Loading...
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
26struct sk_buff;
27
28enum sw_flow_mac_proto {
29 MAC_PROTO_NONE = 0,
30 MAC_PROTO_ETHERNET,
31};
32#define SW_FLOW_KEY_INVALID 0x80
33
34/* Store options at the end of the array if they are less than the
35 * maximum size. This allows us to get the benefits of variable length
36 * matching for small options.
37 */
38#define TUN_METADATA_OFFSET(opt_len) \
39 (FIELD_SIZEOF(struct sw_flow_key, tun_opts) - opt_len)
40#define TUN_METADATA_OPTS(flow_key, opt_len) \
41 ((void *)((flow_key)->tun_opts + TUN_METADATA_OFFSET(opt_len)))
42
43struct ovs_tunnel_info {
44 struct metadata_dst *tun_dst;
45};
46
47struct vlan_head {
48 __be16 tpid; /* Vlan type. Generally 802.1q or 802.1ad.*/
49 __be16 tci; /* 0 if no VLAN, VLAN_CFI_MASK set otherwise. */
50};
51
52#define OVS_SW_FLOW_KEY_METADATA_SIZE \
53 (offsetof(struct sw_flow_key, recirc_id) + \
54 FIELD_SIZEOF(struct sw_flow_key, recirc_id))
55
56struct ovs_key_nsh {
57 struct ovs_nsh_key_base base;
58 __be32 context[NSH_MD1_CONTEXT_SIZE];
59};
60
61struct sw_flow_key {
62 u8 tun_opts[IP_TUNNEL_OPTS_MAX];
63 u8 tun_opts_len;
64 struct ip_tunnel_key tun_key; /* Encapsulating tunnel key. */
65 struct {
66 u32 priority; /* Packet QoS priority. */
67 u32 skb_mark; /* SKB mark. */
68 u16 in_port; /* Input switch port (or DP_MAX_PORTS). */
69 } __packed phy; /* Safe when right after 'tun_key'. */
70 u8 mac_proto; /* MAC layer protocol (e.g. Ethernet). */
71 u8 tun_proto; /* Protocol of encapsulating tunnel. */
72 u32 ovs_flow_hash; /* Datapath computed hash value. */
73 u32 recirc_id; /* Recirculation ID. */
74 struct {
75 u8 src[ETH_ALEN]; /* Ethernet source address. */
76 u8 dst[ETH_ALEN]; /* Ethernet destination address. */
77 struct vlan_head vlan;
78 struct vlan_head cvlan;
79 __be16 type; /* Ethernet frame type. */
80 } eth;
81 /* Filling a hole of two bytes. */
82 u8 ct_state;
83 u8 ct_orig_proto; /* CT original direction tuple IP
84 * protocol.
85 */
86 union {
87 struct {
88 __be32 top_lse; /* top label stack entry */
89 } mpls;
90 struct {
91 u8 proto; /* IP protocol or lower 8 bits of ARP opcode. */
92 u8 tos; /* IP ToS. */
93 u8 ttl; /* IP TTL/hop limit. */
94 u8 frag; /* One of OVS_FRAG_TYPE_*. */
95 } ip;
96 };
97 u16 ct_zone; /* Conntrack zone. */
98 struct {
99 __be16 src; /* TCP/UDP/SCTP source port. */
100 __be16 dst; /* TCP/UDP/SCTP destination port. */
101 __be16 flags; /* TCP flags. */
102 } tp;
103 union {
104 struct {
105 struct {
106 __be32 src; /* IP source address. */
107 __be32 dst; /* IP destination address. */
108 } addr;
109 union {
110 struct {
111 __be32 src;
112 __be32 dst;
113 } ct_orig; /* Conntrack original direction fields. */
114 struct {
115 u8 sha[ETH_ALEN]; /* ARP source hardware address. */
116 u8 tha[ETH_ALEN]; /* ARP target hardware address. */
117 } arp;
118 };
119 } ipv4;
120 struct {
121 struct {
122 struct in6_addr src; /* IPv6 source address. */
123 struct in6_addr dst; /* IPv6 destination address. */
124 } addr;
125 __be32 label; /* IPv6 flow label. */
126 union {
127 struct {
128 struct in6_addr src;
129 struct in6_addr dst;
130 } ct_orig; /* Conntrack original direction fields. */
131 struct {
132 struct in6_addr target; /* ND target address. */
133 u8 sll[ETH_ALEN]; /* ND source link layer address. */
134 u8 tll[ETH_ALEN]; /* ND target link layer address. */
135 } nd;
136 };
137 } ipv6;
138 struct ovs_key_nsh nsh; /* network service header */
139 };
140 struct {
141 /* Connection tracking fields not packed above. */
142 struct {
143 __be16 src; /* CT orig tuple tp src port. */
144 __be16 dst; /* CT orig tuple tp dst port. */
145 } orig_tp;
146 u32 mark;
147 struct ovs_key_ct_labels labels;
148 } ct;
149
150} __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
151
152static inline bool sw_flow_key_is_nd(const struct sw_flow_key *key)
153{
154 return key->eth.type == htons(ETH_P_IPV6) &&
155 key->ip.proto == NEXTHDR_ICMP &&
156 key->tp.dst == 0 &&
157 (key->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
158 key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT));
159}
160
161struct sw_flow_key_range {
162 unsigned short int start;
163 unsigned short int end;
164};
165
166struct sw_flow_mask {
167 int ref_count;
168 struct rcu_head rcu;
169 struct list_head list;
170 struct sw_flow_key_range range;
171 struct sw_flow_key key;
172};
173
174struct sw_flow_match {
175 struct sw_flow_key *key;
176 struct sw_flow_key_range range;
177 struct sw_flow_mask *mask;
178};
179
180#define MAX_UFID_LENGTH 16 /* 128 bits */
181
182struct sw_flow_id {
183 u32 ufid_len;
184 union {
185 u32 ufid[MAX_UFID_LENGTH / 4];
186 struct sw_flow_key *unmasked_key;
187 };
188};
189
190struct sw_flow_actions {
191 struct rcu_head rcu;
192 size_t orig_len; /* From flow_cmd_new netlink actions size */
193 u32 actions_len;
194 struct nlattr actions[];
195};
196
197struct sw_flow_stats {
198 u64 packet_count; /* Number of packets matched. */
199 u64 byte_count; /* Number of bytes matched. */
200 unsigned long used; /* Last used time (in jiffies). */
201 spinlock_t lock; /* Lock for atomic stats update. */
202 __be16 tcp_flags; /* Union of seen TCP flags. */
203};
204
205struct sw_flow {
206 struct rcu_head rcu;
207 struct {
208 struct hlist_node node[2];
209 u32 hash;
210 } flow_table, ufid_table;
211 int stats_last_writer; /* CPU id of the last writer on
212 * 'stats[0]'.
213 */
214 struct sw_flow_key key;
215 struct sw_flow_id id;
216 struct cpumask cpu_used_mask;
217 struct sw_flow_mask *mask;
218 struct sw_flow_actions __rcu *sf_acts;
219 struct sw_flow_stats __rcu *stats[]; /* One for each CPU. First one
220 * is allocated at flow creation time,
221 * the rest are allocated on demand
222 * while holding the 'stats[0].lock'.
223 */
224};
225
226struct arp_eth_header {
227 __be16 ar_hrd; /* format of hardware address */
228 __be16 ar_pro; /* format of protocol address */
229 unsigned char ar_hln; /* length of hardware address */
230 unsigned char ar_pln; /* length of protocol address */
231 __be16 ar_op; /* ARP opcode (command) */
232
233 /* Ethernet+IPv4 specific members. */
234 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
235 unsigned char ar_sip[4]; /* sender IP address */
236 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
237 unsigned char ar_tip[4]; /* target IP address */
238} __packed;
239
240static inline u8 ovs_key_mac_proto(const struct sw_flow_key *key)
241{
242 return key->mac_proto & ~SW_FLOW_KEY_INVALID;
243}
244
245static inline u16 __ovs_mac_header_len(u8 mac_proto)
246{
247 return mac_proto == MAC_PROTO_ETHERNET ? ETH_HLEN : 0;
248}
249
250static inline u16 ovs_mac_header_len(const struct sw_flow_key *key)
251{
252 return __ovs_mac_header_len(ovs_key_mac_proto(key));
253}
254
255static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
256{
257 return sfid->ufid_len;
258}
259
260static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
261{
262 return !ovs_identifier_is_ufid(sfid);
263}
264
265void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
266 const struct sk_buff *);
267void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
268 unsigned long *used, __be16 *tcp_flags);
269void ovs_flow_stats_clear(struct sw_flow *);
270u64 ovs_flow_used_time(unsigned long flow_jiffies);
271
272int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
273int ovs_flow_key_update_l3l4(struct sk_buff *skb, struct sw_flow_key *key);
274int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
275 struct sk_buff *skb,
276 struct sw_flow_key *key);
277/* Extract key from packet coming from userspace. */
278int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
279 struct sk_buff *skb,
280 struct sw_flow_key *key, bool log);
281
282#endif /* flow.h */
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
40/* Store options at the end of the array if they are less than the
41 * maximum size. This allows us to get the benefits of variable length
42 * matching for small options.
43 */
44#define TUN_METADATA_OFFSET(opt_len) \
45 (FIELD_SIZEOF(struct sw_flow_key, tun_opts) - opt_len)
46#define TUN_METADATA_OPTS(flow_key, opt_len) \
47 ((void *)((flow_key)->tun_opts + TUN_METADATA_OFFSET(opt_len)))
48
49struct ovs_tunnel_info {
50 struct metadata_dst *tun_dst;
51};
52
53#define OVS_SW_FLOW_KEY_METADATA_SIZE \
54 (offsetof(struct sw_flow_key, recirc_id) + \
55 FIELD_SIZEOF(struct sw_flow_key, recirc_id))
56
57struct sw_flow_key {
58 u8 tun_opts[IP_TUNNEL_OPTS_MAX];
59 u8 tun_opts_len;
60 struct ip_tunnel_key tun_key; /* Encapsulating tunnel key. */
61 struct {
62 u32 priority; /* Packet QoS priority. */
63 u32 skb_mark; /* SKB mark. */
64 u16 in_port; /* Input switch port (or DP_MAX_PORTS). */
65 } __packed phy; /* Safe when right after 'tun_key'. */
66 u8 tun_proto; /* Protocol of encapsulating tunnel. */
67 u32 ovs_flow_hash; /* Datapath computed hash value. */
68 u32 recirc_id; /* Recirculation ID. */
69 struct {
70 u8 src[ETH_ALEN]; /* Ethernet source address. */
71 u8 dst[ETH_ALEN]; /* Ethernet destination address. */
72 __be16 tci; /* 0 if no VLAN, VLAN_TAG_PRESENT set otherwise. */
73 __be16 type; /* Ethernet frame type. */
74 } eth;
75 union {
76 struct {
77 __be32 top_lse; /* top label stack entry */
78 } mpls;
79 struct {
80 u8 proto; /* IP protocol or lower 8 bits of ARP opcode. */
81 u8 tos; /* IP ToS. */
82 u8 ttl; /* IP TTL/hop limit. */
83 u8 frag; /* One of OVS_FRAG_TYPE_*. */
84 } ip;
85 };
86 struct {
87 __be16 src; /* TCP/UDP/SCTP source port. */
88 __be16 dst; /* TCP/UDP/SCTP destination port. */
89 __be16 flags; /* TCP flags. */
90 } tp;
91 union {
92 struct {
93 struct {
94 __be32 src; /* IP source address. */
95 __be32 dst; /* IP destination address. */
96 } addr;
97 struct {
98 u8 sha[ETH_ALEN]; /* ARP source hardware address. */
99 u8 tha[ETH_ALEN]; /* ARP target hardware address. */
100 } arp;
101 } ipv4;
102 struct {
103 struct {
104 struct in6_addr src; /* IPv6 source address. */
105 struct in6_addr dst; /* IPv6 destination address. */
106 } addr;
107 __be32 label; /* IPv6 flow label. */
108 struct {
109 struct in6_addr target; /* ND target address. */
110 u8 sll[ETH_ALEN]; /* ND source link layer address. */
111 u8 tll[ETH_ALEN]; /* ND target link layer address. */
112 } nd;
113 } ipv6;
114 };
115 struct {
116 /* Connection tracking fields. */
117 u16 zone;
118 u32 mark;
119 u8 state;
120 struct ovs_key_ct_labels labels;
121 } ct;
122
123} __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
124
125struct sw_flow_key_range {
126 unsigned short int start;
127 unsigned short int end;
128};
129
130struct sw_flow_mask {
131 int ref_count;
132 struct rcu_head rcu;
133 struct list_head list;
134 struct sw_flow_key_range range;
135 struct sw_flow_key key;
136};
137
138struct sw_flow_match {
139 struct sw_flow_key *key;
140 struct sw_flow_key_range range;
141 struct sw_flow_mask *mask;
142};
143
144#define MAX_UFID_LENGTH 16 /* 128 bits */
145
146struct sw_flow_id {
147 u32 ufid_len;
148 union {
149 u32 ufid[MAX_UFID_LENGTH / 4];
150 struct sw_flow_key *unmasked_key;
151 };
152};
153
154struct sw_flow_actions {
155 struct rcu_head rcu;
156 size_t orig_len; /* From flow_cmd_new netlink actions size */
157 u32 actions_len;
158 struct nlattr actions[];
159};
160
161struct flow_stats {
162 u64 packet_count; /* Number of packets matched. */
163 u64 byte_count; /* Number of bytes matched. */
164 unsigned long used; /* Last used time (in jiffies). */
165 spinlock_t lock; /* Lock for atomic stats update. */
166 __be16 tcp_flags; /* Union of seen TCP flags. */
167};
168
169struct sw_flow {
170 struct rcu_head rcu;
171 struct {
172 struct hlist_node node[2];
173 u32 hash;
174 } flow_table, ufid_table;
175 int stats_last_writer; /* NUMA-node id of the last writer on
176 * 'stats[0]'.
177 */
178 struct sw_flow_key key;
179 struct sw_flow_id id;
180 struct sw_flow_mask *mask;
181 struct sw_flow_actions __rcu *sf_acts;
182 struct flow_stats __rcu *stats[]; /* One for each NUMA node. First one
183 * is allocated at flow creation time,
184 * the rest are allocated on demand
185 * while holding the 'stats[0].lock'.
186 */
187};
188
189struct arp_eth_header {
190 __be16 ar_hrd; /* format of hardware address */
191 __be16 ar_pro; /* format of protocol address */
192 unsigned char ar_hln; /* length of hardware address */
193 unsigned char ar_pln; /* length of protocol address */
194 __be16 ar_op; /* ARP opcode (command) */
195
196 /* Ethernet+IPv4 specific members. */
197 unsigned char ar_sha[ETH_ALEN]; /* sender hardware address */
198 unsigned char ar_sip[4]; /* sender IP address */
199 unsigned char ar_tha[ETH_ALEN]; /* target hardware address */
200 unsigned char ar_tip[4]; /* target IP address */
201} __packed;
202
203static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
204{
205 return sfid->ufid_len;
206}
207
208static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
209{
210 return !ovs_identifier_is_ufid(sfid);
211}
212
213void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
214 const struct sk_buff *);
215void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
216 unsigned long *used, __be16 *tcp_flags);
217void ovs_flow_stats_clear(struct sw_flow *);
218u64 ovs_flow_used_time(unsigned long flow_jiffies);
219
220int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
221int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
222 struct sk_buff *skb,
223 struct sw_flow_key *key);
224/* Extract key from packet coming from userspace. */
225int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
226 struct sk_buff *skb,
227 struct sw_flow_key *key, bool log);
228
229#endif /* flow.h */