This patch adds a desired folio order in file_ra_state in order to get
non-zero-order allocation for readahead.

Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
---
 include/linux/fs.h | 1 +
 mm/filemap.c       | 2 +-
 mm/readahead.c     | 2 +-
 3 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/include/linux/fs.h b/include/linux/fs.h
index c895146c1444..8233b166139b 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1167,6 +1167,7 @@ struct file_ra_state {
 	unsigned int async_size;
 	unsigned int ra_pages;
 	unsigned short order;
+	unsigned short desired_order;
 	unsigned short mmap_miss;
 	loff_t prev_pos;
 };
diff --git a/mm/filemap.c b/mm/filemap.c
index 13f0259d993c..6aa0e26c5d17 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -3332,7 +3332,7 @@ static struct file *do_sync_mmap_readahead(struct vm_fault *vmf)
 		ra->start = max_t(long, 0, vmf->pgoff - ra->ra_pages / 2);
 		ra->size = ra->ra_pages;
 		ra->async_size = ra->ra_pages / 4;
-		ra->order = 0;
+		ra->order = ra->desired_order;
 	}
 
 	fpin = maybe_unlock_mmap_for_io(vmf, fpin);
diff --git a/mm/readahead.c b/mm/readahead.c
index 3a4b5d58eeb6..5194211dfdee 100644
--- a/mm/readahead.c
+++ b/mm/readahead.c
@@ -621,7 +621,7 @@ void page_cache_sync_ra(struct readahead_control *ractl,
 	ra->size = min(contig_count + req_count, max_pages);
 	ra->async_size = 1;
 readit:
-	ra->order = 0;
+	ra->order = ra->desired_order;
 	ractl->_index = ra->start;
 	page_cache_ra_order(ractl, ra);
 }
-- 
2.52.0.487.g5c8c507ade-goog

On Fri, Nov 21, 2025 at 01:48:11AM +0000, Jaegeuk Kim wrote:
> This patch adds a desired folio order in file_ra_state in order to get
> non-zero-order allocation for readahead.

Why?  Does it not ramp up quickly enough for you?

I see a lot of patches from you that say what the patch does but very
little about why.  Please explain yourself more.

On 11/21, Matthew Wilcox wrote:
> On Fri, Nov 21, 2025 at 01:48:11AM +0000, Jaegeuk Kim wrote:
> > This patch adds a desired folio order in file_ra_state in order to get
> > non-zero-order allocation for readahead.
> 
> Why?  Does it not ramp up quickly enough for you?
> 
> I see a lot of patches from you that say what the patch does but very
> little about why.  Please explain yourself more.

Yeah, let me revisit the patches and post another version that uplifts
code into mm/fs with what I'm trying to address.