Re: [PATCH 07/16] fsverity: don't issue readahead for non-ENOENT errors from __filemap_get_folio

[Eric Biggers <ebiggers@kernel.org>]

On Mon, Jan 26, 2026 at 11:11:02AM -0800, Darrick J. Wong wrote:
> On Mon, Jan 26, 2026 at 05:50:53AM +0100, Christoph Hellwig wrote:
> > Issuing more reads on errors is not a good idea, especially when the
> > most common error here is -ENOMEM.
> > 
> > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > ---
> >  fs/verity/pagecache.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > diff --git a/fs/verity/pagecache.c b/fs/verity/pagecache.c
> > index 1efcdde20b73..63393f0f5834 100644
> > --- a/fs/verity/pagecache.c
> > +++ b/fs/verity/pagecache.c
> > @@ -22,7 +22,8 @@ struct page *generic_read_merkle_tree_page(struct inode *inode, pgoff_t index,
> >  	struct folio *folio;
> >  
> >  	folio = __filemap_get_folio(inode->i_mapping, index, FGP_ACCESSED, 0);
> > -	if (IS_ERR(folio) || !folio_test_uptodate(folio)) {
> > +	if (PTR_ERR(folio) == -ENOENT ||
> > +	    !(IS_ERR(folio) && !folio_test_uptodate(folio))) {
> 
> I don't understand this logic at all.  If @folio is actually an
> ERR_PTR, then we dereference the non-folio to see if it's not uptodate?
> 
> I think (given the previous revisions) that what you want is to initiate
> readahead if either there's no folio at all (ENOENT) or if there is a
> folio but it's not uptodate?  But not if there's some other error
> (ENOMEM, EL3HLT, EFSCORRUPTED, etc)?
> 
> So maybe you want:
> 
> 	folio = __filemap_get_folio(...);
> 	if (!IS_ERR(folio)) {
> 		if (folio_test_uptodate(folio))
> 			return folio_file_page(folio);
> 		folio_put(folio);
> 	} else if (PTR_ERR(folio) == -ENOENT) {
> 		return ERR_CAST(folio);
> 	}
> 
> 	if (num_ra_pages > 1)
> 		page_cache_ra_unbounded(&ractl, num_ra_pages, 0);
> 	folio = read_mapping_folio(inode->i_mapping, index, NULL);
> 	if (IS_ERR(folio))
> 		return ERR_CAST(folio);
> 
> 	return folio_file_page(folio);
> 
> <confused>

That version is wrong too: the condition 'PTR_ERR(folio) == -ENOENT' is
backwards.

This code gets replaced later in the series anyway.  For this patch, we
could simply insert two lines:

	folio = __filemap_get_folio(inode->i_mapping, index, FGP_ACCESSED, 0);
+	if (IS_ERR(folio) && folio != ERR_PTR(-ENOENT))
+		return folio;

Then for the final version in generic_readahead_merkle_tree(), one
option would be:

	struct folio *folio;

	folio = __filemap_get_folio(inode->i_mapping, index, FGP_ACCESSED, 0);
	if (folio == ERR_PTR(-ENOENT) ||
	    (!IS_ERR(folio) && !folio_test_uptodate(folio))) {
		DEFINE_READAHEAD(ractl, NULL, NULL, inode->i_mapping, index);

		page_cache_ra_unbounded(&ractl, nr_pages, 0);
	}
	if (!IS_ERR(folio))
		folio_put(folio);

Or as a diff from this series:

-	if (PTR_ERR(folio) == -ENOENT ||
-	    !(IS_ERR(folio) && !folio_test_uptodate(folio))) {
+	if (folio == ERR_PTR(-ENOENT) ||
+	    (!IS_ERR(folio) && !folio_test_uptodate(folio))) {

(Note that PTR_ERR() shouldn't be used before it's known that the
pointer is an error pointer.)

- Eric