Loading...
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#include <net/ip_tunnels.h>
36#include <net/dst_metadata.h>
37
38struct sk_buff;
39
40enum 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
55struct ovs_tunnel_info {
56 struct metadata_dst *tun_dst;
57};
58
59struct 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
68struct 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 union {
89 struct {
90 __be32 top_lse; /* top label stack entry */
91 } mpls;
92 struct {
93 u8 proto; /* IP protocol or lower 8 bits of ARP opcode. */
94 u8 tos; /* IP ToS. */
95 u8 ttl; /* IP TTL/hop limit. */
96 u8 frag; /* One of OVS_FRAG_TYPE_*. */
97 } ip;
98 };
99 struct {
100 __be16 src; /* TCP/UDP/SCTP source port. */
101 __be16 dst; /* TCP/UDP/SCTP destination port. */
102 __be16 flags; /* TCP flags. */
103 } tp;
104 union {
105 struct {
106 struct {
107 __be32 src; /* IP source address. */
108 __be32 dst; /* IP destination address. */
109 } addr;
110 struct {
111 u8 sha[ETH_ALEN]; /* ARP source hardware address. */
112 u8 tha[ETH_ALEN]; /* ARP target hardware address. */
113 } arp;
114 } ipv4;
115 struct {
116 struct {
117 struct in6_addr src; /* IPv6 source address. */
118 struct in6_addr dst; /* IPv6 destination address. */
119 } addr;
120 __be32 label; /* IPv6 flow label. */
121 struct {
122 struct in6_addr target; /* ND target address. */
123 u8 sll[ETH_ALEN]; /* ND source link layer address. */
124 u8 tll[ETH_ALEN]; /* ND target link layer address. */
125 } nd;
126 } ipv6;
127 };
128 struct {
129 /* Connection tracking fields. */
130 u16 zone;
131 u32 mark;
132 u8 state;
133 struct ovs_key_ct_labels labels;
134 } ct;
135
136} __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
137
138struct sw_flow_key_range {
139 unsigned short int start;
140 unsigned short int end;
141};
142
143struct sw_flow_mask {
144 int ref_count;
145 struct rcu_head rcu;
146 struct list_head list;
147 struct sw_flow_key_range range;
148 struct sw_flow_key key;
149};
150
151struct sw_flow_match {
152 struct sw_flow_key *key;
153 struct sw_flow_key_range range;
154 struct sw_flow_mask *mask;
155};
156
157#define MAX_UFID_LENGTH 16 /* 128 bits */
158
159struct sw_flow_id {
160 u32 ufid_len;
161 union {
162 u32 ufid[MAX_UFID_LENGTH / 4];
163 struct sw_flow_key *unmasked_key;
164 };
165};
166
167struct sw_flow_actions {
168 struct rcu_head rcu;
169 size_t orig_len; /* From flow_cmd_new netlink actions size */
170 u32 actions_len;
171 struct nlattr actions[];
172};
173
174struct flow_stats {
175 u64 packet_count; /* Number of packets matched. */
176 u64 byte_count; /* Number of bytes matched. */
177 unsigned long used; /* Last used time (in jiffies). */
178 spinlock_t lock; /* Lock for atomic stats update. */
179 __be16 tcp_flags; /* Union of seen TCP flags. */
180};
181
182struct sw_flow {
183 struct rcu_head rcu;
184 struct {
185 struct hlist_node node[2];
186 u32 hash;
187 } flow_table, ufid_table;
188 int stats_last_writer; /* CPU id of the last writer on
189 * 'stats[0]'.
190 */
191 struct sw_flow_key key;
192 struct sw_flow_id id;
193 struct sw_flow_mask *mask;
194 struct sw_flow_actions __rcu *sf_acts;
195 struct flow_stats __rcu *stats[]; /* One for each CPU. First one
196 * is allocated at flow creation time,
197 * the rest are allocated on demand
198 * while holding the 'stats[0].lock'.
199 */
200};
201
202struct arp_eth_header {
203 __be16 ar_hrd; /* format of hardware address */
204 __be16 ar_pro; /* format of protocol address */
205 unsigned char ar_hln; /* length of hardware address */
206 unsigned char ar_pln; /* length of protocol address */
207 __be16 ar_op; /* ARP opcode (command) */
208
209 /* Ethernet+IPv4 specific members. */
210 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
211 unsigned char ar_sip[4]; /* sender IP address */
212 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
213 unsigned char ar_tip[4]; /* target IP address */
214} __packed;
215
216static inline u8 ovs_key_mac_proto(const struct sw_flow_key *key)
217{
218 return key->mac_proto & ~SW_FLOW_KEY_INVALID;
219}
220
221static inline u16 __ovs_mac_header_len(u8 mac_proto)
222{
223 return mac_proto == MAC_PROTO_ETHERNET ? ETH_HLEN : 0;
224}
225
226static inline u16 ovs_mac_header_len(const struct sw_flow_key *key)
227{
228 return __ovs_mac_header_len(ovs_key_mac_proto(key));
229}
230
231static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
232{
233 return sfid->ufid_len;
234}
235
236static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
237{
238 return !ovs_identifier_is_ufid(sfid);
239}
240
241void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
242 const struct sk_buff *);
243void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
244 unsigned long *used, __be16 *tcp_flags);
245void ovs_flow_stats_clear(struct sw_flow *);
246u64 ovs_flow_used_time(unsigned long flow_jiffies);
247
248int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
249int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
250 struct sk_buff *skb,
251 struct sw_flow_key *key);
252/* Extract key from packet coming from userspace. */
253int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
254 struct sk_buff *skb,
255 struct sw_flow_key *key, bool log);
256
257#endif /* flow.h */
1/*
2 * Copyright (c) 2007-2011 Nicira Networks.
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/kernel.h>
23#include <linux/netlink.h>
24#include <linux/openvswitch.h>
25#include <linux/spinlock.h>
26#include <linux/types.h>
27#include <linux/rcupdate.h>
28#include <linux/if_ether.h>
29#include <linux/in6.h>
30#include <linux/jiffies.h>
31#include <linux/time.h>
32#include <linux/flex_array.h>
33#include <net/inet_ecn.h>
34
35struct sk_buff;
36
37struct sw_flow_actions {
38 struct rcu_head rcu;
39 u32 actions_len;
40 struct nlattr actions[];
41};
42
43struct sw_flow_key {
44 struct {
45 u32 priority; /* Packet QoS priority. */
46 u16 in_port; /* Input switch port (or USHRT_MAX). */
47 } phy;
48 struct {
49 u8 src[ETH_ALEN]; /* Ethernet source address. */
50 u8 dst[ETH_ALEN]; /* Ethernet destination address. */
51 __be16 tci; /* 0 if no VLAN, VLAN_TAG_PRESENT set otherwise. */
52 __be16 type; /* Ethernet frame type. */
53 } eth;
54 struct {
55 u8 proto; /* IP protocol or lower 8 bits of ARP opcode. */
56 u8 tos; /* IP ToS. */
57 u8 ttl; /* IP TTL/hop limit. */
58 u8 frag; /* One of OVS_FRAG_TYPE_*. */
59 } ip;
60 union {
61 struct {
62 struct {
63 __be32 src; /* IP source address. */
64 __be32 dst; /* IP destination address. */
65 } addr;
66 union {
67 struct {
68 __be16 src; /* TCP/UDP source port. */
69 __be16 dst; /* TCP/UDP destination port. */
70 } tp;
71 struct {
72 u8 sha[ETH_ALEN]; /* ARP source hardware address. */
73 u8 tha[ETH_ALEN]; /* ARP target hardware address. */
74 } arp;
75 };
76 } ipv4;
77 struct {
78 struct {
79 struct in6_addr src; /* IPv6 source address. */
80 struct in6_addr dst; /* IPv6 destination address. */
81 } addr;
82 __be32 label; /* IPv6 flow label. */
83 struct {
84 __be16 src; /* TCP/UDP source port. */
85 __be16 dst; /* TCP/UDP destination port. */
86 } tp;
87 struct {
88 struct in6_addr target; /* ND target address. */
89 u8 sll[ETH_ALEN]; /* ND source link layer address. */
90 u8 tll[ETH_ALEN]; /* ND target link layer address. */
91 } nd;
92 } ipv6;
93 };
94};
95
96struct sw_flow {
97 struct rcu_head rcu;
98 struct hlist_node hash_node[2];
99 u32 hash;
100
101 struct sw_flow_key key;
102 struct sw_flow_actions __rcu *sf_acts;
103
104 spinlock_t lock; /* Lock for values below. */
105 unsigned long used; /* Last used time (in jiffies). */
106 u64 packet_count; /* Number of packets matched. */
107 u64 byte_count; /* Number of bytes matched. */
108 u8 tcp_flags; /* Union of seen TCP flags. */
109};
110
111struct arp_eth_header {
112 __be16 ar_hrd; /* format of hardware address */
113 __be16 ar_pro; /* format of protocol address */
114 unsigned char ar_hln; /* length of hardware address */
115 unsigned char ar_pln; /* length of protocol address */
116 __be16 ar_op; /* ARP opcode (command) */
117
118 /* Ethernet+IPv4 specific members. */
119 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
120 unsigned char ar_sip[4]; /* sender IP address */
121 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
122 unsigned char ar_tip[4]; /* target IP address */
123} __packed;
124
125int ovs_flow_init(void);
126void ovs_flow_exit(void);
127
128struct sw_flow *ovs_flow_alloc(void);
129void ovs_flow_deferred_free(struct sw_flow *);
130void ovs_flow_free(struct sw_flow *flow);
131
132struct sw_flow_actions *ovs_flow_actions_alloc(const struct nlattr *);
133void ovs_flow_deferred_free_acts(struct sw_flow_actions *);
134
135int ovs_flow_extract(struct sk_buff *, u16 in_port, struct sw_flow_key *,
136 int *key_lenp);
137void ovs_flow_used(struct sw_flow *, struct sk_buff *);
138u64 ovs_flow_used_time(unsigned long flow_jiffies);
139
140/* Upper bound on the length of a nlattr-formatted flow key. The longest
141 * nlattr-formatted flow key would be:
142 *
143 * struct pad nl hdr total
144 * ------ --- ------ -----
145 * OVS_KEY_ATTR_PRIORITY 4 -- 4 8
146 * OVS_KEY_ATTR_IN_PORT 4 -- 4 8
147 * OVS_KEY_ATTR_ETHERNET 12 -- 4 16
148 * OVS_KEY_ATTR_8021Q 4 -- 4 8
149 * OVS_KEY_ATTR_ETHERTYPE 2 2 4 8
150 * OVS_KEY_ATTR_IPV6 40 -- 4 44
151 * OVS_KEY_ATTR_ICMPV6 2 2 4 8
152 * OVS_KEY_ATTR_ND 28 -- 4 32
153 * -------------------------------------------------
154 * total 132
155 */
156#define FLOW_BUFSIZE 132
157
158int ovs_flow_to_nlattrs(const struct sw_flow_key *, struct sk_buff *);
159int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
160 const struct nlattr *);
161int ovs_flow_metadata_from_nlattrs(u32 *priority, u16 *in_port,
162 const struct nlattr *);
163
164#define TBL_MIN_BUCKETS 1024
165
166struct flow_table {
167 struct flex_array *buckets;
168 unsigned int count, n_buckets;
169 struct rcu_head rcu;
170 int node_ver;
171 u32 hash_seed;
172 bool keep_flows;
173};
174
175static inline int ovs_flow_tbl_count(struct flow_table *table)
176{
177 return table->count;
178}
179
180static inline int ovs_flow_tbl_need_to_expand(struct flow_table *table)
181{
182 return (table->count > table->n_buckets);
183}
184
185struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *table,
186 struct sw_flow_key *key, int len);
187void ovs_flow_tbl_destroy(struct flow_table *table);
188void ovs_flow_tbl_deferred_destroy(struct flow_table *table);
189struct flow_table *ovs_flow_tbl_alloc(int new_size);
190struct flow_table *ovs_flow_tbl_expand(struct flow_table *table);
191struct flow_table *ovs_flow_tbl_rehash(struct flow_table *table);
192void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow);
193void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow);
194u32 ovs_flow_hash(const struct sw_flow_key *key, int key_len);
195
196struct sw_flow *ovs_flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *idx);
197extern const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1];
198
199#endif /* flow.h */