xref: /linux/net/openvswitch/flow.h (revision e7d759f31ca295d589f7420719c311870bb3166f)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Copyright (c) 2007-2017 Nicira, Inc.
4  */
5 
6 #ifndef FLOW_H
7 #define FLOW_H 1
8 
9 #include <linux/cache.h>
10 #include <linux/kernel.h>
11 #include <linux/netlink.h>
12 #include <linux/openvswitch.h>
13 #include <linux/spinlock.h>
14 #include <linux/types.h>
15 #include <linux/rcupdate.h>
16 #include <linux/if_ether.h>
17 #include <linux/in6.h>
18 #include <linux/jiffies.h>
19 #include <linux/time.h>
20 #include <linux/cpumask.h>
21 #include <net/inet_ecn.h>
22 #include <net/ip_tunnels.h>
23 #include <net/dst_metadata.h>
24 #include <net/nsh.h>
25 
26 struct sk_buff;
27 
28 enum sw_flow_mac_proto {
29 	MAC_PROTO_NONE = 0,
30 	MAC_PROTO_ETHERNET,
31 };
32 #define SW_FLOW_KEY_INVALID	0x80
33 #define MPLS_LABEL_DEPTH       3
34 
35 /* Bit definitions for IPv6 Extension Header pseudo-field. */
36 enum ofp12_ipv6exthdr_flags {
37 	OFPIEH12_NONEXT = 1 << 0,   /* "No next header" encountered. */
38 	OFPIEH12_ESP    = 1 << 1,   /* Encrypted Sec Payload header present. */
39 	OFPIEH12_AUTH   = 1 << 2,   /* Authentication header present. */
40 	OFPIEH12_DEST   = 1 << 3,   /* 1 or 2 dest headers present. */
41 	OFPIEH12_FRAG   = 1 << 4,   /* Fragment header present. */
42 	OFPIEH12_ROUTER = 1 << 5,   /* Router header present. */
43 	OFPIEH12_HOP    = 1 << 6,   /* Hop-by-hop header present. */
44 	OFPIEH12_UNREP  = 1 << 7,   /* Unexpected repeats encountered. */
45 	OFPIEH12_UNSEQ  = 1 << 8    /* Unexpected sequencing encountered. */
46 };
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 	(sizeof_field(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_CFI_MASK set otherwise. */
64 };
65 
66 #define OVS_SW_FLOW_KEY_METADATA_SIZE			\
67 	(offsetof(struct sw_flow_key, recirc_id) +	\
68 	sizeof_field(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 			u8     proto;	/* IP protocol or lower 8 bits of ARP opcode. */
103 			u8     tos;	    /* IP ToS. */
104 			u8     ttl;	    /* IP TTL/hop limit. */
105 			u8     frag;	/* One of OVS_FRAG_TYPE_*. */
106 		} ip;
107 	};
108 	u16 ct_zone;			/* Conntrack zone. */
109 	struct {
110 		__be16 src;		/* TCP/UDP/SCTP source port. */
111 		__be16 dst;		/* TCP/UDP/SCTP destination port. */
112 		__be16 flags;		/* TCP flags. */
113 	} tp;
114 	union {
115 		struct {
116 			struct {
117 				__be32 src;	/* IP source address. */
118 				__be32 dst;	/* IP destination address. */
119 			} addr;
120 			union {
121 				struct {
122 					__be32 src;
123 					__be32 dst;
124 				} ct_orig;	/* Conntrack original direction fields. */
125 				struct {
126 					u8 sha[ETH_ALEN];	/* ARP source hardware address. */
127 					u8 tha[ETH_ALEN];	/* ARP target hardware address. */
128 				} arp;
129 			};
130 		} ipv4;
131 		struct {
132 			struct {
133 				struct in6_addr src;	/* IPv6 source address. */
134 				struct in6_addr dst;	/* IPv6 destination address. */
135 			} addr;
136 			__be32 label;			/* IPv6 flow label. */
137 			u16 exthdrs;	/* IPv6 extension header flags */
138 			union {
139 				struct {
140 					struct in6_addr src;
141 					struct in6_addr dst;
142 				} ct_orig;	/* Conntrack original direction fields. */
143 				struct {
144 					struct in6_addr target;	/* ND target address. */
145 					u8 sll[ETH_ALEN];	/* ND source link layer address. */
146 					u8 tll[ETH_ALEN];	/* ND target link layer address. */
147 				} nd;
148 			};
149 		} ipv6;
150 		struct {
151 			u32 num_labels_mask;    /* labels present bitmap of effective length MPLS_LABEL_DEPTH */
152 			__be32 lse[MPLS_LABEL_DEPTH];     /* label stack entry  */
153 		} mpls;
154 
155 		struct ovs_key_nsh nsh;         /* network service header */
156 	};
157 	struct {
158 		/* Connection tracking fields not packed above. */
159 		struct {
160 			__be16 src;	/* CT orig tuple tp src port. */
161 			__be16 dst;	/* CT orig tuple tp dst port. */
162 		} orig_tp;
163 		u32 mark;
164 		struct ovs_key_ct_labels labels;
165 	} ct;
166 
167 } __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
168 
169 static inline bool sw_flow_key_is_nd(const struct sw_flow_key *key)
170 {
171 	return key->eth.type == htons(ETH_P_IPV6) &&
172 		key->ip.proto == NEXTHDR_ICMP &&
173 		key->tp.dst == 0 &&
174 		(key->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
175 		 key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT));
176 }
177 
178 struct sw_flow_key_range {
179 	unsigned short int start;
180 	unsigned short int end;
181 };
182 
183 struct sw_flow_mask {
184 	int ref_count;
185 	struct rcu_head rcu;
186 	struct sw_flow_key_range range;
187 	struct sw_flow_key key;
188 };
189 
190 struct sw_flow_match {
191 	struct sw_flow_key *key;
192 	struct sw_flow_key_range range;
193 	struct sw_flow_mask *mask;
194 };
195 
196 #define MAX_UFID_LENGTH 16 /* 128 bits */
197 
198 struct sw_flow_id {
199 	u32 ufid_len;
200 	union {
201 		u32 ufid[MAX_UFID_LENGTH / 4];
202 		struct sw_flow_key *unmasked_key;
203 	};
204 };
205 
206 struct sw_flow_actions {
207 	struct rcu_head rcu;
208 	size_t orig_len;	/* From flow_cmd_new netlink actions size */
209 	u32 actions_len;
210 	struct nlattr actions[];
211 };
212 
213 struct sw_flow_stats {
214 	u64 packet_count;		/* Number of packets matched. */
215 	u64 byte_count;			/* Number of bytes matched. */
216 	unsigned long used;		/* Last used time (in jiffies). */
217 	spinlock_t lock;		/* Lock for atomic stats update. */
218 	__be16 tcp_flags;		/* Union of seen TCP flags. */
219 };
220 
221 struct sw_flow {
222 	struct rcu_head rcu;
223 	struct {
224 		struct hlist_node node[2];
225 		u32 hash;
226 	} flow_table, ufid_table;
227 	int stats_last_writer;		/* CPU id of the last writer on
228 					 * 'stats[0]'.
229 					 */
230 	struct sw_flow_key key;
231 	struct sw_flow_id id;
232 	struct cpumask *cpu_used_mask;
233 	struct sw_flow_mask *mask;
234 	struct sw_flow_actions __rcu *sf_acts;
235 	struct sw_flow_stats __rcu *stats[]; /* One for each CPU.  First one
236 					   * is allocated at flow creation time,
237 					   * the rest are allocated on demand
238 					   * while holding the 'stats[0].lock'.
239 					   */
240 };
241 
242 struct arp_eth_header {
243 	__be16      ar_hrd;	/* format of hardware address   */
244 	__be16      ar_pro;	/* format of protocol address   */
245 	unsigned char   ar_hln;	/* length of hardware address   */
246 	unsigned char   ar_pln;	/* length of protocol address   */
247 	__be16      ar_op;	/* ARP opcode (command)     */
248 
249 	/* Ethernet+IPv4 specific members. */
250 	unsigned char       ar_sha[ETH_ALEN];	/* sender hardware address  */
251 	unsigned char       ar_sip[4];		/* sender IP address        */
252 	unsigned char       ar_tha[ETH_ALEN];	/* target hardware address  */
253 	unsigned char       ar_tip[4];		/* target IP address        */
254 } __packed;
255 
256 static inline u8 ovs_key_mac_proto(const struct sw_flow_key *key)
257 {
258 	return key->mac_proto & ~SW_FLOW_KEY_INVALID;
259 }
260 
261 static inline u16 __ovs_mac_header_len(u8 mac_proto)
262 {
263 	return mac_proto == MAC_PROTO_ETHERNET ? ETH_HLEN : 0;
264 }
265 
266 static inline u16 ovs_mac_header_len(const struct sw_flow_key *key)
267 {
268 	return __ovs_mac_header_len(ovs_key_mac_proto(key));
269 }
270 
271 static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
272 {
273 	return sfid->ufid_len;
274 }
275 
276 static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
277 {
278 	return !ovs_identifier_is_ufid(sfid);
279 }
280 
281 void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
282 			   const struct sk_buff *);
283 void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
284 			unsigned long *used, __be16 *tcp_flags);
285 void ovs_flow_stats_clear(struct sw_flow *);
286 u64 ovs_flow_used_time(unsigned long flow_jiffies);
287 
288 int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
289 int ovs_flow_key_update_l3l4(struct sk_buff *skb, struct sw_flow_key *key);
290 int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
291 			 struct sk_buff *skb,
292 			 struct sw_flow_key *key);
293 /* Extract key from packet coming from userspace. */
294 int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
295 				   struct sk_buff *skb,
296 				   struct sw_flow_key *key, bool log);
297 
298 #endif /* flow.h */
299