xref: /kernel/linux/linux-5.10/net/openvswitch/flow.h (revision 8c2ecf20)
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#define MPLS_LABEL_DEPTH       3
34
35/* Store options at the end of the array if they are less than the
36 * maximum size. This allows us to get the benefits of variable length
37 * matching for small options.
38 */
39#define TUN_METADATA_OFFSET(opt_len) \
40	(sizeof_field(struct sw_flow_key, tun_opts) - opt_len)
41#define TUN_METADATA_OPTS(flow_key, opt_len) \
42	((void *)((flow_key)->tun_opts + TUN_METADATA_OFFSET(opt_len)))
43
44struct ovs_tunnel_info {
45	struct metadata_dst	*tun_dst;
46};
47
48struct vlan_head {
49	__be16 tpid; /* Vlan type. Generally 802.1q or 802.1ad.*/
50	__be16 tci;  /* 0 if no VLAN, VLAN_CFI_MASK set otherwise. */
51};
52
53#define OVS_SW_FLOW_KEY_METADATA_SIZE			\
54	(offsetof(struct sw_flow_key, recirc_id) +	\
55	sizeof_field(struct sw_flow_key, recirc_id))
56
57struct ovs_key_nsh {
58	struct ovs_nsh_key_base base;
59	__be32 context[NSH_MD1_CONTEXT_SIZE];
60};
61
62struct sw_flow_key {
63	u8 tun_opts[IP_TUNNEL_OPTS_MAX];
64	u8 tun_opts_len;
65	struct ip_tunnel_key tun_key;	/* Encapsulating tunnel key. */
66	struct {
67		u32	priority;	/* Packet QoS priority. */
68		u32	skb_mark;	/* SKB mark. */
69		u16	in_port;	/* Input switch port (or DP_MAX_PORTS). */
70	} __packed phy; /* Safe when right after 'tun_key'. */
71	u8 mac_proto;			/* MAC layer protocol (e.g. Ethernet). */
72	u8 tun_proto;			/* Protocol of encapsulating tunnel. */
73	u32 ovs_flow_hash;		/* Datapath computed hash value.  */
74	u32 recirc_id;			/* Recirculation ID.  */
75	struct {
76		u8     src[ETH_ALEN];	/* Ethernet source address. */
77		u8     dst[ETH_ALEN];	/* Ethernet destination address. */
78		struct vlan_head vlan;
79		struct vlan_head cvlan;
80		__be16 type;		/* Ethernet frame type. */
81	} eth;
82	/* Filling a hole of two bytes. */
83	u8 ct_state;
84	u8 ct_orig_proto;		/* CT original direction tuple IP
85					 * protocol.
86					 */
87	union {
88		struct {
89			u8     proto;	/* IP protocol or lower 8 bits of ARP opcode. */
90			u8     tos;	    /* IP ToS. */
91			u8     ttl;	    /* IP TTL/hop limit. */
92			u8     frag;	/* One of OVS_FRAG_TYPE_*. */
93		} ip;
94	};
95	u16 ct_zone;			/* Conntrack zone. */
96	struct {
97		__be16 src;		/* TCP/UDP/SCTP source port. */
98		__be16 dst;		/* TCP/UDP/SCTP destination port. */
99		__be16 flags;		/* TCP flags. */
100	} tp;
101	union {
102		struct {
103			struct {
104				__be32 src;	/* IP source address. */
105				__be32 dst;	/* IP destination address. */
106			} addr;
107			union {
108				struct {
109					__be32 src;
110					__be32 dst;
111				} ct_orig;	/* Conntrack original direction fields. */
112				struct {
113					u8 sha[ETH_ALEN];	/* ARP source hardware address. */
114					u8 tha[ETH_ALEN];	/* ARP target hardware address. */
115				} arp;
116			};
117		} ipv4;
118		struct {
119			struct {
120				struct in6_addr src;	/* IPv6 source address. */
121				struct in6_addr dst;	/* IPv6 destination address. */
122			} addr;
123			__be32 label;			/* IPv6 flow label. */
124			union {
125				struct {
126					struct in6_addr src;
127					struct in6_addr dst;
128				} ct_orig;	/* Conntrack original direction fields. */
129				struct {
130					struct in6_addr target;	/* ND target address. */
131					u8 sll[ETH_ALEN];	/* ND source link layer address. */
132					u8 tll[ETH_ALEN];	/* ND target link layer address. */
133				} nd;
134			};
135		} ipv6;
136		struct {
137			u32 num_labels_mask;    /* labels present bitmap of effective length MPLS_LABEL_DEPTH */
138			__be32 lse[MPLS_LABEL_DEPTH];     /* label stack entry  */
139		} mpls;
140
141		struct ovs_key_nsh nsh;         /* network service header */
142	};
143	struct {
144		/* Connection tracking fields not packed above. */
145		struct {
146			__be16 src;	/* CT orig tuple tp src port. */
147			__be16 dst;	/* CT orig tuple tp dst port. */
148		} orig_tp;
149		u32 mark;
150		struct ovs_key_ct_labels labels;
151	} ct;
152
153} __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
154
155static inline bool sw_flow_key_is_nd(const struct sw_flow_key *key)
156{
157	return key->eth.type == htons(ETH_P_IPV6) &&
158		key->ip.proto == NEXTHDR_ICMP &&
159		key->tp.dst == 0 &&
160		(key->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
161		 key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT));
162}
163
164struct sw_flow_key_range {
165	unsigned short int start;
166	unsigned short int end;
167};
168
169struct sw_flow_mask {
170	int ref_count;
171	struct rcu_head rcu;
172	struct sw_flow_key_range range;
173	struct sw_flow_key key;
174};
175
176struct sw_flow_match {
177	struct sw_flow_key *key;
178	struct sw_flow_key_range range;
179	struct sw_flow_mask *mask;
180};
181
182#define MAX_UFID_LENGTH 16 /* 128 bits */
183
184struct sw_flow_id {
185	u32 ufid_len;
186	union {
187		u32 ufid[MAX_UFID_LENGTH / 4];
188		struct sw_flow_key *unmasked_key;
189	};
190};
191
192struct sw_flow_actions {
193	struct rcu_head rcu;
194	size_t orig_len;	/* From flow_cmd_new netlink actions size */
195	u32 actions_len;
196	struct nlattr actions[];
197};
198
199struct sw_flow_stats {
200	u64 packet_count;		/* Number of packets matched. */
201	u64 byte_count;			/* Number of bytes matched. */
202	unsigned long used;		/* Last used time (in jiffies). */
203	spinlock_t lock;		/* Lock for atomic stats update. */
204	__be16 tcp_flags;		/* Union of seen TCP flags. */
205};
206
207struct sw_flow {
208	struct rcu_head rcu;
209	struct {
210		struct hlist_node node[2];
211		u32 hash;
212	} flow_table, ufid_table;
213	int stats_last_writer;		/* CPU id of the last writer on
214					 * 'stats[0]'.
215					 */
216	struct sw_flow_key key;
217	struct sw_flow_id id;
218	struct cpumask cpu_used_mask;
219	struct sw_flow_mask *mask;
220	struct sw_flow_actions __rcu *sf_acts;
221	struct sw_flow_stats __rcu *stats[]; /* One for each CPU.  First one
222					   * is allocated at flow creation time,
223					   * the rest are allocated on demand
224					   * while holding the 'stats[0].lock'.
225					   */
226};
227
228struct arp_eth_header {
229	__be16      ar_hrd;	/* format of hardware address   */
230	__be16      ar_pro;	/* format of protocol address   */
231	unsigned char   ar_hln;	/* length of hardware address   */
232	unsigned char   ar_pln;	/* length of protocol address   */
233	__be16      ar_op;	/* ARP opcode (command)     */
234
235	/* Ethernet+IPv4 specific members. */
236	unsigned char       ar_sha[ETH_ALEN];	/* sender hardware address  */
237	unsigned char       ar_sip[4];		/* sender IP address        */
238	unsigned char       ar_tha[ETH_ALEN];	/* target hardware address  */
239	unsigned char       ar_tip[4];		/* target IP address        */
240} __packed;
241
242static inline u8 ovs_key_mac_proto(const struct sw_flow_key *key)
243{
244	return key->mac_proto & ~SW_FLOW_KEY_INVALID;
245}
246
247static inline u16 __ovs_mac_header_len(u8 mac_proto)
248{
249	return mac_proto == MAC_PROTO_ETHERNET ? ETH_HLEN : 0;
250}
251
252static inline u16 ovs_mac_header_len(const struct sw_flow_key *key)
253{
254	return __ovs_mac_header_len(ovs_key_mac_proto(key));
255}
256
257static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
258{
259	return sfid->ufid_len;
260}
261
262static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
263{
264	return !ovs_identifier_is_ufid(sfid);
265}
266
267void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
268			   const struct sk_buff *);
269void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,
270			unsigned long *used, __be16 *tcp_flags);
271void ovs_flow_stats_clear(struct sw_flow *);
272u64 ovs_flow_used_time(unsigned long flow_jiffies);
273
274int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key);
275int ovs_flow_key_update_l3l4(struct sk_buff *skb, struct sw_flow_key *key);
276int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
277			 struct sk_buff *skb,
278			 struct sw_flow_key *key);
279/* Extract key from packet coming from userspace. */
280int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
281				   struct sk_buff *skb,
282				   struct sw_flow_key *key, bool log);
283
284#endif /* flow.h */
285