minor sb clearing improvements

minor sb clearing improvements

[Christoph Hellwig]

Hi all,

this series moves btrfs off the deprecated write_one_page API that
requires ->writepage to be implemented (in this case on the block
device inode), and while doing so also avoids a pointless sb read
for zoned file systems.

Diffstat:
 volumes.c |   50 +++++++++++++++++++++++++-------------------------
 1 file changed, 25 insertions(+), 25 deletions(-)

[PATCH 1/2] btrfs: don't read the disk superblock for zoned devices in btrfs_scratch_superblocks

[Christoph Hellwig]

For zoned devices, btrfs_scratch_superblocks just resets the sb zones,
which means there is no need to even read the previous superblock.
Split the code to read, zero and write the superblock for convention
devices into a separate helper so that it isn't called for zoned
devices.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/btrfs/volumes.c | 51 +++++++++++++++++++++++-----------------------
 1 file changed, 26 insertions(+), 25 deletions(-)

diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index b386cb4b9aaa1..2dd7d2c5b0d80 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -2005,42 +2005,43 @@ static u64 btrfs_num_devices(struct btrfs_fs_info *fs_info)
 	return num_devices;
 }
 
+static void btrfs_scratch_superblock(struct btrfs_fs_info *fs_info,
+				     struct block_device *bdev, int copy_num)
+{
+	struct btrfs_super_block *disk_super;
+	struct page *page;
+	int ret;
+
+	disk_super = btrfs_read_dev_one_super(bdev, copy_num, false);
+	if (IS_ERR(disk_super))
+		return;
+	memset(&disk_super->magic, 0, sizeof(disk_super->magic));
+	page = virt_to_page(disk_super);
+	set_page_dirty(page);
+	lock_page(page);
+	/* write_on_page() unlocks the page */
+	ret = write_one_page(page);
+	if (ret)
+		btrfs_warn(fs_info,
+			"error clearing superblock number %d (%d)",
+			copy_num, ret);
+	btrfs_release_disk_super(disk_super);
+}
+
 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info,
 			       struct block_device *bdev,
 			       const char *device_path)
 {
-	struct btrfs_super_block *disk_super;
 	int copy_num;
 
 	if (!bdev)
 		return;
 
 	for (copy_num = 0; copy_num < BTRFS_SUPER_MIRROR_MAX; copy_num++) {
-		struct page *page;
-		int ret;
-
-		disk_super = btrfs_read_dev_one_super(bdev, copy_num, false);
-		if (IS_ERR(disk_super))
-			continue;
-
-		if (bdev_is_zoned(bdev)) {
+		if (bdev_is_zoned(bdev))
 			btrfs_reset_sb_log_zones(bdev, copy_num);
-			continue;
-		}
-
-		memset(&disk_super->magic, 0, sizeof(disk_super->magic));
-
-		page = virt_to_page(disk_super);
-		set_page_dirty(page);
-		lock_page(page);
-		/* write_on_page() unlocks the page */
-		ret = write_one_page(page);
-		if (ret)
-			btrfs_warn(fs_info,
-				"error clearing superblock number %d (%d)",
-				copy_num, ret);
-		btrfs_release_disk_super(disk_super);
-
+		else
+			btrfs_scratch_superblock(fs_info, bdev, copy_num);
 	}
 
 	/* Notify udev that device has changed */
-- 
2.30.2

Re: [PATCH 1/2] btrfs: don't read the disk superblock for zoned devices in btrfs_scratch_superblocks

[Johannes Thumshirn]

On 21.11.22 18:48, Christoph Hellwig wrote:
> Split the code to read, zero and write the superblock for convention
                                              conventional ~^

Otherwise,
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>

[PATCH 2/2] btrfs: stop using write_one_page in btrfs_scratch_superblock

[Christoph Hellwig]

write_one_page is an awkward interface that expects the page locked
and ->writepage to be implemented.  Just mark the sb dirty, put
the page and then call the proper bdev helper to sync the range.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/btrfs/volumes.c | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)

diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 2dd7d2c5b0d80..ddf172ba67972 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -2009,23 +2009,22 @@ static void btrfs_scratch_superblock(struct btrfs_fs_info *fs_info,
 				     struct block_device *bdev, int copy_num)
 {
 	struct btrfs_super_block *disk_super;
-	struct page *page;
+	const size_t len = sizeof(disk_super->magic);
+	u64 bytenr = btrfs_sb_offset(copy_num);
 	int ret;
 
-	disk_super = btrfs_read_dev_one_super(bdev, copy_num, false);
+	disk_super = btrfs_read_disk_super(bdev, bytenr, bytenr);
 	if (IS_ERR(disk_super))
 		return;
-	memset(&disk_super->magic, 0, sizeof(disk_super->magic));
-	page = virt_to_page(disk_super);
-	set_page_dirty(page);
-	lock_page(page);
-	/* write_on_page() unlocks the page */
-	ret = write_one_page(page);
+	memset(&disk_super->magic, 0, len);
+	set_page_dirty(virt_to_page(disk_super));
+	btrfs_release_disk_super(disk_super);
+
+	ret = sync_blockdev_range(bdev, bytenr, bytenr + len - 1);
 	if (ret)
 		btrfs_warn(fs_info,
 			"error clearing superblock number %d (%d)",
 			copy_num, ret);
-	btrfs_release_disk_super(disk_super);
 }
 
 void btrfs_scratch_superblocks(struct btrfs_fs_info *fs_info,
-- 
2.30.2

Re: minor sb clearing improvements

[David Sterba]

On Mon, Nov 21, 2022 at 06:47:47PM +0100, Christoph Hellwig wrote:
> Hi all,
> 
> this series moves btrfs off the deprecated write_one_page API that
> requires ->writepage to be implemented (in this case on the block
> device inode), and while doing so also avoids a pointless sb read
> for zoned file systems.
> 
> Diffstat:
>  volumes.c |   50 +++++++++++++++++++++++++-------------------------
>  1 file changed, 25 insertions(+), 25 deletions(-)

Added to misc-next, thanks.