[PATCH] Btrfs: set dead flag on the right root when destroying snapshot

[PATCH] Btrfs: set dead flag on the right root when destroying snapshot

[Filipe David Borba Manana]

We were setting the BTRFS_ROOT_SUBVOL_DEAD flag on the root of the
parent of our target snapshot, instead of setting it in the target
snapshot's root.

This is easy to observe by running the following scenario:

    mkfs.btrfs -f /dev/sdd
    mount /dev/sdd /mnt

    btrfs subvolume create /mnt/first_subvol
    btrfs subvolume snapshot -r /mnt /mnt/mysnap1

    btrfs subvolume delete /mnt/first_subvol
    btrfs subvolume snapshot -r /mnt /mnt/mysnap2

    btrfs send -p /mnt/mysnap1 /mnt/mysnap2 -f /tmp/send.data

The send command failed because the send ioctl returned -EPERM.
A test case for xfstests follows.

Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
---
 fs/btrfs/ioctl.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 362720a..482cad5 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -2314,7 +2314,7 @@ static noinline int btrfs_ioctl_snap_destroy(struct file *file,
 	spin_lock(&dest->root_item_lock);
 	root_flags = btrfs_root_flags(&root->root_item);
 	if (root->send_in_progress == 0) {
-		btrfs_set_root_flags(&root->root_item,
+		btrfs_set_root_flags(&dest->root_item,
 				root_flags | BTRFS_ROOT_SUBVOL_DEAD);
 		spin_unlock(&dest->root_item_lock);
 	} else {
@@ -2417,7 +2417,7 @@ out_unlock:
 	if (err) {
 		spin_lock(&dest->root_item_lock);
 		root_flags = btrfs_root_flags(&root->root_item);
-		btrfs_set_root_flags(&root->root_item,
+		btrfs_set_root_flags(&dest->root_item,
 				root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
 		spin_unlock(&dest->root_item_lock);
 	}
-- 
1.9.1

[PATCH v2] Btrfs: set dead flag on the right root when destroying snapshot

[Filipe David Borba Manana]

We were setting the BTRFS_ROOT_SUBVOL_DEAD flag on the root of the
parent of our target snapshot, instead of setting it in the target
snapshot's root.

This is easy to observe by running the following scenario:

    mkfs.btrfs -f /dev/sdd
    mount /dev/sdd /mnt

    btrfs subvolume create /mnt/first_subvol
    btrfs subvolume snapshot -r /mnt /mnt/mysnap1

    btrfs subvolume delete /mnt/first_subvol
    btrfs subvolume snapshot -r /mnt /mnt/mysnap2

    btrfs send -p /mnt/mysnap1 /mnt/mysnap2 -f /tmp/send.data

The send command failed because the send ioctl returned -EPERM.
A test case for xfstests follows.

Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
---

V2: Add missing replacements of 'root' with 'dest'.

 fs/btrfs/ioctl.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 362720a..38f2169 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -2312,16 +2312,16 @@ static noinline int btrfs_ioctl_snap_destroy(struct file *file,
 	 * again is not run concurrently.
 	 */
 	spin_lock(&dest->root_item_lock);
-	root_flags = btrfs_root_flags(&root->root_item);
-	if (root->send_in_progress == 0) {
-		btrfs_set_root_flags(&root->root_item,
+	root_flags = btrfs_root_flags(&dest->root_item);
+	if (dest->send_in_progress == 0) {
+		btrfs_set_root_flags(&dest->root_item,
 				root_flags | BTRFS_ROOT_SUBVOL_DEAD);
 		spin_unlock(&dest->root_item_lock);
 	} else {
 		spin_unlock(&dest->root_item_lock);
 		btrfs_warn(root->fs_info,
 			"Attempt to delete subvolume %llu during send",
-			root->root_key.objectid);
+			dest->root_key.objectid);
 		err = -EPERM;
 		goto out_dput;
 	}
@@ -2416,8 +2416,8 @@ out_up_write:
 out_unlock:
 	if (err) {
 		spin_lock(&dest->root_item_lock);
-		root_flags = btrfs_root_flags(&root->root_item);
-		btrfs_set_root_flags(&root->root_item,
+		root_flags = btrfs_root_flags(&dest->root_item);
+		btrfs_set_root_flags(&dest->root_item,
 				root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
 		spin_unlock(&dest->root_item_lock);
 	}
-- 
1.9.1

Re: [PATCH v2] Btrfs: set dead flag on the right root when destroying snapshot

[David Sterba]

On Sun, May 25, 2014 at 03:55:44AM +0100, Filipe David Borba Manana wrote:
> We were setting the BTRFS_ROOT_SUBVOL_DEAD flag on the root of the
> parent of our target snapshot, instead of setting it in the target
> snapshot's root.
> 
> This is easy to observe by running the following scenario:
> 
>     mkfs.btrfs -f /dev/sdd
>     mount /dev/sdd /mnt
> 
>     btrfs subvolume create /mnt/first_subvol
>     btrfs subvolume snapshot -r /mnt /mnt/mysnap1
> 
>     btrfs subvolume delete /mnt/first_subvol
>     btrfs subvolume snapshot -r /mnt /mnt/mysnap2
> 
>     btrfs send -p /mnt/mysnap1 /mnt/mysnap2 -f /tmp/send.data
> 
> The send command failed because the send ioctl returned -EPERM.
> A test case for xfstests follows.
> 
> Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>

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

Thanks for catching it, I was so focused on verifying the locks to do
what they're supposed to do and missed the typo, not that the variable
names help to avoid confusion.