Skip to contentSkip to conversation

bcache: Use a folio

PATCH Applied [email protected] 3 participants 4 messages Jul 2, 2025

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 bcache: Use a folio

From: Matthew Wilcox (Oracle) <[email protected]>

Retrieve a folio from the page cache instead of a page. Removes a hidden call to compound_head(). Then be sure to call folio_put() instead of put_page() to release it. That doesn't save any calls to compound_head(), just moves them around.

---
 drivers/md/bcache/super.c | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)
drivers/md/bcache/super.c +11 -11
index 2ea490b9d370..1492c8552255 100644
@@ -168,14 +168,14 @@ static const char *read_super(struct cache_sb *sb, struct block_device *bdev,
 {
 	const char *err;
 	struct cache_sb_disk *s;
-	struct page *page;
+	struct folio *folio;
 	unsigned int i;
 
-	page = read_cache_page_gfp(bdev->bd_mapping,
-				   SB_OFFSET >> PAGE_SHIFT, GFP_KERNEL);
-	if (IS_ERR(page))
+	folio = mapping_read_folio_gfp(bdev->bd_mapping,
+			SB_OFFSET >> PAGE_SHIFT, GFP_KERNEL);
+	if (IS_ERR(folio))
 		return "IO error";
-	s = page_address(page) + offset_in_page(SB_OFFSET);
+	s = folio_address(folio) + offset_in_folio(folio, SB_OFFSET);
 
 	sb->offset		= le64_to_cpu(s->offset);
 	sb->version		= le64_to_cpu(s->version);
@@ -272,7 +272,7 @@ static const char *read_super(struct cache_sb *sb, struct block_device *bdev,
 	*res = s;
 	return NULL;
 err:
-	put_page(page);
+	folio_put(folio);
 	return err;
 }
 
@@ -1366,7 +1366,7 @@ static CLOSURE_CALLBACK(cached_dev_free)
 	mutex_unlock(&bch_register_lock);
 
 	if (dc->sb_disk)
-		put_page(virt_to_page(dc->sb_disk));
+		folio_put(virt_to_folio(dc->sb_disk));
 
 	if (dc->bdev_file)
 		fput(dc->bdev_file);
@@ -2216,7 +2216,7 @@ void bch_cache_release(struct kobject *kobj)
 		free_fifo(&ca->free[i]);
 
 	if (ca->sb_disk)
-		put_page(virt_to_page(ca->sb_disk));
+		folio_put(virt_to_folio(ca->sb_disk));
 
 	if (ca->bdev_file)
 		fput(ca->bdev_file);
@@ -2593,7 +2593,7 @@ static ssize_t register_bcache(struct kobject *k, struct kobj_attribute *attr,
 	if (!holder) {
 		ret = -ENOMEM;
 		err = "cannot allocate memory";
-		goto out_put_sb_page;
+		goto out_put_sb_folio;
 	}
 
 	/* Now reopen in exclusive mode with proper holder */
@@ -2667,8 +2667,8 @@ static ssize_t register_bcache(struct kobject *k, struct kobj_attribute *attr,
 
 out_free_holder:
 	kfree(holder);
-out_put_sb_page:
-	put_page(virt_to_page(sb_disk));
+out_put_sb_folio:
+	folio_put(virt_to_folio(sb_disk));
 out_blkdev_put:
 	if (bdev_file)
 		fput(bdev_file);
Signature · 1 line
2.39.5
maintainer
#1 lore raw
Quoted 6 lines · On 7/1/25 8:48 PM, [email protected] wrote:

From: Matthew Wilcox (Oracle) <[email protected]>

Retrieve a folio from the page cache instead of a page. Removes a hidden call to compound_head(). Then be sure to call folio_put() instead of put_page() to release it. That doesn't save any calls to compound_head(), just moves them around.

Really needs a better subject line... I can do that while applying, however. How about:

bcache: switch from pages to folios in read_super()

or something like that.

Signature · 1 line
Jens Axboe
#2 lore raw
Quoted 12 lines

2025年7月2日 21:49,Jens Axboe <[email protected]> 写道:

Quoted 6 lines · On 7/1/25 8:48 PM, [email protected] wrote:

From: Matthew Wilcox (Oracle) <[email protected]>

Retrieve a folio from the page cache instead of a page. Removes a hidden call to compound_head(). Then be sure to call folio_put() instead of put_page() to release it. That doesn't save any calls to compound_head(), just moves them around.

Quoted 4 lines

Really needs a better subject line... I can do that while applying, however. How about:

bcache: switch from pages to folios in read_super()

Yeah, this subject is good IMHO.

Thanks.

Coly Li

maintainer Applied
#3 lore raw
In reply to [email protected]
Quoted 6 lines · On Wed, 02 Jul 2025 10:48:48 +0800, [email protected] wrote:

Retrieve a folio from the page cache instead of a page. Removes a hidden call to compound_head(). Then be sure to call folio_put() instead of put_page() to release it. That doesn't save any calls to compound_head(), just moves them around.

Applied, thanks!

[1/1] bcache: Use a folio

      commit: 88b4b873551f0901dd0314d37ca2a420ec6c2686

Best regards,

Signature · 1 line
Jens Axboe