On Thu, 23 Jan 2020, KP Singh wrote: > > > If you want to put mutable hook handling in the infrastructure > > you need to make it general mutable hook handling as opposed to > > BPF hook handling. I don't know if that would be acceptable for > > all the reasons called out about dynamic module loading. > > We can have generic mutable hook handling and if an LSM doesn't --> provide a mutable security_hook_heads, it would not allow dynamic > hooks / dynamic module loading. > > So, in practice it will just be the BPF LSM that allows mutable hooks > and the other existing LSMs won't. I guess it will be cleaner than > calling the BPF hooks directly from the LSM code (i.e in security.c) I'm inclined to only have mutable hooks for KRSI, not for all LSMs. This is a special case and we don't need to provide this for anyone else. Btw, folks, PLEASE trim replies. -- James Morris <jmorris@xxxxxxxxx>
