xref: /linux/net/openvswitch/flow.h (revision 4246b92cf9fb32da8d8b060c92d8302797c6fbea)
1 /*
2  * Copyright (c) 2007-2017 Nicira, Inc.
3  *
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.
7  *
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.
12  *
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
16  * 02110-1301, USA
17  */
18 
19 #ifndef FLOW_H
20 #define FLOW_H 1
21 
22 #include <linux/cache.h>
23 #include <linux/kernel.h>
24 #include <linux/netlink.h>
25 #include <linux/openvswitch.h>
26 #include <linux/spinlock.h>
27 #include <linux/types.h>
28 #include <linux/rcupdate.h>
29 #include <linux/if_ether.h>
30 #include <linux/in6.h>
31 #include <linux/jiffies.h>
32 #include <linux/time.h>
33 #include <linux/flex_array.h>
34 #include <linux/cpumask.h>
35 #include <net/inet_ecn.h>
36 #include <net/ip_tunnels.h>
37 #include <net/dst_metadata.h>
38 
39 struct sk_buff;
40 
41 enum sw_flow_mac_proto {
42 	MAC_PROTO_NONE = 0,
43 	MAC_PROTO_ETHERNET,
44 };
45 #define SW_FLOW_KEY_INVALID	0x80
46 
47 /* Store options at the end of the array if they are less than the
48  * maximum size. This allows us to get the benefits of variable length
49  * matching for small options.
50  */
51 #define TUN_METADATA_OFFSET(opt_len) \
52 	(FIELD_SIZEOF(struct sw_flow_key, tun_opts) - opt_len)
53 #define TUN_METADATA_OPTS(flow_key, opt_len) \
54 	((void *)((flow_key)->tun_opts + TUN_METADATA_OFFSET(opt_len)))
55 
56 struct ovs_tunnel_info {
57 	struct metadata_dst	*tun_dst;
58 };
59 
60 struct vlan_head {
61 	__be16 tpid; /* Vlan type. Generally 802.1q or 802.1ad.*/
62 	__be16 tci;  /* 0 if no VLAN, VLAN_TAG_PRESENT set otherwise. */
63 };
64 
65 #define OVS_SW_FLOW_KEY_METADATA_SIZE			\
66 	(offsetof(struct sw_flow_key, recirc_id) +	\
67 	FIELD_SIZEOF(struct sw_flow_key, recirc_id))
68 
69 struct sw_flow_key {
70 	u8 tun_opts[IP_TUNNEL_OPTS_MAX];
71 	u8 tun_opts_len;
72 	struct ip_tunnel_key tun_key;	/* Encapsulating tunnel key. */
73 	struct {
74 		u32	priority;	/* Packet QoS priority. */
75 		u32	skb_mark;	/* SKB mark. */
76 		u16	in_port;	/* Input switch port (or DP_MAX_PORTS). */
77 	} __packed phy; /* Safe when right after 'tun_key'. */
78 	u8 mac_proto;			/* MAC layer protocol (e.g. Ethernet). */
79 	u8 tun_proto;			/* Protocol of encapsulating tunnel. */
80 	u32 ovs_flow_hash;		/* Datapath computed hash value.  */
81 	u32 recirc_id;			/* Recirculation ID.  */
82 	struct {
83 		u8     src[ETH_ALEN];	/* Ethernet source address. */
84 		u8     dst[ETH_ALEN];	/* Ethernet destination address. */
85 		struct vlan_head vlan;
86 		struct vlan_head cvlan;
87 		__be16 type;		/* Ethernet frame type. */
88 	} eth;
89 	/* Filling a hole of two bytes. */
90 	u8 ct_state;
91 	u8 ct_orig_proto;		/* CT original direction tuple IP
92 					 * protocol.
93 					 */
94 	union {
95 		struct {
96 			__be32 top_lse;	/* top label stack entry */
97 		} mpls;
98 		struct {
99 			u8     proto;	/* IP protocol or lower 8 bits of ARP opcode. */
100 			u8     tos;	    /* IP ToS. */
101 			u8     ttl;	    /* IP TTL/hop limit. */
102 			u8     frag;	/* One of OVS_FRAG_TYPE_*. */
103 		} ip;
104 	};
105 	u16 ct_zone;			/* Conntrack zone. */
106 	struct {
107 		__be16 src;		/* TCP/UDP/SCTP source port. */
108 		__be16 dst;		/* TCP/UDP/SCTP destination port. */
109 		__be16 flags;		/* TCP flags. */
110 	} tp;
111 	union {
112 		struct {
113 			struct {
114 				__be32 src;	/* IP source address. */
115 				__be32 dst;	/* IP destination address. */
116 			} addr;
117 			union {
118 				struct {
119 					__be32 src;
120 					__be32 dst;
121 				} ct_orig;	/* Conntrack original direction fields. */
122 				struct {
123 					u8 sha[ETH_ALEN];	/* ARP source hardware address. */
124 					u8 tha[ETH_ALEN];	/* ARP target hardware address. */
125 				} arp;
126 			};
127 		} ipv4;
128 		struct {
129 			struct {
130 				struct in6_addr src;	/* IPv6 source address. */
131 				struct in6_addr dst;	/* IPv6 destination address. */
132 			} addr;
133 			__be32 label;			/* IPv6 flow label. */
134 			union {
135 				struct {
136 					struct in6_addr src;
137 					struct in6_addr dst;
138 				} ct_orig;	/* Conntrack original direction fields. */
139 				struct {
140 					struct in6_addr target;	/* ND target address. */
141 					u8 sll[ETH_ALEN];	/* ND source link layer address. */
142 					u8 tll[ETH_ALEN];	/* ND target link layer address. */
143 				} nd;
144 			};
145 		} ipv6;
146 	};
147 	struct {
148 		/* Connection tracking fields not packed above. */
149 		struct {
150 			__be16 src;	/* CT orig tuple tp src port. */
151 			__be16 dst;	/* CT orig tuple tp dst port. */
152 		} orig_tp;
153 		u32 mark;
154 		struct ovs_key_ct_labels labels;
155 	} ct;
156 
157 } __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
158 
159 static inline bool sw_flow_key_is_nd(const struct sw_flow_key *key)
160 {
161 	return key->eth.type == htons(ETH_P_IPV6) &&
162 		key->ip.proto == NEXTHDR_ICMP &&
163 		key->tp.dst == 0 &&
164 		(key->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
165 		 key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT));
166 }
167 
168 struct sw_flow_key_range {
169 	unsigned short int start;
170 	unsigned short int end;
171 };
172 
173 struct sw_flow_mask {
174 	int ref_count;
175 	struct rcu_head rcu;
176 	struct list_head list;
177 	struct sw_flow_key_range range;
178 	struct sw_flow_key key;
179 };
180 
181 struct sw_flow_match {
182 	struct sw_flow_key *key;
183 	struct sw_flow_key_range range;
184 	struct sw_flow_mask *mask;
185 };
186 
187 #define MAX_UFID_LENGTH 16 /* 128 bits */
188 
189 struct sw_flow_id {
190 	u32 ufid_len;
191 	union {
192 		u32 ufid[MAX_UFID_LENGTH / 4];
193 		struct sw_flow_key *unmasked_key;
194 	};
195 };
196 
197 struct sw_flow_actions {
198 	struct rcu_head rcu;
199 	size_t orig_len;	/* From flow_cmd_new netlink actions size */
200 	u32 actions_len;
201 	struct nlattr actions[];
202 };
203 
204 struct flow_stats {
205 	u64 packet_count;		/* Number of packets matched. */
206 	u64 byte_count;			/* Number of bytes matched. */
207 	unsigned long used;		/* Last used time (in jiffies). */
208 	spinlock_t lock;		/* Lock for atomic stats update. */
209 	__be16 tcp_flags;		/* Union of seen TCP flags. */
210 };
211 
212 struct sw_flow {
213 	struct rcu_head rcu;
214 	struct {
215 		struct hlist_node node[2];
216 		u32 hash;
217 	} flow_table, ufid_table;
218 	int stats_last_writer;		/* CPU id of the last writer on
219 					 * 'stats[0]'.
220 					 */
221 	struct sw_flow_key key;
222 	struct sw_flow_id id;
223 	struct cpumask cpu_used_mask;
224 	struct sw_flow_mask *mask;
225 	struct sw_flow_actions __rcu *sf_acts;
226 	struct flow_stats __rcu *stats[]; /* One for each CPU.  First one
227 					   * is allocated at flow creation time,
228 					   * the rest are allocated on demand
229 					   * while holding the 'stats[0].lock'.
230 					   */
231 };
232 
233 struct arp_eth_header {
234 	__be16      ar_hrd;	/* format of hardware address   */
235 	__be16      ar_pro;	/* format of protocol address   */
236 	unsigned char   ar_hln;	/* length of hardware address   */
237 	unsigned char   ar_pln;	/* length of protocol address   */
238 	__be16      ar_op;	/* ARP opcode (command)     */
239 
240 	/* Ethernet+IPv4 specific members. */
241 	unsigned char       ar_sha[ETH_ALEN];	/* sender hardware address  */
242 	unsigned char       ar_sip[4];		/* sender IP address        */
243 	unsigned char       ar_tha[ETH_ALEN];	/* target hardware address  */
244 	unsigned char       ar_tip[4];		/* target IP address        */
245 } __packed;
246 
247 static inline u8 ovs_key_mac_proto(const struct sw_flow_key *key)
248 {
249 	return key->mac_proto & ~SW_FLOW_KEY_INVALID;
250 }
251 
252 static inline u16 __ovs_mac_header_len(u8 mac_proto)
253 {
254 	return mac_proto == MAC_PROTO_ETHERNET ? ETH_HLEN : 0;
255 }
256 
257 static inline u16 ovs_mac_header_len(const struct sw_flow_key *key)
258 {
259 	return __ovs_mac_header_len(ovs_key_mac_proto(key));
260 }
261 
262 static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
263 {
264 	return sfid->ufid_len;
265 }
266 
267 static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
268 {
269 	return !ovs_identifier_is_ufid(sfid);
270 }
271 
272 void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
273 			   const struct sk_buff *);
274 void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
275 			unsigned long *used, __be16 *tcp_flags);
276 void ovs_flow_stats_clear(struct sw_flow *);
277 u64 ovs_flow_used_time(unsigned long flow_jiffies);
278 
279 int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
280 int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
281 			 struct sk_buff *skb,
282 			 struct sw_flow_key *key);
283 /* Extract key from packet coming from userspace. */
284 int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
285 				   struct sk_buff *skb,
286 				   struct sw_flow_key *key, bool log);
287 
288 #endif /* flow.h */
289