bpf, devmap: Fix premature entry free on destroying map

dev_map_free() waits for flush_needed bitmap to be empty in order to
ensure all flush operations have completed before freeing its entries.
However the corresponding clear_bit() was called before using the
entries, so the entries could be used after free.

All access to the entries needs to be done before clearing the bit.
It seems commit a5e2da6 ("bpf: netdev is never null in
__dev_map_flush") accidentally changed the clear_bit() and memory access
order.

Note that the problem happens only in __dev_map_flush(), not in
dev_map_flush_old(). dev_map_flush_old() is called only after nulling
out the corresponding netdev_map entry, so dev_map_free() never frees
the entry thus no such race happens there.

Fixes: a5e2da6 ("bpf: netdev is never null in __dev_map_flush")
Signed-off-by: Toshiaki Makita <toshiaki.makita1@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

kernel/bpf/devmap.c

@@ -291,10 +291,10 @@ void __dev_map_flush(struct bpf_map *map)
 		if (unlikely(!dev))
 			continue;
 
-		__clear_bit(bit, bitmap);
-
 		bq = this_cpu_ptr(dev->bulkq);
 		bq_xmit_all(dev, bq, XDP_XMIT_FLUSH, true);
+
+		__clear_bit(bit, bitmap);
 	}
 }