On Thu, 28 Apr 2022 at 13:30, Miklos Szeredi <miklos@xxxxxxxxxx> wrote: > So I guess the proper fix would be to introduce a version of > lookup_one_len() without inode_permission()... OTOH, we do have CAP_DAC_READ_SEARCH already in the syscall path and knfsd won't be using mnt_userns, so just passing init_user_ns should be fine as a quick fix. I'm in the process of applying these patches, so if there's no objection, I'll make this change. Thanks, Miklos
