2010-04-22 06:17:14

by Xiao Guangrong

[permalink] [raw]
Subject: [PATCH 10/10] KVM MMU: optimize sync/update unsync-page

invlpg only need update unsync page, sp->unsync and sp->unsync_children
can help us to find it

Now, a gfn may have many shadow pages, when one sp need be synced, we
write protect sp->gfn and sync this sp but we keep other shadow pages
asynchronous

So, while gfn happen page fault, let it not touch unsync page, the unsync
page only updated at invlpg/flush TLB time

Signed-off-by: Xiao Guangrong <[email protected]>
---
arch/x86/kvm/mmu.c | 10 ++++++----
1 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index f092e71..5bdcc17 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -2299,10 +2299,11 @@ static void paging_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
for_each_shadow_entry(vcpu, gva, iterator) {
level = iterator.level;
sptep = iterator.sptep;
+ sp = page_header(__pa(sptep));

if (is_last_spte(*sptep, level)) {
-
- sp = page_header(__pa(sptep));
+ if (!sp->unsync)
+ break;

if (!sp->role.cr4_pae)
offset = sp->role.quadrant << PT64_LEVEL_BITS;;
@@ -2320,7 +2321,7 @@ static void paging_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
break;
}

- if (!is_shadow_present_pte(*sptep))
+ if (!is_shadow_present_pte(*sptep) || !sp->unsync_children)
break;
}

@@ -2744,7 +2745,8 @@ void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,

restart:
hlist_for_each_entry_safe(sp, node, n, bucket, hash_link) {
- if (sp->gfn != gfn || sp->role.direct || sp->role.invalid)
+ if (sp->gfn != gfn || sp->role.direct || sp->role.invalid ||
+ sp->unsync)
continue;
pte_size = sp->role.cr4_pae ? 8 : 4;
misaligned = (offset ^ (offset + bytes - 1)) & ~(pte_size - 1);
--
1.6.1.2