2 * Copyright (c) 2007-2012 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 #include <linux/in_route.h>
23 #include <linux/inetdevice.h>
24 #include <linux/jhash.h>
25 #include <linux/list.h>
26 #include <linux/kernel.h>
27 #include <linux/version.h>
28 #include <linux/workqueue.h>
29 #include <linux/rculist.h>
30 #include <net/route.h>
41 * ovs_tnl_rcv - ingress point for generic tunnel code
43 * @vport: port this packet was received on
44 * @skb: received packet
45 * @tos: ToS from encapsulating IP packet, used to copy ECN bits
47 * Must be called with rcu_read_lock.
49 * Packets received by this function are in the following state:
50 * - skb->data points to the inner Ethernet header.
51 * - The inner Ethernet header is in the linear data area.
52 * - skb->csum does not include the inner Ethernet header.
53 * - The layer pointers are undefined.
55 void ovs_tnl_rcv(struct vport *vport, struct sk_buff *skb,
56 struct ovs_key_ipv4_tunnel *tun_key)
60 skb_reset_mac_header(skb);
63 if (likely(ntohs(eh->h_proto) >= ETH_P_802_3_MIN))
64 skb->protocol = eh->h_proto;
66 skb->protocol = htons(ETH_P_802_2);
70 skb_clear_rxhash(skb);
74 if (unlikely(compute_ip_summed(skb, false))) {
79 ovs_vport_receive(vport, skb, tun_key);
82 struct rtable *find_route(struct net *net,
83 __be32 *saddr, __be32 daddr, u8 ipproto,
87 /* Tunnel configuration keeps DSCP part of TOS bits, But Linux
88 * router expect RT_TOS bits only. */
90 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39)
91 struct flowi fl = { .nl_u = { .ip4_u = {
94 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20)
97 .tos = RT_TOS(tos) } },
98 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20)
103 if (unlikely(ip_route_output_key(net, &rt, &fl)))
104 return ERR_PTR(-EADDRNOTAVAIL);
105 *saddr = fl.nl_u.ip4_u.saddr;
108 struct flowi4 fl = { .daddr = daddr,
110 .flowi4_tos = RT_TOS(tos),
111 .flowi4_mark = skb_mark,
112 .flowi4_proto = ipproto };
114 rt = ip_route_output_key(net, &fl);
120 static bool need_linearize(const struct sk_buff *skb)
124 if (unlikely(skb_shinfo(skb)->frag_list))
128 * Generally speaking we should linearize if there are paged frags.
129 * However, if all of the refcounts are 1 we know nobody else can
130 * change them from underneath us and we can skip the linearization.
132 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++)
133 if (unlikely(page_count(skb_frag_page(&skb_shinfo(skb)->frags[i])) > 1))
139 static struct sk_buff *handle_offloads(struct sk_buff *skb)
143 forward_ip_summed(skb, true);
145 if (skb_is_gso(skb)) {
146 struct sk_buff *nskb;
148 nskb = __skb_gso_segment(skb, 0, false);
156 } else if (get_ip_summed(skb) == OVS_CSUM_PARTIAL) {
157 /* Pages aren't locked and could change at any time.
158 * If this happens after we compute the checksum, the
159 * checksum will be wrong. We linearize now to avoid
162 if (unlikely(need_linearize(skb))) {
163 err = __skb_linearize(skb);
168 err = skb_checksum_help(skb);
173 set_ip_summed(skb, OVS_CSUM_NONE);
181 /* Compute source UDP port for outgoing packet.
182 * Currently we use the flow hash.
184 u16 ovs_tnl_get_src_port(struct sk_buff *skb)
189 struct sw_flow_key *pkt_key = OVS_CB(skb)->pkt_key;
190 u32 hash = jhash2((const u32 *)pkt_key,
191 sizeof(*pkt_key) / sizeof(u32), 0);
193 inet_get_local_port_range(&low, &high);
194 range = (high - low) + 1;
195 return (((u64) hash * range) >> 32) + low;
198 int ovs_tnl_send(struct vport *vport, struct sk_buff *skb,
199 u8 ipproto, int tunnel_hlen,
200 void (*build_header)(const struct vport *,
209 struct sk_buff *nskb;
212 saddr = OVS_CB(skb)->tun_key->ipv4_src;
213 rt = find_route(ovs_dp_get_net(vport->dp),
215 OVS_CB(skb)->tun_key->ipv4_dst,
217 OVS_CB(skb)->tun_key->ipv4_tos,
224 tunnel_hlen += sizeof(struct iphdr);
226 min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
228 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
230 if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
231 int head_delta = SKB_DATA_ALIGN(min_headroom -
235 err = pskb_expand_head(skb, max_t(int, head_delta, 0),
242 nskb = handle_offloads(skb);
253 skb_clear_rxhash(skb);
256 struct sk_buff *next_skb = skb->next;
262 if (unlikely(vlan_deaccel_tag(skb)))
266 skb_push(skb, tunnel_hlen);
267 skb_reset_network_header(skb);
268 skb_set_transport_header(skb, sizeof(struct iphdr));
271 skb_dst_set(skb, dst_clone(&rt_dst(rt)));
273 skb_dst_set(skb, &rt_dst(rt));
275 /* Push Tunnel header. */
276 build_header(vport, skb, tunnel_hlen);
278 /* Push IP header. */
281 iph->ihl = sizeof(struct iphdr) >> 2;
282 iph->protocol = ipproto;
283 iph->daddr = OVS_CB(skb)->tun_key->ipv4_dst;
285 iph->tos = OVS_CB(skb)->tun_key->ipv4_tos;
286 iph->ttl = OVS_CB(skb)->tun_key->ipv4_ttl;
287 iph->frag_off = OVS_CB(skb)->tun_key->tun_flags &
288 TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
290 * Allow our local IP stack to fragment the outer packet even
291 * if the DF bit is set as a last resort. We also need to
292 * force selection of an IP ID here with __ip_select_ident(),
293 * as ip_select_ident() assumes a proper ID is not needed when
294 * when the DF bit is set.
297 __ip_select_ident(iph, skb_dst(skb), 0);
299 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
301 err = ip_local_out(skb);
302 if (unlikely(net_xmit_eval(err)))
305 sent_len += frag_len;