On Mon, Apr 23, 2018 at 12:21 PM, Miklos Szeredi <miklos@xxxxxxxxxx> wrote: > We could fix it to use file_inode()->i_sb instead of > f_path.dentry->d_sb after reverting the overlay specific hack, and > that would fix the freeze bypass bug and would be correct for all > filesystems. But I wonder how many such issues we have where > discrepancy between f_path.dentry and file_inode() matters. OTOH, any such issues should already have surfaced. Just need to be careful when reverting VFS patches that they don't regress in the face of f_path containing the overlay path. So we are probably better to keep using overlay path in real file universally, just to keep the code simpler. Thanks, Miklos
