arm64/kvm: Enable eager hugepage splitting if HDBSS is available

[PATCH] arm64/kvm: Enable eager hugepage splitting if HDBSS is available

Posted by Leonardo Bras 1 week, 1 day ago

FEAT_HDBSS speeds up guest memory dirty tracking by avoiding a page fault
and saving the entry in a tracking structure.

That may be a problem when we have guest memory backed by hugepages or
transparent huge pages, as it's not possible to do on-demand hugepage
splitting, relying only on eager hugepage splitting.

So, at stage2 initialization, enable eager hugepage splitting with
chunk = PAGE_SIZE if the system supports HDBSS.

Signed-off-by: Leonardo Bras <leo.bras@arm.com>
---
 arch/arm64/kvm/mmu.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index 070a01e53fcb..bdfa72b7c073 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -993,22 +993,26 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
 
 	mmu->last_vcpu_ran = alloc_percpu(typeof(*mmu->last_vcpu_ran));
 	if (!mmu->last_vcpu_ran) {
 		err = -ENOMEM;
 		goto out_destroy_pgtable;
 	}
 
 	for_each_possible_cpu(cpu)
 		*per_cpu_ptr(mmu->last_vcpu_ran, cpu) = -1;
 
-	 /* The eager page splitting is disabled by default */
-	mmu->split_page_chunk_size = KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT;
+	 /* The eager page splitting is disabled by default if system has no HDBSS */
+	if (system_supports_hacdbs())
+		mmu->split_page_chunk_size = PAGE_SIZE;
+	else
+		mmu->split_page_chunk_size = KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT;
+
 	mmu->split_page_cache.gfp_zero = __GFP_ZERO;
 
 	mmu->pgd_phys = __pa(pgt->pgd);
 
 	if (kvm_is_nested_s2_mmu(kvm, mmu))
 		kvm_init_nested_s2_mmu(mmu);
 
 	return 0;
 
 out_destroy_pgtable:
-- 
2.53.0

Re: [PATCH] arm64/kvm: Enable eager hugepage splitting if HDBSS is available

Posted by Tian Zheng 6 days, 14 hours ago

On 3/26/2026 2:20 AM, Leonardo Bras wrote:
> FEAT_HDBSS speeds up guest memory dirty tracking by avoiding a page fault
> and saving the entry in a tracking structure.
>
> That may be a problem when we have guest memory backed by hugepages or
> transparent huge pages, as it's not possible to do on-demand hugepage
> splitting, relying only on eager hugepage splitting.
>
> So, at stage2 initialization, enable eager hugepage splitting with
> chunk = PAGE_SIZE if the system supports HDBSS.
>
> Signed-off-by: Leonardo Bras <leo.bras@arm.com>
> ---
>   arch/arm64/kvm/mmu.c | 8 ++++++--
>   1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index 070a01e53fcb..bdfa72b7c073 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -993,22 +993,26 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
>   
>   	mmu->last_vcpu_ran = alloc_percpu(typeof(*mmu->last_vcpu_ran));
>   	if (!mmu->last_vcpu_ran) {
>   		err = -ENOMEM;
>   		goto out_destroy_pgtable;
>   	}
>   
>   	for_each_possible_cpu(cpu)
>   		*per_cpu_ptr(mmu->last_vcpu_ran, cpu) = -1;
>   
> -	 /* The eager page splitting is disabled by default */
> -	mmu->split_page_chunk_size = KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT;
> +	 /* The eager page splitting is disabled by default if system has no HDBSS */
> +	if (system_supports_hacdbs())
> +		mmu->split_page_chunk_size = PAGE_SIZE;
> +	else
> +		mmu->split_page_chunk_size = KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT;
> +
>   	mmu->split_page_cache.gfp_zero = __GFP_ZERO;
>   
>   	mmu->pgd_phys = __pa(pgt->pgd);
>   
>   	if (kvm_is_nested_s2_mmu(kvm, mmu))
>   		kvm_init_nested_s2_mmu(mmu);
>   
>   	return 0;
>   
>   out_destroy_pgtable:


Thanks again for sending this patch. I'll integrate it into the next 
version and run some tests.

Re: [PATCH] arm64/kvm: Enable eager hugepage splitting if HDBSS is available

Posted by Leonardo Bras 6 days, 7 hours ago

On Fri, Mar 27, 2026 at 03:40:30PM +0800, Tian Zheng wrote:
> 
> On 3/26/2026 2:20 AM, Leonardo Bras wrote:
> > FEAT_HDBSS speeds up guest memory dirty tracking by avoiding a page fault
> > and saving the entry in a tracking structure.
> > 
> > That may be a problem when we have guest memory backed by hugepages or
> > transparent huge pages, as it's not possible to do on-demand hugepage
> > splitting, relying only on eager hugepage splitting.
> > 
> > So, at stage2 initialization, enable eager hugepage splitting with
> > chunk = PAGE_SIZE if the system supports HDBSS.
> > 
> > Signed-off-by: Leonardo Bras <leo.bras@arm.com>
> > ---
> >   arch/arm64/kvm/mmu.c | 8 ++++++--
> >   1 file changed, 6 insertions(+), 2 deletions(-)
> > 
> > diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> > index 070a01e53fcb..bdfa72b7c073 100644
> > --- a/arch/arm64/kvm/mmu.c
> > +++ b/arch/arm64/kvm/mmu.c
> > @@ -993,22 +993,26 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
> >   	mmu->last_vcpu_ran = alloc_percpu(typeof(*mmu->last_vcpu_ran));
> >   	if (!mmu->last_vcpu_ran) {
> >   		err = -ENOMEM;
> >   		goto out_destroy_pgtable;
> >   	}
> >   	for_each_possible_cpu(cpu)
> >   		*per_cpu_ptr(mmu->last_vcpu_ran, cpu) = -1;
> > -	 /* The eager page splitting is disabled by default */
> > -	mmu->split_page_chunk_size = KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT;
> > +	 /* The eager page splitting is disabled by default if system has no HDBSS */
> > +	if (system_supports_hacdbs())
> > +		mmu->split_page_chunk_size = PAGE_SIZE;
> > +	else
> > +		mmu->split_page_chunk_size = KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT;
> > +
> >   	mmu->split_page_cache.gfp_zero = __GFP_ZERO;
> >   	mmu->pgd_phys = __pa(pgt->pgd);
> >   	if (kvm_is_nested_s2_mmu(kvm, mmu))
> >   		kvm_init_nested_s2_mmu(mmu);
> >   	return 0;
> >   out_destroy_pgtable:
> 
> 
> Thanks again for sending this patch. I'll integrate it into the next version
> and run some tests.
> 
> 

Awesome, thanks!
Leo