Re: [PATCH PATCH v2 v2 2/6] bpf: Add bpf_perf_event_aux_pause kfunc

[Leo Yan]

Hi Eduard,

On Mon, Jul 21, 2025 at 03:38:59PM -0700, Eduard Zingerman wrote:
> On Fri, 2025-07-18 at 16:25 +0100, Leo Yan wrote:
> 
> [...]
> 
> > +__bpf_kfunc int bpf_perf_event_aux_pause(void *p__map, u64 flags, u32 pause)
> > +{
> > +	struct bpf_map *map = p__map;
> > +	struct bpf_array *array = container_of(map, struct bpf_array, map);
> 
> Verifier makes sure that p__map is a not null pointer to an object of
> type bpf_map, but it does not guarantee that the object is an instance
> of bpf_array.
> You need to check map->type, same way bpf_arena_alloc_pages() does.

Makes sense. Will do.

> > +	unsigned int cpu = smp_processor_id();
> > +	u64 index = flags & BPF_F_INDEX_MASK;
> > +	struct bpf_event_entry *ee;
> > +	int ret = 0;
> > +
> > +	/* Disabling IRQ avoids race condition with perf event flows. */
> > +	guard(irqsave)();
> > +
> > +	if (unlikely(flags & ~(BPF_F_INDEX_MASK))) {
> > +		ret = -EINVAL;
> > +		goto out;
> > +	}
> > +
> > +	if (index == BPF_F_CURRENT_CPU)
> > +		index = cpu;
> > +
> > +	if (unlikely(index >= array->map.max_entries)) {
> > +		ret = -E2BIG;
> > +		goto out;
> > +	}
> > +
> > +	ee = READ_ONCE(array->ptrs[index]);
> > +	if (!ee) {
> > +		ret = -ENOENT;
> > +		goto out;
> > +	}
> > +
> > +	if (!has_aux(ee->event))
> > +		ret = -EINVAL;

I should refactor a bit for removing "goto out" and directly return
error cases.

Thanks for review and your suggestion in another reply.

Leo

> > +
> > +	perf_event_aux_pause(ee->event, pause);
> > +out:
> > +	return ret;
> > +}
> 
> [...]