[PATCH 2/7] kvm mmu: infrastructure changes for multiple huge page support
From: Joerg Roedel
Date: Fri Mar 27 2009 - 10:34:26 EST
This patch includes most of the necessary changes to the KVM SoftMMU for
supporting more than one huge page size. The changes in this patch
include:
* introduce 'enum kvm_page_size' which is used to represent the page
size used
* change boolean is_largepage_backed() function to backing_size()
which returns the largest page size KVM can use to map a gfn
* change the other largepage flags to 'enum kvm_page_size'
Signed-off-by: Joerg Roedel <joerg.roedel@xxxxxxx>
---
arch/x86/include/asm/kvm_host.h | 17 ++++--
arch/x86/kvm/mmu.c | 111 +++++++++++++++++++++------------------
arch/x86/kvm/paging_tmpl.h | 22 ++++----
3 files changed, 83 insertions(+), 67 deletions(-)
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 8351c4d..f268f99 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -52,11 +52,13 @@
#define UNMAPPED_GVA (~(gpa_t)0)
/* shadow tables are PAE even on non-PAE hosts */
-#define KVM_HPAGE_SHIFT 21
-#define KVM_HPAGE_SIZE (1UL << KVM_HPAGE_SHIFT)
-#define KVM_HPAGE_MASK (~(KVM_HPAGE_SIZE - 1))
+#define KVM_2M_PAGE_SHIFT 21
+#define KVM_2M_PAGE_SIZE (1UL << KVM_2M_PAGE_SHIFT)
+#define KVM_2M_PAGE_MASK (~(KVM_2M_PAGE_SIZE - 1))
-#define KVM_PAGES_PER_HPAGE (KVM_HPAGE_SIZE / PAGE_SIZE)
+#define KVM_PAGES_PER_2M_PAGE (KVM_2M_PAGE_SIZE / PAGE_SIZE)
+
+#define KVM_PAGES_PER_HPAGE KVM_PAGES_PER_2M_PAGE
#define DE_VECTOR 0
#define DB_VECTOR 1
@@ -263,6 +265,11 @@ struct kvm_mmu {
u64 *pae_root;
};
+enum kvm_page_size {
+ KVM_PAGE_SIZE_4k = (1 << 12),
+ KVM_PAGE_SIZE_2M = (1 << 21),
+};
+
struct kvm_vcpu_arch {
u64 host_tsc;
int interrupt_window_open;
@@ -310,7 +317,7 @@ struct kvm_vcpu_arch {
struct {
gfn_t gfn; /* presumed gfn during guest pte update */
pfn_t pfn; /* pfn corresponding to that gfn */
- int largepage;
+ enum kvm_page_size page_size;
unsigned long mmu_seq;
} update_pte;
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index b625ed4..3a57c17 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -385,8 +385,8 @@ static int *slot_largepage_idx(gfn_t gfn, struct kvm_memory_slot *slot)
{
unsigned long idx;
- idx = (gfn / KVM_PAGES_PER_HPAGE) -
- (slot->base_gfn / KVM_PAGES_PER_HPAGE);
+ idx = (gfn / KVM_PAGES_PER_2M_PAGE) -
+ (slot->base_gfn / KVM_PAGES_PER_2M_PAGE);
return &slot->lpage_info[idx].write_count;
}
@@ -426,11 +426,11 @@ static int has_wrprotected_page(struct kvm *kvm, gfn_t gfn)
return 1;
}
-static int host_largepage_backed(struct kvm *kvm, gfn_t gfn)
+static enum kvm_page_size host_page_size(struct kvm *kvm, gfn_t gfn)
{
struct vm_area_struct *vma;
- unsigned long addr;
- int ret = 0;
+ unsigned long addr, size;
+ enum kvm_page_size ret = KVM_PAGE_SIZE_4k;
addr = gfn_to_hva(kvm, gfn);
if (kvm_is_error_hva(addr))
@@ -438,28 +438,31 @@ static int host_largepage_backed(struct kvm *kvm, gfn_t gfn)
down_read(¤t->mm->mmap_sem);
vma = find_vma(current->mm, addr);
- if (vma && is_vm_hugetlb_page(vma))
- ret = 1;
+ if (vma) {
+ size = vma_kernel_pagesize(vma);
+ if (size >= KVM_PAGE_SIZE_2M)
+ ret = KVM_PAGE_SIZE_2M;
+ }
up_read(¤t->mm->mmap_sem);
return ret;
}
-static int is_largepage_backed(struct kvm_vcpu *vcpu, gfn_t large_gfn)
+static enum kvm_page_size backing_size(struct kvm_vcpu *vcpu, gfn_t gfn)
{
struct kvm_memory_slot *slot;
- if (has_wrprotected_page(vcpu->kvm, large_gfn))
- return 0;
+ if (has_wrprotected_page(vcpu->kvm, gfn))
+ return KVM_PAGE_SIZE_4k;
- if (!host_largepage_backed(vcpu->kvm, large_gfn))
- return 0;
+ if (host_page_size(vcpu->kvm, gfn) < KVM_PAGE_SIZE_2M)
+ return KVM_PAGE_SIZE_4k;
- slot = gfn_to_memslot(vcpu->kvm, large_gfn);
+ slot = gfn_to_memslot(vcpu->kvm, gfn);
if (slot && slot->dirty_bitmap)
- return 0;
+ return KVM_PAGE_SIZE_4k;
- return 1;
+ return KVM_PAGE_SIZE_2M;
}
/*
@@ -467,17 +470,18 @@ static int is_largepage_backed(struct kvm_vcpu *vcpu, gfn_t large_gfn)
* Note: gfn must be unaliased before this function get called
*/
-static unsigned long *gfn_to_rmap(struct kvm *kvm, gfn_t gfn, int lpage)
+static unsigned long *gfn_to_rmap(struct kvm *kvm, gfn_t gfn,
+ enum kvm_page_size psize)
{
struct kvm_memory_slot *slot;
unsigned long idx;
slot = gfn_to_memslot(kvm, gfn);
- if (!lpage)
+ if (psize == KVM_PAGE_SIZE_4k)
return &slot->rmap[gfn - slot->base_gfn];
- idx = (gfn / KVM_PAGES_PER_HPAGE) -
- (slot->base_gfn / KVM_PAGES_PER_HPAGE);
+ idx = (gfn / KVM_PAGES_PER_2M_PAGE) -
+ (slot->base_gfn / KVM_PAGES_PER_2M_PAGE);
return &slot->lpage_info[idx].rmap_pde;
}
@@ -491,7 +495,8 @@ static unsigned long *gfn_to_rmap(struct kvm *kvm, gfn_t gfn, int lpage)
* If rmapp bit zero is one, (then rmap & ~1) points to a struct kvm_rmap_desc
* containing more mappings.
*/
-static void rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn, int lpage)
+static void rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn,
+ enum kvm_page_size psize)
{
struct kvm_mmu_page *sp;
struct kvm_rmap_desc *desc;
@@ -503,7 +508,7 @@ static void rmap_add(struct kvm_vcpu *vcpu, u64 *spte, gfn_t gfn, int lpage)
gfn = unalias_gfn(vcpu->kvm, gfn);
sp = page_header(__pa(spte));
sp->gfns[spte - sp->spt] = gfn;
- rmapp = gfn_to_rmap(vcpu->kvm, gfn, lpage);
+ rmapp = gfn_to_rmap(vcpu->kvm, gfn, psize);
if (!*rmapp) {
rmap_printk("rmap_add: %p %llx 0->1\n", spte, *spte);
*rmapp = (unsigned long)spte;
@@ -559,6 +564,7 @@ static void rmap_remove(struct kvm *kvm, u64 *spte)
pfn_t pfn;
unsigned long *rmapp;
int i;
+ enum kvm_page_size psize;
if (!is_rmap_pte(*spte))
return;
@@ -570,7 +576,8 @@ static void rmap_remove(struct kvm *kvm, u64 *spte)
kvm_release_pfn_dirty(pfn);
else
kvm_release_pfn_clean(pfn);
- rmapp = gfn_to_rmap(kvm, sp->gfns[spte - sp->spt], is_large_pte(*spte));
+ psize = is_large_pte(*spte) ? KVM_PAGE_SIZE_2M : KVM_PAGE_SIZE_4k;
+ rmapp = gfn_to_rmap(kvm, sp->gfns[spte - sp->spt], psize);
if (!*rmapp) {
printk(KERN_ERR "rmap_remove: %p %llx 0->BUG\n", spte, *spte);
BUG();
@@ -636,7 +643,7 @@ static int rmap_write_protect(struct kvm *kvm, u64 gfn)
int write_protected = 0;
gfn = unalias_gfn(kvm, gfn);
- rmapp = gfn_to_rmap(kvm, gfn, 0);
+ rmapp = gfn_to_rmap(kvm, gfn, KVM_PAGE_SIZE_4k);
spte = rmap_next(kvm, rmapp, NULL);
while (spte) {
@@ -658,7 +665,7 @@ static int rmap_write_protect(struct kvm *kvm, u64 gfn)
}
/* check for huge page mappings */
- rmapp = gfn_to_rmap(kvm, gfn, 1);
+ rmapp = gfn_to_rmap(kvm, gfn, KVM_PAGE_SIZE_2M);
spte = rmap_next(kvm, rmapp, NULL);
while (spte) {
BUG_ON(!spte);
@@ -719,7 +726,7 @@ static int kvm_handle_hva(struct kvm *kvm, unsigned long hva,
retval |= handler(kvm,
&memslot->lpage_info[
gfn_offset /
- KVM_PAGES_PER_HPAGE].rmap_pde);
+ KVM_PAGES_PER_2M_PAGE].rmap_pde);
}
}
@@ -1676,7 +1683,7 @@ static int mmu_need_write_protect(struct kvm_vcpu *vcpu, gfn_t gfn,
static int set_spte(struct kvm_vcpu *vcpu, u64 *shadow_pte,
unsigned pte_access, int user_fault,
- int write_fault, int dirty, int largepage,
+ int write_fault, int dirty, enum kvm_page_size psize,
int global, gfn_t gfn, pfn_t pfn, bool speculative,
bool can_unsync)
{
@@ -1709,7 +1716,7 @@ static int set_spte(struct kvm_vcpu *vcpu, u64 *shadow_pte,
spte |= shadow_nx_mask;
if (pte_access & ACC_USER_MASK)
spte |= shadow_user_mask;
- if (largepage)
+ if (psize > KVM_PAGE_SIZE_4k)
spte |= PT_PAGE_SIZE_MASK;
if (mt_mask) {
if (!kvm_is_mmio_pfn(pfn)) {
@@ -1727,7 +1734,8 @@ static int set_spte(struct kvm_vcpu *vcpu, u64 *shadow_pte,
if ((pte_access & ACC_WRITE_MASK)
|| (write_fault && !is_write_protection(vcpu) && !user_fault)) {
- if (largepage && has_wrprotected_page(vcpu->kvm, gfn)) {
+ if (psize > KVM_PAGE_SIZE_4k &&
+ has_wrprotected_page(vcpu->kvm, gfn)) {
ret = 1;
spte = shadow_trap_nonpresent_pte;
goto set_pte;
@@ -1765,7 +1773,7 @@ set_pte:
static void mmu_set_spte(struct kvm_vcpu *vcpu, u64 *shadow_pte,
unsigned pt_access, unsigned pte_access,
int user_fault, int write_fault, int dirty,
- int *ptwrite, int largepage, int global,
+ int *ptwrite, enum kvm_page_size psize, int global,
gfn_t gfn, pfn_t pfn, bool speculative)
{
int was_rmapped = 0;
@@ -1781,7 +1789,7 @@ static void mmu_set_spte(struct kvm_vcpu *vcpu, u64 *shadow_pte,
* If we overwrite a PTE page pointer with a 2MB PMD, unlink
* the parent of the now unreachable PTE.
*/
- if (largepage && !is_large_pte(*shadow_pte)) {
+ if (psize > KVM_PAGE_SIZE_4k && !is_large_pte(*shadow_pte)) {
struct kvm_mmu_page *child;
u64 pte = *shadow_pte;
@@ -1795,7 +1803,7 @@ static void mmu_set_spte(struct kvm_vcpu *vcpu, u64 *shadow_pte,
was_rmapped = 1;
}
if (set_spte(vcpu, shadow_pte, pte_access, user_fault, write_fault,
- dirty, largepage, global, gfn, pfn, speculative, true)) {
+ dirty, psize, global, gfn, pfn, speculative, true)) {
if (write_fault)
*ptwrite = 1;
kvm_x86_ops->tlb_flush(vcpu);
@@ -1811,7 +1819,7 @@ static void mmu_set_spte(struct kvm_vcpu *vcpu, u64 *shadow_pte,
page_header_update_slot(vcpu->kvm, shadow_pte, gfn);
if (!was_rmapped) {
- rmap_add(vcpu, shadow_pte, gfn, largepage);
+ rmap_add(vcpu, shadow_pte, gfn, psize);
if (!is_rmap_pte(*shadow_pte))
kvm_release_pfn_clean(pfn);
} else {
@@ -1831,7 +1839,7 @@ static void nonpaging_new_cr3(struct kvm_vcpu *vcpu)
}
static int __direct_map(struct kvm_vcpu *vcpu, gpa_t v, int write,
- int largepage, gfn_t gfn, pfn_t pfn)
+ enum kvm_page_size psize, gfn_t gfn, pfn_t pfn)
{
struct kvm_shadow_walk_iterator iterator;
struct kvm_mmu_page *sp;
@@ -1840,10 +1848,11 @@ static int __direct_map(struct kvm_vcpu *vcpu, gpa_t v, int write,
for_each_shadow_entry(vcpu, (u64)gfn << PAGE_SHIFT, iterator) {
if (iterator.level == PT_PAGE_TABLE_LEVEL
- || (largepage && iterator.level == PT_DIRECTORY_LEVEL)) {
+ || (psize == KVM_PAGE_SIZE_2M &&
+ iterator.level == PT_DIRECTORY_LEVEL)) {
mmu_set_spte(vcpu, iterator.sptep, ACC_ALL, ACC_ALL,
0, write, 1, &pt_write,
- largepage, 0, gfn, pfn, false);
+ psize, 0, gfn, pfn, false);
++vcpu->stat.pf_fixed;
break;
}
@@ -1871,14 +1880,12 @@ static int __direct_map(struct kvm_vcpu *vcpu, gpa_t v, int write,
static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, int write, gfn_t gfn)
{
int r;
- int largepage = 0;
pfn_t pfn;
unsigned long mmu_seq;
+ enum kvm_page_size psize = backing_size(vcpu, gfn);
- if (is_largepage_backed(vcpu, gfn & ~(KVM_PAGES_PER_HPAGE-1))) {
- gfn &= ~(KVM_PAGES_PER_HPAGE-1);
- largepage = 1;
- }
+ if (psize == KVM_PAGE_SIZE_2M)
+ gfn &= ~(KVM_PAGES_PER_2M_PAGE-1);
mmu_seq = vcpu->kvm->mmu_notifier_seq;
smp_rmb();
@@ -1894,7 +1901,7 @@ static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, int write, gfn_t gfn)
if (mmu_notifier_retry(vcpu, mmu_seq))
goto out_unlock;
kvm_mmu_free_some_pages(vcpu);
- r = __direct_map(vcpu, v, write, largepage, gfn, pfn);
+ r = __direct_map(vcpu, v, write, psize, gfn, pfn);
spin_unlock(&vcpu->kvm->mmu_lock);
@@ -2067,9 +2074,9 @@ static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa,
{
pfn_t pfn;
int r;
- int largepage = 0;
gfn_t gfn = gpa >> PAGE_SHIFT;
unsigned long mmu_seq;
+ enum kvm_page_size psize;
ASSERT(vcpu);
ASSERT(VALID_PAGE(vcpu->arch.mmu.root_hpa));
@@ -2078,10 +2085,9 @@ static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa,
if (r)
return r;
- if (is_largepage_backed(vcpu, gfn & ~(KVM_PAGES_PER_HPAGE-1))) {
- gfn &= ~(KVM_PAGES_PER_HPAGE-1);
- largepage = 1;
- }
+ psize = backing_size(vcpu, gfn);
+ if (psize == KVM_PAGE_SIZE_2M)
+ gfn &= ~(KVM_PAGES_PER_2M_PAGE-1);
mmu_seq = vcpu->kvm->mmu_notifier_seq;
smp_rmb();
pfn = gfn_to_pfn(vcpu->kvm, gfn);
@@ -2094,7 +2100,7 @@ static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa,
goto out_unlock;
kvm_mmu_free_some_pages(vcpu);
r = __direct_map(vcpu, gpa, error_code & PFERR_WRITE_MASK,
- largepage, gfn, pfn);
+ psize, gfn, pfn);
spin_unlock(&vcpu->kvm->mmu_lock);
return r;
@@ -2333,7 +2339,7 @@ static void mmu_pte_write_new_pte(struct kvm_vcpu *vcpu,
const void *new)
{
if (sp->role.level != PT_PAGE_TABLE_LEVEL) {
- if (!vcpu->arch.update_pte.largepage ||
+ if (vcpu->arch.update_pte.page_size != KVM_PAGE_SIZE_2M ||
sp->role.glevels == PT32_ROOT_LEVEL) {
++vcpu->kvm->stat.mmu_pde_zapped;
return;
@@ -2383,7 +2389,7 @@ static void mmu_guess_page_from_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
u64 gpte = 0;
pfn_t pfn;
- vcpu->arch.update_pte.largepage = 0;
+ vcpu->arch.update_pte.page_size = KVM_PAGE_SIZE_4k;
if (bytes != 4 && bytes != 8)
return;
@@ -2412,9 +2418,10 @@ static void mmu_guess_page_from_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
return;
gfn = (gpte & PT64_BASE_ADDR_MASK) >> PAGE_SHIFT;
- if (is_large_pte(gpte) && is_largepage_backed(vcpu, gfn)) {
- gfn &= ~(KVM_PAGES_PER_HPAGE-1);
- vcpu->arch.update_pte.largepage = 1;
+ if (is_large_pte(gpte) &&
+ backing_size(vcpu, gfn) != KVM_PAGE_SIZE_4k) {
+ gfn &= ~(KVM_PAGES_PER_2M_PAGE-1);
+ vcpu->arch.update_pte.page_size = KVM_PAGE_SIZE_2M;
}
vcpu->arch.update_pte.mmu_seq = vcpu->kvm->mmu_notifier_seq;
smp_rmb();
diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
index 855eb71..9fbd049 100644
--- a/arch/x86/kvm/paging_tmpl.h
+++ b/arch/x86/kvm/paging_tmpl.h
@@ -241,7 +241,7 @@ static void FNAME(update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *page,
pt_element_t gpte;
unsigned pte_access;
pfn_t pfn;
- int largepage = vcpu->arch.update_pte.largepage;
+ enum kvm_page_size psize = vcpu->arch.update_pte.page_size;
gpte = *(const pt_element_t *)pte;
if (~gpte & (PT_PRESENT_MASK | PT_ACCESSED_MASK)) {
@@ -260,7 +260,7 @@ static void FNAME(update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *page,
return;
kvm_get_pfn(pfn);
mmu_set_spte(vcpu, spte, page->role.access, pte_access, 0, 0,
- gpte & PT_DIRTY_MASK, NULL, largepage,
+ gpte & PT_DIRTY_MASK, NULL, psize,
gpte & PT_GLOBAL_MASK, gpte_to_gfn(gpte),
pfn, true);
}
@@ -270,7 +270,8 @@ static void FNAME(update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *page,
*/
static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr,
struct guest_walker *gw,
- int user_fault, int write_fault, int largepage,
+ int user_fault, int write_fault,
+ enum kvm_page_size psize,
int *ptwrite, pfn_t pfn)
{
unsigned access = gw->pt_access;
@@ -290,12 +291,13 @@ static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr,
level = iterator.level;
sptep = iterator.sptep;
if (level == PT_PAGE_TABLE_LEVEL
- || (largepage && level == PT_DIRECTORY_LEVEL)) {
+ || (psize == KVM_PAGE_SIZE_2M &&
+ level == PT_DIRECTORY_LEVEL)) {
mmu_set_spte(vcpu, sptep, access,
gw->pte_access & access,
user_fault, write_fault,
gw->ptes[gw->level-1] & PT_DIRTY_MASK,
- ptwrite, largepage,
+ ptwrite, psize,
gw->ptes[gw->level-1] & PT_GLOBAL_MASK,
gw->gfn, pfn, false);
break;
@@ -368,7 +370,7 @@ static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gva_t addr,
int write_pt = 0;
int r;
pfn_t pfn;
- int largepage = 0;
+ enum kvm_page_size psize = KVM_PAGE_SIZE_4k;
unsigned long mmu_seq;
pgprintk("%s: addr %lx err %x\n", __func__, addr, error_code);
@@ -396,10 +398,10 @@ static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gva_t addr,
if (walker.level == PT_DIRECTORY_LEVEL) {
gfn_t large_gfn;
- large_gfn = walker.gfn & ~(KVM_PAGES_PER_HPAGE-1);
- if (is_largepage_backed(vcpu, large_gfn)) {
+ large_gfn = walker.gfn & ~(KVM_PAGES_PER_2M_PAGE-1);
+ if (backing_size(vcpu, large_gfn) != KVM_PAGE_SIZE_4k) {
walker.gfn = large_gfn;
- largepage = 1;
+ psize = KVM_PAGE_SIZE_2M;
}
}
mmu_seq = vcpu->kvm->mmu_notifier_seq;
@@ -418,7 +420,7 @@ static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gva_t addr,
goto out_unlock;
kvm_mmu_free_some_pages(vcpu);
shadow_pte = FNAME(fetch)(vcpu, addr, &walker, user_fault, write_fault,
- largepage, &write_pt, pfn);
+ psize, &write_pt, pfn);
pgprintk("%s: shadow pte %p %llx ptwrite %d\n", __func__,
shadow_pte, *shadow_pte, write_pt);
--
1.5.6.4
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/