Re: [RFC PATCH bpf-next 1/4] bpf: Allow trampoline re-attach

["Toke Høiland-Jørgensen" <toke@redhat.com>]

Jiri Olsa <jolsa@redhat.com> writes:

> On Sat, Apr 03, 2021 at 11:21:55AM -0700, Alexei Starovoitov wrote:
>> On Sat, Apr 03, 2021 at 01:24:12PM +0200, Toke Høiland-Jørgensen wrote:
>> > >  	if (!prog->aux->dst_trampoline && !tgt_prog) {
>> > > -		err = -ENOENT;
>> > > -		goto out_unlock;
>> > > +		/*
>> > > +		 * Allow re-attach for tracing programs, if it's currently
>> > > +		 * linked, bpf_trampoline_link_prog will fail.
>> > > +		 */
>> > > +		if (prog->type != BPF_PROG_TYPE_TRACING) {
>> > > +			err = -ENOENT;
>> > > +			goto out_unlock;
>> > > +		}
>> > > +		if (!prog->aux->attach_btf) {
>> > > +			err = -EINVAL;
>> > > +			goto out_unlock;
>> > > +		}
>> > 
>> > I'm wondering about the two different return codes here. Under what
>> > circumstances will aux->attach_btf be NULL, and why is that not an
>> > ENOENT error? :)
>> 
>> The feature makes sense to me as well.
>> I don't quite see how it would get here with attach_btf == NULL.
>> Maybe WARN_ON then?
>
> right, that should be always there
>
>> Also if we're allowing re-attach this way why exclude PROG_EXT and LSM?
>> 
>
> I was enabling just what I needed for the test, which is so far
> the only use case.. I'll see if I can enable that for all of them

How would that work? For PROG_EXT we clear the destination on the first
attach (to avoid keeping a ref on it), so re-attach can only be done
with an explicit target (which already works just fine)...

-Toke