Re: [PATCH RFC 09/14] mm/page-flags: check page/folio->private instead of PG_private

On Fri Jul 31, 2026 at 10:13 PM EDT, Zi Yan wrote:
> After the changes of the prior commits, page/folio->private != NULL is now
> equivalent to checking PG_private.
>
> Stop checking PG_private on pages and folios and use page/folio->private
> instead, except swapcache and hugetlb folios, because the former uses a
> field (swp_entry_t swap) overlapping with ->private and the latter sets its
> flags in ->private. Exclude swapcache and hugetlb when the code is meant to
> check PG_private only.
>
> folio_set/clear_private() and Set/ClearPagePrivate() become no-ops.
> PG_private is no longer checked at page free time.
>
> KPF_PRIVATE exposes PG_private to userspace. Change its code logic to check
> folio->private != NULL and exclude non-pagecache, swapcache, hugetlb, and
> anon folios. One minor semantic change, for orphaned pagecache folios
> (mapping == NULL) with fs-private data will no longer have KPF_PRIVATE.
>
> Assisted-by: Claude:claude-opus-4-8
> Assisted-by: Codex:gpt-5
> Signed-off-by: Zi Yan <[email protected]>
> To: Andrew Morton <[email protected]>
> To: David Hildenbrand <[email protected]>
> To: Steven Rostedt <[email protected]>
> To: Masami Hiramatsu <[email protected]>
> To: Lorenzo Stoakes <[email protected]>
> To: "Matthew Wilcox (Oracle)" <[email protected]>
> To: Jan Kara <[email protected]>
> To: Johannes Weiner <[email protected]>
> Cc: "Liam R. Howlett" <[email protected]>
> Cc: Vlastimil Babka <[email protected]>
> Cc: Mike Rapoport <[email protected]>
> Cc: Suren Baghdasaryan <[email protected]>
> Cc: Michal Hocko <[email protected]>
> Cc: Mathieu Desnoyers <[email protected]>
> Cc: Zi Yan <[email protected]>
> Cc: Baolin Wang <[email protected]>
> Cc: Nico Pache <[email protected]>
> Cc: Ryan Roberts <[email protected]>
> Cc: Dev Jain <[email protected]>
> Cc: Barry Song <[email protected]>
> Cc: Lance Yang <[email protected]>
> Cc: Usama Arif <[email protected]>
> Cc: Matthew Brost <[email protected]>
> Cc: Joshua Hahn <[email protected]>
> Cc: Rakie Kim <[email protected]>
> Cc: Byungchul Park <[email protected]>
> Cc: Gregory Price <[email protected]>
> Cc: Ying Huang <[email protected]>
> Cc: Alistair Popple <[email protected]>
> Cc: Qi Zheng <[email protected]>
> Cc: Shakeel Butt <[email protected]>
> Cc: Kairui Song <[email protected]>
> Cc: Axel Rasmussen <[email protected]>
> Cc: Yuanchu Xie <[email protected]>
> Cc: Wei Xu <[email protected]>
> Cc: [email protected]
> Cc: [email protected]
> Cc: [email protected]
> Cc: [email protected]
> ---
>  fs/proc/page.c                 |  6 +++++-
>  include/linux/mm.h             | 11 ++++++-----
>  include/linux/page-flags.h     | 26 +++++++++++++++++++++-----
>  include/trace/events/pagemap.h |  4 +++-
>  mm/huge_memory.c               |  4 +++-
>  mm/migrate.c                   |  3 ++-
>  mm/page-writeback.c            |  5 ++++-
>  mm/vmscan.c                    |  3 ++-
>  8 files changed, 46 insertions(+), 16 deletions(-)
>
> diff --git a/fs/proc/page.c b/fs/proc/page.c
> index 260772b20bd99..abfa6f7d890cc 100644
> --- a/fs/proc/page.c
> +++ b/fs/proc/page.c
> @@ -232,7 +232,11 @@ u64 stable_page_flags(const struct page *page)
>  
>  	u |= kpf_copy_bit(k, KPF_RESERVED,	PG_reserved);
>  	u |= kpf_copy_bit(k, KPF_OWNER_2,	PG_owner_2);
> -	u |= kpf_copy_bit(k, KPF_PRIVATE,	PG_private);
> +	/* preserve the original KPF_PRIVATE semantics by excluding non pagecache folios */
> +	if (folio->mapping && !folio_test_anon(folio) &&
> +	    (folio_get_private(folio) && !folio_test_swapcache(folio) &&
> +	     !folio_test_hugetlb(folio)))
> +		u |= BIT_ULL(KPF_PRIVATE);

Based on the discussion in THP cabal meeting, KPF_PRIVATE will be
removed in v2, since kernel page flags are not stable.

>  	u |= kpf_copy_bit(k, KPF_PRIVATE_2,	PG_private_2);
>  	u |= kpf_copy_bit(k, KPF_OWNER_PRIVATE,	PG_owner_priv_1);
>  	u |= kpf_copy_bit(k, KPF_ARCH,		PG_arch_1);
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 7fabe6c66b4b7..ebc035ac26ccc 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -2961,9 +2961,9 @@ static inline bool folio_maybe_mapped_shared(struct folio *folio)
>   * @folio: the folio
>   *
>   * Calculate the expected folio refcount, taking references from the pagecache,
> - * swapcache, PG_private and page table mappings into account. Useful in
> - * combination with folio_ref_count() to detect unexpected references (e.g.,
> - * GUP or other temporary references).
> + * swapcache, private data (folio->private != NULL) and page table mappings into
> + * account. Useful in combination with folio_ref_count() to detect unexpected
> + * references (e.g., GUP or other temporary references).
>   *
>   * Does currently not consider references from the LRU cache. If the folio
>   * was isolated from the LRU (which is the case during migration or split),
> @@ -3003,8 +3003,9 @@ static inline int folio_expected_ref_count(const struct folio *folio)
>  	if (!folio_test_anon(folio)) {
>  		/* One reference per page from the pagecache. */
>  		ref_count += !!folio->mapping << order;
> -		/* One reference from PG_private. */
> -		ref_count += folio_test_private(folio);
> +		/* One reference from filesystem private data. */
> +		ref_count += !!folio->private && !folio_test_hugetlb(folio) &&
> +			     !folio_test_swapcache(folio);
>  	}

Sashiko asked:

folio_expected_ref_count() can be called in lockless contexts, does
reading folio->private directly introduce a data race? And whether a
READ_ONCE() should be used for folio->private.


Answer:

Yes, we expect that and caller should recheck refcount after they hold
the folio lock. Use data_race() for folio->private reading. While at it,
add data_race() for folio->mapping too.




-- 
Best Regards,
Yan, Zi