xref: /linux/net/openvswitch/flow.h (revision e0bf6c5ca2d3281f231c5f0c9bf145e9513644de)
1 /*
2  * Copyright (c) 2007-2014 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 
36 struct sk_buff;
37 
38 /* Used to memset ovs_key_ipv4_tunnel padding. */
39 #define OVS_TUNNEL_KEY_SIZE					\
40 	(offsetof(struct ovs_key_ipv4_tunnel, tp_dst) +		\
41 	 FIELD_SIZEOF(struct ovs_key_ipv4_tunnel, tp_dst))
42 
43 struct ovs_key_ipv4_tunnel {
44 	__be64 tun_id;
45 	__be32 ipv4_src;
46 	__be32 ipv4_dst;
47 	__be16 tun_flags;
48 	u8   ipv4_tos;
49 	u8   ipv4_ttl;
50 	__be16 tp_src;
51 	__be16 tp_dst;
52 } __packed __aligned(4); /* Minimize padding. */
53 
54 struct ovs_tunnel_info {
55 	struct ovs_key_ipv4_tunnel tunnel;
56 	const void *options;
57 	u8 options_len;
58 };
59 
60 /* Store options at the end of the array if they are less than the
61  * maximum size. This allows us to get the benefits of variable length
62  * matching for small options.
63  */
64 #define TUN_METADATA_OFFSET(opt_len) \
65 	(FIELD_SIZEOF(struct sw_flow_key, tun_opts) - opt_len)
66 #define TUN_METADATA_OPTS(flow_key, opt_len) \
67 	((void *)((flow_key)->tun_opts + TUN_METADATA_OFFSET(opt_len)))
68 
69 static inline void __ovs_flow_tun_info_init(struct ovs_tunnel_info *tun_info,
70 					    __be32 saddr, __be32 daddr,
71 					    u8 tos, u8 ttl,
72 					    __be16 tp_src,
73 					    __be16 tp_dst,
74 					    __be64 tun_id,
75 					    __be16 tun_flags,
76 					    const void *opts,
77 					    u8 opts_len)
78 {
79 	tun_info->tunnel.tun_id = tun_id;
80 	tun_info->tunnel.ipv4_src = saddr;
81 	tun_info->tunnel.ipv4_dst = daddr;
82 	tun_info->tunnel.ipv4_tos = tos;
83 	tun_info->tunnel.ipv4_ttl = ttl;
84 	tun_info->tunnel.tun_flags = tun_flags;
85 
86 	/* For the tunnel types on the top of IPsec, the tp_src and tp_dst of
87 	 * the upper tunnel are used.
88 	 * E.g: GRE over IPSEC, the tp_src and tp_port are zero.
89 	 */
90 	tun_info->tunnel.tp_src = tp_src;
91 	tun_info->tunnel.tp_dst = tp_dst;
92 
93 	/* Clear struct padding. */
94 	if (sizeof(tun_info->tunnel) != OVS_TUNNEL_KEY_SIZE)
95 		memset((unsigned char *)&tun_info->tunnel + OVS_TUNNEL_KEY_SIZE,
96 		       0, sizeof(tun_info->tunnel) - OVS_TUNNEL_KEY_SIZE);
97 
98 	tun_info->options = opts;
99 	tun_info->options_len = opts_len;
100 }
101 
102 static inline void ovs_flow_tun_info_init(struct ovs_tunnel_info *tun_info,
103 					  const struct iphdr *iph,
104 					  __be16 tp_src,
105 					  __be16 tp_dst,
106 					  __be64 tun_id,
107 					  __be16 tun_flags,
108 					  const void *opts,
109 					  u8 opts_len)
110 {
111 	__ovs_flow_tun_info_init(tun_info, iph->saddr, iph->daddr,
112 				 iph->tos, iph->ttl,
113 				 tp_src, tp_dst,
114 				 tun_id, tun_flags,
115 				 opts, opts_len);
116 }
117 
118 #define OVS_SW_FLOW_KEY_METADATA_SIZE			\
119 	(offsetof(struct sw_flow_key, recirc_id) +	\
120 	FIELD_SIZEOF(struct sw_flow_key, recirc_id))
121 
122 struct sw_flow_key {
123 	u8 tun_opts[255];
124 	u8 tun_opts_len;
125 	struct ovs_key_ipv4_tunnel tun_key;  /* Encapsulating tunnel key. */
126 	struct {
127 		u32	priority;	/* Packet QoS priority. */
128 		u32	skb_mark;	/* SKB mark. */
129 		u16	in_port;	/* Input switch port (or DP_MAX_PORTS). */
130 	} __packed phy; /* Safe when right after 'tun_key'. */
131 	u32 ovs_flow_hash;		/* Datapath computed hash value.  */
132 	u32 recirc_id;			/* Recirculation ID.  */
133 	struct {
134 		u8     src[ETH_ALEN];	/* Ethernet source address. */
135 		u8     dst[ETH_ALEN];	/* Ethernet destination address. */
136 		__be16 tci;		/* 0 if no VLAN, VLAN_TAG_PRESENT set otherwise. */
137 		__be16 type;		/* Ethernet frame type. */
138 	} eth;
139 	union {
140 		struct {
141 			__be32 top_lse;	/* top label stack entry */
142 		} mpls;
143 		struct {
144 			u8     proto;	/* IP protocol or lower 8 bits of ARP opcode. */
145 			u8     tos;	    /* IP ToS. */
146 			u8     ttl;	    /* IP TTL/hop limit. */
147 			u8     frag;	/* One of OVS_FRAG_TYPE_*. */
148 		} ip;
149 	};
150 	struct {
151 		__be16 src;		/* TCP/UDP/SCTP source port. */
152 		__be16 dst;		/* TCP/UDP/SCTP destination port. */
153 		__be16 flags;		/* TCP flags. */
154 	} tp;
155 	union {
156 		struct {
157 			struct {
158 				__be32 src;	/* IP source address. */
159 				__be32 dst;	/* IP destination address. */
160 			} addr;
161 			struct {
162 				u8 sha[ETH_ALEN];	/* ARP source hardware address. */
163 				u8 tha[ETH_ALEN];	/* ARP target hardware address. */
164 			} arp;
165 		} ipv4;
166 		struct {
167 			struct {
168 				struct in6_addr src;	/* IPv6 source address. */
169 				struct in6_addr dst;	/* IPv6 destination address. */
170 			} addr;
171 			__be32 label;			/* IPv6 flow label. */
172 			struct {
173 				struct in6_addr target;	/* ND target address. */
174 				u8 sll[ETH_ALEN];	/* ND source link layer address. */
175 				u8 tll[ETH_ALEN];	/* ND target link layer address. */
176 			} nd;
177 		} ipv6;
178 	};
179 } __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
180 
181 struct sw_flow_key_range {
182 	unsigned short int start;
183 	unsigned short int end;
184 };
185 
186 struct sw_flow_mask {
187 	int ref_count;
188 	struct rcu_head rcu;
189 	struct list_head list;
190 	struct sw_flow_key_range range;
191 	struct sw_flow_key key;
192 };
193 
194 struct sw_flow_match {
195 	struct sw_flow_key *key;
196 	struct sw_flow_key_range range;
197 	struct sw_flow_mask *mask;
198 };
199 
200 #define MAX_UFID_LENGTH 16 /* 128 bits */
201 
202 struct sw_flow_id {
203 	u32 ufid_len;
204 	union {
205 		u32 ufid[MAX_UFID_LENGTH / 4];
206 		struct sw_flow_key *unmasked_key;
207 	};
208 };
209 
210 struct sw_flow_actions {
211 	struct rcu_head rcu;
212 	u32 actions_len;
213 	struct nlattr actions[];
214 };
215 
216 struct flow_stats {
217 	u64 packet_count;		/* Number of packets matched. */
218 	u64 byte_count;			/* Number of bytes matched. */
219 	unsigned long used;		/* Last used time (in jiffies). */
220 	spinlock_t lock;		/* Lock for atomic stats update. */
221 	__be16 tcp_flags;		/* Union of seen TCP flags. */
222 };
223 
224 struct sw_flow {
225 	struct rcu_head rcu;
226 	struct {
227 		struct hlist_node node[2];
228 		u32 hash;
229 	} flow_table, ufid_table;
230 	int stats_last_writer;		/* NUMA-node id of the last writer on
231 					 * 'stats[0]'.
232 					 */
233 	struct sw_flow_key key;
234 	struct sw_flow_id id;
235 	struct sw_flow_mask *mask;
236 	struct sw_flow_actions __rcu *sf_acts;
237 	struct flow_stats __rcu *stats[]; /* One for each NUMA node.  First one
238 					   * is allocated at flow creation time,
239 					   * the rest are allocated on demand
240 					   * while holding the 'stats[0].lock'.
241 					   */
242 };
243 
244 struct arp_eth_header {
245 	__be16      ar_hrd;	/* format of hardware address   */
246 	__be16      ar_pro;	/* format of protocol address   */
247 	unsigned char   ar_hln;	/* length of hardware address   */
248 	unsigned char   ar_pln;	/* length of protocol address   */
249 	__be16      ar_op;	/* ARP opcode (command)     */
250 
251 	/* Ethernet+IPv4 specific members. */
252 	unsigned char       ar_sha[ETH_ALEN];	/* sender hardware address  */
253 	unsigned char       ar_sip[4];		/* sender IP address        */
254 	unsigned char       ar_tha[ETH_ALEN];	/* target hardware address  */
255 	unsigned char       ar_tip[4];		/* target IP address        */
256 } __packed;
257 
258 static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
259 {
260 	return sfid->ufid_len;
261 }
262 
263 static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
264 {
265 	return !ovs_identifier_is_ufid(sfid);
266 }
267 
268 void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
269 			   const struct sk_buff *);
270 void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
271 			unsigned long *used, __be16 *tcp_flags);
272 void ovs_flow_stats_clear(struct sw_flow *);
273 u64 ovs_flow_used_time(unsigned long flow_jiffies);
274 
275 int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
276 int ovs_flow_key_extract(const struct ovs_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(const struct nlattr *attr,
281 				   struct sk_buff *skb,
282 				   struct sw_flow_key *key, bool log);
283 
284 #endif /* flow.h */
285