On Tue, Jan 7, 2020 at 7:43 PM Josef Bacik <josef@xxxxxxxxxxxxxx> wrote:
>
> We were using btrfs_i_size_write(), which unconditionally jacks up
> inode->disk_i_size. However since clone can operate on ranges we could
> have pending ordered extents for a range prior to the start of our clone
> operation and thus increase disk_i_size too far and have a hole with no
> file extent.
>
> Fix this by using the btrfs_ordered_update_i_size helper which will do
> the right thing in the face of pending ordered extents outside of our
> clone range.
>
> Signed-off-by: Josef Bacik <josef@xxxxxxxxxxxxxx>
Reviewed-by: Filipe Manana <fdmanana@xxxxxxxx>
Looks good, thanks.
> ---
> fs/btrfs/ioctl.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
> index 8ec61f3f0291..291dda3b6547 100644
> --- a/fs/btrfs/ioctl.c
> +++ b/fs/btrfs/ioctl.c
> @@ -3332,8 +3332,10 @@ static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
> */
> if (endoff > destoff + olen)
> endoff = destoff + olen;
> - if (endoff > inode->i_size)
> - btrfs_i_size_write(BTRFS_I(inode), endoff);
> + if (endoff > inode->i_size) {
> + i_size_write(inode, endoff);
> + btrfs_ordered_update_i_size(inode, endoff, NULL);
> + }
>
> ret = btrfs_update_inode(trans, root, inode);
> if (ret) {
> --
> 2.23.0
>
--
Filipe David Manana,
“Whether you think you can, or you think you can't — you're right.”
