Re: [PATCH bpf-next 5/6] bpf: Improve tracing recursion prevention mechanism

[Alexei Starovoitov]

On Mon, Apr 17, 2023 at 6:49 PM Yafang Shao <laoar.shao@xxxxxxxxx> wrote:
>
> On Tue, Apr 18, 2023 at 4:15 AM Alexei Starovoitov
> <alexei.starovoitov@xxxxxxxxx> wrote:
> >
> > On Mon, Apr 17, 2023 at 03:47:36PM +0000, Yafang Shao wrote:
> > > diff --git a/kernel/bpf/trampoline.c b/kernel/bpf/trampoline.c
> > > index f61d513..3df39a5 100644
> > > --- a/kernel/bpf/trampoline.c
> > > +++ b/kernel/bpf/trampoline.c
> > > @@ -842,15 +842,21 @@ static __always_inline u64 notrace bpf_prog_start_time(void)
> > >  static u64 notrace __bpf_prog_enter_recur(struct bpf_prog *prog, struct bpf_tramp_run_ctx *run_ctx)
> > >       __acquires(RCU)
> > >  {
> > > -     rcu_read_lock();
> > > -     migrate_disable();
> > > -
> > > -     run_ctx->saved_run_ctx = bpf_set_run_ctx(&run_ctx->run_ctx);
> > > +     int bit;
> > >
> > > -     if (unlikely(this_cpu_inc_return(*(prog->active)) != 1)) {
> > > +     rcu_read_lock();
> > > +     bit = test_recursion_try_acquire(_THIS_IP_, _RET_IP_);
> >
> > and bpf will prevent ftrace to run and vice versa.
> > Not a good idea.
> >
> > One bpf prog will prevent different bpf prog to run since they share current task.
> > Not a good idea either.
>
> That shouldn't happen. test_recursion_try_acquire() uses a
> per-task_struct value. One single task_struct can't run in parallel,
> right?
> Note that the bpf program running in softirq or irq context won't be
> prevented by it.
> IIUC, the bpf program should run in serial in one single task, right?
> That said, one bpf program can only run after another bpf program
> finished in the same task?

bpf progs can nest in the same task.