[PATCH] kvm mmu: optimizations when tdp is in use

[PATCH] kvm mmu: optimizations when tdp is in use

[Gui Jianfeng]

In case of using tdp, checking write protected page isn't needed and
quadrant also no need to be calculated.

Signed-off-by: Gui Jianfeng <guijianfeng@cn.fujitsu.com>
---
 arch/x86/kvm/mmu.c |    7 +++++--
 1 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 0bb9f17..ce4bbd3 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -495,10 +495,13 @@ static int mapping_level(struct kvm_vcpu *vcpu, gfn_t large_gfn)
 	max_level = kvm_x86_ops->get_lpage_level() < host_level ?
 		kvm_x86_ops->get_lpage_level() : host_level;
 
+	if (tdp_enabled)
+		goto done;
+
 	for (level = PT_DIRECTORY_LEVEL; level <= max_level; ++level)
 		if (has_wrprotected_page(vcpu->kvm, large_gfn, level))
 			break;
-
+done:
 	return level - 1;
 }
 
@@ -1346,7 +1349,7 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
 	if (role.direct)
 		role.cr4_pae = 0;
 	role.access = access;
-	if (vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
+	if (!tdp_enabled && vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
 		quadrant = gaddr >> (PAGE_SHIFT + (PT64_PT_BITS * level));
 		quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
 		role.quadrant = quadrant;
-- 
1.6.5.2

Re: [PATCH] kvm mmu: optimizations when tdp is in use

[Avi Kivity]

On 05/27/2010 11:06 AM, Gui Jianfeng wrote:
> In case of using tdp, checking write protected page isn't needed and
> quadrant also no need to be calculated.
>
> Signed-off-by: Gui Jianfeng<guijianfeng@cn.fujitsu.com>
> ---
>   arch/x86/kvm/mmu.c |    7 +++++--
>   1 files changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
> index 0bb9f17..ce4bbd3 100644
> --- a/arch/x86/kvm/mmu.c
> +++ b/arch/x86/kvm/mmu.c
> @@ -495,10 +495,13 @@ static int mapping_level(struct kvm_vcpu *vcpu, gfn_t large_gfn)
>   	max_level = kvm_x86_ops->get_lpage_level()<  host_level ?
>   		kvm_x86_ops->get_lpage_level() : host_level;
>
> +	if (tdp_enabled)
> +		goto done;
> +
>   	for (level = PT_DIRECTORY_LEVEL; level<= max_level; ++level)
>   		if (has_wrprotected_page(vcpu->kvm, large_gfn, level))
>   			break;
> -
> +done:
>   	return level - 1;
>   }
>
>    

We also use ->write_count to prevent mapping the end of a 
non-large-page-aligned memslot with a large spte.

Undocumented in mmu.txt, I'll post a patch.

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

Re: [PATCH] kvm mmu: optimizations when tdp is in use

[Marcelo Tosatti]

On Thu, May 27, 2010 at 04:06:34PM +0800, Gui Jianfeng wrote:
> In case of using tdp, checking write protected page isn't needed and
> quadrant also no need to be calculated.
> 
> Signed-off-by: Gui Jianfeng <guijianfeng@cn.fujitsu.com>
> ---
>  arch/x86/kvm/mmu.c |    7 +++++--
>  1 files changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
> index 0bb9f17..ce4bbd3 100644
> --- a/arch/x86/kvm/mmu.c
> +++ b/arch/x86/kvm/mmu.c
> @@ -495,10 +495,13 @@ static int mapping_level(struct kvm_vcpu *vcpu, gfn_t large_gfn)
>  	max_level = kvm_x86_ops->get_lpage_level() < host_level ?
>  		kvm_x86_ops->get_lpage_level() : host_level;
>  
> +	if (tdp_enabled)
> +		goto done;
> +

This is wrong. write_count is initialized for alignment purposes, not 
only write protected pages. See __kvm_set_memory_region in
virt/kvm/kvm_main.c.

>  	for (level = PT_DIRECTORY_LEVEL; level <= max_level; ++level)
>  		if (has_wrprotected_page(vcpu->kvm, large_gfn, level))
>  			break;
> -
> +done:
>  	return level - 1;
>  }
>  
> @@ -1346,7 +1349,7 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
>  	if (role.direct)
>  		role.cr4_pae = 0;
>  	role.access = access;
> -	if (vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
> +	if (!tdp_enabled && vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
>  		quadrant = gaddr >> (PAGE_SHIFT + (PT64_PT_BITS * level));
>  		quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
>  		role.quadrant = quadrant;
> -- 
> 1.6.5.2
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] kvm mmu: optimizations when tdp is in use

[Gui Jianfeng]

Marcelo Tosatti wrote:
> On Thu, May 27, 2010 at 04:06:34PM +0800, Gui Jianfeng wrote:
>> In case of using tdp, checking write protected page isn't needed and
>> quadrant also no need to be calculated.
>>
>> Signed-off-by: Gui Jianfeng <guijianfeng@cn.fujitsu.com>
>> ---
>>  arch/x86/kvm/mmu.c |    7 +++++--
>>  1 files changed, 5 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
>> index 0bb9f17..ce4bbd3 100644
>> --- a/arch/x86/kvm/mmu.c
>> +++ b/arch/x86/kvm/mmu.c
>> @@ -495,10 +495,13 @@ static int mapping_level(struct kvm_vcpu *vcpu, gfn_t large_gfn)
>>  	max_level = kvm_x86_ops->get_lpage_level() < host_level ?
>>  		kvm_x86_ops->get_lpage_level() : host_level;
>>  
>> +	if (tdp_enabled)
>> +		goto done;
>> +
> 
> This is wrong. write_count is initialized for alignment purposes, not 
> only write protected pages. See __kvm_set_memory_region in
> virt/kvm/kvm_main.c.

thanks, avi also pointed this out.

Gui

> 
>>  	for (level = PT_DIRECTORY_LEVEL; level <= max_level; ++level)
>>  		if (has_wrprotected_page(vcpu->kvm, large_gfn, level))
>>  			break;
>> -
>> +done:
>>  	return level - 1;
>>  }
>>  
>> @@ -1346,7 +1349,7 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
>>  	if (role.direct)
>>  		role.cr4_pae = 0;
>>  	role.access = access;
>> -	if (vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
>> +	if (!tdp_enabled && vcpu->arch.mmu.root_level <= PT32_ROOT_LEVEL) {
>>  		quadrant = gaddr >> (PAGE_SHIFT + (PT64_PT_BITS * level));
>>  		quadrant &= (1 << ((PT32_PT_BITS - PT64_PT_BITS) * level)) - 1;
>>  		role.quadrant = quadrant;
>> -- 
>> 1.6.5.2
>> --
>> To unsubscribe from this list: send the line "unsubscribe kvm" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
> 

-- 
Regards
Gui Jianfeng