Re: [PATCH 11/12] KVM: arm64: nv: Implement HW access flag management in stage-2 SW PTW

[Marc Zyngier <maz@kernel.org>]

On Wed, 12 Nov 2025 18:34:05 +0000,
Oliver Upton <oupton@kernel.org> wrote:
> 
> Give the stage-2 walk similar treatment to stage-1: update the access
> flag during the table walk and do so for any walk context.
> 
> Signed-off-by: Oliver Upton <oupton@kernel.org>
> ---
>  arch/arm64/kvm/mmu.c    |  3 +++
>  arch/arm64/kvm/nested.c | 46 ++++++++++++++++++++++++++++++++++-------
>  2 files changed, 41 insertions(+), 8 deletions(-)
> 
> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
> index 96f1786c72fe..b23d3dc8865e 100644
> --- a/arch/arm64/kvm/mmu.c
> +++ b/arch/arm64/kvm/mmu.c
> @@ -2012,6 +2012,9 @@ int kvm_handle_guest_abort(struct kvm_vcpu *vcpu)
>  		u32 esr;
>  
>  		ret = kvm_walk_nested_s2(vcpu, fault_ipa, &nested_trans);
> +		if (ret == -EAGAIN)
> +			return 1;
> +
>  		if (ret) {
>  			esr = kvm_s2_trans_esr(&nested_trans);
>  			kvm_inject_s2_fault(vcpu, esr);
> diff --git a/arch/arm64/kvm/nested.c b/arch/arm64/kvm/nested.c
> index b2fb3d7c9c19..7293769dacf5 100644
> --- a/arch/arm64/kvm/nested.c
> +++ b/arch/arm64/kvm/nested.c
> @@ -124,13 +124,14 @@ int kvm_vcpu_init_nested(struct kvm_vcpu *vcpu)
>  }
>  
>  struct s2_walk_info {
> -	void	     *data;
> -	u64	     baddr;
> -	unsigned int max_oa_bits;
> -	unsigned int pgshift;
> -	unsigned int sl;
> -	unsigned int t0sz;
> -	bool	     be;
> +	void		*data;
> +	u64		baddr;
> +	unsigned int	max_oa_bits;
> +	unsigned int	pgshift;
> +	unsigned int	sl;
> +	unsigned int	t0sz;
> +	bool		be;
> +	bool		ha;
>  };
>  
>  static u32 compute_fsc(int level, u32 fsc)
> @@ -220,6 +221,22 @@ static int read_guest_s2_desc(phys_addr_t pa, u64 *desc, struct s2_walk_info *wi
>  	return 0;
>  }
>  
> +static int swap_guest_s2_desc(phys_addr_t pa, u64 old, u64 new,
> +			      struct s2_walk_info *wi)
> +{
> +	struct kvm_vcpu *vcpu = wi->data;
> +
> +	if (wi->be) {
> +		old = cpu_to_be64(old);
> +		new = cpu_to_be64(new);
> +	} else {
> +		old = cpu_to_be64(old);
> +		new = cpu_to_be64(new);
> +	}
> +
> +	return __kvm_at_swap_desc(vcpu->kvm, pa, old, new);
> +}
> +
>  /*
>   * This is essentially a C-version of the pseudo code from the ARM ARM
>   * AArch64.TranslationTableWalk  function.  I strongly recommend looking at
> @@ -233,7 +250,7 @@ static int walk_nested_s2_pgd(phys_addr_t ipa,
>  	int first_block_level, level, stride, input_size, base_lower_bound;
>  	phys_addr_t base_addr;
>  	unsigned int addr_top, addr_bottom;
> -	u64 desc;  /* page table entry */
> +	u64 desc, new_desc;  /* page table entry */
>  	int ret;
>  	phys_addr_t paddr;
>  
> @@ -326,6 +343,17 @@ static int walk_nested_s2_pgd(phys_addr_t ipa,
>  		return 1;
>  	}
>  
> +	if (wi->ha)
> +		new_desc |= KVM_PTE_LEAF_ATTR_LO_S2_AF;

Same issue here (new_desc not initialised).

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.