[PATCH v2 0/2] fsverity: remove fsverity_verify_page()

[PATCH v2 0/2] fsverity: remove fsverity_verify_page()

[Eric Biggers]

This series removes the non-large-folio-aware function
fsverity_verify_page(), which is no longer needed.

Changed in v2:
    - Made one specific part of f2fs_verify_cluster() large-folio-aware.

Eric Biggers (2):
  f2fs: use fsverity_verify_blocks() instead of fsverity_verify_page()
  fsverity: remove fsverity_verify_page()

 fs/f2fs/compress.c       | 6 +++++-
 fs/verity/verify.c       | 4 ++--
 include/linux/fsverity.h | 6 ------
 3 files changed, 7 insertions(+), 9 deletions(-)


base-commit: 3e48a11675c50698374d4ac596fb506736eb1c53
-- 
2.53.0

[PATCH v2 1/2] f2fs: use fsverity_verify_blocks() instead of fsverity_verify_page()

[Eric Biggers]

Replace the only remaining caller of fsverity_verify_page() with a
direct call to fsverity_verify_blocks().  This will allow
fsverity_verify_page() to be removed.

Make it large-folio-aware by using the page's offset in the folio
instead of 0, though the rest of f2fs_verify_cluster() and f2fs
decompression as a whole still assumes small folios.

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Eric Biggers <ebiggers@kernel.org>
---
 fs/f2fs/compress.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/fs/f2fs/compress.c b/fs/f2fs/compress.c
index 006a80acd1de..11c4de515f98 100644
--- a/fs/f2fs/compress.c
+++ b/fs/f2fs/compress.c
@@ -1811,15 +1811,19 @@ static void f2fs_verify_cluster(struct work_struct *work)
 	int i;
 
 	/* Verify, update, and unlock the decompressed pages. */
 	for (i = 0; i < dic->cluster_size; i++) {
 		struct page *rpage = dic->rpages[i];
+		struct folio *rfolio;
+		size_t offset;
 
 		if (!rpage)
 			continue;
+		rfolio = page_folio(rpage);
+		offset = folio_page_idx(rfolio, rpage) * PAGE_SIZE;
 
-		if (fsverity_verify_page(dic->vi, rpage))
+		if (fsverity_verify_blocks(dic->vi, rfolio, PAGE_SIZE, offset))
 			SetPageUptodate(rpage);
 		else
 			ClearPageUptodate(rpage);
 		unlock_page(rpage);
 	}
-- 
2.53.0

Re: [PATCH v2 1/2] f2fs: use fsverity_verify_blocks() instead of fsverity_verify_page()

[Eric Biggers]

On Sat, Feb 14, 2026 at 01:18:29PM -0800, Eric Biggers wrote:
> Replace the only remaining caller of fsverity_verify_page() with a
> direct call to fsverity_verify_blocks().  This will allow
> fsverity_verify_page() to be removed.
> 
> Make it large-folio-aware by using the page's offset in the folio
> instead of 0, though the rest of f2fs_verify_cluster() and f2fs
> decompression as a whole still assumes small folios.
> 
> Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
> Signed-off-by: Eric Biggers <ebiggers@kernel.org>
> ---
>  fs/f2fs/compress.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/f2fs/compress.c b/fs/f2fs/compress.c
> index 006a80acd1de..11c4de515f98 100644
> --- a/fs/f2fs/compress.c
> +++ b/fs/f2fs/compress.c
> @@ -1811,15 +1811,19 @@ static void f2fs_verify_cluster(struct work_struct *work)
>  	int i;
>  
>  	/* Verify, update, and unlock the decompressed pages. */
>  	for (i = 0; i < dic->cluster_size; i++) {
>  		struct page *rpage = dic->rpages[i];
> +		struct folio *rfolio;
> +		size_t offset;
>  
>  		if (!rpage)
>  			continue;
> +		rfolio = page_folio(rpage);
> +		offset = folio_page_idx(rfolio, rpage) * PAGE_SIZE;
>  
> -		if (fsverity_verify_page(dic->vi, rpage))
> +		if (fsverity_verify_blocks(dic->vi, rfolio, PAGE_SIZE, offset))
>  			SetPageUptodate(rpage);
>  		else
>  			ClearPageUptodate(rpage);
>  		unlock_page(rpage);

Let me know if you'd prefer that we verified the whole folio here
instead.  Either way, the behavior will be still incorrect if this
function is passed a large folio (which it's not).  Either we'd mark the
whole folio up-to-date after verifying only one page in it, or we'd
access pages that were not in the array of pages passed to the function.

- Eric

Re: [PATCH v2 1/2] f2fs: use fsverity_verify_blocks() instead of fsverity_verify_page()

[Linus Torvalds]

On Sat, 14 Feb 2026 at 13:50, Eric Biggers <ebiggers@kernel.org> wrote:
>
> Let me know if you'd prefer that we verified the whole folio here
> instead.

This looks good to me.

And hopefully some day that "rpages" becomes "rfolio" (and this can
all go away and it becomes fsverity_verify_folio() and simpler).

It does look like the "cluster size" thing could maybe be made to
simply be the size of one folio for those things, and then being a
single folio of size "PAGE_SIZE << i_log_cluster_size" might simplify
other code too.

Then instead of walking multiple pages, you'd always have exactly one
folio (just different sizes depending on cluster size).

But that's just from a very quick look, and I might mis-understand the
code I saw...

                      Linus

Re: [PATCH v2 1/2] f2fs: use fsverity_verify_blocks() instead of fsverity_verify_page()

[Matthew Wilcox]

On Sat, Feb 14, 2026 at 01:50:08PM -0800, Eric Biggers wrote:
> On Sat, Feb 14, 2026 at 01:18:29PM -0800, Eric Biggers wrote:
> > +++ b/fs/f2fs/compress.c
> > @@ -1811,15 +1811,19 @@ static void f2fs_verify_cluster(struct work_struct *work)
> >  	int i;
> >  
> >  	/* Verify, update, and unlock the decompressed pages. */
> >  	for (i = 0; i < dic->cluster_size; i++) {
> >  		struct page *rpage = dic->rpages[i];
> > +		struct folio *rfolio;
> > +		size_t offset;
> >  
> >  		if (!rpage)
> >  			continue;
> > +		rfolio = page_folio(rpage);
> > +		offset = folio_page_idx(rfolio, rpage) * PAGE_SIZE;
> >  
> > -		if (fsverity_verify_page(dic->vi, rpage))
> > +		if (fsverity_verify_blocks(dic->vi, rfolio, PAGE_SIZE, offset))
> >  			SetPageUptodate(rpage);

Yeah, no.

		if (fsverity_verify_blocks(dic->vi, rfolio,
				folio_size(rfolio), 0));
			folio_mark_uptodate(rfolio);

> >  		else
> >  			ClearPageUptodate(rpage);

This never needed to be here.  The folio must already be !uptodate.
Just delete these two lines.

> >  		unlock_page(rpage);

folio_unlock(rfolio);

Re: [PATCH v2 1/2] f2fs: use fsverity_verify_blocks() instead of fsverity_verify_page()

[Eric Biggers]

On Sun, Feb 15, 2026 at 03:38:20AM +0000, Matthew Wilcox wrote:
> On Sat, Feb 14, 2026 at 01:50:08PM -0800, Eric Biggers wrote:
> > On Sat, Feb 14, 2026 at 01:18:29PM -0800, Eric Biggers wrote:
> > > +++ b/fs/f2fs/compress.c
> > > @@ -1811,15 +1811,19 @@ static void f2fs_verify_cluster(struct work_struct *work)
> > >  	int i;
> > >  
> > >  	/* Verify, update, and unlock the decompressed pages. */
> > >  	for (i = 0; i < dic->cluster_size; i++) {
> > >  		struct page *rpage = dic->rpages[i];
> > > +		struct folio *rfolio;
> > > +		size_t offset;
> > >  
> > >  		if (!rpage)
> > >  			continue;
> > > +		rfolio = page_folio(rpage);
> > > +		offset = folio_page_idx(rfolio, rpage) * PAGE_SIZE;
> > >  
> > > -		if (fsverity_verify_page(dic->vi, rpage))
> > > +		if (fsverity_verify_blocks(dic->vi, rfolio, PAGE_SIZE, offset))
> > >  			SetPageUptodate(rpage);
> 
> Yeah, no.
> 
> 		if (fsverity_verify_blocks(dic->vi, rfolio,
> 				folio_size(rfolio), 0));
> 			folio_mark_uptodate(rfolio);
> 
> > >  		else
> > >  			ClearPageUptodate(rpage);
> 
> This never needed to be here.  The folio must already be !uptodate.
> Just delete these two lines.
> 
> > >  		unlock_page(rpage);
> 
> folio_unlock(rfolio);

Sure.  This kind of scope creep is why I wanted to just do the
straightforward conversion for now.

- Eric

[PATCH v2 2/2] fsverity: remove fsverity_verify_page()

[Eric Biggers]

Now that fsverity_verify_page() has no callers, remove it.

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Eric Biggers <ebiggers@kernel.org>
---
 fs/verity/verify.c       | 4 ++--
 include/linux/fsverity.h | 6 ------
 2 files changed, 2 insertions(+), 8 deletions(-)

diff --git a/fs/verity/verify.c b/fs/verity/verify.c
index 31797f9b24d0..3e38749fbc82 100644
--- a/fs/verity/verify.c
+++ b/fs/verity/verify.c
@@ -431,12 +431,12 @@ EXPORT_SYMBOL_GPL(fsverity_verify_blocks);
  * verification, then bio->bi_status is set to an error status.
  *
  * This is a helper function for use by the ->readahead() method of filesystems
  * that issue bios to read data directly into the page cache.  Filesystems that
  * populate the page cache without issuing bios (e.g. non block-based
- * filesystems) must instead call fsverity_verify_page() directly on each page.
- * All filesystems must also call fsverity_verify_page() on holes.
+ * filesystems) must instead call fsverity_verify_blocks() directly.  All
+ * filesystems must also call fsverity_verify_blocks() on holes.
  */
 void fsverity_verify_bio(struct fsverity_info *vi, struct bio *bio)
 {
 	struct fsverity_verification_context ctx;
 	struct folio_iter fi;
diff --git a/include/linux/fsverity.h b/include/linux/fsverity.h
index fed91023bea9..6de3ddf0b148 100644
--- a/include/linux/fsverity.h
+++ b/include/linux/fsverity.h
@@ -280,16 +280,10 @@ static inline bool fsverity_verify_folio(struct fsverity_info *vi,
 					 struct folio *folio)
 {
 	return fsverity_verify_blocks(vi, folio, folio_size(folio), 0);
 }
 
-static inline bool fsverity_verify_page(struct fsverity_info *vi,
-					struct page *page)
-{
-	return fsverity_verify_blocks(vi, page_folio(page), PAGE_SIZE, 0);
-}
-
 /**
  * fsverity_file_open() - prepare to open a verity file
  * @inode: the inode being opened
  * @filp: the struct file being set up
  *
-- 
2.53.0