Re: [PATCHv7 bpf-next 1/3] xdp: add a new helper for dev map multicast support

On Tue, Jul 14, 2020 at 11:52:14PM +0200, Toke Høiland-Jørgensen wrote:

quoted

+bool dev_in_exclude_map(struct bpf_dtab_netdev *obj, struct bpf_map *map,
+			int exclude_ifindex)
+{
+	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
+	struct bpf_dtab_netdev *dev;
+	struct hlist_head *head;
+	int i = 0;
+
+	if (obj->dev->ifindex == exclude_ifindex)
+		return true;
+
+	if (!map || map->map_type != BPF_MAP_TYPE_DEVMAP_HASH)
+		return false;

The map type should probably be checked earlier and the whole operation
aborted if it is wrong...

Yes, I have already checked it in the helper, there should no need to double
check. I will remove this check.

quoted

+
+	for (; i < dtab->n_buckets; i++) {
+		head = dev_map_index_hash(dtab, i);
+
+		dev = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
+					    struct bpf_dtab_netdev,
+					    index_hlist);
+
+		if (dev && dev->idx == exclude_ifindex)
+			return true;
+	}

This looks broken; why are you iterating through the buckets? Shouldn't
this just be something like:

return __dev_map_hash_lookup_elem(map, obj->dev->ifindex) != NULL;

Ah, yes, I forgot this. I will update the code.

Thanks
Hangbin