xref: /linux/net/openvswitch/flow.h (revision ca64d84e93762f4e587e040a44ad9f6089afc777)
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 /* Store options at the end of the array if they are less than the
36  * maximum size. This allows us to get the benefits of variable length
37  * matching for small options.
38  */
39 #define TUN_METADATA_OFFSET(opt_len) \
40 	(sizeof_field(struct sw_flow_key, tun_opts) - opt_len)
41 #define TUN_METADATA_OPTS(flow_key, opt_len) \
42 	((void *)((flow_key)->tun_opts + TUN_METADATA_OFFSET(opt_len)))
43 
44 struct ovs_tunnel_info {
45 	struct metadata_dst	*tun_dst;
46 };
47 
48 struct vlan_head {
49 	__be16 tpid; /* Vlan type. Generally 802.1q or 802.1ad.*/
50 	__be16 tci;  /* 0 if no VLAN, VLAN_CFI_MASK set otherwise. */
51 };
52 
53 #define OVS_SW_FLOW_KEY_METADATA_SIZE			\
54 	(offsetof(struct sw_flow_key, recirc_id) +	\
55 	sizeof_field(struct sw_flow_key, recirc_id))
56 
57 struct ovs_key_nsh {
58 	struct ovs_nsh_key_base base;
59 	__be32 context[NSH_MD1_CONTEXT_SIZE];
60 };
61 
62 struct sw_flow_key {
63 	u8 tun_opts[IP_TUNNEL_OPTS_MAX];
64 	u8 tun_opts_len;
65 	struct ip_tunnel_key tun_key;	/* Encapsulating tunnel key. */
66 	struct {
67 		u32	priority;	/* Packet QoS priority. */
68 		u32	skb_mark;	/* SKB mark. */
69 		u16	in_port;	/* Input switch port (or DP_MAX_PORTS). */
70 	} __packed phy; /* Safe when right after 'tun_key'. */
71 	u8 mac_proto;			/* MAC layer protocol (e.g. Ethernet). */
72 	u8 tun_proto;			/* Protocol of encapsulating tunnel. */
73 	u32 ovs_flow_hash;		/* Datapath computed hash value.  */
74 	u32 recirc_id;			/* Recirculation ID.  */
75 	struct {
76 		u8     src[ETH_ALEN];	/* Ethernet source address. */
77 		u8     dst[ETH_ALEN];	/* Ethernet destination address. */
78 		struct vlan_head vlan;
79 		struct vlan_head cvlan;
80 		__be16 type;		/* Ethernet frame type. */
81 	} eth;
82 	/* Filling a hole of two bytes. */
83 	u8 ct_state;
84 	u8 ct_orig_proto;		/* CT original direction tuple IP
85 					 * protocol.
86 					 */
87 	union {
88 		struct {
89 			u8     proto;	/* IP protocol or lower 8 bits of ARP opcode. */
90 			u8     tos;	    /* IP ToS. */
91 			u8     ttl;	    /* IP TTL/hop limit. */
92 			u8     frag;	/* One of OVS_FRAG_TYPE_*. */
93 		} ip;
94 	};
95 	u16 ct_zone;			/* Conntrack zone. */
96 	struct {
97 		__be16 src;		/* TCP/UDP/SCTP source port. */
98 		__be16 dst;		/* TCP/UDP/SCTP destination port. */
99 		__be16 flags;		/* TCP flags. */
100 	} tp;
101 	union {
102 		struct {
103 			struct {
104 				__be32 src;	/* IP source address. */
105 				__be32 dst;	/* IP destination address. */
106 			} addr;
107 			union {
108 				struct {
109 					__be32 src;
110 					__be32 dst;
111 				} ct_orig;	/* Conntrack original direction fields. */
112 				struct {
113 					u8 sha[ETH_ALEN];	/* ARP source hardware address. */
114 					u8 tha[ETH_ALEN];	/* ARP target hardware address. */
115 				} arp;
116 			};
117 		} ipv4;
118 		struct {
119 			struct {
120 				struct in6_addr src;	/* IPv6 source address. */
121 				struct in6_addr dst;	/* IPv6 destination address. */
122 			} addr;
123 			__be32 label;			/* IPv6 flow label. */
124 			union {
125 				struct {
126 					struct in6_addr src;
127 					struct in6_addr dst;
128 				} ct_orig;	/* Conntrack original direction fields. */
129 				struct {
130 					struct in6_addr target;	/* ND target address. */
131 					u8 sll[ETH_ALEN];	/* ND source link layer address. */
132 					u8 tll[ETH_ALEN];	/* ND target link layer address. */
133 				} nd;
134 			};
135 		} ipv6;
136 		struct {
137 			u32 num_labels_mask;    /* labels present bitmap of effective length MPLS_LABEL_DEPTH */
138 			__be32 lse[MPLS_LABEL_DEPTH];     /* label stack entry  */
139 		} mpls;
140 
141 		struct ovs_key_nsh nsh;         /* network service header */
142 	};
143 	struct {
144 		/* Connection tracking fields not packed above. */
145 		struct {
146 			__be16 src;	/* CT orig tuple tp src port. */
147 			__be16 dst;	/* CT orig tuple tp dst port. */
148 		} orig_tp;
149 		u32 mark;
150 		struct ovs_key_ct_labels labels;
151 	} ct;
152 
153 } __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
154 
155 static inline bool sw_flow_key_is_nd(const struct sw_flow_key *key)
156 {
157 	return key->eth.type == htons(ETH_P_IPV6) &&
158 		key->ip.proto == NEXTHDR_ICMP &&
159 		key->tp.dst == 0 &&
160 		(key->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
161 		 key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT));
162 }
163 
164 struct sw_flow_key_range {
165 	unsigned short int start;
166 	unsigned short int end;
167 };
168 
169 struct sw_flow_mask {
170 	int ref_count;
171 	struct rcu_head rcu;
172 	struct sw_flow_key_range range;
173 	struct sw_flow_key key;
174 };
175 
176 struct sw_flow_match {
177 	struct sw_flow_key *key;
178 	struct sw_flow_key_range range;
179 	struct sw_flow_mask *mask;
180 };
181 
182 #define MAX_UFID_LENGTH 16 /* 128 bits */
183 
184 struct sw_flow_id {
185 	u32 ufid_len;
186 	union {
187 		u32 ufid[MAX_UFID_LENGTH / 4];
188 		struct sw_flow_key *unmasked_key;
189 	};
190 };
191 
192 struct sw_flow_actions {
193 	struct rcu_head rcu;
194 	size_t orig_len;	/* From flow_cmd_new netlink actions size */
195 	u32 actions_len;
196 	struct nlattr actions[];
197 };
198 
199 struct sw_flow_stats {
200 	u64 packet_count;		/* Number of packets matched. */
201 	u64 byte_count;			/* Number of bytes matched. */
202 	unsigned long used;		/* Last used time (in jiffies). */
203 	spinlock_t lock;		/* Lock for atomic stats update. */
204 	__be16 tcp_flags;		/* Union of seen TCP flags. */
205 };
206 
207 struct sw_flow {
208 	struct rcu_head rcu;
209 	struct {
210 		struct hlist_node node[2];
211 		u32 hash;
212 	} flow_table, ufid_table;
213 	int stats_last_writer;		/* CPU id of the last writer on
214 					 * 'stats[0]'.
215 					 */
216 	struct sw_flow_key key;
217 	struct sw_flow_id id;
218 	struct cpumask cpu_used_mask;
219 	struct sw_flow_mask *mask;
220 	struct sw_flow_actions __rcu *sf_acts;
221 	struct sw_flow_stats __rcu *stats[]; /* One for each CPU.  First one
222 					   * is allocated at flow creation time,
223 					   * the rest are allocated on demand
224 					   * while holding the 'stats[0].lock'.
225 					   */
226 };
227 
228 struct arp_eth_header {
229 	__be16      ar_hrd;	/* format of hardware address   */
230 	__be16      ar_pro;	/* format of protocol address   */
231 	unsigned char   ar_hln;	/* length of hardware address   */
232 	unsigned char   ar_pln;	/* length of protocol address   */
233 	__be16      ar_op;	/* ARP opcode (command)     */
234 
235 	/* Ethernet+IPv4 specific members. */
236 	unsigned char       ar_sha[ETH_ALEN];	/* sender hardware address  */
237 	unsigned char       ar_sip[4];		/* sender IP address        */
238 	unsigned char       ar_tha[ETH_ALEN];	/* target hardware address  */
239 	unsigned char       ar_tip[4];		/* target IP address        */
240 } __packed;
241 
242 static inline u8 ovs_key_mac_proto(const struct sw_flow_key *key)
243 {
244 	return key->mac_proto & ~SW_FLOW_KEY_INVALID;
245 }
246 
247 static inline u16 __ovs_mac_header_len(u8 mac_proto)
248 {
249 	return mac_proto == MAC_PROTO_ETHERNET ? ETH_HLEN : 0;
250 }
251 
252 static inline u16 ovs_mac_header_len(const struct sw_flow_key *key)
253 {
254 	return __ovs_mac_header_len(ovs_key_mac_proto(key));
255 }
256 
257 static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
258 {
259 	return sfid->ufid_len;
260 }
261 
262 static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
263 {
264 	return !ovs_identifier_is_ufid(sfid);
265 }
266 
267 void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
268 			   const struct sk_buff *);
269 void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
270 			unsigned long *used, __be16 *tcp_flags);
271 void ovs_flow_stats_clear(struct sw_flow *);
272 u64 ovs_flow_used_time(unsigned long flow_jiffies);
273 
274 int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
275 int ovs_flow_key_update_l3l4(struct sk_buff *skb, struct sw_flow_key *key);
276 int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
277 			 struct sk_buff *skb,
278 			 struct sw_flow_key *key);
279 /* Extract key from packet coming from userspace. */
280 int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
281 				   struct sk_buff *skb,
282 				   struct sw_flow_key *key, bool log);
283 
284 #endif /* flow.h */
285