Re: [PATCH bpf-next 2/3] bpf: add BPF_PROG_TEST_RUN support for flow dissector

[Stanislav Fomichev <sdf@fomichev.me>]

On 01/23, Alexei Starovoitov wrote:
> On Tue, Jan 22, 2019 at 01:23:14PM -0800, Stanislav Fomichev wrote:
> > The input is packet data, the output is struct bpf_flow_key. This should
> > make it easy to test flow dissector programs without elaborate
> > setup.
> > 
> > Signed-off-by: Stanislav Fomichev <sdf@google.com>
> > ---
> >  include/linux/bpf.h |  3 ++
> >  net/bpf/test_run.c  | 75 ++++++++++++++++++++++++++++++++++++++++++++-
> >  net/core/filter.c   |  1 +
> >  3 files changed, 78 insertions(+), 1 deletion(-)
> > 
> > diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> > index e734f163bd0b..701ef954a258 100644
> > --- a/include/linux/bpf.h
> > +++ b/include/linux/bpf.h
> > @@ -397,6 +397,9 @@ int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
> >  			  union bpf_attr __user *uattr);
> >  int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
> >  			  union bpf_attr __user *uattr);
> > +int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
> > +				     const union bpf_attr *kattr,
> > +				     union bpf_attr __user *uattr);
> >  
> >  /* an array of programs to be executed under rcu_lock.
> >   *
> > diff --git a/net/bpf/test_run.c b/net/bpf/test_run.c
> > index fa2644d276ef..ecad72885f23 100644
> > --- a/net/bpf/test_run.c
> > +++ b/net/bpf/test_run.c
> > @@ -16,12 +16,26 @@
> >  static __always_inline u32 bpf_test_run_one(struct bpf_prog *prog, void *ctx,
> >  		struct bpf_cgroup_storage *storage[MAX_BPF_CGROUP_STORAGE_TYPE])
> >  {
> > +	struct bpf_skb_data_end *cb;
> > +	struct sk_buff *skb;
> >  	u32 ret;
> >  
> >  	preempt_disable();
> >  	rcu_read_lock();
> >  	bpf_cgroup_storage_set(storage);
> > -	ret = BPF_PROG_RUN(prog, ctx);
> > +
> > +	switch (prog->type) {
> > +	case BPF_PROG_TYPE_FLOW_DISSECTOR:
> > +		skb = (struct sk_buff *)ctx;
> > +		cb = (struct bpf_skb_data_end *)skb->cb;
> > +		ret = __skb_flow_bpf_dissect(prog, ctx, &flow_keys_dissector,
> > +					     cb->qdisc_cb.flow_keys);
> > +		break;
> > +	default:
> > +		ret = BPF_PROG_RUN(prog, ctx);
> > +		break;
> > +	}
> 
> What is the benefit of this minimal code reuse?
> It seems to me bpf_test_run_one() gets slower for all,
> since prog type needs to be checked before every prog run.
> The point of bpf_prog_ops->test_run was to avoid this overhead.
> Are you somehow expecting flow_dissector prog using cgroup local storage?
> I don't think that's possible.
Agreed. Initially I didn't want to re-implement the logic of
bpf_test_run. But now that you mention that it's mostly about cgroup local
storage, I think I can do a simple loop inside of
bpf_prog_test_run_flow_dissector. Thanks, will follow up with another
series!