Re: [PATCH v4 1/7] KVM: arm/arm64: Share common code in user_mem_abort()

[Marc Zyngier]

Hi Punit,

On 05/07/18 15:08, Punit Agrawal wrote:
> The code for operations such as marking the pfn as dirty, and
> dcache/icache maintenance during stage 2 fault handling is duplicated
> between normal pages and PMD hugepages.
> 
> Instead of creating another copy of the operations when we introduce
> PUD hugepages, let's share them across the different pagesizes.
> 
> Signed-off-by: Punit Agrawal <punit.agrawal@xxxxxxx>
> Cc: Christoffer Dall <christoffer.dall@xxxxxxx>
> Cc: Marc Zyngier <marc.zyngier@xxxxxxx>
> ---
>  virt/kvm/arm/mmu.c | 68 +++++++++++++++++++++++++++-------------------
>  1 file changed, 40 insertions(+), 28 deletions(-)
> 
> diff --git a/virt/kvm/arm/mmu.c b/virt/kvm/arm/mmu.c
> index 1d90d79706bd..dd14cc36c51c 100644
> --- a/virt/kvm/arm/mmu.c
> +++ b/virt/kvm/arm/mmu.c
> @@ -1398,6 +1398,21 @@ static void invalidate_icache_guest_page(kvm_pfn_t pfn, unsigned long size)
>  	__invalidate_icache_guest_page(pfn, size);
>  }
>  
> +static bool stage2_should_exec(struct kvm *kvm, phys_addr_t addr,
> +			       bool exec_fault, unsigned long fault_status)

I find this "should exec" very confusing.

> +{
> +	/*
> +	 * If we took an execution fault we will have made the
> +	 * icache/dcache coherent and should now let the s2 mapping be
> +	 * executable.
> +	 *
> +	 * Write faults (!exec_fault && FSC_PERM) are orthogonal to
> +	 * execute permissions, and we preserve whatever we have.
> +	 */
> +	return exec_fault ||
> +		(fault_status == FSC_PERM && stage2_is_exec(kvm, addr));
> +}
> +
>  static void kvm_send_hwpoison_signal(unsigned long address,
>  				     struct vm_area_struct *vma)
>  {
> @@ -1431,7 +1446,7 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
>  	kvm_pfn_t pfn;
>  	pgprot_t mem_type = PAGE_S2;
>  	bool logging_active = memslot_is_logging(memslot);
> -	unsigned long flags = 0;
> +	unsigned long vma_pagesize, flags = 0;
>  
>  	write_fault = kvm_is_write_fault(vcpu);
>  	exec_fault = kvm_vcpu_trap_is_iabt(vcpu);
> @@ -1451,7 +1466,8 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
>  		return -EFAULT;
>  	}
>  
> -	if (vma_kernel_pagesize(vma) == PMD_SIZE && !logging_active) {
> +	vma_pagesize = vma_kernel_pagesize(vma);
> +	if (vma_pagesize == PMD_SIZE && !logging_active) {
>  		hugetlb = true;
>  		gfn = (fault_ipa & PMD_MASK) >> PAGE_SHIFT;
>  	} else {
> @@ -1520,28 +1536,34 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
>  	if (mmu_notifier_retry(kvm, mmu_seq))
>  		goto out_unlock;
>  
> -	if (!hugetlb && !force_pte)
> +	if (!hugetlb && !force_pte) {
> +		/*
> +		 * Only PMD_SIZE transparent hugepages(THP) are
> +		 * currently supported. This code will need to be
> +		 * updated to support other THP sizes.
> +		 */
>  		hugetlb = transparent_hugepage_adjust(&pfn, &fault_ipa);
> +		if (hugetlb)
> +			vma_pagesize = PMD_SIZE;
> +	}
> +
> +	if (writable)
> +		kvm_set_pfn_dirty(pfn);
>  
> -	if (hugetlb) {
> +	if (fault_status != FSC_PERM)
> +		clean_dcache_guest_page(pfn, vma_pagesize);
> +
> +	if (exec_fault)
> +		invalidate_icache_guest_page(pfn, vma_pagesize);
> +
> +	if (hugetlb && vma_pagesize == PMD_SIZE) {
>  		pmd_t new_pmd = pfn_pmd(pfn, mem_type);
>  		new_pmd = pmd_mkhuge(new_pmd);
> -		if (writable) {
> +		if (writable)
>  			new_pmd = kvm_s2pmd_mkwrite(new_pmd);
> -			kvm_set_pfn_dirty(pfn);
> -		}
>  
> -		if (fault_status != FSC_PERM)
> -			clean_dcache_guest_page(pfn, PMD_SIZE);
> -
> -		if (exec_fault) {
> +		if (stage2_should_exec(kvm, fault_ipa, exec_fault, fault_status))
>  			new_pmd = kvm_s2pmd_mkexec(new_pmd);

OK, I find this absolutely horrid... ;-)

The rest of the function deals with discrete flags, and all of a sudden
we have a function call with a bunch of seemingly unrelated parameters.
And you are repeating it for each vma_pagesize...

How about something like:

	bool needs_exec;

	[...]

	needs_exec = exec_fault || (fault_status == FSC_PERM &&
				    stage2_is_exec(kvm, fault_ipa);

And then you just check needs_exec to update the pte/pmd. And you drop
this helper.

> -			invalidate_icache_guest_page(pfn, PMD_SIZE);
> -		} else if (fault_status == FSC_PERM) {
> -			/* Preserve execute if XN was already cleared */
> -			if (stage2_is_exec(kvm, fault_ipa))
> -				new_pmd = kvm_s2pmd_mkexec(new_pmd);
> -		}
>  
>  		ret = stage2_set_pmd_huge(kvm, memcache, fault_ipa, &new_pmd);
>  	} else {
> @@ -1549,21 +1571,11 @@ static int user_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
>  
>  		if (writable) {
>  			new_pte = kvm_s2pte_mkwrite(new_pte);
> -			kvm_set_pfn_dirty(pfn);
>  			mark_page_dirty(kvm, gfn);
>  		}
>  
> -		if (fault_status != FSC_PERM)
> -			clean_dcache_guest_page(pfn, PAGE_SIZE);
> -
> -		if (exec_fault) {
> +		if (stage2_should_exec(kvm, fault_ipa, exec_fault, fault_status))
>  			new_pte = kvm_s2pte_mkexec(new_pte);
> -			invalidate_icache_guest_page(pfn, PAGE_SIZE);
> -		} else if (fault_status == FSC_PERM) {
> -			/* Preserve execute if XN was already cleared */
> -			if (stage2_is_exec(kvm, fault_ipa))
> -				new_pte = kvm_s2pte_mkexec(new_pte);
> -		}
>  
>  		ret = stage2_set_pte(kvm, memcache, fault_ipa, &new_pte, flags);
>  	}
> 

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...