Re: [PATCH v9 3/6] KVM: mmu: Improve handling of non-refcounted pfns

[Sean Christopherson <seanjc@google.com>]

On Mon, Sep 11, 2023, David Stevens wrote:
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index c2e0ddf14dba..2ed08ae1a9be 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -1185,10 +1185,31 @@ struct kvm_follow_pfn {
>  	bool atomic;
>  	/* Try to create a writable mapping even for a read fault */
>  	bool try_map_writable;
> +	/* Usage of the returned pfn will be guared by a mmu notifier. */
> +	bool guarded_by_mmu_notifier;
> +	/*
> +	 * When false, do not return pfns for non-refcounted struct pages.
> +	 *
> +	 * TODO: This allows callers to use kvm_release_pfn on the pfns
> +	 * returned by gfn_to_pfn without worrying about corrupting the
> +	 * refcounted of non-refcounted pages. Once all callers respect
> +	 * is_refcounted_page, this flag should be removed.
> +	 */
> +	bool allow_non_refcounted_struct_page;
>  
>  	/* Outputs of __kvm_follow_pfn */
>  	hva_t hva;
>  	bool writable;
> +	/*
> +	 * True if the returned pfn is for a page with a valid refcount. False
> +	 * if the returned pfn has no struct page or if the struct page is not
> +	 * being refcounted (e.g. tail pages of non-compound higher order
> +	 * allocations from IO/PFNMAP mappings).
> +	 *
> +	 * When this output flag is false, callers should not try to convert
> +	 * the pfn to a struct page.
> +	 */
> +	bool is_refcounted_page;

Idea.  Hopefully a good one.  Rather than tracking a bool, what if we track:

	struct page *refcounted_page;

and then make kvm_xxx_page_clean() wrappers around inner helpers that play nice
with NULL pages, e.g.

  static inline void kvm_release_page_clean(struct page *page)
  {
  	if (!page)
		return

  	__kvm_release_page_clean(page);
  }

Then callers of __kvm_follow_pfn() can do:

	kvm_release_page_clean(fault->refcounted_page);

instead of

 	if (fault->is_refcounted_page)
		kvm_release_page_clean(pfn_to_page(fault->pfn));