In preparation for moving preparedness out of the folio flags, pass
the struct file* or struct inode* down to kvm_gmem_mark_prepared,
as well as the offset within the gmem file.  Introduce new functions
to unprepare page on punch-hole, and to query the state.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 virt/kvm/guest_memfd.c | 27 ++++++++++++++++++++-------
 1 file changed, 20 insertions(+), 7 deletions(-)

diff --git a/virt/kvm/guest_memfd.c b/virt/kvm/guest_memfd.c
index 3ea5a7597fd4..416e02a00cae 100644
--- a/virt/kvm/guest_memfd.c
+++ b/virt/kvm/guest_memfd.c
@@ -107,18 +107,28 @@ static int __kvm_gmem_prepare_folio(struct kvm *kvm, struct kvm_memory_slot *slo
 	return 0;
 }
 
-static inline void kvm_gmem_mark_prepared(struct folio *folio)
+static void kvm_gmem_mark_prepared(struct file *file, pgoff_t index, struct folio *folio)
 {
 	folio_mark_uptodate(folio);
 }
 
+static void kvm_gmem_mark_range_unprepared(struct inode *inode, pgoff_t index, pgoff_t npages)
+{
+}
+
+static bool kvm_gmem_is_prepared(struct file *file, pgoff_t index, struct folio *folio)
+{
+	return folio_test_uptodate(folio);
+}
+
 /*
  * Process @folio, which contains @gfn, so that the guest can use it.
  * The folio must be locked and the gfn must be contained in @slot.
  * On successful return the guest sees a zero page so as to avoid
  * leaking host data and the up-to-date flag is set.
  */
-static int kvm_gmem_prepare_folio(struct kvm *kvm, struct kvm_memory_slot *slot,
+static int kvm_gmem_prepare_folio(struct kvm *kvm, struct file *file,
+				  struct kvm_memory_slot *slot,
 				  gfn_t gfn, struct folio *folio)
 {
 	unsigned long nr_pages, i;
@@ -147,7 +157,7 @@ static int kvm_gmem_prepare_folio(struct kvm *kvm, struct kvm_memory_slot *slot,
 	index = ALIGN_DOWN(index, 1 << folio_order(folio));
 	r = __kvm_gmem_prepare_folio(kvm, slot, index, folio);
 	if (!r)
-		kvm_gmem_mark_prepared(folio);
+		kvm_gmem_mark_prepared(file, index, folio);
 
 	return r;
 }
@@ -231,6 +241,7 @@ static long kvm_gmem_punch_hole(struct inode *inode, loff_t offset, loff_t len)
 		kvm_gmem_invalidate_begin(gmem, start, end);
 
 	truncate_inode_pages_range(inode->i_mapping, offset, offset + len - 1);
+	kvm_gmem_mark_range_unprepared(inode, start, end - start);
 
 	list_for_each_entry(gmem, gmem_list, entry)
 		kvm_gmem_invalidate_end(gmem, start, end);
@@ -682,7 +693,7 @@ __kvm_gmem_get_pfn(struct file *file, struct kvm_memory_slot *slot,
 	if (max_order)
 		*max_order = 0;
 
-	*is_prepared = folio_test_uptodate(folio);
+	*is_prepared = kvm_gmem_is_prepared(file, index, folio);
 	return folio;
 }
 
@@ -704,7 +715,7 @@ int kvm_gmem_get_pfn(struct kvm *kvm, struct kvm_memory_slot *slot,
 	}
 
 	if (!is_prepared)
-		r = kvm_gmem_prepare_folio(kvm, slot, gfn, folio);
+		r = kvm_gmem_prepare_folio(kvm, file, slot, gfn, folio);
 
 	folio_unlock(folio);
 	if (r < 0)
@@ -781,8 +792,10 @@ long kvm_gmem_populate(struct kvm *kvm, gfn_t start_gfn, void __user *src, long
 
 		p = src ? src + i * PAGE_SIZE : NULL;
 		ret = post_populate(kvm, gfn, pfn, p, max_order, opaque);
-		if (!ret)
-			kvm_gmem_mark_prepared(folio);
+		if (!ret) {
+			pgoff_t index = gfn - slot->base_gfn + slot->gmem.pgoff;
+			kvm_gmem_mark_prepared(file, index, folio);
+		}
 
 put_folio_and_exit:
 		folio_put(folio);
-- 
2.43.5

On Fri, Nov 08, 2024, Paolo Bonzini wrote:
> In preparation for moving preparedness out of the folio flags, pass
> the struct file* or struct inode* down to kvm_gmem_mark_prepared,
> as well as the offset within the gmem file.  Introduce new functions
> to unprepare page on punch-hole, and to query the state.

...

> +static bool kvm_gmem_is_prepared(struct file *file, pgoff_t index, struct folio *folio)
> +{
> +	return folio_test_uptodate(folio);
> +}
> +
>  /*
>   * Process @folio, which contains @gfn, so that the guest can use it.
>   * The folio must be locked and the gfn must be contained in @slot.
>   * On successful return the guest sees a zero page so as to avoid
>   * leaking host data and the up-to-date flag is set.
>   */
> -static int kvm_gmem_prepare_folio(struct kvm *kvm, struct kvm_memory_slot *slot,
> +static int kvm_gmem_prepare_folio(struct kvm *kvm, struct file *file,
> +				  struct kvm_memory_slot *slot,
>  				  gfn_t gfn, struct folio *folio)
>  {
>  	unsigned long nr_pages, i;
> @@ -147,7 +157,7 @@ static int kvm_gmem_prepare_folio(struct kvm *kvm, struct kvm_memory_slot *slot,
>  	index = ALIGN_DOWN(index, 1 << folio_order(folio));
>  	r = __kvm_gmem_prepare_folio(kvm, slot, index, folio);
>  	if (!r)
> -		kvm_gmem_mark_prepared(folio);
> +		kvm_gmem_mark_prepared(file, index, folio);
>  
>  	return r;
>  }
> @@ -231,6 +241,7 @@ static long kvm_gmem_punch_hole(struct inode *inode, loff_t offset, loff_t len)
>  		kvm_gmem_invalidate_begin(gmem, start, end);
>  
>  	truncate_inode_pages_range(inode->i_mapping, offset, offset + len - 1);
> +	kvm_gmem_mark_range_unprepared(inode, start, end - start);
>  
>  	list_for_each_entry(gmem, gmem_list, entry)
>  		kvm_gmem_invalidate_end(gmem, start, end);
> @@ -682,7 +693,7 @@ __kvm_gmem_get_pfn(struct file *file, struct kvm_memory_slot *slot,
>  	if (max_order)
>  		*max_order = 0;
>  
> -	*is_prepared = folio_test_uptodate(folio);
> +	*is_prepared = kvm_gmem_is_prepared(file, index, folio);
>  	return folio;
>  }
>  
> @@ -704,7 +715,7 @@ int kvm_gmem_get_pfn(struct kvm *kvm, struct kvm_memory_slot *slot,
>  	}
>  
>  	if (!is_prepared)
> -		r = kvm_gmem_prepare_folio(kvm, slot, gfn, folio);
> +		r = kvm_gmem_prepare_folio(kvm, file, slot, gfn, folio);

This is broken when the next patch comes along.  If KVM encounters a partially
prepared folio, i.e. a folio with some prepared pages and some unprepared pages,
then KVM needs to zero only the unprepared pages.  But kvm_gmem_prepare_folio()
zeroes everything.

static int kvm_gmem_prepare_folio(struct kvm *kvm, struct file *file,
				  struct kvm_memory_slot *slot,
				  gfn_t gfn, struct folio *folio)
{
	unsigned long nr_pages, i;
	pgoff_t index;
	int r;

	nr_pages = folio_nr_pages(folio);
	for (i = 0; i < nr_pages; i++)
		clear_highpage(folio_page(folio, i));