xref: /linux/net/openvswitch/flow.h (revision e60e1ee60630cafef5e430c2ae364877e061d980)
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 #include <net/nsh.h>
39 
40 struct sk_buff;
41 
42 enum sw_flow_mac_proto {
43 	MAC_PROTO_NONE = 0,
44 	MAC_PROTO_ETHERNET,
45 };
46 #define SW_FLOW_KEY_INVALID	0x80
47 
48 /* Store options at the end of the array if they are less than the
49  * maximum size. This allows us to get the benefits of variable length
50  * matching for small options.
51  */
52 #define TUN_METADATA_OFFSET(opt_len) \
53 	(FIELD_SIZEOF(struct sw_flow_key, tun_opts) - opt_len)
54 #define TUN_METADATA_OPTS(flow_key, opt_len) \
55 	((void *)((flow_key)->tun_opts + TUN_METADATA_OFFSET(opt_len)))
56 
57 struct ovs_tunnel_info {
58 	struct metadata_dst	*tun_dst;
59 };
60 
61 struct vlan_head {
62 	__be16 tpid; /* Vlan type. Generally 802.1q or 802.1ad.*/
63 	__be16 tci;  /* 0 if no VLAN, VLAN_TAG_PRESENT set otherwise. */
64 };
65 
66 #define OVS_SW_FLOW_KEY_METADATA_SIZE			\
67 	(offsetof(struct sw_flow_key, recirc_id) +	\
68 	FIELD_SIZEOF(struct sw_flow_key, recirc_id))
69 
70 struct ovs_key_nsh {
71 	struct ovs_nsh_key_base base;
72 	__be32 context[NSH_MD1_CONTEXT_SIZE];
73 };
74 
75 struct sw_flow_key {
76 	u8 tun_opts[IP_TUNNEL_OPTS_MAX];
77 	u8 tun_opts_len;
78 	struct ip_tunnel_key tun_key;	/* Encapsulating tunnel key. */
79 	struct {
80 		u32	priority;	/* Packet QoS priority. */
81 		u32	skb_mark;	/* SKB mark. */
82 		u16	in_port;	/* Input switch port (or DP_MAX_PORTS). */
83 	} __packed phy; /* Safe when right after 'tun_key'. */
84 	u8 mac_proto;			/* MAC layer protocol (e.g. Ethernet). */
85 	u8 tun_proto;			/* Protocol of encapsulating tunnel. */
86 	u32 ovs_flow_hash;		/* Datapath computed hash value.  */
87 	u32 recirc_id;			/* Recirculation ID.  */
88 	struct {
89 		u8     src[ETH_ALEN];	/* Ethernet source address. */
90 		u8     dst[ETH_ALEN];	/* Ethernet destination address. */
91 		struct vlan_head vlan;
92 		struct vlan_head cvlan;
93 		__be16 type;		/* Ethernet frame type. */
94 	} eth;
95 	/* Filling a hole of two bytes. */
96 	u8 ct_state;
97 	u8 ct_orig_proto;		/* CT original direction tuple IP
98 					 * protocol.
99 					 */
100 	union {
101 		struct {
102 			__be32 top_lse;	/* top label stack entry */
103 		} mpls;
104 		struct {
105 			u8     proto;	/* IP protocol or lower 8 bits of ARP opcode. */
106 			u8     tos;	    /* IP ToS. */
107 			u8     ttl;	    /* IP TTL/hop limit. */
108 			u8     frag;	/* One of OVS_FRAG_TYPE_*. */
109 		} ip;
110 	};
111 	u16 ct_zone;			/* Conntrack zone. */
112 	struct {
113 		__be16 src;		/* TCP/UDP/SCTP source port. */
114 		__be16 dst;		/* TCP/UDP/SCTP destination port. */
115 		__be16 flags;		/* TCP flags. */
116 	} tp;
117 	union {
118 		struct {
119 			struct {
120 				__be32 src;	/* IP source address. */
121 				__be32 dst;	/* IP destination address. */
122 			} addr;
123 			union {
124 				struct {
125 					__be32 src;
126 					__be32 dst;
127 				} ct_orig;	/* Conntrack original direction fields. */
128 				struct {
129 					u8 sha[ETH_ALEN];	/* ARP source hardware address. */
130 					u8 tha[ETH_ALEN];	/* ARP target hardware address. */
131 				} arp;
132 			};
133 		} ipv4;
134 		struct {
135 			struct {
136 				struct in6_addr src;	/* IPv6 source address. */
137 				struct in6_addr dst;	/* IPv6 destination address. */
138 			} addr;
139 			__be32 label;			/* IPv6 flow label. */
140 			union {
141 				struct {
142 					struct in6_addr src;
143 					struct in6_addr dst;
144 				} ct_orig;	/* Conntrack original direction fields. */
145 				struct {
146 					struct in6_addr target;	/* ND target address. */
147 					u8 sll[ETH_ALEN];	/* ND source link layer address. */
148 					u8 tll[ETH_ALEN];	/* ND target link layer address. */
149 				} nd;
150 			};
151 		} ipv6;
152 		struct ovs_key_nsh nsh;         /* network service header */
153 	};
154 	struct {
155 		/* Connection tracking fields not packed above. */
156 		struct {
157 			__be16 src;	/* CT orig tuple tp src port. */
158 			__be16 dst;	/* CT orig tuple tp dst port. */
159 		} orig_tp;
160 		u32 mark;
161 		struct ovs_key_ct_labels labels;
162 	} ct;
163 
164 } __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
165 
166 static inline bool sw_flow_key_is_nd(const struct sw_flow_key *key)
167 {
168 	return key->eth.type == htons(ETH_P_IPV6) &&
169 		key->ip.proto == NEXTHDR_ICMP &&
170 		key->tp.dst == 0 &&
171 		(key->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
172 		 key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT));
173 }
174 
175 struct sw_flow_key_range {
176 	unsigned short int start;
177 	unsigned short int end;
178 };
179 
180 struct sw_flow_mask {
181 	int ref_count;
182 	struct rcu_head rcu;
183 	struct list_head list;
184 	struct sw_flow_key_range range;
185 	struct sw_flow_key key;
186 };
187 
188 struct sw_flow_match {
189 	struct sw_flow_key *key;
190 	struct sw_flow_key_range range;
191 	struct sw_flow_mask *mask;
192 };
193 
194 #define MAX_UFID_LENGTH 16 /* 128 bits */
195 
196 struct sw_flow_id {
197 	u32 ufid_len;
198 	union {
199 		u32 ufid[MAX_UFID_LENGTH / 4];
200 		struct sw_flow_key *unmasked_key;
201 	};
202 };
203 
204 struct sw_flow_actions {
205 	struct rcu_head rcu;
206 	size_t orig_len;	/* From flow_cmd_new netlink actions size */
207 	u32 actions_len;
208 	struct nlattr actions[];
209 };
210 
211 struct flow_stats {
212 	u64 packet_count;		/* Number of packets matched. */
213 	u64 byte_count;			/* Number of bytes matched. */
214 	unsigned long used;		/* Last used time (in jiffies). */
215 	spinlock_t lock;		/* Lock for atomic stats update. */
216 	__be16 tcp_flags;		/* Union of seen TCP flags. */
217 };
218 
219 struct sw_flow {
220 	struct rcu_head rcu;
221 	struct {
222 		struct hlist_node node[2];
223 		u32 hash;
224 	} flow_table, ufid_table;
225 	int stats_last_writer;		/* CPU id of the last writer on
226 					 * 'stats[0]'.
227 					 */
228 	struct sw_flow_key key;
229 	struct sw_flow_id id;
230 	struct cpumask cpu_used_mask;
231 	struct sw_flow_mask *mask;
232 	struct sw_flow_actions __rcu *sf_acts;
233 	struct flow_stats __rcu *stats[]; /* One for each CPU.  First one
234 					   * is allocated at flow creation time,
235 					   * the rest are allocated on demand
236 					   * while holding the 'stats[0].lock'.
237 					   */
238 };
239 
240 struct arp_eth_header {
241 	__be16      ar_hrd;	/* format of hardware address   */
242 	__be16      ar_pro;	/* format of protocol address   */
243 	unsigned char   ar_hln;	/* length of hardware address   */
244 	unsigned char   ar_pln;	/* length of protocol address   */
245 	__be16      ar_op;	/* ARP opcode (command)     */
246 
247 	/* Ethernet+IPv4 specific members. */
248 	unsigned char       ar_sha[ETH_ALEN];	/* sender hardware address  */
249 	unsigned char       ar_sip[4];		/* sender IP address        */
250 	unsigned char       ar_tha[ETH_ALEN];	/* target hardware address  */
251 	unsigned char       ar_tip[4];		/* target IP address        */
252 } __packed;
253 
254 static inline u8 ovs_key_mac_proto(const struct sw_flow_key *key)
255 {
256 	return key->mac_proto & ~SW_FLOW_KEY_INVALID;
257 }
258 
259 static inline u16 __ovs_mac_header_len(u8 mac_proto)
260 {
261 	return mac_proto == MAC_PROTO_ETHERNET ? ETH_HLEN : 0;
262 }
263 
264 static inline u16 ovs_mac_header_len(const struct sw_flow_key *key)
265 {
266 	return __ovs_mac_header_len(ovs_key_mac_proto(key));
267 }
268 
269 static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
270 {
271 	return sfid->ufid_len;
272 }
273 
274 static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
275 {
276 	return !ovs_identifier_is_ufid(sfid);
277 }
278 
279 void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
280 			   const struct sk_buff *);
281 void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
282 			unsigned long *used, __be16 *tcp_flags);
283 void ovs_flow_stats_clear(struct sw_flow *);
284 u64 ovs_flow_used_time(unsigned long flow_jiffies);
285 
286 int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
287 int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
288 			 struct sk_buff *skb,
289 			 struct sw_flow_key *key);
290 /* Extract key from packet coming from userspace. */
291 int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
292 				   struct sk_buff *skb,
293 				   struct sw_flow_key *key, bool log);
294 
295 #endif /* flow.h */
296