[PATCH 1/2] KVM: MMU: remove mmu_seq verification on pte write path

[PATCH 1/2] KVM: MMU: remove mmu_seq verification on pte write path

[Xiao Guangrong]

The mmu_seq verification can be removed since we get the pfn in the
protection of mmu_lock

Signed-off-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
---
 arch/x86/include/asm/kvm_host.h |    2 +-
 arch/x86/kvm/mmu.c              |   16 +++++-----------
 arch/x86/kvm/paging_tmpl.h      |    4 +---
 3 files changed, 7 insertions(+), 15 deletions(-)

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 35f81b1..ea9fea3 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -258,7 +258,7 @@ struct kvm_mmu {
 			 struct kvm_mmu_page *sp);
 	void (*invlpg)(struct kvm_vcpu *vcpu, gva_t gva);
 	void (*update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
-			u64 *spte, const void *pte, unsigned long mmu_seq);
+			   u64 *spte, const void *pte);
 	hpa_t root_hpa;
 	int root_level;
 	int shadow_root_level;
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 22fae75..2841805 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -1206,7 +1206,7 @@ static void nonpaging_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
 
 static void nonpaging_update_pte(struct kvm_vcpu *vcpu,
 				 struct kvm_mmu_page *sp, u64 *spte,
-				 const void *pte, unsigned long mmu_seq)
+				 const void *pte)
 {
 	WARN_ON(1);
 }
@@ -3163,9 +3163,8 @@ static void mmu_pte_write_zap_pte(struct kvm_vcpu *vcpu,
 }
 
 static void mmu_pte_write_new_pte(struct kvm_vcpu *vcpu,
-				  struct kvm_mmu_page *sp,
-				  u64 *spte,
-				  const void *new, unsigned long mmu_seq)
+				  struct kvm_mmu_page *sp, u64 *spte,
+				  const void *new)
 {
 	if (sp->role.level != PT_PAGE_TABLE_LEVEL) {
 		++vcpu->kvm->stat.mmu_pde_zapped;
@@ -3173,7 +3172,7 @@ static void mmu_pte_write_new_pte(struct kvm_vcpu *vcpu,
         }
 
 	++vcpu->kvm->stat.mmu_pte_updated;
-	vcpu->arch.mmu.update_pte(vcpu, sp, spte, new, mmu_seq);
+	vcpu->arch.mmu.update_pte(vcpu, sp, spte, new);
 }
 
 static bool need_remote_flush(u64 old, u64 new)
@@ -3229,7 +3228,6 @@ void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
 	struct kvm_mmu_page *sp;
 	struct hlist_node *node;
 	LIST_HEAD(invalid_list);
-	unsigned long mmu_seq;
 	u64 entry, gentry, *spte;
 	unsigned pte_size, page_offset, misaligned, quadrant, offset;
 	int level, npte, invlpg_counter, r, flooded = 0;
@@ -3271,9 +3269,6 @@ void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
 		break;
 	}
 
-	mmu_seq = vcpu->kvm->mmu_notifier_seq;
-	smp_rmb();
-
 	spin_lock(&vcpu->kvm->mmu_lock);
 	if (atomic_read(&vcpu->kvm->arch.invlpg_counter) != invlpg_counter)
 		gentry = 0;
@@ -3345,8 +3340,7 @@ void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
 			if (gentry &&
 			      !((sp->role.word ^ vcpu->arch.mmu.base_role.word)
 			      & mask.word))
-				mmu_pte_write_new_pte(vcpu, sp, spte, &gentry,
-						      mmu_seq);
+				mmu_pte_write_new_pte(vcpu, sp, spte, &gentry);
 			if (!remote_flush && need_remote_flush(entry, *spte))
 				remote_flush = true;
 			++spte;
diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
index c639779..74f8567 100644
--- a/arch/x86/kvm/paging_tmpl.h
+++ b/arch/x86/kvm/paging_tmpl.h
@@ -325,7 +325,7 @@ no_present:
 }
 
 static void FNAME(update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
-			      u64 *spte, const void *pte, unsigned long mmu_seq)
+			      u64 *spte, const void *pte)
 {
 	pt_element_t gpte;
 	unsigned pte_access;
@@ -342,8 +342,6 @@ static void FNAME(update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
 		kvm_release_pfn_clean(pfn);
 		return;
 	}
-	if (mmu_notifier_retry(vcpu, mmu_seq))
-		return;
 
 	/*
 	 * we call mmu_set_spte() with host_writable = true because that
-- 
1.7.4

[PATCH 2/2] KVM: MMU: optimize pte write path if don't have protected sp

[Xiao Guangrong]

Simplely return from kvm_mmu_pte_write path if no shadow page is
write-protected, than we can avoid to walk all shadow pages and hold
mmu-lock

Signed-off-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
---
 arch/x86/kvm/mmu.c |    8 ++++++++
 1 files changed, 8 insertions(+), 0 deletions(-)

diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 2841805..ccac6c4 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -3219,6 +3219,11 @@ static void kvm_mmu_access_page(struct kvm_vcpu *vcpu, gfn_t gfn)
 		set_bit(PT_ACCESSED_SHIFT, (unsigned long *)spte);
 }
 
+static bool need_track_pte_changed(struct kvm_vcpu *vcpu)
+{
+	return !tdp_enabled || mmu_is_nested(vcpu);
+}
+
 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
 		       const u8 *new, int bytes,
 		       bool guest_initiated)
@@ -3233,6 +3238,9 @@ void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
 	int level, npte, invlpg_counter, r, flooded = 0;
 	bool remote_flush, local_flush, zap_page;
 
+	if (!need_track_pte_changed(vcpu))
+		return;
+
 	zap_page = remote_flush = local_flush = false;
 	offset = offset_in_page(gpa);
 
-- 
1.7.4

Re: [PATCH 2/2] KVM: MMU: optimize pte write path if don't have protected sp

[Avi Kivity]

On 03/28/2011 04:30 AM, Xiao Guangrong wrote:
> Simplely return from kvm_mmu_pte_write path if no shadow page is
> write-protected, than we can avoid to walk all shadow pages and hold
> mmu-lock
>
> Signed-off-by: Xiao Guangrong<xiaoguangrong@cn.fujitsu.com>
> ---
>   arch/x86/kvm/mmu.c |    8 ++++++++
>   1 files changed, 8 insertions(+), 0 deletions(-)
>
> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
> index 2841805..ccac6c4 100644
> --- a/arch/x86/kvm/mmu.c
> +++ b/arch/x86/kvm/mmu.c
> @@ -3219,6 +3219,11 @@ static void kvm_mmu_access_page(struct kvm_vcpu *vcpu, gfn_t gfn)
>   		set_bit(PT_ACCESSED_SHIFT, (unsigned long *)spte);
>   }
>
> +static bool need_track_pte_changed(struct kvm_vcpu *vcpu)
> +{
> +	return !tdp_enabled || mmu_is_nested(vcpu);
> +}
> +

What happens if this vcpu is nested but another vcpu is not?

Or if we have some indirect pages, but currently we're in non-nested mode?

Suggest instead to keep track of the number of indirect pages across the 
guest.  If it's zero, we can exit early.

>   void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
>   		       const u8 *new, int bytes,
>   		       bool guest_initiated)
> @@ -3233,6 +3238,9 @@ void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
>   	int level, npte, invlpg_counter, r, flooded = 0;
>   	bool remote_flush, local_flush, zap_page;
>
> +	if (!need_track_pte_changed(vcpu))
> +		return;
> +
>   	zap_page = remote_flush = local_flush = false;
>   	offset = offset_in_page(gpa);
>


-- 
error compiling committee.c: too many arguments to function

Re: [PATCH 1/2] KVM: MMU: remove mmu_seq verification on pte write path

[Avi Kivity]

On 03/28/2011 04:29 AM, Xiao Guangrong wrote:
> The mmu_seq verification can be removed since we get the pfn in the
> protection of mmu_lock

Applied, thanks.

-- 
error compiling committee.c: too many arguments to function