net: Convert caif_net_ops

Init method just allocates memory for new cfg, and
assigns net_generic(net, caif_net_id). Despite there is
synchronize_rcu() on error path in cfcnfg_create(),
in real this function does not use global lists,
so it looks like this synchronize_rcu() is some legacy
inheritance. Exit method removes caif devices under
rtnl_lock().

There could be a problem, if someone from foreign net
pernet_operations dereference caif_net_id of this net.
It's dereferenced in get_cfcnfg() and caif_device_list().

get_cfcnfg() is used from netdevice notifiers, where
they are called under rtnl_lock(). The notifiers can't
be called from foreign nets pernet_operations. Also,
it's used from caif_disconnect_client() and from
caif_connect_client(). The both of the functions work
with caif socket, and there is the only possibility
to have a socket, when the net is dead. This may happen
only of the socket was created as kern using sk_alloc().
Grep by PF_CAIF shows we do not create kern caif sockets,
so get_cfcnfg() is safe.

caif_device_list() is used in netdevice notifiers and exit
method under rtnl lock. Also, from caif_get() used in
the netdev notifiers and in caif_flow_cb(). The last item
is skb destructor. Since there are no kernel caif sockets
nobody can send net a packet in parallel with init/exit,
so this is also safe.

So, these pernet_operations are safe to be async.

Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/net/caif/caif_dev.c b/net/caif/caif_dev.c
index e0adcd123f48a1a4f66028bbf731132ed8e7ff17..7a78268cc57242a5c3a26c963abb621dededeb6a 100644 (file)
--- a/net/caif/caif_dev.c
+++ b/net/caif/caif_dev.c
@@ -544,6 +544,7 @@ static struct pernet_operations caif_net_ops = {
        .exit = caif_exit_net,
        .id   = &caif_net_id,
        .size = sizeof(struct caif_net),
+       .async = true,
 };
 
 /* Initialize Caif devices list */