[PATCH] Btrfs: all super blocks of the replaced disk must be scratched

[PATCH] Btrfs: all super blocks of the replaced disk must be scratched

[Anand Jain]

In a normal scenario when sys-admin replaces a disk, the
expeted is btrfs will release the disk completely.

However the below test case gives a wrong impression that
replaced disk is still is in use.

$ btrfs rep start /dev/sde /dev/sdg4 /btrfs
$ mkfs.btrfs /dev/sde
/dev/sde appears to contain an existing filesystem (btrfs).
Error: Use the -f option to force overwrite.

---
 fs/btrfs/volumes.c | 33 +++++++++++++++++++++++++--------
 1 file changed, 25 insertions(+), 8 deletions(-)

diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 3c44800..06717a5 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -6286,16 +6286,33 @@ int btrfs_scratch_superblock(struct btrfs_device *device)
 {
 	struct buffer_head *bh;
 	struct btrfs_super_block *disk_super;
+	int i;
+	u64 bytenr;
 
-	bh = btrfs_read_dev_super(device->bdev);
-	if (!bh)
-		return -EINVAL;
-	disk_super = (struct btrfs_super_block *)bh->b_data;
+	for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
+		bytenr = btrfs_sb_offset(i);
+		if (bytenr + BTRFS_SUPER_INFO_SIZE >=
+				i_size_read(device->bdev->bd_inode))
+			break;
 
-	memset(&disk_super->magic, 0, sizeof(disk_super->magic));
-	set_buffer_dirty(bh);
-	sync_dirty_buffer(bh);
-	brelse(bh);
+		bh = __bread(device->bdev, bytenr / 4096,
+					BTRFS_SUPER_INFO_SIZE);
+		if (!bh)
+			continue;
+
+		disk_super = (struct btrfs_super_block *)bh->b_data;
+		if (btrfs_super_bytenr(disk_super) != bytenr ||
+				btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
+			brelse(bh);
+			continue;
+		}
+
+		memset(&disk_super->magic, 0, sizeof(disk_super->magic));
+
+		set_buffer_dirty(bh);
+		sync_dirty_buffer(bh);
+		brelse(bh);
+	}
 
 	return 0;
 }
-- 
1.8.5.3

Re: [PATCH] Btrfs: all super blocks of the replaced disk must be scratched

[David Sterba]

On Mon, Mar 17, 2014 at 07:58:06PM +0800, Anand Jain wrote:
> In a normal scenario when sys-admin replaces a disk, the
> expeted is btrfs will release the disk completely.
> 
> However the below test case gives a wrong impression that
> replaced disk is still is in use.
> 
> $ btrfs rep start /dev/sde /dev/sdg4 /btrfs
> $ mkfs.btrfs /dev/sde
> /dev/sde appears to contain an existing filesystem (btrfs).
> Error: Use the -f option to force overwrite.

(your signed-off is missing)

Reviewed-by: David Sterba <dsterba@suse.cz>

I've noticed that btrfs_rm_device also cleans only the first superblock
signature, so it should do the same.

I can't think of any drawback where all the superblocks are scratched,
should be safe.

Re: [PATCH] Btrfs: all super blocks of the replaced disk must be scratched

[Anand Jain]

On 21/03/14 02:21, David Sterba wrote:
> On Mon, Mar 17, 2014 at 07:58:06PM +0800, Anand Jain wrote:
>> In a normal scenario when sys-admin replaces a disk, the
>> expeted is btrfs will release the disk completely.
>>
>> However the below test case gives a wrong impression that
>> replaced disk is still is in use.
>>
>> $ btrfs rep start /dev/sde /dev/sdg4 /btrfs
>> $ mkfs.btrfs /dev/sde
>> /dev/sde appears to contain an existing filesystem (btrfs).
>> Error: Use the -f option to force overwrite.
>
> (your signed-off is missing)
>
> Reviewed-by: David Sterba <dsterba@suse.cz>
>
> I've noticed that btrfs_rm_device also cleans only the first superblock
> signature, so it should do the same.

  A separate patch was generated for this
    btrfs: btrfs_rm_device() should zero mirror SB as well


Thanks, Anand


> I can't think of any drawback where all the superblocks are scratched,
> should be safe.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>