xref: /openbmc/linux/net/openvswitch/flow.h (revision 9a8f3203)
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/cpumask.h>
34 #include <net/inet_ecn.h>
35 #include <net/ip_tunnels.h>
36 #include <net/dst_metadata.h>
37 #include <net/nsh.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_CFI_MASK 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 ovs_key_nsh {
70 	struct ovs_nsh_key_base base;
71 	__be32 context[NSH_MD1_CONTEXT_SIZE];
72 };
73 
74 struct sw_flow_key {
75 	u8 tun_opts[IP_TUNNEL_OPTS_MAX];
76 	u8 tun_opts_len;
77 	struct ip_tunnel_key tun_key;	/* Encapsulating tunnel key. */
78 	struct {
79 		u32	priority;	/* Packet QoS priority. */
80 		u32	skb_mark;	/* SKB mark. */
81 		u16	in_port;	/* Input switch port (or DP_MAX_PORTS). */
82 	} __packed phy; /* Safe when right after 'tun_key'. */
83 	u8 mac_proto;			/* MAC layer protocol (e.g. Ethernet). */
84 	u8 tun_proto;			/* Protocol of encapsulating tunnel. */
85 	u32 ovs_flow_hash;		/* Datapath computed hash value.  */
86 	u32 recirc_id;			/* Recirculation ID.  */
87 	struct {
88 		u8     src[ETH_ALEN];	/* Ethernet source address. */
89 		u8     dst[ETH_ALEN];	/* Ethernet destination address. */
90 		struct vlan_head vlan;
91 		struct vlan_head cvlan;
92 		__be16 type;		/* Ethernet frame type. */
93 	} eth;
94 	/* Filling a hole of two bytes. */
95 	u8 ct_state;
96 	u8 ct_orig_proto;		/* CT original direction tuple IP
97 					 * protocol.
98 					 */
99 	union {
100 		struct {
101 			__be32 top_lse;	/* top label stack entry */
102 		} mpls;
103 		struct {
104 			u8     proto;	/* IP protocol or lower 8 bits of ARP opcode. */
105 			u8     tos;	    /* IP ToS. */
106 			u8     ttl;	    /* IP TTL/hop limit. */
107 			u8     frag;	/* One of OVS_FRAG_TYPE_*. */
108 		} ip;
109 	};
110 	u16 ct_zone;			/* Conntrack zone. */
111 	struct {
112 		__be16 src;		/* TCP/UDP/SCTP source port. */
113 		__be16 dst;		/* TCP/UDP/SCTP destination port. */
114 		__be16 flags;		/* TCP flags. */
115 	} tp;
116 	union {
117 		struct {
118 			struct {
119 				__be32 src;	/* IP source address. */
120 				__be32 dst;	/* IP destination address. */
121 			} addr;
122 			union {
123 				struct {
124 					__be32 src;
125 					__be32 dst;
126 				} ct_orig;	/* Conntrack original direction fields. */
127 				struct {
128 					u8 sha[ETH_ALEN];	/* ARP source hardware address. */
129 					u8 tha[ETH_ALEN];	/* ARP target hardware address. */
130 				} arp;
131 			};
132 		} ipv4;
133 		struct {
134 			struct {
135 				struct in6_addr src;	/* IPv6 source address. */
136 				struct in6_addr dst;	/* IPv6 destination address. */
137 			} addr;
138 			__be32 label;			/* IPv6 flow label. */
139 			union {
140 				struct {
141 					struct in6_addr src;
142 					struct in6_addr dst;
143 				} ct_orig;	/* Conntrack original direction fields. */
144 				struct {
145 					struct in6_addr target;	/* ND target address. */
146 					u8 sll[ETH_ALEN];	/* ND source link layer address. */
147 					u8 tll[ETH_ALEN];	/* ND target link layer address. */
148 				} nd;
149 			};
150 		} ipv6;
151 		struct ovs_key_nsh nsh;         /* network service header */
152 	};
153 	struct {
154 		/* Connection tracking fields not packed above. */
155 		struct {
156 			__be16 src;	/* CT orig tuple tp src port. */
157 			__be16 dst;	/* CT orig tuple tp dst port. */
158 		} orig_tp;
159 		u32 mark;
160 		struct ovs_key_ct_labels labels;
161 	} ct;
162 
163 } __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
164 
165 static inline bool sw_flow_key_is_nd(const struct sw_flow_key *key)
166 {
167 	return key->eth.type == htons(ETH_P_IPV6) &&
168 		key->ip.proto == NEXTHDR_ICMP &&
169 		key->tp.dst == 0 &&
170 		(key->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
171 		 key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT));
172 }
173 
174 struct sw_flow_key_range {
175 	unsigned short int start;
176 	unsigned short int end;
177 };
178 
179 struct sw_flow_mask {
180 	int ref_count;
181 	struct rcu_head rcu;
182 	struct list_head list;
183 	struct sw_flow_key_range range;
184 	struct sw_flow_key key;
185 };
186 
187 struct sw_flow_match {
188 	struct sw_flow_key *key;
189 	struct sw_flow_key_range range;
190 	struct sw_flow_mask *mask;
191 };
192 
193 #define MAX_UFID_LENGTH 16 /* 128 bits */
194 
195 struct sw_flow_id {
196 	u32 ufid_len;
197 	union {
198 		u32 ufid[MAX_UFID_LENGTH / 4];
199 		struct sw_flow_key *unmasked_key;
200 	};
201 };
202 
203 struct sw_flow_actions {
204 	struct rcu_head rcu;
205 	size_t orig_len;	/* From flow_cmd_new netlink actions size */
206 	u32 actions_len;
207 	struct nlattr actions[];
208 };
209 
210 struct flow_stats {
211 	u64 packet_count;		/* Number of packets matched. */
212 	u64 byte_count;			/* Number of bytes matched. */
213 	unsigned long used;		/* Last used time (in jiffies). */
214 	spinlock_t lock;		/* Lock for atomic stats update. */
215 	__be16 tcp_flags;		/* Union of seen TCP flags. */
216 };
217 
218 struct sw_flow {
219 	struct rcu_head rcu;
220 	struct {
221 		struct hlist_node node[2];
222 		u32 hash;
223 	} flow_table, ufid_table;
224 	int stats_last_writer;		/* CPU id of the last writer on
225 					 * 'stats[0]'.
226 					 */
227 	struct sw_flow_key key;
228 	struct sw_flow_id id;
229 	struct cpumask cpu_used_mask;
230 	struct sw_flow_mask *mask;
231 	struct sw_flow_actions __rcu *sf_acts;
232 	struct flow_stats __rcu *stats[]; /* One for each CPU.  First one
233 					   * is allocated at flow creation time,
234 					   * the rest are allocated on demand
235 					   * while holding the 'stats[0].lock'.
236 					   */
237 };
238 
239 struct arp_eth_header {
240 	__be16      ar_hrd;	/* format of hardware address   */
241 	__be16      ar_pro;	/* format of protocol address   */
242 	unsigned char   ar_hln;	/* length of hardware address   */
243 	unsigned char   ar_pln;	/* length of protocol address   */
244 	__be16      ar_op;	/* ARP opcode (command)     */
245 
246 	/* Ethernet+IPv4 specific members. */
247 	unsigned char       ar_sha[ETH_ALEN];	/* sender hardware address  */
248 	unsigned char       ar_sip[4];		/* sender IP address        */
249 	unsigned char       ar_tha[ETH_ALEN];	/* target hardware address  */
250 	unsigned char       ar_tip[4];		/* target IP address        */
251 } __packed;
252 
253 static inline u8 ovs_key_mac_proto(const struct sw_flow_key *key)
254 {
255 	return key->mac_proto & ~SW_FLOW_KEY_INVALID;
256 }
257 
258 static inline u16 __ovs_mac_header_len(u8 mac_proto)
259 {
260 	return mac_proto == MAC_PROTO_ETHERNET ? ETH_HLEN : 0;
261 }
262 
263 static inline u16 ovs_mac_header_len(const struct sw_flow_key *key)
264 {
265 	return __ovs_mac_header_len(ovs_key_mac_proto(key));
266 }
267 
268 static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
269 {
270 	return sfid->ufid_len;
271 }
272 
273 static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
274 {
275 	return !ovs_identifier_is_ufid(sfid);
276 }
277 
278 void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
279 			   const struct sk_buff *);
280 void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
281 			unsigned long *used, __be16 *tcp_flags);
282 void ovs_flow_stats_clear(struct sw_flow *);
283 u64 ovs_flow_used_time(unsigned long flow_jiffies);
284 
285 int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
286 int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
287 			 struct sk_buff *skb,
288 			 struct sw_flow_key *key);
289 /* Extract key from packet coming from userspace. */
290 int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
291 				   struct sk_buff *skb,
292 				   struct sw_flow_key *key, bool log);
293 
294 #endif /* flow.h */
295