1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3 * Copyright (c) 2007-2014 Nicira, Inc.
4 */
5
6 #ifndef DATAPATH_H
7 #define DATAPATH_H 1
8
9 #include <asm/page.h>
10 #include <linux/kernel.h>
11 #include <linux/mutex.h>
12 #include <linux/netdevice.h>
13 #include <linux/skbuff.h>
14 #include <linux/u64_stats_sync.h>
15 #include <net/ip_tunnels.h>
16
17 #include "conntrack.h"
18 #include "flow.h"
19 #include "flow_table.h"
20 #include "meter.h"
21 #include "vport-internal_dev.h"
22
23 #define DP_MAX_PORTS USHRT_MAX
24 #define DP_VPORT_HASH_BUCKETS 1024
25 #define DP_MASKS_REBALANCE_INTERVAL 4000
26
27 /**
28 * struct dp_stats_percpu - per-cpu packet processing statistics for a given
29 * datapath.
30 * @n_hit: Number of received packets for which a matching flow was found in
31 * the flow table.
32 * @n_miss: Number of received packets that had no matching flow in the flow
33 * table. The sum of @n_hit and @n_miss is the number of packets that have
34 * been received by the datapath.
35 * @n_lost: Number of received packets that had no matching flow in the flow
36 * table that could not be sent to userspace (normally due to an overflow in
37 * one of the datapath's queues).
38 * @n_mask_hit: Number of masks looked up for flow match.
39 * @n_mask_hit / (@n_hit + @n_missed) will be the average masks looked
40 * up per packet.
41 * @n_cache_hit: The number of received packets that had their mask found using
42 * the mask cache.
43 */
44 struct dp_stats_percpu {
45 u64 n_hit;
46 u64 n_missed;
47 u64 n_lost;
48 u64 n_mask_hit;
49 u64 n_cache_hit;
50 struct u64_stats_sync syncp;
51 };
52
53 /**
54 * struct dp_nlsk_pids - array of netlink portids of for a datapath.
55 * This is used when OVS_DP_F_DISPATCH_UPCALL_PER_CPU
56 * is enabled and must be protected by rcu.
57 * @rcu: RCU callback head for deferred destruction.
58 * @n_pids: Size of @pids array.
59 * @pids: Array storing the Netlink socket PIDs indexed by CPU ID for packets
60 * that miss the flow table.
61 */
62 struct dp_nlsk_pids {
63 struct rcu_head rcu;
64 u32 n_pids;
65 u32 pids[];
66 };
67
68 /**
69 * struct datapath - datapath for flow-based packet switching
70 * @rcu: RCU callback head for deferred destruction.
71 * @list_node: Element in global 'dps' list.
72 * @table: flow table.
73 * @ports: Hash table for ports. %OVSP_LOCAL port always exists. Protected by
74 * ovs_mutex and RCU.
75 * @stats_percpu: Per-CPU datapath statistics.
76 * @net: Reference to net namespace.
77 * @max_headroom: the maximum headroom of all vports in this datapath; it will
78 * be used by all the internal vports in this dp.
79 * @upcall_portids: RCU protected 'struct dp_nlsk_pids'.
80 *
81 * Context: See the comment on locking at the top of datapath.c for additional
82 * locking information.
83 */
84 struct datapath {
85 struct rcu_head rcu;
86 struct list_head list_node;
87
88 /* Flow table. */
89 struct flow_table table;
90
91 /* Switch ports. */
92 struct hlist_head *ports;
93
94 /* Stats. */
95 struct dp_stats_percpu __percpu *stats_percpu;
96
97 /* Network namespace ref. */
98 possible_net_t net;
99
100 u32 user_features;
101
102 u32 max_headroom;
103
104 /* Switch meters. */
105 struct dp_meter_table meter_tbl;
106
107 struct dp_nlsk_pids __rcu *upcall_portids;
108 };
109
110 /**
111 * struct ovs_skb_cb - OVS data in skb CB
112 * @input_vport: The original vport packet came in on. This value is cached
113 * when a packet is received by OVS.
114 * @mru: The maximum received fragement size; 0 if the packet is not
115 * fragmented.
116 * @acts_origlen: The netlink size of the flow actions applied to this skb.
117 * @cutlen: The number of bytes from the packet end to be removed.
118 * @probability: The sampling probability that was applied to this skb; 0 means
119 * no sampling has occurred; U32_MAX means 100% probability.
120 */
121 struct ovs_skb_cb {
122 struct vport *input_vport;
123 u16 mru;
124 u16 acts_origlen;
125 u32 cutlen;
126 u32 probability;
127 };
128 #define OVS_CB(skb) ((struct ovs_skb_cb *)(skb)->cb)
129
130 /**
131 * struct dp_upcall - metadata to include with a packet to send to userspace
132 * @cmd: One of %OVS_PACKET_CMD_*.
133 * @userdata: If nonnull, its variable-length value is passed to userspace as
134 * %OVS_PACKET_ATTR_USERDATA.
135 * @portid: Netlink portid to which packet should be sent. If @portid is 0
136 * then no packet is sent and the packet is accounted in the datapath's @n_lost
137 * counter.
138 * @egress_tun_info: If nonnull, becomes %OVS_PACKET_ATTR_EGRESS_TUN_KEY.
139 * @mru: If not zero, Maximum received IP fragment size.
140 */
141 struct dp_upcall_info {
142 struct ip_tunnel_info *egress_tun_info;
143 const struct nlattr *userdata;
144 const struct nlattr *actions;
145 int actions_len;
146 u32 portid;
147 u8 cmd;
148 u16 mru;
149 };
150
151 /**
152 * struct ovs_net - Per net-namespace data for ovs.
153 * @dps: List of datapaths to enable dumping them all out.
154 * Protected by genl_mutex.
155 */
156 struct ovs_net {
157 struct list_head dps;
158 struct work_struct dp_notify_work;
159 struct delayed_work masks_rebalance;
160 #if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
161 struct ovs_ct_limit_info *ct_limit_info;
162 #endif
163
164 /* Module reference for configuring conntrack. */
165 bool xt_label;
166 };
167
168 /**
169 * enum ovs_pkt_hash_types - hash info to include with a packet
170 * to send to userspace.
171 * @OVS_PACKET_HASH_SW_BIT: indicates hash was computed in software stack.
172 * @OVS_PACKET_HASH_L4_BIT: indicates hash is a canonical 4-tuple hash
173 * over transport ports.
174 */
175 enum ovs_pkt_hash_types {
176 OVS_PACKET_HASH_SW_BIT = (1ULL << 32),
177 OVS_PACKET_HASH_L4_BIT = (1ULL << 33),
178 };
179
180 extern unsigned int ovs_net_id;
181 void ovs_lock(void);
182 void ovs_unlock(void);
183
184 #ifdef CONFIG_LOCKDEP
185 int lockdep_ovsl_is_held(void);
186 #else
187 #define lockdep_ovsl_is_held() 1
188 #endif
189
190 #define ASSERT_OVSL() WARN_ON(!lockdep_ovsl_is_held())
191 #define ovsl_dereference(p) \
192 rcu_dereference_protected(p, lockdep_ovsl_is_held())
193 #define rcu_dereference_ovsl(p) \
194 rcu_dereference_check(p, lockdep_ovsl_is_held())
195
196 static inline struct net *ovs_dp_get_net(const struct datapath *dp)
197 {
198 return read_pnet(&dp->net);
199 }
200
201 static inline void ovs_dp_set_net(struct datapath *dp, struct net *net)
202 {
203 write_pnet(&dp->net, net);
204 }
205
206 struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no);
207
208 static inline struct vport *ovs_vport_rcu(const struct datapath *dp, int port_no)
209 {
210 WARN_ON_ONCE(!rcu_read_lock_held());
211 return ovs_lookup_vport(dp, port_no);
212 }
213
214 static inline struct vport *ovs_vport_ovsl_rcu(const struct datapath *dp, int port_no)
215 {
216 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
217 return ovs_lookup_vport(dp, port_no);
218 }
219
220 static inline struct vport *ovs_vport_ovsl(const struct datapath *dp, int port_no)
221 {
222 ASSERT_OVSL();
223 return ovs_lookup_vport(dp, port_no);
224 }
225
226 /* Must be called with rcu_read_lock. */
227 static inline struct datapath *get_dp_rcu(struct net *net, int dp_ifindex)
228 {
229 struct net_device *dev = dev_get_by_index_rcu(net, dp_ifindex);
230
231 if (dev) {
232 struct vport *vport = ovs_internal_dev_get_vport(dev);
233
234 if (vport)
235 return vport->dp;
236 }
237
238 return NULL;
239 }
240
241 /* The caller must hold either ovs_mutex or rcu_read_lock to keep the
242 * returned dp pointer valid.
243 */
244 static inline struct datapath *get_dp(struct net *net, int dp_ifindex)
245 {
246 struct datapath *dp;
247
248 WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_ovsl_is_held());
249 rcu_read_lock();
250 dp = get_dp_rcu(net, dp_ifindex);
251 rcu_read_unlock();
252
253 return dp;
254 }
255
256 extern struct notifier_block ovs_dp_device_notifier;
257 extern struct genl_family dp_vport_genl_family;
258
259 void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key);
260 void ovs_dp_detach_port(struct vport *);
261 int ovs_dp_upcall(struct datapath *, struct sk_buff *,
262 const struct sw_flow_key *, const struct dp_upcall_info *,
263 uint32_t cutlen);
264
265 u32 ovs_dp_get_upcall_portid(const struct datapath *dp, uint32_t cpu_id);
266
267 const char *ovs_dp_name(const struct datapath *dp);
268 struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, struct net *net,
269 u32 portid, u32 seq, u8 cmd);
270
271 int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb,
272 const struct sw_flow_actions *, struct sw_flow_key *);
273
274 void ovs_dp_notify_wq(struct work_struct *work);
275
276 int action_fifos_init(void);
277 void action_fifos_exit(void);
278
279 /* 'KEY' must not have any bits set outside of the 'MASK' */
280 #define OVS_MASKED(OLD, KEY, MASK) ((KEY) | ((OLD) & ~(MASK)))
281 #define OVS_SET_MASKED(OLD, KEY, MASK) ((OLD) = OVS_MASKED(OLD, KEY, MASK))
282
283 #define OVS_NLERR(logging_allowed, fmt, ...) \
284 do { \
285 if (logging_allowed && net_ratelimit()) \
286 pr_info("netlink: " fmt "\n", ##__VA_ARGS__); \
287 } while (0)
288 #endif /* datapath.h */
289
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.