On Fri, Oct 28, 2016 at 5:25 PM, Miklos Szeredi <miklos@xxxxxxxxxx> wrote: > On Wed, Oct 26, 2016 at 9:34 PM, Amir Goldstein <amir73il@xxxxxxxxx> wrote: >> With overlayfs, it is wrong to compare file_inode(inode)->i_sb >> of regular files with those of non-regular files, because the >> former reference the real (upper/lower) sb and the latter reference >> the overlayfs sb. >> >> Move the test for same super block after the sanity tests for >> clone range of directory and non-regular file. > > Better: compare ->f_path.dentry->d_sb instead of file_inode()->i_sb. > We don't want to be mixing files that come from overlayfs and ones > that come from the underlying layers. That's not a good option. When source file is in lower and dest file is in upper, then clone range should go forward iff both lower and upper inodes are on the same sb. The test as it is checks for this properly. > > BTW, would it be worthwhile adding clone_file_range() support to overlayfs? > There is nothing to add. It works quite well because clone_file_range works on inode level. In fact, the entire 'clone' group of xfstests passes with overlayfs of lower and upper on the same XFS filesystem (with reflink support). Amir. -- To unsubscribe from this list: send the line "unsubscribe linux-unionfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
