On Wed, Nov 4, 2020 at 12:03 PM KP Singh <kpsingh@xxxxxxxxxxxx> wrote:
>
> [...]
>
> > > Ahh. Yes. That should do it. Right now I don't see concerns with safety
> > > of the bpf_spin_lock in bpf_lsm progs.
> >
> > What about sleepable lsm hooks? Normally we wouldn't expect to sleep with
> > a spinlock held. Should we have a check to ensure programs bpf_spin_lock
> > are not also sleepable?
>
> Thanks. Yes, I added that to my patch:
>
> diff --git a/kernel/bpf/bpf_lsm.c b/kernel/bpf/bpf_lsm.c
> index 61f8cc52fd5b..93383df2140b 100644
> --- a/kernel/bpf/bpf_lsm.c
> +++ b/kernel/bpf/bpf_lsm.c
> @@ -63,6 +63,10 @@ bpf_lsm_func_proto(enum bpf_func_id func_id, const
> struct bpf_prog *prog)
> return &bpf_task_storage_get_proto;
> case BPF_FUNC_task_storage_delete:
> return &bpf_task_storage_delete_proto;
> + case BPF_FUNC_spin_lock:
> + return &bpf_spin_lock_proto;
> + case BPF_FUNC_spin_unlock:
> + return &bpf_spin_unlock_proto;
> default:
> return tracing_prog_func_proto(func_id, prog);
> }
> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> index 314018e8fc12..8892f7ba2041 100644
> --- a/kernel/bpf/verifier.c
> +++ b/kernel/bpf/verifier.c
> @@ -9739,6 +9739,23 @@ static int check_map_prog_compatibility(struct
> bpf_verifier_env *env,
> return -EINVAL;
> }
>
> + if (map_value_has_spin_lock(map)) {
> + if (prog_type == BPF_PROG_TYPE_SOCKET_FILTER) {
> + verbose(env, "socket filter progs cannot use
> bpf_spin_lock yet\n");
> + return -EINVAL;
> + }
> +
> + if (is_tracing_prog_type(prog_type)) {
> + verbose(env, "tracing progs cannot use
> bpf_spin_lock yet\n");
> + return -EINVAL;
> + }
> +
> + if (prog->aux->sleepable) {
> + verbose(env, "sleepable progs cannot use
> bpf_spin_lock\n");
I think this can still be "yet" as it's doable; we can disable/enable
preemption in the helpers
and then have the verifier track that no sleepable helper is called
when a spin lock is held.
I would, however, prefer if we do it in a subsequent patch.
- KP
> + return -EINVAL;
> + }
> + }
> +
