Re: [PATCH net-next 2/3] bpf: introduce BPF_MAP_TYPE_STACK_TRACE

From: Peter Zijlstra
Date: Thu Feb 25 2016 - 11:50:22 EST


On Thu, Feb 25, 2016 at 08:42:40AM -0800, Alexei Starovoitov wrote:
> On 2/25/16 6:23 AM, Peter Zijlstra wrote:
> >>+ id = hash & (smap->n_buckets - 1);
> >Its not at all clear where the corresponding rcu_read_lock() is at.
> >
> >>>+ bucket = rcu_dereference(smap->buckets[id]);
>
> bpf programs of all types are always executing under rcu_read_lock().
> This is fundamental for maps and majority of the helpers
> that's why there is no WARN_ON(rcu_read_lock_held) in this helper,
> since we already have it in many other places.
> The rcu_read_lock() for kprobe type is in trace_call_bpf().

OK, just not clear from reading this patch in isolation I suppose.