hw()) {
- WARN(1, "IOMMU setup failed, DMAR can not resume!\n");
+ printk(KERN_WARNING "IOMMU setup failed, DMAR can not resume!\n");
return -EIO;
}
@@ -2916,7 +3220,7 @@ int __init intel_iommu_init(void)
* Check the need for DMA-remapping initialization now.
* Above initialization will also be used by Interrupt-remapping.
*/
- if (no_iommu || (swiotlb && !iommu_pass_through) || dmar_disabled)
+ if (no_iommu || swiotlb || dmar_disabled)
return -ENODEV;
iommu_init_mempool();
@@ -2936,14 +3240,7 @@ int __init intel_iommu_init(void)
init_timer(&unmap_timer);
force_iommu = 1;
-
- if (!iommu_pass_through) {
- printk(KERN_INFO
- "Multi-level page-table translation for DMAR.\n");
- dma_ops = &intel_dma_ops;
- } else
- printk(KERN_INFO
- "DMAR: Pass through translation for DMAR.\n");
+ dma_ops = &intel_dma_ops;
init_iommu_sysfs();
@@ -2952,31 +3249,6 @@ int __init intel_iommu_init(void)
return 0;
}
-static int vm_domain_add_dev_info(struct dmar_domain *domain,
- struct pci_dev *pdev)
-{
- struct device_domain_info *info;
- unsigned long flags;
-
- info = alloc_devinfo_mem();
- if (!info)
- return -ENOMEM;
-
- info->segment = pci_domain_nr(pdev->bus);
- info->bus = pdev->bus->number;
- info->devfn = pdev->devfn;
- info->dev = pdev;
- info->domain = domain;
-
- spin_lock_irqsave(&device_domain_lock, flags);
- list_add(&info->link, &domain->devices);
- list_add(&info->global, &device_domain_list);
- pdev->iommu = info;
- spin_unlock_irqrestore(&device_domain_lock, flags);
-
- return 0;
-}
-
static void iommu_detach_dependent_devices(struct intel_iommu *iommu,
struct pci_dev *pdev)
{
@@ -3004,7 +3276,7 @@ static void iommu_detach_dependent_devices(struct intel_iommu *iommu,
}
}
-static void vm_domain_remove_one_dev_info(struct dmar_domain *domain,
+static void domain_remove_one_dev_info(struct dmar_domain *domain,
struct pci_dev *pdev)
{
struct device_domain_info *info;
@@ -3030,6 +3302,7 @@ static void vm_domain_remove_one_dev_info(struct dmar_domain *domain,
info->dev->iommu = NULL;
spin_unlock_irqrestore(&device_domain_lock, flags);
+ iommu_disable_dev_iotlb(info);
iommu_detach_dev(iommu, info->bus, info->devfn);
iommu_detach_dependent_devices(iommu, pdev);
free_devinfo_mem(info);
@@ -3080,6 +3353,7 @@ static void vm_domain_remove_all_dev_info(struct dmar_domain *domain)
spin_unlock_irqrestore(&device_domain_lock, flags1);
+ iommu_disable_dev_iotlb(info);
iommu = device_to_iommu(info->segment, info->bus, info->devfn);
iommu_detach_dev(iommu, info->bus, info->devfn);
iommu_detach_dependent_devices(iommu, info->dev);
@@ -3135,12 +3409,11 @@ static struct dmar_domain *iommu_alloc_vm_domain(void)
return domain;
}
-static int vm_domain_init(struct dmar_domain *domain, int guest_width)
+static int md_domain_init(struct dmar_domain *domain, int guest_width)
{
int adjust_width;
init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
- spin_lock_init(&domain->mapping_lock);
spin_lock_init(&domain->iommu_lock);
domain_reserve_special_ranges(domain);
@@ -3154,6 +3427,7 @@ static int vm_domain_init(struct dmar_domain *domain, int guest_width)
domain->iommu_count = 0;
domain->iommu_coherency = 0;
+ domain->iommu_snooping = 0;
domain->max_addr = 0;
/* always allocate the top pgd */
@@ -3194,8 +3468,6 @@ static void iommu_free_vm_domain(struct dmar_domain *domain)
static void vm_domain_exit(struct dmar_domain *domain)
{
- u64 end;
-
/* Domain 0 is reserved, so dont process it */
if (!domain)
return;
@@ -3203,14 +3475,12 @@ static void vm_domain_exit(struct dmar_domain *domain)
vm_domain_remove_all_dev_info(domain);
/* destroy iovas */
put_iova_domain(&domain->iovad);
- end = DOMAIN_MAX_ADDR(domain->gaw);
- end = end & (~VTD_PAGE_MASK);
/* clear ptes */
- dma_pte_clear_range(domain, 0, end);
+ dma_pte_clear_range(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
/* free page tables */
- dma_pte_free_pagetable(domain, 0, end);
+ dma_pte_free_pagetable(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
iommu_free_vm_domain(domain);
free_domain_mem(domain);
@@ -3226,7 +3496,7 @@ static int intel_iommu_domain_init(struct iommu_domain *domain)
"intel_iommu_domain_init: dmar_domain == NULL\n");
return -ENOMEM;
}
- if (vm_domain_init(dmar_domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) {
+ if (md_domain_init(dmar_domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) {
printk(KERN_ERR
"intel_iommu_domain_init() failed\n");
vm_domain_exit(dmar_domain);
@@ -3253,7 +3523,6 @@ static int intel_iommu_attach_device(struct iommu_domain *domain,
struct intel_iommu *iommu;
int addr_width;
u64 end;
- int ret;
/* normally pdev is not mapped */
if (unlikely(domain_context_mapped(pdev))) {
@@ -3261,8 +3530,9 @@ static int intel_iommu_attach_device(struct iommu_domain *domain,
old_domain = find_domain(pdev);
if (old_domain) {
- if (dmar_domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE)
- vm_domain_remove_one_dev_info(old_domain, pdev);
+ if (dmar_domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE ||
+ dmar_domain->flags & DOMAIN_FLAG_STATIC_IDENTITY)
+ domain_remove_one_dev_info(old_domain, pdev);
else
domain_remove_dev_info(old_domain);
}
@@ -3284,12 +3554,7 @@ static int intel_iommu_attach_device(struct iommu_domain *domain,
return -EFAULT;
}
- ret = domain_context_mapping(dmar_domain, pdev, CONTEXT_TT_MULTI_LEVEL);
- if (ret)
- return ret;
-
- ret = vm_domain_add_dev_info(dmar_domain, pdev);
- return ret;
+ return domain_add_dev_info(dmar_domain, pdev, CONTEXT_TT_MULTI_LEVEL);
}
static void intel_iommu_detach_device(struct iommu_domain *domain,
@@ -3298,7 +3563,7 @@ static void intel_iommu_detach_device(struct iommu_domain *domain,
struct dmar_domain *dmar_domain = domain->priv;
struct pci_dev *pdev = to_pci_dev(dev);
- vm_domain_remove_one_dev_info(dmar_domain, pdev);
+ domain_remove_one_dev_info(dmar_domain, pdev);
}
static int intel_iommu_map_range(struct iommu_domain *domain,
@@ -3318,7 +3583,7 @@ static int intel_iommu_map_range(struct iommu_domain *domain,
if ((iommu_prot & IOMMU_CACHE) && dmar_domain->iommu_snooping)
prot |= DMA_PTE_SNP;
- max_addr = (iova & VTD_PAGE_MASK) + VTD_PAGE_ALIGN(size);
+ max_addr = iova + size;
if (dmar_domain->max_addr < max_addr) {
int min_agaw;
u64 end;
@@ -3336,8 +3601,11 @@ static int intel_iommu_map_range(struct iommu_domain *domain,
}
dmar_domain->max_addr = max_addr;
}
-
- ret = domain_page_mapping(dmar_domain, iova, hpa, size, prot);
+ /* Round up size to next multiple of PAGE_SIZE, if it and
+ the low bits of hpa would take us onto the next page */
+ size = aligned_nrpages(hpa, size);
+ ret = domain_pfn_mapping(dmar_domain, iova >> VTD_PAGE_SHIFT,
+ hpa >> VTD_PAGE_SHIFT, size, prot);
return ret;
}
@@ -3345,15 +3613,15 @@ static void intel_iommu_unmap_range(struct iommu_domain *domain,
unsigned long iova, size_t size)
{
struct dmar_domain *dmar_domain = domain->priv;
- dma_addr_t base;
- /* The address might not be aligned */
- base = iova & VTD_PAGE_MASK;
- size = VTD_PAGE_ALIGN(size);
- dma_pte_clear_range(dmar_domain, base, base + size);
+ if (!size)
+ return;
+
+ dma_pte_clear_range(dmar_domain, iova >> VTD_PAGE_SHIFT,
+ (iova + size - 1) >> VTD_PAGE_SHIFT);
- if (dmar_domain->max_addr == base + size)
- dmar_domain->max_addr = base;
+ if (dmar_domain->max_addr == iova + size)
+ dmar_domain->max_addr = iova;
}
static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
@@ -3363,7 +3631,7 @@ static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
struct dma_pte *pte;
u64 phys = 0;
- pte = addr_to_dma_pte(dmar_domain, iova);
+ pte = pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT);
if (pte)
phys = dma_pte_addr(pte);
@@ -3403,3 +3671,4 @@ static void __devinit quirk_iommu_rwbf(struct pci_dev *dev)
}
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2a40, quirk_iommu_rwbf);
+
diff --git a/drivers/pci/iova.c b/drivers/pci/iova.c
index 9c11867..4a55838 100644
--- a/drivers/pci/iova.c
+++ b/drivers/pci/iova.c
@@ -22,7 +22,6 @@
void
init_iova_domain(struct iova_domain *iovad, unsigned long pfn_32bit)
{
- spin_lock_init(&iovad->iova_alloc_lock);
spin_lock_init(&iovad->iova_rbtree_lock);
iovad->rbroot = RB_ROOT;
iovad->cached32_node = NULL;
@@ -205,7 +204,6 @@ alloc_iova(struct iova_domain *iovad, unsigned long size,
unsigned long limit_pfn,
bool size_aligned)
{
- unsigned long flags;
struct iova *new_iova;
int ret;
@@ -219,11 +217,9 @@ alloc_iova(struct iova_domain *iovad, unsigned long size,
if (size_aligned)
size = __roundup_pow_of_two(size);
- spin_lock_irqsave(&iovad->iova_alloc_lock, flags);
ret = __alloc_and_insert_iova_range(iovad, size, limit_pfn,
new_iova, size_aligned);
- spin_unlock_irqrestore(&iovad->iova_alloc_lock, flags);
if (ret) {
free_iova_mem(new_iova);
return NULL;
@@ -381,8 +377,7 @@ reserve_iova(struct iova_domain *iovad,
struct iova *iova;
unsigned int overlap = 0;
- spin_lock_irqsave(&iovad->iova_alloc_lock, flags);
- spin_lock(&iovad->iova_rbtree_lock);
+ spin_lock_irqsave(&iovad->iova_rbtree_lock, flags);
for (node = rb_first(&iovad->rbroot); node; node = rb_next(node)) {
if (__is_range_overlap(node, pfn_lo, pfn_hi)) {
iova = container_of(node, struct iova, node);
@@ -402,8 +397,7 @@ reserve_iova(struct iova_domain *iovad,
iova = __insert_new_range(iovad, pfn_lo, pfn_hi);
finish:
- spin_unlock(&iovad->iova_rbtree_lock);
- spin_unlock_irqrestore(&iovad->iova_alloc_lock, flags);
+ spin_unlock_irqrestore(&iovad->iova_rbtree_lock, flags);
return iova;
}
@@ -420,8 +414,7 @@ copy_reserved_iova(struct iova_domain *from, struct iova_domain *to)
unsigned long flags;
struct rb_node *node;
- spin_lock_irqsave(&from->iova_alloc_lock, flags);
- spin_lock(&from->iova_rbtree_lock);
+ spin_lock_irqsave(&from->iova_rbtree_lock, flags);
for (node = rb_first(&from->rbroot); node; node = rb_next(node)) {
struct iova *iova = container_of(node, struct iova, node);
struct iova *new_iova;
@@ -430,6 +423,5 @@ copy_reserved_iova(struct iova_domain *from, struct iova_domain *to)
printk(KERN_ERR "Reserve iova range %lx@%lx failed\n",
iova->pfn_lo, iova->pfn_lo);
}
- spin_unlock(&from->iova_rbtree_lock);
- spin_unlock_irqrestore(&from->iova_alloc_lock, flags);
+ spin_unlock_irqrestore(&from->iova_rbtree_lock, flags);
}
diff --git a/include/linux/dma_remapping.h b/include/linux/dma_remapping.h
index 5619f85..194f569 100644
--- a/include/linux/dma_remapping.h
+++ b/include/linux/dma_remapping.h
@@ -39,4 +39,12 @@ static inline int iommu_calculate_max_sagaw(struct intel_iommu *iommu)
extern int dmar_disabled;
+#ifdef CONFIG_DMA_API_DEBUG
+extern void debug_dma_dump_mappings(struct device *dev);
+#else
+static inline void debug_dma_dump_mappings(struct device *dev)
+{
+}
+#endif /* CONFIG_DMA_API_DEBUG */
+
#endif
diff --git a/include/linux/iova.h b/include/linux/iova.h
index 228f6c9..76a0759 100644
--- a/include/linux/iova.h
+++ b/include/linux/iova.h
@@ -28,7 +28,6 @@ struct iova {
/* holds all the iova translations for a domain */
struct iova_domain {
- spinlock_t iova_alloc_lock;/* Lock to protect iova allocation */
spinlock_t iova_rbtree_lock; /* Lock to protect update of rbtree */
struct rb_root rbroot; /* iova domain rbtree root */
struct rb_node *cached32_node; /* Save last alloced node */
distrib
>
Scientific%20Linux
>
5x
>
x86_64
>
by-pkgid
>
fc11cd6e1c513a17304da94a5390f3cd
>
files
>
2795
