xref: /openbmc/linux/net/openvswitch/flow.h (revision 2891f2d5)
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 <net/inet_ecn.h>
35 #include <net/ip_tunnels.h>
36 #include <net/dst_metadata.h>
37 
38 struct sk_buff;
39 
40 enum sw_flow_mac_proto {
41 	MAC_PROTO_NONE = 0,
42 	MAC_PROTO_ETHERNET,
43 };
44 #define SW_FLOW_KEY_INVALID	0x80
45 
46 /* Store options at the end of the array if they are less than the
47  * maximum size. This allows us to get the benefits of variable length
48  * matching for small options.
49  */
50 #define TUN_METADATA_OFFSET(opt_len) \
51 	(FIELD_SIZEOF(struct sw_flow_key, tun_opts) - opt_len)
52 #define TUN_METADATA_OPTS(flow_key, opt_len) \
53 	((void *)((flow_key)->tun_opts + TUN_METADATA_OFFSET(opt_len)))
54 
55 struct ovs_tunnel_info {
56 	struct metadata_dst	*tun_dst;
57 };
58 
59 struct vlan_head {
60 	__be16 tpid; /* Vlan type. Generally 802.1q or 802.1ad.*/
61 	__be16 tci;  /* 0 if no VLAN, VLAN_TAG_PRESENT set otherwise. */
62 };
63 
64 #define OVS_SW_FLOW_KEY_METADATA_SIZE			\
65 	(offsetof(struct sw_flow_key, recirc_id) +	\
66 	FIELD_SIZEOF(struct sw_flow_key, recirc_id))
67 
68 struct sw_flow_key {
69 	u8 tun_opts[IP_TUNNEL_OPTS_MAX];
70 	u8 tun_opts_len;
71 	struct ip_tunnel_key tun_key;	/* Encapsulating tunnel key. */
72 	struct {
73 		u32	priority;	/* Packet QoS priority. */
74 		u32	skb_mark;	/* SKB mark. */
75 		u16	in_port;	/* Input switch port (or DP_MAX_PORTS). */
76 	} __packed phy; /* Safe when right after 'tun_key'. */
77 	u8 mac_proto;			/* MAC layer protocol (e.g. Ethernet). */
78 	u8 tun_proto;			/* Protocol of encapsulating tunnel. */
79 	u32 ovs_flow_hash;		/* Datapath computed hash value.  */
80 	u32 recirc_id;			/* Recirculation ID.  */
81 	struct {
82 		u8     src[ETH_ALEN];	/* Ethernet source address. */
83 		u8     dst[ETH_ALEN];	/* Ethernet destination address. */
84 		struct vlan_head vlan;
85 		struct vlan_head cvlan;
86 		__be16 type;		/* Ethernet frame type. */
87 	} eth;
88 	/* Filling a hole of two bytes. */
89 	u8 ct_state;
90 	u8 ct_orig_proto;		/* CT original direction tuple IP
91 					 * protocol.
92 					 */
93 	union {
94 		struct {
95 			__be32 top_lse;	/* top label stack entry */
96 		} mpls;
97 		struct {
98 			u8     proto;	/* IP protocol or lower 8 bits of ARP opcode. */
99 			u8     tos;	    /* IP ToS. */
100 			u8     ttl;	    /* IP TTL/hop limit. */
101 			u8     frag;	/* One of OVS_FRAG_TYPE_*. */
102 		} ip;
103 	};
104 	u16 ct_zone;			/* Conntrack zone. */
105 	struct {
106 		__be16 src;		/* TCP/UDP/SCTP source port. */
107 		__be16 dst;		/* TCP/UDP/SCTP destination port. */
108 		__be16 flags;		/* TCP flags. */
109 	} tp;
110 	union {
111 		struct {
112 			struct {
113 				__be32 src;	/* IP source address. */
114 				__be32 dst;	/* IP destination address. */
115 			} addr;
116 			union {
117 				struct {
118 					__be32 src;
119 					__be32 dst;
120 				} ct_orig;	/* Conntrack original direction fields. */
121 				struct {
122 					u8 sha[ETH_ALEN];	/* ARP source hardware address. */
123 					u8 tha[ETH_ALEN];	/* ARP target hardware address. */
124 				} arp;
125 			};
126 		} ipv4;
127 		struct {
128 			struct {
129 				struct in6_addr src;	/* IPv6 source address. */
130 				struct in6_addr dst;	/* IPv6 destination address. */
131 			} addr;
132 			__be32 label;			/* IPv6 flow label. */
133 			union {
134 				struct {
135 					struct in6_addr src;
136 					struct in6_addr dst;
137 				} ct_orig;	/* Conntrack original direction fields. */
138 				struct {
139 					struct in6_addr target;	/* ND target address. */
140 					u8 sll[ETH_ALEN];	/* ND source link layer address. */
141 					u8 tll[ETH_ALEN];	/* ND target link layer address. */
142 				} nd;
143 			};
144 		} ipv6;
145 	};
146 	struct {
147 		/* Connection tracking fields not packed above. */
148 		struct {
149 			__be16 src;	/* CT orig tuple tp src port. */
150 			__be16 dst;	/* CT orig tuple tp dst port. */
151 		} orig_tp;
152 		u32 mark;
153 		struct ovs_key_ct_labels labels;
154 	} ct;
155 
156 } __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
157 
158 static inline bool sw_flow_key_is_nd(const struct sw_flow_key *key)
159 {
160 	return key->eth.type == htons(ETH_P_IPV6) &&
161 		key->ip.proto == NEXTHDR_ICMP &&
162 		key->tp.dst == 0 &&
163 		(key->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
164 		 key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT));
165 }
166 
167 struct sw_flow_key_range {
168 	unsigned short int start;
169 	unsigned short int end;
170 };
171 
172 struct sw_flow_mask {
173 	int ref_count;
174 	struct rcu_head rcu;
175 	struct list_head list;
176 	struct sw_flow_key_range range;
177 	struct sw_flow_key key;
178 };
179 
180 struct sw_flow_match {
181 	struct sw_flow_key *key;
182 	struct sw_flow_key_range range;
183 	struct sw_flow_mask *mask;
184 };
185 
186 #define MAX_UFID_LENGTH 16 /* 128 bits */
187 
188 struct sw_flow_id {
189 	u32 ufid_len;
190 	union {
191 		u32 ufid[MAX_UFID_LENGTH / 4];
192 		struct sw_flow_key *unmasked_key;
193 	};
194 };
195 
196 struct sw_flow_actions {
197 	struct rcu_head rcu;
198 	size_t orig_len;	/* From flow_cmd_new netlink actions size */
199 	u32 actions_len;
200 	struct nlattr actions[];
201 };
202 
203 struct flow_stats {
204 	u64 packet_count;		/* Number of packets matched. */
205 	u64 byte_count;			/* Number of bytes matched. */
206 	unsigned long used;		/* Last used time (in jiffies). */
207 	spinlock_t lock;		/* Lock for atomic stats update. */
208 	__be16 tcp_flags;		/* Union of seen TCP flags. */
209 };
210 
211 struct sw_flow {
212 	struct rcu_head rcu;
213 	struct {
214 		struct hlist_node node[2];
215 		u32 hash;
216 	} flow_table, ufid_table;
217 	int stats_last_writer;		/* CPU id of the last writer on
218 					 * 'stats[0]'.
219 					 */
220 	struct sw_flow_key key;
221 	struct sw_flow_id id;
222 	struct sw_flow_mask *mask;
223 	struct sw_flow_actions __rcu *sf_acts;
224 	struct flow_stats __rcu *stats[]; /* One for each CPU.  First one
225 					   * is allocated at flow creation time,
226 					   * the rest are allocated on demand
227 					   * while holding the 'stats[0].lock'.
228 					   */
229 };
230 
231 struct arp_eth_header {
232 	__be16      ar_hrd;	/* format of hardware address   */
233 	__be16      ar_pro;	/* format of protocol address   */
234 	unsigned char   ar_hln;	/* length of hardware address   */
235 	unsigned char   ar_pln;	/* length of protocol address   */
236 	__be16      ar_op;	/* ARP opcode (command)     */
237 
238 	/* Ethernet+IPv4 specific members. */
239 	unsigned char       ar_sha[ETH_ALEN];	/* sender hardware address  */
240 	unsigned char       ar_sip[4];		/* sender IP address        */
241 	unsigned char       ar_tha[ETH_ALEN];	/* target hardware address  */
242 	unsigned char       ar_tip[4];		/* target IP address        */
243 } __packed;
244 
245 static inline u8 ovs_key_mac_proto(const struct sw_flow_key *key)
246 {
247 	return key->mac_proto & ~SW_FLOW_KEY_INVALID;
248 }
249 
250 static inline u16 __ovs_mac_header_len(u8 mac_proto)
251 {
252 	return mac_proto == MAC_PROTO_ETHERNET ? ETH_HLEN : 0;
253 }
254 
255 static inline u16 ovs_mac_header_len(const struct sw_flow_key *key)
256 {
257 	return __ovs_mac_header_len(ovs_key_mac_proto(key));
258 }
259 
260 static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
261 {
262 	return sfid->ufid_len;
263 }
264 
265 static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
266 {
267 	return !ovs_identifier_is_ufid(sfid);
268 }
269 
270 void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
271 			   const struct sk_buff *);
272 void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
273 			unsigned long *used, __be16 *tcp_flags);
274 void ovs_flow_stats_clear(struct sw_flow *);
275 u64 ovs_flow_used_time(unsigned long flow_jiffies);
276 
277 int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
278 int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
279 			 struct sk_buff *skb,
280 			 struct sw_flow_key *key);
281 /* Extract key from packet coming from userspace. */
282 int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
283 				   struct sk_buff *skb,
284 				   struct sw_flow_key *key, bool log);
285 
286 #endif /* flow.h */
287