On Tue, Mar 16, 2021 at 08:38:00PM +0000, David Howells wrote:

Matthew Wilcox [off-list ref] wrote:

quoted

So ... a page with both flags cleared should have a refcount of N.
A page with one or both flags set should have a refcount of N+1.
...
How is a poor filesystem supposed to make that true?  Also btrfs has this
problem since it uses private_2 for its own purposes.

It's simpler if it's N+2 for both patches set.  Btw, patch 13 adds that - and
possibly that should be merged into an earlier patch.

So ...

static inline int page_has_private(struct page *page)
{
	unsigned long flags = page->flags;
	return ((flags >> PG_private) & 1) + ((flags >> PG_private_2) & 1);
}

perhaps?