[syzbot] [lsm?] general protection fault in hook_inode_free_security

[Christian Brauner]

> bypass?  Shouldn't we call all the inode_free_security() hooks in
> inode_free_by_rcu()?  That would mean to reserve an rcu_head and then
> probably use inode->i_rcu instead.

Note that you can't block in call_rcu(). From a cursory look at the
implementers of the hook it should be fine though.