Skip to contentSkip to conversation

Add readpages support for block devices

PATCH In review Matthew Wilcox 3 participants 3 messages Dec 18, 2010 – Dec 19, 2010 VFS

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Add readpages support for block devices

From: Shane Michael Matthews <[email protected]>

Using mpage_readpages instead of multiple calls to blkdev_readpage reduces the CPU utilisation by 35% when using dd to clone a drive.

---
 fs/block_dev.c |    7 +++++++
 1 files changed, 7 insertions(+), 0 deletions(-)
fs/block_dev.c +7 -0
index 4230252..55cf932 100644
@@ -318,6 +318,12 @@ static int blkdev_readpage(struct file * file, struct page * page)
 	return block_read_full_page(page, blkdev_get_block);
 }
 
+static int blkdev_readpages(struct file *file, struct address_space *mapping,
+			struct list_head *pages, unsigned nr_pages)
+{
+	return mpage_readpages(mapping, pages, nr_pages, blkdev_get_block);
+}
+
 static int blkdev_write_begin(struct file *file, struct address_space *mapping,
 			loff_t pos, unsigned len, unsigned flags,
 			struct page **pagep, void **fsdata)
@@ -1634,6 +1640,7 @@ static int blkdev_releasepage(struct page *page, gfp_t wait)
 
 static const struct address_space_operations def_blk_aops = {
 	.readpage	= blkdev_readpage,
+	.readpages	= blkdev_readpages,
 	.writepage	= blkdev_writepage,
 	.sync_page	= block_sync_page,
 	.write_begin	= blkdev_write_begin,
Signature · 1 line
1.7.2.3
#1 lore raw

We already used to have this and Jens reverted it in commit 172124e220f1854acc99ee394671781b8b5e2120. I can't find a better explanation than his rather brief commit message, but he might not more.

#2 lore raw
Quoted 3 lines · Christoph Hellwig

We already used to have this and Jens reverted it in commit 172124e220f1854acc99ee394671781b8b5e2120. I can't find a better explanation than his rather brief commit message, but he might not more.

The discussion goes here

http://www.gossamer-threads.com/lists/linux/kernel/1085325?do=post_view_threaded#1085325

The problem happens on ext3 fs, non-4k aligned partition and 64K page size. Jeff also find it hurts dump performance. Then it's reverted since no one seems to have the time to dig further.

Thanks,
Fengguang