xref: /linux/drivers/net/ipvlan/ipvlan_l3s.c (revision 44f57d78)
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 
7 static unsigned int ipvlan_netid __read_mostly;
8 
9 struct ipvlan_netns {
10 	unsigned int ipvl_nf_hook_refcnt;
11 };
12 
13 static 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);
33 out:
34 	return addr;
35 }
36 
37 static 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 	}
85 out:
86 	return skb;
87 }
88 
89 static const struct l3mdev_ops ipvl_l3mdev_ops = {
90 	.l3mdev_l3_rcv = ipvlan_l3_rcv,
91 };
92 
93 static 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 	len = skb->len + ETH_HLEN;
105 	ipvlan_count_rx(addr->master, len, true, false);
106 out:
107 	return NF_ACCEPT;
108 }
109 
110 static const struct nf_hook_ops ipvl_nfops[] = {
111 	{
112 		.hook     = ipvlan_nf_input,
113 		.pf       = NFPROTO_IPV4,
114 		.hooknum  = NF_INET_LOCAL_IN,
115 		.priority = INT_MAX,
116 	},
117 #if IS_ENABLED(CONFIG_IPV6)
118 	{
119 		.hook     = ipvlan_nf_input,
120 		.pf       = NFPROTO_IPV6,
121 		.hooknum  = NF_INET_LOCAL_IN,
122 		.priority = INT_MAX,
123 	},
124 #endif
125 };
126 
127 static int ipvlan_register_nf_hook(struct net *net)
128 {
129 	struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid);
130 	int err = 0;
131 
132 	if (!vnet->ipvl_nf_hook_refcnt) {
133 		err = nf_register_net_hooks(net, ipvl_nfops,
134 					    ARRAY_SIZE(ipvl_nfops));
135 		if (!err)
136 			vnet->ipvl_nf_hook_refcnt = 1;
137 	} else {
138 		vnet->ipvl_nf_hook_refcnt++;
139 	}
140 
141 	return err;
142 }
143 
144 static void ipvlan_unregister_nf_hook(struct net *net)
145 {
146 	struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid);
147 
148 	if (WARN_ON(!vnet->ipvl_nf_hook_refcnt))
149 		return;
150 
151 	vnet->ipvl_nf_hook_refcnt--;
152 	if (!vnet->ipvl_nf_hook_refcnt)
153 		nf_unregister_net_hooks(net, ipvl_nfops,
154 					ARRAY_SIZE(ipvl_nfops));
155 }
156 
157 void ipvlan_migrate_l3s_hook(struct net *oldnet, struct net *newnet)
158 {
159 	struct ipvlan_netns *old_vnet;
160 
161 	ASSERT_RTNL();
162 
163 	old_vnet = net_generic(oldnet, ipvlan_netid);
164 	if (!old_vnet->ipvl_nf_hook_refcnt)
165 		return;
166 
167 	ipvlan_register_nf_hook(newnet);
168 	ipvlan_unregister_nf_hook(oldnet);
169 }
170 
171 static void ipvlan_ns_exit(struct net *net)
172 {
173 	struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid);
174 
175 	if (WARN_ON_ONCE(vnet->ipvl_nf_hook_refcnt)) {
176 		vnet->ipvl_nf_hook_refcnt = 0;
177 		nf_unregister_net_hooks(net, ipvl_nfops,
178 					ARRAY_SIZE(ipvl_nfops));
179 	}
180 }
181 
182 static struct pernet_operations ipvlan_net_ops = {
183 	.id   = &ipvlan_netid,
184 	.size = sizeof(struct ipvlan_netns),
185 	.exit = ipvlan_ns_exit,
186 };
187 
188 int ipvlan_l3s_init(void)
189 {
190 	return register_pernet_subsys(&ipvlan_net_ops);
191 }
192 
193 void ipvlan_l3s_cleanup(void)
194 {
195 	unregister_pernet_subsys(&ipvlan_net_ops);
196 }
197 
198 int ipvlan_l3s_register(struct ipvl_port *port)
199 {
200 	struct net_device *dev = port->dev;
201 	int ret;
202 
203 	ASSERT_RTNL();
204 
205 	ret = ipvlan_register_nf_hook(read_pnet(&port->pnet));
206 	if (!ret) {
207 		dev->l3mdev_ops = &ipvl_l3mdev_ops;
208 		dev->priv_flags |= IFF_L3MDEV_RX_HANDLER;
209 	}
210 
211 	return ret;
212 }
213 
214 void ipvlan_l3s_unregister(struct ipvl_port *port)
215 {
216 	struct net_device *dev = port->dev;
217 
218 	ASSERT_RTNL();
219 
220 	dev->priv_flags &= ~IFF_L3MDEV_RX_HANDLER;
221 	ipvlan_unregister_nf_hook(read_pnet(&port->pnet));
222 	dev->l3mdev_ops = NULL;
223 }
224