Re: [PATCH 2/3] KVM: MMU: don not retry #PF for nonpaging guest

[Gleb Natapov <gleb@redhat.com>]

On Fri, Nov 05, 2010 at 01:39:18PM +0800, Xiao Guangrong wrote:
> On 11/04/2010 06:35 PM, Gleb Natapov wrote:
> > On Thu, Nov 04, 2010 at 06:32:42PM +0800, Xiao Guangrong wrote:
> >> nonpaing guest's 'direct_map' is also true, retry #PF for those
> >> guests is useless, so use 'tdp_enabled' instead
> >>
> > nonpaging guest will not attempt async pf. 
> 
> Ah, my mistake, but why we can not attempt async pf for nonpaging guest?
> 
We can, but we do not expect to run many nonpaging guests I guess :)

> > And by checking tdp_enabled
> > here instead of direct_map we will screw nested ntp.
> > 
> 
> It looks like something broken: apfs can generated in L2 guest (nested ntp guest)
> and be retried in L1 guest.
> 
Why is this a problem? apf will be generate on direct map even when L2
guest is running so it should be OK to prefault it into direct map on
completion.

> Below patch fix it and let nonpaging guest support async pf. I'll post it properly
> if you like. :-)
> 
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 7f20f2c..606978e 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -600,6 +600,7 @@ struct kvm_x86_ops {
>  struct kvm_arch_async_pf {
>  	u32 token;
>  	gfn_t gfn;
> +	bool softmmu;
>  };
>  
>  extern struct kvm_x86_ops *kvm_x86_ops;
> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
> index f3fad4f..48ca312 100644
> --- a/arch/x86/kvm/mmu.c
> +++ b/arch/x86/kvm/mmu.c
> @@ -2286,7 +2286,10 @@ static int kvm_handle_bad_page(struct kvm *kvm, gfn_t gfn, pfn_t pfn)
>  	return 1;
>  }
>  
> -static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, int write, gfn_t gfn)
> +static bool try_async_pf(struct kvm_vcpu *vcpu, bool no_apf, gfn_t gfn,
> +			 gva_t gva, pfn_t *pfn, bool write, bool *writable);
> +
> +static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, int write, gfn_t gfn, bool no_apf)
>  {
>  	int r;
>  	int level;
> @@ -2307,7 +2310,9 @@ static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, int write, gfn_t gfn)
>  
>  	mmu_seq = vcpu->kvm->mmu_notifier_seq;
>  	smp_rmb();
> -	pfn = gfn_to_pfn_prot(vcpu->kvm, gfn, write, &map_writable);
> +
> +	if (try_async_pf(vcpu, no_apf, gfn, v, &pfn, write, &map_writable))
> +		return 0;
>  
>  	/* mmio */
>  	if (is_error_pfn(pfn))
> @@ -2594,7 +2599,7 @@ static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
>  	gfn = gva >> PAGE_SHIFT;
>  
>  	return nonpaging_map(vcpu, gva & PAGE_MASK,
> -			     error_code & PFERR_WRITE_MASK, gfn);
> +			     error_code & PFERR_WRITE_MASK, gfn, no_apf);
>  }
>  
>  static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn)
> @@ -2602,6 +2607,7 @@ static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn)
>  	struct kvm_arch_async_pf arch;
>  	arch.token = (vcpu->arch.apf.id++ << 12) | vcpu->vcpu_id;
>  	arch.gfn = gfn;
> +	arch.softmmu = mmu_is_softmmu(vcpu);
>  
>  	return kvm_setup_async_pf(vcpu, gva, gfn, &arch);
>  }
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 2044302..d826d78 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -6172,9 +6172,10 @@ EXPORT_SYMBOL_GPL(kvm_set_rflags);
>  
>  void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu, struct kvm_async_pf *work)
>  {
> +	bool softmmu = mmu_is_softmmu(vcpu);
>  	int r;
>  
> -	if (!vcpu->arch.mmu.direct_map || is_error_page(work->page))
> +	if (softmmu || work->arch.softmmu || is_error_page(work->page))
>  		return;
>  
>  	r = kvm_mmu_reload(vcpu);
> diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h
> index 2cea414..48796c7 100644
> --- a/arch/x86/kvm/x86.h
> +++ b/arch/x86/kvm/x86.h
> @@ -55,6 +55,11 @@ static inline bool mmu_is_nested(struct kvm_vcpu *vcpu)
>  	return vcpu->arch.walk_mmu == &vcpu->arch.nested_mmu;
>  }
>  
> +static inline bool mmu_is_softmmu(struct kvm_vcpu *vcpu)
> +{
> +	return !tdp_enabled || mmu_is_nested(vcpu);
> +}
> +
>  static inline int is_pae(struct kvm_vcpu *vcpu)
>  {
>  	return kvm_read_cr4_bits(vcpu, X86_CR4_PAE);

--
			Gleb.