6.1-stable review patch. If anyone has any objections, please let me know. ------------------ From: Omar Sandoval <osandov@xxxxxx> commit 3324d0547861b16cf436d54abba7052e0c8aa9de upstream. Sweet Tea spotted a race between subvolume deletion and snapshotting that can result in the root item for the snapshot having the BTRFS_ROOT_SUBVOL_DEAD flag set. The race is: Thread 1 | Thread 2 ----------------------------------------------|---------- btrfs_delete_subvolume | btrfs_set_root_flags(BTRFS_ROOT_SUBVOL_DEAD)| |btrfs_mksubvol | down_read(subvol_sem) | create_snapshot | ... | create_pending_snapshot | copy root item from source down_write(subvol_sem) | This flag is only checked in send and swap activate, which this would cause to fail mysteriously. create_snapshot() now checks the root refs to reject a deleted subvolume, so we can fix this by locking subvol_sem earlier so that the BTRFS_ROOT_SUBVOL_DEAD flag and the root refs are updated atomically. CC: stable@xxxxxxxxxxxxxxx # 4.14+ Reported-by: Sweet Tea Dorminy <sweettea-kernel@xxxxxxxxxx> Reviewed-by: Sweet Tea Dorminy <sweettea-kernel@xxxxxxxxxx> Reviewed-by: Anand Jain <anand.jain@xxxxxxxxxx> Signed-off-by: Omar Sandoval <osandov@xxxxxx> Reviewed-by: David Sterba <dsterba@xxxxxxxx> Signed-off-by: David Sterba <dsterba@xxxxxxxx> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> --- fs/btrfs/inode.c | 22 +++++++++++++--------- 1 file changed, 13 insertions(+), 9 deletions(-) --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -4689,6 +4689,8 @@ int btrfs_delete_subvolume(struct inode u64 root_flags; int ret; + down_write(&fs_info->subvol_sem); + /* * Don't allow to delete a subvolume with send in progress. This is * inside the inode lock so the error handling that has to drop the bit @@ -4700,25 +4702,25 @@ int btrfs_delete_subvolume(struct inode btrfs_warn(fs_info, "attempt to delete subvolume %llu during send", dest->root_key.objectid); - return -EPERM; + ret = -EPERM; + goto out_up_write; } if (atomic_read(&dest->nr_swapfiles)) { spin_unlock(&dest->root_item_lock); btrfs_warn(fs_info, "attempt to delete subvolume %llu with active swapfile", root->root_key.objectid); - return -EPERM; + ret = -EPERM; + goto out_up_write; } 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); - down_write(&fs_info->subvol_sem); - ret = may_destroy_subvol(dest); if (ret) - goto out_up_write; + goto out_undead; btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP); /* @@ -4728,7 +4730,7 @@ int btrfs_delete_subvolume(struct inode */ ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 5, true); if (ret) - goto out_up_write; + goto out_undead; trans = btrfs_start_transaction(root, 0); if (IS_ERR(trans)) { @@ -4794,15 +4796,17 @@ out_end_trans: inode->i_flags |= S_DEAD; out_release: btrfs_subvolume_release_metadata(root, &block_rsv); -out_up_write: - up_write(&fs_info->subvol_sem); +out_undead: if (ret) { spin_lock(&dest->root_item_lock); 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); - } else { + } +out_up_write: + up_write(&fs_info->subvol_sem); + if (!ret) { d_invalidate(dentry); btrfs_prune_dentries(dest); ASSERT(dest->send_in_progress == 0);