1// SPDX-License-Identifier: GPL-2.0-or-later
2/* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com>
3 */
4
5#include "ipvlan.h"
6
7static unsigned int ipvlan_netid __read_mostly;
8
9struct ipvlan_netns {
10	unsigned int ipvl_nf_hook_refcnt;
11};
12
13static struct ipvl_addr *ipvlan_skb_to_addr(struct sk_buff *skb,
14					    struct net_device *dev)
15{
16	struct ipvl_addr *addr = NULL;
17	struct ipvl_port *port;
18	int addr_type;
19	void *lyr3h;
20
21	if (!dev || !netif_is_ipvlan_port(dev))
22		goto out;
23
24	port = ipvlan_port_get_rcu(dev);
25	if (!port || port->mode != IPVLAN_MODE_L3S)
26		goto out;
27
28	lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
29	if (!lyr3h)
30		goto out;
31
32	addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
33out:
34	return addr;
35}
36
37static struct sk_buff *ipvlan_l3_rcv(struct net_device *dev,
38				     struct sk_buff *skb, u16 proto)
39{
40	struct ipvl_addr *addr;
41	struct net_device *sdev;
42
43	addr = ipvlan_skb_to_addr(skb, dev);
44	if (!addr)
45		goto out;
46
47	sdev = addr->master->dev;
48	switch (proto) {
49	case AF_INET:
50	{
51		struct iphdr *ip4h = ip_hdr(skb);
52		int err;
53
54		err = ip_route_input_noref(skb, ip4h->daddr, ip4h->saddr,
55					   ip4h->tos, sdev);
56		if (unlikely(err))
57			goto out;
58		break;
59	}
60#if IS_ENABLED(CONFIG_IPV6)
61	case AF_INET6:
62	{
63		struct dst_entry *dst;
64		struct ipv6hdr *ip6h = ipv6_hdr(skb);
65		int flags = RT6_LOOKUP_F_HAS_SADDR;
66		struct flowi6 fl6 = {
67			.flowi6_iif   = sdev->ifindex,
68			.daddr        = ip6h->daddr,
69			.saddr        = ip6h->saddr,
70			.flowlabel    = ip6_flowinfo(ip6h),
71			.flowi6_mark  = skb->mark,
72			.flowi6_proto = ip6h->nexthdr,
73		};
74
75		skb_dst_drop(skb);
76		dst = ip6_route_input_lookup(dev_net(sdev), sdev, &fl6,
77					     skb, flags);
78		skb_dst_set(skb, dst);
79		break;
80	}
81#endif
82	default:
83		break;
84	}
85out:
86	return skb;
87}
88
89static const struct l3mdev_ops ipvl_l3mdev_ops = {
90	.l3mdev_l3_rcv = ipvlan_l3_rcv,
91};
92
93static unsigned int ipvlan_nf_input(void *priv, struct sk_buff *skb,
94				    const struct nf_hook_state *state)
95{
96	struct ipvl_addr *addr;
97	unsigned int len;
98
99	addr = ipvlan_skb_to_addr(skb, skb->dev);
100	if (!addr)
101		goto out;
102
103	skb->dev = addr->master->dev;
104	skb->skb_iif = skb->dev->ifindex;
105#if IS_ENABLED(CONFIG_IPV6)
106	if (addr->atype == IPVL_IPV6)
107		IP6CB(skb)->iif = skb->dev->ifindex;
108#endif
109	len = skb->len + ETH_HLEN;
110	ipvlan_count_rx(addr->master, len, true, false);
111out:
112	return NF_ACCEPT;
113}
114
115static const struct nf_hook_ops ipvl_nfops[] = {
116	{
117		.hook     = ipvlan_nf_input,
118		.pf       = NFPROTO_IPV4,
119		.hooknum  = NF_INET_LOCAL_IN,
120		.priority = INT_MAX,
121	},
122#if IS_ENABLED(CONFIG_IPV6)
123	{
124		.hook     = ipvlan_nf_input,
125		.pf       = NFPROTO_IPV6,
126		.hooknum  = NF_INET_LOCAL_IN,
127		.priority = INT_MAX,
128	},
129#endif
130};
131
132static int ipvlan_register_nf_hook(struct net *net)
133{
134	struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid);
135	int err = 0;
136
137	if (!vnet->ipvl_nf_hook_refcnt) {
138		err = nf_register_net_hooks(net, ipvl_nfops,
139					    ARRAY_SIZE(ipvl_nfops));
140		if (!err)
141			vnet->ipvl_nf_hook_refcnt = 1;
142	} else {
143		vnet->ipvl_nf_hook_refcnt++;
144	}
145
146	return err;
147}
148
149static void ipvlan_unregister_nf_hook(struct net *net)
150{
151	struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid);
152
153	if (WARN_ON(!vnet->ipvl_nf_hook_refcnt))
154		return;
155
156	vnet->ipvl_nf_hook_refcnt--;
157	if (!vnet->ipvl_nf_hook_refcnt)
158		nf_unregister_net_hooks(net, ipvl_nfops,
159					ARRAY_SIZE(ipvl_nfops));
160}
161
162void ipvlan_migrate_l3s_hook(struct net *oldnet, struct net *newnet)
163{
164	struct ipvlan_netns *old_vnet;
165
166	ASSERT_RTNL();
167
168	old_vnet = net_generic(oldnet, ipvlan_netid);
169	if (!old_vnet->ipvl_nf_hook_refcnt)
170		return;
171
172	ipvlan_register_nf_hook(newnet);
173	ipvlan_unregister_nf_hook(oldnet);
174}
175
176static void ipvlan_ns_exit(struct net *net)
177{
178	struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid);
179
180	if (WARN_ON_ONCE(vnet->ipvl_nf_hook_refcnt)) {
181		vnet->ipvl_nf_hook_refcnt = 0;
182		nf_unregister_net_hooks(net, ipvl_nfops,
183					ARRAY_SIZE(ipvl_nfops));
184	}
185}
186
187static struct pernet_operations ipvlan_net_ops = {
188	.id   = &ipvlan_netid,
189	.size = sizeof(struct ipvlan_netns),
190	.exit = ipvlan_ns_exit,
191};
192
193int ipvlan_l3s_init(void)
194{
195	return register_pernet_subsys(&ipvlan_net_ops);
196}
197
198void ipvlan_l3s_cleanup(void)
199{
200	unregister_pernet_subsys(&ipvlan_net_ops);
201}
202
203int ipvlan_l3s_register(struct ipvl_port *port)
204{
205	struct net_device *dev = port->dev;
206	int ret;
207
208	ASSERT_RTNL();
209
210	ret = ipvlan_register_nf_hook(read_pnet(&port->pnet));
211	if (!ret) {
212		dev->l3mdev_ops = &ipvl_l3mdev_ops;
213		dev->priv_flags |= IFF_L3MDEV_RX_HANDLER;
214	}
215
216	return ret;
217}
218
219void ipvlan_l3s_unregister(struct ipvl_port *port)
220{
221	struct net_device *dev = port->dev;
222
223	ASSERT_RTNL();
224
225	dev->priv_flags &= ~IFF_L3MDEV_RX_HANDLER;
226	ipvlan_unregister_nf_hook(read_pnet(&port->pnet));
227	dev->l3mdev_ops = NULL;
228}
229