On Sun, Jun 11, 2023 at 10:12 PM Miklos Szeredi <miklos@xxxxxxxxxx> wrote: > > On Sun, 11 Jun 2023 at 19:52, Amir Goldstein <amir73il@xxxxxxxxx> wrote: > > > Is it because getting f_real_path() and file_dentry() via d_real() > > is more expensive? > > and caching this information in file_fake container would be > > more efficient? > > > > I will restore the file_fake container and post v3. > > I simply dislike the fact that ->d_real() is getting more complex. > I'd prefer d_real to die, which is unfortunately not so easy, as > you've explained. > > But if we can make it somewhat less complex (remove the inode > parameter) instead of more complex (add a vfsmount * parameter) then > that's already a big win in my eyes. > OK, I can relate to that. Here is file_fake restored, now with fsnotify fix also tested: https://github.com/amir73il/linux/commits/ovl_fake_path IIUC, you would now want to change file_dentry(f) to using f_real_file(f)->dentry and get rid of the inode argument to d_real(). Do you want that change also in v3 or should we save some fun for later and just fix fsnotify for now? Thanks, Amir.
